Several proteases, including the bacterial serine protease subtilisins, require the assistance of the N-terminal pro-sequence of precursors to produce active, mature enzymes. Upon completion of folding, the pro-sequence is autocatalytically degraded because it is not necessary for the activity or stability of folded, mature cognates of the original enzymes. Therefore, the pro-sequence functions as an intramolecular chaperone that guides correct folding of the protease domain. Interestingly, Shinde et al. proposed a new theory of "protein memory" in which an identical polypeptide can fold into an altered conformation with different secondary structure, stability and specificities through a mutated pro-sequence [Shinde et al. (1997) Nature 389:520-522]. We also showed that the autoprocessing efficiency was improved by modifications in the pro-sequence of mutant subtilisins with altered substrate specificity. Further, the pro-sequence from a subtilisin homologue was found to chaperone the intramolecular folding of denatured subtilisin. These results indicate that engineering of the pro-sequence, i.e., site-directed and/or random mutagenesis, chimeras and gene shuffling between members of the family, would be a useful method for improving the functions of autoprocessing proteases. Conventional protein engineering techniques have thus far employed mutagenesis in the protease domain to modify the enzymatic properties. This new approach, which we term "pro-sequence engineering", is not only an important tool for studying the mechanism of protein folding, but also a promising technology for creating unique proteases with various beneficial properties.