Glutathione S-Transferase Pi Has at Least Three Distinguishable Xenobiotic Substrate Sites Close to Its Glutathione-binding Site*

Benzyl isothiocyanate (BITC), present in cruciferous vegetables, is an efficient substrate of human glutathione S-transferase P1-1 (hGST P1-1). BITC also acts as an affinity label of hGST P1-1 in the absence of glutathione, yielding an enzyme inactive toward BITC as substrate. As monitored by using BITC as substrate, the dependence of k of inactivation (KI) of hGST P1-1 on [BITC] is hyperbolic, with KI = 66 ± 7 μm. The enzyme incorporates 2 mol of BITC/mol of enzyme subunit upon complete inactivation. S-Methylglutathione and 8-anilino-1-naphthalene sulfonate (ANS) each yield partial protection against inactivation and decrease reagent incorporation, whereas S-(N-benzylthiocarbamoyl)glutathione or S-methylglutathione + ANS protects completely. Mapping of proteolytic digests of modified enzyme by using mass spectrometry reveals that Tyr103 and Cys47 are modified equally. S-Methylglutathione reduces modification of Cys47, indicating this residue is at/near the glutathione binding region, whereas ANS decreases modification of Tyr103, suggesting this residue is at/near the BITC substrate site, which is also near the binding site of ANS. The Y103F and Y103S mutant enzymes were generated, expressed, and purified. Both mutants handle substrate 1-chloro-2,4-dinitrobenzene normally; however, Y103S exhibits a 30-fold increase in Km for BITC and binds ANS poorly, whereas Y103F has a normal Km for BITC and Kd for ANS. These results indicate that an aromatic residue at position 103 is essential for the binding of BITC and ANS. This study provides evidence for the existence of a novel xenobiotic substrate site in hGST P1-1, which can be occupied by benzyl isothiocyanate and is distinct from that of monobromobimane and 1-chloro-2,4 dinitrobenzene.

Glutathione S-transferases (GSTs, 1 EC 2.5.1.18) are a family of multifunctional enzymes that participate in the biotransformation of xenobiotics by catalyzing the addition of the tripeptide GSH to various electrophilic compounds (1,2). Other GST functions include detoxification of lipid and DNA hydroperoxide by their intrinsic peroxidase activity (2), isomerization of certain steroids, and intracellular transport of various hydro-phobic nonsubstrate ligands such as bile acids, bilirubin, and a number of drugs (1).
GSTs are distributed in a wide range of organisms from mammals to Escherichia coli (3). The catalytic diversity of mammalian cytosolic GSTs, which can exist as homo-or heterodimers, arises in part from the existence of at least eight distinct classes (named Alpha, Kappa, Mu, Omega, Pi, Sigma, Theta, and Zeta) (1,4). The glutathione-binding site (G-site) and the catalytic mechanism of these enzymes have been the targets of many investigations involving chemical modification (5)(6)(7)(8), site-directed mutagenesis (9 -14), and x-ray crystallographic analysis (15)(16)(17)(18)(19)(20)(21). On the contrary, the electrophilic substrate-binding site (H-site) of GSTs is more complex and is class-specific. Crystallographic studies have shown that the H-site is quite dissimilar among different classes of GSTs (15)(16)(17)(18)(19)(20)(21). For example, in GST M1-1, the H-site is a hydrophobic cavity, whereas in GST P1-1 (15), the H-site is half-hydrophobic and half-hydrophilic with functionally important water molecules (16). Only a few amino acid residues have been identified in the H-site of GSTs, and relatively little is known about the key determinants of xenobiotic substrate specificity.
Among the various GSTs, the Pi class (GST P1-1) has attracted attention because it is overexpressed in a variety of malignancies, including lung, colon, ovary, esophagus, and stomach cancers (22)(23)(24)(25). Furthermore, a number of studies have shown a correlation between overexpression of GST P1-1 and the development of resistance toward various anti-cancer drugs, such as Adriamycin, cisplatinum, melphalan, and etoposide in resistant tumor cells (22)(23)(24)(25). The task of defining the xenobiotic substrate specificity of Pi class GST becomes a critical component in the rational design of highly potent GST Pi-selective inhibitors that may increase the effectiveness of commonly used anti-cancer agents. Previously, we showed there is a site for monobromobimane in the xenobiotic region of GST Pi distinct from that occupied by 1-chloro-2,4-dinitrobenzene, supporting the concept that this isozyme accomplishes its ability to react with a diversity of substrates in part by harboring distinct xenobiotic substrate sites (26).
Benzyl isothiocyanate (BITC), found abundantly in cruciferous vegetables (e.g. cabbage, cauliflower, and broccoli), contains a central carbon that can undergo facile addition reactions with N-, O-, or S-based nucleophiles, including the sulfhydryl group of glutathione (27)(28)(29). It has been shown that BITC is a substrate of GSTs, with the Pi class being the most efficient catalyst (28). Consequently, locating its binding site in GST Pi should further contribute to the understanding of substrate specificity in relation to other GSTs. In this paper, we report the results of the affinity labeling of human GST P1-1 using BITC. Covalent modification of hGST P1-1 by BITC causes differential loss of activity toward two substrates, suggesting the existence of a novel binding site for xenobiotic substrates in the Pi isozyme distinct from that of monobromo-bimane and 1-chloro-2,4-dinitrobenzene. A preliminary version of this work has been presented (30).
Plasmids and Mutagenesis-The full-length cDNA for human glutathione S-transferase P1-1 was encoded in a pUC120 plasmid, as described by Manoharan et al. (31), and was a gift from W. E. Fahl (University of Wisconsin, Madison). Site-directed mutagenesis was performed using the Stratagene QuikChange kit. The following oligonucleotides and their complements were used to incorporate the mutations (position of the mutation underlined): Y103F, 5Ј-CTC CGC TGC AAA TTC ATC TCC CTC ATC TAC ACC; Y103S, 5Ј-CTC CGC TGC AAA TCC ATC TCC CTC ATC. Mutations were confirmed by DNA sequencing (forward sequencing primer, 5Ј-CCG CCC TAC ACC GTG GTC TAT TTC CCA GTT, and reverse sequencing primer, 5Ј-CTG TTT CCC GTT GCC ATT GAT GGG GAG GTT), which was carried out at the University of Delaware Center for Agricultural Biotechnology using an ABI Prism model 377 DNA sequencer (PE Biosystems).
