Computational model of flower pattern evolution predicts spontaneous emergence of boundary cell types across the petal epidermis.

Petal patterns play an important role in the reproductive success of flowering plants by attracting pollinators and protecting against environmental factors. Some transcription factors (TFs) involved in petal epidermal cell differentiation have been identified, but little is known about the upstream processes that pre-pattern the petal surface to establish their expression domains. Here, we have developed a computational model of petal pattern evolution to investigate this pre-patterning phase, selecting for gene regulatory network (GRNs) that divide the petal into proximal and distal domains to create a bullseye pattern. We found that bullseye evolution was often accompanied by the spontaneous emergence of a third cell type at the boundary between proximal and distal regions, which we validated experimentally in Hibiscus trionum. These boundary cells appeared despite not being explicitly selected for, and arose more often when gene expression was modelled as a noisy process, suggesting they buffer against developmental variability. Our results illuminate the early steps of petal pattern formation and demonstrate that novel cell types can arise spontaneously from selection on other cell types when developmental robustness is considered.