Potential conflict of interest: Nothing to report.
Supported by the Cooperation in Cancer Research of the German Cancer Research Center and Israeli Ministry of Science, Culture and Sport (Ca-130 to P.A., J.H., and E.P., and Ca-147 to P.A. and E.P.), the Federal Ministry of Science, Education and Art (MWK Excellence Cluster Initiative to A.B., P.A., and J.H.), the German Research Foundation (SFB Transregio 77, to P.A., J.H., T.L., and A.V.), and the Dietmar Hopp Foundation (to J.H.). Financial support by the Helmholtz Alliance Preclinical Comprehensive Cancer Center.
The receptor for advanced glycation endproducts (RAGE) is a multiligand receptor and member of the immunoglobulin superfamily. RAGE is mainly involved in tissue damage and chronic inflammatory disorders, sustaining the inflammatory response upon engagement with damage-associated molecular pattern molecules (DAMPs) such as S100 proteins and high-mobility group box 1 (HMGB1). Enhanced expression of RAGE and its ligands has been demonstrated in distinct tumors and several studies support its crucial role in tumor progression and metastasis by still unknown mechanisms. Here we show that RAGE supports hepatocellular carcinoma (HCC) formation in the Mdr2−/− mouse model, a prototype model of inflammation-driven HCC formation, which mimics the human pathology. Mdr2−/− Rage−/− (dKO) mice developed smaller and fewer HCCs than Mdr2−/− mice. Interestingly, although in preneoplastic Mdr2−/− livers RAGE ablation did not affect the onset of inflammation, premalignant dKO livers showed reduced liver damage and fibrosis, in association with decreased oval cell activation. Oval cells expressed high RAGE levels and displayed reduced proliferation upon RAGE silencing. Moreover, stimulation of oval cells with HMGB1 promoted an ERK1/2-Cyclin D1-dependent oval cell proliferation in vitro. Finally, genetic and pharmacologic blockade of RAGE signaling impaired oval cell activation in an independent mouse model of oval cell activation, the choline deficient ethionine-supplemented dietary regime. Conclusion: Our data identified a novel function of RAGE in regulating oval cell activation and tumor development in inflammation-associated liver carcinogenesis. (Hepatology 2013)
The receptor for advanced glycation endproducts (RAGE), originally identified as a receptor for advanced glycation endproducts (AGEs), is nowadays considered a pattern-recognition receptor, able to bind different ligands such as high-mobility group box 1 (HMGB1), members of the S100 protein family, and amyloid β peptides.1–3 High constitutive RAGE expression is restricted to the lung,4 while other tissues display low expression levels on vascular endothelial cells, dendritic cells, neutrophils, monocytes/macrophages, lymphocytes, neurons, and cardiomyocytes.3 RAGE engagement promotes the activation of proinflammatory responses and increases the expression of the receptor itself. As a consequence, RAGE has been shown to play an important role in different acute and chronic inflammatory diseases, sepsis, and late diabetic complications.5, 6
Strong up-regulation of RAGE and its ligands were found in different tumors and experimental evidence supports a critical role for RAGE and its ligands in tumorigenesis.3, 6 In fact, blockade of HMGB1-RAGE interaction resulted in decreased tumor growth and metastasis in mouse xenografts.7 Recent findings unraveled a crucial role for RAGE in chemically induced inflammation-driven skin and colitis-associated carcinogenesis.8, 9 In these settings Rage−/− mice displayed reduced leukocyte recruitment and cytokine production during the tumor promotion phase, suggesting that RAGE is a key player in the establishment of a proinflammatory tumor microenvironment.6
In the liver, several reports demonstrated that the HMGB1-RAGE axis influences tissue damage and inflammatory responses under pathological conditions. Indeed, RAGE blockade promoted accelerated survival and decreased liver damage, necrosis, and fibrosis in mouse models of ischemia/reperfusion, partial hepatectomy, and acetaminophen- and carbon tetrachloride (CCl4)-induced liver damage.10–13 Moreover, increased HMGB1 serum levels and its cytoplasmic relocation in hepatocytes were observed in models of ischemia/reperfusion as well as in patients with liver failure and chronic hepatitis B infection.14–18 Finally, increased RAGE and HMGB1 levels were identified in human hepatocellular carcinoma (HCC), suggesting an important role for RAGE signaling in HCC development.19–21 However, knowledge of the molecular mechanism by which RAGE signaling contributes to the pathogenesis of HCC is limited, as it still remains controversial which liver cell compartments express RAGE and how they are affected by its blockade.22
To address the role played by RAGE in HCC development we took advantage of the multidrug resistance 2 knockout (Mdr2−/−) mouse, a prototype of inflammation-associated HCC development. In this model chronic cholestasis, hepatitis, and fibrosis foster HCC formation, mimicking the clinical progression of the human disease.23 We demonstrate that RAGE ablation impairs tumor development, accompanied by a dramatic reduction of oval cell (OC) activation in the preneoplastic state. OC represent liver progenitor cells that are activated in states of severe and chronic damage24 and, as we demonstrate, express high levels of RAGE. Importantly, we observed increased OC proliferation in vitro upon treatment with the RAGE ligand HMGB1. In mice fed a choline deficient ethionine-supplemented diet (CDE), prominent OC activation is greatly diminished upon either genetic loss of RAGE or pharmacological blockade of RAGE signaling.
