eLife Assessment

In this useful study, the authors perform voltage imaging of CA1 pyramidal cells in head-fixed mice running on a track while local field potentials (LFPs) are recorded. The authors conclude that synchronous ensembles of neurons are differentially associated with different types of LFP patterns, namely theta and ripples. However, evidence for the claims remains incomplete, due to caveats of the experimental approach that were not acknowledged and strong claims that are based on a sparse data set.

Reviewer #1 (Public review):

Summary:

For many years, there has been extensive electrophysiological research investigating the relationship between local field potential patterns and individual cell spike patterns in the hippocampus. In this study, using state-of-the-art imaging techniques, they examined spike synchrony of hippocampal cells during locomotion and immobility states. In contrast to conventional understanding of the hippocampus, the authors demonstrated that hippocampal place cells exhibit prominent synchronous spikes locked to theta oscillations.

Strengths:

The voltage imaging used in this study is a highly novel method that allows recording not only suprathreshold-level spikes but also subthreshold-level activity. With its high frame rate, it offers time resolution comparable to electrophysiological recordings.

Reviewer #2 (Public review):

Summary:

This study employed voltage imaging in the CA1 region of the mouse hippocampus during the exploration of a novel environment. The authors report synchronous activity, involving almost half of the imaged neurons, occurred during periods of immobility. These events did not correlate with SWRs, but instead, occurred during theta oscillations and were phased locked to the trough of theta. Moreover, pairs of neurons with high synchronization tended to display non-overlapping place fields, leading the authors to suggest these events may play a role in binding a distributed representation of the context.

Strengths:

Technically this is an impressive study, using an emerging approach that allow single-cell resolution voltage imaging in animals, that while head-fixed, can move through a real environment. The paper is written clearly and suggests novel observations about population-level activity in CA1.

Weaknesses:

The evidence provided is weak, with the authors making surprising population-level claims based on a very sparse data set (5 data sets, each with less than 20 neurons simultaneously recorded) acquired with exciting, but less tested technology. Further, while the authors link these observations to the novelty of the context, both in the title and text, they do not include data from subsequent visits to support this. Detailed comments are below:

(1) My first question for the authors, which is not addressed in the discussion, is why these events have not been observed in the countless extracellular recording experiments conducted in rodent CA1 during exploration of novel environments. Those data sets often have 10x the neurons simultaneously recording compared to these present data, thus the highly synchronous firing should be very hard to miss. Ideally, the authors could confirm their claims via the analysis of publicly available electrophysiology data sets. Further, the claim of high extra-SWR synchrony is complicated by the observation that their recorded neurons fail to spike during the limited number of SWRs recorded during behavior- again, not agreeing with much of the previous electrophysiological recordings.<br /> (2) The authors posit that these events are linked to the novelty of the context, both in the text, as well as in the title and abstract. However they do not include any imaging data from subsequent days to demonstrate the failure to see this synchrony in a familiar environment. If these data are available it would strengthen the proposed link to novelty is they were included.<br /> (3) In the discussion the authors begin by speculating the theta present during these synchronous events may be slower type II or attentional theta. This can be supported by demonstrating a frequency shift in the theta recording during these events/immobility versus the theta recording during movement.<br /> (4) The authors mention in the discussion that they image deep layer PCs in CA1, however this is not mentioned in the text or methods. They should include data, such as imaging of a slice of a brain post-recording with immunohistochemistry for a layer specific gene to support this.

Comments on revisions:

I have no further major requests and thank the authors for the additional data and analyses.

Reviewer #3 (Public review):

Summary:

In the present manuscript, the authors use a few minutes of voltage imaging of CA1 pyramidal cells in head-fixed mice running on a track while local field potentials (LFPs) are recorded. The authors suggest that synchronous ensembles of neurons are differentially associated with different types of LFP patterns, theta and ripples. The experiments are flawed in that the LFP is not "local" but rather collected the other side of the brain.

Strengths:

The authors use a cutting-edge technique.

Weaknesses:

The two main messages of the manuscript indicated in the title are not supported by the data. The title gives two messages that relate to CA1 pyramidal neurons in behaving head-fixed mice: (1) synchronous ensembles are associated with theta (2) synchronous ensembles are not associated with ripples. The main problem with the work is that the theta and ripple signals were recorded using electrophysiology from the opposite hemisphere to the one in which the spiking was monitored. However, both rhythms exhibit profound differences as a function of location.

Theta phase changes with the precise location along the proximo-distal and dorso-ventral axes, and importantly, even reverses with depth. Because the LFP was recorded using a single-contact tungsten electrode, there is no way to know whether the electrode was exactly in the CA1 pyramidal cell layer, or in the CA1 oriens, CA1 radiatum, or perhaps even CA3 - which exhibits ripples and theta which are weakly correlated and in anti-phase with the CA1 rhythms, respectively. Thus, there is no way to know whether the theta phase used in the analysis is the phase of the local CA1 theta.

Although the occurrence of CA1 ripples is often correlated across parts of the hippocampus, ripples are inherently a locally-generated rhythm. Independent ripples occur within a fraction of a millimeter within the same hemisphere. Ripples are also very sensitive to the precise depth - 100 micrometers up or down, and only a positive deflection/sharp wave is evident. Thus, even if the LFP was recorded from the center of the CA1 pyramidal layer in the contralateral hemisphere, it would not suffice for the claim made in the title.

Author response:

The following is the authors’ response to the original reviews.

We thank all reviewers for their thorough and thoughtful comments. We have carefully addressed each point raised, conducting new experiments and analyses to strengthen the manuscript. Below is a summary:

· Synchronous ensembles in new experiments: New experiments demonstrated synchronous ensembles during immobility in a novel environment (Figure 3-figure supplement 2) and revealed a significant reduction in such synchrony following familiarization training (Figure 4D).

· Ripple-associated activity: We detected a much larger number of ripple events to confirm (a) the suppression of CA1PC spiking during ripples (Figure 4Ai) and (b) that synchronous ensembles mostly occur outside ripples (Figure 3-figure supplement 3). Additionally, spiking suppression was accompanied by decreased subthreshold membrane potentials (Figure 4Bi, Ci). Ripple-associated spiking and membrane potential dynamics shifted toward higher firing rates and more depolarization after familiarization training (Figure 4).

· Public data analysis: Analysis of publicly available data identified thetaassociated synchronous ensembles, demonstrating the generalizability of our findings across different experimental conditions (Supplementary Figure 5).

· Neuron morphology and algorithm validation: Images of recorded neurons after experiments confirmed their intact morphology. We also provided details on validating spike detection algorithms (Methods and Supplementary Figure 1).

· Cell soma locations: New data and analyses illustrate the distribution of cells labeled at different embryonic days along the radial axis of the pyramidal layer (Supplementary Figure 1).

· Analyses testing the robustness of synchronous ensembles: Additional analyses examined the impact of complex bursts and thetaphase locking, confirming the robustness of synchronous ensembles detection (Supplementary Figures 3 and 4).

· Additional analyses and figures: We conducted further analyses and created new figures to address all remaining concerns (Response to Reviewer Figures 1-6).

We believe these revisions have significantly enhanced the paper, and we sincerely thank all reviewers for their invaluable feedback.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

For many years, there has been extensive electrophysiological research investigating the relationship between local field potential patterns and individual cell spike patterns in the hippocampus. In this study, using state-ofthe-art imaging techniques, they examined spike synchrony of hippocampal cells during locomotion and immobility states. In contrast to conventional understanding of the hippocampus, the authors demonstrated that hippocampal place cells exhibit prominent synchronous spikes locked to theta oscillations.

Strengths:

The voltage imaging used in this study is a highly novel method that allows recording not only suprathreshold-level spikes but also subthreshold-level activity. With its high frame rate, it offers time resolution comparable to electrophysiological recordings. Moreover, it enables the visualization of actual cell locations, allowing for the examination of spatial properties (e.g., Figure 4G).

We thank the reviewer for recognizing the strength of our study.

Weaknesses:

There is a notable deviation from several observations obtained through conventional electrophysiological recordings. Particularly, as mentioned below in detail, the considerable differences in baseline firing rates and no observations of ripple-triggered firing patterns raise some concerns about potential artifacts from imaging and analsyis, such as cell toxicity, abnormal excitability, and false detection of spikes. While these findings are intriguing if the validity of these methods is properly proven, accepting the current results as new insights is challenging.

We appreciate the reviewer’s insightful comments regarding the apparent deviation of our observation from conventional understanding, which we address in the following sections.

Reviewer #1 (Recommendations For The Authors):

(1) I am not particularly inclined to strongly adhere to conventional insights, but the findings obtained through this imaging method seem significantly different from those known from conventional electrophysiological recordings. For instance, there are noticeable differences in several basic firing characteristics. First, the average firing rates of 2.3-4.3 Hz (Line 97) appear higher than the distribution of firing frequencies reported in many electrophysiological recordings of pyramidal cells (e.g., Mizuseki et al., Cell Rep, 2013).

We understand that some of our findings differ from conventional insights. However, it is important to emphasize that many of our observations align closely with prior electrophysiological recordings. For instance, individual neurons in our study exhibit expected modulation by locomotion, spatial locations, novelty, and theta oscillations, all of which are hallmarks of normal hippocampal physiology.

Regarding the firing rates, it is important to highlight the heterogeneity of the firing rates, which range from 0.01 to 10 Hz, with a skewed distribution toward lower frequencies(1). While our values (2.3-4.3Hz) are higher than those reported by Mizuseki et al. (2013)(1) in rats, our recordings were obtained from mice and aligned with studies using mice, including firing rates of 2.1 Hz reported by McHugh et al. (1996) and 2.4-2.6 Hz by Buzsaki et al. (2003)(2,3).

In addition, our recordings were performed in a novel environment, which is known to enhance the firing rates of the hippocampal neurons(4). Consistent with this, our new recordings in a familiar environment demonstrate significantly lower firing rates (see below).

Results (line 279)

“Mean firing rates were significantly reduced in the familiar group compared to the novel group (Familiar group: 1.1 to 5.2 Hz (25^th-75^th percentiles), median=2.3 Hz, n\=66 cells, 6 sessions, 4 mice; Novel group: 1.7 to 6.0 Hz (25^th-75^th percentiles), median=4.2 Hz, n\=111 cells, 6 sessions, 6 mice, p\=0.0083, Wilcoxon signed-rank test).”

Second, while this finding suggests that spike synchrony is entirely unrelated to ripple-triggered events, it is indeed difficult to comprehend (researchers who have analyzed electrophysiological data, at the very least, should have experienced some degree of correlation between ripples and spikes).

We thank the reviewer for raising this important point. We, too, found it surprising that population synchrony appears largely unrelated to ripples. To ensure the robustness of this observation, we conducted new experiments under conditions optimized for ripple detection to (a) confirm that the lack of positive correlation is also observed under conditions where we can detect more ripples and (b) demonstrate that our imaging methods can detect a higher correlation between ripples and spikes in a familiar environment (see details below).

Results (line 251)

“It was puzzling that these CA1PCs exhibited robust spiking activities outside of ripples yet generated few spikes during ripples. To further investigate neuronal activities during ripples, we established a recording condition that allowed us to capture more ripple episodes. Specifically, we immobilized mice in a tube to promote behaviors favoring ripple generation. The mice were habituated to head fixation in a tube in a room distinct from the one where imaging experiments were conducted. On the imaging day, the mice were introduced to the recording room and head-fixed under the microscope for the first time.

CA1PCs were labeled in utero on embryonic day (E) 14.5 (n\=56 cells from 3 sessions in 3 mice) and E17.5 (n\=55 cells from 3 sessions in 3 mice) and imaged in adult brains. Both neuronal populations exhibited prominent peaks in their grand average CCGs and significantly higher synchronous event rates compared to jittered data (Figure 3-figure supplement 2A, B). Approximately 40% of the recorded neurons participated in synchronous ensembles, indicating robust synchronous activity involving a substantial proportion of the recorded cells (Figure 3-figure supplement 2C).

In total, 1052 synchronous ensembles and 174 ripple episodes were detected across these imaging sessions. Consistent with findings from walking animals, few synchronous ensembles occurred during ripples when animals were immobilized in a tube (Figure 3-figure supplement 3A, B). Moreover, no distinguishable ripple oscillations were observed in synchronous events, and the average firing rates during ripple episodes were near zero (Figure 3-figure supplement 3C, D). At the single-cell level, 90% of neurons showed significant negative spiking modulation during ripples, with ripple modulation indexes close to -1, indicating strong suppression of spiking (Figure 4Ai). This suppression extended to subthreshold membrane potentials, as nearly all cells exhibited decreased fluorescence during ripples compared to baseline (Figure 4Bi, Ci). These results demonstrate that spiking activity and subthreshold membrane potentials are robustly suppressed during ripples.

Contextual novelty plays a critical role in shaping hippocampal neuronal activities. To assess its influence, we trained mice to become familiar with the imaging procedure and the recording environment over five days and recorded CA1PC activities on the final day. Mean firing rates were significantly reduced in the familiar group compared to the novel group (Familiar group:

1.1 to 5.2 Hz (25^th-75^th percentiles), median=2.3 Hz, n\=66 cells, 6 sessions, 4 mice; Novel group: 1.7 to 6.0 Hz (25^th-75^th percentiles), median=4.2 Hz, n\=111 cells, 6 sessions, 6 mice, p\=0.0083, Wilcoxon signed-rank test). Additionally, 15% of the neurons in the familiar group exhibited significantly positive spiking modulation by ripples, while fewer cells showed negative modulation compared to the novel group (Figure 4A). During ripples, neurons in the novel group predominantly displayed hyperpolarizing membrane voltage responses, whereas a subset of neurons in the familiar group exhibited prominent depolarizing responses (Figure 4B). The mean fluorescence changes in the familiar group shifted toward depolarization compared to the novel group (Figure 4C). Finally, synchronous event frequencies were significantly lower in the familiar context, indicating weaker synchronous activities under familiar conditions (Figure 4D). These results demonstrate that hippocampal neuronal activities, particularly synchronous ensembles, are strongly influenced by contextual novelty.”

Third, the fact that more than 40% of cells frequently exhibit synchronous firing other than during ripples has not been reported before, and if it were the case, many electrophysiologists would have likely noticed it. Overall, the excitability of cells seems too high.

We thank the reviewer for raising this point. As discussed above, the reported spike rates are within the range expected from the previous electrophysiology recordings in mice, especially given that we record cells in a novel environment. In addition, our jittering procedure ensures that the observed synchrony exceeds what could be expected from the given level of spike rates alone. These analyses support the robustness of our observations.

As mentioned below, there are concerns about experimental artifacts and analytical issues from optical imaging.

(2) Method: In surgery, the cortical tissue above the hippocampus was aspirated, which is a general method for in vivo calcium imaging from the hippocampus. Furthermore, they use a CAG promoter to express the sensors. To my knowledge, this promoter is excessively strong and may sometimes be toxic to cells. In addition, for imaging, they use DMSO and Pluronic F-127, which are relatively toxic materials (please describe their concentrations). These conditions might be damaging to hippocampal neurons.

We thank the reviewer for raising these comments. As the reviewer mentioned, cortical aspiration is a general method for in vivo imaging from the hippocampus and has been employed in numerous studies, including behavioral and systems-level investigations(5-15). For example, place cells are routinely recorded in both familiar and novel environments using this method and other approaches. Additionally, synchronous population activities have been observed and studied in the hippocampus both with and without cortical aspiration(6,15-18). These findings demonstrate that the hippocampal neuronal network generates place cells and synchronous activities regardless of whether the cortical tissue above it has been aspirated.

DMSO and Pluronic F-127 are used as solvents for dissolving the JF₅₅₂HaloTag ligand, and the resulting solution is injected into the bloodstream rather than directly into brain tissue. The concentrations of these reagents in the dye solution are now described in the text (see below). Assuming a blood volume of 2 ml in adult mice, the final concentrations of DMSO and Pluronic F-127 in the bloodstream are estimated to be 1% upon injection and then decrease rapidly while they are metabolized and excreted out of the body. Moreover, the effective concentrations in the brain tissue would be even lower. These low concentrations have been demonstrated to have minimal impact on cells and tissue(19-22).

Methods (line 616)

“JF₅₅₂-HaloTag ligand (a generous gift from Dr. Luke Lavis) was first dissolved in DMSO (20 μl, Sigma) and then diluted in Pluronic^TM F-127 (20 μl, P3000MP, Invitrogen) and PBS to achieve a final concentration of 0.83 mM of JF₅₅₂-HaloTag ligand. The solution was then injected intravenously through the retro-orbital sinus. Imaging sessions were initiated 3 hours after the injection of the JF₅₅₂-HaloTag ligand.”

We understand that the CAG promoter may sometimes be toxic to cells if it drives high expression. However, it is important to note that we injected highly diluted virus (20x, final titer: 2.7x10¹² GC/ml) to avoid excessive expression levels. This titer was determined from serial dilution experiments to ensure an optimal expression level free from toxicity (see below). The same titer was used in a previous study(23) to label CA1 interneurons, which exhibited physiological spike rates and synchrony (see Abdelfattah 2023, Neuron, Figure 8). Furthermore, Voltron expression does not significantly affect key cellular properties, including membrane resistance, membrane capacitance, resting membrane potentials, spike amplitudes, and spike width (see Abdelfattah 2019, Science, Supplementary Figures 11 and 12). In our recordings, individual neurons exhibit the expected modulation by locomotion, spatial locations, novelty, and theta oscillations. We now include images of the recorded neurons to demonstrate their intact morphology and healthy appearance following imaging experiments (Supplementary Figure 1A, B), further supporting minimal cytotoxic effects.

Methods (line 577)

“A serial dilution experiment was conducted to determine an optimal titer of the virus carrying Voltron2 genes, minimizing cell toxicity, for use in this and in previous imaging experiments. A fine injection pipette (tip diameter 10-60 um) was used to inject AAV2/1-CAG-flex-Voltron2-ST (2.7x10¹² GC/ml, a generous gift from Dr. Eric Schreiter and the GENIE team at HHMI Janelia Research Campus) into the exposed regions at a depth of 200 μm (up to six injection sites and 100-200 nL of viral suspension).”

(3) Another concern is the relatively low number of cells simultaneously recorded during imaging compared to typical hippocampal imaging such as Inscopix which often records several hundred cells. In this study, however, this number is 20 or fewer. This is likely because the visualized cells at baseline were limited to this extent. It is possible that these cells represent particularly too strong sensor expression, which may facilitate visualization and high signal-to-noise ratio in voltage imaging. Consequently, there is a possibility of abnormal activity occurring in these cells.

The Inscopix studies use calcium imaging, which has a temporal resolution that is too slow to resolve fast synchrony central to our study. To enable highspeed voltage imaging at 2000 frames per second, we employed strategies to achieve sparse labeling and carefully limited the number of labeled cells to minimize out-of-focus contamination. In our analysis, we applied a criterion to include only cells separated by 70 μm or longer, reducing the potential for channel cross-talk among nearby neurons. These criteria limited the number of simultaneously imaged cells in our experiments. To address this issue, we have now included new data from 12 additional animals with 177 neurons to support our findings.

Furthermore, despite the limited number of simultaneously imaged cells, population synchrony beyond what could be expected by chance can be detected using rigorous statistical procedures. As discussed earlier, neuronal activities were within the expected range; they were modulated by animals’ locomotion (Figure 2 and Supplementary Figure 2), exhibited place tuning, and were significantly reduced when the recording context became familiar, supporting the normal physiology of the recorded cells.

(4) Analysis: There are some criteria for detecting spikes (described in the Methods), but there are concerns about whether these criteria truly extract only spike activity. When examining the traces in Figure 1 and Figure 2, there appear to be some activities that show fluorescence increases up to the level of putative spikes. How can we determine that these are indeed subthreshold changes? Conversely, some activities detected as spikes may also be subthreshold synaptic potential (this possibility concerns me). There is a need for more precise validation of spike detection analysis to ensure its accuracy.

Regarding spike detection, we used validated algorithms(23-25) to ensure robust and reliable spike identification. Spiking activity was first separated from slower subthreshold potentials using high-pass filtering. This approach prevents slow fluorescence increases from being misinterpreted as spikes, even if their amplitude is large. We benchmarked this detection algorithm in our recent publication (Huang et al., 2024)(24), demonstrating its high sensitivity and specificity in spike detection (see the figure below). While we acknowledge that a small number of spikes, particularly those occurring later in a burst, might be missed due to their smaller amplitudes (as illustrated in Figures 1 and 2 of the manuscript), we anticipate that any missed spikes would lead to a decrease, rather than an increase, in synchrony between neurons. Overall, we are confident that spike detection is performed in a rigorous and reliable manner.

Method (line 670)

“Previous studies have described and validated the procedure for imaging preprocessing and spike detection. In short, the fluorescence intensities of individual neurons were calculated by averaging the fluorescence intensities of pixels from the same ROIs. Bleaching was corrected by calculating the baseline fluorescence (F₀) at each time point as an average of the fluorescence intensities within ±0.5 seconds around the time point. The dF/F was calculated as the F₀ minus the fluorescence intensity of the same time point divided by F₀. Positive fluorescence transients were detected to identify spikes from the high-passed dF/F traces created by subtracting the dF/F traces from the median-filtered version with a 5-ms window. To simulate the noise of recordings, high-passed dF/F traces were inverted, and the amplitudes of the transients detected from the inverted traces were used to construct a noise distribution of the spike amplitudes. A threshold was set by comparing the amplitudes of the detected transients with the noise distribution of the spike amplitudes to minimize the sum of type I and type II errors. Spikes were first detected when transients were larger than the threshold. Then, spike amplitudes smaller than half of the top 5% spike amplitudes were excluded. The signal-to-noise ratio (SNR) was calculated for each neuron as a ratio of the averaged spike amplitudes over the standard deviation of the high-passed dF/F traces, excluding points 2 ms before and 4 ms after each detected spike to estimate the quality of the recordings.”

(5) If the authors aim to establish this new physiological phenomenon, it is necessary to compare it with electrophysiological data or verify if similar phenomena can be detected from electrophysiological data. Recently, various datasets have been made publicly available (e.g. CRCNS and Mendeley data), and these should be easily verifiable without the need for conducting experiments.

We thank the reviewer for the suggestion. To address this, we analyzed a publicly available dataset (hc-11 on CRCNS), which contains hippocampal recordings from rats navigating novel mazes for water rewards. Using our algorithm, we detected significant population synchrony in the dataset (Supplementary Figure 5A). The synchronous event rates were 6.4-fold higher than those in jittered controls, demonstrating the reliability of our findings.

Additionally, these synchronous events mostly occurred in the absence of ripples and were coupled to theta oscillations (Supplementary Figure 5B-D). These results not only validate our findings using independent datasets but also highlight the generalizability of synchronous ensembles as a distinct network phenomenon relevant to hippocampal function.

Results (line 366)

“To further investigate synchronous ensembles across different datasets, we analyzed publicly available hippocampal recordings ‘hc-11’ from the CRCNS repository, where rats navigated novel mazes for water rewards (see Method). Using our algorithm, we identified a significant number of synchronous ensembles during the first three minutes of novel navigation. On average, the rates of synchronous events were 6.4-fold higher than those detected in jittered controls (mean event rate: 2.0 ± 0.3 Hz for the original data vs. 0.32 ± 0.03 Hz for jittered data, n \= 8 sessions, p \= 0.0078, W \= 36, Wilcoxon signedrank test; Supplementary Figure 5A). To assess whether ripple oscillations were associated with these synchronous ensembles, we analyzed ripple event rates and their relationship to population synchrony. During this period, ripple events were infrequent (mean ripple rate: 0.02 ± 0.01, n \= 8 sessions), and ripple power peaked during ripple episodes but remained low at the timings of population synchrony (Supplementary Figure 5B). Nevertheless, LFP traces aligned to population synchrony revealed prominent theta oscillations (Supplementary Figure 5C). Synchronous ensembles were modulated by LFP theta oscillation (modulation strength: 0.30 ± 0.04, n \= 8 sessions, p < 0.001), and the timings of individual ensembles were consistently locked to the preferred phase of each session, suggesting a functional coupling of synchronous ensembles to theta oscillations important for information processing (Supplementary Figure 5D).”

(6) Please describe exact statistical information (e.g. statistical values, degree of freedom, and test types) throughout the manuscript.

Statistical values, degree of freedom and test types have been included in the manuscript. Please see below an example in the manuscript:

Result (line 96)

“Consistent with previous studies, neurons labeled on E14.5 located more on the deep side of the pyramidal layer than those labeled on E17.5 (t₍₆₀₁₎=22.8, p<0.0001, Student’s t-test; Supplementary Figure 1C, D).”

Minor comment - Figure 2A legend: what is "gray rectangles"?

We apologize for the inconsistency in nomenclature in the figure legends. We have now corrected this issue and consistently use the term “gray vertical bars” to indicate the timings and durations of synchronous events throughout the article.

Reviewer #2 (Public Review):

Summary:

This study employed voltage imaging in the CA1 region of the mouse hippocampus during the exploration of a novel environment. The authors report synchronous activity, involving almost half of the imaged neurons, occurred during periods of immobility. These events did not correlate with SWRs, but instead, occurred during theta oscillations and were phasedlocked to the trough of theta. Moreover, pairs of neurons with high synchronization tended to display non-overlapping place fields, leading the authors to suggest these events may play a role in binding a distributed representation of the context.

We thank the reviewer for a thorough and thoughtful review of our paper.

Strengths:

Technically this is an impressive study, using an emerging approach that allows single-cell resolution voltage imaging in animals, that while head-fixed, can move through a real environment. The paper is written clearly and suggests novel observations about population-level activity in CA1.

We thank the reviewer for pointing out the technical strength and the novelty of our study.

Weaknesses:

The evidence provided is weak, with the authors making surprising population-level claims based on a very sparse data set (5 data sets, each with less than 20 neurons simultaneously recorded) acquired with exciting, but less tested technology. Further, while the authors link these observations to the novelty of the context, both in the title and text, they do not include data from subsequent visits to support this. Detailed comments are below:

We understand the reviewer’s concerns regarding the dataset size. In the revised manuscript, we have included additional data to further strengthen our conclusions and provide a more robust dataset. Specifically, we expanded our analysis by increasing the number of sessions and neurons recorded, ensuring that the findings are more representative and less likely to be influenced by sample sizes.

Moreover, synchronous ensembles exceeding what could be expected by chance were detected in all examined data, validating our claims regarding population synchrony. We have also carefully considered the potential impact of the technology used in our experiments and included additional validation and comparison with results from other studies employing complementary techniques to support the reliability of our conclusions.

Regarding the link to novelty, we have included data from subsequent visits, as suggested by the reviewer. These new data demonstrate that the observed changes in synchronous ensembles are context-dependent and significantly influenced by novelty. This confirms the novelty-related effects observed during initial visits and further supports the conclusions drawn in the manuscript. Please see below for our detailed replies to each of the reviewer’s points.

(1) My first question for the authors, which is not addressed in the discussion, is why these events have not been observed in the countless extracellular recording experiments conducted in rodent CA1 during the exploration of novel environments. Those data sets often have 10x the neurons simultaneously recording compared to these present data, thus the highly synchronous firing should be very hard to miss. Ideally, the authors could confirm their claims via the analysis of publicly available electrophysiology data sets. Further, the claim of high extra-SWR synchrony is complicated by the observation that their recorded neurons fail to spike during the limited number of SWRs recorded during behavior- again, not agreeing with much of the previous electrophysiological recordings.

We thank the reviewer for raising these important questions. To address the first question, it is possible that synchronous ensembles were not previously detected in extracellular recordings due to differences in detection methods or analysis approaches. To investigate this further, we analyzed a publicly available dataset (hc-11 on CRCNs), which contains hippocampal recordings from rats navigating novel mazes for water rewards. Using our algorithm, we detected robust synchronous ensembles in the dataset (Supplementary Figure 5). The rates of synchronous events were significantly higher than those in jittered controls, demonstrating the reliability and generalizability of these synchronous ensembles.

Results (line 366)

“To further investigate synchronous ensembles across different datasets, we analyzed publicly available hippocampal recordings ‘hc-11’ from the CRCNS repository, where rats navigated novel mazes for water rewards (see Method). Using our algorithm, we identified a significant number of synchronous ensembles during the first three minutes of novel navigation. On average, the rates of synchronous events were 6.4-fold higher than those detected in jittered controls (mean event rate: 2.0 ± 0.3 Hz for the original data vs. 0.32 ± 0.03 Hz for jittered data, n \= 8 sessions, p \= 0.0078, W \= 36, Wilcoxon signedrank test; Supplementary Figure 5A). To assess whether ripple oscillations were associated with these synchronous ensembles, we analyzed ripple event rates and their relationship to population synchrony. During this period, ripple events were infrequent (mean ripple rate: 0.02 ± 0.01, n \= 8 sessions), and ripple power peaked during ripple episodes but remained low at the timings of population synchrony (Supplementary Figure 5B). Nevertheless, LFP traces aligned to population synchrony revealed prominent theta oscillations (Supplementary Figure 5C). Synchronous ensembles were modulated by LFP theta oscillation (modulation strength: 0.30 ± 0.04, n \= 8 sessions, p < 0.001), and the timings of individual ensembles were consistently locked to the preferred phase of each session, suggesting a functional coupling of synchronous ensembles to theta oscillations important for information processing (Supplementary Figure 5D).”

To address the second question, we conducted new experiments under conditions optimized for ripple generation. Specifically, we recorded neurons in mice head-fixed in a novel environment, resulting in 174 ripple episodes across six sessions. Consistent with our original findings, spiking rates were significantly suppressed and membrane potentials were hyperpolarized during ripples (Figure 4Ai-Ci of the manuscript). Despite this suppression, the same neurons exhibit rich synchronous activities outside of ripples (Figure 3-figure supplement 3 of the manuscript). These results confirm that these synchronous ensembles are distinct from ripple-related neuronal activity and strengthen our claim that the observed synchronous ensembles represent a distinct physiological phenomenon, consistent across different datasets and experimental conditions.

Results (line 251)

“It was puzzling that these CA1PCs exhibited robust spiking activities outside of ripples yet generated few spikes during ripples. To further investigate neuronal activities during ripples, we established a recording condition that allowed us to capture more ripple episodes. Specifically, we immobilized mice in a tube to promote behaviors favoring ripple generation. The mice were habituated to head fixation in a tube in a room distinct from the one where imaging experiments were conducted. On the imaging day, the mice were introduced to the recording room and head-fixed under the microscope for the first time.

CA1PCs were labeled in utero on embryonic day (E) 14.5 (n\=56 cells from 3 sessions in 3 mice) and E17.5 (n\=55 cells from 3 sessions in 3 mice) and imaged in adult brains. Both neuronal populations exhibited prominent peaks in their grand average CCGs and significantly higher synchronous event rates compared to jittered data (Figure 3-figure supplement 2A, B). Approximately 40% of the recorded neurons participated in synchronous ensembles, indicating robust synchronous activity involving a substantial proportion of the recorded cells (Figure 3-figure supplement 2C).

In total, 1052 synchronous ensembles and 174 ripple episodes were detected across these imaging sessions. Consistent with findings from walking animals, few synchronous ensembles occurred during ripples when animals were immobilized in a tube (Figure 3-figure supplement 3A, B). Moreover, no distinguishable ripple oscillations were observed in synchronous events, and the average firing rates during ripple episodes were near zero (Figure 3-figure supplement 3C, D). At the single-cell level, 90% of neurons showed significant negative spiking modulation during ripples, with ripple modulation indexes close to -1, indicating strong suppression of spiking (Figure 4Ai). This suppression extended to subthreshold membrane potentials, as nearly all cells exhibited decreased fluorescence during ripples compared to baseline (Figure 4Bi, Ci). These results demonstrate that spiking activity and subthreshold membrane potentials are robustly suppressed during ripples.”

(2) The authors posit that these events are linked to the novelty of the context, both in the text, as well as in the title and abstract. However, they do not include any imaging data from subsequent days to demonstrate the failure to see this synchrony in a familiar environment. If these data are available it would strengthen the proposed link to novelty if they were included.

Following the reviewer’s suggestion, we record neuronal activities in a familiar context to test the proposed link between synchronous activity and contextual novelty. We found that synchronous activity levels were significantly lower in the familiar context compared to the novel context, demonstrating that synchronous activity is strongly modulated by contextual novelty (Figure 4D of the manuscript). These findings provide further support for a link of the synchronous ensembles to novel environmental contexts.

Result (line 277)

“Contextual novelty plays a critical role in shaping hippocampal neuronal activities. To assess its influence, we trained mice to become familiar with the imaging procedure and the recording environment over five days and recorded CA1PC activities on the final day. Mean firing rates were significantly reduced in the familiar group compared to the novel group (Familiar group:

1.1 to 5.2 Hz (25^th-75^th percentiles), median=2.3 Hz, n\=66 cells, 6 sessions, 4 mice; Novel group: 1.7 to 6.0 Hz (25^th-75^th percentiles), median=4.2 Hz, n\=111 cells, 6 sessions, 6 mice, p\=0.0083, Wilcoxon signed-rank test). Additionally, 15% of the neurons in the familiar group exhibited significantly positive spiking modulation by ripples, while fewer cells showed negative modulation compared to the novel group (Figure 4A). During ripples, neurons in the novel group predominantly displayed hyperpolarizing membrane voltage responses, whereas a subset of neurons in the familiar group exhibited prominent depolarizing responses (Figure 4B). The mean fluorescence changes in the familiar group shifted toward depolarization compared to the novel group (Figure 4C). Finally, synchronous event frequencies were significantly lower in the familiar context, indicating weaker synchronous activities under familiar conditions (Figure 4D). These results demonstrate that hippocampal neuronal activities, particularly synchronous ensembles, are strongly influenced by contextual novelty.”

(3) In the discussion the authors begin by speculating the theta present during these synchronous events may be slower type II or attentional theta. This can be supported by demonstrating a frequency shift in the theta recording during these events/immobility versus the theta recording during movement.

We thank the reviewer for the suggestion. As the reviewer points out, we did observe a frequency shift in synchrony-associated theta during immobility compared to locomotion (see Figure 5B, red vs. blue curves). We have now highlighted this result in the discussion section. Please refer to the text below.

Discussion (line 471)

“On the other hand, type 2 theta, or attentional theta, is slightly slower and is blocked by muscarinic receptor antagonists, emerging during states of arousal or attention, such as when entering a new environment. Consistent with these distinctions, the peak of the power spectrum density shows a distinctively slower theta frequency during immobility compared to locomotion (Figure 5B).”

(4) The authors mention in the discussion that they image deep-layer PCs in CA1, however, this is not mentioned in the text or methods. They should include data, such as imaging of a slice of a brain post-recording with immunohistochemistry for a layer-specific gene to support this.

We thank the reviewer for the constructive suggestion. In response, we have added images of slices from both E14.5 and E17.5 brains and analyzed soma locations along the radial axis of the pyramidal layer. The results are included in the main text, Methods, and Supplementary Figure 1 of the manuscript (see below).

Result (line 96)

“Consistent with previous studies, neurons labeled on E14.5 located more on the deep side of the pyramidal layer than those labeled on E17.5 (t₍₆₀₁₎=22.8, p<0.0001, Student’s t-test; Supplementary Figure 1C, D).”

Methods (line 563)

“The injection resulted in Cre expression among neurons born on the day of injection, with earlier injection labeling neurons located on the deeper side of the cell layer.”

Reviewer #3 (Public Review):

Summary:

In the present manuscript, the authors use a few minutes of voltage imaging of CA1 pyramidal cells in head-fixed mice running on a track while local field potentials (LFPs) are recorded. The authors suggest that synchronous ensembles of neurons are differentially associated with different types of LFP patterns, theta and ripples. The experiments are flawed in that the LFP is not "local" but rather collected in the other side of the brain, and the investigation is flawed due to multiple problems with the point process analyses. The synchrony terminology refers to dozens of milliseconds as opposed to the millisecond timescale referred to in prior work, and the interpretations do not take into account theta phase locking as a simple alternative explanation.

We appreciate the reviewer’s feedback and acknowledge the concerns raised. However, we believe these concerns can be effectively addressed without compromising the validity of our conclusions. With this in mind, we respectfully disagree with the assessment that our experiments and investigation are flawed. Please allow us to address these concerns and offer additional context to support the validity of our study.

Weaknesses:

The two main messages of the manuscript indicated in the title are not supported by the data. The title gives two messages that relate to CA1 pyramidal neurons in behaving head-fixed mice: (1) synchronous ensembles are associated with theta (2) synchronous ensembles are not associated with ripples.

There are two main methodological problems with the work: (1) experimentally, the theta and ripple signals were recorded using electrophysiology from the opposite hemisphere to the one in which the spiking was monitored. However, both signals exhibit profound differences as a function of location: theta phase changes with the precise location along the proximo-distal and dorso-ventral axes, and importantly, even reverses with depth. And ripples are often a local phenomenon - independent ripples occur within a fraction of a millimeter within the same hemisphere, let alone different hemispheres. Ripples are very sensitive to the precise depth - 100 micrometers up or down, and only a positive deflection/sharp wave is evident.

We acknowledge the reviewer’s consideration regarding the collection of LFP from the contralateral hemisphere. While we acknowledge the limitation of this design, we believe these contralateral LFP recordings still provide valuable insights into the dynamics of synchronous ensembles. Despite potential variations in theta phases due to differences in recording locations and depths, the occurrence and amplitudes of theta oscillations are generally wellcoordinated across hemispheres (Buzsaki et al., 2003, Fig 5)(3). The presence of prominent contralateral LFP theta activity around the times of synchronous ensembles in our study (Figure 5A of the manuscript) strongly supports our conclusion about their association with theta oscillations, even with LFP collected from the opposite hemisphere.

Additionally, we explicitly noted in the manuscript that the “preferred phases” varied between sessions, likely reflecting variability in recording locations (see below). Thus, we believe the concern about theta phase variability has already been adequately addressed in the current manuscript.

Result (line 321)

“Although the preferred phases varied from session to session due to differences in recording sites along the proximal-distal axis of the hippocampus, the timings of individual ensembles were consistently locked to the preferred phase of each session (Figure 5C).”

While we acknowledge that ripple oscillations can sometimes occur locally, the majority of ripples occur synchronously in both hemispheres (up to 70%)(3,26), as demonstrated both in the literature (Szabo et al., 2022, Supplementary Figure 2) and by data from our lab (Huang et al., 2024, Figure S6). As a result, using contralateral LFP to infer ripple occurrence on the ipsilateral side is a well-established practice in the field, commonly employed by many studies published in reputable journals(26-29). Given the high co-occurrence of both theta and ripple oscillations across hemispheres, we maintain that the two main messages of our manuscript are supported by data, despite the concern regarding phase discrepancy mentioned by the reviewer.

(2) The analysis of the point process data (spike trains) is entirely flawed. There are many technical issues: complex spikes ("bursts") are not accounted for; differences in spike counts between the various conditions ("locomotion" and "immobility") are not accounted for; the pooling of multiple CCGs assumes independence, whereas even conditional independence cannot be assumed; etc.

We acknowledge the reviewer’s concern regarding spike train analysis. Complex bursts or differences in behavioral conditions can indeed lead to variations in spike counts, which could potentially affect the detection of synchronous ensembles. However, our jittering procedure is specifically designed to account for variations in spike counts. Notably, while the jittered spike trains retain the same spike count variations, we observed 7.8 times more synchronous events in our data compared to the jitter controls (Figure 1G of the manuscript). This indicates that the specific spike timings in the original data - disrupted in the jitter data – are responsible for the observed synchrony.

To further address the concern that complex bursts might influence the observed synchrony, we performed additional analyses in which we excluded all later spikes in bursts, considering only single spikes and the first spikes of bursts. Importantly, this procedure did not affect the rate or size of synchronous ensembles and did not significantly alter the grand-average CCG (Supplementary Figure 3). These results explicitly demonstrate that complex bursts do not significantly impact the analysis of synchronous ensembles.

Result (line 131)

The observed population synchrony was not attributable to spikes in complex bursts, as synchronous event rates did not differ significantly with or without the inclusion of later spikes in bursts (Supplementary Figure 3).

Beyond those methodological issues, there are two main interpretational problems: (1) the "synchronous ensembles" may be completely consistent with phase locking to the intracellular theta (as even shown by the authors themselves in some of the supplementary figures).

We agree with the reviewer that the synchronous ensembles are indeed consistent with theta phase locking. However, it is important to note that theta phase locking alone does not necessarily imply population synchrony. In fact, previous research has demonstrated that theta phase locking can “reduce” population synchrony(30). Thus, the presence of theta phase locking cannot be considered a simple alternative explanation for the synchronous ensembles.

The idea that theta phase locking does not necessarily lead to population synchrony is illustrated in Author response image 1A. In this example, while all three neurons are perfectly locked to specific theta phases, no synchrony among neurons is evident. In contrast, our data align with the scenario depicted in Figure 4B, where spikes occur not only at specific theta phases but also in the same cycles, thereby facilitating population synchrony.

Author response image 1.

Illustrative diagram of the relationship between theta phase coupling and population synchrony. Illustration of theta phase coupling with low population synchrony. Illustration of population synchrony with theta phase coupling.

To directly assess the contribution of theta phase locking to synchronous ensembles, we performed a new analysis in which the specific theta cycles during which neurons spike were randomized while keeping the spike phases unchanged. This manipulation disrupts spike co-occurrence while preserving theta phase locking, allowing us to test whether theta phase locking alone can explain the population synchrony. We found that theta-cycle randomization significantly reduced the rate of synchronous events by 4.5 folds (Supplementary Figure 4). This new analysis demonstrates that theta phase locking alone cannot account for the population synchrony observed in our data.

Result (line 358)

“Correlated intracellular theta and theta-phase locking of the synchronous ensembles raise the question of whether population synchrony among CA1PCs extends beyond synchrony derived from these effects. To address this, we analyzed population synchrony after randomizing the theta cycles during which neurons spiked, while keeping their theta phases unchanged. Supplementary Figure 4 illustrates a significant reduction in synchronous event rates following theta cycle randomization. The finding indicates spiking at specific theta cycles plays a major role in driving population synchrony.”

(2) The definition of "synchrony" in the present work is very loose and refers to timescales of 20-30 ms. In previous literature that relates to synchrony of point processes, the timescales discussed are 1-2 ms, and longer timescales are referred to as the "baseline" which is actually removed (using smoothing, jittering, etc.).

Regarding the timescale of synchronous ensembles, we acknowledge that it varies considerably across studies and cell types. However, it is important to note that a timescale of dozens or even hundreds of milliseconds is commonly used in the context of synchrony terminology for CA1 pyramidal neurons(6,31-33). In fact, a timescale of 20-30 ms is considered particularly important for information transmission and storage in CA1, as it aligns with the membrane time constant of pyramidal neurons, the period of hippocampal gamma oscillations, and the time window for synaptic plasticity. Therefore, we believe this timescale is highly relevant and consistent with established practices in the field.

Reviewer #3 (Recommendations For The Authors):

(1) L19-20: "these synchronous ensembles were not associated with ripple oscillations" - this is a main fallacy in the present work (ripples are from the other side; there are not enough ripples to obtain sufficient statistical power to even test the hypothesis; etc.). The sentence should be removed.

As we have addressed in the public review, most ripples occur synchronously in both hemispheres(3,26). Many studies have used contralateral LFP to infer ripple occurrence on the ipsilateral side(26-29). Moreover, our new data now support the dissociation between synchronous ensembles and ripples with a much larger number of ripples and rigorous statistical testing (Figure 3-figure supplement 3 of the manuscript). These findings support our conclusion that synchronous ensembles are not associated with ripple oscillations.

Result (line 266)

“In total, 1052 synchronous ensembles and 174 ripple episodes were detected across these imaging sessions. Consistent with findings from walking animals, few synchronous ensembles occurred during ripples when animals were immobilized in a tube (Figure 3-figure supplement 3A, B). Moreover, no distinguishable ripple oscillations were observed in synchronous events, and the average firing rates during ripple episodes were near zero (Figure 3-figure supplement 3C, D). At the single-cell level, 90% of neurons showed significant negative spiking modulation during ripples, with ripple modulation indexes close to -1, indicating strong suppression of spiking (Figure 4Ai). This suppression extended to subthreshold membrane potentials, as nearly all cells exhibited decreased fluorescence during ripples compared to baseline (Figure 4Bi, Ci). These results demonstrate that spiking activity and subthreshold membrane potentials are robustly suppressed during ripples.”

(2) L135/Figure 1: panel C and elsewhere: show the same traces after removing (clipping) the spikes. You may be able to see the intracellular theta nicely, which may be very strongly synchronized between neurons and could then be supplemented by ticks (as in conventional raster plots). This will allow a clearer visualization of the spiking and their relations with Vm.

We have created the plot as suggested (Author response image 2). As demonstrated in our figures (Figure 5 in the manuscript), the subthreshold membrane potentials of individual neurons are strongly correlated and coherent at theta frequency, consistent with the reviewer’s viewpoint.

Author response image 2.

Fluorescence traces of 19 simultaneously recorded cells with truncated spikes replaced by dots. Horizontal scale bar: 25 ms; vertical scale bar: -3%.

(3) Related to the above comment, in general, a much more robust approach with the present dataset may be to derive an estimate of the LFP from the intracellular records. Extracellular theta is related to intracellular theta (approximately the negative), and extracellular ripples co-occur with intracellular high-frequency oscillations. However, because the precise transfer function (TF) between the two is not well established, ground truth data should first be collected. This may be done by voltage imaging of even a single neuron in parallel with an extracellular glass pipette placed in near proximity of the same cell, at the same depth. Such datasets have been collected in the past, so it may be sufficient to contact those authors and derive the TF from existing data. Alternatively, new experiments may be required. It is possible that the TF will not be well defined - in which case there are two options: (1) limit the analyses to the relation between spikes in Vm, or (2) record new datasets with true LOCAL field potentials in every case.

We thank the reviewer for the insightful suggestion. Establishing a precise TF between intracellular and extracellular recordings is indeed crucial when exact phase information is required to draw conclusions. However, our goal is to understand the occurrence of specific network oscillation states surrounding these synchronous ensembles, rather than pinpointing the precise phase at which they occur. Therefore, we believe that the strong bilateral cooccurrence of both theta and ripple oscillations provides a practical and valid foundation for supporting our objective.

While the approach suggested by the reviewer is an excellent idea, conducting simultaneous voltage imaging and local LFP recording is currently not feasible due to technical constraints associated with the implanted glass windows. Nevertheless, we recognize the potential value of this approach and plan to incorporate it into future experimental designs, which could provide further insights into the specific oscillatory phases associated with population synchrony.

(4) L135/Figure 1: panel D and elsewhere: Account for second-order spike train statistics (e.g., bursts). The simplest way to do this is to remove all spikes that are not the first spike in a burst. Otherwise, the zero-lag bin of a pair-wise CCG will be filled with counts that are due e.g., to the first spike of the second neuron co-occurring with the last spike in a burst of the first neuron. In other words, without accounting for bursts, sequential activity may be interpreted as synchrony.

We thank the reviewer for this insightful comment. As recommended, we have performed the suggested analysis by removing all spikes that are not the first spike in a burst (Supplementary Figure 3). The results demonstrate that, even after removing the subsequent spikes in bursts, the rates of synchronous events remain unchanged compared to the original data, and the sizes of the synchronous ensembles are also unaffected. These findings indicate that our conclusions are robust and not confounded by the presence of later spikes within bursts.

Result (line131)

“The observed population synchrony was not attributable to spikes in complex bursts, as synchronous event rates did not differ significantly with or without the inclusion of later spikes in bursts (Supplementary Figure 3).”

(5) L135/Figure 1: panel D and elsewhere: Related to the previous comment: the "grand average" CCG of a single neuron with all the other simultaneouslyrecorded neurons is prone to a peak at zero lag ("synchrony") even if all pairs of neurons have pure mono-synaptic connections (e.g., at a 2 ms time lag). This is because neuron1 (N1) may precede N2, whereas then N3 may precede N2. In such a case, the pooled CCG will have two peaks - at e.g., 2 ms and -2 ms. However, if bursts occur (as is the case in CA1 and Figure 1C), there will also be non-zero counts around zero lag, which will accumulate as well. Together, these will build up to a peak around zero - even without any theta phase locking or any other alternative correlations.

Please see our reply to comment #6 below.

(6) L135/Figure 1: panel D and elsewhere: refrain from averaging "grand averages" over neurons. This problem is distinct from the above (where e.g., N2-N1 is averaged with N2-N3). In any case, all visualizations and measures should be derived from individual (pair-wise) CCGs, and not "grand averages"

We thank the reviewer for the detailed comments and appreciate the opportunity to clarify our methods and analyses related to population synchrony. In response to the suggestion to replace grand average CCGs with pairwise CCGs, we have now included a heatmap to visualize individual pairwise CCGs for all recorded neuronal pairs that meet our inclusion criteria (497 pairs, Author response image 3). The heatmap provides a comprehensive view of the temporal relationships between neuron pairs.

Author response image 3.

Color-coded plot of pairwise CCGs for all cell pairs that meet our inclusion criteria.

While we have chosen to keep the grand-average CCGs, we emphasize that they are served only to summarize the overall temporal scale of the population synchrony. Importantly, our conclusions regarding synchronous ensembles are not based on grand-average CCGs. Instead, we assess population synchrony using a rigorous approach: we compute spike counts across the population in 25-ms sliding windows and compare these counts to those derived from jittered data, where spike timings are randomly shifted by ±75 ms while preserving the overall spike count distribution. Synchrony is identified when the original spike counts exceed those from the jittered data by more than 4 standard deviations. This approach accounts for the potential accumulation of zero-lag counts arising from mixed mono-synaptic connections or bursting, as noted by the reviewer. By perturbing spike timings and preserving spike count distributions, our method identifies synchrony beyond what is expected by chance, ensuring robust and artifact-free conclusions.

(7) L135/Figure 1: panel D and elsewhere: after deriving measures (peak lag, FWHM, synchrony strength, etc.) from individual pairwise CCGs, show the measures as a function of the spike counts. For a pair of neurons N1-N2, derive the geometric mean spike count (or the mean, or the max). For instance, if there are 500 pairs of neurons, show e.g., pairwise synchrony strength as a function of the spike count geometric mean. While little correlation is expected when the timescale is small (1-2 ms), the "synchrony" effect at a timescale of 20-30 ms is expected to be very strongly related to the spike counts. Because the spike counts may differ between the lower and higher speed "states", many results reported in the present manuscript may be an epiphenomenon of that relationship.

We thank the reviewer for these valuable comments. In response, we analyzed pairwise synchronization strengths as a function of spike counts geometric mean of neuron pairs, as suggested. As shown in Author response image 4, the CCG peak counts in the original data (red dots) increase with the spike count geometric mean, consistent with the expected trend. However, this trend is also captured by the jitter control (black dots), which reflects synchrony levels expected by chance given the spike count levels.

Importantly, the normalized synchronization strengths - defined as the ratio of CCG peak counts in the original data to the jitter control – are not positively correlated with spike counts and remain significantly greater than 1 across all spike count levels (Author response image 5). This demonstrates synchrony beyond what could be explained by spike count variations alone.

While we understand the potential influence of state-dependent spike count variations, our jittering approach effectively controls for this by removing chance-level synchrony that could arise from these variations. This ensures that the observed synchrony reflects genuine neuronal interactions rather than an epiphenomenon of spike count variations between states.

Author response image 4.

Plot of peak spike counts of pairwise CCGs (red) and mean spike counts from jittered data (black) against geometric means of pair spike counts.

Author response image 5.

Plot of normalized synchronization strengths against spike count geometric means.

(8) L135/Figure 1: show all CCGs in a color matrix.

We have generated a color matrix visualization of all pairwise CCGs, as recommended (Author response image 3). This visualization highlights the consistency of our results across neuron pairs.

(9) L168/Figure 2: the LFPO is nearly irrelevant - it is from the other hemisphere, and it is unclear whether the depth is the same as in the "deep" (closer to the brain surface) imaging plain used for the voltage recordings.

As previously explained, the LFPO is relevant because it reveals the occurrence of theta and ripple states, which are highly synchronous across both hemispheres and serve as reliable indicators of network states relevant to our findings.

(10) L222/Figure 3: The ripple-related analyses are completely irrelevant - ripples are a local phenomenon, and recording from the other hemisphere is completely irrelevant.

We thank the reviewer’s suggestions. As we have explained in the public review, as well as in the reviewer’s comments #1 and #3, the occurrences of theta and ripple oscillations are well-coordinated across hemispheres. As our analyses only depend on the occurrences of these oscillations, our conclusions regarding the association of the synchronous ensembles with theta but not ripple oscillations are supported by data.

(11) L292/Figure 4, panels A-E: please trigger Vm on the same-neuron spikes, not on the "synchrony events". This will already explain most of the observations. Some of this is already shown in the supplementary figures.

As the reviewer correctly noted, we have already presented data triggered on same-neuron spikes in Figure 5-figure supplement 1C and D. The reason we show synchrony-triggered LFP and subthreshold Vm in the figure is to highlight the network dynamics during synchronous events. This approach provides a broader perspective on how neural networks function and interact during periods of synchrony, offering insights beyond individual neuron activity

(12) L351/Figure 5, panel C: typo - should read "strength"

The typo has been corrected.

(13) L351/Figure 5: show "spatial tuning correlation" vs. inter-soma distance (as in Fig. 4G). This may explain part (if not all) of the observations

We have followed the reviewer’s suggestion and generated the plot (Author response image 6). Consistent with the literature, the plot demonstrates that the spatial tuning correlations of place cell pairs exhibit little relationship with their inter-soma distances.

Author response image 6.

Plot of spatial tuning correlation vs. inter-soma distance (Spearman correlation coefficient=0.06, p\=0.54, n\=91 pairs).

(14) L937/Figure S3: panel A: the ripples here appear to be recorded from the top part of the layer, i.e., the electrode is not in the center of the layer. Panel B: add statistical testing.

We agree with the reviewer that this is possible, as we aimed to place our LFP electrodes in the stratum pyramidale. Regarding panel B of the figure, we verified the quality of LFP recordings by acquiring data from subsequent sessions following the initial imaging sessions. The detection of ripples in the same animals during these later sessions indicates that the absence of ripples during the first sessions is not due to deterioration in LFP recording quality. However, due to the small sample size, the statistical power is insufficient to demonstrate significance (n\=5 sessions, p\=0.06, Wilcoxon signed-rank test). Nevertheless, our conclusions are not contingent upon achieving statistical significance in this test.

(15) L944/Figure S4: The "R=1" is very likely to be an outcome of n=1 spike. In other words, estimates of phase are unreliable when the spike count is very low. This is related to the problem referred to in Comment #7 above.

We understand that phase estimates can be unreliable when the spike counts are low. We now highlight that this effect has been taken into account by a shuffling procedure that assesses the significance of phase modulation, and by excluding neurons with nonsignificant modulation strengths. Neurons with low spike count or inconsistent spike phases are typically excluded due to the non-significant strength of phase modulation.

Method (line 828)

“The significance of the modulation strength was tested by shuffling the spike timings and recalculating the modulation strength a thousand times to generate a distribution based on the shuffled spike timings. The original modulation strength was then compared to the distribution, with significance determined if it exceeded the 95% confidence interval of the shuffled values.

Significant modulation strengths were plotted and compared across groups.”

(16) L944/Figure S4: Putting the spike count issue (Comment #15) aside for a moment, the analyses in this figure are actually valid - they are carried out at the single-neuron level, with respect to the local (same-neuron) Vm. These findings provide a key alternative explanation to the observations purported in the main figures: (1) if spiking is locked to intracellular theta (occurring at the peak of Vm); and if (2) intra-cellular (Vm) theta is locked to extracellular theta (antiphase); and if (3) extracellular theta is similar for nearby neurons (the imaged neurons), then synchrony is a necessary outcome. The key question is then whether there is any EXTRA synchrony between the CA1PC - beyond that which necessarily derives from (1)+(2)+(3).

We acknowledge the reviewer’s perspective. However, the factors (1)+(2)+(3) alone do not account for the synchrony we observed. As the reviewer points out (and as discussed in our response to the public review and in Supplementary Figure 4), theta phase locking does not necessarily imply population synchrony. To demonstrate that population synchrony extends beyond the contribution of (1)+(2)+(3), we performed an analysis where the theta cycles in which neurons spike were randomized, while the theta phases remained unchanged (Supplementary Figure 4). The analysis revealed that randomizing the theta cycles while preserving theta phases significantly reduces population synchrony. This finding indicates that spiking in specific theta cycles plays a major role in driving population synchrony.

Result (line 358)

“Correlated intracellular theta and theta-phase locking of the synchronous ensembles raise the question of whether population synchrony among CA1PCs extends beyond synchrony derived from these effects. To address this, we analyzed population synchrony after randomizing the theta cycles during which neurons spiked, while keeping their theta phases unchanged. Supplementary Figure 4 illustrates a significant reduction in synchronous event rates following theta cycle randomization. The finding indicates spiking at specific theta cycles plays a major role in driving population synchrony.”

(17) L944/Fig. S4: Why 71 neurons in AB and only 59 in CD?

In the previous version, panels A and B included 71 neurons, as we collected data from 71 cells across 5 mice (see the text below).

Result (line 93)

“…in total, 71 cells imaged from 5 fields of view in 5 mice; Figure 1B and

Supplementary Figure 1A and 1B).”

In the current version, we only include neurons with significant modulation strengths, reducing the number of cells from 71 to 65 in panel A and from 71 to 54 in panel B.

Methods (line 828)

“The significance of the modulation strength was tested by shuffling the spike timings and recalculating the modulation strength a thousand times to generate a distribution based on the shuffled spike timings. The original modulation strength was then compared to the distribution, with significance determined if it exceeded the 95% confidence interval of the shuffled values. Significant modulation strengths were plotted and compared across groups.”

“Figure 5-figure supplement 1 Figure legend (line 1231)

Polar plot comparing subVm theta modulation between spikes participating in synchronous ensembles (sync spikes) and spikes not participating in synchronous ensembles (other spikes) during immobility. Each dot represents the averaged modulation of a cell. Cells with modulation strengths that are not significant are excluded in the plot and in the comparison.”

For panels C and D, we excluded neurons with four or fewer triggering events from the analysis, which reduced the number of cells from 71 to 59 (see the second text paragraph below).

Method (line 835)

“We extracted segments of fluorescence traces using a ±300 ms time window centered on the spike timings. To examine variations in fluorescence waveforms triggered by spikes within and outside synchronous events, we categorized the fluorescence traces based on whether the spikes occurred within or outside these events. Subsequently, we performed pairwise comparisons of the fluorescence values from the same neuron, concentrating on spikes occurring during corresponding behavioral states. Neurons with four or fewer triggering events in any of these categories were omitted from the analysis.”

(1) Mizuseki, K. & Buzsaki, G. Preconfigured, skewed distribution of firing rates in the hippocampus and entorhinal cortex. Cell Rep 4, 1010-1021 (2013). https://doi.org:10.1016/j.celrep.2013.07.039

(2) McHugh, T. J., Blum, K. I., Tsien, J. Z., Tonegawa, S. & Wilson, M. A. Impaired hippocampal representation of space in CA1-specific NMDAR1 knockout mice. Cell 87, 1339-1349 (1996). https://doi.org:10.1016/s0092-8674(00)81828-0 3

(3) Buzsaki, G. et al. Hippocampal network patterns of activity in the mouse. Neuroscience 116, 201-211 (2003). https://doi.org:10.1016/s03064522(02)00669-3

(4) Karlsson, M. P. & Frank, L. M. Network dynamics underlying the formation of sparse, informative representations in the hippocampus. J Neurosci 28, 14271-14281 (2008). https://doi.org:10.1523/JNEUROSCI.4261-08.2008

(5) Dombeck, D. A., Harvey, C. D., Tian, L., Looger, L. L. & Tank, D. W. Functional imaging of hippocampal place cells at cellular resolution during virtual navigation. Nat Neurosci 13, 1433-1440 (2010). https://doi.org:10.1038/nn.2648

(5) Malvache, A., Reichinnek, S., Villette, V., Haimerl, C. & Cossart, R. Awake hippocampal reactivations project onto orthogonal neuronal assemblies. Science 353, 1280-1283 (2016). https://doi.org:10.1126/science.aaf3319

(7) Sheffield, M. E. J., Adoff, M. D. & Dombeck, D. A. Increased Prevalence of Calcium Transients across the Dendritic Arbor during Place Field Formation. Neuron 96, 490-504 e495 (2017). https://doi.org:10.1016/j.neuron.2017.09.029

(8) Adam, Y. et al. Voltage imaging and optogenetics reveal behaviour-dependent changes in hippocampal dynamics. Nature 569, 413-417 (2019). https://doi.org:10.1038/s41586-019-1166-7

(9) Go, M. A. et al. Place Cells in Head-Fixed Mice Navigating a Floating RealWorld Environment. Front Cell Neurosci 15, 618658 (2021). https://doi.org:10.3389/fncel.2021.618658

(10) Geiller, T. et al. Local circuit amplification of spatial selectivity in the hippocampus. Nature 601, 105-109 (2022). https://doi.org:10.1038/s41586021-04169-9

(11) Rolotti, S. V. et al. Local feedback inhibition tightly controls rapid formation of hippocampal place fields. Neuron 110, 783-794 e786 (2022). https://doi.org:10.1016/j.neuron.2021.12.003

(12) Pettit, N. L., Yap, E. L., Greenberg, M. E. & Harvey, C. D. Fos ensembles encode and shape stable spatial maps in the hippocampus. Nature 609, 327-334 (2022). https://doi.org:10.1038/s41586-022-05113-1

(13) Hainmueller, T. & Bartos, M. Parallel emergence of stable and dynamic memory engrams in the hippocampus. Nature 558, 292-296 (2018). https://doi.org:10.1038/s41586-018-0191-2

(14) Gauthier, J. L. & Tank, D. W. A Dedicated Population for Reward Coding in the Hippocampus. Neuron 99, 179-193 e177 (2018). https://doi.org:10.1016/j.neuron.2018.06.008

(15) Grosmark, A. D., Sparks, F. T., Davis, M. J. & Losonczy, A. Reactivation predicts the consolidation of unbiased long-term cognitive maps. Nat Neurosci 24, 1574-1585 (2021). https://doi.org:10.1038/s41593-021-00920-7

(16) Farrell, J. S., Hwaun, E., Dudok, B. & Soltesz, I. Neural and behavioural state switching during hippocampal dentate spikes. Nature 628, 590-595 (2024). https://doi.org:10.1038/s41586-024-07192-8

(17) McHugh, S. B. et al. Offline hippocampal reactivation during dentate spikes supports flexible memory. Neuron 112, 3768-3781 e3768 (2024). https://doi.org:10.1016/j.neuron.2024.08.022

(18) Gava, G. P. et al. Organizing the coactivity structure of the hippocampus from robust to flexible memory. Science 385, 1120-1127 (2024). https://doi.org:10.1126/science.adk9611

(19) Galvao, J. et al. Unexpected low-dose toxicity of the universal solvent DMSO. FASEB J 28, 1317-1330 (2014). https://doi.org:10.1096/fj.13-235440

(20) Yuan, C. et al. Dimethyl sulfoxide damages mitochondrial integrity and membrane potential in cultured astrocytes. PloS one 9, e107447 (2014). https://doi.org:10.1371/journal.pone.0107447

(21) Modrzynski, J. J., Christensen, J. H. & Brandt, K. K. Evaluation of dimethyl sulfoxide (DMSO) as a co-solvent for toxicity testing of hydrophobic organic compounds. Ecotoxicology 28, 1136-1141 (2019). https://doi.org:10.1007/s10646-019-02107-0

(22) Hoyberghs, J. et al. DMSO Concentrations up to 1% are Safe to be Used in the Zebrafish Embryo Developmental Toxicity Assay. Front Toxicol 3, 804033 (2021). https://doi.org:10.3389/ftox.2021.804033

(23) Abdelfattah, A. S. et al. Sensitivity optimization of a rhodopsin-based fluorescent voltage indicator. Neuron (2023). https://doi.org:10.1016/j.neuron.2023.03.009

(24) Huang, Y. C. et al. Dynamic assemblies of parvalbumin interneurons in brain oscillations. Neuron 112, 2600-2613 e2605 (2024). https://doi.org:10.1016/j.neuron.2024.05.015

(25) Abdelfattah, A. S. et al. Bright and photostable chemigenetic indicators for extended in vivo voltage imaging. Science 365, 699-704 (2019). https://doi.org:10.1126/science.aav6416

(26) Szabo, G. G. et al. Ripple-selective GABAergic projection cells in the hippocampus. Neuron 110, 1959-1977 e1959 (2022). https://doi.org:10.1016/j.neuron.2022.04.002

(27) Dudok, B. et al. Alternating sources of perisomatic inhibition during behavior. Neuron 109, 997-10¹² e1019 (2021). https://doi.org:10.1016/j.neuron.2021.01.003

(28) Terada, S. et al. Adaptive stimulus selection for consolidation in the hippocampus. Nature 601, 240-244 (2022). https://doi.org:10.1038/s41586021-04118-6

(29) Geiller, T. et al. Large-Scale 3D Two-Photon Imaging of Molecularly Identified CA1 Interneuron Dynamics in Behaving Mice. Neuron 108, 968-983 e969 (2020). https://doi.org:10.1016/j.neuron.2020.09.013

(30) Mizuseki, K. & Buzsaki, G. Theta oscillations decrease spike synchrony in the hippocampus and entorhinal cortex. Philos Trans R Soc Lond B Biol Sci 369, 20120530 (2014). https://doi.org:10.1098/rstb.2012.0530

(31) Csicsvari, J., Hirase, H., Mamiya, A. & Buzsaki, G. Ensemble patterns of hippocampal CA3-CA1 neurons during sharp wave-associated population events. Neuron 28, 585-594 (2000). https://doi.org:10.1016/s08966273(00)00135-5

(32) Harris, K. D., Csicsvari, J., Hirase, H., Dragoi, G. & Buzsaki, G. Organization of cell assemblies in the hippocampus. Nature 424, 552-556 (2003). https://doi.org:10.1038/nature01834

(33) Yagi, S., Igata, H., Ikegaya, Y. & Sasaki, T. Awake hippocampal synchronous events are incorporated into offline neuronal reactivation. Cell Rep 42, 112871 (2023). https://doi.org:10.1016/j.celrep.2023.112871

eLife Assessment

This important study substantially advances our understanding of the neural circuits that regulate social behavior by identifying a population of hypothalamic neurons in the preoptic area that promote social interactions following short-term isolation. The evidence supporting the authors' claims is solid, with well-designed experiments using validated activity-dependent tagging and manipulation methods, though some differences in outcomes between experiments highlight limitations of the tagging approach. The work will be of broad interest to neuroscientists studying social behavior, neural circuit function, and hypothalamic mechanisms and will represent a meaningful contribution to the field.

Reviewer #2 (Public review):

Summary:

This study reveals that short-term social isolation increases social behavior at a reunion, and a population of hypothalamic preoptic area neurons become active after social interaction following short-term isolation (POAsocial neurons). Effectively utilizing a TRAP activity-dependent labeling method, the authors inhibit or activate the POAsocial neurons and find that these neurons are involved in controlling various social behaviors, including ultrasonic vocalization, investigation, and mounting in both male and female mice. This work suggests a complex role for the POA in regulating multiple aspects of social behavior, beyond solely controlling male sexual behaviors.

Strengths

While a few studies have shown that optogenetic activation of the POA in females promotes vocalization and mounting behavior similar to the effects observed in males, these were results of artificially stimulating POA neurons, and whether POA neurons play a role in naturally occurring female social behaviors was unknown. This paper clearly demonstrates that a population of POA neurons is necessary for naturally evoked female social vocalizations and mounting behaviors.

Weaknesses

The authors used various gain-of-function and loss-of-function methods to identify the function of POAsocial neurons. However, there were inconsistent results among the different methodologies. As the authors describe in the manuscript, these inconsistencies are potentially due to limitations of the TRAP activity-dependent labeling method; however, different approaches will be necessary to clarify these issues.

Overall, this paper is well-written and provides valuable new data on the neural circuit for female social behaviors and the potentially complex role of POA in social behavior control.

Reviewer #3 (Public review):

Summary:

The mechanisms by which short-term isolation influences the brain to promote social behavior remain poorly understood. The authors observed that acute isolation enhanced social behaviors, including increased investigation, mounting, and ultrasonic vocalizations (USVs). These effects were evident in same-sex interactions among females and in male-female interactions. Concurrently, cFos expression in the preoptic area (POA) of the hypothalamus was selectively elevated in single-housed females. To further investigate, the authors used an innovative tagging strategy (TRAP2) to manipulate these neurons. Overall, the study identifies a population of hypothalamic neurons that promote various aspects of social behavior after short-term isolation, with effects that are sex- and context-dependent.

Strengths:

Understanding the neural circuit mechanisms underlying acute social isolation is an important and timely topic. By employing state-of-the-art techniques to tag neurons active during specific behavioral epochs, the authors identified the preoptic area (POA) as a key locus mediating the effects of social isolation. The experimental design is sound, and the data are of high quality. Notably, the control experiments, which show that chemogenetic inactivation of other hypothalamic regions (AH and VMH) does not affect social behavior, strongly support the specificity of the POA's role within the hypothalamus. Through a combination of behavioral assays, activity-dependent neural tagging, and circuit manipulation techniques, the authors provide compelling evidence for the POA's involvement in behaviors following social isolation. These findings represent a valuable contribution to understanding how hypothalamic circuits adapt to the challenges of social isolation.

Weaknesses:

The authors conducted several circuit perturbation experiments, including chemogenetics, ablation, and optogenetics, to investigate the effects of POA-social neurons. They observed that the outcomes of these manipulations varied depending on whether the intervention was chronic (e.g., ablations) or acute (e.g., DREADDs), potentially due to compensatory mechanisms in other brain regions. Furthermore, their additional experiments revealed that the robustness of the manipulations was influenced by the heterozygosity or homozygosity of TRAP2 animals. While these findings suggest that POA neurons contribute to multiple behavioral responses to social isolation, further experiments are needed to clarify their precise roles.

Reviewer #4 (Public review):

Summary:

Using immunostaining for the immediate early gene Fos, and employing TRAP2-mediated chemogenetic and optogenetic perturbations, the authors provide evidence that neurons in the preoptic hypothalamus, identified as 'POA-social neurons,' promote social behaviors in mice - particularly in socially isolated (or deprived) mice, who exhibit an increased motivation for social investigations.

Strengths:

The focus on female-female social interactions is a valuable contribution to the field, as these interactions are less studied and the underlying neural mechanisms are less understood. The authors should be commended for their comprehensive approach in performing and reporting multiple perturbation experiments, including optogenetics, chemogenetics, and ablation. The authors also deserve recognition for their thoughtful discussion of the nuances in the phenotypes observed across these various perturbation experiments.

Weaknesses:

A limitation of the paper, however, is the insufficient clarification of the specific functions of these POA-social neurons. In my interpretation of the results, the neurons may be crucial for motivated social behaviors in females and motivated mounting of females in males, regardless of whether the test mice are housed singly or in groups. For group-housed mice, the motivation to interact with stimulus mice was likely low in their behavioral paradigm, which may explain the reduced interactions observed in the resident-intruder assay and why these neurons were not tagged (TRAPed) in that setting. Tagging these neurons in singly housed mice following a social interaction, followed by imaging in a group setting where motivated social behaviors do occur, could elucidate whether these neurons are specifically activated during social interactions in socially deprived mice or are generally crucial for motivated social behaviors in any setting. I understand that such calcium imaging may be beyond the scope of this version of the paper, but incorporating these results in a future version would significantly enhance the paper's impact. Depending on the outcomes of such experiments, the title 'Short-term social isolation acts on hypothalamic neurons to promote social behaviors in a sex- and context-dependent manner' may need to be revised to more accurately reflect the findings.

Author response:

The following is the authors’ response to the original reviews.

Reviewer 1:

The main weaknesses of the paper are a lack of significance in key findings, and relatedly, concluding effects from insignificant findings. Additional elements could be improved to help strengthen this overall well-rounded and intriguing set of results.

In the original manuscript, we reported that chemogenetic silencing of POA-social neurons (previously called POA-iso neurons; more details on rationale for renaming below in our responses to reviewer recommendations) tended to reduce mounting in both single-housed female and single-housed male mice, although these effects were non-significant. We have added samples to both datasets and now report that chemogenetic silencing of POA-social neurons significantly reduces the proportion of trials with mounting in both sexes (Fig. 2C and Fig. 6G). 

We have also included new analyses to test whether optogenetic activation of POAsocial neurons in group-housed females promotes social investigation (in addition to USV production, as reported in the original manuscript). We now report that optogenetic activation of POA-social neurons significantly increases the probability of social investigation (Fig. 4E-F) and significantly increases the duration of social investigation bouts (Fig. 4G). 

Additional recommendations from the reviewer are addressed in detail below. Thank you for your critical and insightful feedback.

Reviewer 2:

All the activity-dependent labeling experiments with TRAP mice, including the subsequent neural activity manipulation experiments (Figures 2, 3, 4, 5E-F), were conducted by labeling neurons only in socially isolated animals, not group-housed animals. The authors labeled neurons after 30-minute social interactions, raising the possibility that the labeled neurons simply represent a "social interaction/behavior population" (mediating mounting and USVs in females and males) rather than a set of neurons specific to social isolation.

I strongly recommend including experimental groups that involve labeling neurons after 30minute social interactions in group-housed female or male mice and inhibit TRAPed neurons after social isolation or activate TRAPed neurons after group housing. If manipulating the grouphoused TRAP neurons has similar effects to manipulating the isolated TRAP neurons, it would suggest the current labeling paradigm is not isolating neurons specific to the effect of social isolation per se. Rather, the neurons may mediate more general social interaction or motivationrelated activities. Given the known role of POA in male mating behavior, a group-housed TRAP experiment in males with a female visitor is especially important for understanding the selectivity of the labeled cells.

Without proper controls, referring to the labeled neurons as "POAiso" neurons is potentially misleading. The data thus far suggests these neurons may predominantly reflect a "POA social behavior" population rather than a set of cells distinctly responsive to isolated housing.

We agree with the reviewer that the POA neurons we are studying regulate the production of social behaviors in females and males, rather than representing a set of cells distinctly responsive to single housing. To more clearly reflect our thinking, we have changed the name of the neurons from “POA-iso neurons” to “POA-social neurons”. Thank you for this helpful criticism.

Our Fos data are consistent with the idea that the POA may regulate social behaviors in group-housed females (not just single-housed females). Namely, we found that counts of Fospositive POA neurons are significantly related to rates of social investigation (p = 0.01) and tend to be related to USV rates (p = 0.05) in group-housed females that engaged in same-sex interactions (Fig. S1C). We now include two new sets of experiments aimed at further testing the idea this idea. 

First, we include 2 control groups in which TRAPing sessions were performed in grouphoused females following same-sex interactions. We find that chemogenetic silencing of grouphoused-TRAPed POA neurons fails to reduce social behaviors in females that are subsequently single-housed and given a same-sex social interaction (Fig. 5A-D), and that optogenetic activation of group-housed-TRAPed POA neurons fails to promote female social behavior (Fig. 5E-H). At face value, these findings do not support the idea that the POA contains neurons that regulate social behaviors in group-housed females.

However, one important caveat is that group-housed females engage in low rates of social behaviors (low investigation time, no mounting, and few USVs), and thus TRAP-based labeling may not work efficaciously in these mice. There may be POA neurons that regulate social behaviors in group-housed females but that do not upregulate Fos following production of relatively low rates of social behaviors. To test this idea, we also include females in which POA neurons are chemogenetically silenced using a viral strategy that does not depend on activitydependent labeling. In this new experiment, we report that silencing of POA neurons significantly reduces USV production in group-housed females (Fig. 5J-L) and significantly reduces social investigation, mounting, and USV production when these same females are retested following single-housing (Fig. 5M-O). Together, these experiments suggest that the POA may regulate the production of social behaviors during same-sex interactions in group-housed females, but that these effects may be difficult to detect in some cases given the low rates at which group-housed females engage in social behaviors during same-sex interactions relative to single-housed females.

Finally, we want to highlight an additional new dataset that supports the idea that POAsocial neurons regulate social behaviors, rather than encoding the “state” of social isolation. We now include a control group for the chemogenetic silencing of female POA-social neurons, in which females were single-housed but were not given a social interaction prior to 4-OHT treatment (N = 5 non-social controls). Rates of social behaviors were subsequently unaffected following CNO delivery in these females (Fig. S2D-G). These new data support the conclusion that POA-social neurons regulate the production of social behaviors, rather than encoding the state of social isolation. 

Reviewer 3:

While the authors should be commended for performing and reporting multiple circuit perturbation experiments (e.g., chemogenetics, ablation), the conflicting effects on behavior are hard to interpret without additional experiments. For example, chemogenetic silencing of the POA neurons (using DREADDs) attenuated all three behavioral measures but the ablation of the same POA neurons (using CASPACE) decreased mounting duration without impacting social investigation or USV production. Similarly, optogenetic activation of POA neurons was sufficient to generate USV production as reported in earlier studies but mounting or social investigation remained unaffected. 

Do these discrepancies arise due to the efficiency differences between DREADD-mediated silencing vs. Casp3 ablation? Or does the chemogenetic result reflect off-manifold effects on downstream circuitry whereas a more permanent ablation strategy allows other brain regions to compensate due to redundancy? It is important to resolve whether these arise due to technical reasons or whether these reflect the underlying (perhaps messy) logic of neural circuitry. Therefore, while it is clear that POA neurons likely contribute to multiple behavioral readouts of social isolation, understanding their exact roles in any greater detail will require further experiments.

We have added new analyses to consider the possibility that optogenetic activation of female POA-social neurons promotes social investigation. In the original manuscript, we analyzed the duration of social investigation bouts in POA-social-ChR2 females according to whether they overlapped with laser stimulation or whether they did not overlap. We realized that we made an error in this first analysis and inadvertently included social investigation bouts that occurred during the first 5 minutes of the social sessions, prior to any laser stimulation. Because these earlier bouts tend to be longer duration than later bouts, this mistake washed out the effect of laser stimulation on social bout duration. After correcting that error, we now report that optogenetic activation of female POA-social neurons lengthens social investigation bout duration (Fig. 4G). Inspired by this interesting finding, we also included analyses of the probability of social investigation following laser stimulation (Fig. 4E-F; excluding laser stimulations that were preceded by social investigation in the pre-laser baseline period). These analyses support the conclusion that optogenetic activation of POA-social neurons promotes both USV production and social investigation in group-housed females.  

The majority of the females that we used in our TRAP2-based ablation experiments were heterozygous for TRAP2 (N = 11 of 15 POA-social-caspase subjects were TRAP2;Ai14 females), whereas all females used in our chemogenetic silencing experiments were homozygous for TRAP2. To test whether a more effective ablation of POA-social neurons might drive decreases in social investigation and USV production, we set up additional TRAP2 homozygous POA-social-caspase females and directly compare the effects of ablation between the two genotypes (Fig. S3; N = 11 hets in total and N = 9 homozygotes in total). These experiments revealed that effects on mounting were more pronounced following POA-social ablation in TRAP2 homozygotes vs. heterozygotes, but that neither group exhibited decreased social investigation or USV production following 4-OHT treatment.

To ask whether caspase-mediated ablation in TRAP2 homozygotes was effective in eliminating neural activity associated with social behaviors in females, we performed Fos immunostaining in a subset of the POA-social-caspase TRAP2 homozygotes following a samesex interaction. We found that POA Fos expression was robustly reduced in these females relative to control group-housed and control single-housed females that also engaged in samesex interactions, down to levels seen in group-housed and single-housed females that did not engage in a social interaction (comparison shown in Fig. S3D; control female data same as in Fig. 1). Moreover, the remaining POA Fos in these TRAP2 homozygotes was no longer positively correlated to social investigation or USV production (Fig. S3E-F). Together, these findings lead us to favor the interpretation suggested by the reviewer below, that permanent ablation of POA-social neurons leads to compensation from other brain regions due to redundancy. In addition, our finding that optogenetic activation of POA-social neurons promotes both USV production and social investigation supports the idea that POA-social neurons directly regulate these behaviors. We agree with the reviewer that additional work is needed to understand the complex sex- and context-dependent role played by the POA in the regulation of mouse social behaviors.

Recommendations for the Authors:

Reviewer 1 Recommendations:

(1) The largest issue is that many of the stated "key" behavioral findings are not statistically significant.

(1a) Figure 2C is not significant and Figure 5G is not significant

We have added N = 5 POA-social-hM4Di females, N = 3 POA-social-hM4Di males, and N = 3 POA-social-GFP males to the dataset. The decrease in mounting following chemogenetic silencing of POA-social neurons is now statistically significant in both sexes (p < 0.05 for both; see current Figs. 2C and 6G). We also simplified our statistical analysis of mounting in these experiments to consider the proportion of trials with and without resident-initiated mounting on saline vs. CNO days, using McNemar’s test for paired proportions. 

(1b) Mounting graphs are completely omitted in Figure 4. 

Given that mounting was only observed infrequently in POA-social-ChR2 females, we simply report this information in the Results text (lines 382-388). In our prior summary of the mounting results, we reported that mounting was observed in a total of 3 trials from 2 females, but we inadvertently included information from a duplicate trial from one of the POA-socialChR2 females in this summary (all other analyses of the POA-social-ChR2 females included one trial per female). We have corrected that error and now report that we observed mounting following laser stimulation in 1 trial from 1 POA-social-ChR2 female. We have expanded our consideration of potential effects of optogenetic activation of POA-social neurons on social investigation and include these new analyses as part of Figure 4 (Fig. 4E-G), following the existing analyses of USV production.

(1c) Figure 3C shows a reduction of mounting following the ablation of POA (although no stats on the graph to denote significance), but this ablation approach can't resolve whether POA is required to encode the state produced by the short period of isolation, and/or whether it needs to be online at test.

We have now added an asterisk in Fig. 3C to denote a p value less than 0.05. Thank you for catching our oversight.

We designed our activity-dependent labeling experiments to TRAP and express viruses in POA neurons that increase their activity in conjunction with the production of social behaviors in single-housed females. We believe our findings our most consistent with the conclusion that these neurons regulate the production of social behaviors, rather than encoding the state of social isolation, and we have renamed these neurons as “POA-social” neurons to better reflect our thinking.

We also now include control experiments (albeit chemogenetic inhibition, not caspase ablation) in which the TRAP2 strategy is used to express hM4Di in the POA of single-housed females that do not experience a social interaction prior to 4-OHT delivery (non-social controls, Fig. S2D-G). We report that chemogenetic inhibition of these neurons does not decrease social behavior in single-housed females during a subsequent same-sex interaction (p > 0.05 for saline vs. CNO rates of social investigation, mounting, and USVs). These additional findings support the idea that the activity of POA-social neurons is related to the production of social behaviors rather than to the state of social isolation. 

The reviewer is correct that our ablation approach cannot resolve the question of whether POA-social neuronal activity is required online during testing, but our reversible chemogenetic inhibition experiments provide evidence that the activity of POA-social neurons is required online at the time of testing to regulate social behavior.

(1d) A similar issue is seen regarding investigation (a general lack of significance with most of the LOF and GOF manipulations).

As reported in the original manuscript, we find that chemogenetic inhibition of POAsocial neurons reduces social investigation in females, while caspase-mediated ablation of female POA-social neurons does not. Our original caspase dataset used mostly but not all TRAP2 heterozygous females (N = 11 TRAP2 heterozygotes (TRAP2;Ai14), generated by crossing TRAP2 mice with Ai14 mice, for the purpose of visualizing the absence of tdTomato labeling to estimate spread of the caspase virus; and N = 4 TRAP2 homozygotes). By adding to the TRAP2 homozygous caspase dataset and comparing the effects on female social behavior of ablation of POA-social neurons in TRAP2 heterozygous vs. TRAP2 homozygous females, we

now provide evidence that the attenuation of mounting is more efficacious in TRAP2 homozygous females than in heterozygotes (Fig. S3B). Nonetheless, we fail to see effects on social investigation and USV production, even when caspase ablation of POA-social neurons is performed in TRAP2 homozygous females (Fig. S3A,C). 

In spite of the lack of effect on these behaviors, we show that caspase-mediated ablation of POA-social neurons in TRAP2 homozygous females leads to a dramatic reduction in social interaction-induced Fos expression in the POA. POA Fos expression in these caspase females is reduced to the levels seen in control group-housed and single-housed females that are not given social interactions and are significantly lower than Fos expression in group-housed and single-housed females that are given a same-sex interaction (Fig. S3D). Moreover, the remaining POA Fos expression in the caspase females is no longer related to rates of social investigation (Fig. S3E), as is normally the case in group-housed and single-housed control females (Fig. S1C, left). Together, these data support the idea that some type of neuronal compensation outside of the POA is occurring following ablation of POA-social neurons, and this compensation permits normal levels of USV production and social investigation.

As in the original manuscript, we report that chemogenetic inhibition of POA-social neurons in male mice reduces mounting but does not reduce social investigation (or USV production). We now include quantification of social behaviors produced by male and female POA-social-hM4Di mice in the TRAPing sessions that preceded 4-OHT delivery (Fig. S5). These measurements show that males spent significantly more time than females engaged in mounting, and we speculate that this bias in TRAPing session behavior might have led to a bias in TRAP-mediated viral labeling of male POA neurons that regulate mounting, at the expense of male POA neurons that regulate social investigation (or USV production).

We have added new analyses to consider the possibility that optogenetic activation of female POA-social neurons promotes social investigation. In the original manuscript, we analyzed the duration of social investigation bouts in POA-social-ChR2 females according to whether they overlapped with laser stimulation or whether they did not overlap. We realized that we made an error in this first analysis and inadvertently included social investigation bouts that occurred during the first 5 minutes of the social sessions, prior to any laser stimulation. Because these earlier bouts tend to be longer duration than later bouts, this mistake washed out the effect of laser stimulation on social bout duration. After correcting that error, we now report that optogenetic activation of female POA-social neurons lengthens social investigation bout duration (Fig. 4G). Inspired by this encouraging finding, we also included analyses of the probability of social investigation following laser stimulation (Fig. 4E-F; excluding laser stimulations that were preceded by social investigation in the pre-laser baseline period). These analyses support the conclusion that optogenetic activation of POA-social neurons promotes both USV production and social investigation in group-housed females.

(2) In Figure 1 and elsewhere, the authors use a Mann-Whitney U test, which should be used for non-parametric data, but in other places, they use statistical tests for normally distributed data. Why? How was the normality of distributions tested?

We tested the normality of data distributions using the Shapiro-Wilk test. Parametric tests were used for analyses that contained normally distributed data, and non-parametric tests were used for analyses that contained non-normally distributed data. This information is included in the Methods (lines 997-1000), and full details of statistical analyses can be found in Table S1.

(3) The method for "trapping" neurons that are part of the short-term isolation ensemble has some caveats that have not been adequately addressed. First, 4-OHT was administered after social interaction, but before 24 hours of isolation, making it unclear exactly WHAT is being trapped.

i) Is it neurons that encode the recent 3-day iso experience? (seems unlikely, as this would have been hours after the end of that iso window)

We now include a group of control females to directly test this possibility (Fig. S2D-G). These TRAP2 females were single-housed for 3 days but were not given a social interaction prior to 4-OHT treatment (N = 5 non-social controls). Presumably, POA neurons TRAPed in these females might encode the experience of short-term isolation. However, we found that chemogenetic inactivation of these TRAPed neurons during a subsequent same-sex interaction failed to decrease social behaviors in single-housed females (Fig. S2E-G; p > 0.05 for CNO vs. saline rates of social investigation, mounting, and USV production). These control experiments support the idea that we are TRAPing neurons whose activity is related to the production of social behaviors, and we have renamed the neurons as “POA-social” neurons to reflect this thinking.

ii) Is it neurons that encode the recent behavior impacted by the 3-day iso? (this seems to be the goal, but the authors do not provide evidence that the time course of their injection is efficient enough to recruit the recently activated neurons, nor do they provide evidence that opening the trapping window directly after the behavior is better than directly before)

We opted to perform IP injections of 4-OHT immediately following the behavior session, rather than behavior, due to concern that handling the mice and delivering IP injections prior to behavior sessions would stress the mice, leading to lower rates of social behaviors. The nonsocial female hM4Di experiments described above support the idea that we are TRAPing neurons related to the production of social behaviors, as the reviewer suggests. 

iii) Is it trapping neurons active during the subsequent 24 hours of isolation? (seems possible, but this would mean that the authors are looking at a different population of neurons than they claim).

If chemogenetic silencing of POA neurons that were TRAPed following 3-days of social isolation but in the absence of a social interaction (N = 5 non-social controls, Fig. S2D-G) does not alter social behaviors, there is no compelling reason to hypothesize that TRAPing POA neurons activated following the 24 hours of social isolation that follow a social interaction would do so. Moreover, in the original study characterizing the TRAP2 mice (DeNardo et al., 2019), the authors performed experiments to characterize the time course of TRAPing relative to 4-OHT treatment and concluded that the majority of TRAPing occurs within a 6-hour window centered around the 4-OHT injection.

(4) Relatedly, the authors seem to find a fair bit of variability in their TRAP-mediated experiments. This begs the question - are the effects of their GOF and LOF approaches

i) dependent on the iso-behaviors that were "trapped" for each animal (in other words, how does behavior at test 1 correlate with behavior at test 2)? 

To test the reviewer’s idea, we compared rates of TRAPing session behaviors for the POA-social-hM4Di females to the subsequent effects of neuronal silencing on these behaviors (calculated as (CNO behavior – saline behavior). These correlations are shown in Fig. S2A-C and are all non-significant. We also include below for the reviewer the same types of correlations for the other datasets in our study (loss-of-function experiments: female POAsocial-caspase, male POA-social-hM4Di; and gain-of-function experiments: female POA-socialChR2).

Author response image 1.

The only loss-of-function experiment comparison in the above figure that reveals a negative and significant correlation is the mounting comparison for the POA-social-hM4Di males (time spent mounting during TRAPing session vs. (CNO time spent mounting -saline time spent mounting). This significant correlation likely reflects that fact that (1) no males mounted in the CNO session and (2) that mounting rates for individual males are relatively consistent over time (in comparison to female mounting, which is more variable; see Author response image 2 below of TRAPing session vs. saline mounting in male vs. female POA-social-hM4Di experiments). The correlation between TRAPing session and testing session mounting is significant for the POA-social-ChR2 females, but despite the significant correlation, we would want to see more instances of optogenetically-elicited mounting to make any claim about its relationship to TRAPing session behavior.

Author response image 2.

Nonetheless, we agree with the reviewer’s intuition that one would expect the effects of POA activity manipulations on different behaviors to scale with rates at which these behaviors were performed during the TRAPing session. We speculate that variability in the TRAPing process might have obscured such a relationship. There is inevitable variability in the exact body cavity placement of IP injections, which can affect drug absorption, and another point is that we delivered a fixed volume of 4-OHT (10 mg/mL 4-OHT in 150 uL filtered corn oil) to all mice in the study, regardless of their weight, which likely added variability in TRAPing efficacy from animal to animal. This detail was reported inaccurately in the Methods, and that error has been corrected (line 920). With regard to our male POA-social-hM4Di dataset, we find that these males spend more time mounting during their TRAPing sessions than female POA-socialhM4Di (Fig. S5; males also spent less time investigating and tended to produce fewer USVs than females), a fact that we hypothesize may have led to a bias toward TRAPing mountingrelated POA neurons in male subjects. In addition, however, the fact that male mice typically weigh more than females and would have received a slightly lower effective dosage of 4-OHT may also have contributed to the weaker effects on behavior in the male POA-social-hM4Di experiments relative to the female POA-social-hM4Di experiments.

We also want to highlight that interpreting correlations for females between time spent mounting during the TRAPing session and time spent mounting during the test sessions can be complicated. For example, we see 2 cases in the female POA-social-hM4Di dataset in which the female did not mount in the TRAPing session, and then mounted on the saline day (12s and 10s total mounting for those 2 females) but not on the CNO day. One interpretation of the data from these 2 females is that mounting on the TRAPing day is not required to attenuate mounting on the later test days. However, female mounting behavior itself is variable, both across different females and across different tests of a given female, as noted above. If we consider all singlehoused females included in our dataset for which we quantified control behavioral data (i.e., behavior trials from unmanipulated females and TRAPing sessions from females that were later manipulated), we find that mounting is not observed in ~30% of the females (24 of 83). In ongoing behavioral experiments not included in this manuscript, we are investigating factors that regulate female mounting following single-housing. In that dataset, we also see little evidence that female mounting in one social interaction predicts mounting in a subsequent interaction

(i.e., there don’t appear to stable “high mounters” and “low mounters” following single housing). Thus, the small number of cases in which females did not mount in the TRAPing session and then displayed mounting on the CNO only day are difficult to interpret. 

Two additional considerations are that TRAPing may not be equally efficacious for POA neurons that regulate different behaviors, and that different behaviors may be differentially sensitive to perturbations of the POA. Previous elegant calcium imaging work has shown that different subsets of Esr1+ POA neurons exhibit activity that is “tuned” to specific behaviors (sniffing vs. mounting in males interacting with females; Yang et al., 2023). However, it is possible that these subsets of neurons display differential levels of Fos expression following the production of their preferred behavior and that some behavior-related subsets may thus be more easily TRAPed than others. It may also be the case that some behaviors are more easily disrupted by POA activity manipulations than others (e.g., perturbation in a smaller percentage of behavior-related POA neurons may be required to disrupt some behaviors relative to others). 

Despite these caveats, we have two lines of evidence that the effects of chemogenetic silencing of POA-social neurons depends on the behaviors produced during the TRAPing sessions.

(1) Social behavior is required during the TRAPing session to see subsequent effects on social behavior following chemogenetic silencing of TRAPed POA neurons. In control females that were single-housed but were not given a social interaction prior to 4OHT treatment, social behaviors are not reduced by chemogenetic silencing of TRAPed POA neurons (Figs. S2D-G).

