The Roles of Glycine Residues in the ATP Binding Site of Human Brain Hexokinase*

Mutants of hexokinase I (Arg539→ Lys, Thr661 → Ala, Thr661 → Val, Gly534 → Ala, Gly679 → Ala, and Gly862 → Ala), located putatively in the vicinity of the ATP binding pocket, were constructed, purified to homogeneity, and studied by circular dichroism (CD) spectroscopy, fluorescence spectroscopy, and initial velocity kinetics. The wild-type and mutant enzymes have similar secondary structures on the basis of CD spectroscopy. The mutation Gly679 → Ala had little effect on the kinetic properties of the enzyme. Compared with the wild-type enzyme, however, the Gly534 → Ala mutant exhibited a 4000-fold decrease in k cat and the Gly862 → Ala mutant showed an 11-fold increase inK m for ATP. Glucose 6-phosphate inhibition of the three glycine mutants is comparable to that of the wild-type enzyme. Inorganic phosphate is, however, less effective in relieving glucose 6-phosphate inhibition of the Gly862 → Ala mutant, relative to the wild-type enzyme and entirely ineffective in relieving inhibition of the Gly534 → Ala mutant. Although the fluorescence emission spectra showed some difference for the Gly862 → Ala mutant relative to that of the wild-type enzyme, indicating an environmental alteration around tryptophan residues, no change was observed for the Gly534 → Ala and Gly679 → Ala mutants. Gly862 → Ala and Gly534 → Ala are the first instances of single residue mutations in hexokinase I that affect the binding affinity of ATP and abolish phosphate-induced relief of glucose 6-phosphate inhibition, respectively.

Hexokinase catalyzes the phosphorylation of glucose, using ATP as a phosphoryl donor. Four isoforms of hexokinase exist in mammalian tissue (1). Hexokinase isoforms I, II, and III have molecular weights of approximately 100,000 and are monomers under most conditions. Amino acid sequences of isoforms I-III are 70% identical (2). Moreover the N-and Cterminal halves of isoforms I-III have similar amino acid sequences, probably as a result of gene duplication and fusion (3)(4)(5)(6)(7). Hexokinase isoform IV (glucokinase) has a molecular weight of 50,000, similar to that of yeast hexokinase. Glucokinase exhibits (as does yeast hexokinase) significant sequence similarity to the N-and C-terminal halves of isoforms I-III.
Despite sequence similarities, the functional properties of hexokinase isoforms differ significantly. Isoform I (hereafter, brain hexokinase or hexokinase I) governs the rate-limiting step of glycolysis in brain and red blood cells (8,9). The reaction product, glucose 6-phosphate (Glu-6-P 1 ), inhibits both isoforms I and II (but not isoform IV) at micromolar levels. Inorganic phosphate (P i ), however, reverses Glu-6-P inhibition of only hexokinase I. The C-terminal domain of hexokinase I possesses catalytic activity, whereas the N-terminal domain is involved in the P i -induced relief of product inhibition (10). In contrast, both the C-and N-terminal halves exhibit comparable catalytic activity in isoform II (11). Thus, among hexokinase isoforms, brain hexokinase exhibits unique regulatory properties in that physiological levels of P i can reverse inhibition due to physiological levels of Glu-6-P (12)(13)(14).
The crystal structures of yeast hexokinase (15)(16)(17) are the basis for a model of mammalian glucokinase and its glucose binding site (18). The C-terminal domain of human brain hexokinase and its ATP binding site has been modeled in our laboratory based on similarities among the ATP-binding domains of actin, heat shock protein, and glycerol kinase (19). The model for the complex of ATP with hexokinase I puts a number of residues in the vicinity of ATP, of which Thr 680 , Asp 532 , and Arg 539 have been the focus of directed mutations and investigations of initial rate kinetics (19,20). These residues evidently stabilize the transition state, but do not influence the binding affinity of ATP (19,20).
This study presents the results of CD, fluorescence, and kinetic investigations of Gly 534 3 Ala, Arg 539 3 Lys, Thr 661 3 Ala, Thr 661 3 Val, Gly 679 3 Ala, and Gly 862 3 Ala mutants of brain hexokinase. The mutation Gly 862 3 Ala causes an order of magnitude increase in the K m for ATP, the first instance of a mutation in hexokinase I that has had a significant influence on the binding affinity of ATP. The mutation Gly 534 3 Ala reduces k cat by three orders of magnitude, but more significantly abolishes P i -induced relief of Glu-6-P inhibition. Mutation at Gly 534 is the first instance whereby a single mutation in hexokinase I has had a significant impact on the amelioration of product inhibition by P i .

