Karst ecosystems often contain extraordinary biodiversity, but the complex underground aquifers of karst regions present challenges for assessing and conserving stygobiont diversity and investigating their evolutionary history. We examined the karst-obligate salamanders of the Eurycea neotenes species complex in the Edwards Plateau region of central Texas using population genomics data to address questions about population connectivity and the potential for gene exchange within the underlying aquifer system. The Eurycea neotenes species complex has historically been divided into three nominal species, but their status, and spatial extent of species ranges, have remained uncertain. We discovered evidence of extensive admixture within the species complex and with adjacent lineages. We observed relatively low levels of differentiation among all sampling localities which supports the hypothesis of recent divergence. Nominal taxonomy, aquifer region and geography accounted for a modest amount of the overall population genomic variation, but these predictors were largely confounded and difficult to disentangle. Importantly, the taxonomy of the three nominal species does not reflect the admixture apparent in clustering analyses. Inference of migration events revealed a complex pattern of gene exchange, suggesting that Eurycea salamanders have a dynamic history of dispersal through the aquifer system. These results highlight the need for greater understanding of how stygobiont populations are connected via dispersal and gene exchange through karst aquifers. These results also highlight the applicability of population genomics data as a powerful lever for investigating connectivity among populations in systems where direct detection of dispersal paths is difficult, as in underground, aquatic systems.

Matrix of posterior genotype probabilities for 607 salamanders at 16,094 SNP loci generated by genotyping-by-sequencing methods (similar to ddRAD.seq methods). The hierarchical Bayesian clustering algorithm ENTROPY was used to generate these genotype probabilities. A locality file is included along with a script and matrix of predictors for Redundancy Analysis and script and data object for Bayesian Linear Mixed Model are included.