Tempo and mode of winter diapause evolution in butterflies

Quantifying the tempo and mode via modern phylogenetic comparative methods can provide key insights into how selection and constraints shape trait evolution on a macroevolutionary time scale. Here we elucidate the evolution of hibernation (winter) diapause, a complex and defining life-history trait that allows temporal escape from harsh winters in temperate regions in many insects, including our model system, butterflies. Butterflies can diapause in all major life-stages and availability of global-scale phylogenies makes them an ideal model system for studying diapause evolution. First, using a thorough literature survey, we scored the developmental stage of hibernation diapause (egg, larva, pupa, adult) vs. absence of diapause. We find that larval diapause is most common while pupal, egg, and adult diapause are relatively rare. Next, we determined that the loss of diapause occurred at a much higher rate, and that gains primarily occurred from the non-diapause state. While ancestral state estimation at deeper nodes remained uncertain, we found consistent patterns for some families and strong evidence for extensive convergence in diapause evolution. Contrary to expectations, we find no support for increased gain of diapause during the Eocene-Oligocene glaciation (~35 MYA). Overall, the evolution of diapause in butterflies has a complex history, has evolved convergently, and has likely pre-dated the major glaciation event consistent with the deep history of diapause evolution in insects. This study advances our understanding of the evolution of a complex and important life-history trait and establishes a macroevolutionary base for future studies on the ultimate and proximate basis of diapause evolution.

Methods
The data comprises the developmental stage in which butterflies diapause (i.e. egg, larva, pupa, adult) or do not diapause, and was collected by literature search. The diapause classification was scored at the genus level. Next, using the phylogenetic comparative methods, mainly models for modelling the evolution of discrete traits or Mk models, a range of analyses were performed, for example, quantifying the rate of transition between states, ancestral state estimation, rate through time, etc. Please refer to the Methods section in the paper for more details. All analyses data analyses were performed in R.