Potential conflict of interest: Nothing to report.
We recently reported that mice with a T cell–restricted expression of a dominant negative form of transforming growth factor β receptor type II (dnTGFβRII) spontaneously develop autoimmune cholangitis that resembles human primary biliary cirrhosis (PBC), including antimitochondrial antibodies (AMAs) and extensive portal CD4+ and CD8+ lymphocytic infiltrates. On the basis of these data, we performed a series of experiments to determine whether the pathology was secondary to direct dnTGFβRII disruption of the liver and/or alternatively the appearance of autoreactive T cells. First, using dnTGFβRIIRag1−/− mice, we noted a normal hepatic and biliary structure. Hence, we performed a rigorous series of adoptive transfer studies, transferring Ly5.1+ unfractionated spleen cell CD4+ or CD8+ T cells from dnTGFβRII mice into B6/Rag−/− (Ly 5.2) recipients. In unmanipulated dnTGFβRII mice, there was a marked increase in CD4+ and CD8+ T cell biliary infiltrates with AMA. Indeed, B6/Rag−/− recipients of dnTGFβRII unfractionated cells develop features of liver disease similar to PBC, suggesting that splenic loss of self-tolerance alone is sufficient to cause disease in this model and therefore that there is no specific abnormality in the biliary targets required for appearance of disease. More importantly, adoptive transfer of CD8+ but not CD4+ T cells into B6/Rag−/− mice led to liver histopathology remarkably similar to PBC, emphasizing a prominent role for CD8 T cell–mediated pathogenesis. In contrast, B6/Rag−/− recipients of CD4+ T cells from dnTGFβRII mice predominantly developed inflammatory bowel disease associated with higher levels of serum interferon γ and tumor necrosis factor α. Conclusion: These data suggest that in this model of PBC, autoreactive CD8+ cells destroy bile ducts. (HEPATOLOGY 2008.)
Primary biliary cirrhosis (PBC) is a progressive autoimmune cholangiolitis with destruction of intrahepatic bile ducts, cholestasis, and progressive development of fibrosis, cirrhosis, and eventual liver failure.1–3 Serologically, PBC is characterized by the presence of antimitochondrial antibodies (AMAs) to autoantigens of the family of 2-oxo-acid dehydrogenase complexes located in the inner mitochondrial membrane,4, 5 including the pyruvate dehydrogenase E2 complex (PDC-E2), E3-binding protein, ketoglutaric acid dehydrogenase E2 complex (OGDC-E2), and branched-chain 2-oxo-acid dehydrogenase E2 complex (BCOADC-E2).6 AMA is a key diagnostic marker in PBC and can precede disease by several years.7 There are also autoreactive T cells that target mitochondrial autoantigens.8–10 However, the precise nature of the autoimmune effector mechanisms of biliary ductular damage remains unclear, partly because of the inherent difficulty in studying primary cellular events during the early, asymptomatic phase of PBC and the general limitations of human-based research.11–15
We recently reported that mice with a T cell–restricted expression of a dominant negative form of transforming growth factor β receptor type II (dnTGFβRII) spontaneously developed periductular aggregates in the liver with loss of self-tolerance to mitochondrial proteins leading to autoimmune-mediated liver pathology that closely resembles the histopathological features of human PBC, including the development of significant levels of autoantibodies with specificity against PDC-E2, BCOADC-E2, and OGDC-E2 and extensive CD4+ and CD8+ lymphocytic infiltrates in portal tracts associated with biliary destruction.16 We reasoned that the pathology that develops in these mice could be due either to a structural or functional change induced in liver tissue because of disruption of the transforming growth factor β receptor type II (TGFβRII) pathway or alternatively to a breakdown of self-tolerance. Such thoughts prompted us to carry out studies to distinguish between these possibilities. We performed adoptive transfer studies in which unfractionated or highly enriched populations of either CD4+ or CD8+ T cells from dnTGFβRII mice were adoptively transferred into recombinase-deficient (Rag1−/−) mice that lack a diversified B and T cell receptor repertoire. Results from these studies demonstrated that whereas the CD8+ T cell recipient mice demonstrated a significant expansion of donor T cells and portal tract infiltrates, the CD4+ T cell population expanded but did not home to or aggregate within the portal tracts. These data imply that in this murine model of PBC, the CD8+ T cell is the pathogenic effector pathway.
dnTGFβRII mice17 were bred onto the congenic C57BL/6-Ly5.1-Pep3b mice (B6 Ly5.1, bought from The Jackson Laboratory, Bar Harbor, ME) at the University of California animal facility (Davis, CA). Female dnTGFβRII (Ly5.1) mice and control littermates were used at 4 to 22 weeks of age. dnTGFβRII and wild-type littermates were fed sterile rodent Helicobacter Medicated Dosing System (three-drug combination) diets (Bio-Serv, Frenchtown, NJ) and maintained in individually ventilated cages under specific pathogen-free conditions. B6/Rag1−/− mice (Ly5.2) were obtained from The Jackson Laboratory. dnTGFβRIIRag1−/− mice were generated by breeding dnTGFβRII mice with B6/Rag1−/− mice. All studies were performed with approval from the University of California Animal Care and Use Committee.
