AUTOIMMUNE, CHOLESTATIC AND BILIARY DISEASE

Loss of Tolerance in C57BL/6 Mice to the AutoantigenE2 Subunit of Pyruvate Dehydrogenase by a

Xenobiotic with Ensuing Biliary Ductular DiseaseKanji Wakabayashi,1,2 Zhe-Xiong Lian,1 Patrick S.C. Leung,1 Yuki Moritoki,1 Koichi Tsuneyama,3 Mark J. Kurth,4

Kit S. Lam,5 Katsunori Yoshida,1 Guo-Xiang Yang,1 Toshifumi Hibi,2 Aftab A. Ansari,6 William M. Ridgway,7

Ross L. Coppel,8 Ian R. Mackay,9 and M. Eric Gershwin1

There have been important advances in defining effector mechanisms for several human auto-immune diseases. However, for most human autoimmune diseases, the induction stage is less welldefined and there are very few clues on etiology. Our laboratory has focused on defining themolecular basis of autoantibody recognition and epitope modification in primary biliary cirrho-sis (PBC). Our work has demonstrated that antibodies to mitochondria, the hallmark of disease,are directed against a very conserved site of pyruvate dehydrogenase, the E2 subunit of pyruvatedehydrogenase (PDC-E2). We have also demonstrated that several chemical xenobiotics, chosenbased on quantitative structural activity relationship analysis and rigorous epitope analysis, whencoupled to the lysine residue that normally binds the lipoic acid cofactor of PDC-E2, reacts aswell or better to PBC sera than native autoantigen. In the present studies, we immunizedC57BL/6 mice with one such xenobiotic, 2-octynoic acid, coupled to bovine serum albumin andwe followed the mice for 24 weeks. Animals were studied for appearance of histologic lesions aswell as appearance of antibodies to PDC-E2, serum levels of tumor necrosis factor–� and inter-feron-�, and splenic and liver lymphoid phenotyping by flow cytometry. Mice immunized with2-octynoic acid manifest autoimmune cholangitis, typical mitochondrial autoantibodies, in-creased liver lymphoid cell numbers, an increase in CD8� liver infiltrating cells, particularlyCD8� T cells that coexpress CD44, and finally an elevation of serum tumor necrosis factor–�and interferon-�. Conclusion: these data provide a persuasive argument in favor of an environ-mental origin for human PBC. (HEPATOLOGY 2008;48:531-540.)

There have been important advances in definingmechanisms required for the effector stage of var-ious human autoimmune diseases, including the

autoreactive populations involved in damage to the biliary

epithelial cell (BEC) in primary biliary cirrhosis (PBC).On the other hand, for most autoimmune diseases, evenat the induction stage, contributing genetic influencesand environmental contributions are less well under-

Abbreviations: 2OA, 2-octynoic acid; 6BH, 6-bromohexanoate; AMA, antimitochondrial antibody; BEC, biliary epithelial cell; BSA, bovine serum albumin; CFA,Complete Freund’s Adjuvant; ELISA, enzyme-linked immunosorbent assay; HRP, horseradish peroxidase; IFN-�, interferon �; Ig, immunoglobulin; IL, interleukin; mAb,monoclonal antibody; NHS, N-hydroxysuccinimide; PBC, primary biliary cirrhosis, PBS, phosphate-buffered saline; PDC-E2, E2 subunit of pyruvate dehydrogenase; RT,room temperature; Th1, T helper 1; Th2, T helper 2; TNF-� tumor necrosis factor �.

From the 1Division of Rheumatology, Allergy and Clinical Immunology, University of California at Davis, Davis, CA; 2Department of Internal Medicine, KeioUniversity School of Medicine, Tokyo, Japan; 3Department of Pathology (I), University of Toyama School of Medicine, Toyama, Japan; 4Department of Chemistry,University of California at Davis, Davis, CA; 5University of California at Davis Cancer Center, Division of Hematology and Oncology, University of California at Davis,Sacramento, CA; 6Department of Pathology, Emory University School of Medicine, Atlanta, GA; 7Division of Rheumatology and Immunology, University of PittsburghSchool of Medicine, Pittsburgh, PA; 8Department of Microbiology, Monash University, Clayton, Australia; 9Department of Biochemistry and Molecular Biology, MonashUniversity, Clayton, Australia.

Received March 17, 2008; accepted April 15, 2008.Supported by the National Institutes of Health, grants DK39588 and DK067003.Address reprint requests to: M. Eric Gershwin, M.D., Division of Rheumatology, Allergy and Clinical Immunology, University of California at Davis School of Medicine,

451 Health Sciences Drive, Suite 6510, Davis, CA 95616. E-mail: megershwin@ucdavis.edu; fax: 530-752-4669.Copyright © 2008 by the American Association for the Study of Liver Diseases.Published online in Wiley InterScience (www.interscience.wiley.com).DOI 10.1002/hep.22390Potential conflict of interest: Nothing to report.