Materials and Methods
Animals and Animal Work
Animals were maintained in a specific pathogen-free environment and experiments were performed with aged-matched male mice. The procedures for performing animal experiments were in accordance with the principles and guidelines of the Arbeitsgemeinschaft der Tierschutzbeauftragten in Baden-Württemberg and were approved by the Regierungspräsidium of Karlsruhe, Germany.
Mdr2−/−25 and Rage−/−26 animals were described previously. Mdr2+/+ and Mdr2+/− mice were used as controls. For diethylnitrosamine (DEN) treatment, 15-day-old male C57Bl/6 mice were injected intraperitoneally with 10 mg/kg DEN and sacrificed 6 and 12 months after injection. CDE diet was performed on 5-week-old male C57Bl/6 mice as described.27 After 1 week of treatment mice were randomized and injected intraperitoneally with sRAGE (100 μg) or saline every 2 days for 14 days and thereafter sacrificed. Alanine aminotransferase (ALT) activity was measured using an Olympus AU 400 System (measurement range: 3-1,000 U/L). The HMGB1 enzyme-linked immunosorbent assay (ELISA) was done according to the manufacturer's instruction (Shino-Test, Tokyo, Japan). Isolation of liver cell fractions was performed, following two-step collagenase perfusion, by 20%/50% discontinuous Percoll gradient followed by MACS separation with CD45 MicroBeads to deplete CD45-positive cells from the OC fraction. Purity of OC and hepatocyte fractions was determined by morphology analysis, histologic analysis of hematoxylin and eosin (H&E)-stained cytospins and expression analysis by quantitative polymerase chain reaction (qPCR).
Digital Images and Statistical Analysis
All images were acquired on a Leica DMLB microscope and processed using Photoshop CS5 (Adobe, Munich, Germany). Error bars represent standard deviation (SD) except where indicated. Pairwise comparisons between continuous data were done using unpaired two-tailed Student t test.
advanced glycation endproducts
bipotential murine oval liver
choline deficient ethionine-supplemented diet
high mobility group box 1
multidrug resistance protein 2
receptor for advanced glycation endproducts
Hepatocellular Carcinogenesis Impaired in Mdr2−/− Rage−/− Double Knockout Mice
To define the role of RAGE in inflammation-driven tumor development, we crossed Rage−/− mice with the Mdr2−/− mouse strain.23, 25Mdr2−/− Rage−/− double knockout (dKO) mice were viable and produced offspring in a Mendelian ratio. At 15 months of age, control, Rage,−/− Mdr2−/−, and dKO mice (n = 10 for each group) were sacrificed and livers were subjected to histological analysis. Control and Rage−/− livers did not present any focal lesion, while Mdr2−/− mice had enlarged livers that developed multiple HCCs and dysplastic nodules (Fig. 1A, and data not shown). Pathological grading of tumors from Mdr2−/− mice ranged from well differentiated (G1), moderately (G2), up to poorly differentiated (G3), according to the Armed Forces Institute of Pathology grading system. In contrast, dKO mice developed mainly dysplastic nodules (Fig. 1A,B) and only two dKO mice exhibited a single HCC classified as moderately differentiated (G2). Interestingly, while the percentage of mice without any detectable lesion was comparable between Mdr2−/− (28%) and dKO (30%) mice, most Mdr2−/− mice (61%) developed HCCs, whereas the majority of dKO mice (50%) exhibited only premalignant dysplastic nodules (Fig. 1B). In particular, dKO mice showed fewer and smaller liver lesions that did not exceed 12 mm in diameter, whereas lesions in Mdr2−/− mice were bigger in size (up to 20 mm in diameter) and in number (Fig. 1C). Furthermore, dKO mice showed significantly less multifocal tumorigenesis compared to Mdr2−/− mice (Fig. 1D). In contrast, when mice were treated with DEN, which is an alkylating agent causing DNA strand breaks promoting mutations and subsequent HCC formation in a cirrhosis-free manner,28–30 we could not detect any significant difference in tumor number, size, and multiplicity between wildtype (WT) and Rage−/− mice at 12 months after injection (Supporting Fig. 1).
