Glucose Phosphorylation SITE-DIRECTED MUTATIONS WHICH IMPAIR THE CATALYTIC FUNCTION OF HEXOKINASE*

Recent studies from this and other laboratories have resulted in the cloning and sequencing of hexokinases from a variety of tissues including yeast, human kidney, rat brain, rat liver, and mouse hepatoma. Significantly, studies on the hepatoma enzyme conducted in this laboratory (Arora, K.K., Fanciulli, M., and Pedersen, P.L. (1990) J. Biol. Chem. 265, 6481-6488) resulted also in its overexpression in Escherichia coli in active form. We have now used site-directed mutagenesis for the first time in studies of hexokinase to evaluate the role of amino acid residues predicted to interact with either glucose or ATP. Four amino acid residues (Ser-603, Asp-657, Glu-708, and Glu-742) believed to interact with glucose were mutated to alanine or glycine, whereas a lysine residue (Lys-558) thought to be directly involved in binding ATP was mutated to either methionine or arginine. Of all the mutations in residues believed to interact with glucose, the Asp-657----Ala mutation is the most profound, reducing the hexokinase activity to a level less than 1% of the wild type. The relative Vmax values for Ser-603----Ala, Glu-708----Ala, and Glu-742----Ala enzymes are 6, 10, and 6.5%, respectively, of the wild-type enzyme. Glu-708 and Glu-742 mutations increase the apparent Km for glucose 50- and 14-fold, respectively, while the Ser-603----Ala mutation decreases the apparent Km for glucose 5-fold. At the putative ATP binding site, the relative Vmax for Lys-558----Arg and Lys-558----Met enzymes are 70 and 29%, respectively, of the wild-type enzyme with no changes in the apparent Km for glucose. No changes were observed in the apparent Km for ATP with any mutation. These results support the view that all 4 residues predicted to interact with glucose from earlier x-ray studies may play a role in binding and/or catalysis. The Asp-657 and Ser-603 residues may be involved in both, while Glu-708 and Glu-742 clearly contribute to binding but are not essential for catalysis. In contrast, Lys-558 appears to be essential neither for binding nor catalysis.

I To whom correspondence should be addressed. Hexokinase (ATP:D-hexose 6-phosphotransferase, EC 2.7.1.1) is the first enzyme of the glycolytic pathway that commits glucose to catabolism by catalyzing its phosphorylation to Glc-6-P with MgATP. In rapidly growing tumor cells, the enzyme hexokinase is markedly elevated and 50430% of the total activity is bound to the outer mitochondrial membrane (1)(2)(3), where it has preferred access to mitochondrially generated ATP (3). Of the four known hexokinases distinguished on the basis of charge (Types I, 11,111, and IV), the major form found in tumors exhibits properties in common with both the Type I and I1 isozymes (4).
Recently, we have reported the primary structure of tumor mitochondrial hexokinase deduced from a full-length cDNA clone isolated from a highly glycolytic mouse hepatoma cell line (5). The tumor enzyme consists of 918 amino acids and has a mass of 102,272 daltons which is very close to that of the recently cloned and sequenced Type I hexokinases from rat brain (6) and human kidney (7). These three enzymes are twice the size of hexokinase Type IV (also called glucokinase) and yeast hexokinase. Similar to the Type I rat brain and human kidney hexokinases (6,7), the tumor enzyme appears to have arisen from a gene duplication-gene fusion event resulting in a gene encoding a hexokinase with approximately two halves, each with a mass of 50 kDa which share close to 68% amino acid sequence homology (5). Significantly, the Cterminal half of the brain enzyme has been shown to account for all its catalytic activity (8), whereas the N-terminal half is predicted to be involved in product inhibition by Glc-6-P (6,9).
Based on amino acid sequence comparisons of tumor, brain, kidney, and liver hexokinases with that of the yeast enzyme, the x-ray structure of which is known (10,l l), putative glucose and ATP-binding domains have been predicted (12). The glucose-binding domains, found in both the C-and N-terminal halves, include residues thought to interact directly with glucose in the yeast enzyme, i.e. Ser-158, Asp-211, Glu-269, and Glu-302 (11). These residues correspond, respectively, to Ser-603, Asp-657, Glu-708, and Glu-742 of the C-terminal half of the tumor enzyme (5).
