Vacuolar H⁺-ATPase Determines Daughter Cell Fates through Asymmetric Segregation of the Nucleosome Remodeling and Deacetylase Complex

Joint Public Review:

Xie et al. propose that the asymmetric segregation of the NuRD complex is regulated in a V-ATPase-dependent manner, and plays a crucial role in determining the differential expression of the apoptosis activator egl-1 and thus critical for the life/death fate decision.

While the model is very intriguing, the reviewers raised concerns regarding the rigor of the method. One issue is with statistics (either insufficient information or inadequate use of statistics), and second is the concern that the asymmetry observed may be caused by one cell dying (resulting in protein degradation, RNA degradation etc). We recommend that the authors address these issues.

Major #1:

There are still many misleading statements/conclusions that are not rigorously tested or that are logically flawed. These issues must be thoroughly addressed for this manuscript to be solid.

1. Asymmetry detected by scRNA seq vs. imaging may not represent the same phenomenon, thus should not be discussed as two supporting pieces of evidence for the authors' model, and importantly each method has its own flaw. First, for scRNA seq, when cells become already egl-1 positive, those cells may be already dying, and thus NuRD complex's transcripts' asymmetry may not have any significance. The data presented in FigS1D, E show that there are lots of genes (6487 out of 8624) that are decreased in dying cells. Thus, it is not convincing to claim that NuRD asymmetry is regulated by differential RNA amount.

2. Regarding NuRD protein's asymmetry, there are still multiple issues. Most likely explanation of their asymmetry is purely daughter size asymmetry. Because one cell is much bigger than the other (3 times larger), NuRD components, which are not chromatin associated, would be inherited to the bigger cell 3 times more than the smaller daughter. Then, upon nuclear envelope reformation, NuRD components will enter the nucleus, and there will be 3 times more NuRD components in the bigger daughter cell. It is possible that this is actually the underling mechanism to regulate gene expression differentially, but this possibility is not properly acknowledged. Currently, the authors use chromatin associated protein (Mys-1) as 'symmetric control', but this is not necessarily a fair comparison. For NuRD asymmetry to be meaningful, an example of protein is needed that is non-chromatin associated in mitosis, distributed to daughter cells proportional to daughter cell size, and re-enter nucleus after nuclear envelope formation to show symmetric distribution. And if daughter size asymmetry is the cause of NuRD asymmetry, other lineages that do not undergo apoptosis but exhibit daughter size asymmetry would also show NuRD asymmetry. The authors should comment on this (if such examples exist, it is fine in that in those cell types, NuRD asymmetry may be used for differential gene expression, not necessarily to induce cell death, but such comparison provides the explanation for NuRD asymmetry, and puts the authors finding in a better context).

3. For the analysis of protein asymmetry between two daughters in Fig S4C, the method of calibration is unclear, making it difficult to interpret the results.

4. As for pHluorin experiments, the authors were asked to test the changes in fluorescence observed are due to changes in pH or changes in the amount of pHluorin protein. They need to add a ratio-metric method in this manuscript. A brief mention to Page 12 line 12 is insufficient to clarify this issue.

Major #2:

Some issues surrounding statistics must be resolved.

1. Fig. 1FG, 2D, 3BDEG, 5BD and 6B used either one-sample t-test or unpaired two-tailed parametric t-test for statistical comparison. These t-tests require a verification of each sample fitting to a normal distribution. The authors need to describe a statistical test used to verify a normal distribution of each sample.

2. Fig. 2D, 3D, and 3G have very small sample size (N=3-4, N=6, N=3, respectively), it is possible that a normal distribution cannot be verified. How can the authors justify the use of one-sample t-test and unpaired parametric t-test ?

3. Statistical comparison in Fig. 2D and Fig. 6B should be re-assessed. For Fig. 2D, the authors need to compare the intensity ratio of HDA-1/LIN53 between sister cells dying within 35 min and those over 400 min. For Fig. 6B, they need to compare the intensity ratio of VHA-17 between DMSO- and BafA1- treated cells at the same time point after anaphase.

Author Response

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

Public Reviews:

Throughout the study, there is insufficient information about how experiments were performed and how often (imaging, pull-downs etc), how data was acquired, modified and analysed (especially imaging data, see below), how statistical analyses were done and what is presented in the figures (single planes or maximum intensity projections etc). This makes it difficult to evaluate the data and results.

