Kinetic and Structural Characterization of the Glutathione-binding Site of Aldose Reductase*

reductase (AR), a member of the aldo-keto reductase superfamily, has been implicated in the etiology of secondary diabetic complications. However, the physiological functions of AR under euglycemic conditions unclear. We have recently demonstrated that, in intact heart, AR catalyzes the reduction of the glutathione conjugate of the lipid peroxidation product 4-hy-droxy- trans -2-nonenal Biol. Chem. 10893–10900), consistent with a possible role of AR in the metabolism of glutathione conjugates of aldehydes. Herein, we present several lines of evidence suggesting that the active site of AR forms a specific gluta-thione-binding

Aldose reductase (AR) 1 is an NADPH-dependent aldo-keto reductase that catalyzes the reduction of a wide variety of aldehydes, including glucose (for review, see Ref. 1). Several lines of evidence indicate that increased flux of hexoses via the AR-catalyzed pathway is one of the underlying causes of tissue injury and dysfunction associated with hyperglycemic states such as diabetes mellitus and galactosemia. In experimental models of diabetes and galactosemia, pharmacological inhibition of AR attenuates, prevents, and/or delays several pleiotropic complications (1,2). Conversely, in transgenic animals, lens-specific up-regulation of AR accelerates sugar cataract (3), providing evidence for a critical role of this enzyme in the genesis of hyperglycemic injury. Nonetheless, clinical trials with AR inhibitors have yielded uncertain results (4,5), and the long-term efficacy of these drugs in treating diabetic complications remains to be demonstrated. Although the variable results obtained in clinical trials of AR inhibitors may be due, in part, to their poor specificity and an inappropriate dosing schedule, a major impediment to their long-term clinical acceptance relates to a lack of understanding of the normal physiological role of AR.
Recent evidence suggests that AR and its isoforms may be involved in several physiological functions such as cell growth and/or differentiation. The expression of AR as well as a similar murine aldo-keto reductase, FR-1 (fibroblast growth factorregulated protein-1), is enhanced by growth factors such as fibroblast growth factor and epidermal growth factor (6 -8). Moreover, an AR-related protein is one of the most prominent antigens up-regulated during hepatocarcinogenesis (9). In addition, AR may be involved in several cell type-specific functions such as osmoregulation (10), fructose (11) and tetrahydrobioprotein (12) synthesis, and the metabolism of corticosteroids (13). However, the most general role of AR may be detoxification of reactive aldehydes. Both AR and FR-1 display high catalytic efficiency with unbranched saturated and unsaturated aldehydes (14 -17). Since such aldehydes are the major bioactive end products of lipid peroxidation (18) and reduction diminishes their bioactivity, these enzymes may be important components of the cellular antioxidant defenses.
In vitro studies indicate that the high efficiency of AR in catalyzing the reduction of lipid-derived aldehydes is, in part, due to the high hydrophobicity of the enzyme's active site, which interacts favorably with long alkyl chains (17). Furthermore, because the active site residues of AR do not form extensive hydrogen bonds with the substrates, the enzyme appears to be ill-suited for the reduction of hydrophilic aldehydes such as glucose. These characteristics of the enzyme suggest that a more general role of AR may be reduction of medium-to longchain aldehydes. However, in most cells, reduction to alcohol is a minor metabolic fate of aldehydes which are readily oxidized to acids. For instance, oxidation to 4-hydroxynonanoic acid accounts for 40 -50% of the metabolism of 4-hydroxy-trans-2nonenal, whereas the alcohol 1,4-dihydroxynonene represents only a minor (Ͻ10%) component (18,19). Additionally, unsaturated aldehydes, because of their high electrophilicity, readily form covalent adducts with glutathione. As a result, in most glutathione-competent cells, unsaturated aldehydes generated by lipid peroxidation are likely to be readily conjugated with glutathione, and aldehyde oxidoreductases are likely to be presented with an glutathione-aldehyde conjugate, rather than the free aldehyde. Indeed, our studies show that in addition to reducing 4-hydroxy-trans-2-nonenal, AR is also an efficient catalyst for the reduction of the glutathione adduct of 4-hydroxy-trans-2-nonenal (14,17). Nonetheless, the specificity of the enzyme for glutathione-aldehyde conjugates of varying structure has not been examined, and the role of glutathione in facilitating aldehyde binding to the active site of AR remains poorly defined. This study was therefore undertaken to identify the kinetic and structural features of AR that determine its interaction with glutathione conjugates.