RAGE Ablation Does Not Affect Inflammatory Cell Recruitment
As Rage−/− mice displayed impaired inflammation in two mouse models of tumorigenesis,8, 9 the impact of RAGE ablation on recruitment of inflammatory cells was analyzed in the premalignant phase in Mdr2−/− mice. Liver sections from 3- and 6-month-old Mdr2−/− and dKO mice displayed progressive periductular inflammation and a broad rim of periductular extracellular matrix, which was detected neither in control nor in Rage−/− animals (Supporting Fig. 2A). Staining of tissue sections from 3- and 6-month-old control, Mdr2−/−, and dKO livers for the panleukocyte marker CD45, the neutrophil marker myeloperoxidase (Fig. 2A,B), and the T-cell marker CD3 (data not shown) revealed highly increased levels of immune cells in livers of both Mdr2−/− and dKO mice as compared to controls. However, no significant difference was found between Mdr2−/− and dKO liver sections, suggesting that RAGE deficiency had no major impact on the recruitment of inflammatory cells to the liver in the Mdr2−/− mouse. On the contrary, premalignant WT and Rage−/− mice 6 months after DEN injection did not show any evident sign of liver inflammation as measured by H&E and immunohistochemical staining for either CD45, myeloperoxidase, or CD3-positive cells (Supporting Fig. 3A and data not shown).
At 3 months of age liver and liver/body weight measurements revealed a slight increase in Mdr2−/− and dKO mice as compared to controls, which became significant after 6 months. However, no significant reduction was found in the liver weight of dKO compared to Mdr2−/− mice (Supporting Fig. 2B). Finally, qPCR analysis of tumor necrosis factor alpha (TNF-β), interleukin (IL)−1, IL-6, and several other cyto- and chemokines revealed comparable transcript levels between Mdr2−/− and dKO livers (Fig. 2C; Supporting Fig. 4). In summary, our results demonstrated that RAGE expression is dispensable for the onset and maintenance of inflammation in the Mdr2−/− model.
Impaired Liver Damage and OC Activation in Mdr2−/− Rage−/− Double Knockout Mice
At 3 months of age, both Mdr2−/− and dKO mice exhibited an increased compensatory proliferation of hepatocytes as compared to controls, while the amount of proliferating cell nuclear antigen (PCNA)-positive hepatocytes was significantly reduced in 6-month-old dKO mice as compared to matched Mdr2−/− mice (Supporting Fig. 5A). However, we did not observe any difference in Caspase-3 activation in control, Mdr2−/−, and dKO mice (Supporting Fig. 5B). Quantification of serum samples of 3- and 6-month-old mice showed significantly higher ALT levels in Mdr2−/− and dKO mice as compared to controls. However, this increase in ALT levels was more pronounced in Mdr2−/− animals as compared to dKO mice (Fig. 3A). Similar results were observed for aspartate transaminase (AST) levels (data not shown). Fibrosis analysis by Sirius Red histochemistry of Mdr2−/− liver sections revealed strong periportal and septal fibrosis both at 3 and 6 months of age. Interestingly, dKO livers displayed only a mild fibrosis at 3 months that was slightly increased at 6 months of age (Fig. 3B). Impaired fibrosis in dKO livers was further confirmed by qPCR analysis for Collagen1β1 expression (Fig. 3C).
