How does migration effect evolution

The average size of these regions was Each region included one to four variants, all of which occurred in non-coding regions Supplementary file 5.

Accordingly, the selection in genomic regions that we identified here is probably involved in the transition from migrant to resident phenotypes. Genetic elements, scaffolds and genes discussed in the text are highlighted.

Local neighbour joining trees for regions under selection in a the resident continent population on Super-Scaffold 99, and b medium-distance NW population on Super-Scaffold Selection is indicated by longer branch lengths in each population than is the case in global trees built using data from all genomic regions Figure 4—figure supplement 1. Panels to the right of the trees show the corresponding frequency of haplotypes in each population of the tree.

Haplotype clusters are colour coded colours of haplotype clusters do not correspond to the population colour coding used in other figures , and frequencies are plotted along the Y axis. Haplotype frequency plots show the near fixation of a single dominating haplotype in a resident continent yellow and b medium-distance NW populations blue.

The location in bp of these regions on each Super-Scaffold is shown below these panels and the resident continent group is only included to root the tree in panel b , and thus has no haplotype frequencies. Results from hapFLK include the size, the population where the signal was found and genes within the region. Estimates of PBS were re-estimated using island populations vs. PBS is an F ST - based statistic that estimates allele frequency differences between three or more populations.

This parameter can be elevated by linked purifying selection or background selection within populations that is unrelated to positive selection in our case selection related to migration. Linked selection should increase haplotype lengths at genomic regions that are under positive selection. As noted already, population structure and linked selection can elevate differentiation between populations. In addition, linked purifying selection would be expected to increase PBS in all populations i.

We conservatively excluded these populations from our initial analyses because their sample sizes are small and because genetic drift can affect estimates of differentiation in island populations. Nevertheless, the island populations are also resident and thus these estimates could help to validate the genomic regions that were identified as being under selection in resident populations on the continent. Table 1a summarizes these results, noting which genomic regions exhibited elevated values of PBS on islands.

Of particular interest, PBS was elevated in all three island populations at the genomic region on Super-Scaffold 99 Figure 5b. These estimates are only elevated in the resident continent phenotype, ruling out a role for linked selection in generating this signature in residents. Colours correspond to Figure 1a with yellow showing data for resident continent birds. Note that it is possible that the signatures of positive selection that we document here reflect selection based on different ecological variables involved with the colonization of areas further south on the continent, but at least in the case of Super-Scaffold 99, we believe that this is rather unlikely as most ecological variables biotic and abiotic are quite distinct between islands and the continent and between the islands themselves Cropper, ; Valente et al.

The transition to residency is shared, probably representing one of the only shared selection pressures experienced by all of these populations. Note that the lack of consistent results for other regions under selection in the resident continent population does not preclude the potential importance of these regions as, for example, genetic drift on islands would affect which genetic variants were present on islands for selection to act on.

Our finding that only a few genomic regions under selection contain genes and that the strongly associated SNPs identified by CAVIAR are in non-coding regions could suggest that cis-regulatory changes are important for the transition from migration to residency.

In fact, previous work with monarch butterflies identified 55 conserved microRNAs that are differentially expressed between summer and migratory butterflies Zhan et al. Future work using techniques aimed at identifying binding sites for transcription factors e. Specifically, Ruegg et al. Four of these genes are transcription factors whose motifs are in the libraries searched by HOMER: three basic helix-loop-helix transcription factors bHLH Clock , Npas2 , and Bmal1 and one basic leucine zipper domain Nfil3.

This motif could disrupt or weaken transcription factor binding Kasowski et al. This is also the genomic region that showed elevated PBS in both resident continent and island populations Figure 4a , Figure 5. Clock , Npas2 and Bmal1 are involved in maintaining circadian rhythms. Circadian rhythms synchronize circannual clocks, which are important cues controlling seasonal migratory behaviour Gwinner, ; Visser et al.

Concerning the actual identity of genes within regions that are under selection, several have functions that could be related to the transition from migration to residency. For example, LOC located on in the genomic region under selection on Super-Scaffold 12, the region with a bHLH motif mentioned above; Table 1a has been annotated as a probable G-protein coupled receptor that mediates the function of neuropeptide Y NPY.

