Functional Interaction between the N- and C-terminal Halves of Human Hexokinase II*

Mammalian hexokinases (HKs) I–III are composed of two highly homologous ∼50-kDa halves. Studies of HKI indicate that the C-terminal half of the molecule is active and is sensitive to inhibition by glucose 6-phosphate (G6P), whereas the N-terminal half binds G6P but is devoid of catalytic activity. In contrast, both the N- and C-terminal halves of HKII (N-HKII and C-HKII, respectively) are catalytically active, and when expressed as discrete proteins both are inhibited by G6P. However, C-HKII has a significantly higherK i for G6P (K i G6P) than N-HKII. We here address the question of whether the highK i G6P of the C-terminal half (C-half) of HKII is decreased by interaction with the N-terminal half (N-half) in the context of the intact enzyme. A chimeric protein consisting of the N-half of HKI and the C-half of HKII was prepared. Because the N-half of HKI is unable to phosphorylate glucose, the catalytic activity of this chimeric enzyme depends entirely on the C-HKII component. The K i G6P of this chimeric enzyme is similar to that of HKI and is significantly lower than that of C-HKII. When a conserved amino acid (Asp209) required for glucose binding is mutated in the N-half of this chimeric protein, a significantly higher K i G6P (similar to that of C-HKII) is observed. However, mutation of a second conserved amino acid (Ser155), also involved in catalysis but not required for glucose binding, does not increase theK i G6P of the chimeric enzyme. This resembles the behavior of HKII, in which a D209A mutation results in an increase in the K i G6P of the enzyme, whereas a S155A mutation does not. These results suggest an interaction in which glucose binding by the N-half causes the activity of the C-half to be regulated by significantly lower concentrations of G6P.

6-phosphate (G6P). Type I, II, and III isozymes of mammalian HKs are ϳ100 kDa in size and are inhibited by the reaction product, G6P. Type IV HK (also known as glucokinase) is similar to yeast HK in that it has a relative mass of ϳ50 kDa and is insensitive to inhibition by physiologic concentrations of G6P. The deduced amino acid sequences of HKI-III reveal internal similarities between their N-and C-terminal halves and between each of these and yeast HK and glucokinase (1). This observation supports the hypothesis that the 100-kDa mammalian HKs evolved from the duplication and fusion of an ancestral 50-kDa HK (1,2).
The N-and C-terminal halves of HKI show a marked functional difference despite the similarity of their amino acid sequences. The C-terminal half of HKI is catalytically active, whereas the N-terminal half is inactive (3)(4)(5). Whereas G6P binds to both halves of HKI (3,5), the G6P regulatory site of HKI is thought to be in the N-half of the intact enzyme, and the C-half binding site is latent (1,6). In contrast, we have shown that both the N-and C-terminal halves of human and rat HKII (N-HKII and C-HKII, respectively) have catalytic activity and that each is inhibited by G6P (7). However, the N-and Cterminal halves of HKII possess different kinetic characteristics: N-HKII has a slightly higher affinity for ATP than does C-HKII, and its K i for G6P is about 20 -30-fold lower than that of C-HKII (7). Furthermore, the K iG6P value of the intact 100-kDa HKII enzyme is similar to that of N-HKII (7), an observation confirmed by Tsai and Wilson (8).
Based on the crystal structure of yeast HK (9 -12) and recently confirmed by a similar analysis of mammalian HKI (13)(14)(15)(16), glucose is thought to bind to a cleft in the open conformation of the enzyme. This binding results in closure of the cleft and initiates the catalytic sequence. Asp 211 is an important residue in yeast hexokinase. As the base catalyst, it binds the 6-hydroxyl group of glucose, and this receives the phosphate moiety of ATP. Mutation of the corresponding residue to alanine in glucokinase (D205A) (17), the C-half of HKI (D209A) (18), and the N-and C-halves of HKII (D209A and D657A) (7) results in a complete loss of activity. In contrast, the carbonyl group of Ser 158 in yeast HK interacts with the 3-hydroxyl group of glucose only in the closed conformation (1). Mutation of the corresponding amino acid to alanine in the C-half of HKI (S155A) (5) and the N-and C-halves of HKII in the intact enzyme (S155A and S603A) reduces the V max of the enzyme to about 10% of that of the wild type (8). 2 Thus, although the D209A mutation in mammalian HKs results in a loss of glucose binding and catalytic activity, the S158A mutation maintains glucose binding but reduces the V max .
