Potential conflict of interest: Nothing to report.
Anti-liver-kidney microsome type 1 (LKM1) autoantibodies directed against the cytochrome P450 2D6 (CYP2D6) are considered specific markers of type 2 autoimmune hepatitis, but are also found in 5% of sera from patients chronically infected by hepatitis C virus (HCV). Molecular mimicry between HCV proteins and CYP2D6 has been proposed to explain the emergence of these autoantibodies. Anti-LKM1 autoantibodies from hepatitis C–infected patients were affinity-purified against immobilized CYP2D6 protein and used to screen a phage display library. CYP2D6 conformational epitopes were identified using phage display analysis and the identification of statistically significant pairs (SSPs). Cross-reactivity between CYP2D6 and HCV protein candidates was tested by immunoprecipitation. Nineteen different clones were isolated, and their sequencing resulted in the mapping of a conformational epitope to the region of amino acids 254-288 of CYP2D6. Candidate HCV proteins for molecular mimicry included: core, E2, NS3 and NS5a. Affinity-purified autoantibodies from HCV+/LKM1+ patients immunoprecipitated either NS3, NS5a, or both, and these reactivities were specifically inhibited by immobilized CYP2D6. In conclusion, HCV+/LKM1+ sera recognize a specific conformational epitope on CYP2D6 between amino acids 254 to 288, the region that contains the major linear epitope in type 2 autoimmune hepatitis patients. Cross-reactivity due to molecular mimicry at the B-cell level was shown between the CYP2D6 and the HCV NS3 and NS5a proteins and could explain the presence of anti-LKM1 in patients chronically infected with HCV. Further investigation of the role played by this molecular mimicry in HCV-infected patients may lead to more specific strategies for diagnosis and treatment. (HEPATOLOGY 2005.)
The number of people infected by hepatitis C virus (HCV) worldwide is estimated at 170 million, with a global prevalence of 3%.1, 2 Several autoimmune phenomena have been reported to be frequently associated with chronic HCV infection, namely, rheumatoid symptoms, keratoconjunctivitis sicca, lichen planus, glomerulonephritis, mixed cryoglobulinemia, and circulating autoantibodies.3–7 When serious liver injury occurs in chronic hepatitis C infection, an autoimmune mechanism may contribute to hepatic destruction.8
The presence of Liver Kidney Microsomes type 1 (LKM1) autoantibodies in chronic HCV-infected patients is puzzling. Although anti-LKM1 autoantibodies are considered to be a specific serological marker of type 2 autoimmune hepatitis (AIH-2), 5% to 10% of patients with chronic HCV hepatitis are also LKM1 positive.9–11 LKM1 autoantibodies are known to be specifically directed against cytochrome P450IID6 (CYP2D6),12, 13 a protein located on the cytoplasmic side of the endoplasmic reticulum of hepatocytes. Circulating autoantibodies in sera from patients with AIH-2 recognize linear epitopes on the CYP2D6 protein, whereas most HCV+/LKM1+ sera (70%) are only directed against conformational epitopes of this autoantigen.14, 15 Recently, a new method that combines phage display analysis and a bioinformatics analysis to identify statistically significant pairs (SSPs)16 has eased the study and identification of conformational epitopes. This method has been used to precisely map a B-cell conformational epitope in human immunodeficiency virus (HIV)16 and has opened the way to properly study these discontinuous epitopes, which composed the vast majority of recognized sequences.17
The origin of LKM1 autoantibodies in both HCV infection and AIH is unknown. Their presence cannot be solely attributed to the release of sequestered hepatocyte proteins, because anti-LKM1 autoantibodies are completely absent in sera from patients with other liver diseases.18 Specific major histocompatibility complex loci have been described in association with the presence of LKM1 autoantibodies in both AIH-2 patients19 and HCV chronically infected patients.20 Molecular mimicry between HCV proteins and CYP2D6 could explain the emergence of these autoantibodies. The hepatitis C virus is made of single-stranded RNA that is translated into one polyprotein that is cleaved by cellular and viral proteases into 10 structural (core, E1, E2) and nonstructural (P7, NS2, NS3, NS4a, NS4b, NS5a, NS5b) proteins.21 The aim of our study was to characterize the reactivity of LKM1+ autoantibodies in HCV chronically infected patients and test for possible molecular mimicry between HCV proteins and CYP2D6, in the hope of shedding light on an otherwise poorly understood process by which LKM1 autoantibodies are produced in both HCV-infected and autoimmune hepatitis patients.
