Condition-dependence of the sexually dimorphic transcriptome in Drosophila melanogaster

Sexually dimorphic traits are by definition exaggerated in one sex, which may arise from a history of sex-specific selection – in males, females, or both. If this exaggeration comes at a cost, exaggeration is expected to be greater in higher condition individuals (condition-dependent). Although studies using small numbers of morphological traits are generally supportive, this prediction has not been examined at a larger scale. We test this prediction across the trancriptome by determining the condition-dependence of sex-biased (dimorphic) gene expression. We find that high-condition populations are more sexually dimorphic in transcription than low-condition populations. High condition populations have more male-biased genes and more female-biased genes, and a greater degree of sexually dimorphic expression in these genes. Also, condition-dependence in male-biased genes was greater than in a set of unbiased genes. Interestingly, male-biased genes expressed in the testes were not more condition-dependent than those in the soma. By contrast, increased female-biased expression under high condition may be have occurred because of the greater contribution of the ovary-specific transcripts to the entire mRNA pool. We did not find any genomic signatures distinguishing the condition-dependent sex-biased genes. The degree of condition-dependent sexual dimorphism (CDSD) did not differ between the autosomes and the X-chromosome. There was only weak evidence that rates of evolution correlated with CDSD. We suggest that the sensitivity of both female-biased genes and male-biased genes to condition may be akin to the overall heightened sensitivity to condition that life-history and sexually selected traits tend to exhibit. Our results demonstrate that through condition-dependence, early life experience has dramatic effects on sexual dimorphism in the adult transcriptome. There were 8 biologically distinct samples. Each was replicated 6 times for a total of 48 biological samples on 24 arrays. There was no reference or control sample as a loop design was used. Each of the 48 samples are represented separately.