Ribulose 1,5-bisphosphate carboxylase/oxygenase (Rubisco) occupies a critical position in photosynthetic CO2-fixation and consequently has been the focus of intense study. Crystal-structure-guided site-directed mutagenesis studies have met with limited success in engineering kinetic improvements to Rubisco, highlighting our inadequate understanding of structural constraints at the atomic level that dictate the enzyme's catalytic chemistry. Bioselection provides an alternative random mutagenic approach that is useful for identifying and elucidating imperceptible structure-function relationships. Using the dimeric Form II Rubisco from Rhodospirillum rubrum, its gene (rbcM) was randomly mutated and introduced under positive selection into Escherichia coli cells metabolically engineered to be dependent on Rubisco to detoxify its substrate ribulose 1,5-bisphosphate. Thirteen colonies displaying improved fitness were isolated, and all were found to harbor mutations in rbcM at one of two codons, histidine-44 or aspartate-117, that are structurally adjacent amino acids located about 10 A from the active site. Biochemical characterization of the mutant enzymes showed the mutations reduced their CO2/O2 specificity by 40% and decreased their carboxylation turnover rate by 20-40%. Structural analyses showed histidine-44 and aspartate-117 form a hydrogen bond in R. rubrum Rubisco and that the residues are conserved among other Form II Rubiscos. This study demonstrated the utility of directed evolution in E. coli for identifying catalytically relevant residues (in particular nonobvious residues disconnected from active site residues) and their potential molecular interactions that influence Rubisco's catalytic chemistry.