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 manuscript the authors show a huge ambition to catalog biological functions of Plasmodium kinesins. This was done by generating transgenic cell lines where kinesins were deleted and/or tagged with GFP that served as a tool to gather as much biological information on each kinesin isoform. On one side I find this manuscript highly impressive in terms of the amounts of data and information. In particular, the cell biology and microscopy results are of high quality and certainly provide useful information to the research community. I am fairly convinced that most results genuinely represents the individual biological aspects of the kinesins in the best possible way. Unfortunately, I have major reservations about the presentation of these results in the compiled manuscript. In my view the authors were overambitious about the volume and diversity of data that they wished to present, which opened a lot of questions about the depth and quality of each of the experimental effort. There is 10 figures which is highly nonstandard for a scientific publication to start with and yet there is, in my view, major gaps in some results descriptions, data presentations e.t.c. Perhaps, because of this huge ambition the data are presented in a highly superficial manner often lacking negative and positive controls. Unfortunately that creates many doubts about the overall quality of the results and as such the interpretations. In my view the authors might be well advised to separate this large body of work into several publications each focusing on more tangible biological problem in the more in-depth manner. This would give the reader (me) better confidence about the validity of the statements made in this manuscript.

I can give few examples of such discrepancies but cannot account for all.

1.The authors created GFP-tagged transgenic cell lines for each of the 9 kinesins and generated life cell images for each of the line across multiple stages of the entire plasmodium life cycle. This is an impressive amount of work and data. It is certainly useful to see that in life cell imaging the different kinesins isoforms can be detected in different sets of developmental stages some diffused in the cytoplasm and some associated with the nucleus. Even though these results are impressive, there are based solely on life cell imaging that rely on a certain level of detection limit and GFP visibility. One can imagine that a kinesin may still be expressed in a developmental stage and not detected by life cell imaging. I believe that some other detection methods such a western blot, immunoprecipitation e.t.c. should be provided to truly demonstrate that an individual isoform of a kinesin IS of ISNOT expressed. Without that the Figure 1B is overstated. Moreover, the authors claim that the punctuate signal in the nucleus corresponds to spindle. I do not see any supporting evidence for this in this figure.

2.For the analysis of kinesin 5 and 8x the authors note two types of experiments. First they created a "cross" between the two cells lines. Second, the authors carry out ChIP-Seq to show that the proteins localize to the centromere. This could be an impressive result unfortunately there is very little if any information about it. Genetic crosses in Plasmodium are not standard techniques that one can assume works all the time. I believe there should be more evidence that the presented images come from a true genetic cross. The least the authors could show that the florescence signal for both channels come from genuine integrations of the GFP proteins to their target kinesins by PCR or genome sequencing. Similarly for the chip-seq, there is a need to provide much detailed information about the entire results with a particular clarity about the position of the peaks in respect to projected centromeres. In addition the ChIP-Seq analysis should be supported by data along with positive and negative controls to truly show the kinesins associations with the centromeres.

3.In the middle part, the author present rater impressive analyses of several kinesin deletion trains and their effect on the development of the mosquito stages. In particular, they demonstrate the effect of kinesin 20 on ookinete development. Yet in the next paragraph they present RNA seq analysis of the kinesin 20 deletion on gametocyte induction, in which kinesis 20 should not have any effect; judging from the presented phenotypic assay. This experiment seems out of context as it is unclear why this assay was done and what is the outcome. The authors identified a small group of differentially expressed genes seemingly unrelated to neither kinesin function nor gametocyte induction. This experiment does not make sense to me in the context of the rest of the paper.

Significance

As mentioned above, these three examples represent some of the discrepancies not necessarily about the data quality and fidelity but rather a confusing character of the entire study. From this perspective I have two types of problems with this manuscript. First, while reading this manuscript, lacking key controls and detailed description of some of the analyses, made me loose interest as well as confidence in other parts of the studies which may or may not be solid. Second, I struggle to see the key purpose of the presentation. Instead the manuscript seems to be a compilation of very diverse data some of which are interesting but other out of context, confusing and not connected to the rest of the study.

Overall I wish to reiterate that I believe that there are a lot of very good experimental results in this study but unfortunately many of these get lost in the overall presentation that is often superficial or out of context. My general impression is that the authors are trying to show too much, "too fast" and as such many of the presented results remain questionable. The author are likely able to correct all these discrepancies but this might not be possible to do in ne 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 #2

Evidence, reproducibility and clarity

Summary: In this study, Zeeshan et al used live-cell imaging, ultrastructure expansion microscopy, and electron microscopy, gene deletion, genetic knockdown, RNA-seq, ChIP-seq analyses, and matrigel substrate to examine the subcellular localization and the function of Plasmodium kinesins throughout the P. berghei life cycle. They find that Kinesin-13 is the only kinesin essential for both asexual blood stages and sexual stages.

This manuscript represents a lot of work by the authors. The data appear rigorous and well-executed. The data are clearly presented and the writing is clear. I have only minor comments that may improve the reader's comprehension.

Major comments:

Figure 2C:

The ChIP-seq experiments examined the kinesin-5 and -8x binding site at the chromosome at 6 mpa. Did the authors do any tests at other time points post-activation?

Figure 4:

The authors conclude that kinesin-x3 and kinesin-x4 are non-essential for the P.berghei life cycle. Does deletion of kinesin-x4 affect the length of the flagella?

Oocyte size: To the non-specialist, it is difficult to reconcile the images in panel E with the conclusions in panel A. Based on the images, it looks like only knocking out of kinesin-8x seems to affect oocyst size. Can the authors clarify and provide graphs of the quantification of oocyte size?

Minor comments:

line 190: typo, kinesin-x4

Figure 3: what do the arrows mean?

Figure 4F:

1. typo, scale bar, um.

2. Does deletion of kinesin-5 show a significant difference?

Significance

The study provides comprehensive information on the diverse subcellular location and functions of P. berghei kinesins throughout the P. berghei life cycle. That is useful to exploit the therapeutic targets against malaria.

The main findings are that kinesin-13 genetic knockdown affected MT dynamics during spindle formation and axoneme assembly in male gametocytes and subpellicular MT organization in ookinetes. In addition, Kinesin-13 shows different binding to kinetochores during the gametogenesis and ookinete development, suggesting other proteins may regulate kinesin-13 binding to kinetochores at various stages. The underlying mechanism will help to better understand the role of kinesin-13 in the parasite life cycle.

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 study led by Dr. Zeeshan analyzes nine mouse Plasmodium parasites kinesin by determining their expression pattern and subcellular location in various stages of the parasites in the mammalian and mosquito host. The genetic and phenotypic analyses of all nine kinesins indicate that most are critical for parasite development in the mosquito host, except for Kinesin 13 being the only kinesin essential during the intraerythrocytic development cycle. The authors presented an in-depth analysis on kinesin 13 and 20, using an impressive pallet of molecular techniques such as promotor swapping, chromatin immunoprecipitation, and global transcriptomic analysis using RNAseq, as well as numerous microscopy techniques such as live fluorescence imaging, expansion microscopy, and electron microscopy. This comprehensive study provides an outstanding amount of data on Kinesins in Plasmodium parasites that would be best showcased with a rethinking of the manuscript structure and a more insightful discussion section that directed most of my comments in the review the manuscript. I believe no additional experiments are needed assuming that the authors will link Kinesin 13 and or 20 to the IMC formation in future work.

Major Comments:

•The current manuscript shows the " Location and function of Plasmodium kinesins" as the title suggests; however, I strongly recommend the authors consider alternative storytelling focusing on Kinesin 13 and 20. The author provides in-depth phenotypical analysis resulting in the most innovating and exciting data. In addition, the discussion section from lines 592 to 634 was fascinating compared to the following section (see details comments for Discussion section below).

•The following significant comments are related to figures where I believe a restructuration is most needed to bring clarity to the paper."

•Figure 1. I suggest the authors move Figure 4A to figure 1; Figure1C should move to supplementary information except for Kinesin 13 and 20 data to center the paper's focus on these two proteins. I would also present the kinesin data in the current Figure4A not by numeric order but by biological relevance. All the "normal" together and so on

•Figure 2: Kinesin 5 and 8X have the same results. I suggest the authors present only one in the same manuscript and place the other one in Supplementary information. I would recommend adding the little schematic used in Figure1C to help the reader quickly identify the parasite stages presented in the figures.

•Figure4: Panels B to E should be a supplementary information

•Figure 5: Panels H to J should be supplementary information, and I strongly recommend the authors to present data by stages; therefore, I would remove panels F and G and replace them with Figure 6A, the expansion microscopy represents the data in Figure 4B, C, D, and E beautifully.

•Figure 6B: It is challenging to identify the layout between WT and delta-kinesin 20. All annotations on the EM data cover the data itself. I recommend drawing a representative schematic to guide the reader for identification of ultrastructure.

•Figure 8: Panel C and D should be supplementary information and replaced by the accurate colocalization data of Kinesin 13 presented in Supp figure 5. In addition, comment line 442 is also actual for the ookinete. The true colocalization is with tubulin in male gamete and gametocytes in figure 5A/B.

•Figure 9: Panel F to J go to supplementary information and replace with the data in figure 10.

•Figure 10: Could be a great abstract figure in the current state. As a model figure, I would recommend incorporating more details

Minor Comments:

I will address my following minor comment by Line number rather than section:

Figure 1C: It is unclear if the black square is an actual picture or a black square. I would suggest the authors present the absence of data by a white square or a bar.

Line 96: " a final synchronized round of S-phase" The classical mitotic terminology is poorly used in the field of Plasmodium mitosis due to the absence of canonical cell cycle checkpoint. I would recommend the authors rephrase as " a final synchronized round of DNA replication."

Line 149-151: Could the authors indicate what stage of the life cycle the work was done?

Line 161: Missing space between the word "parasite and cell"

Line 163: " These findings will inform a strategy ..." Could the authors explain in greater detail how the study is informative for targeting MT motors for therapeutic. I would argue with the authors that it is an overstatement since the paper did not provide structural data on kinesin as a foundation for drug discovery.

Line 368: What was the reasoning for examining whether other kinesin genes' expression is misregulated in deltaKinesin 20?

Line 515: Could the authors define what is a nuclear pole?

Line: 576 - 579: The authors mention the absence of the IFT component for flagellum assembly due to the assembly of the axoneme in the cytoplasm. It is known that kinesin-2 is required for the anterograde transport in organism building cilia and flagella using IFT. In the current study, kinesin 2 is not part of the nine kinesins; therefore, it is unclear why the authors made these comments and did not reflect on them. I would suggest removing it or comment it.

Line 546-560: this entire section of discussion would be best in a review paper. It is a well-written summary of the current literature with no discussion related to the data on the present study; therefore, I suggest the authors remove it from the discussion.

Line 561 - 571: Great summary of the Kinesin-13 work without discussion.

Line 572: What do the authors mean by " these findings"?

Line 573 - 589: The authors miss the opportunity to elaborate on how the depletion of kinesin protein could impact the global transcriptome. Are we looking at downstream effects? I strongly recommend the authors resolve the lack of discussion related to the RNAseq data in the study.

Significance

This study is a tremendous amount of work done rigorously and will advance our knowledge in the biology of Plasmodium parasites. We are in urgent need to develop innovative ways to block the replication and transmission of Plasmodium spp. and it can happen only through advancing our knowledge in the basic biology of the parasite.

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-2021-01015

Corresponding author(s): Jordan, Raff

1. General Statements [optional]

We thank the reviewers for their thoughtful and constructive comments and have now revised our manuscript accordingly. We apologise that it has taken so long to send in these revisions, but this is in part because both first authors have now left the lab.

2. Point-by-point description of the revisions

Reviewer #1

This reviewer was generally supportive. They note that it is unfortunate that our data suggests the CP110/Cep97 complex does not play a major part in controlling daughter centriole growth—although we believe that this is an important negative result—but feel that other aspects of our data are interesting. They requested no further experiments, but did comment that it would be interesting to determine when g-tubulin is incorporated into growing centrioles. Unfortunately, we cannot test this as the centrioles in these embryos recruit large amounts of g-tubulin to their PCM, so we cannot specifically assay the small amount of protein in the centriolar fraction.

Reviewer #2

Major Points:

__Figure 1: The reviewer notes that Sas-4 and CP110 have antagonistic roles in promoting/repressing centriole growth and asks if Sas-4 is involved in promoting centriole elongation and whether it also oscillates. __It is unclear if Sas-4 directly promotes centriole elongation in flies. We have previously shown that centriolar Sas-4 levels do oscillate during S-phase, but with a timing that is distinct from CP110/Cep97 (Novak et al., Curr. Biol., 2014). These observations do not shed much light on the potential antagonistic relationship between CP110/Cep97 and Sas-4, so we do not comment on this here.

Figure S1B: The reviewer requests that we image the centrioles with greater laser intensity to test whether some residual CP110 or Cep97 protein can be recruited in the absence of the other protein. The quantification of this data suggests that some residual CP110 or Cep97 can still be recruited to centrioles in the absence of the other (Graphs, Figure S1B,C), so we do not think it necessary to repeat this experiment at higher laser intensity to further test this point. We now state that the centriolar recruitment of one protein may not be completely dependent of the other (p6, para.2).

Figure 3: The reviewer questions whether the reduction in CP110/Cep97 levels at the mother centriole that we observe during S-phase could be due to photobleaching. This is an interesting point that we now analyse in more detail (p8, para.2). We do not think the decrease in mother centriolar CP110/Cep97 levels is due to photobleaching as our new analysis (which includes more data points during mitosis) strongly suggests that centriolar levels on the mother rise again at the start of the next cycle (New Figure 3C,D).

The reviewer asks whether the CP110/Cep97 oscillations occur at the tip of the growing centriole, and whether we can use super-resolution imaging to address this. A large body of evidence indicates that CP110/Cep97 are highly concentrated at centriole distal tips, and all our experiments suggest that it is this fraction that is oscillating. In Figure 3, for example, we use Airy-scan super-resolution imaging to follow the oscillation on Mother and Daughter centrioles in living embryos. Although the resolution in these experiments is not as high as we can achieve using 3D-SIM in fixed cells, it seems reasonable to assume that the dots of fluorescence we observe oscillating on these centrioles (Fig. 3) are the same fluorescent dots we observe localised at the distal tips of the mother and daughter using 3D-SIM in fixed cells (Fig. 1A).

The reviewer requests additional quantification of the western blots shown in Figure S1 that we use to judge relative expression levels. As we now describe in more detail in the M&M, these ECL blots are very sensitive, but highly non-linear, so we usually estimate relative expression levels by comparing serial dilutions of the different fractions (see, for example, Figure 1B, Franz et al., JCB, 2013). As we now clarify, the key point is not precisely by how much these proteins are over- or under-expressed, but that we observe a similar oscillatory behaviour when they are either over- or under-expressed.

__The reviewer points out that our statement that the CP110/Cep97 oscillation is entrained by the Cdk/Cyclin oscillator (CCO) is too strong as it is based only on a correlation. __We agree and apologise for this overstatement. To address this, we have now perturbed the CCO by halving the dose of Cyclin B (New Figure 5E—H). This extends S-phase length and we now show that the period of the CP110/Cep97 oscillation is also extended. This suggests that the CCO directly influences the period of the CP110/Cep97 oscillation.

The reviewer notes that our conclusion that the centriole cartwheels are longer or shorter when CP110 or Cep97 are absent or overexpressed, respectively, is based only on Sas-6-GFP fluorescence intensity. They ask if this fluorescence intensity perfectly reflects cartwheel length, and if we can confirm these conclusions using EM. Sas-6 is the main structural component of the cartwheel, so the amount of Sas-6 at the centriole should be proportional to cartwheel length, and we have published two papers that support this conclusion and that use the incorporation of Sas-6 as a proxy to measure cartwheel length (Aydogan et al., JCB, 2018; Aydogan et al., Cell, 2020). Importantly, our previous EM studies support our conclusions about the relationship between cartwheel length and CP110/Cep97 levels: the centrioles in wing-disc cells are slightly longer in the absence of CP110 and slightly shorter when CP110 is overexpressed (Franz et al., JCB, 2013). The new findings reported here provide a potential explanation for this EM data, which was puzzling at the time.

Minor Points:

Figure 1C: The reviewer noted that our schematic illustrations in this Figure could be misleading____. We agree and have now redrawn them.

Reviewer #3

Major points:

The reviewer requested that we clarify our use of the term oscillation, pointing out that oscillations are repetitive variations in levels/activity over time, whereas the “oscillations” we describe here occur during each round of centriole assembly. This is a fair point, and one that is often debated in the oscillation field, with many believing that too many biological processes are termed “oscillations”, when they are not truly driven by the passage of time. To avoid any ambiguity, we now no longer describe the behaviour of CP110/Cep97 as an oscillation (although, for ease of discussion, we still use the term in this letter).

The reviewer thought that the data we show in Figure 1 was not relevant as we largely analyse centrioles in living embryos whereas the data in Figure 1 is derived from fixed wing-disc cells—and similar fixed-cell data has been shown in previous studies. The reviewer suggests we use super-resolution methods to analyse Cp110/Cep97 dynamics in the syncytial embryo, and show this relative to Sas-6 and Plk4. They ask if Plk4 and CP110/Cep97 colocalise at any time. While CP110/Cep97 localisation has been analysed by super-resolution microscopy previously (e.g. Yang et al., Nat. Comm., 2018; LeGuennec et al., Sci. Adv., 2020), CP110/Cep97 was a minor part of these studies and our data is the first to show that this complex sits as a ring on top of the centriole MTs in fly centrioles (that lack the complex distal and sub-distal appendages present in the previously analysed systems). As this localisation is important in thinking about how CP110/Cep97 might influence centriole MT growth, we would like to include it. We cannot show this detail in living embryos as the movement of the centrioles reduces resolution and we cannot observe the ring structure.

Although we do use Airy-scan super-resolution microscopy to study CP110/Cep97 dynamics in living embryos (Figure 3), we cannot do this in two colours (to compare these dynamics to Sas-6 or Plk4 dynamics) as red-fluorescent proteins bleach too quickly. We now show the relative dynamics of CP110/Cep97 and Plk4 recruitment using standard resolution microscopy (New Figure S2). While it is well established that Plk4 and CP110/Cep97 are concentrated at opposite ends of centrioles, they are all recruited to the nascent site of daughter centriole assembly, effectively “colocalising” at this timepoint. This could provide an opportunity for the crosstalk we observe here, and we now mention this possibility (p17, para.1).

The Reviewer questioned whether the loading of Sas-6-GFP onto centrioles can be used as a proxy for cartwheel length, pointing out that Sas-6 could load into centrioles in a way that does not change the cartwheel structure, and that EM is required to test this. As described in our response to Reviewer #2, Sas-6 is the main structural component of the cartwheel, and we have published two papers that use the incorporation of Sas-6 into the cartwheel as a proxy to measure cartwheel length (Aydogan et al., JCB, 2018; Aydogan et al., Cell, 2020). While we cannot exclude that Sas-6 might also associate with the cartwheel in a way that does not involve its incorporation into the cartwheel, it is not clear how EM might address this question. Moreover, even if such a fraction existed, it should not affect our conclusions—as long as Sas-6 is binding to the cartwheel in some way, then the amount bound should remain proportional to the length of the cartwheel. Perhaps the reviewer is suggesting that we perform an EM time course of cartwheel growth to back up our conclusions from the Sas-6 incorporation assay? If so, we think this impractical. The changes in cartwheel length shown in Figure 6 are revealed from analysing several thousand images of centrioles compared at precise relative time points. Such an analysis cannot be done in fixed embryos by EM.

Similar to the point above, the reviewer notes that we use the length of the cartwheel to infer centriole MT length, but we never directly measure MT length. They suggest we perform either an EM analysis or use MT markers to directly measure the kinetics of centriole MT growth. In flies (and many other organisms), the centriole MTs grow to the same length as the centriole cartwheel (Gonzalez, JCS, 1998), so we can be confident that the final length of the cartwheel reflects the final length of the centriole MTs. Moreover, we previously measured the distance between the mother centriole and the GFP-Cep97 cap that sits at the distal tip of the centriole MTs as a proxy for centriole MT length, and found that the inferred kinetics of MT growth were similar to the kinetics of cartwheel growth (inferred from Sas-6 incorporation) (Aydogan et al., 2018). This manual analysis was very time consuming, and we have tried to implement computational analysis methods, but so far without success. For similar reasons to those described in the point above, it is not feasible to accurately measure centriole MT growth kinetics by EM (nobody has been able to do this). Moreover, the centrosomes in these embryos are associated with too much tubulin and the centriole MTs are not yet modified (e.g. by acetylation) as the cycles are so fast—so we cannot directly stain the centriole MTs in fixed embryos. We have now toned down our conclusions about MT length throughout the paper, and we make it clear that we cannot directly measure this.

All of the experiments shown here are performed in the presence of endogenous untagged proteins, and the reviewer wonders if recruitment dynamics might be influenced by competition for binding from the endogenous protein. We have compared the behaviour of many centriole and centrosome proteins in the presence and absence of the untagged WT protein. In all cases, less tagged-protein binds to centrioles/centrosomes in the presence of untagged protein, presumably due to competition. Apart from this, however, we usually observe no real difference in overall dynamics and in Reviewer Figure 1 (see below) we show that CP110-GFP and GFP-Cep97 both oscillate even in the absence of any endogenous protein. As we feel this result is not very surprising, we do not show it in the manuscript.

The reviewer correctly noted that our data was not strong enough to conclude that the CP110/Cep97 oscillation is influenced by the CCO. This was also raised by Reviewer #2 and, as described above (p2, para.3 above), we have now performed additional experiments to more directly demonstrate this point (new Figure 5G—H).

The reviewer requests more discussion of why our conclusion that CP110/Cep97 levels oscillate on the growing daughter centrioles during S-phase is different to that reached by Dobbelaere et al, (Curr. Biol., 2020), who conclude that Cep97-GFP only starts to incorporate into the new daughter centrioles late in S-phase when the daughters are fully grown. We have discussed this discrepancy with these authors and they kindly shared their reagents with us (so our endogenous Cep97-GFP oscillation data comes from the same line they used in their experiments), but we have not come to a clear conclusion on this point. We have shown robust oscillations for CP110 and Cep97 by quantifying many hundreds of centrioles using multiple transgenes (both over- and under-expressed) in multiple backgrounds. Cep97 dynamics were a very minor part of the Dobbelaere et al., study, and they analysed a much smaller number of centrioles. We now briefly mention this discrepancy (p9, para.1), but do not discuss it in detail as we have no definitive explanation for it.

The reviewer requests more experiments or more discussion to address the mechanism(s) of crosstalk between CP110/Cep97 and Plk4, and they suggest several avenues for further investigations. These are excellent ideas, and we are working hard on these approaches. These are all long-term experiments, however, and we feel it is important that the field be made aware of these surprising findings as soon as possible, as others may be better-placed to provide mechanistic insight into how this system ultimately works. We now briefly mention some of the future directions the reviewer highlights in the Discussion.

The reviewer thought we should highlight the previous publications showing that Plk4-induced centriole amplification requires CP110 and that Plk4 can phosphorylate CP110. These studies (Kleylein-Sohn et al, Dev. Cell, 2007; Lee et al., Cell Cycle, 2017) were mentioned, but we now discuss them more prominently (p17, para.2).

Minor Points:

The reviewer raised a number of minor concerns that we have now addressed: (1) We discuss the model the reviewer suggests; (2) we no longer state that the crosstalk between CP110/Cep97 and Plk4 is unexpected; (3) We have clarified our description of the shift in timing of the peak levels of CP110/Cep97, which we no longer refer to as an oscillation; (4) We define mNG as monomeric Neon Green; (5) We have changed our schematics in Figure 1 as suggested by the reviewer; (6) We have corrected the mistake in the legend to Figure 8.

Reviewer #4

Major points:

1. The reviewer noted that the amplitude of the CP110/Cep97 oscillations depended on protein expression levels, so the oscillations might not reflect the behaviour of the endogenous proteins. They requested that we either repeat our experiments with CRISPR knock-in alleles, or conduct experiments with the lines driven by the endogenous promotors but in their respective mutant backgrounds. We have not generated CRISPR knock-ins for CP110/Cep97, but have done so for many other centriole/centrosome proteins (>8) and found that most such lines are expressed at higher or lower levels than the endogenous allele (and sometimes very significantly so). This is also true for our standard transgenic lines, where genes are expressed from their endogenous promoters, but are randomly integrated into the genome. The blots in Figure 4 show that CP110-GFP and GFP-Cep97 expressed from a ubiquitin (u) promoter or from their endogenous promoters (e) are expressed at ~2-5X higher or ~2-5X lower levels than the endogenous proteins, respectively. As we observe CP110/Cep97 oscillations in all cases, it seems unnecessary to generate new CRISPR knock-ins (that are also likely to be somewhat over- or under-expressed) to show this again. As the reviewer asks, we show that Cep97-GFP and CP110-GFP still oscillate in in the absence of the endogenous proteins (Reviewer Figure 1). As this does not seem a surprising result, we do not show this in the main manuscript. In the same point the reviewer requests that we use antibody staining in fixed embryos to show that the untagged proteins also oscillate. Analysing protein dynamics is much harder in fixed embryos, as the levels of fluorescent staining are more variable and we can only approximately infer relative timing, rather than precisely measuring it (as we can in living embryos). Moreover, as both proteins in the CP110/Cep97 complex exhibit a very similar oscillatory behaviour when tagged with either GFP or RFP (e.g. Figure 2C), and this behaviour is distinct to that observed with several other GFP- or RFP-tagged centriole proteins (e.g. Novak et al., Curr. Biol., 2014; Conduit et al., eLife, 2015; Aydogan et al., JCB, 2018; Aydogan et al., Cell, 2020) it seems very unlikely that this behaviour is induced by the GFP (or RFP) tag.

The reviewer also suggests that we show the data with the endogenous promoter before we show the data with the ubiquitin promoter. As we now explain better (and show in Figure 4), this seems unnecessary as the proteins expressed from the ubiquitin promotor are probably actually expressed at levels that are more similar to the endogenous protein.

The reviewer questions whether the oscillations we observe might be due to the centrioles simply moving up and down in the embryo during the cell cycle, and they suggest we monitor Asl behaviour to rule this out. We have previously shown that Asl-GFP levels do not oscillate; they remain constant throughout the cell cycle on old-mother centrioles, and grow approximately linearly throughout S-phase on new-mother centrioles (see Figure 1D in Novak et al., Curr. Biol., 2014).

We were not sure we understood this point properly, so we copy the reviewers comment in full here: ____The authors mention (for instance on p. 3) that the inner cartwheel and the surrounding microtubules assemble at opposite ends of the daughter centriole. However, my understanding is that the short centrioles present in the fly embryo have an inner cartwheel that extends throughout the organelle, such that it might be moot to make a distinction between the two ends in this case. Moreover, it is also my understanding that this inner cartwheel is itself surrounded by microtubules, so that microtubule assembly might not be expected to occur strictly at the distal end no matter what. The reviewer is correct that Drosophila centrioles are short (~150nm) and that the cartwheel extends throughout the centriole. We think the reviewer is suggesting that it may not be relevant therefore whether the cartwheel and centriole MTs grow from opposite ends—as the activities that govern their growth may not be spatially separated? However, because cartwheels grow preferentially from the proximal-end (Aydogan et al., JCB 2018) while centriole MTs are assumed to grow preferentially from the distal (plus) end, there is an intrinsic problem in ensuring they grow to the same size—no matter how short or long the centrioles are. The reviewer is correct that one possible solution to this problem is that the centriole MTs actually grow from their minus ends, but this is not widely accepted (or even proposed). We have tried to explain this issue more clearly throughout the revised manuscript.

The reviewer points out that the schematic illustrations in Figure 1A and 1C are inaccurate and unhelpful. We agree and have now redrawn these.

The reviewer asks that we provide information about the eccentricities of the centrioles in the different datasets used to calculate the protein distributions shown in Figure 1, particularly as the data for Sas-4-GFP and Sas-6-GFP were obtained previously using a different microscope modality, making comparisons complicated. The point that comparing distance measurements across different datasets is difficult is an important one, and we now state that such comparisons should be treated with caution. However, we have not provided information on the distribution of centriole eccentricities in the different experiments as it wasn’t clear to us how this information could be used to make such comparisons more accurate (presumably the reviewer is suggesting we could apply a correction factor to each dataset?). The very tight overlap in the positioning of CP110/Cep97 fusions (Figure 1C) strongly suggests that any difference in the average centriole eccentricities of the different populations of centrioles analysed, which are already tightly selected for their en-face orientation (i.e. eccentricity

The reviewer requested that we show the “noisy data” we obtained during mitosis that we excluded from our analysis in Figure 3. As we now explain in more detail (p8, para.2), there are two reasons why the data for mitosis in this experiment is “noisy”: (1) The protein levels on the centrioles are low in mitosis and the centrioles are more mobile, so they are hard to track; (2) The Asl-mCherry marker used to identify the mother centriole starts to incorporate into the daughter (now new mother) centriole during mitosis, making it difficult to unambiguously distinguish mothers and daughters. As a result, we cannot track and assign mother/daughter identity to very many centrioles during mitosis—although we now include some extra data-points during mitosis for the centrioles where we could do this (revised Figure 3C,D). Importantly, it is clear that this “noisy” data hides no surprises: one can see (Figure 3C,D) that the signal on the centrioles is simply low during mitosis and then starts to rise again as the embryos enter the next cycle. This is confirmed in the normal resolution data (Figure 2B,C; Movies S1 and S2) where we can track many more centrioles due to the wider field of view and because we do not have to discard centrioles in mitosis that we cannot unambiguously assign as mothers or daughters.

The reviewer requests that we conduct a super-resolution Airy-scan analysis of CP110/Cep97 driven from their endogenous promoters (eCP110 or eCep97) to ensure that the oscillations we see with these lines (shown in Figure 4C,D) are also occurring at the daughter centriole—as we already show for the oscillations observed with the uCP110 and uCep97 lines (shown in Figure 4C,D, and analysed at super-resolution on the Airy-scan in Figure 3). This is technically very challenging as super-resolution techniques require a lot of light and the centriole signal in the eCP110/Cep97 embryos is very dim compared to uCP110/Cep97 embryos (Figure 4C,D). We have managed to do this for eCep97-GFP and confirmed that—even in these embryos that express Cep97-GFP at much lower levels than the endogenous protein (Figure 4A)—the “oscillation” is primarily on the daughter (Reviewer Figure 2). As this data is very noisy, and as the ubiquitin uCP110/Cep97 lines express these fusions at levels that are closer to endogenous levels (Figure 4A,B), we do not show this data in the main text.

The reviewer also asks for clarification as to why we use the Airy-scan for some experiments and 3D-SIM for others. As we now explain (p8, para.1), 3D-SIM has better resolution than the Airy-scan, but it takes more time and requires more light—so we cannot use it to follow these proteins in living embryos. Thus, for tracking CP110/Cep97 throughout S-phase in living embryos we had to use the Airy-scan.

The reviewer questions why in some experiments we analyse the behaviour of 100s of centrioles, whereas in others the numbers are much smaller (1-14 in Figure 3—note, the reviewer quoted this number as coming from Figure 4, but it actually comes from Figure 3, so we have assumed they mean Figure 3). We apologise for not explaining this properly. The super-resolution experiments in Figure 3 are performed on a Zeiss Airy-scan system, which has a much smaller field of view than the conventional systems we use in other experiments. Thus, we inherently analyse a much smaller number of centrioles in these experiments. In addition, as explained in point 6 above, in these experiments we need to analyse mother and daughter centrioles independently, and in many cases we cannot unambiguously make this assignment, so these centrioles have to be excluded from our analysis.

The reviewer questions why we selected the 10 brightest centrioles for the analysis shown in Figure S1B,C (note, the reviewer states Figure S2 here, but it is the data shown in Figure S1B,C that is selected from the 10 brightest centrioles, so we assume this is the relevant Figure). We apologise for not explaining this properly. In these mutant embryos very little CP110-GFP localises to centrioles in the absence of Cep97, and vice versa, so we cannot track centrioles using our usual pipeline and instead have to select centrioles using the Asl-mCherry signal. As the difference between the WT and mutant embryos is so striking, we simply selected the brightest 10 centrioles (based on Asl-mCherry levels) in both the WT and mutant embryos for quantification. We could select more centrioles, or select centrioles based on different criteria, but our main conclusion—that the centriolar localisation of one protein is largely dependent on the other—would not change.

The reviewer also questioned why we performed the analysis shown in Figure S2 (new Figure S3) during S-phase of nuclear cycle 14, when the rest of the manuscript focuses on nuclear cycles 11-13. We apologise for not explaining this properly. In cycles 11-13 centriolar CP110/Cep97 levels rise and fall during S-phase, whereas both proteins reach a sustained plateau during the extended S-phase (~1hr) of nuclear cycle 14—making it easier to analyse CP110/Cep97 levels in embryos when their centriole levels are maximal. We now explain this.

The reviewer requests that we quantify the western blots shown in Figure 4 in the same way we do in figure 8. To do this we would need to perform multiple repeats of these blots and we did not perform these because the blots shown in Figure 4 largely recapitulate already published data (Franz et al., JCB, 2013; Dobbelaere et al., Curr. Biol., 2020). Moreover, as described in our response to Reviewer #2, these ECL blots are very sensitive, but highly non-linear, so we always compare multiple serial dilutions of the different extracts to try to estimate relative levels of protein expression. We now explain this in the M&M.

The reviewer suggests the data shown in Figure 8 is a “straw man”: we really want to test whether modulating CP110/Cep97 levels modulates centriolar Plk4 levels, but instead we test how they modulate cytoplasmic Plk4 levels. The language here is harsh, as it suggests that our intention was to mislead readers into thinking that we have addressed a relevant question by addressing a different, irrelevant, one. We apologise if we have missed something, but we believe we do perform exactly the experiment that the reviewer thinks we should be doing—quantifying how centriolar Plk4 levels change when we modulate the levels of CP110 or Cep97 (Figure 7). It is clear from this data that modulating the levels of CP110/Cep97 does indeed modulate the centriolar levels of Plk4. In Figure 8 we seek to address whether this change in centriolar Plk4 levels occurs because global Plk4 levels in the embryo are affected—a very reasonable hypothesis, which this experiment addresses quite convincingly (although negatively).

Minor Points:

The reviewer highlights a small number of mistakes and omissions, all of which have been corrected.

Finally, we would like to thank the reviewers again for their detailed comments and suggestions. We hope that you and they will agree that the changes we have made in response to these comments have substantially improved that manuscript and that it is suitable for publication in The Journal of Cell Science.

Sincerely,

Jordan Raff

__Reviewer Figure 1. CP110/Cep97 dynamics remain cyclical even when Cep97-GFP and CP110-GFP are expressed from their endogenous promotors in the absence of any endogenous protein. __Graphs show how the levels (Mean±SEM) of centriolar CP110/Cep97-GFP change during nuclear cycle 12 in (A) Cep97-/- embryos expressing eCep97-GFP or (B) CP110-/- embryos expressing eCP110-GFP. CS=Centrosome Separation, NEB=Nuclear Envelope Breakdown. N≥11 embryos per group, average of n≥15 centrioles per embryo.

__Reviewer Figure 2. ____The cyclical recruitment of Cep97-GFP expressed from its endogenous promoter occurs largely at the growing daughter centriole. __The graph quantifies the fluorescence intensity (Mean±SD) acquired using Airy-scan microscopy of eCep97-GFP on mother (dark green) and daughter (light green) centrioles in individual embryos over Cycle 12. CS = Centrosome Separation, NEB = Nuclear Envelope Breakdown. Data was averaged from 3 embryos as the number of centriole pairs that could be measured was relatively low (total of 2-8 daughter and mother centrioles per time point; in part due to the much dimmer signal of eCep97-GFP in comparison to uGFP-Cep97).

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

The authors report that CP110 and Cep97 localize near the distal end of centrioles in Drosophila embryos. CP110 and Cep97 tagged with GFP exhibit an oscillatory distribution, with levels on the daughter centriole being maximal in mid S-phase. These oscillations correlate with cell cycle progression. The authors also show that modulating CP110 or Cep97 levels changes the rate at which Sas6-GFP incorporates in the daughter centriole, as well as aspects of the previously reported oscillatory behavior of Plk4.

These results could be of potential interest if the stated conclusions were backed up by more convincing data than that which is provided at present. The issues delineated hereafter must be addressed in full before further consideration of the manuscript.

Major points

1) The oscillatory amplitude of CP110/Cep97 tagged with GFP is much smaller when expression is driven by the endogenous promoters than upon overexpression (see Figure 4), raising the possibility that oscillation might not reflect, or only reflect in part, the behavior of the endogenous proteins. To address this issue, the authors could GFP tag the endogenous loci using CRISPR/Cas9. If this is too demanding, they should at the minimum conduct experiments with the extant lines driven by the endogenous promoters, but in the background of the available CP110 or Cep97 null mutants. Moreover, the authors should stain staged wild-type embryos with antibodies against CP110 and Cep97 to ensure that the mild oscillations reported in Figure 4 do not merely reflect the behavior of the tagged proteins, for example due to the presence of GFP. Related to this point, the authors should considering showing first the data with CP110-GFP GFP-Cep97 driven from the endogenous promoters (current Figure 4), perhaps relegating the results upon overexpression (current Figure 2) to a Supplementary Figure.

2) In repeating the above experiments, the authors must ensure that potential mild oscillations do not simply reflect the fact that centrioles are located at a slightly different distance from the coverslip as a function of cell cycle stage. This could be addressed for example by simultaneously imaging a mother centriole marker such as Asl-mCherry.

Other important points

3) The authors mention (for instance on p. 3) that the inner cartwheel and the surrounding microtubules assemble at opposite ends of the daughter centriole. However, my understanding is that the short centrioles present in the fly embryo have an inner cartwheel that extends throughout the organelle, such that it might be moot to make a distinction between the two ends in this case. Moreover, it is also my understanding that this inner cartwheel is itself surrounded by microtubules, so that microtubule assembly might not be expected to occur strictly at the distal end no matter what.

4) Partially related to the point above, the schematic representations in Figure 1 are somewhat confusing. The schematic in Figure 1A represents CP110/Cep97 strictly at the distal end of the centriole, yet the actual immunofluorescence data on the left suggests that CP110/Cep97 are in fact present very close to Asl-mCherry. This apparent difference must be resolved. Moreover, Figure 1C seems to indicate that all the depicted proteins are present throughout the centriole, which I guess is not what the authors wanted to convey.

5) For the quantification of the data reported in Figure 1, the authors considered only centrioles for which CP110/Cep97 ring eccentricity was less than 1.2, to ensure that only near top views are considered (see p. 23). This is entirely reasonable, but the authors should report the distribution of eccentricities in the data set for the two proteins, and compare them to those of the Sas6-GFP and Sas4-GFP data set, all the more since the latter two were obtained previously with a different microscope modality, potentially complicating thorough comparisons. A slight difference in the fraction of centrioles with a slight tilt could easily skew the data when dealing with such small dimensions.

6) In Figure 3, the authors chose not to report the "Noisy data" observed during mitosis. While it is understandable that the data is noisier at this stage, it must nevertheless be reported, as this may have bearing on assessing oscillations between cycles 12 and 13.

7) The authors should conduct Airy-scan analyses of CP110/Cep97 oscillations driven from the endogenous promoters, to ensure that the variations across the cell cycle reported in Figure 4 reflect changes in the daughter centriole. Moreover, it was not clear why the authors used the Airy-scan for some super-resolution experiments and 3D-SIM for others.

8) Why are solely 1-14 centrioles per embryo considered in the experiments reported in Figure 4 as compared to over 100 per embryo in Figure 2? And how were these centrioles chosen? This needs to be explained, justified and, potentially, rectified.

9) Likewise, why are only the 10 brightest centriole pairs in each embryo retained for the analysis reported in Figure S2? And would the conclusion differ if more centrioles than that were included? Moreover, S phase of cycle 14 is analyzed in Figure S2 for Sas6-GFP, whereas the remainder of the manuscript analyzes CP110/Cep97 during cycles 11 through 13 (with an emphasis on cycle 12). This must be resolved.

10) The Western blots in Figure 4A, 4B, as well as in Figure S1A, should be quantified in the same manner as those in Figure 8C, to achieve a better assessment of the differences in protein levels between conditions.

11) The set up for the experiment reported in Figure 8 comes across as a straw man. What one would really like to find out is whether levels of Plk4 at centrioles are modulated by CP110/Cep97 levels, as the authors themselves acknowledge. Since this does not appear to be feasible, the authors set out to test whether cytoplasmic levels of Plk4 differ, finding that this is not the case. Since this experiment does not address what should be tested, it could be reported as a Supplementary Figure, not as the last main figure of the manuscript.

Minor points

• The authors forgot to mention the Tang et al. paper (doi: 10.1038/ncb1889) when referring to Sas-4/CPAP (for instance on p. 4).
• On p. 9, the authors conclude that the "recruitment of CP110/Cep97 to centrioles is regulated by the CCO". Figure 5 shows that the two correlate, not that the latter regulates the former. A related comment holds for the discussion (bottom of p. 13).
• It is not clear why the authors sometimes report SDs (Figure 7) and sometimes SEMs (Figure 3), or fail to report what is being shown (Figure 2). This needs to be clarified.
• The legend of Figure 8A mentions Pie charts and other things that are not featured in the current rendition of the figure.

Significance

These results could be of potential interest if the stated conclusions were backed up by more convincing data than that which is provided at present.

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 study uses nuclear cycles 11-13 of Drosophila embryos to show the dynamics of the distal centriole localizing CP110/Cep97 complex during the predicted time of MT assembly during new centriole assembly. Continuing from prior work from this group, the authors find that the increase and decrease in CP110/Cep97 at new centrioles correlates with the timing of Cdk/Cyclin oscillations (CCO). The authors find that increased or decreased levels of CP110/Cep97 changes the dynamics of SAS6 and Plk4 levels. The authors suggest that there is crosstalk between the distal localizing CP110/Cep97 complex and the proximal localizing Plk4 and SAS6 proteins required for early centriole assembly.

MAJOR

Overall, the results are potentially interesting but I believe that there a number of instances in this manuscript where the conclusions need to either be strengthened with further experiments or toned down to reveal exactly what is shown in the manuscript.

CP110/Cep97 OSCILLATIONS

Because oscillations are repetitive variation in levels/activity with time, I think the manuscript needs to either use other terms that accurately describe what is measuring here or it should be defined what the authors are calling an oscillation. CP110/Cep97 only increases and then decreases during a single new centriole assembly and maturation event and I think that this should be clearly describe it this way.

LOCALIZATION OF CP110/Cep97 TO DISTAL END OF CENTRIOLES

Based on the existing published studies, it is clear that CP110/Cep97 localizes to the distal end of centrioles. Figure 1 does not show distal centriole localization in daughter centrioles of the syncytium that are the subject of this manuscript though. Its shows radial localization in the mother centriole of the fly wing. Figure 1 therefore has not relevance to the rest of the manuscript and has already been shown in prior studies.

My suggestion would be that this figure should study the dynamic localization of CP110/Cep97 at daughter centrioles during new centriole assembly in the syncytium. Moreover, this should localize these proteins relative to SAS6 and Plk4 that are the subject of the manuscript. Are there localization dynamic changes during the oscillation? Are there times when these proteins do co-localize?

SAS6 AND CW CONCLUSIONS

The current manuscript routinely equates SAS6 levels to cartwheel growth. This is overstated and EM is required to understand whether this is truly impacting the actual cartwheel structure. Loading more sas6 protein doesn't necessarily mean the cartwheel structure changed.

CONNECTION BETWEEN OSCILLATIONS AND MT GROWTH?

Much as above, the manuscript infers MT growth without ever showing it. How does all of this relate to centriole length and growth dynamics.? Page 8 refers to prior work but it seems like this is necessary with EM or MT markers. Having this comparison seems important to the conclusion that MTs do not stop growing when CP110/Cep97 levels reach a threshold level at the distal end.

The following statement is overstated when the data for MT growth are not even presented in this study. "...our findings essentially rule out the possibility that centriole MTs stop growing when a threshold level of CP110/Cep97 accumulates at the centriole distal end." To make such arguments in this study the manuscript would need to include EM and / or MT staining.

ENDOGENOUS UNTAGGED PROTEIN AFFECTING DYNAMICS?

The manuscript shows protein dynamics under conditions of both overexpressed and expression under the endogenous promoter. However, I believe that both of these conditions are also in the presence of untagged protein expression.(?). If so, is it possible that the dynamics represent competition for binding relative to the endogenous, untagged protein? I think this point should at least be discussed.

CP110/Cep97 "INFLUENCED" BY CCO

While I agree that it is likely to be the case that CP110/Cep97 rise and fall at the daughter centriole correlates with CCO, this study does not directly test if CCO changes impact CP110/Cep97 dynamics. Stating that "CP110/Cep97 oscillation is strongly influenced by the activity of the core Cdk/Cyclin cell cycle oscillator (CCO)" is overstated. Is does correlate though.

DISTINCTION FROM PRIOR STUDIES

Dobbelaere 2020 argue that CP110/Cep97 gets to the centriole distal end in late S phase. How could this be considering the data presented in this study? Need discussion of this point. Could Dobbelaere be following the dynamics of the core / basal levels and missed the dynamics that are found in this study? I think a discussion of the Cep97 functions needs to be provided.

MECHANISM OF CROSS TALK

How two apparently spatially separated complexes influence each other should be more mechanistically addressed through either or both experimentation and / or discussion. Obviously the impact of this study would greatly benefit by showing how they are associated and influence each other. CP110 is a phospho target of Plk4. Does this occur in the fly syncytium? Do these interact? What is the timing of the interaction and phosphorylation? Are the changes to SAS6 levels actually the result of Plk4 changes? At this point, these concepts are not tested.

BACKGROUND

In its current form the prior results that 1) Plk4-induced centriole amplification requires CP110 and 2) Plk4 phosphorylates CP110 is important for centriole assembly in some systems is not highlighted in this manuscript as further support for the model of interplay between CP110/Cep97, Plk4 and SAS6.

REPRODUCTION OF DATA

I believe that the data and methods are of high quality and described in such a way that they can be reproduced.

MINOR

ALTERNATIVE MODEL

Because CP110 is a target of Plk4, I wonder if the elevated expression of CP110 sequesters Plk4 away from its cartwheel functions (Ana2/STIL/SAS5 phosphorylation) and this is therefore affecting SAS6 levels?

OVERSTATED CROSSTALK

The text states a "...reveals an unexpected crosstalk between proteins that are usually thought to influence the proximal end of the CW and the distal end of centriole MTs." This is true but there are enough data in the literature to suggest that CP110/Cep97 influence centriole assembly that would indicate that this is not "unexpected".

PAGE 11 - SHIFT IN PEAK

I could not find the data clearly showing that there was a shift in "the Plk4 oscillation to later in S-phase". Are the authors referring to the plateau in levels? Please explain further.

WHAT IS "Plk4-NG"?

I assume Neon Green but I don't see the definition.

FIGURE 2

A schematic of the system used for image averaging would help the reader to understand that these "oscillations" represent the mother and daughter centriole together and that each "oscillation" represents one event of the daughter centriole only increasing in CP110/CEP97 levels and then decreasing after peak intensity.

FIGURE 5 and 8

I think these could be supplemental images. I was unable to figure this out but something is wrong with the legend in Figure 8. (A) is referencing items that I cannot find in the figure.

Significance

This study's advance is an expansion of the authors' prior work showing that during the fly nuclear cycles centriole assembly proteins increase and then reduce in what the authors call an oscillation. Here they show that the CP110/Cep97 complex also oscillates and somehow influences the levels of Plk4 and SAS4 that typically reside at the proximal end of the centriole. This is consistent with prior work indicating that, in some systems, CP110/Cep97 influence centriole duplication and assembly.

I believe that with additional experiments to strengthen the conclusions and toned down concluding statements this will be of interest to the centriole, centrosome, and cilia community. My research expertise is also in this community but I am not a Drosophila researcher. I do appreciate the beauty of this system that the authors use.

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

In this study, Aydogan, Hankins, and colleagues, present an interesting work that follows up on their article "An Autonomous Oscillation Times and Executes Centriole Biogenesis" published last year in Cell. In this new study, they analyzed the distal complex consisting of CP110/Cep97 in the centriole of Drosophila embryos. They first demonstrated their oscillatory recruitment at the distal tip of the daughter centriole and they proposed that this protein complex is implicated in the control of centriole growth timing. They also demonstrated the importance of the crosstalk between CP110/Cep97 and Plk4 and its impact on cartwheel growth. This paper proposes a compelling model explaining how centriole growth is regulated. This manuscript is very well written and the data is of high quality. However, some point needs to be clarified before publication:

Major points:

• Figure 1: Since SAS-4 and CP110/CEP97 are only 5nm apart, SAS-4/CPAP is thought to have an antagonistic function to CP110 in the regulation of centriolar growth, and Plk4 can phosphorylate CPAP (DOI: 10.1038/emboj.2010.118), do the authors think that SAS-4 might also be involved in cartwheel/centriole elongation? Does SAS-4 oscillate?
• Figure S1B: The reduction in the intensity of CP110 in Cep97-/- and of Cep97 in CP110-/- is very obvious, nevertheless it is surprising that the cytoplasmic background, even reduced, is not visible, the images are completely dark. Would it be possible to image with a higher laser power or boost the intensity to see if a small amount is present at centrioles?
• Figure 3: The authors indicate that "uGFP-CP110 or uGFP-CEP97 levels remained relatively constant on the mother". However, the intensity clearly decreases over time. Can the authors explain this result, is it due to photobleaching?
• Do the oscillations of CP110 and Cep97 occur at or around the tip of the growing centriole? Would it be possible to use super-resolution at different stages of the S-phase to answer this question?
• The authors indicate that the level of overexpression of CP110-GFP and Cep97-GFP is 2.5X compared to their endogenous proteins (based on the western blot in Figure S1). Nevertheless, it seems that the overexpression of CP110 is more important. Quantification is necessary here.
• The authors proposed that "The CP110/Cep97 oscillation is entrained by the Cdk/Cyclin cell cycle" because they observed a strong and significant correlation between the timing of the CP110/Cep97 peak and S-phase length for both uGFP-Cep97 and uCP110-GFP at all nuclear cycles. It seems to me that this correlation is not sufficient for this statement. If it is not possible to inhibit the CCO to check its impact on CP110/Cep97, this statement should be mitigated.
• Figure 6: According to your results, cartwheels are longer in absence of CP110 or CEP97 and opposite in overexpression situations. Does the intensity perfectly reflect the length of the cartwheel? is the centriole longer? Could you confirm your observation on cartwheel/centriole length using electron microscopy?

Minor points:

• Figure 1C: as the authors show that CP110 and Cep97 are localizing at the distal end of the centriole, I suggest that they place CP110 and Cep97 distally and not at the level of the cartwheel, this representation can be misleading and suggest that CP110 and Cep97 are part of the cartwheel/MT connection.

Significance

The results presented are new and quite unexpected. This work allows a better understanding of phenotypes previously observed. I believe that this work will have an important impact in the field as it brings a whole new vision on the regulation of centriole growth. This article is primarily aimed at centriole/centrosome/cilia fields but may be of interest to a broader cell biology audience.

My field of expertise is centriole/cilia biology

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 is a continuation of the previous articles of the authors (Aydogan et al., JCB 217:1233, 2018; Aydogan et al., Cell 181:1566, 2020). They reported that Plk4 initiates and times the growth of the cartwheel at the proximal end during early divisions of the Drosophila embryos. In this manuscript, they investigated roles of the CP110/Cep97 complex in the centriole growth control at the distal end of the centriole. The daughter centriole levels of the CP110/Cep97 complex oscillate in S phase in a similar manner to those of Plk4. The CP110/Cep97 oscillation is entrained by the core Cdk/Cyclin cell cycle oscillator but not by Plk4. Rather, the centriolar levels of Plk4 increased in the CP110 and Cep97 deletion embryos. The experiments seem to be carefully carried out, data are nicely presented, and manuscript is clearly written.

Significance

I agree with their interpretation that the CP110/Cep97 oscillation does not appear to play a major part in determining the period of daughter centriole growth during early divisions of the Drosophila embryos. The CP110/Cep97 complex seems to have a limited role in the centriole length control. The CP110/Cep97 complex may be important to prevent centrioles from over-elongating after the initial growth of centrioles.

As suggested in the manuscript, phosphorylation may be a regulatory mechanism for CP110 behaviors at the centrioles. It was previously reported that CP110 is a substrate of the cell cycle kinases, such as Cdk2 (Chen et al., Dev Cell 3:339, 2002) and Plk4 (Lee et al., Cell Cycle 16:1225, 2017). Phosphorylation may be required for recruitment or removal of CP110 at the centrioles. Nonetheless, it is hard to interpret the functional significance of the S phase oscillation of the CP110/Cep97 complex with the data in the manuscript.

It is unfortunate to conclude that the CP110/Cep97 complex may not be a major player for controlling the centriole growth. However, the manuscript includes other interesting observations. For example, they presented data supporting that the SAS6 protein is added at the proximal side of the centrioles, which is opposite to the microtubule growth. Microtubules in the daughter centrioles may assemble at the minus end rather than the plus end. It would be interesting to determine when γ-tubulins are recruited to the growing centrioles.

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

This manuscript was evaluated at Review Commons by four individual reviewers. There was a consensus amongst reviewers that the localization behavior of altORF peptide to the Golgi is a compelling observation and that, with some additional characterization, would provide an effective cell biological tool for use in labeling and studying the Golgi. Our primary goal for this paper was to explore this surprising alternative protein hidden within the sequence of a centromere component and to establish this peptide as a cell biological tool that can be used to study the Golgi. However, the reviewers also highlighted some interesting open questions regarding the nature of this peptide. Below we summarize these core comments our current changes and plans

• Where within the Golgi does the peptide localize? In the work currently included in the paper, we demonstrate that the altORF peptide robustly colocalizes with markers for the Golgi (GM130/TGN46), but not with markers for the Endoplasmic Reticulum (KDEL). However, the resolution at which we imaged the localization of the peptide was not sufficient to determine in which compartment of the Golgi the peptide resides. To address reviewer comments on the specificity of the peptide’s localization within the Golgi, we will attempt to use higher resolution imaging such as confocal or spinning disk microscopy to attempt to better resolve this.
• How does the peptide target to the Golgi? In this manuscript, we show that the localization of the altORF peptide relies on a Cysteine residue present within in a minimal 10 amino acid sequence. Through treatment with 2-Bromopalmitate (2-BP; a palymityltransferase inhibitor) to disrupt its localization, our work suggests that the peptide is palmitoylated. In addition to this observation, the reviewers asked for an additional demonstration that this peptide is palmitoylated in cells. To test this, we have attempted to identify this modification using mass spectrometry of the isolated (IP) GFP tagged peptide from cells. However, we were unable to identify peptides that coincide with the modified peptide cysteine residue. Secondly, we have attempted to identify the modification using Click-chemistry labeling strategy, but this has proved to be technically challenging and infeasible. As an alternative approach for the revised version, we will attempt to perform hydroxylamine treatment followed by SDS-PAGE analysis to determine whether this results in a shift in migration of the GFP tagged altORF, as suggested by a reviewer, to provide additional evidence that the peptide is modified.
• Can this peptide be used to ectopically target proteins to the Golgi? The reviewers asked whether the altORF peptide can be used to ectopically target proteins to the Golgi. In this manuscript, we demonstrate that the peptide sequence is sufficient to target both GFP and the Halo tag (two very different proteins) to the Golgi, and can be tagged at either terminus of the peptide, suggesting that it can be used as a powerful strategy to recruit other proteins to the outer surface of the Golgi. We have emphasized this point in the updated version that is included in this revision.
• Does this peptide alter Golgi structure? For this peptide to provide a useful cell biological marker, it would be preferential for it not to alter cellular physiology. Our work demonstrates that expression of the altORF peptide does not affect the growth of cultured cells. For this updated version, we have performed additional analysis to test whether induced expression of the altORF peptide alters the structure of the Golgi or the localization of other Golgi-associated proteins. Based on a qualitative analysis of these cells, we do not detect any obvious changes in Golgi organization or morphology. This is now included as Supplemental Figure 2D.
• • Is this peptide expressed in human cells? *We have analyzed published ribosome profiling data that suggests that this altORF can be translated, although it is produced to a much lower degree than the full-length CENP-R protein. The short length of the peptide as well as the nature of the amino acid sequence makes this peptide highly challenging to identify via mass spec. It is also possible that this peptide would be expressed in different cell types in the human body, but not robustly expressed in cultured cells. We believe that these are beyond the scope of this paper. However, we now comment on these important points in the updated version.
• • Is this peptide “functional”? *Based on our deliberate analysis of the evolutionary conservation of this altORF within the CENPR transcript, it is clear that this peptide acquired the ability to localize to the Golgi only recently during evolution (only old world primates have this capacity). We believe that this peptide represents a great example of evolution in action, with minor sequence changes resulting in the acquisition of a new capacity and trait. However, as this peptide is not broadly conserved across mammals, it is unlikely to facilitate a core biological function that can be analyzed in cell culture. It is certainly possible that this peptide would contribute to a feature of human biology on the organismal level, but it is not feasible to test this experimentally. The functional nature of this peptide, and particularly the recent evolutionary acquisition of this novel trait are interesting points that we have now commented on in the updated manuscript (text changes in blue).

