DISCUSSION
Our whole-genome data reveal that invasive starling populations show moderate genetic diversity across North America, rapid population growth after a genetic bottleneck, negligible geographically structured genetic differentiation, significant but quite weak isolation-by-distance, and intriguing associations of genetic variation with environmental parameters that might result from adaptive processes. Genetic diversity is typically predicted to be low in invasive populations (Dlugosch & Parker, 2008), and although RADseq tends to estimate lower genetic diversity than the true diversity, this bias is minimal for taxa known to be genetically depauperate (Cabe, 1998; Cariou et al., 2016). Within invasion genetics, one area of recent focus involves the spatial dynamics of invasion, and we can explore how genetic variation varies across the post-invasion range of starlings in North America. Although there is a significant (but low magnitude) signal of isolation-by-distance, hierarchical AMOVAs find that variation within and among individuals explains observed differentiation better than variation among populations. There is no evidence of population structure, and while models indicate some subtle spatial patterns of genetic variation, these model-based inferences likely reflect sampling artifacts (Supplementary Information). Finally, after the initial founder effect, the effective population size has grown by a ratio of 1:100. These patterns are consistent with the expectation that extensive gene flow—as shown by extremely low FST among populations (Table 2)—maintains high connectivity across North American starling populations. When interpreted within the context of a complementary exploration of movements inferred from feather isotope assays, this genomic survey confirms that dispersal and migration continue to influence the genetic variation generated in just over a century of range expansion.
Long-distance dispersal may be common in introduced starlings on multiple continents, where rates of dispersal are determined by demographic changes and environmental quality. In South Africa, invasive starlings disperse when their natal environment becomes crowded or unsuitable, as at the leading edge of their range expansion (Huiet al., 2012). In North America, the effective population size (Ne) of the present-day invasive population has expanded dramatically, with models indicating that current Ne is even larger than the Ne of the founding population. This explosive population growth from hundreds to millions of individuals may have encouraged long-distance dispersal away from the dense populations in eastern North America. In general, multiple lines of evidence have shown that the dispersal rates and distances of juvenile starlings in North America are remarkably high (Dolbeer, 1982; Cabe, 1999; Werner, Fischer, & Hobson, 2020). Considered in concert with similar findings in the invasive South African and Australian starling populations (Huiet al., 2012; Phair et al., 2017), it is clear that dispersal similarly plays an ongoing and major role in the connectivity of North American starling populations.
Although it is generally important, dispersal has not been uniform across time or space in North American starlings. One open question in invasion genetics is whether expanding populations maximize spatiotemporal fitness via spatial sorting and incipient adaptation (Williams et al. 2019). Empirical studies of some invasions suggest that spatial sorting is common, including in common mynas (Berthouly-Salazaret al. , 2012), pumpkinseed fish (Ashenden et al., 2017), and cane toads (Brown et al. , 2014). More empirical work is needed to verify the conditions in which spatial sorting can lead to lasting shifts in fitness (Lee, 2011; Phillips & Perkins, 2019; Williams et al. , 2019). However, we do know that adaptive dispersal strategies can facilitate range expansion in Western bluebirds (Duckworth, 2008) and invasive beetles
(Lombaert et al., 2014; Ochocki & Miller, 2017) among other species. For North American starlings, we can combine historical records, isotopic and genetic evidence to make inferences about the possibility of spatial sorting.
Feather isotope data from the same starlings sampled in this genetic survey suggests that movements are not uniform continent-wide, and that both juvenile dispersal and regional differences in migration could influence genetic variation (Werner, Fischer, & Hobson, 2020). If starling movement was the primary driver of differentiation, then we would expect FST among populations to be directly related to the number of birds collected and assigned to that population. Starlings in the western U.S. appear to have differentiated subtly from their eastern counterparts based on the higher FST between Arizona—and to some degree, New Mexico and Colorado—and all other sampling locations (Table 2). Birds collected in these southwestern states are also assigned to those states by discriminant function analysis (Werner, Fischer, & Hobson, 2020); for example, 67% of starlings collected in New Mexico were also assigned to that state (Table 2). We suggest that birds assigned to the same state where they were collected may reside in that state year-round, but we note that collecting feathers once in the lifespan of the bird does not allow us to determine the bird’s lifelong migration and dispersal. The highest rates of pairwise differentiation occur between states where one of those “populations” shows a relatively high collection-state assignment, which suggests that starling movement among sampling locations does impact the genetic variation. This comparison is further supported by the low but significant isolation-by-distance across the North American range; however, environmental pressures explain genetic variation better than a simple model that accounts for geographic distance. In addition, assignment to a state other than the collection state seems to have only a minor effect on FST, even though we might expect that movement among geographically distant sampling locations might lead to gene flow among those locations. In Arizona—where FST is highest but still relatively weak—Werner et al.’s model suggests that 25% of birds originated in New Hampshire, which is approximately the percentage of birds assigned to Arizona (28%, Table 2). Unsurprisingly, geography, environment, and demography all appear to influence genetic variation within North American starling, and explicit tests for selection can help to clarify the relative weight of each factor.
