Diego Vergani, Immunology Group, Institute of Hepatology, University College London, Gower Street Campus, Harold Samuel House, 69–75 Chenies Mews, London WC1E 6HX, UK Tel: 44 207 679 6515. Fax: 44 207 380 0405 e-mail: D.Vergani@ucl.ac.uk
Abstract: The pathogenesis of autoimmune liver disease and autoimmunity associated with chronic viral hepatitis remains poorly understood. One of the major hurdles to a deeper understanding of these pathological processes is the absence of clearly defined inductive mechanisms, which, if identified and characterised, could guide clinical strategies for their prevention or allow therapeutic intervention. Molecular mimicry leading to crossreactive autoimmune responses has gained strong experimental support in the past decade. A fundamental premise of this hypothesis is the involvement of a mimicking environmental trigger. In view of the numerous viral and bacterial agents epidemiologically linked to autoimmune liver diseases, we and others have proposed molecular mimicry to be an important mechanism in these diseases. We also propose similar crossreactive mechanisms to operate in the generation of autoimmunity in viral hepatitis. This review focuses on molecular mimicry at the level of the B-cell, as few data on T-cell crossreactivity in liver disease are thus far available.
The immune system has evolved to discriminate between endogenous or ‘self’ tissues and exogenous and neoplastic components collectively termed ‘non-self’. A dominant selective pressure in this process has been the threat to self-integrity posed by infectious pathogens. The virtually infinite antigenic variety of pathogens that the immune system must respond to has resulted in an innate set of responses that allow the triggering of immunity to commonly encountered pathogens, and an adaptive system that adopts an anticipatory approach in responding to more antigenically exotic or complex challenges.
The central problem for the adaptive immune system is to generate T- and B-lymphocytes, its two most important components, that can specifically recognise a potentially infinite number of non-self-antigens, without any prior information as to their structure. This is achieved by randomly generating a large number of T- and B-cell specificities (via their respective antigen receptors – the T-cell receptor (TCR) and the antibody) that are then able to clonally expand and recruit effector mechanisms on recognition of their specific antigen. It is, however, becoming clear that even this system cannot cope with the extent of non-self-antigenic diversity, and in the past decade convincing evidence for crossreactivity as an inherent property of immune ontogeny has emerged. This has been studied primarily in the context of T-lymphocytes, where it is clear that altered peptide ligands (APL), peptides similar in structure to the peptide antigen which is initially encountered, are able to induce both stimulatory and inhibitory T-cell responses and, indeed, endogenous APLs operate in selecting the T-cell repertoire in the thymus (1). This implies that a single T-cell, rather than responding to a single antigen specificity, is able to crossreactively respond to a number of antigens, thus expanding the antigenic specificities of the immune system to a level that reflects the antigenic diversity of the external environment.
This inherent potential for crossreactivity, whilst allowing efficient responses to a vast array of pathogens also imbues the immune system with the potential to crossreact with self, leading to autoimmunity. This concept has been termed ‘molecular mimicry’, where immune responses to external pathogens become directed towards structurally similar self-components (2). Molecular mimicry has been demonstrated to be a dominant mechanism in the pathogenesis of autoimmune disease both in experimental models and in the human setting at the level of both T- and B-cells (3–6). We and others have described immunological crossreactivity between pathogen and self to operate in the aetiopathogenesis of autoimmune liver disease and autoimmunity associated with chronic viral hepatitis. We describe below the main concepts and emerging evidence for molecular mimicry in these disease processes.
Liver–kidney microsomal type 1 antibodies and molecular mimicry
Autoimmune hepatitis type 2 (AIH2) is characterised by severe mononuclear cell infiltration of the hepatic portal tracts and ensuing interface hepatitis, serum elevation of liver enzymes and early progression to cirrhosis. The immunological hallmark of AIH2 is liver–kidney microsomal type 1 (LKM1) antibody (7). The target of this autoantibody is cytochrome P4502D6, a member of the hepatic P450 enzyme family (8). Due to the very strong association between LKM1 antibodies and AIH2, and in view of a recent report demonstrating surface expression of CYP2D6 on hepatocytes (rendering them amenable to recognition by these antibodies), LKM1 antibodies are thought to be involved in the pathogenesis of AIH2 (9). Intriguingly, LKM1 antibodies are also found in up to 10% of patients with hepatitis C virus (HCV) infection, and appear to correlate with increased disease severity and adverse reactions to interferon (IFN) treatment (10). Moreover, a minority of LKM1+ AIH2 patients are also infected with HCV, although the significance of HCV infection in this setting is unclear (11). There are important clinical reasons for differentiating those with LKM1+ AIH2 and those with LKM1+ HCV infection, as immunosuppression, the first-line treatment for AIH2, may be harmful if administered to patients infected with HCV (12). Conversely, IFN may potentiate autoimmune disease if administered to patients with autoimmune hepatitis (13).
