SEVERAL hundred million tons of toxic mercurials are dispersed in the biosphere1. Microbes can detoxify organo-mercurials and mercury salts through sequential action of two enzymes, organomercury lyase2 and mercuric ion reductase (MerA) 3-5. The latter, a homodimer with homology to the FAD-dependent disulphide oxidoreductases6, catalyses the reaction NADPH + Hg(II) → NADP+ + H+Hg(0), one of the very rare enzymic reactions with metal substrates. Human glutathione reductase7,8 serves as a reference molecule for FAD-dependent disulphide reductases and between its primary structure9 and that of MerA from Tn501 (Pseudomonas), Tn21 (Shigella), pI258 (Staphylococcus) and Bacillus, 25-30% of the residues have been conserved10,11. All MerAs have a C-terminal extension about 15 residues long but have very varied N termini. Although the enzyme from Streptomyces lividans has no addition, from Pseudomonas aeruginosa Tn5Ol and Bacillus sp. strain RC607 it has one and two copies respectively of a domain of 80-85 residues, highly homologous to MerP, the periplasmic component of proteins encoded by the mer operon11. These domains can be proteolytically cleaved off without changing the catalytic efficiency3. We report here the crystal structure of MerA from the Gram-positive bacterium Bacillus sp. strain RC607. Analysis of its complexes with nicotinamide dinucleotide substrates and the inhibitor Cd(II) reveals how limited structural changes enable an enzyme to accept as substrate what used to be a dangerous inhibitor. Knowledge of the mode of mercury ligation is a prerequisite for understanding this unique detoxification mechanism.