Protein Purification-hGST P1-1 was expressed in E. coli JM105. Cells were grown at 37°C, and when the A 600 nm reached 0.4, the cells were induced with 1 mM isopropyl thio-␤-D-galactopyranoside. After induction, the cells were grown at 25°C for 24 h, at which time they were harvested by centrifugation at 10,000 ϫ g for 20 min. The pellets were then frozen at Ϫ80°C. Cells were resuspended in 20 mM Tris, 100 mM NaCl, 1 mM EDTA, 1 mM DTT (pH 7.5, ϳ100 ml for 6 liters of culture), followed by sonication for 6 min using a sonicator (Ultrasonic, Inc at 20 kHz and 475 watts). This suspension was then centrifuged for 25 min at 10,000 ϫ g. Wild-type and mutant enzymes were purified using affinity chromatography on S-hexylglutathione-agarose. Briefly, the column was eluted with 20 mM Tris buffer (pH 7.5) containing 0.1 M NaCl, 1 mM EDTA, and 1 mM DTT, followed by the same buffer, to elute any weakly bound proteins. The GST was eluted with 20 mM Tris buffer (pH 7.5) containing 1 mM EDTA, 1 mM DTT, 0.2 M NaCl, plus 2.5 mM S-hexylglutathione and was dialyzed into 0.l M KPO 4 buffer (pH 6.5) containing 1 mM EDTA. The purity of the enzymes was determined using SDS-PAGE and N-terminal sequencing, performed on an Applied Biosystems Procise Sequencing System.
Enzyme Assays-Enzymatic activity toward CDNB was measured in a total volume of 1.0 ml using a Hewlett-Packard 8453 spectrophotometer by monitoring the formation of the conjugate of CDNB (3 mM) and glutathione (2.5 mM) at 340 nm (⌬⑀ ϭ 9600 M Ϫ1 cm Ϫ1 ) in 0.1 M potassium phosphate buffer (pH 6.5) at 25°C, according to the method of Habig et al. (32). All measurements were corrected for the spontaneous nonenzymatic rate of formation of the conjugate of glutathione and CDNB.
Other Substrates Utilized to Assay for Enzymatic Activity Were BITC and mBBr-For BITC, enzymatic activity was measured in a total volume of 1.0 ml at 274 nm (⌬⑀ ϭ 9250 M Ϫ1 cm Ϫ1 ) by monitoring the formation of the conjugate of BITC (400 M) and glutathione (450 M) in 0.1 M potassium phosphate buffer (pH 6.5) at 25°C, according to the method of Kolm et al. (28). For mBBr, the enzymatic activity was measured in a total volume of 0.9 ml using a PerkinElmer Life Sciences MPF-3 fluorescence spectrophotometer (excitation at 395 nm and emission at 480 nm) by monitoring the formation of the conjugate of mBBr (100 M) and glutathione (600 M) in 0.1 M potassium phosphate buffer (pH 6.5) at 25°C, according to the method of Hulbert and Yakubu (33). A known amount of glutathione-bimane was used as a fluorescence standard to calibrate and calculate the product formation. Lower concentrations of glutathione were chosen for these substrates because of the relatively large spontaneous nonenzymatic rate of reactions of product formation.
To determine the apparent K m value of glutathione, a range of glutathione concentrations (0.01 to 2 mM) was investigated at a constant CDNB concentration (3 mM). Similarly, the apparent K m value for CDNB was determined from a range of concentrations of CDNB (0.01-3.5 mM) at a constant glutathione concentration (2.5 mM) in 0.1 M potassium phosphate buffer (pH 6.5). For BITC, the apparent K m value was determined from a range of concentrations (4 -400 M) at a constant glutathione concentration (450 M). For the determination of the K m value of mBBr, the glutathione concentration was maintained at 600 M, whereas mBBr varied from 10 to 200 M. Data were analyzed by fitting directly to the Michaelis-Menten equation using a nonlinear curve fitting program (SigmaPlot from SPSS).
Reaction of hGST P1-1 with BITC-hGST P1-1 (0.3 mg/ml) was incubated in 90 mM phosphate buffer (pH 6.5) at 25°C with various concentrations of BITC by the addition of appropriate stock solutions of BITC in acetonitrile. The concentration of BITC was determined from the absorbance at 245 nm using ⑀ 245 nm ϭ 10,780 M Ϫ1 cm. The volume of acetonitrile was maintained at 10% of the total volume of the reaction mixture. In control experiments, the enzyme was incubated under the same conditions including 10% acetonitrile but without BITC. In every case, aliquots of the reaction mixture were removed at specified times and assayed for enzymatic activities using either CDNB or BITC as substrate. For both assays, aliquots of the reaction mixture were diluted 10-fold with 0.1 M potassium phosphate buffer (pH 6.5) at 25°C and were assayed by the addition of 20 l to the cuvette. The rate constants for reaction of the enzyme with mBBr were calculated from semi-logarithmic plots of E/E 0 versus time, in accordance with the pseudo first-order kinetic equation: ln(E t /E 0 ) ϭ Ϫk obs t, where E 0 is the activity of the enzyme at time 0; E t represents the activity at a given time (t); and k obs is the observed pseudo first-order rate constant.
In the preparation of modified and control enzyme, excess reagent was separated from the enzyme by the gel centrifugation method of Penefsky (34) in which aliquots (0.5 ml) of the reaction mixture at a given time were applied to a 5-ml column of Sephadex G-50 equilibrated with 0.1 M potassium phosphate buffer (pH 6.5) and centrifuged. The protein concentration in the filtrate was determined using the Bio-Rad protein reagent, which is based on the dye-binding method of Bradford (35). Purified GST Pi was used as the protein standard, and absorbance at 600 nm was measured using a Bio-Rad model 2550 RIA reader.