Sera were obtained from patients (2 females and 1 male, average age: 63 ± 12 years) chronically infected by HCV 1a who were shown to be LKM1+ by indirect immunofluorescence using rat or human liver and kidney sections and by western blot against human liver microsomal fractions as described.18, 22 HCV infection was diagnosed by second-generation recombinant immunoblot assay or polymerase chain reaction. Sera from patients with type 2 autoimmune hepatitis (LKM1+), but not infected by HCV, and pooled sera from healthy individuals (LKM1−) and HCV-infected (LKM1−) patients were used as control. This study conforms to current ethical guidelines and was approved by the local ethical committee.
Human CYP2D6 and HCV Constructions.
The complete CYP2D6 cDNA (GenBank accession number: AF009606), obtained from Dr U. Meyer (Basel, Switzerland), was sub-cloned into pGem3 (Pharmacia Biotech, Uppsala, Sweden) and pMalcrI (New England Biolabs, Beverly, MA) vectors at their EcoRI sites. The HCV polyprotein cDNA (provided by Dr D. Moradpour, Freiburg, Germany) was used to amplify the different specific proteins (E2, Core, NS3, and NS5a). The polymerase chain reaction product of each HCV protein was cloned into the pCRII-TOPO TA cloning vector (Invitrogen, Carlsbad, CA) and used to chemically transfect TOP10 competent cells (Invitrogen).
Affinity Purification of LKM1+/HCV+ Sera.
The MBP/CYP2D6 fusion product was produced in vitro (pMAL Protein Fusion and Purification System; New England Biolabs). The protein obtained was coupled to a resin (NHS-activated Sepharose 4 Fast Flow; Amersham Life Sciences, Oakville, Canada) and free sites properly blocked by an ethanolamine solution. HCV+/LKM1+ patients' sera were diluted 1:10 in PBS (1 mL final volume) then loaded 3 times on 120 μL of the prepared affinity column. The column was then washed and the autoantibodies eluted with an acid solution (glycine-HCl, pH 2.5), washed, and eluted again with a basic solution (Triethylamine pH 11.5). The eluted fractions were concentrated to 900 μL by centrifugation in a Microcon YM-30 filter column (Millipore, Billerica, MA).
Western Blot Analysis.
Proteins of the human liver microsomal fraction22 were separated by electrophoresis on a 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred onto nitrocellulose filters (Amersham). The membrane was blocked with 5% powder milk in TBST (10 mmol/L Tris-HCl, pH 7.4, 150 mmol/L NaCl, 0.05% Tween 20). The nitrocellulose filters were then incubated for 2 hours at room temperature with the purified HCV+/LKM1+ sera and control sera in TBST/1% powder milk. The secondary antibody used was conjugated to peroxidase (Biosource International, Camarillo, CA), diluted to 1:10,000 in TBST/1% powder milk. Bound peroxidase was detected with chemiluminescence blotting substrate (Boehringer Mannheim, Germany), according to the manufacturer's instructions.
Immunoprecipitation of In Vitro Translated Proteins From CYP2D6 and HCV DNA Constructions.
The proteins coded by the TOPO vectors were individually synthesized in vitro (TnT Coupled Reticulocyte Systems; Promega, Madison, WI) and labeled with 35S-methionine (Amersham Life Sciences). Sera (5 μL) were individually incubated with 50,000 cpm of these marked proteins, 2 mg protein A-Sepharose (Sigma-Aldrich Canada, Oakville, Canada) and 500 μL of a proper incubating buffer with agitation for 2 hours at 4°C.23 The protein A-associated immunoglobulins were washed 6 times.23 The remaining pellet was then denatured in a loading buffer followed by 5 minutes' boiling before loading on the denaturing SDS-PAGE. After electrophoresis, the gel was fixed, Coomassie colored, washed, and dried before being exposed overnight against an autoradiography film (BiomaxMR; Kodak, Rochester, NY).
Mimotope Identification Through Phage Display.
The purified and concentrated autoantibodies were used as the selecting matrix with a phage display library (Ph.D.-7 Phage Display Peptide Library Kit; New England Biolabs) with 150 μL for every selection cycle. The retained phages, eluted with CYP2D6 fusion protein (20 μg) and purified, were then sequenced (Sequenase Version 2.0 DNA Sequencing Kit, USB, Cleveland, OH). Nineteen different phages selected by the 3 purified sera after the third cycle were pooled at equal titers and used as the peptide bank for a fourth cycle with each of the purified sera at two temperature conditions, 4°C and room temperature (20°C). The phages from this cycle were also purified and sequenced. The sequences obtained were then compared between themselves and to the CYP2D6 (GeneBank acc.num.: P10635) and HCV-1a (GeneBank acc. num.:AAA45676) sequences using the Vector NTI program (InforMax, Carlsbad, CA).