531

stood. Our laboratory has been studying the putative eti-ologies for the initiation of PBC for the last decade withthe objective to establish a small animal model that wouldfacilitate studies of the mechanisms that are the basis ofthis disease. During these studies, our laboratory has iden-tified several potential environmental initiators, particu-larly bacteria1-3 and chemical xenobiotics.4-11 In agenetically susceptible host, these initiators can disruptimmune tolerance to the dominant autoantigen of PBC,the mitochondrially localized pyruvate dehydrogenaseE-2 subunit (PDC-E2), which sets in train a multilineagecellular and humoral immune response against this andrelated autoantigens. This same multilineage immune re-sponse then becomes the pathogenic and effector agent byvirtue of the unique immunobiology of the BEC, as dis-cussed and presented elsewhere in detail.6 The fine map-ping of the epitopes recognized by the humoral andcellular autoimmune response against PDC-E2 has led tothe observation that these sites are highly conserved mo-lecular sequences flanking lipoic acid binding sites and theresponses to these are among the most directed and spe-cific responses in human autoimmunity, and includes re-activity of not only B cells, but also CD4 and CD8 T cellresponses.6

These data encouraged us to screen PBC sera againstPDC-E2 molecules that carried chemical modifications ofeither the lysine residue (173K) or the lipoic acid cofactorthat attaches to 173K. We first demonstrated that severalchemical xenobiotics, when coupled via 173K to the dom-inant autoantigenic peptide of PDC-E2, led to the forma-tion of a neoantigen that surprisingly reacted at least aswell or even better with PBC sera than did the nativeautoantigen;4,9,11 in other words, anti-PDC-E2 antibod-ies from patients with PBC could recognize xenobioticmodified PDC-E2 peptides that mimic lipoic acid, andsuch recognition often includes a higher titer reactivitythan to the native autoantigen. Second, we have recentlyperformed more detailed quantitative structure-activityrelationship analysis and have identified 2-octynoic acid(2OA) as an even better xenobiotic candidate for anti-genic modification of the PDC-E2 peptide.4,11 Third, weascertained that immunization of rabbits with a particularxenobiotic chemical, 6-bromohexanoate (6BH), coupledto bovine serum albumin (BSA), could break tolerance toPDC-E2 as judged by production of antimitochondrialantibodies (AMA) as seen in patients with PBC.5,8

Fourth, and finally, we recently demonstrated that immu-nization with the compound 6BH, coupled to BSA, led tothe development of histological lesions typical of autoim-mune cholangitis with the concurrent appearance ofAMA in guinea pigs, albeit with a long latency of some 18months.7

In the present studies, we immunized mice with 2OAcoupled to BSA and we observed the appearance of anti-PDC-E2 together with histological lesions typical of au-toimmune cholangitis. However, of interest is our findingthat immunization of mice with 2OA-BSA leads to auto-immune PBC-like disease following a latency periodwithin just a month, rather than the latency of nearly 18months for guinea pigs immunized with 6BH. These dataillustrate the capacity of chemical xenobiotics when con-jugated to a potential autoepitope site to induce a PBC-like disease model in mice. Also the data provide apersuasive argument in favor of an environmental (xeno-biotic) origin for human PBC.

Materials and MethodsMurine Immunization

Female C57BL/6J (B6) mice (n � 13 per each group)at 8 to 9 weeks of age were obtained from The JacksonLaboratory (Bar Harbor, ME) and maintained in venti-lated cages under specific pathogen-free conditions. Eachmouse was immunized with a mixture of 2-octynoic acidBSA conjugate (2OA-BSA; 100 �g/25�L) intraperitone-ally in Complete Freund’s Adjuvant (CFA; Sigma-Al-drich, St. Louis, MO) containing 10 mg/mL ofMycobacterium tuberculosis strain H37Ra and subse-quently boosted every 2 weeks with 2OA-BSA and In-complete Freund’s Adjuvant (IFA; Sigma-Aldrich). Ascontrols, a comparable numbers of female B6 mice wereimmunized with BSA coupled to CFA (100 �g/mouse)and boosted using the identical protocol. Sera were col-lected before immunization and thereafter every 4 weeksprior to xenobiotic immunization for definition of anti-PDC-E2. Sera were also analyzed for the cytokines, tumornecrosis factor–� (TNF-�) and interferon-� (IFN-�), asnoted below. Animals were sacrificed 12 or 24 weeks afterimmunization, liver, salivary gland, thyroid, lung, kidney,small intestine, and colon tissues were collected for histo-logical evaluation. Livers and spleens were collected at thesame time points and studied by flow cytometry for cel-lular phenotypes using the protocol below.

Preparation of Immunogen2OA was purchased from Sigma-Aldrich and conju-

gated to BSA as follows. First, 2OA (NHS) was dissolvedin dry dimethyl ether. N-hydroxysuccinimide was thenadded and the solution was cooled to 0°C and stirred for20 minutes. Dicyclohexylcarbodiimide was then addedand the mixture was allowed to warm to ambient temper-ature over night. The solution was filtered, concentratedby rotoevaporation under reduced pressure, redissolvedwith ethyl ether, washed with water, NaHCO3 (1 M),brine, dried under magnesium sulfate, filtered, and con-

532 WAKABAYASHI ET AL. HEPATOLOGY, August 2008

centrated. The product was then purified using flash chro-matography (30% ethyl acetate/hexane). NHS-activated2OA was dissolved in dimethyl sulfoxide and then coupledto the lysine residues of BSA (EMD Chemicals, Gibbstown,NJ) as described8 (Fig. 1). The solution was allowed to reactfor 3 hours followed by dialysis with phosphate-bufferedsaline (PBS).

Detection of Anti-PDC-E2

Immunoblotting. A total of 10 �g of the purified re-combinant PDC-E2 was resolved on a 10% Novex minigel (Invitrogen, Carlsbad, CA). The proteins were electro-blotted to nitrocellulose membrane (Whatman, Dassel,Germany) and cut into 2-mm strips. Each strip was thenblocked for 1 hour at room temperature (RT) with caseinblocker (Piece Biotechnology, Rockford, IL). The stripswere incubated for 1 hour with either mouse sera (1:200)or an AMA-positive human PBC control serum (1:1,000)or a mouse anti-PDC-E2 monoclonal antibody (mAb),2H412,13 (1:50) diluted in PBS containing 0.05%Tween-20 (PBS-T) and 0.05% casein blocker (FisherBiotech, Fair Lawn, NJ). These strips were then washedsix times for 5 minutes in PBS-T. Secondary antibody waseither horseradish peroxidase (HRP)-conjugated goat an-

ti-mouse immunoglobulin (Ig)G, IgA, IgM or HRP-con-jugated goat anti-human IgG, IgA, IgM (Zymed, SanFrancisco, CA) diluted 1:10,000 in the same solution asthe primary antibody. These strips were exposed for1 hour at RT to the respective secondary antibodiesfollowed by six PBS-T wash cycles and exposed for 5minutes to chemiluminescent substrate (Pierce Biotech-nology). Reactive components were visualized with aFluor Tech 8900 gel doc system (Alpha Innotech, SanLeandro, CA) equipped with a chemiluminescent filter.

