Use of Molecular Simulation for Mapping Conformational CYP2E1 Epitopes*

The identification of the epitopes recognized by autoantibodies against cytochrome P450s (CYPs) associated with drug-induced hepatotoxicity is difficult because of their conformational nature. In the present investigation, we used a novel approach based on the analysis of the whole molecule antigenic capacity following single amino acid substitutions to identify the conformational epitopes on CYP2E1. A molecular model of CYP2E1 was generated based on the CYP2C5 crystal structure, and potential motifs for amino acid exchanges were selected by computer simulation in the surface of α helices and β sheets. Fourteen modified, apparently correctly folded CYP2E1 variants were produced in Escherichia coli and evaluated in immunoprecipitation experiments using sera with anti-CYP2E1 autoreactivity from 10 patients with halothane hepatitis and 12 patients with alcoholic liver disease. Ala substitution of Glu-248 and Lys-251 as well as of Lys-324, Lys-342, Lys-420, and Phe-421 severely decreased or abolished CYP2E1 recognition by the majority of both the halothane hepatitis and alcoholic liver disease sera, whereas the other substitutions had only minor effects. Based on the structural model, these substitutions identified two distinct epitopes on the CYP2E1 surface corresponding to the G-helix and an area formed by juxtaposition of the J′ and K″ helices, respectively. The combined use of molecular modeling and single amino acid mutagenesis is thus a useful approach for the characterization of conformational epitopes recognized by autoantibodies.

ketone bodies (1). Studies by Bourdi and et al. (2) and Eliasson and Kenna (3) have shown that patients suffering from halothane hepatitis develop autoantibodies specifically targeting CYP2E1. Similar autoantibodies are also detectable in anesthesiologists exposed to halogenated anesthetic gases (4). We have reported that chronic intragastric alcohol-fed rats develop IgG directed toward CYP2E1, and their titers correlate with the extent of hepatic injury (5). These observations have been confirmed in humans, showing that high titers of anti-CYP2E1 autoantibodies are present in about 40% of patients with advanced alcoholic liver disease but not in heavy drinkers without liver damage (6). In the former, the presence of anti-CYP2E1 IgG correlated with the extent of lymphocyte infiltration in liver biopsies (7), suggesting a possible contribution of autoimmune mechanisms in the pathogenesis of alcohol liver injury.
These observations, along with the evidence indicating that CYP2E1 autoantibodies target functionally active CYP2E1 present on the outer layer of hepatocyte plasma membrane (3), prompted us to investigate the epitope specificity of CYP2E1 autoantibodies associated to halothane hepatitis or alcoholic liver disease to understand more about their formation and to get information of value for the development of more specific diagnostic tests. For this purpose, we used a novel approach by analyzing the whole molecular antigenic capacity following single amino acid substitutions designed using a structural model of CYP2E1 generated with the crystal CYP2C5 structure as a template.

EXPERIMENTAL PROCEDURES
Patients-For this study, the sera of 10 patients with unexplained hepatitis following multiple halothane anesthesia and negative for markers of hepatitis viruses or evidence of exposure to hepatotoxic drugs or alcohol were used along with 12 sera from patients with severe alcoholic liver disease (ALD). Sera from 10 patients with halothane hepatitis were generously provided by Dr. J. G. Kenna, Imperial College School of Medicine, London, UK. Their properties have been described in previously published studies (22,23). In brief, halothane hepatitis was defined clinically as severe, otherwise unexplained, histologically confirmed hepatitis occurring within 4 weeks after halothane exposure in patients with normal preoperative liver function.
The diagnosis of ALD was based on clinical, ecographical, and laboratory criteria. Liver biopsies were available for all the ALD patients and showed the classical features of micronodular cirrhosis, hepatocyte ballooning degeneration with Mallory's bodies, and inflammatory infiltrates. The patients with alcohol abuse were negative for serum markers for hepatitis virus and for the presence of hepatitis C virus RNA. The reactivity of the sera with CYP2E1 was preliminarily assessed by enzyme-linked immunosorbent assay using as antigen recombinant human CYP2E1 (Oxford Biochemicals Inc., Oxford, MI) (300 ng/well) as reported previously (6) and was at least three times higher (1:100 dilution) than those of 50 control sera.