Synthesis of S-(N-Benzylthiocarbamoyl)glutathione-S-(N-Benzylthiocarbamoyl)-glutathione was prepared by reaction of BITC with an excess concentration of glutathione. In a typical reaction, 100 mM of BITC in acetonitrile (0.5 ml) was combined with 300 mM of glutathione (0.5 ml) (pH 6.5-7). The reaction mixture was allowed to stand at room temperature for 1 h. The solution was then lyophilized and redissolved in water. S-(N-Benzylthiocarbamoyl)glutathione was purified by HPLC using a Varian 5000 LC equipped with a Vydac C18 column and a UV-100 detector. The solvent system used was H 2 O (solvent A) and acetonitrile (solvent B). The column was equilibrated with solvent A. After 30 min at 0% solvent B, a linear gradient was run to 100% solvent B in 100 min at a flow rate of 1 ml/min. The effluent was monitored at 220 nm and S-(N-benzylthiocarbamoyl)glutathione eluted at 11% solvent B. For comparison, glutathione elutes at 0% solvent B. Product purity was assessed by thin layer chromatography on cellulose plates with a fluorescent indicator using n-butyl alcohol/acetic acid/ water (20:8:5) as the solvent system. A single spot (which was ninhydrin positive) was observed, with an R f value of 0.65. For comparison, oxidized and reduced glutathione exhibited spots at R f 0.06 and 0.40, respectively.
S-(N-Benzylthiocarbamoyl)glutathione exhibits an ultraviolet absorption spectrum with a peak at 245 nm, which is not present in the spectrum of the precursor glutathione. The extinction coefficient was measured as 9950 M Ϫ1 cm Ϫ1 in phosphate buffer (pH 6.5) at the max of 245 nm on the basis of the S-(N-benzylthiocarbamoyl)glutathione concentration determined from the absorbance at 274 nm (⌬⑀ ϭ 9250 M Ϫ1 cm Ϫ1 ) (28).
Measurement of Incorporation of BITC into hGST P1-1-BITC and its amino acid conjugates display an absorption peak at 245 nm, whereas the absorption spectrum of hGST P1-1 is close to a minimum at 245 nm; therefore, the determination of an absorption coefficient at this wavelength should allow accurate incorporation measurements of BITC to hGST P1-1. Based on the method of Kolm et al. (28), BITC (20 M) was allowed to react with either 1 mM glutathione or 1 mM L-Gly-Tyr in 100 mM potassium phosphate (pH 8.5) containing 2% (v/v) acetonitrile. The absorbance at 245 nm was measured after completion of the reactions (30 min), and the molar extinction coefficients were calculated as 9950 and 11,760 M Ϫ1 cm Ϫ1 for S-(Nbenzylthiocarbamoyl)glutathione and O-(N-benzylthiocarbamoyl)tyrosine, respectively, yielding an average extinction coefficient of 10,850 M Ϫ1 cm Ϫ1 .
hGST P1-1 (0.3 mg/ml) was incubated for the indicated time with 100 M BITC, with or without the addition of protectants under standard reaction conditions. Excess reagents were removed by gel filtration columns, and the protein concentration was determined by the Bio-Rad method, as described above. The amount of reagent incorporated was determined from the absorbance at 245 nm using ⑀ 245 nm ϭ 10,850 Preparation of Modified hGST P1-1 for Kinetic Studies-hGST P1-1 (0.3 mg/ml) was incubated for 40 min with 100 M BITC in the presence of 5 mM S-methylglutathione. Excess reagents were removed by gel filtration, and the protein concentration was determined by the Bio-Rad dye-binding method. The prepared modified enzyme was frozen quickly and stored at Ϫ80°C.
Measurement of ANS Binding to hGST P1-1-Equilibrium binding studies with ANS were determined by measuring the enhancement of ANS fluorescence when the ligand (0 -900 M) binds to native and modified wild-type hGST P1-1, Y103F, and Y103S mutant enzymes (16 M), as described by Bico et al. (36). Excitation was at 390 nm and emission at 480 nm. Fluorescence measurements were made in 0.1 M phosphate buffer (pH 6.5) at 25°C.
Trypsin Digestion of Modified hGST P1-1-hGST P1-1 (0.3 mg/ml) was incubated for 40 min with 100 M BITC with or without the addition of ligands under standard reaction conditions. Excess reagent was removed by gel filtration as described above. Solid urea (to give 6 M as the final concentration) and N-ethylmaleimide (final concentration, 10 mM) were added to the enzyme and incubated for 30 min at 25°C. The enzyme solution was then dialyzed overnight against 50 mM ammonium bicarbonate (pH 7.8) (4 liters).
The enzyme solution was lyophilized and then resolubilized in 250 l of 8 M urea in 50 mM ammonium bicarbonate (pH 7.8). This solution was incubated at 37°C for 2 h to denature the protein. Ammonium bicarbonate (750 l, 50 mM (pH 7.8)) was then added to the solution to dilute the urea to 2 M. Trypsin (5% (w/w)) was added, and the enzyme sample was incubated for 2 h at 37°C. A second aliquot of the trypsin solution was added, and incubation was continued for another 2 h at 37°C.
HPLC Separation of Modified Peptides-The tryptic digest was injected onto a Varian 5000 LC HPLC (Varian, Walnut Creek, CA) equipped with a Phenomenex C18 reverse-phase column (0.46 ϫ 25 cm). At a flow rate of 1 ml/min, the peptides were separated by a 300-min linear gradient from 0 to 60% solvent B (solvent A, 0.1% trifluoroacetic acid in water; solvent B, 0.075% trifluoroacetic acid in acetonitrile), followed by a 40-min gradient to 80% solvent B, and finally a 20-min gradient to 100% solvent B. The eluate was monitored at 220 and 245 nm with 1-ml fractions collected.
Peaks I and II from the trypsin digest were pooled separately, lyophilized, and redissolved in 1 ml of 20 mM ammonium bicarbonate (pH 7.8). Chymotrypsin (2.5% w/w) was added to the peak I solution, and thermolysin (2.5% w/w) was added to the peak II solution. Both solutions were incubated at 25°C for 2 h. The redigested samples were applied to the HPLC (Vydac C18 column) and eluted at a flow rate of 1 ml/min using the same gradient as before.
Analysis of Isolated Peptides-The purified labeled peptides from the HPLC separation (as described previously) were lyophilized and redissolved in 0.1% trifluoroacetic acid/H 2 O. Peptide molecular weights were determined using a ThermoFinnigan model LCQ mass spectrometer equipped with an electrospray attachment.
