A Homology Model for Rat Mu Class Glutathione S-Transferase 4-4

28

Chem. Res. Toxicol. 1996, 9, 28-40

A Homology Model for Rat Mu Class Glutathione S-Transferase 4-4 Marcel J. de Groot,†,‡ Nico P. E. Vermeulen,*,‡ Diana L. J. Mullenders,†,‡ and Gabrie¨lle M. Donne´-Op den Kelder†,‡ Leiden/Amsterdam Center for Drug Research (LACDR), Divisions of Molecular Toxicology and Medicinal Chemistry, Department of Pharmacochemistry, Vrije Universiteit, De Boelelaan 1083, 1081 HV Amsterdam, The Netherlands Received May 11, 1995X

Glutathione S-transferases (GSTs) are an important class of phase II (de)toxifying enzymes, catalyzing the conjugation of glutathione (GSH) to electrophilic species. Recently, a number of cytosolic GSTs was crystallized. In the present study, molecular modeling techniques have been used to derive a three-dimensional homology model for rat GST 4-4 based upon the crystal structure of rat GST 3-3, both members of the mu class. GST 3-3 and GST 4-4 isoenzymes share a sequence homology of 88%. GST 4-4 distinguishes itself from GST 3-3 in being much more efficient and stereoselective in the nucleophilic addition of GSH to epoxides and R,βunsaturated ketones. GST 3-3, however, is much more efficient in catalyzing nucleophilic aromatic substitution reactions. In this study, several known substrates of GST 4-4 were selected and their GSH conjugates docked into the active site of GST 4-4. GSH conjugates of phenanthrene 9(S),10(R)-oxide and 4,5-diazaphenanthrene 9(S),10(R)-oxide were docked into the active site of both GST 3-3 and GST 4-4. From these homology modeling and docking data, the difference in stereoselectivity between GST 3-3 and GST 4-4 for the R- and S-configured carbons of the oxirane moiety could be rationalized. The data acquired from a recently derived small molecule model for GST 4-4 substrates were compared with the results of the present protein homology model of GST 4-4. The energy optimized positions of the conjugates in the protein model agreed very well with the original relative positions of the substrates within the substrate model, confirming the usefulness of small molecule models in the absence of structural protein data. The protein homology model, together with the substrate model, will be useful to further rationalize the substrate selectivity of GST 4-4, and to identify new potential GST 4-4 substrates.

Introduction 2.5.1.18)1

Glutathione S-transferases (GSTs: EC catalyze the nucleophilic attack of the tripeptide glutathione (GSH) to electrophilic substrates, resulting in either addition or substitution reactions depending on the nature of the substrate (1). The primary function of these enzymes is generally considered to be detoxification of both endogenous and xenobiotic compounds, although in some cases the conjugation to GSH may also lead to toxification (1-3). In general, GST isoenzymes can accommodate a broad range of electrophilic substrates. However, some isoenzymes display a relatively high stereoselectivity (4, 5). Mammalian cytosolic GST isoenzymes have been classified into four major families, alpha, mu, pi, and theta * To whom correspondence should be addressed. † Division of Molecular Toxicology. ‡ Division of Medicinal Chemistry. X Abstract published in Advance ACS Abstracts, December 1, 1995. 1 Abbreviations: BPDE, benzo[a]pyrene 7(S),8(R)-diol 9(S),10(R)epoxide; CDNB, 1-chloro-2,4-dinitrobenzene; diazaPO, 4,5-diazaphenanthrene 9(S),10(R)-oxide; G-site, binding site for GSH; GS-BPDE, (binding mode of) glutathione conjugate of BPDE; GS-DNB, (binding mode of) glutathione conjugate of CDNB; GS-diazaPO, (binding mode of) glutathione conjugate of diazaPO; GS-PO, (binding mode of) glutathione conjugate of PO; GST, glutathione S-transferase; H-bond, hydrogen bond; H-site, substrate binding site; IS, interaction site in the substrate; pIS, protein interaction site; PO, phenanthrene 9(S),10(R)-oxide; R-conjugate, conjugate resulting from GSH conjugation at the R-configured carbon of the oxirane moiety; RHF, restricted Hartree Fock; S-conjugate, conjugate resulting from GSH conjugation at the S-configured carbon of the oxirane moiety; RMS, root mean square; STO, Slater type orbital.

0893-228x/96/2709-0028$12.00/0

(6, 7), based upon differences in their molecular masses, isoelectric points, and other properties. Cytosolic GSTs occur as homo- and heterodimeric proteins comprised of two subunits (8). Heterodimers can only be formed between subunits of the same family. The primary sequence homology between subunits of the same family is generally higher than 70%, while interfamily subunit homologies are generally between 30% and 40% (1, 9). Each GST isoenzyme subunit contains an active site consisting of two binding sites, one for the cofactor GSH (G-site) and one for the electrophilic substrate (H-site) (8, 10). Over the last four years, a number of crystal structures have been reported for a variety of mammalian and nonmammalian GST isoenzymes from various classes: human alpha class GSTA1-1 (9), rat liver mu class GST 3-3 (11-14), human muscle tissue mu class GSTM2-2 (15), pig lung pi class GST (16, 17), human placenta pi class GST (18), mouse liver pi class GST (19), a GST from Australian sheep blowfly (20), a GST from Schistosoma japonica (21), and a squid sigma class GST (22). These crystal structures reveal that all isoenzymes share a number of general structural features, but differ considerably in detail. In order to derive insight in the threedimensional properties of GST isoenzymes not (yet) crystallized, homology models can be build based on known crystal structures of the same enzyme class (23). Only 60% of the Caucasian population possesses fully active human GSTM1-1, which is highly active (together with human class pi GST isoenzymes (4)) in catalyzing © 1996 American Chemical Society

Homology Model for Rat GST 4-4

the conjugation of potentially toxic and carcinogenic alkene and arene oxides, such as styrene 7,8-oxide, benzo[a]pyrene 4,5-oxide, and benzo[a]pyrene 7(R),8(S)-diol 9(S),10(R)-epoxide (BPDE) (24, 25). The polymorphism of this isoenzyme is genetically determined and is caused by the absence of the GST 1 allele. Homozygous individuals for this null allele display a very low activity, or even a complete absence of activity, with respect to conjugation of trans-stilbene oxide (25-27). From various points of view it would be interesting to build and further investigate the protein structure of human GSTM1-1. However, as yet, insufficient tools are available to validate a homology model for human GSTM1-1. The aim of the present study is to derive and validate a homology model for rat GST 4-4 that will reveal the amino acids possibly responsible for protein-substrate interactions and possibly the specific catalytic behavior of this isoenzyme. A second objective is to explain the experimentally observed enhanced stereoselectivity of GST 4-4 in the conjugation of aromatic epoxides when compared to GST 3-3. For several reasons, rat GST 4-4 was chosen in the present study for structural investigations: (1) a high sequence homology exists with human GSTM1-1 (88%);2 (2) a large correspondence in substrate selectivity is observed between human GSTM1-1 and rat GST 4-4 (4, 30); (3) mutation studies (31) on rat mu class GST 3-3 have identified several important residues in the active site of this closely related isoenzyme; some mutations led to a conversion of the metabolic selectivity of GST 3-3 to that of GST 4-4; (4) recently, a small molecule substrate model for rat GST 4-4 has been derived (32), which is able to assist in verifying and validating a homology protein model of GST 4-4. The current homology model of GST 4-4 was derived from the crystal structure of rat mu class GST 3-3 (11). Several substrates from the recently derived small molecule substrate model (32) were docked as their corresponding S-conjugates into the homology model of GST 4-4 in order to validate the homology model.

Computational Methods The crystal structure of rat mu class GST 3-3 (11) was retrieved from the Brookhaven Protein Data Bank (33, 34) (reference code 1GST). The sequence of GST 4-4 was retrieved from the Swiss-Protein Database (reference code GTB2 rat (35, 36)).3 The homology building was carried out using Quanta version 4.0 (37) implemented on a Silicon Graphics Personal Iris workstation and IBM RS6000 workstations. All calculations were carried out using CHARMm version 22.0 (38, 39) on IBM RS6000 workstations. The programs SETOR (40, 41) (on a Silicon Graphics Personal Iris) and ChemX (42) (on IBM RS6000 workstations) were used for visualization. Alignment. The alignment of GST 3-3 and GST 4-4 was performed using Quanta and is shown in Figure 1. Both isoenzymes consist of 2 subunits comprising 217 amino acids. Homology Building. The backbone coordinates of all residues (2 × 217 amino acids) of GST 3-3 were used directly for building the GST 4-4 model. The coordinates of identical side chains (2 × 191 amino acids) were copied to the GST 4-4 model. The coordinates of the remaining side chains were calculated with CHARMm using standard atomic distances and bond angles (39). After a search for close contacts (interresidue 2 Homology determined by aligning primary sequences of rat class mu GST 4-4 (code: GTB2 rat) and human class mu GST 1-1 (code: GTM1 human) retrieved from the Swiss-Protein Database with the PSQ routines of the CAMMSA program package (28, 29). 3 Contains also mutant W146S.

