Both intrinsic and extrinsic forces work together to shape connectivity and genetic variation in populations across the landscape. Here we explored how geography, breeding system traits, and environmental factors influence the population genetic patterns of Triodanis perfoliata, a widespread mix-mating annual plant in the contiguous US. By integrating population genomic data with spatial analyses and modeling the relationship between breeding system and genetic diversity, we illustrate the complex ways in which these forces shape genetic variation. Specifically, we used 4,705 single nucleotide polymorphisms to assess genetic diversity, structure, and evolutionary history among 18 populations. Populations with more obligately selfing flowers harbored less genetic diversity (π: R² = 0.63, P = 0.01, n = 9 populations), and we found significant population structuring (F_ST = 0.48). Both geographic isolation and environmental factors played significant roles in predicting the observed genetic diversity: we found that corridors of suitable environment appear to facilitate gene flow between populations, and that environmental resistance is correlated with increased genetic distance between populations. Last, we integrated our genetic results with species distribution modeling to assess likely patterns of connectivity among our study populations. Our landscape and evolutionary genetic results suggest that T. perfoliata experienced a complex demographic and evolutionary history, particularly in the center of its distribution. As such, there is no singular mechanism driving this species’ evolution. Together, our analyses support the hypothesis that breeding system, geography, and environmental variables shape the patterns of diversity and connectivity of T. perfoliata in the US. 

In late spring and early summer of 2017, leaf tissues were collected in the field from 18 populations of T. perfoliata (total = 76 individuals; range= 1-6 individuals/population) spanning the contiguous US, and from 6 individuals of T. biflora from southeast Missouri (Midwestern US) to serve as an outgroup for phylogenetic analyses. We used a CTAB protocol (Doyle & Doyle, 1987) to extract high-quality genomic DNA from silica-dried leaf tissue. Subsequently, RADSeq (Restriction site Associated DNA Sequencing) was performed at Floragenex, Inc. (Eugene, OR, USA) to identify genetic variants (Eaton, 2014). The restriction enzyme Sbf1 generated short fragments prior to the addition of sequencing adapters, and all samples were analyzed on the same flow cell with Illumina 1x91bp sequencing. After sequencing, quality control and sequence alignment were conducted using Bowtie (Langmead & Salzberg, 2012), BWA (Li H, 2011), and Velvet (Zerbino, 2011) and variant calling were performed using Samtools (Li et al., 2009). The final dataset consists of variant calls with a minimum sequencing depth of 15x, minimum Phred score of 20, and no more than 10% of missing genotypes.

A total of 9,716,774 raw reads were generated, of which 9,657,413 passed quality filters. These were used to build 5,646,126 provisional clusters, i.e., groups of sequencing reads that likely cover the same position in multiple samples, each with a minimum cluster depth of 5x and maximum cluster depth of 1500x. After read alignment and quality assessment, this yielded a final assembly that was approximately 5.2 Mb in length, consisting of 56,6649 contigs, each with a length of 92 bp. An average of 38.9% of the sequence reads from each sample aligned to a single position in this assembly. Variant calling yielded 4,705 single nucleotide polymorphic (SNP) sites observed >90% of the sequenced individuals of T. perfoliata.