Sex-biased gene expression in fish – a milestone

Sylvestre et al. describe a fantastic dataset for exploring sex-specific patterns of expression in wild-caught sticklebacks, with over 30 individuals of each sex sequenced for RNA of brain, gill and liver. They describe substantial heterogeneity between tissues in the amount and functional nature of sex-biased genes in the three tissues, and confirm that dosage compensation has not evolved in this group in response to Y chromosome degeneration. 

I have some questions below about data acquisition and analysis, which I think are worth clarifying in the manuscript. 

* In response to the specific questions asked by PCI: I found the title a bit disconnected from the article, since gills are a small part of the results, and most of the sex-bias is found in the liver. But that's a matter of personal taste, I also understand why you would chose to emphasize the most surprising result.

* "We collected adult anadromous three-spined sticklebacks from tide pools of the St Lawrence
120 River at Baie de l’Isle verte (48.009961, −69.407070). "
--> Do you have a sense of how old the individuals were? Would age heterogeneity affect the results?

* "In fish, global dosage compensation has rarely been found (Darolti et al., 2019) and
81 we still lack knowledge about the extent of the evolution of dosage compensation in this highly
82 diverse group."
--> this sounds slightly confusing, as it could mean that global compensation has not been found in fish with differentiated sex chromosomes, or that it has not been studied because there are not so many fish species with differentiated sex chromosomes. Could you clarify?

* "Identification of Shared Genes Between X and Y"
--> I could not understand how this was done. How exactly do the custom scripts work? What information is used to know if you have X and/or Y-derived transcripts? Are X- and Y-linked genes already annotated? Could you make the methods more explicit? 

* "We
246 had a percent mapping of uniquely mapped reads to the reference genome ranging from 44.3% to
247 88.61% (median 76.76%)."
--> This seems low. Could you provide numbers for each sample (perhaps this is already in a supplementary table, I could not find these), and perhaps repeat analyses with only samples with good mapping rates as an additional line of evidence that the conclusions are robust? More importantly, are there systematic differences in mapping rates between tissues? These could affect your power to detect sex-bias.

* On a similar note, while the PCA is reassuring, could you also provide correlation values between replicates? Or more generally a measure of heterogeneity within the sexes as well as between them? From the PCA it seems that liver is highly heterogeneous independent of sex, brain is highly homogeneous, and gills are somewhere in between, so sex-bias may just mirror general patterns of heterogeneity.

* Along those lines, would it be possible to color males and females in the PCA?

* Do you have any kind of mininum expression filtering? If not it could be reassuring to see that patterns hold when very low expression genes are removed, or you could justify why this was not deemed useful.

* "After filtering, we identified five overexpressed genes in the liver but none
248 in the gills or the brain."
--> overexpressed relative to what? Could you define this more specifically?

* 418 Sex-biased Gene Expression on Sexual Chromosome Mostly Reflects Gene Loss in Non-
419 Recombining Regions
It would be helpful to know how many genes are being considered in each of the categories (diploid, hemiploid X, hemiploid Y)

* Fig 5: Why do you have all and diploid in the PAR? Should all PAR genes not be diploid?

Review of the manuscript «Sex-biased gene expression across tissues reveals unexpected differentiation in the gills of the threespine stickleback» 

In the manuscript untitled «Sex-biased gene expression across tissues reveals unexpected differentiation in the gills of the threespine stickleback», the authors explore the patterns of sex-biased gene expression in brain, liver and gills during the reproductive period in males and females of the sexually dimorphic threespine stickleback. Confirming previous results, they found that expression in the liver is extremely sex-biased, whereas the brain is not, and that there is a lack of global dosage compensation in this young sex chromosome system. They also found that gills exhibit high levels of differentiation, an interesting result as sex is rarely considered in physiological and ecotoxicological studies of gill responses in fishes. They outline a few genes and gene functions that are particularly sex-biased, and thus likely determinant in sexual dimorphism in this species. Those descriptive  results bring some insights into the role of sex biased expression in sexual dimorphism in sticklebacks. It would be even more apparent if the authors added a brief description, for instance in the introduction, of the sexually dimorphic phenotypic traits observed in the focal species for the non-specialist reader. The manuscript is well written, the topic is well introduced, the used methodology is sound (notably the choice of sampling during the reproduction period, and to use the QuanSeq 3’ UTR technology for sequencing) and the analyses are clear and well conducted (only a few details of the material and method require some clarification) and the conclusions adequately supported by the results.