Isolation and Flow Cytometric Analysis of Liver T Cells.
dnTGFβRII and control B6 mice were sacrificed at serial ages, and liver lymphoid cells were isolated as described previously.18 Briefly, phosphate-buffered saline (PBS)–perfused livers were passed through a 100-μm nylon cell strainer (BD Bioscience, Bedford, MA). Lymphocytes were separated from hepatocytes by centrifugation with Histopaque-1.077 (Sigma-Aldrich, St. Louis, MO). Single cell suspensions were washed, and viability was confirmed using trypan blue exclusion. For flow cytometry, Fc receptors were blocked by incubation with the 2.4G2 monoclonal antibody (mAb; eBioscience, San Diego, CA) and cells stained with fluorochrome-conjugated antibodies (Abs) specific for CD3, CD4, CD8, NK1.1, and CD19 (eBioscience) at 4°C in PBS/0.2% bovine serum albumin for 30 minutes. NK1.1 and CD19 Abs were used to exclude natural killer T cells and B cells, respectively. Stained cells were subjected to multiple-color flow analyses using a FACScan flow cytometer updated by Cytec Development (Fremont, CA) to allow for 5-color analysis.
Adoptive Cell Transfer.
Spleen cells were collected from 10- to 12-week-old dnTGFβRII (Ly5.1) or B6 Ly5.1 control mice and used for transfer. For T cell adoptive transfer, the CD4+ or CD8+ T cells were purified by positive selection with CD4 or CD8 microbeads (Miltenyi Biotec, Auburn, CA). Eight- to 10-week-old female B6/Rag1−/− (Ly5.2) mice were used as recipients. Body weight and survival of recipients were measured at serial time points post-transplantation. The numbers of donor-derived cells (Ly5.1+Ly5.2−) in the liver and spleen were calculated on the basis of flow cytometric analysis of stained cells with fluorochrome-conjugated Abs specific for Ly5.1 or Ly5.2 (BioLegend, San Diego, CA).19 The tissue sections from a variety of organs were prepared for routine hematoxylin and eosin (H&E) staining (hematoxylin from DakoCytomation, Glostrup, Denmark, and eosin from American Master Tech Scientific, Lodi, CA).
The liver was removed and immediately fixed with a 10% solution of formalin overnight at room temperature (RT). After deparaffinization, sections were incubated in a Decloaking Chamber (Biocare Medical, Concord, CA; set point 1 123°C for 2 minutes, set point 2 85°C for 10 s, set point limit 10°C) and soaked in a 3% H2O2 methanol solution for 5 minutes, then 15 minutes in a 1× Universal blocking solution (BioGenex, San Ramon, CA), and 20 minutes in 10% goat serum to prevent nonspecific staining. Rat mAbs against mouse CD4 (1/200 dilution) and CD8 (1/200 dilution) were applied for 1 hour at RT in a moist chamber. After three washes with 0.1% Tween 20 in PBS for 5 minutes, horseradish peroxidase–conjugated polyclonal rabbit anti-rat immunoglobulin (DakoCytomation; 1/100) was applied as secondary Ab for 1 hour at RT in a moist chamber. After three washes with 0.1% Tween 20 in PBS, the sections were developed with 3,3′-diaminobenzidine (DakoCytomation) and counterstained with Mayer's hematoxylin (DakoCytomation).
Enzyme-Linked Immunosorbent Assay (ELISA) for AMA.
Serum anti–PDC-E2 (AMA) was measured by an ELISA with purified recombinant PDC-E2.20 Briefly, PDC-E2 at 10 μg/mL in carbonate buffer (pH 9.6) was coated onto 96-well ELISA plates at 4°C overnight, washed 5 times with PBS containing 0.05% Tween-20 (FisherBiotech, Fair Lawn, NJ), and blocked with 3% skim milk in PBS for 30 minutes. One hundred microliters of the diluted sera to be tested (1:100) was added for 1 hour at RT, and plates were rewashed. One hundred microliters of horseradish peroxidase–conjugated anti-mouse immunoglobulin M, G, or A (1:2000; Zymed, San Francisco, CA) was added to each well for 1 hour at RT, and the microtiter wells were thence rewashed. Immunoreactivity was detected by the measurement of the optical density (OD) at 450 nm after exposure for 15 minutes to 100 μL of ABTS peroxidase substrate (KPL, Gaithersburg, MD).20
Measurement of Serum Cytokines.