Because the K iG6P of HKII is similar to that of N-HKII and significantly different from that of C-HKII, we focused the present analysis on how the activity of C-HKII is regulated by G6P in the intact enzyme. Mutations that have different effects on glucose binding in the N-half of HKII result in different valves of K iG6P . We here show that the regulation of the C-half by G6P in the intact HKII is influenced by the N-half of the molecule.

EXPERIMENTAL PROCEDURES
Construction of NICII Chimeric Proteins-The human hexokinase I cDNA was provided by Dr. Graeme Bell (Howard Hughes Medical Institute, University of Chicago). A DNA fragment that encodes the N-half of HKI was prepared by PCR using B2392 (5Ј-GCCAGCAT-GATCGCCGCGCA-3Ј) as the 5Ј oligonucleotide and A2670 (5Ј-TTACT-GCCGGTGCTGCTGGG-3Ј) as the 3Ј oligonucleotide. A2670 contains a stop codon at position 1407, which is the putative fusion site between the N-and C-terminal halves of HKI. The pGef plasmid (7), constructed by insertion of a bacteriophage f1 origin into the pGex-3x vector (Amersham Pharmacia Biotech), was linearized at a SmaI restriction site located adjacent to the glutathione S-transferase (GST) open reading frame. The PCR product that contains the N-half of HKI was then ligated into pGef to obtain pGST⅐N-HKI. This plasmid encodes amino acids 1-469 of human HKI fused to the C terminus of GST.
An NcoI site is conveniently located near the fusion site of the N-and C-terminal halves of both human HKI and HKII. NcoI cleaves at nucleotide 1366 in the open reading frame of both HKI and HKII. pGST⅐N-HKI was digested with NcoI and BssHII (a cleavage site about 1.2 kbp 5Ј of the GST start codon) to produce a fragment of ϳ3.2 kbp that contains the coding sequence of GST and the N-terminal half of HKI. This fragment was purified by size separation through agarose gel electrophoresis and was isolated using Spin-x tubes (Costar, Cambridge, MA). A GST⅐HKII expression plasmid, pGST⅐HKII (7), was also digested with NcoI and BssHII, and the larger fragment of about 5.2 kbp was isolated. These two fragments were ligated together to generate the plasmid pGST⅐NHKI-CHKII, which encodes the chimeric protein GST⅐NICII.
PCR-based Mutagenesis of Hexokinase Constructs-Site-directed mutations were generated by two separate PCR reactions. The primers, templates, and PCR conditions used for the construction of the hexokinase mutants are illustrated in Table I. The PCR products were isolated by gel electrophoresis. Equal amounts (approximately 100 ng) of the two purified PCR products for each construct were mixed, denatured, annealed, and used as a template for a third PCR reaction with the outer primers, as described. By using these two primers, it was possible to selectively amplify a template that consisted of annealed PCR products 1 and 2 with the desired mutation. In the case of pGST⅐HKII-S155A, the final PCR product was digested with BamHI and SalI and was used to replace the wild-type BamHI-SalI restriction fragment in the pGST⅐HKII plasmid. For pGST⅐HKII-S603A, the third PCR product was digested with XhoI and ClaI and was used to replace the corresponding wild-type fragment of pGST⅐HKII. In the case of pGST⅐NHKI-D209A and pGST⅐NHKI-S155A, the final PCR products were digested with SmaI and SalI and used to replace the corresponding wild-type fragment of pGST⅐N-HKI. pGST⅐NHKI-D209A and pGST⅐NHKI-S155A were digested with BssHII and NcoI, and the smaller fragment was used to replace the corresponding fragment of pGST⅐HKII to generate the plasmids pGST⅐NICII-D209A and pGST⅐NICII-S155A. These plasmids encode the chimeric proteins GST⅐NICII-D209A and GST⅐NICII-S155A. The PCR fragments inserted into the wild-type plasmids were validated by DNA sequencing.