It has been hypothesized that the effects of NPY may extend to seasonal changes in energy balance that are important for migration, including hyperphagia and fat deposition Boswell and Dunn, Beyond its role in energy balance, NPY also facilitates learning and memory via the modulation of hippocampal activity and has an effect on circadian rhythms, reproduction, and the contraction of vascular smooth muscles.

It has been suggested that a common genetic mechanism or major regulator may control migratory traits Liedvogel et al. A protein such as NPY, or the transcription factors that bind the bHLH motif identified in the prior analysis, could fill this role. So far, we have considered all three migratory traits exhibited by blackcaps together propensity, orientation and distance and our results relate mostly to residents. The elevated population differentiation that we noted between resident and migratory birds could reduce our power to identify selection that is specific to migrants Fariello et al.

Accordingly, we ran a second set of analyses excluding resident birds and examining migratory orientation and distance independently. Results for nSL can be found in Supplementary file 5. The list of genes in genomic regions that are under selection in this analysis focusing on orientation is small, but it also includes genes with functions that are strongly related to the phenotype they are associated with.

This gene codes for a transmembrane protein that helps to regulate the Wnt signalling pathway. This pathway plays a role in embryonic development and has been shown to influence feather and beak morphogenesis, along with feather molt Yu et al. NW migrants have rounder wings and more narrow beaks than southern migrants Rolshausen et al.

Differences in the timing of migration probably mean that birds also molt at different times. This has not been evaluated directly in comparisons between migrants, but variation in molt patterns have been documented between NW migrants and birds that are resident on the continent de la Hera et al.

Delmore and Liedvogel identified a region on chromosome 4 and Lundberg et al. None of these regions overlap with those under selection in our study on blackcaps. It is tempting to suggest that migration may be controlled by similar genes across broad taxonomic scales, with early results from candidate genes e. Nevertheless, several studies have failed to document an association with Clock, and a comparison of our results with those of Delmore and Liedvogel and Lundberg et al.

This is an important finding as it has long been hypothesized that there may be a shared genetic mechanism for migration, not only in birds but also in other taxonomic groups Liedvogel et al. This fact is evident in Figure 4 , in which the regions under selection still include haplotypes from a different cluster, and it could suggest that selection is acting on shared genetic variation i.

The idea that transitions between migratory phenotypes have been facilitated by shared genetic variation has been around for quite some time in the blackcap literature, particularly as rapid transitions have been observed and include the evolution of a new NW migratory route in the past 70 years. Shared variation can facilitate these rapid changes as these variants are already present in the population and have been tested by selection Barrett and Schluter, The fact that regions under selection are quite narrow Table 1 also supports a role for shared genetic variation Barrett and Schluter, and we provide further evidence below.

First, we estimated the genetic distance between one haplotype in each cluster and an ancestral sequence that we derived using WGS from the two most closely related sister taxa, hill babbler Pseudoalcippe abyssinica , an African resident and garden warbler Sylvia borin , a long distance migrant Voelker and Light, Using the region on Super-Scaffold 73 that shows selection in NW migrants Figure 4b , we predicted that if haplotypes in the light blue cluster were present in the population already, they should exhibit similar levels of divergence from the ancestral sequence as haplotypes from all other clusters.

This is precisely what we found; genetic distance from the ancestral sequence was similar for haplotypes from all clusters differences for the NW haplotype vs. We reran this analysis limiting our data to synonymous substitutions in predicted coding regions i. Specifically, we identified six synonymous substitutions between all three medium-distance migrant populations and both garden warblers and hill babblers, suggesting that there is no difference in the age of these haplotypes.

To follow up on the former analysis, we constructed a maximum likelihood ML tree using sequence data from the region under selection on Super-Scaffold We built this tree using data from all continental blackcaps, garden warblers, and hill babblers, using the willow warbler as an outgroup, and compared this tree to a consensus tree summarizing ML trees constructed for each scaffold in the blackcap reference genome i. Supporting previous phylogenetic work in the system, garden warblers and hill babblers formed a sister clade to blackcaps in the consensus tree, and relationships among blackcaps were largely unresolved.

