The Andean and Atlantic forests are separated by the open vegetation corridor, which acts as a geographic barrier. However, these forests experienced cycles of connection and isolation in the past, which shaped the phylogeographic patterns of their biotas. We analysed the evolutionary history of the Rufous-capped Antshrike (Thamnophilus ruficapillus), a species with a disjunct distribution in the Atlantic and Andean forests and thus an appropriate model to study the effect of the open vegetation corridor and the Andes on the diversification of the Neotropical avifauna. We performed a phylogenetic/phylogeographic analysis, including the five subspecies, using mitochondrial and nuclear genomic DNA, and studied their differences in vocalizations and plumage coloration. Both the mitochondrial and nuclear DNA evidenced a marked phylogeographic structure with three differentiated lineages that diverged without signs of gene flow in the Pleistocene (1.0–1.7 Mya): one in the Atlantic Forest and two in the Andean forest. However, the two Andean lineages do not coincide with the two disjunct areas of distribution of the species in the Andes. Vocalizations were significantly different between most subspecies, but their pattern of differentiation was discordant with that of the nuclear and mitochondrial DNA. In fact, we did not find song differentiation between the subspecies of the Atlantic Forest and that of the northwestern Bolivian Andes, even though they differ genetically and belong to different lineages. Consistently, no differences were found in plumage coloration between the subspecies of the Atlantic Forest and that of the southern Andes. Our results suggest a complex evolutionary history in this species, which differentiated both due to dispersion across the open vegetation corridor, likely during a period of connection between the Andean and Atlantic forests, and the effect of the Bolivian Altiplano as a geographic barrier. In both cases, Pleistocene climatic oscillations appear to have influenced the species diversification.

We extracted genomic DNA from fresh tissue samples using DNeasy tissue extraction kit (Qiagen, CA, USA). To generate the genomic data we used double-digest restriction site-associated DNA sequencing (ddRADseq) following the protocol of Peterson et al. (2012) with modifications described in Thrasher et al. (2018). Briefly, for each sample we isolated ~500 ng of DNA, at a standardized concentration of 20 ng/µl, and digested it with the restriction enzymes SbfI High Fidelity (8 base bp recognition site; 5’-CCTGCAGG-3’) and MspI (4 base bp recognition site; 5’-CCGG-3’) (New England BioLabs, MA, USA). The digested DNA was ligated to P1 and P2 adapters on both ends (sequences are available in Peterson et al. 2012). We size-selected groups of 20 uniquely barcoded samples (index groups), retaining fragments between 400 and 700 bp using Blue Pippin (Sage Science, MA, USA). To incorporate the full Illumina TruSeq primer sequences and unique indexing primers into each library, we performed low cycle number PCR with Phusion High-Fidelity DNA Polymerase (New England BioLabs), with the following thermocycling profile: 98°C for 30 sec followed by 11 cycles at 98°C for 5 sec, 60°C for 25 sec, and 72°C for 10 sec with a final extension at 72°C for 5 min. We visualized the product of this amplification on a 1% agarose gel and performed a final 0.73 AMPure cleanup to eliminate DNA fragments smaller than 200 bp. Finally, sequencing was performed on an Illumina HiSeq 2500 lane at the Cornell University Biotechnology Resource Center as part of a larger sequencing batch, obtaining single-end 150-bp sequences, with an average mean sequencing depth of 151 reads per locus per individual (range: 104–220).

We demultiplexed, trimmed, filtered reads, assembled loci and called single nucleotide polymorphism (SNPs) with Ipyrad 0.7.28 (Eaton and Overcast 2020). We discarded sequences when a single base had a Phred quality score below 10 or more than 5% of bases had a Phred quality score below 20. Additional filtering was applied to only retain reads that did not have enzyme cleavage sites, adaptor sequences or index sequences. We used the default settings within Ipyrad to assemble loci and call SNPs: cluster threshold was set to 0.85, maximum fraction of heterozygous bases allowed in consensus sequences was set to 0.05, maximum number of SNPs allowed in a final locus was set to 0.2, maximum number of ambiguous sites was set to 0.05 and maximum number of shared polymorphic sites in a locus was set to 0.5. Finally, we only kept loci that were present in at least 80% of the samples.