Immunoprecipitation of [ 35 S]Methionine-labeled CYP2E1-[ 35 S]Methionine-labeled wild-type CYP2E1 and the N-terminal and C-terminal deleted forms were produced from the respective pGEM4z plasmid cDNA (1 g) using a rabbit reticulocyte "in vitro" translation/transcription system TNT T7-quick coupled T translation/transcription kit (Promega, Madison, WI) and 10 mol/liter [ 35 S]methionine (1,000 Ci/mmol) according to the manufacturer's instructions. The N-terminal moiety contained amino acids 1-222, and the C-terminal part contained the amino acids 223-493, i.e. starting with the sequence just prior to the G-helix. These constructs were kindly provided by Dr. Inger Johansson, Karolinska Institutet. For immunoprecipitation experiments, 5 l of the transcription/translation mixture were diluted (1:20) with RIPA buffer (50 mmol/liter Tris/HCl, 1% Nonidet P-40, 0.25% sodium deoxycholate, 150 mmol/liter NaCl, and 1 mmol/l EDTA, pH 7.4) and preincubated 1 h at 4°C with 50 l of protein A-Sepharose CL4B beads (50% w/v suspension in phosphate-buffered saline) (Amersham Biosciences). After centrifugation at 5,000 rpm for 5 min, 95 l of the supernatant were added to 5 l of human sera (dilution 1:20 in RIPA buffer) and first incubated 12 h at 4°C in an orbital shaker then 2 h at 4°C in the presence of protein A-Sepharose CL4B beads (50 l of 50% w/v suspension in phosphate-buffered saline). As a positive control, a polyclonal rabbit anti-CYP2E1 (dilution 1:100 in RIPA buffer) was used. Immunocomplexes bound to protein A-Sepharose were recovered by centrifugation, washed three times with 1.5 ml of phosphate-buffered saline, and solubilized in 40 l of SDS buffer, pH 6.8 (4% w/v sodium dodecyl sulfate, 0.2 mol of Tris/HCl, 26% v/v glycerol). One aliquot (15 l) was added to scintillation fluid and used for radioactivity determination. The other was boiled 5 min, centrifuged, and used for SDS-PAGE electrophoresis. The recovery of radioactive CYP2E1 was calculated as (dpm sample Ϫ dpm background)/(dpm positive control Ϫ dpm background) ϫ 1,000, where dpm background was the radioactivity recovered in the absence of added serum. SDS-PAGE electrophoresis was performed for 45 min at 200 V using 4% stacking and 10% resolving gels. The proteins were transferred to Hybond-C Extra nitrocellulose gel (Amersham Biosciences) and exposed to autoradiography using x-ray films (Eastman Kodak Co.).
CYP2E1 Computer Simulation-A structural model of human CYP2E1 was generated by the first approach mode using the SWISS-MODEL automated comparative protein modeling server (www. expasy.ch/swissmod/SWISS-MODEL.html) (24) and the template coordinates of CYP2C5 (code ExPDB 1DT6A) (25). Energy minimization of the model was performed using GROMOS96 force field algorithm. Root mean square errors and atomic distances were calculated using Deep-View-SwissPdbViewer software (www.expasy.org/spdbv/) (24). Graphical representation was made by the use of Rastop (version 2.0.3) by Philippe Valadon (La Jolla, CA).
Site-directed Mutagenesis of CYP2E1-The CYP2E1 was cloned into the expression plasmid pCWori ϩ (26) between the restriction sites NdeI and HindIII. Additional bases encoding 6 C-terminal histidines were added, and nucleotides encoding the first 18 amino acids were removed to optimize the expression (27). The mutant CYP2E1s were generated by using the QuikChange™ XL site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's instructions. The forward and reverse DNA primers used for each mutation are listed in Table I. XL-1 Blue supercompetent Escherichia coli strain was transfected with the different plasmids by heat shock and selected on LB agar plates containing ampicillin. Single colonies were further expanded by overnight culture at 37°C in 5 ml of LB medium plus ampicillin (50 g/ml). Plasmid DNA was isolated using QIAprep spin miniprep columns (Qiagen Inc., Valencia CA), and plasmid concentration was evaluated spectrophotometrically. The correct sequence of the insert was confirmed by automated DNA sequencing with the ABI PRISM dye terminator cycle sequencing kit (Applied Biosystems, Foster City, CA).
Prokaryote Expression of Mutated and Wild-type CYP2E1-For the expression of wild-type and mutated CYP2E1 variants, 70 l of XL1 Blue supercompetent E. coli cells were transfected with 1 l of the respective plasmid solution as described above, and colonies were selected on LB agar plates with ampicillin. Single colonies were further grown for 16 h at 37°C and 10 h at 4°C on LB medium ampicillin. Two ml of these cultures were used to seed 198 ml of LB agar ampicillin medium supplemented with 1 mmol/liter thiamine, 0.5 mmol/liter ␦-aminolevulinic acid, 25 mol/liter FeCl 3 and incubated at 30°C under continuous shaking (25). Bacterial growth was monitored spectrophotometrically until reaching 0.7 OD units before adding 0.5 mol/liter imidazole and 1 mmol/liter isopropyl ␤-D-thiogalactoside. After a further 24 h of culture at 30°C, CYP expression was monitored in whole cells by recording absorption spectrum at 450 nm in the presence of reducing agent and CO (28).