By contrast, garden warblers were more closely related to blackcaps than were hill babblers in the tree built using data from the region on Super-Scaffold 73 Figure 6b. In addition, the medium-distance NW population in which positive selection is acting in this particular region occurs at the base of the blackcap clade. Recall that garden warblers are obligate migrants whereas hill babblers are residents, supporting the suggestion that haplotypes favoured in the NW phenotype were already present in the population before divergence, perhaps even in ancestral populations.

Unfortunately we do not have data from any closely related species to determine how old this haplotype is i. Node numbers indicate the number of scaffolds in which populations were partitioned into two sets. In a final analysis, we compared the site frequency spectrum SFS for the region on Super-Scaffold 73 to SFSs estimated for random sequences of the same length from throughout the genome.

SFSs for the random sequences are similar to expectations under neutrality, with a preponderance of alleles at low frequencies. Greater variance in SFSs are expected when selection makes use of standing variation because alleles have been recombining onto different backgrounds in ancestral populations Przeworski et al.

We conclude our study by examining the genetic architecture of migratory distance. BSLMMs are a form of genome-wide association analysis that includes a term for other factors that influence the phenotype and are correlated with genotype e.

Combined with the haplotype identified in the hapFLK analysis, which provides a signature of positive selection in short-distance migrants on Super-Scaffold 17 Table 1a , these loci represent good candidates for controlling migratory distance, but future analyses with a larger sample size are needed to confirm the robustness of this finding.

Direct information on migratory distance could also inform this analysis by allowing us to code the phenotype as continuous. Early research on blackcaps was pivotal for demonstrating the existence of a genetic basis of migration and studying its evolution. This is due in large part to the tractability of this species and its variability in migratory behaviour. Here, we have expanded this study system beyond phenotypic and marker-based approaches, launching it into the genomic era and conducting one of the most comprehensive genome-wide analyses of migration to date.

There is evidence for past gene flow between migratory and resident populations on the European continent but comparison of the contemporary structure of these populations suggests that gene flow may be limited. This is certainly the case for resident island birds. It has been suggested that one single genetic mechanism controls migratory traits and may be shared across broad taxonomic groups.

We do not find evidence for one common genetic mechanism across species here, and no protein-coding change is shared across the three focal traits propensity, distance and orientation that we examined in unison. Future work on gene expression may identify major regulators that control multiple migratory traits, and both NPY and bHLH transcription factors are good candidates.

Combined with the additional results that we presented here such as the importance of standing genetic variation , this information is vital for understanding how predictable the evolution of migration and other complex behavioural traits may be. Blackcaps have not only been relevant to work on the evolution and genetics of migration. Early work in this system suggested that differences in migration might serve as reproductive isolating barriers early in speciation.

For example, hybrids were shown to exhibit intermediate orientation behaviour that was predicted to be inferior because it would bring hybrids over large ecological barriers that pure forms avoid Helbig, b. More recently, it was shown that NW migrants arrive on the breeding grounds earlier than SW migrants, and that these birds mate assortatively on the basis of arrival time, helping to reduce gene flow between phenotypically distinct groups Bearhop et al.

The role of migration in speciation has gained considerable traction in recent years Rolshausen et al. Blood samples from two male blackcaps from the Mooswald breeding population at Freiburg im Breisgau, Germany, classified as medium-distance SW migrants on the basis of morphometrics and isotope signatures were used to assemble the reference genome.

Briefly, genomic DNA from one individual was used to sequence Illumina sequencing libraries fragment and mate pair libraries with insert sizes of 2, 5 and 10 kb. This assembly was improved several ways e.

These maps were used to super-scaffold HTS scaffolds. Statistics for the final assembly and each stage can be found in Supplementary file 2. We used SatsumaSynteny Grabherr et al. Gene prediction was performed using a de novo testis transcriptome of blackcaps and cDNAs from three avian species zebra finch, chicken and flycatchers from the ensembl database. Following MAKER, we obtained the predicted protein sequences to annotate genes functionally using blastp and Interproscan.

We obtained whole genome resequencing WGS data from male blackcaps including WGS data from the two individuals used to generate the reference genomes. High molecular weight DNA was extracted from blood withdrawn from the brachial vein, following a standard salt extraction protocol.

Individual samples were collected across the European breeding range including three island populations Canary Islands, Cape Verde, and Azores and covering the entire range of migratory phenotypes. Birds were sampled during the breeding season unless indicated otherwise.

