Escherichia coli infection induces autoimmune cholangitis and anti-mitochondrial antibodies in non-obese diabetic (NOD).B6 (Idd10/Idd18) mice



Several epidemiological studies have demonstrated that patients with primary biliary cirrhosis (PBC) have a higher incidence of urinary tract infections (UTI) and there is significant homology of the immunodominant mitochondrial autoantigen, the E2 component of the pyruvate dehydrogenase complex (PDC-E2), between mammals and bacteria. Previous work has demonstrated that non-obese diabetic (NOD).B6 Idd10/Idd18 infected with Novosphingobium aromaticivorans developed liver lesions similar to human PBC. It was postulated that the biliary disease was dependent upon the presence of the unique N. aro glycosphingolipids in activating natural killer T (NK T) cells. To address this issue, we infected NOD.B6 Idd10/Idd18 mice with either Escherichia coli, N. aro or use of a phosphate-buffered saline (PBS) vehicle control and serially followed animals for the appearance of liver pathology and anti-mitochondrial autoantibodies (AMA). Of striking importance, the biliary disease of E. coli-infected mice was more severe than N. Aro-infected mice and the titre of AMA was higher in E. coli-infected mice. Furthermore, the immunopathology did not correlate with the ability of bacterial extracts to produce antigen-dependent activation of NK T cells. Our data suggest that the unique glycosphingolipids of N. aro are not required for the development of autoimmune cholangitis. Importantly, the data highlight the clinical significance of E. coli infection in a genetically susceptible host, and we suggest that the appearance of autoimmune cholangitis is dependent upon molecular mimicry. These data highlight that breach of tolerance to PDC-E2 is probably the first event in the natural history of PBC in genetically susceptible hosts.


It is becoming increasingly clear that the appearance of autoimmunity is dependent upon a combination of genetic predisposition and environmental factors [1-3]. Further, a number of microbial infections have been postulated to trigger a cascade of immunological events in genetically susceptible hosts that lead to a breach of tolerance to self-antigens [4-8]. Although multiple mechanisms have been proposed involving both innate and adaptive responses, all depend upon the concept of molecular mimicry [9-12]. Indeed, this discussion is important because in human primary biliary cirrhosis (PBC), several epidemiological studies have demonstrated an increased incidence of urinary tract infections (UTIs) [13, 14].

The serological hallmark of PBC is the presence of anti-mitochondrial autoantibodies (AMA), considered the most specific diagnostic marker of PBC, but also among the most highly directed specific autoantibodies in human immunopathology [15, 16]. The autoantigens have been identified as the E2 subunits of the 2-oxo-acid dehydrogenase complexes (2OADC-E2), including the E2 subunits of the pyruvate dehydrogenase complex (PDC-E2), branched chain 2-oxo-acid dehydrogenase complex (BCOADC-E2), 2-oxo-glutarate dehydrogenase complex (OGDC-E2) [16-18] and the E3 binding protein of dihydrolipoamide dehydrogenase [19]. The AMA target antigens are all localized within the inner mitochondrial matrix and catalyze the oxidative decarboxylation of 2-oxo-acid acid substrates [20]. Biochemically, the 2OADC-E2 has a common functional domain containing a single or multiple lipoyl groups. The immunodominant epitopes recognized by AMA are mapped within the lipoyl domains of these target antigens [21, 22]. In patients with PBC, T helper (CD4+) T cells and cytotoxic (CD8+) T cells are present in portal tracts around damaged bile ducts [23]. Both PDC-E2 specific CD4 and CD8 autoreactive T cells have been identified in PBC, and are highly enriched in the liver versus peripheral blood. Interestingly, the autoreactive CD4 and CD8 T cell epitopes in patients with PBC also map within the lipoyl domain and overlap with the B cell epitope [24-27].