Chem. Res. Toxicol., Vol. 9, No. 1, 1996 29 distances smaller than 1.5 Å (39)), these contacts were removed via a conformational search in which the number of close contacts was minimized. Within the conformational analysis, all internal side chain axes of the residue under investigation were rotated (see Figure 2) in steps of 60° and with a cutoff distance of 3.0 Å (i.e., interacting atoms more than 3.0 Å apart were not taken into account). Using Quanta/CHARMm, the lysine and arginine residues were by default protonated (positively charged) and the aspartic and glutamic acids deprotonated (negatively charged), while the histidines were protonated at one nitrogen of the imidazole ring (yielding a neutral heterocycle), as expected at physiological pH (43). This resulted in a total charge of +4 and 0 for the GST 3-3 and GST 4-4 dimers, respectively (see sequences in Figure 1), which was smoothed over carbons and nonpolar hydrogens (38). Geometry Optimizations and RMS Deviations. In general, geometry optimizations of the protein structures were performed with the CHARMm forcefield (39) using a five stage optimization procedure with increasing conformational freedom at each stage of the calculation: (1) the coordinates of all nonhydrogens were fixed, (2) only backbone atoms were fixed, (3 and 4) a harmonic constraint of 100 kcal‚mol-1‚ Å-1 and 10 kcal‚mol-1‚ Å-1, respectively, was applied to the backbone atoms, and (5) all restraints were removed and the complete structure was freely optimized. At each stage of the optimization process, the steepest descend method (44) followed by the conjugate gradient method (44, 45) was used. All calculations were done with the “all hydrogens” definition of CHARMm (i.e., all hydrogens were explicitly present in the calculations (39)), while hydrogen bonds were explicitly considered in the calculations.4 In general, root mean square (RMS) deviations in the atomic positions were used to evaluate structural differences. The crystal structure of GST 3-3 containing GSH in the G-site (11) was compared with the structure after optimization, in order to determine the influence of the geometry optimization procedure (CHARMm forcefield (38, 39)) on the crystal structure coordinates. Both the homology model of GST 4-4 and the crystal structure of GST 3-3 were geometry optimized in the presence and absence of GSH (for GSH, a conformation was chosen as found in the crystal structure of GST 3-3 (11)). The influence of GSH on the conformation of both protein structures was determined. Structural differences were also determined between the geometry optimized structures of GST 4-4 (with and without GSH) and the geometry optimized structures of GST 3-3. Possible effects of dimer formation on the geometry of the separate GST subunits were investigated by comparison of the geometry of a separately geometry optimized subunit 4 with the geometry of the same subunit in the geometry optimized GST 4-4 dimer. In order to evaluate the influence of waters of crystallization on the protein model of GST 4-4, all water molecules present in the crystal structure of GST 3-3 were included in the protein model of GST 4-4. A geometry optimization was carried out on GST 4-4 with GSH bound at the putative G-site and with 240 water molecules present within 3.5 Å from the protein. Several GST 4-4 structures containing conjugates (see below) were geometry optimized. Docking of GSH Conjugates. The recently derived substrate model for GST 4-4 (32) assigned possible amino acids for interaction with the substrates in the active site of GST 4-4 (Figure 3). The substrate model indicated the possible presence of three Lewis acid sites in the protein (notably, protein interaction sites 1, 2, and 3 (pIS1,2,3) in Figure 3) and a region in the protein responsible for aromatic interactions (pIS4 in Figure 3). Four key substrates forming the basis for this 4 CHARMm parameters used for all calculations were as follows: NSTE 5000, NPRI 25, TOLG 0.10, STEP 0.020, TOLS 0.00, TOLENR 0.00. The hydrogen bond parameters used were as follows: ALL, ACCE, IHBFRQ 50, CTONHB 3.5, CTOFHB 4, CUTHB 4.5, CTONHA 50, CTOFHA 70, CUTHBA 90. The nonbonded parameters used were as follows: RDIE, EPS 4.0, INBFRQ 50, CTONNB 10.50, CTOFNB 11.50, CUTNB 12.00, VSWITCH, SWITCH.

30

Chem. Res. Toxicol., Vol. 9, No. 1, 1996

de Groot et al.

Figure 1. Sequence alignment of GST 3-3 and GST 4-4 (78% identity, 88% homology). The 47 different amino acids are indicated in bold. Chemically similar amino acids are underlined.

Figure 2. Conformation analysis of peptide side chains (see text). Freely rotatable axes are indicated by arrows. substrate model (notably, BPDE, 1-chloro-2,4-dinitrobenzene (CDNB), phenanthrene 9(S),10(R)-oxide (PO), and 4,5-diazaphenanthrene 9(S),10(R)-oxide (diazaPO), see Figure 4) were docked into the homology protein model of GST 4-4: the glutathionyl part of the GSH conjugates was not changed with respect to the original position of GSH in the active site, whereas the substrate part of the GSH conjugates was optimally orientated in the active site by rotations around the axes indicated in Figure 2. The docking procedure for the conjugates consisted of five stages: (1) The geometry of the substrate was taken from the

Figure 3. Schematic representation of a substrate model for the active site of GST 4-4. Epoxide oxygen (IS1), C8-hydroxyl group (IS2), and C7-hydroxyl group (IS3) interact with protein interaction sites pIS1, pIS2, and pIS3, respectively. Aromatic interaction region (IS4) possibly interacts with aromatic amino acids in the protein (pIS4). Taken from ref 32. substrate model (32) and converted to an S-methyl conjugate with the molecular modeling package ChemX (42). (2) The geometry of the S-methyl conjugate was geometry optimized at the RHF (restricted Hartree Fock) level using the STO-3G (Slater type orbital comprised of 3 Gaussians (46)) minimal basis set of the ab initio quantum chemical program package GAMESSUK (47, 48) implemented on IBM RS6000 workstations. (3) The


Chem. Res. Toxicol., Vol. 9, No. 1, 1996 31

Figure 4. Selected GST 4-4 substrates which were docked in the active site of the derived GST 4-4 protein model after conjugation to GSH (Figure 6a-d) (see text for details). S-methyl group of the geometry optimized S-methyl conjugate was matched with the side chain of the cysteine residue from GSH taken from the geometry optimized GST 4-4 model with two GSH molecules, using the “molecular similarity” routines of Quanta. (4) One of the S-methyl groups was removed, and the respective GSH conjugate was formed. The overall charge of the conjugate was set to -1 (smoothed over carbons and nonpolar hydrogens). (5) The conjugate was docked in the GST 4-4 protein model (from which GSH was removed). A conformational analysis was then carried out around the freely rotatable axes of the cysteine-substrate part of the GSH conjugate and/or specific amino acids in the protein (Figure 2) in order to find optimal orientations for the substrate part of the conjugate in the H-site of GST 4-4. Determination of Binding and Interaction Energies. The binding energy of GSH to the isoenzymes was determined by comparing the energy of the geometry optimized protein containing GSH in both G-sites with the separate energies of the geometry optimized protein without GSH and the energy of 2 geometry optimized GSH molecules, i.e.:

Ebind(GSH) ) E(protein + 2GSH)opt E(protein)opt - 2E(GSH)opt where Ebind(GSH) is the binding energy of 2 GSH molecules to the dimer. In a similar way, the conjugate binding energies could be determined, and the interaction energy between the two subunits of GST 4-4.