Enzyme-Linked Immunosorbent Assay. This assaywas performed as described.14 Purified recombinantPDC-E2 antigen at 10 �g/mL in carbonate buffer (pH9.6) was coated onto 96-well enzyme-linked immunosor-bent assay (ELISA) plates at 4°C overnight, washed fivetimes with PBS-T, and blocked with 3% skim milk inPBS for 1 hour. A total of 100 �L of the diluted sera(1:500) was added to the wells and incubated for 1 hour atRT followed by PBS-T washes, then 100 �L of HRP-conjugated anti-mouse IgG, IgA or IgM (1:2,000)(Zymed) was added to each well for 1 hour at RT followedby another set of PBS-T washes. Immunoreactivity wasdetermined by measuring the optical density at 450 nmwith an Emax precision microplate reader (Molecular De-vices, Sunnyvale, CA) after incubation with 100 �L ofTMB substrate (Becton-Dickinson [BD] Biosciences,San Jose, CA) for 30 minutes.

CytokinesSerum TNF-� and IFN-� levels were measured using

the mouse inflammatory cytometric bead array (CBA) kit(Mouse T helper 1 [Th1]/T helper 2 [Th2] Cytokine Kitand Mouse Inflammation Kit; BD Biosciences) as de-scribed.15 Samples were diluted 1:5 with Assay Diluentincluded with the kit. Following the preparation, sampleswere analyzed using a FACScan flow cytometer (BD Im-munocytometry System, San Jose, CA) and BD CBASoftware.

Microscopy of Tissue Sections

Light Microscopy. Immediately after sacrifice, tissue(liver, thyroid, salivary gland, lung, kidney, small intes-tine, and colon) were harvested, fixed in 10% bufferedformalin, embedded in paraffin, and cut into 5-�m sec-tions. Liver sections were deparaffinized, stained with he-matoxylin and eosin, and evaluated under lightmicroscopy.

Immunohistochemistry. Phenotypic analysis of the in-trahepatic cell infiltrates was performed utilizing rat anti-mouse CD4 and anti-mouse CD8 antibodies (1:20 dilution;BioLegend, San Diego, CA). After deparaffinization, sec-tions were soaked in TRS buffer (pH 9.0 for CD4, pH 6.1

Fig. 1. Structures of lipoic acid and 2OA-BSA. 2OA was conjugatedwith bovine serum albumin (BSA) and used for immunization. “Z”represents any lysine residue of BSA. NHS, N-hydroxysuccinimide.

HEPATOLOGY, Vol. 48, No. 2, 2008 WAKABAYASHI ET AL. 533

for CD8; Dako Cytomation, Carpinteria, CA) containing0.1% Triton X-100 (Sigma-Aldrich) and placed in a De-cloaking Chamber (Biocare Medical, Concord, CA) (Step 1at 123°C for 2 minutes, Step 2 85°C for 10 seconds, SP limit10°C) and then soaked in Universal blocking solution (Bio-Genex, San Roman, CA) for 15 minutes. Primary antibodieswere applied overnight at 4°C in a moist chamber. Afterthree washes with PBS for 5 minutes, sections were soaked in3% H2O2 methanol solution for 5 minutes and then HRP-conjugated polyclonal rabbit anti-rat Ig (1:100; Dako Cyto-mation) was applied as a secondary antibody for 1 hour atRT in a moist chamber. After three washes with PBS, thesections were developed with the DAB peroxidase substratekit (Vector Laboratories, Burlingame, CA) and counter-stained with hematoxylin (Dako Cytomation).

Flow CytometryImmediately after sacrifice, livers and spleens were col-

lected from B6 mice immunized with 2OA-BSA or BSA.Livers were first perfused with PBS containing 0.2% BSA,passed through a nylon mesh, and resuspended in PBS/0.2% BSA. Hepatocytes were removed as pellets after cen-trifugation at 75 � g for 1 minute and the remainder ofthe cells were collected. Spleen tissue was disrupted be-tween two glass slides and suspended in PBS/0.2% BSA.Lymphocytes from suspended liver and spleen cells wereisolated using Accu-Paque (density: 1.086; AccurateChemical & Scientific Corp., Westbury, NY) gradient.After centrifugation, cells at the interface were washedwith PBS/0.2% BSA, and the viability of cells was con-firmed by Trypan Blue dye (Invitrogen) exclusion. Cellpreparations were incubated with mAb 2.4G2 for Fc re-ceptor blocking (BioLegend, San Diego, CA) and thenincubated at 4°C with a combination of fluorochrome-conjugated antibodies, including anti-T cell receptor �fluorescein isothiocyanate (anti-TCR� FITC; eBio-sciences, San Diego, CA), anti-CD4 allophyocyanin(APC)-Cy7 (BioLegend), anti-CD8a phycoerythrin(PE)-Cy5 (eBiosciences), and anti-CD19 Alexa Fluor 647(eBiosciences). Multicolor flow analyses were performedusing a FACScan flow cytometer (BD Biosciences) up-graded by Cytec Development (Fremont, CA) to allowfor five-color analysis. Acquired data were analyzed withCELLQUEST software (BD Biosciences).

