Changing genetic covariance
In contrast to genetic variance, we observed a clear trend in changing genetic covariances. Specifically, the correlation between Z11-16:OH and Z11-16:OAc changed in both sign and magnitude. Since both Z11-16:OH and Z11-16:OAc reduce heterospecific mate attraction (Vetter and Baker, 1983; Vickers and Baker, 1997), and both traits covary with fitness and are condition-dependent (Blankers et al., 2021), the negative genetic correlation in the starting population could be the result of a trade-off that is selected against when artificial selection favors higher rates of acetates.
Tracking changes in G matrices through time, we observed another interesting phenomenon: after selection, the genetic variance in Z11-16:OAc was more strongly associated with genetic (g) PC2 compared to before selection, while genetic variance described by gPC1 increased relative to the other gPCs. Thus, more of the genetic variance was allocated in fewer dimensions of genetic variation, while the component under selection was uncoupled from the major dimension of genetic variance. We acknowledge that the observed changes are subtle and that they cannot be unequivocally attributed to selection. However, we observed similar patterns of G matrix evolution in the high and low line and these patterns fit the prediction that genetic variances and covariances are reoriented to align the phenotype with the dominant direction of selection (Melo and Marroig, 2014). These changes in the G matrix can likely facilitate a more specific response where only Z11-16:OAc evolves in the absence of indirect selection responses: relative to the selection regime, the modularity (i.e. the genetic independence of trait modules) of the pheromone increased.