Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #4

Evidence, reproducibility and clarity

Summary:

The manuscript by Liang et al. shows that, while overall translation is reduced under hypoxic conditions, translation of Ldh mRNA substantially increases. This increase is demonstrated to depend upon the Ldh 3' UTR and the variant translation factor eIF4EHP. The manuscript further shows that eIF4EHP associates with polysomal fractions and with Ldh mRNA under hypoxic conditions and that it is enriched in cytoplasmic foci apparently distinct from stress granules and P bodies in hypoxia. Finally, the paper provides evidence that loss of eIF4EHP worsens the survival rate of flies in reduced oxygen conditions.

Major comments:

The experimental data largely support the conclusions that are drawn but I have some important reservations.

1. In human cells eIF4E2 (the eIF4EHP orthologue) can activate translation in hypoxia by forming a complex with HIF2alpha, RBM4, and eIF4G3. The paper states in the discussion that inhibition of the Drosophila eIF4G3 counterparts eIF4G2 and NAT1 does not affect translation under hypoxia. However, the data behind this important conclusion are not shown, nor are details given as to how eIF4G2 and NAT1 were 'inhibited', whether or not both were inhibited at the same time, and whether the fly RBM4 orthologue Lark was investigated. While the mechanism of how eIF4EHP activates translation must be different in Drosophila from that in human cells because of the absence of HIF2alpha, not much more can be concluded than that in the absence of explicit experimental data.

2. Fig 3B shows a large increase in LDH levels under hypoxic conditions in eIF4EHP KO cells, which is inconsistent with the narrative description of this result and the paper's conclusions. Ratios of LDH in normoxic and hypoxic conditions for eIF4EHP KO and control cells should be quantitated from multiple experiments, compared, and tested for statistical significance.

3. Statistical significance should also be shown for the data in Fig 3E. I note that the empty vector control reduces puromycin incorporation by quite a lot.

4. The number of survivors is very low for both control and eIF4EHP KD flies in Fig 7A. Are these comparisons statistically significant?

Minor comments:

1. It should be explained why eIF4E6 KO was included in Fig 3.

2. I assume that 6% oxygen was used for the viability experiment in Fig 7A because the lower concentration of 1% used in all other experiments would be lethal to both control and KD flies. This reasoning should be made explicit.

3. There is a great deal of sloppiness about nomenclature. The FlyBase term for the molecule of interest is eIF4EHP. Within this manuscript I found it referred to as eIF4EHP, eiF4EHP, eIF4HP, eiF4HP, and ei4EHP. This requires correction.

4. The references are also sloppy and presented in several different formats.

Significance

As mentioned above, it is established in human cells that the orthologue of eIF4EHP can activate translation under hypoxic conditions. So the basic result in this manuscript simply confirms that the same phenomenon occurs in flies. As mentioned in the previous section, in human cells eIF4EHP activates translation in a 3' UTR dependent manner as part of a complex containing HIF2alpha, RBM4, and eIF4G3. While a different mechanism is likely to exist in Drosophila, it is not elucidated by the experimental data presented in this paper. Therefore I find the work to be of limited significance.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Exposure to hypoxia induces dramatic metabolic changes in metazoan cells, and a major reprogramming of gene expression occurs to adapt to decreased energy production, such as global repression of protein synthesis, with exception of a select subset of genes whose functions are required during hypoxia. Inspired by work from a mammalian study, the authors of this manuscript investigated how a certain mRNA, in this case Ldh, is selectively translated under hypoxia in Drosophila melanogaster. The authors discovered that the 3'UTR of Ldh mediates its translational activation in hypoxia. Furthermore, they identify eIF4EHP as a critical component for this hypoxia induced translation, and show that its function is important for fly survival and development under hypoxic conditions. The authors present extensive, compelling results to support their conclusions. However, we still find this work can be improved in several aspects:

1. The authors showed the importance of the 3'UTR of Ldh in regulating Ldh translation under hypoxia, and they mention in the discussion that further investigation is needed to understand the nature of this regulation. Mutational analysis of the 3'UTR sequence could help to extend the paper and enhance its impact.

2. One obvious shortcoming of this paper is that the functions of the Ldh 3' UTR and o eIF4EHP are not connected by experimental tests. Experiments aimed at determining the functional relation of the 3' UTR and eIF4EHP could enhance the paper and deliver a more complete story about the mechanism of selective translation under hypoxia.

3. The authors refer to human cell studies showing that HIF2a is involved in mRNA translation in hypoxic conditions. They mentioned in the discussion that Drosophila Sima has not been identified to interact with eIF4EHP. To test whether Sima regulates to Ldh translation, the authors could test the involvement of Sima in experiments from Fig 6.

4. In Fig 3A-B, induction of LDH by hypoxia is almost completely blocked by eiF4EHP knockdown in fly heads but not in S2 cells. We wonder whether this indicates tissue specific regulation of LDH translation, such that there might be alternative cap-binding proteins in S2 cells. Please comment.

5. In Fig 3E, the authors present results of puromycin incorporation with eIF4EHP KD under hypoxia. It is actually not clear what the function of eIF4EHP is under normoxic conditions. The authors shall include a control of puromycin incorporation with eIF4EHP KD under 21% O2. In addition, the authors should include statistical tests for the comparisons.

6. In Fig 4C, the authors did Western Blots to detect eIF4EHP, and find that there is protein signal under the condition with eIF4EHP KD + D. persimilis eIF4EHP overexpression. It is unclear whether the antibody detects both eIF4EHPs in Drosophila melanogaster and D. persimilis, or whether, alternatively, expressing the D. persimilis version induces cells to express endogenous eIF4EHP. Please comment.

7. In Fig 7A, we don't see the point of including results of flies carrying balancers as control. Direct comparisons with mCherry-RNAi should suffice. Also, presenting the percentage of hatched embryos can be misleading. We would suggest the authors present the absolute numbers of embryos examined and indicate the number of larvae that hatched.

8. In addition (Fig 7) "birth" is not an appropriate term for insects. Please state whether the numbers indicated larvae "hatched" from eggs, or adult flies "eclosed" from pupae. Please use these terms in the text, figure and figure legends.

9. In Fig 6c, a statistical test is missing.

10. The authors sometimes refer to knockdown as KO, please be accurate.

Significance

Exposure to hypoxia induces dramatic metabolic changes in metazoan cells, and a major reprogramming of gene expression occurs to adapt to decreased energy production, such as global repression of protein synthesis, with exception of a select subset of genes whose functions are required during hypoxia. Inspired by work from a mammalian study, the authors of this manuscript investigated how a certain mRNA, in this case Ldh, is selectively translated under hypoxia in Drosophila melanogaster. The authors discovered that the 3'UTR of Ldh mediates its translational activation in hypoxia. Furthermore, they identify eIF4EHP as a critical component for this hypoxia induced translation, and show that its function is important for fly survival and development under hypoxic conditions. These results are unique and significant, and should be of interest to researchers who work on hypoxia, stress responses in general, and mRNA translation. As noted above (#1, #2) the paper could go further mechanistically, to determine the relative roles of the LDH 3' UTR and eIF4EHP. But it already has an extensive array of impressive translation data.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

When cells are stressed, such as during hypoxia, most of the canonical mRNA translation is shut off. Nonetheless, stress-response genes need to be translated. One such example is the lactate dehydrogenase (Ldh) mRNA, which encodes for an enzyme needed by cells to resist hypoxia. Therefore, these mRNA employ non-canonical mechanisms of translation to escape the general repression. Here, Liang et al. study the molecular mechanism how the Ldh mRNA is translated in Drosophila during hypoxia. They discover that the Ldh 3'UTR contains sequences that enable efficient translation. They identify the non-canonical initiation factor eIF4HP as a factor required for Ldh translation, and show that it binds the Ldh mRNA. Interestingly, they show that the ability of the Ldh 3'UTR to promote translation during hypoxia depends on eIF4HP. They go on to show that unlike other canonical translation initiation factors, during hypoxia eIF4HP remains in polysomes and does not form translationally-repressed aggregates such as stress granules or P-bodies. Finally, Liang et al show that eIF4HP is required in flies to efficiently resist hypoxia.

Major comments:

This is a very nice and solid study. Overall, I found the data and the conclusions very convincing. Also nice is how they cover the topic broadly, starting at the molecular level with the 3'UTR of the Ldh mRNA and ending with physiological assays such as resistance of flies to hypoxia. In particular I found the experiments presented in Fig 2C and 4A to be very nice, showing that the 3'UTR is responsible for the resistance to translation inhibition, and that this is mediated by eiF4EHP. I have only a few minor comments (below) to strengthen further the study, but I think it could be published also without these additional experiments.

Minor comments:

I have only two minor suggestions for experiments that would further strengthen the conclusions presented in the paper:

• Fig 2A - it would be good to show the equivalent luciferase assays at 21% oxygen, to test whether the elevated activity of the Ldh 3'UTR is something specific to the hypoxic condition, or whether this is always the case, also under non-stressed conditions.

• Similarly, in Fig 6C it would be good to show the equivalent CLIP data for 21% oxygen. Presumably, sinc eIF4HP is not in polysomes in the normoxic condition, there should be no enrichment (or little enrichment) for the Ldh mRNA.

• For Fig 7B - it's a bit confusing to label the y-axis as "flies escaping" to mean flies that climb past a certain limit. I suggest relabeling the axis to something like "flies climbing past threshold".

Significance

• The mechanisms of translation that one can read when opening a standard molecular biology textbook are the canonical mechanisms that take place in cells when they are not stressed. Cells, however, often experience stress. For instance, animals in the wild are exposed to heat or cold stress, cancer cells in a tumor are exposed to hypoxia, low amino acids, low sugar, etc. In such conditions, the canonical mRNA translation systems are shut down, and non-canononical mechanisms are remain (or are turned on). So it is a very interesting topic to understand how these non-canonical mRNA translation mechanisms function. This study contributes significantly to this topic by identifying the molecular mechanism how Ldh is translated during hypoxia. Ldh is an important gene because it is the last step of anaerobic glycolysis in animals, needed for cells to produce ATP when oxidative phosphorylation in mitochondria is incapable of functioning. Hence, understanding how the Ldh mRNA is translated is important for cell metabolism, cell viability, and organismal physiology. Furthermore, discovering that a non-canonical form of eIF4E, called eIF4EHP in Drosophila, or eIF4E2 in humans, is responsible for this translation is an important finding that opens up new avenues for future research, asking more broadly which parts of the translatome depend on this factor for their translation. The fact that eIF4EHP knockdown flies fare less well than control flies in response to hypoxia shows that eIF4EHP is playing an important role.

• I think these findings will be interesting for a broad audience studying mRNA translation, hypoxia, cell stress responses, cell and organismal metabolism, and organismal physiology.

• My expertise is in Drosophila development, mRNA translation and tissue growth.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The manuscript by Gueydan and collaborators examines the role of eiF4EHP in Drosophila. The major conclusion of the paper is that eIF4EHP believed to be a translation repressor in fact drives protein synthesis during hypoxia. The bulk of the study is focused on the lactate dehydrogenase (LDH) mRNA, a mRNA that undergoes efficient translation during hypoxia thereby going against the grain of the general protein synthesis inhibition by low oxygen tension. While the paper is somewhat confirmatory of work published by other groups that eIF4E2 (homolog of eIF4EHP) drives hypoxic translation, the work shown here is done in whole organism that adds what I consider to be another significant layer of evidence to the story. The paper also provides additional evidence that cells are equipped with multiple stimuli-specific cap-binding translation initiation factors that produce adaptive translatomes. The work is convincing, the paper well-written and the demonstration of eIF4EHP role in whole organism is important. Nonetheless, I do have a few comments that should be addressed to strengthen the conclusions of the authors.

1. The work focuses mainly of LDH. It would be a missed opportunity to show the effect of eIF4EHP in hypoxic Drosophila on puromycin incorporation. The work on S2 cells shown in Fig. 3E, while convincing, only reproduces what is already known. Puromycin incorporation in larvae.

2. In mammalian cells, blocking de novo transcription does not affect protein accumulation of eIF4E2 translated mRNA, even if these mRNA do not increase during hypoxia. It would be important to silence sima and test if this could block increased translation of ldh or other target in hypoxic conditions. This would suggest that sima is both a transcription and translation factor that evolved in HIF1a and HIF2a, the latter being a hypoxic translation regulator. This can be compared to cells treated with a general transcription inhibitor. These experiments would broaden the impact of the work

3. Fig 5C. It is unclear what are the conclusions of the authors are on the lack of co-localization between poly(A)RNA and eIF4EHP. In principle, eIF4EHP should co-localize with poly(A)mRNA. Perhaps a proximity ligation assay should be done to clarify this question. The data shown in F5C is not convincing one way or the other and needs clear cut experiments. Idem for 5A and B. PLA would provide convincing results.

4. Fig 6C is important but the data is hard to understand. First, the Y-axis is "enrichment relative to input". Is this hypoxia vs normoxia? Is the Y-axis log (probably)? The best would be to show that normoxia and hypoxia and see if eIF4EHP binds to ldh mRNA in both conditions perhaps acting as a translational repressor in normoxia and translational activator in hypoxia. Also, the authors could pool monosome and polysome factions and see in which fractions eIF4EHP binds to ldh mRNA. Finally, do the authors think that rpl32 remains associated with ldh mRNA in normoxia and hypoxia?

5. The authors should integrate the emerging concept of adaptive cap-binding translation factors in the discussion. It is still generally assumed that inhibition of eIF4E results automatically to cap-independent protein synthesis. The discovery that eIF3D, 4E2 and 4E3, amongst others, can initiate stimuli-specific translation should be discussed in the context of the work from this paper.

Significance

While the paper is somewhat confirmatory of work published by other groups that eIF4E2 (homolog of eIF4EHP) drives hypoxic translation, the work shown here is done in whole organism that adds what I consider to be another significant layer of evidence to the story

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2022-01680

Corresponding author(s): Woo Jae, Kim

1. General Statements The goal of this study is to provide the groundwork for future studies of genetically controlled neuronal regulation of ‘interval timing’ through the provision of a behavioral paradigm. Interval timing, or the sense of time in the seconds to hours range, is important in foraging, decision making, and learning in humans via activation of cortico-striatal circuits. Interval timing requires completely distinct brain processes from millisecond or circadian timing. In summary, interval timing allows us to subjectively sense the passage of physical time, allowing us to integrate action sequences, thoughts, and behavior, detect developing trends, and predict future consequences.

Many researchers have tried to figure out how animals, including humans, can estimate time intervals with such precision. However, most investigations have been conducted in the realm of psychology rather than biology thus far. Because the study of interval timing was limited in its ability to intervene in the human brain, many psychologists concentrated on developing convincing theoretical models to explain the known occurrence of interval timing.

To overcome the limits of studying interval timing in terms of genetic control, we have reported that the time investment strategy for mating in Drosophila males can be a suitable behavioral platform to genetically dissect the principle of brain circuit mechanism for interval timing. For example, we previously reported that males prolong their mating when they have previously been exposed to rivals (Kim, Jan & Jan, "Contribution of visual and circadian neural circuits to memory for prolonged mating induced by rivals" Nature Neuroscience, 2012), and this behavior is regulated by visual stimuli, clock genes, and neuropeptide signaling in a subset of neurons (Kim, Jan & Jan, “A PDF/NPF Neuropeptide Signaling Circuitry of Male Drosophila melanogaster Controls Rival-Induced Prolonged Mating” Neuron, 2013).

Throughout their lives, all animals must make decisions in order to optimize their utility function. Male reproductive success is determined by how many sperms successfully fertilize an egg with a restricted number of investment resources. To optimize male reproductive fitness, a time investment strategy has been devised. As a consequence, we believe that the flexible responses of mating duration to different environmental contexts in Drosophila males might be an excellent model to investigate neural circuits for interval timing.

One of the most well-known features of human interval timing is the association of different sensory inputs with perception of time intervals, which influences our estimate of time intervals. Therefore, the first step toward comprehending the neural regulation of interval timing is to dissect the role that numerous sensory inputs play in determining the time duration. In this article, we discuss a different time-investment strategy adopted by males, called "Shorter-Mating-Duration" (SMD). According to our findings, male Drosophila with more sexual experience had shorter mating duration. During our investigation into the sensory inputs for SMD behavior, we found a small number of cells that express sugar receptors and pheromone receptors (ppk25 and ppk29) and thus transmit the multisensory information from females in order to generate memories of sexual experiences, which will determine the final decision of mating duration.

Our discovery of sensory integration mechanisms associated with complex behavioral trait in male Drosophila at the brain circuit and genetic network levels will be a huge step forward in our knowledge of interval timing behavior.

Description of the planned revisions

REVIEWER #1

1. • Overall I think this would be difficult for a general audience as the rationale and explanation of experiments needs to be clearer. * Answer: During the revision process, we will make our text more legible for wide audiences.

REVIEWER #2

1. • 'The knockdown of LUSH, an odorant-binding protein' Lush is expressed in trichoid sensilla in olfactory organs , from the beginning, they exclude the role of olfaction and later one they said 'suggesting that the expression of the pheromone sensing proteins LUSH and Snmp1 in Gr5a-positive gustatory neurons is critical for generating SMD behavior.' ? Therefore, I recommend If available, please provide a reference for the statement in the Methods section that the Orco1 line was "validated via electrophysiology", or include the electrophysiology data itself in this manuscript as supplementary figure. Ideally, positive behavioral controls for this line would also be included in the manuscript. * Answer: We value the reviewer's concern. LUSH has been discovered as an odorant-binding protein; nevertheless, current research suggests that LUSH may be involved in the sensing of additional pheromones to cVA, implying the presence of a lush-independent cVA detection mechanism [1]. Billeter et al. demonstrated in their paper that LUSH detects a female stimulatory chemical and modifies male mating latency (Fig. 2 of Billeter at al.). As Billeter et al. stated, our present understanding of pheromonal recognition in Drosophila is insufficient, and we concur. As a result, we attempted to validate the expression of Snmp1 in the male leg by experiments (Fig. 7I-J) performing sncRNA seq analysis on the Fly SCope dataset, as shown in Fig. 12. As demonstrated in Fig.12, Snmp1 and LUSH is higly expressed fly leg and wing system. Future study will look at the roles of Snmp1 and LUSH in female pheromone sensing, as well as PPK receptors.

Following the reviewer's advice, we will repeat the electrophysiologically validated Orco2 mutant phenotype with proper control and attach it when we submit the complete revision to the journal.

• What is this (GustDx6)? I suggest using Poxn mutant line. *

Answer: We value the reviewer's recommendation. We believe we have previously demonstrated that the Gr5a-mediated gustatory pathway is essential for the generation of sensory input for SMD behavior, but we will test the Poxn mutant and Poxn-RNAi to replace the GustDx6 mutant result.

Description of the revisions that have already been incorporated in the transferred manuscript

REVIEWER #1

2. My copy of this ms does not have page numbers or line numbers, this makes it extremely difficult to identify where I am making queries/ suggestions. I don't know whether this is a decision of the journal or authors, but please change this in the future.* Answer: We put page numbers and line numbers.

3. A general point, there is simply too much in this ms. It covers too much ground and so doesn't give proper descriptions, discuss the consequences of the data fully or integrate properly with existing literature. Quantity does not equal impact. *

Answer: We appreciate the reviewer's insight. We have previously separated this document from our original preprint [2] in response to a prior reviewer's advice; we believe we have included too much data, which may confuse readers. As a result, we will delete all of the mechanosensory/thermosensory receptor screening data from our present paper and write a second manuscript on sensory integration for the production of SMD behavior. We also removed the most of sncRNA seq data analysis except Fig.12 which confirms our finding in a single diagram.

• Results paragraph 1 says that white mutant background had no effect "unlike that of LMD behavior as reported previously", ignoring that there has been a contrary report that extension of mating duration after exposure to a rival does not involve visual cues and so is not affected by the white mutation (Bretman et al 2011 Curr Biol). *

Answer: We recognize that there is a conflicting report concerning white mutation on LMD behavior, however because we are now reporting SMD rather than LMD behavior, we deleted the statement comparing white mutant results to earlier reports, as shown below;

“thus suggesting that the effect of the white mutant genetic background was not evident.” (line 97)

• A general point in the methodology, it's not very helpful just to say "as in a previous study" without giving at least a brief idea of what that was (e.g. the explanation of egg counting procedures).

A "sperm depletion" assay is described in the results that I cannot find any methodology for. *

Answer: We thank the reviewer for allowing us to clarify our lacking methodologies for a better comprehension of our manuscript.

We included the egg counting procedure to the EXPERIMENTAL PROCEDURES section to further illustrate our approach of egg laying assay as below;

“In short, wild type females mated with naïve or experienced males were transferred to a fresh new vial and allowed to lay eggs for 24 hr at 25°C. After 24 hr of egg laying, number of eggs were counted under the stereomicroscope. After we count the number of eggs, we kept vials in 25°C incubator and counted the total number of progenies ecolsed from them.” (line 956-960)

We included “Sperm Depletion from Males” section in EXPERIMENTAL PROCEDURES as below;

“To deplete sperm from males, 40 virgin Defexel6234 females which lacks SPR and shows multiple mating with males (Yang 2009) were placed in a vial containing four CS males for indicated time (2 h, 4 h, 8 h, and 24 h).” (line 880)

• Was the "excessive mating" with SPR females actually observed, or inferred from previous work? Needs to be clear. In what way do virgins expressing fruitless behave like mated females? It is so unclear how all the evidence in this paragraph leads to the conclusion that both cues from females and successful copulation. Especially as in the next paragraph experience with feminized females (with which the focal males cannot copulate) elicits the response.

It might be helpful to combine the results into a table, so it is easy to see under which conditions males reduce mating duration. *

Answer: We modified the sentence describing SPR mutant female experiment and added references as below;

“Sexual experiences with sex peptide receptor (SPR) mutant females which exhibit a delayed post-mating response and multiple mating with males [3] had no additional effect on SMD (Fig. 2I).” (line 135)

We clarify in which extent, fru>UAS-mSP virgin females behave like mated females as below;

“Virgin females behave like mated females by expressing a membrane-bound version of male sex-peptide in fruitless-positive neurons, hence rejecting the male's copulation attempt.” (line 136)

In the instance of feminized males, we assume that these feminine males can give adequate signals for inducing SMD and eliminated the term "successful copulation" since we are unsure if males can copulate these feminized males or not, despite the fact that males can mount and mate with them (Fig. 2O-P).

Tables S1 and S2 describe the conditions, genotypes, and descriptions of an experiments illustrated in Fig. 2. We believe that these tables may assist general audiences in comprehending our experimental design.

• Why are no statistics reported in the results? Identifying sig diffs on figures is not sufficient. I'm very sceptical that "mating duration of males showed normal distribution" for all comparisons, but then it's also difficult to identify which were analysed in this way (if statistics were properly reported this would not be an issue). *

Answer: We described our statistical analysis with mating duration previously [4–7] and followed the statistical analysis of copulation duration assay reported by Crickmore et al., published in CELL (2013) and NEURON (2020) [8,9]. To further validate our statistical analysis, we added estimation statistics which focuses on the effect size of one's experiment/intervention, as opposed to significance testing [10]. We already described our statistical analysis in EXPERIMENTAL PROCEDURES section in details. We also described our statistical analysis for mating duration will be same in all other figures in the Fig.1 legend.

We appreciate the reviewer's recommendation that the normal distribution of our mating duration data be validated. As a consequence, we performed the normailty test with Graphpad prism and added the histogram and QQ plot results to Fig. S1M and N. Table S3 also contains the results of the normality and lognormality tests.

• Gr5a/ Gr66a mediate acceptance/ avoidance of what? Why would you hypothesise these in particular to be involved? *

Answer: We accidentally left out the citation for that phrase and updated it with Wang et al.'s CELL (2004) paper. Wang et al. wrote in their article about taste representations in the Drosophila brain, “Our behavioral studies reveal that Gr5a cells recognize sugars and mediate acceptance/attractive behaviors whereas Gr66a cells recognize bitter compounds and mediate avoidance…. This suggests that Gr5a cells may be “acceptance” cells rather than “sweet” cells…. Our expression and behavioral studies reveal that Gr5a marks cells that recognize sugars and mediate taste acceptance, whereas Gr66a marks cells that recognize bitter compounds and mediate avoidance.” [11]

As a result, we hypothesize that Gr5a and Gr66a-positive cells influence acceptance or avoidance of "taste." We also changed certain sentences to make them clearer, as seen below;

“Of the various gustatory receptors, Gr5a marks cells that recognize sugars and mediate taste acceptance, whereas Gr66a marks cells that recognizes bitter compounds and mediates avoidance.” (line 173)

• As Orco was not found to affect the behaviour, why test Or67d? *

Answer: We appreciate the reviewer bringing this to our attention. We omitted the Or67d result from the present manuscript to simplify it and make it easier for readers to grasp.

• "Mate guarding" suddenly appears in the modelling section. Can a difference of a couple of minutes in a mating duration of 15-20min really be considered mate guarding? A similar variation in response to rival males is not considered mate guarding, but is linked to adjustments in ejaculate expenditure (admittedly not in a very straight forward way). Surely in a system like this the benefits arise more from how many females the male can mate with in a given time? How does this model relate to any of the previous models of mate guarding?

In this section the work of Linklater et al 2007 is important, they showed progeny declined over successive matings, and related this to exhaustion of Acps rather than sperm. I would urge the authors to consider that what they observe does not necessarily have an adaptive explanation. *

Answer: We have defined “mate guarding” in the text now. The costs and benefits of mate guarding have been extensively studied in insects and demonstrated to shape the optimal mating duration of males. In our experiment, we cannot specify whether the shortened mating duration was caused by the adjustments in ejaculate expenditure or a shorted stay after the ejaculation. Instead, our model has a general assumption that the costs of mate guarding increase linearly at the same rate in both pre- and post-ejaculation periods, which is highlighted in the model text.

There exist many models for the optimal mating duration (earlier models include Grafen and Ridley, 1983. A model of mate guarding. J. Theor. Biol. 102: 549 – 567 [12]). While our model was not built upon a novel theoretical approach (it was built based on the classical Charnov’s marginal value theorem equation), our model was developed specifically for generating testable predictions for the observed SMD behaviors.

We have rephrased the text as follow;

“This model assumes that (i) the shortened (or prolonged) mating duration is controlled by males and shaped by a trade-off between the benefit of mate guarding (remaining with the female both before and after the sperm ejaculation) and opportunistic costs (e.g. searching for another mate).” (line 970)”

• I can't find a data accessibility statement. *

Answer: We added it in the manuscript.

• That said, a current grand challenge in understanding behaviour is discovering the mechanisms that enable individuals to respond plastically to changing environments. This speaks directly to that challenge. However, this behavioural observation is not novel, as claimed. Generally the idea of refractoriness is widely known, and specifically the reduction in mating duration over successive matings in D. melanogaster was shown by Linklater et al 2007 Evolution. Moreover, the time between exposure to females has been shown to be important. Linklater et al 2007 gave males mating attempts in quick succession and observed the decrease in mating duration, whereas given recovery time of 3 days, males either mate equally as long, or even longer across their life course (Bretman et al 2011 Proc B, Bretman et al 2013 Evolution). These papers should be discussed, and more broadly the work understood in the light of previous knowledge. The behaviour does not need to be novel for this manuscript to make a significant contribution to the field. *

Answer: We believe the reviewer highlighted relevant past research that examined the influence of female experiences on mating duration. We agree with the reviewer that SMD behavior does not have to be original in order to contribute significantly to the field. As a result, we examined past reports and updated the introduction as follows;

“It has been reported that previous sexual experience with females influences the mating duration of male D. melanogaster [15,16,34]; however, the neural circuits and physiology underlying this behavior have not been deeply investigated. Here, we report the sensory integration mechanisms by which sexually experienced males exhibit plastic behavior by limiting their investment in copulation time; we refer to this behavior as "shorter mating duration (SMD)."” (line 85)

• Both in the introduction and discussion the extended mating duration in response to rivals is raised. A great deal of work has been done on this plasticity and yet the way this is written implies just two papers from these authors (whilst referencing others elsewhere). *

Answer: We agree with the reviewer. In the introductory and discussion sections, we cited as many key publications explaining the plastic responses of male mating duration as we could.

__REVIEWER #2

__

1. • Summary: The submitted manuscript reports that Drosophila melanogaster males use information derived from their previous sexual experiences from multiple sensory inputs to optimize their investment in mating. They refer to this plasticity as 'shorter-mating duration (SMD)'. SMD requires sexually dimorphic taste neurons. They identified several neurons in the male foreleg and midleg that express specific sugar, pheromone and mechanosensory receptors. Unfortunately, several aspects of the study design and methods used are inappropriate. Although the statistical approaches used are appropriate, the results are questionable. The discussion and conclusions are therefore too speculative in my view and overstretch the implications of the results as presented. Below I explain each one of these concerns about the study design, methods and results in detail as follows.* Answer: We appreciate the reviewer's assessment, especially the statement that our statistical approaches were appropriate. We will revise our manuscript in response to the reviewer's suggestions.

2. The conclusions (as the authors point out) hinge on small (often extremely small) effect sizes. This is not an insurmountable problem, so long as the assays are robust across trials. Unfortunately, they are not-the variation in the baseline for control replicates is often as large as, or larger than, the effects from which the conclusions are derived. Given the extreme experimental challenges of small effect size combined with large intertrial variability, it is notable that the authors do not report any likely false negative or false positive data, as would be frequently expected under these conditions. One explanation for the reproducibility of statistical effect seen across many experiments despite these experimental hurdles is manipulation of sample size. The authors acknowledge the extreme variability in sample size offer seemingly harmless explanations, but a closer look shows how problematic this practice is. For example, see Figure 1 (I, J, L) there is a big different between naive and experience males? *

Answer: We value the reviewer's feedback. Several research have been conducted to investigate the mating duration of male fruit fly. For example, our lab [2,13–15] and others [13–30] have regularly reported that previous rival exposure increases male fruit fly mating duration. Bretman A et al. utilized 49-59 males in their studies to compare the variations in mating duration between circumstances. Crickmore et al. also reported the effect of mating duration differences caused by genetic or experimental modification [8]. They utilized 10-18 male flies in their study to compare the variations in mating duration across circumstances, as shown in Figs. 1G (n=15-18) and 2A (n=10-27). All of these findings indicate that our mating duration sample size is sufficient to examine the effect size variations between the naive and experienced conditions. To confirm our statistical analysis further, we incorporated estimate statistics, which focus on the effect size of one's experiment/intervention rather than significance tests [10]. We have already detailed our statistical analyses under the EXPERIMENTAL PROCEDURES section. We conducted hundreds of mating duration assays using this configuration and confirmed that all of our results are reproducible in a blind test. As a result, we believe our mating duration assay has been validated by other groups' findings, several analytic tools, and numerous blind tests conducted by us. We appreciate the reviewers' concerns, but our data meets the reproducibility requirements.

• I am not sure if you keep using the same control with different experiments (that is okay if those exp is done in the same time) as in figure 1 B, I,J,K,L.But I don't think you did Fig 1B in the same time with Fig 1I, J, K,L. *

Answer: We appreciate the reviewer's feedback. Yes, all of our tests comparing the differences in mating duration between naive and experienced conditions were conducted under the same conditions and at the same time. We replaced Fig.1B with new data (n=49-51) obtained lately in a new lab in China. As previously stated, SMD behavior could be reproduced by the same Canton S genotype in different locations by different experimenters.

• It will be clear if you mention in the text how much reduction in percent happened in copulation duration when the males had previous sexual experience? *

Answer: We appreciate the reviewer’s suggestion and added in the manuscript as follow;

“We found that the mating duration of various wild-type and w1118 naïve males are significantly longer (wild type 15.7~15.8%, w1118 12.4%) than that of sexually experienced males (Fig. 1B-D, Fig. S1A)” (line 99)

• 'Drosophila simulans, the sibling species of D. melanogaster also exhibits SMD, thus suggesting that SMD is conserved between close species of D. melanogaster (Fig. S1B).'. If you want come with this conclusion, you need to test D. erecta, D. sechelia and D. yakuba. *

Answer: We appreciate the reviewer's feedback. We removed the D. simulans data because it is not required for the conclusion of this manuscript. In future research, we will look on the conservation of SMD behavior between species.

• The authors mention that Gr66a is salt. This is not 100% correct. GR66a is expressed in many bitter sensing neurons and is required for the physiological and behavioral responses to many bitter compounds. check this reference DOI:https://doi.org/10.1016/j.cub.2019.11.005. *

Answer: We made the following changes and cited the article reviewer's suggestion.

“Of the various gustatory receptors, Gr5a marks cells that recognize sugars and mediate taste acceptance, whereas Gr66a marks cells that recognizes bitter compounds and mediates avoidance (Wang et al, 2004; Dweck & Carlson, 2020).” (line 180)

• Drosophila melanogaster mating duration is between 21- 23 mins. I never saw copulation duration in normal condition (control) 10-15 mins as in figure fig 2E, Fig 7 C,E,F, Fig 8 E and fig 12 G . To the best of my knowledge, of all of the papers on copulation duration, the only one that ascribes a shortened duration to manipulations of the female is Rideout...Goodwin Nature Neuroscience 2010, who argue that this shortening results from markedly increased female activity/agitation during mating, leading the male to terminate early. *

Answer: We appreciate the reviewer's feedback. Copulation duration in Drosophila melanogaster male is extremely variable and has been reported to be approximately 20 minutes. However, as other groups documented, male copulation duration can range from 10-15 minutes depending on sperm completion (Fig. 1a-c of Bretman A et al.) [30] and genetic background (Fig. 1C, Fig. 2E, Fig. 5D, and Fig. 7A and E of Crickmore et al) [8]. And, as previously stated, males dominate copulation duration [8,30], not females, and we always utilized the same genotype of females for mating duration experiment. As a result, we believe that these rather short mating duration outcomes are the product of a distinct genetic background. Because we employed the same genotype of males while altering the female experience condition, we believe our mating duration results are all equivalent and comparable.

• In some experiments, the authors test very few number of replicates which is not convinced me to their conclusion as example Fig 2F and Fig 12 E. Why you test 100, 103 replicates in this exp fig 10 F? How you compare 47 replicates against 9 replicates in fig S10 I? *

Answer: We appreciate the reviewer's input. As we previously stated in response to Reviewer Question 2, the n number of males exhibited in Figs. 2F and 12E is statistically significant. To corroborate findings with replication, we examined 100, 103 duplicates of Fig. 10F, which represents pyx-RNAi screening results. The results of Fig. S10I are screening data, and we cannot rule out the possibility that TrpA1 knockdown in Gr5a neurons affects the mating success of sexually experienced males. We only placed it there because it was screening results and the differences between naive and experienced conditions were substantial despite the small sample size. However, we deleted Fig. 10F and Fig. S10I data from the current paper in response to Reviewer #1's advice, thus it will not be an issue for the manuscript's conclusion.

• 'Next, to decipher whether DEG/NaC channel-expressing pheromone sensing neurons require the function of OBP, we expressed lush-RNAi using ppk23-, ppk25- and ppk29-GAL4 drivers to knockdown LUSH in each channel-expressing neuron. The knockdown of LUSH in ppk25- and ppk29-GAL4 labeled cells, but not in ppk23-GAL4 labeled cells, led to a disturbance in SMD behavior, thus suggesting that LUSH functions in ppk25- and ppk29-positive neurons to detect pheromones and elicit SMD behavior (Fig. 9G-I). The knockdown of SNMP1 in ppk29-GAL4- labeled neurons also inhibited SMD behavior (Fig. 9J), thus suggesting that SNMP1 also functions in ppk29-positive neurons to induce SMD behavior.' What about ppk25? **

*

Answer: As indicated by the reviewer, we included ppk25-GAL4/snmp1-RNAi data in Fig. S9I, indicating that snmp1 expression in ppk25-positive cells is similarly implicated in SMD behavior.

• There are no page or line numbers throughout the ms! *

Answer: We included page and line numbers.

• The use of subheadings in the results section makes reading much easier.*

Answer: We added subheadings in the results section.

• 'We found that the mating duration of various wild-type and w 1118 naïve males are significantly longer than that of sexually experienced males (Fig. 1B-D, Fig. S1A)' . I think you should change various wild type to CS and WT Berlin as in legend and figure 1B,C .*

Answer: The revised sentence is as follows:

“We found that the mating duration of Canton S, WT-Berlin, Oregon-R, and w1118 naïve males are significantly longer (wild type 15.7~15.8%, w1118 12.4%) than that of sexually experienced males (Fig. 1B-D, Fig. S1A)” (line 102)

• Suggested exp , Fig S1E-H , they might test 2,6, 12 hours males separation from females to test exactly when this behavior change over time. *

Answer: We value the reviewer's recommendation. As seen in Fig. S4B of Kim et al., we have previously conducted experiments for examining the memory circuit of SMD [6]. Briefly, the male with a shorter mating duration recovers completely after 12 to 24 hours of isolation from females. As we are currently preparing the memory section of the SMD study, this information will be included in a future manuscript.

• General comment in figures, you could remove the common y axis as example in figure 1 B,C,D , difference between means and mating duration. *

Answer: We welcome the reviewer's idea, however in this situation we believe that the y axis of each data set is independent from one another and will thus retain the originals. We feel this would be more useful for the general audiences.

• You might move the number of replicates to the legend. *

Answer: We appreciate the reviewer's idea, however we feel that adding more information to the graphic will aid the general audience in comprehending our statistics.

• Latin name should be italic as example Drosophila simulans.*

Answer: We fixed it.

Description of analyses that authors prefer not to carry out

N/A

References

1. Billeter J-C, Levine JD. The role of cVA and the Odorant binding protein Lush in social and sexual behavior in Drosophila melanogaster. Frontiers Ecol Evol. 2015;3: 75. doi:10.3389/fevo.2015.00075
2. Kim WJ, Lee SG, Schweizer J, Auge A-C, Jan LY, Jan YN. Sexually experienced male Drosophila melanogaster uses gustatory-to-neuropeptide integrative circuits to reduce time investment for mating. Biorxiv. 2016; 088724. doi:10.1101/088724
3. Yang C, Rumpf S, Xiang Y, Gordon MD, Song W, Jan LY, et al. Control of the Postmating Behavioral Switch in Drosophila Females by Internal Sensory Neurons. Neuron. 2009;61: 519–526. doi:10.1016/j.neuron.2008.12.021
4. Kim WJ, Jan LY, Jan YN. Contribution of visual and circadian neural circuits to memory for prolonged mating induced by rivals. Nat Neurosci. 2012;15: 876–883. doi:10.1038/nn.3104
5. Kim WJ, Jan LY, Jan YN. A PDF/NPF Neuropeptide Signaling Circuitry of Male Drosophila melanogaster Controls Rival-Induced Prolonged Mating. Neuron. 2013;80: 1190–1205. doi:10.1016/j.neuron.2013.09.034
6. Kim WJ, Lee SG, Auge A-C, Jan LY, Jan YN. Sexually satiated male uses gustatory-to-neuropeptide integrative circuits to reduce time investment for mating. Biorxiv. 2016; 088724. doi:10.1101/088724
7. Wong K, Schweizer J, Nguyen K-NH, Atieh S, Kim WJ. Neuropeptide relay between SIFa signaling controls the experience-dependent mating duration of male Drosophila. Biorxiv. 2019; 819045. doi:10.1101/819045
8. Crickmore MA, Vosshall LB. Opposing Dopaminergic and GABAergic Neurons Control the Duration and Persistence of Copulation in Drosophila. Cell. 2013;155: 881–893. doi:10.1016/j.cell.2013.09.055
9. Thornquist SC, Langer K, Zhang SX, Rogulja D, Crickmore MA. CaMKII Measures the Passage of Time to Coordinate Behavior and Motivational State. Neuron. 2020;105: 334-345.e9. doi:10.1016/j.neuron.2019.10.018
10. Claridge-Chang A, Assam PN. Estimation statistics should replace significance testing. Nat Methods. 2016;13: 108–109. doi:10.1038/nmeth.3729
11. Wang Z, Singhvi A, Kong P, Scott K. Taste Representations in the Drosophila Brain. Cell. 2004;117: 981–991. doi:10.1016/j.cell.2004.06.011
12. Grafen A, Ridley M. A model of mate guarding. J Theor Biol. 1983;102: 549–567. doi:10.1016/0022-5193(83)90390-9
13. Kim WJ, Jan LY, Jan YN. A PDF/NPF Neuropeptide Signaling Circuitry of Male Drosophila melanogaster Controls Rival-Induced Prolonged Mating. Neuron. 2013;80: 1190–1205. doi:10.1016/j.neuron.2013.09.034
14. Kim WJ, Jan LY, Jan YN. Contribution of visual and circadian neural circuits to memory for prolonged mating induced by rivals. Nat Neurosci. 2012;15: 876–883. doi:10.1038/nn.3104
15. Wong K, Schweizer J, Nguyen K-NH, Atieh S, Kim WJ. Neuropeptide relay between SIFa signaling controls the experience-dependent mating duration of male Drosophila. Biorxiv. 2019; 819045. doi:10.1101/819045
16. Bretman A, Fricke C, Chapman T. Plastic responses of male Drosophila melanogaster to the level of sperm competition increase male reproductive fitness. Proc Royal Soc B Biological Sci. 2009;276: 1705–1711. doi:10.1098/rspb.2008.1878
17. Bretman A, Westmancoat JD, Chapman T. Male control of mating duration following exposure to rivals in fruitflies. J Insect Physiol. 2013;59: 824–827. doi:10.1016/j.jinsphys.2013.05.011
18. Bretman A, Gage MJG, Chapman T. Quick-change artists: male plastic behavioural responses to rivals. Trends Ecol Evol. 2011;26: 467–473. doi:10.1016/j.tree.2011.05.002
19. Lizé A, Doff RJ, Smaller EA, Lewis Z, Hurst GDD. Perception of male–male competition influences Drosophila copulation behaviour even in species where females rarely remate. Biol Letters. 2012;8: 35–38. doi:10.1098/rsbl.2011.0544
20. Rouse J, Bretman A. Exposure time to rivals and sensory cues affect how quickly males respond to changes in sperm competition threat. Anim Behav. 2016;122: 1–8. doi:10.1016/j.anbehav.2016.09.011
21. Bretman A, Fricke C, Hetherington P, Stone R, Chapman T. Exposure to rivals and plastic responses to sperm competition in Drosophila melanogaster. Behav Ecol. 2010;21: 317–321. doi:10.1093/beheco/arp189
22. Rouse J, Watkinson K, Bretman A. Flexible memory controls sperm competition responses in male Drosophila melanogaster. Proc Royal Soc B Biological Sci. 2018;285: 20180619. doi:10.1098/rspb.2018.0619
23. Maguire CP, Lizé A, Price TAR. Assessment of Rival Males through the Use of Multiple Sensory Cues in the Fruitfly Drosophila pseudoobscura. Plos One. 2015;10: e0123058. doi:10.1371/journal.pone.0123058
24. Bretman A, Westmancoat JD, Gage MJG, Chapman T. COSTS AND BENEFITS OF LIFETIME EXPOSURE TO MATING RIVALS IN MALE DROSOPHILA MELANOGASTER. Evolution. 2013;67: 2413–2422. doi:10.1111/evo.12125
25. Bretman A, Fricke C, Westmancoat JD, Chapman T. Effect of competitive cues on reproductive morphology and behavioral plasticity in male fruitflies. Behav Ecol. 2016;27: 452–461. doi:10.1093/beheco/arv170
26. Price TAR, Lizé A, Marcello M, Bretman A. Experience of mating rivals causes males to modulate sperm transfer in the fly Drosophila pseudoobscura. J Insect Physiol. 2012;58: 1669–1675. doi:10.1016/j.jinsphys.2012.10.008
27. Bretman A, Westmancoat JD, Gage MJG, Chapman T. Males Use Multiple, Redundant Cues to Detect Mating Rivals. Curr Biol. 2011;21: 617–622. doi:10.1016/j.cub.2011.03.008
28. Fowler EK, Leigh S, Rostant WG, Thomas A, Bretman A, Chapman T. Memory of social experience affects female fecundity via perception of fly deposits. Bmc Biol. 2022;20: 244. doi:10.1186/s12915-022-01438-5
29. Dore AA, Rostant WG, Bretman A, Chapman T. Plastic male mating behavior evolves in response to the competitive environment*. Evolution. 2021;75: 101–115. doi:10.1111/evo.14089
30. Bretman A, Fricke C, Chapman T. Plastic responses of male Drosophila melanogaster to the level of sperm competition increase male reproductive fitness. Proc Royal Soc B Biological Sci. 2009;276: 1705–1711. doi:10.1098/rspb.2008.1878

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

The submitted manuscript reports that Drosophila melanogaster males use information derived from their previous sexual experiences from multiple sensory inputs to optimize their investment in mating. They refer to this plasticity as 'shorter-mating duration (SMD)'. SMD requires sexually dimorphic taste neurons. They identified several neurons in the male foreleg and midleg that express specific sugar, pheromone and mechanosensory receptors. Unfortunately, several aspects of the study design and methods used are inappropriate. Although the statistical approaches used are appropriate, the results are questionable. The discussion and conclusions are therefore too speculative in my view and overstretch the implications of the results as presented. Below I explain each one of these concerns about the study design, methods and results in detail as follows.

Major comments:

1. The conclusions (as the authors point out) hinge on small (often extremely small) effect sizes. This is not an insurmountable problem, so long as the assays are robust across trials. Unfortunately, they are not-the variation in the baseline for control replicates is often as large as, or larger than, the effects from which the conclusions are derived. Given the extreme experimental challenges of small effect size combined with large intertrial variability, it is notable that the authors do not report any likely false negative or false positive data, as would be frequently expected under these conditions. One explanation for the reproducibility of statistical effect seen across many experiments despite these experimental hurdles is manipulation of sample size. The authors acknowledge the extreme variability in sample size offer seemingly harmless explanations, but a closer look shows how problematic this practice is. For example, see Figure 1 (I, J, L) there is a big different between naive and experience males?
2. I am not sure if you keep using the same control with different experiments (that is okay if those exp is done in the same time) as in figure 1 B, I,J,K,L.But I don't think you did Fig 1B in the same time with Fig 1I, J, K,L.
3. It will be clear if you mention in the text how much reduction in percent happened in copulation duration when the males had previous sexual experience?
4. 'Drosophila simulans, the sibling species of D. melanogaster also exhibits SMD, thus suggesting that SMD is conserved between close species of D. melanogaster (Fig. S1B).'. If you want come with this conclusion, you need to test D. erecta, D. sechelia and D. yakuba.
5. The authors mention that Gr66a is salt. This is not 100% correct. GR66a is expressed in many bitter sensing neurons and is required for the physiological and behavioral responses to many bitter compounds. check this reference DOI:https://doi.org/10.1016/j.cub.2019.11.005.
6. Drosophila melanogaster mating duration is between 21- 23 mins. I never saw copulation duration in normal condition (control) 10-15 mins as in figure fig 2E, Fig 7 C,E,F, Fig 8 E and fig 12 G . To the best of my knowledge, of all of the papers on copulation duration, the only one that ascribes a shortened duration to manipulations of the female is Rideout...Goodwin Nature Neuroscience 2010, who argue that this shortening results from markedly increased female activity/agitation during mating, leading the male to terminate early.
7. In some experiments, the authors test very few number of replicates which is not convinced me to their conclusion as example Fig 2F and Fig 12 E
8. Why you test 100, 103 replicates in this exp fig 10 F?
9. How you compare 47 replicates against 9 replicates in fig S10 I?
10. 'The knockdown of LUSH, an odorant-binding protein' Lush is expressed in trichoid sensilla in olfactory organs , from the beginning, they exclude the role of olfaction and later one they said 'suggesting that the expression of the pheromone sensing proteins LUSH and Snmp1 in Gr5a-positive gustatory neurons is critical for generating SMD behavior.' ? Therefore, I recommend If available, please provide a reference for the statement in the Methods section that the Orco1 line was "validated via electrophysiology", or include the electrophysiology data itself in this manuscript as supplementary figure. Ideally, positive behavioral controls for this line would also be included in the manuscript.
11. 'Next, to decipher whether DEG/NaC channel-expressing pheromone sensing neurons require the function of OBP, we expressed lush-RNAi using ppk23-, ppk25- and ppk29-GAL4 drivers to knockdown LUSH in each channel-expressing neuron. The knockdown of LUSH in ppk25- and ppk29-GAL4 labeled cells, but not in ppk23-GAL4 labeled cells, led to a disturbance in SMD behavior, thus suggesting that LUSH functions in ppk25- and ppk29-positive neurons to detect pheromones and elicit SMD behavior (Fig. 9G-I). The knockdown of SNMP1 in ppk29-GAL4- labeled neurons also inhibited SMD behavior (Fig. 9J), thus suggesting that SNMP1 also functions in ppk29-positive neurons to induce SMD behavior.' What about ppk25?

Minor comments:

1. There are no page or line numbers throughout the ms!
2. The use of subheadings in the results section makes reading much easier.
3. 'We found that the mating duration of various wild-type and w 1118 naïve males are significantly longer than that of sexually experienced males (Fig. 1B-D, Fig. S1A)' . I think you should change various wild type to CS and WT Berlin as in legend and figure 1B,C .
4. Suggested exp , Fig S1E-H , they might test 2,6, 12 hours males separation from females to test exactly when this behavior change over time.
5. What is this (GustDx6)? I suggest using Poxn mutant line.
6. General comment in figures, you could remove the common y axis as example in figure 1 B,C,D , difference between means and mating duration.
7. You might move the number of replicates to the legend.
8. Latin name should be italic as example Drosophila simulans.

Referees cross-commenting

I found the comments of the other reviewer reasonable and fair. I agree that the time for fixing all these comments is about six months.

Significance

The idea of the work is interesting, but the design of experiments in some places is inappropriate (see above). The discussion mainly depends on doi: 10.1016/j.neuron.2013.09.034. I study chemoreception and sexual communication in Drosophila and insect vectors of human disease.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

My copy of this ms does not have page numbers or line numbers, this makes it extremely difficult to identify where I am making queries/ suggestions. I don't know whether this is a decision of the journal or authors, but please change this in the future.

Overall I think this would be difficult for a general audience as the rationale and explanation of experiments needs to be clearer.

Results paragraph 1 says that white mutant background had no effect "unlike that of LMD behavior as reported previously", ignoring that there has been a contrary report that extension of mating duration after exposure to a rival does not involve visual cues and so is not affected by the white mutation (Bretman et al 2011 Curr Biol).

A general point in the methodology, it's not very helpful just to say "as in a previous study" without giving at least a brief idea of what that was (e.g. the explanation of egg counting procedures).

A "sperm depletion" assay is described in the results that I cannot find any methodology for.

Was the "excessive mating" with SPR females actually observed, or inferred from previous work? Needs to be clear. In what way do virgins expressing fruitless behave like mated females? It is so unclear how all the evidence in this paragraph leads to the conclusion that both cues from females and successful copulation. Especially as in the next paragraph experience with feminized females (with which the focal males cannot copulate) elicits the response.

It might be helpful to combine the results into a table, so it is easy to see under which conditions males reduce mating duration.

Why are no statistics reported in the results? Identifying sig diffs on figures is not sufficient. I'm very sceptical that "mating duration of males showed normal distribution" for all comparisons, but then it's also difficult to identify which were analysed in this way (if statistics were properly reported this would not be an issue).

Gr5a/ Gr66a mediate acceptance/ avoidance of what? Why would you hypothesise these in particular to be involved?

As Orco was not found to affect the behaviour, why test Or67d?

"Mate guarding" suddenly appears in the modelling section. Can a difference of a couple of minutes in a mating duration of 15-20min really be considered mate guarding? A similar variation in response to rival males is not considered mate guarding, but is linked to adjustments in ejaculate expenditure (admittedly not in a very straight forward way). Surely in a system like this the benefits arise more from how many females the male can mate with in a given time? How does this model relate to any of the previous models of mate guarding?

In this section the work of Linklater et al 2007 is important, they showed progeny declined over successive matings, and related this to exhaustion of Acps rather than sperm. I would urge the authors to consider that what they observe does not necessarily have an adaptive explanation.

I can't find a data accessibility statement.

Significance

A general point, there is simply too much in this ms. It covers too much ground and so doesn't give proper descriptions, discuss the consequences of the data fully or integrate properly with existing literature. Quantity does not equal impact.

That said, a current grand challenge in understanding behaviour is discovering the mechanisms that enable individuals to respond plastically to changing environments. This speaks directly to that challenge.

However, this behavioural observation is not novel, as claimed. Generally the idea of refractoriness is widely known, and specifically the reduction in mating duration over successive matings in D. melanogaster was shown by Linklater et al 2007 Evolution. Moreover, the time between exposure to females has been shown to be important. Linklater et al 2007 gave males mating attempts in quick succession and observed the decrease in mating duration, whereas given recovery time of 3 days, males either mate equally as long, or even longer across their life course (Bretman et al 2011 Proc B, Bretman et al 2013 Evolution). These papers should be discussed, and more broadly the work understood in the light of previous knowledge. The behaviour does not need to be novel for this manuscript to make a significant contribution to the field.

Both in the introduction and discussion the extended mating duration in response to rivals is raised. A great deal of work has been done on this plasticity and yet the way this is written implies just two papers from these authors (whilst referencing others elsewhere).

<![endif]-->

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1:

Minor edits

1. Line 91. Is a bit misleading to say "many other vibrios" possess T3SS. This conveys that this is perhaps the majority, but T3SS in vibrios is at best 50/50. I think best just to delete this sentence.

We deleted this comment, as suggested.

1. Revised to "Thus, in this study, we set out to..." Since the entire paragraph starts with "recent study" I missed that this was summary of new data rather than preview of new results.

The sentence was revised as suggested.

1. Line 503. Correct "xxx-584" or more detail on what this means.

We Thank the reviewer for pointing out this typo._ This refers to the deletion made in tie1, in the region corresponding to nucleotides 485-584 of this gene. The text was corrected accordingly.

1. Line 603. Salmonella should be italicized.

Corrected.

1. The labelling of the figures is pretty complicated with the long genetic designations. Is it reasonable to for example name the ∆vprh/∆hns1 strain with an abbreviation (such as ∆VH)? Or instead create a strain name, common used approaches would be HC## (for Hadar Cohen) or TAU# for Tel Aviv University. If you go this route, be sure to update the strain list. The current method can be followed, the figures are just complicated.

We thank the reviewer for raising this concern. We acknowledge the difficulty in following the many different strains and mutations. Nevertheless, after considering the proposed modifications to the strain names, we believe that they will not add much clarity, and may even cause some confusion. Therefore, we respectfully decided to keep the current nomenclature in place.

Reviewer #2:

Minor edits

1. The authors used a hyperactive T6SS (HNS mutant) to investigate its toxicity. Would the authors be able to use a wild type strain to reproduce the function of T6SS?

We have yet to reveal the external cues that lead to full activation of T6SS3 in vitro. Therefore, in the current study we used genetic tools, such as hns deletion or Ats3 over-expression, to monitor the effect of this system on immune cells. We will dissect the activating conditions in future studies, but we believe that the use of genetic tools should not affect the validity of the results in the current study, nor their timely publication.

1. The authors showed that Tie1 and Tie2 are secreted by T6SS3. It is important to show if they are actually delivered into the host cells during infection. Otherwise it is hard to conclude that they are truly effectors. The primary concern is the lack of in vivo studies to show that Tie1 and Tie2 are actually effectors that play a role in activation of NLRP3 inflammasome._

We present 3 pieces of evidence that, when taken together, support the conclusion that Tie1 and Tie2 are T6SS3 effectors: 1) the proteins are secreted in a T6SS3-dependent manner; 2) their deletion does not hamper overall T6SS3 activity; and 3) their deletion causes the same loss of NLRP3-mediated inflammasome activation and pyroptosis as does inactivation of T6SS3 by deletion of its structural component, tssL3. Although we agree with the reviewer that directly showing delivery of Tie1 and Tie2 into host cells will further strengthen our conclusion, such experiments are quite challenging and difficult to interpret, especially with T6SS effectors that can use diverse mechanisms for secretion through the system. This point was also noted by reviewer #3: “…I believe they were suggesting to demonstrate secretion in host cells. Although this would be nice, it is non-standard and technically not feasible. These types of experiments require genetically fusing the effector with either an enzymatic moiety (e.g. Beta lactamase) or fragment of split GFP. Although such approaches have been previously performed, they often result in either blocked or aberrant secretion due to the presence of the added fragment."

Regarding the reviewer’s comment on the lack of in vivo studies: we agree that these are extremely important, yet they are beyond the scope of the current work, as concurred by reviewers #1 and #3:

Reviewer#1 with regard to Reviewer#2: "I don't think mouse (or aquatic animal) studies are essential for this study. The work contributes nicely to our understanding molecular mechanisms of this T6SS system. As noted in my review, there are many additional lines of study that can be pursued from this work, including animal studies, but this should not preclude publication of this work that is itself an intact unit."

Reviewer#3 regarding reviewer #1's comment on Reviewer#2: "I don't believe that reviewer #2 was suggesting to perform mouse or aquatic animal studies by suggesting in vivo demonstration of secretion…”

Reviewer #3:

Major comments:

1. If the authors believe that GSDME partially compensates in the absence of GSDMD, have they infected a GSDME/GSDMD double knockouts to see if there is an additive effect?

Indeed, this is a very interesting and specific question for the cell death field. We do not currently possess such a GSDME/GSDMD double knockout mouse, and generating one will be a long endeavor. Since its absence does not diminish the importance or the conclusions of the current work, we think that it should not warrant a delay in publication. We do plan to address this question in future studies.

1. It is clear that Ats3 regulates T6SS3, but not the T6SS1; however, there no evidence suggesting that Atg3 does not regulate other gene clusters. For example, have the authors performed RNA seq to compare the transcriptomes of WT and an Ats3 mutant? If not, the authors should refrain using the words "specific activation".

We thank the reviewer for this important note. Indeed, we lack additional data indicating that Ats3’s effect is indeed restricted only to T6SS3. Therefore, we modified the text accordingly and removed mentions of specific T6SS3 activation.

1. In figure 6B, it's unclear why the bacteria infecting cytochalasin D-treated cells grow more than the T6SS3 mutants in the absence of cytochalasin D.

The difference probably stems from the fact that phagocytosis, the major mechanisms by which BMDMs kill bacteria, is hampered in the presence of cytochalasin D, thus allowing bacteria to grow more than when the BMDMs phagocytose them. The results show that in the absence of cytochalasin D, an active T6SS3 counteracts the killing effect by BMDMs with functional phagocytosis.

Minor comments:

1. Figure 1A and other secretion assays: The Western blots include loading control (LC) blots. These are non-standard, non-informative, and not required with the inclusion of the western blots on the "cells" fraction. I would suggest removing these as they may confuse the reader.

We respectfully disagree. Loading controls are standard in bacterial secretion assays, and they are important since they confirm comparable loading and allow proper analysis of the results, especially since we aim to determine whether certain mutations affect the expression of T6SS components. Notably, some groups choose to blot for a cytoplasmic protein (e.g., RpoB in Allsop et al., PNAS, 2017; Liang et al., PLoS Pathogens, 2021) instead of showing overall loaded proteins, as shown in our figures.

1. Line 503: "xxx" should reflect the actual nucleotide nubmers_

We thank the reviewer for pointing out this typo._ This refers to the deletion made in tie1, in the region corresponding to nucleotides 485-584 of this gene. The text was corrected accordingly.

1. Since V. proteolyticus is an aquatic pathogen, have the authors tried to infect corals, fish, and crustaceans (or derived cells) with WT and effector mutants?

This is an interesting point, and indeed we are setting up such systems and we plan to perform such experiments in the future as part of follow up projects. However, these in vivo studies are beyond the scope of the current manuscript, as also noted by the reviewer in the cross-consultation comments: “…my previous comment on infecting aquatic animals or cells derived from them is non-standard and not necessary…”

1. Are the targeted host proteins in this study (performed with murine BMDM) conserved in the natural hosts for V. proteolyticus?

We hypothesize that the conservation is not in the pathway components that are activated upon infection, but rather in the ability of the host cell to sense danger (i.e., to sense the effect of T6SS3 effectors on the host cell or one of its components), which is the role of the NLRP3 inflammasome in mammalian cells. It is well documented that major differences in immune mechanisms exist between mammals and the potential natural marine hosts of V. proteolyticus (e.g., corals, arthropods, and fish); therefore, the conservation at the protein level is low. Nevertheless, basic signaling pathways, such as programed cell death, are conserved between the different phyla. For example, a caspase-1 homolog which was found in arthropods (Chu, B. et al. PLoS One (2014). doi:10.1371/journal.pone.0085343) probably induces an apoptotic-like cell death mechanism, similar to apoptosis in C. elegans. We now provide further discussion on this point in the text (lines 648-659).

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Bacterial type VI secretion systems (T6SSs) are best-characterized in their ability to drive interbacterial competition; however, several systems target host cells and are important for pathogen-host interactions. Based on the presence of T6SS in almost a quarter of sequenced Gram-negative bacterial genomes, including many well-known pathogens, the T6SS is an understudied pathway that is worthy of further inspection. This manuscript by Cohen et al. investigates the role three T6SSs in Vibrio proteolyticus by 1) identifying two of these systems play a role in targeting host cells, 2) characterizing the host pyroptotic pathways targeted by these systems, and 3) identifying two secreted toxins responsible for the host response. Furthermore, this manuscript identifies a specific transcriptional regulator of the T6SS3 and suggests that the host cell responds with an alternative/compensatory pathway upon inhibition of the canonical pyroptosis pathway.

Major comments:

• If the authors believe that GSDME partially compensates in the absence of GSDMD, have they infected a GSDME / GSDMD double knockouts to see if there is an additive effect?
• It is clear that Ats3 regulates T6SS3, but not the T6SS1; however, there no evidence suggesting that Atg3 does not regulate other gene clusters. For example, have the authors performed RNA seq to compare the transcriptomes of WT and an Ats3 mutant? If not, the authors should refrain using the words "specific activation".
• In figure 6B, it's unclear why the bacteria infecting cytochalasin D-treated cells grow more than the T6SS3 mutants in the absence of cytochalasin D.

Minor comments:

• Figure 1A and other secretion assays: The Western blots include loading control (LC) blots. These are non-standard, non-informative, and not required with the inclusion of the the western blots on the "cells" fraction. I would suggest removing these as they may confuse the reader.
• Line 503: "xxx" should reflect the actual nucleotide nubmers
• Since V. proteolyticus is an aquatic pathogen, have the authors tried to infect corals, fish, and crustaceans (or derived cells) with WT and effector mutants?
• Are the targeted host proteins in this study (performed with murine BMDM) conserved in the natural hosts for V. proteolyticus?

Referees cross-commenting

Regarding Reviewer #1's comment on 2022-07-26, I don't believe that reviewer #2 was suggesting to perform mouse or aquatic animal studies by suggesting in vivo demonstration of secretion. (At least, I hope they weren't.) I believe they were suggesting to demonstrate secretion in host cells. Although this would be nice, it is non-standard and technically not feasible. These types of experiments require genetically fusing the effector with either an enzymatic moiety (e.g. Beta lactamase) or fragment of split GFP. Although such approaches have been previously performed, they often result in either blocked or aberrant secretion due to the presence of the added fragment.

Likewise, my previous comment on infecting aquatic animals or cells derived from them is non-standard and not necessary. It would however be interesting to note (either through a supplementary figure or in the text) if the targeted host proteins are conserved between mice and aquatic animals.

Significance

This manuscript provides both novel and innovative details of a previously uncharacterized secretion system. The interested audience includes readers who are interested in host-targeted secretion systems, bacterial effectors, and mechanisms of inflammasome activation. As with any key discovery, there are many avenues of investigation that will follow from this study.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The authors used a hyperactive T6SS (HNS mutant) to investigate its toxicity. Would the authors be able to use a wild type strain to reproduce the function of T6SS?<br /> The authors showed that Tie1 and Tie2 are secreted by T6SS3. It is important to show if they are actually delivered into the host cells during infection. Otherwise it is hard to conclude that they are truly effectors.<br /> The primary concern is the lack of in vivo studies to show that Tie1 and Tie2 are actually effectors that play a role in activation of NLRP3 inflammasome.

Significance

The authors demonstrated in vitro that T6SS of Vibrio plays an important role in inflammasome activation. This is potentially important as this may suggest that T6SS may be a virulence factor. However, as mentioned above, it is important to show in vivo data to demonstrate this.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This paper by Cohen et al described discovery of the function of novel genes in the T6SS operon of Vibrio proteolyticus, a Vibrio isolated from corals. V. proteolyticus also impacts other sea animals. The T6SS3 in particular is found to kill eukarytoic phagocytic cells following engulfment of bacteria into the phagocyte. This strategy of killing phagocytic cells following entry has been shown for other Vibrios. The net goal is protection of the population by the bystander effector. The study first shows that deletion of H-NS (a global negative regulator) stimulates T6SS facilitating ease of work by pushing the system to great cell killing. This allowed them to probe the mechanism of cell death and reveal it as NLRP3 dependent, capase 1 dependent pyroptosis via pore formation by Gasdermin D. Activation of the inflammasome is also linked to cleavage and release of IL-1beta. When GSDMD is absent, there was a slower cell killing by GSDME via capsase 3 activation. The stimulation of this system is additive by two newly recognized T6SS effectors Tie1 and Tie2.

The study is complete, the experiments are well conducted and well controlled. The experiments show reproducibility. The manuscript text is clear, Overall. I suggest no changes in the results or experiments and suggest only a few minor edits of the text.

Minor edits

Line 91. Is a bit misleading to say "many other vibrios" possess T3SS. This conveys that this is perhaps the majority, but T3SS in vibrios is at best 50/50. I think best just to delete this sentence.

Line 102. Revised to "Thus, in this study, we set out to..." Since the entire paragraph starts with "recent study" I missed that this was summary of new data rather than preview of new results.

Line 503. Correct "xxx-584" or more detail on what this means.

Line 603. Salmonella should be italicized.

Figures. The labelling of the figures is pretty complicated with the long genetic designations. Is it reasonable to for example name the ∆vprh/∆hns1 strain with an abbreviation (such as ∆VH)? Or instead create a strain name, common used approaches would be HC## (for Hadar Cohen) or TAU# for Tel Aviv University. If you go this route, be sure to update the strain list. The current method can be followed, the figures are just complicated.

Referees cross-commenting

With regard to Reviewer#2, I don't think mouse (or aquatic animal) studies are essential for this study. The work contributes nicely to our understanding molecular mechanisms of this T6SS system. As noted in my review, there are many additional lines of study that can be pursued from this work, including animal studies, but this should not preclude publication of this work that is itself an intact unit.

Significance

The work is significant in that it links T6SS to a eukaryotic killing system and discovers novel details regarding the mechanisms of death, that may impact our knowledge of other Vibrio T6SS (including V. cholerae) that also target eukaryotic cell actin. There are remaining questions that could be probed, but these are in my opinion major studies that would easily themselves comprise new papers if done properly and thus are not essential for this paper. These include the struture and biochemical activity of Tie1 and Tie2 and the mechanism of caspase-8 independent activation of caspase-3 to then cleave GSDME. Why NLRP3 is required for capase 3 activation is also an open question. I look forward to following this work for some time to come. The authors have revealed very interesting effectors and interesting cell biological process that will merit multiple years and multiple manuscripts to unravel. This work will be of interest to the community interested in bacterial toxin systems (microbial pathogenesis), the bacterial effector mechanism field (biochemistry and cell biology), and the inflammasome activation field (immune systems). The work will be of interest (with essentially no modification) directed at these fields of interest.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1:

Major comments:

In general, the data support the conclusions. I cannot comment on the atomistic simulation experiment as it is outside of my expertise. I had some difficulties interpreting Figure 2 as the contrast in the colour panels made it difficult to assess the different staining patterns. I would recommend changing the blue to cyan for easier visibility. While I agree that there are some differences between Fig 2F and Fig 2G it is not simple for the non-expert to distinguish the gonadal mesoderm from the somatic mesoderm. I think the enlarged panels could do with also showing the overlap in staining, or at least a tracing of the different cell populations so that the gonadal mesoderm can be clearly defined. Please also add some scale bars to the figure. Figure 3 demonstrates clear differences in gonad morphology between male and female mutants but the contrast in the colour panels A-G could also be improved. Panels H-J are very clear.

Response: As suggested by Referees 1 and 3 we have modified the colour channels in all figures. We have also enlarged the figures taking away the uninformative region and focused around the enlarged gonads and added scale bars. For Fig 2F-G, we have added a close up of the region of interest both in colour and in black and white. These changes have increased the contrast and facilitate the data interpretation to non-expert readers.

The rescue experiment in Figure 4 is clearly presented but could the DLC3 mutants in the graph (panel b) please be named similarly to the schematic proteins shown in panel a.

Response: We have changed the names to maintain nomenclature uniformity.

I found the difference between the RhoGAP domain mutants and the StART domain mutants of Cv-c to be clearly defined, and correlate with DLC3 function. This is a very interesting result that indicates multiple molecular functions for the Cv-c /DLC family.

Response: The methods are well described, statistics adequate and the data well described._

Minor Comments:

My only suggestion for the text is to provide a more through description of the StART domain in the introduction.

Response: We have included the following paragraph in the introduction describing the StART domain:

“This family of proteins share different domains: besides the Rho GTPase Activating Protein domain (GAP), they present a protein-protein interacting Sterile Alpha Motif (SAM) at the N terminal end and a Steroidogenic Acute Regulatory protein (StAR)-related lipid transfer (StART) domain at the C terminal. StART domains have been shown in other proteins to be involved in lipid interaction, protein localization and function.”

Reviewer #2:

My only issue with the present study is to how well the present experimental findings in Drosophila translate to humans. As far as I can tell the present studies show that inactivating mutations in Cv-c in Drosophila result in failure of germ cell enclosure by somatic cells into the testis, resulting in sterility. In humans, and in experimental mouse transgenic lines, it has been well established that absence of germ cells does not of itself lead to failure of testis differentiation and onward development, nor does it lead automatically to sex reversal or impairment of masculinization. For the latter to occur, there must be impairment/failure of fetal Leydig cell function such that insufficient androgen is produced to effect genital/bodywide masculinization. Obviously, this will happen if no testis forms as appears to be the case in the new human DLC3 mutant reported in the present manuscript (although detail on this is unfortunately lacking). This appears to be different to the previous published DLC3/STARD8 mutant sisters, in whom the phenotype appears to reflect failure of steroidogenesis. Is the proposal that DLC3/STARD8 plays a role in both testis differentiation and in Leydig cell function (steroidogenesis) or is this due to different DLC3 genes? I think the authors need to address these key issues in their discussion, if only to highlight that there are at present many gaps in our understanding.

The reviewer says:

“As far as I can tell the present studies show that inactivating mutations in Cv-c in Drosophila result in failure of germ cell enclosure by somatic cells into the testis, resulting in sterility.”

Response: This sentence does not represent the spirit of our findings accurately and this probably reflects the fact that we stressed the interaction between somatic mesodermal cells and germ cells in Drosophila which probably concealed that the main defects in Cv-c mutants are caused by the abnormal interaction of the mesodermal cells with germ cells but also among themselves. Our study provides insights about a new conserved pathway required in the mesodermal cells for the maintenance of an already formed testis, and only indirectly can be considered to deal with sterility. We show that Cv-c is required in the mesodermal cells for the correct maintenance of the testis structure, that when it fails leads to the testis dysgenesis which, among other defects, releases the germ cells. We show that in the absence of Cv-c function in the testis, the mesodermal pigment cells do not form a continuous layer around the testis and the ECM surrounding the testis breaks. We also show that the interstitial gonadal cells fail to ensheath the germ cells and as a result of all these the germ cells become dispersed. These perturbations can be partially corrected by expression in the testis mesoderm of human DLC3 or Drosophila Cv-c that in both cases require a functional StART domain. Thus, our results suggest that Cv-c/DLC3 have a fundamental function on the mesodermal testis cells that has been conserved. These results indicate that, as in Drosophila, the primary cause for the gonadal dysgenesis in DLC3 human patients is due to the abnormal maintenance of the testis mesoderm cells, which include both Sertoli and Leydig cells. Thus, our proposal is that DLC3/STARD8 plays a role in testis maintenance through its function in mesodermal cells which will probably affects both Sertoli and Leydig cell function.

To clarify the issue raised by the referee we have modified both, the introduction and the discussion to highlight that although humans and Drosophila diverged millions of years ago there are similarities regarding gonad stabilisation.

We have modified the introduction to clarify this issue:

“Gonadogenesis can be subdivided into three stages: specification of precursor germ cells, directional migration towards the somatic gonadal precursors and gonad compaction. In mammals, somatic cells, i.e. Sertoli cells in male and Granulosa cells in females, play a central role in sex determination with the germ cells differentiating into sperms or oocytes depending on their somatic mesoderm environment. In humans, Primordial Germ Cells (PGCs) are formed near the allantois during gastrulation around the 4th gestational week (GW) and migrate to the genital ridge where they form the anlage necessary for gonadal development (GW5-6). Somatic mesodermal cells are required for both PGCs migration and the formation of a proper gonad. Once PGCs reach their destination, the somatic gonadal cells join them (around GW 7-8 in males, GW10 in females) and provide a suitable environment for survival and self-renewal until gamete differentiation {Jemc, 2011 #413}. Thus, mutations in genes regulating somatic Sertoli and Granulosa support cell function in humans are often associated with complete or partial gonadal dysgenesis in both sexes and sex reversal in males {Zarkower, 2021 #430; Knower, 2011 #418; Brunello, 2021 #399}. Other mesodermal cells, the Leydig cells, also play an important role in the testis by being the primary source of testosterone and other androgens and maintaining secondary sexual characteristics.”

Also we have added a paragraph in the discussion to emphasize this argument:

“We show that in the absence of Cv-c function in the testis, the mesodermal pigment cells do not form a continuous layer around the testis and the ECM surrounding the testis breaks. We also show that the interstitial gonadal cells fail to ensheath the germ cells and as a result of all these the germ cells become dispersed from the testis. These perturbations can be partially corrected by expression in the testis mesoderm of human DLC3 or Drosophila Cv-c that in both cases require a functional StART domain. Thus, our results suggest that Cv-c/DLC3 have a fundamental function on the mesodermal testis cells that has been conserved. These results indicate that, as in Drosophila, the primary cause for the gonadal dysgenesis in DLC3 human patients is due to the abnormal maintenance of the testis mesoderm cells, which include both Sertoli and Leydig cells”.

I would also suggest that the authors highlight another potentially more important spin-off from such studies, namely that understanding of the regulation of DLC3/STARD8 genes, and what might perturb their expression/action would appear to present a whole new area for exploration in relation to testicular dysgenesis/masculinization disorders.

Response: We have modified the last part of the discussion to introduce referee 2’s suggestion:

“Our work points to DLC3/Cv-c as a novel gene required specifically in testis formation. Adding DLC3 to the list of genes involved in 46X,Y complete dysgenesis opens up a new avenue to analyse the molecular and cellular mechanisms behind these disorders that could help in diagnosis and the development of future treatments”.

Reviewer #3 :

Major comments:

1. This study has shown the expression pattern of cv-c and the consequence of cv-c mutation on different aspects of gonad development. However, one major comment is there is no quantification of the expression levels as well as the scoring of the mutant phenotypes.
2. In Figure 2, for instance, I recommend that the authors display the quantification of the fluorescence intensity of the cv-c expression under all circumstances (in situ hybridization as well as protein-trap based GFP expression) to better depict the differences among the male vs female gonad.

Response: We don’t think quantifying the stainings will add much to the results. We believe that the changes performed increasing the images’ contrast and their amplification are sufficient to illustrate our statement about cv-c being expressed in testis but not in ovaries.

1. In Figure 3, the authors show the different gonad developmental defects associated with the cv-c mutation. Specifically, the authors show that the gonad mesoderm cells are displaced with the pigment cells failing to ensheath the germ cells. In addition, the authors also suggest that there is an increased frequency of germ cell blebbing, an indication of migratory activity. However, there is no quantification of these findings. I think the authors should display a quantitative estimation of % of the mutant gonad depicting these phenotypes vs the normal gonad to have a perspective of how penetrant the phenotypes are.

Response: As referee suggested, we have quantified bleb phenotype. The results are presented in figure 3, panel J.

1. In Figure 4, the authors attempt to rescue the Cv-C mutation linked gonadal defects by overexpressing different Cv-C protein variants. The rescue experiments are not very clear. The graph shows the % of normal testes under different genotypic combinations. It is not very clear what the authors mean by normal (in what context)? Since the mutation results in different defects of gonad development, I think recommend that represent the rescue in terms of these defects. It would be interesting to see for instance, what happens to the blebbing or germ cell ensheathment phenoype upon rescue. How many % of testes show the rescue as compared to cv-c mutants?

Response: The percentages are quantified considering if the testes have any germ cell outside the gonad. We have added a line to clarify this point in the figure legend: “…quantified as encapsulated gonads with all germ cells inside the testis as assessed by Fisher-test”.

Nevertheless, we are going to quantify the number of ECM breaks and show the results in the reviewed manuscript.

1. Did the authors try cell-specific depletion of cv-c and examined the consequence on gonad development?

Response: cv-c mutants are embryonic lethal because of Cv-c’s widespread requirement on various embryonic tissues during development. Induction of FRT clones in the embryonic testis mesoderm was unsuccessful because of the low number of divisions during embryogenesis. We also tried to knock down cv-c expression with 3 different RNAi lines. Unfortunately, overexpression of these RNAi with different testis Gal4 drivers did not decrease cv-c mRNA levels significantly in the mesoderm or in other tissues where cv-c is expressed. Despite these experiments unsatisfactory outcome, our finding that cv-c is expressed in the testis mesoderm cells, and the fact that we can rescue the testis phenotypes by expressing Cv-c with gonadal mesodermal specific Gal4 lines supports a testis mesoderm requirement of cv-c for its gonadal function.

1. Another major concern is the lack of mechanistic insight of cv-c. For example, how does loss of cv-c result in gonadal dysgenesis? The authors suggested that StART domains regulate via lipid binding. The authors could examine if StART domain function is dependent on lipid-mediated interactions.

Response: We agree with the referee that the molecular characterisation of the StART-mediated GAP-independent Cv-c function we have uncovered in this work is a very interesting finding that should be addressed by future work. However, such biochemical characterisation requires a complex approach to distinguish between the already known StART function regulating the GAP activity shown before (Sotillos Scientific Reports) and the new GAP-independent function we describe in the testes that falls beyond this work.

The central point of this manuscript is the demonstration that both DLC3/Cv-c are involved in male gonad formation, an important conserved function for both of them that had been overlooked by previous publication. Thus, DLC3 should be considered a new gene to be analysed in the future when studying gonadal dysgenesis. A second important point raised by our work is the demonstration that DLC3/Cv-c can perform RhoGAP independent functions, something that had never been described for these proteins.

Not withstanding this, in the revised version, we have added a new supplementary figure (1) related to the StART domain-lipid interaction analysed in-silico. The in-silico model shows that the DLC3-StART domain Ω1-loop structure displays the highest frequency of interaction with the membrane. This loop is conserved in the StART domains of several other STARD proteins and seems to modulate access to the ligand binding cavity. Ω-loops play multiple roles in protein function, often related to ligand binding, stability and folding. In this context, mutations in the proximity of the Ω1-loop, like the ones carried by the patients, may have drastic effects on overall protein stability that could affect the interaction between gonadal precursor cells.

1. Do the cv-c mutants survive to adulthood? If yes, then it would be interesting to know how the adult testis behaves in cv-c mutants. Does it result in sterility?

Response: Unfortunately, all studied cv-c mutants are embryonic lethal.

1. Ensheathment is required for proper germline development and defects in ensheathment can affect soma-germline communication and germline development. Germ cell ensheathment affects the proliferation of germ cells and display defective JAK/STAT signaling. It would be interesting to know if the germ cells in cv-c mutant gonad show the proliferation defect and impaired JAK/STAT signaling.

Response: This is an interesting suggestion. JAK/STAT signalling has a male specific function that could explain why cv-c gonadal defects are male specific. We are going to study how cv-c affects STAT signalling in the male gonad. We are currently preparing stocks combining 10XSTAT::GFP reporter with cv-c mutants and preparing samples for anti-STAT labelling. We will also analyse if embryos lacking STAT activation, activate cv-c expression in the testes.

1. I was also wondering if the authors have examined the number of germ cells in the mutant gonads.

Response: Yes, we have counted the number of germ cells in cv-c mutants and, if anything, there are more. We initially considered that an excess of GC proliferation could be the cause of gonad disruption. However, we have discarded this hypothesis as phospho-histone 3 stainings did not show a significant increase of GC divisions. Moreover, when we blocked cell proliferation in cv-c’ mutant gonads using UAS-p21, the testes phenotype was not rescued. We are unsure what could be responsible for the slight increase of germ cells observed.

1. In addition, I think the quality of the images should be improved.

Response: We have changed the colours used in the confocal images and amplified the relevant regions in all panels. We thank both referees for this suggestion as these changes have improved the figure contrast.

Minor comments:

1. cv-c mRNA in Figure 2 panels (Fig. 2D) should be in italics.

Response: We have changed it.

1. There is no scale bar in Figure panels. In addition, there is no scale bar in the zoomed images in Figure 2. Scale bars should be consistently put in the all the Figures, in particular on the first panels of the Figures.

Response: We have added scale bars to all panels.

1. In the line 677, the manuscript says "arrowhead". There are no arrowheads but the arrows.

Response: Corrected

1. Please be consistent with the labels in Figure panels: Vasa is shown in capital while Eya is not.

Response: Corrected

1. Please be consistent with the labeling of the Figure panels: Figure 3A vs Figure 4a.

Response: Corrected

1. What does the asterisk signify in Figure 2? There is no mention of asterisk in the Figure 2 legend.

Response: The meaning of the asterisk was explained in the figure legend.

1. There is no grey channel (sagittal view) for the panels Figure 3I and J.

Response: We have already included sagittal views in the figure.

1. Please be thorough in labeling the genotypes in Figures. For instance, Figure 4c depict the % of normal testis in cv-c delta StART. However, the correct genotype is twi>Cv-c StART. In addition, in Figure 4c graph, cv-c mut should be cv-cGAPmut.
2. Please be consistent with the depiction of the "START" domain of the protein throughout the manuscript. In figure 4c for instance, it is "START" in the graph while in the figure panel 4i, it is StART.
3. In Figure 4b, it is written DLC3-GA. Did the authors mean DLC3-S993N?
4. In line 723, it should be anti-beta catenin.

Response: As suggested, we have unified figure labelling.

1. The authors have shown two images to suggest that cv-c mutant gonad depict the germ cell blebbing (Figure 3I and J). I think it would be much better to put up a graph showing the number or percentage of cv-c mutant gonads displaying the germ cell blebbing than putting two images with the same information.

Response: We have already done the quantification and added the data as a graph in figure (3J).

1. The previous comment is also true for Figure 6H and I. In both the panels, the authors wish to show discontinuous ECM marked by Perlecan expression in cv-c mutant gonads. I think it would be better to display a score of the number of mutant gonads depicting the discontinuous ECM.

Response: We are repeating stainings to quantify Perlecan disruption in cv-c mutants and we will display the results as a graph in figure 6.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

This manuscript uses genetics, cell biology and confocal imaging to study the loss-of-function phenotypes of Cv-c, the Drosophila ortholog of DLC3. DLC3 mutation has been recently associated with male gonad dysgenesis in DSD patients. This work showed that Cv-c acts in the somatic gonadal cells to regulate male gonad development. Interestingly, the Cv-c mutant phenotypes result in testicular dysgenesis, marked by defective germ cell ensheathment and blebbing. Using different variants of the Cv-c protein and genetics based analyses, the authors further identified that the gonad development is dependent on StART domain but is independent of the GAP domain. Overall, these results should be of interest to researchers in disorders of sexual development, germ cell biology and developmental biology fields. However, at several places in order to reach the conclusions, more rigorous experiments should be performed (see revision details). In conclusion, this work has the potential but requires statistical quantification to support the central claims.

Major comments:

1. This study has shown the expression pattern of cv-c and the consequence of cv-c mutation on different aspects of gonad development. However, one major comment is there is no quantification of the expression levels as well as the scoring of the mutant phenotypes.
2. In Figure 2, for instance, I recommend that the authors display the quantification of the fluorescence intensity of the cv-c expression under all circumstances (in situ hybridization as well as protein-trap based GFP expression) to better depict the differences among the male vs female gonad.
3. In Figure 3, the authors show the different gonad developmental defects associated with the cv-c mutation. Specifically, the authors show that the gonad mesoderm cells are displaced with the pigment cells failing to ensheath the germ cells. In addition, the authors also suggest that there is an increased frequency of germ cell blebbing, an indication of migratory activity. However, there is no quantification of these findings. I think the authors should display a quantitative estimation of % of the mutant gonad depicting these phenotypes vs the normal gonad to have a perspective of how penetrant the phenotypes are.
4. In Figure 4, the authors attempt to rescue the Cv-C mutation linked gonadal defects by overexpressing different Cv-C protein variants. The rescue experiments are not very clear. The graph shows the % of normal testes under different genotypic combinations. It is not very clear what the authors mean by normal (in what context)? Since the mutation results in different defects of gonad development, I think recommend that represent the rescue in terms of these defects. It would be interesting to see for instance, what happens to the blebbing or germ cell ensheathment phenoype upon rescue. How many % of testes show the rescue as compared to cv-c mutants?
5. Did the authors try cell-specific depletion of cv-c and examined the consequence on gonad development?
6. Another major concern is the lack of mechanistic insight of cv-c. For example, how does loss of cv-c result in gonadal dysgenesis? The authors suggested that StART domains regulate via lipid binding. The authors could examine if StART domain function is dependent on lipid-mediated interactions.
7. Do the cv-c mutants survive to adulthood? If yes, then it would be interesting to know how the adult testis behaves in cv-c mutants. Does it result in sterility?
8. Ensheathment is required for proper germline development and defects in ensheathment can affect soma-germline communication and germline development. Germ cell ensheathment affects the proliferation of germ cells and display defective JAK/STAT signaling. It would be interesting to know if the germ cells in cv-c mutant gonad show the proliferation defect and impaired JAK/STAT signaling.
9. I was also wondering if the authors have examined the number of germ cells in the mutant gonads.
10. In addition, I think the quality of the images should be improved.

Minor comments:

1. cv-c mRNA in Figure 2 panels (Fig. 2D) should be in italics.
2. There is no scale bar in Figure panels. In addition, there is no scale bar in the zoomed images in Figure 2. Scale bars should be consistently put in the all the Figures, in particular on the first panels of the Figures.
3. In the line 677, the manuscript says "arrowhead". There are no arrowheads but the arrows.
4. Please be consistent with the labels in Figure panels: Vasa is shown in capital while Eya is not.
5. Please be consistent with the labeling of the Figure panels: Figure 3A vs Figure 4a.
6. What does the asterisk signify in Figure 2? There is no mention of asterisk in the Figure 2 legend.
7. There is no grey channel (sagittal view) for the panels Figure 3I and J.
8. Please be thorough in labeling the genotypes in Figures. For instance, Figure 4c depict the % of normal testis in cv-c delta StART. However, the correct genotype is twi>Cv-c StART. In addition, in Figure 4c graph, cv-c mut should be cv-cGAPmut.
9. Please be consistent with the depiction of the "START" domain of the protein throughout the manuscript. In figure 4c for instance, it is "START" in the graph while in the figure panel 4i, it is StART.
10. In Figure 4b, it is written DLC3-GA. Did the authors mean DLC3-S993N?
11. In line 723, it should be anti-beta catenin.
12. The authors have shown two images to suggest that cv-c mutant gonad depict the germ cell blebbing (Figure 3I and J). I think it would be much better to put up a graph showing the number or percentage of cv-c mutant gonads displaying the germ cell blebbing than putting two images with the same information.
13. The previous comment is also true for Figure 6H and I. In both the panels, the authors wish to show discontinuous ECM marked by Perlecan expression in cv-c mutant gonads. I think it would be better to display a score of the number of mutant gonads depicting the discontinuous ECM.

Significance

The significance of this paper is the identification of a conserved protein that has a conserved function in regulating spermatogenesis. The Drosophila embryonic testis is an ideal system to study the role of cv-c in gonadogenesis. It also sets the stage for studying the potential roles of cv-c in adult testes.

The existing published knowledge about somatic ensheathment of germ stem cells is sufficiently covered and cv-c is a new player in this process.

The audience for this study is developmental biologists and researchers studying disorders of sex development.

Our expertise is somatic sex identity in adult Drosophila gonads and DSD.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

This is an interesting and generally convincing study demonstrating the potential role of DLC3 (STARD8)/Cv-c in testicular development, and in particular its role in early (fetal in human) germ cell development. The genetic manipulations and associated techniques involving Drosophila appear to be state of the art, although I would not class myself as expert enough in such techniques to be able to give a truly informed opinion.

My only issue with the present study is to how well the present experimental findings in Drosophila translate to humans. As far as I can tell the present studies show that inactivating mutations in Cv-c in Drosophila result in failure of germ cell enclosure by somatic cells into the testis, resulting in sterility. In humans, and in experimental mouse transgenic lines, it has been well established that absence of germ cells does not of itself lead to failure of testis differentiation and onward development, nor does it lead automatically to sex reversal or impairment of masculinization. For the latter to occur, there must be impairment/failure of fetal Leydig cell function such that insufficient androgen is produced to effect genital/bodywide masculinization. Obviously, this will happen if no testis forms as appears to be the case in the new human DLC3 mutant reported in the present manuscript (although detail on this is unfortunately lacking). This appears to be different to the previous published DLC3/STARD8 mutant sisters, in whom the phenotype appears to reflect failure of steroidogenesis. Is the proposal that DLC3/STARD8 plays a role in both testis differentiation and in Leydig cell function (steroidogenesis) or is this due to different DLC3 genes? I think the authors need to address these key issues in their discussion, if only to highlight that there are at present many gaps in our understanding.

