An intercross population study reveals genes associated with body size and plumage color in ducks

Comparative population genomics offers an opportunity to discover the signatures of artificial selection during animal domestication, however, their function cannot be directly revealed. We discover the selection signatures using genome-wide comparisons among 40 mallards, 36 indigenous-breed ducks, and 30 Pekin ducks. Then, the phenotypes are fine-mapped based on resequencing of 1026 ducks from an F2 segregating population generated by wild × domestic crosses. Interestingly, the two key economic traits of Pekin duck are associated with two selective sweeps with fixed mutations. A novel intronic insertion most possibly leads to a splicing change in MITF accounted for white duck down feathers. And a putative long-distance regulatory mutation causes continuous expression of the IGF2BP1 gene after birth which increases body size by 15% and feed efficiency by 6%. This study provides new insights into genotype–phenotype associations in animal research and constitutes a promising resource on economically important genes in fowl.

This is a very interesting study that combines selective sweep analysis on two different duck populations. The way that doemstication has proceeded in the duck has allowed a really nice separation of domestication related sweeps from more recent improvement related sweeps. I think this is a very novel addition to the domestication literature, and in fact slightly more could be made of this in the manuscript. The authors also use an F2 intercross to QTL map the traits that have been selected during domestication and use these to verify one of the sweep regions. This represents far greater evidence of function than the usual sweep mapping approach of simply looking at promising annotated genes and pontificating over putative functions that is very limited. The manuscript has some strange paragraph and sentence construction at various places, as well as some odd turns of phrase that could use some additional scrutiny. In general it looks like it has been written in 'Nature' format, but could be better suited changing to a more standard style (intro/ methods/ results/ discussion) that Nature Communications accepts.
Major points • Far too much is made of the KEGG pathway analysis and the authors use the same rather speculative pontificating that many sweep studies have used when looking at the lists of genes that are selected, namely pull out a lot of GO terms and then cherry pick which ones seem to make a nice story. It is always a very weak line of questioning at best, but here when the authors actually go to the trouble of developing an intercross to test these regions later on in the manuscript, it should be greatly reduced. For instance they select two of 16 and then make the case for behaviour and egg production, but really with such poor annotation in the duck genome (and even in far better annotated genomes this approach is generally specious -see an excellent critique in 'Pavlidis, P. et al. A critical assessment of storytelling: GO categories and the importance of validating genomic scans. Mol Biol Evol 29(10):3237-3248, 2012' where they randomly generate selective sweeps in Drosophila and then make plausible stories from the KEGG terms they pull out of these random data sets).
• Similar to the above issue, the gene PPARA is selected due to a rough hypothesis based on weak annotation arguments. There is no additional evidence for this gene above any of the other sweeps, no overlap apparently with any of the linkage-based QTL analysis. This isn't very persuasive, and should be scaled back or dropped entirely.
• For the MITF section, the authors find that they have 21 fixed SNPs in the region, plus a 6kb intronic insertion. They perform some nice experiments with RNA expression to show that although MITF is not affected, downstream targets are. I think that this highlights MITF as the causal gene rather nicely (especially with the large amounts of sequencing they do on mallards with the colour morph). It is not particularly surprising of course (MITF has been demonstrated to affect colour in multiple other species), and they do not really verify that the insertion is indeed the causal mutation. Although further mutation verification experiments (transient transfection, maybe EMSA, etc) seem too onerous to perform given the huge amount of sequencing that has already been performed, they should be more cautious when interpreting causality for the insertion and insert a few more caveats.
