PREMISE: Floral color is a stunning, complex trait that has long served as a model for connecting genetics, development, evolution, and ecology. Nevertheless, few mechanistic models relate flower color to the pigments that produce variation, nor has there been much exploration into theoretically possible flower color variation. Here we explored these topics using an anthocyanin-derived theoretical color-space approach. METHODS: We characterized flower color, floral anthocyanin concentrations, evolutionary history, and biogeography for 51 species of neotropical Ruellia to compare extant color diversity to an anthocyanin-derived theoretical color space and analyzed potential drivers of variation. To build the color space, we utilized reflectance spectrometry, HPLC, double-digest restriction-site-associated next-generation sequencing, and an extensive data set of Ruellia occurrences. RESULTS: An anthocyanin floral color model predicted a significant portion of the observed variation in reflectance spectra. Flowers spanned most of the theoretically possible color space, but with phenotypes clustered at the extreme edges of the space. Species of Ruellia exhibited less biochemical constraint than other well-studied lineages, commonly producing three or more types of anthocyanins (39%), but still showed evidence of constraint. Shared evolutionary history and biogeographical overlap were not strong predictors of color disparity between species pairs. CONCLUSIONS: Anthocyanins were primary predictors of flower color in Ruellia, but a significant portion of variation remained unexplained by our model, implicating additional mechanisms (e.g., co-pigmentation and pH) underlying flower color. Modeling color space provided a powerful framework for quantifying evolutionary constraints, offering insights into the mechanisms shaping phenotypic diversity.
