Research article

Mutational analysis of allosteric activation and inhibition of glucokinase

Bogumil Zelent, Stella Odili, Carol Buettger, Dorothy K. Zelent, Pan Chen, Deborah Fenner, Joseph Bass, Charles Stanley, Monique Laberge, Jane M. Vanderkooi, Ramakanth Sarabu, Joseph Grimsby, Franz M. Matschinsky


GK (glucokinase) is activated by glucose binding to its substrate site, is inhibited by GKRP (GK regulatory protein) and stimulated by GKAs (GK activator drugs). To explore further the mechanisms of these processes we studied pure recombinant human GK (normal enzyme and a selection of 31 mutants) using steady-state kinetics of the enzyme and TF (tryptophan fluorescence). TF studies of the normal binary GK–glucose complex corroborate recent crystallography studies showing that it exists in a closed conformation greatly different from the open conformation of the ligand-free structure, but indistinguishable from the ternary GK–glucose–GKA complex. GKAs did activate and GKRP did inhibit normal GK, whereas its TF was doubled by glucose saturation. However, the enzyme kinetics, GKRP inhibition, TF enhancement by glucose and responsiveness to GKA of the selected mutants varied greatly. Two predominant response patterns were identified accounting for nearly all mutants: (i) GK mutants with a normal or close to normal response to GKA, normally low basal TF (indicating an open conformation), some variability of kinetic parameters (kcat, glucose S0.5, h and ATP Km), but usually strong GKRP inhibition (13/31); and (ii) GK mutants that are refractory to GKAs, exhibit relatively high basal TF (indicating structural compaction and partial closure), usually show strongly enhanced catalytic activity primarily due to lowering of the glucose S0.5, but with reduced or no GKRP inhibition in most cases (14/31). These results and those of previous studies are best explained by envisioning a common allosteric regulator region with spatially non-overlapping GKRP- and GKA-binding sites.

  • enzyme kinetics
  • fluorescence quantum yield; glucokinase (GK)
  • glucokinase mutant (GK mutant)
  • glucokinase regulatory protein (GKRP)
  • glucokinase activator drug (GKA)
  • glucose
  • tryptophan fluorescence (TF)


Substrate activation and allosteric regulation of enzymes and transporters are essential features of homoeostasis in cells and organisms. Haemoglobin is the classical illustration of this principle in the physiological chemistry of oxygen [1]. Glucose activation of GK (glucokinase) and allosteric inhibition by GKRP (glucokinase regulatory protein) or activation by GKAs (glucokinase activator drugs) of the enzyme are other striking examples of these fundamental mechanisms [27], in this case of great importance for glucose homoeostasis and the pharmacotherapy of diabetes mellitus. However, a full molecular understanding of such processes is frequently lacking, even for haemoglobin.

The present study is an attempt to contribute to our knowledge about GK regulation. GK, the low-affinity isoform of hexokinases, plays a critical role in glucose homoeostasis as the glucose sensor in pancreatic β-cells controlling GSIR (glucose-stimulated insulin release) and, equally important, as the regulator of hepatic glycolysis, glycogen synthesis and gluconeogenesis [4,5,7]. Regulation of the enzyme takes place at several levels, including control of gene expression and protein degradation (on a time scale of minutes to hours), but also by substrate activation and allosteric modification (on a time scale of seconds to minutes) [5,710]. A single gene endowed with two promoters directs constitutive GK synthesis in the neuro-endocrine system, including the pancreatic β-cells and gonadotropes (regulated by the upstream promoter) or in the liver (regulated insulin-dependently by the downstream promoter). Acute regulation of GK is mediated by its substrate glucose, by allosteric modifiers (GKRP, GKAs, the bifunctional enzyme F6P-2-P (6-phosphofructo-2 kinase/fructose-2,6-bisphosphatase) kinase/phosphatase and the proapoptotic factor BAD (Bcl-2-associated death promoter) [36,11,12] and by compartmental redistribution within the cell involving the nucleus, mitochondria, hormone granules and the Golgi apparatus [4,12,13]. Activation of GK by its substrate glucose in the physiological range of 4–12 mM is the basis of the enzyme's co-operativity with regard to glucose and is caused by the process explained by a ‘mnemonic’ or ‘ligand-induced slow transition’ mechanism during which GK changes from a low-affinity open conformation to a high-affinity closed conformation [2,14,15]. These two explanations have their corollary in the ‘induced fit’ hypothesis or the concept of a ‘pre-existing equilibrium’. Accumulating evidence favours the idea of a ‘pre-existing equilibrium’ [1619]. The transition process and its reversal is much slower than the catalytic cycle explaining the sigmoidal glucose-dependency of the phosphorylation reaction. It is readily observable by TF (tryptophan fluorescence) recordings [15,20,21]. Superimposed on to this basic kinetic co-operativity with regard to glucose is an allosteric control mechanism involving inhibition by GKRP and activation by GKAs. Inhibition by GKRP is a liver-specific process [3,4]. GKRP is a 65 kDa protein of unknown crystal structure. It is a primarily nuclear protein that binds with GK competitively with glucose and sequesters the inhibited enzyme in the nucleus. Its effectiveness is enhanced by F6P (fructose 6-phosphate), but counteracted by F1P (fructose 1-phosphate) which explains the stimulation of hepatic glycolysis by low levels of fructose. Glucose and F1P dissociate the nuclear GK–GKRP complex, thus initiating translocation from the nucleus to the cytosol and thereby activating glycolysis and glycogen synthesis. The limited structural information is part of the reason that delineation of the interface between GK and GKRP remains disputed. The discovery of GKAs was reported in 2003 [5,6]. All currently known GKAs are pharmacological agents, and the search for postulated physiological activators has been unsuccessful so far. GKAs are non-essential allosteric modifiers of hepatic and islet GK. Countless patented GKA structures have a comparable pharmacophore with variable affinities to a well-defined allosteric binding site of GK. GKAs lower the glucose S0.5 value as much as 10-fold and usually increase the kcat value as much as 2-fold. They markedly increase the ATP Km value at glucose below the glucose S0.5 and have variable strength in lowering the h (Hill coefficient) [5,22]. As a result they enhance GSIR from the pancreas and stimulate hepatic glucose uptake and glycogen synthesis, but curb glucose production. The most compelling evidence for the dominant role of GK in glucose homoeostasis stems from the biochemical genetic analysis of more than 600 mutations in humans [23] discovered in individuals with HI (hyperinsulinism) due to activating mutations and with diabetes mellitus due to inactivating mutations [mild forms when only one allele is affected in MODY (maturity onset diabetes of the young)-2 and severe forms when both alleles are involved in PNDM (permanent neonatal diabetes mellitus)].The recent demonstration of high efficacy of GKAs to lower blood glucose in patients with T2DM (Type 2 diabetes mellitus) underscores further the medical relevance of GK's role in glucose homoeostasis [24].