Novosphingobium aromaticivorans is a bacterial species that has attracted attention with respect to the aetiology of PBC for several reasons. First, N. aro is a unique ubiquitous bacterium that metabolizes xenobiotics. Secondly, there are significant autoantibodies to PDC-E2 that are immunoreactive to N. aro, perhaps because N. aro contains four copies of PDC-E2-like proteins [28, 29]. Furthermore, it has been reported that N. aro-infected mice developed autoantibodies to PDC-E2 and liver histology similar to humans with PBC [30]. The model is postulated to occur because of the unique potential of the N. aro glycosphingolipids in activating natural killer T (NK T) cells. The data also suggested that the non-obese diabetic (NOD).B6 insulin-dependent diabetes susceptibility region (Idd10/Idd18) contains the genetic loci that are important in determining the bile duct lesions in the N. aro-infected mice. More recently, Mohammed et al. reported [31] that the Idd10 region in the NOD.B6 Idd10 mice infected with N. aro developed liver lesions similar to PBC, which correlates with the genotype-dependent expression of cd101, a murine type 1 diabetes candidate gene.

We have explored this issue in more detail; in particular, a rigorous serial study of Escherichia coli-infected mice. We report herein that E. coli-infected NOD.B6 Idd10/Idd18 develop liver lesions strikingly similar to the portal infiltrates of humans with PBC. N. aro-infected mice, as expected, also develop autoimmune cholangitis but, interestingly, the autoantibodies were higher in the E. coli-infected mice. Our data suggest that infection of a genetically susceptible host with the evolutionarily conserved PDC-E2 has the potential to break tolerance and elicit biliary pathology. These data take on further significance in light of the epidemiological data in humans of urinary infections and subsequent development of PBC.

Materials and methods

Bacterial strains

N. aro (ATCC 700278; American Type Culture Collection, Manassas, VA, USA) and E. coli (DH5α, ATCC 25922; American Type Culture Collection) were grown overnight in Mueller Hinton broth (Becton-Dickinson, Franklin Lakes, NJ, USA) and Luria–Bertani broth, respectively, and then inoculated in fresh medium, grown for 8 h (E. coli at 37°C, N. aro at 30°C) to an optical density (OD) of 0·5 at 600 nm, washed and resuspended in sterile phosphate-buffered saline (PBS) for immediate administration to experimental animals or to prepare sonicates for antigen presentation assays. Sphingomonas yanoikuyae (ATCC 51230; American Type Culture Collection) were grown at 30°C in tryptic soy broth.

Animals and bacterium infection

Female NOD.B6 Idd10/Idd18 (lines 7754) mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA) and maintained in individually ventilated cages under specific pathogen-free conditions at the University of California at Davis animal facility. All experimental protocols were approved by the University of California Animal Care and Use Committee. The mice were separated into three groups: 13 were infected with N. aro, 13 were infected with E. coli and six were administered with sterile PBS as controls. Briefly, aliquots of 5 × 107 N. aro, or E. coli in 100 μl PBS were administered intravenously (i.v.) into 6-week-old mice through periorbital venous sinus and once more 14 days thereafter. Blood samples were collected every 2 weeks after inoculation. At 26 weeks after inoculation, animals were killed and liver tissues were harvested for histological analysis (Fig. 1).

Figure 1.

General experimental protocol. The mice were separated into three groups: 13 were infected with Novosphingobium aromaticivorans, 13 were infected with Escherichia coli and six were administered with sterile phosphate-buffered saline (PBS) as controls. Briefly, aliquot of 5 × 107 N. aro or E. coli in 100 μl PBS was administered intravenously (i.v.) into 6-week-old mice through periorbital venous sinus and once 2 weeks thereafter. Blood samples were collected every 2 weeks for anti-mitochondrial autoantibodies (AMA) determination. At 26 weeks after inoculation, animals were killed and liver tissues were harvested for histological analysis.

