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 reports the control of PALB2 - chromatin interaction by the acetylation of a particular lysine-rich domain of the protein called ChAM. This acetylation is shown to be mediated by the acetyltransferases KAT2A/B. Following these investigations, the authors made an effort to place their findings in the context of DNA replication and DNA repair. The proposed model is that the acetylation-dependent interaction of PALB2 with chromatin could ensure the protection of the genome during DNA replication and control DNA repair.

Specific remarks

1 - Based on different experiments, essentially the one shown in Fig. 3B, the authors conclude that the acetylation of the ChAM domain enhances its association with nucleosomes. However, taking into account the experimental setting, this conclusion should be largely tuned down. Indeed, this enhanced acetylation-dependent nucleosome binding was observed when the experiment was carried out in the presence of excess of free naked DNA. Under these conditions, the non-acetylated ChAM fragments became mostly trapped by DNA (clearly shown in Fig. 3C/D), and hence would not be available for nucleosome binding, while the acetylated ChAM fragments would remain available for nucleosome association because of their reduced DNA-binding ability.

Consequently, the acetylation of the ChAM domain would only play a role on the availability of PALB2 for chromatin/nucleosome binding and not directly stimulate nucleosome binding. Therefore, the nucleosome-binding capacity of ChAM by itself should not be dependent on ChAM domain acetylation.

If true, this hypothesis could also be relevant in vivo since the poly-K in the ChAM domain could also non-specifically interact with nuclear RNAs and hence its acetylation, by releasing it from nuclear RNAs, would make it available for chromatin-binding. The importance of RNAs in the regulation of PALB2 nucleosome-binding could be tested in the experiments shown in Fig. 2C and 2E by adding RNase to the pull-down medium (WT +/-RNase or addition of increasing exogenous RNAs).

2 - The real question is as follows. While acetylation makes the protein available for nucleosome binding, which part of the ChAM domain is actually mediating nucleosome binding and whether lysine acetylation could be directly involved in this binding. Another question would be to identify the elements in the nucleosome mediating this interaction, histones (core domain, tails, post-translational modifications, specific histone types), histone-DNA, etc...

3 - Taking into account the authors conclusions on the role of ChAM domain acetylation and its impact on PALB2 mobility, in Figure 4D/E, one should expect a difference of t1/2 when wild-type and 7R mutant are assayed by FRAP. At least the measures of t1/2 in the wild-type should have been more heterogeneous compared to the 7R mutant due to the acetylation of the wild-type PALB2 by the endogenous HATs (the impact of endogenous HATs on the wild-type sequence is shown in Fig. 3F). Could the authors comment on this?

4 - It would be better to remove the data presented in Fig. 5 since, as currently presented, these investigations remain shallow and do not bring much information on what is happening. The presented data are rather confusing since, in the absence of further investigations, it is not clear which one(s) of the mechanisms involved in the control of DNA replication is controlled by PALB2 and many explanations, including artefacts, remain possible.

The manuscript would gain in interest if the authors would devote the functional studies only to the repair part (Fig.6 and 7).

Minor points

5 - High background of non-enzymatic acetylation of PALB2 fragments makes the identification of KAT2A/B specific acetylation not very convincing. The immunoblot detection of acetylation fragments shown in Figure S1 is much more convincing. Therefore, the authors may consider to present Fig S1 as a main Figure and Fig.1B as a supplementary one.

6 - It would be interesting if the authors would comment on why the presence of regions flanking the ChAM domain (Fig. 1A, construct #5) significantly reduces chromatin (Fig. 1B) and nucleosome binding (Fig. 1C).

Significance

This manuscript brings interesting information on the control of the activity of PALB2 by acetylation which depends on specific cellular HATs. By doing so, it also opens the door to envision a role for a metabolic reprogramming of protein acetylation in the control of PALB2 activity, as discussed by the authors.

This is an interesting report but the functional part is rather weak, decreasing the interest of the manuscript as a whole. This report would gain in interest if the authors would develop the first part (Figs 1- 4), while reducing the second part (Figs 5 -7) to the presentation and discussion of the strongest functional data (see my comments).

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Fournier et al. detect acetylation within the chromatin association motif (ChAM) of PALB2 and demonstrate that KAT2 can acetylate these 7 lysine residues within this region. They then generate K to R mutations (7R) or K to Q mutations (7Q) at these sites and perform assays of fluorescence recovery after photobleaching (FRAP) to measure mobility as a measure of chromatin association, RAD51 foci, PALB2 recruitment at sites of laser-induced DNA damage, and sensitivity to olaparib. They find increased mobility of the 7Q mutant of PALB2 but not 7R in the absence of exogenous DNA damage, as well as defects in DNA damage-induced RAD51 foci and resistance to olaparib. On this basis, the authors conclude that acetylation is required for the association of PALB2 with undamaged chromatin and that deacetylation permits mobilization and association with BRCA1 to enable proper DNA repair. While the manuscript is generally well-written, many of the systems are rather elegant, and this study may yield novel insights into the regulation and function of PALB2 in DNA repair, there are some missing experiments to be added and important contradictions that should be resolved in order to fully establish the new model the authors propose.

Major comments:

1.There are some concerns about the interpretation of experiments with the 7R and 7Q mutants of PALB2. For example, in the description of results in Fig. S2C, the authors state "K to R substitutions maintain the charge yet are unable to accept acetylation and hence mimic constitutively non-acetyl lysine". However, in Fig. 4B the association of the 7R mutant with chromatin is similar to WT and in Fig. 7D,E the relative immobility of the 7R mutant is very similar to WT PALB2. Thus, the conclusion that acetylation is required for PALB2 association with damaged undamaged chromatin and for release of PALB2 upon DNA damage does not appear justified. Perhaps the authors need to better consider whether the 7R mutant mimics acetylation because of its charge. Even so, the mutant then maintains the charge normally associated with acetylated PALB2, calling into question whether deacetylation indeed "releases PALB2 from undamaged chromatin".

2.Related to questions of interpreting results utilizing the 7R and 7Q mutants of PALB2, in Fig. 7B,C the 7R mutant but not 7Q supports RAD51 foci and resistance to olaparib similar to WT PALB2. The authors then state in the Discussion that "our work also suggests that caution should be exercised in the use of K to Q substitutions for functional studies of lysine acetylation". Thus, which mutant is giving the correct and reliable results? Perhaps even more importantly, if results with the 7Q mutant are suspect, the conclusion that deacetylation is required for HR (or DNA repair) is suspect because that is the only case where the authors see a defect in RAD51 foci and resistance to olaparib. Similarly, if the 7R mutant "mimics non-acetyl-lysine" then the fact that it has normal RAD51 foci and resistance to olaparib contradicts the conclusion that deacetylation is required for DNA repair.

3.There are multiple concerns about Figs. 5 and S5. In Fig. 5A-C, difference in cell cycle progression after synchronization are relatively small and no rationale/interpretation is given for how this may be related to PALB2 function is given. In Fig. 5D,E differences in the levels of gamma-H2AX as a marker of DNA damage between different forms of PALB2 do not become readily apparent until about 6 or more days after addition of doxycycline. As such, it seems that these could be indirect effects and it is unclear how strongly this supports the importance of PALB2 acetylation in the DNA damage response. In Fig. S5, it is interesting that there are difference in the association of different forms of PALB2 with 3 distinct active loci, but no error bars or measures of statistical significance are given. Further at 2 of the 3 loci, the association of the 7Q mutant is closer to WT than the 7R mutant. Taken together, neither Fig. 5 nor Fig. S5 strongly support the key conclusion that acetylation regulates the association of PALB2 with actively transcribed genes to protect them.

4.Figs. 6D-G and S6A-D conclude that "DNA damage triggers ChAM deacetylation and induces PALB2 mobilization" based upon FRAP experiments utilizing WT PALB2. But there is no control to demonstrate that this is a specific effect driven by the state of PALB2 acetylation. For example, DNA damage mmight cause global acetylation changes resulting in relaxed chromatin in which proteins that are not subject to acetylation-deacetylation also show increased mobility.

5.Fig. 7B shows that the 7Q mutant has diminished RAD51 foci while Fig. S7C,D suggests based upon a different methodology (laser-induced damage) that the 7Q mutant does not affect PALB2 recruitment. Since the issue of recruitment is key to the mechanism proposed, the authors should examine PALB2 foci instead as this may be a more sensitive assay of PALB2 recruitment.

6.The authors state in the last sentence of the Results section that "lysine residues within the ChAM 7K-patch are indispensable for PALB2 function in HR" but never test the mutants for HR using reporter assays. The manuscript would be strengthened by performing such assays.

7.The model for the role of ChAM acetylation in regulating PALB2 function presented in Fig. 7D is not fully supported by the data presented. Critically, while association with RAD51 and BRCA2 is tested in Fig. S7B, the authors hypothesize that deacetylation is required to release PALB2 to enable association with BRCA1 but this is not tested utilizing the mutants. Also, there are some specific points that should be considered in the context of the model. This include how DNA damage may trigger deacetylation, and whether it is the deacetylated state or the process of deacetylation of ChAM that is critical. Also, if acetylation is important for protecting active genes in the absence of DNA damage, is deacetylation necessary to release PALB2 local or global. This is important, because if it is local there needs to be a specific mechanism for local deacetylation, while if deacetylation is global that could result in transcriptionally active genes becoming unprotected.

Minor Comments:

a.Some parts of the Materials and Methods are overly long (such as the subsection on "Protein purification" and "immunofluorescence microscopy") and could be shortened by consolidating experimental details that are largely the same for related processes.

b.In the description of Fig. 1D, the statement "7K-patch, which is common to PALB2 orthologs" is misleading since there is not complete conservation of each lysine residue across each ortholog.

c.Figs. 3E,F and S3B,C perform FRAP in cells with knockdown of KAT2A/B as a surrogate for chromatin association. The authors note that this global reduction in acetylation increases PALB2 diffusion, but there is concern that this experiment is not very informative because the increased mobility may have nothing to do acetylation of PALB2.

Significance

The role of acetylation of PALB2 in the regulation and function of PALB2 was previously unknown. In this context, this is potentially an important manuscript in the DNA damage response and DNA repair field. Notably, these authors previously identified the ChAM motif which promotes chromatin association of PALB2 (JY Bleuyard et al. PMID:22193777) and demonstrated that PALB2 may associate with actively replicating genes aside of its general role in DNA repair (JY Bleuyard et al. PMID:28673974). In this context, the finding that acetylation of the ChAM motif appears to be associated with undamaged chromatin leads to the conclusion that acetylation may specifically regulate chromatin association of PALB2 and represent a switch for association with undamaged chromatin and damaged chromatin.

This work should be of specific interest to those working in the fields of DNA repair and DNA damage responses, as well as those with interests in chromatin biology. This study should be of general interest to individuals interested in cell biology, transcription, and cancer biology and therapeutics.

This reviewer's expertise is in the area of DNA damage responses and DNA repair.

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)):

**General comments**

The manuscript 'Second messenger control of mRNA translation by dynamic ribosome modification' is a very interesting follow up on the research performed by the authors published in 2016. Here, the authors continue their study by determining the impact of the intricate RimABK pathway in Pseudomonas fluorescens on translational reprogramming by controlled modification of ribosomal protein S6 in response to environmental signals. The manuscript is interesting and well written, and the results are sound. However, in my opinion the general conclusion is not supported by experimental data and leaves several potential explanations open. Thus, I suggest to either perform in vitro translation experiments using ribosomes equipped with glutamated S6 to validate translational selectivity, or to soften the language on the working model shown in Figure 12.

The authors would like to thank reviewer 1 for their detailed review of our manuscript. We agree with the reviewer that alternative explanations are possible for the translational changes linked specifically to glutamation, as opposed to rimBK deletion. Our intention when writing the discussion was to clearly distinguish glutamation-specific effects from the large number of indirect translational changes associated with Hfq disruption and other uncharacterised consequences of rimBK deletion. With hindsight, we acknowledge that the discussion and the model in figure 12 should more clearly outline the possible alternative causes for the observed glutamation-specific translational regulation. We have modified the discussion and figure 12 (now figure 10) accordingly.

The reviewer further suggests that we perform in vitro translation experiments using ribosomes equipped with glutamated S6, to prove that glutamation controls translation directly. This is an excellent suggestion that would clarify this important point, and we will certainly attempt it as part of our future analysis of the Rim system. However, at this stage we feel these experiments are beyond the intended scope of this paper, which is to describe the signal inputs and mechanism of the RimABK system and to show evidence for both specific and secondary translational effects of ribosome modification.

**Specific comments**

Figure 1 and S1: The RT-PCR analysis shown here does not allow excluding transcription initiation at alternative promoters downstream of the one determined by 5'-RACE. However, an alternative promoter might contribute to relative ratios between the rimA, rimB, and rimK mRNAs. A Northern blot and/or primer extension analysis could clarify this assumption and would give more detailed insights into the specific regulation.

The reviewer is correct that alternative rim promoters could exist downstream of the amplified 5'-RACE region. To test this hypothesis, we conducted additional RT-PCR experiments to measure expression of rimA (the third gene of the polycistronic rimABK operon) in the same set of conditions as tested for rimK. Relative levels of rimA mRNA do not substantially differ from those seen for rimK, strongly suggesting that the promoter upstream of rimK controls expression of all three rim genes. We have added this dataset to figure S1 and have modified the relevant sections of the text.

Figure 2B: I'm confused by the results shown here! I do only see a reduction of RpsF in the presence of RimA, RimK and cdG. What indicates the modification? Please, explain the interpretation of the result in more detail. Shouldn't the modified RpsF shift due to the addition of glutamate residues?

The uncontrolled activity of RimK acting in the absence of RimB (e.g. the experiment represented in Fig 2B) typically results in a reduction of the unmodified RpsF fraction in the reaction, replaced with RpsF proteins with widely varying numbers of glutamate residues attached to their C-termini. The resulting modified RpsF fraction can appear as a smear of protein density throughout the gel. We have clarified the text surrounding figure 2 to make this more explicit.

Figure 2C: Why does the RpsF modification lead to a supershift? How many glutamate residues are added? Is the smear visible in lane 4 (RpsF+RimK) representing already the slightly modified RpsF protein, which upon addition of RimA results in a supershift? For all SDS-Page analyses shown in the manuscript the validation of the glutamation using the antibodies specific against poly-glutamate would be a great asset to facilitate their interpretation.

Pseudomonas fluorescens RimK appears to have unregulated ligase activity, with many hundreds of glutamates being added to each RpsF protein in the absence of RimB cleavage. In our 2016 paper (Little et al., PLoS Genetics) we use radiolabelled glutamate incorporation and mass spectrometry to show that the supershifted protein smear is composed entirely of RpsF units with C-terminal glutamate tails of varying length. (It is interesting to note that E. coli RimK, which does not have an accompanying RimB protease, can only add 4-15 glutamates to each RpsF protein). We have modified the text slightly to make this clearer.

The reviewer’s suggestion to stain the supershifted RpsF with the poly-E antibody is interesting but would likely only reiterate our published results with radiolabelled glutamate (Little et al. 2016).

Lines 236-238: '...strongly suggesting that the proteomic changes we observe are an active response to modification of ribosomally-associated RpsF proteins.' This is an important suggestion as it allows a flexible and very fast integration of the external signals into a specialized protein synthesis. Thus, it definitely deserves further analysis! Considering that the purified RimA and RimK proteins are available, in vitro modification of RpsF in the context of the purified ribosome would be an important experiment and would greatly increase the quality of the paper. Up to now the selective or specialized translation is pure hypothesis and might also be explained by indirect effects via e.g. increased interaction between the ribosome and HFQ that might mediate interaction with certain mRNAs and thus stimulate their translation.

We agree with the reviewer that direct measurement of translational changes in vitro would tell us a great deal about the mechanism of RimK regulation. This would enable us to confirm whether the glutamation-specific effect is direct, or if it functions through an as-yet uncharacterised indirect mechanism (such as interaction with another translational regulator). As stated above we feel these major experiments are beyond the scope of the current manuscript, although we are keen to do them (as part of a planned structural biology investigation of modified ribosomes). As stated elsewhere in our response, we have extensively revised the discussion text and figure 12 to clarify the limits of our current understanding and highlight the different potential regulatory routes for RpsF glutamation.

Lines 322: '...into a single output: the proportion of all ribosomally-associated RpsF proteins that have C-terminal poly-glutamate tails.' Considering the identification of a group of genes whose translation is altered by rimBK deletion, but not by RpsF glutamation (Class 1, Fig 11B), I would suggest softening this statement. If I interpret the data correctly, they pinpoint to a moonlighting function of the rim-pathway that does not target RpsF!

The genes whose translation is affected by rimBK deletion, but not by RpsF glutamation specifically, include all those genes whose translation is indirectly affected by downstream translational regulators, or through interaction with another affected gene target. As expected, there is substantial overlap between the rimBK and hfq translatomes (Grenga et al. 2017): this analysis can be included in the manuscript as a supplementary table if requested. Importantly, there is very little overlap between the Hfq translatome and those genes that are affected specifically by RpsF glutamation. One possibility is that Hfq interacts with RimK at the ribosome, and the loss of the RimK protein is a major factor in destabilising Hfq function in the ∆**rimK mutant. We have modified figure 12 (now figure 10) and expanded the discussion to include this hypothesis.

While we cannot exclude the possibility that RimK has other cellular targets in SBW25, we think this is unlikely to be a major cause for the results we see here. We have carefully examined the C-terminal peptides of proteins detected in our various proteomic assays and are confident that RpsF is the sole target of RimK in SBW25 under the conditions we tested. We also directly tested RimK interaction with purified Hfq and confirmed that Hfq is not a direct target of RimK modification.

Lines 377-76: '...distinguishing features in the primary or predicted secondary structures of the Rim-mRNAs...' As mentioned already above several indirect options are still open that could confer selectivity to the ribosome.

As stated above, the discussion has been rewritten to more completely reference the possible mechanisms by which RpsF glutamation may lead to translational regulation.

Reviewer #1 (Significance (Required)):

The key concept of the manuscript namely the impact of the intricate RimABK pathway in Pseudomonas fluorescens on translational reprogramming by controlled modification of ribosomal protein S6 in response to environmental signals is novel and will significantly impact the field.

---

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

**Summary**

The main question addressed by this research is how bacteria adapt to rhizospheric niche through the RimK ATPase glutaminase. This enzyme post-transcriptionally modifies the ribosomal protein RpsF in a process of complex regulation. Regulation is mediated by c-di-GMP that is degraded by the phosphodiesterase RimA and the protease RimB exerts a role opposite of RimK. Novel findings include the finding of RimK acting as a four-state ATPase, depending on the binding of RimA, c-di-GMP or both. Another important finding is the opposite roles of RimK and RimB on the glutamation/deglutamation of RpsF and the tendency to a steady state of four glutamate residues in the RpsF protein. The authors also use proteomics to determine the effect of glutamation, specially at low temperature and under nutrient limitation.

We thank the reviewer for their positive review of the manuscript and address their comments below.

**Major comment**

In my opinion, the results obtained with the Hfq regulation by RimK blur the message. I firmly think that the Ms is very solid with the results obtained in relation with the RimABK/RpsF regulation in P. fluorescens shown as a model in the Figure 12. Moreover, in this final model presented by the authors (fig. 12) they not included the results related with Hfq. These results could be part of another paper.

We agree with the reviewer that the Hfq independent effects of RpsF are an exciting finding and should be a major focus of the paper. That said, we feel that the additional work we have done showing how Hfq is affected by RimK should also be retained in the manuscript in some form. Our data (e.g. figure 8) indicate that Hfq is responsible for a large (indirect) fraction of the ∆rimK phenotype, so understanding how it is affected is important to understand how RimK functions. Based on comments from reviewers 2 and 3 we have reviewed the manuscript text (including data on Hfq) to make the narrative as focussed and clear as possible. We have also redesigned figure 12 (now Fig 10) to consider comments from all three reviewers and have changed the text in the discussion to match this.

**Minor comments**

In figure 4A, what is lane 5?

Lane 5 contains RimB without ADP. The figure legend has been modified accordingly, and we thank the reviewer for highlighting this error.

Line 159 change "suppression of RimK band-shifting" by "suppression of RpsF band shifting"

This has been fixed.

Reviewer #2 (Significance (Required)):

The Ms. is very interesting and deeply describes the relation between environmental conditions, c-di GMP second messenger and the RpsF ribosomal protein posttranscriptional modification in order to respond to low temperatures and changes in nutrient availability. The research developed in this manuscript is original and novel in the field and includes new advances in the signal transduction pathways implicated in the regulation of bacteria adaption to the environment. Besides, the research design and technical methodology is original and includes multidisciplinary approaches of interest to the research community in general.

---

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

**Summary** Post-transcriptional control of protein abundance is an important yet poorly examined regulatory process by which bacteria respond to signals found in the environments they grow in. The authors' team have previously identified, described and published details around a novel regulatory pathway involving the ribosomal modification protein RimK, regulator proteins RimA and RimB, and the widespread bacterial second messenger cyclic-di-GMP (cdG). In the current manuscript builds considerably on those previous findings and goes several steps further, through various approaches including protein biochemistry, computational modelling, quantitative proteomics and ribosomal profiling, the authors described how the RimABK pathway as a novel translator system that controls bacterial adaption to the rhizosphere in the bacterium Pseudomonas fluorescens. They show that the system achieves this through specific controlled modification of the ribosomal protein RpsF. I read the article with excitement and overall the manuscript describes an extensive data set that will be of considerable interest to many readers in several fields. However, I have made a few points below that the authors need to take on board and address. If these issues are addressed, I believe it will make the presented data much clearer to the reader, tidy up a few ambiguities and make the article a little more accessible to many non-specialist readers.

We thank the reviewer for their thorough and positive assessment of the manuscript. We address their specific points below.

**Major Comments**

1) The major finding described in the manuscript and the one that will be of significant interest to reader is that a novel post-translational ribosomal modification regulatory mechanism involving Rim system controls bacterial adaption. The second messenger cdG only plays a small part in this complicated process. Therefore, I believe the title needs to be revised to capture the scope and key findings of the manuscript.

We are happy to change the title along the lines suggested by the reviewer. We propose: Control of mRNA translation by dynamic ribosome modification as a new title.

2) The authors present a lot of interesting data; however, I found the manuscript a bit of a dense read. I find the key findings are diluted within the text. I would ask that the authors to make it a little more focused. For example, on the regulatory role of RimK and its influence on Hfq and RpsF has been detailed previously so could be placed in supporting information and briefly mention when required. Also, the experiments on the pvdIJ pathway could be removed or placed in supporting information as they are not the main focus of the manuscript. Fig 5 and 6 could be combined as one figure as well.

We have modified the manuscript throughout to make it clearer, more concise, and to focus as much as possible on new findings rather than reiterating what we showed in our last manuscript. In line with the reviewers’ recommendation, we have moved the pvdIJ data into the supplementary material (Fig S3) and merged figures 5 and 6 into one. In addition, to support our data on the importance of RpsF glutamation for ribosomal regulation we used Western blotting to confirm that RpsF4/10glu variants incorporate normally into SBW25 ribosomes in vivo (added as supplemental data Fig S5).

As stated elsewhere, we feel that key data on the relationship between Hfq and RimK should remain in the main manuscript, although we have reviewed the text thoroughly to try to ensure it is as focussed as possible and have moved some results to supplementary material as suggested.

3) The authors propose a four-state kinetic model for RimK ATPase activity with RimA and cdG (described in Fig2 and Table S1). However, later in the manuscript the authors demonstrate that RimB also stimulates RimK ATPase activity, but this seems to have smaller impact than RimA and cdG (Fig 2E, Fig 3A). Why RimB was not included in the ATPase kinetic model of RimK? Does including the RimB data suggest there might be more conformational states for RimK?

Thank you for raising this point. The reviewer is correct in that this data does indeed suggest another level of ATPase activity of RimK. We have added text to the manuscript to reflect this. We have also extended the supplemental Table S1 to include these equations.

4) The authors claim that the suppressive effect of cdG on RimK was depended on the enzyme activity (PDE domain) of RimA. This was tested using an enzymatically inactive RimA variant (RimA-E47A). However, in Fig 3E the amount of RimA-E47A used in the assay seems to be significantly less than wildtype RimA. Additionally, in Fig 2B, the authors show that addition of cdG also stimulates RpsF modification with or without RimA (lane 4-6). I would ask the authors to clarify these points.

It is difficult to directly compare protein variants due to differences in solubility post-purification. Due to difficulties in purifying this (less soluble) form of RimA, co-purifying contaminants have also probably influenced the determination of RimA-E47A concentration to some extent. This restricts us to making largely qualitative statements about protein function, as we do here. Despite its poor solubility and low concentration, RimA-E47A is still able to stimulate RimK. Furthermore, the relatively low concentration of RimA-E47A in our assays would render it at least as susceptible to any effects of cdG addition as WT RimA, meaning we can be confident that cdG has no effect on RimK stimulation by this variant.

Our model incorporates direct stimulation of RimK by cdG alongside its effect on RimA. We show evidence for this in this manuscript and in our 2016 paper.

5) The authors claim that high levels of cdG increase the ratio of RimB protease activity to RimK glutamate ligase activity. However, there is no experiment to provide direct evidence to support this. Please tone down the language used or provide evidence. On the same point Fig 6 was not explained in the main text to support this conclusion. Please include an explanation.

The hypothesis that high cdG levels favour RimB activity over RimK stems from the observation that cdG suppresses RimK activity (by abolishing RimA stimulation) but does not affect RimB. We have data showing that increasing cdG levels suppresses RpsF band shifting in vitro in an assay containing all three Rim proteins (Fig 4). However, we agree the hypothesis that cdG controls the ratio of RimB to RimK activity by controlling the activity of RimK currently lacks explicit, direct evidence and we have modified the text to tone down the language.

An explanation for Figure 6 (now 5b) has been added to the manuscript as requested.

6) In some of the figures/images, for example, Fig2B and Fig 3E, RimA is shown as a major band. However, in other figures/images, for example, Fig 2D, Fig 3D, RimA seems to be two bands. The authors should explain the reason for this.

Based on extensive experimentation, we are confident that the second band present in some of our assays is a cleavage product of RimA. This is an experimental artefact that is linked to concentration and protein stability in vitro. We must stress that the presence of an inactive fraction of RimA in our assays does not affect the conclusions we are able to draw from these experiments. A note has been added to the relevant section of the text.

**Minor comments**

• Line 151, should be RpsF band-shifting instead of RimK.

• Fig 4A there is no legend for lane 5, which made it very difficult to understand the data presented.

Please see above. These two minor errors will be fixed.

• The layout of some figures could be improved.

We have revised the layout of several figures, in line with the reviewer’s suggestion.

• If it is possible to have Fig 11 as a Venn diagram or some intuitive diagram, it will help the readers gain access to the data and understand the results.

We respectfully disagree with the reviewer here. We have tried several different presentation styles for these data, but ultimately considered scatter charts to be the most effective, in line with our previous study of Hfq regulation in Pseudomonas (Grenga et al. Frontiers in Microbiology 2017).

Fig 12 is very neatly laid out. However, I don't feel it captures the dynamic nature of the system. I am just wondering if the authors could break it down so that it describes the changes relating to environmental conditions and/or different cdG levels?

Figure 12 (now Figure 10) has been modified to reflect to comments of all three reviewers.

Reviewer #3 (Significance (Required)):

The manuscript provides detailed evidence to demonstrate a dynamic, post-translational ribosomal modification mechanism which is an important feature of prokaryotic (potentially archaeal and eukaryotic) environmental adaptation. This is an exciting manuscript and one many will wish to read. The data provided will be of interest to scientists working in many fields including microbiology, biochemistry and plant pathology.

I have several areas of expertise including genomics, molecular microbiology, small molecule signalling and regulation, micro-host interaction, adaptation,

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

Post-transcriptional control of protein abundance is an important yet poorly examined regulatory process by which bacteria respond to signals found in the environments they grow in. The authors' team have previously identified, described and published details around a novel regulatory pathway involving the ribosomal modification protein RimK, regulator proteins RimA and RimB, and the widespread bacterial second messenger cyclic-di-GMP (cdG). In the current manuscript builds considerably on those previous findings and goes several steps further, through various approaches including protein biochemistry, computational modelling, quantitative proteomics and ribosomal profiling, the authors described how the RimABK pathway as a novel translator system that controls bacterial adaption to the rhizosphere in the bacterium Pseudomonas fluorescens. They show that the system achieves this through specific controlled modification of the ribosomal protein RpsF. I read the article with excitement and overall the manuscript describes an extensive data set that will be of considerable interest to many readers in several fields. However, I have made a few points below that the authors need to take on board and address. If these issues are addressed, I believe it will make the presented data much clearer to the reader, tidy up a few ambiguities and make the article a little more accessible to many non-specialist readers.

Major Comments:

1) The major finding described in the manuscript and the one that will be of significant interest to reader is that a novel post-translational ribosomal modification regulatory mechanism involving Rim system controls bacterial adaption. The second messenger cdG only plays a small part in this complicated process. Therefore, I believe the title needs to be revised to capture the scope and key findings of the manuscript.

2) The authors present a lot of interesting data, however, I found the manuscript a bit of a dense read. I find the key findings are diluted within the text. I would ask that the authors to make it a little more focused. For example, on the regulatory role of RimK and its influence on Hfq and rpsF has been detailed previously so could be placed in supporting information and briefly mention when required. Also, the experiments on the pvdIJ pathway could be removed or placed in supporting information as they are not the main focus of the manuscript. Fig 5 and 6 could be combined as one figure as well.

3) The authors propose a four-state kinetic model for RimK ATPase activity with RimA and cdG (described inFig2 and Table S1). However, later in the manuscript the authors demonstrate that RimB also stimulates RimK ATPase activity but this seems to have smaller impact than RimA and cdG (Fig 2E, Fig 3A). Why RimB was not included in the ATpase kinetic model of RimK? Does including the RimB data suggest there might be more conformational states for RimK?

4) The authors claim that the suppressive effect of cdG on RimK was depended on the enzyme activity (PDE domain) of RimA. This was tested using an enzymatically inactive RimA variant (RimA-E47A). However, in Fig 3E the amount of RimA-E47A used in the assay seems to be significantly less than wildtype RimA. Additionally, in Fig 2B, the authors show that addition of cdG also stimulates RpsF modification with or without RimA (lane 4-6). I would ask the authors to clarify these points.

5) The authors claim that high levels of cdG increase the ratio of RimB protease activity to RimK glutamate ligase activity. However, there is no experiment to provide direct evidence to support this. Please tone down the language used or provide evidence. On the same point

Fig 6 was not explained in the main text to support this conclusion. Please include an explanation.

6) In some of the figures/images, for example, Fig2B and Fig 3E, RimA is shown as a major band. However, in other figures/images, for example, Fig 2D, Fig 3D, RimA seems to be two bands. The authors should explain the reason for this.

Minor comments

1) Line 151, should be RpsF band-shifting instead of RimK.

2) Fig 4A there is no legend for lane 5, which made it very difficult to understand the data presented.

3) The layout of some figures could be improved.

4) If it is possible to have Fig 11 as a Venn diagram or some intuitive diagram, it will help the readers gain access to the data and understand the results.

5) Fig 12 is very neatly laid out. However, I don't feel it captures the dynamic nature of the system. I am just wondering if the authors could break it down so that it describes the changes relating to environmental conditions and/or different cdG levels?

Significance

The manuscript provides detailed evidence to demonstrate a dynamic, post-translational ribosomal modification mechanism which is an important feature of prokaryotic (potentially archaeal and eukaryotic) environmental adaptation. This is an exciting manuscript and one many will wish to read. The data provided will be of interest to scientists working in many fields including microbiology, biochemistry and plant pathology.

I have several areas of expertise including genomics, molecular microbiology, small molecule signalling and regulation, micro-host interaction, adaptation, virulence and pathogenies.

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 main question addressed by this research is how bacteria adapt to rhizospheric niche through the RimK ATPase glutaminase. This enzyme post-transcriptionally modifies the ribosomal protein RspF in a process of complex regulation. Regulation is mediated by c-diGMP that is degraded by the phosphodiesterase RimA and the protease RimB exerts a role opposite of RimK. Novel findings include the finding of RimK acting as a four-state ATPase, depending on the binding of RimA, c-diGMP or both. Another important finding is the opposite roles of RimK and RimB on the glutamation/deglutamation of RpsF and the tendency to a steady state of four glutamate residues in the RspF protein. The authors also use proteomics to determine the effect of glutamation, specially at low temperature and under nutrient limitation.

Major comment:

In my opinion, the results obtained with the Hfq regulation by RimK blur the message. I firmly think that the Ms is very solid with the results obtained in relation with the RimABK/RpsF regulation in P. fluorescens shown as a model in the Figure 12. Moreover, in this final model presented by the authors (fig. 12) they not included the results related with HfQ. These results could be part of another paper.

Minor comments:

In figure 4A, what is lane 5? Line 159 change "suppression of Rimk band-shifting" by "suppression of RpsF band shifting"

Significance

The Ms. is very interesting and deeply describes the relation between environmental conditions, c-di GMP second messenger and the RfsF ribosomal protein posttranscriptional modification in order to respond to low temperatures and changes in nutrient availability. The research developed in this manuscript is original and novel in the field and includes new advances in the signal transduction pathways implicated in the regulation of bacteria adaption to the environment. Besides, the research design and technical methodology is original and includes multidisciplinary approaches of interest to the research community 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

General comments

The manuscript 'Second messenger control of mRNA translation by dynamic ribosome modification' is a very interesting follow up on the research performed by the authors published in 2016. Here, the authors continue their study by determining the impact of the intricate RimABK pathway in Pseudomonas fluorescens on translational reprogramming by controlled modification of ribosomal protein S6 in response to environmental signals. The manuscript is interesting and well written, and the results are sound. However, in my opinion the general conclusion is not supported by experimental data and leaves several potential explanations open. Thus, I suggest to either perform in vitro translation experiments using ribosomes equipped with glutamated S6 to validate translational selectivity, or to soften the language on the working model shown in Figure 12.

Specific comments

Figure 1 and S1: The RT-PCR analysis shown here does not allow excluding transcription initiation at alternative promoters downstream of the one determined by 5'-RACE. However, an alternative promoter might contribute to relative ratios between the rimA, rimB, and rimK mRNAs. A Northern blot and/or primer extension analysis could clarify this assumption and would give more detailed insights into the specific regulation.

Figure 2B: I'm confused by the results shown here! I do only see a reduction of RpsF in the presence of RimA, RimK and cdG. What indicates the modification? Please, explain the interpretation of the result in more detail. Shouldn't the modified RpsF shift due to the addition of glutamate residues?

Figure 2C: Why does the RpsF modification lead to a supershift? How many glutamate residues are added? Is the smear visible in lane 4 (RpsF+RimK) representing already the slightly modified RpsF protein, which upon addition of RimA results in a supershift? For all SDS-Page analyses shown in the manuscript the validation of the glutamation using the antibodies specific against poly-glutamate would be a great asset to facilitate their interpretation.

Lines 236-238: '...strongly suggesting that th e proteomic changes we observe are an active response to modification of ribosomally-associated RpsF proteins.' This is an important suggestion as it allows a flexible and very fast integration of the external signals into a specialized protein synthesis. Thus, it definitely deserves further analysis! Considering that the purified RimA and RimK proteins are available, in vitro modification of RpsF in the context of the purified ribosome would be an important experiment and would greatly increase the quality of the paper. Up to now the selective or specialized translation is pure hypothesis and might also be explained by indirect effects via e.g. increased interaction between the ribosome and HFQ that might mediate interaction with certain mRNAs and thus stimulate their translation.

Lines 322: '...into a single output: the proportion of all ribosomally-associated RpsF proteins that have C-terminal poly-glutamate tails.' Considering the identification of a group of genes whose translation is altered by rimBK deletion, but not by RpsF glutamation (Class 1, Fig 11B), I would suggest to soften this statement. If I interpret the data correctly, they pinpoint to a moonlighting function of the rim-pathway that does not target RpsF!

Lines 377-76: '...distinguishing features in the primary or predicted secondary structures of the Rim-mRNAs,...' As mentioned already above several indirect options are still open that could confer selectivity to the ribosome.

Significance

The key concept of the manuscript namely the impact of the intricate RimABK pathway in Pseudomonas fluorescens on translational reprogramming by controlled modification of ribosomal protein S6 in response to environmental signals is novel and will significantly impact the field.

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 study by Wegwitz et al, the authors examine the tumour promoting properties of RNF40 (and the H2B monoubiquitinylation catalysed by it) in Her2 driven breast cancer.

They report, using publicly available data, that increased RNF40 expression is associated with reduced overall and disease-free survival.

Using a mouse model, where they crossed the Erbb2 (mouse Her2) under the control of the MMTV promoter with conditional Rnf40 deletion constructs, the authors found that deletion of Rnf40 simultaneous to Her2 overexpression resulted in a prolonged tumour-free survival, somewhat reduced tumour growth kinetics and tumour incidence.

siRNA silencing of Rnf40 in two Her2 positive breast cancer cell lines resulted in reduced proliferation, clonogenicity and tumour sphere formation and cellular motility.

Transcriptome analysis revealed pathways that could explain the phenotype, like increased apoptosis and actin cytoskeleton regulation. The authors then took further some candidates in the later pathway to investigate the mechanism. They find that Rnf40 loss impacts on actin cytoskeletal dynamics. They also investigate the impact on focal adhesion signalling integrity.

Finally, they investigate the relationship between the transcriptome and H3K4me3 and H2Bub1 landscape in the presence or absence of Rnf40.

The manuscript is convincing regarding the tumour promoting roles of Rnf40, but the key claim that H2B monoubiquitinylation is essential for activation of the Rho/Rock/Limk pathway, where genes are down regulated upon Rnf40 loss resulting in decreased tumourigenicity of cells, is so far not convincing.

"Together, these findings support the hypothesis that the actin regulatory gene network is dependent on direct epigenetic regulation by RNF40 through modulation of H2Bub1 and a trans-histone cross-talk with H3K4me3 levels in HER2-positive BC cells."

Although the correlation is apparent, at this point it's unclear if the phenotype is dependent on the catalytic activity of Rnf40 or it's a non-catalytic effect. Generating a catalytic mutant RNF40 and test it at least in the cell lines studied would be desirable.

We thank the reviewer for this comment and agree that the addition of data with a catalytic mutant RNF40 could strengthen our findings and further clarify mechanisms involved. Thus, in a resubmission we will directly address this point by performing knockdown/rescue experiments with either wildtype or a RING finger mutant RNF40. This will be done by transfecting cells with expression constructs for either wildtype or mutant RNF40 proteins followed by knockdown of endogenous RNF40 using siRNAs targeting the 3’ UTR. Experiments central to our take-home message will be performed (e.g., cell migration, target gene expression, Western blot for H2Bub1, F-actin formation). Together, we hope these experiments will help significantly solidify the message of this paper and further clarify the individual role of RNF40 within the RNF20/40 heterodimer.

**Other comments that need a response:**

1."we investigated RNF40 expression and H2Bub1 levels by immunohistochemical staining of 176 primary BC tumors and 78 brain metastases."

In Fig 1 I can only count 41 primary BC tumours and 73 brain metastases. Numbers don't add up. Also, how is "low" defined as opposed to negative? What is used as controls?

We apologize for this mistake. We corrected the numbers of primary and metastatic HER2-positive specimens used in this study.

2."Moreover, HER2-positive metastatic BC samples showed a particularly high expression of RNF40 compared to primary tumors"

Figure 1 or Fig S1A does not contain data on HER2-positive metastatic BC

We think there might have been a confusion regarding this point. The manuscript does provide information about RNF40 and H2Bub1-staining in primary HER2-positive breast cancer lesions as well as HER2-positive brain cancer metastasis specimens in Fig.1A-C as well in Fig.S1A (representative brain metastases are shown in IHC pictures). This is stated both in the main text as well as in the respective figure legends. However, if for some reason this remains unclear, we would certainly be open to suggestions as to how we can modify the respective sections to improve their clarity.

3."tumors did not display a loss of either RNF40 or H2Bub1 (Fig. 1H) when compared to the adjacent normal mammary epithelium (Fig. S1F)."

I don't understand what I see in Fig S1F, where is the tumour, what is adjacent?

We agree with the reviewer that splitting tumor staining in the main Fig 1 and normal tissues in Fig S1 makes a comparison difficult. We will therefore edit the Fig S1F and provide there an overview of tumor and surrounding normal tissues together with magnifications of the respective areas. This should significantly ease a comparison of both RNF40 and H2Bub1 in tumor and adjacent normal tissues.

4."homozygous loss of Rnf40 (Rnf40fl/fl) resulted in dramatically increased tumor-free survival of MMTV-Erbb2 animals (Fig.1E)." This is overinterpretation of the data, I would not call it dramatic, just significant.

The MMTV-Erbb2 mouse model is a very reliable mouse model for the induction of HER2-positive lesions. In our hands, the tumor incidence in these animals was 100% with a median tumor free survival of 166 days. In comparison, approx. 20% of the Rnf40fl/fl animals (3 out of 14) never developed the disease during the 18 month observation. The animals that still developed lesions had a median tumor free survival of 241 days, which represents a delay of 75 days (45% delay). In light of this, it seems to us that the effect of RNF40 loss on HER2–positive lesions is, indeed, remarkably strong. However, we do not wish to give an impression of over-interpreting or misrepresenting the data. For that reason we modified the wording in the main manuscript according to the reviewer’s suggestion (line 140: “dramatically” was replaced with “pronounced”).

5."loss of Rnf40 led to strongly reduced tumor growth kinetics (Fig.1G)." Is this result significant, I did not see an evaluation of statistical significance in this data.

As suggested by the reviewer, we have performed additional analyses to examine the statistical significance. We have now included the results of these tests in the respective figure.

6."Rnf40fl/fl lesions displayed a heterogeneous pattern of RNF40 expression (Fig.1H), suggesting that the few tumors that did develop in this model were largely caused by an incomplete loss of the Rnf40 allele." If this conclusion is suggested, the authors should check if the "escaper" cells have failed to flox the Rnf40 allele on the genetic/protein level. Otherwise it's not conclusive.

The reviewer brings up an interesting and important point about the heterogeneous loss of RNF40 in “escaper” tumors. Very important to note is that these “escaper” tumors developed significantly later and three animals never developed tumors. Thus, the “escaper” phenotype is rare (at the cellular level) and is likely similar to the selective process that occurs in human tumorigenesis and tumor progression. It is well established through a number of publications that deletion of genes essential for tumorigenesis via Cre-based systems frequently results in a specific selection for the rare instance that the Cre-mediated excision is ineffective. These “escaper” cells can then grow out and proliferate because they do not suffer from deletion of the floxed allele. This effect has also been established when combining MMTV-HER2 and MMTV-Cre. For example, analogous findings were recently published by Costa, et al., in Nature Communications (doi: 10.1038/s41467-019-11510-4) where the MMTV-Cre-mediated deletion of Pak4 resulted in impaired MMTV/HER2 or MMTV-PyMT-driven tumorigenesis, but occasional tumors did appear, which all retained some degree of PAK4 expression. This effect, which we have also seen in our system, was also reported by Miao, et al. in Cancer Research (doi: 10.1158/0008-5472.CAN-11-1015) in 2011. In their work the authors observed that deletion of the Myb gene also impaired MMTV-HER2-driven tumorigenesis and those tumors that developed in Myb flox/flox mice displayed a late onset and invariably retained MYB expression. Similar findings have been reported in a number of other tumor types and with various Cre drivers. Thus, we posit that these findings provide further support for the essential role of RNF40 in HER2-driven tumorigenesis to the extent that rare, RNF40/H2Bub1-expressing “escaper” cells are positively selected for during tumorigenesis and tumor progression.

In order to definitively establish this, we propose performing dual immunofluorescence staining of Rnf40 flox/flox tumors to verify that H2Bub1 is exclusively and universally lost together with RNF40 and that each case of a complete loss of RNF40 also results in a complete loss of detectable H2Bub1 staining. Additionally, we will assess the efficiency of the cre mediated deletion of Rnf40 exons 3 and 4 in Rnf40fl/fl animals by detecting their presence using a conventional PCR approach.

1. Fig S4D - is this clonogenic assay? How many replicates were done, biological technical?

We apologize for the imprecise description of this figure. We edited the manuscript accordingly and included details about the number of replicates.

8."Additionally, treatment with either CYM-5441 (Fig.4J) or …"

Fig 4J is missing! It makes this section rather hard to follow. Fig S4F-G, how many replicates were done, biological technical?

We thank the reviewer for noticing this error. The figures were indeed inappropriately referenced in the text. This error has been corrected.

9."Consistent with our analyses based on changes in H3K4me3 occupancy, genes downregulated upon RNF40 silencing displayed the most prominent decrease in H3K4me3 in the gene body (the 3' end of the peak)"

The impact of these mods changes is hard to judge because they are rather small (I would not use the wording prominent).

As implied by the reviewer, we will replace the word “prominent” with “noticeable”.

Also, are there many other "peak narrowing" genes but they are not downregulated?

The point mentioned here is very interesting. The bioinformatic analyses performed in this study solely focused on the relationship of significantly regulated genes and the H3K4me3 peak narrowing at their TSSs. However, we did not analyze the global regulation of genes showing H3K4me3 peak narrowing near the tSS. As this information might be of high relevance for this study, we propose to investigate this interesting aspect in the revised version of the manuscript.

In fact, our analyses have revealed that a large fraction of genes with H3K4me3 narrowing peaks do not show an appreciable decrease in expression. To better understand the epigenetic features determining the sensitivity of genes to H3K4me3 peak narrowing, we studied the occupancy of several histone marks at differently behaving genes. We discovered that sensitive genes globally present lower occupancy of histone modifications which are known to positively influencing gene transcription. We therefore propose that the epigenetic context (i.e., presence of additional histone modifications) strongly determines whether the loss of H2Bub1, and ensuing narrowing of H3K4me3 near the TSS, results in decreased transcription.

Statistical analysis missing: for example in Fig 2C, Fig 2E, Fig 3G what is n=?, how many technical, biological replicates were analysed?

This information has been added to the revised manuscript.

Fig 4E seems to be a partial duplication of Fig 3D!

The samples in Fig. 3D and 4E originate from independent experiments. In Figure 4D, we indeed provide again PARP and Casp3 signal for siRNF40 samples in order to allow a direct comparison of the effect magnitude between RNF40 depletion and ROCKi treatment.

**Minor:**

Figure referencing: it can be quite confusing to see a different ordering of figures compared to the referencing in the manuscript, for example Fih 1H is referenced in the text before Fig 1F, G. The authors should change the order in the main figures....

We thank the reviewer for pointing this out. We have updated the figure order in the revised manuscript accordingly.

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

In this study, Wegwitz et al propose that the E3 ubiquitin ligase RNF40 is highly expressed in HER2+ breast cancer tumours and correlates with poorer survival, using their own and TCGA data. Contrary to observations suggesting a tumour-suppressive role in other cancers, authors show using RNF40-knockout breast cancer mouse models and in vitro data shat RNF40 promotes tumour growth. RNF40 depletion impairs proliferation, survival and sphere formation by inducing apoptosis. In addition, RNF40 promotes cell migration by upregulating expression of cytoskeletal proteins (ROCK1, VAV3, LIMK2) and their effectors such as phosphorylated cofilin. Authors show elegant partial rescue experiments of the effect of RNF40 depletion on apoptosis and survival.

Given that RNF40 function seems to be context-dependent, findings from this study could have broad significance for other cancers with high RNF40. It also provides some mechanistic data (that should be improved as suggested below) linking this ubiquin ligase to the cytoskeletal machinery and, therefore, control of migration and also proliferation and survival.

Data are well presented and most conclusions are supported by the data. However, there are some gaps at the mechanistic level. Since migration is controlled by RNF40 in vitro, evaluation of metastatic ability in vivo (local invasion for example as suggested below) should be evaluated and would strengthen this part too.

**Major comments**

1. Fig.1A-B, S1A. Specificity of RNF40 antibody should be shown, which could be done quite easily in the tumours from the knockouts. From the datasheets, antibodies recognize human protein only.

We thank the reviewer for this suggestion and apologize for this mistake. The antibody utilized in the IHC studies is actually from Abcam (ab191309) and, in fact, recognizes both species. Table S5 has been corrected accordingly.

It is unclear when the murine tumours were analysed, at endpoint? This should be stated.

We thank the reviewer for this comment. Indeed, all IHC analyses were performed after dissection (endpoint). This information will be added to the manuscript. Kinetic analyses of tumor growth (i.e., Fig. 1G) were performed on the same mouse cohort.

Could authors establish cell lines from the mouse tumours (knockout, partial knockout escapers..)? These could be very useful tools to evaluate key in vitro findings from the study.

The reviewer makes an interesting suggestion. Unfortunately, we were not able to establish cell lines from this model and have since stopped breeding this mouse line (due to the relocation of the principle investigator). However, we did try to generate RNF40-deficient breast cancer cell lines using the CRISPR/Cas9 technology. Interestingly, all attempts failed, supporting the fact that the loss or RNF40 is lethal for the cancer cells. However, to further establish this, for the revision we propose to transfect HCC1954 cells with CRISPR/Cas9 constructs targeting exons 3 and 4, similar to our mouse model. We will then assess the evolution of RNF40-negative cells population over time (i.e., via immunofluorescence staining for H2Bub1). This assay should inform about the expected growth “disadvantage” following RNF40 loss.

Fig.1F-G: since RNF40 controls the cytoskeletal machinery and therefore, migration (Fig. 2G) in the RNF40 knockout tumours, was metastasis (if observed) affected? Or if there was no growth in distant organs detected in the time frame of these experiments, was invasion (and/or pattern of invasion or mode of invasion (morphology of invading cells)) into adjacent tissues affected upon RNF40 depletion? This would add in vivo relevance to the in vitro mechanistic findings, especially since the authors later showed that p-cofilin was also decreased in the RNF40-depleted mouse tumours (Fig.4D).

We agree that the metastasis data from our mouse genetic tumor model would significantly help solidify our findings. Unfortunately, the MMTV-Erbb2 mouse model (overexpressing wildtype Erbb2 gene) only rarely develops distal metastases. In our analyses, we only ever observed two macroscopically visible metastases (one wt/wt and one in an Rnf40flox/flox mouse). However, we feel that the reviewer’s suggestion is a one and will follow this suggestion and attempt to examine possible changes in local invasion of primary tumors into adjacent tissues.

Fig.3: most results using HCC1954 cell line. Key findings should be validated in other cell lines.

We agree with the reviewer about the importance of cross validation of findings using different cell lines. For this purpose, we have now generated data with an additional HER2-positive cell line. These data using the SKBR3 cell line were performed for several of the key experiments. Key findings about phenotypic changes (growth kinetics and colony formation), Ki67 protein levels differences and mRNA regulation of identified regulators of actin cytoskeleton (VAV3, ROCK1, LIMK2 and PFN2) will be included in the revision for both cell lines. Furthermore, as seen in the HCC1954 cell line, an increase of the apoptosis marker cleaved PARP as well as a loss of VAV3 and ROCK1 protein levels was also observed upon RNF40 knockdown in SKBR3 cells. These data will be included in the revised manuscript.

Fig.3A: authors state "both pathways remained intact following RNF40 depletion". However, from those blots, siRNF40 clearly increases pERK and slightly pAKT, which would be unexpected according to previous data in Fig.2. Authors could show quantifications of different blots, or show a more representative blot if increase in pERK was not consistently observed. Was this also seen in SKBR3 cell line?

We thank the reviewer for this comment. Initially, we had anticipated that oncogenic signaling may be decreased in the Rnf40 conditional knockout model. However, much to our surprise, the activity of the downstream signaling actually appears to be increased. In fact, the increase in AKT and ERK1/2 phosphorylation following RNF40 silencing was consistent across different experiments and replicates. While this finding is also consistent with our previous results in an ER-positive system (e.g., see Prenzel, et al., 2011), we do not understand the mechanistic underpinnings of this finding. Importantly though, while consistent, we do not feel that this increase explains the observed phenotype. Nevertheless, to more precisely show the overall change of p-ERK/ERK and p-AKT/AKT, in the revision we will provide a densitometry quantification for both cell lines. We will also modify the sentence to more precisely describe this finding and make the point that since these pathways are not reduced/impaired, they are unlikely to be responsible for the increased apoptosis observed upon RNF40-KD. Western blots assessing p-ERK/ERK and p-AKT/AKT levels in SKBR3 upon RNF40 knock-down will also be added into the supplementary data of the revised manuscript (Fig.S3).

For Fig.3G and Fig.S3A, authors selected genes from this set, how was this done (fold change?). Was expression of the other family members (ROCK2, LIMK1, etc) or of Rho GTPases regulated too?

This information was indeed missing in the manuscript. We have modified the figure legend and the main text accordingly in order to provide the information about the cutoff used in the Enrichr analysis. Regarding the expression of other family members of the actin regulatory network, in the past we performed a more, in depth and focused analysis of our RNA-seq data, broadening our view to other members of the RHO/RAC/CDC42 pathways. While we did identify a few further potentially regulated target genes (e.g. ROS1 or PAK6), these genes were either only weakly expressed or weakly regulated. For this reason, we presumed that these factors could only play a marginal role in the observed phenotype and have focused our attention on the robust part of the signature.

Fig.4B: this may not help, decrease of p-cofilin by Vav3 knockdown is way less dramatic compared to RNF40 depletion or ROCK inh treatment. See comment below regarding other effectors such as Myosin.

Indeed, the consequence of VAV3 loss on p-cofilin levels is less pronounced than the effect observed upon RNF40 knockdown or ROCK1i treatment. Given the fact that RNF40 loss not only affects VAV3 expression, but also has additional direct effects on the expression of other pathway members, this may be expected. We do, however, feel that the VAV3 regulation is likely one component of the effects of RNF40 loss. In addition, it has also been shown that VAV3 is not the only GEF regulating the activity of RHO kinases upstream of ROCK1. Therefore, we would also expect that VAV3 loss only partially reduces ROCK1 activity and therefore only partially phenocopies the effects observed. We will expand the description of these findings in the revised manuscript to reflect these views.