Although x-ray studies using 8-bromo-AMP predict a nucleotide-binding site within the larger of the two lobes in the yeast enzyme (13), there appears to be little sequence homology in this region among various hexokinases. In contrast, a predicted ATP-binding region (12), based on sequence homology with the catalytic site of CAMP-dependent protein kinases (14), lies in the smaller of the two lobes (15,16). This region includes Lys-111 in the yeast enzyme which has been reported to interact directly with an ATP affinity label (15). Also, the ATP analog, TNP-ATP,' has been shown to bind a 50-amino acid peptide of yeast hexokinase which includes Lys-111 (16). This "invariant" lysine residue has been found in all hexokinases sequenced to date (5,6,7,12) and corresponds to Lys-558 in the C-terminal half of the tumor enzyme.
In this report, we have used the technique of site-directed mutagenesis to evaluate the role of amino acid residues predicted to interact with either glucose or ATP in the C-terminal half of tumor hexokinase.

Site-directed Mutations in Hexokinase
EXPERIMENTAL PROCEDURES

Materials
Sources of restriction and DNA-modifying enzymes and other molecular biological chemicals have been previously described (5). Oligonucleotide primers (19-mers) for site-directed mutagenesis were synthesized in the protein/peptide/DNA facility of the Department of Biological Chemistry, Johns Hopkins University School of Medicine. The oligonucleotide-directed in vitro mutagenesis kit was from Amersham Corp.

Methods
Oligonucleotide-directed Mutagenesis of Tumor Hexokinase-Sitedirected mutagenesis was performed according to the method of Taylor et al. (17) using a kit from Amersham Corp. A 2.84-kilobase XbaI-XbaI fragment containing the entire coding region of tumor hexokinase and 50 base pairs of the 3'-untranslated region was subcloned into the polylinker region of M13mp19 at the XbaI site and was used as a template to make site-directed changes. For Ser-603, the 19-mer long mutagenic primer was 5'-CACCTTCGCGT-TTCCCTGC-3', and at the underlined base, codon TCG (Ser) was replaced with GCG (Ala). For Asp-657, the 19-mer primer was 5'-GGTCAACGC&CACCGTGGGC-3', and at the underlined bases, codon GAC (Asp) was changed to GCC (Ala) or GGC (Gly). For Glu-708, the 19-mer primer was 5'-CAATATGGC&ATGGGGGGCC-3', and at the underlined bases, codon GAA (Glu) was changed to GCA (Ala) or GGA (Gly). For Glu-742, the sequence of the 19-mer oligonucleotide was 5'-GGTTTGmGAAGATGATCAG-3', and at the underlined bases, codon GAG (Glu) was changed to GCG (Ala) or GGG (Gly). In the putative ATP-binding domain, Lys-558 was changed to Met or Arg. The 19-mer mutagenic primer was 5'-GCA-CAACAzGATCTACTCC-3', and at the underlined bases the codon AAG for Lys was altered to either ATG for Met or to AGG for Arg. A single oligonucleotide preparation was used to modify one residue to two different residues by adding 50% of both the bases at that particular position during synthesis of these oligonucleotides. Mutations were identified by sequencing of single strand DNA from M13 transformants. DNA sequencing was carried out by the dideoxy chain termination method of Sanger et al. (18) using Sequenase (version 2.0).
In order to transfer the mutants to the expression plasmid pKAHX (5) and to ensure that the site-directed mutations reported in this study were confined to the C-terminal half of the molecule, a replicating form DNA of M13 mutants was cut a t two unique restriction sites, NcoI (at base 1365) and StuI (at base 2553). The 1188-base pair Ncol-StuI fragment was gel-purified and then recloned into the plasmid pKAHX from which the wild-type NcoI-StuI fragment had been removed. The desired mutations in the expression plasmid were reconfirmed by DNA sequence analysis.