LKM1 antibodies recognise linear regions of CYP2D6 in a hierarchical manner in AIH2. The principal linear B-cell epitope, CYP2D6257–269, is recognised by 85% of patients, CYP2D6321–351 by 53% of patients, and two additional minor epitopes, CYP2D6373–389 and CYP2D6410–429, are recognised by 7% and 13%, respectively (14). In addition, we have identified CYP2D6193–212 as a major B-cell epitope recognised by 93% of patients with AIH2 and 50% of those with LKM1+ HCV infection (15). This is in a partial agreement with Klein et al. who described reactivity to CYP2D6196–218 in 68% of patients with AIH2 but in only 18% of the LKM1+ HCV-infected patients (16). We further demonstrated crossreactive antibody recognition of homologous regions of HCV (NS5B HCV2985–2990) and cytomegalovirus (CMV) (EXON CMV130–135) antigens in LKM+ HCV-infected patients recognising CYP2D6 (CYP2D6204–209) (Fig. 1). Others have also invoked crossreactive mechanisms to explain the aetiology of LKM1 antibodies. Manns et al. suggested that reactivity to the major epitope of CYP2D6 recognised by LKM1 antibodies may arise through a crossreactive response with HCV or herpes simplex virus (HSV), as the aa 310–324 of E1 HCV and aa 156–170 of IE175 HSV1 share sequence homology with the immunodominant region, aa 254–271, of CYP2D6 (17) (Fig. 1). Although it is attractive, there has been little experimental evidence to support this hypothesis. We have recently demonstrated highly specific simultaneous antibody reactivity to both CYP2D6254–271 and E1 HCV310–324 in LKM1+ HCV-infected patients (18). Competition studies suggest this ‘double-reactivity’ to be crossreactive (Bogdanos et al., manuscript in preparation).
As LKM1 antibodies appear to crossreact with homologous regions of CYP2D6, HCV and HSV, and CMV, a ‘multi-hit’ mechanism for the generation of these antibodies and possibly AIH2 may be envisaged. In this model multiple exposures to CMV or HSV, common viral pathogen, may establish permissive immunological conditions, by priming a crossreactive subset of T-cells, in the genetically predisposed host. Chronic infection with HCV may then provide the final impetus for the generation of LKM1 autoantibodies through crossreactive mechanisms. Depending on the degree of immunological priming, the degree of genetic susceptibility (particularly at the HLA locus and coding regions for ‘innate’ components of immunity), and antigenic dose of the infecting pathogens, a minority of individuals may progress to autoimmune disease.
Aspects of this model are not without precedent. In rheumatic carditis, multiple exposures to group A streptococcus lead to a spectrum of immunological responses based on molecular mimicry between pathogen and cardiac myosin and laminin (19). Multiple exposures to this pathogen in the susceptible host lead to progressive valvular damage mediated by crossreactive antibodies. Neither does the initial pathogenic insult need to be closely temporally related to the manifestation of autoimmune disease. This is well illustrated in Chagas’ disease caused by the Trypanosoma cruzi. Acute infection with T. cruzi causes a transient myocarditis which resolves completely. Decades later, an autoimmune myocarditis, histologically characterised by mononuclear cell infiltration, in the absence of detectable parasites, develops in the susceptible host, leading eventually to dilated cardiomyopathy. Molecular mimicry between the B13 antigen of T. cruzi and cardiac myocytes is thought to be responsible for the development of the autoimmune response (20). This system illustrates the ‘hit and run’ model for the development of autoimmunity following infection. These models taken together provide conceptual insights into mechanisms that may operate in leading to autoimmunity and autoimmune liver disease.
