Epigallocatechin-3-gallate potently inhibits the in vitro activity of hydroxy-3-methyl-glutaryl-CoA reductase.

Hydroxy-3-methyl-glutaryl-CoA reductase (HMGR) is the rate-controlling enzyme of cholesterol synthesis, and owing to its biological and pharmacological relevance, researchers have investigated several compounds capable of modulating its activity with the hope of developing new hypocholesterolemic drugs. In particular, polyphenol-rich extracts were extensively tested for their cholesterol-lowering effect as alternatives, or adjuvants, to the conventional statin therapies, but a full understanding of the mechanism of their action has yet to be reached. Our work reports on a detailed kinetic and equilibrium study on the modulation of HMGR by the most-abundant catechin in green tea, epigallocatechin-3-gallate (EGCG). Using a concerted approach involving spectrophotometric, optical biosensor, and chromatographic analyses, molecular docking, and site-directed mutagenesis on the cofactor site of HMGR, we have demonstrated that EGCG potently inhibits the in vitro activity of HMGR (Ki in the nanomolar range) by competitively binding to the cofactor site of the reductase. Finally, we evaluated the effect of combined EGCG-statin administration.

nonconventional treatments for hypercholesterolemia, such as novel drugs based on a natural product, tea catechins (mainly gallate ester derivatives) have been tested successfully both in vitro and in vivo as cholesterol-lowering agents (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25). Globally, these fi ndings indicate that catechins exert concerted, multiple-level action involving the upregulation of the LDL receptor ( 15,24 ), the reduction of cholesterol absorption ( 16,26,27 ), and the modulation of both synthetic ( 15,28,29 ) and metabolic pathways ( 24 ). Owing to the complexity of action and the short halflife of HMGR ( 30 ), the direct effect on HMGR activity is very diffi cult to recognize and isolate, and the literature supporting a physiologically signifi cant inhibition of HMGR by catechins is limited ( 22 ). In this work, in order to clarify the direct interplay between epigallocatechin-3gallate (EGCG) and HMGR, supported by experimental evidence on analogous enzymatic systems that point to the presence of a high-affi nity binding site for EGCG on reduced nicotinamide adenine dinucleotide phosphate (NADPH) (31)(32)(33)(34)(35)(36)(37), we sought to identify site-directed binding of EGCG to HMGR by discussing the results of a predictive structural bioinformatic approach. In particular, on the basis of the information derived from computational analysis, we performed both enzymatic and binding studies, the results of which show collectively and unequivocally that EGCG strongly inhibits HMGR activity, and in all likelihood has a profound effect on the synthetic pathway of cholesterol.