We also obtained WGS data for five garden warblers and three hill babblers, the closest sister taxa to blackcaps, sampled during breeding Voelker and Light, We prepared small insert libraries using DNA from each individual and sequenced five samples per lane on NextSeq with paired-end bp reads.

All analyses made use of data from resequencing reads that were aligned to the reference genome using bwa mem Li and Durbin, or stampy in the case of the garden warblers divergence time of 0. BQSR requires a set of known variants. We used the second set of known SNPs common and high-quality from BQSR for this analysis and combined variants from all populations into a single vcf file for subsequent analyses.

All repetitive regions were excluded from our analyses and those focused on demography did not include the Z chromosome. In order to estimate unfolded SFS, we needed an ancestral sequence, or the ancestral state of variants segregating in blackcaps.

This sequence was generated using WGS from garden warblers and hill babblers. We used MSMC2 to infer the demographic history of blackcaps in our dataset. MSMC2 implements the multiple sequentially Markovian coalescent MSMC model, estimating effective population size by time and relative cross-coalescence rates between any two populations. It allows inference of the expansions and contractions of a population and of the extent and timing of population divergence Malaspinas et al.

Specifically, by running a hidden Markov model HMM along all possible pairs of haplotypes, MSMC2 estimates the free parameters for a demography model a series of effective population sizes as a function of segmented time and relative cross-coalescence rates between sequences using a maximum-likelihood approach.

We grouped medium NW, SW and SE and long-distance migrants because they exhibited very little population structure Figure 1 and indistinguishable demographic histories Figure 2—figure supplement 4 ; Figure 2—figure supplement 5 ; Figure 2—figure supplement 6.

We excluded any birds with less than 15x coverage. We used the bamCaller. We generated a global mappability mask file for the reference genome using GEM Derrien et al. Statistical phasing i. The analysis of cross-coalescence rates requires comparisons between groups and we considered all possible combinations of groups for our analysis Schiffels and Wang, In other words, we ran analyses for all 15 possible combinations three between groups on the continent, three between populations on the islands, and nine for comparisons between the three continent groups and three island populations.

For each pairwise combination, we ran the combineCrossCoal. This program permits the inclusion of two or more populations and accounts for both drift within populations different N e and covariance across them hierarchical structuring.

We ran this analysis for the complete dataset including all populations, and for a restricted dataset including only medium-distance migrants. We determined the number of clusters for each dataset separately using fastPHASE Scheet and Stephens, and the cross-validation procedure mentioned earlier. Once hapFLK is estimated, it is normalized using rlm in R, and p-values are computed from the chi-squared distribution. We used a permutation analysis to establish a threshold, beyond which genomic regions would be considered to be experiencing positive selection.

Specifically, we randomly shuffled population labels times, re-estimated hapFLK and p-values, recorded the lowest p-value for each randomization and set the threshold to the fifth percentile across randomizations. Once these regions were identified, we determined which population was experiencing selection by comparing branch lengths for a tree built using data from the entire genome and one built using data from the region under selection.

Note that results from analyses using medium-distance migrants are plotted using the resident phenotype for illustrative purposes, but the analysis was not run using these birds. We include birds from three resident continent populations — Cazalla de la Sierra and Gibraltar in the Iberian Peninsula along with Asni in Morocco only three birds were sampled from this African population, precluding its use in the present analysis; Supplementary file 4.

There is also some evidence in our PCA to show that this heterogeneity has led to some differentiation between populations, as birds from Cazalla de la Sierra exhibit values more similar to migrants on PC2 Figure 1c. Accordingly, to avoid any confounding effects from population structure, we limited our analysis to birds from Gibraltar. Results using Cazalla de la Sierra instead were very similar.

We followed methods described in Rochus et al. PBS is similar to F ST , but can include more than two populations and identifies regions within each population that exhibit differences in allele frequencies. This statistic was originally designed for three populations, but can be expanded to include more populations Zhan et al.

This equation is an example that was applied to resident populations R , where T is log transformed F ST between the populations indicated in exponents:. Recent papers have noted that F ST can be elevated by reductions in within-population variation alone and that there are many factors that can reduce variation within populations, including linked selection in areas of reduced recombination that may result from purifying selection background selection, [ Cruickshank and Hahn, ; Noor and Bennett, ].