We have incorporated additional experimental details to the Materials and Methods section: "Recent advancements in optical and camera technologies permit the acquisition of Z-stacks without perturbing Q cell division or overall animal development. Z-stack images were acquired over a range of -1.6 to +1.6 μm from the focal plane, at intervals of 0.8 μm. The field-of-view spanned 160 μm × 160 μm, and the laser power, as measured at the optical fiber, was approximately 1 mW. ImageJ software (http://rsbweb.nih.gov/ij/) was used to perform image analysis and measurement. Image stacks were z-projected using the average projection for quantification and using the maximum projection for visual display. "

The majority of our experimental procedures adhere to methodologies delineated in our prior publications and other scientific literature. We were pioneers in the development of fluorescence time-lapse live microscopy techniques for capturing Q cell migration and asymmetric division (Ou and Vale, Journal of Cell Biology, 2009; Ou et al., Science, 2010; Chai et al., Nature Protocols, 2012). Our innovative imaging protocol uncovered a novel mode of polarized, non-muscle myosin-II-dependent asymmetric cell division (Ou et al., Science, 2010). Subsequently, we unveiled another previously uncharacterized mechanism of asymmetric cell division dependent on polarized actin polymerization (Chai et al., Cell Discovery, 2022). In the present study, we have significantly refined our imaging and quantification protocols. Different from the single-focal-plane imaging employed in our earlier study by Ou et al. 2009, advancements in optical technologies and camera resolution now enable us to undertake time-lapse imaging across multiple focal planes and track signal differences between the anterior and posterior segments of dividing cells.

There is insufficient information about tools and reporters used. This is misleading and impacts the conclusions that can be made from the results presented. To give an example, in Figure 1D-F, the authors present data that HDA-1::GFP and LIN-53::mNeonGreen (both components of the nucleosome remodeling and deacetylation complex) but not the histone acetyltransferase MYS-1::GFP are 'asymmetrically segregated' during QR.a division. However, the authors do not mention that HDA-1::GFP and LIN-53::mNeonGreen are expressed at endogenous levels (they are CRISPR alleles) whereas MYS-1::GFP is overexpressed (integration of a multi-copy extrachromosomal array). The difference in 'segregation' could therefore be a consequence of different levels of expression rather than different modes of segregation ('asymmetric' versus 'symmetric').

Figure S2 shows overexpressed HDA-1, LIN-53 and CHD-3 are also asymmetrically segregated during ACD of QR.a, which indicates that different levels of expression do not affect the modes of segregation, at least for the NuRD subunits. In the main text, however, we presented the asymmetric segregation of HDA-1::GFP and LIN-53::mNeonGreen using their CRISPR KI alleles.

There is insufficient information about the phenotypes of the animals used (RNAi knock-downs of hda-1, lin-53 RNAi, pig-1 etc). Again this is misleading and impacts the conclusions that can be made. To give some examples,

1. In Figure 3A-G, control RNAi embryos are compared to hda-1 RNAi and lin-53 RNAi embryos. What the authors do not mention is that hda-1 RNAi and lin-53 RNAi embryos have severe developmental defects and essentially cannot be compared to control RNAi embryos. The differences between the embryos can be seen in Figure S7B where bright-field images of control RNAi, hda-1 RNAi and lin-53 RNAi embryos are shown. At the 350 min time point, a normal embryo is visible for the control, a 'ball of cells' embryo for hda-1 RNAi and an embryo that seems to have arrested at an earlier developmental stage (and therefore have much larger cells) for lin-53 RNAi. Because of these pleiotropic phenotypes, it is unclear whether differences seen for example in sAnxV::GFP positive cells (Figure 3A) are the result of a direct effect of hda-1(RNAi) on cell death or whether they are the result of global changes in development and cell fate induced by hda-1(RNAi). hda-1(RNAi) and lin-53(RNAi) embryos are also used for the data shown in Figures S6 and S7, raising the same concerns;

In the submitted manuscript, we mentioned that hda-1 RNAi and lin-53 RNAi caused embryonic lethality and that we could track some of the apoptotic events in hda-1 RNAi embryos arrested between the late gastrulation stage and bean stage. We agree with the reviewers that because of the pleiotropic phenotypes, we cannot distinguish whether sAnxV::GFP positive cells (Figure 3A) are the result of a direct effect of hda-1 (RNAi) on cell death or whether they are the result of global changes in development and cell fate induced by hda-1 (RNAi). We added the sentence to page 9 line 26: “Considering the pleiotropic phenotypes caused by loss of HDA-1, we cannot exclude the possibility that ectopic cell death might result from global changes in development, even though HDA-1 may directly contribute to the life-versus-death fate determination.”