Recombinant DNA Procedures-Unless specified, standard molecular biological protocols described by Sambrook et al. (19) were used.
Overexpression and Partial Purification of Wild-type and Mutant Hexokinase Proteins-Wild-type and mutant hexokinase proteins were overproduced in Escherichia coli under the control of the alkaline phosphatase (phoA) promoter system utilized recently by Arora et al. (5). Using this overexpression system, it was observed that, in response to the stress of overproduction of foreign proteins, bacterial glucokinase was also induced.' Therefore, a procedure was developed to separate the overproduced tumor hexokinase from induced E. coli glucokinase. Briefly, the procedure involved was as follows. 250 ml of a 12-14-h culture overexpressing hexokinase protein was sedimented at 4,000 X g for 10 min and the pellet resuspended in 2.0 ml of lysis buffer (50 mM Tris-HC1, 25% sucrose (w/v), 1 mM EDTA, p H 8.0). T o this suspension, 0.5 ml of lysozyme (10 mg/ml) was added and incubated on ice for 30 min. This was followed by the addition of MgC12, MnCl', and DNase to final concentrations of 10 mM, 1 mM, and 10 wg/ml, respectively, followed by incubation at room temperature for 30 min. Then, 5.0 ml of detergent buffer (0.2 M NaC1, 1.0% sodium deoxycholate (w/v), 1.0% Nonidet P-40 (v/v), 20 mM Tris-HCI, 2 mM EDTA, pH 8.0) was added to the lysate and incubated at room temperature for 10 min. After centrifugation at 12,000 X g for 10 min a t 4 "C, the supernatant was removed and the pellet was completely resuspended in 6.0 ml of a buffer containing 50 mM Tris-HC1 (pH 8.0), 1 mM EDTA (pH 8.0) and 100 mM NaCl. This 'K. K. Arora, C. R. Filburn, and P. L. Pedersen, unpublished results.
suspension was again centrifuged a t 12,000 x g for 10 min a t 4 "C. This procedure of washing the pellet with Tris-HCl/EDTA/NaCl buffer was repeated once again. After this step, >99% of the stressinduced glucokinase activity was eliminated and most of the overexpressed tumor hexokinase was recovered in the "particulate" fraction.
Significantly, about 5% of hexokinase activity present in the particulate fraction could be solubilized by extraction with 0.5% Triton X-100 in 50 mM Tris-HC1 (pH 8.0), 1 mM EDTA, and 100 mM NaC1, and 95% of the tumor hexokinase protein and activity remained in the particulate fraction. The effect of site-directed mutations on hexokinase activity was tested using both the "Triton X-100 extract" and the detergent extracted particulate fractions. Both fractions yielded comparable results, but activity in the particulate fraction was higher in specific activity and much more stable, permitting storage for prolonged periods. Therefore, in this study results from the particulate fraction are reported.
Hexokinase Activity Assay-Hexokinase activity assays for both wild-type and mutant hexokinase protein preparations were performed spectrophotometrically essentially as described previously (2) using 2.3 mM glucose and 10 mM ATP in a coupled assay involving glucose 6-phosphate dehydrogenase and NADP'. The particulate fraction was assayed in the presence of 0.25% Lubrol WX. Hexokinase activities are presented as milliunits, defined as the formation of 1 nmol of NADPH/min. SDS-PAGE and Protein Determination-SDS-polyacrylamide gel electrophoresis was carried out in 1.5-mm-thick 10% acrylamide gels according to the method of Laemmli (20). Protein concentrations were determined by the bicinchoninic acid assay (Pierce Chemical Co.) using bovine serum albumin as a standard.

