Discussion
Migration and environmental change can drive adaptation (Rankin and Burchsted 1992; Charlesworth et al. 2003; Gillespie 2004; Excoffier et al. 2009; Porretta et al. 2012; White et al.2013). Species with somewhat isolated or divided populations are likely to adapt to their differing local environments. Migration can both facilitate and hinder such adaptation, allowing new variation (including potentially beneficial variants) to be spread between populations and preventing inbreeding depression. Strong migration can also import locally nonadaptive variants and prevent the fixation of the most fit variants in local populations. We sought to examine the extent that these processes take place in a species of Drosophila found across four forests separated by large expanses of desert.
We characterized the phylogeographic history of four populations ofDrosophila innubila , a mycophagous species endemic to the Arizonan Sky islands using whole genome resequencing of wild-caught individuals. D. innubila expanded into its current range during or following the previous glacial maximum (Figure 1, Supplementary Figure 1). We find some evidence of local adaptation, primarily in the cuticle development genes and antifungal immune genes (Figure 2, Supplementary Figure 2). Interestingly, there is very little support for population structure across the nuclear genome (Figures 1 & 2, Supplementary Figures 1 & 2), including in the repetitive content (Supplementary Figure 11), but some evidence of population structure in the mitochondria, as found previously in D. innubila(Dyer and Jaenike 2005). This suggests that if gene flow is occurring, it could be primarily males migrating, as is seen in other non-Drosophila species (Rankin and Burchsted 1992; Searle et al. 2009; Ma et al. 2013; Avgar and Fryxell 2014). Based on the polymorphism data available, coalescent times are not deep, and given our estimated population history, this suggests that variants aren’t ancestrally maintained and are instead transmitted through migration between locations (Charlesworth et al. 2003).
Segregating inversions are often associated with population structure and could explain the abnormalities seen on Muller element B here (Supplementary Figures 2-5). Our detection of several putative segregating inversions on Muller element B relative to all other chromosomes (Figure 3A) supports this assertion. However, few of the putative inversions support this hypothesis, in that all are large and common inversions characterized in all populations, suggesting the inversions are not driving the elevated FST. We suspect that the actual causal inversions may not have been characterized due to the limitations of detecting inversions in repetitive regions with short read data (Marzo et al. 2008; Chakrabortyet al. 2017). The elevated FST could also be caused by other factors, such as extensive duplication and divergence on Muller element B being misanalysed as just divergence. In fact, the broken and split read pairs used to detect inversions are very similar to the signal used to detect duplications (Ye et al.2009; Rausch et al. 2012; Chen et al.2016), suggesting some misidentification may have occurred. If a large proportion of Muller B was duplicated, we would see elevated mean coverage of Muller element B in all strains compared to other autosomes, which is not the case (Supplementary Table 1). Further study is necessary to disentangle if inversions or other factors are causing this elevated FST and the selective and/or demographic pressures driving this differentiation. However, it is worth noting thatD. pseudoobscura segregates for inversions on Muller element C and these segregate by population in the same Sky island populations (and beyond) as the populations described here (Dobzhansky and Sturtevant 1937; Dobzhansky et al. 1963; Fuller et al. 2016). Thus, the inversion polymorphism among populations is a plausible area for local adaptation and may provide an interesting contrast to the well-studied D. pseudoobscura inversions.
We find very few signatures of divergence between samples from 2001 and 2017 (Supplementary Figure 8). Though the environment has changed in the past few decades, there may have been little impact on the habitat ofD. innubila in the Chiricahuas, resulting in few changes in selection pressures in this short period of time, unlike most bird and mammal’s species in the same area (Coe et al. 2012). Interestingly, there was an extensive forest fire in 2011 which could plausibly have been a strong selective force but we see no genome-wide signature of such (Arechederra-Romero 2012). Alternatively, seasonal fluctuations in allele frequencies may swamp out directional selection. Excessive allele frequency change is limited to a few genes with no known association to each other, and little overlap with the diverging genes between populations. Some of the genes with elevated FST (and differing in allele frequency) between time points overlap with divergent genes between sexes, primarily at the telomere of the X chromosome (Muller element A, Supplementary Figure 8). In fact, FST is significantly correlated on Muller element A between the two surveys (Pearson’s correlation t = 82.411, p-value = 1.2e-16), even with the 2001-2017 survey only considering male samples, supporting an association between the factors driving divergence between sexes and over time. Given the sex bias of SNPs in this region, this could suggest that a selfish factor with differential effects in the sexes is located on the X chromosome near the telomere (Burt and Trivers 2006). Often these selfish elements also accumulate inversions to prevent the breakdown of synergistic genetic components (Burt and Trivers 2006), and the Muller A telomere appears to have accumulated several inversions (Figure 3A). However, populations of D. innubila are already female-biased due to the male-killing Wolbachia infection found in 30-35% of females (Dyer 2004; Dyer and Jaenike 2005; Jaenike and Dyer 2008). Thus D. innubila could be simultaneously parasitized by both the male-killing Wolbachia and a selfish X chromosome. Alternatively, the strong signals associated with the telomere of the X could be a signature of selection related to the Wolbachiainfection (Unckless 2011b).
Ours is one of few studies that sequences individual wild-caughtDrosophila and therefore avoids several generations of inbreeding that would purge recessive deleterious alleles (Gillespie 2004; Mackay et al. 2012; Pool et al. 2012). The excess of putatively deleterious alleles harkens back to early studies of segregating lethal mutations in populations as well as recent work on humans (Dobzhansky et al. 1963; Marinkovic 1967; Dobzhansky and Spassky 1968; Watanabe et al. 1974; Gao et al. 2015).
To date, most of the genomic work concerning the phylogeography and dispersal of different Drosophila species has been limited to themelanogaster supergroup (Pool et al. 2012; Pool and Langley 2013; Behrman et al. 2015; Lack et al. 2015; Machado et al. 2015), with some work in other Sophophora species (Fulleret al. 2016). This limits our understanding of how non-commensal species disperse and behave, and what factors seem to drive population demography over time. Here we have glimpsed into the dispersal and history of a species of mycophageous Drosophila and found evidence of changes in population distributions potentially due to the changing climate (Survey 2005) and population structure possibly driven by segregating inversions and selfish elements. Because many species have recently undergone range changes or expansions (Excoffier et al. 2009; Porretta et al.2012; White et al. 2013), we believe examining how this has affected genomic variation is important for population modelling and even for future conservation efforts (Excoffier et al.2009; Coe et al. 2012).