Structure/function studies of human beta-cell glucokinase. Enzymatic properties of a sequence polymorphism, mutations associated with diabetes, and other site-directed mutants.

Glucokinase plays a key role in the regulation of glucose metabolism in insulin-secreting pancreatic beta-cells and in the liver. Recent studies have shown that mutations in this enzyme can lead to the development of a form of non-insulin-dependent diabetes mellitus that is characterized by an autosomal dominant mode of inheritance and onset during childhood. Here, we report the catalytic properties of five additional missense mutations associated with diabetes (Glu70-->Lys, Ser131-->Pro, Ala188-->Thr, Trp257-->Arg and Lys414-->Glu), one polymorphism present in both normal and diabetic subjects (Asp4-->Asn), and three site-directed mutations (Glu177-->Lys, Glu256-->Ala, and Lys414-->Ala). The Trp257-->Arg mutation generated an enzyme that had an activity that was less than 0.5% of that for native human beta-cell glucokinase. By contrast, the Glu70-->Lys, Ser131-->Pro, Ala188-->Thr, and Lys414-->Glu mutations had a Vmax that was 20-100% of normal but a Km for glucose that was 8-14-fold greater than the native enzyme. There was no effect of the Asp4-->Asn polymorphism or the Glu177-->Lys substitution on glucokinase activity. The Lys414-->Ala substitution had no effect on Vmax but increased the Km for glucose 2-fold and the Glu256-->Ala substitution caused a approximately 200-fold decrease in Vmax. These studies have led to the identification of additional residues involved in glucokinase catalysis and substrate binding.

Glucokinase plays a key role in the regulation of glucose metabolism in insulin-secreting pancreatic p-cells and in the liver. Recent studies have shown that mutations in this enzyme can lead to the development of a form of non-insulin-dependent diabetes mellitus that is characterized by an autosomal dominant mode of inheritance and onset during childhood. Here, we report the catalytic properties of five additional missense mutations associated with diabetes (G1u70 -+ Lys, Serlsl -+ Pro, Ala1--+ Thr, Trp267 -+ Arg and Lys414 -+ Glu), one polymorphism present in both normal and diabetic subjects (Asp4 -* Asn), and three site-directed mutations -t Lys, Glu2" -*Ala, and Lys414 + Ala). The Trp267 -+ Arg mutation generated an enzyme that had an activity that was less than 0.6% of that for native human P-cell glucokinase. By contrast, the Glu70 + Lys, Serlsl -+ Pro, Ala188+Thr, and Lys414 -+ Glu mutations had a V , , that was 20-1W0 of normal but a K, for glucose that was 8-14-fold greater than the native enzyme. There was no effect of the Asp4 -+ Asn polymorphism or the + Lys substitution on glucokinase activity. The Lys414 -+ Ala substitution had no effect on V , , but increased the K, for glucose 2-fold and the G1u256 -+ Ala substitution caused a -200-fold decrease in Vma. These studies have led to the identification of additional residues involved in glucokinase catalysis and substrate binding.
Glucokinase (ATP:D-hexose 6-phosphotransferase, EC 2.7.1.2.) or hexokinase type IV catalyzes the phosphorylation of ical Institute, National Institutes of Health Grants DK-20595, DK-* These studies were supported in part by the Howard Hughes Med-38354, and DK-44840, National Science Foundation Grant DBM-8608989, Juvenile Diabetes Foundation International, Association Francaise contre les Myopathies (through the Genethon program), Assistance Publique-Hopitaw de Paris, the French Ministry for Research and Technology, Japanese Ministry of Education, Science and Culture (Scientific Research Grants 02304034 and 03671130, and a fellowship from the Deutsche ForschungsgemeinschaR (to M. S.) The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "adverthis fact.
