The function of the hydrophobic residues Leu28, Phe31, Ile50, and Leu54 at the folate binding site in Escherichia coli dihydrofolate reductase (5,6,7,8-tetrahydrofolate:NADP+ oxidoreductase, EC 18.104.22.168) has been studied by a combination of site-specific mutagenesis and reaction kinetics. Studies suggest that the overall protein structure and kinetic sequence for the reaction did not change for the mutant proteins compared to the wild-type enzyme. Two sets of mutated reductases have been constructed. The first set, in which the side chains of the targeted amino acids are spatially well separated (∼8 Å), includes two single mutants (L28Y and L54F) and a double mutant (L28Y-L54F). This set features residues that increased the side chain surface area and the potential for substrate interactions. Unexpectedly, nonadditivity in the free energy changes for the thermodynamics of ligand binding and in the rates of hydride transfer and product release is observed. The progressive increase in dihydrofolate binding is reversed for the sterically more crowded double mutant, with ΔΔG ca. 3 kcal mol−1 less favorable than anticipated. On the other hand, the decrease in the rate constant for hydride transfer noted with the single mutants relative to the wild-type enzyme is reversed for the double mutant, so that ΔΔG‡ is ca. 2 kcal mol−1 more favorable. The second set of mutant proteins includes two double mutants (L28A-F31A and I50A-L54G) in which the selected amino acids are separated by three to four intervening amino acids and a quadruple mutant (L28A- F31A-I50A-L54G) in which the two sets L28A-F31A and I50A-L54G are spatially distinct. This set deleted the side chain surface area to lower the opportunity for substrate interactions. Nonadditivity in the free energy changes associated with key kinetic and thermodynamic parameters is again observed. The decrease in dihydrofolate binding found with the two double mutants is not observed with the quadruple mutant, which binds the substrate with ΔΔG ca. 6.5 kcal moh−1 more favorable than expected. Similarly, the quadruple mutant has a larger rate constant for hydride transfer (−ΔΔG‡ ≃ 1.7 kcal mol−1) than predicted. One interpretation for the nonadditivity is that these residues interact through binding of the folate substrate, which serves to link molecularly remote side chain moieties within the active site. The resiliency of the active site to tolerate such damage suggests that there must be sufficient flexibility and redundancy built into the site and its framework to retain a high population of active enzyme-substrate complexes.