Preparation of recombinant mitochondrial autoantigen

Recombinant human PDC-E2 protein was prepared as described previously [22]. Briefly, overnight E. coli cultures expressing the human PDC-E2 lipoyl domain in plasmid pGEX4T-1 were diluted 1:10 with fresh Lauria–Bertani medium (50 μg/ml ampicillin) until the OD was 0·7–0·8, and induced with 1 mM isopropyl-b-thiogalactopyranoside for an additional 3–4 h at 37°C. Cells were pelleted, resuspended in PBS containing 1% Triton X-100 and 1% Tween-20 (Sigma Chemical Co., St Louis, MO, USA) and sonicated. The sonicated extract was centrifuged at 10 000 g for 15 min at 4°C; the supernatant was collected and incubated with glutathione agarose beads (Sigma) for 2 h at room temperature. Gluthathione agarose beads were washed three times with PBS and the fusion protein was eluted by competition with 50 mM Tris HCl pH 8·0 containing 20 mM reduced glutathione (Sigma). Protein concentrations of the eluate were determined by bicinchoninic acid assay (Thermo Scientific, Tewksbury, MA, USA). Recombinant BCOADC-E2 and OGDC-E2 were purified similarly [22].

Determination of serum AMA by enzyme-linked immunosorbent assay (ELISA)

Serum samples were examined for levels of anti-PDC-E2 antibodies using an ELISA. Briefly, 96-well ELISA plates were coated with 5 μg/ml of purified recombinant PDC-E2 in carbonate buffer (pH 9·6) at 4°C overnight, washed with Tris-buffered saline Tween-20 (TBS-T) and blocked with 5% skimmed milk in TBS for 30 min. Serum samples (diluted 1:500) were added to individual wells of the microtitre plate and incubated for 1 h at room temperature (RT). After washing, horseradish peroxidase-conjugated anti-mouse immunoglobulin (Ig) (A + M + G) (H + L) (1:3000) (Zymed, San Francisco, CA, USA) was added. The plates were incubated for 1 h at RT, then washed. OD450nm was measured after addition of 3,3′,5,5′-tetramethylbenzidine peroxidase substrate (BD Biosciences, San Jose, CA, USA) and incubation at room temperature for 5 min. Previously calibrated positive and negative standards were included with each assay [21, 32].

Detection of serum AMA by immunoblotting

A measured quantity of 20 μg of either recombinant human PDC-E2 protein recombinant BCOADC-E2 or recombinant OGDC-E2 was resolved on 10% sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membrane. The membrane was then cut into 3-mm strips; each carried approximately 0·6 μg of recombinant protein, blocked with 3% non-fat dry milk in PBS for 1 h and then incubated with mouse sera (1:500 dilution) for 1 h. Membranes were then washed four times with PBS containing 0·05% Tween 20, 10 min each, before incubating with horseradish peroxidase-conjugated anti-mouse Ig (Zymed) for 1 h at room temperature. Membranes were then washed with PBS containing 0·05% Tween 20, followed by chemiluminescent detection (Pierce, Rockford, IL, USA) [33].

NK T cell activation assay

The CD1d-reactive NK T cell hybridomas 1·2 and 2C12 have been described previously [34]. Stimulation of T cell hybridomas on CD1d-coated plates was carried out according to published protocols [35]. Briefly, the indicated dilutions of bacterial sonicates were incubated for 24 h in microwells coated with 1·0 μg of mouse CD1d. The strong NK T cell glycolipid agonist α-galactosylceramide was added to CD1d-coated wells as a positive control. After washing, 5 × 104–1 × 105 NK T cell hybridomas were cultured in the plate for 16–20 h, and IL-2 in the supernatant was measured by ELISA (BD PharMingen, San Diego, CA, USA).


Liver tissues were collected immediately from animals upon killing, fixed in 4% paraformaldehyde, embedded in paraffin, cut into 4-μm sections, deparaffinized, stained with haematoxylin and eosin (H&E) and evaluated using light microscopy [36]. Scoring of liver inflammation was performed on coded H&E-stained sections of liver using a set of three indices by a ‘blinded’ pathologist (K.T.); indices including degrees of portal inflammation, parenchymal inflammation and bile duct damage were scored as: 0 = normal, no inflammation (or bile duct damage); 1 = minimal inflammation (or bile duct damage); 2 = mild inflammation (or bile duct damage); 3 = moderate inflammation (or bile duct damage); and 4 = severe inflammation (or bile duct damage).


