The best of both worlds: Shortcutting evolution through adaptive hybridization in hares


 Brown hares (Lepus europaeus Pallas) are able to hybridize with mountain hares (L. timidus Linnaeus) and produce fertile offspring, which results in cross-species gene flow. However, not much is known about the functional significance of this genetic introgression. Using targeted sequencing of candidate loci combined with mtDNA genotyping, we found the ancestral genetic diversity in the brown hare to be small, likely due to founder effect and range expansion, while gene flow from mountain hares constitutes an important source of functional genetic variability. Some of this variability, such as the alleles of the mountain hare thermogenin (uncoupling protein 1, UCP1), is likely of adaptive advantage for brown hares, whereas immunity-related MHC alleles are reciprocally exchanged and maintained via balancing selection. Our study offers a rare example where an expanding species can ease its expansion load through hybridization and obtain beneficial alleles to shortcut evolutionary adaptation to the novel environmental conditions.


Introduction
Understanding of the ecological and evolutionary processes underlying species' dispersal and range limits is a fundamental theme in evolutionary biology 1 . The leading edge of the species range expansion is typically populated by few individuals, which are also living at the boundary of their speci c habitat preferences. Low population densities result in increased genetic drift which typically has deleterious effects on population tness as a result of the so-called expansion load 2 . Together with poor adaptation to the local environment, these expanding populations are vulnerable to local extinctions, which might cause periods of range contraction before the species establishes 1 . Although range expansions are often associated with introduced invasive species, there are also numerous examples of naturally occurring range shifts, especially under current global climate change 3 . This is also the case with brown hares (Lagomorpha: Leporidae, Lepus europaeus Pallas), native to Central Asia and mainland Europe, that are currently expanding their range northwards 4,5 and have reached approximately the 65.3˚N in the Nordic countries 6 . The species arrived in Finland naturally from the south-east as early as the 19th century 7,8 , but has experienced an impressive 300 km range expansion to north in only the last 30 years, quite likely facilitated by climate change 9 .
As brown hares frequently hybridize with the resident mountain hare (Lagomorpha: Leporidae, Lepus timidus Linnaeus) an arctic/subarctic species that has occurred in Fennoscandia for approximately 10,000 years 10 , we hypothesized that hybridization may have substantial evolutionary implications for the two species. In the present study, we compare levels of genetic diversity among the Finnish mountain hares and brown hares with Central European brown hares. Specimens from both Finnish hare species represent a hybrid population, with varying degrees of introgression, whose intensity roughly correlates with the invasion front of the brown hare's northward expansion (Fig. 1). We are particularly interested in investigating whether the Finnish brown hares have obtained signi cant levels of adaptive allelic variants from mountain hares. In principle, the cross-species gene ow has the potential to not only alleviate expansion load but to also allow the expanding species to co-opt bene cial, locally adapted alleles assisting brown hares to secure a foothold in the new environment. Adaptive advantage of the local alleles should manifest as overrepresentation of mountain hare alleles in brown hares compared to the background introgression level.
One of the candidates is thermogenin (uncoupling protein 1, UCP1). UCP1 is a mitochondrial inner membrane channel protein, which channels protons, bypassing ATP-synthase and converting the energy stored in the mitochondrial membrane potential to heat instead of ATP 11 . UCP1-mediated heat production is especially important for small and adolescent mammals 12 . In fact, positive selection on certain UCP1 variants has also been suggested to confer climatic adaptation in humans 13 . As an arctic/boreal species, mountain hare speci c UCP1 alleles are expected to be cold adapted in contrast to those of the brown hare, which originate from a much milder climatic region. Other evolutionarily signi cant genes are those of the major histocompatibility complex (MHC) class II DQA and DQB loci, which are critical for adaptive immunity. As MHC class II receptors are required for the pathogen recognition, a broader range of genetic variants in a population should offer protection against a broader range of pathogens and parasites 14 . As the brown hares at the species' expansion front are likely to show little ancestral genetic variation due to the founder effect, the mountain hare/brown hare hybrids could bene t from the rare allele advantage at the MHC loci [15][16][17] , where the rare alleles have been obtained from the mountain hare. An additional bene t for analyzing the MHC loci is that they typically comprise many alleles in all populations, making them suitable proxies to estimate overall genetic diversity of a population [18][19][20] .
