Genetic diversity and population structure
The evaluation of linkage equilibrium and Hardy-Weinberg equilibrium of loci was done using exact tests with the functions test_LD and test_HW, respectively in the “genepop” package (v.1.1.7; Rousset, 2008). Descriptive summaries for each population across loci were calculated using the divBasic function in the “diveRsity” package (v. 1.9.90; Keenan, McGinnity, Cross, Crozier, & Prodöhl, 2013). Descriptive summaries for each locus across populations were calculated using locus_table function in the “poppr” package (v. 2.8.6; Kamvar, Brooks, & Grünwald, 2015; Kamvar, Tabima, & Grünwald, 2014).
Missing data were assessed using the info_table function in the “poppr” package (v. 2.8.6; Kamvar et al., 2015; Kamvar et al., 2014) and loci with more than 10% data missing and individuals with more than 20% data missing were removed. Descriptive summaries for populations and loci were performed on the entire data set and again on the data set after removing loci and individuals with unacceptable levels of missing data. A population level phylogeny based on the neighbor-joining clustering method using the Prevosti’s genetic distance model was generated with bootstrapped support using 1000 replicates with the aboot function in the “poppr” package (v. 2.8.6; Kamvar et al., 2015; Kamvar et al., 2014) and plotted using the Interactive Tree of Life v4 (Letunic & Bork, 2019).
Model-based putative population clustering was performed using STRUCTURE v2.3.4 (Pritchard, Stephens, & Donnelly, 2000). The number of genetic groups (K) present within the 509 individuals tested was determined running a continuous series of K = 1-22. The program was run with a burn-in of 30,000 and a run-length of 100,000 Markov Chain Monte Carlo (MCMC) replications in 20 independent runs using the LOCPRIOR model (sampling location information included) to account for weak structure signals in the dataset. The most likely number of clusters was determined using the Evanno method (Evanno, Regnaut, & Goudet, 2005) as implemented in STRUCTURE HARVESTER v0.6.94 (Earl & vonHoldt, 2012). The final analysis for K=3 was performed using a burn-in of 50,000 and 500,000 MCMC with 20 independent runs. Runs were summarized using CLUMPP v.1.1.2b (Jakobsson & Rosenberg, 2007) utilizing the Greedy algorithm and visualized with DISTRUCT v1.1 (Rosenberg, 2004).