To examine the bile duct pathology, immunochemical staining was performed with a rabbit polyclonal antibody for cytokeratin (CK) 19, which is an established marker of biliary epithelial cells. Liver sections were immunostained using standard microwave protocol, as described previously [37]. In brief, after deparaffinization and microwave heating for antigen retrieval, rabbit polyclonal antibody against CK19 (Novus Biologicals, Littleton, CO, USA) was applied and incubated under intermittent microwave irradiation. After rinsing with TBS, Envision-peroxidase for rabbit polyclonal antibodies (Dako, Carpenteria, CA, USA) was applied and incubated under intermittent microwave treatment. As a substrate of peroxidase, 3,3′-diaminobenzidine (DAB; Vector, Burlingame, CA, USA) was applied for 5 min. Heamatoxylin was used as a counter-stain.

Statistical analysis

Data are presented as the mean ± standard error of the mean (s.e.m.). Two-sample comparisons were analysed using the two-tailed unpaired t-test. The correlation between two parameters was analysed using Spearman's correlation method. A value of P < 0·05 was considered statistically significant.


Anti-mitochondrial autoantibodies

As shown in Fig. 2a, the levels of anti-PDC-E2, measured as OD values in ELISA using 1:500 diluted serum samples, were significantly higher (P < 0·001) in E. coli-infected mice 4–12 weeks after bacterium infection when compared with the N. aro-infected mice and the uninfected control group. The level of anti-PDC-E2 peaked at 4 weeks after E. coli infection and then gradually decreased to the same level as that of N. aro-infected mice. Anti-PDC-E2 and anti-OGDC-E2 antibodies were detected in the serum of E. coli-infected mice but not N. aro-infected mice, while anti-BCOADC-E2 antibodies were not detected in either group (Fig. 2b). Next we validated the specificity of AMA by immunoblotting, which confirmed the presence of anti PDC-E2 antibodies in both E. coli- and N. aro-infected mice but not in control mice (Fig. 2c). These data indicated that both N. aro and E. coli infection could elicit AMA production, but that E. coli was the more potent stimulus.

Figure 2.

Anti-mitochondrial autoantibodies (AMA) determination of Novosphingobium aromaticivorans, Escherichia coli-infected mice and control mice. (a) Longitudinal analysis of serological anti-E2 subunit of the pyruvate dehydrogenase complex (PDC-E2). Sera obtained from NOD.B6 Idd10/Idd18 mice administrated with either N. aro (n = 13) or E. coli (n = 13) or phosphate-buffered saline (PBS) only (n = 6) were assayed for anti-PDC-E2 antibodies by enzyme-linked immunosorbent assay (ELISA) at 1:500 sera dilution. Note that antibody to recombinant PDC-E2 was significantly higher (P < 0·001) in the E. coli group when compared with the N. aro group at 4, 8 and 12 weeks post-infection. (b) Sera from NOD.B6 Idd10/Idd18 mice administered with either N. aro (n = 13) or E. coli (n = 13) or PBS only (n = 6) were collected at 12 weeks post-infection and assayed for anti-PDC-E2, anti-2-oxo-glutarate dehydrogenase complex (OGDC)–E2 and anti- E2 subunit of the branched chain 2-oxo acid dehydrogenase complex (BCOADC–E2) antibodies by ELISA at 1:500 sera dilution. Note that antibody to recombinant PDC-E2 was significantly higher (P < 0·001) in the E. coli group when compared with the N. aro group at 12 weeks post-infection. (c). Immunoblot analysis of the N. aro, E. coli and PBS groups at 4 weeks post-infection. Sera obtained from NOD.B6 Idd10/Idd18 mice administered with either N. aro (n = 13) or E. coli (n = 13) or PBS only (n = 6) were analysed for anti-PDC-E2 antibodies by immunoblot at 1:500 sera dilution. M: molecular weight in kilodaltons (KDa), 30: 30 KDa, 40: 40 KDa, 50: 50 KDa.