In addition, we analyzed Toll like receptor 2 (TLR2), a gene involved in innate immunity, whose allelic variants have been suggested to explain differences in pathogen resistance between various wild mammal populations [21][22][23] , as well as the succinate dehydrogenase subunit A locus (SDHa). SDHa encodes for a subunit of mitochondrial oxidative phosphorylation (OXPHOS) complex II (CII), which is the only OXPHOS complex encoded entirely by nuclear genes and therefore avoids any nuclear-mitochondrial incompatibility issues 24 . These in turn might cause mitochondrial DNA (mtDNA) driven selection on nuclear encoded mitochondrial genes. As a functionally conserved housekeeping gene, we expect the SDHa locus to be evolutionarily neutral and serve as a proxy for the introgression of nuclear genes in general.
The hybrid zone interaction between mountain hares and brown hares provides an ideal opportunity to test the notion that an expanding species may bene t by adopting locally adapted genetic variants from the resident species, thus shortcutting evolutionary adaptation to the new environment while simultaneously mitigating the effects of expansion load. We hypothesise that the mountain hare UCP1 alleles represent local adaptation that should confer a selection advantage in the Finnish hybrid zone brown hares and that MHC class II allelic diversity should be under balancing selection and thus broadly shared between the two species.

Results
The overall genetic diversity (GD) as computed from the polymorphism data of all loci 25 was 0.86 ± 0.67 for the Finnish mountain hares (n = 147) and 0.65 ± 0.45 for the Finnish hybrid zone brown hares (n = 173). Interestingly, GD was lowest (0.53 ± 0.28) for the allopatric Austrian brown hares (n = 48) used as a reference to detect brown hare speci c variation. Much of the difference in genetic variation among Finnish hares can be attributed to the MHC class II genes, as DQA and DQB loci were dominated by one common allele each in brown hares (Leeu-DQA*006 = 0.49; Leeu-DQB*001 = 0.56), while mountain hares showed more even genotype representation and higher number of alleles compared to brown hares (Fig. 2, Table 1, Table S3). In addition, the levels of heterozygosity (Hz) were higher in mountain hares than brown hares for both MHC class II loci whereas the other nuclear loci showed higher Hz levels in brown hares (Table 1). Remarkable 83% and 70% of the DQA and DQB variation was explained by variation within individuals, whereas for TLR2 or SDHa the variation was mainly between the species ( Table 2). The introgression of mtDNA as well as TLR2, SDHa and neutral SNP alleles was notably different between the species, with 0.60-0.85 of the alleles being unique in brown hares, compared to 0.89-0.98 in mountain hares (Table 1). For both species, sympatric populations showed more allele sharing at all loci compared to allopatric populations (Fig. 2). Allele sequences have been submitted to GenBank and the accession numbers and project data are available online (see Data Accessibility).  For the loci with more than three alleles across the entire dataset, we performed phylogenetic analyses to correlate the clustering of related alleles with their frequencies in the two species and four population groups (Figs. 2 and 3). We also estimated their divergence dates to compare the allele ages with the estimated speciation event circa 5MYA.
UCP1, TLR and SDHa loci showed clear ancestral allele frequency differences (two clades, each dominated by one species) between mountain hares and brown hares (Fig. 2). Notably, the UPC1 locus exhibited an allele appearing in high frequency in the Finnish hybrid zone brown hares, but not detected in either of the allopatric populations (UCP02). Evolutionarily however, this allele is well embedded in the brown hare clade but has presumably been lost or was unsampled in Austrian brown hares.
In contrast, the MHC loci showed high levels of variation as well as considerable allele frequency differences between species and population groups ( Fig. 3A and B). Hybrid zone brown hare individuals retained most of the allelic variants found in both ancestral species and signi cantly higher levels of diversity than allopatric brown hares (compare the within individuals vs between populations variance in Table 2). As well as being a signature of introgression, the shared variation between mountain hares and brown hares is also typical for trans-species polymorphisms (TSP), which is the persistence of allelic lineages beyond speciation events and has often been associated with MHC loci 26 .