tisement" in accordance with 18 U.S.C. Section 1734 solely to indicate $5 To whom correspondence and reprint requests should be addressed Dept. of Physiology and Biophysics, Health Sciences Center, Level 6, Rm. 140, SUNY, Stony Brook, NY 11794-8661. Tel.: 516-444-2287;Fax: 516-444-3432. glucose, the first rate-limiting reaction in glycolysis (1)(2)(3). It is expressed in the pancreatic p-cells and liver and is readily distinguished from other hexokinases in mammalian cells by its smaller size (50 ma), an affinity for glucose ( K , = 8 m) that is lower than other hexokinase isozymes and is in the physiological range of plasma glucose levels, and a relative lack of inhibition by glucose 6-phosphate. These properties ensure a gradient for glucose entry into the hepatocyte and p-cells, especially following a meal when plasma glucose levels are elevated (3).
Recent studies have shown that mutations in glucokinase can lead to the development of an autosomal dominant form of NIDDM' that has an onset in childhood (4)(5)(6)(7)(8)(9)(10). Clinical studies suggest that the threshold for glucose-stimulated insulin secretion in subjects with glucokinase mutations is increased, implying that the mutations affect glucose sensing by the pancreatic p-cells (11). Twenty-three different mutations have been described that are associated with diabetes, of which 16 are missense mutations, four are nonsense mutations, and three are splicing mutations (Fig. 1). We have previously reported the effects of 11 missense mutations on the enzymatic properties of human p-cell glucokinase, and all are associated with a decrease in V, , and/or in affinity for glucose (7). These mutations were in regions of the protein encoded by exons 5-8, and molecular modeling predicts that they are in the active site cleft or surface loops leading into this cleft as well as in a region of the smaller of the two globular domains of the glucokinase molecule that is believed to undergo a substrate-induced conformation change on glucose binding (12-15). Here, we describe the catalytic properties of five additional missense mutations associated with glucokinase-deficient diabetes (G1u70 -j Lys, Ser131 + Pro, Mals8 + Thr, T r p 2 5 7 -j Arg and Lys414 + Glu) that are encoded by exons 2, 4, 5, 7, and 9, respectively, one polymorphism (Asp4 + Asn) that is encoded by exon la and thus is present only in the p-cell form of glucokinase, and which is not associated with diabetes, and three site-directed mutants ( G l~l~~ + Lys, G~u~~~ + Ala and Lys414 + Ala). The results of these studies define additional key residues involved in glucokinase activity.
The abbreviations used are: NIDDM, non-insulin-dependent diabetes mellitus; MODY, maturity-onset diabetes of the young; PMSF, phenylmethylsulfonyl fluoride; DIT, dithiothreitol; kb, kilobase(s); PAGE, polyacrylamide gel electrophoresis; FPLC, fast pressure liquid chromatography. FIG. 1. Schematic structure of the h u m a n glucokinase gene and localization of mutations identified in patients with diabetes. The p-cell isoform of glucokinase is encoded by exons l a and 2-10 (6). In this report, amino acids are designated relative to the sequence of human @-cell glucokinase. Single-letter codes for amino acids are used in this figure. The mutations reported in this paper that were expressed in E. coli and characterized kinetically (Table 11) are in boldface type and include those from French (Glu7" -Lys, Lys414 -f Glu (18) and EXPERIMENTAL PROCEDURES Materials-Mono Q-Sepharose Fast Flow and Sepharose were from Pharmacia LKB Biotechnology Inc. Glucose-6-phosphate dehydrogenase and glucosamine were from Boehringer Mannheim. Glucosamine-Sepharose was prepared as described (18).
Construction ofpET Expression Plasmids for Human P-CeU Ghcokinuse-A 2.6-kb human pancreatic p-cell glucokinase cDNA clone, ph- . was used to generate the construct pEhgk-WT. An NdeI site was generated a t the 5' end using polymerase chain reaction with an oligonucleotide that has a 1-base pair mismatch ( G G C TGG TGT GCA TAT GCT GGA CGA CAG). The insert in the PET 3a expression construct included the protein coding region of the cDNA as well as the 3'-untranslated region.