Expression of Data and Statistical AnalysisData for groups of mouse sera tested by ELISA for

anti-PDC-E2 responses, and for levels of cytokines arecited as means � standard error of the mean. Statisticalanalysis was performed using an unpaired t test to com-pare data for 2OA-BSA and control mice. P values lessthan 0.05 were considered statistically significant.

Results2OA-BSA-Immunized Mice Developed PDC-E2Specific IgG, IgA, and IgM

Immunoblotting. Sera from 100% of mice (13/13)immunized with 2OA-BSA (Fig. 1) reacted with PDC-E2at 8 weeks postimmunization. In contrast, none of the 13sera from the control mice immunized with BSA alonehad any detectable reactivity against PDC-E2 (Fig. 2A).

ELISA. At 4 weeks, serum IgG and IgM antibodies toPDC-E2 significantly increased in mice immunized with2OA-BSA (n � 11) as compared to sera from mice im-munized with BSA (n � 10): optical density (mean �standard error of the mean) 0.33 � 0.12 versus 0.04 �0.01, P � 0.0378 for IgG; and 0.08 � 0.01 versus 0.02 �0.01, P � 0.0019 for IgM. At 8 weeks, IgG, IgA, and IgMantibodies to PDC-E2 were all significantly increased inthe sera from mice immunized with 2OA-BSA comparedwith mice immunized with BSA: 1.23 � 0.23 versus0.03 � 0.00, P � 0.0006 for IgG; 0.55 � 0.14 versus0.00 � 0.00, P � 0.0025 for IgA; and 0.27 � 0.06 versus0.02 � 0.00, P � 0.0006 for IgM (Fig. 2B).

Serological TNF-� and IFN-� Were Elevated in2OA-BSA-Immunized Mice

TNF-� and IFN-� were significantly increased in serafrom the mice immunized with 2OA-BSA compared withlevels of mice immunized with BSA from 4 weeks to 12weeks: 79.0 � 13.7 pg/mL versus 19.2 � 7.5 pg/mL, P �0.0021 for TNF-�; and 14.7 � 3.3 pg/mL versus 0.0 �0.0 pg/mL, P � 0.001 for IFN-�, at 4 weeks (Fig. 3). Onthe other hand, there was no difference in the serum levelsof Th2 prototype cytokines interleukin (IL)-4, IL-5, orIL-10 (data not shown), indicating Th1 cytokine bias.

Presence of Lymphoid Infiltration in the Liver andMild Bile Duct Lesion in 2OA-BSA-ImmunizedMice but Not in BSA-Immunized Mice

Light Microscopy. The portal tracts of mice immu-nized with 2OA-BSA contained a significant amount oflymphoid cell infiltrates of lymphoid cells (Fig. 4B-D).Mild infiltration of lymphocytes or mononuclear cellssurrounding damaged bile ducts was frequently observedin portal areas (Fig. 4B,C). These bile duct lesions resem-bled those of chronic nonsuppurative destructive cholan-gitis as seen in human PBC. Furthermore, ductopenia wasobserved, and epithelioid granulomas were scatteredwithin some portal tracts and also in hepatic parenchyma(Fig. 4C,D); these pathological findings resembled thoseof human PBC. There were also small areas of mild focalnecrosis in hepatic parenchyma. No significant changewas observed in central vein areas. On the other hand, no

534 WAKABAYASHI ET AL. HEPATOLOGY, August 2008

detectable cellular infiltrates and/or bile duct destructionwere present and no parenchymal histopathologicalchange was observed in liver tissues from the control mice(Fig. 4A). Portal infiltrates and bile duct damages were

evaluated with our original scoring: 0 for no infiltrate orno bile duct damage, 1 for mild, and 2 for moderate. Mildportal infiltrates and mild bile duct damages were seen inall the tested mice for 2OA-BSA (n � 6) and none ofthem were found in all the mice for BSA (n � 4) (Fig.4G). At 24 weeks we did not observe evidence of eithersteatosis, eosinophilia, or cholestasis. Liver sections frommice at 24 weeks postimmunization showed similar find-ings as mice at 12 weeks. Additional tissues, including thethyroid, salivary gland, lung, kidney, small intestine, andcolon were examined (data not shown). These sectionswere identical and interpreted as normal in both groups ofanimals.

Immunohistochemistry. Portal area cellular infiltratesin the liver tissues of 2OA-BSA-immunized mice con-tained numerous CD4� and CD8� lymphocytes, whichwere not seen in liver tissues from the control BSA-im-munized mice (Fig. 4E,F). Both CD4� and CD8� lym-phocytes were observed in the same portal tracts andCD8� lymphocytes predominated over CD4� lympho-cytes.

Fig. 3. Levels of TNF-� and IFN-� in sera of mice after immunization.The levels of both Th1 cytokines in sera from mice immunized with2OA-BSA versus BSA-immunized mice were markedly increased. **P �0.01, ***P � 0.001.

Fig. 2. Antibody to PDC-E2. (A) Immunoblot at 8 weeks after immuni-zation with 2OA-BSA and BSA. Recombinant PDC-E2 protein was loaded onsodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) andtransferred to a nitrocellulose membrane. All 13 sera from mice immunizedwith 2OA-BSA versus none of 12 sera from BSA-immunized mice showeddetectable reactivity with a component of �70 kDa. Sera from two PBCpatients and a mouse anti-PDC-E2 mAb, 2H4, were included as positivecontrols. (B) Quantification of anti-PDC-E2 in the sera by ELISA at 2-weekintervals after immunization shows significant increases in optical density(O.D.) values after immunization between responder and control mice. *P �0.05, **P � 0.01, ***P � 0.001.

