Genetics of Adaptation of Island Rattlesnakes

The study of adaptive molecular evolution in natural populations has been severely limited by the difficulty of linking genetic variation to phenotypic variation to fitness effects. Most studies connecting genotype, phenotype, and fitness have used reverse genetic approaches to measure the functional effects of specific mutations in the laboratory because this relationship is difficult to measure in natural populations, particularly for complex traits because of the "many-to-one" mapping of genotype to phenotype. Many of the fundamental features of evolving systems, such as evolvability, epistasis, and pleiotropy, however, may be stronger determinants of evolutionary outcomes in natural populations than in the laboratory because artificial selection and breeding schemes are generally more simplistic relative to selection and demographic effects in natural settings. Snake venoms have emerged as a system for the study of the genetics of adaptation in complex, polygenic traits because of their genetic tractability and role in feeding, digestion, and defense, all of which are directly relevant to fitness. Because venom gene expression is tissue-specific (i.e., no pleiotropic constraints) and toxin protein abundance directly influences venom efficacy, venoms are not inherently biased toward a particular mutational pathway, enabling a systematic comparison of the molecular mechanisms underlying adaptive evolution. Venom phenotypes are manifest only upon injection into another animal, and venom functions are directly measurable through various assays, allowing direct tests of adaptive hypotheses in natural prey populations. In this work, we sought to create a genotype-phenotype-fitness map for the venom system of the eastern diamondback rattlesnake (Crotalus adamanteus) and, for the first time, identify the genetic basis of adaptation for a complex, polygenic trait in natural populations. Crotalus adamanteus is the largest species of rattlesnake and exclusively consumes endotherms. Crotalus adamanteus is historically native to seven states in the southeastern United States but has recently been extirpated from Louisiana, is endangered in North Carolina, and is currently under consideration for listing as threatened under the Endangered Species Act. In Chapter 1, we sequenced the venom-gland transcriptome and integrated mass spectrometry data to construct a transcriptome-proteome map for the venom system. We then used this map to identify significant toxin-gene expression differentiation across the range of C. adamanteus, providing candidate-genes for which to test the functional and evolutionary significance of the identified variation. In Chapter 2, we used a similar approach and identified significant ontogenetic differentiation in toxin gene expression; further analyses determined that ontogenetic effects explained more variation in toxin expression than geographic effects, although both juvenile and adult expression patterns varied geographically, and time-series experiments in lab-raised individuals demonstrated that geographic and ontogenetic expression differentiation were not environmentally induced but rather under genetic control. In Chapter 3, we used in vitro functional assays to verify that the expression differences found in the previous two chapters corresponded to differences in venom function. We found that, overall, the statistical differences in toxin expression outlined in the first two chapters equated to functional differences in toxic activities in a predictable, tractable manner, suggesting that the differences identified in the first two chapters were, in fact, biologically relevant. In Chapter 4, we used a target-enrichment approach to sequence the exons of all identified toxins in the venom-gland transcriptome as well as several thousand neutral loci to ascertain the relative roles of expression versus coding-sequence changes in a trait not inherently biased towards either mutational pathway. We found evidence for adaptive changes at both the expression and sequence levels across the entire range, although expression differentiation did appear to be the more frequent molecular mechanism. But, without functional characterizations of the identified sequence and expression evolution, it was difficult to characterize the relative roles demography, selection, and drift played in generating the identified sequence and expression divergence. Although Chapter 3 did link expression variation to functional variation, these assays were not conducted in the actual target of venoms, natural prey. To address these issues, we examined toxin sequence and expression evolution and estimated venom toxicity (i.e., fitness) in sympatric and allopatric natural prey across an island-mainland population pair in Chapter 5 to, for the first time, construct a genotype-phenotype-fitness map for a complex trait in natural populations. We found that expression differentiation was predominantly, or exclusively, the genetic basis of polygenic adaptation, suggesting that over ecological timescales complex traits may preferentially evolve through mutations affecting expression because more molecular mechanisms exist for altering the amount of protein produced than for altering their functions through their primary sequences. In Chapter 1, we found significant expression differentiation in both high- and low-abundance proteins across the range and over 1 million years of divergence, and in Chapter 4, we found both sequence and expression differentiation across the same temporal and spatial scales. In Chapter 5, however, we only identified expression differentiation, and found that this expression differentiation was restricted to low-expression proteins because of physiological and selective constraints on high-expression proteins. These differences in the molecular mechanism underlying adaptive evolution were most likely the result of temporal constraints on generating beneficial variation; because more molecular mechanisms exist for altering protein amounts than protein function, the probability of generating a beneficial expression variant is greater than the probability of generating a beneficial point mutation in the coding-region of a specific protein, and these differences in probability would be most pronounced over extremely short timescales. Given enough time, however, both mutational pathways and proteins expressed at all levels can generate beneficial variation, and these results provide qualitative predictions regarding the process of adaptation for a complex trait.