Discussion
Heterogeneous environmental conditions across time and space can drive adaptive population divergence among even populations that are only partially reproductively isolated (Hereford, 2009). Here we assessed population structure and functional divergence among six geographically isolated breeding populations of Arctic-breeding snow buntings. Both our neutral (microsatellite DNA loci) and functional (coding-gene SNPs) genetic marker data show substantial population divergence among all populations, likely indicative of some level of reproductive isolation. Furthermore, we demonstrated that the observed population differentiation patterns at selected known-function SNPs likely resulted primarily from stabilizing, but also divergent, selection at the candidate loci. The global divergence analyses showed strong evidence of stabilizing selection across all populations, which is not surprising given the expected canalization of the vital functional gene loci chosen for this study. At the pairwise population comparison level, our functional marker results show signatures of both neutral drift and selection, with functional divergent selection observed at some SNP loci. Such selection effects likely reflect local adaptation of different snow bunting populations to their local environments (e.g., both wintering and migratory route selection pressures).
Although both of our marker types yielded broad spatial divergence patterns separating resident and migratory populations, finer genetic structure differed based on the marker type. A prime example is the lack of divergence between the Alert & Mitivik Island and the Utqiagvik & Svalbard populations, indicative of gene flow between both population pairs. The Alert & Mitivik Island population pair exhibited similar microsatellite and SNP divergence patterns; however, the Utqiagvik & Svalbard population pair curiously exhibited significant SNP divergence, but no divergence with the microsatellite markers. Such a pattern is consistent with a strong selection signature at the SNP loci, despite gene flow. Our observation of gene flow between the Alert and Mitivik Island populations is a new finding, but supports previous work in this species using stable hydrogen isotope analysis and light-level geolocator tracking that suggested two parallel migratory systems for the high and low Eastern Canadian Arctic with Hudson Bay as a migratory divide (Macdonald et al., 2012). Thus, it is possible that the Alert population follows the same migratory route as the Mitivik Island population (i.e., to the West of Hudson Bay, NU, Canada; Macdonald et al., 2012). Additionally, because the Mitivik Island population has been shown to winter in the Canadian provinces of Saskatchewan and Alberta (Macdonald et al., 2012), it is further possible the individuals in these populations winter together, or even mix during spring migration to the breeding grounds. On the other hand, the presence of potential gene flow between the Utqiagvik and Svalbard populations is surprising, given the geographic distance between the two sample sites. Although we do not currently know where birds from the Utqiagvik breeding population overwinter, recent tracking of the Svalbard buntings using light-level geolocators (GLS) indicate they overwinter in the Asian Western Siberian Steppe where they utilize the high abundance of grain croplands and face very little interspecific competition (Snell et al., 2018). This could also be true for individuals breeding at Utqiagvik, providing a potential mechanism for gene flow between the two populations. If true, Svalbard birds would be migrating west in the Fall, and Utqiagvik birds east in the fall, to share wintering grounds in the Asian Western Siberian Steppes. Such cross-hemisphere migration has been demonstrated in a similarly-sized songbird, the northern wheatear (Oenanthe oenanthe ), using GLS (Bairlein et al., 2012). Nevertheless, a detailed migration study is needed for Utqiagvik snow buntings to empirically test the possibility of a shared wintering ground.
While fairly spatially distant snow bunting populations showed genetic connectivity, we surprisingly found significant differentiation between the two non-migratory populations in Alaska (Aleutian and Pribilof islands), based on microsatellite data. These populations exhibited substantial divergence despite being geographically close (Figure 1). Migratory life history is a critical component of genetic population structure; high dispersal rates result in genetically homogeneous populations, whereas restricted dispersal allows for development of genetically differentiated populations (Milgroom, 2015) due to elevated isolation and drift (Arguedas & Parker, 2000; Winker et al., 2000). Our results support this pattern in snow buntings, with both microsatellite and SNP data clustering resident and migratory populations separately. Overall, in addition to identifying significant global population differentiation, the genetic markers used in this study add to our knowledge of migratory connectivity patterns among breeding snow bunting populations. More importantly, our results shed light on the vulnerability of common wintering grounds for some populations should these sites face human-induced stressors such as habitat degradation.
Although local adaptation is predicted in populations experiencing divergent local selection pressures, it is rarely directly demonstrated empirically since it requires common-garden or reciprocal transplant experiments (Kawecki & Ebert, 2004) which are not practical for many wild populations (Blanquart et al., 2013). Avian species have been shown to exhibit population-level patterns of variation in timing of migration and brood initiation (Gu et al., 2021; Wanamaker et al., 2020), body size and mass (Blondel et al., 2006), song (Badyaev et al., 2008), personality (Mouchet et al., 2021), and plumage (Antoniazza et al., 2010) that have been identified as possible locally adapted traits. Although those studies provide strong indirect evidence of local adaptation, they may reflect phenotypic plasticity or even flexibility. In this study, we used SNP gene loci expected to be under selection based on their putative gene function, and as such likely to reflect environmental and ecological differences driving genetic variation among populations (Wellband et al., 2018). Although more than a quarter (28/101) of our SNP markers are missense variants, all were in very strong linkage disequilibrium with the target known-function genes, making our study unique from other SNP-based studies in birds that used random SNPs located in both coding and non-coding regions of the genome (e.g., Tiffin & Ross-Ibarra, 2014; Pardo-Diez et al., 2015). Oura priori choice of candidate gene function improves the likelihood of detecting functional patterns of population differentiation consistent with local adaptation in breeding snow bunting populations. Local adaptation has implications for management and conservation aimed at preserving local genetic diversity, especially as Arctic-migratory species continue to face strong effects of climate change and other anthropogenic stressors worldwide.
