In this biorxiv manuscript, Castel and colleagues explore the sexual and sex determination systems of three closely related hydrothermal vent gastropods. Dissections and histological analyses of 276 individuals revealed that two of these species are gonochoristic, while the third one appears to have both males and intersex individuals. Reanalysis of published RAD-seq data shows that all three species have an XY system and that their sex chromosomes probably share ancestry.

This study provides the first insights into the reproduction of animals living in deep marine vents, which have remained poorly studied, and opens new research perspectives on these animals. I found this very interesting. The methods are standard, the results are sound, and the introduction and discussion were enjoyable to read.

I only have minor comments:

- l140-144: This is a bit confusing. Are these new data?

- l145-165: What samples were used to build Stacks' catalog? Line 158 suggests that samples with missing data were removed, but the final data set still contains 499 samples.

- l145-165: Is there a filter for allele frequency? Knowing whether there are singletons in the final data set would be helpful. A bad quality library could increase SNP count dramatically (e.g., the individual discussed in l313 might be better removed).

- l234-237: Theoretically, performing reciprocal blast using the whole RAD catalog would be better to avoid forcing matches.

- l239: Do we know the sex of the reference genome?

- l245: How were multiple hits handled?

- l255: Given that sex assignment proved error-prone, should the samples be described as "putative" males and "putative" females instead?

- l255-265: Sample sizes differed from those in Table 1 and l168.

- l264-265: Sample sizes differed from those found in l267.

- l303-316: Only PC1 and PC2 are shown. Sexes may show strong clustering on other axes. An AMOVA could eventually be used to determine the total percentage of genetic variation explained by sex in each species and to disentangle sex from geography.

- l313: Cropping a PCA plot seems unconventional. Ideally, it would be better to perform PCA without the problematic individual or choose other axes for plotting.

- l345-347: Did all females have the same genotype at each of the 553 SNPs? 

- l361-362: Could this be driven by a few errors in sex assignment or low coverage in some females?

- l469-470: Could high differentiation also lead to smaller Fst values? Less sex-linked RAD loci will be recovered if the two chromosomes are highly divergent.

- l479: "on the multi-locus PCA" -> on the first two axes of the multi-locus PCA

- l483: The percentage of total genetic variation explained by geography is likely higher (there are other PC axes).

-l556: "share the same sex-linked genomic region" -> share some sex-linked genomic regions.

In this manuscript Castel et al analyse a RAD-seq dataset in three species of gasteropodes and study sex determinism of the species. They newly identify hermaphrodite individuals (females also producing male gametes) in A. kojimai. The study is well conducted and the paper is detailed and well written. I am mostly suggesting further analyses for future investigations or slight modification of the text (especially the discussion on A. kojimai XY system).

I am not a 100% convinced of the XY sex chromosomes in A. kojimai because heterozygosity in males is not equal to 1 and I think maybe the discussion should slightly tone down this conclusion. There could be partial X-Y recombination or a recent breakdown of the non-recombining region in A. kojimai. I understand that the observed patterns were not obtained in autosomal simulations, but could the authors please add more discussion on why heterozygosity is not 1 in A. kojimai males? One explanation may be that the non recombining region in A. kojimai is very small and the SNPs detected as "sex-linked" are in fact only partially sex-linked, outside of the non recombining region but strongly linked to it.

I was wondering if the authors would be interested in using the program SD-pop (Kafer et al 2021)? This would allow to statistically test for the presence of sex chromosomes with a BIC approach in each species. I appreciated the simulation approach taken by the authors to validate their sex-linked SNPs, however the simulations did not include a genotyping error nor duplicated genes, which SDpop does.

Are the Y alleles identified in A. kojimai identical to the Y alleles in other species? Said otherwise, do the three species share Y alleles in the RAD loci that are commonly sex-linked? This would be a further argument to say that the species share the same chromosome pair (otherwise the same chromosome region may have become sex-linked independently in each species, although less likely, this is still possible).

I wonder if the data could allow a GWAS analysis to look for variants specific to hermaphrodites? This may help understand how hermaphroditic sex is determined in A. kojimai.

What are the functions of the genes located in sex-linked scaffolds? This analysis could provide sex-determining gene candidates.

Line 580: if selfing was possible in A. kojimai, wouldn't it be detected with higher Fis values? If selfing was not detected, then there must be a mechanism preventing self-fertilization in hermaphrodite individuals. Or maybe the male gametes are not functional?

line 523 identified AS having.

Pradillon et al: Genetic sex determination in three closely related hydrothermal vent gastropods, one of which has intersex individuals

https://doi.org/10.1101/2023.04.11.536409 

General comments

Here Pradillon et al present new data on the sex determination systems of three gastropod species. Gastropod sex determination systems are severely understudied so this is a welcome contribution to the field in that regard! The paper’s strengths lie in the size and power in the dataset, and in the robust analyses performed to find the sex linked loci. In particular I very much like the simulation analyses to contextualise the Fst and Fis results from the real data. While I suggest a few minor improvements for the analyses (below), I have good confidence in the results and conclusions as they currently stand. 

The main weakness of the paper, as I see it, is the framing of the paper around the observation that “sex is the primary driver of the genetic structure” in these species. I suppose this was the finding that turned the authors onto exploring sex determination in these species, but I do not think this result has much value in the paper. The reason being, it confounds autosomal population structure with sex linkage. The most interesting results for the paper (given the title) are mainly the description of the sex determination systems, rather than the relative strength of signal from sex linkage vs geography in population structure analyses. The latter being difficult to interpret as it could be due to either highly differentiated sex chromosomes, or very weak population structure (likely both of course). 