Abstract Hydroxy-3-methyl-glutaryl-CoA reductase (HMGR)
is the rate-controlling enzyme of cholesterol synthesis, and owing to its biological and pharmacological relevance, researchers have investigated several compounds capable of modulating its activity with the hope of developing new hypocholesterolemic drugs. In particular, polyphenol-rich extracts were extensively tested for their cholesterol-lowering effect as alternatives, or adjuvants, to the conventional statin therapies, but a full understanding of the mechanism of their action has yet to be reached. Our work reports on a detailed kinetic and equilibrium study on the modulation of HMGR by the most-abundant catechin in green tea, epigallocatechin-3-gallate (EGCG). Using a concerted approach involving spectrophotometric, optical biosensor, and chromatographic analyses, molecular docking, and sitedirected mutagenesis on the cofactor site of HMGR, we have demonstrated that EGCG potently inhibits the in vitro activity of HMGR ( K i in the nanomolar range) by competitively binding to the cofactor site of the reductase. Supplementary key words cholesterol • reduced nicotinamide adenine dinucleotide phosphate • epigallocatechin-3-gallate • inhibition Atherosclerotic cardiovascular diseases are frequent in individuals with abnormally high levels of blood cholesterol (1)(2)(3). Early stages of cholesterol biosynthesis in humans are rate-regulated by 3-hydroxy-3-methyl-glutaryl-CoA reductase (HMGR) ( 4, 5 ), a two-domain endoplasmic reticulum-bound enzyme involved in the mevalonate pathway ( 6 ). Because of this central role, HMGR is the target of several hypocholesterolemic drugs, of which statins are the most extensively studied ( 7-9 ) and among the most widely prescribed drugs worldwide, despite some recurrent side-effects (10)(11)(12)(13)(14). Recently, in efforts to identify from the Protein Data Bank ( 44 ). Hydrogen atoms were added to the protein prior to any analysis. Autodock, a software performing a Lamarckian genetic algorithm to explore the binding possibilities of a ligand in a binding pocket ( 45 ), was used with a grid of 48, 48, and 48 points (in the x, y, and z directions) around the HMG-CoA binding site, and with a grid of 54, 50, and 52 points around the NADPH binding site, with a grid spacing of 0.375 Å, a root-mean-square (rms) tolerance of 0.8 Å, and a maximum of 2,500,000 energy evaluations. Other parameters were set to default values ( 46 ). Next, InsightII was used to refi ne the ligand/ receptor models through an energy minimization procedure, using the Discover module of the software with a consistent valence force fi eld and a conjugate gradients algorithm to an rms derivative of 0.001 kcal/mol. Finally, ligand-receptor binding affi nities were calculated using the experimentally validated ( 47 ) scoring functions of the Ludi algorithm, which takes into account fi ve contributions: total number of hydrogen bonds, perturbed ionic interactions, lyophilic interactions, frozen internal degrees of freedom of the ligand, and loss of translational and rotational entropy of the ligand. Binding affi nities were expressed throughout as predicted equilibrium dissociation constants ( K D,p ) derived from the Ludi Score, according to the standard relation (equation 1): and Energy Estimate 3 as scoring function ( 48 ). The output from InsightII, all modeling studies, and images were rendered with PyMOL (Python Molecular Graphics -2006; DeLano Scientifi c LLC, San Carlos, CA). PyMOL was also used to calculate the length of theoretical hydrogen bonds, measured between the hydrogen and its putative binding partner.

Assay of HMG-CoA reductase activity
HMGR catalyses the four-electron reduction of HMG-CoA to CoA and mevalonate. The spectrophotometric assay of HMGR activity was derived from previously published methods ( 7 ), performing enzyme rate measurements with respect to both HMG-CoA and NADPH. Briefl y, HMGR (4 nM) was incubated at 37°C with increasing concentrations of EGCG in 100 mM phosphate buffer containing 1 mM EDTA, 10 mM DTT, and 2% DMSO at pH 6.8 (activity buffer).
HMGR enzymatic activity was continuously monitored after addition of both free and EGCG-preincubated enzyme to assay

Induction and expression of recombinant HMGR
Expression and purifi cation of the human HMGR catalytic subunit were carried out as described previously ( 38 ). Briefl y, the pGEX-cs plasmid was transformed into BL21 Escherichia coli strains containing the pLysC plasmid according to a standard protocol ( 39 ). Bacteria were grown at 37°C in 100 ml Luria-Bertani medium containing ampicillin at 60 µg ml Ϫ 1 and chloramphenicol at 30 µg ml Ϫ 1 to the A600 of 0.6. Overproduction of recombinant protein was induced by adding isopropyl ␤ -Dthiogalactopyranoside (IPTG) to the fi nal concentration of 0.4 mM. Next, cell growth was prolonged for an additional 3 h essentially as described by Parks et al. ( 40 ). A control culture was grown under the same conditions in IPTG-free medium. Fusion protein was purifi ed by GST-glutathione affi nity chromatography at 4°C, and its purity was fi nally checked by SDS-PAGE.

Preparation of liver microsomes
Human liver microsomes were prepared as previously reported ( 41 ). Liver tissue samples (0.5 g) were added to 4.5 ml of cold homogenization buffer (50 mM Tris-HCl buffer, 0.3 M sucrose, 10 mM EDTA, 10 mM DTT, and 50 mM NaCl at pH 7.4, in the presence of 1 mM PMSF and 1 mM TPCK) and homogenized using a bench-top Ultra-Turrax TP 18/10 homogenizer (Janke and Kunkel; Staufen, Germany). The homogenate was centrifuged at 20,000 g for 15 min at 4°C. The supernatant was collected and centrifuged at 100,000 g for 60 min at 4°C. The microsomal pellet was fi nally resuspended in the activity buffer. Total protein concentration was determined according to the method of Lowry et al. ( 42 ).

