A protein engineering approach to delineating which distinct elements of homologous tRNA synthetase architectures are responsible for divergent RNA-amino acid pairing specificities is described. Previously, we constructed a hybrid enzyme in which 23 amino acids from the catalytic domain of Escherichia coli glutaminyl-tRNA synthetase (GlnRS) were replaced with the corresponding residues of human glutamyl-tRNA synthetase (GluRS). The engineered hybrid (GlnRS S1/L1/L2) synthesizes Glu-tRNA(Gln) more than 10(4)-fold more efficiently than GlnRS. Detailed comparison of kinetic parameters between GlnRS S1/L1/L2 and the naturally occurring Methanothermobacter thermautotrophicus GluRS(ND), which is also capable of Glu-tRNA(Gln) synthesis, now shows that both k(cat) and K(m) for glutamate are recapitulated in the engineered enzyme, but that K(m) for tRNA is 200-fold higher. Thus, the simultaneous optimization of paired amino acid and tRNA binding sites found in a naturally occurring enzyme is not recapitulated in a hybrid that is successfully engineered for amino acid complementarity. We infer that the GlnRS architecture has differentiated to match only cognate amino acid-RNA pairs, and that the substrate selection functions do not operate independently of each other. Design and characterization of four additional hybrids identify further residues involved in improving complementarity for glutamate and in communicating between amino acid and tRNA binding sites. The robust catalytic function demonstrated in this engineered system offers a novel platform for exploring the stereochemical origins of coding as a property of the ancient Rossmann fold.