Multilocus genetic structure within and among species
Despite extensive allele and haplotype sharing among white-headed gull species, genetic substructure was observed within and among species across all marker types. Species and populations at high latitude exhibited lower genetic differentiation then their southern counterparts. Furthermore, individuals breeding at northern latitudes clustered together regardless of species designation, consistent with contemporary hybridization. At least two haplotype/allele groups were observed at each locus; however no haplotype/allele group was represented by a single species at any of the loci. Private alleles and haplotypes were observed for most species at most loci; however, private alleles were only observed in two or three nuclear introns, respectively, for L. glaucoides and L. thayeri (Fig. 2).
Genetic evidence for contemporary hybridization among northern populations of Arctic white-headed gulls is corroborated by field reports (e.g., L. glaucescens×L. hyperboreus; L. argentatus×L. hyperboreus; L. argentatus×L. glaucescens; L. glaucescens×L. schistisagus; L. argentatus×L. schistisagus; L. glaucoides×L. thayeri; Olsen and Larsson 2004 and citations therein). Hybridization would be expected to homogenize allelic frequencies by locality, as neutral loci will remain similar because of introgression and recombination (Mallet 2005). Species appear to have been isolated long enough to have accumulated unique mutations, as indicated by the partitions in the nuclear and mtDNA genomes. Therefore, we contend that hybridization has occurred only recently in Arctic white-headed gull evolutionary history, likely from secondary contact following contemporary range expansion. Introgression of species-specific alleles may be maintained through local adaptation to intermediate habitat types where species coexist, as hybrids have been reported to display adaptive traits of both parental species (L. glaucescens×L. occidentalis; Good et al. 2000).
Recent speciation and contemporary hybridization likely both play a role in the magnitude of allele and haplotype sharing observed among white-headed gulls. Of particular interest is the extent of introgression/hybridization occurring at the northern limits of species’ ranges and among white-headed gulls that breed exclusively at high latitudes. Long-term stable hybrid zones have been reported in temperate areas for several white-headed gull taxa (L. occidentalis×L. glaucescens, Bell 1996, 1997; Good et al. 2000; L. argentatus×L. marinus, Crochet et al. 2003) and may be maintained by hybrid superiority at the hybrid zone (Moore 1977). In contrast, white-headed gull species appear to hybridize pervasively throughout northern latitudes (L. argentatus×L. hyperboreus, Vigfúsdóttir et al. 2008; L. argentatus×L. glaucescens, Williamson and Peyton 1963; L. argenatus×L. schistisagus, Olsen and Larsson 2004; L. glaucoides×L. thayeri, Weir et al. 2000; L. glaucescens×L. schistisagus, Olsen and Larsson 2004) and discrete hybrid zones appear to be absent. Differences in the degree of hybridization may be attributable to the stability of the habitat where these areas of secondary contact occur. Stable secondary contact zones for gulls are observed at temperate latitudes, where presumably habitat has remained relatively stable throughout the last glacial maximum, allowing species to diverge in allopatry without coming into secondary contact during interglacial periods. Conversely, Arctic species reside in more stochastic environments, where suitable habitat repeatedly contracted and expanded during the Pleistocene glacial cycles (Hewitt 2004). These highly variable climatic conditions likely resulted in a cycle of isolation during glacial periods and secondary contact during interglacial periods, potentially limiting species divergence and development of pre- and postzygotic isolating mechanisms.
Comparatively higher estimates of population structure observed for mtDNA than nuclear DNA markers are consistent with Haldane's rule. Haldane's rule states that hybrids of the heterogametic sex will experience reduced fitness (i.e., greater inviability or sterility) relative to those of the homogametic sex (Coyne and Orr 2004). In birds, females are heterogametic; therefore, if hybrid females were experiencing a strong disadvantage relative to hybrid males, observed genetic differentiation would be greater in mtDNA than nuclear markers. In a study of mainly European white-headed gull populations, researchers proposed that the large discrepancy in interspecific comparisons (mtDNA estimates were 3.3–14.5 times greater than estimates using microsatellites) between marker types could be attributed to a strong disadvantage for hybrid females (Crochet et al. 2003). We did observe high levels of interpopulation comparisons among species for mtDNA (ΦST= 0.130–0.821; Appendix S3) with no significant comparisons at nuclear markers. However, Haldane's rule would presumably have the greatest influence at secondary contact zones. In these areas, 23% (5/22; Appendix S3) of the comparisons had significant mtDNA estimates. Of those five, only two comparisons (HypYKD × VegCh, HypIc × AeIc) had proportionally greater degree of divergence than can be explained by differences in the effective population size between genomes (Zink and Barrowclough 2008) and maximum possible FST value (Meirmans 2006), although differences were slight (HypYKD × VegCh, FSTExp. = 0.039, FSTObs. = not significant; HypIc × AeIc, FSTExp. = 0.051, FSTObs. = not significant). Therefore, Haldane's rule does not appear to apply to the white-headed gull species studied here.
