CD8 T cells primed in the gut-associated lymphoid tissue induce immune-mediated cholangitis in mice

Authors


  • Potential conflict of interest: Nothing to report.

  • Supported by Deutsche Forschungsgemeinschaft (SFB633A10, Z1).

Abstract

The pathogenesis of primary sclerosing cholangitis (PSC) remains poorly understood. Since PSC predominantly occurs in patients with inflammatory bowel disease, autoimmunity triggered by activated T cells migrating from the gut to the liver is a possible mechanism. We hypothesized that T cells primed in the gut-associated lymphoid tissue (GALT) by a specific antigen migrate to the liver and cause cholangitis when they recognize the same antigen on cholangiocytes. We induced ovalbumin-dependent colitis in mice that express ovalbumin in biliary epithelia (ASBT-OVA mice) and crossed ASBT-OVA mice with mice that express ovalbumin in enterocytes (iFABP-OVA mice). We analyzed T-cell activation in the GALT and crossreactivity to the same antigen in the liver as well as the effects of colitis per se on antigen-presentation and T-cell activation in the liver. Intrarectal application of ovalbumin followed by transfer of CD8 OT-I T cells led to antigen-dependent colitis. CD8 T cells primed in the GALT acquired effector function and the capability to migrate to the liver, where they caused cholangitis in a strictly antigen-dependent manner. Likewise, cholangitis developed in mice expressing ovalbumin simultaneously in biliary epithelia and enterocytes after transfer of OT-I T cells. Dextran sodium sulfate colitis led to increased levels of inflammatory cytokines in the portal venous blood, induced activation of resident liver dendritic cells, and promoted the induction of T-cell-dependent cholangitis. Conclusion: Our data strengthen the notion that immune-mediated cholangitis is caused by T cells primed in the GALT and provide the first link between colitis and cholangitis in an antigen-dependent mouse model. (Hepatology 2014;59:601–611)

Abbreviations
APC

antigen presenting cell

ASBT

apical sodium-dependent bile transporter

BSA

bovine serum albumin

DC

dendritic cell

DSS

dextran sodium sulfate

GALT

gut-associated lymphoid tissue

iFABP

intestinal fatty acid binding protein

OVA

ovalbumin

PSC

primary sclerosing cholangitis

Primary sclerosing cholangitis (PSC) is a chronic cholestatic liver disease that accompanies inflammatory bowel disease (IBD) in ∼5% of patients. The pathogenesis of PSC remains obscure. An autoimmune pathogenesis is supported by the association with specific HLA alleles and other autoimmune conditions.[1] Since PSC develops predominantly in patients with IBD and may arise long after colectomy, it was suggested that long-lived immune cells migrating from the gut to the liver play a role. Indeed, T cells displaying the gut-homing phenotype characterized by expression of integrin α4β7[2] and CCR9[3] are found in livers of patients with PSC but not other liver diseases. Liver antigen-presenting cells (APCs) lack the capability to inscribe this phenotype in T cells,[4] since this mechanism requires retinoic acid and is specifically located in the gut-associated lymphoid tissue (GALT).[5] MAdCAM-1, the receptor for α4β7, as well as CCL25, the chemokine attracting CCR9-positive T cells, are expressed in PSC livers.[3, 6] These findings have strengthened the hypothesis that the enterohepatic migration of T cells is responsible for the induction of cholangitis in the early course of PSC.[7]

Several animal models have been proposed for PSC,[8] all of which allow investigation of specific aspects of the disease, namely, infectious triggers,[9] bile transporter dysfunction,[10] and chemical disruption of the biliary epithelium.[11] However, none of them allows the investigation of the enterohepatic migration of T cells. We have developed mice expressing ovalbumin in cholangiocytes of the larger bile ducts.[12] In these mice, CD8 T cells are primed in liver and liver draining lymph nodes and acquire effector function but do not cause inflammation detectable by elevated liver enzymes. By combining these mice with mice in which antigen-specific activation of T cells is achieved in the GALT, we investigated the hypothesis that immigration of CD8 T cells from the gut is responsible for the induction of immune-mediated cholangitis.

Materials and Methods

Mice and Animal Procedures

ASBT-OVA mice express ovalbumin in cholangiocytes.[12] iFABP-OVA mice expressing ovalbumin in enterocytes of the small intestine[13] were crossed with ASBT-OVA mice to obtain iFABP-OVAxASBT-OVA mice. OT-I mice express a T-cell receptor specific for OVA257-264 in the context of H-2Kb. ASBT-OVA, iFABP-OVA, iFABP-OVAxASBT-OVA, C57BL/6J, and RAG1−/−CD90.1+ OT-I mice were maintained under pathogen-free conditions and used at 6 to 12 weeks of age.