I would also suggest that the authors highlight another potentially more important spin-off from such studies, namely that understanding of the regulation of DLC3/STARD8 genes, and what might perturb their expression/action would appear to present a whole new area for exploration in relation to testicular dysgenesis/masculinization disorders.

Significance

This represents a significant advance in our understanding and identifies model systems in which to gain further insight.

I would also suggest that the authors highlight another potentially more important spin-off from such studies, namely that understanding of the regulation of DLC3/STARD8 genes, and what might perturb their expression/action would appear to present a whole new area for exploration in relation to testicular dysgenesis/masculinization disorders in humans.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

The authors have utilised Drosophila to provide evidence that a DLC3 mutation identified in 46,XY patients with male gonadal dysgenesis is likely causative for the phenotype. They demonstrated that mutation of the Drosophila ortholog of DLC3, Cv-c, was associated with defects in embryonic testis development and that the phenotype could be rescued by expression of wildtype human DLC3 but not the patient variant. DLC3/Cv-c are members of the RhoGAP family of proteins but GAP activity was not required for in the testis while mutation of the StART domain disrupted gonad morphogenesis. Mutations in this domain were associated with human gonadal dysgenesis suggesting a conservation of function during gonad development.

Major comments:

In general, the data support the conclusions. I cannot comment on the atomistic simulation experiment as it is outside of my expertise. I had some difficulties interpreting Figure 2 as the contrast in the colour panels made it difficult to assess the different staining patterns. I would recommend changing the blue to cyan for easier visibility. While I agree that there are some differences between Fig 2F and Fig 2G it is not simple for the non-expert to distinguish the gonadal mesoderm from the somatic mesoderm. I think the enlarged panels could do with also showing the overlap in staining, or at least a tracing of the different cell populations so that the gonadal mesoderm can be clearly defined. Please also add some scale bars to the figure.<br /> Figure 3 demonstrates clear differences in gonad morphology between male and female mutants but the contrast in the colour panels A-G could also be improved. Panels H-J are very clear.<br /> The rescue experiment in Figure 4 is clearly presented but could the DLC3 mutants in the graph (panel b) please be named similarly to the schematic proteins shown in panel a.<br /> I found the difference between the RhoGAP domain mutants and the StART domain mutants of Cv-c to be clearly defined, and correlate with DLC3 function. This is a very interesting result that indicates multiple molecular functions for the Cv-c /DLC family.<br /> The methods are well described, statistics adequate and the data well described.

Minor Comments:

My only suggestion for the text is to provide a more through description of the StART domain in the introduction.

Referees cross-commenting

Quantification of dispersed low level punctate expression levels will be very difficult to achieve and I do not believe will alter the conclusions. I agree that quantification of the percentage of gonads that display the illustrated phenotypes would be beneficial. The manuscript nicely describes a phenotype that has been defined to be due to mutation of the StART domain. Defining the mechanism of how the StART domain regulates protein function would take the study to a new level but may not be necessary for publication of the findings.

Significance

The significance of this work is two-fold.

1. A demonstration that Cv-c and DLC3 have similar functions in regulation of gonad morphogenesis, and that the patient mutations almost certainly correlate with the observed phenotypes.
2. An intriguing demonstration that this family of proteins is not simply functioning as RhoGAPs, and that there is a specific roles for the StART domain in male gonad development.

DLC3 has already been shown to rescue Malpighian tubule defects associated with Cv-c mutants so the functional equivalence of these proteins has already been established. The novelty of the present study relates to a demonstration that the StART domain is specifically required for male gonad morphogenesis, and the utility of experiments in Drosophila to assist in associating a human mutation with a clinical phenotype.

This work should find a willing audience from clinical geneticists as an example of how model organism genetics can assist genetic diagnoses. It will also be of great interest to the field of reproductive biologists who wish to understand how the sex determination pathways are resolved by differential tissue development. The obvious next steps are to determine why mutations in Cv-c only affect male gonads and not females, and what is the specific role of the StART domain in this process.<br /> My field of study has focussed on adult gonad development and regeneration, and I found the manuscript presented in a style that was simple to follow with my suggestion that contrast of some of the colour panels could be improved.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

The rapid syncytial nuclear cycles that occur during the first ~2.5 hours of Drosophila embryogenesis and give rise to the blastoderm are supported by large amounts of maternally deposited histone proteins which are stored in the egg cytoplasm for deposition into replicating DNA during each round of S phase. Although the H2A/H2B storage chaperone Jabba was identified by Michael Welte's lab several years ago, maternal H3/H4 storage chaperones have not been identified. Tirgar et al provide evidence that the Drosophila NASP protein provides histone H3 and H4 storage function during these earliest stages of Drosophila embryogenesis. The data include genetic analyses that NASP function is required maternally, but not zygotically, and molecular analyses that NASP binds H3 and that H3 and H4 levels are reduced in the embryo and late-stage oocytes in the absence of NASP. These data are convincing and support the conclusion that NASP is a maternally acting H3/H4 storage chaperone needed in the early embryo.

Two additional lines of investigation would strengthen this conclusion and perhaps increase the impact and appeal of the manuscript.

The first is a microscopic analysis of the nuclear division cycles in eggs derived from NASP mutant mothers. The authors report DAPI staining and assessment of nuclear cycles, but do not show these data. In fact, the two embryos shown in Figure 4B do not look like DAPI stained embryos-there are no nuclei apparent in the images. Loss of maternal histone causes defects in chromosome morphology that result in characteristic defects such as lagging chromosomes and the failure of sister chromatid segregation leading to fused daughter nuclei (see PMID: 11157774 for an example). These defects should not be difficult to detect via DNA staining or even using fluorescently labeled H2 type histones. Characterizing such defects would lend support to the hypothesis and I think is important for this paper.

We thank the reviewer for their constructive review and feedback. We have switched to Propidium Iodide (PI) staining to increase the signal-to-noise for DNA staining in early embryos. Given the improved signal we see with PI over DAPI, we will be able to provide both improved images of nuclear staining and assay for defects in chromosome morphology as suggested. We will include this data in the revised version of the manuscript. Second, determining the location of NASP in the early embryo might provide further insight into the mechanism of storage. i.e. is NASP located in the cytoplasm rather than the nucleus, perhaps in association with lipid droplets like Jabba? Do the antibodies the authors developed work in IF experiments to ask this question? At the moment what is shown is that NASP is present in 0-2 hour embryos via western blot analysis, supporting the conclusion that it functions in the early embryo as a storage chaperone. This analysis would be nice to have but is not essential in my view.

We have tried to use our antibody to monitor the localization of NASP in the early embryo. Unfortunately, the staining has yet to work. We will continue to alter fixation and permeabilization conditions in the early embryo with the goal of including this data in the revised manuscript. We have, however, been able to monitor NASP localization in Drosophila S2 cultured cells with our antibody. If we are unable to get the antibody staining to work in embryos, we will include the NASP localization data in S2 cells in combination with EdU labeling to mark cells in S phase.

Small points: Is NASP really a maternal effect "lethal"? Some of the eggs do hatch, and so some develop to stages where maternal histones are no longer necessary and zygotic production takes over (i.e. cycle 15). Perhaps consider the language used here.

We see the reviewers point with respect to the term ‘lethal’. We do see a very small fraction of progeny laid by NASPmutant mothers make it to adulthood, although they die shortly after hatching. We’ve removed the term ‘lethal’ and refer to NASP solely as a maternal effect gene. On this point, do NASP mutant females lay the same number of eggs as wild type? i.e. is there a requirement for oogenesis/egg production (other than depositing H3/H4 into the egg), or just for the early zygotic cycles?

We have noticed that NASP mutant mothers have lower fecundity. We have included this data in the revised manuscript as Supplemental Figure 2A.

The first paragraph of the results is redundant with much of the introduction, which I think could do a better job at describing in more detail the syncytial cycles and the special needs they have for histone storage and chaperone function versus the post-blastoderm embryonic cycles and the rest of development. i.e. make a better distinction between the first two hours of embryogenesis versus the rest of embryogenesis, and the when the switch from maternal to zygotic control of development and histone production occurs (cycle 15 at 3-4 hours AED).

We appreciate the reviewer for this suggestion. The manuscript has been edited to be less redundant and include details of embryogenesis as suggested. CROSS-CONSULTATION COMMENTS Seems like all reviewers are in general agreement, particularly about providing additional data regarding chromosome/nuclear behavior in the NASP mutants and NASP localization in the early embryo to increase impact of the study. While rescue of the NASP mutant phenotype with a transgene would be nice, as suggested by referee #2, I don't think it's essential given the genetic approaches employed.

Reviewer #1 (Significance (Required)):

see above

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Tirgar et al. report on a functional characterization of the Drosophila homolog of the histone H3/H4 chaperone NASP. They generated a loss of function allele of NASP by CRISPR/Cas9, which induces a partial maternal effect embryo lethal phenotype. Using quantitative mass spectrometry, they demonstrate that NASP stabilizes reservoirs of H3 and H4 in the early embryo. The manuscript is very clear and confirms the functional importance of maternal NASP for the early embryo. Genetic analyses are well conducted (but see my comments below) and the impact of NASP maternal mutant on H3 and H4 stockpiles is convincingly established by both quantitative mass spectrometry and Western-blotting.

Major comments:

• Although the authors used two independent deficiencies of the NASP genetic region to characterize their NASP CRISPR alleles, it is relatively standard in this type of functional analyses to perform rescue experiments using a transgene expressing the WT protein.

We thank the reviewer for this suggestion. As discussed in the cross consultation, we agree that the use of the two different deficiency lines and the NASP1 CRISPR control are clear lines of evidence that the phenotypical data are due to lack of NASP.

• In WB analyses, NASP appears systematically shorter in the NASP[1]/Df genotype compared to WT. Can the authors comment on this?

While we reproducibly see this change in migration, we can only guess as to why this may be. One possible reason is that the NASP1 mutant protein could be missing a post-translational modification. Proteomic data from Krauchunas et al. (Dev Biol. 2012; PMC3441184) shows that NASP has the potential to be regulated by phosphorylation. Therefore, the NASP1 mutant protein could be missing a phosphorylation. Intriguingly, the 6bp insertion is next to a Thr residue that could affect its ability to be phosphorylated (if it is phosphorylated at all). Since we can only offer speculation, we do not feel comfortable adding this to the manuscript.

• The authors do not mention the centromeric histone H3 variant Cid in their analyses. Do they have evidence that it is not affected by loss of maternal NASP?

We thank the reviewer for raising this great point. Our mass spec data reveals that Cid levels stay the same in the absence of NASP in both embryos and stage 14 egg chambers. We have edited Figures 3D and 3E to include Cid. Unfortunately, we did not identify any Cid-specific peptides in our IP-mass spec data.

• The authors could have chosen to explore in more details the phenotypic defects of embryos derived from NASP mutant mothers. Instead, a single abnormal embryo is shown with no cytological details. This is a bit problematic since an earlier study (Zhang et al 2018, cited in the manuscript) actually provided more phenotypic details of embryos from NASP KD mothers.

This issue was also raised by Reviewer 1. We have switched to Propidium Iodide (PI) staining to increase the signal-to-noise for DNA staining in early embryos. Given the improved signal we see with PI over DAPI, we will be able to provide both improved images of nuclear staining and assay for defects in chromosome morphology as suggested. We will include this data in the revised version of the manuscript. - Similarly, the authors could have used their anti-NASP antibody to analyze the distribution of NASP during cleavage divisions. Does it behave like ASF1, for instance, which enters S phase nuclei at each cycle or does it remain in the cytoplasm? These are relatively simple experiments/analyses that could increase the significance of the study.

This point was also raised by Reviewer 1. We have tried to use our antibody to monitor the localization of NASP in the early embryo. Unfortunately, the staining has yet to work. We will continue to alter fixation and permeabilization conditions in the early embryo with the goal of including this data in the revised manuscript. We have, however, been able to monitor NASP localization in Drosophila S2 cultured cells with our antibody. If we are unable to get the antibody staining to work in embryos, we will include the NASP localization data in S2 cells in combination with EdU labeling to mark cells in S phase.

Minor comments:

• line 60: I suggest to introduce Drosophila in the next sentence, where it seems more appropriate (not all embryos develop "extremely rapidly").

We have edited the second sentence to state “the early Drosophila embryo”.

• line 68: the 50% estimation of free histones does not really make sense without defining the embryonic stage.

We have edited the manuscript to state the specific cell cycle in which there has been 50% free histones measured. - line 89: Are the authors specifically referring to Drosophila NASP?

Yes, we have edited the text to include Drosophila in this instance. - lines 99-106: I found this paragraph redundant with the introduction.

We appreciate this suggestion. It was also pointed out by Reviewer 1. We have made changes to the manuscript to address the redundancy.

• line 142: H3-H4

Thank you for noticing this. We have edited the text to include 4.

• line190-191: It seems to me that data of Figure S2C are already included in Fig. 2E.

The data in FigureS2C was performed with virgin females compared to the data in Figure 2E that was generated with non-virgin mothers. This was important to control the genotype of the embryos.

• line 232: it is surprising that the Zhang et al paper (reporting maternal KD of NASP) is only mentioned here. As a reader, I would certainly prefer to have it presented right from the introduction.

We have edited the manuscript to include this reference in the introduction.

• Figure 4B needs a scale bar.

Figure 4B will be replaced with better images of the embryo stained with PI. It will also include images of chromosome morphology/segregation. We will be sure to include scale bars.

• line 302: Mentioning the identity and function of known H3/H4 histone chaperones acting in the early embryo (ASF1, HIRA, CAF-1, ...) could provide perspective to the present study.

Thank you for this suggestion. We have edited the manuscript to include functions of other histone chaperones in the early embryo to provide context.

• line 304: in contrast to this statement, I found quite surprising and interesting that NASP is not absolutely essential for embryo development considering its role. This should be discussed.

In the absence of Jabba alone, upregulation of translation can compensate for the destabilization of H2A, H2B, and H2Av. It is only when translation is inhibited in embryos laid by Jabba mutant mothers that embryos die (Li.Z, et al. Curr Biol 2013). Therefore, it is possible that translation can partially compensate for the degradation of H3 and H4 in the absence of NASP. This may be why a fraction of embryos laid by NASP mutant mothers are able to hatch and why we still detect some H3 in embryos laid by NASP mutant mothers. We have edited the manuscript to discuss this more in depth.

CROSS-CONSULTATION COMMENTS I fully agree with the other reports. The NASP rescue experiment is just a suggestion but is not essential.

Reviewer #2 (Significance (Required)):

This work clarifies the identity and function of Drosophila NASP and clearly demonstrates that NASP is important for the stabilization of maternal stockpiles of H3 and H4 during early embryo development. The conservation of NASP function as a histone H3/H4 chaperone in Drosophila is not really a surprise but the merit of this study is to establish this assumption as a fact. It also establishes useful tools (mutant lines and antibody) for the fly community interested in this topic. The study however does not provide new insights about the dynamic distribution of NASP and the cytological consequences of its maternal depletion on the amplification of cleavage nuclei.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Summary: Rapid cell cycles in early embryogenesis is driven from maternally supplied stockpiles of RNA and protein, including histones H3 and H4. This study uses sequence homology searches, biochemical approaches (immunoprecipitation and mass spectrometry) and genetics to identify NASP (CG8223) as the H3-H4 chaperone in Drosophila. Using CRISPR technology, the authors generate a NASP mutant fly line and show using genetic crosses that NASP is a maternal lethal gene. Furthermore the study shows that NASP stabilises H3-H4 during oogenesis and embryogenesis and is required for early embryogenesis.

Major comments: The key conclusions of this study are very convincing. For example, the authors use multiple approaches to show H3-H4 specific interactions with NASP and that H3-H4 protein levels are reduced in mutants (Western analyses, quantitative MS). Analysis is carried out on two individual NASP mutant lines (one deletion that produces no protein, one insertion that still produces some protein acting as a control). All experiments are well controlled, executed and presented. Genetic crossing schemes are well presented and statistical analysis of progeny is clear.

• We thank the reviewer for their positive feedback of our manuscript. Minor comments: In Figure 1B - Authors could indicate amino acids shown or are they full length proteins?

We have edited the methods to include specific amino residues that are included for each structure.

In Figure 2B - Authors could (semi) quantify reduction in NASP1 mutant to show this is a gene dose effect?

We have now included the quantification of the Western blot in Figure 2B.

CROSS-CONSULTATION COMMENTS I agree with the other reports. Although I did not indicate it in my original report, I agree that more in depth analysis of nuclear or chromosomal defects in NASP mutant embryos would enhance the study.

Thank you for this suggestion. We are repeating the DNA staining in embryos and will include this new data in the revised version of the manuscript.

Reviewer #3 (Significance (Required)):

Excess soluble histones can be toxic and must be bound to chaperones. Until this study the chaperone responsible for H3-H4 stabilisation in rapidly cycling cells in Drosophila embryos was not known. Moreover, the NASP homolog had not yet been identified in Drosophila nor had its function been characterised. The findings are of interest to Drosophila researchers, the field of chromatin assembly, as well as those interested in early embryogenesis in animals.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

Rapid cell cycles in early embryogenesis is driven from maternally supplied stockpiles of RNA and protein, including histones H3 and H4. This study uses sequence homology searches, biochemical approaches (immunoprecipitation and mass spectrometry) and genetics to identify NASP (CG8223) as the H3-H4 chaperone in Drosophila. Using CRISPR technology, the authors generate a NASP mutant fly line and show using genetic crosses that NASP is a maternal lethal gene. Furthermore the study shows that NASP stabilises H3-H4 during oogenesis and embryogenesis and is required for early embryogenesis.

Major comments:

The key conclusions of this study are very convincing. For example, the authors use multiple approaches to show H3-H4 specific interactions with NASP and that H3-H4 protein levels are reduced in mutants (Western analyses, quantitative MS). Analysis is carried out on two individual NASP mutant lines (one deletion that produces no protein, one insertion that still produces some protein acting as a control). All experiments are well controlled, executed and presented. Genetic crossing schemes are well presented and statistical analysis of progeny is clear.

Minor comments:

In Figure 1B - Authors could indicate amino acids shown or are they full length proteins?

In Figure 2B - Authors could (semi) quantify reduction in NASP1 mutant to show this is a gene dose effect?

Referees cross-commenting

I agree with the other reports. Although I did not indicate it in my original report, I agree that more in depth analysis of nuclear or chromosomal defects in NASP mutant embryos would enhance the study.

Significance

Excess soluble histones can be toxic and must be bound to chaperones. Until this study the chaperone responsible for H3-H4 stabilisation in rapidly cycling cells in Drosophila embryos was not known. Moreover, the NASP homolog had not yet been identified in Drosophila nor had its function been characterised. The findings are of interest to Drosophila researchers, the field of chromatin assembly, as well as those interested in early embryogenesis in animals.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Tirgar et al. report on a functional characterization of the Drosophila homolog of the histone H3/H4 chaperone NASP. They generated a loss of function allele of NASP by CRISPR/Cas9, which induces a partial maternal effect embryo lethal phenotype. Using quantitative mass spectrometry, they demonstrate that NASP stabilizes reservoirs of H3 and H4 in the early embryo. The manuscript is very clear and confirms the functional importance of maternal NASP for the early embryo. Genetic analyses are well conducted (but see my comments below) and the impact of NASP maternal mutant on H3 and H4 stockpiles is convincingly established by both quantitative mass spectrometry and Western-blotting.

Major comments:

• Although the authors used two independent deficiencies of the NASP genetic region to characterize their NASP CRISPR alleles, it is relatively standard in this type of functional analyses to perform rescue experiments using a transgene expressing the WT protein.
• In WB analyses, NASP appears systematically shorter in the NASP[1]/Df genotype compared to WT. Can the authors comment on this?
• The authors do not mention the centromeric histone H3 variant Cid in their analyses. Do they have evidence that it is not affected by loss of maternal NASP?
• The authors could have chosen to explore in more details the phenotypic defects of embryos derived from NASP mutant mothers. Instead, a single abnormal embryo is shown with no cytological details. This is a bit problematic since an earlier study (Zhang et al 2018, cited in the manuscript) actually provided more phenotypic details of embryos from NASP KD mothers.
• Similarly, the authors could have used their anti-NASP antibody to analyze the distribution of NASP during cleavage divisions. Does it behave like ASF1, for instance, which enters S phase nuclei at each cycle or does it remain in the cytoplasm? These are relatively simple experiments/analyses that could increase the significance of the study.

Minor comments:

• line 60: I suggest to introduce Drosophila in the next sentence, where it seems more appropriate (not all embryos develop "extremely rapidly").
• line 68: the 50% estimation of free histones does not really make sense without defining the embryonic stage.
• line 89: Are the authors specifically referring to Drosophila NASP?
• lines 99-106: I found this paragraph redundant with the introduction.
• line 142: H3-H4
• line190-191: It seems to me that data of Figure S2C are already included in Fig. 2E.
• line 232: it is surprising that the Zhang et al paper (reporting maternal KD of NASP) is only mentioned here. As a reader, I would certainly prefer to have it presented right from the introduction.
• Figure 4B needs a scale bar.
• line 302: Mentioning the identity and function of known H3/H4 histone chaperones acting in the early embryo (ASF1, HIRA, CAF-1, ...) could provide perspective to the present study.
• line 304: in contrast to this statement, I found quite surprising and interesting that NASP is not absolutely essential for embryo development considering its role. This should be discussed.

Referees cross-commenting

I fully agree with the other reports.

The NASP rescue experiment is just a suggestion but is not essential.

Significance

This work clarifies the identity and function of Drosophila NASP and clearly demonstrates that NASP is important for the stabilization of maternal stockpiles of H3 and H4 during early embryo development. The conservation of NASP function as a histone H3/H4 chaperone in Drosophila is not really a surprise but the merit of this study is to establish this assumption as a fact. It also establishes useful tools (mutant lines and antibody) for the fly community interested in this topic. The study however does not provide new insights about the dynamic distribution of NASP and the cytological consequences of its maternal depletion on the amplification of cleavage nuclei.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The rapid syncytial nuclear cycles that occur during the first ~2.5 hours of Drosophila embryogenesis and give rise to the blastoderm are supported by large amounts of maternally deposited histone proteins which are stored in the egg cytoplasm for deposition into replicating DNA during each round of S phase. Although the H2A/H2B storage chaperone Jabba was identified by Michael Welte's lab several years ago, maternal H3/H4 storage chaperones have not been identified. Tirgar et al provide evidence that the Drosophila NASP protein provides histone H3 and H4 storage function during these earliest stages of Drosophila embryogenesis. The data include genetic analyses that NASP function is required maternally, but not zygotically, and molecular analyses that NASP binds H3 and that H3 and H4 levels are reduced in the embryo and late-stage oocytes in the absence of NASP. These data are convincing and support the conclusion that NASP is a maternally acting H3/H4 storage chaperone needed in the early embryo.

Two additional lines of investigation would strengthen this conclusion and perhaps increase the impact and appeal of the manuscript.

The first is a microscopic analysis of the nuclear division cycles in eggs derived from NASP mutant mothers. The authors report DAPI staining and assessment of nuclear cycles, but do not show these data. In fact, the two embryos shown in Figure 4B do not look like DAPI stained embryos-there are no nuclei apparent in the images. Loss of maternal histone causes defects in chromosome morphology that result in characteristic defects such as lagging chromosomes and the failure of sister chromatid segregation leading to fused daughter nuclei (see PMID: 11157774 for an example). These defects should not be difficult to detect via DNA staining or even using fluorescently labeled H2 type histones. Characterizing such defects would lend support to the hypothesis and I think is important for this paper.

Second, determining the location of NASP in the early embryo might provide further insight into the mechanism of storage. i.e. is NASP located in the cytoplasm rather than the nucleus, perhaps in association with lipid droplets like Jabba? Do the antibodies the authors developed work in IF experiments to ask this question? At the moment what is shown is that NASP is present in 0-2 hour embryos via western blot analysis, supporting the conclusion that it functions in the early embryo as a storage chaperone. This analysis would be nice to have but is not essential in my view.

Small points:

Is NASP really a maternal effect "lethal"? Some of the eggs do hatch, and so some develop to stages where maternal histones are no longer necessary and zygotic production takes over (i.e. cycle 15). Perhaps consider the language used here. On this point, do NASP mutant females lay the same number of eggs as wild type? i.e. is there a requirement for oogenesis/egg production (other than depositing H3/H4 into the egg), or just for the early zygotic cycles?

The first paragraph of the results is redundant with much of the introduction, which I think could do a better job at describing in more detail the syncytial cycles and the special needs they have for histone storage and chaperone function versus the post-blastoderm embryonic cycles and the rest of development. i.e. make a better distinction between the first two hours of embryogenesis versus the rest of embryogenesis, and the when the switch from maternal to zygotic control of development and histone production occurs (cycle 15 at 3-4 hours AED).

Referees cross-commenting

Seems like all reviewers are in general agreement, particularly about providing additional data regarding chromosome/nuclear behavior in the NASP mutants and NASP localization in the early embryo to increase impact of the study. While rescue of the NASP mutant phenotype with a transgene would be nice, as suggested by referee #2, I don't think it's essential given the genetic approaches employed.

Significance

see above

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

An exciting development in our knowledge about how the Arp2/3 complex controls the assembly of actin networks has come from the discovery that in addition to forming branched networks, Arp2/3 can nucleate linear filaments when it is activated by WISH/DIP/SPIN90. However, despite some excellent work largely done by the Nolen lab in yeast, many questions remain about how Arp2/3-mediated assembly of branched vs. linear actin filament. This is especially true in the complex environment of cells, were synergy and competition of different actin networks is used to control biological processes. Knowing the biochemical and physical properties of these different Arp2/3 assemblies will be key to figuring out how they work in cells. Here Cao et al. use an elegant microfluidics based single filament assay system to perform a comparative analysis of the stability of linear and branched Arp2/3 networks. They find interesting differences in how they respond to stabilizing and destabilizing factors. The most striking differences happens when force or aging is applied- both cause debranching of branched networks but have little effect on Spin90-Arp2/3 nucleated filaments.

We thank the reviewer for their positive comments.

Major comments:

As a comparative study on the stability of branched vs. linear Arp2/3 nucleated filaments, this manuscript is fairly complete. The key conclusions are well supported by rigorous experiments which can be reproduced by others based on the information provided. However, I am not seeing explicit information on performing biological replicates. This should be included in the manuscript. The use of statistics is largely fine; however I question the use of one statistical test on one figure (see minor comments below).

The revised manuscript is now explicit about biological replicates. We now specify the biological repeats of all our experiments in the figure legends, and we now show the results from new repeats in Fig 4 and Supp Fig S2 (please see also our response to the minor comments below, for more details).

I would not ask for additional experiments at this time. However, there is an analysis that would be important for interpreting the authors' claims- branch/filament length at the time of dissociation or destabilization of Arp2/3. This would help address if there was a physical tipping point for each type of structure that could explain potential differences they see. The authors should already have this data and the time to complete it would be negligible in delaying publication.

If we understand correctly, the “physical tipping point” mentioned by the reviewer would be a threshold force, where the Arp2/3-filament interface would become unstable. This is an interesting idea. Indeed, the applied force scales with the length of the filament (or branch), as well as with the flow velocity. In most of our experiments, however, the force applied to SPIN90-Arp2/3 and to branch junctions was kept constant and below 0.2 pN. This was done by exposing the filaments (or branches) to G-actin at the critical concentration, in order to minimize variations of their lengths. Therefore, by design, dissociation events in these experiments take place at the same length, ruling out the existence of a “tipping point”.

Our data provide another test of the reviewer’s hypothesis, thanks to the experiments where we specifically address the question of the impact of force (Fig 5 and Supp Fig S6), by varying length and flow rate. We found that the stability of SPIN90-Arp2/3 linear filaments was unaffected by force, and that debranching was steadily accelerated by force. In both cases, it thus appears that there is no detectable threshold.

One additional major comment is that the manuscript's title and abstract hint that this paper explores the differences in nucleation of branched vs. linear filaments by Arp2/3. However, the only figure that deals explicitly with nucleation in the paper is Figure 1, which is really just a confirmation that the mammalian proteins used in this study perform similarly to their yeast homologues (Balzer et al, Current Biology 2019). The authors might think about rewording the title/abstract to better reflect that paper really explores the differences in the stability of the two networks

This is a fair point. We have now modified the title into “Regulation of branched versus linear Arp2/3-generated actin filaments”.

Minor comments:

1 in 12 men and 1 in 200 women are red/green colorblind. Please change the coloring of the schematics and images so that they can be easily seen by all people. This is especially true of the schematics, which are important for understanding exactly what each assay is measuring.

We thank the reviewer for pointing this out. We have now made the schematics and images in Figs 1A, 2A, 2D and 4D colorblind-friendly.

The Introduction is a bit choppy and unfocused. It was difficult to deduce exactly where the paper was going from it. Please consider re-writing it for better clarity. The Discussion on the other hand was fantastic. Great job on interpreting your results in a larger context.

We have re-written large parts of the Introduction to make it clearer. We are glad the reviewer liked the Discussion, where we have nonetheless made some small changes in response to comments from the other reviewers.

Many figures- while the use of different lightness values of the same color is appreciated in conveying different concentrations of reagents used, there were several instances where it was very hard to read the one on the very bottom (ex. 2B, E; 3A; 5C, G).

We have now changed the colors in these figures, to make them clearer.

Figure 1- since this is a confirmation of previous results performed using the same proteins from other species, the title should reflect that (ex. VCA domains accelerate the nucleation of filaments by mammalian SPIN90-Arp2/3). Also, to me this figure is supplementary to the main message of the paper. The authors might think of moving it to Supplementary Information.

We have modified the title of Figure 1, now specifying “mammalian”, following the reviewer’s suggestion. However, we prefer to keep this figure as a main figure, rather than move it to Supplementary as proposed. Indeed, this figure does more than simply confirm previous results with mammalian proteins, since it compares different VCAs, which is new. These results are important because they are put in perspective with our results on the acceleration of linear filament detachment by different VCAs, later in the manuscript.

Figure 1- If the goal was to verify that G-actin recruitment by VCA was important for Spin90-Arp 2/3 nucleation by performing a competition experiment with profilin, why was the concentration of G-actin AND profilin increased between the experiments in 1B vs. 1C. It makes it hard to directly compare the results.

We now provide new data in Fig 1C, which can be directly compared to Fig 1B (only the profilin concentration was increased). It clearly shows that the effect of VCA disappears when the profilin concentration is increased.

Figure 4B-F- Here, it would be nice to see the distribution of all the individual results, which are hidden by the bar graph. Additionally, the Chi-square test is not the appropriate test for evaluating statistical significance between multiple groups. ANOVA followed by an appropriate post hoc test should be used here.

We now show the individual results in the bar graphs of figure 4. In this situation, we agree that the statistical significance should not be evaluated by a Chi-square test. We now indicate the p-values obtained from a paired t-test, which seems appropriate since we are comparing averages in pairs.

Figure 4G- Please quantify and show reproducibility.

We now show quantified repeats (shown in Fig 4, new panels H and I).

Figure 5- the piconewton forces used for these experiments is in line with measured forces that are applied to actin in cells (ex. Mehida et al, Nature Cell Biology 2021; Jiang et al, Nature 2003). The text would benefit if this was explicitly stated.

We now state this explicitly, when presenting these results.

Reviewer #1 (Significance (Required)):

The real significance of this work is in characterizing the differential stabilities of linear vs. branched Arp2/3 filaments in response to actin-binding proteins, mechanical stress, and aging. While both types of filaments respond similarly to actin-binding proteins, with nuanced differences, the most striking results came from applied force and aging experiments, with Spin90-Arp2/3 filaments being much more resistant to both. This has some very interesting implications for how these two types of assemblies might synergize in cells. Additionally, the results also have some exciting implications for the pointed-end regulation of actin filaments, which is still poorly understood in complex systems. Since the manuscript is A) more of a survey study on the factors that influence filament stability that does not go particularly deep into any particular mechanism of regulation and B) has no direct applicability to how the physical properties of branched and linear Arp2/3 nucleated actin filaments influence actin network activity in cells, the audience will likely by limited to actin enthusiasts. However, the work is still important in both what it reveals and implies.

We thank the reviewer for pointing out the novelty and the importance of our work. We agree that the significance of our paper lies in the characterization of the differential stabilities of linear vs. branched Arp2/3 filaments, in response to different physiological factors. One of the strengths of our approach is that we do not focus on one regulatory mechanism in particular. Rather, we reveal fundamental differences between the Arp2/3-generated filaments and how they can be regulated. Understanding these basic mechanisms is a prerequisite to understand the regulation of entire cytoskeletal networks.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

The quantitative analysis can be improved. It appears that most of the data results from single experiments, with rate values and errors resulting from fitting of single experiments without repetitions. In Fig. 1C legend (p.5) the authors state "These experiments were repeated three times, with similar results", but the data is not used in the analysis and other experiments do not mention this point. This is particularly important for comparisons among different VCAs that are rather similar in nature. In Fig. 1B. N-WASP is more efficient in nucleating SPIN90-Arp2/3 complex-linear filaments followed by WASP and then WASH. In Fig. 2 B,C, N-WASP is the most effective in dissociating SPIN90-Arp2/3 complex linear filaments followed by WASH and then WASP. But in Fig. 2 E,F, WASH is by far the most effective in dissociating branches followed by N-WASP and then WASP. Therefore, the conclusion in the Discussion (p.12) "While these regulatory proteins similarly affect branched and linear Arp2/3-generated filaments, they do so with clear quantitative differences" is not supported by quantification. To remedy this problem the authors should include at least 3 repeats of each experiment in data analysis. Also, they could include an analysis of sequence differences among VCAs and discuss how these may correlate with the observed differences. For instance, one WH2 in WASP vs. two in N-WASP.

Indeed, we argue that the two forms of activated Arp2/3 differ in their sensitivity to different VCA motifs, based on how these VCA motifs rank in their ability to destabilize branched and linear filaments (the VCA motifs also rank differently in their activation and co-activation of Arp2/3 to nucleate branches and linear filaments, but this result does not contribute to our discussion of how proteins interact with the activated Arp2/3). Following the reviewer’s suggestion, we now show repeats of these experiments (new Supp Fig S2), clearly showing that N-WASP is the most effective in dissociating linear filaments while the differences are milder for dissociating branches, with WASH being at least as effective as NWASP. We now also discuss how this observation could relate to differences in sequence between VCAs (Discussion section and new Supp Fig S9).

Also, please note that, following a suggestion from Reviewer 3, we have now performed experiments with the CA-domains of NWASP (new Supp Fig S4C and S4D), which show that the V-domain plays an important role in debranching but plays no role in destabilizing SPIN90-Arp2/3 at filament pointed ends. These new results reinforce our statement that VCA affects branched and linear Arp2/3-generated filaments differently.

Reviewer #2 (Significance (Required)):

Arp2/3 complex is a 7-protein complex implicated in actin filament nucleation and branching. Arp2/3 complex-nucleated branched networks are found at several locations in cells and are responsible for processes such as cell motility.

Cao et al. compare the effect of several proteins on the filament nucleation activity of Arp2/3 complex, and the stabilization or destabilization of actin filament branches as well as linear actin filaments nucleated by SPIN90-Arp2/3 complex. The proteins tested include the VCA regions of three NPFs (N-WASP, WASP, and WASH) that activate Arp2/3 complex, GMF (a debranching protein) and cortactin (a branch stabilizing protein). For the most part, the study uses a single method, microfluidics-TIRF microscopy.

The main findings are:

1. VCA domains enhance nucleation of linear filaments by SPIN90-Arp2/3 complex in the presence of actin monomers.
2. However, VCA domains can also destabilize existing SPIN90-Arp2/3 complex linear filaments and branches, and this effect depends on the presence of of V-domain (WH2 domain that binds actin monomers).
3. The debranching factor GMF also destabilizes SPIN90-Arp2/3 complex linear filaments. Both GMF and VCA generate free pointed ends by dissociating Arp2/3 complex from pointed ends and SPIN90.
4. SPIN90-Arp2/3 complex linear filaments are less susceptible to force and aging than filament branches.
5. Cortactin stabilizes SPIN90-Arp2/3 complex linear filaments to higher degree than it does branches. These are novel and very interesting new observations of significant interest to the actin cytoskeleton field. Therefore, I recommend publication of this paper in EMBO J.

We thank the reviewer for their positive evaluation of our work.

I have one recommendation and one suggestion for improvement:

Major:

1. The quantitative analysis can be improved. It appears that most of the data results from single experiments, with rate values and errors resulting from fitting of single experiments without repetitions. In Fig. 1C legend (p.5) the authors state "These experiments were repeated three times, with similar results", but the data is not used in the analysis and other experiments do not mention this point. This is particularly important for comparisons among different VCAs that are rather similar in nature. In Fig. 1B. N-WASP is more efficient in nucleating SPIN90-Arp2/3 complex-linear filaments followed by WASP and then WASH. In Fig. 2 B,C, N-WASP is the most effective in dissociating SPIN90-Arp2/3 complex linear filaments followed by WASH and then WASP. But in Fig. 2 E,F, WASH is by far the most effective in dissociating branches followed by N-WASP and then WASP. Therefore, the conclusion in the Discussion (p.12) "While these regulatory proteins similarly affect branched and linear Arp2/3-generated filaments, they do so with clear quantitative differences" is not supported by quantification. To remedy this problem the authors should include at least 3 repeats of each experiment in data analysis. Also, they could include an analysis of sequence differences among VCAs and discuss how these may correlate with the observed differences. For instance, one WH2 in WASP vs. two in N-WASP.

This comment is identical to the reviewer’s first paragraph. We copy our answer here again, for convenience:

Indeed, we argue that the two forms of activated Arp2/3 differ in their sensitivity to different VCA motifs, based on how these VCA motifs rank in their ability to destabilize branched and linear filaments (the VCA motifs also rank differently in their activation and co-activation of Arp2/3 to nucleate branches and linear filaments, but this result does not contribute to our discussion of how proteins interact with the activated Arp2/3). Following the reviewer’s suggestion, we now show repeats of these experiments (new Supp Fig S2), clearly showing that N-WASP is the most effective in dissociating linear filaments while the differences are milder for dissociating branches, with WASH being at least as effective as NWASP. We now also discuss how this observation could relate to differences in sequence between VCAs (Discussion section and new Supp Fig S9).

Also, please note that, following a suggestion from Reviewer 3, we have now performed experiments with the CA-domains of NWASP (new Supp Fig S4C and S4D), which show that the V-domain plays an important role in debranching but plays no role in destabilizing SPIN90-Arp2/3 at filament pointed ends. These new results reinforce our statement that VCA affects branched and linear Arp2/3-generated filaments differently.

Minor:

In GST-pull-down experiments (Fig. 4G), the amount of Arp2/3 complex bound is analyzed by Western, which is rather unprecise. Is the amount of Arp2/3 complex so little that it cannot be quantified using regular SDS-PAGE? If that is the case, this would suggest rather low affinity of SPIN90 for Arp2/3 complex. How does this affect the proposed mechanism and experiments in the microfluidics chamber?

Indeed, the amount of pulled-down Arp2/3 is low and difficult to quantify by SDS-PAGE. This is consistent with previous reports which indicate a low affinity of SPIN90 for the Arp2/3 complex (Wagner et al. Current Biology 2013, Balzer et al. eLife 2020). This does not affect our conclusions, which we now confirm by showing quantified repeats of our pull-down experiments (new panels H and I, in Figure 4). In spite of this low affinity, which makes it difficult to saturate SPIN90 with Arp2/3, the SPIN90-Arp2/3 interaction is very stable and allows us to carry out our experiments in the microfluidics chamber over several tens of minutes (as was already the case in our previous study, Cao et al. NCB 2020).

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Summary:

In this study, Cao and collaborators investigate the biochemical and mechanical differences between branched actin filaments nucleated by WASP-activated Arp2/3 complex and linear actin filaments nucleated by SPIN90-activated Arp2/3 complex. They use TIRF microscopy in a microfluidic chamber to show that the mammalian proteins, SPIN90 and WASP (or N-WASP or WAVE), like their yeast homologues, co-activate Arp2/3 complex to nucleate linear actin filaments. Using the same assays, they find the surprising result that the VCA segment of WASP proteins destabilizes the interaction between SPIN90 and Arp2/3 complex in linear actin filaments nucleated by Arp2/3 complex. They then show that VCA also destabilizes actin filament branches. The remainder of the study explores the influence of branch stabilizing/destabilizing proteins or mechanical stress on the stability of the interaction between SPIN90 and Arp2/3 complex on the pointed end of the actin filament. They find that like branch junctions, SPIN90-bound Arp2/3 is destabilized at the end of linear filaments by GMF and stabilized by cortactin. However, unlike branch junctions, SPIN90-Arp2/3 complex is not destabilized on filament ends by piconewton forces or by aging. They conclude that SPIN90- versus VCA-activated Arp2/3 complex adopt similar but non-identical conformations.

Overall, the paper is well written and the experiments, which are very challenging, are rigorously executed. The biochemical results are convincing, novel and unexpected. However, the work could be strengthened by more strongly connecting the biochemical observations to biological implications. In addition, there are some interpretations/conclusions that seem somewhat weakly supported, and the authors should consider revising. Nonetheless, given the quality of the work and the importance of the system, this manuscript will appeal to a broad audience.

We thank the reviewer for their positive comments. We have rewritten parts of the Discussion in order to better connect our observations to implications in cells. We address the concerns regarding our interpretations in the point-by-point, below.

Comments on evidence, reproducibility, clarity and significance:

The differences in the stability of SPIN90-Arp2/3 on linear filaments verses branch junctions led the authors to conclude that SPIN90- versus VCA-activated complexes adopt similar yet non-identical conformations. There are two problems with this conclusion:

1) This conclusion rests on the idea that the biochemical differences can only be due to differences in the "ground state" active conformations of the complex. Another possible scenario would be that the active conformations are the same, but the transition state or intermediate state structures within the debranching reactions are different, thus changing the kinetics of the debranching reactions.

We thank the reviewer for this remark, and we agree that conformational differences may also arise in the intermediate states, during dissociation (of the branch from the mother, or of the linear filaments from SPIN90). We now mention this possibility in our Discussion.

2.) There are already structural data showing conformational differences between the Dip1-bound Arp2/3 complex on the end of a linear filament and Arp2/3 complex at a branch junction. While there are some caveats to comparisons of the structures (e.g., the Dip1 structure includes the fission yeast SPIN90 protein (Dip1) and the fission yeast Arp2/3 complex while the branch junction contains mammalian proteins), these data offer much stronger evidence that the active states adopt (somewhat) different conformations than the data presented here.

We agree that the available structural data (in particular, Ding et al. PNAS 2022, which was not yet published when we submitted our manuscript, and which we now cite) provide a clear indication that active Arp2/3 adopts different conformations in branches and linear filaments. We have modified our text to make this point clearer.

The authors make comparisons between the Fäβler branch junction structure and the Shaaban Dip1-Arp2/3-filament structure. The Fäβler branch junction structure is a low resolution structure (9 angstroms) and should be interpreted with caution (see below). A much higher resolution of a branch junction structure was recently solved (Ding et al, PNAS 2022) and should be used for comparisons between the structures.

Ding et al. PNAS 2022 was not yet published when we submitted our manuscript. We now use it to compare the structures of active Arp2/3, and we have modified the text accordingly.

Pg 14 - The authors say differences between ARPC3-Arp2 and ARPC5-Arp2 contacts in the two structures are likely to cause the differences in interactions with GMF and VCA. Two concerns with this statement are: 1.) The basis for the conclusion that the ARPC5-Arp2 contacts are different (in Fäβler, et al.) is not solid (see Ding, et al) and 2.) The analysis is vague. To reasonably conclude that differences in the contacts would influence GMF and VCA interactions would require mapping out the structural connection between the ARPC3-Arp2 interaction site and the GMF or VCA binding sites. If there is no obvious connection between these sites, the conclusion that the differences in the ARPC3-Arp2 interface cause differences in VCA and GMF binding should be far more circumspect.

We have re-written this part of the Discussion section. In light of the new data by Ding et al., we agree with the reviewer that the conclusion that the ARPC5-Arp2 contacts are different is not solid. Our revised text makes it clear that we are not making any claims involving interactions within the Arp2/3 complex. Our point is simply that recent cryo-EM reports indicate conformational differences in Arp2 and Arp3 between the two activated forms of the Arp2/3 complex and that, since the CA-domain of NPFs bind to Arp2 and Arp3, it appears reasonable to make a connection with our results.

Pg 6. "These observations suggest that the ability of VCA to destabilize Arp2/3-nucleated filaments relies on the availability of its V-domain." It's possible that G-actin binding to V blocks the CA from accessing the branch junction. Therefore, it seems important to test whether N-WASP-CA can destabilize Arp2/3-nucleated actin filaments.

We thank the reviewer for this suggestion. We now present results from new experiments performed with the CA-domain of NWASP (new Supp Fig S4C,D). We find that the V-domain participates in the enhancement of debranching, but that it appears to play no role in the destabilization of SPIN90-Arp2/3 from the pointed end. It thus seems that the reviewer’s proposal is correct, and that G-actin binding to the V-domain blocks the CA-domain from accessing the branch junction. We now propose this interpretation in the text.

Pg 1 - The authors state that "It thus appears that linear and branched Arp2/3-generated filaments respond similarly to regulatory proteins, albeit with quantitative differences". It is worth considering if one should make a blanket statement that linear and branched filaments respond similarly to regulatory proteins when they have tested 3 in total.

We have rephrased this sentence. It now reads “… respond similarly to the regulatory proteins we have tested…”

Pg 3 - "More generally, the stability of SPIN90-Arp2/3 at the pointed end, which is important to understand the reorganization and disassembly of actin filament networks, remains to be established." In some ways this statement not quite accurate because Balzer et al previously showed that Dip1-Arp2/3 complex is very stable at the pointed end. Is the question here whether that stability is also conserved in mammalian systems? If so, that should be more directly stated.

We meant that, beyond observing that SPIN90 remains visible at the pointed end for some time (as in Balzer et al.), a lot remained unknown: its lifetime had not been quantified, and its sensitivity to the factors that affect branch junctions (proteins, aging, mechanical tension) had not been studied. We have rephrased the sentence in the manuscript to clarify this point.

The observation that VCA accelerates debranching and SPIN90-Arp2/3 dissociation is very interesting. However, it is uncertain if this biochemical activity has biological relevance, given that once nucleation occurs, Arp2/3 complex will move away from the membrane. While the authors mention in the discussion that debranching by VCA could be relevant when the network is compressed near the membrane, this argument is not particularly strong. Are there ways to strengthen this argument, or find another impact this finding might have on our understanding of Arp2/3 complex regulation?

We now mention another situation where branch junctions could encounter membrane-bound VCA domains: on the dorsal and ventral membrane surfaces of lamellipodia. We now cite the recent Kage et al. J Cell Science 2022 and Mehidi et al. NCB 2021, where WAVE has been observed in lamellipodia away from the leading edge.

The observation that SPIN90+Arp2/3-nucleated filaments are not sensitive to piconewton forces is also very interesting. The authors focus on the differences in the amount of surface area buried when discussing this result. However, if seems a key factor in the stability of the linear filaments would be the direction of the force relative to the complex and attached filament(s), which would be very different for a branch versus a linear filament. The authors should consider addressing this in their discussion.

The orientation of the applied force is an interesting point. In their study on debranching, Pandit et al. (PNAS 2020) report that their results are not affected by the angle of the applied force relative to the mother filament (their Fig S1D). We now specify this in our manuscript, when introducing our results on mechanical tension. Similarly, we found that anchoring SPIN90 to the coverslip surface by its N-terminus rather than its C-terminus, which likely affects the orientation of the applied force, had no impact on our results (Supp Fig S6A). We have now also added a sentence regarding this aspect in our manuscript, after presenting this result.

Fig 4, D-F: It is unclear how the authors determined which filaments were spontaneously nucleated versus those that were nucleated by SPIN90-Arp2/3 complex in these experiments. In reactions containing SPIN90 and Arp2/3 complex what fraction of the filaments will be spontaneously nucleated?

In our conditions, there is no detectable spontaneous nucleation. In control experiments where we flow in the same concentration of G-actin, in the absence of Arp2/3 or in the absence of SPIN90, we observe no filaments at all on the surface, over several fields of view, after 5 minutes. We now specify this in the Methods section.

Pg 9 - The observation that VCA negatively influences binding of SPIN90 to the complex is unexpected. What implications does this have for understanding how SPIN90 and VCA synergize to activate the complex?

It appears that the outcome depends on the context. The main role of VCA during co-activation of the Arp2/3 complex with SPIN90 seems to be to supply G-actin, as already proposed (Balzer, 2020) and confirmed by our results (Fig 1C). In the absence of G-actin, VCA is more likely to remove Arp2/3 from SPIN90 (Fig 4G,I). Similarly, when a filament is already formed, the presence of G-actin mitigates the removal of SPIN90-Arp2/3 from the pointed end by VCA (Supp Fig S4).

Fig 4B - Why is there greater nucleation when Arp2/3 complex and GMF are added together compared to renucleation in reactions that don't have any GMF? This is surprising, especially considering that GMF decreases binding of Arp2/3 complex to SPIN90.

Indeed, there is a small yet statistically significant difference in the re-nucleation fraction we measured in the presence of Arp2/3, with or without GMF (Fig 4B). This may be due to the different timescales of the two situations. In the absence of GMF, the detachment of filaments is slow and new filaments are nucleated from the initial Arp2/3 complexes, which remained bound to SPIN90 upon detachment of the first filaments. In contrast, in the presence of GMF, detachment is faster and accompanied by the departure of the initial Arp2/3, and a fresh Arp2/3 then binds to SPIN90 to nucleate a new filament. It is thus possible that, in the absence of GMF, a small fraction of the SPIN90 and/or their initially bound Arp2/3 complexes would denature over the time they spend at the bottom of the microchamber at 25°C, thereby leading to a slightly smaller re-nucleation fraction. A similar mechanism could be at play in the experiments with or without VCA, in addition to the enhancement of nucleation by VCA (Fig 4C).

Minor Corrections/Comments

Pg 3 "We show that Arp2/3 nucleation is similarly stabilized by cortactin and destabilized by GMF" Do the authors mean branches and linear filaments nucleated by Arp2/3 complex?

Yes, that is what we meant. This sentence has now been modified.

Pg 6- The cyan 3uM data and legend in figure 2B and E is probably too dim to see clearly.

The colors have been changed to improve readability.

Fig 4 B,C,E,F: It would be best to show the individual data points here if possible.

We now show individual data points in all these figure panels.

Pg 16 Please specify which antibody was used to anchor SPIN90.

The antibodies are Anti-GST for Nter anchoring of GST-SPIN90, and anti-His for Cter anchoring of SPIN90-His. We now specify this in the Methods section.

CROSS-CONSULTATION COMMENTS I agree with the points that the other reviewers raised.

Reviewer #3 (Significance (Required)):

Comments on significance are in the above section.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

In this study, Cao and collaborators investigate the biochemical and mechanical differences between branched actin filaments nucleated by WASP-activated Arp2/3 complex and linear actin filaments nucleated by SPIN90-activated Arp2/3 complex. They use TIRF microscopy in a microfluidic chamber to show that the mammalian proteins, SPIN90 and WASP (or N-WASP or WAVE), like their yeast homologues, co-activate Arp2/3 complex to nucleate linear actin filaments. Using the same assays, they find the surprising result that the VCA segment of WASP proteins destabilizes the interaction between SPIN90 and Arp2/3 complex in linear actin filaments nucleated by Arp2/3 complex. They then show that VCA also destabilizes actin filament branches. The remainder of the study explores the influence of branch stabilizing/destabilizing proteins or mechanical stress on the stability of the interaction between SPIN90 and Arp2/3 complex on the pointed end of the actin filament. They find that like branch junctions, SPIN90-bound Arp2/3 is destabilized at the end of linear filaments by GMF and stabilized by cortactin. However, unlike branch junctions, SPIN90-Arp2/3 complex is not destabilized on filament ends by piconewton forces or by aging. They conclude that SPIN90- versus VCA-activated Arp2/3 complex adopt similar but non-identical conformations.

Overall, the paper is well written and the experiments, which are very challenging, are rigorously executed. The biochemical results are convincing, novel and unexpected. However, the work could be strengthened by more strongly connecting the biochemical observations to biological implications. In addition, there are some interpretations/conclusions that seem somewhat weakly supported, and the authors should consider revising. Nonetheless, given the quality of the work and the importance of the system, this manuscript will appeal to a broad audience.

Comments on evidence, reproducibility, clarity and significance:

The differences in the stability of SPIN90-Arp2/3 on linear filaments verses branch junctions led the authors to conclude that SPIN90- versus VCA-activated complexes adopt similar yet non-identical conformations. There are two problems with this conclusion:

1. This conclusion rests on the idea that the biochemical differences can only be due to differences in the "ground state" active conformations of the complex. Another possible scenario would be that the active conformations are the same, but the transition state or intermediate state structures within the debranching reactions are different, thus changing the kinetics of the debranching reactions.
2. There are already structural data showing conformational differences between the Dip1-bound Arp2/3 complex on the end of a linear filament and Arp2/3 complex at a branch junction. While there are some caveats to comparisons of the structures (e.g., the Dip1 structure includes the fission yeast SPIN90 protein (Dip1) and the fission yeast Arp2/3 complex while the branch junction contains mammalian proteins), these data offer much stronger evidence that the active states adopt (somewhat) different conformations than the data presented here.

The authors make comparisons between the Fäβler branch junction structure and the Shaaban Dip1-Arp2/3-filament structure. The Fäβler branch junction structure is a low resolution structure (9 angstroms) and should be interpreted with caution (see below). A much higher resolution of a branch junction structure was recently solved (Ding et al, PNAS 2022) and should be used for comparisons between the structures.

Pg 14 - The authors say differences between ARPC3-Arp2 and ARPC5-Arp2 contacts in the two structures are likely to cause the differences in interactions with GMF and VCA. Two concerns with this statement are: 1.) The basis for the conclusion that the ARPC5-Arp2 contacts are different (in Fäβler, et al.) is not solid (see Ding, et al) and 2.) The analysis is vague. To reasonably conclude that differences in the contacts would influence GMF and VCA interactions would require mapping out the structural connection between the ARPC3-Arp2 interaction site and the GMF or VCA binding sites. If there is no obvious connection between these sites, the conclusion that the differences in the ARPC3-Arp2 interface cause differences in VCA and GMF binding should be far more circumspect.

Pg 6. "These observations suggest that the ability of VCA to destabilize Arp2/3-nucleated filaments relies on the availability of its V-domain." It's possible that G-actin binding to V blocks the CA from accessing the branch junction. Therefore, it seems important to test whether N-WASP-CA can destabilize Arp2/3-nucleated actin filaments.

Pg 1 - The authors state that "It thus appears that linear and branched Arp2/3-generated filaments respond similarly to regulatory proteins, albeit with quantitative differences". It is worth considering if one should make a blanket statement that linear and branched filaments respond similarly to regulatory proteins when they have tested 3 in total.

Pg 3 - "More generally, the stability of SPIN90-Arp2/3 at the pointed end, which is important to understand the reorganization and disassembly of actin filament networks, remains to be established." In some ways this statement not quite accurate because Balzer et al previously showed that Dip1-Arp2/3 complex is very stable at the pointed end. Is the question here whether that stability is also conserved in mammalian systems? If so, that should be more directly stated.

The observation that VCA accelerates debranching and SPIN90-Arp2/3 dissociation is very interesting. However, it is uncertain if this biochemical activity has biological relevance, given that once nucleation occurs, Arp2/3 complex will move away from the membrane. While the authors mention in the discussion that debranching by VCA could be relevant when the network is compressed near the membrane, this argument is not particularly strong. Are there ways to strengthen this argument, or find another impact this finding might have on our understanding of Arp2/3 complex regulation?

The observation that SPIN90+Arp2/3-nucleated filaments are not sensitive to piconewton forces is also very interesting. The authors focus on the differences in the amount of surface area buried when discussing this result. However, if seems a key factor in the stability of the linear filaments would be the direction of the force relative to the complex and attached filament(s), which would be very different for a branch versus a linear filament. The authors should consider addressing this in their discussion.

Fig 4, D-F: It is unclear how the authors determined which filaments were spontaneously nucleated versus those that were nucleated by SPIN90-Arp2/3 complex in these experiments. In reactions containing SPIN90 and Arp2/3 complex what fraction of the filaments will be spontaneously nucleated?

Pg 9 - The observation that VCA negatively influences binding of SPIN90 to the complex is unexpected. What implications does this have for understanding how SPIN90 and VCA synergize to activate the complex?

Fig 4B - Why is there greater nucleation when Arp2/3 complex and GMF are added together compared to renucleation in reactions that don't have any GMF? This is surprising, especially considering that GMF decreases binding of Arp2/3 complex to SPIN90.

Minor Corrections/Comments

Pg 3 "We show that Arp2/3 nucleation is similarly stabilized by cortactin and destabilized by GMF" Do the authors mean branches and linear filaments nucleated by Arp2/3 complex?

Pg 6- The cyan 3uM data and legend in figure 2B and E is probably too dim to see clearly.

Fig 4 B,C,E,F: It would be best to show the individual data points here if possible.

Pg 16 Please specify which antibody was used to anchor SPIN90.

Referees cross-commenting

I agree with the points that the other reviewers raised.

Significance

Comments on significance are in the above section.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The quantitative analysis can be improved. It appears that most of the data results from single experiments, with rate values and errors resulting from fitting of single experiments without repetitions. In Fig. 1C legend (p.5) the authors state "These experiments were repeated three times, with similar results", but the data is not used in the analysis and other experiments do not mention this point. This is particularly important for comparisons among different VCAs that are rather similar in nature. In Fig. 1B. N-WASP is more efficient in nucleating SPIN90-Arp2/3 complex-linear filaments followed by WASP and then WASH. In Fig. 2 B,C, N-WASP is the most effective in dissociating SPIN90-Arp2/3 complex linear filaments followed by WASH and then WASP. But in Fig. 2 E,F, WASH is by far the most effective in dissociating branches followed by N-WASP and then WASP. Therefore, the conclusion in the Discussion (p.12) "While these regulatory proteins similarly affect branched and linear Arp2/3-generated filaments, they do so with clear quantitative differences" is not supported by quantification. To remedy this problem the authors should include at least 3 repeats of each experiment in data analysis. Also, they could include an analysis of sequence differences among VCAs and discuss how these may correlate with the observed differences. For instance, one WH2 in WASP vs. two in N-WASP.

Significance

Arp2/3 complex is a 7-protein complex implicated in actin filament nucleation and branching. Arp2/3 complex-nucleated branched networks are found at several locations in cells and are responsible for processes such as cell motility.

Cao et al. compare the effect of several proteins on the filament nucleation activity of Arp2/3 complex, and the stabilization or destabilization of actin filament branches as well as linear actin filaments nucleated by SPIN90-Arp2/3 complex. The proteins tested include the VCA regions of three NPFs (N-WASP, WASP, and WASH) that activate Arp2/3 complex, GMF (a debranching protein) and cortactin (a branch stabilizing protein). For the most part, the study uses a single method, microfluidics-TIRF microscopy.

The main findings are:

1. VCA domains enhance nucleation of linear filaments by SPIN90-Arp2/3 complex in the presence of actin monomers.
2. However, VCA domains can also destabilize existing SPIN90-Arp2/3 complex linear filaments and branches, and this effect depends on the presence of of V-domain (WH2 domain that binds actin monomers).
3. The debranching factor GMF also destabilizes SPIN90-Arp2/3 complex linear filaments. Both GMF and VCA generate free pointed ends by dissociating Arp2/3 complex from pointed ends and SPIN90.
4. SPIN90-Arp2/3 complex linear filaments are less susceptible to force and aging than filament branches.
5. Cortactin stabilizes SPIN90-Arp2/3 complex linear filaments to higher degree than it does branches.

These are novel and very interesting new observations of significant interest to the actin cytoskeleton field. Therefore, I recommend publication of this paper in EMBO J. I have one recommendation and one suggestion for improvement:

Major:

1. The quantitative analysis can be improved. It appears that most of the data results from single experiments, with rate values and errors resulting from fitting of single experiments without repetitions. In Fig. 1C legend (p.5) the authors state "These experiments were repeated three times, with similar results", but the data is not used in the analysis and other experiments do not mention this point. This is particularly important for comparisons among different VCAs that are rather similar in nature. In Fig. 1B. N-WASP is more efficient in nucleating SPIN90-Arp2/3 complex-linear filaments followed by WASP and then WASH. In Fig. 2 B,C, N-WASP is the most effective in dissociating SPIN90-Arp2/3 complex linear filaments followed by WASH and then WASP. But in Fig. 2 E,F, WASH is by far the most effective in dissociating branches followed by N-WASP and then WASP. Therefore, the conclusion in the Discussion (p.12) "While these regulatory proteins similarly affect branched and linear Arp2/3-generated filaments, they do so with clear quantitative differences" is not supported by quantification. To remedy this problem the authors should include at least 3 repeats of each experiment in data analysis. Also, they could include an analysis of sequence differences among VCAs and discuss how these may correlate with the observed differences. For instance, one WH2 in WASP vs. two in N-WASP.

Minor:

1. In GST-pull-down experiments (Fig. 4G), the amount of Arp2/3 complex bound is analyzed by Western, which is rather unprecise. Is the amount of Arp2/3 complex so little that it cannot be quantified using regular SDS-PAGE? If that is the case, this would suggest rather low affinity of SPIN90 for Arp2/3 complex. How does this affect the proposed mechanism and experiments in the microfluidics chamber?

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

An exciting development in our knowledge about how the Arp2/3 complex controls the assembly of actin networks has come from the discovery that in addition to forming branched networks, Arp2/3 can nucleate linear filaments when it is activated by WISH/DIP/SPIN90. However, despite some excellent work largely done by the Nolen lab in yeast, many questions remain about how Arp2/3-mediated assembly of branched vs. linear actin filament. This is especially true in the complex environment of cells, were synergy and competition of different actin networks is used to control biological processes. Knowing the biochemical and physical properties of these different Arp2/3 assemblies will be key to figuring out how they work in cells. Here Cao et al. use an elegant microfluidics based single filament assay system to perform a comparative analysis of the stability of linear and branched Arp2/3 networks. They find interesting differences in how they respond to stabilizing and destabilizing factors. The most striking differences happens when force or aging is applied- both cause debranching of branched networks but have little effect on Spin90-Arp2/3 nucleated filaments.

Major comments:

As a comparative study on the stability of branched vs. linear Arp2/3 nucleated filaments, this manuscript is fairly complete. The key conclusions are well supported by rigorous experiments which can be reproduced by others based on the information provided. However, I am not seeing explicit information on performing biological replicates. This should be included in the manuscript. The use of statistics is largely fine; however I question the use of one statistical test on one figure (see minor comments below).

I would not ask for additional experiments at this time. However, there is an analysis that would be important for interpreting the authors' claims- branch/filament length at the time of dissociation or destabilization of Arp2/3. This would help address if there was a physical tipping point for each type of structure that could explain potential differences they see. The authors should already have this data and the time to complete it would be negligible in delaying publication.

One additional major comment is that the manuscript's title and abstract hint that this paper explores the differences in nucleation of branched vs. linear filaments by Arp2/3. However, the only figure that deals explicitly with nucleation in the paper is Figure 1, which is really just a confirmation that the mammalian proteins used in this study perform similarly to their yeast homologues (Balzer et al, Current Biology 2019). The authors might think about rewording the title/abstract to better reflect that paper really explores the differences in the stability of the two networks

Minor comments:

1 in 12 men and 1 in 200 women are red/green colorblind. Please change the coloring of the schematics and images so that they can be easily seen by all people. This is especially true of the schematics, which are important for understanding exactly what each assay is measuring.

The Introduction is a bit choppy and unfocused. It was difficult to deduce exactly where the paper was going from it. Please consider re-writing it for better clarity. The Discussion on the other hand was fantastic. Great job on interpreting your results in a larger context.

Many figures- while the use of different lightness values of the same color is appreciated in conveying different concentrations of reagents used, there were several instances where it was very hard to read the one on the very bottom (ex. 2B, E; 3A; 5C, G).

Figure 1- since this is a confirmation of previous results performed using the same proteins from other species, the title should reflect that (ex. VCA domains accelerate the nucleation of filaments by mammalian SPIN90-Arp2/3). Also, to me this figure is supplementary to the main message of the paper. The authors might think of moving it to Supplementary Information.

Figure 1- If the goal was to verify that G-actin recruitment by VCA was important for Spin90-Arp 2/3 nucleation by performing a competition experiment with profilin, why was the concentration of G-actin AND profilin increased between the experiments in 1B vs. 1C. It makes it hard to directly compare the results.

Figure 4B-F- Here, it would be nice to see the distribution of all the individual results, which are hidden by the bar graph. Additionally, the Chi-square test is not the appropriate test for evaluating statistical significance between multiple groups. ANOVA followed by an appropriate post hoc test should be used here.

Figure 4G- Please quantify and show reproducibility.

Figure 5- the piconewton forces used for these experiments is in line with measured forces that are applied to actin in cells (ex. Mehida et al, Nature Cell Biology 2021; Jiang et al, Nature 2003). The text would benefit if this was explicitly stated.

Significance

The real significance of this work is in characterizing the differential stabilities of linear vs. branched Arp2/3 filaments in response to actin-binding proteins, mechanical stress, and aging. While both types of filaments respond similarly to actin-binding proteins, with nuanced differences, the most striking results came from applied force and aging experiments, with Spin90-Arp2/3 filaments being much more resistant to both. This has some very interesting implications for how these two types of assemblies might synergize in cells. Additionally, the results also have some exciting implications for the pointed-end regulation of actin filaments, which is still poorly understood in complex systems. Since the manuscript is A) more of a survey study on the factors that influence filament stability that does not go particularly deep into any particular mechanism of regulation and B) has no direct applicability to how the physical properties of branched and linear Arp2/3 nucleated actin filaments influence actin network activity in cells, the audience will likely by limited to actin enthusiasts. However, the work is still important in both what it reveals and implies.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We thank all three reviewers for their thoughtful and rigorous critique of our manuscript, which we feel has significantly improved the presentation of our work. Below we detail point-by-point responses to comments made by the three reviewers as well changes we have already made addressing the majority of minor and some major points.

Specification of the eye-field during gastrulation represents the earliest known stage of eye development. Using an optic-vesicle organoid model system, the overall goal of our work is to provide an unbiased characterisation of this critical, early developmental event in mammals and to gain insights into relevant gene regulatory mechanisms. A common theme to some of the reviewer comments is that this work doesn't provide much of an advance to the field and our findings are not particularly original. We feel that these comments are slightly harsh for the following reasons. Firstly, although some of our findings are not unexpected, to our knowledge, this is the first unbiased characterisation of the eye-field in a mammalian model system, and not based on knowledge gained through previous work in other non-mammalian vertebrate systems, e.g. Xenopus. Secondly, by generating both RNA-seq and ATAC-seq from a timecourse of organoid development we have been able to quantify dynamic patterns of gene-expression as the eye-field is established and simultaneously gain insights to the regulatory role of some of the key transcription factors, both of which are not present in the literature. Thirdly, by constructing careful, integrated analyses of our RNA-seq and ATAC-seq datasets we were able to generate specific hypotheses regarding cis-regulation of key genes, which we have then demonstrated are possible to efficiently test within the organoid system. In all, although we have been purposely careful not to overinterpret our results, we feel our work does represent a significant step towards understanding the mammalian eye field and additionally provides important datasets as well as an analysis framework to begin to quantitatively probe the regulatory mechanisms underlying the transition to an ocular fate. Given the relevance of this developmental event to clinical genetics research as well as to developmental biology we are confident that this work represents an important and significant advance to the literature.

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

Summary Owen et al. characterize the transcriptome and chromatin accessibility of mouse retinal organoids at early stages during which eye field-like cells are specified. Since cell specification and differentiation in retinal organoids largely mimic those processes in vivo, retinal organoids are viable models for studying the mechanisms of early eye development. Owen et al. utilize a previously established Rx-GFP cell line, bulk RNA sequencing, and bulk ATAC sequencing to dissect the mechanisms of early eye development in mice. Their findings are generally consistent with previous studies. Overall, the study is interesting for the field, but its conceptual and technical advances are moderate. In addition, a few major points need to be clarified.

Major points 1. The authors did not show any analysis of retinal organoids at stages when Vsx2 is expressed. This is a significant weakness since the chemically defined medium (CDM) used in Owen et al.'s study was previously shown to induce rostral hypothalamic differentiation (Wataya et al., 2008). Related to this notion, several eye-field transcription factors, such as Rax and Six3, are also expressed in the hypothalamus. Therefore, Owen et al. need to demonstrate that organoids in their modified differentiation system efficiently produce Vsx2-positive retinal progenitors, and samples of organoids at stages when Vsx2 is expressed should be included for RNA sequencing. If Vsx2 is not efficiently expressed in their organoids, the interpretation of results will be very different.

We thank the reviewer for their important comments here. There are several reasons why we are confident that our data and conclusions regarding the organoid eye-field are robust. Firstly, our RNA-seq data, in particular the differences between GFP-positive and GFP-negative cells, clearly show a coordinated up-regulation of the set of canonical eye-field TFs (not individually), which previous studies in Xenopus have shown is a prerequisite for differentiation into anterior eye structures (including retina). Secondly, we have checked that some of the later (in development) eye markers, including Vsx2, are differentially up-regulated (DeSeq2, logfc>1.5, FDRIn all, we are very confident that our approach of using the optic-vesicle organoids and generating molecular data from an organoid developmental timecourse (including sorting), is unpicking the ocular-fate transition event that we are interested in.

1. The authors state that "two differentiation medias were used for this work due to the differentiation becoming unstable after the initial experiments had been performed. The organoids used for RNA and ATAC-seq were grown in CDM media and the organoids with mutations introduced in potential CREs were grown in KSR media". Why the differentiation becomes unstable after the initial experiments? Differences in the two media cause additional complexities. Related to this notion, "WT Rx-GFP" in Figure 4B and 4E appears to show a different expression pattern compared to that in Figure 1A.

We were unable to identify the reason behind the destabilisation of differentiation in CDM media after the cell lines had been through CRISPR despite thorough testing. The differentiation of these cell lines was stabilised enough using KSR media such that every batch of organoids grown contained some organoids that expressed GFP in a pattern similar to what we had seen before and we carried on our experiments using this. We recognise that using two different media adds complexity, however we see the same patterns of organoid growth and GFP expression when differentiating untransfected WT Rax-GFP cells in both of these medias. We have edited Fig.S1 to include representative images of organoids grown in KSR media which can be directly compared to those grown in CDM shown in Figure 1A.

The reviewer has pointed out that the WT Rx-GFP organoids in Figure 6B and 6E show a different expression pattern to those in figure 1A. With the addition of the supplemental figure mentioned above it becomes apparent that these differences are not due to the change of media. We have clarified in the text that these WT cells have also been transfected so as to act as appropriate controls that have been treated identically to the CRISPR edited cell lines and that this has affected their differentiation capacity.

1. Is the deletion of Rax and Six6 regulatory elements homozygous? Sanger sequencing or amplicon sequencing is needed to show the deletion.

The deletions are homozygous (we have stated this in the manuscript text) and as suggested we have added a supplementary figure showing the Sanger sequencing traces for the WT and mutant cell lines used in this study.

1. The deletion of Rax and Six6 regulatory elements appears to cause minor changes in the expression of Rax and Six6 (Figure 6C, F). Therefore, the impact of findings in bulk RNA seq and bulk ATAC seq in this study is still unclear.

We have added a sentence to the text underlining that developmental genes are expected to be regulated by multiple enhancers. Our expectation is therefore, that in perturbing a single putative regulatory element for Rax/Six6, we will very likely not see the complete ablation of Rax/Six6 expression.

1. Retinal organoids and sorted cells are composed of heterogeneous cell populations. Bulk RNA seq and bulk ATAC seq do not have the power to dissect the complexity of heterogeneous cell populations. Single-cell RNA seq and single-cell ATAC seq are more powerful for this study.

We agree with the referee about the fact that the organoids are likely composed of relatively heterogeneous cell populations. We have added this limitation of our generated datasets in a “limitations” paragraph in the discussion.

1. Numerous motifs in the JASPAR database are identified using in vitro assays and have not been validated using in vivo assays. Unexpected results in motif analysis could be due to the differences in DNA binding motifs between in vitro and in vivo conditions. This notion should be added in the discussion.

We have added a couple of sentences in the discussion section, highlighting that TF-motif and footprinting analyses of ATAC-seq data provide indirect evidence of TF binding, and to validate these findings experiments such as ChIP-seq or Cut&Run could be performed in the future.

Minor points

Numerous labels in figures are too small.

We have adjusted the size of a number of the figures to increase the size of the labels, which are now mostly the same size as the text in the corresponding figure captions. We are very happy to make further increases in the sizes of figure labels/text upon recommendation.

CROSS-CONSULTATION COMMENTS

My fellow reviewers identify similar major weaknesses and additional points. I agree with the other reviewers' comments.

Reviewer #1 (Significance (Required)): Nature and Significance of the advances In Owen et al.'s study, the Rx-GFP cell line and retinal differentiation protocol were established in previous studies (Wataya et al., 2008; Eiraku et al., 2011); bulk RNA sequencing and bulk ATAC sequencing are standard procedures. Although candidate regulatory elements for early eye development are identified, deletions of two prioritized elements using CRISPR/Cas9 only cause minor changes in the expression of targeted genes. Overall, conceptual and technical advances in Owen et al.'s study are moderate. Compare to existing published knowledge The datasets could be useful for the field, but conceptual and technical advances are moderate.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

The authors grow eye organoids from cells with a reporter driving GFP in the Rax locus, a gene that is expressed in the eye field in many animal model systems. They show that expression of GFP picks up by day 4 and performed FACS sorting of GFP+ cells on day 4 and day 5 organoids to compare gene expression by RNAseq comparing with earlier day organoids. The data shows 37 genes with a differential expression on days 4 and 5, compared to day 3, and enriched in GFP+ cells, which they define as EF-up genes. It is notable that some of these genes had already been identified as canonical eye field gene regulatory network transcription factors. In the same way, they identify a group of differentially expressed regulated genes, EF-down, and state that 'many' of them are involved in pluripotency. However, they do not mention how many, or the proportion of these genes in the whole list.

The number of EF-down genes with GO terms linked to pluripotency has now been added to the text.

It would be useful if they could provide the number to understand how many of these genes are related to pluripotency, the whole list of genes mentioned to be downregulated in a supplementary file.

We appreciate that this list was missing and will include it now as a supplemental file.

The authors also note that genes known to be required for eye specification like Sox2 and Otx2 are not differentially expressed across the day 3-4 timepoint (Ln 190). However, this is not surprising considering that both genes are broadly expressed in the anterior neural ectoderm and required for its specification, which should be noted by the authors.

We have amended the aforementioned sentence to reflect this: “It is noteworthy that Sox2 and Otx2, known to be crucial in eye development are not differentially expressed across this critical time-point (Fig.2A), consistent with these genes being more broadly expressed in the anterior neuroectoderm in vivo.”

The authors then go on and cluster the EF-up, EF-down and genes deferentially expressed between days 2 and 3, and identify 6 discreet trajectory groups. From this analysis, they identify a third group of genes which shows a peak on day 3 but whose expression falls on days 4 and 5. It is interesting to see that this group includes Wnt and Fgf morphogenes. The authors should provide a list of the genes in the different clusters for the readers to inspect and analyse.

We note that there was a typo in the original manuscirpt and the genes that were clustered were the EF-up and EF-down genes. This typo has been fixed and the requested information is now available in a supplementary file.

Aiming to generate insight into the cis-regulatory elements that regulate of the genes the authors found differentially expressed in their model system they performed a series of ATAC-seq experiments. When linking the genomic regions with differential ATAC-seq accessibility to gene locus using the GREAT analysis, they identified association to 22 of the EF-up and 161 of the EF-down genes. This suggests a functional link between the ATAC-seq genomic regions and the gene regulation of the differentially expressed genes.

The authors later screened the ATAC-seq regions of increased accessibility for TF binding motifs and found that these regions were enriched with motifs for EFTF genes Rax, Lhx2 and Pax6. When assessing motifs in the ATAC-seq regions in EF-up TADs, Rax and Lhx2 motifs scored highly associated to open chromatin positions. Authors also observe a positive gene expression-accessibility correlation between in Pax6, Lhx2, Six3 and Otx2, and suggest this could mean these genes activate transcription of the EF-up group of genes. The same analysis, but focusing on EF-down genes, suggests that EFTFs repress the expression of EF-down genes which include those involved in pluripotency.

Further interrogating the ATAC-seq data, the authors use TOBIAS footprinting analysis to identify changes in TF binding in EF-TADs and EF-up motifs. Remarkably, whole genome analysis reveals that the largest increase in motif binding corresponds to EF-up genes Rax, Pax6 and Lhx2. The authors then narrow down on specific gene regulation by studying the ATAC-seq data within the TAD of Rax and Six6. However, they do not explain the rationale for which these two genes were highlighted, and why Pax6 or Lhx2 were excluded. This explanation should be added to the manuscript.

We have expanded this section of the manuscript to explain that Rax and Six6 were prioritised due to the GFP readout of Rax expression and Six6 being located in a smaller and thus less complex TAD than Pax6, Six3 and Lhx2 after the initial analysis was performed for all five TADs.

The analysis identifies three regulatory elements in the Rax TAD and two for Six6. They then go on and study one putative regulatory element of each gene and generate CRISPR deletions in cell lines. The rationale for the choice of these particular elements is not clear, nor if the cell lines are the same used for the RNAseq experiments. This information should be explicit in the results and in the methods section.

The manuscript has been updated to include the rationale behind our choice of the regulatory elements deleted.

The authors mention that the CRISPR cell lines are "considerably more variable" (Ln 822) compared to the previously studied organoids and suggest that no conclusions can be driven from GFP expression or morphology alone. However, they do not specify which is the variable trait. This information should be added to the text.

We have amended the text to include that the organoids are more variable in terms of the OV like structures produced and GFP expression level.

The authors also miss out on specifying the time stage of the organoids in figure 6 which should be stated.

We thank the reviewer for pointing this out and have updated the manuscript to contain the stage of the organoids.

Regardless, the wildtype organoids in figure 6 and figure S7 show a very different morphology and GFP expression compared to those in figure 1, suggesting that the conclusions from this last set of experiments are not reliable or comparable to those in figure 1. This, together with the fact that different reagents were used to grow the organoids for the RNAseq and the CRISPR experiments, is a weakness of this work that must be addressed.

We recognise this weakness however our amendments detailed above in response to reviewer 1’s comments, including adding a figure showing WT organoids grown in the KSR media that closely resemble the organoids in Fig.1A, removes the uncertainty that it is the change in media producing these differences in morphology and GFP expression.

Our aim in this section was to specifically test the hypotheses regarding the regulatory nature of the distal genomic regions identified by our intra-TAD analyses of ATAC-seq data. To do this it was important to compare organoids derived from wildtype and mutant cells that had been subjected to the same growth conditions and genomic-editing protocols. The stress associated with the latter is what we expect has resulted in the differences in morphology and GFP expression compared with the original Fig1. organoids (which have not been through this procedure).

The last part of the results section belongs to the discussion as no results generated by the researchers are included.

Although no new data was generated for this section, we have used the data generated in our work, together with existing ChIP-seq datasets to construct a new plausible hypothesis regarding the activation of Rax-expression through changes in TF-binding at an enhancer displaying little/no change in accessibility. As this section ties in with previous results sections discussing the regulation of eye-field genes, we feel it belongs in the results section rather than in the discussion.

The discussion in this paper is a good opportunity to state the limitation of this study.

As requested, we have added a paragraph discussing the main limitations to our study in the discussion section.

Major comments to address

1. One of the main issues identified is that the morphology of the control conditions in the CRISPR experiments (Fig.6) do not look is that those used for the RNAseq experiments (Fig.1) and the authors should address this issue. The fact that CDM media was used on the RNA extraction and ATACseq experiments and then KSR media was used for the CRISPR experiments is worrying and makes one wonder whether the second set of experiments is at all comparable to the first. This should be somehow controlled carefully by at least replicating one set of RNA experiments with the KSR media.

We have addressed this in response to the reviewer’s summary above. Unfortunately, it is not possible for us to replicate the RNA experiments in the KSR media due to the research group closure upon Professor FitzPatrick’s retirement.

1. The requirement of Wnt signalling inhibition has been well established as a requirement for forebrain specification, including the eye field. Considering the link of the Wnt/beta-catenin pathway to eye specification and that TCFs, the transcription factors that mediate Wnt pathway transcription regulation, have known and well-studied DNA motifs, it is surprising that authors do not include the analysis of TCF motifs in their study. Also considering that TCF7l1 (TCF3, old nomenclature) has recently been shown to be cell-autonomously required for the expression of rx3 (Rax homologue) in zebrafish. One would expect TCFs to be included in the analysis as it was done with Sox2 and Otx2, which were studied due to the known relevance in forebrain specification rather than from the direct analysis of the differential gene expression experiments.

We thank the referee for their valuable comment here. Our current analyses indeed do not consider TCFs and are therefore likely incomplete. We plan to address this by further analysing our data to quantify the patterns and effects of the TCF genes, and will appropriately amend our manuscript to reflect our findings.

Minor comments to address

1. The authors should clearly state the day timepoint used in the organoids experiments in the results section and figure legends, not just in the methods.

We have updated the text and figure legends to include the time point of all organoids.

1. The report by Agnes et al Development 2022 should be cited in the introduction as it is an excellent paper related to this topic, including a comprehensive analysis of the EFTFs expression pattern.

We thank the reviewer for pointing us to this very interesting paper. Although we feel it doesn’t fit in with our introduction that is currently tailored to the set of genes that has historically defined the eye-field (and which was discovered in non-mammalian models), we do recognise that the 3D organisation of the eye-field and in particular the patterns of gene-expression defining different regions of this is important to disentangle in mammalian systems. We have therefore inserted a reference to the Agnes at al 2022 study on the dimorphic teleost in our extended discussion.

1. Ln 41. Mutations in these genes do not always cause severe bilateral eye malformations. Probably best to moderate and mention that they 'can' cause these malformations.

As suggested we have softened this sentence to: “ Mutations in at least three of the genes encoding orthologs of the Xenopus EFTF can cause severe bilateral eye malformations in humans (OTX2, PAX6 and RAX) (Fitzpatrick and van Heyningen, 2005).”

1. Ln 146. Authors mention that in vitro organoid systems "closely mimic the in vitro regulatory dynamics". This statement should be moderated as we do not know if this is true. In fact, one of the positive aspects of this study is that it contributes to supporting this statement.

We agree with the referee regarding the strength of this original statement. We have changed this to:

“We have exploited a reproducible, in vitro organoid model system enabling us to generate data from this cell-state transition and through computational analysis gain a quantitative understanding of the underlying regulatory mechanisms.”

1. Ln 150. Rax homologue Rx3 is also expressed in cells that give rise to the hypothalamus in zebrafish and cavefish, and probably in Xenopus too. It could well be the case in mice too.

We have corrected this to indicate that Rax is also expressed in the hypothalamus in mice.

1. I do not think the GO term data adds much to Figure 1. If possible, I would move it to the supplementary section.

We have moved the GO visualisations to supplementary, Fig.S2.

1. It should be made clear which set of experiments was performed as biological replicates and which did not.

We have added details on the number of replicates used in each experiment.

1. Based on the heatmap in Fig1A, expression of Rax is significant in GFP- cells at days 4 and 5. The authors should comment or discuss this.

We have amended the text and supplemental methods section to include more details of our FACS protocol. The limitations of our sorting procedure include the fact that cells are not sorted into pure GFP expressing and non-expressing populations. Rather the GFP negative sample may contain some cells with low Rax expression or cells that have just begun to express Rax that were not excluded by our sorting. Our aim was to collect sufficient numbers of cells for each condition and separate out cells that expressed GFP to get a more uniform population of cells to study. It is also of note that the heatmap shows Rax expression by day 3. Although it was not detectable by imaging there were around 100 cells per organoid that FACS marked as GFP positive but were retained within the day 3 sample to ensure we had a complete picture of the gene expression at this time point.

1. Ln 99 of materials and methods mentions that the sorting of GFP+ was performed "when possible". The authors should state the differences in the conditions in the different experiments.

This has been expanded to detail exactly how cells were sorted.

1. The sentence closing the first section of the results (Ln 270) is an overstatement and should be moderated. I cannot see how the results shown in this section on their own could reflect and drive solid conclusions on brain cell fate specification.

We agree with the referee and have changed this sentence to: “In summary, these first analyses of RNA-seq data generated from the timecourse of optic vesicle organoid development, show that this is a robust and relevant model system with which to study the gene dynamics underlying mammalian eye field specification.”

1. Appropriate citations should be added to back up the argument that opens the second part of the results section (starting Ln 279).

We have added several citations that discuss and review the current knowledge regarding gene regulation via TF-binding at accessible cis-regulatory elements.

1. Ln 342-343. I suggest being consistent and using the EF-up or EF-down nomenclature on the whole manuscript unless referring to a different subset of genes.

We have modified the text to consistently use “EF-up” or “EF-down” terminology.

1. Ln 692 Refers to Fig.S4F, but this figure has only panels A-D.

This was a typo and has been corrected in the text.

1. Figures 6B and E and the figure legend do not indicate the differences between the panels, or the time stage of the experiments.

The figure legend has been updated to include these details.

CROSS-CONSULTATION COMMENTS I agree with the comments and suggestions made by the other two reviewers, which identify similar and also specific issues in the manuscript. I believe they are all pertinent and should be acknowledged before re-submitting.

Reviewer #2 (Significance (Required)): The manuscript by Owen et al, presents the analysis of in vitro eye vesicle organoids derived from mouse ESCs at stages equivalent to when the eye field is specified in vivo. This work is pertinent and necessary as detailed data on gene expression in early eye organoids was missing in the field and is necessary for the interpretation of experiments in these systems.

Although the computational data provided in this manuscript is based on consensus TF motifs, the functional relevance of the specific motifs must be proven before being able to drive any significant conclusions, and one should be moderate about the conclusion that can be driven from this kind of analysis. Still, the analysis put forward is a good reference and starting point for future functional studies. One possible limitation of this study is that the quantification of the expression of genes is based on the RNAseq data, and the expression data should be further confirmed using a proper quantitative method like qPCR.

This study will be of interest to the audience studying eye development and disease in animal model systems and humans.

My lab studies the genetic, cellular, and molecular aspects of eye specification, development and disease in zebrafish, and study mutations identified patients with eye globe defects.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Studies in Xenopus embryos have established that the specification of the eye field requires a core set of transcription factors (TFs) that impose eye identity to anterior neural plate progenitors. In this manuscript the authors have used mouse embryonic stem cells-derived optic vesicle organoid to ask if the acquisition of mammalian eye identity requires the same set of TFs. They further use different genomic approaches to identify the cis-regulatory elements involved in the expression of these genes and analyses the consequences of altering the sequence of some of the identified regulatory elements. Their results confirm that in mammals the acquisition of eye field identity requires the upregulation of the expression of the same core set of TFs described in Xenopus, with a particularly important role for three of them: Rax, Pax6 and Lhx2. This upregulation is associated to the downregulation of pluripotency genes.

This is a generally well-performed study, that indeed involves a large amount work and adds the identification of several cis-regulatory elements controlling the expression of this core set of already identified eye field TFs. However, conceptually the study does not add much to what is already known and the authors do not offer any very original conclusion from their study. They have generated a large amount of information that likely could allow them to go beyond what is known. For example, they could enlarge the composition of the gene regulatory network that controls eye field specification, given than one of their argument is that their analysis can predict the composition of such a network. Perhaps, they could also address some of the questions that are posed in the discussion. This will strengthen the manuscript and valorize their work.

Additional points that could be taken into consideration are the following:

1) According to the text, the authors identify only 53 CREs with decreased chromatin accessibility (ATACseq signal) between the 3 day and 5 days timepoints, versus the 7752 CREs with increased signal. However, this contrasts with the proportion of genes upregulated/ downregulated in their RNAseq analysis (37 vs 448) and with the notion that specification of the eye field involves the concomitant repression of other neural fates. This also suggests that at least an important fraction of the dynamic ATACseq peaks associated with 161 of the 448 downregulated genes increase their accessibility and allow the recruitment of transcriptional repressors. However, the role of TF binding and chromatin accessibility dynamics on gene repression is poorly discussed and the authors need to provide some interpretation of these observations. Also, authors interpret the fact that the presence of BS for EF downregulated genes, such as En2 and GATA6, correlates with increased chromatin accessibility as a consequence of the fact that TFBS can be bound by different TF paralogs but do not seem to consider that these TFs have been reported to work as transcriptional repressors, so that their downregulation could well explain the changes in chromatin accessibility.

