Introduction

Primary biliary cirrhosis (PBC) is an autoimmune cholestatic disease, which is characterized by bile duct loss leading to fibrosis, cirrhosis and subsequent liver failure [1–5]. The disease shows a remarkable female predominance, mainly affecting middle aged women and at times affects several members within a family [6–11]. The disease can run an asymptomatic course. The complaints of symptomatic patients include non-specific symptoms such as fatigue, pruritus, sicca symptomatology and arthralgias [1–5, 12]. Patients at more advanced stage of the disease suffer from symptoms of liver decompensation and portal hypertension [1–5, 12]. The disease progresses at slow pace, but the progress rates are difficult to predict [1–5, 12–14].

The widely accepted diagnostic criteria for PBC are based on biochemically-evident cholestasis, disease-specific autoantibody serology and histological features consistent with PBC [1–5, 12, 13, 15, 16]. Cholestatic markers assisting the diagnosis include increased levels of alkaline phosphatase and γGT [1–5]. More than 90 % of the patients have detectable anti-mitochondrial antibodies (AMA), while PBC-specific anti-nuclear antibodies (ANA) are present in up to 50 % of the affected cases [15–23]. The mere presence of AMA predicts the development of PBC in asymptomatic patients, as a significant proportion of acholestatic women with positive AMA tests develop clinical features of PBC over time [6, 14–17, 24–26].

PBC-specific AMA targets the 2-oxo-acid dehydrogenase complexes, and in particular the E2 subunit of the pyruvate dehydrogenase complex (PDC) in approximately 80–90 % of AMA-positive cases [20, 27–38]. The E2 subunits of branched-chain 2-oxo-acid dehydrogenase complex (BCOADC) and 2-oxo-glutarate dehydrogenase complex (OGDC) are targeted to a lesser extent, while reactivity to the E1α and E1β subunits of PDC has been considered less significant [20, 27–36]. Most of the authors agree that neither the titre nor the presence of AMA bears clinical significance [15–17, 39–41]. The detection of AMA is usually performed by indirect immunoflourescence assay (IFA). Immunoblotting was more popular in the past, but in recent years has been replaced by ELISA or line/dot blot immunoassays [15–17, 40–51]. ANA specifically found in PBC recognise the nuclear body proteins such as sp100, promyelocytic leukaemia, sp140, and small ubiquitin-like modifiers, or nuclear envelope antigens such as gp210 and nucleoporin p62 [15, 16, 19, 52–56]. Reactivity to nuclear body proteins gives a multiple nuclear dot pattern and anti-nuclear membrane antibodies give a nuclear rim pattern by IFA [15, 16, 18, 57, 58]. When AMA are absent, ANA specific for the disease appear to be more frequent. These ANA appear to have prognostic and clinical significance, and can be found in asymptomatic women and members of their family [18, 48, 52, 53, 57, 59–68].

The aetiopathogenesis of PBC remains poorly understood, and the mechanisms responsible for the development of AMA and ANA are theoretical [21, 69–72]. Whether PBC-specific autoantibodies are pathogenic is not known [17, 55, 73, 74]. While some believe that these autoantibodies are just epiphenomena, others consider them as the indicators of autoantigen-specific processes involving CD4 and CD8+ T lymphocyte autoreactive populations which are able to inflict tissue damage [17, 55, 71, 73–76]. The possibility that the mechanisms responsible for the induction of PBC-specific autoantibodies are also responsible for the pathogenesis of PBC has not been excluded [77, 78].

Most agree that infections by themselves are not sufficient to induce disease and that the other environmental agents such as toxins, as well as genetic factors are also important [28, 52, 65, 66, 73, 79–104]. The striking female predominance of PBC has also been the focus of increased interest in the recent years [78, 94, 105–108]. E. coli has been amongst the first infectious agents to be considered a likely trigger of the disease [28, 65, 70, 77, 82, 89–91, 109, 110]. Early epidemiological studies have shown that patients with PBC suffer more frequently than other patients with liver disease from recurrent urinary tract infections (UTI), which are more prevalent in PBC patients than in patients with other liver diseases [88, 90, 111]. More recent studies have replicated these findings [112, 113]. Within the last decade or so, Novosphingobium aromaticivorans has been considered a serious candidate and has been pathogenetically linked to the disease. Most work to support this view has come from one group that has provided serological and other experimental data to support the role of this agent with the development of the disease [52, 79, 82, 98, 99]. Other infectious triggers have also been studied, but the number of publications per agent is considerably outsized by that related to Novosphingobium aromaticivorans and E.coli [114]. Nevertheless, compared to other autoimmune liver diseases such as autoimmune hepatitis, the list of potentially harmful infectious agents linked to PBC is noteworthy [86, 95, 114]. Most of the infectious agents included in this list have led to studies failing to provide positive associations with PBC, and the involved triggers have not been studied extensively due to the profound lack of scientific interest from competing groups.