Synthesis of Glutathione Conjugates-GSH was incubated with individual compounds at a 1:1 molar ratio in 0.1 M potassium phosphate (pH 7.4) at room temperature. The reaction was monitored by following the decrease in the max of aldehydes. Free GSH content was monitored by 5,5Ј-dithiobis(2-nitrobenzoic acid). The conjugates were purified by HPLC using a Rainin reverse-phase ODS C 18 column (1 ϫ 30 cm) pre-equilibrated with 0.1% trifluoroacetic acid in water (solvent A) at a flow rate of 1 ml/min. Approximately 1 ml of conjugate (1-10 mol) was applied to the column, and the eluant was monitored at 220 nm. The conjugates were eluted using a gradient consisting of solvent A and solvent B (acetonitrile) at a flow rate of 1 ml/min. The gradient was established such that solvent B reached 10% in 20 min and 25% in 35 min and held at 25% for 30 min. In an additional 10 min, solvent B reached 60%. In this system, the conjugates eluted with a retention time of 30 -45 min.
Amino Acid Analysis-The concentration of the glutathione conjugates was determined by amino acid analysis. After lyophilization, the conjugates were hydrolyzed in a vapor-phase automated hydrolyzer (Perkin-Elmer). The resulting amino acids were derivatized with phenyl isothiocyanate to form phenyl isothiocyanate-derivatives, which were extracted and transferred to an on-line 130/A HPLC apparatus for analysis using an Applied Biosystems 420/H PTC amino acid analyzer/ hydrolyzer. In each assay, 100 -300 pmol (0.1-0.2%) of the sample were analyzed, which is well within the range of sensitivity of the instrument (25 pmol to 2 nmol).
Electrospray Ionization/Mass Spectrometry (ESI/MS)-The molecular identity of the GS-aldehyde conjugates was established by electrospray mass spectrometry. The electrospray mass spectrometry experiments were performed on a Micromass ZMD single quadrapole electrospray mass spectrometer. The tuning and calibration solution consisted of polypropylene glycol 2000 in water/methanol (50:50, v/v) containing 0.1% acetic acid. Additional calibration around m/z 300 was performed by using GSH, which displayed a well resolved peak at m/z 307. The capillary, cone, and extractor were operated at 3.5, 40, and 3 V, respectively, with 40 p.s.i. N 2 at a flow rate of 0.5 liter/min. The source block and desolvation temperatures were set at 80 and 200°C, respectively. Typically, a 10 -20 M solution of the conjugates was prepared in acetonitrile/water/acetic acid (50:50:0.1, v/v/v) and injected into the ion source of the spectrometer using a Harvard syringe pump at a flow rate of 5-10 l/min. The mass spectrometer was set to scan from m/z 100 to m/z 750 with a step size of 0.5, a dwell time of 2 ms, and a scan speed of 45. When dimers of dehydrated ions were observed, the cone voltage and the source block temperature were varied to optimize formation of the parent molecular ion.
Reduction of Recombinant Enzyme-Before each experiment, stored recombinant AR was reduced by incubation with 0.1 M DTT at 37°C for 1 h in 0.1 M potassium phosphate (pH 7.0). Excess DTT was removed by gel filtration using a Sephadex G-25 column (PD-10) pre-equilibrated with N 2 -saturated 0.1 M potassium phosphate (pH 7.0) containing 1 mM EDTA. All operations were carried out at 4°C to prevent oxidation of the enzyme. The DTT-free, reduced enzyme was stored under N 2 and used within 1 h.
Enzyme Assay-Enzyme activity was measured at room temperature in a 1-ml reaction system containing 0.1 M potassium phosphate, 1 mM EDTA (pH 7.0), 10 mM glyceraldehyde, and 0.15 mM NADPH. The reaction was monitored by measuring the disappearance of NADPH at 340 nm using a Gilford Response II spectrophotometer or a Varian Cary 100 Bio spectrophotometer. One unit of enzyme activity is defined as the amount of enzyme required to oxidize 1 mol of NADPH/min. The control cuvette (blank) contained all components of the reaction mixture except the enzyme. The substrate concentration was varied over a range extending from 0.2 to 5-7 times the K m . Initial velocity was measured at seven to nine different concentrations of each substrate.