It is unlikely that this process affects our results because recombination rate should elevate estimates of PBS in all populations, but this is not the case Figure 5a. Regardless, we followed methods from Vijay et al. Vijay et al.

F ST in focal populations would have to extend beyond that in non-focal populations to be considered important in generating the trait of interest. We used the same approach for PBS. In this last analysis, we focus on the affects that selection can have within a population instead. Specifically, selective sweeps can reduce variation at both the locus under selection and its neighbours Smith and Haigh, Local reductions in variation result in the presence of extended regions of haplotype homozygosity within phenotypes long haplotypes at high frequency.

In this way, nSL does not require a genetic map and is more robust to variation in not only recombination rate but also mutation rate. For this analysis, we used selscan v. We ran the data through fastPHASE first to phase haplotypes using 50 iterations of the EM algorithm, sampling haplotypes from the posterior distribution and using same number of clusters identified for hapFLK.

Alignment files are available upon request. Specifically, we used findMotifsGenome. HOMER includes known motifs for thousands of transcription factors mostly for model organisms ; we chose to focus on candidate transcription factors identified by previous studies as having an association with migration Ruegg et al. In our final analysis on migratory distance, we limited our dataset to short-, medium- and long-distance migrants.

They also include a term for factors that influence the phenotype and are correlated with genotype e. This is because our focal variable here distance is ordinal in nature and this fact would have been lost in hapFLK. We could not code this variable as continuous because the average distance individuals in each population travel on migration is not exactly known. All other data are included in the manuscript and supporting files.

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses. What follows is the decision letter after the first round of review. Thank you for submitting your work entitled "The evolutionary history and genomics of European blackcap migration" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor.

The reviewers have opted to remain anonymous. This manuscript compiles an enormous dataset to address questions about the genetic basis of migratory behavior in a classic model system in the field.

The reviewers and reviewing editor were enthusiastic about the topic and united in finding the dataset extremely impressive. They also thought several key results of the paper were important. In particular, the data add to growing evidence that there is not a common genetic mechanism underlying migratory behavior. Furthermore, the attempt to decompose a complex behavior into component parts propensity, distance, and orientation was novel and interesting.

Despite this enthusiasm for the work, a number of significant issues were raised about the current presentation of the data and some of the analyses. These major concerns are synthesized below. After extensive discussion, the revisions required for the paper to be acceptable were deemed to be too substantial to be completed in the two month period allowed for revisions at eLife. We have therefore recommended rejection of this manuscript in its current state.

Given the great promise of the dataset, however, the reviewers encourage submission of a significantly revised manuscript as a new submission in the future. The Introduction does not set up clear questions and gives little background information on why the blackcap system is uniquely well-suited for addressing questions about the genetic basis of migratory behavior.

The Introduction also does not set up the key findings of the paper. It is recommended that the authors re-focus the overall framing of the paper to highlight their key results, remove tangential analyses see below for specifics , and generally streamline and contextualize results within the existing literature throughout the entire paper.

These analyses were not clearly integrated with the rest the paper. The inference that migration is the ancestral state and was lost on islands is generally reasonable, as that is a pattern found in other taxa. The authors should also acknowledge the limitations of their inference more openly, given the weak support in their data.

Further, the overall contribution of the phylogeographic analysis to the main conclusions of the paper should be clarified throughout the text. To more accurately date divergence times, the authors should use a demographic modeling approach based on the joint site frequency spectrum e. Alternatively, the authors could refocus that section on variable demography rather than the timing of splits. Demographic trajectories diverging earlier than previously estimated split times could then be a "suggestive" point, but shouldn't be the primary finding of that analysis.

It might be useful to refer to Mazet et al. Particularly in the transition to residency section, outliers cannot be separated from population structure, and could therefore be due to a variety of factors unrelated to migration. The authors suggest that finding an elevated region on super scaffold 99 in both island and resident populations indicates this region may be associated with the transition to residency.

This analysis is rather unconvincing, as these shared elevated regions could be due to shared ancestry. It is also not clear if the extensive subsequent discussion of candidate loci focuses only on this shared elevated region on scaffold 99, or looks at all outliers.

