Date:
Location:
Title: Quantification of fish locomotor performance to understand adaptive evolution
Abstract: Adaptive evolution is often understood as a connection from genotype to phenotype to performance to fitness. While each of these elements represent valuable fields of study when considered individually, it is only by combining all of them that we can connect genotype to fitness, and therefore mechanistically connect evolution to natural selection. This integrative approach has led to many insights into the dynamics of evolution, but so far has only been applied to functionally simple systems. One major reason for this is that there has historically been a disconnect between the traits whose genetic basis has been elucidated and those that are involved in complex functions. I posit that this disconnect exists in part because functionally relevant traits tend to be genetically quantitative, and therefore measuring the functional relevance of individual genes or developmental pathways requires high precision measurements. Since most functional studies are conducted on live animals with small sample sizes, it is often difficult to obtain such high precision measures of organismal performance. However, with recent advances in robotics and statistical analysis this barrier is quickly falling.
In this dissertation I use studies of fish locomotion to show how modern advances in functional biology can facilitate the study of adaptive evolution. In the introduction I explore the current state of research linking genotype to phenotype to performance to fitness and assert that new methodology is needed in order to facilitate such studies in functionally complex systems. In particular, I argue that biological robotics, structural equation modeling, and simultaneous multi-modal data collection methods carry the most promise. I then give examples of how each of these methods can be applied to make functionally complex traits evolutionarily tractable. In chapter 1 I give a more comprehensive example of how biological robotics can be used to quantify fine-scale performance variation. I combine a simple flapping robotic system with multivariate modeling to isolate the effects of fin position and relative fin motion on several locomotor performance metrics in a biomimetic model. I find that both the position and timing of the dorsal fin relative to the caudal fin impact swimming performance, with as much as a 35% increase in swimming speed possible if the two fins are optimally aligned. In chapter 2 I extend this approach to show how computational fluid dynamics can be combined with biological robotics to further elucidate performance variation. Here I use tuna inspired models to examine the role of the caudal peduncle, the structure connecting a fish’s body to its tail fin, to ask whether actively controlled traits could be used to alter swimming performance. I find that although it is possible to change the timing of the tail relative to the body, there is a tradeoff between thrust production and power consumption as you increase the phase lag of the tail. When I examine the effect of varied stiffness around this optimal tradeoff point, I find that there was no one set of parameters that outperformed other configurations. This highlights the necessity of active control to tailor the exact motion of the tail to the behavior at hand. Finally, in chapter 3 I use structural equation modeling to measure the effect of a fibrosis immune response on the escape swimming performance of threespine stickleback (Gasterosteus aculeatus). By measuring fibrosis levels, body stiffness, body kinematics, and escape performance I am able to build a hierarchical structural model in which each of these metrics is accounted for. I can then estimate the total effect of fibrosis on escape performance by taking the product of successive correlations in the path model. Through this method I find that in addition to reducing parasite load, fibrosis is associated with increased swimming performance during the linear acceleration phase of an escape. I also compare this result to those obtained with classical multivariate statistics to demonstrate that these results could not be obtained without the use of structural equation modeling. Together, these chapters demonstrate how modern methods can be used to increase the resolution of performance measurements, allowing us to better characterize functional variation as it relates to evolutionary outcomes.
Committee: George Lauder (Advisor), Andy Biewener, Stephanie Pierce