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

The manuscript reports the characterisation of a 37 amino acid alternative open reading frame (altORF) within the RNA of the centromere protein, CENP-R. The resulting peptide, when expressed in different cell lines fused to GFP, localises on the Golgi complex, exposed on the cytosolic face of Golgi membranes. It remains associated with the Golgi complex under conditions inducing fragmentation or dispersal of the Golgi complex such as mitosis and BFA. The authors identify in aa 5-14 the minimal Golgi targeting motif and in cysteine 11 a key aa for the targeting. They suppose that palmitoylation may be involved in Golgi targeting as palmitoylation inhibitors prevent its Golgi targeting. The data are clearly presented and sustain the conclusions.

Significance

Though the identification of a Golgi targeting motif is of potential interest, the manuscript appears to be at a preliminary stage as it fails to provide any data on the possible function of the altORF of CENP-R palmitoylation or even evidence for its existence in the cells used in the manuscript. The authors appear to be aware of the limits of their study as they conclude their study led to the identification of an "easy-to-use Golgi labeling construct". Also in this scenario, however, some key information are missing: the actual sub-Golgi localisation of the probe, its possible impact on Golgi structure and function, and the formal proof that it is palmitoylated.

Referess cross-commenting

I see all the reviewers agree that the manuscript has major limits. Overcoming these limits wold require years if one had to provide proofs for the existence and for the physiological relevance of this alternative ORF, and months to provide the missing information that have been highlighted by the reviewers to consider "just" the technical aspect of this altORF as a possible Golgi reporter/targeting sequence.

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, the authors characterized a potential alternative open reading frame close to the CENP-R open reading frame that had previously been found by ribosome profiling. Its 37 amino acid peptide sequence was included in a proteomic database and is conserved in primates. Transfection of different cell lines with the GFP-tagged peptide was used for immunofluorescence and proteolytic cleavage by a cytosolic protease was used to show that it localizes to the cytoplasmic face of Golgi membranes throughout the cell cycle and Brefeldin treatment had no influence on fragmentation or reformation of the Golgi stacks. The specific localization could be confirmed using different cell lines. The use of numerous truncation mutants allowed to narrow down the minimal Golgi recognition sequence to a 10 amino acid stretch including a species-specific conserved cysteine that required palmitoylation. From these data and comparison with similar sequences in other species the authors determined a consensus sequence for this Golgi targeting sequence in primates.

Major comments:

1. Without ultrastructural analysis it is always difficult to judge whether a localization is limited to just one organelle. Immunofluorescence alone gives no clear answer in particular when organelles differ in size and form from cell to cell. In particular when the authors claim that the peptide may serve as a marker. For example when you are working on secretion it is important to distinguish membranes derived from ER exit sites (ERES), the ER-Golgi intermediate compartment (ERGIC), the Golgi itself and Golgi-derived vesicles. I therefore recommend to add a subcellular fractionation by which numerous fractions can be analyzed by a gel in parallel using markers for all the above mentioned different membrane origins.
2. Is it possible to confirm the in vivo existence of this peptide? There are probably no specific antibodies available, but it should be possible to detect the peptide in enriched Golgi membrane fractions by mass spectrometry.
3. It would be interesting to reveal the potential in vivo role of this peptide, when it exists. The authors failed to identify potential interaction partners by IP-MS, so I wonder whether its role may be different by controlling the Golgi association of other well known Golgi interactors like GM130, Golgin or GORASP proteins. Is their Golgi association altered in the presence of the peptide?
4. Finally the authors determined a consensus sequence which they claim to be a Golgi targeting sequence, however when this is true one would expect that there are other proteins in the cell that use this consensus sequence as targeting sequence. The authors only show that the consensus is conserved among the same alternative open reading frame in primates, but to serve as a Golgi targeting sequence it should be possible to identify unrelated other proteins using this consensus by bioinformatics. What happens when an otherwise differently addressed protein is attached to this Golgi sequence, is it mislocalized?

Minor comments:

There are a couple of typos and smaller issues - In the Introduction line 2 the citation is missing and skip the "a" in line 7. - In the Results and Discussion section page 5, line 5 "In our ongoing work, we..." - In the same section close to the end in the second from the last paragraphs Figure 5B should be Figure 5C - In the Methods section check the temperature specifications: 4{degree sign}C or 37{degree sign}C, not 4C or 37C - Also in the Methods: there are no secondary antibodies recognizing complete animals (antiMouse or antiRabbit)! The antibodies are directed either against IgG or IgM (e.g. anti-Mouse IgG) - Some subscripts are missing: MgCl2 not MgCl2, NaN3 not NaN3 - Also on the last page of the Methods section the antibody is specific for TGN46, not TGN146 - Last paragraph: for concentrations use μM not uM (also in the Fig.4 legend) - The end of the second from the last sentence is missing. - In the References, is the citation for the Samandi et al. manuscript correct, just one number? - Legend to Suppl. Fig 3: Golgi (capital letter), (~) is missing in figure - Suppl. Fig1B use Courier also for peptide sequence, this will omit alignment problems

Significance

Overall, this study is interesting and may provide a helpful tool for cell biologists working on trafficking projects (like myself) in particular because a general Golgi targeting sequence is missing. For techniques like RUSH (Retention using specific hooks) which can be used to synchronize secretory protein traffic reliable and highly specific targeting sequences are required. I am supportive of this study, however, to be useful for the audience the authors need to provide more examples demonstrating the targeting efficiency.

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 have identified a alternately translated region with in the mRNA of CENP-R that encodes a small 37 aa peptide that localizes to the perinuclear Golgi region. The main premise here is the this peptide can be used as a novel Golgi marker. The peptide seems to localize peripherally to membranes and interacts with cis and TGN elements based on light microscopy. Mutational analysis indicates that a cysteine residue within a 10 aa region is critical and defines a minimal consensus sequence required for Golgi localization. Evidence is presented based on inhibition by 2-bromopalmitate that C11 is palmitoylated.

Significance

If this peptide probe is to be used as a Golgi-specific marker, there are several major issues that have not been addressed. The first is whether it actually binds to Golgi elements and if so, what are the specific elements? The light microscopy images are not of high enough resolution to determine if the peptide interacts with cis or TGN Golgi. The BFA experiments suggest it interacts with the TGN or some other associated vesicular compartment since staining fragments into vesicles and does not get integrated into the ER (Fig. 2B). The authors would have to use higher resolution confocal imaging or, more preferably, immuno-EM to identify exactly where the peptide is located.

The second issue is the conclusion that the peptide is palmitoylated, which is only based on partial loss of 'Golgi' staining after 2-BP treatment (Fig. 4D). More conclusive evidence is required such as incorporation of radiolabelled or click-palmitate probes into peptide, or band shift after hydroxylamine treatment. In regard to the last point, the protein seems to migrate as a doublet on SDS-PAGE (Fig. 2D) suggesting some type of modification or cleavage that is not commented on.

Lastly, I would be unlikely to use this as a Golgi probe for the reasons described above, as well as the fact that there is nothing known about the biological function of the peptide (this is potentially the most interesting aspect that is seemingly ignored). If you express the peptide what impact does it have on Golgi structure and function? I could envision that its binding to a Golgi element(s) could affect one of myriad functions that rely on Golgi activity.

Referees cross-commenting

This is more of a technical report that does not address the function of the peptide within the Golgi complex. Without this information, and identification of the compartments involved, I don't see the advantage of the probe compared to other methods. As one reviewer mentioned, this seems to be a preliminary study that is difficult to assess given the limited and ambiguous results.

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 authors note that we currently lack a robust targeting signal to direct proteins to the cytoplasmic face of Golgi membranes. The presented work clearly identifies a novel Golgi targeting sequence rich in aromatic/hydrophobic/basic residues and with a key critical cysteine (C11). One can imagine a situation where the non-cysteine residues provide an underlying affinity for cell membranes and thereby allow access to membrane-associated zDHHC S-acyltransferases. I guess the unknown question is whether Golgi specificity comes from the amino acid sequence per se (mediating specific interaction with components of Golgi membranes) or instead by specific recognition of the cysteine by Golgi-localised zDHHC enzymes. It might be worth discussing this in the paper although this should not detract from the main focus/message of the paper- the identification of a Golgi targeting peptide. Data is compelling and support the conclusions of the paper. Although much of the data is not quantified, the data provided is convincing.

Significance

Interesting advance for researchers in the general membrane trafficking area and S-acylation field. Provides new information that can be used to target proteins of interest to the Golgi. I note that restriction of an S-acylated peptide at the Golgi is unusual as S-acylation is usually followed by trafficking to the plasma membrane. My expertise is in S-acylation and protein trafficking

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>

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>

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

1. General Statements

*The reviewers are enthusiastic. They agree with the claims made and comment favorably with regard to the impact as well on the short- and long-term potential for translation. All three go out of their way to emphasize positive aspects. A variety of questions were raised and we submit a complete revision with point-by-point replies that addresses all of these. This includes addressing tumor organoid (tumoroid) plasticity (reviewer #1) and composition/heterogeneity (reviewer #3) by incorporating single cell data as well as other analyses. We thank the reviewers for the thorough feedback. The additional data, analyses and clarifications strengthen the study. *

To keep the rebuttal as short as possible we have only copied the reviewers’ concerns/questions, not the favorable comments. The copied remarks are in highlighted. Our replies are in italics. Each question is accompanied by a reply and a brief description of changes made in response.

2. Point-by-point description of the revisions

__Reviewer #1, Major Comment #1: “__The authors provided a foundational validation of their organoids through various methods, and their protocol stands to impact the field of RMS biology. To validate the organoids as recapitulating the primary human tumors, the authors perform analysis on the bulk organoid and bulk human primary tumors. The authors showed through sequencing efforts that the bulk mutational profile and transcriptional profiles do not dramatically change between the parent tumor and organoids. This analysis was done well; however, the authors fail to rigorously illustrate that the organoids maintain tumor cell heterogeneity of the primary human tumors. To rigorously validate the organoid system, the authors should illustrate the organoid culture conditions do not alter the heterogeneity of cells (cell plasticity) compared to that of the primary tumor. A formal assessment of the cellular plasticity in the organoids to the primary tumor would determine how the organoid system either maintains or shifts the cancer cell plasticity because of changes in microenvironment (Oncogene, 2020, 39: 2055-2068). The addition scRNA-seq would illustrate whether the organoids maintain the same populations as the primary tumor or bias for the propagation of specific cell populations at a single cell level and provide more rigorous information about every cell type present.”

Reply: The reviewer’s question is whether the tumor cells in the tumoroid culture have the same degree of plasticity and are therefore as heterogeneous in culture as they are in the tumors that they are derived from. We agree that evaluating the heterogeneity of tumor cells in the tumoroid culture is desirable. This would ensure that the procedure has not simply selected for a single type of tumor cell. We have therefore generated single-cell RNA sequencing (scRNA-seq) data of tumoroid cells as suggested. It is important to point out that a complete inventory of RMS tumor cell heterogeneity by scRNA-seq has not been published as yet. Such an undertaking, i.e., scRNA-seq of a large cohort of RMS tumors, is an entire study in itself and lies outside the scope of this study. It would also not be feasible due to limited sample material for many of the tumors used here. Nevertheless, as is being alluded to by the reviewer, there is ample evidence of tumor cell heterogeneity in primary RMS tumors based on previous studies using immunohistochemistry (for example the well-known heterogeneity in expression of RMS marker proteins such as Myogenin, MyoD1 and Desmin). As shown in new Fig. 2D, when cultured as tumoroid models, examples from both of the main tumor types (FP-RMS sample RMS127 and FN-RMS sample RMS444) show a large degree of heterogeneity in expression of the known, heterogeneously expressed tumor cell markers Myogenin (MYOG gene), MyoD1 (MYOD1 gene) and Desmin (DES gene). Comparison with the cell cycle marker Ki-67 (MKI67 gene) shows that this heterogeneity is not due the cells being present in different cell cycle phases. Tumor cell heterogeneity in the tumoroid culture is further indicated by the heterogeneous CNV patterns derived from the tumoroid scRNA-seq data (new Suppl. Fig. 1B).

Both the CNV analysis and the scRNA-seq marker gene expression indicate that the tumoroid culture conditions neither stringently select for a single type of tumor cell, nor drive the tumor cells into a uniform expression pattern phenotype, consistent with maintaining plasticity, even after the 7 (RMS127) and 5 (RMS444) passages. These are good indications of retained plasticity/heterogeneity. Additionally, we make it clear in the revised version that a more exhaustive answer would benefit from having a complete cohort of tumor scRNA-seq data to first determine the degree of heterogeneity exhibited by RMS tumors for all genes.

The related question of tumoroid cellular composition, with regard to the presence of non-tumor cells, is addressed in response to reviewer #3, major comment #1.

Changes: Addition of a new Fig. 2D and a new Suppl. Fig. 1B with figure captions. Additional text in the Results and the Discussion sections. Additions to the Methods for the generation and analysis of the scRNA-seq data.

Reviewer #1, Major Comment #2: “The authors took great strides to show that the organoids respond to therapeutics similarly to primary tumors. However, Figure 5A could be more transparent with more data labelled in the graph instead of just in the app and the implications of the variable responses could have been explored in the discussion section. Furthermore, for this model to be clinically relevant for pharmacokinetic studies, propagation in mice needs to be shown.”

Reply:

• We have made Fig. 5A more transparent by adding the drug names.
• The different response between FN-RMS and FP-RMS subtypes for certain drugs is known and the implication that the models reflect this is discussed more thoroughly now as suggested.
• We agree that animal experiments are imperative for pharmacokinetic studies of new drugs. However, most of the drugs that were included here, especially the ones highlighted, have already been evaluated in early phase clinical trials in adults and/or children. The pharmacokinetic data for humans is therefore already available for these drugs, making additional animal studies for pharmacokinetics of these drugs redundant. For future studies, various types of animal studies are likely to be required and we make this clear in the Discussion, also emphasizing that in general, tumoroid models do propagate in mice. To address this specifically for RMS, we have started a collaboration to generate PDX mouse models derived from RMS tumor and tumoroid samples in parallel. Anecdotally we can state here that at least 50% propagate. However, since we wish to investigate this systematically and with a complete set of tumoroid models, it is not prudent to wait for these results before publishing the current study. This would delay making the protocols, findings and tumoroid models available to the scientific community and as our (and many other groups’) work exemplifies, tumoroid models can yield important findings on their own. Changes: Drug names added to rows in Fig. 5A. The Discussion has been expanded to include the differential response of tumor subtypes and tumoroids to different drugs and to include the uses (including pharmacokinetics) of different types of models has been expanded.

Reviewer #1, Major Comment #3: “Figure 1 is well put together to graphically demonstrate the process by which organoids were obtained and manipulated. Figure 1B, however, as a graphical summary is a little confusing, and the information would be greatly enhanced by the addition of a comprehensive table. Furthermore, additional information could be added to the table to make it a more inclusive and impactful addition to the paper.”

Reply: We agree.

Changes: A new Table 1 has been added as a separate file with a corresponding revised legend in the main document.

Reviewer #1, Major Comment #4: “It is quite impactful that the authors were able to actively engineer the organoids with CRISPR/Cas9 and accurately delete TRP53, but controls were not represented in the figure. The experiment should have included a sgRNA targeting a pan-essential gene as a positive control and a non-targeting sgRNA as a negative control. We recommend addition of both controls to the experiment outlined in Figure 6 to increase the validity and rigor of the data presented.”

Reply:* We respectfully note that all appropriate controls were done. This included a non-targeting sgRNA as negative control (see Methods lines 1137 to 1140). As also explained in Fig. 6A, the strategy for generating a P53 knock-out involved selection through nutlin-3 exposure, whereby cells wildtype for P53 are selected against. As described (Methods lines 1144 to 1146), cells transfected with the non-targeting sgRNA plasmid indeed died upon nutlin-3 exposure. A sgRNA against a pan-essential gene was not included in this strategy since the nutlin-3 already kills all cells with a wildtype P53. Finally, we draw attention to the fact that the success of the approach was assessed by Western Blotting (Fig. 6B) and Sanger sequencing (Suppl. Fig. 6A). *

Changes: None.

Reviewer #1, Major Comment #5: “Although the authors provide an insight into a useful preclinical RMS model, the paper lacks mechanistic insight besides cursory description of the model.”

Reply: Insight into a wide variety of different molecular and cellular mechanisms will be exciting to explore in future studies. This publication is indeed focused on describing an approach that works for RMS, and therefore showing for the first time that this works systematically for mesenchymal-derived tumors. In addition, the study describes key characteristics of the tumoroid models that are important to establish their validity as models and that are essential to demonstrate before making the tumoroid models available to the wider scientific community in order to perform the further mechanistic analyses. The word cursory is in contrast to the many positive comments made by this reviewer and the other two reviewers with regard to the extensive characterization.

Changes: None.

Reviewer #1, Minor Comment #1: “Figure 3C and 4B are not transparent in their labels and could be altered so that every line has an associated gene in the publication. Furthermore, there are sample specific differences that could be explored in the discussion.”

Reply: We agree.

Changes: Gene names have been added for every row in both figures. The Discussion now incorporates the observed differences.

Reviewer #1, Minor Comment #2: “In Supplementary Figure 1, higher magnification inserts are needed to get a closer look at the IHC. Furthermore, the white balance is not the same in all the images and needs to be corrected prior to publication. The difference in white balance can clearly be seen in the last panels depicting IHC for RMS335, where the MYOD1 staining has a yellow background whereas the H&E staining has a white background.”

Reply: We agree.

Changes: Higher magnification inserts have now been provided throughout Suppl. Fig. 1A. The white balance has been corrected.

Reviewer #1, Minor Comment #3____: “The authors mentioned in line 202 that some of their organoids contain the novel fusion of PAX3 and WWTR1, but this fusion is not indeed novel as it has previously been seen in biphenotypic sinonasal sarcoma (Am J Surg Pathol 2019, 43:747-754).”

Reply: We rephrased this to clarify that this is the first report of such a fusion in RMS, rather than in general.

Changes: The corresponding sentence has been rephrased.

Reviewer #1, Comment within the Significance Statement: “The authors state that this is the first system to use organoids but should recognize the advances demonstrated by Manzella et al. (Nat Commun, 2020, 11:4629). Additionally, the authors state that this is the first demonstration of pre-clinical models harboring FGFR4V550L mutations; this fails to recognize the prior reported work by several groups (Chen et al., Cancer Cell, 2013, 24:710-24; Manzella et al., Nat Commun, 2020, 11:4629; McKinnon et al., Oncogene, 2018, 37:2630-2644).”

Reply:* We had in fact already recognized the advances described by Manzella et al. which was referenced in two places in the original submission (current lines 100 and 388). We thank the reviewer for pointing out the previous work done on an RMS cell line that harbored an FGFR4 p.V550L mutation. *

Changes: We rephrased the corresponding passages concerning the FGFR4 mutation.

We thank reviewer #1 for all the comments. This has resulted in many improvements.

Reviewer #2: W____e thank reviewer #2 for the positive comments. There are no major/minor queries to address.

Reviewer #3, Major Comment #1: “The authors describe the models derived as organoids/tumoroids implying that multiple cell types are represented potentially recreating the tumor microenvironment. Can the authors comment more specifically and demonstrate the extent to which cell types in addition to the tumor cells are represented, viable and are organized through analyses of the original and tumoroid sections (extend fig 2C/supplementary fig) and via analyses of the RNAseq data?”

Reply: We use the term tumor organoid or tumoroids as coined by the field in general. This indeed indicates a degree of self-organization such as the three-dimensional growth in spheres and the propagation of a heterogeneous population of tumor cells (see comment #1, reviewer 1) for example. In general, however, tumoroids do not include growth of a non-tumor cell microenvironment inter-woven with the (different types of) tumor cells. Exceptions to this are very early passage tumoroids that are not yet stable and which may still contain non-tumor cells, or specialized co-culture conditions that are currently being actively sought to allow for co-culture of tumor cells within a non tumor cell microenvironment. It is therefore not anticipated that late passage tumoroid models will have non-tumor cells. The basis of the technology is that the defined set of growth factors in the medium mimics the tumor stimulating conditions of the non-tumor cell microenvironment. Since the mixed presence of tumor and non-tumor cells generally gives rise to one (frequently the non-tumour cell) outgrowing the other, it is often considered the hallmark of an unsuccessful tumoroid.

The reviewer therefore raises an important point that we have failed to make clear. We have addressed this in two ways. We emphasize that the scRNA-seq data that are now included in response to reviewer #1, comment #1 do not indicate the presence of any non-tumor cells (as expected). In addition, this aspect of tumor organoid technology is explained better in the Introduction.

Changes: The results section has been expanded with the description of the scRNA-seq data emphasizing the expected lack of non-tumor cells and the introductory section on tumor organoid technology has been improved to make it clear that currently this generally involves growth of different types of tumor cells only.

Reviewer #3, Major Comment #2: “Does the quantification from the RT-qPCR analyses for the MYOD1, MYOG and Desmin of the models match that in the samples from which they were derived? Does the RNAseq that was performed on tumor and the culture at the time of the drug screen tie in with this?”

Reply: The answer is yes. The figure below shows tumor and tumoroid bulk RNA seq of those genes also analyzed by RT-qPCR (i.e., DES, MYOG, and MYOD1). Note that this is also the same stage as for the drug screening. As can be appreciated, the expression of these markers is generally very comparable between tumors and the derived tumoroid models. Note that this also constitutes a nice independent (albeit indirect) verification of the similar degree of heterogeneity issue raised by Reviewer #1 (comment #1). Expression of the markers was lower in the tumoroid models of RMS000HQC and RMS000ETY compared to the primary tumor. In line with this, expression of these genes was also already lower in the early passages of the culture as determined by RT-qPCR (Fig. 2A). Nevertheless, copy-number analysis inferred from whole-genome sequencing showed that the resulting tumoroid models are indeed tumor cells (Suppl. Fig. 2A top panel and Suppl. Fig. 2B lower panel).

We therefore conclude that the expression of probed marker genes is generally comparable between tumor and tumoroid and that early passage RT-qPCR based expression analysis of these markers can be reflective of the expression in the fully established model.

*- Rebuttal letter includes corresponding figure here - *

Changes: None. The expression data are already available within the interactive browser-based Shiny App.

Reviewer #3, Major Comment #3: “How do the frequencies of SNVs compare with recent studies? Or are the numbers in the risk groups not appropriately represented?”

Reply: The SNV frequencies are quite comparable to recent studies, with similar differences between risk groups, all as depicted in the new Suppl. Fig 2E. The SNV frequency was calculated from our WGS data following the procedure from the most recent report in pediatric cancer (https://www.biorxiv.org/content/10.1101/2021.09.28.462210v1). Across tumor and tumoroid models we found a somatic mutation frequency of SNVs with a VAF of >0.3 ranging from 0.03 to 1.92 mut/MB (median 0.70 mut/MB) which is comparable to the reported somatic mutation frequency in the afore-mentioned study (median 0.9 mut/MB in RMS). Concerning the risk groups, a recent study (https://pubmed.ncbi.nlm.nih.gov/31699828/) found a significant difference in the tumor mutational burden between fusion-negative (FN) and fusion-positive (FP) RMS (2.6 mut/MB vs. 1.0 mut/MB, respectively) with a higher mutational burden associated with poorer outcome. In our study, the FN-RMS tumoroid models also show a higher mutation frequency compared to the FP-RMS tumoroid models (FN 4 vs. FP 15, p = 0.02, Wilcoxon). Such a difference is also found between the original tumors but without statistical difference (FN 4 vs. FP 15, p = 0.15, Wilcoxon) likely related to the small sample sizes. This underscores the representative nature of the tumoroid models and is of obvious interest to include. We have made the appropriate changes.

Changes: To include these analyses in the manuscript, we added a new Suppl. Fig. 2E with corresponding Suppl. Fig. legend and a new paragraph in the main text.

Reviewer #3, Minor Comment #1: “The number of models and success rates would be useful to indicate in the abstract.”

Reply: We agree.

Changes: We added this information to the abstract.

Reviewer #3, Minor Comment #2: “It would be helpful to define the SBS1, 5,and 18 in the figure legends. Do the age related signatures in any way correlate with patient age or aggressivness of tumors?”

Reply:

• Agreed. The definitions of SBS1, 5, and 18 have now been included the legends of Fig. 3B and 4A.
• The age-related signatures SBS1 (but not SBS5) shows a weak albeit significant correlation with patient age only in RMS tumoroid models but not in RMS tumors. Furthermore, concerning aggressiveness, FP-RMS tumors and tumoroid models show a significantly higher contribution of SBS1 (but not SBS5) to their overall somatic mutation frequency compared to FN-RMS tumors and tumoroid models. However, since FP-RMS tumor samples were obtained from older patients (median 14 years versus median 6 years in FN-RMS tumor samples), this observation could also be related to the patient-age and not primarily to the fusion-status. The heterogeneity of samples (e.g., primary therapy-naïve samples versus relapse and therefore pre-treated samples) and the relatively low sample number could be explanations for the lack of a stronger correlation in general. Changes: Added definitions of SBS1, 5, and 18 in the legends of Fig. 3B and 4A. Added text in the Results section to indicate the observed correlations.

Reviewer #3, Minor Comment #3: “Page 13 line 300 just because the RH30 cell line has TP53 mutation doesn't mean that it was acquired in culture - unless there is specific evidence that supports this.”

Reply:* We thank the reviewer for this rectification. To our knowledge, there is indeed no specific evidence that this cell line acquired the TP53 mutation during culturing or whether the mutation was already present in the primary tumor the cell line was derived from. *

Changes: The corresponding statement has been removed.

We thank reviewer #3 for all the comments. This has resulted in many improvements.

Besides the changes described above, additional minor changes were made:

*We have moved the interactive, browser-based Shiny app to a server that is managed by our institute instead of having it hosted on shinyapps.io. We include the new URL in line 556. *

The data upload to the European Genome-Phenome Archive (EGA) of the data from the initial submission has been completed and the raw sequencing data can now be accessed. The data upload of the scRNA-seq data generated for the revision is currently ongoing. We have therefore adapted and renamed the “Bulk sequencing data availability” section in the Methods in the manuscript (lines 1043 to 1050).

We updated the code available at https://github.com/teresouza/rms2018-009* following the additional analyses performed for the revision. *

Supplementary Table 1: The values for row “RMS000FLV” for columns “sample_body_site” and “primary_site_specific” were corrected as this tumor was located in the upper leg and not the upper arm of the patient. Furthermore, we added patient numbers as in the new Table 1 and corrected spelling errors. This does not change any of the conclusions in the manuscript.

Figure 6A: The protein “P53” was spelled without capital “P” in the initial version. We corrected this.

We included the recently described Zebrafish RMS PDX models (https://pubmed.ncbi.nlm.nih.gov/31031007/) in the Discussion of RMS models. See lines 507 to 510.

With the addition of Fig. 2D, the figure legends of Fig. 2A and 2B were moved to the side (Fig. 2A) or below (Fig. 2B) the figure. With the addition of the single-cell copy-number plots, Suppl. Fig. 1 was divided in Suppl. Fig. 1A and 1B.

Some of the original scale bars in Fig. 2C and Suppl. Fig. 1A were incorrectly labelled and this has now been corrected. This does not change any of the conclusions.

Minor corrections in the sections Affiliations, Financial support, Author contributions and Conflict of Interests.

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

Meister et al., describe their methodology in establishing what they term organoid or tumoroid 2D/3D cultures derived from samples of rhabdomyosarcoma (RMS) patient tumours. They have success to varying degrees across the subtypes with greater success in those more clinically aggressive. Their analyses of markers, the somatic genetics and gene expression profiles suggest that they are largely representative of RMS and the tumor samples from which they were derived. Their utility in drug screening and manipulation by knocking out TP53 by CRISP/Cas9 is also demonstrated. The conclusion is that this represents a useful approach for generating patient derived models and a unique resource for preclinical analyses and other research into RMS.

This is a major piece of work that is well written and presented. The link to interrogate the data worked. I have only a few comments.

Major comments

The authors describe the models derived as organoids/tumoroids implying that multiple cell types are represented potentially recreating the tumor microenvironment. Can the authors comment more specifically and demonstrate the extent to which cell types in addition to the tumor cells are represented, viable and are organized through analyses of the original and tumoroid sections (extend fig 2C/supplementary fig) and via analyses of the RNAseq data?

Does the quantification from the RT-qPCR analyses for the MYOD1, MYOG and Desmin of the models match that in the samples from which they were derived? Does the RNAseq that was performed on tumour and the culture at the time of the drug screen tie in with this?

How do the frequencies of SNVs compare with recent studies? Or are the numbers in the risk groups not appropriately represented?

Minor comments

The number of models and success rates would be useful to indicate in the abstract.

It would be helpful to define the SBS1, 5,and 18 in the figure legends. Do the age related signatures in any way correlate with patient age or aggressivness of tumors?

Page 13 line 300 just because the RH30 cell line has TP53 mutation doesn't mean that it was acquired in culture - unless there is specific evidence that supports this.

Significance

The significance of this study is in describing how a relatively large number of models of RMS were established plus increasing awareness of the biobank resource and associated data that has been created. The approach, although used in more ad hoc reports of smaller numbers of RMS, represents a useful development for mesenchymal tumors versus the more established development of such models in epithelial cancers. Although a lower success rate than xenografts, it is a useful and practical cost-effective alternative for preclinical testing and research. Likely interest to a speciaclist audience for those involved in the RMS, sarcoma and pediatric cancer field.

Referees cross-commenting

OK with the balance of comments for the authors to address. I think the extent to which they are prepared to address the heterogeneity issue, and the results of this for the models, is likely to affect the impact of their 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

This manuscript describes the possibility to generate a collection of pediatric rhabdomyosarcoma (RMS) tumor organoid models comprising broad spectrum subtypes from highly aggressive to extremely rare. The authors were able very successfully establish 19 RMS models from 46 pediatric RMS patient samples with 41 % efficiency. All RMS tumoroid models were thoroughly characterized and retained the molecular characteristics of the tumor they are derived from as well as they displayed genetic stability over time. Most of the tested tumors showed long-term propagation potential, reaching passage 40 and remaining stable. Though, establishing time for RMS tumoroid models varied with a median time from acquisition of the tumor sample to successful drug screening being 81 days, highly aggressive tumors were established in as little as 27 days. Also, authors shown us in elegant manner the suitability of RMS tumoroid models for research in two specific ways: via drug screening and CRISPER/Cas9 genome editing.

Significance

In summary, the author's work made significant progress in 3D culture and tumor organoid models of mesenchymal origin, being the first collection of tumoroid models from mesenchymal malignant tumors and the second thoroughly characterized tumoroid collection specific for pediatric cancers. Without doubt, biobanked collection of RMS tumoroids will be useful for drug screening as well as molecular editing. Also, these models will be a useful resource for future research and in preclinical and clinical testing therapeutics for RMS. In the future, organoids generated from patients with RMS may lead to precise and personalized treatment.

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:

Meister et al. set out to develop a new organoid preclinical model of rhabdomyosarcoma (RMS). The authors comment that this system would be beneficial for preclinical modeling because it has the ability to maintain the tumor's molecular characteristics. The authors then proved that organoids derived from multiple RMS subtypes resembled their parent tumors using RT-qPCR for characteristic markers, histopathology, copy number profiles, mutational signature analyses, and transcriptional profiling. Importantly, the authors performed long term studies to show that the organoids remain stable over multiple passages and do not change their mutational landscape dramatically. Finally, the authors tested their organoids with known RMS therapeutics and for their ability to be engineered with the CRISPR/Cas9 system. Not surprisingly, the authors found their organoids sensitive to known RMS therapeutics and were successfully able to generate TP53-/- organoids with CRISPR/Cas9, underscoring this organoid system in translatable use. This report nicely describes a method for the establishment of human RMS organoid culture systems that can be leveraged for preclinical testing.

Major Comments:

1. The authors provided a foundational validation of their organoids through various methods, and their protocol stands to impact the field of RMS biology. To validate the organoids as recapitulating the primary human tumors, the authors perform analysis on the bulk organoid and bulk human primary tumors. The authors showed through sequencing efforts that the bulk mutational profile and transcriptional profiles do not dramatically change between the parent tumor and organoids. This analysis was done well; however, the authors fail to rigorously illustrate that the organoids maintain tumor cell heterogeneity of the primary human tumors. To rigorously validate the organoid system, the authors should illustrate the organoid culture conditions do not alter the heterogeneity of cells (cell plasticity) compared to that of the primary tumor. A formal assessment of the cellular plasticity in the organoids to the primary tumor would determine how the organoid system either maintains or shifts the cancer cell plasticity because of changes in microenvironment (Oncogene, 2020, 39: 2055-2068). The addition scRNA-seq would illustrate whether the organoids maintain the same populations as the primary tumor or bias for the propagation of specific cell populations at a single cell level and provide more rigorous information about every cell type present.
2. The authors took great strides to show that the organoids respond to therapeutics similarly to primary tumors. However, Figure 5A could be more transparent with more data labelled in the graph instead of just in the app and the implications of the variable responses could have been explored in the discussion section. Furthermore, for this model to be clinically relevant for pharmacokinetic studies, propagation in mice needs to be shown.
3. Figure 1 is well put together to graphically demonstrate the process by which organoids were obtained and manipulated. Figure 1B, however, as a graphical summary is a little confusing, and the information would be greatly enhanced by the addition of a comprehensive table. Furthermore, additional information could be added to the table to make it a more inclusive and impactful addition to the paper.
4. It is quite impactful that the authors were able to actively engineer the organoids with CRISPR/Cas9 and accurately delete TRP53, but controls were not represented in the figure. The experiment should have included a sgRNA targeting a pan-essential gene as a positive control and a non-targeting sgRNA as a negative control. We recommend addition of both controls to the experiment outlined in Figure 6 to increase the validity and rigor of the data presented.
5. Although the authors provide an insight into a useful preclinical RMS model, the paper lacks mechanistic insight besides cursory description of the model.

Minor Comments

1. Figure 3C and 4B are not transparent in their labels and could be altered so that every line has an associated gene in the publication. Furthermore, there are sample specific differences that could be explored in the discussion.
2. In Supplementary Figure 1, higher magnification inserts are needed to get a closer look at the IHC. Furthermore, the white balance is not the same in all the images and needs to be corrected prior to publication. The difference in white balance can clearly be seen in the last panels depicting IHC for RMS335, where the MYOD1 staining has a yellow background whereas the H&E staining has a white background.
3. The authors mentioned in line 202 that some of their organoids contain the novel fusion of PAX3 and WWTR1, but this fusion is not indeed novel as it has previously been seen in biphenotypic sinonasal sarcoma (Am J Surg Pathol 2019, 43:747-754).

Significance

As has been mentioned previously, this research is impactful to the field of RMS biology because the authors were successfully able to use organoid technology, which has not previously been reported. The authors do a great job of listing current RMS modelling techniques and explaining how their system addresses the pitfalls of the others. Furthermore, this protocol could be expanded to the development of other organoid systems for other sarcomas. The rhabdomyosarcoma field and larger sarcoma community would be keenly interested in this work. It is clear that this system has the potential for use in pre-clinical settings as well as in high-throughput screens, but further validation and increased rigor is required on both fronts.

It is astounding and the authors should be complimented that they were able to show a median time from patient to drug screen was 81 days! This has enormous potential such as rapid translation of therapies and personalized medicine. That said, the authors must first refine the heterogeneity of the organoids and demonstrate how the organoids reflect the phenotypic and cellular plasticity of the parent tumors. Furthermore, the authors ought to be careful when making priority claims. The authors state that this is the first system to use organoids but should recognize the advances demonstrated by Manzella et al. (Nat Commun, 2020, 11:4629). Additionally, the authors state that this is the first demonstration of pre-clinical models harboring FGFR4V550L mutations; this fails to recognize the prior reported work by several groups (Chen et al., Cancer Cell, 2013, 24:710-24; Manzella et al., Nat Commun, 2020, 11:4629; McKinnon et al., Oncogene, 2018, 37:2630-2644).

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>

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>

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

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

In this manuscript by Wang and colleagues, the authors analyse single-cell RNA-seq (scRNAseq) data by applying transition path theory to infer gene regulatory network (GRN) changes along the transition (reaction coordinate, trajectory) between free energy stable states (i.e. cell types). The work aims to understand how stable cell types, and their regulatory programs (combination of active and repressed genes) switches during differentiation/reprogramming/response (i.e. cell phenotypic transition/CPT). The premise of the work is to assess whether genes within GRNs undergo step-wise repression, state-change and activation (& vice-versa; analogous to SN1) or concurrently regulate gene expression (analogous to SN2). The GRNs are inferred based on highly variable genes and their expression dynamics from RNA velocity over CPT, across 3 scRNA-seq datasets.

The authors first analyse public scRNA-seq dataset of 3003 human A549 adenocarcinomic basal epithelial cells treated with TGF-b for 0hrs, 8hrs, 1 day and 3 days (4 timepoints). The authors select two stable states (Day0-untreated; Epithelial and Day 3-treatment; Mesenchymal) using local kernel densities and set transition paths using Dijkstra shortest path, dividing state space into Voronoi cells (i.e. reaction coordinate value), and constructed single-cell GRNs based on RNA velocity differences (n=500 genes) and a linear model (from Qiu et al). This GRN is based on expression and velocity estimates, and does not distinguish direct from indirect regulation. Calculating interaction frequency (edges) across two stable states over 4 louvain clusters, the authors find global increase in effective edges that correlates with increased active genes; but with variable trend within inter-cluster edges. To quantify the concerted GRN changes between clusters, the authors utilise a "frustration" score (from Tripathi et al 2020). The average frustration score increases and peaks at day 1 treatment, followed by a decline over terminal stable state (day 3-treatment); similar to interaction frequency trends. The author also separately measure network heterogeneity and repeat analysis using alternative transition matrix. The authors conclude that EMT proceeds through concerted regulation of multiple genes first with an increase in inter-cluster edges, frustration and heterogeneity followed by a decrease into final stable state. The authors apply the analysis to scRNA-seq data from (i) pancreatic endocrine differentiation where Ngn3-low progenitors give rise to Ngn3-high, then Fev-high and into glucagon producing a-endocrine cells; (ii) dendate gyrus; radial glial cell differentiation into nIPCs, neuroblast, immature granule and mature granule cells. In both cases, the authors observe concerted regulation with initial increase in inter-community edges, heterogeneity during differentiation followed by decrease towards final stable state. **

The study and ideas in the manuscript are interesting and the methods would be potentially be useful. However, there are a few specific and general comments stated below, which the authors should try to address.

1 • P4: "RC increases first and reaches a peak when cells were treated with TGF-β for about one day, then decreases (Fig. 1G)". It would be better to label the figure with the treatment information. *

Reply: Thanks for your advice. In the revised manuscript, we analyzed two additional datasets, and moved the EMT result in the supplemental Fig. EV8. In the new Fig. 1d, we marked the cell types along the reaction coordinate.

*2 • Fig. 1G and EV1D: Why are the trends different? *

Reply: In the original figures, ____Fig____.1g is the frustration score and EV1D shows the variation of pseudo-Hamiltonian along the reaction coordinate. The frustration score is the focus of this work. We also calculated the pseudo-Hamiltonian since it has been used in the literature. However, we realized that showing both of the results might lead to confusion, so we deleted all pseudo-Hamiltonian results in the revised manuscript.

* 3 • How is the appropriate community/cluster/Louvain resolution selected? This can have a major impact on number of cell states, types and transition path from initial to final state. *

Reply: The number of cell states, types and transition path from initial to final state____ are not determined from the community/cluster/Louvain analyses. For the EMT data, we assume most cells in the initial treatment time are epithelial cells, and those in the final time point are mesenchymal cells. For other datasets, we followed the original publications to assign cell types based on known marker expression.

The Louvain method was applied to coarse grain the gene regulation network, and it does not affect the number of cell states, types and transition path, which were determined separately. To address the reviewer’s question, we also use the Leiden method to adjust the resolution ____(1)____. The resolution does not affect the result. The results are added to Fig. EV12. We tried three different resolution values 0.8,1.0 and 1.2. The number of inter-community edges consistently shows the trend that it increases first then decreases.

Figure EV12 Cell-specific variation of the number of effective inter-community edges between communities calculated with different resolution parameter values for dentate gyrus neurogenesis (a), pancreatic endocrinogenesis (b), and bone marrow marrow hematopoiesis (c). Each dot represents a cell and the color represents the number of inter-community edges____.

• * What effect does the Louvain resolution have on e.g. frustration scores? * Reply: The resolution of community division algorithm doesn’t affect the frustration scores, since the frustration score is based on the gene-gene interactions instead of community assignment.

• * The authors match resolution to samples/timepoints/known prior cell types i.e. 3-4 communities. However it is unclear whether this is enough to describe entire differentiation/transition process. * Reply: This is a good question. In one above reply we have explained how the cell types were determined____. We also agree with the reviewer that these coarse-grained communities cannot reflect the overall heterogeneity and dynamics of the whole process. Notice in most of our analyses (e.g., reaction coordinate and transition paths), we treated the transition as continuous and the distribution of single cell data points in all datasets cover the whole space that involved in cell phenotype transition. The coarse-grained analyses are for further mechanistic insights on how gene regulatory networks are reorganized during the transition process.

• * Gene selection: The selection based on minimum 20 counts as highly expressed genes is arbitrary and dependent on sequencing depth. Perhaps the authors could show distribution of gene counts for the datasets and have a data-driven filtering criteria * Reply: Thanks for the advice. The number 20 is a default value suggested in the package (scVelo) we use, and in another package dynamo the default number is 30. Following the reviewer’s suggestion (together with the next question on the influence of all highly variable genes), we looked for a data-drive filtering criterion. The method has been described in different tools ____(2-4)____. We first grouped the genes into 20 bins by their mean expression values, and____ scaled their dispersions by subtracting the mean of dispersions and dividing standard deviation of dispersions____. Figure EV9 shows the distribution of the minimum shared counts. ____As one can see, most genes counts are larger than 10, and using a smaller value causes error in the following velocity analysis. Therefore we set the minimum shared counts as 10 in the new results.

Figure EV9 Shared counts distribution of the datasets. (a) Dentate gyrus neurogenesis; (b) Pancreatic endocrinogenesis; (c) Bone marrow hematopoiesis.

• * The choice of 500 variable genes (for human A549 cells) is also quite arbitrary. Perhaps, the authors could compare how additional genes (all highly variable genes) affects their analysis and interpretation. * Reply: ____Thanks. Following previous question on shared counts and ____data-driven filtering criteria____,____ we take all the highly variable genes into consideration. The details of gene selection and binarization are given in the Materialss and Methods (Materials and Methods 2) section.

• * How are other factors (sequencing depth, genes detected, #of cell types, multiple branches) affects the connectivity between communities at different phases of transition/development? * Reply: This is a good question. The A549 EMT dataset has a sequence depth of 40000-50000. The ____dentate gyrus neurogenesis dataset____ has a sequence depth of 56,700 reads. A saturation depth would be close to 1,000,000, but there is a compromise between cell number and depth. There are genes that are not detected even under the saturation reads setting. That is why the preprocessing is needed. On the other hand, the network we inferred include both direct and indirect interaction, so the influence of sequence depth and gene number detected can be reduced to a certain extent. We used a random subset of the selected gene and performed the same analyses. The results are consistent with what we obtained using all the genes (Fig. EV11b). With the new gene selection criteria (Materials and Method 2), our analyses are not related with the number of cell types.

We did analysis on another beta branch of pancreatic endocrinogenesis data. The other branches show the same results (Fig. EV4). There are two additional branches in the pancreatic endocrinogenesis dataset. It has been reported that the RNA velocity estimation for the epsilon branch is incorrect ____(3)____. There are too few cells in the delta branch for reliable analyses. Therefore we didn’t present results for these two branches.

Figure EV4 Analyses on the branch of glucagon producing β-cells in pancreatic endocrinogenesis.

(a) Transition graph based on RNA velocity.

(b) The RCs and corresponding Voronoi cells. The large colored dots represent the RC points (start from blue and ends in red). The small dots represent cells with color as cell type.

(c) Frustration score along the RCs.

(d) Cell-specific variation of effective intercommunity regulation. Each dot represents a cell. Color represents the number of effective intercommunity edges within each cell in the GRN.

• • Are the velocity graph, transition matrix and further shortest path estimation derived in a reduced latent space, and if so, how much (nPCs) and what impact does it have. Presumably, the density estimation is not performed in expression space. Reply: Yes. ____The calculation of transition matrix is based on neighbor information. The calculation of neighbors was in the reduced latent space in scVelo and Dynamo. We performed the same analysis by varying number of principal components. The results are similar because the first several components account for large proportion of variance. Figure R1 shows the results of dentate gyrus neurogenesis with the number of principal components being 10, 20 and 30, respectively. In the revised manuscript, we delete the step of using density estimation constrain to simplify the procedure. __Figure R1 Frustration scorer along RCs (left) and cell specific variation of number of effective intercommunity edges (Each dot represents a cell and color represents the number of effective intercommunity edges) in the GRN within each cell (right) when using different number of PCs in analyses (dentate gyrus neurogenesis): (a) number of PCs is 10.*__

(b) number of PCs is 20. (c) number of PCs is 30

* - The figure legends and labels were hard to read. These should be improved for better readability. *

Reply: Thanks. We modified the figure legends and labels.

* - A suggestion would be move the initial results section to methods and highlight the biological interpretation. *

Reply: Thanks for your advice. We moved large part of this section to the Materials and Methods.

*The authors could highly which GRN and representative genes/edge pairs are highest ranked within inter-community and to overall final stable states. *

Reply: Thanks. We list some representative gene pairs in the Table. EV 2&EV 3 &EV 4 for different datasets. And we performed gene enrichment analysis for each community.

* - How does the GRN inference compare to current state-of-the-art GRN inference scRNA-seq methods? *

Reply: we used the method GRISLI to perform the same analysis ____(5)____. The results are similar to what obtained with our current method (Figure EV6). We want to emphasize that the focus of this work is not on another GRN inference method, but discussing some general principles of GRN reorganization during a cell phenotypic transition process.

Figure EV6 Analyses of datasets of dentate gyrus neurogenesis (a), pancreatic endocrinogenesis (b), and hematopoiesis (c) based on GRN inferred with GRISLI.

(a) Frustration score along the RCs of dentate gyrus neurogenesis (left) and cell-specific variation of the number of inter-community edges (right). Each dot represents a cell and color represents the number of inter-community edges in GRN within each cell.

(b) Same as in panel (a), except for pancreatic endocrinogenesis.

(c) Same as in panel (a), except for hematopoiesis.

* - How do extremely noisy/stochastic genes vary in metrics between final stable states? How are the metrics affected by number of cells and stochasticity of expression within a given cluster/community. *

Reply: To address this question, we selected two genes, Id2 and Cdkn1c, with high variance and compare their distributions in the initial and final states. ____The gene distributions show significant shift between the Ngn3 low EP cells and Alpha cells (Fig. R2 a &b left).____ Then we randomly selected a subset (half) of cells and compared the distributions of these high-variance genes in the sub-population (Fig. R2 a&b right). The results are similar to the full-set results.

Fig. R2 Comparison of gene distribution in the initial and final states in pancreatic endocrinogenesis. (a) Comparison of the distribution of gene Id2 at the initial and final states (left), and in the randomly selected sub-population at the initial and final states (right). (b) Comparison of the distribution of Cdkn1c at the initial and final states (left), and in the randomly selected sub-population at the initial and final states (right).

* - Given that the author's approach includes both direct and indirect genes effects, the authors could further prune genes based on existing TF databases or protein-protein validated networks. *Reply: This is a good suggestion. We will work on this idea in future work. As we mentioned, due to constrains of data quality, only tens of transcription factors can be analyzed in these dataset. We list some regulations of transcription factors inferred with current method in Table EV1.

• *It is unclear which GRNs are already known and which ones are novel and biologically relevant * Reply: We compare some regulations inferred with the method and compare these interactions w____ith some references in Table. EV1____.

* - It would be good for authors to comment when there are multiple bifurcations instead of A-B transitions. Particularly in datasets with multiple discrete stable states. *Reply: This is a good question.____ In our analysis, we focus on the transition from one stable state to another stable state. For transition process with multiple bifurcations like____ the pancreatic endocrinogenesis, the results are similar across different branches. For the transition that goes through multiple discrete stable states, for example, a transition from state A____à____B____à____C, we expect to observe two peaks in the frustration score and the number of inter-community edges. We added some discussions in the Discussion section.

• *Another suggestion would be to highlight gene expression of selected markers based on f-regression and mi over the trajectory * Reply: As we modified the criteria of gene selection, we plotted trajectories of some high-variance genes versus the reaction coordinate obtained with different datasets in Fig. EV10 based on current criteria.

Figure EV10 ____Typical trajectories of high variance genes versus RCs of dentate gyrus neurogenesis (a), pancreatic endocrinogenesis (b) and bone marrow ____hematopoiesis ____(c).

* - If possible, a proof of principle could be re-analysis of a perturbation scRNA-seq dataset (e.g. where one path/transition path is stalled) *

Reply: Thanks. This is a really a good suggestion. We will perform more systematic studies in future work.

* Reviewer #1 (Significance (Required)): Nature and significance of advance: The study and ideas in the manuscript are interesting and the methods would be potentially be useful to community. Compare to existing published knowledge: *

*Audience: Predominantly computational audience *

*Your Expertise: PI with background in experimental, computational biology and expertise in single-cell genomic tools and developmental biology *

*

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

Understanding the cellular and molecular basis of cell type or cell state transitions occurring during development or reprogramming is a fundamental challenge. scRNA-seq has provided a window into gene expression programs across thousands of cells undergoing such transitions. Wang and colleagues leverage scRNA-seq and develop an approach to reverse engineer gene regulatory network underlying cells along a path from one cell type/state to another, and characterize community-level properties of this network associated with various stages of the cell phenotype transition. The study is innovative and rigorous, and their results point to how intercommunity interactions increase and then decrease, indicating a concerted regulatory rewiring that orchestrates transitions. Application of their approach to three different datasets also shows that this trend is consistent across three different transitions and maybe a general trend. However, there are some major and minor concerns that need to be addressed.

**Major comments and questions**

1. The analogy to SN1 and SN2 mechanisms of chemical bond formation is very nice.
2. What is the basis for the two statements made in paragraph 3 of Introduction (beginning with "A question arises ...") about transitions being sequential or concurrent? Please *Reply: Thanks. We added references in this paragraph.

* 2.1. Provide references to previous experimental and computational studies that have investigated developmental and reprogramming gene expression programs. *

Reply: Thanks. We added a paragraph in the Introduction.

*

2.2. Describe specific examples of findings that support the two possible transitions highlighted here. Why couldn't transitions happen through an entirely gradual process involving changes to overlapping subsets of genes. *

Reply: Thanks. In the review paper of Naomi Moris et. al., they proposed the hypothesis that cell phenotype transition is similar to a chemical reaction ____(6)____. Thus we extrapolate this hypothesis and test it in our study. For the example of SN1 mechanism, ____Kalkan et al. showed that mouse embryonic stem cells can exit from ____naïve pluripotency____ but remain uncommitted ____(7)____.

Just like the SN1 and SN2 mechanisms are two extremes in chemical reactions and there are cases lie in between, for cell phenotypic transitions we agree with the reviewer that such gradual process may exist. Actually the result in Fig. EV4d shows that the frustration score remains flat for the Fev+ ____à____ Beta transition, suggesting a possible gradual process. With the analyses provided in this work, such as the reaction coordinate, frustration score, heterogeneity, and inter-/intra- community edges, one may perform more systematic studies on a larger number of datasets and enumerate/classify possible patterns of transitions.

• Please make plots of the number of effective intra-community edges vs. number of active genes to support the statement that these two numbers are correlated. *

Reply: We plotted the corresponding intra-community active genes and calculated its correlation coefficient with the number of effective intra-community edges in dentate gyrus neurogenesis (Fig. EV1d). ____The correlation coefficients are 0.91,0.96, 0.99 and 0.96 for community 0, 1, 2 and 3 separately.

* A bunch of notations are not clear:

4.1. What is the "r" in "strongest intercommunity interactions at r = 10 (Fig. 1F)"? Is it the same as the "r" mentioned in the Methods section? *

Reply: r____ is the index number of the discretized reaction coordinate. We added it when we define the reaction coordinate. We modified the conflict usage of r in Materials and Method 4.

4.2. What is "s_i" in "cell-specific effective matrix, Fbar_ij = (2*s_i - 1)*F_ij"? Also, that description of F_ij, f_ij, and H should be moved to the Methods section, and a more high-level, intuitive description should instead be included in this Results paragraph. Reply: represent the binarized gene expression state. is 0 for when gene is in low expression level (silence) and is 1 when gene is in high express level (active). We modified this part following your advice.

* How were the h_f and h_m thresholds chosen? *

Reply: and are based on the distribution of each dataset. Following suggestions from another reviewer, we modified this part. All the highly variable genes were selected and the genes were binarized with the Silverman’s bandwidth method and ____K____means (Materials and Methods 2).

* What is the "density of each single cell" ("⍴_t")? The formulation of the penalty of the distance between cells i and j (the expression with -logP_ij...) is unclear. What is the intuition behind it? What is r? How were the values of r (0.5 and 0.8) chosen? *

Reply: The probability density of cells in the expression space is based on the kernel density estimation. Intuitively, a region in the expression space with more cells is more likely passed by more cell trajectories. The values are based on the distribution of kernel density estimation in different datasets.

In the modified manuscript, we used trajectory simulation and deleted this assumption for simplification.

* One of the reasons the authors state to justify the choice of PLSR is "In the scRNA dataset, the number of genes is often comparable to or larger than the number of cells." This is not true most of the time. In nearly all recent studies, the number of cells is way larger than the number of genes measured. *

Reply: The PLSR method definitely can be used for the data whose number of cells is larger than the number of genes. Also the PLSR method was applied on cells that are the k nearest neighbors of each reaction coordinate, which are a subset of the whole dataset (Materials and Methods 5). While we mainly presented results with the PLSR method, in this revised manuscript we also added results with another method of GRISLI (Materials and Methods 9). The results are similar with what we obtained with PLSR.

* There is a fleeting reference to a nice previous finding that supports their observations: "several lines of evidence support that EMT proceeds through a concerted mechanism. Indeed, both in vivo and in vitro studies have identified intermediate states of EMT that have co-expressed epithelial and mesenchymal genes (Pastushenko et al, 2018; Zhang et al, 2014)". The authors should thoroughly survey the literature related to EMT transition, development of pancreatic endocrine cells, and development of the granule cell lineage in dentate gyrus, to find more previously identified molecular/cellular features relevant to cell state/type transitions, compared and contrasted with findings from this study. *

Reply: Thanks. We added references on these cell phenotype transitions and modified the corresponding part. We do want to point out that the main focus of this work is that all these processes share a common feature of transient increase of intercommunity interactions.

* What is the "dynamo" package, which is supposed to contain a Python notebook? As of now, the code and data have not been made available. Both need to be released along with thorough documentation on how to run the code to reproduce the analyses described here. *Reply: Thanks. Dynamo is a python package accompanying our recent publication ____(8)____. We uploaded the code on Github and added the link of Dynamo.

* **Minor comments and questions**

1. Replace "confliction" throughout the manuscript with "conflict" or "conflicting" as appropriate. *

Reply: Thanks. We modified them.