Bioinformatic analysis
To identify the most probable binding site for EGCG on HMGR, preliminary molecular docking analyses were performed on a Pentium4/Linux Red Hat-based platform using Autodock 4.0 and InsightII (release 2005) software. The X-ray crystal structure of human HMGR [pdb entry: 3CCT ( 43 )] was retrieved mixtures containing NADPH and HMG-CoA. To clarify the mechanism of the interaction, the apparent dissociation constant ( K d,app ) of the preformed HMGR-EGCG complex on the NADPH binding site was determined by measuring the decrease in catalytic activity upon addition of increasing levels of EGCG at different cofactor concentrations in the range 30-270 M.
After 20 min preincubation (more-prolonged preincubation periods did not affect HMGR activity further; Fig. 3 F), residual activities were measured at 340 nm by continuously monitoring the disappearance of NADPH after the initiation of the reaction. The residual activity ( a i ) was expressed as the ratio of the initial velocities of the product formation in the presence ( v 0,i ) and in the absence ( v 0 ) of a given EGCG con- The experimental dataset was constituted by a set of residual activities ( a i ) measured at increasing EGCG concentrations [ I i ].  For each cofactor concentration (m = 4), the residual activity was measured at increasing EGCG concentrations (n = 10), and the resulting n×2×m matrix was globally analyzed by Marquardt-Levenberg nonlinear least squares fi tting procedure ( 50 ) using equations 3 and 4 as fi tting functions.

Chromatographic assay of HMGR activity
HMGR residual activity assays were performed upon 20 min preincubation of the enzyme (0.4 µM) with increasing levels of EGCG (0-6.54 mM). The preformed EGCG-HMGR complex was added to 1.55 µM HMG-CoA and 2.68 mM NADPH

Biosensor binding studies
The CMD surface was rinsed and equilibrated at 37°C with PBS-T [10 mM Na 2 HPO 4 , 2.7 mM KCl, 138 mM NaCl, 0.05% (v/v) Tween-20, pH 7.4] in the vibrostirred sensing chamber. Prior to activation, the surface was washed with detergent-free PBS, pH 7.4, to avoid a possible 'mask' effect of carboxyl groups. Next, the CMD surface was standardly activated with an equimolar EDC/NHS mixture ( 52 ). Then, HMG-CoA reductase (0.2 mg/ml) dissolved in the immobilization buffer [10 mM CH 3 COONa, pH 5.5, chosen on the basis of the isoelectric point (pI) of the enzyme (pI = 6.8)] was added and incubated for 10 min. Nonspecifi cally-bound ligand was removed by PBS wash, whereas nonreacted carboxylic groups were deactivated with 1 M ethanolamine, pH 8.5. Finally, EGCG was added at increasing concentrations. Raw data were analyzed with Fast Fit software (Fison Applied Sensor Technology; Affi nity Sensors); the software makes it possible to use both mono-exponential and bi-exponential models ( 53 ) to fi t experimental data. Kinetic raw data were globally fi tted ( 54, 55 ) using a standard monophasic time course equation (equation 5): where k ass and k diss are the association and dissociation rate constants, respectively, R t is the response at time t , R 0 is the initial response, and R eq is the maximal response at equilibrium for a given EGCG concentration (equation 6): and R max is the extent of binding at asymptotically high concentrations of [ L ].

Production of G807D mutant
The G at position 2470 on the sequence of HMGR ( 56 ) was replaced with an A to change the amino acid residue from Gly to Asp. The site-directed mutagenesis was carried out by the Quick Change Kit (Stratagene) using pGEX-cs plasmid as template ( 40 ) and the mutagenic oligos HMG-F11M (5 ′ -GGAACGGT GGGT G-ATGGGA CCAACCTAC-3 ′ ) and HMG-R22M (5 ′ -GTAGGTT GGT-CCCATC A CCC ACCGTTCC-3 ′ ). The mutation was confi rmed by DNA sequencing according to the dideoxy chain termination reaction ( 39 ).

