Postcopulatory sexual selection results from variation in competitive fertilization success among males, and comprises powerful evolutionary forces that operate after the onset of mating [1, 2]. Theoretical advances in the field of sexual selection addressing the build-up and co-evolutionary consequences of genetic coupling [3-5], motivate the hypothesis that indirect postcopulatory sexual selection may promote evolution of male secondary sexual traits—those traits traditionally ascribed to mate choice and male fighting [6, 7]. A crucial prediction of this hypothesis is genetic covariance between trait expression and competitive fertilization success, which has been predicted to arise, for example, when traits subject to pre- and postcopulatory sexual selection are under positive correlational selection [8]. We imposed bidirectional artificial selection on male ornament (sex comb) size in Drosophila bipectinata, and demonstrate increased competitive fertilization success as a correlated evolutionary response to increasing ornament size. Transcriptional analyses revealed that levels of specific seminal fluid proteins repeatedly shifted in response to this selection, suggesting that properties of the ejaculate rather than the enlarged sex comb itself contributed fertilizing capacity. We used ultraprecise laser surgery to reduce ornament size of high line males, and found that their fertilizing superiority persisted despite the size reduction, reinforcing the transcriptional results. The data support the existence of positive genetic covariance between a male secondary sexual trait and competitive fertilization success, and suggest the possibility that indirect postcopulatory sexual selection may under certain conditions magnify net selection on ornamental trait expression.  

Raw data from two main experiments are provided, measuring responses in competitive fertilization success. The response in both cases is the proportion offspring of a given doubly-mated female fathered by the second male to mate. For the competitive fertilization data set, we used a REML model contained the following terms: selection treatment (high, low and control, fixed effect), replicate line (1, 2 and 3, treated as a random effect nested within selection treatment), and six mean-centered covariates. For the phenotypic engineering data set, we modeled fertilization success using a generalized linear model with a binomial error structure and logit link function, where the number of fertilized eggs laid by each female after her second mating was the response and the total number of eggs laid the binomial denominator. The factor, in this case, was treatment, and there were 5 mean-centered covariates.