Data Analysis-Individual saturation curves used to obtain V max and K m were fitted to the general Michaelis-Menten equation using a nonlinear iterative fitting program (21) where v is the enzyme velocity and S is the substrate concentration. Data conforming to linear uncompetitive and noncompetitive or competitive inhibition were fitted to Equations 2-4, respectively, where I is the inhibitor concentration and K ii and K is are the intercept and slope inhibition constants, respectively. In all cases, the best fit to the data was chosen on the basis of the standard error of the fitted parameter and the lowest values of , which is defined as sum of the squares of the residuals divided by the degree of freedom, i.e. n Ϫ 1, where n represents the number of velocity measurements. Data are expressed as means Ϯ S.E. Molecular Modeling and Molecular Dynamics Calculations-The structure of GS-propanal was constructed from the coordinates of glutathione (Protein Data Bank code 1gra) (22) and 2-cyclopropylmethylenepropanal (Protein Data Bank code 1hrn) (23), with the starting conformation of glutathione the same as that in the crystal structure of glutathione reductase. The chemical structure of GS-propanal is shown in Fig. 1. For AR, the 1.76-Å structure complexed with NADP ϩ and glucose 6-phosphate was used (Protein Data Bank code 2aq) (24) as the starting model. GS-propanal was positioned in the active site of AR using program O (25). The aldehyde moiety of GS-propanal was positioned such that the carbonyl oxygen of the aldehyde is 2.9 Å from both the NE2 atom of His 110 and the hydroxyl group of Tyr 48 and is in the same position as the carboxylate oxygen of zopolrestat (25). This resulted in the carbonyl moiety being parallel to the nicotinamide ring of NADPH. Two distinct starting orientations of GS-propanal, termed orientations 1 and 2, were examined. The propanal group was oriented similarly in both models, and the positioning of the GS-propanal residues ␥Glu 1 and Gly 3 was switched in the two models. The energy of the GS-propanal⅐AR complex was minimized to reduce both steric strain and close van der Waals contacts. Molecular dynamics calculations were then performed for 0.5 ps at 300 K to enable the bound GSpropanal to sample a large part of the conformational space within the AR active site. A 200-step conjugate gradient energy minimization was performed after the molecular dynamics calculation to generate the final models of the GS-propanal⅐AR complex. All energy minimization and molecular dynamics calculations were performed with the X-PLOR program (50). Force-field terms for GS-propanal were constructed from analogous amino acids parameters. To maintain the geometry of the catalytic site, the aldehyde moiety of GS-propanal was constrained with weak distance restraints (nuclear Overhauser effect Restraints) to be within 2.9 Å of the hydroxyl oxygen of Tyr 48 and the NE2 atom of His 110 .

RESULTS
In the first series of experiments, the effect of glutathione conjugation on the catalytic efficiency of AR in reducing unsat-urated aldehydes was examined. For this, glutathione conjugates of acrolein, trans-2-hexenal, trans-2-nonenal, and trans, trans-2,4-decadienal were prepared by incubating the alkenals with GSH at room temperature and neutral pH. The conjugates formed were purified by HPLC, and their concentration was determined by amino acid analysis. In some experiments, [ 3 H]GSH was used to conjugate alkenals; and after HPLC purification, the concentration of the conjugates was calculated based upon the radioactivity. Upon ESI ϩ /MS, the conjugates displayed intense molecular ions with m/z values corresponding to a 1:1 adduct between glutathione and the alkenal (Table  I). Appropriate dilutions of the conjugates were used for the measurement of AR activity. Since conjugation removes unsaturation at C-3, the kinetic constants obtained with the conjugates were compared with the corresponding aldehyde saturated at this position. Under the experimental conditions used, the aldehydes and their conjugates followed Michaelis-Menten kinetics. To ensure that no side reactions or unexpected interactions contaminated the calculated values of the kinetic parameter, the products generated by AR catalysis were examined. For this, the ESI ϩ /MS spectra of the glutathione conjugates before and after reduction by AR were compared. As shown in Fig. 2 for GS-propanal, incubation with AR led to an increase in the m/z value of the conjugate from 364.4 to 366.4, consistent with the reduction of the aldehyde to an alcohol. No significant formation of other ions was detected. Single reduction products were also obtained with other glutathione conjugates (data not shown).