Reaction of Human Sera with Mutated CYP2E1s-The recognition of mutated and wild-type CYP2E1 by the human sera was estimated in immunoprecipitation experiments using 10 pmol of CYP2E1 solubilized in RIPA buffer and 5 l of the different sera (1:20 dilution in RIPA buffer) as described above. Immunocomplexes bound to protein A-Sepharose beads were solubilized in 40 l of SDS buffer, pH 6.8 (4% w/v sodium dodecyl sulfate, 0.2 mol/liter Tris/HCl, 26% v/v glycerol) and used for SDS-PAGE electrophoresis (45 min at 200 V using 4% stacking and 10% resolving gels). The proteins were then transferred to Hybond-C Extra nitrocellulose gel (Amersham Biosciences), and the membranes were probed with a monoclonal mouse IgG toward the His 6 tag tail (Amersham Biosciences). The antibody binding was revealed by horseradish peroxidase-conjugated anti-mouse immunoglobulins (Bio-Rad) using Western Lightning chemiluminescence reagent Plus (PerkinElmer Life Sciences) and x-ray film (Kodak). The intensities of the bands obtained with the differently mutated CYP2E1s were measured by videodensitometry and expressed as the percentage of the intensities of the bands with wild-type CYP2E1.

RESULTS
Initial experiments confirmed that the sera from patients with halothane-induced hepatitis (HAL) and ALD recognizing recombinant human CYP2E1 in enzyme-linked immunosorbent assays immunoprecipitated [ 35 S]methioninelabeled CYP2E1 translated in vitro using the rabbit reticulocyte system. To discriminate between conformational and linear epitopes, we prepared CYP2E1 constructs with the N-terminal (222 amino acids) or C-terminal (271 amino acids) moieties. However, the reactivity of these human sera against these two truncated forms were undetectable (not shown), indicating that the anti-CYP2E1 autoantibodies in both HAL and ALD are directed toward conformational epitopes. This conclusion was strengthened by the observation that the human sera failed to recognize CYP2E1 in Western blots (not shown).
To characterize these conformational epitopes, we decided to study the effects of single amino acid substitutions on the antigenic capacity of the whole molecule. A computer-simulated structure of CYP2E1 was generated using the SWISS-MODEL automated comparative protein modeling server (www.expasy.ch/swissmod/SWISS-MODEL.html) (24) and the crystal structure of rabbit CYP2C5, which share 59% amino acid sequence homology with CYP2E1 (29). The three-dimensional model of CYP2E1 obtained was quite similar to the structure of CYP2C5 (root mean square error 0.017 nm) with the exception of a loop between Lys-277 and Tyr-285 and a helix between Pro-213 and Pro-222 (Fig. 1). This latter region was replaced by the corresponding sequence from CYP2C3 to get CYP2C5 suitable for crystallization (25). Our CYP2E1 simulation was also similar (root mean square errors ranging from 0.170 and 0.185 nm) with other CYP2E1 models based on The CYP2E1 containing these mutations showed adsorption spectra comparable to wild-type CYP2E1 and were used in further experiments.
CYP2C5 and bacterial CYPs (joneslab.wsu.edu). Using the CYP2E1 structural model, sequences containing charged amino acids facing the outer surface of the protein were selected in the ␣ helices and ␤ sheets located on the external portions of the molecule. We focused our attention to protruding areas suitable for antibody binding, whereas sequences that were in troughs or not apparently accessible were excluded. Since most of the epitopes so far identified in CYPs largely involve the C-terminal portion of the molecule (15)(16)(17)(18)(19)(20)(21), the sequence between amino acids 200 and 493 was investigated. Preliminary experiments using purified rat or rabbit CYP2E1 showed that human anti-CYP2E1 IgG cross-reacted with the enzyme from these species in enzyme-linked immunosorbent assays (not shown). Therefore, amino acid sequences of human CYP2E1 not in common with the rat and rabbit CYP2E1 orthologues were excluded. The potential antigenic capability of the selected sequences was then confirmed by computer assessment of the antigenic index according to Jameson and Wolf (30) (GCG Wisconsin package; Accelrys Inc.). To produce major changes in the configuration of the possible epitopes without disrupting the tertiary structure of the molecule, we decided to insert a neutral amino acid having a low steric hindrance such as alanine in place of charged residues of lysine, arginine, and glutamic acid. The effects of these mutations on CYP2E1 antigenic potential was evaluated using the GCG Wisconsin software. As a result, 20 single amino acid substitutions were selected (Table I).