Expression, Purification, and Activity Measurements of Hexokinase Proteins-The expression and purification of GST⅐HK fusion proteins were carried out as described previously (7). Kinetic parameters were determined by a glucose phosphorylation assay that uses tracer [U-14 C]glucose as substrate, as described previously (7). The K m values for glucose and ATP were determined by measuring the phosphorylation of [U-14 C]glucose as a function of added glucose (0.1-3 mM) and ATP (0.2-10 mM). The K iG6P values were determined, with G6P ranging from 0 to 10 mM, at a glucose concentration of 0.1 mM and an ATP concentration of 10 mM.
Analysis of Data-A rate law for hexokinases was taken as the model of the enzyme mechanism for the analysis of the kinetic data (19). Garfinkel et al. derived this formula according to the data of Grossbard and Schimke (20) and assumed a random order ternary complex mechanism and assumed that G6P is a competitive inhibitor of ATP and a noncompetitive inhibitor of glucose. A sensitive and precise tracer technique allowed us to measure small amounts of enzyme activity at low substrate and high inhibitor concentrations (21). To use the glucose tracer data directly in the determination of the kinetic parameters, we modified the hexokinase rate law by dividing both sides of the equation by the concentration of glucose to correct for the specific activity of the tracer. The modified equation is as follows.
In this equation, [Glc], [ATP], and [G6P] refer to the concentrations of Glc, ATP, and G6P, respectively, and K mGlc , K mATP , and K iG6P are the respective kinetic constants for glucose, ATP, and G6P. V max is the maximum velocity, and v is the velocity of glucose phosphorylation. Global Analysis-Kinetic data were analyzed using a nonlinear least squares curve-fitting program called Globals Unlimited TM (Department of Physics, University of Illinois, Urbana, IL). This global analysis program was originally developed for the analysis of multiple sets of fluorescence spectroscopic data (22,23) and has been used to analyze other complex systems, such as coupled glucose transport and phosphorylation (24). The practical aspects of this approach have been reviewed (25). Global analysis was used to generate nonlinear least square fits to the kinetic parameters of the HKII enzymes based on the hexokinase rate law (see above). The kinetic parameters (i.e. K mGlc , K mATP , and K iG6P ) were determined by analyzing the measured activities as a function of changes in each of the substrate and inhibitor concentrations.

RESULTS AND DISCUSSION
The kinetic parameters of HKII, its two halves (N-HKII and C-HKII), and the D209A and D657A mutant enzymes are shown in Table II. In general, these are similar to the apparent values obtained previously by simple regression analysis (7). The exception is the K iG6P , which is 10-fold lower on average, because the current method of analysis allowed for extrapolation to very low levels of the competitive ligand ATP.
The D209A mutation in HKII results in complete inactivation of the N-half of HKII (7), so the activity of the D209A enzyme is entirely contributed by its C-half. As noted with C-HKII, this mutant enzyme has a 20 -30-fold higher K iG6P than HKII and N-HKII. Inactivation of the C-half of HKII (D657A) results in an enzyme with a K iG6P similar to that of N-HKII. However, neither the S155A mutation, which also inactivates the N-half of the enzyme, nor the S603A mutation affect the K iG6P of the intact enzyme (Table II). The K iG6P values for HKII, S155A, and S603A are between 39 and 45 M, which is in agreement with results reported by Tsai and Wilson (8). These data suggest that S155A and D209A have different roles in HKII, particularly with respect to the regulation of the catalytic activity of the C-half by G6P. Previous studies have shown that the D209A mutation in mammalian hexokinases results in a loss of glucose binding and catalytic activity, whereas the S155A mutation maintains glucose binding but reduces the V max (5,7,17,18). Therefore, the N-half of the D209A enzyme may remain in an open conformation, because the glucose-Asp 209 contact is thought to trigger a conformational change from an open to a closed form. In contrast, the N-half of the S155A enzyme would undergo the conformational change associated with glucose binding, because the glucose-Ser 155 contact occurs after the conformational change. We therefore hypothesize that a conformational change of the Nhalf is required for inhibition of the C-half at relatively low G6P concentrations.