(2) To directly test whether mounting in the TRAPing session is required to see attenuation of mounting during subsequent chemogenetic silencing of POA-social neurons, we performed control experiments in which single-housed females interacted with a female visitor that was placed under a cup during the TRAPing session prior to 4-OHT treatment. Mounting was not possible in this context, and we also found that females produced lower rates of USVs during the TRAPing session relative to single-housed females engaged in free social interaction. However, subject females spent more time engaged in social investigation of the visitor relative to single-housed females engaged in free social interactions (see Author response image 3 below).

Author response image 3.

Unfortunately, none of the experimental females in this cohort displayed mounting in the CNO or saline sessions. Given that we could use this dataset to address the intended question, we did not include it in the manuscript. However, it is quite interesting that female subjects displayed higher than normal social investigation and lower than normal USV production in their TRAPing sessions (relative to single-housed females engaged in free interactions), and subsequently, chemogenetic inhibition of TRAPed POA neurons decreased social investigation but did not decrease USV production (Author response image 4 below). 

Author response image 4.

Together, we think our data support the idea that the POA neurons that are TRAPed are related to the social behaviors performed by the animals, but these relationships may be complex and difficult to detect from comparisons across animals within a single experimental group.

And/or are they

ii) influenced by the spread or amount of virus for each animal? These correlations could help shed light on what exactly is being trapped - is it specific behaviors or is it the "state" of shortterm isolation?

Our control experiments with females that were single-housed but did not receive a social interaction prior to 4-OHT treatment provide evidence that the production of social behaviors is required to see subsequent effects on behavior following chemogenetic inhibition of TRAPed POA neurons (Figs. S2D-G).

The same volume of virus was injected across all activity manipulation experiments (200 nL). Because of the trajectory of our POA viral injections (performed at a slight rostral angle relative to vertical), we did sometimes see viral labeling that spread into the AH caudal to the POA. For this reason, we included the AH TRAPed control group (Fig. 2), to rule out the possibility that viral spread into the AH could account for the effects of chemogenetic silencing of POA-social neurons on female social behaviors. Also because of the injection angle used, we don’t see substantial viral spread rostral to our injection coordinates. In short, there isn’t systematic variability in the targeting or spread of our POA viral injections that can account for variability in the effects on USV production and social investigation of our LOF and GOF manipulations (female hM4Di and female ChR2 experiments).

In older lesion studies in male rodents and birds, there is some support for the idea that rostral vs. caudal POA neurons differentially regulate appetitive vs. consummatory sexual behaviors (as reviewed in Balthazart and Ball, 2007). However, all of our viral injections were placed in what that review paper would have considered ‘caudal’ POA. We also note that more recent imaging studies have reported that subsets of POA neurons are differentially tuned to male sniffing vs. male mounting (Yang et al.,2023), and these subsets must be relatively co-localized given that they are imaged in the same field of view. Whether distinct subsets of POA neurons regulate the production of different female social behaviors, and if so, how these subsets are localized within the POA, remains an important question for future study.

(5) The authors label their region of interest as the "POA" but images throughout (e.g. their fos image, Figure 1E), look more like the MPO. Why label it POA?

The POA neurons in our study are found in a band that spans the medial POA, as well as a bit of the lateral POA. To avoid over-specifying, we call this region the POA more generally.

(6) In all the experiments, mice are isolated and then re-group housed with siblings. Do all the siblings in the group belong to the same experimental group, or are siblings naïve? This may be critical to help determine whether some of the effects observed may be "group" effects.

In general, multiple (although not always all) mice in a cage belonged to the same experimental group. In our inhibitory DREADDs experiments, it is unclear how that could drive our observed effects on behavior, given that home cage behavior would only be expected to differ for a given mouse in the time period following their CNO session. 

For the female POA-social-caspase mice, we cannot rule out the possibility that their home cage behaviors differed in the time period following 4-OHT treatment and re-grouphousing and prior to post-4-OHT behavior measurements. However, given that the only social behavior affected by ablation of POA-social neurons was mounting, and that rates of mounting would be expected to be very low in group-housed females within home cages, it is unclear how our experimental result could be attributed to group effects.

If by “group” effects the reviewer means “litter” effects, we include a plot below that shows the CNO vs. saline behaviors for the POA-social-hM4Di females, separated by cage ID. There is no evidence that the effects of chemogenetic silencing of POA-social-hM4Di females are being driven by only certain cages (only social investigation and USVs are shown, because mounting was uniformly low (1 of 17 females mounted) in the CNO session).

Author response image 5.

(7) For chemogenetic experiments, the authors state that CNO and Saline were given in a counterbalanced order (eg line 189). Did the authors see any order effects?

We did not see order effects, and we can include plots of those data below for the female and male POA-social-hM4Di groups, with mice plotted according to which treatment they received first.

Author response image 6.

(8) In the control experiments in Figure 2 where VMH or AH are chemogenetically silenced, it isn't clear whether these groups include mice that were subjected to 3 days of isolation. Please clarify.

Yes, these female groups were also subjected to 3 days of isolation (first prior to the TRAPing session, and for a second time prior to the onset of the CNO/saline testing sessions). That information has been clarified in the Results section (line 214) and in the Methods (lines 935-938).

(9) Line 312. The title for this section, "POA neurons increase their activity....." is somewhat misleading. It sounds like the authors imaged trapped neurons. I think what they mean is that more POA neurons are activated following opposite-sex interactions with males.

Thanks for this catch. We have modified the section title, as well as the title of the first results sub-section.

(10) Figure 5A, right panels. The authors fail to find an increase in the investigation of male-male pairs following the short-term isolation of one. This contrasts with the main finding in Matthews et al., 2016 Cell, where short periods of isolation are said to promote pro-social behaviors. The authors could comment on this discrepancy in their discussion (eg difference in testing apparatus/test type? Difference in the number of days of isolation? etc.).

In current Fig. 6A, there is no significant interaction between the two main effects, but each main effect is significant: single-housed males spend more time investigating partners than group-housed males, and males spend more time investigating female partners than male partners. The significant main effect of housing condition is consistent with the findings of Matthews et al., 2016 and is included within the Results (lines 486-492). 

(11) Figure 5F, the authors seem to have a main effect of virus (more overall investigation in dreadds mice). Nothing about this is addressed.

We sometimes see differences in social behavior between cohorts of males when they are tested at different times and, correspondingly, with different groups of female social partners. Our POA-social-hM4Di and POA-social-GFP males were set-up and tested at largely non-overlapping times. We have added a brief note to the Results section to include this information (lines 535-539).

Reviewer 2 Recommendations:

(1) (C)ritical control experiments are missing to support this claim (that a population of preoptic hypothalamic neurons contribute to the effects of short-term social isolation on the social behaviors of female mice).  

(1a) All the activity-dependent labeling experiments with TRAP mice, including the subsequent neural activity manipulation experiments (Figures 2, 3, 4, 5E-F), were conducted by labeling neurons only in socially isolated animals, not group-housed animals. The authors labeled neurons after 30-minute social interactions, raising the possibility that the labeled neurons simply represent a "social interaction/behavior population" (mediating mounting and USVs in females and males) rather than a set of neurons specific to social isolation behaviors of mice)… The data thus far suggests these neurons may predominantly reflect a "POA social behavior" population rather than a set of cells distinctly responsive to isolated housing.

We agree with the reviewer that the POA neurons we are studying regulate the production of social behaviors in females and males, rather than representing a set of cells distinctly responsive to single housing. To more clearly reflect our thinking, we have changed the name of the neurons from “POA-iso neurons” to “POA-social neurons”. Thank you for this helpful criticism.

Our Fos data are consistent with the idea that the POA may regulate social behaviors in group-housed females (not just single-housed females). Namely, we found that counts of Fospositive POA neurons are significantly related to rates of social investigation (p = 0.01) and tend to be related to USV rates (p = 0.05) in group-housed females that engaged in same-sex interactions (Fig. S1C). We now include two new sets of experiments aimed at further testing the idea this idea. 

First, we include 2 control groups in which TRAPing sessions were performed in grouphoused females following same-sex interactions. We find that chemogenetic silencing of these group-housed-TRAPed POA neurons fails to reduce social behaviors in females that are subsequently single-housed and given a same-sex social interaction (Fig. 5A-D; GH-TRAPed POA hM4Di females), and that optogenetic activation of group-housed-TRAPed POA neurons fails to promote female social behavior (Fig. 5E-H; GH-TRAPed POA ChR2 females). At face value, these findings do not support the idea that the POA contains neurons that regulate social behaviors in group-housed females.

However, one important caveat is that group-housed females engage in low rates of social behaviors (low investigation time, no mounting, and few USVs), and thus TRAP-based labeling may not work efficaciously in these mice. There may be POA neurons that regulate social behaviors in group-housed females but that do not upregulate Fos following production of relatively low rates of social behaviors. To test this idea, we also include females in which POA neurons are chemogenetically silenced using a viral strategy that does not depend on activitydependent labeling. In this new experiment, we report that silencing of POA neurons significantly reduces USV production in group-housed females (Fig. 5J-L) and significantly reduces social investigation, mounting, and USV production when these same females are retested following single-housing (Fig. 5M-O).

(2) Please add strain background information of subject animals in the methods.

This information has been added to the Animals section within the Methods (lines 788802).

Responses to Reviewer 3 Recommendations:

(1a) (T)he conflicting effects on behavior are hard to interpret without additional experiments….Similarly, optogenetic activation of POA neurons was sufficient to generate USV production as reported in earlier studies but mounting or social investigation remained unaffected. 

We have added new analyses to consider the possibility that optogenetic activation of female POA-social neurons promotes social investigation. In the original manuscript, we analyzed the duration of social investigation bouts in POA-social-ChR2 females according to whether they overlapped with laser stimulation or whether they did not overlap. We realized that we made an error in this first analysis and inadvertently included social investigation bouts that occurred during the first 5 minutes of the social sessions, prior to any laser stimulation. Because these earlier bouts tend to be longer duration than later bouts, this mistake washed out the effect of laser stimulation on social bout duration. After correcting that error, we now report that optogenetic activation of female POA-social neurons lengthens social investigation bout duration (Fig. 4G). Inspired by this interesting finding, we also included analyses of the probability of social investigation following laser stimulation (Fig. 4E-F; excluding laser stimulations that were preceded by social investigation in the pre-laser baseline period). These analyses support the conclusion that optogenetic activation of POA-social neurons promotes both USV production and social investigation in group-housed females.

(1b) Do these discrepancies (between hM4Di and caspase) arise due to the efficiency differences between DREADD-mediated silencing vs. Casp3 ablation? Or does the chemogenetic result reflect off-manifold effects on downstream circuitry whereas a more permanent ablation strategy allows other brain regions to compensate due to redundancy? It is important to resolve whether these arise due to technical reasons or whether these reflect the underlying (perhaps messy) logic of neural circuitry.  

The possibility that the difference in effects on behavior between chemogenetic silencing and caspase ablation at face value seems inconsistent with the findings of previous experiments, in which ablation of large numbers of POA neurons failed to reduce USV production in male mice (POA lesions in Bean et al., 1981; ablation of VGAT+ POA neurons by Gao et al., 2018). These findings stand in contrast to those using chemogenetic silencing of large numbers of POA neurons, which report reduced USV production in male mice (VGAT+/Esr1+ in Karigo et al., 2021; Esr1+ in Chen et al., 2021).

However, it is the case that the majority of the females that we used in our TRAP2-based ablation experiments were heterozygous for TRAP2 (N = 11 of 15 POA-social-caspase subjects were TRAP2;Ai14 females), whereas all females used in our chemogenetic silencing experiments were homozygous for TRAP2. To test whether a more effective ablation of POAsocial neurons might drive decreases in social investigation and USV production, we set up additional TRAP2 homozygous POA-social-caspase females and directly compare the effects of ablation between the two genotypes (Fig. S3; N = 11 hets in total and N = 9 homozygotes in total). These experiments revealed that effects on mounting were more pronounced following POA-social ablation in TRAP2 homozygotes vs. heterozygotes, but that neither group exhibited decreased social investigation or USV production following 4-OHT treatment.

To ask whether caspase-mediated ablation in TRAP2 homozygotes was effective in eliminating neural activity associated with social behaviors in females, we performed Fos immunostaining in a subset of the POA-social-caspase TRAP2 homozygotes following a samesex interaction. We found that POA Fos expression was robustly reduced in these females relative to control group-housed and control single-housed females that also engaged in samesex interactions, down to levels seen in group-housed and single-housed females that did not engage in a social interaction (comparison shown in Fig. S3D; control female data same as in Fig. 1). Moreover, the remaining POA Fos in these TRAP2 homozygotes was no longer positively correlated to social investigation or USV production (Fig. S3E-F). Together, these findings lead us to favor the interpretation suggested by the reviewer below, that permanent ablation of POA-social neurons leads to compensation from other brain regions due to redundancy.

Given the negative results above, we favor this possibility and indicate so in our Discussion. In addition, our finding that optogenetic activation of POA-social neurons promotes both USV production and social investigation supports the idea that POA-social neurons directly regulate these behaviors. We agree with the reviewer that additional work is needed to understand the complex sex- and context-dependent role played by the POA in the regulation of mouse social behaviors.

(2) L 49: Please define Mesolimbic circuitry the first time it is mentioned.

We have added a definition (lines 52-53).

(3) L 210: In Figure 2C, the mounting duration baseline (saline) distribution seems lower than the same experimental baseline in Figures 1C and 3C. Does this reflect natural variability in the behavioral assay and might this be mitigated by additional sampling of animals?

Yes, there is substantial variability in the display of mounting behavior by single-housed females, including in the proportion of trials with mounting as well as in the total duration of mounting. In the revised manuscript, we have simplified our analysis of mounting in our TRAPbased experiments to quantify the proportion of trials with mounting, rather than considering the total time spent mounting. After adding N = 5 additional females to the POA-social-hM4Di dataset, we now report a statistically significant decrease in the proportion of trials with mounting following chemogenetic silencing of POA-social neurons (Fig. 2C; McNemar’s test for paired proportions). 

(4) L 310: The authors claim that "These findings suggest that a subset of POAiso neurons overlap with GABAergic, PAG-projecting POA neurons that have been demonstrated in previous work to promote USVs via disinhibition of excitatory PAG neurons important to USV production (Chen et al., 2021; Michael et al., 2020)." I think the data reported suggests the opposite since only 18.3% of all POA->PAG neurons are cFos+. Perhaps better rephrased as "A subset (18.3%) of POA->PAG neurons are labelled by cFos and that is sufficient to drive the production of USVs". Is it surprising?

We modified the phrasing (lines 468-469), but a bit differently than suggested above, because although we suspect that optogenetic activation of the PAG-projecting neurons within the larger population of POA-social neurons is responsible for eliciting USV production, we did not technically demonstrate this to be the case in the current dataset. 

We do find it surprising that so few (only ~20%) of PAG-projecting POA neurons upregulate Fos following female-female interactions marked by high rates of USV production. Even though optogenetic activation of PAG-projecting POA neurons elicits USV production, our finding suggests that the majority of PAG-projecting POA neurons may not play a role in regulating vocalization. In future work, it may be useful to apply an intersectional approach to further understand how the POA regulates USV production (for example, measure or manipulate activity selectively in projection-defined subsets of POA-social neurons).

(5) Given the considerable prior evidence of POA->PAG circuit in promoting USVs, it is hard to understand why chemogenetic inactivation of POA neurons in males affects mounting but not USV production (Figures 5F-H). Any potential explanation for this discrepancy?

We have two ideas about this surprising result. First, we examined the TRAPing session social behaviors of female and male POA-social-hM4Di mice. We found that male POA-socialhM4Di mice spent more time than female subjects mounting during the TRAPing sessions, and conversely, males spent less time investigating visitors and tended to produce fewer USVs than female subjects (Fig. S5). Given that our labeling method is activity-dependent, one possibility is that this bias in behavior is reflected in a bias toward labeling of POA neurons related to mounting.  

Second, each mouse in the TRAP2-based hM4Di datasets received an IP injection of the same amount of 4-OHT (150 nL of 10 mg/mL 4-OHT in filtered corn oil) not adjusted for weight of the mouse. This information was not reported accurately in the Methods, and we have adjusted that section accordingly (line 920). As a result, because male mice typically weigh more than females and would have received a lower effective dosage of 4-OHT, another possibility is that TRAPing in males was less efficient than in females and accounts for the less complete effects on social behaviors. We have added language to the Results to discuss these possibilities (lines 540-560).

(6) L 472: Typo. "we found that short-term isolation exerts more robust on the effects of male behavior during subsequent interactions with females than during interactions with males."

Thank you for catching this mistake.

eLife Assessment

The authors provide a thorough investigation of the interaction of megakaryocytes (MK) with their associated extracellular matrix (ECM) during maturation; they provide evidence that the existence of a dense cage-like pericellular structure containing laminin γ1 and α4 and collagen IV is key to fixing the perisinusoidal localization of MK and preventing their premature intravasation. Adhesion of MK to this ECM cage is dependent on integrin beta1 and beta3 expressed by MK. This strong and solid conclusion is based on the use of state-of-the art techniques such as the use of primary murine bone marrow MK cultures, mice lacking ECM receptors, namely integrin beta1 and beta3 null mice, as well as high-resolution 2D and 3D imaging. The study provides valuable insight into the role of cell-matrix interactions in MK maturation and provides an interesting model with practical implications for the fields of hemostasis and thrombosis.

Reviewer #1 (Public review):

The authors report on a thorough investigation of the interaction of megakaryocytes (MK) with their associated ECM during maturation. They report convincing evidence to support the existence of a dense cage-like pericellular structure containing laminin γ1 and α4 and collagen IV, which interacts with integrins β1 and β3 on MK and serves to fix the perisinusoidal localization of MK and prevent their premature intravasation. As with everything in nature, the authors support a Goldilocks range of MK-ECM interactions - inability to digest the ECM via inhibition of MMPs leads to insufficient MK maturation and development of smaller MK. This important work sheds light on the role of cell-matrix interactions in MK maturation, and suggests that higher-dimensional analyses are necessary to capture the full scope of cellular biology in the context of their microenvironment.

There are several outstanding questions that this work does not address.

Major:

The authors postulate a synergistic role for Itgb1 and Itgb3 in the intravasation phenotype, because the single KOs did not replicate the phenotype of the DKO. However, this is not a correct interpretation in the opinion of this reviewer. The roles appear rather to be redundant. Synergistic roles would rather demonstrate a modest effect in the single KO with potentiation in the DKO.

Furthermore, the experiment does not explain how these integrins influence the interaction of the MK with their microenvironment. It is not surprising that attachment will be impacted by the presence or absence of integrins. However, it is unclear how activation of integrins allows the MK to become "architects for their ECM microenvironment" as the authors posit. A transcriptomic analysis of control and DKO MKs may help elucidate these effects.

Integrin DKO have a 50% reduction in platelets counts as reported previously, however laminin α4 deficiency only leads to 20% reduction in counts. This suggests a more nuanced and subtle role of the ECM in platelet growth. To this end, functional assays of the platelets in the KO and wildtype mice may provide more information.

There is insufficient information in the Methods Section to understand the BM isolation approach. Did the authors flush the bone marrow and then image residual bone, or the extruded bone marrow itself as described in PMID: 29104956?

The references in the Methods section were very frustrating. The authors reference Eckly et al 2020 (PMID: 32702204) which provides no more detail but references a previous publication (PMID: 24152908), which also offers no information and references a further paper (PMID: 22008103), which, as far as this reviewer can tell, did not describe the methodology of in situ bone marrow imaging.

Therefore, this reviewer cannot tell how the preparation was performed and, importantly, how can we be sure that the microarchitecture of the tissue did not get distorted in the process?

Reviewer #2 (Public review):

Summary:

This study makes a significant contribution to understanding the microenvironment of megakaryocytes (MKs) in the bone marrow, identifying an extracellular matrix (ECM) cage structure that influences MK localization and maturation. The authors provide compelling evidence for the presence of this ECM cage and its role in MK homeostasis, employing an array of sophisticated imaging techniques and molecular analyses. While the work is innovative and impactful, there are several points that require clarification or further data to fully support the conclusions.

Major Strengths:

Novelty: The identification of an ECM cage as a regulator of MK localization and maturation in the bone marrow is a novel and exciting finding.

Imaging Techniques: The use of advanced microscopy to visualize the 3D structure of the ECM cage and its role in MK homeostasis provides a strong visual foundation for the study's claims.

Comprehensive Analysis: The integration of in vivo and ex vivo approaches enhances the significance of the findings, offering valuable insights into the molecular mechanisms involved in ECM cage formation.

Areas for Improvement and Clarifications:

(1) ECM cage imaging:<br /> a) The value or additional information provided by the staining on nano-sections (A) is not clear, especially considering that the thick vibratome sections already display the entirety of the laminin γ1 cage structure effectively. Further clarification on the unique insights gained from each approach would help justify its inclusion.<br /> b) The sMK shown in Supplementary Figure 1C appears to be linked to two sinusoids, releasing proplatelets to the more distant vessels. Is this observation representative, and if so, can further discussion be provided?<br /> c) Freshly isolated BM-derived MKs are reported to maintain their laminin γ1 cage. Are the proportions of MKs with/without cages consistent with those observed in microscopy?

(2) ECM cage formation:<br /> a) The statement "the full assembly of the 3D ECM cage required megakaryocyte interaction with the sinusoidal basement membrane" on page 7 is too strong given the data presented at this stage of the study. Supplemental Figure 1C shows that approximately 10% of pMKs form cages without direct vessel contact, indicating that other factors may also play a role in cage formation.<br /> b) The data supporting the statement that "pMK represent a small fraction of the total MK population" (cell number or density) could be shown to help contextualize the 10% of them with a cage.<br /> c) How "the full assembly of the 3D ECM cage" is defined at this stage of the study should be clarified, specifically regarding the ECM components and structural features that characterize its completion.

(3) Data on MK Circulation and Cage Integrity: Does the cage require full component integrity to prevent MK release in circulation? Are circulating MKs found in Lama4-/- mice? Is the intravasation affected in these mice? Are the ~50% sinusoid associated MK functional?

(4) Methodology:<br /> a) Details on fixation time are not provided, which is critical as it can impact antibody binding and staining. Including this information would improve reproducibility and feasibility for other researchers.<br /> b) The description of 'random length measuring' is unclear, and the rationale behind choosing random quantification should be explained. Additionally, in the shown image, it appears that only the branching ends were measured, which makes it difficult to discern the randomness in the measurements.

(5) Figures:<br /> a) Overall, the figures and their corresponding legends would benefit from greater clarity if some panels were split, such as separating images from graph quantifications.

Reviewer #3 (Public review):

In this manuscript, Masson, Scandola, et al investigate how interactions between megakaryocytes and the extracellular matrix contribute to the regulation of thrombopoiesis using primary murine bone marrow MK cultures, integrin B1/B3 knock-out mice, and high-resolution 2D and 3D imaging. They find that laminin and collagen iv create a 3D "cage" of ECM surrounding MKs and anchor them at the sinusoidal basement membrane, which contributes to MK maturation and proplatelet intravasation into circulation. Deletion of laminin a4 disrupts the localization of MKs and the endothelial basement membrane, reducing the number of MKs associated with the sinusoid while having no effect on MK-associated collagen IV. Deletion of B1/B3 integrin reduces the quantity, localization, and structural organization of multiple ECM components surrounding MKs, and reduces MK adhesion when subject to conditions of sinusoidal flow.

Further, using intravital microscopy of calvarial bone marrow and the pulmonary vasculature, they provide data suggesting that the stabilization of ECM around MKs (either in the BM or lung) prevents MKs from entering circulation as intact cells. Interestingly, deletion of B1 integrin reduces MK coverage in laminin y1, but deletion of both B1 and B3 independently results in increased MK intravasation into the sinusoidal space. Comparison of integrin KO MKs with GPVI KO MKs suggests that ECM cage formation, vessel adhesion, and intravasation are likely dependent on integrin activation/signaling rather than GPVI signals.

Further, they provide data that the balance of ECM synthesis and degradation is essential for MK maturation and also provide data showing that inhibition of ECM turnover (in vivo inhibition of MMPs) results in increased ECM cage components that correspond with reduced MK maturation, and reduced demarcation membrane development.

The conclusions of the paper are supported by the data, but there are some areas that would benefit from clarification or expansion.

(1) The data linking ECM cage formation to MK maturation raises several interesting questions. As the authors mention, MKs have been suggested to mature rapidly at the sinusoids, and both integrin KO and laminin KO MKs appear mislocalized away from the sinusoids. Additionally, average MK distances from the sinusoid may also help separate whether the maturation defects could be in part due to impaired migration towards CXCL12 at the sinusoid. Presumably, MKs could appear mislocalized away from the sinusoid given the data presented suggesting they leaving the BM and entering circulation. Additional data or commentary on intrinsic (ex-vivo) MK maturation phenotypes may help strengthen the author's conclusions and shed light on whether an essential function of the ECM cage is integrin activation at the sinusoid.

(2) The data demonstrating intact MKs inter circulation is intriguing - can the authors comment or provide evidence as to whether MKs are detectable in blood? A quantitative metric may strengthen these observations.

(3) Supplementary Figure 6 - shows no effect on in vitro MK maturation and proplt, or MK area - But Figures 6B/6C demonstrate an increase in total MK number in MMP-inhibitor treated mice compared to control. Some additional clarification in the text may substantiate the author's conclusions as to either the source of the MMPs or the in vitro environment not fully reflecting the complex and dynamic niche of the BM ECM in vivo.

(4) Similarly, one function of the ECM discussed relates to MK maturation but in the B1/3 integrin KO mice, the presence of the ECM cage is reduced but there appears to be no significant impact upon maturation (Supplementary Figure 4). By contrast, MMP inhibition in vivo (but not in vitro) reduces MK maturation. These data could be better clarified in the text, or by the addition of experiments addressing whether the composition and quantity of ECM cage components directly inhibit maturation versus whether effects of MMP-inhibitors perhaps lead to over-activation of the integrins (as with the B4galt KO in the discussion) are responsible for the differences in maturation.

Author response:

Reviewer #1 (Public review):

Point 1. The authors postulate a synergistic role for Itgb1 and Itgb3 in the intravasation phenotype, because the single KOs did not replicate the phenotype of the DKO. However, this is not a correct interpretation in the opinion of this reviewer. The roles appear rather to be redundant. Synergistic roles would rather demonstrate a modest effect in the single KO with potentiation in the DKO.

We agree that the interaction between Itgb1 and Itgb3 appears redundant and we will correct this point in the revised manuscript.

Point 2. The experiment does not explain how these integrins influence the interaction of the MK with their microenvironment. It is not surprising that attachment will be impacted by the presence or absence of integrins. However, it is unclear how activation of integrins allows the MK to become "architects for their ECM microenvironment" as the authors posit. A transcriptomic analysis of control and DKO MKs may help elucidate these effects.

We do not currently understand how α5β1 or αvβ3 integrins activation would contribute to ECM remodeling by megakaryocytes. Integrins are well known key regulators of ECM remodelling (https://doi.org/10.1016/j.ceb.2006.08.009). They can transmit traction force that provoques ECM remodelling (https://doi.org/10.1016/j.bpj.2008.10.009). We will discuss our previous study on the observed reduction in RhoA activation in double knockout (DKO) mice (Guinard et al., 2023,  PMID: 37171626), which likely impact the organization of the ECM microenvironment. Alternatively, integrin signalling contribute to gene expression regulation involved in ECM remodelling (ECM proteins, proteases….). We do agree with the reviewer that the transcriptomic analysis could provide strong evidence; however, it is challenging to perform this analysis in vivo. Isolation of native megakaryocytes (MKs) from DKO mice is challenging due to their reduced numbers, requiring too many mice for sufficient RNA and risk of cell contamination. An alternative approach will be to analyze platelets, which are more abundant and easier to isolate, while still mimicking the characteristics of bone marrow MKs. We will use PCR array technology for selected ECM panels and adhesion molecules (from all players currently known to contribute to ECM remodelling), providing a practical way to address the reviewer's suggestions and provide valuable insights.

Point 3. Integrin DKO have a 50% reduction in platelets counts as reported previously, however laminin α4 deficiency only leads to 20% reduction in counts. This suggests a more nuanced and subtle role of the ECM in platelet growth. To this end, functional assays of the platelets in the KO and wildtype mice may provide more information.

The difference in platelet counts between integrin DKO and laminin α4 KO mice is not fully understood. Although our study specifically focuses on MK-ECM interactions in the bone marrow, we recognize the importance of providing additional information on platelet functionality. To address this, we will use flow cytometry to examine the levels of P-selectin surface expression and fibrinogen binding under basal conditions and after stimulation with collagen-related peptide and TRAP.

Point 4. There is insufficient information in the Methods Section to understand the BM isolation approach. Did the authors flush the bone marrow and then image residual bone, or the extruded bone marrow itself as described in PMID: 29104956?

Additional information on the methodology will be provided to clarify the BM isolation.

Point 5. The references in the Methods section were very frustrating. The authors reference Eckly et al 2020 (PMID: 32702204) which provides no more detail but references a previous publication (PMID: 24152908), which also offers no information and references a further paper (PMID: 22008103), which, as far as this reviewer can tell, did not describe the methodology of in situ bone marrow imaging.

To address this confusion, we will add the reference "In Situ Exploration of the Major Steps of Megakaryopoiesis Using Transmission Electron Microscopy" by C. Scandola et al. (PMID: 34570102), which provides a standardized protocol for bone marrow isolation.

Therefore, this reviewer cannot tell how the preparation was performed and, importantly, how can we be sure that the microarchitecture of the tissue did not get distorted in the process?

Thank you for pointing this out. While we cannot completely rule out the possibility of distortion, we will clarify the precautions taken to minimize it. We utilized a double fixation process immediately after extruding the bone marrow, followed by embedding it in agarose to preserve its integrity as much as possible. We will address this point in greater detail in Methods section of the revised version.

Reviewer #2 (Public review):

Point 1. ECM cage imaging

a) The value or additional information provided by the staining on nano-sections (A) is not clear, especially considering that the thick vibratome sections already display the entirety of the laminin γ1 cage structure effectively. Further clarification on the unique insights gained from each approach would help justify its inclusion.

Ultrathin cryosection allow high-resolution imaging (10x fold increased in Z), facilitating the analysis of signal superposition. This study explores the interactions between MKs and their immediate ECM microenvironment, located at a distance of less than one micrometer, making nano-sections optimal for precise analysis of ECM distribution both within and surrounding MKs. This high-resolution approach has revealed the presence of collagen IV, laminin, fibronectin, and fibrinogen near MKs, More importantly, ultrathin cryosection allow us to clearly show with high resolution the presence of activated integrin in contact with laminin an coll IV fibers (see Fig. 3)

We employed large-volume whole-mount imaging to clarify the overall three-dimensional architecture of the ECM interface, allowing us to identify the cages. Our findings emphasize the role of specific ECM components in facilitating proplatelet passage through the sinusoid barrier, an essential step for platelet production. Further details will be addressed in the revised manuscript.

b) The sMK shown in Supplementary Figure 1C appears to be linked to two sinusoids, releasing proplatelets to the more distant vessels. Is this observation representative, and if so, can further discussion be provided?

This observation is not representative; MKs can also be associated with just one sinusoid.

c) Freshly isolated BM-derived MKs are reported to maintain their laminin γ1 cage. Are the proportions of MKs with/without cages consistent with those observed in microscopy?   

In the revised manuscript, we will include the quantification of the proportion of BM-derived MKs with/without cages.

Point 2.  ECM cage formation

a) The statement "the full assembly of the 3D ECM cage required megakaryocyte interaction with the sinusoidal basement membrane" on page 7 is too strong given the data presented at this stage of the study. Supplemental Figure 1C shows that approximately 10% of pMKs form cages without direct vessel contact, indicating that other factors may also play a role in cage formation.

The reviewer is correct. We will modify the text to reflect a more cautious interpretation of our results.

b) The data supporting the statement that "pMK represent a small fraction of the total MK population" (cell number or density) could be shown to help contextualize the 10% of them with a cage.

New bar graphs will be provided to represent the density of MK in the parenchyma against the total MK in the bone marrow.

c) How "the full assembly of the 3D ECM cage" is defined at this stage of the study should be clarified, specifically regarding the ECM components and structural features that characterize its completion.

We recognize that the term ' full assembly' of the 3D ECM cage can be misleading, as it might suggest different stages of cage formation, such as a completed cage, one that is in the process of formation, or an incomplete cage. Since we have not yet studied this concept, we will eliminate the term "full assembly" from the manuscript to avoid any confusion. Instead, we will simply mention the presence of a cage.

Point 3. Data on MK Circulation and Cage Integrity: Does the cage require full component integrity to prevent MK release in circulation? Are circulating MKs found in Lama4-/- mice? Is the intravasation affected in these mice? Are the ~50% sinusoid associated MK functional?  

These are very valid points. We will answer all these questions by performing a detailed analysis of MK localization, vessel association and intravascular MK detection using IF and high-resolution EM imaging of Lamα4^-/- mice. Additionally, we will analyze data from Lamα4-/- bone marrow explants to assess the capacity of MKs to extend proplatelets.

Point 4. Methodology

a) Details on fixation time are not provided, which is critical as it can impact antibody binding and staining. Including this information would improve reproducibility and feasibility for other researchers.

We will added this information in the methods section.

b) The description of 'random length measuring' is unclear, and the rationale behind choosing random quantification should be explained. Additionally, in the shown image, it appears that only the branching ends were measured, which makes it difficult to discern the randomness in the measurements.

The random length measurement method uses random sampling to provide unbiased data on laminin/collagen fibers in a 3D cage. Contrary to what the initial image might have suggested, measurements go beyond just the branching ends; they include intervals between various branching points throughout the cage.

To clarify this process, we will outline these steps: 1) acquire 3D images, 2) project onto 2D planar sections, 3) select random intersection points for measurement, 4) measure intervals using ImageJ software, and 5) repeat the process for a representative dataset. This will better illustrate the randomness of our measurements.

Point 5.  Figures

a) Overall, the figures and their corresponding legends would benefit from greater clarity if some panels were split, such as separating images from graph quantifications.

Following the reviewer’s suggestion, we will fully update all the Figures and separate images from graph quantifications.

Reviewer #3 (Public review):

Point 1. The data linking ECM cage formation to MK maturation raises several interesting questions. As the authors mention, MKs have been suggested to mature rapidly at the sinusoids, and both integrin KO and laminin KO MKs appear mislocalized away from the sinusoids. Additionally, average MK distances from the sinusoid may also help separate whether the maturation defects could be in part due to impaired migration towards CXCL12 at the sinusoid. Presumably, MKs could appear mislocalized away from the sinusoid given the data presented suggesting they leaving the BM and entering circulation. Additional data or commentary on intrinsic (ex-vivo) MK maturation phenotypes may help strengthen the author's conclusions and shed light on whether an essential function of the ECM cage is integrin activation at the sinusoid.

The hypothesis of MK migration towards CXCL12 is interesting, although it has recently been challenged by Stegner et al. (2017), who found that MKs are primarily sessile. However, we cannot exclude this possibility. To address the reviewer's concerns, we will quantify the distance of MKs from the sinusoids. This could help to determine whether the maturation defects are due to impaired migration towards CXCL12 at the sinusoids or other factors, such as the ECM cage.

We would appreciate some clarification regarding the second point raised by the reviewer. Is the question  specifically addressing whether the ECM cage has an effect on the activation of integrins in the sinusoids? If so, we will use immunofluorescence (IF) to investigate the relationship between the presence of an ECM cage and the activation of integrins on the surface of endothelial cells within the sinusoids. Thank you for your guidance on this matter.

Point 2. The data demonstrating intact MKs inter circulation is intriguing - can the authors comment or provide evidence as to whether MKs are detectable in blood? A quantitative metric may strengthen these observations.

We will conduct flow cytometry experiments and prepare blood smears to determine whether intact MKs are detectable in blood.

Point 3. Supplementary Figure 6 - shows no effect on in vitro MK maturation and proplt, or MK area - But Figures 6B/6C demonstrate an increase in total MK number in MMP-inhibitor treated mice compared to control. Some additional clarification in the text may substantiate the author's conclusions as to either the source of the MMPs or the in vitro environment not fully reflecting the complex and dynamic niche of the BM ECM in vivo.

This is a valid point. We will revise the text to include further clarification.

Point 4.  Similarly, one function of the ECM discussed relates to MK maturation but in the B1/3 integrin KO mice, the presence of the ECM cage is reduced but there appears to be no significant impact upon maturation (Supplementary Figure 4). By contrast, MMP inhibition in vivo (but not in vitro) reduces MK maturation. These data could be better clarified in the text, or by the addition of experiments addressing whether the composition and quantity of ECM cage components directly inhibit maturation versus whether effects of MMP-inhibitors perhaps lead to over-activation of the integrins (as with the B4galt KO in the discussion) are responsible for the differences in maturation.

These are very good questions, but they are difficult to assess in situ. To approach this, we will perform in vitro experiments :

(1) We will vary collagenIV and laminin411 concentrations in the culture conditions to determine how this affects MK maturation ; and

(2) We will assess the integrin activation states on cultured MKs treated with MMP inhibitors to determine if MMP inhibitors could influence MK maturation through over-activation of integrins.

eLife Assessment

In this valuable study, the authors used rats to determine the receptor for a food-related perception (kokumi) that has been characterized in humans. They employ a combination of behavioral, electrophysiological, and immunohistochemical results to provide solid support for their conclusion that ornithine-mediated kokumi effects are mediated by the GPRC6A receptor. They complemented the rat data with some human psychophysical data. The results are intriguing, but there are some deficiencies.

Reviewer #1 (Public review):

Summary:

This paper contains what could be described as a "classic" approach towards evaluating a novel taste stimuli in an animal model, including standard behavioral tests (some with nerve transections), taste nerve physiology, and immunocytochemistry of taste cells of the tongue. The stimulus being tested is ornithine, from a class of stimuli called "kokumi" (in terms of human taste); these kokumi stimuli appear to enhance other canonical tastes, increasing what are essentially hedonic attributes of other stimuli. The mechanism for ornithine detection is thought to be GPRC6A receptors expressed in taste cells. The authors showed evidence for this in an earlier paper with mice; this paper evaluates ornithine taste in a rat model, and comes to a similar conclusion, albeit with some small differences between the two rodent species.

Strengths:

The data show effects of ornithine on taste/intake in laboratory rats: In two-bottle and briefer intake tests, adding ornithine results in higher intake of most, but all not all stimuli tested. Bilateral chorda tympani (CT) nerve cuts or the addition of GPRC6A antagonists decreased or eliminated these effects. Ornithine also evoked responses by itself in the CT nerve, but mainly at higher concentrations; at lower concentrations it potentiated the response to monosodium glutamate. Finally, immunocytochemistry of taste cell expression indicated that GPRC6A was expressed predominantly in the anterior tongue, and co-localized (to a small extent) with only IP3R3, indicative of expression in a subset of type II taste receptor cells.

Weaknesses:

As the authors are aware, it is difficult to assess a complex human taste with complex attributes, such as kokumi, in an animal model. In these experiments they attempt to uncover mechanistic insights about how ornithine potentiates other stimuli by using a variety of established experimental approaches in rats. They partially succeed by finding evidence that GPRC6A may mediate effects of ornithine when it is used at lower concentrations. In the revision they have scaled back their interpretations accordingly. A supplementary experiment measuring certain aspects of the effects of ornithine added to Miso soup in human subjects is included for the express purpose of establishing that the kokumi sensation of a complex solution is enhanced by ornithine; however, they do not use any such complex solutions in the rat studies. Moreover, the sample size of the human experiment is (still) small - it really doesn't belong in the same manuscript with the rat studies.

Reviewer #2 (Public review):

Summary:

The authors used rats to determine the receptor for a food-related perception (kokumi) that has been characterized in humans. They employ a combination of behavioral, electrophysiological, and immunohistochemical results to support their conclusion that ornithine-mediated kokumi effects are mediated by the GPRC6A receptor. They complemented the rat data with some human psychophysical data. I find the results intriguing, but believe that the authors overinterpret their data.

Strengths:

The authors provide compelling evidence that ornithine enhances the palatability of several chemical stimuli (i.e., IMP, MSG, MPG, Intralipos, sucrose, NaCl, quinine). Ornithine also increases CT nerve responses to MSG. Additionally, the authors provide evidence that the effects of ornithine are mediated by GPRC6A, a G-protein-coupled receptor family C group 6 subtype A, and that this receptor is expressed primarily in fungiform taste buds. Taken together, these results indicate that ornithine enhances the palatability of multiple taste stimuli in rats and that the enhancement is mediated, at least in part, within fungiform taste buds. This is an important finding that could stand on its own. The question of whether ornithine produces these effects by eliciting kokumi-like perceptions (see below) should be presented as speculation in the Discussion section.

Weaknesses:

I am still unconvinced that the measurements in rats reflect the "kokumi" taste percept described in humans. The authors conducted long-term preference tests, 10-min avidity tests and whole chorda tympani (CT) nerve recordings. None of these procedures specifically model features of "kokumi" perception in humans, which (according to the authors) include increasing "intensity of whole complex tastes (rich flavor with complex tastes), mouthfulness (spread of taste and flavor throughout the oral cavity), and persistence of taste (lingering flavor)." While it may be possible to develop behavioral assays in rats (or mice) that effectively model kokumi taste perception in humans, the authors have not made any effort to do so. As a result, I do not think that the rat data provide support for the main conclusion of the study--that "ornithine is a kokumi substance and GPRC6A is a novel kokumi receptor."

Why are the authors hypothesizing that the primary impacts of ornithine are on the peripheral taste system? While the CT recordings provide support for peripheral taste enhancement, they do not rule out the possibility of additional central enhancement. Indeed, based on the definition of human kokumi described above, it is likely that the effects of kokumi stimuli in humans are mediated at least in part by the central flavor system.

The authors include (in the supplemental data section) a pilot study that examined the impact of ornithine on variety of subjective measures of flavor perception in humans. The presence of this pilot study within the larger rat study does not really mice sense. While I agree with the authors that there is value in conducting parallel tests in both humans and rodents, I think that this can only be done effectively when the measurements in both species are the same. For this reason, I recommend that the human data be published in a separate article.

The authors indicated on several occasions (e.g., see Abstract) that ornithine produced "synergistic" effects on the CT nerve response to chemical stimuli. "Synergy" is used to describe a situation where two stimuli produce an effect that is greater than the sum of the response to each stimulus alone (i.e., 2 + 2 = 5). As far as I can tell, the CT recordings in Fig. 3 do not reflect a synergism.

Reviewer #3 (Public review):

Summary:

In this study the authors set out to investigate whether GPRC6A mediates kokumi taste initiated by the amino acid L-ornithine. They used Wistar rats, a standard laboratory strain, as the primary model and also performed an informative taste test in humans, in which miso soup was supplemented with various concentrations of L-ornithine. The findings are valuable and overall the evidence is solid. L-Ornithine should be considered to be a useful test substance in future studies of kokumi taste and the class C G protein coupled receptor known as GPRC6A (C6A) along with its homolog, the calcium-sensing receptor (CaSR) should be considered candidate mediators of kokumi taste. The researchers confirmed in rats their previous work on Ornithine and C6A in mice (Mizuta et al Nutrients 2021).

Strengths:

The overall experimental design is solid based on two bottle preference tests in rats. After determining the optimal concentration for L-Ornithine (1 mM) in the presence of MSG, it was added to various tastants including: inosine 5'-monophosphate; monosodium glutamate (MSG); mono-potassium glutamate (MPG); intralipos (a soybean oil emulsion); sucrose; sodium chloride (NaCl; salt); citric acid (sour) and quinine hydrochloride (bitter). Robust effects of ornithine were observed in the cases of IMP, MSG, MPG and sucrose; and little or no effects were observed in the cases of sodium chloride, citric acid; quinine HCl. The researchers then focused on the preference for Ornithine-containing MSG solutions. Inclusion of the C6A inhibitors Calindol (0.3 mM but not 0.06 mM) or the gallate derivative EGCG (0.1 mM but not 0.03 mM) eliminated the preference for solutions that contained Ornithine in addition to MSG. The researchers next performed transections of the chord tympani nerves (with sham operation controls) in anesthetized rats to identify a role of the chorda tympani branches of the facial nerves (cranial nerve VII) in the preference for Ornithine-containing MSG solutions. This finding implicates the anterior half-two thirds of the tongue in ornithine-induced kokumi taste. They then used electrical recordings from intact chorda tympani nerves in anesthetized rats to demonstrate that ornithine enhanced MSG-induced responses following the application of tastants to the anterior surface of the tongue. They went on to show that this enhanced response was insensitive to amiloride, selected to inhibit 'salt tastant' responses mediated by the epithelial Na+ channel, but eliminated by Calindol. Finally they performed immunohistochemistry on sections of rat tongue demonstrating C6A positive spindle-shaped cells in fungiform papillae that partially overlapped in its distribution with the IP3 type-3 receptor, used as a marker of Type-II cells, but not with (i) gustducin, the G protein partner of Tas1 receptors (T1Rs), used as a marker of a subset of type-II cells; or (ii) 5-HT (serotonin) and Synaptosome-associated protein 25 kDa (SNAP-25) used as markers of Type-III cells.

At least two other receptors in addition to C6A might mediate taste responses to ornithine: (i) the CaSR, which binds and responds to multiple L-amino acids (Conigrave et al, PNAS 2000), and which has been previously reported to mediate kokumi taste (Ohsu et al., JBC 2010) as well as responses to Ornithine (Shin et al., Cell Signaling 2020); and (ii) T1R1/T1R3 heterodimers which also respond to L-amino acids and exhibit enhanced responses to IMP (Nelson et al., Nature 2001). These alternatives are appropriately discussed and, taken together, the experimental results favor the authors' interpretation that C6A mediates the Ornithine responses. The authors provide preliminary data in Suppl. 3 for the possibility of co-expression of C6A with the CaSR.

Weaknesses:

The authors point out that animal models pose some difficulties of interpretation in studies of taste and raise the possibility in the Discussion that umami substances may enhance the taste response to ornithine (Line 271, Page 9).

One issue that is not addressed, and could be usefully addressed in the Discussion, relates to the potential effects of kokumi substances on the threshold concentrations of key tastants such as glutamate. Thus, an extension of taste distribution to additional areas of the mouth (previously referred to as 'mouthfulness') and persistence of taste/flavor responses (previously referred to as 'continuity') could arise from a reduction in the threshold concentrations of umami and other substances that evoke taste responses.

The status of one of the compounds used as an inhibitor of C6A, the gallate derivative EGCG, as a potential inhibitor of the CaSR or T1R1/T1R3 is unknown. It would have been helpful to show that a specific inhibitor of the CaSR failed to block the ornithine response.

It would have been helpful to include a positive control kokumi substance in the two bottle preference experiment (e.g., one of the known gamma glutamyl peptides such as gamma-glu-Val-Gly or glutathione), to compare the relative potencies of the control kokumi compound and Ornithine, and to compare the sensitivities of the two responses to C6A and CaSR inhibitors.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public review):

Summary:

This paper contains what could be described as a "classic" approach towards evaluating a novel taste stimuli in an animal model, including standard behavioral tests (some with nerve transections), taste nerve physiology, and immunocytochemistry of the tongue. The stimulus being tested is ornithine, from a class of stimuli called "kokumi", which are stimuli that enhance other canonical tastes, increasing essentially the hedonic attributes of these other stimuli; the mechanism for ornithine detection is thought to be GPRC6A receptors expressed in taste cells. The authors showed evidence for this in an earlier paper with mice; this paper evaluates ornithine taste in a rat model.

Strengths:

The data show the effects of ornithine on taste: in two-bottle and briefer intake tests, adding ornithine results in a higher intake of most, but not all, stimuli tests. Bilateral nerve cuts or the addition of GPRC6A antagonists decrease this effect. Small effects of ornithine are shown in whole-nerve recordings.

Weaknesses:

The conclusion seems to be that the authors have found evidence for ornithine acting as a taste modifier through the GPRC6A receptor expressed on the anterior tongue. It is hard to separate their conclusions from the possibility that any effects are additive rather than modulatory. Animals did prefer ornithine to water when presented by itself. Additionally, the authors refer to evidence that ornithine is activating the T1R1-T1R3 amino acid taste receptor, possibly at higher concentrations than they use for most of the study, although this seems speculative. It is striking that the largest effects on taste are found with the other amino acid (umami) stimuli, leading to the possibility that these are largely synergistic effects taking place at the tas1r receptor heterodimer.

We would like to thank Reviewer #1 for the valuable comments. Our basis for considering ornithine as a taste modifier stems from our observation that a low concentration of ornithine (1 mM), which does not elicit a preference on its own, enhances the preference for umami substances, sucrose, and soybean oil through the activation of the GPRC6A receptor. Notably, this receptor is not typically considered a taste receptor. The reviewer suggested that the enhancement of umami taste might be due to potentiation occurring at the TAS1R receptor heterodimer. However, we propose that a different mechanism may be at play, as an antagonist of GPRC6A almost completely abolished this enhancement. In the revised manuscript, we will endeavor to provide additional information on the role of ornithine as a taste modifier acting through the GPRC6A receptor.

Reviewer #2 (Public review):

Summary:

The authors used rats to determine the receptor for a food-related perception (kokumi) that has been characterized in humans. They employ a combination of behavioral, electrophysiological, and immunohistochemical results to support their conclusion that ornithine-mediated kokumi effects are mediated by the GPRC6A receptor. They complemented the rat data with some human psychophysical data. I find the results intriguing, but believe that the authors overinterpret their data.

Strengths:

The authors examined a new and exciting taste enhancer (ornithine). They used a variety of experimental approaches in rats to document the impact of ornithine on taste preference and peripheral taste nerve recordings. Further, they provided evidence pointing to a potential receptor for ornithine.

Weaknesses:

The authors have not established that the rat is an appropriate model system for studying kokumi. Their measurements do not provide insight into any of the established effects of kokumi on human flavor perception. The small study on humans is difficult to compare to the rat study because the authors made completely different types of measurements. Thus, I think that the authors need to substantially scale back the scope of their interpretations. These weaknesses diminish the likely impact of the work on the field of flavor perception.

We would like to thank Reviewer #2 for the valuable comments and suggestions. Regarding the question of whether the rat is an appropriate model system for studying kokumi, we have chosen this species for several reasons: it is readily available as a conventional experimental model for gustatory research; the calcium-sensing receptor (CaSR), known as the kokumi receptor, is expressed in taste bud cells; and prior research has demonstrated the use of rats in kokumi studies involving gamma Glu-Val-Gly (Yamamoto and Mizuta, Chem. Senses, 2022).

We acknowledge that fundamentally different types of measurements were conducted in the human psychophysical study and the rat study. Kokumi can indeed be assessed and expressed in humans; however, we do not currently have the means to confirm that animals experience kokumi in the same way that humans do. Therefore, human studies are necessary to evaluate kokumi, a conceptual term denoting enhanced flavor, while animal studies are needed to explore the potential underlying mechanisms of kokumi. We believe that a combination of both human and animal studies is essential, as is the case with research on sugars. While sugars are known to elicit sweetness, it is unclear whether animals perceive sweetness identically to humans, even though they exhibit a strong preference for sugars. In the revised manuscript, we will incorporate additional information to address the comments raised by the reviewer. We will also carefully review and revise our previous statements to ensure accuracy and clarity.

Reviewer #3 (Public review):

Summary:

In this study, the authors set out to investigate whether GPRC6A mediates kokumi taste initiated by the amino acid L-ornithine. They used Wistar rats, a standard laboratory strain, as the primary model and also performed an informative taste test in humans, in which miso soup was supplemented with various concentrations of L-ornithine. The findings are valuable and overall the evidence is solid. L-Ornithine should be considered to be a useful test substance in future studies of kokumi taste and the class C G protein-coupled receptor known as GPRC6A (C6A) along with its homolog, the calcium-sensing receptor (CaSR) should be considered candidate mediators of kokumi taste.

Strengths:

The overall experimental design is solid based on two bottle preference tests in rats. After determining the optimal concentration for L-Ornithine (1 mM) in the presence of MSG, it was added to various tastants, including inosine 5'-monophosphate; monosodium glutamate (MSG); mono-potassium glutamate (MPG); intralipos (a soybean oil emulsion); sucrose; sodium chloride (NaCl); citric acid and quinine hydrochloride. Robust effects of ornithine were observed in the cases of IMP, MSG, MPG, and sucrose, and little or no effects were observed in the cases of sodium chloride, citric acid, and quinine HCl. The researchers then focused on the preference for Ornithine-containing MSG solutions. The inclusion of the C6A inhibitors Calindol (0.3 mM but not 0.06 mM) or the gallate derivative EGCG (0.1 mM but not 0.03 mM) eliminated the preference for solutions that contained Ornithine in addition to MSG. The researchers next performed transections of the chord tympani nerves (with sham operation controls) in anesthetized rats to identify the role of the chorda tympani branches of the facial nerves (cranial nerve VII) in the preference for Ornithine-containing MSG solutions. This finding implicates the anterior half-two thirds of the tongue in ornithine-induced kokumi taste. They then used electrical recordings from intact chorda tympani nerves in anesthetized rats to demonstrate that ornithine enhanced MSG-induced responses following the application of tastants to the anterior surface of the tongue. They went on to show that this enhanced response was insensitive to amiloride, selected to inhibit 'salt tastant' responses mediated by the epithelial Na+ channel, but eliminated by Calindol. Finally, they performed immunohistochemistry on sections of rat tongue demonstrating C6A positive spindle-shaped cells in fungiform papillae that partially overlapped in its distribution with the IP3 type-3 receptor, used as a marker of Type-II cells, but not with (i) gustducin, the G protein partner of Tas1 receptors (T1Rs), used as a marker of a subset of type-II cells; or (ii) 5-HT (serotonin) and Synaptosome-associated protein 25 kDa (SNAP-25) used as markers of Type-III cells.

Weaknesses:

The researchers undertook what turned out to be largely confirmatory studies in rats with respect to their previously published work on Ornithine and C6A in mice (Mizuta et al Nutrients 2021).

The authors point out that animal models pose some difficulties of interpretation in studies of taste and raise the possibility in the Discussion that umami substances may enhance the taste response to ornithine (Line 271, Page 9). They miss an opportunity to outline the experimental results from the study that favor their preferred interpretation that ornithine is a taste enhancer rather than a tastant.

At least two other receptors in addition to C6A might mediate taste responses to ornithine: (i) the CaSR, which binds and responds to multiple L-amino acids (Conigrave et al, PNAS 2000), and which has been previously reported to mediate kokumi taste (Ohsu et al., JBC 2010) as well as responses to Ornithine (Shin et al., Cell Signaling 2020); and (ii) T1R1/T1R3 heterodimers which also respond to L-amino acids and exhibit enhanced responses to IMP (Nelson et al., Nature 2001). While the experimental results as a whole favor the authors' interpretation that C6A mediates the Ornithine responses, they do not make clear either the nature of the 'receptor identification problem' in the Introduction or the way in which they approached that problem in the Results and Discussion sections. It would be helpful to show that a specific inhibitor of the CaSR failed to block the ornithine response. In addition, while they showed that C6A-positive cells were clearly distinct from gustducin-positive, and thus T1R-positive cells, they missed an opportunity to clearly differentiate C6A-expressing taste cells and CaSR-expressing taste cells in the rat tongue sections.

It would have been helpful to include a positive control kokumi substance in the two-bottle preference experiment (e.g., one of the known gamma-glutamyl peptides such as gamma-glu-Val-Gly or glutathione), to compare the relative potencies of the control kokumi compound and Ornithine, and to compare the sensitivities of the two responses to C6A and CaSR inhibitors.

The results demonstrate that enhancement of the chorda tympani nerve response to MSG occurs at substantially greater Ornithine concentrations (10 and 30 mM) than were required to observe differences in the two bottle preference experiments (1.0 mM; Figure 2). The discrepancy requires careful discussion and if necessary further experiments using the two-bottle preference format.

We would like to thank Reviewer #3 for the valuable comments and helpful suggestions. We propose that ornithine has two stimulatory actions: one acting on GPRC6A, particularly at lower concentrations, and another on amino acid receptors such as T1R1/T1R3 at higher concentrations. Consequently, ornithine is not preferable at lower concentrations but becomes preferable at higher concentrations. For our study on kokumi, we used a low concentration (1 mM) of ornithine. The possibility mentioned in the Discussion that 'the umami substances may enhance the taste response to ornithine' is entirely speculative. We will reconsider including this description in the revised version. As the reviewer suggested, in addition to GPRC6A, ornithine may bind to CaSR and/or T1R1/T1R3 heterodimers. However, we believe that ornithine mainly binds to GPRC6A, as a specific inhibitor of this receptor almost completely abolished the enhanced response to umami substances, and our immunohistochemical study indicated that GPRC6A-expressing taste cells are distinct from CaSR-expressing taste cells (see Supplemental Fig. 3). We conducted essentially the same experiments using gamma-Glu-Val-Gly in Wistar rats (Yamamoto and Mizuta, Chem. Senses, 2022) and compared the results in the Discussion. The reviewer may have misunderstood the chorda tympani results: we added the same concentration (1 mM) used in the two-bottle preference test to MSG (Fig. 5-B). Fig. 5-A shows nerve responses to five concentrations of plain ornithine. In the revised manuscript, we will strive to provide more precise information reflecting the reviewer’s comments.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) The behavioral effects found with the CPRC6A antagonists are not entirely convincing, as the antagonist is seemingly just mixed up in the solution with the stimuli. There are no control experiments demonstrating that the antagonists do not have a taste themselves.

We mixed the antagonists into both liquids used in the two-bottle preference test to eliminate any potential taste effects of the antagonists themselves. In the electrophysiological experiments, the antagonist was incorporated into the solution after confirming that it did not elicit any appreciable response in the taste nerve.

(2) The effects of ornithine found with quinine did not have a satisfying explanation - if there is some taste cell-taste cell modulation that accounts for the taste enhancement, why is the quinine less aversive? Why is it not enhanced like the other compounds?

The effects of ornithine on quinine responses remain difficult to explain. A previous study (Tokuyama et al., Chem Pharm Bull, 2006) proposed that ornithine prevents bitter substances from binding to bitter receptors, although this hypothesis lacks definitive evidence. In the present study, our findings suggest that the binding of quinine to bitter receptors is essential, as another agonist, gallate, also enhanced the preference for quinine, but this effect was abolished by EGCG, a GPRC6A antagonist (see Supplemental Fig. 2).

(3) Unless I am missing something, there appears to be no quantitative analysis of the immunocytochemical data, just assertions.

We have made quantitative analyses in the revised text, and the following sentences have been added: “Approximately 11% of GPRC6A-positive cells overlapped with IP3R3 (9 double-positive cells/80 GPRC6A-positive cells), while approximately 8.3% of IP3R3-positive cells expressed GPRC6A (9 double-positive /109 IP3R3-positive cells). In addition, GPRC6A-positive cells were unlikely to colocalize with a-gustducin, another marker for a subset of type II cells, in single taste cells (0 double-positive cell/93 GPRC6A-positive cells). Regarding type III cell markers, GPRC6A-positive cells were unlikely to colocalize with 5-HT in single taste cells (0 double-positive cell/75 GPRC6A-positive cells).”

(4) The hallmarks of Kokumi taste include descriptors such as "thickness", and "mouthfeel", which sound like potential somatosensory attributes. Perhaps the authors should consider this possibility for at least some of the effects found.

The term kokumi, a Japanese word, refers to a phenomenon in which the flavor of complexly composed food is enhanced through certain processes, making them more delicious. To date, kokumi has been described using the representative terms thickness, mouthfulness, and continuity, originally introduced in the first paper on kokumi by Ueda et al. (1990). However, these terms are derived from Japanese and may not fully convey the nuances of the original language when translated into these simple English words. In particular, thickness is often interpreted as referring to physical properties such as viscosity or somatosensory sensations. Since kokumi inherently lacks somatosensory elements, this revised paper adopts alternative terms and explanations for the three components of kokumi to prevent misunderstanding and confusion.

Therefore, to clarify that kokumi attributes are inherently gustatory, thickness is replaced with intensity of whole complex tastes (rich flavor with complex tastes), emphasizing the synergistic effects of a variety of tastes rather than the mere enhancement of a single flavor. Mouthfulness is clarified as not referring to mouthfeel (the tactile sensation a food gives in the mouth) but rather as spread of taste and flavor throughout the oral cavity, describing how the flavor fills the mouth. Continuity is replaced with persistence of taste (lingering flavor).

(5) I don't think the human experiment (S1) belongs to the paper, even as a supplementary bit of data. It's only 17 subjects, they are all female, and we don't know anything about how they were selected, even though it states they are all students/staff at Kio. Were any of them lab members? Were they aware of the goals of the experiment? Could simply increasing the amount of solute in the soup make it seem thicker? This (sparse) data seems to have been shoehorned into the paper without enough detail/justification.

Despite the reviewer’s suggestion, we would like to include the human experiment because the rationale of the present study is to confirm, through a human sensory test, that the kokumi of a complex solution (in this case, miso soup) is enhanced by the addition of ornithine. This is followed by basic animal experiments to investigate the underlying mechanisms. Therefore, this human study serves an important role.

The total number of participants increased to 22 (19 women and three men) following an additional experiment with 5 new participants. New results have been shown in Supplemental Figure 1 with statistical analyses. The rewritten parts are as follows:

We recruited 22 participants (19 women and three men, aged 21-28 years) from Kio University who were not affiliated with our laboratory, including students and staff members. All participants passed a screening test based on taste sensitivity. According to the responses obtained from a pre-experimental questionnaire, we confirmed that none of the participants had any sensory abnormalities, eating disorders, or mental disorders, or were taking any medications that may potentially affect their sense of taste. All participants were instructed not to eat or drink anything for 1 hour prior to the start of the experiment. We provided them with a detailed explanation of the experimental procedures, including safety measures and personal data protection, without revealing the specific goals of the study.

(6) The introduction could be more concise - for example, when describing Kokumi stimuli such as ornithine and its possible receptors, the authors do not need to add the detail about how this stimulus was deduced from adding clams to the soup. Details like this can be reserved for the discussion.

Thank you for this comment. We have tried to shorten the Introduction.

(7) Line 86: awkward phrasing - this doesn't need to be a rhetorical question.

We have deleted the sentence.

(8) Supplementary Figure 1: The labels on the figure say "Miso soup in 1 mM Orn" when the Orn is dissolved into the soup.

Thank you for pointing out our mistake. We have changed the description, such as “1 mM Orn in miso soup”.

Reviewer #2 (Recommendations for the authors):

Major concerns

(1) The impact of "kokumi" taste ligands on food perception appears to be profound in humans. This observation is fascinating because it implies that molecules like ornithine impact a variety of flavor perceptions, some of which are non-gustatory in nature (e.g., spread, mouthfulness and harmony). What remains unclear is whether "kokumi" ligands produce analogous sensations in rodents. If they don't, then rodents are an inappropriate model system for studying the impact of kokumi on flavor perceptions. The authors fail to address this key issue, and uncritically assume that kokumi ligands produce sensations like thickness, mouthfulness, and continuity in rodents. For this reason, the authors' reference to GPRC6A as a kokumi receptor is inappropriate.

Thank you very much for the valuable comments. The term kokumi refers to a phenomenon in which the flavor of complexly composed foods is enhanced through certain processes, making them more delicious. It is an important concept in the field of food science, which studies how to make prepared dishes more enjoyable. Kokumi is also considered a higher-order, profound cognitive function evaluated by humans who experience a wide variety of foods. However, it is unclear whether animals, particularly experimental animals, can perceive kokumi in the same way humans do.

To date, kokumi has been described using the representative terms thickness, mouthfulness, and continuity, originally introduced in the first paper on kokumi by Ueda et al. (1990). However, these terms are derived from Japanese and may not fully convey the nuances of the original language when translated into these simple English words. In particular, thickness is often interpreted as referring to physical properties such as viscosity or somatosensory sensations. Since kokumi inherently lacks somatosensory elements, this revised paper adopts alternative terms and explanations for the three components of kokumi to prevent misunderstanding and confusion.

Therefore, to clarify that kokumi attributes are inherently gustatory, thickness is replaced with intensity of whole complex tastes (rich flavor with complex tastes), emphasizing the synergistic effects of a variety of tastes rather than the mere enhancement of a single flavor. Mouthfulness is clarified as not referring to mouthfeel (the tactile sensation a food gives in the mouth) but rather as spread of taste and flavor throughout the oral cavity, describing how the flavor fills the mouth. Continuity is replaced with persistence of taste (lingering flavor).

Rodents are thought to possess basic taste functions similar to humans, such as the expression of taste receptors, including kokumi receptors, in taste cells. Regardless of whether rodents can perceive kokumi, findings from studies on rodents may provide insights into aspects of the kokumi concept as experienced by humans.

Indeed, the results of this study indicate that ornithine enhances umami, sweetness, fat taste, and saltiness, leading to the enhancement of complex flavors—referred to as intensity of whole taste. The activation of various taste cells, resulting in the enhancement of multiple tastes, may contribute to the sensation of flavors spreading throughout the oral cavity. Furthermore, the strong enhancement of MSG and MPG suggests that glutamate contributes to the mouthfulness and persistence of taste characteristic of kokumi.

(2) A related concern is that the authors did not make any measurements that model kokumi sensations documented in the literature. For example, they would need to develop behavioral/electrophysiological measurements that reflect the known effects of kokumi ligands on flavor perception (i.e., increases in intensity, spread, continuity, richness, harmony, and punch). For example, ornithine is thought to produce more "punch" (i.e., a more rapid rise in intensity). This could be manifested as a more rapid rise in peripheral taste response or a more rapid fMRI response in the taste cortex. Alternatively, ornithine is thought to increase "continuity" (i.e., make the taste response more persistent). This response would presumably be manifested as a peripheral taste response that adapts more slowly or a more persistent fMRI response. As it stands, the authors have documented that ornithine increases (i) the preference of rats for some chemical stimuli, but not others; and (ii) the response of the CT nerve to some but not all taste stimuli.

In animal experiments, it is challenging to examine each attribute of kokumi. The increase of complex tastes can be investigated through behavioral experiments and neural activity recordings. However, phenomena such as spread or harmony, which arise from profound human judgments, are difficult to validate in animal studies.

While it was possible to examine persistence through neural responses to tastants, all stimuli were rinsed at 30 seconds after onset of stimulation, so the exact duration of persistence was not investigated. However, since the MSG response was enhanced approximately 1.5 times with the addition of ornithine, it is strongly suggested that the duration might also have been prolonged.

Regarding punch, no differences were observed in the neural responses when ornithine was added, likely because the phasic response already had a rapid onset.

In the context of fMRI studies, there has been a report that adding glutathione to mixtures of umami and salt solutions increases responses (Goto et al. Chem Senses, 2016). However, research specifically examining the attributes of kokumi has not yet been reported.

(3) The quality of the SNAP-25 immunohistochemistry is poor (see Figure 7D), with lots of seemingly nonspecific staining in and outside the taste bud.

The quality of the SNAP-25 is not poor. It is known that SNAP-25 labels not only type III cells but also the dense network of intragemmal nerve fibers (Tizzano et al., Immunohistochemical Analysis of Human Vallate Taste Buds. Chem Senses.40:655-60, 2015). Therefore, lots of seemingly nonspecific staining is due to intense SNAP-25-immunoreactivity of the nerve fibers.

(4) The authors need to drastically scale back the scope of their conclusions. What they can say is that ornithine appears to enhance the taste responses of rats to a variety of taste stimuli and that this effect appears to be mediated by the GPRC6A receptor. They cannot use their data to address kokumi effects in humans, as they have not attempted to model any of these effects. Given the known problems with pharmacological blocking agents (e.g., nonspecificity), the authors would significantly strengthen their case if they could generate similar results in a GPRC6A knockout mouse.

Our research approach begins with confirming in humans that the addition of ornithine to complex foods (such as miso soup) induces kokumi. Based on this confirmation, we conduct fundamental studies using animal models to investigate the peripheral taste mechanisms underlying the expression of kokumi.

It is possible that the key to kokumi expression lies in the enhancement of desirable tastes (particularly umami) and the suppression of unpleasant tastes. Moving forward, we will deepen our fundamental research on the action of ornithine mediated through GPRC6A, including studies using knockout mice.

(5) The introduction is too long. Much of the discussion of kokumi perception in humans should either be removed or shortened considerably.

Following the reviewer’s suggestion, the introduction has been shortened.

(6) I recommend that the authors break up the Methods and Results sections into different experiments. This would enable the authors to provide separate rationales for each procedure. For instance, the authors conducted a variety of different behavioral procedures (e.g., long- and short-term preference tests, and preference tests with and without GPRC6A receptor antagonists).

Rather than following the reviewer’s suggestion, we have added subheadings to describe the purpose of each experiment. This approach would help readers better understand the experimental flow, as each experiment is relatively straightforward.

(7) The inclusion of the human data is odd for two reasons. First, the measurements used to assess the impact of ornithine on flavor perception in humans were totally different than those used in rats. This makes it impossible to compare the human and rat datasets. Second, the human study was rather limited in scope, had small effect sizes, and had a lot of individual variation. For these reasons, the human data are not terribly helpful. I recommend that the authors remove the human data from this paper, and publish them as part of a more extensive study on humans.

Despite the reviewer’s suggestion, we would like to include the human experiment because the rationale of the present study is to confirm, through a human sensory test, that the kokumi of a complex solution (in this case, miso soup) is enhanced by the addition of ornithine. This is followed by basic animal experiments to investigate the underlying mechanisms. Therefore, this human study serves an important role. The considerable variation in the scores suggests that evaluating the three kokumi attributes is challenging and likely influenced by differences in judgment criteria among participants.

The total number of participants increased to 22 (19 women and three men) following an additional experiment with 5 new participants. New results have been shown in Supplemental Figure 1 with statistical analyses. The rewritten parts are as follows:

We recruited 22 participants (19 women and three men, aged 21-28 years) from Kio University who were not affiliated with our laboratory, including students and staff members. All participants passed a screening test based on taste sensitivity. According to the responses obtained from a pre-experimental questionnaire, we confirmed that none of the participants had any sensory abnormalities, eating disorders, or mental disorders, or were taking any medications that may potentially affect their sense of taste. All participants were instructed not to eat or drink anything for 1 hour prior to the start of the experiment. We provided them with a detailed explanation of the experimental procedures, including safety measures and personal data protection, without revealing the specific goals of the study.

(8) While the use of English is generally good, there are many instances where the English is a bit awkward. I recommend that the authors ask a native English speaker to edit the text.

Thank you for this comment. The text has been edited by a native English speaker.

Minor concerns

(1) Lines 13-14: The authors state that "the concept of 'kokumi' has garnered significant attention in gustatory physiology and food science." This is an exaggeration. Kokumi has generated considerable interest in food science but has yet to generate much interest in gustatory physiology.

We have rewritten this part: “The concept of “kokumi” has generated considerable interest in food science but kokumi has not been well studied in gustatory physiology.”

(2) Line 20: The use of "specific taste" is unclear in this context. The authors indicate (in Figure 5A) that 1 mM ornithine generates a CT nerve response. They also reveal (in Figure 1A) that rats do not prefer 1 mM ornithine over water. The results from a preference test do not provide insight into whether a solution can be tasted; they merely demonstrate a lack of preference for that solution. Based on these data, the authors cannot infer that 1 mM ornithine cannot be tasted.

We agree with the reviewer’s comment. Ornithine at 1 mM concentration may have a weak taste because this solution elicited a small neural response (Fig. 5-A). We have rewritten the text: “… at a concentration without preference for this solution.”

(3) Line 44: Sensory information from foods enters the oral and the nasal cavity.

The nasal cavity has been added.

(5) Lines 59: The terms "thickness", "mouthfulness" and "continuity" are not intuitive in English, and may reflect, at least in part, a failure in translation. The word thickness implies a tactile sensation (e.g., owing to high viscosity), but the authors use it to indicate a flavor that is more intense and onsets more quickly. The word mouthfulness is supposed to indicate that a flavor is experienced throughout the oral cavity. The problem here is that this happens with all tastants, independent of the presence of substances like ornithine. Indeed, taste buds occur in a limited portion of the oral epithelium, but we nevertheless experience tastes throughout the oral cavity, owing to a phenomenon called tactile referral (see the following reference: Todrank and Bartoshuk, 1991, A taste illusion: taste sensation localized by touch" Physiology & Behavior 50:1027-1031). The word continuity does not imply that the taste is long-lasting or persistent.

These three attributes were originally introduced by Ueda et al. (1990), who translated Japanese terms describing the profound characteristics of kokumi, which are deeply rooted in Japanese culinary culture. However, these simply translated terms have caused global misunderstanding and confusion, because they sound like somatosensory rather than gustatory descriptions. Therefore, to clarify that kokumi attributes are inherently gustatory, in the revised version we use the terms “intensity of whole complex tastes (rich flavor with complex tastes)” instead of thickness, “mouthfulness (spread of taste and flavor throughout the oral cavity),” and “persistence of taste (lingering flavor)” instead of continuity.

The results of this study indicate that ornithine enhances umami, sweetness, fat taste, and saltiness, leading to the enhancement of complex flavors—referred to as intensity of whole taste. The activation of various taste cells, resulting in the enhancement of multiple tastes, may contribute to the sensation of flavors spreading throughout the oral cavity. Furthermore, the strong enhancement of MSG and MPG suggests that glutamate contributes to the mouthfulness and persistence of taste characteristic of kokumi.

(6) Figure legends: The authors provide results of statistical comparisons in several of the figures. They need to explain what statistical procedures were performed. As it stands, it is impossible to interpret the asterisks provided.

We have explained statistical procedures in each Figure legend.

(7) I did not see any reference to the sources of funding or any mention of potential conflicts of interest.

We have added the following information:

Funding: JSPS KAKENHI Grant Numbers JP17K00935 (to TY) and JP22K11803(to KU).

Declaration of interests: The authors declare that they have no competing interests.

Reviewer #3 (Recommendations for the authors):

(1) I suggest that the authors increase their level of interest in glutathione and gamma-glutamyl peptides. This might include an appropriate gamma-glutamyl control substance in the two-bottle preference study (see Public Review). It might also include more careful attention to the work that identified glutathione as an activator of the CaSR (Wang et al., JBC 2006) and the nature of its binding site on the CaSR which overlaps with its site for L-amino acids (Broadhead et al., JBC 2011). This latter article also identified S-methyl glutathione, in which the free-SH group is blocked, as a high-potency activator of the CaSR. It would be expected to show comparable potency to gamma-glu-Val-Gly in assays of kokumi taste.

We have appropriately referenced glutathione and gamma-Glu-Val-Gly, potent agonists of CaSR, where necessary. In our previous study (Yamamoto and Mizuta, Chem Senses, 2022), we examined the additive effects of these substances on basic taste stimuli in rodents, and the results were compared in greater detail with those obtained from the addition of ornithine in the present study. We have also discussed the potential binding of ornithine to other receptors, including CaSR and T1R1/T1R3 heterodimers.

(2) Figures:

-None of the figures were labelled with their Figure numbers. I have inferred the Figure numbers from the legends and their positions in the pdf.

We are sorry for this inconvenience.

- The labelling of Figure 1 and Figure 2 are problematic. In Figure 1 it should be made clear that the horizontal axes refer to the Ornithine concentration. In Figure 2 it should be made clear that the horizontal axes refer to the tastant concentrations (MSG, IMP, etc) and that the Ornithine concentrations were fixed at either zero or 1.0 mM.

We are sorry for the lack of information about the horizontal axes. We have explained the horizontal axes in figure legends in Figs. 1 and 2. The labelling of both figures has also been modified to make this clear.

- Figure 3B: 'Control' should appear at the top of this panel since the panels that follow all refer to it.

Following the reviewer’s suggestion, we have added ‘Control’ at the top of Figure 3B.

- Figure 5A. Provide a label for the test substance, presumably Ornithine.

Yes, we have added ‘Ornithine’.

- Figure 7 would be strengthened by the inclusion of immunohistochemistry analyses of the CaSR.

We are sorry that we did not analyze immunohistochemistry for the CaSR because a previous study precisely had analyzed the CaSR expression on taste cells in rats. We have analyzed co-expression of GPRC6A and CaSR (see Supplemental Figure 3).

(3) Other Matters:

- Line 38: list the five basic taste modalities here.

Yes, we have included the five basic taste modalities here.

- Line 107: 'even if ... kokumi ... is less developed in rodents' - if there is evidence that kokumi is less developed in rodents it should be cited here.

We cannot cite any references here because no studies have compared the perception of kokumi between humans and rodents.

- Line 308: 'recently we conducted experiments in rats using gallate ...' - the authors appear to imply that they performed the research in Reference 43, however, I was unable to find an overlap between the two lists of authors.

We are not doing a similar study as the research in Reference 43 (40 in the revised paper). Following the result that gallate is an agonist of GPRC6A as shown by Reference 43, we were interested in doing similar behavioral experiments using gallate instead of ornithine.

The sentences have been rewritten to avoid misunderstanding.

- Line 506: the sections are said to be 20 mm thick - should this read 20 micrometers?

Thank you. We have changed to 20 micrometers.

eLife Assessment

In this important study, the authors use zebrafish to examine protein absorption in the gut. Using a combination of imaging and single-cell RNA-seq, they characterize a population of lysosome-rich enterocytes that are important for protein uptake. They find that the microbiome impacts the ability of these cells to uptake protein. The RNA-seq provides a rich dataset for future functional experiments, which makes a convincing case for the importance of these cells.

Reviewer #1 (Public review):

The Bagnat and Rawls groups' previous published work (Park et al., 2019) described the kinetics and genetic basis of protein absorption in a specialized cell population of young vertebrates termed lysosome-rich enterocytes (LREs). In this study they seek to understand how the presence and composition of the microbiota impacts the protein absorption function of these cells and reciprocally, how diet and intestinal protein absorption function impact the microbiome.

Strengths of the study include the functional assays for protein absorption performed in live larval zebrafish, which provides detailed kinetics on protein uptake and degradation with anatomic precision, and the gnotobiotic manipulations. The authors clearly show that the presence of the microbiota or of certain individual bacterial members slows the uptake and degradation of multiple different tester fluorescent proteins.

To understand the mechanistic basis for these differences, the authors also provide detailed single-cell transcriptomic analyses of cells isolated based on both an intestinal epithelial cell identity (based on a transgenic marker) and their protein uptake activity. The data generated from these analyses, presented in Figures 3-5, are valuable for expanding knowledge about zebrafish intestinal epithelial cell identities, but of more limited interest to a broader readership. Some of the descriptive analysis in this section is circular because the authors define subsets of LREs (termed anterior and posterior) based on their fabp2 expression levels, but then go on to note transcriptional differences between these cells (for example in fabp2) that are a consequence of this initial subsetting.

Inspired by their single-cell profiling and by previous characterization of the genes required for protein uptake and degradation in the LREs, the authors use quantitative hybridization chain reaction RNA-fluorescent in situ hybridization to examine transcript levels of several of these genes along the length of the LRE intestinal region of germ-free versus mono-associated larvae. They provide good evidence for reduced transcript levels of these genes that correlate with the reduced protein uptake in the mono-associated larval groups.

The final part of the study (shown in Figure 7) characterized the microbiomes of 30-day-old zebrafish reared from 6-30 days on defined diets of low and high protein and with or without homozygous loss of the cubn gene required for protein uptake. The analysis of these microbiomes notes some significant differences between fish genotypes by diet treatments, but the discussion of these data does not provide strong support for the hypothesis that "LRE activity has reciprocal effects on the gut microbiome". The most striking feature of the MDS plot of Bray Curtis distance between zebrafish samples shown in Figure 7B is the separation by diet independent of host genotype, which is not discussed in the associated text. Additionally, the high protein diet microbiomes have a greater spread than those of the low protein treatment groups, with the high protein diet cubn mutant samples being the most dispersed. This pattern is consistent with the intestinal microbiota under a high protein diet regimen and in the absence of protein absorption machinery being most perturbed in stochastic ways than in hosts competent for protein uptake, consistent with greater beta dispersal associated with more dysbiotic microbiomes (described as the Anna Karenina principle here: https://pubmed.ncbi.nlm.nih.gov/28836573/). It would be useful for the authors to provide statistics on the beta dispersal of each treatment group.

Overall, this study provides strong evidence that specific members of the microbiota differentially impact gene expression and cellular activities of enterocyte protein uptake and degradation, findings that have a significant impact on the field of gastrointestinal physiology. The work refines our understanding of intestinal cell types that contribute to protein uptake and their respective transcriptomes. The work also provides some evidence that microbiomes are modulated by enterocyte protein uptake capacity in a diet-dependent manner. These latter findings provide valuable datasets for future related studies.

Reviewer #2 (Public review):

Summary:

The authors set out to determine how the microbiome and host genotype impact host protein-based nutrition.

Strengths:

The quantification of protein uptake dynamics is a major strength of this work and the sensitivity of this assay shows that the microbiome and even mono-associated bacterial strains dampen protein uptake in the host by causing down-regulation of genes involved in this process rather than a change in cell type.

The use of fluorescent proteins in combination with transcript clustering in the single cell seq analysis deepens our understanding of the cells that participate in protein uptake along the intestine. In addition to the lysozome-rich enterocytes (LRE), subsets of enteroendocrine cells, acinar, and goblet cells also take up protein. Intriguingly, these non-LRE cells did not show lysosomal-based protein degradation; but importantly analysis of the transcripts upregulated in these cells include dab2 and cubn, genes shown previously as being essential to protein uptake.

The derivation of zebrafish mono-associated with single strains of microbes paired with HCR to localize and quantify the expression of host protein absorption genes shows that different bacterial strains suppress these genes to variable extents.

The analysis of microbiome composition, when host protein absorption is compromised in cubn-/- larvae or by reducing protein in the food, demonstrates that changes to host uptake can alter the abundance of specific microbial taxa like Aeramonas.

Weaknesses:

The finding that neurons are positive for protein uptake in the single-cell data set is not adequately discussed. It is curious because the cldn:GFP line used for sorting does not mark neurons and if the neurons are taking up mCherry via trans-synaptic uptake from EECs, those neurons should be mCherry+/GFP-; yet methods indicate GFP+ and GFP+/mCherry+ cells were the ones collected and analyzed.

Reviewer #3 (Public review):

Summary:

Childers et al. address a fundamental question about the complex relationship within the gut: the link between nutrient absorption, microbial presence, and intestinal physiology. They focus on the role of lysosome-rich enterocytes (LREs) and the microbiota in protein absorption within the intestinal epithelium. By using germ-free and conventional zebrafishes, they demonstrate that microbial association leads to a reduction in protein uptake by LREs. Through impressive in vivo imaging of gavaged fluorescent proteins, they detail the degradation rate within the LRE region, positioning these cells as key players in the process. Additionally, the authors map protein absorption in the gut using single-cell sequencing analysis, extensively describing LRE subpopulations in terms of clustering and transcriptomic patterns. They further explore the monoassociation of ex-germ-free animals with specific bacterial strains, revealing that the reduction in protein absorption in the LRE region is strain-specific.

Strengths:

The authors employ state-of-the-art imaging to provide clear evidence of the protein absorption rate phenotype, focusing on a specific intestinal region. This innovative method of fluorescent protein tracing expands the field of in vivo gut physiology.

Using both conventional and germ-free animals for single-cell sequencing analysis, they offer valuable epithelial datasets for researchers studying host-microbe interactions. By capitalizing on fluorescently labelled proteins in vivo, they create a new and specific atlas of cells involved in protein absorption, along with a detailed LRE single-cell transcriptomic dataset.

Weaknesses:

While the authors present tangible hypotheses, the data are primarily correlative, and the statistical methods are inadequate. They examine protein absorption in a specific, normalized intestinal region but do not address confounding factors between germ-free and conventional animals, such as size differences, transit time, and oral gavage, which may impact their in vivo observations. This oversight can lead to bold conclusions, where the data appear valuable but require more nuance.

The sections of the study describing the microbiota or attempting functional analysis are elusive, with related data being overinterpreted. The microbiome field has long used 16S sequencing to characterize the microbiota, but its variability due to experimental parameters limits the ability to draw causative conclusions about the link between LRE activity, dietary protein, and microbial composition. Additionally, the complex networks involved in dopamine synthesis and signalling cannot be fully represented by RNA levels alone. The authors' conclusions on this biological phenomenon based on single-cell data need support from functional and in vivo experiments.

Author response:

Public Reviews: 

Reviewer #1 (Public review): 

The Bagnat and Rawls groups' previous published work (Park et al., 2019) described the kinetics and genetic basis of protein absorption in a specialized cell population of young vertebrates termed lysosome-rich enterocytes (LREs). In this study they seek to understand how the presence and composition of the microbiota impacts the protein absorption function of these cells and reciprocally, how diet and intestinal protein absorption function impact the microbiome. 

Strengths of the study include the functional assays for protein absorption performed in live larval zebrafish, which provides detailed kinetics on protein uptake and degradation with anatomic precision, and the gnotobiotic manipulations. The authors clearly show that the presence of the microbiota or of certain individual bacterial members slows the uptake and degradation of multiple different tester fluorescent proteins. 

To understand the mechanistic basis for these differences, the authors also provide detailed single-cell transcriptomic analyses of cells isolated based on both an intestinal epithelial cell identity (based on a transgenic marker) and their protein uptake activity. The data generated from these analyses, presented in Figures 3-5, are valuable for expanding knowledge about zebrafish intestinal epithelial cell identities, but of more limited interest to a broader readership. Some of the descriptive analysis in this section is circular because the authors define subsets of LREs (termed anterior and posterior) based on their fabp2 expression levels, but then go on to note transcriptional differences between these cells (for example in fabp2) that are a consequence of this initial subsetting. 

Inspired by their single-cell profiling and by previous characterization of the genes required for protein uptake and degradation in the LREs, the authors use quantitative hybridization chain reaction RNA-fluorescent in situ hybridization to examine transcript levels of several of these genes along the length of the LRE intestinal region of germ-free versus mono-associated larvae. They provide good evidence for reduced transcript levels of these genes that correlate with the reduced protein uptake in the mono-associated larval groups. 

The final part of the study (shown in Figure 7) characterized the microbiomes of 30-day-old zebrafish reared from 6-30 days on defined diets of low and high protein and with or without homozygous loss of the cubn gene required for protein uptake. The analysis of these microbiomes notes some significant differences between fish genotypes by diet treatments, but the discussion of these data does not provide strong support for the hypothesis that "LRE activity has reciprocal effects on the gut microbiome". The most striking feature of the MDS plot of Bray Curtis distance between zebrafish samples shown in Figure 7B is the separation by diet independent of host genotype, which is not discussed in the associated text. Additionally, the high protein diet microbiomes have a greater spread than those of the low protein treatment groups, with the high protein diet cubn mutant samples being the most dispersed. This pattern is consistent with the intestinal microbiota under a high protein diet regimen and in the absence of protein absorption machinery being most perturbed in stochastic ways than in hosts competent for protein uptake, consistent with greater beta dispersal associated with more dysbiotic microbiomes (described as the Anna Karenina principle here: https://pubmed.ncbi.nlm.nih.gov/28836573/). It would be useful for the authors to provide statistics on the beta dispersal of each treatment group. 

Overall, this study provides strong evidence that specific members of the microbiota differentially impact gene expression and cellular activities of enterocyte protein uptake and degradation, findings that have a significant impact on the field of gastrointestinal physiology. The work refines our understanding of intestinal cell types that contribute to protein uptake and their respective transcriptomes. The work also provides some evidence that microbiomes are modulated by enterocyte protein uptake capacity in a diet-dependent manner. These latter findings provide valuable datasets for future related studies. 

We thank the reviewer for their thorough and kind assessment. We appreciate the suggestion for edits and for pointing out areas that need further clarification.

One point that clearly needs further explanation is the use fabp6 (referred to as fabp2 by the reviewer) to define anterior LREs and their gene expression pattern. which includes high levels of fabp6. This was deemed by the reviewer as a “circular argument”.  We would like to clarify that the rationale for using fabp6 as anchor is that we had previously reported overlap between fabp6 and LREs (see Fig.6C-E in Wen et al. PMID: 34301599) and thus were able here to define fabp6’s spatial pattern in relation to other LRE markers and the neighboring ileocyte population using transgenic markers and HCR. Thus, far from being a circular argument, using fabp6 allowed us to identify other markers that are differentially expressed between anterior and posterior LREs, which share a core program that we highlight in our study. In the revised manuscript we will clarify this point.

We will also add the analysis suggested for the 16S rRNA gene sequencing data, include statistics on beta dispersal, and expand the discussion of these data as suggested.

Reviewer #2 (Public review): 

Summary: 

The authors set out to determine how the microbiome and host genotype impact host protein-based nutrition. 

Strengths: 

The quantification of protein uptake dynamics is a major strength of this work and the sensitivity of this assay shows that the microbiome and even mono-associated bacterial strains dampen protein uptake in the host by causing down-regulation of genes involved in this process rather than a change in cell type. 

The use of fluorescent proteins in combination with transcript clustering in the single cell seq analysis deepens our understanding of the cells that participate in protein uptake along the intestine. In addition to the lysozome-rich enterocytes (LRE), subsets of enteroendocrine cells, acinar, and goblet cells also take up protein. Intriguingly, these non-LRE cells did not show lysosomal-based protein degradation; but importantly analysis of the transcripts upregulated in these cells include dab2 and cubn, genes shown previously as being essential to protein uptake. 

The derivation of zebrafish mono-associated with single strains of microbes paired with HCR to localize and quantify the expression of host protein absorption genes shows that different bacterial strains suppress these genes to variable extents. 

The analysis of microbiome composition, when host protein absorption is compromised in cubn-/- larvae or by reducing protein in the food, demonstrates that changes to host uptake can alter the abundance of specific microbial taxa like Aeramonas. 

Weaknesses: 

The finding that neurons are positive for protein uptake in the single-cell data set is not adequately discussed. It is curious because the cldn:GFP line used for sorting does not mark neurons and if the neurons are taking up mCherry via trans-synaptic uptake from EECs, those neurons should be mCherry+/GFP-; yet methods indicate GFP+ and GFP+/mCherry+ cells were the ones collected and analyzed. 

We thank the Reviewer for the kind and positive assessment of our work, for suggestions to improve the accessibility and clarity of the manuscript, and for pointing out an issue related to a neuronal population that needs further clarification.

We confirm that there is a population of neurons that express cldn15la (and cldn15la:GFP). They are not easily visualized by microscopy because IECs express this gene at a relatively much higher level. However, the endogenous cldn15la transcript can be found in a recently published dataset (PMID: 35108531) as well as in ours. We will add a Discussion point to clarify this issue.

Reviewer #3 (Public review): 

Summary: 

Childers et al. address a fundamental question about the complex relationship within the gut: the link between nutrient absorption, microbial presence, and intestinal physiology. They focus on the role of lysosome-rich enterocytes (LREs) and the microbiota in protein absorption within the intestinal epithelium. By using germ-free and conventional zebrafishes, they demonstrate that microbial association leads to a reduction in protein uptake by LREs. Through impressive in vivo imaging of gavaged fluorescent proteins, they detail the degradation rate within the LRE region, positioning these cells as key players in the process. Additionally, the authors map protein absorption in the gut using single-cell sequencing analysis, extensively describing LRE subpopulations in terms of clustering and transcriptomic patterns. They further explore the monoassociation of ex-germ-free animals with specific bacterial strains, revealing that the reduction in protein absorption in the LRE region is strain-specific. 

Strengths: 

The authors employ state-of-the-art imaging to provide clear evidence of the protein absorption rate phenotype, focusing on a specific intestinal region. This innovative method of fluorescent protein tracing expands the field of in vivo gut physiology. 

Using both conventional and germ-free animals for single-cell sequencing analysis, they offer valuable epithelial datasets for researchers studying host-microbe interactions. By capitalizing on fluorescently labelled proteins in vivo, they create a new and specific atlas of cells involved in protein absorption, along with a detailed LRE single-cell transcriptomic dataset. 

Weaknesses: 

While the authors present tangible hypotheses, the data are primarily correlative, and the statistical methods are inadequate. They examine protein absorption in a specific, normalized intestinal region but do not address confounding factors between germ-free and conventional animals, such as size differences, transit time, and oral gavage, which may impact their in vivo observations. This oversight can lead to bold conclusions, where the data appear valuable but require more nuance. 

The sections of the study describing the microbiota or attempting functional analysis are elusive, with related data being overinterpreted. The microbiome field has long used 16S sequencing to characterize the microbiota, but its variability due to experimental parameters limits the ability to draw causative conclusions about the link between LRE activity, dietary protein, and microbial composition. Additionally, the complex networks involved in dopamine synthesis and signalling cannot be fully represented by RNA levels alone. The authors' conclusions on this biological phenomenon based on single-cell data need support from functional and in vivo experiments. 

We thank the reviewer for their assessment and for pointing out some areas that need to be explained better and/or discussed further.

The reviewer mentions some potential confounding factors (ie., size differences, transit time, oral gavage) in the gnotobiotic experiments. We would like to convey that these aspects have been addressed in our experimental design and will be clarified in our full in the revised manuscript by adding information to Methods or by adding data statements. Briefly: 1-larval sizes were recorded and found to be similar between GF and monoassociated larvae. A statement will be added to text.; 2-while intestinal transit time has been reported to be affected by microbes in larval zebrafish (PMIDs: 16781702, 28207737, 33352109) and is a topic of interest, it does not represent a confounding factor for our experiments. In our assay, luminal cargo is present at high concentrations throughout the gut and is not limiting at any point during the assay; 3-gavage, which is necessary for quantitative assays, is indeed an experimental manipulation that may somehow alter the subjects (the same is true for microscopy and virtually any research method). However, any potential effects of gavage manipulation would not explain differences between GF and CV animals or alter our conclusions about microbial or dietary effects. We will elaborate on this in the revised Discussion.

