Genome-wide tests for introgression between cactophilic Drosophila implicate a role of inversions during speciation

Abstract

Models of speciation-with-gene-flow have shown that the reduction in recombination between alternative chromosome arrangements can facilitate the fixation of locally adaptive genes in the face of gene flow and contribute to speciation. However, it has proven frustratingly difficult to show empirically that inversions have reduced gene flow and arose during or shortly after the onset of species divergence rather than represent ancestral polymorphisms. Here, we present an analysis of whole genome data from a pair of cactophilic fruit flies, Drosophila mojavensis and D. arizonae, which are reproductively isolated in the wild and differ by several large inversions on three chromosomes. We found an increase in divergence at rearranged compared to colinear chromosomes. Using the density of divergent sites in short sequence blocks we fit a series of explicit models of species divergence in which gene flow is restricted to an initial period after divergence and may differ between colinear and rearranged parts of the genome. These analyses show that D. mojavensis and D. arizonae have experienced postdivergence gene flow that ceased around 270 KY ago and was significantly reduced in chromosomes with fixed inversions. Moreover, we show that these inversions most likely originated around the time of species divergence which is compatible with theoretical models that posit a role of inversions in speciation with gene flow.

Keywords: Drosophila arizonae; Drosophila mojavensis; Speciation with gene flow; divergence genomics; inversions.

Figure 1

Alternative scenarios of species divergence: (1) strict divergence without gene flow, (2) isolation with migration (IM), and (3) isolation with initial migration (IIM). (4) An inversion that predates the species split should be associated with an older population divergence time τ₀.

Figure 2

Mean per site divergence in 500 kb sliding windows. Divergence between D. arizonae and D. mojavensis is nearly identical for allopatric (black) and sympatric (red) comparisons. Divergence between the two D. mojavensis lines is shown in blue. Known inversion breakpoints on chromosome 2 (Guillen and Ruiz 2012) and the X (Runcie and Noor 2009) are indicated by solid, vertical lines, the position of the unmapped breakpoints on chromosome 3 by dashed, gray lines. All scaffolds are oriented with the centromere to the left (origin).

Figure 3

The distribution of divergent sites (k) between D. arizonae and sympatric D. mojavensis in 250 bp (left) and 500 bp (right) intergenic blocks. Colinear chromosomes 4 and 5 are shown in black, the inverted chromosomes 2 and 3 in blue and green, respectively. Points are joined for clarity. The expected distributions under the best supported model inferred from the data (Table 3) are shown as dashed lines.

Figure 4

(A) Marginal support ($\Delta lnL$ relative to the maximum likelihood solution) for the rates of gene flow (M) between D. arizonae and D. mojavensis (sympatric comparison) estimated in colinear (black) and rearranged (chromosome 2, blue; chromosome 3, green) autosomes under the IIM model. (B) Marginal support for the onset of species divergence (τ₀) and the cessation of gene flow (τ₁) (black) and the divergence time between D. mojavensis populations in Baja California and Sonora (red). The age of the inversion 2q falls within the estimated onset of divergence between D. arizonae and D. mojavensis (turquoise). The horizontal line defines 95% confidence intervals of parameter estimates.

Figure 5

Schematic of the speciation history of D. arizonae and D. mojavensis. The onset of divergence around 1.3 MY was followed by a prolonged period of gene flow that ceased before the divergence of the different populations of D. mojavensis. Inversion 2q arose in D. mojavensis during the onset of divergence (blue star) and is the first in a cascade of three overlapping inversions on chromosome 2 that became fixed in D. mojavensis (adapted from Guillen and Ruiz (2012)).