1. The authors do not mention what the impact of Baf A1 treatment is on animals; however, the images provided in Figure 5E indicate that Baf A1 treatment causes pleiotropic effects in L1 larvae.

We have carefully checked the BafA1 treated animals, but have not been able to detect any visible defect in Baf A1 treated animals under a 25× dissection microscope at the given dosage and duration of treatment. We also searched for the published images or literature and did not find pleiotropic effects on the animal level at this dosage and duration; however, we agree with the reviewers that perturbation of pH homeostasis in lysosomes by BafA1 will certainly generate pleiotropic cellular defects. We discussed the issue below:

"Although BafA1-mediated disruption of lysosomal pH homeostasis is recognized to elicit a wide array of intracellular abnormalities, we found no evidence of such pleiotropic effects at the organismal level with the dosage and duration of treatment employed in this study."

There is a lack of adequate controls. Because of this, some of the data presented must be considered as preliminary. To give some examples:

1. Controls are lacking for the data shown in Figure 3D-G (i.e. genes other than egl-1). Since hda-1 RNAi has a pleiotropic effect and most likely affects H3K27 acetylation genome-wide, this is critical. Based on what is shown, it is unclear whether the results presented are specific to egl-1 or not;

In figure 3F, we added F23B12.1 and sru-43 as the controls of egl-1. We added “while the H3K27ac level of genes adjacent to egl-1 showed no significant changes” to Page 10 line 22 in the revised text.

1. The co-IP and mass spec data shown in Figure 4A, C and Figure S8 also lack a critical control, which is GFP only. Because of this, it is unclear whether subunits of the V-ATPase bind to HDA-1 or GFP. The co-IP and mass spec data forms the basis of Figures 5 and 6 as well as Figure S9. Data presented in these figures therefore has to be considered preliminary as well.

In the co-IP and mass spec shown in Figure 4A, we used ACT-4::GFP as the negative control, which can preclude V-ATPase subunits that bind to GFP. In Figure 4C, we used anti-V1A (V-ATPase V1 domain A subunit) antibody to confirm the interaction between V1A and HDA-1. In Figure S8B, we also used ACT-4::GFP as a control, showing other NuRD subunits bind to HDA-1 rather than GFP.

Inappropriate methods are used. For this reason, some of the data again must be considered preliminary. To give some examples:

1. In Figure 5A, B, the authors used super-ecliptic pHluorin to look at changes in pH in the daughter cells. However, the authors used quenching of super-ecliptic pHluorin fluorescence rather than a ratio-metric method to 'measure' changes in pH. Because of this, it is unclear whether the changes in fluorescence observed are due to changes in pH or changes in the amount of pHluorin protein. Figure 5A, B forms the basis for the experiments presented in the remaining parts of Figure 5 as well as in Figure 6 and Figure S9;

Bafilomycin A1 inhibits the activity of V-ATPase, presumably preventing the pumping of protons into the apoptotic daughter cell. It is more likely that the apoptotic daughter cell becomes less acidic and more neutral after the treatment of Baf1A, although we cannot exclude the possibility that the changes in fluorescence could be due to changes in the amount of pHluorin protein. A ratio-metric method to measure changes in pH will be further used to distinguish the two possibilities.

We added “although we cannot exclude the possibility that the changes in fluorescence could be due to changes in the amount of pHluorin protein.” to Page 12 line 12 in the revised text.