Molecular Modeling-Modeling was conducted using the program Insight II from Biosym Technologies on a Silicon Graphics work station. The molecular models of BITC and bromosulfophthalein were built and energy-minimized using the Builder module of the Insight II program. The approach of the BITC molecule to Tyr 103 was modeled by positioning BITC at a site close to Tyr 103 and the bromosulfophthalein site of human GST Pi (PDB codes 19GS (47) and 18GS (48)) by sequentially rotating and translating it along the x, y, and z axes. The intermolecular energy in terms of both van der Waals' and electrostatic interactions as well as the interatomic distance between the oxygen atom of Tyr 103 and the reactive central carbon atom of BITC were continuously monitored for conformations, with reasonable distances and potential energies constituting possible productive interactions for chemical modification of the tyrosyl hydroxyl group. The molecular model of the enzyme modified by BITC at Tyr 103 was built and energy-minimized using the Builder and Homology modules of the Insight II program. Another possible conformation for the molecule was constructed by rotating BITC so that the reactive central carbon atom would face the thiol group of glutathione. The BITC was also docked at a site close to Cys 47 of hGST P1-1. The manually docked complex as well as BITC-modified enzyme models were submitted to the Discover® program from Biosym for extensive energy minimization using steepest descent and conjugate gradient methods to relieve residual van der Waals' overlaps and optimize the structures.

RESULTS
BITC as a Substrate for Human GST P1-1-The formation of the glutathione conjugate with BITC is catalyzed by hGST P1-1, as determined by Kolm et al. (28). As part of this study, kinetic parameters for the enzymatic reaction have been confirmed, and the catalysis follows normal Michaelis-Menten kinetics. In hGST P1-1, the apparent K m values for BITC and CDNB were 68 Ϯ 5 and 650 Ϯ 39 M, respectively, and the apparent V max values were 70 Ϯ 4 and 65 Ϯ 6 mol/min/mg, respectively. Thus, the enzyme has a higher apparent affinity for BITC than for CDNB, the more traditional substrate for GST, but the two substrates have similar V max values.
Because benzyl isothiocyanate is similar in structure to phenyl isothiocyanate, which is used in N-terminal polypeptide sequencing, the reaction of BITC with the tripeptide glutathione may be considered to involve either the free N-terminal group or the thiolate group. To address this question, hGST P1-1 was added to 0.1 M phosphate buffer at pH 6.5 containing 400 M BITC and 450 M S-methylglutathione, a glutathione derivative which has its thiolate group blocked. No formation of product was detected spectrophotometrically at 274 nm. We conclude that BITC must react with the thiolate of GSH under these conditions. Previously, we found that S-(hydroxyethyl)bimane functions as a competitive inhibitor of monobromobimane (K I ϭ 56 M) (26). We have now found that S-(hydroxyethyl)bimane does not function as an inhibitor when BITC is the substrate, indicating BITC and mBBr occupy distinct substrate sites of hGST P1-1. This conclusion is based on the observation that addition of 50 or 100 M of S-(hydroxyethyl)bimane does not change either the K m or V max value for BITC. The nonsubstrate ligand ANS has been shown to function as a noncompetitive inhibitor with respect to other substrates (36), but its ability to inhibit BITC has not previously been tested. The addition of various concentrations of ANS does not appreciably change the K m value for BITC but decreases V max ; therefore, we conclude that ANS functions as a noncompetitive inhibitor with respect to BITC yielding a K I of 6.1 Ϯ 0.1 M.
Inactivation of Human GST P1-1 by BITC-The effect of incubating homogeneous hGST P1-1 (0.3 mg/ml) with 100 M BITC at pH 7.0 and 25°C in the absence of glutathione was studied. The concentration of BITC was determined from the absorbance at 245 nm using ⑀ 245 nm ϭ 10,780 M Ϫ1 cm Ϫ1 . The reaction was followed by assaying aliquots for their residual activity using either CDNB or BITC as substrates. This results in a time-dependent inactivation of the enzyme as assayed using BITC as substrate (k obs ϭ 0.067 Ϯ 0.007 min Ϫ1 ), with a lesser effect on the use of CDNB as substrate for the enzyme (k obs ϭ 0.032 Ϯ 0.005 min Ϫ1 ) (Fig. 1). The control enzyme, incubated under the same conditions but in the absence of the reagent, showed no change in activity during the same period. As monitored by using BITC as substrate, the dependence of the k value of inactivation of hGST P1-1 on [BITC] is hyperbolic, with a K I of 66 Ϯ 7 M and k max ϭ 0.118 Ϯ 0.007 min Ϫ1 (Fig. 2). This type of curve is typical for an affinity label, suggesting that an enzyme-reagent complex is formed reversibly prior to the irreversible modification of the enzyme (37). In contrast, when monitored using CDNB as substrate, the k value for inactivation exhibits a linear dependence on [BITC] with a second-order rate constant of 308 Ϯ 11 min Ϫ1 M Ϫ1 (Fig.  2). These results suggest there are at least two different target sites for BITC.
Effect of Substrate Analogs on the Rate of Inactivation of hGST P1-1 by BITC-The ability of various substrate analogs to protect against inactivation of the enzyme by 100 M BITC was examined to evaluate the site(s) at which BITC reacts.
Because the rate of inactivation of the enzyme was different when assayed using either substrate BITC or CDNB, protection against inactivation was monitored separately with either BITC or CDNB as substrates. The results, given in Table I, are expressed as k ϩL /k ϪL , where k ϩL is the rate constant for inactivation in the presence of a particular ligand, and k ϪL is the rate constant for inactivation in the absence of a particular ligand. When assayed by CDNB, all glutathione analogs afford substantial protection against enzyme inactivation (Table I,  lines 2-4). However, electrophilic substrate analogs do not provide any protection (Table I, lines 5 and 6). In fact, the addition of dinitrophenol to the reaction mixture doubles the rate of inactivation (Table I, line 5) when compared with the absence of ligand, suggesting a reactive group in the enzyme is further exposed upon the addition of dinitrophenol. 8-Anilino-1-naphthalene sulfonate and bromosulfophthalein are two compounds known to bind to nonsubstrate-binding sites in hGST P1-1 (36); the addition of these ligands to the reaction mixture has little or no effect on the rate constant (Table I, lines 7 and 8). Product analogs, which contain a glutathionyl moiety, also provide effective protection against enzyme inactivation as assayed by CDNB (Table I, lines 12 and  13). These results indicate that inactivation of the enzyme as assayed by CDNB is governed by reaction of BITC at the glutathione substrate site.