We acknowledge that microbiota composition is prone to relatively high degrees of interindividual and interexperimental variation, and that measuring microbiota composition using 16S rRNA gene sequencing is accompanied by inherent technical limitations such as limited taxonomic resolution, primer bias, etc.  It is important to note that comparable assays such as shotgun metagenomic DNA sequencing are not currently suitable for samples such as larval zebrafish or their dissected digestive tracts where the relative superabundance of host DNA prevents adequate coverage of microbial DNA. However, 16S rRNA gene sequencing remains a mainstream assay in the larger microbial ecology field, has proven effective at revealing important impacts of environmental factors on the gut microbiota (PMIDs: 21346791, 31409661, 31324413). Our results here also illustrate how 16S rRNA gene sequencing can be a useful method to detect perturbations to the zebrafish gut microbiome. Reproducing previous findings, we detected in our samples many of the core zebrafish microbiota taxa that have been identified by other studies (PMIDs: 26339860, 21472014, 17055441). To increase the robustness of our results, we included several biological replicates for each condition, co-housed genotypes and included large sample sizes to minimize environmental variation between groups. Importantly, replicates housed in different tanks showed similar results. We will emphasize these points in the revised Discussion. To further underscore this in the revised manuscript, we will add a beta diversity plot and statistical analysis showing that the microbiome was not significantly affected by our experimental replicates.

Regarding dopamine pathways, we thank the reviewer for pointing out that the language we used in our interpretation of this and other pathways enriched in our scRNAseq data was too strong. In the revised manuscript, we will soften those conclusions, and instead indicate that these may be areas worthy of future dedicated investigation.

Finally, the reviewer mentions the use of inadequate statistical methods for some analyses but without specifying or indicating alternative analyses. Only the need to justify the use of two-way ANOVA was made explicit. In this point, we respectfully disagree and would like to emphasize that we use statistical methods that are standards in the field. We will nevertheless add a justification for the use of two-way ANOVA where appropriate. Briefly, the two-way ANOVA test was used to compare fluorescence profiles of gavages cargoes or HCR probes at each level along the length of the LRE region. This test accounts for differences in fluorescence between experimental conditions at each level (binned 30 μm areas) along the LRE region (~300 μm). This test allows us to capture differences in fluorescence between experimental conditions while accounting for heterogeneity in the LRE region.

eLife Assessment

This study presents fundamental findings that could redefine the specificity and mechanism of action of the well-studied Ser/Thr kinase IKK2 (a subunit of inhibitor of nuclear factor kappa-B kinase (IkB) that propagates cellular response to inflammation). Solid evidence supports the claim that IKK2 exhibits dual specificity that allows tyrosine autophosphorylation and the authors further show that auto-phosphorylated IKK2 is involved in an unanticipated relay mechanism that transfers phosphate from an IKK2 tyrosine onto the IkBa substrate. The findings are a starting point for follow-up studies to confirm the unexpected mechanism and further pursue functional significance.

Reviewer #1 (Public review):

IKK is the key signaling node for inflammatory signaling. Despite the availability of molecular structures, how the kinase achieves its specificity remains unclear. This paper describes a dynamic sequence of events in which autophosphorylation of a tyrosine near the activate site facilitates phosphorylation of the serine on the substrate via a phosphor-transfer reaction. The proposed mechanism is conceptually novel in several ways, suggesting that the kinase is dual specificity (tyrosine and serine) and that it mediates a phospho-transfer reaction. While bacteria contain phosphorylation-transfer enzymes, this is unheard of for mammalian kinases. However, what the functional significance of this enzymatic activity might remain unaddressed.

The revised manuscript adequately addresses all the points I suggested in the review of the first submission.

Reviewer #2 (Public review):

The authors investigate the phosphotransfer capacity of Ser/Thr kinase IκB kinase (IKK), a mediator of cellular inflammation signaling. Canonically, IKK activity is promoted by activation loop phosphorylation at Ser177/Ser181. Active IKK can then unleash NF-κB signaling by phosphorylating repressor IκBα at residues Ser32/Ser26. Noting the reports of other IKK phosphorylation sites, the authors explore the extent of autophosphorylation.

Semi-phosphorylated IKK purified from Sf9 cells, exhibits the capacity for further autophosphorylation. Anti-phosphotyrosine immunoblotting indicated unexpected tyrosine phosphorylation. Contaminating kinase activity was tested by generating a kinase-dead K44M variant, supporting the notion that the unexpected phosphorylation was IKK-dependent. In addition, the observed phosphotyrosine signal required phosphorylated IKK activation loop serines.

Two candidate IKK tyrosines were examined as the source of the phosphotyrosine immunoblotting signal. Activation loop residues Tyr169 and Tyr188 were each rendered non-phosphorylatable by mutation to Phe. The Tyr variants decreased both autophosphorylation and phosphotransfer to IκBα. Likewise, Y169F and Y188F IKK2 variants immunoprecipitated from TNFa-stimulated cells also exhibited reduced activity in vitro.

The authors further focus on Tyr169 phosphorylation, proposing a role as a phospho-sink capable of phosphotransfer to IκBα substrate. This model is reminiscent of the bacterial two-component signaling phosphotransfer from phosphohistidine to aspartate. Efforts are made to phosphorylate IKK2 and remove ATP to assess the capacity for phosphotransfer. Phosphorylation of IκBα is observed after ATP removal, although there are ambiguous requirements for ADP.

Strengths:

Ultimately, the authors draw together the lines of evidence for IKK2 phosphotyrosine and ATP-independent phosphotransfer to develop a novel model for IKK2-mediated phosphorylation of IκBα. The model suggests that IKK activation loop Ser phosphorylation primes the kinase for tyrosine autophosphorylation. With the assumption that IKK retains the bound ADP, the phosphotyrosine is conformationally available to relay the phosphate to IκBα substrate. The authors are clearly aware of the high burden of evidence required for this unusual proposed mechanism. Indeed, many possible artifacts (e.g., contaminating kinases or ATP) are anticipated and control experiments are included to address many of these concerns. The analysis hinges on the fidelity of pan-specific phosphotyrosine antibodies, and the authors have probed with two different anti-phosphotyrosine antibody clones. Taken together, the observations are thought-provoking, and I look forward to seeing this model tested in a cellular system.

Weaknesses:

Multiple phosphorylated tyrosines in IKK2 were apparently identified by mass spectrometric analyses. LC-MS/MS spectra are presented, but fragments supporting phospho-Y188 and Y325 are difficult to distinguish from noise. It is common to find non-physiological post-translational modifications in over-expressed proteins from recombinant sources. Are these IKK2 phosphotyrosines evident by MS in IKK2 immunoprecipitated from TNFa-stimulated cells? Identifying IKK2 phosphotyrosine sites from cells would be especially helpful in supporting the proposed model.

Reviewer #3 (Public review):

Summary:

The authors investigate the kinase activity of IKK2, a crucial regulator of inflammatory cell signaling. They describe a novel tyrosine kinase activity of this well-studied enzyme and a highly unusual phosphotransfer from phosphorylated IKK2 onto substrate proteins in the absence of ATP as a substrate.

Strengths:

The authors provide an extensive biochemical characterization of the processes with recombinant protein, western blot, autoradiography, protein engineering and provide MS data now.

Weaknesses:

The identity and purity of the used proteins has improved in the revised work. Since the findings are so unexpected and potentially of wide-reaching interest - this is important. Similar specific detection of phospho-Ser/Thr vs phospho-Tyr relies largely on antibodies which can have varying degrees of specificity. Using multiple antibodies and MS improves the quality of the data.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

The model of phosphotransfer from Y169 IKK to S32 IkBa is compelling and an important new contribution to the field. In fact, this model will not be without controversy, and publishing the work will catalyze follow-up studies for this kinase and others as well. As such, I am supportive of this paper, though I do also suggest some shortening and modification.

We appreciate the reviewers candid response on the difficulty of this study and the requirement of follow-up studies to confirm a direct transfer of the phosphate. We also have edited the manuscript to make it shorter.

Generally, the paper is well written, but several figures should be quantified, and experimental reproducibility is not always clear. The first 4 figures are slow-going and could be condensed to show the key points, so that the reader gets to Figures 6 and 7 which contain the "meat" of the paper.

We have indicated the experimental reproducibility in the methodology section against each assay. We have shortened the manuscript corresponding to sections describing figures 1-4. However, when we talked to some of our colleagues whose expertise do not align with kinases and IKK, we realized that some description were necessary to introduce them to the next figures. Additionally, we added Fig. S6 indicating that the radiolabelled phospho-IKK2 Y169F is unable to transfer its own phosphate group(s) to the substrate IkBa.

Reviewer #2 (Public Review):

Phosphorylation of IκBα is observed after ATP removal, although there are ambiguous requirements for ADP.

We agree with the reviewer that this observation is puzzling. We hypothesize that ADP is simultaneously regulating the transfer process likely through binding to the active site.

It seems that the analysis hinges on the fidelity of pan-specific phosphotyrosine antibodies.

We agree with the reviewer. To bolster our conclusion, we used antibodies from two different sources. These were Monoclonal mouse anti-Phospho-Tyrosine (catalogue number: 610000) was from BD Biosciences or from EMD Millipore (catalogue no. 05-321X).

The analysis often returns to the notion that tyrosine phosphorylation(s) (and critical active site Lys44) dictate IKK2 substrate specificity, but evidence for this seems diffuse and indirect. This is an especially difficult claim to make with in vitro assays, omitting the context of other cellular specificity determinants (e.g., localization, scaffolding, phosphatases).

We agree with the concerns that the specificity could be dependent on other cellular specificity determinants and toned down our claims where necessary. However, we would like to point out that the specificity of IKK2 towards S32 and S36 of IkBa in cells in response to specific stimuli is well-established. It is also well-established that its non-catalytic scaffolding partner NEMO is critical in selectively bringing IkBa to IKK from a large pool of proteins. The exact mechanism of how IKK2 choose the two serines amongst many others in the substrate is not clear.

Multiple phosphorylated tyrosines in IKK2 were apparently identified by mass spectrometric analyses, but the data and methods are not described. It is common to find non-physiological post-translational modifications in over-expressed proteins from recombinant sources. Are these IKK2 phosphotyrosines evident by MS in IKK2 immunoprecipitated from TNFa-stimulated cells? Identifying IKK2 phosphotyrosine sites from cells would be especially helpful in supporting the proposed model.

Mass spectrometric data for identification of phosphotyrosines from purified IKK2 is now incorporated (Figure S3A). Although we have not analyzed IKK2 from TNF-a treated cells in this study, a different study of phospho-status of cellular IKK2 indicated tyrosine phosphorylation (Meyer et al 2013).

Reviewer #3 (Public Review):

The identity and purity of the used proteins is not clear. Since the findings are so unexpected and potentially of wide-reaching interest - this is a weakness. Similar specific detection of phospho-Ser/Thr vs phospho-Tyr relies largely on antibodies which can have varying degrees of specificity.

We followed a stringent purification protocol of several steps (optimized for the successful crystallization of the IKK2) that removed most impurities (PMID: 23776406, PMID: 39227404). The samples analysed with ESI MS did not show any significant contaminating kinase from the Sf9 cells.

Sequence specific phospho-antibodies used in this study are very well characterized and have been used in the field for years (Basak et al 2007, PMID: 17254973). We agree on the reviewer’s concerns on the pan-specific phospho-antibodies. Since phospho-tyrosine detection is the crucial aspect of this study, we minimized such bias by using pan-specific phosphotyrosine antibodies from two independent sources.

Reviewer #1 (Recommendations For The Authors):

I understand that Figure 3 shows that K44M abolishes both S32/26 phosphorylation and tyrosine phosphorylation, but not PEST region phosphorylation. This suggests that autophosphorylation is reflective of its known specific biological role in signal transduction. But I do not understand why "these results strongly suggest that IKK2-autophosphorylation is critical for its substrate specificity". That statement would be supported by a mutant that no longer autophosphorylates, and as a result shows a loss of substrate specificity, i.e. phosphorylates non-specific residues more strongly. Is that the case? Maybe Darwech et al 2010 or Meyer et al 2013 showed this.

Later figures seem to address this point, so maybe this conclusion should be stated later in the paper.

We have now clarified this in the manuscript and moved the comment to the next section. We have consolidated the results in Figure 3 and 4 in the previous version into a single figure in Figure. The text has also been modified accordingly.

Page 10: mentions DFG+1 without a proper introduction. The Chen et al 2014 paper appears to inform the author's interest in Y169 phosphorylation, or is it just an additional interesting finding? Does this publication belong in the Introduction or the Discussion?

The position of Y169 at the DFG+1 was intriguing and the 2014 article Chen et al further bolstered our interest in this residue to be investigated. We think this publication is important in both sections. 

To understand the significance of Figure 4D, we need a WT IKK2 control: or is there prior literature to cite? This is relevant to the conclusion that Y169 phosphorylation is particularly important for S32 phosphorylation.

We have now added a new supplementary figure where activities of WT and Y169F IKK2 towards WT and S32/S36 mutants are compared (Figure S3F). At a similar concentration, the activity of WT-IKK2 is many fold higher than that of YtoF mutants (Fig. 4C). The experiments were performed simultaneously, although samples were loaded on different gels but otherwise processed in a similar way. The corresponding data is now included in the manuscript as Figure S3F.

The cold ATP quenching experiment is nice for testing the model that Y169 functions as a phospho sink that allows for a transfer reaction. However, there is only a single timepoint and condition, which does not allow for a quantitative analysis. Furthermore, a positive control would make this experiment more compelling, and Y169F mutant should show that cold ATP quenching reduces the phosphorylation of IkBa.

We thank the reviewer for appreciating our experimental design, and pointing out the concerns. We kept the ATP-time point as the maximum of the non-competition experiment. Also, we took 50mM ATP to compare its competition with highest concentration of ADP used. The idea behind using the maximum time and ATP (comparable to ADP) was to capture the effect of competitive-effect of ATP, if any, that would be maximal in the given assay condition in comparison with the phospho-transfer set up in absence of cold ATP. We agree that finer ranges of ATP concentration and time points would have enabled more quantitative analyses. We have now included data where different time intervals are tested (Figure S5D).

Why is the EE mutant recognized by anti-phospho-serine antibodies? In Figure 2F.

We anticipate Serine residues besides those in the activation loop to be phosphorylated when IKK2 is overexpressed and purified from the Sf9 cells. Since Glu (E) mimics phospho-Ser, the said antibody cross reacts with the IKK2-EE that mimics IKK2 phosphorylated at Ser177 and 181.

Figure 7B is clear, but 7C does not add much.

We have now removed the Fig. 7C in the current version. Figure 7 is now renumbered as Figure 6 that does not contain the said cartoon.  

Reviewer #2 (Recommendations For The Authors):

Regarding the specificity arguments (see above in public review), the authors note that NEMO is very important in IKK specificity, and - if I'm understanding correctly - most of these assays were performed without NEMO. Would the IKK2-NEMO complex change these conclusions?

NEMO is a scaffolding protein whose action goes beyond the activation of the IKK-complex. In cells, NEMO brings IkBa from a pool of thousands of proteins to its bonafide kinase when the cells encounter specific signals. In other words, NEMO channels IKK-activity towards its bonafide substrate IkBa at that moment. Though direct proof is lacking, it is likely that NEMO present IkBa in the correct pose to IKK such that the S32/S36 region of IkBa is poised for phosphorylation. The proposed mechanism in the current study further ensures the specificity and fidelity of that phosphorylation event. We believe this mechanism will be preserved in the IKK-NEMO complex unless proven otherwise. As shown below, IKK2 undergoes tyrosine autophosphorylation in presence of NEMO.

Author response image 1.

The work primarily focuses on Y169 as a candidate target for IKK autophosphorylation. This seems reasonable given the proximity to the ATP gamma phosphate. However, Y188F more potently disrupted IκBα phosphorylation. The authors note that this could be due to folding perturbations, but this caveat would also apply to Y169F. A test for global fold perturbations for both Tyr mutants would be helpful.

Y188 is conserved in S/T kinases and that in PKA (Y204) has been studied extensively using structural, biochemical and biophysical tools. It was found in case of PKA that Y204 participates in packing of the hydrophobic core of the large lobe. Disruption of this core structure by mutation allosterically affect the activity of the kinase. We also observed similar engagement of Y188 in IKK2’s large lobe, and speculated folding perturbations in analogy with the experimental evidence observed in PKA. What we meant was mutation of Y188 would allosterically affect the kinase activity. Y169 on the other hand is unique at that position, an no experimental evidence on the effect of phospho-ablative mutation of this residue exist in the literature. Hence, we refrained from speculating its effect on the folding or conformational allostery, however, such a possibility cannot be ruled out. 

I struggled to follow the rationalization of the results of Figure 4D, the series of phosphorylation tests of Y169F against IκBα with combinations of phosphoablative or phosphomimetic variants at Ser32 and Ser36. This experiment is hard to interpret without a direct comparison to WT IKK2.

We agree with the reviewer’s concerns. Through this experiment we wanted to inform about the importance of Tyr-phosphorylation of IKK2 in phosphorylating S32 of IκBα which is of vital importance in NF-kB signaling. We have now provided a comparison with WT-IKK2 in the supplementary Figure S3F. We hope this will help bring more clarity to the issue.

MD simulations were performed to compare structures of unphosphorylated vs. Ser-phosphorylated (p-IKK2) vs. Ser+Tyr-phosphorylated (P-IKK2) forms of IKK2. These simulations were performed without ATP bound, and then a representative pose was subject to ADP or ATP docking. The authors note distortions in the simulated P-IKK2 kinase fold and clashes with ATP docking. Given the high cellular concentration of ATP, it seems more logical to approach the MD with the assumption of nucleotide availability. Most kinase domains are highly dynamic in the absence of substrate. Is it possible that the P-IKK2 poses are a result of simulation in a non-physiological absence of bound ATP? Ultimately, this MD observation is linked to the proposed model where ADP-binding is required for efficient phospho-relay to IκBα. Therefore, this observation warrants scrutiny. Perhaps the authors could follow up with binding experiments to directly test whether P-IKK2 binds ADP and fails to bind ATP.

We thank that reviewer for bringing up this issue. This is an important issue and we must agree that we don’t fully understand it yet. We took more rigorous approach this time where we used three docking programs: ATP and ADP were docked to the kinase structures using LeDock and GOLD followed by rescoring with AutoDock Vina. We found that ATP is highly unfavourable to P-IKK2 compared to ADP. To further address these issues, we performed detailed MM-PBSA (Molecular Mechanics Poisson-Boltzmann Surface Area) analyses after MD-simulation to estimate binding free energies and affinities of ADP and ATP for each of the three differently phosphorylated states of IKK2. These analyses (Figure S4 E and F) clearly indicate that phosphorylated IKK2 have much higher preference for ADP over ATP. However, it does not negate ATP-binding by P-IKK2 in a different pose that may not support kinase activity.

We could not perform any binding experiment because of the following reason. We incubated FL IKK2 WT with or without cold ATP for 30mins, and then incubated these samples with ³²P-ATP and analysed the samples by autoradiography after resolving them on a 10% SDS-PAGE. We found that even after pre-incubation of the kinase with excess cold ATP it still underwent autophosphorylation when radioactive ATP was added as shown below. This prevented us from doing direct binding experiment with ATP as it would not represent true binding event. We also noticed that after removal of bulk ATP post autophosphorylation, phosphorylated IKK2 is capable of further autophosphorylation when freshly incubated with ATP. We have not been able to come up with a condition that would only account for binding of ATP and not hydrolysis. 

Author response image 2.

The authors could comment on whether robust phosphorylation of NEMO was expected (Figure 1D). On a related note, why is NEMO a single band in the 1D left panel and double bands on the right?

No, we did not expect robust phosphorylation of NEMO. However, robust phosphorylation of NEMO is observed only in the absence of IκBα. In presence of IκBα, phosphorylation of NEMO goes down drastically. These were two different preparations of NEMO. When TEV-digestion to remove His-tag is incomplete it gives two bands as the tagged and untagged versions cannot be separated in size exclusion chromatography which is the final step.

Page 14, line 360. "...observed phosphorylation of tyrosine residue(s) only upon fresh ATP-treatment..." I'm not sure I understand the wording here (or the relevance of the citation). Is this a comment on unreported data demonstrating the rapid hydrolysis of the putative phosphotyrosine(s)? If so, that would be helpful to clarify and report in the supporting information.

In our X-ray crystallographic studies with phosphorylated IKK2 we failed to observe any density of phosphate moiety. Furthermore, this IKK2 showed further autophosphorylation when incubated with fresh ATP. These two observations lead us to believe that some of the autophosphorylation are transient in nature. However, quantitative kinetic analyses of this dephosphorylation have not been performed.

Figure S3 middle panel: The PKA substrate overlaid on the IKK2 seems sterically implausible for protein substrate docking. Is that just a consequence of the viewing angle? On a related note, Figure S3 may be mislabeled as S4 in the main text).

It is a consequence of the viewing angle. Also, we apologize for this inadvertent mislabelling. It has been corrected in the current version.

Reviewer #3 (Recommendations For The Authors):

The detection of phosphorylated amino acids relies largely on antibodies which can have a varying degree of specificity. An alternative detection mode of the phospho-amino acids for example by MS would strengthen the evidence.

We agree with the concern of specificity bias of antibodies. We tried to minimize such bias by using two different p-Tyr antibodies as noted previously and also in the methodology section. We were also able to detect phospho-tyrosine residues by MS/MS analyses, representative spectra are now added (Figure S3A).

IKK2 purity - protocol states "desired purity". What was the actual purity and how was it checked? MS would be useful to check for the presence of other kinases.

Purity of the recombinantly purified IKK2s are routinely checked by silver staining. A representative silver stained SDS-PAGE is shown (Figure S1C). It may be noted that, there’s a direct correlation of expression level and solubility, and hence purification yield and quality with the activity of the kinase. Active IKK2s express at much higher level and yields cleaner prep. In our experience, inactive IKKs like K44M give rise to poor yield and purity. We analysed K44M by LC MS/MS to identify other proteins present in the sample. We did not find any significant contaminant kinase the sample (Figure S1D). The MS/MS result is attached.

Figure 1C&D: where are the Mw markers? What is the size of the band? What is the MS evidence for tyrosine phosphorylation?

We have now indicated MW marker positions on these figures.

MS/MS scan data for the two peptides containing pTyr169 and pTyr188 are shown separately (Figure S3A).

Figure 2F: Why is fresh ATP necessary? Why was Tyr not already phosphorylated? The kinetics of this process appear to be unusual when the reaction runs to completion within 5 minutes ?

As stated earlier, we believe some of the autophosphorylation are transient in nature. We think the Tyr-phosphorylation are lost due to the action of cellular phosphatases. We agree with the concern of the reviewer that, the reaction appears to reach completion within 5 minutes in Fig 2F. We believe it is probably due to the fact that the amount of kinase used in this study exceeds the linear portion of the dynamic range of the antibody used. Lower concentration of the kinase do show that reaction does not reach completion until 60mins as shown in Fig. 2A.

Figure 3: Can the authors exclude contamination with a Tyr kinase in the IKK2-K44M prep? The LC/MS/MS data should be included.

We have reanalysed the sample on orbitrap to check if there’s any Tyr-kinase or any other kinase contamination. We used Spodoptera frugiperda proteome available on the Uniprot website for this analysis. These analyses confirmed that there’s no significant kinase contaminant present in the fraction (Figure S1D).

What is the specificity of IKK-2 Inhibitor VII? Could it inhibit a contaminant kinase?

This inhibitor is highly potent against IKK2 and the IKK-complex, and to a lesser extent to IKK1. No literature is available on its activity on other kinases. In an unrelated study, this compound was used alongside MAPK inhibitor SB202190 wherein they observed completely different outcomes of these two inhibitors (Matou-Nasri S, Najdi M, AlSaud NA, Alhaidan Y, Al-Eidi H, Alatar G, et al. (2022) Blockade of p38 MAPK overcomes AML stem cell line KG1a resistance to 5-Fluorouridine and the impact on miRNA profiling. PLoS ONE 17(5):e0267855. https://doi.org/10.1371/journal.pone.0267855). This study indirectly proves that IKK inhibitor VII does not fiddle with the MAPK pathways. We have not found any literature on the non-specific activity of this inhibitor.

Figure 6B: the band corresponding to "p-IkBa" appears to be similar in the presence of ADP (lanes 4-7) or in the absence of ADP but the presence of ATP (lane 8).

Radioactive p-IκBα level is more when ADP is added than in absence of ADP. In presence of cold ATP, radioactive p-IκBα level remains unchanged. This result strongly indicate that the addition of phosphate group to IκBα happens directly from the radioactively labelled kinase that is not competed out by the cold ATP.

eLife Assessment

Understanding how the divisome is assembled in Chlamydia trachomatis, a bacterial pathogen, is crucial since this bacterium has a non-canonical cell wall and lacks the master regulator of cell division, FtsZ. This important study shows that a DNA translocase, FtsK, is an early and essential component of the Chlamydia trachomatis divisome. The evidence presented is convincing, leveraging the elegant use of genetics and fluorescence microscopy. As this role of FtsK is distinct relative to most other bacteria, these findings should be of significant interest to bacterial cell biologists.

Reviewer #1 (Public review):

Summary:

In this work, Harpring et al. investigated divisome assembly in Chlamydia trachomatis serovar L2 (Ct), an obligate intracellular bacterium that lacks FtsZ, the canonical master regulator of bacterial cell division. They find that divisome assembly is initiated by the protein FtsK in Ct by showing that it forms discrete foci at the septum and future division sites. Additionally, knocking down ftsK prevents divisome assembly and inhibits cell division, further supporting their hypothesis that FtsK regulates divisome assembly. Finally, they show that MreB is one of the last chlamydial divisome proteins to arrive at the site of division and is necessary for the formation of septal peptidoglycan rings but does not act as a scaffold for division assembly as previously proposed.

Strengths:

The authors use microscopy to clearly show that FtsK forms foci both at the septum as well as at the base of the progenitor cell where the next septum will form. They also show that the Ct proteins PBP2, PBP3, MreC, and MreB localize to these same sites suggesting they are involved in the divisome complex.

Using CRISPRi the authors knockdown ftsK and find that most cells are no longer able to divide and that PBP2 and PBP3 no longer localized to sites of division suggesting that FtsK is responsible for initiating divisome assembly. They also performed a knockdown of pbp2 using the same approach and found that this also mostly inhibited cell division. Additionally, FtsK was still able to localize in this strain however PBP3 did not suggest that FtsK acts upstream of PBP2 in the divisome assembly process while PBP2 is responsible for the localization of PBP3.

The authors also find that performing a knockdown of ftsK also prevents new PG synthesis further supporting the idea that FtsK regulates divisome assembly. They also find that inhibiting MreB filament formation using A22 results in diffuse PG, suggesting that MreB filament formation is necessary for proper PG synthesis to drive cell division.

Overall the authors propose a new hypothesis for divisome assembly in an organism that lacks FtsZ and use a combination of microscopy and genetics to support their model that is rigorous and convincing. The finding that FtsK, rather than a cytoskeletal or "scaffolding" protein is the first division protein to localize to the incipient division site is unexpected and opens up a host of questions about its regulation. The findings will progress our understanding of how cell division is accomplished in bacteria with non-canonical cell wall structure and/or that lack FtsZ.

Weaknesses:

No major weaknesses were noted in the data supporting the main conclusions.

Reviewer #2 (Public review):

Summary:

Chlamydial cell division is a peculiar event, whose mechanism was mysterious for many years. C. trachomatis division was shown to be polar and involve a minimal divisome machinery composed of both homologues of divisome and elongasome components, in absence of an homologue of the classical division organizer FtsZ. In this paper, Harpring et al., show that FtsK is required at an early stage of the chlamydial divisome formation.

Strengths:

The manuscript is well-written and the results are convincing. Quantification of divisome component localization is well performed, number of replicas and number of cells assessed are sufficient to get convincing data. The use of a CRISPRi approach to knock down some divisome components is an asset and allows a mechanistic understanding of the hierarchy of divisome components

Weaknesses

Despite advances in the understanding of the importance of FtsK for chlamydial division, this manuscript does not show by which mechanism FtsK specifically localizes at the division site and how it mediates recruitment of other divisome members. Moreover, the potential link with DNA partitioning is not addressed.

Reviewer #3 (Public review):

Summary:

The obligate intracellular bacterium Chlamydia trachomatis (Ct) divides by binary fission. It lacks FtsZ, but still has many other proteins that regulate synthesis of septal peptidoglycan, including FtsW and FtsI (PBP3) as well as divisome proteins that recruit and activate them, such as FtsK and FtsQLB. Interestingly, MreB is also required for division of Ct cells, perhaps by polymerizing to form an FtsZ-like scaffold. Here, Harpring et al. show that MreB does not act early in division and instead is recruited to a protein complex that includes FtsK and PBP2/PBP3. This indicates that Ct cell division is organized by a chimera between conserved divisome and elongasome proteins. Their work also shows convincingly that FtsK is the earliest known step of divisome activity, potentially nucleating the divisome as a single protein complex at the future division site. This is reminiscent of the activity of FtsZ, yet fundamentally different.

Strengths:

The study is very well written and presented, and the data are convincing and rigorous. The data underlying the proposed localization dependency order of the various proteins for cell division is well justified by several different approaches using small molecule inhibitors, knockdowns, and fluorescent protein fusions. The proposed dependency pathway of divisome assembly is consistent with the data and with a novel mechanism for MreB in septum synthesis in Ct.

Weaknesses:

The authors have addressed the weaknesses brought up in my previous review.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public review):

Summary:

In this work, Harpring et al. investigated divisome assembly in Chlamydia trachomatis serovar L2 (Ct), an obligate intracellular bacterium that lacks FtsZ, the canonical master regulator of bacterial cell division. They find that divisome assembly is initiated by the protein FtsK in Ct by showing that it forms discrete foci at the septum and future division sites. Additionally, knocking down ftsK prevents divisome assembly and inhibits cell division, further supporting their hypothesis that FtsK regulates divisome assembly. Finally, they show that MreB is one of the last chlamydial divisome proteins to arrive at the site of division and is necessary for the formation of septal peptidoglycan rings but does not act as a scaffold for division assembly as previously proposed.

Strengths:

The authors use microscopy to clearly show that FtsK forms foci both at the septum as well as at the base of the progenitor cell where the next septum will form. They also show that the Ct proteins PBP2, PBP3, MreC, and MreB localize to these same sites suggesting they are involved in the divisome complex.

Using CRISPRi the authors knock down ftsK and find that most cells are no longer able to divide and that PBP2 and PBP3 no longer localized to sites of division suggesting that FtsK is responsible for initiating divisome assembly. They also performed a knockdown of pbp2 using the same approach and found that this also mostly inhibited cell division. Additionally, FtsK was still able to localize in this strain, however PBP3 did not, suggesting that FtsK acts upstream of PBP2 in the divisome assembly process while PBP2 is responsible for the localization of PBP3.

The authors also find that performing a knockdown of ftsK also prevents new PG synthesis further supporting the idea that FtsK regulates divisome assembly. They also find that inhibiting MreB filament formation using A22 results in diffuse PG, suggesting that MreB filament formation is necessary for proper PG synthesis to drive cell division.

Overall the authors propose a new hypothesis for divisome assembly in an organism that lacks FtsZ and use a combination of microscopy and genetics to support their model that is rigorous and convincing. The finding that FtsK, rather than a cytoskeletal or "scaffolding" protein is the first division protein to localize to the incipient division site is unexpected and opens up a host of questions about its regulation. The findings will progress our understanding of how cell division is accomplished in bacteria with non-canonical cell wall structure and/or that lack FtsZ.

Weaknesses:

No major weaknesses were noted in the data supporting the main conclusions. However, there was a claim of novelty in showing that multiple divisome complexes can drive cell wall synthesis simultaneously that was not well-supported (i.e. this has been shown previously in other organisms). In addition, there were minor weaknesses in data presentation that do not substantially impact interpretation (e.g. presenting the number of cells rather than the percentage of the population when quantifying phenotypes and showing partial western blots instead of total western blots).

We agree with the weaknesses identified by the reviewer. We removed the statements in the Results and Discussion that multiple independent divisome complexes can simultaneously direct PG synthesis. We presented the data in Figs. 3-5 as % of the cells in the population, and complete western blots are shown in Supp. Fig. S1.

Reviewer #2 (Public review):

Summary:

Chlamydial cell division is a peculiar event, whose mechanism was mysterious for many years. C. trachomatis division was shown to be polar and involve a minimal divisome machinery composed of both homologues of divisome and elongasome components, in the absence of an homologue of the classical division organizer FtsZ. In this paper, Harpring et al., show that FtsK is required at an early stage of the chlamydial divisome formation.

Strengths:

The manuscript is well-written and the results are convincing. Quantification of divisome component localization is well performed, number of replicas and number of cells assessed are sufficient to get convincing data. The use of a CRISPRi approach to knock down some divisome components is an asset and allows a mechanistic understanding of the hierarchy of divisome components.

Weaknesses:

The authors did not analyse the role of all potential chlamydial divisome components and did not show how FtsK may initiate the positioning of the divisome. Their conclusion that FtsK initiates the assembly of the divisome is an overinterpretation and is not backed by the data. However, data show convincingly that FtsK, if perhaps not the initiator of chlamydial division, is definitely an early and essential component of the chlamydial divisome.

The following statement has been included in the Discussion (pg. 16 of the revised manuscript)  “Although we focused our study on a subset of the divisome and elongasome proteins that Chlamydia expresses (bolded in Fig. 6G), our results support our conclusion that chlamydial budding is dependent upon a hybrid divisome complex and that FtsK is required for the assembly of this hybrid divisome. At this time, we cannot rule out that other proteins act upstream of FtsK to initiate divisome assembly in this obligate intracellular bacterial pathogen.”

We will soon be submitting another manuscript that addresses how FtsK specifies the site of divisome assembly. This work is too extensive to be included in this manuscript.

Reviewer #3 (Public review):

Summary:

The obligate intracellular bacterium Chlamydia trachomatis (Ct) divides by binary fission. It lacks FtsZ, but still has many other proteins that regulate the synthesis of septal peptidoglycan, including FtsW and FtsI (PBP3) as well as divisome proteins that recruit and activate them, such as FtsK and FtsQLB. Interestingly, MreB is also required for the division of Ct cells, perhaps by polymerizing to form an FtsZ-like scaffold. Here, Harpring et al. show that MreB does not act early in division and instead is recruited to a protein complex that includes FtsK and PBP2/PBP3. This indicates that Ct cell division is organized by a chimera between conserved divisome and elongasome proteins. Their work also shows convincingly that FtsK is the earliest known step of divisome activity, potentially nucleating the divisome as a single protein complex at the future division site. This is reminiscent of the activity of FtsZ, yet fundamentally different.

Strengths:

The study is very well written and presented, and the data are convincing and rigorous. The data underlying the proposed localization dependency order of the various proteins for cell division is well justified by several different approaches using small molecule inhibitors, knockdowns, and fluorescent protein fusions. The proposed dependency pathway of divisome assembly is consistent with the data and with a novel mechanism for MreB in septum synthesis in Ct.

Weaknesses:

The paper could be improved by including more information about FtsK, the "focus" of this study. For example, if FtsK really is the FtsZ-like nucleator of the Ct divisome, how is the Ct FtsK different sequence-wise or structurally from FtsK of, e.g. E. coli? Is the N-terminal part of FtsK sufficient for cell division in Ct like it is in E. coli, or is the DNA translocase also involved in focus formation or localization? Addressing those questions would put the proposed initiator role of FtsK in Ct in a better context and make the conclusions more attractive to a wider readership.

We will be submitting another manuscript soon that details the conserved domain organization of FtsK from different bacteria, and the role of the various domains of chlamydial FtsK (including the N-terminus and the C-terminal translocase domain) in directing its localization in dividing Chlamydia. We have added text to the discussion (pg. 16 of the revised manuscript) that describes the sequence homology of chlamydial FtsK to FtsK from E. coli.

Another weakness is that the title of the paper implies that FtsK alone initiates divisome assembly. However, the data indicate only that FtsK is important at an early stage of divisome assembly, not that it is THE initiator. I suggest modifying the title to account for this--perhaps "FtsK is required to initiate....".

We agree with the reviewer and modified the title to “FtsK is Critical for the Assembly of the Unique Divisome Complex of the FtsZ-less Chlamydia trachomatis”. We have also modified the text throughout to indicate that FtsK is required for the assembly of the hybrid divisome of Chlamydia. 

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Suggestions for improvement (mostly minor):

(1) For several of the graphs, the authors plot the number of cells with a given phenotype on the y-axis, but then describe percentages of cells in the text. It would make it clearer if all the graphs had the percentage of cells on the y-axis instead.

We have modified the figures to indicate the percentage of cells on the y-axis with a given phenotype.

(2) In Figures 3, 4, and 5 the authors show separate graphs for plus/minus drug or inducer. These should be on the same graph as they are directly comparing these two different conditions. Having them on separate graphs makes it less clear whether these differences are significant between the two conditions

We modified Fig. 4 to show +/- inducer in ftsk and pbp2 knockdown strains in the same graph.  Regarding Figures 3 and 5, we believe the figures in the original submission effectively demonstrate the +/- drug conditions, so these figures remain unchanged in the revised manuscript.

(3) In Figure 2 the authors show microscopy of the colocalization of FtsK with several other divisome proteins from Ct. Quantification of the colocalization of FtsK with these other proteins would provide a more holistic understanding of their colocalization and help further support their argument that FtsK initiates the assembly of the divisome.

Supp. Fig. S4A of the revised manuscript contains images showing the colocalization of FtsK with the fusions at the septum and the base of dividing cells, and the colocalization of FtsK with the fusions that are only at the base of dividing cells. Supp. Fig. S4B quantified the percentage of dividing cells where FtsK overlaps the localization of each of the fusions at the septum, at the septum and the base, and at the base alone.

(4) In Figure 6 the authors mention that the PG ring was at a slight angle relative to the MOMP-stained septum. What is the significance of this? The authors mention it several times but do not explain its relevance to divisome assembly. It is not really evident in the images presented.

We mention in the discussion pgs. 17-18 of the revised manuscript that “The relevance of the angled orientation of PG and MreC rings relative to the MOMP-stained septum in division intermediates is unclear. However, it appears to be a conserved feature of the cell division process and may arise because the divisome proteins are often positioned slightly above or below the plane of the MOMP-stained septum. The positioning of divisome proteins above or below the septum is indicated in Figs. 1 and 2.

We included cartoons in Fig. 6C of the revised manuscript to assist the reader in visualizing the angled orientation of the PG ring relative to the MOMP-stained septum.

(5) In line 270 the authors claim that "these are the first data in any system to suggest that septal PG synthesis/modification is simultaneously directed by multiple independent divisome complexes." However, their experiments do not demonstrate that multiple divisome complexes are active at the same time. They show that multiple foci of FtsK etc. are present at sites where PG synthesis has occurred, but that does not necessarily mean that each focus/complex was actively synthesizing PG at the same time. Moreover, similar approaches were used to support a claim that septal PG synthesis is directed by multiple discrete divisome complexes previously (e.g. in Figure 1 of Bisson-Filho et al. 2017 (PMID: 28209898) in Bacillus subtilis and in Perez et al 2021 (PMID: 33269494) in Streptococcus pneumoniae). This claim is not central to the main conclusions of the study and could just be removed.

This statement has been removed from the Results and the Discussion.

(6) In Figure 6B the authors see three distinct FtsK foci. Why is this the only place in the manuscript where they see three foci? They mentioned previously that they saw foci at the septum and at the base of the progenitor mother cell, but why are there three foci here?

The vast majority of dividing cells displayed one foci at the septum and/or the base.  Representative images were chosen that reflected the localization profiles observed in the majority of cells. While we observed cells with  multiple foci, as shown in Figure 6C, these cells were relatively rare   (~2% of cells for all the divisome proteins in 3 independent experiments).  Since  the number of cells with multiple foci were relatively rare, we chose to group these cells with the cells that had single foci at the septum, the septum and base, or base alone categories in the quantification shown in Fig. 2C. This is stated in the legend of Fig. 2 of the revised manuscript.

(7) The Discussion section is lacking a couple of things that would put the data in a broader context. Can the authors speculate on how FtsK knows how to find the division site? I.e. what might be upstream of FtsK localization? Additionally, the authors do not talk about the FtsK sequence or domains at any point in the paper. Does Ct FtsK have a similar sequence/structure to FtsKs from other bacteria? Are there any differences in sequence/structure that might tell us about its function in Ct?

We will be submitting another manuscript soon that examines how the site of assembly of the divisome is defined in dividing Chlamydia. This manuscript will also define the localization of the different sub-domains of chlamydial FtsK during cell division.  For this manuscript, we added a paragraph in the Discussion (pg. 16 of the revised manuscript) that states the domain organization is conserved in FtsK proteins from different bacteria. This paragraph includes information regarding the % sequence identity of the C-terminus and the N-terminus of chlamydial FtsK when compared to E. coli FtsK.

(8) For Supplementary Figure S1B-C. The authors should show the full blots rather than just the single band of the protein of interest to show that the antibodies are specific. Additionally, the authors should include a loading control to show that they loaded the same amount of protein for each sample.

We have included the full blots in Supp. Fig. S1 of the revised manuscript. We do not see the need for including a loading control for these blots because we are not making arguments about the relative level of the proteins that were assayed. We only use the blots to show that the fusion proteins are primarily a single species of the predicted molecular mass.

(9) In Supplementary Figure S4A the authors use RT-qPCR to measure ftsK and pbp2 transcript levels. Since they have antibodies against these proteins, they should also include Western blots to show that the proteins are not being produced when targeted using CRISPRi.

We have included data in Supp. Fig. S5E of the resubmission that indicates foci of FtsK and PBP2 could not be detected following the knockdown of ftsk and pbp2. We feel that these data support our conclusion that the induced expression of dCas12 in the the ftsk and pbp2 knockdown strains results in the downregulation of the endogenous FtsK and PBP2 polypeptides.

(10) In lines 261-262 the authors say that "PG organization was the same or differed at the septum." What is the PG organization being compared to? Same or different from what?

We agree with the reviewer that the text in lines 261-262 in the original submission was confusing.  The text has been modified.

(11) Lines 201-215 the authors refer to Supplementary Figure S3 throughout this section, but they should refer to Supplementary Figure S4.

This has been corrected.

Reviewer #2 (Recommendations for the authors):

I am not convinced that this paper shows that FtsK initiates the assembly of the divisome since the authors did not analyse the role and localization of all other chlamydial divisome components. Out of the ten homologues of divisome and elongasome components encoded by C. trachomatis genome, only five are investigated in this study. There is no explanation about how these five were chosen.

We state on pg. 16 of the revised manuscript that “Although we focused our study on a subset of the divisome and elongasome proteins that Chlamydia expresses (bolded in Fig. 6G), our results support our conclusion that chlamydial budding is dependent upon a hybrid divisome complex and that FtsK is required for the assembly of this hybrid divisome. At this time, we cannot rule out that other proteins act upstream of FtsK to initiate divisome assembly in this obligate intracellular bacterial pathogen.

Results convincingly indicate that FtsK is an early divisome component, but proofs are lacking to indicate that it initiates the divisome formation. Indeed, the authors do not show how FtsK would be the first protein to selectively accumulate at a given location to initiate the divisome formation. For this reason, the model they propose at the end of their study is not backed by sufficient data, to my opinion.

We agree with the reviewer that our data does not show that FtsK initiates divisome assembly. The title of the manuscript has been modified to “FtsK is Critical for the Assembly of the Unique Divisome Complex of the FtsZ-less Chlamydia trachomatis” and the text throughout has been modified to indicate that FtsK is the first protein we assayed that associates with nascent divisomes at the base of dividing cells. We will soon be submitting another manuscript that details how FtsK is recruited to a specific site to initiate nascent divisome assembly, This work is too extensive to be included in this manuscript.

There are also discrepancies in the number of cells analysed to quantify the localization of divisome components, ranging from 50 to 250 cells. The authors could better explain why there are such variations.

There were differences in the number of cells analyzed in the various experiments, but in every instance the effect of inhibitors (A22 and mecillinam) or ftsk and pbp2 knockdown on divisome assembly was statistically significant.

There are a few mistakes in the text regarding figure numbering (Figure S4 is mentioned as S3 in the text). Figures 5B and D are not specifically cited.

These mistakes have been corrected in the revised manuscript.

Line 261-262: the sentence starting "Our imaging analysis.." is not clear to me.

We agree with the reviewer that the text in lines 261-262 was confusing.  The text has been modified (pg. 14 of the revised manuscript).

Line 270-271: there are insufficient proofs to say that there are multiple independent divisome complexes. This is in my opinion an overinterpretation of the data, since there is no proof that these complexes are independent.

This statement has been removed from the text.

A few details are lacking in the figure legends:

Figure 2C: when was the expression of the different mCherry and 6xHis constructs induced?

The onset and length of the induction of the fusions have been included in the legend of Fig. 2.

Bars are sometimes mentioned as uM and should be um. Bars sizes, number of replicates, and/or meaning of the error bars are lacking in legends of Figures S2, S3, and S4

This has been corrected in the revised manuscript.

The consistency of Figures could be improved between Figures 3A, 4A, B, and 5A. The results of treated cells could be always shown as dark grey. It would help the reader.

We have used consistent coloring in Figs. 3-5 to indicate the treated cells.

Reviewer #3 (Recommendations for the authors):

(1) Lines 113-118: do Ct cells increase in size as they get closer to starting division? If so, could a pseudo-time course (demograph) be done to bolster the evidence that the base foci formed mainly in predivisional cells and not newborn cells? This evidence might be more convincing than the data in Figures 1F and G.

Chlamydial cells in the population were heterogeneous in size at the timepoint we are studying. This observation is consistent with previous reports in the literature (Liechti et al.,2021). While we agree that a pseudo-time course could potentially bolster the evidence about when FtsK foci appear, we believe our current analysis sufficiently demonstrates that basal foci of FtsK appear prior to the appearance of new buds at the base of dividing cells.

(2) Figure 3E: It looks like MreC localization to foci doesn't strictly require MreB polymerization. Is this known for E. coli or other species?

To our knowledge, MreC assembly into a filament has not been shown to be dependent upon MreB in other bacteria.  In Caulobacter crescentus, MreC forms a helical structure that is not dependent upon MreB or MreB filament formation (Dye et al., 2005. PNAS; Divakaruni et al., 2005. PNAS).

(3) Figure 5E: why is nearly half of PBP2 and PBP3 still localized to foci at the membrane even after treatment with mecillinam? This suggests, as the authors mention, that mecillinam reduces the efficiency of localization to the divisome but does not eliminate it. Any ideas why?

At this time, we do not know why inhibiting the catalytic activity of PBP2 with mecillinam does not fully prevent the association of PBP2 with the chlamydial divisome. We have included a statement in the Results (pg. 13 of the revised manuscript) that inhibiting the catalytic activity of PBP2 prevents it from efficiently associating with or maintaining its association with polarized divisome complexes.

(4) Line 262-263: This sentence is confusing-please rephrase. The same as what? Differed from what?

We agree with the reviewer. The wording in lines 262-263 of the original submission has been modified.  

(5) Lines 265-267 and Figure 6: Adding cartoon schematics might help readers visualize cell orientations in Fig. 6 (especially 6B).

Cartoons have been added to Fig. 6C (Fig. 6B in the original submission) to orient the reader.

(6) Line 294-298: Do the authors think that the residual 5-10% of PG foci after FtsK knockdown is due to the ability of residual FtsK to organize divisomes?

We show that knockdown of FtsK is not complete, and while we cannot be certain, it is likely, that the PG foci detected in FtsK knockdown cells is due to the ability of the residual FtsK to organize divisomes that direct PG synthesis.

(7) Do the authors have any evidence that FtsK foci are mobile like treadmilling FtsZ?

We have not performed real-time imaging studies, and we currently have no evidence that FtsK foci are mobile.

(8) FtsK foci here are reminiscent of mobile foci formed by the FtsK-like SpoIIIE at the Bacillus subtilis sporulation septum. This might be a good idea to mention in the Discussion. Is it possible that Ct FtsK is also involved in coordinating chromosome partitioning through the developing septum? (That is another reason why it would be useful to know if the translocase domain was dispensable for localization/activity).

We are currently preparing another manuscript that documents the contribution of the various domains of FtsK to its localization profile and whether the division defect in ftsk knockdown cells can be suppressed by specific subdomains of FtsK. This manuscript not only will include these data, it will also include experiments that address how the site of polarized budding is defined. In the revised manuscript, we have included a description of how the domain organization of chlamydial FtsK is similar to E. coli FtsK (pg. 16 of revision). Chlamydial FtsK also has a similar domain organization as SpoIIIE from B. subtilis. The C-terminal catalytic domain of SpoIIIE is 45% identical to chlamydial FtsK. The N-terminus of SpoIIIE is predicted to encode 4 transmembrane spanning helices, like chlamydial FtsK. However, the N-terminus of SpoIIIE shares no sequence homology with the N-terminus of chlamydial FtsK.  We have not included the similar domain organization of SpoIIIE and chlamydial FtsK in the revised manuscript.

(9) It seems that FtsK foci localize to a particular spot opposite from the active septum, although how this spot is specified is not clear. Is there any geometric clue for FtsK's localization like there is for Min-specified FtsZ localization?

As mentioned above, we are currently preparing another manuscript that documents our efforts to understand how the site of polarized budding is defined.  This analysis is too extensive to include in this study.

(10) As mentioned in the Summary, do the authors know whether the N-terminal membrane binding part of FtsK (FtsKn) sufficient for localization/divisome assembly in Ct as it is in other species? Oullette et al. 2012 showed that FtsKn could interact with MreB in BACTH.

We are currently preparing another manuscript that documents the contribution of the various domains of FtsK to its localization profile.

(11) The previous BACTH result with MreB and FtsKn implies that this interaction is direct, yet the current data suggest that this is not the case. Can the authors comment on this? Is this due to bridging effects inherent in the BACTH system?

We have not presented any data to indicate that FtsK and MreB do not interact. We have only shown that FtsK localization is not dependent upon MreB filament formation (Fig. 3).

(12) The FtsZ-independent role of FtsK in nucleating the divisome suggests that Ct FtsK may differ from other FtsKs structurally - can this be explored, perhaps with AlphaFold 3?

As mentioned above, we have included a paragraph in the discussion of the revised manuscript (pg. 16 of the revised manuscript) that states the domain organization of chlamydial FtsK is similar to E.coli FtsK. This conserved domain organization is evident when we view the structures of the proteins using Alphafold.

(13) Typo on line 559: should be HeLa.

This has been corrected.

eLife Assessment

In this study, the authors use the zebrafish to investigate how the microbiome affects a specialized gut cell called the lysosome rich enterocyte. They use a combination of functional assays for protein absorption, gnotobiotic manipulations and single-cell RNA-seq. The findings in the paper are considered important and the results are convincing.

Reviewer #1 (Public review):

The Bagnat and Rawls groups' previous published work (Park et al., 2019) described the kinetics and genetic basis of protein absorption in a specialized cell population of young vertebrates termed lysosome rich enterocytes (LREs). In this study they seek to understand how the presence and composition of the microbiota impacts the protein absorption function of these cells and reciprocally, how diet and intestinal protein absorption function impact the microbiome.

Strengths of the study include the functional assays for protein absorption performed in live larval zebrafish, which provides detailed kinetics on protein uptake and degradation with anatomic precision, and the gnotobiotic manipulations. The authors clearly show that the presence of the microbiota or of certain individual bacterial members slows the uptake and degradation of multiple different tester fluorescent proteins.

To understand the mechanistic basis for these differences, the authors also provide detailed single-cell transcriptomic analyses of cells isolated based on both an intestinal epithelial cell identity (based on a transgenic marker) and their protein uptake activity. The data generated from these analyses, presented in Figures 3-5, are valuable for expanding knowledge about zebrafish intestinal epithelial cell identities, but of more limited interest to a broader readership. Some of the descriptive analysis in this section is circular because the authors define subsets of LREs (termed anterior and posterior) based on their fabp6 expression levels, but then go on to note transcriptional differences between these cells (for example in fabp6) that are a consequence of this initial subsetting.

Inspired by their single-cell profiling and by previous characterization of the genes required for protein uptake and degradation in the LREs, the authors use quantitative hybridization chain reaction RNA-fluorescent in situ hybridization to examine transcript levels of several of these genes along the length of the LRE intestinal region of germ-free versus mono-associated larvae. They provide good evidence for reduced transcript levels of these genes that correlate with the reduced protein uptake in the mono-associated larval groups.

The final part of the study (shown in Figure 7) characterized the microbiomes of 30-day-old zebrafish reared from 6-30 days on defined diets of low and high protein and with or without homozygous loss of the cubn gene required for protein uptake. The analysis of these microbiomes notes some significant differences between fish genotypes by diet treatments, but the discussion of these data does not provide strong support for the hypothesis that "LRE activity has reciprocal effects on the gut microbiome". The most striking feature of the MDS plot of Bray Curtis distance between zebrafish samples shown in Figure 7B is the separation by diet independent of host genotype, which is not discussed in the associated text. Additionally, the high protein diet microbiomes have a greater spread than those of the low protein treatment groups, with the high protein diet cubn mutant samples being the most dispersed. This pattern is consistent with the intestinal microbiota under a high protein diet regimen and in the absence of protein absorption machinery being most perturbed in stochastic ways than in hosts competent for protein uptake, consistent with greater beta dispersal associated with more dysbiotic microbiomes (described as the Anna Karenina principle here: https://pubmed.ncbi.nlm.nih.gov/28836573/). It would be useful for the authors to provide statistics on the beta dispersal of each treatment group.

Overall, this study provides strong evidence that specific members of the microbiota differentially impact gene expression and cellular activities of enterocyte protein uptake and degradation, findings that have a significant impact on the field of gastrointestinal physiology. The work refines our understanding of intestinal cell types that contribute to protein uptake and their respective transcriptomes. The work also provides some evidence that microbiomes are modulated by enterocyte protein uptake capacity in a diet-dependent manner. These latter findings provide valuable datasets for future related studies.

Comments on revisions:

I suggest that the authors clarify the level of protein in the standard fish food and how this relates to the protein levels in the high protein and low protein diets used in their microbiome study.

Reviewer #2 (Public review):

Summary:

The authors set out to determine how the microbiome and host genotype impact host protein-based nutrition.

Strengths:

The quantification of protein uptake dynamics is a major strength of this work and the sensitivity of this assay shows that the microbiome and even mono-associated bacterial strains dampen protein uptake in the host by causing down-regulation of genes involved in this process rather than a change in cell type.

The use of fluorescent proteins in combination with transcript clustering in the single cell seq analysis deepens our understanding of the cells that participate in protein uptake along the intestine. In addition to the lysozome-rich enterocytes (LRE), subsets of enteroendocrine cells, acinar, and goblet cells also take up protein. Intriguingly, these non-LRE cells did not show lysosomal-based protein degradation; but importantly analysis of the transcripts upregulated in these cells include dab2 and cubn, genes shown previously as being essential to protein uptake.

The derivation of zebrafish mono-associated with single strains of microbes paired with HCR to localize and quantify the expression of host protein absorption genes shows that different bacterial strains suppress these genes to variable extents.

The analysis of microbiome composition, when host protein absorption is compromised in cubn-/- larvae or by reducing protein in the food, demonstrates that changes to host uptake can alter the abundance of specific microbial taxa like Aeramonas.

Reviewer #3 (Public review):

Childers et al. address a fundamental question about the complex relationship within the gut: the link between nutrient absorption, microbial presence, and intestinal physiology. They focus on the role of lysosome-rich enterocytes (LREs) and the microbiota in protein absorption within the intestinal epithelium. By using germ-free and conventional zebrafishes, they demonstrate that microbial association leads to a reduction in protein uptake by LREs. Through impressive in vivo imaging of gavaged fluorescent proteins, they detail the degradation rate within the LRE region, positioning these cells as key players in the process. Additionally, the authors map protein absorption in the gut using single-cell sequencing analysis, extensively describing LRE subpopulations in terms of clustering and transcriptomic patterns. They further explore the monoassociation of ex-germ-free animals with specific bacterial strains, revealing that the reduction in protein absorption in the LRE region is strain-specific.

Strengths:

- The authors employ state-of-the-art imaging to provide clear evidence of the protein absorption rate phenotype, focusing on a specific intestinal region. This innovative method of fluorescent protein tracing expands the field of in vivo gut physiology.<br /> - Using both conventional and germ-free animals for single-cell sequencing analysis, they offer valuable epithelial datasets for researchers studying host-microbe interactions. By capitalizing on fluorescently labelled proteins in vivo, they create a new and specific atlas of cells involved in protein absorption, along with a detailed LRE single-cell transcriptomic dataset.<br /> - Their robust and convincing microbiota analysis puts forward a diet-dependent mechanism of community change upon low-protein diet, intricately linked with the host.

Comments on revisions:

The authors have improved the manuscript following the revision work. No further recommendations.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public review):

The Bagnat and Rawls groups' previous published work (Park et al., 2019) described the kinetics and genetic basis of protein absorption in a specialized cell population of young vertebrates termed lysosome-rich enterocytes (LREs). In this study they seek to understand how the presence and composition of the microbiota impacts the protein absorption function of these cells and reciprocally, how diet and intestinal protein absorption function impact the microbiome.

Strengths of the study include the functional assays for protein absorption performed in live larval zebrafish, which provides detailed kinetics on protein uptake and degradation with anatomic precision, and the gnotobiotic manipulations. The authors clearly show that the presence of the microbiota or of certain individual bacterial members slows the uptake and degradation of multiple different tester fluorescent proteins.

To understand the mechanistic basis for these differences, the authors also provide detailed single-cell transcriptomic analyses of cells isolated based on both an intestinal epithelial cell identity (based on a transgenic marker) and their protein uptake activity. The data generated from these analyses, presented in Figures 3-5, are valuable for expanding knowledge about zebrafish intestinal epithelial cell identities, but of more limited interest to a broader readership. Some of the descriptive analysis in this section is circular because the authors define subsets of LREs (termed anterior and posterior) based on their fabp2 expression levels, but then go on to note transcriptional differences between these cells (for example in fabp2) that are a consequence of this initial subsetting.

Inspired by their single-cell profiling and by previous characterization of the genes required for protein uptake and degradation in the LREs, the authors use quantitative hybridization chain reaction RNA-fluorescent in situ hybridization to examine transcript levels of several of these genes along the length of the LRE intestinal region of germ-free versus mono-associated larvae. They provide good evidence for reduced transcript levels of these genes that correlate with the reduced protein uptake in the mono-associated larval groups.

The final part of the study (shown in Figure 7) characterized the microbiomes of 30-day-old zebrafish reared from 6-30 days on defined diets of low and high protein and with or without homozygous loss of the cubn gene required for protein uptake. The analysis of these microbiomes notes some significant differences between fish genotypes by diet treatments, but the discussion of these data does not provide strong support for the hypothesis that "LRE activity has reciprocal effects on the gut microbiome". The most striking feature of the MDS plot of Bray Curtis distance between zebrafish samples shown in Figure 7B is the separation by diet independent of host genotype, which is not discussed in the associated text. Additionally, the high protein diet microbiomes have a greater spread than those of the low protein treatment groups, with the high protein diet cubn mutant samples being the most dispersed. This pattern is consistent with the intestinal microbiota under a high protein diet regimen and in the absence of protein absorption machinery being most perturbed in stochastic ways than in hosts competent for protein uptake, consistent with greater beta dispersal associated with more dysbiotic microbiomes (described as the Anna Karenina principle here: https://pubmed.ncbi.nlm.nih.gov/28836573/). It would be useful for the authors to provide statistics on the beta dispersal of each treatment group.

Overall, this study provides strong evidence that specific members of the microbiota differentially impact gene expression and cellular activities of enterocyte protein uptake and degradation, findings that have a significant impact on the field of gastrointestinal physiology. The work refines our understanding of intestinal cell types that contribute to protein uptake and their respective transcriptomes. The work also provides some evidence that microbiomes are modulated by enterocyte protein uptake capacity in a diet-dependent manner. These latter findings provide valuable datasets for future related studies.

We thank the Reviewer for their thorough and kind assessment. We appreciate the suggestion for edits and for pointing out areas that needed further clarification.

One point in need of further explanation is the use fabp6 (referred to as fabp2 by the reviewer) to define anterior LREs and their gene expression pattern, which includes high levels of fabp6, something that was deemed a “circular argument” by the reviewer.  The rationale for using fabp6 as a reference is that we were able to define its spatial pattern in relation to other LRE markers and the neighboring ileocyte population using transgenic markers (Lickwar et al., 2017; Wen et al., 2021). Thus, far from being a circular argument, using fabp6 allowed us to identify other markers that are differentially expressed between anterior and posterior LREs, which share a core program that we highlight in our study. In the revised manuscript, we clarified this point (lines 166 – 169).

We followed the Reviewer’s suggestion to test if LRE activity and dietary protein affected beta dispersal. Our analyses revealed that beta dispersion was not significantly different between our experimental conditions. We added details about this analysis (lines 384 – 386) and a new supplemental figure panel (Figure S7C).

Reviewer #2 (Public review):

Summary:

The authors set out to determine how the microbiome and host genotype impact host protein-based nutrition.

Strengths:

The quantification of protein uptake dynamics is a major strength of this work and the sensitivity of this assay shows that the microbiome and even mono-associated bacterial strains dampen protein uptake in the host by causing down-regulation of genes involved in this process rather than a change in cell type.

The use of fluorescent proteins in combination with transcript clustering in the single cell seq analysis deepens our understanding of the cells that participate in protein uptake along the intestine. In addition to the lysozome-rich enterocytes (LRE), subsets of enteroendocrine cells, acinar, and goblet cells also take up protein. Intriguingly, these non-LRE cells did not show lysosomal-based protein degradation; but importantly analysis of the transcripts upregulated in these cells include dab2 and cubn, genes shown previously as being essential to protein uptake.

The derivation of zebrafish mono-associated with single strains of microbes paired with HCR to localize and quantify the expression of host protein absorption genes shows that different bacterial strains suppress these genes to variable extents.

The analysis of microbiome composition, when host protein absorption is compromised in cubn-/- larvae or by reducing protein in the food, demonstrates that changes to host uptake can alter the abundance of specific microbial taxa like Aeramonas.

Weaknesses:

The finding that neurons are positive for protein uptake in the single-cell data set is not adequately discussed. It is curious because the cldn:GFP line used for sorting does not mark neurons and if the neurons are taking up mCherry via trans-synaptic uptake from EECs, those neurons should be mCherry+/GFP-; yet methods indicate GFP+ and GFP+/mCherry+ cells were the ones collected and analyzed.

We thank the Reviewer for the kind and positive assessment of our work, for suggestions to improve the accessibility and clarity of the manuscript, and for pointing out an issue related to a neuronal population that needed further clarification.

It turns out that there is a population of neurons that express cldn15la. They are not easily visualized by microscopy because IECs express this gene much more highly. However, the endogenous cldn15la transcripts can be found in neurons as shown in a recently published dataset (PMID: 35108531) as well as in this study We added a discussion point to clarify this issue (lines 463 – 465).

Reviewer #3 (Public review):

Summary:

Childers et al. address a fundamental question about the complex relationship within the gut: the link between nutrient absorption, microbial presence, and intestinal physiology. They focus on the role of lysosome-rich enterocytes (LREs) and the microbiota in protein absorption within the intestinal epithelium. By using germ-free and conventional zebrafishes, they demonstrate that microbial association leads to a reduction in protein uptake by LREs. Through impressive in vivo imaging of gavaged fluorescent proteins, they detail the degradation rate within the LRE region, positioning these cells as key players in the process. Additionally, the authors map protein absorption in the gut using single-cell sequencing analysis, extensively describing LRE subpopulations in terms of clustering and transcriptomic patterns. They further explore the monoassociation of ex-germ-free animals with specific bacterial strains, revealing that the reduction in protein absorption in the LRE region is strain-specific.

Strengths:

The authors employ state-of-the-art imaging to provide clear evidence of the protein absorption rate phenotype, focusing on a specific intestinal region. This innovative method of fluorescent protein tracing expands the field of in vivo gut physiology.

Using both conventional and germ-free animals for single-cell sequencing analysis, they offer valuable epithelial datasets for researchers studying host-microbe interactions. By capitalizing on fluorescently labelled proteins in vivo, they create a new and specific atlas of cells involved in protein absorption, along with a detailed LRE single-cell transcriptomic dataset.

Weaknesses:

While the authors present tangible hypotheses, the data are primarily correlative, and the statistical methods are inadequate. They examine protein absorption in a specific, normalized intestinal region but do not address confounding factors between germ-free and conventional animals, such as size differences, transit time, and oral gavage, which may impact their in vivo observations. This oversight can lead to bold conclusions, where the data appear valuable but require more nuance.

The sections of the study describing the microbiota or attempting functional analysis are elusive, with related data being overinterpreted. The microbiome field has long used 16S sequencing to characterize the microbiota, but its variability due to experimental parameters limits the ability to draw causative conclusions about the link between LRE activity, dietary protein, and microbial composition. Additionally, the complex networks involved in dopamine synthesis and signalling cannot be fully represented by RNA levels alone. The authors' conclusions on this biological phenomenon based on single-cell data need support from functional and in vivo experiments.

We thank the Reviewer for their assessment and for pointing out some areas that needed to be explained better and/or discussed.

The Reviewer mentions some potential confounding factors (ie., size differences, transit time, oral gavage) in the gnotobiology experiments. We would like to convey that these aspects have been addressed in our experimental design and are now clarified in the revised manuscript: 1- larval sizes were recorded and found to be similar between GF and monoassociated larvae (Figure S6A); 2- while intestinal transit time may be affected by microbes and is a topic of interest, in our assay luminal mCherry cargo is present at high levels throughout the gut and is not limiting at any point during the experiment; 3- gavage, which is necessary for quantitative assays, is indeed an experimental manipulation that may somehow alter the subjects (the same is true for microscopy and virtually any research method). However, it cannot explain differences between GF and CV or alter our conclusions via microbial or dietary effects. We now elaborate the former point in the revised discussion (line 426). A new panel has been added for Fig.S6 to show that standard length was similar in GF and monoassociated larvae (Figure S6A).

We are aware that microbial community composition is often highly variable between experiments and this necessitates adequately high biological replication and inclusion of internal controls to allow conclusions to be drawn. Nevertheless, studies evaluating the utility of 16S rRNA gene sequencing have found that this analysis reveals important impacts of environmental factors on the gut microbiome (PMIDs: 21346791, 31409661, 31324413). Our results provide further evidence that 16S rRNA gene sequencing remains a useful method to detect perturbations to the zebrafish gut microbiome. Reproducing previous findings, we detected many of the core zebrafish microbiota strains in our samples that have been identified by other studies (PMIDs: 26339860, 21472014, 17055441). To ensure the robustness of our results, we included several biological replicates for each condition, co-housed genotypes and included large sample sizes to minimize environmental variability between groups. In response to this reviewer concern, we have added a supplemental beta diversity plot and statistical analyses showing that the microbiomes in our larvae were significantly different from the diets or tank water (Figure S7A). This analysis shows that the host environment influenced microbial community composition (lines 376 – 378). We also added an additional supplemental panel and performed analysis showing that the experimental replicates (i.e., different tanks) were not a significant source of variation in this study (lines 378 – 380) (Figure S7B). This result underscores that the microbiota in these larvae were influenced by both the host and diet.

Regarding dopamine pathways, we acknowledge that it involves complex biology that will require dedicated studies. In this work, we simply point out gene expression patterns we find interesting as they may inform future studies.

Finally, the Reviewer mentions the use of inadequate statistical methods for some analyses without specifying or indicating alternative analyses, only the need to justify the use of two-way ANOVA is made explicit. In this point, we respectfully disagree and would like to emphasize that we use statistical methods that are standard in the field (PMID: 37707499). We nevertheless added a justification for the use of two-way ANOVA where appropriate (lines 635-637, 653-654, 773-776). The two-way ANOVA test was to compare fluorescence profiles of gavages cargoes or HCR probes along the length of the LRE region. This test accounts for differences in fluorescence between experimental conditions in segments (30 μm) along the LRE region (~300 μm). This allows us to capture differences in fluorescence between experimental conditions while accounting for heterogeneity in the LRE region. Please see our comment below for more information about our use of the 2-way ANOVA.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Please provide in the materials and methods the strain identifiers and sources of the bacteria used in the study.

Thank you for the suggestions. Strain identifiers and source information were added to the methods (lines 576-579).

Reviewer #2 (Recommendations for the authors):

(1) This is a very satisfying and thorough analysis of the reciprocal influence of diet, microbiome, and host genotype on protein absorption by the host. Below I make suggestions that mainly relate to making the paper more accessible to a broader audience.

(2) Line 233 Starts a section that reports the findings of the scRNA dataset. The writing is inconsistent with respect to how the genes are listed: whether abbreviation only or spelled out followed by abbreviation. I prefer the latter. For example, slc10a2 is a bile acid Na cotransporter but for those not in the know, they would have to look this up. Perhaps adding a supplementary table that provides a gene list of those discussed in the text with abbreviation/spelled-out, and KEGG terms.

Thank you for pointing out inconsistent gene labeling. We have revised the text with spelled out gene names followed by abbreviations.

(3) Line 461 Where did the neurons come from when you were sorting cldn+ cells?

Neuronal expression of cldn15la was detected in our data and other published datasets (PMID: 37995681, 35108531). We added a note to the text clarifying that neuronal cells can express cldn15la (lines 463-465).

(4) Line 561 1x tricaine should be converted to percentage in solution or concentration throughout.

The tricaine concentration was 0.2 mg/mL. We added this detail to the methods (line 596).

(5) Line 612 Please clarify how normalizations are carried out: is it to the peak value in the germ-free condition? CV never reaches 1.

AUC values were normalized to the peak value in the GF condition at 60 minutes PG. We clarified this step in the methods (lines 618-619).

(6) Line 654-663 I think mCherry here should be mTourquoise?

Thank you for catching this typo. We corrected it in the text.

(7) In Figure 1 Please consider adding a color so that magenta does not represent BOTH germ-free AND mCherry.

Due to the many colors of fluorescent proteins and HCR probes in this paper, we were not able to find an alternative plot line color to represent GF.

(8) In Figure 2 I suggest consistency with respect to the order you present GF/CV

Figure 1 GF->CV

Figure 2 CV->GF

My preference is GF->CV

Images in Figure 2 were re-ordered following reviewer’s recommendation.

Here, 20 minute time point also appears qualitatively different between GF and CV.

There can be slight differences in LREs between individuals. These images were selected because they represented the average differences in the amount of mTurquoise degradation activity that occurred between 20 – 60 minutes post-flushing in the GF and CV conditions.

In Figure 3E Figure legend refers to being able to see BSA in vacuoles. The image should be modified to show this- currently too small.

In response, we enlarged the confocal microscopy images showing DQ red BSA in the LRE region (Figure 3E). We added a panel with confocal microscopy images of the LREs in 6 dpf larva gavaged with DQ red BSA (Figure S3F). These images show that DQ red BSA fluorescence was localized to the LRE lysosomal vacuole.

In Figure 5D, Posterior LRE should be pink not green in the key to the right of the heatmap.

Thank you for catching this error. We have corrected the colors (Figure 5D).

Reviewer #3 (Recommendations for the authors):

(1) Introduction and context:

Expand the introduction to include more background on microbial-mediated protein absorption, with references to relevant findings in Drosophila. This will provide a stronger foundation for the study's contributions to the field.

Thank you for this suggestion. We added information about microbe-mediated amino acid harvest in Drosophila to the introduction (lines 49-53).

(12) Methodological suggestions:

Measure and report differences between germ-free (GF) and conventional (CV) animals, such as transit time, to account for potential confounding factors in protein absorption dynamics.

We respectfully assert that a transit assay is not required for this study and could actually create confusion as an effect in transit time could be interpreted as a contributing factor when it is in fact not the case due to the experimental design. This is because the concentration of luminal protein was equivalent in GF and CV larvae (Figure S1E), so the LREs had equal saturating access to those proteins in both conditions. Furthermore, we showed the microbiota did not degrade fluorescent protein (Figure S1F). Therefore, we feel confident that there was lower protein uptake in the LREs of CV larvae because the microbiome exerted regulatory effects on LRE activity.

Provide detailed information on the gating strategy used for single-cell sorting to enhance the dataset's utility and support claims about cell changes.

The methods we used for sorting cells were previously described (PMID: 31474562). In this manuscript, we describe them under the heading “Fluorescence activated cell sorting for single cell RNA-sequencing.”

Explain the "GeneRatio" metric in figure legends for clarity.

The GeneRatio is the ratio of genes associated with each individual GO term to the number of genes associated with the domain. An explanation was added to the caption (Figure S3C).

(13) Visual and statistical improvements:

Include images of labeled peptidases within lysosome-rich enterocytes (LREs) to reinforce findings.

Thank you for the suggestion. We added images of labeled peptidases in the LRE region (Figure S6E-D).

For Panels 4-F and 5-D, consider using violin plots of selected genes to improve clarity and emphasize major ideas.

In Figure 4F, the heatmap shows multiple genes were upregulated in mCherry-positive cells. We tried the plotting suggested by the reviewer and felt that violin plots could not convey this message as clearly. Likewise, the heatmap in Figure 5D effectively shows the gradient of expression between ileocytes, anterior and posterior LREs.

Strengthen statistical analysis by employing more rigorous methods and justifying their selection, such as using two-way ANOVA where appropriate.

The two-way ANOVA was used to quantify protein uptake or HCR probe fluorescence along the length of the LRE region. This statistical test allowed us to compare differences in fluorescence between experimental conditions in multiple LRE segments (see Authoer response image 1 below for example). As our assays show, the LRE region is heterogenous with segments showing different levels of activity and gene expression. The two-way ANOVA is appropriate because it allows us to account for this heterogeneity by comparing fluorescence across multiple segments.

Author response image 1.

Our figures display these fluorescent levels in line plots (above, left) rather than bar plots (above, right). The results are easier to visualize interpret in line plots, and they display the fluorescence profiles in greater detail.

(14) Technical corrections:

Correct figure references: Figure 5 about tryptophan metabolism should be 5A, S5G-S5H.

We corrected the figure references.

Line 518: Spell out "heterozygotes" instead of using "gets".

We changed the term from “hets” to “heterozygotes.”

(15) Revise Figure S2 citation to match the actual figure labeling.

We corrected the text to indicate “Figure S2” rather than “Figure S2A.”

Additional manuscript modification

· Figure panels 3B-C, S3A-B, 4A-C: Two cluster were relabeled with improved descriptors based on our updated annotations. The clusters “Pharynx-esophagus-cloaca 1” (PEC1) and PEC2 were relabeled as “Pharynx-cloaca 1” and “Pharynx-cloaca 2.”

eLife Assessment

The important manuscript presents convincing evidence of temporal correlations during specific oscillatory activity between the prefrontal cortex, thalamic nucleus reuniens, and the hippocampus, in naturally sleeping animals. Such correlations represent solid evidence to support the notion that the thalamic nucleus reuniens participates in the hippocampal and prefrontal cortex dialogue subserving memory processes.

Reviewer #1 (Public review):

Summary:

In this study, Basha and colleagues aim to test whether the thalamic nucleus reuniens can facilitate the hippocampus/prefrontal cortex coupling during sleep. Considering the importance of sleep in memory consolidation, this study is important to understand the functional interaction between these three majorly involved regions. This work suggests that the thalamic nucleus reuniens has a functional role in synchronizing the hippocampus and prefrontal cortex. Therefore, it paves the way to new perspectives in order to decipher the neuronal and circuit mechanisms underlying such processes.

Strengths:

The authors have used an interdisciplinary approach to determine how the thalamic nucleus reuniens can impact on the cortico-hippocampal dialogue during sleep. They performed recording in naturally sleeping cats, and analysed the correlation between the main slow wave sleep oscillatory hallmarks: slow waves, spindles, and hippocampal ripples, and with reuniens' neurons firing. They also associated intracellular recordings to assess the reuniens-prefrontal connectivity, and computational models of large networks in which they determine that the coupling of oscillations is modulated by the strength of hippocampal-thalamic connections.<br /> Since the literature regarding fundamentals in nucleus reuniens anatomy in cats is much thinner as compared to what is available in rodents, the authors have performed complementary functional anatomy experiments in anesthetised cats in order to support the functional results (aka excitatory links between reuniens and its targets).

Weaknesses:

The authors have used cats as animal models to study the hippocampo-cortical dialogue. Whereas this is a very interesting and well-done study, it will address a more limited audience, as the majority of research is conducted on rodents, which have a different functional anatomy. Hence, the mechanisms of triggering of thalamic spindles would therefore be far less relevant in rodents, unless shown otherwise in the future.

Reviewer #2 (Public review):

Summary:

The interplay between the medial prefrontal cortex and ventral hippocampal system is critical for many cognitive processes, including memory and its consolidation over time. A prominent idea in recent research is that this relationship is mediated at least in part by the midline nucleus reuniens with respect to consolidation in particular. Whereas the bulk of evidence has focused on neuroanatomy and the effects of temproary or permanent lesions of the nucleus reuniens, the current work examined the electrophysiology of these three structures and how they inter-relate, especially during sleep, which is anticipated to be critical for consolidation. They provide evidence from intercellualr recordings of the bi-directional functional connectivity among these structures. There is an emphasis on the interactions between these regions during sleep, especially slow wave sleep. They provide evidence, in cats, that cortical slow waves precede reuniens slow waves and hippocampal sharp-wave ripples, which may reflect prefrontal control of the timing of thalamic and hippocampal events, They also find evidence that hippocampal sharp wave ripples trigger thalamic firing and precede the onset of reuniens and medial prefrontal cortex spindles. The authors suggest that the effectiveness of bidirectional connections between the reuniens and the (ventral) CA1 is particularly strong during non-rapid eye movement sleep in the cat. This is a very interesting, complex study on a highly topical subject.

Strengths:

An excellent array of different electrophysiological techniques and analyses are conducted. The temporal relationships described are novel findings that suggest mechanisms behind the interactions between the key regions of interest. These may be of value for future experimental studies to test more directly their association with memory consolidation.

Weaknesses:

The number of findings provided is complex and some readers may struggle to follow all the details. The fact that bidirectional connections exist in the model system is not new per se. How and why the specific findings add to existing literature could still be presented with a little more impact. However, I am not sure this can be done more easily than is currently presented. I leave that to the authors to consider.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this study, Basha and colleagues aim to test whether the thalamic nucleus reuniens can facilitate the hippocampus/prefrontal cortex coupling during sleep. Considering the importance of sleep in memory consolidation, this study is important to understand the functional interaction between these three majorly involved regions. This work suggests that the thalamic nucleus reuniens has a functional role in synchronizing the hippocampus and prefrontal cortex.

Strengths:

The authors performed recordings in naturally sleeping cats, and analysed the correlation between the main slow wave sleep oscillatory hallmarks: slow waves, spindles, and hippocampal ripples, and with reuniens' neurons firing. They also associated intracellular recordings to assess the reuniens-prefrontal connectivity, and computational models of large networks in which they determined that the coupling of oscillations is modulated by the strength of hippocampal-thalamic connections.

Thank you for your positive evaluation.

Weaknesses:

The authors' main claim is made on slow waves and spindle coupling, which are recorded both in the prefrontal cortex and surprisingly in reuniens. Known to be generated in the cortex by cortico-thalamic mechanisms, the slow waves and spindles recorded in reuniens show no evidence of local generation in the reuniens, which is not anatomically equipped to generate such activities. Until shown differently, these oscillations recorded in reuniens are most likely volume-conducted from nearby cortices. Therefore, such a caveat is a major obstacle to analysing their correlation (in time or frequency domains) with oscillations in other regions.

(1) We fully agree with the reviewer that reuniens likely does not generate neither slow waves nor spindles. We do not make such claim, which we clearly stated in the discussion (lines 319-324). We propose that Reuniens neurons mediate different forms of activity. In the model, we introduced MD nucleus only because without MD we were unable to generate spindles. While the slow waves and spindles are generated in other thalamocortical regions, the REU neurons show these rhythms due to long-range projections from these regions to REU as has been shown in the model.

(2) Definitely, we cannot exclude some influence of volume conductance on obtained LFP recordings in REU nucleus. However, we show modulation of spiking activity within REU by spindles. Spike modulation cannot be explained by volume conductance but can be explained by either synaptic drive (likely the case here) or some intrinsic neuronal processes (like T-current).

(3) In our REU recordings for spike identification we used tetrode recordings. If slow waves and spindles are volume conducted, then slow waves and spindles recorded with tetrodes should have identical shape. Following reviewer comment, we took these recordings and subtracted one channel from another. The difference in signal during slow waves is in the order 0.1 mV. Considering that the distance between electrodes is in the order of 20 um, such a difference in voltage is major and can only be explained by local extracellular currents, likely due to synaptic activities originating in afferent structures.

Finally, the choice of the animal model (cats) is the best suited one, as too few data, particularly anatomical ones regarding reuniens connectivity, are available to support functional results.

(1) Thalamus of majority of mammals (definitely primates and carnivores, including cats) contain local circuit interneurons (about 30 % of all neurons). A vast majority of studies in rodents (except LGN nucleus) report either absence or extremally low (i.e. Jager P, Moore G, Calpin P, et al. Dual midbrain and forebrain origins of thalamic inhibitory interneurons. eLife. 2021; 10: e59272.) number of thalamic interneurons. Therefore, studies on other species than rodents are necessary, and bring new information, which is impossible to obtain in rodents.

(2) Cats’ brain is much larger than the brain of mice or rats, therefore, the effects of volume conductance from cortex to REU are much smaller, if not negligible. The distance between REU and closest cortical structure (ectosylvian gyrus) in cats is about 15 mm.

(3) Indeed, there is much less anatomical data on cats as opposed to rodents. This is why, we performed experiments shown in the figure 1. This figure contains functional anatomy data. Antidromic responses show that recorded structure projects to stimulated structure. Orthodromic responses show that stimulated structure projects to recorded structure.

Reviewer #2 (Public Review):

Summary:

The interplay between the medial prefrontal cortex and ventral hippocampal system is critical for many cognitive processes, including memory and its consolidation over time. A prominent idea in recent research is that this relationship is mediated at least in part by the midline nucleus reuniens with respect to consolidation in particular. Whereas the bulk of evidence has focused on neuroanatomy and the effects of temproary or permanent lesions of the nucleus reuniens, the current work examined the electrophysiology of these three structures and how they inter-relate, especially during sleep, which is anticipated to be critical for consolidation. They provide evidence from intercellular recordings of the bi-directional functional connectivity among these structures. There is an emphasis on the interactions between these regions during sleep, especially slow-wave sleep. They provide evidence, in cats, that cortical slow waves precede reuniens slow waves and hippocampal sharp-wave ripples, which may reflect prefrontal control of the timing of thalamic and hippocampal events, They also find evidence that hippocampal sharp wave ripples trigger thalamic firing and precede the onset of reuniens and medial prefrontal cortex spindles. The authors suggest that the effectiveness of bidirectional connections between the reuniens and the (ventral) CA1 is particularly strong during non-rapid eye movement sleep in the cat. This is a very interesting, complex study on a highly topical subject.

Strengths:

An excellent array of different electrophysiological techniques and analyses are conducted. The temporal relationships described are novel findings that suggest mechanisms behind the interactions between the key regions of interest. These may be of value for future experimental studies to test more directly their association with memory consolidation.

We thank this reviewer for very positive evaluation of our study.

Weaknesses:

Given the complexity and number of findings provided, clearer explanation(s) and organisation that directed the specific value and importance of different findings would improve the paper. Most readers may then find it easier to follow the specific relevance of key approaches and findings and their emphasis. For example, the fact that bidirectional connections exist in the model system is not new per se. How and why the specific findings add to existing literature would have more impact if this information was addressed more directly in the written text and in the figure legends.

Thank you for this comment. In the revised version, we will do our best to simplify presentation and more clearly explain our findings.

Reviewing Editor (Recommendations for Authors):

Please discuss the ability of reuniens to generate spindles?

We briefly discussed this in previous version. We now extended the discussion (p. 18).

For population data, how many cats were used in acute and chronic experiments, where does the population data originate in Fig. 2? How repeatable were the findings across animals? Was histology verified in each animal?

As previously stated in the beginning of method section we totally used 20 cats: 16 anesthetized (or acute) and 4 non-anesthetized (or chronic). We added number of cats in appropriate places in the result section. Population data in figure 2 comes from 48, 49 or 52 recording sessions (depending on the type of analysis, and indicated in the figure legend) from 4 chronic cats; we clarified this information in the legend. Results were highly repeatable across animals. Histology was verified in all chronic and acute animals, we added a sentence in the method section.

Explanation of figures is very poor, values in figures should be reported in results so they can be compared in the context of the description.

In this revised version, we report most numbers present in figures and their legend to the main text (result section).

The depth of the recording tungsten electrodes are meaningless without the AP and ML coordinates given how heterogenous mPFC is. What is the ventromedial wall of the mPFC in the cat?

We added the ML and AP coordinates in the method section. We corrected ventromedial wall for ventroposterior part of the mPFC.

What are the two vertical lines in 1F?

This was an error while preparing the figure. The panel was corrected.

Line 90 mean +-SD of what? There are no numbers.

Thanks, we now indicate the values.

Panel 2L does not show increased spindling in reuniens prior to PFC as indicated in the results, please explain. It does show SWR in the hippocampus prior to spindles, what is the meaning of such a time relationship?

Panel 2L did show an increased spindling reuniens prior to mPFC, but indeed at the time scale shown, it was not very clear. In this revised manuscript, we added an inset zooming around time zero to make this point clearer.

Panel 2L indeed show an increase in SWR prior to the increase in spindle in both Reuniens and mPFC.

As stated in the discussion, ‘We found that hippocampal SWRs trigger thalamic firing and precede the onset of reuniens and mPFC spindles, which points to SWRs as one of candidate events for spindle initiation.’

It is unclear what the slow waves of PFC mean, these represent filtered PFC lfp, but is this a particular oscillation? They continue to occur during the spindle, while the slow waves supposedly trigger the spindle. Please explain and clarify.

We recently published a review article involving several scientists studying both human and animal sleep that has inserted Box. 1 (Timofeev I, Schoch S, LeBourgeois M, Huber R, Riedner B, Kurth S. Spatio-temporal properties of sleep slow waves and implications for development. Current Opinion in Physiology. 2020; 15: 172–182). In this box among other terms, we provide current definition of slow waves vs slow oscillation. Briefly, if slow waves are repeated with a given rhythm, they typically form slow oscillation. However, if they occur in isolation or are not rhythmic, they remain slow waves, but cannot be called slow oscillation.

Regarding relation of spindles and slow oscillation. We are currently systematically analyzing data on spindles and slow waves obtained from head-restrained and freely behaving cats. One of the main findings is that a majority of ‘cortical’ spindles are local. Local to the extent that spindles can occur in alternation in two neighboring cortical cells. Largely, LFP sleep spindles occur more or less synchronously within suprasylvian gyrus of cats where indeed a large majority of them was triggered by slow waves. The synchrony between LFP spindles in suprasylvian vs other other cortical areas is much less clear. So, it is not surprizing that spindles in one bran region can occur when there is a slow wave present in some other brain region. Something of a kind was also shown in human (Mölle M, Bergmann TO, Marshall L, Born J. Fast and slow spindles during the sleep slow oscillation: disparate coalescence and engagement in memory processing. Sleep. 2011; 34 (10): 1411-1421).

In this regard, we are not ready to include modifications in the manuscript.

Line 134, where is spindle amplitude shown? Plots report power within the spindle frequency band, which obviously captures more than just spindles.

No, plots of figure 3 B, C show the phase-amplitude coupling (PAC) strength. These were calculated with detected spindles, therefore, while we cannot exclude some false spindle detections, we are confident that the false spindle detections are at a negligible level. We modified text and instead of spindle amplitude, we describe SW-spindle amplitude coupling. This reflects our analysis with exactitude.

The discussion must include the medio dorsal nucleus which is the largest thalamic input to the prefrontal cortex and also receives input from the hippocampus. In particular, the case must be made for why reuniens would play a more important or different role than MD? (For example: Occurrence of Hippocampal Ripples is Associated with Activity Suppression in the Mediodorsal Thalamic Nucleus - PMC (nih.gov)).

We cited the suggested study. We cannot say whether reuniens plays a more or less important role. What is clear is that hippocampal ripples at the onset of spindles trigger increased firing in both MD and reuniens. Our extracellular recordings (Fig. 4, K) suggest that the increased firing is associated with spike-bursts. We also have a parallel unpublished study done on anesthetized mice showing SWR triggered inhibitory potentials in both reuniens and MD that reverses around -65mV - -70 mV. Because the majority of SWR occurred at the onset of cortical up state, a relative role of cortico-thalamic vs hippocampo-thalamic drive is not easy to separate. We hope, we will convincingly do this in our forthcoming study, with the limitation that it was done on anesthetized mice.

Reviewer #1 (Recommendations For The Authors):

I strongly encourage the authors to perform current source density analyses on the LFP signals recorded in the nucleus reuniens to make sure that the observed oscillations are indeed locally generated. So far, the anatomical organisation in reuniens cannot support the local generation of oscillations, such as spindles and slow wave. At least in rodents (the cat reuniens does not seem too different, until shown differently), there were no oscillators found in reuniens, and at least not arranged like in cortical areas, allowing the summation in time, and particularly space, of rhythmic input currents. Bipolar recordings with pairs of twisted electrodes might also be useful to assess the local existence of spindles and slow waves.

Current source density calculation is possible when one knows the exact distance between recording sites. As we used tetrodes made with 4 twisted platinum-iridium wires, we know more or less the range of distance between recording sites, but not the exact distance between any given pair of electrodes.

Then, the physical distance between the reuniens and any cortical structure is about 8-9 mm. Therefore, with such distances, volume conductance is expected to be negligible. If slow waves and spindles are volume conducted, then slow waves and spindles recorded with tetrodes should have identical shape. Following reviewer comment, we took these recordings and subtracted one channel from another. The difference in signal during slow waves is in the order 0.1 mV. Considering that the distance between electrodes is in the order of 20 um, such a difference in voltage is major and can only be explained by local extracellular currents, likely due to synaptic activities originating in afferent structures.

Below, we plotted the voltage of one channel of the tetrode versus another channel of the same tetrode. If the signal was simply volume conducted, one would expect to see the vast majority of points on the x=y line (red).

Author response image 1.

Below is a segment of mPFC LFP recording (upper black trace), mPFC LFP filtered for spindle frequency (7-15 Hz) and the spindle detected (black lines above the filtered trace. Then two LFP traces from a tetrode in the Reuniens (orange and light blue) are overlayed. The second trace (Blue) from bottom represents the substraction of Reuniens 1 minus Reuniens 2 channel, and just below (lower Blue trace) is this susbtraction trace filtered for spindle frequency (7-15 Hz) showing clear voltage difference in the spindle range between the two electrodes. Note also that around time 179-179.5 s, there is clear spindle oscillation in the mPFC recording which is not present in the Reuniens recordings.

Author response image 2.

Therefore, we are convinced that in our recordings, volume conductance did not play any significant role.

Another concern regarding delays between events, like slow waves, measured between two regions (as exemplified by Figure 3). It appears that the delays were calculated from the filtered signal. Figure 3G shows a delay between the peak of the mPFC slow wave between the raw and the filtered signal, which might be artifactual of the processing. It is though not (or less) visible for the reuniens recording. Such mismatch might explain the observed differences in delays.

Thanks for this comment. We recomputed the analysis using the original signal (smoothed) and obtained very similar results. Panels H and I of figure 3 were updated using the new analysis performed on original signal.

The overall analyses of LFP-triggered reuniens MUA activity lack of statistics (at least z-scored firing to normalise the firings).

Fig. 2 H and I are representative examples for histograms; statistical data are shown in circular plots as explained in the legend. Fig. 2 L, shows populational data and we provide now standard error. Fig. 4 C and D show individual example. Fig. 4 E shows histograms of activity of all identified putative single units. Units that show significant modulation are displayed above white line. Fig. 4 F shows populational data for significantly modified units.  

A last point of detail in the model, which surprisingly shows reuniens to excitatory hippocampal cells' connectivity. Recent literature reports that reuniens only connect hippocampal interneurons, and not principal cells (at least in rodents, I could not find any report in cats). I wonder how changing this parameter would affect the results of the computational investigation, particularly the results shown in Figure 6.

There are several studies in the literature showing a direct excitation from the Reuniens to pyramidal cells in the CA1, here are three of them:

Goswamee, P., et al. (2021). "Nucleus Reuniens Afferents in Hippocampus Modulate CA1 Network Function via Monosynaptic Excitation and Polysynaptic Inhibition." Frontiers in Cellular Neuroscience 15.

Dolleman-Van der Weel MJ, Lopes da Silva FH, Witter MP (1997) Nucleus Reuniens Thalami Modulates Activity in Hippocampal Field CA1 through Excitatory and Inhibitory Mechanisms. The Journal of Neuroscience 17:5640.

Dolleman-van der Weel MJ, Lopes da Silva FH, Witter MP (2017) Interaction of nucleus reuniens and entorhinal cortex projections in hippocampal field CA1 of the rat. Brain Structure and Function 222:2421-2438.

Because this is not a review paper, we opted to not cite all the papers describing connectivity between mPFC, hippocampus and thalamus.

Reviewer #2 (Recommendations For The Authors):

I respectively suggest that the earlier (public) comments listed above should be addressed. In addition, it would be useful to make it clearer when non-rapid eye movement sleep was being addressed and when rapid eye movement was being addressed. Is it of value to use a single term instead of adding "slow wave sleep" or else clarify when either term is used? The addition of more subheadings might help. Moreover, the relative contribution/value of evidence from these two sleep states was not addressed or was not very clear.

We tried to make it clearer when NREM and when REM was analysed.

We replaced slow-wave sleep with NREM sleep in the figure 5 title.

We added several subheadings in the discussion.

Relative contribution of NREM vs REM sleep was not addressed? Sorry but we do not clearly understand your question. Figs. 2 and 3 deal mainly with NREM sleep (Fig 2.B has an example of REM sleep). Fig. 4 essentially describes results obtained during REM sleep.

I was not sure if the Abstract summarised the key take-home messages from the large amount of evidence provided. Some choices are needed, of course, but "evidence of bidirectional connectivity" struck me as less novel than other evidence provided. Given the huge amount of findings provided, which is commendable, it is still useful to present it perhaps in a more digestible fashion. For example, the headings or the first sentence(s) below headings could indicate the aim or the outcome of the specific method/analysis/findings.

