|Home | About | Journals | Submit | Contact Us | Français|
Hepatitis C virus (HCV) infection is a leading cause of liver disease worldwide. Alpha interferon (IFN-α) therapy of chronic hepatitis C leads to a sustained response in 10 to 20% of patients only. The mechanisms of viral persistence and the pathogenesis of hepatitis C are poorly understood. We established continuous human cell lines, allowing the tightly regulated expression of the entire HCV open reading frame under the control of a tetracycline-responsive promoter. Using this in vitro system, we analyzed the effect of HCV proteins on IFN-induced intracellular signaling. Expression of HCV proteins in these cells strongly inhibited IFN-α-induced signal transduction through the Jak-STAT pathway. Inhibition occurred downstream of STAT tyrosine phosphorylation. Inhibition of the Jak-STAT pathway was not restricted to IFN-α-induced signaling but was observed in leukemia inhibitory factor-induced signaling through Stat3 as well. By contrast, tumor necrosis factor alpha-induced activation of the transcription factor NF-κB was not affected. Interference of HCV with IFN-α-induced signaling through the Jak-STAT pathway could contribute to the resistance to IFN-α therapy observed in the majority of patients and may represent a general escape strategy of HCV contributing to viral persistence and pathogenesis of chronic liver disease.
Since its discovery in 1989 (4, 22), the hepatitis C virus (HCV) has emerged as the major etiologic agent responsible for most cases of transfusion-associated and sporadic non-A, non-B hepatitis (1). Most HCV-infected individuals develop chronic disease which may progress to liver cirrhosis and eventually hepatocellular carcinoma (2, 45). With an estimated more than 100 million carriers, HCV infection is one of the most important causes of liver disease worldwide. Vaccine development is hampered by the lack of in vitro propagation systems for HCV and the high genetic variability of this single-stranded RNA virus. Currently, alpha interferon (IFN-α) and IFN-α–ribavirin combination therapy are the only approved therapies of HCV infection (18, 29). However, the sustained response rate of IFN-α monotherapy is 10 to 20% and of combination therapy 30 to 40% only (29, 33).
The mechanisms underlying HCV resistance to IFN treatment are not understood. Viral proteins could interfere with IFN induced intracellular signal transduction, thereby inhibiting induction of a number of antiviral effector proteins. Alternatively, the virus could have developed defense strategies against these cellular effector mechanisms. Several IFN-induced effector proteins have been characterized, among them PKR, Mx, 2′-5′ oligoadenylate synthetase, and RNase L (39). The IFN-induced double-stranded RNA-activated protein kinase (PKR) phosphorylates the α-subunit of the eukaryotic translation initiation factor 2 (eIF-2α), thereby inhibiting protein synthesis (25). Recently, repression of the catalytic activity of PKR by the HCV nonstructural 5A (NS5A) protein has been found by biochemical, transfection, and yeast functional analyses (11, 12). Likewise, a reduced basal and induced 2′-5′ oligoadenylate synthetase activity was found in peripheral blood lymphocytes from patients with persistent HCV viremia (32). The relevance of these observations for the natural history of HCV infection is not yet clear, but viral defense strategies targeting the effector mechanisms of IFN-induced antiviral activities could play an important role in viral pathogenesis.