Fig.4C: does ROCK inh reduce RNF40 levels? It may from the immunofluorescence picture.

We thank the reviewer for this comment. In fact, we have examined this possibility. However, no significant changes in RNF40 protein levels were observed upon RKI-1447. If helpful, we can provide Western blot data demonstrating this in the supplemental figures.

Fig.4H-I: the sphingosine 1-phosphate receptor-3 agonist (CYM-5441) partially rescued the effects of RNF40. Since S1P signalling involves Rho GTPase activation -presumably downstream of VAV3 -which is a GEF for Rho, Rac and Cdc42- and upstream of ROCK, LIMK, was activity of these Rho GTPases affected upon RNF40 depletion? This would strengthen the mechanism.

The reviewer points at an interesting aspect of the actin regulation. Indeed we expect that the reduction of VAV3 levels upon RNF40 loss would significantly influence the activity of the downstream client GTPases. However, the measurement of RHO-GTPase activity is tricky and expensive. Furthermore, as mentioned in the previous comment (#7, part 1) VAV3 is only one component of the four major genes encoding critical actin cytoskeleton regulatory proteins regulated upon RNF40 loss, and the only factor upstream of RHO-GTPases. The reduction of downstream ROCK1, LIMK2 and PFN2 levels also influence the activity of this pathway downstream of RHO-GTPase activity. We therefore focused our efforts on assessing F-actin and p-cofilin levels as these may provide more sensitive readouts about the consequence of RNF40 loss on this signaling cascade. However, if the reviewer considers this information as indispensable, we would attempt to investigate changes in Rho-GTPase activity by commercially available Active Rho Detection Kits, although this will significantly delay the resubmission of a revised manuscript.

Related to this, was Myosin II activity (phosphorylated MLC2) affected -since its upstream regulators, especially ROCK are controlled by RNF40?

We thank the reviewer for this insightful suggestion. To address this possibility, we will test this hypothesis for the revised manuscript as suggested by performing Western blot analysis for phosphorylated MLC2.

Fig. S5E:

Authors should consider presenting data of decreased histone methylation of cytoskeleton regulators in main Fig. 5, since this is an important conclusion of this part.

As suggested we will shift the information currently presented in figure S5E to the main figure 5.

Statistics should be revised throughout the manuscript. Comparisons of more than 2 groups should be performed with ANOVA or similar multiple comparison test (instead of t-test).

We thank reviewer for this comment. We replaced statistical tests with the appropriate ANOVA in the respective graphs and updated the legends accordingly.

**Minor comments**

Statement of significance mentions "Anti-HER2-therapy resistance", but this is a misleading since the paper does not deal with therapy resistance. Or are the cell lines used in the study resistant to anti-HER2?

We thank the reviewer for this suggestion. While resistance to anti-HER2 therapy remains one of the major clinical challenges in the treatment of HER2 positive BC lesions, we agree that our data do not strictly address this point. Thus, we have modified the sentence accordingly.

In line with this, authors could add some lines of thought on how RNF40 could be targeted in the clinical/pre-clinical context, which could inform further translational studies.

This is a great suggestion. In the revised manuscript we will include additional text to specifically address this point.

Line 117 "Moreover, HER2-positive metastatic BC samples showed a particularly high expression of RNF40 compared to primary tumors".

Perhaps rephrase, was it that the expression level (intensity) was higher or that the % of positive cells/tumours was higher in the brain mets?

This is a critical point that we will consider in the revised manuscript. We have modified the sentence accordingly to read, “Moreover, the incidence of RNF40-high specimens was higher in HER2-positive brain metastases compared to primary tumors (Fig.1A-B)”.

Fig.1D and S1C. while S1C shows TCGA data, it is unclear which set of patients is Fig.1D (since text says publicly available data, line 118-119), are these their own set of patients (used in Fig.1A-B)? This should be specified in the text, legend.

The origin of the data shown in Fig.1D for relapse free survival of RNF40high and RNF40low patients (KM plotter) is mentioned in the figure legend (kmplot.com) and in the material and method section. However, since this was not apparent, to increase the readability, we have now added a statement about the publically available database of origin for every output graph in the main text as well in the legend and supplementary material.

Line 122. Authors should be careful with this conclusion so far, a correlation between expression and cancer stage/survival does not necessarily mean a tumor suppressive/supportive role.

We thank the review for this comment. We agree with this statement. Therefore, we have carefully rephrased this sentence as following, “In summary, these data demonstrate that RNF40 expression is maintained in HER2-driven primary metastatic BC and that its expression correlates with poor prognosis in these patients.”

Fig. S4E: there are missing labels in graph (control, siRNF40).

The labels have been added.

Panels in some figures are discussed in text randomly and not following same order. For example, Fig.1 (after panel D, then panel H, then back to E, F, G), S3, 4A-C,

We reorganized the order of the different panels of Fig1 to increase the readability. We further screened the main text for similar problems and modified the respective figures accordingly.

Fig.1E: I would suggest changing the line colours, so Rnf40wt-wt line is red and the fl-fl is black, therefore it is similar to panel D (high Rnf40 red, low in black).

We thank the reviewer for this suggestion. Accordingly, we have now indicated low RNF40 expression in red (Figure 1D, 1E and S1B) in the same way that we have indicated RNF40 expression throughout the rest of the study.

Supp Videos: for reviewers and readers, it would help that video has a label while it plays, otherwise after downloading it, video name does not tell whether it is control or RNAi.

As suggested we have renamed the video files of for each condition and added a label informing about the identity of the sample while the video plays.

Reviewer #2 (Significance (Required)):

Given that RNF40 function seems to be context-dependent, findings from this study could have broad significance for other cancers with high RNF40, or even in other pathological contexts -if any- that cursed with high RNF40.

It also provides some mechanistic data (that should be improved as suggested in comments) linking this ubiquitin ligase to the cytoskeletal machinery and, therefore, control of migration and also proliferation and survival. This will also advance the field.

Area of expertise

Actin-myosin cytoskeleton, Rho GTPases-ROCK, cancer, metastasis, cell signalling

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

**Summary:**

Wegwitz and colleagues present extensive and detailed data focussed on the role of the E3 ubiquitin ligase and ring finger protein RNF40 in HER-2 associated breast cancer. It is clear that the role of RNF40 and its major substrate histone H2B (monoubiquitination of histone H2B at lysine 120; H2Bub1) as part of a complex with RNF20, is not a simple one in the context of malignancy. This group, and others, have previously reported on the intriguing role of RNF40 that can in certain circumstances function to suppress tumorigenesis, and in other circumstances function to support tumorigenesis. While H2Bub1 has been shown to be lost in many different malignancies, these investigators show that in HER-2 associated breast cancer, this is not the case. In fact, the results presented in this study show that RNF40-mediated H2Bub1 is important for the expression of genes involved in the actin cytoskeleton and the downstream FAK signalling cascade. Supporting this, mining of a public database showed that RNF40 mRNA was high in HER-2 associated breast cancers and was correlated with a worse prognosis (overall and RFS, relapse free survival). The investigators also used a mouse model (MMTV-Erbb2) generating a tri-transgenic (MMTV-Erbb2; MMTV-Cre; Rnf40flox) that allowed breast tissue specific overexpression of HER2 at the same time as KO of Rnf40, so mimicking the human disease. In fact, mouse tumours recapitulated the human results, including in disease free survival (lower with higher Rnf40), and with less tumours seen when there was less Rnf40 (the Rnf40 floxed tumours appeared heterogenous in staining patterns for Rnf40 and H2Bub1, supporting the concept of "escaper" cells, positive for both Rnf40 and H2Bub1 that would be positively selected during tumorigenesis).

The authors also took a cell biology approach to studying HER2 positive breast cancer and RNF40 using two HER2 positive cell lines (HCC1954 and SKBR3). RNF40 was down-regulated using siRNA and numerous functional studies showed that targeting RNF40 suppressed behaviours consistent with tumorigenesis (proliferation, migration, clonogenic survival, spheroid formation, growth kinetics). Furthermore, down-regulation of RNF40 in the HCC1954 cell line followed by GSEA identified gene signatures associated with apopotosis and the actin cytoskeleton regulatory pathway (e.g. ROCK1, VAV3, LIMK2, PFN2). They further showed that phospo - cofilin (that occurs downstream of ROCK1) was reduced in RNF40 down-regulated cells, also implicated in regulation of the actin cytoskeleton. Phalloidin staining for F-actin showed disruption of the cytoskeleton in RNF40 down-regulated cells. Additionally, the ROCK1 inhibitor, RKI-1447, showed similar effects to depletion of RNF40.

The authors then sought to determine whether the RNF40 associated gene expression in HER2 positive cells was in fact happening through H2Bub1 and the active histone mark H3K4me3 it has been reported to cross-talk with. RNF40 regulated genes (up or down-regulated) showed lower levels of H2Bub1 occupancy compared to non-regulated genes. H3K4me3 was lost in most genes influenced by RNF40 down-regulation, including genes associated with the actin regulatory pathway. The overall conclusion is that RNF40 is a major epigenetic regulator of the actin regulatory gene network in HER 2 positive breast cancer and could be a therapeutic target.

**Major comments: major issues affecting the conclusions.**

(1) What is happening at the gene level for both H2Bub1 and H3K4me3 in the context of RNF20 down-regulation is complex and would benefit from inclusion of a schematic, or a series of schematics describing different scenarios, as the text is quite difficult to follow.

This is a very constructive proposition. We will attempt to follow this suggestion in order to simplify the message of the respective section by providing to schematic illustrations depicting the cascade of events occurring upon RNF40 loss in the cancer cells.

It is not entirely clear that the changes seen in H3K4me3 are a direct result of cross-talk with H2Bub1 (some literature reports that there is no cross-talk between these histone marks for instance). It is also not entirely clear how the other histone marks investigated support the main discoveries of the paper. The authors need to consider this in the way that they present the data and their interpretation of it.

The reviewer addresses an important point about the mechanistic aspect of the RNF40-dependent epigenetic regulation. We and others have shown that RNF40-mediated H2B monoubiquitination is a central step for activation of the COMPASS complex and the TSS-proximal broadening of H3K4me3 (PMID:31733991, 19410543, 22505722, 28209164). However, the situation certainly is not as straight forward as it is in yeast, where the vast majority of H3K4 trimethylation is H2Bub1-dependent. To what degree global H3K4me3 levels are dependent upon the H2B ubiquitin ligases RNF20 and RNF40 appears to vary, depending upon the investigated system (probably the variation in the literature referred to by the reviewer). However, in our work, we reproducibly see widespread H3K4me3 peak narrowing specifically on RNF40-dependent genes, in a context-dependent manner (i.e., genes displaying these effects are different according to the system investigated). To support and consolidate the central function of the H2Bub1-H4K4me3 crosstalk in our system, we propose to perform rescue experiments: siRNAs targeting the 3’UTR of RNF40 will be co-transfected with an expression construct encoding for either a wild type or a ΔRING (catalytic inactive) form of RNF40 lacking the endogenous 3’UTR. The ability of ectopically expressed wild-type, but not catalytic inactive RNF40, to rescue the expression of the identified actin cytoskeleton genes and downstream signaling should provide a solid argument to support the hypothesis of our study. We will also include additional discussion about the potential different H3K4 methyltransferases that may potentially be involved.

(2) RNF40 is known to work in a complex with RNF20 to monoubiquitinate histone H2B at lysine 120 (H2Bub1). In experiments where RNF40 has been down-regulated, did the authors also note down-regulation of RNF20 (as has been previously reported).

This is an interesting question from the reviewer. We indeed observed a consistent reduction of RNF20 protein levels upon RNF40 knockdown (and vice versa) in different cell systems, including the HER2-positive cell lines HCC1954 and SKBR3.

Is the data presented likely to be the result of abrogation of the complex rather than RNF40 specifically?

Although particularly difficult to answer, the use of a catalytic mutant in key experiments should at least partially shed light on this aspect (as proposed in the answer to Reviewer #3’s question 1). In that case, the complex integrity can be maintained while specifically abrogating RNF40 ubiquitin ligase activity.

While I am not asking for experiments to be repeated with down-regulation of RNF20, some consideration of this needs to be included in the Discussion. Is RNF20 also highly expressed in HER2 positive breast cancer (TCGA, KM Plotter data).

We absolutely agree with the reviewer’s point of view. As an obligate binding partner of RNF40, RNF20 indisputably plays an important function in the phenotype caused through RNF40 loss. We will therefore carefully further discuss this aspect in the revised manuscript. Preliminary analyses based on the TCGA dataset point at a high expression of RNF20 in HER2-positive lesions. Furthermore, survival analysis of HER2+ BC patients based on the same dataset showed that patients with high RNF20 expression harbor an unfavorable prognosis, similar to what we have seen with RNF40. We may therefore implement these expression and survival data in the revised manuscript.

**Minor comments: important issues that can confidently be addressed.**

(3) It would appear that immunohistochemistry for RNF40 and H2Bub1 on human samples is only reported as "low" or "high". This is perhaps not dealing with the full spectrum of IHC scores, such as completely absent, although the methods do note a "null" value (no detectable staining). Were there no "null" results? Please define the criteria for "low" or "high".

Indeed, specimens lacking H2Bub1 or RNF40 staining were attributed the “null” scoring. However, while we have observed null staining in other BC subtypes (e.g., see Bedi, et al., 2015), none of the HER2 positive BC samples were found to be negative for either RNF40 or for H2Bub1. However, for the revision, we will provide representative examples of null-stained tumor specimens (from other BC subtypes) for RNF40 and H2Bub1 from the same tissue microarray.

(4) I think there might be some confusion in labelling of Fig 1A and B as the legend states that all breast cancers are on the left and the HER-2 positive on the right, for each of primary tumours and brain mets, but I think one is under the other? Labelling should be checked in this figure.

We apologize for this mistake. This has been corrected in the figure legend accordingly.

(5) What this IHC data doesn't show is whether RNF40 and H2Bub1 levels are always correlated in individual tumours (i.e. RNF40:H2Bub1, high:high OR low:low OR null:null). Can the authors please include and comment on this data.

The reviewer has made a very interesting point here. We will comment on this point in the revision.

(6) Please include overall survival data (KM Plotter) as a panel in figure 1, alongside RFS for RNF40 expression levels (currently in Supplementary).

We added the OS as well the RFS data from the same database next to each other in the main figure.

(7) Spheroid formation looks to only be shown in a single cell line (HCC1954). Was the other cell line not suitable for spheroid studies? Some comment should be made and care taken not to "overclaim" as text notes two cell lines.

SKBR3 are unfortunately not suitable for tumor sphere formation assay. We may provide instead a soft agar assay with SKBR3 cells upon. If needed, we may replace the SKBR3 cell line with BT474 for this specific experiment

(8) It would have been interesting to see results of a GSEA in the mouse mammary tumours as a complement to human. Is there a reason why this wasn't undertaken?

Rnf40fl/fl tumors present a large fraction of “escaper” cancer cells retaining RNF40 expression. For this reason, bulk sequencing of such tumors would likely only provide a “diluted” molecular signature consequent to RNF40 loss. For this reason this experiment has not been done.

(9) Conclusions are made about RNF40 in HER2 positive cells only in the context of H2Bub1 and H3K4me3. Without having conducted similar experiments in HER2 negative breast cancer cell lines / models, it is difficult to draw the conclusion that this is HER2 positive specific. Can the authors either soften some of their conclusions along this line, or consider repeating some of their data in HER2 negative models.

The scope of the study has deliberately been set on HER2-positive malignancies, because former studies already extensively studied the impact of H2Bub1 loss in TNBC and Luminal BC (PMID 28157208, 18832071). We will therefore modify the manuscript text accordingly and soften the appropriate sections as suggested by the reviewer.

(10) RNF40 likely has substrates other than histone H2B. There is a report describing interactions with RNF40 (STARING) and syntaxin for e.g., (Chin et al., 2002 J Biol Chem 277:35071-9). Can the authors please comment on other potential substrates of RNF40 in light of their data that focuses only on its epigenetic role as a regulator of the actin cytoskeleton.

Our study was mainly focused only on the gene expression program driven by RNF40 in HER2+ BC. We therefore do not know nor have we focused on other novel non-histone substrates. We will, however, allude to this possibility in a revised manuscript.

Reviewer #3 (Significance (Required)):

Nature and Significance of the Advance:

Clinically, this work provides a significant advance in that it is zeroing in on HER2 positive breast cancer and generating fundamental data that could underpin development of a new therapy for this malignancy. Conceptually, it is expanding knowledge of how the E3 ubiquitin ligase RNF40 is functioning as an epigenetic modifier of a specific type of malignancy by being important for the actin cytoskeleton.

Work in Context of Existing Literature:

As acknowledged by the authors, this work builds on a previous publication of theirs (Xie et. al., 2017 "RNF40 regulates gene expression in an epigenetic context-dependent manner." Genome Biol). They have other recent papers on RNF40 (Schneider et al., 2019 "The E3 ubiquitin ligase RNF40 suppresses apoptosis in colorectal cancer cells", Clin Epigenetics; Kosinsky et al., 2019 "Loss of RNF40 decreases NF-kappaB activity in colorectal cancer cells and reduces colitis burden in mice", J Crohns Colitis). H2Bub1 is one of the least well studied histone modifications and as such, this study of one of its key histone writers, RNF40, is significant in elucidating the significance of this histone mark.

Audience:

This paper will suit a discovery-based science audience interested in epigenomic regulation of malignancy. Further, it will suit those looking for new drug development strategies for malignancy.

My Field of Expertise:

Basic scientist with expertise in epigenetic/epigenomic regulation in malignancy; cell and molecular biology. I felt capable of reviewing all aspects of this 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

Summary:

Wegwitz and colleagues present extensive and detailed data focussed on the role of the E3 ubiquitin ligase and ring finger protein RNF40 in HER-2 associated breast cancer. It is clear that the role of RNF40 and its major substrate histone H2B (monoubiquitination of histone H2B at lysine 120; H2Bub1) as part of a complex with RNF20, is not a simple one in the context of malignancy. This group, and others, have previously reported on the intriguing role of RNF40 that can in certain circumstances function to suppress tumorigenesis, and in other circumstances function to support tumorigenesis. While H2Bub1 has been shown to be lost in many different malignancies, these investigators show that in HER-2 associated breast cancer, this is not the case. In fact, the results presented in this study show that RNF40-mediated H2Bub1 is important for the expression of genes involved in the actin cytoskeleton and the downstream FAK signalling cascade. Supporting this, mining of a public database showed that RNF40 mRNA was high in HER-2 associated breast cancers and was correlated with a worse prognosis (overall and RFS, relapse free survival). The investigators also used a mouse model (MMTV-Erbb2) generating a tri-transgenic (MMTV-Erbb2; MMTV-Cre; Rnf40flox) that allowed breast tissue specific overexpression of HER2 at the same time as KO of Rnf40, so mimicking the human disease. In fact, mouse tumours recapitulated the human results, including in disease free survival (lower with higher Rnf40), and with less tumours seen when there was less Rnf40 (the Rnf40 floxed tumours appeared heterogenous in staining patterns for Rnf40 and H2Bub1, supporting the concept of "escaper" cells, positive for both Rnf40 and H2Bub1 that would be positively selected during tumorigenesis). The authors also took a cell biology approach to studying HER2 positive breast cancer and RNF40 using two HER2 positive cell lines (HCC1954 and SKBR3). RNF40 was down-regulated using siRNA and numerous functional studies showed that targeting RNF40 suppressed behaviours consistent with tumorigenesis (proliferation, migration, clonogenic survival, spheroid formation, growth kinetics). Furthermore, down-regulation of RNF40 in the HCC1954 cell line followed by GSEA identified gene signatures associated with apopotosis and the actin cytoskeleton regulatory pathway (e.g. ROCK1, VAV3, LIMK2, PFN2). They further showed that phospo - cofilin (that occurs downstream of ROCK1) was reduced in RNF40 down-regulated cells, also implicated in regulation of the actin cytoskeleton. Phalloidin staining for F-actin showed disruption of the cytoskeleton in RNF40 down-regulated cells. Additionally, the ROCK1 inhibitor, RKI-1447, showed similar effects to depletion of RNF40. The authors then sought to determine whether the RNF40 associated gene expression in HER2 positive cells was in fact happening through H2Bub1 and the active histone mark H3K4me3 it has been reported to cross-talk with. RNF40 regulated genes (up or down-regulated) showed lower levels of H2Bub1 occupancy compared to non-regulated genes. H3K4me3 was lost in most genes influenced by RNF40 down-regulation, including genes associated with the actin regulatory pathway. The overall conclusion is that RNF40 is a major epigenetic regulator of the actin regulatory gene network in HER 2 positive breast cancer and could be a therapeutic target.

Major comments: major issues affecting the conclusions.

(1) What is happening at the gene level for both H2Bub1 and H3K4me3 in the context of RNF20 down-regulation is complex and would benefit from inclusion of a schematic, or a series of schematics describing different scenarios, as the text is quite difficult to follow. It is not entirely clear that the changes seen in H3K4me3 are a direct result of cross-talk with H2Bub1 (some literature reports that there is no cross-talk between these histone marks for instance). It is also not entirely clear how the other histone marks investigated support the main discoveries of the paper. The authors need to consider this in the way that they present the data and their interpretation of it.

(2) RNF40 is known to work in a complex with RNF20 to monoubiquitinate histone H2B at lysine 120 (H2Bub1). In experiments where RNF40 has been down-regulated, did the authors also note down-regulation of RNF20 (as has been previously reported). Is the data presented likely to be the result of abrogation of the complex rather than RNF40 specifically? While I am not asking for experiments to be repeated with down-regulation of RNF20, some consideration of this needs to be included in the Discussion. Is RNF20 also highly expressed in HER2 positive breast cancer (TCGA, KM Plotter data).

Minor comments: important issues that can confidently be addressed.

(3) It would appear that immunohistochemistry for RNF40 and H2Bub1 on human samples is only reported as "low" or "high". This is perhaps not dealing with the full spectrum of IHC scores, such as completely absent, although the methods do note a "null" value (no detectable staining). Were there no "null" results? Please define the criteria for "low" or "high".

(4) I think there might be some confusion in labelling of Fig 1A and B as the legend states that all breast cancers are on the left and the HER-2 positive on the right, for each of primary tumours and brain mets, but I think one is under the other? Labelling should be checked in this figure.

(5) What this IHC data doesn't show is whether RNF40 and H2Bub1 levels are always correlated in individual tumours (i.e. RNF40:H2Bub1, high:high OR low:low OR null:null). Can the authors please include and comment on this data.

(6) Please include overall survival data (KM Plotter) as a panel in figure 1, alongside RFS for RNF40 expression levels (currently in Supplementary).

(7) Spheroid formation looks to only be shown in a single cell line (HCC1954). Was the other cell line not suitable for spheroid studies? Some comment should be made and care taken not to "overclaim" as text notes two cell lines.

(8) It would have been interesting to see results of a GSEA in the mouse mammary tumours as a complement to human. Is there a reason why this wasn't undertaken?

(9) Conclusions are made about RNF40 in HER2 positive cells only in the context of H2Bub1 and H3K4me3. Without having conducted similar experiments in HER2 negative breast cancer cell lines / models, it is difficult to draw the conclusion that this is HER2 positive specific. Can the authors either soften some of their conclusions along this line, or consider repeating some of their data in HER2 negative models.

(10) RNF40 likely has substrates other than histone H2B. There is a report describing interactions with RNF40 (STARING) and syntaxin for e.g., (Chin et al., 2002 J Biol Chem 277:35071-9). Can the authors please comment on other potential substrates of RNF40 in light of their data that focuses only on its epigenetic role as a regulator of the actin cytoskeleton.

Significance

Nature and Significance of the Advance:

Clinically, this work provides a significant advance in that it is zeroing in on HER2 positive breast cancer and generating fundamental data that could underpin development of a new therapy for this malignancy. Conceptually, it is expanding knowledge of how the E3 ubiquitin ligase RNF40 is functioning as an epigenetic modifier of a specific type of malignancy by being important for the actin cytoskeleton.

Work in Context of Existing Literature:

As acknowledged by the authors, this work builds on a previous publication of theirs (Xie et. al., 2017 "RNF40 regulates gene expression in an epigenetic context-dependent manner." Genome Biol). They have other recent papers on RNF40 (Schneider et al., 2019 "The E3 ubiquitin ligase RNF40 suppresses apoptosis in colorectal cancer cells", Clin Epigenetics; Kosinsky et al., 2019 "Loss of RNF40 decreases NF-kappaB activity in colorectal cancer cells and reduces colitis burden in mice", J Crohns Colitis). H2Bub1 is one of the least well studied histone modifications and as such, this study of one of its key histone writers, RNF40, is significant in elucidating the significance of this histone mark.

Audience:

This paper will suit a discovery-based science audience interested in epigenomic regulation of malignancy. Further, it will suit those looking for new drug development strategies for malignancy.

My Field of Expertise:

Basic scientist with expertise in epigenetic/epigenomic regulation in malignancy; cell and molecular biology. I felt capable of reviewing all aspects of this 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 study, Wegwitz et al propose that the E3 ubiquitin ligase RNF40 is highly expressed in HER2+ breast cancer tumours and correlates with poorer survival, using their own and TCGA data. Contrary to observations suggesting a tumour-suppressive role in other cancers, authors show using RNF40-knockout breast cancer mouse models and in vitro data shat RNF40 promotes tumour growth. RNF40 depletion impairs proliferation, survival and sphere formation by inducing apoptosis. In addition, RNF40 promotes cell migration by upregulating expression of cytoskeletal proteins (ROCK1, VAV3, LIMK2) and their effectors such as phosphorylated cofilin. Authors show elegant partial rescue experiments of the effect of RNF40 depletion on apoptosis and survival.

Given that RNF40 function seems to be context-dependent, findings from this study could have broad significance for other cancers with high RNF40. It also provides some mechanistic data (that should be improved as suggested below) linking this ubiquin ligase to the cytoskeletal machinery and, therefore, control of migration and also proliferation and survival.

Data are well presented and most conclusions are supported by the data. However, there are some gaps at the mechanistic level. Since migration is controlled by RNF40 in vitro, evaluation of metastatic ability in vivo (local invasion for example as suggested below) should be evaluated and would strengthen this part too.

Major comments

1.Fig.1A-B, S1A. Specificity of RNF40 antibody should be shown, which could be done quite easily in the tumours from the knockouts. From the datasheets, antibodies recognize human protein only.

2.It is unclear when the murine tumours were analysed, at endpoint? This should be stated. Could authors establish cell lines from the mouse tumours (knockout, partial knockout escapers..)? These could be very useful tools to evaluate key in vitro findings from the study.

3.Fig.1F-G: since RNF40 controls the cytoskeletal machinery and therefore, migration (Fig. 2G) in the RNF40 knockout tumours, was metastasis (if observed) affected? Or if there was no growth in distant organs detected in the time frame of these experiments, was invasion (and/or pattern of invasion or mode of invasion (morphology of invading cells)) into adjacent tissues affected upon RNF40 depletion? This would add in vivo relevance to the in vitro mechanistic findings, especially since the authors later showed that p-cofilin was also decreased in the RNF40-depleted mouse tumours (Fig.4D).

4.Fig.3: most results using HCC1954 cell line. Key findings should be validated in other cell lines.

5.Fig.3A: authors state "both pathways remained intact following RNF40 depletion". However, from those blots, siRNF40 clearly increases pERK and slightly pAKT, which would be unexpected according to previous data in Fig.2. Authors could show quantifications of different blots, or show a more representative blot if increase in pERK was not consistently observed. Was this also seen in SKBR3 cell line?

6.For Fig.3G and Fig.S3A, authors selected genes from this set, how was this done (fold change?). Was expression of the other family members (ROCK2, LIMK1, etc) or of Rho GTPases regulated too?

7.Fig.4B: this may not help, decrease of p-cofilin by Vav3 knockdown is way less dramatic compared to RNF40 depletion or ROCK inh treatment. See comment below regarding other effectors such as Myosin. Fig.4C: does ROCK inh reduce RNF40 levels? It may from the immunofluorescence picture.

8.Fig.4H-I: the sphingosine 1-phosphate receptor-3 agonist (CYM-5441) partially rescued the effects of RNF40. Since S1P signalling involves Rho GTPase activation -presumably downstream of VAV3 -which is a GEF for Rho, Rac and Cdc42- and upstream of ROCK, LIMK, was activity of these Rho GTPases affected upon RNF40 depletion? This would strengthen the mechanism.

Related to this, was Myosin II activity (phosphorylated MLC2) affected -since its upstream regulators, especially ROCK are controlled by RNF40?

9.Fig. S5E: Authors should consider presenting data of decreased histone methylation of cytoskeleton regulators in main Fig. 5, since this is an important conclusion of this part.

10.Statistics should be revised throughout the manuscript. Comparisons of more than 2 groups should be performed with ANOVA or similar multiple comparison test (instead of t-test).

Minor comments

1.Statement of significance mentions "Anti-HER2-therapy resistance", but this is a misleading since the paper does not deal with therapy resistance. Or are the cell lines used in the study resistant to anti-HER2? In line with this, authors could add some lines of thought on how RNF40 could be targeted in the clinical/pre-clinical context, which could inform further translational studies.

2.Line 117 "Moreover, HER2-positive metastatic BC samples showed a particularly high expression of RNF40 compared to primary tumors". Perhaps rephrase, was it that the expression level (intensity) was higher or that the % of positive cells/tumours was higher in the brain mets?

3.Fig.1D and S1C. while S1C shows TCGA data, it is unclear which set of patients is Fig.1D (since text says publicly available data, line 118-119), are these their own set of patients (used in Fig.1A-B)? This should be specified in the text, legend.

4.Line 122. Authors should be careful with this conclusion so far, a correlation between expression and cancer stage/survival does not necessarily mean a tumor suppressive/supportive role.

5.Fig. S4E: there are missing labels in graph (control, siRNF40).

6.Panels in some figures are discussed in text randomly and not following same order. For example, Fig.1 (after panel D, then panel H, then back to E, F, G), S3, 4A-C,

7.Fig.1E: I would suggest changing the line colours, so Rnf40wt-wt line is red and the fl-fl is black, therefore it is similar to panel D (high Rnf40 red, low in black).

8.Supp Videos: for reviewers and readers, it would help that video has a label while it plays, otherwise after downloading it, video name does not tell whether it is control or RNAi.

Significance

Given that RNF40 function seems to be context-dependent, findings from this study could have broad significance for other cancers with high RNF40, or even in other pathological contexts -if any- that cursed with high RNF40. It also provides some mechanistic data (that should be improved as suggested in comments) linking this ubiquitin ligase to the cytoskeletal machinery and, therefore, control of migration and also proliferation and survival. This will also advance the field.

Area of expertise Actin-myosin cytoskeleton, Rho GTPases-ROCK, cancer, metastasis, cell signalling

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 study by Wegwitz et al, the authors examine the tumour promoting properties of RNF40 (and the H2B monoubiquitinylation catalysed by it) in Her2 driven breast cancer. They report, using publicly available data, that increased RNF40 expression is associated with reduced overall and disease-free survival. Using a mouse model, where they crossed the Erbb2 (mouse Her2) under the control of the MMTV promoter with conditional Rnf40 deletion constructs, the authors found that deletion of Rnf40 simultaneous to Her2 overexpression resulted in a prolonged tumour-free survival, somewhat reduced tumour growth kinetics and tumour incidence. siRNA silencing of Rnf40 in two Her2 positive breast cancer cell lines resulted in reduced proliferation, clonogenicity and tumour sphere formation and cellular motility. Transcriptome analysis revealed pathways that could explain the phenotype, like increased apoptosis and actin cytoskeleton regulation. The authors then took further some candidates in the later pathway to investigate the mechanism. They find that Rnf40 loss impacts on actin cytoskeletal dynamics. They also investigate the impact on focal adhesion signalling integrity. Finally, they investigate the relationship between the transcriptome and H3K4me3 and H2Bub1 landscape in the presence or absence of Rnf40.

The manuscript is convincing regarding the tumour promoting roles of Rnf40, but the key claim that H2B monoubiquitinylation is essential for activation of the Rho/Rock/Limk pathway, where genes are down regulated upon Rnf40 loss resulting in decreased tumourigenicity of cells, is so far not convincing. "Together, these findings support the hypothesis that the actin regulatory gene network is dependent on direct 271 epigenetic regulation by RNF40 through modulation of H2Bub1 and a trans-histone cross-talk with H3K4me3 272 levels in HER2-positive BC cells." Although the correlation is apparent, at this point it's unclear if the phenotype is dependent on the catalytic activity of Rnf40 or it's a non-catalytic effect. Generating a catalytic mutant RNF40 and test it at least in the cell lines studied would be desirable.

Other comments that need a response:

1."we investigated RNF40 expression and 114 H2Bub1 levels by immunohistochemical staining of 176 primary BC tumors and 78 brain metastases." In Fig 1 I can only count 41 primary BC tumours and 73 brain metastases. Numbers don't add up. Also, how is "low" defined as opposed to negative? What is used as controls?

2."Moreover, HER2-positive metastatic BC samples showed a117 particularly high expression of RNF40 compared to primary tumors" Figure 1 or Fig S1A does not contain data on HER2-positive metastatic BC

3."tumors did not display a loss of either 132 RNF40 or H2Bub1 (Fig. 1H) when compared to the adjacent normal mammary epithelium (Fig. S1F)." I don't understand what I see in Fig S1F, where is the tumour, what is adjacent?

4."homozygous loss of Rnf40 (Rnf40fl/fl134 ) resulted in 135 dramatically increased tumor-free survival of MMTV-Erbb2 animals (Fig.1E)." This is overinterpretation of the data, I would not call it dramatic, just significant.

5."loss of Rnf40 led to 139 strongly reduced tumor growth kinetics (Fig.1G)." Is this result significant, I did not see an evaluation of statistical significance in this data.

6."Rnf40fl/fl 142 lesions displayed a 143 heterogeneous pattern of RNF40 expression (Fig.1H), suggesting that the few tumors that did develop in this 144 model were largely caused by an incomplete loss of the Rnf40 allele." If this conclusion is suggested, the authors should check if the "escaper" cells have failed to flox the Rnf40 allele on the genetic/protein level. Otherwise it's not conclusive.

7.Fig S4D - is this clonogenic assay? How many replicates were done, biological technical?

8."Additionally, treatment with either CYM-5441 (Fig.4J) or 225" Fig 4J is missing! It makes this section rather hard to follow. Fig S4F-G, how many replicates were done, biological technical?

9."Consistent with our analyses based on changes in H3K4me3 occupancy, genes downregulated upon RNF40 256 silencing displayed the most prominent decrease in H3K4me3 in the gene body (the 3' end of the peak)" The impact of these mods changes is hard to judge because they are rather small (I would not use the wording prominent). Also, are there many other "peak narrowing" genes but they are not downregulated?

10.Statistical analysis missing: for example in Fig 2C, Fig 2E, Fig 3G what is n=?, how many technical, biological replicates were analysed?

11.Fig 4E seems to be a partial duplication of Fig 3D!

Minor:

Figure referencing: it can be quite confusing to see a different ordering of figures compared to the referencing in the manuscript, for example Fih 1H is referenced in the text before Fig 1F, G. The authors should change the order in the main figures....

Significance

It's an interesting study that associates epigenetic regulation of actin cytoskeletal dynamics in Her2 driven breast cancer.

Reviewer #3

This is a straightforward study addressing the evolutionary divergence of the glucocorticoid receptor. Authors use the GR and MR receptors from elephant shark which represent distinct orthologs of human GR/MR which diverged from a common CR receptor in cartilaginous fishes. The authors address two functional questions regarding 1) agonist/antagonist specificity between ES GR/MR and 2) The functional role of the AB domain (N terminal domain) of the GR/MR which is known to play a specific role in GR transactivation. The study is technically well executed. However, the following should be addressed.

1) While the introduction is informative, it is difficult to follow as the authors describe ligand activities for two receptors, multiple ligands, multiple species and chimeras. While this information is summarized well in Table 1, it does not appear until well into the manuscript. It would help readers, I believe, to be more general in the introduction rather than provide a plethora of ligand specificities.

2) Given that a large component of the study focuses on the functionality of the GR A/B (AF1) activation domain here termed the "NTD" it would seem prudent to have some introduction and/or discussion on the role of this domain in NR's in general. Depending upon which of the NRs is being addressed the AB domain may serve multiple functionalities. For instance, the AB domain domain is a target site for receptor phosphorylation through differing kinase/phosphatase activities. Phosphorylation within the AF-1 domain can significantly affect transcriptional activity and impact ligand dependent and ligand independent activities. For example, Estrogen receptors are phosphorylated at both serine and threonine residues by mitogen activated kinase (MAPK) following growth factor stimulation and enhance transcriptional activity. PPAR and PPAR are additionally phosphorylated within the A/B domain yet exhibit reciprocal transcriptional activation (PPAR) and repression (PPAR). VDR and RXR also have putative phosphorylation sites. In this study, no mention is given to the role of Ab domain phosphorylation or how the functionality of the AB domain might be involved in allosteric interactions that facilitate ligand receptor binding.

3) As stated in comment 2 above, the study could also be greatly enhanced if the authors further conducted experiments to further refine the region of importance within the NTD that facilitates the activation of the GR. Alignment of the two sequences could help infer potential targets and functional mutation studies may provide greater mechanistic insight into the NR functionality and aid making evolutionary inferences in how MR and GR have diverged from human GR/MR. This aspect of the study could also be modeled using a three-dimensional molecular docking approach.

4) The experiments were conducted in HEK cells, which may or may not contain essential coregulators necessary for driving transactivation. It is also highly noticeable that MR activities are significantly attenuated compared to GR. Comment by the authors on both these points would be useful.

5) While the study recognizes the significant differences in EC50 across receptor types and their ligands, little attention is given to the Emax for each of the assays. It seems strange that ES-MR demonstrates a great potency for cortisol and corticosterone than GR however the Emax values for GR are magnitudes greater than MR.

Minor comments:

Why was the AF2 domain left out of Figure 2?

Reviewer #2

The authors report the first characterization of the elephant shark glucocorticoid receptor (GR). In my view, the experiments are a useful contribution to the literature, but the significance of the work as presented is limited.

They have two new findings:

1) The elephant shark GR does not activate in response to progesterone or 19norP, despite the steroids binding to the GR. This contrasts with their previously published characterization of the elephant shark MR (ref #36). GR from other organisms does not activate with progestins, but also does not bind them.

2) The GR N-terminal domain (NTD) dramatically increases the fold activation of the GR, but has no apparent effect on steroid specificity (Figure 4). This is a property of the NTD, as swapping the GR NTD onto the MR ligand-binding domain leads to elevated activity (Figure 6). This behavior matches what has been seen previously for bony-vertebrate GRs, but has not been demonstrated for cartilaginous fish GRs.

They have one finding that is somewhere between new and confirmatory:

3) The elephant shark GR behaves similarly to the previously characterized skate GR (refs #7, #37) in that it responds to both aldosterone and cortisol and has lower sensitivity to steroids than the MR.

Presentation:

I found this paper difficult to read. The introduction was long. It was difficult to tell from the introduction what was previously known and what was new in this paper. The results had little narrative structure, making it difficult to understand why the authors chose to do each experiment. And the discussion did not really explore the implications of their observations.

There are four data figures and a table. Of those, Figs 3, 4, and Table 1 are the same data shown in different ways. Of the data in these figures, the MR bits-about half of the data-have already been published for slightly different constructs of the same proteins (ref #36). The work was observational, with no mechanism – evolutionary, biochemical, physiological, or otherwise – presented.

Specific comments:

1) The authors never discuss the implications of their results for the physiology of elephant sharks. Why should it matter that the elephant shark GR does not respond to progesterone and 19norP? Is this surprising given what we know about GRs from other species?

2) The authors don't "close the loop" on their evolutionary questions regarding steroid specificity. How does their work contribute to our understanding of the evolution of the GR and its function across vertebrates? Can they propose when the progesterone response evolved (or was lost)? Was it gained on the MR lineage or lost on the GR lineage? (One of the papers the authors cite (Bridgham et al, #7) reports that the hagfish CR-which is co-orthologous to MR and GR from jawed vertebrates-responds to progesterone. It seems like this is worth bringing into their discussion). In general, a much more fleshed out discussion of what is known about GR and MR from other cartilaginous, ray-finned and jawless fishes is in order.

3) The authors argue that the importance of the NTD for GR activation, but not MR activation, indicates that the NTD activity evolved after the divergence of MR and GR. It is equally likely, however, that MR NTD lost its ancestral ability to activate. (This could be tested by, for example, characterizing full-length CR from hagfish or lamprey and asking if its NTD is more MR-like or GR-like in function.)

4) In the paragraph starting “Activation of elephant shark GR by aldosterone...”, the authors should probably note that the previously characterized skate GR responds to aldosterone and cortisol, as does the reconstructed ancestor of GR and MR (refs #7, #37). This adds heft to their claim that GR is transitional from MR in elephant shark.

5) The authors motivate their decision to characterize the full-length elephant shark GR by saying that because no full-length elasmobranch GR has been characterized, "the identity of the physiological glucocorticoids in cartilaginous fish is not known." This seems odd, given that, to a first approximation, most GR NTDs amplify the response to all steroids without dramatically altering specificity (see, Figure 4A and C, for example). Is there some reason the authors expect the NTD to alter specificity in this case? Further, all of the data in this manuscript are in vitro: this cannot show whether these steroids are physiological or not.

Minor comments:

In several places, the language the authors chose seems to imply that the elephant shark is ancestral. The sentences should be modified to indicate that the shark gives insight into an ancestral state, but is not itself ancestral.

Reviewer #1

This is a well carried out study of the ligand specificity and also the role of the NTD of elephant shark GR and MR. The study though, conflates two things – the role of the NTD in transactivation (it is well known that the NTD of steroid receptors contains a transcription activation function – TAF1) – and the role of the NTD in ligand binding (allosteric interactions between the NTD and the ligand binding domain; LBD). While it is possible that the NTD exerts an allosteric influence over the LBD, as suggested by the authors, I do not feel that this conclusion is justified by the data presented.

Major points:

1) The introduction conflates the two issues of allosteric interaction between NTD and LBD (e.g., as shown in ref 22) with the existence of a TAF in the NTD (demonstrated in ref 19 for example). The activation domain is autonomous, requiring tethering to DNA by the DNA binding domain of the receptor. This applies to the statement “It is not known when the strong dependence of vertebrate GR on the NTD for activation of gene transcription evolved”. However, it has also been demonstrated (though not in most of the references cited) that there is an allosteric interaction by which the GR NTD alters ligand binding (i.e., affinity) by the LBD, alluded to (albeit not explicitly) earlier on.

This is problematic when it comes to the way the data in Figure 3 are described. Activation of a reporter gene was measured, not activation of the receptor. If the reporter gene had not contained a GRE, there would have been no effect of steroid on the experimental readout, but the receptor would still have been activated by the steroid. What Figure 3 shows is that the GR NTD contains a strong transcriptional activation domain, required for induction of MMTV LTR-luciferase, consistent with previously published data. However, the MR does not have a TAF in the NTD (at least one active at this promoter); deletion of the NTD has no effect on transactivation of the reporter. The data in Figure 3 say nothing about the affinity of the receptors for the ligands. To infer anything about ligand-dependent activation of receptor, a ligand dose-response is required (as in Figure 4).

2) It is not clear if the EC50s reported in Table 1 are sufficiently (significantly) different to each other in order to infer anything about the influence of the NTD on ligand binding. No statistical analysis has been performed on the EC50s.

The similarities of the EC50s for all 4 corticosteroids for the truncated and full-length MR supports the EC50 being determined by the LBD and is consistent with no role for TAF1 at this promoter. For the GR, the EC50s reported in Table 1 derive from the data shown in Fig4. There are very minor differences in EC50 for corticosterone (the highest affinity ligand) between GR-FL, GR-truncated and the MR-GR chimera. This suggests the affinity of the receptor for this ligand is determined by the LBD. Likewise, the affinity of all 3 receptors for cortisol is within experimental error (with the additional caveat that the graphs in Figure 4 do not plateau for cortisol, so the estimate of EC50 is likely to be inaccurate; this caveat also applies to the statement “the EC50 for cortisol increased over 2-fold, and an EC50 for 11-deoxycorticosterone was too low to be calculated”). Aldo also doesn't reach plateau, so the same caveats apply there (and there is <3-fold difference between the EC50 for GR-FL and MR-GR). This I do not feel that the data support the conclusion drawn regarding allosteric signalling between NTD and LBD (“These results indicate that allosteric signaling between the NTD and DBD-LBD in elephant shark GR is critical for its response to corticosteroids, in contrast to elephant shark MR”).

3) The same issue arises in the description of the results of the chimera experiment (starting “Thus, in the GR NTD-MR DBD-LBD chimera...”). The reason that fusion of the GR NTD to MR LBD conferred greater activation of the reporter gene is because it fused a strong activation domain (TAF1) to the MR LBD. This also allowed the prog and 19norprog activation - the fusion of the GR TAF1 to the MR ligand specificity. This was to be expected. Similarly, the MR NTD does not contain a TAF (at least one active at this particular promoter). So it was only to be expected that a fusion of the MR NTD to GR LBD would lack the strong GR TAF.

Minor comments:

The authors might want to discuss RU486 (which binds to both GR and PR) with respect to the experiments shown in Figure 5.

Preprint Review

This preprint was reviewed using eLife’s Preprint Review service, which provides public peer reviews of manuscripts posted on bioRxiv for the benefit of the authors, readers, potential readers, and others interested in our assessment of the work. This review applies only to Version 1 of the preprint: https://www.biorxiv.org/content/10.1101/822718v1.full

Summary

This paper uses the sequences of and experiments involving the mineralocorticoid (MR) and glucocorticoid (GR) receptors from a cartilaginous fish (the elephant shark), a sister taxa to the ray-finned fish and terrestrial vertebrates, to investigate the early evolution of the specificity of these important steroid receptors.

The reviewers appreciate the value of studying the activity and specificity of steroid receptors (SR) from a taxon that diverged from its common ancestor with vertebrates close to 500 million years ago, but identified several important issues that they feel limit the impact of the manuscript. These are described in detail in the individual reviews.

1) In the interpretation of their experiments on the N-terminal domain (NTD), the authors conflate two things: the role of the NTD in transactivation and its role in ligand binding. This leads to a conclusion – that there is an allosteric interaction between the NTD and the ligand binding domain (LBD) – this is not demonstrated.

2) The in vitro characterization of the activity and steroid specificity of elephant shark SRs in an in vitro assay is a useful contribution. However, in the absence of a stronger relationship of these experimental observations to a biochemical mechanism of action, a specific evolutionary scenario, or to elephant shark physiology, the broader significance of these findings is unclear.

3) Some of the statistical analyses and evolutionary analyses need stronger support.

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

Rebuttal_ Preprint RC-2020-00156

We thank the editor for handling our manuscript and both reviewers for their constructive critiques. We provide below a detailed list of results already available and experiments we propose to perform to address the reviewers’ comments and improve the quality of our manuscript.

Reviewer #1

In this manuscript, Obacz et al. investigated the role of IRE1 signaling in regulating the recruitment of myeloid cells in glioblastoma multiforme (GBM) microenvironment. They show that inhibition of IRE1 signaling decreased polynuclear neutrophil (PN) infiltration to GBM tumors in an animal model; conversely, IRE1 activation correlated with higher expression of myeloid cells-attracting chemokines in GBM. They also show that IRE1-XBP1s pathway promotes proinflammatory chemokines in GBM tumor cells through upregulation of UBE2D3, which leads to degradation of the NFκB inhibitor IκB and activation of NFκB downstream signaling. Their finding of a novel IRE1/XBP1s/UBE2D3/NFκB axis is important for understanding the basis of pro-tumoral inflammation in GBM, potentially in other 'immune hot' cancers. The manuscript is well written and the conclusion is well supported by the experiments. However, there are a few critical points that need to be addressed to strengthen their study**.

We thank this reviewer for his/her positive comments on our work and for the suggestions made to improve its relevance

Review#1 point 1: In this study, the authors used the GBM primary cell line RADH87 with stable overexpression of wild-type (WT) IRE1 or a truncated IRE1 variant. The expression of wild-type IRE1 was confirmed by Western analysis (Figure S1D). However, the expression of truncated IRE1 variant was not shown.

Response 1.1. The expression on truncated IRE1 variant (designated as Q780* - 80 KDa) is shown in Fig.S1D, following the expression on wild-type (WT) IRE1 (110 KDa). This point will be indicated in the revised version of the supplemental figure legends.

In addition, without tunicamycin treatment, there was no visible difference in XBP1s expression between the cells expressing WT or the mutant IRE1.

Response 1.2. Under basal condition, XBP1 splicing is indeed limited and therefore, there is no detectable difference in XBP1s expression level between IRE1 WT and Q780*. In contrast, under tunicamycin treatment (acute stress), reduced XBP1 mRNA splicing is observed (Fig.S1D) thus confirming the functionality of the Q780* truncated form. Of note, RNAseq was performed on these cell lines and basal splicing was quantified showing that even though it is an event that occurs at low frequency, it is decreased in cells expressing the Q780* mutant (this information will be added in the revised manuscript, data are available and analyses ongoing).

In the Boyden chamber assay (Figure 1C, D), conditioned medium from these cells were used; it was not described whether the cells were treated (e.g. with tunicamycin) to activate the IRE1 pathway. ** Response 1.3. Cells were not treated with Tunicamycin to excluded the impact of other UPR arms in the induction of cytokines expression/myeloid cells attraction. As a consequence, it is the basal secretome (found in conditioned media) that was used in those experiments to evaluate cell migration. We have now strong evidences that blunting IRE1 signaling (either genetically or pharmacologically) has a strong impact on GBM cells proteome and in particular on their secretome even if under basal conditions (manuscript in preparation). This information together with the fact that basal XBP1 mRNA splicing is reduced in IRE1 signaling deficient (Q780* expressing) cells, indicate that in GBM cells, constitutive IRE1 activity contributes to modulate the composition of their secretome towards chemoattraction of myeloid cells. This point will be further detailed in the results and discussion sections of the revised manuscript.

Review#1 point 2: The evidence that the mRNA expression of UBE2D3 positively correlates with IRE1/XBP1s pathway is weak. First, In Figure 3D, the correlation between the mRNA expression of UBE2D3 and XBP1 does not seem strong. In addition, as XBP1 mRNA level does not reflect IRE1 activation (as opposed to that of XBP1s), the level of XBP1s instead of total XBP1 should be assessed. Furthermore, such correlation should be validated in additional GBM cohorts/datasets.

Response 2. We agree that the correlation between UBE2D3 and XBP1 mRNA levels in TCGA GBM cohort might not be strong. However data presented in Fig3D were significant. Values indicated in green were Pearson’s correlation values (r). This point will be included in the revised figure legends. Moreover, in the revised version of the manuscript we propose to directly correlate the levels of XBP1s mRNA with the expression levels of SYVN1, UBE2D3 and UBE2J1 mRNAs. These data are available from the RNAseq data obtained from the TCGA cohort and already used previously by us (Lhomond et al. Embo Mol Med 2018). In addition, following this observation we have carried out a number of experimental validations using both established and primary GBM cell lines with genetic modifications of XBP1/XBP1s expression as well as ER stress-dependent induction of XBP1s and we clearly demonstrated that XBP1s mRNA levels correlate with UBE2D3 mRNA expression levels (Fig.3G-H, Fig.S2D-E). In addition, in Fig3E using our IRE1 activity signature we have shown a strong correlation between UBE2D3 and XBP1s, which is even more robust than simply correlating the mRNA levels. Data are already available and analyses are ongoing.

Review#1 point 3: The results in Figure 3 indicated that XBP1s acts as a transcriptional regulator of UBE2D3 expression. However, it is not clear whether this effect in GBM cells is direct or indirect. Further experiments such as chromatin immunoprecipitation and reporter assays are required to clarify this point.

Response 3. We agree with this reviewer’s point. Although we have scrutinized the publicly available ChIPseq databases and found UB2D3 among potential XBP1-regulated genes, we did not validate this observation in our model. To address this point we propose to perform ChIP experiments in cells overexpressing a tagged form of XBP1s and validate the presence of UBE2D3 promoter fragments in our experimental system. Moreover, these experiments will also be carried out with endogenous XBP1s (in-house XBP1s antibodies Pluquet et al. Cancer Res. 2013) in our primary GBM lines under basal and ER stress conditions. At last, to further document this, luciferase reporter assays using the UBE2D3 promoter (whose length would be defined based on ChIP experiments and the presence of XBP1s binding sites) upstream the luciferase ORF could be performed. Both ChIP and reporter assays have to be performed.

Review#1 point 4: In addition to UBE2D3, the two other ubiquitin-protein ligases, SYVN1 and UBE2J1, may also be implicated in the degradation of IκB. Did the authors assess their potential role on IκB degradation in their model system?

Response 4. We thank this reviewer for this suggestion. We have previously tested the impact of SYVN1 on IkB degradation with results showing a lot of variation. Indeed even though the trend of our results indicated that SYVN1 silencing appeared to lead to a slight increase in IkB expression, they never reached statistical significance. Variability in the results might be due to the efficacy of SYVN1 silencing and as such we propose to repeat further these experiments with SYVN1 siRNA smart pools to improve silencing efficacy. Moreover, SYVN1 has been shown to also contribute to the ubiquitylation and degradation of IRE1 (Gao et al. Embo Rep 2008; Sun et al. Nat Cell Biol 2015) and has its expression regulated by IRE1 activity (Dibdiakova et al. Neurol Res 2019), it might represent as well a very interesting target to study. Regarding UBE2J1, the situation is less documented. However, it was shown that this E2 works together with SYVN1 in conserved manner to contribute to ERAD (Chen et al. Nat Plants 2016). As such it might also be interesting to test whether the silencing of UBE2J1 impacts on IkB expression. To sum up, we propose to test experimentally whether the silencing of UBE2J1 or SYVN1 or both together impacts on IkB expression (we need to perform the experiments).

Review#1 point 5: The authors only used ectopic expression of relevant proteins to test their hypothesis in U87 and RADH87 cells. It is necessary to validate these findings using siRNAs/inhibitors for IRE1 and UBE2D3 in a GBM cell line that expresses high levels of endogenous IRE1 and UBE2D3.

