The choices we make and the impacts they have: Machine learning and species delimitation in North American box turtles (Terrapene spp.).

Model-based approaches that attempt to delimit species are hampered by computational limitations as well as the unfortunate tendency by users to disregard algorithmic assumptions. Alternatives are clearly needed, and machine-learning (M-L) is attractive in this regard as it functions without the need to explicitly define a species concept. Unfortunately, its performance will vary according to which (of several) bioinformatic parameters are invoked. Herein, we gauge the effectiveness of M-L-based species-delimitation algorithms by parsing 64 variably-filtered versions of a ddRAD-derived SNP data set collected from North American box turtles (Terrapene spp.). Our filtering strategies included: (i) minor allele frequencies (MAF) of 5%, 3%, 1%, and 0% (= none), and (ii) maximum missing data per-individual/per-population at 25%, 50%, 75%, and 100% (= no filtering). We found that species-delimitation via unsupervised M-L impacted the signal-to-noise ratio in our data, as well as the discordance among resolved clades. The latter may also reflect biogeographic history, gene flow, incomplete lineage sorting, or combinations thereof (as corroborated from previously observed patterns of differential introgression). Our results substantiate M-L as a viable species-delimitation method, but also demonstrate how commonly observed patterns of phylogenetic discordance can seriously impact M-L-classification.

Contrasting signatures of introgression in North American box turtle (Terrapene spp.) contact zones.

Hybridization occurs differentially across the genome in a balancing act between selection and migration. With the unprecedented resolution of contemporary sequencing technologies, selection and migration can now be effectively quantified such that researchers can identify genetic elements involved in introgression. Furthermore, genomic patterns can now be associated with ecologically relevant phenotypes, given availability of annotated reference genomes. We do so in North American box turtles (Terrapene) by deciphering how selection affects hybrid zones at the interface of species boundaries and identifying genetic regions potentially under selection that may relate to thermal adaptations. Such genes may impact physiological pathways involved in temperature-dependent sex determination, immune system functioning and hypoxia tolerance. We contrasted these patterns across inter- and intraspecific hybrid zones that differ temporally and biogeographically. We demonstrate hybridization is broadly apparent in Terrapene, but with observed genomic cline patterns corresponding to species boundaries at loci potentially associated with thermal adaptation. These loci display signatures of directional introgression within intraspecific boundaries, despite a genome-wide selective trend against intergrades. In contrast, outlier loci for interspecific comparisons exhibited evidence of being under selection against hybrids. Importantly, adaptations coinciding with species boundaries in Terrapene overlap with climatic boundaries and highlight the vulnerability of these terrestrial ectotherms to anthropogenic pressures.

Sequence-based molecular phylogenetics and phylogeography of the American box turtles (Terrapene spp.) with support from DNA barcoding.

The classification of the American box turtles (Terrapene spp.) has remained enigmatic to systematists. Previous comprehensive phylogenetic studies focused primarily on morphology. The goal of this study was to re-assess the classification of Terrapene spp. by obtaining DNA sequence data from a broad geographic range and from all four recognized species and 11 subspecies within the genus. Tissue samples were obtained for all taxa except for Terrapene nelsoni klauberi. DNA was extracted, and the mitochondrial DNA (mtDNA) cytochrome b (Cytb) and nuclear DNA (nucDNA) glyceraldehyde-3-phosphate-dehydrogenase (GAPD) genes were amplified via polymerase chain reaction and sequenced. In addition, the mtDNA gene commonly used for DNA barcoding (cytochrome oxidase c subunit I; COI) was amplified and sequenced to calculate pairwise percent DNA sequence divergence comparisons for each Terrapene taxon. The sequence data were analyzed using maximum likelihood and Bayesian phylogenetic inference, a molecular clock, AMOVAs, SAMOVAs, haplotype networks, and pairwise percent sequence divergence comparisons. Terrapene carolina mexicana and T. c. yucatana formed a monophyletic clade with T. c. triunguis, and this clade was paraphyletic to the rest of T. carolina. Terrapene ornata ornata and T. o. luteola lacked distinction phylogenetically, and Terrapene nelsoni was confirmed to be the sister taxon of T. ornata. Terrapene c. major, T. c. bauri, and Terrapene coahuila were not well resolved for some of the analyses. The DNA barcoding results indicated that all taxa were different species (>2% sequence divergence) except for T. c. triunguis - T. c. mexicana and T. o. ornata - T. o. luteola. The results suggest that T. c. triunguis should be elevated to species status (Terrapene mexicana), and mexicana and yucatana should be included in this group as subspecies. In addition, T. o. ornata and T. o. luteola should not be considered separate subspecies. The DNA barcoding data support these recommended taxonomic revisions. Because conservation efforts are typically species-based, these results will be important for facilitating successful conservation management strategies.