Inhibition of IFN-induced intracellular signals could prevent cellular antiviral responses at an even earlier point. Indeed, examples of viral interference with IFN signal transduction have been reported. Vaccinia virus encodes a soluble IFN-α/β receptor which neutralizes IFN before it can bind to the cellular receptor (44). Similarly, stable expression of the polymerase gene of hepatitis B virus results in impaired activation of interferon-stimulated gene factor 3 (ISGF3) (9). More recently, human cytomegalovirus was reported to inhibit IFN-γ-induced Jak-STAT signaling, probably by enhancing Jak1 protein degradation (26). Over the past several years, the complete signal transduction pathway from the IFN receptors to the nucleus has been identified (6, 16, 19), and viral interference with IFN-induced signaling can now be studied in detail. IFN-α and IFN-β bind to heterodimeric IFN-α/β receptors consisting of IFN-α receptor I (IFNARI) and IFN-α receptor II (IFNARII) (24). Ligand binding results in activation of two cytoplasmic protein tyrosine kinases associated with IFNARI and IFNARII, Tyk2 and Jak1 (46). The activated kinases then phosphorylate tyrosine residues of the receptors (24, 49). These phosphotyrosines are consecutively bound by the src homology 2 (SH2) domains of signal transducer and activator of transcription 1 (Stat1), Stat2, and Stat3 (17, 37, 42). The signal transducers and activators of transcription (STATs) are then phosphorylated at a conserved tyrosine residue located immediately C-terminal of the SH2 domain (41) and form heterodimers or homodimers through mutual SH2-domain–phosphotyrosine interactions (10, 40, 51). Stat3 and Stat1 form homodimers, designated serum inducible factor A (SIF-A) and SIF-C, respectively, and a Stat1-Stat3 heterodimer, SIF-B, that can be detected by electrophoretic mobility shift assays (EMSAs) by using the oligonucleotide probe m67 derived from the promoter of the c-fos gene (48). Stat1 can also dimerize with Stat2, and this Stat1-Stat2 heterodimer associates with a third DNA binding protein, ISGF3γ-p48, to form ISGF3 (10). ISGF3 binds to a different response element and can be detected by EMSA with the oligonucleotide probe IFN-stimulated response element (ISRE) derived from the promoter of IFN-stimulated gene 15 (35). Binding of these STAT factors to their cognate sequences in the promoter regions of target genes results in enhanced gene transcription. Among others, Stat1, Stat2, and p48 have been identified as IFN-α-induced target genes (23). A number of regulatory mechanisms of the Jak-STAT signal transduction pathway have recently been identified. The activity of the Jak kinases is controlled by receptor-associated phosphatases (50) and by the newly discovered family of suppressors of cytokine signaling (SOCS) (7, 31, 43). Binding of Stat3 dimers to DNA can be inhibited by PIAS3 (protein inhibitor of activated STATs) (5). STATs are deactivated by an as-yet-unknown nuclear phosphatase (15) and by protein degradation through the ubiquitin-proteasome pathway (20). At any of the steps outlined above, viral proteins could interfere with the Jak-STAT pathway and inhibit the induction of antiviral effector proteins.
Since an in vitro propagation system for HCV is not available yet for the study of viral interference with IFN signal transduction, we established U-2 OS human osteosarcoma-derived cell lines stably expressing the tetracycline-controlled transactivator (tTA), together with the entire HCV open reading frame under the control of a tTA-dependent promoter. In this system, expression of the transgene can be tightly regulated by varying the concentration of tetracycline in the culture medium (13). By using these cell lines, IFN-induced signaling through the Jak-STAT pathway was investigated in the absence or presence of proteins derived from this inducible HCV transgene.
Human IFN-α2a (Roferon-A) was a gift from Roche Pharma AG (Reinach, Switzerland). Recombinant human TNF-α and leukemia inhibitory factor (LIF) was purchased from Genzyme Diagnostics (Cambridge, MA), recombinant human IFN-γ was obtained from Sigma (St. Louis, Mo.), and recombinant human IL-4 and IL-6 and leptin came from R&D Systems (Wiesbaden, Germany). Antibodies against Stat1 (sc417) and Stat2 (sc476) were from Santa Cruz Biotechnology, Inc. (Santa Cruz, Calif.), antibodies against Stat3 (06-373) and phosphotyrosine (05-321) came from Upstate Biotechnology (Lake Placid, N.Y.), antibodies against ISGF3γ-p48 were from Transduction Labs (Lexington, Ky.), and antibodies against phosphorylated Stat1 were from New England Biolabs (Beverly, Mass.). Antibodies to the p65 subunit of NF-κB were a kind gift of Lienhard Schmitz (38), antibodies to RelB (sc-226), to c-Rel (sc-70), NF-κB p50 (sc-114), and NF-κB p52 (sc-298) were purchased from Santa Cruz Biotechnology. The monoclonal antibody (MAb) C7-50 to the HCV core protein has been described previously (27). The protein inhibitors phenylmethylsulfonyl fluoride (PMSF), aprotinin, leupeptin, and pepstatin were obtained from Boehringer GmbH, Mannheim, Germany.