RESULTS
Site-directed Mutations of Ser-603, Asp-657, Glu-708, and Glu-742 in the Putative Glucose Binding Domain of Tumor Hexokinase-In order to evaluate the role of amino acid residues predicted to interact with glucose during the hexokinase-catalyzed reaction, Ser-603, Asp-657, Glu-708, and Glu-742 were mutated to either Ala or Gly, as described under "Methods." These latter amino acids cannot participate in hydrogen bonding to the hydroxyl groups of glucose. Kinetic analysis of overexpressed wild-type and mutant enzymes ( Table I, A) shows that all four site-directed mutations reduce hexokinase activity ( Vma.) by at least 90%. A single mutation of Asp-657 to Ala has the greatest effect, reducing the activity to <I% that of the wild-type enzyme. Changing Glu-708, Glu-742, or Ser-603 to Ala reduces V,,, by 10-15-fold or more relative to wild-type enzyme. Similar reductions in hexokinase activity are seen also when Glu-708, Glu-742, or Asp-657 are mutated to Gly (data not shown).
Apparent K,,, values for both hexokinase substrates were determined for wild-type and all mutant enzymes except Asp-657 + Ala, the activity of which was too low to do meaningful kinetics. These values reported also in Table I, A show that substitution of Ala for Glu-708 or Glu-742 produces a marked effect on the glucose concentration dependence for catalysis, which is reflected in 50-and 14-fold increases in the apparent K,,, for glucose, respectively. These mutations have no significant effect on the apparent K,,, for ATP ( Table I, A). In contrast, the substitution of Ser-603 for Ala decreases the apparent K,,, for glucose compared with the wild type by 5fold ( Table I, A). Again, there is no effect on apparent K,,, for ATP while Vmax is markedly decreased. The effect of Glu-708 and Glu-742 mutations is much more pronounced when the ratio of relative Vmax and apparent K,,, for glucose is compared with the wild-type enzyme (Table I, A). Site-directed Mutation of Lys-558 in the Putative ATPbinding Domain of Tumor Hexokinase-The ATP consensus sequence of protein kinases has a conserved lysine located

Effwt of si&-directed mutations within the C-trrminal half of tumor hexokinasr on its kinetic proprrtirs
Procedures for preparation of the wild-t.ype and mutant particulate fractions, and hexokinase activity determination are descrihed under "Methods." Kinetic parameters were determined using the Woolf-Augustinsson-Hofstee Plot ( u uersus u / S ) ( 2 5 ) . The values shown are averages of at least two independent experiments. Relative VmaX values were calculated assuming the rate for wild-type enzyme equal to 100. The average specific activity for the wild-t.ype enzyme was 205 nmol/min/mg protein and the percent deviation from the mean did not exceed 10%. am. aooarent.  11-14 residues "downstream" (C-terminal direction) from the glycine-rich core sequence (12,14). The "invariant" lysine corresponds to residue 558 of the tumor enzyme (5) and was changed in this study to Met in order to maintain the size and shape of Lys while removing its positive charge. Lys-558 was mutated also to Arg, which has the same charge as Lys but differs in side chain length. Vmnx values for Lys-558 + Argand Lys-558 4 Met enzymes are 70 and 29%, respectively, relative to that observed for the wild-type enzyme (Table I,  B). The substitution of Lys-558 to Met or Arg reduces hexokinase activity by reducing Vmax because in both cases, the apparent K , for either glucose or ATP is indistinguishable from that of the wild-type enzyme (Table I, B). These results suggest that the proposed Lys-558 in the predicted ATPbinding doamin may not have the crucial role in catalysis that homologous lysines in other "kinases" have (14). Alternatively, another lysine residue within the vicinity of Lys-558 may be performing a role in binding or catalysis. The other likely candidates may be Lys-549 or Lys-550. In future studies, these residues should be mutated and tested for their roles in catalysis.