Serum samples were collected from recipient B6/Rag1−/− mice at serial time points after T cell transfer and stored at −70°C until assayed. Concentrations of interferon γ (IFN-γ) and tumor necrosis factor α (TNF-α) were measured simultaneously with a mouse inflammatory cytometric bead array (CBA) kit (BD Biosciences); samples were analyzed on a FACScan flow cytometer (BD Immunocytometry Systems) using CBA software (BD Biosciences).
An unpaired, two-tailed Student t test was used for comparison of the weight, cell number, and frequencies of cell surface markers and cytokine production in sera. Differences were considered statistically significant when P < 0.05.
CD4+ and CD8+ T Cell Expansion in the Liver of dnTGFβRII Mice.
We first analyzed the characteristics of T cell liver infiltrates in the liver of dnTGFβRII mice compared with B6 littermates. Total CD3+ T cell counts in the liver were significantly increased at 8 weeks of age with maximal levels occurring at week 12 (Fig. 1A). Although both CD4+ and CD8+ T cell populations were expanded in the liver of dnTGFβRII mice, the CD8/CD4 ratio was significantly higher in dnTGFβRII mice from 12 weeks of age (Fig. 1A). The cellular infiltration in the liver histologically also peaked at 12 weeks (data not shown). Hence, infiltration of T cells and an increase in the CD8/CD4 ratio appear to be key components in dnTGFβRII liver disease. Since immunohistology demonstrated both CD4+ and CD8+ T cell infiltrates in affected portal tracts (Fig. 1B), we performed adoptive transfer studies to identify the role of each of these cell lineages in bile duct cell damage.
Transfer of dnTGFβRII Splenocytes Results in Disease in Rag1−/− Recipient Mice.
A standard adoptive transfer protocol was used to determine whether PBC-like changes seen in the liver of dnTGFβRII mice could be transferred. Thus, 1 × 107 splenocytes from dnTGFβRII (Ly5.1) or control B6 Ly5.1 mice were intravenously transferred into B6/Rag1−/− (Ly5.2) mice. As shown in Fig. 2A, the recipients began to lose body weight 4 weeks after transfer and developed diarrhea (as do dnTGFβRII mice that are not raised under specific pathogen-free conditions),17 which was associated with a significant increase in lymphocyte counts in the liver of recipients of dnTGFβRII cells as compared with recipients of control B6 Ly5.1 mice (Fig. 2B). The majority of cells infiltrating the liver were donor-derived Ly5.1+ CD3+ T cells (data not shown). On histological assessment, there was a marked lymphoid cell infiltration that surrounded the portal tracts and central veins in the liver, accompanied by inflammatory infiltrates in the colon and lungs (Fig. 3). Because spontaneous production of AMAs was detected in sera of dnTGFβRII mice,16 we also measured autoantibody levels to PDC-E2 in the sera of these spleen cell recipient mice. Serum AMAs of the dnTGFβRII spleen cell recipients were positive 4 weeks after transfer (Fig. 2C). Thus, a PBC-like disease could be transferred into B6/Rag1−/− recipients by adoptive transfer of unfractionated dnTGFβRII splenocytes.
We next backcrossed the dnTGFβRII mice to B6/Rag1−/− mice that lack a repertoire of mature T and B cells. dnTGFβRIIRag1−/− mice were studied at 20 weeks of age, and tissues were examined histologically (Fig. 4). No lymphocyte infiltration or inflammation was evident in the liver, intestine, or lung. The normal liver histology, similar to that of B6/Rag1−/− mice, supports our view that autoreactive T cells cause bile duct damage in the dnTGFβRII mice; that is, there is no intrinsic liver defect.
Transfer of CD4+ and CD8+ Cells into Rag1−/− Recipients.
