The mutability of enzyme active-site shape determinants

Abstract

Investigations of enzyme action typically focus on elucidating the catalytic roles of hydrogen bonding interactions between polar active-site residues and substrate molecules. Less clear is the importance of non-hydrogen bonding contacts to enzymatic rate accelerations. To investigate the importance of such interactions in a model system, six residues that participate in van der Waals contacts with substrate glucose within the active site of Escherichia coli glucokinase were individually randomized via site-directed mutagenesis. In vivo selection in a glucokinase-deficient bacterium was employed to identify amino acid substitutions that were complicit with enzyme activity. The results suggest that small residues, such as alanine and glycine, are largely immutable, whereas larger amino acids are more tolerant of diverse substitution patterns. Surprisingly, a glucokinase variant that contains glycine in place of six non-hydrogen bonding contacts retains ∼1% of the wild-type activity. These findings establish non-hydrogen bonding shape determinants as highly appealing targets for widespread substitution during efforts to redesign the catalytic properties of natural enzymes.

Enzyme active sites are highly complementary to the structure of their cognate substrates. This complementarity is generated by a combination of polar residues that form hydrogen bonds with the bound substrate, and non-hydrogen bonding interactions that create a solvent-excluded template of the substrate's van der Waals surface. Numerous investigations of enzyme catalysis have established the critical importance of hydrogen bonding contacts between polar active-site residues and the substrate molecule (Kraut et al. 2003; Wolfenden 2003; Warshel et al. 2006). In contrast, no systematic investigation of the role of van der Waals contacts in enzyme action appears to have been presented in the literature. In principle, such active-site constituents could provide a number of beneficial attributes to a protein catalyst. For example, these residues could supply structural integrity to the active site and provide a framework from which to display polar side chains. In addition, van der Waals contacts could assist in the proper positioning of substrates with respect to one another, as well as with respect to key catalytic residues. Nonpolar, non-hydrogen bonding contacts could also prevent the undesired access of solvent water to substrates and/or reactive intermediates. Historically, it has been difficult to experimentally assess the impact of such factors upon enzyme catalyzed chemical transformations.

Escherichia coli glucokinase catalyzes the ATP-dependent phosphorylation of glucose in the first step of glycolysis (Fraenkel 1986). The glucokinase reaction is a paradigm for bisubstrate enzyme catalysis (Jencks 1969; Fry et al. 1980). This transformation requires the precise positioning of the substrate's O6 atom for in-line attack at the γ-phosphoryl group of ATP. The crystal structure of the bacterial enzyme reveals a series of non-hydrogen bonding contacts between bound glucose and active-site residues that occupy the primary shell of the binding site (▶; Lunin et al. 2004). Specifically, one side of the glucose binding site is constructed from the polypeptide backbone formed via three consecutive residues: Ala-64, Cys-65, and Pro-66. A pair of neighboring glycine residues, Gly-138 and Gly-140, provides the opposite face of the glucose-binding cleft. The hydrophobic side chain of Phe-101, which is adjacent to the critical active-site base Asp-100, donates a sixth nonpolar constituent. Upon glucose binding, the Phe-101 side chain moves ∼2.5 Å into a position where it forms close contacts with the reactive O6 atom of glucose. Together, these six residues combine to form a solvent-excluded binding site that is highly complementary to the glucose molecule (▶). Given these considerations, the bacterial glucokinase system provides an interesting test case for investigating the mutability of non-hydrogen bonding interactions within an enzyme active site.

Figure 1.

The structure of Escherichia coli glucokinase. (A) Glucokinase in complex with substrate glucose showing six non-hydrogen bonded active-site residues and the critical active-site base, Asp-100. (B) Space-filling model of the glucokinase active site depicting the structural complementarity generated by non-hydrogen bonding contacts. Images were created with the program PyMOL (Delano Scientific) and PDB entry 1SZ2.

Results and Discussion

Combinatorial mutagenesis was coupled with functional selection to assess the mutability of six shape determinants in the E. coli glucokinase active site. A summary of the permissible amino acid substitutions for each non-hydrogen bonding contact is provided in ▶. Of the three consecutive amino acids whose peptide backbone form one side of the glucose binding site, only Ala-64 showed significant restriction on amino acid identity. Only residues with small side chains, such as glycine, serine, and cysteine, were permissible at this position. In contrast, 14 of 20 possible residues were observed to effectively substitute for Cys-65. Similarly, 15 different residues were observed to replace Pro-66, a finding that is somewhat surprising given the constrained geometry often imposed by proline residues upon polypeptide structures (Schimmel and Flory 1968). These results indicate that the main chain atoms, not the side chains, are the most important structural features of the amino acids that occupy positions 65 and 66 in the glucokinase polypeptide.

Figure 2.

Mutability of the glucokinase active site. Permissible amino acid substitutions observed at six active-site residues in glucokinase identified after randomization and in vivo selection for functional enzymes: (white = nonpolar; yellow = polar; red = acidic; blue = basic). The numbers in parentheses represent the sequence conservation percentage of each residue as determined from an analysis of 45 diverse glucokinases that display expectation values of ≤10⁻⁴⁵ when compared with the primary sequence of E. coli glucokinase.

