Discussion
The combination of mitochondrial and nuclear DNA, complemented with data on vocalizations and plumage coloration, allowed us to study the evolutionary history of T. ruficapillus . Our findings provide a good starting point towards understanding the history of diversification of this species, the relative patterns of genotypic and phenotypic variation, and the role played by the open vegetation corridor and the Andes Mountains in this process.
First, the mitochondrial DNA results suggest that the split betweenT. doliatus and the lineage that includes T. ruficapillusand T. torquatus (which are sister species; Brumfield and Edwards 2007) occurred 1.65 Mya, within the Pleistocene. This indicates that the origin of T. ruficapillus took place even more recently within the Pleistocene. The subsequent diversification within the species generated a phylogeographic structure with three main genetic lineages: one in the Atlantic Forest (T. r. ruficapillus ), another one including the southernmost Andean subspecies (T. r. cochabambae ), and finally one lineage formed by the three remaining Andean subspecies (T. r. subfasciatus , T. r. marcapatae and T. r. jaczewskii ). The various analyses based on both the mitochondrial DNA and the nuclear genome recovered these three genetic clusters (although we note that for genomic analyses we were only able to include T. r. subfasciatus as a representative of the northern Andean lineage).
The relationship between these three main lineages is not clear and it can actually be considered a polytomy. The Bayesian and MP mitochondrial reconstructions, as well as the nuclear genomic phylogeny, suggested that the first split within the species took place between the lineage from the Atlantic Forest and an Andean lineage, with a subsequent separation within the latter. However, the time-calibrated ultrametric phylogeny inferred that the first split occurred between the two Andean lineages. Surprisingly, the topology of this ultrametric phylogeny indicates that shortly after the Andean split, the colonization of the lowland areas of the Atlantic Forest occurred from the northern Andean lineage and not from the lineage that inhabits northwestern Argentina and southern Bolivia, which is much geographically closer to the Atlantic Forest. This discordance in the relationship among linages could be due to the temporal proximity of both splits, a result common to all analyses. In particular, the time-calibrated ultrametric phylogeny based on mitochondrial DNA suggested that the first split within T. ruficapillus (in this reconstruction the split between the two Andean lineages) occurred approximately 1.3 Mya, and the colonization of the Atlantic Forest approximately 1 Mya. The G-PhoCS analysis, which was based on nuclear DNA, suggested that the two splits between lineages took place approximately 1.7 million generations ago. If we consider that the generation time for passerines could be estimated to be around 1 year on average (Wang 2004, Cabanne et al. 2008), these two results (from the analyses of mitochondrial and nuclear genomic DNA) are relatively concordant. Consistent with the temporal proximity of these two deepest splits and their different sequence of occurrence depending on the reconstruction methodology, one should note that the support of the nodes that define the relationship of these three lineages was very low in all mitochondrial reconstructions (63% bootstrap support in the MP analysis, 0.63 posterior probability in the Bayesian phylogeny and 0.58 posterior probability in the time-calibrated phylogeny; see Figure 1c and Figure S3). Regardless of the subtle difference in the timing of the splits and the uncertainty in their order, our results are congruent in suggesting that the main events of diversification of T. ruficapillus that generated its current patterns of population diversity emerged in a short time period during the Pleistocene, shortly after the origin of the species.
Taking together our genetic and genomic results, including the lack of gene flow between the Atlantic Forest and Andean forest populations, we can confirm the role of the open vegetation corridor as a geographic barrier for T. ruficapillus . Nevertheless, the origin of the allopatric populations of this species should be explained as a result of dispersal through the open vegetation corridor and not as a vicariance event due to its establishment. This is because the Atlantic Forest and the Andean forest became isolated by this barrier before the Pleistocene (Costa 2003, Trujillo-Arias et al. 2017, Cabanne et al. 2019, Lavinia et al. 2019) and therefore before the origin of T. ruficapillus and the split between its Atlantic and Andean lineages (which is dated to approximately 1-1.7 Mya by our analyses). This suggests a relevant role of the Pleistocene climatic cycles on the diversification of T. ruficapillus , since the dispersion through the open vegetation corridor could have occurred during one of the periods of connection between the Atlantic and Andean forests. If the first split within the species occurred between the lineage from the Atlantic Forest and the ancestral lineage from the Andean forest (as suggested by most of our analyses, including the phylogeny based on genomic DNA), it is difficult to establish the direction of the dispersion and colonization. On the other hand, if the first split occurred within the Andes (as suggested by the time-calibrated Bayesian analysis), it could be considered that the species originated in the mountain range and after its first split between the Andean lineages the dispersion occurred to the east, allowing the colonization of the Atlantic Forest. In this second scenario, the Bayesian phylogeny indicates that the colonization occurred from the northern lineage of the Andean forest, suggesting that it could have taken place across the Cerrado as proposed by Silva (1994), and not across the Chaco as hypothesized by Nores (1992). We note that this is only a possibility since our lack of sampling on the northern portion of the species distribution in the Atlantic Forest prevents us from drawing robust conclusions about the route/s of connection. Moreover, the fact thatT. doliatus (the sister species of T. ruficapillus ) is widely distributed to the north of T. ruficapillus and also has populations in the east and west of the Neotropics, precludes us from assuming which is the area in which these two species may have split, which could have helped distinguishing among hypothesized colonization scenarios. Further studies with a more geographically comprehensive sampling of T. ruficapillus and the inclusion of T. doliatus are needed to fully elucidate this matter.