Further, the population structure is perhaps of some interest too here, but in the present state it is obscured by the sex linkage. Now that the sex linked loci are known, they can be removed them, and the population structure analyses can be re-run. This will undoubtedly give more accurate and high-resolution results, perhaps beyond the basin-level.

Thus, my suggestion would be to move Figs 3, S4, S5 and table 1 to the supplementary, and streamline the main text to focus on the sex determination systems. 

Hopefully these are constructive, as they are intended to be! And I look forward to seeing the follow-ups in A. kojimai, there is clearly lots of interesting stuff to be found there!

Best Wishes

Daniel Jeffries 

Specific comments to authors

L. 156 (and supp. mat. ) - The rationale for the high M value makes sense, but I am concerned that this will lead to over-merging loci. To address this concern, perhaps the authors could include a plot of M value vs heterozygosity (which increases when overmerging) in a subset of test samples. 

Also it would be good to see the test results for M=10 for A. kojami given that this is the value you chose for this species. 

L. 162 - Could you give some coverage statistics for the final dataset? And again a plot of coverage vs heterozygosity would be useful to tell whether coverage is high enough for accurate SNP calling. 

L. 170 - Were these pop gene analyses performed after removing the sex-linked loci? Although there are not many, there may be few loci overall in the dataset that are informative for some of these calculations, thus a few loci may have an effect. 

L. 196 - “heterogamous” should be “heterogametic”

L. 204. If I understand correctly, here you are simulating a dataset in order to estimate the expected variance around Fst and Fis for their dataset. I like this approach, and agree that the biases in heterozygosity could certainly help find loci proximate to, but not within the completely sex linked region. My one thought here is that the results of sampling from such a simulated dataset could depend on the distribution of simulated allele frequencies. Did the authors consider simulating this dataset based on the site frequency spectrum of the real dataset? 

L. 240 - could you give a rough estimate of the distance to this reference if it is known? Timetree.org is good for this. 

L. 245 - I couldn’t find it in the Stacks methods descriptions - did you assemble the 1st and 2nd end reads or treat each end as a separate RADtag? Having ~300bp fragments to map will greatly improve alignment rates to distant genomes. 

L. 286 - small typo - in the figure your oocyte abbreviation is “oo” but in the legend you use “ovo” 

L. 313 - when you say the anomalous sample in the A. kojami PCA is “not shown” in fig 3, does this mean you removed it and re-ran the PCA, or just removed before plotting the PCA? The former would be better. Also it looks like you removed 2 samples, not just one here (the blue Manus male in the top right as well). Same for Fig. S5. And do you have any idea why these samples are anomalous? You might check coverage for these samples, this is often the culprit. 

L. 348 - at the risk of seeming too self promoting, perhaps you would find our recent analyses useful here (Jeffries, D. L., Mee, J. A., & Peichel, C. L. (2022). Identification of a candidate sex determination gene in Culaea inconstans suggests convergent recruitment of an Amh duplicate in two lineages of stickleback. Journal of Evolutionary Biology. https://doi.org/10.1111/jeb.14034). Specifically we used some simple probability calculations to assign a p-value to each locus, using exactly the same heterozygosity analyses that you do here. I think that would be useful, especially in A. strummeri & A. kojami. 

L. 377 - Could you add here how much overlap there was between the loci identified by your simulation-based approach and the heterozygosity approach? 

L. 411 - Did you specifically look at how the species-specific sex-linked loci look in species where they are not sex linked? For example, locating the 348 markers that are sex linked in A. boucheti on the A. strummeri plot in Fig. 5 would perhaps give some indication as to whether you missed them due to power issues, or that they are truly autosomal in this species. 

Also, if I understand correctly, you are talking in terms of RAD loci here. So you are able to say that the same genomic regions are sex linked among samples. But are we looking at the same SNPs within these tags? I.e. is it shared ancestral polymorphism that you are seeing, or independently derived sex linked SNPs within the same genomic region. Knowing this might help you narrow down which scaffold/region contains the sex determination locus. 

L. 436 - First, this statement is not accurate, as it is only true for the sex chromosome. Instead the statement should read something like “In PCA analyses, signals of sex linkage within the genome were stronger than signals of geographic isolation”. 

L. 456 - sex chromosomes are “sufficiently differentiated”, or population structure is sufficiently weak. 

L. 469/470 - or that population structure is stronger. 

L. 496-505 - please avoid repeating results in the discussion

L. 511,512 - I’m not sure how much we can infer about the recombination on the sex chromosome in A. kojimai. Given the propensity for intersex individuals, the lack of sex linked identified could just be linked to errors in assigning samples to male or female/hermaphrodites, which would greatly reduce power to identify sex linked loci. Again here it would be good to see how the loci identified in A. boucheti behave in this species. You might even try clustering A. kojimai samples based on the A. boucheti loci.

L. 513-516 - I agree with your conclusion here, that some of these loci could represent polymorphism between Y chromosomes in the population. But keep in mind (and maybe mention) that the most common genotyping error is missing heterozygous calls, which would also give the same pattern. 

L. 530 - As I mentioned for the results section, I think a more detailed comparison of the Fst/Fis analyses against the heterozygosity analysis would be useful here. 

L. 549 - I think “unique” should be “single” here 

L. 552 - do those 3 loci co-localise on a scaffold?