CYP2D6 Conformational Epitope Analysis.
The 19 mimotopes obtained by phage display were used to find SSPs for the identification of conformational epitopes on the CYP2D6 protein. According to the process described by Enshell-Seijffers et al.,16 mimotope sequences were simplified by using a degenerate amino acid code that groups amino acids with similar antibody-binding characteristics (R,K = B; E,D = J;S,T-O; L,V,I = U;Q,N = X; W,F = Z). The amino acid pairs were then ranked by their occurrence in the different mimotopes. The theoretical occurrences (based on randomness) of the different pairs was then calculated, and the over-represented amino acid pairs were identified as SSPs. The crystallographic model of the CYP2D624 (provided by Dr Lewis, University of Surrey, UK) was used to search for tandem and bridged SSPs on the protein surface. Amino acids of a pair had to be within 8Å of each other. The SSPs then formed clusters on the surface of the protein (a cluster is defined as a pair that is within 8Å of a pair which is within 8Å of a second pair, etc.). The region that contains a cluster of SSPs on the surface of a protein was considered a putative epitope.
Immunoabsorption of Purified HCV+/LKM1+ Sera.
To confirm the results obtained by immunoprecipitation, 50 μL purified HCV+/LKM1+ sera was adsorbed on either 100 μL control resin (GST-Sepharose) or 100 μL CYP2D6-sepharose. The unabsorbed sera were then used to immunoprecipitate (as described above) CYP2D6 (control), NS3, or NS5a proteins.
Protein Structural Analysis.
The protein structure was analyzed with the Swiss PDB Viewer 3.7 program,25 and the final images were rendered with POVRAY version 3.5.
Analysis of HCV+/LKM1+ Patients Sera Purification.
The sera of the 3 HCV+ patients had been previously identified as LKM1+ by immunofluorescence and western blot. These results were confirmed by immunoprecipitation with radiolabeled CYP2D6, thereby indicating that these sera recognize both native and denatured CYP2D6. The different fractions obtained after the affinity purification of anti-LKM1 autoantibodies using a CYP2D6-sepharose column are shown in Fig. 1. Affinity-purified LKM1 antibodies were mainly reactive against the native form of CYP2D6 by immunoprecipitation compared with western blot (Fig. 1A-B), thereby suggesting that the purification against native CYP2D6 led to the selection of antibodies recognizing conformational epitopes of the molecule. The antibodies directed against linear epitopes were lost during the purification, possibly because linear epitopes can be hidden in a correctly folded CYP2D6 protein. The purified autoantibodies were of the immunoglobulin G isotype (data not shown).
Identification of a Possible Molecular Mimicry Between Mimotopes and HCV Proteins.
Nineteen mimotopes were identified using the purified LKM1+ sera and the phage display technique. Based on the presumption that a conformational epitope could have a partial homology with the antigen amino acids sequence, the mimotope sequences were compared with HCV protein sequences (Fig. 2). Four proteins (core, E2, NS3, and NS5a) were found as having homologies with several mimotopes.
Identification of LKM1 Conformational Epitopes on the CYP2D6.
The phage display technique allows the detection of both linear and conformational epitopes.26, 27 The mimotope sequences were analyzed using a novel approach16 that aims at mapping conformational epitopes through the identification of SSPs of amino acids. Using this approach, 3 SSPs were identified (Fig. 3): OO, OU, JO. These sequences use a degenerate code that groups amino acids with similar side chains according to their antibody-binding characteristics (Fig. 3B). The amino acid pairs were ranked by their occurrence in the different mimotopes (Fig. 3C) and compared with their theoretical occurrences (based on randomness) (Fig. 3D). The over-represented pairs were identified as statistically significant pairs (SSPs). These 3 SSPs were then mapped on the CYP2D6 crystallographic model as tandem and bridged amino acids (within 8 Å of each other). These identified pairs (OO, OU, JO) formed a unique cluster in the region of amino acids 254 to 288 (Fig. 4A). This cluster is the only one that encompasses all 3 amino acids pairs within 8 Å of one another. These amino acids are on the surface of the protein and span a region of 13 Å by 28 Å (364 Å2) (Fig. 4B). The region of amino acids 254 to 288 is contained within a homogenous region of negative electrostatic potential (Fig. 4C). These facts indicate that the region of amino acids 254 to 288 of the CYP2D6 contains a putative conformational epitope mimicked by the mimotopes identified with the affinity-purified HCV+/LKM1+ sera.