Results and Discussion Alignment of GST 3-3 and GST 4-4. As indicated in Figure 1, GST 3-3 and GST 4-4 differ in 47 out of the 217 amino acid residues, resulting in 78% amino acid identity and 88% amino acid homology. Most differences involve single amino acids (no neighboring changes), except for three consecutive residues 128-130, i.e., K128E, T129G, and I130L, and a few pairs of residues (Table 1). Suitability of the CHARMm Forcefield for GSTs. In order to test the influence of the CHARMm forcefield on the conformation of GSTs, RMS deviations in the atomic positions were calculated between the original GST 3-3 crystal structure (11) (PDB reference code 1GST (33, 34)) and the geometry optimized structure of GST 3-3. An RMS value of 0.85 Å was obtained for the backbone and 0.83 Å for R-carbons. Furthermore, the RMS deviations between the crystal structures of GST 3-3 (11-13) (PDB reference codes 1GST (resolution 2.2 Å), 2GST (resolution 1.8 Å), 3GST (resolution 1.9 Å), 4GST (resolution 1.9 Å), and 5GST (resolution 2.0 Å) (33, 34)) and each of the others (1GST, 2GST, 3GST, 4GST,

Figure 5. SETOR representation (40, 41) of the homology protein model for GST 4-4 containing GSH in both subunits. R-Helices are indicated in either blue or red to differentiate between the two subunits, while β-sheets are depicted in white or green. GSH molecules are colored yellow or purple.

and 5GST) were found to be 0.45-0.80 Å for the nonhydrogens, 0.16-0.47 Å for the backbone and 0.14-0.45 Å for the R-carbons. These results are in line with the RMS values given by Ji et al. of 0.13 and 0.21 Å for the R-carbons between 2GST and 3GST, and 2GST and 1GST, respectively (13). All values remain well below the RMS values reported for GST crystal structures from different classes (1.87-2.09 Å for R-carbons (9)). The above results indicate that the effect of the CHARMm forcefield on the crystal structure of GST 3-3 is of a similar order of magnitude as the differences observed between crystals of the same structure. Therefore, the forcefield was concluded to be suitable for geometry optimizations of this class of protein structures. Homology Building. GST 4-4 was built using the sequence alignment with GST 3-3 as given in Figure 1. Due to amino acid differences, some side chain interactions in the three-dimensional protein model of GST 4-4 are different when compared to GST 3-3 (see Table 1). Several amino acid differences (22 out of 47) are found at the surface of the protein structure (indicated in Table 1 with an *). The final homology built protein model for GST 4-4 containing GSH in both G-sites is given in Figure 5. Specific interactions will be discussed below. Structural Influence of GSH on the Homology Model of GST 4-4. In the crystal structure of GST 3-3, GSH is bound (in both G-sites) in an extended conformation, with its backbone facing the cavity between the two protein subunits and the sulfur pointing toward the subunit to which it is bound (11). The coordinates of GSH were copied directly from the GST 3-3 crystal structure to the homology model of GST 4-4, thereby retaining the same conformation. The sulfurs of the two GSH molecules are 19.70 Å apart in both protein structures. Eight amino acids from one subunit (Trp7, Arg42, Trp45, Lys49, Asn58, Leu59, Gln71, and Ser72) and one amino acid from the other subunit (Asp105) are responsible for eleven hydrogen bonds formed between each GSH molecule and the protein. For both GST 3-3 and GST 4-4 the possible

32


de Groot et al.

Table 1. Amino Acid Differences between GST 3-3 and GST 4-4 and Their Structural Implications amino acid differences GST 3-3 f GST 4-4a,b I N V T P L S E R A N R M A H A I V V N M I L N K Q F

3* 8* 91 131 15 19 25* 29* 31* 33* 47* 67*( 76 80 84* 96 98 99 103 106 108(1 1111 113 116* 121* 122 126*

T D I A A F T D K S S H L G N V V I A T L A V S R K Y

K T I R D V A I Y H L A S Y T S L Q S N

128* 129 130 144 150 152 159 162 166 167* 184 195* 200* 202* 205* 2091 211* 213* 215* 216*

E G L Q N I V V H R V D G F K A M F N P

structural element (11)c

structural influence of amino acid differences between GST 4-4 and GST 3-3d,e

β1 gain of H-bond to M34 β-turn R1 R1 β2 β2 β4 R3 R3 R4 R4 R4 R4 R4 R4 R4 R4 β-turn R5a R5a R5a

second residue of R1-helix close to F157 Loss of H-bonds to L20 and Y27 Loss of H-bond to R31 Loss of H-bonds to E29, Y196 and R201 Loss of H-bond to Y78

Loss of H-bonds to Q109 and Y166, gain of H-bond to Y126

loss of H-bond to D118, H-bond to K122 instead of Q122 H-bonds to S116 and D118 instead of N116 and D118 Y facing charged H166 instead of F facing Y166, loss of H-bond to T129, gain of H-bonds to T106 and K133 loss of H-bonds to F126 and K133

R5b β-turn

R6 R6 R6 R6 R7 R8 β-turn β-turn cis-Pro-turn β-turn β-turn

loss of H-bond to F140, gain of H-bond to F147

H facing charged Y126 instead of Y facing F126, loss of H-bond to N106 gain of H-bond to D175 gain of H-bond to S199 loss of H-bond to Y160 and D164 gain of H-bond to N215 loss of H-bonds to Y115 and L211 loss of H-bonds to S209 and Q213 loss of H-bond to L211 gain of H-bond to K205 loss of internal H-bond to N216 and H-bond to K210

a Residues indicated with: (*) are at the surface of the protein structure; (() are at subunit interface; (1) are within 5 Å from the center of the H-site. b Nonpolar amino acids: A, C, F, G, I, L, M, P, V, W. Uncharged polar amino acids: H, N, Q, S, T, Y. Negatively charged amino acids: D, E. Positively charged amino acids: K, R. c Both in GST 3-3 and GST 4-4. d Hydrogen bonds were determined after energy optimization of the structures with and without GSH in the G-site. e Structural information concerns GST 4-4 relative to GST 3-3.

influence of GSH on the conformation of the protein was investigated. In case of GST 3-3 the RMS deviation between the protein structures with and without GSH is 0.14 Å for all atoms and 0.09 Å for both the R-carbons and the complete GST backbone. For GST 4-4 the RMS deviation between the protein structures with and without GSH was 0.20 Å for all atoms and 0.13 Å for R-carbons and the complete GST backbone. These small RMS values indicate that in the absence of GSH both the GST 3-3 and the GST 4-4 proteins optimize to a low energy structure which closely resembles the protein structures containing GSH. The binding energy of GSH to GST 3-3 was found to be larger than that of GSH to GST 4-4 (-79.7 and -75.7 kcal/mol, respectively, see Table 2). Waters of crystallization were copied from the crystal structure of GST 3-3 to the homology model of GST 4-4 and hardly influenced the GST 4-4 model (RMS 0.53 Å for all atoms).

Table 2. Results of Geometry Optimizations of GST 3-3 and GST 4-4 with and without GSH in G-Site G-site

Etotal (kcal/mol)

total H-bonds in complex

Ebindinga (kcal/mol)

empty GSH

GST 3-3 (Dimer) -5535.8 1138 -5716.0 1182

NAb -79.7

empty GSH

GST 4-4 (Dimer) -5469.3 1094 -5641.4 1139

NAb -75.7

GSH

Subunit 4 (Monomer) -2725.3 554

NAb

a Binding energy for binding one GSH molecule (more negative ) stronger binding). b Not applicable.

Structural Influence of the Subunits onto Each Other. As GSTs occur as dimers, the interactions between the two subunits of GST 4-4 are important and were therefore investigated. In the GST 4-4 protein



Table 3. Possible Hydrogen Bonding and Stacking Interactions between the Two Subunits of the Geometry Optimized Structure of GST 4-4a,b subunit A/B

subunit B/A

Lys49 Asn58 Gln71 Asn73 Ala74 Tyr78 Arg81 Asp105

Hydrogen Bonds Glu132 Gln102, Asp105 Gln102, Asp105 Asn101 Asn101 Asp97 Glu90, Arg93, Asp97 amino terminus of GSH

Phe56

Aromatic Stacking Phe140

a For symmetry reasons, only half of the interactions are given. Furthermore, due to small conformational differences between the subunits, a small number of interactions (4) are not classified as an H-bond. In total B-A + A-B interactions: 22 H-bonds expected; 18 found due to conformational differences. b Derived from geometry optimized structures in the presence and absence of GSH.

model, 18 hydrogen bonds between residues of the different subunits are present (Table 3). Furthermore, a stacking interaction between Phe56 and Phe140 is observed, and a hydrogen bond between the amino group of GSH in one subunit and an aspartic acid (Asp105) in the other subunit (Table 3). To investigate the conformational influence of the subunits of GST 4-4 onto each other, a single subunit containing GSH was geometry optimized. The RMS value with respect to a corresponding subunit within a dimer appears to be 0.23 Å for the R-carbons, 0.24 Å for the complete GST backbone, and an overall RMS deviation of 0.33 Å for all atoms of GST. This small RMS value indicates very few conformational changes occur upon removal of a subunit. The energy gained by forming a dimer from two subunits/monomers is 190.8 kcal/mol (Table 2: Etotal(dimer) - 2Etotal(monomer)). Molecular dynamics calculations are probably necessary to obtain a more accurate and detailed picture of the conformational influence (if any) of GST subunits onto each other. With the geometry optimizations performed in this study, interactions which are not present in the local energy minimum of the protein structure cannot be considered. Structural Comparison between GST 4-4 and GST 3-3. The present study indicates that the overall structure of the GST 4-4 homology model is similar to the structure of GST 3-3. The secondary structure elements (R-helices and β-sheets) of the two proteins are identical. The homology model of GST 4-4 with R-helices and β-sheets indicated is given in Figure 5. The RMS deviation between the geometry optimized structure of GST 3-3 and the homology model for GST 4-4 is 0.50 Å for the R-carbons, 0.52 Å for all backbone atoms, and 0.70 Å for all equivalent atoms. These values are in line with the calculated RMS deviations between the crystal structure and geometry optimized structure of GST 3-3 (0.14-0.45 Å for R-carbons and 0.45-0.80 Å for all atoms, see above). Therefore, the overall structure of the homology model of GST 4-4 can be concluded to closely resemble the crystal structure of GST 3-3. The differences observed in their catalytic behavior are probably caused by local effects of the protein environment, which are difficult to assess with theoretical energy minimization calculations; however, molecular dynamics and quantum chemical calculations can possibly be used to assess these local effects.