Response 5. We propose to test the effect of SYVN1 and UBE2J1 silencing on IkB expression in U87 and RADH87 cells in the revised version of the manuscript (see above). In addition to address this reviewer’s comment, we propose to use U87 and RADH87 cells overexpressing IRE1 (Lhomond et al. 2018) and treat them with MKC886, or with siUBE2D3 or with both and evaluate whether in those conditions the NFkB pathway is affected. These experiments should be carried out relatively easily provided that all the recombinant cell lines, drugs and siRNA are already available.

Review#1 point 6: In Figure 3I: The protein expression of UBE2D3 should be shown.

Response 6: We had included control experiments with UBE2B3 expression in FigS3B in the initial version of the manuscript. We will include UBE2D3 expression for Fig3I in the revised version of the manuscript (these data are already available).

Review#1 point 7: In the right panel of Figure 3I: What do the labels #1, 2, 5 mean? Clear descriptions should be provided in the figure legend.

Response 7. Those labels correspond to different RADH87 cell lines stably overexpressing UBE2D3 protein. The validation of UBE2D3 expression using Western blotting will be included in FigS3B of the revised version of the manuscript (data are already available).

Review#1 point 8: In Figure S1D: The expression levels of the truncated IRE1 variant should be shown.

Response 8. The expression on truncated IRE1 variant (designated as Q780*) is shown in Fig.S1D, following the expression on wild-type (WT) IRE1. This point will be indicated in the revised version of the supplemental figure legends.

======================================================================

Reviewer #2

In the current study, the authors generate evidence supporting a novel pathway downstream of IRE1α/XBP1s in GBM cells involving the activation of an E2-ubiquitin ligase, UBE2D3. In order to do this, they use a combination of patient derived and established cell lines engineered to overexpress IRE1 mutants, XBP1s or UBE2D3. They claim that UBE2D3 is upregulated downstream of XBP1s in GBM cells, and functions to activate NF-kB through the degradation of IkB, thus promoting CXCL2/IL-6/IL-8 production and the subsequent recruitment of monocytes and polymorphonuclear (PN) cells to the tumor microenvironment. However, the article has major shortcomings that need to be addressed before considering its publication

We thank this reviewer for his/her constructive comments on our work.

Review#2 point 1: Fig. 1: Classification of immune cells infiltrating GBM. The characterization of immune infiltrate in GBM is too simplistic. Monocytes, monocyte-derived macrophages and microglia are treated as equivalents along the text (IBA1+), making the story hard to follow. At least in mice, these populations can be easily distinguished based on CD45/CD11b/Ly6C expression (see for example Zhihong Chen et al., Cancer Research, 2017). Can the authors further analyze which of those population are actually affected under IRE1 deficiency and/or UBE2D3 overexpression? On the other hand, it is rather questionable that all CD11b negative cells are exclusively T cells, as suggested in Fig 1B. Can the authors provide evidence and/or references to support their gating strategies?

Response 1: We thank the reviewer for this comment. Our objective was to test the impact of IRE1 modulation on the infiltration of myeloid cells in the tumor, and we did not plan to describe this effect on the complete and detailed infiltrating myeloid populations in GBM which could represent a full study on its own. However, to address this reviewer’s critique we propose to complete the characterization of the myeloid population in our mouse model using IHC by adding Ly6C staining for macrophages and granulocytes. We did not select flow cytometry approach to explore this point as suggested by the reviewer (Cheng, Cancer Res, 2017), but instead IHC was preferred as we thought that the localization of the infiltrated immune cells was important to evaluate (periphery vs. core of the tumor). The information about the localization of immune cells is already available and will be added to the revised manuscript. Concerning the second point raised by this reviewer, the strategy to characterize the immune population in human GBM specimen was to combine CD45 and CD11b markers as previously described by Hussain et al. Neuro-Oncol 2006 and Parney et al. J Neurosurg 2009. Moreover, the analysis of additional markers allowed us to confirm that CD45+ CD11b+ cells were mainly monocytic cells (that also co-expressed CD14, CD168, CD64 and HLA-DR); CD45 low CD11b high cells were granulocytes (CD66B, CD15 and CD16); and CD45 high CD11b low cells were mainly CD3+ T cells. These data are already available and will be added to the revised manuscript.

Review#2 point 2: Fig. 1: RADH IRE1 Q780\ model - Can the authors further validate the IRE1 deficiency of their model cell line RADH87 IRE1Q780*? It appears to have severely reduced IRE1 levels when compared to the RAD87-IRE1WT cell line (figS1D). Furthermore, the WT and not the truncated form seems to be predominantly expressed. Intriguingly, XBP1 is still being spliced after tunicamycin treatment in this mutant line. All these results differ significantly from the U87-Q780* cell line originally published by Lhomond et al., 2018. Can the authors comment on these differences? Was there a mixture in cell lines? *

Response 2: We agree with the reviewer that the level of IRE1Q780* expression on RADH87 cells is lower than the IRE1WT expression (Fig.S1D). As observed by this reviewer, XBP1 was still spliced in Q780* cells but XBP1s expression was reduced as shown in Figure S1D. This is mostly due to the ratio between the expression endogenous IRE1 and that of Q780*, which as previously shown (Lhomond et al; 2018) acts as a dominant negative and preempts endogenous IRE1 signaling. The differences observed are also probably due to the methods used, indeed we measured XBP1 and XBP1s mRNA expression in U87 cells (Lhomond et al. 2018), whereas XBP1s protein expression was used with RADH87 cells (introducing the RNA translation parameter that was not monitored in U87 cells). Differences could be also linked to the cell lines as we used the U87 immortalized and RADH87 primary cell lines.

Review#2 point 3: Fig. 1: Impact of IRE1 inhibition on recruitment of myeloid cells to the TME. The experiment in figure 1E-F, which is the only in vivo evidence supporting a role of IRE1 signaling on myeloid cell recruitment, is very hard to interpret. The authors show no evidence that IRE1 is being inhibited under the treatment and if so, up to which extent. Furthermore, what are the cells targeted by MKC in this setting? The differences in the infiltration of PN cells seem very slight, nothing is mentioned regarding the number of mice per group, or the statistical analysis performed. I would suggest performing a simpler experiment to demonstrate an intrinsic effect of IRE1 signaling in GBM cells, comparing the recruitment of myeloid cells in tumors generated by GL261 cells expressing WT vs deficient forms of IRE1.

Response 3: The mouse model used in the paper is fully described in (Le Reste BioRxiv 2020 - doi: https://doi.org/10.1101/841296) and all the details about the procedures can be found in this manuscript. This model was developed to recapitulate in mice the standard of care for GBM patient including surgical resection. In addition, drug delivery in brain tumors is often an issue due to the blood-brain barrier. Therefore, the IRE1 inhibitor was delivered locally after resection of the tumor, exposing both tumor and stromal cells. To quantify the myeloid cell recruitment in Fig1E-F, at least thirty random fields from tumor tissue and at least thirty random fields from tumor periphery were quantified for control (PLUG) and MKC-treated group (2 mice/group). The number of positive cells in tumor tissue and tumor periphery were then pulled together for statistical analyses. The significance of the differences in myeloid cells recruitment between control (PLUG) and MKC-treated group was estimated using unpaired student t-test. At least 8 tumors of each group were analyzed comprising 2 to 3 sections of each and 10 fields per section. In addition, we have also performed the experiments using GL261 cells knockout for IRE1, the data are already available and could be possibly added to the revised manuscript.

Review#2 point 4: Fig. 2: Correlation between IRE1 signature and cytokine/chemokine signature. In the IRE1 signature as determined in the EMBO Mol Med paper (and to which the authors continuously refer) 6 out of 38 (15%) of the genes correspond to cytokines and/or chemokines (Il6, Il1b, Cxcl2, Cxcl5 and Ccl20) (Lhomond et al., 2018). Besides the fact that it is very unclear how this signature was obtained in the first place, it is rather surprising that in the current paper the authors correlate this "IRE1 activity" signature with the same or other cytokines/chemokines mRNA levels and come to the conclusion that there is a high correlation (fig 2A). Isn't this to be expected? Can the authors clearly explain how the IRE1 signature was determined and prove that their "IRE1 signature" is, in fact, representing IRE1 activity? For instance, it is important to cross validate their results by using an independent signature of IRE1 activity (e.g. ChipSeq XBP1s targets, Chen et al., 2014)?

Response 4: We thank this reviewer for asking for precisions about the procedure. The IRE1 signature was fully described in Lhomond et al. 2018 and was obtained from transcriptome datasets obtained from U87 modified for IRE1 activity (Pluquet et al., 2013). IRE1 was validated on GBM patients and appeared as an important tool to evaluate IRE1 activity in tumor specimen not only in GBM but also in other tumor types (Rubio-Patiño C, Cell Metab 2018). Furthermore, IRE1 activity was also directly linked to the pro-inflammatory tumor cell secretome in various studies such as Logue et al. 2018. As indicated by this, some cytokines/chemokines studied in this work were indeed part of the IRE1 signature and correlation between this signature and their expression was indeed expected. However the other main cytokines/chemokines studied here were not present in the IRE1 signature indicating that IRE1 could have been involved in the regulation of their expression. As proposed by reviewer#2, we will include in the revised version of the manuscript the analysis of cytokines/chemokines from the dataset ChipSeq XBP1s targets (Chen et al. 2014), although this study was performed on breast tumors.

Review#2 point 5: Fig 2: XBP1s controlling cytokines/chemokines expression in GBM cells - As suggested by the data on fig1C-D and fig2E, IRE1 appears to be constitutively active in GBM cells, as IRE1 deficiency is sufficient to cause a defect in chemokine production. However, as shown in fig S1D, XBP1s protein was not detected under basal conditions, suggesting that the deficiency in chemokine production in IRE1-deficient cell lines is XBP1s-independent. Can the authors further discuss these results?

Response 5: We thank the reviewer for commenting this point. We think that indeed IRE1 is constitutively active in GBM cells. As we have tested XBP1s protein expression in untreated and tunicamycin-treated RAD87 cells (FigS1D), and we will also provide real time qPCR data to show the presence of basal XBP1s mRNA levels (data already available). We agree that the way we presented the results are misleading and undermine the basal expression of XBP1s. This will be fixed in the revised manuscript.

Review#2 point 6: Fig 3: IRE1/XBP1s/UBE2D3/NF-kB axis - Authors must show the activation status of NF-kB in parental U87 cells (Fig3A), as this is a critical evidence to support that IRE1a-deficient U87-DN cells are defective in chemokine production due to an impairment in NF-kB signaling. In addition, even when tunicamycin treatment induce XBP1s and UBE2D3 (figS2D) it does not induce IkB degradation nor NF-kB phosphorylation in parental U87 and RADH87 cells (figS3C) as one should expect if IRE1/XBP1s/UBE2D3/NF-kB pathway is operating in these cells. How can this be explained? Only after XBP1s or UBE2D3 overexpression, NF-kB signaling appears to be affected.

Response 6: As shown in Fig3A, U87 cells deficient for IRE1 signaling (DN) exhibit decreased NFkB signaling as exemplified by decreased phospho-NFkB and phospho-IkB compared to control U87 cells proficient for IRE1 signaling. In our manuscript, we mainly focused on the activation of the IRE1/XBP1s/UBE2D3/NFkB signaling axis under basal condition. One could speculate that tunicamycin treatment leads to a strong stress response that others mechanisms are activated that overwhelm the IRE1/XBP1s/UBE2D3 pathway we are describing herein. For instance, it has been demonstrated that the IRE1/JNK signaling was linked to NFkB activation upon acute ER stress (Tam et al. PLoS One. 2012;7(10):e45078; Schmitz et al. Biomedicines. 2018 Jun; 6(2): 58.) and furthermore PERK activation upon thapsigargin or tunicamycin treatment was also found to promote NFkB activation (Deng et al. DOI: 10.1128/MCB.24.23.10161-10168.2004; Fan et al. Cell Death Discov. 2018 Feb 12;4:15). We believe that the pathway we describe here might be linked to constitutive activation of IRE1 signaling (proper to tumor cells) rather than acute activation of this pathway and be compatible with sustained proliferation. To further document this point, we have already generated data about the phosphorylation status of NFKB in GL261 cells KO for IRE1 compared to the parental cells (data will be provided in the revised version of the manuscript). In addition we are currently investigating the correlation between IRE1 activity signature and that of NFkB as defined previously (Jin et al. Cancer Res. 2014 May 15; 74(10): 2763–2772.), results should be available shortly and will be added in the revised manuscript.

Review#2 point 7: Fig 4: UBE2D3 and MIB1 – The authors should discuss better what is the possible interaction between UBE2D3 and MIB1. As shown in fig4G, silencing of MIB1 cause a severe increase in UBE2D3 protein levels but this is not commented in the text.

Response 7: We thank the reviewer for this comment. We believe that MIB1 might also controls the expression of UBE2D3. The data are already available and will be included in the revised version of the manuscript.

Review#2 point 8: Fig 6: Chemokines driving recruitment of myeloid cells to UBE2D3 overexpressing tumors. A formal demonstration that GL261-UBE2D3 tumors recruit higher numbers of MM and PNs through an enhanced production of CXCL2, IL-6 and/or IL-8 is lacking. For instance, they could compare the infiltration of myeloid cells in GL261-UBE2D3 vs GL261-UBE2D3-CXCL2KO tumors.

Response 8: To address this point, we propose to test the expression of these cytokines/chemokines in the GL261 tumors after resection using ELISA. These experiments could be carried out in IRE1 KO tumors, in UBE2D3 overexpressing tumors and performed for instance using perfusion of CXCL2, IL6 or IL8 neutralizing antibodies or cells KO for these chemokines. These experiments could be performed but might lead to inconclusive results (not statistically significant) if there is redundancy between the roles of those chemokines. As such, we think that we could provide in vitro information about the respective roles of these chemokines in recruiting MM and PNs but that at present stage the in vivo demonstration is to premature.

Review#2 point 9: Authors must provide replicates of the blots to sustain their claims: FigS1D, Fig3A, Fig3I, Fig4G.

Response 9: Replicates and quantifications are already available and will be provided in the revised version of the manuscript.

Review#2 point 10: The authors should include a better description of the methods regarding bioinformatic analysis. For instance, which genes where used for MM/PN/T cell signatures in fig1A/S1A?.

Response 10: We thank the reviewer#2. This information is available and a complete description will be included in the revised version of the manuscript.

Review#2 point 11: Missing statistical significance on fig 2C and fig 6A to support their claims.

Response 11: Statistical values will be included in the revised manuscript.

Review#2 point 12: Fig2F is presented in the text as mRNA levels but in the figure as protein levels.

Response 12: This point will be fixed in the revised version of the manuscript.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary: your understanding of the study and its conclusions.

In the current study, the authors generate evidence supporting a novel pathway downstream of IRE1α/XBP1s in GBM cells involving the activation of an E2-ubiquitin ligase, UBE2D3. In order to do this, they use a combination of patient derived and established cell lines engineered to overexpress IRE1 mutants, XBP1s or UBE2D3. They claim that UBE2D3 is upregulated downstream of XBP1s in GBM cells, and functions to activate NF-kB through the degradation of IkB, thus promoting CXCL2/IL-6/IL-8 production and the subsequent recruitment of monocytes and polymorphonuclear (PN) cells to the tumor microenvironment. However, the article has major shortcomings that need to be addressed before considering its publication

Major comments: major issues affecting the conclusions.

-Fig. 1: Classification of immune cells infiltrating GBM The characterization of immune infiltrate in GBM is too simplistic. Monocytes, monocyte-derived macrophages and microglia are treated as equivalents along the text (IBA1+), making the story hard to follow. At least in mice, these populations can be easily distinguished based on CD45/CD11b/Ly6C expression (see for example Zhihong Chen et al., Cancer Research, 2017). Can the authors further analyze which of those population are actually affected under IRE1 deficiency and/or UBE2D3 overexpression? On the other hand, it is rather questionable that all CD11b negative cells are exclusively T cells, as suggested in Fig 1B. Can the authors provide evidence and/or references to support their gating strategies?

-Fig. 1: RADH IRE1 Q780 model Can the authors further validate the IRE1 deficiency of their model cell line RADH87 IRE1Q780? It appears to have severely reduced IRE1 levels when compared to the RAD87-IRE1WT cell line (figS1D). Furthermore, the WT and not the truncated form seems to be predominantly expressed. Intriguingly, XBP1 is still being spliced after tunicamycin treatment in this mutant line. All these results differ significantly from the U87-Q780* cell line originally published by Lhomond et al., 2018. Can the authors comment on these differences? Was there a mixture in cell lines?

-Fig. 1: Impact of IRE1 inhibition on recruitment of myeloid cells to the TME. The experiment in figure 1E-F, which is the only in vivo evidence supporting a role of IRE1 signaling on myeloid cell recruitment, is very hard to interpret. The authors show no evidence that IRE1 is being inhibited under the treatment and if so, up to which extent. Furthermore, what are the cells targeted by MKC in this setting? The differences in the infiltration of PN cells seem very slight, nothing is mentioned regarding the number of mice per group, or the statistical analysis performed. I would suggest performing a simpler experiment to demonstrate an intrinsic effect of IRE1 signaling in GBM cells, comparing the recruitment of myeloid cells in tumors generated by GL261 cells expressing WT vs deficient forms of IRE1.

-Fig. 2: Correlation between IRE1 signature and cytokine/chemokine signature In the IRE1 signature as determined in the EMBO Mol Med paper (and to which the authors continuously refer) 6 out of 38 (15%) of the genes correspond to cytokines and/or chemokines(Il6, Il1b, Cxcl2, Cxcl5 and Ccl20) (Lhomond et al., 2018). Besides the fact that it is very unclear how this signature was obtained in the first place, it is rather surprising that in the current paper the authors correlate this "IRE1 activity" signature with the same or other cytokines/chemokines mRNA levels and come to the conclusion that there is a high correlation(fig 2A). Isn't this to be expected? Can the authors clearly explain how the IRE1 signature was determined and prove that their "IRE1 signature" is, in fact, representing IRE1 activity? For instance, it is important to cross validate their results by using an independent signature of IRE1 activity (e.g. ChipSeq XBP1s targets, Chen et al., 2014)?

-Fig 2: XBP1s controlling cytokines/chemokines expression in GBM cells As suggested by the data on fig1C-D and fig2E, IRE1 appears to be constitutively active in GBM cells, as IRE1 deficiency is sufficient to cause a defect in chemokine production. However, as shown in fig S1D, XBP1s protein was not detected under basal conditions, suggesting that the deficiency in chemokine production in IRE1-deficient cell lines is XBP1s-independent. Can the authors further discuss these results?

-Fig 3: IRE1/XBP1s/UBE2D3/NF-kB axis Authors must show the activation status of NF-kB in parental U87 cells (Fig3A), as this is a critical evidence to support that IRE1a-deficient U87-DN cells are defective in chemokine production due to an impairment in NF-kB signaling. In addition, even when tunicamycin treatment induce XBP1s and UBE2D3 (figS2D) it does not induce IkB degradation nor NF-kB phosphorylation in parental U87 and RADH87 cells (figS3C) as one should expect if IRE1/XBP1s/UBE2D3/NF-kB pathway is operating in these cells. How can this be explained? Only after XBP1s or UBE2D3 overexpression, NF-kB signaling appears to be affected.

-Fig 4: UBE2D3 and MIB1 The authors should discuss better what is the possible interaction between UBE2D3 and MIB1. As shown in fig4G, silencing of MIB1 cause a severe increase in UBE2D3 protein levels but this is not commented in the text.

-Fig 6: Chemokines driving recruitment of myeloid cells to UBE2D3 overexpressing tumors.<br> A formal demonstration that GL261-UBE2D3 tumors recruit higher numbers of MM and PNs through an enhanced production of CXCL2, IL-6 and/or IL-8 is lacking. For instance, they could compare the infiltration of myeloid cells in GL261-UBE2D3 vs GL261-UBE2D3-CXCL2KO tumors.

Minor comments: important issues that can confidently be addressed.

Authors must provide replicates of the blots to sustain their claims: FigS1D, Fig3A, Fig3I, Fig4G. The authors should include a better description of the methods regarding bioinformatic analysis. For instance, which genes where used for MM/PN/T cell signatures in fig1A/S1A? Missing statistical significance on fig 2C and fig 6A to support their claims. Fig2F is presented in the text as mRNA levels but in the figure as protein levels.

Significance

Significance

In general, there is a clear interest both from academia and pharma companies to understand the role of the UPR in tumor biology and how the UPR shapes the immune compartment. This is highly relevant as the UPR is a novel drug target in cancer therapy, but unfortunately many inconsistent data are around. However, as the paper is now, it will not contribute to clarify these inconsistencies.

Compare to existing published knowledge.

Unfortunately, there are many studies around with inconclusive results and strong claims based on poorly validated tools.

Audience. Tumor immunologists, UPR field

Your expertise. Role of the UPR in immune cells and anti-tumor 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

Summary

In this manuscript, Obacz et al. investigated the role of IRE1 singaling in regulating the recruitment of myeloid cells in glioblastoma multiforme (GBM) microenvironment. They show that inhibition of IRE1 signaling decreased polynuclear neutrophil (PN) infiltration to GBM tumors in ananimal model; conversely, IRE1 activation correlated with higher expression of myeloid cells-attracting chemokines in GBM. They also show that IRE1-XBP1s pathway promotes proinflammatory chemokines in GBM tumor cells through upregulation of UBE2D3, which leads to degradation of the NFκB inhibitor IκB and activation of NFκB downstream signaling. Their finding of a novel IRE1/XBP1s/UBE2D3/NFκB axis is important for understanding the basis of pro-tumoral inflammation in GBM, potentially in other 'immune hot' cancers. The manuscript is well written and the conclusion is well supported by the experiments. However, there are a few critical points that need to be addressed to strengthen their study.

Major comments:

1.In this study, the authors used the GBM primary cell line RADH87 with stable overexpression of wild-type (WT) IRE1 or a truncated IRE1 variant. The expression of wild-type IRE1 was confirmed by Western analysis (Figure S1D). However, the expression of truncated IRE1 variant was not shown. In addition, without tunicamycin treatment, there was no visible difference in XBP1s expression between the cells expressing WT or the mutant IRE1. In the Boyden chamber assay (Figure 1C, D), conditioned medium from these cells were used; it was not described whether the cells were treated (e.g. with tunicamycin) to activate the IRE1 pathway.

2.The evidence that the mRNA expression of UBE2D3 positively correlates with IRE1/XBP1s pathway is weak. First, In Figure 3D, the correlation between the mRNA expression of UBE2D3 and XBP1 does not seem strong. In addition, as XBP1 mRNA level does not reflect IRE1 activation (as opposed to that of XBP1s), the level of XBP1s instead of total XBP1 should be assessed. Furthermore, such correlation should be validated in additional GBM cohorts/datasets.

3.The results in Figure 3 indicated that XBP1s acts as a transcriptional regulator of UBE2D3 expression. However, it is not clear whether this effect in GBM cells is direct or indirect. Further experiments such as chromatin immunoprecipitation and reporter assays are required to clarify this point.

4.In addition to UBE2D3, the two other ubiquitin-protein ligases, SYVN1 and UBE2J1, may also be implicated in the degradation of IκB. Did the authors assess their potential role on IκB degradation in their model system?

5.The authors only used ectopic expression of relevant proteins to test their hypothesis in U87 and RADH87 cells. It is necessary to validate these findings using siRNAs/inhibitors for IRE1 and UBE2D3 in a GBM cell line that expresses high levels of endogenous IRE1 and UBE2D3.

Minor comments:

1.In Figure 3I: The protein expression of UBE2D3 should be shown.

2.In the right panel of Figure 3I: What do the labels #1, 2, 5 mean? Clear descriptions should be provided in the figure legend.

3.In Figure S1D: The expression levels of the truncated IRE1 variant should be shown.

Significance

In this manuscript, the authors report some of the molecular mechanisms by which IRE1-XBP1s signaling controls GBM immune infiltration. They show that a novel IRE1/UBE2D3 signaling axis, mediated by XBP1s, regulates NF-κB activation, which subsequently promotes pro-inflammatory responses and the recruitment of immune/inflammatory cells to the tumor site. This study provides significant new information on the role of IRE1 in GBM. The findings also establish a basis for potential new approaches to improve the efficacy of current immunotherapies, also in other cancer types, which needs to be further explored.

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 report, van Schaik et al., modified an established CUT and RUN method and combined it with previously used DamID to identify Lamin Associated Domains (LADs) with better temporal resolution. Previous DamID experiments labeled locations where lamin proteins were present within a 5-25 hour window while the new technique, pA-DamID, labels DNA within a 30 minute window providing better temporal resolution. The authors used this technique to identify LADs at multiple stages of the cell cycle and applied this protocol to different cell types. The authors FIND differences when comparing data sets between cell cycle time points and cell lines.

We thank the reviewer for the helpful comments.

**Major points:**

1) The data sets generated and displayed in this manuscript seem incomplete. In Figure 1G, the authors compare lamin B2 vs. lamin B1 generated LADs in HAP-1 cells and lamin A/C vs lamin B2 LADs in hTERT-RPE cells. In figure S4, panel C compares lamin B1 and lamin B2 in K562 cells and lamin B2 and lamin A/C in hTERT-RPE cells. It would have been informative to have a complete dataset for lamin B1, lamin B2, and lamin A/C identified LADs in all cell lines analyzed. The information provided from these datasets would be useful to the scientific community.

We did not think it was necessary to generate every lamin pA-DamID data set in every cell line, given that previous DamID studies indicated that lamins A, B1 and B2 give the same genome-wide pattern {Meuleman, 2013, 23124521; Kind, 2014; 24717229}. However, we agree with the reviewer that the missing data sets lead to a sense of incompleteness and might distract the reader from the main message of the manuscript. We suggest to generate Lamin B1 pA-DamID in hTERT-RPE cells– provided that the current Corona virus shutdown will not prevent us from doing this experiment. Doing so, we 1) have a complete lamin data set in hTERT-RPE cells, which we study in most detail in this manuscript 2) can compare all lamins within the same cell type 3) can compare all Lamin B1 DamID data to the corresponding Lamin B1 pA-DamID data.

2) The authors discovered that LADs reposition during progression through the cell cycle. It would have been interesting to know whether these changes have transcriptional consequences? One could perform RNA-SEQ experiments to discover if LAD occupancy results in transcriptional changes and choose a few genes to confirm the findings with RT-PCR. Is this the same for lamin B1, lamin B2, and lamin A/C occupied LADs? Analyze if there are any genomic features such as CTCF or transcription factor binding sites that correlate with the loss of LADs.

In the first part of this point, the reviewer suggests to look at transcriptional consequences of changes in NL interactions. To address this point, we require some measure of nascenttranscription during the cell cycle, which is not available in any of the studied cell lines. A potential experiment would be to map polymerase occupancy with pA-DamID / CUT&RUN or run-on transcription with any other method at the synchronized time points. However, this experiment is not trivial and we feel that this goes beyond the scope of this manuscript, which focuses on the development of pA-DamID and the m6A-Tracer with a proof-of-principle example of NL binding dynamics during the cell cycle.

In the second part of this point, the reviewer asks whether changes in NL binding correlate with genomic features such as CTCF binding sites or transcription factor binding sites. In the manuscript, we already include correlations with various active features (active gene density / replication timing) (Fig. 3E-G, 4C-E), that generally correlate well with transcription factor binding. We have added CTCF peaks as comparison (Fig. S7F).

3) The authors state that using H3K27me3/H3K9me3 in pa-DamID showed no enrichment. This is surprising considering that both modifications are enriched in heterochromatin and at the nuclear periphery. It appears that the peripheral enrichment is masked by the larger overall internal pool. The authors should discuss this observation and comment on the sensitivity of the method to detect local enrichment versus the global levels of a protein or modification in pa-DamID.

We believe that H3K27me3 and H3K9me3 histone modifications show the expected pattern in their distribution in the nucleus. However, due to the peripheral mask slightly extending beyond the cell boundaries, the calculated peripheral enrichment is underestimated. This has been better described in the figure legend.There is a small enrichment at the nuclear periphery compared to diffuse Dam and untargeted pA-Dam (Fig. 1B/1C/1F). To further support the pA-DamID data quality of these histone modifications, we have added a comparison with ENCODE ChIP-seq data tracks in K562 cells (Fig. S3C).

**Minor points:**

Figure 1: Change colors for Figure 1F and Figure 2D. The colors are hard to discern.

Figure 2B: Please mark which antibody was used for this analysis.

Figure 2C: Please also overlay data from pA-DamID lamin A/C experiments.

Figure 4: Please mention which antibody was used for the pA-DamID experiments used to generate this dataset.

Figure 5: Please mention which antibody was used for the pA-DamID experiments used to generate this dataset.

Figure S5 C and D: Please mention which antibody was used for the pA-DamID experiments.

We have made edits to address the minor comments above. However, we do not have Lamin A/C data in HAP-1 and K562 cells to add to Fig. 2C.

Reviewer #1 (Significance (Required)):

The major contribution of this manuscript is the description of an improved method to map LADs. This is a valuable contribution. By using this new method, the findings of this paper provide some new insight in LAD dynamics throughout the cell cycle although the experiments are largely phenomenological. This is a technically sound study.

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

The paper describes a new method for detecting Lamin associated DNA domains, which allows better time resolution than classical DamId. It is a good idea and its functionality is demonstrated in tissue culture cells. There are minor insights but it is important that we advance the field with new and better technologies, thus this version amply suffices to give evidence of that.

We thank the reviewer for the positive feedback.

Reviewer #2 (Significance (Required)):

The audience is all persons working on chromatin organization in the nucleus, which is a large audience. The data are clear as they basically are proof of principle for a new technique. There is nothing major to request as revision. They might cite papers on damID in worms and tissue specific applications of this in living organisms, as this is likely to be the situation that is most interesting in the long run. The resolution (in bp) would be interesting to know and validate.

We have extended the discussion on new applications of pA-DamID.

We now compare data quality and resolution between DamID and pA-DamID, focusing on the mapping of NL interactions (Fig. S4D-E).These plots indicate similar data quality and resolution between the two methods.

I have no other major revisions to request.

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

In the manuscript by Schaik et al (Van Steensel laboratory) the authors describe a very clever approach to identifying Lamina Associated Domains (LADs) using the principles of the 'cut-n-run' strategy. Specifically, they engineer the Dam methyltransferase used in canonical DamID in frame with a protein A moiety capable of interacting with an antibody (in this case lamins--B1, B2 and A/C). After permeabilization, cells are incubated with antibodies, then pA-Dam purified protein--for a brief time window--is added to mark associated DNA with GmATC. This technique is a valuable contribution to the field, particularly since, as the authors point out,an advantage of pA-DamID is that the labeled DNA can also be visualized in situ using the m6A-Tracer, before this DNA is sequenced. This allows for validation of findings and is highly amenable to cell sorting technologies. In addition, this technology allows for a time-resolved measure of LADs not currently available by standard DamID. The authors apply this technology to four different cell types. They noted that the 'maps' generated by the is technology differed from canonical DamID at very specific regions (small LADs in very localized regions) . They then embark on a series of experiments to show that these differences arise from cell cycle -related differences that are differentially picked up by the methods--with the pA-DamID allowing for dissection of more discrete cell cycle stages/configurations. In general they find an initial preference for sub-telomeric LADs to associate with the nuclear lamina fist, then more centromeric. There is some data suggesting loss/gain of LADs in specific regions/with specific features. The manuscript is well written and the data well presented. However, there are some points that need to be addressed . Overall, there is some oversimplification or omission of previous data in the field, a lack of clarity in how some of the data was interpreted, and some areas where clarification and/or additional analyses would be helpful. I sincerely hope the authors find the following critiques to be useful. Thank you for the opportunity to review your very nice work.

We thank the reviewer for the constructive and very detailed comments, these have been extremely helpful in improving the manuscript.

**Introduction:**

**Microscopy studies found that telomeres are enriched near the NL in early G1 phase, leading to the hypothesis that telomeres may assist in NL reassembly onto chromatin [13].**

● There have been numerous studies identifying the timing and disposition of INM proteins and Lamins at m the end of mitosis (during NE reformation). Why are you citing just this one? (e.g. ​Thomas Dechat et al., 2004; T. Dechat et al., 2000; Ellenberg et al., 1997; Haraguchi et al., 2001​)

We have expanded the introduction to better cover previous work on the reforming NL (paragraph 2) and initial genomic interactions with the NL (paragraph 3).

**Furthermore, during S-phase B-type lamins have been found to transiently overlap with replication foci in the nuclear interior, at least in some cell types [20]**

● While, technically, this has indeed been reported, this study is from 1994 and has not been repeated. The cells used in this study (3T3 fibroblasts) are widely used and others have not noted this phenomenon. Soften this.

**Other studies have indicated that lamins are important for DNA replication [reviewed in 21].**

● Likewise, direct roles for lamins in replication are controversial (acknowledged in the small section of the cited review on the role of lamins in replication).

● Perhaps combine the two sentences above to soften the implication that this is a "known" role of B-type lamins. e.g. "A handful of studies have implicated a role for B-type lamins in replication, but the direct role of the lamina in this process remains unclear. Nonetheless, ......"

This is a very good suggestion by the reviewer. We agree that literature has been controversial and should be approached with care. We have followed the advice and changed this.

**Results:**

**So far, the cell cycle dynamics of genome - NL interactions have primarily been studied by microscopy. While these studies have been highly informative, they were often limited to a ​few selected loci.​**

● Please cite your own study (Kind et al.) and other recent papers (Luperchio et al.-​https://www.biorxiv.org/content/10.1101/481598v1​;; Zhang et al., Nature-​https://www.nature.com/articles/s41586-019-1778-y​;) in which they were either 1) not limited to a few selected loci and/or 2) not microscopy-directed studies? There is an argument to be made here for the resolution (time and b.p.) you have achieved through your studies that these studies did not.

To our knowledge, there have been no high-throughput microscopy studies of many individual loci performed studying this. Microscopy has been performed of collective sequences (i.e. all LADs (Kind, 2013 and indeed Luperchio, 2018)), which provide additional insights but lack sequence information in the images. We have expanded the introduction to better acknowledge these microscopy studies that are not limited to single loci.We feel that observations on LAD domain clustering (Luperchio) and B compartment formation (Zhang) are better suited for the Discussion, given that these observations are not directly related to genome – NL contact dynamics. We already discussed B compartment formation in the discussion, but now also include the observed LAD domain clustering. Also, we have discussed data resolution in more detail in the results (see reviewer #2).

How does this data correlate with TSA-seq, another antibody-based method developed by the Belmont lab, but collaboratively developed for use in identifying LADs (ie Dam alternative) with the Van Steensel group?​ I can imagine there are numerous advantages to this approach (radius of "labeling" being one).

TSA-seq provides a different perspective on genome – NL interactions, given its distance dependence rather than contact. We have added a comparison with TSA-seq to the Discussion.

**When Dam-Lamin B1 is expressed in vivo for 5-25 hours during interphase, LADs that interact with the NL become progressively labeled, eventually resulting in a layer of labeled chromatin of up to ~1 μm thick [8]. This is because LADs are in dynamic contact with the NL. We expected that in pA-DamID this layer would be thinner, because the NL-tethered Dam is only activated for 30 minutes. In addition, permeabilization depletes small molecules including ATP and thus prevents active DNA remodeling in the nucleus [26]. Indeed, pA-DamID yields a m6A layer that is ~2.5 fold thinner than the layer in cells that express Dam-Lamin B1 in vivo (Fig. S2A-C). This is not an artifact due to collapse of chromatin onto the NL caused by the permeabilization, because permeabilization of cells expressing Dam-Lamin B1 in vivo did not significantly reduce the thickness of the m6A layer compared to directly fixed cells (Fig. S2C). The thin layer of labeled DNA obtained by pA-DamID points to an improved temporal resolution of pA-DamID compared to conventional DamID.**

● I think this requires a bit more care. Your previous work clearly demonstrates LADs are dynamic. Others in the field have shown that these domains are also constrained within the larger sub-chromosomal compartment (self-interaction) of LADs (e.g. Luperchio 2018) within a chromosome. So, this is truly a temporal "snapshot" that may miss some regions of LADs that are less directly (or more dynamically) associated with the lamina, but still compartmentalized into the larger LAD sub-chromosomal compartment. It is unclear if the treatment used for this study perturbs these LAD-lamina​ dynamic​ interactions--one can imagine that the LADs are much less mobile generally under the protocol described in your supplemental information. In other words, ​LADs don't collapse, nor do they behave in the same way they would after permeabilization​. The technique has compromised some of that --which is actually fine for most of the purposes in this manuscript, but this needs to be discussed.

As the reviewer points out, there are fundamental differences between DamID and pA-DamID in their m6A deposition that should be clear from the text. We elaborated on this in the comparison between pA-DamID and DamID.

● In addtion, imaging data showing dam-LaminB1/2 plus m6A-tracer is missing (figure S2). This should be included. Is the intensity of the "tracer" similar between conditions? If so, were the exposures kept constant in all images? This is important since the decay rate is highly related to intensity of signal.

We are afraid that this figure has been misinterpreted. We have changed the figure labels and legend to explain it better. The HT1080 Dam-Lamin B1 clonal cells (new clone kindly supplied by Jop Kind) still showed significant variation in m6A-Tracer intensity per cell, suggesting different expression levels of Dam-Lamin B1. To create optimal images for halfway decay estimation, laser settings were changed between images. This has now been mentioned more clearly in the methods.

**In some cell types, especially in​ HCT116 and ​hTERT-RPE​ cells, we noted local discrepancies between the two methods (Fig. 2A,bottom panel). These differences involve mostly regions with low signals in DamID that have higher signals in pA-DamID. However, such differences are not obvious in HAP-1 and K562 cells.**

● Only HCT116 data is shown in the indicated figure. hTERT-RPE cells are shown in the accompanying supplemental figure and use a different antibody (lamin B2) as the target for the pA-Dam.

We have changed the pointer to include the supplementary figure.

(See reviewer #1 for a similar comment.)We agree that the comparison between Lamin B1 DamID and Lamin B2 pA-DamID in hTERT-RPE cells leads to sense of incompleteness and confusion. We suggest to generate Lamin B1 pA-DamID data in hTERT-RPE cells to solve this – provided that the current Corona virus shutdown will not prevent us from doing this experiment.

This brings up another point: the data (log2 ratio schema) shown in figure 2 is for HCT116 lamin B1 pA-Dam. Yet, the subsequent studies for transient/building interactions during G1 and into S (Figure 3) are done in hTERT-RPE cells using lamin B2. To be consistent, data from lamin B2 should be used in both figures (it seems lamin B2 data is available for all cell types). The comparison of Dam-Lamin B1 can be addressed in the Venn overlays (as they are now) and in the supplements. The hTERT-RPE data should be in Figure 2 since it is followed up on in the subsequent figure (ie it fails to meet the definition of being relegated to 'supplemental' data).

As written in the response above, we suggest to generate Lamin B1 pA-DamID data in hTERT-RPE data.This will allow us to make a more consistent story and address these comments.

**suggesting that the separation of LADs and inter-LADs becomes progressively more pronounced after mitosis. Nevertheless....**

● This is overstated, especially given the previously mentioned work (Luperchio, Zhang). More accurate to say LADs ​association with the nuclear lamina becomes more pronounced​. LADs (predominantly B-compartment) and inter-LADs (predominantly A-compartment) show much earlier separation from each other. This may be distinct from association with the lamina. This is an important distinction as it may lead to different hypotheses regarding mechanisms of LAD targeting/association with the lamina.

We agree that this is an overinterpretation of our data. We have changed the phrasing to make it more accurate.

**Progression from prometaphase to late telophase in HeLa cells takes about 1 hour [33], suggesting that this timepoint captures the initial interactions with the reforming NL. Remarkably, the majority of these interactions is shared with later time points, indicating that most LADs can interact with the NL throughout interphase and are defined (and positioned at the NL) very soon after mitosis.**

● There is wide variability in this number, some cells rapidly exit, others take significantly longer. This number is an average (and, for what it's worth, based on a very compromised cancer cell line). The "interactions' mapped are likely reflecting the ensembe measurements of the many cells that have transited into G1. Also, this statement seemingly directly contradicts the premise of many of your following data/interpretations of a sort of step-wise wave or prefered interactions from telomere proximal toward centromeric regions. This also disagrees with your previous work (Kind et al) and more recent work regarding positioning to the NL very soon after mitosis. Again, this is BULK (many cells of a continuum of configurations) versus single cell observations. This is overstated.

We felt that there was a need to explain why we interpret the 1h time point as the initial interactions with the NL and included this reference, but the reviewer is correct that this number can vary greatly between cell types and conditions. We have removed the reference and now include FACS and imaging data supporting this claim directly.

We have changed the phrasing of these results to make our interpretation clearer.

**We next looked into characteristics of the dynamic LADs. At early time points, LADs with decreasing interactions do not have lower pA-DamID scores than stable LADs, suggesting that their ​detachment from the NL is not simply due to weak initial ​binding**

● The methods used here are dynamic proximity measures. Words like "binding" and "attachment" should be avoided (use interacting, associated, etc )

Good point. We have replaced all occurrences of these words.

**LAD dynamics are linked to telomere distance and LAD size in multiple cell types**

● Perhaps I am missing something, but I find relatively little data showing centromere-proximal LADs across cell cycle stages (referring here to Log2 ratio plots similar to what is shown for telomere-proximal LADs, Supplemental figure 6 is the only place where this is obvious.).

To better illustrate the inverse dynamics of telomeres and centromeres in hTERT-RPE cells, we have changed Fig. 3B to a full chromosome overview.

● In addtion, it seems to me that you are arguing in this and the preceding section for the following parameters: intensity of the LAD region. ie small, telomere-proximal, more euchromatic, AND less "intensely" associated.

● What is a "small" LAD? 100 kb or less? In Figure 2 (HCT1016, log 2 ratios), the original observation that leads into a discovery of changing NL associations through the cell cycle, the LAD that changes appears to be at least average size. Perhaps a "small" LAD adjacent to an "average" LAD. Nor do the signals appear to be all that low. There are regions within this sub-chromosomal plot that do appear to be "small" "low intensity" LADs. I am uncertain what parameters are defining these attributes. Are the cut-offs the same between cell types (ie is there a rule here?).

We do not set any cut-offs for any features that we compare with. We took the strategy to define stable and dynamics LADs (Fig. 3C) and ask whether there are differences in feature distributions, including LAD size, replication timing and other features. As you can see in Fig. 3E, LADs with decreasing NL interaction are smaller than stable or increasing LADs. This strategy is consistent between cell lines. To assist the reader in following our reasoning, we have added LAD domains and their differential status to Fig. 3B.

● The rules outlined above seem to break down across the different cell types. In particular, the number of active genes per Mb seems to have very little correlation overall with LADs that change. In addition, it is very unclear if "LAD size" is really a readout of both size AND intensity of interactions (understanding that this is not necessarily a direct quantitative measure of interactions).

This comment reflects our reasoning why we added a comparison between cell types in Fig. 4. Indeed, we find no general trend that active gene density correlates with LADs with decreasing NL interactions in every cell type. In contrast, LADs with decreasing NL interactions are consistently close to telomeres and smaller in size than stable or increasing LADs. We made it clearer that LAD size solely reflects the genomic size in basepairs.

**Correlation of pA- determined LADs that change into G1/S with B-compartment sub-types**

● There is certainly Hi-C data on most (all?) of the cell types analyzed in this manuscript. It would be very useful for the authors to parse out how the gain/loss LADs correlate with the B1, B2. A1, A2 (etc) compartment classifications. This may help to address the point above.

We have now included a comparison with Hi-C sub-compartments (Fig. 4F).

**Nucleosomal pattern of pA-DamID digestion/amplification (figure S3)**

● Onset of apoptosis needs to be ruled out. The nucleosomal (laddering) pattern could be due to DNA getting cleaved through programmed cell death pathways after permeabilization. These fragments could easily be amplified by the subsequent DamID protocol.

Amplification of apoptotic fragments, if present, is visible in DamID assays using the negative controls. Every library preparation, we include one or more negative controls in which we omit DpnI. If apoptotic fragments are present in this negative control, these can ligate to the DamID adapter and result in amplification, which we consistently do notsee. We have added a supplementary figure that shows this (Fig. S3A).

**Definition of 'bulk' assays**

● All of the assays were done in bulk. Some were synchronized, some were not. This is important since the implication is that anything not 'bulk' is single-cell. Throughout the manuscript and in the figures, please refer to the conditions as 'synchronized' versus 'unsynchronized'

The reviewer is correct that our terminology is wrong. We changed all occurrences of “bulk” to “unsynchronized”.

**Much of supplemental Figure 6 should be in a main figure**

● It is puzzling why the first (and most easily seen/interpreted) description of LAD organization relative to telomeres/centromeres after exit from mitosis is relegated to supplemental figures. It is a foundational experiment(s) for the paper.

We have changed the zoomed-in Fig. 3B with a chromosome overview that better captures this main observation. We see the remainder of Fig. S6 as technical controls and details of the experiment that are useful to include but not necessary as main figure.

**pA-Dam is possibly influenced by cell-cycle related chromatin accessibility (particularly at mitotic exit)**

● During the transition from mitosis to early G1, there are dynamic changes to chromatin state that are directly coupled to the cell cycle. A recent report, for instance, highlights that interactions of antibodies (or other proteins) with H3K9me2/3 modifications is likely influenced by phosphorylation of histone tails. The dynamics of histone modification/chromatin state possibly occluding or interfering with the interpretation of the results must be discussed.

Similar to DamID, pA-DamID utilizes a Dam-control to measure DNA accessibility and control for this. We show that a change in pA-DamID score is due to changes in NL reads, while the Dam reads do not change (Fig. S6F). In other words, we find no evidence that a change in chromatin state impacts the accessibility as measured by our Dam-control and thereby influences the results. We now repeat this observation in the discussion.

Reviewer #3 (Significance (Required)):

N/A

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 Schaik et al (Van Steensel laboratory) the authors describe a very clever approach to identifying Lamina Associated Domains (LADs) using the principles of the 'cut-n-run' strategy. Specifically, they engineer the Dam methyltransferase used in canonical DamID in frame with a protein A moiety capable of interacting with an antibody (in this case lamins--B1, B2 and A/C). After permeabilization, cells are incubated with antibodies, then pA-Dam purified protein--for a brief time window--is added to mark associated DNA with GmATC. This technique is a valuable contribution to the field, particularly since, as the authors point out,an advantage of pA-DamID is that the labeled DNA can also be visualized in situ using the m6A-Tracer, before this DNA is sequenced. This allows for validation of findings and is highly amenable to cell sorting technologies. In addition, this technology allows for a time-resolved measure of LADs not currently available by standard DamID. The authors apply this technology to four different cell types. They noted that the 'maps' generated by the is technology differed from canonical DamID at very specific regions (small LADs in very localized regions) . They then embark on a series of experiments to show that these differences arise from cell cycle -related differences that are differentially picked up by the methods--with the pA-DamID allowing for dissection of more discrete cell cycle stages/configurations. In general they find an initial preference for sub-telomeric LADs to associate with the nuclear lamina fist, then more centromeric. There is some data suggesting loss/gain of LADs in specific regions/with specific features. The manuscript is well written and the data well presented. However, there are some points that need to be addressed . Overall, there is some oversimplification or omission of previous data in the field, a lack of clarity in how some of the data was interpreted, and some areas where clarification and/or additional analyses would be helpful. I sincerely hope the authors find the following critiques to be useful. Thank you for the opportunity to review your very nice work.

Introduction:

Microscopy studies found that telomeres are enriched near the NL in early G1 phase, leading to the hypothesis that telomeres may assist in NL reassembly onto chromatin [13].

● There have been numerous studies identifying the timing and disposition of INM proteins and Lamins at m the end of mitosis (during NE reformation). Why are you citing just this one? (e.g. ​Thomas Dechat et al., 2004; T. Dechat et al., 2000; Ellenberg et al., 1997; Haraguchi et al., 2001​)

Furthermore, during S-phase B-type lamins have been found to transiently overlap with replication foci in the nuclear interior, at least in some cell types [20]

● While, technically, this has indeed been reported, this study is from 1994 and has not been repeated. The cells used in this study (3T3 fibroblasts) are widely used and others have not noted this phenomenon. Soften this.

Other studies have indicated that lamins are important for DNA replication [reviewed in 21].

● Likewise, direct roles for lamins in replication are controversial (acknowledged in the small section of the cited review on the role of lamins in replication).

● Perhaps combine the two sentences above to soften the implication that this is a "known" role of B-type lamins. e.g. "A handful of studies have implicated a role for B-type lamins in replication, but the direct role of the lamina in this process remains unclear. Nonetheless, ......"

Results:

So far, the cell cycle dynamics of genome - NL interactions have primarily been studied by microscopy. While these studies have been highly informative, they were often limited to a ​few selected loci.​

● Please cite your own study (Kind et al.) and other recent papers (Luperchio et al.-​https://www.biorxiv.org/content/10.1101/481598v1&#x200B;; Zhang et al., Nature-​https://www.nature.com/articles/s41586-019-1778-y&#x200B;) in which they were either 1) not limited to a few selected loci and/or 2) not microscopy-directed studies? There is an argument to be made here for the resolution (time and b.p.) you have achieved through your studies that these studies did not.

● How does this data correlate with TSA-seq, another antibody-based method developed by the Belmont lab, but collaboratively developed for use in identifying LADs (ie Dam alternative) with the Van Steensel group?​ I can imagine there are numerous advantages to this approach (radius of "labeling" being one).

When Dam-Lamin B1 is expressed in vivo for 5-25 hours during interphase, LADs that interact with the NL become progressively labeled, eventually resulting in a layer of labeled chromatin of up to ~1 μm thick [8]. This is because LADs are in dynamic contact with the NL. We expected that in pA-DamID this layer would be thinner, because the NL-tethered Dam is only activated for 30 minutes. In addition, permeabilization depletes small molecules including ATP and thus prevents active DNA remodeling in the nucleus [26]. Indeed, pA-DamID yields a m6A layer that is ~2.5 fold thinner than the layer in cells that express Dam-Lamin B1 in vivo (Fig. S2A-C). This is not an artifact due to collapse of chromatin onto the NL caused by the permeabilization, because permeabilization of cells expressing Dam-Lamin B1 in vivo did not significantly reduce the thickness of the m6A layer compared to directly fixed cells (Fig. S2C). The thin layer of labeled DNA obtained by pA-DamID points to an improved temporal resolution of pA-DamID compared to conventional DamID.

● I think this requires a bit more care. Your previous work clearly demonstrates LADs are dynamic. Others in the field have shown that these domains are also constrained within the larger sub-chromosomal compartment (self-interaction) of LADs (e.g. Luperchio 2018) within a chromosome. So, this is truly a temporal "snapshot" that may miss some regions of LADs that are less directly (or more dynamically) associated with the lamina, but still compartmentalized into the larger LAD sub-chromosomal compartment. It is unclear if the treatment used for this study perturbs these LAD-lamina​ dynamic​ interactions--one can imagine that the LADs are much less mobile generally under the protocol described in your supplemental information. In other words, ​LADs don't collapse, nor do they behave in the same way they would after permeabilization​. The technique has compromised some of that --which is actually fine for most of the purposes in this manuscript, but this needs to be discussed.

● In addtion, imaging data showing dam-LaminB1/2 plus m6A-tracer is missing (figure S2). This should be included. Is the intensity of the "tracer" similar between conditions? If so, were the exposures kept constant in all images? This is important since the decay rate is highly related to intensity of signal.

In some cell types, especially in​ HCT116 and ​hTERT-RPE​ cells, we noted local discrepancies between the two methods (Fig. 2A,bottom panel). These differences involve mostly regions with low signals in DamID that have higher signals in pA-DamID. However, such differences are not obvious in HAP-1 and K562 cells.

● Only HCT116 data is shown in the indicated figure. hTERT-RPE cells are shown in the accompanying supplemental figure and use a different antibody (lamin B2) as the target for the pA-Dam.

● This brings up another point: the data (log2 ratio schema) shown in figure 2 is for HCT116 lamin B1 pA-Dam. Yet, the subsequent studies for transient/building interactions during G1 and into S (Figure 3) are done in hTERT-RPE cells using lamin B2. To be consistent, data from lamin B2 should be used in both figures (it seems lamin B2 data is available for all cell types). The comparison of Dam-Lamin B1 can be addressed in the Venn overlays (as they are now) and in the supplements. The hTERT-RPE data should be in Figure 2 since it is followed up on in the subsequent figure (ie it fails to meet the definition of being relegated to 'supplemental' data).

suggesting that the separation of LADs and inter-LADs becomes progressively more pronounced after mitosis. Nevertheless....

● This is overstated, especially given the previously mentioned work (Luperchio, Zhang). More accurate to say LADs ​association with the nuclear lamina becomes more pronounced​. LADs (predominantly B-compartment) and inter-LADs (predominantly A-compartment) show much earlier separation from each other. This may be distinct from association with the lamina. This is an important distinction as it may lead to different hypotheses regarding mechanisms of LAD targeting/association with the lamina.

Progression from prometaphase to late telophase in HeLa cells takes about 1 hour [33], suggesting that this timepoint captures the initial interactions with the reforming NL. Remarkably, the majority of these interactions is shared with later time points, indicating that most LADs can interact with the NL throughout interphase and are defined (and positioned at the NL) very soon after mitosis.

● There is wide variability in this number, some cells rapidly exit, others take significantly longer. This number is an average (and, for what it's worth, based on a very compromised cancer cell line). The "interactions' mapped are likely reflecting the ensembe measurements of the many cells that have transited into G1. Also, this statement seemingly directly contradicts the premise of many of your following data/interpretations of a sort of step-wise wave or prefered interactions from telomere proximal toward centromeric regions. This also disagrees with your previous work (Kind et al) and more recent work regarding positioning to the NL very soon after mitosis. Again, this is BULK (many cells of a continuum of configurations) versus single cell observations. This is overstated.

