Genetic variation among the five species
To assess genetic diversity within each species, we calculated nucleotide polymorphism, haplotype polymorphism (Hd), nucleotide diversity (π/kb), polymorphic sites (θ/kb), and Tajima’s D (Table 2). Except for S. torqueola , four of the species have comparable π values, ranging from 2.73-4.52(/kb). I. mcclellandii , A. hueti and P. pectoralis have significantly high θ values (5.68 – 8.51/kb) compared to L. lutea (1.22/kb) and S. torqueola (1.8/kb). Hd values are similar in all five species (0.83 to 0.972). Tajima’s D is significantly negative in I. mcclellandii and L. lutea , suggesting substantial recent population expansion or a selective sweep in these species (Tajima 1989). Population expansion is also supported by their radiating haplotype networks (Figure 2). S. torqueola , despite wide regional sampling and a relatively large number of individuals (n = 39), appears to differ substantially from the other species in variation. It has a relatively low number of SNPs (single versus double digits) and low values of π and θ. Its Tajima’s D is close to zero, signaling little recent population expansion. As a resident of moderate elevation (700-1500m), S. torqueola also presents a highly homogenous population composition suggestive of substantial gene flow (Figure 2E).
Except for S. torqueola , the species generally display comparable haplotype diversity, complexity, and patterns of recent population expansion. Most of their substitutions are synonymous, and their networks contain conspicuous patterns of haplotype radiation from highly diverse ancestral clusters (e.g., Figure 2A-D). In all 5 species’ networks, divergence among haplotypes of 2-5 bp is common between populations, strongly indicating rapid substitution combined with homogenous haplotype distribution in populations following isolation. Some of the species included a conspicuous core haplotype distinct from derived haplotypes, e.g., the blue haplotypes in I. mcclellandiiand L. lutea (Figure 2A and 2C). This pattern is less notable in other circumstances, e.g., the green cluster in L. lutea (Figure 2C). Under the conventional substitution rate of 10-8/site/year (Nguyen and Ho 2016), this pattern suggests divergence times of 0.1~0.3 Ma among diverged haplotypes in all species.
Interestingly, the haplotype network of L. lutea suggests this species went through a strong bottleneck, followed by rapid radiation of each of its haplotype groups (Figure 2C). The two star-like clusters are separated by a 2 bp substitution. One of these substitutions (at bp 1033) is non-synonymous, causing an amino acid change from Isoleucine (assumed to be the ancestral state by reference to the outgroup,L. argentauris , in the green cluster) to Valine (in the blue cluster). The high haplotype diversity in the ancestral cluster and the radiating haplotype network in the derived cluster suggest recent expansion of the new non-synonymous substitution, which is homogenously distributed among the sampled (blue) populations.
Genetic clustering returned different results for the 5 species. Overall, all species were confidently clustered by the gmyc(Supporting information). I. mcclellandii split into 3 maximum likelihood clusters, while A. hueti split into 4 clusters,L. leiothrix split into 6 clusters, P. pectoralis into 15 clusters, and S. torqueola into 5 clusters. None of the clusters show strong alias to populations except for P. pectoralis . InP. pectoralis , the clustering divided populations into smaller closely related groups (Supporting information). Gmyc also inferred the threshold time of the split. The threshold time of I. mcclellandii, A. hueti, L. lutea and S. torqueola are in the approximate range of 0.2 MYA, while the inferred split time of P. pectoralis is later than 23 000 and 12 000 years ago (Supporting information). The clustering result is supported with significant likelihood ratio test. Homogeneous ancestry is an indication of simultaneous demographic changes. While for P. pectoralis, the short threshold time and fine arranged population clusters indicate that they probably were dispersed recently, e.g., in the last glacial cycle.