ARTIFICIAL SELECTION DIVERSIFIES BIRDSONG THROUGH POLYGENIC CHANGES ACROSS SENSORY AND MOTOR SYSTEMS
Vocal learning, or the ability to acquire and produce contextualized sounds through social interaction, underlies human speech yet has evolved in a few animal lineages, including songbirds. In songbirds, learned song is produced by a dedicated neural circuit and depends on auditory feedback for its acquisition and maintenance. Roughly four centuries of artificial selection on the domestic canary (Serinus canaria) have produced song breeds whose vocalizations diverge as much as those of distinct species, offering a rare intraspecific system in which to identify the genetic and neural changes underlying vocal divergence. Here we show that selection for song reshaped the canary song control system. Compared with breeds selected for plumage or body form, song breeds show increased neural progenitor proliferation and breed-specific changes in neuronal recruitment, sensorimotor brain region volume, and auditory sensitivity. Genome-wide analyses reveal a polygenic architecture underlying song divergence, enriched for transcriptional regulation and synaptic transmission genes. Among these, we find a coding variant in ADGRV1, the gene mutated in human Usher syndrome type 2C, associated with low-frequency singing and elevated auditory thresholds in canary. Disrupting ADGRV1 in zebrafish alters hair cell dynamics, linking adgrv1 mutation to a peripheral sensory mechanism. These findings demonstrate that learned vocal behavior can diversify rapidly through changes spanning sensory, neurogenic, and motor systems, with hearing itself as a substrate for selection.