To determine which of the T cell lineages generated PBC-like liver disease in mice, 1 × 106 CD4+ or CD8+ T cells from the spleens of dnTGFβRII (Ly5.1) mice were injected intravenously into B6/Rag1−/− (Ly5.2) mice. Mice that were recipients of CD4+ T cells displayed wasting starting at 4 weeks after transfer (Fig. 5A) similar to that of mice receiving unfractionated spleen cells from dnTGFβRII mice (Fig. 2). These CD4+ T cell recipient mice had 100% mortality by 16 weeks following cell transfer, whereas mice that received CD8+ T cells survived more than 30 weeks (Fig. 5B). To compare the differences in the cellular infiltrates between mice that received CD4+ or CD8+ T cells, a nested group of recipient mice was killed at 8 weeks. As shown in Fig. 5A, the body weight of recipients of CD4+ T cells fell significantly, and these animals became ill at this time. However, these effects on the animals were not due to demonstrable liver pathology. Rather, flow cytometric analysis demonstrated that the proportion and absolute number of intrahepatic T lymphocytes were significantly higher in recipients of CD8+ T cells compared to CD4+ T cell recipients (Fig. 5C and data not shown). Recipients of CD4+ T cells develop more severe intestinal disease and die without permitting longer term observations. Recipient lungs appear similar in both CD4+ and CD8+ T cell groups.
Since both CD4+ and CD8+ T cells can produce inflammatory cytokines, we measured the production of inflammatory cytokines in the serum of recipients. Both IFN-γ production and TNF-α production were significantly higher in mice that were recipients of CD4+ T cells than in mice that were recipients of CD8+ T cells (Fig. 6), and this indicated that dnTGFβRII CD4+ T cells were more potent in inducing systemic inflammation following adoptive transfer.
Activated CD8+ T Cells Induce PBC-Like Changes in the Liver.
The liver histology of human PBC is characterized by CD4+ and CD8+ T cell infiltrates associated with selective small bile duct destruction. We addressed the role of T cells in induction of liver disease by analysis of the cellular infiltrates in the livers of Rag1−/− mice that were recipients of CD4+ or CD8+ T cells from dnTGFβRII mice. Immunohistological staining of the T cell infiltrates, shown in Fig. 7, indicates that recipients of CD8+ T cells had more severe inflammatory cell infiltrates in the liver than recipients of CD4+ T cells (Fig. 7A,B). CD4+ T cell infiltrates primarily surrounded central veins and hepatic lobules, and focal necrosis and granulomas were found within the intralobular area (Fig. 7E). In contrast, recipients of CD8+ T cells demonstrate that most CD8+ T cell infiltrates surrounded portal tracts but not central veins (Fig. 7F). Although CD4+ T cell infiltrates were evident in some portal tract areas in the CD4+ T cell recipient mice, small biliary ducts were not their primary targets (Fig. 7C). In contrast, moderate to severe lymphoid cell infiltrates were detected within the portal tracts 8 weeks after CD8+ T cell transfer. The small bile ducts in portal tracts that entrapped the infiltrating CD8+ T cells were often seen to be damaged and undergoing destruction (Fig. 7B). In some portal tracts, bile ducts had disappeared, and significant granuloma had developed (Fig. 7D). Interestingly, histological examination of the colon and lungs demonstrated substantial lymphocytic infiltrates in recipients of CD4+ T cells. In contrast, recipients of CD8+ T cells had only mild lymphoid cell infiltrates in these organs (Fig. 8).
We have previously demonstrated that dnTGFβRII mice deprived of regulatory TGF-β signaling restricted to T cells develop liver lesions similar to those of human PBC16; this is indicative of T cells being the likely effectors of autoimmune cholangitis. Here we have transferred splenic lymphocytes isolated from dnTGFβRII mice into Rag1−/− mice; the recipient mice rapidly develop a liver pathology similar to that seen in the dnTGFβRII mice. Further, in dnTGFβRIIRag1−/− mice that lack mature T and B cells, there was no evidence of liver pathology indicating the need for immunocompetent T or B cells or both for the PBC-like phenotype seen in dnTGFβRII mice. Finally, after adoptive transfer of CD4+ and CD8+ T cells from dnTGFβRII mice, both cell types proliferated in the liver of recipients, but only the CD8+ T cell population underwent significant expansion, and it was the major constituent of the portal tract infiltrates. Histologically, the bile duct epithelial cells appeared to be the target of the transferred CD8+ T cells. Accordingly, and on the basis of these observations, we suggest that impairment of the TGF-β signaling pathway in the CD8+ cytotoxic T cell population is essential for the development of autoimmune biliary epithelial cell damage in this model. In contrast, autoreactive CD4+ T cells in this model are primarily responsible, via the production of proinflammatory cytokines, for systemic inflammation. In our cell transfer experiments, we addressed the role of CD8+ T cells by studying the histopathology of recipient lung, intestine, and liver, and as described herein, the biliary changes were specific for CD8+ T cells.
