Population structure of S. corvina subspecies
We assessed genetic structure between individuals of S. corvinausing three complementary approaches. First, we estimate spatial genetic structure between samples using estimated effective migration surfaces (EEMS; Petkova et al., 2015). This software superposes a triangular grid of a predetermined number of demes (i.e., nodes) within the geographic area of interest and assigns the georeferenced samples to the closest deme. Then, EEMS estimates pairwise genetic distances between demes to identify paths of above- or below-average genetic similarity, while correcting for geographic distance. Deviations from the average genetic distance for a given geographic distance are inferred to result from increased or reduced gene flow. Therefore, EEMS finds geographic regions that act as barriers or corridors to gene flow. We explored the spatial structure using an arbitrary number of 200, 300, and 400 demes, and completed three independent runs per number of demes with different seeds to assess consistency and the effect of demes density (Petkova et al., 2015).
Second, we used the program ADMIXTURE (Alexander et al., 2009) to test for the most likely number of genetically distinct population clusters (K ) within Costa Rica and Panama, and we evaluated different population assignments and ancestry by identifying pure and admixed individuals. We ran ADMIXTURE for 100 iterations (-C), 200 bootstraps (-B), and 10 cross validations (-cv), with K values from 1 to 10. We found that K = 3 (see Results) was the most likely number of genetic populations and that they matched, in general, the three recognized subspecies. We therefore refer to pure individuals withq values > 0.95 by their respective subspecies’ name (i.e., corvina , hoffmanni , and hicksii ). Then, we classified hybrids as those individuals with q < 0.02 from the third genetic cluster and in which q from the second genetic cluster was at least 2x the value from the third genetic cluster. Admixed individuals with q > 0.1 from the third cluster were considered three-way hybrids (as in Scordato et al., 2017) and were removed from further analyses, together with the remaining individuals in which q from the second and third populations were similar, because they cannot be confidently assigned for pairwise comparisons. We calculated average Weir and Cockerham’sFST per SNP between pure individuals (i.e.,corvina vs hicksii , hoffmanni vs hicksii , corvina vs hoffmanni ) in VCFTOOLS to estimate genomic differentiation between subspecies pairs. Finally, we decomposed the genetic variation by a principal component analysis using the R package ADEGENET (Jombart, 2008), and visualized individuals based on the pure and hybrids categories described above based on q values from ADMIXTURE.