With each of the aldehydes tested, the catalytic efficiency (k cat /K m ) was significantly higher for the conjugate than for the corresponding free aldehyde (Table I). The increases in the catalytic efficiency were, however, variable. A greater enhancement (Ͼ1000-fold) of the catalytic efficiency was observed with the small-chain aldehyde acrolein than with intermediatechain aldehydes such as hexenal and nonenal (3.5-5-fold). The medium-chain unsaturated conjugate GS-trans-4-decenal was reduced with a 16-fold greater efficiency than trans-4-decenal. These results suggest that conjugation with glutathione enhances the catalytic efficiency of AR, particularly for smallchain or unsaturated aldehydes.
To examine the specificity of glutathione in facilitating the aldehyde binding/catalysis at the active site, we measured the kinetic parameters of AR with conjugates in which the peptide backbone was systematically varied while retaining the Cys-aldehyde bond. For this analysis, propanal adducts were used because GS-propanal displayed the greatest enhancement of catalytic efficiency over propanal. The conjugates were prepared by incubating the indicated thiol peptides with acrolein. The propanal conjugates formed intense molecular ions upon ESI ϩ /MS, which accounted for Ͼ90% of the signal. Some of the conjugates examined, e.g. those with cysteine and Cys-Gly, formed 1:2 adducts (two aldehydes conjugated with the amino acid or the peptide). These adducts were not used for further analysis. The remaining conjugates displayed strict 1:1 stoichiometry. Of these, some, e.g. Gly-Cys-Gly and ␥Glu-Cys-Glu, displayed additional peaks (data not shown) that were assigned to dehydrated forms of the conjugate. However, since no peaks corresponding to 1:2 adducts were observed, these conjugates were tested as potential AR substrates. The m/z values of the parent peptides and synthesized conjugates are listed in Table II. Modification of the N-terminal glutamate (␥Glu 1 ) had profound effects on the catalytic efficiency with which the enzyme catalyzed the reduction of the conjugate. As compared with GS-propanal (in which the amide bond is between the ␥-carboxyl of glutamate and the amine of cysteine), the propanal conjugate of ␣Glu-Cys-Gly (in which cysteine is linked to the ␣-carboxyl of Glu) was reduced with much higher efficiency by the enzyme, indicating sensitivity of the active site to the geometry of the N-terminal amide linkage. However, removal of the ␥-carboxyl of Glu or substitution of Glu with Gly diminished catalytic efficiency by 50 -65% (compare ␥-aminobutyric acid-Cys-Gly and Gly-Cys-Gly with GSH), suggesting that the presence of Glu at the N terminus enhances catalytic efficiency due to specific interaction between the enzyme and the ␥-carboxyl group of Glu. The catalytic efficiency was equally sensitive to alterations in C-terminal glycine. Although ␥Glu-Cys-Glu was reduced with efficiency slightly lower than GSH, substitution of the C-terminal glycine with alanine led to a marked reduction in activity and poor catalytic activity. However, transposition of glycine to the N terminus led to a peptide whose conjugate was more efficiently reduced than GSH, indicating that an ␣-amide linkage at the N terminus facilitates catalysis. This is consistent with the higher catalytic efficiency  observed with ␣Glu-Cys-Gly than with GSH. Moreover, removal of the C-terminal glycine, which led to a smaller molecule, enhanced catalysis. The high catalytic efficiency observed with dipeptides was comparable when either glycine or glutamate was linked with cysteine (cf. N-acetyl-Cys-Gly and Nacetyl-Cys-Glu), suggesting that smaller molecules with less structural constraints during binding are more efficiently reduced by the enzyme. A further decrease in the length of the peptide backbone led, however, to a sharp decline in the catalytic efficiency, and the smallest adduct, N-acetylcysteine-propanal, was among the poorest substrates tested, indicating that residues linked to cysteine interact specifically with the active site, providing additional anchoring and stabilization. The catalytic efficiency was also decreased upon increasing the size of the peptide backbone. Thus, ␥Glu-Cys-Ala-Gly was reduced much less efficiently than GSH, and no improvement was evident when the N-terminal glutamate of the tetrapeptide was linked via the ␣-carboxyl rather than the ␥-carboxyl.