The lower FST values observed at nuclear fragment assays (i.e., microsatellites) relative to mtDNA sequence data among species may be attributable to fragment length homoplasy, through not identifying unique alleles because fragments of the same length may have different sequences or may have mutated back to the ancestral state. Both types of homoplasy could pose problems when assessing population structure with fragment analysis based on detecting allelic frequency differences among populations. Most interpopulation comparisons among species have higher RST than FST estimates, indicating that the mutation process, and therefore homoplasy, is having an effect on estimators of population subdivision. Caution should be taken when interpreting pairwise population comparisons of allelic variance among species. Rousset (1996) showed that there are no simple effects of homoplasy on estimators of population differentiation (FST and RST) for loci evolving under the SMM and island model of migration, making it difficult to assess potential biases in estimates. However, we observed similar signals of population structure based on microsatellite and nuclear intron data. Therefore, the differences in the degree of population structure observed between the genomes studied here may be more attributable to introgression reducing the rate of lineage sorting in the nuclear genome and, to a lesser extent, fragment length homoplasy, although experimental evidence is needed to test this hypothesis.
Pleistocene refugia and comparisons with other taxa
Two distinct mtDNA haplotype groups were observed within L. argentatus, L. canus, and L. hyperboreus, the three sampled taxa with circumpolar distributions, consistent with other studies on white-headed gulls (Fig. 2G; Liebers et al. 2004; Sternkopf et al. 2010). A pattern of at least two allele groups was also observed for the nuclear intron loci (Fig. 2A–F). Concordance in haplotype and allele groups suggests that white-headed gulls were subdivided into at least two refugia that persisted for extended periods of time during the Pleistocene. Furthermore, substructuring observed within mtDNA corresponds to locality. The small central haplotype group is represented by L. argentatus individuals from Iceland and Tromsø and L. hyperboreus individuals from Iceland, Greenland, Svalbard, and a single individual from northern Alaska (Fig. 2G, population data not shown). All populations of L. argentatus are represented in the large haplotype group; however, only North American and Greenland L. hyperboreus individuals (and a single individual from Iceland) are observed within this group (Fig. 2G, population data not shown). These findings differ from those of Sternkopf et al. (2010), who identified genetic similarity between L. hyperboreus and L. a. smithsonianus (the North American subspecies); we did not observe a haplotype group restricted to North America in L. argentatus. The presence of a primarily Scandinavian/Greenland/Iceland haplotype group indicates the restriction of at least L. argentatus and L. hyperboreus into a high-latitude refugium in the North Atlantic/Arctic Ocean, possibly Spitsbergen Bank or northwest Norway. Furthermore, given the restricted geographical distribution of L. hyperboreus haplotypes within the central clade, the Scandinavian/Greenland/Iceland refugium was likely isolated from other L. hyperboreus populations and did not substantially contribute to the postglacial colonization of North America and Europe.
Despite the presence of species with distributions restricted to northern latitudes, suggestive of restriction to Pleistocene refugia, we were unable to identify glacial refugia for the white-headed gulls studied here based on the coalescent. White-headed gulls are characterized by strong dispersal ability and a propensity to hybridize in areas of secondary contact, as is reflected in contemporary accounts of long-range colonization and subsequent hybridization (Pierotti 1987; Olsen and Larsson 2004; Vigfúsdóttir et al. 2008). The tendency for hybridization at areas of secondary contact is very strong: 16 of the 18 species (89%) are reported to hybridize in nature (Pierotti 1987; Olsen and Larsson 2004 and citations therein) and appear to be free of postzygotic barriers to hybridization (Shields 1987; Snell 1991). Hybridization and subsequent introgression may have erased the genetic legacy of the Pleistocene for white-headed gulls; reductions in effective population sizes associated with the restriction to glacial refugia were not observed, even for populations currently located in glaciated areas. Alternatively, historic Arctic white-headed gull populations that were restricted south and north of the ice sheets likely followed habitat made available by the retreating glacial ice sheets to present day locations. Short movements from refugia would have allowed these historical populations to retain genetic diversity because effective population sizes would not be reduced (Hewitt 1996), especially if colonization occurred over a long period. However, it is unlikely that all populations studied here colonized slowly subsequent to glacial retreat, given the dispersal and colonization ability of Arctic white-headed gulls. Therefore, a more parsimonious explanation is that the strong tendency for hybridization in this group erased the genetic signature of Pleistocene refugia.