We next looked into characteristics of the dynamic LADs. At early time points, LADs with decreasing interactions do not have lower pA-DamID scores than stable LADs, suggesting that their ​detachment from the NL is not simply due to weak initial ​binding

● The methods used here are dynamic proximity measures. Words like "binding" and "attachment" should be avoided (use interacting, associated, etc )

LAD dynamics are linked to telomere distance and LAD size in multiple cell types

● Perhaps I am missing something, but I find relatively little data showing centromere-proximal LADs across cell cycle stages (referring here to Log2 ratio plots similar to what is shown for telomere-proximal LADs, Supplemental figure 6 is the only place where this is obvious.).

● In addtion, it seems to me that you are arguing in this and the preceding section for the following parameters: intensity of the LAD region. ie small, telomere-proximal, more euchromatic, AND less "intensely" associated.

● What is a "small" LAD? 100 kb or less? In Figure 2 (HCT1016, log 2 ratios), the original observation that leads into a discovery of changing NL associations through the cell cycle, the LAD that changes appears to be at least average size. Perhaps a "small" LAD adjacent to an "average" LAD. Nor do the signals appear to be all that low. There are regions within this sub-chromosomal plot that do appear to be "small" "low intensity" LADs. I am uncertain what parameters are defining these attributes. Are the cut-offs the same between cell types (ie is there a rule here?).

● The rules outlined above seem to break down across the different cell types. In particular, the number of active genes per Mb seems to have very little correlation overall with LADs that change. In addition, it is very unclear if "LAD size" is really a readout of both size AND intensity of interactions (understanding that this is not necessarily a direct quantitative measure of interactions).

Correlation of pA- determined LADs that change into G1/S with B-compartment sub-types

● There is certainly Hi-C data on most (all?) of the cell types analyzed in this manuscript. It would be very useful for the authors to parse out how the gain/loss LADs correlate with the B1, B2. A1, A2 (etc) compartment classifications. This may help to address the point above.

Nucleosomal pattern of pA-DamID digestion/amplification (figure S3)

● Onset of apoptosis needs to be ruled out. The nucleosomal (laddering) pattern could be due to DNA getting cleaved through programmed cell death pathways after permeabilization. These fragments could easily be amplified by the subsequent DamID protocol.

Definition of 'bulk' assays

● All of the assays were done in bulk. Some were synchronized, some were not. This is important since the implication is that anything not 'bulk' is single-cell. Throughout the manuscript and in the figures, please refer to the conditions as 'synchronized' versus 'unsynchronized'

Much of supplemental Figure 6 should be in a main figure

● It is puzzling why the first (and most easily seen/interpreted) description of LAD organization relative to telomeres/centromeres after exit from mitosis is relegated to supplemental figures. It is a foundational experiment(s) for the paper.

pA-Dam is possibly influenced by cell-cycle related chromatin accessibility (particularly at mitotic exit)

● During the transition from mitosis to early G1, there are dynamic changes to chromatin state that are directly coupled to the cell cycle. A recent report, for instance, highlights that interactions of antibodies (or other proteins) with H3K9me2/3 modifications is likely influenced by phosphorylation of histone tails. The dynamics of histone modification/chromatin state possibly occluding or interfering with the interpretation of the results must be discussed.

Significance

N/A

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 describes a new method for detecting Lamin associated DNA domains, which allows better time resolution than classical DamId. It is a good idea and its functionality is demonstrated in tissue culture cells. There are minor insights but it is important that we advance the field with new and better technologies, thus this version amply suffices to give evidence of that.

Significance

The audience is all persons working on chromatin organization in the nucleus, which is a large audience. The data are clear as they basically are proof of principle for a new technique. There is nothing major to request as revision. They might cite papers on damID in worms and tissue specific applications of this in living organisms, as this is likely to be the situation that is most interesting in the long run. The resolution (in bp) would be interesting to know and validate.

I have no other major revisions to request.

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 report, van Schaik et al., modified an established CUT and RUN method and combined it with previously used DamID to identify Lamin Associated Domains (LADs) with better temporal resolution. Previous DamID experiments labeled locations where lamin proteins were present within a 5-25 hour window while the new technique, pA-DamID, labels DNA within a 30 minute window providing better temporal resolution. The authors used this technique to identify LADs at multiple stages of the cell cycle and applied this protocol to different cell types. The authors FIND differences when comparing data sets between cell cycle time points and cell lines.

Major points:

1) The data sets generated and displayed in this manuscript seem incomplete. In Figure 1G, the authors compare lamin B2 vs. lamin B1 generated LADs in HAP-1 cells and lamin A/C vs lamin B2 LADs in hTERT-RPE cells. In figure S4, panel C compares lamin B1 and lamin B2 in K562 cells and lamin B2 and lamin A/C in hTERT-RPE cells. It would have been informative to have a complete dataset for lamin B1, lamin B2, and lamin A/C identified LADs in all cell lines analyzed. The information provided from these datasets would be useful to the scientific community.

2) The authors discovered that LADs reposition during progression through the cell cycle. It would have been interesting to know whether these changes have transcriptional consequences? One could perform RNA-SEQ experiments to discover if LAD occupancy results in transcriptional changes and choose a few genes to confirm the findings with RT-PCR. Is this the same for lamin B1, lamin B2, and lamin A/C occupied LADs? Analyze if there are any genomic features such as CTCF or transcription factor binding sites that correlate with the loss of LADs.

3) The authors state that using H3K27me3/H3K9me3 in pa-DamID showed no enrichment. This is surprising considering that both modifications are enriched in heterochromatin and at the nuclear periphery. It appears that the peripheral enrichment is masked by the larger overall internal pool. The authors should discuss this observation and comment on the sensitivity of the method to detect local enrichment versus the global levels of a protein or modification in pa-DamID.

Minor points:

Figure 1: Change colors for Figure 1F and Figure 2D. The colors are hard to discern.

Figure 2B: Please mark which antibody was used for this analysis.

Figure 2C: Please also overlay data from pA-DamID lamin A/C experiments.

Figure 4: Please mention which antibody was used for the pA-DamID experiments used to generate this dataset.

Figure 5: Please mention which antibody was used for the pA-DamID experiments used to generate this dataset.

Figure S5 C and D: Please mention which antibody was used for the pA-DamID experiments.

Significance

The major contribution of this manuscript is the description of an improved method to map LADs. This is a valuable contribution. By using this new method, the findings of this paper provide some new insight in LAD dynamics throughout the cell cycle although the experiments are largely phenomenological. This is a technically sound study.

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

Learn more at Review Commons

---

Reply to the reviewers

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

Reply to the Reviewers:

We would like to thank the reviewers for taking the time to provide us with insightful and constructive comments, which helped us in improving the manuscript. We have performed additional experiments and data analysis while also improving the presentation of our analysis. In addition, we have noted one experiment (bulk RNA-sequencing of pax2a-Low and pax2a-High thyrocyte population) for further revision to address the concerns raised by Reviewer #1.

Please find below our point-by-point response. Please note that the figure references in the response refer to the ‘Revised Manuscript’, ‘Revised Figures’ and ‘Revised Supplementary Data’ file.

Reviewer #1:

Gillotay et al. performed an unbiased profiling of the zebra fish thyroid gland and captured different cell populations by single-cell gene expression analysis. Using bioinformatic tools, they identified seven clusters corresponding to expected, but also poorly characterized, sub-populations, such as non-follicular epithelial cells. They also found two transcriptionally distinct types of thyrocytes and validated this heterogeneity using a new transgenic Pax2a reporter line. Using this tool, they identified and located Pax2a-low and -high thyrocytes within thyroid follicles. Finally, they highlighted a dense intercellular signaling network based on ligands expressed by the diverse sub-populations present in the thyroid and receptors expressed by the thyrocytes. This is a descriptive work calling for more in-depth analyses.

The authors thank the reviewer #1 for acknowledging the strengths of the manuscript and for stating that it could be of interest to a larger audience. We appreciate the reviewer’s advice on supportive experiments and for improving the clarity of the analysis and presentation. We have tried to experimentally address the concerns of the reviewer and hope this provides in-depth substantiation of our observations.

**Major comments on main conclusions :**

Conclusion 1 : Identification of 7 clusters

• This reviewer was particularly surprised by the relative abundance of the different sub-populations identified. For example, 267 thyrocytes out of 6249 cells from the thyroid gland is less than 5% of the total thyroid cell number. In comparison, authors identified three times more immune cells or non-follicular epithelial cells. Authors should comment on these numbers, on the dissection (contaminants?) and dissociation procedure.

Do these relative abundance reflect the proportion of thyrocytes, immune, stromal cells they normally observe in adult thyroid sections ? It would be interesting to have a better resolution for figure 6A in order to evaluate the number of nuclei stained only with DAPI as compared to nuclei stained with DAPI and surrounded by E-cadherin staining. Based on the image, this reviewer seriously doubts that follicular cells represent less than 5% of the total cell number in this organ, considering that the colloid is a cell-free zone.

We fully agree with the reviewer statement. We have clarified the nature of the dissociated tissue in Results, Methods and Fig. 1 C- G. In this, we have performed 3D confocal imaging of the region utilized for single-cell RNA-Seq. (Fig. 1C). Further, we quantified percentage of thyrocytes in transverse sections across the dissociated region (Fig. 1 D – E). Our results demonstrate presence of 5.9 ± 1.9 % thyrocytes in the region. Lastly, we provide the FACS plots of the cells utilized for single-cell RNA-Seq. In this, we obtained around 4% thyrocytes among the live cells. Taken together, our quantifications of the thyrocyte proportion in the tissue matches well with the percentage of thyrocytes obtained in the single-cell atlas, suggesting lack of thyrocyte loss during the procedure.

We agree with the reviewer’s observation that the follicle lumen represents a cell-free zone. However, it is worth noting that the cells surrounding the follicles, in particular gills and stroma, have a higher cell density. Additionally, the zebrafish thyroid follicles sit loosely on the ventral aorta, thus making it difficult to manually separate the follicles from the surrounding tissue without destroying the organ. We avoided injuring the organ to minimize cell-death associated with manual dissection.

• When using the webtool developed on the thyrocyte population, one can notice that only a small fraction of the thyrocyte population expresses common thyroid-specific genes such as Tpo or duox. Was this expected ? It would be interesting to comment on this observation and confirm using standard localization technique to demonstrate that this is real and not due to the sequencing. Is it specific to the zebrafish ? On the other hand, Tg is expressed in most thyrocytes, but surprisingly also in all the clusters at a fairly good level. This should be commented... Is it normal ? due to the sequencing quality ? or clustering ?

We believe these are technical issues related to single-cell sequencing and thank the reviewer for their insight in seeing this.

The non-uniform expression of thyroid-specific marker genes (Tpo and duox) likely represents dropout effects, possibly due to the low expression levels of these genes. To address this issue, we propose to perform bulk RNA-Seq. of thyrocytes (segregated by pax2a expression levels). Bulk RNA-Seq. is more sensitive than single-cell RNA-Seq. and should provide the expression levels of these genes in the thyrocytes.

The expression of Tg in non-thyrocyte population likely represents cross-contamination of free RNAs released from ruptured cells. Since Tg mRNA is highly expressed in the thyroid follicular cells, release of the mRNA from a few injured cells would contaminate the cell suspension, leading to its detection in non-thyroid cells. However, the expression represents background noise signal. To test this, we utilized DecontX, a recently developed approach for background correction (S. Yang et al. 2020). In this, the expression of a gene is modelled as a mixture of expression in the expected population plus background expression. With this, we could robustly reduce Tg mRNA expression in non-thyrocytes (Supp. Fig. 3). This supports our hypothesis that the Tg expression in non-thyrocytes likely represents cross-contamination of mRNA from ruptured cells.

• In order to validate and locate the different populations identified in the thyroid, this reviewer suggests to perform in situ hybridization or immunostaining, based on the specific marker genes identified in each cluster. This experiment could lead to the precise identification of the different sub-populations and their respective localization. These experiments would also help in the interpretation of the cellular interaction network.

We have characterized the different cell types surrounding the thyroid follicles using various reporter lines. The data is presented in Fig. 4.

Conclusion 2 : two distinct types of thyrocytes

• This is an interesting observation. However, from a non-expert it is difficult to understand why the authors propose two populations. Based on the points distribution (Figure 4A), this reviewer would rather identify 3 or 4 clusters but not the two shown in red and blue.... Did the authors impose two populations for the clustering ? Did they perform a permutation test to confirm the pax2a significant fold change seen between clusters is not a false positive generated by the clustering ? Could they show, as a supplementary file, the same graph with points colored based on Pax2a expression ?

We concur with the reviewer that the number of potential clusters in thyrocyte population might be more than two. In-fact, there is no upper limit to the diversity present in the thyrocyte population. However, our message in the manuscript is that the population is not homogenous, and there are at-least two populations based on pax2a expression level. In this regard, we do believe that we have only scratched the tip of the iceberg and further investigation is needed to completely answer this issue. Nonetheless, we are the first to demonstrate genetic heterogeneity among the population.

The clustering of thyrocytes followed the guidelines suggested by Seurat package. For thyrocytes, we utilized the principal components displaying significant deviation from uniform distribution. For cluster identification, we utilized a resolution of 0.3, which was same as the one utilized for clustering the entire organ (further details on this provided as a response to Minor Concern #3 by Reviewer #1). A plot of pax2a expression, along with tg expression, is provided as Supp. Fig. 8.

Finally, to strengthen our observation, we conducted independent analysis of transcriptional diversity in the population using a recently developed method called ROGUE (Ratio of Global Unshifted Entropy) (Liu et al. 2019). The method provides genes that display transcriptional heterogeneity within the cell population. Assessment is based on expression entropy, a measure of the degree of uncertainty, or promiscuity, in the expression of a gene (Teschendorff and Enver 2017). For this, we utilized raw counts of thyrocytes so as to provide an alternative analysis of our data. The analysis, presented as Fig. 6D, demonstrates significant entropy, or transcriptional heterogeneity, for pax2a and cathepsin B (ctsba). The full list of genes displaying transcriptional heterogeneity in thyrocytes is provided as Supp. Table 4.

• This reviewer was also surprised by the relatively "heavy" approach (generation of the Pax2a reporter line) used to demonstrate the existence of two types of thyrocytes. Knowing that the reporter line was validated with a very good Pax2a antibody... The use of the reporter line is a bit short. The authors could for example validate the two populations of Figure 4A using the Pax2a FACS-sorted cells and RT-qPCR.

We completely agree with the suggestion of the author and plan to perform bulk RNA-sequencing using the pax2a reporter line to corroborate our results. This is the advantage provided by the generation of knock-in line. In addition, we have performed antibody staining against endogenous pax2a protein (Fig. 8 E – F), which validates the transcriptional heterogeneity observed in our single-cell RNA-Seq. data.

• In addition, data available in the sequencing dataset could be used to prove that the two populations are really active thyrocytes. This reviewer would suggest to present a table with the expression level of common and thyroid-specific genes such as TshR, Nis, Tpo, Duox, Tg, Pax2, TTF1 and other known transcription factors in the two populations to demonstrate that these two types of cells are indeed thyrocytes. Finally, image quality (Figure 6) could be improved and high-magnification images with several thyrocyte marker could be shown to convince the readers.

We strongly agree with the reviewer that this is a very important concern to address. To address this, we have taken three steps:

We have included the expression level of tg in the two populations in Fig. 6C and Supp. Fig. 8. We performed antibody staining against pax2a on thin section obtained from Tg(tg:nls-EGFP) animals (Fig. 8 E-F). In this, we observed pax2a-Low cells with tg reporter expression, suggesting that they are indeed differentiated thyrocytes. We plan to perform bulk RNA-sequencing of cells from pax2a-Low and pax2a-High population. This will allow us to validate the transcriptional differences observed by single-cell RNA-Seq., and allow us to demonstrate expression of thyroid-specific markers genes that are missing from our dataset (for instance, duox).

Conclusion 3 : cellular interaction network

• Most of the interactions revealed by the analysis seem to belong to the extracellular matrix and not to classical ligands such a Wnt, TGFb, FGF, PDGF,..... could the authors comment on this ? Considering that both endothelial cells and epithelial cells assemble their own basement membrane, the analysis will obviously reveal interactions between endothelial cells and epithelial cells....

We appreciate that the reviewer pointed this out. The enrichment of ligands related to extracellular matrix, and not growth factors, likely represents the homeostatic nature of the organ. Growth at 2 and 8 mpf is low (if not absent). Correspondingly, gene expression related to development and cell-cycle might be reduced. As stated in response to the next concern, extending the atlas to juvenile stage (1 mpf) would be beneficial to understand the regulators of cell-cycle.

However, to improve the cellular interaction network, we have incorporated physical information from the characterization of cell-populations surrounding the thyroid follicles (Fig. 4). Our experiments suggested that stromal, gills and NFE do not physically contact the thyrocytes. Thus, interactions based on ligands incorporated into the cell membrane were removed for these cell-populations.

**Minor comments to improve the Ms :**

• Could you explain how from 2 x 12 000 FACS-sorted live-cells (from six animals at each stage; 2 mpf and 8 mpf) you obtain 6249 cells (pooled of 2 mpf and 8 mpf), and why the two stages were first sorted separately and then pooled (?), as no differential analysis is carried out for the two stages.

The number of cells obtained for analysis represents cells that were successfully incorporated into droplets during library preparation and generated high-quality data that passed quality control (Supp. Fig. 2). FACS sorted cells were utilized for droplet generation using 10X Chromium that encapsulates cells with single-Poisson distribution (Zheng et al. 2017). This leads to approximately 50% cell capture rate, which is the ratio of the number of cells detected by sequencing and the number of cells loaded. Thus, we obtained 13,106 sequenced cells from 24,000 input cells (54.6 % cell capture rate). Further, the quality control criteria removed 6,857 low-quality cells (52.3 % dropout rate). We chose a stringent cut-off for quality control so as to remove technical artefacts from the analysis. We have added these detail to the Result section.

We pooled the two samples as the stages represent the range of homeostasis in zebrafish. We decided not to include differential expression between the two stages as the number of cells in multiple clusters were too low for individual stages (less than 100), and thus not trustworthy. It future, it would be of interest to extend the analysis for rapidly-growing juvenile (less than 1 mpf) and old-age (greater than 1.5 ypf animals) animals and to perform single-cell or bulk RNA-Seq. with high cell numbers. We have mentioned this drawback in the discussion section.

• Which method was used for the graph-based clustering ? KNN ? Louvain ? Random walk ?

The details have been added to the Method section. Specifically, the clustering was performed using graph-based method, shared nearest neighbour (SNN), which is default for Seurat 2.3 package.

• How did you define the numbers of clusters ?

The number of clusters were defined by using the first five principal components as they displayed significant deviation from uniform distribution as accessed by JackStraw analysis. Further, a resolution of 0.3 was used in Seurat as the clusters generated by this parameter could be annotated using a cell-type specific marker from literature.

• Figure 4B, the color-code for the expression level would help the reader.

The color code has been added (Fig. 6B in revision).

• Figure 4C, violin plot for Pax2a: why do we find cells that do not express this gene in the two populations ? The same is true for tbx2a and ahnak ... is the clustering optimal ?

The detected expression of pax2a depends on its biological expression and technical dropout rate. Thus, the pax2a-High cluster also contain cells with no detectable expression of pax2a. Similar detection dynamics can be expected for other genes.

We have experimentally validated the variability in pax2a expression using antibody staining for endogenous pax2a protein in tg:nls-EGFP transgenic line (Fig. 8 E-F). With this, we can validate the presence of pax2a heterogeneity within thyrocytes.

• Figure 4C, blue violin plot for ptp4a3 does not seem to fit with the distribution of the points.

Due to the high number of cells that do not express ptp4a3, the cells collapse on each other at the bottom of the graph, thus making the violin plot seem different from the distribution. However, the plots were made with Seurat without changing any parameters.

• what is the function of tbx2a, ahnak, ptp4a3 and dusp5, which are not mentioned in the text.

The genes have been implicated in regulation of cell proliferation. However, we acknowledge that the evidence based on a handful of genes needs to be strengthened. For this, we have removed the figure panels from the revision, and will instead identify genetic markers based on bulk RNA-sequencing analysis of pax2a-High and pax2a-Low population.

• Line 195: "pax2amKO2 reporter expression perfectly overlapped with PAX2A antibody staining". This reviewer would be more cautious as the images (Fig. 5 C, D and F) do not show a perfect colocalization: one can observe only blue or only red staining.

We have edited the text to “The pax2apax2a-T2A-mKO2 (abbreviated as pax2amKO2) reporter expression overlapped with PAX2A antibody staining in a majority of regions at 9.5 hours post-fertilization.” The regions with single colors in Fig. 5C (Fig. 7C in revision) are due to differences in staining intensity between different regions.

• Line 246: it is proposed to "study the functional and replicative differences among the two sub-populations of thyrocytes". This reviewer completely agrees and the suggestion made (vide supra) to use the datasets to assemble a table with the expression level of common and thyroid-specific genes such as TshR, Nis, Tpo, Duox, Tg, Pax2, TTF1 and other known transcription factors in the two populations could already give some indications on the functionality of these two types of cells. Expression of genes involved in cell-cycle control and/or apoptosis would be another possibility to better characterize the two populations. Lastly, the authors could perform the comparative analysis of ligand-receptor pairs between these two sub-populations to examine if they differentially interact with their environment.

We agree with the reviewer for the need to characterize the two populations in detail. Using the current dataset, we are limited to the 265 genes differentially expressed between the two thyrocyte states. Thus, we propose to perform bulk RNA-sequencing of the two populations to obtain a better picture of the cellular identity. In this, we will perform sequencing of each population in triplicates. With this, we will avoid the dropout effects suffered by the single-cell analysis. The experiment would demonstrate the differentiation status of each cell population, and provide insights into other pathways active within each population. Further, we will identify ligands and receptors that are enriched in a particular population.

Text improvement:

Intro: thyroglobulin (TG) appears twice (line 46 and 49)

Results: Fig. 5 (not 8) (line 203 and 205)

Figure 3: stromal? (not skeletal)

Figure 4: fold change scale is missing

Figure 5: Thyroid gland (Gland)

Figure sup 2: Fluorescence-activated cell sorting (FACS) of zebrafish thyroid gland

Figure sup 3: number of unique molecular identifier

Figure sup 4: "are present in the zebrafish"

We thank the reviewer for pointing these errors. We have edited them.

Reviewer #1 (Significance (Required)):

The work performed by Gillotay et al. is clearly novel but descriptive. It provides a useful database to propose hypotheses to further study the thyroid gland. The single-cell RNA-seq analysis brings a molecular appreciation of the various thyroid cell populations, thyrocyte heterogeneity and intercellular signaling network. Although focused on the thyroid, results will be of interest to a larger audience than the thyroid community, especially the demonstration (if further and better validated) of thyrocyte heterogeneity and the intercellular communication possibilities.

In response to comments by Reviewer 1, we plan to perform bulk RNA-Sequencing of pax2a-Low and pax2a-High thyrocytes. We believe that this will help address all the remaining concerns of the reviewer.

Reviewer #2:__ __

**Summary**

In this manuscript the authors present a single-cell transcriptome atlas of the zebrafish thyroid gland (possibly also including some adjacent tissues depending on how the dissection was performed, see comments below). The atlas includes cell clusters that are expected to be found in the thyroid of any higher vertebrate species (thyrocytes, blood vessels, lymphatic vessels, immune cells and fibroblasts), but also musculature/gills and a less well-defined population of non-follicular epithelium. The data will be made publicly available as a resource, by what seems to be a user-friendly web-interface (more accessible to a broader audience than customary raw sequencing data deposition, that I suppose will also be provided). The results are used to describe putative autocrine or paracrine signaling networks. The authors are able to further subdivide the thyrocyte cell cluster into two sub-populations with different transcriptomic features. Interestingly, these populations differ in their expression of for instance the key transcription factor pax2a, which is further demonstrated by the use of a novel zebrafish reporter strain.

In general, this is a clearly interesting, novel, nicely structured and well written manuscript and the data presented seems to be of high quality.

We would like to thank reviewer 2 for the constructive comments. We are glad that the reviewer finds our work interesting, novel and of high quality. We appreciate the reviewer’s advice on additional experiments, analysis and on improving the clarity of the text. We have addressed all the concerns raised, and hope that our revised manuscript satisfies the reviewer.

**Major comments** Key conclusions are convincing and performed with scientific rigor. As will be further discussed below the seemingly superficial description of the extent of tissue that was subjected to transcriptome analysis makes it a bit difficult to assess reproducibility outside the authors' lab.

We acknowledge the lack of clarity in the description of the tissue utilized for single-cell analysis. We have corrected this providing a detailed step-by-step protocol for dissecting the organ in Methods Section, titled ‘Dissection of the zebrafish thyroid gland’. Additionally, using immunofluorescence based imaging of the region and FACS, we estimate the proportion of thyroid follicular cells within the region. The results are presented in Fig. 1 C – G.

As far as I can see very little methodological detail or information is provided about how the dissection of the thyroid region was performed, more than that "the thyroid gland was collected" or that "we dissected out the entire thyroid gland". This is essential to understand the significance of the cell populations that are described based on the transcriptomic data. The section "Tissue collection" describes dissection of the thyroid for whole-mount imaging. From Fig. 6A it seems that substantial amounts of non-thyroid tissue are included in this dissection. Is it the same kind of dissection that was performed for transcriptomics? Was the string of thyroid follicles shelled out from their surroundings or was some kind of en bloc dissection, including other neighboring tissues, performed (as suggested from the transcriptome cell populations data including e.g. gill transcripts)? In the latter case it would be good if the authors discuss in more detail what other tissues or structures that are expected to also be included in the dissected tissue and transcriptomic data.

In response to this concern of the reviewer, we have improved the clarity of the text in Results and Methods section. We have added the following text to the Result section, “The zebrafish thyroid gland is composed of follicles scattered in the soft tissue surrounding the ventral aorta (Fig. 1 A, B). Ventral aorta extends from the outflow tract of the zebrafish heart and carries blood from the ventricle to the gills. Dissection of the ventral aorta associated region (detailed in Methods section) provided us with tissue that included the thyroid follicles and parts of zebrafish gills (Fig. 1C). Using Tg(tg:nls-EGFP) transgenic line, which labels thyrocytes with nuclear green fluorescence (Fig. 1D), we estimated presence of 5.9 ± 1.9 % thyrocytes within the dissociated region (Fig. 1E).”

Further, the Methods section now defines a step-by-step protocol for dissociation (‘Dissection of the zebrafish thyroid gland’).

In addition, we have improved the characterization of the dissected region, as stated in response to the next comment.

It would facilitate understanding if the thyroid is outlined in Fig. 1A as well as the region that was dissected for downstream single cell sequencing.

A whole-mount 3D reconstruction of the region is presented in Fig. 1C. A transverse section from the region is presented as Fig. 1D, while quantification of the percentage of thyrocytes in the transverse sections is provided in Fig. 1E.

The clusters seem logical given what cell types that could be expected in the region (but depending on how dissection was performed). The only exception is cluster 7 (non-follicular epithelium; NFE). I do understand that relative sizes of the clusters do not necessarily reflect the endogenous relative abundance of different cell types, as I guess they may be more or less prone to enzymatic dissociation, survival etc. Nevertheless, the number of cells in the NFE cluster (831 cells) seems sizeable relative to the number of thyrocytes (267 cells). In my opinion, a major weakness of the current manuscript is that little detail is provided about this cell population and that no attempt to at least spatially localize these cells is presented.

Detailed characterization of the cell-populations surrounding the thyroid follicles is now presented in Fig. 4. In addition, we have quantified the percentage of thyrocytes in the region (Fig. 1 E), which demonstrates that thyrocytes represent 5.9 ± 1.9 % of the cell-population. Additionally, we have presented FACS analysis of the dissociated region as Fig. 1 F - G, which corroborates the imaging based quantification. Both quantifications are in the same range as the proportion of thyrocytes identified in the single-cell analysis (4.27 % - 267 out of 6249 cells). Thus, we do not believe that cell-loss had a big impact on the relative abundance of thyrocytes in the single-cell atlas.

A detailed characterization of NFE cells is provided in response to the next three comments, which includes visualization of the population using TP63 / p63 antibody staining in Fig. 4D. Particularly, Fig. 4D contains 72 thyrocytes and 302 TP63+ nuclei, thereby pointing to the higher relative abundance of NFE in the region.

The NFE cells are characterized by TP63 expression and the authors speculate that they might show homology to main cells of solid cell nests. From previous zebrafish literature it seems like what is supposedly ultimobranchial bodies (or ultimobranchial glands more similar to those in avian species) are located rather distant from the thyroid follicles (Alt et al 2006). Is it possible that these structures are included in the dissection that has been performed? As solid cell nests are supposed to be ultimobranchial body remnants (with calcitonin positive and calcitonin negative epithelial cells) it would be good if the authors discuss in more detail what is known about the ultimobranchial bodies in zebrafish, if they are located inside the zebrafish thyroid, in an anatomical region that is included in the dissected tissue of this study or in a region that is likely not included.

As stated by the reviewer, the ultimobranchial bodies lie distant to the thyroid gland. They lie as a pair of follicles on top of the sinus venous (Alt et al. 2006), which is a blood vessel that delivers blood to the atrium. In contrast the thyroid follicles sit loosely on top of ventral aorta that connects to the ventricle via the outflow tract (Fig. 1B). Thus, the collection of the ultimobranchial bodies is not expected. This is also corroborated by the absence of calcitonin (calca) expression in the NFE (Supp. Table 1). This has been added to Discussion section.

In higher vertebrates, P63 expression is typically seen in basal cells of stratified epithelia (as for instance in the esophagus), in myoepithelial cells, in the urothelium and in the thymus. Is it possible that the TP63 expressing NFE population corresponds to cells of the zebrafish thymus (that might perhaps explain the seemingly large immune cell population in cluster 4)? Could TP63 expressing NFE cells represent the esophagus (if included in the dissection)? As so little detail is provided about the dissection procedure this opens up for speculation and it would be good if the authors discuss these possibilities, as some of them might perhaps be unlikely or even impossible given regional anatomy of the zebrafish and how the dissection was performed.

The dissection region is now characterised in detail in Fig. 1 C – E. The presence of immune cells (macrophages) in the proximity of thyroid follicles is validated in Fig. 4B. The presence of NFE is presented as Fig. 4D, and explained in detail in response to the next comment.

To gain better understanding of the sizeable TP63 expressing NFE population the authors briefly mention the possibility of future studies utilizing a TP63 reporter. If a reporter line is not available, immunofluorescent detection of P63 (as presented for PAX2A in Fig. 5 and E-cadherin in Fig. 6) would be desirable to provide more insight into the location of the NFE population. Given the proficiency the authors demonstrate in this manuscript when it comes to zebrafish imaging, at least whole-mount immunostaining of P63 in the region that was dissected for transcriptome analysis seems clearly feasible, both with respect to resources and time needed (perhaps in the range of 1-3 months).

To clarify the presence of NFE cells, we have followed the suggestion of the reviewer and performed immunostaining against TP63. The result is presented as Fig. 4D. The staining was performed on thin (8 µm) sections, allowing us to ensure uniform antibody penetration. As depicted in the image, TP63+ cells are part of the gills. This population possibly represents a progenitor population, similar to the TP63+ basal layer in the zebrafish (Guzman et al. 2013) and mammalian (A. Yang et al. 1999) epithelium. Additionally, a sub-set of TP63+ cells were observed in the region between the thyroid follicles and gills. Our data provides an exciting opportunity for an in-depth study of these cells in future, particularly using tp63-regulatory region driven transgenic reporter and Cre lines.

**Minor comments** It is a little bit confusing that different color coding for the various cell populations is used in Fig. 3B as compared to Figs. 1 and 2.

The color coding for Fig. 3B (Fig. 5B in revised manuscript) has been modified in accordance to the once used in Fig. 1 and 2.

In the discussion of the intercellular interaction network (Fig. 3B) the authors clearly point out that anatomical barriers are not modelled. Nevertheless, it would be more informative if this description was able to sort out ligands that are secreted, from ligands that are not secreted and would require physical interaction between thyrocytes and a different cell population for signaling to occur.

We have now improved the intercellular network to resemble the putative physical interactions. By characterizing the different cell-populations present in the atlas (Fig. 4), we recognized that gills, stromal and NFE are not in physical proximity of the thyrocytes. Thus, these three cell-populations would not be able to communicate via ligands attached to the cell membrane. Hence, we have pruned the network to remove cell-membrane attached ligands for these three cell-populations. Only secreted ligands were considered for the mentioned cell-populations. In accordance, the figure and Supplementary Table 2 has been updated.

The authors describe thyroid solid cell nests as "... lumen containing irregular structures located between the thyroid lobes in mammals...". Solid cell nests of the thyroid in higher vertebrates (e.g. humans and dogs) are located within the thyroid lobes and not between the lobes (i.e. the right and left thyroid lobes). Moreover, the authors write that "Recently, epithelial cells have been reported in a structure called the Solid Cell Nests (SCN) of the thyroid..." and give reference to a paper from 1988. If that is recent or not might be a matter of opinion, but to the best of my knowledge, solid cell nests were describe already by Getzowa in 1907 and I suppose that the epithelial identity (or at least epithelioid morphology) has been appreciated for long.

We thank the reviewer for pointing this out. We have added the reference to the original study by Dr. Sophia Getzowa identifying SCN (Getzowa 1907). As the original study is in German, we have also added a reference to a recent article attributing the discovery of SCN to Dr. Getzowa and performing immunohistochemistry analysis of the tissue (Ríos Moreno et al. 2011). Notably, the authors note the presence of TP63 staining, along with the absence of TG and Calcitonin staining, in SCN main cells – similar to the expression profile of NFE in our atlas. Finally, we have edited the text to accurately describe their location in the mammalian thyroid gland.

Reviewer #2 (Significance (Required)): The authors provide a single-cell transcriptomic atlas of the zebrafish thyroid gland. To the best of my knowledge this is certainly a unique resource. In that sense it provides novel and significant information that will surely facilitate our further understanding of thyroid biology. It will surely be of great interest and value to the thyroid community, but probably also to a wider audience interested in e.g. zebrafish biology, endodermal biology and the biology of endocrine glands. Their finding and direct demonstration of transcriptional heterogeneity within the thyrocyte population is very interesting, also in different contexts of thyroid disease. However, I leave it to other referees to comment on the conceptual uniqueness of the current manuscript (i.e. a single-cell transcriptomic atlas of a zebrafish organ). Does it provide conceptually unique information, or does it add to an expanding collection of single-cell transcriptomic atlases of zebrafish organs?

References:

Alt, Burkhard, Saskia Reibe, Natalia M. Feitosa, Osama A. Elsalini, Thomas Wendl, and Klaus B. Rohr. 2006. “Analysis of Origin and Growth of the Thyroid Gland in Zebrafish.” Developmental Dynamics 235 (7): 1872–83. https://doi.org/10.1002/dvdy.20831.

Getzowa, Sophia. 1907. “Über Die Glandula Parathyreodeaa, Intrathyreoideale Zellhaufen Derselben Und Reste Des Postbranchialen Körpers.” Virchows Archiv Für Pathologische Anatomie Und Physiologie Und Für Klinische Medizin 188 (2): 181–235. https://doi.org/10.1007/BF01945893.

Guzman, A., J. L. Ramos-Balderas, S. Carrillo-Rosas, and E. Maldonado. 2013. “A Stem Cell Proliferation Burst Forms New Layers of P63 Expressing Suprabasal Cells during Zebrafish Postembryonic Epidermal Development.” Biology Open 2 (11): 1179–86. https://doi.org/10.1242/bio.20136023.

Liu, Baolin, Chenwei Li, Ziyi Li, Xianwen Ren, and Zemin Zhang. 2019. “ROGUE: An Entropy-Based Universal Metric for Assessing the Purity of Single Cell Population.” BioRxiv, January, 819581. https://doi.org/10.1101/819581.

Ríos Moreno, María José, Hugo Galera-Ruiz, Manuel De Miguel, María Inés Carmona López, Matilde Illanes, and Hugo Galera-Davidson. 2011. “Inmunohistochemical Profile of Solid Cell Nest of Thyroid Gland.” Endocrine Pathology 22 (1): 35–39. https://doi.org/10.1007/s12022-010-9145-4.

Teschendorff, Andrew E., and Tariq Enver. 2017. “Single-Cell Entropy for Accurate Estimation of Differentiation Potency from a Cell’s Transcriptome.” Nature Communications 8 (1): 15599. https://doi.org/10.1038/ncomms15599.

Yang, A, R Schweitzer, D Sun, M Kaghad, N Walker, R T Bronson, C Tabin, et al. 1999. “P63 Is Essential for Regenerative Proliferation in Limb, Craniofacial and Epithelial Development.” Nature 398 (6729): 714–18. https://doi.org/10.1038/19539.

Yang, Shiyi, Sean E. Corbett, Yusuke Koga, Zhe Wang, W Evan Johnson, Masanao Yajima, and Joshua D. Campbell. 2020. “Decontamination of Ambient RNA in Single-Cell RNA-Seq with DecontX.” Genome Biology 21 (1): 57. https://doi.org/10.1186/s13059-020-1950-6.

Zheng, Grace X. Y., Jessica M. Terry, Phillip Belgrader, Paul Ryvkin, Zachary W. Bent, Ryan Wilson, Solongo B. Ziraldo, et al. 2017. “Massively Parallel Digital Transcriptional Profiling of Single Cells.” Nature Communications 8 (1): 14049. https://doi.org/10.1038/ncomms14049.

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 the authors present a single-cell transcriptome atlas of the zebrafish thyroid gland (possibly also including some adjacent tissues depending on how the dissection was performed, see comments below). The atlas includes cell clusters that are expected to be found in the thyroid of any higher vertebrate species (thyrocytes, blood vessels, lymphatic vessels, immune cells and fibroblasts), but also musculature/gills and a less well-defined population of non-follicular epithelium. The data will be made publicly available as a resource, by what seems to be a user-friendly web-interface (more accessible to a broader audience than customary raw sequencing data deposition, that I suppose will also be provided). The results are used to describe putative autocrine or paracrine signaling networks. The authors are able to further subdivide the thyrocyte cell cluster into two sub-populations with different transcriptomic features. Interestingly, these populations differ in their expression of for instance the key transcription factor pax2a, which is further demonstrated by the use of a novel zebrafish reporter strain. In general, this is a clearly interesting, novel, nicely structured and well written manuscript and the data presented seems to be of high quality.

Major comments

Key conclusions are convincing and performed with scientific rigor. As will be further discussed below the seemingly superficial description of the extent of tissue that was subjected to transcriptome analysis makes it a bit difficult to assess reproducibility outside the authors' lab.

As far as I can see very little methodological detail or information is provided about how the dissection of the thyroid region was performed, more than that "the thyroid gland was collected" or that "we dissected out the entire thyroid gland". This is essential to understand the significance of the cell populations that are described based on the transcriptomic data. The section "Tissue collection" describes dissection of the thyroid for whole-mount imaging. From Fig. 6A it seems that substantial amounts of non-thyroid tissue are included in this dissection. Is it the same kind of dissection that was performed for transcriptomics? Was the string of thyroid follicles shelled out from their surroundings or was some kind of en bloc dissection, including other neighboring tissues, performed (as suggested from the transcriptome cell populations data including e.g. gill transcripts)? In the latter case it would be good if the authors discuss in more detail what other tissues or structures that are expected to also be included in the dissected tissue and transcriptomic data.

It would facilitate understanding if the thyroid is outlined in Fig. 1A as well as the region that was dissected for downstream single cell sequencing.

The clusters seem logical given what cell types that could be expected in the region (but depending on how dissection was performed). The only exception is cluster 7 (non-follicular epithelium; NFE). I do understand that relative sizes of the clusters do not necessarily reflect the endogenous relative abundance of different cell types, as I guess they may be more or less prone to enzymatic dissociation, survival etc. Nevertheless, the number of cells in the NFE cluster (831 cells) seems sizeable relative to the number of thyrocytes (267 cells). In my opinion, a major weakness of the current manuscript is that little detail is provided about this cell population and that no attempt to at least spatially localize these cells is presented.

The NFE cells are characterized by TP63 expression and the authors speculate that they might show homology to main cells of solid cell nests. From previous zebrafish literature it seems like what is supposedly ultimobranchial bodies (or ultimobranchial glands more similar to those in avian species) are located rather distant from the thyroid follicles (Alt et al 2006). Is it possible that these structures are included in the dissection that has been performed? As solid cell nests are supposed to be ultimobranchial body remnants (with calcitonin positive and calcitonin negative epithelial cells) it would be good if the authors discuss in more detail what is known about the ultimobranchial bodies in zebrafish, if they are located inside the zebrafish thyroid, in an anatomical region that is included in the dissected tissue of this study or in a region that is likely not included.

In higher vertebrates, P63 expression is typically seen in basal cells of stratified epithelia (as for instance in the esophagus), in myoepithelial cells, in the urothelium and in the thymus. Is it possible that the TP63 expressing NFE population corresponds to cells of the zebrafish thymus (that might perhaps explain the seemingly large immune cell population in cluster 4)? Could TP63 expressing NFE cells represent the esophagus (if included in the dissection)? As so little detail is provided about the dissection procedure this opens up for speculation and it would be good if the authors discuss these possibilities, as some of them might perhaps be unlikely or even impossible given regional anatomy of the zebrafish and how the dissection was performed.

To gain better understanding of the sizeable TP63 expressing NFE population the authors briefly mention the possibility of future studies utilizing a TP63 reporter. If a reporter line is not available, immunofluorescent detection of P63 (as presented for PAX2A in Fig. 5 and E-cadherin in Fig. 6) would be desirable to provide more insight into the location of the NFE population. Given the proficiency the authors demonstrate in this manuscript when it comes to zebrafish imaging, at least whole-mount immunostaining of P63 in the region that was dissected for transcriptome analysis seems clearly feasible, both with respect to resources and time needed (perhaps in the range of 1-3 months).

Minor comments

It is a little bit confusing that different color coding for the various cell populations is used in Fig. 3B as compared to Figs. 1 and 2.

In the discussion of the intercellular interaction network (Fig. 3B) the authors clearly point out that anatomical barriers are not modelled. Nevertheless, it would be more informative if this description was able to sort out ligands that are secreted, from ligands that are not secreted and would require physical interaction between thyrocytes and a different cell population for signaling to occur.

The authors describe thyroid solid cell nests as "... lumen containing irregular structures located between the thyroid lobes in mammals...". Solid cell nests of the thyroid in higher vertebrates (e.g. humans and dogs) are located within the thyroid lobes and not between the lobes (i.e. the right and left thyroid lobes). Moreover, the authors write that "Recently, epithelial cells have been reported in a structure called the Solid Cell Nests (SCN) of the thyroid..." and give reference to a paper from 1988. If that is recent or not might be a matter of opinion, but to the best of my knowledge, solid cell nests were describe already by Getzowa in 1907 and I suppose that the epithelial identity (or at least epithelioid morphology) has been appreciated for long.

Significance

The authors provide a single-cell transcriptomic atlas of the zebrafish thyroid gland. To the best of my knowledge this is certainly a unique resource. In that sense it provides novel and significant information that will surely facilitate our further understanding of thyroid biology. It will surely be of great interest and value to the thyroid community, but probably also to a wider audience interested in e.g. zebrafish biology, endodermal biology and the biology of endocrine glands. Their finding and direct demonstration of transcriptional heterogeneity within the thyrocyte population is very interesting, also in different contexts of thyroid disease. However, I leave it to other referees to comment on the conceptual uniqueness of the current manuscript (i.e. a single-cell transcriptomic atlas of a zebrafish organ). Does it provide conceptually unique information, or does it add to an expanding collection of single-cell transcriptomic atlases of zebrafish organs?

Own field of expertise: thyroid gland biology, endoderm 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

Gillotay et al. performed an unbiased profiling of the zebra fish thyroid gland and captured different cell populations by single-cell gene expression analysis. Using bioinformatic tools, they identified seven clusters corresponding to expected, but also poorly characterized, sub-populations, such as non-follicular epithelial cells. They also found two transcriptionally distinct types of thyrocytes and validated this heterogeneity using a new transgenic Pax2a reporter line. Using this tool, they identified and located Pax2a-low and -high thyrocytes within thyroid follicles. Finally, they highlighted a dense intercellular signaling network based on ligands expressed by the diverse sub-populations present in the thyroid and receptors expressed by the thyrocytes. This is a descriptive work calling for more in-depth analyses.

Major comments on main conclusions :

Conclusion 1 : Identification of 7 clusters

• This reviewer was particularly surprised by the relative abundance of the different sub-populations identified. For example, 267 thyrocytes out of 6249 cells from the thyroid gland is less than 5% of the total thyroid cell number. In comparison, authors identified three times more immune cells or non-follicular epithelial cells. Authors should comment on these numbers, on the dissection (contaminants?) and dissociation procedure. Do these relative abundance reflect the proportion of thyrocytes, immune, stromal cells they normally observe in adult thyroid sections ? It would be interesting to have a better resolution for figure 6A in order to evaluate the number of nuclei stained only with DAPI as compared to nuclei stained with DAPI and surrounded by E-cadherin staining. Based on the image, this reviewer seriously doubts that follicular cells represent less than 5% of the total cell number in this organ, considering that the colloid is a cell-free zone.
• When using the webtool developed on the thyrocyte population, one can notice that only a small fraction of the thyrocyte population expresses common thyroid-specific genes such as Tpo or duox. Was this expected ? It would be interesting to comment on this observation and confirm using standard localization technique to demonstrate that this is real and not due to the sequencing. Is it specific to the zebrafish ? On the other hand, Tg is expressed in most thyrocytes, but surprisingly also in all the clusters at a fairly good level. This should be commented... Is it normal ? due to the sequencing quality ? or clustering ?
• In order to validate and locate the different populations identified in the thyroid, this reviewer suggests to perform in situ hybridization or immunostaining, based on the specific marker genes identified in each cluster. This experiment could lead to the precise identification of the different sub-populations and their respective localization. These experiments would also help in the interpretation of the cellular interaction network.

Conclusion 2 : two distinct types of thyrocytes

• This is an interesting observation. However, from a non-expert it is difficult to understand why the authors propose two populations. Based on the points distribution (Figure 4A), this reviewer would rather identify 3 or 4 clusters but not the two shown in red and blue.... Did the authors impose two populations for the clustering ? Did they perform a permutation test to confirm the pax2a significant fold change seen between clusters is not a false positive generated by the clustering ? Could they show, as a supplementary file, the same graph with points colored based on Pax2a expression ?
• This reviewer was also surprised by the relatively "heavy" approach (generation of the Pax2a reporter line) used to demonstrate the existence of two types of thyrocytes. Knowing that the reporter line was validated with a very good Pax2a antibody... The use of the reporter line is a bit short. The authors could for example validate the two populations of Figure 4A using the Pax2a FACS-sorted cells and RT-qPCR.
• In addition, data available in the sequencing dataset could be used to prove that the two populations are really active thyrocytes. This reviewer would suggest to present a table with the expression level of common and thyroid-specific genes such as TshR, Nis, Tpo, Duox, Tg, Pax2, TTF1 and other known transcription factors in the two populations to demonstrate that these two types of cells are indeed thyrocytes. Finally, image quality (Figure 6) could be improved and high-magnification images with several thyrocyte marker could be shown to convince the readers.

Conclusion 3 : cellular interaction network

• Most of the interactions revealed by the analysis seem to belong to the extracellular matrix and not to classical ligands such a Wnt, TGFb, FGF, PDGF,..... could the authors comment on this ? Considering that both endothelial cells and epithelial cells assemble their own basement membrane, the analysis will obviously reveal interactions between endothelial cells and epithelial cells....

Minor comments to improve the Ms :

• Could you explain how from 2 x 12 000 FACS-sorted live-cells (from six animals at each stage; 2 mpf and 8 mpf) you obtain 6249 cells (pooled of 2 mpf and 8 mpf), and why the two stages were first sorted separately and then pooled (?), as no differential analysis is carried out for the two stages.
• Which method was used for the graph-based clustering ? KNN ? Louvain ? Random walk ?
• How did you define the numbers of clusters ?
• Figure 4B, the color-code for the expression level would help the reader.
• Figure 4C, violin plot for Pax2a: why do we find cells that do not express this gene in the two populations ? The same is true for tbx2a and ahnak ... is the clustering optimal ?
• Figure 4C, blue violin plot for ptp4a3 does not seem to fit with the distribution of the points.
• what is the function of tbx2a, ahnak, ptp4a3 and dusp5, which are not mentioned in the text.
• Line 195: "pax2amKO2 reporter expression perfectly overlapped with PAX2A antibody staining". This reviewer would be more cautious as the images (Fig. 5 C, D and F) do not show a perfect colocalization: one can observe only blue or only red staining.
• Line 246: it is proposed to "study the functional and replicative differences among the two sub-populations of thyrocytes". This reviewer completely agrees and the suggestion made (vide supra) to use the datasets to assemble a table with the expression level of common and thyroid-specific genes such as TshR, Nis, Tpo, Duox, Tg, Pax2, TTF1 and other known transcription factors in the two populations could already give some indications on the functionality of these two types of cells. Expression of genes involved in cell-cycle control and/or apoptosis would be another possibility to better characterize the two populations. Lastly, the authors could perform the comparative analysis of ligand-receptor pairs between these two sub-populations to examine if they differentially interact with their environment.

Text improvement: Intro: thyroglobulin (TG) appears twice (line 46 and 49) Results: Fig. 5 (not 8) (line 203 and 205) Figure 3: stromal? (not skeletal) Figure 4: fold change scale is missing Figure 5: Thyroid gland (Gland) Figure sup 2: Fluorescence-activated cell sorting (FACS) of zebrafish thyroid gland Figure sup 3: number of unique molecular identifier Figure sup 4: "are present in the zebrafish"

Significance

The work performed by Gillotay et al. is clearly novel but descriptive. It provides a useful database to propose hypotheses to further study the thyroid gland. The single-cell RNA-seq analysis brings a molecular appreciation of the various thyroid cell populations, thyrocyte heterogeneity and intercellular signaling network. Although focused on the thyroid, results will be of interest to a larger audience than the thyroid community, especially the demonstration (if further and better validated) of thyrocyte heterogeneity and the intercellular communication possibilities.

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 work reports finding a molecular genetic basis for individual differences in behavior in different strains of outbred mice, even including individual behavioral differences between mice of the same inbred genetically isogenic strain. The authors were able to measure copy numbers for the tandemly repeated intronic snoRNA clusters SNORD115 and SNORD116 and found correlation with measures of anxiety in open-field test and elevated plus maze. Expression data for previously proposed targets of these snoRNAs are also provided.

We note that this description represents only part of the achievements in our paper. The key of our paper is that we have not only used "different strains of outbred mice", but in addition one very different species of mouse (Apodemus) and a Guinea pig species. We believe that the test in very different species with very different genetic backgrounds is the crucial proof for the specificity of the effect.

**Major comments:** 1.The techniques to measure copy numbers are challenging and the authors' conclusion that ddPCR was their method of choice is convincing. They were able to obtain limited optical mapping (Bionano zephyr) data, only for SNORD116 and only in mouse, but these data are useful to corroborate those obtained with ddPCR. 2.Figure 3 reports single copy numbers for individuals that are presumably heterozygous. Do we have to assume that the numbers reported represent the larger alleles since the ddPCR method does not allow to distinguish two different size alleles, as was shown for optical mapping?

The numbers are derived from the whole genome DNA, i.e. represent the cumulative copy number of both alleles. We have updated the text to make this clear.

3.The analyses reported do not take into account the specific parental origin of the alleles used in the regression analyses. Since PWSCR-specific SNORDs are only expressed from the paternal chromosomes, this generates some uncertainty about the whole dataset.

We had explained why it is not possible to distinguish the two alleles with the current technology. Hence, it is evidently also not possible to determine which allele comes from the paternal side. In fact, given that we showed that copy numbers can change every generation, even the knowledge of which chromosome is the paternal one would not predict its copy number. Accordingly, it lies in the nature of the whole phenomenon that this uncertainty is given. It is therefore just the more surprising that we still can observe correlations that are much stronger than has been shown for natural alleles of any genetic locus implicated in behavioral traits so far.

4.Lines 353-365: The ankrd11 exon-specific RNAseq data are confusing and too preliminary. More work needs to be done to resolve the splice variants in this region and their relationship to SNORD116 copy numbers. Alternatively lines 356-361 should be deleted.

We have included the data to show that the mechanism must be different from the one that is seen for Htr2c. This difference is clearly documented and we should therefore like to retain it. What is missing is to show the actual mechanism by which SNORD116 causes the alternative splicing. This will require more biochemical approaches that go beyond the current study.

5.In all tested rodents, higher SNORD copy number was correlated with higher relative anxiety score. In the human samples, however, higher anxiety scores were associated with lower copy numbers. These apparently contradictory results are not mentioned in the abstract, and are not satisfactory explained in the text.

We have decided to leave the human data out from the current manuscript. First, the behavioral tests for the rodents are indeed not directly comparable with the questionnaire scores in humans. Second, in human genetics one usually asks the results to be confirmed in an independent study. hence, we plan to extend the human data, but to present them eventually in a follow-up paper.

6.Extension to other species would be desirable but was limited by availability of genomic data: Results are presented for wood mouse only for SNORD115 and for the guinea pig for SNORD116.

Given that we show a strong correlation between SNORD115 and SNORD116 copy numbers for those species where the information is available for both loci, we do not think that this is a major limitation of our study.

**Minor comments:** 1.The authors present skull shape data related to SNORD116 copy numbers, but fail to consider how these data are relevant to the craniofacial abnormalities reported in an ankrd11 mutation. Barbaric et al (2008) reported a dominant ENU- induced mutation caused shortened snouts, wider skull, deformed nasal bones, reduced BMD, reduced osteoblast activity and reduced leptin levels. This phenotype was traced to a heterozygous missense mutation (conserved glutamate to lysine change) in an HDAC binding site. They postulated that the mutation fails to recruit HDACs to a transcription complex and to inhibit hormone-receptor activated gene transcription. What is the postulated link between this mechanism and the here reported skull shape data correlated with SNORD copy number variation?

The described missense mutation is located in the differentially spliced exon, i.e. a direct functional link is given. We have added this information to the text and compared the direct phenotypic effects from their study and our study.

