2.4 Population structure analysis
To explore the phylogeographic structure of ribotypes/chlorotypes, gene diversity within populations (hS ), gene diversity (hT ), index of gene diversity of total populations (NST ) and index of genetic differentiation between populations (GST ) were calculated using the HAPLONST (Pons & Petit, 1996). This software was also used to compute the test statistic U that compares the values of NST and GST , which indicates the presence of phylogeographic structure ifNST is higher than GST(Pons & Petit, 1996).
Since the cpDNA is maternally inherited in Primulina , there is no recombination among loci and structure analysis makes no sense for such dataset. Population structure of nrDNA sequences was inferred using the Bayesian clustering procedure implemented in the software STRUCTURE v.2.3.4 (Evanno et al. , 2005), that identifies the most probable number (K ) of genetic clusters of origin of the sampled individuals, and assigns individuals to clusters. We used Markov Chain Monte Carlo (MCMC) iterations as implemented in the software to explore the parameter space considering individual memberships to the Kclusters, ranging from K = 1 (null hypothesis of panmixia) toK = 10 (the total number of populations sampled). Three independent runs were performed with an admixture model at 105 MCMC iterations and a 105burn-in period. The most likely number of population groups (K , indicating the number of true clusters in the data) and the model values (△K ) according to the second-order rate of Change of clusterK that best fit the data was calculated in Structure Harvester (Earl & vonHoldt, 2012). The graphical representation of results was accomplished in the CLUMPAK server (http://clumpak.tau.ac.il/index. html).
The genetic differentiation (Fst ) and gene flow (Nm ) between populations and among different regions for nrDNA and cpDNA sequences separately. An Analysis of Molecular Variance (AMOVA) was conducted on nrDNA and cpDNA sequences separately to test genetic differentiation within populations, among regions and among populations within regions using GENALEX v. 6.503 (Peakall & Smouse, 2012). To test whether there was local genetic variation attributable to isolation by distance (IBD), Neiʼs genetic distances (Nei, 1973) calculated with by MEGA v. 6.5 and geographic distances (in km) between all pair-wise combinations of the ten populations sampled were subjected to a Mantel test (Mantel, 1967) in GENALEX v. 6.503. The genealogical relationships between ribotypes/chlorotypes were inferred from the Median-Joining network (MJ) of NETWORK v4.6.1.0 (http://www.fluxus-Engineering.com/). In order to identify and quantify potential genetic discontinuities and biogeographic boundaries between populations from both nrITS and cpDNA datasets, we calculated the Monmonierʼs maximum-difference algorithm in Barrier v.2.2 (Manni et al., 2004). The robustness of these barriers was assessed by bootstrap, as in the Barrier v.2.2.