For the induction of ovalbumin-dependent colitis (OVA-colitis), anesthetized mice were treated intrarectally with 30% (v/v) ethanol followed after 30 minutes by application of 40 mg ovalbumin (OVA) or bovine serum albumin (BSA) as a control. Simultaneously, 6 × 106 naïve OT-I cells were transferred intravenously. Endoscopy was performed using a miniature endoscope system (Karl Storz, Germany) at day 4.

Chronic dextran sodium sulfate (DSS) colitis was induced by administration of three cycles of 2% DSS in drinking water for 1 week repeated after 3 weeks.[14] Endoscopy was performed at week 6.

All animals received humane care according to institutional criteria. Animal procedures were approved by the LAGeSo Berlin.

T-cell Preparation

Naïve OT-I CD8 T cells were isolated from spleen and lymph nodes of OT-I mice using the negative selection kit from Miltenyi Biotec (Germany). Purity of preparations was above 90%.

To isolate intrahepatic lymphocytes, livers were perfused with phosphate-buffered saline (PBS)/0.5% BSA and passed through a 70-μm nylon mesh. After centrifugation at 50g, nonparenchymal cells were separated on a 40/70% Easycoll gradient (Biochrom, Germany) at 850g for 20 minutes, followed by red blood cell lysis. Lamina propria lymphocytes were isolated as described.[15]

In Vivo Homing Assay

CFSE-labeled OT-I T cells were injected intravenously into iFABP-OVA mice. After 3 days, gut-activated lymphocytes were isolated from mesenteric lymph nodes by sorting for CFSE-positive cells. Naïve and gut-activated OT-I T cells were incubated for 1 hour at 37°C with 20 μCi/mL 51Chromium (Amersham-Buchler, Germany) in complete RPMI. After exclusion of dead cells by gradient centrifugation (Nycodenz 17%) 0.8-1 × 106 cells were injected intravenously into wild-type mice. After 3 hours, organs were harvested for measurement of radioactivity using a γ-counter (Wallac, Finland).

Flow Cytometry

Lymphocytes were stained with anti-CD4, anti-CD8, anti-CD62L, anti-CD90.1, anti-CD44, anti-LPAM1 (α4β7), anti-CD29 (β1), anti-CXCR3, anti-CCR5, anti-CCR9, anti-CD11c, anti-CD40, anti-CD80, anti-CD86 antibodies (eBiosciences, USA). For intracellular staining cells were fixed with 0.5% formaldehyde, permeabilized in 0.5% saponine, and stained with anti-interferon-gamma (IFN-γ) or anti-interleukin (IL)17A in 0.5% saponine for 15 minutes at 4°C. After washing, cells were analyzed on a FacsCalibur/FacsCanto using the CellQuest/Aria software (BD Biosciences, USA).

Alanine Aminotransferase and Cytokine Measurement

Blood samples were centrifuged at 6,000g for 10 minutes, and plasma was stored at −20°C until analysis on a MODULAR analyzer (Roche, Germany). Cytokines within the portal vein plasma were determined by CBA multiplex kit (BD Biosciences) according to the manufacturer's recommendations.

Real-Time Reverse-Transcription Polymerase Chain Reaction (RT-PCR)

Total liver RNA was extracted using the FastRNA Pro green kit (MP Biomedicals, France) and treated with DNaseI (Sigma-Aldrich). Reverse transcription was performed with the SuperScriptIII First-strand synthesis system (Invitrogen), and real-time PCR was performed with SSoFast EVAGreen supermix (Biorad, Germany) using an ABI Prism 7500 Detection System (Applied Biosystems, USA) with primers Rpl4 5′-GTGGGCATGTGGGCC GTTTCT-3′, 5′-GCGATGAATCTTCTTGCGTGGT GC-3′; Icam-1 5′-AGCCTCCGGACTTTCGATCTT-3′, 5′-AGAGGCAGGAAACAGGCCTT-3′; MadCam-1 5′-CCCTACCAGCTCAGCAGAGGACA-3′, 5′-ACCC GGGCTACACCCTCGTC-3′; Vcam-1 5′-GGGGGCC AAATCCACGCTTGT-3′, 5′-AGGGAATGAGTAGAC CTCCACCTGG-3′, and Ccl25 5′-CTGTGAGATTC TACTTCCGCCA-3′, 5′-TCCAGTGGACTAGCCTT TTCCTA-3′. Expression was analyzed relative to the housekeeping gene Rpl4.

