The Evolutionary Dynamics of the Lion Panthera leo Revealed by Host and Viral Population Genomics

Antunes, Agostinho; Troyer, Jennifer L.; Roelke, Melody E.; Pecon-Slattery, Jill; Packer, Craig; Winterbach, Christiaan; Winterbach, Hanlie; Hemson, Graham; Frank, Laurence; Stander, Philip; Siefert, Ludwig; Driciru, Margaret; Funston, Paul J.; Alexander, Kathy A.; Prager, Katherine C.; Mills, Gus; Wildt, David; Bush, Mitch; O'Brien, Stephen J.; Johnson, Warren E.

Patterns of gene flow and divergence among populations were described using a variety of tests. First, to visualize subtle relationships among individual autosomal genotypes, threedimensional factorial correspondence analyses [23] (FCA) were performed in GENETIX [49] , which graphically projects the individuals on the factor space defined by the similarity of their allelic states. Second, neighbor-joining (NJ) analyses implementing the Cavalli-Sforza & Edwards' chord genetic distance [56] (D CE ) were estimated in PHYLIP 3.6 [57] , and the tree topology support was assessed by 100 bootstraps. Third, the difference in average H O and A was compared among population groups using a twosided test in FSTAT 2.9.3.2 [58] , which allows to assess the significance of the statistic OS x using 1,000 randomizations. Four, the equilibrium between drift and gene flow was tested using a regression of pairwise F ST on geographic distance matrix among all populations for host nDNA(microsatellites)/mtDNA and FIV Ple data. A Mantel test [59] was used to estimate the 95% upper probability for each matrix correlation. Assuming a stepping stone model of migration where gene flow is more likely between adjacent populations, one can reject the null hypothesis that populations in a region are at equilibrium if (1) there is a nonsignificant association between genetic and geographic distances, and/or (2) a scatterplot of the genetic and geographic distances fails to reveal a positive and monotonic relationship over all distance values of a region [60] . We also evaluated linearized F ST [i.e. F ST /(12F ST )] [24] among populations. We tested two competing models of isolation-by-distance, one assuming the habitat to be arrayed in an infinite one-dimensional lattice and another assuming an infinite two-dimensional lattice. Both models showed that genetic differentiation increased with raw and logtransformed Euclidean distances, respectively [24] . We determined the confidence interval value of the slope of the regression for the nDNA data using a non parametric ABC bootstrap [61] in GENEPOP 4.0 [62] .