We recently described the development of sclerosing

Transcription

1 GASTROENTEROLOGY 2004;127: Regurgitation of Bile Acids From Leaky Bile Ducts Causes Sclerosing Cholangitis in Mdr2 (Abcb4) Knockout Mice PETER FICKERT,* ANDREA FUCHSBICHLER, MARTIN WAGNER,* GERNOT ZOLLNER,* ARTHUR KASER, HERBERT TILG, ROBERT KRAUSE,* FRANK LAMMERT, CORD LANGNER, KURT ZATLOUKAL, HANNS ULRICH MARSCHALL, HELMUT DENK, and MICHAEL TRAUNER* Departments of *Medicine; and Pathology, Medical University, Graz, Austria; Department of Medicine, Medical University, Innsbruck, Austria; Department of Medicine III, University Hospital Aachen, Aachen, Germany; Department of Medicine, Karolinska University Hospital Huddinge, Stockholm, Sweden Background & Aims: Because the mechanisms leading to bile duct damage in sclerosing cholangitis are unknown, we aimed to determine the pathogenesis of bile duct injury in multidrug resistance gene (Mdr2) (Abcb4) knockout mice (Mdr2 / ) as a novel model of the disease. Methods: Mdr2 / and wild-type controls (Mdr2 / ) were studied at 2, 4, and 8 weeks of age. Liver histology, ultrastructure, immunofluorescence microscopy (to study inflammatory cells, tight junction protein ZO-1, basement membrane protein laminin, fluorescence-labeled ursodeoxycholic acid), immunohistochemistry (for -smooth muscle actin, nitrotyrosine), sirius red staining, bacterial cultures of intra-abdominal organs, and polymerase chain reaction (PCR) for Helicobacter bilis DNA were compared between both genotypes. Hepatic cytokine expression was determined by reverse-transcription PCR. Results: Bile ducts of Mdr2 / showed disrupted tight junctions and basement membranes, bile acid leakage into portal tracts, induction of a portal inflammatory (CD11b, CD4-positive) infiltrate, and activation of proinflammatory (tumor necrosis factor [TNF]-, interleukin [IL]-1 ) and profibrogenic cytokines (transforming growth factor [TGF]- 1). This resulted in activation of periductal myofibroblasts, leading to periductal fibrosis, separating the peribiliary plexus from bile duct epithelial cells and, finally, causing atrophy and death of the bile duct epithelium. Bacterial translocation was not increased and H. bilis was not detectable in Mdr2 /. Conclusions: Sclerosing cholangitis in Mdr2 / mice is a multistep process with regurgitation of bile from leaky ducts into the portal tracts, leading to induction of periductal inflammation, followed by activation of periductal fibrogenesis, finally causing obliterative cholangitis owing to atrophy and death of bile duct epithelial cells. We recently described the development of sclerosing cholangitis in mice with targeted disruption of the multidrug resistance gene (Mdr2) (Abcb4). 1 Mdr2 / mice develop macroscopic and microscopic features (i.e., extra- and intrahepatic biliary strictures and dilatations, onion-skin type periductal fibrosis, and focal fibroobliteration of bile ducts) comparable with the morphologic characteristics of primary and secondary sclerosing cholangitis in humans. 1 3 Under physiologic conditions biliary phospholipids are transported into bile via the canalicular phospholipid flippase Mdr2 and subsequently form mixed phospholipid bile acid micelles, thus protecting cholangiocytes from bile acid induced cell injury. 4,5 The lack of biliary phospholipids in Mdr2 / therefore could result in toxic bile acid induced bile duct damage, ultimately leading to sclerosing cholangitis. 1,6,7 However, the pathogenetic concept of toxic bile causing bile duct injury in Mdr2 / still remains to be proven and the detailed mechanisms leading to sclerosing cholangitis in this model system remain to be determined. In addition to the potential relevance of the Mdr2 / model for studying the pathogenesis of sclerosing cholangitis, there is increasing interest in this model because mutations of its human orthologue MDR-3 cause a wide clinical spectrum of liver disease ranging from neonatal cholestasis to adult liver diseases (e.g., cholestasis of pregnancy, biliary sludge and cholelithiasis, liver cirrhosis) Accordingly, this study was designed to gain insights into the pathophysiologic mechanisms leading to sclerosing cholangitis in Mdr2 /. This information could be valuable for (1) providing fundamental insights into the pathogenesis of sclerosing cholangitis, (2) deepening Abbreviations used in this paper: -SMA, -smooth muscle actin; BECs, bile duct epithelial cells; IL, interleukin; KCs, Kupffer cells; Mdr, multidrug resistance gene; Mdr2 /, Mdr2 (Abcb4) knock-out mice; Mdr2 /, wild-type controls; PCR, polymerase chain reaction; PSC, primary sclerosing cholangitis; TGF- 1, transforming growth factor ; TNF-, tumor necrosis factor ; ZO-1, zonula occludens protein by the American Gastroenterological Association /04/$30.00 doi: /j.gastro

2 262 FICKERT ET AL. GASTROENTEROLOGY Vol. 127, No. 1 our understanding of liver diseases caused by Mdr2/ MDR3 mutations, and (3) developing novel treatment strategies for these disorders. Materials and Methods Animals Mdr2 / knock-out and Mdr2 / wild-type mice (FVB/N background) were obtained from Jackson Laboratory (Jackson Laboratory, Bar Harbor, ME). Mice were housed with a 12-hour light-dark cycle and permitted ad libitum consumption of water and a standard mouse diet (Sniff, Soest, Germany). Livers from 2-, 4-, and 8-week-old male mice (n 5 10 in each group owing to the small amounts of liver tissue in 2- and 4-week-old mice) were excised after cervical dislocation under general anesthesia (400 mg avertin/kg/body weight, intraperitoneally; Sigma-Aldrich, Steinheim, Germany). The experimental protocols were approved by the local Animal Care and Use Committee according to criteria outlined in the Guide for the Care and Use of Laboratory Animals. Routine Serum Biochemistry and Bile Acid Measurements Analysis of alanine transaminase, alkaline phosphatase, and bilirubin by routine clinical chemistry was performed on a Hitachi 917 analyzer (Boehringer Mannheim, Mannheim, Germany). For determination of total serum bile acid levels, a commercial 3 -hydroxysteroid dehydrogenase assay (Ecoline S ; DiaSys, Holzheim, Germany) was used. Liver Histology For conventional light microscopy, livers were fixed in 4% neutral-buffered formaldehyde solution and embedded in paraffin. Sections (4- m thick) were stained with H&E. For visualization of intrahepatic cholesterol crystals, tissue was cryostat sectioned at 4- m thickness and air-dried. Crystals were visualized by polarizing light microscopy as previously described. 