The characterization of continuous human cell lines, termed UHCV, which allow the tightly regulated expression of the entire HCV open reading frame, has been described in detail elsewhere (28). In brief, U-2 OS human osteosarcoma cells were transfected with the tTA (8). One of the cell lines obtained, UTA-6, was used as a founder cell line for further transfections and as a control cell line in our experiments. UTA-6 cells were then stably transfected with a plasmid, allowing expression of the complete HCV genotype 1a open reading frame derived from pBRTM/HCV1-3011 (14) under the transcriptional control of a tTA-dependent promoter. UHCV-11 and UHCV-32 are well-characterized clones without any HCV gene expression detectable by Northern and Western blot analyses in the presence of 1 μg of tetracycline per ml and with high expression of HCV proteins in the absence of tetracycline. UHCV-26 and UHCV-35 are additional independent clones with different levels of maximal HCV protein expression. UGFP-9 cells allow the tightly regulated expression of the green fluorescent protein (GFP) in the UTA-6 cell background (30). In all of these cells, steady-state expression levels are reached 24 to 48 h after tetracycline withdrawal. Cells were cultured in Dulbecco’s modified Eagle medium supplemented with 10% fetal calf serum, 50 U of penicillin G per ml, 50 μg of streptomycin per ml, 500 μg of G418 per ml, 1 μg of puromycin (for UHCV and UGFP-9 cells but not for UTA-6 cells) per ml, and 1 μg of tetracycline per ml. For induction of viral protein expression, cells were cultured in the absence of tetracycline for 24 h (except where indicated differently) before cytokine treatment. Cells were stimulated for 30 min with 500 U of IFN-α per ml, for 15 min with 10 ng of LIF per ml, or for 15 min with tumor-necrosis factor alpha (TNF-α) at the indicated concentrations or as indicated in the Results and Discussion sections and in the figure legends.
Whole-cell lysates and Western blotting were performed as described elsewhere (17). For detection of HCV core protein in lysates prepared for EMSA, cytoplasmic extracts were analyzed by immunoblot as described earlier (27). For immunoprecipitation experiments, 200 μl of whole-cell extracts were incubated at 4°C for 2 h with 2 μl of antiserum specific for Stat1 (sc-346) or for Stat3 (sc-482) (both from Santa Cruz Biotechnology). After precipitation with protein A-agarose (Upstate Biotechnology), samples were washed three times with 800 μl of lysis buffer buffer (50 mM Tris, pH 8; 280 mM NaCl; 0.5% NP-40; 0.2 mM EDTA; 2 mM EGTA; 10% glycerol; 100 μM Na3VO4; 1 mM dithiothreitol [DTT], 1 mM PMSF, 2 μg of aprotinin per ml, 1 μg of leupeptin per ml, 1 μg of pepstatin per ml) and once with phosphate-buffered saline. Pellets were then boiled for 2 min in 2× sample loading buffer and analyzed by Western blot.
After treatment with cytokines, cells were lysed in low salt buffer (20 mM HEPES, pH 7.9; 10 mM KCl; 1 mM EDTA; 1 mM EGTA; 0.2% NP-40; 10% glycerol; 0.1 mM Na3VO4; 1 mM PMSF; 1 mM DTT; 2 μg of aprotinin, 1 μg of leupeptin, and 1 μg of pepstatin per ml) at 4°C for 10 min. After centrifugation for 5 min at 3,000 × g, supernatants (i.e., cytoplasmic extracts) were immediately frozen on dry ice, and pellets were extracted with high-salt buffer (same as low-salt buffer except for the addition of 420 mM NaCl and 20% glycerol) for 30 min at 4°C. Samples were cleared by centrifugation at 12,000 × g at 4°C. Supernatants were aliquoted, frozen on dry ice, and stored at −70°C. For EMSA, nuclear (2 μl) or cytoplasmic (4 μl) extracts were incubated for 20 to 30 min at 20°C in a mixture containing 20 mM HEPES (pH 7.9), 4% Ficoll, 1 mM MgCl2, 40 mM KCl, 0.1 mM EGTA, 0.5 mM DTT, and 160 μg of poly(dI-dC)-poly(dI-dC) per ml (Sigma) with 1 ng of 32P-labeled oligonucleotides. Samples were separated on a 4.5% nondenaturing polyacrylamide gel at 400 V for 3 h at 4°C. Gels were then dried and exposed to BioMax MR (Kodak) films for 6 h to 3 days. The following oligonucleotides corresponding to STAT response element sequences were used: ISRE, 5′-GAAAGGGAAACCGAACTGAAGC-3′; SIE-m67 (mutated serum inducible element), 5′-CATTTCCCGTAAATCAT-3′; βCAS, 5′-GATTTCTAGGAATTCAATCC-3′; and C, 5′-CACTTCCCAAGAACAGA-3′. For detection of NF-κB shifts, the NF-κB consensus oligonucleotide (5′AGTTGAGGGGACTTTCCCAGGC-3′) was used. For supershift experiments, 1 μl of 1:10-diluted antisera were added to the samples as indicated.