Histology and Immunohistochemistry

Left-sided colon and liver were fixed in 4% formaldehyde for 24 hours at 4°C and embedded in paraffin. Four-μm sections were stained with hematoxylin/eosin (H&E). Histological analysis was performed in a blinded fashion.[16]

For immunohistochemistry, deparaffinized 4-μm sections were incubated for 30 minutes with anti-CD3 or anti-MPO7 (#N1580, #A0398, Dako, Denmark), followed by biotinylated donkey anti-rabbit antibody (Dianova, Germany) and the streptavidin alkaline phosphatase kit (Dako) with Fast Red as chromogen. For detection of CK19, sections were boiled in citrate buffer for 5 minutes, incubated with anti-CK19 (EPNCIR127B, Abcam, UK) for 60 minutes, followed by the Envision peroxidase rabbit kit with DAB as chromogen (Dako). For detection of CD8+ cells, cryosections were incubated for 30 minutes at room temperature with anti-CD8 (Ly-2; eBioscience) followed by detection with biotinylated antirat antibody (Dianova), the streptavidin-AP kit using the Fast Red chromogen (Dako), and hematoxylin counterstaining. Positive cells were quantified as the sum of 10 high-power fields (0.237 mm2) for each section. Evaluations were performed in a blinded manner.

Statistical Analysis

Results are shown as mean ± SD. The Wilcoxon-Mann-Whitney test (U test) or analysis of variance (ANOVA) with Bonferroni's post-hoc test for multiple comparisons were used to compare the results for statistical significance; P < 0.05 was considered significant. Statistical analyses were performed using GraphPad Prism 5 (GraphPad Software, USA).

Results

Induction of Antigen-Dependent Colitis in Mice

To establish an animal model in which CD8 T cells are primed in the GALT in an antigen-dependent manner, we treated mice intrarectally with ethanol to disturb the mucosal barrier followed by intrarectal administration of OVA. Naïve OT-I cells were transferred intravenously and their proliferative response was assessed after 24 hours. Proliferating OT-I cells were present in lymph nodes, spleen, and liver (Supporting Fig. 1A) of mice after treatment with OVA but not BSA. After restimulation with the peptide antigen IFN-γ was produced by OT-I cells isolated from spleen, mesenteric lymph node, liver, and lamina propria (Supporting Fig. 1B) of mice treated with OVA but not BSA. No IL-17A secretion was detected (data not shown). To further address the functional capacity of the in vivo primed T cells, we conducted an in vivo cytolysis assay. Specific lysis of target cells was observed in mice after intrarectal application of OVA but not BSA (Supporting Fig. 1C). Cytolysis was observed in spleen and mesenteric lymph nodes but not in the liver, possibly related to the fact that effector and target cells locate to different intrahepatic compartments and contradictory to their effector function displayed in vitro.

Next we analyzed whether antigen-dependent priming of CD8 T cells in the GALT leads to colitis. Intrarectal treatment with OVA in conjunction with transfer of naïve OT-I cells led to transient moderate weight reduction (Fig. 1A). Endoscopy revealed mild to moderate signs of colitis in animals treated with OVA, characterized by increased granularity, vascularization, reduced translucency of the colon wall, and fibrin deposits (Fig. 1B). Histological assessment displayed signs of colitis including infiltrates of mononuclear cells in the lamina propria and edema of the villi (Fig. 1C). Immunohistochemistry revealed an increased content of CD8 T cells in epithelium and lamina propria of mice treated with OVA and increased staining for myeloperoxidase (MPO). Semiquantitative evaluation confirmed significant infiltrates consisting of CD8-positive cells, increased MPO staining, and a mean histological colitis score of 2.5 (±0.8) versus 0.2 (±0.4) in BSA control (Fig. 1D-F).

Figure 1.

Induction of antigen-dependent colitis in mice. Wild-type mice were treated intrarectally with ethanol followed by OVA or BSA, and naïve OT-I cells were transferred intravenously. (A) Dotplots depict weight after induction of OVA colitis relative to initial weight (n ≥ 32). (B) Representative endoscopic view from mini-endoscopy of BSA- or OVA-treated mice at day 4 and endoscopic score (n = 8). (C) H&E staining and immunohistochemistry for CD8 (red) and MPO (red) from colon sections, magnification 100×, insert 400×. (D) Semiquantitative analysis of histopathology. (E) CD8+ cells per 10 high-power fields. (F) MPO+ cells per 10 high-power fields. Mean ± SD (n = 6), Mann-Whitney U test *P < 0.05, **P < 0.01.

Taken together, we established an antigen-dependent model of colitis induced by CD8 T cells primed in the GALT.