Hybrid individuals shared MHC alleles that occurred only in pure brown hares and alleles that only occurred in individuals assigned as "pure" mountain hares ( Figure S1A and B). Only two alleles each from DQA (Leeu-DQA*001 and Leeu-DQA*010) and from DQB (Leeu-DQB*001 and Leeu-DQB*007) were shared across allopatric and sympatric populations. There were however, nine additional DQA alleles and 13 DQB alleles that were shared only between hybrids and either parental species. We take this as evidence that, despite detecting the expected signal of TSP, there is additional strong introgression of MHC alleles in both directions from the parental lineages. While many of the DQA alleles were previously known 27,28 , all of the DQB variants were novel discoveries in mountain hares and only three had been seen previously in brown hares (Table S3).
Detection of loci under selection using coalescent simulations for locus-speci c F-statistics 29 , showed DQA, DQB and UCP1 were strongly selected for in both species (Table S4) Fitting with the role of UCP1 in cold tolerance, mountain hare speci c alleles are more common in hybrids in the higher latitudes (Fig. 4B), although this tendency is not statistically signi cant.

Discussion
The Finnish brown hare population has been growing in size and increasing its range northwards during the past few decades 6 . This development offers an interesting natural setting to study the population genetics and evolutionary adaptation of expanding species. The small pioneering populations at the edges of the distribution are in uenced by founder effect and genetic drift, causing the edge populations to have lower genetic diversity than in the core populations -a phenomenon known as expansion load 2 .
As asymmetric gene ow from mountain hare to brown hare is well established [30][31][32][33][34] , we sought to see how this introgression contributes to the genetic makeup of the Finnish brown hares and whether the mountain hare alleles could have adaptive signi cance for the brown hares.
From our analysis, it was apparent that much of the genetic makeup of Finnish brown hares is heavily in uenced by the introgression of alleles from mountain hare (  Figs. 2 and 3). If only the alleles that are shared between Austrian and Finnish brown hares (Fig. 2, Table S2) were taken into account, the GD of Finnish brown hares would be 0.40 ± 0.37. This is notably lower value than observed for the Finnish (0.65 ± 0.45) or Austrian (0.53 ± 0.28) brown hare populations. It should be pointed out that this is an estimate, as the Austrian population is not directly ancestral for the Finnish brown hares and because the origin of the MHC alleles cannot be traced based on their phylogeny. As another tell-tale sign of a founder effect, the most diverse loci DQA and DQB loci were dominated by one common allele each in brown hares (Leeu-DQA*006 = 0.49; Leeu-DQB*001 = 0.56), while mountain hares showed more even genotypic representation and higher number of alleles (Fig. 3, Table 1, Table S3), resulting also in higher levels of heterozygosity in mountain hares than brown hares for both MHC class II loci. In contrast, the other nuclear loci showed higher levels of Hz in Finnish brown hares, which can be explained by an asymmetric ow of alleles at these loci from mountain hares to brown hares, as alleles dominant in mountain hare were commonly present in brown hares but not vice versa (Table 1, Figs. 2 and 3, Table S3).
Taken together, these ndings show that brown hares commonly obtain genetic variants from mountain hares and suggest that this plays a role in suppressing the effects of expansion load. Interestingly, some mountain hare alleles were also detected, albeit very rarely, also in the allopatric Austrian brown hare population (Figs. 2 and 3).
Our prime candidates to detect potentially meaningful adaptive introgression in hares included the UCP1 gene, which could have a potential role in adaptation for a colder climate 35,36 , MHC class II loci DQA and DQB, which confer general pathogen resistance and are known to be under strong balancing selection favoring genetic diversity in various chordates (e.g. 28,37−39 and TLR2, required for innate immunity 40 . SDHa was chosen as potential neutral locus together with mitochondrial DNA. The conducted outlier analyses detected moderate balancing selection acting on UCP1 and strong balancing selection at both MHC loci (Fig. 4A), while the maintenance of introgressed alleles at TLR2 and SDHa likely re ects neutral processes similar to those at SNP loci ). This is also evident in the AMOVA analysis, which showed that the majority of the variation in neutral alleles was between the species, while the loci under selection showed signi cant within individual variation (Table 2). However, the fact that the Bayesian outlier analysis suggests balancing selection for UCP1 is quite likely an artefact caused by the dynamics of the hybridization and selection progress. Alleles under purifying selection are identi ed from their tendency for xation, while in the case of the Finnish brown hares, there is a constant input of ancestral alleles through hybridization, which are encountered by differential selection pressures. Revealingly, the mountain hare speci c alleles were represented in Finnish brown hares at higher levels than observed for the neutral loci (Fig. 2, Table 1, Table S3). This results in rather even distribution of mountain hare alleles in the Finnish brown hares (Fig. 4B), which mimics the effects of balancing selection. In fact, it is well known that demographic factors of a population can speci cally increase the false discovery of balancing selection in Bayesian outlier analysis 41 .
Our interpretation is that although some brown hare alleles were detected also in mountain hares, the dominance of mountain hare alleles in both species suggests that selection is favoring locally adapted alleles at the expense of outbred alleles per se. This is evident also from the fact that the Austrian allopatric brown hares have a surprising allelic diversity at the UCP1 locus, including four alleles that are shared with the Finnish hares and six private alleles, while the UCP01 allele is almost the dominant allele in both species in Finland (Fig. 2).
Adaptive thermogenesis in brown fat is dependent on the UCP1 and is essential for all neonate mammals 42 . At birth, the newly born mammal needs to rapidly adapt from the maternal body temperature to the environmental temperature and this process is developmentally controlled. We chose to investigate the variation in UCP1 speci cally as both mountain hares and brown hares do not have nests or lairs, but leave their young unattended after birth and typically feed the leverets only once per day. This means that especially the spring generation, often born on snow, is exposed to cold and variable weather conditions of a boreal climate and is independent of the mother for shelter or thermoregulation. Under these circumstances, any cold adapted alleles for NST are likely to make a difference for early season breeding for brown hares. Therefore, we nd it credible that positive selection explains the excess of mountain hare UCP1 alleles in the Finnish brown hare population. As most of Finland is at the northern edge of brown hare distribution, a steep allelic gradient of the mountain hare UCP1 in the brown hare population is not necessarily expected (Fig. 4B). However, there can be also trade-offs with the cold-adapted alleles, such metabolic ine ciency 43 , which could balance out the bene ts when food is scarce or poor quality. Unfortunately, we can currently only hypothesize about the physiological differences between the different alleles, but these might be possible to investigate in future studies.
While balancing selection favoring allelic diversity is well established for MHC loci, the directionality of the gene ow between the two species is di cult to dissect. As pointed out, while Finnish brown hares have a dominant allele in both DQA and DQB loci, the same alleles are also abundant in mountain hares (Fig. 3). Similarly, there is evidence for gene ow from brown hare to mountain hare at the studied nuclear loci and mtDNA, although this is dwarfed by the introgression to the opposite direction (Table 1, Fig. 2). 26,44 , which are known to occur even among different genera of leporids 45 . For example, TSP could explain the existence of common brown hare DQA and DQB alleles in the Finnish mountain hare population (Fig. 3, Table S3) but without samples from highly isolated populations e.g., in the arctic islands, the distinction from recent introgression cannot be easily resolved. Similarly, our phylogenetic analysis was inconclusive as the DQA and DQB alleles do not cluster as species-speci c (or one species dominated) clades, unlike UCP1. The timing of the nodes on the tree also gave rather young approximations for the ages of the clades when compared to the speciation estimate. This is likely due to the fact that the applicability of fast evolving sequences as molecular clocks saturates at deeper divergences 46 . However, the allopatric Austrian brown hares had four private DQA alleles and seven DQB allelles, while sharing only four and three alleles respectively with the Finnish hares (Fig. 3). We take this as evidence that the alleles shared among the Finnish and Austrian populations could represent TSPs or a very ancient introgression event, whereas the Finnish hares have universal allele sharing via bidirectional gene ow, as is also seen for the other investigated loci. The fact that there were no DQA or DQB private alleles in the allopatric Finnish mountain hares (Fig. 3), suggests that the Finnish mountain hares form a uniform, genetically connected population 34 and that any bene cial alleles spread quickly in it 47 . Similarly, balancing selection involving novel allele advantage has been suggested to have promoted the introgression of MHC alleles also in other vertebrates [48][49][50][51] . As a follow-up, it would be interesting to correlate the MHC genotypes with susceptibility to speci c pathogens or parasites.

MHC loci are also notable for examples of trans-species polymorphisms
In Finland, the distribution of brown hares is rather precisely limited by 150 days of snow cover, a boundary that has steadily retreated northward during the last three decades due to climate change 6 .
Although capable of replacing mountain hares through direct competition 4,52 , brown hares are dependent on anthropogenic habitats in Finland and they are thought to be unable to penetrate the boreal forest ecosystems. The situation may be more complex regionally however, as brown hares can exploit road networks in their expansion and there are indications that the species may adapt to forested habitats in Nordic countries 53 . Our data give some insight into how adaptation may be occurring at the genomic level and underlines the advantage conveyed by the ability to co-opt genic variants from the resident species to open up otherwise inaccessible habitats. Climate change may facilitate this process by relaxing local selection pressures, such as the ones caused by the snow cover.
The recent decline in mountain hare numbers in Southern Finland is likely due to not only increased competition with the adapting brown hare, but also because of the shortening of the snow-covered season 33 , resulting in the camou age mismatch of the white winter pelage and increasing predation mortality 54 . Further work incorporating coat color loci 55 and gene expression may be informative on this front and indicate if plasticity exists which might assist the persistence of mountain hares. Although the mountain hare has shown remarkable resilience in the past 56 , the extent of the combined effects of the current climate changes are likely to be complex and their outcome for the species is unknown.

Sample collection and DNA isolation
The Finnish hare samples were obtained from hunters and the Finnish Food Authority, as well as from a GPS/GSM telemetry study 57 . The 48 Austrian brown hare samples came from a previous study of MHC diversity in central Europe (Smith et al. 2011).
To estimate the allelic diversity and degree of introgression between the two species, we performed targeted sequencing of candidate loci likely to be under selection in hares in a total of 368 individuals: 22 allopatric mountain hares from northern Finland, 48 allopatric brown hares from the species core range in Austria and 298 sympatric specimens of both species from within the hybrid zone in central and southern Finland (Table S1). Because there are no allopatric populations of brown hares in Finland, the lowland eastern Austrian brown hares were chosen to monitor brown-hare speci c allelic variation that would be free from mountain hare genetic in uence. As these hares otherwise represent a rather distant population of brown hares compared to the Finnish ones, the Austrian hares have not been included in the general population genetic analyses (genetic diversity, population structure). We also had previously obtained 9 genotype data for 6833 SNP loci for 36 specimens of the Finnish hares (Table S1), which was used here to detect selection signatures (see below).
All specimens have been initially identi ed using morphological features. The species differ considerably in their habitat preferences and identi cation during most of the hunting season and are easily distinguished due to the white winter pelage of mountain hares. DNA from ear muscle biopsies was isolated using Chelex® 100 (Bio-Rad) method 58 , following manufacturers' recommendations. The maps of sample locations were drawn in R v4.0.2 59 using ggplot2 60 Fig. 1, Table S1) using a previously published PCR-RFLP method 32 . Brie y, the PCR product of the mitochondrial Cytb locus was digested with DdeI, which cuts mountain hare sequence twice and brown hare sequence four times. The distinct sized restriction fragments are then unequivocally identi able on agarose gel.

Next generation sequencing of candidate loci
The ve candidate loci analyzed in this study are: a fragment of exon 2 including the antigen binding motif of the major histocompatibility complex (MHC) class II loci DQA (218 bp), and DQB (210 bp), which have been previously reported to show both spatial as well as temporal positive selection in hares 28,37 ; a 372 bp fragment of the Toll like receptor 2 corresponding to the genes for Toll-interleukin-1 receptor domain protein (TLR2), involved in innate immunity; a 358 bp fragment of the uncoupling protein 1 (UCP1); and nally a 135 bp fragment of the neutral locus, succinate dehydrogenase complex subunit A (SDHa). As an additional control for neutrality, we had previous genotype data for 6833 neutral SNPs from 36 specimens 9 (Table S1). Brie y, library preparation was performed by rstly amplifying each sample using the primer pairs from Table S2. A second round of PCRs was performed to attach unique DNA barcodes to all samples and achieve compatibility with Illumina ow cell chemistry. PCR products were then puri ed, and after estimation of amplicon concentrations, all samples were normalized, pooled and sent to the Microsynth (AG) for sequencing on an Illumina Miseq as 250 bp paired-end. Initial amplicon data processing was achieved as outlined in Biedrzycka et al. 61 and Sebastian et al. 62 using the different amplicon sequencing analysis tools available at: http://evobiolab.biol.amu.edu.pl/amplisat/. Allele calls for the low polymorphism loci (TLR2 and SDHa) were validated via Sanger sequencing in both directions, using the same primers as for their ampli cation. Genotype data for each sample is provided in Table S1. Allele frequencies and the GenBank access numbers for the sequences are given in Table S3.