EXPERIMENTAL PROCEDURES
Materials-Affi-Gel Blue and Bio-gel hydroxyapatite came from Bio-Rad. The Transformer TM site-directed mutagenesis kit (2nd version) was a product of CLONTECH. The Magic Minipreps DNA purification system was a product of Promega. Oligonucleotide synthesis and nucleotide sequencing was done by the Iowa State University nucleic acid facility. NruI and XhoI were obtained from New England Biolabs and Promega, respectively. The pET-11a plasmid was purchased from No-* This work was supported by research Grant NS 10546 from the National Institutes of Health and Grant MCB-9603595 from the National Science Foundation. This is Journal Paper J-17668 of the Iowa Agriculture and Home Economics Experiment Station, Ames, IA, Project 3191. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  hjfromm@iastate.edu. 1 The abbreviation used is: Glu-6-P, glucose 6-phosphate.
Expression of Human Brain Hexokinase-The cDNA for human hexokinase I was cloned into the expression vector pET-11a to generate pET-11a-HKI (21). pET-11a-HKI was transformed into E. coli strain ZSC13, which does not contain endogenous hexokinase. A 100-ml culture of the transformed E. coli was grown overnight in LB medium plus 40 mg/liter ampicillin and then added to 10 liters of the same medium. The culture was grown in a fermentor at 37°C to early log phase (A 600 ϭ 0.4) after which isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 0.4 mM to induce the T7 RNA polymerase gene. The culture was grown with stirring (200 rpm) and a filtered air flow of 5 p.s.i. for 24 h at 22°C.
Purification of Wild-type and Mutant Brain Hexokinase-The wildtype and mutant forms of hexokinase were purified as described elsewhere (20). The Gly 862 3 Ala mutant, however, did not bind to Affi-Gel Blue and was purified instead by DEAE anion-exchange chromatography.
Site-directed Mutagenesis-Mutagenesis was performed by following the instructions provided with the Transformer TM site-directed mutagenesis kit (2nd version) from CLONTECH. The oligonucleotides used for mutagenesis are as follows: 5Ј-CTGGATCTTGCAGGAACC-3Ј for Gly 534 3 Ala, 5Ј-CT-CATTGTTGCGACCGGC-3Ј for Gly 679 3 Ala, 5Ј-GTGGACGCGACACTC-TAC-3Ј for Gly 862 3 Ala, 5Ј-CAGTGGGCGCAATGATGACC-3Ј for Thr 661 3 Ala, 5Ј-CAGTGGGCGTCATGATGACC-3Ј for Thr 661 3 Val, 5Ј-ACCAATT-TCAAAGTGCTG-3Ј for Arg 539 3 Lys, where the altered codons are underlined. The selective oligonucleotide sequence was chosen at the NruI restriction site in the vector pET-11a and designed to change the NruI site to a XhoI site. The sequence of the selective oligonucleotide is 5Ј-CAGCCTCGC-CTCGAGAACGCCAG-3Ј, where the underline represents the XhoI site. Mutations were verified by fluorescent dideoxy chain termination DNA sequencing.
Hexokinase Assay-Hexokinase activity was determined spectrophotometrically as described previously (21). The kinetic parameters depicted in Table I were obtained from initial rate data obtained from two or more experiments. The substrate concentrations in the kinetic experiments were varied from K m /2 to 5K m . At least three concentrations of the inhibitor 1,5-anhydroglucitol-6-phosphate, from below to above its K i , were used to evaluate its effect on the kinetics of the brain hexokinase enzyme.
Methods-1,5-anhydroglucitol-6-phosphate was prepared as described elsewhere (22). Protein concentration was determined by the method of Bradford (23) using bovine serum albumin as a standard. CD spectra were recorded using a Jasco J710 CD spectrometer as described elsewhere (20).
Fluorescence Emission Spectra-The wild-type and mutant enzymes were dialyzed against 20 mM Hepes buffer (pH 7.0) containing 1 mM of ␤-mercaptoethanol. The enzyme concentration was 66.7 g/ml. The fluorescent intensity was recorded over the wavelength range from 300 -350 nm, using an excitation wavelength of 290 nm.

Purification of the Wild-type and Mutant Human Brain
Hexokinase-The use of a 10-liter fermentor enhanced enzyme yield compared with 2-liter growth flasks. 20 liters of culture provided 70 mg of pure hexokinase I. The lack of retention of the Gly 862 3 Ala mutant on Affi-Gel Blue, a nucleotide affinity column, is consistent with the elevated K m for ATP exhibited by this mutant (see below). The wild-type and mutant enzymes were more than 95% pure on the basis of SDS-polyacrylamide gel electrophoresis (data not shown).