Comments: 

1. A non-specialist reader would greatly benefit from a brief description of the phenotypic traits that are under sexual dimorphism in the threespine stickleback (for instance in the introduction at around line 84). It would also be interesting to link the gene ontology observations to those phenotypic traits. For instance, you found several GO enrichment terms related to immune system in SBG, are there any indication that male and female show different immune response in some context?

2. I really appreciated that the a priori expectations were clearly stated at the end of the introduction. It would be even better if they were more deeply discussed later in the manuscript: for instance the expectation that the brain would be highly sex-biased during the reproduction period (line 106-107) is eventually not met in the results : what could be some possible explanations?

3. Maybe briefly explain the principle and advantages of the use of the QuanSeq 3’ UTR technology to quantify gene expression level, and how this might impact comparisons with other studies based on classic RNAseq. 

4. In the “Alignment and expression counts” section of the material and method:

-What HTSeq parameters/functions have been used ? 

-Multi-mapping reads have to be filtered out to obtain reliable read counts. However, for X-linked genes still retaining a Y copy outside the PAR, I expect reads could map on both the X and Y copy in males. How was this handled for the quantification of expression of XY gametologs? A related question: how was X and Y allele expression measured for those gametologs?

5. For the non-specialists, briefly explain what the different normalization methods in different analyzes normalize for (within sample/within dataset/between samples comparisons) : cpm, Vst (it should be stated that it stands for Variance stabilizing transformation), and the the (average? Median?) ratio normalization factor.

6. Foie/ Cerveau / Branchie are kept in french in supplementary table 1 and 2.

7. At line 121: under five, seven and ten minutes respectively → after death?

8. At line 119: What year were the samples collected?

9. At line 122: We disrupted samples in

10. At line 182: (lien ?) was kept in the text → https://zenodo.org/records/11477976

11. At line 312: (0.78%) of expressed genes) → (0.78% of expressed genes)

12. At line 361: Is it chromosome V, as stated in the text, or chromosome VII, as stated in Figure 2 legend, which is enriched in male-biased genes?

13. At line 366: total gene expression

14. At line 429-431: Precise the statistics that are stated there.  

15. At line 450-453: “In sticklebacks, hemizigous genes tend to be dosage insensitive, meaning that protein quantities are independent from its expression level (Peichel et al., 2020).” → I would alleviate this sentence, as the Peichel et al., 2020’ result only indicate that genes with a retained X and Y copy were more likely to exhibit haploinsufficiency based on orthology with human genes. 

“This suggests that there is no selective pressure to evolve dosage compensation and is corroborated by the fact that conserved genes are dosage-sensitive and evolving under purifying selection (White et al., 2015).” → I am not sure I understand the second part of this sentence: is it that as Y-linked genes show signs of degeneration, they are likely to be under relaxed selective pressure and thus to be dosage-insensitive?

16. At line 477-484: (p-value=2.10-2)

Sex biased genes are widely found throughout the animal kingdom as potentially the result of resolved sexual conflict. This work inspected the pattern of sex biased genes in gill, brain and liver from the wild population of threespine sticklebacks, a model species for teleost sex chromosome evolution. The species has been previously reported to evolved three recombination suppression events (evolutionary strata) between the XY chromosomes. This work found liver harbors many more sex biased genes than the other tissues, with also many sex biased genes unexpectedly in the gill. Overall I found the work interesting and the conclusions are generally consistent with what one would expect or what the previous results reported. Since there are several works (Kitano et al. 2010, Leder et al. 2010, Kaitetzidou et al. 2022) did a similar study but with different techiniques, or sample sources, I suggest deeper comparison with them, particulary in the regard of lab vs. wild collected samples, as stated in the introduction by the authors. Other aspects that need to be addressed is the techinical aspect of the study, including the definition of sex biased genes, that could lead to very different conclusions. For example, the authors first stated over 5000 sex biased genes in the liver, but then it was reduced to around 300 if the authors applied the same cutoff with one of the previous study, if I understand it correctly. Another contradiction is the feminization pattern in the PAR, that the author found a different result compared to a previous study, due to the usage of autosome or outgroup orthologs, but which one is more appropriate? Finally, a general issue is that the paper is organized in a way that the authors reported the number or general patterns of sex biased genes, then decribe invidiual genes in these tissues. Somehow the reader cannot have a coherent conclusion from these pattern throughout the paper. For example, why gill, a tissue that is not expect to have many sex biased genes show so many in this study? This has not been discussed before and newly found in this study. The author might look deeper into this, just to draw a difference from all previous studies. I have many detailed commments that I marked in the pdf uploaded together with these comments.