We rewrote abstract and we also added some conclusion to highlight major findings and their meaning.

It is more common to use NRe or Re, rather than REU.

We avoided using RE as, for decades, we used RE to abbreviate the thalamic reticular nucleus in several publications. In this revised version, we spell at full - Reuniens.

Line 49 mentions "short-term" memory. Please specify this more clearly as it is otherwise ambiguous. Also, line 303.

We rephrased the sentence: In particular, the hierarchical coupling of slow waves, spindles and SWRs is thought to play a key role in memory consolidation.

Line 303 was likely about the ventromedial wall: we corrected that sentence.

Line 62: the word, "required" (for memory function) is too strong because there is evidence that it is not always required.

We modified the sentence for plays a major role.

The focus within the medial prefrontal cortex could be specified more clearly / earlier.

The mPFC is mentioned in the second sentence of the abstract and in the first sentence of the introduction.

Line 134: The heading states "determine" and then mentions modulation. These terms may not be interchangeable or they need clarification.

We changed it to slow wave-spindle amplitude coupling. This represents exactly our analysis.

Line 204: Does "cortical network" mean prefrontal cortex network"?

Yes, as described in lines 192-193, the two cortical networks (N1 and N2) of the model represent the mPFC layer 5 and 6 respectively.

Lines 283 to 289: These were not very clear to me.

These lines described the potential mechanisms for the responses to hippocampal and reuniens stimulation recorded intracellularly (results in figure 1). We modified this paragraph for clarity.

Line 296: Specify the "claim".

We modified the sentence for “[…] provides supporting evidence for this claim that nucleus Reuniens might synchronize the activity of ventral hippocampus and mPFC.”

The discussion naturally focuses on the thalamic nucleus reuniens, but also occasionally mentions the thalamic mediodorsal nucleus. The distinction, assuming this is highly relevant, could be expressed more clearly (direct comparison with their previous papers).

We never published a study on the mediodorsal nucleus. We do have some unpublished results from recordings in the MD nucleus and they reveal the presence of an inhibitory component at the beginning of cortical active states, therefore behaving in a similar way to first order nuclei. It is then possible that spindles recorded in the reuniens are actually generated in the MD nucleus and then transmitted to Reuniens through the thalamic reticular nucleus, as both MD and reuniens are connected to the rostral thalamic reticular nucleus. We added some discussion about this.

Figure 1B: Do the authors have any additional evidence of the placements in the reuniens, because the photo provided suggests a large area beyond the reuniens boundary. Also, please confirm is the CEM between Rh and Re in the cat (I think the Rh and Re are adjacent in the rat).

Figure 1B is from an electrolytic lesion, which is necessarily bigger than the tip of the electrode. Therefore the center of the electrolytic lesion indicates where the electrode tip was located which is well within the reuniens nucleus.

Also, yes CE (Nucleus centralis thalami, pars medialis) is located between the reuniens and rhomboid in cats. This can be found in two cat atlas:  

Reinoso-Suárez, F. (1961). Topographischer Hirnatlas der Katze für experimental-physiologische Untersuchungen (Merck).

Berman AL, Jones EG (1982) The Thalamus and Basal Telencephalon of the Cat: A Cytoarchitectonic Atlas with Stereotaxic Coordinates: University of Wisconsin Press.

The first mention of hippocampus in the figure legends should remind the reader by stating "ventral hippocampus".

In this revised version, we added “ventral” in several instances both in the main text and in figure legend.

Figure 2: It seems unusual to mention "unusually short NREM". Presumably, things are the same otherwise - if so, perhaps mention that, especially if some of the effects reflect an "unusual" episode.

We display this particular segment because we want to show continuous recording in which still individual elements characterizing specific states are still visible.

Some effects look like they are strong and others perhaps weaker. If so, how do these impact the final conclusions?

Sorry, we did not understand clearly what is meant here by the reviewer. In general, if any effect has statistically significant difference (old fashion 0.05) we consider it as significant. Any other cases are described on individual basis.

Perhaps "MAD" should be in full on the first occasion, if not already.

It was spelled out at line 659, but we now spell it out also in the results section and in figure 2 legend.

Methods: the key question is the use of rodent recordings to classify cat recordings. It would be good to have a reference indicating that this can be directly used for cats, which may have different sleep cycles and patterns compared to rats.

We did not use rodent recordings to classify cat recordings, however we did used a state detection script that was developed with rodent recordings. As mentioned in the method section, we adapted the script to cat mPFC recordings and then manual corrections were made to correctly detect REM episodes. Respectfully, our lab investigates sleep-wake in non-anesthetized animals for a few decades; we developed state detection algorithm in mice, cats, marmosets when needed (to analyse months of recordings), and we have an extensive expertise in identifying states of vigilance from electrophysiological recordings.

eLife Assessment

This work presents an important mouse model for a liver-specific depletion of the Survival Motor Neuron (SMN) protein, where the liver retains 30% of functional full-length SMN protein. The authors provide a profile of phenotypic changes in liver-specific SMN depleted mice with convincing evidence supporting their claims.

Reviewer #1 (Public review):

Summary:

This manuscript presents a comprehensive exploration of the role of liver-specific Survival Motor Neuron (SMN) depletion in peripheral and central nervous system tissue pathology through a well-constructed mouse model. This study is pioneering in its approach, focusing on the broader physiological implications of SMN, which has traditionally been associated predominantly with spinal muscular atrophy (SMA).

Strengths:

(1) Novelty and Relevance: The study addresses a significant gap in understanding the role of liver-specific SMN depletion in the context of SMA. This is a novel approach that adds valuable insights into the multi-organ impact of SMN deficiency.

(2) Comprehensive Methodology: The use of a well-characterized mouse model with liver-specific SMN depletion is a strength. The study employs a robust set of techniques, including genetic engineering, histological analysis, and various biochemical assays.

(3) Detailed Analysis: The manuscript provides a thorough analysis of liver pathology and its potential systemic effects, particularly on the pancreas and glucose metabolism.

(4) Clear Presentation: The manuscript is well written. The results are presented clearly with well-designed figures and detailed legends.

Weaknesses:

(1) Limited Time Points: The study primarily focuses on a single time point (P19). This limits the understanding of the temporal progression of liver and pancreatic pathology in the context of SMN depletion. Longitudinal studies would provide a better understanding of disease progression.

(2) Incomplete Recombination: The mosaic pattern of Cre-mediated excision leads to variability in SMN depletion, which complicates the interpretation of some results. Ensuring more consistent recombination across samples would strengthen the conclusions.

After the revision, the authors addressed the reviewers' questions by extending their analyses to include P60 mice, conducting both liver and pancreatic analyses, and adding a comprehensive panel of metabolic hormones related to glucose metabolism in animals at P19 and P60. They also corrected all errors identified during the initial review process and expanded the discussion to clarify raised issues. All my questions have now been addressed.

Reviewer #2 (Public review):

Summary:

Marylin Alves de Almeida et al. developed a novel mouse cross via conditionally depleting functional SMN protein in the liver (AlbCre/+;Smn2B/F7). This mouse model retains a proportion of SMN in the liver, which better recapitulates SMN deficiency observed in SMA patients and allows further investigation into liver-specific SMN deficiency and its systemic impact. They show that AlbCre/+;Smn2B/F7 mice do not develop an apparent SMA phenotype as mice did not develop motor neuron death, neuromuscular pathology or muscle atrophy, which is observed in the Smn2B/- controls. Nonetheless, at P19 and P60, these mice develop mild liver steatosis, and interestingly, this conditional depletion of SMN in the liver impacts cells in the pancreas.

Strengths:

The current model has clearly delineated the apparent metabolic perturbations which involve a significantly increased lipid accumulation in the liver and pancreatic cell defects in AlbCre/+;Smn2B/F7 mice at P19 and P60. Standard methods like H&E and Oil Red-O staining show that in AlbCre/+;Smn2B/F7 mice, their livers closely mimic the livers of Smn2B/- mice, which have the full body knockout of SMN protein. Unlike previous work, this liver-specific conditional depletion of SMN is superior in that it is not lethal to the mouse, which allows an opportunity to investigate the long-term effects of liver-specific SMN on the pathology of SMA.

Weaknesses:

Given that SMA often involves fatty liver, dyslipidemia and insulin resistance, using the current mouse model, the authors could have explored the long-term effects of liver-specific depletion of SMN on metabolic phenotypes beyond P19, as well as systemic effects like glucose homeostasis. Given that the authors also report pancreatic cell defects, the long-term effect on insulin secretion and resistance could be further explored. This has been addressed in the revised manuscript. The mechanistic link between a liver-specific SMN depletion and apparent pancreatic cell defects has been made clearer.

Discussion:

This current work explores a novel mouse cross in order to specifically deplete liver SMN using an Albumin-Cre driver line. This provides insight into the contribution of liver-specific SMN protein to the pathology of SMA, which is relevant for understanding metabolic perturbations in SMA patients. Nonetheless, given that SMA in patients involve a systemic deletion or mutation of the SMN gene, the authors could emphasize the utility of this liver-specific mouse model, as opposed to using in vitro models, which have been recently reported (Leow et al, 2024, JCI).

Comments on current version:

No further suggestions. Previous recommendations have been addressed by the authors.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public review):

Summary:

This manuscript presents a comprehensive exploration of the role of liver-specific Survival Motor Neuron (SMN) depletion in peripheral and central nervous system tissue pathology through a well-constructed mouse model. This study is pioneering in its approach, focusing on the broader physiological implications of SMN, which has traditionally been associated predominantly with spinal muscular atrophy (SMA).

Strengths:

(1) Novelty and Relevance: The study addresses a significant gap in understanding the role of liver-specific SMN depletion in the context of SMA. This is a novel approach that adds valuable insights into the multi-organ impact of SMN deficiency.

(2) Comprehensive Methodology: The use of a well-characterized mouse model with liver-specific SMN depletion is a strength. The study employs a robust set of techniques, including genetic engineering, histological analysis, and various biochemical assays.

(3) Detailed Analysis: The manuscript provides a thorough analysis of liver pathology and its potential systemic effects, particularly on the pancreas and glucose metabolism.

(4) Clear Presentation: The manuscript is well written. The results are presented clearly with well-designed figures and detailed legends.

We thank the reviewer for their positive comments. They had some concerns for us to consider (see below). We provide a point-by-point response to their comments.

Weaknesses:

(1) Limited Time Points: The study primarily focuses on a single time point (P19). This limits the understanding of the temporal progression of liver and pancreatic pathology in the context of SMN depletion. Longitudinal studies would provide a better understanding of disease progression.

We thank the reviewer for the suggestion. We extended our analysis to include P60 mice and performed both liver and pancreatic analyses at this time point to address this suggestion.

(2) Incomplete Recombination: The mosaic pattern of Cre-mediated excision leads to variability in SMN depletion, which complicates the interpretation of some results. Ensuring more consistent recombination across samples would strengthen the conclusions.

The variability in Cre-mediated excision is inherently stochastic, influenced by factors such as Cre expression levels, timing of recombination, and the accessibility of the target locus in individual cells. Achieving complete consistency across samples is particularly challenging, especially given the complexity of our breeding scheme, which occasionally results in litters without any animals of the desired genotype. Importantly, our study not only demonstrates that liver-specific SMN depletion results in liver alterations and pancreatic dysfunction but also highlights the limitations and challenges associated with this mouse model. By doing so, we aim to provide valuable insights for other researchers considering similar approaches in future studies.

Reviewer #2 (Public review):

Summary:

Marylin Alves de Almeida et al. developed a novel mouse cross via conditionally depleting functional SMN protein in the liver (AlbCre/+;Smn2B/F7). This mouse model retains a proportion of SMN in the liver, which better recapitulates SMN deficiency observed in SMA patients and allows further investigation into liver-specific SMN deficiency and its systemic impact. They show that AlbCre/+;Smn2B/F7 mice do not develop an apparent SMA phenotype as mice did not develop motor neuron death, neuromuscular pathology or muscle atrophy, which is observed in the Smn2B/- controls. Nonetheless, at P19, these mice develop mild liver steatosis, and interestingly, this conditional depletion of SMN in the liver impacts cells in the pancreas.

Strengths:

The current model has clearly delineated the apparent metabolic perturbations which involve a significantly increased lipid accumulation in the liver and pancreatic cell defects in AlbCre/+;Smn2B/F7 mice at P19. Standard methods like H&E and Oil Red-O staining show that in AlbCre/+;Smn2B/F7 mice, their livers closely mimic the livers of Smn2B/- mice, which have the full body knockout of SMN protein. Unlike previous work, this liver-specific conditional depletion of SMN is superior in that it is not lethal to the mouse, which allows an opportunity to investigate the long-term effects of liver-specific SMN on the pathology of SMA.

We thank the reviewer for their positive comments. They had some concerns for us to consider (see below). We provide a point-by-point response to their comments (review comments in black, our response in red).

Weaknesses:

Given that SMA often involves fatty liver, dyslipidemia and insulin resistance, using the current mouse model, the authors could have explored the long-term effects of liver-specific depletion of SMN on metabolic phenotypes beyond P19, as well as systemic effects like glucose homeostasis. Given that the authors also report pancreatic cell defects, the long-term effect on insulin secretion and resistance could be further explored. The mechanistic link between a liver-specific SMN depletion and apparent pancreatic cell defects is also unclear.

We extended our analysis to include P60 mice and performed both liver and pancreatic analyses at this time point to address this suggestion. In addition, we discussed the liver-pancreas axis in the Discussion.

Discussion:

This current work explores a novel mouse cross in order to specifically deplete liver SMN using an Albumin-Cre driver line. This provides insight into the contribution of liver-specific SMN protein to the pathology of SMA, which is relevant for understanding metabolic perturbations in SMA patients. Nonetheless, given that SMA in patients involve a systemic deletion or mutation of the SMN gene, the authors could emphasize the utility of this liver-specific mouse model, as opposed to using in vitro models, which have been recently reported (Leow et al, 2024, JCI). Authors should also discuss why a mild metabolic phenotype is observed in this current mouse model, as opposed to other SMA mouse models described in literature.

We appreciate the reviewer’s insightful comment. We have thoroughly addressed this suggestion in the Discussion section, particularly in lines 284-298; 309-322 and 334-359.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) Longitudinal Studies: Conducting studies at maybe one more time points postnatally to provide a clearer picture of how liver-specific SMN depletion affects tissue pathology over time.

We extended our analysis to include P60 mice and performed both liver and pancreatic analyses at this time point to address this suggestion.

(2) Functional Assays: Incorporate glucose tolerance tests, insulin sensitivity tests, and more detailed metabolic profiling to better understand the physiological consequences of liver-specific SMN depletion on glucose metabolism and pancreatic function.

We sincerely thank the reviewer for this suggestion. We have included a full panel of metabolic hormones associated with glucose metabolism from animals at P19 and P60. These new data, along with additional figures, have now been provided in our revised manuscript.

(3) Mechanism: Discuss the molecular pathways affected by SMN depletion in the liver and pancreas. Mechanistic studies including transcriptomic or proteomic analyses to identify dysregulated pathways will help.

We appreciate the reviewer’s insightful comment. We have thoroughly addressed this suggestion in the Discussion section, particularly in lines 284-298 and 334-359.

(4) Typos in the abstract: beta cells secret insulin and alpha cells produce gulcagon. 

Thank you for catching this error. It has been corrected to reflect that insulin is produced by beta cells and glucagon by alpha cells.

(5) Efficiency and specificity of the Alb-Cre: if possible, cross the Alb-Cre with the Rosa26 reporter line to test the efficiency and specificity of the Alb-Cre.

We agree that this would provide valuable insights. However, initiating a new breeding program to generate the required genotypes would take over a year and is beyond the scope of this study. To address this in part, we performed Cre immunostaining of the liver, pancreas, and spinal cord at P19, as well as the liver at P60. These results, now included in Supplemental Figure 1, demonstrate liver-specific expression and variability across hepatocytes.

Reviewer #2 (Recommendations for the authors):

The title of this manuscript is potentially misleading. The manuscript largely investigates the involvement of SMN protein on peripheral organs such as the liver, muscles, neuromuscular junction, and the pancreas. Yet, the title could be interpreted that the peripheral nervous system or central nervous system is the main focus. The title should be edited to indicate key terms such as "motor neuron and peripheral tissue pathology".

Thank you for pointing this out. We have revised the title to better represent the study’s focus. It is now “Impact of liver-specific survival motor neuron (SMN) depletion on central nervous system and peripheral tissue pathology”.

Suggestions:

Please clarify and explain clearly the various mouse lines (Smn2B/+, Smn2B/- and +/+; Smn2B/F7 ) used as controls as the nomenclature used is confusing. In addition, authors could consider the use of a wild-type mouse line to be used as a control to validate changes in AlbCre/+;Smn2B/F7 mice.

We have now provided clarification on mouse line nomenclature in the Results section (lines 104–124). Full-body heterozygous mice (_Smn_2B/+) are used as controls due to their slightly reduced SMN protein levels and absence of phenotypic changes compared to wild-type mice.

Given that the main phenotype implicated by the liver-specific depletion of SMN protein in AlbCre/+;Smn2B/F7 mice is pancreatic abnormalities (changes in alpha- and beta- cell numbers and blood glucose levels), authors should expand further on the pancreatic phenotype.  

We added a full panel of metabolic hormones related to glucose metabolism in animals at P19 and P60. Furthermore, this has been discussed in detail in lines 284-298 and 334-344 of the Discussion.

A pancreas-specific depletion of SMN would provide this current manuscript with a better understanding of the role of SMN in regulating SMA pathology and provide more definitive conclusions on the contribution of liver-specific SMN depletion on normal pancreatic function.

We agree that this would be very informative. However, to do this would require initiation of a new breeding program that will take more than a year to arrive at the right genotypes. Although valuable, it is beyond the scope of the present study.

The authors should also delineate the role of hepatic SMN in pancreatic function, and how the intrinsic liver-specific loss of SMN directly impacts the pancreas. Currently, literature demonstrates that the fatty liver phenotype in SMA patients is a primary SMN-dependent hepatocyte-intrinsic liver defect associated with mitochondrial and other hepatic metabolism implications (see Leow et al, 2024 J Clin Invest). Given that the authors describe that SMN protein levels are not altered in the pancreas of AlbCre/+;Smn2B/F7 mice at P19, the authors ought to clarify how pancreas development and function is impacted in this mouse model, whether in-utero or postnatally. This could potentially underscore the cross-talk between liver SMN and pancreas function.

We have discussed the relationship between hepatic SMN and pancreatic function in the Discussion at lines 284-298 and 334-359.

Authors should also perform some metabolic tolerance tests to both oral glucose and insulin at an older age (e.g. P60) to study their homeostasis in these mice. These would help to substantiate the authors' conclusion and provide the paper with a greater level of novelty.

We thank the reviewer for this suggestion. A full panel of metabolic hormones related to glucose metabolism at P19 and P60 has been included, supported by additional figures that enhance the manuscript's novelty and depth.

Authors mentioned in the Discussion in lines 238 to 240: "Altogether, our findings underscore the necessity of conducting further investigations at later time points to unveil potential modifications in other pathways and their repercussions on liver physiology". Please elucidate the effects of longer term liver-specific depletion of SMN beyond P19, such as the onset of NAFLD or a diabetic phenotype due to pancreatic dysfunctions.

We extended our data to include P60 mice and performed liver and pancreatic analyses at these time points. The observed effects were transient, possibly due to the stochastic nature of Cre expression.

In addition, while AlbCre/+;Smn2B/F7 mice had similar weight gain trends as controls, it does appear that AlbCre/+;Smn2B/F7 mice weigh more than their controls by P60 (Figure 9C). This data would provide more convincing evidence of the metabolic defects observed in these mice.

As per the reviewer’s suggestion, we included new data (Figure 9D) showing % weight gain at P60 normalized to basal weight at P7. However, no statistically significant differences were detected.

Other than protein quantification, authors should perform immunohistochemistry or in-situ hybridization of SMN and imaging of AlbCre/+;Smn2B/F7 organs to validate the loss of liver-specific SMN. It is unclear from western blots that the expression of SMN is only in hepatocytes.

We thank the reviewer for the suggestion. Unfortunately, SMN antibodies have not produced reliable tissue immunostaining. To address this, we performed Cre immunostaining of the liver, pancreas, and spinal cord at P19, and the liver at P60, which demonstrated liver-specific expression. These results are now included in Supplemental Figure 1.

Authors should consider re-wording lines 228 through 231: "While our current analysis did not reveal significant differences in AlbCre/+;Smn2B/F7 mice, the observed upward trend in transferrin and HO levels suggests ongoing changes in iron metabolism, which may not be fully manifested at P19". Alternatively, a higher number of mouse samples would allow them to qualify this statement. Authors should also consider comparing levels of liver biomarkers such as ALT and AST, to check for liver homeostatic function.

We have removed speculative statements to avoid unsupported claims.

Recommendations:

The methods and additional details to generate the AlbCre/+;Smn2B/F7 should be explained better in section 2.1 of the Results. It is potentially confusing as to why these mice had to carry both 2B and F7 alleles. Additionally, the role of the F7 allele is not deliberately clear in the Introduction.

Additional details are now included in the Introduction (lines 87-90) and the Results section (lines 104-124).

Authors should refer to Leow et al 2024 (J Clin Invest) and discuss how their current findings compare with their hepatocyte-intrinsic SMN deficiency IPSCs model.<br /> We note a previous publication (Deguise et al 2021 Cell Mol Gastroenterol Hepatol) by the authors which characterized the Smn2B/- mouse model and its NAFLD/NASH features. From our understanding, the Smn2B/- mouse model appears to recapitulate SMA phenotype well, such as the early onset of hepatic steatosis and neurological conditions. As a follow-up to this publication, authors should discuss why this current study of a liver-specific SMN depletion is important and relevant to the study of SMA pathology.

We thank the reviewer for the insightful suggestions. We have included a discussion of these findings and their relevance to the study of SMA pathology in lines 284-298 and 309-322.

Minor corrections:

Abstract (line 32) reads: "a decrease in insulin producing alpha-cells and an increase in glucagon producing beta-cells". The authors should clarify and correct as insulin producing beta-cells and glucagon producing alpha-cells.

Thank you for catching the error. We corrected the description of insulin- and glucagon-producing cells.

Please clarify the number and gender of mice used for weight tracking and motor function experiments up to P60 (Figure 9C). It would be inappropriate if male and female mice were plotted together. If so, authors should stratify data by gender.

We thank the reviewer for the suggestion. Unfortunately, we did not stratify the animals by sex due to the unequal and insufficient number of males and females in our study. To address this, we normalized weight gain to each animal’s starting weight, and no significant differences were observed (now shown in Figure 9D).

The number of figures should be reduced. We recommend merging Figures 1 and 2 (generation of AlbCre/+;Smn2B/F7 mouse line and validation) and Figures 3 and 4 (liver function). Figures 5 through 9 may be supplemental figures instead.

We thank the reviewer for the suggestions. We merged Figures 1 and 2, and Figures 3 and 4, as requested. However, we would prefer to keep the other figures within the main results as they assess the impact of liver-specific depletion of SMN on other pathologies within the mouse model.

Standardize the use of asterisks and reporting p-values in Figure 2. All other figures in the manuscript utilize asterisks, but Figures 2C', 2D' and 2E' use p-values across comparisons.

P-values were included only when they approached statistical significance, providing additional clarity to the results.

It is unclear what the white arrow in Figure 7A indicates.

It is meant to point out the absence of an innervating axon. Please see Figure 5 legend, lines 801-802.

Note spelling errors in Figures 8B and 8C: 'Muscle flber'.

Thank you for catching this. We have corrected the typo to indicate muscle fiber instead.

Please clarify if muscle fiber size should be indicated as µm2 instead of µ2 in Figures 8B and 8C, as written in Materials and Methods under line 394.

Thank you for catching this. We corrected the typo to indicate µm2 instead.

eLife Assessment

This important body of work uses state-of-the-art quantitative methods to characterize and compare behaviors across five different fish species to understand which features are conserved and which ones are differentiated. The convincing results from this study will be of interest to ethologists and also have potential utility in understanding the neural mechanisms leading to these behaviors.

Reviewer #3 (Public review):

Summary:

This paper uses 2D pose estimation and quantitative behavioral analyses to compare patterns of prey capture behavior used by six species of freshwater larval fish, including zebrafish, medaka, and four cichlids. The convincing comparison of tail and eye kinematics during hunts reveals that cichlids and zebrafish use binocular vision and similar hunting strategies, but that cichlids make use of an expanded set of action types. The authors also provide convincing evidence that medaka instead use monocular vision during hunts. This finding has important implications for the evolution of distinct distance estimation algorithms used by larval teleost fish species during prey capture.

Strengths:

The quality of the behavioral data is solid and the high frame rate allowed for careful quantification and comparison of eye and tail dynamics during hunts. The statistical approach to assess eye vergence states (Figure 2B) is elegant, the cross-species comparison of prey location throughout each hunt phase is well done (Figure 3B-D), and the demonstration that swim bout tail kinematics from diverse species can be embedded in a shared "canonical" principal component space to explain most of the variance in 2D postural dynamics for each species (Figure 4A-C) provides a simple and powerful framework for future studies of behavioral diversification across fish species.

Author response:

The following is the authors’ response to the original reviews.

Reviewer 1:

(1) The overall conclusion, as summarized in the abstract as "Together, our study documents the diversification of locomotor and oculomotor adaptations among hunting teleost larvae" is not that compelling. What would be much more interesting would be to directly relate these differences to different ecological niches (e.g. different types of natural prey, visual scene conditions, height in water column etc), and/or differences in neural circuit mechanisms. While I appreciate that this paper provides a first step on this path, by itself it seems on the verge of stamp collecting, i.e. collecting and cataloging observations without a clear, overarching hypothesis or theoretical framework.

There are limited studies on the prey capture behaviors of larval fishes, and ours is the first to compare multiple species systematically using a common analysis framework. Our analysis approach could have uncovered a common set of swim kinematics and capture strategies shared by all species; but instead, we found that medaka used a monocular strategy rather than the binocular strategy of cichlids and zebrafish. Our analysis similarly could have revealed first-feeding larvae of all species go through a “bout” stage, which was previously proposed as important for sensorimotor decision making (Bahl et al., 2019), but instead we found that medaka and some cichlids have more continuous swimming from an early life stage. Finally, the rate at which prey capture kinematics evolves is not known. Our approach could have revealed rapid diversification of feeding strategies in cichlids (similarly to how adult feeding behavior evolves), but instead we found smaller differences within cichlids than between cichlids and medaka.

(2) The data to support some of the claims is either weak or lacking entirely.

Highlighted timestamps in videos, new stats in fig 1H and fig 2, updated supplementary figures now provide additional support for claims.

- It would be helpful to include previously published data from zebrafish for comparison.

We appreciate the suggestion. Mearns et al. (2020) provided a comprehensive account of prey capture in zebrafish larvae in an almost identical setup with similar analyses. We do not feel it is necessary to recount all the findings in that paper here. There are many studies on prey capture in zebrafish from the past 20 years, and reproducing these here would not add anything to that extensive pre-existing literature.

- Justification is required for why it is meaningful to compare hunting strategies when both fish species and prey species are being varied. For instance, artemia and paramecia are different sizes and have different movement statistics.

We added text explaining why different food was chosen for medaka/cichlids. There is no easy way to stage match fishes as evolutionarily diverged as cichlids, medaka, and zebrafish. Size is a reasonable metric within a species, but there is no guarantee that sizematched larvae of two different species are at the same level of maturity. Therefore, we thought the most appropriate stage to address is when larvae first start feeding, as this enables us to study innate prey capture behavior before any learning or experience-dependent changes have taken place. Given that zebrafish, medaka and cichlid larvae are different sizes when they first start feeding, it was necessary to study their hunting behavior to different prey items.

- It would be helpful in Figure 1A to add the abbreviations used elsewhere in the paper. I found it slightly distracting that the authors switch back and forth in the paper between using "OL" and "medaka" to refer to the same species: please pick one and then remain consistent.

Medaka is the common name for the japanese rice fish, O. latipes. Cichlilds do not have common names are only referred to by their scientific names. Since readers are more likely to be familiar with the common name, medaka, we now use medaka (OL) throughout the manuscript, which we hope makes the text clearer.

- The conceptual meaning of behavioral segmentation is somewhat unclear. For zebrafish, the bouts already come temporally segmented. However in medaka for instance, swimming is more continuous, and the segmentation is presumably more in terms of "behavioral syllables" as have been discussed for example mouse or drosophila behavior (in the last row of Figure S1 it is not at all obvious why some of the boundaries were placed at their specific locations). It's not clear whether it's meaningful to make an equivalence between syllables and bouts, and so whether for instance Figure 1H is making an apples-to-apples comparison.

We clarified the text to say we are comparing syllables, rather than bouts.

- The interpretation of 1H is that "medaka exhibited significantly longer swims than cichlids"; however this is not supported by the appropriate statistical test. The KS test only says that two probability distributions are different; to say that one quantity is larger than another requires a comparison of means.

Updated Fig 1H; boostrap test (difference of medians) and re plotted data as violin plots.

(2) The data to support some of the claims is either weak or lacking entirely.

Highlighted timestamps in videos, new stats in fig 1H and fig 2, updated supplementary figures now provide additional support for claims.

- I think the evidence that there are qualitatively different patterns of eye convergence between species is weak. In Figure 2A I admire the authors addressing this using BIC, and the distributions are clearly separated in LA (the Hartigan dip test could be a useful additional test here). However for LO, NM, and AB the distributions only have one peak, and it's therefore unclear why it's better to fit them with two Gaussians rather than e.g. a gamma distribution. Indeed the latter has fewer parameters than a two-gaussian model, so it would be worthwhile to use BIC to make that comparison. The positions of the two Gaussians for LO, NM, and AB are separated by only a handful of degrees (cf LA, where the separation is ~20 degrees), which further supports the idea that there aren't really two qualitatively different convergence states here.

Added explanation to text.

- Figure S2 is unfortunately misleading in this regard. I don't claim the authors aimed to mislead, but they have made the well-known error of using colors with very different luminances in a plot where size matters (see e.g.

https://nam12.safelinks.protection.outlook.com/?url=https%3A %2F%2Fwww.r-project.org%2Fconferences%2FDSC2003%2FProceedings%2FIhaka.pdf&data=05%7C02%7Cdme arns%40princeton.edu%7C17ae2b44f0f246f15ddd08dc9b8e2 01c%7C2ff601167431425db5af077d7791bda4%7C0%7C0%7

C638556282750568814%7CUnknown%7CTWFpbGZsb3d8ey

JWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJ XVCI6Mn0%3D%7C0%7C%7C%7C&sdata=Ll4J4Xo39JEtKb %2FNnRWNoyedZAu5aAOMq0lHJCwsfXI%3D&reserved=0).

Thus, to the eye, it appears there's a big valley between the red and blue regions, but actually, that valley is full of points: it's really just one big continuous blob.

Kernel density estimation of eye convergence angles were added to Figure S2. The point we wish to make is that there is higher density when both eyes are rotated invwards (converged) in cichlids, but not medaka (O. latipes). The valley between converged and unconverged states being full of points is due to (1) slight variation with placement of key points in SLEAP, which blurs the boundary between states and (2) the eye convergence angle must pass through the valley in order to become converged, so necessarily there are points in between the two extremes of eye convergence.

- In Figure 2D please could the authors double-check the significance of the difference between LO and NM: they certainly don't look different in the plot.

Thank for for flagging this. We realize the way we previously reported the stats was open to misinterpretation. We have updated figure 2C, D and F to use letters to indicate statistical groupings, which hopefully makes it clearer which species are statistically different from each other.

- In Figure 2G it's not clear why AB is not included. It is mentioned that the artemia was hard to track in the AB videos, but the supplementary videos provided do not support this.

The contrast of the artemia in the AB videos is sufficiently different from the other cichlid videos that our pre-trained YOLO model fails. Retraining the model would be a lot of extra work and we feel like a comparison of three species is sufficient to address the sensorimotor transformations that occur over the course of prey capture in cichlids.

- The statement "Zebrafish larvae have a unique swim repertoire during prey capture, which is distinct from exploratory swim bouts" is not supported by the work of others or indeed the authors' own work. In Figure 4F all types of bouts can occur at any time, it's just the probability at which they occur that varies during prey capture versus other times (see also Mearns et al (2020) Figure S4B).

The point is well taken that there probably is not a hard separation between spontaneous and prey capture swims based on tail kinematics alone, which is also shown in Marques et al. (2018). However, we think that figure 2I of Mearns et al., which plots the probability of swims being drawn from different parts of the behavior space during prey capture (eyes converged) or not (eyes unconverged), shows that the repertoire of swims during the two states is substantially different. Points are blue or red; there are very few pale blue/pale red points in that figure panel. Figure S4B is showing clustered data, and clustering is a notoriously challenging problem for which there exists no perfect solution (Kleinberg, 2002). The clusters in Mearns et al. incorporated information about transition structure, as this was necessary for obtaining interpretable clusters for subsequent analyses. However, a different clustering approach could have yielded different boundaries, which may have shown more (or less) separation of bout types during prey capture/exploratory swimming. Therefore, we have updated the text to say that zebrafish perferentially perform different swim types during prey capture and exploration, and re-interpreted the behavior of cichlids similarly.

- More discussion is warranted of the large variation in the number of behavioral clusters found between species (11-32). First, how much is this variation really to be trusted? I appreciate the affinity propogation parameters were the same in all cases, but what parameters "make sense" is somewhat dependent on the particular data set. Second, if one does believe this represents real variation, then why? This is really the key question, and it's unsatisfying to merely document it without trying to interpret it.

Extended paragraph with more interpretation.

- What is the purpose of "hovers"? Why not stay motionless? Could it be a way of reducing the latency of a subsequent movement? Is this an example of the scallop theorem?

Added a couple of sentences speculating on function.

- I'm not sure "spring-loaded" is a good term here: the tension force of a coiled tail is fairly negligible since there's little internal force actively trying to straighten it.

Rewrote this part to highlight that fish spring toward the prey, without the implication that tension forces in the tail are responible for the movement. However, we are not aware of any literature measuring passive forces within the tail of fishes. Presumably the notochord is relatively stiff and may provide an internal force trying to straighten the tail.

- There are now several statements for which no direct evidence is presented. We shouldn't have to rely on the author's qualitative impressions of what they observed: show us quantitative analysis.

* "often hover"

* "cichlids often alternate between approaches and hover swims"

* "over many hundreds of milliseconds"

* "we have also observed suction captures and ram-like attacks"

* "may swim backwards"

* "may expel prey from their mouth"

* "cichlid captures often occur in two phases"

Added references to supplementary videos with timestamps to highlight these behaviors.

- I don't find it plausible that sated fish continue hunting prey that they know they're not going to eat just for the practice.

Removed the speculation.

- In Figure 3 is it not possible to include medaka, based on the hand-tracked paramecia?

The videos are recorded at high frame rate, so it would be a lot of additional work to track these manually. Furthermore, earlier in prey capture it is very difficult to tell by watching videos which prey the medaka are tracking, especially as single paramecia can drift in and out of focus in the videos. Since there is no eye convergence, it is very difficult to ascertain for certain when tracking a given prey begins. In Fig 4, it was only possible to track paramecia by hand since it is immediately prior to the strike and from the video it is possible to see which paramecium the fish targeted. Our analyses of heading changes was performed over the 200 ms prior to a strike, which we think is a conservative enough cutoff to say that fish were probably pursuing prey in this window (it is shorter than the average behavioral syllable duration in medaka).

- Figure 3 (particularly 3D) suggests the interesting finding that LA essentially only hunt prey that is directly in front of them (unlike LO and NM, the distribution of prey azimuth actually seems to broaden slightly over the duration of hunting events).

This is worthy of discussion.

We offer a suggestion for the many instances of prey capture being initiated in the central visual field in LA later in the manuscript when we discuss spitting behavior. We have added text to make this point earlier in the manuscript. The increase in azimuthal range at the end of prey capture may be due to abort swims (e.g. supp. vid. 1, 00:21). The widening of azimuthal angles is present in LO and NM also and is not unique to LA.

- The reference Ding et al (2016) is not in the reference list.

Wrong paper was referenced. Should be Ding 2019, which has been added to bibliography.

- I am not convinced that medaka exhibit a unique side-swing behavior. I agree there is this tendency in the example movie, however, the results of the quantification (Figure 4) are underwhelming. First, cluster 5 in 4K appears to include a proportion of cases from LA and AB. These proportions may be small, but anything above zero means this is not unique to medaka. Second, the heading angle (4N) starts at 4 degrees for LA and 8 degrees for medaka. This difference is genuine but very small, much smaller than what's drawn in the schematic (4M). I'm not sure it's justifiable to call a difference of 4 degrees a qualitatively different strategy.

We have changed the text to highlight that side swing is highly enriched in medaka. Comparing 4J to 3B we would argue that there is a qualitative difference in the strategy used to capture prey in the cichlid larvae we study here and medaka. We agree that further work is required to understand distance estimation behaviors in different species. In this manuscript, we use heading angle as a proxy for how prey position might change on the retina over a hunting sequence. But as the heading and distance are changing over time, the actual change in angle on the retina for prey may be much larger than the ~8 degree shift reported here. The actual position of the prey is also important here, which, for reasons mentioned above, we could not track. Given the final location of prey in the visual field prior to the strike (Fig 4J), the most parsimonious explanation of the data is that the prey is always in the monocular visual field. In cichlids, the prey is more-or-less centered in the 200 ms preceding the strike. While it is true theat the absolute difference in heading is 4 degrees, when converted to an angular velocity (4N, right), the medaka (OL) effectively rotate twice as fast as LA (20 deg/s vs 40 deg/s), which we think is a substantial difference and evidence of a different targeting strategy.

- 4K: This is referred to in the caption as a confusion matrix, which it's not.

Fixed.

- 4N right panel: how many fish contributed to the points shown?

Added to figure legend (n=113, LA; n=36, OL). Same data in left and right panels.

- In the Discussion it is hypothesized that medaka use their lateral line in hunting more than in other species. Testing this hypothesis (even just compared to one other species) would be fairly straightforward, and would add significant interest to the paper overall.

We agree that this is an interesting experiment for follow up studies, but it is beyond the scope of the current manuscript as we do not have the appropriate animal license for this experiment.

Reviewer 2:

The paper is rather descriptive in nature, although more context is provided in the discussion. Most figures are great, but I think the authors could add a couple of visual aids in certain places to explain how certain components were measured.

Added new supplemental figure (Supp Fig 2)

Figure 1B- it could be useful to add zebrafish and medaka to the scientific names (I realize it's already in Figure A but I found myself going back and forth a couple of times, mostly trying to confirm that O. latipes is medaka).

Added common names to 1B, sprinkled reminders of OL/medaka throughout text.

Figure 1G. I wasn't sure how to interpret the eye angle relative to the midline. Can they rotate their eyes or is this due to curvature in the 'upper' body of the fish? Adding a schematic figure or something like that could help a reader who is not familiar with these methods. Related to this, I was a bit confused by Figure 2A. After reading the methods section, I think I understand - but I little cartoon to describe this would help. It also reminds the reader (especially if they don't work with fish) that fish eyes can rotate. I also wanted to note that initially, I thought convergence was a measure of how the two eyes were positioned relative to the prey given the emphasis given on binocular vision, and only after reading certain sections again did I realize convergence was a measure of eye rotation/movement.

New supplemental figure explaining how eye tracking is performed

Figure 3. It was not immediately clear to me what onset, middle, and end represented - although it is explained in the caption. I think what tripped me up is the 'eye convergence' title in the top right corner of Figure 3A.

Updated figure with schematic illustrating that time is measured relative to eye convergence onset and end.

The result section about attack swim, S-strike, capture spring, etc. was a bit confusing to read and could benefit from a couple of concise descriptions of these behaviors. For example, I am not familiar with the S strike but a couple of paragraphs into this section, the reader learns more about the difference between S strike vs. attack swim. This can be mentioned in the first paragraph when these distinct behaviors are mentioned.

Added description of behavior earlier in text.

Figure 4. Presents lots of interesting data! I wonder if using Figure 1E could help the reader better understand how these measurements were taken.

New supplemental figure added, explaining how tail tracking is performed.

I probably overlooked this, but I wonder why so many panels are just focused on one species.

Added explanation to the text.

Is the S-shaped capture strategy the same as an S strike?

Clarified in text to say "S-strike-like". This is a description of prey capture from adult largemouth bass in New et al. (2002). From the still frames shown in that paper, the kinematics looks similar to an S-strike or capture spring. The important point we wish to make is that tail is coiled in an S-shape prior to a strike, which indicates this that a kinematically similar behavior exists fishes beyond just larval cichlids and zebrafish.

At the end of the page, when continuous swimming versus interrupted swimming is discussed, please remind the reader that medaka shows more continuous swimming (longer bouts).

Added "while medaka swim continuously with longer bouts ("gliding")".

After reading the discussion, it looks like many findings are unique. For example, given that medaka is such a popular model species in biology, it strikes me that nobody has ever looked into their hunting movements before. If their findings are novel, perhaps they should state so it is clear that the authors are not ignoring the literature.

We have highlighted what we believe to be the novelty of our findings (first description of prey capture in larval cichlids and medaka). To our knowledge, we are first to describe hunting in medaka; but there is an extensive literature on medaka dating back to the early 20th century, some of which is only published in Japanese. We have done our best to review the literature, but we cannot rule out that there are papers that we missed. No English language article or review we found mentions literature on hunting behavior in medaka larvae.

Reviewer 3:

More evidence is needed to assess the types of visual monocular depth cues used by medaka fish to estimate prey location, but that is beyond the scope of this compelling paper. For example, medaka may estimate depth through knowledge of expected prey size, accommodation, defocus blur, ocular parallax, and/or other possible algorithms to complement cues from motion parallax.

Added sentence to discussion highlighting that other cues may also contribute to distance estimation in cichlids and medakas. Follow-up studies will require new animal license.

None. It's quite nice, timely, and thorough work! For future work, one could use 3D pose estimation of eye and prey kinematics to assess the dynamics of the 2D image (prey and background) cast onto the retina. This sort of representation could be useful to infer which monocular depth cues may be used by medaka during hunting.

Great suggestion for follow up studies. Bolton et al. and Mearns et al. both find changes in z associated with prey capture, and it would be interesting to see how other fish species use the full 3-dimensional water column during prey capture, especially considering the diversity of hunting strategies in adult cichlids (ranging from piscivorous species, like LA, to algar grazers).

In Figure 4N, you use "change in heading leading up to a strike as a proxy for the change in visual angle of the prey for cichlids and medaka." This proxy makes sense, but you also have the eye angles and (in some cases) the prey positions. One could estimate the actual change in visual angle from this information, which would also allow one to measure whether the fish are trying to stabilize the position of the prey on a high-acuity patch of the retina during the final moments of the hunt. This information may also shed light on which monocular depth cues are used.

As addressed in comment to reviewer 1, this would require actually manually tracking individual paramecia over hundreds of frames. It is not possible to determine exactly when hunting begins in medaka, and it is prone to errors if medaka switch between targets over the course of a hunting episode. This question is better addressed with psychophysics experiments in embedded animals where it is possible to precisely control the stimulus, but this requires new animal licenses and is beyond the scope of this paper.

In Figure 5, you could place the prey object a little farther from the D. rerio fish for the S-strike diagram.

Fixed.

Figure 4F legend should read "...at the peak of each bout."

Fixed.

eLife Assessment

This study presents solid evidence for distinct neurotransmitter release modalities between two subclasses of dopaminergic neurons in the olfactory bulb, highlighting an important finding that dendritic neurotransmitter release in anaxonic neurons and axonal neurotransmitter release in axon-bearing neurons, and GABAergic self-inhibition in anaxonic neurons emphasizes the functional differences between these neuronal subtypes. However, some experiments looked incomplete with a relatively small sample size (low n). The conclusion would benefit significantly from additional validations.

Reviewer #1 (Public review):

Summary:

Dorrego-Rivas et al. investigated two different DA neurons and their neurotransmitter release properties in the main olfactory bulb. They found that the two different DA neurons in mostly glomerular layers have different morphologies as well as electrophysiological properties. The anaxonic DA neurons are able to self-inhibit but the axon-bearing ones are not. The findings are interesting and important to increase the understanding both of the synaptic transmissions in the main olfactory bulb and the DA neuron diversity. However, there are some major questions that the authors need to address to support their conclusions.

(1) It is known that there are two types of DA neurons in the glomerular layer with different diameters and capacitances (Kosaka and Kosaka, 2008; Pignatelli et al., 2005; Angela Pignatelli and Ottorino Belluzzi, 2017). In this manuscript, the authors need to articulate better which layer the imaging and ephys recordings took place, all glomerular layers or with an exception. Meanwhile, they have to report the electrophysiological properties of their recordings, including capacitances, input resistance, etc.

(2) It is understandable that recording the DA neurons in the glomerular layer is not easy. However, the authors still need to increase their n's and repeat the experiments at least three times to make their conclusion more solid. For example (but not limited to), Fig 3B, n=2 cells from 1 mouse. Fig.4G, the recording only has 3 cells.

(3) The statistics also use pseudoreplicates. It might be better to present the biology replicates, too.

(4) In Figure 4D, the authors report the values in the manuscript. It is recommended to make a bar graph to be more intuitive.

(5) In Figure 4F and G, although the data with three cells suggest no phenotype, the kinetics looked different. So, the authors might need to explore that aside from increasing the n.

(6) Similarly, for Figure 4I and J, L and M, it is better to present and analyze it like F and G, instead of showing only the after-antagonist effect.

Reviewer #2 (Public review):

Summary:

This study provides novel insights into the neurotransmitter release mechanisms employed by two distinct subclasses of dopaminergic neurons in the olfactory bulb (OB). The findings suggest that anaxonic neurons primarily release neurotransmitters through their dendrites, whereas axon-bearing neurons predominantly release neurotransmitters via their axons. Furthermore, the study reveals that anaxonic neurons exhibit self-inhibitory behavior, indicating that closely related neuronal subclasses may possess specialized roles in sensory processing.

Strengths:

This study introduces a novel and significant concept, demonstrating that two closely related neuron subclasses can exhibit distinct patterns of neurotransmitter release. Therefore, this finding establishes a valuable framework for future investigations into the functional diversity of neuronal subclasses and their contributions to sensory processing. Furthermore, these findings offer fundamental insights into the neural circuitry of the olfactory bulb, enhancing our understanding of sensory information processing within this critical brain region.

Weaknesses:

While this study offers novel insights, it is hindered by several limitations. The experimental approaches sometimes lack comprehensive justification and often rely on citations without providing adequate explanatory context. The small sample sizes (n values) compromise the statistical reliability and generalizability of the findings. Furthermore, the reliance on synaptophysin-based presynaptic structures raises concerns regarding whether these structures represent functional synapses. These shortcomings highlight the need for further refinement and additional data to substantiate the study's central conclusions. Addressing these concerns would improve the rigor and impact of the study's findings while ensuring the validity of its conclusions.

eLife Assessment

The authors adapt a previously-established method that permits detection of in vivo extracellular levels of two distinct enkephalin opioid peptides in response to stressful experiences in mice. The present study highlights the potential of measuring actual peptides by microdialysis-LC-MS. They use this approach in conjunction with fiber photometric calcium imaging to correlate enkephalin neuron activity and enkephalin release in response to repeated stress, providing convincing evidence that this improved approach can provide new insights into opioid signaling in-vivo. This important study provides a means to understand various behavioral states controlled by endogenous opioids and the nucleus accumbens, including hedonic and stress responses, in health and disease. This work will be of broad interest to the neuroscientific community.

Reviewer #1 (Public review):

The present study by Mikati et al describes an improved method for in-vivo detection of enkephalin release and examines the impact of stress on activation of enkephalin neurons and enkephalin release in the nucleus accumbens (NAc). The authors refine their pipeline to measure met and leu enkephalin using liquid chromatography and mass spectrometry. The authors subsequently measure met and leu enkephalin in the NAc during stress induced by handling and fox urine, in addition to calcium activity of enkephalinergic cells using fiber photometry. The authors conclude that this improved tool for measuring enkephalin reveals experimenter handling stress-induced enkephalin release in the NAc that habituates and is dissociable from calcium activity of these cells, whose activity doesn't habituate. The authors subsequently show that NAc enkephalin neuron calcium activity does habituate to fox urine exposure, is activated by a novel weigh boat, and that fox urine acutely causes increases in met-enk levels, in some animals, as assessed by microdialysis. This study highlights a new approach to monitor two distinct enkephalins and more robust analytical approach for more sensitive detection of neuropeptides. The authors also provide a pipeline that potentially could aid in detection of other neuropeptides and increase our understanding of endogenous opioid neuropeptidergic control in health and disease.

Reviewer #2 (Public review):

Summary:

The authors aimed to improve the detection of enkephalins, opioid peptides involved in pain modulation, reward, and stress. They used optogenetics, microdialysis and mass spectrometry to measure enkephalin release during acute stress in freely moving rodents. Their study provided detection and quantitation of enkephalins due to implementation of previously reported derivatization reaction combined with improved sample collection and offered insights into the dynamics and relationship between Met- and Leu-Enkephalin in the Nucleus Accumbens shell during stress.

Strengths:

A strength of this work is the quantitative Enk measurements resulted from an optimized microdialysis technique coupled with established derivatization approach and sensitive and quantitative nLC-MS measurements. This approach allowed basal and stimulated peptide release with higher temporal resolution, lower detection thresholds, and native-state endogenous peptide measurement.

Weaknesses:

The optimization of the previously published customizable microdialysis probe and the Met-Enk oxidation derivatization are included in the results, but these adjustments seem more like refinements or practical adaptations rather than significant innovations.

Another consideration is the use of log transformation for quantitation of peptides. Transforming data to achieve normality for parametric tests does not guarantee that all assumptions of normality are met, especially in small and variable datasets reported here. Visual checks like QQ-plots can help verify the appropriateness of transformations. In biological contexts, log transformation can obscure the relationship between measured values and underlying processes.

Reviewer #3 (Public review):

Summary:

This important paper describes improvements to the measurement of enkephalins in vivo using microdialysis and LC-MS. The key improvement is oxidation of met- to prevent having a mix of reduced and oxidized methionine the sample which make quantification more difficult. It then shows measurements of enkephalins in the nucleus accumbens in two different stress situations-handling and exposure to predator odor. It also reports the ratio of released met- and leu-enkephalin matching that expected from digestion of proenkephalin. Measurements are also made by photometry of Ca2+ changes for the fox odor stressor. Some key takeaways are: 1) reliable measurement of met-enkephalin, significance of directly measuring peptides as opposed to proxy measurements, and the opening of a new avenue into research of enkephalins due to stress based on these direct measurements.

Strengths:

- Improved methods for measurement of enkephalins in vivo<br /> - Compelling examples of using this method<br /> - Opening a new area of looking at stress responses through the lens of enkephalin concentrations

Comments on revisions:

This revision has been improved upon in most ways. As I mentioned in the original review, there is a great deal of work here on showing the capability of measuring met- and leu-enk in different contexts. There is a technical improvement in the control of met oxidation which likely improves the detection of met-enk.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

Thank you for your constructive feedback and recognition of our work. We followed your suggestion and improved the accuracy of the language used to interpret some of our findings. 

Summary:

The present study by Mikati et al demonstrates an improved method for in-vivo detection of enkephalin release and studies the impact of stress on the activation of enkephalin neurons and enkephalin release in the nucleus accumbens (NAc). The authors refine their pipeline to measure met and leu enkephalin using liquid chromatography and mass spectrometry. The authors subsequently measured met and leu enkephalin in the NAc during stress induced by handling, and fox urine, in addition to calcium activity of enkephalinergic cells using fiber photometry. The authors conclude that this improved tool for measuring enkephalin reveals experimenter handling stress-induced enkephalin release in the NAc that habituates and is dissociable from the calcium activity of these cells, whose activity doesn't habituate. The authors subsequently show that NAc enkephalin neuron calcium activity does habituate to fox urine exposure, is activated by a novel weigh boat, and that fox urine acutely causes increases in met-enk levels, in some animals, as assessed by microdialysis.

Strengths:

A new approach to monitoring two distinct enkephalins and a more robust analytical approach for more sensitive detection of neuropeptides. A pipeline that potentially could help for the detection of other neuropeptides.

Weaknesses:

Some of the interpretations are not fully supported by the existing data or would require further testing to draw those conclusions. This can be addressed by appropriately tampering down interpretations and acknowledging other limitations the authors did not cover brought by procedural differences between experiments.

We have taken time to go through the manuscript ensuring we are more detailed and precise with our interpretations as well as appropriately acknowledging limitations. 

Reviewer #2 (Public Review):

Thank you for your constructive and thorough assessment of our work. In our revised manuscript, we adjusted the text to reflect the references you mentioned regarding the methionine oxidation procedure. Additionally, we expanded the methods section to include the key details of the statistical tests and procedures that you outlined. 

Summary:

The authors aimed to improve the detection of enkephalins, opioid peptides involved in pain modulation, reward, and stress. They used optogenetics, microdialysis, and mass spectrometry to measure enkephalin release during acute stress in freely moving rodents. Their study provided better detection of enkephalins due to the implementation of previously reported derivatization reaction combined with improved sample collection and offered insights into the dynamics and relationship between Met- and Leu-Enkephalin in the Nucleus Accumbens shell during stress.

Strengths:

A strength of this work is the enhanced opioid peptide detection resulting from an improved microdialysis technique coupled with an established derivatization approach and sensitive and quantitative nLC-MS measurements. These improvements allowed basal and stimulated peptide release with higher temporal resolution, lower detection thresholds, and native-state endogenous peptide measurement.

Weaknesses:

The draft incorrectly credits itself for the development of an oxidation method for the stabilization of Met- and Leu-Enk peptides. The use of hydrogen peroxide reaction for the oxidation of Met-Enk in various biological samples, including brain regions, has been reported previously, although the protocols may slightly vary. Specifically, the manuscript writes about "a critical discovery in the stabilization of enkephalin detection" and that they have "developed a method of methionine stabilization." Those statements are incorrect and the preceding papers that relied on hydrogen peroxide reaction for oxidation of Met-Enk and HPLC for quantification of oxidized Enk forms should be cited. One suggested example is Finn A, Agren G, Bjellerup P, Vedin I, Lundeberg T. Production and characterization of antibodies for the specific determination of the opioid peptide Met5-Enkephalin-Arg6-Phe7. Scand J Clin Lab Invest. 2004;64(1):49-56. doi: 10.1080/00365510410004119. PMID: 15025428.

Thank you for highlighting this. It was not our intention to imply that we developed the oxidation method, rather that we were able improve the detection of metenkephalin by oxidation of the methionine without compromising the detection resolution of leu-enkephalin, enabling the simultaneous detection of both peptides. We have addressed this is the manuscript and included the suggested citation. 

Another suggestion for this draft is to make the method section more comprehensive by adding information on specific tools and parameters used for statistical analysis:

(1) Need to define "proteomics data" and explain whether calculations were performed on EIC for each m/z corresponding to specific peptides or as a batch processing for all detected peptides, from which only select findings are reported here. What type of data normalization was used, and other relevant details of data handling? Explain how Met- and Leu-Enk were identified from DIA data, and what tools were used.

Thank you for pointing out this source of confusion. We believe it is because we use a different DIA method than is typically used in other literature. Briefly, we use a DIA method with the targeted inclusion list to ensure MS2 triggering as opposed to using large isolation widths to capture all precursors for fragmentation, as is typically done with MS1 features. For our method, MS2 is triggered based on the 4 selected m/z values (heavy and light versions of Leu and Met-Enkephalin peptides) at specific retention time windows with isolation width of 2 Da; regardless of the intensity of MS1 of the peptides. 

(2) Simple Linear Regression Analysis: The text mentions that simple linear regression analysis was performed on forward and reverse curves, and line equations were reported, but it lacks details such as the specific variables being regressed (although figures have labels) and any associated statistical parameters (e.g., R-squared values). 

Additional detail about the linear regression process was added to the methods section, please see lines 614-618. The R squared values are also now shown on the figure. 

‘For the forward curves, the regression was applied to the measured concentration of the light standard as the theoretical concentration was increased. For plotting purposes, we show the measured peak area ratios for the light standards in the forward curves. For the reverse curves, the regression was applied to the measured concentration of the heavy standard, as the theoretical concentration was varied.’

(3) Violin Plots: The proteomics data is represented as violin plots with quartiles and median lines. This visual representation is mentioned, but there is no detail regarding the software/tools used for creating these plots.

We used Graphpad Prism to create these plots. This detail has been added to the statistical analysis section. See line 630.

(4) Log Transformation: The text states that the data was log-transformed to reduce skewness, which is a common data preprocessing step. However, it does not specify the base of the logarithm used or any information about the distribution before and after transformation.

We have added the requested details about the log transformation, and how the data looked before and after, into the statistical analysis section. We followed convention that the use of log is generally base 10 unless otherwise specified as natural log (base 2) or a different base. See lines 622-625

‘The data was log10 transformed to reduce the skewness of the dataset caused by the variable range of concentrations measured across experiments/animals. Prior to log transformation, the measurements failed normality testing for a Gaussian distribution. After the log transformation, the data passed normality testing, which provided the rationale for the use of statistical analyses that assume normality.’

(5) Two-Way ANOVA: Two-way ANOVA was conducted with peptide and treatment as independent variables. This analysis is described, but there is no information regarding the software or statistical tests used, p-values, post-hoc tests, or any results of this analysis.

Information about the two-way ANOVA analysis has been added to the statistical analysis section. Additionally, more detailed information has been added to the figure legends about the statistical results. Please see lines 625-628.

‘Two-way ANOVA testing with peptide (Met-Enk or Leu-Enk) and treatment (buffer or stress for example) as the two independent variables. Post-hoc testing was done using Šídák's multiple comparisons test and the p values for each of these analyses are shown in the figures (Figs. 1F, 2A).’ 

(6) Paired T-Test: A paired t-test was performed on predator odor proteomic data before and after treatment. This step is mentioned, but specific details like sample sizes, and the hypothesis being tested are not provided.

The sample size is included in the figure legend to which we have included a reference. We have also included the following text to highlight the purpose of this test. See lines 628-630

A paired t-test was performed on the predator odor proteomic data before and after odor exposure to test that hypothesis that Met-Enk increases following exposure to predator odor  (Fig. 3F). These analyses were conducted using Graphpad Prism.

(7) Correlation Analysis: The text mentions a simple linear regression analysis to correlate the levels of Met-Enk and Leu-Enk and reports the slopes. However, details such as correlation coefficients, and p-values are missing.

We apologize for the use of the word correlation as we think it may have caused some confusion and have adjusted the language accordingly. Since this was a linear regression analysis, there is no correlation coefficient. The slope of the fitted line is reported on the figures to show the fitted values of Met-Enk to Leu-Enk. 

(8) Fiber Photometry Data: Z-scores were calculated for fiber photometry data, and a reference to a cited source is provided. This section lacks details about the calculation of zscores, and their use in the analysis. 

These details have been added to the statistical analysis section. See lines 634-637

‘For the fiber photometry data, the z-scores were calculated as described in using GuPPy which is an open-source python toolbox for fiber photometry analysis. The z-score equation used in GuPPy is z=(DF/F-(mean of DF/F)/standard deviation of DF/F) where F refers to fluorescence of the GCaMP6s signal.’

(9) Averaged Plots: Z-scores from individual animals were averaged and represented with SEM. It is briefly described, but more details about the number of animals, the purpose of averaging, and the significance of SEM are needed.

We have added additional information about the averaging process in the statistical analysis section. See lines 639-643.

‘The purpose of the averaged traces is to show the extent of concordance of the response to experimenter handling and predator odor stress among animals with the SEM demonstrating that variability. The heatmaps depict the individual responses of each animal. The heatmaps were plotted using Seaborn in Python and mean traces were plotted using Matplotlib in Python.’

A more comprehensive and objective interpretation of results could enhance the overall quality of the paper.

We have taken this opportunity to improve our manuscript following comments from all the reviewers that we hope has resulted in a manuscript with a more objective interpretation of results. 

Reviewer #3 (Public Review):

Thank you for your thoughtful review of our work. To clarify some of the points you raised, we revised the manuscript to include more detail on how we distinguish between the oxidized endogenous and standard signal, as well as refine the language concerning the spatial resolution. We also edited the manuscript regarding the concentration measurements. We conducted technical replicates, so we appreciate you raising this point and clarify that in the main text. 

Summary:

This important paper describes improvements to the measurement of enkephalins in vivo using microdialysis and LC-MS. The key improvement is the oxidation of met- to prevent having a mix of reduced and oxidized methionine in the sample which makes quantification more difficult. It then shows measurements of enkephalins in the nucleus accumbens in two different stress situations - handling and exposure to predator odor. It also reports the ratio of released met- and leu-enkephalin matching what is expected from the digestion of proenkephalin. Measurements are also made by photometry of Ca2+ changes for the fox odor stressor. Some key takeaways are the reliable measurement of met-enkephalin, the significance of directly measuring peptides as opposed to proxy measurements, and the opening of a new avenue into the research of enkephalins due to stress based on these direct measurements.

Strengths:

-Improved methods for measurement of enkephalins in vivo.

-Compelling examples of using this method.

-Opening a new area of looking at stress responses through the lens of enkephalin concentrations.

Weaknesses:

(1) It is not clear if oxidized met-enk is endogenous or not and this method eliminates being able to discern that.

We clarified our wording in the text copied below to provide an explanation on how we distinguish between the two. Even after oxidation, the standard signal has a higher m/z ratio due to the presence of the Carbon and Nitrogen isotopes as described in the Chemicals section of the methods ‘For Met Enkephalin, a fully labeled L-Phenylalanine (¹³C₉, ¹⁵N) was added (YGGFM). The resulting mass shift between the endogenous (light) and heavy isotope-labeled peptide are 7Da and 10Da, respectively.’, so they can still be differentiated from the endogenous signal. We have clarified the language in the results section. See lines 82-87. 

‘After each sample collection, we add a consistent known concentration of isotopically labeled internal standard of Met-Enk and Leu-Enk of 40 amol/sample to the collected ISF for the accurate identification and quantification of endogenous peptide. These internal standards have a different mass/charge (m/z) ratio than endogenous Met- and Leu-Enk. Thus, we can identify true endogenous signal for Met-Enk and Leu-Enk (Suppl Fig. 1A,C) versus noise, interfering signals, and standard signal (Suppl. Fig. 1B,D).’

(2) It is not clear if the spatial resolution is really better as claimed since other probes of similar dimensions have been used.

Apologies for any confusion here. To clarify we primarily state that our approach improves temporal resolution and in a few cases refer to improved spatiotemporal resolution, which we believe we show. The dimensions of the microdialysis probe used in these experiments allow us to target the nucleus accumbens shell and as well as being smaller – especially at the membrane level - than a fiber photometry probe. 

(3) Claims of having the first concentration measurement are not quite accurate.

Thank you for your feedback. To clarify, we do not claim that we have the first concentration measurements, rather we are the first to quantify the ratio of Met-Enk to Leu-Enk in vivo in freely behaving animals in the NAcSh. 

(4) Without a report of technical replicates, the reliability of the method is not as wellevaluated as might be expected.

We have added these details in the methods section, please see lines 521-530. 

‘Each sample was run in two technical replicates and the peak area ratio was averaged before concentration calculations of the peptides were conducted. Several quality control steps were conducted prior to running the in vivo samples. 1) Two technical replicates of a known concentration were injected and analyzed – an example table from 4 random experiments included in this manuscript is shown below. 2) The buffers used on the day of the experiment (aCSF and high K+ buffer) were also tested for any contaminating Met-Enk or Leu-Enk signals by injecting two technical replicates for each buffer. Once these two criteria were met, the experiment was analyzed through the system. If either step failed, which happened a few times, the samples were frozen and the machines were cleaned and restarted until the quality control measures were met.’

Recommendations For The Authors:

Reviewer #1 (Recommendations For The Authors):

• The authors should provide appropriate citations of a study that has validated the Enkephalin-Cre mouse line in the nucleus accumbens or provide verification experiments if they have any available.

Thank you for your comment. We have added a reference validating the Enk-Cre mouse line in the nucleus accumbens to the methods section and is copied here. 

D.C. Castro, C.S. Oswell, E.T. Zhang, C.E. Pedersen, S.C. Piantadosi, M.A. Rossi, A.C. Hunker, A. Guglin, J.A. Morón, L.S. Zweifel, G.D. Stuber, M.R. Bruchas, An endogenous opioid circuit determines state-dependent reward consumption, Nature 2021 598:7882 598 (2021) 646–651. https://doi.org/10.1038/s41586-02104013-0.

• Better definition of the labels y1,y2,b3 in Figures 1 and S1 would be useful. I may have missed it but it wasn't described in methods, results, or legends.

Thank you for this comment. We have added this information to Fig.1 legend ‘Y1, y2, b3 refer to the different elution fragments resulting from Met-Enk during LC-MS.

• It is interesting that the ratio of KCl-evoked release is what changes differentially for Met- vs Leu. Leu enk increases to the range of met-enk. There is non-detectable or approaching being non-detectable leu-enk (below the 40 amol / sample limit of quantification) in most of the subjects that become apparent and approach basal levels of met-enkephalin. This suggests that the K+ evoked response may be more pronounced for leu-enk. This is something that should be considered for further analysis and should be discussed.

Thank you for this astute observation, and you make a great point. We have added some discussion of this finding in the results and discussion sections see lines 111112 and lines 253-257. 

‘Interestingly, Leu-Enk showed a greater fold change compared to baseline than did Met-Enk with the fold changes being 28 and 7 respectively based on the data in Fig.1F.’

‘We also noted that Leu-Enk showed a greater fold increase relative to baseline after depolarization with high K+ buffer as compared to Met-Enk. This may be due to increased Leu-Enk packaging in dense core vesicles compared to Met-Enk or due to the fact that there are two distinct precursor sources for Leu-Enk, namely both proenkephalin and prodynorphin while Met-Enk is mostly cleaved from proenkephalin (see Table 1 [48]).’

• For example in 2E, it would be helpful to label in the graph axis what samples correspond to the manipulation and also in the text provide the reader with the sample numbers. The authors interpret the relationship between the last two samples of baseline and posthandling stress as the following in the figure legend "the concentration released in later samples is affected; such influence suggests that there is regulation of the maximum amount of peptide to be released in NAcSh. E. The negative correlation in panel d is reversed by using a high K+ buffer to evoke Met-Enk release, suggesting that the limited release observed in D is due to modulation of peptide release rather than depletion of reserves." However, the correlations are similar between 2D and E and it appears that two mice are mediating the difference between the two groups. The appropriate statistical analysis would be to compare the regressions of the two groups. Statistics for the high K+ (and all other graphs where appropriate) need to be reported, including the r2 and p-value.

Thank you for your constructive critique. To elucidate the effect of high K+, we have plotted the regression line and reported the slope for Fig. 2E. Notably, the slope is reduced by a factor of 2 and appears to be driven by a large subset of the animals. The statistics for the high K+ graph are shown on the figure (Fig 1F) which test the hypothesis of whether high K+ leads to the release of Leu-Enk and Met-Enk respectively compared to baseline with aCSF. We have added the test statistics to the figure legend for additional clarity. Fig. 1G has no statistics because it is only there to elucidate the ratio between Met-Enk and Leu-Enk in the same samples. We did not test any hypotheses related to whether there are differences between their levels as that is not relevant to our question. The correlation on the same data is depicted in Fig. 1H, and we have added the R² value per your request. 

• The interpretation that handling stress induces enkephalin release from microdialysis experiments is also confounded by other factors. For instance, from the methods, it appears that mice were connected and sample collection started 30 min after surgery, therefore recovery from anesthesia is also a confounding variable, among other technical aspects, such as equilibration of the interstitial fluid to the aCSF running through the probe that is acting as a transmitter and extracellular molecule "sink". Did the authors try to handle the mice post hookup similar to what was done with photometry to have a more direct comparison to photometry experiments? This procedural difference, recording from recently surgerized animals (microdialysis) vs well-recovered animals with photometry should be mentioned in addition to the other caveats the authors mention.

Thank you for your comment. We are aware of this technical limitation, and it is largely why we sought to conduct the fiber photometry experiments to get at the same question. As you requested, we have included additional language in the discussion to acknowledge this limitation and how we chose to address it by measuring calcium activity in the enkephalinergic neurons, which would presumably be the same cell population whose release we are quantifying using microdialysis. See lines 262-273.  

‘Our findings showed a robust increase in peptide release at the beginning of experiments, which we interpreted as due to experimenter handling stress that directly precedes microdialysis collections. However, there are other technical limitations to consider such as the fact that we were collecting samples from mice that were recently operated on. Another consideration is that the circulation of aCSF through the probe may cause a sudden shift in oncotic and hydrostatic forces, leading to increased peptide release to the extracellular space. As such, we wanted to examine our findings using a different technique, so we chose to record calcium activity from enkephalinergic neurons - the same cell population leading to peptide release. Using fiber photometry, we showed that enkephalinergic neurons are activated by stress exposure, both experimenter handling and fox odor, thereby adding more evidence to suggest that enkephalinergic neurons are activated by stress exposure which could explain the heightened peptide levels at the beginning of microdialysis experiments.’

• The authors should provide more details on handling stress manipulation during photometry. For photometry what was the duration of the handling bout, what was the interval between handling events, and can the authors provide a description of what handling entailed? Were mice habituated to handling days before doing photometry recording experiments?

Thank you for your suggestion. We have addressed all of your points in the methods section. See lines 564-570. 

‘The handling bout which mimicked traditional scruffing lasted about 3-5 seconds. The mouse was then let go and the handling was repeated another two times in a single session with a minimum of 1-2 minutes between handling bouts. Mice were habituated to this manipulation by being attached to the fiber photometry rig, for 3-5 consecutive days prior to the experimental recording. Additionally, the same maneuver was employed when attaching/detaching the fiber photometry cord, so the mice were subjected to the same process several times.’

• For the novel weigh boat experiments, the authors should explicitly state when these experiments were done in relation to the fox urine, was it a different session or the same session? Were they the same animals? Statements like the following (line 251) imply it was done in the same animals in the same session but it should be clarified in the methods "We also showed using fiber photometry that the novelty of the introduction of a foreign object to the cage, before adding fox odor, was sufficient to activate enkephalinergic neurons."

As shown in supplementary figure 4, individual animal data is shown for both water and fox urine exposure (overlaid) to depict whether there were differences in their responses to each manipulation – in the same animal. And yes, you are correct, the animals were first exposed to water 3 times in the recording session and then exposed to fox urine 3 times in the same session. We have added that to the methods section describing in vivo fiber photometry. See lines 575-576.  

• Statistical testing would be needed to affirm the conclusions the authors draw from the fox urine and novel weigh boat experiments. For example, it shows stats that the response attenuates, that it is not different between fox urine and novel (it looks like the response is stronger to the fox urine when looking at the individual animals), etc. These data look clear but stats are formally needed. Formal statistics are also missing in other parts of the manuscript where conclusions are drawn from the data but direct statistical comparisons are not included (e.g. Fig 2.G-I).

The photometry data is shown as z-scores which is a formal statistical analysis. ANOVA would be inappropriate to run to compare z-scores. We understand that this is erroneously done in fiber photometry literature, however, it remains incorrect. The z-scores alone provide all the information needed about the deviation from baseline. We understand that this is not immediately clear to readers, and we thank you for allowing us to explain why this is the case. We have added test statistics to figure legends where hypothesis testing was done and p-values were reported. 

• Did the authors try to present the animals with repeated fox urine exposure to see if this habituates like the photometry?

No, we did not do that experiment due to the constrained timing within which we had to run our microdialysis/LC-MS timeline, but it is a great point for future exploration. 

• It would be useful to present the time course of the odor experiment for the microdialysis experiment.

The timeline is shown in Fig.1a and Fig.3e. To reiterate, each sample is 13 minutes long.

• Can the authors determine if differences in behavior (e.g. excessive avoidance in animals with with one type of response) or microdialysis probe location dictate whether animals fall into categories of increased release, no release, or no-detection? From the breakdown, it looks like it is almost equally split into three parts but the authors' descriptions of this split are somewhat misleading (line 210). " The response to predator odor varies appreciably: although most animals show increased Met-Enk release after fox odor exposure, some show continued release with no elevation in Met-Enk levels, and a minority show no detectable release".

Thank you for your constructive feedback. We do not believe the difference in behavior is correlated with probe placement. The hit map can be found in suppl. Fig 3 and shows that all mice included in the manuscript had probes in the NAcSh. We purposely did not distinguish between dorsal and ventral because of our 1 mm membrane would make it hard to presume exclusive sampling from one subregion. That is a great point though, and we have thought about it extensively for future studies. We have edited the language to reflect the almost even split of responses for Met-Enk and appreciate you pointing that out. 

• Overall, given the inconsistencies in experimental design and overall caveats associated, I think the authors are unable to draw reasonable conclusions from the repeated stressor experiments and something they should either consider is not trying to draw strong conclusions from these observations or perform additional experiments that provide the grounds to derive those conclusions.

We have included additional language on the caveats of our study, and our use of a dual approach using fiber photometry and microdialysis was largely driven by a

desire to offer additional support of our conclusions. We expected pushback about our conclusions, so we wanted to offer a secondary analysis using a different technique to test our hypothesis. To be honest the tone of this comment and content is not particularly constructive (especially for trainees) nor does it offer a space to realistically address anything. This work took multiple years to optimize, it was led by a graduate student, and required a multidisciplinary team. As highlighted, we believe it offers an important contribution to the literature and pushes the field of peptide detection forward.  

Reviewer #2 (Recommendations For The Authors):

A more comprehensive and objective interpretation of results could enhance the overall quality of the paper. The manuscript contains statements like "we are the first to confirm," which can be challenging to substantiate and may not significantly enhance the paper. It's essential to ensure that novelty statements are well-founded. For example, the release of enkephalins from other brain regions after stress exposure is well-documented but not addressed in the paper. Similarly, the role of the NA shell in stress has been extensively studied but lacks coverage in this manuscript.

We have edited the language to reflect your feedback. We have also included relevant literature expanding on the demonstrated roles of enkephalins in the literature. We would like to note that most studies have focused on chronic stress, and we were particularly interested in acute stress. See lines 129-134.

‘These studies have included regions such as the locus coeruleus, the ventral medulla, the basolateral nucleus of the amygdala, and the nucleus accumbens core and shell. Studies using global knockout of enkephalins have shown varying responses to chronic stress interventions where male knockout mice showed resistance to chronic mild stress in one study, while another study showed that enkephalin-knockout mice showed delayed termination of corticosteroid release. [33,34]’ 

Finally, not a weakness but a clarification suggestion: the method description mentions the use of 1% FA in the sample reconstitution solution and LC solvents, which is an unusually high concentration of acid. If this concentration is intentional for maintaining the peptides' oxidation state, it would be beneficial to mention this in the text to assist readers who might want to replicate the method.

This is correct and has been clarified in the methods section

Reviewer #3 (Recommendations For The Authors):

-The Abstract should state the critical improvements that are made. Also, quantify the improvements in spatiotemporal resolution.

Thank you for your comment. We have edited the abstract to reflect this. 

- The use of "amol/sample" as concentration is less informative than an SI units (e.g., pM concentration) and should be changed. Especially since the volume used was the same for in vivo sampling experiments.

Thank you for your comment. We chose to report amol/sample because we are measuring such a small concentration and wanted to account for any slight errors in volume that can make drastic differences on reported concentrations especially since samples are dried and resuspended.  

-Please check this sentence: "After each collection, the samples were spiked with 2 µL of 12.5 fM isotopically labeled Met-Enkephalin and Leu-Enkephalin" This dilution would yield a concentration of ~2 fM. In a 12 uL sample, that would be ~0.02 amol, well below the detection limit. (note that fM would femtomolar concentration and fmol would be femtomoles added).

-"liquid chromatography/mass spectrometry (LC-MS) [9-12]"... Reference 9 is a RIA analysis paper, not LC-MS as stated.

Thank you for catching these. We have corrected the unit and citation. 

-Given that improvements in temporal resolution are claimed, the lack of time course data with a time axis is surprising. Rather, data for baseline and during treatment appear to be combined in different plots. Time course plots of individuals and group averages would be informative.

Due to the expected variability between individual animal time course data, where for example, we measure detectable levels in one sample followed by no detection, it was very difficult to combine data across time. Therefore, to maximize data inclusion from all animals that showed baseline measurements and responses to individual manipulations, we opted to report snapshot data. Our improvement in temporal resolution refers to the duration of each sample rather than continuous sampling, so those two are unrelated. Thank you for your feedback and allowing us to clarify this.

- I do not understand this claim "We use custom-made microdialysis probes, intentionally modified so they are similar in size to commonly used fiber photometry probes to avoid extensive tissue damage caused by traditional microdialysis probes (Fig. 1B)." The probes used are 320 um OD and 1 mm long. This is not an uncommon size of microdialysis probes and indeed many are smaller, so is their probe really causing less damage than traditional probes?

Thank you for your comment. We are only trying to make the point that the tissue damage from these probes is comparable to commonly used fiber photometry probes. We only point that out because tissue damage is used as a point to dissuade the usage of microdialysis in some literature, and we just wanted to disambiguate that. We have clarified the statement you pointed out.  

-The oxidation procedure is a good idea, as mentioned above. It would be interesting to compare met-enk with and without the oxidation procedure to see how much it affects the result (I would not say this is necessary though). It is not uncommon to add antioxidants to avoid losses like this. Also, it should be acknowledged that the treatment does prevent the detection of any in vivo oxidation, perhaps that is important in met-enk metabolism?

The comparison between oxidized and unoxidized Met-Enk detection is in figure 1C. 

-It would be a best practice to report the standard deviation of signal for technical replicates (say near in vivo concentrations) of standards and repeated analysis of a dialysate sample to be able to understand the variability associated with this method. Similarly, an averaged basal concentration from all rats.

Thank you for your comment. We have included a table showing example quality control standard injections from 4 randomly selected experiments included in the manuscript that were run before and after each experiment and descriptive statistics associated with these technical replicates. We also added some detail to the methods section to describe how quality control is done. See lines 521-530. 

‘Each sample was run in two technical replicates and the peak area ratio was averaged before concentration calculations of the peptides were conducted. Several quality control steps were conducted prior to running the in vivo samples. 1) Two technical replicates of a known concentration were injected and analyzed – an example table from 4 random experiments included in this manuscript is shown below. 2) The buffers used on the day of the experiment (aCSF and high K+ buffer) were also tested for any contaminating Met-Enk or Leu-Enk signals by injecting two technical replicates for each buffer. Once these two criteria were met, the experiment was analyzed through the system. If either step failed, which happened a few times, the samples were frozen and the machines were cleaned and restarted until the quality control measures were met.’

EDITORS NOTE

Should you choose to revise your manuscript, please include full statistical reporting including exact p-values wherever possible alongside the summary statistics (test statistic and df) and 95% confidence intervals. These should be reported for all key questions and not only when the p-value is less than 0.05.

Thank you for your suggestion. We have included more detail about statistical analysis in the figure legends per this comment and reviewer comments.

eLife Assessment

This study provides valuable observations indicating that human pyramidal neurons propagate information as fast as rat pyramidal neurons despite their larger size. Convincing evidence demonstrates that this property is due to several biophysical properties of human neurons. This study will be of interest to neurophysiologists.

Reviewer #1 (Public review):

The propagation of electrical signals within neuronal circuits is tightly regulated by the physical and molecular properties of neurons. Since neurons vary in size across species, the question arises whether propagation speed also varies to compensate for it. The present article compares numerous speed-related properties in human and rat neurons. They found that the larger size of human neurons seems to be compensated by a faster propagation within dendrites but not axons of these neurons. The faster dendritic signal propagation was found to arise from wider dendritic diameters and greater conductance load in human neurons. In addition, the article provides a careful characterization of human dendrites and axons, as the field has only recently begun to characterize post-operative human cells. There are only a few studies reporting dendritic properties and these are not all consistent, hence there is added value of reporting these findings, particularly given that the characterization is condensed in a compartmental model.

Strengths

The study was performed with great care using standard techniques in slice electrophysiology (pharmacological manipulation with somatic patch-clamp) as well as some challenging ones (axonal and dendritic patch-clamp). Modeling was used to parse out the role of different features in regulating dendritic propagation speed. The finding that propagation speed varies across species is novel as previous studies did not find a large change in membrane time constant nor axonal diameters (a significant parameter affecting speed). A number of possible, yet less likely factors were carefully tested (Ih, membrane capacitance). The main features outlined here are well known to regulate speed in neuronal processes. The modeling was also carefully done to verify that the magnitude of the effects is consistent with the difference in biophysical properties. Hence, the findings appear very solid to me.

Weaknesses

The role of diameter in regulating propagation speed is well known in the axon literature.

Comment on the revised version: the authors have now made clearer that the role of diameter was well known in the manuscript.

Reviewer #2 (Public review):

Summary:

In this paper, Oláh and colleagues introduce new research data on the cellular and biophysical elements involved in transmission within the pyramidal circuits of the human neocortex. They gathered a comprehensive set of patch-clamp recordings from human and rat pyramidal neurons to compare how the temporal aspect of neuronal processing is maintained in the larger human neocortex. A range of experimental techniques have been used, including two-photon guided dual whole-cell recordings, electron microscopy, complemented by theoretical and computational methods.

The authors find that synaptically connected pyramidal neurons within the human neocortex have longer intercellular path lengths. They go on to show that the short soma to soma latencies is not due to propagation velocity along the axon but instead reflects a higher propagation speed of synaptic potentials from dendrite to soma. Next, in a series of extensive computational modeling studies focusing on the synaptic potentials, the authors show that the shorter latency may be explained by larger diameters, affecting the cable properties and resulting is relatively faster propagation of EPSPs in the human neuron. The manuscript is well-written, and the physiological experiments and in-depth theoretical steps for the simulations are clear. Whether passive cable properties of the dendrites alone are responsible for higher velocities remains to be further investigated. Based on the present data the contribution of active membrane properties cannot be excluded.

Strengths:

The authors used complex 2P-guided dual whole-cell recordings in human neurons. In combination with detailed reconstructions, these approaches represent the next steps in unravelling the information processing in human circuits.

The computational modelling and cable theory application to the experimentally constrained simulations provide an integrated view of the passive membrane properties of human neurons.

Weaknesses:

There are concerns with the statistical analyses of the experimental data. The two-way analyses are not supporting that the backpropagation speed in human neurons is more affected by TTX-induced or after ZD remains higher. Significance of interaction is required, and the authors make errors in the interpretation and application of separate additional t-tests. Whether the cable properties alone are the main explanation for speeding the electrical signaling in human pyramidal neurons deserves further studies.

Comments on the latest version:

In my previous review I suggested the author read upon the need to perform two-way ANOVA for their experiments. Although I am glad this has now been done, I'm surprised to read the interpretation remains flawed and we are not provided with all the analyses. We need to know all the covariates and results of the post-hoc comparisons. What is written in the results is incomplete.

One cannot perform two-way ANOVA and subsequently performing t-tests on computed differences. Figures 3C and F are irrelevant. Instead, we need to know the Bonferroni post-hoc results for all comparisons.

If there is an interaction significance then the authors will have to conclude the Na+ channels have a larger contribution to the backpropagation.

Line187 "It therefore be argued that HCN channels may contribute to the higher conduction velocities in human dendrites, but do not by themselves explain the differences between the two species."

One wonders whether supplementary figures are required.

Reviewer #3 (Public review):

Summary:

This study indicates that connections across human cortical pyramidal cells have identical latencies despite a larger mean dendritic and axonal length between somas in human cortex. A precise demonstration combining detailed electrophysiology and modeling, indicates that this property is due to faster propagation of signals in proximal human dendrites. This faster propagation is itself due to a slightly thicker dendrite, to a larger capacitive load, and to stronger hyperpolarizing currents. Hence, the biophysical properties of human pyramidal cells are adapted such that they do not compromise information transfer speed.

Strengths:

The manuscript is clear and very detailed. The authors have experimentally verified a large number of aspects that could affect propagation speed and have pinpointed the most important one. This paper provides an excellent comparision of biophysical properties between rat and human pyramidal cells. Thanks to this approach a comprehensive description of the mechanisms underlying the acceleration of propagation in human dendrite is provided.

Weaknesses:

The weaknesses I had identified have been addressed by the authors.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

The propagation of electrical signals within neuronal circuits is tightly regulated by the physical and molecular properties of neurons. Since neurons vary in size across species, the question arises whether propagation speed also varies to compensate for it. The present article compares numerous speed-related properties in human and rat neurons. They found that the larger size of human neurons seems to be compensated by a faster propagation within dendrites but not the axons of these neurons. The faster dendritic signal propagation was found to arise from wider dendritic diameters and greater conductance load in human neurons. In addition, the article provides a careful characterization of human dendrites and axons, as the field has only recently begun to characterize post-operative human cells. There are only a few studies reporting dendritic properties and these are not all consistent, hence there is the added value of reporting these findings, particularly given that the characterization is condensed in a compartmental model.

Strengths:

The study was performed with great care using standard techniques in slice electrophysiology (pharmacological manipulation with somatic patch-clamp) as well as some challenging ones (axonal and dendritic patch-clamp). Modeling was used to parse out the role of different features in regulating dendritic propagation speed. The finding that propagation speed varies across species is novel as previous studies did not find a large change in membrane time constant or axonal diameters (a significant parameter affecting speed). A number of possible, yet less likely factors were carefully tested (Ih, membrane capacitance). The main features outlined here are well-known to regulate speed in neuronal processes. The modeling was also carefully done to verify that the magnitude of the effects is consistent with the difference in biophysical properties. Hence, the findings appear very solid to me.

Weaknesses:

The role of diameter in regulating propagation speed is well-known in the axon literature.

We thank the reviewer for this comment. This is indeed true. The paper does not claim that this is new – we just refereed to Waxman’s book to acknowledge this established effect. Our main emphasize is on the impact of dendritic (rather than axonal) diameter – highlighting the faster EPSP speed near the input synapse and converging to steady-state value further away from the soma and using this to explore the impact of differences in dendritic diameter of rat vs. human on EPSP latency and velocity. We now made this point clearer in the revised text.

Reviewer #2 (Public Review):

Summary:

In this paper, Oláh and colleagues introduce new research data on the cellular and biophysical elements involved in transmission within the pyramidal circuits of the human neocortex. They gathered a comprehensive set of patch-clamp recordings from human and rat pyramidal neurons to compare how the temporal aspect of neuronal processing is maintained in the larger human neocortex. A broad range of experimental, theoretical, and computational methods are used, including two-photon guided dual whole-cell recordings, electron microscopy, and computational simulations of reconstructed neurons.

Recordings from synaptically connected pyramidal neurons revealed longer intercellular path lengths within the human neocortex. Further, by using dual whole-cell recordings from somadendrite and soma-axon locations, they found that short latencies from soma to soma can be partly attributed to an increased propagation speed for synaptic potentials, but not for the propagation of action potentials along the axon.

Next, in a series of extensive computational modeling studies focusing on the synaptic potentials, the authors observe that the short-latency within large human pyramidal neural circuits may have a passive origin. For a wide array of local synaptic input sites, the authors show that the conductance load of the dendrites, electrically coupled to a large diameter apical dendrite, affects the cable properties. The result is a relatively faster propagation of EPSPs in the human neuron.

The manuscript is well-written and the physiological experiments and biophysical arguments are very well explained. I appreciated the in-depth theoretical steps for the simulations. That passive cable properties of the dendrites are causing a higher velocity in human dendrites is interesting but there is a disconnect between the experimental findings and the model simulations. Based on the present data the contribution of active membrane properties cannot be dismissed and deserves further experiments.

See our response below

Strengths:

The authors present state-of-the-art 2P-guided dual whole-cell recordings in human neurons. In combination with detailed reconstructions, these approaches represent the next steps in unravelling the information processing in human circuits.

The computational modeling based on cable theory and experimentally constrained simulations provides an excellent integrated view of the passive membrane properties.

Weaknesses:

There are smaller and larger issues with the statistical analyses of the experimental data which muddles the interim conclusions.

That the cable properties alone are the main explanation for speeding the electrical signaling in human pyramidal neurons appears inconsistent with the experimental data.

This is an excellent point – we indeed performed analysis on only passive cases – highlighting (and now also ranking) the impact of the various morpho-electrical properties of the neurons on the differences in signal latency in human vs. rats. We did explored (not shown) the effect of active channels in the dendrites (including the h-current); as expected the results strongly depend on channel density and their spatial distribution over the dendritic tree. As we do not know these parameters for the modelled cells, we decided to remain focus on the impact of passive/morphological parameters. We also note that the experimental results (page 4-5 in manuscript) show minor contribution of h-current emphasizing that the passive properties have the main role in differentiating human and rats. differences between human and rat. 

Some of the electrophysiological experiments require further control experiments to make robust conclusions.

Reviewer #3 (Public Review):

Summary:

This study indicates that connections across human cortical pyramidal cells have identical latencies despite a larger mean dendritic and axonal length between somas in the human cortex. A precise demonstration combining detailed electrophysiology and modeling indicates that this property is due to faster propagation of signals in proximal human dendrites. This faster propagation is itself due to a slightly thicker dendrite, a larger capacitive load, and stronger hyperpolarizing currents. Hence, the biophysical properties of human pyramidal cells are adapted such that they do not compromise information transfer speed.

Strengths:

The manuscript is clear and very detailed. The authors have experimentally verified a large number of aspects that could affect propagation speed and have pinpointed the most important one. This paper provides an excellent comparison of biophysical properties between rat and human pyramidal cells. Thanks to this approach a comprehensive description of the mechanisms underlying the acceleration of propagation in human dendrite is provided.

Weaknesses:

Several aspects having an impact on propagation speed are highlighted (dendritic diameter, ionic channels, capacitive load) and there is no clear ranking of their impact on signal propagation speed. It seems that the capacitive load plays a major role, much more than dendritic diameter for which only a 10% increase is observed across species. Both aspects actually indicate that there is an increase in passive signal propagation speed with bigger cells at least close to the soma. This suggests that bigger cells are mechanically more rapid. An intuitive reason why capacitive load increases speed would also help the reader follow the demonstration.

We thank the referee for both these excellent points. In response to them:

(i) We now performed a new comprehensive statistical analysis and show the ranking of the effect of the different morphological/cable factors on EPSP propagation. This analysis appears in both Supp. Table 5& 6, Fig. S16 and also in the main text as follows:

To rank the impact of the various factors affecting EPSP propagation latency in human and rat neurons, we conducted a comprehensive statistical analysis using two complementary approaches: the generalized linear model (GLM) (Kiebel & Holmes, 2007) as well as SHAP (SHapley Additive exPlanations) (Lundberg & Lee, 2017) based on fitting Gradient Tree Boosting  (Friedman, 2002)model. We began by fitting a GLM without interaction terms among the factors affecting EPSP latency (Suppl. Table 5). This enables us to quantify the primary individual factors affecting EPSP propagation. Our analysis revealed the following ranking order: 1) physical distance of synapses from soma had the strongest effect; 2) species differences; 3) conductance load, as demonstrated by our “hybrid cells” manipulation; 4) radii of the apical dendrite, affecting the cables’ space constant, λ; and 5) the specific cable parameters, as revealed when using per-cell fitted parameters versus uniform cable parameters, was minimal. We next performed GLM analysis with interaction terms showing that, as expected, there are significant interactions between the factors affecting EPSP latency (Suppl. Table 6). To further validate the above ranking while incorporating the interactions between the various factors affecting EPSP latency, we performed a SHAP analysis. Notably, even with interactions included, the ranking of the factors affecting signal propagation are aligned with the results from the analysis based on the GLM without interaction terms (see Fig S.16).

(ii) As for the intuitive explanation required by the referee. We added the following paragraph In the Discussion:

The intuitive reason for this enhancement is that the large conductance load (the “leaky end” boundary conditions) more effectively “steals” the synaptic (axial) current (like water pouring faster into a large pool). The more mathematical intuition is that the large soma (sink) adds fast time constants to the system (see also related explanation in Fig. 4 in Eyal et al., 2014).

We thank the editors for considering and revising our manuscript for publication in eLife. We appreciate the positive appreciation of the work and the critical points raised by the reviewers. We have responded in detail to all the excellent comments from all reviewers. We believe that these revisions have significantly improved the quality of our study.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

There are two points that could improve the reading experience of this nice manuscript. These should be easily addressed with minor re-phrasing.

Credit to conduction velocity literature. Less widely known in the dendrite literature, in the axon literature, the relationship between propagation speed and process diameter is well established. I thought the two articles cited (Jack Noble Tsien and Agmon-Snir & Segev) were not as direct in the treatment of this relationship. The work of Stephen Waxman, for instance, made clear how axon diameter tightly controls propagation speed (see for instance the Scholarpedia entry by Swadlow and Waxman). In my opinion, this is a widely known piece of work, that is part of some introductory books to neuroscience. While the article does not claim they found this relationship, parts of the presentation are better understood if we ignore this well-known fact. I am referring to the abstract, intro, and the beginning of results where 'larger' is presented as synonymous with 'slower'. For instance 'to compensate for the increase neurons' size' (abstract) or 'the increase in size of dendrites and axons might come with a cost of longer signal propagation times' only makes sense if 'size' refers to spatial extent, not diameter.

We thank for this valid point; leaving out axon diameter references was not intentional. We have now added the suggested reference to our manuscript. In the size comparisons, we have only pointed out the obvious size differences between the body and the dendritic processes. We have reworded sentences with size comparisons.

In Abstract (lines 1-6):

Human-specific cognitive abilities depend on information processing in the cerebral cortex, where neurons are significantly larger, their processes are longer and sparser compared to rodents. We found that, in synaptically-connected layer 2/3 pyramidal cells (L2/3 PCs), soma-tosoma signal propagation delay is similar in humans and rodents. Thus, to compensate for the increase in neurons’s longer processes, membrane potential changes must propagate faster in human axons and/or dendrites.

In section “Effect of dendritic thickness” in Results we have modified it as follows:

The relationship between conduction velocity and axon diameter is well known for small myelinated and unmyelinated axons (Waxman and Bennett, 1972). Anatomical features of neuronal processes dendrites also have a major influence on signal propagation properties 5,19, thus …

Waxman, S. G. and Bennett, M. V. L. Relative conduction velocity of small myelinated and nonmyelinated fibres in the central nervous system. Nature New Biol., 238217-219, 1972.