We thank the reviewer for their interesting comments here. We have added short discussions on both main points above (EF-down genes linked to peaks with increasing accessibility and En2/Gata as transcriptional repressors) in the text related to the analysis of our ATAC-seq data. The notion that a loss of repression leading to the activation of gene-expression is indeed a very exciting one and one that we have thought about in the context of the switch-on of the eye-field TFs. This certainly deserves further future work, however in the present study we wanted to be careful not to overinterpret our data. To robustly gain insights into the loss of repression, experiments such as En1/Gata6 ChIP-seq would be very useful, though we are unable to perform these in the near future.

2) ATACseq signal analysis is an indirect measure of TF binding. The authors demonstrate the predictive nature of this analysis of TF dynamics and have use an available Sox2 ChIP dataset. However, this does not allow assessing dynamic changes in the occupancy of this TF and its correlation with ATACseq. Therefore, at least for few of the TF stressed in this work (e.g. Sox2 and Otx2 and for which good antibodies exist) they could attempt ChIP-seq analysis. This would considerably strengthen the work and provide support to an idea that the authors have particularly emphasized in their manuscript.

We agree with the referee that not having generated ChIP-seq data does not allow us to validate some of the hypotheses and evidence provided by the computational analysis of our ATAC-seq data – we have added a discussion of this limitation in the discussion section of our manuscript. We do note however, as observed in Bentsen et al, 2020, that compared to simple TF-motif occurrence analyses, TF-footprinting analyses (such as those we have performed) yield results on putative TF binding that are much closer to more direct measurements of TF binding via e.g. ChIP-seq. We fully agree that it would be very interesting to perform ChIP-seq/Cut&Cut experiments on the organoid system for a set of interesting TFs identified in our study. Unfortunately, because the lab of Prof FitzPatrick has now closed, it is not possible for us to perform further wet-lab experiments in the very near future. However, we plan to further explore the literature to try to find additional publicly-available ChIP-seq datasets (including for Otx2) which would help reinforce some of the hypotheses we make, and will report any relevant findings in our final manuscript.

3) Previous studies (i.e. 10.1242/dev.067660; 10.1093/hmg/ddt562) have shown the importance of gene dosage in eye field specification and repression of other fates. These studies could be included in the discussion, which, in its current version is a quite brief and leaves out many of the reported analysis.

We thank the referee for pointing us to this very relevant question – we have added this to the further research questions in the discussion.

CROSS-CONSULATION COMMENTS

The comments from the other reviewers complement the aspects that we have underscored and should be fully considered as they will contribute to improve the manuscript.

Reviewer #3 (Significance (Required)): This is a generally well-performed study, that indeed involves a large amount work and adds the identification of several cis-regulatory elements controlling the expression of this core set of already identified eye field TFs. However, conceptually the study does not add much to what is already known and the authors do not offer any very original conclusion from their study. They have generated a large amount of information that likely could allow them to go beyond what is known.

Developmental neurobiologists, genome

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Studies in Xenopus embryos have established that the specification of the eye field requires a core set of transcription factors (TFs) that impose eye identity to anterior neural plate progenitors. In this manuscript the authors have used mouse embryonic stem cells-derived optic vesicle organoid to ask if the acquisition of mammalian eye identity requires the same set of TFs. They further use different genomic approaches to identify the cis-regulatory elements involved in the expression of these genes and analyses the consequences of altering the sequence of some of the identified regulatory elements. Their results confirm that in mammals the acquisition of eye field identity requires the upregulation of the expression of the same core set of TFs described in Xenopus, with a particularly important role for three of them: Rax, Pax6 and Lhx2. This upregulation is associated to the downregulation of pluripotency genes.

This is a generally well-performed study, that indeed involves a large amount work and adds the identification of several cis-regulatory elements controlling the expression of this core set of already identified eye field TFs. However, conceptually the study does not add much to what is already known and the authors do not offer any very original conclusion from their study. They have generated a large amount of information that likely could allow them to go beyond what is known. For example, they could enlarge the composition of the gene regulatory network that controls eye field specification, given than one of their argument is that their analysis can predict the composition of such a network. Perhaps, they could also address some of the questions that are posed in the discussion. This will strengthen the manuscript and valorize their work. Additional points that could be taken into consideration are the following:

1. According to the text, the authors identify only 53 CREs with decreased chromatin accessibility (ATACseq signal) between the 3 day and 5 days timepoints, versus the 7752 CREs with increased signal. However, this contrasts with the proportion of genes upregulated/ downregulated in their RNAseq analysis (37 vs 448) and with the notion that specification of the eye field involves the concomitant repression of other neural fates. This also suggests that at least an important fraction of the dynamic ATACseq peaks associated with 161 of the 448 downregulated genes increase their accessibility and allow the recruitment of transcriptional repressors. However, the role of TF binding and chromatin accessibility dynamics on gene repression is poorly discussed and the authors need to provide some interpretation of these observations. Also, authors interpret the fact that the presence of BS for EF downregulated genes, such as En2 and GATA6, correlates with increased chromatin accessibility as a consequence of the fact that TFBS can be bound by different TF paralogs but do not seem to consider that these TFs have been reported to work as transcriptional repressors, so that their downregulation could well explain the changes in chromatin accessibility.
2. ATACseq signal analysis is an indirect measure of TF binding. The authors demonstrate the predictive nature of this analysis of TF dynamics and have use an available Sox2 ChIP dataset. However, this does not allow assessing dynamic changes in the occupancy of this TF and its correlation with ATACseq. Therefore, at least for few of the TF stressed in this work (e.g. Sox2 and Otx2 and for which good antibodies exist) they could attempt ChIP-seq analysis. This would considerably strenghen the work and provide support to an idea that the authors have particularly emphasized in their manuscript.
3. Previous studies (i.e. 10.1242/dev.067660; 10.1093/hmg/ddt562) have shown the importance of gene dosage in eye field specification and repression of other fates. These studies could be included in the discussion, which, in its current version is a quite brief and leaves out many of the reported analysis.

Referees cross-commenting

The comments from the other reviewers complement the aspects that we have underscored and should be fully considered as they will contribute to improve the manuscript

Significance

This is a generally well-performed study, that indeed involves a large amount work and adds the identification of several cis-regulatory elements controlling the expression of this core set of already identified eye field TFs. However, conceptually the study does not add much to what is already known and the authors do not offer any very original conclusion from their study. They have generated a large amount of information that likely could allow them to go beyond what is known.

Developmental neurobiologists, genome

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The authors grow eye organoids from cells with a reporter driving GFP in the Rax locus, a gene that is expressed in the eye field in many animal model systems. They show that expression of GFP picks up by day 4 and performed FACS sorting of GFP+ cells on day 4 and day 5 organoids to compare gene expression by RNAseq comparing with earlier day organoids. The data shows 37 genes with a differential expression on days 4 and 5, compared to day 3, and enriched in GFP+ cells, which they define as EF-up genes. It is notable that some of these genes had already been identified as canonical eye field gene regulatory network transcription factors. In the same way, they identify a group of differentially expressed regulated genes, EF-down, and state that 'many' of them are involved in pluripotency. However, they do not mention how many, or the proportion of these genes in the whole list. It would be useful if they could provide the number to understand how many of these genes are related to pluripotency, the whole list of genes mentioned to be downregulated in a supplementary file. The authors also note that genes known to be required for eye specification like Sox2 and Otx2 are not differentially expressed across the day 3-4 timepoint (Ln 190). However, this is not surprising considering that both genes are broadly expressed in the anterior neural ectoderm and required for its specification, which should be noted by the authors.

The authors then go on and cluster the EF-up, EF-down and genes deferentially expressed between days 2 and 3, and identify 6 discreet trajectory groups. From this analysis, they identify a third group of genes which shows a peak on day 3 but whose expression falls on days 4 and 5. It is interesting to see that this group includes Wnt and Fgf morphogenes. The authors should provide a list of the genes in the different clusters for the readers to inspect and analyse. Aiming to generate insight into the cis-regulatory elements that regulate of the genes the authors found differentially expressed in their model system they performed a series of ATAC-seq experiments. When linking the genomic regions with differential ATAC-seq accessibility to gene locus using the GREAT analysis, they identified association to 22 of the EF-up and 161 of the EF-down genes. This suggests a functional link between the ATAC-seq genomic regions and the gene regulation of the differentially expressed genes.

The authors later screened the ATAC-seq regions of increased accessibility for TF binding motifs and found that these regions were enriched with motifs for EFTF gens Rax, Lhx2 and Pax6. When assessing motifs in the ATAC-seq regions in EF-up TADs, Rax and Lhx2 motifs scored highly associated to open chromatin positions. Authors also observe a positive gene expression-accessibility correlation between in Pax6, Lhx2, Six3 and Otx2, and suggest this could mean these genes activate transcription of the EF-up group of genes. The same analysis, but focusing on EF-down genes, suggests that EFTFs repress the expression of EF-down genes which include those involved in pluripotency.

Further interrogating the ATAC-seq data, the authors use TOBIAS footprinting analysis to identify changes in TF binging in EF-TADs and EF-up motifs. Remarkably, whole genome analysis reveals that the largest increase in motif binging corresponds to EF-up genes Rax, Pax6 and Lhx2. The authors then narrow down on specific gene regulation by studying the ATAC-seq data within the TAD of Rax and Six6. However, they do not explain the rationale for which these two genes were highlighted, and why Pax6 or Lhx2 were excluded. This explanation should be added to the manuscript. The analysis identifies three regulatory elements in the Rax TAD and two for Six6. They then go on and study one putative regulatory element of each gene and generate CRISPR deletions in cell lines. The rationale for the choice of these particular elements is not clear, nor if the cell lines are the same used for the RNAseq experiments. This information should be explicit in the results and in the methods section. The authors mention that the CRISPR cell lines are "considerably more variable" (Ln 822) compared to the previously studied organoids and suggest that no conclusions can be driven from GFP expression or morphology alone. However, they do not specify which is the variable trait. This information should be added to the text. The authors also miss out on specifying the time stage of the organoids in figure 6 which should be stated. Regardless, the wildtype organoids in figure 6 and figure S7 show a very different morphology and GFP expression compared to those in figure 1, suggesting that the conclusions from this last set of experiments are not reliable or comparable to those in figure 1. This, together with the fact that different reagents were used to grow the organoids for the RNAseq and the CRISPR experiments, is a weakness of this work that must be addressed.

The last part of the results section belongs to the discussion as no results generated by the researchers are included. The discussion in this paper is a good opportunity to state the limitation of this study.

Major comments to address

1. One of the main issues identified is that the morphology of the control conditions in the CRISPR experiments (Fig.6) do not look is that those used for the RNAseq experiments (Fig.1) and the authors should address this issue. The fact that CDM media was used on the RNA extraction and ATACseq experiments and then KSR media was used for the CRISPR experiments is worrying and makes one wonder whether the second set of experiments is at all comparable to the first. This should be somehow controlled carefully by at least replicating one set of RNA experiments with the KSR media.
2. The requirement of Wnt signalling inhibition has been well established as a requirement for forebrain specification, including the eye field. Considering the link of the Wnt/beta-catenin pathway to eye specification and that TCFs, the transcription factors that mediate Wnt pathway transcription regulation, have known and well-studied DNA motifs, it is surprising that authors do not include the analysis of TCF motifs in their study. Also considering that TCF7l1 (TCF3, old nomenclature) has recently been shown to be cell-autonomously required for the expression of rx3 (Rax homologue) in zebrafish. One would expect TCFs to be included in the analysis as it was done with Sox2 and Otx2, which were studied due to the known relevance in forebrain specification rather than from the direct analysis of the differential gene expression experiments.

Minor comments to address

1. The authors should clearly state the day timepoint used in the organoids experiments in the results section and figure legends, not just in the methods.
2. The report by Agnes et al Development 2022 should be cited in the introduction as it is an excellent paper related to this topic, including a comprehensive analysis of the EFTFs expression pattern.
3. Ln 41. Mutations in these genes do not always cause severe bilateral eye malformations. Probably best to moderate and mention that they 'can' cause these malformations.
4. Ln 146. Authors mention that in vitro organoid systems "closely mimic the in vitro regulatory dynamics". This statement should be moderated as we do not know if this is true. In fact, one of the positive aspects of this study is that it contributes to supporting this statement.
5. Ln 150. Rax homologue Rx3 is also expressed in cells that give rise to the hypothalamus in zebrafish and cavefish, and probably in Xenopus too. It could well be the case in mice too.
6. I do not think the GO term data adds much to Figure 1. If possible, I would move it to the supplementary section.
7. It should be made clear which set of experiments was performed as biological replicates and which did not.
8. Based on the heatmap in Fig1A, expression of Rax is significant in GFP- cells at days 4 and 5. The authors should comment or discuss this.
9. Ln 99 of materials and methods mentions that the sorting of GFP+ was performed "when possible". The authors should state the differences in the conditions in the different experiments.
10. The sentence closing the first section of the results (Ln 270) is an overstatement and should be moderated. I cannot see how the results shown in this section on their own could reflect and drive solid conclusions on brain cell fate specification.
11. Appropriate citations should be added to back up the argument that opens the second part of the results section (starting Ln 279).
12. Ln 342-343. I suggest being consistent and using the EF-up or EF-down nomenclature on the whole manuscript unless referring to a different subset of genes.
13. Ln 692 Refers to Fig.S4F, but this figure has only panels A-D.
14. Figures 6B and E and the figure legend do not indicate the differences between the panels, or the time stage of the experiments.

Referees cross-commenting

I agree with the comments and suggestions made by the other two reviewers, which identify similar and also specific issues in the manuscript. I believe they are all pertinent and should be acknowledged before re-submitting.

Significance

The manuscript by Owen et al, presents the analysis of in vitro eye vesicle organoids derived from mouse ESCs at stages equivalent to when the eye field is specified in vivo. This work is pertinent and necessary as detailed data on gene expression in early eye organoids was missing in the field and is necessary for the interpretation of experiments in these systems.

Although the computational data provided in this manuscript is based on consensus TF motifs, the functional relevance of the specific motifs must be proven before being able to drive any significant conclusions, and one should be moderate about the conclusion that can be driven from this kind of analysis. Still, the analysis put forward is a good reference and starting point for future functional studies. One possible limitation of this study is that the quantification of the expression of genes is based on the RNAseq data, and the expression data should be further confirmed using a proper quantitative method like qPCR.

This study will be of interest to the audience studying eye development and disease in animal model systems and humans.

My lab studies the genetic, cellular, and molecular aspects of eye specification, development and disease in zebrafish, and study mutations identified patients with eye globe defects.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary

Owen et al. characterize the transcriptome and chromatin accessibility of mouse retinal organoids at early stages during which eye field-like cells are specified. Since cell specification and differentiation in retinal organoids largely mimic those processes in vivo, retinal organoids are viable models for studying the mechanisms of early eye development. Owen et al. utilize a previously established Rx-GFP cell line, bulk RNA sequencing, and bulk ATAC sequencing to dissect the mechanisms of early eye development in mice. Their findings are generally consistent with previous studies. Overall, the study is interesting for the field, but its conceptual and technical advances are moderate. In addition, a few major points need to be clarified.

Major points

1. The authors did not show any analysis of retinal organoids at stages when Vsx2 is expressed. This is a significant weakness since the chemically defined medium (CDM) used in Owen et al.'s study was previously shown to induce rostral hypothalamic differentiation (Wataya et al., 2008). Related to this notion, several eye-field transcription factors, such as Rax and Six3, are also expressed in the hypothalamus. Therefore, Owen et al. need to demonstrate that organoids in their modified differentiation system efficiently produce Vsx2-positive retinal progenitors, and samples of organoids at stages when Vsx2 is expressed should be included for RNA sequencing. If Vsx2 is not efficiently expressed in their organoids, the interpretation of results will be very different.
2. The authors state that "two differentiation medias were used for this work due to the differentiation becoming unstable after the initial experiments had been performed. The organoids used for RNA and ATAC-seq were grown in CDM media and the organoids with mutations introduced in potential CREs were grown in KSR media". Why the differentiation becomes unstable after the initial experiments? Differences in the two media cause additional complexities. Related to this notion, "WT Rx-GFP" in Figure 4B and 4E appears to show a different expression pattern compared to that in Figure 1A.
3. Is the deletion of Rax and Six6 regulatory elements homozygous? Sanger sequencing or amplicon sequencing is needed to show the deletion.
4. The deletion of Rax and Six6 regulatory elements appears to cause minor changes in the expression of Rax and Six6 (Figure 6C, F). Therefore, the impact of findings in bulk RNA seq and bulk ATAC seq in this study is still unclear.
5. Retinal organoids and sorted cells are composed of heterogeneous cell populations. Bulk RNA seq and bulk ATAC seq do not have the power to dissect the complexity of heterogeneous cell populations. Single-cell RNA seq and single-cell ATAC seq are more powerful for this study.
6. Numerous motifs in the JASPAR database are identified using in vitro assays and have not been validated using in vivo assays. Unexpected results in motif analysis could be due to the differences in DNA binding motifs between in vitro and in vivo conditions. This notion should be added in the discussion.

Minor points

Numerous labels in figures are too small.

Referees cross-commenting

My fellow reviewers identify similar major weaknesses and additional points. I agree with the other reviewers' comments.

Significance

Nature and Significance of the advances

In Owen et al.'s study, the Rx-GFP cell line and retinal differentiation protocol were established in previous studies (Wataya et al., 2008; Eiraku et al., 2011); bulk RNA sequencing and bulk ATAC sequencing are standard procedures. Although candidate regulatory elements for early eye development are identified, deletions of two prioritized elements using CRISPR/Cas9 only cause minor changes in the expression of targeted genes. Overall, conceptual and technical advances in Owen et al.'s study are moderate.

Compare to existing published knowledge

The datasets could be useful for the field, but conceptual and technical advances are moderate.

Audience

Developmental biologists, stem cell biologists, vision researchers.

Your expertise

Developmental biology, stem cell biology, vision research

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to provide a response at this time.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The authors have utilised single-cell RNA seq to profile cells in the Drosophila female germline cells and somatic cells within the ovary. Using a candidate approach, the authors assessed whether some of the candidate genes identified in this study play a role in germ cell differentiation. Amongst these, they found that an uncharacterized gene called eggplant is required for differentiation. On rich diets, Eggplant expression is reduced. Furthermore, Mmps and Timp are required to regulate Eggplant expression level.

Major comments:

Overall, the conclusions are supported by experimental evidence. The sc-RNA seq provides an important resource for the community, and some of the reagents generated such as the antibody against Eggplant or the tagged lines would provide valuable resource for future studies. The main criticism I have for this manuscript is that it lies half way between a resource paper and a mechanistic paper. If it was just a resource paper, then the sc data should suffice. Instead, the authors studied the effects of Eggplant and MMP/TIMP on GSC differentiation, however, not enough mechanism was gained through these studies.

1. It is not clear what kind of protein Eggplant is, and its mode of action is unclear, despite that it was shown by the authors that Eggplant is required for GSC differentiation. The authors further showed that animlas fed a high yeast diet promotes GSC proliferation and increased egg production by inhibiting the expression of eggpl. The authors suggested that this may lie downstream of insulin signaling, however no genetic experiments were conducted. In order to support this line of conclusions, essential experiments should include: is there a genetic interaction between Eggplant and insulin signaling? Is this happening autonomously within the GSCs or at the GSC niche? In the discussion the authors mentioned that since egg production can occur upon Eggpl knockdown even on normal diet, thus protein is not involved in this process. So, what is the mechanism?
2. Similarly, the authors mentioned that MMPs and TIMP are involved in GSC differentiation, and they discussed that it could be mediated via tissue stiffness. However, no experimental data was presented to support this. It would be good to offer some mechanistic insights. TIMP and Eggplant genetic interaction should be done.

Minor comments:

Line 376 and Line 437 are repetitive in explaining why the gene is called eggplant. I did not find the pictures in Figure7 F,G and Figure 8 B informative. Some quantifications should also proof the same point.

Referees cross-commenting

I also agree with the comments from both reviewers 1 and 2.

Significance

• The significance of this study lies in presentation of a thorough characterization of gene expression at the single cell level in the GSC and in identification of novel regulators of GSC differentiation.
• My field of expertise is Drosophila stem cells and nutrition/metabolism.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Brief summary

The authors conducted single cell RNA sequencing on adult fly ovaries to survey the transcriptomic profile of individual ovarian cells. The datasets resulted in classification of 24 discrete cellular populations, including different types of stem cells, progenitors and differentiated cells. Genes that differentially expressed during germline differentiation were selected and then examined with RNA interference to determine their specific roles in regulating germline differentiation. Meanwhile, using the single cell transcriptomic profiles acquired from specific types of germ cells, gene regulatory networks were further characterized. Among the genes examined, the gene eggplant (eggpl) was found to act as a novel regulator mediating germline differentiation with its protein expression predominantly detected in germline stem cells (GSCs), CBs and early cysts. Interestingly, since the protein expression of Eggplant was decreased upon rich dietary intake, eggpl may function as a novel molecular link coupling the nutritional status and the process of germline differentiation during fly oogenesis.

Specific comments

Major points:

1. The authors chose a specific group of differentially upregulated genes (in GSCs) and performed germline specific RNA interference (RNAi) to determine their function in regulating germline differentiation. The RNAi results were then compiled and presented in Figure 3B-C. However, details shall be provided for the readers/reviewers to understand the phenotypic analyses. Specifically, as stated in text that "we found that RNAi knockdown of 19 upregulated genes induced disruption of GSCs/CB homeostasis. Of these, 12 genes were classified as "changes to the number of CSC/CB", 6 genes as "empty germarium" and 4 genes exhibited "differentiation defects"... ", there was neither specific descriptions nor representative examples (with proper labelling) explaining the bases of such classification. It was particularly difficult to appreciate the assorted phenotypes presented in Figure 3C without proper descriptions. Not to mention that the immunostaining needs to be optimized in some of the representative pictures. Meanwhile, it was not clearly stated in the text about how 19 candidate genes were then classified into (12+6+4=22) genes. Along the same line, it will be helpful to include a supplementary table of summarizing the genes tested and their corresponding phenotypes.
2. The authors also performed germline specific RNAi to knockdown the expression of 21 out of 39 most highly expressing genes in GSCs to determine their function in regulating oogenesis. Some of the RNAi effects were presented in Figure S2B. Similarly, the effects of RNAi experiments will be better appreciated if some more descriptions on specific RNAi phenotypes are added. Also, inclusion of a supplementary table summarizing the RNAi lines utilized in this experiment is required.
3. The authors performed in situ hybridization and immunostaining to discover that the eggpl transcripts were enriched in GSCs and CBs (Bam-GFP- germ cells), while the Eggpl::GFP expression (in the knock-in line, KI) was detected throughout region 1 within germarium. The authors then concluded that protein level of Eggpl was "actually highest in cells that have no detectable eggpl transcripts". However, as shown in Figure 5F, the highest level of Eggpl::GFP appeared in the cell that contains a spectrosome, suggesting that is either a GSC or a CB obtaining highest Eggpl::GFP (KI) expression. Furthermore, as shown in Figure 4E, similar level of Eggpl protein expression indicated by immunostaining was found in GSCs and CBs, indicating a nice coupling between the mRNA and protein level of eggpl found in GSCs/CBs. Therefore, the authors need to simultaneously perform in situ hybridization and immunostaining of eggpl to visualize the expression of eggpl transcripts and protein within the same germarium and to directly determine whether the statement that "protein level of Eggpl was actually highest in cells that have no detectable eggpl transcripts" holds true.
4. The genetic experiments done by the authors show that disruption of eggpl led to an increase in the average number of Brdu+ cysts, the size of the ovary and ultimately the number of eggs produced. Interestingly, such phenotypes were reminiscent to the effects caused by rich dietary intakes. The authors interpret these results by proposing that eggpl negatively regulates GSC proliferation in regulating oogenesis. However, as shown in Figure 5J, the number of GSCs was not significantly altered upon eggpl RNAi but was increased when over-expressing Eggpl::GFP, indicating that eggpl promotes the proliferation of GSCs. To be able to determine the function of eggpl in regulating GSC proliferation during oogenesis and to support the stamen that "We also found that cell cycle in germ cell cysts was accelerated" (in introduction), the authors need to perform clonal analyses to monitor the progression rate of eggpl mutant cells (of both germline clones and follicle cell clones) compared to control cells. An accelerated rate of germline development shall be predicted by increased GSC proliferation. Moreover, to rule out the possibility that changes in the dynamics of germline apoptosis can affect the number of Brdu+ cysts observed upon eggpl loss, the authors should also perform experiments assaying the number of apoptotic cells in the germarium when manipulating the expression level of eggpl.
5. The authors examined the expression level of Eggpl::GFP (KI) in different feeding conditions and found a reverse correlation between Eggpl::GFP (KI) protein level and the quality of dietary condition. Moreover, the enlarged ovary seen upon eggpl loss was recapitulated when over-expressing Timp and Mmp1/2. Interestingly, the authors also found that overexpression of Timp and Mmp1/2 led to lower protein expression of Eggpl::GFP (KI). The authors interpret these results by proposing that eggpl mediates the GSC differentiation via MMP-dependent Timp regulation pathway. However, the authors should perform experiments to directly examine if eggpl and Timp (and Mmp1/2) interact genetically to regulate GSC proliferation/egg production while monitoring the dietary status.

Minor points:

1. Please clarify what you mean in this sentence in the introduction as the message is not clear: "Mei-P26 suppress transcripts that promote differentiation in CB by antagonizing miRNA pathway
2. In Figure 3B, for the genes that were listed for more than once (CG42250, CG10295, CG6904 and CG15845), please note that those are different RNAi lines.
3. In Figure S2A, for the genes that were listed for more than once, please note that those are different RNAi lines.
4. If applicable, more than one RNAi line of each candidate gene should be examined to validate the specific effects caused by gene KD.
5. As shown in Figure S2B, knocking down CG31666 and CG12743 led to impaired oogenesis. Please add their gene name (CG31666 (chinmo) and CG12743 (otu)) for acknowledging their well-documented role in regulating oogenesis.
6. How was the RNAi generally performed. Please add into material and method.
7. In Figure 7A, it is difficult to tell how many of the BrdU+ cysts was shown in Eggpl[1] mutant germarium and whether it is qualitatively/quantitively different from control.
8. The authors compared their scRNA-Seq dataset with other 4 public available Drosophila ovary scRNA-Seq datasets and concluded that comparable features were found among these datasets by UMAP analysis. The authors should also comment on what are the potential differences among these adult ovary scRNA-Seq datasets in the discussion section. Providing such information will not only highlight the importance of this dataset and will also be beneficial to the fly community for future research.

Referees cross-commenting

I also agree with the comments from both reviewers 1 and 3.

Significance

In summary, the characterization of single-cell atlas of adult fly ovary from this study adds additional datasets for future investigation on cell-specific differentiation in vivo. Meanwhile, the identification of a novel molecular link between nutritional status and germline differentiation is of broad interest to readers in the field of dietary/metabolic control on stem cell function. However, there are specific concerns about how the results were presented and interpreted while some important controls are missing. I believe this article would be much improved if the points mentioned can be effectively addressed.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript, Sun et al. conduct single-cell RNAseq on germline stem cells in the Drosophila ovary to better understand the transcriptional profile of this small, but important, population of stem cells. This resulted in the identification of a subset of genes and gene networks proposed to be involved in early germ cell development and differentiation. A subsequent RNAi screen of these candidate genes indicated that an uncharacterized one (renamed eggplant) was expressed at the transcriptional level in GSCs and at the protein level in the germarium. RNAi-mediated knockdown or CRISPR/Cas9-mediated knockout of eggplant resulted in an accelerated cell cycle in germ cell cysts and flies with overall larger ovaries. These results were not observed, however, when flies were fed a rich yeast-based diet. The authors also draw a connection between the MMP-Timp pathway and the regulation of eggplant in ovarian germ cells. Collectively, the findings presented in this manuscript detail novel modulators of germ cell differentiation and proliferation that are regulated by nutrient availability. While the data presented are intriguing (especially in the first part of the manuscript), there are several major and minor issues that should be addressed prior to publication.

Major issues:

1. The bioinformatic analysis of the scRNA-seq data is elegant and well done. However, the data from the follow-up studies on eggplant are mostly observational, correlative, and not convincing. There is little to no mechanism described regarding the function of eggplant. This may be beyond the scope of this paper, but it should at least be addressed in more detail in the Discussion section.
2. Along the lines of point 1, what is the amino acid sequence of eggplant? What is its predicted molecular weight? Does it contain any predicted domains that may hint at its function in the germarium? The answers to these questions should be included in the manuscript.
3. There is no validation of the novel eggplant antibody that was generated. This needs to be demonstrated in order to be confident that the antibody is working as expected to interpret the immunofluorescence data.
4. In Figure 8, the connection between MMPs, TIMPs, and eggplant is very weak. It is unclear how the regulation is occurring. Is it direct or indirect? And by what mechanism? Also need to show representative images to go along with the graphs. But as is, this is the weakest data in the manuscript and needs more clarification.

Minor issues:

1. It is unclear what it is being shown in Figure 5M.
2. In Figure 7B, how was the egg-laying assay performed. This information was not included in the Materials and Methods section.
3. There are some grammar issues in the Introduction section.

Referees cross-commenting

My comments align closely with those of Reviewer #3. I also agree that the specific points raised by Reviewer #2 should be addressed to strengthen the manuscript.

Significance

The significance of this work lies in the transcriptional profiling of the Drosophila ovarian germline stem cells. These cells are scarce and are notoriously difficult to isolate and study. This work makes headway in that regard, laying the foundation for future studies, and as mentioned, the bioinformatic analysis of the scRNA-seq data is the strength of the study.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity):

Summary<br /> Authors show that overexpression of bHLH transcription factor Dpn in the medullary neurons of the Drosophila optic lobe results in the dedifferentiation of these neurons back into the NBs. These dedifferentiated NBs acquire and maintain mid-temporal identity, express Ey and Slp, and show delayed onset of tTF Tailless (Tll), leading to an excess of neurons of mid-temporal fate at the expense of late temporal fate neurons and glial cells. The dedifferentiated NBs are stalled in the cell cycle and fail to undergo terminal differentiation. Over expression of tTF Dicheate (D) or promoting G1/S transition pushed these NBs to late stages of the temporal series, partly rescuing the neuronal diversity and causing their terminal differentiation. They also show that the dedifferentiation of NBs by Notch hyper-activation also exhibited stalled temporal progression, which is restored by D overexpression.<br /> Authors suggest that cell cycle regulation and tTF are primary to the proliferation and termination profile of dedifferentiated NBs.<br /> Using these conclusions, the authors emphasize the need to recreate the right temporal profile and ensure appropriate cell cycle progression to use dedifferentiated NSC for regenerative purposes or prevent tumorigenesis originating from differentiated cell types.

Major comments:<br /> - Are the key conclusions convincing?<br /> Most conclusions are convincing; however, some issues are pointed out below.

• Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

The authors have overexpressed Dpn and shown that medulla neurons dedifferentiate to NBs, similar to the loss of function phenotype seen for the Nerfin-1 of which Dpn is a target. They also show that temporal series progression defect is also seen in the case of dedifferentiated NB generated by Notch over-activation.<br /> Using these two examples, the authors suggest that for dedifferentiated NSC, which are to be used for the regenerative purpose, one needs to recreate the right temporal profile and ensure cell cycle progression occurs appropriately. Authors also claim that to prevent tumorigenesis originating from differentiated cell types, one needs to recreate the right temporal profile and ensure cell cycle progression occurs appropriately.

While I agree with this, I think this is an overreaching conclusion based on just these two examples. If they could show the same for one more method of dedifferentiation (For, e.g. Lola) happening in medulla neurons which happens by a mechanism independent of Nerfin-1, Dpn, Notch axis, the argument will become more convincing and broad.

We will characterise the temporal identity, termination and cellular identity of Lola-Ri induced ectopic neuroblasts. If these parameters are disrupted, we will overexpress D to assess whether this can trigger the progression of the temporal series.

Also when authors mention N mediated dedifferentiation, they need to inform that Dpn is a direct target of Notch in NBs (Doi. 10.1016/j.ydbio.2011.01.019), they do so in the discussion, but mentioning it here gives a broader context to the reader.

We will include that Dpn is a target of Notch when first mentioned.

Another important point that needs a mentioned here is that conclusions are based on dedifferentiation happening in the medulla neurons, which are considered less stable since they lack Prospero. Therefore whether this conclusion can be generalized for all the tumors arising from dedifferentiation in the CNS (eg, those arising from NICD activation in the central brain or thoracic region of the VNC) is another concern. Maybe authors can consider making a more conservative claim.<br /> Generalizing this conclusion to Prospero expressing NBs lies outside the scope of the current study and cannot be addressed here because central brain Type-I NBs use a different set of tTFs.

We will make a more conservative claim and clarify all of our conclusion are medulla neuron-specific.

Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.<br /> Experiments with Lola knockdown/mutants in medulla neurons can be done quickly, in my opinion, and will substantiate this claim.<br /> Another obvious question that comes to mind is if medulla neurons dedifferentiate on overexpression of Dpn, does the same happen in nerfin-1 mutant clones as well? And if yes, why has the author not done similar experiments for nerfin-1 mutants.

We will assess the temporal identity of neuroblasts in nerfin-1 mutant clones.

Please show Ey staining in Fig-2 if possible, it will also help to add a line on why Slp was used as marker for mid tTFs instead of Ey.

Ey is shown in Fig-2 (D-D’’) already. Slp is used as a marker of mid tTFs as Ey is expressed also in neurons thus would also be present in deep sections of control clones, whereas Slp is not expressed in neurons. We therefore used Slp as a proxy for mid-temporal identity throughout our study. We will include this text in our revision.

In Model shown in last figure Dpn is shown to repress D and activate Slp. Can authors show that Dpn overexpression represses D and activate Slp either by antibody staining or by RT PCR.

In Figure 2H, we have shown in clones that overexpression of Dpn induced a significant increase of Slp. In Figure S3B-B’’, we have shown that Dpn overexpression causes an upregulation of Slp at 6 hr APF. We can think we have pretty convincingly shown that Dpn overexpression activates Slp.

For Dichaete, our existing data shows that Dpn overexpression did not significantly alter D expression. To assess if using a stronger driver might allow us to see some changes, we will induced dedifferentiation via Dpn overexpression using the Eyeless-Gal4 driver. In this experiment, we will quantify the amount of D upon Dpn overexpression. Depending on this result, we will revise our conclusion on whether Dpn overexpression represses D.

Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.<br /> Experiments with Lola and nerfin-1 mutants can be done in a few months. I cannot comment on the cost involved.<br /> - Are the data and the methods presented in such a way that they can be reproduced?<br /> Yes

Are the experiments adequately replicated and statistical analysis adequate?<br /> Replication and statistical analysis are fine. The activated Notch experiments show only three data points in all the experiments. It will be good to increase this number.

We will repeat Notch experiments to increase the n number for these experiments.

Minor comments:<br /> - Specific experimental issues that are easily addressable.<br /> There is a problem with Fig-5F (both 5E and 5F have % EdU in clone/ % Mira in the clone as y-axis), I do not understand how the Fig-5F let them conclude that D overexpression increases the rate of neuronal production.

In the text we said: “We found that D overexpression did not significantly increase neuronal production, suggesting that it is likely that cell cycle progression lies upstream or in parallel to the temporal series, to promote the generation of neurons.”

In one place, the authors conclude, "Together, this data suggests that it is likely that cell cycle progression lies upstream of the temporal series, to promote the generation of neurons". Authors should consider adding "medulla NBs" at the end of the sentence since cell cycle progression being upstream of temporal series is already known in Type-I NBs, as pointed out by authors as well (Ameele and Brand 2019).

We will add “medulla NBs” to the end of this sentence.

In the discussion authors says that "Our data support the possible links between cell cycle progression and the expression of temporal regulators controlling NB proliferation and cellular diversity". This is new information, as the 2019 study did not show how cell diversity changes with a changed tTF profile. I think the authors should elaborate on this point to highlight how this is different from what is already known from the 2019 study (done in the context of Type-I NBs).<br /> Maybe they need to highlight that the cell cycle directs/regulates the progression of temporal series compared to the earlier observation where temporal series was shown to be downstream of the cell cycle.

We will expand in discussion to discuss the link between cell cycle/tTFs.

In fig-3J in clones even after 24 AHS, Dpn continues to be overexpressed but these cells undergo terminal differentiation, can authors comment why is it so?<br /> In one place authors say, "To better assess the cumulative effect of the neurons made throughout development, EyOK107-GAL4 was used to drive the expression of Dpn" maybe some background on why use this specific GAL4.<br /> Also a line about why GMR31HI08-GAL4 eyOK107-GAL4 and and eyR16F10-GAL4 were used.

While Dpn is overexpressed, it progresses through the temporal series at a slower pace due to a delay in cell cycle progression, as well as delayed onset of D, these NBs still eventually reach the terminal temporal identity, and are thus about to undergo terminal differentiation. We will include an additional piece of data that shows NBs induced by Dpn overexpression do eventually turn on Tll.

Are prior studies referenced appropriately ?<br /> Yes, but in a few places, some references can be added.<br /> An important point that needs to be mentioned for the context is the medulla neurons do not use Prospero for terminal differentiation and are thus considered less stable (DOI: 10.1242/dev.14134

We beg to disagree with the reviewer in terms of Pros is not required for terminal differentiation of medulla neuroblasts. Li et al., 2013 shows that nuclear Pros is found in the oldest NBs. We do agree that differentiated state of medulla neurons is less stable, possibly owing to absence of Pros, and we will include that in our discussion.

In discussion, the authors say that "It would be interesting to explore whether N similarly acts on these target genes to specify cell fate and proliferation profiles of dedifferentiated NBs." There is a study looking at Notch targets in NB hyperplasia (DOI: 10.1242/dev.126326); whether that study shows if any of the cell cycle genes are downstream of activated Notch, needs a mention here.<br /> Also, when authors mention N mediated dedifferentiation, they need to inform that Dpn is a direct target of Notch in NBs (Doi. 10.1016/j.ydbio.2011.01.019). They do so in the discussion, but mentioning it in the introduction or results will give a broader context to the reader.

We will discuss the study looking at N targets in NB hyperplasia in the discussion of the revised manuscript.

We will mention that Dpn is a target of Notch in the results section.

Another gene that needs a mention is "Brat", which regulates both Dpn and Notch, and causes dedifferentiation and tumors in CNS, I think this gene and its interaction with Dpn and Nerfin and Notch needs to be discussed either in the introduction or discussion.

We will comment on Brat in the discussion.

Are the text and figures clear and accurate?<br /> The main figures are not labeled. Therefore, it was very annoying to deduce the specific figure numbers.<br /> There are 1 or 2 places where figure calling is wrong in the text.<br /> The Image Fig-5I shows cycD and CDK4 at the G2-M transition; while the text says it supports G1/S, which is indeed the case, the figure needs modification.

We thank the reviewers for identifying these mistakes, and will correct them.

Do you have suggestions that would help the authors improve the presentation of their data and conclusions?<br /> The presentation is okay, in my opinion.

Reviewer #1 (Significance):

• Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.

The factors leading to dedifferentiation of the neurons have been identified previously by groups of Chris Doe (mldc, DOI: 10.1242/dev.093781), Andrea brand (10.1016/j.devcel.2014.01.030.) as well as the authors of this paper (10.1101/gad.250282.114, 10.1016/j.celrep.2018.10.038.). However, many questions remained unaddressed regarding such NB generated from neuronal dedifferentiation. For example, whether these cells contribute to native cell diversity of the CNS, undergo timely differentiation or their progeny cells incorporated into appropriate circuits is not well understood. Successful execution of these phenomena is critical for generating functional CNS and such insights are crucial for understanding the origin of tumorigenesis in CNS or employing dedifferentiated NSC for regenerative purposes.

This study is an overexpression-based study, however, some of the results give significant conceptual insights into the tumors arising out of the dedifferentiation of the neurons. It also gives insights into the fact that the dedifferentiated cells need to be carefully examined for the temporal factor profile before they can be employed for regeneration or any therapy targeting them.<br /> However, in my opinion, they need to test this idea at least in one more system of neuronal dedifferentiation, preferably independent of the nerfin-1/Notch/Dpn axis to generalize this claim.

• Place the work in the context of the existing literature (provide references, where appropriate).<br /> Cerdic Maurange's group had looked at the role of temporal factors and identified the early phase of malignant susceptibility in Drosophila in 2016 (doi: 10.7554/eLife.13463). Andrea Brand's group has shown in a 2019 paper that cell cycle progression is essential for temporal transition in NBs (doi: 10.7554/eLife.47887). Both these studies were in the context of Type-I NBs, which express Prospero, which is crucial for the differentiation of the neurons.<br /> Previously the authors have studied type-I NBs and shown by Targeted DamID that Dpn is Nerfin-1 target. They also show that Nerfin-1 mutants show dedifferentiation of neurons. They follow up on this observation in medulla neurons, where they find that Dpn overexpression results in their dedifferentiation into medulla NBs. Medulla NBs differ from Type-I NBs in using a separate set of tTFs. Also, Type-I NB and neurons arising from them use Prospero for terminal differentiation, while medulla neurons do not express Prospero and are therefore considered less stable (DOI: 10.1242/dev.141341).

The importance of the study lies in the results that show that the NB arising out of dedifferentiation of medulla neurons takes up mid-temporal fate. These NBs are stalled in Slp expressing mid-temporal stage unless the cell cycle is promoted by overexpression of cell cycle genes regulating G1/S transition.<br /> Authors also show that overexpression of D promotes the progression of temporal series in these dedifferentiated NBs, which could partly rescue neuronal diversity and result in terminal differentiation. Thus D plays an important role in determining the type of neurons these NBs generated. This suggests that knowing the tTF profile of these types of dedifferentiated NBs is vital if these cells were to be used for regenerative purposes. Authors further claimed that cell cycle regulation and tTFs are critical determinants of the proliferation and termination profile of dedifferentiated NBs.

• State what audience might be interested in and influenced by the reported findings.<br /> The study will be of broader interest to researchers interested in central nervous system patterning, regeneration, and cancer biology.

• Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.<br /> Drosophila, central nervous system patterning and cell fate determination of neural stem cells.

Reviewer #2 (Evidence, reproducibility and clarity):

Stem cells can divide asymmetrically to self-renew the stem cell while generating differentiating sibling cells. To restrict the number and type of differentiating sibling cells, stem cells often undergo terminal differentiation. Terminally differentiated cells can dedifferentiate and revert to a stem cell like fate. However, the underlying molecular mechanisms are incompletely understood in vivo.<br /> Here, Veen et al., use Drosophila neural stem cells (called neuroblasts) to investigate how terminal differentiation is regulated. Neuroblasts faithfully produce the correct number and type of neuronal cells through temporal patterning and regulated terminal differentiation. The authors show that misexpression of the bHLH transcription factor Deadpan (Dpn) induces ectopic neuroblasts, which predominantly express mid-temporal transcription factors at the expense of late-temporal transcription factors. As a consequence, these ectopic neuroblasts also fail to produce Repo positive glial cells and are stalled in their cell cycle progression. The authors provide evidence that promoting cell cycle progression and overexpression of the transcription factor Dichaete (D) is sufficient to restore the temporal transcription factor series, neuronal diversity and timely neuroblast differentiation.

This is an interesting study that will be of interest to the stem cell field. However, I encourage the authors to consider the following critiques:

1. Explain the rationale for the three different neuronal/NB drivers (GMR31HI08-GAL4, eyOK107-GAL4, eyR16F10-GAL4. How are they expressed?

We will include an expression analysis of EyOK107-GAL4 and eyR16F10-GAL4. GMR31HI08-GAL4 expression analysis was previously published (Vissers et al., 2018). We will explain in the text the benefits of each driver.

1. The rationale for the Edu experiment (Figure S1I) is not clear. Why is this a measure for the production of neuronal progeny? For the correct interpretation of these results, the authors should also provide control clones or Edu experiments of regular neuroblasts.

We will repeat this experiment and mark the progeny with the neuronal marker Elav, to demonstrate that they are neurons. Additionally, we will add the control to this figure.

1. How was % of Mira (Figure 1K and below) or the % of tTFs (Figure 2H onward) quantified? For instance, Figure 2C-G often shows clonal signal that is not highlighted with the dashed lines and the corresponding tTF intensity does not match the intensity in the outlined clone (eg. Figure 2D-D'; a large optic lobe clone is negative for Ey. Figure 2E-E'; an unmarked clone is negative for Slp).<br /> Similarly, the Hth signal is very weak to begin with so it is unclear how this was quantified. How was determined what constitutes real signal vs. background noise?<br /> Additional explanations in the methods section is needed to assess the robustness of the data.

We will expand the methods section and mention that we used similar thresholding in antibody staining between control and uas dpn in all instances, so even if the antibody is weaker (eg hth) it is consistently quantified. Additionally, we can increase the intensity of Ey in Figure 2D-2D’, as it is expressed at low levels.

1. This sentence should be rephrased: 'As the tumour cell-of-origin can define the competence of tumour NBs to undergo malignancy (Farnsworth et al., 2015; Narbonne-Reveau et al., 2016), we next tested whether the temporal identity of the dedifferentiated NBs were conferred by the age of the neurons they were derived from.'<br /> The connection between tumorigenicity and temporal identity is not really clear and should be briefly reintroduced for this paragraph.

We will rephrase this sentence and further introduce this concept when talking about tumour cell of origin and competence.

1. Figure 2I-N: The experimental outline in I and J should be grouped with the corresponding images to clarify what is compared. Also, there are no images for the control clones, which make a comparison difficult. The images are also too small. I cannot really see the Hth, or Slp signal in the small clones shown in Figure 2K-L".

We will split figure 2 into two images. The first image including A-H and the control data. And the second including I-Q and the control data. This will increase the size of the images. Additionally, we will group I and J with corresponding data.

1. Figure 3H: It is not clear why there are only a small group of Nbs that are positive for Mira. Please explain.

Most NBs have terminated by this time point, we will explain this within the text.

1. Figure 3K-M: Please explain how the Toy signal was measured and quantified.

We will expand the methods section and explain how Toy quantification is made.

1. The TaDa data set is very interesting but the following might be an overstatement: "We found that Dpn directly binds to slp1 as well as the Sox-family TF dichaete (D) which is expressed in medulla NBs after slp1 (Li et al., 2013) (Figure S6 A-B)."<br /> More direct binding assays might be needed to show that Dpn directly binds to slp1 and D. If this is already shown, clarify the sentence to indicate what is published and what is extracted from the data shown here.<br /> Also, what is the rationale for this statement: "Consistent with the model that D represses Slp-1..."?

The DamID data do actually show that Dpn binds (i.e. there is a statistically significant peak at FDR<0.01) directly at these loci (see the TaDa supp fig A & B). Whether it’s doing anything functional or not, we can’t say, but our data shows that Dpn directly binds to slp1 and D. We will clarify the sentence to indicate this in our revision.

1. This might be an overinterpretation: D overexpression in UAS-Dpn NBs promoted their pre-mature cell cycle exit at 6 hrs APF using eyR16F10-GAL4. The data shows loss of Mira signal, which could occur through different mechanisms.

Our data already shows that these NBs express Tll, the terminal temporal transcription factor (Figure 4F). In addition, we show that there is an increase in Tll+ and Repo+ progeny (Figure 4K, L). Together, this suggests that D overexpression promotes the progression of the temporal series. However, it is possible that Mira+ cells can disappear via cell death. We will assess this possibility by staining for cell death marker Dcp1 at 6hr APF.

Reviewer #2 (Significance):

These appear to be novel and significant findings that will enhance our understanding of the temporal progression and terminal differentiation program of neural stem cells in vivo.<br /> I think the findings will be of interest to cell, developmental cell and stem cell biologists.

My primary expertise is in the cell biology of fly neural stem cells and asymmetric cell division of neuroblasts. Although I am not intimately familiar with the differentiation and differentiation literature, I consider the findings reported here relevant and impactful.

Reviewer #3 (Evidence, reproducibility and clarity):

The discoveries that the author describe in this manuscript are very specific to dedifferentiated neuroblasts created by UAS-dpn transgene overexpression. Dpn is endogenously expressed in optic lobe neuroblast throughout larval stage, which makes understanding how Dpn regulates gene expression based on the authors results (suppression of cell-cycle genes, and promotion of a specific temporal state) confusing.

Our data relate specifically to gene regulation by Dpn in a dedifferentiated context, and do not seek to understand Dpn regulation in wt neuroblasts. The reviewer is assuming our scope is greater here: we’re not trying to claim that we know what Dpn is doing in wt NBs, and it’s not surprising that ectopic effects in neurons may be different to wt NBs.

To assess whether the mechanisms described apply to more than Dpn overexpression, we will also assess whether the temporal series progression is affected in Lola RNAi and Nerfin-1 mutant.

Therefore, this manuscript does not advance our understanding of regulation of temporal identity and cell cycle progression in optic lobe neuroblasts during normal neurogenesis.<br /> The author's state:<br /> "However, beyond the fact that misexpression of these factors and pathways caused the formation of ectopic NBs, whether these dedifferentiated NBs faithfully produce the correct number and types of neurons or glial cells, or undergo timely terminal differentiation, has not been assessed. These characteristics are key determinants of overall CNS size and function, thus are important parameters when considering whether dedifferentiation leads to tumourigenesis or can be appropriately utilized for regenerative purposes."<br /> at the end of introduction. If this is a true primary goal of this study, the authors should describe it in abstract. Otherwise, readers will lose enthusiasm to read this manuscript in abstract and no longer read the following sections.

We will add this to the abstract.

Results<br /> 1. The authors should describe the expression pattern of all three of the Gal4 drivers used. While there are dotted outlines in the supplemental figure, there should be a description in the main text for the expression pattern of these lines which described with temporal state of NBs these lines are expressed in, and whether they are also expressed in the neurons or not.

We will include expression analysis of all three drivers in a supplementary figure and explain in the text the benefit of each driver.

1. The authors claim that overexpression of Dpn in the medulla region causes "dedifferentiation." The data provided however is not sufficient to conclude that dedifferentiation is occurring. The GAL4s used all drive in the NBs, and so it is unclear if the ectopic NBs ever became mature neurons. In addition, the lack of ectopic NBs in the clonal analysis 16hrs AHS does not prove that ectopic NBs at 24hrs AHS must have come from "mature neurons." To demonstrate dedifferentiation, the authors should use a driver system that is specific to mature neurons, and then overexpress dpn and look for mira+ cells. Currently, the authors data does not prove that mature neurons dedifferentiatiate into ectopic NBs upon Dpn OE.

We have conducted lineage tracing (G-Trace) analysis of the medulla neuron driver GMR31H08-GAL4 which we utilise in our study, this driver is predominantly expressed within the medulla neurons (real time) except for a few GMCs present in the lineage. Therefore, the Mira positive cells induced via Dpn overexpression are most likely from dedifferentiation (We will include this data in a supplemental figure in our revised manuscript).

To further support this, we will use GMR31H08-GAL4 with a Gal80ts, to restrict the timing to dedifferentiation induction to 3rd instar, so that the driver is restricted to neurons. Similar strategy to induce dedifferentiation was utilised in DOI: 10.1242/dev.141341 and DOI: 10.1016/j.devcel.2014.01.030.

1. What is a conclusion of fig 2C-H?

Fig 2C-H assess the expression of tTFs in UAS-dpn induced ectopic NBs. We will make these conclusions clearer in the text.

1. "As the tumor cell-of-origin can define the competence of tumor NBs to undergo malignancy identity of the dedifferentiated NBs were conferred by the age of the neurons they were derived from". This sentence is confusing. What are the authors investigating in the following experiment? Do they want to see ectopic NBs keep their early identity like Chinmo in ventral cord tumor NB? Or tll-positive NB's progenies can dedifferentiate to ectopic NB, but this ectopic neuroblast is not able to keep proliferation in pupal stage? It is hard to understand the connection of this sentence and the following experiment.

We will rephrase this sentence and further introduce this concept when talking about tumour cell of origin and competence. Additionally, we will make the connection to the experiments which follow it clearer.

1. The DamID experiment described used wor-gal4 as a driver, which means the Dpn binding profile generated is coming from not only optic lobe NBs, but central brain NBs and VNC NBs as well. In Magadi et al. (2020), the authors profiled Dpn binding in CNS hyperplasia, and found that dpn strongly bound Nerfin-1 and gcm. However, it does not bind cell cycle genes in this context. How do the authors know that the region that they claim are bound by dpn are bound in medulla NBs? The authors should also include tracks to show dpn binding at Nerfin-1, as well as the other tTFs (hth, ey, tll, and gcm). Providing this data will help to understand if Dpn binding is specific to the mid-temporal genes, as Dpn expression is known to be expressed in all medulla NBs regardless of temporal state.

We agree with the reviewer that the profile is not specific to medulla NBs. To assess Dpn binding profiles specifically in the medulla NBs, we will use the recently-published NanoDam technique (https://doi.org/10.1016/j.devcel.2022.04.008) for profiling GFP-fusion proteins, with a medulla specific driver (eyR16F10-GAL4) and Dpn-GFP (recombineered locus under endogenous control). This should inform us whether the target genes we have identified are relevant in the medulla.

We will include the tracks of the other transcription factors.

1. Currently, the DamID data does not help to interpret the Dpn overexpression phenotype at all. Inside of flip-out clone, some cells show Slp-1 expression while others showed D expression. The authors explain that Slp-1 and D suppress their expression to each other. But the DamID data indicate that both Slp-1 and D are Dpn target genes. If this is true, why did they observe the mosaic expression pattern inside of the same clone.

We observed that high levels of Slp-1 is correlated with low levels of D. This suggest to us that the initial stochastic differences accounts for where Slp-1 is high is where D is low, and vice versa.

1. The authors hypothesized if Dpn activated Slp-1directly. Does this mean that Dpn directly activate transcription of Slp-1? It is well known that Dpn is transcriptional repressor. Hes family proteins form a homodimer or heterodimer with another Hes protein and interacts Gro, which recruits a Histon deacetylase protein. The author's claim does not fit to the model what we currently believe. In addition, the authors claimed that Dpn inhibits cell cycle gene transcription directly. This is inconsistent to their claim that Dpn directly activate Slp-1 expression. If the authors want to claim that Dpn has two different functions in this context, the authors must demonstrate it by experimental results.

We will discuss these models in the Discussion, and make our claims more conservative, as we do not have direct experimental evidence to prove or disprove the model that Dpn is acting as an activator in this context.

1. Related to the above question, I wondered if the authors guess Dpn activate or repress D transcription by binding to D promoter region because they claimed that Dpn activate Slp-1, while suppress cell cycle genes.

We will make our claims more conservative, and discuss this point further in the Discussion.

1. I am confused to the claim that Dpn suppress cell cycle genes expression. Dpn overexpression induces dedifferentiation of neuron into NB and re-entry into the cell cycle. If Dpn suppress cell cycle genes how can the dedifferentiated cell re-enter into the cell cycle?

The data points towards that Dpn overexpression has two separate roles in regulating the cell cycle. Ofcourse dedifferentiation requires a commitment of neurons into the cell cycle (this we think is still happening), however, we think once these cells have turned on NB markers, they have limited ability to progress through the cell cycle. We will discuss this point in the Discussion.

1. Figure 6 looked redundant because we know Dpn is a direct target of Notch. It is obvious that an upstream factor overexpression can induce the identical phenotype to the phenotype induced by overexpression of a downstream factor.

A direct target does not necessarily infer the same phenotype. To assess whether the mechanisms apply to other dedifferentiation models, we will add Lola-RNAi and Nerfin-1 data to our revised manuscript.

Minor comments:<br /> 1. Typo in main text: "GMR31HI08-GAL4" should be "GMR31H08-GAL4"<br /> 2. In figure 1E-H the dotted line regions indicated the clones are not shown in the merge image. Please include<br /> 3. Typo in discussion paragraph 2: "temporal series was no sufficient to rescue cycle cycle progression"

We will correct these typos.

Reviewer #3 (Significance):

Insights into the developmental capacity of dedifferentiated stem cells will likely lead to novel strategy to replenish cells lost due to aging, injury and diseases in regenerative medicine.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The discoveries that the author describe in this manuscript are very specific to dedifferentiated neuroblasts created by UAS-dpn transgene overexpression. Dpn is endogenously expressed in optic lobe neuroblast throughout larval stage, which makes understanding how Dpn regulates gene expression based on the authors results (suppression of cell-cycle genes, and promotion of a specific temporal state) confusing. Therefore, this manuscript does not advance our understanding of regulation of temporal identity and cell cycle progression in optic lobe neuroblasts during normal neurogenesis.

The author's state:

"However, beyond the fact that misexpression of these factors and pathways caused the formation of ectopic NBs, whether these dedifferentiated NBs faithfully produce the correct number and types of neurons or glial cells, or undergo timely terminal differentiation, has not been assessed. These characteristics are key determinants of overall CNS size and function, thus are important parameters when considering whether dedifferentiation leads to tumourigenesis or can be appropriately utilized for regenerative purposes."<br /> at the end of introduction. If this is a true primary goal of this study, the authors should describe it in abstract. Otherwise, readers will lose enthusiasm to read this manuscript in abstract and no longer read the following sections.

Results

1. The authors should describe the expression pattern of all three of the Gal4 drivers used. While there are dotted outlines in the supplemental figure, there should be a description in the main text for the expression pattern of these lines which described with temporal state of NBs these lines are expressed in, and whether they are also expressed in the neurons or not.
2. The authors claim that overexpression of Dpn in the medulla region causes "dedifferentiation." The data provided however is not sufficient to conclude that dedifferentiation is occurring. The GAL4s used all drive in the NBs, and so it is unclear if the ectopic NBs ever became mature neurons. In addition, the lack of ectopic NBs in the clonal analysis 16hrs AHS does not prove that ectopic NBs at 24hrs AHS must have come from "mature neurons." To demonstrate dedifferentiation, the authors should use a driver system that is specific to mature neurons, and then overexpress dpn and look for mira+ cells. Currently, the authors data does not prove that mature neurons dedifferentiatiate into ectopic NBs upon Dpn OE.
3. What is a conclusion of fig 2C-H?
4. "As the tumor cell-of-origin can define the competence of tumor NBs to undergo malignancy identity of the dedifferentiated NBs were conferred by the age of the neurons they were derived from". This sentence is confusing. What are the authors investigating in the following experiment? Do they want to see ectopic NBs keep their early identity like Chinmo in ventral cord tumor NB? Or tll-positive NB's progenies can dedifferentiate to ectopic NB, but this ectopic neuroblast is not able to keep proliferation in pupal stage? It is hard to understand the connection of this sentence and the following experiment.
5. The DamID experiment described used wor-gal4 as a driver, which means the Dpn binding profile generated is coming from not only optic lobe NBs, but central brain NBs and VNC NBs as well. In Magadi et al. (2020), the authors profiled Dpn binding in CNS hyperplasia, and found that dpn strongly bound Nerfin-1 and gcm. However, it does not bind cell cycle genes in this context. How do the authors know that the region that they claim are bound by dpn are bound in medulla NBs? The authors should also include tracks to show dpn binding at Nerfin-1, as well as the other tTFs (hth, ey, tll, and gcm). Providing this data will help to understand if Dpn binding is specific to the mid-temporal genes, as Dpn expression is known to be expressed in all medulla NBs regardless of temporal state.
6. Currently, the DamID data does not help to interpret the Dpn overexpression phenotype at all. Inside of flip-out clone, some cells show Slp-1 expression while others showed D expression. The authors explain that Slp-1 and D suppress their expression to each other. But the DamID data indicate that both Slp-1 and D are Dpn target genes. If this is true, why did they observe the mosaic expression pattern inside of the same clone.
7. The authors hypothesized if Dpn activated Slp-1directly. Does this mean that Dpn directly activate transcription of Slp-1? It is well known that Dpn is transcriptional repressor. Hes family proteins form a homodimer or heterodimer with another Hes protein and interacts Gro, which recruits a Histon deacetylase protein. The author's claim does not fit to the model what we currently believe. In addition, the authors claimed that Dpn inhibits cell cycle gene transcription directly. This is inconsistent to their claim that Dpn directly activate Slp-1 expression. If the authors want to claim that Dpn has two different functions in this context, the authors must demonstrate it by experimental results.
8. Related to the above question, I wondered if the authors guess Dpn activate or repress D transcription by binding to D promoter region because they claimed that Dpn activate Slp-1, while suppress cell cycle genes.
9. I am confused to the claim that Dpn suppress cell cycle genes expression. Dpn overexpression induces dedifferentiation of neuron into NB and re-entry into the cell cycle. If Dpn suppress cell cycle genes how can the dedifferentiated cell re-enter into the cell cycle?
10. Figure 6 looked redundant because we know Dpn is a direct target of Notch. It is obvious that an upstream factor overexpression can induce the identical phenotype to the phenotype induced by overexpression of a downstream factor.

Minor comments:

1. Typo in main text: "GMR31HI08-GAL4" should be "GMR31H08-GAL4"
2. In figure 1E-H the dotted line regions indicated the clones are not shown in the merge image. Please include
3. Typo in discussion paragraph 2: "temporal series was no sufficient to rescue cycle cycle progression"

Significance

Insights into the developmental capacity of dedifferentiated stem cells will likely lead to novel strategy to replenish cells lost due to aging, injury and diseases in regenerative medicine.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Stem cells can divide asymmetrically to self-renew the stem cell while generating differentiating sibling cells. To restrict the number and type of differentiating sibling cells, stem cells often undergo terminal differentiation. Terminally differentiated cells can dedifferentiate and revert to a stem cell like fate. However, the underlying molecular mechanisms are incompletely understood in vivo.<br /> Here, Veen et al., use Drosophila neural stem cells (called neuroblasts) to investigate how terminal differentiation is regulated. Neuroblasts faithfully produce the correct number and type of neuronal cells through temporal patterning and regulated terminal differentiation. The authors show that misexpression of the bHLH transcription factor Deadpan (Dpn) induces ectopic neuroblasts, which predominantly express mid-temporal transcription factors at the expense of late-temporal transcription factors. As a consequence, these ectopic neuroblasts also fail to produce Repo positive glial cells and are stalled in their cell cycle progression. The authors provide evidence that promoting cell cycle progression and overexpression of the transcription factor Dichaete (D) is sufficient to restore the temporal transcription factor series, neuronal diversity and timely neuroblast differentiation.

This is an interesting study that will be of interest to the stem cell field. However, I encourage the authors to consider the following critiques:

1. Explain the rationale for the three different neuronal/NB drivers (GMR31HI08-GAL4, eyOK107-GAL4, eyR16F10-GAL4. How are they expressed?
2. The rationale for the Edu experiment (Figure S1I) is not clear. Why is this a measure for the production of neuronal progeny? For the correct interpretation of these results, the authors should also provide control clones or Edu experiments of regular neuroblasts.
3. How was % of Mira (Figure 1K and below) or the % of tTFs (Figure 2H onward) quantified? For instance, Figure 2C-G often shows clonal signal that is not highlighted with the dashed lines and the corresponding tTF intensity does not match the intensity in the outlined clone (eg. Figure 2D-D'; a large optic lobe clone is negative for Ey. Figure 2E-E'; an unmarked clone is negative for Slp).<br /> Similarly, the Hth signal is very weak to begin with so it is unclear how this was quantified. How was determined what constitutes real signal vs. background noise?<br /> Additional explanations in the methods section is needed to assess the robustness of the data.
4. This sentence should be rephrased: 'As the tumour cell-of-origin can define the competence of tumour NBs to undergo malignancy (Farnsworth et al., 2015; Narbonne-Reveau et al., 2016), we next tested whether the temporal identity of the dedifferentiated NBs were conferred by the age of the neurons they were derived from.'<br /> The connection between tumorigenicity and temporal identity is not really clear and should be briefly reintroduced for this paragraph.
5. Figure 2I-N: The experimental outline in I and J should be grouped with the corresponding images to clarify what is compared. Also, there are no images for the control clones, which make a comparison difficult. The images are also too small. I cannot really see the Hth, or Slp signal in the small clones shown in Figure 2K-L".
6. Figure 3H: It is not clear why there are only a small group of Nbs that are positive for Mira. Please explain.
7. Figure 3K-M: Please explain how the Toy signal was measured and quantified.
8. The TaDa data set is very interesting but the following might be an overstatement: "We found that Dpn directly binds to slp1 as well as the Sox-family TF dichaete (D) which is expressed in medulla NBs after slp1 (Li et al., 2013) (Figure S6 A-B)."<br /> More direct binding assays might be needed to show that Dpn directly binds to slp1 and D. If this is already shown, clarify the sentence to indicate what is published and what is extracted from the data shown here.<br /> Also, what is the rationale for this statement: "Consistent with the model that D represses Slp-1..."?
9. This might be an overinterpretation: D overexpression in UAS-Dpn NBs promoted their pre-mature cell cycle exit at 6 hrs APF using eyR16F10-GAL4. The data shows loss of Mira signal, which could occur through different mechanisms.

Significance

These appear to be novel and significant findings that will enhance our understanding of the temporal progression and terminal differentiation program of neural stem cells in vivo.<br /> I think the findings will be of interest to cell, developmental cell and stem cell biologists.

My primary expertise is in the cell biology of fly neural stem cells and asymmetric cell division of neuroblasts. Although I am not intimately familiar with the differentiation and differentiation literature, I consider the findings reported here relevant and impactful.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary

Authors show that overexpression of bHLH transcription factor Dpn in the medullary neurons of the Drosophila optic lobe results in the dedifferentiation of these neurons back into the NBs. These dedifferentiated NBs acquire and maintain mid-temporal identity, express Ey and Slp, and show delayed onset of tTF Tailless (Tll), leading to an excess of neurons of mid-temporal fate at the expense of late temporal fate neurons and glial cells. The dedifferentiated NBs are stalled in the cell cycle and fail to undergo terminal differentiation. Over expression of tTF Dicheate (D) or promoting G1/S transition pushed these NBs to late stages of the temporal series, partly rescuing the neuronal diversity and causing their terminal differentiation. They also show that the dedifferentiation of NBs by Notch hyper-activation also exhibited stalled temporal progression, which is restored by D overexpression.<br /> Authors suggest that cell cycle regulation and tTF are primary to the proliferation and termination profile of dedifferentiated NBs.<br /> Using these conclusions, the authors emphasize the need to recreate the right temporal profile and ensure appropriate cell cycle progression to use dedifferentiated NSC for regenerative purposes or prevent tumorigenesis originating from differentiated cell types.

Major comments:

• Are the key conclusions convincing?

Most conclusions are convincing; however, some issues are pointed out below.<br /> - Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

The authors have overexpressed Dpn and shown that medulla neurons dedifferentiate to NBs, similar to the loss of function phenotype seen for the Nerfin-1 of which Dpn is a target. They also show that temporal series progression defect is also seen in the case of dedifferentiated NB generated by Notch over-activation.<br /> Using these two examples, the authors suggest that for dedifferentiated NSC, which are to be used for the regenerative purpose, one needs to recreate the right temporal profile and ensure cell cycle progression occurs appropriately. Authors also claim that to prevent tumorigenesis originating from differentiated cell types, one needs to recreate the right temporal profile and ensure cell cycle progression occurs appropriately.

While I agree with this, I think this is an overreaching conclusion based on just these two examples. If they could show the same for one more method of dedifferentiation (For, e.g. Lola) happening in medulla neurons which happens by a mechanism independent of Nerfin-1, Dpn, Notch axis, the argument will become more convincing and broad.<br /> Also when authors mention N mediated dedifferentiation, they need to inform that Dpn is a direct target of Notch in NBs (Doi. 10.1016/j.ydbio.2011.01.019), they do so in the discussion, but mentioning it here gives a broader context to the reader.

Another important point that needs a mentioned here is that conclusions are based on dedifferentiation happening in the medulla neurons, which are considered less stable since they lack Prospero. Therefore whether this conclusion can be generalized for all the tumors arising from dedifferentiation in the CNS (eg, those arising from NICD activation in the central brain or thoracic region of the VNC) is another concern. Maybe authors can consider making a more conservative claim.<br /> Generalizing this conclusion to Prospero expressing NBs lies outside the scope of the current study and cannot be addressed here because central brain Type-I NBs use a different set of tTFs.<br /> - Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

Experiments with Lola knockdown/mutants in medulla neurons can be done quickly, in my opinion, and will substantiate this claim.<br /> Another obvious question that comes to mind is if medulla neurons dedifferentiate on overexpression of Dpn, does the same happen in nerfin-1 mutant clones as well? And if yes, why has the author not done similar experiments for nerfin-1 mutants.<br /> Please show Ey staining in Fig-2 if possible, it will also help to add a line on why Slp was used as marker for mid tTFs instead of Ey.<br /> In Model shown in last figure Dpn is shown to repress D and activate Slp. Can authors show that Dpn overexpression represses D and activate Slp either by antibody staining or by RT PCR.<br /> - Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

Experiments with Lola and nerfin-1 mutants can be done in a few months. I cannot comment on the cost involved.<br /> - Are the data and the methods presented in such a way that they can be reproduced?

Yes<br /> - Are the experiments adequately replicated and statistical analysis adequate?

Replication and statistical analysis are fine. The activated Notch experiments show only three data points in all the experiments. It will be good to increase this number.

Minor comments:

• Specific experimental issues that are easily addressable.

There is a problem with Fig-5F (both 5E and 5F have % EdU in clone/ % Mira in the clone as y-axis), I do not understand how the Fig-5F let them conclude that D overexpression increases the rate of neuronal production.

In one place, the authors conclude, "Together, this data suggests that it is likely that cell cycle progression lies upstream of the temporal series, to promote the generation of neurons". Authors should consider adding "medulla NBs" at the end of the sentence since cell cycle progression being upstream of temporal series is already known in Type-I NBs, as pointed out by authors as well (Ameele and Brand 2019).

In the discussion authors says that "Our data support the possible links between cell cycle progression and the expression of temporal regulators controlling NB proliferation and cellular diversity". This is new information, as the 2019 study did not show how cell diversity changes with a changed tTF profile. I think the authors should elaborate on this point to highlight how this is different from what is already known from the 2019 study (done in the context of Type-I NBs).<br /> Maybe they need to highlight that the cell cycle directs/regulates the progression of temporal series compared to the earlier observation where temporal series was shown to be downstream of the cell cycle.

In fig-3J in clones even after 24 AHS, Dpn continues to be overexpressed but these cells undergo terminal differentiation, can authors comment why is it so?<br /> In one place authors say, "To better assess the cumulative effect of the neurons made throughout development, EyOK107-GAL4 was used to drive the expression of Dpn" maybe some background on why use this specific GAL4.<br /> Also a line about why GMR31HI08-GAL4 eyOK107-GAL4 and and eyR16F10-GAL4 were used.<br /> - Are prior studies referenced appropriately ?

Yes, but in a few places, some references can be added.<br /> An important point that needs to be mentioned for the context is the medulla neurons do not use Prospero for terminal differentiation and are thus considered less stable (DOI: 10.1242/dev.141341).<br /> In discussion, the authors say that "It would be interesting to explore whether N similarly acts on these target genes to specify cell fate and proliferation profiles of dedifferentiated NBs." There is a study looking at Notch targets in NB hyperplasia (DOI: 10.1242/dev.126326); whether that study shows if any of the cell cycle genes are downstream of activated Notch, needs a mention here.<br /> Also, when authors mention N mediated dedifferentiation, they need to inform that Dpn is a direct target of Notch in NBs (Doi. 10.1016/j.ydbio.2011.01.019). They do so in the discussion, but mentioning it in the introduction or results will give a broader context to the reader.<br /> Another gene that needs a mention is "Brat", which regulates both Dpn and Notch, and causes dedifferentiation and tumors in CNS, I think this gene and its interaction with Dpn and Nerfin and Notch needs to be discussed either in the introduction or discussion.<br /> - Are the text and figures clear and accurate?

The main figures are not labeled. Therefore, it was very annoying to deduce the specific figure numbers.<br /> There are 1 or 2 places where figure calling is wrong in the text.<br /> The Image Fig-5I shows cycD and CDK4 at the G2-M transition; while the text says it supports G1/S, which is indeed the case, the figure needs modification.<br /> - Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

The presentation is okay, in my opinion.

Significance

• Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.

The factors leading to dedifferentiation of the neurons have been identified previously by groups of Chris Doe (mldc, DOI: 10.1242/dev.093781), Andrea brand (10.1016/j.devcel.2014.01.030.) as well as the authors of this paper (10.1101/gad.250282.114, 10.1016/j.celrep.2018.10.038.). However, many questions remained unaddressed regarding such NB generated from neuronal dedifferentiation. For example, whether these cells contribute to native cell diversity of the CNS, undergo timely differentiation or their progeny cells incorporated into appropriate circuits is not well understood. Successful execution of these phenomena is critical for generating functional CNS and such insights are crucial for understanding the origin of tumorigenesis in CNS or employing dedifferentiated NSC for regenerative purposes.

This study is an overexpression-based study, however, some of the results give significant conceptual insights into the tumors arising out of the dedifferentiation of the neurons. It also gives insights into the fact that the dedifferentiated cells need to be carefully examined for the temporal factor profile before they can be employed for regeneration or any therapy targeting them.<br /> However, in my opinion, they need to test this idea at least in one more system of neuronal dedifferentiation, preferably independent of the nerfin-1/Notch/Dpn axis to generalize this claim.<br /> - Place the work in the context of the existing literature (provide references, where appropriate).

Cerdic Maurange's group had looked at the role of temporal factors and identified the early phase of malignant susceptibility in Drosophila in 2016 (doi: 10.7554/eLife.13463). Andrea Brand's group has shown in a 2019 paper that cell cycle progression is essential for temporal transition in NBs (doi: 10.7554/eLife.47887). Both these studies were in the context of Type-I NBs, which express Prospero, which is crucial for the differentiation of the neurons.<br /> Previously the authors have studied type-I NBs and shown by Targeted DamID that Dpn is Nerfin-1 target. They also show that Nerfin-1 mutants show dedifferentiation of neurons. They follow up on this observation in medulla neurons, where they find that Dpn overexpression results in their dedifferentiation into medulla NBs. Medulla NBs differ from Type-I NBs in using a separate set of tTFs. Also, Type-I NB and neurons arising from them use Prospero for terminal differentiation, while medulla neurons do not express Prospero and are therefore considered less stable (DOI: 10.1242/dev.141341).

The importance of the study lies in the results that show that the NB arising out of dedifferentiation of medulla neurons takes up mid-temporal fate. These NBs are stalled in Slp expressing mid-temporal stage unless the cell cycle is promoted by overexpression of cell cycle genes regulating G1/S transition.<br /> Authors also show that overexpression of D promotes the progression of temporal series in these dedifferentiated NBs, which could partly rescue neuronal diversity and result in terminal differentiation. Thus D plays an important role in determining the type of neurons these NBs generated. This suggests that knowing the tTF profile of these types of dedifferentiated NBs is vital if these cells were to be used for regenerative purposes. Authors further claimed that cell cycle regulation and tTFs are critical determinants of the proliferation and termination profile of dedifferentiated NBs.<br /> - State what audience might be interested in and influenced by the reported findings.

The study will be of broader interest to researchers interested in central nervous system patterning, regeneration, and cancer biology.<br /> - Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

Drosophila, central nervous system patterning and cell fate determination of neural stem cells.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We propose three revisions, that have not yet been included in the current manuscript:

1. All three reviewers comment on the data in figure 7, in which the application of the sensor is shown. We agree that the number of cells is low, and we plan to repeat this experiment to increase the number of cells, and better demonstrate the usefulness of the new probe. We note that the improved Cdc42 sensor is used in a recent preprint (see figures 7 and 9 of: https://www.biorxiv.org/content/10.1101/2022.06.22.497207v2.full), clearly showing the potential of the probe for detection of Cdc42 and increasing our confidence that we can generate higher quality data.
2. The ratio of expression of the different components was not quantified. We have these data and we will (re)analyze it and present the results (related to Reviewer #3, point2).
3. We will reanalyze the images to ensure that representative images are depicted in the manuscript (related to Reviewer #2, point 3).

   Reviewer #1

1) It is not clear why RhoA data were included in this manuscript (Fig. 1), since they seem irrelevant to the primary topic addressed.

We have cell-based data from our previous (published) work that we can use to check whether these results align with the mass-spec data. To make this point clearer we add “We looked first into GBDs for Rho, to compare the results of the mass spectrometry screen with the results of our cell-based assays”.

2) It is not clear what cell type was used when screening for p67phox. The expression of this component of the NADPH oxidase is restricted to a few specific cell types.

That’s a relevant point and therefore observation that p67phox is not detected is perhaps not surprising. We removed this statement.

3) There is precious little quantitation of the colocalization or translocation of the probes throughout the manuscript. It is difficult to assess the validity of the conclusions in the absence of analysis of the statistical significance of the colocalization.

In figure 2, which is an initial screen, there is only a qualitative assessment. However, for the promising candidates, there is a quantitative assessment in Figures 3B and 4 B as to which extent the candidates colocalize with the nuclear localized target. From the rank order and individual datapoints the best performing binder can be inferred.

4) It is not clear why translocation to mitochondria was used in some experiments and translocation to the nucleus in others.

To clarify, we have added text: ”We have previously used nuclear localized, constitutive active Rho GTPases, but these are not accessible for larger proteins that cannot enter the nucleus”

5) In the S1P experiments, it is difficult to ascertain whether increased fluorescence resulted from membrane folding/ruffling or is actually a consequence of localized activation of receptors. Why does the fluorescence decrease progressively over 1500 seconds? Isn't maximal receptor activation accomplished much sooner?

This experiment suffered from bleaching. We will redo the experiment to get higher number of cells and to improve the data.

Reviewer #2

Major comments

1. Statistical tests are missing in most of the figures. If the principal purpose of this work is to compare the performance of candidate peptides, the quantitative comparison is essential. If the purpose is just to report another relocation probe, then, more application data may be necessary.

We will improve the quality of the application data. As for statistics, we have added the effect size to figures 5C-F and figure 6A. To explain this (not so common) statistic we add to the materials and methods: “The effect size that quantifies the difference and its distribution was calculated with the web tool ‘PlotsofDifferences’”.

1. The criteria for selecting the best peptide should be clearly described. Is it just by inspection or based on any quantitative data? We know that quantification of colocalization is a difficult task. Therefore, it depends on the aim of this work whether the authors are asked to show quantitative data or not. If a strict comparison of peptides is aimed at, the expression level of each target peptide should be at a comparable level. It will be also required whether the design of each probe guarantees the proper folding to bind to GTPases.

There are two stages for the selection. First, we did a qualitative analysis of colocalization (shown in figure 2). Based on the results (“Candidates colocalizing with the mitochondrial tagged Rho GTPase were further tested for their potential as localization-based sensors”), we generated smaller biosensor candidates of which binding to a nuclear target was quantitatively analyze (figures 3B and 4B). As the expression level is an important factor, we ascertained potential candidates were expressed at roughly the same level in the nuclear accumulation assay.

1. About the images of cells: When a fluorescent image is presented, we assume it represents all other cells. Please check all images whether they are truly representing the data. For example, in Fig. S3 the nuclei of ABI1-expressing cells look weird, and the nucleus of CYRI-A is very large. If this is true, the reason why ABI1 and CYRI-A should be excluded from the candidate is not the relocation efficiency but the undesired effect on cell physiology. For the screening of the peptides, this information is also very important. With that, this paper becomes more valuable for scientists.

We agree that this is an important point. We will reanalyze the data as indicated in the ‘planned revisions’.

1. Please examine the order of panels. For example, the result of mScarlet is on the top in Fig3, but at the bottom in Fig4. Such inconsistency would disturb readers.

We thank the reviewer for this suggestion and we changed figure 4.

1. The label should be consistent throughout the paper. For example, in Fig. 5A, Lck-FRB-mTurquoise2 is labeled as Lck-FRB (without the fluorescent protein's name). WASp(CRIB)-mScarlet-I-WASp(CRIB) is labeled as WASp(CRIB)-mScar-WASp(CRIB) (with fluorescent protein's name). Moreover, the same peptide is labeled as mSca-1xWASp(CRIB) in Panel B. Such inconsistency is confusing.

We agree, we have updated figure 5A by adding the abbreviations of the fluorescent proteins. Please note that WASp(CRIB)-mSca-WASp(CRIB), mSca-1xWASp(CRIB) and mSca-2xWASp(CRIB) are three different constructs. In the first one the CRIB domains are sandwiching the fluorescent protein and in the third one they are in tandem downstream of the fluorescent protein.

1. Quantitative insight would improve this work. For example, in Fig. 7, the reason why the authors believe that the probe worked is the accumulation of probe at the tip of lamellipodia and the decrease in cytoplasmic intensity. This reviewer does not think the accumulation of the probe in the small area of the lamellipodia explains the massive decrease of cytoplasmic signals. Probably, a substantial amount of the probe is relocated to the plasma membrane, not limited to the lamellipodia.

Minor comments

We propose to repeat the experiment shown in figure 7 and to improve the quality of the data.

1. Introduction, "FRET signal is typically measured with a wide field microscope.": This reviewer does not agree with this statement. Confocal and two-photon microscopes have also been used widely.

Fair point. We changed the text to “when the FRET signal is measured with a wide field microscope”

Introduction, "G-protein activating proteins (GAP)": It should read as "GTPase-activating proteins (GAPs)"

Thanks, corrected.

TRIF should read as TIRF.

All instances have been corrected.

Fig.1: To the best of this reviewer's knowledge, PKN1 was first used as the RhoA target peptide by Yoshizaki et al in 2003. J Cell Biol 162, 223-232. They also examined mDia, Rhoteki, and Rhophilin as the target peptides. Pak1 was first used as the Rac1 probe by Kraynov et al. Science 290, 333-337, 2000. Use of Pak1 as the Cdc42 probe was reported by Itoh et al. Mol Cell Biol 22, 6582-659, 2002. This reviewer believes that the priority of the first report should be respected.

We changed part of the introduction to:

High scoring proteins for interacting with constitutively active RhoA(Q63L) included ANLN part of the AniRBD Rho location sensor (Piekny and Glotzer, 2000), PKN1 part of aRho FRET sensor (Yoshizaki et al., 2003) and RTKN part of the rGBD Rho location sensor (Benink and Bement, 2005; Mahlandt et al., 2021) (Fig. 1A,B). This suggested that proteins with a high score in the mass spectrometry screen are potentially suitable as Rho GTPase activity biosensor. Indeed, the GBDs used for Cdc42 location sensors from, PAK1 used in the PBD location sensor (Itoh et al., 2002; Petrie et al., 2012) and N-WASP similar to WASp used in the wGBD location sensor (Benink and Bement, 2005) showed a high score in the screen (Fig. 1A,B).

Discussion:

Another challenge is the Rho GTPase specificity of the relocation-based sensor. For example, Pak1(CRIB) was first used in a Rac1 FRET sensor (Kraynov et al., 2000)____. ThenPak1(CRIB) has been utilized in Cdc42 FRET sensors and in an intensiometric Cdc42 sensor (Hanna et al., 2014; Itoh et al., 2002; Kim et al., 2019). However, Pak1(CRIB), also named PBD sensor, has then been reintroduced by Weiner and colleagues as a Rac1 specific location-based sensor and is often used in neutrophil HL60 cells (Brunetti et al., 2022; Graziano et al., 2019; Le et al., 2021; Weiner et al., 2007).

We also updated the tables in Figure 1.

Fig. 1: Why do the authors omit other promising candidates shown in panel 1B? Please describe the reason for the choice.

We took into account the availability of plasmid DNA, as also explained in the manuscript: “candidate GBDs were selected from top 30 scores of the mass spectrometry screen, that were specific for one Rho GTPase and their DNA was available on addgene”

Fig. 1B: Be consistent to use either "Name" or "Uni Prot name" in Panel A.

We updated figure 1.

Fig. 2: Please include information on TOMM20. The readers may not read the paper by Gillingham et al.

We added an explanation: “To this end, a fusion with TOMM20 was used for mitochondrial localization.”

Fig3 and 4: The authors should show the images of control H2A.

We provide the data for control H2A in figures 3B and 4B.

In Fig3B and 4B, "Cdc42/Rac1 affinity" would be misleading, because the control dots represent their authentic localization rather than "Cdc42/Rac1 affinity".

We agree, we have updated figure 3B and 4B.

Fig. 4: More explanation of this figure is required.

We added text: “Hence, the sensor candidate can freely partition between Rac and Cdc42 binding.”

Fig. 5: More explanation about the FKBP-FRB system will be helpful.

We changed the text to: “The system used rapamycin induced heterodimerization of the two domains FRB and FKBP to recruit the DHPH domain of the Cdc42 specific GEF ITSN1 to the plasma membrane, where it induces activity of the endogenous Cdc42”

Fig. 6: It is rather surprising to see that control-mScarlet also responds to Rac1 activation. What is the explanation for this observation?

We agree and have no explanation.

Fig. 7: A single champion data may not be convincing to prove the usefulness of this probe.

We agree and propose to repeat the experiment.

Reviewer #3

1) The discussion comparing different types of biosensors missed important points. Although the advantages of localization biosensors listed by the authors are correct, they gave the impression that these should simply be an improved replacement for FRET biosensors. There are times when FRET biosensors provide clear advantages. Unlike other proteins, Rho GTPases are well suited for localization sensors because the activated conformation, and only the activated conformation, localizes to the membrane. For diffuse or 3D localization FRET can provide better quantification. Subtle features such as gradients are not easily quantified over a background of unattached domain. The authors state that localization biosensors have enhanced spatial resolution, but this is not explained.

We agree that our introduction is biased towards a preference for relocation based biosensors. However, having used both approaches, we see that both strategies have pro’s and cons. Therefore, we removed the claim for higher resolution and we added: “Still, the ratiometric mode of imaging FRET sensors is beneficial for detection of gradients or activity in 3D imaging”.

2) Throughout the paper, the ratio between the GTPase and the domain, and the overall expression level of each, was not sufficiently examined. The results in many cases would be dependent on both these factors (was a large excess of domain used? Was there insufficient domain to bind the GTPase and provide a signal? Did this vary for different domains, and therefore produce the differences observed? A lack of apparent binding specificity could be produced by high domain expression.)

This is an important point. We will re-analyze the data and include a figure where we add the binding efficiency versus the expression level.

3) In the nuclear exclusion assay, some GTPases were excluded from the nucleus and others not. This was true even without expression of the domains. When GTPases were excluded from the nucleus, domains were eliminated from contention, even when this was true without domain. The authors could at least mention that these domains may be viable.

Correct, and we have added this text: “we cannot exclude that these would be viable Cdc42 sensor candidates”

4) In the multiplexing experiment, only two cells were imaged. In one cell RhoA activity was inversely correlated with Cdc42 activity. In the other cell it was not. It seems there is insufficient information to reach firm conclusions.

We agree and in the revision plan we indicate that we will repeat this experiment to increase the number of cells.

Minor points:

• There appear to be errors in naming mutants. Q60L is used for constitutively active Rac, but Q61L is likely meant. H2A-mTurquoise2-Rac1(G12V)-ΔCaaX is used when it likely should be H2A-mTurquoise2-Rac1(Q61L)-ΔCaaX. There are other examples -- a careful check of these names throughout the manuscript would be valuable.

Thanks for spotting this. Q60L is changed to Q61L. Note that the Rac1(G12V) is correct as it also is a constitutive active Rac1.

• Intro-Paragraph 1-line 5: change present to presence

• Intro-Paragraph 5- line 7: use them instead of theme.

Thanks, both corrected.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript was clearly written, but the introduction or discussion could provide a more balanced view of the significance. There are important experiments that are required to support the conclusions regarding selectivity and differences between the domains, particularly the role of expression level and the ratio of expressed proteins. Our review is summarized here:

The authors set out to improve localization-based biosensors for Rac1 and Cdc42 by identifying domains that bind selectively to the activated conformation of Rac1 and Cdc42. Using screens based on mitochondrial and nuclear relocalization, together with mass spec and proximity biotinylation, several potential candidates were identified. The work concludes with the identification of a WASp CRIB domain that can be used as a useful Cdc42 relocalization biosensor. It was applied in a well-executed proof of principle study demonstrating multiplexed imaging of RhoA and Cdc42 localization biosensors.

1. The discussion comparing different types of biosensors missed important points. Although the advantages of localization biosensors listed by the authors are correct, they gave the impression that these should simply be an improved replacement for FRET biosensors. There are times when FRET biosensors provide clear advantages. Unlike other proteins, Rho GTPases are well suited for localization sensors because the activated conformation, and only the activated conformation, localizes to the membrane. For diffuse or 3D localization FRET can provide better quantification. Subtle features such as gradients are not easily quantified over a background of unattached domain. The authors state that localization biosensors have enhanced spatial resolution, but this is not explained.
2. Throughout the paper, the ratio between the GTPase and the domain, and the overall expression level of each, was not sufficiently examined. The results in many cases would be dependent on both these factors (was a large excess of domain used? Was there insufficient domain to bind the GTPase and provide a signal? Did this vary for different domains, and therefore produce the differences observed? A lack of apparent binding specificity could be produced by high domain expression.)
3. In the nuclear exclusion assay, some GTPases were excluded from the nucleus and others not. This was true even without expression of the domains. When GTPases were excluded from the nucleus, domains were eliminated from contention, even when this was true without domain. The authors could at least mention that these domains may be viable.
4. In the multiplexing experiment, only two cells were imaged. In one cell RhoA activity was inversely correlated with Cdc42 activity. In the other cell it was not. It seems there is insufficient information to reach firm conclusions.

Minor points:

• There appear to be errors in naming mutants. Q60L is used for constitutively active Rac, but Q61L is likely meant. H2A-mTurquoise2-Rac1(G12V)-ΔCaaX is used when it likely should be H2A-mTurquoise2-Rac1(Q61L)-ΔCaaX. There are other examples -- a careful check of these names throughout the manuscript would be valuable.
• Intro-Paragraph 1-line 5: change present to presence
• Intro-Paragraph 5- line 7: use them instead of theme.