In the present study we have selected 31 GK mutants to explore the nature of the interaction between GK and its best-known binding partners glucose, GKRP and GKAs. Enzyme kinetics and TF were used to investigate the actions of glucose, GKRP and GKA on the function and structure of GK. Highly relevant previously published results were also considered and contributed critically to this exploration. The results of this investigation allowed us to conceptualize an allosteric regulatory region of the enzyme with clearly defined separate binding sites for GKRP and GKAs contrasted with the remote substrate binding domain with distinct contact areas for glucose and MgATP. The structural visualization of the glucose-induced slow transition provides plausible corollaries for facilitating GKRP dissociation from and GKA binding to GK.



D(+)-Glucose of 95% purity by GC, was supplied by Sigma–Aldrich. The non-fluorescent GKA (RO0274375-000) used was discovered, synthesized and characterized by Hoffmann-La Roche [25]. Water used for all solutions was first deionized using Millipore reverse osmosis and then glass-distilled.

Selection of GK mutants

We selected 31 GK mutants from our database of more than 100 well-characterized mutants based primarily on kinetic criteria rather than being guided by the GK-linked clinical phenotypes, because the correlation between molecular features of the mutants and clinical manifestation is complicated by cell-biological factors such as functional and structural instability [26]. The goal was to cover as wide a range as practical from strongly inhibited to strongly activated enzymes. For this purpose we used the GK AI [activity index; (kcat/S0.5h)×(2.5/2.5+ATP Km)] [2729] and achieved a 600-fold range from about 0.1–60 divided into two groups: 13 enzymes with AIs from 0.09 to 1.60, including the control (AI=1.37), and 19 enzymes with AIs from 2.08 to 57.3. This selection of enzymes featured the range of responsiveness to the allosteric modifiers GKA and GKRP from refractory to fully reactive required by the research plan. Extremes of mutational inactivation were avoided because very low D-glucose affinities and/or catalytic capacities decrease the quality of the biochemical analysis. Finally, proteins with sufficient yields and high stability during storage were chosen to allow extensive kinetic studies and fluorescence analysis. Most of these enzymes have been studied and partially characterized before using their GST fusion proteins (but not pure GK) in studies on the biochemical genetics of GK linked hyper- and hypo-glycaemia in humans (wild-type enzyme, V62M, S64P, S64Y, T65I, G68V, G68K, G72R, V91L, M197I, C213R, Y214C, C252Y, S263P, M298K, S336L, V389L, K414E, E442K, V452L, 454-Ala,V455M and A456V) [23,2731]. Of these, seven were activating or inactivating mutants that had been designed in the course of the present or previous studies (M197E, M197I/A379T, M197L, Y214A, Y214A/V452A, Y215A and A379T). Two were recently identified in mice made diabetic by ENU (N-ethyl-N-nitrosourea) treatment, K140E and P417R [32]. Finally, one was an incidental mutant that had been discovered in the course of GK mutagenesis experiments (A379T) [30].

Enzyme purification

Recombinant wild-type and mutant human β-cell GKs were generated and expressed as GST fusion proteins in Escherichia coli as described previously [21]. GST–GK fusion proteins were cleaved with Factor Xa and submitted to another round of purification by removing GST with glutathione–agarose and Factor Xa with benzamidine–Sepharose 6B following the manufacturer's protocols. Point mutations were introduced into the pGEX-3X vector using the QuikChange site-directed mutagenesis kit (Stratagene). All mutants were transformed into E. coli cells and verified by DNA sequencing [21].

All mutant GK proteins were expressed in significant amounts as indicated by the yields in terms of mg of protein per litre of growth medium and were found to be essentially pure as demonstrated by the presence of a single band at 50 or 75 kDa on PhastGel electrophoresis for GK or GST–GK respectively. Purified GST protein was stored at −80°C, either with 50 mM glucose or in the absence of glucose in a medium containing 25 mM Tris/HCl, 100 mM KCl, 30% glycerol and 2.5 mM DTT (dithiothreitol) (pH 7.6) [21]. Immediately following cleavage and removal of GST, the enzyme was stored in buffer lacking glycerol, but containing 50 mM glucose. Glucose was removed by dialysis and the enzyme (usually concentrated about 2-fold) was then stored in glucose and glycerol-free buffer until used for kinetic analysis or spectrofluorimetry. Recombinant human GKRP was prepared as described previously [27,28].