Two or four dendritic factors? The study identifies two major dendritic factors influencing the propagation speed (diameter and load), however the end of the results highlights four factors. I did not understand how factor 2 was different than factor 1. Neither did I understand how factor 4 was different from the other factors. There seemed to be a little redundancy here that could be streamlined.

We thank the reviewer for pointing this out. We now have changes the respective text, added the ranking statistics (see above) to assess the effect of the different parameters on signal propagation in dendrites.

Microcircuits? The study found that the changes in speed arise from the dendrites rather than the axons, as such it seems it would be more precise to replace 'microcircuits' with 'dendrites'.

We are thankful for this suggestion. We change the title to Accelerated signal propagation speed in human neocortical dendrites.

Typos

P3 line 24 'find significant difference the propagation'.

P6 line 35 'how morphological differences' it would be useful to specify which morphological difference here.

Corrected.

Reviewer #2 (Recommendations For The Authors):

(1) The statistical analyses should be changed. T-testing populations and comparing visual differences of differences ("human minus rats") is a common but egregious error in the field of neurosciences (see doi:10.1038/nn.2886). The conclusion that HCN channels "... do not by themselves explained the differences between the two species" (lines 174-176) is not compelling. The design of the experiments presented in Figure 3 is paired recordings and the addition of a blocker (ZD7288 or TTX cocktail). These are classic 2 x 2 factorial designs (species x drug). The authors will need to perform a repeated-measured analysis of variance (RM-ANOVA) and provide information on the interaction significance. Please revise the figures and improve statistical reporting. Post-hoc comparisons of the velocity populations are required to support the idea of whether h-channels are explaining the observed differences.

Thank you for drawing our attention to this error. The statistical analysis of the pharmacological experiments was re-performed as suggested. After the 2-way ANOVA with repeated measures and Bonferroni post-hoc correction, we can indeed find significant differences only in the control group, namely that the propagation speed of bAPs in human dendrites was significantly higher. The implementation of the proposed statistical analysis demonstrates that the administration of ZD has no statistically significant effect on the propagation speed of human or rat dendrites. The treatment with TTX cocktail resulted in a significant difference in signal propagation in humans but not in rodents. However the trend is discernible and the P = 0.0588 value is close to the widely accepted 0.05 threshold. After the TTX cocktail treatment, the speed of signal propagation did not differ significantly between the two species. However, on average, the human dendrites remained faster. These alterations in P-values do not affect our primary conclusions. The MS text has been modified accordingly.

(2) Although ZD7288, in my opinion, influences the bAP (see point #1) the authors subsequently leave the h-current unblocked in the experiments in Figures 3D, E. Here, they use sodium, potassium, and calcium currents as well as synaptic conductances. I am puzzled why (in line 188) they claim the dendrites are "passive" although the data show h-currents are contributing to the shape of the bAP in human neurons. In line 196 they conclude voltage-gated conductances have a "minor" contribution and passive properties a main role. Please revise conclusions or provide better experimental support.

Thank you for this point. We meant to refer to the state in which no action potential can be generated, although the word 'passive' might be misleading in this context; we rephrase these sentences in the MS accordingly.

(3) A major concern is the injection of an AP in voltage-clamp mode. Although this is the right choice and I'm in support of the experiment, it is technically challenging to space clamp the soma and fully recapitulate the speed and amplitude of a 100 mV depolarization. The voltage drop in peak amplitude as well as the increased delay between the baseline AP (current clamp) and AP in blocker conditions (voltage clamp) could be fully explained by switching between current- and voltage-clamp modes. In additional control experiments, the authors should add a second voltage follower electrode (CC) at the soma showing whether the authors can preserve the original AP (from CC) in VC/blocker condition. It may well be they need to adjust the injection protocol.

Our experiments were designed to replicate the work of Stuart et al. (1994), in which they compared the attenuation of active and passive backpropagating signals. When they blocked Na+ channels with TTX they injected simulated action potentials in voltage-clamp mode. They concluded that TTX-sensitive Na+ channels cause somatic action potential entry into the dendritic compartment. They found a comparable attenuation of the backward propagating action potential in the dendrites in control conditions (~70 %). 

We performed control recordings based on the reviewer’s suggestion (Author response image 1).

Author response image 1.

Injection of the previously recorded AP (blue) in VC mode produced a completely similar somatic AP in CC mode (orange). The slight temporal delay between the two signal caused by the different position of the pipettes on the cell body.  The right panel shows the plot of the two peak-aligned APs as a function of each other, close to the blue ‘equality’ line. We concluded that the original AP is well preserved in VC/blocker condition.

(5) From the paragraph entitled "Modeling EPSP propagation in dendrites" and onwards the authors make countless conclusions based on theory and modelling results but without any statistical support. Multiple neurons are used thus it is rather straightforward to provide numerical support for the assertions. For example, but this is not an exhaustive list, how should we interpret that latency ranges are different (line 240, line 253) etc.? Or were the estimated Cm values of human and rat neurons (0.6 versus 1.1) significantly different? And if so, how does this align with the Cm estimates in the nucleated patch experiments?

We thank the referee for this comment and now added a set of statistical analyses. The results appear now throughout the whole theoretical paper in revised article. In particular with respect to Figs. 6&7 where we now show that, indeed, our various manipulations (e.g., hybrid vs. original cells) as well as the cable parameters (Cm, Rm) are indeed significantly different between human and rats whereas the membrane time constant is not significantly different between human and rat. As for Cm in human. Our limited sample size shows significant difference between human and rat. Yet, the range of values for Cm that we found in our modeling study does fall within the experimental range reported in the present study.

Minor

Line 44. The "simulated EPSP" example in Figure 2C is not a command waveform for an EPSC. Line 526 in the methods states that also ramp currents were used. Please revise to clarify the main text.

Thank you for bringing this discrepancy to our attention. In the experiments, we used ramp injections. We have made this clear in the main text as follows: ”... we tested orthodromic or forward propagating signal propagation velocity by injecting short-duration current ramps to simulate EPSP (sEPSP) signals in the dendrites and recorded the resultant subthreshold voltage response in the soma”

Line 522. The authors state the recordings were all carried out "in current clamp mode" but detailed VC method information is lacking. Did they use series resistance compensation?

We did not use series resistance compensation.

Line 479 From which region(s) where human "neocortical slices" sampled? Please add this information.

We have added regions of origin to the Methods section: frontal (n = 21), temporal (n = 20), parietal (n = 20), and occipital (n = 1).  

Please show higher temporal resolution example traces, for example in Figure 3. Differences are at the micrometer scale, but APs are shown at the millisecond scale. Hard to judge the quality of the data. Showing the command potentials (inset Figure 3D, E) is misleading (see major point #3).

In response to the reviewer's request, we have redrawn the example traces in Figure 3.

Please check the labeling of figures. There is information missing. For example, in Figure 5 A to C I am missing information and the units of the axes.

In the black plots on the right side of panels B and C, the y-axis shows the thickness measurements for the given dendrite stacked on top of each other and the x-axis shows the measurement values, the units for the x-axis are µm as mentioned in the figure legend.

Line 981 "scalebars" should read scale bars."

Line 986 "bootstraped" should read "bootstrapped".

Done.

Are the dendritic diameters increased for all basal and apical higher-order branches? It is unclear how the model simulations were built on diameters of primary and higher-order branches.

In our modelling study we took the actual diameter of the reconstructed PCs in both proximal and higher order branches. We did compare per-distance differences in diameter – but it is automatically incorporated into the computation of the basal load (“equivalent cables” in Figs 6&8).

The velocity calculation for axonal propagation (yielding a ~0.9 m/s conduction velocity, Figure 2B) is incorrect. Using the peak of the action potentials between soma and axon misses the fact that action potentials start earlier and spatially distally from the soma in the axon. Please revise the calculation to include the temporal delay and actual distance travelled by the forward propagating action potential.

Thank you for this question. We are aware that the AP is generated at the AIS and that it is located between the two recording electrodes and we have to take into account that the signal propagates from the AIS to the soma and this may shorten the delay in the system. To the best of our knowledge, there is no experimental evidence of the location of the AP generation site on the AIS in layer 2-3 pyramidal cells in the human neocortex, so we assumed that it is located 35 microns from the soma, and that the propagation speed from the AIS to the two directions is the same. Consequently, we have corrected our propagation velocity values as follows:

“For the axon bleb recordings we assumed that the axon initial segment (AIS) of the cells are 35 µm from the axon hillock, and the APs propagate to forward (to the bleb) and backward (to the soma) at the same speed. For the correction of the AIS we used the following formula: (2)

where vcorr is the corrected propagation speed for AIS position, l is the axonal distance between the soma and the axon bleb, t is the latency between the two measuring point, ais is the assumed position of the AIS alongside the axon (35 µm).”

What explains the strongly attenuated axonal action potential at the bleb? Is this representative?

The strongly attenuated axonal action potential at the bleb can be explained by a few key factors:

(1) Membrane Integrity: Bleb formation often indicates some level of membrane damage or alteration. This can disrupt the normal ionic gradients across the membrane, leading to a failure in generating or propagating action potentials effectively.

(2) Current Leakage: Bleb formation may create additional pathways for ion leakage, which can dissipate the electrical current that would normally propagate the action potential. This leakage reduces the overall amplitude of the action potential.

Line 275 "To our delight", please rephrase.

Corrected.

Reviewer #3 (Recommendations For The Authors):

- In Figure 1, the number of cells used to assess intersomatic distance is quite low. A larger number of neuron pairs should be analyzed to be more representative. Or at least an explanation of why such a low sampling can be conclusive.

We appreciate the reviewer’s concerns on sample sizes of the first set of experiments, where the anatomical pathways were measured through the synapses of coupled cells with electrophysiological recordings. We acknowledge that this is a limitation of our study. However, in this series of experiments, we simply wanted to experimentally confirm already known results which consisted of two parts: first, that in humans the dendrites and axons of neurons are longer, and second, that they have the same time delay in terms of synaptic latency. 

The reported similarity in synaptic latencies is consistent with the results of a recent study by Campagnola et al. (2022) showing that EPSP latencies of local connections between layer 2/3 pyramidal cells are in the same range in humans and mice (human median latency = 1.73 ms vs. mouse median latency = 1.49 ms). We came to the same conclusion in our previous work where we compared pyramidal basket cell synaptically coupled pairs in human and rat pairs (Molnár et al. 2016). 

On the other hand, we report interspecific differences in cable pathways from soma to soma, again consistent with the literature suggesting that the length of pyramidal neural processes is longer in humans than in rodents (see Supplementary Figure 1 and e.g. Berg et al. 2021).

From a practical point of view the collection of experimental data in this hard won experiment is particularly difficult. The electrophysiological recording of a connected pair with an appropriate pre- and postsynaptic series resistance, where human tissue samples are limited, is the first step here. To obtain information about the path of the signals between pre- and postsynaptic cells, an anatomical reconstruction is required. This requires a) a high-quality recovery of postsynaptic dendrites and presynaptic axons, b) successful tracing of all potential contact points between presynaptic axons and postsynaptic dendrites back to the pre- and postsynaptic soma. The difficulty of the latter point in particular arises from the fact that parts of the presynaptic axonal arbor are myelinated and the success of biocytin-based tracing depends on the length of the myelinated axon branches. The success/failure of complete axonal tracing only becomes apparent at the end of these efforts.

- The author should provide an intuitive explanation of why capacitive load accelerates propagation in the dendrite.

See answer above  

- The author should more clearly rank the contribution of each difference between rat and human neurons. The 10% increase in dendritic diameter which affects velocity only via a square root seems a very weak contribution. This should be clarified.

We now added a set of statistical methods to perform such a ranking in the theoretical part of this study, as described above (and in a new paragraph, attached above) in the revised article. 

References

Eyal, G., Mansvelder, H. D., de Kock, C. P. J., & Segev, I. (2014). Dendrites impact the encoding capabilities of the axon. Journal of Neuroscience, 34(24), 8063–8071. https://doi.org/10.1523/JNEUROSCI.5431-13.2014

Friedman, J. H. (2002). Stochastic gradient boosting. In Computational Statistics & Data Analysis (Vol. 38). www.elsevier.com/locate/csda

Kiebel, S. J., & Holmes, A. P. (2007). The General Linear Model. In K. Friston, J. Ashburner, S. Kiebel, T. Nichols, & P. William (Eds.), Statistical Parametric Mapping (pp. 101–125). Academic Press.

Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Proceedings of the 31st International Conference on Neural Information Processing Systems, 4768–4777.

eLife Assessment

This study presents valuable findings to the field interested in inattentional blindness (IB), the phenomenon that participants fail to notice salient stimuli when their attention is directed elsewhere. This study reveals that participants who indicate no awareness of unexpected stimuli through yes/no questions ("did you notice anything unusual?"), may still show above-chance sensitivity to specific properties of these stimuli through follow-up forced-choice questions (e.g., regarding its location or color). By introducing absent trials where no IB stimulus is presented, the authors show that this is because participants are generally conservative and biased to report not noticing in inattentional blindness experiments. The evidence supporting these conclusions is convincing, the samples sizes are large and the analysis protocol is novel.

Reviewer #1 (Public review):

The results of these experiments support a modest but important conclusion: If sub-optimal methods are used to collect retrospective reports, such as simple yes/no questions, inattentional blindness (IB) rates may be overestimated by up to ~8%.

(1) In experiment 1, data from 374 subjects were included in the analysis. As shown in figure 2b, 267 subjects reported noticing the critical stimulus and 107 subjects reported not noticing it. This translates to a 29% IB rate if we were to only consider the "did you notice anything unusual Y/N" question. As reported in the results text (and figure 2c), when asked to report the location of the critical stimulus (left/right), 63.6% of the "non-noticer" group answered correctly. In other words, 68 subjects were correct about the location while 39 subjects were incorrect. Importantly, because the location judgment was a 2-alternative-forced-choice, the assumption was that if 50% (or at least not statistically different than 50%) of the subjects answered the location question correctly, everyone was purely guessing. Therefore, we can estimate that ~39 of the subjects who answered correctly were simply guessing (because 39 guessed incorrectly), leaving 29 subjects from the non-noticer group who were correct on the 2AFC above and beyond the pure guess rate. If these 29 subjects are moved from the non-noticer to the noticer group, the corrected rate of IB for Experiment 1 is 20.86% instead of the original 28.61% rate that would have been obtained if only the Y/N question was used. In other words, relying only on the "Y/N did you notice anything" question led to an overestimate of IB rates by 7.75% in Experiment 1.

In the revised version of their manuscript, the authors provided the data that was missing from the original submission, which allows this same exercise to be carried out on the other 4 experiments. Using the same logic as above, i.e., calculating the pure-guess rate on the 2AFC, moving the number of subjects above this pure-guess rate to the non-noticer group, and then re-calculating a "corrected IB rate", the other experiments demonstrate the following:

Experiment 2: IB rates were overestimated by 4.74% (original IB rate based only on Y/N question = 27.73%; corrected IB rate that includes the 2AFC = 22.99%)

Experiment 3: IB rates were overestimated by 3.58% (original IB rate = 30.85%; corrected IB rate = 27.27%)

Experiment 4: IB rates were overestimated by ~8.19% (original IB rate = 57.32%; corrected IB rate for color* = 39.71%, corrected IB rate for shape = 52.61%, corrected IB rate for location = 55.07%)

Experiment 5: IB rates were overestimated by ~1.44% (original IB rate = 28.99%; corrected IB rate for color = 27.56%, corrected IB rate for shape = 26.43%, corrected IB rate for location = 28.65%)

*note: the highest overestimate of IB rates was from Experiment 4, color condition, but the authors admitted that there was a problem with 2AFC color guessing bias in this version of the experiment which was a main motivation for running experiment 5 which corrected for this bias.

Taken as a whole, this data clearly demonstrates that even with a conservative approach to analyzing the combination of Y/N and 2AFC data, inattentional blindness was evident in a sizeable portion of the subject populations. An important (albeit modest) overestimate of IB rates was demonstrated by incorporating these improved methods.

(2) One of the strongest pieces of evidence presented in this paper was the single data point in Figure 3e showing that in Experiment 3, even the super subject group that rated their non-noticing as "highly confident" had a d' score significantly above zero. Asking for confidence ratings is certainly an improvement over simple Y/N questions about noticing, and if this result were to hold, it could provide a key challenge to IB. However, this result can most likely be explained by measurement error.

In their revised paper, the authors reported data that was missing from their original submission: the confidence ratings on the 2AFC judgments that followed the initial Y/N question. The most striking indication that this data is likely due to measurement error comes from the number of subjects who indicated that they were highly confident that they didn't notice anything on the critical trial, but then when asked to guess the location of the stimulus, indicated that they were highly confident that the stimulus was on the left (or right). There were 18 subjects (8.82% of the high-confidence non-noticer group) who responded this way. To most readers, this combination of responses (high confidence in correctly judging a stimulus feature that one is highly confident in having not seen at all) indicates that a portion of subjects misunderstood the confidence scales (or just didn't read the questions carefully or made mistakes in their responses, which is common for experiments conducted online).

In the authors' rebuttal to the first round of peer review, they wrote, "it is perfectly rationally coherent to be very confident that one didn't see anything but also very confident that if there was anything to be seen, it was on the left." I respectfully disagree that such a combination of responses is rationally coherent. The more parsimonious interpretation is that a measurement error occurred, and it's questionable whether we should trust any responses from these 18 subjects.

In their rebuttal, the authors go on to note that 14 of the 18 subjects who rated their 2AFC with high confidence were correct in their location judgment. If these 14 subjects were removed from analysis (which seems like a reasonable analysis choice, given their contradictory responses), d' for the high-confidence non-noticer group would most likely fall to chance levels. In other words, we would see a data pattern similar to that plotted in Figure 3e, but with the first data point on the left moving down to zero d'. This corrected Figure 3e would then provide a very nice evidence-based justification for including confidence ratings along with Y/N questions in future inattentional blindness studies.

(3) In most (if not all) IB experiments in the literature, a partial attention and/or full attention trial is administered after the critical trial. These control trials are very important for validating IB on the critical trial, as they must show that, when attended, the critical stimuli are very easy to see. If a subject cannot detect the critical stimulus on the control trial, one cannot conclude that they were inattentionally blind on the critical trial, e.g., perhaps the stimulus was just too difficult to see (e.g., too weak, too brief, too far in the periphery, too crowded by distractor stimuli, etc.), or perhaps they weren't paying enough attention overall or failed to follow instructions. In the aggregate data, rates of noticing the stimuli should increase substantially from the critical trial to the control trials. If noticing rates are equivalent on the critical and control trials, one cannot conclude that attention was manipulated in the first place.

In their rebuttal to the first round of peer review, the authors provided weak justification for not including such a control condition. They cite one paper that argues such control conditions are often used to exclude subjects from analysis (those who fail to notice the stimulus on the control trial are either removed from analysis or replaced with new subjects) and such exclusions/replacements can lead to underestimations of inattentional blindness rates. However, the inclusion of a partial or full attention condition as a control does not necessitate the extra step of excluding or replacing subjects. In the broadest sense, such a control condition simply validates the attention manipulation, i.e., one can easily compare the percent of subjects who answered "yes" or who got the 2AFC judgment correct during the critical trial versus the control trial. The subsequent choice about exclusion/replacement is separate, and researchers can always report the data with and without such exclusions/replacements to remain more neutral on this practice.

If anyone were to follow-up on this study, I highly recommend including a partial or full attention control condition, especially given the online nature of data collection. It's important to know the percent of online subjects who answer yes and who get the 2AFC question correct when the critical stimulus is attended, because that is the baseline (in this case, the "ceiling level" of performance) to which the IB rates on the critical trial can be compared.

Reviewer #2 (Public review):

In this study, Nartker et al. examine how much observers are conscious of using variations of classic inattentional blindness studies. The key idea is that rather than simply ask observers if they noticed a critical object with one yes/no question, the authors also ask follow-up questions to determine if observers are aware of more than the yes/no questions suggest. Specifically, by having observers make forced choice guesses about the critical object, the authors find that many observers who initially said "no" they did not see the object can still "guess" above chance about the critical object's location, color, etc. Thus, the authors claim, that prior claims of inattentional blindness are mistaken and that using such simple methods has led numerous researchers to overestimate how little observers see in the world. To quote the authors themselves, these results imply that "inattentionally blind subjects consciously perceive these stimuli after all... they show sensitivity to IB stimuli because they can see them."

Before getting to a few issues I have with the paper, I do want to make sure to explicitly compliment the researchers for many aspects of their work. Getting massive amounts of data, using signal detection measures, and the novel use of a "super subject" are all important contributions to the literature that I hope are employed more in the future.

Main point 1: My primary issue with this work is that I believe the authors are misrepresenting the way people often perform inattentional blindness studies. In effect, the authors are saying, "People do the studies 'incorrectly' and report that people see very little. We perform the studies 'correctly' and report that people see much more than previously thought." But the way previous studies are conducted is not accurately described in this paper. The authors describe previous studies as follows on page 3:

"Crucially, however, this interpretation of IB and the many implications that follow from it rest on a measure that psychophysics has long recognized to be problematic: simply asking participants whether they noticed anything unusual. In IB studies, awareness of the unexpected stimulus (the novel shape, the parading gorilla, etc.) is retroactively probed with a yes/no question, standardly, "Did you notice anything unusual on the last trial which wasn't there on previous trials?". Any subject who answers "no" is assumed not to have any awareness of the unexpected stimulus.

If this quote were true, the authors would have a point. Unfortunately, I do not believe it is true. This is simply not how many inattentional blindness studies are run. Some of the most famous studies in the inattentional blindness literature do not simply as observes a yes/no question (e.g., the invisible gorilla (Simons et al. 1999), the classic door study where the person changes (Simons and Levin, 1998), the study where observers do not notice a fight happening a few feet from them (Chabris et al., 2011). Instead, these papers consistently ask a series of follow-up questions and even tell the observers what just occurred to confirm that observers did not notice that critical event (e.g., "If I were to tell you we just did XYZ, did you notice that?"). In fact, after a brief search on Google Scholar, I was able to relatively quickly find over a dozen papers that do not just use a yes/no procedure, and instead as a series of multiple questions to determine if someone is inattentionally blind. In no particular order some papers:

(1) Most et al. (2005) Psych Review<br /> (2) Drew et al. (2013) Psych Science<br /> (3) Drew et al. (2016) Journal of Vision<br /> (4) Simons et al. (1999) Perception<br /> (5) Simons and Levin (1998) Perception<br /> (6) Chabris et al. (2011) iPerception<br /> (7) Ward & Scholl (2015) Psych Bulletin and Review<br /> (8) Most et al. (2001) Psych Science<br /> (9) Todd & Marois (2005) Psych Science<br /> (10) Fougnie & Marois (2007) Psych Bulletin and Review<br /> (11) New and German (2015) Evolution and Human Behaviour<br /> (12) Jackson-Nielsen (2017) Consciousness and cognition<br /> (13) Mack et al. (2016) Consciousness and cognition<br /> (14) Devue et al. (2009) Perception<br /> (15) Memmert (2014) Cognitive Development<br /> (16) Moore & Egeth (1997) JEP:HPP<br /> (17) Cohen et al. (2020) Proc Natl Acad Sci<br /> (18) Cohen et al. (2011) Psych Science

This is a critical point. The authors' key idea is that when you ask more than just a simple yes/no question, you find that other studies have overestimated the effects of inattentional blindness. But none of the studies listed above only asked simple yes/no questions. Thus, I believe the authors are mis-representing the field. Moreover, many of the studies that do much more than ask a simple yes/no question are cited by the authors themselves! Furthermore, as far as I can tell, the authors believe that if researchers do these extra steps and ask more follow-ups, then the results are valid. But since so many of these prior studies do those extra steps, I am not exactly sure what is being criticized.

To make sure this point is clear, I'd like to use a paper of mine as an example. In this study (Cohen et al., 2020, Proc Natl Acad Sci USA) we used gaze-contingent virtual reality to examine how much color people see in the world. On the critical trial, the part of the scene they fixated on was in color, but the periphery was entirely in black and white. As soon as the trial ended, we asked participants a series of questions to determine what they noticed. The list of questions included:

(1) "Did you notice anything strange or different about that last trial?"<br /> (2) "If I were to tell you that we did something odd on the last trial, would you have a guess as to what we did?"<br /> (3) "If I were to tell you we did something different in the second half of the last trial, would you have a guess as to what we did?"<br /> (4) "Did you notice anything different about the colors in the last scene?"<br /> (5) We then showed observers the previous trial again and drew their attention to the effect and confirmed that they did not notice that previously.<br /> In a situation like this, when the observers are asked so many questions, do the authors believe that "the inattentionally blind can see after all?" I believe they would not say that and the reason they would not say that is because of the follow-up questions after the initial yes/no question. But since so many previous studies use similar follow-up questions, I do not think you can state that the field is broadly overestimating inattentional blindness. This is why it seems to me to be a bit of a straw-man: most people do not just use the yes/no method.

Main point 2: Let's imagine for a second that every study did just ask a yes/no question and then would stop. So, the criticism the authors are bringing up is valid (even though I believe it is not). I am not entirely sure that above chance performance on a forced choice task proves that the inattentionally blind can see after all. Could it just be a form of subliminal priming? Could there be a significant number of participants who basically would say something like, "No I did not see anything, and I feel like I am just guessing, but if you want me to say whether the thing was to the left or right, I will just 100% guess"? I know the literature on priming from things like change and inattentional blindness is a bit unclear, but this seems like maybe what is going on. In fact, maybe the authors are getting some of the best priming from inattentional blindness because of their large sample size, which previous studies do not use.<br /> I'm curious how the authors would relate their studies to masked priming. In masked priming studies, observers say the did not see the target (like in this study) but still are above chance when forced to guess (like in this study). Do the researchers here think that that is evidence of "masked stimuli are truly seen" even if a participant openly says they are guessing?

Main point 3: My last question is about how the authors interpret a variety of inattentional blindness findings. Previous work has found that observers fail to notice a gorilla in a CT scan (Drew et al., 2013), a fight occurring right in front of them (Chabris et al., 2011), a plane on a runway that pilots crash into (Haines, 1991), and so forth. In a situation like this, do the authors believe that many participants are truly aware of these items but simply failed to answer a yes/no question correctly? For example, imagine the researchers made participants choose if the gorilla was in the left or right lung and some participants who initially said they did not notice the gorilla were still able to correctly say if it was in the left or right lung. Would the authors claim "that participant actually did see the gorilla in the lung"? I ask because it is difficult to understand what it means to be aware of something as salient as a gorilla in a CT scan, but say "no" you didn't notice it when asked a yes/no question. What does it mean to be aware of such important, ecologically relevant stimuli, but not act in response to them and openly say "no" you did not notice them?

Overall: I believe there are many aspects of this set of studies that are innovative and I hope the methods will be used more broadly in the literature. However, I believe the authors misrepresent the field and overstate what can be interpreted from their results. While I am sure there are cases where more nuanced questions might reveal inattentional blindness is somewhat overestimated, claims like "the inattentionally blind can see after all" or "Inattentionally blind subjects consciously perceive thest stimuli after all" seem to be incorrect (or at least not at all proven by this data).

Author response:

The following is the authors’ response to the current reviews.

Responses to Reviewer #1:

We thank the reviewer for these additional comments, and more generally for their extensive engagement with our work, which is greatly appreciated. Here, we respond to the three points in their latest review in turn.

The results of these experiments support a modest but important conclusion: If sub-optimal methods are used to collect retrospective reports, such as simple yes/no questions, inattentional blindness (IB) rates may be overestimated by up to ~8%.

It is true, of course, that we think the field has overstated the extent of IB, and we appreciate the reviewer characterizing our results as important along these lines. Nevertheless, we respectfully disagree with the framing and interpretation the reviewer attaches to them. As explained in our previous response, we think this interpretation — and the associated calculations of IB overestimation ‘rates’ — perpetuates a binary approach to perception and awareness which we regard as mistaken.

A graded approach to IB and visual awareness 

Our sense is that many theorists interested in IB have conceived of perception and awareness as ‘all or nothing’: You either see a perfectly clear gorilla right in front of you, or you see nothing at all. This is implicit in the reviewer’s characterization of our results as simply indicating that fewer subjects fail to see the critical stimulus than previously assumed. To think that way is precisely to assume the orthodox binary position about perception, i.e., that any given subject can neatly be categorized into one of two boxes, saw or didn’t see.

Our perspective is different. We think there can be degraded forms of perception and awareness that fall neatly into neither of the categories “saw the stimulus perfectly clearly” or “saw nothing at all”. On this graded conception, the question is not: “What proportion of subjects saw the stimulus?” but: “What is the sensitivity of subjects to the stimulus?” This is why we prefer signal detection measures like d′ over % noticing and % correct. This powerful framework has been successful in essentially every domain to which it has been applied, and we think perception and visual awareness are no exception. We understand that the reviewer may not think the same way about this foundational issue, but since part of our goal is to promote a graded approach to perception, we are keen to highlight our disagreement here and so resist the reviewer’s interpretation of our results (even to the extent that it is a positive one!).

Finally, we note that given this perspective, we are correspondingly inclined to reject many of the summary figures following below in Point (1) by the reviewer. These calculations (given in terms of % noticing and not noticing) make sense on the binary conception of awareness, but not on the SDT-based approach we favor. We say more about this below. 

(1) In experiment 1, data from 374 subjects were included in the analysis. As shown in figure 2b, 267 subjects reported noticing the critical stimulus and 107 subjects reported not noticing it. This translates to a 29% IB rate if we were to only consider the "did you notice anything unusual Y/N" question. As reported in the results text (and figure 2c), when asked to report the location of the critical stimulus (left/right), 63.6% of the "non-noticer" group answered correctly. In other words, 68 subjects were correct about the location while 39 subjects were incorrect. Importantly, because the location judgment was a 2-alternative-forced-choice, the assumption was that if 50% (or at least not statistically different than 50%) of the subjects answered the location question correctly, everyone was purely guessing. Therefore, we can estimate that ~39 of the subjects who answered correctly were simply guessing (because 39 guessed incorrectly), leaving 29 subjects from the nonnoticer group who were correct on the 2AFC above and beyond the pure guess rate. If these 29 subjects are moved from the non-noticer to the noticer group, the corrected rate of IB for Experiment 1 is 20.86% instead of the original 28.61% rate that would have been obtained if only the Y/N question was used. In other words, relying only on the "Y/N did you notice anything" question led to an overestimate of IB rates by 7.75% in Experiment 1.

In the revised version of their manuscript, the authors provided the data that was missing from the original submission, which allows this same exercise to be carried out on the other 4 experiments.  

(To briefly interject: All of these data were provided in our public archive since our original submission and remain available at https://osf.io/fcrhu. The difference now is only that they are included in the manuscript itself.)

Using the same logic as above, i.e., calculating the pure-guess rate on the 2AFC, moving the number of subjects above this pure-guess rate to the non-noticer group, and then re-calculating a "corrected IB rate", the other experiments demonstrate the following:

Experiment 2: IB rates were overestimated by 4.74% (original IB rate based only on Y/N question = 27.73%; corrected IB rate that includes the 2AFC = 22.99%)

Experiment 3: IB rates were overestimated by 3.58% (original IB rate = 30.85%; corrected IB rate = 27.27%)

Experiment 4: IB rates were overestimated by ~8.19% (original IB rate = 57.32%; corrected IB rate for color* = 39.71%, corrected IB rate for shape = 52.61%, corrected IB rate for location = 55.07%)

Experiment 5: IB rates were overestimated by ~1.44% (original IB rate = 28.99%; corrected IB rate for color = 27.56%, corrected IB rate for shape = 26.43%, corrected IB rate for location = 28.65%)

*note: the highest overestimate of IB rates was from Experiment 4, color condition, but the authors admitted that there was a problem with 2AFC color guessing bias in this version of the experiment which was a main motivation for running experiment 5 which corrected for this bias.

Taken as a whole, this data clearly demonstrates that even with a conservative approach to analyzing the combination of Y/N and 2AFC data, inattentional blindness was evident in a sizeable portion of the subject populations. An important (albeit modest) overestimate of IB rates was demonstrated by incorporating these improved methods.

We appreciate the work the reviewer has put into making these calculations. However, as noted above, such calculations implicitly reflect the binary approach to perception and awareness that we reject. 

Consider how we’d think about the single subject case where the task is 2afc detection of a low contrast stimulus in noise. Suppose that this subject achieves 70% correct. One way of thinking about this is that the subject fully and clearly sees the stimulus on 40% of trials (achieving 100% correct on those) and guesses completely blindly on the other 60% (achieving 50% correct on those) for a total of 40% + 30% = 70% overall. However, this is essentially a ‘high threshold’ approach to the problem, in contrast to an SDT approach. On an SDT approach — an approach with tremendous evidential support — on every trial the subject receives samples from probabilistic distributions corresponding to each interval (one noise and one signal + noise) and determines which is higher according to the 2afc decision rule. Thus, across trials, they have access to differentially graded information about the stimulus. Moreover, on some trials they may have significant information from the stimulus (perhaps, well above their single interval detection criterion) but still decide incorrectly because of high noise from the other spatial interval. From this perspective, there is no nonarbitrary way of saying whether the subject saw/did not see on a given trial. Instead, we must characterize the subject’s overall sensitivity to the stimulus/its visibility to them in terms of a parameter such as d′ (here, ~ 0.7).

We take the same attitude to the subjects in our experiments (and specifically to our ‘super subject’). Instead of calculating the proportion of subjects who saw or failed to see the stimulus (with some characterized as aware and some as unaware), we think the best way to characterize our results is that, across subjects (and so trials also), there was differential graded access to information from the stimulus, and this is best represented in terms of the group-level sensitivity parameter d′. This is why we frame our results as demonstrating that subjects traditionally considered inattentionally blind exhibit significant residual visual sensitivity to the critical stimulus.

(2) One of the strongest pieces of evidence presented in this paper was the single data point in Figure 3e showing that in Experiment 3, even the super subject group that rated their non-noticing as "highly confident" had a d' score significantly above zero. Asking for confidence ratings is certainly an improvement over simple Y/N questions about noticing, and if this result were to hold, it could provide a key challenge to IB. However, this result can most likely be explained by measurement error.

In their revised paper, the authors reported data that was missing from their original submission: the confidence ratings on the 2AFC judgments that followed the initial Y/N question. The most striking indication that this data is likely due to measurement error comes from the number of subjects who indicated that they were highly confident that they didn't notice anything on the critical trial, but then when asked to guess the location of the stimulus, indicated that they were highly confident that the stimulus was on the left (or right). There were 18 subjects (8.82% of the high-confidence non-noticer group) who responded this way. To most readers, this combination of responses (high confidence in correctly judging a stimulus feature that one is highly confident in having not seen at all) indicates that a portion of subjects misunderstood the confidence scales (or just didn't read the questions carefully or made mistakes in their responses, which is common for experiments conducted online).

In the authors' rebuttal to the first round of peer review, they wrote, "it is perfectly rationally coherent to be very confident that one didn't see anything but also very confident that if there was anything to be seen, it was on the left." I respectfully disagree that such a combination of responses is rationally coherent. The more parsimonious interpretation is that a measurement error occurred, and it's questionable whether we should trust any responses from these 18 subjects.

In their rebuttal, the authors go on to note that 14 of the 18 subjects who rated their 2AFC with high confidence were correct in their location judgment. If these 14 subjects were removed from analysis (which seems like a reasonable analysis choice, given their contradictory responses), d' for the high-confidence non-noticer group would most likely fall to chance levels. In other words, we would see a data pattern similar to that plotted in Figure 3e, but with the first data point on the left moving down to zero d'. This corrected Figure 3e would then provide a very nice evidence-based justification for including confidence ratings along with Y/N questions in future inattentional blindness studies.

We appreciate the reviewer’s highlighting of this particular piece of evidence as amongst our strongest. (At the same time, we must resist its characterization as a “single data point”: it derives from a large pre-registered experiment involving some 7,000 subjects total, with over 200 subjects in the relevant bin — both figures being far larger than a typical IB experiment.) We also appreciate their raising the issue of measurement error.

Specifically, the reviewer contends that our finding that even highly confident non-noticers exhibit significant sensitivity is “most likely … explained by measurement error” due to subjects mistakenly inverting our confidence scale in giving their response. In our original reply, we gave two reasons for thinking this quite unlikely; the reviewer has not addressed these in this revised review. First, we explicitly labeled our confidence scale (with 0 labeled as ‘Not at all confident’ and 3 as ‘Highly confident’) so that subjects would be very unlikely simply to invert the scale. This is especially so as it is very counterintuitive to treat “0” as reflecting high confidence. More importantly, however, we reasoned that any measurement error due to inverting or misconstruing the confidence scale should be symmetric. That is: If subjects are liable to invert the confidence scale, they should do so just as often when they answer “yes” as when they answer “no” – after all the very same scale is being used in both cases. This allows us to explore evidence of measurement error in relation to the large number of high-confidence “yes” subjects (N = 2677), thus providing a robust indicator as to whether subjects are generally liable to misconstrue the confidence scale. Looking at the number of such high confidence noticers who subsequently respond to the 2afc question with low confidence (a pattern which might, though need not, suggest measurement error), we found that the number was tiny. Only 28/2677 (1.05%) of high-confidence noticers subsequently gave the lowest level of confidence on the 2afc question, and only 63/2677 (2.35%) subjects gave either of the two lower levels of confidence. For these reasons, we consider any measurement error due to misunderstanding the confidence scale to be extremely minimal.

The reviewer is correct to note that 18/204 (9%) subjects reported both being highly confident that they didn't notice anything and highly confident in their 2afc judgment, although only 14/18 were correct in this judgment. Should we exclude these 14? Perhaps if we agree with the reviewer that such a pattern of responses is not “rationally coherent” and so must reflect a misconstrual of the scale. But such a pattern is in fact perfectly and straightforwardly intelligible. Specifically, in a 2afc task, two stimuli can individually fall well below a subject’s single interval detection criterion — leading to a high confidence judgment that nothing was presented in either interval. Quite consistent with this, the lefthand stimulus may produce a signal that is much higher than the right-hand stimulus — leading to a high confidence forced-choice judgment that, if something was presented, it was on the left. (By analogy, consider how a radiologist could look at a scan and say the following: “We’re 95% confident there’s no tumor. But even on the 5% chance that there is, our tests completely rule out that it’s a malignant one, so don’t worry.”) 

(3) In most (if not all) IB experiments in the literature, a partial attention and/or full attention trial is administered after the critical trial. These control trials are very important for validating IB on the critical trial, as they must show that, when attended, the critical stimuli are very easy to see. If a subject cannot detect the critical stimulus on the control trial, one cannot conclude that they were inattentionally blind on the critical trial, e.g., perhaps the stimulus was just too difficult to see (e.g., too weak, too brief, too far in the periphery, too crowded by distractor stimuli, etc.), or perhaps they weren't paying enough attention overall or failed to follow instructions. In the aggregate data, rates of noticing the stimuli should increase substantially from the critical trial to the control trials. If noticing rates are equivalent on the critical and control trials, one cannot conclude that attention was manipulated in the first place.

In their rebuttal to the first round of peer review, the authors provided weak justification for not including such a control condition. They cite one paper that argues such control conditions are often used to exclude subjects from analysis (those who fail to notice the stimulus on the control trial are either removed from analysis or replaced with new subjects) and such exclusions/replacements can lead to underestimations of inattentional blindness rates. However, the inclusion of a partial or full attention condition as a control does not necessitate the extra step of excluding or replacing subjects. In the broadest sense, such a control condition simply validates the attention manipulation, i.e., one can easily compare the percent of subjects who answered "yes" or who got the 2AFC judgment correct during the critical trial versus the control trial. The subsequent choice about exclusion/replacement is separate, and researchers can always report the data with and without such exclusions/replacements to remain more neutral on this practice.

If anyone were to follow-up on this study, I highly recommend including a partial or full attention control condition, especially given the online nature of data collection. It's important to know the percent of online subjects who answer yes and who get the 2AFC question correct when the critical stimulus is attended, because that is the baseline (in this case, the "ceiling level" of performance) to which the IB rates on the critical trial can be compared.

We agree with the reviewer that future studies could benefit from including a partial or full attention condition. They are surely right that we might learn something additional from such conditions. 

Where we differ from the reviewer is in thinking of these conditions as “controls” appropriate to our research question. This is why we offered the justification we did in our earlier response. When these conditions are used as controls, they are used to exclude subjects in ways that serve to inflate the biases we are concerned with in our work. For our question, the absence of these conditions does not impact the significance of the findings, since such conditions are designed to answer a question which is not the one at the heart of our paper. Our key claim is that subjects who deny noticing an unexpected stimulus in a standard inattentional blindness paradigm nonetheless exhibit significant residual sensitivity (as well as a conservative bias in their response to the noticing question); the presence or absence of partial- or full-attention conditions is orthogonal to that question.

Moreover, we note that our tasks were precisely chosen to be classic tasks widely used in the literature to manipulate attention. Thus, by common consensus in the field, they are effective means to soak up attention, and have in effect been tested in partial- and full-attention control settings in a huge number of studies. Second, we think it very doubtful that subjects in a full-attention trial would not overwhelmingly have detected our critical stimuli. The reviewer worries that they might have been “too weak, too brief, too far in the periphery, too crowded by distractor stimuli, etc.” But consider E5 where the stimulus was a highly salient orange or green shape, present on the screen for 5 seconds. The reviewer also suggests that subjects in the full-attention control might not have detected the stimulus because they “weren't paying enough attention overall”. But evidently if they weren’t paying attention even in the full-attention trial this would be reason for thinking that there was inattentional blindness even in this condition (a point made by White et al. 2018) and certainly not a reason for thinking there was not an attentional effect in the critical trial. Lastly, the reviewer suggests that a full-attention condition would have helped ensure that subjects were following instructions. But we ensured this already by (as per our pre-registration) excluding subjects who performed poorly in the relevant primary tasks.

Thus, both in principle and in practice, we do not see the absence of such conditions as impacting the interpretation of our findings, even as we agree that future work posing a different research question could certainly learn something from including such conditions.

Responses to Reviewer #2:

We note that this report is unchanged from an earlier round of review, and not a response to our significantly revised manuscript. We believe our latest version fully addresses all the issues which the reviewer originally raised. The interested reader can see our original response below. We again thank the reviewer for their previous report which was extremely helpful.

—-

The following is the authors’ response to the original reviews.

eLife Assessment

This study presents valuable findings to the field interested in inattentional blindness (IB), reporting that participants indicating no awareness of unexpected stimuli through yes/no questions, still show above-chance sensitivity to specific properties of these stimuli through follow-up forced-choice questions (e.g., its color). The results suggest that this is because participants are conservative and biased to report not noticing in IB. The authors conclude that these results provide evidence for residual perceptual awareness of inattentionally blind stimuli and that therefore these findings cast doubt on the claim that awareness requires attention. Although the samples are large and the analysis protocol novel, the evidence supporting this interpretation is still incomplete, because effect sizes are rather small, the experimental design could be improved and alternative explanations have not been ruled out.

We are encouraged to hear that eLife found our work “valuable”. We also understand, having closely looked at the reviews, why the assessment also includes an evaluation of “incomplete”. We gave considerable attention to this latter aspect of the assessment in our revision. In addition to providing additional data and analyses that we believe strengthen our case, we also include a much more substantial review and critique of existing methods in the IB literature to make clear exactly the gap our work fills and the advance it makes. (Indeed, if it is appropriate to say this here, we believe one key aspect of our work that is missing from the assessment is our inclusion of ‘absent’ trials, which is what allows us to make the crucial claims about conservative reporting of awareness in IB for the first time.) Moreover, we refocus our discussion on only our most central claims, and weaken several of our secondary claims so that the data we’ve collected are better aligned with the conclusions we draw, to ensure that the case we now make is in fact complete. Specifically, our two core claims are (1) that there is residual sensitivity to visual features for subjects who would ordinarily be classified as inattentionally blind (whether this sensitivity is conscious or not), and (2) that there is a tendency to respond conservatively on yes/no questions in the context of IB. We believe we have very compelling support for these two core claims, as we explain in detail below and also through revisions to our manuscript.

Given the combination of strengthened and clarified case, as well as the weakening of any conclusions that may not have been fully supported, we believe and hope that these efforts make our contribution “solid”, “convincing”, or even “compelling” (especially because the “compelling” assessment characterizes contributions that are “more rigorous than the current state-of-the-art”, which we believe to be the case given the issues that have plagued this literature and that we make progress on).

Reviewer #1 (Public review):

Summary:

In the abstract and throughout the paper, the authors boldly claim that their evidence, from the largest set of data ever collected on inattentional blindness, supports the views that "inattentionally blind participants can successfully report the location, color, and shape of stimuli they deny noticing", "subjects retain awareness of stimuli they fail to report", and "these data...cast doubt on claims that awareness requires attention." If their results were to support these claims, this study would overturn 25+ years of research on inattentional blindness, resolve the rich vs. sparse debate in consciousness research, and critically challenge the current majority view in cognitive science that attention is necessary for awareness.

Unfortunately, these extraordinary claims are not supported by extraordinary (or even moderately convincing) evidence. At best, the results support the more modest conclusion: If sub-optimal methods are used to collect retrospective reports, inattentional blindness rates will be overestimated by up to ~8% (details provided below in comment #1). This evidence-based conclusion means that the phenomenon of inattentional blindness is alive and well as it is even robust to experiments that were specifically aimed at falsifying it. Thankfully, improved methods already exist for correcting the ~8% overestimation of IB rates that this study successfully identified.

We appreciate here the reviewer’s recognition of the importance of work on inattentional blindness, and the centrality of inattentional blindness to a range of major questions. We also recognize their concerns with what they see as a gap between our data and the claims made on their basis. We address this in detail below (as well as, of course, in our revised manuscript). However, from the outset we are keen to clarify that our central claim is only the first one the reviewer mentions — and the one which appears in our title — namely that, as a group, participants can successfully report the location, color, and shape of stimuli they deny noticing, and thus that there is “Sensitivity to visual features in inattentional blindness”. This is the claim that we believe is strongly supported by our data, and all the more so after revising the manuscript in light of the helpful comments we’ve received.

By contrast, the other claims the reviewer mentions, concerning awareness (as opposed to residual sensitivity–which might be conscious or unconscious) were intended as both secondary and tentative. We agree with the referee that these are not as strongly supported by our data (and indeed we say so in our manuscript), whereas we do think our data strongly support the more modest — and, to us central — claim that, as a group, inattentionally blind participants can successfully report the location, color, and shape of stimuli they deny noticing. 

We also feel compelled to resist somewhat the reviewer’s summary of our claims. For example, the reviewer attributes to us the claim that “subjects retain awareness of stimuli they fail to report”; but while that phrase does appear in our abstract, what we in fact say is that our data are “consistent with an alternative hypothesis about IB, namely that subjects retain awareness of stimuli they fail to report”. We do in fact believe that our data are consistent with that hypothesis, whereas earlier investigations seemed not to be. We mention this only because we had used that careful phrasing precisely for this sort of reason, so that we wouldn’t be read as saying that our results unequivocally support that alternative.

Still, looking back, we see how we may have given more emphasis than we intended to some of these more secondary claims. So, we’ve now gone through and revised our manuscript throughout to emphasize that our main claim is about residual sensitivity, and to make clear that our claims about awareness are secondary and tentative. Indeed, we now say precisely this, that although we favor an interpretation of “our results in terms of residual conscious vision in IB … this claim is tentative and secondary to our primary finding”. We also weaken the statements in the abstract that the reviewer mentions, to better reflect our key claims.

Finally, we note one further point: Dialectically, inattentional blindness has been used to argue (e.g.) that attention is required for awareness. We think that our data concerning residual sensitivity at least push back on the use of IB to make this claim, even if (as we agree) they do not provide decisive evidence that awareness survives inattention. In other words, we think our data call that claim into question, such that it’s now genuinely unclear whether awareness does or does not survive inattention. We have adjusted our claims on this point accordingly as well.

Comments:

(1) In experiment 1, data from 374 subjects were included in the analysis. As shown in figure 2b, 267 subjects reported noticing the critical stimulus and 107 subjects reported not noticing it. This translates to a 29% IB rate, if we were to only consider the "did you notice anything unusual Y/N" question. As reported in the results text (and figure 2c), when asked to report the location of the critical stimulus (left/right), 63.6% of the "non-noticer" group answered correctly. In other words, 68 subjects were correct about the location while 39 subjects were incorrect. Importantly, because the location judgment was a 2-alternative-forced-choice, the assumption was that if 50% (or at least not statistically different than 50%) of the subjects answered the location question correctly, everyone was purely guessing. Therefore, we can estimate that ~39 of the subjects who answered correctly were simply guessing (because 39 guessed incorrectly), leaving 29 subjects from the nonnoticer group who may have indeed actually seen the location of the stimulus. If these 29 subjects are moved to the noticer group, the corrected rate of IB for experiment 1 is 21% instead of 29%. In other words, relying only on the "Y/N did you notice anything" question leads to an overestimate of IB rates by 8%. This modest level of inaccuracy in estimating IB rates is insufficient for concluding that "subjects retain awareness of stimuli they fail to report", i.e. that inattentional blindness does not exist.

In addition, this 8% inaccuracy in IB rates only considers one side of the story. Given the data reported for experiment 1, one can also calculate the number of subjects who answered "yes, I did notice something unusual" but then reported the incorrect location of the critical stimulus. This turned out to be 8 subjects (or 3% of the "noticer" group). Some would argue that it's reasonable to consider these subjects as inattentionally blind, since they couldn't even report where the critical stimulus they apparently noticed was located. If we move these 8 subjects to the non-noticer group, the 8% overestimation of IB rates is reduced to 6%.

The same exercise can and should be carried out on the other 4 experiments, however, the authors do not report the subject numbers for any of the other experiments, i.e., how many subjects answered Y/N to the noticing question and how many in each group correctly answered the stimulus feature question. From the limited data reported (only total subject numbers and d' values), the effect sizes in experiments 2-5 were all smaller than in experiment 1 (d' for the non-noticer group was lower in all of these follow-up experiments), so it can be safely assumed that the ~6-8% overestimation of IB rates was smaller in these other four experiments. In a revision, the authors should consider reporting these subject numbers for all 5 experiments.

We now report, as requested, all these subject numbers in our supplementary data (see Supplementary Tables 1 and 2 in our Supplementary Materials).

However, we wish to address the larger question the reviewer has raised: Do our data only support a relatively modest reduction in IB rates? Even if they did, we still believe that this would be a consequential result, suggesting a significant overestimation of IB rates in classic paradigms. However, part of our purpose in writing this paper is to push back against a certain binary way of thinking about seeing/awareness. Our sense is that the field has conceived of awareness as “all or nothing”: You either see a perfectly clear gorilla right in front of you, or you see nothing at all. Our perspective is different: We think there can be degraded forms of awareness that fall into neither of those categories. For that reason, we are disinclined to see our results in the way that the reviewer suggests, namely as simply indicating that fewer subjects fail to see the stimulus than previously assumed. To think that way is, in our view, to assume the orthodox binary position about awareness. If, instead, one conceives of awareness as we do (and as we believe the framework of signal detection theory should compel us to), then it isn’t quite right to think of the proportion of subjects who were aware, but rather (e.g.) the sensitivity of subjects to the relevant stimulus. This is why we prefer measures like d′ over % noticing and % correct. We understand that the reviewer may not think the same way about this issue as we do, but part of our goal is to promote that way of thinking in general, and so some of our comments below reflect that perspective and approach.

For example, consider how we’d think about the single subject case where the task is 2afc detection of a low contrast stimulus in noise. Suppose that this subject achieves 70% correct. One way of thinking about that is that the subject sees the stimulus on 40% of trials (achieving 100% correct on those) and guesses blindly on the other 60% (achieving 50% correct on those) for a total of 40% + 30% = 70% overall. However, this is essentially a “high threshold” approach to the problem, in contrast to an SDT approach. On an SDT approach (an approach with tremendous evidential support), on every trial the subject receives samples from probabilistic distributions corresponding to each interval (one noise and one signal + noise) and determines which is higher according to the 2afc decision rule. Thus, across trials they have access to differentially graded information about the stimulus. Moreover, on some trials they may have significant information from the stimulus (perhaps, well above their single interval detection criterion) but still decide incorrectly because of high noise from the other spatial interval. From this perspective, there is no non-arbitrary way of saying whether the subject saw/did not see on a given trial. Instead, we must characterize the subject’s overall sensitivity to the stimulus/its visibility to them in terms of a parameter such as d′ (here, ~ 0.7).

We take the same attitude to our super subject. Instead of saying that some subjects saw/failed to see the stimuli, instead we suggest that the best way to characterize our results is that across subjects (and so trials also) there was differential graded access to information from the stimulus best represented in terms of the group-level sensitivity parameter d′.

We acknowledge that (despite ourselves) we occasionally fell into an all-too-natural binary/high threshold way of thinking, as when we suggested that our data show that “inattentionally blind subjects consciously perceive these stimuli after all” and “the inattentionally blind can see after all." (p.17) We have removed such problematic phrasing as well as other problematic phrasing as noted below.

(2) Because classic IB paradigms involve only one critical trial per subject, the authors used a "super subject" approach to estimate sensitivity (d') and response criterion (c) according to signal detection theory (SDT). Some readers may have issues with this super subject approach, but my main concern is with the lack of precision used by the authors when interpreting the results from this super subject analysis.

Only the super subject had above-chance sensitivity (and it was quite modest, with d' values between 0.07 and 0.51), but the authors over-interpret these results as applying to every subject. The methods and analyses cannot determine if any individual subject could report the features above-chance. Therefore, the following list of quotes should be revised for accuracy or removed from the paper as they are misleading and are not supported by the super subject analysis: "Altogether this approach reveals that subjects can report above-chance the features of stimuli (color, shape, and location) that they had claimed not to notice under traditional yes/no questioning" (p.6)

"In other words, nearly two-thirds of subjects who had just claimed not to have noticed any additional stimulus were then able to correctly report its location." (p.6)

"Even subjects who answer "no" under traditional questioning can still correctly report various features of the stimulus they just reported not having noticed, suggesting that they were at least partially aware of it after all." (p.8)

"Why, if subjects could succeed at our forced-response questions, did they claim not to have noticed anything?" (p.8)

"we found that observers could successfully report a variety of features of unattended stimuli, even when they claimed not to have noticed these stimuli." (p.14)

"our results point to an alternative (and perhaps more straightforward) explanation: that inattentionally blind subjects consciously perceive these stimuli after all... they show sensitivity to IB stimuli because they can see them." (p.16)

"In other words, the inattentionally blind can see after all." (p.17)

We thank the reviewer for pointing out how these quotations may be misleading as regards our central claim. We intended them all to be read generically as concerning the group, and not universally as claiming that all subjects could report above-chance/see the stimuli etc. We agree entirely that the latter universal claim would not be supported by our data. In contrast, we do contend that our super-subject analysis shows that, as a group, subjects traditionally considered intentionally blind exhibit residual sensitivity to features of stimuli (color, shape, and location) that they had all claimed not to notice, and likewise that as a group they could succeed at our forced-choice questions. 

To ensure this claim is clear throughout the paper, and that we are not interpreted as making an unsupported universal claim we have revised the language in all of the quotations above, as follows, as well as in numerous other places in the paper.

“Altogether this approach reveals that subjects can report above-chance the features of stimuli (color, shape, and location) that they had claimed not to notice under traditional yes/no questioning” (p.6) => “Altogether this approach reveals that as a group subjects can report above-chance the features of stimuli (color, shape, and location) that they had all claimed not to notice under traditional yes/no questioning” (p.6)

“Even subjects who answer “no” under traditional questioning can still correctly report various features of the stimulus they just reported not having noticed, suggesting that they were at least partially aware of it after all.” (p.8) => “... even subjects who answer “no” under traditional questioning can, as a group, still correctly report various features of the stimuli they just reported not having noticed, indicating significant group-level sensitivity to visual features. Moreover, these results are even consistent with an alternative hypothesis about IB, that as a group, subjects who would traditionally be classified as inattentionally blind are in fact at least partially aware of the stimuli they deny noticing.” (p.8)

“Why, if subjects could succeed at our forced-response questions, did they claim not to have noticed anything?” (p.8) => “Why, if subjects could succeed at our forcedresponse questions as a group, did they all individually claim not to have noticed anything?” (p.8)

“we found that observers could successfully report a variety of features of unattended stimuli, even when they claimed not to have noticed these stimuli.” (p.14) => “we found that groups of observers could successfully report a variety of features of unattended stimuli, even when they all individually claimed not to have noticed those stimuli.” (p.14)

“our results point to an alternative (and perhaps more straightforward) explanation: that inattentionally blind subjects consciously perceive these stimuli after all... they show sensitivity to IB stimuli because they can see them.” (p.16) => “our results just as easily raise an alternative (and perhaps more straightforward) explanation: that inattentionally blind subjects may retain a degree of awareness of these stimuli after all.” (p.16) Here deleting: “they show sensitivity to IB stimuli because they can see them.”

“In other words, the inattentionally blind can see after all.” (p.17) => “In other words, as a group, the inattentionally blind enjoy at least some degraded or partial sensitivity to the location, color and shape of stimuli which they report not noticing.” (p.17)

In one case, we felt the sentence was correct as it stood, since it simply reported a fact about our data:

“In other words, nearly two-thirds of subjects who had just claimed not to have noticed any additional stimulus were then able to correctly report its location.” (p.6)

After all, if subjects were entirely blind and simply guessed, it would be true to say that 50% of subjects would be able to correctly report the stimulus location (by guessing).

In addition to these and numerous other changes, we also added the following explicit statement early in the paper to head-off any confusion on this point: “Note that all analyses reported here relate to this super subject as opposed to individual subjects”. 

(3) In addition to the d' values for the super subject being slightly above zero, the authors attempted an analysis of response bias to further question the existence of IB. By including in some of their experiments critical trials in which no critical stimulus was presented, but asking subjects the standard Y/N IB question anyway, the authors obtained false alarm and correct rejection rates. When these FA/CR rates are taken into account along with hit/miss rates when critical stimuli were presented, the authors could calculate c (response criterion) for the super subject. Here, the authors report that response criteria are biased towards saying "no, I didn't notice anything". However, the validity of applying SDT to classic Y/N IB questioning is questionable.

For example, with the subject numbers provided in Box 1 (the 2x2 table of hits/misses/FA/CR), one can ask, 'how many subjects would have needed to answer "yes, I noticed something unusual" when nothing was presented on the screen in order to obtain a non-biased criterion estimate, i.e., c = 0?' The answer turns out to be 800 subjects (out of the 2761 total subjects in the stimulus-absent condition), or 29% of subjects in this condition.

In the context of these IB paradigms, it is difficult to imagine 29% of subjects claiming to have seen something unusual when nothing was presented. Here, it seems that we may have reached the limits of extending SDT to IB paradigms, which are very different than what SDT was designed for. For example, in classic psychophysical paradigms, the subject is asked to report Y/N as to whether they think a threshold-level stimulus was presented on the screen, i.e., to detect a faint signal in the noise. Subjects complete many trials and know in advance that there will often be stimuli presented and the stimuli will be very difficult to see. In those cases, it seems more reasonable to incorrectly answer "yes" 29% of the time, as you are trying to detect something very subtle that is out there in the world of noise. In IB paradigms, the stimuli are intentionally designed to be highly salient (and unusual), such that with a tiny bit of attention they can be easily seen. When no stimulus is presented and subjects are asked about their own noticing (especially of something unusual), it seems highly unlikely that 29% of them would answer "yes", which is the rate of FAs that would be needed to support the null hypothesis here, i.e., of a non-biased criterion. For these reasons, the analysis of response bias in the current context is questionable and the results claiming to demonstrate a biased criterion do not provide convincing evidence against IB.

We are grateful to the reviewer for highlighting this aspect of our data. We agree with several of these points. For example, it is indeed striking that — given the corresponding hit rate — a false alarm rate of 29% would be needed to obtain an unbiased criterion. At the same time, we would respectfully push back on other points above. In our first experiment that uses the super-subject analysis, for example, d′ is 0.51 and highly significant; to describe that figure, as the reviewer does, as “slightly above zero” seemed not quite right to us (and all the more so given that these experiments involve very large samples and preregistered analysis plans). 

We also respectfully disagree that our data call into question the validity of applying SDT to classic yes/no IB questioning. The mathematical foundations of SDT are rock solid, and have been applied far more broadly than we have applied them here. In fact, in a way we would suggest that exactly the opposite attitude is appropriate: rather than thinking that IB challenges an immensely well-supported, rigorously tested and broadly applicable mathematical model of perception, we think that the conflict between our SDT-based model of IB and the standard interpretation constitutes strong reason to disfavor the standard interpretation. Several points are worth making here.

First, it is already surprising that 11.03% of our subjects in E2 (46/417) and 7.24% of our subjects in E5 (200/2761) E5 reported noticing a stimulus when no stimulus was present. But while this may have seemed unlikely in advance of inquiry, this is in fact what the data show and forms the basis of our criterion calculations. Thus, our criterion calculations already factor in a surprising but empirically verified high false alarm rate of subjects answering “yes” when no stimulus was presented and were asked about their noticing. (We also note that the only paper we know of to report a false alarm rate in an IB paradigm, though not one used to calculate a response criterion, found a very consistent false alarm rate of 10.4%. See Devue et al. 2009.)

Second, while the reviewer is of course correct that a common psychophysical paradigm involves detection of a “threshold-level”/faint stimulus in noise, it is widely recognized that SDT has an extremely broad application, being applicable to any situation in which two kinds of event are to be discriminated (Pastore & Scheirer 1975) and being “almost universally accepted as a theoretical account of decision making in research on perceptual detection and recognition and in numerous extensions to applied domains” quite generally (Estes 2002, see also: Wixted 2020). Indeed, cases abound in which SDT has been successfully applied to situations which do not involve near threshold stimuli in noise. To pick two examples at random, SDT has been used in studying acceptability judgments in linguistics (Huang and Ferreira 2020) and the assessment of physical aggression in childstudent interactions (Lerman et al. 2010; for more general discussion of practical applications, see Swets et al. 2000). Given that the framework of SDT is so widely applied and well supported, and that we see no special reason to make an exception, we believe it can be relied on in the present context.

Finally, we note that inattentional blindness can in many ways be considered analogous to “near threshold” detection since inattention is precisely thought to degrade or even abolish awareness of stimuli, meaning that our stimuli can be construed as near threshold in the relevant sense. Indeed, our relatively modest d′ values suggest that under inattention stimuli are indeed hard to detect. Thus, even were SDT more limited in its application, we think it still would be appropriate to apply to the case of IB.

(4) One of the strongest pieces of evidence presented in the entire paper is the single data point in Figure 3e showing that in Experiment 3, even the super subject group that rated their non-noticing as "highly confident" had a d' score significantly above zero. Asking for confidence ratings is certainly an improvement over simple Y/N questions about noticing, and if this result were to hold, it could provide a key challenge to IB. However, this result hinges on a single data point, it was not replicated in any of the other 4 experiments, and it can be explained by methodological limitations. I strongly encourage the authors (and other readers) to follow up on this result, in an in-person experiment, with improved questioning procedures.

We agree that our finding that even the super-subject group that rated their non-noticing as “highly confident” had a d' score significantly above zero is an especially strong piece of evidence, and we thank the reviewer for highlighting that here. At the same time, we note that while the finding is represented by a single marker in Figure 3e, it seemed not quite right to call this a “single data point”, as the reviewer does, given that it derives from a large pre-registered experiment involving some 7,000 subjects total, with over 200 subjects in the relevant bin — both figures being far larger than a typical IB experiment. It would of course be tremendous to follow up on this result – and we certainly hope our work inspires various follow-up studies. That said, we note that recruiting the necessary numbers of in person subjects would be an absolutely enormous, career-level undertaking – it would involve bringing more than the entire undergraduate population at our own institution, Johns Hopkins, into our laboratory! While those results would obviously be extremely valuable, we wouldn’t want to read the reviewer’s comments as implying that only an experiment of that magnitude — requiring thousands upon thousands of in-person subjects — could make progress on these issues. Indeed, because every subject can only contribute one critical trial in IB, it has long been recognized as an extremely challenging paradigm to study in a sufficiently well-powered and psychophysically rigorous way. We believe that our large preregistered online approach represents a major leap forward here, even if it involves certain trade-offs.

In the current Experiment 3, the authors asked the standard Y/N IB question, and then asked how confident subjects were in their answer. Asking back-to-back questions, the second one with a scale that pertains to the first one (including a tricky inversion, e.g., "yes, I am confident in my answer of no"), may be asking too much of some subjects, especially subjects paying half-attention in online experiments. This procedure is likely to introduce a sizeable degree of measurement error.

An easy fix in a follow-up study would be to ask subjects to rate their confidence in having noticed something with a single question using an unambiguous scale:

On the last trial, did you notice anything besides the cross?

(1): I am highly confident I didn't notice anything else

(2): I am confident I didn't notice anything else

(3): I am somewhat confident I didn't notice anything else

(4): I am unsure whether I noticed anything else

(5): I am somewhat confident I noticed something else

(6): I am confident I noticed something else

(7): I am highly confident I noticed something else

If we were to re-run this same experiment, in the lab where we can better control the stimuli and the questioning procedure, we would most likely find a d' of zero for subjects who were confident or highly confident (1-2 on the improved scale above) that they didn't notice anything. From there on, the d' values would gradually increase, tracking along with the confidence scale (from 3-7 on the scale). In other words, we would likely find a data pattern similar to that plotted in Figure 3e, but with the first data point on the left moving down to zero d'. In the current online study with the successive (and potentially confusing) retrospective questioning, a handful of subjects could have easily misinterpreted the confidence scale (e.g., inverting the scale) which would lead to a mixture of genuine high-confidence ratings and mistaken ratings, which would result in a super subject d' that falls between zero and the other extreme of the scale (which is exactly what the data in Fig 3e shows).

One way to check on this potential measurement error using the existing dataset would be to conduct additional analyses that incorporate the confidence ratings from the 2AFC location judgment task. For example, were there any subjects who reported being confident or highly confident that they didn't see anything, but then reported being confident or highly confident in judging the location of the thing they didn't see? If so, how many? In other words, how internally (in)consistent were subjects' confidence ratings across the IB and location questions? Such an analysis could help screen-out subjects who made a mistake on the first question and corrected themselves on the second, as well as subjects who weren't reading the questions carefully enough.

As far as I could tell, the confidence rating data from the 2AFC location task were not reported anywhere in the main paper or supplement.

We are grateful to the reviewer for raising this issue and for requesting that we report the confidence rating data from our 2afc location task in Experiment 3. We now report all this data in our Supplementary Materials (see Supplementary Table 3).

We of course agree with the reviewer’s concern about measurement error, which is a concern in all experiments. What, then, of the particular concern that some subjects might have misunderstood our confidence question? It is surely impossible in principle to rule out this possibility; however, several factors bear on the plausibility of this interpretation. First, we explicitly labeled our confidence scale (with 0 labeled as ‘Not at all confident’ and 3 as ‘Highly confident’) so that subjects would be very unlikely simply to invert the scale. This is especially so as it is very counterintuitive to treat “0” as reflecting high confidence. However, we accept that it is a possibility that certain subjects might nonetheless have been confused in some other way.

So, we also took a second approach. We examined the confidence ratings on the 2afc question of subjects who reported being highly confident that they didn't notice anything.

Reassuringly, the large majority of these high confidence “no” subjects (~80%) reported low confidence of 0 or 1 on the 2afc question, and the majority (51%) reported the lowest confidence of 0. Only 18/204 (9%) subjects reported high confidence on both questions. 

Still, the numbers of subjects here are small and so may not be reliable. This led us to take a third approach. We reasoned that any measurement error due to inverting or misconstruing the confidence scale should be symmetric. That is: If subjects are liable to invert the confidence scale, they should do so just as often when they answer “yes” as when they answer “no” – after all the very same scale is being used in both cases. This allows us to explore evidence of measurement error in relation to the much larger number of highconfidence “yes” subjects (N = 2677), thus providing a much more robust indicator as to whether subjects are generally liable to misconstrue the confidence scale. Looking at the number of such high confidence noticers who subsequently respond to the 2afc question with low-confidence, we found that the number was tiny. Only 28/2677 (1.05%) of highconfidence noticers subsequently gave the lowest level of confidence on the 2afc question, and only 63/2677 (2.35%) subjects gave either of the two lower levels of confidence. In this light, we consider any measurement error due to misunderstanding the confidence scale to be extremely minimal.

What should we make of the 18 subjects who were highly confident non-noticers but then only low-confidence on the 2afc question? Importantly, we do not think that these 18 subjects necessarily made a mistake on the first question and so should be excluded. There is no a priori reason why one’s confidence criterion in a yes/no question should carry over to a 2afc question. After all, it is perfectly rationally coherent to be very confident that one didn’t see anything but also very confident that if there was anything to be seen, it was on the left. Moreover, these 18 subjects were not all correct on the 2afc question despite their high confidence (4/18 or 22% getting the wrong answer). 

Nonetheless, and again reassuringly, we found that the above-chance patterns in our data remained the same even excluding these 18 subjects. We did observe a slight reduction in percent correct and d′ but this is absolutely what one should expect since excluding the most confident performers in any task will almost inevitably reduce performance.

In this light, we consider it unlikely that measurement error fully explains the residual sensitivity found even amongst highly confident non-noticers. That said, we appreciate this concern. We now raise the issue and the analysis of high confidence noticers which addresses it in our revised manuscript. We also thank the reviewer for pressing us to think harder about this issue, which led directly to these new analyses that we believed have strengthened the paper.

(5) In most (if not all) IB experiments in the literature, a partial attention and/or full attention trial (or set of trials) is administered after the critical trial. These control trials are very important for validating IB on the critical trial, as they must show that, when attended, the critical stimuli are very easy to see. If a subject cannot detect the critical stimulus on the control trial, one cannot conclude that they were inattentionally blind on the critical trial, e.g., perhaps the stimulus was just too difficult to see (e.g., too weak, too brief, too far in the periphery, too crowded by distractor stimuli, etc.), or perhaps they weren't paying enough attention overall or failed to follow instructions. In the aggregate data, rates of noticing the stimuli should increase substantially from the critical trial to the control trials. If noticing rates are equivalent on the critical and control trials one cannot conclude that attention was manipulated.

It is puzzling why the authors decided not to include any control trials with partial or full attention in their five experiments, especially given their online data collection procedures where stimulus size, intensity, eccentricity, etc. were uncontrolled and variable across subjects. Including such trials could have actually helped them achieve their goal of challenging the IB hypothesis, e.g., excluding subjects who failed to see the stimulus on the control trials might have reduced the inattentional blindness rates further. This design decision should at least be acknowledged and justified (or noted as a limitation) in a revision of this paper.

We acknowledge that other studies in the literature include divided and full attention trials, and that they could have been included in our work as well. However, we deliberately decided not to include such control trials for an important reason. As the referee comments, the main role of such trials in previous work has been to exclude from analysis subjects who failed to report the unexpected stimulus on the divided and/or full attention control trials.

(For example, as Most et al. 2001 write: “Because observers should have seen the object in the full-attention trial (Mack & Rock, 1998), we used this trial as a control … Accordingly, 3 observers who failed to see the cross on this trial were replaced, and their data were excluded from the analyses.") As the reviewer points out, excluding such subjects would very likely have ‘helped' us. However, the practice is controversial. Indeed, in a review of 128 experiments, White et al. 2018 argue that the practice has “problematic consequences” and “may lead researchers to understate the pervasiveness of inattentional blindness". Since we wanted to offer as simple and demanding a test of residual sensitivity in IB as possible, we thus decided not to use any such exclusions, and for that reason decided not to include divided/full attention trials. 

As recommended, we discuss this decision not to include divided/full attention trials and our logic for not doing so in the manuscript. As we explain, not having those conditions makes it more impressive, not less impressive, that we observed the results we in fact did — it makes our results more interpretable, not less interpretable, and so absence of such conditions from our manuscript should not (in our view) be considered any kind of weakness.

(6) In the discussion section, the authors devote a short paragraph to considering an alternative explanation of their non-zero d' results in their super subject analyses: perhaps the critical stimuli were processed unconsciously and left a trace such that when later forced to guess a feature of the stimuli, subjects were able to draw upon this unconscious trace to guide their 2AFC decision. In the subsequent paragraph, the authors relate these results to above-chance forced-choice guessing in blindsight subjects, but reject the analogy based on claims of parsimony.

First, the authors dismiss the comparison of IB and blindsight too quickly. In particular, the results from experiment 3, in which some subjects adamantly (confidently) deny seeing the critical stimulus but guess a feature at above-chance levels (at least at the super subject level and assuming the online subjects interpreted and used the confidence scale correctly), seem highly analogous to blindsight. Importantly, the analogy is strengthened if the subjects who were confident in not seeing anything also reported not being confident in their forced-choice judgments, but as mentioned above this data was not reported.

Second, the authors fail to mention an even more straightforward explanation of these results, which is that ~8% of subjects misinterpreted the "unusual" part of the standard IB question used in experiments 1-3. After all, colored lines and shapes are pretty "usual" for psychology experiments and were present in the distractor stimuli everyone attended to. It seems quite reasonable that some subjects answered this first question, "no, I didn't see anything unusual", but then when told that there was a critical stimulus and asked to judge one of its features, adjusted their response by reconsidering, "oh, ok, if that's the unusual thing you were asking about, of course I saw that extra line flash on the left of the screen". This seems like a more parsimonious alternative compared to either of the two interpretations considered by the authors: (1) IB does not exist, (2) super-subject d' is driven by unconscious processing. Why not also consider: (3) a small percentage of subjects misinterpreted the Y/N question about noticing something unusual. In experiments 4-5, they dropped the term "unusual" but do not analyze whether this made a difference nor do they report enough of the data (subject numbers for the Y/N question and 2AFC) for readers to determine if this helped reduce the ~8% overestimate of IB rates.

Our primary ambition in the paper was to establish, as our title suggests, residual sensitivity in IB. The ambition is quite neutral as to whether the sensitivity reflects conscious or unconscious processing (i.e. is akin to blindsight as traditionally conceived). We were evidently not clear about this, however, leading to two referees coming away with an impression of our claims that is different than we intended. We have revised our manuscript throughout to address this. But we also want to emphasize here that we take our data primarily to support the more modest claim that there is residual sensitivity (conscious or unconscious) in the group of subjects who are traditionally classified as inattentionally blind. We believe that this claim has solid support in our data.

We do in the discussion section offer one reason for believing that there is residual awareness in the group of subjects who are traditionally classified as inattentionally blind. However, we acknowledge that this is controversial and now emphasize in the manuscript that this claim “is tentative and secondary to our primary finding”. We also emphasize that part of our point is dialectical: Inattentional blindness has been used to argue (e.g.) that attention is required for awareness. We think that our data concerning residual sensitivity at least push back on the use of IB to make this claim, even if they do not provide decisive evidence (as we agree) that awareness survives inattention. (Cf. here, Hirshhorn et al. 2024 who take up a common suggestion in the field that awareness is best assessed by using both subjective and objective measures, with claims about lack of awareness ideally being supported by both; our data suggest at a minimum that in IB objective measures do not neatly line up with subjective measures.)

We hope this addresses the referee’s concern that we dismiss the “the comparison of IB and blindsight too quickly”. We do not intend to dismiss that comparison at all, indeed we raise it because we consider it a serious hypothesis. Our aim is simply to raise one possible consideration against it. But, again, our main claim is quite consistent with sensitivity in IB being akin to “blindsight”.

We also agree with the referee that a possible explanation of why some subjects say they do not notice something unusual in IB paradigms, is not because they didn’t notice anything but because they didn’t consider the unexpected stimulus sufficiently unusual. However, the reviewer is incorrect that we did not mention this interpretation; to the contrary, it was precisely the kind of concern which led us to be dissatisfied with standard IB methods and so motivated our approach. As we wrote in our main text: “However, yes/no questions of this sort are inherently and notoriously subject to bias…   For example, observers might be under-confident whether they saw anything (or whether what they saw counted as unusual); this might lead them to respond “no” out of an excess of caution.” On our view, this is exactly the kind of reason (among other reasons) that one cannot rely on yes/no reports of noticing unusual stimuli, even though the field has relied on just these sorts of questions in just this way.

We do not, however, think that this explanation accounts for why all subjects fail to report noticing, nor do we think that it accounts for our finding of above-chance sensitivity amongst non-noticers. This is for two critical reasons. First, whereas the word “unusual” did appear in the yes/no question in our Experiments 1-3, it did not appear in our Experiments 4 and 5 on dynamic IB. (In both cases, we used the exact wording of such questions in the experiments we were basing our work on.) And, of course, we still found significant residual sensitivity amongst non-noticers in Experiments 4 and 5. Second, in relation to our confidence experiment, we think it unlikely that subjects who were highly confident that they did not notice anything unusual only said that because they thought what they had seen was insufficiently unusual. Yet even in this group of subjects who were maximally confident that they did not notice anything unusual, we still found residual sensitivity.

(7) The authors use sub-optimal questioning procedures to challenge the existence of the phenomenon this questioning is intended to demonstrate. A more neutral interpretation of this study is that it is a critique on methods in IB research, not a critique on IB as a manipulation or phenomenon. The authors neglect to mention the dozens of modern IB experiments that have improved upon the simple Y/N IB questioning methods. For example, in Michael Cohen's IB experiments (e.g., Cohen et al., 2011; Cohen et al., 2020; Cohen et al., 2021), he uses a carefully crafted set of probing questions to conservatively ensure that subjects who happened to notice the critical stimuli have every possible opportunity to report seeing them. In other experiments (e.g., Hirschhorn et al., 2024; Pitts et al., 2012), researchers not only ask the Y/N question but then follow this up by presenting examples of the critical stimuli so subjects can see exactly what they are being asked about (recognition-style instead of free recall, which is more sensitive). These follow-up questions include foil stimuli that were never presented (similar to the stimulus-absent trials here), and ask for confidence ratings of all stimuli. Conservative, pre-defined exclusion criteria are employed to improve the accuracy of their IB-rate estimates. In these and other studies, researchers are very cautious about trusting what subjects report seeing, and in all cases, still find substantial IB rates, even to highly salient stimuli. The authors should consider at least mentioning these improved methods, and perhaps consider using some of them in their future experiments.

The concern that we do not sufficiently discuss the range of “improved” methods in IB studies is well-taken. A similar concern is raised by Reviewer #2 (Dr. Cohen). To address the concern, we have added to our manuscript a substantial new discussion of such improved methods. However, although we do agree that these methods can be helpful and may well address some of the methodological concerns which our paper raises, we do not think that they are a panacea. Thus, our discussion of these methods also includes a substantial discussion of the problems and pitfalls with such methods which led us to favor our own simple forced-response and 2afc questions, combined with SDT analysis. We think this approach is superior both to the classic approach in IB studies and to the approach raised by the reviewers.

In particular, we have four main concerns about the follow up questions now commonly used in the field:

First, many follow up questions are used not to exclude people from the IB group but to include people in the IB group. Thus, Most et al. 2001 asked follow up questions but used these to increase their IB group, only excluding subjects from the IB group if they both reported seeing and answered their follow ups incorrectly: “Observers were regarded as having seen the unexpected object if they answered 'yes' when asked if they had seen anything on the critical trial that had not been present before and if they were able to describe its color, motion, or shape." This means that subjects who saw the object but failed to see its color, say, would be treated as inattentionally blind. This has the purpose of inflating IB rates, in exactly the way our paper is intended to critique. So, in our view this isn’t an improvement but rather part of the approach we take issue with.

Second, many follow up questions remain yes/no questions or nearby variants, all of which are subject to response bias. For example, in Cohen’s studies which the reviewer mentions, it is certainly true that “he uses a carefully crafted set of probing questions to conservatively ensure that subjects who happened to notice the critical stimuli have every possible opportunity to report seeing them.” We agree that this improves over a simple yes/no question in some ways. However, such follow up probes nonetheless remain yes/no questions, subject to response bias, e.g.:

(1) “Did you notice anything strange or different about that last trial?”

(2) “If I were to tell you that we did something odd on the last trial, would you have a guess as to what we did?”

(3) “If I were to tell you we did something different in the second half of the last trial, would you have a guess as to what we did?”

(4) “Did you notice anything different about the colors in the last scene?”

Indeed, follow up questions of this kind can be especially susceptible to bias, since subjects may be reluctant to “take back” their earlier answers and so be conservative in responding positively to avoid inconsistency or acknowledgement of earlier error. This may explain why such follow up questions produce remarkable consistency despite their rather different wording. Thus, Simons and Chabris (1999) report: “Although we asked a series of questions escalating in specificity to determine whether observers had noticed the unexpected event, only one observer who failed to report the event in response to the first question (“did you notice anything unusual?'') reported the event in response to any of the next three questions (which culminated in “did you see a ... walk across the screen?''). Thus, since the responses were nearly always consistent across all four questions, we will present the results in terms of overall rates of noticing.” Thus, while there are undoubtedly merits to these follow ups, they do not resolve problems of bias.

This same basic issue affects the follow up question used in Pitts et al. 2012 which the reviewer mentions. Pitts et al. write: “If a participant reported not seeing any patterns and rated their confidence in seeing the square pattern (once shown the sample) as a 3 or less (1 = least confident, 5 = most confident), she or he was placed in Group 1 and was considered to be inattentionally blind to the square patterns.” The confidence rating follow-up question here remains subject to bias. Moreover, and strikingly, the inclusion criterion used means that subjects who were moderately confident that they saw the square pattern when shown (i.e. answered 3) were counted as inattentionally blind (!). We do not think this is an appropriate inclusion criterion.

The third problem is that follow up questions are often free/open-response. For instance, Most et al. (2005) ask the follow up question: "If you did see something on the last trial that had not been present during the first two trials, what color was it? If you did not see something, please guess." This is a much more difficult and to that extent less sensitive question than our binary forced-response/2afc questions. For this reason, we believe our follow up questions are more suitable for ascertaining low levels of sensitivity.

The fourth and final issue is that whereas 2afc questions are criterion free (in that they naturally have an unbiased decision rule), this is in fact not true of n_afc questions in general, nor is it true in general of _delayed n-alternative match to sample designs. Thus, even when limited response options are given, they are not immune to response biases and so require SDT analysis. Moreover, some such tasks can involve decision spaces which are often poorly understood or difficult to analyze without making substantial assumptions about observer strategy. 

This last point (as well as the first) is relevant to Hirshhorn et al. 2024. Hirshhorn et al. write that they “used two awareness measures. Firstly, participants were asked to rate stimulus visibility on the Perceptual Awareness Scale (PAS, a subjective measure of awareness: Ramsøy & Overgaard, 2004), and then they were asked to select the stimulus image from an array of four images (an objective measure: Jakel & Wichmann, 2006).”

While certainly an improvement on simple yes/no questioning, the PAS remains subject to response bias. On the other hand, we applaud Hirshhorn et al.’s use of objective measures in the context of IB which of course our design implements. However, while Hirshhorn et al. 2024 suggest that their task is a spatial 4afc following the recommendation of this design by Jakel & Wichmann (2006), it is strictly a 4-alternative delayed match to sample task, so it is doubtful if it can be considered a preferred psychophysical task for the reasons Jakel & Wichmann offer. Regardless, the more crucial point is that observers in such a task might be biased towards one alternative as opposed to another. Thus, use of d′ (as opposed to percent correct as in Hirshhorn et al. 2024) is crucial in assessing performance in such tasks.

For all these reasons, then, while we agree that the field has taken significant steps to move beyond the simple yes/no question traditionally used in IB studies (and we have revised our manuscript to make this clear); we do not think it has resolved the methodological issues which our paper seeks to highlight and address, and we believe that our approach contributes something additional that is not yet present in the literature. We have now revised our manuscript to make these points much more clearly, and we thank the reviewer for prompting these improvements.

Reviewer #2 (Public review):

In this study, Nartker et al. examine how much observers are conscious of using variations of classic inattentional blindness studies. The key idea is that rather than simply asking observers if they noticed a critical object with one yes/no question, the authors also ask follow-up questions to determine if observers are aware of more than the yes/no questions suggest. Specifically, by having observers make forced choice guesses about the critical object, the authors find that many observers who initially said "no" they did not see the object can still "guess" above chance about the critical object's location, color, etc. Thus, the authors claim, that prior claims of inattentional blindness are mistaken and that using such simple methods has led numerous researchers to overestimate how little observers see in the world. To quote the authors themselves, these results imply that "inattentionally blind subjects consciously perceive these stimuli after all... they show sensitivity to IB stimuli because they can see them."

Before getting to a few issues I have with the paper, I do want to make sure to explicitly compliment the researchers for many aspects of their work. Getting massive amounts of data, using signal detection measures, and the novel use of a "super subject" are all important contributions to the literature that I hope are employed more in the future.

We really appreciate this comment and that the reviewer found our work to make these important contributions to the literature. We wrote this paper expecting not everyone to accept our conclusions, but hoping that readers would see the work as making a valuable contribution to the literature promoting an underexplored alternative in a compelling way. Given that this reviewer goes on to express some skepticism about our claims, it is especially encouraging to see this positive feedback up top!

Main point 1: My primary issue with this work is that I believe the authors are misrepresenting the way people often perform inattentional blindness studies. In effect, the authors are saying, "People do the studies 'incorrectly' and report that people see very little. We perform the studies 'correctly' and report that people see much more than previously thought." But the way previous studies are conducted is not accurately described in this paper. The authors describe previous studies as follows on page 3:

"Crucially, however, this interpretation of IB and the many implications that follow from it rest on a measure that psychophysics has long recognized to be problematic: simply asking participants whether they noticed anything unusual. In IB studies, awareness of the unexpected stimulus (the novel shape, the parading gorilla, etc.) is retroactively probed with a yes/no question, standardly, "Did you notice anything unusual on the last trial which wasn't there on previous trials?". Any subject who answers "no" is assumed not to have any awareness of the unexpected stimulus.

If this quote were true, the authors would have a point. Unfortunately, I do not believe it is true. This is simply not how many inattentional blindness studies are run. Some of the most famous studies in the inattentional blindness literature do not simply as observes a yes/no question (e.g., the invisible gorilla (Simons et al. 1999), the classic door study where the person changes (Simons and Levin, 1998), the study where observers do not notice a fight happening a few feet from them (Chabris et al., 2011). Instead, these papers consistently ask a series of follow-up questions and even tell the observers what just occurred to confirm that observers did not notice that critical event (e.g., "If I were to tell you we just did XYZ, did you notice that?"). In fact, after a brief search on Google Scholar, I was able to relatively quickly find over a dozen papers that do not just use a yes/no procedure, and instead as a series of multiple questions to determine if someone is inattentionally blind. In no particular order some papers (full disclosure: including my own):

(1) Most et al. (2005) Psych Review

(2) Drew et al. (2013) Psych Science

(3) Drew et al. (2016) Journal of Vision

(4) Simons et al. (1999) Perception

(5) Simons and Levin (1998) Perception

(6) Chabris et al. (2011) iPerception

(7) Ward & Scholl (2015) Psych Bulletin and Review

(8) Most et al. (2001) Psych Science

(9) Todd & Marois (2005) Psych Science

(10) Fougnie & Marois (2007) Psych Bulletin and Review

(11) New and German (2015) Evolution and Human Behaviour

(12) Jackson-Nielsen (2017) Consciousness and cognition

(13) Mack et al. (2016) Consciousness and cognition

(14) Devue et al. (2009) Perception

(15) Memmert (2014) Cognitive Development

(16) Moore & Egeth (1997) JEP:HPP

(17) Cohen et al. (2020) Proc Natl Acad Sci

(18) Cohen et al. (2011) Psych Science

This is a critical point. The authors' key idea is that when you ask more than just a simple yes/no question, you find that other studies have overestimated the effects of inattentional blindness. But none of the studies listed above only asked simple yes/no questions. Thus, I believe the authors are mis-representing the field. Moreover, many of the studies that do much more than ask a simple yes/no question are cited by the authors themselves! Furthermore, as far as I can tell, the authors believe that if researchers do these extra steps and ask more follow-ups, then the results are valid. But since so many of these prior studies do those extra steps, I am not exactly sure what is being criticized.

To make sure this point is clear, I'd like to use a paper of mine as an example. In this study (Cohen et al., 2020, Proc Natl Acad Sci USA) we used gaze-contingent virtual reality to examine how much color people see in the world. On the critical trial, the part of the scene they fixated on was in color, but the periphery was entirely in black and white. As soon as the trial ended, we asked participants a series of questions to determine what they noticed. The list of questions included:

(1) "Did you notice anything strange or different about that last trial?"

(2) "If I were to tell you that we did something odd on the last trial, would you have a guess as to what we did?"

(3) "If I were to tell you we did something different in the second half of the last trial, would you have a guess as to what we did?"

(4) "Did you notice anything different about the colors in the last scene?"

(5) We then showed observers the previous trial again and drew their attention to the effect and confirmed that they did not notice that previously.

In a situation like this, when the observers are asked so many questions, do the authors believe that "the inattentionally blind can see after all?" I believe they would not say that and the reason they would not say that is because of the follow-up questions after the initial yes/no question. But since so many previous studies use similar follow-up questions, I do not think you can state that the field is broadly overestimating inattentional blindness. This is why it seems to me to be a bit of a strawman: most people do not just use the yes/no method.

We appreciate this reviewer raising this issue. As he (Dr. Cohen) states, his “primary issue” concerns our discussion of the broader literature (which he worries understates recent improvements made to the IB methodology), rather than, e.g., the experiments we’ve run. We take this concern very seriously and address it comprehensively here.

A very similar issue is identified by Reviewer #1, comment (7). To review some of what we say in reply to them: To address the concern we have added to our manuscript a substantial new discussion of such improved methods. However, although we do agree that these methods can be helpful and may well address some of the methodological concerns which our paper raises, we do not think that they are a panacea. Thus, our discussion of these methods also includes a substantial discussion of the problems and pitfalls with such methods which led us to favor our own simple forced-response and 2afc questions, combined with SDT analysis. We think this approach is superior both to the classic approach in IB studies and to the approach raised by the reviewers.

In particular, we have three main concerns about the follow up questions now commonly used in the field:

First, many follow up questions are used not to exclude subjects from the IB group but to include subjects in the IB group. Thus, Most et al. (2001) asked follow up questions but used these to increase their IB group, only excluding subjects from the IB group if they both reported seeing and failed to answer their follow ups correctly: “Observers were regarded as having seen the unexpected object if they answered 'yes' when asked if they had seen anything on the critical trial that had not been present before and if they were able to describe its color, motion, or shape." This means that subjects who saw the object but failed to describe it in these respects would be treated as inattentionally blind. This is problematic since failure to describe a feature (e.g., color, shape) does not imply a complete lack of information concerning that feature; and even if a subject did lack all information concerning these features of an object, this would not imply a complete failure to see the object. Similarly, Pitts et al. (2012) asked subjects to rate their confidence in their initial yes/no response from 1 = least confident to 5 = most confident, and used these ratings to include in the IB group those who rated their confidence in seeing at 3 or less. This is evidently problematic, since there is a large gap between being under confident that one saw something and being completely blind to it. More generally, using follows up to inflate IB rates in such ways raises precisely the kinds of issues our paper is intended to critique. So in our view this isn’t an improvement but rather part of the approach we take issue with.

Second, many follow up questions remain yes/no questions or nearby variants, all of which are subject to response bias. For example, in the reviewer’s own studies (Cohen et al. 2020, 2011; see also: Simons et al., 1999; Most et al., 2001, 2005; Drew et al., 2013; Memmert, 2014) a series of follow up questions are used to try and ensure that subjects who noticed the critical stimuli are given the maximum opportunity to report doing so, e.g.:

(1) “Did you notice anything strange or different about that last trial?”

(2) “If I were to tell you that we did something odd on the last trial, would you have a guess as to what we did?”

(3) “If I were to tell you we did something different in the second half of the last trial, would you have a guess as to what we did?”

(4) “Did you notice anything different about the colors in the last scene?”

We certainly agree that such follow up questions improve over a simple yes/no question in some ways. However, such follow up probes nonetheless remain yes/no questions, intrinsically subject to response bias. Indeed, follow up questions of this kind can be especially susceptible to bias, since subjects may be reluctant to “take back” their earlier answers and so be conservative in responding positively to avoid inconsistency or acknowledgement of earlier error. This may explain why such follow up questions produce remarkable consistency despite their rather different wording. Thus, Simons and Chabris (1999) report: “Although we asked a series of questions escalating in specificity to determine whether observers had noticed the unexpected event, only one observer who failed to report the event in response to the first question (“did you notice anything unusual?'') reported the event in response to any of the next three questions (which culminated in “did you see a ... walk across the screen?''). Thus, since the responses were nearly always consistent across all four questions, we will present the results in terms of overall rates of noticing.” Thus, while there are undoubtedly merits to these follow ups, they do not resolve problems of bias.

It is also important to recognize that whereas 2afc questions are criterion free (in that they naturally have an unbiased decision rule), this is not true of n_afc nor delayed _n-alternative match to sample designs in general. Performance in such tasks thus requires SDT analysis – which itself may be problematic if the decision space is not properly understood or requires making substantial assumptions about observer strategy.

Third, and finally, many follow up questions are insufficiently sensitive (especially with small sample sizes). For instance, Todd, Fougnie & Marois (2005) used a 12-alternative match-tosample task (see similarly: Fougnie & Marois, 2007; Devue et al., 2009). And Most et al. (2005) asked an open-response follow-up: “If you did see something on the last trial that had not been present during the first two trials, what color was it? If you did not see something, please guess.” These questions are more difficult and to that extent less sensitive than binary forced-response/2afc questions of the sort we use in our own studies – a difference which may be critical in uncovering degraded perceptual sensitivity.

For all these reasons, then, while we agree that the field has taken significant steps to move beyond the simple yes/no question traditionally used in IB studies (and we have revised our manuscript to make this clear); we do not think it has resolved the methodological issues which our paper seeks to highlight and address, and we believe that our approach of using 2afc or forced-response questions combined with signal detection analysis is an important improvement on prior methods and contributes something additional that is not yet present in the literature. We have now revised our manuscript to make these points much clearer.