Significance

The assays and approach to finding domains for biosensors were novel and interesting. The end result was not as surprising as one might have hoped, but the approach alone made the paper worthwhile.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Mahlandt et al report Rho GTPase relocation sensors. First, the authors picked up candidate peptides based on the Mass-Spec data reported by Sean Munro's laboratory. The authors repeated the experiments to confirm the binding of peptides to mitochondria-targeted Cdc42 and Rac1 and narrowed down the candidate peptides by binding to nuclear Cdc42. The specificity of binding to Rac1 and Cdc42 was also tested. Eventually, they concluded that dimeric Tomato-WASp(CRIB) is the best sensor for Cdc42, which could detect S1P-induced Cdc42 activation in primary endothelial cells. The effort to improve the relocation sensors should be evaluated highly. This reviewer has some suggestions to improve this paper.

Major comments:

1. Statistical tests are missing in most of the figures. If the principal purpose of this work is to compare the performance of candidate peptides, the quantitative comparison is essential. If the purpose is just to report another relocation probe, then, more application data may be necessary.
2. The criteria for selecting the best peptide should be clearly described. Is it just by inspection or based on any quantitative data? We know that quantification of colocalization is a difficult task. Therefore, it depends on the aim of this work whether the authors are asked to show quantitative data or not. If a strict comparison of peptides is aimed at, the expression level of each target peptide should be at a comparable level. It will be also required whether the design of each probe guarantees the proper folding to bind to GTPases.
3. About the images of cells: When a fluorescent image is presented, we assume it represents all other cells. Please check all images whether they are truly representing the data. For example, in Fig. S3 the nuclei of ABI1-expressing cells look weird, and the nucleus of CYRI-A is very large. If this is true, the reason why ABI1 and CYRI-A should be excluded from the candidate is not the relocation efficiency but the undesired effect on cell physiology. For the screening of the peptides, this information is also very important. With that, this paper becomes more valuable for scientists.
4. Please examine the order of panels. For example, the result of mScarlet is on the top in Fig3, but at the bottom in Fig4. Such inconsistency would disturb readers.
5. The label should be consistent throughout the paper. For example, in Fig. 5A, Lck-FRB-mTurquoise2 is labeled as Lck-FRB (without the fluorescent protein's name). WASp(CRIB)-mScarlet-I-WASp(CRIB) is labeled as WASp(CRIB)-mScar-WASp(CRIB) (with fluorescent protein's name). Moreover, the same peptide is labeled as mSca-1xWASp(CRIB) in Panel B. Such inconsistency is confusing.
6. Quantitative insight would improve this work. For example, in Fig. 7, the reason why the authors believe that the probe worked is the accumulation of probe at the tip of lamellipodia and the decrease in cytoplasmic intensity. This reviewer does not think the accumulation of the probe in the small area of the lamellipodia explains the massive decrease of cytoplasmic signals. Probably, a substantial amount of the probe is relocated to the plasma membrane, not limited to the lamellipodia.

Minor comments:

1. Introduction, "FRET signal is typically measured with a wide field microscope.": This reviewer does not agree with this statement. Confocal and two-photon microscopes have also been used widely.
2. Introduction, "G-protein activating proteins (GAP)": It should read as "GTPase-activating proteins (GAPs)"
3. TRIF should read as TIRF.
4. Fig.1: To the best of this reviewer's knowledge, PKN1 was first used as the RhoA target peptide by Yoshizaki et al in 2003. J Cell Biol 162, 223-232. They also examined mDia, Rhoteki, and Rhophilin as the target peptides. Pak1 was first used as the Rac1 probe by Kraynov et al. Science 290, 333-337, 2000. Use of Pak1 as the Cdc42 probe was reported by Itoh et al. Mol Cell Biol 22, 6582-659, 2002. This reviewer believes that the priority of the first report should be respected.
5. Fig. 1: Why do the authors omit other promising candidates shown in panel 1B? Please describe the reason for the choice.
6. Fig. 1B: Be consistent to use either "Name" or "Uni Prot name" in Panel A.
7. Fig. 2: Please include information on TOMM20. The readers may not read the paper by Gillingham et al.
8. Fig3 and 4: The authors should show the images of control H2A.
9. In Fig3B and 4B, "Cdc42/Rac1 affinity" would be misleading, because the control dots represent their authentic localization rather than "Cdc42/Rac1 affinity".
10. Fig. 4: More explanation of this figure is required.
11. Fig. 5: More explanation about the FKBP-FRB system will be helpful.
12. Fig. 6: It is rather surprising to see that control-mScarlet also responds to Rac1 activation. What is the explanation for this observation?
13. Fig. 7: A single champion data may not be convincing to prove the usefulness of this probe.

Significance

1. The authors have screened many peptides, which may serve as the relocation sensor for Rho-family GTPases.
2. There are precedent relocation sensors, a part of which is listed in Fig. 1A. This work discloses an improved relocation biosensor.
3. Cell biologists who is working on Cdc42 will be interested in this probe.
4. Expertise of this reviewer: Signal transduction, Fluorescence microscopy.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript Mahlandt et al. report efforts to generate sensitive fluorescent biosensors to monitor the activation of Rac and Cdc42. While biosensors for these GTPases have been designed and used by others, some have poor specificity while others require the measurement of FRET, which is technically more complex and usually yields low signal to noise ratios. The efforts of Mahlandt et al. are therefore well justified and commendable, and follow on the heels of their successful development of a Rho-selective probe (J. Cell Sci., 2021).

The authors used existing data of interacting proteins to identify candidate domains that could be adapted for use as Rac or Cdc42 activation indicators. They proceeded to select the most promising candidates, assessed their specificity and tried to improve their efficiency by generating tandem constructs that are expected to have higher avidity. They identify CYRA-I as a positive Rac interactor, but the affinity and selectivity of this construct were not deemed to be sufficient and this line of enquiry was not pursued further. In contrast, the WASp-CRIB was found to be sufficiently Cdc42-specific and its avidity was improved by generating a construct tagged with dimeric Tomato fluorescent protein. Unfortunately, while data using overexpression of active Cdc42 are convincing and clear, the results obtained under physiological conditions -stimulating cells with S1P- show very modest recruitment. The general usefulness of the probe is therefore questionable.

There are also a number of specific issues:

1. It is not clear why RhoA data were included in this manuscript (Fig. 1), since they seem irrelevant to the primary topic addressed.
2. It is not clear what cell type was used when screening for p67phox. The expression of this component of the NADPH oxidase is restricted to a few specific cell types.
3. There is precious little quantitation of the colocalization or translocation of the probes throughout the manuscript. It is difficult to assess the validity of the conclusions in the absence of analysis of the statistical significance of the colocalization.
4. It is not clear why translocation to mitochondria was used in some experiments and translocation to the nucleus in others.
5. In the S1P experiments, it is difficult to ascertain whether increased fluorescence resulted from membrane folding/ruffling or is actually a consequence of localized activation of receptors. Why does the fluorescence decrease progressively over 1500 seconds? Isn't maximal receptor activation accomplished much sooner?

Significance

While the purpose and intent of the study are commendable, the results are far from convincing and the probes designed do not represent a sufficient improvement over existing biosensors. The lack of quantitative and statistical analyses is problematic, as is it is difficult to assess the significance of the results.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

1. General Statements [optional]

We thank the reviewers for fair and constructive comments, and for good suggestions for how to further improve the manuscript. We also appreciate the comments on novelty and potential importance for future clinical approaches for CADASIL. Below we provide our point-by-point responses to the reviewers’ comments. We start with a description of planned revisions, followed by describing changes already carried out in the transferred manuscript.

2. Description of the planned revisions

We plan to experimentally address the following points raised by the reviewers.

First, we will address the ASMA staining in the arterioles by isolating brain microvessels from the TgN3R182C150 mouse and WT mouse and strain for ASMA and perlecan/PDGFRb (stain pericytes and VSMC) to in more detail visualize that ASMA is only expressed in the arterioles and not in the pericytes in the capillaries (reviewer 1, point 4).

Second, we will assess the safety and toxicity of the active immunotherapy (reviewer 2, point 4). Specifically, we plan to analyse the structure and morphology of the kidneys from sham and vaccinated mice for renal damage by histology, H/E staining as well as creatinine levels in the serum. If the histology shows signs of damage (such as necrosis, apoptosis or granules in the tubules), we will stain the tissue with KIM1 for acute renal damage and caspase 3 for apoptosis. Moreover, we will analyse the inflammation C-Reactive Protein (CRP) marker in the serum by western blot analysis and/or ELISA and finally Neurofilament Light chain protein in the serum by SIMOA analysis to monitor if the vaccination is causing any neurodegeneration.

Third, we will improve the image quality for Figure 5A (reviewer 1, point 5).

Reviewer #1 (Evidence, reproducibility and clarity):

Summary:

This study have used immunogenic aggregates formed by recombinant Notch3 fragments EGF 1-5 that contain CADASIL NOTCH3 R133C and wild type NOTCH3 to inoculate a CADASIL mouse model TgN3R182C as an active immunisation therapy of CADASIL. The vaccinated mice showed a decreased deposition of NOTCH3 around brain capillaries, reduced blood NOTCH3 ECD and microglia activation, suggesting a potential novel therapy for future treatment of CADASIL. The lack of impact on retinal vasculature, body weight, and general behaviour of the treated mice indicate the safety of the therapy.

We thank the reviewer for the positive comments and for finding it to be a potential novel and safe therapy for CADASIL.

Major comments:

The results are interesting and the manuscript was carefully written and presented. However, the reduction of NOTCH3 in blood samples and the deposition around capillaries were so modest, plus there was no significant change of NOTCH3 deposition in arterioles. This questions the real effectiveness of this immunisation therapy when translating to clinical patients.

We thank the reviewers for the positive words on the manuscript and understand the concerns over the lack of effects in the arterioles. We still believe that observing NOTCH3 ECD reductions around capillaries and in the blood is a significant step forward and supports further endeavors in the area of active immunization, especially considering that the treatment is well tolerated with no overt NOTCH3-related toxicity effects. With this said, we of course realize that there would still be a long way towards clinical use, but all therapy developments have to start somewhere, and future studies (from us and others) can build on the data presented here. There are a number of possible explanations as to why there was not a significant reduction of NOTCH3 around the arterioles. It might be due to that the TgN3R182C150 model is a mild disease model which has a NOTCH3 accumulation onset around 6-7 months of age that starts in the capillaries and at later ages spreads to the arterioles, and that a different time axis or longer immunization period would also lead to a reduction of NOTCH3 in the arterioles, which can be addressed in future experiments. We however believe it was worthwhile to first explore an early immunization protocol, as it may pave the way to a treatment strategy that can be used early on the disease progress, maybe already at the pre-symptomatic phase.

Minor comments:

• Figure EV1 is redundant as Figure 2C has the same information.

We agree the figures are redundant and have deleted Figure EV1 in our new version and renamed EV2 and EV3 to EV1 and EV2, respectively.

• P5 last line "there was a prominent loss of monomeric NOTCH3 EGF1-5 when WT and R133C fragments were mixed as compared to incubating them separately (Fig 2C)". Why the aggregation is more obvious when mutant protein fragment mixed with wild type comparing mutant fragment alone?

The reviewer raises an interesting point. While we do not strictly know why a mixing of mutated and wildtype fragments promote aggregation, our observations are in line with previous reports. Duering et al. Hum. Mol. Genet, 2011 used a single-particle approach SIFT (scanning for intensely fluorescent targets) to study the co-aggregation of WT and mutant NOTCH3 EGF1-5 proteins and reported that mixtures of WT and mutant proteins showed dual colour high-intensity signals, representing de novo aggregates. They also observed WT/R133C multimer formation over three days as compared to five days used in our study. It is also of note that most CADASIL cases are heterozygous for the NOTCH3 mutation, which suggests that wildtypeand mutant NOTCH3 proteins co-exist in vivo, which also have been confirmed by co-immunoprecipitation experiments from cell lysates by Opherk et al. Hum. Mol. Genet. 2009. In the light of these previous reports, we performed a variety of ratio between WT and R133C NOTCH3 EGF1-5 proteins and opted for a 1:1 mixture since it showed the highest amount of multimers in relation to monomers after incubation.

• How was the dosage or concentration of aggregated protein (0.5 mg/ml) used for immunisation determined?

The dosage concentration was calculated by measuring the purified protein with a BCA absorbance assay. The protein was then mixed with the adjuvant or PBS to the desired concentration. This is now more clearly described in the Materials and Methods.

• P7 line 1-2, while using ASMA as a marker for arteries/arterioles, didn't the author see any expression of ASMA in pericytes (capillaries)?

We have used a direct fluorescent conjugated ASMA antibody for this mouse strain and another mouse strain with same mutation but higher NOTCH3 expression (described in Rutten et al, Acta Neuropathol Commun, 2015) as well as for the wildtype mice. In all cases, we find that pericytes have no or very low ASMA expression. This is in line with a report from Ghezali et al. Ann Neurol 2018, in which they used perlecan as a marker for pericytes in the rat Notch3 R169C mutation model, since ASMA did not stain the capillaries. The lack of ASMA staining in pericytes is in agreement with our previous large-scale single cell RNA-seq study of the mouse brain vasculature, where we find very little ASMA (ACTA2) gene expression in pericytes, while it is abundant in the vascular smooth muscle cells (Vanlandewijck et al., Nature 2018).

With this said, we will in the revised version include images on isolated brain microvessels from the N3TgR182C150 mouse model and WT that are stained with ASMA and perlecan/PDGFRb to clearly show that ASMA only stains the arterioles and not the capillaries, see Point 2 “Description of planned revisions” part above.

• In Figure 5A, the NOTCH3 ECD signal looks like similar between Shame and Vaccinated, although the quantifications seem significant in Figure 5B. The author may discuss the robustness of the quantitation method used and therefore the conclusion (i.e., active immunization with NOTCH3 EGF1-5 WT/R133C aggregates specifically reduces the amount of NOTCH3 aggregates around cerebral capillaries.).

We agree that Figure 5A is not clearly showing the differences between vaccinated and sham although highly significant. We have included more representative images and we have also improved the description of the quantification in the method part (page 18-19).

• In Figure 8B, what does the "% of microglia area" mean and how was it calculated?

The % of microglia area represents the percentage that the CD68 staining occupies per microglia area (stained by Iba1), meaning that the vaccinated mice contain more activated microglia compared to sham. This was calculated by measuring the area of the CD68 staining that colocalizes with the microglia. In the transferred version, we have improved the description of this in the text (page 18-19).

• Figure 9A doesn't seem to support the conclusion "There was also a trend towards more NOTCH3 ECD deposits inside or in close vicinity to microglia in vaccinated TgN3R182C150 mice compared to control or sham-vaccinated TgN3R182C150 mice, although the difference did not reach statistical significance", as it is not so convincing that the signals of the NOTCH3 ECD staining co-localise with microglia in the vaccinated sample. Besides, what's the meaning or functional significance about "% of microglia area", "number/1000 um2 microglia", and "average size per microglia"?

Thank you for the comments and we understand the concern, given the highly complex morphology and dynamics of microglia, which makes it difficult to describe potential differences in a straightforward manner. In the first graph in Figure 9B we show that the vaccinated mice had a higher percentage of microglia with NOTCH3 ECD deposits. However, these deposits did not occupy more area of the microglia (% microglia area, second graph), or showed a significant increase in the number of deposits per microglia area (number/1000 µm2 microglia, third graph). The average size of these deposits did not increase in a significant way per microglia area (average size per microglia µm2, fourth graph) in comparison to sham and non-vaccinated mice. We have rephrased the text to better indicate that the proportion of microglia with NOTCH3 deposits increased, while not the area or number of deposits within a microglial cell (Results, page 9).

Reviewer #1 (Significance):

CADASIL is the most common genetic small vessel disease that leads to cognitive defect and eventual vascular dementia. Current, there is no specific treatment available. Similar to a number of neurodegenerative conditions caused by protein accumulations like Alzheimer's disease, Parkinson disease and Huntington disease etc, NOTCH3 protein accumulation represents a key pathological change in CADASIL and therefore a drug target. The approach of active immunisation therapy described in this paper demonstrated a novel method for the treatment of this condition. Although the effectiveness of the therapy in the transgenic mouse model of CADASIL was yet highly impressive, this paper provides a proof-of-principle that the active immunisation is more or less functional and, most importantly, tolerable. One other advantage of this approach is that the immunisation therapy is not restricted to a specific NOTCH3 mutation. After further development this strategy could potentially benefit patients in the future.

This paper may be interested by researchers working on diseases that are caused by specific protein accumulation or aggregations.

My expertise is in the area of studying the molecular mechanisms of CADASIL and other genetic small vessel diseases.

We thank the reviewer for these comments and that the manuscript provides proof-of-principle for a novel strategy towards a safe and tolerable therapy for future treatment of CADASIL patients. As such, we believe the study will be a useful platform and resource for future studies from us and others.

Reviewer #2 (Evidence, reproducibility and clarity):

The article submitted by Daniel V. Oliveira et al to Review Commons and titled "NOTCH3 active immunotherapy reduces NOTCH3 deposition in brain capillaries in a CADASIL mouse model" describes a novel active immunization therapy against the aggregated NOTCH3 mutant protein associated to CADASIL pathology. This strategy induces a significant reduction in NOTCH3 deposition around brain capillaries, increase of microglia activation and lowering of serum levels of NOTCH3, that demonstrates the potential clinical value of the therapy. In addition, the authors report that the therapy is safe and tolerable.<br /> The study is well written, very clear and the results are very promising.

We thank the reviewer for positive comments on our work and pleased that you find the manuscript well and clearly written and containing promising data.

Same minor comments need to be clarified before publication.<br /> 1. In the introduction is indicated that ".....Aducanumab recently being approved for treatment of AD by the Food and Drug Association (Budd Haeberlein et al, 2017; Mintun et al, 2021; Tolar et al, 2020)." In the discussion, Page 10, "In the quest for AD therapies, emerging encouraging results suggest a clinically meaningful effect from immunotherapies aimed at clearing Ab-amyloid which recently also have led to the first FDA approval of a passive vaccine for treatment of AD (Demattos et al, 2012; Golde et al, 2009; Sevigny et al, 2016).<br /> References should be updated with recent clinical studies about the efficacy of Aducanumab.

This is a valid point. We have updated the manuscript accordingly, and included a recent study on the effect of Aducanumab (Knopman et al. Alzheimers Dement. 2021).

1. Schedule used to induce active immunization should be justified or referenced.

We appreciate this comment and realize that we did not explain the rationale for the immunization scheme in the original version. We used a similar immunization schedule as used by Kontsekova et al. Alzheimer's Research & Therapy 2014 in their study of active immunization of a tau peptide in a transgenic rat Alzheimer disease model. This is now mentioned in the text on page 6. The slight modifications of the Kontsekova et al., scheme are also described in the Materials and Methods, page 15 (we made some slight modifications in order to be compatible to mouse and also immunized with an aggregated protein of 25kD instead of a 2 kD tau peptide).

1. CADASIL is associated with high risk of stroke, dementia and migraine. Did the authors check the brain of TgN3R182C150 mice by MRI, for instance? This analysis could be interesting in order to increase the clinical impact and the translational value of the therapy.

Thank you for pointing this out. We have previously reported MRI data on a more severe CADASIL mouse model (TgN3R182C350), which harbors the same transgene but expressed at a 2.3 times higher level compared to the mouse model used in this study (Rutten et al. Acta Neuropathol. Commun. 2015 and Gravesteijn et al. Transl. Stroke Res. 2020). We did not observe any consistent differences on MRI or behavior between the TgN3R182C350 model and controls at 20 months of age. From this data we do not expect to see any differences on MRI in our milder model at 7 months of age. We have however included a short description of the previous MRI data in the transferred version on page 12.

1. It is indicated that the therapy did not affect to the vascular structure of the retina, suggesting that endogenous Notch signaling was not affected by vaccination and therefore the therapy is safe; however, additional tox analysis should be included to confirm the biocompatibility, such as blood pressure analysis, inflammation markers, renal and hepatic damage marker.

Thank you for pointing out this important issue regarding the safety aspects for our active immunotherapy. We will address this by additional experiments of the renal damage by histology and inflammation markers and neuronal damage markers in the serum with western blotting/ELISA and SIMOA, which we have written in more detail in Point 2 “Description of planned revisions” part above.

The suggestions for blood pressure and hepatic damage analysis are valid but would require that we restart novel series of experiments from scratch, as we currently do not have immunized mice up and running. Such an experiment would take several months, and with the subsequent analysis, most likely more than a year. We believe this time axis is too long, and we therefore respectfully suggest waiving these experiments. Also, we unfortunately did not collect livers from the first round of animals, so analysis of livers would have a similarly long time axis.

Reviewer #2 (Significance):

The study is well written, very clear and the results are very promising. Some references related with the effect of Aducanumab in AD should ne updated. Safety and Tox analysis of the therapy should be complemented with additional assays.

We thank the reviewer for these positive comments. We have updated the manuscript with a new reference regarding the effect of Aducanumab. For the safety and tox analysis we will do additional experimental analysis which we have addressed above in “Description of planned revisions”.

3. Description of the revisions that have already been incorporated in the transferred manuscript

Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. If no revisions have been carried out yet, please leave this section empty.

We deleted Figure EV1 in our new version and renamed EV2 and EV3 to EV1 and EV2, respectively. (Reviewer 1, point 1).

We provide an improved description of how the dosage concentration was calculated by measuring the purified protein with a BCA absorbance assay (Reviewer 1, point 3) (Materials and Methods, page 15).

We have improved the description of the quantification of the immunohistochemistry in the method part (Reviewer 1, point 5) (Materials and Methods, page 18-19).

We have improved the description of microglia in the quantification in the method part (Reviewer 1, point 6 and 7) (Materials and Methods, page 18-19).

We have rephrased the text to better indicate that the proportion of microglia with NOTCH3 deposits increased, while not the area or number of deposits within a microglial cell (Results, page 9).

We have updated the manuscript accordingly, and included a recent study on the effect of Aducanumab (Knopman et al. Alzheimers Dement. 2021). (Reviewer 2, point 1).

We have provided a rationale and a more detailed description of the immunization protocol. (Reviewer 2, point 2) (Materials and Methods, page 15).

We have included a short description of the previous MRI data in the transferred version. (Reviewer 2, point 3), (Discussion, page 12)

We have included the following new references (Vanlandewijck et al., Nature 2018, Knopman et al. Alzheimers Dement. 2021, Kontsekova et al. Alzheimer's Research & Therapy 2014, Gravesteijn et al. Transl. Stroke Res. 2020).

4. Description of analyses that authors prefer not to carry out

Please include a point-by-point response explaining why some of the requested data or additional analyses might not be necessary or cannot be provided within the scope of a revision. This can be due to time or resource limitations or in case of disagreement about the necessity of such additional data given the scope of the study. Please leave empty if not applicable.

The suggestions for blood pressure and hepatic damage analysis are valid but would require that we restart novel series of experiments from scratch, as we currently do not have immunized mice up and running. Such an experiment would take several months, and with the subsequent analysis, most likely more than a year. We believe this time axis is too long, and we therefore respectfully suggest waiving these experiments. Also, we unfortunately did not collect livers from the first round of animals, so analysis of livers would have a similar time axis.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The article submitted by Daniel V. Oliveira et al to Review Commons and titled "NOTCH3 active immunotherapy reduces NOTCH3 deposition in brain capillaries in a CADASIL mouse model" describes a novel active immunization therapy against the aggregated NOTCH3 mutant protein associated to CADASIL pathology. This strategy induces a significant reduction in NOTCH3 deposition around brain capillaries, increase of microglia activation and lowering of serum levels of NOTCH3, that demonstrates the potential clinical value of the therapy. In addition, the authors report that the therapy is safe and tolerable.<br /> The study is well written, very clear and the results are very promising.

Same minor comments need to be clarified before publication.

1. In the introduction is indicated that ".....Aducanumab recently being approved for treatment of AD by the Food and Drug Association (Budd Haeberlein et al, 2017; Mintun et al, 2021; Tolar et al, 2020)." In the discussion, Page 10, "In the quest for AD therapies, emerging encouraging results suggest a clinically meaningful effect from immunotherapies aimed at clearing Ab-amyloid which recently also have led to the first FDA approval of a passive vaccine for treatment of AD (Demattos et al, 2012; Golde et al, 2009; Sevigny et al, 2016).<br /> References should be updated with recent clinical studies about the efficacy of Aducanumab.
2. Schedule used to induce active immunization should be justified or referenced.
3. CADASIL is associated with high risk of stroke, dementia and migraine. Did the authors check the brain of TgN3R182C150 mice by MRI, for instance? This analysis could be interesting in order to increase the clinical impact and the translational value of the therapy.
4. It is indicated that the therapy did not affect to the vascular structure of the retina, suggesting that endogenous Notch signaling was not affected by vaccination and therefore the therapy is safe; however, additional tox analysis should be included to confirm the biocompatibility, such as blood pressure analysis, inflammation markers, renal and hepatic damage marker, ....

Significance

The study is well written, very clear and the results are very promising. Some references related with the effect of Aducanumab in AD should ne updated. Safety and Tox analysis of the therapy should be complemented with additional assays.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

This study have used immunogenic aggregates formed by recombinant Notch3 fragments EGF 1-5 that contain CADASIL NOTCH3 R133C and wild type NOTCH3 to inoculate a CADASIL mouse model TgN3R182C as an active immunisation therapy of CADASIL. The vaccinated mice showed a decreased deposition of NOTCH3 around brain capillaries, reduced blood NOTCH3 ECD and microglia activation, suggesting a potential novel therapy for future treatment of CADASIL. The lack of impact on retinal vasculature, body weight, and general behaviour of the treated mice indicate the safety of the therapy.

Major comments:

The results are interesting and the manuscript was carefully written and presented. However, the reduction of NOTCH3 in blood samples and the deposition around capillaries were so modest, plus there was no significant change of NOTCH3 deposition in arterioles. This questions the real effectiveness of this immunisation therapy when translating to clinical patients.

Minor comments:

• Figure EV1 is redundant as Figure 2C has the same information.
• P5 last line "there was a prominent loss of monomeric NOTCH3 EGF1-5 when WT and R133C fragments were mixed as compared to incubating them separately (Fig 2C)". Why the aggregation is more obvious when mutant protein fragment mixed with wild type comparing mutant fragment alone?
• How was the dosage or concentration of aggregated protein (0.5 mg/ml) used for immunisation determined?
• P7 line 1-2, while using ASMA as a marker for arteries/arterioles, didn't the author see any expression of ASMA in pericytes (capillaries)?
• In Figure 5A, the NOTCH3 ECD signal looks like similar between Shame and Vaccinated, although the quantifications seem significant in Figure 5B. The author may discuss the robustness of the quantitation method used and therefore the conclusion (i.e., active immunization with NOTCH3 EGF1-5 WT/R133C aggregates specifically reduces the amount of NOTCH3 aggregates around cerebral capillaries.).
• In Figure 8B, what does the "% of microglia area" mean and how was it calculated?
• Figure 9A doesn't seem to support the conclusion "There was also a trend towards more NOTCH3 ECD deposits inside or in close vicinity to microglia in vaccinated TgN3R182C150 mice compared to control or sham-vaccinated TgN3R182C150 mice, although the difference did not reach statistical significance", as it is not so convincing that the signals of the NOTCH3 ECD staining co-localise with microglia in the vaccinated sample. Besides, what's the meaning or functional significance about "% of microglia area", "number/1000 um2 microglia", and "average size per microglia"?

Significance

CADASIL is the most common genetic small vessel disease that leads to cognitive defect and eventual vascular dementia. Current, there is no specific treatment available. Similar to a number of neurodegenerative conditions caused by protein accumulations like Alzheimer's disease, Parkinson disease and Huntington disease etc, NOTCH3 protein accumulation represents a key pathological change in CADASIL and therefore a drug target. The approach of active immunisation therapy described in this paper demonstrated a novel method for the treatment of this condition. Although the effectiveness of the therapy in the transgenic mouse model of CADASIL was yet highly impressive, this paper provides a proof-of-principle that the active immunisation is more or less functional and, most importantly, tolerable. One other advantage of this approach is that the immunisation therapy is not restricted to a specific NOTCH3 mutation. After further development this strategy could potentially benefit patients in the future.

This paper may be interested by researchers working on diseases that are caused by specific protein accumulation or aggregations.

My expertise is in the area of studying the molecular mechanisms of CADASIL and other genetic small vessel diseases.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Author response (Tane at al: RC-2022-01646)

Reviewer #1 (Evidence, reproducibility and clarity (Required)): * Comments The work described in this manuscript starts with an in-silico analysis of the primary amino-acid sequence of CAP-H proteins that reveals the presence in vertebrate orthologs of an N-terminal extension of ~80 amino acids in length which contains 19 serine or threonine residues and also, in its centre, a stretch of conserved basic amino acids predicted to form a helix. These features suggest a regulatory module. Using xenopus egg extracts depleted of endogenous condensins and supplemented with recombinant condensin I holocomplexes, either wildtype or mutants, the authors show that deleting the N-terminal tail of CAP-H, or just the central helix (CH), increases the association condensin I with chromatin in mitotic egg extracts and accelerates the formation of mitotic chromosomes. Interestingly, they also show that deleting the N-tail enables a substantial amount of condensin I to associate with chromatin in interphase extracts and to form chromosome-like structures, while WT condensin I cannot. Using in vitro assays and naked DNA as substrate, the authors further show that removing the N-terminal tail of CAP-H improves both the topological (salt-resistant) association of condensin I with DNA and it loop extrusion activity. These experiments appear to me as are clear and robust. They convincingly reveal that N-tail of human CAP-H hinders the binding of condensin I to DNA and both its loop-extrusion and chromosome-shaping activities. However, the mechanism through which such hindrance is achieved remains elusive (see major comments 1-3). A complementary part of the work tackles the important question of the cell cycle control of such counteracting effect. Using newly-designed antibodies against two phospho-serine residues within the tail, the authors provide evidence that the tail is phosphorylated in mitosis-specific manner. This points towards phosphorylation as a biological mean to modulate the effect of the tail on condensin's binding during the cell cycle. In support to this idea, using non-phosphorylatable or phosphomimic substitutions of all the serine and threonine residues within the tail (n =19), including one substitution within the CH domain (Ser 70), the authors show that non-phosphorylatable mutations (H-N19A) or phosphomimic mutations (H-N19D) respectively reduce or improve condensin I binding to chromatin in mitotic egg extracts. This suggests that the phosphorylation of the N-terminal tail in mitosis might relieve its negative effect on condensin I binding to chromatin. The weaknesses I see in this part of the study concern (1) the validation of the phospho-antibodies, which appears to me as insufficiently described (major comment 4), (2) the possibility the bulk changes in amino acids (n=19 out of 80) could impact the folding of the central helix (minor comment X) and (3) that some substitutions could impact the binding of condensin I by different mechanisms (minor comment X).

Major comments:

1. On the model. The authors propose that the N-tail could stabilise an interaction between the N-terminal part of CAP-H and SMC2's neck, which would restrain the transient opening of a DNA entry gate within the ring, necessary for the topological engagement of DNA and loop formation. Although the model is sound, I see no direct data that support it in the manuscript. Such model predicts that a CAP-H protein containing or not the N-terminal tail (or the central helix) should exhibit different binding strengths to SMC2 in vitro. It seems to me that the authors could easily test this prediction using the recombinant proteins they produced in the context of this study. *

Response

We thank the reviewer for pointing out this important issue. To test whether the CAP-H N-tail indeed contributes to the stabilization of the SMC2-kleisin gate, we set up a highly sophisticated functional assay described by Hassler et al (2019). The authors used this assay to demonstrate that an N-terminal fragment of kleisin (engineered to be cleaved by TEV protease) is released from the rest of the condensin complex in an ATP-dependent (i.e., head-head engagement-dependent) manner. We reasoned that this assay is most powerful to prove our hypothesis in a mechanistically relevant context. We envisioned that the CAP-H fragment lacking its N-tail can readily be released whereas the CAP-H fragment retaining its N-tail is more difficult to be released (because of the postulated stabilization of the SMC2-CAP-H interaction). Despite substantial efforts in making TEV-cleavable constructs and in testing various releasing conditions, we have not been able to recapitulate the ATP-dependent release even with the holo(H-dN) construct. Thus, unfortunately, this trial enabled us to neither prove nor disprove our hypothesis.

We are fully aware that the full reconstitution of ATP-dependent and phosphorylation-stimulated gate-opening reaction in vitro is a very important direction in the future. It is beyond the scope of the current study, however.

2. On ATP-hydrolysis. Given the importance of ATP hydrolysis for the engagement of condensin into a topological mode of association with DNA and for its loop extrusion activity, I suggest that the authors measure the impact of the N-tail and of the CH domain on the rate of ATP hydrolysis by condensin I holocomplexes. I suppose that it can be relatively easily done (PMID: 9288743) using the recombinant WT and mutant versions they purified in the course of this study.

Response

We appreciate this constructive comment. In fact, we did a preliminary experiment and found that ATPase activities (either in the absence or presence of DNA) were not significantly different between holo(WT) and holo(H-dN). We were not surprised with this result because our previous study on condensin II indicated that enhanced ATP hydrolysis by a class of mutant complexes is not directly coupled to their enhanced association with chromosomes (Yoshida et al., 2022, eLife). We consider that other functional assays, such as the topological loading assay and the loop extrusion assay shown in the current manuscript, are more informative assays to address ATP-dependent activities of the condensin complexes.

3. A conundrum with previous work? In Kimura et al. Science 1998 (PMID: 9774278), the lab of Tatsuya Hirano has shown that xenopus condensin I purified from mitotic egg extracts induces the supercoiling of plasmid DNA in vitro, but fails to do so when it is purified from interphase egg extracts. This echoes the inhibitory effect of the N-tail of the topological binding of condensin I described in the current manuscript. However, using a gel shift assay, Kimura et al. 1998 also provide evidence that interphase and mitotic condensin I bind plasmid DNA in vitro with similar efficiencies. At first sight, this prior observation seems to contradict the idea that the N-tail of CAP-H restrains DNA binding unless it is phosphorylated in mitosis. Is it possible that the in vitro binding assays used in Kimura et al. 1998 and in this work might assess different modes of binding? I suggest that this apparent conundrum should to be discussed.

Response

We thank the reviewer for following our early studies. As discussed below, we are confident that our conclusion reported in the current study by no means contradicts our previous observations.

We reason that the confusion expressed by the reviewer stems from intrinsic, technical limitations of the gel-shift assay. Such limitations become apparent especially when it is applied to the functional analyses of complicated protein machines such as condensins. For instance, the DNA-binding activity of condensin I detected by the gel-shift assay is neither ATP-dependent nor phosphorylation-dependent (Kimura and Hirano, 1997; Kimura et al., 1998). It is fundamentally different from the ATP-dependent activities detected by the topological loading and loop extrusion assays reported in the current study (It remains unknown whether the two activities are stimulated by mitotic phosphorylation). Thus, the DNA-binding activity detected by the gel-shift assay does not reflect “productive” DNA interactions that depend on ATP hydrolysis in vitro. We therefore consider that gel-shift analyses of holo(WT) and holo(H-dN) would not produce any useful information.

*Related to that, could it be possible for the authors to assess the impact of the N-tail on the salt-sensitive binding of condensin to DNA, i.e. by reproducing the topological binding assay but omitting the high salt washes? I guess this information could be useful to fully apprehend the impact of the N-tail on the binding of condensin. *

Response

When we set up the topological loading assay, we actually tested a low-salt wash condition that the reviewer suggests here. Although a much higher level of DNA recovery was observed with the low-salt condition than with the high-salt wash condition, no nucleotide dependency was detectable with the low-salt condition. Moreover, no difference in DNA recovery between holo(WT) and holo(H-dN) was observed. Thus, the low-condition condition allowed us to detect the “bulk” DNA-binding activity that is equivalent to that detected by the gel-shift assay. These results were fully consistent with the discussion above.

4. Validation of phospho-antibodies and by extension showing the phosphorylation of the tail. The newly-designed phospho-serine antibodies used in this study to show that the N-tail is phosphorylated at serine 17 and serine 76 in mitosis (Fig. EV3) are, in my view, not characterized enough. This piece of data is key to substantiate the idea that the tail is phosphorylated in mitosis. Yet, showing that these antibodies detect epitopes on WT condensin I from mitotic egg extracts but not on the H-N19A counterpart, nor on WT condensin I from interphase extracts, does not demonstrate the phospho-specificity of such antibodies. I suggest that a PPase treatment should be conducted to assess the phospho-specificity of these antibodies. Moreover, since the lab of Tatsuya Hirano has the know-how to deplete Cdc2/CDK1 from xenopus egg extract, such strategy could/should be used to further validate the antibodies and assess to which extent the N-tail is phosphorylated in a Cdc2-dependent manner.

Response

We have performed two sets of experiments to confirm the specificity of the phosphoepitopes recognized by anti-hHP1 and anti-hHP2. In the first set, we performed a phosphatase treatment assay. Holo(WT) that had been preincubated with Dcond M-HSS was immunoprecipitated using an antibody against hCAP-G, treated with l protein phosphatase in the presence or absence of phosphatase inhibitors, and analyzed by immunoblotting using anti-hHP1 and anti-hHP2. The results (now shown in Supplementary Fig 3C) demonstrated that the epitopes recognized by anti-hHP1 and anti-hHP2 are sensitive to phosphatase treatment. In the second set, we performed a phosphopeptide competition assay. Holo(WT) that had been preincubated with Dcond M-HSS was immunoprecipitated and subjected to immunoblotting. The membranes were triplicated and probed with anti-hHP1 in the presence of no competing peptide, hHP1 or hHP2. Similarly, another set of triplicated membranes was probed with anti-hHP2 in the presence of no competing peptide, hHP1 or hHP2. We found that the signal recognized by anti-hHP1 competed with hHP1, but not with hHP2, and that the signal recognized by anti-hHP2 competed with hHP2, but not with hHP1. The results (now shown in Supplementary Fig 3D) demonstrated the sequence specificity of the phosphoepitopes recognized by the two antibodies. The procedures for these experiments have been described in Materials and Methods.

Because Cdk1 depletion from M-HSS creates an HSS equivalent to I-HSS, we do not consider that the suggested experiment will provide additional information.

*Minor comments:

1. The impact of the 19 mutations, A or D, introduced within the tail on the folding of the central helix? The idea that the negative effect of the N-tail is relieved by phosphorylation is based on the chromatin binding phenotypes exhibited by the H-N19D or H-N19A mutant holocomplexes, in which 19 amino-acids out of 80 have been modified, include one in the central helix. The authors also provide evidence that the central helix (CH) located within the tail plays a key role in the negative regulation of condensin I binding. Thus, I wonder to which extent the folding of the central helix could be impacted by the mutations introduced in the tail and notably the one within the central helix itself. Could the author assess the structure of mutated tails using Alpho-fold and/or discuss this point? *

Response

According to the reviewer’s suggestion, we performed structure predictions using Alphafold2, and found that neither the N19A nor N19D mutations alter the original prediction of helix formation that was made for the wild-type CH sequence. A conventional secondary structure prediction using Jpred4 reached the same conclusion.

2. Phosphorylation of serine 70 in the central helix by Aurora-B kinase? A prior study by Tada et al. (PMID: 21633354) has shown (1) that serine 70 of the N-tail of hCAP-H is phosphorylated by Aurora-B kinase, (2) that the mutation S70A reduces the binding of condensin I to chromatin in HeLa cells and (3) that hCAP-H interacts with histone H2A in an Aurora-B dependent manner. This draws a picture in which the phosphorylation of Ser70 by Aurora-B would improve condensin I binding to chromatin by promoting an interaction between hCAP-H and histone H2A/nucleosomes. Intriguingly, Ser 70 in Tada et al. correspond to the serine residue located within the conserved central helix analysed in this study, and this Ser70 residue is mutated in the H-N19D or H-N19A holocomplexes that show reduced chromatin binding in this study. This raises the question as what could be the contribution of the S70A or S70D substitution to the chromatin binding phenotypes shown by the H-N19D or H-N19A holocomplexes. This is not discussed in the manuscript, and the authors do not cite this earlier work (PMID: 21633354) in their manuscript. Is there any reason for that? I suggest it should be cited and discussed.

Response

We thank the reviewer for bringing up this issue. In many respects, we do not trust the data reported by Tada et al (2011) and the resultant model they proposed. Previous and emerging lines of evidence reported from our own and other laboratories indicate that histones compete with condensins for DNA binding, strongly arguing against the possibility that histone H2A acts as a “chromatin receptor” for condensins. We formally discussed and criticized the Tada 2011 model in our previous publications, which included Shintomi et al (2015) NCB, Shintomi et al (2017) Science, Hirano (2016) Cell and Kinoshita/Hirano (2017) COCB. We thought that those were enough. That said, we also consider that the reviewer is right. The current study demonstrates that the deletion of the CAP-H N-tail accelerates, rather than decelerates, condensin I loading, providing an additional line of evidence that argues against the Tada model. A critical comparison between the Tada model and our current model would benefit the readers. In the revised manuscript, we have added the following discussion:

In terms of the regulatory role of the CAP-H N-tail, it would be worthy to discuss the model previously proposed by Tada et al (2011). According to their model, aurora B-mediated phosphorylation of the CAP-H N-tail allows its direct interaction with the histone H2A N-tail, which in turn triggers condensin I loading onto chromatin. Accumulating lines of evidence, however, strongly argue against this model: (i) aurora B is not essential for single chromatid assembly in Xenopus egg extracts (MacCallum et al., 2002) or in a reconstitution assay (Shintomi et al., 2015); (ii) the H2A N-tail is dispensable for condensin I-dependent chromatid assembly in the reconstitution assay (Shintomi et al., 2015); (iii) even whole nucleosomes are not essential for condensin I-mediated assembly of chromatid-like structures (Shintomi et al., 2017). The current study demonstrates that the deletion of the CAP-H N-tail accelerates, rather than decelerates, condensin I loading, providing an additional piece of evidence against the model proposed by Tada et al (2011).

3. Other minor comments - Please provide a microscope image of DNA loop in Fig. 4D.

Response

In the revised manuscript, we have provided a set of time-lapse images of loop extrusion events catalyzed by holo(WT) and holo(H-dN) in Fig 4E.

*- The authors do not compare the kleisin of condensin I with the one of condensin II with respect to the features tackled in this work. Given the different behaviours condensin I and II, such comparison could be informative for the readers. *

Response

We thank the reviewer for this constructive comment. In the revised manuscript, we have added the following statement:

It should also be added that CAP-H2, the kleisin subunit of condensin II, lacks the N-terminal extension that corresponds to the CAP-H N-tail. Thus, the negative regulation by the kleisin N-tail reported here is not shared by condensin II.

*- The authors do not reference the work of Robellet et al. Genes & Dev (2015) (PMID: 25691469) on the regulation of condensin binding in budding yeast by an SMC4 phospho-tail. I suggest that the analogy should be discussed. *

Response

According to the reviewer’s comment, we have added the following statements at the beginning of Discussion.

Previous studies showed that mitotic phosphorylation of Cut3/SMC4 regulates the nuclear import of condensin in fission yeast (Sutani et al. 1999) and that phosphorylation of Smc4/SMC4 slows down the dynamic turnover of condensin on mitotic chromosomes in budding yeast (Robellet et al. 2015; Thadani et al. 2018). In the current study, we have focused on the phosphoregulation of vertebrate condensin I by its kleisin subunit CAP-H.

- In the introduction section, lane 5, the sentence "Many if not all eukaryotic species have two different condensin complexes" appears inappropriate since budding and fission yeast cells possess a single condensin complexes, similar to condensin I in term of primary amino-acid sequence.

Response

We thought that the original wording “Many if not all” was good enough to imply that some species, which include budding yeast and fission yeast, have only a single condensin complex. To make it clear, however, we have modified the sentence in the revised manuscript as follows:

Many eukaryotic species have two different condensin complexes although some species including fungi have only condensin I.

*- page 4; typo: motif I and V bind to the SMC neck and the SMC4 cap regions, respectively. Should read SMC2 neck. *

Response

The reviewer is right. It should read the SMC2 neck. Corrected.

*- Are the data and the methods presented in such a way that they can be reproduced? YES - Are the experiments adequately replicated and statistical analysis adequate? YES - Are prior studies referenced appropriately? Not all of them (see above) - Are the text and figures clear and accurate? YES

CROSS-CONSULTATION COMMENTS I consider the comments from all reviewers as helpful for the authors.

Reviewer #1 (Significance (Required)):

Summary Condensins are genome organisers of the family of SMC ATPase complexes and are best characterized as the drivers of mitotic chromosome assembly (condensation). It is acknowledged that condensins shape mitotic chromosomes by massively associating with DNA upon mitotic entry (loading step) and by folding chromatin fibres into arrays of loops, most likely through an ATP-dependent extrusion of DNA into loops, as seen in vitro. What remains unclear, however, are the mechanisms by which condensins load onto DNA and fold it into arrays of loops in vivo, and how these reactions are coupled with the cell cycle, i.e. restricted mostly to mitosis. Condensins are ring shaped pentamers that change conformation upon ATP-hydrolysis. In vitro studies suggest that condensins bind DNA via ATP-hydrolysis-independent, direct electrostatic contacts between condensin subunits and DNA. Such electrostatic contacts are salt-sensitive in in-vitro assays. Upon ATP-hydrolysis, condensins engage into an additional mode of binding that is resistant to high salt concentration and likely to be topological in nature. Both modes of association are necessary to form DNA loops. Vertebrates possess two types of condensin complexes, condensin I and II, each composed of a same SMC2-SMC4 ATPase core but associated with two different sets of three non-SMC subunits; a kleisin and two HEAT-repeat proteins. Condensin II is nuclear during interphase and stably binds chromatin upon mitotic entry, while condensin I is located in the cytoplasm during interphase and binds chromatin in a dynamic manner upon nuclear envelope breakdown. How the spatiotemporal control of condensin I and II is achieved remains poorly understood. Previous studies have shown that the phosphorylation of condensin I by mitotic kinases, such as CDK1, Aurora-B and Polo, play a positive role in its binding to chromatin and/or its functioning, but the underlying mechanisms remain to be characterised. In this manuscript, Shoji Tane and colleagues provide good evidence that the N-terminal tail of the human kleisin subunit of condensin I, hCAP-H, serves as a regulatory module for the cell-cycle control of condensin I binding to chromatin and chromosome shaping activity. The authors clearly show that the N-tail of CAP-H hinders the binding of condensin I to chromatin in xenopus egg extracts and, using in vitro assays, that the N-tail also hinders the topological association of condensin I with DNA and its loop extrusion activity. The authors provide additional data suggesting that the phosphorylation of the N-tail of CAP-H, in mitosis, relieves its inhibitory effect on condensin I binding. Based on their findings, Tane et al. propose a model suggesting that the N-terminal tail of CAP-H constitutes a gate keeper that maintains condensin-rings in a closed conformation that is unfavourable for topological binding to DNA, and whose locking effect is relieved in mitosis by phosphorylation.

Taken as a whole, this work has the potential to reveal a molecular basis for the cell cycle regulation of condensin I in vertebrate cells and as such to significantly improve our understanding of the integrated functioning condensin I. The characterisation of the inhibitory effect of the N-tail on condensin binding to chromatin and to naked DNA in vitro is well described, the data are clear and robust and the results convincing. On the other hand, some of the data on the phospho-regulation appear to me as are more debatable and I think that some of the results described here deserve to be discussed in the context of previous works. Finally, I see no data in the manuscript that directly supports the mechanistic model proposed by the authors, while it seems to me that such model could have been easily tested exprimentally. Thus, I suggest that Tane and colleagues should perform a couple of relatively easy experiments to strengthen their claims and that a few omitted prior studies on the topic should be referenced and discussed. *

Reviewer #2 (Evidence, reproducibility and clarity (Required)): * The manuscript reveals that the N-terminal region of CAPH could play a role in regulating condensin I activity, using a range of in vitro methods. They propose that the N-terminal region of CAPH inhibits complex activity, and this is turned off upon deletion or phosphorylation, by using truncations, phospho-mimics or phospho-deficient mutations. While the results are interesting to the field, and helps to address the question as to how condensin complexes are controlled in a cell cycle dependent manner, some key data and controls are necessary to ensure the conclusion is robust.

Main comments

• What is meant by "unperturbed I-HSS" on page 7, ie membrane containing versus membrane free or condensin depleted? *

Response

We apologize for having created unnecessary confusion. We meant that the “unperturbed I-HSS” is the “undepleted I-HSS”. As far as the issue of membrane-containing vs membrane-free is concerned, we explicitly mentioned that “we used membrane-free I-HSS in the following experiments” several lines above. In the revised manuscript, we have revised the statement accordingly.

In many of the protein gels eg figure 4B, the bands for SMC2 and 4 are more intense that the non-SMC components. The method for protein purification also does not include a size exclusion step to ensure sample homogeneity. Authors should perform some sort of quality control checks on samples such as analytical gel filtration or mass photometry to ensure stoichiometry/homogeneity. This is particularly important for samples eg the N19A, where activity is reduced compared to wild-type as poor protein behaviour could create false negative results.

Response

As the reviewer is fully aware, the reconstitution and purification of multiprotein complexes, such as condensins, is by no means an easy task. We notice that many groups in the field share common concerns about sample homogeneity and subunit stoichiometry, and that these concerns cannot completely be eliminated even after size exclusion chromatography. Because the current study handles a large number of mutant complexes, we consider that the purification by two-step column chromatography is the most practical approach. We do not notice any abnormal behaviors of holo(H-N19A) in the processes of expression and purification. It is also important to emphasize that the H-N19D mutations cause the completely opposite phenotype. Taken all together, we are confident of our current conclusions.

That said, in the revised manuscript, we have added the following statement in Results:

Although we cannot rule out the possibility that the introduction of multiple mutations into the N-tail causes unforeseeable adverse effects on protein conformations, these results supported the idea that ….

• Loop extrusion assays in figure 4D-G shows no example data i.e. no pictures or videos of loops being formed. These should also be included.*

Response

In the revised manuscript, we have provided a set of time-lapse images of loop extrusion events catalyzed by holo(WT) and holo(H-dN) in Fig 4E.

• Given the mutant holo(H-dN) has higher activity than wild-type, a negative control such as holo(H-dN) without ATP or holo(H-dN) ATPase deficient mutant should also be measured in loop extrusion assays, to ensure the activity is derived from the ATPase activity.*

Response

In the revised manuscript, we have added loop formation data for both holo(WT) and holo(H-dN) in the absence or presence of ATP (Supplementary Fig 5). We are confident that both complexes support loop extrusion strictly in an ATP-dependent manner.

• According to the methods, this work performs the same loop extrusion assay as described in Kinoshita et al, 2022, however, in Kinoshita et al, wild type condensin I makes loops in 30-50% of DNA molecules, where in this study the percentage is less than half that. Can the author please explain the discrepancy given the same method was used?*

Response

First of all, we wish to remind the reviewer that the holo(WT) constructs used in the two studies are not identical: CAP-H was N-terminally HaloTagged in all constructs used in Kinoshita et al (2022), whereas the same subunit was C-terminally HaloTagged in the pair of constructs used in the current study. Because we wanted to compare the activities between the full-length CAP-H and N-terminally deleted version of CAP-H (H-dN), we reasoned that it would be inappropriate to put the HaloTag to the N-terminus of CAP-H. The difference in the constructs could explain the observed discrepancy, even if it might not be the sole reason.

The design of the constructs was accurately described in each manuscript, but the statements were not very explicit about the positions of the HaloTag. To clarify this issue, we have added the following sentences in the revised manuscript:

Note that the HaloTag was fused to the C-terminus of CAP-H in the current study because we wanted to investigate the effect of the N-terminal deletion of CAP-H. We used N-terminally HaloTagged CAP-H constructs in our previous study (Kinoshita et al., 2022).

• In the concluding statement the author suggests "Upon mitotic entry, multisite phosphorylation of the N-tail relieves the stabilization, allowing the opening of the DNA entry gate, hence, the loading of condensin I onto chromosomes." This seems unlikely as fusion the N-terminus of the of the kleisin to the C-terminus of SMC2 is able to function for yeast (Shaltiel et al 2022) and condensin II (Houlard et al 2021), and equivalently in cohesin (Davidson et al 2019).*

Response

We appreciate the reviewer’s concern. In our view, however, the issue of the “DNA-entry gate” remains under debate in the SMC field (e.g., Higashi et al [2020] Mol Cell; Taschner and Gruber [2022] bioRxiv). For instance, Shaltiel et al (2022) demonstrated that neck-gate fusion constructs can support in vitro activities including topological loading under certain conditions, but also showed that such constructs greatly reduce the cell viability, leaving the possibility that the gate opening is required for some physiological functions.

That said, it is true that the data reported in the current manuscript do not exclude the possibility that the SMC2 neck-kleisin interface is not used as a DNA entry gate for condensin I loading. In the revised manuscript, we have added the following statement in Discussion:

Although our model predicts that the SMC2 neck-kleisin interface is used as a DNA entry gate, we are aware that several studies reported evidence arguing against this possibility (e.g., Houlard et al [2021]; Shaltiel et al [2022]). Our current data do not exclude other models.

*Reviewer #2 (Significance (Required)):

This is an interesting story that reveals new insights about condensin regulation.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

This paper reveals a role of an N-terminal extension of CAP-H in the regulation of condensin-I activity in Xenopus extracts using biochemical reconstitution experiments. The authors demonstrate that a motif in the N-terminal tail that is conserved in vertebrates acts as an inhibitor of condensin I activity. Using several mutant constructs, it is shown that the inhibition by this motif is in turn counteracted by the phosphorylation of neighbouring serine and threonine residues in mitosis, presumably at least in part by CdK. Mutants that have lost this inhibition are able to condense chromatin into chromatid-like structures more efficiently and to some degree even in interphase extracts. Moreover, one such mutant is characterized in detail by biochemical and biophysical experiments and shown to have increased capacity in salt-stable DNA loading and in DNA loop extrusion.

Major comments: This is a beautiful and thorough study that is presented in a clear and concise manner. The main conclusions are well justified. No additional experiments are needed to support them. Replication and statistical analysis appear adequate. The final model is however largely speculative. Recent work has indicated that loading of yeast condensin does not require gate opening. The authors may thus want to include alternative scenarios or remain more vague. *

Response

This comment is related to the last comment of Reviewer#2. See above for our response.

*The H-N19A mutant has a loss of function phenotype (possibly due to folding problem caused by 19 point mutations rather than lack of phosphorylation), the authors could consider to rescue the phenotype by also including the CH motif mutations in this construct (or make an explanatory statement in the text). *

Response

We understand the reviewer’s logic here, but overlaying additional mutations into the H-N19A mutations could cause an unforeseeable effect, potentially making the interpretation of the outcome complicated.

We also wish to point out that it may be inappropriate to regard the phenotype exhibited by holo(H-N19A) as a simple loss-of-function phenotype. This is because the opposite, accelerated loading phenotype exhibited by holo(H-dN) can be regarded as a consequence of loss of negative regulation. Like holo(H-dN), the phosphomimetic mutant complex holo(H-N19D) displayed an accelerated loading phenotype, fully supporting our conclusion. In the revised manuscript, we have added the following statement in Results:

Although we cannot rule out the possibility that the introduction of multiple mutations into the N-tail causes unforeseeable adverse effects on protein conformations, these results supported the idea that ….

*Albeit not necessary for the main conclusions, the authors could possibly significantly strengthen their study by testing for binding partners of the N-tail and the CH motif by running AlphaFold predictions against the condensin I subunits. *

Response

We appreciate this constructive comment. We attempted to predict possible interactions between SMC2 and a CAP-H fragment containing its N-tail and motif I using

ColabFold (Mirdita et al., 2022, Nat. Methods). The algorism excellently predicted the proper folding of the CAP-H motif I and its interaction with the SMC2 neck. Under this condition of predictions, however, the N-tail remained largely disordered (except for the CH), and no interaction with any part of SMC2 was predicted. The same was true when the N19D mutations were introduced in the N-tail sequence. Thus, this trial did not provide much information about the potential interaction target(s) of the CAP-H N-tail.

*The efficiency of depletion of condensin subunits from I-HSS extracts is not documented (in contrast to M-HSS extracts - figure EV1C). While any condensin remaining in these extracts might not be active (or interfering), the authors may want to at least comment on this in the text. *

Response

We check the efficiency of immunodepletion every time by immunoblotting and confirm that a high level of depletion is achieved from both M-HSS and I-HSS. According to the reviewer’s comment, the following statement was placed in Materials and Methods:

The efficiency of immunodepletion was checked every time by immunoblotting. An example of immunodepletion from M-HSS was shown in Supplemental figure 1C. We also confirmed that a similar efficiency of immunodepletion was achieved from I-HSS.

*The authors should include information on the quantification of chromatid morphology. Is the analysis based on chromatids taken from the same images/imaging session, from technical replicates, biological replicates? *

Response

In the revised manuscript, we have added statements on image presentation and experimental repeats in the appropriate figure legends and methods section. During the revision process, we repeated the experiments shown in Supplementary Fig 2, and obtained the same results. In the revised manuscript, the original set of data has been replaced with the new set of data along with panel C (Quantification of the intensity of mSMC4 per DNA area).

Minor comment: The colour scheme in Figure 5A is confusing. Use less colour? The orange and red colours are moreover quite similar.

Response

According to the reviewer’s comment, we have modified Figure 5A.

*Reviewer #3 (Significance (Required)):

The findings provide new insights into how condensin-I activity is restricted outside of mitosis. It was previously assumed that this regulation was (largely) due to the exclusion of condensin I from the nucleus prior to nuclear envelope breakdown. This study shows that another pathway is contributing to the regulation and implies that controlling condensin I activity is more important than previously appreciated. Whether all residual nuclear condensin I is inactivated, remains to be determined. The physiological impact of loss of autoinhibition on chromosome segregation and cell cycle progression also remains to be uncovered. The observed effects are robust and appear significant. Future research on condensin I using reconstitution will likely benefit from being able to control or eliminate the self-inhibition.

This reviewer has expertise on the biochemistry and structural biology of SMC protein complexes.

Reviewer #4 (Evidence, reproducibility and clarity (Required)):

Mitotic chromosome formation is a cell cycle-regulated process that is crucial for eukaryotic genome stability. The chromosomal condensin complex promotes chromosome condensation, but the temporal control over condensin function is only scantly understood. In this impressive manuscript, "Cell cycle-specific loading of condensin I is regulated by the N-terminal tail of its kleisin subunit", Tane and colleagues provide important new insight into the cell cycle-regulation of condensin. The authors identify a kleisin N-tail that acts as a negative regulator of condensin's DNA interactions. Removal of this N-tail, or its cell cycle-dependent phosphorylation, relieves inhibition and activates condensin. This is a simple, yet very important story, that advances our molecular understanding of chromosome formation. The experiments are performed to a very high technical standard and support the authors conclusions. This manuscript is highly suitable for publication in any molecular biology journal, once the authors have considered the following points.

1. Introduction. a) The authors could better explain their own prior work (Kimura et al. 1998), which has identified the condensin XCAP-D2 and XCAP-H as the targets of phosphoregulation. The current manuscript explains the role of XCAP-H phosphorylation. *

Response

According to the reviewer’s comment, we have added the following sentence in Introduction:

The major targets of mitotic phosphorylation identified in these studies included the CAP-D2 and CAP-H subunits.

1. b) Given the limited knowledge about condensin cell cycle regulation, it seems prudent to provide a brief summary of what is known. Fission yeast Smc4 phosphorylation regulates condensin nuclear import (Sutani et al. 1999), while budding yeast Smc4 phosphorylation slows down the dynamic turnover of the condensin complex on chromosomes (Robellet et al. 2015 and Thadani et al. 2018).

Response

We appreciate this constructive comment. According to the reviewer’s comment, we have added the following statements at the beginning of Discussion.

Previous studies showed that mitotic phosphorylation of Cut3/SMC4 regulates the nuclear import of condensin in fission yeast (Sutani et al. 1999) and that phosphorylation of Smc4/SMC4 slows down the dynamic turnover of condensin on mitotic chromosomes in budding yeast (Robellet et al. 2015 and Thadani et al. 2018). In the current study, we have focused on the phosphoregulation of vertebrate condensin I by its kleisin subunit CAP-H.

2. Extracts were mixed with mouse sperm nuclei. If there is a reason why mouse rather than Xenopus sperm nuclei were used, this would be interesting to know.

Response

The original motivation for introducing mouse sperm nuclei into Xenopus egg extracts was to test the functional contribution of nucleosomes to mitotic chromosome assembly. When mouse sperm nuclei are incubated with an extract depleted of the histone chaperone Asf1, the assembly of octasomes can be suppressed almost completely. Remarkably, we found that even under this “nucleosome-depleted” condition, mitotic chromosome-like structures can be assembled in a manner dependent on condensins (Shintomi et al., 2017, Science). Xenopus sperm nuclei cannot be used in this type of experiment because they endogenously retain histones H3 and H4 and are therefore competent in assembling octasomes even in the Asf1-depleted extract. During this study, we realized that the use of mouse sperm nuclei in Xenopus egg extracts provides additional and deep insights into the basic mechanisms of mitotic chromosome assembly. For instance, the functional contribution of condensin II to chromosome assembly could be observed more prominently when mouse sperm nuclei are used as a substrate than when Xenopus sperm nuclei are used (Shintomi et al., 2017, Science). We suspected that the slow kinetics of nucleosome assembly on the mouse sperm substrate creates an environment in favor of condensin II’s action. For these reasons, our laboratory now extensively uses mouse sperm nuclei for the functional analyses of condensin II (Yoshida et al., 2022. eLife) and other purposes (Kinoshita et al., 2022, JCB). Yoshida et al (2022) used experimental approaches analogous to the current study, and found that the deletion of the CAP-D3 C-tail, causes accelerated loading of condensin II. One of the long-term goals in our laboratory is to critically compare and contrast the actions of condensin I and condensin II in mitotic chromosome assembly. Thus, the use of the same substrate in the two complementary studies can be fully justified.

During the preparation of this response, we realized that the readers would benefit from a brief statement about the comparison between condensin I and condensin II. In the revised manuscript, we have added the following statement in Discussion:

It should also be added that CAP-H2, the kleisin subunit of condensin II, lacks the N-terminal extension that corresponds to the CAP-H N-tail. Thus, the negative regulation by the kleisin N-tail reported here is not shared by condensin II. Interestingly, however, a recent study from our laboratory has shown that the deletion of the CAP-D3 C-tail causes accelerated loading of condensin II onto chromatin (Yoshida et al., 2022). It is therefore possible that condensins I and II utilize similar IDR-mediated regulatory mechanisms, but they do so in different ways.

3. Page 5. "we next focused on the conserved helix (CH) [...], that is enriched with basic amino acids." Based on the provided sequence alignment, the helix contains an equal number of both basic and acidic residues. Is it correct to characterize this helix as positively charged?

Response

The reviewer is right. In the revised manuscript, we have used a more neutral expression as follows:

we next focused on the conserved helix (CH) [...], that contains conserved basic amino acids.

4. To prevent N-tail phosphorylation, the authors create a (H-N19A) allele, referring to Cdk promiscuity. Cdk cooperation with other mitotic kinases can also be expected. Nevertheless, in case the authors created a variant with only the 4 Cdk consensus sites mutated, it would be interesting to know its consequences.

Response

We consider that this is a reasonable question. In our early experiments, we noticed that introduction of multiple SP/TP sites in the non-SMC subunits of condensin I including CAP-H caused a relatively mild phenotype in mitotic chromosome assembly in Xenopus egg extracts. Then we found that the deletion of the CAP-H N-tail caused a very clear, accelerated loading phenotype, prompting us to focus on the regulatory function of the CAP-H N-tail. As the reviewer correctly points out, the current study does not pinpoint the number and position of target sites involved in the proposed phosphoregulation by the CAP-H N-tail. We wish to address this important issue in the near future, along with reconstitution of the phosphoregulation using purified components.

5. Fig EV3A, a second region of mitotic condensin phosphorylation is XCAP-D2. The authors state that XCAP-D2 phosphorylation does not impact on condensin function in their assays. This is very relevant to the current paper, so it would be good to see the Yoshida et al. 2022 Elife publication (in press) as an accompanying manuscript.

Response

We thank the reviewer for pointing out this issue, but it is not necessarily clear to us what the reviewer requests. In the original manuscript, we cited Yoshida et al (2022) in Discussion as follows:

Recent studies from our laboratory showed that the deletion of the CAP-D2 C-tail, which also contains multiple SP/TP sites (Supplementary Figure 3A), has little impact on condensin I function as judged by the same and related add-back assays using Xenopus egg extracts (Kinoshita et al, 2022; Yoshida et al, 2022).

We believe that the current statement is good enough.

6. One of the authors' most striking results is chromosome formation in interphase egg extracts using condensin (H-dN). At the same time, condensin (H-dN) is unable to support DNA supercoiling or chromosome reconstitution with recombinant components. More emphasis might be placed on this important piece of information, and possible reasons should be discussed. Can Cdk-treatment restore condensin (H-dN) biochemical activity? If not, then condensin (H-dN) might have lost more than just an inhibitory N-tail. The cohesin N-tail is thought to fulfil a positive role during DNA loading (Higashi et al. 2020). Could it be that the condensin N-tail encompasses both positive and negative roles?

Response

We were also surprised to find that holo(H-dN) gains the ability to assemble mitotic chromosome-like structures in interphase extracts. It should be stressed, however, that the formation of mitotic chromosome-like structures in I-HSS requires a much higher concentration (150 nM) than the standard concentration used in M-HSS (35 nM). Thus, the deletion of the CAP-H N-tail alone cannot make the condensin I complex fully active in I-HSS. We think that the negative regulation by the CAP-H N-tail is not the sole mechanism responsible for the very tight cell cycle regulation of condensin I function. We emphasize this important point by mentioning that “our results uncover one of the multilayered mechanisms that ensure cell cycle-specific loading of condensin I onto chromosomes” in Summary.

At the end of Discussion, we describe the limitations of the current study: “we have so far been unsuccessful in using these recombinant complexes to recapitulate positive DNA supercoiling or chromatid reconstitution, both of which require proper Cdk1 phosphorylation in vitro”. We are fully aware that full reconstitution of phosphorylation-dependent activation of condensin I in vitro is one of the most important directions in the future.

Although we currently do not have any evidence to suggest that the H N-tail has a positive role, we do not exclude such a possibility.

7. Here comes my main question for the authors (though I should stress that I do not expect an answer for publication in a Review Commons journal). The authors now have a unique opportunity to gain key mechanistic insight into condensin by answering the question, 'how does the kleisin N-tail inhibit condensin'? The authors allude to a model in which the N-tail interacts with Smc2 to close/obstruct the kleisin N-gate, through which the DNA likely enters the condensin ring. Can the authors biochemically recapitulate an interaction between an isolated N-tail (or N-terminal section of XCAP-H) and Smc2? Does Cdk phosphorylation alter this interaction?

Response

This comment is related to Comment #1 of Reviewer#1. See above for our response.

*Minor points. 8. The condensin loop extrusion results would benefit from a supplementary movie or time-series, to illustrate the comparison. Details of how loop rate, duration and sizes were assessed should be added to the methods section. *

Response

In the revised manuscript, we have provided a set of time-lapse images of loop extrusion events catalyzed by holo(WT) and holo(H-dN) in Fig 4E. We have also added the following explanations for how the parameters of loop extrusion reactions were assessed in Materials and Methods:

To determine the loop size, the fluorescence intensity of the looped DNA was divided by that of the entire DNA molecule for each image, and multiplied by the length of the entire DNA molecule (48.5 kb). The loop rate was obtained by averaging the increase in looped DNA size per second. The loop duration was calculated by measuring the time from the start of DNA loop formation until the DNA loop became unidentifiable.

9. Figure EV3A legend, "hHP4" should probably read "hHP2".

Response

The reviewer is right. It should read hHP2. Corrected.

*Reviewer #4 (Significance (Required)):

see above *

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #4

Evidence, reproducibility and clarity

Mitotic chromosome formation is a cell cycle-regulated process that is crucial for eukaryotic genome stability. The chromosomal condensin complex promotes chromosome condensation, but the temporal control over condensin function is only scantly understood. In this impressive manuscript, "Cell cycle-specific loading of condensin I is regulated by the N-terminal tail of its kleisin subunit", Tane and colleagues provide important new insight into the cell cycle-regulation of condensin. The authors identify a kleisin N-tail that acts as a negative regulator of condensin's DNA interactions. Removal of this N-tail, or its cell cycle-dependent phosphorylation, relieves inhibition and activates condensin. This is a simple, yet very important story, that advances our molecular understanding of chromosome formation. The experiments are performed to a very high technical standard and support the authors conclusions. This manuscript is highly suitable for publication in any molecular biology journal, once the authors have considered the following points.

1. Introduction.

   • a) The authors could better explain their own prior work (Kimura et al. 1998), which has identified the condensin XCAP-D2 and XCAP-H as the targets of phosphoregulation. The current manuscript explains the role of XCAP-H phosphorylation.
   • b) Given the limited knowledge about condensin cell cycle regulation, it seems prudent to provide a brief summary of what is known. Fission yeast Smc4 phosphorylation regulates condensin nuclear import (Sutani et al. 1999), while budding yeast Smc4 phosphorylation slows down the dynamic turnover of the condensin complex on chromosomes (Robellet et al. 2015 and Thadani et al. 2018).
   • Extracts were mixed with mouse sperm nuclei. If there is a reason why mouse rather than Xenopus sperm nuclei were used, this would be interesting to know.
   • Page 5. "we next focused on the conserved helix (CH) [...], that is enriched with basic amino acids." Based on the provided sequence alignment, the helix contains an equal number of both basic and acidic residues. Is it correct to characterize this helix as positively charged?
   • To prevent N-tail phosphorylation, the authors create a (H-N19A) allele, referring to Cdk promiscuity. Cdk cooperation with other mitotic kinases can also be expected. Nevertheless, in case the authors created a variant with only the 4 Cdk consensus sites mutated, it would be interesting to know its consequences.
   • Fig EV3A, a second region of mitotic condensin phosphorylation is XCAP-D2. The authors state that XCAP-D2 phosphorylation does not impact on condensin function in their assays. This is very relevant to the current paper, so it would be good to see the Yoshida et al. 2022 Elife publication (in press) as an accompanying manuscript.
   • One of the authors' most striking results is chromosome formation in interphase egg extracts using condensin (H-dN). At the same time, condensin (H-dN) is unable to support DNA supercoiling or chromosome reconstitution with recombinant components. More emphasis might be placed on this important piece of information, and possible reasons should be discussed. Can Cdk-treatment restore condensin (H-dN) biochemical activity? If not, then condensin (H-dN) might have lost more than just an inhibitory N-tail. The cohesin N-tail is thought to fulfil a positive role during DNA loading (Higashi et al. 2020). Could it be that the condensin N-tail encompasses both positive and negative roles?
   • Here comes my main question for the authors (though I should stress that I do not expect an answer for publication in a Review Commons journal). The authors now have a unique opportunity to gain key mechanistic insight into condensin by answering the question, 'how does the kleisin N-tail inhibit condensin'? The authors allude to a model in which the N-tail interacts with Smc2 to close/obstruct the kleisin N-gate, through which the DNA likely enters the condensin ring. Can the authors biochemically recapitulate an interaction between an isolated N-tail (or N-terminal section of XCAP-H) and Smc2? Does Cdk phosphorylation alter this interaction?

Minor points.

1. The condensin loop extrusion results would benefit from a supplementary movie or time-series, to illustrate the comparison. Details of how loop rate, duration and sizes were assessed should be added to the methods section.
2. Figure EV3A legend, "hHP4" should probably read "hHP2".

Significance

see above

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This paper reveals a role of an N-terminal extension of CAP-H in the regulation of condensin-I activity in Xenopus extracts using biochemical reconstitution experiments. The authors demonstrate that a motif in the N-terminal tail that is conserved in vertebrates acts as an inhibitor of condensin I activity. Using several mutant constructs, it is shown that the inhibition by this motif is in turn counteracted by the phosphorylation of neighbouring serine and threonine residues in mitosis, presumably at least in part by CdK. Mutants that have lost this inhibition are able to condense chromatin into chromatid-like structures more efficiently and to some degree even in interphase extracts. Moreover, one such mutant is characterized in detail by biochemical and biophysical experiments and shown to have increased capacity in salt-stable DNA loading and in DNA loop extrusion.

Major comments:

This is a beautiful and thorough study that is presented in a clear and concise manner. The main conclusions are well justified. No additional experiments are needed to support them. Replication and statistical analysis appear adequate. The final model is however largely speculative. Recent work has indicated that loading of yeast condensin does not require gate opening. The authors may thus want to include alternative scenarios or remain more vague.

The H-N19A mutant has a loss of function phenotype (possibly due to folding problem caused by 19 point mutations rather than lack of phosphorylation), the authors could consider to rescue the phenotype by also including the CH motif mutations in this construct (or make an explanatory statement in the text).

Albeit not necessary for the main conclusions, the authors could possibly significantly strengthen their study by testing for binding partners of the N-tail and the CH motif by running AlphaFold predictions against the condensin I subunits.

The efficiency of depletion of condensin subunits from I-HSS extracts is not documented (in contrast to M-HSS extracts - figure EV1C). While any condensin remaining in these extracts might not be active (or interfering), the authors may want to at least comment on this in the text.

The authors should include information on the quantification of chromatid morphology. Is the analysis based on chromatids taken from the same images/imaging session, from technical replicates, biological replicates?

Minor comment:

The colour scheme in Figure 5A is confusing. Use less colour? The orange and red colours are moreover quite similar.

Significance

The findings provide new insights into how condensin-I activity is restricted outside of mitosis. It was previously assumed that this regulation was (largely) due to the exclusion of condensin I from the nucleus prior to nuclear envelope breakdown. This study shows that another pathway is contributing to the regulation and implies that controlling condensin I activity is more important than previously appreciated. Whether all residual nuclear condensin I is inactivated, remains to be determined. The physiological impact of loss of autoinhibition on chromosome segregation and cell cycle progression also remains to be uncovered. The observed effects are robust and appear significant. Future research on condensin I using reconstitution will likely benefit from being able to control or eliminate the self-inhibition.

This reviewer has expertise on the biochemistry and structural biology of SMC protein complexes.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The manuscript reveals that the N-terminal region of CAPH could play a role in regulating condensin I activity, using a range of in vitro methods. They propose that the N-terminal region of CAPH inhibits complex activity, and this is turned off upon deletion or phosphorylation, by using truncations, phospho-mimics or phospho-deficient mutations.

While the results are interesting to the field, and helps to address the question as to how condensin complexes are controlled in a cell cycle dependent manner, some key data and controls are necessary to ensure the conclusion is robust.

Main comments

• What is meant by "unperturbed I-HSS" on page 7, ie membrane containing versus membrane free or condensin depleted?
• In many of the protein gels eg figure 4B, the bands for SMC2 and 4 are more intense that the non-SMC components. The method for protein purification also does not include a size exclusion step to ensure sample homogeneity. Authors should perform some sort of quality control checks on samples such as analytical gel filtration or mass photometry to ensure stoichiometry/homogeneity. This is particularly important for samples eg the N19A, where activity is reduced compared to wild-type as poor protein behaviour could create false negative results.
• Loop extrusion assays in figure 4D-G shows no example data i.e. no pictures or videos of loops being formed. These should also be included.
• Given the mutant holo(H-dN) has higher activity than wild-type, a negative control such as holo(H-dN) without ATP or holo(H-dN) ATPase deficient mutant should also be measured in loop extrusion assays, to ensure the activity is derived from the ATPase activity.
• According to the methods, this work performs the same loop extrusion assay as described in Kinoshita et al, 2022, however, in Kinoshita et al, wild type condensin I makes loops in 30-50% of DNA molecules, where in this study the percentage is less than half that. Can the author please explain the discrepancy given the same method was used?
• In the concluding statement the author suggests "Upon mitotic entry, multisite phosphorylation of the N-tail relieves the stabilization, allowing the opening of the DNA entry gate, hence, the loading of condensin I onto chromosomes." This seems unlikely as fusion the N-terminus of the of the kleisin to the C-terminus of SMC2 is able to function for yeast (Shaltiel et al 2022) and condensin II (Houlard et al 2021), and equivalently in cohesin (Davidson et al 2019)

Significance

This is an interesting story that reveals new insights about condensin regulation.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Comments

The work described in this manuscript starts with an in-silico analysis of the primary amino-acid sequence of CAP-H proteins that reveals the presence in vertebrate orthologs of an N-terminal extension of ~80 amino acids in length which contains 19 serine or threonine residues and also, in its centre, a stretch of conserved basic amino acids predicted to form a helix. These features suggest a regulatory module. Using xenopus egg extracts depleted of endogenous condensins and supplemented with recombinant condensin I holocomplexes, either wildtype or mutants, the authors show that deleting the N-terminal tail of CAP-H, or just the central helix (CH), increases the association condensin I with chromatin in mitotic egg extracts and accelerates the formation of mitotic chromosomes. Interestingly, they also show that deleting the N-tail enables a substantial amount of condensin I to associate with chromatin in interphase extracts and to form chromosome-like structures, while WT condensin I cannot. Using in vitro assays and naked DNA as substrate, the authors further show that removing the N-terminal tail of CAP-H improves both the topological (salt-resistant) association of condensin I with DNA and it loop extrusion activity. These experiments appear to me as are clear and robust. They convincingly reveal that N-tail of human CAP-H hinders the binding of condensin I to DNA and both its loop-extrusion and chromosome-shaping activities. However, the mechanism through which such hindrance is achieved remains elusive (see major comments 1-3).

A complementary part of the work tackles the important question of the cell cycle control of such counteracting effect. Using newly-designed antibodies against two phospho-serine residues within the tail, the authors provide evidence that the tail is phosphorylated in mitosis-specific manner. This points towards phosphorylation as a biological mean to modulate the effect of the tail on condensin's binding during the cell cycle. In support to this idea, using non-phosphorylatable or phosphomimic substitutions of all the serine and threonine residues within the tail (n =19), including one substitution within the CH domain (Ser 70), the authors show that non-phosphorylatable mutations (H-N19A) or phosphomimic mutations (H-N19D) respectively reduce or improve condensin I binding to chromatin in mitotic egg extracts. This suggests that the phosphorylation of the N-terminal tail in mitosis might relieve its negative effect on condensin I binding to chromatin. The weaknesses I see in this part of the study concern (1) the validation of the phospho-antibodies, which appears to me as insufficiently described (major comment 4), (2) the possibility the bulk changes in amino acids (n=19 out of 80) could impact the folding of the central helix (minor comment X) and (3) that some substitutions could impact the binding of condensin I by different mechanisms (minor comment X).

Major comments:

1. On the model. The authors propose that the N-tail could stabilise an interaction between the N-terminal part of CAP-H and SMC2's neck, which would restrain the transient opening of a DNA entry gate within the ring, necessary for the topological engagement of DNA and loop formation. Although the model is sound, I see no direct data that support it in the manuscript. Such model predicts that a CAP-H protein containing or not the N-terminal tail (or the central helix) should exhibit different binding strengths to SMC2 in vitro. It seems to me that the authors could easily test this prediction using the recombinant proteins they produced in the context of this study.
2. On ATP-hydrolysis. Given the importance of ATP hydrolysis for the engagement of condensin into a topological mode of association with DNA and for its loop extrusion activity, I suggest that the authors measure the impact of the N-tail and of the CH domain on the rate of ATP hydrolysis by condensin I holocomplexes. I suppose that it can be relatively easily done (PMID: 9288743) using the recombinant WT and mutant versions they purified in the course of this study.
3. A conundrum with previous work? In Kimura et al. Science 1998 (PMID: 9774278), the lab of Tatsuya Hirano has shown that xenopus condensin I purified from mitotic egg extracts induces the supercoiling of plasmid DNA in vitro, but fails to do so when it is purified from interphase egg extracts. This echoes the inhibitory effect of the N-tail of the topological binding of condensin I described in the current manuscript. However, using a gel shift assay, Kimura et al. 1998 also provide evidence that interphase and mitotic condensin I bind plasmid DNA in vitro with similar efficiencies. At first sight, this prior observation seems to contradict the idea that the N-tail of CAP-H restrains DNA binding unless it is phosphorylated in mitosis. Is it possible that the in vitro binding assays used in Kimura et al. 1998 and in this work might assess different modes of binding? I suggest that this apparent conundrum should to be discussed. Related to that, could it be possible for the authors to assess the impact of the N-tail on the salt-sensitive binding of condensin to DNA, i.e. by reproducing the topological binding assay but omitting the high salt washes? I guess this information could be useful to fully apprehend the impact of the N-tail on the binding of condensin.
4. Validation of phospho-antibodies and by extension showing the phosphorylation of the tail. The newly-designed phospho-serine antibodies used in this study to show that the N-tail is phosphorylated at serine 17 and serine 76 in mitosis (Fig. EV3) are, in my view, not characterized enough. This piece of data is key to substantiate the idea that the tail is phosphorylated in mitosis. Yet, showing that these antibodies detect epitopes on WT condensin I from mitotic egg extracts but not on the H-N19A counterpart, nor on WT condensin I from interphase extracts, does not demonstrate the phospho-specificity of such antibodies. I suggest that a PPase treatment should be conducted to assess the phospho-specificity of these antibodies. Moreover, since the lab of Tatsuya Hirano has the know-how to deplete Cdc2/CDK1 from xenopus egg extract, such strategy could/should be used to further validate the antibodies and assess to which extent the N-tail is phosphorylated in a Cdc2-dependent manner.

Minor comments:

1. The impact of the 19 mutations, A or D, introduced within the tail on the folding of the central helix? The idea that the negative effect of the N-tail is relieved by phosphorylation is based on the chromatin binding phenotypes exhibited by the H-N19D or H-N19A mutant holocomplexes, in which 19 amino-acids out of 80 have been modified, include one in the central helix. The authors also provide evidence that the central helix (CH) located within the tail plays a key role in the negative regulation of condensin I binding. Thus, I wonder to which extent the folding of the central helix could be impacted by the mutations introduced in the tail and notably the one within the central helix itself. Could the author assess the structure of mutated tails using Alpho-fold and/or discuss this point?
2. Phosphorylation of serine 70 in the central helix by Aurora-B kinase? A prior study by Tada et al. (PMID: 21633354) has shown (1) that serine 70 of the N-tail of hCAP-H is phosphorylated by Aurora-B kinase, (2) that the mutation S70A reduces the binding of condensin I to chromatin in HeLa cells and (3) that hCAP-H interacts with histone H2A in an Aurora-B dependent manner. This draws a picture in which the phosphorylation of Ser70 by Aurora-B would improve condensin I binding to chromatin by promoting an interaction between hCAP-H and histone H2A/nucleosomes. Intriguingly, Ser 70 in Tada et al. correspond to the serine residue located within the conserved central helix analysed in this study, and this Ser70 residue is mutated in the H-N19D or H-N19A holocomplexes that show reduced chromatin binding in this study. This raises the question as what could be the contribution of the S70A or S70D substitution to the chromatin binding phenotypes shown by the H-N19D or H-N19A holocomplexes. This is not discussed in the manuscript, and the authors do not cite this earlier work (PMID: 21633354) in their manuscript. Is there any reason for that? I suggest it should be cited and discussed.

3. Other minor comments

   • Please provide a microscope image of DNA loop in Fig. 4D
   • The authors do not compare the kleisin of condensin I with the one of condensin II with respect to the features tackled in this work. Given the different behaviours condensin I and II, such comparison could be informative for the readers.
   • The authors do not reference the work of Robellet et al. Genes & Dev (2015) (PMID: 25691469) on the regulation of condensin binding in budding yeast by an SMC4 phospho-tail. I suggest that the analogy should be discussed.
   • In the introduction section, lane 5, the sentence "Many if not all eukaryotic species have two different condensin complexes" appears inappropriate since budding and fission yeast cells possess a single condensin complexes, similar to condensin I in term of primary amino-acid sequence.
   • page 4; typo: motif I and V bind to the SMC neck and the SMC4 cap regions, respectively. Should read SMC2 neck.

4. Are the data and the methods presented in such a way that they can be reproduced? YES

5. Are the experiments adequately replicated and statistical analysis adequate? YES
6. Are prior studies referenced appropriately? Not all of them (see above)
7. Are the text and figures clear and accurate? YES

Referees cross-commenting

I consider the comments from all reviewers as helpful for the authors.

Significance

Summary

Condensins are genome organisers of the family of SMC ATPase complexes and are best characterized as the drivers of mitotic chromosome assembly (condensation). It is acknowledged that condensins shape mitotic chromosomes by massively associating with DNA upon mitotic entry (loading step) and by folding chromatin fibres into arrays of loops, most likely through an ATP-dependent extrusion of DNA into loops, as seen in vitro. What remains unclear, however, are the mechanisms by which condensins load onto DNA and fold it into arrays of loops in vivo, and how these reactions are coupled with the cell cycle, i.e. restricted mostly to mitosis. Condensins are ring shaped pentamers that change conformation upon ATP-hydrolysis. In vitro studies suggest that condensins bind DNA via ATP-hydrolysis-independent, direct electrostatic contacts between condensin subunits and DNA. Such electrostatic contacts are salt-sensitive in in-vitro assays. Upon ATP-hydrolysis, condensins engage into an additional mode of binding that is resistant to high salt concentration and likely to be topological in nature. Both modes of association are necessary to form DNA loops. Vertebrates possess two types of condensin complexes, condensin I and II, each composed of a same SMC2-SMC4 ATPase core but associated with two different sets of three non-SMC subunits; a kleisin and two HEAT-repeat proteins. Condensin II is nuclear during interphase and stably binds chromatin upon mitotic entry, while condensin I is located in the cytoplasm during interphase and binds chromatin in a dynamic manner upon nuclear envelope breakdown. How the spatiotemporal control of condensin I and II is achieved remains poorly understood. Previous studies have shown that the phosphorylation of condensin I by mitotic kinases, such as CDK1, Aurora-B and Polo, play a positive role in its binding to chromatin and/or its functioning, but the underlying mechanisms remain to be characterised. In this manuscript, Shoji Tane and colleagues provide good evidence that the N-terminal tail of the human kleisin subunit of condensin I, hCAP-H, serves as a regulatory module for the cell-cycle control of condensin I binding to chromatin and chromosome shaping activity. The authors clearly show that the N-tail of CAP-H hinders the binding of condensin I to chromatin in xenopus egg extracts and, using in vitro assays, that the N-tail also hinders the topological association of condensin I with DNA and its loop extrusion activity. The authors provide additional data suggesting that the phosphorylation of the N-tail of CAP-H, in mitosis, relieves its inhibitory effect on condensin I binding. Based on their findings, Tane et al. propose a model suggesting that the N-terminal tail of CAP-H constitutes a gate keeper that maintains condensin-rings in a closed conformation that is unfavourable for topological binding to DNA, and whose locking effect is relieved in mitosis by phosphorylation.

Taken as a whole, this work has the potential to reveal a molecular basis for the cell cycle regulation of condensin I in vertebrate cells and as such to significantly improve our understanding of the integrated functioning condensin I. The characterisation of the inhibitory effect of the N-tail on condensin binding to chromatin and to naked DNA in vitro is well described, the data are clear and robust and the results convincing. On the other hand, some of the data on the phospho-regulation appear to me as are more debatable and I think that some of the results described here deserve to be discussed in the context of previous works. Finally, I see no data in the manuscript that directly supports the mechanistic model proposed by the authors, while it seems to me that such model could have been easily tested exprimentally. Thus, I suggest that Tane and colleagues should perform a couple of relatively easy experiments to strengthen their claims and that a few omitted prior studies on the topic should be referenced and discussed.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We really appreciate the reviewers’ insightful comments, which help improve the quality of this work. We have responded to the reviewers’ questions/comments point by point in the following text and made the corresponding changes in the revised manuscript. Lastly, we added one more figure (Fig. 7) with lineage tracing experiments demonstrating the conversion of id2a+ liver ductal cells to hepatocytes in extreme hepatocyte loss condition.

Reviewer #1 (Evidence, reproducibility and clarity):

Mi and Andersson describe a method for creating efficient 3' knock-ins in zebrafish using a combination of end-modified dsDNA and Cas9/gRNA RNPs. They tested their method on four genetic loci where they introduced Cre recombinase endogenously, and obtained high F0 mosaicism and germline transmission. The authors included fluorescent proteins with self-cleaving peptides to determine that endogenous expression patterns are observed. By crossing their knock-in Cre lines with lineage tracing reporter lines, the authors temporally traced lineage divergences in zebrafish liver and pancreas.

The authors should clarify the following points before I can recommend publication:

Overall, I suggest that the authors consider paring down their figures. Throughout the paper, multiple figure panels convey the same point but for different genes. Furthermore, many construct configurations are shown that are not used in the subsequent panels. For example, the mNeonGreen only (no Cre) constructs and the EGFP constructs are largely not used in downstream experiments. The authors could pick the important constructs and show the relevant data, and summarize all their other constructs in one supplementary figure. The authors also jump around in different parts of the paper with regards to using iCre or CreERT2 and ubi:Switch or ubi:CSHm. It's not clear to me why they're doing that? It makes the paper hard to follow. For example, why use iCre - it's not temporal if I understand correctly (and I'm not sure what improved Cre is - could they reference a paper and include a small explanation) so CreERT2 seems suitable especially for their temporal lineage tracing experiments. Why not limit the description to CreERT2 in the main text/figures? Also, isn't ubi:Switch and ubi:CSHm pretty similar except the latter is nuclear mCherry due to H2B? Why not only focus on ubi:CSHm experiments? I found the paper to be unnecessarily long and think it would benefit from editing to describe the most important concepts and experiments.