Multiple autoimmunity: immune crossreaction and bad AIRE
A feature of AIH2 is the frequent coexistence of other autoimmune diseases, in particular type 1 diabetes, Addison’s disease, hypoparathyroidism and autoimmune thyroiditis (21). We have demonstrated crossreactive LKM1 antibody responses between CYP2D6321–351, the second most frequently recognised epitope of CYP2D6, and structurally similar regions of carboxypeptidase H, an autoantigen in type 1 diabetes, and 21-hydroxylase, the major autoantigen in Addison’s disease (22) (Fig. 1). We propose that autoimmunity once induced against one self-antigen may spread via molecular mimicry to other homologous self-antigens, and in genetically predisposed individuals lead to overt autoimmune disease (Fig. 2). Although intermolecular epitope spreading is a feature of other autoimmune disease such as type 1 diabetes and connective tissue disorders, the mechanistic explanation for this phenomenon has relied on molecular association or tissue colocalisation for the diversification of autoimmunity. Our hypothesis is able to explain the spread of autoimmunity to anatomically distant tissues through immunological crossreactivity.
Rarely, multiple autoimmune diseases, including AIH2, and mucocutaneous fungal infection, occur in the same patient in the autoimmune polyglandular syndrome type 1 (APS1) (23). This is an autosomal recessive disease, caused by heterogeneous mutations at the autoimmune regulator (AIRE) locus on the long arm of chromosome 21 (24). The AIRE gene-product is expressed in the thymus, peripheral lymphoid tissues, and fetal liver (25). Although structural analysis suggests AIRE functions as a nuclear transcriptional regulator, its precise function/s have yet to be fully characterised. There is some evidence for AIRE regulating HLA class II expression, as cells expressing AIRE also exhibit high HLA class II expression (24). These findings allow the possibility of AIRE contributing to the modulation of thymic selection and, by increasing the surface density of rare self-epitopes presented by HLA class II molecules on thymocytes, results in the selection of autoreactive T-cells. These T-cells would then become activated in the periphery under similar conditions of high HLA expression, as is found in endocrine tissue undergoing autoimmune attack.
As activated T-cells are able to traverse endothelial barriers and therefore traffic through mutiple organs, autoreactive T-cells activated in a particular organ, could expand and infiltrate other organs if epitopes crossreactive to the activating epitope were encountered. As activated T-cells provide ‘help’ for B-cell immunoglobulin isotype class switching, it is conceivable that the above crossreactive lymphocyte response would ultimately manifest as a class-switched crossreactive antibody response. Such a mechanism would explain our finding of crossreactive antibody responses to structurally similar autoantigens in different tissues in the context of AIH2 and associated endocrine autoimmunity. The role of AIRE mutations in AIH2, especially when associated with other autoimmune disease, therefore deserves urgent address.
Microbes, mitochondria and mimicry
Primary biliary cirrhosis (PBC) is a chronic cholestatic liver disease characterised by mononuclear cell infiltration and progressive destruction of the small interlobular bile ducts, eventually leading to cirrhosis (26). Virtually all patients with PBC (96%) have antibodies reactive to mitochondria (AMA) (26). The principal target of these antibodies is the E2 component of the mitochondrial antigen – pyruvate dehydrogenase (PDC-E2), although other major autoantigens such as 2-oxoglutarate dehydrogenase and branched chain 2-oxo acid dehydrogenase have also been characterised (27). In view of the remarkable sequence conservation of PDC-E2 across species, Burroughs et al. suggested molecular mimicry between microbial PDC-E2, that of E. coli in particular, and human PDC-E2 could give rise to a crossreactive immunological response manifesting as AMA (28).