* Paragraph two of the Introduction (beginning with "Another example of transitions ...") is missing multiple references, esp. for the last four sentences. *

Reply: Thanks. We added references.

* There are direct quotes from previous papers like "predicts the future state of individual cells on a timescale of hours". The authors are highly encouraged to check for usage of exact phrasing using available text software such as iThenticate. *

Reply____: ____Thanks a lot for pointing out this severe mistake. We re-edited the manuscript and checked with iThenticate. *

*

• "Each community contains both E and M genes": what does this mean? *

Reply: The E (M) genes are defined as those genes that are active or have high expression levels in epithelial (mesenchymal) state or sample. As we reorganized the manuscript, we add this explanation for all datasets in the caption of Fig.1i.*

*

• Reference to Qui 2021 is missing in the "Path analysis" subsection under Methods. *

Reply: We added it in the Methods.

* Fix: "transition between the cells that their sample time points are successive" in Methods. *

Reply: Thanks. ____We modified it.

* In Methods, under "Network inference", it is "partial least square regression" (not *least* s square). *

Reply: Thanks. We modified it.

* Figure 1: The cyan, magenta, and lime in 1C are very hard to see and, perhaps, the grey of the points can be made lighter. Also, change the red and green colors for the arrows in 1I to something else. These colors are not colorblind-friendly. *

Reply: Thanks. We re-plotted the figures and changed the colormap.*

*

• Periods and commas are missing at several places. Reply: Thanks. We modify these and re-edit the manuscript.

Reviewer #2 (Significance (Required)):

The study uses RNA-velocity calculated from scRNA-seq data in an inventive way to characterize paths that reflect cell phenotype transitions. Then, a sparse gene regulatory network is reverse engineered from the data and the community structure within this network is examined at various stages along the transition to make observations about inter- and intra-community regulation and network "frustration". However, the study lacks the context of existing literature in terms of previous work studying cell transitions both experimentally and computationally. Adding this context (as suggested in the comments) will considerably improve the utility and significance of the findings. Overall, this study will be of broad interest to researchers interested in development and reprogramming as well as computational scientists developing and applying methods for scRNA-seq data analysis, trajectory inference, and network reconstruction. All the comments and questions raised here are based on my background and expertise in omics data (including scRNA-seq) analysis and network biology.

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

The authors analyze three datasets of Single cell RNA velocity measured during phenotypic transition. They infer the gene regulatory network in each case and characterize the transition between the initial and final expression states (in which different sets of genes are expressed). Their motivating question was to find whether during such transitions first genes characterizing the initial state are no longer expressed and only then the genes associated with the final state start expressing or alternatively there is gradual transition through an intermediate state in which subsets of both initial and final state genes are transiently expressed.

They define a measure of regulatory frustration representing the mismatch between regulatory signals a gene receives and its current expression state. They conclude that phenotypic transitions involve transient interactions between otherwise non-interacting gene modules and a temporary increase of gene frustration, which is relaxed once the final expression state is reached.

The study uses of advanced inference and machine learning methods.

I find the question studied in this manuscript interesting, opening avenue to further questions and studies and relevant to different scientific communities. Personally I think that the focus of the paper should be the exposition of the methods used this manuscript would benefit from a longer format, but that depends of course on the journal they are aiming at. *

*

Statistical analysis is missing. Especially since the authors mention the potential of over-fitting due to large number of genes (on the order of the number of cells) - I think the authors should provide a sensitivity analysis testing how sensitive are the conclusions to the choice of cells or genes by applying the methods to subsets of the cells / genes. *

Reply: Thanks. For the subset of cells, we randomly selected cells from the dataset and performed the analyses (Fig. EV11a). For the subset of genes, we selected a subset of genes randomly and performed the analyses (Fig. EV 11b). We found the results are not affected. We also perform another statistical analysis by varying the value of resolution in community detection algorithm. And we found that the conclusion on variation of inter-community edges is not affected (Fig. EV12).

Figure EV11 Statistical analyses of dentate gyrus neurogenesis. Each dot represents a cell and color represents the number of inter-community edges.

(a) Frustration score along the RCs (left) and cell-specific variation of the number of inter-community edges (right) of a randomly selected sub-population of 2000 cells (from a total of 3184 cells);

(b) Frustration score along the RCs (left) and cell-specific variation of the number of inter-community edges) (right) of cells on the space of 400 randomly selected genes (from a total of 678 genes).

*What is the meaning of the distribution in the frustration plots? *

Reply: For each cell we calculated a frustration score. Therefore for cells in each Voronoi cell (which is a geometric cell, don’t be confused with the biological “cells”) along the reaction coordinate (Fig.1d, Fig. 2b &2g), we obtained a distribution of the frustration scores.*

In general, the conclusions are well-justified, but I think some statements in the discussion are inaccurate: "intercommunity interactions of a GRN are indeed minimized' - are they minimal or are they only lower at the stable states? There are two stable states - for which of them is intercommunity interaction lower? *

Reply: Thank. We agree with the reviewer and modified the writing. Comparing with the transition state, the number of intercommunity interactions is less for the stable states. ____The datasets' quality are not high enough for us to investigate whether ____"intercommunity interactions of a GRN are indeed minimized”.*

It is written in the discussion that 'for all three datasets frustration decreases with differentiation', but then Fig. 1g shows the opposite (final state is more frustrated than initial state). It is interesting to discuss the differences between the datasets analyzed in that respect and what could cause transition to a more frustrated state. I suggest that the authors also refer in the discussion to related questions and possible follow-up studies, such as: what determines the duration of the phenotypic transition? A relevant number is the switching time of a single gene. *

Reply: Good suggestion. Compared to other datasets, we found that the result of EMT shows larger variances. The relative difference of the frustration score is also affected by the GRN inference algorithm. For example, the difference between initial and final frustration scores of the pancreatic endocrinogenesis is more significant when using the GRISLI method (Figure EV6b). Given these, the trend that the frustration scores in the transition states transiently increase keep consistent.

Our conclusion is limited by the quality of the data. So we delete this part of discussion in the manuscript.

Qiu et al. have shown that splicing-based ____RNA velocities are relative, while metabolic-labeling-based RNA velocities are more quantitative and accurate____(8)____. We will re-analyze this problem if data with metabolic labeling becomes available.

* The authors mention at the end that the networks can often reach multiple final states from a common initial states. Do such transitions share some of their path (and in particular the intermediate frustrated state)? Given the intermediate connected state, it would be interesting to characterize the network stability to perturbations. *

Reply: This is a very important question. To reliably address these questions, we need higher quality data. We plan to characterize the network stability to perturbations in future studies, while in our recent paper using a full nonlinear modeling framework____(8)____, we performed in silico perturbations.

* While interesting, the manuscript itself is unfortunately hard to read and would benefit from major editing, including better exposition of the science and language editing. *

Reply: Thanks. We revised the manuscript extensively.*

Methods: Description of PCA and 'revised finite temperature string method' are missing in the Methods section. *

Reply:____ Thanks. PCA is used in RNA velocity analysis for dimension reduction. We added this in Materials and Methods 3. The revised string method is in Materials and Methods ____4.

*

Some examples:

Figure captions are very short and often non-informative. Some variables are not defined (or only defined later on) and the reader then needs to guess their meaning: it took me a while to understand what is 'r' in Fig. 1f and what 'r=10' (p. 4) means. *

Reply: Thanks. ____r____ represents the index number of reaction coordinates. We added this in the manuscript where we define reaction coordinates.*

p. 4: what are 'f' (as opposed to F) and 's_ij' and 's_j' (expression states?) Or is fs_ij one variable? What does a Hamiltonian of a cell mean (p. 4, bottom)? *

Reply: is the regulation of gene ____j on gene i, and is the expression state of gene i (0 for silence, and 1 for active expression). is the frustration value of regulation from gene j to gene i.

The pseudo Hamiltonian value is proposed in the literature as an analogy of ____the magnetic systems following the work of Boolean model in EMT ____(9)____. A high Hamiltonian value indicates that the cell is in an unstable state. In the original manuscript we included this quantity since it has been discussed in the literature. However we found it causes confusion and is not necessary for our discussions, so we removed the pseudo-Hamiltonian results in the revised manuscript. * P. 4: how are 'E and M genes' defined? *

Reply: The E (M) genes are defined as those genes that are active or have high expression levels at the epithelial (mesenchymal) state or sample. We explained our general strategy in the caption of Fig.1i . * What does 'network heterogeneity' (p. 5) mean? *

Reply: Network heterogeneity measures how homogenously the connections are distributed among the genes____(10)____. A high heterogeneity ____means that some genes have high degree of connectivity (the so-called hubs), while some have low degree of connectivity.

*

Fig. 1 is too tiny and hard to read and details are missing. *

Reply: Thanks. We modified this figure and caption.*

A glossary for all the acronyms used would be very helpful. *

Reply: Thanks. We added glossary in the manuscript.*

Language (some examples):

p. 5 bottom: Another system is on development... invitro -> in vitro

p. 6: 'measure on developmental potential' -> measure of... *

Reply: Thanks. We modified these and re-edited the whole manuscript.*

Reviewer #3 (Significance (Required)):

This study presents a methodological advance in demonstrating the application of data analysis methods to study developmental phenotypic transitions. High throughput measurements and computation power available today enable putting to test theoretical conjectures, as made by Waddington. I think this is a promising line of research, which could be used to further develop the computational methods as well as to further our understanding of developmental transitions and potentially develop associated mathematical modeling frameworks.

This study should be of interest to a diverse readership composed of developmental biologists as well as to quantitative biologists and CS researchers applying optimization techniques and data analysis methods to high-throughput biological data.

I am not an expert on the computational methods applied in this manuscript and hence cannot assess their correct use and statistical analysis.

*

1. Traag VA, Waltman L, & van Eck NJ (2019) From Louvain to Leiden: guaranteeing well-connected communities. Scientific Reports 9(1):5233.
2. Stuart T, et al. (2019) Comprehensive Integration of Single-Cell Data. Cell 177(7):1888-1902.e1821.
3. Bergen V, Lange M, Peidli S, Wolf FA, & Theis FJ (2020) Generalizing RNA velocity to transient cell states through dynamical modeling. Nature Biotechnology 38(12):1408-1414.
4. Wolf FA, Angerer P, & Theis FJ (2018) SCANPY: large-scale single-cell gene expression data analysis. Genome Biology 19(1):15.
5. Aubin-Frankowski P-C & Vert J-P (2020) Gene regulation inference from single-cell RNA-seq data with linear differential equations and velocity inference. Bioinformatics (Oxford, England) 36(18):4774-4780.
6. Moris N, Pina C, & Arias AM (2016) Transition states and cell fate decisions in epigenetic landscapes. Nature reviews. Genetics 17(11):693-703.
7. Kalkan T, et al. (2017) Tracking the embryonic stem cell transition from ground state pluripotency. Development 144(7):1221-1234.
8. Qiu X, et al. (2022) Mapping Transcriptomic Vector Fields of Single Cells. Cell 185(4):690-711.
9. Font-Clos F, Zapperi S, & La Porta CAM (2018) Topography of epithelial–mesenchymal plasticity. Proceedings of the National Academy of Sciences 115(23):5902-5907.
10. Gao J, Barzel B, & Barabási A-L (2016) Universal resilience patterns in complex networks. Nature 530(7590):307-312.

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 analyze three datasets of Single cell RNA velocity measured during phenotypic transition. They infer the gene regulatory network in each case and characterize the transition between the initial and final expression states (in which different sets of genes are expressed). Their motivating question was to find whether during such transitions first genes characterizing the initial state are no longer expressed and only then the genes associated with the final state start expressing or alternatively there is gradual transition through an intermediate state in which subsets of both initial and final state genes are transiently expressed.

They define a measure of regulatory frustration representing the mismatch between regulatory signals a gene receives and its current expression state. They conclude that phenotypic transitions involve transient interactions between otherwise non-interacting gene modules and a temporary increase of gene frustration, which is relaxed once the final expression state is reached.

The study uses of advanced inference and machine learning methods.

I find the question studied in this manuscript interesting, opening avenue to further questions and studies and relevant to different scientific communities. Personally I think that the focus of the paper should be the exposition of the methods used this manuscript would benefit from a longer format, but that depends of course on the journal they are aiming at.

Statistical analysis is missing. Especially since the authors mention the potential of over-fitting due to large number of genes (on the order of the number of cells) - I think the authors should provide a sensitivity analysis testing how sensitive are the conclusions to the choice of cells or genes by applying the methods to subsets of the cells / genes.

What is the meaning of the distribution in the frustration plots?

In general, the conclusions are well-justified, but I think some statements in the discussion are inaccurate: "intercommunity interactions of a GRN are indeed minimized' - are they minimal or are they only lower at the stable states? There are two stable states - for which of them is intercommunity interaction lower?

It is written in the discussion that 'for all three datasets frustration decreases with differentiation', but then Fig. 1g shows the opposite (final state is more frustrated than initial state). It is interesting to discuss the differences between the datasets analyzed in that respect and what could cause transition to a more frustrated state. I suggest that the authors also refer in the discussion to related questions and possible follow-up studies, such as: what determines the duration of the phenotypic transition? A relevant number is the switching time of a single gene.

The authors mention at the end that the networks can often reach multiple final states from a common initial states. Do such transitions share some of their path (and in particular the intermediate frustrated state)? Given the intermediate connected state, it would be interesting to characterize the network stability to perturbations. While interesting, the manuscript itself is unfortunately hard to read and would benefit from major editing, including better exposition of the science and language editing.

Methods: Description of PCA and 'revised finite temperature string method' are missing in the Methods section.

Some examples:

Figure captions are very short and often non-informative. Some variables are not defined (or only defined later on) and the reader then needs to guess their meaning: it took me a while to understand what is 'r' in Fig. 1f and what 'r=10' (p. 4) means.

p. 4: what are 'f' (as opposed to F) and 's_ij' and 's_j' (expression states?) Or is fs_ij one variable? What does a Hamiltonian of a cell mean (p. 4, bottom)?

P. 4: how are 'E and M genes' defined?

What does 'network heterogeneity' (p. 5) mean?

Fig. 1 is too tiny and hard to read and details are missing.

A glossary for all the acronyms used would be very helpful.

Language (some examples):

p. 5 bottom: Another system is on development... invitro -> in vitro

p. 6: 'measure on developmental potential' -> measure of...

Significance

This study presents a methodological advance in demonstrating the application of data analysis methods to study developmental phenotypic transitions. High throughput measurements and computation power available today enable putting to test theoretical conjectures, as made by Waddington. I think this is a promising line of research, which could be used to further develop the computational methods as well as to further our understanding of developmental transitions and potentially develop associated mathematical modeling frameworks.

This study should be of interest to a diverse readership composed of developmental biologists as well as to quantitative biologists and CS researchers applying optimization techniques and data analysis methods to high-throughput biological data.

I am not an expert on the computational methods applied in this manuscript and hence cannot assess their correct use and statistical analysis.

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

Understanding the cellular and molecular basis of cell type or cell state transitions occurring during development or reprogramming is a fundamental challenge. scRNA-seq has provided a window into gene expression programs across thousands of cells undergoing such transitions. Wang and colleagues leverage scRNA-seq and develop an approach to reverse engineer gene regulatory network underlying cells along a path from one cell type/state to another, and characterize community-level properties of this network associated with various stages of the cell phenotype transition. The study is innovative and rigorous, and their results point to how intercommunity interactions increase and then decrease, indicating a concerted regulatory rewiring that orchestrates transitions. Application of their approach to three different datasets also shows that this trend is consistent across three different transitions and maybe a general trend. However, there are some major and minor concerns that need to be addressed.

Major comments and questions

1. The analogy to SN1 and SN2 mechanisms of chemical bond formation is very nice.
2. What is the basis for the two statements made in paragraph 3 of Introduction (beginning with "A question arises ...") about transitions being sequential or concurrent? Please

2.1. Provide references to previous experimental and computational studies that have investigated developmental and reprogramming gene expression programs.

2.2. Describe specific examples of findings that support the two possible transitions highlighted here. Why couldn't transitions happen through an entirely gradual process involving changes to overlapping subsets of genes.

1. Please make plots of the number of effective intra-community edges vs. number of active genes to support the statement that these two numbers are correlated.
2. A bunch of notations are not clear:

4.1. What is the "r" in "strongest intercommunity interactions at r = 10 (Fig. 1F)"? Is it the same as the "r" mentioned in the Methods section?

4.2. What is "s_i" in "cell-specific effective matrix, Fbar_ij = (2s_i - 1)F_ij"? Also, that description of F_ij, f_ij, and H should be moved to the Methods section, and a more high-level, intuitive description should instead be included in this Results paragraph.

1. How were the h_f and h_m thresholds chosen?
2. What is the "density of each single cell" ("⍴_t")? The formulation of the penalty of the distance between cells i and j (the expression with -logP_ij...) is unclear. What is the intuition behind it? What is r? How were the values of r (0.5 and 0.8) chosen?
3. One of the reasons the authors state to justify the choice of PLSR is "In the scRNA dataset, the number of genes is often comparable to or larger than the number of cells." This is not true most of the time. In nearly all recent studies, the number of cells is way larger than the number of genes measured.
4. There is a fleeting reference to a nice previous finding that supports their observations: "several lines of evidence support that EMT proceeds through a concerted mechanism. Indeed, both in vivo and in vitro studies have identified intermediate states of EMT that have co-expressed epithelial and mesenchymal genes (Pastushenko et al, 2018; Zhang et al, 2014)". The authors should thoroughly survey the literature related to EMT transition, development of pancreatic endocrine cells, and development of the granule cell lineage in dentate gyrus, to find more previously identified molecular/cellular features relevant to cell state/type transitions, compared and contrasted with findings from this study.
5. What is the "dynamo" package, which is supposed to contain a Python notebook? As of now, the code and data have not been made available. Both need to be released along with thorough documentation on how to run the code to reproduce the analyses described here.

Minor comments and questions

1. Replace "confliction" throughout the manuscript with "conflict" or "conflicting" as appropriate.
2. Paragraph two of the Introduction (beginning with "Another example of transitions ...") is missing multiple references, esp. for the last four sentences.
3. There are direct quotes from previous papers like "predicts the future state of individual cells on a timescale of hours". The authors are highly encouraged to check for usage of exact phrasing using available text software such as iThenticate.
4. "Each community contains both E and M genes": what does this mean?
5. Reference to Qui 2021 is missing in the "Path analysis" subsection under Methods.
6. Fix: "transition between the cells that their sample time points are successive" in Methods.
7. In Methods, under "Network inference", it is "partial least square regression" (not least s square).
8. Figure 1: The cyan, magenta, and lime in 1C are very hard to see and, perhaps, the grey of the points can be made lighter. Also, change the red and green colors for the arrows in 1I to something else. These colors are not colorblind-friendly.
9. Periods and commas are missing at several places.

Significance

The study uses RNA-velocity calculated from scRNA-seq data in an inventive way to characterize paths that reflect cell phenotype transitions. Then, a sparse gene regulatory network is reverse engineered from the data and the community structure within this network is examined at various stages along the transition to make observations about inter- and intra-community regulation and network "frustration". However, the study lacks the context of existing literature in terms of previous work studying cell transitions both experimentally and computationally. Adding this context (as suggested in the comments) will considerably improve the utility and significance of the findings. Overall, this study will be of broad interest to researchers interested in development and reprogramming as well as computational scientists developing and applying methods for scRNA-seq data analysis, trajectory inference, and network reconstruction. All the comments and questions raised here are based on my background and expertise in omics data (including scRNA-seq) analysis and network biology.

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 by Wang and colleagues, the authors analyse single-cell RNA-seq (scRNAseq) data by applying transition path theory to infer gene regulatory network (GRN) changes along the transition (reaction coordinate, trajectory) between free energy stable states (i.e. cell types). The work aims to understand how stable cell types, and their regulatory programs (combination of active and repressed genes) switches during differentiation/reprogramming/response (i.e. cell phenotypic transition/CPT). The premise of the work is to assess whether genes within GRNs undergo step-wise repression, state-change and activation (& vice-versa; analogous to SN1) or concurrently regulate gene expression (analogous to SN2). The GRNs are inferred based on highly variable genes and their expression dynamics from RNA velocity over CPT, across 3 scRNA-seq datasets.

The authors first analyse public scRNA-seq dataset of 3003 human A549 adenocarcinomic basal epithelial cells treated with TGF- for 0hrs, 8hrs, 1 day and 3 days (4 timepoints). The authors select two stable states (Day0-untreated; Epithelial and Day 3-treatment; Mesenchymal) using local kernel densities and set transition paths using Dijkstra shortest path, dividing state space into Voronoi cells (i.e. reaction coordinate value), and constructed single-cell GRNs based on RNA velocity differences (n=500 genes) and a linear model (from Qiu et al). This GRN is based on expression and velocity estimates, and does not distinguish direct from indirect regulation. Calculating interaction frequency (edges) across two stable states over 4 louvain clusters, the authors find global increase in effective edges that correlates with increased active genes; but with variable trend within inter-cluster edges. To quantify the concerted GRN changes between clusters, the authors utilise a "frustration" score (from Tripathi et al 2020). The average frustration score increases and peaks at day 1 treatment, followed by a decline over terminal stable state (day 3-treatment); similar to interaction frequency trends. The author also separately measure network heterogeneity and repeat analysis using alternative transition matrix. The authors conclude that EMT proceeds through concerted regulation of multiple genes first with an increase in inter-cluster edges, frustration and heterogeneity followed by a decrease into final stable state. The authors apply the analysis to scRNA-seq data from (i) pancreatic endocrine differentiation where Ngn3-low progenitors give rise to Ngn3-high, then Fev-high and into glucagon producing -endocrine cells; (ii) dendate gyrus; radial glial cell differentiation into nIPCs, neuroblast, immature granule and mature granule cells. In both cases, the authors observe concerted regulation with initial increase in inter-community edges, heterogeneity during differentiation followed by decrease towards final stable state.

The study and ideas in the manuscript are interesting and the methods would be potentially be useful. However, there are a few specific and general comments stated below, which the authors should try to address.

• P4: "RC increases first and reaches a peak when cells were treated with TGF-β for about one day, then decreases (Fig. 1G)". It would be better to label the figure with the treatment information. • Fig. 1G and EV1D: Why are the trends different? • How is the appropriate community/cluster/Louvain resolution selected? This can have a major impact on number of cell states, types and transition path from initial to final state. • What effect does the Louvain resolution have on e.g. frustration scores? • The authors match resolution to samples/timepoints/known prior cell types i.e. 3-4 communities. However it is unclear whether this is enough to describe entire differentiation/transition process. • Gene selection: The selection based on minimum 20 counts as highly expressed genes is arbitrary and dependent on sequencing depth. Perhaps the authors could show distribution of gene counts for the datasets and have a data-driven filtering criteria • The choice of 500 variable genes (for human A549 cells) is also quite arbitrary. Perhaps, the authors could compare how additional genes (all highly variable genes) affects their analysis and interpretation. • How are other factors (sequencing depth, genes detected, #of cell types, multiple branches) affects the connectivity between communities at different phases of transition/development? • Are the velocity graph, transition matrix and further shortest path estimation derived in a reduced latent space, and if so, how much (nPCs) and what impact does it have. Presumably, the density estimation is not performed in expression space.

• The figure legends and labels were hard to read. These should be improved for better readability.
• A suggestion would be move the initial results section to methods and highlight the biological interpretation. The authors could highly which GRN and representative genes/edge pairs are highest ranked within inter-community and to overall final stable states.
• How does the GRN inference compare to current state-of-the-art GRN inference scRNA-seq methods?
• How do extremely noisy/stochastic genes vary in metrics between final stable states? How are the metrics affected by number of cells and stochasticity of expression within a given cluster/community.
• Given that the author's approach includes both direct and indirect genes effects, the authors could further prune genes based on existing TF databases or protein-protein validated networks.
• It is unclear which GRNs are already known and which ones are novel and biologically relevant
• It would be good for authors to comment when there are multiple bifurcations instead of A-B transitions. Particularly in datasets with multiple discrete stable states.
• Another suggestion would be to highlight gene expression of selected markers based on f-regression and mi over the trajectory
• If possible, a proof of principle could be re-analysis of a perturbation scRNA-seq dataset (e.g. where one path/transition path is stalled)

Significance

Nature and significance of advance: The study and ideas in the manuscript are interesting and the methods would be potentially be useful to community.

Compare to existing published knowledge: -

Audience: Predominantly computational audience

Your Expertise: PI with background in experimental, computational biology and expertise in single-cell genomic tools and developmental biology

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>

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>

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 are very grateful to the three referees for their constructive comments and suggestions which have helped improve the quality of our manuscript.

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

In the publication HAT-field: a very cheap, robust and quantitative point-of-care serological test for Covid-19 by Joly and Ribes the authors describe an adaption and an improved protocol to their previously published haemagglutination based test to detect antibodies to SARS-CoV-2 in patient blood (Towsend et al., 2021). In detail, they analyzed the effect of several adaptions including buffer optimization, plate coating, usage of patient whole blood instead of washed RBCs and plasma. Additionally they tested different temperatures and stability of the reagents, namely the nanobody-RBD construct IH4-RBD. For validation they compared their optimized HAT-field assay with Jurkat-S&R as a FACS-based assay.

Major comments:

Introduction: This section is rather short and could benefit from a broader overview of currently established methods and assays to detect appropriate immune responses against SARS-CoV-2. The author are advised to summarize the current literature in the field more comprehensively and not only focus on their own work.

Response: Hundreds of different tests to monitor immune responses against SARS-CoV-2 have been described to date, and the literature on these various tests is vast, with new articles coming out almost on a daily basis. We would not feel either that the introduction of our rather technical paper would benefit from being lengthened by such a review of the current literature, or even competent to carry out such a summary. Following the referee’s suggestion, we have, however, introduced a new sentence and given three references providing relatively recent overviews on the subject of immune-monitoring.

Cross-reactivity with IH4-RBD. In Figure 6, the authors highlight the samples in red and orange that showed cross-reactivity with IH4-RBD. In their discussion, however, the authors state that only 2 of 60 (3%) were cross-reactive. In making this statement, they ignore the proportion of cross-reactive samples that were also positive in the Jurkat S&R assay. Therefore, the authors should acknowledge in the discussion that the actual number of cross-reactive samples was higher.

Response*: The statement in the discussion about 2 cross reactive samples out of 60 concerns the results obtained after an incubation of one hour under normal gravity, and not the two red dots in each of the three graphs of figure 6, which correspond to the two negative samples which gave false-positive results in HAT plasma titrations after spinning (Figure 6C), for which we correctly state in the discussion that 12 samples showed cross-reactivity on IH4 alone. The data presented in Figure 6B corresponds to HAT-field after spinning, for which we correctly state in the discussion that 5 out of 60 showed cross-reactivity (4 orange dots + 1 red dot, the second red dot having a score of 0, in accordance with the fact that this sample showed no cross reaction on IH4 alone in HAT-field after spinning). *

*To try to prevent this possible confusion, we have now clarified what data we are referring to at the start of that paragraph in the discussion. *

Quantitative Assay. Since the HAT assay does not allow determination of the absolute number of antibodies reactive to SARS-CoV-2 in the blood samples, the authors should refrain from claiming that the HAT-field is a quantitative assay.

Response*: Since immune sera are inherently polyclonal, they contain a multitude of different types of antibodies of different affinities and avidities, and we are not aware of any technique that allows to determine the “absolute number” of antibodies directed against a given antigen in such samples. *

*For many serological tests, including ELISA and the initial protocol of HAT, serum or plasma titrations are used as a means to obtain what is widely considered as a quantitative evaluation of the amounts of antibodies in blood samples. Even FACS-based assays such as the Jurkat-S&R-flow test we have used, are commonly considered as quantitative but those only provide relative results and not absolute numbers. *

We perceive that the close correlations we find between the results of the HAT-field protocol and those of the Jurkat-S&R-flow test as well as with serum titrations using the standard HAT protocol warrants considering the results of HAT-field as being as quantitative as those obtained with all those other tests.

Morphological read out For field application, the morphological description of the observed deposits ("teardrop" vs. "button") could be problematic and might lead to bias depending on the user. Thus, the authors should provide a clearer description for phenotype classification.

Response: We have now introduced a specific paragraph detailing how to score HAT assays in the Methods section, as well as a new figure providing a graphic description of positive, partial and negative RBCs deposits.

Minor comments: Title: the authors should remove "very"

Response*: We have now removed the word ‘very’ from the title, and thank the referee for this helpful suggestion. *

By the way: What are the costs of IH4-RBD for a 96 well plate? Who will produce this reagent? Is the sequence of the IH4 fully disclosed?

Response*: As specified in our original paper (see Townsend et al. 2021), the plasmid coding for the IH4-RBD is available upon request from Alain Townsend (Oxford, UK). Furthermore, his laboratory funded the production of 1 gram of the IH4-RBD reagent by a commercial company, and professor Townsend has been graciously sending aliquots of 1 mg of this reagent, which suffice for several thousand tests, to all the laboratories that have requested it from him. *

*In its initial format, HAT only required 100 ng of IH4-RBD per well, corresponding to a cost of 0.0027 £ per well. For the HAT-field protocol, 5 times more reagent is needed, thus bringing the cost of the reagent to 1.5 cts per test, to which one would have to add a similar cost for the IH4-reagent alone. This would thus bring the cost of the two reagents to approximately 3 cts, which is still lower than the price of any of the cheap disposable plasticware necessary for the test (lancet, pipet, plastic tube and portion of a plate). *

The sequence of the IH4 nanobody is indeed fully disclosed (see figure 1 of Townsend et al. 2021), and has actually been protected by a patent ( US9879090B2 ). Whilst IH4 can be used freely for research purposes, licensing rights would have to be taken into consideration by any health authority wishing to use the technique broadly, or for any commercial distribution.

The usage of the CR3022 as positive control for neutralizing antibodies should be reconsidered since this antibody does not confer viral neutralization. Other well describe antibodies blocking the ACE2:RBD interface might be better suited.

Response*: CR3022 was the one that we had at our disposal, but other mAbs can certainly be used instead of as positive controls, and this is actually indicated in the detailed HAT-field protocol provided. Since the use of a positive control is only to ensure that the IH4-RBD has not been degraded and works as well as expected, and that any negative samples are not due to a very rare glycophorin mutation that could prevent IH4 from binding to it at the surface of RBCs, we are not sure why using a mAb with neutralizing activity would necessarily be better than the CR3022 mAb. *

Figure 2: Please state the concentration of IH4-RBD used. As stated in the figure legends for Figure 2 B, the authors should show the result all 4 replicates (incl. SD)

Response: The concentration of IH4-RBD was 1 m*g/ml, i.e. the normal concentration for standard HAT tests. This was already indicated in the Methods section, but has now been added to the legend of Figure 2. *

Whilst 4 experiments were indeed carried out, which all gave similar results, i.e. showed that using PBS-N3 or PBN did not hinder HAT performance, but could instead result in a slight increase in HAT sensitivity, those various experiments were not all exact replicates of the experiment shown on figure 2. Furthermore, performing of those various experiments was spread over a period of over a year, using different reagents, thus precluding numerical comparisons between the various results. We have clarified this issue by rewording the final statement to “Comparable results were obtained in four similar experiments.”

Figure 3: Although the authors showed stability of IH4-RBD at 2 µg/ml they do not provide data for the stabilities at higher dilutions. As the authors suggest to predistribute the IH4-RBD in plates they should at least discuss this issue.

We thank the referee for raising this valid point, which has now been discussed in the paragraph entitled “Practical considerations for performing HAT assays” in the Methods section: “One aspect that will have to be considered for the design and use of such individual strips of wells will be to ensure that, upon storage, the various dilutions of IH4-RBD are as stable in such strips as the working stocks of IH4-RBD (2 mg/ml) tested in Figure 3.”

Figure 6/Supplementary Figure 1 and 3 The presentation of the data is not accurate, as many of the points (samples) are obviously identically positioned in the graph. The authors should choose a different representation of their data. E.g. they could adjust the size of the points to the number of overlapping samples.

Response: We thank the referee for raising this issue, which was also pointed to by referee #2. This apparent inaccuracy is due to the fact that, on these plots, the scales for both x and Y axes used discrete values, which indeed results in multiple points overlapping on top of one another. This was resolved by adding numbers next to the positions where several dots overlapped

Wording / text length In the current manuscript the text is very long. Thus, the authors should shorten it to report the essential findings more appropriately. Additionally they should check for correct English wording.

Response*: We thank the referee for this remark, which helped us realize that the excessive length of the manuscript was mostly due to an extensive discussion of highly technical and practical points. The corresponding paragraphs were indeed out of place in the general discussion, and have not been deleted but have been moved to the Methods section since we feel that they contain very important information for people who would actually start to performing HAT assays. *

Reviewer #1 (Significance (Required)):

In summary, the authors describe the HAT-field test as a simple PoC test for the detection of SARS-CoV-2 antibodies in patients. Because of its ease of use and robustness, the test appears to be particularly well suited for use in countries with underdeveloped health care or limited testing facilities, as also reported previously. The value of this manuscript lies mainly in the detailed description of the protocol and its validation. In this context, the adaptations described are certainly useful and helpful from a practical point of view, but do not provide significant new scientific insights. In light of these considerations, we recommend that this work be submitted to an appropriate journal specializing in the publication of such methods

Expertise The reviewers have established and published different serological assays to monitor immune responses against SARS-CoV-2

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

In this paper, the authors developed a feasible protocol for an affordable point-of-care serological test for SARS-CoV-2. This method was adapted from the HAT plasma titration test that the authors previously published. Specifically, the test utilizes a 96-well plate pre-coated with the RBD of SARS-CoV-2 spike glycoprotein fusing to a red blood cell targeting nanobody (IH4). By adding microliters amount of the blood or plasma samples to the plate, it allows the detection of antibodies against RBD by measuring the level of hemagglutination. In the current upgraded protocol (so called HAT-field), the authors made major modifications including optimizations of buffer and experimental protocol and the use of pre-titrated IH4-RBD on the plate, which collectively helped to lower the sample consumptions, improved the stability and the sensitivity of detection, and made the test more user-friendly under non-clinical settings.

Major comments: My major concerns are related to the robustness and quantitative capability of this approach. Specifically: It seems that multiple variables may impact the results. These include volume of droplets, the presence/absence of serum IH4 or BSA cross-reactive antibodies, and the amount (%) of red blood cells which may vary substantially among samples. Could you find a way to normalize the results (e.g., the discrepancy shown in Figure 6) instead of only leaving them as false-positives or false-negative?

Response*: Regarding the volume of the droplets, in other words, the amount of blood collected and used in an assay, two sentences in the manuscript underline the fact that this is not a critical variable: *

In the Results section “the precise volume of blood collected is not critical; it may vary by as much as 30% with no detectable influence on the results.”

In the discussion: “On this subject, we have found that increasing the amount of whole blood per well (in other words using blood that is less dilute) has very little influence over the HAT-field results, and, if anything, adding more blood can sometimes reduce the sensitivity, albeit never by more than 1 dilution.”

Consequently the % of RBCs in samples seem unlikely to influence the HAT-field scores significantly. This is supported by the fact that, although men tend to have higher hematocrits than women, we have not noticed any detectable difference between men and women in the correlation of the HAT-field scores with those of the Jurkat-S&R-flow test.

We are not sure that we fully understand what discrepancy shown in Figure 6 the referee is pointing to, but if it is about the increase in the number of samples found to be cross reacting on IH4 alone when the sensitivity increases, in the discussion, we propose to perform tests using titrations of the IH4 nanobody alone simultaneously to using the IH4-RBD reagent, so as to minimize the number of samples that would be identified as false positives if only one concentration of IH4 alone was used as negative control. Comparing the titers obtained with IH4-RBD and IH4 alone will then provide some level of normalization for the samples cross reacting on IH4. As for the hypothetical presence of antibodies cross reacting on BSA alluded to by the referee, since such antibodies would not bind to RBCs, we do not think they would affect the HAT results.

Second, the score of the HAT-field ranges from 0 - 8. However, based on the current manuscript, it is not clear how the scoring and scaling works. How is the noise (non-specific antibody signal) defined here?

Response: We have now introduced a specific paragraph and a new figure detailing how to score HAT assays in the Methods section.

In addition, it is unclear how to translate the HAT-field score into a meaningful measure of protection by serum antibodies.

Response*: Documenting the correlation between HAT-field scores and levels of protection against SARS-CoV-2 infections and/or Covid-19 severity would indeed be extremely interesting. This would, however, require setting up a large scale clinical trial carried out over several months. This type of work could only be carried out by a large consortium including clinicians or even preferably a national health agency. This was, however, far beyond the reach of this initial project, which was based on the work of a single person on a shoestring budget. *

Can you provide more evidence to demonstrate that the test is quantitative? For example, performing additional orthogonal experiments to better validate the scoring and generate a correlation function?

Response*: Inasmuch as it would have been very interesting to perform additional serological tests from commercial sources on the samples of our cohort, such tests are all very expensive (e.g. ca. 500 € for one ELISA plate). This was in fact the main reason for developing the Jurkat-S&R-flow test in the first place, since it is much cheaper, more modular, and at least as sensitive as ELISA (see Maurel Ribes et al. 2021). The funds for this whole project came from a single 15 k€ grant obtained from the ANR, and we simply did not have access to the funds, or to the human resources to carry out such experiments based on commercial serological tests. *

Minor comments: Figure 6: are all results included? To me, it does not seem that all 60 samples data were included in the plot.

Response: We thank the referee for raising this issue, which was also pointed to by referee #1. This apparent inaccuracy is due to the fact that the scales for both x and Y axes used discrete values, which results in multiple points overlapping on top of one another. This was resolved by adding numbers next to the positions where several dots overlapped.

There are several redundant statements in the discussion and results section. Please make the text more concise.

Response: The discussion has now been shortened considerably, mostly by moving the paragraphs pertaining to technical considerations to the Methods section.

Reviewer #2 (Significance (Required)):

The current paper is built upon the improvement of previous published work. In addition, there are similar approaches that have been published. It was unclear if the current method is superior to other works.

Response: Whilst we have made no statement regarding whether the method we describe is superior to other methods, we are pretty confident that very few alternatives will be as frugal and simple as the HAT-field protocol described here. As alluded to in the final paragraph of the discussion, two recent reports have described that HAT could be performed on cards rather than in V-shaped wells, with semi-quantitative results being obtained in minutes. If such card-based approaches turn out to provide sensitivity and reliability comparable to those of the HAT-field protocol, they will certainly represent very interesting alternatives. As stated in our manuscript, we would be very interested if the comparative evaluation of the two approaches could be carried out by one or several independent third party.

My research involves the development of antiviral antibody therapeutics. This method may be used as a point-of-care tool for the measurement of serologic response to RBD in less developed countries. However, due to the high vaccination rate and large infected populations, the overall needs for such detection drastically decrease. The significance of the work and utilities of the test may expand with more experiments related to the variants.

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

This paper describes a low-cost robust and quantitative serological test based on haemgglutination, which could be used in resource limited settings for evaluating population-based and vaccine induced immunity. Neutralising antibodies to the receptor binding domain (RBD) on the SARS CoV-2 spike protein are an immunological correlate of protection. The HAT has a single reagent the RBD domain of SARS CoV-2 linked to a monomeric anti-erythrocyte single domain nanobody. When human polyclonal serum antibodies bind to the RBD they cross-link and agglutinate human red blood cells, resulting in haemagglutination which can be read visually.

This paper thoroughly evaluate the stability of the HAT reagents used to measure human and monoclonal antibodies examining the robustness of the HAT reagent. It provides a comprehensive protocol for conducting field based HAT with limited reagents. The test can evaluate is subjects have been infected using a simple finger prick to detect RBD specific antibodies. The field HAT can also be used to define people that can be susceptible to reinfection or in need of vaccination, With the use of RBDs from the variants of concern the test can be rapidly adapted to evaluate antibodies as new variants arise to evaluate surrogate correlates of protection to allow timely evaluation of vaccine effectiveness and predict the need for vaccine booster doses. The data are very comprehensively presented with good figures demonstrating the most appropriate buffer to store the IH4-RBD reagent and the robustness of the HAT over time at different temperatures. No additional experiments are needed and suitable numbers of replicates are included. All data, methods and reagents are comprehensively described.

Minor comments: The paper is well written but rather long in places and may have benefited from being more succinct.

Response: The excessive length of the manuscript was mostly due to an extensive discussion of highly technical and practical points. The discussion has now been shortened considerably, mostly by moving the paragraphs pertaining to technical considerations to the Methods section.

Panels in figures could be labelled as A, B, C etc to help in identifying the correct panel..

Response: We thank the referee for this helpful suggestion, which we have followed.

I would avoid the use of experiment and project and refer to next we confirmed... or in this paper or our results show Please make sure all abbreviation are defined upon first use. Perhaps include early in the paper that most of the work was conducted with the Wuhan RBD

Response: We thank the referee for these helpful suggestions, which we have followed to the best of our abilities. The abstract now contains a mention of the fact the work on optimizing the protocol was carried out with the IH4-RBD carrying the Wuhan version.

Figure 2: I would suggest placing either a solid line between the two halves of the plates to make it easier for the reader to differentiate between the two antibodies. It also would have been easier to read if the bottom PBS, PBS-N3 and PBN were at 45 degree angle. In B include the serum name (e.g. serum 197).

Response: We thank the referee for these helpful suggestions, which we have followed.

Legend to figure 4: please include the serum numbers after covid-19 patients. Perhaps include arrows to demonstrate the dilutions of serum and IH4-RBD in the figure.

Page 6 it might be easiest to use the same times as in figure 6 and use for example more than one year in the discussion

Response: We thank the referee for these helpful suggestions, which we have all followed.

Legend figure 6 perhaps replace dots with circles page 10 include the R values from figure 6 in the description of results.

Response: We are grateful to the referee for these helpful suggestions, but have not followed them since we do not feel that these changes would be real improvements.

Page 12 of note perhaps this can be moved to the methods ?

Response: This, and several other paragraphs of the Discussion, have now been moved to the Methods section.

Supplementary figure 2 A can be seen, is something missing here?

Response: An s was indeed missing : “A can be seen” corrected to “As can be seen “

*

Reviewer #3 (Significance (Required)):

This paper describes a simple rapid field test for evaluating antibodies to the receptor binding domain of the spike of SARS CoV-2 using the Wuhan and delta variant. Whilst high income countries can provide booster doses and extensive testing (either lateral flow or RT-PCR based) and contact racing to control the waves of the pandemic, low income countries have had limited access to Covid vaccine and the extent of previous waves of the pandemic in the populations are unknown.

This paper describes a robust and simple test for investigating human antibodies to SARS-CoV-2 which could be performed in resource limited settings providing a very useful tool for monitoring infection in the community and potentially for prioritising this scarce COVID-19 vaccines available.

This study builds upon the work conducted on the HAT and has extensively studied and optimised the test so that it could be used globally. This paper provides a comprehensive protocol and has simplified the test to ensure it could be used in LMICs.

This paper would be of great interest to a wide scientific audience who are interested in a rapid low-cost test to evaluate population based and vaccine induced immunity.

Reviewer: serological assays for use in virology and vaccinology. Suitable competence to review the whole paper *

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 describes a low-cost robust and quantitative serological test based on haemgglutination, which could be used in resource limited settings for evaluating population-based and vaccine induced immunity. Neutralising antibodies to the receptor binding domain (RBD) on the SARS CoV-2 spike protein are an immunological correlate of protection. The HAT has a single reagent the RBD domain of SARS CoV-2 linked to a monomeric anti-erythrocyte single domain nanobody. When human polyclonal serum antibodies bind to the RBD they cross-link and agglutinate human red blood cells, resulting in haemagglutination which can be read visually.

This paper thoroughly evaluate the stability of the HAT reagents used to measure human and monoclonal antibodies examining the robustness of the HAT reagent. It provides a comprehensive protocol for conducting field based HAT with limited reagents. The test can evaluate is subjects have been infected using a simple finger prick to detect RBD specific antibodies. The field HAT can also be used to define people that can be susceptible to reinfection or in need of vaccination, With the use of RBDsfrom the variants of concern the test can be rapidly adapted to evaluate antibodies as new variants arise to evaluate surrogate correlates of protection to allow timely evaluation of vaccine effectiveness and predict the need for vaccine booster doses. The data are very comprehensively presented with good figures demonstrating the most appropriate buffer to store the IH4-RBD reagent and the robustness of the HAT over time at different temperatures. No additional experiments are needed and suitable numbers of replicates are included. All data, methods and reagents are comprehensively described.

Minor comments:

The paper is well written but rather long in places and may have benefited from being more succinct.

Panels in figures could be labelled as A, B, C etc to help in identifying the correct panel..

I would avoid the use of experiment and project and refer to next we confirmed... or in this paper or our results show

Please make sure all abbreviation are defined upon first use.

Perhaps include early in the paper that most of the work was conducted with the Wuhan RBD

Figure 2: I would suggest placing either a solid line between the two halves of the plates to make it easier for the reader to differentiate between the two antibodies. It also would have been easier to read if the bottom PBS, PBS-N3 and PBN were at 45 degree angle. In B include the serum name (e.g. serum 197).

Legend to figure 4: please include the serum numbers after covid-19 patients. Perhaps include arrows to demonstrate the dilutions of serum and IH4-RBD in the figure.

Page 6 it might be easiest to use the same times as in figure 6 and use for example more than one year in the discussion Legend figure 6 perhaps replace dots with circles page 10 include the R values from figure 6 in the description of results.

Page 12 of note preps this can be moved to the methods Supplementary figure 2 A can be seen, is something missing here?

Significance

This paper describes a simple rapid field test for evaluating antibodies to the receptor binding domain of the spike of SARS CoV-2 using the Wuhan and delta variant. Whilst high income countries can provide booster doses and extensive testing (either lateral flow or RT-PCR based) and contact racing to control the waves of the pandemic, low income countries have had limited access to Covid vaccine and the extent of previous waves of the pandemic in the populations are unknown.

This paper describes a robust and simple test for investigating human antibodies to SARS-CoV-2 which could be performed in resource limited settings providing a very useful tool for monitoring infection in the community and potentially for prioritising this scarce COVID-19 vaccines available.

This study builds upon the work conducted on the HAT and has extensively studied and optimised the test so that it could be used globally. This paper provides a comprehensive protocol and has simplified the test to ensure it could be used in LMICs.

This paper would be of great interest to a wide scientific audience who are interested in a rapid low-cost test to evaluate population based and vaccine induced immunity.

Reviewer: serological assays for use in virology and vaccinology. Suitable competence to review the whole paper

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

In this paper, the authors developed a feasible protocol for an affordable point-of-care serological test for SARS-CoV-2. This method was adapted from the HAT plasma titration test that the authors previously published. Specifically, the test utilizes a 96-well plate pre-coated with the RBD of SARS-CoV-2 spike glycoprotein fusing to a red blood cell targeting nanobody (IH4). By adding microliters amount of the blood or plasma samples to the plate, it allows the detection of antibodies against RBD by measuring the level of hemagglutination. In the current upgraded protocol (so called HAT-field), the authors made major modifications including optimizations of buffer and experimental protocol and the use of pre-titrated IH4-RBD on the plate, which collectively helped to lower the sample consumptions, improved the stability and the sensitivity of detection, and made the test more user-friendly under non-clinical settings.

Major comments:

My major concerns are related to the robustness and quantitative capability of this approach.

Specifically:

It seems that multiple variables may impact the results. These include volume of droplets, the presence/absence of serum IH4 or BSA cross-reactive antibodies, and the amount (%) of red blood cells which may vary substantially among samples. Could you find a way to normalize the results (e.g., the discrepancy shown in Figure 6) instead of only leaving them as false-positives or false-negative? Second, the score of the HAT-field ranges from 0 - 8. However, based on the current manuscript, it is not clear how the scoring and scaling works. How is the noise (non-specific antibody singal) defined here? In addition, it is unclear how to translate the HAT-field score into a meaningful measure of protection by serum antibodies. Can you provide more evidence to demonstrate that the test is quantitative? For example, performing additional orthogonal experiments to better validate the scoring and generate a correlation function?

Minor comments:

Figure 6: are all results included? To me, it does not seem that all 60 samples data were included in the plot.

There are several redundant statements in the discussion and results section. Please make the text more concise.

Significance

The current paper is built upon the improvement of previous published work. In addition, there are similar approaches that have been published. It was unclear if the current method is superior to other works. My research involves the development of antiviral antibody therapeutics. This method may be used as a point-of-care tool for the measurement of serologic response to RBD in less developed countries. However, due to the high vaccination rate and large infected populations, the overall needs for such detection drastically decrease. The significance of the work and utilities of the test may expand with more experiments related to the variants.

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 the publication HAT-field: a very cheap, robust and quantitative point-of-care serological test for Covid-19 by Joly and Ribes the authors describe an adaption and an improved protocol to their previously published haemagglutination based test to detect antibodies to SARS-CoV-2 in patient blood (Towsend et al., 2021). In detail, they analyzed the effect of several adaptions including buffer optimization, plate coating, usage of patient whole blood instead of washed RBCs and plasma. Additionally they tested different temperatures and stability of the reagents, namely the nanobody-RBD construct IH4-RBD. For validation they compared their optimized HAT-field assay with Jurkat-S&R as a FACS-based assay.

Major comments:

Introduction: This section is rather short and could benefit from a broader overview of currently established methods and assays to detect appropriate immune responses against SARS-CoV-2. The author are advised to summarize the current literature in the field more comprehensively and not only focus on their own work.

Cross-reactivity with IH4-RBD. In Figure 6, the authors highlight the samples in red and orange that showed cross-reactivity with IH4-RBD. In their discussion, however, the authors state that only 2 of 60 (3%) were cross-reactive. In making this statement, they ignore the proportion of cross-reactive samples that were also positive in the Jurkat S&R assay. Therefore, the authors should acknowledge in the discussion that the actual number of cross-reactive samples was higher.

Quantitative Assay. Since the HAT assay does not allow determination of the absolute number of antibodies reactive to SARS-CoV-2 in the blood samples, the authors should refrain from claiming that the HAT-field is a quantitative assay.

Morphological read out For field application, the morphological description of the observed deposits ("teardrop" vs. "button") could be problematic and might lead to bias depending on the user. Thus, the authors should provide a clearer description for phenotype classification.

Minor comments:

Title: the authors should remove "very" By the way: What are the costs of IH4-RBD for a 96 well plate? Who will produce this reagent? Is the sequence of the IH4 fully disclosed?

The usage of the CR3022 as positive control for neutralizing antibodies should be reconsidered since this antibody does not confer viral neutralization. Other well describe antibodies blocking the ACE2:RBD interface might be better suited.

Figure 2: Please state the concentration of IH4-RBD used. As stated in the figure legends for Figure 2 B, the authors should show the result all 4 replicates (incl. SD)

Figure 3: Although the authors showed stability of IH4-RBD at 2 µg/ml they do not provide data for the stabilities at higher dilutions. As the authors suggest to predistribute the IH4-RBD in plates they should at discuss this issue.

Figure 6/Supplementary Figure 1 and 3 The presentation of the data is not accurate, as many of the points (samples) are obviously identically positioned in the graph. The authors should choose a different representation of their data. E.g. they could adjust the size of the points to the number of overlapping samples.

Wording / text length In the current manuscript the text is very long. Thus, the authors should shorten it to report the essential findings more appropriately. Additionally they should check for correct English wording.

Significance

In summary, the authors describe the HAT-field test as a simple PoC test for the detection of SARS-CoV-2 antibodies in patients. Because of its ease of use and robustness, the test appears to be particularly well suited for use in countries with underdeveloped health care or limited testing facilities, as also reported previously. The value of this manuscript lies mainly in the detailed description of the protocol and its validation. In this context, the adaptations described are certainly useful and helpful from a practical point of view, but do not provide significant new scientific insights. In light of these considerations, we recommend that this work be submitted to an appropriate journal specializing in the publication of such methods

Expertise The reviewers have established and published different serological assays to monitor immune responses against SARS-CoV-2

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>

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

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

Summary:

Estrach and colleagues seek to identify the ECM components that are key to regulating hair follicle stem cell (HFSC) activation using the highly-characterized mouse hair follicle as a model. They first use a targeted approach to examine key ECM components expressed by HFSC and find that Fibronectin (FN) is highly expressed. Further, wholemount analysis of the hair follicle reveals a meshwork of FN enveloping the hair follicle. They hypothesize that FN is a fundamental regulator of hair follicle (HF) cycling and then proceed to carry out longterm studies required to examine hair follicle cycling and knockout FN with two different HFSC Cre lines (Lrig1 and Krt19), as well as integrin coreceptor SLC3A2. They clearly show that absence of Fibronectin (FN) and SLC3A2 is detrimental to hair follicle stem cell activation and cycling (FN) and hair follicle identity (SLC3A2).

Overall comments:

The authors use the tail hair follicles as a model similarly to the highly-characterized, synchronous back skin hair follicles. However, the tail hair follicles are asynchronous (Braun et al. 2003, PMID: 12954714), thus reporting the age of the mouse from which the tail whole mounts came from is not sufficient to claim a HF cycle disorder - HF should be imaged in an unbiased manner and subsequently quantified for phase. The manuscript would greatly benefit from including more information in the figure legends, such as age of mice, number of mice and HF quantified, as well as what the error bars represent. Further, in samples where many HF were counted per mouse, these should be averaged and then the average per mouse displayed; super plots would be great to use here.

Major comments:

1. In Figure 1, the use of tail whole mount images indeed provides striking display of the fibronectin meshwork that envelops the hair follicle. However, addition of a marker of the regenerative phase (e.g. proliferation) and resting phase would provide more convincing evidence that this is the particular phase of the hair cycle that you have captured, especially given my overall comment regarding the asynchronous nature of the tail HF cycle.
2. The authors show that FN is expressed in early-mid anagen and conclude that FN is a regenerative signal. This claim should be substantiated with FN staining on more time points across the HF cycle to substantiate the argument that it is a regeneration-specific signal, found only in the telogen-anagen transition.
3. Lrig1-cre and K19-cre-mediated FN knockout result in HF that are thinner at D158 - this is not immediately apparent from histological sections. Can you use your thick sections to give better perspective?
4. The authors measure the width of the infundibulum from lightsheet microscope images. It is a bit difficult to position whole tissues using this technique, and the images that are shown are not from the same perspective, and thus measurement of the width is not accurate from these images. I suggest either removing this analysis or using more comparable images. Further, if this is a true phenotype, can you speculate on what the thickened infundibulum might mean?
5. The authors then show mislocalization of Lrig1+ cells to the infundibulum in absence of FN. Are other stem markers localized to the infundibulum or outside of the bulge? Further, what might the mislocalization of Lrig1+ cells might mean?
6. Please explain your conclusion after Figure 3i and at the end of the manuscript that states that FN is required for stem cell anchorage. I think that a very plausible explanation is that FN is required for stem cell function and identity, but anchorage of the SC lacks sufficient evidence. Further, your only evidence to support the anchoring theory only comes from expression of Lrig1 in FN knockout and no other markers. Are they also mislocalized? Please either tone down this conclusion on SC anchorage or provide stainings for more SC markers to show mislocalization in absence of FN.
7. In Figure 3l-o, you examine proliferation on the control vs the conditional deletion of FN in D30 and D158. However, in D30, these tissues are not at all directly comparable since one is obviously in anagen and the knockout in telogen. You must compare the anagen knockout sample, although this occurs a bit later than the control. Further, how was the infundibulum distinguished from the bulb in these control images?
8. In Figure 3P, you carry out RT-qPCR on whole skin to detect HFSC markers. This should have been carried out on sorted epithelial cells as isolation of whole back skin introduces bias to the system in that the number of stem cells may artificially look different in skin that is in anagen vs skin that is in telogen as the anagen skin has a different proportion of SC to progenitor cells to dermal cells. This concern is also similar to point 9 - the control and FN knockout at D30 are not comparable given that they are in different phases of the hair cycle.
9. Figure 4a these images need to be of the whole mouse - it is not possible to determine what we are looking at or where - there is not even a scale bar.
10. After Figure 4, you argue that because fibronectin expression resolves from healing dermis is the reason that hair follicles do not form, and site Dekonick and Blanpain (PMID: 30602767) - however this review makes no mention of the dynamics of fibronectin in wound healing. Further, evidence from Driskell et al (2013, PMID: 30602767) would suggest that it is the fibroblast population that responds to the wound that determines whether HF regenerate. And further, very large wounds do regenerate HF (Ito et al PMID: 30602767). In addition, this would all be fibroblast-derived FN, as opposed to the current study which examines keratinocyte-derived FN. Please reconsider this argument.
11. The authors knockout SLC3A2, an integrin coreceptor that is localized to the plasma membrane. They show a very similar, yet more severe phenotype to the Lrig1- and K19- mediated knockout of FN. Given the bidirectional communication that SLC3A2 is responsible for, can you reconcile whether the defects in the HF cycle and the HFSC are a result of outside-in or inside-out signaling? Further, is it possible that integrin function regulated by SLC3A2 is necessary for more than FN assembly? This could be especially relevant given that your targeted screen also identified Col17A1, which is well known to be required for HFSC function (Matsumura et al., PMID: 26912707)
12. It is intriguing that in the absence of HFSC-derived SLC3A2 that no FN network forms. Is FN expressed or is the assembly perturbed in the absence of properly functioning integrins? The authors conclude that the signaling cascade flows from fibronectin to integrin to SLC3A2, but do not test where the FN phenotype arises in the SLC3A2 knockout - is it due to aberrant assembly of the FN meshwork or a change in transcriptional or translational levels?
13. In the grafting assay in Supplemental Figure 3, keratinocytes undergo a de novo hair follicle morphogenesis - is Lrig1 expression maintained in order to carry out cre-mediated deletion? Further, the fibroblasts in this assay may adopt a wound-like phenotype, expressing FN, which you earlier claim to be required for hair follicle production in wounds. Yet in the absence of epithelial FN, no HF form. Can the authors reconcile this?