Close to the GSH binding site (G-site) in GST 4-4, the hydrophobic substrate binding site (H-site) can be found. The pocket in GST 4-4 comprises the side chains of Tyr6, Trp7, Ile9, and Leu12 on one side and Ala111, Tyr115, Phe208, and Ala209 on the other side. Other residues (within 5 Å from the substrate part of docked conjugates) are Arg10, Gly11, Ala13, Met34, Arg42, Arg107, Leu108, Leu110, Gln165, and Ile207. All residues originate from the same subunit. The active site region (with docked conjugates) is depicted in Figure 6a-d. The H-site cavity in GST 4-4 is somewhat larger, less congested, and more hydrophobic than in GST 3-3 due to three residue differences (V9I, I111A, and S209A, see Figure 1). The distance between the H-sites of the two subunits in the homology model of GST 4-4 is approximately 20-25 Å . Docking Substrates and GSH Conjugates in the Substrate Binding Site of GST 4-4. Optimal orientations of four substrates (Figure 4) containing the main characteristics of the recently derived substrate model for GST 4-4 (32) were established within the present homology model for GST 4-4. Substrates BPDE, PO, and diazaPO are conjugated to GSH via a nucleophilic addition reaction (49-51), while CDNB is conjugated via a nucleophilic aromatic substitution reaction (1). As mentioned under Computational Methods, the substrates were coupled to the cysteine sulfur of GSH, yielding the conjugated products. For each GSH conjugate several conformations (binding modes) were generated by keeping the GSH moiety fixed (in the G-site). Also, the side chains of some residues in the protein model were allowed several orientations (Figure 2) in order to create specific interactions between the conjugates and the protein, resulting in different binding modes for the GSH conjugates. Benzo[a]pyrene 7(R),8(S)-Diol 9(S),10(R)-Epoxide (BPDE; Table 4 and Figure 6a). For the GSH conjugate of BPDE, four possible binding modes were generated within GST 4-4 and geometry optimized with CHARMm. Binding mode 1 (GS-BPDE1) results from the conformation analysis as mentioned under Computational Methods and has the least close contacts with the protein. In binding mode 2, the orientation of the conjugate is unchanged, but Arg42 has been reoriented to provide a hydrogen bond with the substrate part of GS-BPDE2. In binding mode 3 (GS-BPDE3), the protein is unchanged but the conjugate has been reoriented to provide a hydrogen bond with Arg42. Binding mode 4 (GS-BPDE4) is intermediate between binding mode 2 and 3, changing both the orientation of the conjugate and Arg42, and directing the side chain of Arg42 to both hydroxyl groups of the conjugate. Table 4 summarizes the hydrogen bonding interactions between the substrate part of the GSH conjugate and specific amino acids in the active site of GST 4-4. Trp7 is involved in hydrogen bonding to the substrate part of the conjugate in all four binding modes (GS-BPDE1 to 4), while Tyr6 and/or Arg42 are additional hydrogen bonding partners. Tyr115 points toward the substrate moiety of the conjugate, while Phe208 can interact with its aromatic region. Other residues important to mention (31) are Ile9 and Ala111, which are approximately 3.6 ( 0.4 Å and 5.5 ( 1.5 Å removed from the substrate part of the conjugate; Figure 6a presents the binding mode with the most favorable binding and total energy. The differences in binding energies calculated for all four binding modes of GS-BPDE are relatively small

34


de Groot et al.

Figure 6. SETOR representation (40, 41) of the active site area of GST 4-4, containing (panel a (upper left)) GS-BPDE2, (panel b (upper right)) GS-DNB, (panel c (lower left)) GS-PO1 (R), or (panel d (lower right)) GS-diazaPO2 (R). Colors used are as follows: Conjugates: cyan; Tyr6: yellow; Trp7: pink; Arg42: blue; Tyr115: orange; Phe208: red; other amino acids: white. Table 4. Results of Geometry Optimizations of GST 4-4 Containing GS-BPDE in One Active Site and GSH in the Other Active Site conjugate in one of the active sites/binding mode

Etotal (kcal/mol)

H-bonds in complex


GS-BPDE1

-5636.2

1138

-118.6

GS-BPDE2c

-5645.3

1139

-127.8

GS-BPDE3d

-5637.4

1145

-119.8

GS-BPDE4d

-5640.8

1146

-123.2

specific H-bond interactions between protein structure and substrate part of the docked conjugateb Tyr6‚‚‚7-hydroxyl group (2×) Trp7‚‚‚8-hydroxyl group Tyr6‚‚‚7-hydroxyl group Trp7‚‚‚8-hydroxyl group Arg42‚‚‚8-hydroxyl group Arg42‚‚‚9-hydroxyl group Trp7‚‚‚7-hydroxyl group Arg42‚‚‚7-hydroxyl group (2×) Trp7‚‚‚7-hydroxyl group Arg42‚‚‚7-hydroxyl group (2×)

a Binding energy for binding one conjugate molecule (more negative ) stronger binding). b In all binding modes, Tyr115 is pointing at the substrate part of the conjugate; Phe208 can interact with the aromatic region of the substrate. c Displayed in Figure 6a. d Tyr6 points toward the aromatic part of the conjugate.

(within 10 kcal/mol). Increasing the number of hydrogen bonds formed between the substrate part of this conju-

gate and the protein structure from 3 (in binding modes 1, 3, and 4) to 4 (in binding mode 2) apparently enhances



Table 5. Results of Geometry Optimizations of GST 4-4 Containing GS-DNB in One Active Site and GSH in the Other Active Site Etotal (kcal/mol)

H-bonds in complex


GS-DNB1b

-5697.9

1167

-141.8

GS-DNB2b,c GS-DNB3 GS-DNB4 GS-DNB5 and H2Od GS-DNB6 and H2Ob

-5662.3 -5680.4 -5684.7 -5687.3 -5689.7

1153 1160 1143 1151 1163

-106.2 -124.3 -128.6 -131.2 -133.6

conjugate in one of the active sites/binding mode

specific H-bond interactions between protein structure and substrate part of the docked conjugate Tyr6‚‚‚p-nitro group Ala209‚‚‚p-nitro group Arg42‚‚‚o-nitro group (2×) none none none Arg107‚‚‚p-nitro group (2×)

a Binding energy for binding one conjugate (more negative ) stronger binding). b Tyr115 points to the p-nitro group of the substrate part of the conjugate. c Displayed in Figure 6b. d Tyr115 points to the o-nitro group of the substrate part of the conjugate.

Table 6. Results of Geometry Optimizations Containing GS-PO or GS-diazaPO in Both Active Sites conjugate in both active sites/binding modea

Etotal (kcal/mol)

Ebindingb (kcal/mol)

H-bonds in complex

GS-PO1 (R) GS-PO1 (S) GS-diazaPO1 (R)

-5758.8 -5755.5 -5650.1

GST 3-3 1199 1181 1149

GS-diazaPO1 (S)

-5617.1

1153

-102.9 -110.5 -45.9 -36.4

specific H-bond interactions between protein structure and substrate part of the docked conjugate none none Tyr6‚‚‚N (in ring) Tyr115‚‚‚O (from epoxide) Tyr6‚‚‚N (in ring)

GST 4-4

c

GS-PO1 (R)c GS-PO1 (S) GS-diazaPO1 (R)

-5698.0 -5685.4 -5680.6

1158 1153 1163

-105.7 -108.6 -94.4

GS-diazaPO1 (S)

-5712.6

1154

-117.4

GS-diazaPO2 (R, rot)d

-5681.5

1176

-94.8

GS-diazaPO2 (S, rot)