In contrast to catalytic efficiency, which was greatly enhanced by several substitutions, none of the changes in the glutathione backbone enhanced the turnover number of the enzyme, which displayed the highest catalytic rate with the GS conjugate. Moreover, in general, di-and tetrapeptides displayed much lower catalytic rates than the tripeptides, although the catalytic cycle with N-acetylcysteine-propanal was faster, presumably in part due to less tight binding.
To further probe the nature of the glutathione-binding site of AR, we examined whether reduction of glutathione-aldehyde adducts is inhibited by glutathione analogs. We reasoned that if GSH specifically binds to the active site, the analogs of glutathione should inhibit catalytic reduction of GS-aldehyde conjugates. Several analogs of glutathione representing fragments of the glutathione molecule or S-derivatives of GSH were examined. Although most of the glutathione analogs were obtained from commercial sources, the glutathione adduct of NEM along with the well studied conjugate of GSH and iodoacetic acid and GS-menadione were synthesized in-house as described under "Experimental Procedures." The synthesized analogs were purified by HPLC and analyzed by ESI ϩ /MS. As shown in Fig. 3, glutathione conjugates of iodoacetic acid, NEM, and menadione displayed predominant monoisotopic molecular ions at m/z 366.2, 433.3, and 478.0, respectively. Since these values correspond to [M ϩ H] ϩ ions of these conjugates and no additional peaks with higher or lower m/z values were observed, we conclude that our procedure of synthesis and purification yields predominantly a 1:1 glutathione adduct of these electrophiles. Several of the glutathione analogs examined displayed inhibition constants in the submillimolar range (Table III). In most cases, the nature of inhibition by the glutathione analogs was uncompetitive versus the aldehyde substrate. Although for a simple, one-substrate/one-product reaction scheme, this would suggest that the inhibitors bind to the E⅐S complex, previous experiments have shown that several AR inhibitors such as sorbinil, tolrestat, and zopolrestat that do not display competitive inhibition (26,27) bind exclusively to the active site of the enzyme (27)(28)(29)(30). This is because the rate-limiting step in AR catalysis is the slow isomerization of the E⅐NADP binary complex (31). Since, under steady state, this form represents 90 -95% of the enzyme (32), most activesite inhibitors preferentially bind to the E⅐NADP complex. As expected from the complete velocity equation of the enzyme reaction scheme (27,29), such inhibitors do not display competitive inhibition patterns versus the aldehyde, which binds exclusively to the E⅐NADPH binary complex. The binding of uncompetitive inhibitors to the active site of the enzyme has been directly demonstrated by x-ray analysis of E⅐NADP crystals bound to zopolrestat (25) or sorbinil or tolrestat (33). The active-site binding of glutathione conjugates is further supported by the observation that two of these compounds, i.e. S-(dicarboxyethyl)glutathione and S-(p-nitrobenzyl)glutathione, displayed competitive inhibition versus the aldehyde substrate (Table III), suggesting preferential binding of these conjugates to E⅐NADPH. Reasons for the selectivity of the inhibitors for E⅐NADPH over E⅐NADP complexes are not clear, but may relate to the greater structural similarity of the competitive inhibitors to the aldehyde substrate. The lowest K i in

. ESI ؉ /MS of glutathione analogs of N-ethylmaleimide (A), menadione (B), and iodoacetic acid (C).
Each of these compounds was reacted individually at a 1:1 molar ratio with glutathione at pH 7.4. The conjugate formed was purified by reverse-phase HPLC and injected into the spectrometer using acetonitrile/water/acetic acid (50:50:0.1, v/v/v) as the flow injection solvent. IAA, iodoacetic acid. this series of inhibitors was obtained with GS-menadione, consistent with the high affinity of the AR active site for the phenol derivatives due to efficient ring stacking (25).