Response: Thank you for your constructive and helpful comments. We do agree that sometimes the schematic constructs seem redundant. This is because the krt4, nkx6.1, and id2a genes have similar gRNA targeting sites (all spanning over the stop codon). However, we prefer to keep these schematic constructs as we have all the statistical results showing the knock-in efficiency in the subsequent figure panels. Such layout can allow readers to make comparisons and better understand the efficacy of this method. However, combined with the comments from the second reviewer, we indeed need to add more detailed information, including the sequence and the length of the short left and right homologous arms in the schematics, to enable the readers to follow this strategy more easily. Meanwhile, we added a new supplementary figure with the sequences of the long left and right homologous arms, as well as the genetic cassettes/point mutations for krt92 knock-in (Figure EV1).

As for the color switch lines we used, we appreciate your comments and replaced Fig. 5E-G with new fluorescent images using zebrafish larvae carrying the ubb:CSHm transgene. For most of the lineage tracing experiments in this study, we used Tg(ubb:CSHm) as the H2BmCherry is more stable, located in the nucleus, and the fluorescence intensity is stronger than in Tg(ubb:Switch). However, for the lineage tracing experiments in the liver injury model, we believe that Tg(ubb:switch) is a better option than Tg(ubb:CSHm). In the absence of a hepatocyte specific far-red reporter line, we can distinguish the hepatocytes derived from the id2a+ origin using the Tg(ubb:Switch) line, as the cells with Cre recombination express mCherry in the cytoplasm; i.e. we can tell the cell types based on the cell morphology in combination with the ductal anti-vasnb staining. This strategy was previously used by Dr. Donghun Shin’s group in their 2014 Gastroenterology paper (Figure 4B, DOI: 10.1053/j.gastro.2013.10.019). Therefore, we still kept the ubb:switch in the Fig. 1F schematic, and we have elaborated on why we chose Tg(ubb:switch) line for the id2a+ cell conversion experiments in Fig. 7 and Figure EV14.

The iCre we used is a codon-improved Cre (iCre). The original cDNA sequence was from pDIRE (Addgene plasmid #26745; provided by Dr. Rolf Zeller, University of Basel) (Osterwalder et al., 2010).

At the beginning of this project, we actually didn’t know whether there were any differences between iCre and CreERT2 in labelling of the cells of interest. Here, using both the iCre and CreERT2 lines, we for the first time, formally show the developmental lineage path of nkx6.1-expressing cells in the zebrafish pancreas. Our data suggested that the early nkx6.1-expressing cells are multipotent pancreatic progenitors giving rise to all three major cell types in the pancreas (endocrine, ductal and acinar cells, shown by nkx6.1 knock-in iCre) and gradually the nkx6.1-expressing cells become restricted in the ductal/endocrine lineages (shown by the nkx6.1 knock-in CreERT2 treated with 4-OHT at different timepoints). In addition, we also aim to use these knock-in lines for multiple studies in which we need to perform many quantitative experiments. As expected, we are unable to reach 100% labeling using the knock-in CreERT2 lines, even if we treated the larvae with very high concentration of 4-OHT over a long period of time. This means that the CreERT2 induced recombination will introduce more variation for quantitative experiments (for instance, the number of regenerated beta-cells from the ductal origin). As we were quite confident with the efficiency of this knock-in strategy, we decided to make both iCre and CreERT2 lines in krt4, nkx6.1, and id2a locus and just observe how they performed. We often use iCre knock-in lines for lineage tracing experiments, because the iCre lines reach near 100% labeling efficiency. Such iCre lines are particularly useful if they only label terminally differentiated cell types. Thus, the near 100% labeling efficiency in iCre lines can be of great help for initial experiments, which later can be confirmed by temporal labeling using CreERT2 lines.

1. Could the authors describe the purpose of the 5'AmC6 modification earlier in the paper? I didn't see much text about it until the discussion. It seems that the speculation is that it provides end protection and prevents degradation (based on in vitro studies in human). This should be inserted into the introduction as a reader might be wondering about this and won't find an answer until near the end. Also, is this the first in vivo use of this modification for knock-ins? If so, that should be highlighted in the text.

Response: This is a helpful comment. In the revised manuscript, we elaborate more on why we chose 5'AmC6 modification in our donors. To our knowledge, this is the first time this 5’ modification is used in vivo, however, bulky 5’modification (5'Biotin - 5x phosphorothioate bonds) has been used in medaka (DOI: https://doi.org/10.7554/eLife.39468.001, 2018 Elife, as we previously referenced). The cell division rate is much faster in zebrafish embryos compared with medaka embryos during early development, so we speculate that such modification might be of more importance in zebrafish to achieve early integration. Another advantage is that the 5'AmC6 modification is commercially available, allowing researchers to prepare the donor dsDNA in a handy fashion. We have now expanded on these details and advantages in the introduction.

1. The authors do not show any sequencing data confirming that their insert was knocked-in as designed with no disruption to the immediate upstream and downstream endogenous sequences. Can they sequence the loci to confirm?

Response: This is indeed a question we frequently get – thank you for making us relay this information more clearly! We have put the raw Sanger sequencing data in a public repository (mentioned in the Data Availability section), and included the sequencing primers in the method paragraph. Now we also refer to this data in the discussion section in conjunction to highlighting that the integrations were correctly placed in the loci. If you think there are better ways to show the sequencing results, please let us know.

1. I found the descriptions of the long and short HA to be confusing when describing the results, especially since the first tested gene krt92 only has long and all subsequent ones are short. The discussion made it more clear that short HA is more efficient and applicable when gRNAs span the stop codon. Perhaps that wasn't possible with krt92, but the authors could prevent the confusion by clearly stating the design requirements of long and short HA and that they wanted to test which is more efficient before starting to describe the data. I also didn't see a description of what the length difference between long and short HA is? How short is short HA?

Response: This is a great question that is well worth discussing. In the revised manuscript, we changed the order in which the parts are described, with nkx6.1 knock-in in front of krt4 knock-in. Here we explain why we would like to do that:

At the beginning of this project, we did not know if the 5’ modified dsDNA could be an effective donor. To test our hypothesis, we chose the krt92 gene as our first target, as this is a keratin protein and expressed in the epithelial cells. We can easily detect the fluorescence in the epithelial cells (most notably in the skin), which allow us to sort the F0 mosaic embryos with high percentage of integration. Notably, from our experience, the most difficult part of the knock-in method is the sorting step (usually performed during 1-3 dpf). This is because the fluorescence signal is highly dependent on the endogenous gene expression level and is usually dimmer with an overall integration efficiency that is lower compared to canonical transgenesis. Therefore, we thought that targeting an epithelial cell marker would be informative and help us to evaluate the validity and reproducibility of the method. If it worked, then we could move on targeting genes expressing in more restricted tissues or cell types. For krt92 gene, the gRNA targets the region upstream of the stop codon. To prevent the cleavage of the donor template, we had to introduce several point mutations and at the same time keep the amino acid sequence intact. However, such mutations can restrict the knock-in and lower the integration efficiency when using shorter arms (due to the sequence mismatch).

After we managed to make the krt92 knock-in, our next question was, what about using a gRNA spanning over the stop codon region? In this way, we don’t need to introduce point mutations on neither the left nor the right homologous arm. Also, for the purpose of our biological study, the nkx6.1 were on top of our gene list for lineage tracing experiments and we luckily identified that there is very good gRNA targeting this locus. After we successfully made the nkx6.1 knock-in, we were thinking that we could simplify the protocol even further, i.e. switching to short homologous arms so that we can prepare the donor by a one-step PCR instead of making complicated constructs. We tested that hypothesis in nkx6.1, krt4, and id2a sites and obtained very promising efficiency. Also, we did some further testing with dsDNA without the 5’ modifications and showed that the 5’ modifications indeed greatly increased integration efficiency. Therefore, although the short homologous arm method is a highlight here, we also point out that it was not planned from the beginning. In the revised manuscripts, we want to convey our method in a logical way and show how we modify the method in a step-by-step fashion.

Moreover, with regards to the comments from the second reviewer, we now added the length of the homologous arms as well as the mutation site on the schematics. We chose short homologous arm because in previous literature it was suggested that short homologous arms (36-48 bp, which we now write out in both the results and the methods) can promote microhomology-mediated end joining (doi: 10.1096/fj.201800077RR). We also noticed that the recent Geneweld method (DOI: 10.7554/eLife.53968) also adheres to a similar length for homology mediated integration. In this study, HAs even shorter than 36 bp also perform well.

1. The authors state that they could not use in situs to confirm krt92 endogenous and knock-in expression overlap, but rather say that they match based on data from an intestine scRNA-seq dataset. Can they elaborate on this? Which clusters/cell types show overlap? Furthermore, is there any krt92:GFP transgenic line that can be used as a reference for expression as well? This point is also applicable for krt4 described in Fig.2

Response: We appreciated this point. In the beginning, we contacted Molecular Instruments to synthesize krt92 HCR3.0 in situ hybridization probes. However, the technical staff there told us that they are unable to make specific probes due to high sequence similarity to other keratin protein families. We can see that the sequence similarity mostly occurs in the middle of krt92 genes, and the HCR3.0 probes rely on a probe set (preferably 20-30 probes with different sequences) to target the mRNA.

The scRNA-seq data that we referenced are from 10X platform, which is based on a 3’enrichment methodology. The reads mapping to krt92 genes are mostly located on the 3’ end. This is good as there is much less similarity to other cytoskeleton genes in the 3’ end of the gene. Unfortunately, there is no krt92 transgenic lines available, so we relied on the single-cell data to correlate expression patterns in this case.

There are two zebrafish intestine single-cell data sets available, with the following links:

(1): https://singlecell.broadinstitute.org/single_cell/study/SCP1675/zebrafish-intestinal-epithelial-cells-wt-and-fxr?genes=krt92#study-visualize

(2): https://singlecell.broadinstitute.org/single_cell/study/SCP1623/zebrafish-intestine-conventional-and-germ-free-conditions?genes=krt92#study-visualize

We can see that krt92 is widely expressed in different types of intestinal epithelial cells (absorptive enterocytes, secretory enteroendocrine/goblet cells and ionocyte).

For the krt4 gene, we now added the HCR3.0 in situ hybridization and immunofluorescence for both krt4 knock-in EGFP-t2a-CreERT2 lines and the Tg(krt4:EGFP-rpl10a) transgenic line (a construct from Anna Huttenlocher, https://www.addgene.org/128839/, which has been widely used to label skin cells). The results are shown in Figure EV9. We show that krt4 has very high expression in the intestinal bulb and hindgut based on the HCR3.0 in situ. The Immunofluorescence of the krt4 knock-in fully recapitulate the krt4 expression pattern in the intestine, while there is almost no fluorescence signal in Tg(krt4:EGFP-Mmu.Rpl10a). We believe this is another advantage of using the knock-in method, over transgenics, for cellular labeling and lineage tracing. Classical transgenics often rely on short promoters of the proximal/enhancer region upstream of ATG with various length (arbitrarily or based on clues from motif analysis/DNA methylation sites). However, different tissues/cell types tend to use different cis-_regulatory elements and the chromatin structure/enhancer-promoter loops might differ dramatically among different cell types. It is hard to predict the exact region of the regulatory sequences that is sufficient for driving the gene expression in a certain cell type. Thus, such reasoning consolidates with that our knock-in lines recapitulate the endogenous _krt4 gene expression. Therefore, we believe that the knock-in based genetic lineage tracing will become the standard in the zebrafish field, as theoretically it avoids both the lack of relevant expression and leakage problems of transgenics.

1. I think Figure 2A needs the dotted lines on the last construct to be fixed (points to p2A)

Response: Thank you for noticing! This was due to a bug in the IBS software, and we changed it manually using Adobe Illustrator in the revised manuscript.

1. There are a few instances where the authors describe performing 4-OHT treatment for long period (e.g. over a 20 hour or 24 hour period). Is fresh 4-OHT added after a certain amount of time or is it a one-time addition? Is such long periods of 4-OHT required or has maximal recombination already occurred within a few hours after addition of 4-OHT?

Response: For 4-OHT treatment, we referred to the method described by Dr. Christian Mosimann (DOI: 10.1371/journal.pone.0152989). We actually tried different conditions (dosage, duration, refresh or not). This is particularly important for the knock-in CreERT lines because the level of CreERT2 is highly dependent upon the endogenous gene expression level. In our case, the nkx6.1 and id2a are transcriptional regulators and relatively lowly expressed compared with structural proteins. We maximized the labeling efficiency by using the highest concentration and longest duration suggested for 4-OHT treatment. The 4-OHT was stored in -20 ℃ and it would become less effective after 30 days of storage. Therefore, we first incubated the 4-OHT in 65 ℃ for 10 min (as recommended by Dr. Christian Mosimann) in order to convert it to a bioactive form. Next, we treated the zebrafish embryos with 4-OHT using a final concentration of 20 μM for 24 hours. We didn’t refresh the 4-OHT since there was no significant difference compared with a one-time addition. Moreover, using higher dosage or longer treatment time can lead to less survival and increased deformity rate. 20 μM 4-OHT treatment for shorter time periods (6 or 12 hours) can cause high labeling variability (some larvae have good labeling while others not). In the end, after several rounds of experiments, we settled on 20 μM 4-OHT treatment for 24 hours as it can reach the highest labeling efficiency, lower variability, and good survival.

1. For Figures 4-6 where confocal images of lineage tracing experiments are shown, there is no indication of how many times the experiments were repeated, how many sections were images, how many animals used, how many cells counted. All of this information should be included in the figure legends and plots should be added showing quantification and statistical analysis (where appropriate).

Response: The reviewer makes a good point and we have now added the number of larvae used and statistical results for the quantitative experiments. The quantification of experiments in Figure 3E-H (originally Figure 4E-H) are shown in Figure EV6D using box/dotplot. We randomly selected 3 secondary islets of different sizes (large, middle, and small) from each juvenile fish (n=5) and pooled the number of mCherry/ins double positive cells and ins positive cells together. The quantification of the lineage-tracing efficiency in the experiments in Figure 6 are shown in Figure EV13.

1. Figure 4 C, C' - I'm not sure what to look for. Is the message that there is no Cherry positive cells that are vasnb negative when labelling is done at 8 somite? But the vasnb positive cells that are also Cherry positive remain? The vasnb staining seems much weaker/harder to see in C C' compared to B, B'. As mentioned above, these data should be quantified and statistical significance indicated.

Response: Thank you for pointing this out; the second reviewer made a similar point. We redid the experiments using zebrafish larvae carrying the ptf1α:EGFP transgene to indicate the acinar cells (Figure 3B-D, Figure EV4G). We also quantified the results and performed statistical testing.

1. I recommend the authors include a short section in the discussion comparing the efficiency of their method to other knock-in strategies used in zebrafish. This is an important claim of the paper yet it is not clear how much better it is (if at all) in terms of frequency of F0 mosaicism and identification of founders relative to other methods. I do appreciate the relative simplicity of the molecular steps of construct design/generation.

Response: This is indeed important. It is also tricky since we are unable to make head-to-head comparisons between different methods as we are targeting different genetic loci and do not have the other methods up and running in our lab. However, the general comparison is based on the statistics shown in the hallmark papers describing these other methods, regardless of which genes were selected for targeting. In the discussion, we added a list of points that are novel/improved with our method versus previous ones, including that: 1) we simplify the knock-in methodology circumventing complicated molecular cloning; 2) we have very high germline transmission rate, which means that one morning of injection is often enough to get a founder; and the expression of fluorescence proteins avoids tedious work in identifying founders, which also saves a lot of space in the fish facility; 3) our lines can be applied for multiple utilities; 4) the method does not disrupt the endogenous gene product. We believe this is critical for the field of developmental biology, regenerative medicine, and disease modeling in zebrafish – and perhaps a similar 3’ knock-in based lineage-tracing method can become commonly used to delineate the cell differentiation and plasticity during homeostatic and diseased conditions in additional organisms.

Reviewer #1 (Significance):

Overall, the study contributes a new knock-in strategy in zebrafish that appears to be more user-friendly and results in high germline transmission. The authors also identify nkx6.1+ ductal cells as progenitors of endocrine cells in the pancreas highlighting the biological applications of their method. I think this study represents an important advancement in zebrafish genetics and will have future impact in lineage tracing during development, regeneration, and disease.

Reviewer #2 (Evidence, reproducibility and clarity):

Summary:

Here, the authors present a strategy where they performed knock-in at the level of the STOP codon, taking care of not perturbing the coding region. They integrate cassettes coding for fluorescence protein and Cre recombinase, which are separated from the endogenous gene and each other by two self-cleavable peptides.

The cassettes are done by PCR with primers with 5' AmC6 modifications and they test short (36 to 46 bp) or long homologous arms (~950bp). For nkx6.1 gene, they observed a dramatic increase of recombination efficiency when injecting the donors with short Homology arms compared to long arms suggesting that short arms could be used. Indeed, short arms used with krt4 and id2a allow them to obtain K.I lines.

The techniques described here look promising. Indeed, even if the proportion of F0 showing adequate reporter expression is low (usually about 2%), the percentages of founders among these mosaic F0 were quite high (between 50% and 100%). And this is the most important aspect as it is usually the most time-consuming aspect of the work.

Major comment:

The authors claim that the knock-in lines can precisely reflect the endogenous gene expression, as visualized by optional fluorescent proteins. But are the authors sure that the integration of the cassettes coding for fluorescence protein and Cre recombinase, which are separated from the endogenous gene and each other by two self-cleavable peptides, will not affect the level of expression of the targeted genes . Indeed, it has been shown that sometimes self-cleavable peptides could affect the expression of the genes of the cassette like for example in this reference ([https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8034980]. Therefore it is important that the authors check whether the cassette affect the level of expression of the targeted gene if they want to claim that the knock-in lines precisely reflect the endogenous gene expression.

Response: Thank you for your insightful comments. With regards to the endogenous gene expression, we now use qPCR for further validation. We added the qPCR results to the supplement material (Figure EV15) in the revised manuscript. In brief, we pooled 4 larvae in one tube per biological replicate and have 4 biological replicates for each knock-in line. We didn’t see a significant change in the endogenous expression for any gene. In addition, we have grown up homozygous knock-in lines to adulthood and they are fertile without any overt phenotype.

The highlighted reference is dealing with a cardiomyocyte specific transgenic line, and we assume figure 3-Supplementary figure 1 is what the reviewer is referring to. The altered level of erbb2 expression might be due to the experimental conditions (no treatment or 3 days post treatment). Also, it is possible multiple transgenic insertions occur, as well as gene silencing at some insertion sites. However, such issues would not present, or very limited, with knock-in methods.

Minor comments:

General points:

I believed that the authors should improve the presentation of their data. Indeed, based on what they present, it would be impossible for me to reproduce their technique. Indeed, it is not clear at all how they design the short and long arm, where they are exactly located, which mutations they have done (for fig1), where is located the guide RNA compared to the STOP codon and the HA arms. Graphics that exactly place all these sequences are absolutely required to understand the strategy used and should be placed in figure 1, 2, 3 and 4.

Response: Thank you for these comments. In the revised version, we added the sequence information of the short homologous arms in each of the schematics. As for the krt92 gene, we added the sequence information in the first supplement results (Figure EV1) with the genetic cassettes and point mutation information. We list all the primer information in the methods. Also, we have uploaded our vector templates in the public repository (as listed in the Data availability section). Lastly, we added a key resource table in the supplement file with all the detailed information of reagents for the ease of reproducibility (including all the primers sequences used). We are also willing to share our constructs with the scientific community upon request.

Specific points:

Introduction:

"In zebrafish, the NHEJ-mediated methods have been intensively investigated in 5'knock-in upstream of ATG using donor plasmid containing in vivo linearization site flanking the insertion sequences (11,12,17-20). The 3' knock-in method has also been examined using circular plasmid as the donor with either long or short homologous arms (HAs) flanked by in vivo linearization sites (14, 21-23). Recently, intron-based and exon-based knock-in approaches have remarkably expanded the knock-in toolbox by targeting genetic loci beyond the 5' or 3' end (8-10,13,24-26)."<br /> This part should be explained better in order that the readers could really understand the differences between these old studies and this new one. And really insist on what is the novelty of their technique.

Response: Good points. In the revised version, we elaborated more on the previous discoveries, the major challenges, the knowledge gap in zebrafish knock-in methodology, and what is novel and improved with our new technique. Please, see clarifications and the expanded text in both the introduction and discussion.

Results:

Page 4: To my opinion, the first paragraph should be removed and the technique directly explained based on krt92 strategy as this paragraph does not allow to understand the technique. As indicated above, figure 1 should indicate more clearly the location of the long arms and which mutations they have done and where is located the guide RNA.

Figure 1G: The expression in the skin is far from obvious and the image should be improved (for example with some inset).

Response: Thank you for the comments. We added a new supplementary figure (Figure EV1) and show the sequences of left and right homologous arms, the genetic cassettes, as well as the point mutations with different background color highlight. We added the insets to show the magnified regions of interest. Also, we added the images from the fluorescent microscope used for sorting, to show the EGFP signals in live zebrafish embryos (Figure EV2D and Figure EV8D).

Figure 3E: The authors say that "cells expressing nkx6.1 (displayed by the green fluorescence) were located on the ventral side of the spinal cord whereas H2BmCherry positive cells, which include all the progenies of nkx6.1+ cells after the iCre recombination, resided in both the ventral and dorsal parts of spinal cord". This differential expression in the spinal cord is not obvious and a more closer view should be provided.

Response: Thank you for the comment. First, we changed the order and now describe all nkx6.1 content in Figure 2 and 3 and the krt4 content in Figure 4. We added insets to show the magnified regions and better display the expression pattern of the two fluorescence proteins in Figure 2E-G. One can now clearly see from the magnified insets that the green signals driven by the endogenous nkx6.1 gene are present in the ventral part of the spinal cord, while the red signals are present in both the ventral and the dorsal side of the spinal cord.

Fig S4H: The authors say that" using lineage tracing, we could trace back all three major cell types in the pancreas (acinar, ductal and endocrine cells) to nkx6.1 lineage (Figure 3H-H',Supplementary Figure S4G, H)". While this is obvious for endocrine, the colocalisation with ela3l:GFP is not obvious and the figure should be improved.

Response: This is a very good point, and the first reviewer gave similar suggestions. In the revised version (shown in Figure EV4H and I), we added the insets to show the magnified regions to better display the expression pattern of two fluorescence proteins. The ela3l reporter line is using a short promoter to drive the expression of H2B-EGFP (doi: 10.1242/dmm.026633). However, this short promoter cannot reach 100% labeling of acinar cells, so we also use the ptf1α:EGFP transgene for further validation (new Figure EV4G). Both transgenic reporter lines showed many EGFP and mCherry double-positive cells, indicating that these acinar cells are derived from a nkx6.1-expressing origin. Here we did not use the anti-GFP antibody, as our color switch lines contains CFP and anti-GFP antibody can also recognize CFP. However, the GFP signal is strong enough to show the expression. We hope the additional experiments and insets clarifies this point.

Page 8: the authors say that "The immunostaining at 6 dpf showed that both intrapancreatic ductal cells and a portion of acinar cells can be lineage traced when the 4-OHT treatment started at the 6 somite stage (Figure 4B and B'). The identification of the acinar cells has been done based on the absence of the ductal marker vasnb. To trace efficiently the acinar cells, this should be done with an acinar marker.

Response: Another good point also mentioned by reviewer one. We redid the analyses using zebrafish larvae containing the ptf1α:EGFP transgene to indicate the acinar cells and the co-expression pattern with the lineage-tracing (the data is shown in new Figure 3B-D).

Reviewer #2 (Significance):

I do not have enough expertise in the KI field to evaluate whether this strategy is really novel and as mentioned above, the authors should better explain what is really the novelty of their strategy.

Response: In our answers to the comments of the first reviewer, we elaborated more on the points that are novel/improved with our method vs previous methods, as reiterated here:

“…including that: 1) we simplify the knock-in methodology circumventing complicated molecular cloning; 2) we have very high germline transmission rate, which means that one morning of injection is often enough to get a founder; and the expression of fluorescence proteins avoids tedious work in identifying founders, which also saves a lot of space in the fish facility; 3) our lines can be applied for multiple utilities; 4) the method does not disrupt the endogenous gene product.”

Moreover, the first reviewer asked about the difference between the krt4 knock-in and krt4 transgenics, and based on the in situ data, we showed that our krt4 knock-in can fully recapitulate the endogenous gene expression, while the krt4 transgenics can hardly label the intestinal bulb and hindgut. This might be due to that different tissues/cell types may depend on different _cis-_regulatory elements to drive the gene expression. The chromatin structure and the enhancer/promoter loop might also differ dramatically among different tissues. Therefore, the transgenics might be useful for one type of cells, while they might be not useful at all for other cell types. In the future, we believe that, similar to the mouse field, the 3’ knock-in based lineage tracing methods might become the standard method in the zebrafish field, to delineate cellular differentiation and plasticity during homeostatic and diseased conditions.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

Here, the authors present a strategy where they performed knock-in at the level of the STOP codon, taking care of not perturbing the coding region. They integrate cassettes coding for fluorescence protein and Cre recombinase, which are separated from the endogenous gene and each other by two self-cleavable peptides.<br /> The cassettes are done by PCR with primers with 5' AmC6 modifications and they test short (36 to 46 bp) or long homologous arms (~950bp). For nkx6.1 gene, they observed a dramatic increase of recombination efficiency when injecting the donors with short Homology arms compared to long arms suggesting that short arms could be used. Indeed, short arms used with krt4 and id2a allow them to obtain K.I lines.<br /> The techniques described here look promising. Indeed, even if the proportion of F0 showing adequate reporter expression is low (usually about 2%), the percentages of founders among these mosaic F0 were quite high ( between 50% and 100%). And this is the most important aspect as it is usually the most time-consuming aspect of the work

Major comment:

The authors claim that the knock-in lines can precisely reflect the endogenous gene expression, as visualized by optional fluorescent proteins. But are the authors sure that the integration of the cassettes coding for fluorescence protein and Cre recombinase, which are separated from the endogenous gene and each other by two self-cleavable peptides, will not affect the level of expression of the targeted genes . Indeed, it has been shown that sometimes self-cleavable peptides could affect the expression of the genes of the cassette like for example in this reference (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8034980/) . Therefore it is important that the authors check whether the cassette affect the level of expression of the targeted gene if they want to to claim that the knock-in lines precisely reflect the endogenous gene expression

Minor comments:

General points :

I believed that the authors should improve the presentation of their data. Indeed, based on what they present, it would be impossible for me to reproduce their technique. Indeed, it is not clear at all how they design the short and long arm, where they are exactly located, which mutations they have done (for fig1), where is located the guide RNA compared to the STOP codon and the HA arms. Graphics that exactly place all these sequences are absolutely required to understand the strategy used and should be placed in figure 1, 2, 3 and 4.

Specific points :

Introduction :

"In zebrafish, the NHEJ-mediated methods have been intensively investigated in 5'knock-in upstream of ATG using donor plasmid containing in vivo linearization site flanking the insertion sequences (11,12,17-20). The 3' knock-in method has also been examined using circular plasmid as the donor with either long or short homologous arms (HAs) flanked by in vivo linearization sites (14, 21-23). Recently, intron-based and exon-based knock-in approaches have remarkably expanded the knock-in toolbox by targeting genetic loci beyond the 5' or 3' end (8-10,13,24-26)."<br /> This part should be explained better in order that the readers could really understand the differences between these old studies and this new one. And really insist on what is the novelty of their technique.

Results :

Page 4 : To my opinion, the first paragraph should be removed and the technique directly explained based on krt92 strategy as this paragraph does not allow to understand the technique. As indicated above, figure 1 should indicate more clearly the location of the long arms and which mutations they have done and where is located the guide RNA.

Figure 1G : The expression in the skin is far from obvious and the image should be improved (for example with some inset) .

Figure 3E : The authors say that "cells expressing nkx6.1 (displayed by the green fluorescence) were located on the ventral side of the spinal cord whereas H2BmCherry positive cells, which include all the progenies of nkx6.1+ cells after the iCre recombination, resided in both the ventral and dorsal parts of spinal cord". This differential expression in the spinal cord is not obvious and a more closer view should be provided.

Fig S4H: The authors say that" using lineage tracing, we could trace back all three major cell types in the pancreas (acinar, ductal and endocrine cells) to nkx6.1 lineage (Figure 3H-H',Supplementary Figure S4G, H)". While this is obvious for endocrine, the colocalisation with ela3l:GFP is not obvious and the figure should be improved.

Page 8: the authors say that "The immunostaining at 6 dpf showed that both intrapancreatic ductal cells and a portion of acinar cells can be lineage traced when the 4-OHT treatment started at the 6 somite stage (Figure 4B and B'). The identification of the acinar cells has been done based on the absence of the ductal marker vasnb. To trace efficiently the acinar cells, this should be done with an acinar marker.

Significance

I do not have enough expertise in the KI field to evaluate whether this strategy is really novel and as mentioned above, the authors should better explain what is really the novelty of their strategy.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Mi and Andersson describe a method for creating efficient 3' knock-ins in zebrafish using a combination of end-modified dsDNA and Cas9/gRNA RNPs. They tested their method on four genetic loci where they introduced Cre recombinase endogenously, and obtained high F0 mosaicism and germline transmission. The authors included fluorescent proteins with self-cleaving peptides to determine that endogenous expression patterns are observed. By crossing their knock-in Cre lines with lineage tracing reporter lines, the authors temporally traced lineage divergences in zebrafish liver and pancreas.

The authors should clarify the following points before I can recommend publication:

1. Overall, I suggest that the authors consider paring down their figures. Throughout the paper, multiple figure panels convey the same point but for different genes. Furthermore, many construct configurations are shown that are not used in the subsequent panels. For example, the mNeonGreen only (no Cre) constructs and the EGFP constructs are largely not used in downstream experiments. The authors could pick the important constructs and show the relevant data, and summarize all their other constructs in one supplementary figure. The authors also jump around in different parts of the paper with regards to using iCre or CreERT2 and ubi:Switch or ubi:CSHm. It's not clear to me why they're doing that? It makes the paper hard to follow. For example, why use iCre - it's not temporal if I understand correctly (and I'm not sure what improved Cre is - could they reference a paper and include a small explanation) so CreERT2 seems suitable especially for their temporal lineage tracing experiments. Why not limit the description to CreERT2 in the main text/figures? Also, isn't ubi:Switch and ubi:CSHm pretty similar except the latter is nuclear mCherry due to H2B? Why not only focus on ubi:CSHm experiments? I found the paper to be unnecessarily long and think it would benefit from editing to describe the most important concepts and experiments.
2. Could the authors describe the purpose of the 5'AmC6 modification earlier in the paper? I didn't see much text about it until the discussion. It seems that the speculation is that it provides end protection and prevents degradation (based on in vitro studies in human). This should be inserted into the introduction as a reader might be wondering about this and won't find an answer until near the end. Also, is this the first in vivo use of this modification for knock-ins? If so, that should be highlighted in the text.
3. The authors do not show any sequencing data confirming that their insert was knocked-in as designed with no disruption to the immediate upstream and downstream endogenous sequences. Can they sequence the loci to confirm?
4. I found the descriptions of the long and short HA to be confusing when describing the results, especially since the first tested gene krt92 only has long and all subsequent ones are short. The discussion made it more clear that short HA is more efficient and applicable when gRNAs span the stop codon. Perhaps that wasn't possible with krt92, but the authors could prevent the confusion by clearly stating the design requirements of long and short HA and that they wanted to test which is more efficient before starting to describe the data. I also didn't see a description of what the length difference between long and short HA is? How short is short HA?
5. The authors state that they could not use in situs to confirm krt92 endogenous and knock-in expression overlap, but rather say that they match based on data from an intestine scRNA-seq dataset. Can they elaborate on this? Which clusters/cell types show overlap? Furthermore, is there any krt92:GFP transgenic line that can be used as a reference for expression as well? This point is also applicable for krt4 described in Fig.2
6. I think Figure 2A needs the dotted lines on the last construct to be fixed (points to p2A)
7. There are a few instances where the authors describe performing 4-OHT treatment for long period (e.g. over a 20 hour or 24 hour period). Is fresh 4-OHT added after a certain amount of time or is it a one-time addition? Is such long periods of 4-OHT required or has maximal recombination already occurred within a few hours after addition of 4-OHT?
8. For Figures 4-6 where confocal images of lineage tracing experiments are shown, there is no indication of how many times the experiments were repeated, how many sections were images, how many animals used, how many cells counted. All of this information should be included in the figure legends and plots should be added showing quantification and statistical analysis (where appropriate).
9. Figure 4 C, C' - I'm not sure what to look for. Is the message that there is no Cherry positive cells that are vasnb negative when labelling is done at 8 somite? But the vasnb positive cells that are also Cherry positive remain? The vasnb staining seems much weaker/harder to see in C C' compared to B, B'. As mentioned above, these data should be quantified and statistical significance indicated.
10. I recommend the authors include a short section in the discussion comparing the efficiency of their method to other knock-in strategies used in zebrafish. This is an important claim of the paper yet it is not clear how much better it is (if at all) in terms of frequency of F0 mosaicism and identification of founders relative to other methods. I do appreciate the relative simplicity of the molecular steps of construct design/generation.

Significance

Overall, the study contributes a new knock-in strategy in zebrafish that appears to be more user-friendly and results in high germline transmission. The authors also identify nkx6.1+ ductal cells as progenitors of endocrine cells in the pancreas highlighting the biological applications of their method. I think this study represents an important advancement in zebrafish genetics and will have future impact in lineage tracing during development, regeneration and disease.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We thank the Reviewers for their comments. Below we have the Reviewers’ comments and our responses.

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

In this work, the authors claim that their machine learning approach can be combined with a biophysical model to predictably engineer sensors. The concept is interesting, but there are many issues that must be addressed before considering its publication.

1. It is surprising that their citations are too biased. They keep citing nonrelevant papers from several groups while omitting many key papers regarding genetic sensors and circuits in the field. Some can be justified (e.g., Voigt lab's reports), but others (e.g., reports on dynamic controllers too often) would not be relevant.

There are hundreds (possibly thousands) of papers that have been published on genetic sensors. Most of those papers report only qualitative results (e.g., genetic sensor implemented in a new host organism or demonstrated to sense a ligand of interest).

The purpose of this manuscript is to demonstrate methods for quantitative engineering of genetic sensors. Specifically, the manuscript is focused on quantitative tuning of the genetic sensor dose-response curve. So, in deciding which previous papers to cite, we chose several review articles (to cover the many, many qualitative results), any previous papers we could find that reported strategies for tuning the dose-response curve of genetic sensors (the Voigt lab’s reports and others), and any papers we could find that discussed reasons/applications for quantitative tuning of a genetic sensor dose-response curve (e.g., dynamic controllers).

We added a new paragraph to the beginning of the Results section to explain this focus on quantitative tuning (and to clearly state which statistic we use for assessing accuracy – see response to next comment; lines 72-83 in the revised manuscript).

We would also like to add more relevant citations as suggested by the reviewer, but that is difficult based on the reviewer’s comment, which just indicates that we have omitted many “key” papers. For the central focus of this manuscript, we think the “key” papers are those that describe methods to tune the dose-response curve of genetic sensors, and we have done our best to cite all of those that we could find. So, we ask the reviewer to please suggest some specific papers that they consider to be “key” that we should cite, or at least some more specific definition of what they think constitutes a “key” paper that should be cited.

It is very unclear which statistical analysis has been done for their work.

The main statistical metric used in the manuscript is the fold-accuracy. The fold-accuracy was defined in the previous version of the manuscript, but we agree that it could have been stated more clearly. So, we have moved the definition of fold-accuracy to the (new) first paragraph of the Results section, and identified it as “…the primary statistic we will use to assess different methods.” (line 77 of the revised manuscript)

There are many practical sensors for real applications, but their work focuses on IPTG-responsive sensors or circuits. I was wondering whether this work would have significant impacts on the field or the advancement of knowledge.

Similarly, it is questionable that their approach is generalizable.

Currently, there is only one published dataset that can be used for the methods described in this manuscript, for IPTG-responsive LacI variants.

However, previous work (cited in our manuscript) has shown that directed evolution can be used to qualitatively “improve” a wide range of genetic sensors beyond LacI. Furthermore, some of those previous studies used a single round of mutagenesis and libraries with diversity similar to the size of the LacI dataset (104 to 105 variants). Based on that, we think it is highly likely that our in silico selection approach will generalize to other sensor proteins.

With regard to the ML methods used in our manuscript, we showed in the initial publication describing the LANTERN method that the approach is generalizable to different types of proteins and protein functions (LacI sensor protein, GFP fluorescence protein, SARS Cov-2 spike-binding protein). So, we don’t see any reason to question the generalizability of that approach to other sensor proteins.

We have edited the Discussion section of the manuscript to include these points regarding the generalizability of our approach (lines 340-350 in the revised manuscript).

Due to the biased literature review, it is unclear to me whether this work is novel.

The majority of relevant literature on genetic sensor engineering is qualitative in nature and is not particularly comparable to the work here. We have tried to emphasize this in the introduction and discussion. We have searched the relevant literature extensively, and we have only found a small number of papers that describe quantitative methods to tune the dose-response of genetic sensors. Furthermore, there are only a few that contain any kind of quantitative assessment of that tuning. We have cited all of those papers and included specific discussions and comparisons between them and our results.

If the reviewer knows of any specific papers that we missed we would be happy to include them in our literature review.

I am unsure whether their correlation is sufficiently high.

This comment is too vague to address.

Again, we ask the reviewer for more specific information: What “correlation” are you referring to? And what is “sufficiently high”?

We have provided statistics on the accuracy of our methods, as discussed above.

Is EC50 the only important parameter? Or is it really relevant for real applications where the expression levels would change due to RBS changes, context effects, metabolic burdens, circuit topologies, etc.?

EC50 is not the only important parameter. That is why we also demonstrate the ability to quantitatively tune other aspects of the dose-response (e.g., G∞).

In any real application of genetic sensors, the EC50 will have to be engineered to have a quantitatively specified value (within some tolerance). So, yes, it really is relevant.

There is an important question about the effect of context however, and perhaps that is what the reviewer is really asking: If we engineer a genetic sensor that has a given EC50 in the context used for the large-scale measurement, will we be able to use that genetic sensor in a different context where, because of the change in context, its EC50 may be different?

This is one of the outstanding challenges in the field, to be able to predict the effect of a change in context. But for genetic sensors, there are several previous publications that demonstrate promising routes to quantitatively predict the effect of context on genetic sensor function.

So, we have added a paragraph to the Discussion section addressing this point and citing the relevant previous publications (lines 315-339 in the revised manuscript).

There are many reports on mutations or part-variants and their impacts on circuit behaviors. Those papers have not been cited. This is another omission.

As discussed in response to Comment 1, above, there are many hundreds of such papers. It would not be practical or appropriate for us to cite all of them. However, there are only a few that contain any kind of quantitative assessment of the predictability of mutational effects or of efforts to use mutations to engineer sensors to meet a quantitative specification. We have done our best to cite and discuss all of those. Again, if the reviewer knows of any specific additional papers that we should cite, please tell us.

CROSS-CONSULTATION COMMENTS

In general, I agree with the other reviewer. Its significance would be too incremental.

Reviewer #1 (Significance (Required)):

See above.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

This paper proposes two approaches for forward-design of genetically encoded biosensors. Both methods rely on a large scale dataset published earlier by the authors in Mol Syst Biol, containing ~65k lacI sequences and their measured dose response curves. One approach, termed 'in silico selection', is proposed as a way to find variants of interest according to phenotypic traits such as the dynamic range and IC50 of the biosensor dose-response curve. The second approach uses machine learning to regress the dynamic range, IC50 and others from the lacI sequences themselves - the ML regressor can then be used to predict phenotypes of new variants not present in the original dataset. The ML algorithm has been published by the same authors in a recent PNAS paper.

The manuscript has serious flaws and seems too preliminary/incremental:

1) The 'in silico selection' method corresponds to a simple lookup table. This is a perfectly acceptable method for sequence design, but the attempt to portray this as a new method or 'multiobjective optimization' is highly misleading. Also, the analogy between 'in silico selection' and darwinian evolution or directed evolution are inappropriate, because both latter approaches rely on iterative selection through fitness optimization and randomization of variants. The 'in silico selection' approach in contrast is one-shot and does not use randomization.

We agree with some of the reviewer’s points here. In making the analogy to directed evolution, we wanted to give the reader a connection to something familiar, but the reviewer is correct that the analogy is imperfect. The “lookup table” description is much better, and probably a familiar idea to most readers. So, we edited the relevant paragraph to describe in vitro selection as the use of the large-scale dataset as a lookup table instead of making the analogy to directed evolution. We thank the reviewer for this suggestion.

However, we disagree with the reviewer with regard to “multi-objective optimization.” We clearly demonstrate in Figures 3 and 4 that we can simultaneously tune multiple aspects the dose-response curve to meet quantitative specifications. If the reviewer is aware of any previous publications that they think provide a better demonstration of multi-objective engineering of biological function, please let us know; we would like to cite those papers appropriately.

Also, the reviewer is incorrect in stating that our in silico selection approach does not use randomization. The randomization occurs as part of the large-scale measurement. This is clearly stated in the second paragraph of the Results section.

2) The ML approach is a minor extension to what they already published in PNAS 2022. One could imagine an extra figure in that paper would be able to contain all ML results in this new manuscript. A couple of comments about the actual method: a) it seems unlikely to work on sequences of lengths relevant to applications, because it relies on gaussian processes that are known to scale poorly in high dimensions. b) The notion of 'interpretable ML' is misleading and quite different to what people in interpretable AI understand. Moreover, the connection between the three latent variables, which provide the 'interpretability', and biophysical models seems to come from their earlier PNAS work and this specific dataset, but there is no indication that such connection exists in other cases. Although this is somewhat acknowledged in L192-195, the text tends to portray the connection with biophysical models as something generalizable.

The ML results presented in this manuscript are specifically aimed to quantitatively assess the accuracy of the ML predictions for the parameters of a genetic sensor dose-response curve. So, we think those results belong in the current manuscript.

The reviewer’s comment on Gaussian processes and dimensionality is clearly contradicted by the results presented in this manuscript and in our previous publication describing the ML method: The ML method works quite well for “sequences of lengths relevant to applications,” including LacI (360 amino acids), the SARS-Cov2 receptor binding domain (200 amino acids), and GFP (250 amino acids). The reason for this is that the Gaussian process is only applied on the low-dimensional latent space learned by the ML method.

The reviewer’s comment on “interpretable ML” is not relevant to this manuscript but is instead a criticism aimed at our previous publication on the ML method.

The generalizability of this approach is an open question. The same could be said for most other publications describing new methods, since most of those publications include demonstrations with only a small number of specific systems. After re-reading the relevant portions of the manuscript, we disagree with the reviewer’s suggestion that we have exaggerated the potential generalizability of the approach. For example, in the last sentence of the Results paragraph, we state, “Although imperfect, this initial test of linking an interpretable, data-driven ML model to a biophysical model to engineer genetic sensors shows promise…” And, in the Discussion section, “The use of interpretable ML modeling in conjunction with a biophysical model also has the potential to become a useful engineering approach… But more rigorous methods would be needed…”

Other comments:

3) There are quite a few reduntant figures, eg Figure 1 contains too many heatmaps of the same variables. Fig 2B and C are redundant as the contain the same information. Altogether figures feel bloated and could have been compressed much more.

We disagree. The sub-panels of Figure 1 show different 2-D projections of the multi-dimensional data that are relevant to specific aspects of the results in Figs. 2-4.

Admittedly, Fig 2C shows the residuals from Fig 2B, which is in some sense the “same information.” But it is quite common, in papers focused on quantitative results, to have one sub-panel showing a comparison between predicted and actual and a second sub-panel showing the residuals.

4) Fig 2A and 3A have problems: the blue & orange lines (Fig 2A) and blue & green lines (Fig 3A) have a kink just before the second dot from the left. Such kinks cannot have been produced by a Hill function. This kind of errors cast doubt on the overall legitimacy and reproducibility of the results.

The kinks in the curves are a consequence of the use of the “symmetrical log” scale on the x-axis, which allows the zero-IPTG and non-zero-IPTG data to be shown on the same plot while showing the non-zero-IPTG data on a logarithmic scale. That symmetrical log axis uses a log scale for large x values, and a linear scale for smaller x values. The kink appears at the transition between the log and linear scales. We have re-plotted all of the figures showing dose-response curves to move the log-linear transition to overlap with the axis break.

CROSS-CONSULTATION COMMENTS

I agree with the other reviewer's comments, particularly on the lack of statistical analyses.

See our response to Reviewers #1, comment 2, above.

Reviewer #2 (Significance (Required)):

The work addresses a timely subject but is too incremental.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

This paper proposes two approaches for forward-design of genetically encoded biosensors. Both methods rely on a large scale dataset published earlier by the authors in Mol Syst Biol, containing ~65k lacI sequences and their measured dose response curves. One approach, termed 'in silico selection', is proposed as a way to find variants of interest according to phenotypic traits such as the dynamic range and IC50 of the biosensor dose-response curve. The second approach uses machine learning to regress the dynamic range, IC50 and others from the lacI sequences themselves - the ML regressor can then be used to predict phenotypes of new variants not present in the original dataset. The ML algorithm has been published by the same authors in a recent PNAS paper.

The manuscript has serious flaws and seems too preliminary/incremental:

1. The 'in silico selection' method corresponds to a simple lookup table. This is a perfectly acceptable method for sequence design, but the attempt to portray this as a new method or 'multiobjective optimization' is highly misleading. Also, the analogy between 'in silico selection' and darwinian evolution or directed evolution are inappropriate, because both latter approaches rely on iterative selection through fitness optimization and randomization of variants. The 'in silico selection' approach in contrast is one-shot and does not use randomization.
2. The ML approach is a minor extension to what they already published in PNAS 2022. One could imagine an extra figure in that paper would be able to contain all ML results in this new manuscript. A couple of comments about the actual method: a) it seems unlikely to work on sequences of lengths relevant to applications, because it relies on gaussian processes that are known to scale poorly in high dimensions. b) The notion of 'interpretable ML' is misleading and quite different to what people in interpretable AI understand. Moreover, the connection between the three latent variables, which provide the 'interpretability', and biophysical models seems to come from their earlier PNAS work and this specific dataset, but there is no indication that such connection exists in other cases. Although this is somewhat acknowledged in L192-195, the text tends to portray the connection with biophysical models as something generalizable.

Other comments:

1. There are quite a few reduntant figures, eg Figure 1 contains too many heatmaps of the same variables. Fig 2B and C are redundant as the contain the same information. Altogether figures feel bloated and could have been compressed much more.
2. Fig 2A and 3A have problems: the blue & orange lines (Fig 2A) and blue & green lines (Fig 3A) have a kink just before the second dot from the left. Such kinks cannot have been produced by a Hill function. This kind of errors cast doubt on the overall legitimacy and reproducibility of the results.

Referees cross-commenting

I agree with the other reviewer's comments, particularly on the lack of statistical analyses.

Significance

The work addresses a timely subject but is too incremental.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this work, the authors claim that their machine learning approach can be combined with a biophysical model to predictably engineer sensors. The concept is interesting, but there are many issues that must be addressed before considering its publication.

1. It is surprising that their citations are too biased. They keep citing nonrelevant papers from several groups while omitting many key papers regarding genetic sensors and circuits in the field. Some can be justified (e.g., Voigt lab's reports), but others (e.g., reports on dynamic controllers too often) would not be relevant.
2. It is very unclear which statistical analysis has been done for their work.
3. There are many practical sensors for real applications, but their work focuses on IPTG-responsive sensors or circuits. I was wondering whether this work would have significant impacts on the field or the advancement of knowledge.
4. Similarly, it is questionable that their approach is generalizable.
5. Due to the biased literature review, it is unclear to me whether this work is novel.
6. I am unsure whether their correlation is sufficiently high.
7. Is EC50 the only important parameter? Or is it really relevant for real applications where the expression levels would change due to RBS changes, context effects, metabolic burdens, circuit topologies, etc.?
8. There are many reports on mutations or part-variants and their impacts on circuit behaviors. Those papers have not been cited. This is another omission.

Referees cross-commenting

In general, I agree with the other reviewer. Its significance would be too incremental.

Significance

See above.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This study aims to gain a better understanding of PDLCs and their associated cementum and alveolar bone. The study provides a very clear results for differential expression of Plap-1 and IBSP the periodontal fibroblast and associated cementoblasts and osteoblasts.

The most infesting is the generation of reporter mice for identification of Plap-1+ cells. The generation of this mice lined allowed then to gain insight to the regeneration of periodontium as well as heterogeneity in of Plap-1+ cells.

Minor issues:

1. Many abbreviation in the papers have to be better defined. (Spp1, Bgn, Sparc, Col1a. also DN and DP in legends to Figure 3.
2. Legends to all figure can be written more clearly.
3. Statement in the result (line 27 and 28) cement oblasts and osteoblasts were aligned ..... should be eliminated as the figure 1A does not allow appreciation of such features. Also, the statement does not add anything to the manuscript and its results.
4. The statement on Page 5 (line 1, 2) the protein distribution of Plap-1 needs to be described.
5. Line 14 and 15 on page 5. It should be noted that very few/if any cells are co-expressing Ibsp and td-tomato. The number is so few that brings questions to the conclusion.

Major/important issues to be addressed:

1. The authors have very nicely and clearly shown that Ibsp is expressed by cementoblasts and osteoblasts but not by PDL fibroblasts. Therefore, the lineage tracing experiments after PDL injury should be followed by examination of Ibsp in cementoblasts and osteoblasts originating from the Plap-1+ cells.
2. It is also important to know what is the percentage of Plap-1+/Ly6a+ cells.
3. The author should include a stronger statement for the possible role of Plap-1+/Ly6a+ cells (not Plap-1+ alone) as a source pf progenitors for periodontium.

Significance

by providing new markers and new transgenic animal model, the paper makes an important and significant contribution to the field

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

• The authors describe in a richly illustrated manuscript periodontal ligament associated protein-1 Plap-1 as a periodontal fibroblast (PDLC) associated molecule that has the possibility to differentiate both into cementoblasts lining the tooth surface and into osteoblasts lining the alveolar bone.

• In the introduction, please not only refer to higher expression of Plap-1 in certain tissues, but also refer to the function, as revealed by Sakashita et al (also on the periodontium/susceptibility to periodontitis). Apart from the fact that it is a 43 kDa ECM protein, subtype of the leucine-rich etc., it is also important to briefly sum-up -if scientific data allow - the function of the protein.

• Page 5, line 2: have the authors investigated Plap-1 in tissues other than the periodontal ligament? These experiments seem essential to demonstrate the uniqueness of Plap-1, possibly as a confirmation of the Sakashita et al. paper of 2021. It is always a good habit to confirm previous work in a next study.

• Page 5 on lineage tracing with Tomato: please spend a few lines on the essence of the experiment, either on page 5 or in the legend of Fig. 2, or both. You will thus keep the readers involved who are not familiar with lineage tracing.
• Page 6 and 7: the description of the protocol is very valuable. It is also important the cell numbers of the various cell types were described in great detail (Fig 3). So, authors have now used cells derived from extracted teeth, which is world-wide a sample of convenience. However, after extraction, half of the PDLCs are likely attached to the alveolar bone of the tooth socket. Have the authors ever considered to harvest these cells? In principle, and biologically, PDLC cells at the site of the alveolar bone could be the more osteogenic cells. The PDLC that are attached to tooth could in principle be quite different, being anti-osteogenic and anti-osteoclastogenesis-stimulating.
• Page 6, line 8: indicate what CD51+ cells are. In corresponding figure 3, explain the abbreviations in the X-axis in the legend.
• Page 7 line 1: The proerythroblasts in the PDL are a surprise to me! I assumed that the bone marrow would be the natural niche. Authors are also encouraged to highlight plasma cell specific RNAs in their atlas, since these are quite abundant in periodontitis lesions.
• In figure 4, its seems to me that the stromal cells in 4C are scattered more or less in 3 domains. This idea is strengthened when interpreting 4E Plap-1 and lbsp. Could the authors specify these domains?
• On Page 7: again for the not-so-informed reader: briefly, in the first sentence, describe the phenomenon of RNA velocity.
• Page 7, line 20: delete "were".
• In figure 5A it could be helpful to put the numbers in the figure as well.

• In figure 6G, it seems like that some osteocytes are positive, which means that they were derived from the TomatoRed cells within 7 days. That is quite remarkable and should gain some attention. Probably use a white arrow and specific mentioning in the legend.

• In the discussion, I miss a clear link and comparison with human periodontal ligament. Is all this mouse specific or are some of these aspects also present in the human periodontal ligament? One study comes to mind that has actually studied gene expression of Plap-1 etc. in PDLC and in alveolar bone derived cells: Loo-Kirana R, et al., Frontiers in Cell and Developmental Biology, 2021: DOI: 10.3389/fcell.2021.709408. But there is bound to be other studies as well. A brief mirroring of these findings with other studies would be in place.

CROSS-CONSULTATION COMMENTS

I have read and seen the comments of the other reviewers. They are more or less in line with mine, and I have nothing to add.

Significance

Authors identify Flap-1 postive cells as key cells contributing to stem cell ness of the periodontium. With advanced techniques using GFP and Tomato Rd mice they are able to show a kind of hierarchy in cell differentiation. They also describe the presence of all kind of cells in the peridoontal ligament as well as the capacity of the Plap-1 positive cells to contribute to regeneration. It is a very valuable addition to existing literature.

Audience: those, basic scientist but also dentists in general for whom the biology of the periodontal ligament is crucial.

My expertise: periodontal ligament specialist, but more the human part. I use PDL to study osteogenesis and osteoclastogenesis, in presence of bacterial products, inflammatory and anti-inflammatory reagents.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Comments:

• In the presented study the authors attempt to perform an in vivo characterisation of the cellular hierarchies and cellular dynamics of the periodontal ligament (PDL), a largely unknown compartment of the periodontal tissue which function is to support mammalian teeth. Periodontal diseases are one of the major causes of adult tooth loss, hence understanding the cellular dynamics of this compartment is of particular interest. To do this, the authors develop a novel traceable mouse line based on the Plap-1 marker, which they identify specifically labels periodontal fibroblasts. The developed Plap-1-GFP-2A-CreER mice are then crossed onto an inducible Rosa26-tdTomato to enable in vivo tracing of Plap-1 PDL fibroblasts. This tool is of significant relevance as it allows for the first time to analyse the cellular fate of the elusive populations that constitute PDL under normal homeostatic and regenerative conditions. This mouse model also enables the authors to sort the cells of interest (as marked by GFP) to perform single-cell RNA sequencing analysis, providing further knowledge on the cellular heterogeneity of PDL cells.

• The work presented in this article is of interest for the periodontal stem cell field and more generally the mesenchymal stem cell field. In particular, through the development of new tools, including a novel lineage traceable mouse line amenable for lineage tracing studies, the authors provide knew knowledge advancing our understanding on the populations and hierarchies that constitute the PDL. Having said this, I find this study rather descriptive and, in certain cases, the significance of the results are somewhat overinterpreted. For instance, lineage tracing studies are rather vague... based on the colocalization of a widely induced traceable fluorophore and markers present in the relevant cell populations, or even just histological positioning of cells; something that entails numerous technical implications and potential artefacts. It would be more convincing to titrate down the tamoxifen levels used to induce Plap-1 traceable mice, in order to track how single-cell derived clones actually contribute to the formation of other PDL populations, and validate this using the relevant markers at critical time points. Quantification of clonal distribution would also provide a deeper understanding of the process.

• Another rather technical, but critical, aspect is the need for further validation of their new mouse model. Particularly, in order to interpret any prospective data on clonal dynamics, it is first important to know whether their new Cre system is tightly regulated, or whether there is leakage in the absence of Tamoxifen induction. Imaging of aged un-induced animals would help clarify this point.

• Finally, the scRNA-seq is rather superficial, a more in-depth analysis would be required to support the statements based on hierarchies and trajectories proposed by the authors.

• Despite all this, I believe the authors have the tools and data to address most of the aspects discussed below, which would make the study sound and result of advancement in the relevant field.

Significance

See above

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2022-01490

Corresponding author(s): Cariboni, Anna; Howard, Sasha R

[Please use this template only if the submitted manuscript should be considered by the affiliate journal as a full revision in response to the points raised by the reviewers.

• *

If you wish to submit a preliminary revision with a revision plan, please use our "Revision Plan" template. It is important to use the appropriate template to clearly inform the editors of your intentions.]

1. General Statements [optional]

This section is optional. Insert here any general statements you wish to make about the goal of the study or about the reviews.

2. Point-by-point description of the revisions

This section is mandatory. *Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. *

• *

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

The current manuscript in question is well written and of general interest to the reproductive neuroendocrinology field. Overall it is a well written and substantiated.

Reply: We thank the reviewer for his/her positive and supportive comments on our manuscript.

The primary problem with the paper is the data derived from the microarray. While the experimental design included replicates (n = 3), although weak, the actual microarray data was based on a single data point. A major weakness. This experiment should be repeated using more up-to-date approaches such as RNA-seq or left out of the manuscript, because this data set is compromised due to the data collection procedure.

Reply: We thank the Reviewer for raising these points, which we wish to clarify. We respectfully disagree that the microarray data generated in this study is not valuable. The transcriptomic analysis of immortalized cells was performed on 3 biological replicates (specifically, RNA was extracted from n=3 samples, obtained from each cell line at 3 different passages) and run as 3 independent samples (for a total of 6, 3 for GN11 cells and 3 for GT1-7 cells). For the primary embryonic GFP-GnRH neurons, given the difficulty of isolating with FACS a sufficient number of GFP+ cells from each embryo due their very small number (around 1000 GnRH neurons/head), we had to pool sorted cells from 2-3 embryos for each time-point. Thus, although the primary cell microarrays were run on one sample for each time point, the RNA was not derived from one embryo only, but from at least 2/3 embryos.

Nevertheless, to overcome the issue of low number of replicates for the primary embryonic cells, we revised our manuscript by re-running our analyses, using as the starting dataset the analyses obtained from immortalized cells, which were based on a ‘true’ n=3 of biological replicates. In this context, we filtered DEGs from this microarray using logFC>2 and adj. p-value1) found in primary GFP-GnRH neurons. We believe that this revised analysis is statistically more powerful, as the core bioinformatic analyses were performed on triplicate samples, with a second filtering step to take advantage of biologically relevant data obtained from n=1 primary GFP-GnRH neurons to confirm in vivo the expression of selected genes. Whilst RNAseq offers wider coverage of the genome and has advantages over microarray, we do not believe that this renders unimportant the data generated from these unique experiments and the novel genomic discoveries it facilitated.

In line with this, our work may be considered as a proof-of-principle that transcriptomic profiles from rodent GnRH neurons can be exploited at different levels, including the possibility to identify novel GD candidate genes. Overall, our work also highlights the existence of similarities between two immortalized GnRH neuron cell lines with primary GnRH neurons, which was so far demonstrated by several functional studies, but not at molecular level.

The manuscript has been now edited as per the above amendments (see first and second paragraph of Results section, lines 86-135).

__CROSS-CONSULTATION COMMENTS __Notwithstanding the importance of neuroligin 3 during glutaminergic synaptogenesis, I agree with the reviewers on both points. Further screenings of the patient's family members should be done and the microarray data should be removed or potentially moved to a supplementary status.

Reply: we thank the reviewer for their comments and, accordingly with their suggestion, we revised the filtering strategy starting from immortalized cells microarray and therefore moved a substantial part of the microarray data from primary GFP+ neurons as supplementary data. We also unsuccessfully tried to collect information of the brother from case 2 and investigated datasets from both the DECIPHER and 100,000 genome projects, but have been limited to two cases for which we have familial consent to publish.

Reviewer #1 (Significance (Required)): The paper is of significance based on the neuroligin 3 data, which is indicative of abnormal synaptogenesis. However, these defects seem to only have a limited effect on the functionality of GnRH neuron system and do not seem to cause elimination of GnRH neurons themselves. Nevertheless these data do open end a new direction that may help explain some dysfunctions in reproductive health.

Reply: we thank the reviewer for their comments and agree that our findings have the potential to facilitate new avenues for the investigation of reproductive disorders.

Reviewer #2 (Evidence, reproducibility and clarity (Required)): Oleari et al performed comparative transcriptome analysis on the different developmental stages of GnRH neurons, as well as two immortalized GnRH neuronal cells GT1-7 and GN11 which represent mature and immature GnRH neurons. As a results, they identified a panel of differentially expressed genes (DEG). They further used top DEGs as candidate disease-related genes for GnRH-deficiency (GD), a disorder characterized with absent of delayed puberty and infertility. To this end, they found two loss-of-function mutations in NLGN3 in patients with GD combined with autism. This study provide a resource for the identification of novel GD-associated genes, and suggest an intrinsic connection between GD and other neurodevelopmental diseases, such as autism. I only have some minor concerns.

1. According to the pedigree, both probands (case 1 and 2) inherited their NLGN3 mutations from their unaffected mother, consistent with an X-linked recessive inheritance. However, only "parent" was used in the manuscript, therefore, it is not clear if this "parent" is the probands' mother or father. __Reply: __Thank you for this comment. We were limited to the use of non-gendered terminology due to medRxiv policies. We have now amended the text and changed ‘parent’ to ‘mother’, lines 161, 173, 179, 185 and 730. We also integrated this sentence highlighting the X-linked pattern of inheritance: “Sanger sequencing of the probands’ mothers confirmed them to be the heterozygous carrier in each family, consistent with an X-linked recessive inheritance pattern.”, lines 185-186.

It is suggested to integrate Figure 2 as a panel in Figure 1.

__Reply: __We thank the reviewer for this suggestion. Due to our revision of first two Results paragraphs, we have now edited the Figures and the filtering flowchart has been added in Figure 2.

What is the meaning of Peak LH and Peak FSH, and how are they measured in Table 2?

Reply: This refers to peak value obtained after standard protocol GnRH stimulation testing with 100mcg GnRH (Gonadorelin) as an IV bolus and measurement of serum LH and FSH at 0, 20 and 60 minutes intervals. (e.g. Harrington et al., 2012, doi:10.1210/jc.2012-1598). This clarification has been added to the text in Table 2 legend (lines 681-683).

A genotyping for the elder brother of Case 2 will be a strong evidence to support NLGN3 as a GD-associated gene.

__Reply: __We thank the reviewer for this important point. In view of this issue, we have strived to collect DNA from this individual. Unfortunately, despite trying repeatedly to contact the family of proband 2, it has not been practically possible to collect these extra data from this family.

We also identified a third case via a public database with central hypogonadism who carried a stop-gain variant in NLGN3, but unfortunately the family did not release their consent for publishing this case.

The authors claimed neither probands carried deleterious variants in known GD genes. It is suggested to indicate the exclusion criteria (which genes? How do they define a variant is deleterious?)

Reply: We thank this reviewer for raising this important point of clarification. Inclusion criteria for variants in known GD genes (updated gene list available in Supplemental Table 3) were as per Saengkaew et al., 2021 (doi: 10.1530/EJE-21-0387): “Only variants that met the ACMG criteria for pathogenicity, likely pathogenicity, or variants of uncertain significance (VUS) were retained in the analysis”. We have added this sentence in the manuscript, lines 150-151.

Please also include a sequence chromatogram for proband 2.

Reply: We thank the reviewer for their comment. We added the chromatograms for proband 2 and his heterozygous mother in revised Figure 3.

CROSS-CONSULTATION COMMENTS I agree with Reviewer 3, the genetics is not very strong, as NLGN3 mutations were only found in one GD case from their cohort and one pre-pubertal case from the literature. It will be nice to analyze the genotype and phenotype of Case 2's older brother. Further, it is important to screen NLGN3 rare sequencing variants in larger GD cohorts.

Reply: We thank the reviewer for their comment, but respectfully disagree with this assertion. The second case is not from the literature, but is a second case found thanks to GeneMatcher, an international tool that allows researchers to collaborate on novel gene discovery. We have also explored other cohorts that were available to us, including the DECIPHER and 100,000 genome project, but have been limited to two cases for which we have familial consent to publish. We anticipate that further international patient cohorts will be screened following the publication of this manuscript (added in Discussion section, lines 306-308). As described above, despite trying repeatedly to contact the family of proband 2, it has not been practically possible to collect these extra data from this family.

Reviewer #2 (Significance (Required)): This study provides a resource for the identification of novel GD-associated genes, and suggest an intrinsic connection between GD and other neurodevelopmental diseases, such as autism. It may welcome by researchers and clinicians in the filed of neurodevelopment.

Reply: We thank the reviewer for their positive and supportive comments.

__Reviewer #3 (Evidence, reproducibility and clarity (Required)):

__Summary: Oleari et al used murine GnRH1, and immortalized GnRH cell lines (GT1-7, Gn11) to define genes of interest in GnRH development and used this list to filter exome sequencing data from patients with some evidence for GnRH Deficiency.

Title: I am concerned that the title of the paper overstates the results and conclusions.

Intro: use of "candidate causative genes" overstates the evidence presented.

__Reply: __We thank the reviewer for their comment and have revised the title to reflect the findings of the study. We have also edited the sentence in the abstract reporting "candidate causative genes" as follows: “Here, we combined bioinformatic analyses of primary embryonic and immortalized GnRH neuron transcriptomes with exome sequencing from GD patients to identify candidate genes implicated in GD pathogenesis”, lines 40-43.

Results: The transcriptomic profile of the developing human GnRH neuron has been published via in vitro differentiation protocols twice (Lund et al 2020, and Keen et al 2021). Gene set data is publicly available. This should be explicitly compared in results not relegated to discussion -- two or three examples it not enough to say mouse can be used instead of human.

__Reply: __We thank the reviewer for this comment. We apologize if our sentence in the Discussion was misleading, as we did not intend to make a conclusion on the similarities of the two datasets/cell types, neither to suggest the use of rodent instead of human.

Although we are aware that differences among species might exist, mouse/rodent models including immortalized cells have been instrumental to understand the molecular mechanisms of GnRH neuron development and to predict candidate genes. Indeed, our aim was to demonstrate that transcriptomic profiles of rodent GnRH neurons could be integrated with exome sequencing data from human patients to reveal novel candidate genes.

Therefore, the aim of our study was different to that of the Lund and Keen publications. Further, caution should be exercised in any deeper comparative analyses with our transcriptomes, for following reasons: first, the GnRH neurons generated from human iPSC and cultured for 20 and 27 days cannot be objectively defined for their ‘age’ in order to be then compared to immortalized or primary embryonic GnRH neurons; second, in these datasets a different and more extensive transcriptomic technique has been used (RNAseq vs microarrays).

There was no intention to relegate to the discussion the possible similarities with other transcriptomic datasets, but we felt that these comparative analyses were beyond the scope of our work.

However, following the Reviewer’s suggestion, we have tried to make comparative analyses with the publicly available datasets from Lund et al 2020 and Keen et al 2021, and with a paper just published (Wang et al 2022), as follows.

In Lund et al. paper, GnRH-like neurons were obtained from human iPSCs by dual SMAD inhibition and FGF8 treatment. We selected data obtained from cells treated with FGF8 and cultured for 20 days and 27 days for comparison with our early and late genes, respectively.

Because the authors of this paper did not publish the full list of differentially expressed genes (DEGs) from this specific comparison (20 vs 27days) and we were not able to retrieve it upon request, we used the normalized counts of these samples (available at ArrayExpress repository) to compare the two experimental groups with DESeq (Bioconductor release 3.15). To increase stringency of our analysis, we considered as differentially expressed those genes which displayed both an adjusted p-value of less than 0.05 and an absolute fold change of >2. The number of DEGs obtained was different and greater (5981) than from the published data, and this large number of genes may, by chance alone, contain a large fraction of any gene dataset (including the genes that we found with our analysis). For this reason, this particular comparison in this dataset cannot be informative or useful.

Next, we considered the dataset from Keen et al. In this paper, the authors have tested different differentiation protocols to obtain GnRH-like neurons from human wild-type or mCherry embryonic stem cells (hESC). They transcriptomically profiled hESC-mCherry-derived GnRH neurons at 8,15 and 25 days of culture.

Again, although we cannot precisely define the matching embryonic stage of cells cultured for 8, 15 or 25 days, we compared the lists of DEGs from immortalized GnRH neurons (GN11vsGT1-7) with the transcriptomic profiles of mCh-hESC at day 15 vs day 8 and mCh-hESC at day 25 vs day 15, respectively. We considered as differentially expressed the genes that displayed both an adjusted p-value of less than 0.05 and logFC>2. We found that the majority of the genes that were differentially expressed in one dataset were not in the other. However, the few genes that were differentially expressed in both datasets demonstrated a good correlation, i.e. the same expression trend. Although this latter approach was more fruitful, by suggesting a partial similarity between primary GFP-GnRH neurons and hESCs-derived GnRH neurons at day 25 vs day 15 time-point, we do not feel that we could draw significant and reliable conclusions.

Further, if we compare these two datasets obtained by RNAseq from hiPSC and hESC, even by taking into account the large amount of DEGs found in our re-analysis of Lund et al., 2021 raw data, a relatively small number of common DEGs were found. These data also suggest that there is transcriptomic heterogeneity even among human-derived GnRH neurons.

In addition to these two datasets, while our manuscript was under revision, a new paper was published, in which the authors dissected iPSC-derived GnRH neuron transcriptome with RNA-seq at single cell level (Wang et al., 2022, doi:10.1093/stmcls/sxac069). Again, although the same concerns may apply in comparing this dataset with ours and raw data of DEGs were not publicly available in this case, we compared the expression trends of our 29 candidates with gene expression trajectories identified in this work. As a result, 24/29 candidate genes, including NLGN3, were found to have an expression trend consistent with our dataset. The few remaining genes exhibited an opposite trend (2/29) or were not found in available data from this work (3/29). As this is a purely qualitative analysis, we do not feel it would be appropriate to include it in the Results section, but have included commentary on these comparative dataset analyses in the Discussion section (lines 247-257). A future study could be designed to mine the raw data from all the available transcriptomic profiles of developing GnRH neurons, but this is beyond the scope of our current manuscript.

The authors need to comment on other GnRH1 expression in the brain of developing rodent and if they think the GnRH1 sorted neurons are just "GnRH Neurons" associated with reproduction (Parhar et al 2005) due to microdissection.

__Reply: __We thanks the reviewer for raising this point of clarification. We have carefully selected by microdissection nasal areas from E14, nasal and basal forebrain areas from E17 and basal forebrain from E20 rat embryos (see revised Methods, lines 325-327). We are therefore confident that what we have obtained is RNA from ‘reproductive’ GnRH neurons only.

Questions about Cases/Missing Phenotypic Information: 1) Case 1: the patient underwent increased testicular volume on testosterone therapy -- testosterone therapy does not increase testicular volume. Has this patient undergone or been assessed for reversal of his hypogonadism?

__Reply: __We thank the reviewer for their comment. The patient had minimal testicular development on testosterone (from 10ml to 12ml) but did not increase testes volume beyond 12mls, consistent with a partial HH phenotype. He has had two trial periods of 3-4 months off testosterone treatment and during these periods had both low serum testosterone concentrations and symptoms of hypogonadism (tiredness, low energy and reduced muscle strength).

2) Case 2: Is too young to be classified as having a pubertal defect. Microphallus is mentioned but what size, was this diagnosed at birth and treated? I think the case for GD is overstated in the results and discussion (especially with the discussion of small testes).

Reply: We thank the reviewer for requesting these clarifications. The patient has not received any treatment for his microphallus (2.5 cm length in mid-childhood). We agree that this case is too young to be classified as having a pubertal defect, but the presence of microphallus and small testes volume in infancy and early childhood, in association with low gonadotrophins and absent erections, are well recognized as red flag signs for hypogonadotropic hypogonadism (Swee & Quinton, 2019, doi:10.3389/fendo.2019.00097). We added this information to the Results section, lines 175-177.

Genetic Information: Since this was a candidate gene search -- what other candidate genes were uncovered in these probands?

Reply: The revised list of 29 candidate genes were screened in the two probands from our study using the whole exome sequencing datasets for these individuals, and only the variants of interest in NLGN3 described in the manuscript were found.

By searching for mutations of the revised list of candidate genes in our GD cohort, we identified nonsense variants only in NLGN3 and no splice variants. We also found few rare and predicted damaging missense variants in this gene list identified. Indeed, two rare (MAF 25) missense variants were identified in the genes PLXNC1 and CLSTN2 in two further probands (now summarized in Supplemental table 4). We have not identified further probands with PLXNC1 or CLSTN2 variants of interest from additional cohorts and thus at present we have not yet taken these gene variants further for molecular characterization, but we will examine the relevance of this gene variant in future work.

Do the probands have a clear explanation for their developmental disability other than the gene noted?

__Reply: __We thank the reviewer for raising this point. Proband exomes were also screened for genes related to developmental delay and no other causal gene variant were identified. We added this information in the text, lines 183-185.

I would encourage the authors to update Table 3: they are missing IHH/KS genes such as GLI3, SEMA7A, SOX2, STUB1, TCF12. I suggest they update the Table and analyses.

Reply: we thank the reviewer for highlighting this point. Since we performed a new analysis, we also performed a new candidate gene prioritization using a more up-to-date gene list to instruct ToppGene (please see revised Supplemental table 3).

CROSS-CONSULTATION COMMENTS Dear Reviewer #2, I am concerned that the paper presents only a single case of GD to support the scientific work. What do you think?

__Reply: __We would like to highlight that, as we describe above, GD can be diagnosed prior to pubertal age in individuals with red flag phenotypic signs and biochemical evidence of hypogonadism.

Dear Reviewer #1: In addition to the weakness in the microarray data, what do you think about the authors using publicly available data from human GnRH neuron transcriptomics for analysis?

__Reply: __please see the above discussion on the comparison with publicly available datasets.

Reviewer #3 (Significance (Required)):

There is not high significance to this paper: This is not the first article with GnRH transcriptomes. I would argue the human data is more relevant. Developmental disability has been previously linked the GnRH deficiency (as even cited in this paper) The article presents one case of GnRH deficiency, and one pre-pubertal case -- providing some modest evidence for a candidate gene, NLGN3.

__Reply: __We would like to rebuff this assessment of the paper’s significance. To our knowledge, this is the first report of transcriptomes from primary GnRH neurons isolated at key embryonic developmental time points. Other published reports refer to iPSC-derived or adult GnRH neurons (Keen et al., 2021; Lund et al., 2020; Wang et al., 2022; Vastagh et al., 2016 and 2020).

Similarly, the association of central hypogonadism with developmental disabilities have been reported in registry-based studies, but few causative genes have been identified, nor patient variants functionally validated in order to investigate the molecular biology underpinning this association. In the Discussion, in the light of a recent paper (Manfredi-Lozano et al., 2022, doi: 10.1126/science.abq4515), we also postulate that NLGN3 might be required for neuritogenesis of extra-hypothalamic projections of GnRH neurons thus contributing to the pathogenesis of NDD (lines 294-300).

Regarding to human data, we would like to acknowledge that we had a third case that we were not able to publish due to family consent. NLGN3 deficiency is likely to be a rare disorder, but that should not obviate the impact of investigating the molecular etiology – indeed, many insights into human biology have come from private mutations in rare disease.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

Oleari et al used murine GnRH1, and immortalized GnRH cell lines (GT1-7, Gn11) to define genes of interest in GnRH development and used this list to filter exome sequencing data from patients with some evidence for GnRH Deficiency.

Title: I am concerned that the title of the paper overstates the results and conclusions.

Intro: use of "candidate causative genes" overstates the evidence presented.

Results:

The transcriptomic profile of the developing human GnRH neuron has been published via in vitro differentiation protocols twice (Lund et al 2020, and Keen et al 2021). Gene set data is publicly available. This should be explicitly compared in results not relegated to discussion -- two or three examples it not enough to say mouse can be used instead of human.

The authors need to comment on other GnRH1 expression in the brain of developing rodent and if they think the GnRH1 sorted neurons are just "GnRH Neurons" associated with reproduction (Parhar et al 2005) due to microdissection.

Questions about Cases/Missing Phenotypic Information:

1. Case 1: the patient underwent increased testicular volume on testosterone therapy -- testosterone therapy does not increase testicular volume. Has this patient undergone or been assessed for reversal of his hypogonadism?
2. Case 2: Is too young to be classified as having a pubertal defect. Microphallus is mentioned but what size, was this diagnosed at birth and treated? I think the case for GD is overstated in the results and discussion (especially with the discussion of small testes).

Genetic Information:

Since this was a candidate gene search -- what other candidate genes were uncovered in these probands? Do the probands have a clear explanation for their developmental disability other than the gene noted?

I would encourage the authors to update Table 3: they are missing IHH/KS genes such as GLI3, SEMA7A, SOX2, STUB1, TCF12. I suggest they update the Table and analyses.

Referees cross-commenting

Dear Reviewer #2, I am concerned that the paper presents only a single case of GD to support the scientific work. What do you think?

Dear Reviewer #1: In addition to the weakness in the microarray data, what do you think about the authors using publicly available data from human GnRH neuron transcriptomics for analysis?

Significance

There is not high significance to this paper: This is not the first article with GnRH transcriptomes. I would argue the human data is more relevant. Developmental disability has been previously linked the GnRH deficiency (as even cited in this paper) The article presents one case of GnRH deficiency, and one pre-pubertal case -- providing some modest evidence for a candidate gene, NLGN3.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Oleari et al performed comparative transcriptome analysis on the different developmental stages of GnRH neurons, as well as two immortalized GnRH neuronal cells GT1-7 and GN11 which represent mature and immature GnRH neurons. As a results, they identified a panel of differentially expressed genes (DEG). They further used top DEGs as candidate disease-related genes for GnRH-deficiency (GD), a disorder characterized with absent of delayed puberty and infertility. To this end, they found two loss-of-function mutations in NLGN3 in patients with GD combined with autism. This study provide a resource for the identification of novel GD-associated genes, and suggest an intrinsic connection between GD and other neurodevelopmental diseases, such as autism. I only have some minor concerns.

1. According to the pedigree, both probands (case 1 and 2) inherited their NLGN3 mutations from their unaffected mother, consistent with an X-linked recessive inheritance. However, only "parent" was used in the manuscript, therefore, it is not clear if this "parent" is the probands' mother or father.
2. It is suggested to integrate Figure 2 as a panel in Figure 1.
3. What is the meaning of Peak LH and Peak FSH, and how are they measured in Table 2?
4. A genotyping for the elder brother of Case 2 will be a strong evidence to support NLGN3 as a GD-associated gene.
5. The authors claimed neither probands carried deleterious variants in known GD genes. It is suggested to indicate the exclusion criteria (which genes? How do they define a variants is deleterious?)
6. Please also include a sequence chromatogram for proband 2.

Referees cross-commenting

I agree with Reviewer 3, the genetics is not very strong, as NLGN3 mutations were only found in one GD case from their cohort and one pre-pubertal case from the literature. It will be nice to analyze the genotype and phenotype of Case 2's older brother. Further, it is important to screen NLGN3 rare sequencing variants in larger GD cohorts.

Significance

This study provide a resource for the identification of novel GD-associated genes, and suggest an intrinsic connection between GD and other neurodevelopmental diseases, such as autism. It may welcome by researchers and clinicians in the filed of neurodevelopment.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The current manuscript in question is well written and of general interest to the reproductive neuroendocrinology field. Overall it is a well written and substantiated.

The primary problem with the paper is the data derived from the microarray. While the experimental design included replicates (n = 3), although weak, the actual microarray data was based on a single data point. A major weakness. This experiment should be repeated using more up-to-date approaches such as RNA-seq or left out of the manuscript, because this data set is compromised due to the data collection procedure.

Referees cross-commenting

Notwithstanding the importance of neuroligin 3 during glutaminergic synaptogenesis, I agree with the reviewers on both points. Further screenings of the patient's family members should be done and the microarray data should be removed or potentially moved to a supplementary status.

Significance

The paper is of significance based on the neuroligin 3 data, which is indicative of abnormal synaptogenesis. However, these defects seem to only have a limited effect on the functionality of GnRH neuron system and do not seem to cause elimination of GnRH neurons themselves. Nevertheless these data do open end a new direction that may help explain some dysfunctions in reproductive health.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #4

Evidence, reproducibility and clarity

Summary:

In this manuscript, the authors use both cellular and single molecule assays to compare the motility properties of three human kinesin-6 proteins: MKLP1, MKLP2, and KIF20B. The presented data indicate that the three motors are primarily non-processive as single molecules, but are capable of driving plus-end directed motility in ensembles. The ability of kinesin-6 motors to move organelles in cells was also tested using an FRB-FKBP induced dimerization assay. MKLP1 and KIF20B were more capable of driving the dispersion of both peroxisomes and golgi than MKLP2. These data are interpreted to indicate that MKLP2 ensembles exhibit less force than MKLP1 or KIF20B.

Major Comments:

1. The single molecule results from analyses of MKLP2 presented in this study contrast significantly with those presented in Adriaans et al. 2020. This previous study presented evidence that full-length MKLP2 moves processively (1.1 um run-length at 150 nm/s) as single molecules, and that this processivity is enhanced by binding to the chromosome passenger complex. The current study addresses these differences to some extent by indicating that a fraction of MKLP2 motors displayed slow, processive movement. However, the kymographs of MKLP2 in the two studies still look quite different (e.g. frequency of processive movement, pausing, velocity), and further explanation would be useful for understanding the apparent conflict in conclusions regarding MKLP2 motility. Does the use of a truncated MKLP2 construct in the current study change the behavior of the protein in the motility assay?
2. The organelle dispersion assays shown in Figures 4 and 5 rely on co-transfection of motor-FRB and targeting-FKBP constructs. The extent of dispersion could be affected by expression levels of either construct in a particular cell. Controls indicating that similar expression levels were compared across experimental groups should be included.
3. The authors speculate about the contributions of kinesin-6 extensions in the neck-linker and presence or absence of the N-latch residue to the motility properties observed. However, these predictions are not tested experimentally.

Minor Comments:

1. In the legend for Figure 2B- kymographs are of fluorescent microtubules?

Significance

The presented work provides an assessment of human kinesin-6 motors in a number of different motility assays. These motors play key roles during cell division and cytokinesis, and the multifaceted investigation of kinesin-6 motility presented in this manuscript complements previous studies that examined the same motors in one type of assay or assessed the activity of kinesin-6 motors from other organisms. The work, therefore, provides a framework for future structure, function studies that will be of interest to the molecular motor, cell division, and cytoskeleton fields.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript, "Differences in single-motor and multi-motor motility properties across the kinesin-6 family," by Poulos, et al, is a comprehensive study of the motility properties of the kinesin-6 motor proteins.

The work has a high number of molecules, filaments, and cells used to have statistical significance and reliability of the results.

Positive aspects<br /> 1. Kinesin-6 family is an important class of motors that needed to be investigated in a systematic manner.<br /> 2. The work was performed using highly reproducible assays that revealed novel information about this family of motor proteins.<br /> 3. The data was presented in a clear and cogent manner, making the paper highly accessible to non-experts.

Negative aspects

My only suggestion is that the figures be discussed in order to make it a bit easier for the reader to follow. This is a minor suggestion.

Significance

Employing single molecule motility assays, gliding assays, and cellular transport assays, this study elucidates the physical abilities of this elusive and essential family of kinesin motors. Excitingly, it was shown that the kinesin 6 motors can move infrequently as single motors and more frequently as multi-motors. In cells, they can also transport cargos except MKLP2. These results clearly demonstrate that kinesin-6 motors have motility and can even move large objects under high load in multi-motor configurations. The work is well-articulated and should be accessible to a wide audience.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

In this manuscript, Poulos and colleagues perform experiments aimed at understanding the in vitro behavior of kinesin-6 family motor proteins. The results, which are well supported by the experiments, show that the truncated motors show mostly diffusive behavior in single molecule assays. Surface bound motor domains drive microtubule sliding at low velocity. Assays in which motors were coupled to organelles, using rapamycin induced dimerization of the FRB-tagged motor construct and either Golgi or lysosome proteins with a FKBP domain, further revealed that the MKLP1 and Kif20B could drive transport of both peroxisomes and Golgi, although MKLP2 could not transport the high load cargo (Golgi) and showed limited transport of peroxisomes.

Comments.

These experiments are important to show the properties of the kinesin-6 family motor domain; however, the general lack of robust single molecule processive motility supports the idea that in vivo, these motors contribute to cellular processes as part of complexes with other proteins (Central spindlin (MKLP1), Chromosome passenger complex (MKLP2)) which enhance motility. In fact others have shown that motorclustering is important for plus end accumulation of MKLP1 (using C.elegans proteins). More recently Adriaans etal showed that for MKLP2 that MKLP2 is a processive plus end directed motor, using purified homodimeres of full length protein. Importantly, addition of a recombinant CPC further increased processivity. What remains unclear is why imaging of MKLP2 in cells shows predominantly diffuse behavior with only a fraction of events showing directed motility. The authors might discuss this concept in more detail - how motility is impacted by binding partners and/or regions outside of the motor domain for some kinesin families. Alternatively, they could demonstrate the changes in motility by extending the study with longer constructs and additional components.

The authors used truncated proteins for their assay, but also tested longer constructs. They state that the behavior was similar in single molecule assays, so they focus on the truncated motors. However, figure S2 looks like there are move processive motility events for MKLP1 and MKLP2, which is more in line with some other results (i.e. Adriaans et al, for MLP2). Can the authors comment on this?

The authors perform assays to study organelle motility. What is already known about kinesin-6 motor contribution to this process? For MKLP1 and 2, the best studied role is anaphase and cytokinesis.

Significance

Overall, the work is well done; however, the main result is that motor domains of kinesin-6 show mostly diffuse motility. This strongly suggests that other binding partners or other parts of the motor are needed for processive motion. The mechanism responsible for mixture of directed and diffuse motion observed in cells remains unclear. The advance from these studies is not major, for the current manuscript. The authors could submit to a journal that supports publication of work that is well-executed, and an important part of the larger picture, but that is not a major advance. Alternatively, they could continue to address the additional cellular machanisms that may contribute to regulation of processive motion in dividing cells.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This manuscript reports motility characteristics and load-bearing properties of three human kinesin-6 family proteins that function during late telophase/cytokinesis of mitosis. The authors report single molecule and multiple motor motility assays, and vesicle dispersion assays for the three motors. Because the kinesin motors are important for normal division, their motility characteristics are of interest to workers in the mitosis field. However, data presentation in this manuscript could be greatly improved, along with interpretations of functional differences based on kinesin-6 motility properties.

Major points are the following:

1. Quantitation and presentation of the data throughout the manuscript should be improved.

The criteria used for identifying fluorescent spots as single motors are not given. This is typically based on photobleaching experiments and fluorescence intensity measurements - the authors should show these data to validate that the motility reported is due to single motors.

A table should be included that shows the single molecule motility parameters that were analyzed and compared for the three motors, rather than just the dwell times for the assays shown in Fig. 1. Other motility characteristics should include run lengths, binding rates, detachment rates, and velocity. The percentage of time that the single motors move directionally, diffuse, or remain stationary should also be given.<br /> The authors refer to imaging rates (1 frame/50ms, p. 5), but do not state the total time of the assays, making the statements uninterpretable, as it is not clear what would be expected without knowledge of the total assay time. The authors also state that a slower imaging rate (1 frame/2 sec) was used to detect slow processive motility, but the logic underlying this statement is not clear, as a longer assay time should reveal the slow processive movement irrespective of the imaging rate. These statements should be clarified.<br /> The authors give the data for the dwell times in single motor assays and velocities in multiple motor assays as the mean + SEM, but the SD rather than SEM should be reported for these assays, given that the data are for individual single motors or individual gliding microtubules. The authors state the number of replicate experiments for the assays, but they should also state the number of data points that were obtained for each replicate. Further, they should evaluate the significance of differences in their data by giving P values obtained using appropriate statistical tests and indicate whether the differences among the motors are significant.<br /> The percentages of processive events (p. 5) are most likely dependent on the amount of inactive or denatured protein in a given preparation, rather than a motility property of the motor protein - this could be determined by analysis of whether the percentages differ from preparation to preparation of each motor and whether the mean+SD of the preparations of a given motor differs from the other motors. The statements by the authors on p. 8 that "the majority of proteins do not undergo unidirectional processive motility as single molecules but rather diffuse along the surface of the microtubule for several seconds" and "It is presently unclear why only a subset of kinesin-6 molecules are capable of directional motility (Figure 1 ..." are not meaningful, as they do not take into account the percentages of the kinesin-6 proteins that are inactivated or denatured during protein preparation.<br /> Again, given that inactive motors are produced during preparation of the proteins, it is not clear what the frequency of processive motility events means. If the authors think that the frequency of processive motility events is informative and a characteristic of each motor, they should present controls showing frequencies of processive motility events for specific well characterized motors. For example, does a control of kinesin-1 show 100% or only 95% processive motility events?<br /> For the multiple motor gliding assays, velocities are shown in Fig. 2 without controls demonstrating the dependence of the velocities on motor concentration in the assays - the gliding assays require dilution experiments to show that the velocities are within the linear range of motor concentration and do not fall within the range of higher concentrations in which motor gliding velocity is inhibited or lower motor concentrations in which the density of motors on the surface is too low to support processive movement. These control experiments of motor concentration vs velocity for the gliding assays should be shown for each of the three motors that was assayed. The authors should state whether the gliding velocities that were determined correspond to the Vmax for each of the motors that was assayed.