In contrast, when assayed using substrate BITC, glutathione derivatives offer much less protection (Table I, lines 2-4). The 5 mM concentrations used are sufficient to saturate the glutathione site, yet k ϩL /k ϪL for S-methylglutathione does not decrease below 0.75. The protective effect increases only when the alkyl chain length is increased (e.g. Table I, line 3) or if Smethylglutathione is combined with other ligand analogs (e.g. Table I, lines 9 -11). These results suggest that affinity labeled BITC must have another target site aside from the glutathione substrate site.
The xenobiotic substrate analogs of CDNB and mBBr do not protect against inactivation by BITC, indicating that the affinity label BITC does not react at these substrate sites (Table I,  lines 5 and 6). Conversely, the nonsubstrate hydrophobic ligands, such as 8-anilino-1-naphthalene sulfonate or bromosulfophthalein, provide partial protection against inactivation when used alone (Table I, (Table I, line  11). The best protection is provided by the combination of S-methylglutathione and ANS (Table I, line 11) and the substrate product S-(N-benzylthiocarbamoyl)glutathione (Table I, line 13), suggesting that the second reaction occurs at or near the nonsubstrate hydrophobic site that is not far from the glutathione substrate site.
Incorporation of BITC by hGST P1-1-hGST P1-1 was incubated with 100 M BITC in the absence or presence of added protectants. Subsequently, the modified enzymes were isolated, and the incorporation of BITC was measured from its characteristic absorbance at 245 nm as described under "Experimental Procedures." The incorporation of BITC into hGST P1-1 was measured as a function of time (0 -40 min) in the absence of added ligands, and the reagent incorporation was plotted as a function of residual activity at the same time (Fig.  3). Extrapolation to totally inactive enzyme yields about 2 mol of reagent/mol of subunit. These results suggest that inactivation results from modification of at most two amino acid residues.
The effect of the addition of ligands on the incorporation of BITC into hGST P1-1 was investigated. Incubation of hGST P1-1 with 100 M BITC for 40 min affords a modified enzyme that is 8% active and contains 1.72 mol of reagent/mol of subunit. The addition to the reaction mixture of either Smethylglutathione or ANS yields enzyme that retains a little more activity (15 and 22%, respectively) and exhibits small decreases in incorporation to no less than 1.23 mol of reagent/ mol of subunit. In contrast, when both S-methylglutathione and ANS are present in the reaction mixture, a dramatic decrease in incorporation is observed (0.31 mol of reagent/mol of subunit), and 92% enzymatic activity is retained. These results support the proposal that the BITC reaction occurs at two locations. Reaction at either one of these reaction sites is sufficient to cause complete inactivation of the enzyme.
Properties of BITC-modified hGST P1-1-The catalytic properties of the modified enzyme were investigated by using CDNB, BITC, and mBBr as substrates. As reported earlier, reaction of BITC occurs at two different sites. S-Methylglutathione affords partial protection against inactivation, suggesting that one of the reaction sites may be at or near the glutathione binding region and the other at a different site in the   FIG. 1. Inactivation of hGST P1-1 by BITC. A solution of hGST P1-1 (0.3 mg/ml) was incubated with 100 M BITC in 90 mM potassium phosphate buffer (pH 6.5) at 25°C. Enzymatic activity (E) was measured using as substrates either 1-chloro-2,4-dinitrobenzene (•) or BITC (Ⅺ), as described under "Experimental Procedures." From this graph, a pseudo first-order rate constant of 0.032 Ϯ 0.004 min Ϫ1 was calculated from the time-dependent decrease in residual activity measured by the CDNB assay and 0.067 Ϯ 0.003 min Ϫ1 when E was measured from the BITC assay. As a control, enzyme was incubated without BITC under similar conditions and assayed for activity toward BITC (E). enzyme. Therefore, to study catalytic properties of modified enzyme, hGST P1-1 was incubated with 100 M BITC at a single time (90 min) in the presence of S-methylglutathione to protect the glutathione-binding site, while leaving the other target site vulnerable to reaction. Table II compares the percent inactivation as monitored by various substrates. The residual activity is different with respect to all substrates. The enzyme retains almost full activity with respect to CDNB as substrate but is nearly completely inactive with respect to BITC. In contrast, the enzyme retains 40% residual activity with respect to mBBr. These results suggest that the sites occupied by BITC, CDNB, and mBBr are distinct.
It seemed possible that BITC modification of one site of the enzyme could indirectly alter the mBBr and CDNB sites. Thus, the kinetic constants for mBBr and CDNB were determined for control and modified hGST P1-1. As shown in Table II, the modification results in 60% inactivation when monitored by mBBr as a substrate; however, the K m value of mBBr for hGST P1-1 does not change appreciably as a result of the modification with BITC (see Table V for the K m value for wild-type enzyme). In the case of CDNB as a substrate, modification of the enzyme with BITC does not have an effect on V max (data not shown) but results in a 7-fold increase of K m when compared with the K m value of native enzyme (see Table V) of 650 Ϯ 39 M. For comparison, the kinetic constants of glutathione were also de-termined for modified enzyme, and they were similar to those of control enzyme (data not shown). These results suggest that reaction with BITC at one site has an indirect effect on the properties of the xenobiotic substrate region of residual active enzyme.
In previous studies, we found that monobromobimane does not occupy the same site as the hydrophobic compound ANS (26). However, ANS provides partial protection against BITC modification (Table I); therefore, we tested the ability of this 90-min BITC-modified enzyme, which contains 1.83 mol of reagent/mol of subunit, to bind ANS, using the fluorescence enhancement of enzyme-bound ANS. Fig. 4A compares the binding abilities of control versus modified enzyme for ANS at a range of concentrations from 3 to 45 M. Results from the wild-type unmodified enzyme indicate there is a high affinity binding site for ANS with a K D value of 6.4 Ϯ 0.4 M, a value similar to that found previously (26). In the case of the modified enzyme the fluorescence intensity is similar to that of free ANS, indicating that modification with BITC eliminates the high affinity binding site for ANS. These results suggest the binding site of ANS may be at or near the target site for BITC.
