Investigating the genetic architecture and adaptive relevance of complex traits in Cape Verde Arabidopsis

Understanding how organisms adapt to new environments is a key goal of evolutionary biology. Populations subject to abrupt environmental change must adapt quickly to avoid extinction. Small populations are especially vulnerable to habitat changes, confronting high extinction risk due to limited genetic variation and low efficiency of selection. Theory predicts that the age of a population and its long- term effective size should influence adaptation and trait architecture. Here, we investigate the mechanisms of adaptation after a sudden shift to a more arid climate using natural populations of Arabidopsis thaliana in Cape Verde (CVI). CVI Arabidopsis is found on two islands (Santo Antão and Fogo) and represents diverged, monophyletic lineages based on the near absence of shared polymorphisms with each other or the continent. Time to flowering was reduced in parallel on the islands, causing a consequent increase in fitness, and allowing adaptation to the arid CVI. This change was mediated by convergent de novo loss of function of two core flowering time genes: FRI in Santo Antão and FLC in Fogo. Our results reveal a case where expansion of the new populations coincided with the emergence and proliferation of these novel variants, consistent with models of rapid adaptation and evolutionary rescue. We further contrast the genetic architecture of flowering time in the recently formed small Ne Arabidopsis lineages from Cape Verde with their much older, larger Ne progenitor – the Moroccan population. We find that polygenicity is severely reduced in the colonizing populations and effect sizes of candidate loci are exponentially distributed, consistent with fitness measures showing evidence for directional selection in the islands. In addition to the major effect variants FRI K232X and FLC R3X, we identify candidate variants from core flowering time pathways as well as those that indirectly affect flowering time, including nutrient processing and light sensing. Surprisingly we find no effect of the well- known Cvi-0-EDI (CRY2 V367M) variant in the natural population. Our results provide a particularly clear empirical example of the effect of demographic history has on trait architecture