Cholangitis induction in liver of N. aro- and E. coli-infected NOD.B6-Idd10/Idd18 mice

Next we examined the livers of N. aro- and E. coli-infected mice by histological and immunohistochemical staining. Although AMA were detectable as early as 4 weeks after bacterium infection, significant pathological changes in liver were not detected before 19 weeks after either N. aro or E. coli inoculation. However, by 26 weeks following infection, striking portal inflammation accompanied by granuloma formation was present in livers of both N. aro- and E. coli-infected mice, but not in the uninfected control group. Significant biliary cell damage was also detected in both E. coli- and N. aro-infected mice (Fig. 3). To further determine the extent of bile duct damage, we performed immunohistochemical staining for CK19 to visualize biliary epithelial cells among lymphoid aggregation. As shown in Fig. 4, varying degrees of biliary cell damage were found in either E. coli- or N. aro-infected mice, but not in the control mice. In both infected groups, while some bile ducts are nearly intact with mild lymphoid aggregation (blue arrows), in some portal tracts the biliary epithelial cells were completely obliterated (red arrows). These results indicate that E. coli infection is sufficient to induce cholangitis in the biliary disease-prone NOD.B6-Idd10/Idd18 mice.

Figure 3.

Histological analysis of liver sections of Novosphingobium aromaticivorans, Escherichia coli-infected mice and negative control mice at 26 weeks post-immunization. (a) Striking portal inflammation accompanied by granuloma formation were present in livers of both N. aro- and E. coli-infected mice, but not in the uninfected control group. Significant biliary cell damage was also detected in both E. coli- and N. aro-infected mice (blue arrow). (b) Port inflammation score and bile duct damage score of N, aro, E. coli-infected mice and negative control mice at 26 weeks post-immunization. Moderate to severe inflammatory cells infiltration is observed around damaged bile ducts [chronic non-suppurative destructive cholangitis (CNSDC)-like biliary damage] and portal tracts. Portal inflammation and bile duct damage were examined in individual animals at 26 weeks, three to 10 animals from each group were examined (N. aro n = 10, E. coli n = 10, phosphate-buffered saline (PBS) only n = 3). Scored as: 0 = normal, no inflammation (or bile duct damage); 1 = minimal inflammation (or bile duct damage); 2 = mild inflammation (or bile duct damage); 3 = moderate inflammation (or bile duct damage); and 4 = severe inflammation (or bile duct damage).

Figure 4.

Cytokeratin (CK)19 staining for the biliary epithelial cells and bile ductules cells in liver sections of Novosphingobium aromaticivorans, Escherichia coli-infected mice and negative control mice at 26 weeks post-immunization. In intact portal tract, biliary epithelial cells and proliferating bile ductules expressed CK19. In one portal tract of N. aro group mice, bile duct is nearly intact with mild lymphoid aggregation (a, blue arrow), while bile duct was lost in neighbouring portal tract (b, red arrow). In portal tract of E. coli-infected mice, bile duct was destroyed by lymphocytes. Only a few epithelial cells of the E. coli group mice remained among marked lymphoid aggregation (c, red arrow). Biliary epithelial cells of E. coli group mice was completely lost. In one portal tract of E. coli group mice, bile duct is nearly intact with mild lymphoid aggregation (c, blue arrow), while bile duct was destroyed in neighbouring portal tract (d, red arrow). CK19 is useful marker to highlight control biliary epithelial cells among lymphoid aggregation (e,f, blue arrow).

NK T cell antigen content in bacterial sonicates

We have previously used an antigen-presenting cell (APC)-free assay to identify microbes that have antigens for NK T cells [38, 39]. In this assay, microwells are coated with soluble mouse CD1d molecules and incubated either with antigen preparations or total bacterial sonicates. The plates are then cultured with NK T cell hybridomas and interleukin (IL)-2 release, which provides a bioassay for T cell antigen receptor engagement, was quantitated. As can be seen in Fig. 5, sonicates of S. yanoikuyae, which are known to have glycosphingolipid antigens for NK T cells [40], produced IL-2 release from several NK T cell hybridomas. By contrast, E. coli sonicates, which do not have such antigens, did not produce hybridoma IL-2 release. Although related to Sphingomonas spp., N. aro sonicates also did not produce IL-2 secretion by NK T cells. Therefore, it is unlikely that N. aro has significant quantities of a glycolipid antigen capable of activating NK T cells.

