To develop a general approach to designing cofactor-binding sites for catalytic antibodies, we characterized structural patterns in the binding sites of antibodies and zinc enzymes. Superposition of eight sets of antibody light- and heavy-chain variable domains identified structurally conserved sites within the sequence-variable complementarity determining regions. The pattern for catalytic zinc sites included two ligands close in sequence, a sequence-distant ligand, and a main-chain hydrogen bond joining two ligands. In both the light- and heavy-chain variable domains, the stereochemistry of five structurally conserved sites general to all known antibody structures matched that of the zinc ligands of carbonic anhydrase: three residues on two hydrogen-bonded antiparallel beta-strands. For one such general site, an antibody model replacing residue 34 on the first complementarity determining region of the light chain (L1) and residues 89 and 91 on the third complementarity determining region of the light chain (L3) with histidine ligands formed a zinc-binding site with an open coordination position at the bottom of the antibody binding pocket. For the anti-fluorescein antibody 4-4-20, this L1-L3 site placed the zinc ion about 4 A from the bound fluorescein, an indicator for metal binding. This predicted zinc-binding mutant was created in the single-chain variable domain construct, expressed, and found by fluorescence quenching to bind metal ion with an affinity constant of 10(6) M-1. Thus, our template-based multisite design proved successful for remodeling an antibody to contain a cofactor-binding site, without requiring further mutagenesis and screening. Combination of a specific light or heavy chain containing a catalytic metal site with a library of complementary chains raised to potential substrates or transition state analogs should greatly improve the production of catalytic antibodies with desired activities and specificities.