Previous studies in humans implicate both CD4+ and CD8+ T cell reactivities in the destruction of biliary epithelial cells.2, 6, 21, 22 Hitherto, the elucidation of early events in the induction of periductular inflammation and autoimmunity in PBC has been hampered by the lack of suitable animal models.23-25 Herein, using our unique model,16 we found that populations of both CD4+ and CD8+ T cells were expanded in the liver, but with a CD8/CD4 ratio biased in favor of CD8. This finding is interesting in view of previous reports of an increase in CD8+ precursors of cytotoxic T lymphocytes (CTLs) in the blood of patients in early stages of PBC compared with advanced stages of PBC and a 10- to 15-fold increase of autoantigen-specific CTLs within the liver compared with peripheral blood.8 The increased number of CD8+ CTLs in dnTGFβRII mice is in agreement with findings in PBC. In addition, adoptively transferred CD4+ T cells induced severe inflammation in the intestine and lungs, whereas adoptively transferred CD8+ T cells induced inflammation predominantly in periductular regions of the liver that was histologically similar to what is observed in human PBC. Hence, adoptive transfer of CD8+ T cells to Rag1−/− mice not only provides an informative model for human PBC but also has implications for understanding TGF-β signaling and autoimmunity in general and for developing novel immunotherapies to prevent target organ damage.26, 27 We should emphasize the ready detection of granulomas within the portal tracts in mice that received adoptive transfer of CD8+ T cells. However, at least within the time period of 16 weeks studied herein, we did not find evidence of fibrosis. Future studies are focused on longer term survival with additional serial histopathology in order to identify the natural history and progression of disease. We should also note that the biliary system of the mouse is significantly different from that of humans and that the distinction between small and large bile ducts within the mouse is not as clear. At this point, we cannot clearly determine whether the biliary duct lesions in these mice are specific for small bile ducts versus large bile ducts.
Our hypothesis is that autoreactive dnTGFβRII CD8+ T cells have their preferred target cells in the liver. Thus, histologically, transferred CD4+ T cells induced lymphoid cell infiltration that was mild in the liver but more severe in the gut and lungs. Moreover, in regard to the liver location, CD4+ T cells were noted predominantly around the central vein and within the liver lobules (Fig. 7); the bile ducts were relatively spared. In contrast, CD8+ T cells infiltrated the portal tract and specifically targeted bile duct cells (Fig. 7). Comparably, in another autoimmune setting, that of diabetic insulitis of nonobese diabetic mice, autoaggressive CD8+ T cells destroy the insulin-secreting β cells of pancreatic islets.21, 28, 29 In contrast, autoaggressive CD4+ T cells are the effectors of inflammatory bowel disease.30, 31 Furthermore, for TGFβ1−/− mice on a major histocompatibility complex (MHC) class I–deficient background (β2-microglobulin–null mice), survival is much longer than for TGFβ1−/− mice on an MHC class II–deficient background.32 Thus, abnormal expression and presentation of self-antigens by MHC class II molecules to autoimmune CD4+ T cells may be more potent effectors of systemic inflammation and autoimmunity, likely by the release of proinflammatory cytokines, and CD8+ T cells may be the more potent effectors of tissue-specific cytotoxicity.33, 34
Previous work has demonstrated that while the levels of invariant natural killer T (iNKT) cells in the peripheral blood of PBC patients is lower than those of healthy individuals, the frequency of iNKT cells in PBC liver is significantly higher than in blood. In addition, the frequency of iNKT cells in the liver is significantly higher in PBC than controls.35, 36 These results imply that in addition to CD4 and CD8 T cells, iNKT cells may play a contributory role in the liver injury of PBC. To confirm this and using our dnTGFβRII murine model, we recently noted that iNKT cells exacerbate liver injury in the early stages of murine PBC by secreting IFN-γ, which leads to NK cells and subsequently T and B lymphocyte activation and recruitment in liver.37 Future studies in this model include multiple areas but, on the basis of these observations, will use CD8+ knockout mice as another means of studying the pathways of immune injury. Such studies will also allow us to address the mechanisms of CD8+ T cell cytotoxicity and the relationships with Fas/Fas ligand, the perforin/Granzyme B pathway, and interactions with other cell populations, including the role of innate immunity. Finally, we should emphasize that the use of these animals allows studies of antigen-specific CD8 T cell frequencies as well as their homing characteristics, research possibilities that are not available with human tissue. Clearly, there are multiple similarities but also differences in this murine model of autoimmune cholangitis, and further work will allow us to define the extent to which these data can be extrapolated to humans for potential therapeutic intervention.