Serine hydroxymethyltransferase (SHMT) is a member of the fold type I family of vitamin B6-dependent enzymes, a group of evolutionarily related proteins that share the same overall fold. The reaction catalysed by SHMT, the transfer of Cbeta of serine to tetrahydropteroylglutamate (H4PteGlu), represents in the cell an important link between the breakdown of amino acids and the metabolism of folates. In the absence of H4PteGlu and when presented with appropriate substrate analogues, SHMT shows a broad range of reaction specificity, being able to catalyse at appreciable rates retroaldol cleavage, racemase, aminotransferase and decarboxylase reactions. This apparent lack of specificity is probably a consequence of the particular catalytic apparatus evolved by SHMT. An interesting question is whether other fold type I members that normally catalyse the reactions which for SHMT could be considered as 'forced errors', may be close relatives of this enzyme and have a catalytic apparatus with the same basic features. As shown in this study, l-threonine aldolase from Escherichia coli is able to catalyse the same range of reactions catalysed by SHMT, with the exception of the serine hydroxymethyltransferase reaction. This observation strongly suggests that SHMT and l-threonine aldolase are closely related enzymes specialized for different functions. An evolutionary analysis of the fold type I enzymes revealed that SHMT and l-threonine aldolase may actually belong to a subgroup of closely related proteins; fungal alanine racemase, an extremely close relative of l-threonine aldolase, also appears to be a member of the same subgroup. The construction of three-dimensional homology models of l-threonine aldolase from E. coli and alanine racemase from Cochliobolus carbonum, and their comparison with the SHMT crystal structure, indicated how the tetrahydrofolate binding site might have evolved and offered a starting point for further investigations.