Population structure, phylogenetic tree and demographic history modelling based on ddRAD markers
We obtained reduced-representation genomic data for 18 specimens ofT. ruficapillus collected in 11 localities from Bolivia, Argentina and Uruguay that belong to the subspecies T. r. ruficapillus , T. r. cochabambae and T. r. subfasciatus(Figure 1a, Table S1) and represent the three clades obtained with mitochondrial DNA. The de novo assembly produced a total of 6,165 RADseq loci (13,485 SNPs) present in at least 80% of the individuals. This data set was used to perform the analyses described below.
Nuclear genomic variation was congruent with the results described for mitochondrial DNA, showing a clear differentiation between the three lineages. First, individuals clustered in the PCA in three clearly differentiated groups which corresponded to each of the subspecies (Figure 2a). The Structure analysis also supported the three main groups, as K = 3 was the model with the highest likelihood (Figure S4) and the three genomic clusters correspond to the mitochondrial lineages (Figure 2b). Only one individual of T. r. ruficapillus from Uruguay presented mixed genomic content, but the proportion of its genome belonging to the subfasciatus cluster was very low (less than 5%). The rest of the individuals did not show admixture in their genome, suggesting a lack of gene flow among these three subspecies.
The maximum likelihood reconstruction recovered a topology with the same three clades obtained in the Structure and PCA analyses, which also coincide with the mitochondrial lineages (Figure 2c). All nodes had maximum bootstrap support. The first split within the species occurred between T. r. ruficapillus (from the Atlantic Forest) and the subspecies from the Andean forest (T. r. cochabambae and T. r. subfasciatus ), with a subsequent separation between these Andean subspecies. This topology is congruent with those obtained in the MP and Bayesian analysis with mitochondrial data (but differ from the topology of the calibrated ultrametric tree; see Figure 1c and Figure S3).
Our demographic analysis in G-PhoCS was based on the topology of the nuclear phylogeny, which indicated that the initial split occurred between the eastern T. r. ruficapillus and the Andean subspecies (T. r. cochabambae and T. r. subfasciatus ), with a subsequent separation between the latter. The result, however, indicated that the three lineages split almost at the same time (Figure 3):T. r. ruficapillus diverged from the other two subspecies almost 1,7 million generations ago (Mga) (95% confidence interval: 1,5-2,0 Mga) and the two Andean subspecies diverged from each other only 440 generations after this first split. The effective population size of each of the subspecies is between one third and half of the ancestral effective population size estimated for the species (~1.5M individuals). However, the most noticeable result in this regard is the very small estimated ancestral effective population size of the lineage leading to T. r. subfasciatus andT. r. cochabambae , which was estimated to be only 313 individuals. This result suggests that this population could have experienced a bottleneck after (or during) its split from T. r. ruficapillus , expanding afterwards to reach the current effective population size of the Andean subspecies (~0.4M and ~0.55M individuals; Figure 3). This is consistent with the fact that Andean populations could have been isolated in refugia during the glacial periods after their split from T. r. ruficapillus (Chaves et al. 2011). Finally, this analysis suggests very low gene flow between subspecies (Figure 3), a result consistent with the PCA and Structure analyses.