2.The observed co-variation of copy numbers between the two SNORD clusters could indicate a duplication involving the entire region, This could be tested by determining the dosage of IPW, UBE3a and Snrpn genes.

While this is a theoretical possibility, it was not described in the literature before. Also, in our systematic survey of copy number variation in mouse populations (Pezer et al. 2015) we did not find a deviation of these genes from expected diploid copy number in any of the populations analysed.

3.Line 129 "the RNA coding region" and Line 148 "snoRNA coding parts" (and elsewhere) does seems correct, as by definition, this is non-coding RNA. The region they are referring to could be called the "processed C/D box snoRNA". The mechanism that generates these C/D box snoRNAs is well established: the "genes" are located in introns of host genes - and after transcription - the spliced out introns are exonucleolytically trimmed to the functional sizes. Both SNORD115 and 116 clusters are within a large transcript that originates from the SNRPN promoter of the paternal allele.

We adjusted the wording to make clear that we refer to the mature RNAs.

4.Figure 2 does not show data on skull shape as claimed in the legend.

We apologize - this was a carry-over from an older version of this figure. The skull shape analysis had been moved to a new figure in the current version of the manuscript.

5.S1 Figure: Snprn should be Snrpn

Thank you for spotting the error - we have corrected this

Reviewer #1 (Significance (Required)): This provocative work proposes the regulation of behavioral variance by dosage changes of a regulatory RNA. The dosage changes are apparently caused by dynamic and frequent alteration in copy number. This is a novel concept and worthy of publicizing. Extensive data documentation is provided for others to analyze and possibly replicate. The data potentially throw light on the function of the tandemly repeated imprinted snoRNA clusters in the PWS critical region. Novel aspects of this work include the discovery of copy number variation of these snoRNAs; and validation of a target of SNORD116: Ankrd11 is one of many potential targets of SNORD116 that was previously computationally predicted, this paper reports experimental evidence for this interaction. The work would be of interest to researchers in behavioral evolution, non-coding RNA function, epigenetics and overall genome evolution. Define your field of expertise with a few keyword: Molecular genetic disorders, Prader-Willi syndrome, mouse models

Reviewer #2 (Evidence, reproducibility and clarity (Required)): **Summary** Maryam Keshavarz et al. aimed at seeking the molecular basis underlying individual behavioral variance within populations. Focusing on the Prader-Willi Syndrome (PWS) gene complex, which has been well recognized being associated with neurodevelopmental disorders, anxiety and metabolic issues, the authors found that the levels of PWS region's small nucleolar RNAs SNORD 115/116 of individual animals correlated with these individuals' behavioral test scores. The variations in transcript processing of some anxiety-associated target genes also revealed correlation with SNORD 115/116 copy numbers. Authors implicated that the copy numbers of SNORD 115/116 within PWS plausibly influenced behavioral variances among individuals. • Authors first validated that the droplet digital PCR (ddPCR) was suitable for quantifying variations in copy numbers of genomic clusters. Their ddPCR data showed confident correspondence with reads calculation of whole-genome-seq dataset. Also, ddPCR showed good replicability and congruent tissue-to-tissue similarity. • Authors found the ddPCR-measured SNORD copy numbers from several mice populations showed significant regression with SNORD RNA levels, respectively. Also, the anxiety profiling using Open Field Test and Elevated Plus Maze test indicated a significant regression between SNORD copy numbers and anxiety profiling scores, namely individual mouse with higher copy numbers received higher relative anxiety scores. Some other representative genes outside PWS, such as Sfi1 and Cwc22, failed to show such copy number-anxiety score regression. • Authors applied RNA-seq of individual mice with different SNORD 115/116 copy numbers and analyzed potential target gene regions. They found the level of alternative splice-resulted exon Vb of gene Htr2c, a serotonin receptor, was positively correlated with SNORD 115 copy number. Also, an alternative splicing product of gene Ankrd11, a chromatin receptor regulating GABA receptor, was found to positively correlated with SNORD 116 copy number. Positive correlation to SNORD copy numbers also occurred to some Htr2c and Ankrd downstream genes. • Authors used a landmark-based analysis to score mice craniofacial features and found the scores were in relationship with SNORD 116 copy numbers. • Authors also found significant regression between SNORD copy numbers and behavioral evaluations in other rodents. In humans, the Tridimensional Personality Questionnaire (TPQ) based evaluation also showed correlation with SNORD 115 and 116 copy numbers. **Major comments** The study mainly revealed important correlations between copy numbers of 2 small nucleolar RNAs and cognitive behavioral variance of different individual animal. Although very useful and important findings, the study did not provide any evidence about the causality between SNORD 115/116 and the observed behavioral phenotypes. For instance, • #1: the behavioral observations (i.e. anxiety) may not be merely regulated by the PWS gene complex.

It is already well understood that the respective behavioral observations have a polygenic basis. But our data show that the SNORD copy numbers act as major modulators of the behavior.

• #2: the paper did not show if manipulations on mouse SNORD 115/116 could affect target genes as well as the consequential behavioral phenotypes.

A direct interaction between SNORD115 and its target gene HTr2c has previously been shown in cell culture experiments. Further, we show that the commonly used inbred mouse strain C57Bl6 carries already different copy number alleles that would not be different from artificial manipulation of the copy number. There is a long tradition in mouse genetics to accept also spontaneous alleles as genetic proof, not only the alleles that were created by artificial intervention.

Further, as also pointed out in response to reviewer 1, in the absence of the possibility to do a direct genetic manipulation in a given genetic background, we use the comparative analysis between different genetic backgrounds to prove causality.

Reviewer #2 (Significance (Required)): Authors provided a potential molecular basis regulating the PWS region, which is a genomic imprinted gene complex and related to many neurodevelopmental diseases in mammals, including humans. Considerably cost-saving than whole-genome deep-seq, the application of droplet digital PCR on copy number (esp. in stretching regions) measurement can overcome some technical difficulties, for example, qPCR has limit in resolution when differentiating subtle variance in copy numbers; the Nanopore seq and current mapping algorithm show difficulties when placing the internal repeats also. Authors proposed SNORD copy number as a potential explanation to the individual-to-individual variance within the same species or even the same population.

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

Maryam Keshavarz et al. aimed at seeking the molecular basis underlying individual behavioral variance within populations. Focusing on the Prader-Willi Syndrome (PWS) gene complex, which has been well recognized being associated with neurodevelopmental disorders, anxiety and metabolic issues, the authors found that the levels of PWS region's small nucleolar RNAs SNORD 115/116 of individual animals correlated with these individuals' behavioral test scores. The variations in transcript processing of some anxiety-associated target genes also revealed correlation with SNORD 115/116 copy numbers. Authors implicated that the copy numbers of SNORD 115/116 within PWS plausibly influenced behavioral variances among individuals.

• Authors first validated that the droplet digital PCR (ddPCR) was suitable for quantifying variations in copy numbers of genomic clusters. Their ddPCR data showed confident correspondence with reads calculation of whole-genome-seq dataset. Also, ddPCR showed good replicability and congruent tissue-to-tissue similarity.

• Authors found the ddPCR-measured SNORD copy numbers from several mice populations showed significant regression with SNORD RNA levels, respectively. Also, the anxiety profiling using Open Field Test and Elevated Plus Maze test indicated a significant regression between SNORD copy numbers and anxiety profiling scores, namely individual mouse with higher copy numbers received higher relative anxiety scores. Some other representative genes outside PWS, such as Sfi1 and Cwc22, failed to show such copy number-anxiety score regression.

• Authors applied RNA-seq of individual mice with different SNORD 115/116 copy numbers and analyzed potential target gene regions. They found the level of alternative splice-resulted exon Vb of gene Htr2c, a serotonin receptor, was positively correlated with SNORD 115 copy number. Also, an alternative splicing product of gene Ankrd11, a chromatin receptor regulating GABA receptor, was found to positively correlated with SNORD 116 copy number. Positive correlation to SNORD copy numbers also occurred to some Htr2c and Ankrd downstream genes.

• Authors used a landmark-based analysis to score mice craniofacial features and found the scores were in relationship with SNORD 116 copy numbers.

• Authors also found significant regression between SNORD copy numbers and behavioral evaluations in other rodents. In humans, the Tridimensional Personality Questionnaire (TPQ) based evaluation also showed correlation with SNORD 115 and 116 copy numbers.

Major comments

The study mainly revealed important correlations between copy numbers of 2 small nucleolar RNAs and cognitive behavioral variance of different individual animal. Although very useful and important findings, the study did not provide any evidence about the causality between SNORD 115/116 and the observed behavioral phenotypes. For instance,

• #1: the behavioral observations (i.e. anxiety) may not be merely regulated by the PWS gene complex.

• #2: the paper did not show if manipulations on mouse SNORD 115/116 could affect target genes as well as the consequential behavioral phenotypes.

Significance

Authors provided a potential molecular basis regulating the PWS region, which is a genomic imprinted gene complex and related to many neurodevelopmental diseases in mammals, including humans.

Considerably cost-saving than whole-genome deep-seq, the application of droplet digital PCR on copy number (esp. in stretching regions) measurement can overcome some technical difficulties, for example, qPCR has limit in resolution when differentiating subtle variance in copy numbers; the Nanopore seq and current mapping algorithm show difficulties when placing the internal repeats also.

Authors proposed SNORD copy number as a potential explanation to the individual-to-individual variance within the same species or even the same population.

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 study aims to develop tools for yeast researchers to automatically segment and classify yeast colonies. The machine learning method enables rapid classification compared to manual counting.

\*MAJOR CONCERNS:***

Please include additional details about the types of images that must be captured for segmentation and categorization. It is important to provide details of what level of magnification might be needed during image capture. We anticipate that providing clear protocols for altering thresholds to classify colonies might be one way to overcome this challenge

That’s correct. Details on image acquisition, such as the level of magnification, are important to obtain accurate results.

To address this, we provide a detailed protocol in our companion article on ProtocolExchange: https://protocolexchange.researchsquare.com/article/nprot-7305/v1

We have updated the manuscript to include this link.

While the program crops colonies and segments them accurately, there is no spatial information of where these colonies are located in the image. This loss of spatial information limits the ability to use this platform to identify colonies of interest following experiments such as a genetic screen.

In principle it would be possible to retain the location of each cropped colony in the form of (x,y) coordinates in pixels. This could be included in a future release. However, we doubt the utility of such information for a genetic screen, unless identification of a positive hit could be linked to robotic picking of the identified colony (which would certainly go beyond the scope of this work). In reality, researchers will pick positive hits manually anyways, making our pipeline superfluous for such an application. We emphasize that we have developed our pipeline for large-scale quantification of red/white color assays. Here the pipeline makes a huge difference, as compared to manual counting.

The inability to accurately recognize sectored colonies as sectored (rather than red) is a significant limitation to the usage of this program for quantitative assays. While differentiating between red and white colonies is useful, the conclusion by the authors about its value for quantitative assays is limited unless variegation can be accurately defined. The authors should either soften this conclusion or qualify what quantitative measurements might mean given the limitations of their classification program. This somewhat diminished our overall enthusiasm.

The reviewer correctly points out that our algorithm shows lower accuracy when differentiating between red and variegating colonies than when differentiating between white and non-white colonies (including red, variegating and pink). Given this observation, we initially focused on predicting white vs. non-white colonies with our tool. However, the output of our pipeline also includes more granular predictions of numbers of white, red, pink and variegating colonies. We therefore leave it up to the user to decide which level of granularity is more appropriate, taking into account the tradeoff between granularity and accuracy. In particular, we note that for the colonies we tested, splitting the non-white category of predictions into red, variegating and pink resulted in a decrease of sensitivity from 0.98 for the non-white category to 0.86-0.88 for the individual categories, while the corresponding specificity showed a smaller reduction from 1.0 to 0.97-0.98. Considering that a lack of any predictions for the red, pink, and variegating categories effectively prohibits the researcher from detecting them at all, even a reduced sensitivity may be better than nothing and therefore acceptable in this case. In order to make this clearer in the text, we provide a more detailed comparison of performance metrics between levels of prediction, which may help to guide the user’s decision.

This program must be benchmarked with other colony classifiers. Cell Profiler is an example of a popular yeast colony segmentation program. How does this machine learning based tool compare with other colony segmentation and categorization programs. One possibility is to include an additional figure that compares their program with clear benchmarks. The outcome of effort based on benchmarking is not as important since we believe it is useful to have many alternatives for yeast segmentation and categorization. We think this revision would be essential to the manuscript and would add significant value.

We have used other approaches and were not satisfied with the outcomes. Hence, we developed our own pipeline, specifically designed to accurately distinguish red from white colonies and quantify such assays at a large scale.

When using CellProfiler we could not reliably distinguish variegating, pink, and red colonies. White colonies show up in the Red, Blue and Green channel, Red colonies mainly in the Red Channel. Therefore, variegating, pink and red colonies can be distinguished from white by reduced Blue and Green values, which is indirect and caused several issues. One of the problems was reflection of the flash during image acquisition, giving two reflective white patches on each colony that differed in pixel size depending on the magnification and colony size. We tried to prevent reflection with a ‘tent’, which reduced but could not eliminate the reflection. Therefore, the MaxIntensity of the Green/Blue channel was always the same of each colony, impeding classification. Furthermore, most red/pink colonies had a slim white rim, which was sometimes bigger/smaller and the relative area of rim to colony depends on the colony size, which made it impossible to tell a bit variegating from red by the output values from CellProfiler.

If deemed useful by the editors, we will be happy to mention this in the manuscript. A systematic comparison with other classifiers seems to be a bit of an overkill though. As stated by this reviewer, the outcome of such comparison would not matter much. It is important that the community has several approaches to choose from, so that the best solution can be found for each specific application.

\*MINOR CONCERN***

The program currently saved cropped images of each segmented colony. This takes up a lot of storage space. It might be useful to provide an option to save or not save these cropped images. This flexibility will be valuable for users but does not detract from the major conclusions of the manuscript.

While we appreciate that the need to save individual images of cropped colonies may be a drawback for some users, in the current implementation it is not possible to avoid this step. One could imagine a scenario in which all cropped images were stored in RAM prior to classification rather than written to a computer’s disk; however, we believe that most users would have more limitations on the availability of RAM than on disk storage, therefore making this option also not feasible.

The authors have provided excellent examples of colonies they believe are red, white or sectored. More accurately defining a pink colony would be valuable for users of this program. How much of red is classified as pink by this program?

As the reviewer points out, it is difficult to give an objective definition of a pink colony. In this case, we relied exclusively on subjective expert annotations to define which colonies were pink (as well as for all other categories).

We acknowledge that this may introduce some error into the model, as there may be some overlap between red and pink colonies or between pink and variegating colonies; however, this problem also exists in the case of manual annotation. As shown in Figure 1d, for the colonies we tested, 4 out of a total of 55 colonies annotated as red by an expert were predicted as pink by our algorithm. We would like to emphasize that our pipeline alleviates biases between different researchers who would be annotating colony color manually, therefore improving reproducibility. Such biases could be subjective or objective, such as different monitors used to inspect the images.

Providing an example data set with the protocol would be helpful for users with limited Python experience. In combination with their protocol on Protocol exchange, this would serve as a valuable resource for novices in programming.

We agree with the reviewer’s suggestion and will be happy to provide an example dataset used in the manuscript. We will defer to the journal’s guidelines as to the best way to share these raw images.

One technical issue of the program is that the program tries to open all files in the specified folder even if they aren't jpg. This causes problems if there are additional or hidden files in the folder and the program cannot process the additional files.

We appreciate the reviewer pointing out this issue and have fixed it in a new version of the code.

Reviewer #1 (Significance (Required)):

This manuscript describes a machine learning approach to segment and categorize yeast colonies based on a red/white selection assay. The approach has been implemented using Python which makes this widely accessible to many researchers. Their detailed protocol on Protocol Exchange is a valuable resource which made it possible for us to evaluate its performance. The program meets its goals of reducing user time via manual counting. It is also reasonably accurate in discriminating between red and white colonies based on our initial tests. However, there are several important concerns that the authors will need to address before this manuscript can become a valuable resource for the yeast community. It is important to note that our framework is one where we have a great interest in quantitative yeast genetics but cannot evaluate the strengths and weakness of the computational approach. So much of the review is focussed on what would be needed to make this tool more user appropriate.

Reviewer #2

(Evidence, reproducibility and clarity (Required)):

\*Summary:***

Carl et al present an application of a deep learning-based image analysis able to segment and classify individual yeast colonies by their phenotype in a special plate. They evaluated the method and show that it provides the accuracy similar to the one achieved by experts' manual classification.

\*Major comments:***

The key conclusions are convincing. The evaluation is performed on 3 datasets showing different properties (strong presence of phenotype, almost lack of the phenotype, gradual change of the phenotype).

The claims are carefully formulated. The deep learning methodology (training, validation, using modern technologies such as transfer learning, Unet, augmentation) is carefully designed and carried out. The evaluation is sound. The limitations are discussed.

For a short paper as it's formulated currently, no additional experiments are necessary.

The methods are implemented and are available on GitHub.

However, I'd strongly recommend to share also the data used in the paper, to make possible the reproduction of the results as well as to be used as examples for future users.

As stated above, we agree with the reviewer’s suggestion and will be happy to provide an example dataset used in the manuscript. We will defer to the journal’s guidelines as to the best way to share these raw images.

No replicates are provided unfortunately. The manuscript would benefit from showing results from replicates, especially because they should be easily obtainable.

It is not clear to us to which experiment the reviewer is referring. All of the results presented in Figure 2 did include replicates, as detailed in the figure legend.

\*Minor comments:***

I'm not familiar with the state of the art to judge on whether prior studies are referenced.

The text and fitures are very clear and well formulated.

Reviewer #2 (Significance (Required)):

Despite the conceptual innovation is average, the method is well-developed and seems to be very useful for yeast analysis.

I'm not an expert in the application area to judge the state of the art. The carried out deep learning methodology is top notch.

The manuscript can be interesting and useful for experts using the described assay for yeast.

My expertise is in omics, image analysis, and machine learning.

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:

Carl et al present an application of a deep learning-based image analysis able to segment and classify individual yeast colonies by their phenotype in a special plate. They evaluated the method and show that it provides the accuracy similar to the one achieved by experts' manual classification.

Major comments:

The key conclusions are convincing. The evaluation is performed on 3 datasets showing different properties (strong presence of phenotype, almost lack of the phenotype, gradual change of the phenotype).

The claims are carefully formulated. The deep learning methodology (training, validation, using modern technologies such as transfer learning, Unet, augmentation) is carefully designed and carried out. The evaluation is sound. The limitations are discussed.

For a short paper as it's formulated currently, no additional experiments are necessary.

The methods are implemented and are available on GitHub.

However, I'd strongly recommend to share also the data used in the paper, to make possible the reproduction of the results as well as to be used as examples for future users.

No replicates are provided unfortunately. The manuscript would benefit from showing results from replicates, especially because they should be easily obtainable.

Minor comments:

I'm not familiar with the state of the art to judge on whether prior studies are referenced.

The text and fitures are very clear and well formulated.

Significance

Despite the conceptual innovation is average, the method is well-developed and seems to be very useful for yeast analysis.

I'm not an expert in the application area to judge the state of the art. The carried out deep learning methodology is top notch.

The manuscript can be interesting and useful for experts using the described assay for yeast.

My expertise is in omics, image analysis, and machine learning.

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 study aims to develop tools for yeast researchers to automatically segment and classify yeast colonies. The machine learning method enables rapid classification compared to manual counting.

MAJOR CONCERNS:

Please include additional details about the types of images that must be captured for segmentation and categorization. It is important to provide details of what level of magnification might be needed during image capture. We anticipate that providing clear protocols for altering thresholds to classify colonies might be one way to overcome this challenge

While the program crops colonies and segments them accurately, there is no spatial information of where these colonies are located in the image. This loss of spatial information limits the ability to use this platform to identify colonies of interest following experiments such as a genetic screens.T

The inability to accurately recognize sectored colonies as sectored (rather than red) is a significant limitation to the usage of this program for quantitative assays. While differentiating between red and white colonies is useful, the conclusion by the authors about its value for quantitative assays is limited unless variegation can be accurately defined. The authors should either soften this conclusion or qualify what quantitative measurements might mean given the limitations of their classification program. This somewhat diminished our overall enthusiasm.

This program must be benchmarked with other colony classifiers. Cell Profiler is an example of a popular yeast colony segmentation program. How does this machine learning based tool compare with other colony segmentation and categorization programs. One possibility is to include an additional figure that compares their program with clear benchmarks. The outcome of effort based on benchmarking is not as important since we believe it is useful to have many alternatives for yeast segmentation and categorization. We think this revision would be essential to the manuscript and would add significant value.

MINOR CONCERN

The program currently saved cropped images of each segmented colony. This takes up a lot of storage space. It might be useful to provide an option to save or not save these cropped images. This flexibility will be valuable for users but does not detract from the major conclusions of the manuscript.

The authors have provided excellent examples of colonies they believe are red, white or sectored. More accurately defining a pink colony would be valuable for users of this program. How much of red is classified as pink by this program?

Providing an example data set with the protocol would be helpful for users with limited Python experience. In combination with their protocol on Protocol exchange, this would serve as a valuable resource for novices in programming.

One technical issue of the program is that the program tries to open all files in the specified folder even if they aren't jpg. This causes problems if there are additional or hidden files in the folder and the program cannot process the additional files.

Significance

This manuscript describes a machine learning approach to segment and categorize yeast colonies based on a red/white selection assay. The approach has been implemented using Python which makes this widely accessible to many researchers. Their detailed protocol on Protocol Exchange is a valuable resource which made it possible for us to evaluate its performance. The program meets its goals of reducing user time via manual counting. It is also reasonably accurate in discriminating between red and white colonies based on our initial tests. However, there are several important concerns that the authors will need to address before this manuscript can become a valuable resource for the yeast community. It is important to note that our framework is one where we have a great interest in quantitative yeast genetics but cannot evaluate the strengths and weakness of the computational approach. So much of the review is focussed on what would be needed to make this tool more user appropriate.

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 grateful to all three reviewers for their careful analysis of the manuscript, and for their constructive comments. Two common critiques were:

(1) that assaying origin firing via an independent method would strengthen the conclusions, and (2) that additional analysis of ribonucleotide incorporation to exclude the retention of lagging-strand primers would allow us to definitively determine whether Pol ɛ plays a role in lagging-strand synthesis.

We will include experiments to address both critiques in a revised manuscript. To independently verify changes in origin efficiency, we will sequence nascent BrdU-containing DNA across a time course from cells released into S-phase: we will also use the last timepoint of our Okazaki sequencing analysis to control for potential cell-cycle differences. To further test the contribution of Pol ɛ and ascertain whether lagging-strand primers are retained, we will analyze ribonucleotide incorporation in both wild-type and pol2-M644L (Pol ɛ ribonucleotide hypo-incorporating) strains. We address individual specific comments and our planned revisions in more detail below.

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

This study examined the consequences of limiting levels of DNA polymerase d (Pol d) in yeast. The authors first reported multiple genome instability consequences following lowered Pol d level, including defect in S phase progression, growth defect, elevated spontaneous DNA damage, accumulation of ssDNA and activation of replication and DNA damage checkpoint. These observations are solid but not unexpected. By genome wide analysis using the Okazaki fragment (OF) mapping and ribonucleotide mapping (for polymerase usage), the authors claim a few potentially novel and striking observations that lowered Pol d differentially impact efficiencies of early vs. late origins, and that lowered Pol d results in Pol e participating in lagging strand synthesis around late origins. However, I remained unconvinced based on the data presented. These observations need to be further substantiated and alternative interpretations should be considered.

\*Main concerns:** *

One of major conclusions the authors tried to make is that the early vs. late origins are differentially affected by low level of Pol d. First, they used OF mapping data to examine origin efficiency. Asynchronous "Cultures were treated with IAA for 2h before the addition of rapamycin for 1h to deplete DNA ligase I (Cdc9) from the nucleus via anchor-away". IAA concentrations used were of 0, 0.2 mM, 0.6 mM, and 1 mM. The problem is that Figure S1 clearly showed that treating asynchronous cultures with >0.1 mM of IAA for as short as 30 min significantly alters the cell cycle profiles, mainly resulting in accumulation of S phase cells, to different extent. Presumably Okazaki fragments accumulated from these cultures suffering from the synchronizing effect may not be representative of the real change in global replication profile. For instance, it is not difficult to predict that the Okazaki fragments enrichment may be skewed towards late origins if more cells are accumulated in mid S phase following Pol d depletion. For this reason, I don't believe the result is conclusive. The experiment may be re-designed for samples at different time points following release from G1.

We agree that altered cell-cycle profiles might affect the number of Okazaki fragments sequenced in late vs early replicating regions of the genome. As noted by reviewer 3 in cross-commenting, these differences should not affect origin efficiency calculations as these are based on the ratio of reads on each strand (and therefore normalized). To more directly address this question, we will calculate origin firing efficiencies from the final timepoint of the arrest-release experiments shown in Figure 4 as suggested by the reviewer. We will also analyze origin efficiency using BrdU over a time course.

This concern also should preclude the authors from drawing conclusion about Pol e usage on lagging strands based on comparison between HydEn-seq data and OF mapping data shown in Figure 6. In fact, the rNMP incorporation change is very similar between early and late origins. The only evidence that the author rely on is the discrepancy in OF data between the two groups origins, which makes the reliability if origin efficiency measurement the central piece of data in this study. Thus, alternative approaches should also be considered to map origin efficiencies.

As noted above, we agree that an independent method of tracking origin firing efficiencies would be helpful to strengthen our conclusions. To this end, we will analyze time courses of BrdU incorporation from cultures released into S-phase.

Even if Pol e strand bias is lowered at late origins, as the authors tend to believe, there are still alternative models other than Pol e being used for lagging strand synthesis. For instance, if TLS polymerases are used on lagging strands, it could result in more ribonucleotide incorporation on the lagging strand, as they are lower-fidelity polymerases. Alternatively, if Pol d scarcity leads to more Pol e synthesis or lower RNA primer processing, it might also contribute to more apparent ribonucleotide incorporation on the lagging strands.

We feel that the widespread use of TLS polymerases is unlikely, especially given the data in figure 6A that show no growth or viability change upon deletion of all three TLS polymerases in the Pol ∂ depletion strain, even at very low levels of Pol ∂. We agree with the reviewer that our data do not conclusively rule out increased retention of lagging-strand primers – as we state in the text. We aim to test this possibility by analyzing ribonucleotide strand bias in a pol2-M644L strain that incorporates fewer ribonucleotides than the wild-type Pol ɛ. In this case, increased lagging-strand primer retention would lead to a lagging-strand bias of ribonucleotides upon Pol ∂ depletion, while increased Pol ɛ usage would not. An analogous experiment with wild-type POL2 is potentially harder to interpret because the wild-type polymerase is the predominant source of ribonucleotides in a wild-type strain (Nick McElhinny et al, 2010 - PMID:20729855), but we now have the data for this strain in hand and ready to analyze.

In Figure S5, the two HydEn-seq replicates are very different, where replicate1 shows very low strand bias. I suspect perhaps the strain used for replicate 1 does not contain pol2-M644G or rnh202 deletion.

The change in ribonucleotide incorporation is indeed substantially stronger in one replicate than the other. We have additional time-course data from a Pol ∂ depletion showing that ribonucleotide strand bias decreases over time as Pol ∂ is depleted, and will include this in a revised manuscript.

Reviewer #1 (Significance (Required)):

Given that different aspects of Pol d deficiency have been implicated in various human diseases and cancer, this type of analysis is of interest to the field.

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

\*Summary** *

In this manuscript the authors explore the basis of the deleterious effects of reduced levels of the replicative Polymerase delta. This polymerase plays several important roles in the replication process: it synthesizes most of the lagging strand, but also extends the first primer during the synthesis of the lagging strand, and it contributes to the removal of the RNA and most of the DNA synthesized by primase/Polymerase alpha during Okazaki fragment maturation. In this study, the authors systematically analyze the impact of Pol delta depletion in S. cerevisiae. They use a degron-tagged allele to modulate the levels of the polymeraes and apply mainly NGS methods and classical genetics to explore the consequences for survival, checkpoint signalling and replication features such as fork speed, origin firing efficiency and Okazaki fragment length and distribution. They report that Pol delta depletion leads to a checkpoint activation via the Rad9-dependent damage signalling pathway (but not the Mrc1-dependent replisome-associated signalling) and an accumulation of single-stranded DNA. Phosphorylation of histone H2A is taken as a marker of DNA double-strand breaks, and from the observation that deletion of recombination factors, but not end-joining factors aggravate the fitness of Pol delta-depleted cells they conclude that homologous recombination is responsible for the repair of these breaks. Analysis of replication by Okazaki fragment sequencing indicates a slight decrease in the firing efficiency of early/efficient origins, but an increase in the firing efficiency of late origins. They also observe a reduction in fork speed, although they are not able to attribute this to either a globally slower fork movement or an increase in the stalling of individual forks. They find that Pol delta depletion does not change the size of Okazaki fragments, but causes defects in the nick translation during Okazaki fragment maturation. Finally, they use NGS technology to show that the leading strand Polymerase epsilon engages in lagging strand replication particularly at late origins when Pol delta is depleted. From their observations, the authors develop a model where depletion of Pol delta primarily affects late replicating regions. They explain this by invoking a stable association of Pol delta with early replisomes, which sequesters the enzyme, thus causing an under-supply at replisomes that assemble later during S phase. This then leads to the involvement of Pol epsilon on the lagging strand. Based on the observation that fork speed and Okazaki fragment maturation are both affected, they propose that these two reactions normally compete for Pol delta, suggesting that optimal replication would require two molecules of the polymerase per fork.

\*Major comments** *

The experiments shown here are largely clean and well controlled, and the manuscript is written nicely and well-structured.

Compared to the Okazaki fragment analysis, the treatment of double-strand breaks appears somewhat cursory and remains inconclusive. Phosphorylation of H2A seems insufficient evidence for double-strand breaks, as other structures could also give rise to that signal. These lesions should be detected in a more direct manner, e.g. pulsed-field gel electrophoresis. The authors also don't provide a mechanism by which such breaks would emerge. Related to the minor effect of the ku mutant, I am wondering about the altered appearance of the colonies in Figure 2F (concerning both ku70 and rad51) - what is different about these, and could their „denser" appearance explain the slight suppression effect observed?

We agree that our treatment of double-strand breaks is limited: consistent with comments from all three reviewers about which aspects of this work are most novel, we intend to focus as much as possible on replication enzymology here. We will tone down the language around double-strand breaks in the manuscript.

Concerning the damage signalling: it is surprising to see a damage signal at concentrations of IAA that do not lead to a significant depletion of Pol delta yet (0.05 mM). At this point, it is hard to imagine DSBs to form. Could the authors explain this discrepancy?

We note that, as observed in Figure 1A and to a slightly lesser extent in Fig. 2E, Pol ∂ levels are already substantially reduced in 0.05 mM IAA. This reduction appears sufficient to induce damage

The notion that late origin firing is enhanced despite checkpoint activation is counterintuitive. Do the authors think that this effect overcomes the suppression of late origins that is normally associated with checkpoint activation? It would be helpful to test whether abolishing that phenomenon (e.g. by a mec1-100 mutant) would enhance the effect and render late origins even more active.

We thank the reviewer for this excellent suggestion: we will test the effects of a mec1-100 mutant and include the results in a revised manuscript.

It would be important to characterise the fork speed defect better, using alternative methods rather than just relying on Okazaki fragments. A differentiation between slower fork progression and more frequent fork stalling would be relevant and might help to evaluate the contribution of Pol epsilon. This might be accomplished by DNA fibre analysis. Alternatively, BrdU incorporation could serve to observe replication over the entire genome rather than only in the vicinity of replication origins. It would also be important to differentiate fork speeds in early versus late replicating regions - according to the authors' model, the defects should be most obvious in the late regions (Fig. 4 concerns only early origins).

As noted above in our response to reviewer 1, we will use BrdU incorporation to independently verify changes in fork speed and origin firing. Analysis of fork speed in late-replicating regions is challenging regardless of the methodology used, due to contributions from converging forks, but we will try to do this

Figure S3: Considering the differences in cell cycle progression, it would make more sense to compare equivalent stages of the cell cycle / S phase rather than identical time points.

We can include this analysis, although the changes in cell cycle progression and origin firing efficiency make such comparisons somewhat subjective

Considering that the Okazaki fragment analysis was done with non-synchronised cultures, is it possible that the skew in the cell cycle profile could influence the Okazaki fragment pattern?

(copy-pasted from our response to a similar query by reviewer 1)… We agree that altered cell-cycle profiles might affect the number of Okazaki fragments sequenced in late vs early replicating regions of the genome. As noted by reviewer 3 in cross-commenting, these differences should not affect origin efficiency calculations as these are based on the ratio of reads on each strand (and therefore normalized). To more directly address this question, we will calculate origin firing efficiencies from the final timepoint of the arrest-release experiments shown in Figure 4 as suggested by the reviewer.

Would it be possible to monitor not only total Pol delta levels, but also the level of Pol delta bound to the chromatin? It is shown that the level of Pol delta is depleted in the whole cell extracts, but it would be interesting to see how severe the depletion is on the chromatin.

We agree that the relative fraction of chromatin-bound vs free Pol ∂ is an interesting question, and will attempt this experiment. However, we note that extensive depletion of Pol3 makes it hard to detect by Western blot, so the results are likely to be most informative at modest depletion levels. Regardless, these data should give us an idea of the size of the ‘free’ Pol ∂ pool in cells with normal or mildly reduced Pol ∂.

Figure 6 is confusing and should be clarified: - Figure 6B: assigning the Watson and Crick strands in the schematic would make that figure easier to understand; - Figure 6B-C: the axes are labeled as 'Fraction of rNMP on Watson strand', but would it not make more sense if they were labeled 'Fraction of rNMP in Crick strand'? - Figure 6D-E: the right side scale is labelled as 'increase in rNMP on Crick strand' while in the figure legend is says it is 'change in the fraction of ribonucleotides mapping to the Watson strand. That description should be clarified; - Figure 6D: using 'Change in Okazaki fragments strand bias' to label the black line (description in the box above the figure) instead 'Change in Okazaki strand bias' would be more appropriate; - Figure 6F: the authors seem to have subtracted strand bias measured for Okazaki fragments from the strand bias measured for rNMP. It is valid to subtract these biases from each other, considering that the two structures arise independently and with different frequencies?

We can make changes to figure 6 as suggested. Regarding the validity of subtracting strand biases, we think this is sufficient to give at least a semi-quantitative view of Pol ɛ usage since both of our sequencing approaches produce quantitative readouts that directly report on replication direction or polymerase usage, respectively.

\*Minor comments:** *

Can the authors conclude that Pol delta deficiency/ incompleteness of lagging strand synthesis affects the nucleosome deposition onto DNA? (Figure 5-A)

We cannot rule out that this is occurring, and we agree that this is an interesting question for future studies. But the changes that we observe Okazaki fragment terminus location are very similar to our previously published observations from cells lacking Rad27 function, consistent with decreased nick translation.

Why did the authors use rnh202Δ and not a mutant in the catalytic subunit of RNase H2?

Deletion of any subunit of the heterotrimeric RNase H2 complex completely abolishes its function in yeast, so RNH202 was a somewhat arbitrary choice

An extra control might be useful: comparing POL3-AID rnh202Δ with the POL3-AID pol2M644G rnh202Δ triple mutant could further confirm that the observed effect is Pol epsilon-dependent.

We agree (see also our response to reviewer 1). In addition to the wild-type, we will analyze pol2-M644L – a mutant in which Pol ɛ incorporates fewer than normal ribonucleotides. An increase in ribonucleotide density on the lagging strand in pol2-M644L would support increased primer retention on the lagging strand.

Figure 2H: It would be good to see the cell cycle distribution corresponding to the western blot images.

We can include this

Various spelling, grammar or precision of expression issues: - Pg. 4, line 4: endonucleolytically instead of nucleolytically. - Pg. 6, line 10: Remove 'was'. - Pg. 6, line 12: Remove 'in vivo' from the subtitle. - Pg. 6, line 14: 'an C-terminal' should be 'a C-terminal' - Pg 16, line 13: Phrasing implies that the synthesis of both leading and lagging strands by Pol delta in regions in the vicinity of replication origins is essential - please quote any paper testing its essentiality. - Please follow standard yeast genotype nomenclature, remove ';' when listing the yeast genotypes (e.g. POL3-AID mec1Δ sml1Δ instead of POL3-AID;mec1Δ;sml1Δ- example from Figure 2-B). - Concentrations of IAA are missing in few places (e.g. legend of Figure 1-C, page 24). - Figure 1A: add the label 'IAA (mM)' - Figure 2G: pleae provide a shorter exposure of the H4 blot in addition to the one shown. - Figure 6: adding a schematic presenting the events at actively and passively replicating late origins (and the predictions about leading and lagging strand bias) would help to understand the figure. - The format of the references is inconsistent. - 'On Watson/Crick strand' should be replaced with 'in Watson/Crick strand' We will fix typos, etc

Reviewer #2 (Significance (Required)):

This is a nice piece of work that provides in vivo confirmation of several observations that have been made in purified recombinant systems. In that sense, the overall novelty is limited, but this type of study is still important to do, as biochemical assays do not always reflect what is happening in cells, and this study gives insight into basic activities of the replisome. The participation of Pol epsilon in lagging strand synthesis is an interesting observation. Overall, the study will be of interest for the DNA replication field. My own expertise is in replication, predominantly in yeast. I have experience in NGS analysis of replication as well as in genetic analysis of the DNA damage response. I therefore feel competent to evaluate all aspects of the manuscript.

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

This manuscript describes the consequences of reducing the cellular concentration of a pol-delta subunit in S. cerevisiae. Pol-delta plays multiple cellular roles at both the replication fork (it is one of two DNA polymerases responsible for lagging strand synthesis) and during repair synthesis after DNA damage. The authors combine genetic and genomic methodologies to characterise how reduction in pol-delta concentration impacts on cellular fitness and specifically lagging strand synthesis. Overall it is technically a well executed study that is clearly presented and the data are predominantly appropriately interpreted.

\*I have a number of major comments:** *

1) The authors apply OK-seq (a methodology first developed by the senior author and therefore they are clearly experts at this) to study the consequence of pol-delta depletion on genome replication. OK-seq requires isolation of Okazaki fragments and this in turn requires removal of DNA ligase (Cdc9) - the authors achieve this with the anchor away system. My concern is that in these experiments the authors are depleting two factors required for lagging strand synthesis: pol-delta and ligase; it is unclear to me how the authors can determine the relative contribution of each depletion to the observed phenotypes. Could some of the observed phenotypes (e.g. fork slowing, 5' and 3' ends of Okazaki fragments, etc) be a consequence of the double depletion, rather than just pol-delta depletion as concluded by the authors? The authors present this as a method to determine genome replication timing, but really it is an assay to look at fork direction. Given the need for an addition mutation in OK-seq (cdc9), I would encourage the authors to consider a more direct assay for replication dynamics upon pol-delta depletion, such as a copy number measure (or BrdU-ip) of DNA replication or DNA combing - these methods don't require Cdc9 depletion and could therefore ensure that observed phenotypes are a consequence of pol-delta depletion (rather than the double depletion).

As outlined in our response to the first two reviewers, we will do BrdU-IP experiments. We agree that the double depletion may have an effect on fork speed, and that BrdU-IP will allow us to test this possibility. However, we note that our analysis of Okazaki fragment initiation/processing requires the depletion of Cdc9, so for this we are limited to looking at differences between Cdc9 depletion alone vs Cdc9 depletion + Pol3 depletion.

2) One major conclusion reached by the authors is that pol-epsilon can contribute to lagging strand synthesis upon pol-delta depletion (at least during late replication). This conclusion comes from the authors use of HydEn-seq to measure rNTP incorporation from which the contribution of a polymerase (pol-epsilon in this case) to strand synthesis can be determined. In a manner analogous to OK-seq, this requires the introduction of additional mutations (both in the polymerase and by the removal of RNaseH activity). The authors interpretation that pol-epsilon can play a role in lagging strand synthesis is dependent upon there being no temporal change in pol-delta strand-displacement activity, despite continued pol-delta depletion through S phase. It is not clear to me that the data presented in Fig 5 & 6 has the sensitivity to conclude this (and the OK-seq data is also subject to the potential bias of the double depletion of pol-delta and Cdc9). I feel that a necessary control to support this conclusion, would be to undertake the HydEn-seq experiment in the absence of the pol-epsilon mutation (just pol-delta depletion in the absence of RNaseH activity). This would allow the authors to measure any increase is residual rNTPs (likely from pol-alpha primase) on the lagging strand as a consequence of pol-delta depletion and determine whether they are equally likely in early and late S phase.

As discussed in our response to the first two reviewers, we will analyze analogous data from both a POL2 wild-type and a pol2-M644L strain that incorporates fewer ribonucleotides than the wild-type.

\*The following comments are more minor:** *

-for the experiment in Fig S1B, the growth in 1.0 mM IAA is somewhat surprising given how sick the cells appear on equivalent plates. I couldn't find in the methods a description of the experimental conditions.

The cells grow very slowly in 1 mM IAA (doubling time doubles). We think this is quite consistent with the poor growth on plates

-there is considerable variability in the S phase kinetics from bulk DNA analysis (flow cytometry) when comparing Fig 1C, 2D, S3. Fig 1C appears to be the exception, with all the other figures showing poor S phase progress by comparison. It would be useful for the authors to recognise these differences and comment upon them. E.g. they appear to all be identical experiments, but are there experimental differences that could explain the different kinetics?

We see some variability in our release, but generally cells enter S-phase at around 30 minutes. The release in figure S3 was carried out at 25 ˚C rather than 30 ˚C, which accounts for the additional delay in these data

-Fig 2F, why is the rad51-deletion less severe that rad52-deletion - should they not be identical?

We agree that these should logically be very similar: we do not know why the two mutants behave slightly differently at some (but not all) IAA concentrations

-Fig 2H - could the authors show the flow cytometry (in a supplemental figure) for this experiment?

We can show this

-Fig 3B-E: OEM is described as a measure of origin efficiency - how is possible for this to have negative values?

OEM describes Okazaki fragment strand bias around previously identified origins. If such an origin does not fire in our strain background, a negative OEM can result.

-pp9: "Analysis of Okazaki fragment strand bias across the genome suggested that the average direction of replication was relatively similar at most loci across all Pol3 depletion conditions". The authors data is quantitative and they should be able to quantify how similar their data are across the various conditions, rather than making a qualitative statement: "relatively similar".

We apologize and can re-phrase this. The intention of this statement is simply to draw the reader’s attention to the fact that global distributions of Okazaki fragments are not completely altered (e.g. only 1-2 origins per chromosome) as a prelude to the more quantitative analysis that follows in figure 3.

-pp9: "origin firing efficiency in S. cerevisiae correlates strongly with replication timing"; it would be useful for the authors to support this statement with a citation.

We will add 1-2 citations to support this statement

-Fig 4A: it would help the reader if the authors could show 'zoomed in' examples of the points that the authors make (in addition to the whole chromosome view): slowed fork progression, reduced early origin activity, increased late origin activity (e.g. an origin that is normally passively replicated, that upon pol-delta depletion is no longer passively replicated and therefore becomes more efficient), etc.

We agree that this would be helpful, and can add examples in the supplement

-pp11: "An analogous global decrease in replication-origin firing efficiency has been observed in Pol ∂-deficient human fibroblasts" - but the authors are reporting a global increase in origin firing efficiency (Fig 3B).

We can re-phrase this.

-the nucleosomal ladder in Fig 5A is only weakly apparent from the gel and not particularly apparent from the density trace, this makes it's disappearance upon IAA treatment hard to interpret. Is the weak nucleosomal ladder what the authors had anticipated (in the absence of IAA)?

We do not expect a weaker nucleosomal ladder than normal in the absence of IAA. In our experience these gels just sometimes give better ladders than others, and we hope that the traces help with interpretation

-I found the effects being described by the authors in Fig 5B & C difficult to see, particularly for the transcription factors. Furthermore, why are these data differently normalised to those in Fig 4B & C (median vs. maximum)

In figure 4 we normalize to maximum since all DNA should eventually be replicated, and we therefore think that showing coverage relative to a maximum value of 1 is most informative. In figure 5 we compare distributions of termini around obstacles, and therefore feel that normalizing to the median is a more appropriate way to compare enrichment around a given meta-element. The shapes of the graphs would be unchanged by choosing a different normalization point. In order to make changes easier to see, we can make the lines thinner in figure 5 and/or change the y-axis scale.

-the final sentence of the results section returns to an analysis of the OK-seq data and is essentially a temporally segregated analysis (Fig S6) otherwise equivalent to that presented in Fig 5B. Given the importance placed on these data by the authors in the interpretation of the HydEn-seq data, I feel that it would help the reader to have been presented with these data earlier in the results section.

We can move these data up

-p22: OK-seq methods. The authors should indicate the culture conditions for these experiments.

We can include this

-p22: Computational analyses: the authors should indicate which reference genome assembly they used.

We can include this

-Fig 6B & C: the y-axis labels are confusing - do the authors mean Crick strand here?

Oops. Yes, we do. We thank the reviewer for catching this

\*REFEREE CROSS COMMENTING:** *

All three reviewers seems to be in broad agreement about this manuscript. There is one significant concern raised by the other reviewers that I'd missed: that some of the Okazaki fragment analysis was done with non-synchronised cultures. I agree with this concern, however I don't think that there is necessarily a problem with the alternative explanation suggested by reviewer #1 ('Okazaki fragments enrichment may be skewed towards late origins'). While the accumulation of S phase cells might well be expected to lead to a bias towards isolating more Okazaki fragments from around late origins, the authors calculate the fraction of reads (i.e. Okazaki fragments) mapping to each strand. The potential presence of more late S phase cells would give greater sequence coverage over late replicating regions, but alone would not alter the fraction of reads mapping to each strand. However, I agree that interpretation of this experiment is not as simple as suggested by the authors and there may well be alternative explanations along the lines suggested by reviewer #1.

There was a subsequent Okazaki fragment experiment performed with synchronised cells (Fig 4) and it might be possible to use these data to assess any differential impact on early vs late origins.

We agree, and will do this analysis

Reviewer #3 (Significance (Required)):

My expertise is in DNA replication and genome stability, particularly replication timing and replication origin function.

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 describes the consequences of reducing the cellular concentration of a pol-delta subunit in S. cerevisiae. Pol-delta plays multiple cellular roles at both the replication fork (it is one of two DNA polymerases responsible for lagging strand synthesis) and during repair synthesis after DNA damage. The authors combine genetic and genomic methodologies to characterise how reduction in pol-delta concentration impacts on cellular fitness and specifically lagging strand synthesis. Overall it is technically a well executed study that is clearly presented and the data are predominantly appropriately interpreted.

I have a number of major comments:

1) The authors apply OK-seq (a methodology first developed by the senior author and therefore they are clearly experts at this) to study the consequence of pol-delta depletion on genome replication. OK-seq requires isolation of Okazaki fragments and this in turn requires removal of DNA ligase (Cdc9) - the authors achieve this with the anchor away system. My concern is that in these experiments the authors are depleting two factors required for lagging strand synthesis: pol-delta and ligase; it is unclear to me how the authors can determine the relative contribution of each depletion to the observed phenotypes. Could some of the observed phenotypes (e.g. fork slowing, 5' and 3' ends of Okazaki fragments, etc) be a consequence of the double depletion, rather than just pol-delta depletion as concluded by the authors? The authors present this as a method to determine genome replication timing, but really it is an assay to look at fork direction. Given the need for an addition mutation in OK-seq (cdc9), I would encourage the authors to consider a more direct assay for replication dynamics upon pol-delta depletion, such as a copy number measure (or BrdU-ip) of DNA replication or DNA combing - these methods don't require Cdc9 depletion and could therefore ensure that observed phenotypes are a consequence of pol-delta depletion (rather than the double depletion).

2) One major conclusion reached by the authors is that pol-epsilon can contribute to lagging strand synthesis upon pol-delta depletion (at least during late replication). This conclusion comes from the authors use of HydEn-seq to measure rNTP incorporation from which the contribution of a polymerase (pol-epsilon in this case) to strand synthesis can be determined. In a manner analogous to OK-seq, this requires the introduction of additional mutations (both in the polymerase and by the removal of RNaseH activity). The authors interpretation that pol-epsilon can play a role in lagging strand synthesis is dependent upon there being no temporal change in pol-delta strand-displacement activity, despite continued pol-delta depletion through S phase. It is not clear to me that the data presented in Fig 5 & 6 has the sensitivity to conclude this (and the OK-seq data is also subject to the potential bias of the double depletion of pol-delta and Cdc9). I feel that a necessary control to support this conclusion, would be to undertake the HydEn-seq experiment in the absence of the pol-epsilon mutation (just pol-delta depletion in the absence of RNaseH activity). This would allow the authors to measure any increase is residual rNTPs (likely from pol-alpha primase) on the lagging strand as a consequence of pol-delta depletion and determine whether they are equally likely in early and late S phase.

The following comments are more minor:

-for the experiment in Fig S1B, the growth in 1.0 mM IAA is somewhat surprising given how sick the cells appear on equivalent plates. I couldn't find in the methods a description of the experimental conditions.

-there is considerable variability in the S phase kinetics from bulk DNA analysis (flow cytometry) when comparing Fig 1C, 2D, S3. Fig 1C appears to be the exception, with all the other figures showing poor S phase progress by comparison. It would be useful for the authors to recognise these differences and comment upon them. E.g. they appear to all be identical experiments, but are there experimental differences that could explain the different kinetics?

-Fig 2F, why is the rad51-deletion less severe that rad52-deletion - should they not be identical?

-Fig 2H - could the authors show the flow cytometry (in a supplemental figure) for this experiment?

-Fig 3B-E: OEM is described as a measure of origin efficiency - how is possible for this to have negative values?

-pp9: "Analysis of Okazaki fragment strand bias across the genome suggested that the average direction of replication was relatively similar at most loci across all Pol3 depletion conditions". The authors data is quantitative and they should be able to quantify how similar their data are across the various conditions, rather than making a qualitative statement: "relatively similar".

-pp9: "origin firing efficiency in S. cerevisiae correlates strongly with replication timing"; it would be useful for the authors to support this statement with a citation.

-Fig 4A: it would help the reader if the authors could show 'zoomed in' examples of the points that the authors make (in addition to the whole chromosome view): slowed fork progression, reduced early origin activity, increased late origin activity (e.g. an origin that is normally passively replicated, that upon pol-delta depletion is no longer passively replicated and therefore becomes more efficient), etc.

-pp11: "An analogous global decrease in replication-origin firing efficiency has been observed in Pol ∂-deficient human fibroblasts" - but the authors are reporting a global increase in origin firing efficiency (Fig 3B).

-the nucleosomal ladder in Fig 5A is only weakly apparent from the gel and not particularly apparent from the density trace, this makes it's disappearance upon IAA treatment hard to interpret. Is the weak nucleosomal ladder what the authors had anticipated (in the absence of IAA)?

-I found the effects being described by the authors in Fig 5B & C difficult to see, particularly for the transcription factors. Furthermore, why are these data differently normalised to those in Fig 4B & C (median vs. maximum)

-the final sentence of the results section returns to an analysis of the OK-seq data and is essentially a temporally segregated analysis (Fig S6) otherwise equivalent to that presented in Fig 5B. Given the importance placed on these data by the authors in the interpretation of the HydEn-seq data, I feel that it would help the reader to have been presented with these data earlier in the results section.

-p22: OK-seq methods. The authors should indicate the culture conditions for these experiments.

-p22: Computational analyses: the authors should indicate which reference genome assembly they used.

-Fig 6B & C: the y-axis labels are confusing - do the authors mean Crick strand here?

REFEREE CROSS COMMENTING:

All three reviewers seems to be in broad agreement about this manuscript. There is one significant concern raised by the other reviewers that I'd missed: that some of the Okazaki fragment analysis was done with non-synchronised cultures. I agree with this concern, however I don't think that there is necessarily a problem with the alternative explanation suggested by reviewer #1 ('Okazaki fragments enrichment may be skewed towards late origins'). While the accumulation of S phase cells might well be expected to lead to a bias towards isolating more Okazaki fragments from around late origins, the authors calculate the fraction of reads (i.e. Okazaki fragments) mapping to each strand. The potential presence of more late S phase cells would give greater sequence coverage over late replicating regions, but alone would not alter the fraction of reads mapping to each strand. However, I agree that interpretation of this experiment is not as simple as suggested by the authors and there may well be alternative explanations along the lines suggested by reviewer #1.

There was a subsequent Okazaki fragment experiment performed with synchronised cells (Fig 4) and it might be possible to use these data to assess any differential impact on early vs late origins.

Significance

My expertise is in DNA replication and genome stability, particularly replication timing and replication origin function.

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 the authors explore the basis of the deleterious effects of reduced levels of the replicative Polymerase delta. This polymerase plays several important roles in the replication process: it synthesizes most of the lagging strand, but also extends the first primer during the synthesis of the lagging strand, and it contributes to the removal of the RNA and most of the DNA synthesized by primase/Polymerase alpha during Okazaki fragment maturation. In this study, the authors systematically analyze the impact of Pol delta depletion in S. cerevisiae. They use a degron-tagged allele to modulate the levels of the polymeraes and apply mainly NGS methods and classical genetics to explore the consequences for survival, checkpoint signalling and replication features such as fork speed, origin firing efficiency and Okazaki fragment length and distribution. They report that Pol delta depletion leads to a checkpoint activation via the Rad9-dependent damage signalling pathway (but not the Mrc1-dependent replisome-associated signalling) and an accumulation of single-stranded DNA. Phosphorylation of histone H2A is taken as a marker of DNA double-strand breaks, and from the observation that deletion of recombination factors, but not end-joining factors aggravate the fitness of Pol delta-depleted cells they conclude that homologous recombination is responsible for the repair of these breaks. Analysis of replication by Okazaki fragment sequencing indicates a slight decrease in the firing efficiency of early/efficient origins, but an increase in the firing efficiency of late origins. They also observe a reduction in fork speed, although they are not able to attribute this to either a globally slower fork movement or an increase in the stalling of individual forks. They find that Pol delta depletion does not change the size of Okazaki fragments, but causes defects in the nick translation during Okazaki fragment maturation. Finally, they use NGS technology to show that the leading strand Polymerase epsilon engages in lagging strand replication particularly at late origins when Pol delta is depleted. From their observations, the authors develop a model where depletion of Pol delta primarily affects late replicating regions. They explain this by invoking a stable association of Pol delta with early replisomes, which sequesters the enzyme, thus causing an under-supply at replisomes that assemble later during S phase. This then leads to the involvement of Pol epsilon on the lagging strand. Based on the observation that fork speed and Okazaki fragment maturation are both affected, they propose that these two reactions normally compete for Pol delta, suggesting that optimal replication would require two molecules of the polymerase per fork.