14 The sections were coded and examined by 2 pathologists (H.D., C.L.), who were unaware of the animals genotype. Immunofluorescence Microscopy for Tight Junction Protein Zonula Occludens-1, Leukocyte Markers CD4, CD8, CD11b, and Basement Membrane Protein Laminin Immunofluorescence staining for zonula occludens-1 (ZO-1, dilution, 1:100, Anawa; Zurich, Switzerland) was performed on acetone-fixed ( 20 C for 10 min) cryosections of liver tissue as described previously. 15 In addition, slides also were incubated with monoclonal rat antibodies anti-cd4 (Clone RM 4-5, dilution 1:30; BD PharMingen, Heidelberg, Germany), anti-cd8 (Clone , dilution 1:10; BD PharMingen), and monoclonal rat anti-cd11b (Clone M 1/70, dilution 1:50; BD PharMingen), as well as the polyclonal rabbit anti 50-K-160 (dilution 1:50) recognizing cytokeratins 8 and Double immunofluorescence labeling also was performed on acetone-fixed ( 20 C for 10 min) cryosections of liver tissue using rabbit antilaminin (dilution, 1:400; E-Y Laboratories, San Mateo, CA) combined with the polyclonal rabbit anti 50-K-160 (dilution 1:50). 16 Secondary antibodies were tetramethylrhodamine isothiocyanate conjugated antirat (Dianova, Hamburg, Germany) and fluorescein isothiocyanate conjugated anti-rabbit (Dako, Glostrup, Denmark) immunoglobulins. Immunofluorescence Microscopy for Fluorescent Ursodeoxycholic Acid To visualize potential leakage of bile ducts, fluorescent-labeled ursodeoxycholic acid (UDCA; kindly provided by Dr. Allan Hofmann, University of California, San Diego, La Jolla, CA) was injected into the inferior vena cava of 2-weekold Mdr2 / and Mdr2 /. After laparotomy under general anesthesia (400 mg avertin/kg/body weight, intraperitoneally), the inferior vena cava was cannulated using a 27-gauge needle. Thereafter, fluorescent UDCA (0.25 mol/kg/min solved in physiologic saline solution) was given continuously over 4 minutes. After another 2 minutes the needle was removed, and livers were excised and immediately snap-frozen in liquid nitrogen. Unfixed cryosections of liver tissue were mounted on slides, air-dried, and immediately examined without mounting with a laser scanning microscope (LSM 510, Zeiss, Jena, Germany). Immunohistochemistry for -Smooth Muscle Actin, Kupffer Cells, and Nitrotyrosine Immunohistochemistry for -smooth muscle actin ( - SMA) was performed on microwave-treated (0.01 mol/l citrate buffer, ph 6.0) paraffin sections (4- m thick) by using the monoclonal mouse anti -SMA (dilution 1:2500; Sigma, St. Louis, MO). Binding of the antibody was detected using the ABC system (Dako) using -amino-9-ethyl-carbazole (Dako) as substrate. Kupffer cells (KCs) were detected by staining paraffin sections with an antibody recognizing the F4/80 antigen expressed by macrophages (dilution 1:50; Serotec, Oxford, UK). In brief, 4- m paraffin sections were deparaffinated, rehydrated, and digested with 0.1% protease (type XXIV; Sigma) for 10 minutes. Endogenous peroxidase was blocked with 1% H 2 O 2 in methanol for 10 minutes. Specific binding of the F4/80 antibody was detected by using a biotinylated anti-rat IgG (Dako) and the ABC-System (Dako) with amino- 9-ethyl-carbazole as substrate. The anti-nitrotyrosine antibody (dilution 1:50; Abcam, Cambridgeshire, UK) also was detected using the ABC system and amino-9-ethyl-carbazole as substrate after microwave treatment (0.01 mol/l citrate buffer, ph 6.0). Electron Microscopy For electron microscopy, liver tissue was fixed in 2.5% glutaraldehyde in 0.1 mol/l cacodylate buffer, ph 7.3. After

3 July 2004 SCLEROSING CHOLANGITIS IN MDR2 KNOCKOUT MICE 263 postfixation in 1% OsO 4 the tissue was dehydrated and embedded in Agar-100 resin (Agar Scientific, Essex, UK). Sections of 50 nm were contrasted with uranyl acetate and lead citrate and studied in a CM 100 electron microscope (Philips, Eindhoven, Netherlands). 17 Bacterial Cultures of Intra-Abdominal Organs Organs of Mdr2 / and Mdr2 / mice were cultured as described previously. 18 Briefly, animals were killed under general anesthesia (400 mg avertin/kg/body weight, intraperitoneally) by cervical dislocation and the abdominal cavity was entered using sterile procedures. The liver, spleen, and small and large intestines were removed, weighed, placed in sterile tubes, and homogenized with 9 volumes of brain heart infusion using sterile ground-glass stoppers. After manual grinding, 100 L of the homogenates were transferred into tubes containing 0.9 ml of brain heart infusion. From these dilutions, 100- L aliquots out of the homogenized dilutions were plated onto blood agar, McConkey agar (BioMerieux, Marcy l Etoile, France) and Schaedler agar plates (Becton Dickinson, Heidelberg, Germany). All agar plates were incubated at 37 C for 48 hours under aerobic conditions, Schaedler agar plates were incubated at 37 C for 48 hours under anaerobic conditions. The rest of the tissue brain heart infusion homogenates were incubated for 48 hours under aerobic conditions and 100 L then was subcultured on blood agar and McConkey agar. Quantitative culture results were determined by the number of CFU/g, calculated from the cultured dilutions of the organ homogenates. Bacteria were identified by using standard microbiologic procedures and the API system (BioMerieux). Detection of the Helicobacter bilis DNA in Liver Tissue Homogenates Pairs of primers for amplification of a H. bilis specific 638-base DNA fragment (H. bilis primers 11 and 12) were designed automatically according to previously published sequences (VBC-Genomics Bioresearch, Vienna, Austria). 