Characterization of Mutants by Kinetics and Spectroscopy-The kinetic parameters of wild-type and mutant enzymes are in Table I. Compared with the wild-type enzyme, the Gly 534 3 Ala mutant showed a 4000-fold decrease in k cat , and 4-, 5-, and 3-fold increases in the K m (or K i ) values for glucose, ATP, and 1,5-anhydroglucitol-6-phosphate (an analog that mimics Glu-6-P (24) and which can be used in the hexokinase-glucose-6phosphate dehydrogenase coupled spectrophotometric assay), respectively. The Gly 862 3 Ala mutant showed an 11-fold increase in the K m for ATP relative to the wild-type enzyme (Table I), a 2-fold increase in the K m for glucose, an 18-fold decrease in k cat , but little change in the K i for 1,5-anhydroglucitol-6-phosphate. The mutation of Gly 679 to alanine did not change kinetic properties, but the Gly 679 3 Ile mutant was relatively insoluble (20). P i reverses Glu-6-P inhibition of the Gly 679 3 Ala mutant and the wild-type enzyme to the same extent, whereas P i has a modest or no effect on the inhibition of the Gly 862 3 Ala and Gly 534 3 Ala mutants, respectively (Fig. 1).
The fluorescence emission spectra of the wild-type, Gly 534 3 Ala and Gly 679 3 Ala enzymes were identical, whereas the spectrum for the Gly 862 3 Ala mutant differed from that of the wild-type enzyme (Fig. 2). This difference is probably due to changes in the local environment of tryptophan residues. However, the CD spectra for the wild-type and all the mutant enzymes are identical (data not shown), suggesting no global conformational differences among these proteins.
The Thr 661 3 Ala mutant showed a 4-fold increase in the K i for 1,5-anhydroglucitol-6-phosphate. The Thr 661 3 Val mutant exhibited a 9-fold decrease in k cat relative to the wild-type enzyme. Other kinetic parameters for the two mutants were unaltered relative to those of the wild-type enzyme. The Arg 539 3 Lys mutant showed a 12-fold decrease in k cat , and little change in the K m for either ATP or glucose or the K i for 1,5-anhydroglucitol-6-phosphate. DISCUSSION Arg 539 putatively interacts with the polyphosphoryl portion of ATP and stabilizes the transition state (19,20). The Arg 539 3 Lys mutant is 10-fold more active than the Arg 539 3 Ile mutant (20), suggesting the importance of the positive charge at position 539. However, as the Arg 539 3 Lys mutant reported here is still 12-fold less active than wild-type hexokinase I, specific hydrogen bond interactions of the arginyl side chain are of equal importance in stabilizing the transition state. We have suggested, on the basis of previous work, that Arg 539 may form salt bridges with oxygen atoms of the ␣and ␤-phosphoryl groups of ATP (19). The observed properties of the Lys 539 The values shown are the mean Ϯ S.D. The kinetic analysis was done with the computer program MINITAB (19). Figures in parentheses represent p value % differences between the mutant and wild-type kinetic parameters. a p value % differences between the mutant and wild-type kinetic parameters is 0.05. mutant are consistent with that suggestion. Thr 661 is 10 Å away from the ␤-phosphoryl group of ATP in our model, however, the ␥-oxygen atom of Thr 661 is 3 Å from Asp 657 , which is putatively the catalytic base in the abstraction of a proton from the 6-hydroxyl group of glucose (25). The Thr 661 3 Ala mutant has kinetic properties similar to those of the wild-type enzyme, but mutation of Thr 661 to valine causes a 9-fold decrease in k cat , probably by introducing an unfavorable nonbonded contact that perturbs Asp 657 .
Consensus sequences, Gly-X-X-Gly-X-Gly-Lys-(Ser/Thr) in mononucleotide-binding proteins (26), Gly-X-Gly-X-X-Gly in dinucleotide-binding proteins (26), and Y-Gly-X-Gly-X-(Phe/ Tyr)-Gly-X-Val, where Y is a hydrophobic residue for protein kinases (27), are rich in conserved glycines. In the mononucleotide-binding protein, p21 H-ras , the dihedral angles of the polypeptide chain require glycine (28). For dinucleotide-binding proteins (29), the second glycine of the consensus sequence provides space for the polyphosphoryl moiety, and the first and third glycines satisfy conformational constraints of the polypeptide chain. The glycine-rich sequences of protein kinases participate in nucleotide binding, substrate recognition, and enzyme catalysis (30,31). Yeast hexokinase, actin, hsc70, and glycerol kinase, however, are without a consensus sequence for nucleotide binding. Instead, the residues associated with nucleotide binding are scattered throughout the primary structure, but come together at single sites in the context of the folded polypeptide chains (32). Interestingly, the ATP binding domains of yeast hexokinase, actin, hsc70, and glycerol kinase are also rich in glycine residues.
