The role of natural selection in the evolution of trait complexity can be characterized by testing hypothesized links between complex forms and their functions across species. Predatory venoms are traits composed of multiple proteins that collectively function to incapacitate prey. Venom complexity fluctuates considerably over evolutionary timescales, with apparent increases and decreases in complexity, yet the evolutionary causes of this variation is unclear. Here, we tested alternative hypotheses for the link between venom complexity and ecological sources of selection related to diet in the largest clade of front-fanged venomous snakes in North America: the rattlesnakes, copperheads, cantils, and cottonmouths Crotalus, Sistrurus and Agkistrodon. We generated independent transcriptomic and proteomic measures of venom complexity and then estimated prey diversity using the past century's extensive natural history studies on these snakes. We then conducted comparative tests relating different measures of predator venom complexity and prey community diversity using the first, genome-scale, dated phylogenies for this clade of snakes. Strikingly, phylogenetic diversity of prey was more strongly correlated to venom diversity than was species diversity, implicating prey species divergence, rather than the number of lineages alone, in the evolution of venom complexity. This positive relationship was observed within three of the four largest toxin gene families in viper venom. Given documented examples of taxonomic specificity of venoms spanning several levels of divergence, we suggest that the phylogenetic diversity of prey measures functionally-relevant divergence in the molecular targets of venom, a claim supported by diversity in the coagulation cascade targets of snake venom serine proteases. Our results support the general concept that the evolved diversity of species in an ecological community is more important than their overall number in determining evolutionary patterns in predator trait complexity.

Venom-gland transcriptomes were subject to Trinity assembly and BUSCO v.3 search for orthologous genes. Alignments were produced using MAFFT and trimmed with trim-al. See manuscript for additional details.