To test for a possible interference of HCV proteins with IFN signal transduction, UHCV-11 and UHCV-32 cells were analyzed for IFN-α-induced ISGF3 formation after derepression of the HCV cDNA. UTA-6 cells, a U-2 OS derived cell line expressing the tTA but lacking the viral transgene served as a negative control. Subconfluent cell monolayers were cultured for 24 h in medium with or without tetracycline. Western blot analysis with an HCV core-specific MAb revealed HCV protein expression in derepressed UHCV-11 and UHCV-32 cells but not in cells cultured in the presence of tetracycline (Fig. (Fig.1C).1C). As expected, no viral proteins are expressed in UTA-6 cells irrespective of the presence or absence of tetracycline in the culture medium. Cells then either were left untreated or were stimulated for 30 min with 500 U of human IFN-α per ml. Nuclear extracts were prepared and tested for ISGF3 DNA binding activity by EMSA by using the ISRE as an oligonucleotide probe (Fig. (Fig.1A).1A). ISGF3 induction after IFN-α treatment was detectable in UTA-6 cells irrespective of the culture conditions (Fig. (Fig.1A,1A, lanes 2 and 4). In UHCV-11 and UHCV-32 cells, however, ISGF3 induction was readily detectable only in cells where viral protein expression has been repressed by tetracycline (Fig. (Fig.1A,1A, lanes 6 and 10). If these cells were cultured in the absence of tetracycline, IFN-α-induced ISGF3 shift activity was inhibited (Fig. (Fig.1A,1A, lanes 8 and 12). Supershift experiments with antisera specific for Stat1 (Fig. (Fig.1A,1A, lane 14), Stat2 (Fig. (Fig.1A,1A, lane 15), and Stat3 (Fig. (Fig.1A,1A, lane 16) confirmed the identity of the induced shift as ISGF3. These findings were confirmed in two independent cell lines inducibly expressing the HCV open reading frame, UHCV-26 and UHCV-35.
In many cells, IFN-α not only induces ISGF3 but also the formation of Stat1 homodimers (designated SIF-C), Stat1-Stat3 heterodimers (SIF-B), and Stat3 homodimers (SIF-A). These shifts can also be observed in our UTA-6 and UHCV cells. A low-level, constitutive Stat3 activation was present even in untreated cells, whereas Stat1 was not detected in these control samples (Fig. (Fig.1B,1B, lanes 1, 3, 5, 7, 9, and 11). A clear induction of SIF-A, SIF-B, and SIF-C was observed in IFN-α-treated UTA-6 cells. In UHCV-11 and UHCV-32 cells, however, expression of viral proteins after derepression of the viral transgene again inhibited the induction of SIF shifts (Fig. (Fig.1B,1B, lanes 8 and 12), although to a somewhat lesser degree than ISGF3. The identity of these shifts was confirmed by supershift experiments with antisera specific for Stat1, Stat2, and Stat3 (Fig. (Fig.2B,2B, lanes 14, 15, and 16, respectively).
In an additional set of experiments, UGFP-9 cells, which allow the tightly regulated expression of GFP as a nonrelevant control protein, were cultured for 24 h in the presence or absence of tetracycline and were then either left untreated or stimulated for 30 min with 500 U of human IFN-α per ml. As shown in Fig. Fig.1D,1D, ISGF3 formation was not affected by the expression of GFP, thereby further confirming the specificity of the inhibition of Jak-STAT signaling observed in UHCV cells.
We next performed a time course experiment with UHCV-32 cells. After derepression of the HCV transgene by withdrawal of tetracycline, viral proteins become detectable after 4 h and then accumulate to reach a maximum expression level after 24 to 48 h (Fig. (Fig.2B).2B). IFN-α-induced ISGF3 shows an inverse correlation to viral protein levels (Fig. (Fig.2A).2A). In the first h after derepression of the transgene, ISGF3 is still detectable, then becomes weaker, and is finally completely absent after 24 h. The results from this time course experiment were confirmed in dose-response assays, where cells were cultured in the presence of 0.01 or 0.005 μg of tetracycline per ml. The tetracycline concentration-dependent protein expression levels also inversely correlated with the intensity of the ISGF3 gel shift (data not shown).