HEPATOLOGY, Vol. 48, No. 2, 2008 WAKABAYASHI ET AL. 535

Disruption of CD4/CD8 Ratio in the Liver of 2OA-BSA-Immunized Mice

The composition of the lymphoid cell populationwithin the liver and spleen of the immunized mice wasanalyzed at 12 weeks postimmunization. The frequencyof CD4� cells were decreased associated with a concom-

itant increase in the frequency of CD8� in liver tissuesfrom the 2OA-BSA-immunized mice as compared withcontrol BSA-immunized mice. The frequency of bothCD4� and CD8� cells were, on the other hand, decreasedin the spleen (Fig. 5A). Therefore CD4/CD8 ratios in the2OA-BSA-immunized mice liver were significantly de-creased: 2.03 � 0.08 versus 12.50 � 1.63, P � 0.0015;whereas there was no difference in the spleen (Fig. 5B).

The total lymphoid cell number was significantly in-creased in liver tissues from the 2OA-BSA immunizedmice as compared with control BSA-immunized mice at12 weeks: 4.18 � 0.24 versus 1.88 � 0.14 (�106), P �0.0013; whereas there was no difference in the spleen(Fig. 6A). Although the frequency of CD4� T cells in theliver was decreased, the absolute number of CD4� T cellswas not decreased at 12 weeks because of the increase oftotal cell number in the liver. On the other hand, there

Fig. 5. Lymphoid cell population in liver and spleen from immunizedmice at 12 weeks. Cells were stained with mAbs to T cell receptor �(TCR�), CD4, and CD8. Analysis was performed with gating of TCR��

cells. The CD4/CD8 ratio was significantly decreased in liver but not inspleen from mice immunized with 2OA-BSA versus control BSA-immu-nized mice. **P � 0.01.

Fig. 4. Microscopic examination of liver tissue from immunized andcontrol mice. (A-D) Light microscopy of stained sections of liver at 12weeks after immunization of mice with BSA. (A) Shows no abnormalities,and (B-D) after immunization with 2OA-BSA shows mild infiltration oflymphocytes in portal tracts, particularly surrounding damaged intralob-ular bile ducts (B,C; green *). (D) Bile duct loss and epithelioid granu-lomas (yellow arrow) were observed in some portal tracts, and (C)granulomas were found also in hepatic parenchyma. P, portal vein.(hematoxylin and eosin [H&E] �400: scale bar, 100 �m.) (E,F) Immu-nohistochemical analysis of liver sections from 2OA-BSA-immunizedmice at 12 weeks after immunization using mAbs to (E) CD4 and (F)CD8. Both CD4� and CD8� T lymphocytes were distributed in portaltracts with CD8� T cells predominating over CD4� T cells. (�400: scalebar, 100 �m.) (G) Portal infiltrates and bile duct damages were scoredat 12 weeks after immunization: 0 for none, 1 for mild, and 2 formoderate. Both portal infiltrates and bile duct damages were observed inall the tested mice for 2OA-BSA (n � 6) while none of them were foundin all the mice for BSA (n � 4).

536 WAKABAYASHI ET AL. HEPATOLOGY, August 2008

was no difference in either the frequency or absolute num-bers of CD4� T cells in the spleen (Fig. 6B). Importantly,the absolute number of CD8� T cells was markedly in-creased only in the liver compared with control mice:

0.47 � 0.05 versus 0.06 � 0.01 (�106), P � 0.0015, andthere was no difference in the spleen: 8.82 � 1.71 versus11.58 � 1.68 (�106), P � 0.3114 (Fig. 6C). Interest-ingly, the absolute number of CD19� B cells in the liverwas also increased: 1.29 � 0.02 versus 0.35 � 0.06(�106), P � 0.0001, but there was no difference in thespleen (Fig. 6D). The frequency of CD8� T cells thatcoexpressed CD44, a marker of mouse memory T cells,was increased over controls in the liver: 88.70 � 2.04versus 76.47 � 3.50, P � 0.0233, but there was no dif-ference in the spleen.

DiscussionThe data reported herein demonstrates several major

advances since our earlier work on induction, using xeno-biotics, of AMAs and biliary pathology in older guineapigs.7 First, our use of quantitative structure-activity rela-tionship has allowed us to choose 2OA instead of 6BH asa more likely candidate for disease induction. Second, thelatency time for appearance of AMAs and biliary pathol-ogy is now measured in weeks, rather than a year orlonger. Third, the use of a murine model, with the abilityto track and follow infiltrates with mAbs, is much greaterthan with guinea pigs. However, the model still has dis-advantages, including the failure to see fibrosis and theinability in mice to measure cholestatic enzymes. We sus-pect that longer periods of observation, and likely addi-tional host factors, are required to eventually developmore severe destructive cholangitis. These observationsnotwithstanding, these data are strikingly important forstudies of tolerance and study of the early events of auto-immune cholangitis. Finally, and before putting thesedata into the context of human PBC, we should empha-size that we believe that many chemicals may be involvedin similar induction of disease and the use of 2OA hasproven to be a convenient method to break tolerance.This is discussed in more detail in a recent review.6

PBC is an autoimmune disease with a long latencyperiod during which there is a progressive loss of self-tolerance and seropositivity for the signature antibody,anti-PDC-E2. Valid animal models of PBC would greatlyfacilitate investigation of this loss of self-tolerance. Thoserecently described include genetically manipulated mousestrains, including NOD.c3c4 mice,16,17 transforminggrowth factor–� receptor II dominant-negative mice,18-20

and IL-2 receptor ��//� mice.15 Such mouse modelsspontaneously develop selective features of PBC. How-ever, to date, there has been no accepted murine model ofhuman PBC based on experimental immunization.Claimed models are described, including B6 mice in-jected with either polyinosinic:polycytidylic acid,21 orwith purified PDC-E2 in lipopolysaccharide or recombi-

