Abstract
A number of genes encoding bacterial glycosyltransferases have been sequenced during the last few years, but their low sequence similarity has prevented a straightforward grouping of these enzymes into families. The sequences of several bacterial α-mannosyltransferases have been compared using current alignment algorithms as well as hydrophobic cluster analysis (HCA). These sequences show a similarity which is significant but too low to be reliably aligned using automatic alignment methods. However, a region spanning approx. 270 residues in these proteins could be aligned by HCA, and several invariant amino acid residues were identified. These features were also found in several other glycosyltransferases, as well as in proteins of unknown function present in sequence databases. This similarity most probably reflects the existence of a family of proteins with conserved structural and mechanistic features. It is argued that the present IUBMB classification of glycosyltransferases could be complemented by a classification of these enzymes based on sequence similarities analogous to that which we proposed for glycosyl hydrolases [Henrissat, B. (1991) Biochem. J. 280, 309–316].