Although no significant inhibition was observed with individual amino acids, the dipeptides ␥Glu-Cys and Cys-Gly were inhibitory. Glutathione itself was less inhibitory than the dipeptides. Nevertheless, even the weak inhibition by GSH (K i ϭ 1.5 mM) may be physiologically significant since most cells maintain an intracellular GSH concentration of 1-10 mM. Since GSH is easily oxidized, we were concerned that inhibition by GSH could be partially accounted for by irreversible inactivation of the enzyme by GSSG. We (34) and others (35) have previously reported that GSSG causes reversible S-thiolation of Cys 298 , located at the active site of the enzyme. To rule out inactivation of AR by GSSG contamination in the GSH solutions, we measured the catalytic activity in the presence of 5 mM GSH and varying amounts of AR (5-25 g) in the 1-ml reaction system. The resultant activity versus concentration curves intercepted the activity axis at zero (data not shown), indicating no significant covalent modification of the enzyme. These observations suggest that reduced glutathione causes reversible enzyme inhibition, which is not accounted for by inactivation due to covalent modification by GSSG arising from spontaneous oxidation of GSH in solution. That inhibition by GSH was not due to covalent modification is also supported by the observation that the enzyme was inhibited by glutathionesulfonic acid with equal efficacy. The sulfonic acid derivative of glutathione does not have a free sulfhydryl; thus, inhibition by this compound, as well as glutathione, suggests that binding of glutathione to the enzyme is independent of the oxidation state of cysteine, but is due to specific recognition of Glu and Gly by the active-site residues. The addition of a long flexible alkyl chain to GSH seemed to slightly increase the extent of inhibition. Nonetheless, no consistent pattern of inhibitory efficacy emerged upon systematically varying the length of the S-alkyl chain.
The low energy conformations of GS-propanal bound in the AR active site are shown in Fig. 4. The AR side chains occupy essentially the same positions in orientations 1 and 2. In both orientations, the propanal moiety is positioned in the hydro-phobic cleft of the AR active site such that no steric clashes occur between the substrate and the enzyme. The sulfur atom of GS-propanal is within 5 Å of the sulfur atom of Cys 298 in both orientations 1 and 2. Table IV lists all contacts between GSpropanal and AR in the two energy-minimized orientations. The interaction energies between GS-propanal and AR are similar in the two orientations examined (Table V). The calculated interaction energies for orientations 1 and 2 are Ϫ33.6 and Ϫ28.8 kcal/mol, respectively. The individual contributions of ␥Glu 1 and Gly 3 of GS-propanal to the interaction energy of orientation 1 are Ϫ11.6 and Ϫ8.0 kcal/mol, respectively. The individual contributions of ␥Glu 1 and Gly 3 of GS-propanal to the interaction energy of orientation 2 are Ϫ13.0 and Ϫ3.0 kcal/mol, respectively. In both orientations, the interaction energy contributed by ␥Glu 1 is much greater than that contributed by Gly 3 . The interaction energy between ␥Glu 1 and AR is greatest in orientation 2, where ␥Glu 1 is positioned to bind to Leu 301 and Ser 302 . In contrast, the interaction energy between Gly 3 and AR is greatest in orientation 1. However, in orientation 1, Gly 3 is positioned to bind to Leu 301 and Ser 302 . Thus, Leu 301 and Ser 302 likely stabilize the binding of both GS and other peptide adducts to AR. DISCUSSION Although AR catalyzes the reduction of glucose during hyperglycemia, the euglycemic role of the enzyme remains obscure. Recent studies show that relative to glucose, mediumchain hydrophobic aldehydes are reduced much more efficiently by AR (14,15,17). These aldehydes, particularly the 4-hydroxyalkenals, are predominant end products of lipid peroxidation. However, due to ␣,␤-unsaturation, the alkenals readily form adducts with glutathione (18,19). Thus, in glutathione-competent cells, these conjugates, rather than free aldehydes, are likely to be the more abundant products of lipid peroxidation. Although our studies show that AR is also an efficient catalyst for the reduction of such glutathione-aldehyde adducts (14,17), the specificity of the enzyme for glutathiolated substrates and the selectivity of the AR active site for glutathione has not been systematically examined. As discussed below, several lines of evidence presented herein suggest that the active site of AR forms a specific glutathione-binding domain and that the glutathiolation enhances the range and the selectivity of the enzyme for aldehydes of diverse structure.