EGCG/pravastatin combined effect on HMG-CoA reductase activity
The combined effect of pravastatin and EGCG on the enzymatic activity of HMGR was evaluated, monitoring the effect of increasing levels of the statin in the range 0.6-54 µM both in the presence and in the absence of a fi xed concentration of the catechin ([EGCG] = 66 µM). The data on the effect of EGCG on pravastatin binding to HMGR were analyzed according to a classic linkage relation: in detail, we considered a macromolecule M (namely HMGR) having two heterotropically associated binding sites, targeted by pravastatin (X) and EGCG (Y), respectively. The macromolecule M can exist in four states: M, MX, MY, and MXY, only M being productive ( Fig. 1 ). Therefore, the inhibition of HMGR by pravastatin in the presence of EGCG is described by a general phenomenological equation ( 57 ) (equation 7): where ␤ 1 , 1 , and ␤ 2 are the stepwise equilibrium association constants for the equilibria described in Fig. 1 .

Bioinformatic analysis
Docking studies of EGCG and NADPH on the X-ray crystal structure of the human HMGR disclosed new insights regarding both the interaction strength and the binding geometry of the complex. The cofactor binding to its own site revealed a lower affi nity ( K D,p = 220 nM). This difference can be essentially attributed to the lower entropic contribution due to conformational degrees of freedom; in fact, the docking of NADPH into its own site forced the ligand and the binding pocket into morerestricted conformations with respect to EGCG (1 and 4 conformations per lowest energy cluster, respectively), resulting in a lower value of conformational entropy and an unfavorable contribution to the total free energy of binding ( 7 ).
Additionally, EGCG was docked into the substrate binding site (predicted K D,p = 2.01 µM), with a resulting 30-fold lower affi nity with respect to the NADPH binding site, and a VdW contribution of Ϫ 8.76 kcal/mol (electrostatic ener-site was calculated to be the most likely to accommodate an EGCG molecule ( Fig. 2 ), with a dominant nonelectrostatic energy contribution. In fact, as derived from the Ludi Score (equation 1), EGCG showed a high affi nity for the cofactor binding site ( K D,p = 71 nM), and thus is a strong candidate for NADPH competitor. In detail, the predicted EGCG-HMGR complex was characterized by the formation of 11     gies were comparable), supporting the hypothesis of EGCG essentially as an NADPH site-directed ligand.

Effect of EGCG on HMGR activity
A preincubation period (20 min) was required to allow the establishment of equilibrium between the enzyme, the inhibitor, and the enzyme-inhibitor complex ( Fig. 3 F ), in particular at EGCG concentrations lower than 1 µM (for equal treatment time, a 20 min preincubation was performed throughout). The initial velocity of HMGR was manifestly dependent on EGCG, as unequivocally shown in the residual activity plot; in particular, in the presence of EGCG, we obtained hyperbolic inhibition isotherms, whose steepness decreased at increasing cofactor concentrations ( Fig. 3 ). At saturating NADPH, the apparent dissociation constants ( K d,app ) of the complex HMG-CoA/EGCG increased linearly with cofactor concentrations, unambiguously suggesting the competitive binding of tea catechin on the NADPH binding site ( Fig. 3 ) ( 49 ). The experimentally measured equilibrium dissociation constant for the EGCG/HMG-CoA complex at the cofactor site ( K d = 77.0 ± 12.0 nM) was in excellent agreement with the computationally predicted value. Comparative analysis showed that the inhibition of HMGR activity measured without preincubation was signifi cantly lower, in agreement with the time required for the establishment of the equilibrium between HMGR and EGCG ( Fig. 3 F).
No relevant differences in K d were observed when experiments were repeated at different HMG-CoA concentrations.

