Inferring the dynamics of selective constraints across complex ontogenies

James G DuBose

Most organisms undergo a series of complex phenotypic changes throughout their life cycles that allow them to meet the demands of different niches throughout ontogeny. Theory suggests the significant coordination required to undergo such ontogenetic transitions can impose evolutionary constraints on variation to developmental programs. This produces patterns known as developmental hourglasses, which are periods during development where phenotypic and/or genetic variation is constrained between taxa relative to other points throughout ontogeny. Although empirical support for developmental hourglasses has been well established in animal embryonic development, it remains unclear if other non-animal or non-embryonic development programs exhibit such patterns. Furthermore, more recent investigations of developmental hourglasses have largely relied on relating gene age and sequence divergence to their temporal expression profiles across development, an approach highly susceptible to noise due to historical contingency and developmental system drift. Likewise, more recent investigations have described more complex fluctuations in the strength of selective constraints across ontogeny, suggesting our understanding may be improved by more nuanced and flexible approaches for quantifying developmental constraints across ontogeny. To this end, here I present a theoretical and empirical framework for leveraging population-level variation in developmental gene/trait expression dynamics to infer the strength of selective constraints across an ontogeny. I first provide theoretical precedence for this approach using an extended geometric model, which suggests that patterns of variation in gene/trait expression within populations are stable and recapitulate the latent ontogenetic selective dynamics. I then describe an empirical approach for inferring these latent ontogenetic selective dynamics and use a simple simulation to illustrate its utility. Finally, I utilize this approach to infer the dynamics of selective constraints from population-level transcriptional data of various stages across the monarch butterfly Danaus plexippus metamorphosis, which suggests both the transitions from larva to pupa and from pupa to adult constitute developmental hourglasses.