Genetic diversity and population structure based on mitochondrial markers
We analyzed the concatenated mitochondrial DNA sequences (COI + cyt b = 1,130 bp) of 24 individuals of T. ruficapillus belonging to the five subspecies (Table S1). The specimens were collected in 16 localities from Peru, Bolivia, Argentina and Uruguay, covering a wide range of the species distribution (Figure 1a). Our results showed considerable variation within T. ruficapillus and, in particular, the subspecies that inhabits the Atlantic Forest, T. r. ruficapillus , clearly differed from the four Andean subspecies, with a mean genetic differentiation between 2.6% and 3.6% (Table 1). There was also considerable variation among the Andean subspecies, withT. r. cochabambae showing the highest genetic divergence both with respect to T. r. ruficapillus and with the remaining Andean subspecies (Table 1). Lower levels of differentiation were observed between T. r. subfasciatus , T. r. marcapatae and T. r. jaczewskii and in fact we found almost no differentiation (0.2%) between T. r. subfasciatus and T. r. marcapatae .
Both the haplotype network (Figure 1b) and the AMOVA (Table 1) show a marked phylogeographic structure within T. ruficapillus , with three clearly differentiated lineages: one formed by the subspecies present in the Atlantic Forest (T. r. ruficapillus ), another one by the subspecies distributed in the Andes of southern Bolivia and northwestern Argentina (T. r. cochabambae ), and the last one containing the remaining Andean subspecies from northern Bolivia to Peru (see Figure 1b). Consistently, ΦST values from the AMOVA were high and significant between the subspecies of these three different lineages (ΦST 0.7-0.83), but were not significant between the three subspecies of the northern lineage (T. r. subfasciatus , T. r. marcapatae and T. r. jaczewskii ; see Table 1). The haplotype network confirmed a lack of differentiation and shared haplotypes between T. r. subfasciatusand T. r. marcapatae , and low genetic differentiation between these two subspecies and T. r. jaczewskii , despite its geographically isolated and disjunct distribution.
Phylogenetic analyses and diversification dating based on mitochondrial DNA
Bayesian and maximum parsimony (MP) reconstructions recovered T. ruficapillus as monophyletic in relation to T. doliatus with maximum statistical support (Figure S3). The time-calibrated ultrametric phylogeny (Figure 1c) shows that the split between them occurred 1.65 million years ago (Mya) (95% HDP: 0.82-2.27). These reconstructions show three clades that correspond to the lineages described above: one includes the subspecies from the Atlantic Forest (T. r. ruficapillus ), another one is formed by T. r. cochabambae and the third one by T. r. subfasciatus , T. r. jaczewskii andT. r. marcapatae (see Figure 1c and Figure S3). In the case of the Bayesian and MP reconstructions the oldest split was between the subspecies from the Atlantic Forest and those from the Andean forest (Figure S3), with low values of both posterior probability and bootstrap support between the two Andean clades (0.63 and 63%, respectively). However, the time-calibrated ultrametric phylogeny indicated that the first split within the species, which occurred in the Pleistocene approximately 1.29 Mya (95% HPD: 0.71-2.10 Mya), was between the southernmost Andean subspecies (T. r. cochabambae ) and the rest of the subspecies (Figure 1c). This analysis suggested a more recent divergence between the subspecies from the Atlantic Forest (T. r. ruficapillus ) and the clade that includes the rest of the Andean subspecies (T. r. subfasciatus , T. r. jaczewskii andT. r. marcapatae ), approximately 0.97 Mya (95% HPD: 0.52-1.57 Mya). This split, however, also had low support (posterior probability: 0.58; Figure 1c).