(2R)-Methyl[2-3H]malonyl-CoA was used as the substrate for methylmalonyl-CoA epimerase from Propionibacterium shermanii, under conditions where the (2S)-methylmalonyl-CoA product was removed enzymically as fast as it was formed, and the fate of the label was monitored at different extents of reaction. Very little, if any, tritium is found attached to the C-2 position in the (2S)-epimer product (isolated as propionyl-CoA). Evidently, the hydrogen atom of the new C-H bond in the product is essentially solvent-derived. The rate of tritium release into the solvent is lower than the rate of product formation, and shows a primary kinetic tritium-isotope effect on kcat./Km of 2.3 +/- 0.1. The specific radioactivity of the remaining substrate rises slowly during the epimerase-catalysed reaction, and this provides an independent estimate of the primary kinetic tritium-isotope effect on kcat./Km of 1.6 +/- 0.5. These results, taken together, indicate that the mechanistic pathway of the epimerase-catalysed reaction resembles that established for proline racemase [Cardinale & Abeles, (1968) Biochemistry 7, 3970-3978], in which two enzyme bases are involved in catalysis. One base removes the proton from the substrate, the second provides the new proton, and there is no fast isotopic exchange between enzyme-bound intermediates and solvent protons.