Analysis of the kinetic parameters of the HKII enzyme, and its two halves, supports the hypothesis that the N-half influences the regulation of the C-half by G6P in the intact enzyme. The K iG6P of the intact enzyme is almost identical to that of N-HKII, despite the fact that the K iG6P of C-HKII is ϳ30-fold higher than that of the N-half. Detailed analysis of the G6P regulation of C-HKII in the intact enzyme is, however, complicated by the fact that both halves of HKII are active and both are sensitive to inhibition by G6P. We therefore constructed a chimeric enzyme in which C-HKII is the only catalytically active component and tested it for G6P inhibition.
The N-half of HKI is capable of binding glucose and G6P (1, 6), but it is catalytically inactive when expressed as a discrete protein and in the context of the intact HKI enzyme (Refs. 3-5 and data not shown). A chimeric protein consisting of the N-half of HKI and the C-half of HKII (NICII) allowed us to test whether glucose binding to the N-half, per se, is involved in the regulation of the C-half by G6P. In this chimeric enzyme the catalytic activity is entirely due to C-HKII, which has a significantly higher K iG6P than either HKI or HKII. The chimeric enzyme has K mGlc and K mATP values similar to those of C-HKII; however, the K iG6P is significantly lower than that of C-HKII (Tables II and III). These results indicate that the presence of the N-half of HKI causes a significant reduction in the K iG6P of the C-half of HKII in the context of a 100-kDa enzyme.
We then asked whether the D209A or S155A mutations of this chimeric enzyme have an effect on the K iG6P similar to that seen when the same mutations are made in HKII. The D209A mutation was made in the N-half of NICII chimeric enzyme. This resulted in a significant increase in the K iG6P (Table III). The NICII enzyme, with the S155A mutation in the N-half, was then constructed. This mutation, which does not decrease glucose binding in other hexokinases, did not affect the K iG6P of this chimeric enzyme (Table III).
In the NICII chimeric enzyme, the two halves are not independent entities, and the presence of an N-half that is capable of binding glucose causes the C-half to be regulated at lower concentrations of G6P. Therefore, an allosteric interaction that affects the function of the intact enzyme appears to occur between the two halves of HKII. This interaction may change the structure of the C-half, causing a reduction in the K iG6P of the C-half of HKII. Alternatively, G6P bound to the N-half may result in a conformation that directly inhibits the activity of the C-half in the intact enzyme. The available data do not allow us to distinguish between these possibilities.
Regulation of the function of the C-half of HKII, by binding events that occur in the N-half, has precedent based on observations of HKI. A functional interaction between the two halves of HKI is determined by the binding of glucose, G6P, or P i . For example, co-operative binding of glucose to N-and C-half sites in HKI was noted (26), an observation not expected for a monomeric enzyme, particularly in view of the fact that the N-half has no catalytic activity (3)(4)(5). The binding of either G6P or P i results in a loss of cooperative binding of glucose with a corresponding change in the conformation of HKI (26). Also, the binding constant for G6P is substantially reduced in the absence of glucose (26).