Site-directed Mutagenesis of Human p-Cell Glucokinase-In vitro mutagenesis of human p-cell glucokinase was camed out using the Altered Sites In Vitro Mutagenesis System (Promega, Madison, WI). Mutations in the native p-cell glucokinase expression construct were generated by replacing appropriate restriction fragments with the corresponding fragment containing the mutation from the mutagenesis vector. A SacII-BamHI restriction fragment was used for mutations Glu70 -+ Lys and SerI3I -Pro. GIU"~ -Lys, Ala18R + Thr, G h P 6 -, Ala, Trp257 .--) Arg, Lys414 + Ala, and Lys4I4 + Gly were generated by using the ClaI-BamHI sites. An NdeI-BamHI fragment from the mutagenesis vector was used to replace the corresponding fragment from the native p-cell construct to generate the Asp4 -Asn mutant. All mutations were confirmed by DNA sequencing.
Bacterial Expression and Purification ofNative and Mutant Forms of Human p-Cell Glucokinase-Native and mutant human glucokinases were expressed in Escherichia coli using the PET expression system essentially as described previously (7,17,18). Glucokinase was purified 20-fold from extracts of E. coli in four steps, including (NH4),S04 precipitation (4545%). gel filtration on a Sephadex G-100 column, chromatography on a glucosamine-Sepharose column, and FPLC Mono Q-Sepharose chromatography. The latter step was necessary to remove a small amount of low K,,, hexokinase activity in the cell extract that could confound kinetic studies. The purified protein was homogeneous as judged by SDS-PAGE. Native human p-cell glucokinase had a specific activity of about 100 unitdmg, which is similar to that reported for the purified rat liver enzyme (19). The purified recombinant protein was also subjected to NH,-terminal sequence analysis. The protein, analyzed through 12 cycles, gave the expected sequence: Met-Leu-Asp-Asp-

Arg-Ala-Arg-Met-Glu-Ala-Ala-Lys-.
Modeling of Human P-CeU Glucokinase Structure-A model of human p-cell glucokinase was generated by analogy to the known crystal structure of the related yeast hexokinase B (12,13, 1512. The model was built and examined using the program FRODO (20) run on an Evans and Sutherland ESV 10 computer graphics workstation.
Enzyme Assay and Kinetic Analysis-Glucokinase activity and calculations of V , , and K,,, for substrates were performed as described previously (7,17,18).

Expression in E. coli and Purification of Native and Mutant
Forms of Glucokinase-The human p-cell form of glucokinase was expressed in E. coli using the T7 RNA polymerase-based system (21). Easily detectable levels of glucokinase activity * I. Weber, S. J. Pilkis, and R. Hamson, manuscript in preparation.
were found in the soluble fraction of cell extracts after 15 min of induction of expression with isopropyl P-D-thiogalactopyranoside, and activity increased nearly 50-fold after 2 h (data not shown). The native and mutant forms of p-cell glucokinase were expressed a t similar levels (Fig. 2). Like the native enzyme, all the mutants were recovered in the soluble fraction and could be purified to homogeneity. Native human p-cell glucokinase and various mutant forms were purified by a modification of the method ofAndreone et al. (19). Resuspended (NH4)2S04 precipitates (4545%) were gelfiltered on a Sephadex G-100 column, then chromatographed on a glucosamine-Sepharose column (Fig. 3A). After elution of the enzyme from the glucosamine-Sepharose column by 80 m M KC1 and 0.5 M glucose, there were two remaining proteins, which could be separated by FPLC Q-Sepharose column chromatography with a linear KC1 gradient (Fig. 3B). The glucokinase eluted at the higher salt concentration and was homogeneous as judged by the criterion of SDS-PAGE (Fig. 3B). The mutants also bound to the glucosamine-Sepharose and could be purified to near homogeneity by adjusting the KC1 concentration during elution. The purification scheme for the native enzyme is shown in Table I.