The language in this section should tempered, alternative explanations discussed, and it should be made clear which populations are included when identifying candidate loci.

The first section about standing genetic variation dealing with short distance migrants is weakly supported- removal this section is recommended. The second section, if retained, should use methods that have been previously shown through theory or simulation to distinguish selection on standing genetic variation. It is also recommended that the authors articulate clear predictions about expected results if migratory behavior arises from selection on novel vs.

This is particularly important for explaining how phenotypes were assigned to "short-distance SW migrants," which fall into two different clusters on PC2. This is a major gap in the paper. The analysis of distance, if retained, should either apply the same approaches for outlier detections as were used for the analyses of propensity and orientation, or clearly explain use of a different approach. There should also be further justification for classifying distance as a categorical variable.

This paper uses the European blackcap, a classic model system in migration research, to reconstruct the evolution of migratory behavior and identify genomic regions associated with different components of migratory phenotype. To me, the most impactful results are that the authors do not find candidate loci associated with migratory behavior in other systems to be under selection in the blackcap, adding to growing evidence against a common underlying genetic architecture of migratory behavior.

Also important is that the authors looked at multiple components of migratory behavior propensity, orientation, and distance. However, while the dataset and analyses are very impressive, I think that the authors try to do too much at the expense of clarity and a focused message.

Several sections of results are unclear, key information about sampling is missing, there are many tangential and speculative sections, and the paper as a whole does not have a strong unifying framework or question. The Introduction, in my opinion, does not articulate a clear overarching question, and the questions that are laid out in the Introduction are not linked well to the analyses presented in the paper.

For example, the Introduction seems to set up a study focused on the genetic architecture of migration. However, the first part of the paper is a lengthy analysis of population structure and phylogeography that seems unrelated to any topics introduced in the Introduction and is not connected to the background provided on the system.

When we get to the genetic architecture component of the paper, it is not contextualized well with previous knowledge of the system, and it is therefore hard to evaluate the significance of the results. It is also not clear how the earlier pop structure and phylogeography component is related to analyses of selection and genetic architecture.

The only information provided is in the legend of Figure 1, which states that samples were collected on the breeding grounds with the exception of a few collected during the winter or during migration. This is problematic, given that the paper relies on individual-level phenotype assignments for these samples in all analyses.

Explanations of how migratory distance, propensity, and orientation are determined for individual samples are needed. The way the PCA is interpreted is not convincing and is not supported by any citations.

The purpose of this section overall is also not entirely clear- how do these analyses link back to overall questions about genetic architecture of migration? It is unclear to me how any of these data support variation in migratory behavior arising from standing genetic variation. Are the authors suggesting that individuals in the short-distance migrant population have different migratory strategies e. Later evidence for the role of standing genetic variation is more convincing.

The purpose of this section needs to be made clearer. However, these regions could just as reasonably be under other sources of divergent selection between these populations e. Couldn't shared regions under selection in island and resident populations also be due to gene flow and shared ancestry in these populations, as indicated by TREEMIX? Overall, I don't find it very convincing that regions under selection in non-migratory populations are directly associated with migratory propensity.

The authors should either temper this section and discuss alternative explanations, or make the support for their assertion clearer. What groups of samples were actually compared? Why not make these analyses comparable? This last section in general felt somewhat tacked-on and not well integrated into the rest of the paper.

The authors have a very cool dataset, but the parts that are novel and exciting are not highlighted very well throughout. The amount of background information given seems to assume familiarity with the blackcap system, which limits the accessibility of the paper to a broad audience.

I think more effort needs to be put towards explaining why this system is so well-suited for asking these questions, and what new things we learn about migratory behavior from these analyses. In this manuscript, the authors use whole-genome data to study the genetic basis of three migratory traits in European blackcaps: the propensity to migrate, distance, and orientation.

Leveraging the diversity of migratory strategies in blackcaps, the authors document associations between migratory behavior and SNPs in regulatory regions — providing the first genome-wide characterization of migratory behavior in this species.

Not only is this an interesting study system, but the sampling and experimental design also provide a robust foundation for investigating the genomic basis of migratory behavior. Overall, I thought it was an interesting data set and enjoyed reading the manuscript.

