Glucokinase: Structural Analysis of a Protein Involved in Susceptibility to Diabetes*

The first step in the cellular metabolism of glucose in mamma- lian cells is the transfer of phosphate from ATP to the 6-hydroxyl group of the glucose molecule. This reaction is catalyzed by a family of structurally related hexose phosphotransferases that have dis-tinct sizes, kinetic properties, and tissue distributions (1). They can be divided into two general classes. The first class includes hexokinases I, 11, and 111. These enzymes have a size of 100 kDa and high affinity for glucose with K,, values of 20-130 p ~ . Their activities are feedback inhibited by physiological concentrations of glucose 6-phosphate. The second class is represented by hexokinase IV or glucokinase, which has a smaller size (50 kDa) and a K,,, for glucose of 5-8 mM that is -100-fold greater than those of the hexokinases (2). In addition, the activity of glucokinase is not inhibited by phys- iological concentrations of glucose 6-phosphate. Although the des-ignation “glucokinase” reflects well the physiological role of this enzyme, it has the same specificity as the low K,,, hexokinases. The major difference between glucokinase and the hexokinases is in their affinity for the hexose

none of which appears competent to encode a functional protein (12).
The physiological significance of these nonproductive transcripts with respect t o regulation of metabolism or other functions is uncertain.
Glucose transport does not appear t o be rate-limiting for glucose uptake by the liver and p-cells, and thus glucose phosphorylation, the first rate-limiting reaction in glycolysis, is the major factor in the regulation of glucose utilization in these tissues (14,15). The low affinity of glucokinase for glucose ensures that the rate of glucose phosphorylation will be directly proportional t o blood glucose levels and the cellular levels of glucokinase. The kinetic properties of glucokinase and the regulation of glucokinase levels by insulin and glucagon allow glucokinase t o assume a key role in regulating and integrating glucose metabolism in liver (15). In the p-cell, glucokinase functions as a glucose sensor, integrating blood glucose levels and insulin secretion (10). The threshold for glucosestimulated insulin release is about 5 mM glucose with a maximal response at 20 mM and a steep dose-response curve. However, the kinetics of glucose-stimulated insulin secretion cannot be easily related to the glucose concentration dependence of glucokinase, which has a K,, for glucose of 8 m M and also exhibits slow transition (hysteretic) and cooperative behavior, suggesting that glucose phosphorylation and usage may be necessary but not sufficient for insulin secretion (10).
The presence of tissue-specific promoters in the glucokinase gene allows the levels of glucokinase to be independently regulated in liver and p-cells. The most important physiologic regulators of rat hepatic glucokinase activity and gene expression are insulin and glucagon (via its second messenger CAMP); insulin increases and CAMP decreases gene transcription (2, 15,16). In contrast, insulin has no effect on glucokinase levels in 6-cells where regulation appears to be mediated by glucose, probably via a post-transcriptional mechanism (2, 10).
In addition t o transcriptional and post-transcriptional regulation, glucokinase activity may be controlled by other mechanisms. Long chain acyl-CoAs have been shown to be competitive inhibitors of rat liver glucokinase in vitro (171, but the physiological relevance of this observation is uncertain. Rat liver glucokinase has also been reported to be a substrate for CAMP-dependent protein kinase (18). However, there are no consensus phosphorylation sites in the rat liver enzyme. This and the small inhibitory effect of phosphorylation on activity make these studies difficult to interpret. Moreover, no acute effects of glucagon andor insulin administration on hepatic glucokinase activity have been reported under conditions where the phosphorylation state and activity of other enzymes, e.g. pyruvate kinase and 6-phosphofructo-2-kinase/fructose-2,6bisphosphatase, which are CAMP-dependent protein kinase substrates, were readily apparent.' Thus, it is unlikely that glucokinase is regulated by phosphorylation.
Glucokinase activity may also be regulated through interaction with other cellular proteins. Van Shaftingen and co-workers (19,201 have isolated and cloned a 62-kDa rat liver protein that inhibits hepatic glucokinase in the presence of fructose 6-phosphate. The inhibition of glucokinase activity by the "glucokinase regulatory protein" is characterized by an increase in K, for glucose with little change in V,,,,,. The effect of the regulatory protein and palmitoyl-CoA on glucokinase activity is competitive, suggesting that they bind to the same site (20). It has been postulated that this site is an allosteric one and that binding interferes with the glucose-induced conformational change that is essential for catalysis. Although modulation of glucokinase activity by the regulatory protein is an attractive hypothesis, its physiologic relevance is uncertain, particularly in p-cells where existence of the regulatory protein has yet t o be clearly demonstrated.
Mammalian glucokinase, but not the low K,, hexokinases, exhibits a sigmoidal dependence of activity on glucose concentration (21)   reflected in a Hill coefficient of about 1.7 at saturating ATP concentrations. Although there is some evidence for glucokinase monomer/monomer interactions (22), the consensus is that glucokinase is active as a monomer. The sigmoidal rate behavior has been attributed to either formation of a random ternary complex that may exhibit deviations from Michaelis-Menten kinetics even if the reaction is effectively ordered with respect to the net reaction rate (23) or to a "mnemonical" mechanism, which implies glucose binding differentially to two forms of the free enzyme that are not in equilibrium under steady-state conditions (21). A more general mechanism has also been proposed for glucokinase, based on the slow transition mechanism (24). The sigmoidal rate behavior is only observed below 5 m M glucose, and it is not clear whether it plays a physiologically significant role in regulation of glucose phosphorylation in liver or p-cells. The sigmoidal behavior, which has been observed in intact cells (25), may make the enzyme more sensitive to changes in glucose concentrations (10).

