Arginine kinase from the sea cucumber Stichopus japonicus underwent a unique molecular evolution. Unlike the monomeric 40kDa arginine kinases from molluscs and arthropods, Stichopus arginine kinase is dimeric, the same as cytoplasmic isoenzymes of the vertebrate creatine kinases. Its entire amino acid sequence is more similar to creatine kinases than to other arginine kinases, but the guanidino specificity region (GS region) is of the arginine kinase type. To elucidate its unusual evolution, the structure of the Stichopus arginine kinase gene was determined. It consisted of seven exons and six introns, and a part of the exon 2 of the Stichopus gene corresponds to the GS region. Compared with the structure of the human muscle creatine kinase gene (seven exons, six introns), the splice junctions of five introns were conserved exactly between the two genes, suggesting that these introns had been conserved for at least 500 million years. The entire sequence of Stichopus arginine kinase is distinctly included in the creatine kinase cluster in all tree construction methods examined. On the other hand, if the tree is constructed only from sequences corresponding to Stichopus exon 2, it is placed in the arginine kinase cluster. Thus we conclude that Stichopus arginine kinase evolved not from the arginine kinase gene but from the creatine kinase gene, and suggest that its GS region, determining substrate specificity, has been replaced by an arginine kinase type via exon shuffling. In typical arginine kinases four residues, Ser63, Gly64, Val65 and Tyr68 (numbering from the Limulus polyphemus sequence), in the GS region are highly conserved and are associated with substrate binding. Among them, Tyr68 appears to play a crucial role by forming a hydrogen bond with the substrate, and is conserved exactly in all arginine kinases. However, in Stichopus arginine kinase, none of these four conserved residues were present. Nevertheless, the enzyme displays an affinity for the substrate arginine (Km = 0.8mM) comparable with other arginine kinases. This implies that a completely different substrate-binding system has been developed in Stichopus arginine kinase. We propose that the His64 in Stichopus arginine kinase acts as a substitute for the Tyr68 in other arginine kinases, and that the imidazole ring of His64 is hydrogen bonded with the substrate arginine, thus stabilizing it.