Michaelis-Menten parameters
The Michaelis constant ( NADPH m K ) was experimentally determined by plotting the initial rate ( V 0 ) against increasing Fig. 7. Residual activity plot of HMGR in the presence of EGCG ( ), pravastatin ( ᭡ ), and pravastatin and EGCG ( ᭹ ). Raw data for the HMGR inhibition by pravastatin in the presence of EGCG were fi tted to equation 7. , V max , and the catalytic constant K cat for each inhibitor concentration were derived from raw data analysis. As shown in Fig. 4 (

Biosensor binding studies
A more-detailed evaluation of the binding kinetics of EGCG to the enzyme was performed on an IAsys plus biosensor system. The sensing surface containing anchored HMGR was optimized on the basis of different experiments on the variations of enzyme concentration and immobilization buffer composition. Under the experimental conditions described in the Materials and Methods section, readout of 800 arcsec was obtained, corresponding to a partial Langmuir monolayer for a 90 kDa protein (1.30 ng mm Ϫ 2 ); this "conveniently-low" HMGR surface density minimized possible hindering effects during recognition events. Then, EGCG was added at increasing concentrations in the range 0.4-6 µM, and association kinetics were followed up to equilibrium. Dissociation steps were performed by addition of fresh PBS buffer ( Fig. 5A ), whereas baseline recovery was achieved by multiple PBS-T washes at pH 7.4, because detergents decreased the stability of the complex. The longevity of the bio-layer was also tested; surface stability and enzyme activity were retained after at least 50 association/dissociation events. This experimental approach confi rmed the high-affi nity interaction between soluble EGCG and blocked HMG-CoA reductase ( K D = 88.8 ± 31.4 nM), both association ( k ass = 29,500 ± 2,900 M Ϫ 1 s Ϫ 1 ) and dissociation ( k diss = [2.6 ± 0.9]·10 Ϫ 3 s Ϫ 1 ) phases signifi cantly contributing to the stabilization of the complex. Monophasic time courses were always observed upon the addition of soluble EGCG, both in the presence and in the absence of saturating HMG-CoA, whereas presaturation of the cofactor site with 1 mM NADPH prior to EGCG addition substantially annihilated any response.
To assess the validity of the monophasic model in fi tting each time course, a standard F-test procedure was used; the biphasic model was statistically nonsignifi cant at 95% confi dence ( 58 ). Time courses measured at several ligand concentrations were globally analyzed using equations 5 and 6, which share common parameters ( k ass , k diss , and R max ).
The binding response at equilibrium (extent of binding) was calculated for each time course with a fully comparable equilibrium dissociation constant ( * D K = 98.9 ± 23.8 nM). The hyperbolic nature of the saturation plot ( Fig. 5B ) demonstrated the noncooperative binding of EGCG to human HMG-CoA reductase.

Chromatographic assay of HMGR activity
According to the RP-HPLC assay of HMGR activity in both bacterial and human liver microsomes, we observed an evident decrease in both HMG-CoA/NADPH consump-DPH binding site: in this assay, NADPH presaturated enzyme completely lost the ability to bind to EGCG, whereas this ability was effectively fully retained upon presaturation with the HMG-CoA (see supplementary material). Again, the spectrophotometric assay of HMGR activity provided further confi rmation of exclusive competition between EGCG and NADPH for the same site, because the residual activity increased with NADPH; conversely, increases in HMG-CoA concentration did not signifi cantly affect HMGR residual activity. Finally, the increase in apparent equilibrium dissociation constants with NADPH ( Fig. 3 E), and the equilibrium shift between HMGRbound and soluble EGCG upon buffer wash (see supplementary material) clearly demonstrated the reversibility of the interaction.
Site-directed mutagenesis ultimately substantiated these results; in fact, the replacement of glycine at position 807 of the cofactor site by an aspartate residue was crucial ( Fig.  2 B) in almost annihilating the binding affi nity of HMGR for EGCG.
On the strength of these results, the physiological relevance of HMGR inhibition by EGCG can be asserted by considering the collective data available on the in vivo levels of NADPH ([NADPH] + [NADP + ] = 20-370 g/g wet liver weight) ( 59 ) and EGCG ([EGCG] = 0.3-7.5 M in the blood of normal green tea consumers) ( 60 ), and their respective binding affi nities for NADPH binding site ( K m,NADPH = 21 M (7), and K D,EGCG = 77 nM).
Hence, on the basis of these parameters, we can reasonably assert that inhibition of the enzyme occurs under physiological conditions, even irrespective of the possible binding of EGCG to other enzymes having NADPH/ NADP + as cofactor, but not sharing the same binding affi nity. HMGR binding affi nity for EGCG is up to 10,000-fold higher than the other NADPH enzymes previously studied, and owing to the concentration of EGCG in blood/ liver upon oral administration ( 60 ), HMGR is likely to be inhibited by the catechin, even in the presence of the other "EGCG-sequestering agents." Nevertheless, the wide spectrum of EGCG molecular targets opens other scenarios on the modulation of HMGR activity, because EGCG is likely also to indirectly inhibit HMGR by activating AMP-activated protein kinase ( 61,62 ), which in turn is responsible for the phosphorylation and inhibition of several enzymes, including HMGR ( 63 ).
Similarly, according to our results, the high specifi city of EGCG for the NADPH site of HMGR is particularly interesting from a pharmacological point of view, inasmuch as the EGCG-statin synergistic inhibition of HMGR provides consistent information in the perspective of the coadministration of catechin-rich food and beverages during statin treatment of hypercholesterolemic states. Collectively, the results presented here confi rm the role of tea catechins, EGCG in particular, as precursors for the development of novel drugs for the treatment of hypercholesterolemic disorders.
The authors would like to thank Professor J. Deisenhofer for providing the expression plasmid of the human HMGR, and tion and mevalonate/NADP + production rates upon 20 min preincubation of HMGR with increasing EGCG ( Fig.  6 ). In detail, at 2.68 mM NADPH, we obtained comparable values of K d,app for the bacterial-expressed and liver HMGR (91.2 ± 23.4 µM and 109.9 ± 35.3 µM, respectively), in strong agreement with the linear relation between K d,app and NADPH concentration derived from the spectrophotometric inhibition assay.