Minor comments:

1. In Figure 1a, the two populations are Lgr5+ and basal; please define what the basal population is in this experiment.
2. Significative is not a word.
3. In Figure 4 figure legend, there is reference to a grafting experiment but no experiment shown.
4. The authors delete FN in Lrig1+ or K19+ cells starting D19 and harvest at D30, and conclude that the hair follicles do not enter anagen after the second telogen, can you please include the data supporting the statement that mutant HF did not reenter the hair cycle after D65.

Significance

The authors show for the first time that fibronectin is expressed during cutaneous homeostasis and that it is required for normal function of the hair follicle stem cells. This is significant conceptual advance for the field of skin biology because fibronectin is thought to only be present in wounds: derived first from infiltrating serum and second from fibroblasts to act as provisional dermal ECM to support epithelialization during wound-response, which is ultimately resolved upon the conclusion of wound healing (reviewed in: Singer and Clark, PMID: 10471461). Further, FN has also been characterized as an EMT marker during cancerous progression (Lamouille et al, PMID: 24556840). Estrach and colleagues show that fibronectin is actually expressed by hair follicle stem cell keratinocytes and then is assembled into a meshwork that envelops the hair follicle and is in fact necessary for hair follicle stem cell homeostasis. This work would be broadly interesting to the field of stem cell biology as well as those working on extra cellular matrix signaling. My field is epithelial stem cells and more specifically hair follicle development and cycling.

Referee Cross-commenting

I have no disagreement with any of the points raised by the other reviewers. In fact, we seem to agree on the majority of the concerns. This includes the use of the tail wholemount model, the use of Lrig1-cre, selection of timepoint vs phase of the hair cycle, the appropriateness of the link between Fibronectin and SLC3A2, and further significant issues related to display of data and their reproducibility. Further, all of the major comments raised need to be addressed in order to properly evaluate the conclusions that the authors make. In my opinion, none of the comments raised here are unreasonable.

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

In this manuscript, Estrach et al., investigated whether extracellular matrix component fibronectin function in hair follicle regeneration, using a range of approaches including FACS, RT-qPCR, immunofluorescent staining, and mouse genetics. They proposed that fibronectin in Lrig1+ cells was necessary for hair follicle stem cell maintenance and activation, and the fibronectin expression and assembly relayed on the integrin co-receptor SLC3A2.

Significance

Major points:

1. In Figure 1a, the author used Lrig1+GFP and a6 to isolate Lrig1+ cells in the infundibulum junction zone above the sebaceous gland at Day 28. However, in Figure 1f-h, the GFP expression was not only in infundibulum above SG, but also in some inner root sheath cells. Since the Lrig1+ cells do not include the hair follicle stem cells (CD34+ bulge cells), result in Figure 1a does not support fibronectin expression in HFSCs.
2. In Figure 1b-e, the author detected fibronectin expression by IF staining with tail skin whole mount and back skin section. The fibronectin is mainly detected in differentiated cells in the inner root sheath (IRS) in anagen (Figure 1b and 1d), upper IRS in catagen (Figure 1c), hair germ in telogen (Figure 1e), but not in the bulge region (Figure 2i-l). Again, these results do not support fibronectin enrichment in HFSCs either.
3. In Figure 2a-c, the author knocked out fibronectin in Lrig1+ cells with Lrig1-CreERT2, FN fl/fl mice, and then validated the knockout efficiency by IF staining. However, result shows the fibronectin expression was not only depleted in the GFP+ Lrig1+ cells, but also depleted in GFP- inner root sheath and matrix. Similarly in Figure 4n, fibronectin was only knocked out in Lrig1+ cells, however, result showed the fibronectin cannot be detected in any cell types in skin. The author should explain why fibronectin depletion in Lrig1+ cell lead to completely fibronectin depletion.
4. In Figure 2r, by Flow cytometry, the author found a significative reduction of a6+CD34+ SC population when fibronectin is conditional knocked out in Lrig1+ cells. As the Lrig1+ cells and a6+CD34+ HFSCs are two distinct cell populations, the author needs to explain how fibronectin depletion in Lrig1+ cells affect the number and activation of HFSCs population.
5. In Figure 3b-c and 2g-h, the author reported the HF thinning in Lrig1-CreERT cKO mice by back skin HE staining. However, by tail skin wholemount staining, the HF thinning was not observed in those cKO mice (Figure 3f-g; Figure 2k-l). The author needs to explain the discrepency. In addition to this, the low quality of HE staining and poor orientation of HF (Figure 2h and 3b), coupled with lack of quantification, made these results and conclusion unconvincing.

Mini Points:

1. Color information in each IF staining panel is only completely presented in Figure 4. It is incomplete or lost in other 4 Figures.
2. In Figure 1a, FACS profile and representative figures are needed to demonstrate the author isolated the correct population at correct hair follicle stage.

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

In the manuscript, Estrach et al addressed the role of skin epithelial stem cell-derived extracellular matrix in hair follicle regeneration. They found that Lrig1+ epithelial stem cells highly express fibronectin gene compared to other basal epithelial cells. Conditional deletion of fibronectin gene using Lrig1CreERt2-GFP or K19CreER (the latter is expressed in the bulge stem cells) resulted in hair follicle regeneration blockade, change in the expression pattern of Lrig1-GFP. Injection of fibronectin protein into the dermis of the conditional fibronectin mutants (Lrig1CreERt2-GFP) rescued the hair regeneration blockade phenotype. The authors also conditionally deleted SLC3A2, an integrin coreceptor, using the same CreER lines and found a decrease in fibronectin deposition and CD34+ bulge stem cell number. With the results from these mouse genetics and phenotype analysis, they conclude that fibronectin-SLC3A2 cascade finely tunes hair follicle stem cell fate and their tissue regenerative capacity.

Major comments

1 Immunostaining results of fibronectin in the tail epidermal wholemounts are not convincing enough and would require improvements. First, the tail epidermal wholemounts lack the mesenchymal matrix and the basement membrane (Fig. 1b, c, f-h; Fig. 2b, c; Fig. 5d-j; Fig. S1b, c) (Braun et al., 2003 Development (PMID: 12954714)). Fibronectin is localized mainly in the mesenchymal matrix and the basement membrane in the skin and other organs (Stenman and Vaheri, 1978 JEM (PMID: 650151); Couchman et al., 1979 Archives of Dermatological Research (PMID: 393184); Jahoda et al., 1992 J. Anat (PMID: 1294570)), thus this sample preparation method is not appropriate to assess fibronectin tissue distribution. The authors use thick back skin sections, which contain entire skin tissues, thus I would recommend this method. Furthermore, fibronectin antibody signals in the tail epidermal wholemounts are detected in the inner part of the hair follicle epithelium, where there is no expected ECM structure (see Couchman et al. and Jahoda et al. above). Consistently, fibronectin signals are localized inside the Lrig1-GFP+ epithelial basal cells (Fig. 1f-h). Thus the specificity of the fibronectin staining needs to be confirmed. The reviewer understands that the authors provide an image showing the great reduction of fibronectin staining in a D30 tail epidermal wholemount of 4-OHT-treated Lrig1CreERt2GFP,FNfl/fl mice (Fig. 2c). However, as the D65 tail epidermal wholemount from wildtype mice also show many hair follicles without fibronectin signals (Fig. 1c), rigorous assessments would be required.

2 Lrig1+ stem cells have been reported to maintain the upper pilosebaceuos unit, containing the infundibulum and sebaceous gland, but contribute to neither the hair follicle nor the interfollicular epidermis under normal homeostatic condition (Page et al., 2013 Cell Stem Cell (PMID: 23954751)). However, only 11 days after the first 4-OHT treatment on Lrig1CreERt2GFP;FNfl/fl mice, Estrach et al found the defects in hair cycle blockade, reduced cell proliferation in the hair bulb, and significant reduction in fibronectin deposition in entire hair follicle structure. Please explain how the deletion of fibronectin gene in Lrig1+ stem cells, which do not contribute to hair follicle lineages, lead to significant hair regeneration defects in a short period of time. Current data do not well explain a causal relationship between the genetic perturbation and the observed phenotypes.

3 In some experiments (listed below), description about the methods, replication and statistics is not adequate, raising concerns about reproducibility. 3.1 Fig. 1a: data variation for basal cells should be presented. Biological replicate number should also be indicated in the figure legend. 3.2 Fig. 2g, h: hair follicle thinning is described here, but only one HE staining image with only one hair follicle is not enough to support this important claim. 3.3 Fig. 2r, 3i: flow cytometric data should be presented. 3.4 Fig. 4: No biological replicate and reproducibility information are provided. 3.5 Fig. 5j: how many biological replicates and hair follicles were analysed? The authors should also perform statistical tests. 3.6 Fig. S3g, h: information for biological replicates should be described. Statistical tests should be applied to Fig. S3h. 3.7 Fig. 5k-n: only one HE staining panel from each mouse line cannot provide rigorous evidence of the defects, which are not obvious from the HE staining.

4 In Fig. 3j, k, n, hair follicles in the control and 4-OHT treated skin are in different hair cycle phases. Therefore there is a possibility that the difference in their PCNA pattern simply reflects the difference in the cell proliferative activity between different hair cycle phases, but not indicates direct effects from the deletion of fibronectin gene in Lrig1+ cells.

5 To assess the expression levels of signaling-related genes (Fig. 3p, S2), the authors used mRNA extracted from whole skin tissues, which contain all epithelial and mesenchymal cell populations in different hair cycle phases. Thus, the time and spatial resolution of the analysis is low and it also cannot eliminate confounding factors derived from the difference in hair cycle phases between control and cKO.

6 In order to provide the characteristics and purity of the FACS isolated cell populations at D28 (Fig. 1a), their flow cytometry data and some marker gene expression data should be presented (see Page et al., 2013 Cell Stem Cell). This assessment is particularly important for the skin compared to other static organs, as it exhibits dynamic gene expression and tissue structural changes during the hair cycle. It is also important to check whether fibronectin protein accumulates around Lrig1+ stem cells in D28 dorsal skin, where upregulation of fibronectin gene expression was detected. The authors should not use tail epidermal wholemounts for the reason described above.

Minor comments

7 The increase in stem cell marker expressions shown in Fig. 3p contradicts to the reduction in the number of bulge stem cells shown in Fig. 2r and 3i. Please provide an explanation for this apparent discrepancy.

8 Although Fig. 5s-v show reduction of a6+CD34+ bulge cell population, the bulge tissue structure can be observed in Fig. 5p. Please explain how to interpret this apparent discrepancy. They just lost the expression of CD34?

9 Connectivity of the data in the fibronectin cKO with that of SLC3A2 cKO is weak. For example, it could be strengthened if the authors show colocalization of fibronectin and SLC3A2 in vivo.

10 Although the format of the manuscript is free in Review Commons, the Introduction and Discussion of this manuscript are too brief for us to understand the background and significance of this study. So I would recommend the authors to provide more detailed background information and discussion.

11 The authors use the term 'HFSC', but it is unclear which stem cell populations they mention; bulge, Lrig1+ or other stem cell populations?

12 Please provide details of fibronectin protein and antibody used in this study.

13 Due to the short for experimental information for Fig S3a, b, d, e, I cannot evaluate the data, thus several questions are raised. As the SLC3A2 level was significantly reduced in most cells in the plot, I assumed that Lrig1-GFP+ cells were gated before examining the expression level of SLC3A2. However, no information on the procedures for 4OHT treatment, isolation of cells and flow gating strategy is described. In the case of K19CreER mice (Fig. S3d, e), if the authors gated GFP+ cells before analysis, what GFP means in this case?

14 The manuscript wants to be checked for copyediting.

Significance

These findings might provide a conceptual advance into the role of epithelial stem cell-derived extracellular matrix in regulating stem cell behaviour and tissue regeneration. As fibronectin is upregulated in development, wound healing and cancers in many other organs, their findings may point to the importance of fibronectin in activating tissue progenitors and stem cells in these processes. Thus, this manuscript is likely to be of interest to a wide range of readers, not only in skin biology, but also in stem cell, regenerative and matrix biology. The contributions of this paper could be enhanced if the documentation were to be made stronger and more rigorous in a revised manuscript.

Referee Cross-commenting

I totally agree with Reviewer #3's comments in this consultation session. I have no disagreement with any of the points raised by other reviewers.

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]

This section is optional. Insert here any general statements you wish to make about the goal of the study or about the reviews.

We are grateful to the reviewers for their honest opinion regarding this work and plan to address the majority of the comments in a revised version either through new analysis or revision to the text, as we believe these will improve the manuscript by making some of the details clearer. There were few suggestions that will lead to substantiative changes to the findings. Here, we address the most salient critiques, the primary one being related to novelty.

We respectfully disagree, as our detailed analysis of the DNA methylome in Octopus bimaculoides represents a significant advance to understanding how the epigenome is patterned in non-model invertebrates in general, and cephalopods in particular. We acknowledge that the previous report that the octopus methylome resembles the few other invertebrates where low DNA methylation has been found, the finding was part of a multi-organism study last year (de Mendoza et al., 2021), which lacked any detailed investigation. Our study provides the first in depth analysis on methylation patterning, the relationship with transposons and gene expression, and reports the finding of other key epigenetic marks in O. bimaculoides, and in other cephalopods.

In short, we believe our study to be highly novel and that it represent the first analysis of this kind in cephalopods and one of the few existing in non-model invertebrate organisms. In addition, we identify the conservation of the histone code in cephalopods. While this may be expected, this is the first experimental evidence in this class and represents an important step forward to understand the epigenetic regulation of genes and transposons in invertebrates. Finally, we plan to provide an updated transcriptome annotation for O. bimaculoides that will be available for the scientific community as a new valuable resource. We believe these features will make this study highly cited.

We believe that findings like ours will complement several recent studies that extend the epigenetics field out of the current narrow focus on model organisms to understand how epigenetic mechanisms function in diverse animals. This provides new insights regarding the epigenetic mechanism of gene regulation in an emerging invertebrate model.

2. Description of the planned revisions

Insert here a point-by-point reply that explains what revisions, additional experimentations and analyses are planned to address the points raised by the referees.

Reviewer 1 raised the following points that we are planning to address:

*- It is unclear why the authors did not use the original gene models of O. bimaculoides or tried to improve them. By only relying on adult tissue (but the relatively late hatchling stage), they would have omitted most developmentally expressed genes, that are incidentally also the ones that are subjected to extensive spatiotemporal gene regulation (which is also a problem to assess the role of methylation). I think more comparisons with existing gene models and how the newly generated stringtie models should be provided. *

We agree that using as many tissues and developmental stages as possible will expand the octopus transcriptome.

We plan to:

• Add RNA-seq data from stage 15 embryos to improve this.
• Compare the gene model used in the original version of the manuscript (Stringtie model to use in Trinotate for improving the annotation of the genes) to the existing annotation model and report on which has superior performance for annotating the * bimaculoides* transcriptome.
• Extend the annotation of the transcriptome which we undertook in a focused fashion in the first iteration of this manuscript. Reviewer 2 raised the following points that we are planning to address:

*- It is not exactly clear to me why the authors look for expression clusters in the first part of the manuscript? This information, while interesting, does not seem to be used in the methylation analysis. It is also somewhat contradictory because the authors first claim that, based on their GO-term enrichment analysis, that different expression clusters are associated with "complex regulatory mechanisms, potentially based in the epigenome". Yet at the end they conclude that, due to the global and tissue-overarching nature of methylation, this "argues against this epigenetic modification as a player in the dynamic regulation of gene expression". *

We thank the reviewer for pointing out this issue and we plan to clarify the point through changing the text and additional analysis. Since we found that the methylation pattern was stable across tissues, and that it corresponded to gene expression levels regardless of tissues, we concluded that the methylation pattern is not likely relevant for the tissue-specific gene expression pattern reported in Figure 1.

We plan to:

• Ask whether there is a correlation between the gene clusters generated in Figure 1 and the DNA methylation patterns identified in Figure 4. *- At least for the trees that are shown in the main figures it would be great to show support values. *

We thank the reviewer for this request.

We plan to:

• Add full Supplementary information regarding the support values in Supplemental Files for all the trees present in the main Figures. Reviewer 3 raised the following points that we are planning to address:

*- It would be great to see more data on cephalopod TET and MBD structure. For example, it would be interesting to know whether octopus TETs have a CxxC domain or whether MBD proteins harbor functional 5mC - binding domains. *

We agree that it would be of interest to examine the conservation of TET genes to expand upon the initial analysis by Planques et al 2021 showing that O. bimaculoides have one TET homolog, one MBD4 homolog and one MBD1/2/3 homolog. Detailed analysis of MBD4 protein has been already performed in de Mendoza et al. 2021 by using the protein sequence of O. vulgaris, as the MBD4 gene in the O. bimaculoides genome appears truncated.

We plan to:

• add the PFAM domain analysis for TET proteins This will be added as a new figure panel.
• Update the text to include the reference to the identification of MBD4/MECP2 as the invertebrate homologs of vertebrate MBD4. *- Even though RRBS provides limited insight into DNA methylation patterns, the authors could have done more to explore read-level 5mC information. For example, by studying single reads, the authors could deduce the numbers of fully methylated, unmethylated or partially methylated reads. Such analyses might provide valuable insight into potentially different modes of epigenetic inheritance in different tissues i.e are there tissues that favor fully methylated or unmethylated stretches of DNA vs tissues that favor partial methylation? *

We think this is a really interesting point. This has been partially addressed in a previous work (de Mendoza et al., 2021) which found limited to no partially methylated reads in whole-genome bisulfite sequencing from O. bimaculoides brain.

We plan to:

• Use Proportional Discordant Reads on the RRBS data we generated from 30 dpf hatchlings to assess the presence of discordant methylated reads across different tissues to assess possible different modes of epigenetic inheritance. We will use “WSHPackage” in R: https://academic.oup.com/nar/article/48/8/e46/5760751?login=true https://github.com/MPIIComputationalEpigenetics/WSHPackage/blob/master/vignettes/WSH.md#x1-50003.1 This will be added as a new figure panel.

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.

Reviewer 1 raised the following points that we have already addressed:

We addressed all the comments raised by this Reviewer by revising the text, fixing references, typos and improving clarity.

Reviewer 2 raised the following points that we have already addressed:

We addressed all the minor Comments raised by this Reviewer regarding spelling errors and Supplementary Figures.

- The finding that less than 10% of all possible sites are methylated is surprising. I could not (easily) find statistics of RRBS experiment read mapping to the genome.

We have now provided this data and new Supplemental Table 1 (refereed in the text as Table S1).

*- It is very exciting to see methylation of gene bodies and some correlation to their expression levels, but the authors may need to include a disclaimer that the methylation of TEs may go undetected due to the gapness of the genome. In fact, the authors may try to map their data onto a somewhat closely related Octopus sinensis genome sequenced with long reads available at NCBI to confirm overall pattern. It is likely though that due the evolutionary distance only gene bodies will have mapping. *

The thank the reviewer for this suggestion and we included a sentence in the Result session indicating that methylation of TEs may go undetected due to the poor annotation of the octopus genome.

*- The statistical reasoning (and methodology) behind how clusters in Figures 1 and 4 were defined is unclear. In particular, in Figure 4, it seems that the authors had asked the program to give four clusters in total - why was this number chosen? It seems that using the same generic clustering approach as in Figure 1 may benefit or confirm the results in Figure 4. *

We clarified the rationale in the Material and Methods session to describe the bioinformatic analysis. We will put the full code used in the manuscript in our GitHub page (https://github.com/SadlerEdepli-NYUAD/) to have a more comprehensive understanding of the Method used.

Reviewer 3 raised the following points that we have already addressed:

We addressed all the minor comments in the text and figures raised by this reviewer regarding typos and clarity.

*- There is little info on the generated 5mC data. To bolster its value as a resource, the manuscript should have a link to the table describing RRBS metrics. This should include: non-conversion rates, numbers of sequenced and mapped reads, read length and other info that the authors deem useful. *

We have now provided this data in a new Supplemental Table 1 (refereed in the text as Table S1).

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.

Reviewer 1 raised the following points that we are not planning to address:

*- The newly sequence RNA-seq samples are using a ribodepletion protocol (RiboZero) while the other ones are using a polyA selection. This might be a slight problem to compare them quantitatively. Actually in the Figure 1, all 4 newly generated samples group together in the hierarchical clustering. *

We acknowledge the reviewer’s point here and agree that heterogeneity in library prep and batch is a common issue when comparing public available with newly generated datasets. This could account for the clustering of the Ribosomal RNA depleted (i.e. RiboZero) from polyA selected RNA libraries. While this could potentially introduce bias, we do not believe that it substantially alters any of the main findings or the interpretations of this data. Our purpose for carrying out the cluster analysis of transcriptomic data from multiple tissues was to identify distinct gene patterns that defined different tissue types. This was accomplished regardless of the potential confounding variable introduced by different library preparations. In addition, we used TMP which seems to help in the comparison across different samples when used for qualitative analysis such as PCA and cluster analysis (Zhao et al. 2020; DOI: http://www.rnajournal.org/cgi/doi/10.1261/rna.074922.120). Therefore, even if not ideal we think that this approach is still valuable.

*- I am not so sure about the way the authors used z-score normalized logTPMs and applied hierarchical clusters, this most likely would not fully alleviate the impact of expression level on the outcome compared to more advanced form of normalization and clustering. *

We agree with the reviewer that applying z-score or a logTPMs normalization would not fully resolve the technical variance in the direct comparison of libraries generataed with different RNA selection methods. We did not apply z-score on logTPMs but these 2 methods were applied separately: z-score on TPMs in Figure 1B to define the gene clusters and log2(TPM+1) in Figure 4E. We have clarified the text to reflect this.

*- I am not convinced that differences in western blot for histone modification could really provide a clear insight into their regulatory role. *

We agree with the reviewer that Western blotting for histone modifications does not provide deep insight into their regulatory role. However, this is the first description of these marks in any cephalopod, and we believe that reporting a finding from experimental evidence is important, even if the result is aligned with the existing paradigm. Moreover, the marked difference in levels of distinct histone marks across tissues supports the hypothesis that they play a regulatory role. We observed this in mice where difference abundance in western blot correspond to different abundance and enrichment also by ChIP-seq (Zhang et al., 2021 DOI: https://doi.org/10.1038/s41467-021-24466-1). Considering the limited tools available in this species, we still consider this an important finding.

Reviewer 2 raised the following points that we are not planning to address:

*- The finding that less than 10% of all possible sites are methylated is surprising. I could not (easily) find statistics of RRBS experiment read mapping to the genome. I also wonder how much the gap-richness of the genome may affect the overall methylation estimate. If assembly permits, would it make sense to limit the sampled sites to areas where no flanking gaps are present (and sufficient scaffold length is available, maybe excluding very short scaffolds)? *

We added all the statistical values regarding the RRBS in a NEW Supplemental Table 1. We used a single base pair analysis approach (not tiling windows), so the data we extracted is not biased by the length of the scaffolds. This is confirmed by the fact that the DNA methylation value obtained in our RRBS data matches the findings observed in Whole Genome Bisulfite Sequencing (WGBS). Moreover, global DNA methylation values assessed by Slot blot analysis as a technique independent from genome assembly confirmed what observed with RRBS.

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 by Macchi et al describes the epigenome and the transcriptome of Octopus bimaculoides. While the manuscript itself is well written and the data are properly analyzed, it is fair to say that the work itself offers little biological novelty. Nevertheless, I still believe that the datasets and some of the analyses could be useful to researchers studying invertebrate epigenomes and gene regulation.

1. It would be great to see more data on cephalopod TET and MBD structure. For example, it would be interesting to know whether octopus TETs have a CxxC domain or whether MBD proteins harbor functional 5mC - binding domains.
2. There is little info on the generated 5mC data. To bolster its value as a resource, the manuscript should have a link to the table describing RRBS metrics. This should include: non-conversion rates, numbers of sequenced and mapped reads, read length and other info that the authors deem useful.
3. Even though RRBS provides limited insight into DNA methylation patterns, the authors could have done more to explore read-level 5mC information. For example, by studying single reads, the authors could deduce the numbers of fully methylated, unmethylated or partially methylated reads. Such analyses might provide valuable insight into potentially different modes of epigenetic inheritance in different tissues i.e are there tissues that favor fully methylated or unmethylated stretches of DNA vs tissues that favor partial methylation?

Minor comments:

There are a few spelling errors throughout the manuscript. Please check for those: Figure 4F ("Trascrips" instead of transcripts), Schmedtea instead of Schmidtea. There are likely other errors as well.

Page 3 - "intergenome"sounds a bit weird.

The authors might consider citing Planques et al, 2021 (BMC Biol) alongside Mendoza et al when discussing unusually high 5mC levels in the sponge.

Significance

The main points of the paper are: i) a somewhat improved transcriptome, ii) DNA methylation data generated by RRBS that follows a canonical invertebrate pattern (low 5mCG levels present in GBs and absent from repeats), and iii) evolutionary analyses of epigenetic machinery components. While lacking biological novelty, the presented data have a resource value and could likely serve as a decent starting point for further exploration of cephalopod gene regulation. I therefore believe that with some revision the manuscript will merit publication in one of the Review Commons - associated journals.

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 paper by Macchi et al studies DNA methylation patterns in Octopus bimaculoides, describing overall conservation of DNA methylation machinery and genome-wide methylation patterns and their effect on gene expression across broad tissue sampling. As such, the paper comrpises a key advancement in the emerging field of cephalopod (epi)genomics and gene regulation. Despite the difficulties relating to the genome assembly of O. bimaculoides, the authors have done a solid analysis of methylation patterns and the results look generally sound. I have a few points that may help the authors improve their manuscript:

• The finding that less than 10% of all possible sites are methylated is surprising. I could not (easily) find statistics of RRBS experiment read mapping to the genome. I also wonder how much the gap-richness of the genome may affect the overall methylation estimate. If assembly permits, would it make sense to limit the sampled sites to areas where no flanking gaps are present (and sufficient scaffold length is available, maybe excluding very short scaffolds)?
• It is not exactly clear to me why the authors look for expression clusters in the first part of the manuscript? This information, while interesting, does not seem to be used in the methylation analysis. It is also somewhat contradictory because the authors first claim that, based on their GO-term enrichment analysis, that different expression clusters are associated with "complex regulatory mechanisms, potentially based in the epigenome". Yet at the end they conclude that, due to the global and tissue-overarching nature of methylation, this "argues against this epigenetic modification as a player in the dynamic regulation of gene expression".
• It is very exciting to see methylation of gene bodies and some correlation to their expression levels, but the authors may need to include a disclaimer that the methylation of TEs may go undetected due to the gapness of the genome. In fact, the authors may try to map their data onto a somewhat closely related Octopus sinensis genome sequenced with long reads available at NCBI to confirm overall pattern. It is likely though that due the evolutionary distance only gene bodies will have mapping.
• At least for the trees that are shown in the main figures it would be great to show support values.
• The statistical reasoning (and methodology) behind how clusters in Figures 1 and 4 were defined is unclear. In particular, in Figure 4, it seems that the authors had asked the program to give four clusters in total - why was this number chosen? It seems that using the same generic clustering approach as in Figure 1 may benefit or confirm the results in Figure 4.
• In the discussion Scmidtea is misspelled.
• Some supplementary figures have to be exported as spell checker highlights are still present (e.g., in Suppl Fig 4).

Significance

This manuscript is an important step towards understanding the workings of gene regulation at the epi-genomic level in octopus and cephalopods in general

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 focuses on the role of DNA methylation and histone modification in the gene regulation of cephalopods. It complements recently published RNA-seq and MethylSeq datasets with a few extra samples and generally confirms previous findings that DNA methylation does not play an active role in tissue or stage-specific regulation of gene expression in cephalopods (which is the general rule for most non-vertebrates). I don't see any methodological issue serious enough to preclude publication but some details should be strengthened.

• the newly sequence RNA-seq samples are using a ribodepletion protocol (RiboZero) while the other ones are using a polyA selection. This might be a slight problem to compare them quantitatively. Actually in the Figure 1, all 4 newly generated samples group together in the hierarchical clustering.
• It is unclear why the authors did not use the original gene models of O. bimaculoides or tried to improve them. By only relying on adult tissue (but the relatively late hatchling stage), they would have omitted most developmentally expressed genes, that are incidentally also the ones that are subjected to extensive spatiotemporal gene regulation (which is also a problem to assess the role of methylation). I think more comparisons with existing gene models and how the newly generated stringtie models should be provided.
• I am not so sure about the way the authors used z-score normalised logTPMs and applied hierarchical clusters, this most likely would not fully alleviate the impact of expression level on the outcome compared to more advanced form of normalisation and clustering.
• I am not convinced that differences in western blot for histone modification could really provide a clear insight into their regulatory role

Significance

This manuscript reports confirmatory results, partly reanalysing and confirming previous work. I would also like to stress that the methylation results have already been reported and discussed in a previous paper (de Mendoza et al. 2021). I don't have a fundamental problem with this but I also find the paper slightly overambitious and unspecific in its goals. I think it should benefit from being made slightly more concise. I find the part of histone marks is quite overstated. These marks are quite universal in eukaryotes and generally demonstrated to play a regulatory role, the fact that they can be detected in cephalopods by western blot is therefore not really a result.

Comments on the text (difficult without line numbers):

• Intro, first section: it would be good to have a few more references
• "While this has been extremely fruitful in elucidating detailed mechanisms of epigenome patterning, regulation and function, they do not provide a comprehensive understanding of the multiple and varied ways that the epigenome functions." -> sentence is quite confusing and without very clear meaning
• "In contrast, the most common invertebrate model organisms - Caenorhabditis elegans and Drosophila melanogaster - lack DNA methylation entirely. " -> could sound like this is the case for more invertebrates.
• P4 "This is the case in many animals, " -> give examples, it is unclear which examples of TE control by methylation outside vertebrates have been corroborated by data. The paper cited do not deal with methylation in squid
• Evolution has selected for variations in the canonical patterns of methylation -> such explanation could also be consistent with neutralistic explanations
• p19: Schmidtea (type) "such as the planarian Schmedtea mediterranea"

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

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 #2

Evidence, reproducibility and clarity

Summary:

The manuscript entitled "Lhx2 is a progenitor-intrinsic modulator of Sonic Hedgehog signaling during early retinal neurogenesis" by Li et al is a very interesting study in which the importance of Lhx2 is studied in conditional knock-out background to decipher the importance during retinal neurogenesis of developing embryo. The study reveal importance of co-receptors essential to Shh signalling. Data presented are clean and would add depth of knowledge to the literature. The study/manuscript do have some lacunae which are listed below which would be good to address before it is published

Major comments

1. Authors do not seem to have performed a rescue experiment with Lhx2 CKO. If this is absolutely not possible, a conditional overexpression could have given confirmatory clues on the Lhx CKO phenotype discussed. In any case some sort of rescue experiments are essential with respect to Lhx2 as done with Cdon and Gas1
2. Also, overexpression studies with the following genes, Cdon, and Gas1 is interesting in the CKO background. What about their over expression phenotype in Wild-type, Ptch1-CKO, and purmorphamine/Shh-N treated conditions. Alternatively, an ex-vivo or in vitro approach using cultured cells may also prove worthy to prove this point.
3. A logical explanation of Tamoxifen administration at a window E11.5 - E15.5 is good to have in the results section.
4. Authors say ".........that Lhx2-deficient RPCs can respond to recombinant Shh-N at more physiologically relevant concentrations, but their response is still attenuated compared to Lhx2-expressing RPCs." If the Shh-N dose is increased above physiological concentrations in Lhx-CKO conditions does the Gli1 read-out will restore to normalcy ? This would give insight on to the roles of co-receptors such as Cdon, Gas1.

Minor comments

1. In the 'Introduction' the authors write "Pathway activation is not achieved, however, by simple ligand binding to Patched, but also requires one of three co-receptors: Cell Adhesion, Oncogene Regulated (Cdon), Brother of Cdon (Boc), or Growth Arrest Specific 1 (Gas1)." Please give a citation of appropriate literature.
2. In the introduction authors write "In this case, an interaction with Notch signaling is partly responsible, through Lhx2-dependent expression of ligand (Notch1), receptors (Dll1, Dll3), and downstream transcriptional effectors (Hes1, Hes5)" I feel Dll1 and Dll3 are ligand and Notch is receptor. Please check.
3. Figure 1c: Western blots tubulin is less in CKO alongside Gli1. It would be better to do quantification for showing any significant changes.
4. Figure 2A and Result 2: Why harvesting at E14.5 specifically for qPCR/ChIP sequencing, while for RNA sequencing and in situ , it was done E15.5.
5. Figure 5D, E: It is not clear why there were more mCitrine+ cells in CKO explants at 96 hours.
6. Figure 6 C,D,E,F: Does Lhx2 CKO, cause cell death as the levels of vsx2 are low in vehicle in CKO (D,F) as compared to ctrl (C,E).
7. The levels of Gli1 are higher in CKO vehicle (D,F) than the control panel (C,E). This difference in Gli1 expression is more evident D versus C. Does it mean that CKO increases Gli1 expression in explants, which seems to be opposite of what shown in the first results (Figure 1D, E).
8. It is not clear why the increase in Gli1 expression with Shh-N (E,F) is not that much evident than with purmorphamine (C,D) in both control and CKO explants.
9. Figure 7 7-F: The changes in the levels of Cyclin D1 and Hes1 in Ctrl/Lhx2 CKO/ dCKO are not very clear from the data in present form. It would be better to show the changes in their mRNA levels by qPCR analysis. Further, it is not clear how the changes in CyclidD1 and Hes1 levels are proving that Lhx2 acts downstream of Shh signaling.
10. Supplementary table 4: In the page 3 have typo error "centrations at 72 hr"
11. Supplementary Figure 8 is labeled as Supplementary Figure 7, so there are two figures labeled as Supplementary Figure 7.

Significance

Study is significant, and adds more depth to existing knowledge in this science field. Developmental and cell biologists would benefit from this study.

Retina regeneration, Cellular signalling, Epigenetics, One knock outs/knockdowns, transgenics, RNAseq, Microarray, ChIPseq, Cell sorting

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 Li et al., entitled: "Lhx2 is a progenitor-intrinsic modulator of Sonic Hedgehog signaling during early retinal neurogenesis," focuses on an important topic in developmental biology-the regulatory interaction between transcription factors (TFs) and signaling pathways, namely how the TF confers cells' competence to respond to extrinsic cues. The study focuses specifically on Lhx2 regulation of Sonic Hedgehog (HH)-pathway genes in retinal development. The authors approach this complex topic through global transcriptomic analyses combined with elegant functional in-vivo studies, including a systematic examination of the pathway genes through the use of inhibitors and ligand on cKO retinal explants. The results reveal complex regulation by Lhx2 of several HH-pathway genes in the developing mouse retina, including Ptch, Gli1 and the co receptors Cdon and Gas1. The finding that a single TF controls several components of the signaling pathway is interesting. Nevertheless, probably due to the complex activity of Lhx2, which functions on additional targets, it remains unclear how regulation of HH-pathway genes by Lhx2 impacts the eventual phenotype of the Lhx2 cKO retinal progenitor cells (RPCs). In the following, I list the main findings and several comments that need to be addressed:

Fig. 1 presents the experimental system using inducible Hes1-CreERT to mutate Lhx2 on E11.5 and examine the expression of Gli1 and Sonic Hedgehog in the control and mutant.

• The authors should present the distribution of the Lhx2 protein in the control vs. mutant. Considering that the deletion is of only part of the gene (as shown in Fig. 2), it is important to present the loss of the protein as well as the efficiency of Cre activity. • On the figure, add a characterization of the cellular phenotype of the cKO retinas on E15.5 by presenting the expression of markers for ganglion cells and RPCs. Figs. 2 & 3 present the bulk RNA-seq analysis of the Lhx2 cKO retinas, including experimental design, validation and results. The integration of previously published ATAC-seq and ChIP-seq data for Lhx2 point to the direct targets and bound regulatory regions.<br /> • "4 biological repeats per genotype" - Specify if four eyes were sampled from two embryos or from four different embryos. Were the embryos from different litters? • Add GSEA analysis for the HH-pathway genes. Fig. 4 presents a published approach to quantifying the response to HH using a cellular reporter assay (Li et al., 2018), whereas in Fig. 5, availability of HH ligand is evaluated by elegantly implementing the cellular reporter assay. The results suggest that Lhx2 does not regulate ligand availability. • Fig. 4 presents a published approach and thus can be included in Fig. 5.<br /> Fig. 6 presents evidence that in the Lhx2 cKO, the Shh pathway is functional downstream of Smo, because the expression of Gli1 increases in cKO cells following Smo activation (with purmorphamine). Furthermore, the response to Shh-N is shown to be partly attenuated in the Lhx2 cKO retina. <br /> Figure 7 examines whether Ptch deletion can rescue aspects of the Lhx2 phenotype. This was done by comparing the phenotypes of cKOs of Lhx2, Ptch, or both Ptch and Lhx2. The results revealed partial rescue, in the Ptch and Lhx2 cKO, of the expression of Ptch1 and Gli1, but not of the proliferation and premature differentiation phenotypes based on expression of Cyclin D1, EDU, PCNA and Hes1. • Add images of the control to Fig. 7B,C. • Explain how the deletion of Ptch1 was examined. They next investigated regulation of the Ptch co - receptors Cdon and Gas1 by Lhx2 (Figs. 8, 9). Fig. 8 presents the developmental expression pattern of Cdon and Gas1 in the control, and their downregulation in the Lhx2 cKO (although Cdon is maintained in the dorsal optic cup). The results show that Cdon is the co-receptor that is normally expressed in RPCs. GAS1 seems to play a role in the peripheral progenitors destined to ciliary body and iris.<br /> Electroporation of both receptors into the Lhx2 cKO retinas resulted in increased pathway activity (based on Gli1 reporter). • Both Cdon and Gas1 were electroporated into the Lhx2 cKO retina, although Gas1 is not expressed in control RPCs (based on the analysis in previous panels). Explain why both were co-electroporated and the outcome of electroporating only Cdon. • The outcome of electroporation of the co-receptors into control retina should be presented. • It is important to include staining for Lhx2; it is possible that the cells that respond to the co-receptors are those that were not mutated (escapers). Presenting the loss of Lhx2 (or Cre activity through the use of a reporter) and comparing it to the outcome of electroporation into the control retina are therefore required.

Finally, the authors present evidence that Lhx2 cKO, on E13.5 when Cdon is no longer expressed in the RPCs, continues to compromise the HH - pathway genes. This further supports continued regulation of several HH-pathway genes in early and late RPCs.

• The finding that a Lhx2 controls several components of HH pathway could be relevant to Lhx2 activity in patterning of the cortex - I suggest to discuss the possible relevance of the findings to other organs.

Additional comments:

• Fig. 4E: Add explanation of the quantitative analysis. • Fig. 5: Explain how results were normalized based on retinal size (which is significantly smaller in cKO retinas). How many independent experiments were run here? How many different retinas were tested? Were retinas taken from the same mouse considered 'independent'?<br /> • Fig. 8B: Indicate the genotype of the presented tissue. • Fig. 8 A,B should be presented in one panel, in the same orientation. • Fig. 8D: Present the different channels, in addition to the merge image.

Significance

The study focuses on an important topic in developmental biology-the regulatory interaction between transcription factors (TFs) and signaling pathways, namely how the TF confers cells' competence to respond to extrinsic cues. The study focuses specifically on Lhx2 regulation of Sonic Hedgehog (HH)-pathway genes in retinal development. The results reveal complex regulation by Lhx2 of several HH-pathway genes in the developing mouse retina, including Ptch, Gli1 and the co receptors Cdon and Gas1. The finding that a single TF controls several components of the signaling pathway is interesting. Nevertheless, probably due to the complex activity of Lhx2, which functions on additional targets, it remains unclear how regulation of HH-pathway genes by Lhx2 impacts the eventual phenotype of the Lhx2 cKO retinal progenitor cells (RPCs).

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 plan to address the minor comments from both reviewers as here described:

Reviewer 1 – comments____:

1. Figure 1a + 1b: The pattern of H3K9me3 at SVA elements appears to be quite different between the two cell types tested - in iPSCs it appears to be more strongly centered on the SVA with some "spreading" occurring upstream of the element, however in NCCITs it appears to mark downstream of the element and is not clearly seen to occur on the SVA. Is this due to sequencing data from different labs or due to cell - type specific epigenetic marks? The authors should consider showing H3K9me3 marks at a comparable region for example repressed genes or another family of repressed TEs as an internal control as well as showing profile plots to more clearly show the spread of epigenetic marks over the SVAs. As recommended by the Reviewer, in Figure 1 we will show several examples of H3K9me3 marked SVAs in both cell types. We will propose explanations for discrepancies if needed, but acknowledge the possibility of a sequencing artifact, as suggested by the Reviewer.

Figure 1: Given that the data in this paper is mostly from NCCIT cells it would be advisable for conclusions in hiPSCs to be tempered accordingly.

We will temper the conclusions as recommended.

Figure 1c: Which SVAs have neither H3K9me3 nor H3K27ac, what might these represent? Do they have any other notable features, or are they of a specific age?

The (very few) SVAs that have neither of those markers are almost certainly an artifact of mappability issues. We will point this out in the manuscript, including both the text and the figure legend.

Line 128-129: Is the assumption that SVAs are enriched in these region as opposed to dictating the epigenetic landscape as a consequence of their own sequence, is this correct?

This is correct, and we will clarify this in the text.

Figure 2c: Are there any enriched motifs in the repressed SVAs?

We performed this analysis, and several motifs are enriched in the repressed SVAs. We will add this data in the revised manuscript and in Fig. 2C.

Figure 2c: How many of the de-repressed SVAs contain this motif? It would be interesting to know if the YY1/OCT4 binding sites exist within any of the repressed SVAs and then, later, whether they accordingly lack actual binding of the TFs due to the presence of H3K9me3.

We will perform this straightforward analysis using the MEME-suit or HOMER, and include it in the revised version.

Figure 3d: It would be interesting to assess how close the differentially expressed genes are to an LTR5H element, or however many of the SVA-proximal and differentially expressed genes are also LTR5H-proximal?

We will perform this straightforward analysis and include it in the revised version.

Figure 3e: Are genes close only to de-repressed SVAs considered here? Worth specifying in the text. Also, what are the 20% of genes which are upregulated?

Yes, the reviewer is correct, only genes close to de-repressed SVAs are considered. We will specify that in the manuscript, and also add a results paragraph on the 20% of genes that are upregulated.

Figure 4a: Are the binding motifs for YY1 present at both binding sites? Are they the same, is the surrounding sequence the same? Are either of the binding sites present in repressed SVAs/is there any detectable binding of SVA/OCT4 in repressed SVAs? Are the YY1/OCT4 bound SVAs also those marked with H3K27ac in the Wysocka Lab dataset?

The YY1-OCT4 motif is present only where both of the factors bind together. We will specify this in the manuscript (results section). As for the second question: yes these correspond to the regions marked by H3K27ac in the Wysocka dataset. We will clarify this in the manuscript.

Figure 4c: Example used is a SVA-D element, are the YY1/OCT4-bound SVAs within the de-repressed group of a specific age?

No, they are found in all SVA groups (SVA through F). We will specify this in the manuscript (results section).

Figure 4d: are there GO enrichment terms for the genes bound by either YY1 and / or OCT4 different?

We will perform this straightforward analysis using the Ingenuity Pathway Analysis toolkit, and include it in the revised version in the results section.

Figure 5: Concluding figure should address how the SVAs subset in terms of binding, H3K9me3, gene expression changes and TFBS/TF binding - there are a lot of parameters which are assessed within the de-repressed subclass and it would be useful to show somewhere graphically where and when they co-occur or not.

We will edit the model figure to show the mechanism highlighted by the reviewer, as requested.

Finally, it would be helpful to see a discussion about what dictates the absence of H3K9me3 / presence of H3K27ac? Is this due to the TFBS sequence within the element? Further, a discussion on how the TFBS is gained in newer elements / lost in older elements is lacking. While the authors begin by stating that they are going to address what dictates whether an element is co-opted and conclude that it is due to sequence and location, I would suggest that as no conclusion is drawn on how the sequence changes to permit co-option and how the location dictates co-option, it may be worth tempering down the introduction on this point.

We will edit the introduction and discussion, as requested.

Reviewer 2 – comments____:

1) Given the CRISPR sgRNAs also target LTR5Hs, which are also bound by OCT4 in NCCIT cells (PMID: 25896322), it would be helpful to rule out more specifically that the observed effects on gene regulation associated with SVAs are actually due to nearby LTR5Hs copies being similarly repressed by CRISPRi. Depending on those results, it may also be fair to further note in the Discussion that this aspect of sgRNA selection is a potential caveat.

We will edit the discussion as recommended by the Reviewer.

2) The approaches taken here provide surprisingly good locus-specific resolution of histone modifications and TF binding to SVAs using only uniquely mapping reads. An example of this, SVA_D_r153 (a heavily 5' truncated SVA) is provided. It could be really useful to convey the central theme of the study by providing in a main figure another SVA example. Except, show a longer SVA (to demonstrate mappability and perhaps enrichment of reads on specific SVA features) near a protein-coding gene, where the SVA becomes repressed in the CRISPRi approach and the gene is differentially expressed as a result. An IGV-style figure perhaps demonstrating each key component of the work in one figure.

As recommended, we will update figure 4 by adding a full length SVA.

3) Literature. Line 58 - would suggest adding PMID: 27197217 and PMID: 33186547. Line 60 - would add PMID: 33722937. Line 67 - would add PMID: 22053090.

We will add these citations in the manuscript as recommended by the Reviewer.

4) Clarifications: Line 108 - the same 751 SVAs came up in both iPSCs and NCCITs? Line 117 - how many (if any) of the SVAs called as repressed by H3K9me3 were also called as de-repressed by H3K27ac? i.e. are the two histone marks giving completely concordant results for calling SVAs as repressed / de-repressed. Line 215 - no evidence for OCT4 or YY1 binding to any SVAs after CRISPRi at all? We will provide the exact numbers in the manuscript to answer these questions.

5) Finally, a point perhaps best left to the Discussion. Was there any cellular phenotype identified subsequent to the CRISPRi?

We did not identify any obvious cellular phenotype and we will mention this in the discussion, as recommended.

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

Barnada et al. explore the regulatory impact of SVA retrotransposons on gene regulation, focusing on NCCIT cells as a workhorse model of pluripotency. They use published and new H3K9me3 and H3K27ac ChIP-seq datasets to identify repressed and de-repressed SVAs, noting that many are in close proximity to protein coding genes. De-repressed SVAs are enriched for YY1/OCT4 binding sites. The authors then use RNA-seq and ChIP-seq to demonstrate YY1 and OCT4 binding are abrogated by SVA CRISPRi, disrupting gene regulation enacted by the SVAs. These findings highlight an important mechanism by which SVA retrotransposons can regulate genes in pluripotent cells.

This work is well executed. I appreciated the consideration paid to ChIP-seq read mappability and the implementation of CRISPRi followed by additional ChIP-seq. The following comments are intended to clarify the findings, which appear to have been obtained from robust experimental approaches.

Minor issues:

1. Given the CRISPR sgRNAs also target LTR5Hs, which are also bound by OCT4 in NCCIT cells (PMID: 25896322), it would be helpful to rule out more specifically that the observed effects on gene regulation associated with SVAs are actually due to nearby LTR5Hs copies being similarly repressed by CRISPRi. Depending on those results, it may also be fair to further note in the Discussion that this aspect of sgRNA selection is a potential caveat.
2. The approaches taken here provide surprisingly good locus-specific resolution of histone modifications and TF binding to SVAs using only uniquely mapping reads. An example of this, SVA_D_r153 (a heavily 5' truncated SVA) is provided. It could be really useful to convey the central theme of the study by providing in a main figure another SVA example. Except, show a longer SVA (to demonstrate mappability and perhaps enrichment of reads on specific SVA features) near a protein-coding gene, where the SVA becomes repressed in the CRISPRi approach and the gene is differentially expressed as a result. An IGV-style figure perhaps demonstrating each key component of the work in one figure.
3. Literature. Line 58 - would suggest adding PMID: 27197217 and PMID: 33186547. Line 60 - would add PMID: 33722937. Line 67 - would add PMID: 22053090.
4. Clarifications: Line 108 - the same 751 SVAs came up in both iPSCs and NCCITs? Line 117 - how many (if any) of the SVAs called as repressed by H3K9me3 were also called as de-repressed by H3K27ac? i.e. are the two histone marks giving completely concordant results for calling SVAs as repressed / de-repressed. Line 215 - no evidence for OCT4 or YY1 binding to any SVAs after CRISPRi at all?
5. Finally, a point perhaps best left to the Discussion. Was there any cellular phenotype identified subsequent to the CRISPRi?

Significance

This interesting study systematically demonstrates the importance of SVA-mediated gene regulation, mediated by YY1 and OCT4, in a cellular model of pluripotency. It would be of broad interest, particularly to those interested in gene regulation, pluripotency and retrotransposons. I would say most of the findings are new to the literature and that the closest publications I can think of in terms of scope either deal differently with SVA regulation (e.g. PMID: 33722937) or upon different retrotransposons (e.g. PMID: 30070637). The significant advance is therefore both technical and conceptual.

Geoff Faulkner (University of Queensland)

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

Genomic features underlie the co-option of SVA transposons as cis-regulatory elements in human pluripotent stem cells Barnada et al.

Barnada et al., set out to address the question of which factors dictate when and how certain TEs become co-opted to regulate host cellular functions while others do not, citing a lack of global understanding of the mechanisms underpinning this widely occurring phenomenon. To do so they use human SVA elements as a study, interrogating the epigenetic regulation of these elements in NCCITs as a proxy for pluripotent cells to reveal that a subset of younger, human-specific SVAs lack canonical H3K9me3 repressive marks while being enriched for H3K27ac active enhancer marks. The authors show that these SVAs, termed de-repressed SVAs, contain adjacent YY1 and OCT4 binding motifs, are closer to genes and TFBSs than the repressed SVAs and as such propose that they may function as active enhancers. To demonstrate this they generate a CRISPRi NCCIT cell line to epigenetically silence de-repressed SVAs using two gRNAs targeting SVAs genome-wide. Upon activation of dCas9 in this system differential expression of >3000 genes occurs, notably the broad repression of de-repressed SVA-proximal genes which are enriched for GO terms and TFBSs related to gametogenesis. The authors then demonstrate that silencing of naturally de-repressed SVAs disrupts YY1/OCT4 binding which is seen in wildtype NCCITs to occur adjacently at one site in the SVA element with solo-YY1 binding also occurring in a subset of SVAs at a second site. Disruption of YY1/OCT4 binding by targeting H3K9me3 to derepressed SVAs leads to dysregulation of proximal genes providing further evidence that SVAs act as enhancers via YY1/OCT4 binding to regulate nearby cellular genes.

Overall, the manuscript is clearly written and the computational and experimental approaches thorough; the findings are compelling and novel and make a helpful contribution to our current understanding of transposon co-option by host genomes. The demonstration of TFBS located within a subset of newer SVA elements is particularly interesting, with binding disrupted upon epigenetic silencing of these elements by CRISPRi. I have some minor comments and queries about further discussion points:

Minor comments:

1. Figure 1a + 1b: The pattern of H3K9me3 at SVA elements appears to be quite different between the two cell types tested - in iPSCs it appears to be more strongly centred on the SVA with some "spreading" occurring upstream of the element, however in NCCITs it appears to mark downstream of the element and is not clearly seen to occur on the SVA. Is this due to sequencing data from different labs or due to cell - type specific epigenetic marks? The authors should consider showing H3K9me3 marks at a comparable region for example repressed genes or another family of repressed TEs as an internal control as well as showing profile plots to more clearly show the spread of epigenetic marks over the SVAs.
2. Figure 1: Given that the data in this paper is mostly from NCCIT cells it would be advisable for conclusions in hiPSCs to be tempered accordingly.
3. Figure 1c: Which SVAs have neither H3K9me3 nor H3K27ac, what might these represent? Do they have any other notable features, or are they of a specific age?
4. Line 128-129: Is the assumption that SVAs are enriched in these region as opposed to dictating the epigenetic landscape as a consequence of their own sequence, is this correct?
5. Figure 2c: Are there any enriched motifs in the repressed SVAs?
6. Figure 2c: How many of the de-repressed SVAs contain this motif? It would be interesting to know if the YY1/OCT4 binding sites exist within any of the repressed SVAs and then, later, whether they accordingly lack actual binding of the TFs due to the presence of H3K9me3.
7. Figure 3d: It would be interesting to assess how close the differentially expressed genes are to an LTR5H element, or however many of the SVA-proximal and differentially expressed genes are also LTR5H-proximal?
8. Figure 3e: Are genes close only to de-repressed SVAs considered here? Worth specifying in the text. Also, what are the 20% of genes which are upregulated?
9. Figure 4a: Are the binding motifs for YY1 present at both binding sites? Are they the same, is the surrounding sequence the same? Are either of the binding sites present in repressed SVAs/is there any detectable binding of SVA/OCT4 in repressed SVAs? Are the YY1/OCT4 bound SVAs also those marked with H3K27ac in the Wysocka Lab dataset?
10. Figure 4c: Example used is a SVA-D element, are the YY1/OCT4-bound SVAs within the de-repressed group of a specific age?
11. Figure 4d: are the GO enrichment terms for the genes bound by either YY1 and / or OCT4 different?
12. Figure 5: Concluding figure should address how the SVAs subset in terms of binding, H3K9me3, gene expression changes and TFBS/TF binding - there are a lot of parameters which are assessed within the de-repressed subclass and it would be useful to show somewhere graphically where and when they co-occur or not.
13. Finally, it would be helpful to see a discussion about what dictates the absence of H3K9me3 / presence of H3K27ac? Is this due to the TFBS sequence within the element? Further, a discussion on how the TFBS is gained in newer elements / lost in older elements is lacking. While the authors begin by stating that they are going to address what dictates whether an element is co-opted and conclude that it is due to sequence and location, I would suggest that as no conclusion is drawn on how the sequence changes to permit co-option and how the location dictates co-option, it may be worth tempering down the introduction on this point.

Significance

This work represents a novel, comprehensive and significant contribution to the understanding of co-option of transposons into host gene networks and factors underlying these processes.

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 article by Skokan and coworkers studies the regulation of macropinocytosis in the Hydra. They design a clever assay to image the formation of macropinosomes in the ectodermal cells of the Hydra body, by amputating the head and the foot of the animal and then helving it onto a thin glass rod, allowing them to study the dynamics of actin rings formation, associated with uptake of external fluid phase. They also observe the cyclic formation of macropinosomes during the oscillatory contractions of spheroids formed from amputated animals during regeneration. By using agonist and antagonist drugs targeting mechano-sensitive calcium channels, they show that the formation of macropinosomes correlates with the reduction of cell tension. Overall, the article is succint, but clear and convincing. However, in my opinion, two major points should be clarified, if not solved before considering publication.