Site-directed mutagenesis was used to introduce the mutations into cDNA, and the modified CYP2E1s were then ex-pressed in E. coli as His 6 -tagged proteins. The capacity to retain the heme iron on the reduced state, as measured by the spectral absorption at 450 nm of the reduced CYPs (28), was used as an index of the correct protein folding. Fourteen out of 20 mutated CYP2E1s (Table I) showing spectral features comparable with the wild-type protein were used for further experiments. The capacity of the human sera to recognize these CYP2E1 variants was assayed by immunoprecipitation followed by Western blotting. As shown in Fig. 2, about half of the substitutions impaired significantly the ability of the sera to immunoprecipitate CYP2E1. Alanine substitutions of Glu-272, Lys-342, and Lys-420 affected the antibody binding to the highest extent. The effects of Lys-324, Arg-374, Phe-421, and Glu-440 substitutions were less consistent, being evident with 2-3 of the different sera (Fig. 2). Conversely, the Ala substitution of Lys-234, Glu-320, Arg-331, Arg-344, and Glu-346 did not interfere with the antibody recognition or had scattered effects with a single serum (Fig. 2). Although the mutations were able to reduce CYP2E1 antigenicity involved amino acids located far away from each other in the JЈ-helix and the ␤-sheet between K and L helices, respectively, computer simulations revealed that the positions of Lys-342, and Lys-420 on the tertiary structure were rather close (about 2.5-3.0 nm) and identified an area in CYP2E1 surface compatible with the presence of a distinct conformational epitope (Fig. 3). This interpretation was supported by the observation that the substitutions of Lys-324 and Phe-421 (Figs. 2 and 3) also decreased the antigen recognition of some sera. , and the immunocomplexes recovered by protein A-Sepharose beads were estimated by Western blotting using a monoclonal mouse IgG targeting the His 6 tag tail of the CYP2E1s. The intensities of the bands were measured by videodensitometry, and the results were expressed as the percentage of the recovery of modified CYP2E1s as compared with the wild-type protein included as a reference in each blot. The black squares indicate a reduction in the antibody recognition Ն 50%, and the gray squares indicate a reduction between 40 and 50%.
The poor immunoreactivity of three HAL and one ALD sera with CYP2E1 modified by the substitution of Glu-272 in the ␤ sheets between the G and H helices (Fig. 2) prompted us to extend the investigation to include substitutions encompassing the G helix between Lys-243 and Lys-251. Fig. 4 shows that the substitution of Lys-243, Glu-244 did not significantly affect CYP2E1 recognition. However, a further change of Glu-248 and Lys-251 for Ala appreciably reduced the capacity of three out of five of HAL and in four out of five of the ALD sera to immunoprecipitate CYP2E1 (Fig. 4). Structural simulation revealed that the mutated amino acids were closely associated with a distinct area on the enzyme surface (Fig. 3).
