A multidimensional selective landscape drives adaptive divergence between and within closely related Phlox species

Abstract

Selection causes local adaptation across populations within species and simultaneously divergence between species. However, it is unclear if either the force of or the response to selection is similar across these scales. We show that natural selection drives divergence between closely related species in a pattern that is distinct from local adaptation within species. We use reciprocal transplant experiments across three species of Phlox wildflowers to characterize widespread adaptive divergence. Using provenance trials, we also find strong local adaptation between populations within a species. Comparing divergence and selection between these two scales of diversity we discover that one suite of traits predicts fitness differences between species and that an independent suite of traits predicts fitness variation within species. Selection drives divergence between species, contributing to speciation, while simultaneously favoring extensive diversity that is maintained across populations within a species. Our work demonstrates how the selection landscape is complex and multidimensional.

Fig. 1. Conceptual schematic representing divergence across scales of biological diversity in response to selection along axes of ecological variation.

Top panel represents the adaptive divergence between populations of two species shown as blue and red dots on different ecological habitats denoted by red and blue backgrounds. Bottom panel represents alternative scenarios of within-species local adaptation. Each colored point is a population adapted to the gradient of ecological conditions in the habitat represented by color across the background. In the scenario shown at the right (blue to red), the ecological gradient driving within-species local adaptation is parallel to the ecological gradient driving between species adaptive divergence. In the left scenario (blue to yellow) the gradient of within-species adaptation is orthogonal to the gradient driving divergence between species (blue to yellow).

Fig. 2. Geographic and environmental variation of broadly sympatric Phlox species.

A Ecological niche modeling predicts the geographic distributions of P. pilosa pilosa (pilosa; blue) and P. amoena amoena (amoena; red) across eastern North American (longitude and latitude indicated) with sampling locations shown as black diamonds (pilosa) and black circles (amoena). Locations of the common gardens are indicated by colored diamonds (amoena in red, pilosa in blue, deamii in green). B Environmental variation of pilosa and amoena summarized with a principal component analysis. Blue and red points indicate conditions of known populations of pilosa and amoena respectively. Black outlined points are populations sampled for transplant experiments and diamonds are the common garden sites. Representative flowers and leaves (not to scale) and pictures of local common garden site, of amoena (C), P. pilosa deamii (deamii) (D), and pilosa (E). Pictures taken by author B.E. Goulet-Scott.

Fig. 3. Performance of each taxon across three garden environments.

Fitness traits include A proportion of plants without herbivore damage, B total number of fruits, C total number of flowers, D aboveground biomass, and E proportion survived to the end of the experiment. Values plotted are taxon means ± standard error in each garden (n = 321 individuals per garden). The ANOVA evaluation of a mixed model analysis for each trait revealed a significant taxon-by-garden interaction for all traits. F Summary of the effect size of post-hoc contrasts evaluating local adaptation and home-garden advantage for each species. Positive values indicate local species performed superior while negative values indicate local species performed worse. Black points indicate Tukey Test contrasts are significant at p < 0.05. See Table 1 for full model results.

Fig. 4. Distance from the common garden predicts P. pilosa pilosa (pilosa) success indicating local adaptation.

As an example, A the relationship between population effect on total fruit set success and geographic distance for P. amoena amoena (amoena; red) and P. pilosa pilosa (pilosa; blue) populations with R² values from regression models indicated. B Distribution of R² values for regression models of population effect vs. distance measures among populations of amoena (red) and pilosa (blue) grown in all three experimental gardens. Solid points indicate significant evidence of local adaptation, where p < 0.05 in an F-statistic hypothesis testing of the model. Full model details are in Supplementary Data Table 4.

Fig. 5. Trait variation predicts fitness variation between and within species.

Principal components analysis describing phenotypic variation across three Phlox species (A). Points indicate values from individuals grown in the common garden experiment and black arrows indicate loadings of specific traits on the axes of variation. Images of four leaves (to scale) are connected to their points with gray arrows to demonstrate variation in shape along PC1 and area along PC2. B–G Relationship between fitness traits and leaf trait variation along PC1 and PC2. Colored points indicate the fitness values of individuals grown in the pilosa habitat garden. Gray lines indicate the linear model found significant relationships across all species and colored lines show relationships within each species (pilosa in blue, amoena in red, and deamii in green). For PC1, linear models find no relationships within species are significant while all within pilosa and amoena relationships are significant between PC2 and fitness traits. Full results in Supplementary Data Table 8. H Heatmap of reconstituted selection gradients (β values) for three fitness measures on each leaf trait with darker colors indicated higher values, negative values in purple, and positive values in orange. Results are shown for data from all species and just for pilosa for each fitness measure. β values are included in each well.