19 For DNA extraction, liver homogenates from Mdr2 / and controls were placed in sterile tubes and subjected to a cell lysis program (Qiagen, Hilden, Germany). DNA preparations (3 L) were added to a 25 L (final volume) reaction mixture containing Ready-To-Go polymerase chain reaction (PCR) Beads (Amersham Biosciences, Cambridgeshire, UK), and 1 mol/l each of PCR primer, and 18 L sterile water. DNA amplification was performed with the following cycling profile: initial denaturation at 95 C for 1 minute, followed by 33 cycles of amplification (denaturation at 95 C for 1 min, annealing at 56 C for 1 min, and extension at 72 C for 1 min), and ending with a final extension at 72 C for 5 minutes. Amplification products were analyzed by agarose gel electrophoresis. Each PCR was performed twice to confirm the results, and each experiment included a PCR-positive control (ATCC strain H. bilis) and a negative control, consisting of the PCR mixture without DNA. Messenger RNA Analysis for Cytokines and Chemokines: Reverse-Transcription and TaqMan Real-Time PCR RNA was isolated as described previously 17 and transcribed into complementary DNA using Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA) according to the manufacturer s instructions. Real-time PCR was performed on a GeneAmp 5700 Sequence Detection System (Applied Biosystems, Vienna, Austria) as previously described. 15 Reactions were performed in duplicates and repeated twice. PCR products were checked by gel electrophoresis for correct size and quality. For quantification, the standard curve method was used. All data were normalized to glyceraldehyde-3-phosphate dehydrogenase. Primer and probe sequences for murine tumor necrosis factor (TNF)-, interleukin (IL)-6, IL-1, interferon-, and transforming growth factor (TGF)- 1 were used as published previously. 20 Determination of Bile Flow and Biliary Bile Acid Composition Bile flow of 8-week-old Mdr2 / and Mdr2 / was measured as described previously 17 and bile samples were analyzed for biliary bile acids by using gas chromatography (GCMS) mass spectrometry and electrospray-mass spectrometry (ESMS) under previously described conditions. 21 Statistical Analysis In each group, 5 10 animals were studied, the latter owing to the small amounts of liver tissue in 2- and 4-weekold mice. Data are reported as arithmetic means SEM unless stated otherwise. For statistical analysis, Student t test or analysis of variance with Bonferroni posttesting were used as appropriate. A P value of 0.05 was considered significant. Results Severe bile duct epithelial cell (BEC) injury rarely is observed before week 4 in Mdr2 /, 1 a time point when periductal inflammation and fibrosis, the hallmarks of the liver phenotype in this model, already are developed fully. Therefore, we established a detailed time course of the individual pathogenetic steps of sclerosing cholangitis in these animals. Two-Week-Old Mdr2 / Have Leaky Bile Ducts, Leading to Bile Acid Leakage Into the Portal Tract In contrast to wild-type controls (Figure 1A C), tight junction morphology was altered severely in 2-week-old Mdr2 / (Figure 1D F). ZO-1 staining between BECs was discontinuous and tortuous in Mdr2 / (Figure 1E F). These findings prompted the question of whether the basal lamina of the bile duct epithelium still was intact and whether the bile ducts

4 264 FICKERT ET AL. GASTROENTEROLOGY Vol. 127, No. 1 Figure 1. Tight junctions are altered severely in Mdr2 /. Double-immunofluorescence labeling of CK 8/18 (in green) and the tight junction protein ZO-1 (in red) was performed in 2-week-old (A C) Mdr2 / ( / ) and (D F) Mdr2 / ( / ). (B, C) ZO-1 staining of bile ducts in Mdr2 / shows a sharp staining pattern outlining the cell contacts between the BECs. (E, F ) In contrast, Mdr2 / show strikingly altered tight junction morphology characterized by a tortuous, partially disrupted, ZO-1 staining pattern between BECs. Bar 20 m. were leaky in Mdr2 /. As determined by immunofluorescence microscopy for laminin (Figure 2E, F) and electron microscopic studies (Figure 3C, D), the basement membrane of the bile ducts frequently was disrupted in Mdr2 / but remained continuous in wildtype controls (Figures 2B, 2C, 3A, 3B). In line with this finding, intercellular spaces between BECs were found to be markedly widened in Mdr2 /, as revealed by electron microscopy (Figure 3C, D). To determine potential leakage of bile, and bile acids in particular, from leaky ducts into portal tracts, fluorescent UDCA was traced in Mdr2 / and wild-type controls. In UDCA-injected Mdr2 /, a distinct canalicular staining pattern was observed (Figure 4A). In addition, the bile duct lumens and BECs stained positive, suggesting that fluorescent UDCA is secreted rapidly into bile and taken up by BECs in Mdr2 / (Figure 4A). In contrast, the canalicular staining pattern was less sharp in Mdr2 / (Figure 4B), which is in line with severely altered tight junction morphology at the canalicular level (not shown). The periductal area of portal fields and bile infarcts frequently stained positive with fluorescent UDCA in Mdr2 /, suggesting leakage of fluorescent UDCA into portal tracts (Figure 4B). In addition, bile infarcts frequently observed in Mdr2 / (which develop owing to obstructive cholangitis and rupture of the bile duct system at the level of the canal of Herring 1 ) stained positive with fluorescent UDCA, suggesting that the periductal area and hepatocytes surrounding bile infarcts face the highest bile acid concentrations. Serum biochemistry and serum bile acid levels are given in Table 1. Table 1. Serum Biochemistry and Bile Acid Levels in Mdr2 / and Mdr2 / Mice ALT (U/L) AP (U/L) SBA ( mol/l) 2wk Mdr2 / Mdr2 / a a a 4wk Mdr2 / b 4 1 b Mdr2 / a a a,b 8wk Mdr2 / b 7 2 Mdr2 / a,b a a,b NOTE. Values are mean SEM from n 5 8 per group. ALT, alanine transaminase; AP, alkaline phosphatase; SBA, serum bile acids. a P 0.05, Mdr2 / vs. Mdr2 / (Student t test). b P 0.05, different timepoints in one genotype (analysis of variance with Bonferroni posttesting).

