Trait correlation networks invoke easy-to-measure functional traits as proxies for underlying inter-related physiological processes. This has improved understanding of plant strategies across broad taxonomic groups, but their explanatory ability may be more limited within plant taxa (e.g., congeners or conspecifics) or across resource gradients. Panicum virgatum L., is a common perennial C4 grass that occupies habitats varying considerably in mean annual temperature and precipitation. P. virgatum has diverged into two distinct ecotypes: the upland ecotype exhibits a quick return on investment of resources by combining high foliar nutrient content with lower investment in leaf construction; the lowland ecotype exhibits a slow return on investment of resources through low foliar nutrient content and higher investment in leaf construction. Here, we examined whether the correlations among key functional traits change as plants become increasingly water limited and whether these relationships can be explained by gene expression. We studied 10 P. virgatum genotypes, both upland and lowland established in a precipitation manipulation experiment with annual rainfall ranging from the driest 10% of years to the wettest 10% of years for the study site.
Results/Conclusions
Among four important functional traits—plant height, foliar N, leaf mass per area, and one-sided leaf area—two principle components explained 75% of the combined variation in these traits. High PC1 values represented high foliar N, low LMA, and shorter stature; PC2 values represented increased leaf area. Genotypes differed in both PC1 and PC2 scores (PC1: P < 0.001; PC2: P < 0.001), and PC1 and PC2 scores were higher in wet plots than in mean or dry plots (PC1: P < 0.045; PC2: P = 0.001). There were no significant genotype × precipitation effects (PC1: P = 0.07; PC2: P = 0.07). Gene expression corresponded to trait expression: the first ten gene expression PCs, explaining 51% of variation in gene expression, also explained 22% of variance in trait PC1 and 35% of variance in trait PC2 in linear mixed models. Moreover, each trait PC was explained by a different combination of gene expression PCs. Specifically, gene expression PC1, PC2, and PC3 significantly predicted different trait PCs, suggesting that different sets of genes influence these axes of phenotypic variation. Together, this suggests that correlations among functional traits change with water availability and these changes are related to differential gene expression.