We assessed population structure and the spatio-temporal pattern of diversification in the Glossy Antshrike Sakesphorus luctuosus (Aves, Thamnophilidae) to understand the processes shaping the evolutionary history of Amazonian floodplains and address unresolved taxonomic controversies surrounding its species limits. By targeting ultraconserved elements (UCEs) from 32 specimens of S. luctuosus, we identified independent lineages and estimated their differentiation, divergence times and migration rates. We also estimated current and past demographic histories for each recovered lineage. We found evidence confirming that S. luctuosus consists of a single species, comprising at least four populations, with some highly admixed individuals and overall similar levels of migration between populations. We confirmed the differentiation of the Araguaia River basin population (S. l. araguayae), and gathered circumstantial evidence indicating that the taxon S. hagmanni may represent a highly introgressed population between 3 distinct phylogroups of S. luctuosus. Divergence time estimates between populations seem to be recent, occurring during the last 183 kya. Signs of population expansions were detected for populations attributed to subspecies S. l. luctuosus, but the S. l. araguayae population had probably maintained its effective size through time. Our results support that S. luctuosus has had a complex population history, resulting from a high dependence on southeastern “clear-water” habitats and their availability through time. Spatial and demographic expansions towards the western “white water” flooded forests might still be ongoing. Our study reinforces the view that isolation due to absence of suitable habitat has been an important driver of population differentiation within Amazonian flooded forests, but also that differences between várzeas (“white water” floodplains, mostly in southwestern Amazonia) and igapós (“clear- water” floodplains, especially located in the east) should be further explored as powerful drivers of micro-evolution.

Genomic DNA was extracted from tissues of 32 specimens of S. luctuosus from across its known range. We used DNeasy Blood & Tissue kit (Qiagen) for DNA extraction, and Qubit® 2.0 Fluorometer (Life Technologies) to assess quantity and quality of the extracted DNA. Sequence capture and sequencing of Ultra Conserved Elements (hereafter UCEs) were performed according to Faircloth et al. (2012) by RAPiD Genomics (Gainesville, FL, USA). More than 2,300 UCEs and 97 exons were targeted (Harvey, Aleixo, Ribas, & Brumfield, 2017; Zucker et al., 2016).

We followed the PHYLUCE pipeline (Faircloth, 2016, 2017) to first remove adapters, barcodes and low quality sequence regions using Illumiprocessor 2.0.7 (Faircloth, 2013), with the trimming tool Trimmomatic 0.32.1 (Bolger, Lohse, & Usadel, 2014); and to assemble trimmed reads using Trinity (Grabherr et al., 2011). These analyses were performed using default parameters. Contigs were blasted against the probe set of exons and UCEs using phyluce_match_contigs_to_probes. The annotated set was divided by locus, and each was aligned with MAFFT using the scripts phyluce_assembly_get_match_counts, phyluce_assembly_get_fastas_from_match_counts, and phyluce_align_seqcap_align. We used seqcap_pop pipeline to extract one biallelic SNP from each locus (Harvey, Smith, Glenn, Faircloth, & Brumfield, 2016). Briefly, the longest contigs were retrieved from an incomplete matrix, which included all loci identified in all 32 Glossy Antshrike specimens sampled using BWA (Li & Durbin, 2009). These contigs were blasted against the zebra finch genome (Taeniopygia guttata v. 3.2.4, NCBI code: GCF_000151805.1), to annotate loci linked to the Z chromosome. These loci were removed. Then, one biallelic SNP, present in all samples, was randomly chosen from each of the alignments, using Genome Analyses Tool Kit (McKenna et al., 2010) and VCFTools (Danecek et al., 2011). To identify Fst outliers (i.e. loci likely under selection), we first performed a principal component analysis in the SNPs dataset to set a preliminary population structure using PAST3 (Hammer, Harper, & Ryan, 2001), and then we run BayeScan (Foll & Gaggiotti, 2008). Loci putatively under selection were also excluded.

This dataset includes 1008 SNPs.