5 July 2004 SCLEROSING CHOLANGITIS IN MDR2 KNOCKOUT MICE 265 Figure 2. Basement membranes are disrupted frequently in Mdr2 /. Double-immunofluorescence labeling of CK 8/18 (green) and laminin (red) was performed in 2-week-old (A C) Mdr2 / ( / ) and (D F) Mdr2 / ( / ). (C) Double-immunofluorescence of Mdr2 / shows a regular continuous pattern of laminin around bile ducts (bd), hepatic arteries (ha), and the portal vein (pv), and outlining the peribiliary plexus (*). (E, F ) In contrast, laminin staining around bile ducts is discontinuous in Mdr2 /.(F ) Note also that the peribiliary plexus is separated from the biliary epithelium in Mdr2 /. Bar 20 m. Taken together, these findings show that bile duct tight junction morphology as well as basal membrane integrity are altered severely in Mdr2 / and suggest that these changes may lead to leaky bile ducts with regurgitation of bile acids into portal tracts of these animals. Characterization of Inflammatory Response in Mdr2 / The initiation of inflammation plays a critical role in the pathogenesis of sclerosing cholangitis. 2,3 Because we previously have observed pronounced periductal inflammation in Mdr2 /, 1 we aimed to specify further the cellular composition of the inflammatory infiltrate as well as the sources and profiles of cytokines and chemokines. The portal neutrophil granulocyte (CD11b) count was increased significantly in Mdr2 / when compared with wild-type controls and decreased beyond week 4 (Figure 5A). In parallel, the number of CD4- and CD8-positive T cells increased in Mdr2 / from week 4 to week 8 (Figure 5B, C). In contrast, wild-type controls were nearly negative for CD8 markers and the number of CD4-positive cells was significantly lower than in Mdr2 / (Figure 5B, C). The number of KCs was increased significantly in 2-week-old Mdr2 / (Figure 6A, D), but there were no differences in KC count at later time points (Figure 6B, C, E, F). Because portal neutrophil granulocytes and KC activation are observed frequently as a result of bacterial infection and this also has been implicated in the pathogenesis of chronic (sclerosing) cholangitis in rodent models, 3,22 we tested the potential role of bacterial translocation in Mdr2 / in comparison with wild-type controls. Bacterial counts from Mdr2 / livers (range, CFU/g), however, were not statistically different from wild-type controls (range, CFU/g). In addition, we found equal bacterial counts in stomach and small and large intestines, and isolated bacteria did not differ between the genotypes (i.e., Bifidobacterium sp., Staphylococcus epidermidis, and Aerococcus viridans). In addition, H. bilis DNA, a bacterium that might represent a causative cofactor in rodent models for sclerosing cholangitis, 3 could not be detected in liver tissue homogenates from Mdr2 / and wild-type controls. Proinflammatory and profibrogenic cytokines may be critical for the pathogenesis of sclerosing cholangitis. 3,22 For further exploration we compared TNF-, IL-1, IL-6, TGF- 1,

6 266 FICKERT ET AL. GASTROENTEROLOGY Vol. 127, No. 1 Figure 3. Ultrastructural characteristics of sclerosing cholangitis in Mdr2 / comprising pathologic cell-to-cell contacts, disrupted basement membranes, intraepithelial lymphocytes, and periductal edema. Electron microscopy was performed in 2-week-old (A, B) Mdr2 / ( / ) and (C, D) Mdr2 / ( / ). (A) Interlobular bile duct (BD) in Mdr2 / with regular architecture and normal peribiliary plexus. A hepatic artery (HA) branch is near the bile duct. (B) Higher magnification showing regular cell-to-cell contacts between the BECs and a continuous basement membrane is indicated by the arrows.(c) In contrast, the Mdr2 / shows severely altered BECs of a bile duct (BD) with widened intercellular spaces (indicated by arrows). Note the appearance of intraepithelial lymphocytes and also the periductal edematous tissue containing leukocytes and activated fibroblasts. (D) Higher magnification showing severely injured BECs with partially intact basement membrane (indicated by arrows) and disruption of the basement membrane (indicated by arrowheads). Note also that collagen bundles (indicated by stars) are most prominently present near the leak in the basement membrane. *Pathologic cell-to-cell contacts of BECs in Mdr2 /.(A, C) Bar 5 m, (B, D) bar 1 m. Figure 4. Leakage of fluorescent UDCA into the portal tract in Mdr2 /. Fluorescent UDCA was visualized in 2-week-old (A) Mdr2 / and (B, C) Mdr2 / as described in the Materials and Methods section. (A) Fluorescent UDCA stains bile canaliculi and the bile duct lumen as well as the BECs in Mdr2 / liver. (B) Leakage of fluorescent UDCA (indicated by arrows) into the portal tract in Mdr2 / liver. The basal cell border of BECs is indicated by the white ellipse. (C) Bile infarct in Mdr2 / liver impregnated by fluorescent UDCA. These findings suggest that periportal hepatocytes and hepatocytes surrounding bile infarcts face the highest bile acid levels within the liver lobule in Mdr2 /. BD, bile duct. Bar 20 m. and interferon- messenger RNA levels in 2-, 4-, and 8-week-old Mdr2 / and wild-type controls because screening experiments by RNAse protection assay (not shown) have identified these cytokines to be deregulated in 4-week-old Mdr2 /, TNF-, and TGF- 1 messenger RNA levels were increased significantly in 2- and 4-week-old Mdr2 /, whereas IL-1 was overexpressed significantly only in 2-week-old Mdr2 / (Figure 7). The highest levels for IL-6 messenger RNA were observed in 4-week-old Mdr2 / without reaching statistical significance. These findings suggest that overexpression of TNF-, IL-1, IL-6, and TGF- 1 may be most relevant in the early disease course of Mdr2 /. Nitrosative stress also may contribute to the observed inflammatory and fibrogenetic response in Mdr2 /. Because proinflammatory cytokines are capable of inducing nitrosative stress in BECs, we