CD8 T Cells Primed in the GALT Migrate to the Liver

Analysis of T-cell activation revealed that proliferating T cells are present in the livers of mice treated with OVA intrarectally. OT-I cells isolated from the livers of OVA-treated mice displayed an activated phenotype similar to their counterparts isolated from the mesenteric lymph node, characterized by low expression of CD62L and high expression of CD44 (Fig. 2A). In contrast, the majority of lymphocytes isolated from lymph nodes and livers of mice treated with BSA retained a naïve phenotype (CD62L+CD44low).

Figure 2.

Phenotypic analysis of CD8 T cells primed in the GALT. OVA-colitis was induced in wild-type mice. The phenotype of OT-I cells isolated from mesenteric lymph node (mLN) and liver at day 4 was analyzed for expression of (A) CD62L and CD44, (B) integrin α4β7, (C) integrin β1, (D) chemokine receptor CXCR3. Representative dotplots/histograms from two experiments (n = 8) are shown. The percentage of cells in the respective region is depicted. All plots show events gated on lymphocytes (FSC/SSC) and CD90.1. OVA-treated mice, filled gray histogram; BSA-treated mice, white line; isotype control, black line.

Given the fact that the OT-I cells had been primed in the GALT, we wondered whether liver-infiltrating T cells express integrin α4β7. Indeed, a significant proportion of OT-I cells isolated from the livers of OVA-treated mice expressed α4β7, whereas low expression was detected on OT-I cells isolated from BSA-treated animals (Fig. 2B, Table 1). The expression patterns in liver-infiltrating T cells resembled those of OT-I cells isolated from the mesenteric lymph node. In contrast, expression of α4β7 was low on endogenous CD8 T cells and on naïve OT-I cells (1.8%-2.4%, data not shown).

Table 1. Comparison of CD8 T Cells Primed in the Large and in the Small Intestine
  B6 OVA-ColitisB6 Colitis ControliFABP-OVA 
  MeanSDMeanSDMeanSDOVA-Colitis Vs. ControlOVA-Colitis Vs. iFABP-OVA
  %%%%%%PP
  1. Wild-type mice were treated intrarectally with ethanol followed by ovalbumin (OVA) or bovine serum albumin (control), and naïve OT-I cells were transferred intravenously or naïve OT-I cells were transferred into iFABP-OVA mice.

  2. T cells were isolated from the indicated organs on day 4 and analyzed by flow cytometry. Results are from events gated on lymphocytes (FSC/SSC) and CD90.1. Statistical analysis was performed by two-way ANOVA with Bonferroni's post-hoc test (ns, not significant).

  3. IEL, intraepithelial lymphocytes; mLN, mesenteric lymph node; n/a, not applicable (no transgenic cells were present in the IEL compartment of control mice, the number of IEL was too low for an analysis of IFN-γ).

CD44hi CD62Llowspleen35.169.4911.695.6068.8610.76P < 0.001P < 0.001
mLN42.328.099.215.9872.8313.01P < 0.001P < 0.001
liver50.214.3412.313.6673.2810.04P < 0.001P < 0.001
IEL73.176.85n/an/a84.459.89n/aP < 0.001
Integrin α4β7spleen22.037.025.911.9453.679.93P < 0.001P < 0.001
mLN22.384.548.530.6039.4110.16P < 0.001P < 0.001
liver18.573.716.190.6044.9810.37P < 0.01P < 0.001
IEL24.4612.39n/an/a33.1514.86n/ans
Integrin β1spleen72.4510.1224.2010.6760.117.06P < 0.001P < 0.05
mLN60.608.8414.755.8739.6314.98P < 0.001P < 0.001
liver76.778.9110.814.5655.099.24P < 0.001P < 0.001
IEL60.5113.24n/an/a44.8918.99n/aP < 0.05
CXCR3spleen74.0510.9626.9411.7478.599.19P < 0.001ns
mLN74.719.1123.404.7773.0631.42P < 0.001ns
liver68.7511.9129.5010.0483.5811.17P < 0.001ns
IEL79.5717.46n/an/a79.0224.86n/ans
CCR9spleen46.1818.7667.2918.0471.9110.98P < 0.05P < 0.01
mLN47.408.4571.6811.2376.7814.45P < 0.01P < 0.001
liver16.377.0963.4818.7355.9812.93P < 0.001P < 0.001
IEL60.538.95n/an/a95.263.37n/aP < 0.01
CCR5spleen11.456.988.054.1919.0912.11nsns
mLN8.811.967.381.9220.2815.10nsns
liver3.750.823.831.4315.787.22nsP < 0.01
IEL41.9533.66n/an/a81.9314.46n/ans
IFN-γspleen51.8411.5712.737.4252.7013.96P < 0.001ns
mLN41.0910.2314.7011.3745.8216.91P < 0.001ns
liver48.278.039.865.2349.629.94P < 0.001ns
 IELn/an/an/an/an/an/an/an/a