• I really like the GWAS (technically an F2 linkage analysis) combined with the sweep regions to actually try and meaningfully understand what the sweep regions potentially control, and I think this should be brought up earlier in the manuscript (certainly in place of the KEGG analysis). Currently just the top hits of the GWAS are listed, but the authors should provide the full details of all QTL identified in these scans for all traits in a table, showing direction of effect, r-squared, additive, dominance etc (there are plots in figure 3c, but these are very small and hard to read). When considering all these QTL found, is there a significant overlap between the QTL detected and the selective sweeps (i.e. are the QTL regions enriched for selective sweeps). This would be very interesting to see, and may also give insights into other sweep regions and their possible effects. Also, no details are given (though maybe I overlooked them somewhere in the supplementary methods, but they appear to be absent) for the actual analysis of this F2 population. If they just analysed it like a typical GWAS this will be erroneous unless they also control for the family effects using a relatedness matrix. This is a linkage rather than a linkage disequilibrium population, so standard QTL map construction and analysis would probably be best in any case.
• The authors look specifically at a 300kb sweep region that overlaps with a QTL for head weight and various other phenotypes, with only 1 of the 19 genes in the region showing differential expression over a range of time points. They use recombinants within the region to narrow down the candidate region controlling IGF2BP1 down to a 100kb region. This seems persuasive, though they do have rather few of certain recombinant types in several cases (the downstream breakpoint is based on 2 R2 birds and 12 R6 birds). Candidate mutation/ polymorphism still lacking potentially, but this is a very nice example of actually using a sweep region to narrow down a candidate region. Possibly a more in-depth breakdown of this 100kb region using multiple haplotypes in the different mallard and indigenous bird species could help further (or at least illustrate the region). The figure (figure 3) for this section is very busy and could be done with split up into two -it is very hard to read the recombinants plus their phenotype with the current scale for example.
• The Hi-C analysis is poorly explained and the results are not very clear to interpret. It looks like the entire candidate region is associated in 1 large block. Is this true? If so, how useful is this technique for getting down to candidate SNPs, etc? What else can it tell us? I think the authors should expand on this section a little more, especially as space is not really a problem with the manuscript length as it stands.
• I am not entirely sure if FST is the best way to detect domestication sweeps, since won't it also be sensitive to changes in the wild population too? The authors currently use FST and pi in combination I believe, how many sweeps are identified if the authors just use pi as a basis? This isn't too major an issue as they are currently being more selective by using both criteria of course.
Minor points • The second paragraph (lines 47-58) is very poorly written, with some very strange sentences and odd construction in general. They also use some odd turns of phrase, which I suggest they remove (line 52 'noble plumage', line 53 'fatten skin', line 54 'incomparable economic traits').
• For the sweep overlaps, the authors fail to mention how many sweeps are shared between the domestication and improvement scans until much later on in the manuscript. This should be mentioned when the sweeps are first discussed.
• The FST differences (0.10 as compared to 0.07) is taken as evidence for stronger selection during the improvement stage. Would it be possible to test whether this is significant somehow? I.e statistically more sweeps in the improvement stage, etc? As it is, this seems like a rather moderate difference in FST values. I am also not sure that this result can be taken as 'unintentional lower selection intensity to establish a reciprocal relationship with humans'. In fact the meaning of this sentence is unclear, but it implies that domestication has been centred on behavioural traits, though without any evidence to back it up bar male indigenous birds having green heads and lower production traits, it should really be removed.
• Line 172 'All the record traits' rephrase for clarity.