Major comments

The experiments shown here are largely clean and well controlled, and the manuscript is written nicely and well-structured.

Compared to the Okazaki fragment analysis, the treatment of double-strand breaks appears somewhat cursory and remains inconclusive. Phosphorylation of H2A seems insufficient evidence for double-strand breaks, as other structures could also give rise to that signal. These lesions should be detected in a more direct manner, e.g. pulsed-field gel electrophoresis. The authors also don't provide a mechanism by which such breaks would emerge. Related to the minor effect of the ku mutant, I am wondering about the altered appearance of the colonies in Figure 2F (concerning both ku70 and rad51) - what is different about these, and could their „denser" appearance explain the slight suppression effect observed?

Concerning the damage signalling: it is surprising to see a damage signal at concentrations of IAA that do not lead to a significant depletion of Pol delta yet (0.05 mM). At this point, it is hard to imagine DSBs to form. Could the authors explain this discrepancy?

The notion that late origin firing is enhanced despite checkpoint activation is counterintuitive. Do the authors think that this effect overcomes the suppression of late origins that is normally associated with checkpoint activation? It would be helpful to test whether abolishing that phenomenon (e.g. by a mec1-100 mutant) would enhance the effect and render late origins even more active.

It would be important to characterise the fork speed defect better, using alternative methods rather than just relying on Okazaki fragments. A differentiation between slower fork progression and more frequent fork stalling would be relevant and might help to evaluate the contribution of Pol epsilon. This might be accomplished by DNA fibre analysis. Alternatively, BrdU incorporation could serve to observe replication over the entire genome rather than only in the vicinity of replication origins. It would also be important to differentiate fork speeds in early versus late replicating regions - according to the authors' model, the defects should be most obvious in the late regions (Fig. 4 concerns only early origins).

Figure S3: Considering the differences in cell cycle progression, it would make more sense to compare equivalent stages of the cell cycle / S phase rather than identical time points.

Considering that the Okazaki fragment analysis was done with non-synchronised cultures, is it possible that the skew in the cell cycle profile could influence the Okazaki fragment pattern?

Would it be possible to monitor not only total Pol delta levels, but also the level of Pol delta bound to the chromatin? It is shown that the level of Pol delta is depleted in the whole cell extracts, but it would be interesting to see how severe the depletion is on the chromatin.

Figure 6 is confusing and should be clarified:

• Figure 6B: assigning the Watson and Crick strands in the schematic would make that figure easier to understand;
• Figure 6B-C: the axes are labeled as 'Fraction of rNMP on Watson strand', but would it not make more sense if they were labeled 'Fraction of rNMP in Crick strand'?
• Figure 6D-E: the right side scale is labelled as 'increase in rNMP on Crick strand' while in the figure legend is says it is 'change in the fraction of ribonucleotides mapping to the Watson strand. That description should be clarified;
• Figure 6D: using 'Change in Okazaki fragments strand bias' to label the black line (description in the box above the figure) instead 'Change in Okazaki strand bias' would be more appropriate;
• Figure 6F: the authors seem to have subtracted strand bias measured for Okazaki fragments from the strand bias measured for rNMP. It is valid to subtract these biases from each other, considering that the two structures arise independently and with different frequencies?

Minor comments:

Can the authors conclude that Pol delta deficiency/ incompleteness of lagging strand synthesis affects the nucleosome deposition onto DNA? (Figure 5-A)

Why did the authors use rnh202Δ and not a mutant in the catalytic subunit of RNase H2?

An extra control might be useful: comparing POL3-AID rnh202Δ with the POL3-AID pol2M644G rnh202Δ triple mutant could further confirm that the observed effect is Pol epsilon-dependent.

Figure 2H: It would be good to see the cell cycle distribution corresponding to the western blot images.

Various spelling, grammar or precision of expression issues:

• Pg. 4, line 4: endonucleolytically instead of nucleolytically.
• Pg. 6, line 10: Remove 'was'.
• Pg. 6, line 12: Remove 'in vivo' from the subtitle.
• Pg. 6, line 14: 'an C-terminal' should be 'a C-terminal'
• Pg 16, line 13: Phrasing implies that the synthesis of both leading and lagging strands by Pol delta in regions in the vicinity of replication origins is essential - please quote any paper testing its essentiality.
• Please follow standard yeast genotype nomenclature, remove ';' when listing the yeast genotypes (e.g. POL3-AID mec1Δ sml1Δ instead of POL3-AID;mec1Δ;sml1Δ- example from Figure 2-B).
• Concentrations of IAA are missing in few places (e.g. legend of Figure 1-C, page 24).
• Figure 1A: add the label 'IAA (mM)'
• Figure 2G: pleae provide a shorter exposure of the H4 blot in addition to the one shown.
• Figure 6: adding a schematic presenting the events at actively and passively replicating late origins (and the predictions about leading and lagging strand bias) would help to understand the figure.
• The format of the references is inconsistent.
• 'On Watson/Crick strand' should be replaced with 'in Watson/Crick strand'

Significance

This is a nice piece of work that provides in vivo confirmation of several observations that have been made in purified recombinant systems. In that sense, the overall novelty is limited, but this type of study is still important to do, as biochemical assays do not always reflect what is happening in cells, and this study gives insight into basic activities of the replisome. The participation of Pol epsilon in lagging strand synthesis is an interesting observation. Overall, the study will be of interest for the DNA replication field. My own expertise is in replication, predominantly in yeast. I have experience in NGS analysis of replication as well as in genetic analysis of the DNA damage response. I therefore feel competent to evaluate all aspects of the manuscript.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This study examined the consequences of limiting levels of DNA polymerase d (Pol d) in yeast. The authors first reported multiple genome instability consequences following lowered Pol d level, including defect in S phase progression, growth defect, elevated spontaneous DNA damage, accumulation of ssDNA and activation of replication and DNA damage checkpoint. These observations are solid but not unexpected. By genome wide analysis using the Okazaki fragment (OF) mapping and ribonucleotide mapping (for polymerase usage), the authors claim a few potentially novel and striking observations that lowered Pol d differentially impact efficiencies of early vs. late origins, and that lowered Pol d results in Pol e participating in lagging strand synthesis around late origins. However, I remained unconvinced based on the data presented. These observations need to be further substantiated and alternative interpretations should be considered.

Main concerns:

One of major conclusions the authors tried to make is that the early vs. late origins are differentially affected by low level of Pol d. First, they used OF mapping data to examine origin efficiency. Asynchronous "Cultures were treated with IAA for 2h before the addition of rapamycin for 1h to deplete DNA ligase I (Cdc9) from the nucleus via anchor-away". IAA concentrations used were of 0, 0.2 mM, 0.6 mM, and 1 mM. The problem is that Figure S1 clearly showed that treating asynchronous cultures with >0.1 mM of IAA for as short as 30 min significantly alters the cell cycle profiles, mainly resulting in accumulation of S phase cells, to different extent. Presumably Okazaki fragments accumulated from these cultures suffering from the synchronizing effect may not be representative of the real change in global replication profile. For instance, it is not difficult to predict that the Okazaki fragments enrichment may be skewed towards late origins if more cells are accumulated in mid S phase following Pol d depletion. For this reason, I don't believe the result is conclusive. The experiment may be re-designed for samples at different time points following release from G1.

This concern also should preclude the authors from drawing conclusion about Pol e usage on lagging strands based on comparison between HydEn-seq data and OF mapping data shown in Figure 6. In fact, the rNMP incorporation change is very similar between early and late origins. The only evidence that the author rely on is the discrepancy in OF data between the two groups origins, which makes the reliability if origin efficiency measurement the central piece of data in this study. Thus, alternative approaches should also be considered to map origin efficiencies.

Even if Pol e strand bias is lowered at late origins, as the authors tend to believe, there are still alternative models other than Pol e being used for lagging strand synthesis. For instance, if TLS polymerases are used on lagging strands, it could result in more ribonucleotide incorporation on the lagging strand, as they are lower-fidelity polymerases. Alternatively, if Pol d scarcity leads to more Pol e synthesis or lower RNA primer processing, it might also contribute to more apparent ribonucleotide incorporation on the lagging strands.

In Figure S5, the two HydEn-seq replicates are very different, where replicate1 shows very low strand bias. I suspect perhaps the strain used for replicate 1 does not contain pol2-M644G or rnh202 deletion.

Significance

Given that different aspects of Pol d deficiency have been implicated in various human diseases and cancer, this type of analysis is of interest to the field.

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 article, the authors characterize a complex formed by sec22b-stx1-Esyt2. They investigate how such interactions are involved in the modulation of dynamics of the plasma membrane in the context of neuritogenesis. They conclude that the contact sites between the ER and the plasma membrane, mediated by the afore-mentioned complex, contribute the expansion of the plasma membrane.

**Major comments:**

Overall, the article clearly shows that in mammalian cells there is an interaction between sec22b-stx1-Esyt2 which appears to be important for filopodia formation and possibly neuritogenesis in neurons. However, performing additional experiments to better clarify some links and assumptions made by authors could strengthen the article.

The manuscript relies on work performed either on cell lines (HeLA, PC12) or primary neuronal cultures. Although it is clear the value of the findings obtained using cell lines, they should be seen as a complementary rather than an exclusive approach. This is particularly important as the authors often make claim about neuron-related cellular biology.

For instance, the biochemistry-based findings on the interaction and characterization of the protein complex (Figure 1) are all derived from experiments perfomed in Hela or PC12. As the authors have the capacity to culture and manipulate primary neuronal cultures, such findings should be validates in neuronal cells. The authors could also consider performing biochemical experiments (i.e. co-ip) of the endogenous proteins in neuronal cultures or brain tissue.*

->Endogenous Co-IP has been tried in E18 brain tissue. One experiment using brain tissue demonstrated co-immunoprecipitation of endogenous Sec22b and E-Syt2. Unfortunately, repetitions of this experiment failed due to high background in negative control (naïve Rabbit IgG). We agree with the reviewer that this data is worth trying again. We will carry out this co-immunoprecipitation experiment from cultured neurons to answer the reviewer’s request.

The authors do show some evidence regarding the complex in neuronal cells using PLA (proximity ligation assay, figure 2) or super resolution microscopy, however, these findings should be corroborated by stronger findings targeting interaction and not based on simple proximity.

->We agree with this reviewer that PLA is limited in demonstrating the occurrence of a protein complex. We would like to stress that we have used PLA complementarily to immunoprecipitation and that we already have shown STED super-resolution data (Figure 3). In order to strengthen the STED data, we will include more details in the figure, as a supplementary movie and a supplementary spreadsheet with the quantification of the distance between the E-Syt2/Sec22b clusters to the plasma membrane stained using WGA. The STED data demonstrate that 50% of the clusters are closer than 33.6nm to the plasma membrane, a distance in the range of ER-PM contact sites.

A similar critique regards the experiments using RNA-interference of Figure 4. Performing loss-of-function experiments in neuronal cultures would strengthen and complement the results obtained via over-expression approaches shown in Figure 5.

->The loss-of-function experiments in neuronal cultures using siRNA were attempted unsuccessfully. The three E-Syts have largely different cDNA sequences thus three distinct siRNAs must be transfected in order to silence all three simultaneously. This is quite challenging in neuronal cultures and we were never able to get strong silencing of the three E-Syts. In the following points, we plan to carry out further experiments using expression of a fragment of Sec22b (Longin domain). We are confident that this is a better approach to demonstrate the importance of Sec22b/E-Syt interaction.

---

*Given that the authors have already in place all the necessary technology for the suggested biochemical and morphological-related experiments, these could be carried out swiftly within 3-4 months.

**Minor comments:**

The manuscript is really technical and at times tough to follow; it could benefit from key sentences to better guide the reader, particularly if not coming from the specialist field, in the appreciation of the experiments and results.

Authors should submit the manuscript to a severe round of proofreading. There are several inconsistencies and sometimes what looks like internal comments: i.e. in the methods "STED Missing" or the fact that "LTP" is used everywhere but not defined and considering that the targeted audience is most likely neuroscience-based could easily lead to confusion.

*

->We fully agree with this reviewer and apologize for leaving behind such errors. We will carefully proofread the revised ms.

*The experiments appear to have been repeated a sufficient number of times and the statistics seem adequate. It would be advisable to show in dot-plots the findings rather than in bar graphs all findings and not just the morphometrics-relative ones.

*

->We will modify the figures according to this reviewer’s suggestion.

Reviewer #1 (Significance (Required)):

*This work closely follows the excellent previous work from the Galli laboratory. As such, it is mostly incremental from a technical perspective and does not present particularly novel findings. An interesting aspect would be in addressing directly the influence of the described interactions in the lipid transfer between ER and the plasma membrane but in that sense the manuscript falls short. Although it is to be appreciated the functional readouts in terms of neuritogenesis, in the present state the manuscript features findings suitable for a very specific audience.

I believe that the appropriate audience for the present manuscript lies within the neuroscientific community interested in development, specifically neuritogenesis, and/or membrane dynamics. Additionally, it might be interesting also for researchers outside of the neuroscience community and interested in the dynamics between ER and plasma membrane.

*

->We are happy to read the comments of this reviewer. Nevertheless, we would like to stress the importance of deciphering precise molecular mechanisms in any biological process. Here, we are the first to demonstrate an interaction between lipid-transfer proteins E-Syts and ER v-SNARE Sec22b. As an example, the molecular mechanism connecting synaptic SNAREs and synaptotagmin has been the topic of more than 500 publications since seminal articles in the early 1990’s. We think that the first article linking E-Syts to SNAREs cannot be considered as a mere increment from our previous work.

The activity of E-Syts to transfer lipids in vitro has been well established __(1–3) In addition, recent work by the De Camilli lab using Origami showed that reducing the distance between liposomes enhanced the lipid transfer mediated by E-Syt2 (3). Therefore, we did not carry out experiments such as combining SNAREs and E-Syt2 in artificial membranes in vitro because we considered that there would not be much more to demonstrate than what has already been done. Furthermore, we considered the experiments in cells, particularly neurons, much more critical at this point. Demonstrating transfer of glycerophospholipid between ER and PM in cells cannot be performed like other lipids’ transfer at other membrane interfaces for the following reasons: phospholipids are very abundant (4) and they are not modified upon transfer (1)__, there are no specific dyes to detect glycerophospholipids (unlike phosphoinositides), and ER and PM are too close to distinguish if a glycerophospholipid is in one or the other membrane. Such a challenging experiment would require the ability to setup a specific biochemical assay circumventing these constraints. We think that this is out of the scope of the present study focused on the role of E-Syt/Sec22b complex.

Nevertheless, in order to get further insights on this question, we will express WT and mutant E-Syt2, purify the PM using the protocol of Figure 4 in Saheki et al __(1)__, followed by lipidomics analysis. We hope that this approach further supports our idea that E-Syts mediate an important lipid transfer mechanism towards the PM.

* Keywords regarding my expertise: Molecular and Cellular Neuroscience, Morphometrics, Dendrite, Neurons, Dendritogenesis, Biochemistry, Imaging, Microscopy.

---

__

Reviewer #2 (Evidence, reproducibility and clarity (Required)): *This manuscript identifies and characterises a novel interaction between E-Syts and Sec22b and demonstrates that lipid transfer between the ER and PM contributes to the development of filopodia and neuronal expansion. This interaction with E-Syt2 occurs through the Longin domain of Sec22b Sec22b association. The authors suggest a continuum with further interactions with syntaxin1, that mediates neurite outgrowth. Overall I find this study very interesting and convincing. The experimental analysis is well carried out and the claims are well aligned with their results.

I only have minor issues:

Figure 1. Some of the western blots have several bands and it is difficult to know which band is the relevant one. They should be indicated in the fig panel. Further panel E and F are barely readable and should be redrawn with the appropriate line and font size.*

->We will make the changes requested by this reviewer in Figure 1.

• *

Figure 2: is there a difference between the number of dots in axons and dendrites? Can the author elaborate on this aspect as it is not clear from the image presented.

->We could not combine PLA with further staining of MAP2 and TAU. Indeed, to perform PLA, neurons are already double labelled to detect the proteins of interest. At the stage of the neurons used in this study, both axons and dendrites are growing. Therefore, we did not invest in distinguishing between axons and dendrites. Because growth cones are known to be the major sites of membrane growth, we instead distinguished dots within neurites and in growth cones. We will make the other changes requested by this reviewer in Figure 2.

Figure 7: statistical analysis should be indicated from the BoNT/A and BoNT/C as BoNT/A represent an appropriate control cleaving SNAP25 but not Syntaxin.

->We agree with this request and we will add statistical analysis as suggested, using BoNT/A as an additional control.

On top of controlling fusion and neuronal outgrowth, syntaxin has a role in survival and its cleavage leads to neuronal death. Is this pathway mediated by E-Syts interactions?

->We have stated in the ms: “Since exposure to BoNT/C1 at high concentrations and for long incubation periods causes degeneration of neurons in culture __(5,6)__, various concentrations and incubation times were tested, and a 4-hour treatment of neurons with 1nM BoNTs was chosen to avoid such deleterious effects.” Accordingly, we did not see any degeneration in our experimental conditions.__ __

---

Reviewer #2 (Significance (Required)): This papers identifies the molecular mechanism of neuronal outgrowth. It is highly significant. ->We are very grateful to this reviewer for pointing out the high significance of our article.

---

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

__*1. The evidence for the claim that the Sec22b/Stx1 complex and E-Syts colocalize in native cells (neurons) and bind in heterologous cells is strong (3 independent lines of evidence: co-immunoprecipitation, Proximity Ligation (PLA) and STED super-resolution microscopy) However, the current title of the paper makes a claim beyond this interaction/proximity, based on evidence that is obtained with E-Syt over-expression in wildtype cells. The physiological relevance of the effects remain elusive with over-expression in wildtype cells only.

Furthermore, it is plausible that overexpression of membrane binding/bending C2-domains promotes neurite outgrowth and ramifications by a non-specific effect (as shown for copine C2 domains, PMID:25450385 and indirect evidence for synaptotagmins1,2,7).*

* This issue is especially relevant in the light of the fact that loss of all 3 Extended Synaptotagmins does not affect normal mouse development and viability (PMID: 27399837)

It would be more appropriate to choose a more descriptive title*

• *

->We agree with this reviewer that the original title may be too strong and are now proposing the following, more descriptive title:

Role of the Sec22b/E-Syt complex in neurite growth and ramification

We are fully aware that proteins harbouring C2 domains could potentially promote non-specific effects when overexpressed. However, we do not think this is the case here because none of the morphological effects of E-Syt2 expression in neurons and HeLa cells were reproduced by mutants lacking the SMP or the membrane-anchoring domains. Based on work on Copine __(7)__, a cytosolic protein, E-Syt2 still containing 3xC2 domains but lacking the membrane-anchoring domain should have shown a morphological effect if non-specific binding to phosphoinositides was the mechanism of action. We will discuss this point in the ms.

---

• The evidence for the working model that the morphological effects of E-Syt2 depends on the Sec22b/Stx1 complex is not strong. Although plausible, the positive effect on neurite outgrowth (E-Syt2 overexpression) and the negative effects (inhibition by Stx1 cleaveage, Sec22b-Longin or Sec22b extended linker expression) may in fact be independent

  The evidence could be strengthened by PLA measurements in neurons over-expressing Myc-E-Syt2 and Sec22b to assess MSC density. It is predicted that in both conditions, MCS density increases. MCS density by PLA measurements could also be performed in Sec22b-P33 and DLongin overexpressed and BoNT/C1 treated neurons. According to the model, the number of MCS should go down. This is of special interest for BoNT/C1 treatment, as it is important to show that the altered morphology is not purely caused by a pre-state of degeneration that is known to be induced by BoNT/C1. In addition, EM measurements of ER-PM distances might provide an independent line of evidence.*

->We agree with this reviewer that additional experiments could strengthen the description of the molecular mechanism. To this end, we will carry out the following experiments:

1/Co-immunoprecipitation experiments of endogenous Syntaxin, Sec22b and E-Syt2 in cells expressing GFP as control or Longin-GFP to demonstrate that expression of the Longin domain perturbs the association of Sec22b with E-Syt2 and Syntaxin.

2/PLA measuring the association between E-Syt2 and Syntaxin in cells expressing GFP as control or Longin-GFP to demonstrate that expression of the Longin domain perturbs the association between E-Syt2 and Syntaxin using a complementary approach.

---

Unfortunately, membrane-associated, BoNTC1-cleaved syntaxin corresponds to a short fragment undetectable by available antibodies whereas the fragment detected by the antibody after BoNTC1 cleavage lacks the transmembrane domain (Figure 7a). Therefore, we cannot perform PLA in BoNTC1-treated neurons.

---

We are confident that further exploring the mechanism of action of the Longin domain, together with the data already in the ms, will make it very clear that the morphological effects of E-Syt2 depends on the Sec22b/Stx1 complex.

---

---

1. • Link between neurite outgrowth and lipid transfer is weak. The authors argue that functional E-syt/Sec22b/Stx interaction is important for neurite outgrowth by mediating lipid transfer. The only line of evidence they provide is the absence of outgrowth effects in E-syt mutants lacking SMP or membrane spanning domains. However, from the data it is unclear whether these mutants are correctly folded, expressed and/or localized. Additional ICC stainings of the mutants in neurons are necessary to drive this point home. *

• *

->The mutants and siRNA have been already used and validated in Giordano et al. 2013 __(8)__, therefore we did not carry out experiments aiming at basic characterization of these reagents. To answer this request, we will show images of the subcellular localization by ICC of WT and mutant E-Syt2 in the revised Figure 6 or in a Supplementary Figure.

---

In addition, the authors might make the link between neurite outgrowth and lipid transfer stronger by examining PM lipid levels and distribution in control, Myc-E-Syt2 and E-Syt2 mutant neurons.

->We agree with this reviewer that this question is of high relevance. In order to answer this request, we will express WT and mutant E-Syt2, purify the PM using the protocol of Figure 4 in Saheki et al __(1)__, followed by lipidomics analysis. We hope that this approach further supports our idea that E-Syts mediate an important lipid transfer mechanism towards the PM.

1. • There is no clear evidence that E-syt first binds to Sec22b, after which Stx1 leaves SNAP25 and joins the interaction. This should be indicated as speculation.

   * ->We will make it clear that our model in Figure 9 is a hypothetical model.

• An apparent discrepancy exists between the TKD E-syts effects (i.e. reduced MSC density, Fig 4) and the lack of neurite outgrowth defects in TKO E-syts. According to the proposed model, the levels of E-syt correlate with the number of MSCs and thereby neurite outgrowth. Furthermore, to knock down E-Syts, single siRNAs against the three E-syts were used in Fig4. Off-target effects are not controlled in this approach. Using multiple siRNAs and/or siRNA resistant rescues would be required for robust conclusions.

  *

->The mutants and siRNA have already been used and validated in Giordano et al. 2013 __(8)__, therefore we did not carry out experiments aiming at basic characterization of these reagents. In addition, we would like to stress the complexity of carrying out a rescue experiment of a triple KD of proteins.

Statistical analysis is incomplete. It is not clear whether statistical assumptions (e.g. normal distribution) were checked before performing the tests, and whether non-parametric alternatives where used if assumptions were not met.

---

->We thank this reviewer for making this important alert. We would like to stress that we have always checked whether samples followed the normal distribution and made non-parametric tests__. We will include this comment in the methods.__

In Fig4, a T-test is used between multiple groups. This test can only be used when comparing two groups. Number of (independent) measurements is not clear in Fig1, 2, 4

->In all the figure legends the number of repetitions is specified

---

All figures: display all individual data points in all bar graphs (as shown in 5c)

*

*

\*Minor comments:**

1. Inconsistencies on distances in model. Syts are enlongated proteins and thought to be found in MSCs of ~20 nm (Fernandez-Busnadiego, 2015). Trans-SNARE complexes start to interact when the distance between membranes is ~8 nm (Liu, 2007). In the introduction, the authors suggest that incomplete zippering might occur between Stx and Sec22b, resulting in a distance between 10 and 20 nm, which would allow E-Syt localization. In the discussion, however, the authors suggest a model where Sec22b/Stx interaction is important to bring the membranes in ~10 nm distance to enhance LTP activity. Proof for either model is lacking. Please clarify.*

Fig1A: Please clarify the multiple bands? for Stx3 (anti-eGFP).

• *

->These additional bands are recognised by the anti-GFP antibody, the tag being N-terminal, thus they represent proteolytic fragments. We consistently observe these in our experiments.

Fig2: There is no size marker for panels C1-C6

• *

->We will make the appropriate correction.

Fig3: Both proteins seem to show a diffuse pattern. Please specify the validity of measuring average distance. A higher magnification zoom of staining pattern in the growth cone and visualization of the calculation could benefit interpretation.

• *

->We agree with this reviewer that Figure 3 was not optimal to show all the extent of our STED data. In order to strengthen this part, we will include more details in both the figure and as a supplementary movie and supplementary spreadsheet with the quantification of the distance between the E-Syt2/Sec22b clusters to the plasma membrane stained using WGA. The STED data demonstrate that 50% of the clusters are closer than 33.6nm to the plasma membrane, a distance in the range of ER-PM contact sites.

• E-Syt2 and E-Syt3 are used interchangeably throughout the manuscript and E-Syt1 is left out completely. It would help the reader if the authors could elaborate on their interpretation on the similarities and differences in structure and functionality between the three E-Syts.

1. Why is there a red line in Fig 7b?*

__->We added the red line to highlight the shift of SNAP25 band in BoNTA samples. If misleading, it can be removed

Reviewer #3 (Significance (Required)):__

A growing body of literature recognizes the importance of close proximities between membranes, facilitating direct interaction between organelles (Scorrano et al., 2019). Membrane Contact Sites (MCSs) are shown to be important for a wide range of cellular functions, such as lipid and calcium transfer. E-Syts have been recognized as one of the key players in neuronal MCSs, mediating lipid transport (Fernández-Busnadiego et al., 2015). A study published in 2014 by the authors of the current study revealed another two proteins important for MSCs in neurons (Petkovic et al., 2014). ER protein Sec22b and PM SNARE Syntaxin1 were shown to form a non-fusogenic trans-SNARE complex, important for lipid-transfer mediated neurite outgrowth. Gallo and colleagues have now provided important new evidence that these two components (E-Syts and Stx1/Sec22b) are together and may work together at MSCs.

->We thank this reviewer for stressing the importance of our article and agree with the conclusion of __Fernández-Busnadiego et al. (9) on E-Syts being one of the key players in neuronal MCSs, mediating lipid transport. We think that our work is a further key piece of evidence in the demonstration of the importance of E-Syts in neuronal development.__

Bibliography

Saheki Y, Bian X, Schauder CM, Sawaki Y, Surma MA, Klose C, et al. Control of plasma membrane lipid homeostasis by the extended synaptotagmins. Nat Cell Biol. 2016 Apr 11;18(5):504–515. Yu H, Liu Y, Gulbranson DR, Paine A, Rathore SS, Shen J. Extended synaptotagmins are Ca2+-dependent lipid transfer proteins at membrane contact sites. Proc Natl Acad Sci USA. 2016 Apr 19;113(16):4362–4367. Bian X, Zhang Z, Xiong Q, De Camilli P, Lin C. A programmable DNA-origami platform for studying lipid transfer between bilayers. Nat Chem Biol. 2019 Jul 18;15(8):830–837. Alberts B, Johnson A, Lewis J, Raff M. The lipid bilayer. Molecular Biology of …. 2002; Osen-Sand A, Staple JK, Naldi E, Schiavo G, Rossetto O, Petitpierre S, et al. Common and distinct fusion proteins in axonal growth and transmitter release. J Comp Neurol. 1996 Apr 1;367(2):222–234. Igarashi M, Kozaki S, Terakawa S, Kawano S, Ide C, Komiya Y. Growth cone collapse and inhibition of neurite growth by Botulinum neurotoxin C1: a t-SNARE is involved in axonal growth. J Cell Biol. 1996 Jul;134(1):205–215. Park N, Yoo JC, Lee Y-S, Choi HY, Hong S-G, Hwang EM, et al. Copine1 C2 domains have a critical calcium-independent role in the neuronal differentiation of hippocampal progenitor HiB5 cells. Biochem Biophys Res Commun. 2014 Nov 7;454(1):228–233. Giordano F, Saheki Y, Idevall-Hagren O, Colombo SF. PI (4, 5) P2-dependent and Ca2+-regulated ER-PM interactions mediated by the extended synaptotagmins. Cell. 2013; Fernández-Busnadiego R, Saheki Y, De Camilli P. Three-dimensional architecture of extended synaptotagmin-mediated endoplasmic reticulum-plasma membrane contact sites. Proc Natl Acad Sci USA. 2015 Apr 21;112(16):E2004–13.

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

1. The evidence for the claim that the Sec22b/Stx1 complex and E-Syts colocalize in native cells (neurons) and bind in heterologous cells is strong (3 independent lines of evidence: co-immunoprecipitation, Proximity Ligation (PLA) and STED super-resolution microscopy) However, the current title of the paper makes a claim beyond this interaction/proximity, based on evidence that is obtained with E-Syt over-expression in wildtype cells. The physiological relevance of the effects remain elusive with over-expression in wildtype cells only.

Furthermore, it is plausible that overexpression of membrane binding/bending C2-domains promotes neurite outgrowth and ramifications by a non-specific effect (as shown for copine C2 domains, PMID:25450385 and indirect evidence for synaptotagmins1,2,7).

This issue is especially relevant in the light of the fact that loss of all 3 Extended Synaptotagmins does not affect normal mouse development and viability (PMID: 27399837)

It would be more appropriate to choose a more descriptive title

1. The evidence for the working model that the morphological effects of E-Syt2 depends on the Sec22b/Stx1 complex is not strong. Although plausible, the positive effect on neurite outgrowth (E-Syt2 overexpression) and the negative effects (inhibition by Stx1 cleaveage, Sec22b-Longin or Sec22b extended linker expression) may in fact be independent

The evidence could be strengthened by PLA measurements in neuronsover-expressing Myc-E-Syt2 and Sec22b to assess MSC density. It is predicted that in both conditions, MCS density increases. MCS density by PLA measurements could also be performed in Sec22b-P33 and Longin overexpressed and BoNT/C1 treated neurons. According to the model, the number of MCS should go down. This is of special interest for BoNT/C1 treatment, as it is important to show that the altered morphology is not purely caused by a pre-state of degeneration that is known to be induced by BoNT/C1. In addition, EM measurements of ER-PM distances might provide an independent line of evidence.

a. Link between neurite outgrowth and lipid transfer is weak. The authors argue that functional E-syt/Sec22b/Stx interaction is important for neurite outgrowth by mediating lipid transfer. The only line of evidence they provide is the absence of outgrowth effects in E-syt mutants lacking SMP or membrane spanning domains. However, from the data it is unclear whether these mutants are correctly folded, expressed and/or localized. Additional ICC stainings of the mutants in neurons are necessary to drive this point home. In addition, the authors might make the link between neurite outgrowth and lipid transfer stronger by examining PM lipid levels and distribution in control, Myc-E-Syt2 and E-Syt2 mutant neurons.

b. There is no clear evidence that E-syt first binds to Sec22b, after which Stx1 leaves SNAP25 and joins the interaction. This should be indicated as speculation.

c. An apparent discrepancy exists between the TKD E-syts effects (i.e. reduced MSC density, Fig 4) and the lack of neurite outgrowth defects in TKO E-syts. According to the proposed model, the levels of E-syt correlate with the number of MSCs and thereby neurite outgrowth. Furthermore, to knock down E-Syts, single siRNAs against the three E-syts were used in Fig4. Off-target effects are not controlled in this approach. Using multiple siRNAs and/or siRNA resistant rescues would be required for robust conclusions.

Statistical analysis is incomplete. It is not clear whether statistical assumptions (e.g. normal distribution) were checked before performing the tests, and whether non-parametric alternatives where used if assumptions were not met. In Fig4, a T-test is used between multiple groups. This test can only be used when comparing two groups. Number of (independent) measurements is not clear in Fig1, 2, 4. All figures: display all individual data points in all bar graphs (as shown in 5c)

Minor comments:

1. Inconsistencies on distances in model. Syts are enlongated proteins and thought to be found in MSCs of ~20 nm (Fernandez-Busnadiego, 2015). Trans-SNARE complexes start to interact when the distance between membranes is ~8 nm (Liu, 2007). In the introduction, the authors suggest that incomplete zippering might occur between Stx and Sec22b, resulting in a distance between 10 and 20 nm, which would allow E-Syt localization. In the discussion, however, the authors suggest a model where Sec22b/Stx interaction is important to bring the membranes in ~10 nm distance to enhance LTP activity. Proof for either model is lacking. Please clarify.
2. Fig1A: Please clarify the multiple bands? for Stx3 (anti-eGFP).
3. There is no size marker for panels C1-C6
4. Fig3: Both proteins seem to show a diffuse pattern. Please specify the validity of measuring average distance. A higher magnification zoom of staining pattern in the growth cone and visualization of the calculation could benefit interpretation.
5. E-Syt2 and E-Syt3 are used interchangeably throughout the manuscript and E-Syt1 is left out completely. It would help the reader if the authors could elaborate on their interpretation on the similarities and differences in structure and functionality between the three E-Syts.
6. Why is there a red line in Fig 7b?

Significance

A growing body of literature recognizes the importance of close proximities between membranes, facilitating direct interaction between organelles (Scorrano et al., 2019). Membrane Contact Sites (MCSs) are shown to be important for a wide range of cellular functions, such as lipid and calcium transfer. E-Syts have been recognized as one of the key players in neuronal MCSs, mediating lipid transport (Fernández-Busnadiego et al., 2015). A study published in 2014 by the authors of the current study revealed another two proteins important for MSCs in neurons (Petkovic et al., 2014). ER protein Sec22b and PM SNARE Syntaxin1 were shown to form a non-fusogenic trans-SNARE complex, important for lipid-transfer mediated neurite outgrowth. Gallo and colleagues have now provided important new evidence that these two components (E-Syts and Stx1/Sec22b) are together and may work together at MSCs.

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 identifies and characterises a novel interaction between E-Syts and Sec22b and demonstrates that lipid transfer between the ER and PM contributes to the development of filopodia and neuronal expansion. This interaction with E-Syt2 occurs through the Longin domain of Sec22b Sec22b association. The authors suggest a continuum with further interactions with syntaxin1, that mediates neurite outgrowth. Overall I find this study very interesting and convincing. The experimental analysis is well carried out and the claims are well aligned with their results.

I only have minor issues:

Figure 1. Some of the western blots have several bands and it is difficult to know which band is the relevant one. They should be indicated in the fig panel. Further panel E and F are barely readable and should be redrawn with the appropriate line and font size. Figure 2: is there a difference between the number of dots in axons and dendrites? Can the author elaborate on this aspect as it is not clear from the image presented. Figure 7: statistical analysis should be indicated from the BoNT/A and BoNT/C as BoNT/A represent an appropriate control cleaving SNAP25 but not Syntaxin. On top of controlling fusion and neuronal outgrowth, syntaxin has a role in survival and its cleavage leads to neuronal death. Is this pathway mediated by E-Syts interactions?

Significance

This papers identifies the molecular mechanism of neuronal outgrowth. It 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

In this article, the authors characterize a complex formed by sec22b-stx1-Esyt2. They investigate how such interactions are involved in the modulation of dynamics of the plasma membrane in the context of neuritogenesis. They conclude that the contact sites between the ER and the plasma membrane, mediated by the afore-mentioned complex, contribute the expansion of the plasma membrane.

Major comments:

Overall, the article clearly shows that in mammalian cells there is an interaction between sec22b-stx1-Esyt2 which appears to be important for filopodia formation and possibly neuritogenesis in neurons. However, performing additional experiments to better clarify some links and assumptions made by authors could strengthen the article.

The manuscript relies on work performed either on cell lines (HeLA, PC12) or primary neuronal cultures. Although it is clear the value of the findings obtained using cell lines, they should be seen as a complementary rather than an exclusive approach. This is particularly important as the authors often make claim about neuron-related cellular biology.

For instance, the biochemistry-based findings on the interaction and characterization of the protein complex (Figure 1) are all derived from experiments perfomed in Hela or PC12. As the authors have the capacity to culture and manipulate primary neuronal cultures, such findings should be validates in neuronal cells. The authors could also consider performing biochemical experiments (i.e. co-ip) of the endogenous proteins in neuronal cultures or brain tissue.

The authors do show some evidence regarding the complex in neuronal cells using PLA (proximity ligation assay, figure 2) or super resolution microscopy, however, these findings should be corroborated by stronger findings targeting interaction and not based on simple proximity.

A similar critique regards the experiments using RNA-interference of Figure 4. Performing loss-of-function experiments in neuronal cultures would strengthen and complement the results obtained via over-expression approaches shown in Figure 5.

Given that the authors have already in place all the necessary technology for the suggested biochemical and morphological-related experiments, these could be carried out swiftly within 3-4 months.

Minor comments:

The manuscript is really technical and at times tough to follow; it could benefit from key sentences to better guide the reader, particularly if not coming from the specialist field, in the appreciation of the experiments and results.

Authors should submit the manuscript to a severe round of proofreading. There are several inconsistencies and sometimes what looks like internal comments: i.e. in the methods "STED Missing" or the fact that "LTP" is used everywhere but not defined and considering that the targeted audience is most likely neuroscience-based could easily lead to confusion.

The experiments appear to have been repeated a sufficient number of times and the statistics seem adequate. It would be advisable to show in dot-plots the findings rather than in bar graphs all findings and not just the morphometrics-relative ones.

Significance

This work closely follows the excellent previous work from the Galli laboratory. As such, it is mostly incremental from a technical perspective and does not present particularly novel findings. An interesting aspect would be in addressing directly the influence of the described interactions in the lipid transfer between ER and the plasma membrane but in that sense the manuscript falls short. Although it is to be appreciated the functional readouts in terms of neuritogenesis, in the present state the manuscript features findings suitable for a very specific audience.

I believe that the appropriate audience for the present manuscript lies within the neuroscientific community interested in development, specifically neuritogenesis, and/or membrane dynamics. Additionally, it might be interesting also for researchers outside of the neuroscience community and interested in the dynamics between ER and plasma membrane.

Keywords regarding my expertise: Molecular and Cellular Neuroscience, Morphometrics, Dendrite, Neurons, Dendritogenesis, Biochemistry, Imaging, Microscopy.

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

Authors’ response to reviewers for manuscript “Bacterial killing by complement requires direct anchoring of Membrane Attack Complex precursor C5b-7” (reference #RC-2019-00125)

Our manuscript entitled “Bacterial killing by complement requires direct anchoring of Membrane Attack Complex precursor C5b-7” has been reviewed by Review Commons. We thank the referees for their interest in our study and are very pleased that the referees consider our findings novel, important and well-designed. Based on the comments given by the referees, we have revised our manuscript and have included two new experimental figures:

-Experimental data validating that our gating strategy with Sytox blue correlates well with bacterial killing on plate (Fig S1-B)

-Experimental data validating that non-bactericidal MAC complexes damage the bacterial OM (Fig. S1-C).

In the response letter below, we respond to the comments raised by the reviewers and explain how we have revised our paper accordingly. All changes into the revised manuscript are clearly highlighted in yellow.

POINT-TO-POINT REPLY

Reviewer #1

(Evidence, reproducibility and clarity (Required)): The paper by Doorduijn et al. addresses a question rarely touched upon in modern studies of the complement system, namely the stability and time-resolved functions of complement component. It extends two earlier reports from the same laboratory, however, with a clear, novel point concerning especially the function of C7. The study embodies several techniques and modes of investigation. From these experiments, the paper contributes significantly to our understanding of the MAC complex is formed and why some bacteria escape this host defense mechanism. Over all the study is very well performed and written. I have only a few major comments.

Reviewer #1 raises 3 points:

POINT 1. The AFM pictures shown in Fig. 6D are of outstanding quality. However, it is a disappointment that the outcome of complement incubation was shown only for a complement-resistant E. coli strain. Would it be possible to show the location on the bacterial surface of MAC complexes, or holes, on a complement-susceptible strains? Comparing the visual outcome for such bacteria with locally formed MAC versus C7 replenished would be quite interesting and perhaps important.

ANSWER 1. Since Fig. 6D represents AFM images of MAC on complement-susceptible E. coli bacteria, we assume that the reviewer is asking why we did not perform AFM experiments on complement-resistant strains? To address this question, it is important to note that we have thus far not succeeded in robustly visualizing MAC complexes under conditions at which bacteria were not killed by MAC complexes (Heesterbeek et al., EMBO J, 2019). While non-bactericidal MAC complexes are present on the bacterial surface as demonstrated with C9 deposition by flow cytometry, we hypothesize that they are not well inserted into the membrane (demonstrated by sensitivity to trypsin) and therefore difficult to resolve by AFM. This is consistent with previous AFM experiments on related pore-forming proteins (Leung et al, 2014, 2017), in which inserted pores were readily detected on supported lipid bilayers, but mobile, non-inserted pores were harder to resolve due to the invasiveness of the AFM measurement and/or insufficient temporal resolution. In the revised manuscript we now better clarify this in line 298-301.

POINT 2. The flow cytometric analysis of bacterial killing is somewhat simplistic. Usually, staining of BOTH live and dead bacteria is performed. This permits better gating of the relevant populations. Specifically, the gating seems to fit the population in Fig. S1 only poorly, with the gate in some cases simply dividing what otherwise appears to uniform population ("C9 at t=0")

ANSWER 2. In the revised manuscript, we have now included additional data demonstrating that our gating strategy with Sytox blue correlates well with bacterial killing on plate (new Fig S1-B referred to in line 78-79, 92-93, 96-98 and Supplementals text line 21-24 shows cfu data for Sytox data of Fig 2D). These data correspond with our earlier findings showing that cells gated to be positive for Sytox blue are indeed the relevant population of dead cells (Heesterbeek, EMBO J, 2019). We disagree with the reviewer that the use of a ‘live’ stain is of added value here. Because the outer membrane of Gram-negative bacteria is also a permeability barrier for ‘live’ stains like Syto9, MAC-dependent outer membrane perforation also results in increase in ‘live’ stain during the process of bacterial lysis (also described in Stiefel et al, BMC Microbiol, 2015 PMID: 25881030). We have therefore chosen to only use the Sytox stain in this study as this is a very reliable marker for killing.

POINT 3. The cited literature is, in general, pertinent and comprehensive. I was surprised, however, that none of the many contributions to field of MAC formation by AF Esser was cited. For instance, the studies over C9 conformation (PMID: 2475785) seem not far away in topic from some of the points raised in the present paper.

ANSWER 3. The reviewer is correct that the work of AF Esser has indeed focused on the contribution of C9 and C9 polymerization to the lytic activity of the MAC pore. In the revised manuscript, we have therefore now included some of the work done by AF Esser (references 34, 36 and 37) and have discussed this in our discussion (line 305-309). However, it is important to note that much of the work on the importance of C9 polymerization by AF Esser has been performed erythrocytes and single-membrane particles (also the suggested paper by the reviewer). Translation of these studies to the role of C9 conformation and polymerization on bacterial killing is therefore limited, although it does provide clues to what differences might cause the discrepancy observed between lysis of erythrocytes and bacterial killing by MAC pores.

Reviewer #1 (Significance (Required)): Insight into the concept of locally formed MAC complexes is lacking and the paper clearly adds novel and quantitative data to this point. The paper probably mostly reaches out to an audience interested in the complement system and researchers interested in large protein complexes with conformational changes as part of their function. My own interest lies with complement-mediated protection against bacteria with a special focus on pattern recognition and protein-bacterial surface interactions.

Reviewer #2

(Evidence, reproducibility and clarity (Required)): Doorduijn et al. present a study illustrating the importance of rapid C7 interaction with C5b6 for MAC-dependent killing of complement sensitive bacteria. The absence of direct C7 interaction results in a MAC which i) doesn't kill the bacteria, and ii) is sensitive to trypsin. The authors have step by step investigated this issue by using common in vitro-methods with different strains of bacteria, serum, and/or purified complement proteins. Bacterial killing is evaluated by sytox blue influx in flow cytometry. I like this work. The experimental strategy is sound, and the conclusions are convincing are based on the presented data. The data and the methods presented in such a way that they can be reproduced. I have no concerns regarding the design, execution or conclusions.

Reviewer #2 RAISES 3 POINTS

POINT 1. My only criticism is on the number of replicates and following statistical analysis: • Overall, the experiments are conducted only three times. With the, in general, large differenced seen between the condition, this may still be acceptable. However, the statistic testing using only N=3 is of low value.

ANSWER 1. As the reviewer pointed out, with these in vitro studies where the experimental conditions are highly controlled it is common practice to perform three independent experiments when the differences are large.

POINT 2. The authors have sometimes used paired testing, and sometimes unpaired. For example, Fig. 5A-B is based on paired testing, whereas data in C, which are based on A-B, is tested using unpaired testing. Why so is unclear to me.

ANSWER 2. We thank the reviewer for this comment, as also for Fig. 5C we should have used a paired analysis, so we have done so accordingly in the revised manuscript (line 725-726).

POINT 3. Further on in Fig. 5 A-B., ANOVA with Tukey multiple comparison tests is used, which implements testing between all conditions; still, only significance is reported for blue vs. red. If the intention was to only test red vs. blue, a t-test would be better.

ANSWER 3. As with the previous comment, we have now performed a paired t-test since we only intended to compare C7 at t=0 vs C7 at t=60 in Fig. 5A-B (line 725-726). Moreover, for consistency in our statistical analyses we’ve also applied this to Fig. 3C (line 706-707).

Reviewer #2 (Significance (Required)): As far as I understand, the presented data is of high significance for the conceptual understanding of the buildup of MAC for bacterial killing on Gram-negative bacteria. I work partly with complement but is not an expert on the terminal pathway.

Reviewer #3

(Evidence, reproducibility and clarity (Required)): The study is a follow-up on the paper the same group of scientists published in EMBO J last year. That paper showed that rapid interaction between C5b6 and C7 is necessary for effective killing of Gram negative bacteria. The follow-up this paper makes is to make that case for a series of E. coli strains, showing as part of this that strains of clinical isolate E. coli resistant to complement attack prevent the rapid C5b6-C7 interaction. The story goes that C5 convertase engagement on the surface of targeted bacteria is the necessary context for effective C5>C5b conversion and thence interaction with C6 and C7. The rapid interaction with C7 is necessary because it prevents release/shedding of C5b6 from the bacterial cell surface. Overall, the conclusions seem justified - that C5b6 interaction with C7 stabilises its interaction with the surface and is needed to prevent C5b6 shedding. But this observation needs a mechanical or biophysical framework to be understood properly.

Reviewer #3 RAISES 5 POINTS

POINT 1. The authors do not observe non-bactericidal MAC pores/non-lytic MAC by AFM and so I think in this study there is no evidence for their existence. Their depiction in Figure 8b is therefore misleading and I think should be deleted. Indeed, the authors do not know what the structure of the non-bactericidal MAC pores could be, so depicting them in this specific way isn't appropriate. They have no idea what they might be like, if they exist.

ANSWER 1. We agree with the referee that we do not know the structure of a non-bactericidal MAC pore, and have therefore deleted the speculative structures in Fig. 8B (explained in line 784-785). Although we have no structural information, we do think that non-bactericidal MAC pores exist and our revised manuscript now includes new data to better explain this (Fig S1-C). While our initial manuscript showed that a delayed interaction between C5b6 and C7 results in MAC complexes that cannot perturb the bacterial inner membrane, we now show that these MAC complexes effectively damage the outer membrane (evidenced by leakage of mCherry from the periplasmic space (Fig. S1-C, explained in line 97-98 and Supplementals text line 24-31). This leads us to conclude that there are pores formed in the outer membrane that are not capable of damaging the inner membrane. We think that within this context we can name these ‘non-bactericidal MAC pores’.

POINT 2. This brings me to another point: it is really unclear to me from this study how the authors envisage the inner bacterial membrane be damaged by MAC attack. Do MAC pores formed in the OM deliver MAC components to the IM? Or what happens - is the damage to the IM indirect? The reason why this is relevant to the possibility of non-bactericidal MAC pores is that it could be these are inserted just like bactericidal pores into the OM but the IM attack is deficient in some way.

ANSWER 2. Although we agree with the reviewer that exact mechanism by which MAC pores perturb the inner membrane is unanswered, we think this is beyond the scope of this paper which mainly deals with the time-resolved functions of MAC assembly. However, to meet the referees’ critique, we have now more clearly addressed this question in our discussion and speculate on several mechanisms by which the MAC pore could induce bacterial inner membrane damage (line 277 - 288). In short, we hypothesize that OM damage could indirectly trigger IM damage by affecting regulation of osmosis, overall cell envelope stability and/or envelope stress.

POINT 3. (Significance (Required)). I am intrigued by the difference between MAC assembly on erythrocytes and bacteria. What do the authors believe to be the basis of this difference? It would help understanding of the significance of their work if they could make this clear. Without this kind of attempted explanation the results seem phenomenological - an observation has been made but why this observation occurs, what the important environmental difference is between erythrocyte membranes and the outer membranes of Gram negative bacteria is not addressed. I am looking for some kind of biophysical explanation - specific lipid properties, for example.

ANSWER 3. We agree that this is intriguing and in our revised manuscript we have included different hypotheses on why MAC assembly on erythrocytes and bacteria could be different. Although differences in composition between the erythrocyte membrane and outer membrane can definitely play a role, our data suggest that the difference is mainly a consequence of the fact that Gram-negative bacteria have two membranes (the outer and inner membrane). In the revised manuscript, the newly added figure (Fig S1-C) supports this, since this figure reveals that MAC pores generated from C5b6 that is generated in the absence of C7 can still damage the bacterial OM. However, despite observing OM damage by measuring leakage of a periplasmic protein, this does not lead to bacterial killing and IM damage. Since we here observe that rapid interaction between C5b6 and C7 is required for bacterial killing and IM damage, we think that efficient anchoring of C5b-7 is primarily relevant in damaging the bacterial IM and subsequently causing bacterial cell death. Finally, we have also mentioned this more specifically in the discussion of the revised manuscript (line 277-288).

POINT 4. Related, at the end of the Results section the authors say "Altogether, these data indicate that complement-resistant E. coli can prevent complement-dependent killing by MAC pores by preventing efficient anchoring of C5b-234 7 and insertion of MAC pores into the bacterial cell envelope." My immediate response was: 'How? The Discussion needs to consider this.' But it doesn't.

ANSWER 4. In the revised manuscript, we now explain this more extensively (line 322 – 326). In short, we hypothesize that the composition of the OM, mostly in terms of capsular polysaccharides and lipopolysaccharides, could affect this. We have added additional references supporting the role of these components in complement resistance in multiple Gram-negative species (reference 45-48 and 50).

POINT 5. I was confused by the term "metastable lipophilic domain" at line 262 on page 10. Do the authors mean the MACPF domain?

ANSWER 5. We have now more explicitly named this in our discussion as being the MACPF domain and have further elaborated what we meant by metastable (line 263-267).

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 study is a follow-up on the paper the same group of scientists published in EMBO J last year.

That paper showed that rapid interaction between C5b6 and C7 is necessary for effective killing of Gram negative bacteria. The follow-up this paper makes is to make that case for a series of E. coli strains, showing as part of this that strains of clinical isolate E. coli resistant to complement attack prevent the rapid C5b6-C7 interaction.

The story goes that C5 convertase engagement on the surface of targeted bacteria is the necessary context for effective C5>C5b conversion and thence interaction with C6 and C7. The rapid interaction with C7 is necessary because it prevents release/shedding of C5b6 from the bacterial cell surface.

Overall, the conclusions seem justified - that C5b6 interaction with C7 stabilises its interaction with the surface and is needed to prevent C5b6 shedding. But this observation needs a mechanical or biophysical framework to be understood properly.

The authors do not observe non-bactericidal MAC pores/non-lytic MAC by AFM and so I think in this study there is no evidence for their existence. Their depiction in Figure 8b is therefore misleading and I think should be deleted. Indeed, the authors do not know what the structure of the non-bactericidal MAC pores could be, so depicting them in this specific way isn't appropriate. They have no idea what they might be like, if they exist.

This brings me to another point: it is really unclear to me from this study how the authors envisage the inner bacterial membrane be damaged by MAC attack. Do MAC pores formed in the OM deliver MAC components to the IM? Or what happens - is the damage to the IM indirect? The reason why this is relevant to the possibility of non-bactericidal MAC pores is that it could be these are inserted just like bactericidal pores into the OM but the IM attack is deficient in some way.

Significance

I am intrigued by the difference between MAC assembly on erythrocytes and bacteria. What do the authors believe to be the basis of this difference? It would help understanding of the significance of their work if they could make this clear. Without this kind of attempted explanation the results seem phenomenological - an observation has been made but why this observation occurs, what the important environmental difference is between erythrocyte membranes and the outer membranes of Gram negative bacteria is not addressed. I am looking for some kind of biophysical explanation - specific lipid properties, for example.

Related, at the end of the Results section the authors say "Altogether, these data indicate that complement-resistant E. coli can prevent complement-dependent killing by MAC pores by preventing efficient anchoring of C5b-234 7 and insertion of MAC pores into the bacterial cell envelope." My immediate response was: 'How? The Discussion needs to consider this.' But it doesn't.

I was confused by the term "metastable lipophilic domain" at line 262 on page 10. Do the authors mean the MACPF domain?

