To date, rational redesign of glycosidase active-site clefts has been mainly limited to the removal of essential functionalities rather than their introduction. The glycoside hydrolase family 26 endo-beta-1,4-mannanase from the soil bacterium Cellulomonas fimi depolymerizes various abundant plant mannans. On the basis of differences in the structures and hydrolytic action patterns of this wild-type (but recombinantly expressed) enzyme and a homologous mannanase from Cellvibrio japonicus, two nonconserved amino acid residues at two distal glycone-binding subsites of the C. fimi enzyme were substituted, Ala323Arg at subsite -2 and Phe325Ala at subsite -3, to achieve inverted mannosyl affinities in the respective subsites, mimicking the Ce. japonicus enzyme that has an Arg providing mannosyl interactions at subsite -2. The X-ray crystal structure of the C. fimi doubly substituted mannanase was determined to 2.35 A resolution and shows that the introduced Arg323 is in a position suitable for hydrogen bonding to mannosyl at subsite -2. We report steady-state enzyme kinetics and hydrolysis-product analyses using anion-exchange chromatography and a novel rapid mass spectrometric profiling method of (18)O-labeled products obtained using H(2)(18)O as a solvent. The results obtained with oligosaccharide substrates show that although the catalytic efficiency (k(cat)/K(m)) is wild-type-like for the engineered enzyme, it has an altered hydrolytic action pattern that stems from promotion of substrate binding at subsite -2 (due to the introduced Arg323) and demotion of it at subsite -3 (to which removal of Phe325 contributed). However, k(cat)/K(m) decreased approximately 1 order of magnitude with polymeric substrates, possibly caused by spatial repositioning of the substrate at subsite -3 and beyond for the engineered enzyme.