Reviewer #1 (Remarks to the Author):
Q1: The paper by Zhou et al. describes extensive genomic studies in duck and provide some very interesting findings in relation to genes that have likely been under selection during the formation of the domestic Pekin duck. Unfortunately, the quality of the English language used throughout the manuscript is rather poor and often results in ambiguous description that are not very clear.

Response:
Thank you very much for your positive comments and constructive suggestions. We added section headings and subheadings in the results to promote the logical and legible flow of the article. The manuscript has been professionally edited by an English editing service agency, American Journal Experts (AJE).
Q2: Furthermore, while the authors present some compelling evidence for selective sweeps based on their Fst analyses, I would have preferred if the authors also had used more formal analyses to identify selective sweep regions using e.g. iHS or EHH methods. I am also not convinced about the Fst thresholds used by the authors (see figure 1d and on page 5, lines 8-9). How were these thresholds calculated, and did the authors taken into account multiple testing.

Response:
We apologize for the unclear description in the methods. The selective sweeps were identified using both FST and the ln π ratio with the threshold defined by the methods used by Qiu et al. (Nature Communications 2015, 6:10283); these methods were used to calculate the significance of each window with the Z test.
We also performed a permutation test (Huang et al. 2012) 193225 0.222753,0.220095,0.219102, 0.204254,0.193225,0.192792, 0.191981,0.190458,0.19019, 0   We also revised this statement more specifically as follows: "Our study not only unraveled the leucism mechanism of Pekin duck but also suggested that functional mutations of a critical gene play a prominent role in duck phenotypic evolution." To avoid false long-distance regulation results due to assembly errors, we focused on using Hi-C data to show the genome assembly quality in the revision. We also checked the end of chromosome 28 with emphasis. Additionally, we added collinearity analyses to confirm the results, which indicated that chicken and duck have very conservative collinearity (see Supplementary Figure 15). We further revised the text as follows: "Considering inversion can cause an illusion of long-distance regulation, we checked the end of chromosome 28 using Hi-C data and collinearity analyses ( Supplementary Fig. 15) and found a high-quality scaffold order assignment and no obvious inversion."

Supplementary Figure 15. Chromosome interaction mapping (Hi-C) and
collinearity analysis confirm the scaffold assembly of the end of chromosome 28.    Figure 12 and Supplementary Table 18). Then, we performed real-time PCR and validated the differential expression of IGF2BP1 in heart, lung, spleen, kidney, and brain as well between mallard and Pekin duck (Figure 4a). We noticed that IGF2BP1 was not differentially expressed in breast muscle, and we found that the IGF2BP1 locus genotype was not associated with breast muscle (P=4.75×10 -9 , the whole genome Bonferroni significance threshold is P<1×10 -9 )) in the F2 population. Second, the IGF2BP1 expression level in liver was strongly associated with body weight (P = 9.1×10 -8 , Figure 4c) and body size-related traits in the F2 population (Supplementary Figure 14).

Response:
We apologize for the confusing figure. We separated Figure 1a containing the map of duck sampling and the principal component plot as Fig 1a and Fig 1b (see   below figure).
(QTL) that shows strong association....." Page 8, line 11: "casual" should be "causal" Q7: Also, no details are given (though maybe I overlooked them somewhere in the supplementary methods, but they appear to be absent) for the actual analysis of this F2 population. If they just analysed it like a typical GWAS this will be erroneous unless they also control for the family effects using a relatedness matrix. This is a linkage rather than a linkage disequilibrium population, so standard QTL map construction and analysis would probably be best in any case. potentially, but this is a very nice example of actually using a sweep region to narrow down a candidate region. Possibly a more in-depth breakdown of this 100kb region using multiple haplotypes in the different mallard and indigenous bird species could help further (or at least illustrate the region). The figure (figure 3) for this section is very busy and could be done with split up into twoit is very hard to read the recombinants plus their phenotype with the current scale for example. Figure 3 into Figure 3 and What else can it tell us? I think the authors should expand on this section a little more, especially as space is not really a problem with the manuscript length as it stands.

Response: We apologize that we over-interpreted the Hi-C results. We agree with the reviewer that the identified topologically associated domain (TAD)-like regions
provide weak evidence of an interaction between IGF2BP1 and their long distance cis regulation region. We also tried to detect the interaction between the two regions using the new method, PSYCHIC (Ron et al. 2017 To avoid the false long-distance regulation results due to assembly errors, we focused on using Hi-C data to show the genome assembly quality in the revision (Supplementary Figure 15). We also emphasized the end of chromosome 28. We also revised the text as follows: "it implies that domestication has been centered on behavioral traits" and deleted the following description in the revision.
Q13: Line 172 'All the record traits' rephrase for clarity.

