Automated behavioral profiling using neural networks reveals differences in stress-like behavior between cave and surface-dwelling Astyanax mexicanus

Abstract

Behavioral stress responses allow animals to quickly adapt to local environments and are critical for survival. Stress responses provide an ideal model for investigating the evolution of complex behaviors due to their conservation across species, critical role in survival, and integration of behavioral and physiological components. The Mexican cavefish (Astyanax mexicanus) has evolved dramatically different stress responses compared to river-dwelling surface fish morphs, providing a model to investigate the neural and evolutionary basis of stress-like responses. Surface morphs inhabit predator-rich environments whereas cave-dwelling morphs occupy predator-free habitats. While these key ecological variables may underlie differences in stress responses, the complexity of the behavioral differences has not been thoroughly examined. By leveraging automated pose-tracking and machine learning tools, we quantified a range of behaviors associated with stress, including freezing, bottom-dwelling, and hyperactivity, during a novel tank assay. Surface fish exhibited heightened stress responses characterized by prolonged bottom-dwelling and frequent freezing, while cavefish demonstrated reduced stress behaviors, marked by greater exploration and minimal freezing. Analysis of F2 hybrids revealed that a subset of behaviors, freezing and bottom-dwelling, co-segregated, suggesting shared genetic or physiological underpinnings. Our findings illustrate the power of computational tools for high-throughput behavioral phenotyping, enabling precise quantification of complex traits and revealing the genetic and ecological factors driving their evolution. This study provides a framework for understanding how integrated behavioral and physiological traits evolve, offering broader insights into the mechanisms underlying the diversification of animal behavior in natural systems.