Again, the velocities given on p. 6 should include the SD and evaluation of the significance of the differences among the motors by obtaining P values.

Proteins for motility assays: Western blots of the purified proteins should be shown as a supplemental figure.

How are the motility characteristics of the three motors related to their spindle functions? This is the central point of the manuscript but is not clearly stated.

1. Functional assays should be relevant to motor function.

Given that the kinesin-6 motors under study are mitotic spindle motors that do not normally transport vesicles, it is not clear why the authors chose to show load dependence using peroxisome and Golgi dispersion assays, rather than assays of spindle function. The authors interpret peroxisomes and Golgi to differ in dispersion load, but this appears to be based on interpretations from assays of highly processive motors, kinesin-1 and myosin V, that function in vesicle trafficking, rather than quantitative data from appropriate controls showing that peroxisomes and Golgi can be dispersed by spindle motors that bear different loads. The problems inherent in the use of these assays for spindle motors are evidenced by the authors' observations on p. 6 that MKLP1- mNG-FRB and KIF20-mNG-FRB in midbodies could not be localized to peroxisomes by rapamycin. There are no data presented showing the dependence of dispersion on protein expression/presence in the cytoplasm, making the dispersion assays difficult to interpret.

The kinesin-6 motor functional tests would be more relevant if they involved mitotic spindle assays, rather than peroxisome or Golgi dispersion assays. It is not clear how the loads involved in peroxisome or Golgi dispersion are related to kinesin motor function in the spindle. What are the implications of low- vs high-load motors in the spindle? How do the authors envision that motor loads in spindles relate to loads borne by vesicle transport motors?

Minor points needed for clarity and reproducibility of the data:

Methods

Plasmids<br /> "MKLP1(1-711) lacks the insert present in KIF23 isoform 1" - the insert present in KIF23 isoform 1 but missing in MKLP1 (1-711) should be depicted/pointed out in Fig. S1 and information provided as to its predicted or actual structure.

"KIF20B contained the protein sequence conflict E713K and natural variations N716I and H749L "- the sites of these changes should be indicated in Fig. S1 and information provided as to their effects on predicted or actual structure.<br /> Protein purification: "MKLP1(1-711)-3xFLAG-Avi was cloned by stitching four oligonucleotide primer sequences together into a digested MKLP1(1-711)-Avitag plasmid" - please explain what this means: what do the four oligonucleotide primer sequences correspond to? if they are the 3xFLAG-Avi tags, why were four sequences stitched together instead of three?<br /> The figures showing the kymographs should include labeled X and Y axes, rather than scale bars.

The significance of the statement that "All motors displayed similar behaviors when tagged with Halo and Flag tags" is not clear, as the Halo and Flag tags were also C-terminal tags, like the 3xmCit tag.

The figures (Fig. 3-5) that contain grey-scale cell depictions would be more readily interpretable by others if they were labeled with the authors' classification of the dispersion phenotype.

Significance

This manuscript reports motility characteristics and load-bearing properties of three human kinesin-6 family proteins that function during late telophase/cytokinesis of mitosis. The authors report single molecule and multiple motor motility assays, and vesicle dispersion assays for the three motors. Because the kinesin motors are important for normal division, their motility characteristics are of interest to workers in the mitosis field. However, data presentation in this manuscript could be greatly improved, along with interpretations of functional differences based on kinesin-6 motility properties.

My expertise: motors, motor function in division, motility assays, microtubules

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

2RE: Review Commons Refereed Preprint #RC-2022-01651

Overall author response to reviewers:

We appreciate the enthusiasm of reviewer 3, who remarks on how our study’s toolset allows us to tackle a longstanding question in the field. We also appreciate the fair criticisms of reviewers 1 and 2.

We have taken into account all the comments of the reviewers by adding new data and new figures, but also by better underlining the limitations of our study design. Here, we underline the main points (see below for points to points answer).

We have addressed the reviewer’s concerns by including sex-specific data in the supplemental for each figure. We also hope the reviewers can see the value in not making this manuscript about sex-specific lifespan effects given our arguments below. We believe that additional experiments using a co-housing design, with far more replication per sex per genotype, would be needed to make strong claims about sex-specific effects. In designing our study, we included both sexes to avoid biasing our overall results to only one sex. In our study, we did not design our experiments in a way that allows sex*genotype interactions to be distinguished from vial effects. We believe our current evidence is sufficient to make claims strictly at the genotype level. But at the sex-specific level, not only is our study not designed for a fair comparison of the sexes by virtue of keeping them in independent vials (with independent microbiomes developing over long timeframes), but such comparisons would also be less robust, effectively based on only half the data per treatment compared to comparisons at the genotype level.

Of note, we do report Cox mixed model statistics in the figures for the interested reader. We just focused on median lifespans for data presentation clarity, and to ensure we focused only on strong trends we could be confident were not statistical Type I errors (falsely rejecting a true null hypothesis). While relying on median lifespan for insights is non-standard, focusing on and showing median lifespans also allows us to better display inter-experiment variation by showing individual data points reflecting each experiment. Sum survival curves with error bars/shading would force us to make an arbitrary decision about what error range to display (SE? SD? 95%CI?), and would not permit the reader to see inter-experiment variation. Median lifespan graphs also let us show just how repeatable our experiments were. We also chose to analyse median lifespan to make it easier to consider the effect of multiple hypothesis testing, so we could better ensure that trends in our data are really striking enough to be worth comment, particularly given the genetics caveats we draw attention to (despite our isogenization efforts, e.g. DefSK3).

We hope with the revisions we provide, reviewers 1 and 2 can, if not agree, at least acknowledge our concern with claiming many sex*genotype interactions. Such effects are lost if we look at other genotypes containing those mutations (e.g. GrA vs. GrAC, or GrB vs. GrBC). It is a challenge to use the fly lines we have generated, which systematically combines 8+ mutant loci, requiring constant reflection. Taking into consideration the complexities of life span studies, we prefer to focus our attention only on the key and robust results of our study.

Nevertheless, in the revised version, we also now provide additional data, and we soften our language regarding the impact of nora virus on aging. We have included new data with additional genotypes that were nora-infected to reinforce our claim that nora virus impacts lifespan. These data greatly strengthen the correlation of nora virus and lifespan reduction. We now also make it explicit that our statement that nora virus infection shortens lifespan is strictly correlation-based, and does not reflect intentional infection experiments. Of note, our study confirms a previous observation done by Ayebed et al. (2009) in the lab of Dan Hultmark, though we find a more striking effect in the DrosDel iso w1118 background.

Overall, we believe we have reinforced the claims of our manuscript after the revisions and addressed the constructive comments of the reviewers. Changes from the original manuscript are highlighted in yellow.

---

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

Summary:

Whether and how age-associated immune hyperactivation affects the ageing process is a key question in the ageing and immunology fields. Using a CRISPR-based knockout approach, Hanson and Lemaitre address the effect of antimicrobial peptides (AMPs), essential immune effectors, on the ageing process. They compared the lifespan and climbing ability of flies deficient for groups of, or individual, AMPs to their isogenic control and the null mutants of the major immune pathways IMD and Toll. Although deletion of individual AMPs had limited effects, the authors detected an association between deletion of a group of AMPs on bacterial proliferation, and examined possible causality using antibiotic treatment. Interestingly, this causal link was missing in the IMD deficient strain, suggesting a potential AMP-independent mechanism of innate immune signalling in the ageing process. The topic is ambitious and exciting; however, the work has some quite serious technical limitations that would need to be addressed with further data analysis and experimentation.

We appreciate Reviewer 1 and Reviewer 2’s concerns regarding the sex-specific effects, which is an important consideration in aging studies. We de-emphasized sex specific effects for reasons outlined in the supplemental text, which we will reiterate here: Per experiment, per genotype, we used ~20 males in one vial and ~20 females in another vial. As a result, the interaction of sex*experiment within a genotype is effectively the same as vial effects. We have to be especially careful of vial effects in our study because we are studying immune-deficient flies susceptible to stochastic microbiome dysbiosis, causing vial stickiness and mass mortality events. Moreover, these mass mortality events are not randomly distributed over aging: they were most likely to occur on Mondays. This is because microbes accrue in the vial over 3 days during the weekend, but only over 2 day time periods for flies flipped between Monday-Friday. As a result, in genotypes that suffer dysbiosis (many AMP group mutants), dysbiosis vial effects can change that experiment’s male/female relative lifespan by ~7 days just by stochastic chance. This issue likely affects other aging studies less, as microbiome dysbiosis is a uniquely important consideration in our immune-deficient flies, and we emphasized this point by noting mass mortality events in Figure 3 as “{“ annotations in the survival curves. We have added these considerations in the revised version (new lines 209-214).

Microbe density in vials is also affected by remaining fly density. As an example of how this can greatly impact the final data at the sex*genotype level: let’s say 4 males died or were censored over the first 50 days randomly in a given experiment. That male vial (16 flies) is less dense than its paired female vial (20 flies), and would have persistent lower microbe load every flip as fewer flies are transferring microbes to the next vial through feces. Then on say… day 60 (a Monday) of that experiment there’s a mass mortality event. Specifically, at day 60 the female vial, which is more densely populated and has had days-to-weeks of higher microbe transfer rate during flipping, has 11/20 females die at once. On the other hand, the males pull through without mass mortality owing to their lower vial density. Suddenly, the female median lifespan in that experiment is 60. The male mortality is most likely to pass the median the next Monday, when another microbiota-induced mass mortality event occurs. Exactly when these critical weekends occurred was stochastic by experiment, but the fact that they occurred was consistent in all experiments in AMP mutant stocks. In this example, female median lifespan would be 60, while males would likely be 67 or so… Importantly, this isn’t a genuine sex-specific effect, it’s variance from how early stochastic mortality affects late-experiment vial density due to microbiome development. It is a vial effect that can look like a sex*genotype effect given our study design, which is exacerbated here because we work on immune deficient flies given that can exhibit microbiome dysbiosis (Marra et al., 2021; mBio).

Given the importance of AMPs in combatting infection, we chose to keep sexes separate to avoid random mating, which can introduce stochastic sexually-transmitted infections. We would also then have had to worry about differences in mating rate amongst genotypes impacting the likelihood of transmitting infections, and differently imposing stress on females by having constant suitors even when they were not receptive. A future study using a co-housing setup would be better equipped to tackle microbiome vial effects. We have now performed the study with sexes kept separately, that can provide a reference for that future study. A future study would also benefit from focusing on fewer genotypes, with far more replication per sex per genotype, to ensure AMP*sex*lifespan effects were robust.

We have now provided supplementary analyses of the genotype-specific ratios of male/female lifespans for each figure (new Table S1-S4). By using the ratio of male lifespan to female lifespan, we focus on how within-genotype comparisons work. We want to emphasize the stochasticity of this ratio depending on experiment batch, even within genotype. For instance, the iso w1118 wild-type male/female lifespan ratio was 0.87 for the six experiment in Figure 2, or 0.93 for the four experiments in Figure 3. It’s a reminder that any collection of experiments is just a sampling of the true population mean. Still, we do see some striking departures in the paper from these ratios. For instance, if we focused on striking departures from this sexual dimorphism, we could comment on a couple genotypes in Figure 3 (AMP groups figure): Group B had a ratio of 1.04 (Only females had reduced lifespan, while males remained comparable to iso w1118). Yet in Group AB the ratio is 0.91, with both sexes having proportionally reduced lifespan. And then Group BC, the ratio is 0.86, with both sexes living to wild-type lifespans. ∆AMP10 and ∆AMP14, which contain all Group B mutations, also have ratios of 0.87 and 0.93 respectively. We also did not see a striking effect in individual mutations from Group B, which all have lifespan and male/female ratios comparable to wild-type. What are we to make of Group B alone then? Is it a genuine effect that combining the Group B mutations uniquely reduces female lifespan? Or is this simply an outlier, perhaps caused by vial effects, that is not supported by all other individual or combinatory genotypes that include Group B mutations?

With all that said, the supplemental tables now included for each figure note these data trends, and includes our cautious view of how to best interpret them. We prefer to merge the sexes in the main article to simplify data display, which is not unique to our study (e.g. see Kounatidis et al. 2017; Cell Rep). We feel our study is not robust at the sex*genotype level to make assertions on sex-specific effects.

Major comments:

1. Figure 1A shows a large increase in fly lifespan in response to deletion of 8 AMPs on Chr II. However, the labelling on the figure shows that data from different experimental runs and from the two sexes have been combined to produce the result, and this problem runs through subsequent results. This is not an acceptable approach. Males and females have different average lifespans and may respond differently to the deletion. Repeats of the same lifespan measurement at different times often give significantly different absolute lifespans between runs, even if the differences between experimental treatment are consistent. The procedure used to construct this Figure therefore lumps heterogeneous data, which in this case is both biologically and, more generally, statistically invalid. We need to see an analysis where the results for the sexes and for different runs of the experiment are disaggregated. To assess the effect of the deletion, comparison should be made only within a single replicate of the experiment, and within each sex. It needs to be made clear how many experimental runs were done and with how many flies in each, since replicability is important with these labile traits. Data should be appropriately analysed, for instance with Log Rank test, or, if the assumptions of the test are fulfilled, with Cox Proportional Hazards. The authors mention analysis of median lifespans, but it is not clear what experimental data the medians were derived from, and this is not a gold standard approach to analysis of lifespan data. This problem also applies to the comparison made between lifespans of iso w1118 flies in Figure 1A and 1B - these were evidently measured in different experiments, so direct comparison cannot be made.

We hope the above response reassures the reviewer about our statistical approach, which we agree is non-standard. We feel it is less likely to result in false-positive claims (Type I error) given the many genotypes we screen and the high stochasticity of individual vials affecting some AMP mutant genotypes more than others (i.e. statistically: unequal variance, requiring an alternate data treatment). Still, we did perform Cox mixed models, and we report those genotype-specific p-values in the figures. But we prefer to make inferences only based on trends robust enough to show in median lifespan comparisons.

1. Figure 2A reports data from mutants that were not backcrossed into a standard genetic background. The authors were aware that this can be a problem because they took care of it with their own mutants, but the data reported in this Figure 2A could be entirely attributable to difference in genetic background between stocks rather than the mutants themselves.

Agreed. The reviewer may be confused regarding our intent in showing these genotypes. We used lines as used in other papers to show how control genotypes behave in our hands. This makes our study easier to compare to the broader literature encompassing immunity*aging studies. The intent is not to test if those mutations’ effects are true in another background. They are there to provide context for our lab settings compared to other studies. They are inter-study calibrator controls.

We have added some text in the study to better highlight the utility of including these control genotypes for making our work more comparable for the field at large (e.g. Line 298, Line 515).

1. Sexual dimorphism is widely reported in many ageing and immunology studies (reviewed in Belmonte et al., Front. Immunol., 2020 and Garschall et al., F1000Res., 2018). This is a problem for the data discussed on line 317, where it is suggested that Group A mutations are shorter lived than the control group because Group A carries the DefSK3 mutation. However, if the results are split by sex, the lifespan difference between Group A vs control is solely contributed by the shorter-lived male flies in Group A. The female flies in Group A live as long as the female control flies. However, the lifespans of both genders of the DefSK3 mutant in Figure 2B are all shorter than control flies. There appears to be a complicated interaction, with diverse function of individual AMPs during ageing, which cannot be summarised by the statement that "deleting single AMP genes has no effect on lifespan" on line 295.

The reviewer is correct that Group A shows a male-specific lifespan reduction. But taking the example of Group B above, we could argue that Group B uniquely reduces female lifespan; except when Group B mutations are found alone, or small combinations (e.g. Dro-AttAB, DptSK1), or in the Group AB, Group BC, ∆AMP10, and ∆AMP14 backgrounds. The same contradictory results are true of the Group A result the reviewer highlights, as Group AC has the opposite effect on male/female lifespan ratios and wild-type lifespan, and ∆AMP10 and ∆AMP14 have normal male/female lifespan ratios.

1. The authors claimed nora virus regulates fly lifespan but they do not produce any direct evidence that this is the case. Bleaching can remove many bacteria and viruses, including many pathogens. To establish the causal role of nora virus on lifespan, reinfection studies with a range of microbes including nora virus would be needed.

We absolutely agree. These experiments were not intended to test for the effect of nora at the outset. Nora is only highlighted because its presence is strictly correlated, in our hands, with strongly reduced lifespan and also the bloating phenotype seen upon aging. We did not previously show other nora-infected data in the manuscript, but in retrospect, we can see that additional context is needed to reassure the reader of why we suspect nora so strongly. We have modified Figure 1B to include nora-positive data from four additional genotypes that were infected with nora at various times in the lab (previously did not show).

To explain the process of how we detected nora in some of our AMP stocks: we randomly screened them for a set of common viruses as part of a research workshop in 2019 hosted by Luis Teixeira. The results of that initial screen were not formally recorded, though we screened all the individual and group AMP mutants available at the time. The viruses we screened for were: Drosophila sigma virus, DCV, DAV, and Drosophila nora virus. Out of the stocks relevant to this study (~20 genotypes), we detected nora virus in iso w1118, AttC, Group C, and Bom. The new data we provide even includes an experiment where AttC with/without nora was included at the same timeframe (re: batch effect concern). We also had instances where OR-R became newly infected with nora virus just from random lab inoculation (not intentional infection).

Importantly, we had noted AttC, Group C and Bom as lines with exceptionally poor lifespan that also showed a bloating phenotype upon aging for ~2 years before we had nora virus on our radar. We were operating on a hypothesis of cryptic microsporidia infection at the time based on a set of chitin stainings of hemolymph that didn’t fully pan out. What convinces us this is nora virus and not a co-infecting virus/microbe is that we only saw the reduced lifespan and bloating upon aging in these confirmed nora-positive lines out of the ~20 we were screening at the time: a 5/5 correlation. We were even able to go back and screen RNA collected from those stocks from earlier experiments to confirm nora infection (informing which experiments we could objectively censor from our analyses). Upon bleaching, we rescued the lifespan of iso w1118, AttC, Bom, and Group C to the levels reported in the main figures. We hope this adds weight to the correlation between nora and lifespan, even if we haven’t done proper reinfection experiments. We also thank the reviewers for commenting, as including these data improves our study by providing more context on the variability of the nora lifespan effect in different genotypes: for instance, we did not collate the nora-infected OR-R data previously, but upon analyzing those data, the negative effect is clearly present but to a lesser extent than iso w1118 (and more in line with what was seen in Habayeb et al. (2009)).

To address the reviewer’s comment, we have also softened our language to be very clear we find only an association, and do not provide a demonstration (new lines 191-193). But the results were sufficiently striking, and poor lifespan associated with bloating was perfectly predictive of nora presence in stocks in our hands. We fully believe the result and feel it is important to include this in the main text to make the field more aware of this important aging study confounding variable.

1. The authors reported a potential AMPs-independent mechanism of the IMD/Relish pathway on ageing, which would be important. However, in Figure 4A and the associated raw data, the authors compared the lifespans of flies from different experiments, and this seems to be a problem. Using ΔAMP14 as an example of a more general problem: the authors assayed conventional feeding ΔAMP14 flies between 07.2021 and 11.2021; however, the antibiotic treated flies were assayed between 12.2021 and 02.2022. There is an obvious batch effect: the median lifespan of ΔAMP14 is 50d on 29.07.2021, 73d on 20.08.2021, 65d on 06.11.2021, and 61d on 10.11.2021. There is therefore a confounding of batch effects with the biological function of AMPs. Some groups seem to have extremely limited sample size, such as only two female flies and six male flies were recorded in "ATM8 23-07-2021 f" in Figure 2A-associated raw data.

We show this inter-experiment variation explicitly by presenting individual experiments as data points in median lifespan graphs. It is not hidden, it is emphasized by our median lifespan data presentation style. The experiments from 06.11.2021 and 10.11.2021 were from independently kept stocks of ∆AMP14, although naturally they come from similar timeframes and would have used food prepared at roughly the same time. If the reviewer feels we should merge the two experiments, we can do that. Merging the experiments and having only 3 entered replicates for conventionally-reared ∆AMP14 changes the one-way ANOVA result in Fig. 4B from “p = .006” to “p = .004,” so we don’t feel having those two experiments entered independently is biasing the analysis. We do feel the parents, and the flies measured, were independent at the rearing level.

ATM8 specifically had poor homozygous viability. This is the only genotype with such a significant departure in sample size per experiment. ATM8 flies are also not relevant to the core message of the study, and their reduced lifespan was extremely clear and in line with previous studies. We prefer to keep the genotype in the study to allow comparison of our lab conditions to studies using ATM8. We have added this note in the Fig. 2 caption (Line 1029).

Regarding how much can be attributed to batch effects: there are many elements of laboratory studies that can contribute to batch effects. However antibiotic food would undergo independent food preparation from standard food regardless of what we do, and parents and flies would be reared in independent vials from independent food preps regardless of what we do. This really just leaves the seasons as the batch effect, which ought to be controlled for by our incubators controlling temperature and humidity. We do appreciate the valid concern, and that incubators aren’t perfect. But we want the importance of the concern to be viewed in the light that even had we done the experiments at the same time, the conventionally-reared and antibiotic-reared flies would still have been given totally different preparations and conditions. Given this consideration, we will further note:

1) ∆AMP14 flies were already known to suffer dysbiosis with aging (Marra et al., 2021; mBio). Our present study only quantifies the effect of this dysbiosis on lifespan.

2) Dysbiosis is associated with flies gut barrier dysfunction and gut content leakage into the hemolymph (Rera et al., 2012; PNAS)(Clark et al., 2015; Cell Rep).

3) Imd-mediated barrier dysfunction leads to microbiome invasion into the hemolymph (Buchon et al., 2009; Genes Dev).

4) Systemic infection by microbiota-derived Acetobacter kills AMP mutants (Marra et al., 2021; mBio)(Hanson et al., 2022; Proc R Soc B), which suggests microbiome invasion into the hemolymph will kill AMP mutants more readily than wild-type flies.

The result that AMP mutants have improved lifespan in antibiotic conditions is therefore not especially surprising – it is important to actually test this, but it is by no means unexpected. One of the curious findings of our study is that we could not rescue Relish to the same extent. If batch effects were affecting the antibiotic rescue experiments by having different intrinsic lifespans during those months, we would expect this to also be visible in iso w1118 and Relish. Antibiotics (or time period batch effects) did not affect lifespan of our wild-type flies, which was very repeatable across all experiments across many years, agreeing with antibiotic-reared fly data from Ren et al. (2007).

Minor comments:

1. The authors reported that the mutant flies lacking 14 AMPs are short-lived due to dysbiosis. Interestingly antibiotic treatment rescued the shortened lifespan of ΔAMP14 but not RelE20. Considering that some of these AMPs, such as lDef, Dro and Drs, are controlled by both the IMD and Toll pathways. It would be worth exploring if the axenic environment could improve the lifespan of Toll mutant flies, which might point to a distinct function of the Toll pathway on the microbiome and ageing process.
2. Line 169: The authors stated that they screened the existence of common viruses but did not provide the results.
3. Line 179: The authors need to include the quantitative results of nora virus in their 44 stocks.
4. Lines 311-314 should be combined with the next paragraph as they are all about "screening".
5. Line 319: Please indicate the associated panel "Fig. 2B" instead of using "Fig. 2".
6. Lines 353 and 358 and in Figure 3C: The authors should provide the quantification of the sticky food in AMPs mutant and Relish mutant flies.
7. Line 389: Please indicate the associated panel "Fig. 2A-B" instead of using "Fig. 2".
8. Line 435: As discussed above, the authors cannot rule out an effect of individual AMPs on lifespan based on their current data and interpretation.
9. Line 507: "During our study, we experienced a number of challenges to lifespan data interpretation." As discussed in the major comments, unless the authors re-perform and re-analyse the lifespan assays this problem will persist.
10. Line 514: As discussed above, the authors cannot rule out the effect of other pathogens on lifespan.

We appreciate the reviewer’s request for spz axenic lifespans. The effect was more striking in Relish, and so we focused on Relish, which regulates antibacterial peptides in the gut. Repeating the experiment with spz will cause non-essential delays and would not affect our main conclusions that focus on AMPs genes.

Line 179: we have, in Fig. 1C. Negatives are not shown, but positives are included in the “Stocks” column. Is the reviewer requesting the exact stock names? We can provide most of these… although the screen was organized by having lab members submit stocks that were in use, and so the spreadsheet of those results is organized with labels provided by lab members (e.g. FM1, rather than the genotype exactly). So we cannot provide exactly which stocks were positive in our hands, but we also don’t feel this is important for the reader. We can share the spreadsheet of this screen (conducted in Dec 2020) with the reviewer if desired.

We were happy to revise the manuscript according to all other requests/comments, highlighted in yellow, and hope our explanations above are sufficient to give the full weight and meaning of the statement in Line 507.

Reviewer #1 (Significance (Required)):

Significance: The crosstalk between AMPs and ageing is a long-standing contradictory topic in the immunology and ageing field (Loch et al., PLoS One, 2017; Badinloo et al., Arch. Insect Biochem. Physiol., 2018; Garschall et al., F1000Res., 2018). This work is a potential step in determining whether and how AMPs regulate ageing. This research may open a door toward yet uncharacterized AMP-independent mechanisms of innate immune signalling in the ageing process.

I am familiar with Drosophila ageing and its relationship with innate immunity, although I am not an insect immunologist.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Summary: In this study, Hanson and Lemaitre generated AMP single or compound mutants, and subjected them to lifespan analyses. They showed that deletion of all AMPs but not individually (except for Defensin) reduced Drosophila lifespan. The observations that ΔAMP14 suffered microbial dysbiosis and antibiotic treatment rescued the decreased lifespan in ΔAMP14 flies implied a link between AMP-controlled lifespan and host microbiome. Thus, the authors would like to conclude that AMPs contribute to fly lifespan via modulating microbiome.

Comments: This reviewer deeply appreciates that the authors generated AMP mutants (individually or with various combinations). However, these mutants have been reported in their previous publications (eLife, 2019; Genetics, 2021). It seems that the authors carefully analyzed the lifespan of indicated flies, but many pieces of data are inconsistent with previous findings (for instance the lifespan under axenic condition). The explanation about the fly medium (Line 476-477) is not convincing (if yes, who not analyze the lifespan using the same diet as described in previous studies?). The lifespan analyses are of course the most pivotal part in this study, but it is a pity that the results are not shown in an appropriate way, making them rather difficult to interpret. For example, the reviewer is surprised to note that the lifespan of males and females are not analyzed and shown separately (e.g. Figures 1A, 1B, 2A, 2B, 3A, 4A), even the authors declared that they did this before lifespan examination. It is also unclear how many males and females were utilized individually in most assays.

We do not believe the axenic lifespans are inconsistent with previous findings. There have been papers both finding an effect of antibiotics, or finding no effect of antibiotics, on wild-type flies (ex: refs 56 and 57: Ren et al., 2007 and Brummel et al., 2004). Our study would support no major effect. Given our note on nora virus, one might wonder if papers where antibiotics affected lifespan might in part be explained by pathogens (like viruses) that were confounding results in the initial stock, and cleared by the initial bleaching common to those experiments. This would imply that commensal bacteria/fungi community do not have a major role in wild-type lifespans, but rather some stocks are stochastically infected with opportunistic viral pathogens, giving the impression that antibiotic treatment was the key, when in fact it was the bleaching pre-treatment that removed an opportunistic viral pathogen. This is only speculation, but it emphasizes the importance of openly discussing the effect nora virus can have on lifespan for future studies.

Regarding the diet comment: which standard diet should we have chosen? There are many standard diets. We have provided our recipe and we have used our diet for decades. Of note, we did try a set of experiments on a molasses-based food, which showed similar lifespan for wild-type flies in our hands (Response supplemental figure below).

We hope our answers to the first reviewer regarding our care to not include sex*genotype interactions as major claims in our study will satisfy reviewer 2’s concerns.

SEE ATTACHED RESPONSE TO REVIEWERS FOR FIGURE

Response supplemental figure: rearing on molasses food (recipe per lab of Brian McCabe) does not drastically alter wild-type lifespan. In this pilot experiment (n = 1) iso w1118 and OR-R lifespan remain comparable to our standard diet, including the relative lifespans of those wild-types to each other.

Another weakness is the study regarding AMP and microbiome. The authors observed that both ΔAMP14 and RelE20 flies became sticky during aging and their foods in the vials were also sticky and discolored, implying the proliferation of microbiome in the gut and/or the external environment of these flies. To test the microbial load, the authors performed an assay of the bacterial abundance on the fly medium. They further performed antibiotic treatment in the food to remove microbiome as they declared in the study and examined the effect on the lifespan of ΔAMP14 flies. According to the knowledge of the reviewer, antibiotic treatment in food can restrict the microbiome in the gut. The authors have also mentioned in the manuscript the important role of gut microbiota in impacting Drosophila aging and lifespan. Thus, a more direct and widely-utilized way is to dissect the guts for microbial analysis (qPCR, 16S-seq, etc.), which is lacking in this study without reasonable explanations.

We published a paper in mBio on the relationship of AMPs and the microbiome with aging in much greater detail previously (Marra et al., 2021; mBio, ref25). That study did not perform lifespan experiments, but rather compared microbiome communities at 10 and 29 days. The novel aspect of this study is properly following survival. However, the reviewer is correct in their critique that our study is not especially innovative nor are our findings strikingly novel. We do have, we feel, an important set of experimental results to contribute to the field at large.

Reviewer #2 (Significance (Required)):

This study utilized various AMP mutants, but these mutants have been reported in the previous publications, making this study somehow lacking the novelty in this context. The IMD pathway has been shown to be involved in regulating fly lifespan, so the findings in this study are not that surprising. Additionally, this study doesn't show any creative improvements in terms of methodology and model system.

Admittedly, in some ways, much of our study is a “negative results” study. Given the contradictory nature of the literature claiming positive or negative effects of immunity and AMPs on lifespan, we feel this is a particularly valuable contribution to avoid publication bias focusing only on significant results in this controversial field. Still, we believe that the AMP/microbiome/lifespan interactions we uncover in our article will have an impact in the immune-aging field. Thus, while not being fully creative, our article is an important step for better characterizing of immune-aging relationships, which is of broad interest.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Gene knockouts provide definitive loss of function for individual and collective AMPs. This eliminates ambiguity caused by the partial efficiency of RNAi, and it disambiguates the pleiotropy caused by mutants of relish. Survival assays are conducted with and without the resident microbiome. This combined design leads to a clear conclusion: loss if individual (mostly so) AMPs does not affect adult survival either way, demonstrating these peptides likely function redundantly. Knock out of 14 AMPs (but still not all of those encoded in the genome) reduces survival when adults must cope with a microbiome. Depleting this microbiome sends survival back to normal. Perhaps these are not surprising results in that they show immune function is essential when challenged with pathogens, but the results are important because they unambiguously show that AMPs themselves are not a cause of intrinsic aging. This will finally put away that lingering hypothesis.

Overall, I like the scholarship of the work, of how it uses the literature, and its quality experimental execution. Cohort size, replication and survival analyses meet current, high standards. Demographic aging is supplemented with data on climbing rate as a function of age. It is a strength of the study that it is simple but comprehensive.

Reviewer #3 (Significance (Required)):

Aging across animal systems is strongly associated with changes in immunity and inflammation, both innate and adaptive. Overall, we want to understand if such changes are underlying causes of morbidity and mortality with age, or are consequences and compensation to underlying aging and cumulative pathogen exposure. These are difficult questions to address in mammalian systems but amenable using Drosophila which possesses a robust innate immune system. Researchers have used the fly for this end but still have mixed and ambiguous results. Now Hanson and Lemaitre provide a substantial design that fully controls the two essential ingredients: expression of antimicrobial peptides and the microbiome.

We thank the reviewer for their positive assessment. In particular, we appreciate the nod to the importance of the question on intrinsic contributions of AMPs to lifespan. We believe this is the key strength of our study’s mutant approach, which has been challenging to assess using previously-existing tools.

Additional Author changes (not requested by reviewers):

We realised there was a copy/paste error in data used in Figure 2. Specifically, extra OR-R experiments were included in these data for this figure that were not intended to be part of this figure. We have removed these OR-R experiments, which are experiments common to Figure S4 and remain visible in Figure S4. This does not impact any of the conclusions in the manuscript.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Gene knockouts provide definitive loss of function for individual and collective AMPs. This eliminates ambiguity caused by the partial efficiency of RNAi, and it disambiguates the pleiotropy caused by mutants of relish. Survival assays are conducted with and without the resident microbiome. This combined design leads to a clear conclusion: loss if individual (mostly so) AMPs does not affect adult survival either way, demonstrating these peptides likely function redundantly. Knock out of 14 AMPs (but still not all of those encoded in the genome) reduces survival when adults must cope with a microbiome. Depleting this microbiome sends survival back to normal. Perhaps these are not surprising results in that they show immune function is essential when challenged with pathogens, but the results are important because they unambiguously show that AMPs themselves are not a cause of intrinsic aging. This will finally put away that lingering hypothesis.

Overall, I like the scholarship of the work, of how it uses the literature, and its quality experimental execution. Cohort size, replication and survival analyses meet current, high standards. Demographic aging is supplemented with data on climbing rate as a function of age. It is a strength of the study that it is simple but comprehensive.

Significance

Aging across animal systems is strongly associated with changes in immunity and inflammation, both innate and adaptive. Overall, we want to understand if such changes are underlying causes of morbidity and mortality with age, or are consequences and compensation to underlying aging and cumulative pathogen exposure. These are difficult questions to address in mammalian systems but amenable using Drosophila which possesses a robust innate immune system. Researchers have used the fly for this end but still have mixed and ambiguous results. Now Hanson and Lemaitre provide a substantial design that fully controls the two essential ingredients: expression of antimicrobial peptides and the microbiome.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

In this study, Hanson and Lemaitre generated AMP single or compound mutants, and subjected them to lifespan analyses. They showed that deletion of all AMPs but not individually (except for Defensin) reduced Drosophila lifespan. The observations that ΔAMP14 suffered microbial dysbiosis and antibiotic treatment rescued the decreased lifespan in ΔAMP14 flies implied a link between AMP-controlled lifespan and host microbiome. Thus, the authors would like to conclude that AMPs contribute to fly lifespan via modulating microbiome.

Comments:

This reviewer deeply appreciates that the authors generated AMP mutants (individually or with various combinations). However, these mutants have been reported in their previous publications (eLife, 2019; Genetics, 2021). It seems that the authors carefully analyzed the lifespan of indicated flies, but many pieces of data are inconsistent with previous findings (for instance the lifespan under axenic condition). The explanation about the fly medium (Line 476-477) is not convincing (if yes, who not analyze the lifespan using the same diet as described in previous studies?). The lifespan analyses are of course the most pivotal part in this study, but it is a pity that the results are not shown in an appropriate way, making them rather difficult to interpret. For example, the reviewer is surprised to note that the lifespan of males and females are not analyzed and shown separately (e.g. Figures 1A, 1B, 2A, 2B, 3A, 4A), even the authors declared that they did this before lifespan examination. It is also unclear how many males and females were utilized individually in most assays.

Another weakness is the study regarding AMP and microbiome. The authors observed that both ΔAMP14 and RelE20 flies became sticky during aging and their foods in the vials were also sticky and discolored, implying the proliferation of microbiome in the gut and/or the external environment of these flies. To test the microbial load, the authors performed an assay of the bacterial abundance on the fly medium. They further performed antibiotic treatment in the food to remove microbiome as they declared in the study and examined the effect on the lifespan of ΔAMP14 flies. According to the knowledge of the reviewer, antibiotic treatment in food can restrict the microbiome in the gut. The authors have also mentioned in the manuscript the important role of gut microbiota in impacting Drosophila aging and lifespan. Thus, a more direct and widely-utilized way is to dissect the guts for microbial analysis (qPCR, 16S-seq, etc.), which is lacking in this study without reasonable explanations.

Significance

This study utilized various AMP mutants, but these mutants have been reported in the previous publications, making this study somehow lacking the novelty in this context. The IMD pathway has been shown to be involved in regulating fly lifespan, so the findings in this study are not that surprising. Additionally, this study doesn't show any creative improvements in terms of methodology and model system.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Whether and how age-associated immune hyperactivation affects the ageing process is a key question in the ageing and immunology fields. Using a CRISPR-based knockout approach, Hanson and Lemaitre address the effect of antimicrobial peptides (AMPs), essential immune effectors, on the ageing process. They compared the lifespan and climbing ability of flies deficient for groups of, or individual, AMPs to their isogenic control and the null mutants of the major immune pathways IMD and Toll. Although deletion of individual AMPs had limited effects, the authors detected an association between deletion of a group of AMPs on bacterial proliferation, and examined possible causality using antibiotic treatment. Interestingly, this causal link was missing in the IMD deficient strain, suggesting a potential AMP-independent mechanism of innate immune signalling in the ageing process. The topic is ambitious and exciting; however, the work has some quite serious technical limitations that would need to be addressed with further data analysis and experimentation.

Major comments:

1. Figure 1A shows a large increase in fly lifespan in response to deletion of 8 AMPs on Chr II. However, the labelling on the figure shows that data from different experimental runs and from the two sexes have been combined to produce the result, and this problem runs through subsequent results. This is not an acceptable approach. Males and females have different average lifespans and may respond differently to the deletion. Repeats of the same lifespan measurement at different times often give significantly different absolute lifespans between runs, even if the differences between experimental treatment are consistent. The procedure used to construct this Figure therefore lumps heterogeneous data, which in this case is both biologically and, more generally, statistically invalid. We need to see an analysis where the results for the sexes and for different runs of the experiment are disaggregated. To assess the effect of the deletion, comparison should be made only within a single replicate of the experiment, and within each sex. It needs to be made clear how many experimental runs were done and with how many flies in each, since replicability is important with these labile traits. Data should be appropriately analysed, for instance with Log Rank test, or, if the assumptions of the test are fulfilled, with Cox Proportional Hazards. The authors mention analysis of median lifespans, but it is not clear what experimental data the medians were derived from, and this is not a gold standard approach to analysis of lifespan data. This problem also applies to the comparison made between lifespans of iso w1118 flies in Figure 1A and 1B - these were evidently measured in different experiments, so direct comparison cannot be made.
2. Figure 2A reports data from mutants that were not backcrossed into a standard genetic background. The authors were aware that this can be a problem because they took care of it with their own mutants, but the data reported in this Figure 2A could be entirely attributable to difference in genetic background between stocks rather than the mutants themselves.
3. Sexual dimorphism is widely reported in many ageing and immunology studies (reviewed in Belmonte et al., Front. Immunol., 2020 and Garschall et al., F1000Res., 2018). This is a problem for the data discussed on line 317, where it is suggested that Group A mutations are shorter lived than the control group because Group A carries the DefSK3 mutation. However, if the results are split by sex, the lifespan difference between Group A vs control is solely contributed by the shorter-lived male flies in Group A. The female flies in Group A live as long as the female control flies. However, the lifespans of both genders of the DefSK3 mutant in Figure 2B are all shorter than control flies. There appears to be a complicated interaction, with diverse function of individual AMPs during ageing, which cannot be summarised by the statement that "deleting single AMP genes has no effect on lifespan" on line 295.
4. The authors claimed nora virus regulates fly lifespan but they do not produce any direct evidence that this is the case. Bleaching can remove many bacteria and viruses, including many pathogens. To establish the causal role of nora virus on lifespan, reinfection studies with a range of microbes including nora virus would be needed.
5. The authors reported a potential AMPs-independent mechanism of the IMD/Relish pathway on ageing, which would be important. However, in Figure 4A and the associated raw data, the authors compared the lifespans of flies from different experiments, and this seems to be a problem. Using ΔAMP14 as an example of a more general problem: the authors assayed conventional feeding ΔAMP14 flies between 07.2021 and 11.2021; however, the antibiotic treated flies were assayed between 12.2021 and 02.2022. There is an obvious batch effect: the median lifespan of ΔAMP14 is 50d on 29.07.2021, 73d on 20.08.2021, 65d on 06.11.2021, and 61d on 10.11.2021. There is therefore a confounding of batch effects with the biological function of AMPs. Some groups seem to have extremely limited sample size, such as only two female flies and six male flies were recorded in "ATM8 23-07-2021 f" in Figure 2A-associated raw data.

Minor comments:

1. The authors reported that the mutant flies lacking 14 AMPs are short-lived due to dysbiosis. Interestingly antibiotic treatment rescued the shortened lifespan of ΔAMP14 but not RelE20. Considering that some of these AMPs, such as lDef, Dro and Drs, are controlled by both the IMD and Toll pathways. It would be worth exploring if the axenic environment could improve the lifespan of Toll mutant flies, which might point to a distinct function of the Toll pathway on the microbiome and ageing process.
2. Line 169: The authors stated that they screened the existence of common viruses but did not provide the results.
3. Line 179: The authors need to include the quantitative results of nora virus in their 44 stocks.
4. Lines 311-314 should be combined with the next paragraph as they are all about "screening".
5. Line 319: Please indicate the associated panel "Fig. 2B" instead of using "Fig. 2".
6. Lines 353 and 358 and in Figure 3C: The authors should provide the quantification of the sticky food in AMPs mutant and Relish mutant flies.
7. Line 389: Please indicate the associated panel "Fig. 2A-B" instead of using "Fig. 2".
8. Line 435: As discussed above, the authors cannot rule out an effect of individual AMPs on lifespan based on their current data and interpretation.
9. Line 507: "During our study, we experienced a number of challenges to lifespan data interpretation." As discussed in the major comments, unless the authors re-perform and re-analyse the lifespan assays this problem will persist.
10. Line 514: As discussed above, the authors cannot rule out the effect of other pathogens on lifespan.

Significance

The crosstalk between AMPs and ageing is a long-standing contradictory topic in the immunology and ageing field (Loch et al., PLoS One, 2017; Badinloo et al., Arch. Insect Biochem. Physiol., 2018; Garschall et al., F1000Res., 2018). This work is a potential step in determining whether and how AMPs regulate ageing. This research may open a door toward yet uncharacterized AMP-independent mechanisms of innate immune signalling in the ageing process.

I am familiar with Drosophila ageing and its relationship with innate immunity, although I am not an insect immunologist.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

RC-2022-01632

Answers to referees

First of all, we wish to thank the 3 referees for their careful evaluation of the manuscript. We see many issues that they have raised as legitimate and have tried to provide experimental or editorial answers. In contrast, some issues are presently addressed in the context of a future manuscript and we had rather not introduce these studies in the revised version.

Below, one will find the answers and the description of the revisions already introduced in the revised manuscript (questions are recalled in blue italics).

New and modified figures, plus not shown figures and tables are indicated in the text below but could not be pasted in the document and can be found in the Revision plan.

Referee # 1

Evidence, reproducibility and clarity

They then delivered 86/8 and LSBio anti-En1 antibodies, that catch En1 in the cleft and prevent it from being captured by MNs.

Perhaps we were not clear. We did not deliver the antibodies 86/8 and LSBio, we used them for western blots and immunohistochemistry (IHC) to identify EN1 and localize it. We delivered the third antibody, a single-chain anti EN1 antibody (scFvEN1), that captures extracellular EN1 and prevents it from being captured by MNs on the basis of the LSBio staining (Figure 4A-C).

Finally, heterozygotes revealed also a degeneration in dopaminergic neurons within midbrain similar to the one observed in spinal MNs, along with an upregulation of SQTSM1/p62 gene/protein, a factor in MN ageing linked to the classical genes implicated in familial forms of ALS (SOD1, TDP-43, FUS, and C9ORF72).

This is a fair comment/work description, that does not require answers.

Significance

*Major comments: *

It is unclear why levels of intensity for RNAscope were not quantified, and qPCR was preferred for quantifications in Figure 1b. RNAscope is a technique that allows for spatial distribution analysis of the markers and their level of the expression. This data can be easily quantified utilizing the QuPath software which is open access. Same concerns apply to Figure 2a.

Quantitative RT-PCR provides a quantitative measure of gene expression. Since only V1 interneurons (including, Renshaw cells) express EN1, we infer the spatial distribution, although not expression level cell by cell. Figure 2A is an actual counting at 4.5 months of En1+ cells and of Calbindin+ cells (Renshaw cells), both identified by RNAscope. Thus, it is clear that the number of En1-expressing cells (V1 interneurons) is not modified at 4.5 months when muscle weakness and death of aMNs are well advanced (around 70% of the aMNs that will eventually die, are already gone). Long-term survival of V1 interneurons is further demonstrated in Figure 2D (left panel) until 15.5 months, (see also below) whereas total En1expression is reduced by half. Quantification neuron by neuron of the amount of En1 transcribed (RNAscope) would indicate the variation, among interneurons, of En1 transcription in WT and mutant mice. This is interesting per se but would not modify the main information that these neurons do not die in the heterozygote and that En1 transcription does not decrease with time in both WT and mutant genotypes (at least until 15.5 months).

*Antibodies should be validated utilizing a reporter mouse. En1cre mice are commercially available and can be crossed with reporters (TdTomato or YFP mice). Utilizing this tissue En1 antibodies can be easily validated. The EN1 antibody shown in Figure 1c seems unspecific, staining several neuronal populations in the spinal cord. *

Indeed, antibody validation is extremely important. LSBio is commercial (CliniSciences), 86/8 was developed in the laboratory and fully characterized and used in previous studies (e.g. Alvarez-Fischer et al. Nature Neurosci. 14: 1260-1266, 2011; Rekaik et al. Cell Reports 13: 242-250, 2015; Blaudin de Thé et al. EMBO J. 37: e97374, 2018), scFv against EN1 was prepared from the 4G11 hybridoma (Developmental Hybridoma Bank, Iowa City, USA) and validated in previous studies (e.g. Wizenmann et al. Neuron 64: 355-366, 2009). In the present study, the two polyclonal were further validated inseveral ways.

In the WBs we compared ventral midbrain (VMB) and spinal cord (SC) tissues and found similar patterns. Strong evidence for antibody specificity is immunostaining extinction with the antigen and with absence of first antibody, which we carried out.

We have now used LSBio and 86/8 to perform a WB on spinal cord (SC) and ventral midbrain (VMB) extracts with or without the first antibody and we find that the absence of first antibody fully eliminates band staining. The western has been introduced in the revised manuscript in place of the cross immunoprecipitation.

Finally, we have quantified EN1 in the aMNs of the heterozygote at 3 months (before cell death), showing that EN1 content is decreased by approximately 2-fold (LSBio antibody) in both a and gMNs with no change in neuron number. This result demonstrating that EN1 is diluted by approximately twofold (concentration per neuron when all neurons are still present), in addition to further validating the antibody, is itself interesting and has been introduced in the revised manuscript as Supp. Fig. 1A.

Regarding the staining in other neuronal populations, there is always some background, in particular in the tissue treatment conditions used for RNAscope. Furthermore, given the large number and wide distribution of V1 interneurons (Fig. 1A), we cannot preclude that EN1 is present at a low concentration in the extracellular space and in several cell types (discussed in Fig. 9 of the manuscript). This does not weaken the main conclusion that it primarily accumulates in MNs which do not express En1 (RNAscope).

*Investigations of En1 expression in motor neurons from already available omics data sets would support the idea that En1 is expressed in motor neurons. *

The En1 locus is silent in MNs. Microdissection of MNs and proteomic analysis would not be definitive since the interneurons that produce EN1 are in close vicinity of the MNs and since some protein is necessarily present in the extracellular space (where it is trapped by scFvEN1), making contamination unavoidable.

Differentiation between Gamma and Alpha motor neurons should be performed using specific markers as Err3, Wnt7a or NeuN.

This is a possible way to do the distinction, but size criterion in Cresyl violet is supported in the literature (Wu et al. Journal of Biological Chemistry, 287: 27335-27344, 2012; Dutta et al. Experimental Neurology, 309: 193-204, 2018). In our study, it is further validated by the demonstration that, in 9-month-old animals, the results obtained (cell number and specific death of large neurons >300µm2, but not of intermediate size ones 200-299µm2) are replicated by counting ChAT-stained neuron (Figure 2C). It is of particular interest that the number of medium size neurons (also ChAT-positive medium size MNs) does not increase when the number of large size (Cresyl and ChAT-positive) neurons decreases, thus precluding a “shrinkage effect”. Most importantly, the size criterion (Cresyl violet) allows us not to be mistaken by a possible down-regulation of markers in the mutant, independently of cell survival. We provide for the reviewer (Revision plan) but not for publication, the evolution with time of the number of neurons based on size (above 200 µm2) showing clearly that at 15.5 months the large population (>300 µm2) is decreased in the En1-Het, with very little change for neurons between 200 and 300 µm2, and certainly not an increase which would be expected if shrinkage occurred.

We were indeed surprised by this finding and a plausible explanation is that a lower metabolic activity makes interneurons less sensitive to stress than aMNs which have to “fuel” long axons and high firing rates (not the case for gMNs). We propose this explanation in the discussion and make it clearer in our revised version. We agree that it is speculative and that the point raised by the reviewer is very interesting. We hope to address this in the future and have discussed this point.

Since the cells do not die, we did not look for signs of apoptosis.

We analyze lumbar sections from L1 to L5 as now indicated in the methods section in the manuscript

The set of experiments reported in Figure 4 is of difficult interpretation without showing the actual presence of extracellular En1, that could be assessed with protein detection or RNAscope.

This is another interesting suggestion, but we think that it will be difficult to distinguish low extracellular staining due to EN1 diffusion from some unspecific background. Since the scFvEN1 is secreted by astrocytes, it necessarily neutralizes extracellular EN1, resulting in a decrease in the MN content of the protein. This is an experiment with high specificity since the same scFv harboring a Cysteine to Serine point mutation that prevents EN1 recognition (no disulfide bound formation between the light and heavy chains) does not block EN1 capture by MNs (Fig. 4C for IHC and quantifications).

As for extracellular EN1 mRNA identified by RNAscope, we hesitate to embark on the idea as mRNAs are likely secreted in insufficient amounts to be identified, even by RNAscope. The results that we have (no En1 visible by RNAscope in MNs, loss of EN1 in MNs following extracellular scFvEN1 activity, and preferential addressing of injected EN1 to MNs) demonstrate EN1 capture by MNs. Indeed, we cannot completely preclude the transfer of tiny amounts (escaping RNAscope detection in MNs) of En1 mRNA (for example, through extracellular vesicles), but we plead for not considering this hypothesis in the present paper. However, if the reviewer wishes, the possibility can be introduced in the discussion.

Referee 2

Evidence, reproducibility and clarity

In general, most of the experiments shown in this study are well done and convincing. However, the data on p62 upregulation appear correlative and do not allow any conclusions about the mechanism and function how EN-1 modulates motoneuron survival and function. In addition, this study is not very precise on the mechanisms how motoneurons degenerate in this model so that there are only limited insights into the way how EN-1 acts on motoneurons in a physiological manner and under pathophysiological conditions.

This criticism is justified, at least in part, as we agree that p62 upregulation is correlative. However, the fact that the neutralization of extracellular EN1 by the scFv increases p62 expression, is in favor of a causative link. The increase is also seen at 3 months in the En1-Het when all aMNs are still present but not after, which is interesting because, due to aMNs death, surviving MNs receive more EN1, information provided below and now introduced and discussed in the revised manuscript (Supp. Fig. 1B).

As for p62, and as also mentioned by referee 3, Fig. 8 is very hard to follow and we propose to simplify it to make the message clearer:

We have revised Fig. 8C, D in which we focus exclusively on SQTSM1/p62 mean expression (see revision plan)

A second information is that a difference in mean p62 expression between WT and Het is seen only at 3 months in aMNs. For aMNs, we propose that this is due to the fact that they are very sensitive to EN1 dosage (in contrast with gMNs which do not die in the En1-Het). At 3 months, aMNs have only half of their normal EN1 content. Later, at 4.5 months 75% of the aMNs bound to die are already dead (Fig. 2D) and the remaining neurons receive more EN1 (even more so at 9 months), as could be measured (see above Supp. Fig. 1B). We thus can propose an accelerated aging of aMNs at 3 months due to both EN1 decrease and high metabolic activity (higher than in gMNs).

In the case of the scFv, scFvEN1, but not the mutated version induces enhanced mean p62 expression in the 80% surviving aMNs and in gMNs at 7 months (low aMN death in this model, see Fig. 4F). As can be seen also in a newly added figure (Supp. Fig. 2) that has been introduced in the revised manuscript and is shown below, 7-month-old scFv animals and 3- to 3.5-month-old En1-Het have similar phenotypes. This mild scFv phenotype (a-MN death and muscle strength loss) in 7-month-old mice in spite of a huge loss in the EN1 content of MNs (Fig. 4C) suggests that the En1-Het phenotype is not entirely due to the decrease in EN1 transport from V1 interneurons to MNs (see discussion and Fig. 9).

It remains true that we have voluntarily decided not to examine in depth the molecular mechanisms allowing EN1 to exert its protective activity, a decision that we would like to defend and maintain.

A first reason is that in previous papers on mesencephalic dopaminergic (mDA) neurons (Alvarez-Fischer et al. Nature Neurosci. 14: 1260-1266, 2011; Rekaik et al. Cell Reports 13: 242-250, 2015; Blaudin de Thé et al. EMBO J. 37: e97374, 2018), we evaluated several mechanisms involved in EN1 neurotrophic activity and we did not want this study to be a duplication of studies done on a different neuronal population, even if mechanisms might differ in part, between aMNs and mDA neurons. What has interested us more is that, in the two cases, age is an important factor in the unveiling of the degeneration phenotype (mDA neurons start dying at 1.5 months and aMNs at 3 months). It is because of this similarity that we performed the bioinformatic study that has led us to SQTSM1/p62. In this context, it is of interest that mean SQTSM1/p62 expression (variability of expression between neurons is not discussed in the revised version) increases with age in the wild type, thus can be seen as an age marker. It allows us to propose that EN1 extracellular neutralization and the loss of one En1 allele, that increases mean SQTSM1/p62 expression accelerate aging.

A second reason is that the study is oriented toward a possible use of EN1 as a therapeutic protein. This orientation also has to do with the focus on SQTSM1/p62. Indeed, there are probably many pathways downstream of EN1, but in the bioinformatic analysis of genes differentially regulated in WT and En1-Het mDA neurons and also expressed in MNs, SQTSM1/p62 is the only one that interacts with the 4 genes mutated in the major ALS familial forms. In addition, SQTSM1/p62 mutations have been observed in ALS patients (References 41 to 45 in the manuscript).

Finally, the most important point is that the main message of this paper is the discovery of a non-cell autonomous EN1 activity in the spinal cord and of its ability to travel between V1 interneurons and MNs. This specificity best explained by a targeting signal that we have identified is at the basis of the specific addressing to MNs of EN1 intrathecally injected, which also has implications for its potential therapeutic use.

Specific points of criticism

1. • In Fig. 2a, the authors show that EN-1-positive interneurons are not reduced at 4.5 months in the spinal cord. No data are shown for later time points such as 9 months, the corresponding stage when motoneuron loss is observed, or at 16 months which corresponds to the data shown in Fig.1. The argument that there is no reduction of V1 interneurons between 4.5 months and 16 months because there is no decrease of EN-1 expression between 4.5 and 16 months, as shown in Fig. 1b is not convincing. EN-1 expression could change in individual cells, thus compensating for the loss. Data on numbers of EN-1-positive cells at 9 and 16 months should be included, and a potential autocrine effect of EN-1 on V1 interneurons, as observed in midbrain dopaminergic neurons, characterized in more detail. * Fig. 2A illustrates the absence of interneuron loss at 4.5 months, but this set of data is completed by those of Fig. 2D that demonstrate the maintenance of V1 interneuron number until 15.5 months, at least. It can be noted that, in contrast with interneurons, aMNs at 4.5 months have experienced massive cell death (70% approx. of total aMN death at 15.5 months). As a whole, data of Fig. 2 demonstrate that the number of small neurons (100-199 µm2) and intermediate size neurons (200-299 µm2) does not change with age, at least through 15.5 months. This is in strong contrast with large aMNs (>300 µm2). As already explained in our answers to referee 1, size is an excellent marker for the identification of neuronal subtypes and the analysis of survival (See answers to referee 1, justifying the use of neuron size).

2. In Fig. 2e, the authors present data on loss of muscle strength between 4.5 and 15.5 months. They conclude that this reflects gradual neuromuscular strength loss. Since neuromuscular endplates have a very high safety factor, they can maintain full function even if transmitter release is reduced by more than 80%. Therefore, the loss of muscle strength seems to reflect the progressive loss of presynaptic terminals at neuromuscular endplates, rather than a gradual loss of neuromuscular strength. *

We apologize for the semantic confusion. What is measured is a progressive loss of muscle strength due to the progressive loss of presynaptic terminals and not a gradual loss of neuromuscular strength. This is now modified throughout the revised text.

• More detailed data on NMJ morphology should be included. How does EN-1 modulate neuromuscular endplates? Is EN-1 located at neuromuscular endplates after being taken up from motoneurons? Even if the mechanism is indirect, via upregulation of p62 under conditions when EN-1 signaling is reduced, does this situation lead to enhanced localization of p62 at neuromuscular endplates? *

We do not see expression of En1 mRNA or the presence of EN1 protein at the level of the endplate (Supp. Fig. 3 in revision plan)

• The data shown in Fig. 3 on changes in NJM morphology appear incomplete and not convincing. As SV2a is not a good marker for changes in presynaptic compartments since it does not allow conclusions on how many synaptic vesicles are released, additional markers for presynaptic active zones such as Bassoon, Piccolo, Munc-13 should be studied. The analysis of fully occupied endplates appears arbitrary, and the differences are relatively small. Additional EM pictures and quantitative analyses of active zone proteins in the presynaptic compartment would help to support the argument of the authors that presynaptic compartments degenerate before cell bodies are lost in EN-1 +/- mice. *

SV2a and NF staining (it is not only SV2a) at the level of endplates identified by a-Bungarotoxin labeling has been used in a large number of studies (Wahlin et al. J. Comp. Neurol. 506: 822-837, 2008; Hasting et al. Scientific Reports 10: 1-13, 2020; Yahata et al. J. Neurosci. 29: 6276-6284, 2009 ; Jones et al. Cell Reports 21: 2348-2356, 2017) Our goal was not to document the loss of synaptic activity through the use of the three suggested markers, Bassoon, Piccolo and Munc-13. Doing it would force us to initiate experiments taking several months to prepare the material and do a quantitative analysis in the models of EN1 loss of function (En1-Het) and neutralization (scFv), plus rescue by EN1. Nor do we wish to initiate a novel collaboration to produce a quantitative ultrastructural study. We see the latter morpho-functional studies beyond the scope of the manuscript and wish to be given the possibility to present them in a separate study (see below in “Description of the experiments that the authors prefer not to carry out”).

The distinction between fully occupied, partially occupied and denervated endplates is not arbitrary and we apologize for not having sufficiently described the methodology. As illustrated in modified Fig. 3 and explained in Material and Methods, a fully innervated endplate is defined as an endplate in which 80% or more of the green pixels (a-BGT) are covered by a red pixel (SV2a), a partially one is between 20 and 80% and a denervated one below 20% coverage. Thus at 9 months and later ages, close to 30% of the endplates are either partially innervated or denervated. In fact, it is more likely that they are partially innervated since the number of AChR clusters does not change (totally denervated clusters normally dissolve). The 80% threshold for fully innervated was selected to give a margin of security, and it is likely that the percentage of 25 to 30% of partially innervated endplates is an underestimation.

In the Revision plan is presented a table with the calculations and modified Figure 3.

We agree that we were not clear enough in our description and that it may have given the impression that the differences were relatively small. We think that retrograde degeneration is strongly supported by a loss of muscle strength that parallels the decrease in fully occupied endplates (a-BGT, NF, SV2a) and precedes aMN loss by more than 1 month. We have recently contacted an electrophysiology group to establish a collaboration that will allow us to follow functional changes at the level of the spinal cord and of the neuromuscular junction and we see the experiments proposed by the reviewer as complementary to these physiological approaches. Yet, we do not want to ignore the opinion of the reviewer and mention it in the conclusion, on the basis of his/her comment.

• The authors present evidence for a glycosaminoglycan (GAG) binding domain that appears responsible for uptake of EN-1 into motoneurons. However, it is unclear into which cellular compartment EN-1 is taken up after GAG binding on motoneurons. The authors propose this could be an alternative pathway to conventional endosomal uptake. How can the EN-1 that is taken up into cells exert transcriptional effects in motoneurons? As a minimum, more data on the subcellular distribution of endocytosed EN-1 should be included to support current hypotheses and to close the gap from cellular uptake to transcriptional regulation. *

The question is justified since we did not recall until page 12 of the Discussion that EN1 is, as most tested homeoprotein transcription factors, captured by a mechanism distinct from endocytosis. While not yet fully understood, the process involves the formation of inverted micelles that allow for direct targeting to the cytoplasm and from there to the nucleus thanks to the NLS. We now mention in the introduction that EN1 transfer and HP transfer is based on unconventional secretion and internalization processes.

• The differences in p62 expression with age in WT and EN-1 +/- mice as shown in Fig. 8c are not convincing. First, the p = 0.0499 and p = 0.0536 values for differences at 3-4 months of age appear borderline, and it is unclear what the dispersion analysis that is shown really means. Moreover, the question remains how a potential dysregulation of p62 then affects NMJ morphology and function. Is this change in p62 also detectable in presynaptic compartments? *

We agree that p values in the range of 0.05 are not extremely high and this is due to the heterogeneity in SQTSM1/p62 expression, that reflects that of MN populations, and induces a high variance. We also agree that this figure is too complicated and a simplified version has been proposed above (see answers to reviewer 1). To summarize, Fig. 8C shows that in WT animals, with no aMN death (grey) the level of SQTSM1/p62 expression in aMNs and gMNs increases between 3 and 4.5 months and between 4.5 months and 9 months, with significances varying between pThe new Fig. 8 panel D (please see above, answers to referee 1) now includes the results obtained with the scFvs. A phenotype comparison between the two models (En1-Het and scFvEN1) has been introduced in Supp. Fig. 2 (see above).

We have no evidence that EN1 modulates the SQTSM1/p62 promoter directly. The identification of this gene as a target (not necessarily a direct target) of EN1 comes from the bioinformatic analysis described in the manuscript and we were intrigued by the interaction with the 4 main familial ALS mutations and the existence of families with SQTSM1/p62 mutations. This is what led us to analyze its expression in our two models of EN1 loss of function. Although the En1-Het mouse is not an ALS model, the results support the idea that EN1 could be used as a therapeutic protein in several familial and even sporadic forms of the disease. The latter hypothesis is now being tested on MNs derived from iPSCs (sporadic patients, fALS and isogenic variants, and healthy controls). If the data lend weight to our hypothesis, as collaborative and in-house preliminary data suggest, then a complete analysis of EN1 targets in human MNs will be undertaken. Again, we really think that this is out of the scope of this study.

For Fig. 8, we fully agree that it can give headaches and we apologize. Moreover, it induces wrong interpretations (mean intensity increases with age and dispersion between 4.5 and 9 months has a calculated p__Referee #3__

Evidence, reproducibility and clarity

Nevertheless, the connection between EN-1 and p62 is not well developed by the data presented and future readers may be left with many questions regarding how EN-1 and p62 are related (e.g. direct interaction? transcriptional regulation?), whether p62 is indeed the mediator of EN-1 trophic effects, or the significance of the increased levels of p62 for motoneuron disease

The reviewer is right and we have tried to better explain and to simplify. Please see responses to referees 1 and 2.

*Figure 1C: There appears to be EN1 immunoreactivity (green) in several areas of the spinal cord, including dorsal regions. Can the authors clarify what that labeling could be representing? *

Unfortunately, there is always some background staining, in particular in the tissue treatment conditions appropriate for RNAscope. Furthermore, given the large number and wide distribution of V1 interneurons (Fig. 1A), we cannot preclude that EN1 is present at a low concentration in the extracellular space and in several cell types (now represented in Fig. 9). This does not weaken the main conclusion that it primarily accumulates in MNs which do not express En1 (RNAscope).

*Figure 1D: These immunoprecipitation results lack a negative control with irrelevant antibody to confirm that the band shown it's being recognized specifically by the antibodies reacting with the blot. *

Please see the response to reviewer 1 above with the Western blot and the absence of staining on a WB in absence of first antibody (86/8 or LSBio).

F*igure 1E: The intensity of the EN1 labeling in MNs, much stronger than in V1 interneurons, is intriguing, given that MNs do not express engrailed-1 mRNA. One would have expected the opposite. It may help here if it was possible to show that immunoreactivity in MNs is diminished in the het mutant mouse. *

We also were surprised by this intensity higher in MNs than in V1 interneurons, as if the protein was exported rapidly towards the target neurons. We have done the experiment proposed by the referee, found a twofold (approx.) immunoreactivity reduction in En1-Het MNs (see above Supp. Fig. 2A in answers to referee 2). This supplemental figure has been introduced in the revised version. The experiment was done at 3 months when no MN death has yet occurred. Later the neurons “replenish” with EN1, probably because they do not have to share the limited supply with the dead ones (see above answers to referee 2 and Supp. Fig. 2B).

*Figure 2D: There are a few possible problems with these data and their interpretation. First, this reviewer feels that 5 neurons (y-axis) is a rather small number. Are these 5 neurons per what area? From how many mice? I did not find that information in the figure legend. A larger area should be quantified so that we get numbers that are more robust. Second, such differences could also be due to hypotrophy of the MNs, namely, that MN number is the same but they are smaller. *

The differences cannot be attributed to hypotrophy. A first reason is that, at 9 months, the Cresyl violet and ChAT staining give the same results for medium size and large neurons (Fig. 2C). Furthermore, when one counts the cells throughout 15.5 months, the decrease in the number of large neurons is not compensated by an increase in the number of medium size or small ones. The reasoning and a graph, not intended for publication can be found in answers to referee 1.

*Figure 3A: It would be useful that the authors explain how these AChR clusters were defined, visualized and counted. I could not find this information in the Methods. Perhaps this could be done by showing an alpha-BTX image illustrating the clusters. *

We fully agree that the procedure was not well explained and we have introduced a correction in the Material and Methods section. For more details, please see answers to referee 2.

*Figure 3B: As each adult endplate is only innervated by one MN, one would have expected fewer clusters and/or endplates, if indeed MNs are missing in this mouse, rather than endplates that are partially occupied. This could be clarified a bit more explicitly. *

This is true and the ambiguity takes its origin in insufficient explanation of how fully innervated, partially innervated and denervated endplates were defined. Please see above and also in answers to reviewer 2. Modifications have been introduced in the text and in Fig. 3. The referee is right, the absence of change in the number of AChR clusters suggests that there are very few fully denervated endplates and that what is defined as such in the analysis corresponds to partially innervated endplates (see above). This is now discussed in the text.

Figure 6B: Would not be better to do this with a virus, like in the case of the antibody? A more robust effect on MN survival may be attainable and thus strengthen the concept.

This would be another interesting experiment and we are presently exploring this possibility (with preliminary results). The choice of the virus and of the promoters is very important. We are comparing several AAVs, including AAV2, AAV2-TT (which diffuses better) and AAV8. For the promoter, we do not want to express within MNs as the imported protein might have special properties, associated with import. V1 interneurons would be best, but we have to verify if this does not modify V1 physiology. Astrocyte is another option, but with a similar pitfall. This means that we have a long way to go before proposing a “gene therapy” approach.

In addition, in the context of future clinical studies, we were eager, on the basis of the long-lasting activity of the protein already observed in the mesencephalic dopaminergic neurons (Alvarez-Fischer et al. Nature Neurosci. 14: 1260-1266, 2011; Rekaik et al. Cell Reports 13: 242-250, 2015; Blaudin de Thé et al. EMBO J. 37: e97374, 2018), to try a protein therapy in the spinal cord. Interestingly, the effects are also long-lasting in the spinal cord, (12 weeks in the mouse before a second injection is needed) and, according to contacted physicians, intrathecal injections, every second month or even more frequently, could be envisaged in the human. In that case, protein injection is possibly advantageous for the following reasons:

(i) viral particles can travel far and we do not know what would be the side effects.

(ii) the protein is short-lived but specifically addressed to MNs (thanks to the presence of EN1 binding sites at their surface), thus minimizing the issues associated with permanent expression and side effects.

(iii) EN1 is a natural protein normally secreted and the immune system might not be solicited as much as with viral approaches.

*Figure 7A: The protein seems to be mainly in the cytoplasm of those cells (nuclei are dark and unlabeled), which is also unusual for a transcription factor that functions in the nucleus. Also surprising that the protein is gone in 3 days, but has effects over 24 weeks. Any explanation for that? *

The protein is imported and is thus both in the cytoplasm where it exerts an effect on protein translation (Brunet et al. Nature 438: 94-98, 2005; Alvarez-Fischer et al. Nature Neurosci. 14: 1260-1266, 2011; Yoon et al. Cell 148: 752-764, 2012) and in the nucleus where it exerts its transcriptional and “epigenetic activity (see below for the latter). In fact, different antibodies and fixation procedures can favor cytoplasmic or nuclear staining. When nuclear, the dark point at the center, probably the nucleolus is less stained.

Two images illustrating this point are shown in the revision plan.

For the second part of the question, three days are sufficient for a long-lasting activity. This was also observed in the midbrain where the protein restores the epigenetic marks jeopardized by an acute oxidative stress (Rekaik et al. Cell Reports 13: 242-250). This has led to the hypothesis that EN1 has an important action at the level of the structure of the heterochromatin, thus a long-lasting “epigenetic” activity. We are presently working on the latter effects on the chromatin structure using human MNs derived from iPSCs (patients and control).

*Figure 7B: It's not clear what the blue and red bars mean, as this is not explained in the legend. Also, the y-axis says "%Chat+" suggesting they are counting MNs, but in the text they talk about EN-1 capture. If the latter, the y-axes should indicate % EN-1 over Chat, or something like that. In general, better figure legends would improve the experience of the reader. *

In this experiment, we wanted to test the presence of a GAG-binding domain in EN1. To test its potential role in EN1 internalization and localization, we co-injected or not the RK-EN1 with hEN1 protein. Then, we counted the percentage of MNs (%ChAT+) which contain, or not, the hEN1 protein (hEN1+ in red or hEN1- in blue), allowing us to verify if the RK-EN1 alters the internalization of the hEN1 protein. So yes, we are looking at the capture of EN1 by the MNs with or without the RK-peptide (or control peptides). We have modified the text to make the point clearer.

*Statistical analyses: In principle, comparisons of data obtained in studies that involved two variable parameters (such as time and genotype/treatment) should be weighted by a 2-way ANOVA test, which is more stringent since more conditions are being tested simultaneously. Usually a t-test is reserved for a pairwise comparison in an experiment involving only two conditions of the same variable. *

The reviewer is correct. The two-way ANOVA is explained in the Statistical analyses section of the Methods. The analyses were carried out and the results listed in the legends for Figs 2, 3, 4, 6 and Supp. Fig. 1.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This is an interesting and provocative manuscript reporting non-cell autonomous trophic activities of a homeobox protein, a concept pioneered by Dr. Prochiantz since many years ago. The study involves a significant amount of experimental work and the authors are to be congratulated by the scope and ambition of their study. Given previous studies by this laboratory on EN-1 functions in midbrain dopaminergic neurons, the concept advanced in the present paper is not entirely novel, although it is indeed interesting to find EN-1 activities in motoneurons; these were unexpected. Given that this is a non-cell-autonomous effect (EN-1 is made and released by neurons adjacent to MNs), it would have been interesting to explore the conditions under which EN-1 synthesis, release and effects are regulated, whether by lesion, degeneration, etc. But that may be something the authors wish to leave for a future report. It is welcome that an effort was put into trying to mechanistically understand how these trophic effects are mediated. This reviewer understands that this is a major undertaking. Nevertheless, the connection between EN-1 and p62 is not well developed by the data presented and future readers may be left with many questions regarding how EN-1 and p62 are related (e.g. direct interaction? transcriptional regulation?), whether p62 is indeed the mediator of EN-1 trophic effects, or the significance of the increased levels of p62 for motoneuron disease. In its present form, this paper will be welcome, if nothing else by the provocative ideas that it advances. For this, it clearly deserves to be published in a good journal (whatever that means these days). Here below are a few questions and suggestions which the authors may want to take into consideration.

Figure 1C: There appears to be EN1 immunoreactivity (green) in several areas of the spinal cord, including dorsal regions. Can the authors clarify what that labeling could be representing?

Figure 1D: These immunoprecipitation results lack a negative control with irrelevant antibody to confirm that the band shown it's being recognized specifically by the antibodies reacting with the blot.

Figure 1E: The intensity of the EN1 labeling in MNs, much stronger than in V1 interneurons, is intriguing, given that MNs do not express engrailed-1 mRNA. One would have expected the opposite. It may help here if it was possible to show that immunoreactivity in MNs is diminished in the het mutant mouse.

Figure 2D: There are a few possible problems with these data and their interpretation. First, this reviewer feels that 5 neurons (y-axis) is a rather small number. Are these 5 neurons per what area? From how many mice? I did not find that information in the figure legend. A larger area should be quantified so that we get numbers that are more robust. Second, such differences could also be due to hypotrophy of the MNs, namely, that MN number is the same but they are smaller.

Figure 3A: It would be useful that the authors explain how these AChR clusters were defined, visualized and counted. I could not find this information in the Methods. Perhaps this could be done by showing an alpha-BTX image illustrating the clusters.

Figure 3B: As each adult endplate is only innervated by one MN, one would have expected fewer clusters and/or endplates, if indeed MNs are missing in this mouse, rather than endplates that are partially occupied. This could be clarified a bit more explicitly.

Figure 6B: Would not be better to do this with a virus, like in the case of the antibody? A more robust effect on MN survival may be attainable and thus strengthen the concept.

Figure 7A: The protein seems to be mainly in the cytoplasm of those cells (nuclei are dark and unlabeled), which is also unusual for a transcription factor that functions in the nucleus. Also surprising that the protein is gone in 3 days, but has effects over 24 weeks. Any explanation for that? Figure 7B: It's not clear what the blue and red bars mean, as this is not explained in the legend. Also, the y-axis says "%Chat+" suggesting they are counting MNs, but in the text they talk about EN-1 capture. If the latter, the y-axes should indicate % EN-1 over Chat, or something like that. In general, better figure legends would improve the experience of the reader.

Statistical analyses: In principle, comparisons of data obtained in studies that involved two variable parameters (such as time and genotype/treatment) should be weighted by a 2-way ANOVA test, which is more stringent since more conditions are being tested simultaneously. Usually a t-test is reserved for a pairwise comparison in an experiment involving only two conditions of the same variable.

Significance

see above

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Engrailed-1 does not act only in a cell-autonomous way in neural development, but also has non-cell-autonomous functions. These functions depend on the release of this homeoprotein which has been characterized in much detail by previous work of this group. In this paper, they show that EN-1 is expressed in spinal V1 interneurons, both on the RNA and on the protein level. In spinal motoneurons, EN-1 protein but not RNA is detected. Neutralization of extracellular EN-1 with a secreted antibody apparently blocks transfer from these interneurons to motoneurons and causes motoneuron disease symptoms. A similar phenotype is also observed in EN-1 +/- mice. Most importantly, the authors also demonstrate that intrathecal injection of EN-1 into EN-1 +/- mice restores loss of muscle strength and prevents motoneuron death. The authors also show that the autophagy modulator SQTSM1/p62 is expressed at elevated levels in EN-1 +/- mice and in mice after injection of the EN-neutralizing antibody. Since p62 expression also seems to be increased in general during aging in motoneurons, the authors conclude that EN-1 from spinal V1 interneurons is a regulator of motoneuron aging. In general, most of the experiments shown in this study are well done and convincing. However, the data on p62 upregulation appear correlative and do not allow any conclusions about the mechanism and function how EN-1 modulates motoneuron survival and function. In addition, this study is not very precise on the mechanisms how motoneurons degenerate in this model so that there are only limited insights into the way how EN-1 acts on motoneurons in a physiological manner and under pathophysiological conditions.

Specific points of criticism:

1. In Fig. 2a, the authors show that EN-1-positive interneurons are not reduced at 4.5 months in the spinal cord. No data are shown for later time points such as 9 months, the corresponding stage when motoneuron loss is observed, or at 16 months which corresponds to the data shown in Fig.1. The argument that there is no reduction of V1 interneurons between 4.5 months and 16 months because there is no decrease of EN-1 expression between 4.5 and 16 months, as shown in Fig. 1b is not convincing. EN-1 expression could change in individual cells, thus compensating for the loss. Data on numbers of EN-1-positive cells at 9 and 16 months should be included, and a potential autocrine effect of EN-1 on V1 interneurons, as observed in midbrain dopaminergic neurons, characterized in more detail.
2. In Fig. 2e, the authors present data on loss of muscle strength between 4.5 and 15.5 months. They conclude that this reflects gradual neuromuscular strength loss. Since neuromuscular endplates have a very high safety factor, they can maintain full function even if transmitter release is reduced by more than 80%. Therefore, the loss of muscle strength seems to reflect the progressive loss of presynaptic terminals at neuromuscular endplates, rather than a gradual loss of neuromuscular strength.
3. More detailed data on NMJ morphology should be included. How does EN-1 modulate neuromuscular endplates? Is EN-1 located at neuromuscular endplates after being taken up from motoneurons? Even if the mechanism is indirect, via upregulation of p62 under conditions when EN-1 signaling is reduced, does this situation lead to enhanced localization of p62 at neuromuscular endplates?
4. The data shown in Fig. 3 on changes in NJM morphology appear incomplete and not convincing. As SV2a is not a good marker for changes in presynaptic compartments since it does not allow conclusions on how many synaptic vesicles are released, additional markers for presynaptic active zones such as Bassoon, Piccolo, Munc-13 should be studied. The analysis of fully occupied endplates appears arbitrary, and the differences are relatively small. Additional EM pictures and quantitative analyses of active zone proteins in the presynaptic compartment would help to support the argument of the authors that presynaptic compartments degenerate before cell bodies are lost in EN-1 +/- mice.
5. The authors present evidence for a glycosaminoglycan (GAG) binding domain that appears responsible for uptake of EN-1 into motoneurons. However, it is unclear into which cellular compartment EN-1 is taken up after GAG binding on motoneurons. The authors propose this could be an alternative pathway to conventional endosomal uptake. How can the EN-1 that is taken up into cells exert transcriptional effects in motoneurons? As a minimum, more data on the subcellular distribution of endocytosed EN-1 should be included to support current hypotheses and to close the gap from cellular uptake to transcriptional regulation.
6. The differences in p62 expression with age in WT and EN-1 +/- mice as shown in Fig. 8c are not convincing. First, the p = 0.0499 and p = 0.0536 values for differences at 3-4 months of age appear borderline, and it is unclear what the dispersion analysis that is shown really means. Moreover, the question remains how a potential dysregulation of p62 then affects NMJ morphology and function. Is this change in p62 also detectable in presynaptic compartments?
7. Is there any molecular evidence that EN-1 modulates the p62 gene promoter directly? What is the argument to assume that increase in SQTSM1/p62 expression and dispersion is an indicator of aging? The mean intensity, if I understand Fig. 8c correctly, does not significantly increase, it is only the dispersion that changes. In general, the data shown in Fig. 8c are hard to read and interpret. For example, in the right panel, the difference between the dispersion in 4.5 and 9 month old EN +/- mice is indicated as p = 0.06, but marked with 4 stars. The presentation of these data should be changed to make them clearer.

Referees cross-commenting

I agree with all comments from the other reviewers

Significance

This study expands previous work of the authors, in particular work that has been performed and published on the effects of EN-1 on mesencephalic dopaminergic neurons. If adequately revised, it could make an interesting contribution to the general understanding how spinal V1 interneurons act on funcitonality and survival of spinal motoneurons.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The present work focuses on Engrailed 1 (En1), a homeoprotein that is expressed in spinal V1 interneurons that connect to α-motoneurons (MNs). The authors studied its role in neuromuscular strength and MN retention and loss with the aid of different approaches. First, they studied its expression in spinal cord with RNAscope, a novel ISH method that makes possible to detect biomarkers that would be otherwise difficult to study with traditional ISH techniques. They then delivered 86/8 and LSBio anti-En1 antibodies, that catch En1 in the cleft and prevent it from being captured by MNs; moreover, they used a heterozygotic En1 mouse model to reduce En1 levels. The behavioral assessment, studied with grip strength, inverted grip test and hindlimb extensor reflex, showed motor alterations, paralleled by α-MNs loss and with an even stronger phenotype in the heterozygotic mice. This phenotype, however, appeared weeks before the MN loss, so they used NMJ assessment to determine what it seems to be a retrograde degeneration. En1 administered intrathecally was effectively internalized by the MNs and led to a long-term amelioration of the motor impairments and renervation of the NMJ, that needed to be boosted after 12 weeks for a stable therapeutic effect. Finally, heterozygotes revealed also a degeneration in dopaminergic neurons within midbrain similar to the one observed in spinal MNs, along with an upregulation of SQTSM1/p62 gene/protein, a factor in MN ageing linked to the classical genes implicated in familial forms of ALS (SOD1, TDP-43, FUS, and C9ORF72). They authors did not observe degeneration in V1 interneurons. They conclude that En1 might have a role in regulating MN ageing in degenerative motor disorders.

Significance

Overall, the manuscript is well written however, some of the data appears too preliminary for publication. While the potential beneficial effect of En1 intrathecal administration looks promising and worth of publication, it is difficult to understand the mechanism of action. Some of the results are puzzling and require further investigations.

Major comments:

It is unclear why levels of intensity for RNAscope were not quantified, and qPCR was preferred for quantifications in Figure 1b. RNAscope is a technique that allows for spatial distribution analysis of the markers and their level of the expression. This data can be easily quantified utilizing the QuPath software which is open access. Same concerns apply to Figure 2a.

Antibodies should be validated utilizing a reporter mouse. En1cre mice are commercially available and can be crossed with reporters (TdTomato or YFP mice). Utilizing this tissue En1 antibodies can be easily validated. The EN1 antibody shown in Figure 1c seems unspecific, staining several neuronal populations in the spinal cord.

Investigations of En1 expression in motor neurons from already available omics data sets would support the idea that En1 is expressed in motor neurons.

Differentiation between Gamma and Alpha motor neurons should be performed using specific markers as Err3, Wnt7a or NeuN.

How can the authors explain the lack of loss of En1 interneurons in the En1-Het mice? Do spinal En1 interneurons show any signs of apoptosis (e.g., cleaved caspase 3 marker)? Which levels of the spinal cord were used for interneuron quantifications? Segments between L1 and L3 would be preferable.

The set of experiments reported in Figure 4 is of difficult interpretation without showing the actual presence of extracellular En1, that could be assessed with protein detection or RNAscope.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

1. General Statements

We would like to thank the editor for handling our manuscript entitled, “Mouse SAS‑6 is required for centriole formation in embryos and integrity in embryonic stem cells”, and the reviewers for the insightful comments and suggestions to improve our work. We aim for our manuscript to be considered for a “Short Report” format. As such, we would like to emphasize that we did not focus on the in vivo part of our study, where the Sas-6 mutant mouse embryos resemble our previously published Sas-4 mutants, as pointed out by the reviewers, because both mutants lack centrioles. In our opinion, the novelty of our work is evident in the discovery that mouse embryonic stem cells (mESCs) lacking SAS-6 are still able to form centrioles, albeit mostly abnormal, which is also shared by the reviewers. This is in contrast to Sas-4 mutant mESCs for example, which lack centrioles (Xiao*, Grzonka* et al, EMBO Reports 2021), and human cultured human cell lines without SAS-6, which have been shown to lose centrosomes. We are in the process of editing the manuscript and performing additional experiments per the reviewers’ recommendations. Below, we provide a point-by-point description of our revision plan.

2. Description of the planned revisions

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

The article by M. Grzonka and H. Bazzi entitled: Mouse Sas-6 is required for centriole formation in embryos and integrity in embryonic stem cells, describes new findings in novel mouse models of Sas-6 knockouts (KO). This is an interesting study that reports two different mouse Sas6 KO models and the depletion of Sas6 from mouse embryonic stem cells (mESCs). This type of analysis has never been done before and so it reveals and describes a role for Sas-6 in centriole biogenesis in mouse.

We thank the reviewer for highlighting the novelty of our work on the roles of SAS-6 in mice.

• *

The authors compare their analysis with Sas-4 KO and overall found similar results when compared to previous work from H. Bazzi, when Sas-4 was depleted in mouse embryos. Due to the mitotic stopwatch pathway, Sas6KO embryos die during development at extremely early stages and this can be rescued by depletion in p53 and other members of the pathway.

Perhaps, not so surprisingly, these embryos do not contain centrioles, showing that in vivo, Sas6 is absolutely required for centriole duplication. More surprisingly, however, in cultures of mESCs, established and propagated in vitro, Sas-6 crispr induced KO, does not result in lack of centrioles. Instead, abnormal structures that show aberrant morphologies, length, and incapacity to assemble cilia were detected. In principle, this means that centrioles can be assembled independently of Sas-6, even if not in the correct manner.

We again thank the reviewer for astutely pointing out the most surprising finding in our data, which is that mESCs lacking SAS-6 can still form centrioles.

The authors interpret these differences as possible differences in the pathways involved in centriole assembly and propose different requirements in different cell types, within the same species.

I have problems with this interpretation. To me is very difficult to understand, how the "protein" absolutely required for cartwheel assembly at the early stages of centriole biogenesis, can be essential and dispensable at the same time. Although, I may be wrong, I think the authors have not envisage other possibilities to interpret their data, which should be taken into consideration.

We agree with the reviewer that SAS-6 is currently considered in the centrosome field as one of the “core” centriole formation or duplication factors and that it is a major component of the cartwheel scaffold during the early phase of centriole biogenesis. Although, the absence of centrioles in the Sas-6 mutant mouse embryos in vivo supports the essential function of SAS-6, and perhaps the cartwheel, in centriole formation; the mere presence of centrioles in mESCs indicates that SAS-6, and again the cartwheel, is not essential for the existence of centrioles in these cells. Because this is a major finding that we would like to bring across from our study, we will better highlight and clarify it in the new version of the manuscript as described below. In fact, in points #4 and #5, we share the same possible explanation for the difference in the phenotypes between Sas-6 mutant mouse embryos and mESCs as the reviewer.

1) I do not know anything about ESC and ESC cultures. So maybe this is a stupid suggestion. But can't they be derived exactly from the same genetic background of SAS-6KO embryos? Because the way the two (or even 3 as there are 2 mouse KO lines) are generated is different. Why is that?

The reviewer is correctly suggesting that the mESC can be derived from the Sas-6 mutant blastocysts. We have initially derived mESCs from the Sas-6em4/em4 mutants and performed our analyses on the centriole phenotypes in these mutants before realizing that the allele was hypomorphic (SAS-6 staining in Fig. S2F, and the appearance of centrioles at E9 in Fig. S1B). Because the surprising finding in our study is that SAS-6 does not seem to be essential for centriole presence in mESCs, as pointed out by the reviewer, we decided to generate a more convincing Sas-6-/- null allele in mESCs by deleting the entire ORF of Sas-6 (more on this point below). We would also like to direct the attention of the reviewer that we have cultured the blastocysts (E3.5) from the Sas-6em5/em5 null mutants, which as we show lack centrioles at E3.5, and the cells indeed start to form centrioles just 24 h post-culture (Fig. 3C-D).

To build on these findings, we have already taken this a step further and generated a mESC line from the Sas-6em5/em5 mutants. These Sas-6em5/em5 –derived mESCs show CENT2-eGFP-positive centrioles, and we are currently analyzing their number and integrity, similar to our analyses of the CRISPR-generated Sas-6-/- null mESCs.

2) Still on mESCs, are the authors sure that there are no WT Sas-6 mRNAs still present in their ESC cells? Because tiny amounts are maybe sufficient to allow the initial cartwheel structure. In FigS2B, I can see a really faint band, very faint but it is there.

Due to the nature of the surprising finding that Sas-6 mutant mESCs can still form centrioles, we understand the concerns and suggestions of this reviewer and the other reviewers in this regard. We have generated several Sas-6 mutant alleles in mESCs (in exons 2, 4 or 5), in which we used Western blots to check whether they were null alleles or not. We used different commercial (Proteintech cat# 21377-1-AP, Sigma-Aldrich cat# HPA028187 and Santa Cruz cat# SC-81431) and non-commercial (kind gift from Renata Basto, Institute Curie) antibodies. The SAS-6 antibody from the Basto lab gave the most reliable and reproducible results. Using this antibody, and in our own interpretation, we were not able to detect SAS-6 by Western blots in Sas-6 mutant mESCs. We concluded that SAS-6 in mESCs (and mouse embryos, see below) is expressed at low levels. Of note, we always detected centrioles in the different Sas-6 mutant mESCs, even those derived from the Sas-6em5/em5 null mutant blastocysts, which as blastocysts had no detectable centrioles.

For a more definitive knockout in mESCs, we decided to bi-allelically delete the entire Sas-6 ORF DNA from the ATG to the TAA (over 34 Kb of DNA, Fig. S2A). According to the central dogma of molecular biology, when there is no DNA, then there should be no mRNA and no protein. In confirmation of this premise, recent RT-PCR data showed that Sas-6 mRNA is not detectable in these Sas-6-/- null mESCs. Also, immunofluorescence analyses did not detect SAS-6 in these cells. We will add the RT-PCR and immunofluorescence data to the fully revised manuscript. We will also repeat the SAS-6 Western blots to achieve better band resolution.