Response: Thank you for your suggestion. All the recorded traits refer to the carcass traits measured in the F2 population slaughter experiment, such as body weight, head
weight, wing weight, heart weight, liver weight, gizzard weight, leg weight, tarsometatarsus length, and chest width. This was clarified in the main text.   The authors have very much improved the manuscript and addressed most of my comments. The paper now reads very well and I very much like the extensive results presented by the authors. Nevertheless, I still am not convinced by the conclusion from the authors that they have conclusive evidence that a mutation in a regulatory region affecting the IGF2BP1 gene is the causative mutation for the phenotypic differences on growth identified between Peking duck and mallard. On page 8, line 25 they refer to the Suppl. Table 14. However this table only shows the primers used for the qPCR but NOT the expression levels for the 10 different haplotypes. I could not find any information about the actual gene expression within the documents provided. Furthermore, the 100 kb region contains many other genes (HOX genes) and the authors do not show a direct correlation between the phenotype and the gene expression of IGF2BP1. In their cross, there is much LD and they cannot rule out that other genes underlie their observed phenotypic difference. In my view the authors need to (1) provide the gene expression data for the 10 haplotypes and (2) tone down their claims that a mutation in a regulatory sequence for the IGF2BP1 is the causative mutation.

List of the updated figures Current Version Last Version
In the discussion (page 10, line 20) the authors refer this as "Another putative new mutation in the long-distance regulatory region". Assuming that this region is indeed a long-distance regulator (of which I am not convinced), than they still do not identify the actual mutation within the 100 kb of sequence.
The English language has now been very much improved, but I have identified a few minor mistakes which are listed below. Page 2, line 5: "segregated for fine mapping". To me this still doesn't sound correct Page 3, line 6: "fowl" should be "fowls" Page 3, last sentence: Z and W chromosomes (plural) Page 4, line 4: replace "for a mean depth" by "at a mean depth". I also think that the word "coverage" is redundant and should be removed. Page 6, line 9: "association" instead of "associate" Page 7, line 5: replace "was" by "is" Page 7, line 23: change to: "the carcass traits measured in the ...." Page 10, line 21-22: "which is fixed ......productive performance". This still is a weird sentence: "fixed under continual selection as a major effect QTL"? and "reinvented production performance"? Page 11, line 15: replace "shared" by "for sharing" Reviewer #2 (Remarks to the Author): I have now looked over the revisions and most have been met to my satisfaction. A few rather minor points are indicated below, however.
Q3. For the sentence that has been revised "the MITF-M isoform has almost no expression in ducks, strongly suggesting that the splicing changes in MITF were most likely caused by insertion and resulted in white plumage in Pekin ducks", reword to the following: "the MITF-M isoform has almost no expression in ducks, suggesting that the splicing changes in MITF were most likely caused by insertion and resulted in white plumage in Pekin ducks. However functional verification of this potentially causal mechanism is still required." Q5. I couldn't supplementary dataset 4 for the full GWAS results, but in any case, these should be given in a table in the supplementary section. Q10. I don't think the authors properly understood my question. As I understand it, they classed sweeps as having to meet both FST and pi thresholds. However, how many sweeps would be detected if they just used pi (i.e. where a lot of sweeps discarded as they met the pi requirements but not the FST thresholds)?
Q12. The authors show that the difference in FSTs between domestication and improvement stages is significant, but this p-value should be inserted into the manuscript text.
Finally, the language still needs to be tidied up quite a bit, but I leave this to the editor to determine by how much. Apart from this the manuscript has been greatly improved by the revisions.
Reviewer #3 (Remarks to the Author): This interesting manuscript by Zhou et al. investigates the effects of two stages artificial selection on the genome of ducks, one during their domestication and one during a more recent improvement of certain economically advantageous characteristics. Using population genomics analyses and an F2 intercross between the wild and improved breeds, the authors identify candidate divergence regions and proceed to map genetic variations affecting expression of two genes related to specific selected phenotypes.
