Abstract
Antagonistic coevolution occurs across a heterogeneous landscape of reciprocal selection, in which species interactions and their fitness consequences vary from one location to the next. While reciprocal selection drives local adaptation, broader landscape patterns of coevolution arise in the often-overlooked context of geographically structured populations, connected by gene flow and ancestry. I investigated how the genetic structure of populations, and their evolutionary histories, contribute to the geographic mosaic of arms race coevolution between common garter snakes (<i>Thamnophis sirtalis</i>) and their toxic prey, rough-skinned newts (<i>Taricha granulosa</i>). Garter snakes in western North America evolved resistance to the deadly tetrodotoxin (TTX) in newts as a result of mutations to the skeletal muscle sodium channel (Na<sub>V</sub>1.4) that disrupt toxin-binding. To characterize the evolutionary history of predator populations, I phylogenetically reconstructed the historical order of mutations to Na<sub>V</sub>1.4 in two geographically-distinct hotspots of coevolution. Both lineages of <i>Th. sirtalis</i> convergently evolved resistance by passing through a repeated first-step mutation to the channel pore, suggesting the initial change had permissive effects on the subsequent evolution of resistance. I investigated how constraints might bias the evolution of Na<sub>V</sub>1.4 towards repeatable outcomes, and I found that negative trade-offs arise once TTX-resistant mutations have accumulated, disrupting channel excitability and muscle performance. The evolutionary trajectory of Na<sub>V</sub>1.4 strikes a balance between TTX resistance and the maintenance of channel function, and these pleiotropic effects seem to underlie variation in resistance across the landscape. Finally, I tested how geographic population structure also contributes to phenotypic divergence in the coevolutionary mosaic. I first showed that allopatric populations of <i>Ta. granulosa</i> have low, but unexpectedly variable levels of TTX, indicating that factors other than selection imposed by <i>Th. sirtalis</i> contribute to toxicity. In sympatry, phenotypic divergence in TTX toxicity is tightly correlated with population genetic structure, implying that neutral processes like drift and gene flow—not reciprocal selection—determine variation in toxicity. In contrast, TTX resistance of <i>Th. sirtalis</i> deviates from neutral expectations and tracks prey toxicity, such that mosaic variation in resistance and toxicity are both predicted by the population structure of <i>Ta. granulosa</i>. This research highlights how two coevolving species are unlikely to undergo a symmetrical response to reciprocal selection. Asymmetries arise at the level of populations, for example, due to drift or contingency. I demonstrate that unique population-level processes occur in both predator and prey, and their combinatory effects explain mosaic variation in species interactions across the landscape, limiting the fixation of species-level traits.