-5706.8

1152

-114.6

Tyr115‚‚‚O (from epoxide) (2×) Tyr115‚‚‚O (from epoxide) Tyr6‚‚‚N (in ring) Tyr115‚‚‚O (from epoxide) (2×) Tyr6‚‚‚N (in ring) Tyr115‚‚‚N (in ring) Arg42‚‚‚N (in ring) (2×) Arg42‚‚‚N (N in other ring) (2×) Tyr115‚‚‚O (from epoxide) Arg42‚‚‚N (in ring) (2×) Arg42‚‚‚N (other N in ring) (2×) Tyr115‚‚‚O (from epoxide)

a (R) ) R-conjugate, (S) ) S-conjugate. b Binding energy for binding one conjugate molecule (more negative ) stronger binding). Displayed in Figure 6c. d Displayed in Figure 6d.

the binding energy with approximately 5-9 kcal/mol (Table 4). 1-Chloro-2,4-dinitrobenzene (CDNB; Table 5 and Figure 6b). For this substrate, six binding modes were established. Binding mode 1 (GS-DNB1) was obtained as described in Computational Methods (smallest number of contacts with the protein). In binding mode 2 (GSDNB2), Arg42 was oriented toward the o-nitro group of GS-DNB. Binding modes 3 and 4 were derived from the orientation of trinitrobenzene in the crystal structure of GST 3-3 (12) (PDB reference code 4GST (33, 34)) since CDNB can be superimposed on trinitrobenzene in two different orientations (GS-DNB3 and GS-DNB4). Although a crystal structure of GST 3-3 complexed with CDNB is available (12) (PDB reference code 5GST (33, 34)), these data were not used since the CDNB conjugate was found outside the hydrophobic cavity possibly due to a slow release of the conjugate out of the GST 3-3 isoenzyme (52). Two additional binding modes (GSDNB5 and GS-DNB6) were established in which an integral water molecule, present in the crystal structure of GST 3-3 (12), was also included in the protein model of GST 4-4. Table 5 summarizes the various hydrogen bonding interactions for GS-DNB (1-6). In binding modes 1 and 2 we observe a functional role for Tyr6 or Arg42, which is taken over by Arg107 in binding mode 6. The three remaining binding modes do not reveal hydrogen bonds between the protein structure and the substrate part of the conjugate. Again, Tyr115 seems to interact with the

substrate part of the conjugate in most binding modes. The water of crystallization (in binding modes 5 and 6) is hydrogen bonded to Arg107 (just as in GST 3-3) and to the backbone atoms of Gly11 and Gln165. Rotation of the side chain of Arg107, which results in the loss of one hydrogen bond with the water molecule and gain of two hydrogen bonds with GS-DNB, is slightly favored (2.4 kcal/mol, binding modes 5 and 6 in Table 5). On average, Phe208 is 4.0 ( 1.0 Å removed from the p-nitro group, while Tyr6 is on average 4.6 ( 1.6 Å distant from this substituent. In some binding modes (GS-DNB 1, 5, and 6), Ile9 is expected to influence conjugation (distance 3.7 ( 0.8 Å). The possibly important Ala111 (31) lies within a range of 5.6 ( 2.2 Å from the substrate part of the CDNB conjugate. Figure 6b presents binding mode 2 (GS-DNB2). The differences in the binding energies between the six conformations are again small. No clear relationship seems to exist between the observed binding energies and the number of hydrogen bonds between the substrate part of the conjugate and the protein structure (Table 5). Phenanthrene 9(S),10(R)-Oxide (PO; Table 6 and Figure 6c). The conjugation of PO with GSH is catalyzed by both rat mu class GST 3-3 and GST 4-4 enzymes. The conjugation catalyzed by GST 4-4 is reported to be highly stereoselective (g99% at the R-configured carbon of the oxirane moiety (51)), while the observed stereoselectivity for GST 3-3 is much lower (43% (51)). In an attempt to explain the differences in stereoselectivity, the

36


compounds conjugated at either the R- or S-configured carbon of the oxirane moiety (the so-called R- and S-conjugates) were docked in both the homology structure of GST 4-4 and the geometry optimized crystal structure of GST 3-3. For both conjugates in the two isoenzymes, only binding mode 1 (GS-PO1; as indicated in Computational Methods) was generated. Figure 6c shows the R-conjugate in the active site of the GST 4-4 model. Table 6 reveals that in both isoenzymes the R-conjugate binds less tightly than the S-conjugate (-102.9 vs -110.5 kcal/mol for GST 3-3; -105.7 vs -108.6 kcal/mol for GST 4-4). This possibly indicates that the R-conjugate (experimentally the most abundant product (51)) will leave the respective active site more easily. Table 6 also gives the total energies of the complexes which support the observed stereoselective differences between GST 4-4 and GST 3-3: the complex containing the R-conjugate is favored over the complex containing the S-conjugate (-5758.8 vs -5755.5 and -5698.0 vs -5685.4 kcal/mol, for GST 3-3 and GST 4-4, respectively). The resulting calculated total energy differences between the R- and S-conjugate are -3.3 and -12.6 kcal/mol for GST 3-3 and GST 4-4, respectively. Using the relationship ∆G ) -RT ln(kR/kS), with kR being the fraction of R-conjugate, kS the fraction of S-conjugate, and ∆G approximated by the calculated total energy differences (∆E) between R- and S-conjugate in both cases (GST 3-3 and GST 4-4), a stereoselectivity of ∼100% would be predicted in favor of the R-conjugate (at 311K). Since the experimentally observed values are much lower (43% R-conjugate for GST 3-3 (51), corresponding to ∆G°obs ≈ 0.2 kcal/mol; 99% R-conjugate for GST 4-4 (51), corresponding to ∆G°obs ≈ -2.8 kcal/mol), the calculated energy differences can at best indicate trends for the stereoselective effects. We do not suggest that calculated total energy differences are generally valid determinants for stereoselectivity predictions, because substrate binding and the activation barriers for subsequent steps in the catalytic process will have to be taken into consideration as well. In the GST 4-4 complex both the R-conjugate and the S-conjugate of PO are hydrogen bonded to Tyr115. This hydrogen bond is absent in the GST 3-3 complex. Recently, crystal structures of GST 3-3 complexed with either of the two conjugates of PO became available (13) (PDB reference code 2GST and 3GST, respectively (33, 34)). The positions for the R- and S-conjugates in the crystal structures largely agree with the ones found in the present docking studies. In GST 4-4 Phe208 is on average 4.5 ( 1.1 Å removed from the aromatic substrate portion of the conjugate and only in binding mode 3 in a correct orientation to stabilize the aromatic rings. Ile9 in GST 4-4 is approximately 1.7 Å closer to the conjugates than Val9 in GST 3-3 while Ala111 in GST 4-4 is 0.6 Å further removed from the conjugates than Ile111 in GST 3-3. In both isoenzymes, Tyr6 and Tyr115 point (from opposite directions) to the aromatic regions of both conjugates. 4,5-Diazaphenanthrene 9(S),10(R)-Oxide (DiazaPO; Table 6 and Figure 6d). The observed stereoselectivity of GST 4-4 toward conjugation of diazaPO (78% (51)) was lower than that observed for PO (g99% (51)). GST 3-3 also showed a somewhat diminished stereoselectivity for the R-conjugate of diazaPO compared to PO (34% vs 43% (51)). The Lewis basic nitrogens present in diazaPO causing a different binding mode have been suggested to be the source for this diminished stereoselectivity (32). To verify this assumption, two

de Groot et al.