Mutations in Glucokinase Are Associated with Non-insulin-dependent Diabetes Mellitus
Recent studies have shown that mutations in glucokinase are associated with a form of non-insulin-dependent diabetes mellitus (NIDDM)' termed maturity onset diabetes of the young or MODY ( Fig. 1) (26)(27)(28)(29)(30). We have estimated that one in 2400 individuals in the United States is heterozygous for mutations in the glucokinase gene and as a consequence has an increased risk of developing diabetes (29). Mutations in this gene may be the most common cause of NIDDM identified to date and remain the only instance where the diabetic phenotype has been linked to a gene of glucose metabolism. Glucokinase mutations appear to cause diabetes mellitus by a gene dosage mechanism. Since all subjects with glucokinase mutations have one normal allele the cellular levels of glucokinase activity are 50% or more of normal levels depending upon the specific mutation present (27,28). The decreased levels of glucokinase activity in the p-cells are associated with an increase in the threshold for glucose-stimulated insulin secretion (31). They would also be expected to result in decreased glucose uptake and metabolism by liver with concomitant enhanced glucose production, and studies to test this prediction are in progress. The developing story of glucokinase deficiency diabetes provides a paradigm for future genetic, molecular, and clinical studies of the more frequently encountered forms of NIDDM.

Model for Structure of Human Glucokinase
Crystal structures are not available for glucokinase or any of the mammalian low K, hexokinases, but the A and B isozymes of yeast hexokinase have been studied by x-ray crystallography (32,33). Yeast hexokinase folds into two domains separated by a large cleft leading to the active site. The amino acid sequences of human glucokinase and yeast hexokinase are 31% identical, and residues in the active site and on the surface of the cleft separating the two * T h e abbreviations used are: NIDDM, non-insulin-dependent diabetes mellitus; MODY, maturity onset diabetes of the young. lobes are highly conserved (27,34). Thus, it was possible to build a molecular model of human glucokinase based on the crystal structure of yeast hexokinase B (34) (Fig. 2). This model suggests that the glucose-binding site is formed by glucokinase residues Lys-56, Asn-204, Asp-205, Gly-229, Asn-231, Glu-256, Gln-287, and Glu-290, which make interconnecting hydrogen bond interactions with each other and with the hydroxyls of glucose, while no interactions are predicted for the glucose ring oxygen atom 0-5 (Fig. 3).