AR Reduces Glutathione Conjugates More Efficiently than Free Aldehydes-As shown in Table I, for the range of aldehydes tested (C 3 to C 10 ), the glutathione conjugate was a better substrate than the corresponding free aldehyde. One possible mechanism by which glutathiolation enhances the catalytic efficiency may be by increasing the extent of hydrophobic/van der Waals interactions between the active site and the substrate. Thus, the high activity with the glutathione conjugates may be simply because these substrates are larger in size and thus provide more opportunity for hydrophobic interactions. However, this seems unlikely. With free aldehydes of varying chain length, the increase in catalytic efficiency was not proportional to the chain length, suggesting that a simple increase in the size of the substrate cannot completely account for the increase in catalytic activity achieved by glutathiolation. Alternatively, the higher efficiency of the enzyme with the glutathione conjugates may be due to specific recognition of the glutathione backbone at the active site. As discussed below, the latter possibility is more consistent with our results. Regardless of the mechanism, efficient reduction of glutathione conjugates of AR suggests that in vivo glutathiolation may be critical in recruiting aldehydes (e.g. acrolein) that would otherwise escape reduction by AR due to structural incompatibility with the active site, thus extending the range of aldehydes accessible to AR.
The Structure of Glutathione Is a Sensitive Determinant of Catalytic Efficiency-Structure-activity relationships suggest that both the N-terminal glutamate and C-terminal glycine are specifically recognized by the active-site residues of AR since substitution of the C-terminal glutamate with glycine or of the N-terminal glycine with alanine led to a significant decrease in catalytic efficiency. Such specificity of AR is comparable to that of most GST isoenzymes, except GST 8-8, which is more stringent and has appreciable activity with only GSH (36). Like , AR was found to accommodate analogues with ␣-glutamate at the C terminus and ␤-alanine at the N terminus and displayed high activity with glutathione analogs such as ␣Glu-Cys-Gly and N-acetylcysteine.
The specificity of AR for glutathione appears to stem partly from the free carboxyl group of the C-terminal glutamate since the decarboxylated glutathione conjugate (␥-aminobutyric acid-Cys-Gly-propanal) was reduced poorly by AR compared with glutathionyl-propanal (Table II). Moreover, the progressively lower efficiencies that were observed as additional groups were added to the C terminus of the Glu-Cys substrate (␥Glu-Cys Ͼ ␥Glu-Cys-Gly Ͼ ␥Glu-Cys-Glu Ͼ ␥Glu-Cys-Ala-Gly) suggest that unfavorable steric interactions with larger molecules interfere with the optimal binding of the entire molecule.
Reduction of Glutathione Conjugates by AR Is Inhibited by Glutathione Analogs-Inhibition of AR catalysis by analogs of glutathione conjugates and glutathione also suggests specific recognition of glutathione by the AR active site. As indicated under "Results," inhibition by glutathione analogs appears to be due to active-site binding. Although both uncompetitive and competitive profiles were obtained, the difference appears to be due to differential preference of the analogs for the E⅐NADP or E⅐NADPH binary complexes. Reasons for such selectivity was not apparent from the present data, but may relate to greater structural similarity of the non-glutathione moiety of the competitive inhibitors to the aldehyde substrate. Furthermore, the relative efficacy with which glutathione analogs inhibit AR provides additional means to assess the specificity of AR for glutathione. It has been previously shown that S-carboxymethylglutathione and S-propylglutathione inhibit glyoxalase I with K i values of 1.3 and 0.3 mM, respectively (39). The corresponding K i values of dicarboxyethylglutathione and S-propylglutathione for AR are 0.19 and 0.59 mM (Table III), indicating that the affinity of AR for glutathione analogs is comparable to that of glyoxalase. However, the most potent inhibition of AR was observed with GS-menadione, which apparently results from the favorable interaction between the aromatic ring of the conjugate with the hydrophobic residues of the AR active site. Reduction of such glutathione conjugates by AR may be of relevance during increased vitamin K exposure. Significantly, AR was also inhibited by free glutathione, with a K i of 1.5 mM (Table III). Since most aerobic cells maintain an intracellular concentration of glutathione between 1 and 10 mM, these