Other studies that improve on the standard methodology

This reviewer adds something else, however: A very helpful list of 18 papers which include follow ups and that he believes overcome many of the issues we raise in our paper. To just state our reaction bluntly: We are familiar with every one of these papers (indeed, one of them is a paper by one of us!), and while we think these are all very valuable contributions to the literature, it is our view that none of these 18 papers resolves the worries that led us to conduct our work.  

Here we briefly comment on the relevant pitfalls in each case. We hope this serves to underscore the importance of our methodological approach.

(1) Most et al. (2005) Psych Review

Either a 2-item or 5-item questionnaire was used. The 2-item questionnaire ran as follows:

(1) On the last trial, did you see anything other than the 4 circles and the 4 squares (anything that had not been present on the original two trials)? Yes No 

(2) If you did see something on the last trial that had not been present during the original two trials, please describe it in as much detail as possible.

This clearly does not substantially improve on the traditional simple yes/no question. Moreover, the second question (as well as being open-ended) was used to include additional subjects in the IB group, in that participants were counted as having seen the object only if they responded “yes” to Q1 and in addition “were able to report at least one accurate detail” in response to Q2. In other words, either a subject says “no” (and is treated as unaware), or says “yes” and then is asked to prove their awareness, as it were. If anything, this intensifies the concerns we raise, by inflating IB rates. 

The 5-item questionnaire looked like this: 

(1) On the last trial, did you see anything other than the black and white L’s and T’s (anything that had not been present on the first two trials)?

(2) If you did see something on the last trial that had not been present during the first two trials, please describe it.

(3) If you did see something on the last trial that had not been present during the first two trials, what color was it? If you did not see something, please guess. (Please indicate whether you did see something or are guessing)

(4) If you did see something during the last trial that had not been present in the first two trials, please draw an arrow on the “screen” below showing the direction in which it was moving. If you did not see something, please guess. (Please indicate whether you did see something or are guessing)

(5) If you did see something during the last trial that had not been present during the first two trials, please circle the shape of the object below [4 shapes are presented to choose from]. If you did not see anything, please guess. (Please indicate whether you did see something or are guessing)

Q5 was not used for analysis purposes. (It suffers from the second issue raised above.) Q1 is the traditional y/n question. Qs 2&3 are open ended. It is unclear how responses to Q4 were analyzed (at the limit it could be considered a helpful, forced-choice question – though it again would suffer from the second issue raised above). However, as noted with respect to the 2-item questionnaire, these responses were not used to exclude people from the IB group but to include people in it. So again, this approach does not in any way address the issues we are concerned about, and if anything, only makes them worse. 

(2)  Drew et al. (2013) Psych Science

All follow ups were yes/no: “we asked a series of questions to determine whether they noticed the gorilla: ‘Did the final trial seem any different than any of the other trials?’, ‘Did you notice anything unusual on the final trial?’, and, finally, ‘Did you see a gorilla on the final trial?’”. So, this paper essentially implements the standard methodology we mention (and criticize). 

(3)  Drew et al. (2016) Journal of Vision

Follow up questions were used, but the reported procedure does not provide sufficient details to evaluate them (we are only told: “After the final trial, they were asked: ‘On that last trial of the task, did you notice anything that was not there on previous trials?’ They then answered questions about the features of the unexpected stimulus on a separate screen (color, shape, movement, and direction of movement).”). It is not clear that these follow ups were used to exclude any subjects from the analysis. Finally, given that the unexpected object could be the same color as the targets/distractors, it is clear that biases would have been introduced which would need to be considered (but which were not).

(4)  Simons & Chabris (1999) Perception

All follow ups were yes/no: “observers were … asked to provide answers to a surprise series of additional questions. (i) While you were doing the counting, did you notice anything unusual on the video? (ii) Did you notice any- thing other than the six players? (iii) Did you see anyone else (besides the six players) appear on the video? (iv) Did you see a gorilla [woman carrying an umbrella] walk across the screen? After any “yes'' response, observers were asked to provide details of what they noticed. If at any point an observer mentioned the unexpected event, the remaining questions were skipped.” As noted previously, the analyses in fact did not use these questions to exclude subjects since answers were so consistent.

(5)  Simons and Levin (1998) Perception

This is a change detection paradigm, not a study of inattentional blindness. And in any case, one yes/no follow up was used: “Did you notice that I'm not the same person who approached you to ask for directions?”

(6)  Chabris et al. (2011) iPerception

Two yes/no questions were asked: “we asked whether the subjects had seen anything unusual along the route, and then whether they had seen anyone fighting.” It seems that follow up questions (a request to describe the fight) were asked only of those who said yes.

This is in fact a common procedure – follow up questions only being asked of the “yes” group. As discussed, it is sometimes used to increase rates of IB, compounding the problem we identify in our paper. So this is another example of a follow-up question that makes the problem we identify worse, not better.

(7) Ward & Scholl (2015) Psych Bulletin and Review

Two yes/no questions were used: “...observers were asked whether they noticed ‘anything … that was different from the first three trials’ — and if so, to describe what was different. They were then shown the gray cross and asked if they had noticed it—and if so, to describe where it was and how it moved. Only observers who explicitly reported not noticing the cross were counted as ‘nonnoticers’ to be included in the final sample (N = 100).” In each case, combining the traditional noticing question with a request to describe and identify may have induced conservative response biases in the noticing question, since a subject might consider being able to describe or identify the unexpected stimulus a precondition of giving a positive answer to the noticing question.

(8) Most et al. (2001) Psych Science

The same 5-item questionnaire discussed above in relation to Most et al. (2005) was used: 

(1) On the last trial, did you see anything other than the black and white L’s and T’s (anything that had not been present on the first two trials)?

(2)   If you did see something on the last trial that had not been present during the first two trials, please describe it.

(3) If you did see something on the last trial that had not been present during the first two trials, what color was it? If you did not see something, please guess. (Please indicate whether you did see something or are guessing)

(4) If you did see something during the last trial that had not been present in the first two trials, please draw an arrow on the “screen” below showing the direction in which it was moving. If you did not see something, please guess. (Please indicate whether you did see something or are guessing)

(5) If you did see something during the last trial that had not been present during the first two trials, please circle the shape of the object below [4 shapes are presented to choose from]. If you did not see anything, please guess. (Please indicate whether you did see something or are guessing)

Q5 was not used for analysis purposes. (It suffers from the second issue raised above.) Q1 is the traditional yes/no question. Qs 2&3 are open ended. It is unclear how responses to Q4 were analyzed (at the limit it could be considered a helpful, forced-choice question – though it again would suffer from the second issue raised above). However, as noted with respect to the two item questionnaire in Most et al. 2005, these responses were not used to exclude people from the IB group but to include people in it. So again this approach does not in any way address the issues we are concerned about, and if anything only makes them worse.

(9) Todd, Fougnie & Marois (2005) Psych Science

“participants were probed with three questions to determine whether they had detected the critical stimulus ... .The first question assessed whether subjects had seen anything unusual during the trial; they responded ‘‘yes’’ or ‘‘no’’ by pressing the appropriate key on the keyboard. The second question asked participants to select which stimulus they might have seen among 12 possible objects and symbols selected from MacIntosh font databases. The third question asked participants to select the quadrant in which the critical stimulus may have appeared by pressing one of four keys, each of which corresponded to one of the quadrants.”

These follow ups were used to include people in the IB group: “In keeping with previous studies (Most et al., 2001), participants were considered to have detected the critical stimulus successfully if they (a) reported seeing an unexpected stimulus and (b) correctly selected its quadrant location.” In line with our third point about sensitivity, the object identity test transpired to be “too difficult even under full-attention conditions … Thus, performance with this question was not analyzed further.”

(10) Fougnie & Marois (2007) Psych Bulletin and Review

Same exact methods and problems as with Todd & Marois (2005) Psych Science, just discussed.

(11) New and German (2015) Evolution and Human Behaviour

“After the fourth trial containing the additional experimental stimulus, the participant was asked, “Did you see anything in addition to the cross on that trial?” and which quadrant the additional stimulus appeared in. They were then asked to identify the stimulus in an array which in Experiment 1 included two variants chosen randomly from the spider stimuli and the two needle stimuli. Participants in Experiment 2 picked from all eight stimuli used in that experiment.”

Our second concern about response biases and the need for appropriate SDT analysis of the 4/8 alternative tasks applies to all these questions. We also note that analyses were only performed on groups separately (those who detected/failed to detect, those who located/failed to locate, and those who identified/failed to identify) and on the group which did all three/failed to do any one of the three. Especially in light of the fact that some subjects could clearly detect the stimulus without being able to identity it (e.g.), the most stringent test given our concerns (which were not obviously New and German’s comparative concerns), would be to consider the group which could not detect, identify or localize.

(12) Jackson-Nielsen (2017) Consciousness and cognition

This is a very interesting example of a follow-up which used a 3-AFC recognition test:

“participants were immediately asked, ‘‘which display looks most like what you just saw?’ from 3 alternatives”. However, though such an objective test is definitely to be preferred in our view to an open-ended series of probes, the 3-AFC test administered clearly had issues with response biases, as discussed, and actually yielded significantly below chance performance in one of the experiments.

(13) Mack et al. (2016) Consciousness and cognition

The follow ups here were essentially yes/no combined with an assessment of surprise. Participants were asked to enter letters into a box, and if they did so “were immediately asked by the experimenter whether they had noticed anything different about the array on this last trial and if they did not, they were told that there had been no letters and their responses to that news were recorded. Clearly, if they expressed surprise, this would be compelling evidence that they were unaware of the absence of the letters. Those observers who did not enter letters and realized there were no letters present were considered aware of the absence.” So, this again has all of the same problems we identify, considering subjects unaware because they expressed surprise.

(14) Devue et al. (2009) Perception

An 8-alternative task was used. The authors were primarily interested in a comparative analysis and so did not use this task to exclude subjects. We note that an 8 alternative task is very demanding – compare the 12-alternative task used in Todd, Fougnie & Marois (2005). There was an attempt to investigate biases in a separate bias trial, however SDT measures were not used.

(15) Memmert (2014) Cognitive Development

“After watching the video and stating the number of passes, participants answered four questions (following Simons & Chabris, 1999): (1) While you were counting, did you perceive anything unusual on the video? (2) Did you perceive anything other than the six players? (3) Did you see anyone else (besides the six players) appear on the video? (4) Did you notice a gorilla walk across the screen? After any “yes” reply, children were asked to provide details of what they noticed. If at any point a child mentioned the unexpected event, the remaining questions were omitted.” All of these follow-up questions are yes/no judgments, used to determine awareness in exactly the way we critique as problematic.

(16) Moore & Egeth (1997) JEP:HPP

This study (which includes one of us, Egeth, as author) did use forced choice questions. In one case, the question was 2-alternative, in the other it was 4-alternative. In the latter case, SDT would have been appropriate but was not used. In the former case, it may have been that a larger sample would have revealed evidence of sensitivity to the background pattern (as it stood 55% answered the 2-alternative question correctly). Although these results have been replicated, unfortunately the replication in Wood and Simons 2019 used a 6-alternative recognition task and this was not analyzed using SDT. We also note that the task is rather difficult in this study. Wood and Simons report: “Exclusion rates were much higher than anticipated, primarily due to exclusions when subjects failed to correctly report the pattern on the full-attention trial; we excluded 361 subjects, or 58% of our sample.”

(17) Cohen et al. (2020) Proc Natl Acad Sci

While this paper improves over a simple yes/no question in some ways, especially in that it used the follow up questions to exclude subjects from the unaware (IB) group, the follow up probes nonetheless remain yes/no questions, subject to response bias, e.g.:

(1) “Did you notice anything strange or different about that last trial?”

(2) “If I were to tell you that we did something odd on the last trial, would you have a guess as to what we did?”

(3) “If I were to tell you we did something different in the second half of the last trial, would you have a guess as to what we did?”

(4) “Did you notice anything different about the colors in the last scene?”

Follow up questions of this kind can be especially susceptible to bias, since subjects may be reluctant to “take back” their earlier answers and so be conservative in responding positively to avoid inconsistency or acknowledgement of earlier error. This may explain why such follow up questions can produce remarkable consistency despite their rather different wording. 

(18) Cohen et al. (2011) Psych Science

Here are the probes used in this study:

(1) Did you notice anything different on that trial?

(2) Did you notice something different about the background stream of images?

(3) Did you notice that a different type of image was presented in the background that was unique in some particular way?

(4) Did you see an actual photograph of a natural scene in that stream?

(5) If I were to tell you that there was a photograph in that stream, can you tell me what it was a photograph of?

Qs 1-4 are yes/no. Q5 is yes/no with an open-ended response. After this, a 5 or 6-alternative recognition test was administered. So again, this faces the same issues, since y/n questions are subject to bias in the way we have described, and many-alternative tests are more problematic than 2afc tests.

In summary

We really appreciate the care that went into compiling this list, and we agree that these papers and the improved methods they contain are relevant. But as hopefully made clear above, the approaches in each of these papers simply don’t solve the foundational issues our critique is aimed at (though they may address other issues). This is why we felt our new approach was necessary. And we continue to feel this way even after reading and incorporating these comments from Dr. Cohen.

Nevertheless, there is clearly lots for us to do in light of these comments. And so as noted earlier we have now added a very substantial new section to our discussion section to more fairly and completely portray the state of the art in this literature. This is really to our benefit in the end, since we now not only better acknowledge the diverse approaches present, but also set up ourselves to make our novel contribution exceedingly clear.

Main point 2: Let's imagine for a second that every study did just ask a yes/no question and then would stop. So, the criticism the authors are bringing up is valid (even though I believe it is not). I am not entirely sure that above chance performance on a forced choice task proves that the inattentionally blind can see after all. Could it just be a form of subliminal priming? Could there be a significant number of participants who basically would say something like, "No I did not see anything, and I feel like I am just guessing, but if you want me to say whether the thing was to the left or right, I will just 100% guess"? I know the literature on priming from things like change and inattentional blindness is a bit unclear, but this seems like maybe what is going on. In fact, maybe the authors are getting some of the best priming from inattentional blindness because of their large sample size, which previous studies do not use.

I'm curious how the authors would relate their studies to masked priming. In masked priming studies, observers say the did not see the target (like in this study) but still are above chance when forced to guess (like in this study). Do the researchers here think that that is evidence of "masked stimuli are truly seen" even if a participant openly says they are guessing?

We’re grateful to the reviewer for raising this question. As we say in response to Reviewer #1, our primary ambition in the paper is to establish, as our title suggests, residual sensitivity in IB. The ambition is quite neutral as to whether the sensitivity reflects conscious or unconscious processing (i.e. is akin to blindsight as traditionally conceived, or what the reviewer here suggests may be happening in masked priming). Since we were evidently insufficiently clear about this we have revised our manuscript in several places to clarify that we take our data primarily to support the more modest claim that there is residual sensitivity (conscious or unconscious) in the group of subjects who are traditionally classified as inattentionally blind. We believe that this claim has much more solid support in our data than our secondary and tentative suggestion about awareness.

This said, we do consider masked priming studies to be susceptible to the critique that performance may reflect degraded conscious awareness which is unreported because of conservative response criteria. There is good evidence that response criteria tend to be conservative near threshold (Björkman et al. 1993; see also: Railo et al. 2020), including specifically in masked priming studies (Sand 2016, cited in Phillips 2021). So, we consider it a perfectly reasonable hypothesis that subjects who say they feel they are guessing in fact have conscious access to a degraded signal which is insufficient to reach a conservative response criterion but nonetheless sufficient to perform above chance in 2afc detection. Of course, we appreciate that this hypothesis is controversial, so it is not one we argue for in our paper (though we are happy to share our feelings about it here).

Main point 3: My last question is about how the authors interpret a variety of inattentional blindness findings. Previous work has found that observers fail to notice a gorilla in a CT scan (Drew et al., 2013), a fight occurring right in front of them (Chabris et al., 2011), a plane on a runway that pilots crash into (Haines, 1991), and so forth. In a situation like this, do the authors believe that many participants are truly aware of these items but simply failed to answer a yes/no question correctly? For example, imagine the researchers made participants choose if the gorilla was in the left or right lung and some participants who initially said they did not notice the gorilla were still able to correctly say if it was in the left or right lung. Would the authors claim "that participant actually did see the gorilla in the lung"? I ask because it is difficult to understand what it means to be aware of something as salient as a gorilla in a CT scan, but say "no" you didn't notice it when asked a yes/no question. What does it mean to be aware of such important, ecologically relevant stimuli, but not act in response to them and openly say "no" you did not notice them?

Our view is that in such cases, observers may well have a “degraded” percept of the relevant feature (gorilla, plane, fight etc.). But crucially we do not suggest that this percept is sufficient for observers to recognize the object/event as a gorilla, plane, fight etc. Our claim is only that, in our studies at least, observers (as a group) do have enough information about the unexpected stimuli to locate them, and discriminate certain low level features better than chance. Crudely, it may be that subjects see the gorilla simply as a smudge or the plane as a shadowy patch etc. (One of us who is familiar with the gorilla CT scan stimuli notes that the gorilla is in fact rather hard to see even when you know which slide it is on, suggesting that they are not as “salient” as the reviewer suggests!) 

More precisely, in the paper we write that in our view perhaps “...unattended stimuli are encoded in a partial or degraded way. Here we see a variety of promising options for future work to investigate. One is that unattended stimuli are only encoded as part of ensemble representations or summary scene statistics (Rosenholtz, 2011; Cohen et al., 2016). Another is that only certain basic “low-level” or “preattentive” features (see Wolfe & Utochkin, 2019 for discussion) can enter awareness without attention. A final possibility consistent with the present data is that observers can in principle be aware of individual objects and higher-level features under inattention but that the precision of the corresponding representations is severely reduced. Our central aim here is to provide evidence that awareness in inattentional blindness is not abolished. Further work is needed to characterize the exact nature of that awareness.” We hope this sheds light on our perspective while still being appropriately cautious not to go too far beyond our data.

Overall: I believe there are many aspects of this set of studies that are innovative and I hope the methods will be used more broadly in the literature. However, I believe the authors misrepresent the field and overstate what can be interpreted from their results. While I am sure there are cases where more nuanced questions might reveal inattentional blindness is somewhat overestimated, claims like "the inattentionally blind can see after all" or "Inattentionally blind subjects consciously perceive thest stimuli after all" seem to be incorrect (or at least not at all proven by this data).

Once again, we would like to thank this reviewer for his feedback, which obviously comes from a place of tremendous expertise on these issues. We appreciate his assessment that our studies are innovative and that our methodological advances will be of use more broadly. We also hear the reviewer loud and clear about the passages in question, which on reflection we agree are not as central to our case as the other claims we make (regarding residual sensitivity and conservative responding), and so we have now edited them accordingly to refocus our discussion on only those claims that are central and supported. Thank you for making our paper stronger!

Reviewer #3 (Public review):

Summary:

Authors try to challenge the mainstream scientific as well as popularly held view that Inattentional

Blindness (IB) signifies subjects having no conscious awareness of what they report not seeing (after being exposed to unexpected stimuli). They show that even when subjects indicate NOT having seen the unexpected stimulus, they are at above chance level for reporting features such as location, color or movement of these stimuli. Also, they show that 'not seen' responses are in part due to a conservative bias of subjects, i.e. they tend to say no more than yes, regardless of actual visibility. Their conclusion is that IB may not (always) be blindness, but possibly amnesia, uncertainty etc.

We just thought to say that we felt this was a very accurate summary of our claims, and in ways underscore the modesty we had hoped to convey. This is especially true of the reviewer’s final sentence: “Their conclusion is that IB may not (always) be blindness, but possibly amnesia, uncertainty etc.”; as we noted in response to other reviewers, our claim is not that IB doesn’t exist, that subjects are always conscious of the stimulus, etc.; it is only that the cohort of IB subjects show sensitivity to the unattended stimulus in ways that suggest they are not as blind as traditionally conceived. Thank you for reading us as intended!

Strengths:

A huge pool of (25.000) subjects is used. They perform several versions of the IB experiments, both with briefly presented stimuli (as the classic Mack and Rock paradigm), as well as with prolonged stimuli moving over the screen for 5 seconds (a bit like the famous gorilla version), and all these versions show similar results, pointing in the same direction: above chance detection of unseen features, as well as conservative bias towards saying not seen.

We’re delighted that the reviewer appreciated these strengths in our manuscript!

Weaknesses:

Results are all significant but effects are not very strong, typically a bit above chance. Also, it is unclear what to compare these effects to, as there are no control experiments showing what performance would have been in a dual task version where subjects have to also report features etc for stimuli that they know will appear in some trials

The backdrop to the experiments reported here is the “consensus view” (Noah & Mangun, 2020) according to which inattention completely abolishes perception, such that subjects undergoing IB “have no awareness at all of the stimulus object” (Rock et al., 1992) and that “one can have one’s eyes focused on an object or event … without seeing it at all” (Carruthers, 2015). In this context, we think our findings of significant above-chance sensitivity (e.g., d′ = 0.51 for location in Experiment 1; chance, of course, would be d′ = 0 here) are striking and constitute strong evidence against the consensus view. We of course agree that the residual sensitivity is far lower than amongst subjects who noticed the stimulus. For this reason, we certainly believe that inattention has a dramatic impact on perception. To that extent, our data speak in favor of a “middle ground” view on which inattention substantially degrades but crucially does not abolish perception/explicit encoding. We see this as an importantly neglected option in a literature which has overly focused on seen/not seen binaries (see our section ‘Visual awareness as graded’).

Regarding the absence of a control condition, we think those conditions wouldn’t have played the same role in our experiments as they typically play in other experiments. As Reviewer #1 comments, the main role of such trials in previous work has been to exclude from analysis subjects who failed to report the unexpected stimulus on the divided and/or full attention control trials. As Reviewer #1 points out, excluding such subjects would very likely have ‘helped’ us. However, the practice is controversial. Indeed, in a review of 128 experiments, White et al. 2018 argue that the practice has “problematic consequences” and “may lead researchers to understate the pervasiveness of inattentional blindness". Since we wanted to offer as simple and demanding a test of residual sensitivity in IB as possible, we thus decided not to use any exclusions, and for that reason decided not to include divided/full attention trials.

As recommended, we discuss this decision not to include divided/full attention trials and our logic for not doing so in the manuscript. As we explain, not having those conditions makes it more impressive, not less impressive, that we observed the results we in fact did — it makes our results more interpretable, not less interpretable, and so absence of such conditions from our manuscript should not (in our view) be considered any kind of weakness.

There are quite some studies showing that during IB, neural processing of visual stimuli continues up to high visual levels, for example, Vandenbroucke et al 2014 doi:10.1162/jocn_a_00530 showed preserved processing of perceptual inference (i.e. seeing a kanizsa illusion) during IB. Scholte et al 2006 doi: 10.1016/j.brainres.2005.10.051 showed preserved scene segmentation signals during IB. Compared to the strength of these neural signatures, the reported effects may be considered not all that surprising, or even weak.

We agree that such evidence of neural processing in IB is relevant to — and perhaps indeed consistent with — our picture, and we’re grateful to the reviewer for pointing out further studies along those lines. Previously, we mentioned a study from Pitts et al., 2012 in which, as we wrote, “unexpected line patterns have been found to elicit the same Nd1 ERP component in both noticers and inattentionally blind subjects (Pitts et al., 2012).” We have added references to both the studies which the reviewer mentions – as well as an additional relevant study – to our manuscript in this context. Thank you for the helpful addition.

We do however think that our studies are importantly different to this previous work. Our question is whether processing under IB yields representations which are available for explicit report and so would constitute clear evidence of seeing, and perhaps even conscious experience. As we discuss, evidence for this kind of processing remains wanting: “A handful of prior studies have explored the possibility that inattentionally blind subjects may retain some visual sensitivity to features of IB stimuli (e.g., Schnuerch et al., 2016; see also Kreitz et al., 2020, Nobre et al., 2020). However, a recent meta-analysis of this literature (Nobre et al., 2022) argues that such work is problematic along a number of dimensions, including underpowered samples and evidence of publication bias that, when corrected for, eliminates effects revealed by earlier approaches, concluding “that more evidence, particularly from well-powered pre-registered experiments, is needed before solid conclusions can be drawn regarding implicit processing during inattentional blindness” (Nobre et al., 2022).” Our paper is aimed at addressing this question which evidence of neural processing can only speak to indirectly.

Recommendations for the authors:  

Reviewer #1 (Recommendations for the authors):

(1) Please report all of the data, especially the number of subjects in each experiment that answered Y/N and the numbers of subjects in each of the Y and N groups that guessed a feature correctly/incorrectly on the 2AFC tasks. And also the confidence ratings for the 2AFC task (for comparison with the confidence ratings on the Y/N questions).

We now report all this data in our (revised) Supplementary Materials. We agree that this information will be helpful to readers.

(2) Consider adding a control condition with partial attention (dual task) or full attention (single task) to estimate the rates of seeing the critical stimulus when it's expected.

This is the only recommendation we have chosen not to implement. The reason, as we explain in detail above (especially in response to Reviewer #1 comment 5), is that this would not in fact be a “control condition” in our studies, and indeed would only inflate the biases we are concerned with in our work. As the referee comments, the main role of such trials in previous work has been to exclude from analysis subjects who failed to report the unexpected stimulus on the divided and/or full attention control trials. And the practice is controversial: Indeed, in a review of 128 experiments, White et al. 2018 argue that the practice has “problematic consequences” and “may lead researchers to understate the pervasiveness of inattentional blindness" (emphasis added). So, our choice not to have such conditions ensures an especially stringent test of our central claim. Not having those conditions (and their accompanying exclusions) makes our results more interpretable, not less interpretable, and so the absence of such conditions from our manuscript should not (in our view) be considered any kind of weakness.

We have added a paragraph to our “Design and analytical approach” section explaining the logic behind our deliberate decision not to include divided or full attention trials in our experiments. (For even fuller discussion, see our response to Reviewer #1’s comment 5 above.)

(3) Consider revising the interpretations to be more precise about the distinction between the super subject being above chance versus each individual subject who cannot be at chance or above chance because there was only a single trial per subject.

We have now done this throughout the manuscript, as discussed above. We have also added a substantive additional discussion to our “Design and analytical approach” section discussing what should be said about individual subjects in light of our group level data.

This was a very helpful point, and greatly clarifies the claims we wish to make in the paper. Thank you for this comment, which has certainly made our paper stronger.

Reviewer #2 (Recommendations for the authors):

I would be curious to hear the authors' response to two points:

(1) What do they have to say about prior studies that do more than just ask yes/no questions (and ask several follow-ups)? Are those studies "valid"?

A very substantial new discussion of this important point has been added. As you will see above, we comment on every one of the 18 papers this reviewer raised (as well as the general argument made); we contend that while many of these papers improve on past methodology in various ways, most in fact do “just ask yes/no questions”, and none of them makes the methodological advance we offer in our manuscript. However, this discussion has helped us clarify that very advance, and so working through this issue has really helped us improve our paper and make its relation to existing literature that much clearer. Thank you for raising this crucial point.

(2) Do the authors think it is possible that in many cases, people are just guessing about a critical item's location or color and this is at least in part a form of priming?

We have clarified our discussion in numerous places to further emphasize that our main point concerns above-chance sensitivity, not awareness. Given this, we take very seriously the hypothesis that something like priming of a kind sometimes proposed to occur in cases of blindsight or other putative cases of unconscious perception could be what is driving the responses in non-noticers.

Reviewer #3 (Recommendations for the authors):

(1) Control dual task version with expected stimuli would be nice

We have added a paragraph to our “Design and analytical approach” section explaining the logic behind our deliberate decision not to include divided or full attention trials, which would not in fact be a “control” task in our experiments. For full discussion, see our response to Reviewer 3 above, as well as our summary here in the Recommendations for Authors section in responding to Reviewer 1, recommendation (2).

(2) Please do a better job in discussing and introducing experiments about neural signatures during IB.

A discussion of Vandenbroucke et al. 2014 and Scholte et al. 2006 has been added to our discussion of neural signatures in IB, as well as an additional reference to an important early study of semantic processing in IB (Rees et al., 1999). Thank you for these very helpful suggestions!

eLife Assessment

This work provides an important framework for understanding the primary causes of disease. While the theoretical results rely on strong assumptions about the underlying causal mechanisms, the authors provide solid empirical evidence that the framework is robust to modest violations of these assumptions.

Reviewer #1 (Public review):

Summary:

This manuscript seeks to estimate the causal effect of genes on disease. To do so, they introduce a novel algorithm, termed the Root Causal Strength using Perturbations (RCSP) algorithm. RCSP uses perturb-seq to first estimate the gene regulatory network structure among genes, and then uses bulk RNA-seq with phenotype data on the samples to estimate causal effects of genes on the phenotype conditional on the learned network structure. The authors assess the performance of RCSP in comparison to other methods via simulation. Next, they apply RCSP to two real human datasets: 513 individuals age-related macular degeneration and 137 individuals with multiple sclerosis.

Strengths:

The authors tackle an important and ambitious problem - the identification of causal contributors to disease in the context of a causal inference framework. As the authors point out, observational RNA-seq data is insufficient for this kind of causal discovery, since it is very challenging to recover the true underlying graph from observational data; interventional data are needed. However, little perturb-seq data has been generated with annotated phenotype data, and much bulk RNA-seq data has already been generated, so it is useful to propose an algorithm to integrate the two as the authors have done.

The authors also offer substantial theoretical exposition for their work, bringing to bear both the literature on causal discovery as well as literature on the genetic architecture of complex traits. They also benchmark RCSP under multiple challenging simulation settings, including an analysis of RCSP when the underlying graph is not a DAG.

Weaknesses:

The notion of a "root" causal gene - which the authors define based on a graph theoretic notion of topologically sorting graphs - requires a graph that is directed and acyclic. It is the latter that constitutes an important weakness here - it simply is a large simplification of human biology to draw out a DAG including hundreds of genes and a phenotype Y and to claim that the true graph contains no cycles. For example - consider the authors' analysis of T cell infiltration in multiple sclerosis (MS). CD4+ effector T cells have the interesting property that they are stimulated by IL2 as a growth factor; yet IL2 also stimulates the activation of (suppressive) regulatory T cells. What does it mean to analyze CD4+ regulation in disease with a graph that does not consider IL2 (or other cytokine) mediated feedback loops/cycles? To the authors' credit, in the supplementary materials they do consider a simulated example with a cyclic underling causal graph, finding that RCSP performed well comparison to an implementation of the additive noise model (ANM), LiNGAM, CausalCell, and two simpler approaches based on linear regression.

I also encourage the authors to consider more carefully when graph structure learned from perturb-seq can be ported over to bulk RNA-seq. Consider again the MS CD4+ example - the authors first start with a large perturb-seq experiment (Replogle et al., 2022) performed in K562 cells. To what extent are K562 cells, which are derived from a leukemia cell line, suitable for learning the regulatory structure of CD4+ cells from individuals with an MS diagnosis? Presumably this structure is not exactly correct - to what extent is the RCSP algorithm sensitive to false edges in this graph? The authors perform an analysis of this scenario in Supplementary Figure 4, which shows that RCSP is robust to some degree of departure from the underlying true structure. And although challenging - it would be ideal for the RCSP to model or reflect the challenges in correctly identifying the regulatory structure.

It should also be noted that in most perturb-seq experiments, the entire genome is not perturbed, and frequently important TFs (that presumably are very far "upstream" and thus candidate "root" causal genes) are not expressed highly enough to be detected with scRNA-seq. In that context - perhaps slightly modifying the language regarding RCSP's capabilities might be helpful for the manuscript - perhaps it would be better to describe it has an algorithm for causal discovery among a set of genes that were perturbed and measured, rather than a truly complete search for causal factors. Perhaps more broadly - it would also benefit the manuscript to devote slightly more text to describing the kinds of scenarios where RCSP (and similar ideas) would be most appropriately applied - perhaps a well-powered, phenotype annotated perturb-seq dataset performed in a disease relevant primary cell.

Reviewer #2 (Public review):

Summary:

This paper presents a very interesting use of a causal graph framework to identify the "root genes" of a disease phenotype. Root genes are the genes that cause a cascade of events that ultimately leads to the disease phenotype, assuming the disease progression is linear.

Strengths:

- The methodology has a solid theoretical background.<br /> - This is a novel use of the causal graph framework to infer root causes in a graph

Comments on revisions:

The authors addressed all of my comments.

Author response:

The following is the authors’ response to the original reviews.

Reviewer 1:

(1) The notion of a “root” causal gene - which the authors define based on a graph theoretic notion of topologically sorting graphs - requires a graph that is directed and acyclic. It is the latter that constitutes an important weakness here - it simply is a large simplification of human biology to draw out a DAG including hundreds of genes and a phenotype Y and to claim that the true graph contains no cycles.

We agree that real causal graphs in biology often contain cycles. We now include additional experimental results with cyclic directed graphs in the Supplementary Materials. RCSP outperformed the other algorithms even in this setting, but we caution the reader that the theoretical interpretation of the RCS score may not coincide with a root causal effect when cycles exist:

“We also evaluated the algorithms on directed graphs with cycles. We generated a linear SEM over ρ + 1 = 1000 variables in . We sampled the coefficient matrix β from a Bernoulli (1/(p − 1)) distribution but did not restrict the non-zero coefficients to the upper triangular portion of the matrix. We then proceeded to permute the variable ordering and weight each entry as in the Methods for the DAG. We repeated this procedure 30 times and report the results in Supplementary Figure 3.

RCSP again outperformed all other algorithms even in the cyclic case. The results suggest that conditioning on the surrogate ancestors also estimates the RCS well even in the cyclic case. However, we caution that an error term E_i can affect the ancestors of when cycles exist. As a result, the RCS may not isolate the causal effect of the error term and thus not truly coincide with the notion of a root causal effect in cyclic causal graphs.”

(2) I also encourage the authors to consider more carefully when graph structure learned from Perturb-seq can be ported over to bulk RNA-seq. Presumably this structure is not exactly correct - to what extent is the RCSP algorithm sensitive to false edges in this graph? This leap - from cell line to primary human cells - is also not modeled in the simulation. Although challenging - it would be ideal for the RCSP to model or reflect the challenges in correctly identifying the regulatory structure.

We now include additional experimental results, where we gradually increased the incongruence between the DAG modeling the Perturb-seq and the DAG modeling the bulk RNA-seq using a mixture of graphs. The performance of RCSP degraded gradually, rather than abruptly, with increasing incongruence. We therefore conclude that RCSP is robust to differences between the causal graphs representing Perturb-seq and bulk RNA-seq:

“We next assessed the performance of RCSP when the DAG underlying the Perturb-seq data differs from the DAG underlying the bulk RNA-seq data. We considered a mixture of two random DAGs in bulk RNA-seq, where one of the DAGs coincided with the Perturb-seq DAG and second alternate DAG did not. We instantiated and simulated samples from each DAG as per the previous subsection. We generated 0%, 25%, 50%, 75%, and 100% of the bulk RNA-seq samples from the alternate DAG, and the rest from the Perturb-seq DAG. We ideally would like to see the performance of RCSP degrade gracefully, as opposed to abruptly, as the percent of samples derived from the alternate DAG increases.

We summarize results in Supplementary Figure 4. As expected, RCSP performed the best when we drew all samples from the same underlying DAG for Perturb-seq and bulk RNA-seq. However, the performance of RCSP also degraded slowly as the percent of samples increased from the alternate DAG. We conclude that RCSP can accommodate some differences between the underlying DAGs in Perturb-seq and bulk RNA-seq with only a mild degradation in performance.”

(3) It should also be noted that in most Perturb-seq experiments, the entire genome is not perturbed, and frequently important TFs (that presumably are very far “upstream” and thus candidate “root” causal genes) are not expressed highly enough to be detected with scRNA-seq. In that context - perhaps slightly modifying the language regarding RCSP’s capabilities might be helpful for the manuscript - perhaps it would be better to describe it as an algorithm for causal discovery among a set of genes that were perturbed and measured, rather than a truly complete search for causal factors. Perhaps more broadly it would also benefit the manuscript to devote slightly more text to describing the kinds of scenarios where RCSP (and similar ideas) would be most appropriately applied - perhaps a well-powered, phenotype annotated Perturb-seq dataset performed in a disease relevant primary cell.

We now clarify that Perturb-seq can only identify root causal genes among the perturbed set of genes in the Discussion:

“Modern genome-wide Perturb-seq datasets also adequately perturb and measure only a few thousand, rather than all, gene expression levels. RCSP can only identify root causal genes within this perturbed and measured subset.”

We now also describe the scenario where RCSP can identify root causal genes well in the Introduction:

“Experiments demonstrate marked improvements in performance, when investigators have access to a large bulk RNA-seq dataset and a genome-wide Perturb-seq dataset from a cell line of a disease-relevant tissue.”

Reviewer 2:

(1) The process from health-to-disease is not linear most of the time with many checks along the way that aim to prevent the disease phenotype. This leads to a non-deterministic nature of the path from health-to-disease. In other words, with the same root gene perturbations, and depending on other factors outside of gene expression, someone may develop a phenotype in a year, another in 10 years and someone else never. Claiming that this information is included in the error terms might not be sufficient to address this issue. The authors should discuss this limitation.

The proposed approach accommodates the above non-deterministic nature. The error terms of model factors that are outside of gene expression. We model the relation from gene expression to Y as probabilistic rather than deterministic because , where E_Y introduces stochasticity. Thus, two individuals with the same instantiations of the root causes may develop disease differently. We now clarify this in Methods:

“The error terms model root causes that are outside of gene expression, such as genetic variation or environmental factors. Moreover, the relation from gene expression to Y is stochastic because , where E_Y introduces the stochasticity. Two individuals may therefore have the exact same error term values over but different instantiations of Y.”

(2) The paper assumes that the network connectivity will remain the same after perturbation. This is not always true due to backup mechanisms in the cells. For example, suppose that a cell wants to create product P and it can do it through two alternative paths: Path #1: A → B → P, Path #2: A → C → P. Now suppose that path #1 is more efficient, so when B can be produced, path #2 is inactive. Once the perturbation blocks element B from being produced, the graph connectivity changes by activation of path #2. I did not see the authors taking this into consideration, which seems to be a major limitation in using Perturb-seq results to infer conductivities.

We agree that backup mechanisms can exist and therefore now include additional experimental results, where we gradually increased the incongruence between the DAG modeling the Perturb-seq and the DAG modeling the bulk RNA-seq using a mixture of graphs. The performance of RCSP degraded gradually, rather than abruptly, with increasing incongruence. We therefore conclude that RCSP is robust to differences between the causal graphs representing Perturb-seq and bulk RNA-seq:

“We next assessed the performance of RCSP when the DAG underlying the Perturb-seq data differs from the DAG underlying the bulk RNA-seq data. We considered a mixture of two random DAGs in bulk RNA-seq, where one of the DAGs coincided with the Perturb-seq DAG and second alternate DAG did not. We generated 0%, 25%, 50%, 75%, and 100% of the bulk RNA-seq samples from the alternate DAG, and the rest from the Perturb-seq DAG. We ideally would like to see the performance of RCSP degrade gracefully, as opposed to abruptly, as the percent of samples derived from the alternate DAG increases.

We summarize results in Supplementary Figure 4. As expected, RCSP performed the best when we drew all samples from the same underlying DAG for Perturb-seq and bulk RNA-seq. However, the performance of RCSP also degraded slowly as the percent of samples increased from the alternate DAG. We conclude that RCSP can accommodate some differences between the underlying DAGs in Perturb-seq and bulk RNA-seq with only a mild degradation in performance.”

(3) There is substantial system heterogeneity that may cause the same phenotype. This goes beyond the authors claim that although the initial gene causes of a disease may differ from person to person, at some point they will all converge to changes in the same set of “root genes.” This is not true for many diseases, which are defined based on symptoms and lab tests at the patient level. You may have two completely different molecular pathologies that lead to the development of the same symptoms and test results. Breast cancer with its subtypes is a prime example of that. In theory, this issue could be addressed if there is infinite sample size. However, this assumption is largely violated in all existing biological datasets.

The proposed method accommodates the above heterogeneity. We do not assume that the root causes affect the same set of root causal genes. Instead the root causes and root causal genes may vary from person to person. We write in the Introduction:

“The problem is further complicated by the existence of complex disease, where a patient may have multiple root causal genes that differ from other patients even within the same diagnostic category... We thus also seek to identify patient-specific root causal genes in order to classify patients into meaningful biological subgroups each hopefully dictated by only a small group of genes.”

The root causal genes may further affect different downstream genes at the patient-specific level. However root causal genes tend to have many downstream effects so that virtually every gene expression level becomes correlated with Y. We now clarify this by describing the omnigenic root causal model in the Introduction as follows:

“Finally, application of the algorithm to two complex diseases with disparate pathogeneses recovers an omnigenic root causal model, where a small set of root causal genes drive pathogenesis but impact many downstream genes within each patient. As a result, nearly all gene expression levels are correlated with the diagnosis at the population level.”

(4) Were the values of the synthetic variables Z-scored?

Yes, all variables were z-scored. We now clarify this in Methods:

“We also standardized all variables before running the regressions to prevent gaming of the marginal variances in causal discovery (Reisach et al., 2021; Ng et al., 2024).”

(5) The algorithm seems to require both RNA-seq and Perturb-seq data (Algorithm 1, page 14). Can it function with RNA-seq data only? What will be different in this case?

The algorithm cannot function with observational bulk RNA-seq data only. We included Perturb-seq because causal discovery with observational RNA-seq data alone tends to be inaccurate and unstable, as highlighted by the results of CausalCell. We further emphasize that we do not rely on d-separation faithfulness in Methods, which is typically required for causal discovery from observational data alone:

“We can also claim the backward direction under d-separation faithfulness. We however avoid making this additional assumption because real biological data may not arise from distributions obeying d-separation faithfulness in practice.”

(6) Synthetic data generation: how many different graphs (SEMs) did they start from? (30?) How many samples per graph? Did they test different sample sizes?

We now clarify that we generate 30 random SEMs, each associated with a DAG. We used 200 samples for the bulk RNA-seq to mimic a relatively large but common sample size. We also drew 200 samples for each perturbation or control in the Perturb-seq data. We did not consider multiple sample sizes due to the time required to complete each run. Instead, we focused on a typical scenario where investigators would apply RCSP. We now write the following in the Methods:

“We drew 200 samples for the bulk RNA-seq data to mimic a large but common dataset size. We introduced knockdown perturbations in Perturb-seq by subtracting an offset of two in the softplus function: . We finally drew 200 samples for the control and each perturbation condition to generate the Perturb-seq data. We repeated the above procedure 30 times.” We also include the following in Results:

“We obtained 200 cell samples from each perturbation, and another 200 controls without perturbations. We therefore generated a total of 2501 × 200 = 500,200 single cell samples for each Perturb-seq dataset. We simulated 200 bulk RNA-seq samples.”

(7) The presentation of comparative results (Supplementary Figures 4 and 7) is not clear. No details are given on how these results were generated. (what does it mean “The first column denotes the standard deviation of the outputs for each algorithm?”) Why all other methods have higher SD differences than RCSP? Is it a matter of scaling? Shouldn’t they have at least some values near zero since the authors “added the minimum value so that all histograms begin at zero?”

Each of these supplementary figures contains a 6 by 3 table of figures. By the first column, we mean column one (with rows 1 through 6) of each figure. The D-RCS and D-SD scores represent standard deviations of the RCS and SD scores from zero of each gene, respectively. We can similarly compute the standard deviation of the outputs of the algorithms. We now clarify this in the Supplementary Materials:

“The figure contains 6 rows and 3 columns. Similar to the D-RCS, we can compute the standard deviation of the output of each algorithm from zero for each gene. The first column in Supplementary Figure 7 denotes the histograms of these standard deviations across the genes.”

Many histograms do not appear to start at zero because the bars are too small to be visible. We now clarify this in the Supplementary Materials as well:

“Note that the bars at zero are not visible for many algorithms, since only a few genes attained standard deviations near the minimum.”

(8) Why RCSP results are more like a negative binomial distribution and every other is kind of normal?

All other methods have higher standard deviations than RCSP because they fail to compute an accurate measure of the root causal effect. Recall that, just like a machine has a few root causal problems, only a few root casual genes have large root causal effects under the omnigenic root causal model. The results of RCSP look more like a negative binomial distribution because most RCS scores are concentrated around zero and only a few RCS scores are large – consistent with the omnigenic root causal model. The other algorithms fail to properly control for the upstream genes and thus attain large standard deviations for nearly all genes. We now clarify these points in the Supplementary Materials as follows:

“If an algorithm accurately identifies root causal genes, then it should only identify a few genes with large conditional root causal effects under the omnigenic root causal model. The RCSP algorithm had a histogram with large probability mass centered around zero with a long tail to the right. The standard deviations of the outputs of the other algorithms attained large values for nearly all genes. Incorporating feature selection and causal discovery with CausalCell introduced more outliers in the histogram of ANM. We conclude that only RCSP detected an omnigenic root causal model.”

(9) What is the significance of genes changing expression “from left to right” in a UMAP plot? (e.g., Fig. 3h and 3g)

The first UMAP dimension captured the variability of the RCS scores for most root causal genes. As a result, we could focus our analysis on the black cluster in Figure 3 (g) with large RCS scores in the subsequent pathway enrichment analysis summarized in Figure 3 (j). If two dimensions were involved, then we would need to analyze at least two clusters (e.g., black and pink), but this was not the case. We now clarify this in Results:

“The RCS scores of most of the top genes exhibited a clear gradation increasing only from the left to the right hand side of the UMAP embedding; we plot an example in Figure 3 (h). We found three exceptions to this rule among the top 30 genes (example in Figure 3 (i) and see Supplementary Materials). RCSP thus detected genes with large RCS scores primarily in the black cluster of Figure 3 (g). Pathway enrichment analysis within this cluster alone yielded supra-significant results on the same pathway detected in the global analysis...”

(10) The authors somewhat overstate the novelty of their algorithm. Representation of GRNs as causal graphs dates back in 2000 with the work of Nir Friedman in yeast. Other methods were developed more recently that look on regulatory network changes at the single sample level which the authors do not seem to be aware (e.g., Ellington et al, NeurIPS 2023 workshop GenBio and Bushur et al, 2019, Bioinformatics are two such examples). The methods they mention are for single cell data and they are not designed to connect single sample-level changes to a person’s phenotype. The RCS method needs to be put in the right background context in order to bring up what is really novel about it.

We agree that many methods already exist for uncovering associational, predictive (Markov, neighborhood) and causal gene regulatory networks. We now cite the above papers. However, the novelty in our manuscript is not causal graph discovery, but rather estimation of root causal effects, detection of root causal genes, and the proposal of the omnigenic root causal model. We now clarify this in the

Introduction:

“Many algorithms focus on discovering associational or predictive relations, sometimes visually represented as gene regulatory networks (Costa et al., 2017; Ellington et al., 2023). Other methods even identify causal relations (Friedman et al., 2000; Wang et al., 2023; Wen et al., 2000; Buschur et al., 2000), but none pinpoint the first gene expression levels that ultimately generate the vast majority of pathogenesis. Simply learning a causal graph does not resolve the issue because causal graphs do not summarize the effects of unobserved root causes, such as unmeasured environmental changes or variants, that are needed to identify all root causal genes. We therefore define the Root Causal Strength (RCS) score...”

Reviewer 3:

(1) Several assumptions of the method are problematic. The most concerning is that the observational expression changes are all causally upstream of disease. There is work using Mendelian randomization (MR) showing that the opposite is more likely to be true: most differential expression in disease cohorts is a consequence rather than a cause of disease (Porcu et al., 2021). Indeed, the oxidative stress of AMD has known cellular responses including the upregulation of p53. The authors need to think carefully about how this impacts their framework. Can the theory say anything in this light? Simulations could also be designed to address robustness.

Strictly speaking, we believe that differential expression in disease most likely has a cyclic causal structure: gene expression causes a diagnosis or symptom severity, and a diagnosis or symptom severity lead to treatments and other behavioral changes that perturb gene expression. For example, revTMWR in Porcu et al. (2021) uses trans-variants that are less likely to directly cause gene expression and instead directly cause a phenotype. However, TWMR as proposed in Porcu et al. (2019) instead uses cis-eQTLs and finds many putative causal relations from gene expression to phenotype. Thus, both causal directions likely hold.

RCSP uses disease-relevant tissue believed to harbor gene expression levels that cause disease. However, RCSP theoretically cannot handle the scenario where Y is a non-sink vertex and is a parent of a gene expression level because modern Perturb-seq datasets usually do not perturb or measure Y. We therefore empirically investigated the degree of error by running experiments, where we set Y to a non-sink vertex, so that it can cause gene expression. We find that the performance of RCSP degrades considerably for gene expression levels that contain Y as a parent. Thus RCSP is sensitive to violations of the sink target assumption:

“We finally considered the scenario where Y is a non-sink (or non-terminal) vertex. If Y is a parent of a gene expression level, then we cannot properly condition on the parents because modern Perturbseq datasets usually do not intervene on Y or measure Y . We therefore empirically investigated the degradation in performance resulting from a non-sink target Y, in particular for gene expression levels where Y is a parent. We again simulated 200 samples from bulk RNA-seq and each condition of Perturbseq with a DAG over 1000 vertices, an expected neighborhood size of 2 and a non-sink target Y . We then removed the outgoing edges from Y and resampled the DAG with a sink target. We compare the results of RCSP for both DAGs in gene expression levels where Y is a parent. We plot the results in Supplementary Figure 5. As expected, we observe a degradation in performance when Y is not terminal, where the mean RMSE increased from 0.045 to 0.342. We conclude that RCSP is sensitive to violations of the sink target assumption.”

(2) A closely related issue is the DAG assumption of no cycles. This assumption is brought to bear because it is required for much classical causal machinery, but is unrealistic in biology where feedback is pervasive. How robust is RCSP to (mild) violations of this assumption? Simulations would be a straightforward way to address this.

We agree that real causal graphs in biology often contain cycles. We now include additional experimental results with cyclic directed graphs in the Supplementary Materials. RCSP outperformed the other algorithms even in this setting, but we caution the reader that the theoretical interpretation of the RCS score may not coincide with a root causal effect when cycles exist:

“We also evaluated the algorithms on directed graphs with cycles. We generated a linear SEM over p + 1 = 1000 variables in . We sampled the coefficient matrix β from a Bernoulli (1/(p − 1)) distribution but did not restrict the non-zero coefficients to the upper triangular portion of the matrix. We then proceeded to permute the variable ordering and weight each entry as in the Methods for the DAG. We repeated this procedure 30 times and report the results in Supplementary Figure 3.

RCSP again outperformed all other algorithms even in the cyclic case. The results suggest that conditioning on the surrogate ancestors also estimates the RCS well even in the cyclic case. However, we caution that an error term E_i can affect the ancestors of , when cycles exist. As a result, the RCS may not isolate the causal effect of the error term and thus not truly coincide with the notion of a root causal effect in cyclic causal graphs.”

(3) The authors spend considerable effort arguing that technical sampling noise in X can effectively be ignored (at least in bulk). While the mathematical arguments here are reasonable, they miss the bigger picture point that the measured gene expression X can only ever be a noisy/biased proxy for the expression changes that caused disease: 1) Those events happened before the disease manifested, possibly early in development for some conditions like neurodevelopmental disorders. 2) bulk RNA-seq gives only an average across cell-types, whereas specific cell-types are likely “causal.” 3) only a small sample, at a single time point, is typically available. Expression in other parts of the tissue and at different times will be variable.

We agree that many other sources of error exist. The causal model of RNA-expression in Methods corresponds to a single snapshot in time for each sample. We now clarify this in the Methods as follows:

“We represent a snapshot of a biological causal process using an SEM over obeying Equation (3).”

We thus only detect the root causal genes in a single snapshot in time for each sample in bulk RNA-seq. If we cannot detect the root causal effect in a gene due to the signal washing out over time as in (1), or if the root causal effect in different cell types cancel each other out to exactly zero in bulk as in (2), then we cannot detect those root causal genes even with an infinite sample size.

(4) While there are connections to the omnigenic model, the latter is somewhat misrepresented. The authors refer to the “core genes” of the omnigenic model as being at the end (longitudinal) of pathogenesis. The omnigenic model makes no statements about temporal ordering: in causal inference terminology the core genes are simply the direct causes of disease.

We now clarify that we use the word pathogenesis to mean the causal cascade from root causes to the diagnosis. In this case, the direct causes of the diagnosis correspond to the end of pathogenesis, while the root causes correspond to the beginning. For example, if , with Y a diagnosis, then X₁ is a root causal gene while X₂ is a core (direct causal) gene. We now clarify this in the Introduction:

“Root causes of disease correspond to the most upstream causes of a diagnosis with strong causal effects on the diagnosis. Pathogenesis refers to the causal cascade from root causes to the diagnosis. Genetic and non-genetic factors may act as root causes and affect gene expression as an intermediate step during pathogenesis. We introduce root causal gene expression levels – or root causal genes for short – that correspond to the initial changes to gene expression induced by genetic and non-genetic root causes that have large causal effects on a downstream diagnosis (Figure 1 (a)). Root causal genes differ from core genes that directly cause the diagnosis and thus lie at the end, rather than at the beginning, of pathogenesis (Boyle et al., 2017).”

(5) A key observation underlying the omnigenic model is that genetic heritability is spread throughout the genome (and somewhat concentrated near genes expressed in disease relevant cell types). This implies that (almost) all expressed genes, or their associated (e)SNPs, are “root causes”.

We now clarify that genetic heritability can be spread throughout the genome in the omnigenic root causal model as well in the Discussion:

“Further, each causal genetic variant tends to have only a small effect on disease risk in complex disease because the variant can directly cause Y or directly cause any causal gene including those with small root causal effects on Y ; thus, all error terms that cause Y can model genetic effects on Y. However, the root causal model further elaborates that genetic and non-genetic factors often combine to produce a few root causal genes with large root causal effects, where non-genetic factors typically account for the majority of the large effects in complex disease. Many variants may therefore cause many genes in diseases with only a few root causal genes.”

We finally add Figure 5 into the Discussion as a concrete example illustrating the omnigenic root causal model:

(6) The claim that root causal genes would be good therapeutic targets feels unfounded. If these are highly variable across individuals then the choice of treatment becomes challenging. By contrast the causal effects may converge on core genes before impacting disease, so that intervening on the core genes might be preferable. The jury is still out on these questions, so the claim should at least be made hypothetical.

We clarify that we do not claim that root causal genes are better treatment targets than core genes in terms of magnitudes of causal effects on the phenotype. For example, in the common cold with a virus as the root cause, giving a patient an antiviral will eliminate fever and congestion, but so will giving a decongestant and an antipyretic. We only claim that treating root causal genes can eliminate disease near its pathogenic onset, just like giving an antiviral can eliminate the viral load and stop pathogenesis. We write the following the Introduction:

“Treating root causal genes can modify disease pathogenesis in its entirety, whereas targeting other causes may only provide symptomatic relief... Identifying root causal genes is therefore critical for developing treatments that eliminate disease near its pathogenic onset.”

We also further clarify in the Discussion that root causal genes account for deleterious causal effects not captured by the diagnosis Y:

“We finally emphasize that the root causal model accounts for all deleterious effects of the root causal genes, whereas the core gene model only captures the deleterious effects captured by the diagnosis Y. For example, the disease of diabetes causes retinopathy, but retinopathy is not a part of the diagnostic criteria of diabetes. As a result, the gene expression levels that cause retinopathy but not the diagnosis of diabetes are not core genes, even though they are affected by the root causal genes.”

We do agree that root causal genes may differ substantially between patients, although it is unclear if the heterogeneity is too great to develop treatments.

(7) The closest thing to a gold standard I believe we have for “root causal genes” is integration of molecular QTLs and GWAS, specifically coloc/MR. Here the “E” of RCSP are explicitly represented as SNPs. I don’t know if there is good data for AMD but there certainly is for MS. The authors should assess the overlap with their results. Another orthogonal avenue would be to check whether the root causal genes change early in disease progression.

Colocalization and Mendelian randomization unfortunately cannot identify root causal effects because they all attempt, either heuristically (colocalization) or rigorously (MR), to identify variants that cause each gene expression level rather than variants that directly cause each gene expression level and thus make up the error terms. We therefore need new methods that can identify direct causal variants in order to assess overlap.

We checked whether root causal genes change early in disease progression using knowledge of pathogenesis. In particular, oxidative stress induces pathogenesis in AMD, and RCSP identified root causal genes involved in oxidative stress in AMD:

“The pathogenesis of AMD involves the loss of RPE cells. The RPE absorbs light in the back of the retina, but the combination of light and oxygen induces oxidative stress, and then a cascade of events such as immune cell activation, cellular senescence, drusen accumulation, neovascularization and ultimately fibrosis (Barouch et al., 2007). We therefore expect the root causal genes of AMD to include genes involved in oxidative stress during early pathogenesis. The gene MIPEP with the highest D-RCS score in Figure 3 (d) indeed promotes the maturation of oxidative phosphorylation-related proteins (Shi et al., 2011). The second gene SLC7A5 is a solute carrier that activates mTORC1 whose hyperactivation increases oxidative stress via lipid peroxidation (Nachef et al., 2021; Go et al., 2020). The gene HEATR1 is involved in ribosome biogenesis that is downregulated by oxidative stress (Turi et al., 2018). The top genes discovered by RCSP thus identify pathways known to be involved in oxidative stress.”

Similarly, T cell infiltration across the blood brain barrier initiates pathogenesis in MS, and RCSP identified root causal genes involved in this infiltration:

“Genes with the highest D-RCS scores included MNT, CERCAM and HERPUD2 (Figure 4 (d)). MNT is a MYC antagonist that modulates the proliferative and pro-survival signals of T cells after engagement of the T cell receptor (Gnanaprakasam et al., 2017). Similarly, CERCAM is an adhesion molecule expressed at high levels in microvessels of the brain that increases leukocyte transmigration across the blood brain barrier (Starzyk et al., 2000). HERPUD2 is involved in the endoplasmic-reticulum associated degradation of unfolded proteins (Kokame et al., 2000). Genes with the highest D-RCS scores thus serve key roles in known pathogenic pathways of MS.”

(8) The available Perturb-seq datasets have limitations beyond on the control of the authors. 1) The set of genes that are perturbed. The authors address this by simply sub-setting their analysis to the intersection of genes represented in the perturbation and observational data. However, this may mean that a true ancestor of X is not modeled/perturbed, limiting the formal claims that can be made. Additionally, some proportion of genes that are nominally perturbed show little to no actual perturbation effect (for example, due to poor guide RNA choice) which will also lead to missing ancestors.

We now clarify that Perturb-seq can only identify root causal genes among the adequately perturbed set of genes in the Discussion:

“Modern genome-wide Perturb-seq datasets also only adequately perturb and measure a few thousand, rather than all, gene expression levels. RCSP can only identify root causal genes within this perturbed and measured subset.”

(9) The authors provide no mechanism for statistical inference/significance for their results at either the individual or aggregated level. While I am a proponent of using effect sizes more than p-values, there is still value in understanding how much signal is present relative to a reasonable null.

We now explain that RCSP does not perform statistical inference in Methods because it is not clear how to define the appropriate cut-off for the RCS score under the null distribution:

“We focus on statistical estimation rather than statistical inference because Φ_i > 0 when E_i causes Y under mild conditions, so we reject the null hypothesis that Φ_i \= 0 for many genes if many gene expression levels cause Y. However, just like a machine typically breaks down due to only one or a few root causal problems, we hypothesize that only a few genes have large RCS scores Φ_i ≫ 0 even in complex disease.”

(10) I agree with the authors that age coming out of a “root cause” is potentially encouraging. However, it is also quite different in nature to expression, including being “measured” exactly. Will RCSP be biased towards variables that have lower measurement error?

We tested the above hypothesis by plotting sequencing depth against the D-RCS scores of each gene. We observed a small negative correlation between sequencing depth and D-RCS scores, indicating the D-RCS scores are slightly biased upwards with low sequencing depth. However, genes with the largest D-RCS scores exhibited a wide variety of sequencing depths in both MS and AMD, suggesting that sequencing depth has minimal effect on the largest D-RCS scores. We now explain these results for AMD in the Supplementary Materials:

“Theorem 1 states that RCS scores may exhibit bias with insufficient sequencing depth. The genes with large D-RCS scores may therefore simply have low sequencing depths. To test this hypothesis, we plotted sequencing depth against D-RCS scores. Consistent with Theorem 1, we observed a small negative correlation between D-RCS and sequencing depth (ρ \= −0.16, p=2.04E-13), and D-RCS scores exhibited greater variability at the lowest sequencing depths (Supplementary Figure 8). However, genes with the largest D-RCS scores had mean sequencing depths interspersed between 20 and 3000. We conclude that genes with the largest D-RCS scores had a variety of sequencing depths ranging from low to high.”

We also report the results for MS:

“We plot sequencing depth against the D-RCS scores of each gene similar to the AMD dataset. We again observed a small negative correlation (ρ \= −0.136, p_<_2.2E-16), indicating that genes with low sequencing depths had slightly higher D-RCS scores on average (Supplementary Figure 12). However, genes with the largest D-RCS scores again had a variety of sequencing depths. We conclude that sequencing depth has minimal correlation with the largest D-RCS scores.”

(11) Finally, it’s a stretch to call K562 cells “lymphoblasts.” They are more myeloid than lymphoid.

We now clarify that K562 cells are undifferentiated blast cells that can be induced to differentiate into lymphoblasts in Results:

“We next ran RCSP on 137 samples collected from CD4+ T cells of multiple sclerosis (MS; GSE137143) as well as Perturb-seq data of 1,989,578 undifferentiated blast cells that can be induced to differentiate into lymphoblasts, or the precursors of T cells and other lymphocytes.”

eLife Assessment

This valuable study investigates how inter-organ communication between the tracheal stem cells and the fat body plays a key role in the directed migration of tracheal stem cells in Drosophila. The evidence supporting the claims of the authors is solid. The work would be of interest to researchers in the fields of developmental biology and cancer biology.

Joint public review:

Summary

In this manuscript, Dong et al. study the directed cell migration of tracheal stem cells in Drosophila pupae. The authors study how the directionality of these cells is regulated along the dorsal trunk. They show that inter-organ communication between the tracheal stem cells and the nearby fat body plays a role in posterior migration. They provide compelling evidence that Upd2 production in the fat body and JAK/STAT activation in the tracheal stem cells play a role. Moreover, they show that JAK/STAT signalling might induce the expression of apicobasal and planar cell polarity genes in the tracheal stem cells which appear to be needed to ensure unidirectional migration. Finally, the authors suggest that trafficking and vesicular transport of Upd2 from the fat body towards the tracheal cells might be important.

Strengths

The manuscript is well written and presents extensive and varied experimental data to show a link between Upd2-JAK/STAT signaling from the fat body and tracheal progenitor cell migration. The authors provide convincing evidence that the fat body, located near the trachea, secretes vesicles containing the Upd2 cytokine and that affecting JAK-STAT signaling results in aberrant migration of some of the tracheal stem cells towards the anterior. Using ChIP-seq as well as analysis of GFP-protein trap lines of planar cell polarity genes in combination with RNAi experiments, the authors show that STAT92E likely regulates the transcription of planar cell polarity genes and some apicobasal cell polarity genes in tracheal stem cells which appear to be needed for unidirectional migration. The work presented here provides some novel insights into the mechanism that ensures polarized migration of tracheal stem cells, preventing bidirectional migration. This might have important implications for other types of directed cell migration in invertebrates or vertebrates including cancer cell migration. Overall, the authors have substantially improved their manuscript since the first submission but there are still some weaknesses.

Weaknesses

Overall, the manuscript lacks insights into the potential significance of the observed phenotypes and of the proposed new signaling model. Most of our concerns could be dealt with by adjusting the text (explaining some parts better and toning down some statements).

(1) Directional migration of tracheal progenitors is only partially compromised, with some cells migrating anteriorly and others maintaining their posterior migration, a quite discrete phenotype. The strongest migration defects quantified in graphs (e.g. 100 μm) are not shown in images, since they would be out of frame, it would be beneficial to see them. In addition, the consequence of defects in polarized migration on tracheal development is not clear and data showing phenotypes on the final trachea morphology in pupae are not explained nor linked to the previous phenotypes.

(2) Some important information is lacking, such as the origin of mutant and UAS-RNAi lines, which are not reported in the material and methods. For instance, mutants for components of the JAK-STAT pathway are used but not described. Are they all viable at the pupal stage? Otherwise, pupae would not be homozygous mutants. From the figure legend, it seems that the Stat92EF allele has been used, which is a point mutation, thus not leading to an absence of protein. If the hopTUM allele has been used, as mentioned in the legend, it is a gain-of-function allele. Thus, the authors should not conclude that "The aberrant anterior migration of tracheal progenitors in the absence of JAK/STAT components led to impairment of tracheal integrity and caused melanization in the trachea (Figure 3-figure supplement 1E-I)".

(3) The authors observe that tracheal progenitors display a polarized distribution of Fat that is controlled by JAK-STAT signaling. However, this conclusion is made from a single experiment using only 3 individuals with no statistics. This is insufficient to support the claim that "JAK/STAT signaling promotes the expression of genes involved in planar cell polarity leading to asymmetric localization of Fat in progenitor cells", as mentioned in the abstract, or that "the activated tracheal progenitors establish a disciplined migration through the asymmetrical distribution of polarity proteins which is directed by an Upd2-JAK/STAT signaling stemming from the remote organ of fat body."

(4) The authors demonstrate that Upd2 is transported through vesicles from the fat body to the tracheal progenitors. It remains somewhat unclear in the proposed model how Upd2 activates JAK-STAT signaling. Are vesicles internalized, as it seems to be proposed, and thus how does Upd2 activate JAK-STAT signaling intracellularly? Or is Upd2 released from vesicles to bind Dome extracellularly to activate the JAK-STAT pathway? Moreover, it is not clear nor discussed what would be the advantage of transporting the ligand in vesicles compared to classical ligand diffusion.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public review):

Summary:

In this manuscript, Dong et al. study the directed cell migration of tracheal stem cells in Drosophila pupae. The migration of these cells which are found in two nearby groups of cells normally happens unidirectionally along the dorsal trunk towards the posterior. Here, the authors study how this directionality is regulated. They show that inter-organ communication between the tracheal stem cells and the nearby fat body plays a role. They provide compelling evidence that Upd2 production in the fat body and JAK/STAT activation in the tracheal stem cells play a role. Moreover, they show that JAK/STAT signalling might induce the expression of apicobasal and planar cell polarity genes in the tracheal stem cells which appear to be needed to ensure unidirectional migration. Finally, the authors suggest that trafficking and vesicular transport of Upd2 from the fat body towards the tracheal cells might be important.

Strengths:

The manuscript is well written. This novel work demonstrates a likely link between Upd2JAK/STAT signalling in the fat body and tracheal stem cells and the control of unidirectional cell migration of tracheal stem cells. The authors show that hid+rpr or Upd2RNAi expression in a fat body or Dome RNAi, Hop RNAi, or STAT92E RNAi expression in tracheal stem cells results in aberrant migration of some of the tracheal stem cells towards the anterior. Using ChIP-seq as well as analysis of GFP-protein trap lines of planar cell polarity genes in combination with RNAi experiments, the authors show that STAT92E likely regulates the transcription of planar cell polarity genes and some apicobasal cell polarity genes in tracheal stem cells which appear to be needed for unidirectional migration. Moreover, the authors hypothesise that extracellular vesicle transport of Upd2 might be involved in this Upd2-JAK/STAT signalling in the fat body and tracheal stem cells, which, if true, would be quite interesting and novel.

Overall, the work presented here provides some novel insights into the mechanism that ensures unidirectional migration of tracheal stem cells that prevents bidirectional migration. This might have important implications for other types of directed cell migration in invertebrates or vertebrates including cancer cell migration.

Weaknesses:

It remains unclear to what extent Upd2-JAK/STAT signalling regulates unidirectional migration. While there seems to be a consistent phenotype upon genetic manipulation of Upd2-JAK/STAT signalling and planar cell polarity genes, as in the aberrant anterior migration of a fraction of the cells, the phenotype seems to be rather mild, with the majority of cells migrating towards the posterior.

We agree that the phenotype is mild, as perturbing JAK/STAT signaling in the progenitors specifically affects the coordinated migration of the cells rather than alters their direction or completely blocks migration. Our data indicate that inter-organ communication ensures coordinated behavior of the progenitor cells, although the differential responses exhibited by individual cells represent an interesting unresolved issue that awaits future in-depth investigation.

While I am not an expert on extracellular vesicle transport, the data presented here regarding Upd2 being transported in extracellular vesicles do not appear to be very convincing.

We performed additional PLA experiments which support the interaction between Upd2 and the core components of extracellular vesicles (revised Figure 8). Furthermore, we performed electron microscopy to visualize the Lbm-containing vesicles in fat body (Figure 8-figure supplement 1D).

These data are now provided in the revised manuscript.

Major comments:

(1) The graphs showing the quantification of anterior (and in some cases also posterior migration) are quite confusing. E.g. Figure 1F (and 5E and all others): These graphs are difficult to read because the quantification for the different conditions is not shown separately. E.g. what is the migration distance for Fj RNAi anterior at 3h in Fig5E? Around -205micron (green plus all the other colors) or around -70micron (just green, even though the green bar goes to -205micron). If it's -205micron, then the images in C' or D' do not seem to show this strong phenotype. If it's around -70, then the way the graph shows it is misleading, because some readers will interpret the result as -205. Moreover, it's also not clear what exactly was quantified and how it was quantified. The details are also not described in the methods. It would be useful, to mark with two arrowheads in the image (e.g. 5 A' -D') where the migration distance is measured (anterior margin and point zero).

Overall, it would be better, if the graph showed the different conditions separately. Also, n numbers should be shown in the figure legend for all graphs.

We apologize for those inappropriate presentation and insufficient description and thank you for kindly pointing them out. We used different colors to represent different genotypes, and the columns were superimposed. we chose to show the quantification in different conditions separately in the revised Figures. The anterior migration distance for Fj RNAi is around 70 µm.

We now provided detailed description in the revised methods. For migration distance measurement, we took snapshots at 0hr\ 1hr\ 2hr and 3hr, and measured the distance from the starting point (the junction of TC and DT) to the leading edge of progenitor clusters. The velocity formula: v=d (micrometer)/t (min). As you kindly suggested, we indicated the anterior margin and point zero in the corresponding panels. We have added n number in the legends.

(2) Figure 2-figure supplement 1: C-L and M: From these images and graph it appears that Upd2 RNAi results in no aberrant anterior migration. Why is this result different from Figures 2D-F where it does?

The fat body-expressing lsp2-Gal4 was used in Figure 2-figure supplement 1C-L and Figure 2D-F, while trachea specific btl-Gal4 was used in Figure 2-figure supplement 1K-L. The lsp2-Gal4-driven but not btl-Gal4-driven upd2RNAi causes aberrant anterior migration, suggesting that fat bodyderived Upd2 plays a role. We have further clarified this in the text.

(3) Figure 5F: The data on the localisation of planar cell polarity proteins in the tracheal stem cell group is rather weak. Figure 5G and J should at least be quantified for several animals of the same age for each genotype. Is there overall more Ft-GFP in the cells on the posterior end of the cell group than on the opposite side? Or is there a more classic planar cell polarity in each cell with FtGFP facing to the posterior side of the cell in each cell? Maybe it would be more convincing if the authors assessed what the subcellular localisation of Ft is through the expression of Ft-GFP in clones to figure out whether it localises posteriorly or anteriorly in individual cells.

We staged the animals, measured several animals for each genotype and provided the quantifications in the revised manuscript. The level of Ft-GFP is higher in the cells at the frontal edge. We tried to examine the expression of Ft-GFP at single-cell level. However, this turned out to be technically difficult because the tracheal stem cells are not regularly arranged as epithelial cells and the proximal-distant axis of the tracheal stem cells remains unclear. We thus decided to measure the fluorescence signal of groups of stem cells along the DT regardless of their individual polarity within cells.

(4) Regarding the trafficking of Upd2 in the fat body, is it known, whether Grasp65, Lbm, Rab5, and 7 are specifically needed for extracellular vesicle trafficking rather than general intracellular trafficking? What is the evidence for this?

In our experiments, knocking down rab5, rab7, grasp65 or lbm in trachea using btl-Gal4 did not cause abnormality in the disciplined migration, which excludes their intracellular contribution in the trachea (Figure 7-figure supplement 1). Perturbation of Grasp65 or Lbm in fat body increased intracellular upd2-containing vesicles, indicating that intracellular production is functional (Figure 6J). The Grasp65 is specifically required for Upd2 production. Lbm, Rab5 and Rab7 are important of vesicle trafficking. Our conclusion does not pertain to extracellular or intracellular compartment.

(5) Figure 8A-B: The data on the proximity of Rab5 and 7 to the Upd2 blobs are not very convincing.

The confocal images indicate the proximity of Rab5 and Rab7 to the Upd2 vesicles. We interpret the proximity together with the results from Co-IP and PLA data (Figure 8E-K).

(6) The authors should clarify whether or not their work has shown that "vesicle-mediated transport of ligands is essential for JAK/STAT signaling". In its current form, this manuscript does not appear to provide enough evidence for extracellular vesicle transport of Upd2.

Lbm belongs to the tetraspanin protein family that contains four transmembrane domains, which are the principal components of extracellular vesicles. We show that Lbm interacts with Upd2. The JAK/STAT signaling depends on the Upd2 in the fat body as well as vesicle trafficking machinery. Furthermore, we performed electron microscopy and show the presence of Lbm-containing vesicles in fat body (Figure 8-figure supplement 1D).

(7) What is the long-term effect of the various genetic manipulations on migration? The authors don't show what the phenotype at later time points would be, regarding the longer-term migration behaviour (e.g. at 10h APF when the cells should normally reach the posterior end of the pupa). And what is the overall effect of the aberrant bidirectional migration phenotype on tracheal remodelling?

We observed that the integrity of tracheal network especially the dorsal trunk was impaired, which may be due to incomplete regeneration (Figure 3-figure supplement1E-I).

(8) The RNAi experiments in this manuscript are generally done using a single RNAi line. To rule out off-target effects, it would be important to use two non-overlapping RNAi lines for each gene.

We validated the phenotype using several independent RNAi alleles.

Reviewer #2 (Public review):

Summary:

This work by Dong and colleagues investigates the directed migration of tracheal stem cells in Drosophila pupae, essential for tissue homeostasis. These cells, found in two nearby groups, migrate unidirectionally along the dorsal trunk towards the posterior to replenish degenerating branches that disperse the FGF mitogen. The authors show that inter-organ communication between tracheal stem cells and the neighboring fat body controls this directionality. They propose that the fat body-derived cytokine Upd2 induces JAK/STAT signaling in tracheal progenitors, maintaining their directional migration. Disruption of Upd2 production or JAK/STAT signaling results in erratic, bidirectional migration. Additionally, JAK/STAT signaling promotes the expression of planar cell polarity genes, leading to asymmetric localization of Fat in progenitor cells. The study also indicates that Upd2 transport depends on Rab5- and Rab7-mediated endocytic sorting and Lbm-dependent vesicle trafficking. This research addresses inter-organ communication and vesicular transport in the disciplined migration of tracheal progenitors.

Strengths:

This manuscript presents extensive and varied experimental data to show a link between Upd2JAK/STAT signaling and tracheal progenitor cell migration. The authors provide convincing evidence that the fat body, located near the trachea, secretes vesicles containing the Upd2 cytokine. These vesicles reach tracheal progenitors and activate the JAK-STAT pathway, which is necessary for their polarized migration. Using ChIP-seq, GFP-protein trap lines of planar cell polarity genes, and RNAi experiments, the authors demonstrate that STAT92E likely regulates the transcription of planar cell polarity genes and some apicobasal cell polarity genes in tracheal stem cells, which seem to be necessary for unidirectional migration.

Weaknesses:

Directional migration of tracheal progenitors is only partially compromised, with some cells migrating anteriorly and others maintaining their posterior migration.

Our results suggest that Upd2-JAK/STAT signaling is required for the consistency of disciplined migration. Although only a few tracheal progenitors display anterior migration, these cells lose the commitment of directional movement. We acknowledge that the phenotype is moderate.

Additionally, the authors do not examine the potential phenotypic consequences of this defective migration.

We examined the long-term effects of the aberrant migration and observed an impairment of tracheal integrity and melanized tracheal branches (Figure 3-figure supplement1E-I).

It is not clear whether the number of tracheal progenitors remains unchanged in the different genetic conditions. If there are more cells, this could affect their localization rather than migration and may change the proposed interpretation of the data.

We examined the progenitor cell number in bidirectional movement samples and control group. The results show that cell number does not exhibit a significant difference between control and bidirectional movement groups (Figure 3-figure supplement 1).

Upd2 transport by vesicles is not convincingly shown.

We performed additional PLA experiments to further support the interaction between Upd2 and the core components of extracellular vesicles. Furthermore, we performed electron microscopy and show the presence of Lbm-containing vesicles in fat body (Figure 8-supplement 1D). Additional experiments such as colocalization and Co-IP assay and better quantification are provided in the revised manuscript (see revised Figure 8).

Data presentation is confusing and incomplete.

We used different colors to represent different genotypes, and the columns were superimposed. we changed the graphs to show the quantification in different conditions separately. We revised data presentation to avoid confusing.

Reviewer #3 (Public review):

Summary:

Dong et al tackle the mechanism leading to polarized migration of tracheal progenitors during Drosophila metamorphosis. This work fits in the stem cell research field and its crucial role in growth and regeneration. While it has been previously reported by others that tracheal progenitors migrate in response to FGF and Insulin signals emanating from the fat body in order to regenerate tracheal branches, the authors identified an additional mechanism involved in the communication of the fat body and tracheal progenitors.

Strengths:

The data presented were obtained using a wide range of complementary techniques combining genetics, molecular biology, quantitative, and live imaging techniques. The authors provide convincing evidence that the fat body, found in close proximity to the trachea, secrete vesicles containing the Upd2 cytokine that reach tracheal progenitors leading to JAK-STAT pathway activation, which is required for their polarized migration. In addition, the authors show that genes regulating planar cell polarity are also involved in this inter-organ communication.

Weaknesses:

(1) Affecting this inter-organ communication leads to a quite discrete phenotype where polarized migration of tracheal progenitors is partially compromised. The study lacks data showing the consequences of this phenotype on the final trachea morphology, function, and/or regeneration capacities at later pupal and adult stages. This could potentially increase the significance of the findings.

Regarding your kind suggestion, we examined the long-term effects of the aberrant migration and observed the impairment of tracheal integrity and melanized tracheal branches (Figure 3-figure supplement1E-I).

(2) The conclusions of this paper are mostly well supported by data, but some aspects of data acquisition and analysis need to be clarified and corrected, such as recurrent errors in plotting of tracheal progenitor migration distance that mislead the reader regarding the severity of the phenotype.

We used different colors to represent different genotypes, and the columns were superimposed. we changed the graphs to show the quantification in different conditions separately. We thank you for kindly pointing it out.

(3) The number of tracheal progenitors should be assessed since they seem to be found in excess in some genetic conditions that affect their behavior. A change in progenitor number could lead to crowding, thus affecting their localization rather than migration capacities, thereby changing the proposed interpretation. In addition, the authors show data suggesting a reduced progenitor migration speed when the fat body is affected, which would also be consistent with a crowding of progenitors.

We examined the cell number in bidirectional movement samples and control group. We examined cell number and cell proliferation and observed that there was no significance between control and bidirectional movement groups (Figure 3-figure supplement 2).

(4) The authors claim that tracheal progenitors display a polarized distribution of PCP proteins that is controlled by JAK-STAT signaling. However, this conclusion is made from a single experiment that is not quantified and for which there is no explanation of how the plot profile measurements were performed. It also seems that this experiment was done only once. Altogether, this is insufficient to support the claim. Finally, a quantification of the number of posterior edges presenting filopodia rather than the number of filopodia at the anterior and posterior leading edges would be more appropriate.

We staged the animals, measured several animals for each genotype and provided the quantifications in the revised manuscript. The level of Ft-GFP is higher in the cells at the frontal edge. We tried to examine the expression of Ft-GFP at single-cell level. However, this turned out to be difficult due to the fact that the tracheal stem cells are not regularly patterned as epithelial cells and the proximaldistant axis of tracheal stem cells is not well defined. We thus decided to measure the fluorescence signal of groups of stem cells along the DT regardless of their individual polarity.

(5) The authors demonstrate that Upd2 is transported through vesicles from the fat body to the tracheal progenitors where they propose they are internalized. Since the Upd2 receptor Dome ligand binding sites are exposed to the extracellular environment, it is difficult to envision in the proposed model how Upd2 would be released from vesicles to bind Dome extracellularly and activate the JAK-STAT pathway. Moreover, data regarding the mechanism of the vesicular transport of Upd2 are not fully convincing since the PLA experiments between Upd2 and Rab5, Rab7, and Lbm are not supported by proper positive and negative controls and co-immunoprecipitation data in the main figure do not always correlate to the raw data.

We use molecular modeling to show that Upd2 and Lbm intermingle, and Upd2 is not entirely encapsulated in vesicles (Figure 8-supplement 1E). We performed PLA experiments using the animals not expressing upd2-Cherry as negative control (Figure 8 E-J). We corrected the Co-IP panel and apologize for this error.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Minor comments:

(1) Figure 1-figure supplement 1: E: How was the migration velocity assessed? By live imaging individual cells or following the cell front of the group? Over what time period? Do the data points in the graph correspond to individual cells or the cell group? It would be important to show confocal images that go along with this quantification.

We took snapshots of pupae at 0hr\ 1hr\ 2hr and 3hr, and measured the distance covered by the migrating progenitor cells from the start place (the junction of TC and DT) to the leading edge of progenitor groups. We then calculated the migration rate by v=d (micrometer)/t (min). As the progenitor cells revolve around and migrate along the DT, tracking single tracheoblast through intact cuticle is technically challenging. We have therefore measured the leading edge as a proxy to the whole cell group. We agree with you that time-lapse imaging is favorable for analysis of migration.

(2) Figure 1-figure supplement 1: F: Why is there Gal80ts in the genotype? (and in Figure 1H). Also, what pupal age was used for this quantification?

Expression of hid and rpr in L3 stage impaired fat body integrity and adipocyte abundance, and caused lethality. Gal80ts was used for controlling the expression of rpr.hid. The pupal at 0hr APF were used in EdU experiment.

(3) Figure 2C: what is shown in the 6 columns (why 3 each for control and rpr/hid)?

We conducted 3 replicates of each group for control and rpr.hid.

(4) In the methods, several Drosophila stocks are listed as 'source:" from a particular person (e.g. Dr Ma). Please list the real source of this stock, e.g. Bloomington stock number, or the lab and publication in which the stock was originally made.

We provide the information on these stocks in the revised methods.

(5) The SKOV3 carcinoma cell and S2 cell work is not described in the methods.

We added detailed description of this experiment in the revised method-Cell culture and transfection. 

(6) Figure 6 (F) 'Bar graph plots the abundance of Upd2-mCherry-containing vesicles in progenitors.' What does abundance mean? What was quantified, the number of vesicles, or the mean intensity? This is also not mentioned in the methods.

We counted the number of Upd2-mCherry-containing vesicles in fat body cells and trachea progenitors and added the description of measurement in the method.

(7) There are a few language mistakes throughout the manuscript. E.g.

(a) Line 117 and other places: Language: 'fat body' should be 'the fat body'.

We thank you for pointing out these errors and corrected it accordingly.

(b) Line 1276 Language mistakes: 'Video 1 3D-view of confocal image stacks of tracheal progenitors and fat body. Scale bar: 100 μm. Genotypes: UAS-mCD8-GFP/+;lsp2-Gal4,P[B123]-RFP-moe/+.' :stacks and genotypes should be singular.

We fixed these errors and thank you for kindly pointing them out. We also proofread the entire manuscript to assure accuracy.

(8) In general, it is hard to figure out the exact genotypes used in experiments. This is mostly not written very clearly in the figure legends. E.g. Figure 2: genotype for A-C missing in figure legend (is B from control animals?)

We added genotypes in the figure legends. For Figure 2, A and C lsp2-Gal4,P[B123]-RFP-moe/+ for control, UAS-rpr-hid/+;Gal80ts/+;lsp2-Gal4,P[B123]-RFP-moe/+ for rpr.hid; B from control animals.

Reviewer #2 (Recommendations for the authors):

Major comments:

(1) The phenotype resulting from Upd2 downregulation by RNAi is subtle and shown by unconvincing images. In addition, these phenotypes are analyzed using only one RNAi line.

We used two independent alleles of upd2RNAi from THFC (THU1288 and THU1331), and observed similar phenotype. For RNAi experiments, we always use multiple independent alleles.

(2) The authors should analyze the phenotypic consequences of directional migration changes. Is there an effect on tracheal remodeling?

We observed that the integrity of tracheal network especially the dorsal trunk was impaired and that melanized tracheal branches were present, which may be due to incomplete regeneration (Figure 3figure supplement1E-I).

(3) The number of tracheal progenitors should be quantified, as some genetic conditions may affect cell numbers, as is apparent in some panels.

We examined cell number and cell proliferation and observed that there was no significance between control and bidirectional movement groups (Figure 3-figure supplement 1).

(4) The data on PCP protein distribution are unconvincing, unquantified, and insufficient to support one of the main conclusions of the study, which is stated in the abstract: "JAK/STAT signaling promotes the expression of genes involved in planar cell polarity, leading to asymmetric localization of Fat in progenitor cells."

We staged the animals, measured several animals for each genotype and provided the quantifications in the revised manuscript. The level of Ft-GFP is higher in the cells at the frontal edge. We tried to examine the expression of Ft-GFP at single-cell level. However, this turned out to be difficult due to the fact that the tracheal stem cells are not regularly patterned as epithelial cells and the proximaldistant axis of tracheal stem cells is not well defined. We thus decided to measure the fluorescence signal of groups of stem cells along the DT regardless of their individual polarity.

Minor comments:

(1) Language should be revised. In many places in the manuscript, starting in line 113, "fat body" should be "the fat body".

Thank you for pointing out this error. We corrected it accordingly.

(2) Genotypes used in experiments should be described.

We added all the genotypes. We proofread the entire manuscript to complete the figure legends for genotypes.

(3) Line 67, the reference to "The progenitor cells reside in Tr4 and Tr5 metameres and start to move along the tracheal branch" should include (Chen and Krasnow, Science 2014).

We added the reference in the manuscript.

(4) Line 1081, Figure 7 Legend. "Bar graph plots the abundance of Upd2-mCherry-containing vesicles" Abundance is the number of vesicles? The graph displays the average number of vesicles? Please explain and describe the quantification.

The bar graph represents the number of Upd2-mCherry-containing vesicles in different conditions. We quantified the number of vesicles per area.

(5) Figure 1 (I-J) What is shown on the panels? Progenitors marked with? This information is not present in the figure or figure legend. Same for Figure 2 (D-E).

Figure 1I-J show the vector of migrating progenitors. We added the information in the legends. The tracheal cells were labeled by nls-mCherry in Figure 1I-J. In Figure 2D-E, the progenitors were marked with P[B123]-RFP-moe.

(6) Figure 3 Q, Stat92E-GFP values in the graph are not well-explained. What do the numbers in the y-axis refer to?

y-axis represents the intensity of Stat92E-GFP normalized to control. We have changed the y-axis label to ‘normalized Stat92E-GFP intensity’ in the legends.

(7) In general, figures and figure legends must be revised. Sometimes stainings are not well-defined, some scale bars are missing and plots do not say what the values are.

We apologized for inadequate information and have revised the figures and legends accordingly.

Reviewer #3 (Recommendations for the authors):

Several points should be addressed by the authors in order to improve their manuscript.

Major points:

(1) The phenotype obtained from decreasing the inter-organ signaling is quite discrete. It is further weakened by the fact that the images chosen to illustrate the measures are not really convincing. No image at 1h APF shows any clear anterior migration. Based on the scale, most of the images at 3h APF do not show a striking difference compared to the control, and in any case, stronger phenotypes would be missed anteriorly since they would thus be out of frame. In addition, at 3h APF, progenitors migrating anteriorly from Tr5 position get mixed with those migrating posteriorly from Tr4 so it is not clear how measurements were made. Given that most phenotypes are observed upon the use of RNAis, it is possible that phenotypes are weak due to persistent gene expression. Using null clones for dome, hop, or stat in progenitors could therefore aggravate the phenotypes and support further the significance of the study. Finally, assessing the consequences of compromised fat body-tracheal communication on trachea morphology, function, and regeneration later in pupal development and on adult flies would also help strengthen the importance of the findings.

We agree with you that anteriorly migrated Tr5 progenitors adjoining Tr4 progenitor hinders measurements and that mutants may give stronger phenotype than RNAi lines. We only measured Tr4 progenitors (instead of Tr5) when assessing anterior migration. Thus, we performed experiments using mutant alleles, which gave aberrant migration of tracheal progenitors (Figure 3-figure supplement1A-D). We can now show that the integrity of tracheal network especially dorsal trunk was impaired, which may be due to incomplete regeneration (Figure 3-figure supplement1E-I).

(2) Although the authors did not observe defects in tracheal progenitor proliferation, progenitors seem to be present in excess in some key genetic background (e.g, upon expression of rpr.hid, statRNAi, Rab-RNAi or in the presence of BFA). This excess could be the result of another mechanism than proliferation (recruitment of extra progenitors since it is not clear how they originate, defect in apoptosis...) and could impact the localization of progenitors, those being pushed anteriorly as a consequence of crowding. A proper characterization of tracheal progenitor number would thus help to discriminate between defects in migration or crowding. This point could also be addressed by performing individual tracking of tracheal progenitors, to find out whether each progenitor is indeed migrating in the wrong direction or if the movement assessed by the global tracking method that is used is just a consequence of progenitor excess.

We examined the cell number in bidirectional movement samples and control group. The results show that there was no significance between control and bidirectional movement groups (Figure 3figure supplement 1). We also tried to follow every progenitor, but were unable to obtain convincing results with P[B123]-RFP-moe, as tracking single tracheoblast through intact cuticle is technically challenging.

(3) Regarding the ChIP-seq experiment, an explanation of why choosing the "establishment of planar polarity" family should be provided since data indicate a quite low GeneRatio. Indeed, the "cell adhesion" family seems a more obvious candidate, which would be further supported by the fact that the JAK-STAT pathway has been shown to affect cell adhesion components such as ECadherin and FAK (Silver and Montell 2001, Mallart et al 2024). Also, have these known targets of JAK-STAT signaling been found in the ChIP-seq data? Since filopodia polarization is affected in tracheal progenitors when JAK-STAT signaling is decreased, the same question also applies to enabled, which is involved in filopodia formation and has been recently identified as a target of JAK-STAT signaling.

As you kindly suggested, we tested a number of cell adhesion-related genes such as E-Cadherin (shg), fak, robo2 and enabled (ena). We did not observe an apparent aberrancy in the migration of tracheal progenitors (Figure 5-supplement 1J).

(4) Data investigating PCP protein distribution is not convincing, not quantified, and not sufficient to draw one of the main conclusions of the study, which is even written in the abstract "JAK/STAT signaling promotes the expression of genes involved in planar cell polarity leading to asymmetric localization of Fat in progenitor cells."

We better quantified the abundance of Ft in in the progenitors in the frontal edge and those lagging behind. The traces plot multiple replicates in the figures. The level of Ft-GFP is higher in the cells at the frontal edge.

(5) Overall, the figures together with their caption and/or the material and methods section lack some important information for the reader to fully understand the data. In addition, some errors are found in multiple plots throughout the article and must be corrected. Here are some examples:

According to your suggestion, we revised legends and methods section to include sufficient information.

(a) Migration distance plots from Figure 3E do not match the data presented in the source data file. It seems that, when creating the plot, instead of superimposing the bars, bars were stacked. This should be corrected for all migration distance plots from Figure 3E onward, including in supplementary figures.

We apologized for misleading representation. We revised it accordingly and show the quantification in different conditions separately.

(b) The number of analyzed flies and/or clusters of tracheal progenitors from different flies should be stated for all quantification or observations made on images. This information is lacking for all migration distance plots, for progenitor migration tracking (Figure 1 I, J), for DIPF reporter in Figure 2J, for plot profiles (Figure 5G, J), for Upd2-Rab5/Rab7/Lbm co-detections, PLA, CoIP, and lbm-pHluorin experiments. This also applies to RNA seq, ChIP seq, and surface proteomics, for which the number of pupae and number of replicates is not indicated.

We changed the graphs to show the quantification and n number in different conditions separately.

We also added the n number of replicates in methods.

(c) How quantifications were performed is not sufficiently explained. For example, the reference point for migration distance measurement is not defined, and neither is whether the measures were made on fixed or live imaging samples. In fluorescence intensity measurements and Upd2 vesicle counting, information on whether measures were made on a single z slice or on a projection of several z slices should be stated together with what ROI and which FIJI tool for quantification were used. For plot profiles, the same information regarding z slices misses together with how the orientation, the thickness, and the length of the line were chosen, and again the number of times the experiment was conducted should be mentioned and error bars should appear on graphs.

We thank this reviewer for the suggestions which help clarify the methodology of our experiments and improve presentation of our data. We have made the changes according to the suggestions and modified our methods section and the related figures to incorporate these changes.

For measuring the migration distance of tracheal progenitors, we took snapshots of living pupae at 0hr\ 1hr\ 2hr and 3hr APF, and measured the migration distance of tracheal progenitors from the start place (the junction of TC and DT) to the leading edge of progenitor groups.

For the measurements of fluorescent intensity of stat92E-GFP and DIPF, we took z-stack confocal images of samples and quantified the fluorescent intensity using FIJI. Specifically, intensity was quantified for regions of interest, using the Analysis and Measurement tools. To quantify Upd2mCherry vesicles, z-stack confocal images of fat body were taken and the cell counting function of FIJI was used to measure the vesicle number.

To quantify the fluorescent intensity of in vivo tagged Ds, Ft and Fj proteins, a single z slice was used. The expression level of the protein was assessed as the integrated fluorescent intensity normalized to area.