1. The authors' description of how some images were modified before quantitative analysis raises concerns. The figures of concern are particularly Figure 1 and Figure S4, where background subtraction with denoising and deconvolution was used. Background subtraction, with denoising and deconvolution is an image manipulation that enhances the contrast between background and what looks like foreground. Therefore, background subtraction should be applied primarily in experiments involving image segmentation not fluorescence intensity measurement. Not being provided any information by the authors about the kind of subtraction that was made, this processing could lead to an uneven subtraction across the image, which can easily lead to artefacts. Since the fluorescence intensity in the smaller daughter cell is lower, and thus closer to background, the algorithm the authors used may have misinterpreted the grey value information in the smaller daughter cell pixels. This could have led to an asymmetric subtraction of background in the two daughter cells, leading to a stronger subtraction in the smaller daughter cell. Ultimately, their processing could have artificially increased the intensity asymmetry between the two daughter cells in all their results.

As mentioned earlier, the imaging and quantification methods of this manuscript have been routinely used in our previous publications or studies from many other labs (Gräbnitz F, et al., Cell Rep. 2023; Herrero E, et al., Genetics. 2020; Roubinet C, et al., Curr Biol. 2021). Background subtraction is a standard procedure to quantify cellular fluorescence intensities. The fluorescence intensity of the slide background was measured from a region without worm bodies, of the same size as the region of interest. We have added how we measured the background to page 19 Line 24: “The fluorescence intensity of the slide background was measured from a region without worm bodies, of the same size as the region of interest.”

The imaging data is of low quality (for example Figures 1, 2, 5, 6; Figures S2, S3, S5, S6, S9). Since much of the study and the findings are based on imaging, this is a major concern. Critical parameters are not mentioned (number of sections in z-stack, size of the field-of-view, laser power used etc), which makes it difficult to understand what was done and what one is looking at.

Fluorescence images of neuroblast asymmetric cell division in developing C. elegans larvae has historically presented considerable challenges. Our recent methodological advancements have facilitated live imaging in this intricate system with improved resolution. In the revised manuscript, we have elucidated the specific z-stack parameters, field-of-view dimensions, and laser power settings employed: "Z-stack images were acquired over a range of -1.6 to +1.6 μm from the focal plane, at intervals of 0.8 μm. The field-of-view spaned 160 μm × 160 μm, and the laser power, as measured at the optical fiber, was approximately 1 mW."

To give some specific examples,

1. The images shown in Figure 2B are of very low quality with severe background from neighbouring cells. In addition, the outline of the cells (plasma membrane) or the nuclei of the daughter cells is unknown. Based on this it is not clear how the authors could have measured 'Fluorescence intensity ratio between sister nuclei' in an accurate and unbiased way (what is clear from these images is that there is an increase in HDA-1::GFP signal in ALL surviving daughters (asymmetric and symmetric divisions) post cytokinesis but not in the daughter cell that is about to die (asymmetric and unequal division));

We employed live-cell imaging in conjunction with automated cell lineage tracing algorithms (Du et al., Cell, 2014) to scrutinize NuRD asymmetry in embryos from the two- or four-cell stage up to the 350-cell stage. This sophisticated approach was initially pioneered by Dr. Zhirong Bao at Sloan Kettering and subsequently refined by Dr. Zhuo Du during Dr. Du's postdoctoral training in Dr. Bao's laboratory. This advanced imaging pipeline enables the scientific community to quantify cellular fluorescence intensity in an automated fashion, thereby substantially mitigating manual intervention and bias.

1. The images in Figure 6A and Figure S9A on VHA-17 segregation and its colocalization to ER and lysosome segregation during QR.a division are of very low quality and it is unclear to the reviewer how such images were used to obtain the quantitative data shown.

In some cases, there is a discrepancy between what is shown in figures and what the authors state in the text. To give some examples:

1. On page 7, the authors state "..., we found that nuclear HDA-1 or LIN-53 asymmetry gradually increased from 1.1-fold at the onset of anaphase to 1.5 or 1.8-fold at cytokinesis, respectively (Figure 1D-E)." Looking at the images for HDA-1 and LIN-53 in Figure 1D, the increase in the ratio mainly occurs between 4 min and 6 min, which is post cytokinesis and NOT prior to cytokinesis;

Thank the reviewer for pointing out this. The nuclear HDA-1 or LIN-53 asymmetry increased to 1.5 or 1.8-fold 6 min after the onset of anaphase, when QR.a just completes cytokinesis. Therefore, We change the sentence “we found that nuclear HDA-1 or LIN-53 asymmetry gradually increased from 1.1-fold at the onset of anaphase to 1.5 or 1.8-fold at cytokinesis, respectively (Figure 1D-E).” to “we found that nuclear HDA-1 or LIN-53 asymmetry gradually increased from 1.1-fold at the onset of anaphase to 1.5 or 1.8-fold upon the completion of cytokinesis, respectively (Figure 1D-E).”