Isolation and Characterization of Peptides from Modified hGST P1-1-hGST P1-1 (0.3 mg/ml) was inactivated for 40 min by 100 M BITC at pH 6.5 and 25°C. The resulting modified enzyme with 8% residual activity toward BITC was isolated, incubated with urea and N-ethylmaleimide to block unreacted cysteine residues, dialyzed, and digested with trypsin. The resulting peptides were separated by HPLC, using a C18 reversephase column equilibrated with 0.1% trifluoroacetic acid using an acetonitrile gradient (Fig. 5). Two peptide regions exhibit the characteristic thiocarbamate absorbance of 245 nm. Peaks I and II eluted at 15 and 39% acetonitrile respectively (Fig. 5B).  The inactivation reaction was conducted at 25°C in 90 mM potassium phosphate buffer, pH 6.5. k ϩL /k ϪL was determined by the ratio of initial inactivation rate with ligand present (k ϩL ) to that observed in the absence of ligand (k ϪL ), where k ϪL (BITC) ϭ 0.067 Ϯ 0.009 min Ϫ1 and k ϪL (CDNB) ϭ 0.032 Ϯ 0.005 min Ϫ1 . Both peaks were subjected to electrospray mass spectrometry to obtain evidence of chemical modification. Peak I yielded peptides of 1534.9 and 1684.7 atomic mass units. The 1534.9 atomic mass units is equivalent to the unmodified form of the peptide YISLIYTNYEAGK, corresponding to residues 103-115, whereas the 1684.7 atomic mass units corresponds to the predicted mass of the same peptide modified by BITC. In peak II, peptides of 1079.9 and 1229.6 atomic mass units were detected. The 1079.9 atomic mass units is comparable with the predicted mass of the unmodified form of the peptide AS-CLYGQLPK corresponding to residues 45-54, whereas 1229.6 atomic mass units are identical to the predicted mass of the same peptide modified by BITC.
Both modified peptides corresponding to peaks I and II have many potential amino acid targets for BITC modification. Thus, they were treated with other proteases to identify further the modified amino acids. Peptide peak I was redigested with chymotrypsin, whereas peptide peak II was redigested with thermolysin, and both were subjected to the same HPLC solvent system as before. Peak Ia, which absorbed at 245 nm, eluted at 10% acetonitrile, earlier than peak I. Peak IIa, which also absorbed at 245 nm, eluted at 6% acetonitrile, earlier than peak II. Table III shows the results obtained when these two peaks were subjected to mass spectrometric analysis. Masses containing BITC in peak Ia correspond to Tyr (332.1 atomic mass units), 103 YISLIY 108 (940.5 atomic mass units). These data indicate that the modified peptide resulting from peak Ia is 103 YISLIY 108 . Other peptides seen in peak Ia were most likely the result of fragmentation during the mass spectrometry analysis. Although 940.5 atomic mass units does not distinguish between reaction of BITC with Tyr 103 or Tyr 108 , peptides I and Ia were both resistant to sequencing by Edman degradation, indicating that Tyr 103 had been derivatized previously by BITC, thereby preventing reaction with phenylisothiocyanate. Furthermore, Tyr 108 of GST Pi had been subjected previously to mutagenesis (38). Ahn et al. (38) studied Y108A and Y108F, showing little change in the kinetic parameters for CDNB as substrate. In peak IIa, masses containing BITC correspond to Cys 47 (271.3 atomic mass units) and 45 ASC 47 (429.6 atomic mass units). From these data, the modified peptide from peak IIa is 45 ASC 47 . Similar to peak Ia, other peptides seen in peak IIa were most likely the result of fragmentation from the mass spectrometry analysis. This information, in conjunction with the masses of the intact peptides and the primary sequence of the enzyme, clearly defines Tyr 103 and Cys 47 as the targets of BITC, and their modifications are responsible for the loss of enzymatic activity.
Effect of Protectants on Labeled Peptide Peaks- Table IV  presence of ligands. Approximately equal amounts of BITC are found in peaks I and II in the absence of protecting ligands. Compared with the absence of ligands, a markedly lower incorporation was found in peptide II when S-methylglutathione was present in the reaction mixture (Table IV, line 2), indicating this peptide is within the glutathione binding region. Comparison of lines 1 and 3 in Table IV reveals the major decrease of BITC incorporation is in peptide I when the enzyme is protected by ANS. This observation suggests that peptide I is primarily within the ANS binding region, and this site appears to be at or near the BITC substrate-binding site. When either S-(N-benzylthiocarbamoyl)glutathione (Table IV, (Table IV, line 4) are added to the reaction mixture, there is a substantial decrease in BITC incorporation into both peptides I and II, indicating the glutathione binding region and the BITC site are not far apart.
Kinetic Properties of Mutant Enzymes-Because the affinity labeling study points to Tyr 103 as the residue responsible for inactivation toward BITC as substrate, mutant enzymes were constructed with substitutions at position 103 to provide insights into the role of this residue in catalysis. Nishihira et al. (39), Park et al. (14), and Ricci et al. (40) have previously studied mutant enzymes with replacements for Cys 47 ; therefore, this residue was not studied further.
Because Tyr 103 has both an aromatic and a hydroxyl group, two mutants were constructed as follows: Y103F, to retain the aromatic moiety while eliminating the hydroxyl group, and Y103S, to retain the hydroxyl group while eliminating the aromatic component. These mutant enzymes were expressed at 25°C in E. coli strain JM105 and then isolated and purified to homogeneity using S-hexylglutathione-agarose affinity chromatography.
Four different substrates, including BITC, were studied. It was expected that significant differences would be observed for substrate BITC in the mutants as compared with the wild-type enzyme if Tyr 103 is involved exclusively in the BITC substrate site; furthermore, it was anticipated that the kinetic parameters for the other substrates would be similar for mutant and wild-type enzymes. The actual data are shown in Table V. For wild-type hGST P1-1, the V max for mBBr is ϳ1/3 that for all the other substrates, whereas the K m for CDNB is about 10 times that of the other substrates (Table V). For both mutants, the K m and V max values of CDNB and GSH (with CDNB as substrate) do not change appreciably. In contrast, for Y103S, the K m for BITC is ϳ30 times that of the wild-type enzyme, but the V max does not change. Y103F retains the same kinetic characteristics as wild-type enzyme for BITC as substrate. These results demonstrate that Tyr 103 contributes to the affinity of the enzyme for BITC possibly through Pi-Pi stacking interactions between the rings of Tyr 103 and BITC. In the case of mBBr, there was an unexpected decrease in V max for both mutants with an unchanged K m value, suggesting that mBBr and BITC may share the same site; however, this is unlikely because we have shown that S-(hydroxyethyl)bimane, an mBBr analog, failed to protect against inactivation by affinity label BITC and did not inhibit the reaction of BITC as a substrate. We propose that BITC binds at a discrete site of hGST P1-1 (see "Discussion" for further consideration of this issue).