Major points:

1. the function of macropinocytosis in the Hydra is not known. The author postulates that it could be linked to a regulation of membrane area during animal contractions. However, one may wonder if the membrane cell surface really changes during contractions. I wonder if another explanation is possible: most of the organisms leaving in fresh water require an efficient mechanism to remove excess water that comes in the cells through osmosis. The hydra regular contractile movement are part of this, and I am wondering the macropinocytosis could be linked to this mechanism. Would the author be able to apply osmotic shocks, in particular hypertonic shocks, and see how it changes the formation rate and the dynamics of macropinosomes? On the reverse, in paralyzed animal, I am wondering if macropinosomes are still formed? Results from these experiments may give a clue about the function of macropinocytosis in the Hydra.
2. Because of the role of Piezo and other mechano-sensitive calcium channels, the author conclude that the factor that limits macropinocytosis is membrane tension. However, unless I am mistaken, actin cytoskeleton has also been involved in mechano-sensing channels, it could be that cortical tension, rather than membrane tension is playing a regulatory role. A direct proof of membrane tension (by measuring it) changes would be required to conclude as the authors do. The role of membrane tension versus macropinocytosis could be directly assessed using membrane tension probes such as FliptR or flipper probes. Otherwise, a less clearly defined term, that combines both cortical tension and membrane tension, such as cell surface tension or cell tension would be preferable.
3. Number of macropinocytic cups(actin rings) per cell is used as a readout for rate of macropinocytosis. Yet in addition to the number of cups parameters the diameter increases in certain conditions such as GdCi3. It would ideally be interesting to show the changes in diameter of cups and how this varies per in different conditions. For example, in videos of Jedi1 treated body columns the cups seem bigger in size. Supporting experiments of monitoring macropinosomes via dextran uptake assays needs to be performed for quantifications a rate of change in macropinocytosis is proposed. Alternatively, dextran beads of different molecular sizes with different fluorophores could also be used to assess the differences in rate and volume of uptake via macropinocytosis in various conditions of this study.
4. If membrane tension is altered upon dissecting Hydra fragments, would it make sense to study potential changes in macropinocytosis within the regenerating body column? Such as differences in actin ring formation in cells close to wound edges versus equatorial regions of regenerating body columns and spheroids?
5. Reasoning for selection of Piezo as molecular target over other stretch activated channels has not been provided. Piezo activators have been used, on the contrary depletion of Piezo via RNAi could be performed in intact animals to assess increased macropinocytosis. Furthermore, rate of macropinocytosis could be assessed in body columns generated from Piezo depleted animals. This would further support the direct role of Piezo in the process.

Minor points:

1. The authors report differences in macropinocytosis based on different parts of the animal (Fig.S1), upon treating intact animals with GdCI3 (Fig.2C) how does this vary? Do the differences still persist in spite of increased macropinocytosis?
2. Hydra are animals with an elongated body column. Dissecting body columns of different lengths could give rise to spheroids of different volume, these could then be inflated to establish a comparative volume study with different volumes and macropinocytosis.
3. For all graphical representation it would also be ideal to state the p-values for each significant comparison to better appreciate differences instead of stars.
4. For better understanding of figures, highlight in graph legends and figure panels which tissue sample has been used i.e intact animal, body column or spheroid.
5. A graphical representation could be given for the comparison of macropinocytic cups between intact hydra versus body column samples with statistical analysis. To appreciate the claims made by the authors regarding the trend of more cups being observed in body columns versus intact animals, as only the mean values are stated in the text.
6. In Fig.1F, there exist streaks of dextran distinctly outlining apical membranes of cell sets in the hydra epithelia, what are these suggestive of?
7. Fig.S1 No p-values

Significance

Overall, the work is of interest for several research communities. The significance could be increase by providing a few more experiments about the physiological role of macropinocytosis in the Hydra.

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, Skokan et al. develop a platform of cnidarian Hydra vulgaris, a powerful model for cellular self-assembly and organismal regeneration, to enable visualization of macropinocytosis in living tissue. Utilizing this system and small molecule perturbation, authors discover that macropinocytosis occurs constitutively at the ectoderm across the entire body axis of Hydra, and is constrained by membrane tension through stretch-activated channels and the downstream calcium influx.

Major Comments:

The manuscript is clearly written and logically organized, and the imaging results are properly quantified. With the logical interpretation, adequate biological repeats and statistical analysis, the method and data in this manuscript are clear and compelling. The major concern is the missing physiological significance of macropinocytosis induced by membrane relaxation in Hydra, if any.

Suggested experiments:

1. In Fig 2, the importance of SAC and Ca2+ for macropinocytosis are addressed. However, only one SAC inhibitor was used, whereas Ca2+ concentration in Ionomycin treated Hydro remained high even after 60 min when macropinocytic cup density had recovered (Fig 2E and G). As the authors mentioned in the DISCUSSION, other SAC transported cations may be involved in and thus need to be tested. Simply, the medium depleted of specific cation or water containing specific cation could be used to monitor the requirement of each cation on Jedi2 treated Hydra.
2. In Fig 3, the authors demonstrate that increased membrane tension leads to higher Ca2+ concentration and less macropinocytic cups in Hydras. The SAC inhibitors and EDTA (or calcium free buffer) used in Fig2 should be applied in the inflated regenerative spheroids to confirm that membrane tension inhibits macropinocytosis via SAC and Ca2+.
3. The authors observed an increase of macropinocytic cups in both amputated Hydra and regenerative spheroids than intact animal (0.186 and ~0.3 compared to 0.015 cups per cell, Fig S1, 2E, 3C). Would the inflation or inhibition of macropinocytosis perturb spheroid regeneration or polarization/sorting? Authors have discussed several potential biological functions of macropinocytosis in Hydra, including tension homeostasis and surface remodeling that are important during spheroid regeneration. It will be worthy to examine if mild membrane tension increase or SAC activation would delay the sorting process of regenerating Hydra tissues.

Minor points:

1. Fig 2C is the quantification results of 2B but include three sets of data (labeled as 1, 2, and 3) without explanation.
2. Would amputation of one tentacle lead to local or global Ca2+ reduction and macropinocytosis in a Hydra?

Significance

Macropinocytosis is an evolutionary conserved, from amoeba to human, and versatile endocytic route critical for mammalian immune and cancer cells for antigen surveillance and nutrient uptake. Despite ample understanding of macropinocytosis in cultured cells has been made, the function and mechanism of macropinocytosis at the organ or organismal level remains poorly studied. Therefore, this work is intriguing and timely to support the physiological occurence of macropinocytosis from the tissue and evolutionary aspects.

Macropinocytosis is critical process for membrane trafficking, cell signaling, immune surveliance and cancer cell growth, and Hydra vulgaris is a powerful model organism for regeneration biology, neuro biology and marine biology. Therefore, audiences from these fields will be interested and influenced by this report studying developing a new method for visualizing macropinocytosis in living Hydra.

I am a cell biologist studying the regulation of membrane remodeling and trafficking upon mechanical or biochemical stimuli. Due to my unfamiliar with Hydra as a model organism, the details of suggested experiments may need to be adjusted.

Referees cross-commenting

I agree with other reviewers and think their comments important and valid. This manuscript will be more clear and compelling after addressing these questions.

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 the manuscript by Skokan et al, the authors demonstrate a constitutive and robust program of macropinocytosis in the outer epithelial layer of the cnidarian Hydra vulgaris. While the model system is less tractable than others including mammalian cell types from a genetic stand-point, the authors have devised a neat approach to visualizing the planar epithelium in live organisms and provide clear evidence for macropinocytosis by a tissue monolayer in vivo. This model also supports the ancient conservation of macropinocytosis, supporting the studies in Dictyostelium, and may represent early modes of nutrient acquisition in complex fluid environments. Using probes for the cytoskeleton, fluid phase indicators, and mechanical and pharmacological interventions, the authors describe how stretch-activated calcium channels inhibit micropinocytosis. In general, while the manuscript is concisely written, and the available data are compelling, much more rigorous experimentation is required to make such a conclusion. In addition, the physiological importance for mechanical stretch in orchestrating the arrest of macropinocytosis remains unclear. Conceivably, this may be involved in the regulation of membrane tension since macropinocytosis (high membrane turnover) would demand that cells have a high rate of membrane recycling to compensate. Below, I have outlined some approaches that the authors could take to improve the study without demanding them to utilize additional model systems, which I think would be outside the scope of the work.

Major comments:

Most importantly, the role of Ca2+ entry via stretch-activated channels and how this would inhibit macropinocytosis remains unclear. In fact, the findings are somewhat counterintuitive since stretch applied to the monolayer would increase membrane tension while Ca2+ influx would support membrane delivery and exocytosis, thereby restoring tensional homeostasis.

In Fig 3, the authors demonstrate that applied stretch to the epithelium increases cytosolic Ca2+ and decreases membrane tension as expected. But whether the Ca2+ influx is required for the loss of macropinocytosis is not clear. This can be tested by either chelating Ca2+ transients in the cytosol or depleting the cells of Ca2+ by inhibiting ER-resident Ca2+ pumps and removing Ca2+ from the medium. In fact, if the authors think that extracellular Ca2+ is the only issue to arresting macropinocytosis, substituting Ca2+ for another divalent cation (or removing all divalent cations from the medium, should the epithelium be amenable to it for short periods of time) could be employed.

The connection between [Ca2+]cyto and macropinocytosis is established by Jedi and ionomycin. In the case of ionomycin, the large and sustained increase [Ca2+]cyto, well beyond what could be expected in physiological conditions, leads to the loss of plasma membrane PIP2, PIP3, and membrane associated F-actin. Jedi1/2 are certainly more targeted, but it is difficult to attribute their effects to Piezo in this system. More worryingly, the Ca2+ influx in response to Jedi2 and especially Jedi1 occurs maximally after 10 min of exposure. Yet, the authors show the complete loss of macropinocytic cups after 10 min (Fig 2E). It's difficult to reconcile that the Ca2+ is the issue.

The authors do not quantify macropinocytosis beyond Figure 1. Instead, they use "macropinocytic cups" as their surrogate for bona fide, sealed macropinosomes. Macropinocytosis can occur at different scales and different rates, so the authors should instead use the 70 kDa dextran as the gold standard in Figure 2. And as part of gold standard approaches, the authors would appease the macropinocytosis field if they tested the requirement for PI3K and Na H+ exchangers in Figure 1.

The appearance of the GCAMP6s in Figure 2F before given Jedi2 is interesting. Aside from the Ca2+ signal that appears where the Hydra has been severed, the Ca2+ through the epithelium appears very heterogeneous. Does this Ca2+ signal oscillate in the cells and/or across the epithelium? Since the authors are able to image the cytoskeleton and Ca2+ in this system, it would be interesting to determine any correlations in their kinetics.

Minor comments:

At this point, minor comments may be less useful to the authors since some of the more major suggestions are likely to impact the overall breadth of the work.

Significance

The work represents a technical advance and new system to consider macropinocytosis, albeit with limited mechanistic insights owed to some intrinsic challenges and remaining experimentation.

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 all the reviewers for their time and for their positive and constructive review of our study. We are happy that they all regard this as a highly significant piece of work. We have addressed some of their suggestions in our updated preprint and indicate below where we are planning further revisions.

Reply to Reviewer 1 Point 1

The reviewer pointed out a possibility that the Golgi polarisation leads to local/centre-most regional E-cadherin junction “maturation”, then contribute to AMIS seeding. To address this suggestion, we did fluorescence recovery after photobleaching (FRAP) using a mESC line that expresses E-cadherin-GFP in the updated manuscript. We compared the recovery speed and rate in the centre-most region and side regions to discuss whether E-cadherin junctions have different stability at these regions. What we found is that though the E-cadherin and E-cadherin-GFP protein level is at the same level at the two regions in mESC doublets (Figure S3), the mobile fraction of E-cadherin-GFP is lower in the centre-most region than the side regions (Figure 3 I, J). This implies that E-cadherin junctions in the centre-most region are more stable. We have included corresponding description of this data in Results, Methods and Discussion. We will also include equivalent data from non-mitomycin c treated control cells in the final manuscript.

Still, we do not know whether the more stable E-cadherin junctions were due to the Golgi polarization, but we have included the possibility of Golgi polarisation leads to local E-cadherin maturation in our Discussion in the transferred manuscript as follows:

“In addition, a recent study of chick neural tube polarisation (where N-Cadherin is the dominant Cadherin) has demonstrated that the interaction of β-catenin with pro-N-cadherin in the Golgi apparatus is necessary for the maturation of N-Cadherin, which is in turn important for apicobasal polarity establishment (Herrera et al, 2021). This provides the possibility that the polarised Golgi apparatus that we observe in the mESC clusters might be directionally delivering mature E-cadherin to the central-most region of cell-cell contact.”

Reply to Reviewer 1 Point 2

The reviewer suggests it would be interesting to know whether there is a role for the proteins JAM-1 or Nectin in AMIS formation and in polarising the Golgi and centrosomes towards the cell-cell contact. Like E-Cadherin, these are transmembrane junctional proteins that are present at the initiation of spot adhesions in epithelial 2D monolayers and are known to be part of a complex network of interactions between PAR-complex, junctional molecules, MAGUK scaffolding proteins and the actin cytoskeleton. Whilst we don’t propose to untangle this network here, we agree that it would be interesting to know more about the potential role of JAM-1 and Nectin in initiating polarity in mESC 3D cultures. However, it is important to note that, regardless of whether JAM-1 and Nectin also play a role in polarisation and AMIS formation, our results already demonstrate that E-cadherin-based adhesions are sufficient to initiate AMIS localisation. For example, our results from figure 4C-E demonstrate that, in a reductionist system of a single cell plated on E-Cadherin covered glass, a centrally located AMIS still forms. Precisely unravelling the mechanisms by which this happens would be better for a future study (which we have now stated in the Discussion).

Nevertheless, we now have new FRAP data (Figure 3I and J), which demonstrates that E-Cadherin is relatively more stable at the central-most point of contact between two adhering cells. This suggests that E-Cadherin is more stably bound via its downstream partners to the internal actin cytoskeleton at this point and may provide at least a partial explanation for why AMIS localisation occurs precisely at this region. We therefore suggest that the most relevant information to our study would be to determine whether either JAM or Nectin proteins are specifically localised at the AMIS, alongside PAR-3 and ZO-1, and might therefore be somehow enabling this stabilisation of E-Cadherin. We therefore plan to carry out IHC stains for JAM-A (new name for JAM-1), which has been found to be present in the mouse inner cell mass, to determine where it is localised within the mESC cell clusters with/without cell division and in WT/Cdh1 KO cells. We will update the supplementary results and discussion accordingly in the final manuscript.

Depending on these results, we might also try to knock down the function of JAM-A, using siRNA. If successful knock down were achieved, we would carry out FRAP to determine whether E-cadherin junctional stability had been altered and would also stain for AMIS markers such as PAR-3 and determine whether Golgi and centrosomes were polarised. However, it is important to note that, although we were able to achieve E-cadherin RNAi to a certain degree, it is not always possible to achieve sufficient knock down of protein by the 24-hour AMIS timepoint. Since the results of these experiments would not alter the impact of our pre-existing data, we do not propose to create new knock out cell lines in the current study. Also, possible redundancy between different paralogs may affect the interpretation of this experiment so we would only include these results if they allowed for clear interpretation.

A previous study (Gao L ,et al. Development. 2017) has already shown that knocking out Afadin (which would therefore disable Nectin junctions) in MDCK cell 3D cultures did not affect initial AMIS formation or localisation, although later cell division orientation and therefore lumen positioning was affected. Afadin was also not localised to the AMIS. Therefore, it is less likely that Nectin is involved in AMIS localisation and while we will stain for its localisation by IHC, we don’t propose to try to knock down its function.

Reply to Reviewer 2 Point 1

The reviewer pointed out using a different mitosis blocker beside Mitomycin C. a) In the updated manuscript, we included one additional drug treatment: Aphidicolin. The results showed the AMIS could form in the centre of cell-cell contacts in Aphidicolin treated, division-blocked cells. AMIS (PAR3, ZO1) and the Golgi network was also polarised towards this point (Figure S1 G-I). In the final manuscript, we will include a full data set with N=3 independent experiments. Though­ the same as Mitomycin C, Aphidicolin is a DNA replication blocker, it confirmed that the AMIS formation upon treatments is not a Mitomycin-only artefact. b) As the reviewer suggested to block mitosis at the M phase, we are testing using microtubule polymerization inhibitors, Nocodazole and Taxol and will include these results if appropriate. However, these treatments will also affect the cytoskeleton, significantly affecting the cell shape and potentially interrupting the cell-cell contact interface. Therefore, it may not be possible to include these experiments.

Reply to Reviewer 2 Point 2

The reviewer suggested to include more examples of movies showing 2 and 4 cell cluster formation in division blocked conditions. We will be happy to provide more examples of the movies included in Figure 2 and Movie 2 in the final submission. The puncta in submitted Movie 2 was not as clear as the in Figure 2D as the reviewer pointed out. This was largely due to the reduce-sized movie in the original submission. We will provide full-resolution movies in the final submission. We do often see the ‘perfect’ 4-cell shape in division-blocked cells (e.g. the last frame of movie 2, shown at timepoint 19:00 in figure 2D). The shape of the clusters appears largely dependent on how many cells fuse together.

Reply to Reviewer 2 Points 3 & 5 and Reviewer 3 Point 2

We appreciate the comments from the reviewers regarding qualifying some of the discussion of our results.

Reviewer 2 points out that E-cadherin is not providing a ‘Symmetry breaking’ step, since cells are eventually able to polarise in the absence of E-cadherin (even though they can’t make an AMIS). We have therefore modified our discussion of this point to read: “Our results therefore suggest that Cadherin-mediated cell-cell adhesion may provide the spatial cue required for AMIS localisation during de novo polarisation.”. The last paragraph of the manuscript now reads: “In summary, our work suggests that Cadherin-mediated cell-cell adhesion is necessary for localising the AMIS during de novo polarisation of epithelial tubes and cavities.”

Reviewer 3 points out that the E-Cadherin molecule by itself is not sufficient to recruit the AMIS proteins to the centre-most region of the cell-cell contacts since E-cadherin is localised all along the cell-cell contact. We have now included a FRAP analysis demonstrating that E-cadherin is more stable in the centre-most region of cell-cell contacts (Figure 3I,J), which supports the role of E-Cadherin in directing AMIS localisation to this centre-most region. Nevertheless, we accept the reviewer’s point that we still do not know the downstream mechanism by which the AMIS is precisely localised to the central region of cell-cell contacts, and we have extended our discussion of this point in the updated manuscript. To clarity the language, we have also altered our results heading and other references to this point to read: “E-Cadherin adhesions are sufficient to initiate AMIS localisation, independent of ECM signalling and cell division”. We believe our experiments with two methods support this claim that the formation of E-cadherin-based adhesions without cell divisions and ECM signals are sufficient to initiate AMIS localisation; in particular Figure 4C-E, in which a centrally located AMIS formed even in a reductionist system of only 1 cell plated on E-cadherin covered glass.

Reply to Reviewer 2 Point 4

The reviewer reasoned that the WT and Cdh1 KO mESC were from different genetic backgrounds. The WT (ES-E14) mESCs were generated from 129P2/Ola mice and the Cdh1 KO mESCs were generated from 129S6/SvEvTacArc mice. To confirm the results acquired based on the two cells lines, we are doing two approaches: 1) As the reviewer suggested, we are using siRNA knock-down of E-cadherin in the Wild-type mESCs (ES-E14) to confirm the results we had of the AMIS absence in the E-cadherin knock-out mESC cultures. As Figure S2C,D now shows, the concentrated PAR3 between two mESCs was largely reduced after E-cadherin knock-down. We will also include Mitomycin-treated conditions in this experiment for the final publication. 2) As an alternative approach, not dependent on RNAi functionality, we have acquired a 129S6/SvEvTacArc background mESC (the W4 line) as the wild-type mESC line that has the same background as the Cdh1 KO mESC line. We are using this line to perform the control experiments of Figure 3A-C to confirm the previous results, which so far are comparable in both the ES-14 and W4 mESC cell lines. Our preliminary data below show the same results as we had with the ES-E14 cells in the current Figure 3A. We will finish the full data set of N = 3 experiments and replace the current Figure 3A-C, S2A data with that from the W4 mESC cell line. In the meanwhile, we have labelled the type of wide type mESC used for each experiment in the manuscript.

Reply to Reviewer 3 Point 1

The reviewer pointed out we should include three independent experiments for our data in Figure 4E. We agree with the reviewer. We are very happy to do the suggested experiments and data analysis and will be able to provide the data of N=3 independent experiments in the final manuscript.

Reply to Review 3 Point 3

We agree with the reviewer. Our current data set of live imaging at day 3 are used to confirm the idea from the fixed images that a wrapping process does happen for lumenogenesis during the Cdh1 KO cyst formation. The current dataset could not exclude the possibility that the hollowing might co-exist. The reviewer therefore suggests including a live movie depicting early stages (before 78:00) of E-Cadherin knock-out cluster development. We did try to collect this data before we first submitted the manuscript but encountered significant technical problems due to the high sensitivity of early stage Cdh1 KO cells to phototoxicity. This meant that we could not image with less than one hour interval nor over longer than 24 hour and were therefore unable to analyse how the cell clusters behave before forming the cup-shaped cavity. We will attempt these experiments again (e.g. imaging from 12-24 hours and 24-36 hours). However, there is a high likelihood that the experiments will not be technically possible, which is why we list them in section 4 of our review plan. Instead, we include the following sentence in our discussion: “We were unable to live-image earlier stages of Cdh1 KO cluster development due to the sensitivity of these cells to phototoxicity so we can’t exclude the possibility that hollowing lumenogenesis occurs in parallel, although our IHC analysis does not indicate that this is the case.”

Reply to reviewers’ minor points

We have revised our texts, made the nomenclature of protein PAR3 consistent, and included the information of antibody suppliers, as the reviewers pointed out. Specific response to Reviewer 2; in p2 and p7, the texts were referring to zebrafish studies, where PAR3 is referred to as Pard3. We have marked it with “Pard3 (PAR-3)” now. We have increased the size of images in figure 5B and inverted the colour to make it more visible. Since this made the figure too big, we moved the ZO1 images to Figure S5A. We will provide a co-staining of mCherry (to label mCherry-PAR6B), Phalloidin and PAR-3 in a more updated manuscript to replace Figure 2A.

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 manuscript, Xuan Liang and collaborators shed light on how the precise localisation of the apical membrane initiation site (AMIS), necessary for organised lumen formation, is directed at the single-cell level. By characterising de novo polarising mouse embryonic stem cells (mESCs) cultured in 3D, the authors have uncovered a division-independent mechanism of de novo polarisation and AMIS localisation based on adhesion molecules. More precisely, they suggest that E-CADHERIN-mediated cell-cell adhesion may provide the symmetry-breaking step required for AMIS localisation during de novo polarisation since this molecule alone is sufficient and necessary to drive correct AMIS localisation. Interestingly, a high proportion of E-Cadherin knock-out (Cdh1 KO) mESC cell clusters do not hollow but instead generate lumen-like cavities via a closure mechanism. Despite not knowing the mechanism involved in the closure of these lumen-like cavities, the role of E-CADHERIN in de novo polarisation would be associated with initial steps in lumen formation (AMIS formation and localisation) but not in later steps where E-Cadherin knock-out mESC cell clusters can still make an apical membrane but do so more slowly than in WT cells and without going through a centralised AMIS stage.

Altogether, this study supports their previously published zebrafish neuroepithelial cell in vivo analysis, which demonstrated the division-independent localisation of Pard3 and ZO-1 at the neural rod primordial midline (Buckley et al., 2013). The authors have provided a novel mechanism of de novo polarisation and AMIS formation that occurs in vivo and in vitro. For this reason, this is a work with great significance that will undoubtedly be of general interest to the readers of Review commons. Nonetheless, several issues should be addressed before the publication of this manuscript.

1. In figure 4, the authors tried to demonstrate that E-CADHERIN is sufficient for AMIS localisation, independent of ECM signalling and cell division. To this end, they cultured individual division-blocked mESCs onto either E-CADHERIN-FC recombinant protein or FIBRONECTIN pre-coated glass and carried out IHC for PAR-3 after 24 hours in culture. They then performed heatmaps and analysed PAR3 intensity (Fig. 4 D, E). Although the data presented are fascinating and show the effects the authors describe, the authors should improve their sample number and repeat this experimental procedure and analysis at least two more times for their results to be consistent (only 15 cells in one experiment have been used to carry out the statistical analysis described previously).
2. The expression of E-cad is necessary for the proteins that define the apical membrane (Par3, Par6, aPKC) to be located in the AMIS. The results are clear and robust. Even so, I do not think this is sufficient, as the authors claim (headline of page 5, Figure 4). It seems clear that something more than Ecad is needed for the localisation of Par3 in the AMIS because, as the authors indicate in the discussion, Ecad is located along the entire cell-cell junction, while par3 is focused on AMIS. There must be something else that is necessary for the location of Par3. Therefore, the experiment in figure 4 does not prove that E-cad is sufficient but confirms that it is necessary for that location. Another series of experiments would have to be carried out to prove that it is sufficient. This must be clearly stated in the final version of the manuscript.
3. It is clear that E-Cadherin knock-out mESC cell clusters open cup-shaped cavities before generating a lumen-like structure. Fig. 5 presents compelling data about this in vivo lumen formation mechanism without hollowing, though they briefly describe this process. Whilst I am conscious that they do not know the mechanism by which such 'closure' occurs and that this would be suitable for another manuscript, I would strongly suggest including a live movie depicting early stages (before 78:00) of E-Cadherin knock-out cluster development. Many queries arise with this piece of data, as it seems that a small lumen could be forming prior to the cup-shaped cavity.

Minor points

1. Fig. 2A: Actin staining could be included to better visualise the spheroids.
2. Fig. 5B is very small, I would recommend them to present it bigger.
3. I would encourage the authors to revise the figures as some have displaced text. (see Fig. 5, 72 hours, Cdh1 KO).

Significance

Altogether, this study supports their previously published zebrafish neuroepithelial cell in vivo analysis, which demonstrated the division-independent localisation of Pard3 and ZO-1 at the neural rod primordial midline (Buckley et al., 2013). The authors have provided a novel mechanism of de novo polarisation and AMIS formation that occurs in vivo and in vitro. For this reason, this is a work with great significance that will undoubtedly be of general interest to the readers of Review commons. Nonetheless, several issues should be addressed before the publication of this manuscript. My lab work in lumen formation in 3D organotypic cultures and organoids

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 importance of cell division and the post-mitotic midbody in the establishment of the apical membrane initiation site (AMIS) is quite well established. However, there are observations hinting to a cell division-independent mechanism of the AMIS formation. The authors hypothesized that cell adhesion involving E-cadherin could direct the site for AMIS localisation during de novo polarisation. As model system the authors used mouse embryo stem cell (mESC) culture in Matrigel, which has been used as an in vitro model for the de novo polarisation of the mouse epiblast. The slow lumen formation in culture allows for a relatively clear separation of the stages of de novo polarisation. This enables to study the initiation of apico-basal polarity of embryonic cells alongside the first cell-cell contacts between isolated cells and small cell clusters. Here, the goal was to determine the role of cell adhesion, and in particular E-cadherin, in mESC AMIS localisation.

Major comments:

1. Mitomycin C is commonly used to block cell division, however, what it does is it blocks DNA replication, and the blockage of cell division is a consequence. It could have other effects than only blocking cell division. What about using a mitosis blocker? It would be good to have a second way in addition to mitomycin C treatment of confirming that the results support the conclusion that cell division is dispensable for AMIS localization, as the full work builds up on that first observation and this experimental setup is carried on through the manuscript.
2. Many clusters at higher cell stage that are shown (e.g. Fig 2C, 3A), look like they are in the perfect 4-cell stage after cell division. Movie 1 does not show that, only 2 cells are clustering. Movie 2 shows how two 2-cell clusters form a 4-cell cluster, however, that does not look as "perfect" as the 4-cell stages shown in the figures, which look as said more like 4-cell stages resulting from cell division. Maybe the authors could provide more movies that show the 2-cell cluster doublets (4-cell clusters)? Also, the puncta relocalization to the cell-cell contacts in the 2-cell cluster doublet is not so very clear in the movie 2. Maybe the authors have more movies for 2-cell cluster doublets?
3. On page 4 the authors observe: "Whilst homogenous control doublets localised PAR-3 to the central region of the cell-cell interface, heterogeneous chimeric doublets did not localise PAR-3 centrally (Figure 3D,E). Golgi and centrosome localisation towards the cell-cell interface suggested that the overall axis of polarity was maintained, even in the absence of both cell division and E-CADHERIN (Figure 3F-H & S2C)." This means that the axis of polarity is established without E-cadherin and without the AMIS. So, the cell-cell contact site itself is able to establish polarity, but not localize the AMIS. In line with this, the authors state: "The results also demonstrate that ECM in the absence of E-CADHERIN is insufficient for AMIS localisation." So, without E-cadherin, there is no AMIS, the ECM cannot establish it, but polarity is still established. On p. 6, it is then concluded that E-cadherin and centralised AMIS localisation are not required for apical membrane formation, but that they promote its formation earlier and more efficiently in development. In the discussion, however, the authors then state in a rather contrary manner "Our results therefore suggest that CADHERIN-mediated cell-cell adhesion may provide the symmetry-breaking step required for AMIS localisation during de novo polarisation." However, cells are polarized and have an apical membrane (Figure 3F-H & S2C) without cadherin and the AMIS, so how can this be the symmetry-breaking step for de novo polarization? Later on, in line with the earlier statement that E-cadherin and the AMIS location are not required for apical membrane formation, the authors then refine the previous statement: "the role of E-CADHERIN in de novo polarisation is specifically to localise the AMIS, which enables the integration of individual cell apical domains to a centralised region preceding lumen hollowing." This seems to be more likely to me than the CADHERIN-mediated cell-cell adhesion as the symmetry-breaking step. I therefore disagree in this point with their overall summary on p. 8, and find this a bit confusing.
4. As Wild-type mESCs (ES-E14) were purchased from Cambridge Stem Cell Institute and Cdh1 KO mESCs were gifted from a lab, it would be good to (genetically) characterize the cell lines because, apparently, they do not have the same origin and the KO cells were not derived from the parental mESCs. Alternatively, a control experiment with knockdown of Cdh1 in the purchased mESCs could be done, even if that would not lead to a complete knockdown, to make sure that the observed effects are the same as with the Cdh1 KO cell line.
5. The existing data is carefully analyzed with appropriate statistics. Replicates are sufficient. The conclusions are not yet fully justified, as discussed.

Minor comments:

• Fig S1D: It is labeled "Pard3". While also correct, it should be consistent, i.e. PAR3.
• P. 5 remove (slightly)
• P. 2, p. 7 "Pard3", replace with PAR3
• The antibody lists already contain catalog numbers in case of the primary antibodies, but no suppliers. They should be added, and also specified for the secondary antibodies for better reproducibility. In general, the text and figures are well prepared and of high quality. The citations appear appropriate.

Significance

The work describes the conceptual novelty of cell adhesion as alternative mechanism to cell division for AMIS localization, and in particular E-Cadherin as being required for AMIS positioning. It is still unclear why the AMIS is centered and the localization of cadherin is equal along cell-cell contacts (Fig 2C, S1E). How do Cadherin localization dynamics look like during the clustering of two cells? During cell division in a MDCK cyst (which is where my expertise lies), cell adhesion has to be partially removed during cytokinesis and abscission, and then be installed again, basically like a new cell-cell contact. Thus, could E-cadherin focus ("trap") the "AMIS initiation seed", rather than direct binding of PAR3 /PAR6 to cadherin as discussed by the authors, since E-cadherin is localized along the whole contact site and not centred? Could the unknown "apical seed" (which in cell division is the midbody) be trapped by cell adhesion? Could this be a common mechanism between cell division- and cell adhesion-driven AMIS localization? This finding could therefore have an even broader impact. What are the author's thoughts? While my speculation might be wrong, it might be worth hypothesizing on the connection between the role of E-cadherin in the two ways of AMIS localization.

Another novelty is the observation that polarity and cavities form later on in development independently of E-cadherin and an AMIS. This type of mechanism should be discussed further and put more into perspective with the literature.

The work describes a new mechanism which could be of broad importance in developmental biology. I therefore think that this work is highly significant.

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:

Formation of tubes in a developing organism may arise from the closure of a pre-existing polarized epithelium or from de novo polarization and cavity formation in group of dividing cells. The concept of apical membrane initiation site (AMIS) refers to the fact that polarity proteins as PAR3 accumulate at a point where the apical membrane will be created. This accumulation occurs as early as the two cell stage. Previous reports have demonstrated the importance of the division process in defining this AMIS, however, in the present work the authors in vitro 3D cultures of mESC to report a mitosis independent mechanism that creates an AMIS, induces the polarization of groups of two or more cells, and permits the formation of a central cavity. The report shows that the mechanism is fully dependent on the polarized accumulation of E-cadherin at the cell membrane in contact with the other cells. Moreover, the mechanism does not require mitosis or interaction with the extracellular matrix.

Major comments:

The main objective of the work is to demonstrate that AMIS creation and cavity formation can be mitosis independent and that it is dependent on the accumulation of E-cadherin at the midline between two cells in contact. To demonstrate these objectives, the authors perform 3D cultures of mESC. To rule out the requirement of mitosis the authors perform cultures that are treated with mitomycin C and the purify single cells that are cultured again. The authors show time-laps experiments demonstrating that individual cells that do not dived create an AMIS when they contact one to each other. With this cultures they demonstrate that the process does not require an interaction with ECM (provided by the matrigel) but requires E-cadherin, to demonstrate, that they use E-cadherin KO cells (the same line where E-cadherin has been deleted). The work is well written and the objectives very clear. The technology used and the experiments done are adequate and sufficient to accomplish the proposed objectives and the results obtained clearly support the conclusions reached. The methods are well explained and transparent to be reproduced elsewhere and the number of replicas and the statistical methods applied seem corrects to me, although I am just a biologist, not a mathematician. Although the objectives of the work, that are: to demonstrate that AMIS formation can be independent of mitosis and that AMIS requires E-cadherin, there are parts of the results that could be farther studied or at least discussed more thoroughly. Firstly, the authors show that in non-dividing cells an AMIS is formed at the first contact site between the two cells, they also show that in the absence of E-cadherin the cell maintains the polarization of centrioles and Golgi apparatus, in spite that no AMIS is formed, this indicates that the deposition of E-cadherin at the midline membrane is part of a more global polarization event that most likely is initiated by the a directional activity of the Golgi apparatus that may direct the delivery of mature E-cadherin in that particular direction, initiating or maintaining the basis for an AMIS, since recent work (already cited in the manuscript) has demonstrated the importance of cadherin maturation for polarity establishment and maintenance (Herrera et al, 2021), the actual results should be farther discussed in this context. Secondly, it was previously shown that in different epithelia, upon cell-cell contact, the aPKC complex (that includes Par3 and Par6) is recruited early to the contact site where with the participation of Cdc42, aPKC is activated generating an initial spot-like adherent junction (AJs) (Suzuki et al., 2002). In that case it is thought to be mediated by a direct interaction between the first PDZ domain of PAR-3 and the C-terminal PDZdomain-binding sequences of immunoglobulin-like cell adhesion molecules: JAM-1 and nectin-1/3 (Fig. 3) (Ebnet et al., 2001; Itoh et al., 2001; Takekuni et al., 2003). Thus it wold be interesting to know if AMIS formation in absence of cell division depends on JAM-1 or nectin and whether JAM-/Nectin signalling is sufficient to initiate the Golgi and centriole polarization and which is the mechanism governing it.

Minor comments:

As I mentioned before, the paper is well presented and very clear, yes it is simple, but simple is always better, no complicated graphics or letterings, thank you. Although in my opinion the work is very well written, I have to admit that I am not qualified to evaluate the literary style of the work since English is not my mother tongue, also I have not reviewed typographical errors since I think that is the work of the editorial, not of scientific reviewers. Please include the full reference of all the antibodies used, including the company and not just the catalog number

Quoted references:

Ebnet, K., Suzuki, A., Horikoshi, Y., Hirose, T., Meyer Zu Brickwedde, M. K., Ohno, S. and Vestweber, D. (2001). The cell polarity protein ASIP/PAR-3 directly associates with junctional adhesion molecule (JAM). EMBO J. 20, 3738-3748.

Itoh, M., Sasaki, H., Furuse, M., Ozaki, H., Kita, T. and Tsukita, S. (2001). Junctional adhesion molecule (JAM) binds to PAR-3: a possible mechanism for the recruitment of PAR-3 to tight junctions. J. Cell Biol. 154, 491-497.

Takekuni, K., Ikeda, W., Fujito, T., Morimoto, K., Takeuchi, M., Monden, M. and Takai, Y. (2003). Direct binding of cell polarity protein PAR-3 to cell-cell adhesion molecule nectin at neuroepithelial cells of developing mouse. J. Biol. Chem. 278, 5497-5500

Suzuki, A., Ishiyama, C., Hashiba, K., Shimizu, M., Ebnet, K. and Ohno, S. (2002). aPKC kinase activity is required for the asymmetric differentiation of the premature junctional complex during epithelial cell polarization. J. Cell Sci. 115, 3565-3573.

Significance

The paper describes for the first time that contrary to what was previously believed an AMIS can be generated without a cell division. This is very important because it opens the possibility that the mechanisms that originate the biologic cavities are in fact not really how we believed. The work is of interest of all cell biology scientists, specially working in developmental biology, cancer research.

My particular field of expertise is cell biology and signaling, always applied to particular events as nervous system development or cancer, in particular I am interested in Wnt/b-catenin and Sonic Hedgehog pathways.

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

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

Here, authors confirm that glycolysis is important macrophage defense against mycobacterial infection and describe a central role of pyruvate in linking glycolysis and antimycobacterial mtROS production to control the intracellular burden. Alike previous authors who have demonstrated that the non-pathogenic Bacillus Calmette-Guerin and heat-killed M. tb increase glycolysis, they show that human primary macrophages infected with M. avium increase glycolysis to facilitate mycobacterial control. Rost and coll. show evidence that the killing mechanism act through the production of mtROS by the complex I of the electron transport chain via the engagement of RET. This mechanism acts in parallel to other immunometabolic defense pathways activated in M. avium infected macrophages, such as the production/induction of itaconate via the IRF-IRG1 pathways (Alexandre Gidon 2021). * They give evidence that IL-6 and TNFa are not involved in regulating the pyruvate-mtROS and show chemical evidence that mitochondrial import of pyruvate through MPC activity is necessary to generate a high membrane potential and the subsequent production mtROS. However, the data presented here don’t explain how pyruvate is driving RET and mtROS; if pyruvate targets the electron transport chain directly or is converted (via TCA) to another metabolite that initiates RET and mtROS. Above all, this work brings attention to the possibility of using compounds that specifically engage mtROS production for therapeutic perspectives*

Reviewer #1 (Significance (Required)):

While the data presented here don t explain how pyruvate is driving RET and mtROS; if pyruvate targets the electron transport chain directly or is converted (via TCA) to another metabolite that initiates RET and mtROS, this work merits to be deeply evaluated for potential publication in a RC journal. However, the language must be improved and polished before submission.

We thank the reviewer for appreciating the importance of our findings. We are sorry for any inconveniences the language may have caused and have carefully revised the manuscript with the intention of improving it.

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

Overall evaluation

This study addresses an interesting aspect of host-pathogen relationship, namely how the metabolism of the host impacts directly or indirectly on the metabolism and/or fitness of the pathogen. For example, the generation of ROS in a way independent of NADPH-oxidases has been suggested to play a role in a number of infections. In particular, whether and how such ROS might be part of the cell-autonomous defence against an intracellular bacterial pathogen, in the present case M. avium, is of relevance. Despite these positive points, the study and manuscript suffer from a high number of serious problems both in form and content. The authors are strongly advised to revise the experimental evidence presented, including by performing additional experiments and re-interpreting some of the ones documented, as well as extensively rewrite/reformat the manuscript.

Major comments:

1- Normally, I would list the following criticisms in minor comments, but their accumulation makes them a major point:

1. • The reference style is cumbersome, because they are listed in alphabetical order of the FIRST NAMES of the first authors, which renders it difficult to identify what is cited and when *
2. • Similarly, the citations in the text include the author first names, whoch is unusual and heavy. * We thank the reviewer for bringing this to our attention. It was a mistake and we have now changed the reference style accordingly by replacing first names with initials and listing citations alphabetically from the first author’s last name.

In addition, the authors almost systematically introduce each of the articles cited with a sentence such as "Mills and colleagues did this and that ...". This is sometimes used in articles, but should not be the norm. Usually, this is used to emphasise that a given group has contributed not only substantially, but also on a regular basis to a field for years. One would write Palade and colleagues ..., or Rothman and colleagues ... . But in the present manuscript, this is mentioning first authors and their colleagues. Mills et al is a contribution from the O'Neils laboratory, which speaks to me.

We see the point and have changed some of these sentences accordingly to either write “O’Neill and colleagues …(ref)”, “First-author et al. showed that…. (ref)”, or “Others have shown …(ref)”. Editing was made page 3, lines 49, 51 and 54-56; page 4, lines 74; page 6, line 111; page 9, lines 178, 182, 193 and 197.

2- Page 4, the authors report that the positive control used to shift macrophage metabolism towards glycolysis did not work. This places doubt on the other experiments and conditions.

We thank the reviewer for bringing this point of confusion to our attention. In an Agilent Seahorse assay, which is commonly used to report glycolytic flux in scientific publications, the extracellular acidification rate (ECAR) is used as an indicator of glycolysis. Extracellular acidification occurs as protons are exported from the cell alongside lactate. In figure 1B we show quantitatively (ng/cell/24h) that lactate export is significantly increased in MDMs after LPS challenge, which translates to an increased ECAR on the traditional Seahorse assay. We also show further evidence that LPS treatment does switch the metabolism to glycolysis: the two first intermediates of glycolysis, G6P and F6P, are consumed (Fig. 1D), though between-donor variation leaves the LPS-induced increase in glucose consumption not significant. Overall, we are confident that our NMR and mass spectrometric metabolic profilings, which have been tested in several publications listed in the methods section are reliable and recapitulate previous knowledge. We have rephrased the paragraph on page 4 line 76 to clarify this point.

*3- Page 5. I am not sure to really understand the reasoning behind : "Quantification by targeted mass spectrometry did not reveal a significant accumulation in the intracellular level of pyruvate in macrophages infected with M. avium or treated with LPS when compared to untreated controls (Fig. 2A), suggesting that pyruvate is rapidly metabolized." *

The rationale for performing mass spectrometric quantification of pyruvate was to confirm experimentally that pyruvate is consumed - which we already know indirectly as its reduced product, lactate, is produced and excreted by the infected cells (Figure 1B). The hypothesis is that a proportion of the pyruvate could also enter the mitochondria and TCA cycle as shown by Mills et al.

We have tried clarifying this in the revised manuscript by replacing the original text by “Quantification by targeted mass spectrometry did not reveal a significant accumulation in the intracellular level of pyruvate in macrophages infected with M. avium or treated with LPS when compared to untreated controls (Fig. 2A), confirming that pyruvate is metabolized. Mills et al have demonstrated that during LPS activation, mouse macrophages switch to aerobic glycolysis while repurposing the TCA cycle activity to generate specific immunomodulatory metabolites (Mills EL 2016), which implies that a fraction of the pyruvate formed by glycolysis enters mitochondria.”(page 5 line 90).

4- The metabolomics experiments seem to be performed on a global population of infected and uninfected cells, without any clear mention of the fraction of infected cells, which is potentially low (Fig 1 appears to indicate less than 50%), and very likely variable between experiments. This is a serious confounding factor and likely precludes interpretation of the results?! The percentage of infected cells, at time "zero" and at each time point post-infection has to be quantified in each experiment.

The reviewer is right that the analysis was carried out on a mixed population. However, even with an infection level of 50% this should be sufficient to pick up significant changes in metabolite levels resulting from infection, which is also not seen with LPS treatment that you would assume activates all cells. We have tested another protocol of infection (MOI 10 for 120 min) that yields almost 100% infection with similar results. These data are included as supplementary figure 1 in the revised manuscript (page 6, line 124).

It would also be useful to analyse and graph the total fluorescence (coming from M. avium) per cell and the average fluorescence per cell.

Intracellular growth was quantified by measuring M. avium fluorescence intensity per cell (n>500 cells per donor and per condition) as mentioned in the figure legends. The bar charts represent the average intensity obtained from at least 500 cells per donor and conditions, each point representing an individual donor. We have successfully used this method to analyze and quantify M. avium growth in human primary macrophages (Gidon et al, PLoS Pathogens, 2017; Gidon et al, mBio 2021).

5- Page 6. How can the authors conclude "Overall, this set of data reveals that no major perturbations of the TCA cycle are induced by the infection, excluding a potential antimicrobial property of these TCA intermediates" from their data? Their experiment do not test the potential antimicrobial activity of the metabolites!

We agree with the reviewer that our data cannot preclude any anti-microbial effects of TCA intermediates. We agree that the phrasing is confusing and not as intended and have replaced it with the following sentence “Since we and others have previously found that altered intracellular levels of the TCA cycle-derived metabolite itaconate following an infection was indicative of an anti-microbial function (Gidon et al, mBio 2021; Chen et al, Science, 2020), we conclude that none of the TCA cycle intermediates warranted further investigation to explain the anti-microbial effect of glycolysis.”. (page 6, line 115).

*6- The effect of the chemical inhibitors used has to be evaluated on the growth of bacteria in broth to exclude the possibility that they directly impact them. *

We agree with the reviewer that this is important to control for. We have performed the suggested experiments and the results, showing that none of the different drugs influence M. avium in vitro growth, are included in a supplementary figure 2 in the revised manuscript (page 8, line 158).

*7- Figures. None of the graphs present error bars. In addition, for example for Fig 1A, the number of points correspond each to one donor. But there is mention neither of the number of biological replicates nor of technical replicates. This is absolutely required. *

The number of donors used for each experiment are included in the figure legend. All the experiments were done independently and are therefore biological replicates. Each point represents the value obtained for one independent donor with no technical replicates. Since we show all individual measurements (donors) an error-bar, to our opinion, is not needed. We have now changed the text in the legend to better reflect this information (page 17, lines 353, 357; page 18, line 361; page 20, lines 371, 375, 378, 383, 383; page 23, lines 410, 413, 417, 422).

8- It is unclear whether the effects documented have been measured in the whole population or only in the infected cells. And when they are measured in infected cells and uninfected cells, are these cells from a population in the same well, or from a well containing only uninfected cells?

By nature, antimicrobial effects can only be detected in infected cells therefore all the experiments measuring the effect on intracellular growth, the mitochondrial potential and the production of mitochondrial ROS were measured on infected cells. Control refers to a well containing only uninfected cells.

*9- In Figure 3A, the localisation of M. avium has to be shown. *

We have edited the Figure 3 that now includes images from the M. avium-CFP channel to help identify the infected cells.

*10- The mechanism proposed at the end of the abstract "...this work stresses out that compounds specifically inducing mitochondrial reactive oxygen species could present themself as valuable adjunct treatments." should be tested to close the loop and validate the data and hypothesis. *

We agree with the reviewer, and we are currently finalizing another manuscript on metformin, which is known to induce mitoROS, as a possible Host Directed Therapeutic agent in a mouse model of M. avium infection.

Minor comments:

1- The manuscript does not show any numbering, neither of pages nor of lines, which renders the writing of the review difficult.

We are sorry for the inconvenience. This is now included in the revised manuscript.

*2- The authors write "undirect" instead of indirect. *

We have corrected the mistake in the revised manuscript.

*3- They also use "if" instead of whether quite frequently. *

We thank the reviewer for bringing that detail to our attention. We have changed the manuscript according to the comment.

*4- Page 5, second line "... a 40% increase in cells treated ..." An increase of what? *

Treatment of infected cells with 2-DG increases the fluorescence coming from M. avium, reflecting the increase of the intracellular burden. We have changed the manuscript to make this point clearer (page 4, line 82).

*5- Page 5. The second paragraph belongs to the introduction or the discussion. *

We don’t agree on this point. We feel that it is important to inform the reader on how pyruvate can be used within the cell before showing the results, but we feel it does not fit with the broader introduction on glycolysis. However, if the editor/reviewers disagree with us, we will move this paragraph in introduction.

*6- Page 6. The authors mention that AMP, ADP etc... are nucleosides. But they are nucleotides. *

We changed the manuscript according to the suggestion (page 6, line 120; page 13, line 279; page 20 line 381).

Reviewer #2 (Significance (Required)):

The study explores an interesting question, but in its present state, the conclusions are not sustained by the evidence.

We thank the reviewer for acknowledging the importance of our work. We believe that we have addressed the concerns expressed in the comments.

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

*This manuscript reported that macrophages rely on glycolysis and RET to control M. avium infection and provide molecular evidence linking pyruvate, the end-product of glycolysis, to anti-mycobacterial mtROS production. The advantages of this paper are the clear thinking, from phenomenon to molecular mechanism, strong logic. However, there are also many shortcomings: *

Major comments:

*1.The main shortcoming of this paper is that authors only found macrophages control M. avium infection through glycolysis and RET in vitro. Although they use primary macrophages from healthy donners, not the cells lines, is it consistent in vivo? Authors should use mouse model that challenged with M. avium. Moreover, authors can isolate primary macrophages from patients that infected with M. avium, and compared it with primary macrophages from healthy donners. *

We agree with the reviewer opinion, and we are finalizing another manuscript using Metformin, a drug known to induce mitochondrial ROS, as a Host Directed Therapy in a mouse model. However, dissection of mechanisms involved such as pyruvate import to mitochondria and RET is not possible in vivo. We are not sure of the meaning of the suggested experiment: comparing primary macrophages from mav-infected patients vs healthy donors.

*2.This paper found that macrophages control M. avium infection by producing mitochondrial reactive oxygen species. This is a very interesting observation. How does mitochondrial reactive oxygen species resist mycobacterial infection? *

We thank the reviewer for appreciating our work. Yet, we are not sure what does the reviewer mean by “How does mitochondrial reactive oxygen species resist mycobacterial infection?”. It has been shown in many studies that cellular ROS causes oxidative damage and can be toxic to pathogens (and cells), including mycobacteria (Fang FC, Nature Reviews in Microbiology, 2004; Dryden M, Int. J. Antimicrob. Agents, 2018; Kim et al, J. Microbiology, 2019; Herb and Schramm, Antioxidants, 2021). However, the role of RET-induced mitochondrial ROS is a relatively new concept, that, to the best of our knowledge, has never been demonstrated to be involved in the control of mycobacterial infection nor in human primary macrophages. Conversely, bacteria have evolved defense mechanisms to protect and counteract the production of antimicrobial ROS (Kim et al, J. Microbiology, 2019).

*3.To make this data solid, whether giving pyruvate supplements to patients with mycobacterium infection can alleviate their disease? or it can be tested in mouse model. *

Initiating a clinical study is beyond the scope of this study. Furthermore, even if we could supplement infected mice with pyruvate, there is no guarantee it will get into the cells and further imported into mitochondria to induce the anti-mycobacterial effects shown in the present study. We rather believe that the key for future treatment would be to induce mitochondrial ROS through the use of other, known agonists to strengthen this cell-intrinsic defense mechanism. As stated above, we are finalizing another manuscript using a compound known to induce mitochondrial ROS as Host Directed Therapy in a mouse model.

*4.This work demonstrated that IL-6 and TNF-α could control the intracellular burden of M. avium. Many cytokines are produced by macrophage during infection. Are there other pro-inflammatory cytokines that play a role? *

We agree with the reviewer view that many cytokines influence host defenses to mycobacterial infections in addition to TNF-a and IL-6, e.g., IL-1, IL-10 and interferons. However, some of these are not induced in Mav infected macrophages (IL-1, interferons), and our previous works have shown that TNF-a and IL-6 are consistently induced by the infection (Gidon et al, PLoS Pathogens, 2017) and that they are involved in the control of the intracellular burden (Gidon et al, mBio, 2021). We therefore chose to focus on these.

*5.In Figure 1C, authors did not observe an increase of glutamine consumption in LPS-activated human macrophage which is in contrary to previous published study. How author explain this contrary result? *

We thank the reviewer for bringing this point to our attention. We have previously published the glutamine consumption of multiple myeloma cell lines quantified by the NMR based method described herein, proving it is sensitive enough to detect differences between cell lines at cell densities comparable to those of the seeded MDMs (Abdollahi et al, The FASEB journal, 2021). Hence, we are confident that the applied methodology would detect significant differences in glutamine consumption, given that the cells in question rely on glutamine. Previous observations of glutamine uptake were made using mouse macrophages and it is referenced that human and mouse macrophages do not share the exact same metabolism (Thomas et al, Frontiers in Immunology, 2014; Vijayan et al, Redox Biology, 2019). It’s worth noting that the species-specificity also extend to how macrophages respond to TLRs ligands (Sun et al, Science Signaling, 2016). As this result does not contribute significantly to the mechanism described in our paper, we do not feel the need to discuss it extensively.

Minor comment:

The authors do not provide sufficient information in the Materials and Methods, and figure legends, such as how many times the experiments were repeated? How to measure the concentration of citrate, isocitrate, succinate......

The number of donors used for each experiment are included in the figure legends. All the experiments were done independently and are therefore biological replicates. Each point represents the value obtained for one independent donor with no technical replicates. The concentrations of citrate, isocitrate, succinate and the other TCA cycle intermediates were measured by capillary ion chromatography tandem mass spectrometry, as described in the legend of Figure 2 and in detail in the Materials and Methods section on page 13-14. All metabolite measurements by targeted mass spectrometry are based on validated and published methods from our laboratory (Kvitvang et al, 2014; Stafnes et al, 2018; Røst et al, 2020). We have included more details to the methods section describing mass spectrometric metabolic profiling (page 13-14).

Reviewer #3 (Significance (Required)):

Mycobacteria avium infection is a common and serious kind of inflammation, in which macrophages has been reported to play an important role. Recently metabolic reprogramming of macrophages is proved in many diseases. By using LPS stimulation, the metabolic reprogramming of macrophages has been reported and have been confirmed to play a role during infection. Therefore, it is not so exciting to see this role of metabolic reprogramming in controlling M. avium infection.

We are sorry that our findings did not excite the reviewer, but we strongly disagree that our study does not report any novel findings. Both the significance of mitochondrial ROS in mycobacterial defense and the discovery that pyruvate can induce mitochondrial ROS via RET, are novel findings not shown before to our knowledge. And – as a note – a phenomenon described for LPS and/or in mouse macrophages does not necessarily reflect what happens during any bacterial or viral infections, nor 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 #3

Evidence, reproducibility and clarity

This manuscript reported that macrophages rely on glycolysis and RET to control M. avium infection and provide molecular evidence linking pyruvate, the end-product of glycolysis, to anti-mycobacterial mtROS production. The advantages of this paper are the clear thinking, from phenomenon to molecular mechanism, strong logic. However, there are also many shortcomings:

Major comments:

1. The main shortcoming of this paper is that authors only found macrophages control M. avium infection through glycolysis and RET in vitro. Although they use primary macrophages from healthy donners, not the cells lines, is it consistent in vivo? Authors should use mouse model that challenged with M. avium. Moreover, authors can isolate primary macrophages from patients that infected with M. avium, and compared it with primary macrophages from healthy donners.
2. This paper found that macrophages control M. avium infection by producing mitochondrial reactive oxygen species. This is a very interesting observation. How does mitochondrial reactive oxygen species resist mycobacterial infection?
3. To make this data solid, whether giving pyruvate supplements to patients with mycobacterium infection can alleviate their disease? or it can be tested in mouse model.
4. This work demonstrated that IL-6 and TNF-α could control the intracellular burden of M. avium. Many cytokines are produced by macrophage during infection. Are there other pro-inflammatory cytokines that play a role?
5. In Figure 1C, authors did not observe an increase of glutamine consumption in LPS-activated human macrophage which is in contrary to previous published study. How author explain this contrary result?