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

Doorduijn et al. present a study illustrating the importance of rapid C7 interaction with C5b6 for MAC-dependent killing of complement sensitive bacteria. The absence of direct C7 interaction results in a MAC which i) doesn't kill the bacteria, and ii) is sensitive to trypsin.

The authors have step by step investigated this issue by using common in vitro-methods with different strains of bacteria, serum, and/or purified complement proteins. Bacterial killing is evaluated by sytox blue influx in flow cytometry.

I like this work. The experimental strategy is sound, and the conclusions are convincing are based on the presented data. The data and the methods presented in such a way that they can be reproduced. I have no concerns regarding the design, execution or conclusions.

My only criticism is on the number of replicates and following statistical analysis:

• Overall, the experiments are conducted only three times. With the, in general, large differenced seen between the condition, this may still be acceptable.

• However, the statistic testing using only N=3 is of low value.

• The authors have sometimes used paired testing, and sometimes unpaired. For example, Fig. 5A-B is based on paired testing, whereas data in C, which are based on A-B, is tested using unpaired testing. Why so is unclear to me.

• Further on in Fig. 5 A-B., ANOVA with Tukey multiple comparison tests is used, which implements testing between all conditions; still, only significance is reported for blue vs. red. If the intention was to only test red vs. blue, a t-test would be better.

Significance

As far as I understand, the presented data is of high significance for the conceptual understanding of the buildup of MAC for bacterial killing on Gram-negative bacteria.

I work partly with complement but is not an expert on the terminal pathway.

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 by Doorduijn et al. addresses a question rarely touched upon in modern studies of the complement system, namely the stability and time-resolved functions of complement component. It extends two earlier reports from the same laboratory, however, with a clear, novel point concerning especially the function of C7.The study embodies several techniques and modes of investigation. From these experiments, the paper contributes significantly to our understanding of the MAC complex is formed and why some bacteria escape this host defense mechanism. Over all the study is very well performed and written. I have only a few major comments.

Major Comments

1.The AFM pictures shown in Fig. 6D are of outstanding quality. However, it is a disappointment that the outcome of complement incubation was shown only for a complement-resistant E. coli strain. Would it be possible to show the location on the bacterial surface of MAC complexes, or holes, on a complement-susceptible strains? Comparing the visual outcome for such bacteria with locally formed MAC versus C7 replenished would be quite interesting and perhaps important.

2.The flow cytometric analysis of bacterial killing is somewhat simplistic. Usually, staining of BOTH live and dead bacteria is performed. This permits better gating of the relevant populations. Specifically, the gating seems to fit the population in Fig. S1 only poorly, with the gate in some cases simply dividing what otherwise appears to uniform population ("C9 at t=0")

Minor point

The cited literature is, in general, pertinent and comprehensive. I was surprised, however, that none of the many contributions to field of MAC formation by AF Esser was cited. For instance, the studies over C9 conformation (PMID: 2475785) seem not far away in topic from some of the points raised in the present paper.

Significance

Insight into the concept of locally formed MAC complexes is lacking and the paper clearly adds novel and quantitative data to this point. The paper probably mostly reaches out to an audience interested in the complement system and researchers interested in large protein complexes with conformational changes as part of their function. My own interest lies with complement-mediated protection against bacteria with a special focus on pattern recognition and protein-bacterial surface interactions.

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 reviewers for their useful and constructive comments on our manuscript. All reviewers appreciated that our manuscript provides a good characterization of the KKT2/3 functional domains, especially in solving crystal structures of the KKT2 central domain and revealing the importance of KKT2/3 central domains for their centromere targeting. They also commented that additional experiments (e.g. testing DNA-binding activities using recombinant proteins and examining whether ectopically expressed KKT2 fragments localize at kinetochores transiently) would significantly strengthen the manuscript. In the revised manuscript, we are going to address their comments as follows.

Reviewer #1:

From the information presented, it seems like there are only two possibilities to explain the role of the zinc finger domains in directing centromere targeting. First, this could mediate a protein-protein interaction. The authors attempt to assess this using their mass spec experiments, but this does not absolutely rule this out as this interaction may not persist through their purification procedure (low affinity or requires the presence of DNA, such as for a nucleosome).

Response: We agree with the reviewer’s comment. We will add a sentence to discuss this possibility in the revised manuscript.

Second, this could reflect direct DNA binding by the zinc finger. Although the existing paper is solid and highlights a role for the zinc finger domains in the localization of these proteins, it would be even better if the authors were to at least assess DNA binding in vitro with their recombinant protein. Comparing its behavior to a well characterized DNA-binding zinc finger protein would be powerful for assessing whether direct DNA binding could be responsible for its centromere localization.

Response: We have tested DNA-binding activities for the KKT2 central domain from T. brucei, Bodo saltans, and Perkinsela using a fluorescent polarization assay. We tested three different DNA probes (50 bp each) that were fluorescently-labelled: a 50 bp DNA probe from the CIR147 sequence, which is the unit sequence of centromere repeats of several chromosomes in T. brucei (36% GC content), as well as two random DNA sequences of 25% and 74% GC content. We found that the Perkinsela KKT2a central domain binds these three different DNA probes with similar affinities (Kd ~100 nM), suggesting that the Perkinsela KKT2a central domain binds DNA in a sequence-independent manner. Although we have not been able to obtain reliable results for T. brucei and Bodo saltans proteins thus far (due to quenching of fluorescent signals by these proteins), it is likely that the T. brucei KKT2 central domain also binds DNA in a sequence-independent manner given the similarity of the Znf1 structure/sequence among kinetoplastids. This is consistent with the observation that there is no DNA sequence that is commonly found in the centromere of all chromosomes in T. brucei and other kinetoplastids. We are going to add the DNA-binding assay results for the Perkinsela KKT2a central domain in the revised manuscript. We do not feel it is informative to compare the KKT2 Znf1’s behavior to a well characterized DNA-binding zinc finger protein (that binds specific DNA sequence), because Perkinsela KKT2a binds DNA in a sequence-independent manner.

The code for KKT2 and KKT3 localization is complicated by the multiple regions that contribute to their targeting. This includes both the zinc finger domain that the authors identify here, as well as a second region that appears to act through associations with other constitutive centromere components. Due to this, it feels that there are several aspects of these proteins that are incompletely explored. First, the authors show that the Znf1 mutant in KKT2 localizes apparently normally to centromeres, but is unable to support KKT2 function in chromosome segregation. This suggests that this zinc finger domain could have a separable role in kinetochore function that is distinct from centromere targeting.

Response: We agree with the reviewer that the mechanism of KKT2 kinetochore localization is complicated because there are at least three distinct domains that contribute to its targeting (Figure 2 in the original manuscript), but we showed that the centromere targeting of the ectopically-expressed KKT2 central domain fragment depends on Znf1 (Figure 6B in the original manuscript). Together with the finding that the Znf1-equivalent domain is essential for the localization of the full length KKT3 protein, we think that a function of the KKT2 Znf1 domain is to promote its centromere localization. In the future, it will be critical to understand the molecular mechanism of how the KKT2 central domain localizes specifically at centromeres.

Second, although the authors identify these minimal zinc finger regions as sufficient for centromere localization, they do not test whether this behavior depends on the presence of other KKT proteins. This seems like a very important experiment to test whether recruitment of the zinc finger occurs through other factors, or whether it could act directly through binding to DNA or histones.

Response: We do not have an experimental setup to test whether the centromere localization of KKT2/3 central domains depends on other KKT proteins (i.e. we cannot keep the expression of the central domain to a low level while inducing RNAi constructs at a high level). As an alternative approach, we have been testing the localization dependency of endogenously-tagged full-length KKT2/3 proteins using RNAi against various KKT proteins but our preliminary results have not found any kinetochore protein whose depletion affects the localization of KKT2 or KKT3 at centromeres. Although these results could be explained by inefficient protein depletion, they are consistent with the possibility that KKT2 and KKT3 central domains directly interact with centromere DNA. We could consider adding these data in the revised manuscript, although a significant amount of additional work will be necessary to confirm these results.

• Based on the description of kinetoplastid centromeres that the authors provide, it is actually unclear to whether these are indeed sequence independent. The authors state that "There is no specific DNA sequence that is common to all centromeres in each organism [Trypanosomes and Leishmania], suggesting that kinetoplastids also determine their kinetochore positions in a sequence-independent manner." However, it remains possible that there are features to this DNA that are responsible for defining the centromere. In principle, enriched clustering of a short motif that may elude sequence comparisons could be responsible for specifying these regions. It would be helpful to use caution with this statement, and I would also encourage the Aikyoshi lab to test this directly in future work, such as using strategies to remove a centromere or alter its position. *

Response: We agree with the reviewer that we cannot exclude the possibility that there might be an enrichment of a short motif that promotes the localization of kinetochore proteins. We will discuss this possibility in the revised manuscript.

• It would be helpful to provide a schematic of kinetoplastid kinetochore organization based on their studies to date (possibly in Figure 1) to provide a context for the relationships between the different KKT proteins tested in this paper.*

Response: While we agree with the referee that a model figure would be helpful, we feel that drawing a model for the overall organization of kinetoplastid kinetochores at this stage could be misleading because we still know very little about it. In fact, our published data (e.g. the microtubule-binding kinetochore protein KKT4 localizes at centromeres throughout the cell cycle and has DNA-binding activities) and our unpublished observations suggest that the design principle of kinetoplastid kinetochores may well be fundamentally different from that of canonical kinetochores in other eukaryotes. We therefore would like to obtain more data before drawing a model of kinetoplastid kinetochores. Instead of a model, we are going to include a summary of localization patterns for kinetoplastid kinetochore proteins in Figure 1 to help orient readers.

Reviewer #2: The experiments are in general well presented but some could be better controlled: - localization of KKT2 and KKT3 mutants is never verified to be centromeres, we have to believe the dots in the DAPI region are centromeres.

Response: We have assumed that the KKT2 and KKT3 mutants that had dots very likely localized at centromeres because they behaved similarly to wild-type proteins (i.e. align at metaphase plate in some 2K1N cells and localize at the leading edge of separating chromosomes). We will confirm this assumption by imaging the KKT2/3 mutants with a kinetochore protein marker (e.g. tdTomato-KKT1).

in some cases mutants are made in full-length (FL) background (viability, sometimes localization), but in other cases only in isolated domains. The former should be done for all assays. This is also important to show that central domain of KKT2 and KKT3 is necessary for localization.

Response: It is very laborious to create point mutants in full-length background at an endogenous locus. This is why we first tested a number of mutants in our ectopic expression of truncated (for KKT2) or full-length (for KKT3) proteins to identify the most critical mutations, which were subsequently tested in the endogenous context. Although not included in the original manuscript, we have performed an ectopic expression of additional KKT2 mutants (C597A/C600A, C616A/C619A, C624A/C627A, C640A/C643A, and H656A/C660A) in the full-length protein and found that all of them had apparently normal localization pattern, which is consistent with the results we obtained in the endogenous expression experiments (C576A, D622A, and C640A/C643A: Figure 6c in the original manuscript).

The data of F2 are interpreted to mean that PDB-like domain and middle region get to kinetochores by binding transient KT components, even though KKT2 itself is constitutive. That interpretation would really be strenghthened by showing the KKT2 fragments are now transient also. **

Response: Our observations suggest that these KKT2 fragments indeed localize at centromeres transiently (from S phase to anaphase). We will confirm this result by imaging with a transiently-localized kinetochore protein, KKT1 tagged with tdTomato, and include in the revised manuscript.

The paper could do with some attempts to get to this, based on the presented data. For example, does Znf1 bind centromeric DNA, does it bind nucleosomes, is it essential for recruiting the other KKTs, etc.

Response: As we responded to Reviewer 1, we have found that Perkinsela KKT2a central domain Znf1 has DNA-binding activities. We agree that it will be important to test whether KKT2 binds nucleosomes but it will be necessary for us to reconstitute nucleosomes using recombinant T. brucei histones. It will also be important to test whether KKT2/3 are essential for recruiting other kinetochore proteins but we think that they are beyond the scope of this manuscript.

Reviewer #3: \*Major Comments:** - No page numbers - this makes it difficult to refer to different parts of the text... *

Response: We sincerely apologize for the lack of page numbers in the original manuscript. We will add page numbers and line numbers in the revised manuscript.

Introduction (page 2), fourth-from bottom line: the authors refer here to "regional centromere" but have not defined this term (I assume, as opposed to point-centromeres of budding yeast?). I suggest rephrasing.

Response: We thank the reviewer for pointing it out. We will rephrase it in the revised manuscript.

Page 4, bottom: The discussion of KKT2 kinetochore localization brings up a lot or questions. First, can the authors use an assay like yeast two-hybrid to test for pairwise interactions between KKT2 domains and other kinetochore proteins? This could provide direct functional data on the role of these various domains in kinetochore localization.

Response: Based on the mass spectrometry of immunoprecipitated KKT2 fragments that localized at kinetochores, we are currently trying to identify direct protein-protein interactions between the KKT2 domains and other kinetochore proteins (e.g. does KKT2-DPB directly interact with KKT1, KKT6, or KKT7 proteins?). While we agree that it is important to address these questions, we think that it is beyond the scope of this manuscript because its focus is the characterization of KKT2/3 central domains. As we mentioned in the manuscript, these central domains failed to co-purify with other kinetochore proteins, and the experiment therefore did not give us any clue about how they might localize specifically at centromeres.

Second, if individual domains are being recruited to kinetochores by their non-constitutive binding partners, wouldn't this be evident if the authors looked at localization at different points in the cell cycle, and/or with dual localization tracking the putative binding partners? Could transient localization of some of the domains explain the intermediate localization phenotype observed for some domains in KKT2?

Response: As we responded to Reviewer 2, our observations suggest that these KKT2 fragments indeed localize at centromeres transiently (from S phase to anaphase). We will confirm this result by imaging with a transiently-localized kinetochore protein, KKT1 tagged with tdTomato.

Page 6: The authors note that KKT2 Znf2 bears strong similarity to DNA-binding canonical Zinc fingers, and even note the high conservation of some putative DNA-binding residues. Have the authors tested for DNA binding by this protein?

Response: As we responded to Reviewer 1 and 2, we used a fluorescence polarization assay and found that the Perkinsela KKT2a central domain binds DNA in a sequence-independent manner.

Can the authors at least model DNA binding and see if that would result in a clash, given the packing of Znf2 against the larger Znf1?

Response: As suggested, we superimposed the structure of Bodo saltans KKT2 Znf2 with that of a zinc finger 268 bound to DNA (PDB:1AAY), which shows a possible mechanism by which Znf2 might bind DNA. It also revealed a clash between DNA and Znf1 (in the crystal packing of the solved structure), implying that the position of Znf2 would need to change in order to bind DNA. We will add a supplementary figure showing a hypothetical DNA-binding mechanism by Znf2 and discuss the possibility of a necessary structural change in the Znf1 position to accommodate the DNA binding by Znf2.

\*Minor Comments:** - Page 5: I'm skeptical as to whether these zinc-binding domains, especially Znf1, should really be referred to as "fingers". *

Response: To our knowledge, the word “zinc finger” could be used for any protein that binds one or more zinc ions. Given that we still do not understand the molecular mechanism by which this domain functions, we wanted to use a very general term, Znf1. However, we do appreciate the reviewer’s point that calling this domain as a zinc finger could be misleading, so we will refer Znf1 and Znf2 in the original manuscript as the CL domain (for centromere localizing domain) and a classic C2H2 zinc finger in the revised manuscript.

Page 8: At the beginning of the section describing KKT3 cellular experiments, I think the authors need to make it much more explicit that T. brucei KKT3 shares both Znf1 and Znf2 with KKT2.

Response: We will add the suggested sentence before describing the functional assay for KKT3.

Figure S1A: The gap between lanes in the middle of the major peak is really confusing (it's not even clear that this is two different SDS-PAGE gels next to one another). I initially thought that KKT2 was in both peaks, given the labeling of this figure. I suggest labeling the lanes specifically, or cropping the picture, to avoid confusion.

Response: As suggested, we will prepare an image that shows only those lanes (from two separate gels) that were used for loading protein samples. We also like to retain the whole gel images in the same figure because those gels have rather low background signal (even without any contrast manipulation).

[Note: this preprint has been peer reviewed by eLife. The decision letter after peer review, based on three reviews, follows. The decision was sent on 19 May 2019.]

Summary

This paper describes five cryo-EM structures of ribosomal complexes apparently representing different stages of RF2-catalyzed translation termination. The novel observations here are that the tip of domain 3 of RF2 undergoes a rearrangement from an a-helical conformation to a b-hairpin conformation during termination that likely facilitates exit of the newly synthesized protein from the ribosomal polypeptide exit tunnel and that the ribosome can undergo two thermally activated, spontaneous conformational changes, a relative rotation of the ribosomal subunits and a swiveling of the 'head' domain of the small subunit, during termination that likely facilitate dissociation of RF2 from the ribosome. These are interesting observations that significantly extend our understanding of how class I RFs and ribosome conformational changes drive important steps during termination and, as such, all three reviewers recommended publication provided the following comments are addressed adequately.

Essential Revisions

1) The maps provided through the eLife system seemed to be unsharpened, as they showed very little detail. However, even after sharpening them with a B-factor of -100A2, they still did not show the expected features for their respective resolutions. My suspicion is that FREALIGN has been used to overfit the data. This should be addressed in the revision. It should be indicated whether gold-standard separation of halves of the data sets were used in the final refinements, or whether those were limited to a specific spatial frequency (like was done in the classifications). If the latter, those frequencies should also be stated in the manuscript, and they should be significantly lower than the claimed resolutions.

In addition: a lot of basic cryo-EM information is missing: the authors should include: a) at least one micrograph image b) some representative 2D class averages c) local resolution maps of the five structures. Also, because the density of important parts of the maps seems to be a lot worse than the resolution claimed, it would be good to explicitly mention the local resolution of the important features discussed in the main text. d) for each structure, some zoomed-in figures with the density on top of the molecular model. These figures should be chosen as to validate the resolution claim. For example, in structures I, II and V, the RNA bases should be well separated (they do so at 3.6A), and in structures III and IV beta-strands should be well separated, and many (larger) side chains should be visible. In addition, some panels with density for the most important features of each structure should be shown. e) FSC curves between the refined PDB models and the cryo-EM maps are missing from the manuscript. These should be included. In addition, to evaluate potential overfitting of the models in the maps, for each structure, the authors should also include the FSC curves between a model that was refined in half-map1 versus half-map1, as well as the FSC curve between _thesame model versus half-map2.

2) There appear to be many self-citations, and there are also a few places where relevant citations are missing or are mis-cited. There are too many to list individually, but, just a few examples: Page 4: the only citation for the phrase "recent biophysical and biochemical findings suggest a highly dynamic series of termination events" is a Rodnina paper. There are many, earlier papers from Ehrenberg, Gonzalez, Puglisi, Green, Joseph, etc. that should be cited here. Page 5: The only citation for the sentence "By contrast, biochemical experiments showed..." is a Green paper. There are earlier papers from Ehrenberg characterizing the effects of the GGQ-->GAQ mutations on the ability of RF3 to accelerate the dissociation of class I RFs from termination complexes that should be cited here. Page 5: There's a sentence that refers to X-ray, cryo-EM, and smFRET studies, but only provides citations to two smFRET studies (Casy et al, 2018 and Sternberg et al, 2009); Page 5: Moazed and Noller, 1989 identified and characterized the P/E hybrid state, but they didn't report that a deacylated P-site tRNA 'samples' the P/E hybrid state 'via a spontaneous intersubunit rotation'--that was later work from Noller and Ha; etc. There are several other instances of missing citations or mis-citations. We would ask that the authors review their citations with an eye for excessive self-citations and for missing citations or mis-citations. In this context, "Ensemble-EM" is also cited as a specific method in the introduction (Abeyrathne et al., 2016; Loveland et al., 2017). However, this method is more commonly known as (3D) classification of cryo-EM images, and there are many older, more appropriate citations.

3) The sample imaged is a model sample generated by in vitro assembly with purified components of a termination complex. In order to mimic a bona fide termination complex, a short messenger RNA with a strong Shine-Dalgarno sequence followed by a start codon and immediately after by a stop codon was used (mRNA sequence: 5'-GGC AAG GAG GUA AAA AUG UGA AAAAAA-3'). Similar constructs were used to crystallize termination complexes in the past and it has been proven by smFRET experiments that, at least regarding ribosomal inter-subunit dynamics, this model sample behaves similarly to a real termination complex with a peptide linked to the P site tRNA. However, the nature of this model sample should be apparent for the non-specialist reader, highlighting its similarities with a real termination complex but also its possible limitations, especially regarding the "artificial" nature of having a stop codon so close to the Shine-Dalgarno sequence, a situation that never happens in real mRNAs. The authors should explicitly acknowledge this and discuss its implications in the main text.

4) The authors set up a couple of somewhat 'strawman' arguments in claiming that: (i) there are discrepancies in the X-ray, cryo-EM, and smFRET literature with regard to whether ribosomes can undergo intersubunit rotation while bound to class I RFs or whether the non-rotated conformation of the ribosome is stabilized by bound class I RFs and (ii) class I RFs are able to terminate translation and dissociate from the ribosome without the aid of RF3. In the case of (i), it is obviously possible for class I RF-bound ribosomes to undergo intersubunit rotation while still favoring the non-rotated conformation of the ribosome. Moreover, there are enough differences between the cited studies, both in terms of the experimental conditions as well as the technical limitations associated with the various experimental techniques, that it is easy to rationalize differences with regard to whether the class I RF-bound ribosomes would be expected to undergo intersubunit rotation and/or whether the researchers would have been able to capture/observe intersubunit rotation. In the case of (ii), decades of biochemistry from Buckingham, Ehrenberg, Green, and others had already demonstrated that class I RFs are able to terminate translation and dissociate from the ribosome without the aid of RF3, and that the role of RF3 in termination is to accelerate the spontaneous dissociation of the class I RFs. If the authors want to highlight discrepancies in the literature, they should frame them in the context of differences between the studies, experimental design, limitations of the approaches/techniques in the cited papers that might account for such discrepancies. Re-writing this paragraph also in the light of addressing the missing citations and mis-citations pointed out under (2) will further help in toning these arguments down, which would strengthen the manuscript's scholarship.

5) Class I RFs are post-translationally methylated at the Q residue of the GGQ motif of domain 3 and Buckingham, Ehrenberg, and others have shown that this methylation accelerates and/or facilitates class I-catalyzed termination both in vitro and in vivo. Nonetheless, Svidritskiy et al do not report whether and to what extent their RF2 is methylated. Was RF2 overexpressed in a manner that ensured homogeneous methylation or lack of methylation? If they are overexpressing prfB and not overexpressing prmC, it is likely that they have a mix of methylated and unmethylated RF2. Assuming they are using the wt E. coli prfB gene, then the residue at position 246 is a T, rather than an A or S, and Buckingham has shown that, in the wt T246 background, a lack of methylation at Q252 is either seriously detrimental in richer media or lethal in more minimal media. It was felt that a discussion of this issue was not needed in the main text, but that it would be helpful if the authors would include the important/relevant experimental details in the Methods section, for example, did they use the T246 wt E. coli variant of RF2; and did they overexpress prmC along with prfB?

6) Structure I is denoted and treated as a pre-termination complex, but that does not seem at all possible given that the sample was prepared by incubating a pre-termination complex for 30 min in the presence of excess RF2, conditions that Figure 1-Figure Supplement 3 suggest results in robust termination. Structure I is more likely the non-rotated conformation of a post-termination complex that is in equilibrium with its rotated counterpart, Structure V. Based on my reading of the manuscript, it is likely that the authors understand this point, but are nonetheless using this structure as a mimic/analog of a pre-termination complex. If so, I think this is fine, but the authors should explicitly state that this is what they are doing. Related to this, the authors should clarify the description of their activity assay, show the raw data, and/or report 'Released [S35]-fMet (%)' instead of 'Released [S35]-fMet, CPM' on the y-axis of Figure 1-Figure Supplement 3; as the activity assay is currently described, reported, and plotted, it is impossible to determine whether RF2 is 1% or 99% active in termination.

7) The final sentence of the manuscript reads: "Translation termination and recycling of the release factors and the ribosome therefore rely on the spontaneous ribosome dynamics, triggered by local rearrangements of the universally conserved elements of the peptidyl-transferase and decoding centers". There are a couple of problems with this sentence as written. First, smFRET experiments by Gonzalez, Puglisi, and Rodnina have previously shown that "Translation termination and recycling of the release factors and the ribosome therefore rely on the spontaneous ribosome dynamics" and the relevant articles should therefore be cited here. Moreover, given the data are static structures solved using a sample that is at equilibrium, it is not clear how the authors determined that these spontaneous ribosome dynamics were "triggered by local rearrangements of the universally conserved elements of the peptidyl-transferase and decoding centers". Isn't it equally possible, given the data presented, that the local rearrangements were triggered by the ribosome dynamics?

[Note: this preprint has been peer reviewed by eLife. The decision letter after peer review, based on three reviews, follows. The decision was sent on 24 May 2019.]

Summary

The manuscript from Munkley, Elliott and colleagues shows that the epithelial splicing regulator ESRP2 is transcriptionally upregulated by the androgen receptor (AR), an observation based on a previous study of gene expression changes in response to androgen in the androgen receptor positive LNCaP prostate cancer cell line by some of these investigators. ESRP2 upregulation leads to a series of changes in alternative splicing, including switches with potential effects in disease relapse and metastasis which correlate with disease outcomes. Prostate cancer is driven by androgens via AR, and therapy involves androgen deprivation (ADT) to slow progression. However, it has also been reported that ADT promotes epithelial mesenchymal transition (EMT) (e.g. Sun et al, 2012), which might be related to the common progression to castration resistant prostate cancer following ADT. Munkley et al show that levels of ESRP2 are reduced after androgen deprivation in 7 prostate cancer patients. A number of other analyses using additional cell lines, a xenograft model, and data from other published prostate cancer samples leads to a general proposal that a decrease in ESRP2 expression (but not ESRP1) and some splicing changes associated with its depletion following androgen deprivation may be associated with prostate cancer progression and worse outcomes. One highlighted example is exon 30 in FLNB, skipping of which is associated with metastatic progression in breast cancer.

A number of papers describing roles for ESRP1/2 in various cancers including breast, colorectal, lung, and ovarian carcinomas have yielded conflicting conclusions on the role of ESRPs or epithelial-specific isoforms it regulates, such as CD44, in cancer progression and/or patient outcomes. In some cases ESRPs are proposed to be tumor suppressors, whereas in other cases they are proposed to promote more aggressive cancers (see, for example, Zhang et al., Genes and Dev 33: 166-179 and references therein). As cited by the authors, a recent manuscript reports that duplication and increased expression of ESRP1 (which would largely promote the same splicing events as ESRP2) is associated with more aggressive human prostate cancers. Thus, a central question is whether the current manuscript can provide further clarity regarding the general role of ESRPs (including ESRP2) in cancer, including prostate cancer.

Munkley et al raise the clinically-relevant point that current treatments for prostate cancer might have undesirable side-effects by inhibiting ESRP2 mediated splicing events. Overall, the manuscript is clearly presented. The data documenting the ESRP and AR regulated splicing program, and the restriction of tumor growth by ESRPs (Figs 1-4, 6) are very clear with very nice correlations between responses to ESRP overexpression, knockdown and androgen stimulation.

Essential Revisions

1) A key concern relates to the relative levels and effects of ESRP1 and ESPR2 under conditions of androgen induction or ADT in prostate cells. The authors do a good job documenting that ESRP2 is under transcriptional control of the androgen receptor, while ESRP1 is not, and that there is a 2-fold reduction in ESPR2 expression post-ADT in cancer samples. On the other hand, a) both ESRP 1 and 2 seem down-regulated at the protein level in androgen receptor-negative prostate cancer cells lines (probably by different mechanisms), b) both ESRP1 and 2 mRNAs are up-regulated in tumor samples compared to controls, c) both ESRP1 and ESRP2 are up- regulated in a cohort of metastatic patient samples, d) the correlation between ESRP levels and recurrence free survival is a more significant for ESRP 1 than 2, and e) a number of functional assays from this manuscript and other publications argue that both ESRP1 and ESPR2 can contribute to regulate overlapping targets relevant for epithelial-specific splicing. Therefore one key question that remains is to what extent the androgen-mediated transcriptional regulation of ESRP2 does contribute to splicing regulation in the context of the relative levels / activities of ESRP1: while a number of the results presented show that androgen treatment can promote splicing towards a stronger "epithelial" pattern, the authors should make additional efforts to demonstrate that ablation of ESRP2 alone (in the presence of ESRP1) leads to substantial changes in splicing that would be expected to explain the association of a loss of ESRP2 with worse outcomes, which is an essential point for the validity of their model. For example, an analysis similar to that of Figure 1A for ESRP1 should be included, as well as other experiments aimed to determine whether the activity of ESRP1 can buffer the effects of ATD on ESRP2.

2) There is also a need for clarity in terms of the coherence of the predicted biological effects of the alternative splice site switches and at least one proof-of-principle demonstration that they are relevant for any property of prostate cells relevant to cancer, as it is difficult to draw firm conclusions from the data presented as to whether the regulation of ESRP2 by androgens is definitively associated with prostate cancer progression or outcomes in a positive or negative manner.

a) Figure 5A shows exons that are more included or skipped in prostate cancer vs normal using TCGA data. But only 6 of the 44 ESRP-AR regulated events are highlighted on the plot, two of which do not change significantly, including FLNB which is highlighted in the abstract and is the only event used to test the response to the AR antagonist Casodex. All of the events from Fig 3 should be highlighted in Figure 5A, with ESRP activated and repressed exons clearly distinguished by colour or symbol. The authors should explain -when known- the nature of the differential activities of the isoforms and whether the isoform switch observed in the presence of androgens / mediated by ESRPs is predicted to contribute, repress or be neutral to tumor cell growth, apoptosis, motility, metastasis, etc. and therefore whether a functionally coherent program of alternative splicing is coordinated by ERSPs or whether various contrasting contributions are predicted whose relative significance will depend on context, etc. If not, is it possible to stratify the data e.g. by tumor grade, or by ESRP expression level? Would this for instance, reveal different classes where events such as FLNB do show a difference between cancer and normal in some classes?

b) In Figure 6, why is FLNB e30 the only splicing event monitored for response to Casodex - especially since this is one of the events that is not altered between prostate cancer and normal tissue-? This Figure should be more systematic with more splicing events.

c) Increased inclusion of exon 30 in FLNB (which occurs for example upon androgen stimulation) is consistent with inhibition of EMT (something that could be stated more clearly in the text). But there is no mechanistic model presented as to how a change in FLNB splicing (or other targets) impacts prostate CA. What about the other alternative splicing events highlighted in Figures 4 / 5? Even if FLNB splicing switches have been shown to influence expression of EMT markers in breast cancer cells (Li et al 2018), it will be essential to show that the degree of switch observed in prostate cancer cells (for FLNB or any other gene) has a relevant biological readout.

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

Response to Reviewers Comments

We would like to thank all reviewers for carefully considering our manuscript and providing useful suggestions/ideas. The general consensus was that our study provides an important conceptual advance that reveals a new way of thinking about kinetochore phosphatases. However, in light of our surprising findings, it was suggested that additional experiments would be required to fully validate our conclusions. In particular, it was seen as important to test whether PLK1 can activate MPS1 from the BUB complex and to confirm that PP1 and PP2A are effectively inhibited in situations where MELT dephosphorylation can occur normally (Figure 3).

In general, we agree with these and the other points raised by the reviewers, therefore we plan to address all comments as outlined in detail below.

The major new additions to the final paper will be the following:

1) Experiments to test how BUB-bound PLK1 affects MPS1 activity.

2) Experiments to determine the efficiency of phosphatase inhibition in figure 3.

3) Experiments to test whether maintaining PLK1 at the BUB complex causes SAC silencing defects

4) Evolutionary analysis demonstrating that the PLK1 and PP2A-binding modules have co-evolved in the kinetochore BUB complex. This analysis, which has been performed already, strengthens our manuscript because it provides additional independent evidence for a functional relationship between PLK1 and PP2A on the BUB complex.

---

Reviewer #1

Minor comments:

1) The authors propose that PP1-KNL1 and BUBR1-bound PP2A-B56 continuously antagonise PLK1 association with the BUB complex by dephosphorylating the CDK1 phosphorylation sites on BUBR1 (pT620) and BUB1 (pT609). It is therefore expected that converting these residues to aspartate would increase PLK1 recruitment. It would be interesting to verify if this hypothesis fits with the proposed model.

Response: The general idea to maintain PLK1 at the BUB complex is a good one, but unfortunately polo-box domains do not bind to acidic negatively charged residues. Instead we will attempt to maintain PLK1 at the BUB complex using alternatively approaches (as suggested by reviewer 2).

2) In Figure 1E, are the mean values for BubR1WT+BubWT and BubR1WT+Bub1T609 both normalized to 1? If so, this fails to reveal the contribution of Bub1 T609 for the recruitment of PLK1 when PP2A-B56 is allowed to localize at kinetochores.

Response: The values will be updated and normalised to the BubR1WT+BUB1WT control. We have also performed additional experiments already and overall the results reveal a small reduction in kinetochore PLK1 following BUB1-T609A mutation and a larger reduction upon combined BUBR1-T620A mutation.

3) What underlies the increase in Bub1 levels at unattached kinetochores of siBubR1 cells (Figure S1C?) Is this caused by an increase in Bub1 T609 phosphorylation and consequently unopposed PLK1 recruitment, which consequently increases MELT phosphorylation?

Response: We suspect that PLK1 is not the cause of the increased BUB1 levels because PLK1 kinetochore levels are actually decreased in this situation (Figure S1A).

4) Although the immunoblotting from Figure S1D indicates that BubR1T620A and Bub1T609A are expressed at similar levels as their respective WT counterparts, some degree of single-cell variability is expected to occur. As a complement to Figure 1B,C and Figure S1E,F could the authors plot the kinetochore intensity of BubR1 pT620 and Bub1T609 relative to the YFP-BubR1 and YFP-Bub1 signal, respectively?

Response: There is indeed variability in the level of re-expression of BUBR1/BUB1 on a single cell level, which can at least partially explain the variation on BUBR1-pT620 and BUB1-pT609 observed within in each condition. We can upload these scatter plots at resubmission and include in the supplementary, if required.

5) The authors nicely show that excessive PLK1 levels at the BUB complex are able to maintain MELT phosphorylation and the SAC (independently of MPS1) when KNL1-localised phosphatases are removed (Figures 2A,B). However, it should be noted that PLK1 is able to promote MPS1 activation at kinetochores and so, whether AZ-3146 at 2.5 uM efficiently inhibits MPS1 under conditions of excessive PLK1 recruitment should be confirmed. Can the authors provide a read-out for MPS1 activation status or activity (other than p-MELTs) to exclude a potential contribution of residual MPS1 activity in maintaining the p-MELTs and SAC?

Response: This is a good point because although PLK1 can phosphorylate the MELTs it can also activate MPS1, although it is unknown whether it can do this from the BUB complex. We had left a dotted line in Figure 4B to include this possibility, but we will now test this directly with additional experiments.

6) To examine whether PLK1 removal is the major role of PP1-KNL1 and PP2A-B56 in the SAC or whether they are additionally needed to dephosphorylate the MELTs, the authors monitored MELT dephosphorylation when MPS1 was inhibited immediately after 30-minute of BI2356. This revealed similar dephosphorylation kinetics, irrespective of compromised PP1-KNL1 or PP2A-B56 activity, thus suggesting that these pools of phosphatases are not required to dephosphorylate MELTs. To confirm this and exclude phosphatase redundancy, the authors simultaneously depleted all PP1 and B56 isoforms or treated cells with Calyculin A to inhibit all PP1 and PP2A phosphatases. In both of these situations, the kinetics of MELT dephosphorylation was indistinguishable from wild type cells if MPS1 and PLK1 were inhibited together. These observations led to the conclusion that neither PP1 or PP2A are required to dephosphorylate the MELT motifs. Instead they are needed to remove PLK1 from the BUB complex. This set of experiments is well-designed and the results support the conclusion. However, it would be of value if the authors provide evidence for the efficiency of PP1 and B56 isoforms depletion and for the efficiency of phosphatase inhibition by Calyculin A. An alternative read-out for the activity of PP1 and PP2A-B56 (other than p-MELT dephosphorylation) clearly confirming that both phosphatases are compromised when MPS1 and PLK1 are inhibited together could make a stronger case in excluding the contribution of residual PP1 or PP2A to the observed dephosphorylation of MELT motifs.

Response: This is also a good point. We had attempted many different combinations in Figure 3 to inhibit PP1/PP2A activity as efficiently as possible. This is especially important considering the “negative” results on pMELT are very surprising. However, we will now test how efficiently we have inhibited PP1 and PP2A phosphatase function in these experiments.

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

Major comments:

1) In its current state I am not convinced that the key conclusions are fully supported by the experiments and alternative conclusions/interpretations can be drawn. For example the level of MELT phosphorylation will be determined by the balance of kinase and phosphatase activity and if they do not achieve 100% inhibition of Mps1 in their assays then they are not strictly monitoring dephosphorylation kinetics in their assays. If the combination of Mps1 and Plk1 inhibition then more strongly inhibits Mps1 then dephosphorylation kinetics becomes faster. Thus subtle differences in Mps1 activity under their different conditions could lead to misleading conclusions but in its present state a careful analysis of Mps1 activity is not provided. This lack of complete inhibition also applies to the phosphatases and the experiments in Figure 3E indicates that their Calyculin preparation is not really active as at steady state MELT phosphorylation levels are much less affected than in for instance BubR1 del PP2A (Figure 2A as an example). Thus they likely still have phosphatase activity in the experiment in figure 3E making it difficult to draw the conclusions they do. A more careful analysis of kinase and phosphatase activities in their different perturbations would be recommendable and should be possible within a reasonable time frame.

Response: These are good points and we will now more carefully assess MPS1 and PP1/PP2A activities.

2) A more stringent test of their model would also be needed. What happens if Plk1 is artificially maintained in the Bub complex? The prediction would be that SAC silencing should be severely delayed even when Mps1 is inhibited. This is a straightforward experiment to do that should not take too long. If the polobox can bind phosphoSer then one could also make BubR1 T620S to slow down dephosphorylation of this site (PPPs work slowly on Ser while Cdk1 have almost same activity for Ser and Thr).

Response: These are good suggestions and we will try to see if maintaining PLK1 at the BUB complex produces effects on the SAC.

3) Another issue is the relevance of Plk1 removal under normal conditions. As their quantification shows in figure 1D-E (I think there is something wrong with figure 1E - should likely be Bub1) the contribution of BubR1 T620 and Bub1 T609 to Plk1 kinetochore localisation seems minimal. Thus upon SAC satisfaction there is not really a need to remove Plk1 through dephosphorylation as it is already at wild type levels. It is only in their BubR1 and KNL1 mutants that there is this effect so one has to question the impact in a normal setting. This is consistent with the data in Figure S1D showing no phosphorylation of these sites under unperturbed conditions.

Response: The major finding of this study is that kinetochore phosphatases are primarily needed to supress PLK1 activity on the BUB complex and thereby prevent excessive MELT phosphorylation. The relevance of this continued PLK1 removal under normal conditions is clear, because when it cannot occur (i.e. if the phosphatases are removed) then the SAC cannot be silenced unless PLK1 is inhibited. Therefore, whilst it is true that PLK1 localisation to the BUB complex is low under normal conditions, that is because the phosphatases are working to keep it that way. The relevance of that continual removal is an interesting, but in our opinion, separate question that will require a new body of work to resolve. One possibility is that PLK1 recruitment is a continual dynamic process, that is perhaps coupled to a particular stage in MCC assembly. For example, PLK1 could bind the BUB complex to recruit PP2A to BUBR1, before being immediately removed by PP2A. In this sense, PLK1 binding could still be functionally important even if it is only occurs transiently and steady state PLK1 levels are low. We will add a line to the discussion to highlight that it would be interesting to test PLK1 dynamics on the BUB complex in future.

4) They write that in the absence of phosphatase activity Plk1 becomes capable of supporting SAC independently (of Mps1 is implied). They do not show this - only that MELT phosphorylation is maintained. As Mps1 has other targets required for SAC activity I would rephrase this.

Response: Good point, this will be rephrased.

Reviewer #2 (Significance (Required)):

The advance is clearly conceptual and provides a new way of thinking about the kinetochore localized phosphatases. These phosphatases and the SAC have been immensely studied but this work brings in a new angle. The discussion would benefit from some evolutionary perspectives as the PP1 and PP2A-B56 binding sites are very conserved but the Plk1 docking sites on Bubs less so. This will be of interest to people in the field of cell division and researchers interested in phospho-mediated signaling.

Response: Since the paper was submitted, we performed evolutionary analysis to examine this point. We discovered that the PLK1 docking sites are surprisingly well conserved and, in fact, they appear to have co-evolved within the same region of MAD/BUB along with the PP2A-B56 binding motif. We believe this new data strengthens our manuscript because it argues strongly for an important functional relationship between PLK1 and PP2A. A new figure containing this evolutionary analysis will be included in the final version.

Reviewer #3

Major comments:

1. An important limitation of this study is that KNL1 dephosphorylation at MELT repeats is monitored only by indirect immunofluorescence using phospho-specific antibodies. Thus, reduction of phospho-KNL1 kinetochore signals could be due to protein turnover at kinetochores, rather than to dephosphorylation. This is a serious issue that could be addressed by checking KNL1 dephosphorylation during time course experiments by western blot using phospho-specific antibodies, as previously done (Espert et al., 2014).

Response: This is an important point that we feel is best addressed by examining total KNL1 levels at kinetochores (instead of simply total cellular levels by western blots). The reason is that KNL1 could potentially still be lost from kinetochores even if the total protein is not degraded. In all experiments involving YFP-KNL1 we observe no change in kinetochore KNL1 levels and this data will be included in the final version. We will also perform new experiments to examine total KNL1 levels in the BUBR1-WT/DPP2A situation to test whether KNL1 kinetochore levels are similarly maintained in these cells following MPS1 inhibition.

1. For obvious technical reasons, the shortest time point at which authors compare KNL1 dephosphorylation upon MPS1-PLK1 inhibition is 5 minutes. Based on immunofluorescence data, authors conclude that kinetics of KNL1 dephosphorylation are similar when kinases are inhibited, independent of whether or not kinetochore-bound phosphatases are active. However, in most experiments (e.g. Fig. 3B, 3C, 3E) lower levels of MELT phosphorylation are detected after 5 minutes of kinase inhibition when phosphatases are present than when they are absent, suggesting that phosphatases likely do contribute to KNL1 dephosphorylation. I suspect that differences between the presence and absence of phosphatases might even be more obvious if authors were to look at shorter time points, when phosphatases conceivably accomplish their function. I would therefore suggest that the authors tone down their conclusions, as their data complement but do not disprove the previous model.

Response: We appreciate that small differences can be seen in figure 3B and 3E at the 5-minute timepoint (between the WT and phosphatase inhibited situations). This may reflect a role for the phosphatases in dephosphorylation or in the ability of drugs such as BI-2536 (3B) or Calyculin A (3E) to fully inhibit their targets in the short timeframe. We will perform additional experiments to examine MPS1 and phosphatase activity under these conditions, in response to comments by reviewers 1 and 2. In the final version we will carefully interpret the new and existing data and, if required, modify the conclusions appropriately.

1. In all experiments cells are kept mitotically arrested through nocodazole treatment, which is not quite a physiological condition to study SAC silencing. This could potentially mask the real contribution of phosphatases in MELT dephosphorylation. Indeed, it is possible that higher amounts of phosphatases are recruited to kinetochores during SAC silencing than during SAC signalling (e.g. during SAC signalling Aurora B phosphorylates the RVSF motif of KNL1 to keep PP1 binding at low levels; Liu et al., 2010). What would happen in a nocodazole wash-out? Would phosphatases be dispensable in these conditions for normal kinetics of MELT dephosphorylation and anaphase onset if PLK1 is inhibited?

Response: All SAC silencing assays where performed in nocodazole for 2 main reasons: 1) PP2A-B56, PP1 or PLK1 can all regulate kinetochore-microtubule attachments, and thereby control the SAC indirectly. Therefore, performing our assays in the absence of microtubules allows us to make specific and direct conclusions about SAC regulation; 2) Previous work on pMELT regulation by PP1/PP2A in human cells was also performed following MPS1 inhibition in nocodazole (Espert et al 2014, Nijenhuis et al, 2014). Therefore, we are able to directly compare the contribution of PLK1 to the previously observed phenotypes, which allowed us to conclude that PLK1 has a major influence. Nevertheless, we appreciate the point that the influence of PLK1 could, in theory, be different during a normal mitosis when microtubule attachment can form. Therefore, we will attempt to address whether PLK1 inhibition can bypass a requirement for PP1/PP2A in SAC silencing during an unperturbed mitosis.

Other data are overinterpreted. For instance, the evidence that CDK1-dependent phosphorylation sites in Bub1 and BubR1 is enhanced when PP1 and PP2A-B56 are absent at kinetochores suggests but does not "demonstrate that PP1-KNL1 and BUBR1-bound PP2A-B56 antagonise PLK1 recruitment to the BUB complex by dephosphorylating key CDK1 phosphorylation sites on BUBR1 (pT620) and BUB1 (pT609)(Figure 1F)". Similarly, the claim "when kinetochore phosphatase recruitment is inhibited, PLK1 becomes capable of supporting the SAC independently" referred to Fig. 2C-D is an overstatement, as residual MPS1 kinase could be still active in the presence of the AZ-3146 inhibitor.

Response: These are good points and the indicated statements will be reworded.

Minor comments:

1. In many graphs (Fig. 1A-C, Fig. 2A,C) relative kinetochore intensities are quantified over "CENPC or YFP-KNL1". Authors should clarify when it is one versus the other.

Response: This will be clarified in the axis and in the methods.

1. The drawing in Fig. 1F depicts the action of PP1 and PP2A-B56 in antagonising PLK1 at kinetochores. Thus, the output should be SAC silencing, rather than activation.

Response: The SAC symbol will be removed from the schematic to avoid confusion and because it is not actually the focus of figure 1 anyway.

1. In the Discussion authors speculate that KNL1 dephosphorylation relies on a constitutive phosphatase with unregulated basal activity. Would a phosphatase be needed at all when MPS1 and PLK1 are inhibited? Could phosphorylated KNL1 be actively degraded?

Response: We will insert total KNL1 immunofluorescence quantification so show that KNL1 KT levels are not decreased in this situation. KNL1 remains anchored at kinetochore but the MELTs must be dephosphorylated to remove the BUB complex.

1. What happens to MPS1 when KNL1-bound PP1 and BUBR1-bound PP2A are absent? Do its kinetochore levels increase as observed for PLK1? And what about the kinetochore levels of Bub1 and BubR1?

Response: We have demonstrated previously that BUB1/BUBR1 increase in this situation in line with the pMELTs (Nijenhuis et al 2014;l Smith et al, 2019) – these papers will be referenced in relation to this. We will also address the effect of phosphatase removal on MPS1 activity, in response to comments by reviewers 1 and 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

The Spindle Assembly Checkpoint (SAC) is a conserved surveillance device that responds to errors in kinetochore-microtubule attachments to ultimately prevent the onset of anaphase until all chromosomes are bipolarly attached. Current models of SAC posit that the Mps1 kinase initiates the SAC signalling cascade by phosphorylating the KNL1/Blinkin kinetochore scaffold at MELT repeats, in order to create phospho-docking sites for the hetero-tetrameric BUB complex made by BUB1-BUB3-BUB3-BUBR1. The BUB complex, in turn, promotes the assembly the Mitotic Checkpoint Complex (MCC), which prevents anaphase onset by inhibiting the E3 ubiquitin ligase Anaphase-Promoting Complex bound to its activator Cdc20 (APCCdc20). The polo-like kinase PLK1, which is recruited to kinetochores through its binding to BUBR1, contributes to the robustness of SAC signalling in human cells by cooperating with Mps1 in KNL1/Blinkin phosphorylation and by phosphorylating MPS1 itself, thereby enhancing its catalytic activity. While in human cells MPS1 is the predominant kinase in SAC signalling, aided by PLK1, in other organisms where MPS1 is absent, such as in nematodes, PLK1 functionally replaces MPS1 and is necessary for SAC activation. Once all chromosomes are bipolarly attached, SAC signalling is extinguished. Key to this process are the PP1 and PP2A-B56 phosphatases that antagonise KNL1 phosphorylation by MPS1 and PLK1 and also dephosphorylate the T-loop of MPS1 to lower its catalytic activity. Current models envision that PP1 and PP2A-B56 dephosphorylate the MELT repeats of KNL1 directly. Importantly, this has been formally tested for both PP2A-B56 in human cells (Espert et al., 2014) and PP1 in yeast (London et al., 2012).

In the present manuscript, the above model is challenged with the proposal that the main contribution of PP1 and PP2A-B56 to SAC silencing is to lower the levels of PLK1 at kinetochores, rather than to dephosphorylate KNL1. By interfering with the levels of these opposing kinases and phosphatases at kinetochores the authors describe an interesting interplay that confirms an overlapping function of PLK1 and MPS1 in KNL1 phosphorylation and highlights a role for the phosphatases in dampening PLK1 kinetochore levels. Consistently, inhibition of both Mps1 and PLK1 is sufficient to bring about KNL1 dephosphorylation upon inhibition of both phosphatases at kinetochores. The hypothesis is interesting and experiments are in general carefully designed and performed. It is clear from the presented data that PP1 and PP2A-B56 antagonize PLK1 kinetochore localisation and that the MELT repeats of KNL1 can be dephosphorylated even in the absence of phosphatases, provided that MPS1 and PLK1 are inhibited. However, in my opinion the results do not rule out that phosphatases actually have a primary and direct role in KNL1 dephosphorylation.

Major comments:

1. An important limitation of this study is that KNL1 dephosphorylation at MELT repeats is monitored only by indirect immunofluorescence using phospho-specific antibodies. Thus, reduction of phospho-KNL1 kinetochore signals could be due to protein turnover at kinetochores, rather than to dephosphorylation. This is a serious issue that could be addressed by checking KNL1 dephosphorylation during time course experiments by western blot using phospho-specific antibodies, as previously done (Espert et al., 2014).
2. For obvious technical reasons, the shortest time point at which authors compare KNL1 dephosphorylation upon MPS1-PLK1 inhibition is 5 minutes. Based on immunofluorescence data, authors conclude that kinetics of KNL1 dephosphorylation are similar when kinases are inhibited, independent of whether or not kinetochore-bound phosphatases are active. However, in most experiments (e.g. Fig. 3B, 3C, 3E) lower levels of MELT phosphorylation are detected after 5 minutes of kinase inhibition when phosphatases are present than when they are absent, suggesting that phosphatases likely do contribute to KNL1 dephosphorylation. I suspect that differences between the presence and absence of phosphatases might even be more obvious if authors were to look at shorter time points, when phosphatases conceivably accomplish their function. I would therefore suggest that the authors tone down their conclusions, as their data complement but do not disprove the previous model.
3. In all experiments cells are kept mitotically arrested through nocodazole treatment, which is not quite a physiological condition to study SAC silencing. This could potentially mask the real contribution of phosphatases in MELT dephosphorylation. Indeed, it is possible that higher amounts of phosphatases are recruited to kinetochores during SAC silencing than during SAC signalling (e.g. during SAC signalling Aurora B phosphorylates the RVSF motif of KNL1 to keep PP1 binding at low levels; Liu et al., 2010). What would happen in a nocodazole wash-out? Would phosphatases be dispensable in these conditions for normal kinetics of MELT dephosphorylation and anaphase onset if PLK1 is inhibited?
4. Other data are overinterpreted. For instance, the evidence that CDK1-dependent phosphorylation sites in Bub1 and BubR1 is enhanced when PP1 and PP2A-B56 are absent at kinetochores suggests but does not "demonstrate that PP1-KNL1 and BUBR1-bound PP2A-B56 antagonise PLK1 recruitment to the BUB complex by dephosphorylating key CDK1 phosphorylation sites on BUBR1 (pT620) and BUB1 (pT609)(Figure 1F)". Similarly, the claim "when kinetochore phosphatase recruitment is inhibited, PLK1 becomes capable of supporting the SAC independently" referred to Fig. 2C-D is an overstatement, as residual MPS1 kinase could be still active in the presence of the AZ-3146 inhibitor.

Minor comments:

1. In many graphs (Fig. 1A-C, Fig. 2A,C) relative kinetochore intensities are quantified over "CENPC or YFP-KNL1". Authors should clarify when it is one versus the other.
2. The drawing in Fig. 1F depicts the action of PP1 and PP2A-B56 in antagonising PLK1 at kinetochores. Thus, the output should be SAC silencing, rather than activation.
3. In the Discussion authors speculate that KNL1 dephosphorylation relies on a constitutive phosphatase with unregulated basal activity. Would a phosphatase be needed at all when MPS1 and PLK1 are inhibited? Could phosphorylated KNL1 be actively degraded?
4. What happens to MPS1 when KNL1-bound PP1 and BUBR1-bound PP2A are absent? Do its kinetochore levels increase as observed for PLK1? And what about the kinetochore levels of Bub1 and BubR1?

Significance

The nature of the advance is conceptual. This paper challenges (although I would rather say "integrates") the prevailing model of spindle checkpoint silencing.

The current model of SAC silencing envisions that PP1 and PP2A-B56 phosphatases oppose SAC kinases (Mps1 and Polo kinase) by directly dephosphorylating some of their targets (e.g. the kinetochore scaffold KNL1 and MPS1 itself). This work proposes instead that the main function of the above phosphatases is to keep low levels of the polo kinase PLK1 at kinetochores, which would otherwise boost KNL1 phosphorylation and assembly of SAC complexes.

People working in the fields of mitosis, chromosome segregation, aneuploidy, spindle checkpoint, kinases/phosphatases could be interested by these findings.