Effects of Mutations, Polymorphism, and Site-directed Mutations on Glucokinase Activity-Native recombinant human /3-cell glucokinase had a V,,, of 93 2 8 unitdmg and K , values for glucose and ATP of 8.0 2 2 and 0.20 2 0.10 mM, respectively (Table 11). These values are very similar to those reported for the rat liver form of this enzyme (18,19). The activity of the Trp257 + Arg enzyme was severely impaired and was less than 0.5% that of the native enzyme. The Glu70 + Lys, Ser131 + Pro, Mals8+ Thr, and Lys4I4 + Glu mutations had only a relatively small effect on V, , , and Lys414 + Glu had no effect. However, these mutations all caused large increases in the K , for glucose (8-14-fold). In addition, the Ser131 + Pro mutation had a significantly increased affinity for ATP. None of the other mutations caused a significant change in the K , for ATP. Thus, as noted previously for other missense mutations associated with diabetes, the mutations described here affect V, , andor K , for glucose of the enzyme (17). The polymorphism Asp4 + Asn found in one French family (8) had no effect on activity (Table  11). Lane A of the inset shows SDS-PAGE of the applied protein sample. Lane B shows SDS-PAGE of the pooled fractions containing glucokinase activity and revealed two protein bands, one with a molecular mass of 50 kDa, which corresponds to glucokinase. B, FPLC-Sepharose chromatography of p-cell glucokinase. The sample containing glucokinase activity from the glucosamine-Sepharose column was loaded onto a FPLC Mono Q column a t a flow rate of 0.5 mumin and washed with 5 ml of glucokinase buffer at I mumin. Glucokinase was eluted after 30 min with a salt gradient of 100-600 mM KC1 a t a flow rate of 1 mumin. The inset shows SDS-PAGE electrophoresis of molecular size standards and of the pooled fractions containing glucokinase activity. The purified protein had a subunit molecular mass of 50 kDa and was homogeneous.
We also examined the effects of three site-directed mutations on glucokinase activity (Table 11). GluZ5(j is predicted to form hydrogen bonds with the hydroxyls of glucose. The + Lys mutation identified in a subject with diabetes (7) was inactive as was the site-directed mutant G~u~~~ + Ala. This result indicates the importance of G~u~~~ for enzymatic activity. We generated the Lys414 + Ala mutant to test whether or not the substitution of Lys414 per se was responsible for the abnormal activity of the Lys414 + Glu mutation or if the decreased affinity of this enzyme for glucose was due to the introduction of a negatively charged amino acid. The Lys414 + Ala mutant   exhibited a small although significant increase in K, for glucose with no change in V, , or in K,,, for ATP, suggesting that the introduction of a negatively charged amino acid at this site was primarily responsible for the decreased affinity for glucose of the Lys414 + Glu mutation. The mutations Gly175 + Arg, + Met, and Ala188 + Thr are all associated with a modest reduction in V , , and an increase in K, for glucose. To assess the effect of other mutations in this region of the glucokinase molecule on its enzymatic activity, we generated the + Lys mutation. This mutant enzyme had normal activity, implying that not all residues in this region of the molecule are critical for catalysis and/or substrate binding.
Yeast Hexokinase B Crystal Structure as a Model for Human Glucokinase-The crystal structure of yeast hexokinase B, an enzyme in which 140 (-30% identity) of the residues are identical with those in human p-cell glucokinase (71, is known (12-15). We have used the structure of yeast hexokinase B to predict the structure of human p-cell glucokinase (6, 7)' by assuming that glucokinase will show the same arrangement of a-helices and p-strands as observed in the crystal structure of yeast hexokinase B. The locations of the missense mutations in human glucokinase that are associated with diabetes, including the five characterized in this report, are indicated on the ribbon a-carbon backbone drawing shown in Fig. 4. This model is a representation of the open conformation of glucokinase, which consists of small and large domains separated by a deep cleft.