Minor comment:

The authors do not provide sufficient information in the Materials and Methods, and figure legends, such as how many times the experiments were repeated? How to measure the concentration of citrate, isocitrate, succinate......

Significance

Mycobacteria avium infection is a common and serious kind of inflammation, in which macrophages has been reported to play an important role. Recently metabolic reprogramming of macrophages is proved in many diseases. By using LPS stimulation, the metabolic reprogramming of macrophages has been reported and have been confirmed to play a role during infection. Therefore, it is not so exciting to see this role of metabolic reprogramming in controlling M. avium infection.

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

Overall evaluation

This study addresses an interesting aspect of host-pathogen relationship, namely how the metabolism of the host impacts directly or indirectly on the metabolism and/or fitness of the pathogen. For example, the generation of ROS in a way independent of NADPH-oxidases has been suggested to play a role in a number of infections. In particular, whether and how such ROS might be part of the cell-autonomous defence against an intracellular bacterial pathogen, in the present case M. avium, is of relevance. Despite these positive points, the study and manuscript suffer from a high number of serious problems both in form and content. The authors are strongly advised to revise the experimental evidence presented, including by performing additional experiments and re-interpreting some of the ones documented, as well as extensively rewrite/reformat the manuscript.

Major comments:

1. Normally, I would list the following criticisms in minor comments, but their accumulation makes them a major point:
   • a. The reference style is cumbersome, because they are listed in alphabetical order of the FIRST NAMES of the first authors, which renders it difficult to identify what is cited and when
   • b. Similarly, the citations in the text include the author first names, whoch is unusual and heavy.
   • c. In addition, the authors almost systematically introduce each of the articles cited with a sentence such as "Mills and colleagues did this and that ...". This is sometimes used in articles, but should not be the norm. Usually, this is used to emphasise that a given group has contributed not only substantially, but also on a regular basis to a field for years. One would write Palade and colleagues ..., or Rothman and colleagues ... . But in the present manuscript, this is mentioning first authors and their colleagues. Mills et al is a contribution from the O'Neils laboratory, which speaks to me.
2. Page 4, the authors report that the positive control used to shift macrophage metabolism towards glycolysis did not work. This places doubt on the other experiments and conditions.
3. Page 5. I am not sure to really understand the reasoning behind : "Quantification by targeted mass spectrometry did not reveal a significant accumulation in the intracellular level of pyruvate in macrophages infected with M. avium or treated with LPS when compared to untreated controls (Fig. 2A), suggesting that pyruvate is rapidly metabolized."
4. The metabolomics experiments seem to be performed on a global population of infected and uninfected cells, without any clear mention of the fraction of infected cells, which is potentially low (Fig 1 appears to indicate less than 50%), and very likely variable between experiments. This is a serious confounding factor and likely precludes interpretation of the results?! The percentage of infected cells, at time "zero" and at each time point post-infection has to be quantified in each experiment. It would also be useful to analyse and graph the total fluorescence (coming from M. avium) per cell and the average fluorescence per cell.
5. Page 6. How can the authors conclude "Overall, this set of data reveals that no major perturbations of the TCA cycle are induced by the infection, excluding a potential antimicrobial property of these TCA intermediates" from their data? Their experiment do not test the potential antimicrobial activity of the metabolites!
6. The effect of the chemical inhibitors used has to be evaluated on the growth of bacteria in broth to exclude the possibility that they directly impact them.
7. Figures. None of the graphs present error bars. In addition, for example for Fig 1A, the number of points correspond each to one donor. But there is mention neither of the number of biological replicates nor of technical replicates. This is absolutely required.
8. It is unclear whether the effects documented have been measured in the whole population or only in the infected cells. And when they are measured in infected cells and uninfected cells, are these cells from a population in the same well, or from a well containing only uninfected cells?
9. In Figure 3A, the localisation of M. avium has to be shown.
10. The mechanism proposed at the end of the abstract "...this work stresses out that compounds specifically inducing mitochondrial reactive oxygen species could present themself as valuable adjunct treatments." should be tested to close the loop and validate the data and hypothesis.

Minor comments:

1. The manuscript does not show any numbering, neither of pages nor of lines, which renders the writing of the review difficult.
2. The authors write "undirect" instead of indirect.
3. They also use "if" instead of whether quite frequently.
4. Page 5, second line "... a 40% increase in cells treated ..." An increase of what?
5. Page 5. The second paragraph belongs to the introduction or the discussion.
6. Page 6. The authors mention that AMP, ADP etc... are nucleosides. But they are nucleotides.

Significance

The study explores an interesting question, but in its present state, the conclusions are not sustained by the evidence.

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

Here, authors confirm that glycolysis is important macrophage defense against mycobacterial infection and describe a central role of pyruvate in linking glycolysis and antimycobacterial mtROS production to control the intracellular burden. Alike previous authors who have demonstrated that the non-pathogenic Bacillus Calmette-Guerin and heat-killed M. tb increase glycolysis, they show that human primary macrophages infected with M. avium increase glycolysis to facilitate mycobacterial control. Rost and coll. show evidence that the killing mechanism act through the production of mtROS by the complex I of the electron transport chain via the engagement of RET. This mechanism acts in parallel to other immunometabolic defense pathways activated in M. avium infected macrophages, such as the production/induction of itaconate via the IRF-IRG1 pathways (Alexandre Gidon 2021).

They give evidence that IL-6 and TNFa are not involved in regulating the pyruvate-mtROS and show chemical evidence that mitochondrial import of pyruvate through MPC activity is necessary to generate a high membrane potential and the subsequent production mtROS.

However, the data presented here don t explain how pyruvate is driving RET and mtROS; if pyruvate targets the electron transport chain directly or is converted (via TCA) to another metabolite that initiates RET and mtROS. Above all, this work brings attention to the possibility of using compounds that specifically engage mtROS production for therapeutic perspectives

Significance

While the data presented here don t explain how pyruvate is driving RET and mtROS; if pyruvate targets the electron transport chain directly or is converted (via TCA) to another metabolite that initiates RET and mtROS, this work merits to be deeply evaluated for potential publication in a RC journal. However, the language must be improved and polished before submission.

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 paper tackles an important problem regarding the effect of demographic dependent vaccination protocols on the reduction in the number of deaths with respect to the situation of no vaccination (say J). A compartmental SIRD model with reinfection Y is proposed, stratified in two (age dependent) groups, based on a binary reduction of a given contact map, and given infection fatality risk (IFR). Several countries are then analyzed.

As far as I understand we have a control variable v, parameters of the stratified model (i=1,2) tuned to match IFRi, and a control objective, i.e. minimization of J over one year.

The paper is well written. The final message and some theoretical passages are not completely clear, at least to me. I have the following observations that the authors may want to consider.

We thank the referee for the revision and are very glad that the overall evaluation is positive. Comments and suggestions have been thoroughly addressed, as we discuss in the following.

1) The study of stability of infection free and endemic equilibria should be better developed. The 5 equations can be reduced to 4 (neglecting D) and the characteristic of the reduced Jacobian used to characterize the local asymptotic stability of equilibria, instability, bifurcation points etc... Alternatively, one can use a co-positive Lyapunov function (LF). For instance, if we take the LF V=S+I+Y+R, we get $\dot V=-\mu_I I-\mu_Y Y \le 0$. If $\mu_I$ and $\mu_y$ are strictly positive all equilibria are characterized by (S*,0 0,R*) and D=1-S*-R*. So, I don't understand the phrase after (7,8), notice that Y cannot be zero in finite time. For $\mu_y=0$ then Y* can be nonzero. I guess that closed-form computation of S* and R* is possible as function of the parameters at least in the case v=0. The stability result should be cast in function of the current reproduction number (not explicitated) wrt to S and R.

The authors are invited to have a look at

1.1) Pagliara et al, "Bistability and Resurgent Epidemics in Reinfection Models", IEEE CSLetters, 2018,

for a theoretical analysis of stability on a similar (just a little bit simpler) model.

We appreciate the suggestions of the referee for improvement of this material. We have carried out an in-depth revision of the stability analysis and significantly extended it. The major addition has been, as suggested, a section relating the current reproductive number at equilibrium (we call it the asymptotic reproductive number in the text) to the fixed points of the dynamics for three different scenarios: general model, no vaccination, and zero mortality of reinfected individuals. As Pagliara et al. show in their paper, the connection between the fixed points and the reproductive number is not trivial, but it is possible to derive it through the next-generation matrix technique, as we now do. Additional references regarding this technique have been added. We have included a Table summarizing the stability analysis (page 2 in SI 3) at the end of this new section.

Other modifications include the reduction of 5 equations to 4 for the stability analysis and a clarification of possible equilibria (page 1 of SI 3), rephrasing and correcting our sentence after eqs. (7) and (8). We also attempted to obtain a closed-form computation of S* and R* but, to the best of our knowledge, concluded that it is not possible. We would be happy to pursue any insight in this respect the referee may have.

What said before should be also extended to the stratified model, where a "network" Rt could be defined, see for instance

1.2) L. Stella et al, "The Role of Asymptomatic Infections in the COVID-19 Epidemic via Complex Networks and Stability Analysis", SIAM J Cont. Opt., 2021, (arxiv.org/pdf/2009.03649.pdf)

We thank the referee for pointing out this reference. Following the analysis in Stella et al., we have carried out a stability analysis for the stratified model as well. The results are included in a new section (pages 7-10 in the SI 3).

2) It is not clear whether the free contagion parameters of the model have been fitted on real data (identification from infection and reinfection data). Notice that the interplay between vaccination strategies and NPI is important, see e.g.

*2.1) Giordano et al, Modeling vaccination rollouts, SARS-CoV-2 variants and the requirement for non-pharmaceutical interventions in Italy", Nature Medicine 2021, *

where progressive vaccination in reverse age order is considered together with different enforced NPI countermeasures.

In the first part of our study, parameters are intendedly left free because we aim at describing the generic behavior of the model. Still, we derive several inequalities and relationships between parameter ratios that seem to be sensible attending to what the different classes in the model stand for. This is as described in sections regarding model parameters when the two generic models (SIYRD and S2IYRD) are introduced. The aim is to represent both the generic dependence with some variables and a broad class of contagious diseases, so parameters are mostly free. In agreement with this approach, parameters can be also freely varied in the companion webpage.

In the second part of our study, the model is applied to COVID-19. In that case, we have used parameter values in agreement with observations, as (admittedly poorly) explained in pages 9-10 of the main text. Indeed, not enough information on parameter estimation was provided in the main text, and the SI 2 also needed some additional information. This has been amended. Let us explicitly mention that we have not fitted the dynamics of the model to any actual data set to fix specific values, as Giordano et al. do. In our case, we have first used different demographic data sets to evaluate contact rates and IFRs of the two population groups (these are parameters Mij and Ni in eqs. (7-10)). Secondly, recovery and death rates are estimated through the IFRi values for each age group i and the infectious period of COVID-19, that we fix at dI=13 days. Third, infection rate βSI=R0/dI has been estimated fixing R0=1, since the reproductive number of COVID-19 all over the world fluctuates around this value (Arroyo-Marioli et al. (2020) Tracking R of COVID-19: A new real-time estimation using the Kalman filter, PLoS ONE 16(1):e0244474). The reinfection rate is defined through its relationship with the infection rate, βRI= α1 βSI, where α1 was in the range 0-0.011 at early COVID-19 stages (Murchu et al. (2022), Quantifying the risk of SARS‐CoV‐2 reinfection over time, Rev Med Virol 32:e2260) and seems to be about 3-4 fold larger for the omicron variant (Pulliam et al., Increased risk of SARS-CoV-2 reinfection associated with emergence of the Omicron variant in South Africa, www.medrxiv.org/content/10.1101/2021.11.11.21266068v2). Given the relationships derived among parameters, our only free parameter was α2RY= α2 βRI, and we fixed it to α2=0.5 (i.e., reinfected individuals recover twice as fast as individuals infected for the first time).

Once more, it was not our goal to precisely recover specific trajectories of COVID-19 or to point at possible future scenarios, but to illustrate the dependence of major trends with model parameters. Also, the appearance of new variants requires the reevaluation of parameters. For example, omicron has different IFR (therefore different mortality and recovery rates), a different infectious period, and higher infection and reinfection rates. In this context, the interactive webpage (where we will update demographic profiles and IFR data as they become available) is a useful resource to simulate any situation different from current or past ones.

3) In the model the immunity waning is not explicitly considered (flux from R to S or better from a vaccinated compartment to S). It is clear that this complicates the model. Please discuss why the indirect way the waning is considered here is justified.

3.1) Batistela et al, "SIRSi compartmental model for COVID-19 pandemic with immunity loss", Chaos Soliton and fractals, 2021.

3.2) McMahon et al, "Reinfection with SARS-CoV-2: Discrete SIR (Susceptible, Infected,Recovered) Modeling Using Empirical Infection Data", JMIR Public health and surveillance, 2020.

Though the model does not consider an incoming flux of individuals to compartment S, the existence of a "backward" flux from R to Y yields a transient phenomenology analogous to models with increases in the S class. Indeed, it is these fluxes that cause persistent endemic states; otherwise, the S class is monotonously depleted until infection extinction.

In Batistela's et al. work, the possibility that individuals become reinfected is effectively implemented through a flux between the R and S classes, since only one class of infected individuals is considered and recovered individuals cannot be infected again. In our case, feeding back to S would mean that previous immunity is completely lost or that vaccines are not effective at all for some individuals. This is neither what McMahon et al. conclude when evaluating real data nor what more recent surveys indicate (see for instance the Science Brief published in October 2021 by the CDC, SARS-CoV-2 Infection-induced and Vaccine-induced Immunity, https://www.cdc.gov/coronavirus/2019-ncov/science/science-briefs/vaccine-induced-immunity.html).

This nonetheless, complete immunity waning (feedback to the S class) and reinfections (feedback to a partly immune class experiencing overall lower severity of the disease) are equivalent to a large extent: the trend of COVID-19 seems to indicate that our Y class will be the "new S", and that fully naive individuals would arrive mostly due to demographic dynamics (birth and death processes, as also implemented by Batistela et al.). Summarizing, complete immunity waning is rare in the time scales considered in our simulations, while partial immunity that decreases the severity of the disease (after infection or vaccination) is the rule, in agreement with our choices.

4) Reduction of deaths wrt no vaccination is of course important, but also reduction of stress in hospitals. This is particularly important now with the advent in Europe of the omicron variant. Please discuss on the real message you want to convey to policy makers in the actual scenario of the pandemic.

The model in this work is deliberately simple. Our main goal was to explore the qualitative effects of demographic structure and disease parameters in protocols for vaccine administration. This was the reason to consider a mean-field model in a population structured into two groups. The main conclusion is that optimal vaccination protocols are demography- and disease-dependent. If this is so in our streamlined model, the more it will be in more realistic models, where one should include a finer stratification and, in all likelihood, heterogeneity in contagions. Our main message, therefore, is that there is no unique protocol for vaccine roll-out, valid for all populations and diseases. The abstract has been modified to highlight this conclusion.

Some qualitative considerations also allow us to draw preliminary conclusions on the reduction of stress in hospitals. Since the number of hospital admissions is proportional to the incidence of the disease, the number H of hospitalized individuals can be represented as H=a I + b Y, with a>>b due to the partial immunity of vaccinated or recovered individuals (which belong to class Y upon (secondary) contagion). Therefore, minimizing the burden on the healthcare system amounts to minimizing the number of individuals in the I class. Beyond non-pharmaceutical measures, I is minimized when individuals are transferred as fast as possible to the Y class, that is, maximizing vaccine supply and acceptance. In terms of our model parameters, this entails maximizing v and also θ (the maximum fraction of individuals eventually vaccinated), for instace through devoted awareness campaigns. These ideas have been included in the Discussion section.

Reviewer #1 (Significance (Required)):

The final message and some theoretical passages are not completely clear, at least to me.

Please discuss on the real message you want to convey to policy makers in the actual scenario of the pandemic.

As discussed above, we have modified the manuscript following the advice given by the Reviewer. We think that both the presentation and the theory are clearer now.

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

In this paper, a compartmental model of the propagation of an infection with vaccination and reinfection is studied. The impact that changes in the rates of these two processes have on disease progression and on the number of deaths is analyzed. In order to highlight the overall effect of the demographic structure of populations and the propagation of a given disease among different groups, the population is divided into two subpopulations and the model is extended to the two-dimensional case. In addition to the study of equilibria and their relative stability, the model is then applied in the case of COVID-19. Different vaccination strategies are studied using real demographic data and with a population split between under 80 and over 80 individuals. It is observed that for low vaccination rates, the advisable strategy is to vaccinate the most vulnerable group first, in contrast to the case of sufficiently high rates, where it is appropriate to vaccinate the most connected group first. The simulations show also that with a low fatality ratio, the strategy that yields the greatest reduction in deaths is vaccination of the group with the most contacts, while the situation is reversed for higher fatality ratio.

The model and simulations presented are interesting and valuable. The comparison of the behavior of the model in the 4 different countries is very interesting, as well as the webpage created by the authors.

We thank the referee for the very positive evaluation and are very glad that the study is found interesting and valuable.

As minor comment, I think that the introduction of the model needs a more extensive literature review. For example, there is no mention of the classic SIR model of Kermack and McKendrick (1927) and other works on the introduction to epidemic models, which form the basis of the model presented by the authors.

The referee is right. There is a long history of extensions and applications since Kermack & McKendrick introduced the SIR model that we obviated. This has been amended by adding an introductory paragraph with several new references at the beginning of the Models section, page 3 in the main text.

Reviewer #2 (Significance (Required)):

The model presented by the authors is quite original and simple enough to be suitable to different contexts and scenarios.

Compared to previous work, this paper makes a twofold contribution, as explained by the authors. First, the introduction of reinfections shows the existence of long transients (or quasi-endemic states) that may precede the transition to a truly endemic state predicted for COVID-19. Second, the simplicity of model allows the characterization of systematic effects due to, at least, group size, demographic composition, and IFRs.

I am involved in the study and analysis of epidemic models accompanied by network effects. I think this paper is a good contribution, although preliminary, in the analysis of the vaccination process and in the search for the optimal strategy.

We thank the Reviewer and are glad that our goal, offering a model as simple as possible to obtain meaningful conclusions, is appreciated.

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

In this paper, a compartmental model of the propagation of an infection with vaccination and reinfection is studied. The impact that changes in the rates of these two processes have on disease progression and on the number of deaths is analyzed. In order to highlight the overall effect of the demographic structure of populations and the propagation of a given disease among different groups, the population is divided into two subpopulations and the model is extended to the two-dimensional case. In addition to the study of equilibria and their relative stability, the model is then applied in the case of COVID-19. Different vaccination strategies are studied using real demographic data and with a population split between under 80 and over 80 individuals. It is observed that for low vaccination rates, the advisable strategy is to vaccinate the most vulnerable group first, in contrast to the case of sufficiently high rates, where it is appropriate to vaccinate the most connected group first. The simulations show also that with a low fatality ratio, the strategy that yields the greatest reduction in deaths is vaccination of the group with the most contacts, while the situation is reversed for higher fatality ratio.

The model and simulations presented are interesting and valuable. The comparison of the behavior of the model in the 4 different countries is very interesting, as well as the webpage created by the authors.

As minor comment, I think that the introduction of the model needs a more extensive literature review. For example, there is no mention of the classic SIR model of Kermack and McKendrick (1927) and other works on the introduction to epidemic models, which form the basis of the model presented by the authors.

Significance

The model presented by the authors is quite original and simple enough to be suitable to different contexts and scenarios.

Compared to previous work, this paper makes a twofold contribution, as explained by the authors. First, the introduction of reinfections shows the existence of long transients (or quasi-endemic states) that may precede the transition to a truly endemic state predicted for COVID-19. Second, the simplicity of model allows the characterization of systematic effects due to, at least, group size, demographic composition, and IFRs.

I am involved in the study and analysis of epidemic models accompanied by network effects. I think this paper is a good contribution, although preliminary, in the analysis of the vaccination process and in the search for the optimal 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

The paper tackles an important problem regarding the effect of demographic dependent vaccination protocols on the reduction in the number of deaths with respect to the situation of no vaccination (say J). A compartmental SIRD model with reinfection Y is proposed, stratified in two (age dependent) groups, based on a binary reduction of a given contact map, and given infection fatality risk (IFR). Several countries are then analized.

As far as I understand we have a control variable v, parameters of the stratified model (i=1,2) tuned to match IFRi, and a control objective, i.e. minimization of J over one year.

The paper is well written. The final message and some theoretical passages are not completely clear, at least to me. I have the following observations that the authors may want to consider.

1)The study of stability of infection free and endemic equlibria should be better developed. The 5 equations can be reduced to 4 (neglecting D) and the characteristic of the reduced Jacobian used to characterize the local asymptotic stability of equlibria, instability, biforcation points etc... Alternatively, one can use a co-positive Lyapunov function (LF). For instance, if we take the LF V=S+I+Y+R, we get \dot V=-\mu_I I-\mu_Y Y \le 0. If \mu_I and \mu_y are strictly positive all equilibria are characterized by (S,0 0,R) and D=1-S-R. So, I don't understand the phrase after (7,8), notice that Y cannot be zero in finite time. For \mu_y=0 then Y can be nonzero. I guess that closed-form computation of S and R* is possible as function of the parameters at least in the case v=0. The stability result should be cast in function of the current reproduction number (not explicitated) wrt to S and R. The authors are invited to have a look at

1.1)Pagliara et al, "Bistability and Resurgent Epidemics in Reinfection Models", IEEE CSLetters, 2018,

for a theoretical analysis of stability on a similar (just a little bit simpler) model. What said before should be also extended to the stratified model, where a "network" Rt could be defined, see for instance

1.2)L. Stella et al, "The Role of Asymptomatic Infections in the COVID-19 Epidemic via Complex Networks and Stability Analysis", SIAM J Cont. Opt., 2021, (arxiv.org/pdf/2009.03649.pdf)

2)It is not clear whether the free contagion parameters of the model have been fitted on real data (identification from infection and reinfection data). Notice that the interplay between vaccination strategies and NPI is important, see e.g. 2.1) Giordano et al, Modeling vaccination rollouts, SARS-CoV-2 variants and the requirement for non-pharmaceutical interventions in Italy", Nature Medicine 2021, where progressive vaccination in reverse age order is considered together with different enforced NPI countermeasures.

3)In the model the immunity waning is not explicitly considered (flux from R to S or better from a vaccinated compartment to S). It is clear that this complicates the model. Please discuss why the indirect way the waning is considered here is justified.

3.1)Batistela et al, "SIRSi compartmental model for COVID-19 pandemic with immunity loss", Chaos Soliton and fractals, 2021.

3.2)McMahon et al, "Reinfection with SARS-CoV-2: Discrete SIR (Susceptible, Infected,Recovered) Modeling Using Empirical Infection Data", JMIR Public health and surveillance, 2020.

4)Reduction of deaths wrt no vaccination is of course important, but also reduction of stress in hospitals. This is particularly important now with the advent in Europe of the omicron variant. Please discuss on the real message you want to convey to policy makers in the actual scenario of the pandemic.

Significance

The final message and some theoretical passages are not completely clear, at least to me. Please discuss on the real message you want to convey to policy makers in the actual scenario of the pandemic.

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)):

This article focuses on one possible outcome of protein sequence evolution after duplication, in which the residue distribution at specific positions of a multiple sequence alignment becomes uncoupled from the distribution expected from the phylogeny of the protein family. The authors call these events "residue inversions" and interpret them as the result of functional pressures on family members with diverging cellular roles. Based on a theoretical model of residue evolution after duplication of the coding gene, the authors describe the criteria for categorizing a particular position in a protein as a "residue inversion" and develop an algorithm to identify such events in a multiple alignment. They then apply their approach to the family of Epidermal Growth Factor Receptors in Teleost fishes and identify 19 EGFR positions in a dataset of 88 fish genomes, which satisfy the criteria of "residues inversions". They provide support to the scoring scheme used in their approach through a simulated evolution run and conclude from a comparison of their positions to the ones predicted by SPEER to represent Specificity Determining Sites that the two are largely orthogonal and may therefore complement each other in sequence-based function prediction.

Major comments: 1. Throughout the paper, the functional involvement of positions subject to "residue inversions" is indirect, inferred from the literature, and in parts sparse and tenuous. It therefore remains unclear to what extent the interpretation that "residue inversions" represent functional adaptations is correct. The authors acknowledge this uncertainty in several places, including the Conclusions.

We agree with the reviewer that without experimental validation an uncertainty about the data interpretation remains, however testing protein function on a large scale and in non-model organisms is extremely challenging. Since we were aware of this obstacle, we validate our conclusions in different ways: 1. the theoretical model and the simulated MSA both show a lower chance of observing residue inversions than what we detected in the teleost fish EGFR example. 2. previous literature highlighted an identified inverted residue as the possible cause of sub-functionalization of teleost fish EGFR. 3 We generated the alpha fold models of teleost fish EGFR and performed molecular dynamic simulation of the two copies, in complex with the ligand. In our simulations, we see the same trend that we observe with the inter-paralog inversions at the functional level. The new results have been integrated in line 692-706.

"Residue inversion" is a very unintuitive term, which took me several readings to penetrate and made reading the article difficult. The authors may wish to reconsider this term. Naively, a residue inversion would be the swapping of residues between two positions, such that a residue expected in position A is found in position B, while the residue expected in B is found in A. That is what I suspect most readers will think.

We acknowledged that the terminology might be confusing. We therefore decided to define it as inter-paralog inversion of amino acids throughout all the text.

Is the phenomenon described here just a curiosity, or an important aspect of divergent evolution after duplication? The authors seem to be of two minds about it, calling the phenomenon "rare" in the Abstract, but an "important and understudied outcome of gene duplication" in the Introduction, then hedging again that it "might be rare" in the Conclusions. The benefits of recognizing such positions are also formulated with great caution, for example in lines 309-311: "In summary, the identification of residue inversion event has the potential to improve functional residue predictions".

We agree with the reviewer that we did not yet test the recurrence of this event on a large scale, however this does not exclude that this event is frequent. This work is focused on the observation, characterization, and implications of this event. Considering this comment and the one below we decided to perform a further analysis (see below for more details).

Additionally, the analysis of the frequency of this event at the whole-organism scale on multiple organisms, while interesting, would be out of the scope of this paper, if not just because it requires a totally different (large-scale) approach compared to the one used in here. This type of analysis is also limited by the absence of a database collecting intermediate knowledge that would speed up the initial part of ortholog classification at a broad range.

Finally, by rarity we mean the statistical chance of the event, not considering the effective chance of observing it from the real data. In fact, we rectified in the text using the reviewer’s observation.

OLD VERSION (ppXX):

Our work uncovers a rare event of protein divergence that has direct implications in protein functional annotation and sequence evolution as a whole.

NEW VERSION:

Our analysis shows a new way to investigate an important and understudied outcome of gene duplication.

It would probably strengthen the article substantially if the authors would (I) use their program to scan a large number of multiple alignments in order to establish more reliably how frequent this phenomenon actually is, and whether it is universal or a specifc aspect of eukaryotic, maybe even only vertebrate evolution; and then (II) mapped the positions identified on structural models for the proteins, obtained by homology modeling or AlfaFold prediction, in order to substantiate their potential origin as functional adaptations.

We thank the reviewer for the thoughtful suggestions. (I) we tested the inter-paralog inversion score at the proteome level using a reduced dataset (70) of reference teleost fish proteomes from Uniprot. We obtained 54 proteins that duplicated in the teleost specific whole genome duplication, then we run our pipeline on it. We found that the overall distribution of scores is more similar to the simulated evolution experiment rather than to the EGFR test case. We integrated the new results and discussion in a new paragraph and new figure in line 708-716.

(II) We considered also the analysis requested in the second point. Unfortunately, we could not extract any meaningful data from the AlphaFold models.

Reviewer #1 (Significance (Required)):

A method to improve the functional annotation of proteins in a paralogous family would be very useful, given the abundance of sequence data.

We thank the reviewer for acknowledging the importance of the question that we have addressed.

I am knowledgeable in varios aspects of molecular evolution and functional annotation. I am neither a mathematician, nor a developer of phylogenetic methods, so I cannot judge these aspects of the paper.

---

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

Review of Pascarelli and Laurino titled “Identification of residue inversions in large phylogenies of duplicated proteins”

I find the topic of the paper very exciting and long overdue. Indeed, I was under the impression that the question of parallel evolution in paralogous copies must have been addressed long ago: to my surprise, having looked in depth at the literature, that is only partially so. The manuscript, therefore, addresses a relatively novel and fundamental question of broad interest.

We thank the reviewer for his positive comment.

Having said this, I also found the manuscript to suffer from an identity problem, which in many places encroaches on the underlying quality of the science. I will structure my review into three concerns: the identity issues, the novelty issue and the emergent quality issues from the two.

Identity issues:

The manuscript is primarily dealing with an evolutionary issue – or I am biased to see it this way as an evolutionary researcher myself. Nevertheless, much of the language and terminology of the paper either misuses evolutionary terms or invents new ones in its place with a bias towards a protein chemistry perspective. Specifically, what the authors call “residue inversions” is called “parallel evolution” or “convergent evolution” in the literature. Also, "residues" are typically used for physical amino acids in a structure. If we are talking about sequence level “amon acid” would be a better term. The issue is further confounded by the meaning of “inversion” in genetics as a single mutation that inverts the position of nucleotides (i.e. an “AT” becomes “TA”).

I strongly recommend for the authors to become familiarized with the common usage of existing and widely used terms in evolutionary biology that describe the phylogenetic patterns they see: parallel evolution, convergent evolution, homoplasy, etc, and to use them consistently throughout the manuscript.

The same goes for "mutation", which the authors confuse on two levels: evolutionary and biochemical. Sometimes the authors refer to “mutation” of amino acids (which can be entertained at some level, but from a genetic perspective only nucleotides mutate – in the protein biochemistry field this term is frequently applied to amino acid residues, which is the basis of the identity issue). However, since the authors also use “mutation” to refer to a “substitution” (which is what we call a mutation that has become fixed in evolution) this creates another level of confusion. I urge the authors to change this aspect of the language of the manuscript to better reflect evolutionary concepts.

As part of the language issues I am not sure how meta-functionalization in the author’s view differs either from neofunctionalization or specialization of duplicated genes.

We thank the reviewer to point out the terminology issue, this will also help reaching a broader audience. We clarify the confusion surrounding the terms “mutation” and “residue inversion” by changing the former to “substitution”, while the latter to “inter-paralog inversions” (see also other reviewer comments).

We understand the importance of the usage of the correct term to talk about this event of protein sequences evolution. Therefore, we used convergent and parallel evolution accordingly when we discussed the nuances between Metafunctionalization and parallel evolution in the text, in lines 188 and 399.

Novelty issues:

As I mentioned, the issue of parallel evolution of gene duplications is an extremely interesting topic. I was sure that the people who studied parallel evolution, or those interested in gene duplications, must have published extensively on this. However, my search of the literature revealed only a modest pre-existing effort. Nevertheless, previous efforts are not entirely non-existent and should be cited and discussed in this paper too. The most pertinent example is

https://bmcecolevol.biomedcentral.com/articles/10.1186/s12862-020-01660-1

which has an identical setup from what I can tell (compare Figure 1 in each paper).

This paper was not hard to find using "parallel evolution", thus my focus on the language issues in the previous section.

We thank the reviewer for his suggestion, we included the relevant papers in the text in lines 520-523. Interestingly, the cited paper shows that a comprehensive analysis of the fate of duplicated genes at the sequence level was done. However, in this paper, the ‘fate’ of a paralog is determined by counting the number of sites that support one or the other fate, independently of the orthologous relationship. In our study, we start from the orthologous relationship to pre-determine the fate of the paralogous protein, then we identify the sites that break this assumption. Our type of analysis is deemed to work only where the orthologous relationship is unequivocal. That is the reason why we chose an example with relatively short branch lengths after duplication (the teleost specific duplication). Our rationale is that with a higher genome coverage across organisms, resolving the orthologous relationship will get easier in time. However, our study focuses on a distinct case (asymmetric divergence) where the diverging paralogs converge to the same phenotype. In such a case, neutral substitutions related to the ancestral relationship of a protein can be filtered out to better search for functional adaptations.

Content issues:

The lack of attention to evolutionary concepts, in my opinion, provided some missed opportunities for the authors to attack the problem in a more convincing fashion. Specifically, in the setup to distinguish between parallel evolution of paralogues versus orthologues ("inversion" versus "species-specific adaptation" in the author's text) one must be able to distinguish between the two copies and assign true evolutionary relationship. In practice, that is not always possible based on tree lengths or topologies alone because of confounding factors such as independent duplications or gene conversion events.

I would feel better about the results of this study if the following two things were integrated.

The use of synteny to better determine homologous relationships (declare copies to be true paralogues if they occupy the same syntenic region). To compare the frequency or parallel evolution of paralogues versus orthologues as a null model of the expected number of parallel events in paralogous copies.

We agree that a synteny analysis has to be included. We tested it for the EGFR proteins in fish and the results support the orthologous relationship of EGFRa and EGFRb in the two groups compared (Cypriniformes versus other teleosts). The results were included in the text and in the Supplementary figure in lines 303-305.

The second point targets the way the model derives the expectations: at the author's own admission the model makes a number of unrealistic assumptions, ") equal branch length between the two paralogs; 2) only zero to one mutation can occur in each of the six branches; 3) after a mutation, each residue is equiprobable; 4) no selective pressure; 5) the probability of a mutation on a branch solely depends on the branch length (mutation rate). The authors do not really test the resulting tree on deviation from these assumptions (I am sure that it does not conform) but essentially comparing the occurrence of parallel events in paralogues versus orthologues may solve the problem with a less restrictive set of assumptions (that one expects an equal number of parallel events in paralogues and orthologues unless there is some paralogue-specific selection pressure, which is what the authors are looking for.

We compared the occurrence of the two outcomes in both the simulation and in the real data. In all cases, the two score distributions have a very similar shape, with a 99th percentile score of respectively 0.062 and 0.113. Most sites in an alignment (>99%) are not expected to be inverted and will have scores very close to 0, making the identification of inversions a quest for outliers. Furthermore, in case of the real data, each distribution can be independently affected by different selective pressures that might bias the background distribution. While the inversion in paralogs is expectedly involving few, functional, residues, the inversion in orthologs is expected to have a broad effect. For example, a temperature adaptation might shift the number of polar residues on the protein surface (see for example: https://academic.oup.com/peds/article/13/3/179/1466666). Also, a different protein chosen for analysis might generate a different background distribution of the two events. In the larger dataset, the similarity of the two distributions is even more (99th percentile of 0.07 and 0.08). Because of the shown similarity of the two event distributions, and the possible issues with different selective pressures, we leave the analysis suggested by the reviewer as a post-processing possibly performed by the user. We report a summary of this result born from the reviewer’s observation in line 478.

In summary, I believe that the topic is very interesting, the authors potentially found a new aspect of evolution of a specific gene family. However, in my opinion a major revision is needed to unite this text with the terms in the field, the previous publication and to integrate the two additional analyses I suggested.

Minor Comments:

I started adding these specific comments before generalizing the broader deviation from the common evolutionary language. There are more further along in the manuscript, but in the interest of time I will not articulate them here hoping that the authors will first try a major revision targeting these issues.

Line 64: While neutral mutations help to determine the phylogenetic position of a protein, mutations of functional residues are a signal of functional shifts that might occur independently of the phylogeny. - this is quite misleading. All substitutions (neutral or beneficial) have a phylogenetic signal. In any case, this is discussed here in phylogenetic terms: https://pubmed.ncbi.nlm.nih.gov/10742039/

We corrected the sentence to refer to divergence time instead of phylogenetic signal.

OLD VERSION:

While neutral mutations help to determine the phylogenetic position of a protein, mutations of functional residues are a signal of functional shifts that might occur independently of the phylogeny.

NEW VERSION:

While neutral substitutions are directly proportional to the time of divergence, a change in functional residues could be a signal of a functional shift that might occur independently of the divergence time.

Line 107: "under high evolutionary pressure" - I do not know what evolutionary pressure is nor why it can be high or low.

We corrected the term to “selective pressure”.

OLD VERSION:

Lorin et al. showed that both copies of EGFR might have been retained because they are involved in the complex process of skin pigmentation (40), which is under high evolutionary pressure in most fish.

NEW VERSION:

Lorin et al. showed that both copies of EGFR might have been retained because they are involved in the complex process of skin pigmentation (40), a trait that is under selective pressure in most fish

Line 112 "linearly inherited across orthologs" - linear is a poor choice of a word here. The first thing that comes to my mind is quadratic inheritance as an alternative. Perhaps the authors are looking for "vertical" versus "horizontal" - these are established terms in phylogenetics (think "horizontal gene transfer").

We corrected the term to “vertically inherited”.

OLD VERSION

Therefore, the power to predict functional residues is limited by our ability to track protein function on the phylogenetic tree when it is not linearly inherited by orthologs.

NEW VERSION

Therefore, the power to predict functional residues is limited by our ability to track protein function on the phylogenetic tree when it is not vertically inherited by orthologs.

It is my invariant practice to reveal my identity to the authors,

Fyodor Kondrashov

Reviewer #2 (Significance (Required)):

Addressed in the 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

Review of Pascarelli and Laurino titled "Identification of residue inversions in large phylogenies of duplicated proteins"

I find the topic of the paper very exciting and long overdue. Indeed, I was under the impression that the question of parallel evolution in paralogous copies must have been addressed long ago: to my surprise, having looked in depth at the literature, that is only partially so. The manuscript, therefore, addresses a relatively novel and fundamental question of broad interest.

Having said this, I also found the manuscript to suffer from an identity problem, which in many places encroaches on the underlying quality of the science. I will structure my review into three concerns: the identity issues, the novelty issue and the emergent quality issues from the two.

Identity issues:

The manuscript is primarily dealing with an evolutionary issue - or I am biased to see it this way as an evolutionary researcher myself. Nevertheless, much of the language and terminology of the paper either misuses evolutionary terms or invents new ones in its place with a bias towards a protein chemistry perspective. Specifically, what the authors call "residue inversions" is called "parallel evolution" or "convergent evolution" in the literature. Also, "residues" are typically used for physical amino acids in a structure. If we are talking about sequence level "amon acid" would be a better term. The issue is further confounded by the meaning of "inversion" in genetics as a single mutation that inverts the position of nucleotides (i.e. an "AT" becomes "TA").

I strongly recommend for the authors to become familiarized with the common usage of existing and widely used terms in evolutionary biology that describe the phylogenetic patterns they see: parallel evolution, convergent evolution, homoplasy, etc, and to use them consistently throughout the manuscript.

The same goes for "mutation", which the authors confuse on two levels: evolutionary and biochemical. Sometimes the authors refer to "mutation" of amino acids (which can be entertained at some level, but from a genetic perspective only nucleotides mutate - in the protein biochemistry field this term is frequently applied to amino acid residues, which is the basis of the identity issue). However, since the authors also use "mutation" to refer to a "substitution" (which is what we call a mutation that has become fixed in evolution) this creates another level of confusion. I urge the authors to change this aspect of the language of the manuscript to better reflect evolutionary concepts.

As part of the language issues I am not sure how meta-functionalization in the author's view differs either from neofunctionalization or specialization of duplicated genes.

Novelty issues:

As I mentioned, the issue of parallel evolution of gene duplications is an extremely interesting topic. I was sure that the people who studied parallel evolution, or those interested in gene duplications, must have published extensively on this. However, my search of the literature revealed only a modest pre-existing effort. Nevertheless, previous efforts are not entirely non-existent and should be cited and discussed in this paper too. The most pertinent example is

https://bmcecolevol.biomedcentral.com/articles/10.1186/s12862-020-01660-1

which has an identical setup from what I can tell (compare Figure 1 in each paper).

This paper was not hard to find using "parallel evolution", thus my focus on the language issues in the previous section.

Content issues:

The lack of attention to evolutionary concepts, in my opinion, provided some missed opportunities for the authors to attack the problem in a more convincing fashion. Specifically, in the setup to distinguish between parallel evolution of paralogues versus orthologues ("inversion" versus "species-specific adaptation" in the author's text) one must be able to distinguish between the two copies and assign true evolutionary relationship. In practice, that is not always possible based on tree lengths or topologies alone because of confounding factors such as independent duplications or gene conversion events.

I would feel better about the results of this study if the following two things were integrated.

The use of synteny to better determine homologous relationships (declare copies to be true paralogues if they occupy the same syntenic region). To compare the frequency or parallel evolution of paralogues versus orthologues as a null model of the expected number of parallel events in paralogous copies.

The second point targets the way the model derives the expectations: at the author's own admission the model makes a number of unrealistic assumptions, ") equal branch length between the two paralogs; 2) only zero to one mutation can occur in each of the six branches; 3) after a mutation, each residue is equiprobable; 4) no selective pressure; 5) the probability of a mutation on a branch solely depends on the branch length (mutation rate). The authors do not really test the resulting tree on deviation from these assumptions (I am sure that it does not conform) but essentially comparing the occurrence of parallel events in paralogues versus orthologues may solve the problem with a less restrictive set of assumptions (that one expects an equal number of parallel events in paralogues and orthologues unless there is some paralogue-specific selection pressure, which is what the authors are looking for.

In summary, I believe that the topic is very interesting, the authors potentially found a new aspect of evolution of a specific gene family. However, in my opinion a major revision is needed to unite this text with the terms in the field, the previous publication and to integrate the two additional analyses I suggested.

Minor Comments:

I started adding these specific comments before generalizing the broader deviation from the common evolutionary language. There are more further along in the manuscript, but in the interest of time I will not articulate them here hoping that the authors will first try a major revision targeting these issues.

Line 64: While neutral mutations help to determine the phylogenetic position of a protein, mutations of functional residues are a signal of functional shifts that might occur independently of the phylogeny. - this is quite misleading. All substitutions (neutral or beneficial) have a phylogenetic signal. In any case, this is discussed here in phylogenetic terms: https://pubmed.ncbi.nlm.nih.gov/10742039/

Line 107: "under high evolutionary pressure" - I do not know what evolutionary pressure is nor why it can be high or low.

Line 112 "linearly inherited across orthologs" - linear is a poor choice of a word here. The first thing that comes to my mind is quadratic inheritance as an alternative. Perhaps the authors are looking for "vertical" versus "horizontal" - these are established terms in phylogenetics (think "horizontal gene transfer").

It is my invariant practice to reveal my identity to the authors,

Fyodor Kondrashov

Significance

Addressed in the 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 #1

Evidence, reproducibility and clarity

This article focuses on one possible outcome of protein sequence evolution after duplication, in which the residue distribution at specific positions of a multiple sequence alignment becomes uncoupled from the distribution expected from the phylogeny of the protein family. The authors call these events "residue inversions" and interpret them as the result of functional pressures on family members with diverging cellular roles. Based on a theoretical model of residue evolution after duplication of the coding gene, the authors describe the criteria for categorizing a particular position in a protein as a "residue inversion" and develop an algorithm to identify such events in a multiple alignment. They then apply their approach to the family of Epidermal Growth Factor Receptors in Teleost fishes and identify 19 EGFR positions in a dataset of 88 fish genomes, which satisfy the criteria of "residues inversions". They provide support to the scoring scheme used in their approach through a simulated evolution run and conclude from a comparison of their positions to the ones predicted by SPEER to represent Specificity Determining Sites that the two are largely orthogonal and may therefore complement each other in sequence-based function prediction.

Major comments:

1. Throughout the paper, the functional involvement of positions subject to "residue inversions" is indirect, inferred from the literature, and in parts sparse and tenuous. It therefore remains unclear to what extent the interpretation that "residue inversions" represent functional adaptations is correct. The authors acknowledge this uncertainty in several places, including the Conclusions.
2. "Residue inversion" is a very unintuitive term, which took me several readings to penetrate and made reading the article difficult. The authors may wish to reconsider this term. Naively, a residue inversion would be the swapping of residues between two positions, such that a residue expected in position A is found in position B, while the residue expected in B is found in A. That is what I suspect most readers will think.
3. Is the phenomenon described here just a curiosity, or an important aspect of divergent evolution after duplication? The authors seem to be of two minds about it, calling the phenomenon "rare" in the Abstract, but an "important and understudied outcome of gene duplication" in the Introduction, then hedging again that it "might be rare" in the Conclusions. The benefits of recognizing such positions are also formulated with great caution, for example in lines 309-311: "In summary, the identification of residue inversion event has the potential to improve functional residue predictions".

It would probably strengthen the article substantially if the authors would (I) use their program to scan a large number of multiple alignments in order to establish more reliably how frequent this phenomenon actually is, and whether it is universal or a specifc aspect of eukaryotic, maybe even only vertebrate evolution; and then (II) mapped the positions identified on structural models for the proteins, obtained by homology modeling or AlfaFold prediction, in order to substantiate their potential origin as functional adaptations.

Significance

A method to improve the functional annotation of proteins in a paralogous family would be very useful, given the abundance of sequence data.

I am knowledgeable in varios aspects of molecular evolution and functional annotation. I am neither a mathematician, nor a developer of phylogenetic methods, so I cannot judge these aspects of the paper.

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>

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>

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

Note: Figures we made to respond to the referee comments appear to be not supported by the ReviewCommons system.

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

**Summary:**

The authors characterized a new lncRNA locus named FLAIL that controls flowering time in Arabidopsis thaliana. The functional validation of this locus is strongly supported by the use of several different tools (CRISPR-Cas9 deletions, T-DNA insertion, amiRNA gene silencing, and transgene complementation of KO lines). It is also suggested that FLAIL lncRNA works in trans but not in cis. There are strong observations supporting that FLAIL works in trans.

Moreover, it is suggested that FLAIL regulates gene expression by interacting with distant chromatin loci. This was assessed using RNA-Seq and ChIRP-Seq. Yet, the overlap between DEGs in the flail mutant and FLAIL binding sites at the chromatin is very small, with only 12 genes. From those, only 2 flowering genes' expression was rescued by FLAIL transgene complementation. The final conclusion that FLAIL lncRNA represses flowering by direct inhibition of the 2 flowering genes expression is correlative, and lacks genetic validation.

#1.1 We plan to support the conclusions in the manuscript genetically as the reviewer suggests. We started these experiments yet they will require the timeframe of the full revision.

In addition inspection of the supplementary file shows that the ChIRP analysis was done without filtering for the FDR so that some of the positive hits have an FDR of 0,232.

#1.2 We strengthened the manuscript by implementing and FDR filter of ChIRP-seq results. The distribution of FLAIL binding sites in Fig. S7B and Table S4, and overlapping numbers between DEGs and FLAIL-ChIRP in Fig. S8A were correspondingly updated.

In addition, many of the peaks land in intergenic regions with is not mentioned in the text a graph with the position of the peaks in respect to nearby genes would help.

#1.3 Thank you for the suggestion, we strengthened the manuscript with the requested analysis. We implemented the FDR filter, then we used "tssRegion" in ChIPseeker to set distance to the nearest TSS as (-1000, 1000), then most peaks were located in promoter regions (67.24%) and in intergenic regions with 16.38%. Since many papers present the position of the peaks by ChIPseeker (PMID: 32338596, PMID: 28221134, PMID: 31081251, PMID: 32012197, PMID: 31649032, PMID: 32633672) we also applied a similar method to display a distribution of FLAIL binding loci relative to distance from the nearest TSS in Fig. S7C.

In one sentence, the authors used the right model system and methodology, including advanced techniques, to characterize a new trans-acting lncRNA important for controlling the flowering time in Arabidopsis but lack evidence supporting a mechanism of action that goes beyond the interaction with several chromatin loci.

**minor points:**

line#63-64 the authors say the COLDAIR and ASL work on FLC in cis in my view the original papers suggested/showed they work in trans.

#1.4 We increased precision by changing this sentence to ‘Vernalization-induced flowering associates with several lncRNAs such as ____COOLAIR____, COLDAIR____, ANTISENSE LONG (ASL), and COLDWRAP____ that in cis or in trans locally repress gene expression of FLOWERING LOCUS C (FLC), a key flowering repressor at different stages of vernalization’____.

Fig 1B please add some more protein-coding RNAs for the bio-info analysis for comparison

#1.5 ____done.

Order of Supplementary Fig citation is mixed with S2 coming before S1B

#1.6 Thank you, we ordered all figures by appearance in the text. __

__

It would help the reader to have a schematic of the crisper deletions, T-DNA insertion, and position of primers used for the RT-qPCR.

#1.7 We enhanced our presentation of Fig. 1A. It shows a schematic of them as well as positions of primers.

In the supplementary PDF file, some of the text is missing on page 3 beginning and end of lines.

#1.8 we ensured all text in new submission.

Reviewer #1 (Significance (Required)):

The use of several different tools to validate the biological function of FLAIL locus is a major strength of this work.

The authors propose that flowering time and its gene regulation are controlled by sense FLAIL lncRNAs. However, the sense transcription of FLAIL locus is not detected in wild-type plants by TSS-Seq, TIF-Seq, or plaNET-Seq.

#1.9.1 There appears to be some confusions. Transcription of sense FLAIL can be observed in chr-DRS, TSS-seq, TIF-seq in wild type and even in plaNET-seq in NRPB2-FLAG nrpb2-1 plant. We enhanced presentation of Fig. 1 and provided a more clear description in Line 81-99.

If the authors would have explored further the expression of FLAIL transcripts in different stages of development (vegetative and non-vegetative) and in response to different conditions, it would make their claims on the function of FLAIL lncRNAs more convincing. Additionally, flail mutants could have been obtained in the hen-2 background, since it's there where we can observe FLAIL transcription.

#1.9.2 Thank you for the suggestion. We included additional analyses in ____Fig. S2 for FLAIL transcription level in different tissues and different abiotic stress conditions base on 20,000 publicly available RNA-seq libraries (PMID: 32768600). Although many libraries are non-stranded, this analysis determined that sense FLAIL or total FLAIL (including sense and antisense) is broadly expressed over many tissues and induced in response to many abiotic stresses (Fig. S2A-B), therefore suggesting that FLAIL may be needed broadly in Arabidopsis.

FLAIL locus lays on the proximal promoter region of PORCUPINE (PCP), an important regulator of plant development. As flail mutants, pcp mutants display an early flowering phenotype. The authors show no link between FLAIL and PCP from the overlap between re-analysis of published RNA-Seq data for pcp and RNA-Seq and ChIRP-Seq from the authors. This analysis is not enough to exclude the involvement of PCP from the FLAIL function. PCP expression using RT-qPCR should be performed in flail mutants to further support that FLAIL works independently from PCP.

#1.10 We strengthened this conclusion by adding the requested experiment. PCP transcription level in flail3 mutant was provided by RT-qPCR and RNA-seq in Fig. S11A-B.

This work does not hypothesize any molecular mechanism besides the interaction of FLAIL lncRNAs with several chromatin loci. It was recently proposed in Arabidopsis that a trans-acting lncRNA interacts with distant loci via the formation of R-loops. The authors do not comment on that. This work would benefit in correlating FLAIL binding sites with R-loop-forming regions mapped in Arabidopsis, regardless of the results from this analysis. Additionally, the authors could attempt to look for a motif responsible for FLAIL binding.

Check R-loop forming data R-loops (Santos-Pereira and Aguilera, 2015) in Arabidopsis, determined by DRIP-seq (Xu et al., 2017).

#1.11 Thanks very much for this excellent suggestions.

First, we searched for a consensus DNA motif on FLAIL binding regions by Homer. We determined four commonly enriched DNA sequence motifs among FLAIL target genes (Fig. 4G). Notably, the target genes CIR1 and LAC8 contained consensus sequences that matched to all FLAIL binding motifs (Fig. 4G). These data are consistent with a model where FLAIL binds DNA targets through a sequence complementary mechanism. Functionally important sequences are frequently conserved among evolutionarily distant species, we observed three motifs that appeared to cross-species conserved (Fig. S9), suggesting a potential evolutionarily constrained role.

Second, we indeed identified R-loops peaks on several of FLAIL binding sites by DRIP-seq (Xu et al., 2017). For example, we observed R-loop formation over three FLAIL binding motifs at CIR1 locus and one at LAC8 (Fig. R1), indicating that R-loop formation may also be a factor determining FLAIL binding. Even though R-loop peaks are present at several FLAIL targets, full elucidation if R-loop formation determines FLAIL targeting requires further experimental evidence is beyond the scope of the current manuscript.

Fig. R1 Representative tracks at LAC8 and CIR1 showing R-loop formation by DRIP-seq on Watson strand (w-R loops), Crick strand (c-R loops). Undetectable R-loops after RNAse-H treatment was shown as negative control. Four conserved sequence regions of FLAIL binding motifs were indicated by red arrows at LAC8 and CIR1 loci. Gene annotation was shown at the bottom.

Most of the key conclusions are convincing, except for the flowering time control directly through CIR1 and LAC8, which should be mentioned as speculative

____#1.12____ Thank you for finding most key conclusions convincing. We plan strengthen the manuscript with additional genetic evidence to as part of the full revision.

The words locus and loci are latin and they should be written in italic. The word Brassicaceae, referring to the family should be in italic, and should not be "Brassicaceaes". The word analysis has the wrong spelling.

#1.13 We follow conventions given in Scientific Style and Format: The CBE Manual for Authors, Editors and Publishers (1994) Cambridge University Press, Cambridge, UK, 6th edn. The words locus and loci are common Latin terms and should not be italicized. However, should the format of the final prefer these words in italics we will change it later. We improved consistency of using italics. “Brassicaceaes” was changed to “Brassicaceae”.

"How much time do you estimate the authors will need to complete the suggested revisions: this is difficult to answer as it depends to which level the author would like to take their work. In my view, if all new experiments would have to be started from scratch it is too far away to be estimated.

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

In this ms, the authors identified the FLAIL lncRNA that represses flowering in Arabidopsis from a locus producing sense and antisense transcripts. They use an allelic series involving T-DNA insertions, CRISPR/Cas9 and artificial miRNAs to study the role of FLAIL in flowering. A complementation series of constructs of the flail3 allele allowed them to show that the sense FLAIL lncRNA can act in trans. RNAseq revealed a small group of genes linked to the regulation of flowering whose expression is affected in the mutant and restored in the complementation line. To gain further insight into FLAIL function, the authors used a ChIRPeq approach to test whether the lncRNA can recognize potential target genes along the genome and they could show that FLAIL binds specific genomic regions. Clearly, this paper shows very nice evidence that the FLAIL lncRNA can act in trans to regulate gene expression. Nevertheless, there are certain points that need to be clarified to further support the action of the sense FLAIL transcript.

1.According to Fig. 1 A, the antisense FLAIL is "internal" to the DNA genomic area spanning the sense FLAIL. Hence, with direct RT-qPCR is very difficult to distinguish between these molecules as a minor "RT" activity of the Taq polymerase may lead to detection of low levels of antisense, idem if RDRs may generate low antisense levels. Although I think that the plaNET seq brings strong evidence about the start and ends of these molecules, to measure them by RT-qPCR is not trivial and requires the use of strand-specific RT-PCR using a 5' extension of the oligo and amplification with one oligo of the FLAIL sequence (sense or antisense) and the added oligo.

#2.1.1 Thanks for this good suggestion. We tested both sense and antisense FLAIL transcription using oligo linked gene specific reverse primers for RT and a pair of the linked oligo and gene specific forward primer for qPCR. Primer locations were shown in Fig. 1A and new data were in Fig. 1C-D, Fig. 2B-C, and Fig. S4B-C.

It is not clear how they could distinguish precisely sense and antisense particularly when both RNAs correlate as it is the case here in all alleles (Fig. 1 C and 1D). This should be more explicitly mentioned in the materials and methods section.

#2.1.2 We gave a description of strand specific RT-qPCR method in detail in Line 397-402.

2.In Fig. 2, what are the levels of antisense in the complementing lines with the sense transcript? And reciprocally sense levels in antisense constructs?

#2.2 We added this data in Fig. 2B-C and described in Line 136-143. We indeed observed that sense FLAIL transcripts in the transformed asFLAIL construct or asFLAIL transcripts in the transformed sense FLAIL construct was similar to the control 35S:GUS (Fig. 2B-C), validating that NOS terminator inhibits antisense transcripts. We also noted that the transformed 35S:GUS and sense FLAIL construct expressed higher asFLAIL compared to the flail3 mutant (Fig. 2C). This may be caused by a T-DNA insertion of the resulting transgenic plants.