This species’ general genome-wide panmixia across the continent allows for tests of selection on loci that may be involved in local adaptation. Because genetic diversity and differentiation are exceptionally low, it is relatively easy to identify sites that have differentiated against this background of low genome-wide or population-specific divergence (Dlugosch et al., 2015). We find that almost 200 of the 15,038 RAD loci appear to be under selection using a redundancy analysis. Only 13 of these SNPs overlap with the SNPs identified by a latent-factor mixed model, and there is no overlap between the RDA candidates and the differentiation-based scans (BayeScEnv and Bayescan). It is unsurprising that each test identifies different candidates for selection, because the assumptions underlying each are very different (for more details on our approach to selection testing, see Supplementary Information). Rather than making inferences based on the genes identified by these scans, we instead propose that genotype-environment associations show that changes in precipitation and temperature can explain genetic variation in North American starlings. However, we do suggest that rapid evolution in genes underlying key physiological processes may have supported the starling’s spread across North America (Supplemental Information). Specifically, we hypothesize that aridity and cold temperatures that are not experienced in the starling’s native range exert enough selective pressure on North American starlings to result in incipient local adaptation. While this finding suggests that local adaptation may explain genetic variation within the North American starling invasion, there are several relevant caveats to this approach.
Because there is evidence of a genetic bottleneck in the North American starling invasion, it is possible that artifacts resulting from genetic drift could mimic these patterns of local adaptation. We cannot rule out allele surfing during range expansion as an explanation based on the methods currently available for testing genotype-environment associations in a young and expanding system like this one. However, given how short the timescale of divergence is in North American starlings, such drift effects are more likely if genetic variation is (1) ancestral and (2) structured among populations. Gene flow among starling populations is remarkably high, but we do not know if variants putatively under selection were present in the ancestral population. These alleles under putative selection do not approach fixation, as no putative outlier has an allele frequency greater than 0.28. We note that concordance among environmental and genetic distances (e.g., partial Mantel tests) indicate that spatial autocorrelation complicates our selection inferences. Under these conditions, any evidence for selection is likely to be weak, and as always these selection scans can generate false positives. However, RDA has the highest rate of true positives and lowest of false positives, and although this method has not been tested in such recent expansions, RDA is well-suited to systems where FST is very low (Meirmans, 2015, Forester et al. , 2018, Supplementary Information). Finally, we do not explicitly control for linkage, and some allele frequency shifts could be explained by recombination or by genetic hitchhiking.
Our study focuses on birds collected during the winter, which may limit our inferences about population structure and selection to these wintering populations. As discussed above, isotopic evidence suggests that starlings in the western U.S. tend to move only regionally whereas birds sampled in the eastern U.S. undertake longer movements (Table 2). This in turn suggests that starlings overwintering in the western U.S. are more likely to breed nearby, and thus the environmental conditions may not change as dramatically among wintering and breeding ranges. In addition, the environmental conditions that we expect to drive selection—precipitation and temperature—vary most substantially in the southwestern region: for example, the sampling location in Arizona is consistently warmer (BIO1) and drier (BIO12 and BIO16) than other locations (Figure 2, Supplementary Information). Western populations experience these environmental conditions year-round, which could allow selection to drive advantageous alleles toward fixation. Elsewhere in the U.S., starlings move more freely among states: individuals within each sampling location may come from different breeding populations, and additional sampling could reveal stronger population structure among true breeding populations. However, if our sampling overlooked some true populations, we would expect some signal of population structure. Individual-based tests of population structure—e.g., those that do not define possible populations a priori —do not recover any signals of population structure. This sampling strategy uses the more vagile eastern populations as a comparison to more-resident western populations that may also be under stronger selection. This framing may suggest that population structure in the western U.S.—of which we find no evidence—could explain allele frequency shifts that we infer to be selection, but when we compare the relative importance of geographic and environmental distances in partial Mantel tests, we find that environmental conditions better explain genetic variation. This evidence supports our interpretation of selection as a major driver of genetic variation in North American starlings.