The role of molecular mimicry

The mechanism of molecular mimicry has been widely used as an explanation for infectious-triggered autoimmunity (mainly viral hepatitides B and C), and for the induction of autoimmune liver diseases due to infections (both viral and non-viral) [17, 52, 65, 66, 70, 73, 74, 79–82, 84, 87, 98, 99, 110, 115]. As most epitopic regions of the AMA and ANA targets are known, the mechanism of molecular mimicry has been used as a tool to identify bacterial triggers of PBC that serve as targets of cross-reactive responses [70, 110]. This mechanism implies that because of the amino acid similarity between infectious and self epitopes, responses against the infectious agent will provoke the development of a cross-reactive response against mitochondrial and nuclear autoantigens (Table 1) [21, 69, 74]. These cross-reactive responses operate at the B- and T cell level [70, 116].

Table 1 Molecular mimicry studies involving various infectious agents as the likely triggers of primary biliary cirrhosis. [52, 66, 79, 84, 85, 143, 145, 149]
Full size table

Popular infectious triggers of PBC

N. aromaticivorans

Experimental data have provided evidence that this aerobic, gram-negative bacterium may be involved in the development of PBC. Molecular mimicry involving N. aromaticivorans and mitochondrial autoantigens has been demonstrated [98, 117, 118]. Selmi et al. [98] were the first to identify two N. aromaticivorans proteins with striking amino acid homology with PDC-E2. All PBC patients who recognised PDC-E2 reacted with the bacterial proteins, and molecular testing has led to the isolation of N. aromaticivorans in faecal samples from PBC patients. A subsequent study has shown that N. aromaticivorans contains PDC-E2-like proteins in their cell membranes [119]. Reactivity to N. aromaticivorans prevailed over the reactivity to that against the mimicking E. coli antigen. Another study from Iceland reported that patients with PBC and their first degree relatives had antibodies against N. aromaticivorans [120]. These anti-microbial antibodies had the ability to cross-react with mitochondrial autoantigens. Individuals with antibodies to N. aromaticivorans, but without evidence of PBC, could develop the disease some years later [120]. Although 25 % of both PBC and FDR groups had N. aromaticivorans in faecal samples, only those with PBC had antibodies against them [120].

Animal studies with mice inoculated with N. aromaticivorans provided data to support the presence of IgG and IgA antibodies against bacterial and mammalian PDC-E2 [118]. Also, the inoculated mice developed histological features of PBC in a Natural Killer T cell dependent manner [118]. Of great interest is the fact that in this experimental setting, transfer experiments of T cells from infected to uninfected mice led to the transfer of bile duct pathological alterations, that were preventable by the early administration of antibiotic treatment [118].

E. coli

When Burroughs and Baum [87] formed the hypothesis that molecular mimicry between human and E. coli PDC-E2 is responsible for the induction of AMA and autoreactive lymphocytes reacting with PDC-E2, they did not know what was going to follow. Several studies have been conducted based on the assumption that E. coli infection may trigger PBC (Table 2). Several reports have shown that sera from PBC patients cross-react with bacterial and human PDC-E2 [20, 89–91, 121, 122]. Subsequent immunological findings have been obtained to demonstrate the presence of cross-reactive antibodies targeting human PDC-E2 and E. coli mimics from proteins unrelated to PDC-E2 or other members of the 2-OADC [66, 84]. While protein–protein BLAST search has identified several E-coli sequences sharing homology with human PDC-E2, very few of those were cross-recognised by PBC sera [66, 82, 99, 110].

