Isolation-by-distance, resistance, instability and environment
Pairwise population fixation indices (FST ) confirmed patterns found in admixture and co-ancestry analyses and indicated moderate to high population structuring, ranging from 0.0178 (adjacent localities 1 and 2 of M. maxima ) to 0.661 (distant localities 1 and 6 of Ad. adscendens ; Fig. 1, Table S2). On average, across all localities in each species, A. costaricensisshowed lowest (FST = 0.123) and Ad. adscendens highest genetic differentiation (FST= 0.422). Comparing the two adjacent localities in each species, we found lowest FST in bee-pollinated M. maxima (0.018) and highest in passerine-pollinated A. costaricensis (0.105, Fig. 1).
Mantel’s tests on normalized FST revealed that, in all species, population differentiation strongly associated with geographic distance (IBD, Table 1, Fig. 4). Associations were significant in all species but A. costaricensis (R² 0.75, p = 0.055). In vertebrate-pollinated M. phlomoides, M. sanguinea ,M. tomentosa , and A. costaricensis ,FST also correlated most strongly with geographic distance (IBD). In bee-pollinated Ad. adscendens, correlation was strongest (but not significant) with geographic barriers (IBRTerrain) and historic habitat suitability (IBI, R² 0.796, p = 0.048), while in M. maxima , FSTassociated most strongly with current climatic suitability (R² 0.777, p = 0.023, IBRHabitat). In all species but A. costaricensis , we also detected a significant correlation with Pleistocene climatic instability (IBI, Table 1, Fig. S3, S6, S7). We did not detect any significant correlation with environmental distances (IBE, Fig. 3, Table 1). These patterns were mostly confirmed by multiple matrix regressions, albeit effects of IBI/ IBRHabitatwere not significant in A. adscendens and M. maxima , respectively (Table S19).
These results were mostly confirmed by GDM (Table 2, Table S20 details model fit). In bee-pollinated Ad. adscendens , we detected strong effects of IBD, IBRTerrain and IBE onFST (Fig. S8), although only IBD was retained as significant in the final model (Table 2). In bee-pollinated M. maxima , we found significant effects of current habitat suitability (IBRHabitat) and IBD (Table 2, Fig. S9).FST in hummingbird-bat-rodent-pollinated M. sanguinea and hummingbird-bat-pollinated M. phlomoides was significantly affected only by geographic distance (IBD, Table 2, Fig. S10, S11), while FST in hummingbird-bat-pollinated and M. tomentosa was most strongly affected by climatic instability since the LGM (IBI) and IBD (Table 2, Fig. S12). In A. costaricensis , again, no variable significantly affected FST (Table 2, Fig. S13).