Enzyme kinetics

Kinetic properties of GK were determined as described previously [21,27,28]. GK activity was measured by spectrometry using an NADP+-coupled assay with glucose-6-phosphate dehydrogenase. The reaction medium contained 6 mM MgCl2, 0.1% BSA, 150 mM KCl, 100 mM Hepes, 1 mM NADP+, 5 unit/ml glucose-6-phosphate dehydrogenase, 5 mM ATP and 2 mM DTT. The following modifications to the protocol were made: in protocol A, kinetic studies were usually carried out with 11 glucose dilutions between 0 and 100 mM in the presence of 5 mM ATP; and in protocol B, a glucose concentration at 10×S0.5 was used, but with various ATP concentrations with 1 mM MgCl2 in excess. GKRP inhibition of GK and its mutants was assayed as described previously with minor modifications [27,28]. The assays were run at the corresponding glucose S0.5 and at 70% ATP saturation with 1 mM excess MgCl. Note that the pH of the buffer was 7.1 and the KCl was lowered to 25 mM compared with 150 mM in the routine GK assay. Because of the limited availability of recombinant huGKRP (human GKRP) and the large number of mutants, we determined the relative inhibition at two GK/GKRP molar ratios (0.88 and 1.76) of the near linear portion of the dose–response curve and report the mean value of the effects at nominally 1.26 molar excess of the inhibitor both in the absence and presence of 10 μM sorbitol 6-phosphate.

Spectroscopic measurements

UV and visible absorption spectra were measured using a Hitachi–PerkinElmer U-3000 spectrophotometer. Fluorescence intensity and spectra were measured with a Fluorolog-3-21 Jobin-Yvon Spex Instrument SA equipped with a 450 W xenon lamp for excitation and a cooled R2658P Hamamatsu photomultiplier tube for detection. For all measurements 90° geometry was used. Excitation wavelengths of 295 nm were used to observe fluorescence emission in the 290–500 nm range. Slit width was set to provide a band-pass of 4 nm for excitation and 3 nm for emission. For time-dependent intensity glucose titrations, the band-pass for excitation was 0.5 nm at λex=280 nm and for emission 12 nm at λem=340 nm. A thermostatically-controlled cell holder maintained sample temperature. The glucose concentration titration curves were obtained by adding the sugar stepwise at increasing concentrations to the fluorescence cuvette. The stock concentration was 1 M and the pH was adjusted to 7.3. Stock solutions were prepared freshly before each titration, but given sufficient time to anomerize. Aliquots (1–100 μl) of the stock solution were added to 1 ml of the protein solution which was about 100 μg/ml and contained 5 mM phosphate buffer with 100 mM KCl and 1 mM DTT (pH 7.3).

Determination of TF quantum yield

Fluorescence quantum yields of tryptophan (Φ) were determined for GK wild-type and mutants using the following equation [33]: Embedded Image The subscripts S and R refer to sample and reference standard respectively. AR and AS denote the absorbances at the excitation wavelength (A<0.05 to avoid the inner filter effect). FS and FR denote the integral intensities of the recorded fluorescence spectra measured under identical instrument settings. nS and nR are the refractive indices. Fluorescence quantum yields were determined relative to NATA (N-acetyl-L-tryptophanamide; Φ=0.14) in water as the reference [34]


Experimental aspects: functional comparison of GST–GK with GK after cleaving the fusion protein and pH-dependencies of selected study parameters

Many studies published previously were performed with recombinant GST–GK, and it was assumed that the pure GST-free authentic enzymes would be functionally indistinguishable from the fusion proteins [23,2731]. This assumption remained to be comprehensively tested. In the present investigation we therefore prepared pure GKs by cleaving off the N-terminal GST tag from the wild-type and 31 mutants. We demonstrated more than 95% purity by the presence of a single 50 kDa band on PhastGel (results not shown) and compared the characteristics of GST–GK with the corresponding pure enzymes (see the results of Tables 1 and 2 and Supplementary Tables S1 and S2 at The present paper expands to 43 cases the comparison of kinetics of GST–GK with pure GK following cleavage of the fusion protein. The results of the present study and those of studies published previously show that removal of the N-terminal GST tag has little impact, with the single exception of Y214A/V452A, as was described previously [21]. As Supplementary Figures S1 and S2 (at summarize, the kinetic and binding constants for the two preparations tend to agree, since the kcat, glucose S0.5, ATP Km and the h are, within experimental error, comparable. This observation demonstrates that previously published biochemical genetic information and interpretations on ‘glucokinase disease’ (almost entirely based on the analysis of GST-tagged recombinant human GK preparations) are biologically meaningful and that the present biophysical studies requiring pure enzyme in which the source of the TF signal is limited to the GK molecule are feasible.

View this table:
Table 1 Kinetic parameters of wild-type (WT) and mutant GK

Enzymes were purified as GST fusion proteins, but then cleaved and stored in the presence of 50 mM D-glucose.

View this table:
Table 2 Effect of GKA on recombinant human wild-type and mutant GK (cleaved)

A denotes the maximal fold drug effect on the kcat, B on the glucose affinity and C on the activity index. EC50 values are half-maximal drug concentrations for the effect on the kcat, the glucose affinity and the activity index. n.a., not applicable.

Pure GK was also used to obtain essential information not previously recorded, i.e. the pH-dependencies of the relative quantum yields of the glucose-induced fluorescence increase and of the effect that GKAs have on the glucose Kd values for the wild-type enzyme (see Supplementary Figures S3 and S4 and also Supplementary Tables S3 and S4 at Strong pH-dependencies were observed for both parameters with optima in the physiological range.