Unlike the backbone contacts donated by residues 64–66, two glycine residues on the opposite face of the active site were found to be immutable. No other amino acid could replace either Gly-138 or Gly-140 without reducing catalytic activity below the 10⁴ M⁻¹ sec⁻¹ limit imposed by the selection criteria. One possible explanation for the absolute requirement of glycine at positions 138 and 140 is the conformational flexibility provided by this amino acid. Glycine has a wider range of permissible Ramachandran angles than other amino acids (Ramachandran and Sasisekharan 1968). Nevertheless, the ϕ and ψ values adopted by Gly-138 and Gly-140, determined from the crystal structure of E. coli glucokinase, do not reveal a special conformational requirement of these polypeptide positions (Lunin et al. 2004). Instead, the location of Gly-138 and Gly-140 near the reactive O6 atom of substrate glucose suggests that substitution of either residue may induce a ground-state ES complex that is not conducive to efficient phosphoryl transfer.

The final residue investigated was Phe-101, an amino acid that forms close interactions with the reactive O6 position of substrate glucose (▶). It is reasonable to expect that the hydrophobic nature of this side chain could be important in excluding bulk solvent from the active site. Exclusion of water during the glucokinase reaction is essential to prevent inadvertent ATP hydrolysis (Delafuente and Sols 1970). Solvent exclusion may also enhance the effectiveness of general base catalysis provided by the carboxylate side chain of the neighboring Asp-100. Surprisingly, the current results demonstrate that Phe-101 is quite tolerant of a variety of substitutions. Both polar and nonpolar residues successfully replace Phe-101, with little effect upon catalytic activity. Even lysine, with its positively charged side chain, can substitute for Phe-101. This observation was unexpected, since lysine could, in principle, counteract the negatively charged Asp-100 side chain that is essential for catalysis. The ability of small residues, such as glycine, to substitute for Phe-101 suggests that exclusion of solvent near the site of phosphoryl transfer may be less important than previously anticipated. Alternatively, the active site may be flexible enough to readjust to the smaller glycine and still produce a solvent-excluded binding site. To investigate this point further, a variant in which Phe-101 was replaced with glycine was purified and its ability to hydrolyze ATP in the absence of a carbohydrate substrate was investigated. Exposure of a subsaturating concentration of ATP (0.5 mM) to high enzyme concentrations (up to 32 μM) revealed that the k _cat/K_m value for the ATPase activity of this variant was <10 M⁻¹ sec⁻¹. Thus, glucokinase retains the ability to effectively discriminate between its native substrate and water, even in the absence of the Phe-101 side chain.

To determine whether the extent of mutability of individual active-site residues observed in this functional study might be reflected in sequence conservation data, a search of the NCBI database was conducted. The primary amino acid sequences of 45 diverse bacterial glucokinases that share a high degree of homology with E. coli glucokinase (expectation values of <10⁻⁴⁵) were analyzed. The percentage of these sequences that retained the identity of the E. coli glucokinase amino acid at each non-hydrogen bonding contact is depicted in ▶. Similar to the results of the functional studies, Gly-138 and Gly-140 were conserved in 100% of the sequences analyzed. Ala-64, which the functional studies revealed to be largely, but not totally immutable, was also conserved in 100% of the sequences. By contrast, Phe-101 and Pro-66 were conserved in 96% (43/45) and 89% (40/45) of the sequences investigated. Based on sequence comparisons alone, the most mutable residue was Cys-65. This amino acid was conserved in only 33% of the sequences. Based on functional studies, Cys-65 and Pro-66 were found to be equally amenable to substitution. Overall, the results of the sequence comparison correlate well with the experimental data accumulated from the functional studies, suggesting that amino acid alignments can provide a reasonable initial estimate of the mutability of individual active-site residues.

The data compiled in ▶ reveal that every non-hydrogen bonding contact in the glucokinase active site can individually be replaced with glycine without suffering a significant loss in catalytic activity. What is the effect of simultaneously replacing all of these residues with glycine? To answer that question, site-directed mutagenesis was used to create a variant form of glucokinase in which Ala-64, Cys-65, Pro-66, and Phe-101 were each replaced with glycine. Under the selective growth conditions employed in this study, expression of this variant glk gene efficiently complemented the glucokinase deficiency of BM5340(DE3). To evaluate the effect of the glycine replacements upon the catalytic activity of the enzyme, the recombinant hexa-glycine enzyme was expressed and purified from BM5340(DE3) cells. The kinetic features of the wild-type and variant enzymes are compared in ▶. The value of k _cat for the glycine variant is decreased by 25-fold and the value of K _{m, glucose} is increased by sevenfold. Interestingly, the K_m value for ATP is also elevated in the hexa-glycine variant, although there is no evidence that any of the altered amino acids interact with bound ATP. Overall, the elimination of non-hydrogen bonding contacts in the E. coli glucokinase active site reduces the catalytic efficiency by a measurable, but modest 200-fold. By comparison, the total rate enhancement produced by this enzyme is estimated to exceed 10¹⁰ (R. Wolfenden, pers. comm.).

