Results
To assess the potential for sexual selection to act on pathogen resistance, we first compared the paternity success of infected and sham-treated males in competition with males from a reference strain. We found that infected males had lower competitive mating success than uninfected males, as evidenced by a lower proportion of offspring sired by the focal males (treatment effect: χ2df=1 = 4.45; p = 0.03;Figure 1, Table 1 ). Infected males sired on average 59.2% of progeny in competition with the competitive standard, while uninfected males sired on average 68.5% of progeny. This result indicated that infection harms male mating success and suggested that genetic variation contributing to infection resistance might be favored by sexual selection.
We also verified that the infected males had cleared the pathogen from their gut by the time they were placed with females. Although males harbored many live P. entomophila 4h after the onset of the infection treatment, no live bacteria were detected at 8 or 24 h (Supplemental Figure S1), in agreement with earlier results (Bou Sleimanet al. , 2015). Thus, there was little opportunity for the males to transmit the infection to the females.
We next evolved replicate populations with and without both sexual selection and pathogen for fourteen generations. Over the course of experimental evolution, P. entomophila virulence varied; the pathogen reliably killed a substantial fraction of the relishmutants (mean survival post infection 43% ± 10.7 (s.e.) inrelish mutants, Supplemental Figure S2 ). Survival at 24 hours was lower in experimental populations exposed to the pathogen (+P), averaging 92.4%, than it was in populations not exposed to the pathogen (-P), in which survival was 99.9%.
To compare resistance to P. entomophila in the populations subject to the different regimes, we measured their survival following infection after fourteen generations of experimental evolution and one generation of common garden rearing. In general, females survived less well after infection than males (Figure 2 ). Populations evolved under pathogen pressure (+P evolutionary regimes) showed better post-infection survival than populations evolved without pathogen exposure (-P evolutionary regimes) (pathogen selection effect: χ2df=1 = 8.89; p = 0.002;Figure 2, Table 1 ). A significant three-way interaction between sexual selection, pathogen, and sex (SS*Pathogen*Sex, χ2df=1 = 5.91; p = 0.01) indicates a difference between males and females in how the interaction between sexual selection and pathogen presence affects post-infection survival, which we further explored in sex-specific analyses.
In females, the sex-specific analyses showed that post-infection survival under +P regimes was better than that in the –P regimes (Figure 2A , pathogen selection effect: χ2df=1 = 4.92; p = 0.026), but we detected no effect of sexual selection (χ2df=1 = 0.04; p = 0.82) or any interaction between sexual selection and pathogen (χ2df=1 = 0.93; p = 0.33). In males there was neither a significant effect of sexual selection (Figure 2B , χ2df=1 = 0.094; p = 0.75) nor pathogen (χ2df=1 = 3.40; p = 0.06). However, there was a significant interaction between sexual selection treatment and pathogen presence (χ2df=1 = 4.71; p = 0.029). For the good genes hypothesis to be true in our case, the +SS +P populations should have elevated survivorship compared to -SS +P regimes. However, in our study we see the opposite effect, with the -SS +P regimes surviving significantly better than +SS +P (Figure 2B , Tukey’s post hoc comparison p = 0.02). At the same time, there is no difference between +SS populations evolved with and without pathogen. This difference in the effect of sexual selection that depended on whether pathogen was present or not during the course of experimental evolution is what drives the significant interaction between sexual selection and pathogen.