Effects of point mutations on GK characteristics, including GKA responsiveness

The physiological substrate D-glucose was used for a comprehensive characterization of the 32 pure GK protein species. The studies included basic kinetics, kinetics influenced by GKA activation or GKRP inhibition, and characteristics of substrate binding with TF (Tables 1–5). The results of the wide-ranging analysis summarized in these Tables 1–5 are recorded in the order of the linear sequence of human islet GK as the least-biased database presentation. A full data set for almost every enzyme was obtained with the exception of GKRP inhibition data for S336L because of its prohibitively low kcat and relatively high ATP Km. The responsiveness to GKA activation and GKRP inhibition varied greatly, ranging from normal to totally refractory as anticipated (for enzymes which had been studied before using GST fusion proteins) or predicted by extrapolation based on the nature of the mutation. The comprehensive glucose-binding studies using TF made possible by employing pure GK add a new dimension to the kinetic analysis of this set of mutants. All TF-based glucose-binding curves were hyperbolic, in agreement with previous reports (Supplementary Figure S5 at, but the spectra and the quantum yields of the basal unliganded state and the glucose-induced relative fluorescence increases and maximal intensities varied greatly among the different enzymes. Examples of these glucose titrations and of the fluorescence emission spectra are shown in Supplementary Figures S5 and S6 (at The impact of a particular mutation on the basal fluorescence and the glucose-induced fluorescence increase can be surprisingly small considering the magnitude of the functional change it causes, as seen in the case of the activating V455M (compare the corresponding kinetic and binding constants of Tables 1–5 and Supplementary Figure S6A with Supplementary Figure S6C). This can be contrasted with cases with dramatic changes of basal and glucose-induced fluorescence of enzymes that are kinetically not very far from normal (e.g. G72R and S263P, as apparent from Tables 1–5 and Supplementary Figures S6B and S6D). It is remarkable that only five of the 31 mutants (M197I, M197L, V389L, E442K and V455M) have a normal or close to normal emission maximum of 326–327 nm in the basal state in contrast with the others, which show red shifts to as high as 332–333 nm and that all but two mutants (T65I and C252Y) show a significant blue shift upon glucose saturation (Table 5 and examples in Supplementary Figures 6E and 6F). Previous studies with single tryptophan enzymes suggested that the red shifts of the emission maxima in the basal state are due to the influence of Trp99 and Trp167, but that the process of glucose binding is then responsible for the rebound to the blue by the altered TF spectrum of Trp167 [21]. We arrive at this interpretation from quantitatively analysing the basal and glucose-dependent TF spectra of the three GK mutants which retain only a single tryptophan residue. The TF spectra of the free and glucose-saturated native enzyme can be explained by simple additivity of the individual spectra of the single tryptophan mutants [21]. The result is indeed remarkable in view of the fact that two tryptophan residue substitutions are needed to generate mutants containing only one tryptophan residue, which causes significant functional changes. The results imply that the backbone structures of these mutants are maintained. Taken together, these results strengthen our interpretation that extrapolations from the spectra to the backbone structures of the open (ligand free) and closed (glucose saturated) conformations are valid.

View this table:
Table 3 Effect of a 1.26 molar excess of human GKRP on human wild-type (WT) and mutant GKs in the absence and presence of sorbitol 6-phosphate (S6P) (relative inhibition of GK activity)

n.d., not determined.

View this table:
Table 4 Binding constants in mM and relative fluorescence increase for D-glucose using cleaved GK and GK mutants in the absence and presence of 20 μM GKA measured by TF

Fluorescence increase is indicated in italics. WT, wild-type.

View this table:
Table 5 Average quantum yields of TF of GK and GK mutants

GKA RO0274375 (RO-cpdA) was used in the present study because previous kinetic and biophysical studies on the mechanism of action of GKAs were done with this compound and because it is not fluorescent [25]. We were unable to study GKA–GKRP interactions directly because of a limited supply of the inhibitor. However, GKAs of a great variety of structures were employed in previous studies that are relevant for the present study because this interaction was addressed directly [3539]. In the present study, RO0274375 did not change the TF of wild-type GK in the absence of glucose using standard assay conditions, which is interpreted as evidence that the drug does not bind. This finding is in agreement with those from two other studies using structurally different GKAs and different methodologies [36,37]. In a third study, glucose-independent drug binding was demonstrated at high GKA concentrations based on increased TF of GK with two different GKAs, one of them fluorescent but with fluorescence characteristics allowing differentiation from TF. It was also demonstrated that the EC50 and Kd of GKAs decrease markedly with increasing glucose concentrations [35]. It is perhaps not surprising that GKA structure affects the glucose-dependency of drug binding. However, as important as differences of GKA structures might be, incubation conditions and mutations of GK greatly influence the glucose-dependency of GKA binding. For example, inclusion of 20% glycerol in the buffer or mutation of Trp99 both have a profound impact, allowing RO0274375 binding in the absence of glucose, again judging from the effect on TF (B. Zelent and F. M. Matschinsky, unpublished work). The present conclusion that GKRP and GKA binding to GK are mutually exclusive is very strongly supported by these reports based on studies with other, structurally distinct, GKAs [3639]. Structure–activity relationship studies of GKA–GKRP interaction in affecting GK have to be greatly expanded to develop a full understanding of this biochemically and pharmacologically significant question.