However, nuclear HDA-1 or LIN-53 asymmetry initiates prior to cytokinesis. We started to see the nuclear HDA-1 or LIN-53 asymmetry (1.4 fold for HDA-1 and 1.2 fold for LIN-53 ) 2 min after the onset of anaphase (Figure 1D).

1. These images (Figure 1D) also show that there is an increase in the HDA-1 and LIN-53 signals in the larger daughter cells (QR.ap), which suggests that the increase in ratios (Figure 1E) is the result of increased HDA-1 and LIN-53 synthesis post cytokinesis. However, on top of page 8, the authors state "The total fluorescence of HDA-1, LIN-53 and MYS-1 remained constant during ACDs, suggesting that protein redistribution may establish NuRD asymmetry (Figure S4C)." In Figure S4C, the authors present straight lines for 'relative total fluorescence' for imaging (probably z-stacks) that was done every min over the course of 7 min. If there was no increase in material as the authors claim, they should have seen significant photobleaching over the course of the 7 min and therefore reduced level of 'relative total fluorescence' over time. How the data presented in Figure S4C was generated is therefore unclear. (Despite the fact that the authors claim that the asymmetry seen is not due to new synthesis in the larger daughter cell post cytokinesis, it would be more consistent with the first experiment presented in this study (Figure S1) that shows that there is more hda-1 mRNA in egl-1(-) cells compared to egl-1(+) cells);

Regarding the concern of photo-bleaching, we have meticulously calibrated our imaging system over the past several years. Rigorous controls, qualification, and analyses were scrupulously undertaken during the development of our fluorescence time-lapse imaging system for the investigation of Q cell dynamics, initially established by Dr. Guangshuo Ou in Ron Vale's laboratory—a renowned hub for avant-garde imaging techniques (Ou & Vale, Journal of Cell Biology, 2009; Ou et al., Science, 2010). Remarkably, no discernible photobleaching was observed even during two to three-hour imaging.

We agree that protein turnover, involving both degradation and synthesis, may occur. However, NuRD asymmetric distribution occurred within several minutes after metaphase and QR.a completes cytokinesis ~6min after the onset of anaphase, while GFP protein translation and maturation require ~ 30 min in Q neuroblast (Ou & Vale, Journal of Cell Biology, 2009). Even if hda-1::gfp mRNA is translated during cell division, the nascent GFP-tagged protein will mature long after the completion of cytokinesis. Consequently, we postulate that the influence of newly synthesized GFP-tagged protein during Q cell division is negligible for quantification purposes. It is plausible that the asymmetry in HAD-1 protein distribution is independent of hda-1 mRNA asymmetry.

1. On page 12, the authors state "..., in Baf A1-treated animals, QRaa inherited similar levels of HDA-1::GFP as its sister cell,...". However, looking at the image provided in Figure 5E (0 min), there seems to be a similar ratio of HDA-1::GFP between the daughter cells in DMSO and Baf A1-treated animals.

We have adjusted the images in Figure 5E to show the asymmetry in DMSO-treated control animals. We acknowledge variations among animals. Our quantifications from more than 10 animals show the HDA-1 asymmetry in DMSO-treated animals in Figure 5B.

Recommendations for the authors:

Conclusion 1

"Here, we demonstrate that the nucleosome remodeling and deacetylase (NuRD) complex is asymmetrically segregated into the surviving daughter cell rather than the apoptotic one during ACDs in Caenorhabditis elegans" (Abstract)

Results described on pages 6-9 ("NuRD asymmetric segregation during neuroblast ACDs" and "NuRD asymmetric segregation in embryonic cell lineages") and data shown in Figure S1, Figure 1, Figures S2, S3, S4, S5, Figure 2.