Molecular Mimicry and Cross-reactivity Between HCV Proteins and CYP2D6.
The mimotopes isolated with the affinity-purified HCV+/LKM1+ sera showed sequence homologies with 4 HCV proteins (core, E2, NS3, and NS5a). These homologies were used as a starting point to look for possible molecular mimicry and cross-reactivity between HCV proteins and the CYP2D6. The four candidate proteins were synthesized in vitro and used in an immunoprecipitation assay. The HCV+/LKM1+ sera purified against the CYP2D6-sepharose (P1, P2 and P3) were used to immunoprecipitate the four previously identified HCV proteins (Fig. 5A-B). If these sera purified for CYP2D6 binding antibodies could immunoprecipitate an HCV protein, this would indicate that the same antibody, present in HCV+/LKM1+ sera, binds to both CYP2D6 and to an HCV protein. As is shown in Fig. 5A-B, the purified sera from patient 1 immunoprecipitated the NS3 protein, patient 2 immunoprecipitated NS5a, and patient 3 immunoprecipitated both NS3 and NS5a. Sera from type 2 AIH patients with LKM1 autoantibodies did not immunoprecipitate either NS3 or NS5a (data not shown). This indicates that a molecular mimicry and cross-reactivity at the B cell level exists between the CYP2D6 and the NS3 and NS5a proteins from HCV. To confirm that these HCV proteins (NS3 and NS5a) were, in fact, immunoprecipitated by the same antibody, sera were pre-absorbed on either a control resin (GST-Sepharose) or the CYP2D6-sepharose. The pre-absorption with the GST-Sepharose had no effect on the immunoprecipitation of the CYP2D6 or the NS3 and NS5a proteins, whereas the pre-absorption with CYP2D6 removed the antibodies necessary to immunoprecipitate CYP2D6, NS3, or NS5a (Fig. 5C). Altogether, these results confirm that there is in fact molecular mimicry between CYP2D6 and NS3 or NS5a and that cross-reactivity between antibodies recognizing conformational epitopes on these molecules exists.
Anti-LKM1 antibodies were first described in patients with chronic active hepatitis.28 The advent of tests for the diagnosis of HCV infection showed that some LKM1(+) chronic active hepatitis patients were infected.29 Later, several studies confirmed that LKM1 autoantibodies are confined to AIH- and HCV-infected patients.18, 30 However, the factors triggering and maintaining this particular and specific B-cell autoimmune response are unknown. A molecular mimicry between HCV antigens and CYP2D6 (the autoantigen recognized by anti-LKM1 antibodies) has been frequently stated as a possible mechanism.31, 32 Here we report, through experimental data, that a molecular mimicry exists between HCV proteins and CYP2D6. This cross-reactivity could explain the presence of anti-LKM1 antibodies in HCV-infected patients.
LKM1 autoantibodies present in chronically infected HCV patients are mainly (70%) directed against the native form of CYP2D6.14, 15 Efforts have been made to identify CYP2D6 epitopes recognized by HCV-infected patients, both linear33 and conformational.34 The study by Sugimura et al.34 identified a conformational epitope (aa 321-379 of CYP2D6) recognized by type 2 AIH and HCV-infected patients through western blot analysis. The possibility that this epitope could also be a discontinuous linear epitope cannot be excluded because these type of autoepitopes have been previously described in type 2 AIH patients.35 The identification of conformational epitopes is a difficult endeavor. This is largely because of the difficulty of producing peptides that adopt their native structure, as found in the native protein. A new method has been recently developed that enables the identification of conformational epitopes through the combined use of phage display analysis36 and the identification of SSPs.16 To ensure that the identified mimotopes were CYP2D6-specific epitopes, the HCV+/LKM1+ sera were affinity-purified against a CYP2D6-sepharose. The purified antibodies were reactive only against the native CYP2D6, indicating that, using this method, only antibodies against conformational epitopes were purified.
Even though a limited number of patients were used, the use of affinity-purified sera and the combined approach of phage display analysis and SSP identification allowed the mapping of a conformational epitope in the region of aa 254-288 of CYP2D6. This epitope comprises an area of 364 Å2 and is within a homogenous field of negative electrostatic potential; these characteristics are consistent with those of a conformational epitope. This region of CYP2D6 also contains the major epitope recognized by sera from patients with AIH-2, the aa 257-269.35 This area of high antigenicity35 of the CYP2D6 seems to elicit distinct immune reactivities in HCV and AIH patients; type 2 AIH patients appear to develop autoantibodies against a linear epitope whereas the HCV patients will do the same against a conformational epitope. This infers that different immune processes are at work (different antigen presentation by antigen-presenting cells, activation of different autoreactive clones) and that these mechanisms are independent of CYP2D6 antigenic properties.