Figure 5.

Antigen-dependent activation of natural killer (NK) T cells by Sphingomonas yanoikuyae but not Novosphingobium aromaticivorans or Escherichia coli lysate. The indicated dilutions of sonicate, vehicle and 6 ng alpha-galactosylceramide (α-GalCer) were incubated in wells coated with mCD1d. Stimulation of Va14i NK T hybridomas, 1·2 (a) or 2C12 (b), was determined by the production of interleukin (IL)-2 in the culture supernatant. The error bars indicate the standard error of the mean of triplicate measurements and the data shown are representative of three separate experiments.

Our data also indicate that exposure to N. aro does not induce cholangitis by a unique NK T activating mechanism and we suggest that previous data were probably secondary to molecular mimicry. The challenge for researchers would be to identify genetically at-risk hosts and determine the extent of other secondary factors that may also contribute, perhaps concurrently with microbial infections, to the aetiology of PBC.


In PBC, the highly orchestrated immunological events around the biliary tract, epitope specificity of B and T cells and the homology between the lipoyl domain in microorganisms and humans lend support to the thesis of microbial aetiology. The association of positive serological AMA in PBC patients with recurrent UTIs suggest a bacteria aetiology in PBC [7]. The hypothesis that E. coli is a cause of PBC was first proposed in 1984, based on the higher prevalence of this bacterium in women in PBC when compared with age-matched women with other chronic liver diseases [13]. More recently, Varyani et al. reported that recurrent UTIs are present within 1 year prior to the diagnosis of in 29% of patients in PBC compared to 17% of non-PBC chronic liver disease controls [14]. This hypothesis is also supported by the demonstration of T and B cell cross-reactivity between AMA epitopes and E. coli PDC-E2 sequences [24, 27, 41].

The induction of autoimmune diseases is considered to be the result of complex interactions between genetic traits and environmental factors, including microbial infections [42]. The microbial aetiology of PBC is poorly defined and the pathogenic mechanisms of biliary injury in PBC remain largely unknown. Clues indicating a microbial aetiological component to the pathogenesis of PBC were based largely on experimental evidence of B and T cell cross-reactivity between the major mitochondrial autoantigens and their mimicking microbial antigenic epitopes [27, 29, 43-50]. Additional support comes from epidemiological studies, case reports or molecular evidence of the presence of microbial or viral agents in the liver or bile specimens of patients with PBC [13, 29, 50-52]. Several hypothetical mechanisms, such as bystander activation of autoreactive cells, induction of proinflammatory cytokines by microbial antigens and molecular mimicry between the microorganism and the host have been proposed to explain how microbes initiate autoimmunity [9-12]. Among these hypotheses, the theory of molecular mimicry has been addressed rigorously in PBC, which is based on the shared linear amino acid sequences or a conformational fit (for B cell cross-reactivity), or a motif (for T cell cross-reactivity) between a bacterial antigen and human ‘self'-antigen [2, 28, 44, 53, 54]. An immune response directed against the mimicking microbial determinants may cross-react with the self-protein(s), and such autoreactivity may cause injury in targeted cells leading to cell destruction and, ultimately, autoimmune disease. In line with the theory of molecular mimicry, previously reported AMA cross-reactivity between the human PDC-E2 and its microbial counterparts in E. coli, N. aro and Lactobacillus delbrueckii suggested that PBC could be induced by exposure to these bacterial antigens [28, 42, 44, 55]. In addition, the identification of the 16S rRNA gene in livers from patients with PBC suggests that Propionibacterium acnes could be involved in granuloma formation in PBC [56].