Conclusion 1 is not supported by the results as numerous concerns exist about the data in many of these figures (see above, major weaknesses). A more likely explanation for the authors' observations is that there is synthesis of NuRD post cytokinesis and that asymmetries in the amounts of NuRD observed in the two daughter cells is a consequence of their different cell sizes (QR.ap is 3x as large as QR.aa). This is supported by the finding that the loss of pig-1, which causes 'equal' division resulting in two daughter cells of similar sizes, abolishes the differences in NuRD seen between the daughter cells.

As discussed earlier, GFP protein translation and maturation require ~ 30 min in Q neuroblast (Ou & Vale, Journal of Cell Biology, 2009). Even if there is the synthesis of NuRD post cytokinesis, the nascent GFP-tagged protein will not mature within our imaging timeframe, Therefore, NuRD asymmetry is unlikely to be a result of the synthesis of NuRD post cytokinesis. In addition, We found that MYS-1::GFP was symmetrically segregated into the small apoptotic daughter cells and big surviving daughter cells, suggesting NuRD asymmetry may be irrelevant to cell size asymmetry.

Interestingly, despite the fact that the loss of pig-1 causes 100% of the divisions to be equal by size and symmetric with respect to NuRD amounts, it only causes about 30% of QR.aa cells to inappropriately survive. This demonstrates that there is a correlation between NuRD asymmetry and daughter cell size asymmetry but NOT between NuRD asymmetry and cell death. This also demonstrates that loss of 'NuRD asymmetry' and presence of NuRD in the daughter that should die is NOT sufficient to block its death.

Cordes et al. 2006 (DOI: 10.1242/dev.02447) reported that in pig-1 loss-of-function mutants, <40% of Q.p lineages produce extra neurons because Q.pp cells inappropriately survive. Noticeably, only 30% and 5% Q.p lineages produce extra neurons in ced-3 and egl-1 loss of function single mutant, respectively. pig-1 ced-3 double mutant or pig-1 egl-1 double mutants show a dramatically stronger phenotype than either single mutant, resulting in about 80% of Q.p lineages producing extra neurons. These results suggest that pig-1 functions in parallel to the EGL-1-CED-9-CED-4-CED-3 cell death pathway to promote Q cell apoptosis.

We agree with the reviewer that “loss of 'NuRD asymmetry' and presence of NuRD in the daughter that should die is NOT sufficient to block its death” in pig-1 loss-of-function mutants. However, these results do not rule out the correlation between NuRD asymmetry and cell death. In the pig-1 mutant, the concentration of NuRD in Q.pp might not be high enough to completely block the death pathway. Alternatively, NuRD may be one but not the only factor blocking the cell death pathway.

Lastly, it is imperative to underscore that cellular aberrations observed during early developmental stages frequently undergo compensatory correction during subsequent developmental stages or even initial aging stages. For example, in certain cell migration mutants exhibiting early migration defects, the initial penetrance exceeds 80%; however, the penetrance is mitigated to a mere 30% in adults. Such observations have been corroborated in our prior publications focusing on cell migration dynamics (Wang et al., PNAS, 2013; Zhu et al., Dev Cell, 2016). This appears to be a pervasive phenomenon, echoed by several laboratories specializing in neural development. Sengupta and Blacque’s labs has reported that early aging can ameliorate ciliary phenotypes in C. elegans mutants with compromised intraflagellar transport mechanisms. Accordingly, late developmental stages may act as a compensatory buffer for antecedent developmental abnormalities.

Conclusion 2

"The absence of NuRD triggers apoptosis via the EGL-1-CED-9-CED-4-CED-3 pathway, while an ectopic gain of NuRD enables apoptotic cells to survive." (Abstract) Results described on pages 8-10 ("Loss of the deacetylation activity of NuRD causes ectopic apoptosis" and "NuRD RNAi upregulates the egl-1 expression by increasing its H3K27 aceylation") and data shown in Figure S6, Figure 3, Figure S7 and data shown in Figure 5.