Testing the hypothesis that mutational activation of GK may lead to partial closure of the superopen GK conformation and might thereby interfere with allosteric regulation of the enzyme

It had been noticed previously that certain mutants (C213R, Y214A/V452A and C252Y) have a high basal fluorescence and lack responsiveness to GKA [21]. This observation led to the speculation that a high basal TF could be a manifestation of a partially closed compact conformation of the protein, reducing at the same time the responsiveness to allosteric modification by GKRP and GKAs. The present study is a test of this hypothesis, which offered a constructive framework for interpreting the large mass of experimental data. It was therefore explored whether a correlation exists between the enzyme's TF, both basal and glucose-induced, and GK's catalytic capacity (kcat/glucose S0.5), GKA responsiveness (as measured by the effect of the activator on the glucose Kd and its S0.5) and responsiveness to the physiological inhibitor GKRP (Figure 1 graphically summarizes the results presented in Tables 1–5). The GKA effect on glucose Kd expressed as Kd(D-glucose)/Kd(GKA) and plotted in order of descending effectiveness was used as basis of comparison because this parameter is the simplest and does not involve the complex chemistry of catalysis and thus seemed to be most suitable for this purpose. A comparison of Figures 1(A)–1(E) suggests strongly that the hypothesis is supported by the evidence. In general it is evident from comparing the response pattern depicted in Figure 1(A) with those of Figures 1(B)–1(E) that good correlations or remarkable trends of correlations do exist between the glucose-binding data and the other parameters.

Figure 1 Summary of the experimental data

(A) The GKA responsiveness of all 32 enzymes [wild-type (WT) and 31 mutants, 24 of these disease-causing and the rest incidental or designed]. GKA responsiveness is expressed as the ratio of Kd values in the absence and presence of drug levels near saturating for the wild-type enzyme (20 μM). The results are presented in two groups: normal or retaining a significant response (white bars) compared with uniformly low or absent response (black bars). (B) The results of TF measurements in the absence (white and black bars) and presence (grey bars) of saturating levels of glucose expressed in terms of relative quantum yield. Note that all of GKA refractory mutants have a high basal quantum yield and that the majority of these (11/15) are also GKRP-resistant. It is also noteworthy that there are four outliers in the left-hand group with high basal TF, three of which are known to be structurally unstable (thermolabile). (C) GKA responsiveness based on steady-state activity measurements of enzyme activity in the presence and absence of GKA. The S0.5 value with saturating GKA present was obtained from dose-dependency studies with the drug level extending to saturation instead of using a fixed drug concentration as in (A) and (B). (D) The kcat/S0.5 ratios as a measure of catalytic capacity of the enzymes studied. Note the value is high for all spontaneous mutants causing hyperinsulinaemic hypoglycaemia in heterozygous carriers. (E) Relative GKRP inhibition of GK as measured by steady-state kinetic analysis with GKRP in 1.26-fold molar excess over GK present at 10 nM. These measurements were made in the absence (white and black bars) and in the presence of 10 μM sorbitol 6-phosphate which enhances the effect of GKRP (grey bars). The six outliers are discussed in the main text.

A careful inspection of the proposed correlations of response patterns reveals apparent exceptions (Figure 1). The correlation of the patterns in Figures 1(A) and 1(B) shows four apparent exceptions: C213R, M298K, S263P and K414E. These mutations are all mildly inhibitory and cause an increase of basal fluorescence without altering GKA or GKRP responsiveness significantly. In addition S263P and M298K have been found to be more thermolabile than any other instability mutant known to date [32]. However, this information is insufficient to explain the basal fluorescence increase, nor is it sufficient to conclude that it is an expression of structural compaction. Comparing Figures 1(A) and 1(C) it can be seen that the correlations are apparently less than ideal for V389L and M298K. V389L (an HI-causing activating mutation) and M298K (an inactivating very thermolabile mutant causing MODY) are outside the allosteric activator region and their exceptional sensitivity to GKA has little impact on the conclusions. It is also noteworthy that glucose S0.5 is a parameter that is derived from the far more complex process of chemical catalysis than that expressed by the glucose Kd, which is undoubtedly reflected in subtle differences of their GKA responsiveness. Comparing Figures 1(A) and 1(D) it appears that V62M, T65I, G72R and C252Y (all part of or close to the allosteric activator site) are outliers. However, lack of extensive activation (GKs 2, 5 and 8) or even reduction (GK 20) of catalytic capacity caused by substitution of amino acids in the allosteric activator site does not contradict the conclusions in view of the complex nature of the kcat/S0.5 expression. Interpreting the results of GKRP inhibition studies is perhaps the most complicated and challenging. The results with mutants K140E, M197E, V452L, G68K, Y214C and Y214A do not seem to fit when Figures 1(A) and 1(E) are compared. Decreased effectiveness of GKRP can have different causes: (i) the mutation may interfere with binding of GKRP to one of the two separate contact patches of GK considered here integral to the inhibitor-binding site without influencing the GKA site that regulates conformational transitions. This seems to be the case for K140E and M197E both resulting in a charge change without affecting basal fluorescence or GKA responsiveness; (ii) activating mutations V452L, Y214C and Y214A may be effective in interfering with GKA binding and sufficiently strong to cause partial closure of GK conformation (as apparent by the increased basal fluorescence), but not strong enough to prevent GKRP binding and GK inhibition at low glucose levels. We have no plausible explanation for the exceptional behaviour of G68K. Finally, it should be kept in mind that the different GK conformations considered in the present paper are probably present in a pre-existing equilibrium which could be subtly altered by differential ligand binding [1619]. It seems therefore that, on the whole, the pattern of GKRP inhibition of the present set of mutants is plausibly related to the other patterns. On balance we conclude that the correlations of Figures 1(A–E) are strong enough to support our views about the allosteric mechanisms underlying GKRP inhibition and GKA activation of GK discussed below.