Based on predictions from the glucokinase model derived from the open structure of hexokinase B, two of the mutations, Trp257 + Arg and Lys414 + Glu involve residues that are adjacent to the region of the active site. Trp257 is next to G~u '~~, which is predicted to form hydrogen bond interactions with glucose. Mutation of G~u~~~ to Lys has been reported in a MODY pedigree, and this mutant enzyme has been shown to be inactive, consistent with its role as a glucose-binding residue (12). Site-directed mutation of Glu256 to Ala also resulted in an inactive enzyme (Table 11). Based on its proximity to such an important glucose binding residue, mutation of Trp257 to Arg would be expected to decrease enzyme activity. This residue is W 7 O in yeast hexokinase B, and it forms a part of the internal hydrophobic core of the protein structure. As such, substitution to the positively charged Arg would be predicted to be unfavorable. The large decrease in V , , of the Trp257 + Arg mutation is completely consistent with this proposed structural alteration of the glucose-binding site.
It seems paradoxical that mutations in glucose binding residues (GluZ5(j) or mutations of adjacent residues (Trp257) result in dramatic decreases in V, , but no change in apparent affin-  ity for glucose as measured by K , for glucose. However, since the glucose-induced conformational change and cleft closure are essential for catalysis (141, it is likely that the K,,, for glucose reflects a large amount of nonproductive glucose binding. If the glucose-induced conformational change is compromised, a decrease in V, , would follow. Consistent with these notions, all the mutants bound to a glucosamine column. Based on analogy to the heat shock protein, actin, and sugar kinase structures (22-241, Lys414, which corresponds to Thr422 in the yeast hexokinase B structure, is in a region that has been predicted to represent a common ATP binding domain. In this common domain the phosphate tail of ATP is bound by residues on two P-hairpins, one from each of two subdomains, and by other segments. However, there is at present no experimental evidence for hexokinase to relate the binding of ATP to any specific residues in this domain. Lys414 is actually located in a region postulated to bind the adenosine ring, and this residue is not conserved in this family ofATP-binding proteins (24). The Lys414 Glu mutation resulted in a 10-fold increase in K , for glucose but with no change in K , for ATP or maximal velocity (Table II), and the Lys414 -Ala mutation showed a much smaller increase in K, for glucose, implying that the introduction of a negatively charged residue at this site was primarily responsible for the decreased affinity for glucose. W e previously studied the effects of mutation of T h r Z z 8 , another residue predicted to be involved in ATP binding (51, on glucokinase activity (7). The ThrZz8 -Met mutation has a greatly decreased V, , but no change in affinity for ATP. The results from mutational analysis of both ThrzZs and Lys414 suggest that they do not play a critical role in ATP binding in glucokinase. Both Lys414 and Thr228 are, however, directed toward the active site (Fig. 4), and it is thus not surprising that their substitution affects enzyme activity. Alternatively, the analogous residues in yeast hexakinase B may bind ATP but have another function in glucokinase. Final resolution of this question will require the x-ray crystal structure of glucokinase.
Two other mutations affect residues that could be important for maintaining the correct conformation of the molecule. Malss corresponds to GlnIg5 in yeast hexokinase a-helix 4, which has loops at either end that connect to active site residues L y P 9 and Aspzo". Aspzo5 has been shown to act as a base catalyst in the phosphorylation of glucose (18). Since Malss is partly internal, the larger Thr residue may cause a small distortion of the structure that possibly could be transmitted to the active site residues L y P 9 and Aspzo5 and thereby affect catalysis. The presence of Thr at position 188 also provides the possibility of new hydrogen bond interactions with the nearby side chains of Asp124 and Ser127 in a-helix 3 and the carbonyl oxygens of residues 184 and 185. This might preferentially stabilize one conformation in this region and thus be unfavorable for the conformation change that occurs on glucose binding. Consistent with these predictions the Mals8 -Thr mutation resulted in a decrease in the affinity for glucose and also caused a 50% reduction in Vma.