Since integrin α4 also pairs with β1 to form integrin α4β1, which is implicated in VCAM-1 mediated lymphocyte adhesion, we also analyzed the expression of integrin β1. OT-I cells retrieved from the livers of mice suffering from OVA-colitis up-regulated β1 compared to BSA-treated mice (Fig. 2C, Table 1). Integrin β1 expressing OT-I cells were enriched in the liver compared to mesenteric lymph nodes of OVA-treated mice, while only the minority of OT-I cells isolated from BSA-treated control mice were integrin β1-positive.

CXCR3 was up-regulated on OT-I cells isolated from livers and mesenteric lymph nodes of OVA-treated compared to BSA-treated mice (Fig. 2D, Table 1). In contrast, CCR9 was down-regulated on OT-I T cells isolated from OVA-treated compared to BSA-treated mice. Little CCR5 was detectable on OT-I cells in both groups except for intraepithelial lymphocytes (Table 1).

To investigate the impact of priming in the GALT on migration to the liver we conducted an in vivo homing assay. Three hours after intravenous transfer of naive or gut-activated OT-I T cells into wild-type mice 27.55% of gut-primed compared to 13.94% of naïve T cells had migrated to the liver (Fig. 3).

Figure 3.

Priming in the GALT enhances T-cell migration to the liver. (A) Naïve or gut-activated OT-I T cells isolated from the mesenteric lymph node were labeled with 51chromium and transferred into wild-type mice. Radioactivity was measured in the indicated organs after 3 hours (mLN, mesenteric lymph node; pLN, peripheral lymph node). The percentage of recovered radioactivity is shown from three pooled experiments (n ≥ 4, mean ± SD). Note the different scales for large and small organs. (B) Schematic representation of the percentage of cells migrating into the depicted organs after intravenous transfer.

Our data imply that priming of T cells in the GALT enhances migration to the liver.

T Cells Primed in the GALT Induce Immune-Mediated Cholangitis

We next analyzed whether gut-derived T cells cause inflammation in the liver. To this end, we induced OVA-colitis in ASBT-OVA mice. While priming of CD8 T cells is observed in the liver and in the liver-draining lymph node in ASBT-OVA mice,[12] the kinetics of T-cell priming were faster in ASBT-OVA mice treated with OVA intrarectally than in BSA-treated ASBT-OVA mice (Supporting Fig. 2A). OT-I cells adoptively transferred into ASBT-OVA mice treated with OVA intrarectally acquired an activated phenotype characterized by production of IFN-γ and cytolytic activity (Supporting Fig. 2B,C).

Histological assessment of livers from ASBT-OVA mice treated with OVA intrarectally displayed a mononuclear infiltrate in the portal tracts, characterized as CD3-positive by immunohistochemistry (Fig. 4A). Importantly, such infiltrates were only observed in ASBT-OVA mice treated with OVA intrarectally but not in ASBT-OVA mice treated with BSA or in wild-type mice treated with OVA, confirming the antigen-specific nature of the inflammatory response in the liver and ruling out antigen-independent bystander-hepatitis. The infiltrate mainly consisted of OT-I T cells, as demonstrated by staining for the congenic marker CD90.1, but a small increase of endogenous B cells and regulatory T cells was also noted (Supporting Fig. 3). Determination of plasma alanine aminotransferase (ALT) levels confirmed the presence of liver inflammation with elevated ALT levels in ASBT-OVA mice treated with OVA intrarectally but not in the control groups (Fig. 4B). The transfer of in vitro activated OT-I T cells into ASBT-OVA mice did not induce cholangitis, indicating that activation in the GALT is required (Supporting Fig. 4).

Figure 4.

T cells primed in the GALT induce immune-mediated cholangitis. Wild-type (B6) and ASBT-OVA mice were treated with OVA intrarectally, or wild-type and ASBT mice were treated with the control protein BSA. All mice also received OT-I T cells intravenously. (A) H&E staining, CD3 (red), and CK19 (brown, marking biliary epithelia) immunohistochemistry from liver sections at day 4 (n = 6); magnification 200×. (B) Plasma ALT at day 4 (n = 8). Naïve OT-I cells were transferred into iFABP-OVA, ASBT-OVA, or iFABP-OVAxASBT-OVA mice. Mice were analyzed at day 7. (C) H&E staining, CD3 (red), and CK19 (brown) immunohistochemistry from liver sections (n = 6); magnification 200×. (D) Plasma ALT values (n = 8). ANOVA with Bonferroni's post-hoc test **P < 0.01.

