The genic capture hypothesis, where sexually selected traits capture genetic variation in condition and the condition reflects genome-wide mutation load, stands to explain the presence of abundant genetic variation underlying sexually selected traits. Here we test this hypothesis by applying bidirectional selection to male mating success for 14 generations in replicate populations of Drosophila melanogaster. We then resequenced the genomes of flies from each population. Consistent with the central predictions of the genic capture hypothesis, we show that genetic variance decreased with success selection and increased with failure selection, providing evidence for purifying sexual selection. This pattern was distributed across the genome and no consistent molecular pathways were associated with divergence, consistent with condition being the target of selection. Together, our results provide molecular evidence suggesting that strong sexual selection erodes genetic variation, and that genome-wide mutation-selection balance contributes to its maintenance.