Understanding Tradeoff Functions for Quantitative Traits

Evolutionary models often involve tradeoffs between a beneficial and detrimental trait. In models of continuous traits, the modeler must decide precisely how costs scale with benefits. For instance, in many adaptive dynamics models, accelerating costs (diminishing returns) lead to monomorphic populations, while decelerating costs lead to genetic polymorphism. Decelerating cost functions are therefore a potential driver of genetic diversity in natural systems. This finding was particularly exciting for researchers in genetic resistance, where natural populations often harbor far more genetic variation than traditional models would predict. Unfortunately, measuring evolutionary tradeoffs in vivo is difficult, and our current theory has little to say on which tradeoff functions might be the most biologically plausible. To test how underlying genetic assumption can determine the shape of tradeoff curves, I developed a multi-locus model of quantitative resistance, where individual loci have random pleiotropic effects on resistance and fecundity5. With this model, I generated genotype distributions, and investigated how these distributions could manifest tradeoff functions via their Pareto fronts (the set of genotypes where improvement in one trait can only be accomplished at the detriment of another trait). I then tested whether different forms of epistasis can change the shape of the Pareto front. I consistently found that my model produced accelerating costs, indicating that accelerating costs may be a strong null model for evolutionary tradeoffs. However, when I incorporated my genotype model into an eco-evolutionary model of disease resistance evolution, I found genetic polymorphism in scenarios where traditional adaptive dynamics models would predict a single dominant genotype. By fitting continuous cost curves to our simulated Pareto fronts, I was able to determine that jagged tradeoff functions with locally decelerating regions can generate polymorphisms that equivalent smooth functions do not.

Modeling Population-Level Spillover Risks

Emerging infectious diseases are often caused by cross-species transmission events, or spillovers. While most work on spillover risks has focused on exogenous factors such as land-cover change and phylogenetic distance between ancestral and naive hosts, it is becoming increasingly clear that many species exhibit heritable variation in susceptibility to foreign pathogens for which they share no coevolutionary history. Our current theory struggles to explain how this variation is maintained. To address this, I developed mathematical models to ask whether evolution in response to endemic pathogens can lead to genetic variation in host susceptibility to foreign pathogens. I hypothesized that the presence of multiple resistance pathways, and their differential selection from endemic pathogen pressure might lead to variation in susceptibility to foreign pathogens. In many systems, hosts have resistance pathways targeted towards certain pathogens (specific resistance) and some of which are more broadly effective (general resistance). I used this framework as the basis for my first two papers in the Bruns lab, investigating how eco-evolutionary feedbacks between multiple resistance pathways can recapitulate patterns seen in natural populations. In my first paper from the Bruns Lab, I used adaptive dynamic framework to test whether specific resistance can suppress the evolution of general resistance, finding that without multiple pathogens, specific resistance can indeed suppress general resistance3. When a foreign pathogen is introduced into the model, a subset of the host population maintains general resistance, while others maintain specific resistance, but never both simultaneously. In a follow up paper, I found that when endemic pathogens can coevolve with their hosts, hosts maintain general resistance at a higher frequency, and coevolution increases the genetic diversity of host populations4. I also found that the genetic linkage between the loci underlying general and specific resistance plays an important role in the structure of host resistance. If these loci become tightly linked, a population can have a higher spillover risk, as resistance to foreign and endemic pathogens can become negatively correlated. This leads to a subset of the host population being more susceptible to foreign pathogens than endemic, creating an ecological niche for foreign pathogens to invade and persist.

Deterministic Evolution of Signal Design in Complex Visual Patterns

The primary goal of my doctoral work was to extend the sensory drive and sensory bias hypotheses to explain diversity in complex visual ornaments, using darters as a model system. Darters are a diverse group of North American freshwater fish which have undergone a dramatic radiation over the last 20 million years. Notably, recently diverged darter species often exhibit remarkably different visual patterns. Previous work in this system has demonstrated that females prefer not only conspecific coloration, but also achromatic aspects of patterning found in conspecific male ornaments. To investigate sexual selection for visual patterning, I hypothesized that attractive visual stimuli mimic the visual statistics of natural visual stimuli. If different species of darters occupy visually distinct habitat types, the way in which their brains process visual signals might reflect statistical properties of their environment. These biases could then lead to preferences for male patterning. To test this prediction, I used mathematical image analysis, machine learning techniques, and behavioral assays. I captured high resolution images of 11 species of darters, as well as underwater photographs of their habitats. Using Fourier image analysis, I found a significant correlation between the visual statistics of male darter patterns, and those of their habitats1. I followed these analyses up with convolutional neural network analyses2, and behavioral assays. This body of work suggests that visual preferences originating from heterogeneous visual environments may be a predictable generator of phenotypic diversity in a trait that has been historically difficult to study.