The identification of Tyr 103 as a major target of the affinity label BITC was strengthened by incubating Y103F with 100 M BITC in the presence of S-methylglutathione. The mutant enzyme was not inactivated by BITC, indicating that the phenoxy  group of Tyr 103 is the nucleophilic group that reacts with BITC under these conditions. The ability of mutant enzymes to bind ANS was also determined. Fig. 4B compares the ability of wild-type and mutant enzymes to bind ANS at concentrations ranging from 3 to 900 M (concentrations considerably higher than those shown in Fig. 4A). Results from the wild-type unmodified enzyme indicate there are at least two sites for ANS (K DϪ1 ϭ 6.4 Ϯ 0.4 M, shown in Fig. 4B, inset, and K DϪ2 ϭ 153 Ϯ 20 M shown in Fig.  4B). As shown in Fig. 4A, modification of the enzyme by BITC eliminated the high affinity site for ANS; however, the BITCmodified enzyme still binds ANS at higher concentrations (K DϪ2 ϭ 148 Ϯ 18 M). The binding of ANS to Y103F (K DϪ1 ϭ 3.3 Ϯ 0.3 M and K DϪ2 ϭ 151 Ϯ 9 M) is similar to that of wild-type enzyme, indicating this mutation does not affect the ANS affinity. In contrast, mutation of Tyr 103 to Ser greatly weakens the high affinity site for ANS (see Fig. 4B, inset), whereas the low affinity binding of ANS is not appreciably changed (K DϪ2 ϭ 169 Ϯ 18 M). The effect of these substitutions suggests that Tyr 103 in hGST P1-1 contributes to the tight binding of ANS. A major determinant in the binding of ANS is the aromatic ring of Tyr 103 , rather than its hydroxyl group.

DISCUSSION
Most of the substrates of GSTs that have been studied in detail are synthetic compounds not present in nature (28); however, benzyl isothiocyanate (BITC), a substrate for human GST P1-1, is abundant in edible plants (28). In addition, recent studies (27)(28)(29) suggest BITC offers chemoprotection against tumor formation by inducing the synthesis of GSTs. Thus, locating the binding site of BITC in hGST P1-1 not only provides structural insights into the several functions of the enzyme but contributes to an understanding of the metabolism of benzyl isothiocyanate in humans. BITC is indeed an efficient substrate for hGST P1-1 with a lower apparent K m value than for the more common xenobiotic substrate CDNB. In addition, when added to the enzyme in the absence of glutathione, BITC acts as an affinity label that reacts covalently with the enzyme.
The results of this study of hGST P1-1 demonstrate that there are two target sites for BITC, Tyr 103 and Cys 47 . 8-Anilino-1-napthalene sulfonate decreases incorporation into Tyr 103 , whereas S-methylglutathione decreases the reaction of BITC with Cys 47 .
Previously, we demonstrated through affinity labeling with monobromobimane that Cys 47 is at or near the glutathione- binding site (26). In this study, BITC also targets Cys 47 and the addition of S-methylglutathione decreases the reaction of BITC with Cys 47 , further supporting the location of this residue at or near the glutathione-binding site. We did not conduct mutagenesis at position 47 because others have provided insights into the role of this residue in the enzyme. For example, Nishihira et al. (39) showed that reaction of Cys 47 with the bulky reagent 7-fluoro-4-sulfamoyl-1,2,3-benzodiazole inactivates hGST Pi by steric hindrance of the glutathione-binding site. Park et al. (14) demonstrated that the binding affinity between hGST P1-1 and glutathione decreases 5-fold when Cys 47 is mutated to Ser. In addition, Cacurri et al. (41) reported that the human GST Pi is inactivated by 1-chloro-2,4-dinitrobenzene upon modification of Cys 47 , with protection being provided by S-methylglutathione; and the replacement of Cys 47 with Ala or Ser decreases the affinity for glutathione. These results all support the location of Cys 47 at or near the glutathione-binding site.
This is the first time that Tyr 103 has been implicated in any function of GST Pi. We demonstrated here that labeling of Tyr 103 in hGST P1-1 results in loss of activity toward BITC as a substrate. Most interestingly, ANS, which is a hydrophobic nonsubstrate ligand, protects against inactivation as monitored by BITC as a substrate and decreases incorporation into Tyr 103 . Examination of the kinetic characteristics of Y103F and Y103S enzymes allows evaluation of the role of Tyr 103 in catalysis. The K m value for BITC is dramatically increased in Y103S, suggesting that Tyr 103 is critical for binding of BITC but is not required for catalysis. Such binding may be governed by Pi-Pi stacking interactions becauseY103F exhibits the same affinity for BITC as wild type, whereas elimination of the aromatic group (Y103S) decreases the BITC affinity markedly. The decrease in the V max value for mBBr of both mutants was unexpected. This result could be interpreted to indicate that mBBr and BITC share the same site; however, the failure of S-(hydroxyethyl)bimane, a monobromobimane analog, to protect against inactivation from BITC indicates this is unlikely. Additionally, S-(hydroxyethyl)bimane does not function as a competitive inhibitor of BITC, as it does with mBBr. Also, the mutations at Tyr 103 had different effects on the kinetic parameters for the mBBr and BITC reactions; the V max value in the mBBr reaction is strikingly decreased when either phenylalanine or serine replaces the original tyrosine. If BITC and mBBr shared the same site, then the two substrates would be affected equally by both mutations. Therefore, we propose that BITC binds at a discrete site of hGST P1-1 and that Tyr 103 makes different contributions toward the reactions of the electrophilic substrates bound at these two sites.