These Sas-6-/- mESCs started from a single cell and have been passaged up to 20 times by now without losing centrioles. SAS-6 protein was not detectable at the early passages and the mRNA is still not detectable. This is how knockouts have been and are produced. If this mutant is still not convincing, then we respectfully ask that the reviewers provide their own suggestion on what will be more convincing. In our humble opinion, this Sas-6-/- mESCs line can be used to test the specificity of the antibodies in mouse cells and not the other way around.

3) This last point goes also with the western-blot of Figure S2C- there is still a band, very tiny between the two very tick bands (marked with *). Maybe separating proteins better will help visualizing the real Sas-6 band? Have they used the Sas6 ab in other WBs from the KO embryos, for example? Can they use the Sas6 ab in immunostaining to show if the assembled abnormal centrioles completely lack Sas6. This will allow to distinguish between the hypothesis of having some (even if not much) sas6 left?

The answer to these questions is above in point #2. In addition, we have used the Basto lab antibody for SAS-6 for Western blots on mouse embryos, which detect low levels of SAS-6 in controls and no signal in the mutants. We will repeat the SAS-6 Western blots on mESCs to achieve better band resolution. Using this antibody for immunofluorescence showed that the Sas-6em4/em4 mutant is hypomorphic, whereas the Sas-6em5/em5 mutant showed very low, most likely background, staining (Fig. S1F). For mESCs, we decided to delete the entire Sas-6 ORF DNA in mESCs and generate homozygous Sas-6-/- null mutants. Immunofluorescence analyses did not detect SAS-6 in these cells.

4) Then a more theoretical point? Have the authors considered that the difference is more in the stability of the abnormal structures. Let's say, without a cartwheel and maybe enough PLK4 activity and high level of other centriolar components, the centrioles are abnormally assembled- they have no cartwheel, but they are disassembled very fast in the embryo but not in ESCs?

• *

We agree and share the reviewer’s interpretation for the potential requirement of SAS-6 in vivo to stabilize intermediate structures, that is compensated for by other factors in mESCs. This was not directly discussed in the first version of the manuscript and we will include it in the new version.

5) Even if there is a real difference and without Sas-6 ESCs can make centrioles that are abnormal in structure and function (at least at the cilia assemble level), the choice of words "strictly required", I am not sure it is correct. Because, since Sas-6 is described by many studies as the factor required for cartwheel assembly, which occurs very early in the pathway, this means that in mESCs centrioles can assembled without forming a cartwheel. And so that the cartwheel is actually not required for the initial building block, but more as a structure that maintains the whole centriole in an intact manner?

We agree with the reviewer on the likely requirement of SAS-6, and therefore the cartwheel as a whole, for the symmetry and integrity of the forming centrioles, which is along the same line as in point #4. In our interpretation, “centriole formation” does not necessarily mean centriole “initiation” but rather the presence of the centriole as a structure. We will use more appropriate and specific wording to match our shared interpretation with the reviewer.

6) The authors mentioned that in flies, abnormal Sas-6 structures have been described in certain cell types. Are these mutants, null mutants? In other words, do these structures assembled in a context of no Sas6 or abnormal Sas-6 protein or even low levels of Sas-6?

According to the published report (Figure S3B in Rodrigues-Martins et al, 2007, PMID: 17689959) the fly brains have no detectable DSAS-6 protein. Therefore, we assume that they are Sas-6 null fly mutants. The abnormal centrioles in Sas-6 C. Reinhardtii mutants and Sas-6-/- mESCs null mutants support the conclusion that the main role of SAS-6, and perhaps the cartwheel, is in maintaining the integrity of the forming procentriole.

• *

Other points:

I think the 1st sentence of the abstract appears disconnected from the rest. The same goes for the 1st sentence of the introduction. And also, what is the evidence that pluripotent stem cells rely primarily on the proper assembly of a mitotic spindle? They rely on many other things, not sure this is the first one.

The sentences were meant to highlight the importance of cell division in stem cells. We will adjust the wording in these sentences per the reviewer’s comment to not focus on pluripotency per se.

The authors mention that centrioles are lost in Sas6-/- after "differentiation" of mESCs. The term differentiation is not appropriate, and confusing here. Differentiation normally refer to cells that stopped proliferating and exited the cell cycle, which is not the case here, as NPCs are progenitor cells that keep cycling.

We believe the reviewer is referring to “terminal differentiation”, when the cells exit the cell cycle and adopt their destined cell fates. The word “differentiation” in this context refers to limiting the potency of stem cells into a subset of cell fates such as NPCs, which are proliferating progenitors.

Figure S1: Percent of cells with centrosomes was assessed by a co-staining of gtubulin and Cep164, which mark the mother centrioles. As Cep164 may be absent from centrosomes after lack of centriole maturation in sas6-/- embryos, another combination of staining should be performed to evaluate the percent of cells without centrosomes. gtubulin staining can be seen in Sas6 em5/em5 embryos, while the quantification claims total absence of centrosomes. The authors use the CENT2-eGFP transgenic line to count the number of centrioles in Figure 3, they should do the same in Figure S1.

We will follow the reviewer’s recommendation of counting Cent2-eGFP for the assessment of centrioles in Sas-6em5/em5 (Fig. S1).

The g-tubulin (TUBG) aggregates at the poles of dividing cells are assembled in the absence of centrioles, as shown in Sas-6em5/em5 embryo sections (Fig. S1H). In addition, we have previously observed these pericentriolar material aggregates in Sas-4-/- mutant embryos (Bazzi and Anderson, 2014), which do not contain centrioles in serial transmission electron microscopy. Therefore, we do not refer to them as centrosomes in the absence of centrioles at their core.

Reviewer #1 (Significance (Required)):

This study shows with a novel mouse model the requirement of centrioles during mouse development. It will be relevant to centrosome labs, the novel mouse lines will be useful to many labs working on centrioles, cilia and centrosomes.

My expertise: centrosome biology

We thank the expert reviewer for the critical comments and suggestions, and the positive evaluation of our manuscript.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

• *

Here, Grzonka and Bazzi present their work on characterizing the requirement of SASS6 in mouse embryo development and in embryonic stem cell (mESC) culture. In mouse, female and male gametes lack centrioles, and early divisions occur without centrioles. De novo formation typically happens at the blastocyst stage (~E3.5). The authors generated two SASS6 knock-out strains, SASS6 em4/em4 (frameshift deletion, reported as a severe hypomorphic allele), and SASS6 em5-em5 (frameshift deletion, reported as a null allele). Mutant embryos arrest development at mid-gestation unless the p53, USP28 and USP28 pathway is perturbed. As expected, centrioles do not form in SASS6 -/- mice. However, the authors report that de novo formation of centrioles is facilitated in mESC culture conditions for SASS6 CRISPR knock-out mESCs and mESCs derived from SASS6 em5/em5 blastocysts. Centrioles are lost upon differentiation of SASS6 CRISPR knock-out mESCs into neural progenitor cells (NPCs).

The presented study is relevant for scientists investigating the requirements for centriole formation during embryonic development. Further, it provides insights in possibly different requirements for centriole formation between stages of differentiation, as well as differences in in vivo and in vitro models.

We thank the reviewer for finding our work relevant and insightful into the differential requirements for centriole formation depending on the cell type.

The data represented by Grzonka and Bazzi are robust and support the manuscript and conclusions made. However, the study is predominantly descriptive, and the authors do not test the molecular pathway underlying the de novo formation of centrioles observed in SASS6 -/- mESCs. It is generally believed that de novo formation of centrioles is not possible in SASS6 knock-out cells although work from Wang and Tsou with SASS6 a oligomerization mutant suggests otherwise. A dissection of the specific factors required for the de novo formation of centrioles in the mESC context would provide more insights into de novo centriole assembly in general and would increase the impact of this work. I would support publication of the manuscript if the following points are addressed:

We again thank the reviewer for finding the data robust and support our conclusions and interpretation. We agree with the reviewer that our study opens new questions about how mESCs manage to assemble centrioles in the absence of SAS-6. Together with the phenotypes of the Sas-6 mutant D. melanogaster and C. Reinhardtii, and the SAS-6 oligomerization mutants (but not full SAS-6 mutants) in human cell lines mentioned by the reviewer and cited in our manuscript, the data open new investigations into the exact requirements of SAS-6 and the cartwheel in centriole biogenesis in the different cellular contexts.

1. • One of the main figures, ideally Figure 1, should be dedicated to the characterization of the newly generated mouse strains. This should also be elaborated in the text further. I would like to see a schematic representation of the genomic modifications. The SASS6 stainings of wt and Sas-6 knock-outs (now Figure S1F) should be shown in that context as well as the Figures S2A-C. The authors should discuss why there still appears to be SASS6 protein in the SASS6-em5/em5 Sas-6 stainings visible. Also, the western blot, especially the unspecific bands so close to the SAS-6 protein, should be discussed. Adding qRTPCR results would also be good. Per the reviewer’s requests, we will move the embryo mutant characterization (Fig. S1F) and mESCs (Fig. S2A-C) to the main figures and elaborate the text accordingly. The genomic modifications in mice are described in a detailed tabular format in Table 1 in Materials and Methods. The immunofluorescence staining in Fig. S1F was performed on mouse embryonic sections, which tend to have higher backgrounds than cultured cells; Thus, we attribute the very low percentage of SAS-6 staining in Sas-6em5/em5* mutants to higher background, especially given the lack of centrioles in these mutants at all the stages examined.

For Western blots, we used different antibodies against SAS-6 that were either commercially available (Proteintech cat# 21377-1-AP, Sigma-Aldrich cat# HPA028187 and Santa Cruz cat# SC-81431) or non-commercial (kind gift from Renata Basto, Institute Curie). The SAS-6 antibody from the Basto lab gave the most reliable and reproducible results. Using this antibody, and in our own interpretation, we were not able to detect SAS-6 by Western blots in Sas-6 mutant mESCs (including hypomorphic alleles). We concluded that SAS-6 in mESCs (and mouse embryos, see below) is expressed at low levels. Thus, we decided to use the antibody provided by Renata Basto and shown in current Fig. S2C, although it shows two thick non-specific bands flanking the specific band for SAS-6.

For a more definitive knockout in mESCs, we decided to bi-allelically delete the entire Sas-6 ORF DNA from the ATG to the TAA (over 34 Kb of DNA, Fig. S2A). According to the central dogma of molecular biology, when there is no DNA, then there should be no mRNA and no protein. In confirmation of this premise, recent RT-PCR data showed that Sas-6 mRNA is not detectable in these Sas-6-/- null mESCs. Also, immunofluorescence analyses did not detect SAS-6 in these cells. We will add the RT-PCR and immunofluorescence data to the fully revised manuscript. We will also repeat the SAS-6 Western blots to achieve better band resolution.

In addition, we have used the Basto lab antibody for SAS-6 for Western blots on mouse embryos, which detect low levels of SAS-6 in controls and no signal in the mutants.

• The authors could elaborate on the topic of mESCs as a special in vitro model for centriole biology akin to the more "primitive" origins of life such as algae.*

We will elaborate on the topic of mESCs as a special system for centriole biology to stress the findings that mESCs without SAS-6 can still form centrioles, but also that these cells seem to tolerate centriolar aberrations, such as in Sas-6 mutants, or even the loss of centrioles, as in Sas-4 mutants, without undergoing apoptosis or cell cycle arrest.

• Figure 4 should show timeline of embryo development, include embryo stages (E3.5, E9 etc.), group together mESCs with corresponding embryonic developmental stage. The Figure can indicate when mESCs were derived from SASS6 em5/em5 blastocysts, when they were stained and indicate the number/state of centriole formation observed.*

We will adjust the model in Fig. 4 to accommodate the suggestions of the reviewer, but at the same time try not to overcrowd the model and dilute the main findings of the study.

• The work from Wang and Tsou using SAS-6 oligomerization mutants should be better discussed in the context of the work presented here since centriole assembly was not affected per se but structural defects were observed, like is the case in this study.*

We will elaborate on this finding from Wang et al. In this respect, we will note that the loss of the entire SAS-6 protein in human RPE-1 cells (on a p53-mutant background), leads to the loss of centrioles, but that the deletion of the oligomerization domain of SAS-6 in these cells leads to similar phenotypes to the total loss of SAS-6 in mESCs.

• The observation that the ability of forming centrioles de novo in NPCs derived from ESCs is lost is interesting but the mechanisms underpinning this differentiation remain unclear. The authors at a minimum should speculate on these further.*

We agree with the reviewer and will speculate on this finding further. This comment is along the same line as the difference in phenotype between the cells in the developing mouse embryo and mESCs, where the NPCs are more akin to the in vivo phenotype.

CROSS-CONSULTATION COMMENTS

Looks like we are all pretty much in agreement.

• *

Reviewer #2 (Significance (Required)):

• *

This is a well executed study with no major flaws that builds on similar studies on knocking out centriole components in mouse and other cell types. Although well-executed the study remains descriptive and lacks a clear mechanistic understanding of why de novo centriole assembly is ineffective in NPCs. As it stands the advances this study provides to the centrosome biogenesis field remain incremental.

We thank the reviewer for the compliments about our work and agree that it opens new questions in the field about the precise roles of SAS-6 and the cartwheel in centriole biogenesis.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

In this publication, Grzonka and Bazzi build upon their recent work describing the role of SAS-like protein function in centriole formation during embryonic development. More specifically, they demonstrate that loss of Sas-6 in vivo and in vitro disrupts centriole formation. To this reviewer's surprise, they found that Sas-6 is required for centriole formation in embryos, yet, stem cells form centrioles with disrupted centriole length and ability to template cilia.

• *

We thank the reviewer for highlighting the novel and surprising aspect of our work, which is that Sas-6 mutant mESCs are still able to form centrioles. We would like to stress that SAS-4, from our previously published work, and SAS-6, in this study, are not part of the same protein family and have different structures and roles in centriole formation. The naming has its origin in “Spindle-ASsembly abnormal/defective” mutant screens performed in C. elegans. Although the phenotypes are similar in vivo, due the lack of centrioles in both cases, only mutations in Sas-4, but not in Sas-6, lead to the lack of centrioles in mESCs.

• *

Likely, this occurs from the residual proteins that existed prior to CRISPR-mediated knockout.

• *

Due to the nature of the surprising finding that Sas-6 mutant mESCs can still form centrioles, we understand the concerns and suggestions of this reviewer and the other reviewers in this regard.

For a more definitive knockout in mESCs, we decided to bi-allelically delete the entire Sas-6 ORF DNA from the ATG to the TAA (over 34 Kb of DNA, Fig. S2A). According to the central dogma of molecular biology, when there is no DNA, then there should be no mRNA and no protein. In confirmation of this premise, recent RT-PCR data showed that Sas-6 mRNA is not detectable in these Sas-6-/- null mESCs. Also, immunofluorescence analyses did not detect SAS-6 in these cells. We will add the RT-PCR and immunofluorescence data to the fully revised manuscript. We will also repeat the SAS-6 Western blots to achieve better band resolution.

These Sas-6-/- mESCs started from a single cell and have been passaged up to 20 times by now without losing centrioles. SAS-6 protein was not detectable at the early passages and the mRNA is still not detectable. This is how knockouts have been and are produced. If this mutant is still not convincing, then we respectfully ask that the reviewers provide their own suggestion on what will be more convincing.

• *

Unsurprisingly, they found that Sas-6 loss in the developing mouse activates the 53BP1-USP28-p53 surveillance pathway leading to cell death and embryonic arrest at mid-gestation, similar to their findings in Cenpj knockouts. What remains to be properly elucidated is the mechanistic differences in the requirement for Sas-6 in stem cells versus the embryo, which may be beyond the scope of a short report. As it reads, the manuscript is a compliment to their Sas-4 paper but falls short of novelty and providing large strides in revealing the role of centriolar proteins in developmental processes. Moreover, the advances beyond the requirement for centriole and associated proteins in embryology is missing, therefore enthusiasm is tempered. Below are remaining concerns that must be addressed:

• *

Remaining concerns:

The authors should provide clear description of the embryonic region (neural plate & mesenchym) used to analyze centriole presence or loss in Figures 1 and S1. Was this in the forelimb vs hindlimb regions?

The assessment of centrosomes in Fig. 1 and S1 was performed on cell types in all three germ layers in the sections that were taken from the brachial region (forelimb and heart level). The information will be added to the Materials and Methods section.

Similar to their Cenpj-mouse data, the authors should provide data detailing the mitotic index and activation of the mitotic surveillance pathway beyond just p53 staining. As novelty is not the only criteria for publication, a thorough analysis of the Sas-6 activation of the mitotic purveyance pathway should be provided, including the crosses between Sas-6 and p53, 53bp1 and usp28 knockout crosses to demonstrate the pathway functions similarly to Cenpj loss.

We will perform the additional experiments suggested by the reviewer that are similar to our previous work in Sas-4 mutants (Xiao*, Grzonka* et al, 2021). We will perform these analyses knowing that both Sas-4 and Sas-6 mutants lose centrioles and activate the mitotic surveillance pathway, as the reviewer indicated. In particular, we will quantify the mitotic index in the Sas-6em5/em5 mutants and perform p53 and Cl. CASP3 staining in the double mutants with 53bp1 or Usp28, to show that the pathway has been suppressed in these mutants.

Centriole structure should be assessed in the embryos using EM to assess loss and confirm the structural defects. This would strengthen their argument and be a slight advance to their largely descriptive paper.

Because the Sas-6em5/em5 embryos lack centrioles, as indicated by regular immunofluorescence and Ultrastructure-Expansion Microscopy (U-ExM), using EM would be an attempt to find a structure that does not exist. In our opinion, it would again be a repetition of TEM studies that we have already performed in Sas-4-/- mutant embryos, that lack centrioles (Bazzi and Anderson, 2014). Using U-ExM has advanced the centriole biology field to a level that is approaching EM resolution and, in our opinion, can substitute for EM.

The WB for Sas-6 knockout is not convincing and should be redone. There are validated Sas-6 antibodies available from SCBT and Proteintech. It is not clear that the band is gone or if there's overlap with the non-specific band.

The answer to this comment is shown above. In addition, we have used the Basto lab antibody for SAS-6 for Western blots on mouse embryos, which detect low levels of SAS-6 in controls and no signal in the mutants. We will also repeat the SAS-6 Western blots on mESCs to achieve better band resolution as recommended by the reviewer.

The authors describe the centriolar structural defect in the mESCs in Figure 2C and D, and further characterize the phenotype in S2D-H. Given the role of the SAS6-CEP135-CPAP axis for centriole elongation, it is peculiar that they see elongation upon reduction of CEP135. The authors should find a rationale mechanism to explain their discordant findings. In addition, other centriole distal end components including CEP97 and CP110 should be examined to determine the structural end caping defect in the Sas-6 mESC.

Over 70% of the centrioles in Sas-6-/- mESCs retain CEP135, but the majority of CEP135 signals (over 80%) seem to be abnormally localized. One potential explanation for the elongated centrioles in Sas-6-/- mESCs is that the mis-localization of CEP135 impacts on the integrity of the centriole and results in parts of the centriolar walls being elongated. Per the reviewer’s suggestion, we have performed U-ExM with stainings for CP110 or CEP97, that also regulate centriole capping and elongation. The preliminary data suggest that similar to WT mESCs, they localize to the ends of the abnormal centrioles in Sas-6-/- mESCs. We will quantify the percentage of normally-localized CP110 and CEP97 in Sas-6-/- mESCs and include it along with the data interpretation in the next version of the manuscript.

• *

In Figure 2I, J the authors state the ciliation rate for the WT mESCs was only 11%, could the authors provide an explanation for the low ciliation rate in WT mESCs? Could cells be arrested to increase the ciliation rate? In addition, is there a rational explanation for the loss of centrioles and centrosomes upon differentiation into NPCs?

mESCs ciliation rate has been shown to be generally low (Bangs et al, 2015; Xiao et al., 2021) perhaps because the cells spend most of the cell cycle in the S-phase. mESCs require a high serum percentage and well-defined media for growth and maintenance. In our hands, attempting to arrest the cells by withdrawing serum, or reducing its percentage, resulted in cell death and a change in morphology to the differentiated phenotype (unpublished data). Our data indicate that a pluripotent state in Sas-6-/- mESCs is compatible with centriole formation but differentiation results in the loss of centrioles (for example, NPCs). Therefore, we have refrained from interfering with the cell cycle of mESCs in order to avoid these confounding effects on cellular viability and centriole formation.

Regarding the loss of centrioles upon differentiation of Sas-6-/- mESCs into NPCs, we agree with the reviewer and will speculate on this finding further. This goes along the same line as the difference in phenotype between the cells in the developing mouse embryo and mESCs, where the NPCs are more akin to the in vivo phenotype of Sas-6 mutants. The data suggest that the formation of centrioles in Sas-6-/- mESCs is associated with the in vitro pluripotent phenotype. A more comprehensive and general characterization of centriole duplication in mESCs is a future direction to elucidate their ability to form centrioles without SAS-6.

In figure 3F in the Sas-6−/− NPCs have a box around a cell without centrosomes yet in 3G here are some cells with centrosomes. While the authors are trying to demonstrate the decrease in centrosome in the Sas-6−/− NPCs, they should show the few cell that have centrosomes or centrosome-like structures.

We will add another example for the minority of cells that retain centrosomes upon differentiation of Sas-6-/- mESCs into NPCs.

CROSS-CONSULTATION COMMENTS

• *

As mentioned in my review; while the Sas6 model is new, it does not provide further evidence of why centriole duplication is important in developing mice aside from it causing an abortive mitosis leading to cell death. The discordant phenotype in the mESCs likely arises from residual Sas6, similar to experiments that were performed in flies with Sas-4 depletion. Moreover, the odd centriole phenotype represents a very small number of cells and is likely phenomenological.

In addition, their work from last year demonstrated a clear connection between Cenpj loss leading to the mitotic surveillance pathway activation. They performed double knockouts that partially rescued the survival phenotype. This new work falls short of that publication.

Reviewer #3 (Significance (Required)):

• *

The new publication adds a known component to the list of animal models for centrosome-opathies but fails to provide novel mechanistic insights. Dr. Bazzi's publication on Sas-4 was far more novel at the time of publication due to the multiple mouse crosses that could rescue the phenotypes. The recent publication fails to provide as much evidence or any novel insights into the role of Sas-6 (sufficient to be convincing).

The audience will be limited to centrosome biologists and even then it may not have enough novelty to be compelling. I would recommend with the revisions to be published in a more specialized journal.

*My expertise lies in genetic causes of microcephaly-associated with mutations in centrosome encoding proteins. *

• *

We thank the reviewer for taking the time to evaluate our work and provide helpful comments and suggestions. We would like to emphasize that even if a certain phenotype is expected, the experiment has to be performed to test the hypothesis, which is the case with the Sas-6 mutant embryos phenocopying the Sas-4 mutants. In our opinion, the novelty of our work goes beyond Fig. 1 to the ability of Sas-6-/- null mESCs to form centrioles. This surprising finding opens new avenues of investigation into the precise roles of SAS-6, and the cartwheel, in centriole biogenesis. We are confident that our study will provide a trigger to re-examine these roles in other cell types and organisms.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this publication, Grzonka and Bazzi build upon their recent work describing the role of SAS-like protein function in centriole formation during embryonic development. More specifically, they demonstrate that loss of Sas-6 in vivo and in vitro disrupts centriole formation. To this reviewer's surprise, they found that Sas-6 is required for centriole formation in embryos, yet, stem cells form centrioles with disrupted centriole length and ability to template cilia. Likely, this occurs from the residual proteins that existed prior to CRISPR-mediated knockout. Unsurprisingly, they found that Sas-6 loss in the developing mouse activates the 53BP1-USP28-p53 surveillance pathway leading to cell death and embryonic arrest at mid-gestation, similar to their findings in Cenpj knockouts. What remains to be properly elucidated is the mechanistic differences in the requirement for Sas-6 in stem cells versus the embryo, which may be beyond the scope of a short report. As it reads, the manuscript is a compliment to their Sas-4 paper but falls short of novelty and providing large strides in revealing the role of centriolar proteins in developmental processes. Moreover, the advances beyond the requirement for centriole and associated proteins in embryology is missing, therefore enthusiasm is tempered. Below are remaining concerns that must be addressed:

Remaining concerns:

• The authors should provide clear description of the embryonic region (neural plate & mesenchym) used to analyze centriole presence or loss in Figures 1 and S1. Was this in the forelimb vs hindlimb regions?

• Similar to their Cenpj-mouse data, the authors should provide data detailing the mitotic index and activation of the mitotic surveillance pathway beyond just p53 staining. As novelty is not the only criteria for publication, a thorough analysis of the Sas-6 activation of the mitotic purveyance pathway should be provided, including the crosses between Sas-6 and p53, 53bp1 and usp28 knockout crosses to demonstrate the pathway functions similarly to Cenpj loss.

• Centriole structure should be assessed in the embryos using EM to assess loss and confirm the structural defects. This would strengthen their argument and be a slight advance to their largely descriptive paper.

• The WB for Sas-6 knockout is not convincing and should be redone. There are validated Sas-6 antibodies available from SCBT and Proteintech. It is not clear that the band is gone or if there's overlap with the non-specific band.

• The authors describe the centriolar structural defect in the mESCs in Figure 2C and D, and further characterize the phenotype in S2D-H. Given the role of the SAS6-CEP135-CPAP axis for centriole elongation, it is peculiar that they see elongation upon reduction of CEP135. The authors should find a rationale mechanism to explain their discordant findings. In addition, other centriole distal end components including CEP97 and CP110 should be examined to determine the structural end caping defect in the Sas-6 mESC.

• In Figure 2I, J the authors state the ciliation rate for the WT mESCs was only 11%, could the authors provide an explanation for the low ciliation rate in WT mESCs? Could cells be arrested to increase the ciliation rate? In addition, is there a rational explanation for the loss of centrioles and centrosomes upon differentiation into NPCs?

• In figure 3F in the Sas-6−/− NPCs have a box around a cell without centrosomes yet in 3G here are some cells with centrosomes. While the authors are trying to demonstrate the decrease in centrosome in the Sas-6−/− NPCs, they should show the few cell that have centrosomes or centrosome-like structures.

CROSS-CONSULTATION COMMENTS

As mentioned in my review; while the Sas6 model is new, it does not provide further evidence of why centriole duplication is important in developing mice aside from it causing an abortive mitosis leading to cell death. The discordant phenotype in the mESCs likely arises from residual Sas6, similar to experiments that were performed in flies with Sas-4 depletion. Moreover, the odd centriole phenotype represents a very small number of cells and is likely phenomenological.

In addition, their work from last year demonstrated a clear connection between Cenpj loss leading to the mitotic surveillance pathway activation. They performed double knockouts that partially rescued the survival phenotype. This new work falls short of that publication.

Significance

The new publication adds a known component to the list of animal models for centrosome-opathies but fails to provide novel mechanistic insights. Dr. Bazzi's publication on Sas-4 was far more novel at the time of publication due to the multiple mouse crosses that could rescue the phenotypes. The recent publication fails to provide as much evidence or any novel insights into the role of Sas-6 (sufficient to be convincing).

The audience will be limited to centrosome biologists and even then it may not have enough novelty to be compelling. I would recommend with the revisions to be published in a more specialized journal.

My expertise lies in genetic causes of microcephaly-associated with mutations in centrosome encoding proteins.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Here, Grzonka and Bazzi present their work on characterizing the requirement of SASS6 in mouse embryo development and in embryonic stem cell (mESC) culture. In mouse, female and male gametes lack centrioles, and early divisions occur without centrioles. De novo formation typically happens at the blastocyst stage (~E3.5). The authors generated two SASS6 knock-out strains, SASS6 em4/em4 (frameshift deletion, reported as a severe hypomorphic allele), and SASS6 em5-em5 (frameshift deletion, reported as a null allele). Mutant embryos arrest development at mid-gestation unless the p53, USP28 and USP28 pathway is perturbed. As expected, centrioles do not form in SASS6 -/- mice. However, the authors report that de novo formation of centrioles is facilitated in mESC culture conditions for SASS6 CRISPR knock-out mESCs and mESCs derived from SASS6 em5/em5 blastocysts. Centrioles are lost upon differentiation of SASS6 CRISPR knock-out mESCs into neural progenitor cells (NPCs).

The presented study is relevant for scientists investigating the requirements for centriole formation during embryonic development. Further, it provides insights in possibly different requirements for centriole formation between stages of differentiation, as well as differences in in vivo and in vitro models. The data represented by Grzonka and Bazzi are robust and support the manuscript and conclusions made. However, the study is predominantly descriptive, and the authors do not test the molecular pathway underlying the de novo formation of centrioles observed in SASS6 -/- mESCs. It is generally believed that de novo formation of centrioles is not possible in SASS6 knock-out cells although work from Wang and Tsou with SASS6 a oligomerization mutant suggests otherwise. A dissection of the specific factors required for the de novo formation of centrioles in the mESC context would provide more insights into de novo centriole assembly in general and would increase the impact of this work.

I would support publication of the manuscript if the following points are addressed:

1. One of the main figures, ideally Figure 1, should be dedicated to the characterization of the newly generated mouse strains. This should also be elaborated in the text further. I would like to see a schematic representation of the genomic modifications. The SASS6 stainings of wt and Sas-6 knock-outs (now Figure S1F) should be shown in that context as well as the Figures S2A-C. The authors should discuss why there still appears to be SASS6 protein in the SASS6-em5/em5 Sas-6 stainings visible. Also, the western blot, especially the unspecific bands so close to the SAS-6 protein, should be discussed. Adding qRTPCR results would also be good.

2. The authors could elaborate on the topic of mESCs as a special in vitro model for centriole biology akin to the more "primitive" origins of life such as algae.

3. Figure 4 should show timeline of embryo development, include embryo stages (E3.5, E9 etc.), group together mESCs with corresponding embryonic developmental stage. The Figure can indicate when mESCs were derived from SASS6 em5/em5 blastocysts, when they were stained and indicate the number/state of centriole formation observed.

4. The work from Wang and Tsou using SAS-6 oligomerization mutants should be better discussed in the context of the work presented here since centriole assembly was not affected per se but structural defects were observed, like is the case in this study.

5. The observation that the ability of forming centrioles de novo in NPCs derived from ESCs is lost is interesting but the mechanisms underpinning this differentiation remain unclear. The authors at a minimum should speculate on these further.

CROSS-CONSULTATION COMMENTS

Looks like we are all pretty much in agreement.

Significance

This is a well executed study with no major flaws that builds on similar studies on knocking out centriole components in mouse and other cell types. Although well-executed the study remains descriptive and lacks a clear mechanistic understanding of why de novo centriole assembly is ineffective in NPCs.As it stands the advances this study provides to the centrosome biogenesis field remain incremental.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The article by M. Grzonka and H. Bazzi entitled: Mouse Sas-6 is required for centriole formation in embryos and integrity in embryonic stem cells, describes new findings in novel mouse models of Sas-6 knockouts (KO). This is an interesting study that reports two different mouse Sas6 KO models and the depletion of Sas6 from mouse embryonic stem cells (mESCs). This type of analysis has never been done before and so it reveals and describes a role for Sas-6 in centriole biogenesis in mouse.

The authors compare their analysis with Sas-4 KO and overall found similar results when compared to previous work from H. Bazzi, when Sas-4 was depleted in mouse embryos. Due to the mitotic stopwatch pathway, Sas6KO embryos die during development at extremely early stages and this can be rescued by depletion in p53 and other members of the pathway. Perhaps, not so surprisingly, these embryos do not contain centrioles, showing that in vivo, Sas6 is absolutely required for centriole duplication. More surprisingly, however, in cultures of mESCs, established and propagated in vitro, Sas-6 crispr induced KO, does not result in lack of centrioles. Instead, abnormal structures that show aberrant morphologies, length, and incapacity to assemble cilia were detected. In principle, this means that centrioles can be assembled independently of Sas-6, even if not in the correct manner.

The authors interpret these differences as possible differences in the pathways involved in centriole assembly and propose different requirements in different cell types, within the same species. I have problems with this interpretation. To me is very difficult to understand, how the "protein" absolutely required for cartwheel assembly at the early stages of centriole biogenesis, can be essential and dispensable at the same time. Although, I may be wrong, I think the authors have not envisage other possibilities to interpret their data, which should be taken into consideration.

1) I do not know anything about ESC and ESC cultures. So maybe this is a stupid suggestion. But can't they be derived exactly from the same genetic background of SAS-6KO embryos? Because the way the two (or even 3 as there are 2 mouse KO lines) are generated is different. Why is that?

2) Still on mESCs, are the authors sure that there are no WT Sas-6mRNAs still present in their ESC cells? Because tiny amounts are maybe sufficient to allow the initial cartwheel structure. In FigS2B, I can see a really faint band, very faint but it is there.

3) This last point goes also with the western-blot of Figure S2C- there is still a band, very tiny between the two very tick bands (marked with *). Maybe separating proteins better will help visualizing the real Sas-6 band? Have they used the Sas6 ab in other WBs from the KO embryos, for example? Can they use the Sas6 ab in immunostaining to show if the assembled abnormal centrioles completely lack Sas6. This will allow to distinguish between the hypothesis of having some (even if not much) sas6 left?

4) Then a more theoretical point? Have the authors considered that the difference is more in the stability of the abnormal structures. Let's say, without a cartwheel and maybe enough PLK4 activity and high level of other centriolar components, the centrioles are abnormally assembled- they have no cartwheel, but they are disassembled very fast in the embryo but not in ESCs?

5) Even if there is a real difference and without Sas-6 ESCs can make centrioles that are abnormal in structure and function (at least at the cilia assemble level), the choice of words "strictly required", I am not sure it is correct. Because, since Sas-6 is described by many studies as the factor required for cartwheel assembly, which occurs very early in the pathway, this means that in mESCs centrioles can assembled without forming a cartwheel. And so that the cartwheel is actually not required for the initial building block, but more as a structure that maintains the whole centriole in an intact manner?

6) The authors mentioned that in flies, abnormal Sas-6 structures have been described in certain cell types. Are these mutants, null mutants? In other words, do these structures assembled in a context of no Sas6 or abnormal Sas-6 protein or even low levels of Sas-6?

Other points:

• I think the 1st sentence of the abstract appears disconnected from the rest. The same goes for the 1st sentence of the introduction. And also, what is the evidence that pluripotent stem cells rely primarily on the proper assembly of a mitotic spindle? They relly on many other things, not sure this is the first one.
• The authors mention that centrioles are lost in Sas6-/- after "differentiation" of mESCs. The term differentiation is not appropriate, and confusing here. Differentiation normally refer to cells that stopped proliferating and exited the cell cycle, which is not the case here, as NPCs are progenitor cells that keep cycling.
• Figure S1: Percent of cells with centrosomes was assessed by a co-staining of tubulin and Cep164, which mark the mother centrioles. As Cep164 may be absent from centrosomes after lack of centriole maturation in sas6-/- embryos, another combination of staining should be performed to evaluate the percent of cells without centrosomes. tubulin staining can be seen in Sas6 em5/em5 embryos, while the quantification claims total absence of centrosomes. The authors use the CENT2-eGFP transgenic line to count the number of centrioles in Figure 3, they should do the same in Figure S1.

Significance

This study shows with a novel mouse model the requirement of centrioles during mouse development. It will be relevnat to centrosome labs, the novel mouse lines will be useful to many labs working on centrioles, cilia and centrosomes. My expertise: centrosome biology

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We thank the Reviewers for their useful feedback on our manuscript. We have addressed the Reviewers’ comments and revised our manuscript accordingly. A point-by-point response is provided below.

Reviewer comments:

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

Gopalan et al use quantitative, comprehensive lipid mass spectrometry of mouse brain tissue isolated at various time points in embryonic and postnatal development. They then go on to use the same quantitative analysis of mouse and human stem cells differentiated in vitro into neurons to define the lipid composition of these cultures.

Major Comments:

1. As mentioned above, it is difficult to assess whether the discrepancy in the lipotype acquisition between in vivo mouse brain development and stem cell differentiation is due to metabolic differences in the in vitro differentiation as the authors state or is due to a lack of the stem cells to actually acquire a neuronal phenotype. Perhaps showing more clearly that the protocols for neuronal differentiation work efficiently and/or how they compare to brains dissected would be helpful in stating that the lipotype is different. The protocol referenced here (Bogetofte et al) only gives ~30% TH+ positive DA neurons in their manuscript. What cell type the other 70% of the cells are is something that could be discussed as means of "diluting" out the lipotype seen in these cultures. Perhaps the 30% TH+ DA neurons do attain the "correct" lipotype, but the lipidomic analysis can not detect this due to the contaminating effects of the non-differentiated cells. In this work, it would be nice to see what percentage of the cells differentiate into the expected cell type rather than referencing previous manuscripts. As differentiation protocols and originating cell sources are highly variable and error-prone, it's difficult to know what the lipotype results are actually reporting on. Furthermore, discussion about these differentiation techniques and how well they represent functional neurons is warranted. The papers referenced here don't show 100% differentiation into the phenotypes that are described in this work such that the lipotype finding is not the only suggestion of a "general failure of in vitro neuronal differentiation models". Maybe a discussion of how the lack of ability to attain the neuronal lipotype due to the metabolic deficiencies discussed here could be causative to the inability to full recapitulate the neuronal phenotype is useful for the reader.

We thank the reviewer for this question and suggested experiments. Following the advice, we have now show immunofluorescence data of pan-neuronal markers (i.e., b-tub III or MAP2) in mESC and iPSCs. In agreement with previously published datasets from the Noh and Meyer labs (Gehre et al., 2020; Bogetofte et al.,2019), we show that the protocols we use generate a very high percentage of neurons. We have now included these images and quantifications in our manuscript as Figs. 2B and S6A,B.

From the discussion and work here is unclear why the stearate feeding of the stem cells did not result in an increase in the 18:0-containing sphingolipids. The authors state that the appropriate metabolic pathways are not fully established and go on to look at the CerS expression levels across the differentiation timeline. It appears that the results presented in Fig. S7 counter the authors' interpretation of the lipotype and more discussion here would be nice to clarify this discrepancy.

We thank the reviewer for highlighting this seemingly counterintuitive observation. We have now included a quantification of CerS mRNA from commercially available mouse tissues analysed in Sladitschek and Neveu, 2019 and compared this to the data from Gehre et al, 2020. In the mouse brain tissue, CerS1 expression is upregulated dramatically, while CerS5 and 6 are downregulated (see new panel A in Fig. S8). In contrast, during in vitro differentiation of mESCs, CerS5 is not downregulated and CerS6 is upregulated (Fig. S8B). Accordingly, we have expanded our discussion in the revised manuscript as follows:

“On the other hand, supplementing the cells with stearic acid (18:0) does not result in high levels of 18:0-sphingolipids. It is known that the Cer synthase CerS1 is specific for stearoyl-CoA (Venkataraman et al., 2002), which results in the production of 18:0-sphingolipids, while the synthases CerS5 and CerS6 are responsible for 16:0-sphingolipid production. During brain development, one observes a 35-fold increase in the expression of CerS1 and a downregulation of CerS5 and CerS6 compared to embryonic tissue (Sladitschek & Neveu, 2019) (Fig. S8A). In contrast, during in vitro neuronal differentiation, between day 8 and 12, CerS1 expression increases only by 5-fold and, contrary to expectation, CerS6 expression is upregulated and CerS5 expression is unchanged (Gehre et al., 2020) (Fig. S8B). This could underpin the observation that 16:0-sphingolipids remain elevated whereas brain-specific 18:0-sphingolipids only increase marginally, despite supplementation with stearic acid. Overall, this suggests that appropriate programming of the sphingolipid metabolic machinery is not fully established in stem cell-derived neurons.”

Minor comments:

1. I find the data presentation of the LENA analysis to be difficult to follow (Fig. 1E). In my opinion, the p-value is not the most important bit of information in this graph, though having it on the y-axis with other pertinent information encoded by colors or arrows being disguised. I would rather see the data on the x-axis that is above a certain p-value (denoted in the figure legend) plotted with the direction and magnitude of change shown.

We thank the reviewer for this suggestion. In the revised manuscript, we now plot log2(odds ratio) on the y-axis instead of the p-value. Moreover, we have dimensioned the size and color intensity of each point as function of the p-value (Fig. 1E and 2E, shown below).

In the PCA in Fig 1, what are the loadings that define the variable PC1 and PC2? What is predominantly changing the P21 samples that lead to such a large shift if most of the data shown in the subsequent panels are not changing much between P2 and P21.

In the revised manuscript, we now include a plot of the PCA loadings of the lipids majorly influencing principal components 1 and 2 as supplemental Fig. S3.

Reviewer #1 (Significance (Required)):

This work provides a nice reference for the complex lipidomes in embryonic and postnatal murine brain development. The details of the lipotype changes during development are well laid out and will of no doubt be of great use across a variety of scientific fields. While I found the in vivo data to be compelling, interesting, and useful, the lack of controls for the in vitro stem cell differentiation work makes this particular data set and comparison less useful. Further work to identify the limitations of the stem cell differentiation protocols as a valid comparison to in vivo brain development need to be done and/or the discussion of the direct comparisons between the two toned down.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

The study used a quantitative lipidomics approach which I am very familiar with. The results should be highly reproducible.

Reviewer #2 (Significance (Required)):

The manuscript submitted by Gopalan et al. reported a quantitative and comparative lipidomics study between mouse brain samples from early embryonic to postnatal stages, and rodent and human stem cell-derived neurons. The authors found a couple of very unique characters only existing in brain samples, but not in stem cell-derived neurons, including 22:6-containing glycerophospholipids and 18:0-containing sphingolipids. The authors further found the brain-like lipotypes can only be partially established in stem cell-derived neurons after supplementing brain lipid precursors. These findings clearly suggest that stem cell-derived neurons might not be appropriately used to mechanistically study lipid biochemistry, membrane biology, and biophysics in brains. The study was well designed. and the manuscript was very informative and resourceful. I would suggest to accept the manuscript for publication.

We thank the Reviewer for the positive assessment of our work.

References

Gehre, M., Bunina, D., Sidoli, S., Lübke, M. J., Diaz, N., Trovato, M., Garcia, B. A., Zaugg, J. B., & Noh, K. M. (2020). Lysine 4 of histone H3.3 is required for embryonic stem cell differentiation, histone enrichment at regulatory regions and transcription accuracy. Nature Genetics, 52(3), 273–282. https://doi.org/10.1038/s41588-020-0586-5

Levy, M., & Futerman, A. H. (2010). Mammalian ceramide synthases. IUBMB Life, 62(5), 347–356. https://doi.org/10.1002/iub.319

Sladitschek, H. L., & Neveu, P. A. (2019). A gene regulatory network controls the balance between mesendoderm and ectoderm at pluripotency exit. Molecular Systems Biology, 15(12), 1–13. https://doi.org/10.15252/msb.20199043

Venkataraman, K., Riebeling, C., Bodennec, J., Riezman, H., Allegood, J. C., Cameron Sullards, M., Merrill, A. H., & Futerman, A. H. (2002). Upstream of growth and differentiation factor 1 (uog1), a mammalian homolog of the yeast longevity assurance gene 1 (LAG1), regulates N-stearoyl-sphinganine (C18-(dihydro)ceramide) synthesis in a fumonisin B1-independent manner in mammalian cells. Journal of Biological Chemistry, 277(38), 35642–35649. https://doi.org/10.1074/jbc.M205211200

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The study used a quantitative lipidomics approach which I am very familiar with. The results should be highly reproducible.

Significance

The manuscript submitted by Gopalan et al. reported a quantitative and comparative lipidomics study between mouse brain samples from early embryonic to postnatal stages, and rodent and human stem cell-derived neurons. The authors found a couple of very unique characters only existing in brain samples, but not in stem cell-derived neurons, including 22:6-containing glycerophospholipids and 18:0-containing sphingolipids. The authors further found the brain-like lipotypes can only be partially established in stem cell-derived neurons after supplementing brain lipid precursors. These findings clearly suggest that stem cell-derived neurons might not be appropriately used to mechanistically study lipid biochemistry, membrane biology, and biophysics in brains. The study was well designed. and the manuscript was very informative and resourceful. I would suggest to accept the manuscript for publication.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Gopalan et al use quantitative, comprehensive lipid mass spectrometry of mouse brain tissue isolated at various time points in embryonic and postnatal development. They then go on to use the same quantitative analysis of mouse and human stem cells differentiated in vitro into neurons to define the lipid composition of these cultures.

Major Comments:

1. As mentioned above, it is difficult to assess whether the discrepancy in the lipotype acquisition between in vivo mouse brain development and stem cell differentiation is due to metabolic differences in the in vitro differentiation as the authors state or is due to a lack of the stem cells to actually acquire a neuronal phenotype. Perhaps showing more clearly that the protocols for neuronal differentiation work efficiently and/or how they compare to brains dissected would be helpful in stating that the lipotype is different. The protocol referenced here (Bogetofte et al) only gives ~30% TH+ positive DA neurons in their manuscript. What cell type the other 70% of the cells are is something that could be discussed as means of "diluting" out the lipotype seen in these cultures. Perhaps the 30% TH+ DA neurons do attain the "correct" lipotype, but the lipidomic analysis can not detect this due to the contaminating effects of the non-differentiated cells. In this work, it would be nice to see what percentage of the cells differentiate into the expected cell type rather than referencing previous manuscripts. As differentiation protocols and originating cell sources are highly variable and error-prone, it's difficult to know what the lipotype results are actually reporting on.

Furthermore, discussion about these differentiation techniques and how well they represent functional neurons is warranted. The papers referenced here don't show 100% differentiation into the phenotypes that are described in this work such that the lipotype finding is not the only suggestion of a "general failure of in vitro neuronal differentiation models". Maybe a discussion of how the lack of ability to attain the neuronal lipotype due to the metabolic deficiencies discussed here could be causative to the inability to full recapitulate the neuronal phenotype is useful for the reader. 2. From the discussion and work here is unclear why the stearate feeding of the stem cells did not result in an increase in the 18:0-containing sphingolipids. The authors state that the appropriate metabolic pathways are not fully established and go on to look at the CerS expression levels across the differentiation timeline. It appears that the results presented in Fig S7 counter the authors' interpretation of the lipotype and more discussion here would be nice to clarify this discrepancy.

Minor comments:

1. I find the data presentation of the LENA analysis to be difficult to follow (Fig 1E). In my opinion, the p-value is not the most important bit of information in this graph, though having it on the y-axis with other pertinent information encoded by colors or arrows being disguised. I would rather see the data on the x-axis that is above a certain p-value (denoted in the figure legend) plotted with the direction and magnitude of change shown.
2. In the PCA in Fig 1, what are the loadings that define the variable PC1 and PC2? What is predominantly changing the P21 samples that lead to such a large shift if most of the data shown in the subsequent panels are not changing much between P2 and P21.

Significance

This work provides a nice reference for the complex lipidomes in embryonic and postnatal murine brain development. The details of the lipotype changes during development are well laid out and will of no doubt be of great use across a variety of scientific fields. While I found the in vivo data to be compelling, interesting, and useful, the lack of controls for the in vitro stem cell differentiation work makes this particular data set and comparison less useful. Further work to identify the limitations of the stem cell differentiation protocols as a valid comparison to in vivo brain development need to be done and/or the discussion of the direct comparisons between the two toned down.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We thank the reviewers for their constructive comments, which greatly helped us in steering the manuscript in the right direction. Below is our point to point response to their concerns:

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

The Abeliovich lab has discovered differential degradation rates for mitochondrial proteins by mitophagy, a selective form of autophagy of mitochondria. In previous publications, they identified two protein kinases Pkp1/2 and the protein phosphatase Aup1(Pct6), which regulate phosphorylation patterns of mitochondrial proteins and affect their degradation rates by mitophagy. In the current manuscript, the authors now analyze the role of the pyruvate dehydrogenase complex (PDC) in mitochondria in the regulation of these regulatory link between Pkp1/2 and Aup1 and their substrate proteins. They find that the deletion of pda1, a core subunit of PDC, inhibits the degradation of previously identified mitophagy substrates Mdh1, Aco2, Qcr2, Aco1, and Idp1, while the turnover of mtDHFR-GFP, a model substrate diffusely localized in the mitochondrial matrix, is not affected. These data indicate that tested mitochondrial proteins are excluded from mitochondrial turnover by mitophagy in the absence of Pda1 by as yet unknown sorting mechanisms. The authors next show that the mutation of a known Pkp1/2 and Aup1 phosphorylation site within Pda1 (S313A) drastically reduces mitophagic turnover of Mdh1, which is completely blocked by the additional deletion of AUP1. A mutation blocking Pda1 activity (R322C) also almost completely blocked Mdh1 degradation. However, PDC activity does not generally affect mitophagy, since an inactivating mutation in Lat1 (K75R) promotes mitophagy of Mdh1. Published work has shown that Pkp1/2 and Aup1, which affect the phosphorylation and degradation of Mdh1, physically interact with the PDC. Thus, the authors tested whether the absence of Pda1 affected the phosphorylation state of Mdh1 and thus its turnover. Indeed, Mdh1 showed globally reduced phosphorylation in the absence of Pda1, which was rescued by the overexpression of Pkp1/2 promoting degradation of Mdh1. Consistent with the effects on mitophagic turnover, Mdh1 was phosphorylated at higher level in the presence of Lat1-K75R. The authors analyzed the mitochondrial distribution of Mdh1 in dependence of Pda1. They provide cytological data suggesting that Mdh1 might segregate from a diffuse matrix localized mtRFP reporter, which is suppressed upon Pkp1/2 overexpression. Finally, the authors perform a mitochondrial phospho-proteome analysis and found that the absence of Pda1 affected phosphorylation sites in 8 mitochondrial matrix proteins. From these data the authors propose a model in which the PDC complex controls the activity of the associated protein kinase Pkp1/2 and the phosphates Aup1 by allosteric changes, which in turn regulates the phosphorylation and differential turnover of mitochondrial proteins by mitophagy. They speculate that the PDC and associated factors could resemble large protein kinase complexes as TORC.

(1) The authors postulate a structural role for the PDC in regulating Pkp1/2 and Aup1 for controllingmitophagy turnover of certain substrates. While the physical association of Pkp1/2 and Aup1 with PDC has been shown previously, the authors need to critically test their model and assess (a) whether this physicalinteraction occurs under their experimental conditions and (b) whether the different mutants that affectmitophagy of Mdh1 also affect the physical interaction of Pkp1/2 and Aup1 with PDC. Otherwise, their model, although consistent, remains purely speculative.

Response: We fully agree with the reviewer that this is an important point and we are currently trying to verify these interactions under our working conditions. Naturally, we will then test whether the documented Pdh1-Aup1 and the Pdh1-Pkp1/2 interactions are affected by mutants such as the lat1K75R mutation and the lat1Δ mutation, as insightfully suggested by the reviewer.

(2) It is important to test every mutant that affects Mdh1 turnover for effects on mitophagy in general using the mtDHFR-GFP reporter to be able to conclude specific effects on Mdh1 mitophagy. For example, the deletion of Lat1 has been shown to induce mitophagy of a mitochondrial matrix reporter during nitrogen starvation (Bockler and Westermann, 2013). While the authors observe a complete block of Mdh1 turnover in lat1 deletion cells, they do observe increased Mdh1 degradation in Lat1-K75R cells. In line with this reasoning, the authors have identified a number of Mdh1 variants in a previous publication (Kolitsida et al. 2019) that can be used to further explore the observed phenomenon. For example, the Mdh1- T199A could be used to test whether the expression of Pkp1/2 in pda1 deficient cells has specific or general effects on mitophagy. Furthermore, can Mdh1-T199D, which is turned over independent of Pkp1/2, be degraded in the absence of Pda1?*

Response: We completely agree with the reviewer, and we have tested the effect of the lat1Δ mutations on mtDHFR-GFP (Supplemental Figure 2). It is important to note, however, that while the Bockler and Westermann paper does report an increased red/green fluorescence ratio in lat1Δ cells using mtRosella as a mitophagy reporter, this is part of a general screen, and was never further investigated or validated with additional, more rigorous methods (that paper focused on the role of the HERMES complex in mitophagy). In vivo fluorescence can be affected by a multitude of physical intracellular and intra-organellar factors including local pH, ionic strength, and molecular crowding. Therefore, the result from the screen conducted by Bockler and Westermann - although intriguing- is somewhat preliminary and requires more scrutiny (e.g. validation with a GFP release assay, complementation, etc) before we can make conclusions regarding the effect of the lat1Δ mutation on mitophagy, under their experimental conditions.

Furthermore, can Mdh1-T199D, which is turned over independent of Pkp1/2, be degraded in the absence of Pda1?*

This is now addressed in Figure 3E and 3F. The results are consistent with our hypothesis, namely showing complete or near complete suppression of the phenotype when tested using the phosphomimetic reporter.

(3) The cytological data are not convincing. The imaging quality is low and it is very difficult for the reader to appreciate the suggested segregation of Mdh1-GFP and mtRFP. Thus, it is important to provide cortical sections in order to visualize mitochondrial tubules/networks and Mdh1 distribution. This is particularly important, because the authors mention effects of Pda1 on mitochondrial morphology, which appears to be rescued by Pkp1/2 overexpression.

We are sorry to hear that the reviewer is not convinced by the microscopy data. We agree with the comment that it is not trivial to observe the segregation by eye. We do not intend to say that the segregation is absolute, and the differences we observe in the distributions of the two signals are relative. By that, we mean that a prominent green dot in the cell, which is brighter than other green dots in the same cell, has a red channel counterpart which is not brighter than the other red dots in the cell, or may even be dimmer than the other red dots in the same cell. However, we do provide illustrative examples of segregation with specific arrows pointing to loci of segregation between the two channels. This was not pointed out in the original figure legend and this oversight has now been corrected. More importantly, we performed a mathematical analysis of the overlap between the two signals, and demonstrate a statistically (very) significant difference between the endogenous mitochondrial protein and the artificial mitochondrially-targeted GFP chimera. We would also like to stress that under our conditions (gluconeogenic medium, stationary phase) yeast mitochondria (at least in our genetic background) are not normally found in tubules (this pattern is specifically observed in cells growing in glucose-based medium). Nonetheless, we will attempt to carry out a 3-d reconstruction using our available z-sections, as per the reviewer’s request.

(3 continued) In this context it would be very informative to follow the localization of PDC in mitochondria. Previous work has shown that Pda1 forms punctate structures in mitochondria in proximity to ER-mitochondria contact sites marked by ERMES (Cohen et al. 2014). Is Pda1 itself degraded by mitophagy or is it excluded? Could the physical interaction of Mdh1 with Pda1 explain mitophagy phenotypes if Pda1 is excluded from mitophagy? In other words, could the PDC form a physical unit to segregate and prevent proteins from mitophagy turnover?

Response: It was shown by others that Pda1-GFP localizes to mitochondrial puncta, and we also observe this in our system. We have also briefly looked at the mitophagic efficiency of Pda1-GFP. While it is inefficiently delivered to the vacuole (5% free GFP after 5 days, vs 20% for Mdh1), it is not excluded from mitophagic targeting. We are not sure what the reviewer is referring to as a “physical interaction between Pdh1 and Mdh1”. We do not demonstrate such an interaction, and to our knowledge such an interaction has not been reported and validated in any publication. Even if such an interaction existed, it cannot explain the rescue of the pda1Δ phenotype by the co-overexpression of Pkp1 and 2, nor the effect of the pda1Δ mutation on Mdh1 phosphorylation, or the effect of the pda1Δ mutation on other mitophagy reporters which we used, such as Aco1, Qcr2, Idp1, and Aco2. We agree with the reviewer that there is a possibility that an organizing center, be it the PDC or another complex, may regulate intra-matrix segregation and mitophagic trafficking. However, addressing this question would require significant additional time and resources, and we would beg the referee to defer this question to future investigations, as it is not central to the claims made in the current manuscript.

(4) The conclusions that can be drawn from the analysis of the mitochondrial phospho-proteome independence of Pda1 are rather limited. First, the experimental setup does not distinguish between directeffects of PDC on Pkp1/2 or Aup1 activity and the effects of simply PDC dysfunction on mitochondrial proteins. Along those lines, it is unclear whether the few identified proteins with altered phosphorylation states are indeed targets of Pkp1/2 or Aup1. Thus, to be able to support their conclusion, the authors need to include a number of additional controls/strains.

Response: We agree with the reviewer that the phosphoproteomic analysis is not comprehensive. However, it was the best that we were able to do at the time, and it is unlikely that we could distinguish direct from indirect effects using this approach. The purpose of the experiment was to test whether we could identify more global effects of Pda1 on mitochondrial protein phosphorylation. This was in no way intended to imply a direct effect. Rather, it provides an unbiased map for potentially identifying the signaling network(s) involved. This may also include downstream events outside the mitochondrial matrix and even in the cytoplasm. We previously showed a connection of the Aup1-dependent signaling pathway to the RTG retrograde signaling pathway (Journo et al, 2009), and such a global analysis may allow a future understanding of how intra-matrix events can signal to the cytoplasm. However, we do not make any specific claims regarding which effects are direct and which are not. To address the reviewer’s concern, we have now modified the text to clarify these points (Page 9 line 28-Page 10 line 1).In an effort to improve the evidence for PDC- dependent mitochondrial matrix phosphorylation, we will carry out a phosphoproteomic analysis comparing WT cells to cells expressing the lat1K75R mutation. Since this mutation is expected to increase kinase activity, we expect to obtain a clearer picture that will complement the results obtained with the pda1Δ deletion mutant.

Reviewer #1 (Significance (Required)):

The authors continue to explore the mechanisms underlying their initial observation of differential turnover rates for mitochondrial proteins by mitophagy. This current work specifically builds on the previous publication identifying Pkp1/2 and Aup1 as regulators of the phosphorylation state of specific substrate proteins (Kolitsida et al. 2019). Here the authors now explore how these regulators might be organized and controlled by PDC. Thus, the study adds another layer of regulation. However, the molecular mechanisms of how PDC might regulate Pkp1/2 and Aup1 is not directly addressed. In conclusion, the current study opens up new lines of research that need to be explored to critically test the proposed model. A key question to me is of course how mitochondrial proteins can be excluded from mitophagy and whether the proteins in this study may play a direct role in these mechanisms.

To further address the mechanism by which the PDC affects mitophagy via Aup1, Pkp1, and Pkp2, we will carry out the following experiments: 1) We will verify the Pda1-Pkp1/2 interaction and test for effects of PDC mutants on this interaction, with an emphasis on the lat1K75R mutation 3) We will analyze the effect of the lat1K75R mutation on mitochondrial phosphoproteomics 4) we will analyze the effect of the pda1Δ mutation on the phosphorylation state of additional proteins which exhibit defective mitophagy in the pda1Δ background (Figure 1B). Our basic hypothesis regarding the reviewers’ question, namely the molecular mechanism behind the observed regulation is based on mitochondrial heterogeneity. We suggest that mitochondrial heterogeneity underlies mitophagic selectivity and that factors which affect heterogeneity also affect mitophagic selectivity.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

The manuscript by Kolitsida et al. unravels an interesting link between mitochondrial pyruvatedehydrogenase complex (PDC) and protein sorting for mitophagy-mediated clearance. Using yeast genetics combined with immunoblotting-based mitophagy protein sorting reporter assay, authors show that targeting components of PDC and associated regulatory factors (kinases and phosphatases) interferes with phosphorylation status of at least some mitochondrial matrix proteins and prevents their degradation by autophagic machinery. Experimental design is sound, and statistic evaluation seems appropriate. Although I find this work important and overall valuable, some of the conclusions made by the authors are still preliminary and could require additional experimental testing.

We thank the reviewer for these supportive comments

*In this regard, some of my major suggestions include: *

1. Authors propose that the interaction of dedicated kinases and phosphatases with pyruvate dehydrogenase complex changes their affinity/activity toward the non-PDC substrates (possibly by allosteric regulation). Yet, alternatively, it can also be considered that the PDC complex acts as a docking station that stabilizes the associated regulatory factors inside the mitochondria and allows their additional functions. Indeed, simultaneous overexpression of Pkp1/2 increases the free GFP signal and Mdh1 phosphorylation even in the absence of core E1 PDC component (pda1, figure 3 C, D and figure 4 A, B), suggesting that the interaction of phosphatases and kinases with pda1 is not essential to mediate the phosphorylation of other mitoproteins. Are the protein levels of regulatory PDC factors (Pkp1, Pkp2, Aup1) somewhat altered by depletion of their target E1a subunit (pda1)? Could depletion of pda1 trigger destabilization of these factors and decrease their half-life? This can be tested by western-blotting and combined with cycloheximide chase assay.

We agree with the reviewer that, theoretically, the PDC could be necessary for the stability of Pkp1, Pkp2 and Aup1. However, have now tested this, and found that deletion of PDA1 has no effect on the expression levels of Aup1, Pkp1 and Pkp2 under our conditions (see Supplementary Figure 2).

2. Allosteric regulation of specificity of PDC kinases and phosphatases is an intriguing yet still slightly unbacked hypothesis. To provide further experimental evidence for this explanation, authors could develop the classical in vitro kinase assay for Pkp1/2 on their model substrate (Mdh1) in the absence or presence of the Pda1 subunit. Even more excitingly, authors could isolate the Pkp1/2 from wild-type and Pda1- compromised cells and see whether they are able to phosphorylate their model substrate in vitro. Additionally, does specific inhibition of pyruvate dehydrogenase kinases (e.g., with dichloroacetate) affect the Mdh1 phosphorylation levels?

We present the hypothesis that the PDC allosterically regulates Aup1 and Pkp1/2 , as an explanation for the data. We agree with the reviewer that further experimental work will be necessary to verify this hypothesis. While the Pkp1/2 kinase assays could be useful in this regard, several issues temper our enthusiasm for this experiment:

i) This assay is far from “classical” with respect to this study. Neither of the two yeast mitochondrial kinases (Pkp1 and Pkp2) was previously characterized using in vitro kinase assays. While the mammalian assay conditions may work here, this is not guaranteed.

ii) A negative result will have no value

iii) A positive result (e.g. phosphorylation of recombinant Mdh1 by recombinant Pkp1/2) could arise due to low stringency conditions in the assay

iv) The nature of the kinase activity is unclear. We do not know if it is Pkp1, Pkp2 or a heterodimer of Pkp1 and Pkp2, or whether currently unknown ancillary factors are necessary for activity. This greatly complicates the analysis.

In lieu of the direct in vitro kinase assay, we have carried out an experiment demonstrating that an Mdh1-GFP variant with a phosphomimetic threonine to aspartate mutation at position 199, which we previously demonstrated to suppress both the aup1Δ and Pkp2Δ phenotypes (Kolitsida et al, 2019), can also suppress the pda1Δ phenotype (New Figure 3E, F). We suggest that this result provides sufficient evidence of a phosphorylation cascade linking the PDC, Pkp1/2 and Mdh1 mitophagic trafficking.

In addition, we will also attempt to further support the allosteric regulation hypothesis, by testing whether mutations in LAT1 such as lat1K75R modulate the known Aup1-Pda1 and Pkp1/2-Pda1 interactions (also see response to referee #1).

We greatly appreciate the suggestion to test the effect of DCA on our assays, and this experiment will be carried out in the near future. However DCA has not been shown to affect Pda1 phosphorylation in yeast or to modulate kinase activity of Pkp1/2, and it is not clear whether this approach will work. Many reagents and inhibitors which act in mammalian cells, are unable to cross the yeast cell wall.

3. Are the Pkp1/Pkp2 changing their interactome upon PDC alternation (e.g., Pda1 depletion)? Do their interactome alters upon stimulation of mitophagy? Co-immunoprecipitation or bioID assay could shed light on how the partitioning of the Pkp1/Pkp2/Aup1 functions between the metabolic regulation and protein quality control is maintained.

We agree with the referee that this is an excellent question. Our current hypothesis posits that the established physical interaction between Pkp1/2 and Pda1 is crucial for activity of the kinases towards "third party" clients. To test this we will assay whether the catalytically inactive lat1K75R mutation affects the Pkp1/2 - Pda1 interaction. If we cannot find evidence for such a direct mechanism using these straightforward hypothesis-driven experiments, then we will also analyze effects the pda1Δ mutation on the general interactomes of Pkp1 and Pkp2.

Minor points:

1. The paper would benefit from a brief explanation of the principles of the mitophagy reporter assay used in this study.

Thank you for the suggestion. We have now expanded our explanation of the GFP release assay (see Methods section, Page 13 lines 8-16).

2. MS phosphoproteomics is indeed a great approach to tackle many questions in this study. Surprisingly,the authors did not discuss these data extensively. Although changes in the phosphorylation status of some proposed targets are apparent (as QCR2), many others were not detected (including Aco1/2, Idp1, and model substrate - Mdh1; Figure 1).

In our experiment, the coverage of the mitochondrial phosphoproteome was not complete. Some proteins, such as Mdh1, were observed only in a subset of the replicates, and therefore are not included in the figure. In the revision, we plan to add an analysis of the mitochondrial phosphoproteome in the lat1K75R mutation, and hopefully this will provide further insights.

Furthermore, many significant changes occur in the proteins that do not belong to the mitochondrial matrix compartment (as TOMM20 or YAT1), questioning the direct involvement of Pkp1/Pkp2/Aup1. How do authors interpret these data?

We previously showed that the Aup1 signaling pathway converges with the RTG retrograde signaling pathway, which signals from the mitochondrial matrix to the nucleus (Journo et al, 2009). We would like to suggest that these effects on non-matrix proteins may indicate a possible role in transducing this signal. We have now added this comment to the discussion on Page 9 lines 28- Page 10 line 1.

3. Minor spelling mistakes should be reviewed (e.g., "mutophagic" pg. 7).

Thank you. Fixed.

*Reviewer #2 (Significance (Required)): *

*This work further expands on the previous observation by the authors (Kolitsida et al., 2019) and provides a novel and interesting hypothesis that may potentially impact our understanding of mitophagy selectivity toward particular mito proteome content. Furthermore, it unravels the additional function of PDC kinases and phosphatases behind the well-established regulation of energy metabolism. Therefore, it could interest the broad field of researchers interested in mitochondrial quality control and its interplay with cellular metabolism. *

We thank the reviewer for these encouraging comments.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

In this elegant and creative study, Dr. Abeliovich and his collaborators describe a novel pathway for the selective elimination of specific proteins. The flow of the experiments is logical and the writing is clear. The topic is very timely as the subject of mitochondrial quality control is of great interest. The main strength of this study is the creative ideas and the very out-of-the-box concept of selective elimination of specific proteins. The main shortcoming of this study is the incomplete direct evidence for a mechanism and conclusions that exceed the results. I believe these are fixable and I am sure this study will be well received after these issues are addressed.

We thank the reviewer for this positive assessment of the manuscript.

Specific comments:

Major comments:

1. It is not clear what is the mechanism for the recovery of pkp1 and pkp2 shown in figure 4. Supposedly, it is downstream of mitophagy, by how? Is there any data on that?

We presume the referee is referring to Figure 4A and B, where we demonstrate suppression of the pda1Δ phosphorylation phenotype by co-overexpression of Pkp1 and Pkp2. As indicated in the manuscript, we interpret these results as indicating that Pkp1 and Pkp2 function downstream of the PDC in the regulation of mitophagic trafficking. To further bolster this conclusion we now also show that a phosphomimetic mutation in Mdh1 that suppresses the pkp2Δ mutation as well as the aup1Δ mutation (Kolitsida et al, 2019), is also able to suppress the pda1Δ mutation. In addition, we will use co-IPs to test whether mutations in Lat1, and specifically the lat1K75R mutation which increases mitophagy and Mdh1 phosphorylation, can affect the known Pkp1/2-Pda1 interaction.

2. Pertaining to the experiment shown in figure 4, it is not clear if this was conducted in the presence of low glucose and if this was required for the effect.

As pointed out in the legend to Figure 4, the cells were grown in SL medium. As per the “Methods” section, SL contains, in addition to 2% lactate, 0.1% glucose. This small amount of glucose is always added to gluconeogenic media because initial growth in the total absence of glucose is very sluggish. 0.1% glucose does not induce the crabtree effect (does not inhibit respiration) and is consumed during the initial growth phase.

3. It is not clear if the effect is directly mediated by pct 5&7 acting on MDH1 or if this is an indirect effect. Can this be stated and then addressed experimentally? Are pct5 and pct7 the only phosphatases in the matrix?

The purpose of this experiment was to untangle the apparent paradox wherein Aup1/Ptc6 is the opposing phosphatase countering Pkp1/2 on Pda1, but this is clearly not the case for the signaling pathway uncovered in this study. This finding further differentiates our signaling pathway from the conventional PDC regulation cascade, and provides a previously unidentified function for Ptc5. We do not claim that these effects are direct. There are 3 documented phosphatases in yeast mitochondria: Ptc5, Ptc7 and Ptc6/Aup1. Since we have previously shown that the aup1Δ phenotype is very similar to that of pkp2Δ, it disqualifies Aup1 from being a candidate for the opposing phosphatase acting on Mdh1-GFP. The experiment shown in Figure 4D and E indicates that Ptc5 is the prime candidate for a phosphatase that targets Mdh1-GFP. We now clarify this point in the discussion (Page 10, lines 14-24), and we suggest that a more definitive identification will be made in future studies.

4. Figure 6 uses deletion of PDA1 as a loss of function experiment. However in this case, this is a rather crude approach since the point the authors are trying to make is that it is the regulation rather than the expression levels that is critical. If possible, the authors should try and affect the regulatory site by mutation rather than delete the gene.

We fully agree with this suggestion. We will now test the effect of the lat1K75R mutation on the mitochondrial phosphoproteome, in order to address this deficiency.

Minor comments:

1. The title should include that this was done in yeast. Similarly, yeast should be mentioned in the abstract too.

We have now added this information to the abstract. We do not wish to do the same for the title, as most journals have strict limits on the number of characters in the title.

2. In the introduction, it is stated that mitophagy is a process degrading dysfunctional mitochondria. It will be more accurate to say that it is degrading dysfunctional mitochondria as well as removing mitochondria in cells that shrink or rebuild their mitochondrial proteome.

Thank you. We have now made these changes (Page 3, lines 8-11)

3.Are pkp1 and pkp2 found only in the mitochondria?

As per the literature, Pkp1 and 2 are exclusively mitochondrial

4. In figure 1A/B only the blots of MDH1 are shown. It will be informative to show the blots of the other

proteins (currently in the supplementary)

We have added these blots to Figure 1

5. It is not clear how the reporter in fig 1C is different from the reporter in 1A.

This has now been clarified in the text (Page 4 line 28- Page 5 line 2)

Figure 4C. Please add an image of the colonies.

Thank you. This will be added to Figure 4.

6. It is not clear what the author defines as "increased segregation of Mdh1-GFP relative to generic mtRFP". Please clarify "generic mtRFP" and explain how the relativeness was deducted.

Thank you. We have now dropped the qualification “generic” and explain the difference. We now also explain the protocol for quantifying the overlap between the two signals (Page 8 lines 9-12, Page 16 lines 19-21 and 27-31).

Reviewer #3 (Significance (Required)):

This is a very hot topic and this lab is at the front of it It is for broad cell biology and biochemistry audience

We thank the referee for his/her generous and supportive comments

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this elegant and creative study, Dr. Abeliovich and his collaborators describe a novel pathway for the selective elimination of specific proteins. The flow of the experiments is logical and the writing is clear. The topic is very timely as the subject of mitochondrial quality control is of great interest. The main strength of this study is the creative ideas and the very out-of-the-box concept of selective elimination of specific proteins. The main shortcoming of this study is the incomplete direct evidence for a mechanism and conclusions that exceed the results. I believe these are fixable and I am sure this study will be well received after these issues are addressed.

Specific comments:

Major comments:

1. It is not clear what is the mechanism for the recovery of pkp1 and pkp2 shown in figure 4. Supposedly, it is downstream of mitophagy, by how? Is there any data on that?
2. Pertaining to the experiment shown in figure 4, it is not clear if this was conducted in the presence of low glucose and if this was required for the effect.
3. It is not clear if the effect is directly mediated by pct 5&7 acting on MDH1 or if this is an indirect effect. Can this be stated and then addressed experimentally? Are pct5 and pct7 the only phosphatases in the matrix?
4. Figure 6 uses deletion of PDA1 as a loss of function experiment. However in this case, this is a rather crude approach since the point the authors are trying to make is that it is the regulation rather than the expression levels that is critical. If possible, the authors should try and affect the regulatory site by mutation rather than delete the gene.

Minor comments:

1. The title should include that this was done in yeast. Similarly, yeast should be mentioned in the abstract too.
2. In the introduction, it is stated that mitophagy is a process degrading dysfunctional mitochondria. It will be more accurate to say that it is degrading dysfunctional mitochondria as well as removing mitochondria in cells that shrink or rebuild their mitochondrial proteome. 3.Are pkp1 and pkp2 found only in the mitochondria?
3. In figure 1A/B only the blots of MDH1 are shown. It will be informative to show the blots of the other proteins (currently in the supplementary)
4. It is not clear how the reporter in fig 1C is different from the reporter in 1A. Figure 4C. Please add an image of the colonies.
5. It is not clear what the author defines as "increased segregation of Mdh1-GFP relative to generic mtRFP". Please clarify "generic mtRFP" and explain how the relativeness was deducted.

Significance

This is a very hot topic and this lab is at the front of it It is for broad cell biology and biochemistry audience

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The manuscript by Kolitsida et al. unravels an interesting link between mitochondrial pyruvate dehydrogenase complex (PDC) and protein sorting for mitophagy-mediated clearance. Using yeast genetics combined with immunoblotting-based mitophagy protein sorting reporter assay, authors show that targeting components of PDC and associated regulatory factors (kinases and phosphatases) interferes with phosphorylation status of at least some mitochondrial matrix proteins and prevents their degradation by autophagic machinery. Experimental design is sound, and statistic evaluation seems appropriate. Although I find this work important and overall valuable, some of the conclusions made by the authors are still preliminary and could require additional experimental testing.

In this regard, some of my major suggestions include:

1. Authors propose that the interaction of dedicated kinases and phosphatases with pyruvate dehydrogenase complex changes their affinity/activity toward the non-PDC substrates (possibly by allosteric regulation). Yet, alternatively, it can also be considered that the PDC complex acts as a docking station that stabilizes the associated regulatory factors inside the mitochondria and allows their additional functions. Indeed, simultaneous overexpression of Pkp1/2 increases the free GFP signal and Mdh1 phosphorylation even in the absence of core E1 PDC component (pda1, figure 3 C, D and figure 4 A, B), suggesting that the interaction of phosphatases and kinases with pda1 is not essential to mediate the phosphorylation of other mitoproteins. Are the protein levels of regulatory PDC factors (Pkp1, Pkp2, Aup1) somewhat altered by depletion of their target E1a subunit (pda1)? Could depletion of pda1 trigger destabilization of these factors and decrease their half-life? This can be tested by western-blotting and combined with cycloheximide chase assay.
2. Allosteric regulation of specificity of PDC kinases and phosphatases is an intriguing yet still slightly unbacked hypothesis. To provide further experimental evidence for this explanation, authors could develop the classical in vitro kinase assay for Pkp1/2 on their model substrate (Mdh1) in the absence or presence of the Pda1 subunit. Even more excitingly, authors could isolate the Pkp1/2 from wild-type and Pda1-compromised cells and see whether they are able to phosphorylate their model substrate in vitro. Additionally, does specific inhibition of pyruvate dehydrogenase kinases (e.g., with dichloroacetate) affect the Mdh1 phosphorylation levels?
3. Are the Pkp1/Pkp2 changing their interactome upon PDC alternation (e.g., Pda1 depletion)? Do their interactome alters upon stimulation of mitophagy? Co-immunoprecipitation or bioID assay could shed light on how the partitioning of the Pkp1/Pkp2/Aup1 functions between the metabolic regulation and protein quality control is maintained.

Minor points:

1. The paper would benefit from a brief explanation of the principles of the mitophagy reporter assay used in this study.
2. MS phosphoproteomics is indeed a great approach to tackle many questions in this study. Surprisingly, the authors did not discuss these data extensively. Although changes in the phosphorylation status of some proposed targets are apparent (as QCR2), many others were not detected (including Aco1/2, Idp1, and model substrate - Mdh1; Figure 1). Furthermore, many significant changes occur in the proteins that do not belong to the mitochondrial matrix compartment (as TOMM20 or YAT1), questioning the direct involvement of Pkp1/Pkp2/Aup1. How do authors interpret these data?
3. Minor spelling mistakes should be reviewed (e.g., "mutophagic" pg. 7).

Significance

This work further expands on the previous observation by the authors (Kolitsida et al., 2019) and provides a novel and interesting hypothesis that may potentially impact our understanding of mitophagy selectivity toward particular mito proteome content. Furthermore, it unravels the additional function of PDC kinases and phosphatases behind the well-established regulation of energy metabolism. Therefore, it could interest the broad field of researchers interested in mitochondrial quality control and its interplay with cellular metabolism.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The Abeliovich lab has discovered differential degradation rates for mitochondrial proteins by mitophagy, a selective form of autophagy of mitochondria. In previous publications, they identified two protein kinases Pkp1/2 and the protein phosphatase Aup1(Pct6), which regulate phosphorylation patterns of mitochondrial proteins and affect their degradation rates by mitophagy. In the current manuscript, the authors now analyze the role of the pyruvate dehydrogenase complex (PDC) in mitochondria in the regulation of these regulatory link between Pkp1/2 and Aup1 and their substrate proteins. They find that the deletion of pda1, a core subunit of PDC, inhibits the degradation of previously identified mitophagy substrates Mdh1, Aco2, Qcr2, Aco1, and Idp1, while the turnover of mtDHFR-GFP, a model substrate diffusely localized in the mitochondrial matrix, is not affected. These data indicate that tested mitochondrial proteins are excluded from mitochondrial turnover by mitophagy in the absence of Pda1 by as yet unknown sorting mechanisms.

The authors next show that the mutation of a known Pkp1/2 and Aup1 phosphorylation site within Pda1 (S313A) drastically reduces mitophagic turnover of Mdh1, which is completely blocked by the additional deletion of AUP1. A mutation blocking Pda1 activity (R322C) also almost completely blocked Mdh1 degradation. However, PDC activity does not generally affect mitophagy, since an inactivating mutation in Lat1 (K75R) promotes mitophagy of Mdh1. Published work has shown that Pkp1/2 and Aup1, which affect the phosphorylation and degradation of Mdh1, physically interact with the PDC. Thus, the authors tested whether the absence of Pda1 affected the phosphorylation state of Mdh1 and thus its turnover. Indeed, Mdh1 showed globally reduced phosphorylation in the absence of Pda1, which was rescued by the overexpression of Pkp1/2 promoting degradation of Mdh1. Consistent with the effects on mitophagic turnover, Mdh1 was phosphorylated at higher level in the presence of Lat1-K75R. The authors analyzed the mitochondrial distribution of Mdh1 in dependence of Pda1. They provide cytological data suggesting that Mdh1 might segregate from a diffuse matrix localized mtRFP reporter, which is suppressed upon Pkp1/2 overexpression.

Finally, the authors perform a mitochondrial phospho-proteome analysis and found that the absence of Pda1 affected phosphorylation sites in 8 mitochondrial matrix proteins.

From these data the authors propose a model in which the PDC complex controls the activity of the associated protein kinase Pkp1/2 and the phosphates Aup1 by allosteric changes, which in turn regulates the phosphorylation and differential turnover of mitochondrial proteins by mitophagy. They speculate that the PDC and associated factors could resemble large protein kinase complexes as TORC.

Critical points:

1. The authors postulate a structural role for the PDC in regulating Pkp1/2 and Aup1 for controlling mitophagy turnover of certain substrates. While the physical association of Pkp1/2 and Aup1 with PDC has been shown previously, the authors need to critically test their model and assess (a) whether this physical interaction occurs under their experimental conditions and (b) whether the different mutants that affect mitophagy of Mdh1 also affect the physical interaction of Pkp1/2 and Aup1 with PDC. Otherwise, their model, although consistent, remains purely speculative.
2. It is important to test every mutant that affects Mdh1 turnover for effects on mitophagy in general using the mtDHFR-GFP reporter to be able to conclude specific effects on Mdh1 mitophagy. For example, the deletion of Lat1 has been shown to induce mitophagy of a mitochondrial matrix reporter during nitrogen starvation (Bockler and Westermann, 2013). While the authors observe a complete block of Mdh1 turnover in lat1 deletion cells, they do observe increased Mdh1 degradation in Lat1-K75R cells. In line with this reasoning, the authors have identified a number of Mdh1 variants in a previous publication (Kolitsida et al. 2019) that can be used to further explore the observed phenomenon. For example, the Mdh1-T199A could be used to test whether the expression of Pkp1/2 in pda1 deficient cells has specific or general effects on mitophagy. Furthermore, can Mdh1-T199D, which is turned over independent of Pkp1/2, be degraded in the absence of Pda1?
3. The cytological data are not convincing. The imaging quality is low and it is very difficult for the reader to appreciate the suggested segregation of Mdh1-GFP and mtRFP. Thus, it is important to provide cortical sections in order to visualize mitochondrial tubules/networks and Mdh1 distribution. This is particularly important, because the authors mention effects of Pda1 on mitochondrial morphology, which appears to be rescued by Pkp1/2 overexpression. In this context it would be very informative to follow the localization of PDC in mitochondria. Previous work has shown that Pda1 forms punctate structures in mitochondria in proximity to ER-mitochondria contact sites marked by ERMES (Cohen et al. 2014). Is Pda1 itself degraded by mitophagy or is it excluded? Could the physical interaction of Mdh1 with Pda1 explain mitophagy phenotypes if Pda1 is excluded from mitophagy? In other words, could the PDC form a physical unit to segregate and prevent proteins from mitophagy turnover?
4. The conclusions that can be drawn from the analysis of the mitochondrial phospho-proteome in dependence of Pda1 are rather limited. First, the experimental setup does not distinguish between direct effects of PDC on Pkp1/2 or Aup1 activity and the effects of simply PDC dysfunction on mitochondrial proteins. Along those lines, it is unclear whether the few identified proteins with altered phosphorylation states are indeed targets of Pkp1/2 or Aup1. Thus, to be able to support their conclusion, the authors need to include a number of additional controls/strains.

Significance

The authors continue to explore the mechanisms underlying their initial observation of differential turnover rates for mitochondrial proteins by mitophagy. This current work specifically builds on the previous publication identifying Pkp1/2 and Aup1 as regulators of the phosphorylation state of specific substrate proteins (Kolitsida et al. 2019). Here the authors now explore how these regulators might be organized and controlled by PDC. Thus, the study adds another layer of regulation. However, the molecular mechanisms of how PDC might regulate Pkp1/2 and Aup1 is not directly addressed. In conclusion, the current study opens up new lines of research that need to be explored to critically test the proposed model. A key question to me is of course how mitochondrial proteins can be excluded from mitophagy and whether the proteins in this study may play a direct role in these mechanisms.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1

SUMMARY

The manuscript by Smoak et al., provides an analysis of the Hyr/Iff-like (Hil) genes in Candida species with a strong focus on C. auris. The authors demonstrate a repeated expansion of these genes in unique lineages of fungal species, many of which are associated with stronger clinical disease. There is evidence of selection operating on the gene family in the primary domain used for identification. These genes include a repeat just downstream of that core domain that changes frequently in copy number and composition. The location of these genes tends to cluster at chromosome ends, which may explain some aspects of their expansion. The study is entirely in silico in nature and does not include experimental data.

MAJOR POINTS

Altogether, many of the general findings could be convincing but there are some aspects of the analysis that need further explanation to ensure they were performed correctly. To start, a single Hil protein from C. auris was used as bait in the query to find all Hil proteins in yeast pathogens. Would you get the same outcome if you started with a different Hil protein? What is the basis for using Hil1 as the starting point? It also doesn't make sense to me to remove species just because there are already related species in the list. This may exclude certain evolutionary trends. Furthermore, it would be helpful to know how using domain presence and the conservation of position changes the abundance of the gene family across species? (beginning of results).

We appreciate the reviewer’s criticisms on our strategy for identifying Hil proteins. In response, we have significantly revised our pipeline. In particular, we now combine the search results from three queries: in addition to C. auris Hil1’s Hyphal_reg_CWP domain (XP_028889033), we added the Hyphal_reg_CWP sequences from C. albicans Hyr1 and C. glabrata Hyr1. They were chosen as representatives in the two phylogenetic groups distinct from the one containing C. auris in order to avoid the bias due to the query’s phylogenetic position. Using the same criteria as we did for the original search, we identified three additional hits compared with the original 104 homologs list. In response to the criticism of the arbitrary exclusion of some species, we now include any species from the BLASTP search results as long as it is part of the 332 yeast species studied by Shen et al. 2018 (PMID: 30415838). The reason for this criterion is so that we can use the high-quality species phylogeny generated by Shen et al. 2018 to properly study the gene family evolution by reconciling the gene tree with the species tree. We additionally include the species in the MDR clade closely related to C. auris and used Muñoz et al. 2018 (PMID: 30559369) as the basis for the species phylogeny in the clade. Lastly, we no longer require the particular domain organization in classifying Hil family members. All BLASTP hits satisfying the E-value cutoff of 1x10-5 and query coverage > 50% are included.

A major challenge in the analysis like this one is in dealing with repetitive sequences present in amplified gene families. For example, testing modes of selection on non-conserved sites is fraught. It's not clear if all sites used for these tests are positionally conserved and this should be clarified. Alignments at repeat edges will need to maintain this conservation and relatively good alignments as stated in lines 241-242 are concerning that this includes sequence that does not retain this structure necessary for making predictions of selection.

We appreciate the reviewer’s comment. In the original manuscript, we performed two different types of analyses, one on the conserved and well-aligned Hyphal_reg_CWP domain and another on the rapidly evolving repeat region. For the former, we performed phylogenetic dN/dS analyses using maximum-likelihood, for which a reliable alignment is crucial and is the case here. The Hyphal_reg_CWP domain alignment for C. auris Hil1-Hil8 is shown below and also included as Fig. S7 in the revised manuscript: (figure in the response file)

In the text, we added this sentence to emphasize this point: “We chose to focus on the Hyphal_reg_CWP domain because of its potential importance in mediating adhesion and also because the high-quality alignment in this domain allowed us to confidently infer the evolutionary rates (Fig. S7).”

For the repeat domain, what we did in the original version was to calculate the pairwise dN/dS between individual repeat units found in Hil1 and Hil2. This didn’t require aligning the entire repeat regions in the two proteins, but instead relied on the alignment of the individual ~44 aa repeat units, which were highly conserved (see below). In the revised manuscript, however, we decided to focus our analyses on the Hyphal_reg_CWP domain because of a different concern, namely gene conversions between paralogs could distort the evolutionary history of the repeats (the same concern was addressed for the effector domain using an additional step of detecting recombination breakpoints, but the same analysis would be challenging for the repeat region due to alignment issues).

(figure in the response file)

It's also unclear to me why Figure S12 is here. The parameters for this analysis should be tested ahead of building models so only one set of parameters should be necessary to run the test. The evolutionary tests within single genes and across strains is really nice!

We appreciate the reviewer’s suggestion. Based on the reviewer’s suggestion, we removed Fig. S12 and describe the model set up in the Materials and Methods section. We were not sure if the last point was a comment or a suggestion. We didn’t perform a population level selective sweep scan in C. auris. Such an analysis has in fact been attempted by Muñoz et al. 2021, who identified several members of the Hil family as the top candidates for positive selection (PMID: 33769478). We cited this in our Discussion:

“Lastly, scans for selective sweep in C. auris identified Hil and Als family members as being among the top 5% of all genes, suggesting that adhesins are targets of natural selection in the recent evolutionary history of this newly emerged pathogen (Muñoz et al. 2021).”

A major challenge for expanded gene families is rooting based on the inability to identify a strong similarity match for the full length sequence. The full alignment mentioned would certainly include significant gaps. If those gaps are removed and conserved sites only are used, does it produce the same tree? Inclusion of unalignable sequences would be expected to significantly alter the outcomes of those analysis and may produce some spurious relationships in reconciling with the species trees. Whether or not there are similar issues in the alignment of PF11765 need to be addressed as well. There's nothing in the methods that clarifies site selection.

We appreciate reviewer’s comment and agree with the concern about alignment quality affecting phylogenetic reconstruction. To clarify, all phylogenetic analyses in this work are based on the alignment of the Hyphal_reg_CWP domain, which is well aligned (shown above for the subset of eight homologs in C. auris). Alignment of all 215 homologs is provided for readers to review (shorturl.at/kDEJ3). To clarify this choice, we now include the following in Results:

“To further characterize the evolutionary history of the Hil family, including among closely related Candida lineages, we reconstructed a species tree-aware maximum likelihood phylogeny for the Hil family based on the Hyphal_reg_CWP domain alignment (Fig. 1C, Fig. S2).”

We also included detailed steps for reconstructing the gene tree in Materials and Methods.

To test the effect of gaps in the alignment on phylogenetic tree inference, we used two trimming programs, ClipKit (PMID: 33264284) and BMGE (PMID: 20626897), with author-recommended modes. They resulted in consistent gene tree results. We present the tree based on the ClipKit trimmed alignment in the main results. The root of the gene tree was inferred by jointly maximizing the likelihood scores for the gene tree based on the alignment and the evolution of the gene family within the species tree, using GeneRax (Morel et al. 2020, PMID: 32502238).

Figure 1A: the placement of evolved pathogenesis is a little arbitrary. It's just as feasible that a single event increased pathogenesis in the LCA of C. albicans and C. parapsilosis that was subsequently lost in L. elongisporus. These should be justified or I'd suggest removing. The assignment of Candida species here also seems incomplete. The Butler paper notes both D. hansineii and C. lusitaniae as Candida species whereas they are excluded here.