The Andes Mountains have also played a relevant role in the diversification of T. ruficapillus , as indicated by the existence of two distinct Andean lineages without evidence of gene flow. Surprisingly, these two lineages do not coincide with the two disjunct areas of distribution of this species in the Andes (at least from a mitochondrial perspective; see Fig. 1): one lineage is formed by the populations of T. r. cochabambae (from northwestern Argentina and southern Bolivia) and the other one is constituted by the three remaining Andean subspecies (T. r. subfasciatus , T. r. marcapatae and T. r. jaczewskii from northern Bolivia and Peru, which span both areas of disjunct distribution in the mountain range and only exhibit a shallow differentiation in the case of T. r. jaczewskii , the northern subspecies that is geographically isolated). This pattern suggests that there is a geographic barrier in northern Bolivia, which in fact coincides with phylogeographic breaks found in the same area in various other montane bird species (Cadena et al. 2007, Weir 2009, Chaves et al. 2011, Valderrama et al. 2014, Gutiérrez-Pinto et al. 2019). Given that our analyses indicated that the divergence between these two Andean lineages occurred around 1.3-1.7 Mya, during the Pleistocene, one could hypothesize that Pleistocene climatic oscillations have also played a role in the diversification of this species in the Andes (Weir 2006, 2009, Chaves et al. 2011, Valderrama et al. 2014).
The vocal analyses showed significant differences in one or various song variables between most of the subspecies, which at first glance seems to be consistent with the presence of a marked phylogeographic structure. In fact, only two of the subspecies comparisons did not show vocal differentiation. One of them was between T. r. subfasciatus andT. r. marcapatae , which were also the pair of subspecies with the lowest genetic divergence and shared haplotypes. The second pair of subspecies without song differences was T. r. ruficapillus andT. r. subfasciatus , an unexpected result given that they belong to different mitochondrial lineages and their areas of distribution are disjunct, with the former present in the Atlantic Forest and the latter in the Andean forest of northwestern Bolivia. On the other extreme of our results, T. r. jaczewskii was one of the most vocally differentiated subspecies, significantly differing in its song from all other subspecies. This result could be expected given that T. r. jaczewskii is geographically isolated from the other populations of the species, but contrasts with the relatively shallow mitochondrial DNA differentiation in this subspecies (nuclear DNA could not be analysed), particularly when compared with T. r. subfasciatus and T. r. marcapatae , which in fact are part of its same lineage but significantly differed in their songs.
The finding of these discordances between the genetic/genomic patterns of divergence and vocal differentiation in T. ruficapillus is surprising given that this species is a suboscine, and therefore its song is likely developed without learning (Kroodsma and Konishi 1991, Isler et al. 1998, Touchton et al. 2014), being free from the influence of cultural evolution and possessing a more direct connection with genetic divergence than in oscines (Seddon 2005). In fact, previous intraspecific studies of suboscine song variation have shown more concordant patterns of vocal and genetic structuring (Isler et al. 2005, García et al. 2018, Acero-Murcia et al. 2021, Bukowski et al. 2024), although with some cases of divergent lineages lacking song differentiation in suboscines (García et al. 2018) and other birds with innate songs (Nwanko et al. 2018), as was the case for T. r. ruficapillus and T. r. subfasciatus in this study. Future vocal studies of this species should analyse the putative effect of natural and sexual selection, which could cause different patterns of song variation compared to neutral genetic markers even in species with innate songs (García et al. 2018, 2023). On the other hand, the vocal pattern of variation in this species can also be affected by the adaptation to habitat characteristics and their effect on sound transmission, which could delay or accelerate song differentiation in relation to genetic phylogeographic patterns, depending on the similarity or disparity of the various forest types along the species distribution (Morton 1975, Wiley 1991, Seddon 2005, Tubaro and Lijtmaer 2006).
Finally, and in spite of their allopatry and marked mitochondrial and nuclear DNA differentiation, no significant differences were found between the Atlantic Forest subspecies T. r. ruficapillus and the Andean subspecies T. r. cochabambae in plumage coloration (neither for males nor for females). The lack of correlation between colour differentiation and genetic divergence among intraspecific lineages has been found in previous studies of Neotropical birds (e.g. Trujillo-Arias et al. 2020, Paulo et al. 2023), but analyses including the remaining three subspecies, as well as a larger geographic sampling, are needed to better assess the pattern of colour variation in this species.