different binding modes were investigated for both the R- and the S-conjugate in the GST 4-4 model: one identical to binding mode 1 of PO (GS-diazaPO1), and one rotated over 90° relative to the first binding mode of diazaPO (GS-diazaPO2). Figure 6d presents binding mode 2 of the R-conjugate (GS-diazaPO2 (R)). For comparison, the R-conjugate and S-conjugate were also docked in the geometry optimized structure of GST 3-3 (binding mode 1 (GS-diazaPO1) only). In both binding modes of the conjugate of diazaPO to GST 4-4 (GS-diazaPO1 and GS-diazaPO2) the total energy of the complex of the R-conjugate is approximately 30 kcal/mol less favorable than that of the corresponding S-conjugate. When this value is compared to a value of 12 kcal/mol in favor of the R-conjugate of PO, it is to be expected that the observed stereoselectivity for conjugation at the R-configured carbon of diazaPO is considerably lower than that of PO, although the exact stereoselectivity cannot be predicted from the present calculations (see previous paragraph). To some extent, this preference will be opposed by a facilitated removal of the R-conjugate, i.e., compared to the S-conjugate, the Rconjugate of diazaPO may leave the active site of GST 4-4 more easily than the S-conjugate (Table 6). Our expectation of a reduced stereoselectivity for the Rconjugate is experimentally supported (51). Arg42 stabilizes binding mode 2 of diazaPO (GS-diazaPO2), which might be responsible for the diminished stereoselectivity compared to PO. Our modeling results, however, do not reveal a necessity for this completely different binding mode of diazaPO when compared to PO. The total energy of GST 3-3 with the R-conjugate of PO indicates this complex to be 3.3 kcal/mol more stable than the corresponding complex with the S-conjugate of PO. The total energy of GST 3-3 complexed with the R-conjugate of diazaPO is even 33 kcal/mol more stable than the complex with the S-conjugate of diazaPO. These calculations suggest an enhanced stereoselectivity for the Rconjugate of diazaPO compared to PO in contrast to the experimental findings (51). Therefore, CHARMm total energy differences apparently do not necessary give correct relative values. Neglect of other possibly important factors determining stereoselectivity seems not to be justified. An analysis of the binding modes 1 and 2 of GSdiazaPO in GST 4-4 reveals that three amino acids are responsible for binding the conjugate via hydrogen bonds, i.e., Tyr115 and Tyr6 or Arg42. In all binding modes Tyr6 and Tyr115 are pointing toward the conjugates (from opposite directions), either to the nitrogens, to the oxygens, or to the aromatic rings. Both Ile9 in GST 4-4 and Val9 in GST 3-3 are further removed from the substrate part of the S-conjugate than from the substrate part of the R-conjugate. The distance between conjugate (both R- and S-conjugate) and Ala111 in GST 4-4 is 1.4 Å larger than the distance between conjugate and Ile111 in GST 3-3. Amino Acids Important for Substrate Binding and Catalytic Behavior. Several amino acids possibly important for differences observed between GST 3-3 and GST 4-4 in their catalytic behavior or binding have been reported and were specifically mentioned in previous paragraphs. Some of these amino acids are different in GST 3-3 and GST 4-4 (see alignment in Figure 1). Tyr6. This amino acid has been reported to be responsible for lowering the pKa of bound GSH (53). Our calculations also indicate a role for this amino acid in


stabilizing the conjugates of BPDE, CDNB, and diazaPO (and possibly the substrates) in the active site of GST 4-4 by hydrogen bond interactions. Val9/Ile9. Recently, Shan and Armstrong indicated Ile9 in GST 4-4 to be responsible for the enhanced stereoselectivity of GST 4-4 compared to GST 3-3 (31). This conclusion was based on site-specific mutagenesis experiments, in which several active site residues of GST 3-3 were converted into their GST 4-4 analogues. All GST 3-3 mutants containing an Ile at position 9 showed an enhanced stereoselectivity (comparable to GST 4-4) for conjugation of trans-4-phenyl-3-buten-2-one and phenanthrene 9,10-oxide (Figure 4: PO) to GSH when compared to the GST 3-3 wild type. The influence of Ile9 on conjugates in the active site of GST 4-4 could not be explained. However, our results indicate that in GST 4-4 Ile9 is approximately 1.7 Å closer to the conjugate of PO than Ala9 in GST 3-3. Increased steric restrictions in GST 4-4 might explain the observed (31) increase in stereoselectivity. Thr13/Ala13. An interesting difference between GST 3-3 and GST 4-4 is related to the residue at position 13, a threonine in GST 3-3 and an alanine in GST 4-4 (Figure 1 and Table 1). Thr13 forms an on-face hydrogen bond to Tyr6, enhancing the effect of Tyr6 on the pKa of bound GSH (53), and resulting in a further lowering of the acid dissociation constant (54). Since this residue is not conserved in other rat mu class isoenzymes, this amino acid cannot be the only determinant in lowering the pKa of GSH. However, its effect on the pKa of GSH results in a more efficient deprotonation of GSH and might explain the enhanced rate of conjugation of substrates like CDNB by GST 3-3 when compared to GST 4-4. The role of this amino acid has not been examined in our study. Arg42. One or two Lewis acids capable of hydrogen bond interactions (e.g., Lys, Arg, or His) have been predicted to be present in the active site of GST 4-4 (32) (Figure 3). However, no lysine or histidine residues were found to be present in the active site area of GST 4-4, while several arginine residues were found: Arg10, Arg42, and Arg107. Arg42 appears to be a flexible amino acid in the vicinity of both the G-site and the H-site, located at the surface of the active site cavity and hydrogen bonded to GSH. Our modeling results reveal that Arg42 stabilizes some binding modes of CDNB and diazaPO and, in the case of diazaPO, might be responsible for a diminished stereoselectivity compared to PO. This influence on stereoselectivity can be ascertained by site-directed mutagenesis. To our opinion, changing residue 42 from an arginine to a lysine (retaining the positive charge), an alanine (neutralizing the charge), or a glutamic acid (changing from a positively to a negatively charged side chain) might give valuable information on the importance of this residue. The experimental results will have to be examined very carefully since Arg42 is also capable of forming hydrogen bonds to GSH, stabilizing the structure. Ile111/Ala111. Shan and Armstrong suggested Ala111 to be one of the amino acids responsible for the enhanced catalytic activity of GST 4-4 toward para-substituted 4-phenyl-3-buten-2-ones (31) when compared to GST 3-3 with an Ile at position 111. GST 3-3 mutants containing an Ala at position 111 showed an enhanced catalytic activity, comparable to GST 4-4. The catalytic role of this amino acid could not be deduced from the present modeling study, since information on catalytic activities


cannot be obtained from theoretical interaction studies. Our calculations indicated that the orientation of residue 111 with respect to PO and diazaPO is very similar in both isoenzymes. Tyr115. A study of Barycki and Colman revealed Tyr115 in GST 4-4 to be involved in binding and enzymatic catalysis of certain types of substrates (55). When Tyr115 of GST 4-4 was chemically modified with 4-(fluorosulfonyl)benzoic acid, the 1,4-Michael addition reaction of GSH to trans-4-phenyl-3-buten-2-one and conjugation of GSH to trans-stilbene oxide was nearly abolished, while CDNB was still conjugated. The suggestion above is further supported by results from Katusz and Colman (56), who chemically modified Tyr115 in GST 4-4 with S-(4bromo-2,3-dioxobutyl)glutathione and thereby deactivated GST 4-4. Johnson et al. (52) suggested that the hydroxyl group of Tyr115 in GST 3-3 contributes to the enzymatic catalysis of the oxirane ring opening in phenanthrene 9,10-oxide. Furthermore, Ji et al. (13) indicated the importance of Tyr115 in GST 3-3 for the conjugation reaction between GSH and phenanthrene 9,10-oxide and trans-4-phenyl-3-buten-2-one. Both substrates were suggested to benefit from electrophilic assistance in the stabilization of oxyanions formed by GSH conjugation. All the above mentioned studies indicate the importance of Tyr115 in enzymatic catalysis. At neutral pH the hydroxyl group of Tyr115 is predominantly present in a protonated neutral state (55), and capable of sharing a proton with an epoxide moiety of the substrate, thereby assisting in the ring opening process. In alpha class GSTs (9), a valine (Val111) is present at a position equivalent to the Tyr115 position in mu class GSTs. As expected, alpha class isoenzymes reveal a very low specific activity toward aromatic epoxides (4), again confirming the relevance of Tyr115 in the catalysis of GSH conjugation of these substrates. The present study indicates that Tyr115 could also assist in stabilizing the substrate part of conjugates of PO and diazaPO via hydrogen bond formation. Phe208. This amino acid is situated in the hydrophobic substrate binding site of GST 4-4, and although not proven to be important for catalytic action, it possibly stabilizes aromatic substrates via π-π interactions. Mutating Phe208 to an alanine residue might reveal the importance of an aromatic amino acid in the H-site. In case no significant changes are observed in the catalytic activity and/or substrate binding, the hydrophobicity of the pocket is apparently sufficient and π-π interactions are less important. Ser209. In GST 3-3, Tyr115 is hydrogen bonded to Ser209 (52), which has been suggested to be responsible for closure of the active site channel and to impede the release of the GSH conjugation product of CDNB. In GST 4-4, Ser209 is replaced by an alanine, possibly relieving spatial restrictions of the structural elements containing residues 115 and 209 (52). Mutation studies changing this amino acid into a serine might reestablish the restrictions found in GST 3-3. Superposition of Conjugates and Comparison with the Substrate Model. In order to compare the relative orientation of the substrates from the original substrate model (32) with the relative orientation of the conjugates as derived in this study, the conjugates were superimposed, including all binding modes in the orientations as obtained from the docking studies (Figure 7a). The similarity between the van der Waals volume of the

38


de Groot et al.