Structural information on GK contributes greatly to the understanding of the present results

Current views about GK are greatly influenced by the crystal structures described by Kamata et al. [40] showing a ligand free ‘superopen’ conformation which is contrasted with a ‘closed’ ternary GK–glucose–GKA complex. Previous crystallographic studies have greatly amplified this knowledge base by solving the structures of the binary GK–glucose complex (PDB code 3IDH; also [41]), the ternary GK–glucose–MgANP (ANP is phosphoaminophosphonic acid-adenylate ester) complex (PDB code 3FGU) and finally of the quaternary GK–glucose–MgANP–GKA complex (PDB code 3ID8). This crystallographic evidence suggests strongly that all glucose-containing complexes are practically indistinguishable, as illustrated in the present study by the nucleotide-free complexes when compared with the open structure of Kamata et al. [40] (Figure 2). The evidence from TF measurements also demonstrates that the binary GK–glucose complex does indeed resemble the closed ternary GK–glucose–GKA complex very closely, because the fluorescence characteristics are practically the same (results not shown). On the basis of this background we projected the mutated amino acids shown in the present paper to interfere with GKA action and also of those mutants that impair GK inhibition by GKRP on to the open and closed configurations, the latter with GKA present (Figures 3A, 3B, 4A and 4B) and contrasted them with the projections of mutants ineffective in this regard (Figures 3C, 3D, 4C and 4D). We also included in this graphic representation highly relevant previously published results of a mutational analysis of GKRP inhibition of Xenopus laevis or human GK showing that the amino acids Glu51, Glu52, His141, Lys142, Lys143 and Leu144 are involved in GKRP action and contrasting them with 12 others that are not: A114I, T116Q, M121E, Y125H, V154C, R155H, H156Q, E157T, D158N, T346V, L349R, R353Q [42,43]. The binding site for GKAs has been well delineated and includes the following amino acids: Val62, Arg63, Met210, Ile211, Tyr214, Tyr215, Met235, Val452 and Val455 [5]. The results are striking. The projections show that GKA and GKRP bind to a common region of GK (termed the allosteric regulator region), but interact with clearly separate domains in this region and that glucose binding to the substrate site affects the conformations of these domains differentially in a manner which causes the GKRP-binding patches to separate and presumably dislocate the inhibitor while opening the GKA-binding domain so that the activator has access to its contact amino acids. The concerted motion of these subdomains is depicted in Figure 5.

Figure 2 Three GK structures in equilibrium

GK in the superopen conformation (PDB code 1V4T) (A), in the binary closed structure (PDB code 3IDH) (B) and the conformation of the closed ternary complex (PDB code 1V4S) (C) are shown. The tryptophan residues are shown in red, glucose in green and GKA in blue.

Figure 3 Location of those amino acids which are implicated in GKA and GKRP binding compared with those which are not

(A and B) Amino acids involved in binding. (C and D) Amino acids not involved in binding. The structures are viewed from the side. In addition to the results of the present study (amino acids coloured red, blue and light blue), 18 mutants reported in the literature were used (amino acids coloured grey) [42,43]. It is noteworthy that the activating mutants M197I, V389L and E442K are located outside the allosteric activator site as currently conceptualized. The locations of the three tryptophan residues are indicated. Open and closed (plus glucose in green and GKA in yellow) structures are shown. Most notable are the opening of the GKA-binding site upon addition of glucose (compare the locations of the amino acids in the open with the closed conformation) on one hand and the separation of the two GKRP contact patches of amino acids when glucose plus GKA bind (Glu51/Glu52 from Lys140, His141, Lys142, Lys143, Leu144 and Met197) on the other. It should be realized that any structural change that might be associated with a mutation (e.g. partial cleft closure) is ignored for lack of direct information. Note that Thr346, Leu349 and Arg353 of the X. laevis enzyme are replaced by the corresponding amino acids lysine, tyrosine and serine in human GK.

Figure 4 GKA and GKRP binding residues of GK from an alternative orientation

The results of the present study are in blue or red and those from the literature in grey. The locations of tryptophan residues (W) are indicated. Glucose is green and the GKA yellow. Remarkably, the amino acids not involved in GKA or GKRP binding (action) are randomly scattered in the periphery of the protein and are absent from the ‘allosteric regulator region’. This is most convincing by viewing Figures 3(C) and 4(C). Also notable is the fact that the tops of the large and the small lobes show no evidence of involvement in GKA or GKRP binding (action).

Figure 5 GK GKA binding site relative to glucose and to GKRP binding patches

(A and B) The closed ternary structures. (A) The GKA-binding site relative to glucose (green) and the GKRP-binding patches (orange). The two sites clearly do not overlap. (B) A zoomed in and rotated view of the sites shown in (A). (C) The transition from the open to the closed conformation illustrating the large spatial separation of the two GKRP-binding patches. The displacement of the GKRP-binding patch upon the addition of glucose is clearly shown: the orange patch with glucose, the yellow, without.

