Evolution of animal genitalia: patterns of phenotypic and genotypic variation and condition dependence of genital and non-genital morphology in water strider (Heteroptera: Gerridae: Insecta)

Abstract

Rapid and divergent evolution of male genitalia represents one of the most general evolutionary patterns in animals with internal fertilization, but the causes of this evolutionary trend are poorly understood. Several hypotheses have been proposed to account for genitalic evolution, most prominent of which are the lock-and-key, sexual selection and pleiotropy hypotheses. However, insights into the evolutionary mechanisms of genitalic evolution are hindered by a lack of relevant in-depth studies of genital morphology. We used a biparental progenies breeding design to study the effects of food stress during ontogeny on phenotypic expression of a suite of genital and non-genital morphological traits, both linear traits and multivariate shape indices, in a natural population of the water strider Gerris incognitus. In general, genitalic traits were as variable as non-genital traits, both phenotypically and genotypically. Average narrow-sense heritability of genital traits was 0·47 (SE=0·05). Further, while food stress during development had a large impact on adult morphology, and expression of genitalic traits exhibited significant levels of condition dependence, different genotypes did not significantly differ in their ability to cope with food stress. Genitalic conformation was also both phenotypically and genetically correlated with general morphological traits. These patterns are in disagreement with certain predictions generated by the long-standing lock-and-key hypothesis, but are in general agreement with several other hypotheses of genital evolution. We failed to find any additive genetic components in fluctuating asymmetry of any bilaterally symmetrical traits and the effects on fluctuating asymmetry of food stress during development were very low and insignificant. Some methodological implications of our study are discussed, such as the bias introduced by the non-negativity constraint in restricted maximum likelihood estimation of variance components.