Epidemiological studies have provided some support for persistent or repeated bacterial and mycobacterial infection in the aetiology of PBC (29–31). Of these, the association between E. coli and PBC has received some degree of experimental corroboration. Hopf et al. report an association between rough form mutants of E. coli in faecal samples of patients with PBC, whilst Butler et al. report 69% of sera from patients with recurrent urinary tract infections, especially with E. coli organisms, recognising PDC-E2 (32, 33). Conversely, patients with PBC have a higher incidence of recurrent urinary tract infections (33). More recently, Mayo et al. reported antibody recognition of a carboxyl terminal region of E. coli ClpP, a component of the E. coli intracytoplasmic Clp protease, ClpP177–194, in a third of sera from patients with PBC (34). Importantly, this reactivity was strongly disease-specific as less than 1% of sera from patients with other autoimmune diseases recognised this antigen. However, these workers failed to find significant sequence similarity of E coli ClpP177–194 with the major mitochondrial autoantigens or other self-proteins that would fit with a mechanism of molecular mimicry.
We have recently proposed a mechanistic model, based on molecular mimicry at the level of the T-cell, for the emergence of humoral reactivity to E. coli ClpP177–194, that explains this apparent anomaly (35) (Fig. 3). Previous studies by our group demonstrated reactivity of sera from patients with PBC to a panel of E. coli peptides, derived from proteins unrelated to the oxo-acid dehydrogenase complex family, with sequence homology to the immunodominant epitope of PDC-E2 (PDC-E2213–227) (36). Surprisingly, the E. coli peptide with the highest sequence similarity to PDC-E2213–227, an ATP-binding component of the E coli Clp protease – ClpX280–294, was recognised with the lowest frequency (2%) in our patient sample (n=50). As both ClpP and ClpX are components of the Clp protease complex (37), we propose that this complex is internalised by B-cells with surface Ig receptors specific for ClpP177–194, and subsequently processed via the exogenous antigen-processing pathway. As B-cell microsomal proteases degrade the Clp complex, both ClpX and ClpP peptides are generated and loaded onto HLA class II molecules for presentation to CD4 helper T-cells. Recognition of E. coli ClpX280–294 by specific CD4 T-helper cells would trigger cognate and soluble cytokine signals, providing ‘help’ required for class-switching to IgG and secretion of E. coli ClpP specific IgG antibodies. We identify strong sequence similarity between ClpX280–294 and the immunodominant T-helper epitope of human PDC-E2, PDC-E2213–227, providing the basis for a crossreactive T-cell response to ClpX280–294. Thus, T-cell recognition of E. coli ClpX280–294 may operate in the generation of class-switched antibodies to both E. coli ClpP177–194, and human PDC-E2213–227 (Fig. 3).
Viral hepatitis and non-organ-specific autoantibodies
Autoantibodies to smooth muscle (SMA) and nuclear antigens (ANA) are frequently found, and appear de novo, in the course of chronic hepatitis B virus (HBV) infection, although their origin remains unclear (38). HBV does not directly cause hepatocyte injury but renders hepatocytes vulnerable to immune attack through poorly understood mechanisms. These include liver-specific alterations in cytokine milieu, hepatocyte damage as a bystander effect of T-cell-mediated attack on virus-laden hepatocytes, antibody-mediated lysis of infected hepatocytes, and antibody-dependent T-cell cytoxicity (39). The resulting exposure of normally sequestered self-antigens to immune scrutiny is thought to initiate the autoimmune response giving rise to non-organ-specific autoantibodies. This explanation is unable to fully explain the specific generation of SMA and ANA and the relative paucity of reactivity to other liver-specific antigens.
We have suggested molecular mimicry between HBV and smooth-muscle and nuclear antigens as a mechanism for the generation of these autoantibodies. In this regard, we have demonstrated crossreactive antibody responses between homologous regions of HBV DNA polymerase (HBV-DNA-pol) and four nuclear and two smooth muscle proteins using ELISA methodology (40) (Fig. 1). We further demonstrate crossreactivity between homologous peptide and whole antigen with respect to myosin and caldesmon, and identify these proteins as important autoantigens in chronic HBV infection by immunoblot. Interestingly, a common motif shared between HBV-DNA-pol, smooth-muscle myosin, caldesmon and the polymyositis-sclerosis nuclear antigen, EK[R,K]RLK, allows speculation on the possibility of a humoral response directed against HBV-DNA-pol giving rise, through molecular mimicry, to both nuclear and smooth-muscle specificities. This contention is supported by further experiments demonstrating crossreactivity between caldesmon600–619, smooth-muscle myosin836–855, pm-scl761–780, and HBV-DNA-pol99–118 (40).