A hallmark of chronic and severe liver damage is the activation of OC, the putative liver progenitor cells.24 Since dKO mice displayed impaired liver damage and fibrosis, we analyzed whether RAGE ablation affects OC activation. Liver sections of 3- and 6-month-old control, Mdr2−/−, and dKO mice were stained for the OC markers A6 (Fig. 3D) and pan-CK (Supporting Fig. 6).31–33 Positive staining in control liver sections was restricted to the portal tracts, whereas intense staining for activated OC invading the liver parenchyma was found in 3- and 6-month-old Mdr2−/− livers. Importantly, OC activation was strongly impaired in dKO liver sections. These data demonstrate an obvious delay in the onset of liver damage and in OC activation in the premalignant phase of dKO mice. On the contrary, premalignant WT and Rage−/− mice 6 months after DEN injection revealed neither increased ALT levels nor enhanced fibrosis or OC activation when compared to age-matched untreated WT and Rage−/− mice (Supporting Fig. 3A-C and data not shown).
RAGE Expression and Function on OC Proliferation In Vivo
To define more precisely the role played by RAGE on OC activation, we analyzed RAGE expression in hepatocytes, leukocytes (CD45-positive), and OC isolated from livers of mice fed with a CDE diet, a regime which induces liver injury with a prominent OC reaction.27, 34 qPCR and western blot analyses revealed that RAGE was significantly expressed in inflammatory cells but barely detectable in hepatocytes. Noteworthy, OC showed the highest RAGE transcript levels and RAGE protein was easily detectable (Fig. 4A,B), supporting the assumption that RAGE represents a direct regulator of OC activation.
To confirm this hypothesis, we interfered with RAGE signaling in WT mice, in which an OC response was promoted by a 3-week CDE regime. After the first week of treatment, mice were injected every second day with soluble RAGE (sRAGE, 100 μg/mouse), a RAGE decoy receptor,35 or saline. After 2 weeks of treatment mice were sacrificed and livers were analyzed. Quantification of serum ALT levels and PCNA immunohistochemistry revealed increased liver damage and compensatory proliferation in CDE-treated mice as compared to normal diet controls, which was not affected by the administration of sRAGE (Fig. 5A; Supporting Fig. 7A), confirming that sRAGE treatment had no major impact on CDE-induced tissue damage. In line with our previous data, staining for A6 and pan-CK revealed impaired OC activation on liver sections of CDE-sRAGE as compared to CDE-saline animals (Fig. 5B; Supporting Fig. 8A). An impaired OC activation was also observed in Rage−/− mice as compared to CDE-treated WT mice fed a CDE diet for 4 weeks (Supporting Fig. 8B). However, we could observe neither an increase in apoptosis nor an evident infiltration of CD45-positive cells or fibrotic phenotype in either NaCl- or sRAGE-treated mice fed a CDE or a normal diet (Supporting Fig. 7B-D), supporting the assumption that RAGE-dependent OC activation is independent of RAGE signaling in the activation and/or recruitment of immune cells. Indeed, transplantation experiments with bone marrow cells of either WT (WT→WT) or Rage−/− (Rage→WT) donor mice into WT acceptor mice followed by a CDE diet did not affect OC activation due to Rage ablation in infiltrating immune cells (Supporting Fig. 9D). High reconstitution efficiency was confirmed by detecting the Rage-deficient (GFP-positive)26 immune cells using flow cytometry and co-immunofluorescence staining (Supporting Fig. 9A-C).
HMGB1 Promotes OC Proliferation In Vitro
Next, we quantified the amount of known RAGE ligands in liver and serum samples and detected comparable levels for N-carboxymethyllysine (CML), one of the most abundant AGEs, as well as S100A8 and S100A9 in 3- and 6-month-old control, Mdr2−/−, and dKO mice (data not shown). However, HMGB1 serum levels were significantly elevated in Mdr2−/− and dKO mice as compared to controls both at 3 and 6 months of age (Fig. 6A). Accordingly, immunohistochemical staining for HMGB1 was exclusively nuclear in hepatocytes of controls, whereas Mdr2−/− and dKO liver sections displayed strong HMGB1 expression in infiltrating immune cells, accompanied by HMGB1 cytoplasmic relocation in adjacent hepatocytes (Fig. 6B). These data suggest that activated inflammatory cells promote HMGB1 secretion from hepatocytes and thereby promote liver damage and activation of OC. In accordance, in premalignant WT and Rage−/− mice 6 months after DEN injection, which are devoid of any sign of inflammation and liver damage, serum HMGB1 levels were comparable to untreated mice and HMGB1 was retained in the nucleus of hepatocytes (Supporting Fig. 10A,B).