GST Pi has been implicated in the intracellular transport and storage of a broad range of structurally diverse hydrophobic ligands (36). The hydrophobic nonsubstrate or "ligandin"binding site has been proposed as distinct from the G-and H-sites (36). However, controversy still exists as to whether ligandin sites are separate or the same as those occupied by the xenobiotic substrates (42). We have demonstrated that either bromosulfophthalein or ANS prevents inactivation and that ANS decreases incorporation of BITC into Tyr 103 , raising the possibility that the BITC site overlaps a ligandin site of hGST P1-1. We therefore tested the ability of the BITC-modified enzyme to bind ANS, and we demonstrated that modification with BITC eliminates the high affinity binding site for ANS. The binding site for ANS may thus be near the xenobiotic substrate region. In addition, we tested the ability of Y103F and Y103S enzyme to bind ANS. The ANS binding of wild-type and Y103F enzyme is similar. In contrast, replacement of Tyr 103 by serine weakens binding of ANS to the high affinity site, suggesting that Tyr 103 is an important contributor to this site for ANS and a major component in binding is the aromatic ring of Tyr 103 , rather than its hydroxyl group.
For all of the enzymes (modified, wild-type, or mutant) ANS was bound at higher concentrations. This finding indicates the existence of more than one binding site for ANS. Because ANS is commonly used as a probe to assess the surface hydrophobicity of proteins (43), other sites in hGST P1-1 for ANS binding may be located at hydrophobic patches throughout the enzyme. Similarly, a high affinity and several low affinity sites have been reported recently (44) for a different nonsubstrate ligand, bromosulfophthalein, of the Alpha class GST A1-1. We here demonstrate that the high affinity binding site of ANS on hGST P1-1 is located near the BITC substrate region; however, the location of the low affinity site(s) remains unknown.
Examination of the crystal structure of hGST P1-1 allows us to better understand the experimental results. In Fig. 6A, glutathione is positioned in an orientation suitable for reaction with BITC at the active site. Because the aromatic ring of Tyr 103 is important for binding BITC; the aromatic rings of both Tyr 103 and BITC have been positioned appropriately for Pi-Pi stacking interactions, i.e. the rings should be Ͻ5.6 Å apart (45). The distance between the reactive carbon of BITC and the -SH of glutathione is ϳ2.4 Å, and the distance between the two aromatics is ϳ4 Å. Thus, the catalytic reaction of BITC and glutathione is feasible. The binding site for BSP, but not ANS, in hGST P1-1 is known from a crystal structure (46). Because BSP binds competitively with respect to ANS (36), the BSP-binding site was used to interpret results related to ANS. With BITC positioned to react with glutathione, as in Fig. 6A, it is clear that BITC and BSP bind at distinct sites, thereby accounting for the noncompetitive inhibition by ANS with respect to BITC as substrate.
An alternate mode of docking of BITC is shown in Fig. 6B, in which the reactive isothiocyanate group is oriented close to the -OH of Tyr 103 . The distance between the reactive carbon of BITC and the -OH of Tyr 103 is ϳ2.3 Å, whereas it is about 10.9 Å to the -SH of glutathione. This orientation can account for the protection provided by BSP and ANS against reaction at Tyr 103 . In Fig. 6C, an energy-minimized structure is shown FIG. 7. Overlay of hGST P1-1 complexed with dinitrobenzylglutathione and hGST P1-1 with BITC and mBBr. Ribbon representation of hGST P1-1 complexed with BITC with its reactive carbon facing the -SH of glutathione and complexed with monobromobimane with -CH 2 Br facing the -SH in glutathione (PDB code 19GS) overlaid with hGST P1-1 complexed with dinitrobenzylglutathione (PDB code 18GS). The structures are colored as follows: gold, hGST P1-1 subunit B; white, glutathione; pink, dinitrobenzyl-glutathione; green, carbon; blue, nitrogen; red, oxygen; brown, bromine. Atoms of protein side chains are shown in cyan. The BSP site is designated. that depicts the covalently modified Tyr 103 . We propose that BITC approaches Tyr 103 by the route indicated in Fig. 6B that passes through or close to the site occupied by BSP (Fig. 6B). However, after reaction occurs, the most stable conformation of the covalently modified Tyr 103 places the BITC moiety in the BITC substrate site (Fig. 6C); the location of the modified Tyr 103 thus accounts for the complete loss of activity with BITC as substrate but retention of some activity with either CDNB or monobromobimane as substrate.
BITC also reacts covalently with Cys 47 . In the structure shown in Fig. 6D, BITC has been docked into a site close to Cys 47 ; the distance between the reactive carbon of BITC and -SH of Cys 47 is ϳ1.8 Å. This structure shows that the aromatic ring of BITC overlaps the enzyme-bound glutathione; glutathione and BITC cannot be bound simultaneously, thereby explaining the protection provided by S-methylglutathione against reaction of BITC with Cys 47 .
To visualize whether BITC can really occupy a different location from monobromobimane and the dinitrobenzyl group of CDNB, Fig. 7 shows the structure of hGST P1-1 complexed with dinitrophenyl-glutathione and monobromobimane overlaid on the structure of the enzyme with BITC docked into the substrate site as in Fig. 6A. BITC, monobromobimane, and dinitrophenyl are clearly in three discrete locations in this model, indicating that the site for BITC, can be distinct from that of the other xenobiotic substrates.
Glutathione S-transferases function to protect an organism against potentially toxic compounds. Because an organism may encounter a wide variety of foreign chemicals through its lifetime, there is an advantage to having multiple classes of glutathione S-transferases with diverse substrate specificities. However, it is physiologically even more efficient for a given isozyme to evolve several xenobiotic substrate sites within the H-site to handle the wide array of foreign chemicals.
In summary, the results presented in this paper provide evidence for the existence of a novel binding site for xenobiotic substrates in hGST P1-1, which can be occupied by BITC, and support the existence of at least three discrete binding sites for electrophilic substrates within the xenobiotic substrate region of hGST P1-1. Benzyl isothiocyanate, in addition to serving as a substrate for hGST P1-1, acts as an affinity label. Modification of Tyr 103 occurs from the xenobiotic substrate site occupied by BITC during its reaction with glutathione, whereas reaction with Cys 47 takes place with BITC positioned at or near the glutathione-binding site.