Unlike mammals, fish are able to regenerate axons in their central nervous system. This difference has been partly attributed to the loss/acquisition of inhibitory proteins during evolution. Nogo-A - the longest isoform of the reticulon4 (rtn4) gene product - is commonly found in mammalian myelin where it acts as a potent inhibitor of axonal regeneration. Interestingly, fish RTN4 isoforms were previously reported to lack the most inhibitory Nogo-A-specific region (NSR). Nevertheless, fish axons collapse on contact with mammalian NSR, suggesting that fish possess a functional Nogo-A receptor but not its ligand. To reconcile these findings, we revisited the early evolution of rtn4. Mining of current genome databases established the unequivocal presence of NSR-coding sequences in fish rtn4 paralogues. Further comparative analyses indicate that the common ancestor of fish and tetrapods had an NSR-coding rtn4 gene, which underwent duplication and divergent evolution in bony fish. Our genomic survey also revealed that the cephalochordate Branchiostoma floridae contains a single rtn gene lacking the NSR. Hence, Nogo-A most probably arose independently in the rtn4 gene of a gnathostome ancestor before the split of the fish and tetrapod lineages. Close examination of the NSR uncovered clusters of structural and sequential similarities with neurocan (NCAN), an inhibitory proteoglycan of the glial scar. Notably, the shared presence of transposable elements in ncan and rtn4 genes suggests that Nogo-A originated via insertion of an ncan-like sequence into the rtn4 gene of an early jawed vertebrate with myelinated axons. © 2010 The Author.