A number of known activators of the Jak-STAT pathway were tested on U-2 OS, UTA-6, and UHCV cells. Treatment of cells with 10 ng of LIF per ml for 15 min results in induction of SIF-A (Stat3 homodimers), whereas no STAT shifts could be detected with ISRE, m67, βCas, or C oligonucleotide probes after treatment of cells with IFN-γ, IL-4, IL-6, platelet-derived growth factor, or leptin (data not shown). Interestingly, LIF-induced Stat3 DNA binding was inhibited by viral proteins as well (Fig. (Fig.3A3A and B). We cannot exclude that both IFN-α and LIF signaling through the Jak-STAT pathway are specifically targeted by HCV proteins. However, a more general inhibition of the Jak-STAT pathway not limited to IFN-α and LIF appears a more likely interpretation of our results.
Expression of viral proteins could result in a general disturbance of cellular homeostasis and, consequently, intracellular signaling events. We therefore tested TNF-α-dependent NF-κB induction in UHCV-32 cells as an example of a non-Jak-STAT signal transduction pathway. After its binding to the 75-kDa TNF-receptor II, TNF-α allows rapid nuclear translocation of NF-κB through degradation of IκB inhibitory cytoplasmic retention proteins (3, 47). In the nucleus, NF-κB binds to a decameric DNA sequence element in the promoter region of target genes (3). NF-κB activation, as detected by EMSA with a consensus binding site oligonucleotide, was not inhibited in UHCV-32 cells expressing HCV proteins (Fig. (Fig.4).4).
At which step of the Jak-STAT signal transduction pathway does interference by viral proteins occur? As outlined above, STAT proteins are activated by phosphorylation of a single tyrosine residue immediately downstream from the SH2 domain (41). To test whether the observed inhibition of DNA binding by STAT proteins is caused by impaired STAT activation at the receptor-kinase complex, Stat1 was immunoprecipitated from whole-cell extracts of UHCV-32 cells either stimulated with IFN-α or left untreated after culture in the presence or absence of tetracycline. Phospho-Stat1-specific signals showed the same intensity in repressed and derepressed cells (Fig. (Fig.5A,5A, α-Stat1-P). Stat1 phosphorylation, therefore, was not impaired by HCV proteins. Stat2 phosphorylation was not inhibited either, as demonstrated by coimmunoprecipitation experiments (Fig. (Fig.5A,5A, α-Stat2). Coimmunoprecipitation is an indirect but reliable indicator of Stat2 phosphorylation, because Stat1-Stat2 heterodimers form only if both Stat1 and Stat2 are tyrosine phosphorylated (34). Analysis of immunoprecipitated proteins with antiphosphotyrosine Western blots confirmed these results (data not shown). We concluded from these experiments that activation of STATs through tyrosine phosphorylation at the receptor-kinase complex was not inhibited by HCV proteins. Viral protein expression could also diminish the cellular concentrations of STAT proteins or of ISGF3γ-p48 by either enhanced protein degradation or impaired gene expression. We could not, however, detect any quantitative difference for Stat1, Stat2, Stat3, or ISGF3γ-p48 (Fig. (Fig.5).5).
After their activation at the cell membrane, STATs translocate into the nucleus by an as-yet-unknown mechanism. Viral proteins could inhibit this translocation. We therefore compared the presence and signal strength of ISGF3 and of SIF-A, SIF-B, and SIF-C of cytoplasmic and nuclear extracts. As shown in Fig. Fig.66 for ISGF3, the same degree of inhibition of DNA binding was observed in cytoplasmic extracts as that found with nuclear extracts (Fig. (Fig.1A).1A). Likewise, no difference between cytoplasmic and nuclear extracts was found for SIF-A, SIF-B, and SIF-C shift (data not shown). Our preparation method for cytoplasmic and nuclear extracts might result in some contamination of cytoplasmic extracts with nuclear proteins and vice versa. But in the case of a nuclear import block for STAT complexes with normal configuration and DNA binding capacity, we would expect no inhibition of shift activities in cytoplasmic extracts from derepressed cells, irrespective of contamination with nuclear proteins, or even stronger shift complexes due to retention of STATs in the cytoplasm. We, therefore, concluded that viral proteins do not inhibit nuclear translocation of STATs. Since neither the activation of the STATs nor their nuclear translocation seem to be inhibited, we believe that HCV proteins or cellular proteins induced by the expression of viral proteins in an indirect way most likely interfere with DNA binding of STATs.