It is my opinion that, if these concerns are addressed, this work would make a nice contribution to the existing literature. I would recommend that you narrow this section down to focus on fewer, but more definitive, analyses, as it would strengthen the overall argument.

I think you could remove this section as I am not really sure it adds much to the paper. I think the fact that variant calling in GATK is combined with other programs makes this a robust approach but was wondering why you chose UnifiedGenotyper. May be worth mentioning somewhere in the results.

Figure 5: I think the figure is very aesthetically pleasing, but I am really struggling to understand what is going on in this figure. It may just be a matter of clarifying the legend, or perhaps it is just me. But I am not convinced this figure adds much. Also, I think you mean to say "panel f shows haplotype.

In this study the authors analyze genetic variation in what has historically been the model species for songbird migration and present a set of results covering both phylogeographic history and selection on genes putatively associated with migratory behaviors. Helbig and Berthold's captive breeding studies in black-caps showing that migratory orientation is heritable are classics in ornithology, and an obvious followup question is "what genes are responsible for this behavior?

From that perspective I think one of the most interesting thing here is that this is now the third study fourth if you count the brand new Toews et al. PNAS study looking for associations of specific genomic regions with migratory behavior, and as far as I can tell there are zero overlapping outlier regions across these studies. The authors make this point but I think it could be emphasized, particularly given that much of the interest in migration as a tractable trait for genetic mapping is driven by the captive breeding studies conducted in this species.

In general I thought the new data in this paper was very good, but the results presented cover so much territory that individual analyses sometimes feel rushed and the whole picture is hard to follow. One general issue that should be addressed more directly when describing results for regions putatively associated with migratory behaviors is that migration itself may be tangential to the actual selective force — climatic variation on either the breeding or wintering range, dietary differences across geographic regions, or any other environmental factor varying among populations could create the basis for selection that could be detected by approaches like PBS.

Because migratory phenotypes covary with many of these other environmental factors, it is inevitably going to be difficult to identify genes that drive rather than being driven by aspects of migration like orientation, distance, or phenology.

That being said I think this is an important paper in the area of migration genomics because it is in the only system with truly compelling captive-breeding results from crosses and because the dataset is excellent. What about if PCA is run on just continental, or just migrant populations alone?

Because PCA will represent variance in the full dataset, running it with divergent island populations may mask differentiation among migrant populations.

Probably same issues in the admixture analysis. I'd like to see a supplemental figure showing at least PC when the analysis is run on just continental birds. As it is the clearest evidence seemed to be from the phylogenetic analysis, which relies on poorly supported nodes in a topology constructed using a method that effectively assumes no gene flow.

In addition, the mutation rate and generation times though they look about right relative to other migratory birds don't appear to be pulled from the citation listed Noor and Bennett, — apologies if I missed it buried in there somewhere , and these will directly scale the inferred timing of population size changes. I'd suggest demographic modeling of the joint site frequency spectrum in dadi or moments.

In addition, the end of this section regarding the possible speed of evolution of variation in migratory traits should acknowledge that much of that literature is based on the documented contemporary evolution of the NW migration and not on biogeographic reconstruction of island colonizations. Or is it just that there is one bHLH motif found in an outlier region? If the latter, I'd suggest cutting the second half of this section as the evidence of any specific regulatory element being involved seems quite weak.

This point could be more generally made in the Introduction, as well. I suggest cutting this section. I think it needs a rewrite, and the methods should be better justified. I suggest starting with a paragraph laying out expectations for what you expect given selection from novel vs standing variation see Barret and Schluter, , and using only analyses that have been previously proposed and shown through either theory or simulation to distinguish these processes.

A matrix of pairwise genetic distance over the whole genome? Please provide a little more detail on your implementation of this method. Figure 1C: These colors seem to partially but not entirely match the map, which is confusing especially when referring back to this figure for color references. I'd use different colors than in the map. Figure 5 legend: the letters seem to be off here — please check that e and f are correctly referenced, and discuss panels in order a-e.

Thank you for resubmitting your work entitled "The evolutionary history and genomics of European blackcap migration" for further consideration by eLife.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:. All three reviewers found the manuscript much improved.