A similar project on starlings in the Australian invasion—which colonized that continent nearly concurrently with the North American invasion—found that geographic but not environmental distance explains genetic patterns there (Cardilini et al., 2020). Starlings in the Australian range show substantial population structuring and significant patterns of isolation-by-distance. Earlier work had shown that gene flow among Australian starling populations is low (Rollins et al.,2009), and phylogeographic patterns of mitochondrial sequence variation confirm that starlings on the edge of the expansion front in Western Australia have differentiated from those still living in the introduction site (Rollins et al., 2011). In fact, starlings at the expansion front may have rapidly adapted during the Australian invasion (Rollins et al., 2016): the proportion of adult starlings in Western Australia carrying a novel mitochondrial haplotype has increased rapidly only at this range edge. A genotyping-by-sequencing survey employing a much greater number of SNP markers indicates three population subdivisions in Australia, where geographic distance explains genetic differentiation in starlings better than does environmental variation (Cardilini et al. , 2020). Global FST across all Australian populations is an order of magnitude higher than the equivalent FST index across North America, despite similar areas sampled. In Australia, the strong evidence for isolation-by-distance and founder effects complicate attempts to disentangle selection from drift, yet despite their differences in invasion dynamics, genotype-environment associations reveal signatures of selection in both invasions. On both continents, starling genetic variation can be explained by extremes in temperature and precipitation, and preliminary results of whole-genome resequencing of native and introduced populations confirm that variability in temperature and precipitation may shape observed genetic variation in starlings world-wide (Hofmeister et al., in prep ).
Our results contribute to the growing evidence of rapid adaptation in some expanding populations, even in extremely young systems. Some studies of rapidly expanding invasions find little evidence that adaptation may facilitate this expansion, as in corals (Leydet et al. 2018). However, other work suggests a role for selection in supporting rapid range expansion, such as in experimental studies of flour beetles (Szucs et al., 2017) and empirical work in guppies (Baltazar-Soares et al., 2019). Invasion biologists have long highlighted propagule pressure as a driver of invasion success, but the genetic composition may be just as important as the size of the establishing population (Briski et al., 2018). For example, genetic bottlenecks in monk parakeets, another avian invader now distributed world-wide, do not seem to inhibit invasion success (Edelaaret al., 2015). Pre-adaptation in the native range or selection during transport may facilitate the spread of invasive species, and human commensalism may support establishment and spread, as shown in house sparrows (Ravinet et al., 2018) and common mynas (Cohenet al., 2019), and reviewed across alien bird species (Cardador & Blackburn, 2019). Empirical studies of invaders like the ones described here also show how, in addition to genetic variation, epigenetic shifts and/or plastic changes in gene expression may support the establishment and expansion of invasive species (Marin et al., 2019). In the well-studied house sparrow—a system quite similar to starlings—epigenetic shifts may have supported invasions in Africa (Liebl et al., 2013) but not necessarily in Australia (Sheldonet al., 2018). Taken together, recent work suggests that we should consider a much wider range of demographic and ecological processes that lead to adaptive evolution in invading populations.
Invasive populations allow us to explore the genetic consequences of colonization and establishment in novel environments. On a background of low genetic differentiation and diversity, we find evidence of incipient genotype-environment associations in North American starlings. Here we explore how genetic variation changes across the landscape, but we cannot fully understand gene flow without studies of dispersal and migration of the individuals that carry genes. Our results complement other recent studies that reveal associations between climate variables and particular loci in North American vertebrates (Schweizer et al., 2015; Bay et al., 2018). Finally, we suggest that our study adds to those suggesting that rapid local adaptation can evolve even in dispersive and young populations.