For the measurement of Ft-GFP distribution, a single z slice of the progenitors immediately proximal to the DT was imaged. An arbitrary line was drawn along the migration direction from the starting TC-DT junction to the leading front (the length of the line corresponds to the distribution range of tracheal stem cell clusters). Then, fluorescent intensity along the line was automatically calculated with the imbedded measurement function of Zeiss confocal software.

Minor points:

(1) In several instances, the authors generalize that stem cells migrate to leave their niche, but this is not the case for all stem cells.

The phenomenon that stem cells leave their niche when they are activated is commonly observed. We interpreted the general mechanism from our system of tracheal stem cells. We fully agree with you that it may not be the case for all stem cells. We modified the text accordingly.

(2) Line 122 -a reference paper or an image showing the expression pattern of the lsp2-Gal4 driver is missing.

We added the reference in the manuscript.

(3) Line 136 - The term "traces of individual progenitors" is overstated and should be reformulated as the method used does not seem to be individual cell tracking.

We rephrased accordingly in the revised manuscript.

(4) Line 146 - Fat body and tracheal progenitors are qualified as interdependent organs, in which aspect do tracheal progenitors affect the fat body?

Current knowledge suggests a close inter-organ crosstalk between trachea and fat body: The fly trachea provides oxygen to the body and influences the oxidation and metabolism of the whole body. When the trachea is perturbed, the body is in hypoxia, which causes inflammatory response in adipose tissue as an important immune organ (Shin et al., 2024).

(5) Line 163 - Not all the genes tested are cytokines, so the sentence should be reformulated. In addition, in supplementary Fig2-1 C-J, the KD of hh seems to abolish completely tracheal progenitor migration, which is not commented on.

According to your suggestion, we revised the description on information of the genes tested. We added comments in the revised manuscript regarding phenotypes of hh knockdown. 

(6) Line 180 - Conclusion is made on Dome expression while using a dome-Gal4 construct, which does not necessarily recapitulate the endogenous pattern of dome expression, so it should be reformulated. Ideally, dome expression should be assessed in another way. Also, it is not clear whether GFP is present only in progenitors since images are zoomed.

We revised statement and provided larger view of dome>GFP that shows an enriched expression in the tracheal progenitors (Figure 2-figure supplement 2E), an expression pattern that is consistent with FlyBase.

(7) Line 199 - Is it upd-Gal4 or upd2-Gal4 that is used? Since the conclusion of the experiment is made on upd2, the use of upd-gal4 would not be relevant. If upd2-gal4 is used, it should be corrected. In general, the provenance of the Gal4 lines should be provided. In addition, a strong GFP signal in the trachea is visible on the image in Supplementary Figure 2-2F but not commented on and seems contradictory with the conclusion mentioning that fat body and gut are the main source of Upd2 production.

We removed data obtained from the use of this irrelevant upd-Gal4 line.

(8) Figures:

-  Figure 1 G, H - Scale bar is missing.

We added it accordingly.

-  Figure 1 I, J - The information on the staining is missing.

We added it in the revised manuscript.

-  Figure 2A - Providing explanations of the terms "Count" and "Gene ratio" in the caption would be helpful for readers who are not used to this kind of data. In addition, the color code is confusing since the same color is used for the selected gene family and for high p-values (the same applies to other similar graphs).

Gene ratio refers to the proportion of genes in a dataset that are associated with a particular biological process, function, or pathway. Count indicates the number of genes from input gene list that are associated with a specific GO term. We used redness to indicate a smaller p-value and a higher significance.

-  Figure 2 B, C - What does the color scale represent? What do the columns in C correspond to, different time points, different replicates?

The color scale represents the normalized expression. The columns in C correspond to different replicates of control and rpr.hid.

-  Figure 2 F - The error bars on the 3h APF posterior bars are missing.

We added error bars accordingly.

-  Figure 2 G - The legend "Down-Stable-Up" is in comparison to what?

The control group was generated from the reaction without H2O2. The comparison was relative to the control group.

-  Figure 2 J - The specificity of the DIPF tool that has been created should be validated in other tissues displaying known JAK-STAT activity and/or in conditions of decreased JAK-STAT signaling. In addition, the added value of the tool as compared to the JAK-STAT activity reporter used later, which has been well characterized, is not obvious.

We added the signal of DIPF in fat body and salivary gland, both of which harbor active JAK/STAT signaling (Figure 2-figure supplement 2F-H). As opposed to the well characterized Stat92E-GFP reporter that assays the downstream transcription activity, the DIPF reporter measures the upstream event of receptor dimerization.

-  Figure 3 I-P - Reporter tool validation in Images I-L could be moved to supplementary data. In images M-P, staining of nuclei and/or membranes would be useful to assess cell integrity.

We revised the figures accordingly.

-  Figure 3Q and similar plots in the following figures do not explain the normalization performed and how it can be higher than 1 in control conditions.

In these figures, we normalized the signal relative to control groups, e.g., The value of Stat92E-GFP in btl-GFP control group was set to 1 in the previous Figure 3Q (revised Figure 3-supplementary

Figure B-J).

-  Figure 4C - These representations lack explanations to be fully understood by a broad audience.

The figure showing that Stat92E binding was detected in the promoters and intronic regions (the orange peaks) of genes functioning in distal-to-proximal signaling, such as ds, fj, fz, stan, Vang and fat2. We added the information in figure legend according to your suggestion.

-  Figure 5 K,L - What is the x-axis missing, together with the method of tracking used?

The x-axis refers to time of recording from a t stack series with a time interval of 5 min. We revised method section and provide detailed procedure of this experiment.

-  Figures 6 and 8- The overall figures lack a wider view of the cells/tissues/organs and/or additional staining to understand what is presented.

We showed preparation of fat body. In order to obtain the high resolution of vesicles, we used high magnification. We now added wider views of the tissues under investigation (e.g. Figure 6-figure supplement 1).

-  Figure 6 D,E - The scale bar is missing.

We added it accordingly.

-  Figure 8 O-S - What is the blue staining?

The blue staining shows DAPI-stained nuclei. We have added the information in the legend.

-  PLA experiments can give a lot of non-specific background. What kind of controls have been used in Figure 8 F-J? Negative controls should be done on cells that do not express upd2-mCherry using both antibodies to detect non-specific background, which does not usually appear completely black.

If possible, a positive control using a known protein interacting with Rab5-GFP should be included.

We used the control samples without one of the primary antibodies in previous Figure 8. In the revised Figure 8, we conducted experiment as you suggested with controls that do not express upd2mCherry (Figure 8 E-J).

-  Co-IP experiments - The raw data file for blots is quite hard to read through. Some legends are not facing the right lane and some blots presented in the main figure are difficult to track since several blots are presented in the raw data file. e.g.

(a)  Raw blot for Figure 8 K: the band for mCherry in the IP anti-GFP blot (lane one in K) is not convincing, it is not distinguishable from other aspecific bands. On the reverse IP presented only in raw data, on the input from blot IB anti-mCherry, both lanes present exactly the same bands at 72kb when one of the lanes corresponds to extract from flies not expressing upd2-mCherry.

We thank you for pointing out the incorrect labels. We apologized for the errors and corrected it accordingly.

(b)  Raw blot for Figure 8 L: on the input blot IB anti-GFP, there is a band corresponding to Rab7-GFP in the lane of the extract from flies not expressing Rab7-GFP.

We corrected it.

(c)  Raw data for Figure 8 M: on the last blot, legends are missing above the input Ib anti-GFP blot.

We added the missing legends in the figure.

Shin, M., Chang, E., Lee, D., Kim, N., Cho, B., Cha, N., Koranteng, F., Song, J.J., and Shim, J. (2024). Drosophila immune cells transport oxygen through PPO2 protein phase transition. Nature 631, 350-359.

eLife Assessment

This important work demonstrates the application of Pro-PRIME, a large language model, to engineer VHH antibodies with enhanced stability for extreme industrial environments. The evidence is convincing, showing through two rounds of design and experimental validation that AI-guided approaches can outperform traditional rational design methods. The solid methodology and results establish a foundation for further exploration of LLM-assisted protein engineering.

Joint Public Review:

Summary:

In this manuscript, the model's capacity to capture epistatic interactions through multi-point mutations and its success in finding the global optimum within the protein fitness landscape highlights the strength of deep learning methods over traditional approaches.

Strengths:

It is impressive that the authors used AI combined with limited experimental validation to achieve such significant enhancements in protein performance. Besides, the successful application of the designed antibody in industrial settings demonstrates the practical and economic relevance of the study. Overall, this work has broad implications for future AI-guided protein engineering efforts.

Reviewing Editor's comments on revised version:

The authors extensively addressed conceptual and methodological points raised by reviewers, as well as constructive comments to clarify the narrative. Consequently, the manuscript experienced a qualitative jump on clarity and appeal for the eLife readership.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public review):

(1) Summary:

In this manuscript, the model's capacity to capture epistatic interactions through multi-point mutations and its success in finding the global optimum within the protein fitness landscape highlights the strength of deep learning methods over traditional approaches.

We thank the reviewer for his/her recognition of our model’s potential and advantages.

(2) Strengths:

It is impressive that the authors used AI combined with limited experimental validation to achieve such significant enhancements in protein performance. Besides, the successful application of the designed antibody in industrial settings demonstrates the practical and economic relevance of the study. Overall, this work has broad implications for future AI-guided protein engineering efforts.

We are thankful for the editor’s appreciation on our work, especially acknowledged the practical application of our model.

(3) Weaknesses:

However, the authors should conduct a more thorough computational analysis to complement their manuscript. While the identification of improved multi-point mutants is commendable, the manuscript lacks a detailed investigation into the mechanisms by which these mutations enhance protein properties. The authors briefly mention that some physicochemical characteristics of the mutants are unusual, but they do not delve into why these mutations result in improved performance. Could computational techniques, such as molecular dynamics simulations, be employed to explore the effects of these mutations?

We thank the reviewer for this good question, which allows us to provide a deeper investigation into the mechanisms by which the mutations significantly enhance the alkali-resistance of proteins. By following the reviewer’s suggestion, we have expanded our analysis by incorporating molecular dynamics (MD) simulations to understand the impact of the mutations. As an example, we focused on the representative alkali-resistant mutant, A57D;P29T, and examined its MD simulation results. As shown in Figure S4A, the two-point mutant of A57D;P29T has a Tm increase of around 8 ℃ and a much stronger binding affinity than the WT. Our analysis of the MD trajectories indicates that the A57D;P29T mutant has a more rigid structure than that of WT due to its lower root mean squared deviation (RMSD) of protein (Figure S4B). Furthermore, we calculated the root mean squared fluctuation (RMSF) for each residue, and realized that the mutant displayed less fluctuation at residue 29 but similar flexibility at residue 57. Interestingly, residues at positions 10, 108 and 118 which spatially distant from residues 29 and 57 in the mutant exhibited remarkable weakened fluctuations than those in the WT (Figure S4C), implying a more rigid structure of the mutant contributing to its improved resistance on high temperature and strong alkalinity. However, Figure S4D shows the AlphaFold3 predicted structures of the WT and the mutant are quite similar.

To unveil the origin of change on structural flexibility, we computed the intramolecular interactions, such as salt bridges and hydrogen bonds for both WT and the mutant. We observed that the mutations increased the number of hydrogen bonds between the mutation sites and the rest of the protein (Figure S4E). However, the overall structure of the mutant did not show significant changes, which is also evident from the solvent-accessible surface area (SASA) analysis (Figure S4F). We also analyzed changes in salt bridges and found that although residue 57 mutated to Histidine, no new salt bridges were formed. Additionally, RMSF results showed that residues 10, 108, and 118 became more rigid, but further analysis revealed that there was no significant change in hydrogen bonds or other interactions in these regions. Overall, the MD results suggest that more hydrogen bonds introduced by the mutations of A57D;P29T stabilize the protein, leading to the enhanced alkali resistance observed in the mutant. These results are now presented in Figure S4 and discussed in detail in the revised manuscript.

Specifically, we have added the following discussion in the main text:

“In order to gain deeper insights into the mechanisms by which the identified mutations enhance protein properties, we performed molecular dynamics (MD) simulations on the best alkali-resistant mutant. The simulation results revealed several key observations that help explain the observed improvements in protein stability and alkali resistance. As shown in Figure S4A, the two-point mutant of A57D;P29T has a Tm increase of around 8℃ and a much stronger binding affinity than the WT. Our analysis of the MD trajectories indicates that the A57D;P29T mutant has a more rigid structure than that of WT due to its lower root mean squared deviation (RMSD) of protein (Figure S4B). Furthermore, we calculated the root mean squared fluctuation (RMSF) for each residue, and realized that the mutant displayed less fluctuation at residue 29 but similar flexibility at residue 57. Interestingly, residues at positions 10, 108 and 118 which spatially distant from residues 29 and 57 in the mutant exhibited remarkable weakened fluctuations than those in the WT (Figure S1C), implying a more rigid structure of the mutant contributing to its improved resistance on high temperature and strong alkalinity. However, Figure S4D shows the AlphaFold3 predicted structures of the WT and the mutant are quite similar. To unveil the origin of change on structural flexibility, we computed the intramolecular interactions, such as salt bridges and hydrogen bonds for both WT and the mutant. We observed that the mutations increased the number of hydrogen bonds between the mutation sites and the rest of the protein (Figure S4E). However, the overall structure of the mutant did not show significant changes, which is also evident from the solvent-accessible surface area (SASA) analysis (Figure S4F). We also analyzed changes in salt bridges and found that although residue 57 mutated to Histidine, no new salt bridges were formed. Additionally, RMSF results showed that residues 10, 108, and 118 became more rigid, but further analysis revealed that there were no significant changes in hydrogen bonds or other interactions in these regions. Taken together, these findings suggest that the enhanced alkali resistance of the mutant is likely due to an overall increase in protein stability, rather than a dramatic change in its structural conformation. The MD simulation results, which are detailed in Figure S4, provide a deeper understanding of how specific mutations can improve protein properties and offer valuable insights for future protein engineering applications.”

And we also included the following content in the SI:

“Molecular Dynamics (MD) simulations

The initial structures for molecular dynamics (MD) simulations of both the wild type and the mutant were predicted using AlphaFold3. To simulate experimental conditions, each protein was placed in a cubic water box containing 0.1 M NaCl. The CHARMM27 force field and the TIP4P water model were applied throughout the simulations. After an initial energy minimization of 50,000 steps, the systems were heated and equilibrated for 1 ns in the NVT ensemble at 300 K followed by an additional 1 ns in the NPT ensemble at 1 atm. The production phase then involved 200-ns simulations with periodic boundary conditions, using a 2 fs integration time step. The LINCS algorithm was used to constrain covalent bonds involving hydrogen atoms, while Lennard-Jones interactions were cut off at 10 Å. Electrostatic interactions were computed with the particle mesh Ewald method, using a 10 Å cutoff and a grid spacing of approximately 1.6 Å with a fourth-order spline. Temperature and pressure were regulated by the velocity rescaling thermostat and Parrinello-Rahman algorithm, respectively. All simulations were performed using GROMACS 2020.4 software packages. Both systems have reached equilibrium according to the analyses of root mean squared deviation (RMSD).”

(4) Additionally, the authors claim that their method is efficient. However, the selected VHH is relatively short (<150 AA), resulting in lower computational costs. It remains unclear whether the computational cost of this approach would still be acceptable when designing larger proteins (>1000 AA). Besides, the design process involves a large number of prediction tasks, including the properties of both single-site saturation and multi-point mutants. The computational load is closely tied to the protein length and the number of mutation sites. Could the authors analyze the model's capability boundaries in this regard and discuss how scalable their approach is when dealing with larger proteins or more complex mutation tasks?

In our prior work, we have demonstrated that our method is applicable to larger proteins as well [Jiang et al., Sci. Adv. 10, eadr2641 (2024)]. For instance, when engineering a protein with 1000 amino acids, inferring the fitness of one million mutants using the model on a single 4090 GPU takes approximately 20 hours. However, it remains infeasible to explore all possible mutations when designing multi-point mutants due to the vast space. To address this challenge, we propose the design of a reliable mutant library. In the first round of experiments, we used the model to score all single-point mutations, and then constructed the multi-point mutant library by combining experimentally tested single-point mutations. In this way, even when designing five-point mutants, we only need to score on the order of millions of mutants, making the inference process time-efficient and fully acceptable. As a result, the number of single-point mutations selected for combination into the multi-point mutant library becomes a crucial parameter that affects both inference time and scope. We limited the number of single-point mutations to between 30 and 50 to strike a balance between efficiency and accuracy.

These results are discussed in the revised manuscript. Specifically, we have added the following discussion at the section 2.2 in the main text:

“Although the model inference is fast, it is not feasible to explore all possible mutations when designing multi-point mutants due to the exponential increase in the number of potential combinations. To manage this challenge, we constructed a mutant library based on a two-stage design process. In the first stage, we scored all single-point mutations using the model, and in the second stage, we combined experimentally validated single-point mutations to create the multi-point mutant library. This approach ensures that even when designing multi-point mutants (e.g., five-point mutants), the number of mutants to score remains in the millions, which is computationally efficient and practical. The number of single-point mutations selected for the multi-point mutant library is a key factor influencing both the computational load and the scope of the design space. To maintain a balance between efficiency and accuracy, we limited the number of single-point mutations to between 30 and 50. This strategic approach allows us to achieve both scalability and precision in our protein engineering tasks.”

Reviewer #2 (Public review):

In this paper, the authors aim to explore whether an AI model trained on natural protein data can aid in designing proteins that are resistant to extreme environments. While this is an interesting attempt, the study's computational contributions are weak, and the design of the computational experiments appears arbitrary.

The reviewer’s comments give us an opportunity to further state the novelty of this study. Despite the AI model has been reported in our previous work [Sci. Adv. 10, eadr2641 (2024)], the unnatural physicochemical properties of proteins, to the best of our knowledge, have never been predicted using AI models. Our preceding work [Sci. Adv. 10, eadr2641 (2024)] has demonstrated that the large language model can predict the performances of the mutants on thermostability, catalytic activity, and binding affinity, etc. However, whether the AI models are able to evaluate the unnatural properties of the mutants remains unexplored. Our work has shown that AI models trained on the natural proteins can be used to design the mutants that resistant extreme conditions, such as strong alkalinity, substantially expanding the application of AI for bioengineering. Moreover, our design of the computational experiments was driven by the nature of the task and the availability of experimental data. We employed different strategies for designing single-point and multi-point mutants, specifically using a zero-shot approach for single-point mutations to overcome the challenge of rare data and fine-tuning the model for multi-point mutations to leverage the experimental data of single-point mutations.

(1) The writing throughout the paper is poor. This leaves the reader confused.

The manuscript has been revised accordingly, and we would like to address the reader’s questions if anything is confused.

(2) The main technical issue the authors address is whether AI can identify protein mutations that adapt to extreme environments based solely on natural protein data. However, the introduction could be more concise and focused on the key points to better clarify the significance of this question.

We thank the reviewer for this comment. We have revised the manuscript, particularly the introduction, where we focused on the research questions, methods, and main findings, while removing excessive background information to improve the manuscript’s conciseness and clarity.

“Protein engineering, situated at the nexus of molecular biology, bioinformatics, and biotechnology, focuses on the design of proteins to introduce novel functionalities or enhance existing attributes[1-3]. With the exponential growth of biological data and computational power, protein engineering has experienced a significant shift towards advanced computational methodologies, particularly deep learning, to expedite the design process and unravel complex protein-function relationships[4-9]. However, a significant challenge in industrial protein engineering is designing proteins with inherent resistance to extreme conditions, such as high temperature and extreme pH environments (acidic or alkaline)[17, 18]. Unlike proteins in natural ecosystems, those used in industrial processes often encounter harsh physical and chemical conditions, necessitating exceptional resilience to maintain functionality[19, 20]. Previous efforts to enhance protein resistance have often relied on rational design and mutant library screening. These methods are typically labor-intensive, inefficient, and yield limited improvements[23-26]. Consequently, the industrial demand for proteins resilient to harsh environments poses a notable absence within the training datasets of Artificial Intelligence (AI) models. Exploring whether AI can achieve the evolution of protein resistance to extreme environments is crucial for broadening protein applications and improving modification efficiency.

Recent advances in large-scale protein language models (LLMs) have enabled zero-shot predictions of protein mutants based on self-supervised learning from natural protein sequences. Although AI-guided protein design has been applied to predict the mutants with greater thermostability and higher activity[34-36], it is unexplored whether these models based on the natural protein information can find the mutants that adapt the unnatural extreme environments, such as the alkaline solution with the pH value higher than 13.

Here, we employed a LLM (large language model) developed by our group, the Pro-PRIME model[27], to predict dozens of mutants of a nano-antibody against growth hormone (a VHH antibody), and examined their fitness, including alkali resistance and thermostability, to evaluate their performance under extreme environments.

We utilized the Pro-PRIME model to score saturated single-point mutations of the VHH in a zero-shot setting, and selected the top 45 mutants for experimental testing. Some mutants exhibited improved alkali resistance, while others demonstrated higher thermal stability or affinity. Subsequently, we fine-tuned the Pro-PRIME model to predict dozens of multi-point mutations. As a result, we obtained three multi-point mutants with enhanced alkali resistance, higher thermostability, as well as strong affinity to the targeted protein. Also, the dynamic binding capacity of the selected mutant did not show significant decline after more than 100 cycles, making it suitable for practical application in industrial production. The selected mutant has been used in practical production and lower the cost for over one million dollars in a year. To the best of our knowledge, this is the first protein product developed by a LLM that has been successfully applied in mass production. Due to the Pro-PRIME model's ability to achieve precise predictions of multi-point mutations with reliance on a small amount of experimental data, our two-round design process involved experimental validation of only 65 mutants in two months, demonstrating remarkable high efficiency. Furthermore, we performed a systematic analysis of these findings and determined that the model can yield more valuable predictive outcomes while remaining consistent with rational design principles. Specifically, within the framework of multi-point combinations, the model's incorporation of negative single-point mutations into the combinatorial space led to exceptional results, showcasing its capacity to capture epistatic interactions. Notably, in striving for global optimum, deep learning methods offer distinct advantages over traditional rational design approaches.”

(3) The authors did not develop a new model but instead used their previously developed Pro-PRIME model. This significantly weakens the novelty and contribution of this work.

While it is true that the Pro-PRIME model was previously developed, the novelty and contribution of this work lie in its novel application to design proteins with properties that are not naturally found or are rare in nature. In our original work, the Pro-PRIME model was used to optimize proteins for existing, well-established properties, such as thermal stability, enzymatic activity, and affinity. However, in this study, we extended the model’s capabilities to design proteins that exhibit resilience to extreme environments, such as high pH—properties that are not inherently present in most natural proteins. To our knowledge, no existing model has addressed the challenge of engineering alkali-resistant proteins, nor is there relevant dataset available for training such models.

This shift from optimizing existing characteristics to engineering entirely new properties represents a significant step forward in the field of protein design. By focusing on the design of proteins that can survive and function in harsh, unnatural environments, we have demonstrated the broader applicability of the Pro-PRIME model beyond its initial scope. This expansion of the model's application is a novel contribution that has the potential to accelerate the development of proteins for industrial, agricultural, and biotechnological applications.

Thus, while the Pro-PRIME model itself is not new, its application to the new challenge of engineering proteins with alkali resistance and other novel properties significantly enhances the impact and novelty of this work. Moreover, this work is groundbreaking not only in terms of the model’s novel application but also because no previous studies have specifically targeted alkali resistance or provided data for training models on such extreme properties. Therefore, our approach is unique, marking a new direction in protein engineering.

We have made the following revisions to the conclusions section of the manuscript:

“Through two rounds of evolution, we successfully designed a VHH antibody with strong resistance to extreme environments and enhanced affinity using the Pro-PRIME model. Although rare case can tolerate the extreme pH and saline conditions in our pre-training dataset, the Pro-PRIME model showed impressive performance after supervised learning with limited data, especially on capturing the epistatic effects. The analysis of these 65 mutants revealed that the Pro-PRIME model is adept at exploring the large space of protein fitness, being less susceptible to local optima, and having greater potential to find the global optimum. Our efficient method of designing mutants that consider multiple properties improvement holds promise for industrial application of proteins. Specifically, the VHH antibody has been deployed in practical production and significantly enhancing the efficiency of the entire production line after our design. While the Pro-PRIME model itself has been reported, this work demonstrates its first-time application to the challenge of designing proteins with alkali resistance and other extreme properties that are not found in natural proteins, nor have previous studies addressed or provided data for such applications. This shift from optimizing existing protein properties to engineering entirely new, unnatural traits is a significant advance in the field. This study shows that the AI models, such as Pro-PRIME, can not only guide the evolution of protein thermal stability, enzymatic activity, ligand affinity, etc., but also enable to develop the mutants adapting the harsh unnatural environments, such as extreme pH and concentrated salt, largely expanding its application. The novelty of this work lies in the ability to design and engineer proteins with novel properties, specifically alkali resistance, which is an unprecedented achievement in AI-assisted protein engineering. The great potential of AI model is expected to significantly accelerate the development of proteins for diverse applications in medicine, agriculture, bioengineering, etc.”

(4) The computational experiments are not well-justified. For instance, the authors used a zero-shot setting for single-point mutation experiments but opted for fine-tuning in multiple-point mutation experiments. There is no clear explanation for this discrepancy. How does the model perform in zero-shot settings for multiple-point mutations? How would fine-tuning affect single-point mutation results? The choice of these strategies seems arbitrary and lacks sufficient discussion.

We appreciate the reviewer’s comment regarding the use of zero-shot and fine-tuning settings for single-point and multi-point mutation experiments, and we are grateful for the opportunity to further clarify this aspect of our work.

In the first round of design, we used the zero-shot approach for single-point mutations because the number of possible single-point mutations is limited, and no prior experimental data was available. In the absence of relevant data, the zero-shot approach allows the model to make predictions based on the learned sequence patterns from the pre-trained protein language model. Given that single-point mutations are relatively fewer in number and computationally feasible to evaluate, the zero-shot approach was deemed appropriate for this task.

However, when it comes to designing multi-point mutants, the number of potential combinations increases exponentially, making it computationally impractical to explore all possible mutations in a reasonable timeframe. Furthermore, since we had already obtained some experimental data for single-point mutations in the first round, we fine-tuned the model with this data in the second round to improve the accuracy of predictions for multi-point mutants. Fine-tuning helps the model better capture the specific features that contribute to protein functionality, which are critical when dealing with multi-point mutations where multiple residues interact. This allows the model to produce more reliable and targeted predictions for multi-point mutants, ultimately leading to better design outcomes.

Regarding the model's performance in zero-shot settings for multi-point mutations, we tested this approach, and the results did not align well with the experimental data for multi-point mutants. Specifically, the Spearman correlation coefficient between the zero-shot predictions and experimental results was -0.71, indicating that zero-shot predictions for multi-point mutations were not as accurate as those from the fine-tuned model.

In summary, the choice of using zero-shot for single-point mutations and fine-tuning for multi-point mutations was driven by the nature of the task and the availability of experimental data. Fine-tuning the model improves its predictive performance, especially for more complex multi-point mutation tasks. We have now clarified these choices in the manuscript and have added further discussion on the trade-offs between zero-shot and fine-tuning approaches.

Specifically, we have added the following discussion at the section 2.2 in the main text:

“Note that we employed different strategies for designing single-point and multi-point mutants, specifically using a zero-shot approach for single-point mutations and fine-tuning the model for multi-point mutations. These choices were made based on the distinct characteristics of the two tasks and the availability of experimental data. For single-point mutations, the number of possible mutations is relatively limited, and at the outset, there were no experimental data available. In such cases, the zero-shot setting was chosen because it allows the model to predict the fitness of mutants based solely on the information learned during pre-training on a large protein sequence dataset. Since single-point mutations are computationally manageable, this approach was deemed appropriate to generate initial predictions for protein engineering. However, when designing multi-point mutants, the situation changes significantly. The potential combinations of mutations increase exponentially, and without prior data, it becomes computationally infeasible to evaluate every possible combination within a reasonable timeframe. Moreover, by the time we reached the multi-point mutation design stage, experimental data for several single-point mutations had already been obtained. This data enabled us to fine-tune the model to better capture the specific structural and functional features that contribute to protein stability and resistance, especially in the context of multiple interacting mutations. Fine-tuning improves the model’s accuracy by adjusting its parameters to align more closely with the experimental data, ensuring that the predicted multi-point mutants are more likely to meet the desired engineering goals. After the second round of design, the fitness of the mutants was further improved. In improving alkali resistance, experimental results showed that 15 of the 45 designed mutants exhibited positive responses, yielding a success rate of 30%, close to the 35% success rate achieved in the second round. Compared to the wild type, the best single-point mutant improved alkali resistance by approximately 44.7%, while the best multi-point mutant achieved a 67.7% increase. For thermal stability enhancement, the success rate in the first round was 77.8%, rising to 100% in the second round. The top single-point mutant exhibited a Tm increase of 6.37°C over the wild type, while the best multi-point mutant had a Tm increase of 10.02°C. We also tested the performance of the zero-shot approach for multi-point mutants, and the results showed that this method did not yield satisfactory predictions. The Spearman correlation coefficient between the zero-shot predictions and experimental results for multi-point mutants was -0.71, indicating a significant discrepancy. This further highlights the importance of fine-tuning the model for multi-point mutations, as the fine-tuned model provided more accurate and reliable results. In summary, the choice of zero-shot for single-point mutations and fine-tuning for multi-point mutations was driven by practical considerations regarding computational feasibility and the availability of experimental data. Fine-tuning the model significantly enhances its predictive performance, particularly for complex multi-point mutations where multiple residues interact. We believe this strategy strikes an optimal balance between computational efficiency and predictive accuracy, making it well-suited for practical protein engineering applications.”

eLife Assessment

The preliminary data presented in this manuscript seem valuable. The data are currently incomplete and there are numerous technical concerns, some of which may arise from insufficient description of methodologies used.

Reviewer #1 (Public review):

Summary:

This study investigates the impact of Pink1 loss on glial function and neuronal health in a Drosophila model, highlighting the role of mitochondria-organelle contacts and key genes such as Ccz1, Vps13, Mon1, and Rab7. The work provides insights into cellular processes underlying neurodegenerative diseases, with a focus on glia-neuron interactions. While the findings are promising, the study lacks critical controls, detailed mechanistic evidence, and explanatory figures to strengthen its claims.

Strengths:

(1) The study addresses an important topic in neuroscience, exploring the mechanisms of Pink1 loss, which has implications for Parkinson's disease and neurodegeneration.

(2) The focus on mitochondria-organelle contacts and their regulation by Rab7-mediated pathways is novel and provides a potential mechanism for neuronal dysfunction.

(3) The identification of key genes (Ccz1, Vps13, Mon1, Rab7) and their potential roles in Pink1-related pathways adds valuable knowledge to the field.

(4) The manuscript uses a combination of genetic tools, Drosophila models, and functional assays to approach the problem from multiple angles.

Weaknesses:

(1) Specificity of Mz-Gal4: The study lacks validation of Mz-Gal4 specificity, as it may also drive expression in a few neurons or other types of glia. Additional control experiments using nls-GFP with Elav, Repo, or Draper antibody staining or alternative glial drivers would be helpful.

(2) DLG staining is central to the story but is not well-supported by high-resolution Z-stack imaging, which should be included in the supplementary figures.

(3) The manuscript does not confirm whether the candidate RNAi (Ccz1, Vps13, Mon1, Rab7) directly influence Rab7-mediated membrane trafficking or mitochondria-lysosome contacts in Pink1 mutants.

(4) Using ERG as a readout for EG effects in the antenna is not a direct or appropriate assay. Alternative functional assays relevant to antenna glia should be considered.

(5) A graphical explanation of the interactions and functions of the candidate genes in Pink1 KO mutants is missing. This would greatly enhance the manuscript's clarity.

(6) The study lacks details on sample sizes, effect sizes, and reproducibility, which are necessary for robust conclusions.

(7) There are repeated words on page 3 ("olfactory Olfactory Receptor Neurons") and a lack of explanation in Figure 3C regarding the most up-regulated and down-regulated genes and the significance of large red dots.

Reviewer #2 (Public review):

Summary: This study proposes a novel role for ensheathing glia (EG) in a Pink1-model of Parkinson's disease and shows that this cell population exibits the highest number of DEG in a pre-symptomatic stage. In the olfactory system, there seems to be morphological changes in this cell-type that resembles an 'activated' state and the authors further show that the neuronal loss of Pink1 is responsible for this defect. The authors go on to show that manipulation of Pink1 in EG also leads to some defects in the visual system and in the dopaminergic neurons (DAN) that innervate the mushroom body (MB), and performed a screen based on the 'on-transient' defect of the ERG to identify potential genes that may modulate the function of EG in synaptic regulation. They focus on several genes related to Rab7/Vps13, and performed some additional experiments in the visual system and MB to propose the role of vesicle/lipid trafficking in EG as a important factor for PD pathogenesis.

Strengths: The study proposes functional and mechanistic connections between several genes that have been linked to PD (PINK1, VPS13A/C). I feel that the data presented in Figure 1 and Fig3A-C are performed with rigor and are convincing/novel. The selection of Drosophila to study the questions is also a strength and the lab has extensive experiences in this field and model organism.

Weaknesses: There is one fundamental concern I have with the genetic experiments performed in this paper (especially in Fig 3D and Fig4, see major issue #1), and I feel that there is a bit of a disconnect between the EG 'activation' phenotype the author show in the olfactory system and the other two neuronal systems (visual system, MB DAN) that the authors investigate see major issue #2). Also, there are quite a bit of information that is not provided in the manuscript (see major issues #3 and #4), which makes me difficult to judge the rigor and interpretation of several experiments.

Major Concern #1: A number of lines used in this study are referred to as "RNAi" lines but when I look at the actual genotypes of reagents listed in the table in the METHODS section, many are actually NOT RNAi lines. Quite a few lines, including lines that the authors use as RNAi against Ccz1, Rab7 and Mon1, are gRNA lines for the TKO (TRiP-CRISPR knockout) system. While these reagents can theoretically knock-out these genes in somatic cells if used in combination with UAS-Cas9, there is no mention that UAS-Cas9 was used in this work throughout the manuscript. Hence, when these lines are just crossed to GAL4 with or without the Pink1 mutant, they shouldn't be having any effects. Similarly, the strongest hit from their screen was a TOE (TRiP-CRISPR Over Expression) gRNA against PIG-A, which could allow overexpression of PIG-A if there is a UAS-dCas9::VP64. However, I also do not see any mention that such activator was introduced into the crossing scheme. Considering that 3 of the 4 'hits' from their screen are not RNAi lines, I am quite skeptical of the study. Similarly, except for Vps13, all reagents used in Fig4 are TKO gRNA lines. Therefore, if this experiment was conducted without an UAS-Cas9, most of the data shown here are problematic. Also, note that several of the 'RNAi' lines listed in the Table in the METHODS section are actually MiMIC alleles. While some MiMIC lines could function as strong LOF alleles (if they are inserted in the exon or in an intron of the gene in the same orientation as the gene), some of the lines are not expected to affect gene function (e.g. FASN2 and CG17712, MiMICs are in introns and face the opposite orientation). Hence, the rationale of including these reagents in the screen doesn't make much sense. The description of the modifier screen should be much more detailed in the RESULTS and METHODS section and if the UAS-Cas9/dCas9::VP64 transgenes were not introduced when the TKO/TOE reagents were utilized, what can be concluded?

In addition, for the 4 genes that the authors further study in Fig4, there are many other reagents that the authors can use, including mutant alleles, previously characterized RNAi lines (e.g. Vps13) and dominant negative/constitute active lines (e.g. especially for Rab7). The authors should validate their results with independent reagents to really convincingly show that the same conclusions can be drawn for the Vps13/Rab7 related genes since this is the key takeaway message of this paper.

Also, they do not show whether the manipulation of these genes in a wild-type background (they only show what happens in Pink1 mutants) affect ERG and MB DAN synapse morphology. If these manipulations alone dramatically affect these phenotypes, it would be very difficult to interpret their data.

Major Concern #2: In Figure 1, the authors show some morphological evidence that EG are 'activated' in Pink1 mutants, but whether the same phenomenon occurs in the visual system and in the MB is not shown. Since all of the studies in Fig3D and Fig4 are done in the visual system and MB, it is not clear whether the visual system and MB phenotypes are related to 'activation' of EG.

Also, in the RNA-seq data in Fig1A and Fig3C, is there any molecular evidence that EG are indeed 'activated'? The only evidence that the authors show to state that EG are 'activated' in young Pink1 null animals is based on increased CD8::GFP staining in the olfactory system.

The authors cannot draw a strong conclusion that indeed EG are 'activated' based on these data (e.g. perhaps the expression level of CD8::GFP is just increased). Additional evidence that the EG are 'activated' could be provided by looking at the increase in Draper intensity (as reported by Doherty et al. and MacDonald et al. that the authors cite), not only in the olfactory system, but also in the visual system and in the MB. It would also be informative if the authors can look at morphology of the EG in the visual system and MB to convincingly that the data shown in Fig4 is relevant to EG 'activation'.

Major Concern #3: In Fig3, there is no clear explanation why they focus on the ON transients and ignore the OFF transients, and also why the difference in the depolarization is not quantified in Fig4.

Major Concern #4: While the authors claim that mz709-GAL4 is a EG specific driver, do the authors know that this is indeed true in the tissues and stages that are studied here? The Ito et al,. paper that is cited in the METHOD section has only looked at the expression of this reporter in embryonic and larval stages. The authors need to that the authors should validate their findings with an additional EG specific driver and/or provide additional data that mz709-GAL4 is indeed specific to EG in the adult fly brain and eye. If mz709-GAL4 is expressed in other cell-types, the interpretation of many of the data in this paper becomes quite questionable. I believe the data in Fig3B is suggesting that mz709-GAL4 is indeed specific to glia cells and not expressed in neurons, but whether this driver is truly specific to EG (and not in other glial types), especially in the visual system (including the lamina as well as in the eye), is not obvious.

Author response:

We would like to thank the reviewers and the editors for carefully reading and commenting our manuscript and plan to prepare a revised manuscript. Particularly, we want to thank reviewer 2 for spotting a major oversight regarding the use of the TKO (TRiP-CRISPR knockout) and TOE (TRiP-CRISPR Over Expression) systems and the MiMIC alleles. As the reviewer pointed out, these lines were not used as intended, therefore our results and conclusions regarding the genetic interactions between Pink1 and several of genes in the paper (PIG-A, Rab7, Ccz1, CG10646, Mon1, FASN2, CG17712) that we attempted to target, are incorrect and based on a technical mistake. These results need to be removed from the manuscript.

eLife Assessment

This study reports an approach for restoring sperm motility in mice. The strength lies in in the novelty of the methodology being developed, but the evidence for the success of the method or its mechanism is inadequate. Additional experimental support would be required to support the conclusions of the authors.

Reviewer #1 (Public review):

The authors assess the effectiveness of electroporating mRNA into male germ cells to rescue the expression of proteins required for spermatogenesis progression in individuals where these proteins are mutated or depleted. To set up the methodology, they first evaluated the expression of reporter proteins in wild-type mice, which showed expression in germ cells for over two weeks. Then, they attempted to recover fertility in a model of late spermatogenesis arrest that produces immotile sperm. By electroporating the mutated protein, the authors recovered the motility of ~5% of the sperm; although the sperm regenerated was not able to produce offspring using IVF, the embryos reached the 2-cell state (in contrast to controls that did not progress past the zygote state).

This is a comprehensive evaluation of the mRNA methodology with multiple strengths. First, the authors show that naked synthetic RNA, purchased from a commercial source or generated in the laboratory with simple methods, is enough to express exogenous proteins in testicular germ cells. The authors compared RNA to DNA electroporation and found that germ cells are efficiently electroporated with RNA, but not DNA. The differences between these constructs were evaluated using in vivo imaging to track the reporter signal in individual animals through time. To understand how the reporter proteins affect the results of the experiments, the authors used different reporters: two fluorescent (eGFP and mCherry) and one bioluminescent (Luciferase). Although they observed differences among reporters, in every case expression lasted for at least two weeks.

The authors used a relevant system to study the therapeutic potential of RNA electroporation. The ARMC2-deficient animals have impaired sperm motility phenotype that affects only the later stages of spermatogenesis. The authors showed that sperm motility was recovered to ~5%, which is remarkable due to the small fraction of germ cells electroporated with RNA with the current protocol. The sperm motility parameters were thoroughly assessed by CASA. The 3D reconstruction of an electroporated testis using state-of-the-art methods to show the electroporated regions is compelling.

The main weakness of the manuscript is that although the authors manage to recover motility in a small fraction of the sperm population, it is unclear whether the increased sperm quality is substantial to improve assisted reproduction outcomes. The authors found that the rescued sperm could be used to obtain 2-cell embryos via IVF, but no evidence for more advanced stages of embryo differentiation was provided. The motile rescued sperm was also successfully used to generate blastocyst by ICSI, but the statistical significance of the rate of blastocyst production compared to non-rescued sperm remains unclear. The title is thus an overstatement since fertility was never restored for IVF, and the mutant sperm was already able to produce blastocysts without the electroporation intervention.

Overall, the authors clearly show that electroporating mRNA can improve spermatogenesis as demonstrated by the generation of motile sperm in the ARMC2 KO mouse model.

Reviewer #2 (Public review):

The authors inject, into the rete testes, mRNA and plasmids encoding mRNAs for GFP and then ARMC2 (into infertile Armc2 KO mice) in a gene therapy approach to express exogenous proteins in male germ cells. They do show GFP epifluorescence and ARMC2 protein in KO tissues, although the evidence presented is weak. Overall, the data do not necessarily make sense given the biology of spermatogenesis and more rigorous testing of this model is required to fully support the conclusions, that gene therapy can be used to rescue male infertility.

In this revision, the authors attempt to respond to the critiques from the first round of reviews. While they did address many of the minor concerns, there are still a number to be addressed. With that said, the data still do not support the conclusions of the manuscript.

(1) The authors have not satisfactorily provided an explanation for how a naked mRNA can persist and direct expression of GFP or luciferase for ~3 weeks. The most stable mRNAs in mammalian cells have half-lives of ~24-60 hours. The stability of the injected mRNAs should be evaluated and reported using cell lines. GFP protein's half-life is ~26 hours, and luciferase protein's half-life is ~2 hours.

(2) There is no convincing data shown in Figs. 1-8 that the GFP is even expressed in germ cells, which is obviously a prerequisite for the Armc2 KO rescue experiment shown in the later figures! In fact, to this reviewer the GFP appears to be in Sertoli cell cytoplasm, which spans the epithelium and surrounds germ cells - thus, it can be oft-confused with germ cells. In addition, if it is in germ cells, then the authors should be able to show, on subsequent days, that it is present in clones of germ cells that are maturing. Due to intracellular bridges, a molecule like GFP has been shown to diffuse readily and rapidly (in a matter of minutes) between adjacent germ cells. To clarify, the authors must generate single cell suspensions and immunostain for GFP using any of a number of excellent commercially-available antibodies to verify it is present in germ cells. It should also be present in sperm, if it is indeed in the germline.

Other comments:

70-1 This is an incorrect interpretation of the findings from Ref 5 - that review stated there were ~2,000 testis-enriched genes, but that does not mean "the whole process involves around two thousand of genes"

74 would specify 'male'

79-84 Are the concerns with ICSI due to the procedure itself, or the fact that it's often used when there is likely to be a genetic issue with the male whose sperm was used? This should be clarified if possible using references from the literature, as this reviewer imagines this could be a rather contentious issue with clinicians who routinely use this procedure, even in cases where IVF would very likely have worked

199 Codon optimization improvement of mRNA stability needs a reference; in one study using yeast transcripts, optimization improved RNA stability on the order of minutes (e.g., from ~5 minutes to ~17 minutes); is there some evidence that it could be increased dramatically to days or weeks?

472-3 The reported half-life of EGFP is ~36 hours - so, if the mRNA is unstable (and not measured, but certainly could be estimated by qRT-PCR detection of the transcript on subsequent days after injection) and EGFP is comparatively more stable (but still hours), how does EGFP persist for 21 days after injection of naked mRNA??

Curious why the authors were unable to get anti-GFP to work in immunostaining?

In Fig. 3-4, the GFP signals are unremarkable, in that they cannot be fairly attributed to any structure or cell type - they just look like blobs; and why, in Fig. 4D-E, why does the GFP signal appear stronger at 21 days than 15 days? And why is it completely gone by 28 days? This data is unconvincing. If the authors did a single cell suspension, what types or percentage of cells would be GFP+? Since germ cells are not adherent in culture, a simple experiment could be done whereby a single cell suspension could be made, cultured for 4-6 hours, and non-adherent cells "shaken off" and imaged vs adherent cells. Cells could also be fixed and immunostained for GFP, which has worked in many other labs using anti-GFP.

In Fig. 5, what is the half-life of luciferase? From this reviewer's search of the literature, it appears to be ~2-3 h in mammalian cells. With this said, how do the authors envision detectable protein for up to 20 days from a naked mRNA? The stability of the injected mRNAs should be shown in a mammalian cell line - perhaps this mRNA has an incredibly long half-life, which might help explain these results. However, even the most stable endogenous mRNAs (e.g., globin) are ~24-60 hrs.

527-8 The Sertoli cell cytoplasm is not just present along the basement membrane as stated, but also projects all the way to the lumina

529-30 This is incorrect, as round spermatids are never "localized between the spermatocytes and elongated spermatids" - if elongated spermatids are present, rounds are not - they are never coincident in the same testis section

Fig. 7 To this reviewer, all of the GFP appears to be in Sertoli cell cytoplasm

In Figs 1-8 there is no convincing evidence presented that GFP is expressed in germ cells! In fact, it appears to be in Sertoli cells

Fig. 9 - alpha-tubuline?

Fig. 11 - how was sperm morphology/motility not rescued on "days 3, 6, 10, 15, or 28 after surgery", but it was in some at 21 and 35? How does this make sense, given the known kinetics of male germ cell development?? And at least one of the sperm in the KO in Fig. B5 looks relatively normal, and the flagellum may be out-of-focus in the image? With only a few sperm for reviewers to see, how can we know these represent the population?

Reviewer #3 (Public review):

Summary:

The authors used a novel technique to treat male infertility. In a proof-of-concept study, the authors were able to rescue the phenotype of a knockout mouse model with immotile sperm using this technique. This could also be a promising treatment option for infertile men.

Strengths:

In their proof-of-concept study, the authors were able to show that the novel technique rescues the infertility phenotype of Armc2 knockout spermatozoa. In the revised version of the manuscript, the authors have added data on in vitro fertilisation experiments with Armc2 mRNA-rescued sperm. The authors show that Armc2 mRNA-rescued sperm can successfully fertilise oocytes that develop to the blastocyst stage. This adds another level of reliability to the data.

Weaknesses:

Some minor weaknesses identified in my previous report have already been fixed. The technique is new and may not yet be fully established for all issues. Nevertheless, the data presented in this manuscript opens the way for several approaches to immotile spermatozoa to ensure successful fertilisation of oocytes and subsequent appropriate embryo development.

[Editors' note: The images in Figure 12 do not support the authors' interpretation that 2-cell embryos resulted from in vitro fertilization. Instead, the cells shown appear to be fragmented, unfertilized eggs. Combined with the lack of further development, it seems highly unlikely that fertilization was successful.]

Author response:

The following is the authors’ response to the original reviews.

Public Reviews: 

Reviewer #1 (Public Review): 

The authors assess the effectiveness of electroporating mRNA into male germ cells to rescue the expression of proteins required for spermatogenesis progression in individuals where these proteins are mutated or depleted. To set up the methodology, they first evaluated the expression of reporter proteins in wild-type mice, which showed expression in germ cells for over two weeks. Then, they attempted to recover fertility in a model of late spermatogenesis arrest that produces immotile sperm. By electroporating the mutated protein, the authors recovered the motility of ~5% of the sperm, although the sperm regenerated was not able to produce offspring using IVF.

We actually did not write that “sperm regenerated was not able to produce offspring using IVF” but rather that IVF was not attempted because the number of rescued sperm was too low. To address this important point, the ability of sperm to produce embryos was therefore challenged by two different assisted reproduction technologies, that are IVF and ICSI. To increase the number of motile sperm for IVF experiments, we have injected both testes from one male. We also conducted intracytoplasmic sperm injection (ICSI) experiments, using only rescued sperm, identified as motile sperm with a normal flagellum. The results of these new experiments have demonstrated that the rescued ARMC2 sperm successfully fertilized eggs and produced embryos at the two-cell stage by IVF and blastocysts by ICSI. These outcomes are presented in Figure 12.

This is a comprehensive evaluation of the mRNA methodology with multiple strengths. First, the authors show that naked synthetic RNA, purchased from a commercial source or generated in the laboratory with simple methods, is enough to express exogenous proteins in testicular germ cells. The authors compared RNA to DNA electroporation and found that germ cells are efficiently electroporated with RNA, but not DNA. The differences between these constructs were evaluated using in vivo imaging to track the reporter signal in individual animals through time. To understand how the reporter proteins affect the results of the experiments, the authors used different reporters: two fluorescent (eGFP and mCherry) and one bioluminescent (Luciferase). Although they observed differences among reporters, in every case expression lasted for at least two weeks. 

The authors used a relevant system to study the therapeutic potential of RNA electroporation. The ARMC2-deficient animals have impaired sperm motility phenotype that affects only the later stages of spermatogenesis. The authors showed that sperm motility was recovered to ~5%, which is remarkable due to the small fraction of germ cells electroporated with RNA with the current protocol. The 3D reconstruction of an electroporated testis using state-of-the-art methods to show the electroporated regions is compelling. 

The main weakness of the manuscript is that although the authors manage to recover motility in a small fraction of the sperm population, it is unclear whether the increased sperm quality is substantial to improve assisted reproduction outcomes. The quality of the sperm was not systematically evaluated in the manuscript, with the endpoints being sperm morphology and sperm mobility. 

We would like to thank the reviewers for their comments. As previously stated above, we produced additional rescue experiments and performed CASA, morphology observation, IVF and ICSI with the rescued sperm. The rescued ARMC2 sperm exhibited normal morphology (new figure 11 and Supp Fig 8), motility (figure 11), and fecundity (figure 12).  Whereas sperm from untreated KO males were unable to fertilize egg by IVF, the rescued sperm fertilized eggs in vitro at a significant level (mean 62%, n=5), demonstrating that our strategy improves the sperm quality and assisted reproduction outcome (from 0 to 62%). 

Some key results, such as the 3D reconstruction of the testis and the recovery of sperm motility, are qualitative given the low replicate numbers or the small magnitude of the effects. The presentation of the sperm motility data could have been clearer as well. For example, on day 21 after Armc2-mRNA electroporation, only one animal out of the three tested showed increased sperm motility. However, it is unclear from Figure 11A what the percentage of sperm motility for this animal is since the graph shows a value of >5% and the reported aggregate motility is 4.5%. It would have been helpful to show all individual data points in Figure 11A. 

We provide now in figure 11A, a graph showing the percentage of rescued sperm for all animals. (scatter dot plot). Moreover, we performed additional CASA experiments to analyze in detail sperm motility (Figure 11A2-A3). Individual CASA parameters for motile sperm cells were extracted as requested by reviewer 3 and represented in a new graph (Fig 11 A2). 

The expression of the reporter genes is unambiguous; however, better figures could have been presented to show cell type specificity. The DAPI staining is diffused, and it is challenging to understand where the basement membranes of the tubules are. For example, in Figures 7B3 and 7E3, the spermatogonia seems to be in the middle of the seminiferous tubule. The imaging was better for Figure 8. Suboptimal staining appears to lead to mislabeling of some germ cell populations. For example, in Supplementary Figure 4A3, the round spermatid label appears to be labeling spermatocytes. Also, in some instances, the authors seem to be confusing, elongating spermatids with spermatozoa, such as in the case of Supplementary Figures 4D3 and D4.

Thanks for the comments, some spermatogenic cells were indeed mislabeled as you mentioned. We have therefore readjusted the labeling accordingly. We also changed spermatozoa to mature spermatids. The new sentence is now: “At the cellular level, fluorescence was detectable in germ cells (B1-B3) including Spermatogonia (Sg), Spermatocytes (Scytes),round Spermatids (RStids), mature spermatids (m-Sptids) and Sertoli cells (SC)”. Moreover, to indicate the localization of the basal membrane, we have also labelled myoid cells.

The characterization of Armc2 expression could have been improved as well. The authors show a convincing expression of ARMC2 in a few spermatids/sperm using a combination of an anti-ARMC2 antibody and tubules derived from ARMC2 KO animals. At the minimum, one would have liked to see at least one whole tubule of a relevant stage.  

Thanks for the remark. 

We present now new images showing transversal section of seminiferous tubules as requested (see supp fig 6). In this new figure, it is clear that Armc2 is only expressed in spermatids. We have also added in this figure an analysis of the RNA-seq database produced by Gan's team (Gan, Wen et al. 2013), confirming that ArmC2 expression is predominantly expressed at the elongated spermatid stage. This point is now clearly indicated in the text.

Overall, the authors show that electroporating mRNA can improve spermatogenesis as demonstrated by the generation of motile sperm in the ARMC2 KO mouse model. 

Thank you

Reviewer #2 (Public Review): 

Summary: 

Here, the authors inject naked mRNAs and plasmids into the rete testes of mice to express exogenous proteins - GFP and later ARMC2. This approach has been taken before, as noted in the Discussion to rescue Dmc1 KO infertility. While the concept is exciting, multiple concerns reduce reviewer enthusiasm. 

Strengths: 

The approach, while not necessarily novel, is timely and interesting.  Weaknesses: 

Overall, the writing and text can be improved and standardized - as an example, in some places in vivo is italicized, in others it's not; gene names are italicized in some places, others not; some places have spaces between a number and the units, others not. This lack of attention to detail in the preparation of the manuscript is a significant concern to this reviewer - the presentation of the experimental details does cast some reasonable concern with how the experiments might have been done. While this may be unfair, it is all the reviewers have to judge. Multiple typographical and grammatical errors are present, and vague or misleading statements. 

Thanks for the comment, we have revised the whole manuscript to remove all the mistakes. We have also added new experiments/figures to strengthen the message. Finally, we have substantially modified the discussion.

Reviewer #3 (Public Review):

Summary: 

The authors used a novel technique to treat male infertility. In a proof-of-concept study, the authors were able to rescue the phenotype of a knockout mouse model with immotile sperm using this technique. This could also be a promising treatment option for infertile men. 

Strengths: 

In their proof-of-concept study, the authors were able to show that the novel technique rescues the infertility phenotype in vivo. 

Weaknesses: 

Some minor weaknesses, especially in the discussion section, could be addressed to further improve the quality of the manuscript. 

We have substantially modified the discussion, following the remarks of the reviewers.

It is very convincing that the phenotype of Armc2 KO mice could (at least in part) be rescued by injection of Armc2 RNA. However, a central question remains about which testicular cell types have been targeted by the constructs. From the pictures presented in Figures 7 and 8, this issue is hard to assess. Given the more punctate staining of the DNA construct a targeting of Sertoli cells is more likely, whereas the more broader staining of seminiferous tubules using RNA constructs is talking toward germ cells. Further, the staining for up to 119 days (Figure 5) would point toward an integration of the DNA construct into the genome of early germ cells such as spermatogonia and/or possibly to Sertoli cells. 

Thanks for the comment. We would like to recall the peculiar properties of the non-insertional Enhanced Episomes Vector (EEV) plasmid, which is a non-viral episome based on the Epstein-Barr virus (EBV: Epstein-Barr Virus). It allows the persistence of the plasmid for long period of time without integration. Its maintenance within the cell is made possible by its ability to replicate in a synchronous manner with the host genome and to segregate into daughter cells. This is due to the fact that EEV is composed of two distinct elements derived from EBV: an origin of replication (oriP) and an EpsteinBarr Nuclear Antigen 1 (EBNA1) expression cassette (Gil, Gallaher, and Berk, 2010).   The oriP is a locus comprising two EBNA1-binding domains, designated as the Family of Repeats (FR) and Dyad Symmetry (DS). The FR is an array of approximately 20 EBNA1-binding sites (20 repeats of 30 bp) with high affinity, while the DS comprises four lower-affinity sites operating in tandem (Ehrhardt et al., 2008). 

The 641-amino-acid EBNA1 protein contains numerous domains. The N-terminal domains are rich in glycines and alanines, which enable interaction with host chromosomes. The C-terminal region is responsible for binding to oriP (Hodin, Najrana, and Yates, 2013). The binding of EBNA1 to the DS element results in the recruitment of the origin of replication. This results in the synchronous initiation of extra-chromosomal EEV replication with host DNA at each S phase of the cell cycle (Düzgüneş, Cheung, and Konopka 2018). Furthermore, EBNA1 binding to the FR domain induces the formation of a bridge between metaphase chromosomes and the vector during mitosis. This binding is responsible for the segregation of the EEV episome in daughter cells (Düzgüneş, Cheung, and Konopka 2018). It is notable that EEV is maintained at a rate of 90-95% per cell division.

Because of the intrinsic properties of EEV described above, the presence of the reporter protein at 119 day after injection was likely due to the maintenance of the plasmid, mostly in Sertoli cells, and not to the DNA integration of the plasmid.

Of note, the specificity of EEV was already indicated in the introduction (lines 124-128 clean copy). Nevertheless, we have added more information about EEV to help the readers.  

Given the expression after RNA transfection for up to 21 days (Figure 4) and the detection of motile sperm after 21 days (Figure 11), this would point to either round spermatids or spermatocytes.  These aspects need to be discussed more carefully (discussion section: lines 549-574).

We added a sentence to highlight that spermatids are transfected and protein synthetized at this stage and this question is discussed in details (see lines 677-684 clean copy).

It would also be very interesting to know in which testicular cell type Armc2 is endogenously expressed (lines 575-591)

Thanks for the remarks. We present now new images showing the full seminiferous tubules as requested by reviewer 1 (see supp fig 6). In this new figure, it is clear that Armc2 is only expressed in spermatids. We have also added in this figure an analysis of the RNA-seq database produced by Gan's team (Gan, Wen et al. 2013), confirming that Armc2 is predominantly expressed at the elongated spermatid stage. This point is now clearly indicated in the text. (lines 570-579 clean copy).

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors): 

The article is well-structured and easy to read. Nonetheless, there are typos and mistakes in some places that are distracting to the reader, such as the capitalization of the word "Oligo-" in the title of the manuscript, the use of the word "Materiel" in the title of the Materials and methods and the presence of space holders "Schorr staining was obtained from Merck (XXX)".  Thank you, we corrected the misspelling of "Materials and Methods" and corrected our error: "obtained from Merck (Darmstadt, Germany)". We also carefully corrected the manuscript to remove typos and mistakes.

The discussion is too lengthy, with much repetition regarding the methods used and the results obtained. For example, these are two sentences from the discussion. "The vector was injected via the rete testis into the adult Armc2 KO mice. The testes were then electroporated." I would recommend shortening these passages.

Thanks for your comments, we removed the sentences and we have substantially modified the discussion, following the remarks of the reviewers.

The work is extensive, and many experiments have been done to prove the points made. However, a more in-depth analysis of critical experiments would have benefited the manuscript significantly. A more thorough analysis of sperm mobility and morphology using the CASA system would have been an initial step.

In response to the observations made, additional CASA experiments and sperm motility analysis were conducted, as illustrated in Figure 11 (A2-A3). Individual CASA parameters for motile sperm cells were extracted as suggested and represented in a new graph (Fig 11 A2). We have observed significant differences between WT and rescued sperm. In particular, the VSL and LIN parameters were lower for rescued sperm. Nevertheless, these differences were not sufficient to prevent IVF, maybe because the curvilinear velocity (VCL) was not modified.

In the case of ARMC2 localization, an analysis of the different stages of spermatogenesis to show when ARMC2 starts to be expressed. 

Thanks for the remarks. This is an important remark pointed out by all reviewers. As explained above, we have performed more experiments. We present now new images showing transversal section of seminiferous tubules as requested (see supp fig 6). In this new figure, it is clear that Armc2 is only expressed in spermatid layers. We have also added in this figure an analysis of the RNA-seq database produced by Gan's team (Gan, Wen et al. 2013), confirming that ArmC2 expression is predominantly expressed at the elongated spermatid stage. This point is now clearly indicated in the text. (lines 575579 clean copy).

Finally, exploring additional endpoints to understand the quality of the sperm generated, such as the efficiency of ICSI or sperm damage, could have helped understand the degree of the recovery.

This point was underlined in public review. We paste here our answer: “To address this important point, the ability of sperm to produce embryos was therefore challenged by two different assisted reproduction technologies, that are IVF and ICSI. To increase the number of motile sperm for IVF experiments, we have injected both testes from one male. We also conducted intracytoplasmic sperm injection (ICSI) experiments, using only rescued sperm, identified as motile sperm with a normal flagellum. The results of these new experiments have demonstrated that the rescued ARMC2 sperm successfully fertilized eggs and produced embryos at the two-cell stage by IVF and blastocysts by ICSI. These outcomes are presented in Figure 12.”

Reviewer #2 (Recommendations For The Authors):

38,74 intracellular

Thanks, we changed it accordingly: "Intracytoplasmic sperm injection (ICSI) is required to treat such a condition, but it has limited efficacy and has been associated with a small increase in birth defects" and "such as intracytoplasmic sperm injection (ICSI)".

39 "limited efficacy" Versus what? And for what reason? "small increase in birth defects" - compared to what? 

We changed to “… but it is associated with a small increase in birth defect with comparison to pregnancies not involving assisted conception.”

40 Just thinking through the logic of the argument thus far - the authors lay out that there are people with OAT (true), ICSI must be used (true), ICSI is bad (not convincing), and therefore a new strategy is needed... so is this an alternative to ICSI? And this is to restore fertility, not "restore spermatogenesis"

- because ICSI doesn't restore spermatogenesis. This logic flow needs to be cleaned up some

Thanks we changed it accordingly: “restore fertility.”

45 "mostly"?

Thank you, we removed the word: “We show that mRNA-coded reporter proteins are detected for up to 3 weeks in germ cells, making the use of mRNA possible to treat infertility.”

65 Reference missing. 

We added the following reference Kumar, N. and A. K. Singh (2015). "Trends of male factor infertility, an important cause of infertility: A review of literature." J Hum Reprod Sci 8(4): 191-196.

68 Would argue meiosis is not a reduction of the number of chromosomes - that happens at the ends of meiosis I and II - but the bulk of meiosis is doubling DNA and recombination; would re-word; replace "differentiation" with morphogenesis, which is much more commonly used:

Thank you, we have changed the sentence accordingly: "proliferation (mitosis of spermatogonia), reduction of the number of chromosomes (meiosis of spermatocytes), and morphogenesis of sperm (spermiogenesis)".

70 "almost exclusively" is an odd term, and a bit of an oxymoron - if not exclusively, then where else are they expressed? Can you provide some sense of scale rather than using vague words like "large", "almost", "several", "strongly" and "most...likely" - need some support for these claims by being more specific: 

Thanks for the comment, we changed the sentence: "The whole process involves around two thousand genes, 60% of which are expressed exclusively in the testes."

73 "severe infertility" is redundant - if they are infertile, is there really any more or less about it? I think what is meant is patients with immotile sperm can be helped by ICSI - so just be more specific... 

We changed the transition : “Among infertility disorders, oligo-astheno-teratozoospermia  (OAT) is the most frequent (50 % (Thonneau, Marchand et al. 1991); it is likely to be of genetic origin. Spermatocytograms of OAT patients show a decrease in sperm concentration, multiple morphological defects and defective motility. Because of these combined defects, patients are infertile and can only conceive by IntraCytoplasmic Sperm Injection (ICSI). IntraCytoplasmic Sperm Injection (ICSI) can efficiently overcome the problems faced. However, there are …”

75 "some" is vague - how many concerns, and who has them? Be specific!

Thanks for the comment, we removed the word.

76-7 Again, be specific - "real" has little meaning - what is the increased risk, in % or fold? This is likely a controversial point, so make sure you absolutely support your contention with data .

77 "these"? There was only one concern listed - increased birth defects; and "a number" is vague - what number, 1 or 1,000,000? A few (2-3), dozens, hundreds? 

Thanks for the comment, we have reworded the sentence: “Nevertheless, concerns persist regarding the potential risks associated with this technique, including blastogenesis defect, cardiovascular defect, gastrointestinal defect, musculoskeletal defect, orofacial defect, leukemia, central nervous system tumors, and solid tumors. Statistical analyses of birth records have demonstrated an elevated risk of birth defects, with a 30–40% increased likelihood in cases involving ICSI, and a prevalence of birth defects between 1% and 4%.” We have added a list of references to support these claims.

79-81 So, basically transgenesis? Again, vague terms "widely" - I don't think it's all that widely used yet... and references are missing to support the statement that integration of DNA into patient genomes is widely used. Give specific numbers, and provide a reference to support the contention. 

Thanks for the comment, we removed the word widely and add references.

81-5 Just finished talking about humans, but now it appears the authors have switched to talking about mice - got to let the readers know that! Unless you're talking about the Chinese group that deleted CCR5 in making transgenic humans? 

Your feedback is greatly appreciated. In response to your comments, the sentence in question has been amended to provide a more comprehensive understanding. Indeed, the text refers to experiences carried in mice. The revised wording is as follows: “Given the genetic basis of male infertility, the first strategy, tested in mice, was to overcome spermatogenic failure associated with monogenic diseases by delivery of an intact gene to deficient germ cells (Usmani, Ganguli et al. 2013). 

84-5 "efficiently" and "high" - provide context so the reader can understand what is meant - do the authors mean the experiments work efficiently, or that a high percentage of cells are transfected? And give some numbers or range of numbers - you're asking the readers to take your word for things when you choose adjectives - instead, provide values and let the readers decide for themselves.

Thanks for the comment, we have reworded the sentence: Gene therapy is effective in germ cells, as numerous publications have shown that conventional plasmids can be transferred into spermatogonia in several species with success, allowing their transcription in all cells of the germinal lineage (Usmani, Ganguli et al. 2013, Michaelis, Sobczak et al. 2014, Raina, Kumar et al. 2015, Wang, Liu et al. 2022).

93 Reference at the end of the sentence "most countries"

Thanks, we changed the sentence and added the reference: the new sentence is "… to avoid any eugenic deviations, transmissible changes in humans are illegal in 39 countries (Liu 2020)” (Liu, S. (2020). "Legal reflections on the case of genomeedited babies." Glob Health Res Policy 5: 24

93-4 Odd to say "multiple" and then list only one. 

Thanks for the comment, we have reworded the sentence: “Furthermore, the genetic modification of germ cell lines poses biological risks, including the induction of cancer, off-target effects, and cell mosaicism. Errors in editing may have adverse effects on future generations. It is exceedingly challenging to anticipate the consequences of genetic mosaicism, for instance, in a single individual. (Sadelain, Papapetrou et al. 2011, Ishii 2017).”

97 Is this really a "small" change? Again, would use adjectives carefully - to this reviewer, this is not a small change, but a significant one! And "should be" is not altogether convincing

Thanks for the comment, we have reworded the sentence: “Thanks to this change, the risk of genomic insertion is avoided, and thus there is no question of heritable alterations.”

What chance is there of retrotransposition? Is there any data in the literature for that, after injecting millions of copies of RNA one or more might be reverse transcribed and inserted into the genome?

This is certainly possible and is the putative origin for multiple intronless spermatid-expressed genes: 

The expert poses an interesting question, but one that unfortunately remains unanswered at present. Most papers on mRNA therapy state that there is no risk concerning genomic integration, but no reference is given (for instance see mRNA-based therapeutics: looking beyond COVID-19 vaccines. Lancet. 2024 doi: 10.1016/S0140-6736(23)02444-3). This is an important question, which deserves to be evaluated, but is beyond the scope of this manuscript. Nevertheless is remaining very debating (Igyarto and Qin 2024).

98 Odd to say "should be no risk" and then conclude with "there is no question" - so start the sentence with 'hedging', and then end with certainty - got to pick one or the other.

Thanks for the comment, we have reworded the sentence

99 "Complete" - probably not, would delete:

We removed the word: “The first part of this study presents a characterization of the protein expression patterns obtained following transfection of naked mRNA coding for reporter genes into the testes of mice”

101-2 Reference missing, as are numbers - what % of cases? 

Thank you, we changed the sentence and added the reference: “Among infertility disorders, oligoastheno-teratozoospermia  (OAT) is the most frequent (50 % (Thonneau, Marchand et al. 1991)” Thonneau, P., S. Marchand, A. Tallec, M. L. Ferial, B. Ducot, J. Lansac, P. Lopes, J. M. Tabaste and A. Spira (1991). "Incidence and main causes of infertility in a resident population (1,850,000) of three French regions (1988-1989)." Hum Reprod 6(6): 811-816.

103 Once again, the reference is missing:

We have added these references: (Colpi, Francavilla et al. 2018) (Cavallini 2006)

104-5 Awkward transition.

Thanks, we changed the transition: “The first part of this study presents a characterization of the protein expression patterns obtained following transfection of naked mRNA coding for reporter genes into the testes of mice. The second part is to apply the protocol to a preclinical mouse model of OAT.”

105 Backslash is odd - never seen it used in that way before

Removed

108 "completely infertile" is redundant;

Thank you, we changed it accordingly: “Patients and mice carrying mutations in the ARMC2 gene present a canonical OAT phenotype and are infertile”.

and is a KO mouse really "preclinical"? 

The definition of preclinical research, is research involving the use of animals to ascertain the potential efficacy of a drug, procedure, or treatment. Preclinical studies are conducted prior to any testing in humans. Our KO mouse model has been shown to mimic human infertility. Indeed Armc2-/-mice exhibit a phenotype that is identical to that observed in humans. Our study is in line with this definition. For this reason, we have decided to maintain our current position and to use the term "preclinical" in the article. 

110  Delete "sperm".

Thank you, we changed it accordingly: “The preclinical Armc2 deficient (Armc2 KO) mouse model is therefore a valuable model to assess whether in vivo injection of naked mRNA combined with electroporation can restore spermatogenesis”

111  "Easy"? Really? 

We changed it accordingly: “We chose this model for several reasons: first, Armc2 KO mice are sterile and all sperm exhibit short, thick or coiled flagella [13].”

112-3 "completely immobile" is redundant - either they are immobile or not.

Thank you, we changed it accordingly: “As a result, 100 % of sperm are immobile, thus it should be easy to determine the efficacy of the technique by measuring sperm motility with a CASA system.”

108-33 Condense this lengthy text into a coherent few sentences to give readers a sense of what you sought to accomplish, broadly how it was done, and what you found. This reads more like a Results section

Thanks for the comment, we shortened the text.

Materials and Methods 

The sections appear to have been written by different scientists - the authors should standardize so that similar detail and formatting are used - e.g., in some parts the source is in parentheses with catalog number, in others not, some have city, state, country, others do not... the authors should check eLife mandates for this type of information and provide. 

We are grateful for your feedback. We standardized the text, and if we had missed some, as outlined on the E-Life website, we can finish to format the article once it has been accepted for publication in the journal before sending the VOR.

134 Misspelling

We corrected the misspelling  

142 Just reference, don't need to spell it out.

Thanks, we changed it accordingly: “and the Armc2 KO mouse strain obtained by CRISPR-Cas9 (Coutton, Martinez et al. 2019). Experiments”

150 What is XXX?

We would like to express our gratitude for bringing this error to our attention. We have duly rectified the issue: “obtained from Merck (Darmstadt, Germany).”

157-60 Are enough details provided for readers to repeat this if necessary? Doesn't seem so to this reviewer; if kits were followed, then can say "using manufacturer's protocol", or refer to another manuscript - but this is too vague. 

Thanks, we change it accordingly: After expansion, plasmids were purified with a NucleoBond Xtra Midi kit (740410-50; Macherey-Nagel, Düren, Germany) using manufacturer's protocol.”

165 Again, too few details - how was it purified? What liquid was it in?

Thanks for the comment, the EEV plasmids were purified like all other plasmids. We change the text: “All plasmids,EEV CAGs-GFP-T2A-Luciferase,((EEV604A-2), System Bioscience, Palo Alto, CA, USA), mCherry plasmid ( given by Dr. Conti MD at UCSF, San Francisco, CA, USA) and EEV-Armc2-GFP plasmid (CUSTOM-S017188-R2-3,Trilink,San Diego, USA) were amplified by bacterial transformation” 

170 Seems some words are missing - and will everyone know Dr. Conti by last name alone? Would spell out, and the details of the plasmid must either be provided or a reference given; how was amplification done? Purification? What was it resuspended in? 

Thank for the remark, the mcherry plasmids were purified like all other plasmids. We change the text: “All plasmids,EEV CAGs-GFP-T2A-Luciferase,((EEV604A-2), System Bioscience, Palo Alto, CA, USA), mCherry plasmid ( given by Dr. Conti MD, UCSF, San Francisco, CA, USA) and EEV-Armc2-GFP plasmid (CUSTOM-S017188-R2-3,Trilink,San Diego, USA) were amplified by bacterial transformation”

175 Again, for this plasmid provide more information - catalog number, reference, etc; how amplified and purified, what resuspension buffer?

Thank you for the remark, as We mentioned, we add this sentence for the preparation: “All plasmids, EEV CAGs-GFP-T2A-Luciferase,((EEV604A-2), System Bioscience, Palo Alto, CA, USA), mCherry plasmid (given by Dr. Conti MD at UCSF, San Francisco, CA, USA) and EEV-Armc2-GFP plasmid (CUSTOMS017188-R2-3,Trilink,San Diego, USA) were amplified by bacterial transformation” and we add these sentence “The EEV-Armc2-GFP plasmid used for in vivo testes microinjection and electroporation was synthesized and customized by Trilink (CUSTOM-S017188-R2-3,San Diego, USA).”

183 What sequence, or isoform was used? Mouse or human? 

Thanks, we changed accordingly: “This non-integrative episome contains the mice cDNA sequences of Armc2 (ENSMUST00000095729.11)”

186-7 Provide sequence or catalog number; what was it resolubilized in?

Thanks we changed accordingly “the final plasmid concentration was adjusted to 9 μg μL-1 in water.” We provided the sequence of EEV-Armc2-GFP in supp data 6.

207-219 Much better, this is how the entire section needs to be written! 

237-240 Font

Thanks for the comment, we changed it accordingly

246 Cauda, and sperm, not sperm cells

Thanks for the comment, we changed it accordingly

255-6 Which was done first? Would indicate clearly.

Thanks for the comment, we changed the sentence: “Adult mice were euthanized by cervical dislocation and then transcardiac perfused  with 1X PBS”

281-2 Provide source for software - company, location, etc: 

We changed it accordingly: FIJI software (Opened source software) was used to process and analyze images and Imaris software (Oxford Instruments Tubney Woods, Abingdon, Oxon OX13 5QX, UK) for the 3D reconstructions.  

323 um, not uM. 

Thanks for the comment, we changed our mistake: “After filtration (100 µm filter)”

Results 

369 Weighed.  

Thanks for the comment, we changed our mistake: “the testes were measured and weighed”

371 No difference in what, specifically?

Thanks for the comment, we changed the sentence to: “No statistical differences in length and weight were observed between control and treated testes”

375 "was respected"? What does this mean?

Thanks for the comment, we changed the sentence to “The layered structure of germ cells were identical in all conditions”

378  This is highly unlikely to be true, as even epididymal sperm from WT animals are often defective - the authors are saying there were ZERO morphological defects? Or that there was no difference between control and treated? Only showing 2-3 sperm for control vs treatment is not sufficient.

Your observation that the epididymal spermatozoa from wild-type animals exhibited defective morphology is indeed true. The prevalence of these defects varies by strain, with an average incidence of 20% to 40% (Kawai, Hata et al., 2006; Fan, Liu et al., 2015). To provide a more comprehensive representation, we conducted a Harris-Shorr staining procedure and included a histogram of the percentage of normal sperm in each condition (new figure 2F4). Furthermore, Harris-Shorr staining of the epididymal sperm cells revealed that there were no discernible increases in morphological defects when mRNA and EEV were utilized, in comparison with the control. We add the sentence “At last, Harris-Shorr staining of the epididymal sperm cells demonstrated that there were no increases in morphological defects when mRNA and EEV were used in comparison with the control”.

379  "safe" is not the right word - better to say "did not perturb spermatogenesis". 

Thanks, we changed it accordingly: “these results suggest that in vivo microinjection and electroporation of EEV or mRNA did not perturb spermatogenesis”

382-3 This sentence needs attention, doesn't make sense as written: 

Thanks for the remark, we changed the sentence to: “No testicular lesions were observed on the testes at any post injection time”

389  How long after injection? 

Thanks for the comment, we changed the sentence to: “It is worth noting that both vectors induced GFP expression at one day post-injection”

390  Given the duration of mouse spermatogenesis (~35 days), for GFP to persist past that time suggests that it was maintained in SSCs? How can the authors explain how such a strong signal was maintained after such a long period of time? How stable are the episomally-maintained plasmids, are they maintained 100% for months? And if they are inherited by progeny of SSCs, shouldn't they be successively diluted over time? And if they are inherited by daughter cells such that they would still be expressed 49 days after injection, shouldn't all the cells originating from that SSC also be positive, instead of what appear to be small subsets as shown in Fig. 3H2? Overall, this reviewer is struggling to understand how a plasmid would be inherited and passed through spermatogenesis in the manner seen in these results. 

Thanks for the comment. 

This point was already underlined in public review. We paste here our answer: “The non-insertional Enhanced Episomes Vector (EEV) plasmid is a non-viral episome based on the Epstein-Barr virus (EBV: Epstein-Barr Virus). Its maintenance within the cell is made possible by its ability to replicate in a synchronous manner with the host genome and to segregate into daughter cells. This is due to the fact that EEV is composed of two distinct elements derived from EBV: an origin of replication (oriP) and an Epstein-Barr Nuclear Antigen 1 (EBNA1) expression cassette (Gil, Gallaher, and Berk, 2010).   The oriP is a locus comprising two EBNA1-binding domains, designated as the Family of Repeats (FR) and Dyad Symmetry (DS). The FR is an array of approximately 20 EBNA1-binding sites (20 repeats of 30 bp) with high affinity, while the DS comprises four lower-affinity sites operating in tandem (Ehrhardt et al., 2008). 

The 641-amino-acid EBNA1 protein contains numerous domains.The N-terminal domains are rich in glycines and alanines, which enable interaction with host chromosomes. The C-terminal region is responsible for binding to oriP (Hodin, Najrana, and Yates, 2013a). The binding of EBNA1 to the DS element results in the recruitment of the origin of replication. This results in the synchronous initiation of extra-chromosomal EEV replication with host DNA at each S phase of the cell cycle (Düzgüneş, Cheung, and Konopka 2018a). Furthermore, EBNA1 binding to the FR domain induces the formation of a bridge between metaphase chromosomes and the vector during mitosis. This binding is responsible for the segregation of the EEV episome in daughter cells (Düzgüneş, Cheung, and Konopka 2018b). It is notable that EEV is maintained at a rate of 90-95% per cell division.”

Because of the intrinsic properties of EEV described above, the presence of the reporter protein at 119 day after injection was likely due to the maintenance of the plasmid, mostly in Sertoli cells, and not to the DNA integration of the plasmid.

Of note, the specificity of EEV was already indicated in the introduction. Nevertheless, we have added more information about it to help the readers (lines 124-128 clean copy)  

398 Which "cell types"? 

Your feedback is greatly appreciated, and the sentence in question has been amended to provide a more comprehensive understanding. The revised wording is as follows: These results suggest that GFPmRNA and EEV-GFP targeted different seminiferous cell types, such as Sertoli cells and all germline cells, or that there were differences in terms of transfection efficiency.

409 Why is it important to inject similar copies of EEV and mRNA? Wouldn't the EEV be expected to generate many, many more copies of RNA per molecule than the mRNAs when injected directly?? 

We removed the word importantly. 

415 How is an injected naked mRNA stably maintained for 3 weeks? What is the stability of this mRNA?? Wouldn't its residence in germ cells for 21 days make it more stable than even the most stable endogenous mRNAs? Even mRNAs for housekeeping genes such as actin, which are incredibly stable, have half-lives of 9-10 hours.

We appreciate your inquiry and concur with your assessment that mRNA stability is limited.  It is our hypothesis that the source of the confusion lies in the fact that we injected mRNA coding for the GFP protein, rather than mRNA tagged with GFP. After a three-week observation period, we did not observe the mRNA, but we observed the expression of the GFP protein induced by the mRNA. To draw the reader's attention to this point, we have added the following sentence to the text “It is important to underline that the signal measured is the fluorescence emitted by the GFP. This signal is dependent of both the half-lives of the plasmid/mRNA and the GFP. Therefore, the kinetic of the signal persistence (which is called here expression) is a combination of the persistence of the vector and the synthetized protein. See lines 469-472 clean copy. 

This being said, it is difficult to compare the lifespan of a cellular mRNA with that of a mRNA that has been modified at different levels, including 5’Cap, mRNA body, poly(A)tail modifications, which both increase mRNA stability and translation (see The Pivotal Role of Chemical Modifications in mRNA Therapeutics  (2022) https://doi.org/10.3389/fcell.2022.901510). This question is discussed lines 687698 clean copy

467 "safely" should be deleted

Thanks, we removed the word: “To validate and confirm the capacity of naked mRNA to express proteins in the testes after injection and electroporation”

470  Except that apoptotic cells were clearly seen in Figure 2:

We would like to thank the reviewer for their comment. We agree that the staining of the provided sections were of heterogenous quality. To address the remark, we carried out additional HE staining for all conditions, and we now present testis sections correctly stained obtained in the different condition in Fig. 2 and Supp. 7. Our observations revealed that the number of apoptotic cells remained consistent across all conditions.

471  "remanence"?

We appreciate your feedback and have amended the sentence to provide clear meaning. The revised wording is as follows: “The assessment of the temporal persistence of testicular mCherry fluorescent protein expression revealed a robust red fluorescence from day 1 post-injection, which remained detectable for at least 15 days (Fig. Supp. 3 B2, C2, and D2).”

489 IF measures steady-state protein levels, not translation; should say you determined when ARMC2 was detectable. 

Thanks for the remark, we changed the sentence to: “ By IF, we determined when ARMC2 protein was detectable during spermatogenesis.”

491 Flagella

Thanks for the comment, we changed our mistake: “in the flagella of the elongated spermatids (Fig 9A)”

Discussion 

The Discussion is largely a re-hashing of the Methods and Results, with additional background.

Message stability must be addressed - how is a naked mRNA maintained for 21 days?

As previously stated, it is our hypothesis that the source of the confusion lies in the fact that we injected mRNA coding for the GFP protein, rather than mRNA tagged with GFP. After a three-week observation period, we did not observe the mRNA, but we observed the synthetized GFP protein. This point and the stability of protein in the testis is now discussed lines 677-684 (clean copy).

556 How do the authors define "safe"?

Thanks for the comment, we changed the sentence to be clearer: “Our results also showed that the combination of injection and electroporation did not perturb spermatogenesis when electric pulses are carefully controlled”

563 Synthesized

Thanks, we changed it accordingly

602 Again, this was not apparent, as there were more apoptotic cells in Fig. 2 - data must be provided to show "no effect".

As previously stated, we carried out additional HE staining for all conditions, as can be observed in Fig. 2 . Our observations revealed that the number of apoptotic cells remained consistent across all conditions.

629-30 This directly contradicts the authors' contention in the Introduction that ICSI was unsafe - how is this procedure going to be an advancement over ICSI as proposed, if ICSI needs to be used?? Why not just skip all this and do ICSI then?? Perhaps if this technique was used to 'repair' defects in spermatogonia or spermatocytes, then that makes more sense. But if ICSI is required, then this is not an advancement when trying to rescue a sperm morphology/motility defect.

In light of the latest findings (Fig 12), we have revised this part of the discussion and this paragraph no longer exist.

Nevertheless, to address specifically the reviewer’s remark, we would like to underline that ICSI with sperm from fertile donor is always more efficient than ICSI with sperm from patient suffering of OAT condition. Our strategy, by improving sperm quality, will improve the efficiency of ICSI and at the end will increase the live birth rate resulting from the first fresh IVF cycle.

640-2 What is meant by "sperm organelles" And what examples are provided for sperm proteins being required at or after fertilization? 

This paragraph was also strongly modified and the notion of protein persistence during spermatogenesis was discussed in the paragraph on fluorescent signal duration. See lines 698-705.

651 "Dong team"??

Thanks for the comment, we added the references. 

Figure 2D2 - tubule treated with EEV-GFP appears to have considerably more apoptotic cells - this reviewer counted ~10 vs 0 in control; also, many of the spermatocytes appear abnormal in terms of their chromatin morphology - the authors must address this by staining for markers of apoptosis - not fair to conclude there was no difference when there's a very obvious difference! 

We would like to thank the reviewer for their comment. This point was already addressed. As previously stated, we provide now new testis sections for all condition (see Fig. 2). Our observations revealed that the number of apoptotic cells remained consistent across all conditions.

Figure 2D3 staining is quite different than D1-2, likely a technical issue - looks like no hematoxylin was added? Need to re-stain so results can be compared to the other 2 figures 

As previously stated, we carried out additional HE staining for all conditions, and new images are provided, with similar staining. 

Figure 3 - the fluorescent images lack any context of tubule structure so it is nearly impossible to get a sense of what cells express GFP, or whether they're in the basal vs adluminal compartment - can the authors outline them? Indicate where the BM and lumen are. 

We would like to thank the reviewer for their comment. This figure provides actually a global view of the green fluorescent protein (GFP) expression at the surface of the testis. The entire testis was placed under an inverted epifluorescence microscope, and a picture of the GFP signal was recorded. For this reason, it is impossible to delineate the BM and the lumen. It should be noted that the fluorescence likely originates from different seminiferous tubules.

Author response image 1.

So, for Figure 3 if the plasmid is being uptaken by cells and maintained as an episome, is it able to replicate? Likely not. 

Yes! it is the intrinsic property of the episome, see the detailed explanation provided above about the EEV plasmid

So, initially, it could be in spermatogonia, spermatocytes, and spermatids. As time progressed those initially positive spermatids and then spermatocytes would be lost - and finally, the only cells that should be positive would be the progeny of spermatogonia that were positive - but, as they proliferate shouldn't the GFP signal decline? 

Because EEV is able  to replicate in a synchronous manner with the host genome and to segregate into daughter cells at a level of 90% of the mother cell, the expected decline is very slow.

And, since clones of germ cells are connected throughout their development, shouldn't the GFP diffuse through the intercellular bridges so entire clones are positive? Was this observed? 

We did not perform IF experiments further than 7 days after injection, a time too short to observe what the reviewer suggested. Moreover, if at 1 day after injection, GFP synthesized from injected EEV was found in both germ cells and Sertoli cells (Fig 7), after one week, the reporter proteins were only observable in Sertoli cells. This result suggests that EEV is maintained only in Sertoli cells, thus preventing the observation of stained clones.

Can these sections be stained for the ICB TEX14 so that clonality can be distinguished? Based on the apparent distance between cells, it appears some are clones, but many are not... 

We thank the reviewer for this suggestion but we are not able to perform testis sectioning and costaining experiments because the PFA treatment bleaches the GFP signal. We also tested several GFP antibodies, but all failed.  

Nevertheless, we were able to localize and identify transfected cells thank to the whole testis optical clearing, combined with a measure of GFP fluorescence and three-dimensional image reconstructions. 

For Figure 4, with the mRNA-GFP, why does the 1-day image (which looks similar to the plasmidtransfected) look so different from days 7-21? 

And why do days 7-21 look so different from those days in Fig 3? 

Thank you for your feedback. It is an excellent question. Because of the low resolution of the whole testis epifluorescences imaging and light penetration issue, we decided to carry-out whole testis optical clearing and three-dimensional image reconstructions experiments, in order to get insights on the transfection process. At day 1, GFP synthesized from EEV injection was found in spermatogonia, spermatocytes and Sertoli cells (Fig 7).  After one week, the reporter protein synthesized from injected EEV was only observable in Sertoli cells.

In contrast, for mRNA, on day 1 and day 7 post-injection, GFP fluorescent signal was associated with both Sertoli cells and germ cells. This explains why patterns between mRNA-GFP and EEV-GFP are similar at day 1 and different at day 7 between both conditions. 

Why do the authors think the signal went from so strong at 21 to undetectable at 28? What changed so drastically over those 7 days?

What is the half-life of this mRNA supposed to be? It seems that 21 days is an unreasonably long time, but then to go to zero at 28 seems also odd... Please provide some explanation, and context for whether the residence of an exogenous mRNA for 21 days is expected. 

As previously stated, it is our hypothesis that the source of the confusion lies in the fact that we injected mRNA coding for the GFP protein, rather than mRNA tagged with GFP. After a three-week observation period, we did not observe the mRNA, but we observed the GFP protein produced by the mRNA. The time of observation of the reporter proteins expressed by the respective mRNA molecules (mCherry, luciferase, or GFP) ranged from 15 to 21 days. Proteins have very different turnover rates, with half-lives ranging from minutes to days. Half-lives depend on proteins but also on tissues. As explained in the discussion, it has been demonstrated that proteins involved in spermatogenesis exhibit a markedly low turnover rate and this explains the duration of the fluorescent signal. 

The authors should immunostain testis sections from controls and those with mRNA and plasmid and immunostain with established germ cell protein fate markers to show what specific germ cell types are GFP+

Thank you for your feedback. As previously mentioned, we were unable to perform testis sectioning and co-staining because the PFA treatment bleaches the GFP signal and because we were unable to reveal GFP with an GFP antibody, for unknown reasons.

For the GFP signal to be maintained past 35 days, the plasmid must have integrated into SSCs - and for that to happen, the plasmid would have to cross the blood-testis-barrier... is this expected? 

We are grateful for your observation. 

First, as explained above, we do not think that the plasmid has been integrated. 

Concerning the blood-testing barrier.  It bears noting that electroporation is a technique that is widely utilized in biotechnology and medicine for the delivery of drugs and the transfer of genes into living cells (Boussetta, Lebovka et al. 2009). This process entails the application of an electric current, which induces the formation of hydrophilic pores in the lipid bilayer of the plasma membrane (Kanduser, Miklavcic et al. 2009). The pores remain stable throughout the electroporation process and then close again once it is complete. Consequently, as electroporation destabilizes the cell membrane, it can also destabilize the gap junctions responsible of the blood-testis barrier. This was actually confirmed by several studies, which have observed plasmid transfection beyond the blood-testis barrier with injection into rete testis following electroporation (Muramatsu, Shibata et al. 1997, Kubota, Hayashi et al. 2005, Danner, Kirchhoff et al. 2009, Kanduser, Miklavcic et al. 2009, Michaelis, Sobczak et al. 2014).

Figure 9 - authors should show >1 cell - this is insufficient; also, it's stated it's only in the flagella, but it also appears to be in the head as well. And is this just the principal piece?? And are the authors sure those are elongating vs condensing spermatids? Need to show multiple tubules, at different stages, to make these claims

We have partly answered to this question in the public review; We pastehere  our answer

“We present now new images showing the full seminiferous tubules as requested (see supp fig 6). In this new figure, it is clear that Armc2 is only expressed in spermatids. We have also added in this figure an analysis of the RNA-seq database produced by Gan's team (Gan, Wen et al. 2013), confirming that ArmC2 expression is predominantly expressed at the elongated spermatid stage. This point is now clearly indicated in the text.”

Concerning the localization of the protein in the head, we confirm that the base of the manchette is stained but we have no explanation so far. This point is now indicated in the manuscript.

Figure 10B2 image - a better resolution is necessary

We are grateful for your feedback. We concede that the quality of the image was not optimal. Consequently, We have replaced it with an alternative.

Figure 11 - in control, need to show >1 sperm; and lower-mag images should be provided for all samples to show population-wide effects; showing 1 "normal" sperm per group (white arrows) is insufficient: 

We are grateful for your feedback. We conducted further experiments and provide now additional images in Supp. figure 8.

Reviewer #3 (Recommendations For The Authors): 

In this study, Vilpreux et al. developed a microinjection/electroporation method in order to transfect RNA into testicular cells. The authors studied several parameters of treated testis and compared the injection of DNA versus RNA. Using the injection of Armc2 RNA into mice with an Armc2 knockout the authors were able to (partly) rescue the fertility phenotype. 

Minor points. 

Figure 6 + lines 553+554: might it be that the staining pattern primarily on one side of the testis is due to the orientation of the scissor electrode during the electroporation procedure and the migration direction of negatively charged RNA molecules (Figure 6)? 

Your input is greatly appreciated. We concur that the observed peripheral expression is due to both the electroporation and injection. Accordingly, we have amended the sentence as follows: "The peripheral expression observed was due to the close vicinity of cells to the electrodes, and to a peripheral dispersal of the injected solution, as shown by the distribution of the fluorescent i-particles NIRFiP-180."

Discussion of the safety aspect (lines 601-608): The authors state several times that there are no visible tissue changes after the electroporation procedure. However, in order to claim that this procedure is "safe", it is necessary to examine the offspring born after microinjection/electroporation. 

Your input is greatly appreciated. Consequently, the term "safe" has been replaced with "did not perturb spermatogenesis" in accordance with the provided feedback. Your assertion is correct; an examination of the offspring born would be necessary to ascertain the safety of the procedure. Due to the quantity of motile sperm obtained, it was not possible to produce offspring through natural mating. However, novel Armc2-/--rescued sperm samples have been produced and in vitro fertilization (IVF) and intracytoplasmic sperm injection (ICSI) experiments have been conducted. The results demonstrate that the Armc2-/--rescued sperm can successfully fertilize eggs and produce two-cell embryos by IVF and blastocysts by ICSI. These outcomes are visually represented in Figure 12. The development of embryos up to the blastocyst stage is a step in the right direction.

The discussion section could be shortened. Lines 632-646 are largely a repetition of the introductory section. In addition, the Dong paper (ref. 25) may be interesting; however, this part could also be shortened (lines 647-676). This reviewer would prefer the authors to focus on the technique (different application sites and applied nucleotides) and proof of concept for (partial) phenotype rescue in the knockout mice. 

