Primary biliary cirrhosis, bacteria and molecular mimicry: what's the molecule and where's the mimic?

Abstract

Primary biliary cirrhosis (PBC) is an enigmatic disease. While it is thought that an environmental agent precipitates disease in a genetically susceptible individual, we are still none the wiser about the specific genes and microbes that intersect to cause PBC. For quite some time, bacterial proteins have been touted as the molecular mimic for breaking tolerance to mitochondrial proteins in PBC. This notion has gained considerable momentum, as the majority of patients with PBC make anti-mitochondrial antibodies (AMA) and T-lymphocyte responses to the inner lipoyl domain of the pyruvate dehydrogenase complex (PDC)-E2 and other lipoylated mitochondrial enzyme complexes (1–3). PDC-E2 and related proteins are found in all organisms that use oxidative phosphorylation and it has been hypothesized that infection with a spectrum of several bacteria including Escherichia, Mycobacteria, Chlamydia, Helicobacter and Novosphingobium can trigger PBC by breaking tolerance to human PDC-E2 (1). In this issue of Liver International, Berg and colleagues add Mycobacteria pneumoniae to the long list of pathogens with a potential role in propagating PBC by showing that sera from PBC patients has reactivity against recombinant PDC-E2 encoded by M. pneumoniae (4). The authors also showed that bacterial reactivity was not blocked by excess porcine PDC-E2 in a proportion of patients and some PBC patients recognized M. pneumoniae PDC-E2 but not porcine PDC-E2. Based on these findings the authors hypothesize that infection with M. pneumoniae results in an immune response directed against PDC-E2 present on the bacterial cell membrane to initiate PBC. Many investigators have posited a bacterial etiology of PBC based on PDC-E2 cross-reactivity (1) but is there additional evidence to support the bacterial molecular mimicry theory? Originally, molecular mimicry was adopted as an attractive model for many autoimmune disorders because it provided a link with infection and loss of tolerance to host proteins (5). The hypothesis stemmed from the idea that microbes share immunodominant epitopes with the host. These molecular mimics either allowed the microbe to escape immune recognition or, alternatively, provoked loss of tolerance to the host protein once the microbe was recognized. Although this idea has been circulating for over 40 years, there is no convincing evidence that human autoimmune disorders are actually triggered by molecular mimicry (5). Indeed, only one animal model of herpes virus induced stromal keratitis in mice has linked molecular mimicry with disease and even then, viral replication is still required to perpetuate disease (6). So why has it been so difficult to prove that molecular mimicry triggers human autoimmune diseases? Certainly, it is problematic to demonstrate causality for any microbial infection in chronic, multi-factorial complex disorders. But shouldn't we be looking for alternative models for breaking tolerance now that molecular mimicry hypothesis has been ongoing for four decades without concrete evidence that it triggers human disease? There are additional layers of complexity in establishing the autoimmune pathogenesis of PBC. For example, it is fair to ask how an autoimmune response to a protein located on the inner mitochondrial membrane of nearly all nucleated cells causes an organ-specific disease targeted to biliary epithelium. Some light has been thrown on this mystery by studies showing aberrant expression of large amounts of PDC-E2 on the cell surface of biliary epithelium and lymph node macrophages from patients with PBC (2). It was originally thought that the protein found on biliary epithelium was a molecular mimic of PDC-E2 as not all preparations of AMA reacted with the antigen (7). However, subsequent studies performed on biliary epithelium isolated from PBC patients' livers revealed that these proteins had the same molecular weight as human PDC-E2 and the related PDC-E3 binding protein that is also reactive with AMA (8). The search for bacterial PDC-E2 in the liver has been unrewarding and the consensus is that specific bacteria are not found in the biliary epithelium or the liver of patients with PBC, apart from a couple of reports to the contrary (1). So it appears as though the trigger to break tolerance is actually human PDC-E2 or a modified form of the protein located on the biliary epithelium cell surface rather than a bacterial molecular mimic (2, 3). As bacteria have not been reproducibly detected in the liver of PBC patients, the current molecular mimicry model proposes that patients with PBC suffer recurrent urinary tract or other bacterial infections. In this regard, it is noteworthy that in a large survey of North American patients, a history of urinary tract infections were recorded in 59% of PBC patients and 52% of control subjects (9). The caveat here is that these events only trigger PBC in susceptible patients. The molecular mimicry model predicts that bacterial infection provokes an immune response to the highly conserved bacterial PDC-E2 proteins and the ensuing autoimmune response hones in on the biliary epithelium to cause cholangitis (Fig. 1a). An important consideration here is that the autoimmune response can only target the biliary epithelium because PDC-E2 is exposed on the cell surface, which probably is a critical factor to expose cryptic epitopes to the immune system. Indeed, it has been suggested that the detection of intact PDC-E2 is specific to biliary epithelium undergoing apoptosis due to lack of degraded PDC-E2 linked to reduced glutathiolation in this cell type (10). In this model, the mitochondrial protein expression on biliary epithelium is a proximal or early event in the disease process to permit a specific autoimmune attack on bile ducts. Once established, the autoimmune process is thought to be propagated by epitope spreading, where other antigens become immunogenic (5). Models for loss of tolerance to PDC-E2 and PBC pathogenesis. (a) In a two hit process, (i) biliary epithelial cells undergo apoptosis or other modifications with xenobiotics, for example, and present intact PDC-E2 on the biliary epithelium (3). Then (ii) bacterial infection results bacterial PDC-E2 presentation in antigen presenting cells (APC), which breaks tolerance to human PDC-E2 (arrows represent T-cell help). The subsequent autoimmune response hones in on the aberrant expression of PDC-E2 in biliary epithelium (? represents factors that may precipitate cell surface PDC-E2 in a). (b) In a single hit process, the MMTV-like human betaretrovirus infects biliary epithelium leading to PDC-E2 expression on the cell surface (12). Then either the virus incorporates PDC-E2 while budding from the cell surface, or exits with PDC-E2 in exosomes [not shown in this figure, see Onlamoon et al. (13) for example]. APC then present PDC-E2 and viral proteins, resulting in a bystander immune response to PDC-E2 and an immune response to viral and self-proteins expressed on biliary epithelium. PBC, primary biliary cirrhosis; PDC, pyruvate dehydrogenase complex. One problem with this complex molecular mimicry model is that it requires two unrelated events to cause PBC: the first being the generation of the mitochondrial phenotype with PDC-E2 expression in biliary epithelium and the second a loss of tolerance to PDC-E2 following bacterial infection (Fig. 1a). An alternative mechanism for AMA production has been suggested for xenobiotics that modify the lipoyl domain of PDC-E2 by creating neoantigens, possibly in concert with xenobiotic modifying bacteria such as Novosphingobium aromaticivorans (3). However, this model also provides no explanation of how PBC patients develop the mitochondrial phenotype with increased PDC-E2 in bile ducts. A second problem with the molecular mimicry model is that there is no direct evidence to link specific bacterial pathogens with PBC. For example, in the study reported by Berg and colleagues, a similar proportion of control subjects and patients with PBC had positive serology to immunogenic M. pneumoniae proteins P1 and CARDS TX or PCR evidence of M. pneumoniae in blood (4). Similarly, there are no data to link N. aromaticivorans to PBC except that PBC patients make AMA that react with the bacterial PDC-E2 (3). Another limitation of this model, and the one most difficult to address, is that this model anatomically divorces the infection from the autoimmune process in the liver. Indeed, molecular mimicry was first described as a way of intracellular microbes avoiding immune recognition and animal models of microbial molecular mimicry are associated with microbial infection in the affected organ (5). Also, this purely autoimmune model is not really supported by clinical observations in patients with PBC. Individual immunosuppressive treatments have had little impact on halting the progression of PBC and have not been adopted because of toxicity or lack of efficacy (11). In fact in liver transplant recipients, more potent immunosuppressive regimens using tacrolimus accelerate the onset and severity of recurrence of PBC (11). Another model to consider for the loss of tolerance to PDC-E2 is bystander activation or ‘guilt by association’. In this model, an infectious agent that provokes both PDC-E2 expression on the cell surface and an immune response to the infected cell could theoretically provoke an autoimmune response. In this case, the immune attack would be both anti-microbial and autoimmune. At present there is only one environmental trigger that has been reported to increase the level and location of PDC-E2 in biliary epithelium (12). The human betaretrovirus related to the mouse mammary tumour virus (MMTV) characterized in patients with PBC does not encode proteins resembling mitochondrial antigens and is therefore not linked to molecular mimicry. However, the MMTV-like virus is detected in cells expressing the mitochondrial phenotype in PBC patient's lymph nodes, and has been shown to increase PDC-E2 in biliary epithelium in vitro following infection (12). In this scenario, AMA production is triggered as a bystander effect, whereby T-lymphocyte helper cells recognize the betaretroviral infection and recruit an immune response to PDC-E2 expressed on the cell surface as a bystander effect. Alternatively, the virus incorporates PDC-E2 as it buds from the cell surface, or from exomes in multi-vesicular bodies (13), and both viral and mitochondrial proteins are presented by antigen presenting cells (Fig. 1b). Either way, the immune response to PDC-E2 is triggered as a result as ‘guilt by association’ process accompanying betaretroviral infection. Apart from the AMA reactivity to bacterial PDC-E2 and the unconfirmed report of C. pneumoniae in livers of PBC patients, there is little data to directly link any bacteria with PBC at this time (1). Likewise, the report of detecting human betaretrovirus in a significant proportion of patients with PBC remains unconfirmed and controversial (14). Recently, new animal models of PBC have provided support for both the bacterial and viral hypotheses. For example, in a genetically susceptible non-obese diabetic (NOD) derived mouse strain, AMA production and cholangitis were triggered by infection with N. aromaticivorans (15). Similarly, the NOD.c3c4 mouse that develops spontaneous AMA production and cholangitis was found to have high levels of PDC-E2 and MMTV proteins in the biliary epithelium (16). While these models demonstrate the possibility of bacterial and viral triggers of PBC, there is still a long way to go to understanding the environmental factors that may impact on the development of PBC and other autoimmune disorders. In the same way that genome-wide association scans have begun to shed light on some of the complexity of genetic modifiers for autoimmune diseases, it is hoped that similarly high powered techniques such as a metagenomic approach to detecting novel pathogens with massively parallel sequencing might help to discover the potential pathogens associated with PBC and other autoimmune diseases (17). Once these genetic and environmental factors that cause disease have been unraveled, we can start to make sense of the complexity of disease and devise rational strategies for treatment. In the mean time, more compelling data from clinical studies are required to directly link microbial infection with PBC from all investigators in the field.