results suggest that under physiological conditions, AR catalysis may be regulated by GSH and that the enzyme may be activated by conditions leading to GSH depletion such as long-term diabetes (40) (Fig. 4). Several different crystal structures show that when complexed to proteins, the glutathione molecule adopts an extended conformation (39,42,43). This shape makes an excellent fit to the shape of the AR active site, which is ϳ12 Å deep, 10 Å wide, and 10 Å long. The glutathione moiety of the GS-propanal molecule can bind to the active site of AR in two orientations (orientations 1 and 2) with only minor movements required of the AR residues. In orientation 1, the ␥Glu 1 of GS-propanal forms hydrogen bonds with Trp 20 , Lys 21 , and Val 47 , and Gly 3 forms hydrogen bonds with Leu 301 and Ser 302 (Fig. 4A). In contrast, in orientation 2, the ␥Glu 1 of GS-propanal forms hydrogen bonds with Leu 301 and Ser 302 , and Gly 3 is positioned near Trp 20 , Val 47 , and Tyr 48 . In both orientations, the Cys-propanal moiety of the substrate interacts similarly with His 110 and Tyr 48 . Both the interaction energy and the number of hydrogen bonds between GS-propanal and the active-site residues are greater in orientation 1 than in orientation 2. Thus, orientation 1 is likely to be the preferred mode of GS-propanal binding to AR.
Our molecular modeling studies also suggest that the conformation of glutathione bound to the active site of AR may be similar to that observed with proteins that utilize glutathione as a cofactor. However, the glutathione molecule adopts a variety of conformations. The NMR structure of GSH in solution shows that it is not extended as in the crystalline state, but is folded so as to orient the amine and carboxylate groups in the ␥-glutamic acid and cysteine to accommodate the interaction (39). Some preferred torsion angles exist, and the glycine is less restricted in the and angles (42). For our studies, the starting conformation of glutathione present in the crystal structure of glutathione reductase (44) was used. However, based on root mean square deviations, our energy-minimized model of glutathione bound to the AR active site appears similar to the glutathione moiety in the NMR structure of Escherichia coli glutaredoxin-3. These deviations, calculated using the C-␣ atoms of glutathione, were 0.99, 1.16, and 1.59 compared with the NMR structure of E. coli glutaredoxin (45) and the x-ray structures of glyoxalase I (46) and glutaredoxin-3 (47), respectively. Thus, glutathione bound to the active site of AR appears to be in a conformation that resembles more closely the low energy Y-shape of the molecule rather than the V-form bound to glutaredoxin, glutathione reductase, and glutathione peroxidase (44,47). The glyoxalase-like conformation of glutathione bound to the AR active site is in agreement with the observation that glutathione analogs inhibit AR with efficacy comparable to that of glyoxalase (see above).
The calculations in Table V show that individual amino acid residues of glutathione contribute 3-13 kcal/mol to the binding energy. However, the free energy changes calculated from the kinetic data are much smaller. For instance, the energy difference (⌬⌬G) derived from the comparison between glutathione and ␥Gly-Cys-Glu calculated from the relationship ⌬⌬Gb ϭ ϪRT ln(k cat /K m ) a /(k cat /K m ) b is Ϫ0.2 kcal/mol. Changes in free energy by other substitutions are also within this range, varying from Ϫ0.6 to ϩ 0.6 kcal/mol even when comparing dipeptides and tripeptides. The limiting binding energy of replacement of an amino group by hydrogen is 4.3 kcal/mol. Although this is seldom realized, values of ϳ3 kcal/mol are usually obtained (48). Therefore, the small ⌬⌬G changes calculated from the kinetic data suggest minimal utilization of the substrate binding energy for catalysis and indicate that, during catalysis, the energy for the formation of the transition state is derived mainly from NADPH and not from substrate binding.
In summary, the results of this study suggest that the active site of AR forms a specific glutathione-binding domain that selectively interacts with the glutathione backbone of glutathione-aldehyde conjugates. Moreover, our results indicate that glutathiolation may be an effective means for enhancing the efficiency of aldehyde metabolism and extending the range of aldehydes accessible to AR. Since the active site of AR is similar to other aldo-keto reductases (49), it appears likely that other members of this superfamily are also efficient catalysts for the biotransformation of carbonyl conjugates of glutathione.