Direct binding of viral proteins to STATs could block the DNA binding domain or change the conformation of STAT dimers. We therefore performed a series of coimmunoprecipitation experiments with whole-cell extracts from derepressed UHCV-11 and UHCV-32 cells with antibodies to Stat1 and Stat3 for immunoprecipitation and antibodies to the viral proteins core, E1, NS3, and NS5A for signal detection on Western blots. We could not, however, detect an association between STATs and HCV proteins (data not shown). Either such an interaction is not stable under the conditions used for cell lysis and immunoprecipitation or else viral proteins induce the expression of unknown cellular proteins that interfere with the Jak-STAT pathway.
To examine the effect of HCV proteins on the expression of IFN-α-induced target genes, UGFP-9 and UHCV-11 cells were cultured for 24 h in the presence or absence of tetracycline, stimulated for 8 h with IFN-α, and subsequently examined for the expression of ISGF3γ-p48 and Stat1. As shown in Fig. Fig.7,7, the inhibition of Jak-STAT signaling in UHCV-11 cells resulted in reduced upregulation of p48 and Stat1. The minor induction of these target genes even in the presence of viral proteins (samples 8) is probably the result of the low level of Stat1 homodimers (SIF-C) induced in these cells (Fig. (Fig.1B).1B). Upregulation of IFN-α-induced target genes was unaffected in UGFP-9 cells which inducibly express GFP as a nonrelevant control protein. These observations indicate that the expression of HCV proteins in UHCV cells affects not only IFN signaling but IFN effector functions as well.
Expression of HCV proteins in UHCV cells inhibits signal transduction through the Jak-STAT pathway after stimulation of cells with IFN-α and LIF. It is unlikely that the accumulation of proteins to toxic levels is responsible for this observation for the following reasons: TNF-α-induced signaling through NF-κB was not affected, tyrosine phosphorylation of STATs occurred efficiently, and no cytopathic effects were apparent in these cells under the culture conditions used here even after several days in culture in the absence of tetracycline. In the context of a natural HCV infection, interference with IFN-induced signaling could be a strategy of HCV to escape natural host defense mechanisms. However, the biological and clinical relevance of our results clearly needs to be further addressed once a cell culture system is available, allowing productive HCV infection. In particular, it is presently unknown whether the inhibition of Jak-STAT signaling observed in our cell lines in vitro will be operative at the probably low levels of viral proteins expressed during natural HCV infection in vivo. In chronic HCV infection lasting for decades, however, even a slight impairment of IFN activity could contribute to viral persistence and pathogenesis. In addition, immunohistochemical analyses have shown that viral antigen expression in human liver in chronic hepatitis C is focal, with scattered hepatocytes expressing higher levels of HCV antigens next to negative cells (21, 36). It is possible, therefore, that in natural HCV infection in some hepatocytes viral antigen expression may reach levels similar to those in our in vitro system.
A better understanding of the mechanisms of viral interference with the Jak-STAT pathway in cell lines could help to clarify the role of signaling inhibition by HCV proteins in vivo. In this context, we are presently establishing cell lines inducibly expressing the individual viral proteins in order to identify the viral gene product(s) responsible for interference with IFN-induced signaling. Overall, our data demonstrate that the expression of HCV proteins inhibits IFN-α-induced signaling through the Jak-STAT pathway. This inhibition may contribute to the poor response of HCV-infected individuals to IFN-α therapy and may be a molecular mechanism contributing to HCV persistence and pathogenesis.
Plasmid pBRTM/HCV1-3011 was kindly provided by Charles M. Rice. UTA-6 cells were kindly provided by Christoph Englart. We thank Petra Binninger and Elke Bieck for excellent technical assistance.
This work was supported by grant Mo 799/1-1 from the Deutsche Forschungsgemeinschaft, by the Stiftung für Gastroenterologische Forschung, a grant from Astra Fonds, and a grant from Sandoz-Stiftung zur Förderung der Medizinisch-Biologischen Wissenschaften.