Q1: For one of these genes (IGF2BP1), the selected region appears to be located around 150 kb away from the affected gene itself, thus the authors speculate that this might be due to impaired or altered long-range regulation of IGF2BP1. They then use Hi-C and collinearity analysis to assess the integrity of their genome assembly and check for evidence of large inversions that might create a false suggestion of long-range effects, without finding any. Although the resolution of the Hi-C is quite limited, it is sufficient to validate the quality of the assembly and to support the lack of any large-scale inversions, as the authors appropriately state in the text ("…a high quality scaffold and no obvious inversion"). It should however be noted that this level of resolution cannot guarantee absence of smaller-scale inversions. More importantly, it is unclear what breed they generated their Hi-C library from. The only information I could find in the methods was that "Duck livers were cross-linked…", but were these mallards? Pekin ducks? some of the F2 population? If the analysis was performed in only one of the parental breeds, then the other could still theoretically carry any kind of rearrangements, including an inversion in the IGF2BP1 region (although this is perhaps unlikely given the recombination events observed). In general, there should be more stats about the Hi-C libraries in the methods section or as a supplementary figure/table, indicating for example the total number of reads sequenced, number of reads filtered out at every step, number of reads or read pairs that were used to build the contact maps after all filtering and intra-/inter-chromosomal interaction ratio as an estimator of library quality.
Q2: While it is not necessarily within the scope of the paper, it would have indeed been interesting to see whether the differences in the expression of IGF2BP1 were due to an alteration in the long-range interactions that this gene engages in. At present there is relatively little literature on the impact of 3D genome architecture on genome evolution and showing whether this level of regulation might have had an influence in this relatively well-defined setting of duck genomic evolution would have given an intriguing contribution to the understanding of evolutionary dynamics. That being said, I understand that the authors tried to tackle this problem from their Hi-C data but couldn't find any clear interactions between the IGF2BP1 promoter and the variant region (possibly due to the insufficient resolution) and I appreciate that the poor annotation of the genome can be a big hindrance in this kind of experiment. Addressing this question would probably require the use of 4C using the IGF2BP1 promoter region as viewpoint in animals with opposing phenotypes. The structural information thus obtained could in turn allow for potential examination of the interacting sites for chromatin marks and for the binding of conserved architectural proteins (e.g. CTCF, cohesin) that can be influenced by simple modification of the sequence. This endeavour would require considerable additional sequencing and analysis and possibly even production of new duck-specific antibodies. For these reasons and since the main focus of the paper is indeed different, I am not suggesting that the authors perform these experiments for this work. However, it may be an interesting avenue of research to explore in the future and it should be included as a point of discussion.
Q3 (minor): In the introduction, the white plumage of the Pekin ducks is characterised as "favourable", but the authors never really explain why plumage colour would constitute an economically positive trait. This is a rather accessory point, of course, but given that part of the findings presented concerns a gene controlling plumage colour, clarifying the role that this trait might have had in the selection process might help to better illustrate the logic behind the interpretation of the data.
Q4 (very minor): While I do appreciate some variety in the sometimes dull vocabulary of scientific literature, some of the language in the paper is a bit too hyperbolic. I could also find a typographic error.
-In the introduction I would change the "notable and superb" economic traits to "desirable" or "superior".
-At the end of the discussion, I am not sure what the authors mean when they say that "the creation of new traits occurred quickly and *inconceivably* in the comparatively short domestication process". If they simply want to stress (strongly) that the rate of emergence of these new traits is unexpectedly fast, I would maybe say "the creation of new traits occurred surprisingly quickly, as it was not believed to be possible in the comparatively short domestication process", although I cannot comment on the accuracy of such statement.

