Recent hybrids recapitulate ancient hybrid outcomes

Abstract

Genomic outcomes of hybridization depend on selection and recombination in hybrids. Whether these processes have similar effects on hybrid genome composition in contemporary hybrid zones versus ancient hybrid lineages is unknown. Here we show that patterns of introgression in a contemporary hybrid zone in Lycaeides butterflies predict patterns of ancestry in geographically adjacent, older hybrid populations. We find a particularly striking lack of ancestry from one of the hybridizing taxa, Lycaeides melissa, on the Z chromosome in both the old and contemporary hybrids. The same pattern of reduced L. melissa ancestry on the Z chromosome is seen in two other ancient hybrid lineages. More generally, we find that patterns of ancestry in old or ancient hybrids are remarkably predictable from contemporary hybrids, which suggests selection and recombination affect hybrid genomes in a similar way across disparate time scales and during distinct stages of speciation and species breakdown.

Model-based genotype and ancestry estimation for potential hybrids with mixed-ploidy

“…In the context of recent hybridization or admixture among divergent lineages, estimates of admixture proportions are a fundamental component of analyses of evolutionary processes or of learning the effects of population stratification in genome-wide association studies (Gompert & Harrison & Larson, 2016;Gompert et al, 2017). Hybrids commonly occur between taxa that have sex chromosomes (and mixed-ploidy within genomes of the heterogametic sex) and sex chromosomes may contribute disproportionately to their reproductive isolation (Payseur et al, 2004;Saether et al, 2007;Presgraves, 2008;Macholán et al, 2011;Chaturvedi et al, 2020). Additionally, many species complexes involve interactions and potential hybridization between individuals of different ploidy, including autopolyploids (e.g., Otto & Whitton, 2000;Kolář et al, 2017;Van de Peer et al, 2017).…”

“…Use of the admixture complement model will typically be restricted to low levels of K, because interpretation becomes increasingly complex for K > 2, requiring multi-dimensional plots for combinations of higher K values in the Q matrix (see Figure 2). In empirical study of systems for which K = 2 was well supported, the ancestry complement model has been used to learn about patterns of hybrid matings among Lycaeides butterflies (Gompert et al, 2014b;Chaturvedi et al, 2020), and Catostomus fish (Mandeville et al, 2017(Mandeville et al, , 2019, and in a related model used to study assortative mating among Populus species and their hybrids (Lindtke et al, 2014). The ancestry complement model for diploids is not found in structure, or other structure-like models.…”

“…Our measures of model performance build on previous testing of the model and software for diploids (Gompert et al, 2014b) and emphasize tests of the model extensions to data from polyploid and mixed-ploidy samples. As noted above, the diploid portion of the model has been used previously for several empirical analyses (e.g., Gompert et al, 2014b;Mandeville et al, 2015;Chaturvedi et al, 2020). We do not explicitly test the performance of the ancestry complement model as it has been done previously in Gompert et al (2014b) and has been used subsequently in several studies (e.g., Mandeville et al, 2017;Chaturvedi et al, 2020), to better distinguish among different classes of early generation hybrids and to distinguish recent from more advanced generation hybrids in diploid individuals ( Figure 2).…”

“…These include parametric methods for detecting loci with ancestry that is concordant with the remainder of the genome (e.g., Szymura & Barton, 1986;Gompert & Buerkle, 2011a), or for detecting breakpoints and tracts of ancestry among chromosomal blocks or haplotypes (e.g., Wegmann et al, 2011;Lawson et al, 2012;Sohn et al, 2012;Gompert, 2016). Similarly, researchers have contrasted ancestry and introgression of sex chromosomes relative to ancestry of autosomes in hybrid zones (Harrison & Larson, 2016;Chaturvedi et al, 2020).…”

“…Whereas the model can apply well to cases of contemporary hybridization and population mixtures, model misspecification can lead to incorrect inferences. Hence, the importance of model choice and fit has spurred further development of methods to gauge the appropriateness of the model for individual studies (Gompert et al, 2014b;Garcia-Erill & Albrechtsen, 2019;Chaturvedi et al, 2020).…”