7 July 2004 SCLEROSING CHOLANGITIS IN MDR2 KNOCKOUT MICE 267 Figure 5. Characterization of the inflammatory infiltrate in Mdr2 /. Double-immunofluorescence labeling of (A) CD11b-, (B) CD8-, (C) CD4- positive cells (red) and CK 8/18 (green) was performed in Mdr2 / ( / ) and Mdr2 / ( / ) at 2, 4, and 8 weeks of age. (A) Comparison between Mdr2 / and Mdr2 / showed an increased number of CD11b-positive cells in Mdr2 / that decreased from week 4 to week 8 of life. (B, C) In contrast, there was an increasing number of (B) CD8- and (C) CD4-positive cells in Mdr2 / from week 2 to 8, whereas Mdr2 / livers stained almost negative for CD8 and CD4 markers. Bar 20 m. also investigated the presence of nitrosylated proteins in BECs of Mdr2 /. However. 2- and 4-week-old animals were negative, suggesting that nitrosative stress does not play a major role in the development of bile duct leakage in this model (not shown). Taken together, these findings argue for the concept that bile duct injury with leakage of chemotactic and fibrogenic bile into portal tracts and not bacterial infection (e.g., with H. bilis) or increased bacterial translocation from the intestine may lead to the in-

8 268 FICKERT ET AL. GASTROENTEROLOGY Vol. 127, No. 1 Figure 6. Increased number of KCs in Mdr2 /. Immunohistochemistry for detection of KCs was performed in (A C) Mdr2 / ( / ) and (D F) Mdr2 / ( / ). (A, D) Higher number of KCs in Mdr2 / compared with Mdr2 /, especially in hepatic acinar zone 1 at 2 weeks of age. (B, C, E, F ) No differences in KC number at later age between the genotypes. Original magnification, 40. bd, bile duct. duction of periductal inflammation and fibrosis in Mdr2 /. Induction of Periductal Fibrosis Causes Separation of the Peribiliary Plexus From the Bile Duct Epithelium in Mdr2 / Because cell death of BECs is increasingly observed beyond week 4 in Mdr2 /, 1 we hypothesized that BEC death could be the result of progressive periductal fibrosis initiated much earlier. Onion-skin like periductal fibrosis was already developed fully after 4 weeks in these animals as revealed by H&E (not shown) and sirius red staining (Figure 8C). In addition, there was an increasing number of periductal -SMA positive myofibroblasts in Mdr2 / over time (Figure 8D). Moreover, we observed that the peribiliary plexus was preserved but separated from bile ducts via a thickened fibrotic ring in Mdr2 / (Figures 2E, 8D), whereas the plexus was localized underneath the bile ducts in Mdr2 / (Figures 2B, 8B). We also tested whether destruction of the bile duct epithelium in Mdr2 / could be related to mechanical abrasion via intraductal cholesterol crystals observed in the later time course (i.e., from week 8 of life), predominantly in female Mdr2 / mice, 14 but no evidence for intraductal cholesterol crystals was found on polarization microscopy in male mice investigated in this study (not shown). It is therefore reasonable to assume that BEC death as observed in the later disease course in Mdr2 / may to some extent represent the consequence of separation of the peribiliary plexus from the biliary epithelium, leading to atrophy and death of BECs. Bile Flow, Biliary Bile Acid Concentration, and Individual Biliary Bile Acid Composition in 8-Week-Old Male Mdr2 / When normalized to liver weight, no differences in bile flow were observed between Mdr2 / and wildtype controls (Table 2). The biliary bile acid concentration measured during constant bile flow was approximately 5% that of gallbladder bile reported previously. 14 The bile acid composition in male Mdr2 / was less hydrophobic/toxic as indicated by significantly higher levels of the taurine -conjugates of -muricholic acid and lower levels of cholic acid. Compared with a previous study 14 we found substantial amounts of -muricholic acid, which might be owing to the used gas chromatography-mass spectrometry analysis in the present study. Of note, electrospray-mass spectrometry gave evidence of taurin conjugates only; ions indicating sulfates or glucuronidates were not recorded. In summary, we provide evidence that sclerosing cholangitis in Mdr2 / is a multistep process with regurgitation of toxic bile from leaky bile ducts into the periportal tissue, leading to induction of inflammation (i.e., infiltration of inflammatory cells and induction of proinflammatory cytokines) and activation of periductal

9 July 2004 SCLEROSING CHOLANGITIS IN MDR2 KNOCKOUT MICE 269 Figure 7. Hepatic cytokine levels in Mdr2 /. Real-time PCR for quantification of relative expression levels of (A) TNF-, (B) IL-1, (C) IL-6, (D) interferon-, and (E) TGF-1 was performed in 2-, 4-, and 8-week-old Mdr2 / and Mdr2 /. Data are expressed as fold change of the respective age-matched Mdr2 / (n 5 in each group; mean SEM; *Mdr2 / ) vs. age-matched Mdr2 / (P 0.05)., Mdr2 / ;, Mdr2 /. myofibroblasts leading to periductal fibrosis, finally causing obliterative cholangitis owing to atrophy and cell death of BECs (summarized in Figure 9). Discussion Because understanding of the pathogenesis of bile duct injury is a key prerequisite for the development of effective medical treatment for bile duct disorders including sclerosing cholangitis (e.g., primary sclerosing cholangitis [PSC]), we designed this study in Mdr2 / as a model system to obtain novel insights into the pathogenetic principles of sclerosing cholangitis. 1 We show that sclerosing cholangitis in Mdr2 / is a multistep process initiated by leakage of bile acids from the bile ducts into portal tracts with consecutive periductal inflammation and fibrosis leading to BEC death. These

10 270 FICKERT ET AL. GASTROENTEROLOGY Vol. 127, No. 1 Figure 8. Periductal fibrosis leads to separation of the peribilary plexus from the biliary epithelium in Mdr2 /. Immunohistochemistry for (A, C) -SMA positive cells and (B, D) Sirius red staining for collagen was performed in 4-week-old Mdr2 / ( / ) and Mdr2 / ( / ) (both in red). (A, C) Sirius red staining is enhanced significantly in Mdr2 /, marking a pronounced periductal fibrotic ring. Note also partial desquamation of the BECs (arrowheads). (B) InMdr2 /, only myoepithelial cells of the hepatic artery (ha) and of the peribiliary plexus (arrows) stain positive for -SMA. (D) In contrast, there are numerous -SMA positive myofibroblasts in Mdr2 / surrounding the bile duct. Note also that the peribiliary plexus (arrows) is separated from the biliary epithelium in Mdr2 /. pv, portal vein; ha, hepatic artery, bd, bile duct. Original magnification, 40. results lead to several