Because of the various concerns raised above (major weaknesses) about the data presented in Figure S6, Figure 3, Figure S7 (pleiotropic phenotypes of hda-1 and lin-53 RNAi animals, lack of controls etc), there is no evidence that NuRD has a specific and/or direct effect on egl-1 expression in cells programmed to die or that loss of NuRD causes ectopic egl-1-dependent cell death. The claim that "ectopic gain of NuRD enables apoptotic cells to survive." is based on Figure 5E, where the authors show that Baf A1 treatment causes symmetric NuRD segregation in 11/12 animals and that QR.aa survives in 11/12 animals. However, those data are unconvincing. As mentioned above (major weaknesses), from the low-quality images provided, it is not clear whether there is 'symmetric NuRD segregation' in Baf A1 treated animals, and the conditions of the animals are a concern because of pleiotropic effects of blocking V-ATPase. (I am not convinced I am actually looking at the same region of an L1 larvae in the three animals because the HDA-1::GFP signal seems inconsistent across them.) One process that is affected by a block of V-ATPase is engulfment. The fact that the authors observe that 130 min post-cytokinesis, QR.aa still persists in Baf A1 treated animals could therefore be the result of a delay in engulfment rather than a block in cell death. In addition, the claim that ectopic gain of NuRD enables apoptotic cells to survive contradicts their findings on loss of pig-1 described about ('Conclusion 1').

We acknowledge the limitations of our imaging system; however, as we pointed out earlier that we developed imaging methods and kept improving them. We have tried our best to obtain images from developing C. elegans larvae. On the other hand, we showed that hda-1 RNAi and lin-53 RNAi increase the expression of a subset of genes, including egl-1, either directly or indirectly (Fig. 3C). Figure 3B shows the ectopic cell death caused by loss of NuRD is dependent on EGL-1-CED-9-CED-4-CED-3 pathway. While we cannot exclude several other possibilities while addressing such a complex problem in such a challenging model system, these results provide some evidence supporting that our claim can be one of the possibilities.

Conclusion(s) 3

"We identified the vacuolar H+-adenosine triphosphatase (V-ATPase) complex as a crucial regulator of NuRD's asymmetric segregation. V-ATPase interacts with NuRD and is asymmetrically segregated into the surviving daughter cell. Inhibition of V-ATPase disrupts cytosolic pH asymmetry and NuRD asymmetry" (Abstract)

Results described on pages 10-13 ("V-ATPase regulates asymmetric segregation of NuRD during somatic ACDs") and data shown in Figures 4, 5, 6, Figures S8, S9.

As outlined above (major weaknesses), the evidence that HDA-1 interacts with the V-ATPase complex is preliminary (no GFP control), and I consider the imaging data showing that V-ATPase asymmetrically segregates very low quality and unconvincing (Figure 6). The data on pH changes are also preliminary as the experiment was not done the way it should have (quenching rather than ratiometric). Finally, there are concerns about the results that apparently demonstrate that inhibiting V-ATPase activity disrupts pH asymmetry and NuRD asymmetry (impact of Baf A1 treatment).

As discussed earlier, Bafilomycin A1 inhibits the activity of V-ATPase, presumably preventing the pumping of protons into apoptotic daughter cells. It is more likely that the apoptotic daughter cell becomes less acidic and more neutral after the treatment of Baf1A, although we cannot exclude the possibility that the changes in fluorescence could be due to changes in the amount of pHluorin protein. A ratio-metric method to measure changes in pH will be further used to distinguish the two possibilities.

We added “although we cannot exclude the possibility that the changes in fluorescence could be due to changes in the amount of pHluorin protein.” to Page 12 line 12 in the revised text.

Conclusion 4

"We suggest that asymmetric segregation of V-ATPase may cause distinct acidification levels in the two daughter cells, enabling asymmetric epigenetic inheritance that specifies their respective life-versus-death fates." (Abstract) Discussion and model Figure 6C.

I consider the model premature and not based on any convincing data. In addition, the role of V-ATPase and acidification does not make sense. V-ATPase is involved in the acidification of the lysosomal system (lumen), and it is thought that cytosolic acidification in apoptotic cells is caused by lysosomal leakage. This is not consistent with the authors' model.

This manuscript lacks a section describing details of statistical analyses and the rationale for the chosen test, sample sizes, exclusion criteria, and replication details. Although the sample size is relatively smaller (less than 30), the authors used "unpaired t-test" for most of the tests. They should describe which type of t-test they used (parametric or non-parametric test). They also should provide replication details for non-statistical data set, for example Fig 3F and Fig 4C.

We used the Unpaired two-tailed parametric t-test. We have now added the information for statistic tests in the revised methods and figure legends.