Amadori compounds are ubiquitous in vivo as well as in food and have been implicated in diabetic complications and aging. In recent years, fructosyl amine oxidases (FAOXs) which cleave Amadori products are gaining increasing attention. Until now, however, all FAOXs can only react with small glycated substrates (such as fructosyl amino acids or dipeptides), which has hindered the applications of this new class of enzymes in diagnosis, therapeutics, and detergents. In this study, Aspergillus fumigatus amadoriase II was engineered with the aim to expand its substrate range, using a heat-inducible autolytic vector and fructosyl-polylysine (3-13 lysines) as an intermediate-sized model substrate. After two rounds of directed evolution, a mutant (SII-82) was obtained that showed an 8.78-fold increase in the activity toward fructosyl-polylysine and which also performed several fold better than the wild-type on real gravy stains at concentrations of 10-100 A mu g/ml (parts per million). Mutational analyses revealed useful clues for altering the substrate-binding pocket. This study suggests that it is possible to manipulate fructosyl amine oxidases to accommodate larger substrates, and that mutant SII-82 might serve as a template for further engineering.