To verify the finding that gut-activated OT-I T cells cause inflammation in the liver, we used iFABP-OVA mice as a transgenic model, in which T cells are primed in the GALT.[17] By crossing iFABP-OVA mice with ASBT-OVA mice, we obtained animals that express the same antigen in enterocytes and cholangiocytes. Similar to the results observed in the OVA-colitis model, OT-I T cells adoptively transferred into iFABP-OVA or iFABP-OVAxASBT-OVA mice acquired an activated phenotype characterized by production of IFN-γ and in vivo cytotoxicity (Supporting Fig. 2D-F). An inflammatory infiltrate was observed in the livers of iFABP-OVAxASBT-OVA mice after transfer of OT-I cells but not in either of the single transgenic mice (Fig. 4C). ALT levels were elevated in the plasma of iFABP-OVAxASBT-OVA mice but not in the single transgenic lines (Fig. 4D).

OT-I T cells activated in iFABP-OVA mice or in OVA-colitis mice differed regarding the expression of chemokine receptors and integrins with stronger up-regulation of integrin α4β7 in iFABP-OVA mice and of β1 in OVA-colitis mice. CCR9 was down-regulated more strongly in OVA-colitis mice, while T cells primed in iFABP-OVA mice showed a stronger overall activation (CD44highCD62Llow) (Table 1).

Our data demonstrate that CD8 T cells primed in the GALT migrate to the liver and cause cholangitis in an antigen-specific manner.

Induction of a Proinflammatory Environment in the Liver of Mice Suffering From Colitis

Given the correlation of PSC with IBD, we investigated whether colitis per se causes changes to the local environment of the liver that may trigger autoimmunity. After induction of chronic DSS colitis, endogenous CD4 and CD8 T cells with an activated phenotype accumulated in the liver (Fig. 5A). Some T cells isolated from the livers of mice suffering from chronic DSS colitis expressed integrin α4β7, suggesting that they were initially primed in the GALT. T cells isolated from the livers of mice suffering from DSS colitis expressed higher levels of integrin β1 than T cells from untreated mice. Chronic DSS colitis also induced changes in the liver relevant for T-cell trapping and activation. ICAM-1 and VCAM-1, which serve as binding partners for integrins on activated T cells, were significantly up-regulated in addition to CCL25 (Fig. 5B), while no expression of MAdCAM-1 was detected (data not shown).

Figure 5.

A proinflammatory environment is induced in the liver of mice suffering from colitis. Chronic DSS colitis was induced in wild-type mice. At day 63 hepatic nonparenchymal cells were isolated and analyzed for (A) expression of CD44, integrin α4β7, and integrin β1 on CD4 or CD8 T cells (gated on lymphocytes [FSC/SSC] and CD4 or CD8) and for (C) CD40, CD80, and CD86 on CD11chigh dendritic cells (gated on live cells and CD11chigh). Representative histogram plots from two experiments (n = 6) are shown. The percentage of cells in the respective region is depicted. (B) At day 63 ICAM-1, VCAM-1, and CCL25 expression in the liver was analyzed by RT-PCR (n ≥ 8). (D) At day 63 cytokines were quantified in plasma from portal venous blood by CBA multiplex assay (n ≥ 8). Mann-Whitney U test **P < 0.01, ***P < 0.001.

Considering the key role of dendritic cells (DCs) in balancing immunity versus tolerance, we analyzed the phenotype of liver DCs in mice suffering from chronic DSS colitis. Expression of the costimulatory molecules CD80 and CD40 was up-regulated on liver DCs from mice suffering from chronic colitis compared to control mice, while no change was seen for CD86 (Fig. 5C). Activation of liver APC may be caused by inflammatory cytokines reaching the liver through the portal vein. We analyzed the level of selected cytokines in the portal vein plasma. The concentration of tumor necrosis factor alpha (TNF-α), interferon-gamma (IFN-γ), interleukin (IL)−6, and monocyte chemoattractant protein-1 (MCP-1/CCL2) was significantly elevated in mice suffering from DSS colitis (Fig. 5D).

Taken together, our data demonstrate that colitis per se leads to changes in the local environment of the liver characterized by increased expression of adhesion molecules, accumulation of activated T cells, and activation of DCs.

