A pangenome framework uncovers the role of deletions in repeated evolution of cave-derived traits

Structural variants (SVs) are increasingly recognized as key contributors to adaptive evolution, yet they remain underexplored compared to single-nucleotide variation. To understand how large-scale genomic changes shape repeated evolution, we leveraged multiple levels of sequence data across the powerful evolutionary model system of the Mexican tetra fish (Astyanax mexicanus). Cavefish populations have repeatedly evolved similar and dramatic phenotypes compared to surface fish such as eye loss, pigmentation loss, and sleep loss, suggesting a potential role for loss-of-function variants in cave adaptation. We constructed one of the first pangenome graphs from a naturally evolving vertebrate, enabling comprehensive discovery of SVs among 120 fish from 11 populations. We identified over 2,400 high-confidence cave-specific deletions that are enriched in biological pathways involved in vision, metabolism, and behavior and cluster non-randomly in quantitative trait loci linked to cavefish traits. Strikingly, 67 genes harbor unique deletions between independent cavefish lineages. These reused genes show strong evidence of population-specific selection (99% are under selection, compared to 8-15% in genes lacking SVs), indicating that deletions rose in frequency through repeated positive selection rather than drift. Together, these results reveal that recurrent deletion events have repeatedly contributed to the evolution of cave-adapted phenotypes and highlight deletions as underexplored contributors of adaptive evolution in a system characterized by trait loss.