We previously described the enzymatic properties of two other missense mutations in the region of AlalBS, Gly175 + Arg and -Met (7). To assess the possible effects of other mutations in the region of residues 175-188 on enzymatic activity, we mutated G~u~~~ to Lys. This mutation had no effect on glucokinase activity. The model for the structure of glucokinase predicts that G~u '~~ is a surface residue and substitution with Lys would not be expected to create any steric or electrostatic problems, a prediction consistent with the experimental results.
Ser131 is located in 0-helix 3 in the smaller of the two domains and is far from the active site. The mutation to Pro is predicted to disrupt the a-helix. Glucokinase residue 131 is adjacent to PheIg5, which connects via p-strand 8 to the base catalyst Aspzo5. Therefore changes at residues 131 can easily be transmitted to the active site. This could be the basis for the large change in K, for glucose (14-fold) and the decrease in v, , , observed in this mutation.
The conserved Glu7' lies close to the conserved basic residues Lys458 and Lys459 in the yeast hexokinase B structure, Mutation of Glu7* + Lys would be expected to be electrostatically unfavorable due to the other nearby positively charged residues. Furthermore, Glu70 forms a hydrogen bond interaction with the amide of residue 67, stabilizing the turn and p-strand 2; this interaction would not be possible for Lys. This region is predicted to participate in the substrate-induced conformational change that results in cleft closure so that the destabilizing Glu70 + Lys mutation is probably responsible for the significant changes in enzyme activity observed (Table 11). However, the precise molecular mechanism involved in the increased K, for glucose exhibited by the Gh70 + Lys mutation is not clear, but may reflect the necessity for high glucose concentrations to affect the conformational changdcleft closure.
The kinetic properties of the Asp4 + Asn mutant were unchanged from the native enzyme. This finding is not surprising since Asp4 is at the NHz terminus far from the active site in the region of the hexokinase crystal structure that is not visible and presumed to be disordered and flexible (12-15). The lack of effect of this mutation on enzyme activity is consistent with the fact that it is found in both affected and unaffected members of a family with glucokinase-deficient disorder, but did not segregate with the diabetic phenotype (8). CONCLUSIONS The recent demonstration that mutations in glucokinase can contribute to the development of NIDDM has provided new insight into the etiology of this genetically heterogeneous disorder (4)(5)(6)(7)(8)(9)(10). It has also led to the identification of amino acid residues that are important for glucokinase activity. We have previously proposed that mutations in glucokinase cause an autosomal dominant disorder of glucose metabolism by a gene dosage mechanism since splicing, nonsense, and missense mutations have been identified in patients with diabetes (4)(5)(6)(7)(8). Moreover, since glucokinase is a monomer, it is unlikely that mutations have a dominant-negative effect on enzymatic activity. In support of this hypothesis, no effect on V, , , or on K, for glucose was observed when a n equal amount of native glucokinase was incubated with a "dead" mutant (Trp2s7 + k g ) or with SerI3' + Pro (data not shown). The decreased cellular levels of glucokinase activity in pancreatic @cells are predicted to alter glucose sensing by these cells and thereby increase the threshold for glucose-induced insulin secretion. The altered enzymatic activity of the missense mutations associated with glucokinase-deficient diabetes described in this report is consist-ent with this hypothesis. However, it is also possible that some mutations may affect glucose sensing at the level of the glucose transporter (5) and/or association with the glucokinase regulatory protein (26). Finally, the characterization of the enzymatic properties of these mutations provides a n opportunity to examine structurdfunction relationships in glucokinase. A model for human glucokinase based on the structure of yeast hexokinase B has been particularly useful in this regard (5, 7).2 It is noteworthy that all 16 observed missense mutations associated with the diabetic phenotype are located in regions of conserved sequence in the two enzymes. The strong correlation between the observed effects of changes in amino acid sequence on enzymatic activity and those predicted from structural considerations indicates the utility of this model for considering the effects of amino acid replacements on structurdfunction relationships.