This mechanism envisioning a circumscribed allosteric regulator region with separate but interdependent binding sites for GKAs and GKRP located in the circumscribed domain where the two main lobes of the enzyme are joined together (concept A) as described above in Figures 3 and 4 and summarized in Figure 5, is not compatible with another recent proposal (concept B) which identified different essential GKRP binding patches, one at the tip of the large lobe (Leu350/Asn355) and the other at the inner aspect of the substrate binding area (Leu58/Asn204), even though both mechanisms postulate that glucose binding is responsible for a conformational change that dissociates the inhibitor from the enzyme [44,45]. These contradictory concepts were developed with greatly different research strategies, which may explain the diverging results and conclusions. Concept A is based on the outcome of a mutational analysis of GKA activation and GKRP inhibition using TF and solution kinetic analysis of recombinant GK and a large number of its mutants (mostly identified in MODY-2 and HI patients and from an extensive study with X. laevis GK), whereas concept B is based largely on cell biological evaluation of GK mutants that were designed as guided by random peptide phage display library screening for binding partners of GKRP and GK. The present mechanism is compelling because it was developed from a large database compiled from two independent complementary studies which both used the highly robust methodologies of enzyme kinetic analysis and tryptophan spectrofluorimetry coupled with wide-ranging site-directed mutagenesis of GK and because it is internally highly coherent. It is important that unresolved questions arising from these two mechanistically diverging concepts be conclusively settled by further critical examination.

The interpretation of the present results from extensive TF studies of pure recombinant human islet GK in solution is obviously grounded in the remarkable results and the views of crystallographers, but it leads to a significant progress in the structural characterization of the enzyme. The present TF data suggest strongly that the backbone structure of the binary GK–glucose complex is the same as that of the ternary GK–glucose–GKA complex because the TF spectra are virtually indistinguishable, i.e. the shape of the emission spectra, the wavelengths of the maxima and their quantum yields are the same. The conclusion is warranted because TF is an extremely sensitive indicator of even minor structural changes of GK, as demonstrated in the present and previously published studies. Identity of TF is thus very strong evidence for identity of molecular backbone structure [21,46]. The high sensitivity of TF in this particular case is explained by the ideal locations of the enzyme's three trypytophan residues, particularly of Trp167 which closely reflects structural events in the glucose-binding site and of Trp99 which mirrors dependably even minor local changes in the allosteric activator site, while Trp257 fluorescence is barely affected by most GK modifications. Extensive studies with mutated GKs retaining only one of the three tryptophan residues have been invaluable in the individual characterization of these intramolecular fluorescent probes [21]. TF measurements of the open form of many mutant enzymes in the present study clearly show that activation of the enzyme by single point mutations in the allosteric activator site is frequently associated with a greatly enhanced quantum yield and a red shift of the emission maximum, which is interpreted as evidence for the existence of a compacted partially closed structure of the apoenzyme. We speculate that Trp99 is the predominant localized source of this TF increase and that glucose binding elicits the additional fluorescence augmentation and blue shift of the spectrum by affecting Trp167. It is not unreasonable to speculate that such a partially closed GK structure could normally exist as a low-abundance intermediate conformation in equilibrium with the open apoenzyme and a fully closed structure favoured by formation of the glucose-bound binary complex and that suitable amino acid substitutions change the distribution of these multiple conformations, helping to explain the increase in substrate affinity typical for these mutants. Molecular-dynamics calculations and rapid-mixing studies of the TF increase during the glucose induced slow transition from the open to the closed conformation have generated evidence for the existence of several intermediates [1619]. It is not unreasonable to propose that the high basal TF of activated and perhaps other mutants (e.g. C213R, M298K, S263P and K414E) represent one or more of these hypothetical intermediates in the pathway from the open to the closed conformation. Inclusion of the GK–GKRP interaction in the concept of a pre-existing equilibrium of multiple GK conformations would then also offer explanations for apparent exceptions in the correlations of Figure 1(B) with 1(E) as an expression of effective sequestration of GK in the form of binary GK–GKRP complexes (e.g.G68K, Y214C and Y214A).

General conclusions and future outlook

GK is activated pharmacologically by a novel class of GKAs [5,6] and is inhibited physiologically by the liver-specific GKRP [3,4]. We have used mutational analysis to advance the characterization of the enzyme's binding sites for GKAs and GKRP. The effects of a GKA (RO0274375) and recombinant huGKRP on the steady-state kinetics of recombinant human islet GK were studied, and TF was used to determine the glucose Kd (both in the presence and absence of GKA). TF was also used to extract structural information on GK and its mutants in its open ligand-free form (the apoenzyme), in the closed glucose-bound conformation (the binary enzyme–glucose complex) and in the closed GKA-activated glucose-containing structure (the ternary enzyme–glucose–GKA complex) by measuring the quantum yield and spectral characteristics under these conditions which was essential for the interpretation of the large database (see also [21]). All together we compared 31 mutants with wild-type GK and evaluated highly relevant results on mutant enzymes reported in the literature.

The results of the present study and evidence from the literature support the view that both GKA and GKRP bind to a GK allosteric regulator region [5,6,22,3539], that their binding is mutually exclusive [5,33,35] and that GKA binding is glucose-dependent [5], but GKRP inhibition of GK is competitive with glucose [3,4]. This is based on the finding that point mutations in this regulator region have highly predictable effects on the enzyme conformation, as shown by the TF results, on enzyme kinetics and GK's responsiveness to activator and inhibitor. Two characteristic GK profiles predominate accounting for nearly all mutants: (i) mutants with a normal or close-to-normal GKA response (13/31) usually exhibiting the low basal TF of wild-type enzyme (interpreted to indicate an open relaxed conformation of GK), show some variability in catalytic activity, but are usually strongly inactivated by GKRP; and (ii) mutants which are refractory to a GKA (14/31), usually exhibiting a high basal TF (interpreted as evidence for structural compaction and partial closure of GK), are usually strongly activated (i.e. they have an increased kcat/S0.5 ratio) and show greatly reduced or no inhibition by GKRP.

