Assessment of barriers to gene flow in S. hesperis
The northern and southern S. hesperis lineages identified in this study are largely parapatric in their distributions. It is therefore possible that genetic drift resulting from barriers to dispersal (i.e., vicariance and speciation without selection or mutation-order speciation), rather than ecologically-based divergent selection (i.e., ecological speciation), has been the dominant evolutionary process underlying diversification and reinforcement of the identified lineages (sensu Nosil, 2012). To test for this possibility, we assessed whether there are present-day barriers to dispersal between the northern and southern S. hesperis lineages related to arrangements of suitable habitat. This was accomplished using resistance surfaces parameterized as the inverse of predicted habitat suitability (McRae & Beier 2007; Wang et al. 2008; Storfer et al. 2010; Wang et al. 2012; MacDonald et al. 2020). This approach assumes that organisms are more likely to disperse within suitable habitat and experience high resistance when moving through unsuitable habitat; large stretches of unsuitable habitat thereby pose significant barriers to dispersal (Coyne & Orr 2004; Crispo et al. 2006; McRae 2006; McRae & Beier 2007; Thorpe et al. 2008, 2010; Sánchez-Ramírez et al. 2018). We averaged resistance surfaces of the northern and southern S. hesperis lineages to generate a single resistance surface reflecting the probability that dispersing individuals of the two lineages will come into contact that could result in reciprocal gene flow. Using this single resistance surface, we then calculated pairwise resistance distances between all individuals using the R package gdistance (van Etten 2017) and organized them into a pairwise matrix. These resistance distances are analogous to circuit distances (McRae & Beier 2007), measuring expected random-walk commute time between nodes (i.e., localities of sequenced individuals) in a graph (i.e., resistance surface) (Chandra et al. 1996). We also generated a pairwise matrix of Euclidean distances between all individuals using the R package sp (Pebesma & Bivand 2005). A third pairwise matrix (“lineage distance”) indicated whether individuals belonged to the same lineage (value = 0) or to different lineages (value = 1).
We used a partial mantel test to evaluate whether resistance distances between individuals of different lineages were significantly greater than those between individuals of the same lineage after controlling for Euclidean distance. This analysis effectively evaluates whether significant barriers to dispersal exist between the northern and southern S. hesperis lineages. A significant correlation between resistance distance and lineage distance after partialling out Euclidean distance would suggest that reduced gene flow resulting from barriers to dispersal (vicariance) cannot be ruled out as a principal mechanism reinforcing the genomic integrity of the northern and southern S. hesperis lineages. Alternatively, the absence of a significant correlation between resistance distance and lineage distance after partialling out Euclidean distance would suggest that ecologically-based divergent selection likely contributes to reinforcement of the northern and southern S. hesperis lineages.