Crystal structures of HKI complexed with glucose and G6P (13,14,16) or with glucose and P i (15) were recently reported. These structures provide direct evidence, as first noted in the yeast HK (9 -12), of transitions between open and closed conformations around the glucose binding cleft. The critical role of amino acids Asp 209 and Asp 657 in glucose binding was confirmed (13, 14, 16). Finally, a mutation of Arg 801 , which could

Primers, templates, and PCR conditions used in the construction of hexokinase mutants
Site-directed mutations were generated by two separate PCR reactions. The primers and template used in the PCR reactions, the annealing temperature, and the approximate size of each PCR product are described for each mutant construct. The sequence of the oligonucleotides used in the PCR reactions is shown below each primer name (the underlined letters indicate base substitution). The PCR reaction mixtures contained 10 ng of the template plasmid DNA, 0.  provide a salt link between the two halves of HKI (15), significantly reduces the relief of G6P inhibition by P i (27). Thus, the crystal structures of glucose and P i bound to HKI support the concept of a functional interaction between the two halves of HKs and suggest that it is dependent upon bound glucose.
The characteristics of HKII described in this paper may relate to the important role this enzyme plays in insulin-mediated glucose clearance of skeletal muscle, cardiac muscle, and adipose tissue. The sequential transport and phosphorylation of glucose by mammalian cells ensures that the subsequent metabolic products are retained within the cell. In the basal state, the intracellular glucose concentration is typically low in muscle (28 -31) and in fat (32,33) but rises as the glucose flux increases, either in response to increased extracellular glucose or as a consequence of insulin stimulation. Recent in vivo studies, using a variety of techniques, suggest that glucose phosphorylationmaybedefectiveinindividualswithnon-insulindependent diabetes mellitus (NIDDM). For example, when positron emission tomography is used to measure 2-deoxy-[2 Ϫ18 F]fluoro-D-glucose transport and phosphorylation, insulin infusion increases the rate constant for phosphorylation in normal weight and obese controls but not in NIDDM subjects (34). Bonadonna et al. (35) investigated nondiabetic controls and patients with NIDDM using forearm balance techniques during euglycemic glucose clamps. They found that glucose transport is decreased about 40% but that combined transport and phosphorylation is decreased 85% in patients with NIDDM. Also, in 31 P NMR studies performed during hyperglycemic-hyperinsulinemic clamps, the muscle content of G6P rises with insulin infusion in controls but fails to increase with insulin infusion in NIDDM subjects (36). We conclude that these patients have defective glucose transport/phosphorylation. Finally, results from muscle biopsies in persons with NIDDM from three distinct ethnic backgrounds also led to the suggestion that the regulation of the HKII gene is abnormal in this disease (37).
These observations are interesting with regard to the data reported in this paper. G6P levels rise as glucose utilization increases. In fact, intracellular G6P levels increase from ϳ170 to ϳ1300 M in perfused working rat hearts either in the presence of 10 mM glucose or after 30 min of insulin treatment (38). Glycogen stores in these hearts increase 5-6-fold. If HKII is a rate-determining step in glucose metabolism and is inhibited by low concentrations of G6P (K iG6P of 45 M), why does glucose utilization increase when G6P increases? We suggest that HKII plays a role in insulin-sensitive peripheral tissues similar to the one glucokinase exerts in the liver. Glucokinase, due to its insensitivity to physiological concentrations of G6P, continues to phosphorylate glucose at high levels of G6P, and this results in increased hepatic glucose utilization and storage. Glucokinase, when expressed in muscle cells, has a similar effect on glucose disposal (39). We hypothesize that HKII, because of the low affinity of the C-half for G6P, may serve the same purpose in insulin-sensitive, peripheral tissues. This could be accomplished by the binding of the enzyme to mitochondria and/or by some other effector that results in allosteric modification of the enzyme that effectively uncouples the functional interaction of the N-and C-halves of the enzyme. chimeric enzymes with HKII Each GST ⅐ HK fusion protein was partially purified to remove any endogenous hexokinase activity as described (7). Kinetic parameters were determined using the hexokinase rate law and the global analysis program as described under "Experimental Procedures." The means Ϯ S.E. were determined from three separate preparations of the enzymes and 17 different velocity measurements for each enzyme preparation. The values for HKII are the same as those shown in Table II