To clarify whether HMGB1 exerts a biological effect on OC activation, we took advantage of bipotential murine oval liver (BMOL) cells, an established murine OC line.36, 37 BMOL cells express RAGE and receptor silencing with specific small interfering RNA (siRNA) oligos (siRAGE) caused a substantial reduction in RAGE protein levels and in cell growth as compared to cells transfected with scrambled siRNA oligos (Fig. 6C). However, apoptosis was not affected by RAGE silencing as measured by a caspase activity assay (Supporting Fig. 11).
BMOL cells displayed increased ERK1/2 phosphorylation, Cyclin D1 expression, and cell proliferation following treatment with recombinant HMGB1 (30 ng/mL) (Fig. 6D-F). HMGB1-induced Cyclin D1 expression (Fig. 6E) and BMOL cell growth (Fig. 6F) were attenuated in the presence of the MEK1/2 inhibitor UO126 (10 μM), indicating that ERK1/2-dependent Cyclin D1 expression is, at least in part, responsible for HMGB1-induced activation of BMOL cells.
RAGE has been reported to play an important role in liver injury and inflammation. Indeed, blockade of RAGE signaling increased survival and decreased necrosis and fibrosis in several mouse models of hepatic injury.10–13 Since hepatic damage is a prerequisite to HCC formation,38 we hypothesized that RAGE expression could directly affect hepatocarcinogenesis. To verify this assumption we analyzed the consequence of RAGE ablation in both the chronic injury- and hepatitis-driven Mdr2−/− model23, 25 and in mice treated with the carcinogen DEN, which exhibits its tumorigenic activity via acute DNA damage, stable integration of genomic mutations, and compensatory proliferation of hepatocytes in the absence of chronic inflammation.28, 29
RAGE ablation significantly impaired HCCs formation only in Mdr2−/− mice and residual lesions were mainly classified as premalignant dysplastic nodules, with only two mice developing a single HCC. The comparable percentage of lesion-free mice between Mdr2−/− and dKO livers suggests that RAGE deficiency delays the onset of malignant transformation, further highlighting the role that is played by RAGE in the malignant progression of liver tumors. The fact that Rage−/− mice were not protected from HCC formation after injection of DEN strongly implies that RAGE is not required for carcinogen-induced hepatocyte transformation but becomes essential only in settings of chronic injury and inflammation. In line with this assumption, premalignant WT and Rage−/− mice 6 months after DEN injection did not show obvious signs of inflammation or tissue damage, whereas premalignant Mdr2−/− and dKO mice displayed chronic liver damage, inflammatory infiltrates, and fibrotic deposition.23, 25, 39
RAGE is expressed on leukocytes and endothelial cells and its engagement by its ligands critically contributes to acute and chronic inflammatory responses.3 Furthermore, RAGE deletion hampered the recruitment of inflammatory cells or the secretion of proinflammatory cytokines in inflammation-induced skin and colon cancer mouse models.8, 9 In contrast to these chemically induced tumor models, we could detect neither a significant impairment in the recruitment of inflammatory cells nor a decrease in the expression of proinflammatory cytokines in dKO compared to Mdr2−/− mice. This may be due, at least in part, to compensatory signaling by other damage-associated molecular pattern receptors such as Toll-like receptor 4 (TLR4), which has been shown to play a crucial role in hepatitis.40 Moreover, we cannot exclude the possibility that the impact of RAGE on the establishment of an inflammatory microenvironment depends on the cause and chronological sequence of tissue activation either by chemical agents or altered pathophysiology due to Mdr2 deletion.
We demonstrate that RAGE ablation in Mdr2−/− mice significantly reduced compensatory proliferation, liver damage, and fibrosis. In line with our data, several studies support an involvement of RAGE in the pathogenesis of liver damage.41 However, the underlying molecular mechanism and the most critical cells within the liver that express RAGE under pathological conditions remained elusive.