key implications and questions as described later. What Causes Bile Duct Leakage in Mdr2 /? Our results show severely altered bile duct (ultra) structure with destroyed tight junctions, basement membranes, and abnormal widened intercellular spaces of BECs in Mdr2 /. The altered tight junctions and basement membranes may represent the morphologic basis for the observed leakage of bile acids into portal tracts in Mdr2 /. Leaky tight junctions and concomitant regurgitation of bile acids also may contribute to increased serum bile acid levels observed in 2-week-old Mdr2 / (Table 1). Because bile acids themselves are able to increase tight junction permeability, 29,30 it is reasonable to assume that increased biliary concentrations of nonmicellar, free, potentially toxic bile acids in Mdr2 / may alter tight junction morphology (as suggested by the pathologic ZO-1 staining pattern of bile ducts in Mdr2 / ) and subsequently affect their function. Alternatively, bile acid mediated induction of proinflammatory cytokines and chemokines from macrophages and/or cholangiocytes also may increase paracellular permeability. 31,32 In addition, nitric oxide synthesis is induced in inflamed bile duct epithelia and has been shown to regulate/alter barrier function of epithelial cells including BECs However, our findings do not provide proof for this attractive concept because a marker for nitrosative stress (i.e., immunohistochemistry for nitrotyrosine in BECs) was negative in 2- and 4-week-old Mdr2 /. Thus, to the best of our knowledge it appears likely that abnormal bile composition itself with high concentrations of nonmicellar toxic bile acids may represent the primary cause of bile duct leakiness in these animals. What Causes Portal Inflammation in Mdr2 /? Most cholangiopathies (e.g., PSC and primary biliary cirrhosis) are associated with portal inflammation and the release of cytokines and chemokines by locally

11 July 2004 SCLEROSING CHOLANGITIS IN MDR2 KNOCKOUT MICE 271 obstruction and myeloperoxidase-positive cells in primary biliary cirrhosis livers. 39 Figure 9. Suggested pathogenetic sequence of events for sclerosing cholangitis in Mdr2 / mice. The lack of biliary phospholipid secretion in Mdr2 / leads to increased biliary concentrations of nonmicellarbound, free bile acids. As a result, there is damage of tight junctions (in red) and basement membranes (in green) of bile ducts, leading to leakage of potentially toxic bile acids into the periductal area, inducing inflammation with activation of neutrophil granulocytes as well as CD4- and CD8-positive lymphocytes. Infiltrating leukocytes and autocrine cytokines result in activated periductal myofibroblasts. Activated periductal myofibroblasts form a periductal fibrous ring. The fibrotic ring separates the peribiliary plexus from the BECs, leading to atrophy and, finally, death of BECs, which itself leads to fibroobliteration of the bile duct. infiltrating macrophages and lymphocytes in proximity to the biliary epithelium. 23,32,33 Mdr2 / also develop a pronounced periductal inflammation, 1 changing from a CD11b-rich to a CD4/CD8-rich infiltrate over time as shown by the present study. In addition, we found a higher number of KCs primarily localized in hepatic acinar zone 1 and induced intrahepatic production of proinflammatory and profibrogenic cytokines (e.g., TNF-, IL-1, TGF -1) in Mdr2 /. It appears highly unlikely that the observed portal inflammation in Mdr2 / is of infectious origin because we found no evidence for bacterial translocation in these animals and bacterial counts/cultures were not different from controls. On the contrary, bile and bile acids themselves potentially are chemotactic, suggesting that the observed portal inflammation in these animals may, at least to some extent, be related to bile acid leakage from bile ducts. In addition, bile duct epithelia themselves may produce cytokines (e.g., TNF-, IL-6) and may participate actively in the inflammatory response. 23,32,33 Consequently, the interaction between activated BECs and infiltrating leukocytes may contribute to the observed destruction of basement membranes in Mdr2 / comparable with the recently observed coincidence of bile duct What Causes Onion-Skin Type Periductal Fibrosis in Mdr2 /? Liver fibrosis generally may be considered the consequence of a sustained wound healing response to different types of injury (e.g., viral, autoimmune, druginduced, cholestasis/bile acid induced) Induction of inflammation with leukocyte infiltration and cytokine release are well-known triggers for stellate cell activation and liver fibrogenesis. 40 Each of these events also can be induced directly by bile acids. 31,43 47 Based on the results of our study it is reasonable to assume that the wound in Mdr2 / may be the leaky bile duct. Consequently, periductal fibrosis in Mdr2 / may be, at least in part, the result of bile acid induced inflammation, ultimately resulting in activation of periductal myofibroblasts. In addition, it is attractive to speculate that the observed decrease of serum bile acid levels in Mdr2 / from week 2 to week 8 of life (Table 2) may, at least in part, be related to reduced leakiness of bile ducts as a result of this wound healing process. Of note, no evidence for induction of adaptive hepatic and renal ABC transporter expression (e.g., Mrp3 and Mrp4) nor differences in the expression of the key bile acid synthesis enzymes (e.g., Cyp7a1, Cyp 27) thought to contribute to such an adaptive response to cholestasis were found in these animals over time (Wagner, Fickert, and Trauner, unpublished data). Table 2. Bile Flow, Biliary Bile Acid Concentration, and Composition in 8-Week-Old Male Mdr2 / and Mdr2 / Mdr2 / Mdr2 / Bile flow ( L/min/g liver) Total biliary bile acids (mmol/l) a Relative amounts (%) Allodeoxycholic acid Deoxycholic acid Allocholic acid b -Muricholic acid c b Cholic acid b UDCA b -Muricholic acid b 7 Ketocholic acid Muricholic acid Others NOTE. Values are mean SEM from n 5 per group. a 95% conjungated with taurine as shown by electrospray-mass spectrometry. b P 0.05, Mdr2 / vs. Mdr2 / analyzed with Student t test. c Peak consists of approximately 20% chenodeoxycholic acid.