We removed Figure 1 entirely based on this and another reviewer’s comment. We note that there is broad consensus that opportunistic yeast pathogens have independently arisen multiple times, such as C. auris, C. albicans and C. glabrata. Whether Candida pathogens that are more closely related evolved separately or not are subjects of ongoing research (PMID: 24034898).

It is tricky to include scaffolds in analysis of chromosomal location of the HIL genes. The break in the scaffold may be due to the assc repeats of these proteins alone or other, nearby repeats. Any statistics would be best done to include only known chromosomes or those that are strongly inferred by Munoz, 2021. This will change the display of Figure 7, but is unlikely to change the take home message.

We agree with the reviewer’s concern. In the revised manuscript and with more species included, we now only analyze genomes assembled to a chromosomal level, with the exception of C. auris B8441, which is supported by Muñoz et al. 2021 as having chromosome-length sequences. The revised Figure 7 now only includes these results. We also removed the accompanying supplementary figure that showed results based on scaffold-level assemblies.

MINOR POINTS

Line 18: "spp." Should be "spps."

Addressed throughout the revised manuscript.

Line 41: I might rephrase this as "how pathogenesis arose in yeast..."

Accepted (line 43 in revised manuscript).

I might use a yeast-centric example around line 40 for duplication and divergence. This could include genes for metabolism of different carbon sources in S. cerevisiae.

Accepted (lines 47-48)

The Butler paper referenced on line 51 compared seven Candida species and 9 Saccharomyces species

Changed (line 48)

The autors state no other evolutionary analysis of adhesins has been performed but do not acknowledge this study: https://academic.oup.com/mbe/article/28/11/3127/1047032

We appreciate the reviewer pointing this important reference to us. We now cite it in the introduction (line 64) and discussion (line 340)

The first paragraph of the Results could be condensed

Addressed.

How was the species tree in Figure 1A obtained?

The previous figure 1 is now removed. The species tree used throughout the manuscript is based on Shen et al. 2018 with MDR clade species added, based on Muñoz et al. 2018.

Figure 2: In panel A, "DH" and "SS" are not defined. I'd be careful with use of "non-albicans Candida" in Figure 2B. This usually includes C. tropicalis and C. dubliniensis and may confuse the reader.

We removed the DH and SS labels. Instead, we highlighted three clades, which were defined in previous studies. These are the Candida/Lodderomyces clade (based on NCBI taxonomy database), the MDR clade (e.g., Muñoz et al. 2018, PMID: 30559369) and the glabrata clade (e.g., Gabaldón et al. 2013, PMID: 24034898).

How was the binding domain defined to extract those sequences are produce a phylogeny? In building a ML model, how were parameters chosen?

We now provide the following details in the Materials and Methods section:

“To infer the evolutionary history of the Hil family, we reconstructed a maximum-likelihood tree based on the alignment of the conserved Hyphal_reg_CWP domain. First, we used hmmscan (HmmerWeb version 2.41.2) to identify the location of the Hyphal_reg_CWP domain in each Hil homolog. We used the “envelope boundaries” to define the domain in each sequence, and then aligned their amino acid sequences using Clustal Omega with the parameter {--iter=5}. We then trimmed the alignment using ClipKit with its default smart-gap trimming mode (Steenwyk et al. 2020). RAxML-NG v1.1.0 was compiled and run on the University of Iowa ARGON server with the following parameters on the alignment: raxml-ng-mpi --all --msa $align --model LG+G --seed 123 --bs-trees autoMRE.”

The parameters for the ML tree reconstruction is listed on the last line above. The main parameter was the evolutionary model (LG+G), which accounts for rate variations using a gamma distribution. Other protein evolution models, e.g., VT+I+G, were tested and resulted in nearly identical tree topologies.

Figure 3C/D could be just one panel.

The structure predictions are now reorganized and presented on their own in the new Figure 3.

Can you relate more the fungal hit to the Hil proteins conveyed in lines 152-154?

We appreciate the reviewer’s comment, which referred to CgAwp1 and CgAwp3, whose effector domain structures were reported in a recent study (Reithofer et al. 2021, PMID: 34962966). We now discuss them in relation to the predicted Hyphal_reg_CWP structure, by showing them in Figure 3 and describing them in the Results (lines 181) “crystal structures for the effector domains of two Adhesin-like Wall Proteins (Awp1 and Awp3b) in C. glabrata, which are distantly related to those in the Hil family were recently reported, and the predicted structure of one of C. glabrata’s Hil family members (Awp2) was found to be highly similar to the two solved structures (Reithofer et al. 2021)”

Line 168: Should read "Hence, ..."

The original sentence was removed, but this grammatical error was checked for and corrected.

Label proteins along the top of Figure 4 too.

Accepted (in new Figure 4).

Figure 5: for tests of selection, were sites conserved across the group? What does the black number at each node indicate? Dn and Ds are given as decimals. This is based on what attribute? For panel B, it is unclear what each tip denotes i.e., Hil1_tr6. Hil1 is the gene but what is "tr6"?

In the revised manuscript, we provide the multiple sequence alignment for the Hyphal_reg_CWP domain used for the selection analysis as Fig. S7 to illustrate the level of conservation. The black numbers at the internal nodes are numeric indices used to refer to those nodes. In the revised manuscript, we use some of them to refer to the internal branches, e.g., 12…14 in the legend. In the new Figure 5, we do not list the numeric values of Dn and Ds (aka Ka, Ks). Instead, we use a color gradient to represent the estimated dN/dS ratios. The raw estimates are available in the project github repository. Panel B in the original Figure 5 and other panels related to the evolution of the repeats are now removed.

It's unclear why comparison of the PF11765 domain includes the MRD proteins when those aren't included in the comparison to the repeats alone. Could that skew the comparison due to unequal sample numbers or changed variation frequencies in MDR relative to the other two groups?

These results pertaining to the evolution of the repeats are now removed.

Table 2 doesn't add much. This section could probably be reduced to a few sentences since it's highly speculative (intraspecies variation).

Table 2 is now Table S5. We also simplified the result section in the revised version. While the functional implications of the intraspecific variable number of tandem repeats (VNTR) is speculative, it is founded on two bases: 1) the identification of the VNTR is credible, as the copy number variation is consistent within clades but differ between clades, which is not expected if they are caused by assembly errors; 2) experimental studies in S. cerevisiae for the Flo family strongly supported a direct impact of adhesin length on the adhesive phenotype of the cells (PMID: 16086015).

Table 3 is not needed.

Table 3 is now removed.

Figure 6 - color coding in 6A needs to be explained. I'm assuming this is a taxonomical coding.

In the revised Figure 6A, the coloring scheme is consistent with what we used in Figure 1 based on the three clades, and a legend is provided.

Figure 1B is unnecessary. A Model of the protein indicating domains is sufficient here. Figure 1C needs labels for all termini, not just the pathogenic red branches. The figure doesn't provide clear association between adhesin families and the associated species. This could be omitted, especially since Flo is often associated with Saccharomyces species. Figure 1D is unnecessary.

We have removed the original Figure 1.

SIGNIFICANCE

The work here is sorely needed in expanded gene families and in fungi specifically. No analysis at this level has been performed, to the best of my knowledge, in any fungal associated gene family and certainly not in relationship to pathogenic potential. The authors do a good job in citing the foundational literature upon which their study builds in most cases (one exception is noted above). It would be of general interest to those interested in the evolution of virulence, but the analysis is tricky. This is the biggest drawback I currently have as some of the information to assess the results is missing.

We really appreciate the reviewer's positive comments. We agree and plan to explore the relationship between the adhesin family evolution and virulence phenotypes.

Expertise: gene families, evolution dynamics, human fungal pathogens

Reviewer #2

SUMMARY

Gene duplication and divergence of adhesin proteins are hypothesized to be linked with the emergence of pathogenic yeasts during evolution; however, evidence supporting this hypothesis is limited. Smoak et al. study the evolutionary history of Hil genes and show that expansion of this gene family is restricted to C. auris and other pathogenic yeasts. They identified eight paralogous Hil proteins in C. auris. All these proteins share characteristic domains of adhesin, and the structural prediction supports that their tertiary structures are adhesin-like. Evolutionary analysis of protein domains finds weak evidence of positive selection in the ligand-binding domain, and the central domain showed rapid changes in repeat copy number. However, performed tests cannot unambiguously distinguish between positive selection and relaxed selection of paralogs after gene duplication. Some alternative tests are suggested that may be able to provide more unambiguous evidence. Together with these additional tests, the detailed phylogenetic analyses of Hil genes in C. auris might be able to better support the hypothesis that the expansion and diversification of adhesin proteins could contribute to the evolution of pathogenicity in yeasts.

We appreciate the reviewer’s comments and will address specific points below.

MAJOR COMMENTS

The authors present extensive analyses on the evolution of Hil genes in C. auris. There is significant merit in these analyses. However, the analyses conducted so far are incomplete, lacking proper consideration of other confounding factors. Detailed explanations of our major comments are listed below.

1. First, the authors restricted genes in the Hil family to those only containing the Hyphal_reg_CWP domain. Yet, previous work included genes containing the ligand-binding domain or the repeat domain as Hil genes. More justification is needed whether the author's choice represents the natural evolutionary history of Hil genes appropriately. For instance, are the genes only containing the ligand-binding domain monophyletic or polyphyletic? We recommend including the phylogeny of all the Hil candidate genes, to discern whether evolutionary histories of the repeat domain and ligand-binding domain are congruent. Authors can use this phylogeny as justification to focus only on the ligand-binding domain containing genes.

Butler et al. 2009 (PMID: 19465905) defined the Als family and the Hyr/Iff family as having either the N-terminal effector domain or the intragenic tandem repeats (ITR). Their rationale for the latter was that the ITS sequences were often conserved across species. Upon close inspection (Fig. S19,20 in that paper), however, we found that the ITS tend to be conserved in closely related species, but diverged among more distantly related species. Moreover, proteins in those figures that only contain the ITS and not the ligand-binding domains are all missing either the signal peptide, the GPI-anchor or both. This raises questions as to whether these proteins sharing the ITS sequence alone act as adhesins.

More generally, defining the evolutionary history of proteins with multiple domains is complicated by recombination, which causes different parts (e.g., domains) of the protein to have distinct evolutionary histories. In fact, our study and others show that there exist “chimeras” that combine the effector domain from one adhesin family and the repeat sequence found in another (Zhao et al. 2011, PMID: 21208290, Oh et al. 2019, PMID: 31105652). In these cases, one phylogenetic tree is insufficient to describe the evolutionary history of the whole protein. We chose to define the Hil family by the Hyphal_reg_CWP domain and thus focus on the evolutionary history of this region because 1) while tandem repeat regions also contribute to adhesion in yeasts (Rauceo et al. 2006, PMID: 16936142), the effector domain likely plays a more important role in ligand binding and specificity. Therefore, we believe using the effector domain to define a protein family is more likely to group proteins with similar functional properties than if the repeat sequences were used. Also, while putative fungal adhesins lacking a recognizable ligand-binding domain exist, they are rare (Lipke 2018, PMID: 29772751); 2) The repeat region evolved much more rapidly than the effector domain, as we illustrate in Figures 2, 4 and 6 in our revised manuscript. While some repeat units are highly conserved, e.g., the ~44 aa unit found in Hil1-4 in C. auris and close relatives in the MDR clade, many others are short and degenerate, making it difficult to reliably identify homologs sharing the repeat. Besides, since each protein could contain many distinct repeats, it is not clear how one defines two sequences as belonging to the same family if they share one out of six types of repeats. We acknowledge that this definition leaves out the evolutionary history defined by the tandem repeats, which may reveal intriguing evolutionary dynamics, with functional implications. A recent review for the Als family discussed similar definition challenges and partly supported our choice (Hoyer and Cota, 2016, PMID: 27014205).

In the analysis of positive selection, the authors do not adequately control for the effect of recombination on the evolutionary histories of protein sequences, especially given that Hil genes are rich in repetitive sequences. To account for recombination, GARD, an algorithm detecting recombination, should be performed to detect any recombination breakpoints within a protein domain. If recombination did occur within a protein domain, the authors should treat the unrecombined part as a single unit and use the phylogenetic information of this part to proceed with PAML analysis, instead of using the phylogeny of the entire protein domain. The authors should consider doing GARD analysis for the ligand-binding and repeat domains. For the repeat domain, low BS values in Fig. 5C indicate recombination between repeat units. Thus, the authors should analyze each repeat unit with GARD and re-analyze dN/dS.

We deeply appreciate the reviewers’ criticism here. In the revised manuscript, we removed the analysis of the repeat units and followed the reviewers’ suggestion to carry out GARD analysis on the effector domain, which we now show reveals evidence of intra-domain recombination. Using the inferred breakpoints (Fig. S8), we identified two putatively non-recombining partitions and performed all downstream phylogenetic analyses on them separately. The results are presented in Fig. 5 and Table S6. Compared with the previous result based on the entire Hyphal_reg_CWP domain alignment, the new results reveal clearer patterns, including significantly elevated dN/dS on a subset of the branches. Newly added branch-site test results support a role of positive selection on the effector domain during the expansion of the Hil family in C. auris, suggesting functional diversification following gene duplications.

The authors concluded positive selection in the ligand-binding domain based on the branch-wise model of PAML. Yet, w values were not higher than one, and it's unclear whether the difference in selective pressures the authors claimed here is biologically significant. Overall, what the authors present so far seems to be weak evidence of positive selection but is much more consistent with variation in the degree of purifying selection or evolutionary constraint. Using the site-wise model (m7 vs. m8) in PAML would allow the authors to detect which residues of the ligand-binding domain underwent recurrent positive selection. Combining the evolutionary information of protein residues and the predicted 3D structure will provide molecular insights into the biological impact of rapidly evolving residues. This would be a significant addition and raise the significance of the study, besides providing potentially stronger evidence of positive selection.

We appreciate the reviewers’ criticism and suggestions. In the revised manuscript, we performed site tests comparing models M2a vs M1a, M8 vs M7 and M8a vs M8. For partition 1 (P1-414), all three tests were insignificant. For partition 2 (P697-987), the M2a vs M1a test was insignificant (P > 0.05) but M8 vs M7 and M8a vs M7a were both significant at a 0.01 level, and the omega estimate for the positively selected category was estimated to be ~15. The site tests require all branches to evolve under the same selection regime. To relax this constraint, we performed additional branch-site tests by designating the branches with an estimated dN/dS > 10 as the foreground (based on the free-ratio model estimates). This test was significant for both branches at a 0.01 level and the Bayes Empirical Bayes (BEB) procedure identified a total of 5 residues as having been under positive selection. Although three of the five residues, located in the C-terminus of the Hyphal_reg_CWP domain, are part of the α-crystallin domain, we refrain from drawing any functional conclusions because 1) the BEB procedure is known to be lacking power in identifying positively selected residues and 2) we still lack structure-function relationship for the α-crystallin domain. But we agree and believe that this line of analysis is promising in yielding functional insight into the evolution of the effector domain in the family.

MINOR COMMENTS

1. In Fig 1c, the figure legend should include more specific details: which adhesin proteins are shown here? Please specify species names on the species tree

Figure 1 is removed in the revised manuscript

In Fig 3E, secondary structures are labeled with the wrong colors. Sheet: purple, helix: yellow

In the revised manuscript, the structures of SRRP-BR (original 3E) is now colored in a single color.

What's the ligand-binding activity of the b-solenoid fold? How structurally similar are C. auris PF 11765 domains compared to C. glabrata Awp domains? This information will support the role of adhesin for the ligand-binding domain of Hil genes.

We discuss the ligand-binding activity of the β-solenoid as follows in Discussion:

“The elongated shape and rigid structure of the β-helix are consistent with the functional requirements of adhesins, including the need to protrude from the cell surface and the capacity for multiple binding sites along its length that facilitate adhesion. In some bacterial adhesins, such as the serine rich repeat protein (SRRP) from the Gram-positive bacterium, L. reuterii, a protruding, flexible loop in the β-helix was proposed to serve as a binding pocket for its ligand (Sequeira et al. 2018). Such a feature is not apparent in the predicted structure of the Hyphal_reg_CWP domain. Further studies are needed to elucidate the potential substrate for this domain and its mechanism of adhesion.”

We also compare the structures of the C. auris Hil1/Hil7 Hyphal_reg_CWP domain and the CgAwp1/3 in Figure 3, with this in the legend “(C) Crystal structure of the C. glabrata Awp1 effector domain, which is highly similar to C. auris Hil1 and Hil7, but with the disulfide bond in a different location.”

We added a section in the Discussion to comment on the structure-function relationship based on known β-helix (aka β-solenoid) structures. The main insight comes from similar structures identified through DALI searches, many of which are bacterial and viral surface proteins mediating adhesion. The ligand binding pocket and specificity would require additional structural studies to elucidate.

In lines 248-249, the authors should also consider the influence of evolutionary history. For instance, repeats within the same Hil protein appeared later in evolution, compared to Hil gene duplication, and therefore these repeats experience less time for sequence divergence.

In the revised manuscript, we removed the analyses pertaining to the evolution of the repeat regions due to multiple challenges including alignment, potential of gene conversion and recombination. This is an important and intriguing aspect of adhesin family evolution that we plan to follow up in future work.

Although the bioinformatic evidence of C. auris Hil genes acting as adhesins is strong, it is still worthwhile to discuss the experiments of confirming the function of adhesins.

We agree with the reviewer and acknowledge in the revised manuscript the limitation of our work:

“Future experimental tests of these hypotheses will be important biologically for improving our understanding of the fungal adhesin repertoire, important biotechnologically for inspiring additional nanomaterials, and important biomedically for advancing the development of C. auris-directed therapeutics.”

SIGNIFICANCE

Overall, this study is interesting to investigate the evolutionary history of a crucial virulent gene in C. auris. Such evolutionary understanding will help us identify critical molecular changes associated with the pathogenicity of an organism during evolution, providing insights into the emergence of pathogens and novel strategies to cure fungal infections. The research question is important; however, the current analyses on the positive selection are incomplete, so the conclusion is modest so far. We recommend that the authors re-do the PAML analysis with the above considerations. This work will bring more significance to the mycology field if the functional impact of rapid evolution in protein domains can be supported or inferred.

This manuscript is well-written, and the authors also did a great job specifying all the necessary details in the M&M.

We appreciate the reviewers’ positive comments.

Reviewer #3

Summary:

The manuscript by Smoak et al. provides considerable information gleaned from analysis of HYR/IFF genes in 19 fungal species. A specific focus is on Candida auris. The main conclusion is that this gene family repeatedly expanded in divergent pathogenic Candida lineages including C. auris. Analyses focus on the sequences encoding the protein's N-terminal domain and tracts of repeated sequences that follow. The authors conclude with the hypothesis that expansion and diversification of adhesin gene families underpin fungal pathogen evolution and that the variation among adhesin-encoding genes affects adhesion and virulence within and between species. The paper is easy to read, includes clear and attractive graphics, as well as a considerable number of supplementary data files that provide thorough documentation of the sources of information and their analysis.

We appreciate the positive comment.

MAJOR COMMENTS:

• Are the key conclusions convincing?

Overall, the authors' conclusions are supported by the information they present. However, the overall conclusion is stated as a hypothesis and that hypothesis is not particularly novel. The idea that expansion of gene families associated with pathogenesis occurs in the pathogenic species dates back at least to Butler et al. 2009, who first presented the genome sequences for many of the species considered here.

We appreciate the reviewer’s comment. Our main conclusions are 1) the Hil family is strongly enriched in distinct clades of pathogenic yeasts after accounting for phylogenetic relatedness. This enrichment results from independent duplications, which is ongoing between closely related species; 2) the protein sequence of the Hil family homologs diverged rapidly following gene duplication, driven largely by the evolution of the tandem repeat content, generating large variation in protein length and β-aggregation potentials; 3) there is strong evidence for varying levels of selective constraint and moderate evidence for positive selection acting on the N-terminal effector domain during the expansion of the family in C. auris as our focal species. Based on these observations, we propose that expansion of adhesin gene families is a key preliminary step towards the emergence of fungal pathogenesis.

Indeed, some version of this hypothesis has been proposed by several groups before us. We fully acknowledged this in our previous as well as the revised manuscript, by citing Butler et al. 2009 (PMID: 19465905), Gabaldón et al. 2013, 2016 (PMID: 24034898, 27493146). Our study built on these earlier efforts and extended them by addressing several limitations. First, we performed phylogenetic regression to test for the association between gene family size and the life history trait (pathogen or not) in order to properly account for the phylogenetic relatedness. This was not done in previous studies. Second, most earlier studies didn’t construct a family-wide gene tree to fully investigate the evolutionary history of the family. Gabaldón et al. 2013 did a phylogenetic analysis for the Epa family and a few others within the Nakaseomycetes, revealing highly dynamic expansions. In the present study, we expanded this effort by comprehensively identifying homologs within the Hil family in all yeasts and beyond. Third and perhaps the most important novelty in our study is our detailed analysis of sequence divergence and role of natural selection during the evolution of the family post duplication. This allowed us to present a complete picture of the family’s evolution, not just in its increase in copy number but also its diversification after the duplications, which is a key part of how gene duplications contribute to the evolution of novel traits. As such, we believe our study provides strong support for the above hypothesis.

One key issue with a manuscript of this type is whether genome sequence data are accurate. The authors are not the first research group to take draft genome sequence data at face value and attempt to draw major conclusions from it. The accuracy of public genome data continues to improve, especially with the emergence of PacBio sequencing. Because the IFF/HYR genes contain long tracts of repeated sequences, genome assemblies from short-read data are frequently inaccurate. For example, is it reasonable to have confidence that the number of copies of a tandemly repeated sequence in a specific ORF is exactly 21 (an example taken from Table 2) when each repeat is 40+ amino acids long and highly conserved? Table S6 would benefit from inclusion of the type of sequence data used to construct each draft genome sequence. It is also reasonable to question whether the genome of the type strain is used as a template to construct the draft genomes of the other strains. If that was standard practice, conservation of the repeat copy number among strains might be an artefact. Conservation of repeat sequences to the degree shown is not a feature of the ALS family, a point of contrast between gene families that could be explored in the Discussion.

We appreciate the reviewer’s comment and agree strongly that a key limitation in gene family evolution studies like ours is the quality of the genome assembly. In the original manuscript, we took several steps to ensure the completeness and accuracy of the Hil family homologs, primarily by basing our results on the high quality RefSeq collection of assemblies, and supplementing it with two fungi-specific databases. In the revised manuscript, we performed further quality analyses to assess and correct for inaccuracy in the BLASTP hits. Because RefSeq aims to provide a stable reference, it is often slow in replacing older assemblies with newer ones based on improved technologies. We thus compared the RefSeq hits for species in which a newer, long-read based assembly had become available. The results are documented in Text S1 and in summary, while we did find examples of missing homologs and inconsistent sequences, the problems were isolated to specific species and the inconsistency pertains only to the tandem repeat regions. Regarding the specific example of within-species variable number of tandem repeats (VNTR) in C. auris Hil1-Hil4, we are confident of both the copy number and the sequence variation for two reasons. First, all C. auris strain genomes analyzed in this study were assembled de novo rather than based on a reference genome, and all were long-read based (PacBio) (Table S4). Second, empirically, we found the VNTR identified in Hil1-Hil4 agree among strains within one of the four clades of C. auris while differing between clades (Table S5). Since assembly errors are not expected to produce clade-specific patterns, we believe this is strong evidence for the VNTR identified being real.

We also appreciate the reviewer’s suggestion on discussing the conservation of the repeats as an interesting trait for a group of Hil proteins in comparison to the Als family. We now added a section in Discussion focusing on the special properties of this group of Hil proteins.

• Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

Due to the nature of my comments, this review will not be anonymous. I will include some of the data from my laboratory to further illustrate the point about the quality of draft genome sequences, especially for gene families that contain repeated sequences. My laboratory group has spent the past several years looking at the families of cell wall genes in these species and know that the C. tropicalis genome sequence used in the current analysis is highly flawed. There is even a manuscript from several years ago that documents problems in the assembly (doi: 10.1534/g3.115.017566). There is a new PacBio sequence available that has considerably improved data for this group of genes, but still is not perfect. We designed primers and amplified the various coding regions to verify whether the IFF/HYR were correct in the draft genome sequences. For C. tropicalis, we know that 7 of the genes listed in this paper are broken (i.e. prematurely terminated) giving a false impression of their construction. The current study did not verify any gene sequences, so broken/incomplete genes are a stumbling block for developing conclusions.

We deeply appreciate the reviewer pointing out the flaws in the C. tropicalis genome. Using the PacBio sequence-based new assembly, we were able to confirm the reviewer’s comment on the sequence and annotation error in the RefSeq assembly for C. tropicalis. We listed the comparisons between the two assemblies in Table S8. Because the differences reside outside of the Hyphal_reg_CWP domain, they don’t impact our phylogenetic analyses, which are based on the effector domain alignment. To determine if this is a widespread issue affecting all genome assemblies based on older technologies, and in response to the reviewer’s criticism, we systematically checked the sequences of BLASTP hits based on the RefSeq assemblies against newer, long-read based ones when available. As detailed in Text S1 in the revised manuscript, the problems seen in C. tropicalis were not observed in four other species. While the sample size is small, we believe the issues with C. tropicalis are likely due to a combination of specific issues with the original assembly and special properties of the genome.

Similarly, the recent work from Cormack's lab features a PacBio C. glabrata sequence (doi: 10.1111/mmi.14707). The paper details how the authors focused on accurate assembly of the types of genes studied here. Sequences from the current project should be compared to the PacBio assembly to determine if they provide the same results.

We compared the sequences of the three C. glabrata Hil homologs identified in the RefSeq assembly (GCF_000002545.3) to the best BLAST hits in one of the new Cormack lab assemblies for (BG2 strain, GCA_014217725.1). Two of the three proteins showed identical sequences between the assemblies. One of them is longer in the new assembly than in the RefSeq (1861 vs 1771 aa, XP_447567.2, QHS67215.1). The main difference, however, was the number of hits recovered. Performing BLASTP searches in the new assembly recovered 13 hits vs 3 from the RefSeq assembly, of which 12 were in the subtelomeric region. For this reason, we used the new assembly as the basis for the Hil homologs in our subsequent analyses. To determine if we missed homologs in other genomes due to incomplete subtelomeric regions in the RefSeq assemblies, we repeated the BLASTP search in four other genomes (Text S1). In one of the four species, C. nivariensis, we recovered one more homolog than in the RefSeq. In all other three, we identified the same number (S. cerevisiae: 0, K. lactis: 1, C. albicans: 12), suggesting that the issues seen in C. glabrata is likely specific to this species and its RefSeq assembly.

Another part of the study that deserves additional attention or perhaps altered presentation is the idea that the Iff/Hyr N-terminal domain binds ligands. The literature on the Iff/Hyr proteins is limited. In my opinion, though, the authors of this paper could more completely present the information that is known. The paper by Uppuluri et al. is cited (doi: 10.1371/journal.ppat.1007056), but I did not see any information about their data regarding interaction of C. albicans Hyr1 with bacterial proteins mentioned in the manuscript under review. It is formally possible that the N-terminal domain of Iff/Hyr proteins does not bind a ligand. The current manuscript includes a great deal of speculation on that point, suiting it better to a Hypothesis and Theory format rather than other types of publications.

We appreciate the reviewer’s criticism and suggestion. We made two revisions based on the comments. First, we no longer refer to the Hyphal_reg_CWP domain as ligand-binding. Instead, we refer to it as the effector domain, following existing practices in the field (Lipke 2018, PMID: 29772751, de Groot et al. 2013, PMID: 23397570). Second, during the description of the predicted structure for the domain, we mentioned that it lacks an apparent binding pocket as suggested/identified in other β-solenoid proteins with carbohydrate binding abilities. Therefore, we suggest that the potential substrate and mechanism of binding by this domain remain to be determined with further experiments. We do, however, believe that there is strong evidence for the domain being involved in adhesion. A recent study (Reithofer et al. 2021) presented structural and phenotypic characterization of three Adhesin wall-like proteins (Awp1,2,3) in C. glabrata. In particular, experimental studies of CgAwp2, a Hil family protein, showed that its deletion resulted in the reversion of the hyperadhesive phenotype in one of the C. glabrata strains. Plastic was one of the substrates being evaluated, although, as the reviewer’s work pointed out, adhesion to plastics doesn’t indicate ligand binding, as it can be mediated by non-specific hydrophobic interactions (Hoyer and Cota 2016, PMID: 27014205). Nonetheless, the results presented in Reithofer et al. 2021 and other lines of evidence presented in the current work strongly supported adhesin functions of the Hil family.

Table 1 attempts to offer evidence that the Iff/Hyr N-terminal domain has adhesive function but falls short of convincing the reader. One of the example structural templates is a sugar pyrophosphorylase that seems irrelevant to the current discussion. In the column called "Function", the word adhesin is found several times, but no detail is presented. The only entry that offers an example ligand indicates that the domain binds cellulose which is not likely relevant for mammalian pathogenesis, the main focus of the work. Other functions listed include self-association and cell aggregation--using the N-terminal domain. It is formally possible that Iff/Hyr proteins drive aggregation using the N-terminal domain and beta-aggregation sequences in the repeated region. The authors should develop these ideas further. Discussion of adhesive/aggregative function related to the ALS family can be found in Hoyer and Cota, 2016 (doi: 10.3389/fmicb.2016.00280).

We appreciate the reviewer’s comments. In the revised manuscript, we removed Table 1, which was based on I-TASSER identified templates. Instead, we identified similar structures in the PDB50 database to the AlphaFold2 prediction for the Hyphal_reg_CWP domain in C. auris Hil1 using DALI (Table S3). We described the functional implications based on this list as follows:

“We identified a number of bacterial adhesins with a highly similar β-helix fold but no α-crystallin domain (Table S3), e.g., Hmw1 from H. influenzae (PDB: 2ODL), Tāpirins from C. hydrothermalis (PDB: 6N2C), TibA from enterotoxigenic E. coli (PDB: 4Q1Q) and SRRP from L. reuteri (PDB: 5NY0). For comparison, the binding region of the Serine Rich Repeat Protein 100-23 (SRRP100-23) from L. reuteri was shown in Fig. 3F (Sequeira et al. 2018). Together, these results strongly suggest that the Hyphal_reg_CWP domain in the C. auris Hil family genes mediate adhesion.”

One line of evidence that suggest the Hyphal_reg_CWP domain may have ligand-binding activity is from the L. reuteri SRRP-BR, which is one of the bacterial adhesins identified as having a highly similar β-helical structure (but missing the α-crystallin domain). In Sequeira et al. 2018 (PMID: 29507249), the authors showed via both in-vitro and in-vivo experiments that this domain “bound to host epithelial cells and DNA at neutral pH and recognized polygalacturonic acid (PGA), rhamnogalacturonan I, or chondroitin sulfate A at acidic pH”. However, the predicted structure for the Hyphal_reg_CWP domain in C. auris Hil1 and Hil7 lack a protruding, flexible loop in the β-helix, which was proposed to serve as a binding pocket for the ligand in SRRP-BR. We therefore commented in the text “Such a feature is not apparent in the predicted structure of the Hyphal_reg_CWP domain. Further studies are needed to elucidate the potential substrate for this domain and its mechanism of adhesion.”

We also appreciate the reviewer’s suggestion to discuss the potential role of the Hil proteins in mediating adhesion vs cell aggregation. We now have a section in Discussion that focuses on the potential role of the β-aggregation sequences especially in the subset of Hil proteins led by C. auris Hil1-Hil4, which have an unusually large number of such sequences. We discuss the recent literature suggesting the potential of such features mediating cell-cell aggregation.

The incredibly large number of figures that focus on the repeated sequences in the genes does not appear to include mention of the idea that these regions are frequently highly glycosylated. Knowing how much carbohydrate is added to these sequences in the mature protein would also have bearing on whether the beta-aggregation potential is realized. The Iff/Hyr proteins could stick to other things based on ligand binding (adhesion), hydrophobicity, aggregative activity, etc. Not much is really known about protein function so the conclusions are only speculative. The authors are largely accurate in presenting their conclusions as speculative, but the conclusions are not developed fully and always land on the idea that the N-terminal domain has adhesive function when that aspect clearly is not known.

We appreciate the reviewer’s comment. We have performed N- and O-glycosylataion predictions for the Hil family proteins in C. auris as a focal example and presented the results in Figure 2 of the revised manuscript. Briefly, we found that all eight proteins are predicted to be heavily O-glycosylated (Fig. 2C). N-glycosylation is rare except in Hil5 and Hil6, in regions with a low Ser/Thr content (Fig. 2C). We also deemphasized the ligand-binding ability of the effector domain and its importance in assessing the adhesin function of the Hil family proteins. At the same time, we highlighted other mechanisms as the reviewer pointed out, such as aggregative activities, in our discussion on the potential importance of the large number of β-aggregation motifs.

Another aspect of the analysis that is not mentioned is that several of the species discussed are diploid. What effect does ploidy have on the conclusions? Most draft genomes for diploid species are presented in a haploid display, so are not completely representative of the species. Additionally, some species such as C. parapsilosis are known to vary between strains in their composition of gene families, with varying numbers of loci in different isolates.

We appreciate the reviewer raising this issue. The potential impact of having diploid genomes represented as haploids is twofold. First, if the genome sequencing was performed on a diploid cell sample with some highly polymorphic regions, that would present difficulties to the assembly and could result in poorly assembled sections. Second, either because of the first issue, or because the researchers used the haploid phase of the organism for sequencing, the representative haploid genome will not be “completely representative of the species” as the reviewer suggested. The second problem is not specific to diploids – even for haploids, any single or collection of genomes would represent just a slice of the genetic diversity in the species. We did two things to look into this. First, we analyzed multiple strains in C. auris to reveal both Hil family size variation and also sequence polymorphism, particularly in the tandem repeat region. We also, as part of the quality control, compared and searched assemblies for different strains of some species when available. We agree that characterizing multiple genomes in a species is important for fully revealing the gene pool diversity and could have important consequences on our understanding of the emergence of novel yeast pathogens.

Regarding the first issue, we checked the original publications for two large-scale yeast genome sequencing projects that included 10 of the 32 species in the present study (Dujon et al. 2004, PMID: 15229592 and Butler et al. 2009, PMID: 19465905). In Dujon et al. 2004, the authors stated that haploid cells were used in cases where the species has both haploid and diploid phases. In Butler et al. 2009, the authors said in the Methods that “for highly polymorphic regions of diploid genomes, initial sequence assemblies were iteratively re-assembled in regions of high polymorphism to minimize read disagreement from the two haplotypes while maximizing coverage.”. Therefore, the potential issue of heterozygosity is likely minimal. In addition, many diploid yeasts have large regions in their genomes being homozygous, both as a result of clonal expansion and also due to loss of heterozygosity (LOH), as documented in C. albicans and other Candida species (e.g., PMID: 28080987). Nonetheless, we acknowledge that this issue is yet another challenge to having high-quality, complete genome assemblies. In the discussion, we fully acknowledge the limitation of our study by genome assemblies, and believe that ongoing improvement thanks to the development of long-read technologies will allow more in-depth studies, particularly in the subtelomeric regions and for repeat-rich sequences.

The manuscript concludes that having more genes is better, that the gene family represents diversification that must be driven by its importance to pathogenesis, without recognizing that some species evolve toward lower pathogenesis. This concept could be explored in the Discussion. …My own experience makes me wonder if the authors found any examples of species that provide an exception to the idea that having more genes is better and positively associated with pathogenesis. The parallel between IFF/HYR and ALS genes is made many times in the manuscript. Spathaspora passalidarum, a species that is not pathogenic in humans, but clearly within the phylogenetic group examined here, has 29 loci with sequence similarity to ALS genes. How many IFF/HYR genes are in S. passalidarum?

We appreciate the reviewer’s comment. We will address the two comments above together as they are related. First of all, S. passalidarum is now included in our extended BLAST search list and we identified a total of 3 homologs in this species. When compared with the related Candida/Lodderomyces clade, which includes C. albicans, the Hil family in this species is relatively small (3 vs. >10). More generally, we observe a significant correlation between the Hil family size and the species’ pathogenic potential (Figure 1B and the phylogenetic regression result in the text).

Regarding the first comment, we did identify two species that had a large Hil family (>8 based on C. auris) and yet were not known to infect humans. One of them, M. bicuspidata, has 29 Hil homologs and is interestingly a parasite for freshwater animals, such as Daphnia. The other species, K. africana, has 10 homologs and its ecology is not well described in the literature. With respect to the relationship between adhesin family and pathogenicity, we would like to make two points. First, as mentioned above, we observed a strong correlation between the Hil family size and the pathogen status, after correcting for phylogenetic relatedness, suggesting that expansion of the Hil family is a shared trait among pathogenic species. This doesn’t rule out the possibility that some species may have an expanded adhesin family, such as the example the reviewer mentioned, for reasons other than infecting a human host. Second, a key point in our work is that expansion of the adhesin family is only the first step – the crucial contribution of gene duplications to adaptation is not just in the increase in copy number, but also in providing the raw materials for selection to generate novel phenotypes. On that front, we documented the rapid divergence of the central domains both between and within species, as well as signatures of relaxed selective constraint and positive selection acting on the effector domain following gene duplications in C. auris, both of which support the above theme.

There are several current taxonomies for the species in this region of the tree. The source of the names used in this paper could be specific more completely.

We appreciate the reviewer’s comment. We now gave the complete Latin names for all species in Figure 1 and only use abbreviated names, e.g., C. auris, after the first occurrence. For species with multiple names in the literature, we followed the species name and phylogenetic placement in Shen et al. 2018 (PMID: 30415838).

The Results and Discussion sections are largely redundant. The tone of the paper is conversational, making it easy to read, but there seems little left to say in the Discussion that has not already been mentioned as the background for the various types of analyses. The authors should revise the paper to eliminate discussions of published literature from the Results and expand the Discussion to include some of the themes that have not been mentioned yet.

We appreciate the reviewer’s comment. In the revised manuscript, we have moved discussion points from the Result to the Discussion section. We also overhauled the Discussion to focus on the implications based on, but not already covered, in the Result part, including the points the reviewer suggested, e.g., the implications of the structure on adhesion mechanism.

Another point that the authors do not mention is documented recombination between IFF and ALS genes (doi: 10.3389/fmicb.2019.00781) and the effect of that process on evolution among these gene families.

We appreciate the reviewer’s comment. We now mention this and related observations in Discussion as part of the discussion on the mutational mechanisms for the evolution of the family:

“Diversification of adhesin repertoire within a strain can arise from a variety of molecular mechanisms. For example, chimeric proteins generated through recombination between Als family members or between an Als protein’s N terminal effector domain and an Hyr/Iff protein’s repeat region have been shown (Butler et al. 2009; Zhao et al. 2011; Oh et al. 2019). Some of the adhesins with highly diverged central domains may have arisen in this manner (Fig. S10).”

My reading of the work by Xu et al. 2021 (doi: 10.1111/mmi.14707) does not match the direction of its presentation in the current paper. Oh et al., 2021 (doi: 10.3389/fcimb.2021.794529) discussed that point recently, providing another point for the Discussion in the current paper.

We appreciate the reviewer’s comment and agree that our original reading of Xu et al. 2021 was incorrect. Instead of suggesting a higher mutation rates in the subtelomeric region, the authors instead suggested the evolution of the Epa family in the subtelomere was driven by Break-Induced Replication. We now update our discussion in the following way, also citing Oh et al. 2021

“Finally, as reported by (Muñoz et al. 2021), we found that the Hil family genes are preferentially located near chromosomal ends in C. auris and also in other species examined (Fig 7), similar to previous findings for the Flo and Epa families (Teunissen and Steensma 1995; De Las Peñas et al. 2003; Xu et al. 2020; Xu et al. 2021) as well as the Als genes in certain species (Oh et al. 2021). This location bias of the Hil and other adhesin families is likely a key mechanism for their dynamic expansion and sequence evolution, either via ectopic recombination (Anderson et al. 2015) or by Break-Induced Replication (Bosco and Haber 1998; Sakofsky and Malkova 2017; Xu et al. 2021). Another potential consequence of the subtelomeric location of Hil family members is that the genes may be subject to epigenetic silencing, which can be derepressed in response to stress (Ai et al. 2002). Such epigenetic regulation of the adhesin genes was found to generate cell surface heterogeneity in S. cerevisiae (Halme et al. 2004) and leads to hyperadherent phenotypes in C. glabrata (Castaño et al. 2005).”

I might have missed it, but I could not find what constitutes a BLAST-excluded sequence (Table S7). Additional explanation (or making the explanation easier to find) would help the reader.

We apologize for the inadvertent mistake of leaving out Table S7. In the revised manuscript, we include all hits from species that are part of the 322 species phylogeny in Shen et al. 2018. Thus, we removed the original Table S7.

• Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

• Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

Ideally, validation of all sequences would provide a stronger foundation for the work. However, that request is not realistic in terms of time or resources.

We agree with the reviewer and appreciate the understanding. In the revised manuscript, we performed additional analyses to evaluate the accuracy and correct the sequences of the BLASTP hits from RefSeq database by comparing them to long-read based assemblies when possible. Please see previous replies to reviewers’ comments and Text S1 for details.

• Are the data and the methods presented in such a way that they can be reproduced?

Yes, the data and methods are documented clearly and perhaps too thoroughly in many places. A considerable amount of confidence is placed in sequences that might not be accurate and tracking details down to the amino acid residue may not be reasonable in this context. A disclaimer might help--everyone probably already knows that genome sequences are not perfect but stating that the analysis is only as good as the genome sequence acknowledges that fact.

We appreciate the reviewer’s comment. In the revised manuscript, we tried to strike a balance between providing enough methodological details for the readers to assess the conclusions and yet also keeping the flow of the paper. We also accepted the reviewer’s suggestion by adding a disclaimer in the Discussion:

“we acknowledge the possibility of missing homologs in some species and having inaccurate sequences in the tandem-repeat region. We believe the expected improvements in genome assemblies due to advances in long-read sequencing technologies will be crucial for future studies of the adhesin gene family in yeasts.”

• Are the experiments adequately replicated and statistical analysis adequate?

The idea of replicates does not really apply to this analysis. I think that the species sampled are reasonable to represent the region of the phylogenetic tree on which the analysis is focused. The authors clearly documented their computational methods in an admirable way.

We appreciate the reviewer’s comment.

MINOR COMMENTS:

Figure 1 has elements that would make a nice graphical summary, but most of it should not be part of the final manuscript. For example, Panel A is repeated in Figure 2. It is not clear what Panel C means until the reader gets to Figure 2. Panel D is unnecessary. The image in Panel B is a good graphic. Endothelial adhesion is not mentioned, though. It is also debatable whether the proteins bind directly to plastic or to the body fluids that coat the plastic.

Based on this and another reviewer’s comments, we removed Figure 1 from the revised manuscript.

Compared to Figure 1, the information in Figure 3 is inconsistent. The "central domain" in Panel A is not central to anything as drawn, located at the end of the protein. The figure should be revised to be consistent with the majority of the authors' results.

We appreciate the reviewer’s suggestion. The terminologies used to describe the different parts of a typical yeast adhesin vary in the literature. In the Als family literature, central domain refers to the region after the N-terminal effector domain and before the C-terminal Ser/Thr-rich stalk domain. In the Hil family proteins, there is not a clear distinction between a “central” and a “stalk” region. In Boisramé et al. 2011 (PMID: 21841123), the authors referred to the region between the Hyphal_reg_CWP domain and the GPI-anchor as the central domain. We adopted that use. We realize that this can lead to confusion especially for Als researchers. In some other literature, e.g., Reithofer et al. 2021, this part of the protein is referred to as the B-region. But we couldn’t find wide use of that term. We decided to stay with “central domain” in this work and hope that by defining the term in Figure 2A, we would avoid any confusion within the scope of this work.

Are the low-complexity repeats mentioned in the Figure 4 legend present anywhere else in the C. auris genome or elsewhere among the species used in this study? The answer to that question may also provide evolutionary clues.

We did find one other putative GPI-anchored cell wall protein containing this ~44aa repeat unit, but with a different effector domain (GLEYA, PF10528). This protein (PIS58185.1 in C. auris B8441), appears to be a hybrid between the repeat region of C. auris Hil1 and an N-terminal effector domain of a different family. This result fits the theme of the reviewer’s work in C. albicans and C. tropicalis on the chimeric adhesins formed between the Als and Hyr/Iff families. Due to the scope of the current work, we omitted this finding from the main result.

Figure S1 legend. How was the distance to C. glabrata measured to call it equal?

The original Figure S1 was removed in the revised manuscript. A consistent set of criteria was employed in deciding which BLASTP hits to include as Hil family members.

Figure S4 could be presented better. Both diagonals have the same information. One could be emptied or could alternatively present nucleotide identity.

The original Figure S4 was removed in the revised manuscript.

Italicize the species names in Panel C of Figure S8.

The original Figure S8C is now Figure S9 and we systematically checked to make sure that species Latin names are italicized. Thanks for pointing this out.

Lines 256-257: The paper selectively samples the Iff/Hyr family and does not examine the "entire" family. Please revise.

We appreciate the reviewer’s comment. In the revised manuscript, we no longer selectively sample species. Instead, we only exclude three species that are not part of the 322-yeast species phylogeny in Shen et al. 2018 and Muñoz et al. 2018, namely Diutina rugosa, Kazachstania barnettii and Artibeus jamaicensis. Our extensive BLASTP searches also indicated that the family as defined in this work is specific to the budding yeast subphylum. We therefore believe it is accurate to describe the work as examining the entire Hil family.

• Are prior studies referenced appropriately?

I was disappointed to see that the paper does not reference my laboratory's work at all. When ALS genes are featured so strongly in a report, it seems reasonable to include something we have done over 30+ years. Our most-recent ALS paper (Oh et al., 2021 doi: 10.3389/fcimb.2021.794529) would be a reasonable source for defending the gene numbers used in Figure 2A. Other examples of our work that directly relate to concepts in this paper were mentioned above.

We sincerely apologize for our negligence. We are new to the fungal adhesin field through an accidental finding, and despite our effort to digest the relevant literature, we did unfortunately overlook the extensive work done on the Als family, much of which came from the reviewer’s lab. We have carefully read the papers suggested by the reviewer as well as others, and now have better incorporated prior foundational and insightful work into our result and discussion sections (see previous replies to the reviewer’s comments).

• Are the text and figures clear and accurate?

Suggestions for improvement are incorporated into the comments above.

• Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

Please present Methods and Results in the past tense. I still make the same mistake when I try to get my ideas on the page but proofread one more time and ensure the verb tenses are accurate.

We appreciate the reviewer’s comments and have edited the Methods and Results sections accordingly.

SIGNIFICANCE

• Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.

The paper reads as if it is presenting preliminary data for a grant proposal. Perhaps Prof. He's lab wants to seek functional evidence for the role of the Iff/Hyr proteins. The current paper provides an exhaustive background for such a pursuit. As presented, there is little functional data for these proteins, genome sequences are not 100% accurate, but the trends noted are defendable.

We appreciate the reviewer’s comments. We acknowledge that experimental studies will be needed to prove and further establish the functional importance of our findings. However, we believe our gene family evolutionary studies provided important novel insights and serve as an example for adhesin family evolution.

• Place the work in the context of the existing literature (provide references, where appropriate).

The ideas presented here are similar to those pioneered in the Butler et al. Nature paper in 2009 (doi: 10.1038/nature08064). We now have the benefit of more genome sequences so the analysis can encompass more species. C. auris adds a newer focus on part of the phylogenetic tree that was not previously emphasized. The idea of "more is better" is very simplistic, though. Parallel work for the ALS family shows complexity in gene expression levels, suggesting that some adhesins are poised to make a large contribution while others are likely to have a scant presence on the cell surface. Those concepts are not really explored in the current paper, either. See Hoyer and Cota 2016 (doi: 10.3389/fmicb.2016.00280); Oh et al. (doi: 10.3389/fmicb.2020.594531).

We appreciate the reviewer’s comments and have included a discussion about the potential diversity of the duplicated Hil family proteins, in terms of function and their regulation in the Discussion. Also see our response to the first comment of the reviewer regarding the novelty of our hypothesis and the significance of our findings.

• State what audience might be interested in and influenced by the reported findings.

Potential readers would come from the fields of fungal adhesion and pathogenesis, as well as evolutionary biology.

We appreciate the reviewer’s comments.

• Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

I discovered and named the ALS gene family in C. albicans and have spent 30+ years characterizing it. Most recently, my lab has focused on providing an accurate gene census and validated gene sequences for the cell wall "adhesinome" in the pathogenic Candida species. Some families are expanded and some are not. Some proteins appear only in a few species and demonstrate key roles in host-fungus interactions. There are many nuances to interpretation of what these fungi are doing from the standpoint of cell-surface adhesins and we look forward to exploring these ideas across many genomes, using validated gene sequences. We have a tremendous dataset that might make good fuel for a collaboration with Prof. He, given his enthusiasm for this area of study, as well as his outstanding expertise and perspectives on evolutionary analyses.

We sincerely thank the reviewer for the critical analysis of our manuscript and appreciate the many suggestions for improving the manuscript.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

The manuscript by Smoak et al. provides considerable information gleaned from analysis of HYR/IFF genes in 19 fungal species. A specific focus is on Candida auris. The main conclusion is that this gene family repeatedly expanded in divergent pathogenic Candida lineages including C. auris. Analyses focus on the sequences encoding the protein's N-terminal domain and tracts of repeated sequences that follow. The authors conclude with the hypothesis that expansion and diversification of adhesin gene families underpin fungal pathogen evolution and that the variation among adhesin-encoding genes affects adhesion and virulence within and between species. The paper is easy to read, includes clear and attractive graphics, as well as a considerable number of supplementary data files that provide thorough documentation of the sources of information and their analysis.

Major comments:

• Are the key conclusions convincing?

Overall, the authors' conclusions are supported by the information they present. However, the overall conclusion is stated as a hypothesis and that hypothesis is not particularly novel. The idea that expansion of gene families associated with pathogenesis occurs in the pathogenic species dates back at least to Butler et al. (2009; doi: 10.1038/nature08064) who first presented the genome sequences for many of the species considered here.

One key issue with a manuscript of this type is whether genome sequence data are accurate. The authors are not the first research group to take draft genome sequence data at face value and attempt to draw major conclusions from it. The accuracy of public genome data continues to improve, especially with the emergence of PacBio sequencing. Because the IFF/HYR genes contain long tracts of repeated sequences, genome assemblies from short-read data are frequently inaccurate. For example, is it reasonable to have confidence that the number of copies of a tandemly repeated sequence in a specific ORF is exactly 21 (an example taken from Table 2) when each repeat is 40+ amino acids long and highly conserved? Table S6 would benefit from inclusion of the type of sequence data used to construct each draft genome sequence. It is also reasonable to question whether the genome of the type strain is used as a template to construct the draft genomes of the other strains. If that was standard practice, conservation of the repeat copy number among strains might be an artefact. Conservation of repeat sequences to the degree shown is not a feature of the ALS family, a point of contrast between gene families that could be explored in the Discussion. - Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

Due to the nature of my comments, this review will not be anonymous. I will include some of the data from my laboratory to further illustrate the point about the quality of draft genome sequences, especially for gene families that contain repeated sequences. My laboratory group has spent the past several years looking at the families of cell wall genes in these species and know that the C. tropicalis genome sequence used in the current analysis is highly flawed. There is even a manuscript from several years ago that documents problems in the assembly (doi: 10.1534/g3.115.017566). There is a new PacBio sequence available that has considerably improved data for this group of genes, but still is not perfect. We designed primers and amplified the various coding regions to verify whether the IFF/HYR were correct in the draft genome sequences. For C. tropicalis, we know that 7 of the genes listed in this paper are broken (i.e. prematurely terminated) giving a false impression of their construction. The current study did not verify any gene sequences, so broken/incomplete genes are a stumbling block for developing conclusions.

Similarly, the recent work from Cormack's lab features a PacBio C. glabrata sequence (doi: 10.1111/mmi.14707). The paper details how the authors focused on accurate assembly of the types of genes studied here. Sequences from the current project should be compared to the PacBio assembly to determine if they provide the same results.

Another part of the study that deserves additional attention or perhaps altered presentation is the idea that the Iff/Hyr N-terminal domain binds ligands. The literature on the Iff/Hyr proteins is limited. In my opinion, though, the authors of this paper could more completely present the information that is known. The paper by Uppuluri et al. is cited (doi: 10.1371/journal.ppat.1007056), but I did not see any information about their data regarding interaction of C. albicans Hyr1 with bacterial proteins mentioned in the manuscript under review. It is formally possible that the N-terminal domain of Iff/Hyr proteins does not bind a ligand. The current manuscript includes a great deal of speculation on that point, suiting it better to a Hypothesis and Theory format rather than other types of publications.

Table 1 attempts to offer evidence that the Iff/Hyr N-terminal domain has adhesive function but falls short of convincing the reader. One of the example structural templates is a sugar pyrophosphorylase that seems irrelevant to the current discussion. In the column called "Function", the word adhesin is found several times, but no detail is presented. The only entry that offers an example ligand indicates that the domain binds cellulose which is not likely relevant for mammalian pathogenesis, the main focus of the work. Other functions listed include self-association and cell aggregation--using the N-terminal domain. It is formally possible that Iff/Hyr proteins drive aggregation using the N-terminal domain and beta-aggregation sequences in the repeated region. The authors should develop these ideas further. Discussion of adhesive/aggregative function related to the ALS family can be found in Hoyer and Cota, 2016 (doi: 10.3389/fmicb.2016.00280).

The incredibly large number of figures that focus on the repeated sequences in the genes does not appear to include mention of the idea that these regions are frequently highly glycosylated. Knowing how much carbohydrate is added to these sequences in the mature protein would also have bearing on whether the beta-aggregation potential is realized. The Iff/Hyr proteins could stick to other things based on ligand binding (adhesion), hydrophobicity, aggregative activity, etc. Not much is really known about protein function so the conclusions are only speculative. The authors are largely accurate in presenting their conclusions as speculative, but the conclusions are not developed fully and always land on the idea that the N-terminal domain has adhesive function when that aspect clearly is not known.

Another aspect of the analysis that is not mentioned is that several of the species discussed are diploid. What effect does ploidy have on the conclusions? Most draft genomes for diploid species are presented in a haploid display, so are not completely representative of the species. Additionally, some species such as C. parapsilosis are known to vary between strains in their composition of gene families, with varying numbers of loci in different isolates.

The manuscript concludes that having more genes is better, that the gene family represents diversification that must be driven by its importance to pathogenesis, without recognizing that some species evolve toward lower pathogenesis. This concept could be explored in the Discussion.

The Results and Discussion sections are largely redundant. The tone of the paper is conversational, making it easy to read, but there seems little left to say in the Discussion that has not already been mentioned as the background for the various types of analyses. The authors should revise the paper to eliminate discussions of published literature from the Results and expand the Discussion to include some of the themes that have not been mentioned yet.

My own experience makes me wonder if the authors found any examples of species that provide and exception to the idea that having more genes is better and positively associated with pathogenesis. The parallel between IFF/HYR and ALS genes is made many times in the manuscript. Spathaspora passalidarum, a species that is not pathogenic in humans, but clearly within the phylogenetic group examined here, has 29 loci with sequence similarity to ALS genes. How many IFF/HYR genes are in S. passalidarum?

There are several current taxonomies for the species in this region of the tree. The source of the names used in this paper could be specific more completely.

Another point that the authors do not mention is documented recombination between IFF and ALS genes (doi: 10.3389/fmicb.2019.00781) and the effect of that process on evolution among these gene families.

My reading of the work by Xu et al. 2021 (doi: 10.1111/mmi.14707) does not match the direction of its presentation in the current paper. Oh et al., 2021 (doi: 10.3389/fcimb.2021.794529) discussed that point recently, providing another point for the Discussion in the current paper.

I might have missed it, but I could not find what constitutes a BLAST-excluded sequence (Table S7). Additional explanation (or making the explanation easier to find) would help the reader. - Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation. - Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

Ideally, validation of all sequences would provide a stronger foundation for the work. However, that request is not realistic in terms of time or resources. - Are the data and the methods presented in such a way that they can be reproduced?

Yes, the data and methods are documented clearly and perhaps too thoroughly in many places. A considerable amount of confidence is placed in sequences that might not be accurate and tracking details down to the amino acid residue may not be reasonable in this context. A disclaimer might help--everyone probably already knows that genome sequences are not perfect but stating that the analysis is only as good as the genome sequence acknowledges that fact. - Are the experiments adequately replicated and statistical analysis adequate?

The idea of replicates does not really apply to this analysis. I think that the species sampled are reasonable to represent the region of the phylogenetic tree on which the analysis is focused. The authors clearly documented their computational methods in an admirable way.

Minor comments:

• Specific experimental issues that are easily addressable.

Figure 1 has elements that would make a nice graphical summary, but most of it should not be part of the final manuscript. For example, Panel A is repeated in Figure 2. It is not clear what Panel C means until the reader gets to Figure 2. Panel D is unnecessary. The image in Panel B is a good graphic. Endothelial adhesion is not mentioned, though. It is also debatable whether the proteins bind directly to plastic or to the body fluids that coat the plastic.

Compared to Figure 1, the information in Figure 3 is inconsistent. The "central domain" in Panel A is not central to anything as drawn, located at the end of the protein. The figure should be revised to be consistent with the majority of the authors' results. Structures in Panels C to E would benefit from the "through the spiral" view that is featured in Figure S9. What experimental technique was used to solve the structure in Panel E? Adding that information to the legend would be helpful to the reader. Also, the secondary structure colors seem to be reversed between the legend and domain structure. Adding the coordinates of the domains shown would help the reader to understand their location in the mature protein.

Are the low-complexity repeats mentioned in the Figure 4 legend present anywhere else in the C. auris genome or elsewhere among the species used in this study? The answer to that question may also provide evolutionary clues.

Figure S1 legend. How was the distance to C. glabrata measured to call it equal?

Figure S4 could be presented better. Both diagonals have the same information. One could be emptied or could alternatively present nucleotide identity.

Italicize the species names in Panel C of Figure S8.

Lines 256-257: The paper selectively samples the Iff/Hyr family and does not examine the "entire" family. Please revise. - Are prior studies referenced appropriately?

I was disappointed to see that the paper does not reference my laboratory's work at all. When ALS genes are featured so strongly in a report, it seems reasonable to include something we have done over 30+ years. Our most-recent ALS paper (Oh et al., 2021 doi: 10.3389/fcimb.2021.794529) would be a reasonable source for defending the gene numbers used in Figure 2A. Other examples of our work that directly relate to concepts in this paper were mentioned above. - Are the text and figures clear and accurate?

Suggestions for improvement are incorporated into the comments above. - Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

Please present Methods and Results in the past tense. I still make the same mistake when I try to get my ideas on the page but proofread one more time and ensure the verb tenses are accurate.

Significance

• Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.

The paper reads as if it is presenting preliminary data for a grant proposal. Perhaps Prof. He's lab wants to seek functional evidence for the role of the Iff/Hyr proteins. The current paper provides an exhaustive background for such a pursuit. As presented, there is little functional data for these proteins, genome sequences are not 100% accurate, but the trends noted are defendable. - Place the work in the context of the existing literature (provide references, where appropriate).

The ideas presented here are similar to those pioneered in the Butler et al. Nature paper in 2009 (doi: 10.1038/nature08064). We now have the benefit of more genome sequences so the analysis can encompass more species. C. auris adds a newer focus on part of the phylogenetic tree that was not previously emphasized. The idea of "more is better" is very simplistic, though. Parallel work for the ALS family shows complexity in gene expression levels, suggesting that some adhesins are poised to make a large contribution while others are likely to have a scant presence on the cell surface. Those concepts are not really explored in the current paper, either. See Hoyer and Cota 2016 (doi: 10.3389/fmicb.2016.00280); Oh et al. (doi: 10.3389/fmicb.2020.594531). - State what audience might be interested in and influenced by the reported findings.

Potential readers would come from the fields of fungal adhesion and pathogenesis, as well as evolutionary biology. - Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

I discovered and named the ALS gene family in C. albicans and have spent 30+ years characterizing it. Most recently, my lab has focused on providing an accurate gene census and validated gene sequences for the cell wall "adhesinome" in the pathogenic Candida species. Some families are expanded and some are not. Some proteins appear only in a few species and demonstrate key roles in host-fungus interactions. There are many nuances to interpretation of what these fungi are doing from the standpoint of cell-surface adhesins and we look forward to exploring these ideas across many genomes, using validated gene sequences. We have a tremendous dataset that might make good fuel for a collaboration with Prof. He, given his enthusiasm for this area of study, as well as his outstanding expertise and perspectives on evolutionary analyses.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary

Gene duplication and divergence of adhesin proteins are hypothesized to be linked with the emergence of pathogenic yeasts during evolution; however, evidence supporting this hypothesis is limited. Smoak et al. study the evolutionary history of Hil genes and show that expansion of this gene family is restricted to C. auris and other pathogenic yeasts. They identified eight paralogous Hil proteins in C. auris. All these proteins share characteristic domains of adhesin, and the structural prediction supports that their tertiary structures are adhesin-like. Evolutionary analysis of protein domains finds weak evidence of positive selection in the ligand-binding domain, and the central domain showed rapid changes in repeat copy number. However, performed tests cannot unambiguously distinguish between positive selection and relaxed selection of paralogs after gene duplication. Some alternative tests are suggested that may be able to provide more unambiguous evidence. Together with these additional tests, the detailed phylogenetic analyses of Hil genes in C. auris might be able to better support the hypothesis that the expansion and diversification of adhesin proteins could contribute to the evolution of pathogenicity in yeasts.

Major Comments

The authors present extensive analyses on the evolution of Hil genes in C. auris. There is significant merit in these analyses. However, the analyses conducted so far are incomplete, lacking proper consideration of other confounding factors. Detailed explanations of our major comments are listed below.

1. First, the authors restricted genes in the Hil family to those only containing the Hyphal_reg_CWP domain. Yet, previous work included genes containing the ligand-binding domain or the repeat domain as Hil genes. More justification is needed whether the author's choice represents the natural evolutionary history of Hil genes appropriately. For instance, are the genes only containing the ligand-binding domain monophyletic or polyphyletic? We recommend including the phylogeny of all the Hil candidate genes, to discern whether evolutionary histories of the repeat domain and ligand-binding domain are congruent. Authors can use this phylogeny as justification to focus only on the ligand-binding domain containing genes.
2. In the analysis of positive selection, the authors do not adequately control for the effect of recombination on the evolutionary histories of protein sequences, especially given that Hil genes are rich in repetitive sequences. To account for recombination, GARD, an algorithm detecting recombination, should be performed to detect any recombination breakpoints within a protein domain. If recombination did occur within a protein domain, the authors should treat the unrecombined part as a single unit and use the phylogenetic information of this part to proceed with PAML analysis, instead of using the phylogeny of the entire protein domain. The authors should consider doing GARD analysis for the ligand-binding and repeat domains. For the repeat domain, low BS values in Fig. 5C indicate recombination between repeat units. Thus, the authors should analyze each repeat unit with GARD and re-analyze dN/dS.
3. The authors concluded positive selection in the ligand-binding domain based on the branch-wise model of PAML. Yet, w values were not higher than one, and it's unclear whether the difference in selective pressures the authors claimed here is biologically significant. Overall, what the authors present so far seems to be weak evidence of positive selection but is much more consistent with variation in the degree of purifying selection or evolutionary constraint. Using the site-wise model (m7 vs. m8) in PAML would allow the authors to detect which residues of the ligand-binding domain underwent recurrent positive selection. Combining the evolutionary information of protein residues and the predicted 3D structure will provide molecular insights into the biological impact of rapidly evolving residues. This would be a significant addition and raise the significance of the study, besides providing potentially stronger evidence of positive selection.

Minor Comments

1. In Fig 1c, the figure legend should include more specific details: which adhesin proteins are shown here? Please specify species names on the species tree
2. In Fig 3E, secondary structures are labeled with the wrong colors. Sheet: purple, helix: yellow
3. What's the ligand-binding activity of the b-solenoid fold? How structurally similar are C. auris PF 11765 domains compared to C. glabrata Awp domains? This information will support the role of adhesin for the ligand-binding domain of Hil genes.
4. In lines 248-249, the authors should also consider the influence of evolutionary history. For instance, repeats within the same Hil protein appeared later in evolution, compared to Hil gene duplication, and therefore these repeats experience less time for sequence divergence.
5. Although the bioinformatic evidence of C. auris Hil genes acting as adhesins is strong, it is still worthwhile to discuss the experiments of confirming the function of adhesins.

Significance

Overall, this study is interesting to investigate the evolutionary history of a crucial virulent gene in C. auris. Such evolutionary understanding will help us identify critical molecular changes associated with the pathogenicity of an organism during evolution, providing insights into the emergence of pathogens and novel strategies to cure fungal infections. The research question is important; however, the current analyses on the positive selection are incomplete, so the conclusion is modest so far. We recommend that the authors re-do the PAML analysis with the above considerations. This work will bring more significance to the mycology field if the functional impact of rapid evolution in protein domains can be supported or inferred.

This manuscript is well-written, and the authors also did a great job specifying all the necessary details in the M&M.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The manuscript by Smoak et al., provides an analysis of the Hyr/Iff-like (Hil) genes in Candida species with a strong focus on C. auris. The authors demonstrate a repeated expansion of these genes in unique lineages of fungal species, many of which are associated with stronger clinical disease. There is evidence of selection operating on the gene family in the primary domain used for identification. These genes include a repeat just downstream of that core domain that changes frequently in copy number and composition. The location of these genes tends to cluster at chromosome ends, which may explain some aspects of their expansion. The study is entirely in silico in nature and does not include experimental data.

Major points

Altogether, many of the general findings could be convincing but there are some aspects of the analysis that need further explanation to ensure they were performed correctly.

To start, a single Hil protein from C. auris was used as bait in the query to find all Hil proteins in yeast pathogens. Would you get the same outcome if you started with a different Hil protein? What is the basis for using Hil1 as the starting point? It also doesn't make sense to me to remove species just because there are already related species in the list. This may exclude certain evolutionary trends. Furthermore, it would be helpful to know how using domain presence and the conservation of position changes the abundance of the gene family across species? (beginning of results).

A major challenge in the analysis like this one is in dealing with repetitive sequences present in amplified gene families.

• For example, testing modes of selection on non-conserved sites is fraught. It's not clear if all sites used for these tests are positionally conserved and this should be clarified. Alignments at repeat edges will need to maintain this conservation and relatively good alignments as stated in lines 241-242 are concerning that this includes sequence that does not retain this structure necessary for making predictions of selection.
• It's also unclear to me why Figure S12 is here. The parameters for this analysis should be tested ahead of building models so only one set of parameters should be necessary to run the test. The evolutionary tests within single genes and across strains is really nice!
• A major challenge for expanded gene families is rooting based on the inability to identify a strong similarity match for the full length sequence. The full alignment mentioned would certainly include significant gaps. If those gaps are removed and conserved sites only are used, does it produce the same tree? Inclusion of unalignable sequences would be expected to significantly alter the outcomes of those analysis and may produce some spurious relationships in reconciling with the species trees.
• Whether or not there are similar issues in the alignment of PF11765 need to be addressed as well. There's nothing in the methods that clarifies site selection. Figure 1A: the placement of evolved pathogenesis is a little arbitrary. It's just as feasible that a single event increased pathogenesis in the LCA of C. albicans and C. parapsilosis that was subsequently lost in L. elongisporus. These should be justified or I'd suggest removing. The assignment of Candida species here also seems incomplete. The Butler paper notes both D. hansineii and C. lusitaniae as Candida species whereas they are excluded here. It is tricky to include scaffolds in analysis of chromosomal location of the HIL genes. The break in the scaffold may be due to the assc repeats of these proteins alone or other, nearby repeats. Any statistics would be best done to include only known chromosomes or those that are strongly inferred by Munoz, 2021. This will change the display of Figure 7, but is unlikely to change the take home message.

Minor points

Line 18: "spp." Should be "spps."

Line 41: I might rephrase this as "how pathogenesis arose in yeast..."

I might use a yeast-centric example around line 40 for duplication and divergence. This could include genes for metabolism of different carbon sources in S. cerevisiae.

The Butler paper referenced on line 51 compared seven Candida species and 9 Saccharomyces species The autors state no other evolutionary analysis of adhesins has been performed but do not acknowledge this study: https://academic.oup.com/mbe/article/28/11/3127/1047032

The first paragraph of the Results could be condensed

How was the species tree in Figure 1A obtained?

Figure 2: In panel A, "DH" and "SS" are not defined. I'd be careful with use of "non-albicans Candida" in Figure 2B. This usually includes C. tropicalis and C. dubliniensis and may confuse the reader. How was the binding domain defined to extract those sequences are produce a phylogeny? In building a ML model, how were parameters chosen?

Figure 3C/D could be just one panel.

Can you relate more the fungal hit to the Hil proteins conveyed in lines 152-154?

Line 168: Should read "Hence, ..." Label proteins along the top of Figure 4 too.

Figure 5: for tests of selection, were sites conserved across the group? What does the black number at each node indicate? Dn and Ds are given as decimals. This is based on what attribute? For panel B, it is unclear what each tip denotes i.e., Hil1_tr6. Hil1 is the gene but what is "tr6"?

It's unclear why comparison of the PF11765 domain includes the MRD proteins when those aren't included in the comparison to the repeats alone. Could that skew the comparison due to unequal sample numbers or changed variation frequencies in MDR relative to the other two groups?

Table 2 doesn't add much. This section could probably be reduced to a few sentences since it's highly speculative (intraspecies variation).

Table 3 is not needed.

Figure 6 - color coding in 6A needs to be explained. I'm assuming this is a taxonomical coding.

Figure 1B is unnecessary. A Model of the protein indicating domains is sufficient here. Figure 1C needs labels for all termini, not just the pathogenic red branches. The figure doesn't provide clear association between adhesin families and the associated species. This could be omitted, especially since Flo is often associated with Saccharomyces species. Figure 1D is unnecessary.

Significance

The work here is sorely needed in expanded gene families and in fungi specifically. No analysis at this level has been performed, to the best of my knowledge, in any fungal associated gene family and certainly not in relationship to pathogenic potential. The authors do a good job in citing the foundational literature upon which their study builds in most cases (one exception is noted above). It would be of general interest to those interested in the evolution of virulence, but the analysis is tricky. This is the biggest drawback I currently have as some of the information to assess the results is missing. Expertise: gene families, evolution dynamics, human fungal pathogens

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We would like to thank the Reviewers for their valuable comments and constructive suggestions concerning our manuscript entitled " Drosophila pVALIUM10 TRiP RNAi lines cause undesired silencing of Gateway-based transgenes" (RC-2022-01629).

Please find below our responses to the Reviewers' questions and comments. We have revised the Manuscript following the Reviewers' suggestions. The changes in the Manuscript are indicated in blue.

Reviewer #1 (Evidence, reproducibility and clarity (Required)): ____ This manuscript by Uhlirova and colleagues identified an unwanted off-target effect in the pVALIUM10 TRiP RNAi lines that are commonly used in the fly community. The pVALIUM10 lines use long double-stranded hairpins and are useful vectors for somatic gene knock-down, hence they are widely used.

Here the authors find that any pVALIUM10 TRiP RNAi line can create the silencing of any transgenes that were cloned with the commonly used Gateway system. this is caused by targeting attB1 and attB2 sequences, which are also present in other Drosophila stocks including the transgenic flyORF collection. Hence, this is an important and useful information for the fly community that should be published quickly. All experiments are well documented and well controlled. I only have a few minor comments.

1. I recommend to mention the number of 1800 pVALIUM10 lines in Bloomington in the abstract rather than 11% to make clear that this is an important number of lines. (1800 of 13,698 lines in Bloomiongton are 13 and not 11 per cent?)

We now include the absolute number of pVALIUM10 lines in the manuscript abstract. The percentages have been corrected. Furthermore, we updated/corrected the total number of RNAi lines available from various stock centers in the Discussion, L153-L156.

The status on 23.10.2022

VDRC - 23,411 in total (12,934 GD lines; 9,674 KK lines; 803 shRNA lines)

Bloomington - 13,410 TRiP lines based on pVALIUM vectors (13,674 in total, including 264 non-pVALIUM, and 48 non-fly genes targeting lines)

NIG - 12,365 in total (5,676 TRiP lines; 7,923 NIG RNAi lines)

The authors may consider to call the 'unspecific' silencing effect an 'off-target' effect compared to intended 'on-target'. Such a nomenclature would be more consensus.

We changed the wording in the manuscript as suggested by the reviewer.

Ideally, all the imaging results in Figure 2 and 3 would be quantified. The simple 'V10' label in the Figure 3L and 3M is not the most intuitive, at least it took me a while to figure out what the authors compare.

The labeling in the charts has been changed. We now provide quantifications for the data shown in Figure 2 and 3.

Does the silencing also affect attR sequences? These are present after cassette exchange in many transgenes, most of the time not in the mRNA though, so it might not be so relevant.

A 22 nucleotide stretch of the attB2 site indeed shows a 100% match to the attL2 site. See the example alignment below (availbale in word/PDF version of the Letter). While we did not assess this possibility experimentally, attL sites would likely be susceptible to the same undesirable off-target silencing effects if present in the nascent or mature transcript.

Reviewer #1 (Significance (Required)): This is an important and useful information for the fly community that should be published quickly.

Reviewer #2 (Evidence, reproducibility and clarity (Required)): ____ Stankovic, Csordas, and Uhlirova show that a specific subset of the TRiP RNAi lines available, namely the pVALIUM10 subset, can cause a knockdown of certain co-expressed transgenes that contain attB1 and attB2 sites. The authors demonstrate that while pVALIUM20 or Vienna KK lines for BuGZ or myc RNAi do not affect RNase H1:GFP expression, pVALIUM10 RNAi lines against BuGZ or myc significantly decrease expression of the RNAseH1:GFP transgene. The authors propose that, due to how these RNAi lines were constructed, the siRNA products could be targeting to attB1 and attB2 sites in transgenes that were made using similar methodology. To support this idea, they ubiquitously express mCherry transgenes encoding mRNAs either containing or lacking attB sites. They find that the knockdown of mCherry seen with several different pVALIUM10 RNAi lines is observed with the reporter mRNA containing attB sites, but is suppressed when the attB sites are removed from mCherry mRNA. They also find that the pVALIUM10 RNAi lines reduce the expression of the FlyORF transgene SmD3:HA.

The paper is very clearly written and the data presented is convincing.

Minor suggestions:

1. Figure 3 L+M The labels for the ubi-mcherry and ubiΔattb-mcherry are switched in these graphs (i.e. ubiΔattb-mcherry should be the one with a higher intensity in the pouch compared to the notum).

Figure 3M the labels don't match the RNAi lines used in H-K.

We corrected the labelling in the charts.

Figure 2 and 3. For the images of the transgenes, it seems as if the BuGZ RNAi line has a more drastic effect on RNaseH1 than mCherry, and vice versa for the myc RNAi lines. Did the authors notice a pattern with the decreased expression. Do some of the RNAi lines have a more consistent/severe impact, or might different transgenes be impacted to different extents?

Throughout the study and multiple experimental trials, we did not observe that the BuGZRNAi and mycRNAi silencing efficiency would depend on whether the monitored reporter was RNase H1::GFP or mCherry. What has been reproducible is the differential impact of the three tested mycRNAi lines on ubi-RNaseH1::GFP transgene. While pVALIUM10-based mycRNAi[TRiP.JF01761] reduces RNaseH1::GFP signal Valium20 mycRNAi[TRiP.HMS01538] enhances it and GD mycRNAi[GD2948] has no effect, although the number of replicates for the latter is lower compared to the other tested lines. Why Valium20 mycRNAi[TRiP.HMS01538] increases RNaseH1::GFP signal remains unclear for now.

We would like to refrain from directly quantitatively comparing the effects of phenotypically different RNAi lines on differently tagged mRNAs/proteins. As the RNAseH1::GFP fusion protein is nuclear while the mCherry is cytoplasmic, their distinct subcellular localization and/or turnover rate may give a different overall impression on the change in fluorescence intensity (Boisvert et al, 2012; Mathieson et al, 2018). Another confounding factor is the described roles of Drosophila Myc in regulating transcription, translation, and cell growth (Gallant, 2007).

Line 150 unnecessary comma after Both Line 131 knockdown should be knocked down Line 133 should be "using an additional" Figure legend 1 wing disc should be at least written out when the abbreviation (WD) is first used.

We thank the reviewer for pointing these out, the relevant corrections were performed.

Reviewer #2 (Significance (Required)):

Overall, this manuscript is an informative reminder that RNAi lines can have weaknesses that have not yet been considered, and we appreciate the authors work to inform the fly community about this specific issue. These insights are crucial for fly labs to consider when planning experiments that will use the pVALIUM10 RNAi lines in combination with other transgenesis modalities. The manuscript also provides a cautionary note for the usage of similar resources in other model organisms.

Reviewer #3 (Evidence, reproducibility and clarity (Required)): Summary: In their manuscript "Drosophila pVALIUM10 TRiP RNAi lines cause undesired silencing of Gateway-base transgenes", Stankovic et al. describe off-target silencing of transgenes expressed from Gateway systems when expressed in transgenic RNAi drosophila lines from the VALIUM10 collection. Using fluorescence microscopy and immunostaining, the authors show that this unintended silencing is specific to VALIUM20 lines and is not observed with VALIUM20, KK or GD lines that also allow gene-specific RNAi silencing. This pleiotropic silencing effect was observed in 10 different VALIUM20 lines and affected Gateway-based transgene expressed from an ubiquitous promoter (poly-ubiquitin, ubi) or from Gal4/UAS systems. Finally, the authors identify the molecular basis of VALIUM20 pleiotropic silencing on Gateway transgenes as being due to the presence of short sequences used for PhiC31-based recombination in the Gateway and the VALIUM systems, and could lead to the production of siRNAs against PhiC31 recombination sites in VALIUM10 lines. Using Gateway transgenes lacking the recombination sites (attB1 and attB2), the authors could abrogate silencing of the transgene in VALIUM10 lines, confirming the recombination as shared targets between the Gateway and the VALIUM systems.

Major comments: - The study is well designed and the key conclusions are convincing. - However, the authors provide only fluorescence microscopy data to show decreased transgene expression. To confirm pleiotropic RNAi effect on Gateway transgenes in VALIUM10, the authors should assess silencing with another technique. For instance, expression levels of proteins from Gateway transgenes could be measured by Western blot (e.g.: by assessing protein levels of GFP or other tags present in the Gateway transgenes).

In the manuscript, we present microscopy data as this is the typical use case for fluorescent reporters. The strength of the microscopy, in contrast to Western Blot or RT-qPCR approach, is that it allows us to directly compare the impact of RNAi silencing on cells that express the dsRNA transgene (cell-autonomous) to surrounding neighbor cells. The fluorescent imaging of WDs where all cells express the reporter construct, but only a subset of cells trigger RNAi-mediated silencing, provides spatial resolution and means for normalization while minimizing artifacts that can arise during tissue processing for WB and RT-qPCR. We provide data on GFP and HA-tagged transgenes, respectively, and untagged mCherry expressed from Gateway vectors under ubiquitin or UAS regulatory sequences with the explicit reason to show that the silencing effect is independent of the type of the protein tag or the expression regulator sequence.

In addition, the claim on line 141,"These results strongly indicate that the dsRNA hairpin produced from pVALIUM10 RNAi vectors generates attB1- and attB2-siRNAs" , should be modified. The authors only present fluorescence microscopy data to show decreased transgene expression and do not actually provide data on siRNA expression in the pVALUM20 lines. Therefore, with the current data, the authors should only say that their results suggest that the dsRNA hairpin produced from pVALIUM10 RNAi vectors generates attB1- and attB2-siRNAs.

In order to substantiate their claim about pleiotropic RNAi effects from VALIUM lines on Gateway transgenes due to the production of attB1- and attB2 -siRNAs, the authors should perform an experiment to show attB1- and attB2 -siRNAs production in VALIUM10 lines and not in VALIUM20, KK or GD lines. Deep-sequencing analysis of siRNA (i.e.: miRNA-seq) from tissue expressing the corresponding RNAi transgenes would be an excellent approach to assess siRNA production in multiple samples at once. Alternatively, the authors could search published miRNA-seq datasets from VALIUM10 and other RNAi lines to assess the presence of attB1- and attB2 -siRNAs only in VALIUM10 lines. This would be free and require only a few days of data mining and analysis, if such datasets exist already. Another cheaper and faster approach (if lacking easy access to sequencing platform or bioinformatics capability) would be to perform small RNA northern blots analysis from fly tissues expressing VALIUM10 vs VALIUM20 (or KK or GD lines) and should take only a few days to do as described in doi: 10.1038/nprot.2008.67.

If such experiments or analyses cannot be performed, then the authors can only conclude that their data suggest that the unintended silencing of Gateway transgenes in VALIUM10 is likely due to the production attB1- and attB2 -siRNAs production.

We thank the reviewer for the valuable suggestions on experimental approcahes to identify the exact interfering RNAs produced by the VALIUM10-based RNAi constructs, which can be useful for controlling the specificity of knockdown of transgenes in studies using the resources mentioned in this report.

We believe the fluorescence micrographs and quantifications demonstrate the off-target silencing effects of pVALIUM10-based RNAi lines on transgenic reporters generated using the Gateway LR cloning approach. Furthermore, we provide genetic evidence that removing the attB1 and attB2 sites from the reporter construct, which is otherwise identical to the original transgene (same promoter, same position of insertion, same genetic background), is sufficient to abolish the off-target effect. We would argue that the functional genetic experiments we performed with the original and mutated reporters represent the strongest possible evidence to confirm that silencing is taking effect via the attB sites.

As we do not attempt to detect siRNA complementary to attB1/attB2 sites directly, we have changed the statements in question as per the recommendation of the reviewer.

• The current data and methods are adequately detailed and presented, and the statistical analysis adequate.

Minor comments:

• The current manuscript does not have specific experimental issues.
• Prior studies are referenced appropriately
• Overall the text and figures are clear and accurate except for the following issues with Figure 3 and its legends On lines 396, 397, 399 and 403, the authors refer to "wild-type" ubi-mCherry. This transgene directs the ubiquitous expression of an heterologous reporter gene and thus can not as "wild type". It could instead be referred to as the "original" or "unmodified" transgene.

We removed "wild-type" from the text.

Fig.3 L: the x-axis labels are wrong. Decrease in the mCherry intensity ratio is observed with the ubi-mCherry construct and not in the ubi∆attB-mCherry, where the attB sequences thought to be targeted by the pVALIUM10 have been deleted.

More space should be added between the first row of images (B-G), the second (H-L) and also the third (M-P) to avoid confusion between the labeling of the figures. Finally, to help contextualize their findings and gauging the extent of the risk of using VALIUM10 lines in RNAi screen where a Gateway transgene is involved, the authors could provide information on the overlap between the VALIUM10 collection and VALIUM20, GD and KK collections. Knowing how many genes are uniquely targeted by VALIUM10, could be helpful.

We corrected the Figure panels according Reviewer 1 and 3’s observation.

Of the TRiP pVALIUM-based RNAi stocks currently available in BDSC, 686 genes are targeted exclusively by pVALIUM10 RNAi lines. Considering KK, GD and shRNA transgenic lines from VDRC and NIG RNAi collection, 17 genes remain unique targets for pVALIUM10 lines. The graphical overview of the availbale lines is availbale in the word/PDF file of the Response to Reviewers Letter.

Reviewer #3 (Significance (Required)):

• The manuscript "Drosophila pVALIUM10 TRiP RNAi lines cause undesired silencing of Gateway-base transgenes" by Stankovic et al. is a technical study that sheds light on potential limitations of using common RNAi drosophila lines, namely the VALIUM10 collection.
• The study provides information about very specific genetic screens conditions in Drosophila, that are likely to be rare. A rapid Pubmed search with the following terms: "drosophila TRiP screen" returns only 11 citations, while a similar search with "drosophila CRISPR screen" returns 99 citations. This suggests that in vivo RNAi screen in Drosophila using TRiP RNAi collections might not be as common or powerful as CRISPR-based screens.
• The reported findings might be of interest mostly to a small group of scientists working with Drosophila melanogaster that specifically rely on VALIUM10 lines to perform in vivo RNAi screen in combination with Gateway transgene expression. This very specific combination of parameters is rare, since other RNAi fly stock collections exist (e.g.: VALIUM20, 21, KK, GD...). Furthermore, the advent of CRISPR tools that allows tissue-specific gene knock-out has led to the rapid expansion of CRISPR fly stock collections (https://doi.org/10.7554/eLife.53865). Regardless of the limited scope of the study, this kind information is still valuable, albeit to a very limited audience.
• My relevant fields of expertise for this study are : insect RNAi, RNAi of RNAi screens and drosophila genetics.

We would like to raise some points concerning the above comments.

While TRiP-screen may not be an often-used keyword combination, the use of the TRiP lines is, in fact, ubiquitous in the Drosophila community. The tissue-specific RNA interference is still commonly utilized as a rapid, first-generation screening method that can be performed in a tissue-specific manner, representing one of the key advantages of the Drosophila model. To illustrate, since the submission of our manuscript a new study published by Rylee and co-workers investigated Drosophila pseudopupil formation by screening 3971 TRiP RNAi lines (Rylee et al, 2022). In contrast, genetic screens relying on mutant alleles usually require at least one additional cross, effectively doubling the time of the experiment. In addition, tissue-specific or temporarily restricted knockdown is sometimes required in screens, as full-body loss of function is often lethal or has developmental phenotypes incompatible with assessing gene function later in life.

The use of tissue-specifically driven Cas9 with integrated gRNA-expressing vectors is indeed becoming more common. However, this technique, much like RNA interference, is not without flaws. First, this produces knockout instead of knockdown, which means it has to be induced early in order for the resulting mutation to take effect. Otherwise, the remaining mRNA/protein may prevent the development of a phenotype. Second, the Cas9 must be titrated as high Cas9 levels have adverse phenotypes (Huynh et al, 2018; Meltzer et al, 2019; Poe et al, 2019; Port et al, 2014). Third, in our personal experience, as well as literature reports (Mehravar et al, 2019; Port & Boutros, 2022), indicate that the resulting phenotype can produce mosaics in the tissue.

Although the combination of Gateway-based reporters with TRiP-RNAi lines may seem like a fringe case, there are popular reporters that could be screening targets. Potentially the most well-known is the live cell cycle indicator fly-FUCCI system (Zielke et al, 2014), which allows the analysis of the cell cycle in real-time thanks to the expression of two fluorescently tagged degrons. As FUCCI transgenes were constructed with Gateway recombination, they represent targets of the pVALIUM10 TRiP lines. We now include the fly-FUCCI system as an example in addition to 3xHA-tagged FlyORF collection in the Discussion.

REFERENCES

Boisvert FM, Ahmad Y, Gierlinski M, Charriere F, Lamont D, Scott M, Barton G, Lamond AI (2012) A quantitative spatial proteomics analysis of proteome turnover in human cells. Mol Cell Proteomics 11: M111 011429

Gallant P (2007) Control of transcription by Pontin and Reptin. Trends Cell Biol 17: 187-192

Huynh N, Zeng J, Liu W, King-Jones K (2018) A Drosophila CRISPR/Cas9 Toolkit for Conditionally Manipulating Gene Expression in the Prothoracic Gland as a Test Case for Polytene Tissues. G3 (Bethesda) 8: 3593-3605

Mathieson T, Franken H, Kosinski J, Kurzawa N, Zinn N, Sweetman G, Poeckel D, Ratnu VS, Schramm M, Becher I et al (2018) Systematic analysis of protein turnover in primary cells. Nature Communications 9: 689

Mehravar M, Shirazi A, Nazari M, Banan M (2019) Mosaicism in CRISPR/Cas9-mediated genome editing. Developmental Biology 445: 156-162

Meltzer H, Marom E, Alyagor I, Mayseless O, Berkun V, Segal-Gilboa N, Unger T, Luginbuhl D, Schuldiner O (2019) Tissue-specific (ts)CRISPR as an efficient strategy for in vivo screening in Drosophila. Nature Communications 10: 2113

Poe AR, Wang B, Sapar ML, Ji H, Li K, Onabajo T, Fazliyeva R, Gibbs M, Qiu Y, Hu Y et al (2019) Robust CRISPR/Cas9-Mediated Tissue-Specific Mutagenesis Reveals Gene Redundancy and Perdurance in Drosophila. Genetics 211: 459-472

Port F, Boutros M (2022) Tissue-Specific CRISPR-Cas9 Screening in Drosophila. In: Drosophila: Methods and Protocols, Dahmann C. (ed.) pp. 157-176. Springer US: New York, NY

Port F, Chen HM, Lee T, Bullock SL (2014) Optimized CRISPR/Cas tools for efficient germline and somatic genome engineering in Drosophila. Proc Natl Acad Sci U S A 111: E2967-2976

Rylee J, Mahato S, Aldrich J, Bergh E, Sizemore B, Feder LE, Grega S, Helms K, Maar M, Britt SG et al (2022) A TRiP RNAi screen to identify molecules necessary for Drosophila photoreceptor differentiation. G3 Genes|Genomes|Genetics: jkac257

Zielke N, Korzelius J, van Straaten M, Bender K, Schuhknecht GFP, Dutta D, Xiang J, Edgar BA (2014) Fly-FUCCI: A versatile tool for studying cell proliferation in complex tissues. Cell Rep 7: 588-598