In cases of chronic and severe liver damage, OCs (hepatic progenitor cells) are activated, expand, and invade the liver parenchyma from the portal triad, sustaining liver regeneration and restoring liver homeostasis.42 In mouse models, chronic OC activation was found in the premalignant phase of lymphotoxin-driven liver tumorigenesis32 as well as nonalcoholic steatohepatitis.43 Furthermore, impaired liver tumorigenesis in the TNF receptor type 1 knockout mouse was associated with reduced OC activation.44 Although a causal role of OC in HCC development has not been formally proven, it is assumed that activation of the OC compartment in a setting of chronic injury initiates or promotes HCC development.24, 45–47
We found strong OC activation in the preneoplastic livers of Mdr2−/− mice, which was severely impaired in dKO livers. Notably, we found no OC activation in either WT or Rage−/− mice after DEN treatment during the premalignant phase. These results suggest that RAGE plays a key role during liver malignancy only in settings of chronic inflammation and tissue damage accompanied by OC activation.
There are still discrepancies on the origin of RAGE expression in liver cells.22 Our analysis of RAGE expression levels in isolated hepatocytes, immune cells, and OC identified OCs as the major source for RAGE in the challenged liver and strongly support the assumption that RAGE plays a direct role in OC activation. Indeed, RAGE blockade by means of sRAGE injection impaired OC activation in mice fed a CDE diet, a well-established protocol for OC activation.27, 34 It is worth noting that CDE-induced compensatory proliferation, liver damage, inflammation, and fibrosis were not affected by sRAGE administration, indicating a direct effect of sRAGE on OC activation. This assumption was further supported by bone marrow transfer experiments and impaired OC activation in Rage−/− mice upon CDE treatment, although we cannot completely rule out an involvement of RAGE in resident Kupffer cells.48 In line with these data, RAGE silencing dramatically decreased growth of the OC line BMOL and treatment with the RAGE ligand HMGB1 promoted ERK1/2-cyclin D1-dependent BMOL cell growth.
Several studies demonstrated that the presence of extracellular HMGB1 is causally linked to inflammation and tissue injury.3 In particular, cytoplasmic HMGB1 relocation has been associated with increased serum HMGB1 levels in mouse models of liver injury and with HMGB1 secretion upon lipopolysaccharide (LPS) and TNF treatment in vitro.14–16 In Mdr2−/− and dKO livers, we detected HMGB1-positive infiltrating immune cells and cytoplasmic HMGB1 relocation in adjacent hepatocytes, a prerequisite for its secretion. Accordingly, the HMGB1 concentration was highly increased in sera of dKO and Mdr2−/− mice but remained unaltered in sera of WT and Rage−/− mice 6 months after DEN treatment (data not shown), suggesting an effect of HMGB1 on OC in vivo. Although HMGB1 levels were comparable in Mdr2−/− and dKO mice, liver damage was significantly decreased in dKO mice. This strongly suggests that inflammation is independent of RAGE, while OC activation critically depends on HMGB1-RAGE signaling. Indeed, blockade of RAGE ligands in the CDE model by treatment with the decoy receptor sRAGE or genetic deletion of RAGE reduced OC activation in vivo.
Blockade of HMGB1-RAGE interaction has been shown to effectively reduce liver damage upon acute injury.14, 15, 49 Unfortunately, Hmgb1−/− mice die at birth50 and, hitherto, no conditional Hmgb1−/− mouse has been established to test whether HMGB1 ablation in the Mdr2−/− mouse phenocopies the dKO liver phenotype. Furthermore, it will be of great interest to correlate the expression of RAGE and the abundance of its ligands, in particular HMGB1, with the severity of disease and its clinical outcome in different human liver disorders, and to prove the concept that pharmacological inhibition of RAGE signaling represents a novel strategy for the prevention of HCC development during early stages of liver injury.
We thank Tine Bauer, Angelika Krischke, and Sandra Prokosch for technical support, Prof. George Yeoh (University of Western Australia) and Janina Tirnitz-Parker for BMOL cells, and Valentina Factor (NIH) for the A6 antibody.