Reviewers' comments:
Reviewer #1 (Remarks to the Author): The authors have very much improved the manuscript and addressed most of my comments. The paper now reads very well and I very much like the extensive results presented by the authors.
Response: Thank you for your positive comment. We appreciate your previous suggestions, which have helped us to greatly improve the manuscript.
Q1: Nevertheless, I still am not convinced by the conclusion from the authors that they have conclusive evidence that a mutation in a regulatory region affecting the IGF2BP1 gene is the causative mutation for the phenotypic differences on growth identified between Peking duck and mallard.
On page 8, line 25 they refer to the Suppl. Table 14. However this table only shows the primers used for the qPCR but NOT the expression levels for the 10 different haplotypes. I could not find any information about the actual gene expression within the documents provided. Furthermore, the 100 kb region contains many other genes (HOX genes) and the authors do not show a direct correlation between the phenotype and the gene expression of IGF2BP1. In their cross, there is much LD and they cannot rule out that other genes underlie their observed phenotypic difference.
In my view the authors need to (1) provide the gene expression data for the 10 haplotypes and (2) tone down their claims that a mutation in a regulatory sequence for the IGF2BP1 is the causative mutation.
In the discussion (page 10, line 20) the authors refer this as "Another putative new mutation in the long-distance regulatory region". Assuming that this region is indeed a long-distance regulator (of which I am not convinced), than they still do not identify the actual mutation within the 100 kb of sequence.   Fig. 4a).

Response
We have also toned down our claim and revised it as follows: "a putative long-distance regulatory mutation leading to continuous expression of the IGF2BP1 gene after birth and an increase in body size by 15% and in feed efficiency by 6%" in the abstract and "Another putative new mutation near IGF2BP1, fixed under continual selection, acted as a major-effect QTL and transformed production performance" in the Discussion.
The English language has now been very much improved, but I have identified a few minor mistakes which are listed below.
Page 3, last sentence: Z and W chromosomes (plural) Page 4, line 4: replace "for a mean depth" by "at a mean depth". I also think that the word "coverage" is redundant and should be removed.
Page 11, line 15: replace "shared" by "for sharing" Response: Thank you for patiently reviewing our manuscript and for noting the mistakes. In the revised manuscript, we have corrected each of these mistakes as suggested.

Reviewer #2 (Remarks to the Author):
I have now looked over the revisions and most have been met to my satisfaction. A few rather minor points are indicated below, however.
Q3. For the sentence that has been revised "the MITF-M isoform has almost no expression in ducks, strongly suggesting that the splicing changes in MITF were most likely caused by insertion and resulted in white plumage in Pekin ducks", reword to the following: "the MITF-M isoform has almost no expression in ducks, suggesting that the splicing changes in MITF were most likely caused by insertion and resulted in white plumage in Pekin ducks.
However functional verification of this potentially causal mechanism is still required." Finally, the language still needs to be tidied up quite a bit, but I leave this to the editor to determine by how much. Apart from this the manuscript has been greatly improved by the revisions.

Response:
The manuscript has been carefully edited by a professional English editing service. We are trying our best to meet the language demands of the journal.

Reviewer #3 (Remarks to the Author):
This interesting manuscript by Zhou et al. investigates the effects of two stages artificial selection on the genome of ducks, one during their domestication and one during a more recent improvement of certain economically advantageous characteristics. Using population genomics analyses and an F2 intercross between the wild and improved breeds, the authors identify candidate divergence regions and proceed to map genetic variations affecting expression of two genes related to specific selected phenotypes.
Q1: For one of these genes (IGF2BP1), the selected region appears to be located around 150 kb away from the affected gene itself, thus the authors speculate that this might be due to impaired or altered long-range regulation of IGF2BP1. They then use Hi-C and collinearity analysis to assess the integrity of their genome assembly and check for evidence of large inversions that might create a false suggestion of long-range effects, without finding any.
Although the resolution of the Hi-C is quite limited, it is sufficient to validate the quality of the assembly and to support the lack of any large-scale inversions, as the authors appropriately state in the text ("…a high quality scaffold and no obvious inversion"). It should however be noted that this level of resolution cannot guarantee absence of smaller-scale inversions.
More importantly, it is unclear what breed they generated their Hi-C library from. The only information I could find in the methods was that "Duck livers were cross-linked…", but were these mallards? Pekin ducks? some of the F2 population? If the analysis was performed in only one of the parental breeds, then the other could still theoretically carry any kind of rearrangements, including an inversion in the IGF2BP1 region (although this is perhaps unlikely given the recombination events observed).
In general, there should be more stats about the Hi-C libraries in the methods section or as a supplementary figure/table, indicating for example the total number of reads sequenced, number of reads filtered out at every step, number of reads or read pairs that were used to build the contact maps after all filtering and intra-/inter-chromosomal interaction ratio as an estimator of library quality.