Fig. 6. Numbers of lymphoid cells in liver and spleen from immunizedmice. (A) Total lymphoid cells were significantly increased in the liver but notin the spleen of 2OA-BSA-immunized mice versus control BSA-immunizedmice. (B-D) Cells were stained with anti-TCR�, CD4, CD8, and CD19, andenumerated by flow cytometry. Graphs show data for (B) CD4� T cells, (C)CD8� T cells, and (D) CD19� B cells. CD8� T cells and CD19� B cells weresignificantly increased in the liver but not in the spleen of 2OA-BSA-immunized mice versus control BSA-immunized mice. **P � 0.01.

HEPATOLOGY, Vol. 48, No. 2, 2008 WAKABAYASHI ET AL. 537

nant polypeptides of PDC-E2,22 but these models lackessential diagnostic criteria for human PBC. Further, in agraft versus host disease model,23 there was exemplaryhistopathology but the characteristic AMA reactivity wasnot reliably demonstrated. Finally, in SJL/J mice immu-nized with PDC-E2 emulsified in CFA, there developedan apparent autoimmune cholangitis associated with Tcells that in vitro displayed a Th1/Th2 mixed cytokineprofile after stimulation with the autoantigen,24,25 but un-fortunately a similar histopathology was observed in miceimmunized with control protein (�-casein) emulsified inCFA,26 putting into question the significance of this find-ing. Particular features of the presently described mousemodel, generated by immunization of B6 mice with 2OAconjugated with BSA were that no inflammation wasfound in organs other than liver, no analogous findingswere seen in control B6 mice immunized with BSA inCFA, and antibodies to PDC-E2 were regularly demon-strable by immunoblot and ELISA. The mouse strainused in the present study, C57BL/6J (B6), does not spon-taneously develop biliary disease, but does so only afterimmunization with 2OA-BSA with an anti-PDC-E2 re-sponse.

A number of genetic factors have been suggested aspredisposing both generically to autoimmunity and, morespecifically, to PBC. However, no specific genetic locustsassociated with PBC susceptibility has been demonstrat-ed.27 We should note that the relative risk for a sibling todevelop PBC is significant and, in addition, the concor-dance rate in monozygotic twins in PBC are among thehighest reported in autoimmunity. It is, however, clearthat while genetics are critical, there are important rolesfor either epigenetic events or environmental factors.28

Results from our attempts to develop an animal modelfor human PBC indicate that there are several variablesthat contribute to induction of a “pathological” autoim-mune response versus an autoimmune response with lowto undetectable levels of tissue pathology. These variablesinclude the species of the animal being utilized, the natureof the autoantigenic mimic, and the requirement for anadjuvant. Various clonotypes of self-reactive T cells doescape negative selection within the thymus and, whilepresent in the periphery, do not induce any pathology inthe host.29 Antigenic ignorance (likely based on a naivephenotype), peripheral deletion, induction of anergy, andinfluences of regulatory cells have all been reasoned ascontributors to the failure of such clones of T cells to causeimmunopathology. It is also becoming clear that threesignals could be required for these inert self-reactive Tcells to initiate a pathological response against an autoan-tigen; these include the nature of the antigenic epitopethat leads to the engagement of the appropriate clones of

T cells via the cognate T cell receptor; the costimulatorysignals available; and a third signal that will vary accordingto the lineage and organ within which the autoimmuneresponse is generated. For example, in selected cases, thepresence of IL-12 can efficiently induce such clones ofautoreactive T cells to become activated and initiate au-toimmune pathology.30

In a previous study on PBC, we reported a 10-foldincrease in the relative frequency of PDC-E2-specificCD8� T cells in liver compared with that in peripheralblood,31 and the CD8� T cell frequency in the liver cor-related directly with biliary ductular damage.20,32 Com-parable observations were made for liver tissue from 2OA-BSA-immunized B6 mice in which there were significantnumbers of intrahepatic CD8� T cells (Fig. 6) with adecrease in the CD4/CD8 ratio (Fig. 5) and with an in-crease of CD44high memory T cells (data not shown).These data suggest that CD8� T cells contribute to theimmunopathogenesis of cholangiolitis in 2OA-BSA im-munized B6 mice, but this possibility was not addresseddirectly in the present study. In future work, the effects oftransfer of distinct autoantigen-primed lymphoid cellpopulations within the B6 strain should prove informa-tive. There is obviously a significant body of research thatmust be addressed. We plan, for example, to study a moredetailed phenotypic and functional evaluation of liver in-filtrating cells, dissection of the relationship between theliver infiltrate and biliary duct destruction, and the role ofmajor histocompatibility complex class I and class II ex-pression on bile duct epithelial cells. These data are eitherin progress or are planned. This will include, for example,more detailed phenotypic analysis to dissect the subpopu-lations present in the large pool of CD4–hepatic lympho-cytes, CD8–hepatic lymphocytes, and CD19–hepaticlymphocytes present in the liver of the 2-OA-BSA–in-jected mice.