Phylogenetic analysis of candidate loci
To identify putative species-speci c alleles, phylogenetic analysis was performed for DQA, DQB and UCP1. TLR2 and SDHa were excluded as they had only three alleles each with clear-cut separation between the species (Table S3) Detection of loci under selection was performed using two different outlier detection approaches incorporating genotypes from the candidate loci combined with data from 6833 SNP loci available for a subset of 36 samples genotyped in a parallel study 9 . The rst method was the F-statistics test based on genetic structure 29 , also provided in ARLEQUIN version 3.5.2 68 , with a hierarchical island model applying 50,000 simulations, 100 demes per group for 10 groups. The method was chosen as it allows the population sample to be assigned to different a priori groups (mountain hare, hybrid), allowing for different migrations rates between demes within groups and between groups. The latter is important because of the highly asymmetric nature of the gene ow between the two hare species. The simulation generates a joint distribution of F st versus heterozygosity (Table S4). Loci that fall out of the 99% con dence intervals of the distribution are considered as being under selection. A further test of selection was conducted via the Bayesian approach of Foll and Gaggiotti (2008). The advantage of this approach is that it is amenable to small sample sizes, with a risk of low power but no particular risk of bias due to underrepresentation of sampled demes.
Addressing trans-species polymorphisms in hares Trans-species polymorphism (TSP), the persistence of allelic lineages across species events, should be distinguishable from introgression if the shared alleles also occur in pure parental lines of both species. This is not trivial, as it is di cult to identify isolated populations of Finnish mountain hares or brown hares to select as "pure" parental lines. Nevertheless, we conservatively identi ed 22 individuals that could be considered pure mountain hares: most northerly samples with mountain hare mtDNA and no brown hare characteristic alleles at UCP1, TLR2 or SDHa loci. For brown hares, we used samples from two previous studies 28,69 on lowland Austrian brown hares, whose populations have much longer evolutionary history of isolation from mountain hares. The samples are marked in the Table S1. Under a TSP hypothesis, we would expect little population differentiation based on population assignment due to a large number of allelic variants surviving the Lepus timidus/europaeus species diversi cation event, circa 5 MYA (1.2-10.1 MYA) [70][71][72][73][74] . We performed an Analysis of Molecular Variance (AMOVA) ( Table 2) to test the level of variance within each population using GenAlEX version 6.5 75 with 999 random permutations. BEAST v.2.5.1 76 was used to calculate an ultrametric phylogenetic tree for each of the MHC loci and UCP1, implementing a HKY substitution model, a strict clock model assuming a mutation rate of 10% for DQA and 11% for DQB, and a Yule tree prior. The MCMC chain was run for 10 million steps with tree sampling every 1000 steps, and a 10% burnin fraction was used when calculating the nal MCC tree with common ancestor node heights in TreeAnnotator v.1.10.4. Figure 1 Hybridization among Finnish hares. (A) Distribution of samples used in the study. The level of mtDNA introgression roughly correlates with the expansion front of the brown hare 6. The allopatric mountain hare samples north from the line were assumed as "purebred" for the purpose of the TSP analyses. Both species coexist south of the line, although mountain hare abundance decreases notably towards the south-west. (B) Excerpts from game camera images from the same location in Joensuu, Finland in April 2014. Typical brown hare in the left and mountain hare in the middle. Individual on the right was later captured and con rmed as a rst-generation hybrid (sample 843), having brown hare mtDNA, but being heterozygous for the tested nuclear loci (Table S1). The individual was originally assigned as a brown hare based on morphology and mtDNA genotype Ultrametric tree of UCP1 and allele frequency bar chart of TLR2 and SHDa allele frequencies and mtDNA haplotypes (timidus or europaeus) in hare populations. The bars over the tree nodes represent the 95% con dence interval of the node height. The colors of the allele frequency bar chart represent each of the populations tested. Venn diagrams depict the numbers of unique and shared alleles for each population (same coloring as for the allele frequency bar chart). Private allele numbers in bold. μDQA=10% and μDQB=11%. The bars over the tree nodes represent the 95% con dence interval of the node height. The colors of the allele frequency bar chart represent each of the populations tested. Due to the scale of the allele frequency not all bars are observable, for all allele frequencies refer to Table S3.

Figures
Venn diagrams depict the numbers of unique and shared alleles for each population (same coloring as for the allele frequency bar chart). Private allele numbers in bold. Figure 4