Similarly, non-organ-specific autoantibodies, especially ANA and SMA, are a common feature of hepatitis C virus infection. Using similar methodology we have demonstrated crossreactive antibody responses between a number of HCV antigens and human nuclear and smooth-muscle antigens. Myosin also appears to be an important molecular target in HCV infection, although a different epitope to that in HBV infection is recognised. Antibody determinants of myosin in HBV and HCV infection therefore seem to be governed by crossreactivity to homologous viral sequences. In the context of a recent population-based study correlating non-organ specific autoantibodies (NOSA) with increased hepatic dysfunction and active viral replication, we suggest molecular mimicry resulting in immunological crossreactivity between self and HCV antigens may serve as an immune escape mechanism for the virus, diverting the immune response away from viral targets and towards homologous self-components (41). This would ultimately give rise to ANA and SMA whilst allowing viral persistence.
HCV infection, thyroid autoimmunity and the role of interferon
Autoantibodies against thyroid peroxidase (TPO), commonly associated with thyroid dysfunction, are frequently seen in HCV infection. Progression to overt autoimmune thyroid disease (AITD) occurs in a substantial proportion of these patients (42, 43). Interferon-alpha (IFN-α), used in the treatment of HCV infection, appears to potentiate/exacerbate thyroid autoantibodies/AITD (44, 45). The aetiology of AITD is unknown, although it is clear that mononuclear cell infiltration and T-cell autorecognition of thyroid-specific antigens such as TPO is a feature of autoimmune thyroiditis (46). TPO is not only localised within the follicular cells, but is also highly expressed on the cell surface (47). Moreover, autoantibodies to TPO, typically present at high titre, are capable of lysing thyrocytes in vitro (46). The occurrence of AITD in HCV infection is evidence for a viral component to aetiology in these patients, and has led us to propose molecular mimicry between HCV and thyroid antigens as a likely mechanism for their genesis. In this regard, we have demonstrated crossreactive antibody responses to homologous HCV and TPO peptides in patients with HCV infection (48). Simultaneous recognition of both HCV and homologous regions of TPO was seen in a significant minority of HCV-infected patients (14%) without autoimmune disease, not treated with IFN-α, whereas ‘double reactivity’ to homologous regions of HCV and TPO was higher in HCV patients with AITD (29%). Strikingly, the incidence of double reactivity to HCV/TPO homologues was four times greater in IFN-α-treated HCV patients with AITD as compared to IFN-α-treated patients without AITD (15% vs 60%). Antibody recognition of homologous HCV peptides was crossreactive to both peptide and native TPO as demonstrated by inhibition studies (49). These findings suggest molecular mimicry to operate in producing crossreactive responses to TPO in HCV infection, and that IFN-α appears to markedly exacerbate this process.
Approaching the end of the beginning
Sir Karl Popper propounded that the power of any hypothesis is determined by its testability and, as addendum, suggested this to be a function of the simplicity of its premises. Molecular mimicry leading to crossreactive autoimmunity is an attractive and conceptually simple hypothesis that is testable in all its aspects. That is not to say that the specific mechanisms that lead from molecular mimicry to autoimmunity are without complexity. We feel that molecular mimicry, as an important mechanism in autoimmune liver disease and viral hepatitis, has withstood preliminary tests of plausibility and experimental validation. A consensus for its role in these disease processes has consequently emerged. This review has focused on B-cell molecular mimicry but this implies, by its very existence, that T-cell correlates of this phenomenon must exist. Preliminary demonstrations of molecular mimicry at the cellular level are to date largely confined to PBC, and these show promise in shedding light on the precise structural requirements for mimicry and the functional cellular consequences of these crossreactions. Development of representative animal models for autoimmune hepatitis has proved difficult, but this deserves persistence, as such models allow the most powerful in vivo manipulations of immune components. This approach has led to important demonstrations of molecular mimicry in the aetiopathogenesis of multiple sclerosis (50), Lyme arthritis (4) and herpes stromal keratitis (5). In view of preliminary demonstrations of molecular mimicry as a possible mechanism of autoimmune liver disease in the human setting, we are poised for exciting new insights into the, thus far elusive, aetiopathogenesis of this group of diseases.