12 272 FICKERT ET AL. GASTROENTEROLOGY Vol. 127, No. 1 What Causes BEC Death in Mdr2 /? We previously showed that BEC death is a late event in the disease course of Mdr2 / 1 and now show in the present study that this actually may be related to progressing periductal fibrosis. However, it cannot be fully excluded that, again, nonmicellar toxic bile acids directly leading to chemical/detergent BEC damage might represent a leading mechanism of BEC death in Mdr2 /, although this concept is not supported by the sequence of events brought forward by the current study. Instead, immunohistochemical studies of the peribiliary vascular plexus in patients with PSC have shown that the capillaries are displaced from BECs by expanding fibrotic bands, 48 suggesting that BEC death may be a nutritional problem. The present study provides direct proof for this concept in Mdr2 / by establishing the temporal relationship and sequence of events with periductal fibrosis preceding BEC damage in sclerosing cholangitis. Because damage of BECs as well as full-blown periductal fibrosis are late features in the course of the liver disease observed in Mdr2 /, it is attractive to speculate that atrophy leading to BEC death may result to some extent from relative ischemia and/or malnutrition via a prolonged diffusion distance (e.g., for oxygen and other nutrients). This attractive concept will have to be examined in detail by future studies. What Can We Learn From This Study About Designing Novel Treatment Strategies for Sclerosing Cholangitis? The findings of the present study suggest that bile duct integrity and composition of ductal bile may represent key factors for the development of sclerosing cholangitis in Mdr2 /. If this can be extrapolated to other forms of sclerosing cholangitis it is attractive to speculate that protecting the duct from its potentially toxic content could represent a primary therapeutic goal for sclerosing cholangitis. This therapeutic principle also could explain, at least to some extent, the previously observed beneficial/antifibrotic effects of UDCA not only in Mdr2 / but also in humans with primary biliary cirrhosis and PSC. 1,6,49 During UDCA treatment, the human liver is exposed to lower levels of endogenous (more hydrophobic) bile acids and to an increased concentration of (more hydrophilic) UDCA, rendering the composition of intrahepatic and biliary bile acid less toxic. 50 In addition, UDCA is capable of inducing biliary phospholipid excretion in humans, 51 which also would be anticipated to reduce bile acid damage at the ductal level. 49 It is possible that the recently observed potentially beneficial effects of fibrates and statins in primary biliary cirrhosis patients may result from stimulation of biliary phospholipid secretion and consecutively protection of bile ducts against bile acid induced injury. 52,53 Most interestingly, Mdr2/MDR3 recently has been identified as a target for the nuclear bile acid receptor FXR 54 and may be potently stimulated by therapeutic FXR agonists. Whether these concepts can be applied/extrapolated to PSC remains to be determined. Moreover, interfering with bile duct injury causing periductal fibrosis also should save bile ducts, reduce periductal fibrosis, and, consequently, prevent a vanishing bile duct syndrome if starvation of BECs is indeed involved in the proposed sequence of events. It has to be kept in mind, however, that, at least theoretically, inhibition of the wound healing process (i.e., periductal fibrosis around leaky bile ducts) also could bear potentially harmful effects by permitting ongoing bile acid regurgitation from leaky bile ducts. It is hoped that ongoing interventional studies in Mdr2 / in our laboratory (e.g., testing of antifibrotic agents, modulation of bile composition) will lead to new insights into the complex pathophysiology of bile duct disorders. It is attractive to speculate that immunologically preinjured bile ducts (e.g., in PSC patients) may have increased susceptibility to bile with only reduced phospholipid content. However, it has to be borne in mind that Mdr2 / represent an artificial product/extreme variant completely lacking phospholipid secretion into bile and it remains to be determined whether reduced phospholipid secretion also is involved in the pathogenesis of human PSC or other cholangiopathies. In summary, we herein provide a detailed study of causes and consequences of sclerosing cholangitis in Mdr2 /. We show that tightness of the bile duct system represents the Achilles heel for the development of sclerosing cholangitis in Mdr2 /. This model should stimulate research on the pathogenesis of sclerosing cholangitis because it shares several important morphologic and pathogenetic features with human sclerosing cholangitis (e.g., PSC), and should be suitable to develop and test novel treatment strategies for sclerosing cholangitis. References 1. Fickert P, Zollner G, Fuchsbichler A, Stumptner C, Weiglein AH, Lammert F, Marschall HU, Tsybrovskyy O, Zatloukal K, Denk H. Ursodeoxycholic acid aggravates bile infarcts in bile duct-ligated and Mdr2 knockout mice via disruption of cholangioles. Gastroenterology 2002;123: Angulo P, Lindor KD. Primary sclerosing cholangitis. Hepatology 1999;30: Vierling JM. Animal models for primary sclerosing cholangitis. Best Pract Res Clin Gastroenterol 2001;15: Ruetz S, Gros P. Phosphatidylcholine translocase: a physiological role for the mdr2 gene. Cell 1994;77:

13 July 2004 SCLEROSING CHOLANGITIS IN MDR2 KNOCKOUT MICE Elferink RP, Tytgat GN, Groen AK. Hepatic canalicular membrane 1: the role of mdr2 P-glycoprotein in hepatobiliary lipid transport. FASEB J 1997;11: Van Nieuwkerk CM, Elferink RP, Groen AK, Ottenhoff R, Tytgat GN, Dingemans KP, Van Den Bergh Weerman MA, Offerhaus GJ. Effects of ursodeoxycholate and cholate feeding on liver disease in FVB mice with a disrupted mdr2 P-glycoprotein gene. Gastroenterology 1996;111: Mauad TH, Van Nieuwkerk CM, Dingemans KP, Smit JJ, Schinkel AH, Notenboom RG, Van Den Bergh Weerman MA, Verkruisen RP, Groen AK, Oude Elferink RP. Mice with homozygous disruption of the mdr2 P-glycoprotein gene. A novel animal model for studies of nonsuppurative inflammatory cholangitis and hepatocarcinogenesis. Am J Pathol 1994;145: Jacquemin E, De Vree JM, Cresteil D, Sokal EM, Sturm E, Dumont M, Scheffer GL, Paul M, Burdelski M, Bosma PJ, Bernard O, Hadchouel M, Elferink RP. The wide spectrum of multidrug resistance 3 deficiency: from neonatal cholestasis to cirrhosis of adulthood. Gastroenterology 2001;120: Rosmorduc O, Hermelin B, Poupon R. MDR3 gene defect in adults with symptomatic intrahepatic and gallbladder cholesterol cholelithiasis. Gastroenterology 2001;120: De Vree JM, Jacquemin E, Sturm E, Cresteil D, Bosma PJ, Aten J, Deleuze JF, Desrochers M, Burdelski M, Bernard O, Oude Elferink RP. Mutations in the MDR3 gene cause progressive familial intrahepatic cholestasis. Proc Natl Acad Sci U S A1998;95: Jansen PL, Muller M, Sturm E. Genes and cholestasis. Hepatology 2001;34: Gendrot C, Bacq Y, Brechot MC, Lansac J, Andres C. A second heterozygous MDR3 nonsense mutation associated with intrahepatic cholestasis of pregnancy. J Med Genet 2003;40:e Lucena JF, Herrero JI, Quiroga J, Sangro B, Garcia-Foncillas J, Zabalegui N, Sola J, Herraiz M, Medina JF, Prieto J. A multidrug resistance 3 gene mutation causing cholelithiasis, cholestasis of pregnancy, and adulthood biliary cirrhosis. Gastroenterology 2003;124: Lammert F, Wang DQ, Hillebrandt S, Geier A, Fickert P, Trauner M, Matern S, Paigen B, Carey MC. Spontaneous cholecysto- and hepatolithiasis in mdr2 / mice: a model for low-phospholipidassociated cholelithiasis. Hepatology 2004;39: Wagner M, Fickert P, Zollner G, Fuchsbichler A, Silbert D, Tsybrovskyy O, Zatloukal K, Guo GL, Schuetz JD, Gonzalez FJ, Marschall HU, Denk H, Trauner M. Role of farnesoid X receptor in determining hepatic ABC transporter expression and liver injury in bile duct-ligated mice. Gastroenterology 2003;125: Fickert P, Trauner M, Fuchsbichler A, Stumptner C, Zatloukal K, Denk H. Cytokeratins as targets for bile acid-induced toxicity. Am J Pathol 2002;160: Fickert P, Zollner G, Fuchsbichler A, Stumptner C, Pojer C, Zenz R, Lammert F, Stieger B, Meier PJ, Zatloukal K, Denk H, Trauner M. Effects of ursodeoxycholic and cholic acid feeding on hepatocellular transporter expression in mouse liver. Gastroenterology 2001;121: Wenzl HH, Schimpl G, Feierl G, Steinwender G. Time course of spontaneous bacterial translocation from gastrointestinal tract and its relationship to intestinal microflora in conventionally reared infant rats. Dig Dis Sci 2001;46: Fox JG, Yan LL, Dewhirst FE, Paster BJ, Shames B, Murphy JC, et al. Helicobacter bilis sp. nov., a novel Helicobacter species isolated from bile, livers, and intestines of aged, inbred mice. J Clin Microbiol 1995;33: Overbergh L, Giulietti A, Valckx D, Decallonne R, Bouillon R, Mathieu C. The use of real-time reverse transcriptase PCR for the quantification of cytokine gene expression. J Biomol Tech 2003; 14: Marschall HU, Broome U, Einarsson C, Alvelius G, Thomas HG, Matern S. Isoursodeoxycholic acid: metabolism and therapeutic effects in primary biliary cirrhosis. J Lipid Res 2001;42: Angulo P, Lindor KD. Primary sclerosing cholangitis. Hepatology 1999;30: Spirli C, Fabris L, Duner E, Fiorotto R, Ballardini G, Roskams T, Larusso NF, Sonzogni A, Okolicsanyi L, Strazzabosco M. Cytokine-stimulated nitric oxide production inhibits adenylyl cyclase and camp-dependent secretion in cholangiocytes. Gastroenterology 2003;124: Vos TA, Van Goor H, Tuyt L, Jager-Krikken A, Leuvenink R, Kuipers F, Jansen PL, Moshage H. Expression of inducible nitric oxide synthase in endotoxemic rat hepatocytes is dependent on the cellular glutathione status. Hepatology 1999;29: Vos TA, Gouw AS, Klok PA, Havinga R, Van Goor H, Huitema S, Roelofsen H, Kuipers F, Jansen PL, Moshage H. Differential effects of nitric oxide synthase inhibitors on endotoxin-induced liver damage in rats. Gastroenterology 1997;113: Jaiswal M, Larusso NF, Shapiro RA, Billiar TR, Gores GJ. Nitric oxide-mediated inhibition of DNA repair potentiates oxidative DNA damage in cholangiocytes. Gastroenterology 2001;120: Wallace JL, Miller MJ. Nitric oxide in mucosal defense: a little goes a long way. Gastroenterology 2000;119: Trauner M. When bile ducts say NO: the good, the bad, and the ugly. Gastroenterology 2003;124: Vu DD, Tuchweber B, Raymond P, Yousef IM. Tight junction permeability and liver plasma membrane fluidity in lithocholateinduced cholestasis. Exp Mol Pathol 1992;57: Roma MG, Orsler DJ, Coleman R. Canalicular retention as an in vitro assay of tight junctional permeability in isolated hepatocyte couplets: effects of protein kinase modulation and cholestatic agents. Fundam Appl Toxicol 1997;37: Miyake JH, Wang SL, Davis RA. Bile acid induction of cytokine expression by macrophages correlates with repression of hepatic cholesterol 7alpha-hydroxylase. J Biol Chem 2000;275: Alpini G, McGill JM, Larusso NF. The pathobiology of biliary epithelia. Hepatology 2002;35: Strazzabosco M, Spirli C, Okolicsanyi L. Pathophysiology of the intrahepatic biliary epithelium. J Gastroenterol Hepatol 2000;15: Shibatani N, Yamamoto S, Kubota Y, Tsuji K, Takaoka M, Amoh Y, Matsushita M, Shimatani M, Imai Y, Inoue K. Neutrophil chemotaxis in bile duct-obstructed rats, and effect of internal biliary drainage. Hepatogastroenterology 2002;49: Santoro P, Raimondi F, Annunziata S, Paludetto R, Annella T, Ciccimarra F. Unconjugated bile acids modulate adult and neonatal neutrophil chemotaxis induced in vitro by N-formyl-met-leuphe-peptide. Pediatr Res 2002;51: Kinnman N, Hultcrantz R, Barbu V, Rey C, Wendum D, Poupon R, Housset C. PDGF-mediated chemoattraction of hepatic stellate cells by bile duct segments in cholestatic liver injury. Lab Invest 2000;80: Takaoka M, Kubota Y, Tsuji K, Yamamoto S, Ogura M, Yanagitani K, Shimatani M, Shibatani N, Inoue K. Human neutrophil functions in obstructive jaundice. Hepatogastroenterology 2001;48: Tsuji K, Kubota Y, Yamamoto S, Yanagitani K, Amoh Y, Takaoka M, Ogura M, Kin H, Inoue K. Increased neutrophil chemotaxis in obstructive jaundice: an in vitro experiment in rats. J Gastroenterol Hepatol 1999;14: Wu CT, Eiserich JP, Ansari AA, Coppel RL, Balasubramanian S, Bowlus CL, Gershwin ME, Van De Water J. Myeloperoxidasepositive inflammatory cells participate in bile duct damage in primary biliary cirrhosis through nitric oxide-mediated reactions. Hepatology 2003;38: