Herbarium specimens reveal links between Capsella bursa-pastoris leaf shape and climate

Abstract

Studies into the evolution and development of leaf shape have connected variation in plant form, function, and fitness. For species with consistent leaf margin features, patterns in leaf architecture are related to both biotic and abiotic factors. However, for species with inconsistent leaf margin features, quantifying leaf shape variation and the effects of environmental factors on leaf shape has proven challenging. To investigate leaf shape variation in species with inconsistent shapes, we analyzed approximately 500 digitized Capsella bursa-pastoris specimens collected throughout the continental U.S. over a 100-year period with geometric morphometric modeling and deterministic techniques. We generated a morphospace of C. bursa-pastoris leaf shapes and modeled leaf shape as a function of environment and time. Our results suggest C. bursa-pastoris leaf shape variation is strongly associated with temperature over the C. bursa-pastoris growing season, with lobing decreasing as temperature increases. While we expected to see changes in variation over time, our results show that level of leaf shape variation is consistent over the 100-year period. Our findings showed that species with inconsistent leaf shape variation can be quantified using geometric morphometric modeling techniques and that temperature is the main environmental factor influencing leaf shape variation.

Figure 1.

Overview of herbarium specimen selection, leaf shape types, and leaf shape analysis. (A). Map of the continental United States colored by climate region. Blue points represent herbarium specimen collection locations. (B). Schematic of leaf shape types. The left panel includes a representative of the C bursa-pastoris rosette taken from a herbarium specimen. [A-D]: Shull leaf shape types Simplex, Rhomboidea, Tenius, and Hetersis. [E-O]: Iannetta leaf shape types [E-H]: 1a-1d, [I-J]: 2b-2b, K: 3/4, L: 5, M:6, [N-O]: 7a-7b. (C). Mean leaf shape generated by Generalized Procrustes Analysis. The left leaf (blue outline) is the overall mean leaf shape and the right leaf is each individual leaf outline overlaid together in black with the mean leaf overlaid in blue. (D). Schematic of leaves included in leaf shape analysis, including true landmarks. Outlines of a representative sample of leaves (n = 12) included in this study are presented in blue. The two true landmarks, the leaf tip and leaf base, are represented by purple and orange points respectively. (E). Morphospace of theoretical leaves generated by inverse PCA. The morphospace projects five columns and rows of theoretical leaves generated by inverse PCA from leaf outlines included in this study.

Figure 2.

C. bursa-pastoris leaf morphospace, leaf shape types, circularity, and aspect ratio. (A). Morphospace PCA of leaves as classified by Shull leaf shape types. (B). Morphospace PCA of leaves as classified by Iannetta leaf shape types. (C,E). Graph of circularity (circ) against PC1. Leaves colored by their respective leaf shape type categories: Shull types (C) and Iannetta types (E). (D,F). Graph of aspect ratio (ar) by PC2, leaves colored by their respective leaf shape type categories: Shull types (D) and Iannetta types (F). The blue line represents the fitted linear regression and Tte gray band represents the 95% confidence interval.

Figure 3.

Modeling circularity and aspect ratio. (A). Circularity and aspect ratio exhibit a quadratic relationship. The blue line represents the fitted polynomial regression line. The gray band represents the 95% confidence interval. (B). Effects of climate on circularity and aspect ratio. The blue line represents the linear regression. The first column includes circularity and aspect ratio by the growing season (GS) climate conditions. The second column includes circularity and aspect ratio by the year long (YL) climate conditions. The third column includes circularity and aspect ratio by the date of collection (DOC) climate conditions. The model comparison deltaAIC is included for each climate x time model for both shape descriptors. The best model for explaining variance in circularity (lobing) was the GS model that includes climate region, with a deltaAIC score of 0. The best model for explaining variance in aspect ratio (size) was the DOC model including climate region.

Figure 4.

The relationship between average growing season (GS) temperature and circularity across all samples (left), in the South (middle) and in the Southeast (right). In all panels, the blue line represents the fitted linear regression. The two highest and two lowest circularity values for the south and southeast regions are colored in all three panels and represented by leaf images. Blue = highest circ, yellow = second highest circ, pink = second lowest circ, orange = lowest circ.