StructureiFunction Studies
The hepatic and p-cell forms of glucokinase can be readily expressed in Escherichia coli and easily purified to homogeneity (27)(28)(29). This has facilitated analysis of the effects of mutations associated with NIDDM as well as various site-directed mutants on glucokinase activity. Mutations examined for their effect on the activity of rat liver or human p-cell glucokinase and site-directed mutations are listed in Table I, and the predicted locations of the human mutations in the protein is shown in Fig. 2 (lower panel ). Mutational studies of several glucose-binding residues (Asn-204, Asp-205, and Glu-256) are consistent with the prediction that these residues form hydrogen bonds with glucose since their mutation caused an increase in K, for glucose (Table I). Mutations of glucosebinding residues also caused a decrease in V, , , which is an expected result, since glucose phosphorylation is predicted to be tightly coupled to the substrate-induced change in the conformation of the enzyme (32). The effect of mutation of glucose-binding residues on V, , and K,,, for glucose depended on the nature of the specific mutation and was consistent with these residues hydrogen bonding to the glucose hydroxyls. For example, replacement of Asn-204 with Ser, a residue that can provide hydrogen bonds, resulted in only a small 1.5-fold increase in K,, whereas mutation to Ala resulted in a 30-fold increase in K, (Table I). In contrast to the relatively small effects (a factor of 10-40-fold) on V, , of mutations of glucose-binding residues, mutation of Asp-205, which is predicted to participate directly in catalysis, to Ala was associated with a 1000-fold decrease in V, , , (Table I and Ref. 35).
Glucokinase exhibits cooperative behavior a t glucose concentrations below 5 mM, and the native enzyme has a Hill coefficient of 1.7 (Table I). Only mutations of those residues that are predicted to form hydrogen bond interactions with glucose, e.g. Asn-204, Glu-290, and Glu-256, had an effect on the cooperative behavior of the enzyme. Single mutations of any of these residues completely abolished sigmoid behavior. This implies that the cooperative behavior of the enzyme depends on active site residue interactions with glucose. Interestingly, there is no cooperative behavior with fructose or mannose (22).
Although the determinants for sigmoid behavior and sugar binding to glucokinase are beginning to emerge, the location of the ATP-binding site and the identities of the residues involved in ATP binding are less clear. In the crystal structure of yeast hexokinase, the ATP-binding site was predicted to correspond to the position of a sulfate ion in the large domain (32)(33)(34). The sequences of hexokinase and glucokinase do not contain the consensus ATP-binding motif, GIOMXGK(T/S), nor does the enzyme possess the typical nucleotide binding fold that characterizes tfre ATP-binding site in protein kinases and dehydrogenases. A synthetic 50-amino acid peptide from yeast hexokinase has been reported to bind ATP (36). Glu are in red, other residues in blue, and glucose is in green. Hydrogen This peptide corresponds to glucokinase residues 70-120, which are located in the small domain. In fact, residues 80-83 form a turn at the surface of the cleft not far from the glucose-binding site and may interact with ATP in the closed conformation of the enzyme.
Final identification of the ATP-binding site will require detennination of the three-dimensional structure of glucokinase liganded to ATP andfor ADP.
The recent identification of mutations in glucokinase in patients with NIDDM has also helped to define functionally important residues and regions of the glucokinase molecule (27,28). As shown in Figs. 1 and 2, a preponderance of the mutations are found in exons 5,7, and 8. Mutations of residues that form the glucose-binding site (Val-203, Thr-228, Glu-256, Trp-257) including one predicted to be involved directly in glucose binding (Glu-256) and mutations of residues on the surface of the active site cleft (Gly-261, Glu-279, Gly-299, Glu-300) are associated with reduced enzyme activity (Table I).
The effects of each mutation on the activity and their predicted effects on the structure of glucokinase are consistent with conformationally driven catalysis involving changes in interactions throughout the glucokinase molecule. For example, substitution of Leu-309 by Pro also reduces glucokinase activity. Leu-309 is located in a surface helix far from the active site, and mutation to Pro probably disrupts the overall structure of the enzyme. In addition, mutations ofresidues (Ala-188, Gly-175,Val-182) in the smaller ofthe twolobes of the glucokinase molecule identified a region that is postulated to be involved in the glucose-induced conformational change (27,28). In addition to mutations associated with NIDDM, two amino acid polymorphisms (Asp-4 + Asn and Ala-11 + Thr) have been identified in P-cell glucokinase (28). These polymorphisms are predicted to have no effect on enzymatic activity, an expectation that has been confirmed directly for the Asp-4 --$ Asn substitution.