Table 1.

Kinetic constants of wild-type glucokinase and the hexa-glycine variant

Conclusions

The present results provide new insight into the relationship between fit and function in an enzyme active site that catalyzes a bisubstrate transformation. The high degree of mutability of Phe-101 suggests that this residue does not actively participate in excluding solvent water from the glucokinase active site during phosphoryl group transfer. In general, small residues, such as glycine and alanine, were found to be less replaceable than were larger amino acids. This finding is consistent with expectations based upon simple steric considerations. The fact that the hexa-glycine variant retains activity, both in vitro and in vivo, indicates that non-hydrogen bonding contacts, in general, may be highly tolerant to widespread substitution. Indeed, such active-site shape determinants may prove to be attractive targets for randomization during future attempts to alter the substrate specificities and catalytic activities of existing enzymes in a systematic fashion (Yoshikuni et al. 2006; Glasner et al. 2007; Woycechowsky et al. 2007).

Materials and Methods

Mutagenesis and genetic selection

To assess the mutability of individual non-hydrogen bonding contacts, inactivating mutations were introduced into three discrete positions within the glucokinase coding sequence. QuikChange (Stratagene) site-directed mutagenesis of plasmid pBGM101-glk was conducted to replace the active-site base, Asp-100, with alanine (Miller and Raines 2004). Using a similar strategy, an opal stop codon was inserted in place of the triplet that specifies Cys-65 and Gly-138. These three templates were used in a second round of mutagenesis that employed oligonucleotide primers designed to correct each defect and replace nearby nonpolar residues with the randomized sequence NN(G/T) (Neylon 2004). Following amplification, 10 units of DpnI restriction endonuclease was added to each reaction and the mixture was incubated at 37°C for 1 h to digest template DNA. The randomized libraries were purified with a PCR cleanup kit (Promega) and the resulting DNA was eluted into 10 μL of ddH₂O.

Randomized PCR products were transformed into the glucokinase-deficient E. coli auxotroph, BM5340(DE3), via electroporation. Transformed cells were resuspended in 1 mL of SOC and allowed to recover for 1 h at 37°C. Cell were centrifuged at 3000g and washed twice with 1 mL M9 minimal media before being plated on M9 minimal plates containing glucose (0.005% w/v), ampicillin (150 μg/mL), kanamycin (40 μg/mL), chloramphenicol (25 μg/mL), and IPTG (50 μM) to induce gene expression. Plates were incubated at 37°C and colonies that grew within 48 h were scored as positives. Control experiments with known glucokinase variants indicate that enzymes possessing a k _cat/K_m value of ≥10⁴ M⁻¹ sec⁻¹ enable host cell survival under these growth conditions, whereas variants with lower k _cat/K_m values do not complement the glucokinase deficiency of BM5340(DE3). Colonies were inoculated into 500 μL of LB media containing ampicillin (150 μg/mL), kanamycin (40 μg/mL), and chloramphenicol (25 μg/mL). Following 6 h of growth at 37°C, an additional 4 mL of antibiotic supplemented LB media was added, and the cultures were allowed to grow to saturation overnight. Plasmid DNA was prepared from positive clones and retransformed into BM5340(DE3) to verify the selected phenotype. Randomized library members that survived both rounds of selection were submitted for DNA sequencing.

Construction, expression, and enzymatic assays of the hexa-gly variant

Two sequential rounds of QuikChange (Stratagene) site-directed mutagenesis were performed on the wild-type glk template to replace Ala-64, Cys-65, Pro-66, and Phe-101 with glycine. Following confirmation of successful mutagenesis via DNA sequencing, the hexa-glycine variant was produced in glucokinase-deficient BM5340(DE3) cells to avoid contamination from endogenous glucokinases. The enzyme was produced as a C-terminal hexa-histidine tagged polypeptide and purified using immobilized metal affinity chromatography, as previously described (Miller and Raines 2005). The activity of this enzyme was determined with a spectrophotometric assay that couples the production of glucose 6-phosphate to the reduction of NADP via glucose 6-phosphate dehydrogenase (Cleland 1979). Assay mixtures included Tris (0.1 M, pH 7.6), DTT (1 mM), NADP (0.4 mM), ATP (0.1–10 mM), MgCl₂ (1.1–11 mM), and glucose (0.1–25 mM). Activity assays were performed at 25°C and data were fitted to standard Michaelis-Menten kinetic equations. The ATPase activity of the hexa-glycine variant was estimated by coupling the production of ADP to the oxidation of NADH via the combined action of pyruvate kinase (10 units) and lactate dehydrogenase (10 units) (Easterby 1973). Assays were conducted at 25°C and contained Tris (0.25 M, pH 7.6), NADH (0.5 mM), DTT (1 mM), PEP (5 mM), ATP (0.5 mM), MgCl₂ (1.5 mM), and KCl (5 mM).