With regard to the mechanism of disease induction in2OA-BSA–immunized mice, we hypothesize that a mo-lecular mimicry between the self-antigen, lipoylatedPDC-E2, and xenobiotically modified 2OA-PDC-E2 canbreak tolerance, particularly under the additional influ-ence of CFA in the immunizing inoculum; tolerance fail-ure to PDC-E2 then leads on to PBC-like bile ductdamage. This hypothesis is well supported by the obser-vation that anti-PDC-E2 from patients with PBC canrecognize certain xenobiotically-modified PDC-E2 pep-tides that mimic lipoic acid,9 and that PDC-E2 can beexpressed on the surface of BECs in the course of apopto-sis.12,13,33 We should note that immunization withPDC-E2 alone leads to an antimitochondrial responsethat reacts to mitochondria of other species, but not toitself.34 Moreover rabbits and guinea pigs immunized

538 WAKABAYASHI ET AL. HEPATOLOGY, August 2008

with the lipoic acid mimic 6BH, conjugated to BSA inCFA, did produce AMA, albeit after a long interval.5,7,8

We selected 2OA-BSA for the present study because theinhibition study using human PBC patients’ sera demon-strated that 2OA coupled to the lysine residue (173K) ofthe PDC-E2 peptide inner lipoyl domain is uniqueamong the tested compounds in that it did not cross-reactwith the lipoylated PDC-E2 peptide.4 Importantly, thisxenobiotic compound is not found in nature but can bechemically synthesized and the methyl and ethyl esters arewidely used in cosmetic products such as perfume, lip-stick, soap, detergents, cream, lotion, and many commonfood flavorings.4,11,35 BSA-conjugated 6BH was also in-vestigated as an immunogen in the present study but anti-PDC-E2 were not evident in mice immunized with 6BH-BSA by 6 weeks (data not shown), whereas in miceimmunized with 2OA-BSA, seropositivity was evident asearly as 4 weeks postimmunization. For the purpose ofthis study, 2OA-BSA was used to immunize the mice, butobviously another, or potentially many, chemically equiv-alent environmental xenobiotics may well be involvedwith the initiation of human PBC.

Acknowledgment: We thank Thomas P. Kenny forhelpful advice, Willy M. Hsu for maintaining our mito-chondrial clones, and Nikki Phipps for manuscript prep-aration.

References1. Leung PS, Park O, Matsumura S, Ansari AA, Coppel RL, Gershwin ME.

Is there a relation between Chlamydia infection and primary biliary cir-rhosis? Clin Dev Immunol 2003;10:227-233.

2. Padgett KA, Selmi C, Kenny TP, Leung PS, Balkwill DL, Ansari AA, et al.Phylogenetic and immunological definition of four lipoylated proteinsfrom Novosphingobium aromaticivorans, implications for primary biliarycirrhosis. J Autoimmun 2005;24:209-219.

3. Selmi C, Balkwill DL, Invernizzi P, Ansari AA, Coppel RL, Podda M, et al.Patients with primary biliary cirrhosis react against a ubiquitous xenobi-otic-metabolizing bacterium. HEPATOLOGY 2003;38:1250-1257.

4. Amano K, Leung PS, Rieger R, Quan C, Wang X, Marik J, et al.Chemical xenobiotics and mitochondrial autoantigens in primary bil-iary cirrhosis: identification of antibodies against a common environ-mental, cosmetic, and food additive, 2-octynoic acid. J Immunol 2005;174:5874-5883.

5. Amano K, Leung PS, Xu Q, Marik J, Quan C, Kurth MJ, et al. Xeno-biotic-induced loss of tolerance in rabbits to the mitochondrial autoan-tigen of primary biliary cirrhosis is reversible. J Immunol 2004;172:6444-6452.

6. Gershwin ME, Mackay IR. The causes of primary biliary cirrhosis: conve-nient and inconvenient truths. HEPATOLOGY 2008;47:737-745.

7. Leung PS, Park O, Tsuneyama K, Kurth MJ, Lam KS, Ansari AA, et al.Induction of primary biliary cirrhosis in guinea pigs following chemicalxenobiotic immunization. J Immunol 2007;179:2651-2657.

8. Leung PS, Quan C, Park O, Van de Water J, Kurth MJ, Nantz MH, et al.Immunization with a xenobiotic 6-bromohexanoate bovine serum albu-min conjugate induces antimitochondrial antibodies. J Immunol 2003;170:5326-5332.

9. Long SA, Quan C, Van de Water J, Nantz MH, Kurth MJ, Barsky D, et al.Immunoreactivity of organic mimeotopes of the E2 component of pyru-vate dehydrogenase: connecting xenobiotics with primary biliary cirrhosis.J Immunol 2001;167:2956-2963.

10. Rieger R, Gershwin ME. The X and why of xenobiotics in primary biliarycirrhosis. J Autoimmun 2007;28:76-84.

11. Rieger R, Leung PS, Jeddeloh MR, Kurth MJ, Nantz MH, Lam KS, et al.Identification of 2-nonynoic acid, a cosmetic component, as a potentialtrigger of primary biliary cirrhosis. J Autoimmun 2006;27:7-16.

12. Migliaccio C, Nishio A, Van de Water J, Ansari AA, Leung PS, NakanumaY, et al. Monoclonal antibodies to mitochondrial E2 components defineautoepitopes in primary biliary cirrhosis. J Immunol 1998;161:5157-5163.

13. Migliaccio C, Van de Water J, Ansari AA, Kaplan MM, Coppel RL, LamKS, et al. Heterogeneous response of antimitochondrial autoantibodiesand bile duct apical staining monoclonal antibodies to pyruvate dehydro-genase complex E2: the molecule versus the mimic. HEPATOLOGY 2001;33:792-801.

14. Moteki S, Leung PS, Coppel RL, Dickson ER, Kaplan MM, Munoz S, etal. Use of a designer triple expression hybrid clone for three different lipoyldomain for the detection of antimitochondrial autoantibodies. HEPATOL-OGY 1996;24:97-103.