Table 2 Major findings of the studies investigating the role of E. coli infection in the pathogenesis of primary biliary cirrhosis (PBC)
Full size table

Reactivity to human PDC-E2 is much stronger compared to that against mammalian PDC-E2 [123], and this also appears to be the case for OGDC-E2 [124]. At the peptide level, antibodies against PDC-E2212–226, which is the core epitopic region of human PDC-E2, do not appear to cross-react with E. coli PDC-E2 sequence [52], and this may be due to the lack of significant structural/antigenic mimicry [82]. The core region recognised by the anti-PDC-E2 antibodies significantly overlaps with the CD4 and CD8 T cell epitope on PDC-E2. E.coli and human PDC-E2 cross-react at the CD4 and CD8 T cell levels [84, 85, 125, 126]. Shimoda et al. [84, 85] have shown that the ExDK motif is essential for T cell epitope recognition. It has also been shown that T cell lines specific to the human PDC-E2 autoepitope developed from peripheral blood mononuclear cells or liver infiltrating cells obtained from patients with PBC can also be activated by a motif sharing OGDC-E2 peptide of E. coli [84, 85]. A relationship between evidence of recurrent UTI and the presence of PBC-specific responses against sp100 has been noted [80].

Several studies have reported an increased incidence of recurrent UTI in female PBC patients [88, 112, 127], and E. coli is isolated in most cases [88]. Others failed to find an epidemiological association between E. coli infection and PBC [128]. A recent epidemiological study has suggested a causal relationship between UTIs specific for the disease and PBC. This association preceded the diagnosis of PBC by at least 5 years [129]. Such a relationship has not been noted in the past, and re-enforces the link between E. coli and PBC.

Lactobacillus delbrueckii

L. delbrueckii subsp. bulgaricus is a gram positive rod used in the production of yoghurt, but is also found in the gastrointestinal tract and the vagina in humans. Vaginal lactobacilli maintain a flora which is protective against urinary tract infections. Alterations in the normal bacterial flora may result in bacterial vaginitis and increased UTI incidence. It appears that L. delbrueckii subsp. bulgaricus shares a striking amino acid homology with the core epitopic region on PDC-E2, and the mimicry shared by the infectious and human sequences is unique [52]. Cross-reactive antibody responses against the mimicking sequences belong to the IgG3 subclass and are specifically present in patients with PBC. Reactivity to the lactobacillus mimic can be found in healthy and pathological controls, but belongs exclusively to IgG4 subclass. A case of a German woman, who went on to develop PBC after vaccination with lactobacillus for recurrent vaginitis, has added support to the notion that this infectious agent may cause the disease [79]. The serum of this woman contained antibodies against the lactobacillus mimic, and cross-recognised the human PDC-E2 homologous sequence [79]. In a hypothetical scenario, changes of the normal vaginal flora provide the impetus for microbial infections that contribute to the development of PBC via molecular mimicry or other mechanisms. No other studies have investigated the role of lactobacilli in PBC.

Betaretroviruses

The extent by which betaretroviruses are involved in the development of PBC is currently unknown [130, 131]. Mason’s group has studied this topic in great detail. The initial finding of antibody reactivity to retroviruses has been complemented by the demonstration of viral particles in BECs using electron microscopy [96, 101, 132]. Later on, the same group has reported the cloning of a human betaretrovirus (HBRV) sequence with a remarkable nucleotide homology to the mouse mammary tumour virus (MMTV) through a screening of a PBC biliary epithelium cDNA library [100]. The same group reported the identification of retroviral sequences in a murine model of PBC and the integration of the virus in cholangiocytes from patients with PBC [133]. There is no doubt that Mason’s group has produced an extensive amount of data in support of a potential role of HBRV in the induction of PBC. However, an independent study by Selmi et al. [134] has failed to provide evidence in support of the role of betaretroviruses in the pathogenesis of PBC. These authors found no evidence of infection by a PCR approach [134]. A third group also conducted a study in the same topic and found that the presence of MMTV (MMTV-LV) is not limited to PBC, but it can be found in livers of patients with a wide range of hepatic disorders and not in normal livers [135]. Thus, Johal et al. analyzed 210 liver biopsies and found evidence of MMTV in 25 % of those. In fact, MMTV was present more often in non-PBC diseases livers (27 %) than in PBC livers (12 %), further indicating that if MMTV is probably involved in the induction of a liver disease which is likely to be happening in liver diseases other than PBC [135].

Of relevance to the practising physician, Mason’s group has conducted pilot studies treating PBC patients with Combivir, an antiviral combination treatment consisted of zidovudine and lamivudine. Original data showing some biochemical and histological improvement of patients undergoing treatment [136] have been discarded by subsequent findings, demonstrating lack of efficacy and evidence of resistance to treatment with biochemical rebound [137].