Reduction of Hexokinase Activity Does Not Result from Proteolysis of the Mutant Protein-Following the observation that some of the site-directed changes, especially in the putative glucose-binding domain, result in marked reductions in hexokinase activity, it was tempting to conclude that residues Asp-657, Ser-603, Glu-708, and Glu-742 are essential either in the binding of glucose or its phosphorylation, or both. However, it remained possible that such mutations had destabilized the altered protein, resulting in its degradation via proteolysis. This question was addressed by examining the overexpressed protein products of wild-type and mutant genes. Fig. 1 shows a Coomassie-stained SDS-PAGE gel of these protein samples. All six mutant proteins (lanes 2-7) show a band near 100 kDa of the same intensity as that for the wild-type protein (lanes I and 8). Similar observations were made also for all the mutant proteins using Western blot analysis (results not shown), where the predominant immunoreactive band corresponds to a single species with an electrophoretic mobility comparablei with that of the wild-t-ype enzyme, as shown recently by Arora et al. ( 5 ) . The intensities of all the 100-kDa bands on the blots of comparably loaded gels are nearly equivalent, again suggesting that the observed decrease in hexokinase activities of mutant proteins does not result from proteolysis.
A closer examination of the Coomassie-stained gel (Fig. 1) and immunoblots (not shown) indicates that Asp-657 -. Ala and Glu-708 + Ala proteins run slightly faster than the other mutant and wild-type proteins. This anomalous behavior is not due to a deletion that may have occurred during the sitedirected mutagenesis procedure because a restriction map of the Asp-657 4 Gly mutant, which had minimal activity and the same size as Asp-657 + Ala is identical to that of the wild type. In addition, when an M13 transformant, which lacks the desired mutational change but retains a wild-t-ype sequence, is processed together with the mutant plaques, the resulting overexpressed protein exhibits hexokinase activity identical to that of the wild-type enzyme. Thus, manipulation of the DNA through the steps of the mutagenesis protocol does not produce artifacts that result in reduced activity.

DISCUSSION
Data presented in this paper demonstrate for the first time that all four conserved amino acid residues predicted from xray crystallographic studies of the yeast enzyme (1 1) to be within "contact" distance of glucose may play significant roles in binding and/or catalysis, but apparently in different ways. Catalytic activity is virtually abolished when Asp457 is changed to Ala, while Glu-708 + Ala and Glu-742 -. Ala enzymes retain as much as 10% of the wild-t-ype activity (Table I, A). Significantly, it has been suggested that the residue equivalent to Asp-657 in the yeast enzyme, i .~. Asp-211, is a catalytic base involved in the hexokinase reaction Site-directed Mutations in Hexokinase (10,21). The conversion of Ser-603 to Ala reduces the activity markedly, consistent also with the recent finding (22) that a glucose analog, N-(bromoacety1)-D-glucosamine, labels a region of rat brain hexokinase in the C-terminal half which includes the highly conserved residue Ser-603. The surprising decrease in apparent K, for glucose, but the corresponding and greater decrease in V,,,.,, indicate that Ser-603 influences both glucose binding and catalysis, perhaps by affecting either transfer of the phosphoryl group of ATP to glucose or release of the product Glc-6-P.
Based on x-ray crystallographic studies of yeast hexokinase (10, l l ) , there is a deep central cleft that divides the molecule into two lobes. Upon binding glucose, the smaller lobe rotates 12" relative to the larger lobe, thus partially closing the cleft. This conformational change, from "open" to "closed" forms, induced by glucose is considered to be a necessary step for catalysis. The resultant enzyme-glucose binary complex, upon binding ATP, is predicted to then form a ternary complex which presumably undergoes another conformational change before catalysis occurs (23). It is possible that in both the Glu-708 + Ala and Glu-742 +. Ala mutants the structure of the "open" form is altered such that the affinity for glucose is decreased, as suggested by the increased apparent K,,, values for glucose (Table I, A). Thus, the magnitude of the resulting glucose-induced conformational change may be insufficient to promote a "productive" reaction pathway.
Finally, it seems important to compare results reported here using site-directed mutations with those recently reported on the yeast enzyme in which random mutations were made (24). Of the random mutations made, none correspond t o residues mutated in this study, and of those residues that do affect activity, only Gly-235 and Asn-237 are predicted from studies of the crystal structure of a yeast enzyme-glucose analog complex to be within hydrogen bond distance of the sugar substrate (10). Certainly, additional studies will be necessary to better define the entire glucose-binding domain of hexokinase.