Holarctic species typically exhibit shallow but clear phylogeographic signal, due to fragmentation of the distributions of these species into high-latitude glacial refugia, resulting in genetic and morphological differentiation across ranges (Hewitt 2004). Although Arctic white-headed gulls have morphologically distinct forms across their distribution (i.e., subspecies), limited population genetic subdivision among northern Arctic white-headed gull populations was observed. The combination of limited sorting among species (Liebers et al. 2004; Pons et al. 2005) and morphological diversity of taxa, as is evident in gulls, is suggestive of recent allopatric fragmentation and restriction to multiple glacial refugia. Indeed, four populations of gulls (L. hyperboreus from Greenland, L. glaucescens from Middleton Island, L. argentatus from Iceland, and L. canus from south-central Alaska) have larger time-of-divergence estimates than other sampled populations and coincide with proposed high-latitude glacial refugia for other taxa (Ploeger 1968). Historical and contemporary hybridization among regions, coupled with small effective population sizes, could have overwhelmed the genetic uniqueness accumulated by northern populations during the last glacial maximum, producing a signal of low genetic structure.
Arctic white-headed gulls are less genetically differentiated at the northern limits of their distribution; this pattern was observed both among species that breed exclusively at high latitudes (L. schistisagus, L. glaucoides, L. thayeri, and L. hyperboreus) and within species (e.g., L. argentatus and L. glaucescens). Liebers and Helbig (2002) observed this pattern between L. fuscus and L. cachinnans; L. fuscus exhibited less genetic structure than its southern counterpart L. cachinnans. The authors proposed that differences in the degree of population subdivision indicated that northern gulls are phylogenetically younger than their southern counterparts and were more affected by glacial cycles. In contrast, their southern relatives would have been able to maintain larger long-term stable population sizes during glacial periods. Although Liebers and Helbig's (2002) hypothesis is consistent with observations within Europe, it may not apply to the North American species studied here. In contrast to Europe, high-latitude glacial refugia, notably Beringia, were present in North America during the last glacial maximum and likely promoted genetic diversification in Arctic taxa (Hewitt 2004). As glacial ice sheets retreated, the ranges of temperate species likely expanded northward and came into contact with northern “refugial” populations. Hybridization between “newly arriving” temperate species and northern “refugial” gull species provides an alternate hypothesis for the genetic pattern of decreasing genetic differentiation with increasing latitude.
Arctic white-headed gulls appear to be unique among the Arctic fauna in the spatial distribution of their alleles across species. Previous studies of genetic substructuring among closely related Arctic species reported species-specific clades often correlated with geography (e.g., Tetraoninae, Drovetski 2003; Lemmus spp., Fedorov et al. 2003; Motacilla spp., Pavlova et al. 2003; Lepus spp., Waltari and Cook 2005; Ovis spp., Loehr et al. 2006; Riparia spp., Pavlova et al. 2008). Limited lineage sorting among these seven gull species is noteworthy given that genetic subdivision is regularly observed within individual Arctic breeding species (e.g., Calidris alpina, Wenink et al. 1996; Troglodytes troglodytes, Drovetski 2003; Rangifer tarandus, Flagstad and Røed 2003; Branta canadensis, Scribner et al. 2003; Myodes rutilus, Cook et al. 2004; Microtus oeconomus, Galbreath and Cook 2004; Myopus schisticolor, Fedorov et al. 2008; Pinicola enucleator, Drovetski et al. 2010; Somateria mollissima, Sonsthagen et al. 2011). Aspects of white-headed gull behavioral biology, such as colonization ability and propensity to hybridize, as well as their recent evolutionary history, have likely played a large role in the limited genetic structure observed.