Evolution and mechanism of MEIS2-mediated forelimb specialization in bats
The genomic basis of limb adaptations in tetrapods is thought to be largely driven by changes in gene regulation. However, the mechanisms by which regulatory programs evolve are not well understood. In bats, wing membrane development has been shown to be associated with expression of the transcription factor MEIS2 in the interdigital tissue of the forelimb. However, MEIS2 alone is insufficient to recapitulate wing morphology, suggesting that its regulatory context has also undergone divergence. Here, we integrate functional genomics with sequence-to-function deep learning to dissect both the mechanistic and evolutionary roles of MEIS2 in bat forelimb development. Using models trained on embryonic limb data from bat and mouse, we identify a strong association between MEIS2 binding and the transcription factor TWIST1, a finding which is supported by single-cell transcriptomic analyses. To investigate the evolutionary dimension, we applied these models across more than 100 genomes, including extant bats, closely related species, and reconstructed ancestors. This analysis identified divergences in regulatory regions, which likely contribute to bat-specific forelimb expression of genes that lead to wing morphogenesis. Notably, these changes are prominent in the regulatory domains of the MEIS dimerization partner PBX1, indicating coordinated regulatory evolution. Together, our results demonstrate that the evolution of a complex morphological trait involves coordinated changes in both trans-regulatory environments and cis-regulatory landscapes. More broadly, this study provides a framework for integrating deep learning with comparative and functional genomics to investigate regulatory evolution.