Evolvability and trait function predict phenotypic divergence of plant populations

Abstract

Understanding the causes and limits of population divergence in phenotypic traits is a fundamental aim of evolutionary biology, with the potential to yield predictions of adaptation to environmental change. Reciprocal transplant experiments and the evaluation of optimality models suggest that local adaptation is common but not universal, and some studies suggest that trait divergence is highly constrained by genetic variances and covariances of complex phenotypes. We analyze a large database of population divergence in plants and evaluate whether evolutionary divergence scales positively with standing genetic variation within populations (evolvability), as expected if genetic constraints are evolutionarily important. We further evaluate differences in divergence and evolvability-divergence relationships between reproductive and vegetative traits and between selfing, mixed-mating, and outcrossing species, as these factors are expected to influence both patterns of selection and evolutionary potentials. Evolutionary divergence scaled positively with evolvability. Furthermore, trait divergence was greater for vegetative traits than for floral (reproductive) traits, but largely independent of the mating system. Jointly, these factors explained ~40% of the variance in evolutionary divergence. The consistency of the evolvability-divergence relationships across diverse species suggests substantial predictability of trait divergence. The results are also consistent with genetic constraints playing a role in evolutionary divergence.

Keywords: adaptation; evolvability; genetic constraints; macroevolution.

Fig. 1.

Proportional population divergence of vegetative (green boxes) and floral (blue boxes) traits. The y-axis gives the proportional divergence of an average population, where a value of 1.1 indicates that the trait mean of an average population has evolved to be c. 10% larger or smaller than the grand mean. Thick lines across boxes indicate the median of each trait category, and thick lines within boxes indicate median values for each trait subcategory. Boxes extend from the first to third quartile, range bars extend to 1.5 times the interquartile range, and data points outside this range are shown as open circles. Sample sizes are given in parentheses for each trait subcategory, with the first number giving the number of estimates, and the second number the number of unique studies. Trait categories are defined in SI Appendix, Tables S1 and S2.

Fig. 2.

Univariate evolvability–divergence relationships for floral vs. vegetative traits, selfing vs. mixed-mating vs. outcrossing species, and populations measured in the greenhouse vs. outdoor common garden vs. field. Circle sizes are proportional to the square root of the number of populations studied. Regression lines show the estimated relationships, and solid dots indicate the expected divergence at the median evolvability in each group. See Table 1 for parameter estimates. The y-axis gives the proportional divergence of an average population from the grand mean (Fig. 1).

Fig. 3.

Examples of evolvability–divergence patterns in multivariate phenotype space. The panels explore patterns of divergence for leaf traits among 51 C. tectorum populations, blossom traits among 16 Costa Rican populations of D. scandens s.l., flower-size and floral display traits among 10 populations of L. siphilitica, and floral and vegetative traits among four Scandinavian populations of Arabidopsis lyrata. Lines indicate a 1:1 evolvability–divergence relationship passing through the mean of the data points, illustrating the near-isometric scaling of divergence with evolvability in all four cases. The y-axis gives the proportional divergence of an average population from the grand mean (Fig. 1). Focal directions (“traits”) are defined along the diagonal of the matrices (i.e., the original trait measurements), along the eigenvectors of the G-matrix, along the eigenvectors of the D-matrix, and along the eigenvectors of the mean within-population P-matrix. Coefficients of determination (r²) for the evolvability–divergence relationships are given in parentheses.

Fig. 4.

Summary of evolvability–divergence relationships (log-log regression slopes of population divergence on evolvability). The slopes are estimated for focal phenotypic directions represented by the original traits, and by the eigenvectors of the G-, D-, and P-matrices. (A) Frequency distribution of r² values across individual cases. (B) Scatterplot of slopes estimated for the directions represented by the G and D eigenvectors. Gray circles are individual cases (paired G- and D-matrices), with circle size proportional to the overall r² of the regression (including the original traits and the directions represented by the G and D eigenvectors). The black dots are species medians, and the blue dot is the median of the species medians. (C) Relationship between the slope of the evolvability–divergence relationships and the tightness of the relationship (r²) for each individual case.

Fig. 5.

Patterns of evolvability (e) and conditional evolvability (c) along divergence vectors. Most populations have diverged in directions of above-average evolvability, and more so for populations that have diverged farther from the focal reference population in which a G-matrix was estimated. Horizontal lines give the mean ($\stackrel{-}{e}$), minimum ($e_{\mathit{min}}$), maximum ($e_{\mathit{max}}$) and mean conditional ($\stackrel{-}{c}$) evolvability of the focal-population G-matrix. The minimum and maximum evolvabilities are the evolvabilities along the leading and trailing eigenvectors. In the summary panels (Right), values are scaled proportionally between the mean and the minimum and between the mean and the maximum.