15. Wakabayashi K, Lian ZX, Moritoki Y, Lan RY, Tsuneyama K, ChuangYH, et al. IL-2 receptor alpha(�/�) mice and the development of primarybiliary cirrhosis. HEPATOLOGY 2006;44:1240-1249.

16. Irie J, Wu Y, Wicker LS, Rainbow D, Nalesnik MA, Hirsch R, et al.NOD.c3c4 congenic mice develop autoimmune biliary disease that sero-logically and pathogenetically models human primary biliary cirrhosis. JExp Med 2006;203:1209-1219.

17. Koarada S, Wu Y, Fertig N, Sass DA, Nalesnik M, Todd JA, et al. Geneticcontrol of autoimmunity: protection from diabetes, but spontaneous au-toimmune biliary disease in a nonobese diabetic congenic strain. J Immu-nol 2004;173:2315-2323.

18. Chuang YH, Lian ZX, Yang GX, Shu SA, Moritoki Y, Ridgway WM, et al.Natural killer T cells exacerbate liver injury in a transforming growth factorbeta receptor II dominant-negative mouse model of primary biliary cirrho-sis. HEPATOLOGY 2008;47:571-580.

19. Oertelt S, Lian ZX, Cheng CM, Chuang YH, Padgett KA, He XS, et al.Anti-mitochondrial antibodies and primary biliary cirrhosis in TGF-betareceptor II dominant-negative mice. J Immunol 2006;177:1655-1660.

20. Yang GX, Lian ZX, Chuang YH, Moritoki Y, Lan RY, Wakabayashi K, etal. Adoptive transfer of CD8� T cells from transforming growth factor betreceptor type II (dominant negative form) induces autoimmune cholangi-tis in mice. HEPATOLOGY 2008;47:1974-1982.

21. Okada C, Akbar SM, Horiike N, Onji M. Early development of primarybiliary cirrhosis in female C57BL/6 mice because of poly I:C administra-tion. Liver Int 2005;25:595-603.

22. Ide T, Sata M, Suzuki H, Uchimura Y, Murashima S, Shirachi M, et al. Anexperimental animal model of primary biliary cirrhosis induced by lipo-polysaccharide and pyruvate dehydrogenase. Kurume Med J 1996;43:185-188.

23. Kimura T, Suzuki K, Inada S, Hayashi A, Isobe M, Matsuzaki Y, et al.Monoclonal antibody against lymphocyte function-associated antigen 1inhibits the formation of primary biliary cirrhosis-like lesions induced bymurine graft-versus-host reaction. HEPATOLOGY 1996;24:888-894.

24. Jones DE, Palmer JM, Kirby JA, De Cruz DJ, McCaughan GW, SedgwickJD, et al. Experimental autoimmune cholangitis: a mouse model of im-mune-mediated cholangiopathy. Liver 2000;20:351-356.

25. Jones DE, Palmer JM, Yeaman SJ, Kirby JA, Bassendine MF. Breakdownof tolerance to pyruvate dehydrogenase complex in experimental autoim-mune cholangitis: a mouse model of primary biliary cirrhosis. HEPATOL-OGY 1999;30:65-70.

26. Sasaki M, Long SA, Van De Water J, He XS, Shultz L, Coppel RL, et al.The SJL/J mouse is not a model for PBC. HEPATOLOGY 2002;35:1284-1286.

HEPATOLOGY, Vol. 48, No. 2, 2008 WAKABAYASHI ET AL. 539

27. Tanaka A, Borchers AT, Ishibashi H, Ansari AA, Keen CL, Gershwin ME.Genetic and familial considerations of primary biliary cirrhosis. Am J Gas-troenterol 2001;96:8-15.

28. Selmi C, Mayo MJ, Bach N, Ishibashi H, Invernizzi P, Gish RG, et al.Primary biliary cirrhosis in monozygotic and dizygotic twins: genetics,epigenetics, and environment. Gastroenterology 2004;127:485-492.

29. Sohn SJ, Thompson J, Winoto A. Apoptosis during negative selection ofautoreactive thymocytes. Curr Opin Immunol 2007;19:510-515.

30. Segal R, Dayan M, Zinger H, Habut B, Shearer GM, Mozes E. The effectof IL-12 on clinical and laboratory aspects of experimental SLE in youngand aging mice. Exp Gerontol 2003;38:661-668.

31. Kita H, Matsumura S, He XS, Ansari AA, Lian ZX, Van de Water J, et al. Quan-titative and functional analysis of PDC-E2-specific autoreactive cytotoxic T lym-phocytes in primary biliary cirrhosis. J Clin Invest 2002;109:1231-1240.

32. Lan RY, Cheng C, Lian ZX, Tsuneyama K, Yang GX, Moritoki Y, et al.Liver-targeted and peripheral blood alterations of regulatory T cells inprimary biliary cirrhosis. HEPATOLOGY 2006;43:729-737.

33. Van de Water J, Turchany J, Leung PS, Lake J, Munoz S, Surh CD, et al.Molecular mimicry in primary biliary cirrhosis. Evidence for biliary epi-thelial expression of a molecule cross-reactive with pyruvate dehydrogenasecomplex-E2. J Clin Invest 1993;91:2653-2664.

34. Krams SM, Surh CD, Coppel RL, Ansari A, Ruebner B, Gershwin ME.Immunization of experimental animals with dihydrolipoamide acetyl-transferase, as a purified recombinant polypeptide, generates mito-chondrial antibodies but not primary biliary cirrhosis. HEPATOLOGY

1989;9:411-416.35. Opdyke DL. Monographs on fragrance raw materials. Food Cosmet Toxi-

col 1979;17:357-390.

540 WAKABAYASHI ET AL. HEPATOLOGY, August 2008