We have probed the corresponding glycines by directed mutation, in the expectation that some of these glycines are linked to observed kinetic properties in hexokinase I. The mutation of Gly 534 to alanine produced a dramatic effect on k cat (4000-fold reduction) and modest effects on K m for glucose and ATP. Gly 534 is conserved in sequences of hexokinase, but according to our model (Fig. 3), its main chain torsion angles fall in the allowed region of the Ramachandran plot for alanine. Instead, C␤ of Ala 534 is 3.6 Å from an oxygen of the ␤-phosphoryl group of ATP, but perhaps of greater significance is its 2.4 Å contact with backbone carbonyl 537 of an adjacent ␤-strand. Our model suggests then, the possibility of conformational change in the vicinity of residue 534 to relieve the close contact mentioned above. Such a local conformational change could influence Asp 532 , which on the basis of earlier work (19) plays a critical role in the stabilization of the transition state and may be involved in the binding of Mg 2ϩ . A larger perturbation on the active site due to the mutation of Gly 534 to alanine is not likely, because K m values for substrates are not influenced and CD spectroscopy indicates no change in secondary structure.
The Gly 534 3 Ala mutant represents the first instance whereby the mutation of a single residue has abolished P iinduced relief of Glu-6-P inhibition in hexokinase I. Glu-6-P inhibition of the C-terminal half of hexokinase I (mini-hexokinase) cannot be reversed by P i , implicating the N-terminal domain in the relief of inhibition (20). The mechanism by which P i relieves Glu-6-P inhibition then evidently involves structural elements of both the N-and C-terminal halves of hexokinase I. Furthermore, the loss of P i -induced relief of Glu-6-P inhibition in the Gly 534 3 Ala mutant is linked closely to position 534, as mutations of Asp 532 to lysine and glutamate have little effect on this property (19).
Gly 679 and Gly 862 belong to reverse turns, which pack against each other in our model (Fig. 3). The main chain torsion angles put positions 679 and 862 in unallowed regions of the Ramachandran plot for alanine ( ϭ Ϫ111, ϭ Ϫ133 for Gly 679 ; ϭ 70, ϭ 168 for Gly 862 ). Of the two mutants, Gly 862 3 Ala has the conformation of highest energy. Although the C␤ atoms at positions 679 and 862 probably do not interact with ATP, they make unfavorable contacts in our model with backbone amide 863 (2.6 Å) and backbone amide 679 (2.7 Å), respectively. These unfavorable contacts may not be significant, however, as the mutation of Gly 679 to alanine has no effect on the kinetic properties of the enzyme. Instead, the introduction of alanine at position 862 probably causes conformational changes in main chain torsion angles. Although the fluores-cence spectra of the wild-type and Gly 862 3 Ala enzymes differ (indicating a perturbation in the local environment of tryptophan residues) their CD spectra are identical (indicating no change in secondary structure). Furthermore, the K i for 1,5anhydroglucitol-6-phosphate and the K m for glucose are similar for the Gly 682 3 Ala mutant and the wild-type enzyme. Thus the mutation of Gly 862 to alanine probably has an effect only on residues in the vicinity of position 862. Thr 863 , a residue conserved in hexokinase sequences, interacts with the ribose and base moieties of ATP in our model (Fig. 3). The Gly 682 3 Ala mutant, then, could influence interactions involving the base of ATP by perturbing the conformation or relative position of Thr 683 . The mutation of Gly 862 to alanine increases the K m for ATP by 11-fold without large changes in other kinetic parameters and as such, represents the first mutation, which to our knowledge influences the binding affinity of ATP.
Mutations prepared here and from previous studies (19) show a trend that may be significant to the function of hexokinases in general. Mutations of hexokinase I, which putatively influence interactions involving the polyphosphoryl moiety of ATP, have no effect on K m but a large effect on k cat . Conversely, the Gly 862 3 Ala mutant, which putatively influences interactions at the base moiety, has little effect on k cat but substantial effects on K m for ATP. Interactions involving the base of ATP are important for affinity, but polyphosphoryl-protein interactions are involved in the stabilization of the transition state. Conceivably, hexokinase I diverts energy from favorable interactions between the polyphosphoryl moiety of ATP and the enzyme to promote conformational changes that stabilize the transition state. This phenomenon is not without precedence. In adenylosuccinate synthetase from E. coli mutations involving protein interactions at the base of GTP affect K m (33), whereas interactions between the polyphosphoryl group of GTP and the protein (and Mg 2ϩ ) contribute to the stability of the transition state by driving conformational changes in the active site (34).