Chronic Colitis Promotes the Induction of Immune-Mediated Cholangitis

We tested whether the changes in the local environment of the liver induced by colitis suffice to trigger immune-mediated cholangitis in mice expressing OVA in biliary epithelia. To this end, we induced chronic DSS colitis in ASBT-OVA mice before transferring naïve OT-I cells. OT-I cells proliferated more strongly in ASBT-OVA mice suffering from DSS colitis than in ASBT-OVA mice without colitis, while no proliferation was detected in wild-type mice, no matter whether they suffered from colitis or not (Fig. 6A). Immunohistochemistry confirmed T-cell infiltrates in the livers of ASBT-OVA mice suffering from DSS colitis compared to mice without colitis and wild-type mice with colitis (Fig. 6B). An increase in plasma ALT-levels was observed in ASBT-OVA mice suffering from colitis but neither in ASBT-OVA mice without colitis nor in wild-type mice with colitis (Fig. 6C).

Figure 6.

Chronic colitis promotes the induction of immune-mediated cholangitis. Chronic DSS colitis was induced in wild-type (B6) or ASBT-OVA mice. At day 63 naïve OT-I cells were transferred intravenously. (A) Proliferation of OT-I cells was assessed by CFSE dilution at day 65. The percentage of cells in the respective region is depicted. Representative plots from two experiments (n = 6) are shown. (B) At day 70 H&E staining and CD3 immunohistochemistry (red) from liver sections (n = 6) were performed, magnification 200×. (C) Plasma ALT values (n = 8), ANOVA with Bonferroni's post-hoc test *P < 0.05, **P < 0.01.

Taken together, these data indicate that chronic colitis increases the inflammatory response to antigen present in biliary epithelia.

Discussion

Investigation of the enterohepatic migration of T cells as a cause for autoimmune cholangitis requires an animal model in which the same antigen is present in gut and liver and in which T cells primed in the gut can be traced to the liver, where their immunological effects can be studied. By combining transgenic mice expressing OVA in cholangiocytes with an antigen-dependent model of colitis and with mice expressing the same antigen in the small intestine, we demonstrate that CD8 T cells activated in the GALT cause immune-mediated cholangitis in mice in an antigen-dependent manner. We also show that chronic colitis per se induces a proinflammatory environment in the liver, which leads to increased retention of activated T cells and to enhanced activation of CD8 T cells in the liver.

Our models allow the investigation of the enterohepatic migration of T cells, which directs activated T cells from the GALT to the liver. Frequently, T cells are then rendered tolerant or undergo apoptosis.[18] However, T cells may also elicit immune-mediated liver injury when they encounter their nominal antigen in the liver or cause cytokine-dependent bystander hepatitis.[19] The enterohepatic migration has been proposed as the underlying cause of PSC since it explains key immunological features of PSC, namely, the presence of T cells displaying a gut-homing phenotype in the livers of patients with PSC and the fact that PSC may develop long after proctocolectomy,[7] suggesting that immunologic memory is necessary. This latter finding is difficult to bring in line with the alternative hypothesis that PSC is caused by infectious particles that leak across the damaged mucosal barrier and lead to inflammatory responses in the liver.[20]

Although evidence suggests the presence of an enterohepatic migration of T cells in humans, no mouse model so far existed that allows the investigation of this pathway. We describe mouse models based on the presence of the same antigen in liver and gut, either achieved by topical application or by transgenic expression. In both models antigen-specific activation of CD8 T cells in the GALT is observed with subsequent migration of activated T cells to the liver, where they cause immune-mediated cholangitis in a strictly antigen-dependent manner. Mice that do not express the relevant antigen in their biliary epithelia are protected from cholangitis. Therefore, bystander hepatitis as an unspecific cause of inflammation can be ruled out. Our model is not a model of chronic cholangitis with its sequelae such as strictures or fibrogenesis but a model of immune-mediated cholangitis, in which the initial pathogenic mechanisms leading to autoimmunity can be studied.

Contrary to the clinical situation, in which colitis rather than enteritis is associated with PSC, the antigen could be present in the small or large intestine in our model to allow induction of cholangitis. The discrepancy could be easily explained by the absence of the respective antigen in the small intestine in patients, while transgenic expression allows priming in mice. Given the vast number of bacteria in the colon and their virtual absence in the small intestine, a bacterial antigen would be a prime candidate for the activation of T cells in patients with IBD. OVA-colitis is mild and transient, it neither reflects the chronic nature nor the histopathologic features of ulcerative colitis. However, the sole purpose of our approach was to achieve antigen-specific priming of T cells by an antigen derived from the colon, not the establishment of a novel colitis model. Due to rapid redistribution of OT-I cells we cannot exclude that activation also occurs in other lymphoid organs, although expression of α4β7 strongly suggests that they originate from the GALT. The different phenotype of OT-I cells primed in OVA-colitis and in iFABP-OVA mice should be interpreted with care, since one results from singular topical application while the other results from stable transgenic expression of the antigen. Given the plasticity of T cells it is likely that the imprinted phenotype eventually changes in the absence of antigen.