Only reviewer 1 had a substantial outstanding comment that requires clarification. After consultation, the other two reviewers and the Reviewing Editor concur with the points raised by reviewer 1. Please address the below questions about how individuals were assigned to different phenotype groups in the different analyses. The manuscript is provisionally recommended for acceptance pending satisfactory revision on these points. The authors are also encouraged to consider the minor points raised by all three reviewers.

I find the resubmitted manuscript to be a substantial improvement over the initial submission. The authors are to be commended for their thorough revision and for incorporating reviewer comments. Let me better explain my confusion: in the analyses of population structure Figure 1 Fst plot , there are nine populations with distinct phenotypes: continental residents, short distance sw migrants, long distance SE migrants, medium distance SW, SE, and NE migrants, and the three island populations.

However, in subsequent analyses, these populations are grouped together in different ways that are not clearly explained. For the demographic analysis, the medium- and long- distance migrants are grouped together, despite variation in orientation among these populations. Why is this? Likewise, in the analyses of the genetic basis of traits, the authors describe analyzing all three phenotypes orientation, distance, and propensity ; however, each of these of course has multiple categories e.

In the legend for Figure 3, they note that they are grouping birds into 5 phenotypes for panels A and B and 3 phenotypes for panels B and C. The authors note that islands were excluded and residents limited to a single population, but it is otherwise unclear to me what these different phenotype groups are.

Adding to the confusion is that 1 the analysis of differentiation and selection comes after the demographic modeling, which did not consider orientation in the grouping of phenotypes; 2 Figure 5 also combines the medium and long- distance migrants and removes orientation; and 3 sample sizes are not given for each population used in the comparisons.

Some clarification of exactly which populations and phenotypes are being compared in the analyses, throughout this this section is needed, as well as justification for why medium- and long- distance migrants are combined for some analyses but not others, and clear reporting of sample sizes for each comparison in the main text please remove from legend in Figure 3.

The authors have done a commendable job revising this manuscript. The Introduction is much easier to follow and sets up the study nicely. I appreciate the effort that has gone into addressing reviewer comments. The authors have done a really excellent job with this revision and I'm pleased to recommend acceptance. They have trans- continental breeding distributions in temperate- and high-latitude nearctic forests, and wintering distributions centred on neotropical North and South America.

Migration is a complex mode of dispersal, promoting the colonization of new areas, but also their regular re-colonization and gene flow. Spatial segregation — the linchpin of most speciation theory — becomes less and less likely with increasing migratory tendencies.

Achieving true geographic isolation from other populations, thereby allowing differentiation to occur in the absence of gene flow, seems particularly unlikely among long-distance migrants, whose movements regularly encompass entire continents and oceans. But here we have a conundrum: while migration opens the door to differentiation in new ecological and geographic space, it apparently slams it shut again through denial of geographic isolation and the promotion of gene flow.

Until now, migration was considered to counter differentiation Scenarios proposed to explain migrant speciation have had to invoke geographic isolation and, by implication, mechanisms such as lower historic levels of migration and greater levels of natal philopatry — neither of which fits the evidence 11 , 12 , Although it is true that the origins and losses of migration have occurred independently in many lineages, it is unrealistic to suggest that the associated complex life-history characteristics were somehow held in temporary abeyance across entire lineages or clades.

Recent developments in speciation theory 3 , 4 offer a theoretical framework to escape such ill-fitting scenarios, and species flocks of migrants could provide a testing ground for these theories. Phenotypic evidence in birds suggests that sexual selection may operate only as a distant second to resource competition and perhaps reinforcement adaptations to prevent hybridization in driving speciation events among many migratory animals.

Darwin, C. Google Scholar. Mayr, E. Book Google Scholar. Kondrashov, A. Nature , — Dieckmann, U. Grinnell, J. Auk 39 , — Article Google Scholar. Helbig, A. Ibis , — Richman, A. Evolution 50 , — Groth, J. Montgomery, T. Jr Am. Dilger, W. Auk 73 , — Mengel, R. Living Bird 3 , 9—43 Cox, G. Download references. You can also search for this author in PubMed Google Scholar. Correspondence to Kevin Winker. Reprints and Permissions.

Winker, K. Migration and speciation. Nature , 36 Download citation. Issue Date : 02 March