Reviewer's field of expertise: Cell cycle, mitosis, spindle assembly checkpoint

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 work focuses on the role of kinetochore localized protein phosphatases in the dephosphorylation of MELT motifs and SAC silencing. The focus is on PP1 bound to KNL1 and PP2A-B56 bound to BubR1 and uses largely RNAi rescue experiments in human cell lines combined with immunofluorescence analysis and time-lapse imaging. The authors show that kinetochore localized phosphatases antagonize the localization of the Plk1 mitotic kinase to kinetochores. This is due to the dephosphorylation of BubR1 T620 and Bub1 T609 that are binding sites for Plk1 on the kinetochore. The main conclusion is that if Plk1 kinetochore localisation is prevented then there is no longer a need for kinetochore phosphatases for SAC silencing and MELT dephosphorylation.

Major comments:

1) In its current state I am not convinced that the key conclusions are fully supported by the experiments and alternative conclusions/interpretations can be drawn. For example the level of MELT phosphorylation will be determined by the balance of kinase and phosphatase activity and if they do not achieve 100% inhibition of Mps1 in their assays then they are not strictly monitoring dephosphorylation kinetics in their assays. If the combination of Mps1 and Plk1 inhibition then more strongly inhibits Mps1 then dephosphorylation kinetics becomes faster. Thus subtle differences in Mps1 activity under their different conditions could lead to misleading conclusions but in its present state a careful analysis of Mps1 activity is not provided. This lack of complete inhibition also applies to the phosphatases and the experiments in Figure 3E indicates that their Calyculin preparation is not really active as at steady state MELT phosphorylation levels are much less affected than in for instance BubR1 del PP2A (Figure 2A as an example). Thus they likely still have phosphatase activity in the experiment in figure 3E making it difficult to draw the conclusions they do. A more careful analysis of kinase and phosphatase activities in their different perturbations would be recommendable and should be possible within a reasonable time frame.

2) A more stringent test of their model would also be needed. What happens if Plk1 is artificially maintained in the Bub complex? The prediction would be that SAC silencing should be severely delayed even when Mps1 is inhibited. This is a straightforward experiment to do that should not take too long. If the polobox can bind phosphoSer then one could also make BubR1 T620S to slow down dephosphorylation of this site (PPPs work slowly on Ser while Cdk1 have almost same activity for Ser and Thr).

3) Another issue is the relevance of Plk1 removal under normal conditions. As their quantification shows in figure 1D-E (I think there is something wrong with figure 1E - should likely be Bub1) the contribution of BubR1 T620 and Bub1 T609 to Plk1 kinetochore localisation seems minimal. Thus upon SAC satisfaction there is not really a need to remove Plk1 through dephosphorylation as it is already at wild type levels. It is only in their BubR1 and KNL1 mutants that there is this effect so one has to question the impact in a normal setting. This is consistent with the data in Figure S1D showing no phosphorylation of these sites under unperturbed conditions.

4) They write that in the absence of phosphatase activity Plk1 becomes capable of supporting SAC independently (of Mps1 is implied). They do not show this - only that MELT phosphorylation is maintained. As Mps1 has other targets required for SAC activity I would rephrase this.

5) The method section is extensive and contains sufficient information for reproducing data.

6) Data and statistical analysis is ok.

Significance

The advance is clearly conceptual and provides a new way of thinking about the kinetochore localized phosphatases. These phosphatases and the SAC have been immensely studied but this work brings in a new angle. The discussion would benefit from some evolutionary perspectives as the PP1 and PP2A-B56 binding sites are very conserved but the Plk1 docking sites on Bubs less so. This will be of interest to people in the field of cell division and researchers interested in phospho-mediated signaling.

Field of expertise: kinetochore/phosphatases/bub proteins Jakob Nilsson

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 Cordeiro et al provides a series of compelling evidences to support a provocative conclusion: PP2A-B56 and PP1 are critical for SAC silencing mainly by restraining and extinguishing autonomous kinase activity at kinetochores. This finding challenges the prevailing view of PP2A-B56/PP1-mediated KNL1-MELT dephosphorylation as a major SAC silencing event. This represents a paradigm change in the field and opens an important goal for future research: determine the phosphatases that dephosphorylate the MELTs. In my view this paper delivers an important clarification on how PP1-KNL1 and PP2A-B56 actually drive SAC silencing. This is a nice study and will move the field forward. The manuscript is globally solid, very well written and the conclusions are generally supported by the experimental data. However, I do have some issues with the following points, which in my view, if unaddressed, may leave the conclusion a bit fragile:

Minor comments:

1) The authors propose that PP1-KNL1 and BUBR1-bound PP2A-B56 continuously antagonise PLK1 association with the BUB complex by dephosphorylating the CDK1 phosphorylation sites on BUBR1 (pT620) and BUB1 (pT609). It is therefore expected that converting these residues to aspartate would increase PLK1 recruitment. It would be interesting to verify if this hypothesis fits with the proposed model.

2) In Figure 1E, are the mean values for BubR1WT+BubWT and BubR1WT+Bub1T609 both normalized to 1? If so, this fails to reveal the contribution of Bub1 T609 for the recruitment of PLK1 when PP2A-B56 is allowed to localize at kinetochores.

3) What underlies the increase in Bub1 levels at unattached kinetochores of siBubR1 cells (Figure S1C?) Is this caused by an increase in Bub1 T609 phosphorylation and consequently unopposed PLK1 recruitment, which consequently increases MELT phosphorylation?

4) Although the immunoblotting from Figure S1D indicates that BubR1T620A and Bub1T609A are expressed at similar levels as their respective WT counterparts, some degree of single-cell variability is expected to occur. As a complement to Figure 1B,C and Figure S1E,F could the authors plot the kinetochore intensity of BubR1 pT620 and Bub1T609 relative to the YFP-BubR1 and YFP-Bub1 signal, respectively?

5) The authors nicely show that excessive PLK1 levels at the BUB complex are able to maintain MELT phosphorylation and the SAC (independently of MPS1) when KNL1-localised phosphatases are removed (Figures 2A,B). However, it should be noted that PLK1 is able to promote MPS1 activation at kinetochores and so, whether AZ-3146 at 2.5 uM efficiently inhibits MPS1 under conditions of excessive PLK1 recruitment should be confirmed. Can the authors provide a read-out for MPS1 activation status or activity (other than p-MELTs) to exclude a potential contribution of residual MPS1 activity in maintaining the p-MELTs and SAC?

6) To examine whether PLK1 removal is the major role of PP1-KNL1 and PP2A-B56 in the SAC or whether they are additionally needed to dephosphorylate the MELTs, the authors monitored MELT dephosphorylation when MPS1 was inhibited immediately after 30-minute of BI2356. This revealed similar dephosphorylation kinetics, irrespective of compromised PP1-KNL1 or PP2A-B56 activity, thus suggesting that these pools of phosphatases are not required to dephosphorylate MELTs. To confirm this and exclude phosphatase redundancy, the authors simultaneously depleted all PP1 and B56 isoforms or treated cells with Calyculin A to inhibit all PP1 and PP2A phosphatases. In both of these situations, the kinetics of MELT dephosphorylation was indistinguishable from wild type cells if MPS1 and PLK1 were inhibited together. These observations led to the conclusion that neither PP1 or PP2A are required to dephosphorylate the MELT motifs. Instead they are needed to remove PLK1 from the BUB complex. This set of experiments is well-designed and the results support the conclusion. However, it would be of value if the authors provide evidence for the efficiency of PP1 and B56 isoforms depletion and for the efficiency of phosphatase inhibition by Calyculin A. An alternative read-out for the activity of PP1 and PP2A-B56 (other than p-MELT dephosphorylation) clearly confirming that both phosphatases are compromised when MPS1 and PLK1 are inhibited together could make a stronger case in excluding the contribution of residual PP1 or PP2A to the observed dephosphorylation of MELT motifs.

To summarize, this is a very good paper and will definitely cause an important impact in the field of mitosis.

Significance

This manuscript provides an important conceptual advance for the field of mitosis, specifically to the topic of mitotic checkpoint regulation. It remains elusive how the spindle assembly checkpoint is silenced. While previous studies have shown that PP1-KNL1 and PP2A-B56 contribute to suppress SAC signaling, how they do so is unclear. This study provides important insight into this matter. Cordeiro and colleagues demonstrate that in contrast with previous expectations, PP1 and PP2A promote SAC silencing, not by directly dephosphorylating MELT motifs on KNL1, but instead by removing PLK1 from the Bub complex. The authors find that these phosphatases antagonise CDK1- phosphorylations on BubR1 and Bub1 to dampen PLK1 levels. This activity is crucial to prevent PLK1 from maintaining MELT phosphorylation in an autocatalytic manner, thus (probably) allowing prompt SAC silencing following stable kinetochore-microtubule attachments. The described mechanism extends our view of how the SAC is regulated and should be of interest to those in the field of mitosis. The findings described in this paper allow us to better understand how cells silence the SAC. This is a top priority in the field, as the inability to timely quench SAC signaling can result in chromosome segregation errors. Determining the phosphatases that actually dephosphorylate the MELT motifs will be an essential next step forward

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

The manuscript "Unconventional kinetochore kinases KKT2 and KKT3 have a unique zinc finger that promotes their kinetochore localization" by Marciano et al. describes functional and structural work on two unique kinetochore-localized proteins in kinetoplastids, KKT2 and KKT3. While the kinetochores of most eukaryotes are built on top of a histone H3 variant known as CENP-A (or CenH3), kinetoplastids lack CENP-A. Kinetoplastids also lack homologs of most conserved kinetochore proteins and instead possess an unique complement of kinetochore proteins, as described in earlier work by the lead author, B. Akiyoshi.

The current manuscript follows up this earlier work and seeks to understand how two putative kinases, KKT2 and KKT3, localize to the kinetochores of kinetoplastids. They begin by mapping the regions of both proteins (in Trypanosoma brucei) that are required for kinetochore localization. In both cases, a conserved "central domain" is sufficient for kinetochore localization. They then purify and determine the structure of a KKT2 central domain from a related species (Bodo saltans), and show that it possess two zinc-binding domains, termed Znf1 and Znf2. A more diverged KKT2 from Perkinsela has Znf1, but not Znf2. The authors go on to show that the Znf1 region in particular is important for localization of both KKT2 and KKT3 to kinetochores, and for long-term cell survival, in Trypanosoma brucei.

Major Comments:

• The work is well done, well described, and described in such a way that it should be reproducible.

• No page numbers - this makes it difficult to refer to different parts of the text...

• Introduction (page 2), fourth-from bottom line: the authors refer here to "regional centromere" but have not defined this term (I assume, as opposed to point-centromeres of budding yeast?). I suggest rephrasing.

• Page 4, bottom: The discussion of KKT2 kinetochore localization brings up a lot or questions. First, can the authors use an assay like yeast two-hybrid to test for pairwise interactions between KKT2 domains and other kinetochore proteins? This could provide direct functional data on the role of these various domains in kinetochore localization. Second, if individual domains are being recruited to kinetochores by their non-constitutive binding partners, wouldn't this be evident if the authors looked at localization at different points in the cell cycle, and/or with dual localization tracking the putative binding partners? Could transient localization of some of the domains explain the intermediate localization phenotype observed for some domains in KKT2?

• Page 6: The authors note that KKT2 Znf2 bears strong similarity to DNA-binding canonical Zinc fingers, and even note the high conservation of some putative DNA-binding residues. Have the authors tested for DNA binding by this protein? Can the authors at least model DNA binding and see if that would result in a clash, given the packing of Znf2 against the larger Znf1?

Minor Comments:

• Page 5: I'm skeptical as to whether these zinc-binding domains, especially Znf1, should really be referred to as "fingers"

• Page 8: At the beginning of the section describing KKT3 cellular experiments, I think the authors need to make it much more explicit that T. brucei KKT3 shares both Znf1 and Znf2 with KKT2.

• Figure S1A: The gap between lanes in the middle of the major peak is really confusing (it's not even clear that this is two different SDS-PAGE gels next to one another). I initially thought that KKT2 was in both peaks, given the labeling of this figure. I suggest labeling the lanes specifically, or cropping the picture, to avoid confusion.

Significance

This work is interesting, well done, and described nicely. It highlights how unique and different the kinetochores of kinetoplastid species are, and brings up a number of questions about how these kinetochores are specified and how they function. The structural work is also interesting and well-done. Unfortunately, the work as a whole does not make any strong mechanistic conclusions, leading to a somewhat dissatisfying conclusion.

The work could be significantly strengthened if the authors were able to make a direct functional conclusion about the roles of the Znf regions of KKT2 and/or KKT3, for example detecting DNA binding in vitro, or detecting a specific pairwise interaction between this region and another kinetochore protein.

This work will most likely appeal to researchers in the cell division and kinetochore architecture fields, although since kinetoplastids are so unique the link between this work and most other kinetochore work is unclear. This is in a way exciting: we don't yet know much about how these kinetochores relate to other eukaryotes' kinetochores.

My field of expertise is structural biology and biochemistry, with experience in kinetochore architecture and structure.

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

Kinetoplastids have unconventional kinetochores that lack CENPA nucleosomes that normally dictates the position of the kinetochore in most other eukaryotes. Marciano and colleagues analyse KKT2 and KKT3, two consistutively localized kinetoplastid kinetochore proteins that may contribute to kinetochore positioning on centromeric DNA. They find that in both proteins the central, cysteine-rich domains are sufficient to support centromere localization but that in KKT2 also other domains can do so by themselves. They then obtain crystal structures of the KKT2 central domain from bodo saltans and show it consists of 2 Zinc-finger structures (Znf1 and Znf2) of which the first is conserved in Perkinsella. Mutations of Znf1 and Znf2 in KKT2 and homologous mutations in KKT3 show that Znf1 is crucial for centromere localization and viability, while Znf2 is dispensible for both.

The paper presents a pretty straighforward characerization of functional domains in KKT2 and KKT3 with respect to centromere localization. The authors nicely show a unique Zn-finger structure (Znf1) of KKT2 and show it is crucial for localization. The study does not end up delivering an answer to the questions posed in the manuscript, namely how centromeres and therefore kinetochores are specified in kinetoplastids. The paper could do with some attempts to get to this, based on the presented data. For example, does Znf1 bind centromeric DNA, does it bind nucleosomes, is it essential for recruiting the other KKTs, etc.

The experiments are in general well presented but some could be better controlled:

• localization of KKT2 and KKT3 mutants is never verified to be centromeres, we have to believe the dots in the DAPI region are centromeres.
• in some cases mutants are made in full-length (FL) background (viability, sometimes localization), but in other cases only in isolated domains. The former should be done for all assays. This is also important to show that central domain of KKT2 and KKT3 is necessary for localization.
• The data of F2 are interpreted to mean that PDB-like domain and middle region get to kinetochores by binding transient KT components, even though KKT2 itself is constitutive. That interpretation would really be strenghtened by showing the KKT2 fragments are now transient also.

Significance

The paper presents a pretty straighforward characerization of functional domains in KKT2 and KKT3 with respect to centromere localization. The authors nicely show a unique Zn-finger structure (Znf1) of KKT2 and show it is crucial for localization. The study does not end up delivering an answer to the questions posed in the manuscript, namely how centromeres and therefore kinetochores are specified in kinetoplastids. The paper could do with some attempts to get to this, based on the presented data. For example, does Znf1 bind centromeric DNA, does it bind nucleosomes, is it essential for recruiting the other KKTs, etc.

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

Although most studied eukaryotes display similarities in their overall kinetochore structures to mediate chromosome segregation, kinetoplastid species display highly divergent kinetochores with no clear relationships to canonical kinetochore components. Prior work from the Akiyoshi lab and others has identified kinetochore proteins in Trypanosomes and other kinetoplastids. The identification of these proteins has provided a toolkit to begin to reveal the features that guide the function and assembly of these structures during chromosome segregation. Despite differences in protein composition, all kinetochores must display key properties including their ability to bind to both microtubules and chromosomal DNA. This paper focuses on the mechanisms by which kinetoplastid kinetochore components are targeted to centromere regions, an exciting question due to the apparent DNA sequence-independent nature of these associations. In other eukaryotes, this sequence independent association is specified through the action of histone variants. In contrast, it is unclear how DNA interactions occur in kinetoplastids.

This paper begins by reasoning that the proteins responsible for DNA interactions and defining the location of the centromere would localize persistently to centromeres. Thus, they focus on two constitutively localized proteins with sequence similarity to each other, KKT2 and KKT3. The authors analyze these proteins using a combination of domain analysis to test the localization requirements for these proteins, mass spectrometry analysis of interacting proteins, mutational analysis to test specific residues for localization and function, and most importantly determination of the structure of a kinetochore targeting domain, which reveals a zinc finger structure. The structural work in particular is both interesting and reveals a feature of these proteins that was not obvious based on initial sequence analysis. Overall, this paper appears to be carefully executed, rigorous, and well controlled, but could benefit from additional experiments that would extend the impact of their findings.

1. From the information presented, it seems like there are only two possibilities to explain the role of the zinc finger domains in directing centromere targeting. First, this could mediate a protein-protein interaction. The authors attempt to assess this using their mass spec experiments, but this does not absolutely rule this out as this interaction may not persist through their purification procedure (low affinity or requires the presence of DNA, such as for a nucleosome). Second, this could reflect direct DNA binding by the zinc finger. Although the existing paper is solid and highlights a role for the zinc finger domains in the localization of these proteins, it would be even better if the authors were to at least assess DNA binding in vitro with their recombinant protein. Comparing its behavior to a well characterized DNA-binding zinc finger protein would be powerful for assessing whether direct DNA binding could be responsible for its centromere localization.
2. The code for KKT2 and KKT3 localization is complicated by the multiple regions that contribute to their targeting. This includes both the zinc finger domain that the authors identify here, as well as a second region that appears to act through associations with other constitutive centromere components. Due to this, it feels that there are several aspects of these proteins that are incompletely explored. First, the authors show that the Znf1 mutant in KKT2 localizes apparently normally to centromeres, but is unable to support KKT2 function in chromosome segregation. This suggests that this zinc finger domain could have a separable role in kinetochore function that is distinct from centromere targeting. Second, although the authors identify these minimal zinc finger regions as sufficient for centromere localization, they do not test whether this behavior depends on the presence of other KKT proteins. This seems like a very important experiment to test whether recruitment of the zinc finger occurs through other factors, or whether it could act directly through binding to DNA or histones.
3. Based on the description of kinetoplastid centromeres that the authors provide, it is actually unclear to whether these are indeed sequence independent. The authors state that "There is no specific DNA sequence that is common to all centromeres in each organism [Trypanosomes and Leishmania], suggesting that kinetoplastids also determine their kinetochore positions in a sequence-independent manner." However, it remains possible that there are features to this DNA that are responsible for defining the centromere. In principle, enriched clustering of a short motif that may elude sequence comparisons could be responsible for specifying these regions. It would be helpful to use caution with this statement, and I would also encourage the Aikyoshi lab to test this directly in future work, such as using strategies to remove a centromere or alter its position.
4. It would be helpful to provide a schematic of kinetoplastid kinetochore organization based on their studies to date (possibly in Figure 1) to provide a context for the relationships between the different KKT proteins tested in this paper.

Significance

This paper provides a nice advance in understanding the molecular architecture and functional organization of kinetoplastid kinetochores. As these remain understudied, this work is valuable for revealing the chromosome segregation behaviors in these medically-relevant parasites. In addition, due to the divergence in overall kinetochore function from other eukaryotes, this work will help provide insights into the logic by which kinetochores function and are organized. The existing paper represents a solid advance in understanding the structure and requirements for KKT2 and KKT3 kinetochore targeting through this novel zinc finger domain. However, conducting some of the additional experiments made above, such as testing DNA binding and the requirements for other KKT proteins for zinc finger localization, would allow the authors to make stronger statements and a more impactful advance.

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>

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>

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>

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

---

Author Comments:

Testing Reviewer Comments

• with
• a
• bulleted list

1. and
2. a
3. numbered
4. list

Testing Author rebuttal

• with
• another
• bulleted
• list

1. and
2. another
3. numebred
4. list

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

This reviewer did not leave any comments

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

Learn more at Review Commons

---

Referee #1

This reviewer did not leave any comments

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

---

Author Comments:

Testing Reviewer Comments

• with
• a
• bulleted list

1. and
2. a
3. numbered
4. list

Testing Author rebuttal

• with
• another
• bulleted
• list

1. and
2. another
3. numebred
4. list

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

This reviewer did not leave any comments

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

Learn more at Review Commons

---

Referee #1

This reviewer did not leave any comments

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

---

Author Comments:

Testing Reviewer Comments

• with
• a
• bulleted list

1. and
2. a
3. numbered
4. list

Testing Author rebuttal

• with
• another
• bulleted
• list

1. and
2. another
3. numebred
4. list

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

This reviewer did not leave any comments

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

---

Author Comments:

asdasasdasd

asdasd

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

This reviewer did not leave any comments

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

Learn more at Review Commons

---

Referee #2

Points of Critique

I would urge them to reposition as a descriptive study rather than making too many grand statements about drug sensitivities.### Other Comments

Van Alphen et al. describe phosphotyrosine proteomic profiling of a panel of AML cell lines and two patient-derived AML samples. Subsequent analyses attempt to identify potentially targetable kinases and pathways that would be considered vulnerabilities for drug treatments. Some hypotheses from these analyses are tested by drug treatment. The patient-derived samples are analyzed as a proof of concept and compared to the cell line profiles.

I think that this study is a potentially valuable resource, but might be served better if positioned more as a catalog of pY signaling in AML and less as a drug-targeting effort. The analysis graphs and charts are quite handy, and perhaps they could be served on a more interactive website that could be expanded in the future as the authors continue similar studies. However, I find that many of the conclusions are overstated and that some of the internal logic is inconsistent. Further, while the bioinformatics analyses are carefully planned and well intentioned, I was confused by the inconsistent quantitative metrics used in different parts of the manuscript and curious why a more modern isobaric labeling technique wasn’t used to compare among this relatively small panel of cell lines. Below I offer several points that could be addressed to help to improve this manuscript.

1. The authors claim that sensitivity to drugs predicted by their inferred kinase activity metrics “validates” their predictions. However, all of the drugs tested have demonstrable polypharmacology. How can they be sure the targets being hit that cause loss in viability are the same ones that they have predicted? Also, it seems curious that they only tested quizartinib in predicted FLT3-GoF lines and ponatanib in inferred FGFR-GoF lines. How do we know that these drugs just don’t kill all lines? It would be more convincing to show some lines where these drugs did not cause loss in viability.

2. Along these same lines, phosphoproteomics seems like a long path to identify vulnerabilities in cancer cell lines. Screening drugs on cell lines is cheaper and easier. Indeed, the CTD2 project has a drug screening arm (as did CCLE), and new Cancer Dependency Map screening is enlarging these screens. These projects also have more comprehensive genetic characterization of the cell lines involved. If the logic is that the drug treatments “validate” the predictions of aberrant kinase activity, couldn’t the drug screening be used to make these predictions, to be later validated by phosphoproteomics? Perhaps screen all TKIs against the AML cell lines and see what common targets emerge?

3. I think that claiming that the patient results match the cell line results is a bit of selective interpretation. The first thing I was drawn to is the whopping amounts of MAPK14 phosphorylation identified by their analysis in these samples. MAPK14 - a.k.a. p38 MAPK alpha - also has drugs that target it. If you were making a therapeutic hypothesis, wouldn’t you start with a p38 inhibitor rather than a FLT3 inhibitor? You already knew that the patients had FLT3-ITD, so you’d probably be starting there anyway. While MAPK14 is found to be phosphorylated in the cell lines as well, the degree to which it rises to the top in the patient samples is striking. This also illustrates an issue with drawing inferences from cell lines to real patient samples.

4. On p.15, you state: “P15: “Kinase activity ranking analysis of the FLT3-ITD mutant cell lines MV4-11, MOLM-13, and Kasumi-6, and the V617F JAK2 cell line HEL showed a lower ranking of FLT3 and JAK phosphorylation than expected based on their mutation status, compared to other kinases (position 6-10). Interestingly, other high-ranking kinases in these cell lines were generally located downstream in the FLT3 and JAK2 cellular signaling hierarchy, thereby still implicating FLT3 and JAK2 as primary suspects of driver activity.”

Perhaps this demonstrates the limitations of the approach? Genetics says FLT3 and/or JAK2 are mutated, proxy measurements say its activated, but the way you are estimating direct activity is not so great? Would a targeted panel on activation loop sites be better?

1. Figure 6 is an interesting analysis but how it was generated is unclear. Again, polypharmacology of the drugs make it hard to interpret. Are the graphs for all potential targets? Just some? Weighted by in vitro IC50 concentrations and/or binding affinities?

Minor points:

IC50 is inappropriate nomenclature here, which describes the concentration at which 50% of an enzyme’s activity is inhibited. The authors should substitute EC50 throughout, as they are referring to the concentration at which viability is decreased by 50%.

Ibrutinib is described as a pan-KI - this is confusing and misleading. It is a pretty specific BTK inhibitor.

Please update sup table 5 to show the exact nature of the mutations. Also, why does Kasumi-6 have two different (presumably) allelic ratios for FLT3 (presumed ITD?)?

Methods - p7 “as described elsewhere” - reference needed?

The statistical rationale is not well explained.

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

Other Comments

This manuscript describes phosphotyrosine-focused phosphoproteomics for 16 AML cell lines to obtain molecular profiles of pY towards personal therapy using proper molecular targeting drugs. This is the revised (re-submitted) version and the authors added new data analysis especially on the relationship between kinase-ranking parameters and drug IC50 profiles to kinases. These results indicate the current progress and limitation of phosphoproteomics combined with genomics data and the related computational tools. Overall, the precise descriptions of the experimental procedures as well as the high quality of the experimental datasets are quite useful for researchers in this field. This manuscript should be published after some revisions shown below:

(1) This research group just published a paper on kinase ranking using phosphoproteome datasets, named INKA. Mol Syst Biol. 2019 Apr 12;15(4):e8250. doi: 10.15252/msb.20188250 INKA, an integrative data analysis pipeline for phosphoproteomic inference of active kinases.

INKA seems quite similar to the approaches in this manuscript. The authors should mention about INKA. Especially the parameters in Figure 6 should be described clearly whether these are the same in INKA or not.

(2) Figure 1 as well as Abstract and the first section of RESULTS: the numbers of phosphotyrosine sites and phosphotyrosine peptides should also be described in addition to the current description.

(3) Figure 2: the color for mutation is overlapped with colors for the heatmap. To avoid the misunderstanding, the authors should use the different colors.

Note: The peer reviews in Peerage of Science are judged and scored for accuracy and fairness by other reviewers. The Weight -value indicates that, relative to the best review (Weight=1.00)

---

Review by Peer 4429 (Weight = 1.00)

Introduction: The manuscript evaluates the use of genomic prediction in rice to prevent the accumulation of arsenic in rice grains. This is a food safety issue. Genomic prediction could be an appealing strategy for breeding of rice varieties less prone to accumulate arsenic in grains. Genomic prediction could bridge between current strategies based on land management (genetic improvement is cumulative and permanent) and recently proposed genome editing (for which target causal mutations need to be identified first).

Merits: The study seems original in its proposal of genomic prediction for this particular problem. The authors contextualize in the Introduction the potential interest of genomic prediction against other strategies, including management and genome editing.

The manuscript is quite broad in scope, as it tackles (1) genetic variation of the traits, (2) genome-wide association study GWAS, and (3) genomic prediction.

Despite the low number of significant associations in the GWAS, some of the ones that are detected have annotation terms that could make them interesting candidates for further study.

References are appropriate for the study.

Critique: Because it covers so much ground, the manuscript is quite long and dense. I think it could be softened a little in some sections. Instead it feels a little bit rushed when it comes to genomic prediction, considering that several prediction methods and strategies are used.

While genomic prediction is contextualized against other strategies in the Introduction, some of the results are not discussed as compared with other strategies. For example, there could be a greater effort to discuss the results of GWAS in light of the identification of targets required for genome editing (building on L327-336). There should also be a greater effort in discussing the several methods used for genomic prediction and potentially how genetic architecture from GWAS may help explain the differences between methods; for instance, if genomic prediction is concluded to be the best strategy, which method of all tested is recommended?

I am not totally comfortable with the interpretation that the authors make of the comparison between phenotypic and genomic selection (L346-362). Phenotypic selection is producing 5 to 10% more genetic gain than the genomic (L344-345). This is a large difference that cannot be disregarded. The authors also claim that at equal cost of phenotyping and genotyping, genomic prediction would be preferred. While I agree with the logic that genomic data has the additional benefit that it can be applied to any trait, phenotyping of each of these potentials traits would also be needed with a certain routine to re-train the predictive equation. The authors acknowledge to some extent these points but, because overall phenotypic selection seems to be a better strategy for the specific case of arsenic tolerance and because the suitability of genomic prediction is established as dependent on genotyping costs, the title and conclusions seem a little bit misleading.

It is clear that the paper was written with the Materials and Methods after the Introduction and it was later moved to the end of the manuscript. As a consequence, abbreviations are not properly defined when first read.

Discussion: The manuscript offers a broad perspective on a topic of interest, affecting food safety, and proposes a sensible approach to mitigate it. The study is very detailed about the genetic variation of the traits and GWAS results and overall tackles all important points of discussion. However, it is slightly more vague on the genomic prediction section: several methods and strategies are tested but not described in the Methods section with enough detail and not thoroughly discussed. The authors conclude that genomic prediction would be a more suitable strategy to breed for arsenic-tolerant rice compared to other marker-assisted breeding strategies. However, it seems from the results that genomic prediction still underperforms compared to phenotypic selection and this should be put into context too. This manuscript contains some interesting research and it could be suitable for publication, but some revision is recommended as indicated.

Additional Comments for Authors

• L38: Be explicit. Mitigation of what?

• L59: Please define "Aus genetic group".

• L96: Be explicit. Which three traits?

• Also L96: The distributions in Fig 1 seem to depart from a normal distribution.

• Genomic prediction results: There is an n>p problem here, considering that 100 to 300 accessions but ~20,000 markers were used. Bayes A (one of the methods highlighted as most promising) fits all the markers in every iteration; Bayes B and C fit a pre-defined proportion of markers "pi" (could the authors specify to what value that parameter "pi" was set?); etc.

• Revise English. Several typos and minor grammar errors.

---

Note: The peer reviews in Peerage of Science are judged and scored for accuracy and fairness by other reviewers. The Weight -value indicates that, relative to the best review (Weight=1.00)

---

Review by Peer 1755 (Weight = 1.00)

Introduction: This paper presents a Bayesian model of mating in a fish, that combines behavioural data on encounters and matings with genetic parentage data. It contrasts this model with classical analyses that use only particular facets of these data.

Merits: In my opinion, this paper's most important merits are:

That the model makes conceptual sense, and is presented in a way that is fairly easy to follow.

That the authors share the model code and data. This will make the model a lot more useful for other researchers.

That the paper is well written.

Critique: Despite this, I think there are things that could be clarified or improved:

1. There seems to be a considerable skew in the reproduction data. This is expected, but this comes with a risk violating the assumptions of common statistical models. Does the models used adequately capture this? In particular, the correlation coefficients (Figure 1) must be largely driven by single influential data points.

2. Given the above skew and structure of the data and that the model results extrapolates quite a bit from what was observed, it would be nice to see more through checks and discussion about the validitiy of the model. How well the model can reproduce features of the data? The posterior predictions in Figure 4 seem to indicate that the model fits data rather poorly? But I may be mistaken, and the manuscript does not interpret these results much.

3. I got the JAGS model to run with only minor editing (that is, moving the data generating code to its own file). However, I can't, using the data in the script, recover the scatterplots and Pearson correlations displayed in Figure 1. I assume my analysis (see attached Sweave pdf output) is wrong somehow, suggesting a need for better documentation so that readers such as myself can understand the data. It may help to clarify what variables are what, which samples have been omitted (from what analyses and for what reasons), and store the data in tabular format in addition to the JAGS input format. It would also be a nice addition to have the code used for running the model and summarising the results -- it would save a user quite a bit of effort without much work on behalf of the authors.

4. The sample sizes for data on releasing of gametes are particularly small. One wonders how much information they contribute? Similarly, both observations (line 248) and modelling (line 305-307) suggest that many encounters were not observed. How does this affect conclusions? This ability to deal with incomplete data is highlighted as a feature of the model. Is there arguments or data that show that it is successful?

5. In the Introdution and Abstract, one of the motivations for this approach is to capture effects of interactions of the phenotypes within a pair. But then, "Unfortunately our dataset is too small to properly infer the effect of interaction" (line 428-429). First, previous the focus on this unused feature of the model seems misplaced. Second, it is not clear when a dataset is too small and how you know that (presumably by trying a model not shown?).

6. I think this paper would benefit from more illustration. Figures 1 and 3 are hard to read with small differently shaped symbols, line patterns, and overplotting. I would suggesting making separate plots for males and females to alleviate some of the clutter. Figure 1 b is particularly unreadable. The plots of posteriors are fine, and probably should be in the paper, but I think they should be supplemented with some descriptive graphics that give a feel for the structure of the data and the behaviour of the fish. I would even love to see some visuals of fish mating, maybe stills from the video recordings (or even a supplementary video). Of course, this may be limited by space requirements of the target journal, or nor to the author's taste. But I think you underestimate how cool some of these things are, especially if you aim for a wide audience not well versed in fish mating research.

Discussion: This is likely beyond the scope of this paper, but I feel that a lot of the questions about the model -- does it work on small datasets; does it successfully account for unobserved encounters; how does its parameters relate to the "classical" measures of sexual selection -- could better be answered with simulated than with real data. I sympathise the use of real data: a good biological example is a lot more convincing to biologists than simulations. However, I feel that there are often too many uncertainties in comparing methods on real data. Results of different methods differ, like the "classical" and the new analyses in this study. But which are right?

Additional Comments for Authors

1. The paper would benefit from a two sentence explanation of opportunity for selection, what it measures, and the distinction between opportunity for selection and opportunity for sexual selection.

2. L8-10: The opening of the abstract sets up the paper to be rather technical, jumping directly into marginal sums of matrices. I think you may want to rethink that approach if the goal is too reach, as the author message said, "a wide audience of ecologists and evolutionary biologists".

3. For the same reason, I'd advice against the introduction of a 3-dimensional array on line 34. Even if that is mathematically correct, it is immediately going to be summed to the a parental table. Therefore, the 3-dimensional structure doesn't really contribute much, except act as an obstacle to mathematically less savvy readers.

4. L48-49: "strong link" could be made more precise.

5. Line 123-124: "The experimental setup is the one used in the "constant environment" treatment in Gauthey et al. (2016)." What is the relationship between this work and Guthey et al 2016? Can this be made clearer?

6. Lines 226: "po" is not defined in this section. I think the manuscript would benefit from being checked an extra time for mathematical symbols, when they are defined, how they are referred to, and if they can be spelled out in text to help the reader.

7. Line 270: "Model output" is not a very informative subtitle. I'd suggest dividing the Results into one subsection on the data set, one on the "classical" analyses of sexual selection, and one on the model.

8. Some of the chocies about model structure (specifically, use of informative priors) is discussed in comments in the model code, but not in the Methods. They should be in the Methods too.

---

Review by Peer 1765 (Weight = 0.88)

Introduction: This paper aims to solve a long-standing issue in sexual selection studies in natural populations: that genetic and behavioural data tell us different things about separate stages of sexuals selection and, therefore, often focus on different processes in sexual selection. While behavioural data tend to focus on mate sampling and mate choice, genetic data provide evidence on the resulting mating/reproductive success. This paper makes an important step in trying to combine both types of data in order to analyse the complete process of sexual selection. Such a tool could substantially advance the field of sexual selection in natural populations. I was very enthousiastic about this approach, until I arrived at Figure 4, which shows that the predictions from model the authors suggest does not correlate at all with the observed data from their case study, suggesting the model is possibly very well thought through, but does not represent the data well. Without empirical evidence, I do not see any reason to put the results of the model above those of the classical methods.

Merits: The paper describes the model used in a way that is mostly very clearly understandable for non-modelers, which is important for the general use of the proposed method. Moreover they include a case-study which very nicely links the theory to experimental data.

Critique: The suggested model provides different results from more classical methods of analysing the data. The authors then go on to defend the model as a better way to analyse the data, because they find different results. However, they do not provide evidence that the results from the model fit the data better than the results from the classical analyses. In fact, Figure 4 shows that the model is actually rather bad in predicting observed encounter rates, gamete releases and offspring numbers, because there seems to be no correlation whatsoever between observed and predicted data. For example, many females that did sire large numbers offspring were not predicted to have any offspring according to the model (Fig. 4c). This is not discussed in the paper. I do commend the authors for testing their model on a case study, and combine a theorethical appraoch with an experimental one, but the difference between predicted and observed data should be discussed. The authors could compare the model predictions to the predictions from the classical analyses and see which analyses fit best with the observed data.

Terminology: Encounter rate is a term that is generally reserved for random events depending on population density and sex ratio. However, the way it is used in the case study (which is certainly the most practical for field observations) includes a certain effect of attraction. In most species, males and females do not generally end up close to a spawning ground/ nest without being attracted by some aspect of the individual or this particular nest. The authors are likely aware of this, because they test for an effect of female size on encounter-rate. The fact that they do not find such an effect does not exclude that their may have been attraction to other characteristics of the female or the nest-site. Therefore, I would suggest to use another word for encounter (for example inspection or visit) to avoid confusion between an event where individuals have likely already been attracted to each other (as used in the case study) and a random "encounter". The latter is, however, impossible to quantify in the field, because it is generally impossible to spot whether two individuals have noticed each other and I see no reason to include it in the model.

Discussion: The paper addresses a very important issue in the study of sexual selection: how to combine behavioural and genetic data to study the strength of sexual selection. As the authors rightly argue, both types of data omit important processes in sexual selection and very few studies manage to get both types of data for all (or even most) mating events. The model they suggest would make use of incomplete behavioural and genetic data to explain the underlying processess. Such a model could provide an important tool for sexual selection studies. However, the case study the authors provide suggests that the model is not very good at predicting real case scenarios. Therefore, the autors should investigate how the model could be changed to reflect their experimental data. Doing so would provide an important paper that would be very valuable to the field.

---

Review by Peer 1758 (Weight = 0.85)

Introduction: This manuscript offers a statistical alternative to classical sexual selection gradient analysis by using Bayesian inference that allows accounting for male and female effects simultaneously. Furthermore, the authors highlight that mating success is generally underestimated because it is based on the genetic assignment of offspring. The authors use their own data on the mating behaviour and reproductive output of brown trout to compare the results from classical selection analysis with their Bayesian model and find differences between the two.

Merits: This manuscript is relevant because it highlights limitations of classical sexual selection gradient analysis, and offers a statistical alternative to empiricist with suitable data. I have the following suggestions, which I hope will be useful in revising the authors' original contribution. Also, I welcome that the authors made their research transparent by adding their data and code. However, I want to make clear that I could not review their code because of incompatibilities with JAGS and my software. ​

Critique: The authors statistical alternative is motivated by two shortcomings to (a) account for the interdependence of females and males in sexually reproducing species and (b) getting a grip on the copulatory behaviour instead of inferring it from offspring data. Whilst I agree that (b) is pressing, (a) depends on the mating systems, e.g. in strictly monogamous species, male and female identity overlap and fitting both would not be informative or appropriate for the analysis of sexually selected individual phenotypic traits. Hence, the applicability of the authors' model would profit from information on its suitability for different mating systems, i.e. expand on "a variety of biological systems", l24, in the discussion. Also, the authors approach also relies on empirical data. In other words, the best model does not change that if mating success lacks behavioural observations, and it usually does, we can only make incomplete inferences. In my view, the main contribution of this manuscript is thus to serve as an important reminder of the complexities at play and the importance of comprehensive data collection, rather than a new tool for measuring sexual selection. Also, the pitfalls and shortcomings, (e.g. bias in stochasticity, what is the null model, operational sex ratio) when measuring sexual selection have been comprehensively illustrated here (Klug, Heuschele, Jennions, & Kokko, 2010) and here (Jennions, Kokko, & Klug, 2012). So, I recommend a more inclusive portrait of the matter and attuning with published jargon (e.g. Table 1 in (Klug, Heuschele, Jennions, & Kokko, 2010).

• I advocate that the full results of the linear regression analyses as well as the alternative JAGS model are presented in table format in the main text. Results in the supporting information get missed easily, and plots cannot substitute full estimates.

• The authors could expand more on discussing their most interesting finding, which is the discrepancy between their results using classical regression analyses and Bayesian analysis.

Discussion: This manuscript is motivated by two shortcomings of the classical sexual selection gradient analysis. I agree with the relevance of one of them (i.e. measuring mating success) and yet argue that the relevance of accounting for the additive effects of the sexes for reproductive success is highly dependent on the species mating system, which the authors should address. I also think that the authors should make clearer that their analysis still depends on empiricists collecting data on mating success. I welcome the authors approach to use their own data to compare whether body size of male and female brown trout might be sexually selected. If the authors revise the current version, their manuscript will serve as an important reminder of what to look out for when analysing potentially sexually selected traits.

References Jennions, M. D., Kokko, H., & Klug, H. (2012). The opportunity to be misled in studies of sexual selection. Journal of Evolutionary Biology. http://doi.org/10.1111/j.1420-9101.2011.02451.x

Klug, H., Heuschele, J., Jennions, M. D., & Kokko, H. (2010). The mismeasurement of sexual selection. Journal of Evolutionary Biology. http://doi.org/10.1111/j.1420-9101.2009.01921.x

Schlicht, E., & Kempenaers, B. (2013). Effects of social and extra-pair mating on sexual selection in blue tits (Cyanistes caeruleus). Evolution, 67(5), 1420-1434. http://doi.org/10.1111/evo.12073

Additional Comments for Authors l14: be clearer on "costly" or delete because costs were not measured

l27: add or consider selection gradient, see Table 1 in Klug et al 2010

l44: ambiguous "to do so". Which of the indices exactly?

l52 infertile not unfertile

l53 reference "cost of reproduction"

l64 reference costs

l65 back up the claim of "are essential to understand..."

l68 better name the "fourth definition"

l88-89 reference

l93 define "a pair", e.g. socially monogamous? This could be an opportunity to introduce the mating system you want to target

l109-111 reference?

l113-115 reference?

l116 in brown trout? Please add citation

l120 "a" semi-natural...

l120-123 split into two sentences to improve readability, e.g. This period represents the trout...

l124: chemically communicated?

l129: highly female biased, which might be biological meaningful or a catching bias, please explain. Plus this skew in adult sex ratio will affect the variance in mating success, i.e. "chance variation in mating success is higher when there are fewer potential mates per individual of the focal sex" (Jennions et al 2012), this affects both your statistical approaches but it nowhere mentioned

l132 how did you sex? Molecularly?

l145: one or multiple observers? also "taken" not "took"

l148 any proof? repeatability tests? references for the claim?

l149 say how you dealt with the 30% for analyses

l150 rephrase "the zone", e.g. female nesting/egg release site, etc.

l156 consider "spawning" or gamete release instead of copulating

l159 "degree day" reads misplaced, only use estimate of time after spawning

l172 its

l186 consider making clearer that zero's were included

l247 depending on where you want to submit avoid fish jargon: "redd"

l249 give output of all linear regression analyses in table

l271 I suggest moving these to the main text

l278 why not report Credible Intervals instead of SDs? Also, SDs show high uncertainty in estimates, which should be addressed in the discussion

l333-4 reference

l336 rephrase "to account for..."

l335 give time unit, e.g. over the course of the experiment

l336 Comment: I disagree because sexual selection is commonly referred to as the opportunity for evolutionary change, which is the variance in relative fitness and should consider all reproductively mature adults, hence should be measured among individuals that do and do not interact/mate. Especially the latter is usually omitted, but ignoring unmated individuals in a population will automatically inflate the variance of the successful subset (see also (Schlicht & Kempenaers, 2013)).

l418-19 rephrase, unclear

Plots: General comment: It might be the pdfs but the quality of plots is low and generally offsetting the raw data a bit, e.g. jittering would help viewing individual data points

---

Review by Peer 1761 (Weight = 0.67)

Introduction: The authors point out how the study of mating systems only using behavioural observations or genetic data usually fails to explain accurately the breeding processes and reproductive outcomes, as well as their relationship with sexual selection features.

They propose a model that combines both behavioural and genetic data, and a phenotypic trait linked to sexual selection, using brown trout as model species.

Their model includes several breeding variables behavioural and genetic, and it very adaptable as is able to incorporate other environmental or biological variables if needed.

They show how genetic and behavioural results analyzed separately may differ. Also, how the results from their model and the classic regression analyses to analyse this data also differ, and so, they aim to explain why.

Merits: The model they have built seems flexible enough to be adapted to multiple taxa and systems.

Critique: There is no reference at all about ethics permissions to perform the described experiment. I am quite shocked about this since high numbers of individuals from a wild population were killed.

There is no mention on the conservation status of the species, the permits obtained to carry out the capture and experiment, the effect of the capture system on the ecosystem, or the explanation/justification for the use of lethal methods.

For example, I find electrofishing highly non-targeted and I wonder how was its impact on other non-target fish (and non-fish) species. I believe that assembling a team of fishermen to get the same number of adult specimens would be easy enough to arrange.

My point is not whether the methods were ethically acceptable or not (that is for the journals' ethics committees to decide) but to, at least, justify and explain their use.

Model testing: I understand that in ecology studies usually researchers don't get all behavioural or all genetic data, and that is what the models try to compensate for. However, when testing models in a biological system the ideal situation is to work in a system where almost all information can be collected (ussualy under lab conditions), build a model with all that information, and then subsample the data (as to simulate a real ecological study) to test the model performance.

In this study, however, the initial sampling for the data is quite small, specially for behavioural observations (30min/day). Then, the results from the model are quite different from the results obtained from more classic approaches. The authors offer some hypotheses to explain these differences, but they can't be really tested to see whether the authors' model results are better in explaining the system or not.

All that said, I have to admit that I lack the mathematical background to fully understand and evaluate the model design and performance, and a more qualified researcher should do that.

Discussion: Although the experimental approach to test the validity of the model predictions could have been better, their attempt to combine behavioural and genetic data in mating system studies and relate it to sexual selection is an important step forward in the behavioural field.

Hopefully, more efforts like this will be made to reconcile both aspects of the study of mating systems that rapidly changed from behavioural observations only to genetic analyses only.

---

Review by Peer 1773 (Weight = 0.51)

Introduction: In accordance with traditional approach to estimate the effect of sexual selection on phenotypic trait the number of mates should be regressed on a target phenotypic trait in a separate model for each sex. Such analysis ignores common investment of the sexes into mating success. The authors propose a new approach, which allow combining behavioral and genetic data, thereby enabling to gather information through the successive processes of encounter, gamete release and offspring production.

Merits: The new approach accounted for the three-dimensional structure of the data: males, females and mating occasions. This allowed a qualified definition of mating success and disentangling the joint effects of male and female phenotypes on the different components of reproductive success. Three important features that lack in the traditional approach characterize the authors' model:

1) conditioning of each process (encounter, gamete release and offspring production) on the preceding one,

2) simultaneous estimation of the effect of male and female phenotype,

3) random individual effects.

​The authors tested their model on a brown trout and obtained quite different results for the two approaches.

​The model can be used for a variety of biological systems where behavioral and genetic data are available.

Critique: The model should be tested on a larger sample.

The title of the manuscript is not very successful.

​There is a couple of misprints: p. 7 l. 139 and p. 8 l. 159.

Discussion: This is very important when new algorythms allow to obtain more information from the same set of data. Hopefully, it would be of great importance if the model can be developed to account for real behavior traits in species presenting complex courtship behavior like Drosophila for instance.

[Note: this preprint has been peer reviewed by eLife. The decision letter after peer review, based on three reviews, follows. The decision was sent on 21 May 2019.]

Summary

Masachis, Darfeuille et al. analyse a type I toxin - antitoxin (TA) module of the major human gastric pathogen Helicobacter pylori (Hp). Expression of toxins encoded by Type I modules is controlled by small, labile, cis-encoded antisense RNAs and often also by complicated mRNA metabolism that envolves conserved mRNA folding pathways and/or mRNA processing. Using a combination of elegant and robust in vitro and in vivo methods, the authors first show that that the aapA3/IsoA3 TA system of Hp is regulated in a way very similar to that of the homologous aapA1/IsoA1 system from the same organism (Figs 1 and 2). This initial part of the manuscript sets the stage for the next step, where the authors employ a powerful genetic screen combined with deep sequencing to identify single nucleotide changes that abolish production of the AapA3 toxin (Fig. 3). This principle, which was invented by the authors, is technically robust, intellectually attractive and very powerful, and may yield novel insights that at present cannot be reached by other approaches. In particular, the authors discover that single point mutations outside the toxin gene reading frame suppress toxin gene translation. Focusing on the translation initiation region, they discover two mRNA hairpin structures that, when stabilized by single base changes, reduce translation by preventing ribosome binding (Figs 4-6). They propose that these structures are metastable and form during transcription to keep the toxin translation-rate low, as explained in the model figure (Fig. 7).

Essential Revisions

All of the reviewers thought the quality of the experimental work in the manuscript is outstanding and the conclusions are justified. However, all thought it would be nice to have additional evidence of the proposed metastable structures in the nascent toxin mRNA. While the reviewers understood this might be technically difficult, they agreed that it is worth a try and had the following suggestions.

1) Phylogeny (i.e. nucleotide co-variation in sequence alignments) was previously used to deduce the existence of stem-loop structures not only in ribosomal RNAs but also in mRNAs (e.g., hok mRNAs). Did the Authors consider using this approach to support the existence of the proposed metastable structures in the nascent toxin transcript? This possibility depends on the actual homologous sequences available and is not possible in all cases. If phylogeny indeed supports the existence of the metastable structures, the Authors could look for coupled nucleotide covariations that would support a conserved mRNA folding pathway (that is, one mRNA sequence elements pairs with two or more other elements during the fife-time of the mRNA) . The Authors state in the Discussion that "these local hairpins were previously predicted to form during the co-transcriptional folding pathway of several AapA mRNAs (Arnion et al., 2017)." However, they authors did not explain how these hairpins were predicted. It is worth explaining this central point.

2) Although transient structures are by definition hard to detect, the authors could try in vivo structure probing (DMS) of truncated mRNAs 1-64 and 1-90 to demonstrate the existence of the first and the second metastable structures, respectively.

3) It is preferable to carry out 2D structure predictions on the naturally occurring transcript, not a sub-sequence. 2D structure prediction generated by algorithms such as RNAfold (or Mfold) that are guided by delta-G stability optimisation are sensitive to the sequence context, so the correct sequence needs to be used to be able to draw conclusions. Additionally, the findings presented in Figure 3D could be analyzed a bit further to produce significant, independent evidence for some structure features. Specifically,

Figure 2 caption:

• lines 184 - 186: "2D structure predictions were generated with the RNAfold Web Server (Gruber, Lorenz, Bernhart, Neuböck, & Hofacker, 2008) and VARNA (Darty, Denise, & Ponty, 2009) was used to draw the diagrams."
• Please state clearly whether any of the results of the experimental 2D structure probing were used as input to RNAfold (i.e. as additional constraints to the prediction algorithm).

Figure 3D:

• Please add coloring to the peaks depending on which codon position they overlap (1, 2 or 3) and carefully discuss the corresponding results, also in the context of the 2D structure elements.
• Given that you have a decent number of pair-mutations, analyze them to see whether any correspond to RNA structure base-pairs (and whether any of the pair mutations rescue the base-pair and thus affect the system differently). This would serve as additional, independent evidence of 2D structure probing and predictions.

[Note: this preprint has been peer reviewed by eLife. The decision letter after peer review, based on three reviews, follows. The decision was sent on 17 June 2019.]

Summary

Natural Killer (NK) and the ILC1 subset of innate lymphoid cells share related functions in host defense but have been argued to arise from distinct pathways. Park et al present new evidence challenging this concept. They show that murine Toxoplasma gondii infection promotes the differentiation of NK cells into an ILC1-like cell population which is stable and long-lasting, even after the infection has been cleared. These T. gondii induced cells, unlike Eomes+CD49a- NK cells, are Eomes-CD49a+T-bet+ and therefore resemble ILC1 cells. The authors additionally show that their differentiation involves Eomes down regulation and is STAT-4 dependent, However, in common with NK cells and distinct from ILC1 the T. gondii induced "ILC-like" population circulates to blood and lungs. Finally, the authors employ single cell RNAseq to examine the heterogeneity of the major T. gondii induced innate lymphocyte populations and their NK vs ILC relatedness as assessed by gene expression. Together, their observations establish a previously unappreciated developmental link between NK and ILC1cells in the context of infection.

The 3 reviewers and editor agree that this is an important contribution that sheds new light on the developmental relationship of NK and ILC1 cells, a scientific issue that has received considerable attention in the innate immunity field. Although extensive, most of the criticisms raised can be addressed by revisions to the manuscript. One additional experiment is requested to provide a missing control.

Essential Revisions

All reviewers had a major concern about how this new population of T. gondii induced innate cells should be referred to in the manuscript. Based on the single cell RNAseq data, these cells (cluster 10) are still closer to NK cells than to ILC1s (Figure 5f and Suppl Fig 4e) despite their loss in Eomes expression and acquisition of CD49a expression. Thus, one could easily think of them as "Eomes negative NK" or "ex-NK" cells rather than ILC1s, and to simply refer to them as Eomes-CD49a+ ILC1 cells may be misleading . For this reason, the authors should modify the title of the paper and change their designation throughout the manuscript. We suggest "ILC1-like" as a good descriptor. In addition, although it is clear that the "Eomes negative NK" cells that are generated during T. gondii infection are transcriptionally and epigenetically distinct from the NK cells in the steady state and NK cells after infection (Figure 7 and suppl Figure 6), these "Eomes negative NK" cells referred to as "T. gondii-induced ILC1s" were not directly compared with classical ILC1s. Based on the single cell RNAseq data, these cells may not express many of the ILC1-related signature genes. Therefore, again, the authors need to be cautious in referring to them as ILC1 cells.

A second concern was that the NK 1.1 depletion shown in Supplemental figure 1 was performed with a PBS rather than isotope matched immunoglobulin control which is considered unacceptable. The authors should repeat at least once with proper control Ig to make sure this is not issue. It is not necessary to repeat entire survival curve just experiments shown in A and B and initial survival to make sure there is no death in controls vs. antibody treated.

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>

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>