After reading the response letter and the revised manuscript thoroughly, I am satisfied that my concerns have been thoroughly addressed. In particular, the discussion of dosage compensation in the introduction is much improved, and the results well justified. I am also impressed with the detail devoted to methods in the supplementary material. I believe that this manuscript has the potential to have considerable influence within the field of sex chromosome evolution.

I have read with attention the revised manuscript of Muyle and co-workers entitled “Genomic imprinting mediates dosage compensation in a young plant XY system”. They studied the expression patterns of the sex-linked genes in Silene latifolia, and showed that the reduced expression from the degenerated Y chromosome is compensated by the upregulation of the maternal X chromosome is both sexes. This scenario is reminiscent of the early steps of the X inactivation in mammals proposed by Ohno. I would like to express satisfaction concerning the quality of the manuscript. I really enjoyed reviewing this paper: it is clearly written, competently analysed I guess (I am not a specialist in bioinformatics), and of high interest.

On the whole, I agree with all comments given by the two previous referees and the PCI Evol Biol recommender, and I am satisfied with most responses from the authors. It is a beautiful story which I want to believe, but I have one main concern. The pattern does not seem to be as clear as the main text suggests.

1) For some reasons, X-hemizygous contigs have been analysed separately, and only discussed in supplementary text. They showed that X-hemizygous contigs have poor dosage compensation, and that the parental origin of the X chromosome has limited to no effect on female X-linked gene expression (i.e., no upregulation of the maternal X). These results are strikingly surprising, as we should expect the opposite pattern, with stronger dosage compensation in the oldest strata, since (i) the Y copies have been lost, and thus the balance of gene products needs to be restored, (ii) there was enough time for dosage compensation to be recruited. Indeed, this is what we observe in the mammalian dosage compensation. The authors give some explanations, with which I mostly agree, but they are not completely satisfactory for me, as these hypotheses alone could not explain the complete absence of dosage compensation in this class of genes. More troubling, although this difference in dosage compensation between X hemizygous and X/Y genes is very important to understand the system, it is only discussed in supplementary data. In the same way, the categorie 0 has been removed from the Figures 1 and 2, but kept in the supplementary figures. As a reader, my feeling is that the authors tried to hide this part which does not fit the general conclusions of the paper.

2) The scenario is not that clear for the validated contigs (no significant upregulation of the maternal X), and again this part is downgraded to the supplementary text.

3) Figure 1 shows an excess of expression of female maternal alleles relative to an outgroup in the last two categories only

Hence, my main conclusion is that the results (and the title) should not be overstated.

Muyle et al aim to address an important question in sex chromosome research, the mechanism by which dosage compensation evolves. They study gene expression patterns across sex-linked genes in Silene latifolia, a plant with young sex chromosomes. They show that Y degeneration is associated with upregulation of the maternal X allele, presumably to compensate for unequal gene dose, and provide support for the steps proposed by Ohno in his theory of dosage compensation. Not only does this research provide further insight into the status of dosage compensation in plants, which have been relatively understudied relative to animals, but also represents an important advance in our understanding of the early stages of sex chromosomes evolution. I very much enjoyed reading this paper; it was very clear and represents an important advance. I have some suggestions that should help to improve the study.

The discussion of the steps associated with the evolution of dosage compensation in the introduction are a little confusing. Currently, the introduction reads that equal expression between the sexes is a form of dosage compensation. This has been debated (Mank & Ellegren 2009 Heredity; Melamed et al 2009 Heredity) however, according to the definition proposed by Ohno, expression equality between the sexes is just a side effect of dosage compensation, and maintaining dose between the X and autosomes is the key step. Unequal gene dosage will select for the upregulation of the single X in the heterogametic sex in order to re-establish ancestral expression levels, and then compensatory evolution in the homogametic sex will restore balanced gene expression. Following this, it would be helpful if the introduction were revised to more clearly outline these steps. In addition, it might be useful here to mention briefly why unequal gene dose between the X and autosomes is thought to negatively affect fitness.

It would be useful to have information in the main text about the outgroups and when they diverged from S. latifolia. When did the sex chromosomes in S. latifolia evolve? This would also be helpful in understanding why the authors averaged expression level across the two outgroups – are they equally diverged from S. latifolia? It is difficult to interpret how this may have influenced the results. Can the authors show that expression is S. viscosa and S. vulgaris is highly correlated for all genes and for Y/X genes with Y degeneration?

Figure 1 and Figure 2 appear to be missing. However, the supplementary figures were very clear and detailed. There appears to be a bracket missing on the Y axis.