This will definitively demonstrate the assumption that the T-NOS termination will not allow any expression on the other strand. At present, only one of the lncRNAs is measured in each experiment?

#2.3 We appreciate the next-level reflection of this reviewer, with so many regions initiating cryptic antisense transcription it is an interesting challenge to identify a 3´- terminator that initiates no or poor antisense transcription.

First, previous published data argue that the NOS terminator is largely abolishing initiation of antisense transcription (PMID: 33985972, PMID: 30385760, PMID: 27856735). All these studies address roles of antisense transcription by generating mutations abolishing antisense lncRNA transcription using the NOS terminator sequences.

Second, to satisfy the curiosity of this reviewer, we provide data below that from another manuscript of the lab in preparation. It’s a screenshot of plaNET-seq in fas2-4 NRPB2-FLAG nrpb2-1 mutant carrying a pROK2 construct. The pROK2 T-DNA coincidentally carries a NOS terminator. We mapped plaNET-seq reads to the pROK2 scaffold to display the reads. In pROK2, a NOS promoter activates NPTII expression (red) with NOS terminator as a terminator sequence. No antisense transcription (blue) is detectable by this sensitive method to detect nascent transcripts. Taken together, the selection of the NOS terminator as a region suppressing initiation of antisense transcription represents a valid choice.

Fig. R2 Genome browser screenshot of plaNET-seq at NPTII locus of pROK2 T-DNA vector in fas2-4 NRBP2-FLAG nrpb2-1 mutant. This mutant carries a pROK2 construct, in which a NOS promoter activates NPTII expression with NOS terminator a terminator sequence. Sense strand was shown in red and antisense strand in blue. pROK2 annotation was shown at the bottom.

3.In Fig. 3, it will be important to also show the FLAIL locus in the flail3 mutants (in comparison to the wt) as well as the transgene locus. Here the reads will be strand specific and furthermore this will allow to show that the transgene is not generating antisense transcripts (through RDRs for gene silencing?) and confirm that the sense FLAIL is required for the complementation.

#2.4 Thank you very much for this suggestion. NGS reads for endogenous FLAIL and transgenic FLAIL both map to the FLAIL locus, so we show the FLAIL locus in Fig 3B. This representation shows that sense FLAIL transcripts were significantly reduced in flail3 and rescued in complementation line comparing to wild type. These data argue against the idea of gene silencing and linked antisense production from the transgene. However, RNA-seq suggests that an isoform of asFLAIL appears to accumulate in flail3. Since we fail to identify this accumulation by strand specific RT-qPCR result in flail3 and in CRISPR-deletion lines, this may be an asFLAIL isoform resulting from the T-DNA insertion.

4.In Fig. S5, the expression of FLAIL is shown in the artificial miRNA lines. Is the antisense FLAIL affected "indirectly" by the cleavage of the amiRNA or remains constant? This is likely the case but should be shown.

#2.5 We added this result in Fig. S4C and expression level of asFLAIL remains constant compared to the transformed empty vector control.

5.The ChIRPseq data adds major novelty to the ms and brings new ideas about the way of action of FLAIL. However, are there any common epigenetic states between ChIRP targets (e.g. histone modifications, antisense RNA production, homologies "detected" in the conserved regions between Camelina and Arabidopsis and the target loci? Or others) that may highlight potential mechanisms leading to repression mediated by FLAIL of these loci? There are many databases that could be explored (even during flowering) to search for potential relationships. Although precise description of the mechanism is out of the scope of this ms, this can be discussed in more detail to further expand on the nice data obtained.

#2.6 We searched for a consensus DNA motif on FLAIL binding regions by Homer. We determined four commonly enriched DNA sequence motifs in target genes. Notably, the target genes CIR1 and LAC8 contained consensus sequences that matched to all FLAIL binding motifs (Fig. 4G). These data are consistent with a model where FLAIL binds DNA targets through a sequence complementarity mechanism. Functionally important sequences are frequently conserved among evolutionarily distant species, we observed three motifs that appeared to cross-species conserved (Fig. S9), suggesting a potential evolutionarily constrained role.

**Minor comments:**

6.In Fig. S3, a global alignment between FLAIL and two loci in Arabidopsis and Camelina is sown. What is the extent of homology? How conserved is this sequence at nucleotide level (small or very long?) to support the conservation of this lncRNA. Are there potential structures conserved among these lncRNAs?

#2.7 T____wo consensus regions of ____FLAIL____ sequences among eleven disparate Brassicaceae genomes were shown in Fig. S9. ____Camelina sativa_ shared 98-nucleotide_ conserved sequences with Arabidopsis thaliana. In the future, it will be interesting to explore evolutional conserved structures among Brassicaceae genomes. However, these analyses are beyond the scope of the current manuscript.

7.In Fig. S4B, arrows may help to understand which seeds were selected.

__#2.8 Thanks. Arrows were included.____

__

Reviewer #2 (Significance (Required)):

This paper is a very nice piece of work and demonstrate the action of a long non-coding RNA (lncRNA) in trans on specific targets involved in the regulation of a developmental process, flowering. There is growing evidences that the non-coding genome hides large number of lncRNAs and there is little detailed genetic support for the action of lncRNAs globally. In contrast to many descriptive papers in the field, this ms demonstrates genetically, through an allelic series and complementation experiments, that this lncRNA locus is involved in flowering regulation and that its sense lncRNA recognizes target loci genome-wide, bringing interesting perspectives on potential new mechanisms of transcriptional regulation mediated by non-coding RNAs.

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

In the manuscript by Jin et al authors characterize the FLAIL DNA locus in Arabidopsis (using a wide array of publicly available datasets), which produces a set of sense and anti-sense lncRNAs.

While our work on the FLAIL manuscript was ongoing we published the manuscripts where we presented these novel genomics methods and related data to capture nascent transcription and cryptic isoforms. We shared most data with TAIR, so we are happy to hear that these data are considered publically available.

Authors determined that the sense FLAIL lncRNA (or a set of sense lncRNAs, which isn't fully clear from the way the data are presented) is involved in flowering time in Arabidopsis based on the fact that the several flail mutants lead to the early flowering phenotype and this flowering defect is complemented by transgenic FLAIL DNA, meaning that FLAIL lncRNA acts in trans.

A series of experiments lead us to conclude that the sense isoform of FLAIL is responsible for the effect. We improved the data representation and writing of the manuscript to enhance accessibility.

The T DNA flail3 - mutant results in expression changes (up or down) of 1221 genes, including twenty genes linked to flowering in various ways. Expression of a group of these flowering-related genes could be either fully (for eight genes) or partially (for five genes) rescued by transgenic FLAIL. Authors also conducted the ChIRP-seq to determine which genes are physically bound by FLAIL lncRNA genome-wide. It was found that 210 genes in the genome are bound by FLAIL lncRNA. Comparison of the dataset of differentially expressed genes in the T-DNA flail3 mutant with the ChIRP-seq dataset of genes that are bound by FLAIL lncRNA revealed the 12 overlapping genes.

Among these twelve overlapping genes, four were found to be functionally connected to flowering with expression of these four genes being down in flail3 T-DNA mutant. Two out of these four genes were ruled out from being involved in the regulation of flowering by FLAIL. Authors conclude that the two other genes (Cir1 and Lac8) are responsible for the late flowering phenotype of flail mutants based on the three lines of evidence: (i) these genes expression is reduced in the flail mutant, (ii) FLAIL lncRNA directly interacts with these genes chromatin, (iii) the mutants of these genes were previously reported by others to display early flowering phenotypes too. While I find many of the findings reported in the manuscript very interesting, building a good foundation on which to expand the study and providing a very good leads for follow up experiments, I also have serious concerns about the manuscript in its current form.

Most importantly, this reviewer doesn't think that the mechanism of FLAIL lncRNA action was convincingly demonstrated. The main question would be how FLAIL lncRNA works and this question wasn't fully answered. It is great that FLAIL lncRNA binds directly to the two flowering-related genes, but what does it mean? Does it change any chromatin context of these genes quantitatively or qualitatively to affect the transcription? Or does it bind any components of transcriptional machinery and thus controls the transcriptional output?

#3.1 This manuscript addresses an important question in the field question: what is the evidence for functional elements in non-coding regions of genomes? Despite many efforts, convincing genetic support for these functions often remained limited. In addition to our strong genetic data, we provided new evidence that FLAIL recognizes targets with evolutionally conserved sequence motifs as part of the revision in Fig 4F and Fig. S9. Additionally, we plan to do ChIP-qPCR to identify histone modifications on FLAIL targets.

Additionally, flail3 T-DNA mutant affects the expression of 1221 genes and FLAIL lncRNA physically interact with 210 genes, so how can authors be fully sure that FLAIL lncRNA has only direct effect on these two genes and doesn't also contribute to the regulation of the upstream to Cir1 and Lac8 genes or even components of the transcriptional machinery that regulate these genes?

#3.2 We agree with this opinion. It is the reason why we felt stating this exact conclusion in our previous manuscript was justified. We improved accessibility of our manuscript in the revision, these clarify our model, that the trans-acting lncRNA sense FLAIL can interact with the chromatin regions of its target genes to directly or indirectly regulate gene expression changes involving flowering (Line 274).

Theoretically, doing RNA-seq in the amiR-FLAIL sense lncRNA mutant might have a chance of reducing the number of affected DEGs, making it easier to analyze the FLAIL targets, even if the allele can't be used for complementation experiments.

#3.3 Thanks for this suggestion. We plan to confirm key gene expression changes using amiRNA-FLAIL in full revision.

Also, auhors totally neglect putting the Cir1 and Lac8 genes into the context of flowering regulation, but it is something that needs to be done.

#3.4 ____We discussed roles of CIR1 and LAC8 in flowering regulation in Line 260-272. Flowering is fine-tuned to maximize reproductive success and seed production and by endogenous genetic cues and external environmental stimuli such as photoperiod. Nevertheless, many details of the flowering pathways and their integration remain to be investigated____. CIR1 is a circadian clock gene, induced by light and involved in a regulatory feedback loop that controls a subset of the circadian outputs and thus determines flowering time. Our GO analysis supports that a subset of DEGs are connected to the response to red or far red light that contains among other key flowering genes such as ____phytochrome interacting factor____ 4____ (PIF4) and CONSTANS (CO)____. FLAIL also binds the chromatin region of LAC8. LAC8 is a laccase family member that mainly modulates phenylpropanoid pathway for lignin biosynthesis____. Similar to flail, lac8 mutants flower early. While intermediates in this pathway or dysregulation of lignin-related genes could promote flowering in plants, the molecular connections of reduced LAC8 expression to effects on flowering time will require further investigation.

Lastly, the paper needs to be totally rewritten to be even properly evaluated. In its current state it reads like a very short draft.

#3.5 We reorganized the structure of manuscript, improved clarity and provided new mechanistic evidence in Fig. 4G and Fig. S9 to present a more complete manuscript.

The Abstract is weak, the Introduction is written in a such telegraphic style that it is barely readable, in many places there is no connections between sentences leading to the information appear to be presented as random, even if it isn't.

#3.6- We strengthened the Abstract by providing new evidence and improved for the Introduction.

The Results section is written rather rudimentary with information not being sufficiently provided to describe the results but rather scattered between the Results and Figure legends.

#3.7 Thanks for your suggestions, we described each FLAIL length and all constructs in detail in Results, put a schematic of T-DNA and CRISPR mutants in Fig. 1A, moved comparative genomics data to the end of Results and ensured all figures in order.

The Discussion is the best written part of the manuscript.

Thanks for your appreciation of the Discussion.

The Conclusion section carries no specific information and reads more like a little summary suitable for a review article rather than experimental paper.

#3.8 We agree this opinion, this paragraph fits Discussion better and Conclusion was removed.

Therefore, this reviewer thinks that regardless of how authors will choose to proceed with the current experimental version of the manuscript, it'd be in the authors' best interests to at least fully revise the paper before resubmitting anywhere. I'd also advise authors to seek professional editorial help specifically using an editor with the background in the plant sciences.

Authors might also want to consider moving Fig.3 into the Suppl. as it doesn't carry much weight or significance and perhaps make existing figures more meaningful and comprehensive and by including a better diagram of the locus (e.g., Fig. S1), etc.

#3.9 We thank this helpful suggestion. Fig.3 represents the RNA-seq data. In combination with supporting data in the supplementary material, it gives an easy visual readout of the reproducibility of the findings in replicates of stranded RNA-seq. In a new submission, we moved it to Fig. S5B and highlighted 13 differentially expressed flowering genes as well as sense FLAIL in flail3 that were rescued in complementation line in Fig. 3A. Moreover, we gave screenshots of FLAIL itself and four flowering related FLAIL targets in RNA-seq with a clear schematic representation of each locus. We believe these revisions improve Fig. 3.

It's not practical to list all issues with the writing as the paper requires total re-writing, so I can just make a few suggestions without any specific order to help authors improve the paper:

We are happy to improve our manuscript with the help of the reviewers. We addressed all comments including from reviewer #3 with a constructive spirit. However, since colleagues and reviewers #1 and #2 found the manuscript comprehensible to the point where they could make expert-level comments that illustrate understanding of the manuscript, a total re-writing did not feel like the most constructive suggestion to improve the manuscript.

--There is no statement anywhere that states the goal of the study.

#3.10____ We stated the goal of the study in line 50-69 and we think this is a misunderstanding. We summarized three issues currently exist in characterization of functional lncRNA in the last sentence of the first three paragraphs in Background: 1 in Line 50, the broad range of candidate hypotheses by which lncRNA loci may play functional roles call for multiple approaches to distinguish alternative molecular mechanisms. 2 in Line 59, functional characterization of trans-acting lncRNAs remains a key knowledge gap to understand the regulatory contributions of the non-coding genome. 3 in Line 69, the contribution of trans-acting lncRNAs to the regulation of distant flowering genes is currently unclear. So in the last paragraph of the background, we claimed that our goals are to address these questions through characterization of functional FLAIL lncRNA in flowering repression using multiple genetic approaches and various genomic data.

--No rational is provided on why authors decided to examine this specific genomic locus.

#3.11 For several years, our lab studies the rules and roles of non-coding transcription. We characterized and are characterizing several loci with evidence of non-coding transcription in a range of species. Early experiments suggested that FLAIL functioned in flowering, this manuscript clarifies that the function is executed as trans-acting lncRNA of the sense FLAIL isoform.

--Typically, the significance is in studying the function of lncRNA or a group of lncRNAs produced from a genomic locus, I don't think I ever encountered the instances when it was exciting to study just a specific genomic locus. If the locus does indeed have any significance for initiating the study, it needs to be explained.

#3.12 This study is remarkable in many aspects. We fully discuss key strengths in the discussion. First, we ____exhibit a trans-acting lncRNA FLAIL that represses flowering by promoting the expression of floral repressor genes as discussed in Line 247-281_; Second, in Line 284-306, we informed that this study provide a compelling model about how to apply _series of convincing genetic data____ to functionally characterize lncRNA loci. Third, in Line 307-312, evolutionary conserved FLAIL sequences across species is key to characterize the functional _microhomology in other _Brassicaceae.

--The locus can produce lncRNAs, but it can't harbor them.

#3.13 We clarified this confusion by enhancing ____presentation of Fig. 1 and providing a more clear description of each sequencing method and results in Line 81-99. Although we provided evidence that transcription of both sense and antisense FLAIL are more stable in hen2-2, they were clearly observed in chr-DRS in wild type and plaNET-seq in NRBP2-FLAG nrpb2-1 and sense FLAIL was even detected in TSS-seq and TIF-seq in wild type.

--No length of FLAIL lncRNAs or their range is provided in the first section of Results.

#3.14 We gave the length of sense FLAIL in Line 82 and antisense FLAIL in Line 86.

--On many occasions authors don't state rational for doing experiments, which leads to information often flowing as random.

#3.15 we enhanced clarity of the rational for each experiment and made some connections between sentences to make more fluent. For example, in sentences in Line 99, Line 113, Line 126, Line 159, Line 183, Line 214, and Line 219.

--What do authors mean by the subtitle "FLAIL characterizes a trans-acting lncRNA repressing flowering"? How can lncRNA FLAIL or FLAIL locus characterize lncRNA?

#3.16 We changed it to “FLAIL represses flowering as trans-acting lncRNA” in Line 112.

--Check all figures. E.g., Fig. 3B-E mentions only accession numbers for the genes.

#3.17 The systematic gene IDs are a valid way to represent data, in particular for genomics data since it facilitates cross-comparisons. To make it more accessible we also show systematic names of each gene in Fig. 3A-F, Fig. S6 and Table S3.

--It is not clear where exactly the T-DNA insertion is located relative to sense FLAIL in flail3 mutant (Fig. S4).

#3.18 We moved the schematic to clarify this to revised Fig. 1A and the exact T-DNA insertion site is mentioned in the legend.

--- What is the length of the complementing sense FLAIL lncRNA?

#3.19 We now include the length of the complementing sense and antisense FLAILs in Line 351-352.

--Check the description of each and every construct used and provide explanation for each in the Results. E.g., the pFLAIL:gFLAIL18/88 and pasFLAIL:gasFLAIL18/39 constructs aren't explained in Results, and can only be found in Fig. 2 legends.

#3.20 We described each construct including pFLAIL:gFLAIL18/88 and pasFLAIL:gasFLAIL18/39 constructs in Line 133, amiR-FLAIL-11 and amiR-FLAIL-11 in Line 149.

Reviewer #3 (Significance (Required)):

Tens of thousands of lncRNAs have been identified in various eukaryotes, but their biological roles have been shown only for a small fraction of them, and the mechanisms of their action are delineated for only a very few of them. Most of the advances on the field of lncRNAs are reported in metazoan, while the field of lncRNAs in plants is lagging far behind in terms of knowledge about lncRNAs with assigned biological functions or lncRNAs with delineated mechanisms of action. From this point of view, this reviewer is always excited to see any new functional plant lncRNAs for which either biological or mechanistic functions have been determined, and deems the information on this subject significant. The manuscript's findings are potentially very interesting and present a decent body of work that lays a very solid groundwork for future experiments. My main concern about the manuscript's significance in its current form is the fact that no real solid mechanism of action for the described lncRNA or a set of lncRNAs (?) has been demonstrated. The best mechanistically studied lncRNAs in Arabidopsis are involved in the regulation of flowering time, particularly those that function in the vernalization flowering pathway and to lesser extent in autonomous pathway. The new FLAIL lncRNA or lncRNAs (?) described in this manuscript also appear to regulate the flowering time in Arabidopsis, however more experiments would be needed to provide a definite conclusion about how direct FLAIL's effect is and how exactly it functions. That unfortunately obviously diminishes the significance of the manuscript and makes it potentially interesting only to researches studying flowering in Arabidopsis and even then the manuscript results would be incomplete to make solid conclusion.

Lots of functional phenotype have

Additionally, the manuscript requires complete re-writing.

We thank this reviewer for the appreciation of __a decent body of work and a very solid groundwork for future experiments. We are confident that our revisions make the manuscript more comprehensible to highlight the qualities of our manuscript more accessibly.____

__

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 the manuscript by Jin et al authors characterize the FLAIL DNA locus in Arabidopsis (using a wide array of publicly available datasets), which produces a set of sense and anti-sense lncRNAs. Authors determined that the sense FLAIL lncRNA (or a set of sense lncRNAs, which isn't fully clear from the way the data are presented) is involved in flowering time in Arabidopsis based on the fact that the several flail mutants lead to the early flowering phenotype and this flowering defect is complemented by transgenic FLAIL DNA, meaning that FLAIL lncRNA acts in trans. The T-DNA flail3 mutant results in expression changes (up or down) of 1221 genes, including twenty genes linked to flowering in various ways. Expression of a group of these flowering-related genes could be either fully (for eight genes) or partially (for five genes) rescued by transgenic FLAIL. Authors also conducted the ChIRP-seq to determine which genes are physically bound by FLAIL lncRNA genome-wide. It was found that 210 genes in the genome are bound by FLAIL lncRNA. Comparison of the dataset of differentially expressed genes in the T-DNA flail3 mutant with the ChIRP-seq dataset of genes that are bound by FLAIL lncRNA revealed the 12 overlapping genes.

Among these twelve overlapping genes, four were found to be functionally connected to flowering with expression of these four genes being down in flail3 T-DNA mutant. Two out of these four genes were ruled out from being involved in the regulation of flowering by FLAIL. Authors conclude that the two other genes (Cir1 and Lac8) are responsible for the late flowering phenotype of flail mutants based on the three lines of evidence: (i) these genes expression is reduced in the flail mutant, (ii) FLAIL lncRNA directly interacts with these genes chromatin, (iii) the mutants of these genes were previously reported by others to display early flowering phenotypes too. <br /> While I find many of the findings reported in the manuscript very interesting, building a good foundation on which to expand the study and providing a very good leads for follow up experiments, I also have serious concerns about the manuscript in its current form. Most importantly, this reviewer doesn't think that the mechanism of FLAIL lncRNA action was convincingly demonstrated. The main question would be how FLAIL lncRNA works and this question wasn't fully answered. It is great that FLAIL lncRNA binds directly to the two flowering-related genes, but what does it mean? Does it change any chromatin context of these genes quantitatively or qualitatively to affect the transcription? Or does it bind any components of transcriptional machinery and thus controls the transcriptional output? Additionally, flail3 T-DNA mutant affects the expression of 1221 genes and FLAIL lncRNA physically interact with 210 genes, so how can authors be fully sure that FLAIL lncRNA has only direct effect on these two genes and doesn't also contribute to the regulation of the upstream to Cir1 and Lac8 genes or even components of the transcriptional machinery that regulate these genes? Theoretically, doing RNA-seq in the amiR-FLAIL sense lncRNA mutant might have a chance of reducing the number of affected DEGs, making it easier to analyze the FLAIL targets, even if the allele can't be used for complementation experiments. Also, authors totally neglect putting the Cir1 and Lac8 genes into the context of flowering regulation, but it is something that needs to be done. Lastly, the paper needs to be totally rewritten to be even properly evaluated. In its current state it reads like a very short draft. The Abstract is weak, the Introduction is written in a such telegraphic style that it is barely readable, in many places there is no connections between sentences leading to the information appear to be presented as random, even if it isn't. The Results section is written rather rudimentary with information not being sufficiently provided to describe the results but rather scattered between the Results and Figure legends. The Discussion is the best written part of the manuscript. The Conclusion section carries no specific information and reads more like a little summary suitable for a review article rather than experimental paper.

Therefore, this reviewer thinks that regardless of how authors will choose to proceed with the current experimental version of the manuscript, it'd be in the authors' best interests to at least fully revise the paper before resubmitting anywhere. I'd also advise authors to seek professional editorial help specifically using an editor with the background in the plant sciences. Authors might also want to consider moving Fig.3 into the Suppl. as it doesn't carry much weight or significance and perhaps make existing figures more meaningful and comprehensive and by including a better diagram of the locus (e.g., Fig. S1), etc.

It's not practical to list all issues with the writing as the paper requires total re-writing, so I can just make a few suggestions without any specific order to help authors improve the paper:

--There is no statement anywhere that states the goal of the study.

--No rational is provided on why authors decided to examine this specific genomic locus.

--Typically, the significance is in studying the function of lncRNA or a group of lncRNAs produced from a genomic locus, I don't think I ever encountered the instances when it was exciting to study just a specific genomic locus. If the locus does indeed have any significance for initiating the study, it needs to be explained.

--The locus can produce lncRNAs, but it can't harbor them.

--No length of FLAIL lncRNAs or their range is provided in the first section of Results.

--On many occasions authors don't state rational for doing experiments, which leads to information often flowing as random.

--What do authors mean by the subtitle "FLAIL characterizes a trans-acting lncRNA repressing flowering"? How can lncRNA FLAIL or FLAIL locus characterize lncRNA?

--Check all figures. E.g., Fig. 3B-E mentions only accession numbers for the genes.

--It is not clear where exactly the T-DNA insertion is located relative to sense FLAIL in flail3 mutant (Fig. S4). --- What is the length of the complementing sense FLAIL lncRNA?

--Check the description of each and every construct used and provide explanation for each in the Results. E.g., the pFLAIL:gFLAIL18/88 and pasFLAIL:gasFLAIL18/39 constructs aren't explained in Results, and can only be found in Fig. 2 legends.

Significance

Tens of thousands of lncRNAs have been identified in various eukaryotes, but their biological roles have been shown only for a small fraction of them, and the mechanisms of their action are delineated for only a very few of them. Most of the advances on the field of lncRNAs are reported in metazoan, while the field of lncRNAs in plants is lagging far behind in terms of knowledge about lncRNAs with assigned biological functions or lncRNAs with delineated mechanisms of action. From this point of view, this reviewer is always excited to see any new functional plant lncRNAs for which either biological or mechanistic functions have been determined, and deems the information on this subject significant. The manuscript's findings are potentially very interesting and present a decent body of work that lays a very solid groundwork for future experiments. My main concern about the manuscript's significance in its current form is the fact that no real solid mechanism of action for the described lncRNA or a set of lncRNAs (?) has been demonstrated. The best mechanistically studied lncRNAs in Arabidopsis are involved in the regulation of flowering time, particularly those that function in the vernalization flowering pathway and to lesser extent in autonomous pathway. The new FLAIL lncRNA or lncRNAs (?) described in this manuscript also appear to regulate the flowering time in Arabidopsis, however more experiments would be needed to provide a definite conclusion about how direct FLAIL's effect is and how exactly it functions. That unfortunately obviously diminishes the significance of the manuscript and makes it potentially interesting only to researches studying flowering in Arabidopsis and even then the manuscript results would be incomplete to make solid conclusion. Additionally, the manuscript requires complete re-writing.

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

In this ms, the authors identified the FLAIL lncRNA that represses flowering in Arabidopsis from a locus producing sense and antisense transcripts. They use an allelic series involving T-DNA insertions, CRISPR/Cas9 and artificial miRNAs to study the role of FLAIL in flowering. A complementation series of constructs of the flail3 allele allowed them to show that the sense FLAIL lncRNA can act in trans. RNAseq revealed a small group of genes linked to the regulation of flowering whose expression is affected in the mutant and restored in the complementation line. To gain further insight into FLAIL function, the authors used a ChIRPeq approach to test whether the lncRNA can recognize potential target genes along the genome and they could show that FLAIL binds specific genomic regions. Clearly, this paper shows very nice evidence that the FLAIL lncRNA can act in trans to regulate gene expression. Nevertheless, there are certain points that need to be clarified to further support the action of the sense FLAIL transcript.

1.According to Fig. 1 A, the antisense FLAIL is "internal" to the DNA genomic area spanning the sense FLAIL. Hence, with direct RT-qPCR is very difficult to distinguish between these molecules as a minor "RT" activity of the Taq polymerase may lead to detection of low levels of antisense, idem if RDRs may generate low antisense levels. Although I think that the plaNET seq brings strong evidence about the start and ends of these molecules, to measure them by RT-qPCR is not trivial and requires the use of strand-specific RT-PCR using a 5' extension of the oligo and amplification with one oligo of the FLAIL sequence (sense or antisense) and the added oligo. It is not clear how they could distinguish precisely sense and antisense particularly when both RNAs correlate as it is the case here in all alleles (Fig. 1 C and 1D). This should be more explicitly mentioned in the materials and methods section.

2.In Fig. 2, what are the levels of antisense in the complementing lines with the sense transcript? And reciprocally sense levels in antisense constructs? This will definitively demonstrate the assumption that the T-NOS termination will not allow any expression on the other strand. At present, only one of the lncRNAs is measured in each experiment?

3.In Fig. 3, it will be important to also show the FLAIL locus in the flail3 mutants (in comparison to the wt) as well as the transgene locus. Here the reads will be strand specific and furthermore this will allow to show that the transgene is not generating antisense transcripts (through RDRs for gene silencing?) and confirm that the sense FLAIL is required for the complementation.

4.In Fig. S5, the expression of FLAIL is shown in the artificial miRNA lines. Is the antisense FLAIL affected "indirectly" by the cleavage of the amiRNA or remains constant? This is likely the case but should be shown.

5.The ChIRPseq data adds major novelty to the ms and brings new ideas about the way of action of FLAIL. However, are there any common epigenetic states between ChIRP targets (e.g. histone modifications, antisense RNA production, homologies "detected" in the conserved regions between Camelina and Arabidopsis and the target loci? Or others) that may highlight potential mechanisms leading to repression mediated by FLAIL of these loci? There are many databases that could be explored (even during flowering) to search for potential relationships. Although precise description of the mechanism is out of the scope of this ms, this can be discussed in more detail to further expand on the nice data obtained.

Minor comments:

6.In Fig. S3, a global alignment between FLAIL and two loci in Arabidopsis and Camelina is sown. What is the extent of homology? How conserved is this sequence at nucleotide level (small or very long?) to support the conservation of this lncRNA. Are there potential structures conserved among these lncRNAs?

7.In Fig. S4B, arrows may help to understand which seeds were selected.

Significance

This paper is a very nice piece of work and demonstrate the action of a long non-coding RNA (lncRNA) in trans on specific targets involved in the regulation of a developmental process, flowering. There is growing evidences that the non-coding genome hides large number of lncRNAs and there is little detailed genetic support for the action of lncRNAs globally. In contrast to many descriptive papers in the field, this ms demonstrates genetically, through an allelic series and complementation experiments, that this lncRNA locus is involved in flowering regulation and that its sense lncRNA recognizes target loci genome-wide, bringing interesting perspectives on potential new mechanisms of transcriptional regulation mediated by non-coding RNAs.

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 characterized a new lncRNA locus named FLAIL that controls flowering time in Arabidopsis thaliana. The functional validation of this locus is strongly supported by the use of several different tools (CRISPR-Cas9 deletions, T-DNA insertion, amiRNA gene silencing, and transgene complementation of KO lines). It is also suggested that FLAIL lncRNA works in trans but not in cis. There are strong observations supporting that FLAIL works in trans.

Moreover, it is suggested that FLAIL regulates gene expression by interacting with distant chromatin loci. This was assessed using RNA-Seq and ChIRP-Seq. Yet, the overlap between DEGs in the flail mutant and FLAIL binding sites at the chromatin is very small, with only 12 genes. From those, only 2 flowering genes' expression was rescued by FLAIL transgene complementation. The final conclusion that FLAIL lncRNA represses flowering by direct inhibition of the 2 flowering genes expression is correlative, and lacks genetic validation. In addition inspection of the supplementary file shows that the ChIRP analysis was done without filtering for the FDR so that some of the positive hits have an FDR of 0,232. In addition, many of the peaks land in intergenic regions with is not mentioned in the text a graph with the position of the peaks in respect to nearby genes would help.

In one sentence, the authors used the right model system and methodology, including advanced techniques, to characterize a new trans-acting lncRNA important for controlling the flowering time in Arabidopsis but lack evidence supporting a mechanism of action that goes beyond the interaction with several chromatin loci.

minor points:

line#63-64 the authors say the COLDAIR and ASL work on FLC in cis in my view the original papers suggested/showed they work in trans.

Fig 1B please add some more protein-coding RNAs for the bio-info analysis for comparison

Order of Supplementary Fig citation is mixed with S2 coming before S1B

It would help the reader to have a schematic of the crisper deletions, T-DNA insertion, and position of primers used for the RT-qPCR.

In the supplementary PDF file, some of the text is missing on page 3 beginning and end of lines.

Significance

The use of several different tools to validate the biological function of FLAIL locus is a major strength of this work.

The authors propose that flowering time and its gene regulation are controlled by sense FLAIL lncRNAs. However, the sense transcription of FLAIL locus is not detected in wild-type plants by TSS-Seq, TIF-Seq, or plaNET-Seq. If the authors would have explored further the expression of FLAIL transcripts in different stages of development (vegetative and non-vegetative) and in response to different conditions, it would make their claims on the function of FLAIL lncRNAs more convincing. Additionally, flail mutants could have been obtained in the hen-2 background, since it's there where we can observe FLAIL transcription.

FLAIL locus lays on the proximal promoter region of PORCUPINE (PCP), an important regulator of plant development. As flail mutants, pcp mutants display an early flowering phenotype. The authors show no link between FLAIL and PCP from the overlap between re-analysis of published RNA-Seq data for pcp and RNA-Seq and ChIRP-Seq from the authors. This analysis is not enough to exclude the involvement of PCP from the FLAIL function. PCP expression using RT-qPCR should be performed in flail mutants to further support that FLAIL works independently from PCP.

This work does not hypothesize any molecular mechanism besides the interaction of FLAIL lncRNAs with several chromatin loci. It was recently proposed in Arabidopsis that a trans-acting lncRNA interacts with distant loci via the formation of R-loops. The authors do not comment on that. This work would benefit in correlating FLAIL binding sites with R-loop-forming regions mapped in Arabidopsis, regardless of the results from this analysis. Additionally, the authors could attempt to look for a motif responsible for FLAIL binding.

Most of the key conclusions are convincing, except for the flowering time control directly through CIR1 and LAC8, which should be mentioned as speculative

The words locus and loci are latin and they should be written in italic. The word Brassicaceae, referring to the family should be in italic, and should not be "Brassicaceaes". The word analysis has the wrong spelling.

I was asked "How much time do you estimate the authors will need to complete the suggested revisions: this is difficult to answer as it depends to which level the author would like to take their work. In my view, if all new experiments would have to be started from scratch it is too far away to be estimated.

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-2021-01118

Corresponding author(s): Jun, Nakayama and Kentaro, Semba

1. General Statements

We are grateful to all of the reviewers for their critical comments and insightful suggestions that have helped us considerably improve our paper. As indicated in the responses that follow, we have taken all of these comments and suggestions into account in the revised version of our paper, including the supplementary information.

In the revised manuscript, we focus on the existence of two cancer stem cell-like populations in TNBC xenograft model and patients. The response to each reviewer is described below.

Sincerely,

Jun Nakayama

Kentaro Semba

Department of Life Science and Medical Bioscience

School of Advanced Science and Engineering

Waseda University

E-mail: junakaya@ncc.go.jp or jnakayama.re@gmail.com to JN

ksemba@waseda.jp to KS

2. Point-by-point description of the revisions

Reviewer #1 (Evidence, reproducibility and clarity (Required)): * **Summary:** Nakayama and colleagues use their previously developed automated tissue microdissection punching platform to perform spatial transcriptomics on a breast cancer xenograft model. Using transcriptomics on multiple clumps of 10-30 cells from different regions in a tumor and a lymph node metastasis they identified different cell-type clusters. Two of these clusters expressed different cancer stem cell markers. This led the authors to suggest that two distinct cancer stem cell(-like) populations may exist within one (breast) tumor, which could potentially make tumors more drug-resilient.

**Major comments:** While the quality of the presented sequencing data is good and the manuscript is mostly written in a clear and accessible style, there are some concerns that limit the impact of this story. Most importantly, the manuscript in its present form does not convince me that the MDA-MB-231 xenografts indeed contain two distinct populations of cancer stem(-like) cells.

1.The data obtained are not single cell data, which makes it difficult -if not impossible- to draw conclusions about presence of cancer stem cells. Each data point is the average of 10-30 cells, and the interpretation of the data is severely limited by this. How can the quantification of expression of CD44/MYC/HMGA1 in clumps of 10-30 cells teach us something about the stemness of tumor cells? *

Answer: We would thank the comment. The reviewer’s suggestion is an important point; however, this is technical limitation of spatial transcriptomics technology. Most advanced spatial transcriptomics technologies, e.g. Visium (10x Genomics), also have the same problem. It means that our technology and the advanced technologies are technics to analyze gene expression and characteristics of tissues from 10-30 cells in each spot. Although high resolution spatial transcriptomics has been developed in 2021 [1], it is not generally used yet as described in the comment (Significance) from reviewer1.

From our spatial analysis, we identified that CD44, MYC, and HMGA1 were expressed from human cancer cell. Their expression profiles were distinct among specific parts of the tumor section. To validate the existence of two types of cancer stem-like cells in TNBC tumors, we performed the additional analysis with the public scRNA-seq datasets of high-metastatic MDA-MB-23-LM2 xenograft model (GSE163210) [2]. This study performed scRNA-seq analysis of primary tumor and circulating tumor cells in MDA-MB-231-LM2 xenograft model. We analyzed it with Seurat/R (Figure A-1). As a result of reanalysis, HMGA1 and CD44 expression were confirmed at single-cell resolution (Figure A-2,3). These results verified the existence of two cancer stem cell-like populations (HMGA1-high, CD44-high) in MDA-MB-231 xenograft. Hence, the study of MDA-MB-231 xenograft supported our findings from spatial transcriptomics.

Additionally, we performed the immuno-staining of sections using anti-CD44 antibody and anti-HMGA1 antibody as described in reviewer’s comment 5. As a result, CD44 and HMGA1 were detected in primary tumor sections. There were cells that express either CD44 or HMGA1 and cells that co-express both CD44 and HMGA1 (Figure B). We believe that our findings are solid results because the findings were also validated by other methods.

In the revised manuscript, Figure A are incorporated as Figure 3B-E. Figure B is incorporated as Figure 3A. Hope our new results will be now accepted by the learned Reviewer and Editor.

Figure A-1. Reanalysis of scRNA-seq of metastatic MDA-MB-231 xenograft

Flowchart of the public single-cell RNA-seq (scRNA-seq) reanalysis using GSE163210 datasets.

Figure A-2. UMAP plots of xenograft and CD44/HMGA1 expression

UMAP plot of MDA-MB-231-LM2 xenograft tumors and circulating tumor cells (Left). Expression of CD44 and HMGA1 in the UMAP plot (Right).

Figure A-3. Pie chart of CD44/HMGA1 positive cancer cells in MDA-MB-231 xenograft

Pie chart of cancer stem cell-like population ratio in MDA-MB-231-LM2 xenografts.

Figure B. Fluorescent immuno-staining of MDA-MB-231 primary tumor

Representative images immunostained with CD44 and HMGA1 in primary tumor sections of the MDA-MB-231 xenograft model. Red: HMGA1, Green: CD44, and Blue: Nucleus. Scale bars, 20 μm (left), 10 μm (right). White arrows represent cancer cells that independently expressed or co-expressed.

* 2.Furthermore, the authors should better explain their data analysis strategy with identification of gene expression profiles. It is unclear how they found CD44, MYC, and HMGA1 other than by cherry-picking from the list of cluster markers. *Answer: In this research, to identify the characteristics of clusters, we analyzed differentially expressed genes (DEGs) by ‘FindAllMarkers’ function of Seurat. As a result, ‘Cluster 0’ significantly expressed HMGA1 gene, and ‘cluster 1’ significantly expressed CD44. HMGA1 and CD44 are popular cancer stem cell markers in triple-negative breast cancer [3, 4]. In this study, we focus on metastasis-related genes and cancer stem cell markers (described in introduction section). Therefore, we focus on cancer-stem cell markers in the presented study. Cancer stemness is an important concept in cancer metastasis [5-7]. These results suggested that the existence of two cancer stem cell-like populations could potentially make tumors more drug-resilient in xenograft models and clinical patients.

To improve the manuscript, we revised the description in the revised manuscript (Pages 5-6, Lines 97-105).

* 3.Following up on the above point: I looked in the supplementary tables, but couldn't find MYC. How did the authors conclude that MYC is involved in cluster 1? In fact, when I ran a quick analysis in EnrichR, I saw that putative MYC target genes were strongly enriched among the markers in the HMGA1 cluster, but not the CD44/MYC. That's opposite to what I would expect. *__Answer: __We apologize for our confusing data and description. First, we found the expression of CD44 and HMGA1 in each cluster. Therefore, we performed the up-stream enrichment analysis using gene signatures of FindAllMakers by Metascape. From the result of enrichment analysis, we found the MYC activation in CD44 high-cluster; therefore, we named the cluster “CD44/MYC-high” cluster.

To improve the manuscript, we revised the Figure2, Supplementary Table S3, and manuscript (Pages 5-6, Lines 103-106).

* 4.All data were produced from 1 primary tumor and 1 metastasis. Thus, reproducibility and robustness of the methodology cannot be evaluated. The interpretation of the data could be strengthened when xenografts from at least 3 different mice are shown. *__Answer: __We would thank the suggestion. As the reviewer’s comment, we performed 1 primary tumor and 1 metastasis lesion from a transplanted mouse. Since this experiment take a long time, we tried to validate the findings by other methods (Figure A: scRNA-seq analysis of MDA-MB-231 xenografts, Figure B: Immuno-staining of MDA-MB-231 primary tumor, Figure C: scRNA-seq analysis of TNBC patients).

First, we reanalyzed the public dataset which performed single-cell RNA-seq analysis of MDA-MB-231 xenografted tumor and circulating tumor cells in immunodeficient mice as shown in the answer to comment 1 (Figure A). Next, we performed the immuno-staining of sections using anti-CD44 antibody and anti-HMGA1 antibody as described in reviewer’s comment 5. As results, CD44 and HMGA1 were detected in primary tumor sections. There were cells that express either CD44 or HMGA1 and cells that co-express both CD44 and HMGA1 (Figure B). Next, we performed the reanalysis of 19 scRNA-seq samples from integrated 3 TNBC cohorts (Figure C-1). In a UMAP plot, differences between CD44-positive cancer cell and HMGA1-positive cancer cell were observed; however, these cells did not visually form the specific clusters (Figure C-2). CD44 and HMGA1 expressed globally in the UMAP plot, but CD44 makes some specific clusters (cluster at right side). Additionally, following the comment, we performed the population analysis in each patient (Figure C-3 and C-4). Detection of double-positive population in TNBC patients suggested that the population may be more undifferentiated cancer stem cells diving into both CD44-positive cells and HMGA1-positive cells.

In addition, we reanalyzed primary tumors and metastasis lesions from other mice as a test trial sample (Figure D-1). The microspots including test trial samples showed 3 human clusters which were classified into CD44/MYC, HMGA1, and Marker-low clusters. We believe that our findings are solid results because the findings were also validated by other methods.

In the revised manuscript, Figure A are incorporated as Figure 3B-E. Figure B is incorporated as Figure 3A. Figure C is incorporated as Figure 5. We only showed Figure D in the response to the reviewer’s comment. Hope our new results will be now accepted by the learned Reviewer and Editor.

Figure C-1. Reanalysis of integrated TNBC patients scRNA-seq

A flowchart of the reanalysis of a public scRNA-seq dataset. We downloaded GSE161529, GSE176078, and GSE180286 (scRNA-seq data of 19 TNBC patients). Integrated datasets were analyzed with Seurat. Log normalization, scaling, PCA and UMAP visualization were performed following the basic protocol in Seurat. To extract the cancer cells, cells expressing EPCAM/KRT8 (epithelial marker) were filtered. A UMAP plot of cancer cell from 19 TNBC patients (right).

Figure C-2. CD44/HMGA1 expression in TNBC patients

Expression analysis of CD44 (Expression level > 2) and HMGA1 (Expression level > 2) with UMAP plots.

Figure C-3. CD44/HMGA1-positive cancer cell with UMAP plot

UMAP plots of CD44-high, HMGA1-high, HMGA1/CD44-high, and Negative cancer cells.

Figure C-4. Ratio of CD44/HMGA1-positive cancer cell in each patient

The bar plot showed the ratio of cancer cells that expressed CD44 and HMGA1.

Figure D-1. Analysis of microspots of MDA-MB-231 xenografts including test trial samples

UMAP plots of CD44-high, HMGA1-high, and Marker-low clusters with test trial samples (2 primary tumors and 1 lung metastasis). ‘Primary tumor 1’ has 20 microspots, ‘Primary tumor 2’ has 24 microspots, and ‘lung metastasis’ has 7 microspots. Most microspots of lung metastasis failed extraction of RNA; therefore, these spots classified into Marker-low cluster.

Figure D-2. Expression analysis of CD44, HMGA1, and MYC

Feature plot of CD44-high, HMGA1-high, and Marker-low clusters with test trial samples.

* 5.The only methodology is single cell RNA-sequencing. Immuno-staining on relevant markers such as CD44, MYC, HMGA1 plus human epithelium and cell cycle markers would provide strong additional support for the claims made by the authors, because it's a complementary technique and it allows quantification at single cell resolution. *__Answer: __We would thank the comment. As described in the responses to the reviewer’s comment 1 and 4, we performed the immuno-staining of sections using anti-CD44 antibody and anti-HMGA1 antibody as described in reviewer’s comment 5. As a result, CD44 and HMGA1 were detected in primary tumor sections. There were cells that express either CD44 or HMGA1 and cells that co-express both CD44 and HMGA1 (Figure B).

In the revised manuscript, Figure B is incorporated as Figure 3A.

* 6.Line 173-175. The marker-low cluster look to me simply like spots containing a relatively high amount of dead/dying (tumor) cells. The identity/state of cells in the marker-low cluster should be characterized and discussed more extensively. *__Answer: __We would thank the comment. This suggestion is important. In fact, total count of RNA in the Marker-low cluster decreased as compared to HMGA1-high and CD44/MYC-high (Supplementary Figure S1B). Additionally, Ttr-high mouse cluster also has low total count of RNA (Supplementary Figure S1C).

Following the comment, we described that the Marker-low cluster and Ttr-high cluster have the possibility to include dead/dying cells (Page 13, Lines 268-279).

* 7.Figure 5 and accompanying text in line 182-194; the authors try to infer cell-to-cell interactions using a previously published tool. However, any biological interpretation is lacking. What can be concluded from this analysis? *__Answer: __Initially, algorithms of cell-to-cell interaction were reported with previously published tool [8, 9]; however, in this manuscript, we originally conducted the code for cell-to-cell interaction with the interaction database of the Bader laboratory from Toronto University (https://baderlab.org/CellCellInteractions#Download_Data) as previously described [10, 11]. We aimed to estimate the cell-to-cell interaction in each spot (including 10-30 cells). We think that this analysis will be helpful for discovering the cancer stem cell niche and metastatic niche [6].

However, in the revised manuscript, we focused on the existence of two cancer stem cell-like populations in TNBC xenograft and patients. Therefore, CCI analysis in previous Figure 5 moved to Supplementary Figure S7. Previous Figure 6 is removed from revised manuscript.

* 8.Figure 6. Can the authors please explain more clearly what they mean by "PT" and "Mix" groups? I had a very hard time to understand what the data in figure mean. Again, an overall interpretation at the end (line 211) is lacking. *__Answer: __We apologize for the confusing result. We examined the combinations of human cancer cell cluster and mouse stromal cell cluster. To summarize, there are 10 combinations in the MDA-MB-231 xenograft. The combination groups in only primary tumor were named “PT”; on the other hand, the combination groups in both primary tumor and lymph-node metastasis were named “Mix”. These CCI analysis focused on cluster types of cancer cell and stromal cell. However, according to this revision, our presented study mainly focuses on the existence of two types of cancer stem cell-like population in TNBC xenograft and patients. Therefore, CCI analysis with cluster types was deleted from revised manuscript.

In the revised manuscript, we focused on the existence of two cancer stem cell-like populations in TNBC xenograft and patients. Previous Figure 6 was removed from the revised manuscript.

* 9.Figure 7. I like the effort to align the results with public scRNA-seq data. But although the expression of the cluster-signatures is heterogeneous, there is no evidence for distinct (CSC-like) cell populations. Why don't these HMGA1 vs CD44 signature cells cluster away from each other in the UMAPs? Perhaps the patient-to-patient heterogeneity overwhelms differences within tumors, but in that case the authors could re-run their analysis for each patient separately, to make 6 patient-specific UMAPs. In its present form, this analysis does not convince me that two distinct CSC(-like) populations within one TNBC exist. *Answer: We would thank the comment. To improve the quality of reanalysis of clinical cohorts, we performed the reanalysis of 19 scRNA-seq samples from integrated 3 TNBC cohorts (Figure C-1). In a UMAP plot, there are differences between CD44-positive cancer cells and HMGA1-positive cancer cells; however, these cells did not visually form the specific clusters (Figure C-2). CD44 and HMGA1 were expressed globally in the UMAP plot, but CD44 made some specific clusters (cluster at right side). Additionally, following the comment, we performed the population analysis in each patient (Figure C-3 and C-4). There is double-positive population in TNBC patients suggesting that this population may be more undifferentiated cancer stem cells, dividing into both CD44-positive cells and HMGA1-positive cells.

In the revised manuscript, Figure C is incorporated as Figure 5.

* **Minor comments:** 10.In the Supplemental table 2 noticed that many of the marker genes have adjusted P values well above 0.05 (and even above 0.1). That makes the statistical analysis rather weak. This could especially be problematic since the authors entirely base their main claims on this marker analysis, and I recommend that the authors use more stringent P-value cut-offs in the cluster analysis. *Answer: We would thank the comment. We reshaped the list of differentially expressed genes (DEGs). Significantly expressed genes (adjusted p-value In mouse clusters, the enrichment analysis using significantly DEGs showed that only Tcell-like clusters had a lot of enriched terms. Citric acid (TCA) cycle, chemical stress response, and fatty acid oxidation were enriched in Tcell-like populations (Page 7, Lines 141-144).

In the revised manuscript, enrichment analyses are showed as Supplementary Figure S2 and S3B. We revised the sentence of enrichment analyses (Page 6, Lines 114-121), (Page 7, Lines 141-144). The network visualization of enrichment analysis was removed from the revised manuscript because this result did not support conclusions of the presented study.

* 11.Line 129/130. If I look at figure 3A, I don't see this tendency that the authors describe. Can the authors provide statistical support or visual aid to make their claim more apparent to the reader? *__Answer: __We would thank the suggestion. Following the comment, we performed the statistical analysis of spot position. The spots were categorized outer side (tumor edge) and Inner site (Center of tumor) in the primary tumor section (Figure E-1 upside). We counted the spot numbers of the clusters (Figure E-1 table) and performed statistical test by chi-test. As a result, CD44/MYC clusters significantly resided at outer side of primary tumor (Figure E-1 barplot). On the other hand, the spots in lymph-node metastasis are not readily defined the outer or inner. In addition, cell cycle analysis in the primary tumor and lymph node metastasis was performed with statistical test. As a result, HMGA1-high cluster and CD44/MYC-high cluster significantly proliferated in the lymph node metastasis section (Figure E-2).

Therefore, in the revised manuscript, we revised the sentence of spot position in lymph-node metastasis (Pages 8-9, Lines 159-172). Figure E-1 is incorporated as Figure 4D. Figure E-2 is incorporated as Figure 4F. Hope our new results will be now accepted by the Reviewer and Editor.

Figure E-1. Statistical analysis of spot position

Chi-test was performed by R. *p Figure E-2. Statistical analysis of cell cycle index

Fisher’s exact test was performed by R. *p * 12.Line 217; shouldn't this be 6 patients? I see six clusters and in the original paper six patients are mentioned. *Answer: We would thank the comment. ‘6 patients’ is correct, we revised it. However, in the revised manuscript, we added integrated analysis of TNBC as shown in the answer to comment 9.

Previous reanalysis of clinical scRNA-seq (previous Figure 7) was removed from the revised manuscript. The reanalysis using 3 integrated TNBC cohorts (Figure C) is incorporated as Figure 5.

Reviewer #1 (Significance (Required)): * Conceptual/biological impact: Showing the existence of distinct populations of CSCs within one (breast-)tumor potentially has a high impact on the field of fundamental and translational cancer research. As the authors state, it could be one key reason underlying drug resistance. However, the technology used by the authors does in my view not allow to make such a claim. First and foremost because the technology does not allow analysis at single cell resolution.

Technical impact: The platform used by the authors can be of interest for some applications, but they already published this in Scientic Reports a few years ago. I'm afraid that with the rapid recent developments in the field of spatial single cell transcriptomics (See for example Srivatsan et al Science 2021; 373: 111-117), the technical impact on the field is relatively low.

Audience: Researchers in the field of cancer biology with an interest to perform low-cost molecular analysis at low-resolution spatial-resolved tissue specimens (transcriptomics, but perhaps expanded with bisulfite sequencing, or ATAC sequencing) could be interested in the technology presented in this manuscript.

My expertise: single cell transcriptomics, (cancer) cell cycle, cancer drug resistance, cell plasticity, mouse models. *

**Referee Cross-commenting** I have read the comments and align mostly with reviewer #2. The authors need to improve this manuscript a lot before it's suitable for publication in any of the Review Commons journals. Answer: We are grateful to the reviewers. As indicated in the responses that follow, we have taken all of these comments and suggestions into account in the revised version of our paper, including the supplementary information.

*

*

Reviewer #2 (Evidence, reproducibility and clarity (Required)): * This manuscript uses spatial transcriptomics to perform single cell-like expression analysis between a breast cancer cell line and tumor microenvironment in mice xenografted with these cells. Unfortunately, from the title, abstract, and introduction, it is difficult to understand exactly what the authors are focusing and discussing. It is also unclear the advantage of their technique for evaluating the populations observed within this manuscript. Furthermore, there is very little explanation of the results, and it does not appear to be a scientific logical structure. Hence, this manuscript is not suitable for acceptance in the journal. In order to improve the scientific quality of this study, the following concerns are presented.

**Major concerns:** 1.Is cell-cell interaction (CCI) analysis novel method? If so, please specify detail in the manuscript. If the basic concept and the principle of CCI analysis have not been published, please mention in the discussion section as a limitation that a manuscript on CCI analysis is under submission to the preprint. In addition, please revise the abstract and related text. *__Answer: __Initially, algorithms of cell-to-cell interaction were reported with previously published tool [8, 9]; however, in this manuscript, we originally conducted the code for cell-to-cell interaction with the interaction database of the Bader laboratory from Toronto University (https://baderlab.org/CellCellInteractions#Download_Data) as previously described [10, 11]. We aimed to estimate the cell-to-cell interaction in each spot (including 10-30 cells). We think that this analysis will be helpful for discovering the cancer stem cell niche and metastatic niche [6].

However, in the revised manuscript, we focused on the existence of two cancer stem cell-like populations in TNBC xenograft and patients. Therefore, CCI analysis in previous Figure 5 is moved to Supplementary Figure S7. Previous Figure 6 are removed from the revised manuscript. We revised the description in the manuscript (Page 18, Lines 385-387).

* 2.The reviewer thinks that spatial transcriptomics plays an important role in your manuscript. Please describe the technique in the introduction. *__Answer: __We would thank the comments. Following the comments, we described the spatial technics in Introduction section. We revised the manuscript (Page 4, Lines 63-65) (Page 12, Lines 250-253).

* 3.The classification by expression profile (HMGA1, CD44/MYC and marker-low) lacks an explanation. Authors should mention in detail how these populations were extracted from breast cancer cell lines. *Answer: In this research, to identify the characteristics of clusters, we analyzed differentially expressed genes (DEGs) by FindAllmarkers function of Seurat. As a result, ‘Cluster 0’ significantly expressed HMGA1 gene, and ‘cluster 1’ significantly expressed CD44. Next, we performed the up-stream enrichment analysis using gene signatures of FindAllMakers by Metascape. From result of enrichment analysis, we found the MYC activation in CD44 high-cluster; therefore, we named the cluster “CD44/MYC-high” cluster.

HMGA1 and CD44 are popular cancer stem cell markers in triple-negative breast cancer [3, 4]; therefore, we focus on cancer-stem cell marker in presented study. Cancer stemness is an important concept in cancer metastasis [5-7].These results suggested that the existence of two cancer stem cell-like populations could potentially make tumors more drug-resilient in xenograft model and clinical patient.

To improve the manuscript, we revised the Figure2, Supplementary Table S2 and S4, and manuscript (Pages 5-6, Lines 97-106).

* 4.The description of the results is back and forth and confusing. Please reconsider the flow of the analysis. *__Answer: __We would thank the comment. We reconsidered the description and structure of manuscript. In revised manuscript, we focused on the existence of two cancer stem cell-like populations in TNBC xenograft and patients.

To improve the manuscript, we revised the Figure2 for examination of cluster characteristics by clustering and gene expression profiling. Figure 3 was revised for the validation of two cancer stem cell-like populations in TNBC xenograft model. Figure 4 was revised for the elucidation of spatial characteristics of each cluster. Figure 5 was revised for the validation of two cancer stem cell-like populations in TNBC patients.