Retention of T cells in the liver follows several mechanisms. An antigen-unspecific pathway involves VCAM-1 while an antigen-specific pathway requires ICAM-1.[21] Under inflammatory conditions, MAdCAM-1 is expressed in human liver endothelial cells and may allow attachment of gut-derived α4β7-expressing lymphocytes, leading to aberrant migration to the liver.[22] This mechanism of retention is complemented by the expression of CCL25 in inflamed liver, facilitating trans-migration of CCR9-positive T cells. We show that in mice suffering from chronic colitis VCAM-1, ICAM-1, and CCL25 are up-regulated in the liver, leading to accumulation of GALT-primed α4β7 and α4β1-positive T cells. Since we did not detect MAdCAM-1, the role of this adhesion molecule as an interacting partner for integrin α4β7 remains unclear, at least in mice. Our finding is in keeping with results from another study,[23] which also failed to detect MAdCAM-1 in mouse liver. In our model, interactions with activation- or inflammation-associated adhesion molecules such as VCAM-1 and ICAM-1 seem to suffice for the retention of T cells and the induction of cholangitis.

We show that chronic colitis per se leads to improved costimulatory capacity of liver APCs. The phenotype of APCs, especially DCs, in patients with PSC has not been studied. However, abnormalities in the composition of liver APCs in patients with other autoimmune liver diseases such as primary biliary cirrhosis[24] suggest that the number and activation status of APCs may well determine the outcome in PSC. The activation of liver APCs may be caused by damage of the mucosal barrier and influx to the liver of inflammatory mediators, as reported in the DSS colitis model.[25, 26] As a consequence, the presentation of autoantigens by professional APCs may lead to immunity rather than tolerance. Indeed, DCs loaded with autoantigens suffice to induce autoimmune hepatitis.[27] Likewise, haptenization of autologous proteins and subsequent presentation by APCs to T cells induces chronic cholangitis in rats.[28] Once triggered, cholangitis is perpetuated by the autocrine and paracrine effects of proinflammatory cytokines secreted by cholangiocytes and immune cells as well as by costimulatory molecules up-regulated on cholangiocytes.[29]

In IBD a two-step mechanism is likely, in which damage of the mucosal barrier leads to increased inflammatory activity in the GALT. Inflammatory cytokines are released into the portal vein and cause up-regulation of cell adhesion molecules and chemokines in the liver. At the same time, T cells are primed in the GALT, probably by bacterial antigens. These cells are imprinted with the gut-homing phenotype but also express other integrins typical for activated T cells. Using these receptors, T cells are recruited to the liver, where they may cause bystander hepatitis. However, if such T cells recognize their nominal antigen on bile duct epithelia, autoimmune cholangitis ensues (Fig. 7). Of course, the nature of this antigen is unknown. Transient or chronic infections with bacteria or viruses have been described in association with PSC,[30] and infectious agents may trigger autoimmunity through molecular mimicry. In patients with autoimmune hepatitis and PSC atypical p-ANCA crossreact between human β-tubulin isotype 5 and the bacterial β-tubulin precursor FtsZ, which is present in most bacteria of the intestinal flora.[31] This finding links the humoral immune response to intestinal bacteria in IBD to biliary inflammation in PSC. We provide experimental data that link the cellular immune response originating in the gut to immune-mediated cholangitis.

Figure 7.

Schematic illustration of the proposed model of enterohepatic migration and development of cholangitis. APC, antigen-presenting cell; DC, dendritic cell; PAMP, pathogen-associated molecular pattern.

In summary, we provide novel models that recapitulate the enterohepatic migration of T cells, which leads to immune-mediated cholangitis in an antigen-dependent manner. We show that chronic colitis per se induces proinflammatory changes in the local hepatic environment characterized by an increased retention of activated T cells and the improved costimulatory capacity of APC. We propose a two-step model in which proinflammatory factors increase the antigen-presenting capacity in the liver, which in turn promotes crossreactivity of T cells primed in the gut.

Acknowledgment

We thank Dr. L. Lefrançois (University of Connecticut, Storrs, CT) for generously providing iFABP-OVA mice and Simone Spieckermann for excellent technical assistance.

Ancillary