There is evidence from studies from our laboratory and others implicating NK T cells, an innate-like T lymphocyte population that responds to glycolipids, in the pathogenesis of PBC [28, 30, 31]. Some Sphingomonas spp. bacteria have glycosphingolipid (GSL) in their cell membrane that are potent antigens for NK T cells. It is likely that related bacteria, such as N. aro, also have GSL in their membrane. Although it is therefore appealing to propose that a uniquely active GSL might be present in N. aro to activate NK T cells leading to PBC pathogenesis, our data suggest that such a strong GSL antigen is not present. Some Sphingomonas spp. GSL are not highly antigenic [57], however, and NK T cells can be activated by cytokines such as IL-12 in the absence of a microbial glycolipid antigen [58]. Therefore, the route to PBC following N. aro and E. coli infections may involve NK T cell activation, independent of microbial glycolipid antigens.

Regarding the N. aro-induced severe PBC-like cholangitis in NOD.B6-Idd10/Idd18 mice, Mohammed et al. [31] suggested that allelic variation of the Cd101 gene, located in the Idd10 region, alters the severity of N. aro-induced liver autoimmunity by regulating the susceptibility to liver disease. Expression of the NOD Cd101 allele induces a more tolerogenic milieu in the liver by promoting regulatory T cell (Treg) responses, whereas expression of the B6 Cd101 allele triggers an overzealous T cell response upon infection with N. aro. The loss of CD101 expression on dendritic cells (DCs) drives the enhanced interferon (IFN)-γ and IL-17 production by T cells and subsequently the induction of liver disease upon N. aro infection. Conversely, intravenous inoculation of two different strains of E. coli (DH5α and ATCC25922) or Salmonella into NOD1101 mice could induce transient mild liver inflammation early after inoculation which resolved within a few weeks [30]. In the current study, we show that E. coli also induced severe cholangitis in NOD.B6-Idd10/Idd18 mice. It has been reported that there are six E. coli peptide sequences that mimic the human PDC-E2 autoepitope with six to eight identical amino acid residues [44], which may also account for the E. coli-induced anti-PDCE2 response in the NOD.B6-Idd10/Idd18 mice. The difference in microflora between animal colonies may also partly account for the discrepancies between this study and others [30, 31]. Although the serological antibody reactivity to PDC-E2 is relatively weak in the E. coli-infected mice when compared to sera from patients with PBC [15] or other models of autoimmune cholangitis, including the dominant negative transforming growth factor (dnTGF)-βRII mice and xenobiotic 2-octynonic acid bovine serum albumin (BSA) conjugate-immunized mice [59, 60], initiation of anti-PDC-E2 during the early stage of E. coli infection is sufficient to break tolerance and lead to PBC-like liver pathology in the E. coli-infected mice. It is also interesting to note that frequent inoculation of Streptococcus intermedius could induce chronic non-suppurative destructive cholangitis and autoantibodies in C57BL/6 and BALB/c but not in C3H/HeJ mice [61, 62]. These animal models of microbial-induced PBC-like liver pathology clearly support the thesis of microbial involvement in the aetiology of PBC.

Our current data support previous clinical studies in suggesting a role of E. coli in human PBC. Hopf et al. [63] reported an association between PBC and the presence of rough-form mutants of E. coli in the patients’ fecal samples. In addition, Butler et al. reported reactivity to PDC-E2 in 52% of sera from patients with chronic UTIs [7, 64]. In the first controlled epidemiological analysis for the relationship between E. coli and PBC, Parikh-Patel et al. showed a positive association between PBC and recurrent UTI [65]. A recent epidemiological study on 1032 PBC patients followed-up in 20 tertiary referral centres in the United States and 1041 demographically matched controls confirms earlier studies indicating a connection of UTI with PBC [66]. The discovery of E. coli infection-triggered autoimmunity and liver pathology warrant further consideration in the elucidation of aetiological mechanisms of autoimmune syndromes and may suggest new and simpler ways to diagnose and treat these debilitating diseases. Our data also highlight the importance of microbial infections in autoimmunity either as primary or co-existing secondary inciting events.


This work was supported in part by National Institutes of Health grants DK39588 (M. E. G.) DK067003 (M. E. G.), AI71922 (M. K.) and AI083029 (J. L. V.)


The authors have no financial conflicts of interest.