Figure 7. (Panel a (left panel)) Superposition of the energetically most favorable binding modes of GSH conjugates of various substrates in the orientation as found in the active site of the derived protein model for GST 4-4; (panel b (right panel)) visual comparison of the relative orientation and size of the van der Waals volumes of the derived protein model of GST 4-4 (white) and the recently reported substrate model of GST 4-4 (red) (32)

substrate part of the conjugates and the previously derived substrate model (32) is evident (Figure 7b). Inspection of the protein model reveals that the distance between the oxygens of Tyr6 and Tyr115 is approximately 6.8 Å . All substrate parts of the conjugates are located in between these two residues (compare Figures 6 and 7). Tyr115 is positioned at the access channel of the substrate binding site and has been reported to be capable of electrophilic assistance in epoxide ring opening. Additionally, we showed that Tyr115 may hydrogen bond to the newly formed hydroxyl group in the products of the conjugation reaction. Arg42 is capable of forming hydrogen bonds to atoms situated at about 3.7 Å from the site of conjugation. Tyr6 forms hydrogen bonds to Lewis basic atoms at a distance of about 4.3 Å from the site of conjugation to GSH. Therefore, it seems likely that Tyr115 corresponds to protein interaction site 1 (pIS1, Figure 3) as mentioned in the GST 4-4 substrate model (32), while protein interaction sites 2 and 3 (pIS2 and pIS3, respectively; Figure 3) are likely to correspond to Arg42 and Arg42/Tyr6, respectively. The aromatic region in the substrate model probably interacts with Phe208 (pIS4, Figure 3) (32) and/or other hydrophobic amino acids located nearby.

Summary The present structural model of GST 4-4 derived by homology modeling closely resembles the overall crystal structure of GST 3-3. Furthermore, the orientation of four GSH conjugates docked into the GST 4-4 model appears to be similar to the orientation of substrates as present in the recently derived substrate model for GST 4-4 (32). Most importantly, all amino acids indicated in the literature to be important in the substrate binding 5 Coordinates of the model (with and without GSH and conjugates) are available from the authors upon request.

site of GST 4-4 and/or presumed to be responsible for the enhanced stereoselectivity in the conjugation of aromatic epoxides of GST 4-4 to GSH when compared to GST 3-3 (31) were found in the active site area of the model. Furthermore, the model reveals trends which qualitatively explain the stereoselective conjugation of phenanthrene oxides and diazaphenanthrene oxides on the R-configured carbon of the oxirane moiety by comparing the energies of the conjugate-protein complexes (formation of the conjugate) and the binding energies of the conjugates (product release of the conjugate) for GST 4-4. The observed differences between GST 3-3 and GST 4-4 toward conjugation of these epoxides could not be supported by corresponding calculations, indicating that predictions based on CHARMm total energies remain speculative and should be interpreted with extreme care. Overall, the homology model of GST 4-4 explains various experimental data and agrees very well with the recently proposed substrate model of GST 4-4. The present homology model5 identifies amino acids responsible for specific interactions with the substrate/conjugate, which can be tested experimentally by site-directed mutagenesis experiments.

Acknowledgment. IBM is gratefully acknowledged for providing RS6000 workstations. We thank Drs. E. M. van der Aar for helpful discussions.

References (1) Armstrong, R. N. (1991) Glutathione S-transferases: reaction mechanism, structure, and function. Chem. Res. Toxicol. 4, 131140. (2) van Bladeren, P. J., and van Ommen, B. (1991) The inhibition of glutathione S-transferases: mechanism, toxic consequences and therapeutic benefits. Pharmacol. Ther. 51, 35-46. (3) Commandeur, J. N. M., Stijntjes, G. J., and Vermeulen, N. P. E. (1995) Enzymes and transport systems involved in the formation and disposition of glutathione S-conjugates. Role in bioactivation and detoxication mechanisms of xenobiotics. Pharmacol. Rev. 47, 271-330.

Homology Model for Rat GST 4-4 (4) Mannervik, B., and Danielson, U. H. (1988) Glutathione transferasessstructure and catalytic activity. CRC Crit. Rev. Biochem. 23, 283-337. (5) Vermeulen, N. P. E., and te Koppele, J. M. (1993) Stereoselective biotransformation. Toxicological consequences and implications. In Drug stereochemistry. Analytical methods and pharmacology (Wainer, I. W., Eds) pp 245-280, Marcel Dekker Inc., New York. (6) Mannervik, B., A° lin, P., Guthenberg, C., Jensson, H., Tahir, M. K., Warholm, M., and Jo¨rnvall, H. (1985) Identification of three classes of cytosolic glutathione transferases common to several mammalian species: correlation between structural data and enzymatic properties. Proc. Natl. Acad. Sci. U.S.A. 82, 72027206. (7) Meyer, D. J., Coles, B., Pemble, S. E., Gilmore, K. S., Fraser, G. M., and Ketterer, B. (1991) Theta, a new class of glutathione transferases purified from rat and man liver. Biochem. J. 274, 409-414. (8) Danielson, U. H., and Mannervik, B. (1985) Kinetic independence of the subunits of cytosolic glutathione transferase from the rat. Biochem. J. 231, 263-267. (9) Sinning, I., Kleynwegt, G. J., Cowan, S. W., Reinemer, P., Dirr, H. W., Huber, R., Gilliland, G. L., Armstrong, R. N., Ji, X., Board, P. G., Olin, B., Mannervik, B., and Jones, T. A. (1993) Structure determination and refinement of human alpha class glutathione transferase A1-1, and a comparison with the mu and pi class enzymes. J. Mol. Biol. 232, 192-212. (10) Mannervik, B., Guthenberg, C., Jakobson, I., and Warholm, M. (1978) Glutathione conjugation: reaction mechanism of glutathione S-transferase A. In Conjugation reactions in drug biotransformation (Aitio, A., Ed.) pp 101-110, Elsevier/NorthHolland Biomedical Press, Amsterdam. (11) Ji, X., Zhang, P., Armstrong, R. N., and Gilliland, G. L. (1992) The three-dimensional structure of a glutathione S-transferase from the mu gene class. Structural analysis of the binary complex of isoenzyme 3-3 and glutathione at 2.2-Å resolution. Biochemistry 31, 10169-10184. (12) Ji, X., Armstrong, R. N., and Gilliland, G. L. (1993) Snapshots along the reaction coordinate of an SNAr reaction catalysed by glutathione transferase. Biochemistry 32, 12949-12954. (13) Ji, X., Johnson, W. W., Sesay, M. A., Dickert, L., Prasad, S. M., Ammon, H. L., Armstrong, R. N., and Gilliland, G. L. (1994) Structure and function of the xenobiotic substrate binding site of a glutathione S-transferase as revealed by X-ray crystallographic analysis of product complexes with the diastereomers of 9-(Sglutathionyl)-10-hydroxy-9,10-dihydrophenanthrene. Biochemistry 33, 1043-1052. (14) Fu, J. H., Rose, J., Tam, M. F., and Wang, B. C. (1994) New crystal forms of a µ-class glutathione S-transferase from rat liver. Acta Crystallogr. D50, 219-224. (15) Raghunathan, S., Chandross, R. J., Kretsinger, R. H., Allison, T. J., Penington, C. J., and Rule, G. S. (1994) Crystal structure of human class mu glutathione transferase GSTM2-2. Effects of lattice packing on conformational heterogeneity. J. Mol. Biol. 238, 815-832. (16) Reinemer, P., Dirr, H. W., Ladenstein, R., Scha¨ffer, J., Galley, O., and Huber, R. (1991) The three-dimensional structure of class π glutathione S-transferase in complex with glutathione sulfonate at 2.3 Å resolution. EMBO J. 10, 1997-2005. (17) Dirr, H., Reinemer, P., and Huber, R. (1994) Refined crystal structure of porcine class pi glutathione S-transferase (pGST P11) at 2.1 Å resolution. J. Mol. Biol. 243, 72-92. (18) Reinemer, P., Dirr, H. W., Ladenstein, R., Huber, R., Bello, M. L., Federici, G., and Parker, M. W. (1992) Three-dimensional structure of class π glutathione S-transferase from human placenta in complex with S-hexylglutathione at 2.8 Å resolution. J. Mol. Biol. 227, 214-226. (19) Garcia-Saez, I., Parraga, A., Phillips, M. F., Mantle, T. J., and Coll, M. (1994) Molecular structure at 1.8 Å of mouse liver classPi glutathione S-transferase complexed with S-(p-nitrobenzyl)glutathione and other inhibitors. J. Mol. Biol. 237, 298-314. (20) Wilce, M. C. J., Feil, S. C., Board, P. G., and Parker, M. W. (1994) Crystallization and preliminary X-Ray diffraction studies of a glutathione S-transferase from the Australian sheep blowfly, Lucilia Cuprina. J. Mol. Biol. 236, 1407-1409. (21) McTigue, M. A., Williams, D. R., and Tainer, J. A. (1995) Crystal structures of a schistosomal drug and vaccine target: glutathione S-transferase from Schistosoma japonica and its complex with the leading antischistosomal drug praziquantel. J. Mol. Biol. 246, 21-27. (22) Ji, X., von Rosenvinge, E. C., Johnson, W. W., Tomarev, S. I., Piatigorsky, J., Armstrong, R. N., and Gilliland, G. L. (1995) Three-dimensional structure, catalytic properties, and evolution of a sigma class glutathione transferase from squid, a progenitor of the lens S-crystallins of cephalopods. Biochemistry 34, 53175328.