The sequences obtained through phage display were used to identify candidates for a molecular mimicry between an HCV protein and the CYP2D6. Four HCV proteins have been identified as candidates; core, E2, NS3, and NS5a. By immunoprecipitation with purified sera, and the subsequent confirmation by absorption with a CYP2D6 adsorbing resin, molecular mimicry at the B-cell level between the CYP2D6 and NS3 and NS5a was revealed. These results indicate that the same conformational antibodies that recognize CYP2D6 also recognize NS3, NS3, and NS5a or NS5a only. This description of a conformational molecular mimicry between HCV proteins and the CYP2D6 could explain the presence of LKM1 autoantibodies in patients chronically infected with HCV. The putative regions of NS3 and NS5a that cross-react with CYP2D6 (Fig. 2) are highly conserved in HCV genotypes 1a, 1b, 2, 3, 4, 5, and 6 (data not shown).
To further define the molecular mimicry between CYP2D6 and the HCV proteins (NS3, NS5a), the next logical step would be the structural analysis of these molecules and identification of similar structural motifs. Human cytochrome P450 2D6 has not been crystallized, so only a model of its structure based on a mammalian cytochrome has been produced.24 The crystallographic structure of the NS5a protein has not been elucidated yet, except for 30 amino acids of the membrane anchor domain.37 The tertiary structure of the NS3 protease/helicase has been published.38 The analysis of common structural motifs between these 3 proteins is, therefore, very difficult. Using available methods of structure comparison (VAST algorithm,39 Combinatorial Extension method,40 linear sequence alignments [PAM 10] followed by structural superimpositions, etc.), no common structural domain between the modeled CYP2D6 and the NS3 or with the membrane anchor domain of NS5a were found. The crystallography of NS5a and CYP2D6 should enable a more thorough investigation of common structural motifs between these proteins and enable a better representation of the molecular mimicry identified in this study.
The clinical relevance of these findings cannot be understated. Given the conformational nature of these LKM1 epitopes in HCV-infected patients, testing lacks specificity. Identification of the major structural epitopes recognized by LKM1+ sera could lead to new diagnostic tools using arrays of peptide mimics in ELISA settings that would enable simpler, wider testing of HCV+ patients for LKM1 autoantibodies directed against conformational CYP2D6 epitopes. Currently, to develop a diagnostic test for recognition of conformational epitopes, two steps must be achieved: (1) the identification of a specific epitope, and (2) the development of conformational peptide mimicking complex structural conformations. The development of such peptides relies on elaborate computational processes, a science still in its infancy.41, 42 Testing larger numbers of HCV-infected patients for autoantibodies against the aa 254-288 conformational epitope of CYP2D6 would prove interesting, because their prevalence is unknown, and it could help define their clinical relevance. Both pediatric and adult patients infected with HCV and who have LKM1 autoantibodies have been described as having alanine aminotransferase (ALT) flare ups under treatment, which reportedly resolve on discontinuation of interferon treatment.43–47 This reaction could result from an underlying autoimmune process directed against hepatocytes in LKM1+ patients. Patients that are chronically infected with HCV and have LKM1+ autoantibodies therefore should be carefully monitored during treatment with interferon.
In conclusion, antibodies directed against the CYP2D6 present in HCV+/LKM1+ patient's sera can cross-react with HCV proteins (NS3, NS5a). This cross-reactivity could be the result of molecular mimicry at the B cell level, which through an initial reactivity against viral proteins would lead to reactivity against a self-protein. This specific cross-reactivity with HCV protein is compatible with the observation that only patients infected with HCV develop LKM1 autoantibodies. It implies that it is an HCV-specific effect and explains why a molecular mimicry with HCV proteins has been frequently proposed.48, 49 The conformational epitope targeted by the autoantibodies present in HCV+/LKM1+ patient's sera (aa 254-288) is located in the same region as the major linear epitope recognized by type 2 AIH patients. Further identification of the region targeted by LKM1 antibodies in HCV+ patients and elucidation of the mechanism involved in their production will further our understanding of the autoimmune phenomenon present in these patients. Ultimately, these studies may lead to new and more specific strategies for diagnosis and treatment.
The authors thank Professor Claude C. Roy, MD, for critical reading. Dr. Hugo Soudeyns is a Junior-II Scientist of le Fonds de la Recherche en Santé du Québec.