Mutagenesis of Gly-807 in the cofactor site of HMGR
To further ascertain the EGCG selective binding at the cofactor site HMGR, we constructed a point mutation of Gly 807 . Gly 807 was shown not to infl uence catalysis, because the G807D mutant retained the in vitro activity toward HMG-CoA with respect to wild-type HMGR. Conversely, the substitution was critical in diminishing (and nearly abolishing) the capacity of the mutant to bind EGCG, with a K d,app at least 400-fold higher than wild-type HMGR ( Fig. 6 B).

Combined effect of EGCG/pravastatin on HMG-CoA reductase activity
We explored the effect exerted on HMGR activity by increasing levels of the statin in both the presence and absence of a fi xed concentration of the catechin. Analysis of raw data according to equation 7 showed that the affi nity of pravastatin for HMGR shifted from ␤ 1 (10 8 M Ϫ 1 ) to ␤ 2 (0.47·10 8 M Ϫ 1 ) upon binding of EGCG at the cofactor site, eliciting a mild negative allosteric heterotropic effect ( ␤ 2 / ␤ 1 = 0.47) in the binding of the two inhibitors to HMGR. The inhibition observed upon coadministration of EGCG and pravastatin was higher than the effects exerted by each single inhibitor, but (slightly) lower than the total expected theoretical effect ( Fig. 7 ).

DISCUSSION
Gallate catechins were demonstrated to modulate the expression and activity of cholesterogenic enzymes, but to date, both in vivo and in vitro studies hardly linked the hypocholesterolemic effect of catechin to the direct modulation of HMGR.
The present work showed for the fi rst time that EGCG (commonly the most-abundant tea catechin) potently and reversibly inhibits HMG-CoA reductase from different sources to a comparable extent. The integrated experimental approaches used to derive both equilibrium and kinetic parameters ( Table 3 ) concordantly proved EGCG to be a competitive ligand of HMGR at the cofactor site ( K D in the nanomolar range), emphasizing the signifi cant inference of both association and dissociation events in the stabilization of the complex. In detail, from a thermodynamic point of view, the mono-exponential nature of the binding kinetics, the signifi cant values of both rate and equilibrium constants, and a plateau value for the R max at "saturating" EGCG values ( Fig. 5B ) collectively provided solid evidences for the presence of a highly specifi c binding site for EGCG on HMGR. On the other hand, the biosensor competitive binding assay clearly demonstrated that EGCG selectively overlaps with the NA-