Chem. Res. Toxicol., Vol. 9, No. 1, 1996 39 (23) Cachau, R. E., Erickson, J. W., and Villar, H. O. (1994) Novel procedure for structure refinement in homology modeling and its application to the human class Mu glutathione S-transferases. Protein Eng. 7, 831-839. (24) Dostal, L. A., Guthenberg, C., Mannervik, B., and Bend, J. R. (1988) Stereoselectivity and regioselectivity of purified human glutathione transferases π, R-, and µ with alkene and polycyclic arene oxide substrates. Drug Metab. Dispos. 16, 420-424. (25) Board, P., Coggan, M., Johnston, P., Ross, V., Suzuki, T., and Webb, G. (1990) Genetic heterogeneity of the human glutathione transferases: a complex of gene families. Pharmacol. Ther. 48, 357-369. (26) Roots, I., Brockmo¨ller, J., Drakoulis, N., and Loddenkemper, R. (1992) Mutant genes of cytochrome P-450IID6, glutathione Stransferase class Mu, and arylamine N-acetyltransferase in lung cancer patients. Clin. Invest. 70, 307-319. (27) Comstock, K. E., Widersten, M., Hao, X. Y., Henner, W. D., and Mannervik, B. (1994) A comparison of the enzymatic and physicochemical properties of human glutathione transferase M4-4 and three other human Mu class enzymes. Arch. Biochem. Biophys. 311, 487-495. (28) George, D. G., and Barker, W. C. (1989) PSQ user’s guide, Protein Identification Resource, National Biomedical Research Foundation, Washington. (29) Leunissen, J. A. M., Vergert, J. M. ter, and Gelder, C. W. G. van (1990) CAMMSA user manual version 1.0 ed., CAOS/CAMM Center, University of Nijmegen, Nijmegen. (30) Ketterer, B., Meyer, D. J., and Clark, A. G. (1988) Soluble glutathione transferase isozymes. In Glutathione conjugation: Mechanism and biological significance (Sies, H., and Ketterer, B., Eds.) pp 73-135, Academic Press, London. (31) Shan, S., and Armstrong, R. N. (1994) Rational reconstruction of the active site of a class Mu glutathione S-transferase. J. Biol. Chem. 269, 32373-32379. (32) de Groot, M. J., van der Aar, E. M., Nieuwenhuizen, P. J., van der Plas, R. M., Donne´-Op den Kelder, G. M., Commandeur, J. N. M., and Vermeulen, N. P. E. (1995) A predictive substrate model for rat glutathione S-transferase 4-4. Chem. Res. Toxicol. 8, 649-658. (33) Bernstein, F. C., Koetzle, T. F., Williams, G. J. B., Meyer, E. F., Jr., Brice, M. D., Rodgers, J. R., Kennard, O., Shimanouchi, T., and Tatsumi, M. (1977) The Protein Data Bank: a computerbased archival file for macromolecular structures. J. Mol. Biol. 112, 535-542. (34) Abola, E. E., Bernstein, F. C., Bryant, S. H., Koetzle, T. F., and Weng, J. (1987) Protein Data Bank. In Crystallographic databasessinformation contents, software systems, scientific applications (Allen, F. H., Bergerhoff, G., and Sievers, R., Eds.) pp 107132, Data Commision of the International Union of Crystallography, Bonn/Cambridge/Chester. (35) Ding, G. J. F., Ding, V. D. H., Rodkey, J. A., Bennett, C. D., Lu, A. Y. H., and Pickett, C. B. (1986) Rat liver glutathione Stransferases. DNA sequence analysis of a Yb2 cDNA clone and regulation of the Yb1 and Yb2 mRNAs by phenobarbital. J. Biol. Chem. 261, 7952-7957. (36) A° lin, P., Mannervik, B., and Jo¨rnvall, H. (1986) Cytosolic rat liver glutathione transferase 4-4. Primary structure of the protein reveals extensive differences between homologous glutathione transferases of classes Alpha and Mu. Eur. J. Biochem. 156, 343350. (37) The University of York (1992) Quanta, version 4.0. (38) Harvard College (1992) CHARMm (Chemistry at HARvard Macromolecular mechanics), version 22.0. (39) Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States, D. J., Swaminathan, S., and Karplus, M. J. (1983) CHARMm: A program for macromolecular energy, minimization and dynamics calculations. J. Comput. Chem. 4, 187-217. (40) Evans, S. V. (1993) SETOR: Hardware-lighted three-dimensional solid model representations of macromolecules. J. Mol. Graphics 11, 134-138. (41) Evans, S. V., and University of British Columbia (1995) SETOR, version 4.13.4. (42) Chemical Design Ltd. (1993) ChemX, version January 1993. (43) Stryer, L. (1981) Biochemistry, 2nd ed., W. H. Freeman & Co., New York. (44) Biosym Technologies (1994) Minimization. In Discover 2.9.5/94.0 user guide, Biosym Technologies, San Diego. (45) Schlegel, H. B. (1987) Optimization of equilibrium geometries and transition structures. In Ab Initio Methods in Quantum Chemistry, (Lawley, K. P., Ed.) pp 249-286, John Wiley & Sons, London. (46) Hehre, W. J., Stewart, R. F., and Pople, J. A. (1969) Self consistent molecular orbital methods. I. Use of Gaussian expansions of Slater-type atomic orbitals. J. Chem. Phys. 51, 2657-2664.

40


(47) Dupuis, M., Spangler, D., and Wendoloski, J. (1980) NRCC Program No. QG01 (GAMESS). (48) Guest, M. F., van Lenthe, J. H., Kendrick, J., Schoffel, K., Sherwood, P., Harrison, R. J., with contributions from Amos, R. D., Buenker, R. J., Dupuis, M., Handy, N. C., Hillier, I. H., Knowles, P. J., Bonacic-Koutecky, V., von Niessen, W., Saunders, V. R., and Stone, A. J. (1993) GAMESS-UK, IBM RS6000 version 2.1. (49) Robertson, I. G. C., Jensson, H., Mannervik, B., and Jernstro¨m, B. (1986) Glutathione transferases in rat lung: the presence of transferase 7-7, highly efficient in the conjugation of glutathione with the carcinogenic (+)-7β,8R-dihydroxy-9R,10R-oxy-7,8,9,10tetrahydrobenzo[a]pyrene. Carcinogenesis 7, 295-299. (50) Robertson, I. G. C., and Jernstro¨m, B. (1986) The enzymatic conjugation of glutathione with bay-region diol-epoxides of benzo[a]pyrene, benz[a]anthracene and chrysene. Carcinogenesis 7, 1633-1636. (51) Boehlert, C. C., and Armstrong, R. N. (1984) Investigation of the kinetic and stereochemical recognition of arene and azaarene oxides by isozymes A2 and C2 of glutathione S-transferase. Biochem. Biophys. Res. Commun. 121, 980-986. (52) Johnson, W. W., Liu, S., Ji, X., Gilliland, G. L., and Armstrong, R. N. (1993) Tyrosine 115 participates both in chemical and

de Groot et al.

(53)

(54)

(55)

(56)

physical steps of the catalytic mechanism of a glutathione S-transferase. J. Biol. Chem. 268, 11508-11511. Liu, S., Zhang, P., Ji, X., Johnson, W. W., Gilliland, G. L., and Armstrong, R. N. (1992) Contribution of tyrosine 6 to the catalytic mechanism of isoenzyme 3-3 of glutathione S-transferase. J. Biol. Chem. 267, 4296-4299. Liu, S., Ji, X., Gilliland, G. L., Stevens, W. J., and Armstrong, R. N. (1993) Second-sphere electrostatic effects in the active site of glutathione S-transferase. Observation of an on-face hydrogen bond between the side chain of threonine 13 and the π-cloud of tyrosine 6 and its influence on catalysis. J. Am. Chem. Soc. 115, 7910-7911. Barycki, J. J., and Colman, R. F. (1993) Affinity labeling of glutathione S-transferase, isoenzyme 4-4, by 4-(fluorosulfonyl)benzoic acid reveals Tyr115 to be an important determinant of xenobiotic substrate specificity. Biochemistry 32, 13002-13011. Katusz, R. M., and Colman, R. F. (1991) S-(4-Bromo-2,3dioxobutyl)glutathione: a new affinity label for the 4-4 isoenzyme of rat liver glutathione S-transferase. Biochemistry 30, 11230-11238.

TX950082I

A Homology Model for Rat Mu Class Glutathione S-Transferase 4-4

Recommend Documents