The effects of amino acid substitutions on GK are direct, indirect or may result from a dual mechanism. Mutations of known GKA contact amino acids or close neighbours (V62M, S64P/Y, T65I, S68K/V, G72R, Y214C, Y215A, C252Y and 454-Ala) are probably directly activated by the substitution resulting in structural changes that then cause a higher basal TF and refractoriness to both the activator and inhibitor. GKRP resistance in another group of mutants (E51S, E52K, K140E, H141G, K142P, K143H, L144M and M197E) is also best explained by a localized direct effect of the substituted amino acid, an interpretation supported by the fact that the mutants examined in the present study (K140E and M197E) comprehensively show normal basal TF and GKA reactivity. GKs in a large third group of these mutants have normal basal TF and retain their response to GKA and GKRP. It is noteworthy that TF and GKA pharmacology is not available for the X. laevis GKs which provide an important part of the argument [42,43]. Plausible explanations for a few outliers were offered above.

It is concluded that the GKA and GKRP contact areas are both located in the same GK allosteric regulator region situated on the outer aspects of the protein's hinge between the two lobes, but that these ligand-specific sites do not overlap (Figure 5). They flank two sides of the back of GK. The location of the GKA-binding pocket has been characterized crystallographically and involves at least nine amino acids of the loop between β-1 and β-2, and in the helices α5 and α13 [5,40]. The proposed GKRP-binding area is situated opposite to the GKA site involving at least eight amino acids assembled in two separate contact patches (results of the present study and [42]). In a concerted manner both binding sites undergo very large changes as a result of the glucose-dependent global conformational transition associated with enzyme activation and manifested by a marked increase of TF. In the course of this process GKAs gain access to their receptor site by opening of the Val62–Gly72 loop, which is blocked off in the ligand-free enzyme and, as it seems, partly because Trp99 is moved out of the way [21], whereas GKRP loses its grip on GK because critical contact amino acids are pushed apart (Glu51 and Glu52 are now widely separated from Lys140–Lys144 and Met197). This mechanism explains the mutually exclusive binding of activator and inhibitor and also demonstrates again that glucose is the prime mover.

The experimental results and the discussion of this study contribute to the emerging understanding of structure/function relationships of GK and its physiological and pharmacological regulators and modifiers. However, a complete understanding of the complexities of GK function and structure will remain elusive until the interactions with other significant binding partners are also explored both by solution biochemistry, crystallography and cell-biological methods. Reaching this goal is of high biological and medical importance, because GK is an essential player in glucose homeostasis, serving as glucose sensor in the insulin- producing pancreatic β cells and as regulator with high control strength for hepatic glucose metabolism and also because it is the target of GKAs, a new class of antidiabetic agents with medical potential.


Bogumil Zelent participated in the conceptualization of the research plan, performed all biophysical experiments of the present study and contributed critically to the interpretation of the results. He also designed all tabular and most graphic presentations of the results and participated in a critical manner in the preparation of the manuscript; Stella Odili performed the majority of the enzyme kinetic measurements of the present study; Carol Buettger prepared all recombinant GK proteins and performed the kinetic analysis of a significant fraction of wild type and mutant GKs; Dorothy Zelentc arried out all kinetic studies of GKRP inhibition of GK and contributed essentially to the interpretation of the data; Pan Chen made and validated a large majority of the mutant GK DNAs; Deborah Fenner generated and characterized two unique inactivating GK mutants explored in the present study by using ENU mutagenesis in mice; Joseph Bass is team leader of the ENU mutagenesis program at NWU. He articipated in the conceptualization of the project and contributed to the interpretation of the data and the drafting of the manuscript. Charles Stanley provided cDNAs of unique activating GK mutants discovered in his laboratory and used for this study. Participated in the planning of the study, the interpretation of the results and the preparation of the manuscript; Monique Laberge participated as molecular dynamics expert in the interpretation of the results and the preparation of the graphic illustration of the allosteric regulation of GK; Jane Vanderkooi participated as expert spectroscopist in the conceptualization, realization and interpretative aspect of the present study. She also generated a first rough draft of the manuscript which provided the basis of the final versions of this publication prepared by the senior author. Ramakanth Sarabu is a pharamceutical Chemist at Hoffmann La Roche who prepared and provided the GKA used in the present study. He also critiqued the manuscript prior to submission. Joseph Grimsby was co-discoverer of GKAs while at Hoffmann La Roche. He provided recombinant human GKRP for the present experiments. He participated essentially in the long-term planning and realization of the present study. He critiqued the manuscript prior to submission. Frank Matschinsky originated the project and directed the extensive team effort that was needed to realize the wide-ranging and technically demanding project. He conceptualized the organization of the manuscript and prepared the ultimate draft for submission.


This work was supported by the National Institutes of Health NIDDK (National Institute of Diabetes and Digestive and Kidney Diseases) [grant numbers 22122 (to F.M.M) and 19525 (to J.M.V.)] and the FQRNT (Fonds Quebecois de la Recherche sur la Nature et les Technologies) [grant number 2010-PR-13387 (to M.L.)].

Abbreviations: AI, activity index; ANP, phosphoaminophosphonic acid-adenylate ester; BAD, Bcl-2-associated death promoter; DTT, dithiothreitol; F1P, fructose 1-phosphate; GK, glucokinase: GKA, GK activator drug: GKRP, glucokinase regulatory protein; GSIR, glucose-stimulated insulin release; GST, glutathione transferase; HI, hyperinsulinism; huGKRP, human GKRP; MODY-2, maturity onset diabetes of the young; TF, tryptophan fluorescence


View Abstract