The hallmarks of pyridine nucleotide-dependent dehydrogenase reactions are the stereo- and regiospecific hydride transfer between the nicotinamide coenzyme and the corresponding substrate. When the hydride is delivered from NAD(P)H to reduce the keto-substrate, the site of attack is always at the carbonyl carbon. However, the apparent regioselectivity of the hydride transfer is reversed when difluoromethylene is used as a carbonyl mimic in the NADH-dependent enzyme, TDP-l-rhamnose synthase, which catalyzes the conversion of TDP-6-deoxy-l-lyxo-4-hexulose to TDP-l-rhamnose. The observed reversed regioselectivity can be explained by two mechanisms. One involves the formation of a carbene intermediate followed by a rearrangement involving 1,2-H shift. This mechanistic proposal is theoretically sound and would represent a rare example implicating the intermediacy of a carbene species in an enzyme reaction. However, our results are also consistent with a second mechanism in which the hydride addition to the difluoromethylene moiety occurs at the difluorinated end, opposite from the site predicted on the basis of the reduction of a normal keto functional group. Such a regioselectivity is well precedented in chemical models because nucleophilic addition to fluoroalkenes prefers a route in which the number of fluorines beta to the electron-rich carbon in the transition state is maximized. In this mechanism, the difluoromethylene group may be regarded as a carbonyl mimic with reversed polarity in enzyme catalysis. While further experiments are needed to discriminate between these mechanistic possibilities, the results reported here suggest that the apparent regioselectivity of hydride transfer in a pyridine nucleotide-dependent enzyme can be changed by altering the electrochemical properties of the reaction center.