Unpopular infectious triggers of PBC

Helicobacter species

Several studies have attempted to link Helicobacter pylori with the development of PBC [138–141]. Evidence of anti-helicobacter antibody seropositivity has been found in serum samples from patients with PBC. Also, these antibodies are detectable in bile collected from PBC patients [138–141]. Further experiments have shown that animals immunized with helicobacter species can develop histological features resembling those seen in PBC [142]. However, there is no evidence of molecular mimicry between mitochondrial targets and the major helicobacter antigens [143].

Typical and atypical Mycobacteria

PBC is a granulomatous liver disease, and this has led the researchers to consider the possibility that M.tuberculosis and other mycobacterial infections may cause the disease [1, 144–146]. The presence of relatively low titres of AMA in patients with pulmonary tuberculosis and leprosy has added some support to the pathogenic role of mycobacteria in PBC [146, 147]. Others have considered atypical mycobacteria, such as M. gordonae, as important triggers of the disease, but the importance of the published data and their significance has been questioned [148]. Work conducted in Spain has suggested that patients with PBC contain antibodies reacting with the 65 kDa heat shock protein of Mycobacterium gordonae [144, 145]. Subsequent studies reported the presence of antibodies that cross-recognise human PDC-E2 and mycobacterial hsp65 [149].

Chlamydia pneumoniae

Abdulkarim et al. [86] have obtained striking evidence of C. pneumoniae DNA in explanted liver tissues from PBC, but not from patients with chronic hepatitis C or primary sclerosing cholangitis. A subsequent independent study did not confirm such findings [95]. PCR analysis did not reveal any evidence of C. pneumoniae DNA [95]. However, anti-chlamydial antibodies were found in the great majority of patients with PBC. Leung et al. [95] speculated that anti-chlamydial antibody seropostivity is due to the cross-recognition of mitochondrial autoantigens with the respective antigens from chlamydia.

Epstein–Barr and human cytomegaloviruses

Epstein–Barr virus (EBV) and human cytomegalovirus (HCMV) are two of the most frequently implicated viruses in the pathogenesis of autoimmune diseases [150, 151]. Their role in the pathogenesis of PBC remains obscure [152]. Experimental studies investigating the potential link of EBV with PBC are scarce. An early study investigating the activation status of B-cells of patients with PBC has found a reduced production of immunoglobulins by B-cells stimulated with EBV [153]. The same study has shown that EBV-stimulated cultures of lymphocytes from patients with PBC had diminished suppression of immunoglobulin-secreting cells [153]. A subsequent study has provided evidence of elevated levels of EBV-DNA in peripheral blood mononuclear cells isolated from patients with PBC compared to the levels noted in patients with AIH (61 vs. 19 %) [154]. Immunohistochemical analysis of liver tissues obtained from patients with PBC has also shown increased levels of EBV-DNA compared to those seen in patients with other liver diseases [154]. Increased levels of IgG antibodies against EBV early antigen have been found in patients with PBC compared to healthy controls. However, this finding was not specific for PBC as it was noted in patients with other diseases including patients with systemic lupus erythematosus, rheumatoid arthritis and multiple sclerosis [155].

Another study has shown inability of PDC-E2 mimics originating from EBV to act as a target of cross-reactive antibodies, suggesting that molecular mimicry between EBV and AMA targets is highly unlikely [66].

HCMV is a potent inducer of autoimmunity, but its role in PBC is far from clear. Elevated levels of IgM antibodies against CMV are detectable in patients with PBC compared to controls [155]. Again, such elevated levels are not specifically found in PBC but also in several other autoimmune diseases including patients with systemic lupus erythematosus, antiphospholipid syndrome, polymyositis, Sjögren’s syndrome, systemic sclerosis, Churg-Strauss vasculitis, mixed cryoglobulinemia and other diseases [155].

However, IgG antibodies against CMV are not elevated in patients with PBC compared to healthy controls and several autoimmune diseases [155]. Also, human PDC-E2 responses do not cross-react with their CMV mimics, indicating the lack of a mechanism of molecular mimicry.

Conclusion

Epidemiological and experimental data have provided support to the notion that PBC may be an infectious disease caused by specific bacteria infecting predisposed individuals. Some microbes have been studied more than others. As for other autoimmune diseases, the most prevailing scenarios include those suggesting that molecular mimicry is the mechanism most likely involved in the infectious-triggered autoimmune phenomena seen in patients who develop PBC. Emerging data from animal models of the disease will shed light onto the involvement of infection(s) in the pathogenesis of PBC. Some work has been done on the role of E. coli and N. aromaticivorans, but there is little to suggest that these or other microbial agents are really needed for induction of the disease.