Response: Thank you for your advice. We used a male Pekin duck to generate three independent
Hi-C libraries. We have added this information to the Methods section (see "High-throughput chromosome conformation capture (Hi-C) experiment and sequencing" in the methods section).
Considering the possible rearrangements, we have also toned down our claim and revised "a long-distance regulatory mutation" to "a putative long-distance regulatory mutation" in the abstract. Regarding the limited resolution of Hi-C and its use to validate the quality of the assembly, we have revised our claim as "we found a high-quality scaffold order assignment and no obvious large-scale inversions." In our supplementary materials, we have provided the statistical results of reads filtered out at every step (see Supplementary Table 21). Q2: While it is not necessarily within the scope of the paper, it would have indeed been interesting to see whether the differences in the expression of IGF2BP1 were due to an alteration in the long- That being said, I understand that the authors tried to tackle this problem from their Hi-C data but couldn't find any clear interactions between the IGF2BP1 promoter and the variant region (possibly due to the insufficient resolution) and I appreciate that the poor annotation of the genome can be a big hindrance in this kind of experiment.

Supplementary
Addressing this question would probably require the use of 4C using the IGF2BP1 promoter region as viewpoint in animals with opposing phenotypes. The structural information thus obtained could in turn allow for potential examination of the interacting sites for chromatin marks and for the binding of conserved architectural proteins (e.g. CTCF, cohesin) that can be influenced by simple modification of the sequence. This endeavour would require considerable additional sequencing and analysis and possibly even production of new duck-specific antibodies.
For these reasons and since the main focus of the paper is indeed different, I am not suggesting that the authors perform these experiments for this work. However, it may be an interesting avenue of research to explore in the future and it should be included as a point of discussion.
Response: Thank you for your insightful comments and suggestions. As suggested, we will continue working to identify the causative mutation that leads to continuous expression of the IGF2BP1 gene after hatching, leading to a 15% increase in body size and a 6% increase in feed efficiency.
Q3 (minor): In the introduction, the white plumage of the Pekin ducks is characterised as "favourable", but the authors never really explain why plumage colour would constitute an economically positive trait. This is a rather accessory point, of course, but given that part of the findings presented concerns a gene controlling plumage colour, clarifying the role that this trait might have had in the selection process might help to better illustrate the logic behind the interpretation of the data.
Response: Thank you for noting this ambiguous description. There were two main reasons why we characterized white plumage as an "economic trait". First, the white feathers favored by humans are generally used to make down jackets or quilts, and they are also more suitable for dyeing than black or tan feathers. Second, a lack of pigment increases the appeal of a carcass for culinary use.
We revised the sentence as follows: "Pekin ducks show many striking changes such as white plumage, a favorable feature that meets the demand for white down as a filler for jackets or quilts and makes the carcass easy to clean".
Q4 (very minor): While I do appreciate some variety in the sometimes dull vocabulary of scientific literature, some of the language in the paper is a bit too hyperbolic. I could also find a typographic error.
-In the introduction I would change the "notable and superb" economic traits to "desirable" or "superior".
-At the end of the discussion, I am not sure what the authors mean when they say that "the creation of new traits occurred quickly and *inconceivably* in the comparatively short domestication process". If they simply want to stress (strongly) that the rate of emergence of these new traits is unexpectedly fast, I would maybe say "the creation of new traits occurred surprisingly quickly, as it was not believed to be possible in the comparatively short domestication process", although I cannot comment on the accuracy of such statement.

Response:
Thank you for your detailed revision comments, and please excuse our occasional poor descriptions and spelling errors. Following your suggestions, we have corrected these mistakes in the revised manuscript.