These results indicated that anti-CYP2E1 autoantibodies associated to both halothane hepatitis and alcoholic liver disease recognized similar conformational epitopes and that two of these epitopes were located in, respectively, the G helix and between JЈ and KЉ helices in the C-terminal portion of the molecule. This conclusion was supported by further analysis of additional five HAL and seven ALD sera revealing that these two epitopes were recognized by four out of the five HAL sera and five out of the seven ALD sera (Table II). Altogether, the epitopes identified account for the antigen specificity of 19 out of all 22 (86%) human sera investigated. DISCUSSION In recent years, several investigators have evaluated the epitope specificity of LKM-1 autoantibodies on CYP2D6 by comparing the cross reactivity of human sera with prokaryotic expressed peptides spanning the entire protein sequence. This approach has allowed the identification of several linear    helices on the antigen recognition by additional sera from patients with HAL or ALD Ten pmol of wild-type or mutated CYP2E1s were reacted with 5 l of the different human sera (1:20 dilution), and the immunocomplexes recovered by protein A-Sepharose beads were estimated by western blotting using a monoclonal mouse IgG toward the His 6 tag tail on the CYP2E1s. The intensities of the bands were measured by videodensitometry, and the results were expressed as percentage of the recovery of modified CYP2E1 as compared with the wild-type protein included as reference in each blot. The black squares indicate a reduction in the antibody recognition of CYP2E1 Ն to 50%, and the grey squares indicate a reduction between 40 and 50%. epitopes with different specificity for LKM-1 associated with type II autoimmune hepatitis and chronic hepatitis C (17)(18)(19)(20)(21). Furthermore, it has been shown that some of these epitopes share structural homologies with protein sequences present in type 1 herpes simplex virus and hepatitis C virus (17,21). The use of deleted CYP mutants has also provided important information concerning the structure of different epitopes present in CYP2C9, CYP3A1, and CYP1A2 (15)(16)(17). However, it is well known that B-cell epitopes associated to autoimmune disease are often conformational. For instance, conformational epitopes in CYP2D6 and CYP2C9 are implicated in the autoreactivity of, respectively, LKM-1 and LKM-2 antibodies (15,31). According to Bourdi et al. (2), the sera of patients with either halothane hepatitis or alcoholic liver disease does not recognize denaturated human CYP2E1. The same sera also have low reactivity toward truncated forms of CYP2E1, not allowing the use of deletion mutants for the identification of the epitopes. We have, therefore, devised a new approach based on the analysis of the whole molecule antigenic capacity following single amino acid substitutions that modify the charge distribution on the CYP2E1 surface without affecting the molecule conformation. Hereby, the data obtained demonstrate that Ala substitutions of Lys-324, Lys-342, Lys-420, and Phe-421 identify a conformational epitope formed by the juxtaposition of the JЈ and KЉ helices. The distance between the single amino acids confirms the compatibility with an antigen-antibody binding area. The J, K, and L helices have been implicated as sites of conformational epitopes recognized in CYP2C9 by LKM2 autoantibodies associated with tienilic acid-induced autoimmune hepatitis (15), whereas the K helix of CYP3A4 is considered a major epitope recognized by autoantibodies of patients with hypersensitivity reactions to aromatic anti-convulsing drugs (16). It is interesting to note that both Lys-324 and Lys-342 are comprised in a sequence with good homology to the CYP2D6 321-351 epitope (19), whereas Lys-324 is conserved in the CYP2C9 314 -322 epitope (15). Lys-324 is also close to the CYP2D6 316 -327 sequence that, according to Ma et al. (20), represents a key target for autoantibodies on CYP2D6 surface. Similarly, the CYP2D6 410 -429 epitope (19) has good homology with the sequence containing Lys-420 and Phe-421. This suggests the possibility that the J, K, and L helices might represent an important antigenic area in CYPs, able to give rise to antibodies against both linear and conformational epitopes.
A further conformational epitope corresponding to Glu-248 and Lys-251 is also evident in a distinct area of the G helix that is located on the opposite side of the CYP2E1 surface. This epitope is recognized by 50% of the 22 sera tested, five of them in combination with the epitope in JЈ-KЉ helices. The C-terminal portion of CYP3A4 spanning up to Thr-208 and the region between Thr-208 and Ser-281, which comprises the G helix, is the target of autoantibodies present in a subset of alcoholics (5). However, it is unlikely that anti-CYP3A4 autoantibodies might account for the recognition of CYP2E1 conformational epitopes because they are directed against linear structures, and there is no sequence homology between G helices of CYP2E1 and CYP3A4. We have recently reported (32) that the N-terminal portion of CYP2E1 and particularly the amphipathic amino acids in the B-helix are responsible for electrostatic interactions with negatively charged phospholipids that anchor CYP2E1 to the cell membranes. The orientation of CYP2E1 in relation to the membrane (Fig. 5) shows that the epitopes formed by JЈ and KЉ helices and G helix are both on the outer portion of the molecule and well accessible to antibody binding. This is consistent with previous observations showing that functionally active CYP2E1 is present on the outer layer of hepatocyte plasma membrane, where it is targeted by specific antibodies (3,33,34). Thus, the recognition of these conformational epitopes on the portion of CYP2E1 that faces the extracellular spaces can be involved in triggering antibody-mediated hepatocyte killing.
In conclusion, the results presented demonstrate that anti-CYP2E1 autoantibodies recognize at least two distinct conformational epitopes present in the G-helix and in an area formed by the juxtaposition of JЈ and KЉ helices on the C-terminal portion of the molecule surface. Moreover, we propose the combined use of molecular modeling and site-directed mutagenesis as a novel method to investigate the epitope specificity of conformational anti-CYP autoantibodies associated to liver diseases. FIG. 5. Localization of the two conformational epitopes identified in the G and J-K؆ helices, respectively, in relation to CYP2E1 orientation with respect to the cell membranes proposed by Neve et al. (32).