Future Directions and Concluding Remarks
The discovery that mutations in glucokinase can cause diabetes mellitus has stimulated renewed interest in this enzyme and underscored the critical role played by glucokinase in whole body glucose homeostasis, including regulation of insulin secretion and uptake and metabolism of glucose by the liver. The cis regulatory elements and associated proteins that mediate the effects of insulin and C A M P on gene transcription in liver and the mechanism of the glucose effect on gene expression in p-cells are not known and should, along with the search for the basis of the gene's tissue-specifk expression, be fertile grounds for future research. Likewise, the effect of glucose metabolism on insulin secretion and gene expres-

Minireview: Structural Analysis of Glucokinase
Enzymatic properties of natiue and mutant forms of human Table I p-cell glucokinase Native and mutant forms of human /3-cell glucokinase were expressed in E. coli. Glucokinase activity was determined as described previously (27,28). The values of K for glucose and V, , shown were obtained by fitting the velocity versus subsrrate concentrations m the presence of 5 m~ ATP to a Michaelis-Menten equation by using Sigma Plot. The Hill coefflcient was obtained by fitting the velocity versus substrate concentration curve to V = V, X (K + S"), where n is the Hill coefficient. The K for glucose was lower with t%e latrer equation, but the relative changes were & t h e same. + indicates K and V, , values that are significantly different from the native p-cell enzyme. ?he data are from Refs. 26  sion in pancreatic p-cells are also important areas requiring further investigation.
The availability of a molecular model for human glucokinase has been extremely useful in understanding the effects of human glucokinase mutations on enzymatic activity. Moreover, many of the missense mutations identified in subjects with MODY have provided insights into structure/function relationships of the human glucokinase molecule that would not have been realized by the usual rationales for making mutations based entirely on homology with yeast hexokinase. For example, Ser-131, Gly-175, Val-182, and Leu-309 are located far from the active site, but nonetheless mutation of these residues affects activity and causes diabetes. The characterization of mutations associated with NIDDM as well as specific site-directed mutations has revealed that conformationally driven catalysis exemplified by glucokinase is complex in nature and involves substrate-induced changes in structure throughout the molecule that cannot be understood by studying only those residues involved in glucose binding.
The structural basis for the higher affinity for sugar of hexokinases compared with glucokinase is not known, even though the residues that are predicted to bind glucose are conserved in these proteins. The duplicated NH, and COOH halves (of hexokinase I have been expressed separately in E. coli (37).' Th K,,, for glucose of the COOH-terminal half-molecule was identica \ to that of the intact enzyme and was noncompetitively inhibited by glucose 6-phosphate. In contrast, the NH,-terminal half-molecule was re-ported to be inactive (37).' These findings indicate that the two half-molecules are not identical and that the glucose 6-phosphate inhibitory site was not generated by gene duplication and mutation of active site residues in the NH,-terminal segment. These results also suggest that it is not the size of the protein per se that determines the lower affinity since the COOH-terminal half-molecule has a low affinity for glucose as does the 50-kDa yeast hexokinase. It may be possible to use the model for glucokinase together with site-directed mutagenesis to generate a glucokinase molecule with a higher affinity for glucose. Ultimately, the elucidation of the crystal structure of glucokinase will provide answers to this and other questions about this key regulatory enzyme of glycolysis.