* 5.How did you evaluate the outsides of the samples with very different spot positions in Figure 3A? Please mention your evaluation method in a scientific manner. In particular, authors should clearly indicate the outer evaluation for the metastatic case. *

Answer: We would thank the suggestion. Following the comment, we performed the statistical analysis of spot position. The spots were categorized outer side (tumor edge) and Inner site (Center of tumor) in primary tumor section (Figure E-1 upside). We counted the spot numbers of the clusters (Figure E-1 table) and performed statistical test by chi-test. As a result, CD44/MYC clusters significantly resided at outer side of primary tumor (Figure E-1 bar plot). On the other hand, the spots in lymph-node metastasis are not readily defined the outer or inner. In addition, cell cycle analysis in the primary tumor and lymph node metastasis was performed with statistical test. As a result, HMGA1-high cluster and CD44/MYC-high cluster significantly proliferated in the lymph node metastasis section (Figure E-2).

Therefore, in the revised manuscript, we revised the sentence of spot position in lymph-node metastasis (Pages 8-9, Lines 153-172). Figure E-1 are incorporated as Figure 4D. Figure E-2 are incorporated as Figure 4F. Hope our new results will be now accepted by the Reviewer and Editor.

Figure E-1. Statistical analysis of spot position

Chi-test was performed by R. *p Figure E-2. Statistical analysis of cell cycle index

Fisher’s exact test was performed by R. *p * 6.The spots in primary tumor have few counts derived from mouse stromal/immune cells, as shown in Figure S1A. Nevertheless, Figure 3C shows that mouse stromal/immune cells are evaluated in the same way in primary and metastatic sites. The reviewer thinks that the regions identified as Tcell-like in the metastatic site, where there are many mouse-derived counts, and in the primary, where there are few mouse-derived counts, do not have the same characteristics. If many mouse-derived counts were detected in a spot using the spatial transcriptomics, then there must be many mouse-derived cells in the spot. Please discuss how this expression is evaluated on this technique, which is not a single cell analysis. *__Answer: __We would thank the comment. The reviewer’s suggestion is an important point; however, this suggestion is technical limitation of spatial transcriptomics technology. Most advanced spatial transcriptomics technologies, e.g. Visium (10x Genomics), also have the same problem. It means that our technology and the advanced technologies are technics to analyze gene expression and characteristics of tissues from 10-30 cells in each spot.

In this spatial transcriptome analysis of mouse genes, we first performed the log normalization and scaling. Since Seurat used variable features among the samples for single-cell or spot clustering, we extracted the variable features for detection of clusters using the ‘FindVariableFeatures’ function. PCA and clustering using only mouse genes was performed for detecting the neighboring samples. After the clustering of mouse spots, we identified the character of clusters by finding the gene signatures. As the indication by the reviewer, the detected RNA counts and features are different, so it is difficult to define the exact character and cell type of stromal cells. Theoretically, spatial transcriptomics could only detect some kinds of stromal cells expressing the T-cell marker gene in the spot. Therefore, we named the cluster as “Tcell-like”. Not all of the Tcell-like cluster have the same characteristics or cell types, but they certainly express T-cell marker genes. This is also a technical limitation of spatial transcriptomics. Spatial transcriptomics with higher resolution probably is able to detect the stromal cells as a single-cell resolution, such as the one developed in previous research [1].

In the revised manuscript, we focused on the two types of cancer stem cell-like populations that were validated by other methods (scRNA-seq and Immuno-staining). As the method is not able to define the exact cluster characters, we moved CCI analyses to supplementary figures or removed partly.

We also revised the discussion in the revised manuscript (Pages 13-14, Lines 279-283).

* 7.Please explain how the gene symbols listed in Figure 4A were selected. Also, please indicate the characteristics of the gene groups that are not listed. *__Answer: __We selected the gene signature list from results of ‘FindAllMarker’ function in Seurat. ‘FindAllMarker’ function enables to extract the significantly expressed genes in each cluster. Heatmap in previous Figure 4A was drawn using these marker genes (Adjusted p-value 0.1). Highlighted genes in the heatmap have been reported as cancer-related genes or cell cycle-related genes.

The genes used for drawing heatmap are shown in Supplementary Table S2 and S4.

* 8.Please describe the details of the division and cycle index in lines 141-142. *__Answer: __Cell cycle index is a basic function of Seurat [12] (https://satijalab.org/seurat/archive/v3.1/cell_cycle_vignette.html). A list of cell cycle markers is loaded with Seurat. We can segregate this list into markers of G2/M phase and markers of S phase. We subjected this function into our spatial transcriptomics to estimate the cell cycle in each spot.

We revised the description manuscript (Page 16, Lines 331-332).

* 9.In Line 148-151, the expression and prognosis of TMSB10, CTSD, and LGALS1 is mentioned based on the previous reports. Aren't these findings the result of bulk? Is the HMGA1 cluster that the authors found involved in the prognosis of mice? Please clarify, as it is unclear what you want to discuss. *

Answer: We apologize for our confusing data and description. These highlighted genes (TMSB10, CTSD, LGALS1, CENPK, and CENPN) were extracted as DEGs of human cancer clusters (Supplementary Table S2). Previously, these genes have been reported as cancer-related genes or cell cycle-related genes, described in the manuscript (Page 6, Lines 107-110). To show the other expressed genes in each human cluster, we focused on these genes in the manuscript.

We extracted the gene signatures from DEGs and showed the gene signatures from HMGA1-high cluster correlated to poor prognosis in TNBC patients. Our data suggested that the HMGA1 signatures from the microspot resolution has the potential to be a novel biomarker for diagnosis, and HMGA1-high cancer stem cells may contribute to poor prognosis.

In this revision, since we reperformed DEGs analysis with significant threshold; therefore, survival analysis was reperformed with novel gene signatures with METABRIC TNBC cohorts (Figure F).

To improve the manuscript, we revised the description of DEGs extraction and heatmap (Page 6, Lines 106-112). Hope our Reviewer will approve this revised sentence.

Figure F. Survival analysis with gene signatures of HMGA1-high and CD44/MYC-high

Survival analysis of TNBC patients (claudin-low subtype and basal-like subtype) in METABRIC cohorts by the Kaplan-Meier method. (Left) Survival analysis with the expression of the HMGA1 signatures (High = 151, Low = 247). Shading along the curve indicates 95% confidential interval. Log-rank test, p = 0.012. (Right) Survival analysis with the expression of the CD44/MYC signatures (High = 333, Low = 65). Log-rank test, p = 0.079.

* 10.Please provide details of all statistical tests used in this manuscript and describe significance levels used in the p-values and FDR. *__Answer: __We performed the extraction of differentially expressed genes (DEGs) by ‘FindAllMarkers’ function with MAST method. MAST method identifies differentially expressed genes between two groups of cells using a hurdle model tailored to scRNA-seq data [13]. Adjusted p-value is calculated based on Bonferroni correction using all features in the dataset. In spatial spot analysis, statistical analyses were performed by Chi-test and Fisher’s exact test.

We revised materials and methods section in the manuscript (Page 19, Lines 391-394).

* 11.Please mention CCI score (line 198). *Answer: As described in answer to comment 1, the algorithms of CCI score calculation were performed using previously published tool [8, 9]; however, we originally conducted the code for cell-to-cell interaction with the interaction database of the Bader laboratory from Toronto University (https://baderlab.org/CellCellInteractions#Download_Data). We extracted the genes whose expression value was greater than 2. We selected the combinations representing ligand__-__receptor interactions, in which both ligand genes and receptor genes were expressed in the same spot.

We revised materials and methods section in the manuscript and Supplementary Legends (Page 18, Lines 385-387).

* 12.Lines 204-206 and Figure 6G show specific interaction of ITGB1 and CST3, but it is unclear why only these molecules were extracted. What about the other molecules? At least ITGB1 is not scored in mix5. *Answer: We selected genes that have been reported as cancer-related ones in breast cancer to discuss the interactions in primary tumor and lymph-node metastasis. However, according to this revision, our presented study mainly focused on the existence of two types of cancer stem cell-like population in TNBC xenografts and patients. Therefore, CCI analysis with cluster types moved to supplementary Figure or some were not shown now.

In the revised manuscript, previous Figure 6 is removed.

* 13.HMGA1 signature appears in Line 214, please explain in detail. *__Answer: __As described in answer to comment 7, we selected the gene signature list from results of ‘FindAllMarker’ function. ‘FindAllMarker’ function enables to extract the significantly expressed genes in each cluster. HMGA1 signature genes were selected from significantly differentially expressed genes of HMGA1-high clusters.

We revised the description in the revised manuscript (Pages 9-10, Lines 190-193).

* 14.Authors should discuss how the previously reported bulk expression data used in Figure 7E can be linked to the single-cell-like analysis in this study. *__Answer: __Previous research reported that gene signatures extracted from specific clusters in scRNA-seq study have the potential to be a prognosis marker [14]. We showed the gene signatures from HMGA1-high cluster correlated to poor prognosis in TNBC patients. Our results suggested that the gene signatures from the resolution of microspot (10-30 cells) could have the potential to be prognosis markers. This punching microdissection system enables to extract only the parts of a section that are necessary for diagnosis of cancer and to analyze at low-cost. It could be applied to diagnostics instead of the laser-capture microdissection methods.

We performed additional survival analysis with METABRIC cohorts. As described in this revision, since we reperformed DEGs analysis with significant threshold, survival analysis was reperformed with novel gene signatures with METABRIC TNBC cohorts (Figure F).

In revised manuscript, Figure F were incorporated as Figure 6. The usefulness of gene signatures from microspot resolution was additionally discussed (Page 12, Lines 242-245, 250-253).

* **Minor concerns:** 15.Please describe how the normalized centrality was calculated in UMAP algorithm and explain what this means in the results. __Answer: __The data showed that the expressional diversity in each cluster based on the network centrality of a correlational network with graph theory. The differences in the centrality among the clusters suggested expressional diversity in each (Supplementary Figure 4). Higher centrality represented lower expressional diversity and vice versa*. The detailed method for the calculation of centrality was previously shown to reveal the difference between smokers and never-smokers [10, 11].

We added the description in the Legend (Pages 7-8, Lines 145-150).

* 16.Please mention an explanation for the red X in Figure 1B to the legend. *__Answer: __The red X means failure spot for RNA extraction. We added the description in Figure 1B.

* 17.Please spell out the abbreviations in all figure legends. *__Answer: __We added the abbreviations in the legends of all figures.

* 18.Please explain what is meant by the color of the lines and the size of the circles in Figure 4D. *__Answer: __The network analysis was performed by Metascape (https://metascape.org/gp/index.html#/main/step1) [15]. The node size is proportional to the number of genes belonging to the term, and the node color represents the identity of the cluster. However, as described in the answer to reviewer’s comment 9, we reperformed enrichment analysis with significant DEGs. As a result, only CD44/MYC cluster had a lot of enrichment terms.

Therefore, network visualizations were removed from the revised manuscript.

* 19.Please mention an explanation for the color of the spots in Figure 5D and 5F to the legend. *__Answer: __The color showed the spots categorized into the selected group.

In the revised manuscript, previous Figure 5 was incorporated as Supplementary Figure S7. We added the description in Supplementary Figure S7 and S8 with the legends.

* 20.Is "S51" in Line 148 a typo for "S5A"? *Answer: Thank you. We revised “S5A”.

* 21.Please mention an explanation for the bars in Figure 6D and 6F to the legend. *__Answer: __The bars showed relative CCI scores. As described below, we removed the results of CCI analysis with cluster group (previous Figure 6) in the revised manuscript.

* 22.Please mention an explanation for the colors in Figure 7E to the legend. *__Answer: __The color showed patients’ group based on expression levels of gene signatures. We added the description in the Legend of Figure 6.

*

*

Reviewer #2 (Significance (Required)): * The approach in Figure 5 is interesting, but the rest of the results do not take full advantage of the technology developed by the authors. The structure of the manuscript should be re-examined and new perspectives added. I look forward to the future of the authors' research.

*

Reviewer #3 (Evidence, reproducibility and clarity (Required)): Microtissue transcriptome analysis of triple-negative breast cancer cell line MDA-MB-231 xenograft model using automated tissue microdissection punching techonology revealed that the existence of three cell-type clusters in the primary tumor and axillary lymph node metastasis. The CD44/MYC-high cluster showed aggressive proliferation with MYC expression, the HMGA1-high cluster exhibited HIF1A activation and upregulation of ribosomal processes. The cell-cell-interaction analysis revealed the interaction dynamics generated by the combination of cancer cells and stromal cells in primary tumors and metastases. The gene signature of the HMGA1-high cancer stem cell-like cluster has the potential to serve as a novel biomarker for diagnosis. The key conclusions are convincing. The data and methods are presented in a reproducible way. The experiments are adequately replicated and statistical analysis is adequate. Prior studies are appropriately referenced. The text and figures are clear and accurate. __Answer: __We would thank the valuable comments. As the reviewer mentioned, our findings showed that the existence of two cancer stem cell-like populations has the potential to make tumors more drug-resilient. Our results suggested that the gene signatures from the resolution of microspot (10-30 cells) could have the potential to be prognosis markers. This punching microdissection system enables to extract only the parts of a section that are necessary for diagnosis of cancer and to analyze at low-cost. It could be applied to diagnostics instead of the laser-capture microdissection methods.

In this revision, we focused on the existence of two cancer stem cell-like populations in TNBC xenografts and patients. Following the other reviewer’s comments, we performed the extraction of DEGs with significant threshold; therefore, we revised the results of enrichment analysis but it did not influence our main findings.

To validate the existence of two types of cancer stem-like cells in TNBC tumors, we performed the additional analyses (reanalysis of public scRNA-seq datasets and immuno-staining of MDA-MB-231 primary tumor). These results verified two cancer stem cell-like populations (HMGA1-high, CD44-high) in MDA-MB-231 xenograft and TNBC patients. We believe that our findings are solid results because the findings were also validated by other methods.

Again, we would thank kind reviewing our manuscript.

Reviewer #3 (Significance (Required)): * In the past several studies showed the heterogeneity of cell-cell interactions between cancer cells and stromal cells in situ (Andersson et al, 2021; Wu et al, 2021) and tumor microheterogeneity (Jiang et al, 2016; Liu et al, 2016; Zhang et al, 2020). Spatial transcriptomics methods are important to reveal microheterogeneity of cancer. As a physician working in gynecology and obstetrics in my opinion the results of the study and spatial transcriptomic methods could be relevant to detect new biomarkers for diagnosis and prognosis of breast cancer in future and to find novel therapeutic targets to overcome drug resistance and facilitate curative treatment of breast cancer.

*

References in response letter

1. Srivatsan SR, Regier MC, Barkan E, Franks JM, Packer JS, Grosjean P, et al. Embryo-scale, single-cell spatial transcriptomics. Science. 2021;373(6550):111-7. Epub 2021/07/03. doi: 10.1126/science.abb9536. PubMed PMID: 34210887.
2. Moravec JC, Lanfear R, Spector DL, Diermeier SD, Gavryushkin A. Cancer phylogenetics using single-cell RNA-seq data. bioRxiv. 2021:2021.01.07.425804. doi: 10.1101/2021.01.07.425804.
3. Liu H, Patel MR, Prescher JA, Patsialou A, Qian D, Lin J, et al. Cancer stem cells from human breast tumors are involved in spontaneous metastases in orthotopic mouse models. Proc Natl Acad Sci U S A. 2010;107(42):18115-20. Epub 2010/10/06. doi: 10.1073/pnas.1006732107. PubMed PMID: 20921380; PubMed Central PMCID: PMC2964232.
4. Pegoraro S, Ros G, Piazza S, Sommaggio R, Ciani Y, Rosato A, et al. HMGA1 promotes metastatic processes in basal-like breast cancer regulating EMT and stemness. Oncotarget. 2013;4(8):1293-308. Epub 2013/08/16. doi: 10.18632/oncotarget.1136. PubMed PMID: 23945276; PubMed Central PMCID: PMC3787158.
5. Weiss F, Lauffenburger D, Friedl P. Towards targeting of shared mechanisms of cancer metastasis and therapy resistance. Nat Rev Cancer. 2022. Epub 2022/01/12. doi: 10.1038/s41568-021-00427-0. PubMed PMID: 35013601.
6. Oskarsson T, Batlle E, Massagué J. Metastatic Stem Cells: Sources, Niches, and Vital Pathways. Cell Stem Cell. 2014;14(3):306-21. doi: https://doi.org/10.1016/j.stem.2014.02.002.
7. Turdo A, Veschi V, Gaggianesi M, Chinnici A, Bianca P, Todaro M, et al. Meeting the Challenge of Targeting Cancer Stem Cells. Front Cell Dev Biol. 2019;7:16. Epub 2019/03/06. doi: 10.3389/fcell.2019.00016. PubMed PMID: 30834247; PubMed Central PMCID: PMC6387961.
8. Armingol E, Officer A, Harismendy O, Lewis NE. Deciphering cell-cell interactions and communication from gene expression. Nat Rev Genet. 2021;22(2):71-88. Epub 2020/11/11. doi: 10.1038/s41576-020-00292-x. PubMed PMID: 33168968; PubMed Central PMCID: PMC7649713.
9. Kumar MP, Du J, Lagoudas G, Jiao Y, Sawyer A, Drummond DC, et al. Analysis of Single-Cell RNA-Seq Identifies Cell-Cell Communication Associated with Tumor Characteristics. Cell Rep. 2018;25(6):1458-68.e4. Epub 2018/11/08. doi: 10.1016/j.celrep.2018.10.047. PubMed PMID: 30404002; PubMed Central PMCID: PMCPMC7009724.
10. Watanabe N, Nakayama J, Fujita Y, Mori Y, Kadota T, Shimomura I, et al. Single-cell Transcriptome Analysis Reveals an Anomalous Epithelial Variation and Ectopic Inflammatory Response in Chronic Obstructive Pulmonary Disease. medRxiv. 2020:2020.12.03.20242412. doi: 10.1101/2020.12.03.20242412.
11. Nakayama J, Yamamoto Y. Single-cell meta-analysis of cigarette smoking lung atlas. bioRxiv. 2021:2021.12.09.472029. doi: 10.1101/2021.12.09.472029.
12. Stuart T, Butler A, Hoffman P, Hafemeister C, Papalexi E, Mauck WM, 3rd, et al. Comprehensive Integration of Single-Cell Data. Cell. 2019;177(7):1888-902.e21. Epub 2019/06/11. doi: 10.1016/j.cell.2019.05.031. PubMed PMID: 31178118; PubMed Central PMCID: PMC6687398.
13. Finak G, McDavid A, Yajima M, Deng J, Gersuk V, Shalek AK, et al. MAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data. Genome Biol. 2015;16:278. Epub 2015/12/15. doi: 10.1186/s13059-015-0844-5. PubMed PMID: 26653891; PubMed Central PMCID: PMCPMC4676162.
14. Cheng S, Li Z, Gao R, Xing B, Gao Y, Yang Y, et al. A pan-cancer single-cell transcriptional atlas of tumor infiltrating myeloid cells. Cell. 2021;184(3):792-809.e23. Epub 2021/02/06. doi: 10.1016/j.cell.2021.01.010. PubMed PMID: 33545035.
15. Zhou Y, Zhou B, Pache L, Chang M, Khodabakhshi AH, Tanaseichuk O, et al. Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat Commun. 2019;10(1):1523. Epub 2019/04/05. doi: 10.1038/s41467-019-09234-6. PubMed PMID: 30944313; PubMed Central PMCID: PMC6447622.

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

Microtissue transcriptome analysis of triple-negative breast cancer cell line MDA-MB-231 xenograft model using automated tissue microdissection punching techonology revealed that the existence of three cell-type clusters in the primary tumor and axillary lymph node metastasis. The CD44/MYC-high cluster showed aggressive proliferation with MYC expression, the HMGA1-high cluster exhibited HIF1A activation and upregulation of ribosomal processes. The cell-cell-interaction analysis revealed the interaction dynamics generated by the combination of cancer cells and stromal cells in primary tumors and metastases. The gene signature of the HMGA1-high cancer stem cell-like cluster has the potential to serve as a novel biomarker for diagnosis.

The key conclusions are convincing. The data and methods are presented in a reproducible way. The experiments are adequately replicated and statistical analysis is adequate.

Prior studies are appropriately referenced. The text and figures are clear and accurate.

Significance

In the past several studies showed the heterogeneity of cell-cell interactions between cancer cells and stromal cells in situ (Andersson et al, 2021; Wu et al, 2021) and tumor microheterogeneity (Jiang et al, 2016; Liu et al, 2016; Zhang et al, 2020). Spatial transcriptomics methods are important to reveal microheterogeneity of cancer. As a physician working in gynecology and obstetrics in my opinion the results of the study and spatial transcriptomic methods could be relevant to detect new biomarkers for diagnosis and prognosis of breast cancer in future and to find novel therapeutic targets to overcome drug resistance and facilitate curative treatment of breast cancer.

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 manuscript uses spatial transcriptomics to perform single cell-like expression analysis between a breast cancer cell line and tumor microenvironment in mice xenografted with these cells. Unfortunately, from the title, abstract, and introduction, it is difficult to understand exactly what the authors are focusing and discussing. It is also unclear the advantage of their technique for evaluating the populations observed within this manuscript. Furthermore, there is very little explanation of the results, and it does not appear to be a scientific logical structure. Hence, this manuscript is not suitable for acceptance in the journal. In order to improve the scientific quality of this study, the following concerns are presented.

Major concerns:

1.Is cell-cell interaction (CCI) analysis novel method? If so, please specify detail in the manuscript. If the basic concept and the principle of CCI analysis have not been published, please mention in the discussion section as a limitation that a manuscript on CCI analysis is under submission to the preprint. In addition, please revise the abstract and related text.

2.The reviewer thinks that spatial transcriptomics plays an important role in your manuscript. Please describe the technique in the introduction.

3.The classification by expression profile (HMGA1, CD44/MYC and marker-low) lacks an explanation. Authors should mention in detail how these populations were extracted from breast cancer cell lines.

4.The description of the results is back and forth and confusing. Please reconsider the flow of the analysis.

5.How did you evaluate the outsides of the samples with very different spot positions in Figure 3A? Please mention your evaluation method in a scientific manner. In particular, authors should clearly indicate the outer evaluation for the metastatic case.

6.The spots in primary tumor have few counts derived from mouse stromal/immune cells, as shown in Figure S1A. Nevertheless, Figure 3C shows that mouse stromal/immune cells are evaluated in the same way in primary and metastatic sites. The reviewer thinks that the regions identified as Tcell-like in the metastatic site, where there are many mouse-derived counts, and in the primary, where there are few mouse-derived counts, do not have the same characteristics. If many mouse-derived counts were detected in a spot using the spatial transcriptomics, then there must be many mouse-derived cells in the spot. Please discuss how this expression is evaluated on this technique, which is not a single cell analysis.

7.Please explain how the gene symbols listed in Figure 4A were selected. Also, please indicate the characteristics of the gene groups that are not listed.

8.Please describe the details of the division and cycle index in lines 141-142.

9.In Line 148-151, the expression and prognosis of TMSB10, CTSD, and LGALS1 is mentioned based on the previous reports. Aren't these findings the result of bulk? Is the HMGA1 cluster that the authors found involved in the prognosis of mice? Please clarify, as it is unclear what you want to discuss.

10.Please provide details of all statistical tests used in this manuscript and describe significance levels used in the p-values and FDR.

11.Please mention CCI score (line 198).

12.Lines 204-206 and Figure 6G show specific interaction of ITGB1 and CST3, but it is unclear why only these molecules were extracted. What about the other molecules? At least ITGB1 is not scored in mix5.

13.HMGA1 signature appears in Line 214, please explain in detail.

14.Authors should discuss how the previously reported bulk expression data used in Figure 7E can be linked to the single-cell-like analysis in this study.

Minor concerns:

15.Please describe how the normalized centrality was calculated in UMAP algorithm and explain what this means in the results.

16.Please mention an explanation for the red X in Figure 1B to the legend.

17.Please spell out the abbreviations in all figure legends.

18.Please explain what is meant by the color of the lines and the size of the circles in Figure 4D.

19.Please mention an explanation for the color of the spots in Figure 5D and 5F to the legend.

20.Is "S51" in Line 148 a typo for "S5A"?

21.Please mention an explanation for the bars in Figure 6D and 6F to the legend.

22.Please mention an explanation for the colors in Figure 7E to the legend.

Significance

The approach in Figure 5 is interesting, but the rest of the results do not take full advantage of the technology developed by the authors. The structure of the manuscript should be re-examined and new perspectives added. I look forward to the future of the authors' research.

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:

Nakayama and colleagues use their previously developed automated tissue microdissection punching platform to perform spatial transcriptomics on a breast cancer xenograft model. Using transcriptomics on multiple clumps of 10-30 cells from different regions in a tumor and a lymph node metastasis they identified different cell-type clusters. Two of these clusters expressed different cancer stem cell markers. This led the authors to suggest that two distinct cancer stem cell(-like) populations may exist within one (breast) tumor, which could potentially make tumors more drug-resilient.

Major comments:

While the quality of the presented sequencing data is good and the manuscript is mostly written in a clear and accessible style, there are some concerns that limit the impact of this story. Most importantly, the manuscript in its present form does not convince me that the MDA-MB-231 xenografts indeed contain two distinct populations of cancer stem(-like) cells.

1.The data obtained are not single cell data, which makes it difficult -if not impossible- to draw conclusions about presence of cancer stem cells. Each data point is the average of 10-30 cells, and the interpretation of the data is severely limited by this. How can the quantification of expression of CD44/MYC/HMGA1 in clumps of 10-30 cells teach us something about the stemness of tumor cells?

2.Furthermore, the authors should better explain their data analysis strategy with identification of gene expression profiles. It is unclear how they found CD44, MYC, and HMGA1 other than by cherry-picking from the list of cluster markers.

3.Following up on the above point: I looked in the supplementary tables, but couldn't find MYC. How did the authors conclude that MYC is involved in cluster 1? In fact, when I ran a quick analysis in EnrichR, I saw that putative MYC target genes were strongly enriched among the markers in the HMGA1 cluster, but not the CD44/MYC. That's opposite to what I would expect.

4.All data were produced from 1 primary tumor and 1 metastasis. Thus, reproducibility and robustness of the methodology cannot be evaluated. The interpretation of the data could be strengthened when xenografts from at least 3 different mice are shown.

5.The only methodology is single cell RNA-sequencing. Immuno-staining on relevant markers such as CD44, MYC, HMGA1 plus human epithelium and cell cycle markers would provide strong additional support for the claims made by the authors, because it's a complementary technique and it allows quantification at single cell resolution.

6.Line 173-175. The marker-low cluster look to me simply like spots containing a relatively high amount of dead/dying (tumor) cells. The identity/state of cells in the marker-low cluster should be characterized and discussed more extensively.

7.Figure 5 and accompanying text in line 182-194; the authors try to infer cell-to-cell interactions using a previously published tool. However, any biological interpretation is lacking. What can be concluded from this analysis?

8.Figure 6. Can the authors please explain more clearly what they mean by "PT" and "Mix" groups? I had a very hard time to understand what the data in figure mean. Again, an overall interpretation at the end (line 211) is lacking.

9.Figure 7. I like the effort to align the results with public scRNA-seq data. But although the expression of the cluster-signatures is heterogeneous, there is no evidence for distinct (CSC-like) cell populations. Why don't these HMGA1 vs CD44 signature cells cluster away from each other in the UMAPs? Perhaps the patient-to-patient heterogeneity overwhelms differences within tumors, but in that case the authors could re-run their analysis for each patient separately, to make 6 patient-specific UMAPs. In its present form, this analysis does not convince me that two distinct CSC(-like) populations within one TNBC exist.

Minor comments:

10.In the Supplemental table 2 noticed that many of the marker genes have adjusted P values well above 0.05 (and even above 0.1). That makes the statistical analysis rather weak. This could especially be problematic since the authors entirely base their main claims on this marker analysis, and I recommend that the authors use more stringent P-value cut-offs in the cluster analysis.

11.Line 129/130. If I look at figure 3A, I don't see this tendency that the authors describe. Can the authors provide statistical support or visual aid to make their claim more apparent to the reader?

12.Line 217; shouldn't this be 6 patients? I see six clusters and in the original paper six patients are mentioned.

Significance

Conceptual/biological impact: Showing the existence of distinct populations of CSCs within one (breast-)tumor potentially has a high impact on the field of fundamental and translational cancer research. As the authors state, it could be one key reason underlying drug resistance. However, the technology used by the authors does in my view not allow to make such a claim. First and foremost because the technology does not allow analysis at single cell resolution.

Technical impact: The platform used by the authors can be of interest for some applications, but they already published this in Scientic Reports a few years ago. I'm afraid that with the rapid recent developments in the field of spatial single cell transcriptomics (See for example Srivatsan et al Science 2021; 373: 111-117), the technical impact on the field is relatively low.

Audience: Researchers in the field of cancer biology with an interest to perform low-cost molecular analysis at low-resolution spatial-resolved tissue specimens (transcriptomics, but perhaps expanded with bisulfite sequencing, or ATAC sequencing) could be interested in the technology presented in this manuscript.

My expertise: single cell transcriptomics, (cancer) cell cycle, cancer drug resistance, cell plasticity, mouse models.

Referee Cross-commenting

I have read the comments and align mostly with reviewer #2. The authors need to improve this manuscript a lot before it's suitable for publication in any of the Review Commons journals.

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

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

Reply to the Reviewers

I thank the Referees for their...

Referee #1

1. The authors should provide more information when...

Responses + The typical domed appearance of a hydrocephalus-harboring skull is apparent as early as P4, as shown in a new side-by-side comparison of pups at that age (Fig. 1A). + Though this is not stated in the MS 2. Figure 6: Why has only...

Response: We expanded the comparison

Minor comments:

1. The text contains several...

Response: We added...

Referee #2

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

all OK

Significance

This is a valuable paper that make use of the rapid mitotic cycles of the Drosophila syncytial embryo to study the recruitment of proteins in mitotit centrosome maturation. The synchrony of these cycles make this an excellent experimental system in which to follow the relative timing of recruitment of individual molecules to the centrosome and, while the system may have idiosyncrasies that facilitate rapid cycling, it provides valuable information. This is a significant data set that shows the pulsatile recruitment of Spd2 and Polo kinase peaking in mid S-phase in contrast to the continuous recruitment of Cnn.

The authors carry out some interesting modelling to account for the pulsatile activity of Polo through recruitment to the centriole. As they have previously shown Polo recruitment to be dependent upon S-S/t motifs in An1 and Spd, the authors examine the effects of multiple mutations at these potential recruitment sites. Interestingly they show that mutation of 34 such sites in Ana1 has little effect on recruitment of Polo to old-mother centrioles but perturbs recruitment onto ne mothers. Expression of the multiply mutated Spd-2, on the other hand, perturbs the Polo pulse on both old- and new- mothers. Together this would be in line with their previous suggested role for Ana1 in initially recruiting Polo to centrioles and Spd2 having a role in expanding the PCM.

The modelling carried out by the authors is simple but effective. As with almost any cell cycle model, the models have their short-comings and the authors are largely aware of these. I thought it would be worthwhile to have some more discussion of what activates Polo kinase. It could be partially activated by the Polo-box binding to its receptor site but do other kinases carry out its T-loop phosphorylation? There are plenty of mitotic kinases around and so this could be discussed in greater detail. Moreover, although the pulsatile association of Polo with the centrosome does not have to correspond to pulsatile activity, this is likely. In which case, further discussion of the roles of opposing phosphatases would be in order.

All in all, however, this is a useful paper that comes up with a thorough description of the timing of events of centrosome maturation in Drosophila embryos.

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

Review of 'Mother centrioles generate a local pulse of Polo/PLK1 activity to initiate mitotic centrosome assembly' from Wong et al.

In this paper, Wong et al address the mechanisms of centrosome assembly in flies. They start with the interesting observation that Polo localized at centrosomes oscillates before cells enter mitosis, while Cnn (and with it centrosome maturation) either increases or reaches a plateau. The phenomenon is local, since Polo levels at in the cell are high during mitosis. They propose that the oscillation is driven by a negative feedback loop whereby Polo inhibits its own binding to the centrosome, Ana1 being the most likely relevant receptor. Finally, they discuss the possible meaning of this oscillatory behavior, in the light of the rapidity of the early embryonic cell cycles.

Major comments

1- One can imagine different reasons for the fact that the model displays different dynamics for Cnn and Spd-2/Polo. For example, a major difference may be due to the different dissociation rates of the clusters Cstar and Shat. These are governed by different laws and different parameters (kdis vs kidsCstar1/n). If I understand, both parameters and dependency on Cstar^2 are assumptions. Hence, it would be important to pinpoint which component of the model is more directly responsible for the observed behavior. The analysis should not be limited to the dissociation, but should be extended to the whole model. To this aim, one could test the robustness of the model's parameters. The results of this analysis will also be a prediction of the model.

2- The presence of a positive-feedback loop involving Cnn could offer an alternative and more robust explanation for the slower dynamics of Cnn. Such a loop between Cnn and Spd-2 was proposed by the authors (Conduit, eLife, 2014). I think some comment on this point would be interesting (eg, could the Cnn/Spd-2 loop proposed earlier work in this context? If not, why? If yes, should not this option be explored?).

3- The prediction presented in Figure 6 is very relevant. I wonder how robust this behavior is to changes in parameters values.

4- Additional testing of the model would be important to confirm that the negative feedback loop is actually in place, although I understand experiments may be difficult to be performed. Possible examples: constantly high levels of Polo are expected to decrease its centrosomal localization, is that correct and, if so, testable? Is it possible to delay one cycle, and then observe the decay in Cnn values? This latter experiment, for example, could help to distinguish positive feedback vs slow decay rates. If the experiments are not possible, it may be worth anyway to present some predictions worth testing.

5- The difference between Models 2 and 3 is not clear to me. In mathematical terms, they seem to be basically the same thing: reaction (50)=(33), (51)~(34) given (40) and (52)~(35) again given (40). This is precisely since the model comes with the assumption of a well-stirred system, and thus adding P in solution is not so different from assuming P=Rphat (40). I would have imagined that also Model 2 accounts for the fact that in Spd-2-S16T and Ana1-S347T Polo is recruited slower and for a longer period. Is it not true? If so, is model 3 really needed? More in general, assuming a role for an increase of local concentration of P* is quite a jump, especially given the small distances involved, and the fast diffusion occurring within cells.

Minor points

1-Could the authors use the FRAP data to estimate the different kdis? If so, a comparison with the 20-fold difference used in the model would be useful.

2- p. 6, The authors should state clearly for the worm-uneducated like me whether the fusions were done with the endogenous proteins or not.

3- p.7 Figure 1B, in the text it is referred to display 'levels of peaks' and in the figure and legend we find 'growth period'. Not clear how the two refer to the same quantity.

4- Spd2-mCherry is present in both Figure 1C and D, but with very different amplitudes. Why is that the case?

5- The fact that Polo peaks in mitosis is a key observation. Unfortunately, this is often reported as a personal communication. The authors never tried to produce this piece of data?

6- p.11 It is explained that NM and OM differ for their initial values because the OM starts with some PCM from the previous cycle. However in Figure 3A, for example, the values of Polo at the end of the cycle are identical in the two. Is not this in contrast with the explenation?

Still p11, there is reference to Figure 3C,D, but Figure 3D does not exist, I guess it should be 3A,C.

7- In the formulation of the model (page numbers in Suppl Mat are unfortunately missing..), one citation for the total amount of Polo being large is needed.

8- I do not understand this point: scaled c output is 1, and the initial condition for c=1 also?

9- It has been shown in different systems (from yeast -- haase winey reed, NCB, 2001-- to worms -- McCLeland O-Farrell CB 2008) that centrosome duplication can occur independently from the cell cycle oscillator. I was wondering whether the proposed negative feedback loop may play a role in this phenomenon. This is only a curiosity, which does not need to be addressed.

Significance

The new observation and hypotheses presented in the paper provide a sizeable advance. The presence of an oscillation in Polo, uncoupled from cellular levels, is new, and the model proposes a testable hypothesis to explain it. Some additional experiments to verify the model would strengthen the manuscript.

The work is probably more appropriate for experts in the centrosome field. My primary expertise for this review was in mathematical models.

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:

Embryogenesis is characterized by rapid cell divisions without gap phases. How these cells achieve successive rounds of chromosome segregation in dozens of minutes without failure is of great interest to cell and developmental biologists. A key aspect of rapid divisions is the oscillatory nature of centrosome assembly, which aids in building the mitotic spindle during mitosis, and centrosome disassembly during mitotic exit. Polo kinase activation and localization to the centriole is essential for centrosome dynamics, but its molecular targets, timescales of activation and deactivation, and overall mechanism of action is still not fully determined.

This paper aims to build a mathematical model to tease out the mechanism of Polo recruitment to centrioles and transformation of centrosome scaffold proteins (Spd2/Cep192 and Cnn/CDK5RAP2) from inactive forms to functional, multimeric platforms. The authors posit that several features are critical to describe the dynamics of Polo, Spd-2, and Cnn: 1) a negative feedback loop that releases Polo from a centriole receptor, 2) a kinetic relay that allows Spd-2 to assemble, followed by Cnn, and 3) a disassembly mechanism driven by de-phosphorylation. They validate their model in several ways, most notably by introducing an Ana1 mutant that inhibits Polo binding to centrioles: their model predicts a delay in Polo accumulation which bears out in vivo.

The cell biology experiments of this paper are of high quality and well quantified, and I have no concerns there. However, the mathematical model elevates this study to the next level, and thus deserves greater scrutiny. I'm not concerned that the model doesn't get everything right, or that all of the parameters are correct. This is new territory. I think the value of models is their power to predict, rather than their power to explain existing data. The authors are giving the field a great hypothesis generator which we can use to plan experiments for the next 5 years. Then the model will be updated to be more accurate. Thus, this work represents a significant achievement.

Still, some key validations regarding phosphorylation rates are missing that could be easily tested. Furthermore, the study would be strengthened by greater understanding of the PCM disassembly. mechanism. Addressing these two points will improve my confidence in the mathematical model.

Major Comments.

1. This study builds a model that relies heavily on rates of phosphorylation and de-phosphorylation. Further, de-phosphorylation is assumed to be the key disassembly mechanism, but this has not been rigorously studied in fly embryos. Thus, two critical aspects of the model remain unverified.

Surely, the authors could test how changing phosphorylation rate (kcat S and kcat C) and de-phosphorylation rate (kdis) affects the recruitment and departure of Spd-2, Polo, and Cnn in vivo. This could be achieved by 1) titrating an inhibitor of Polo (e.g., BI-2536) or introducing a mutation in the T-loop of Polo (the equivalent of T210D or T210V in flies; T210D should raise kcat, while T210V should lower kcat; https://doi.org/10.1021/bi602474j), and 2) inhibiting a phosphatase such as PP2A, which is the presumed antagonist of Polo according to several C. elegans studies.

If their model adequately predicts the outcome of these two experiments (changing phosphorylation and dephosphorylation rates), I will be more convinced.

1. The models focus on Polo and Spd-2 pulses during mitosis, but ignore the disassembly phase of Cnn. Do Cnn levels drop during mitotic exit? Can this drop in Cnn be described by any of the authors' proposed models?
2. These models are described as the "simplest possible" yet have many unknown parameters. For example, Model 1 has 12 parameters, none of which have been determined experimentally. How did the authors land on these values? Is it possible that one could alter any combination of these parameters and achieve a similar outcome? Or, if the Kcat of Polo is changed two-fold, does the whole model fall apart (see above)?

Experimentally determining these parameters would greatly strengthen this paper, but I think that would require gargantuan effort that is beyond the scope of the current work. Instead, it is therefore critical to test how robust the model is by probing the parameter space. For example, could the authors show us what the model predicts (e.g., as in Figure 2C) when each parameter is changed by 2-fold? Presumably the authors have already done this, but I would like to see the outcomes.

Minor Comments.

-Figure 1. The authors should include representative images of centrosomes for the plots in panels A,C and D. The x-axes could have more informative labels (e.g., time relative to NEB). - Figure 1A. Much of Figure 1 has already been performed in C. Elegans, yet this fact is not mentioned until the discussion. For example, the pulsed nature of Polo and SPD-2 appearance and disappearance has been reported in C. elegans in Mittasch et al. 2020 and Magescas et al., 2019. These findings, and their implications for evolutionary conservation, should be mentioned in the main text (e.g. page 6 or 7). - Figure 2. It's hard to envision how a scaffold can both flux outward and be structurally strong. The mere fact that there is outward movement of scaffold chunks implies breaking of bonds, which indicates overall structural weakness. Are the authors talking about strength of the entire PCM, or just strength of the chunks? It would be great if the authors could clarify this. -Figure 2. One would think that scaffold flux and strength are anti-correlated. Perhaps this is the case? As far as I'm aware, previous studies of Cnn flux were performed primarily in S-phase, when there is presumably less need for PCM strength. What about during mitosis during chromosome segregation? Does the PCM become stronger during mitosis? Does Cnn flux decrease during mitosis? - Figure 2B. I would prefer a legend in the actual figure indicating what the different symbols mean. I found it difficult glancing back and forth between the text and the figure. - "We also allow the rate of 𝐶∗ disassembly to increase as the size of the 𝐶∗ scaffold increases, which appears to be the case in these embryos (Conduit et al, 2010)."

I can't find any analysis of PCM disassembly in this study. What are the authors referring to as "disassembly"? Do they mean departure of Cnn from the PCM in S-phase? Or, disassembly of the whole PCM during mitotic exit?

-"If the centriole and PCM receptors (Ana1 and Spd-2, respectively) recruit less Polo, the centriole receptor (Ana1) will be inactivated more slowly."

Is Ana1 a known substrate of Polo? This seems highly speculative. The authors should note that deactivation of Ana1 could be through various other mechanisms. Furthermore, Polo could be locally degraded as shown in human cells doi: 10.1083/jcb.200309035.

-"We note that our mathematical models are purposefully minimal to reduce the number of parameters and test possible mechanisms rather than to mimic experimental data." I appreciate this statement.

Significance

This paper aims to build a mathematical model to understand the cyclic nature of centrosome assembly and disassembly in fruit flies. Due to the conserved nature of the components (proteins in the system, such as Polo Kinase, Spd-2, and Cnn/CDK5RAP2), this model could likely be extended to a broad swath of eukaryotes. This approach is quite unique in the centrosome field, as only one other study (Zwicker et al., PNAS 2014) has tried seriously to model the growth kinetics of PCM, the outermost part of a centrosome. The field has been dominated by genetics and cell biology approaches, so implementing a mathematical model will advance the field and generate hypotheses, even if the model is not yet fully fleshed out. This paper represents a significant advance.

This study will be of broad interest to the centrosome field.

Expertise: centrosome biogenesis, mitosis, biophysics

Note: I am not sufficiently qualified to evaluate the mathematics underlying the model.

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

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

Summary: The Non-Photochemical Quenching (NPQ) protects photosystems from energy overloading by excess light exposure. The NPQ consists of multiple factors which function in different time scales and energy levels. One of the factors, qH, has been proposed based on chlorophyll fluorescence lifetime observation and the plastid lipocalin has been identified as the important player to regulate qH. It remains to link the qH phenotype and molecular mechanisms. The authors purify photosynthetic protein complexes from the qH mutants and tried to build a biophysical model to link qH phenomena and protein science based on chlorophyll fluorescence lifetime observation.

Response: Thank you for your constructive comments which we have addressed and complete the manuscript nicely.

Major comments: There are two major issues. One issue is, even many kinetics are presented, but the relationship between these values and qH phenotypes is not clearly stated or connected. One idea is to build the mathematical model(s) to explain these kinetics. The other issue is, lipid composition is not considered. Indeed, this phenomenon is observed or emphasized in low temperatures. Generally thinking, lipid composition bound to photosynthetic complexes would be disturbed or modified its conformation.

Response: ____We have not attempted to build a descriptive model of how the different molecular players in qH operate in the membrane to yield the observed fluorescence kinetics as this is beyond the scope of our study. However, we agree that lipid composition should be considered and we have now added two additional authors, text in the method and result sections and new Fig. 6 and Fig. S9 examining lipid composition of thylakoid extract, LHCII trimer and LHCII/Lhcb monomer fractions. No significant differences can be observed between the qH ON and OFF states in the main chloroplastic lipids.

Minor comments: Some datasets are less biological replicates or not clearly stated about the biological replicate number (Figure 2, Figure 4, Figure 5, Figure S2, Figure S5, and Figure S8). Normally, at least three independent biological replicates are required. Technical replicates are not acceptable.

Response: ____If by biological replicates, you mean three independent plant individuals, we agree that this would be the bare minimum required, and we apologize for the confusion. The definition of biological replicate (also referred to as biological experiment) in our study is each one represents a separate batch of several plant individuals pooled (n = 2 to 8) grown at independent times. Then within each biological experiment, we perform technical replicates (i.e. independent measurements of different aliquots from the same sample) which we believe are acceptable and necessary but we agree not sufficient. For most data, we have at least 2 biological experiments, and up to 3, for assessing the quenched nature of LHCII trimer and not LHCII/Lhcb monomer (Fig. 3). We have rephrased the text so this aspect is clearer and also provide more detail below about the aforementioned figures.

Fig. 2: TCSPC on thylakoids, n=3 technical replicates from 2 independent biological experiments; Two separate thylakoid preparations were made from independently grown plants (leaves from n > or = 5 plants were pooled each time). Fig. 4: CN-PAGE, n=3 technical replicates from 2 independent biological experiments (leaves from n > or = 3 plants were pooled each time). Fig. 5: TCSPC on isolated complexes, n=3 technical replicates from 2 independent biological experiments; Two separate thylakoid preparations were made from independently grown plants (leaves from n = 8 plants were pooled each time). Fig. S2: step solubilization, n=2 technical replicates; here 1 biological experiment was used from n = 8 plants. Fig. S5 contains the biological replicates 1 (n=2 plants) and 2 (n=8 plants) of the representative experiment shown in Fig3, biological replicate 3 (n=8 plants). Fig. S8: HPLC on isolated complexes from 2 independent biological experiments (leaves from n = 8 plants were pooled each time).

In Figure 3 and Figure S3, extend the length of the major tick for each axis. It is hard to distinguish between major tick and minor tick.

Response: Ok, done.

In Figure 3, mark the measured peak wavelength value on the top for readers.

Response: ____Ok, done, added in the legend “with maxima at 679 nm for all samples”.

In Figure 4, Why do not you present chlorophyll kinetics? I suspect it is possible to acquire if you used SpeedZen.

Response: In Figure 4, we present a measurement of fluorescence emission from separated pigment-protein complexes by CN-PAGE, there are no light-induced changes to be measured here hence we do not present chlorophyll fluorescence kinetics.

In Figure 6, decrease the thickness of the border for the bar graph or marker. Markers on the top of the bar graph are not visible.

Response: Ok, done.

Figures S3 and S6, provide the elution volume of protein standard in the chromatograph.

Response: We don’t make any statement regarding the molecular weights of the protein complexes from the chromatograms, so elution volumes of protein standards are not required. Composition of the different peaks were validated by Iwai et al. 2015 (Nat Plants 1: 14008) and further verified here (Fig. S6, S10).

Figures S11 and S12, describe the number of biological replicates.

Response: ____Ok, done (now Fig. S12 and S13).

Reviewer #1 (Significance (Required)): The topic is important for plant physiology especially photosynthesis regulation and biophysical characterization is straightforward to interpret molecular machinery. Other studies are only for chlorophyll observation for the whole plant body, but most importantly, this study is the challenging work on qH characterization with a biochemical approach.

Response: Thank you for your appreciation of our work!

---

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

Authors observed qH in isolated LHCII trimers with Chl fluorescence changes (shorter), and concluded that no single major Lhcb isomers is necessary for qH.

Response: Thank you for your constructive comments which we have now addressed and make the manuscript clearer.

Major concern is: LHCII trimers are divided into S, M, L trimers with different compositions. Authors are requested to interpret their results in terms of L-, M-, S-trimers.

Response: Our solubilization conditions and isolation method don’t allow to distinguish between loosely (L), moderately (M) or strongly (S)-bound trimers, the LHCII trimer fraction is a pool of these trimers. We have now specified this aspect in the discussion and cannot interpret our results further than narrowing down the LHCII trimer as a quenching site. In future work, we will attempt their separation although getting entirely pure fractions of each is technically challenging.

Minor comments are: Authors describes qI as reversible NPQ, but qI with D1 damage is not reversible.

Response: ____D1 can be repaired thereby relaxing qI, see recent article from Nawrocki et al. Sci Adv 2021. We have clarified this point in the introduction.

In page 3 - 2nd paragraph, Authors define components of NPQ one by one, but the definition or recovery kinetics for qH is skipped, And authors suddenly start explaining molecular players of qH without changing paragraph.

Response: We have now clarified that the relaxation kinetics for sustained NPQ including qH are slow (hours to days) and changed paragraph to introduce the molecular players known to be regulating qH.

In Fig. S6, authors tried to confirm the trimer and monomer fractions they used by using Lhcb2 and Lhcb4 antibodies, respectively. But, the distribution of Lhcb2 only in Trimer fraction in WT, which is different from the distribution in other mutants. Contamination of Lhcb4 in Trimer fraction is also of concern. Authors may use BN-PAGE or Ultracentrifugal separation, rather than gel filtration.

Response: Regarding the different distribution of Lhcb2 between WT and mutants, we have now better labeled Fig. S6B so it is clear that WT is non-treated (non-stress condition) and the mutants underwent a cold and high light-treatment (stress condition). This difference may thus be explained by the trimers stability/propensity to be solubilized by the detergent varying between non-stress and stress conditions. It is not a concern as we’re not comparing mutants to WT. Contamination by Lhcb4 in the trimer fraction is neither a concern as its amount is similarly low between the compared samples: soq1 mutant cold HL (qH ON) and soq1 lcnp mutant cold HL (qH OFF). So presence of Lhcb4 cannot account for the observed difference in fluorescence quenching as its quantity does not differ between the ON and OFF states. Importantly, the monomeric fraction, enriched in Lhcb4, does not show fluorescence quenching. We have used CN-PAGE as a complementary approach that showed that LHCII trimers are quenched after a cold and high light-treatment in both WT and soq1 mutants (Fig. 4). These aspects are described page 7 in the results section “qH is observed in isolated major LHCII”. Here we chose not to use BN/CN-PAGE or sucrose gradient ultracentrifugation for the isolation of the trimeric and monomeric fractions for two reasons: they would not be as suitable for TCSPC experiments due to their acrylamide or sucrose content and they would take more time; gel filtration was preferred to limit buffer exchange and time required from plant protein extraction to measurement.

Reviewer #2 (Significance (Required)):

Localization of qH in LHCII trimers is interesting, but not surprizing.

So, authors are recommended to rewrite the significance of their findings.

Response:____ In the last paragraph of the introduction, we have now clarified that this study identifies qH quenching in the LHCII trimers but not in the minor monomeric Lhcbs. Prior to this work, the peripheral antenna as a whole was known to be required for qH, now this study identified the major trimeric LHCII as a quenching site. The novelty and significance of this work is further substantified by the isolation of quenched antenna directly from plants in physiological conditions, as opposed to artificial induction in vitro. Regarding the “surprising” nature of findings in general, please see answer below to reviewer #3.

My expertise: I am working on the movement of L and M trimers in plants under photoinhibitory illumination.

---

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

The studies reported in this manuscript were designed to test the hypothesis that LCNP binds (or modifies) a molecule in the vicinity of (or within) the antenna proteins, under stress conditions. This in turn triggers a conformational change that converts antenna proteins from a light-harvesting to a dissipative state. Experiments were performed to locate the qH quenching site within the peripheral antenna of PSII and determine its sensitivity to Lhcb subunit composition. The authors were able to isolate antenna complexes with active qH that remained quenched after purification. Analysis of these complexes revealed that qH can occur in the major trimeric LHCII complexes. The elegant studies reported in this manuscript have made good use of appropriate molecular techniques and genetic resources. Genome editing and genetic crosses were used to demonstrate that qH is not restricted to inherent regulation of a specific major Lhcb subunit. The data are clearly presented and the data are convincing.

Response: Thank you very much for your appreciation of our work!

Reviewer #3 (Significance (Required)):

The studies reported in this manuscript build on a firm foundation of previous work by these authors and others. The conclusions are based on the analysis of Chl fluorescence lifetimes in intact leaves, thylakoids, and isolated antenna complexes in which qH was "ON" or "OFF". The findings are interesting and incremental in terms of increasing current understanding. However, the data extend our knowledge of the location of qH within the peripheral antenna of PSII. Rather unsurprisingly, the authors highlight the need to preserve thylakoid membrane macroorganisation for a full qH response.

Response: The philosophical concept of findings not being surprising could be discussed at length. To quote a commenter from this blog: ____https://blogs.uw.edu/ajko/2009/09/17/whats-surprising/____, just because one could have guessed the outcome of an experiment is not the same as empirically validating it. We hope you agree. Plus, as Fabrice Rappaport used to say, “we’re never sheltered from a discovery” and it could have been that isolated LHCII with qH ON showed short Chl fluorescence lifetimes similar to observed in leaves. We couldn’t know until we tried!

Data are presented showing that while qH occurs in the trimeric LHCII complexes, it does not require a specific Lhcb subunit and is insensitive to Lhcb composition. However, the discussion is rather speculative because data interpretation is limited by an absence of knowledge regarding what happens to the LHC trimers and qH during isolation of thylakoids and photosynthetic complexes. This point is considered appropriately in the discussion. The authors also acknowledge the existence of additional quenching sites beyond the LHCII trimers that are required for qH.

Response: Indeed, thank you we have addressed these points in the discussion, and have now added new data on the lack of changes in lipid composition in the LHCII trimer with qH ON or OFF. We view this study as an important milestone to obtain knowledge on the molecular origin of qH.

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 studies reported in this manuscript were designed to test the hypothesis that LCNP binds (or modifies) a molecule in the vicinity of (or within) the antenna proteins, under stress conditions. This in turn triggers a conformational change that converts antenna proteins from a light-harvesting to a dissipative state. Experiments were performed to locate the qH quenching site within the peripheral antenna of PSII and determine its sensitivity to Lhcb subunit composition. The authors were able to isolate antenna complexes with active qH that remained quenched after purification. Analysis of these complexes revealed that qH can occur in the major trimeric LHCII complexes. The elegant studies reported in this manuscript have made good use of appropriate molecular techniques and genetic resources. Genome editing and genetic crosses were used to demonstrate that qH is not restricted to inherent regulation of a specific major Lhcb subunit. The data are clearly presented and the data are convincing.

Significance

The studies reported in this manuscript build on a firm foundation of previous work by these authors and others. The conclusions are based on the analysis of Chl fluorescence lifetimes in intact leaves, thylakoids, and isolated antenna complexes in which qH was "ON" or "OFF". The findings are interesting and incremental in terms of increasing current understanding. However, the data extend our knowledge of the location of qH within the peripheral antenna of PSII. Rather unsurprisingly, the authors highlight the need to preserve thylakoid membrane macroorganisation for a full qH response.

Data are presented showing that while qH occurs in the trimeric LHCII complexes, it does not require a specific Lhcb subunit and is insensitive to Lhcb composition. However, the discussion is rather speculative because data interpretation is limited by an absence of knowledge regarding what happens to the LHC trimers and qH during isolation of thylakoids and photosynthetic complexes. This point is considered appropriately in the discussion. The authors also acknowledge the existence of additional quenching sites beyond the LHCII trimers that are required for qH.

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

Authors observed qH in isolated LHCII trimers with Chl fluorescence changes (shorter), and concluded that no single major Lhcb isomers is necessary for qH.

Major concern is: LHCII trimers are divided into S, M, L trimers with different compositions. Authors are requested to interpret their results in terms of L-, M-, S-trimers.

Minor comments are: Authors describes qI as reversible NPQ, but qI with D1 damage is not reversible.

In page 3 - 2nd paragraph, Authors define components of NPQ one by one, but the definition or revoery kinetics for qH is skipped, And authors suddenly start explaining molecular players of qH without changing paragraph.

In Fig. S6, authors tried to confirm the trimer and monomer fractions they used by using Lhcb2 and Lhcb4 antibodies, respectively. But, the distribution of Lhcb2 only in Trimer fraction in WT, which is diferetnf from the distribution in other mutants. Contamination of Lhcb4 in Trimer fraction is also of concern. Authors may use Bn-PAGE or Ultracentrigugal separation, rather than gel filtration.

provide evidences for

Significance

Localization of qH in LHCII trimers is interesting, but not surprizing.

So, authors are recommended to rewrite the significance of their findings.

My expertise: I am working on the movement of L and M trimers in plants under photoinhibitory illumination.