Generally, locally adapted populations are predicted to exhibit significantly higher (for divergent selection) or lower (for stabilizing selection) genetic differentiation than expected under neutral processes (Schlötterer, 2002; Hoban et al., 2016). Consistent with this idea, a high proportion of our candidate loci exhibited evidence of being under selection at candidate functional loci. Only a handful of previous studies have assessed patterns of divergence at both coding (i.e., functional) and non-coding (i.e., presumed neutral) marker loci, to interpret selection patterns in migratory bird species. Furthermore, the majority of those studies used randomly selected genome-wide SNPs and they inferred divergent selection at functional loci based on presumed linkage disequilibrium. For example, Zhan et al. (2015) used a targeted approach comparing thirteen wild populations of saker falcon (Falco cherrug ) across Eurasia using SNP data and inferred that the MHC genes were under directional selection (FST > 0.5), with the remaining candidate SNPs showing signatures of stabilizing selection or drift. Although SNP-based selection studies are becoming more common in migratory bird species (e.g., Ruegg et al., 2014; Bay et al., 2021; Larison et al., 2021; Ruegg et al., 2021), there have only been two such studies on Arctic-breeding migratory birds, both of which employed a random SNP approach and reported no or low levels of selection. For example, Colston-Nepali et al. (2020) used RAD-seq to genotype six breeding colonies of northern fulmar (Fulmarus glacialis ) at 6,614 genome-wide SNPs; however, no outlier loci were identified. Similarly, Tigano et al. (2017) used 2220 genome-wide SNPs across five colonies of Arctic-breeding thick-billed murres and found ~6% outlier SNPs (only 28% of which showed divergent selection). However, random SNP surveys do not have a priori SNP gene function and often it is difficult to assign function. For example, Tigano et al. (2017) found only 6 of their 111 identified outlier SNP loci could be assigned a putative function (GO term). In contrast, we detected strong signatures of stabilizing selection at known-function SNP loci, with some showing evidence of divergent selection in pairwise population comparisons. The high level of stabilizing selection likely results from canalization of the genes associated with our SNP loci as they were selected to reflect critical organismal and cellular functions.
Our candidate SNP loci that exhibited consistent patterns of divergent selection may reflect local adaptation. For example, ACVR2A (divergent between Utqiagvik and Mitivik, and between Utqiagvik and Svalbard population pairs) codes for a receptor that is involved in the induction of adipogenesis and growth (Donaldson et al., 1992). It has been shown that fat reserves aid in thermogenesis, cold endurance (Vézina et al., 2012; Montgomerie & Lyon, 2020) and modulate the adrenocortical response to environmental stress (Wingfield et al., 2004). Those functions not only facilitate successful breeding in Arctic conditions, but also help snow buntings survive challenging conditions (i.e., scarce food resources and cold temperatures) on arrival as they prepare for breeding (Le Pogam et al., 2021). PTPRZ1 (divergent between Pribilof Islands and Svalbard, and between Utqiagvik and Svalbard population pairs) is mainly involved in development of myelinating oligodendrocytes and is thought to play a role in the establishment of contextual memory and learning (The UniProt Consortium, 2015). The role of spatial memory and learning has been explored in passerines for behaviours associated with food hoarding (Hitchcock & Sherry, 1990; Brodin, 1994; Healy & Krebs, 1996; Smulders & DeVoogd, 2000) and vocal communication (Nottebohm, 1999; Zeigler & Marler, 2004). Our candidate gene loci implicated in divergent selection warrant further examination of allelic variation at the loci under selection in snow buntings and possibly other migratory avian species.
Arctic-breeding migratory bird species utilize diverse breeding and over-wintering habitats, resulting in a substantial variation in experienced local abiotic factors such as temperature, wind, precipitation, and snow cover (among others; Martin & Wiebe, 2004; Wingfield et al., 2004). Variation likely drives selection pressures on Arctic-breeding birds that have short breeding times and high energetic demands (Le Pogam et al., 2021), which may contribute to local adaptation (Macdonald et al., 2012; Tigano et al., 2017; Snell et al., 2018). To our knowledge, this is the first study to investigate global population structure and genetic divergence consistent with local adaptation in a circum-polar arctic-breeding songbird. Consistent with our predictions, we observed strong evidence of genetic isolation coupled with some SNP loci showing divergent selection signatures. However, we observed stabilizing selection signatures across most SNP loci, and while high levels of stabilizing selection are reasonable given the nature of our candidate genes, we did not expect such a dominant role for apparent stabilizing selection among our SNP loci. Identifying population genetic structure at a pan-Arctic scale in snow buntings is especially important for their conservation as they face severe effects of climate change in their breeding areas coupled with other anthropogenic stressors in their overwintering areas (Walker et al., 2015). Our analyses of population divergence using both neutral and functional markers provide conservation-related data particularly valuable for species such as the snow bunting, which, because of its migratory life history, experiences diverse management jurisdictions, habitat degradation, and survival challenges.