Your contribution is highly valued. In light of your observations and the latest findings, we have substantially revised the discussion accordingly.

Line 63: oocytes rather than eggs.

We are grateful for your input, but we have decided to retain our current position and to use the term "eggs" rather than "oocytes" in our writing because the definition of an oocyte is a female gametocyte or germ cell involved in reproduction. In other words, oocyte corresponds to a germ cell inside the ovary and after ovulation become an egg.  

Boussetta, N., N. Lebovka, E. Vorobiev, H. Adenier, C. Bedel-Cloutour and J. L. Lanoiselle (2009). "Electrically assisted extraction of soluble matter from chardonnay grape skins for polyphenol recovery." J Agric Food Chem 57(4): 1491-1497.

Cavallini, G. (2006). "Male idiopathic oligoasthenoteratozoospermia." Asian J Androl 8(2): 143-157.

Colpi, G. M., S. Francavilla, G. Haidl, K. Link, H. M. Behre, D. G. Goulis, C. Krausz and A. Giwercman (2018). "European Academy of Andrology guideline Management of oligo-asthenoteratozoospermia." Andrology 6(4): 513-524.

Coutton, C., G. Martinez, Z. E. Kherraf, A. Amiri-Yekta, M. Boguenet, A. Saut, X. He, F. Zhang, M. Cristou-Kent, J. Escoffier, M. Bidart, V. Satre, B. Conne, S. Fourati Ben Mustapha, L. Halouani, O. Marrakchi, M. Makni, H. Latrous, M. Kharouf, K. Pernet-Gallay, M. Bonhivers, S. Hennebicq, N. Rives, E. Dulioust, A. Toure, H. Gourabi, Y. Cao, R. Zouari, S. H. Hosseini, S. Nef, N. Thierry-Mieg, C. Arnoult and P. F. Ray (2019). "Bi-allelic Mutations in ARMC2 Lead to Severe Astheno-Teratozoospermia Due to Sperm Flagellum Malformations in Humans and Mice." Am J Hum Genet 104(2): 331-340.

Danner, S., C. Kirchhoff and R. Ivell (2009). "Seminiferous tubule transfection in vitro to define postmeiotic gene regulation." Reprod Biol Endocrinol 7: 67.

Gan, H., L. Wen, S. Liao, X. Lin, T. Ma, J. Liu, C. X. Song, M. Wang, C. He, C. Han and F. Tang (2013). "Dynamics of 5-hydroxymethylcytosine during mouse spermatogenesis." Nat Commun 4: 1995. Igyarto, B. Z. and Z. Qin (2024). "The mRNA-LNP vaccines - the good, the bad and the ugly?" Front Immunol 15: 1336906.

Ishii, T. (2017). "Germ line genome editing in clinics: the approaches, objectives and global society." Brief Funct Genomics 16(1): 46-56.

Kanduser, M., D. Miklavcic and M. Pavlin (2009). "Mechanisms involved in gene electrotransfer using high- and low-voltage pulses--an in vitro study." Bioelectrochemistry 74(2): 265-271.

Kubota, H., Y. Hayashi, Y. Kubota, K. Coward and J. Parrington (2005). "Comparison of two methods of in vivo gene transfer by electroporation." Fertil Steril 83 Suppl 1: 1310-1318.

Michaelis, M., A. Sobczak and J. M. Weitzel (2014). "In vivo microinjection and electroporation of mouse testis." J Vis Exp(90).

Muramatsu, T., O. Shibata, S. Ryoki, Y. Ohmori and J. Okumura (1997). "Foreign gene expression in the mouse testis by localized in vivo gene transfer." Biochem Biophys Res Commun 233(1): 45-49.

Raina, A., S. Kumar, R. Shrivastava and A. Mitra (2015). "Testis mediated gene transfer: in vitro transfection in goat testis by electroporation." Gene 554(1): 96-100.

Sadelain, M., E. P. Papapetrou and F. D. Bushman (2011). "Safe harbours for the integration of new DNA in the human genome." Nat Rev Cancer 12(1): 51-58.

Thonneau, P., S. Marchand, A. Tallec, M. L. Ferial, B. Ducot, J. Lansac, P. Lopes, J. M. Tabaste and A. Spira (1991). "Incidence and main causes of infertility in a resident population (1,850,000) of three French regions (1988-1989)." Hum Reprod 6(6): 811-816.

Usmani, A., N. Ganguli, H. Sarkar, S. Dhup, S. R. Batta, M. Vimal, N. Ganguli, S. Basu, P. Nagarajan and S. S. Majumdar (2013). "A non-surgical approach for male germ cell mediated gene transmission through transgenesis." Sci Rep 3: 3430.

Wang, L., C. Liu, H. Wei, Y. Ouyang, M. Dong, R. Zhang, L. Wang, Y. Chen, Y. Ma, M. Guo, Y. Yu, Q. Y. Sun and W. Li (2022). "Testis electroporation coupled with autophagy inhibitor to treat nonobstructive azoospermia." Mol Ther Nucleic Acids 30: 451-464.

eLife Assessment

This valuable work presents an interpretable protein-DNA Energy Associative (IDEA) model for predicting binding sites and affinities of DNA-binding proteins. The study provides a detailed description of the method, making it reproducible. However, the generalizability of the prediction model presents certain concerns, and the supporting evidence appears incomplete. Nonetheless, with a thorough re-examination of the training and testing procedures, this model can be widely applicable for predicting genome-wide protein-DNA binding sites.

Reviewer #1 (Public review):

Summary:

This work presents an Interpretable protein-DNA Energy Associative (IDEA) model for predicting binding sites and affinities of DNA-binding proteins. Experimental results demonstrate that such an energy model can predict DNA recognition sites and their binding strengths across various protein families and can capture the absolute protein-DNA binding free energies.

Strengths:

(1) The IDEA model integrates both structural and sequence information, although such an integration is not completely original.

(2) The IDEA predictions seem to have agreement with experimental data such as ChIP-seq measurements.

Weaknesses:

(1) The authors claim that the binding free energy calculated by IDEA, trained using one MAX-DNA complex, correlates well with experimentally measured MAX-DNA binding free energy (Figure 2) based on the reported Pearson Correlation of 0.67. However, the scatter plot in Figure 2A exhibits distinct clustering of the points and thus the linear fit to the data (red line) may not be ideal. As such. the use of the Pearson correlation coefficient that measures linear correlation between two sets of data may not be appropriate and may provide misleading results for non-linear relationships.

(2) In the same vein, the linear Pearson Correlation analysis performed in Figure 5A and the conclusion drawn may be misleading.

(3) The authors included the sequences of the protein and DNA residues that form close contacts in the structure in the training dataset, whereas a series of synthetic decoy sequences were generated by randomizing the contacting residues in both the protein and DNA sequences. In particular, synthetic decoy binders were generated by randomizing either the DNA (1000 sequences) or protein sequences (10,000 sequences) from the strong binders. However, the justification for such randomization and how it might impact the model's generalizability and transferability remain unclear.

(4) The authors performed Receiver Operating Characteristic (ROC) analysis and reported the Area Under the Curve (AUC) scores in order to quantitate the successful identification of the strong binders by IDEA. It would be beneficial to analyze the precision-recall (PR) curve and report the PRAUC metric which could be more robust.

Reviewer #2 (Public review):

Summary:

Zhang et al. present a methodology to model protein-DNA interactions via learning an optimizable energy model, taking into account a representative bound structure for the system and binding data. The methodology is sound and interesting. They apply this model for predicting binding affinity data and binding sites in vivo. However, the manuscript lacks discussion of/comparison with state-of-the-art and evidence of broad applicability. The interpretability aspect is weak, yet over-emphasized.

Strengths:

The manuscript is well organized with good visualizations and is easy to follow. The methodology is discussed in detail. The IDEA energy model seems like an interesting way to study a protein-DNA system in the context of a given structure and binding data. The authors show that an IDEA model trained on one system can be transferred to other structurally similar systems. The authors show good performance in discriminating between binding-vs-decoy sequences for various systems, and binding affinity prediction. The authors also show evidence of the ability to predict genome-wide binding sites.

Weaknesses:

An energy-based model that needs to be optimized for specific systems is inherently an uncomfortable idea. Is this kind of energy model superior to something like Rosetta-based energy models, which are generally applicable? Or is it superior to family-specific knowledge-based models? It is not clear.

Prediction of binding affinity is a well-studied domain and many competitors exist, some of which are well-used. However, no quantitative comparison to such methods is presented. To understand the scope of the presented method, IDEA, the authors should discuss/compare with such methods (e.g. PMID 35606422).

The term "interpretable" has been used lavishly in the manuscript while providing little evidence on the matter. The only evidence shown is the family-specific residue-nucleotide interaction/energy matrix and speculations on how these values are biologically sensible. Recent works already present more biophysical, fine-grained, and sometimes family-independent interpretability (e.g. PMID 39103447, 36656856, 38352411, etc.). The authors should put into context the scope of the interpretability of IDEA among such works.

The manuscript disregards subtle yet important differences in commonly used terminology in the field. For example, the authors use the term "specificity" and "affinity" almost interchangeably (for example, the caption for Figure 3A uses "specificity" although the Methods text describes the prediction as about "affinity"). If the authors are looking to predict specificity, IDEA needs to be put in the context of the corresponding state-of-the-art (PMID 36123148, 39103447, 38867914, 36124796, etc).

It is not clear how much the learned energy model is dependent on the structural model used for a specific system/family. It would be interesting to see the differences in learned model based on different representative PDB structures used. Similarly, the supplementary figures show a lack of discriminative power for proteins like PDX1 (homeodomain family), POU, etc. Can the authors shed some light on why such different performances?

It is also not clear if IDEA's prediction for reverse complement sequences is the same for a given sequence. If so, how is this property being modelled? Either this description is lacking or I missed it.

Reviewer #3 (Public review):

Summary:

Protein-DNA interactions and sequence readout represent a challenging and rapidly evolving field of study. Recognizing the complexity of this task, the authors have developed a compact and elegant model. They have applied well-established approaches to address a difficult problem, effectively enhancing the information extracted from sparse contact maps by integrating artificial sequences decoy set and available experimental data. This has resulted in the creation of a practical tool that can be adapted for use with other proteins.

Strengths:

(1) The authors integrate sparse information with available experimental data to construct a model whose utility extends beyond the limited set of structures used for training.

(2) A comprehensive methods section is included, ensuring that the work can be reproduced. Additionally, the authors have shared their model as a GitHub project, reflecting their commitment to transparency of research.

Weaknesses:

(1) The coarse-graining procedure appears artificial, if not confusing, given that full-atom crystal structures provide more detailed information about residue-residue contacts. While the selection procedure for distance threshold values is explained, the overall motivation for adopting this approach remains unclear. Furthermore, since this model is later employed as an empirical potential for molecular modeling, the use of P and C5 atoms raises concerns, as the interactions in 3SPN are modeled between Cα and the nucleic base, represented by its center of mass rather than P or C5 atoms.

(2) Although the authors use a standard set of metrics to assess model quality and predictive power, some ΔΔG predictions compared to MITOMI-derived ΔΔG values appear nonlinear, which casts doubt on the interpretation of the correlation coefficient.

(3) The discussion section lacks information about the model's limitations and a comprehensive comparison with other models. Additionally, differences in model performance across various proteins and their respective predictive powers are not addressed.

Author response:

Reviewer 1:

Summary: This work presents an Interpretable protein-DNA Energy Associative (IDEA) model for predicting binding sites and affinities of DNA-binding proteins. Experimental results demonstrate that such an energy model can predict DNA recognition sites and their binding strengths across various protein families and can capture the absolute protein-DNA binding free energies.

We appreciate the reviewer’s careful assessment of the paper, and we thank the reviewer for the insightful suggestions and comments.

Strengths:

(1) The IDEA model integrates both structural and sequence information, although such an integration is not completely original. (2) The IDEA predictions seem to have agreement with experimental data such as ChIP-seq measurements.

We appreciate the reviewer’s comments on the strength of the paper.

Weaknesses:

(1) The authors claim that the binding free energy calculated by IDEA, trained using one MAX-DNA complex, correlates well with experimentally measured MAX-DNA binding free energy (Figure 2) based on the reported Pearson Correlation of 0.67. However, the scatter plot in Figure 2A exhibits distinct clustering of the points and thus the linear fit to the data (red line) may not be ideal. As such. the use of the Pearson correlation coefficient that measures linear correlation between two sets of data may not be appropriate and may provide misleading results for non-linear relationships.

We thank the reviewer for the insightful comments and agree that the linear fit between our predictions and the experimental data may not be ideal. The primary utility of the IDEA model is for assessing the relative binding affinities of different DNA sequences. To further support this, we plan to conduct additional statistical analyses that are independent of the linear correlation assumption but instead focus on the ranked order of DNA sequence binding affinities.

(2) In the same vein, the linear Pearson Correlation analysis performed in Figure 5A and the conclusion drawn may be misleading.

We thank the reviewer for the insightful comments. We will perform the same analysis for Figure 5A as detailed in our response to the previous comments.

(3) The authors included the sequences of the protein and DNA residues that form close contacts in the structure in the training dataset, whereas a series of synthetic decoy sequences were generated by randomizing the contacting residues in both the protein and DNA sequences. In particular, synthetic decoy binders were generated by randomizing either the DNA (1000 sequences) or protein sequences (10,000 sequences) from the strong binders. However, the justification for such randomization and how it might impact the model’s generalizability and transferability remain unclear.

We thank the reviewer for the insightful comments. We will perform additional analyses to assess the robustness of our model predictions with respect to the number of randomized decoys. Additionally, we will examine how randomization would potentially affect the model’s generalizability and transferability.

(4) The authors performed Receiver Operating Characteristic (ROC) analysis and reported the Area Under the Curve (AUC) scores in order to quantitate the successful identification of the strong binders by IDEA. It would be beneficial to analyze the precision-recall (PR) curve and report the PRAUC metric which could be more robust.

We agree with Reviewer 1 that more statistical metrics should be used to evaluate our model’s performance. We will include a more robust approach, such as PRAUC, to evaluate our model.

Reviewer 2:

Summary:

Zhang et al. present a methodology to model protein-DNA interactions via learning an optimizable energy model, taking into account a representative bound structure for the system and binding data. The methodology is sound and interesting. They apply this model for predicting binding affinity data and binding sites in vivo. However, the manuscript lacks discussion of/comparison with state-of-the-art and evidence of broad applicability. The interpretability aspect is weak, yet over-emphasized.

We appreciate the reviewer’s excellent summary of the paper, and we thank the reviewer for the insightful suggestions and comments.

Strengths:

The manuscript is well organized with good visualizations and is easy to follow. The methodology is discussed in detail. The IDEA energy model seems like an interesting way to study a protein-DNA system in the context of a given structure and binding data. The authors show that an IDEA model trained on one system can be transferred to other structurally similar systems. The authors show good performance in discriminating between binding-vs-decoy sequences for various systems, and binding affinity prediction. The authors also show evidence of the ability to predict genome-wide binding sites.

We appreciate the reviewer’s strong assessment of the strengths of this paper.

Weaknesses:

An energy-based model that needs to be optimized for specific systems is inherently an uncomfortable idea. Is this kind of energy model superior to something like Rosetta-based energy models, which are generally applicable? Or is it superior to family-specific knowledge-based models? It is not clear.

We thank the reviewer for the insightful comments. We will include predictions by generic protein-DNA energy models, such as the Rosetta-based energy model or family-specific knowledge-based model, to compare with our model performance.

Prediction of binding affinity is a well-studied domain and many competitors exist, some of which are well-used. However, no quantitative comparison to such methods is presented. To understand the scope of the presented method, IDEA, the authors should discuss/compare with such methods (e.g. PMID 35606422).

We thank the reviewer for the insightful comments. In our initial submission, Figure S5 presents a comparison between our model’s prediction and those of an existing method using 10-fold cross-validation. We agree a more comprehensive comparison with other methods is needed and will include a discussion and comparison of the IDEA model’s performance with additional state-of-the-art models.

The term “interpretable” has been used lavishly in the manuscript while providing little evidence on the matter. The only evidence shown is the family-specific residue-nucleotide interaction/energy matrix and speculations on how these values are biologically sensible. Recent works already present more biophysical, fine-grained, and sometimes family-independent interpretability (e.g. PMID 39103447, 36656856, 38352411, etc.). The authors should put into context the scope of the interpretability of IDEA among such works.

We agree that “interpretability” should be discussed in a relevant context. We will discuss the scope of IDEA interoperability within the context of recent works, including those suggested by the reviewers.

The manuscript disregards subtle yet important differences in commonly used terminology in the field. For example, the authors use the term ”specificity” and ”affinity” almost interchangeably (for example, the caption for Figure 3A uses ”specificity” although the Methods text describes the prediction as about ”affinity”). If the authors are looking to predict specificity, IDEA needs to be put in the context of the corresponding state-of-the-art (PMID 36123148, 39103447, 38867914, 36124796, etc).

We really appreciate the reviewer for pointing out our conflation of “specificity” and “affinity” in the manuscript. To clarify, IDEA’s primary function is to predict the binding affinities of protein-DNA pairs in a sequence-specific manner. The acquired binding affinities of target DNA sequences can then be used to assess the specific binding motifs. We will revise our text to clarify this point.

It is not clear how much the learned energy model is dependent on the structural model used for a specific system/family. It would be interesting to see the differences in learned model based on different representative PDB structures used. Similarly, the supplementary figures show a lack of discriminative power for proteins like PDX1 (homeodomain family), POU, etc. Can the authors shed some light on why such different performances?

We thank the reviewer for the insightful comments and agree that the familyspecific energy model could provide insight into the model predictions. We will examine different energy models based on the protein family, and especially investigate whether they can explain the lack of discriminative power for certain proteins.

It is also not clear if IDEA’s prediction for reverse complement sequences is the same for a given sequence. If so, how is this property being modelled? Either this description is lacking or I missed it.

We thank the reviewer for the insightful comments. The IDEA model treats reverse complementary sequences separately. We will provide additional details on how these sequences are modeled.

Reviewer 3:

Summary:

Protein-DNA interactions and sequence readout represent a challenging and rapidly evolving field of study. Recognizing the complexity of this task, the authors have developed a compact and elegant model. They have applied well-established approaches to address a difficult problem, effectively enhancing the information extracted from sparse contact maps by integrating artificial sequences decoy set and available experimental data. This has resulted in the creation of a practical tool that can be adapted for use with other proteins.

We appreciate the reviewer’s excellent summary of the paper, and we thank the reviewer for the insightful suggestions and comments.

Strengths:

(1) The authors integrate sparse information with available experimental data to construct a model whose utility extends beyond the limited set of structures used for training. (2) A comprehensive methods section is included, ensuring that the work can be reproduced. Additionally, the authors have shared their model as a GitHub project, reflecting their commitment to transparency of research.

We appreciate the reviewer’s strong assessment of the strengths of this paper.

Weaknesses:

(1) The coarse-graining procedure appears artificial, if not confusing, given that full-atom crystal structures provide more detailed information about residue-residue contacts. While the selection procedure for distance threshold values is explained, the overall motivation for adopting this approach remains unclear. Furthermore, since this model is later employed as an empirical potential for molecular modeling, the use of P and C5 atoms raises concerns, as the interactions in 3SPN are modeled between C_α and the nucleic base, represented by its center of mass rather than P or C5 atoms.

We appreciate the reviewer’s insightful comments. The selection of P and C5 atoms will augment our model prediction, but the prediction is robust without this selection scheme. We will provide more details on the motivation behind this selection.

Regarding the simulation model, we acknowledge a potential disconnection between the coarse-grained level of the 3SPN model (3 coarse-grained sites per nucleotide) and the data-driven model (1 coarse-grained site per nucleotide). The selection of nucleic bases for molecular interactions in the 3SPN model follows the PI’s previous work [PMID: 34057467] and its code implementation. We will test the simulation model by incorporating interactions between Cff and P atoms. In the future, we will work on implementing IDEA model output for 1-bead-per-nucleotide DNA simulation models.

(2) Although the authors use a standard set of metrics to assess model quality and predictive power, some ∆∆G predictions compared to MITOMI-derived ∆∆G values appear nonlinear, which casts doubt on the interpretation of the correlation coefficient.

We thank the reviewer for the insightful comments and agree that the linear fit between our model’s prediction and the experimental data may not be ideal. The primary utility of the IDEA model is for assessing the relative binding affinities of different DNA sequences. To this end, we plan to perform additional statistical analyses that are independent of the linear correlation assumption but instead focus on the ranked order of DNA sequence binding affinities.

(3) The discussion section lacks information about the model’s limitations and a comprehensive comparison with other models. Additionally, differences in model performance across various proteins and their respective predictive powers are not addressed.

We thank the reviewer for the insightful comments and will compare the performance of the IDEA model with state-of-the-art methods. We will also perform detailed analyses of the learned energy models across different proteins and examine their correlation with the model’s predictive powers.

eLife Assessment

This important study measures the functional specialization of distinct subregions within the mouse posterior parietal cortex (PPC) using mesoscopic two-photon calcium imaging during visual discrimination and choice history-dependent tasks. It presents compelling evidence supporting the existence of functional specialized subregions within the PPC. The work will be of interest to system and computational neuroscientists interested in decision-making, working memory, and multisensory integration.

Reviewer #1 (Public review):

Summary:

This study examined the functional organization of the mouse posterior parietal cortex (PPC) using meso-scale two-photon calcium imaging during visually-guided and history-guided tasks. The researchers found distinct functional modules within the medial PPC: area A, which integrates somatosensory and choice information, and area AM, which integrates visual and choice information. Area A also showed a robust representation of choice history and posture. The study further revealed distinct patterns of inter-area correlations for A and AM, suggesting different roles in cortical communication. These findings shed light on the functional architecture of the mouse PPC and its involvement in various sensorimotor and cognitive functions.

Strengths:

Overall, I find this manuscript excellent. It is very clearly written and built up logically. The subject is important, and the data supports the conclusions without overstating implications. Where the manuscript shines the most is the exceptionally thorough analysis of the data. The authors set a high bar for identifying the boundaries of the PPC subareas, where they combine both somatosensory and visual intrinsic imaging. There are many things to compliment the authors on, but one thing that should be applauded in particular is the analysis of the body movements of the mice in the tube. Anyone working with head-fixed mice knows that mice don't sit still but that almost invariable remains unanalyzed. Here the authors show that this indeed explained some of the variance in the data.

Weaknesses:

I see no major weaknesses and I only have minor comments.

Reviewer #2 (Public review):

Summary:

The posterior parietal cortex (PPC) has been identified as an integrator of multiple sensory streams and guides decision-making. Hira et al observe that dissection of the functional specialization of PPC subregions requires simultaneous measurement of neuronal activity throughout these areas. To this end, they use wide-field calcium imaging to capture the activity of thousands of neurons across the PPC and surrounding areas. They begin by delineating the boundaries between the primary sensory and higher visual areas using intrinsic imaging and validate their mapping using calcium imaging. They then conduct imaging during a visually guided task to identify neurons that respond selectively to visual stimuli or choices. They find that vision and choice neurons intermingle primarily in the anterior medial (AM) area, and that AM uniquely encodes information regarding both the visual stimulus and the previous choice, positioning AM as the main site of integration of behavioral and visual information for this task.

Strengths:

There is an enormous amount of data and results reveal very interesting relationships between stimulus and choice coding across areas and how network dynamics relate to task coding.

Weaknesses:

The enormity of the data and the complexity of the analysis make the manuscript hard to follow. Sometimes it reads like a laundry list of results as opposed to a cohesive story.

Reviewer #3 (Public review):

Summary:

This work from Hira et al leverages mesoscopic 2-photon imaging to study large neural populations in different higher visual areas, in particular areas A and AM of the parietal cortex. The focus of the study is to obtain a better understanding of the representation of different task-related parameters, such as choice formation and short-term history, as well as visual responses in large neural populations across different cortical regions to obtain a better understanding of the functional specialization of neural populations in each region as well as the interaction of neural populations across regions. The authors image a large number of neurons in animals that either perform visual discrimination or a history-dependent task to test how task demands affect neural responses and population dynamics. Furthermore, by including a behavioral perturbation of animal posture they aim to dissociate the neural representation of history signals from body posture. Lastly, they relate their functional findings to anatomical data from the Allen connectivity atlas and show a strong relation between functional correlations on anatomical connectivity patterns.

Strengths:

Overall, the study is very well done and tackles a problem that should be of high interest to the field by aiming to obtain a better understanding of the function and spatial structure of different regions in the parietal cortex. The experimental approach and analyses are sound and of high quality and the main conclusions are well supported by the results. Aside from the detailed analyses, a particular strength is the additional experimental perturbation of posture to isolate history-related activity which supports the conclusion that both posture and history signals are represented in different neurons within the same region.

Weaknesses:

The main point that I found hard to understand was the fairly strong language on functional clusters of neurons while also stating that neurons encoded combinations of different types of information and leveraging the encoding model to dissociate these contributions. Do the authors find mixed selectivity or rather functional segregation of neural tuning in their data? More details on this and some other points are below.

eLife Assessment

This study is important as it highlighted how IL-4 regulates the reactive state of a specific microglial population by increasing the proportion of CD11c+ microglial cells and ultimately suppressing neuropathic pain. It provided convincing evidence on the pain-resolving roles of microglia. However, the authors are encouraged to clarify data interpretation and integrate the study's findings into the existing knowledge about microglia, monocytes, and macrophages.

Reviewer #1 (Public review):

Summary:

Kohno et al. examined whether the anti-inflammatory cytokine IL-4 attenuates neuropathic pain by promoting the emergence of antinociceptive microglia in the dorsal horn of the spinal cord. In two models of neuropathic pain following peripheral nerve injury, intrathecal administration of IL-4 once a day for 3 days from day 14 to day 17 after injury, attenuates hypersensitivity to mechanical stimuli in the hind paw ipsilateral to nerve injury. Such an antinociceptive effect correlates with a higher number of CD11c+microglia in the dorsal horn of the spinal cord which is the termination area for primary afferent fibres injured in the periphery. Interestingly, CD11c+ microglia emerge spontaneously in the dorsal horn in concomitance with the resolution of pain in the spinal nerve model of pain, but not in the spared nerve injury model where pain does not resolve, confirming that this cluster of microglia is involved in resolution pain.

Based on existing evidence that the receptor for IL-4, namely IL-4R, is expressed by microglia, the authors suggest that IL-4R mediates IL-4 effect in microglia including up-regulation of Igf1 mRNA. They have previously reported that IGF-1 can attenuate pain neuron activity in the spinal cord.

Strengths:

This study includes cutting-edge techniques such as flow cytometry analysis of microglia and transgenic mouse models.

Weaknesses:

The conclusion of this paper is supported by data, but the interpretation of some data requires clarification.

Reviewer #2 (Public review):

Summary:

The authors aimed to investigate how IL-4 modulates the reactive state of microglia in the context of neuropathic pain. Specifically, they sought to determine whether IL-4 drives an increase in CD11c+ microglial cells, a population associated with anti-inflammatory responses and whether this change is linked to the suppression of neuropathic pain. The study employs a combination of behavioral assays, pharmacogenetic manipulation of microglial populations, and characterization of microglial markers to address these questions.

Strengths:

The methodological approach in this study is robust, providing convincing evidence for the proposed mechanism of IL-4-mediated microglial regulation in neuropathic pain. The experimental design is well thought out, utilizing two distinct neuropathic pain models (SpNT and SNI), each yielding different outcomes. The SpNT model demonstrates spontaneous pain remission and an increase in the CD11c+ microglial population, which correlates with pain suppression. In contrast, the SNI model, which does not show spontaneous pain remission, lacks a significant increase in CD11c+ microglia, underscoring the specificity of the observed phenomenon. This design effectively highlights the role of the CD11c+ microglial population in pain modulation. The use of behavioral tests provides a clear functional assessment of IL-4 manipulation, and pharmacogenetic tools allow for precise control of microglial populations, minimizing off-target effects. Notably, the manipulation targets the CD11c promoter, which presumably reduces the risk of non-specific ablation of other microglial populations, strengthening the experimental precision. Moreover, the thorough characterization of microglial markers adds depth to the analysis, ensuring that the changes in microglial populations are accurately linked to the behavioral outcomes.

Weaknesses:

One potential limitation of the study is that the mechanistic details of how IL-4 induces the observed shift in microglial populations are not fully explored. While the study demonstrates a correlation between IL-4 and CD11c+ microglial cells, a deeper investigation into the specific signaling pathways and molecular processes driving this population shift would greatly strengthen the conclusions. Additionally, the paper does not clearly integrate the findings into the broader context of microglial reactive state regulation in neuropathic pain.

eLife Assessment

This important study shows that the activity of hypothalamic hypocretin/orexin neurons (HONs) correlates with body movement over multiple behaviors. Sophisticated techniques and analyses showcase this link which appears to be unique to HONs. Evidence for this correlation is, however, incomplete as the confound of arousal with movement needs to be resolved since orexin also plays a key role in arousal. This work should be of interest to scientists studying peptidergic neurons, movement, energy regulation, and brain-body coordination.

Reviewer #1 (Public review):

Summary:

This manuscript by Tesmer and colleagues uses fiber photometry recordings, sophisticated analysis of movement, and deep learning algorithms to provide compelling evidence that activity in hypothalamic hypocretin/orexin neurons (HONs) correlates with net body movement over multiple behaviors. By examining projection targets, the authors show that hypocretin/orexin release differs in projection targets to the locus coeruleus and substantia nigra, pars compacta. Ablation of HONs does not cause differences in the power spectra of movements. The movement-tracking ability of HONs is independent of HON activity that correlates with blood glucose levels. Finally, the authors show that body movement is not encoded to the same extent in other neural populations.

Strengths:

The major strengths of the study are the combination of fiber photometry recordings, analysis of movement in head-fixed mice, and sophisticated classification of movement using deep learning algorithms. The experiments seem to be well performed, and the data are well presented, visually. The data support the main conclusions of the manuscript.

Weaknesses:

The weaknesses are minor, mostly consisting of writing and data visualization throughout the manuscript. To some degree, it is already known that hypocretin/orexin neurons correlate with movement and arousal, although this manuscript studies this correlation with unprecedented sophistication and scale. It is also unfortunate that most of the experiments throughout the study were only performed in male mice.

Taken together, this study is likely to be impactful to the field and our understanding of HONs across behavioral states.

Reviewer #2 (Public review):

Summary:

Despite several methodological strengths, the major and highly significant drawback is the confound of arousal with movement. This confound is not resolved, so the results could be explained by previously established relationships between orexin and arousal/wakefulness.

Strengths:

The authors show that orexin neuron activity is associated with body movement and that this information is conveyed irrespective of the fasted state. They also report differences in different orexin target brain regions for orexin release during movement.

This paper contains an impressive array of cutting-edge techniques to examine a very important brain system, the orexin-hypocretin system. The authors offer an original perspective on the function of this system. The authors showed that orexin neuron activity scales to some degree with the magnitude of body movement change; this is unaffected by a fasted state and seems to be somewhat unique to orexin neurons.

The investigation of other genetically-defined subcortical neuron populations to determine the specificity of findings is also a strength, as is the ability to quantify movement and use deep learning to classify specific behaviors adds sophistication to analysis. The authors also show heterogeneity in orexin projections to specific target nuclei, which is interesting.

The authors "speculate that narcolepsy-cataplexy, caused by HON loss-of-function, is perhaps explained by oscillations into unwanted sleep-states and motor programs due to impaired control loops for wakefulness and movement". This is quite an interesting aspect of their work, and deserving of further study.

Weaknesses:

Despite the strengths, there are several major and minor weaknesses that detract significantly from the study.

Weaknesses - Major

My main concern with this work is the confound of arousal with movement so that correlations with one might reflect a relationship instead with the other. The orexin system is well known to play an important role in arousal, with elevated activity of orexin neurons reported for waking and high arousal. Orexin signaling has also been strongly associated with motivation, which also is associated with arousal and movement. The authors offer no compelling evidence that the relationships they describe between different movements and orexin signaling do not simply reflect the known relationship between arousal and motivation.

The authors could address this concern by including classical arousal measurements, eg, cortical EEG recorded simultaneously with movements. Often, EEG arousal occurs independently of movement, so this could provide one approach to disentangling this confound. The idea that orexin signaling plays a role in arousal rather than movement is supported by their finding that orexin lesions using the orexin-DTR mouse model did not impact movements. In contrast, prior lesion and pharmacologic studies have found that decreased orexin signaling significantly decreases arousal and waking.

Another way they could test their idea would be to paralyze and respirate animals so that orexin activity could be recorded without movement. Alternatively, animals could be trained to remain motionless to receive a reward. Thus, there are several ways to test the overall hypothesis of this work that have not been examined here.

The authors propose that "a simple interpretation of their results is that, via HON movement tracking, the brain creates a "wake up" signal in proportion to movement". This seems to argue for the role of the orexin system in arousal and motivation rather than in movement per se.

There are several studies that have examined the effect of orexin antagonist treatment in rodents on locomotor and other motor activities. These studies have largely found no consistent effect of antagonizing orexin signaling, especially at the OxR1 receptor, on simple motor activity. These studies are not referenced here but should be taken into account in the authors' conclusions.

Figure 3, panel F: I understand HON-DTR is a validated model but a picture of HONs ablation is necessary, including pictures of HONs outputs ablation within the SNc and LC.

The discussion lacks a more extensive paragraph on the distinct signal and role of Ox->SNc and Ox-LC projections.

Reviewer #3 (Public review):

Summary

The study presents an investigation into how hypothalamic orexin neurons (HONs) track body movement with high precision. Using techniques including fiber photometry, video-based movement metrics, and empirical mode decomposition (EMD), the authors demonstrate that HONs encode net body movement consistently across a range of behaviors and metabolic states. They test the ability of HONs to track body movement to that of other subcortical neural populations, from which they distinguish HONs activity from other subcortical neural populations.

Strengths:

The study characterizes HONs activity as key indicators of movement and arousal, and this method may have potential implications for understanding sleep disorders, energy regulation, and brain-body coordination. Overall, I think this is a very interesting story, with novel findings and implications about sensorimotor systems in animals. The manuscript is clearly written and the evidence presented is rigorous. The conclusions are well supported by experimental data with clear statistical analyses.

Weaknesses/suggestions:

There are a couple of issues I think the authors could address to make the paper better and more complete:

(1) The study primarily focuses on steady-state behaviors. It would be interesting if the authors' current dataset allows analyses of HON dynamics during transitions between behavioral states (e.g., resting to running or grooming to sniffing). This could provide additional insights into how HONs adapt to rapid changes in body movement.

(2) Given the established role of HONs in arousal and wakefulness, the study could further investigate how movement-related HON dynamics interact with arousal states. For example, does HON encoding of movement differ during sleep versus wakefulness?

(3) Although HON ablation experiments suggest that HONs do not shape movement frequency profiles. It would be more compelling if the authors could investigate whether HONs contribute to specific types of movements (e.g., fine motor vs. gross motor movements) or modulate movement initiation thresholds.

(4) The heterogeneous movement-related orexin dynamics observed in the LC and SNc raise intriguing questions about the circuit-level mechanisms underlying these differences. Optogenetic or chemogenetic manipulation of these projections could validate the functional implications of these dynamics.

eLife Assessment

This study presents a valuable finding that C238 in vimentin regulates long non-coding RNA XIST to suppress EMT and thereby Xist may be a therapeutic target in breast cancer. The evidence supporting the claims of the authors is solid, although the improvement of data visibility and presentation would have strengthened the study. The work will be of interest to scientists working in the field of BCs.

Reviewer #1 (Public review):

Summary, and Strengths:

The authors and their team have investigated the role of Vimentin Cysteine 328 in epithelial-mesenchymal transition (EMT) and tumorigenesis. Vimentin is a type III intermediate filament, and cysteine 328 is a crucial site for interactions between vimentin and actin. These interactions can significantly influence cell movement, proliferation, and invasion. The team has specifically examined how Vimentin Cysteine 328 affects cancer cell proliferation, the acquisition of stemness markers, and the upregulation of the non-coding RNA XIST. Additionally, functional assays were conducted using both wild-type (WT) and Vimentin Cysteine 328 mutant cells to demonstrate their effects on invasion, EMT, and cancer progression. Overall, the data supports the essential role of Vimentin Cysteine 328 in regulating EMT, cancer stemness, and tumor progression. Overall, the data and its interpretation are on point and support the hypothesis. I believe the manuscript has great potential.

Weaknesses:

Minor issues are related to the visibility and data representation in Figures 2E and 3 A-F.

Reviewer #2 (Public review):

The aim of the investigation was to find out more about the mechanism(s) by which the structural protein vimentin can facilitate the epithelial-mesenchymal transition in breast cancer cells.

The authors focussed on a key amino acid of vimentin, C238, its role in the interaction between vimentin and actin microfilaments, and the downstream molecular and cellular consequences. They model the binding between vimentin and actin in silico to demonstrate the potential involvement of C238, but the outcome is described vaguely. The phenotype of a non-metastatic breast cancer cell line MCF7, which doesn't express vimentin, could be changed to a metastatic phenotype when mutant C238S vimentin, but not wild-type vimentin, was expressed in the cells. Expression of vimentin was confirmed at the level of mRNA, protein, and microscopically. Patterns of expression of vimentin and actin reflected the distinct morphology of the two cell lines. Phenotypic changes were assessed through assay of cell adhesion, proliferation, migration, and morphology and were consistent with greater metastatic potential in the C238S MCF7 cells. Changes in the transcriptome of MCF7 cells expressing wild-type and C238S vimentins were compared and expression of Xist long ncRNA was found to be the transcript most markedly increased in the metastatic cells expressing C238S vimentin. Moreover changes in expression of many other genes in the C238S cells are consistent with an epithelial mesenchymal transition. Tumourigenic potential of MCF7 cells carrying C238S but not wild-type, vimentin was confirmed by inoculation of cells into nude mice. This assay is a measure of the stem-cell quality of the cells and not a measure of metastasis. It does demonstrate phenotypic changes that could be linked to metastasis.

shRNA was used to down-regulate vimentin or Xist in the MCF7 C238S cells. The description of the data is limited in parts and data sets require careful scrutiny to understand the full picture. Down-regulation of vimentin reversed the morphological changes to some degree, but down-regulation of Xist didn't. Conversely, down-regulation of Xist inhibited cell growth, a sign of reversing metastatic potential, but down-regulation of vimentin had no effect on growth. Down-regulation of either did inhibit cell migration, another sign of metastatic reversal. The interpretation of this type of experiment is handicapped when full reversal of expression is not achieved, as was the case in this study.

Overall the study describes an intriguing model of metastasis that is worthy of further investigation, especially at the molecular level to unravel the connection between vimentin and metastasis. The identification of a potential role for Xist in metastasis, beyond its normal role in female cells to inactivate one of the X chromosomes, corroborates the work of others demonstrating increased levels in a variety of tumours in women and even in some tumours in men. It would be of great interest to see where in metastatic cells Xist is expressed and what it binds to.

eLife Assessment

This is an important study that provides CCR7-APEX2 proximity labelling mass spectrometry data that is expected to provide new insights into CCR7 signaling partners and pathways. The study is technically easy to follow and the data is convincing. It will be interesting in the future to have complementary studies in lymphocytes/dendritic cells that endogenously express CCR7. This is of value to the community, and there are likely multiple opportunities to use the APEX2 data set to extend these findings, strengthen some claims, and even explore a new pathway identified in the APEX2 data set.

Reviewer #1 (Public review):

Summary:

Hahn et al use bystander BRET, NanoBiT assays and APEX2 proteomics to investigate endosomal signaling of CCR7 by two agonists, CCL19 and CCL21. The authors suggest that CCR7 signals from early endosomes following internalisation. They use spatial proteomics to try to identify novel interacting partners that may facilitate this signaling and use this data to specifically enhance a Rac1 signaling pathway. The most novel findings are the APEX2 proteomics studies that provide new mechanisms.

Strengths:

(1) The APEX2 resource will be valuable to the GPCR and immunology community. It offers many opportunities to follow up on findings and discover new biology. The authors have used the resource to validate earlier findings in the current manuscript and in previous manuscripts.

(2) The results section is well written and can be followed very easily by the reader.

(3) Some findings verify previous studies (e.g. endomembrane signalling).

Weaknesses:

(1) The findings are interesting although the studies are almost all performed in HEK293 cells. I understand that these are commonly used in GPCR biology and current tools need to be improved in order to perform similar analyses in more relevant cell-lines. Future studies should focus on validating the findings of the current study in physiologically-relevant cell-lines.

Reviewer #2 (Public review):

Summary:

This manuscript describes a comprehensive analysis of signalling downstream of the chemokine receptor CCR7. A comprehensive dataset supports the authors' hypothesis that G protein and beta arrestin signalling can occur simultaneously at CCR7 with implications for continued signalling following receptor endocytosis.

Strengths:

The experiments are well controlled and executed, employing a wide range of assay, using in the main, CCR7 transfectants. Data are well presented, with the authors claims supported by the data. The paper also has an excellent narrative which makes it relatively easy to follow. I think this would certainly be of interest to the readership of the journal.

Weaknesses:

The experiments are currently representative of signalling events in HEK293 transfectants and await verification in more relevant systems e.g. T-cells and dendritic cells.

Appraisal and Discussion

Overall, the authors appear to have achieved their experimental aims and provide substantial evidence that chemokine receptors can stimulate G proteins from within endosomes to regulate signalling pathways involved in cell migration. This builds upon earlier studies from the Legler group which showed that endocytosed CCR7 could activate Rac1 and influence lamellipodia formation. An unbiased mass spectrometry-based proteome profiling approach was used by the authors of this study to identify several candidate proteins which appear to play a role in receptor trafficking and signalling downstream of CCR7. These data may provide clues as to how other chemokine receptors are regulated post endocytosis in various leukocyte subsets.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Hahn et al use bystander BRET, NanoBiT assays, and APEX2 proteomics to investigate endosomal signaling of CCR7 by two agonists, CCL19 and CCL21. The authors suggest that CCR7 signals from early endosomes following internalisation. They use spatial proteomics to try to identify novel interacting partners that may facilitate this signaling and use this data to specifically enhance a Rac1 signaling pathway. Many of the results in the first few figures showing simultaneous recruitment of Barr and G proteins by CCR7 have been shown previously (Laufer et al, 2019, Cell Reports), as has signaling from endomembranes, and Rac1 activation at intracellular sites. The new findings are the APEX2 proteomics studies, which could be useful to the scientific community. Unfortunately, the authors only follow up on a single finding, and the expansion of this section would improve the manuscript.

First of all, we would like to thank the reviewer for helping with the manuscript. The summary is mostly accurate except for the statement that simultaneous recruitment of barr and G protein to CCR7 has been shown before. It should also be noted that it has not been demonstrated that CCR7 activates G proteins from endosomes previously nor has the functional role of this signaling mechanism. However, that CCR7 activity at endomembranes is associated with Rac1 signaling was demonstrated in the Laufer et al. study as the reviewer correctly points out.

Strengths:

(1) The APEX2 resource will be valuable to the GPCR and immunology community. It offers many opportunities to follow up on findings and discover new biology. The resource could also be used to validate earlier findings in the current manuscript and in previous manuscripts. Was there enrichment of early endosomal markers, Barr and Gi as this would provide further evidence for their earlier claims regarding endosomal signaling? Previous studies have suggested signaling from the TGN, so it is possible that the different ligands also direct to different sites. This could easily be investigated using the APEX2 data.

Thank you for your comment. We do in fact observe enrichment of TGN/Golgi markers in response to chemokine stimulation, which we now have highlighted in the manuscript (fourth paragraph on page 7).

(2) The results section is well written and can be followed very easily by the reader.

We are glad that the reviewer found the results section very readable.

(3) Some findings verify previous studies (e.g. endomembrane signalling). This should be acknowledged as this shows the validity of the findings of both studies.

This is correct. We have now included more discussion of previous work related to CCR7 signaling at endomembranes (thirdparagraph on page 10).

Weaknesses:

(1) The findings are interesting although the studies are almost all performed in HEK293 cells. I understand that these are commonly used in GPCR biology and are easy to transfect and don't express many GPCRs at high concentrations, but their use is still odd when there are many cell-lines available that express CCR7 and are more reflective of the endogenous state (e.g. they are polarised, they can perform chemotaxis/ migration). Some of the findings within the study should also be verified in more physiologically relevant cells. At the moment only the final figure looks at this, but findings need to be verified elsewhere.

We thank the reviewer for raising this point and giving us an opportunity to elaborate in further detail. The major goal of our study was to investigate whether CCR7 activates G protein from endosomes, the underlying mechanism, and functions of this potential signaling mechanism. The reason we chose CCR7 as our model receptor was that it belongs to a group of GPCRs, the chemokine receptors, that most often have features associated with the ability to promote endosomal G protein activation (phosphorylation site clusters in the C-terminal region).

Specific detection of G protein activation at distinct subcellular compartments is currently very challenging in truly endogenous systems despite new innovative biosensors that are available (not just related to CCR7, but GPCRs in general). To our knowledge, most if not all studies that detect direct activation of G protein at a specific compartment whether at the plasma membrane, endosome, Golgi, or other compartments, have overexpressed either the receptor, G protein, or both. This is why we choose the HEK293 cell system for most of our experiments, which are easy to manipulate. That being said, we did confirm major findings in an indirect manner using Jurkat T-cells, which express CCR7 endogenously and are physiological relevant. Our hope is that in the future we will be able to use highly sensitive biosensors to directly confirm our findings in such a cell system as the reviewer wisely suggests.

(2) The authors acknowledge that the kinetic patterns of the signals at the early endosome are not consistent with the rates of internalisation. They mention that this could be due to trafficking elsewhere. This could be easily looked at in their APEX2 data. Is there evidence of proximity to markers of other membranes? Perhaps this could be added to the discussion. Similarly, previous studies have shown that CCR7 signaling may involve the TGN. Was there enrichment of these markers? If not, this could also be an interesting finding and should be discussed. It is also possible that the Rab5 reporter is just not as efficient as the trafficking one, especially as in later figures the very convincing differences in the two ligands are not as robust as the differences in trafficking.

Excellent point. We have now highlighted the possibility of CCR7 being further trafficked to the trans-Golgi network (TGN) as possible explanation for the transient translocation of activated CCR7 to the early endosome in Fig. 1G-H (second paragraph on page 3).

Furthermore, in the APEX2 experiment we observe enrichment of proteins involved in lysosomal trafficking (LAMP1, VPS16, VAMP7, WDR91, and PP4P1) by CCL19 stimulation at 25 min, and recycling endosomes/TGN markers (SNX6, RAB7L, and GGA) by CCL21 stimulation at 25 min. In addition to this, several markers of TGN/Golgi (SNX3, COG5, YIF1A, SC22B, and AP3S1) were enriched as well in response to both CCL19 and CCL21 stimulation. We have now included a statement in the manuscript, which describes the likely trafficking of CCR7 to the TGN/Golgi in response to CCL19 and CCL21 stimulation (fourth paragraph on page 7).

(3) In the final sentence of paragraph 2 of the results the authors state that the internalisation is specific to CCR7 as there isn't recruitment to V2R. I'm not sure this is the best control. The authors can only really say it doesn't recruit to unrelated receptors. The authors could have used a different chemokine receptor which does not respond to these ligands to show this.

The point with this control experiment was to demonstrate that the loss of NanoBiT signal in response to CCL19 in CCR7-SmBiT/LgBiT-CAAX expressing cells, but not in V2R-SmBiT/LgBiT-CAAX expressing cells, was a result of bona fide CCR7 internalization rather than potential artifactual effects of CCL19 on the NanoBiT system. Our intent was not to demonstrate specificity of CCL19 among chemokine receptors, which already has been thoroughly tested in previous studies. We have now modified the sentence (second paragraph on page 3) “Moreover, CCL19/CCL21-stimulation of receptor internalization to endosomes is specific to CCR7 as none of the chemokines promote internalization or trafficking to endosomes of the vasopressin type 2 receptor (V₂R)-SmBiT construct (Fig. S1E-F)” to “Moreover, CCL19/CCL21-stimulation did not promote internalization or trafficking to endosomes of the vasopressin type 2 receptor (V₂R)-SmBiT construct, which validates that these chemokines act specifically via the CCR7-SmBiT system (Fig. S1E-F).”

(4) The miniGi-Barr1 and imaging showing co-localisation could be more convincing if it was also repeated in a more physiological cell line as in the final figure. Imaging of CCR7, miniGi, and Barr1 would also provide further evidence that the receptor is also present within the complex.

We agree with the reviewer’s assessment. However, as mentioned above it is currently extremely challenging to detect endogenous G protein coupling/activation to endogenous receptors. In addition, we are not sure if overexpressing fluorophore-tagged receptor, miniG, and barr1 in a physiological-relevant cell line would provide truly physiological conditions as the expression of these proteins still would be artificially high. This is why we chose to conduct these mechanistic experiments in HEK293 cells and then indirectly verify key findings in an endogenous and physiological-relevant cell line.

(5) The findings regarding Rac1 are interesting, although an earlier paper found similar results (Laufer et al, 2019, Cell Reports), so perhaps following up on another APEX2-identified protein pathway would have been more interesting. The authors' statement that Rac1 is specifically activated, and RhoA and Cdc42 are not, is unconvincing from the current data. Only a single NanoBiT assay was used, and as raw values are not reported it is difficult for the reader to glean some essential information. The authors should show evidence that these reporters work well for other receptors (or cite previous studies) and also need evidence from an independent (i.e. non-NanoBiT or BRET) assay.

The major focus of the study was to investigate whether CCR7 can activate G protein after having been internalized into endosomes via formation of CCR7-Gi/o-barr megaplexes, and to dissect out potential functions of said endosomal G protein signaling. To do this, we used CCL19 and CCL21 which stimulate G protein to the same extent but differ in their ability of promote barr recruitment and receptor internalization with CCL19 being superior to CCL21. To this end, we found that CCL19 also promote endosomal G protein activation to a greater extent than CCL21, and therefore, we specifically looked for proteins enriched by CCL19 in our APEX experiment. This led us to some Rho GTPase regulators that were differentially enriched by CCL19 and CCL21. We agree that there were other interesting effectors related to CCR7 biology identified in the APEX experiment such as EYA2, GRIP2, and EI24. However, those proteins were enriched similar by CCL19 and CCL21 challenge, and thus, do not seem to be activated specifically at endosomes. Following the same argument, we also did not observe any difference in the activity of RhoA or Cdc42 when stimulated with CCL19 or CCL21, so we cannot conclude that these signaling proteins are activated specifically in endosomes. On the other hand, Rac1 was stimulated to a larger degree by CCL19 than CCL21, its activity was inhibited by the Gi/o inhibitor PTX and endocytosis inhibitors Dyngo-4a and PitStop2. CCR7-mediated Rac1 signaling was also inhibited by expression of a dominant negative dynamin mutant that inhibits receptor internalization, and Rac1 was not activated by an internalization-deficient CCR7-DS/T mutant. Finally, the involvement of Rac1 in CCR7 mediated chemotaxis of Jurkat T cells was also demonstrated. We believe that these findings together provide strong basis for the claim that endosomal Gi/o protein signaling by CCR7 activates Rac1.

Following the reviewer’s suggestion, we have now included experiments to show that the activation of RhoA, Rac1, and Cdc42 by CXCR4 also can be detected by the NanoBiT biosensors (Fig. S7D-F). We have also added the appropriate references to the original studies where these biosensors were developed in the results section (first paragraph on page 8).

(6) At present, the studies in Figure 7 do not go beyond those in the previous Laufer et al study in which they showed blocking endocytosis affected Rac1 signalling. The authors could show that Rac1 signalling is from early endosomes to improve this, otherwise, it could be from the TGN as previously reported.

The major purpose of Figure 7 was to indirectly confirm findings from HEK293 cells experiments and to tie them to physiological functions. Our experiments using Jurkat T-cells show that CCL19 promote stronger chemotactic response than CCL21 despite similar Gi/o response. In addition, we showed that CCR7-mediated Gi/o activation, receptor endocytosis, as well as Rac1 activity, are required to drive chemotaxis. The Laufer et al. study did not investigate whether CCR7 activates G protein after having been internalized into endosomes via formation of CCR7-Gi/o-barr megaplexes, and thus, did not focus on functional outcomes of this signaling mechanism. Based on this, we believe our work provides new and valuable knowledge to the field.

Reviewer #2 (Public Review):

Summary:

This manuscript describes a comprehensive analysis of signalling downstream of the chemokine receptor CCR7. A comprehensive dataset supports the authors' hypothesis that G protein and beta-arrestin signalling can occur simultaneously at CCR7 with implications for continued signalling following receptor endocytosis.

We would like to thank the reviewer for helping with the manuscript. We agree on all points made and have now updated the manuscript accordingly.

Strengths:

The experiments are well controlled and executed, employing a wide range of assays using - in the main - CCR7 transfectants. Data are well presented, with the authors' claims supported by the data. The paper also has an excellent narrative which makes it relatively easy to follow. I think this would certainly be of interest to the readership of the journal.

We appreciate the positive assessment of strengths.

Weaknesses:

Since the authors show a differential enrichment of RhoGTPases by CCR7 stimulation with CCL19 versus CCL21, I think that they also need to show that the Gi/o coupling of HEK-292-CCR7-APEX2 cells to both CCL19 and CCL21 is not perturbed by the modification. Currently, the authors only show data for CCL19 signalling, which leaves the potential for a false negative finding in terms of CCL21 signalling being selectively impaired. This should be relatively easy to do and should strengthen the authors' conclusions.

We agree with the reviewer and have now included experiments to show that both CCL19- and CCL21-mediated CCR7-APEX2 stimulation leads to Gi/o activation (Fig. S4C). In addition, our proteomics experiments show strong effects of both CCL19 and CCL21 stimulation, which suggest that the receptor is activated by both ligands.

The authors conclude the discussion by suggesting that their findings highlight endosomal signalling as a general mechanism for chemokine receptors in cell migration. I think this is an overreach. The authors chose several studies of CXC chemokine receptors to support their argument that C-terminal truncation or mutation of the C-terminal phosphorylation sites impairs endocytosis and chemotaxis (refs 40-42). However, in some instances e.g. at the related chemokine receptor CCR4, C-terminal removal of these sites impairs endocytosis but promotes chemotaxis (Nakagawa et al, 2014); Anderson et al, 2020). I therefore think that either the final statement needs to be tempered down or the counterargument discussed a little.

We appreciate the reviewer highlighting this point. We have now modified the concluding sentence from “Thus, the findings from our study highlight endosomal G protein signaling by chemokine receptors as a potential general mechanism that regulates key aspects of cell migration” to “Thus, the findings from our study highlight endosomal G protein signaling by some chemokine receptors as a potential mechanism that regulates key aspects of cell migration.” We hope that the temper level of this sentence is more appropriate.

References:

Anderson, C. A. et al. A degradatory fate for CCR4 suggests a primary role in Th2 inflammation. J Leukocyte Biol 107, 455-466 (2020).

Nakagawa, M. et al. Gain-of-function CCR4 mutations in adult T cell leukaemia/lymphoma. Journal of Experimental Medicine 211, 2497-2505 (2014).

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) The results section is well written, although the introduction needs more information on what is known about CCR7 trafficking and endomembrane signaling. I understand this is because the authors wanted to focus on GPCR signaling, but the study will equally be of interest to researchers in the immunology and chemokine fields, and therefore more CCR7-focussed discussion in the introduction would be useful. Similarly, the discussion would benefit from more discussion of previous studies of CCR7 trafficking and endomembrane signaling (in particular the Laufer et al paper) to acknowledge that many of the findings within this paper verify previous studies.

We have now included additional immunology/endomembrane background information about CCR7 at the place where the receptor is introduced (first paragraph on page 3). We have also expanded our discussion of our work in relation to the Laufer et al. study (third paragraph on page 10).

(2) On page 5, the authors state that 'The response to chemokine stimulation was not observed in mock transfected HEK293 cells'. Figure S4D does not have a legend so it is difficult to see what they mean by mock transfected. Do they mean not transfecting with anything or not with the receptor? The better control would be transfecting the reporters but not the receptor. This may have been done, but the wording needs clarifying and S4D needs a legend.

Thanks for pointing this out. We believe the reviewer refers to Figure S2D and we have now highlighted/clarified the legend better. Mock transfected conditions refer to HEK293 cells transfected with the reporter, but not the receptor. This is written in the legend as “(D) Change in luminescence signal generated between SmBiT-barr1 and LgBiT-miniGi in response to 100 nM CCL19 or 100 nM CCL21 in mock transfected HEK293 cells (no CCR7)”, which we believe should be clear to the audience.

(3) The validation of the APEX2 receptor construct relies on a single assay with one ligand. The authors should show that the receptor expresses at the cell surface, is internalised normally, and that both ligands activate the receptor.

We have now included additional data to show that (1) the receptor is expressed at the cell surface, (2) that the CCR7-APEX2 recruits barr1 to the plasma membrane, (3) that this association leads to barr1 translocation to the early endosomes as an indirect measurement of receptor internalization, and (4) that both CCL19- and CCL21-stimulation inhibit forskolin induced cAMP production (Fig.S4A-C, and described in fifth paragraph on page 6).

(4) The APEX2 section is very short, especially as this is novel data. It lacks some important information, e.g. when the authors state that 'we identified a total of 579 proteins', is this in total for both ligands, separately or were some shared? More information on each ligand separately and combined would make this clearer.

We have now specified that the identified total proteins enriched from our APEX2 approach is when the cells are stimulated with either CCL19 or CCL21 (third paragraph on page 7). Furthermore, we have included a Venn diagram in Fig. S5C to show how many proteins were enriched by CCL19 or CCL21 stimulation and how many of those were shared at different time points.

(5) The discussion would benefit from some further work. The current first two paragraphs just reiterate the introduction and don't discuss the current paper so could be removed completely. The Laufer et al study needs much more discussion as they report many of the findings of the current paper (signaling following endocytosis, Rac1 endomembrane signaling) five years ago. The APEX2 findings that are discussed, though interesting, are not followed up by further experimental evidence and there is little discussion of why the two ligands have different responses or what the physiological effects could be.

We appreciate the reviewer’s effort in helping with the discussion. To this end, we have now expanded our discussion of the mentioned paper further as suggested (third paragraph on page 10). We agree that the findings from our APEX experiment are interesting, but the focus of this study relates to proteins enriched specifically at endosomes. Several of the most enriched proteins did not show this localization bias, which is why these proteins were not further investigated.

Minor changes:

(1) The authors should remove the word 'recent' at the start of the first sentence of the third paragraph. Endosomal signaling by GPCRs was described 15 years ago so cannot really be seen as recent anymore.

We have now adjusted the manuscript accordingly.

(2) Tukey defaulted to Turkey in some places.

We thank the reviewer for pointing out these typos, which now have been corrected.

Reviewer #2 (Recommendations For The Authors):

Minor Points:

(1) ACKRs do not couple to G proteins so it is peculiar to see them in this table. I would limit the table to the conventional CCR1-10, CXCR1-6 and XCR1. The ligand for XCR1 is XCL1 which is absent from the table.

We have now modified the table accordingly.

(2) CCL19 (formerly known as ELC) has been long known to be a more efficacious and potent ligand in chemotaxis assays (Bardi et al, 2001). This earlier reference should be added to the citations in the preceding statement on page 10.

This is an important study showing that CCL19 is more efficacious than CCL21 in promoting chemotaxis and that this has been known for decades. We have now included the reference accordingly (reference 59 in second paragraph on page 11).

(3) Figure 6, Panel Q. I think the legends for CCR7 and CCR7 delta ST might be flipped.

We thank the reviewer for pointing out this error. We have now corrected the figure panel.

(4) Figure S5 (or 5) might benefit from simple Venn diagrams showing the numbers of differentially enriched proteins following treatment with the two ligands at different time points.

We have included a Venn diagram in Fig. S5C to show how many proteins were enriched by CCL19 or CCL21 stimulation and how many of those where shared.

Reference:

Bardi, G., Lipp, M., Baggiolini, M. & Loetscher, P. The T cell chemokine receptor CCR7 is internalized on stimulation with ELC, but not with SLC. European Journal of Immunology 31, 3291-3297 (2001).

eLife Assessment

Working with a diverse panel of rice accessions grown in field conditions, this valuable study measures changes in transcript abundance, tests for patterns of selection on gene expression, and maps the genetic basic of variation in gene expression in normal and elevated salinity treatments. The manuscript provides solid evidence that mean gene expression levels are further from the optimum abundance for more genes under the elevated salinity treatment compared to normal treatment, and that a relatively small number of genes are hotspots that harbor genetic variants which affect broader genome-wide patterns of natural variation in gene expression under high salinity conditions. However, the design, clarity, and interpretation of several statistical analyses can be improved, some opportunities for integration among datasets and analyses could yet be realized, and genetic manipulation is required to confirm functional involvement of any specific genes in regulatory networks or organismal traits that confer adaptation to higher salinity conditions. The manuscript will be of interest to evolutionary biologists studying the genetics of complex traits and a resource for plant biologists studying mechanisms of abiotic stress tolerance.

Reviewer #2 (Public review):

The authors investigate the gene expression variation in a rice diversity panel under normal and saline growth conditions to gain insight into the underlying molecular adaptive response to salinity. They present a convincing case to demonstrate that environment stress can induce selective pressure on gene expression, which is in agreement with their earlier study (Groen et al, 2020). The data seems to be a good fit for their study and overall the analytic approach is robust.

(1) The work started by investigating the effect of genotype and their interaction at each transcript level using 3'-end-biased mRNA sequencing, and detect a wide-spread GXE effect. Later, using the total filled grain number as a proxy of fitness, they estimated the strength of selection on each transcript and reported stronger selective pressure in saline environment. However, this current framework rely on precise estimation of fitness and, therefore can be sensitive to the choice of fitness proxy.

(2) Furthermore, the authors decomposed the genetic architecture of expression variation into cis- and trans-eQTL in each environment separately and reported more unique environment specific trans-eQTLs than cis-. The relative contribution of cis- and trans-eQTL depends on both the abundance and effect size. I wonder why the latter was not reported while comparing these two different genetic architectures. If the authors were to compare the variation explained by these two categories of eQTL instead of their frequency, would the inference that trans-eQTLs are primarily associated with expression variation still hold?

(3) Next, the authors investigated the relationship between cis- and trans-eQTLs at transcript level and revealed an excess of reinforcement over compensation pattern. Here, I struggle to understand the motivation for testing the relationship by comparing the effect of cis-QTL with the mean effect of all trans-eQTLs of a given transcript. My concern is that taking the mean can diminish the effect of small trans-eQTLs potentially biasing the relationship towards the large-effect eQTLs.

Comments on latest version:

After the revision, the article has improved substantially. The authors have addressed most of my concerns and suggestions, except for testing the eQTL reinforcement/compensation relationship in the context of genetic architecture. I understand the motivation for testing this relationship at the gene level to determine whether it arises from directional or stabilizing selection, rather than examining it in a cis-trans pairwise fashion. However, I find the definition of this relationship unclear. The authors state in line 824 that "Genes were defined as compensating and reinforcing if they had at least 60% of individuals with opposite and same cis-trans allelic configuration, respectively." In contrast, if I understood correctly, the response to reviewers describes the relationship as reinforcing if the cis-eQTL effect is in the same direction as the mean effect of all the detected trans-eQTLs. I would request that the authors clarify their method of defining this relationship. Also, one should be aware of the fact that this relationship can evolve neutrally. Since there was no formal test performed to say it is otherwise, the authors might need to interpret the relationship carefully.

While the authors explain the possible factors that could lead to the trend of observing widespread genotype-dependent plastic responsse without significant genotype-dependent plasticity for fitness (L142), it is also important to consider the time axis. While filled grain serves as a proxy for fitness over time, gene expression profiles provide only a snapshot at a given time point. Therefore, temporal GxE dynamics may also play a role here.

Also, I am a little surprised by not mentioning anything about the code availability in this manuscript. I would request the authors to incorporate that in the revised version.

Reviewer #3 (Public review):

In this work, the authors conducted a large-scale field trial of 130 indica accessions in normal vs. moderate salt stress conditions. The experiment consists of 3 replicates for each accession in each treatment, making it 780 plants in total. Leaf transcriptome, plant traits, and final yield were collected. Starting from a quantitative genetics framework, the authors first dissected the heritability and selection forces acting on gene expression. After summarizing the selection force acting on gene expression (or plant traits) in each environment, the authors described the difference in gene expression correlation between environments. The final part consists of eQTL investigation and categorizing cis- and trans-effects acting on gene expression.

Building on the group's previous study and using a similar methodology (Groen et al. 2020, 2021), the unique aspect of this study is in incorporating large-scale empirical field works and combining gene expression data with plant traits. Unlike many systems biology studies, this study strongly emphasizes the quantitative genetics perspective and investigates the empirical fitness effects of gene expression data. The large amounts of RNAseq data (one sample for each plant individual) also allow heritability calculation. This study also utilizes the population genetics perspective to test for traces of selection around eQTL. As there are too many genes to fit in multiple regression (for selection analysis) and to construct the G-matrix (for breeder's equation), grouping genes into PCs is a very good idea.

In the previous review, three major points were mentioned. The manuscript was modified, and here I briefly summarize them as a reference for future works:

(1) The separate sections (selection analysis, transcript correlation structure change, and eQTL) could use better integration.<br /> (2) It would be worth considering joint analyses integrating the two environments together.<br /> (3) Whether gene expression PCs or unique expression modules should be used in selection analyses.

Regarding whether to use PCs or WGCNA eigengenes to summarize gene expression for selection analyses, the authors reported that only a few WGCNA eigengenes were under selection, citing this observation as the rationale for choosing PC over eigengenes. However, as the relative false positive-negative rates of these choices likely require another dedicated study to explore, at this stage, it might be premature to state which method is better based on which gives more positive results. On one hand, one could easily imagine that plants screwed up by salinity have erratic genomewide expression and become extreme data points on the PCs, making the PCs a good proxy to correlate with fitness. On the other, it remains to be discussed whether this genomewide screwed-up-ness is what we want to measure in this study or whether we should focus on more dedicated gene modules instead. I suggest the authors acknowledge both possibilities. In this revision, I do not see relevant WGCNA results (as mentioned in the previous response letter) reported.

Figure 4: The observation that chlorophyll a content is under negative selection under BOTH conditions is a bit counterintuitive. The manuscript only mentioned "consistent with the general trend for reduced photosynthesis under salinity stress" (line 329) but did not mention why this increased fitness, even in normal conditions.

Reviewer #4 (Public review):

The manuscript examines how patterns of selection on gene expression differ between a normal field environment and a field environment with elevated salinity based upon transcript abundances obtained from leaves of a diverse panel of rice germplasm. In addition, the manuscript also maps expression QTL (eQTL) that explains variation in each environment. One highlight from the mapping is that a small group of trans-mapping regulators explains some gene expression variation for large sets of transcripts in each environment.

The overall scope of the datasets is impressive, combining large field studies that capture information about fecundity, gene expression, and trait variation at multiple sites. The finding related to patterns indicating increased LD among eQTLs that have cis-trans compensatory or reinforcing effects in interesting in the context of other recent work finding patterns of epistatic selection. The authors have made some changes that address previous comments. However, some analyses in the manuscript remain less compelling or do not make the most from the value of collected data. Although the authors have made several improvements to the precision with which field-specific terminology is applied and to the language chosen when interpreting analytical findings, additional changes to improve these aspects of the manuscript remain necessary.

Selection of gene expression: One strength of the dataset is that gene expression and fecundity were measured for the same genotypes in multiple environments. However, the selection analyses are largely conducted within environments. Addition of phenotypic selection analyses that jointly analyze gene expression across environments and or selection on reaction norms would be worthwhile.

Gene expression trade-offs: The terminology and possibly methods involved in the section on gene expression trade-offs need amendment. I specifically recommend discontinuing reference to the analysis presented as an analysis of antagonistic pleiotropy (rather than more general as trade-offs) because pleiotropy is defined as a property of a genotype, not a phenotype. Gene expression levels are a molecular phenotype, influenced by both genotype and the environment. By conducting analyses of selection within environments as reported, the analysis does not account for the fact that the distribution of phenotypic values, the fitness surface, or both may differ across environments. Thus, this presents a very different situation than asking whether the genotypic effect of a QTL on fitness differs across environments, which is the context in which the contrasting terms antagonistic pleiotropy and conditional neutrality have been traditionally applied. The results reported do not persuasively support the assertion made in the response to reviewers that the terminology is reasonable due to strong coupling between genotype and phenotype. A more interesting analysis would be to examine whether the covariance of phenotype with fitness has truly changed between environments or whether the phenotypic distribution has just shifted to a different area of a static fitness surface.

Biological processes under selection / Decoherence: In the initial review, it was noted that PCA is likely not the most ideal way to cluster genes to generate consolidated metrics for a selection gradient analysis. Because individual genes will contribute to multiple PCs, the current fractional majority-rule method applied to determine whether a PC is under direct or indirect selection for increased or decreased expression comes across as arbitrary and with the potential for double-counting genes. A gene co-expression network analysis could be more appropriate, as genes only belong to one module and one can examine how selection is acting on the eigengene of a co-expression module. Building gene co-expression modules would also provide a complementary and more concrete framework for evaluating whether salinity stress induces "decoherence" and which functional groups of genes are most impacted. Although results of co-expression network analyses are now briefly discussed in the response to reviewers, the findings and their relationship to the PCA/"decoherence" analyses are not reported in the manuscript.

Selection of traits: Having paired organismal and molecular trait data is a strength of the manuscript, but the organismal trait data are underutilized. The manuscript as written only makes weak indirect inferences based on GO categories or assumed gene functions to connect selection at the organismal and molecular levels. After prompted by the initial reviews to test for correspondence between SNPs that explain organismal and gene expression trait variation or co-variance of co-expression module variation and trait variation, the response to reviewers indicates finding negative results. These findings should be included in the manuscript text and discussed.

Genetic architecture of gene expression variation: More descriptive statistics of the eQTL analysis have been included, although additional information about the variation in these measures within environments would be useful. The motivation for featuring patterns of cis-trans compensation specifically for the results obtained under high salinity conditions remains unclear to me. If the lines sampled have predominantly evolved under low salinity conditions, and the hypothesis being evaluated relates to historical experience of stabilizing selection, then evaluating the eQTL patterns under normal conditions provides the more relevant test of the hypothesis.

Lines 280-282: The revised sentence continues to read as an overstatement and merits additional revision with citations.

Lines 379-381: Following revision, it still remains unclear how the interpretation follows from the above analysis; the inference as written goes significantly beyond what may be specifically inferable from the result.

Reviewer #5 (Public review):

Summary:

The researchers examined selection across multiple levels, including gene expression, biological processes, and regulatory mechanisms, with a particular focus on comparing selection between different environmental conditions. They further explored potential evolutionary mechanisms. This is made possible with a comprehensive dataset comprising gene expression data from 130 accessions with three replicates collected in two environments in the field, genomic data from 125 genotypes, and associated physiological traits. The findings have significant implications for understanding the evolution of stress adaptation, and the identified possible genes and pathways for further investigation.

The researchers began by focusing on the selection of gene expression across two environments, comparing the number of genes under selection and the effect sizes, as well as examining how selection in each environment acts on the same individual genes. They then expanded their analysis to consider selection in biological processes, investigating the relationships between selection acting on individual genes within processes and selection acting among different processes.<br /> Additionally, they explored selection at the organismal level by examining traits.

The study further transitioned from analyzing individual gene expression to investigating gene-gene interactions. They briefly examined correlation variation among gene pairs between the two conditions, identifying pairs with rewired interactions that suggest potential selection on gene regulation or the effect of rewiring on tolerance. The researchers then delved into the genetic architecture underlying these patterns by mapping eQTLs. Their comparison of cis- and trans-eQTLs revealed that trans-eQTLs were more variable across conditions. Notably, they identified hotspots representing master regulators that possibly underlie the greater variability of trans-eQTLs across environments. They further discovered that trans-eQTLs are generally under purifying selection (particularly in salt conditions), while cis-eQTLs are under balancing selection, exhibiting higher nucleotide diversity. As for how cis- and trans-eQTL effects combine at the level of individual genes, more are found to be reinforced and the hypothesis of genetic fixation on cis- and trans-eQTL effects combination is further tested.

Strengths:

A key strength of this study is its comprehensive approach, extending beyond the analysis of gene expression to include gene-gene interactions, genetic architectures, and selections of genetic regulation factors. The exploration of gene expression selection through its connection with fitness, as introduced in the researchers' previous work, provides valuable insights into the role of gene expression in adaptation. The study investigates selection across multiple levels of biological responses, including individual gene expression, genes associated with biological processes, gene-gene interactions, and the underlying genetic architecture. The experimental design enables a direct comparison of selection between control and salinity conditions, which sheds light on the effects of stress on selection and the dynamics of adaptation to stress. Additionally, the manuscript is well-written, with a clear connection to current literature. The discussion effectively integrates findings with broader implications, making it a satisfying read.

Weaknesses:

The lack of formal testing for environment-specific selections (e.g., selection of gene expression specifically in salinity stress, PCs, or traits) is a major limitation, as previous reviewers have flagged. Explicit tests of eQTLs variation between conditions are introduced, so similar formal tests should also be introduced in selection sections. For example, a formal test of selections of gene expression might be helpful to solve variance/mean- standardization concerns between two environments.

Additionally, some aspects of the analysis appear somewhat arbitrary and could benefit from further sensitivity testing. Line 203: The concern about bias in detecting more CN than AP, as mentioned by the authors and previously flagged by reviewers, does not seem fully resolved with the current methods given the arbitrary cut-off. Incorporating additional tests suggesting the conclusion is insensitive to the cutoff would be very helpful. Similar is the classification of genes into compensatory and reinforcing categories based on 60% of individuals as a cutoff.

While this study focuses on gene regulation, its connection with the selection of gene expression and biological pathways is not well integrated. In particular, the discovery of eQTLs is not explicitly linked to gene expression selection or biological pathways, leaving this relationship underexplored. Suggestive comments: Currently the summarization of selection is based on eQTLs. It would be interesting to also summarize the selection patterns identified from previous sections based on genes being cis/trans-regulated. Moreover, it might be interesting to see if there is more loss or gain of eQTLs under salt stress and their functions. The current results mentioned variations of eQTLs but not clear if they are loss or gain. E.g., one way is to identify genes related to cis and trans-eQTLs and see their correlation changes with genes being regulated using CILP (also as a way to informatively narrow down gene pairs for CILP).

Similarly, the section on selection at the organismal trait level appears disconnected from the rest of the analysis (e.g., if it is not tested to be related to other features, mentioning why it might not be related would be helpful). Admittedly, the discussion of how biological processes discovered at different levels integrate together is helpful.

Other comments: given there is no comparison between loss of coherence (correlations) and gain of coherence under salinity stress to show the dominant role of decoherence, maybe need to also discuss the genes and processes related to the gain of coherence? This is because the understanding of activation (gain of coherence) of some regulations/processes under stress conditions could also be interesting. It is not clear if decoherence (e.g., lines 293-296) refers to significant correlation changes or just loss of the correlation in salinity stress.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Review):

Summary:

Understanding the mechanisms of how organisms respond to environmental stresses is a key goal of biological research. Assessment of transcriptional responses to stress can provide some insights into those underlying mechanisms. The researchers quantified traits, fitness, and gene expression (transcriptional) response to salinity stress (control vs stress treatments) for 130 accessions of rice (three replicates for each accession), which were grown in the field in the Philippines. This experimental design allowed for many different types of downstream analyses to better understand the biology of the system. These analyses included estimating the strength of selection imposed on transcription in each environment, evaluating possible trade-offs in gene expression, testing whether salinity induces transcriptional decoherence, and conducting various eQTL-type analyses.

Strengths:

The study provides an extensive analysis of gene expression responses to stress in rice and offers some insights into underlying mechanisms of salinity responses in this important crop system. The fact that the study was conducted under field conditions is a major plus, as the gene expression responses to soil salinity are more realistic than if the study was conducted in a greenhouse or growth chamber. The preprint is generally well-written and the methods and results are mostly well-described.

Weaknesses:

While the study makes good use of analyzing the dataset, it is not clear how the current work advances our understanding of gene regulatory evolution or plant responses to soil salinity generally. Overall, the results are consistent with other prior studies of gene expression and studies of selection across environmental conditions. Some of the framing of the paper suggests that there is more novelty to this study than there is in reality. That said, the results will certainly be useful for those working in rice and should be interesting to scientists interested in how gene expression responses to stress occur under field conditions. I detail other concerns I had about the preprint below:

The abstract on lines 33-35 illustrates some of my concerns about the overstatement of the novelty of the current study. For example, is it really true that the role of gene expression in mediating stress response and adaptation is largely unexplored? There have been numerous studies that have evaluated gene expression responses to stresses in a wide range of organisms. Perhaps, I am missing something critically different about this study. If so, I would recommend that the authors reword this sentence to clarify what gap is being filled by this study. Further, is it really the case that none of them have evaluated how the correlational structure of gene expression changes in response to stresses in plants, as implied in lines 263-265? Don't the various modules and PC analyses of gene expression get at this question?

We have re-worded these sentences, and highlighted the novelty of our work.

There were some places in the methods of the preprint that required more information to properly evaluate. For example, more information should be provided on lines 664-668 about how G, E, and GxE effects were established, especially since this is so central to this study. What programs/software (R? SAS? Other?) were used for these analyses? If R, how were the ANOVAs/models fit? What type of ANOVA was used? How exactly was significance determined for each term? Which effects were considered fixed and which were random? If the goal was to fit mixed models, why not use an approach like voom-limma (Law et al. 2014 Genome Biology)? More details should also be added to lines 688-709 about these analyses, including what software/programs were used for these analyses.

We have added more details in the methods. Also, although we could in priciple use voom-limma to fit our mixed model, to be able to partition variance into G, E and G×E, we need to use the function fitExtractVarPartModel (from package VariancePartition) which requires all categorical variables to be modeled as random effects. Therefore, we couldn’t model environment as a fixed effect.

One thing that I found a bit confusing throughout was the intermixing of different terms and types of selection. In particular, there seemed to be some inconsistencies with the usage of quantitative genetics terms for selection (e.g. directional, stabilizing) vs molecular evolution terms for selection (e.g. positive, purifying). I would encourage the authors to think carefully about what they mean by each of these terms and make sure that those definitions are consistently applied here.

We have defined the selection terms used in the study and used these terms consistently throughout the manuscript.

It would be useful to clarify the reasons for the inherent bias in the detection of conditional neutrality (CN) and antagonistic pleiotropy (AP; Lines 187-196). It is also not clear to me what the authors did to deal with the bias in terms of adjusting P-value thresholds for CN and AP the way it is currently written. Further, I found the discussion of antagonistic pleiotropy and conditional neutrality to be a bit confusing for a couple of reasons, especially around lines 489-491. First of all, does it really make sense to contrast gene expression versus local adaptation, when lots of local adaptation likely involves changes in gene expression? Second, the implication that antagonistic pleiotropy is more common for local adaptation than the results found in this study seems questionable. Conditional neutrality appears to be more common for local adaptation as well: see Table 2 of Wadgymar et al. 2017 Methods in Ecology and Evolution. That all said, it is always difficult to conclude that there are no trade-offs (antagonistic pleiotropy) for a particular locus, as the detecting trade-offs may only manifest in some years and not others and can require large sample sizes if they are subtle in effect.

We have now explained the cause of the inherent bias in the detection of CN, and also elaborated on how we deal with this bias. Also, we have edited our discussion and added relevant citations to indicate both conditional neutrality and antagonistic pleiotropy can lead to local adaptations and added the caveat regarding detecting antagonistic pleiotropy.

Reviewer #2 (Public Review):

The authors investigate the gene expression variation in a rice diversity panel under normal and saline growth conditions to gain insight into the underlying molecular adaptive response to salinity. They present a convincing case to demonstrate that environmental stress can induce selective pressure on gene expression, which is in agreement to their earlier study (Groen et al, 2020). The data seems to be a good fit for their study and overall the analytic approach is robust.

(1) The work started by investigating the effect of genotype and their interaction at each transcript level using 3'-end-biased mRNA sequencing, and detecting a wide-spread GXE effect. Later, using the total filled grain number as a proxy of fitness, they estimated the strength of selection on each transcript and reported stronger selective pressure in a saline environment. However, this current framework relies on precise estimation of fitness and, therefore can be sensitive to the choice of fitness proxy.

We now acknowledge this caveat in the discussion.

(2) Furthermore, the authors decomposed the genetic architecture of expression variation into cis- and trans-eQTL in each environment separately and reported more unique environment-specific trans-eQTLs than cis-. The relative contribution of cis- and trans-eQTL depends on both the abundance and effect size. I wonder why the latter was not reported while comparing these two different genetic architectures. If the authors were to compare the variation explained by these two categories of eQTL instead of their frequency, would the inference that trans-eQTLs are primarily associated with expression variation still hold?

We have now also reported the effect sizes for both cis- and trans-eQTLs in the two environments and showed that the trans-eQTLs have higher effect sizes as compared to cis-eQTLs, indicating that they are able to explain higher proportion of variation in transcript abundances in the two environments.

(3) Next, the authors investigated the relationship between cis- and trans-eQTLs at the transcript level and revealed an excess of reinforcement over the compensation pattern. Here, I struggle to understand the motivation for testing the relationship by comparing the effect of cis-QTL with the mean effect of all trans-eQTLs of a given transcript. My concern is that taking the mean can diminish the effect of small trans-eQTLs potentially biasing the relationship towards the large-effect eQTLs.

We wanted to estimate compensating vs reinforcing effects, which essentially entails identifying genes that have opposing directionality of cis and trans-effects. To get the total trans-effect we decided to take the mean effect of trans-eQTLs. This mean was only used to identify the compensating/reinforcing genes and although the mean effects diminishes the effect of small trans-eQTLs, this mean was not used in downstream analyses.

Reviewer #3 (Public Review):

In this work, the authors conducted a large-scale field trial of 130 indica accessions in normal vs. moderate salt stress conditions. The experiment consists of 3 replicates for each accession in each treatment, making it 780 plants in total. Leaf transcriptome, plant traits, and final yield were collected. Starting from a quantitative genetics framework, the authors first dissected the heritability and selection forces acting on gene expression. After summarizing the selection force acting on gene expression (or plant traits) in each environment, the authors described the difference in gene expression correlation between environments. The final part consists of eQTL investigation and categorizing cis- and trans-effects acting on gene expression.

Building on the group's previous study and using a similar methodology (Groen et al. 2020, 2021), the unique aspect of this study is in incorporating large-scale empirical field works and combining gene expression data with plant traits. Unlike many systems biology studies, this study strongly emphasizes the quantitative genetics perspective and investigates the empirical fitness effects of gene expression data. The large amounts of RNAseq data (one sample for each plant individual) also allow heritability calculation. This study also utilizes the population genetics perspective to test for traces of selection around eQTL. As there are too many genes to fit in multiple regression (for selection analysis) and to construct the G-matrix (for breeder's equation), grouping genes into PCs is a very good idea.

Building on large amounts of data, this study conducted many analyses and described some patterns, but a central message or hypothesis would still be necessary. Currently, the selection analysis, transcript correlation structure change, and eQTL parts seem to be independent. The manuscript currently looks like a combination of several parallel works, and this is reflected in the Results, where each part has its own short introduction (e.g., 185-187, 261-266, 349-353). It would be great to discuss how these patterns observed could be translated to larger biological insights. On a related note, since this and the previous studies (focusing on dry-wet environments) use a similar methodology, one would also wonder what the conclusions from these studies would be. How do they agree or disagree with each other?

We acknowledge that the manuscript currently presents some analyses in a somewhat independent manner. Although it would be ideal to have a central hypothesis/message, our study is meant to broadly outline the various responses and fitness effects of salinity stress in rice. Throughout the manuscript, we have also included comparisons between our findings and that of our previous studies on drought stress to highlight any consistent themes or novel insights.

Many analyses were done separately for each environment, and results from these two environments are listed together for comparison. Especially for the eQTL part, no specific comparison was discussed between the two environments. It would be interesting to consider whether one could fit the data in more coherent models specifically modeling the X-by-environment effects, where X might be transcripts, PCs, traits, transcript-transcript correlation, or eQTLs.

We do plan to consider fitting models that explicitly incorporate X-by-environment interactions to provide a more detailed understanding of the genetics of plasticity between the two environments, but it is beyond the scope of this paper. This will be explored in a separate report.

As stated, grouping genes into PCs is a good idea, but although in theory, the PCs are orthogonal, each gene still has some loadings on each PC (ie. each PC is not controlled by a completely different set of genes). Another possibility is to use any gene grouping method, such as WGCNA, to group genes into modules and use the PC1 of each module. There, each module would consist of completely different sets of genes, and one would be more likely to separate the biological functions of each module. I wonder whether the authors could discuss the pros and cons of these methods.

We recognize that individual genes can contribute to multiple PCs, and this is precisely why we choose PCA clustering over WGCNA where one gene can belong to only one module. Our aim was to recognize all biological processes that could be under selection in either environment, and since one gene can be involved in various different processes, we wanted to identify the contribution of these genes to different processes which can be done effectively by a PCA analyses.

Reviewer #4 (Public Review):

The manuscript examines how patterns of selection on gene expression differ between a normal field environment and a field environment with elevated salinity based on transcript abundances obtained from leaves of a diverse panel of rice germplasm. In addition, the manuscript also maps expression QTL (eQTL) that explains variation in each environment. One highlight from the mapping is that a small group of trans-mapping regulators explains some gene expression variation for large sets of transcripts in each environment. The overall scope of the datasets is impressive, combining large field studies that capture information about fecundity, gene expression, and trait variation at multiple sites. The finding related to patterns indicating increased LD among eQTLs that have cis-trans compensatory or reinforcing effects is interesting in the context of other recent work finding patterns of epistatic selection. However, other analyses in the manuscript are less compelling or do not make the most of the value of collected data. Revisions are also warranted to improve the precision with which field-specific terminology is applied and the language chosen when interpreting analytical findings.

Selection of gene expression:

One strength of the dataset is that gene expression and fecundity were measured for the same genotypes in multiple environments. However, the selection analyses are largely conducted within environments. The addition of phenotypic selection analyses that jointly analyze gene expression across environments and or selection on reaction norms would be worthwhile.

We do plan to consider fitting models that explicitly incorporate G×E interactions to provide a more detailed understanding of the genetics of plasticity between the two environments, but it is beyond the scope of this paper. This will be explored in a separate report.

Gene expression trade-offs:

The terminology and possibly methods involved in the section on gene expression trade-offs need amendment. I specifically recommend discontinuing reference to the analysis presented as an analysis of antagonistic pleiotropy (rather than more general trade-offs) because pleiotropy is defined as a property of a genotype, not a phenotype. Gene expression levels are a molecular phenotype, influenced by both genotype and the environment. By conducting analyses of selection within environments as reported, the analysis does not account for the fact that the distribution of phenotypic values, the fitness surface, or both may differ across environments. Thus, this presents a very different situation than asking whether the genotypic effect of a QTL on fitness differs across environments, which is the context in which the contrasting terms antagonistic pleiotropy and conditional neutrality have been traditionally applied. A more interesting analysis would be to examine whether the covariance of phenotype with fitness has truly changed between environments or whether the phenotypic distribution has just shifted to a different area of a static fitness surface.

We recognize that pleiotropy is a property of a genotype, and not phenotype, but since our phenotype (gene expression) is strongly coupled with the genotype, we choose to call trade-offs as antagonistic pleiotropy. That being said, we did test whether the covariance of gene expression with phenotype significantly varies between environments, and found that to indeed be the case.

Biological processes under selection / Decoherence: PCs are likely not the most ideal way to cluster genes to generate consolidated metrics for a selection gradient analysis. Because individual genes will contribute to multiple PCs, the current fractional majority-rule method applied to determine whether a PC is under direct or indirect selection for increased or decreased expression comes across as arbitrary and with the potential for double-counting genes. A gene co-expression network analysis could be more appropriate, as genes only belong to one module and one can examine how selection is acting on the eigengene of a co-expression module. Building gene co-expression modules would also provide a complementary and more concrete framework for evaluating whether salinity stress induces "decoherence" and which functional groups of genes are most impacted.

We recognize that individual genes can contribute to multiple PCs, and this is precisely why we choose PCA clustering over WGCNA where one gene can belong to only one module. Our aim was to recognize all biological processes that could be under selection in either environment, and since one gene can be involved in various different processes, we wanted to identify the contribution of these genes to different processes which can be done effectively by a PCA analyses. But again as pointed out by the reviewer, our PCs did contain contribution (even negligible) of each gene, so to identify the ‘primary’ biological processes represented by the PCs, we chose the majority rule. As for testing decoherence, we agree that a co-expression module analyses would have provided additional support to the specific test performed in our manuscript, but since it would just be additional support, we choose to not add it in the manuscript.

But based on the recommendation of the reviewer(s), we did perform a WGCNA analyses and found a total of 14 and 13 modules in normal and saline conditions, of which 0 and 2 modules (with no significant GO enrichment) were under directional selection. This supports our reasoning of potentially missing on identification of processes under selection.

Selection of traits:

Having paired organismal and molecular trait data is a strength of the manuscript, but the organismal trait data are underutilized. The manuscript as written only makes weak indirect inferences based on GO categories or assumed gene functions to connect selection at the organismal and molecular levels. Stronger connections could be made for instance by showing a selection of co-expression module eigengene values that are also correlated with traits that show similar patterns of selection, or by demonstrating that GWAS hits for trait variation co-localize to cis-mapping eQTL.

We did perform a GWAS for all the traits collected in both normal and saline environment, and only found significant hits for fecundity (in both normal and saline environment) and chlorophyll_a content (in the saline environment). But these regions did not overlap with any candidate genes or cis-mapping eQTL. Hence we choose to mention it in the manuscript. Additionally, using the WGCNA modules, we found that the only two module under selection in the saline environment were not significantly correlated with any of the traits measured.

Genetic architecture of gene expression variation:

The descriptive statistics of the eQTL analysis summarize counts of eQTLs observed in each environment, but these numbers are not broken down to the molecular trait level (e.g., what are the median and range of cis- and trans-eQTLs per gene). In addition, genetic architecture is a combination of the numbers and relative effect sizes of the QTLs. It would be useful to provide information about the relative distributions of phenotypic variance explained by the cis- vs. trans- eQTLs and whether those distributions vary by environment. The motivation for examining patterns of cis-trans compensation specifically for the results obtained under high salinity conditions is unclear to me. If the lines sampled have predominantly evolved under low salinity conditions and the hypothesis being evaluated relates to historical experience of stabilizing selection, then my intuition is that evaluating the eQTL patterns under normal conditions provides the more relevant test of the hypothesis.

We have added the median number of eQTLs per gene in each environment. Additionally, we recognize that genetic architecture is a combination id numbers and effect size, and we have added information regarding the effect sizes of eQTLs by type and by environment as recommend by another reviewer. We did explore the distributions of phenotypic variance explained by the cis- vs. trans- eQTLs as recommended here, and found that trans-eQTLs explain more phenotypic variance than cis-eQTLs in both environments and that the distribution of either type of eQTL does not vary by environment. We are choosing to not add this in the main text due to space limitations. Lastly, we examined the patterns of cis-trans compensation/reinforcement under both normal and salinity conditions and have compared and contrasted the results from both in the main text.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Lines 126: I would recommend citing those who originally developed the 3' end targeted RNA sequencing methods (e.g. Meyer et al 2011 Molecular Ecology).

We have cited the recommended paper.

Lines 128-130: It would be useful to include a description here of what models were fit to the data to partition out G, E, and GxE effects.

Due to space limitations, we have in brief added a sentence to this effect.

Line 139: I would suggest changing "found little" to "no" since the test was not significant.

The sentence has been modified to say no evidence.

Line 313: I think you mean directional selection instead of positive selection.

We have corrected the text

Lines 362-363: Would the authors also expect an enrichment of reinforcing genes for most scenarios where that has been divergent selection, such as local adaptation among populations?

Based on our hypothesis, we would indeed expect an enrichment of reinforcing genes for scenarios of local adaptation where different alleles are maintained in different populations due to local adaptation.

Reviewer #3 (Recommendations For The Authors):

Figures 1d-e are not mentioned in the Results.

The figures have been referenced in appropriate places.

Lines 41-45: Terms such as reinforcement and compensation need to be explained in this specific context. Also "different selection regimes" is a bit broad and vague.

Due to word-count limitation, we are choosing to not elaborate the terms reinforcement and compensation in the abstract (since these are commonly used in the literature, and we have also defined these in the main text). Additionally, we now explicitly state the selection pressures associated with cis and trans eQTLs.

Table 1: Please explain S and C in the footnote.

We have added the recommended footnote

Figures: Some panel labels (a, b, c...) are mingled with the graphs.

We are re-made our figure such that the panel labels do not mingle with the plots.

Lines 588-591: font.

Modified

Lines 620-633: Please describe how these RNAseq libraries were allocated/pooled into different sequencing lanes to avoid potential batch effects among sequencing lanes.

The sequencing was performed on the same Illumina NextSeq 500 machine and we have added the sequencing libraries/pool plan in the methods (lines 688-689). 

Lines 690-692: At the beginning of this paragraph, it was mentioned that the un-standardized coefficients were estimated. But here, it seems like the transcript data were already standardized in the data preparation step. What do lines 687-688 refer to? Further standardizing those estimated coefficients so that the whole distribution has mean=0 and sd=1?

Thank you pointing out our oversight. We checked our scripts and data preparation did not include transcript standardization, and we have removed the above line from the manuscript.

Lines 705-711: Please explain why assigning the positive/negative selection status for each gene is important. "Positive selection" here is defined as genes whose increased expression also increases fitness, but traditionally positive selection was defined as "the derived state is favored over the ancestral state". For a gene whose ancestral expression is high but lower expression increases fitness in this experiment, could we also say this gene is under positive selection? Given that we don't know the ancestral state here, maybe the authors could explain whether this definition is necessary. Also, given that many genes positively or negatively regulate each other in a pathway, it is also unclear whether it is necessary to assign the positive/negative status for a PC using the majority rule (lines 710-711).

We have now defined the different selection terms with respect to our study and use them consistently throughout the manuscript.