The difference in dosage compensation between X hemizygous and X/Y genes is very interesting but only discussed in the supplementary text. I would recommend moving this to the main text and discussing there the potential dosage insensitivity of X hemizygous genes. The authors don’t report whether X genes in general are significantly depleted in ribosomal protein coding genes. This is an important test, is it just X hemizygous genes, is there a significant difference between X/Y genes? Related to this, it would be useful to mention in the figure legends that X hemizygous genes have a mean Y/X expression ratio in males = 0 whereas all other categories are X/Y. For example, on page 3 it says ‘for X/Y contigs, the difference between maternal and paternal X in females increases with Y degeneration’, but clearly this is not the case for genes with mean Y/X expression ratio of 0.

For genes with no Y decay (mean Y/X expression ratio in males = 1-1.5), the Y allele appears to be upregulated in males relative to the ancestral allele. Is this because of sex-specific selection?

It would be useful to mention in the methods why it is important for SEX-DETector that X and Y sequences are assembled in the same contig. This is mentioned in the reference transcriptome section but not explained.

It would be good to mention dosage compensation in Tribolium, which seems to have a parallel mechanism to the pattern the authors find. The female X chromosome is hyper transcribed relative to autosomal expression levels but the X and autosomes in males are compensated. Prince, E. G., Kirkland, D. & Demuth, J. P. Hyperexpression of the X chromosome in both sexes results in extensive female bias of X-linked genes in the Flour Beetle. Genome Biol. Evol. 2, 336–346 (2010).

I assume that the genes shown in Figure 1, Supplementary Figures are genes with unbiased expression (as sex-biased genes were removed). It would be useful to have another supplementary graph showing the magnitude of expression differences between the sexes for these genes. I assume that the expression equality is achieved despite upregulation of the maternal X in females, because the upregulation of the maternal X in males is much greater?

Mule et al. look at allele-specific expression of S. latifolia, a plant whose young sex chromosomes have acquired partial dosage compensation. They find that, contrary to the expectation if this was simply due to general buffering mechanisms, the maternally derived X chromosome of females is also up-regulated relative to the outgroups. They interpret this as evidence for an imprinting-based mechanism of dosage compensation. S. latifolia is a great model in which to study the evolution of dosage compensation, and the examination of paternal and maternal allele expression is an important step forward.

My main concern is that, as far as I understand, the authors did not perform the experiments on the reverse cross (Leuk144-3mother x U1037_father), and therefore cannot control for strain of origin effects. My impression is that this would be essencial for imprinting to be conclusively detected. As it is, the results are very suggestive but difficult to interpret. Since they pick genes that:

-do not have sex-biased expression

-do not differ in expression with the outgroup

-whose expression is lower for the Y copy than the X copy

Then in males these genes will by definition have increased X (maternal) expression. Since the same maternal X (or at least an X from the same strain) is being transmitted to the daughters, increased maternal expression may be expected even if there is no imprinting.

Another issue is that the patterns are not as clear as figure 2 and the main text seem to imply:

*The 0.75-1 category looks very different in figure 1 (no difference between maternal/paternal allele) and figure 2 (very significant difference).

*For 0.5-0.75 there seems to be a difference in figure 1, but not due to overexpression of the maternal allele.

*in flower buds the effect is stronger for the 0.25-0.5 category than 0-0.25.

*in leaf and flower bud an excess of maternal allele expression is detected for the autosomes in females, and in leaf an excess of paternal expression is detected in males for the 0.57-1 category.

*there is generally no significant excess of female maternal allele expression for the validated contigs (figures s4 to s6).

I understand that these could simply be due to noise/lack of power, but they should be addressed more openly in the text. It was particularly confusing that even though flower bud data was used for the sequencing of the parents, and for the reference transcriptome, seedling is the tissue shown in the main figures. This should be justified.

Other:

*It seems surprising to me that there is no differential expression between S. latifolia and its outgroups - such differences are usually found even between strains of the same species.

*Random selection of 200 autosomes: I think an important additional control would be to use the smallest sample size (79) and run a bootstrapping procedure to get a probability of being different from 0 in female.

*"Autosomal normalised expression levels in the two outgroups (S. vulgaris and S. viscosa) were averaged together." --> this does not seem like a standard approach. Was the analysis checked using each separately?

*The numbering in the supplementary methods is inconsistent (e.g. 4) Validation of sex-linked contigs and 4) Expression level estimates).

*Normalized fpkm: was this normalized any further than just calculating fpkm, which is often not sufficient? If not, did you check the distribution of fpkm between samples to exclude biases?

*The text could do with a few more details, even if most of the methods are supplementary. For instance:

-you don't mention the crossing scheme that you used at all.

-In page 2, a sentence explaining how you defined sex biased genes would have been useful.

-The number and type of tissues used is never mentioned in the main text.

-In page 3, it would be useful to have a qualification of the close relatedness of the species.