Phenotypic plasticity in tropical butterflies is linked to climatic seasonality on a macroevolutionary scale

Phenotypic plasticity can be adaptive in fluctuating environments by providing rapid environment-phenotype matching and this applies particularly in seasonal environments. African Bicyclus butterflies have repeatedly colonized seasonal savannahs from ancestral forests around the Late Miocene and many species now exhibit seasonal polyphenism. On a macroevolutionary scale, it can be expected that savannah species will exhibit higher plasticity due to them experiencing stronger environmental seasonality than forest species. We quantified seasonality using environmental niche modelling, and surveyed the degree of plasticity in a key wing pattern element (eyespot size) using museum specimens. We show that species occurring in highly seasonal environments display strong plasticity, while species in less seasonal or aseasonal environments exhibit surprisingly variable degrees of plasticity, including strong to no plasticity. Furthermore, eyespot size plasticity has a moderate phylogenetic signal and the ancestral Bicyclus likely exhibited some degree of plasticity. We propose hypotheses to explain the range of plasticity patterns seen in less seasonal environments, and generate testable predictions for the evolution of plasticity in Bicyclus. Our study provides one of the most compelling cases showing links between seasonality and phenotypic plasticity on a macroevolutionary scale and the potential role of plasticity in facilitating the colonization of novel environments.

Methods
The dataset has two components: (1) phenotypic data- the data was collected by measuring the eyespot size and wing length (which was then used to calculate relative eyespot size) from the photographed museum specimens of African Bicyclus butterflies. (2) distribution data- this is the distribution/location data for the African Bicyclus butterflies collated from museum records and citizen science online platforms. All the data (in .csv and .xlsx) has been processed in R statistical software, except the environmental niche modelling which was carried out using the Maxent software.