Inserm, UMRS 938, Centre de Recherche Saint-Antoine, Paris, France
UPMC, Univ Paris 06, UMRS 938, Centre de Recherche Saint-Antoine, Paris, France
Address reprint requests to: Laura Fouassier, Ph.D., Inserm UMRS 938, Centre de Recherche Saint-Antoine, Faculté de Médecine Pierre et Marie Curie, 27 rue Chaligny, 75571 Paris cedex 12, France. E-mail: email@example.com; fax: +33 1 40 01 13 52.
Potential conflict of interest: Nothing to report.
Intrahepatic cholangiocarcinoma (CCA) is characterized by an abundant desmoplastic environment. Poor prognosis of CCA has been associated with the presence of alpha-smooth muscle actin (α-SMA)-positive myofibroblasts (MFs) in the stroma and with the sustained activation of the epidermal growth factor receptor (EGFR) in tumor cells. Among EGFR ligands, heparin-binding epidermal growth factor (HB-EGF) has emerged as a paracrine factor that contributes to intercellular communications between MFs and tumor cells in several cancers. This study was designed to test whether hepatic MFs contributed to CCA progression through EGFR signaling. The interplay between CCA cells and hepatic MFs was examined first in vivo, using subcutaneous xenografts into immunocompromised mice. In these experiments, cotransplantation of CCA cells with human liver myofibroblasts (HLMFs) increased tumor incidence, size, and metastastic dissemination of tumors. These effects were abolished by gefitinib, an EGFR tyrosine kinase inhibitor. Immunohistochemical analyses of human CCA tissues showed that stromal MFs expressed HB-EGF, whereas EGFR was detected in cancer cells. In vitro, HLMFs produced HB-EGF and their conditioned media induced EGFR activation and promoted disruption of adherens junctions, migratory and invasive properties in CCA cells. These effects were abolished in the presence of gefitinib or HB-EGF-neutralizing antibody. We also showed that CCA cells produced transforming growth factor beta 1, which, in turn, induced HB-EGF expression in HLMFs. Conclusion: A reciprocal cross-talk between CCA cells and myofibroblasts through the HB-EGF/EGFR axis contributes to CCA progression. (Hepatology 2013; 58:2001–2011)
signal transducer and activator of transcription 3
standard uptake value
transforming growth factor beta
optimal Tcf-binding site.
Intrahepatic cholangiocarcinoma (CCA) is a highly fatal tumor that arises from biliary epithelial cells. Worldwide, it accounts for 3% of all primary gastrointestinal malignancies. CCA is the second-most common primary hepatic malignancy after hepatocellular carcinoma (HCC). Its incidence has increased drastically over the past few years, even though factors causing this increase are not clear. CCA has a very poor prognosis because of its late clinical presentation and lack of effective nonsurgical therapies.
The tyrosine kinase receptor, epidermal growth factor receptor (EGFR), binds different ligands, including epidermal growth factor (EGF), heparin-binding epidermal growth factor (HB-EGF), and amphiregulin, which initiate intracellular signaling cascades leading to tumor development and progression. Among EGFR ligands, aberrant expression of HB-EGF has been involved in the development of various cancers, including liver carcinoma.[2-4] EGFR activation disturbs cell–cell adhesion by destabilizing the adherens junction complexes (i.e., E-cadherin/β-catenin) and thus contributes to acquisition of a motile, invasive phenotype. EGFR plays a significant role in CCA malignancy. Activating mutations, sustained activation, and overexpression of EGFR (28%-68%) are associated with a poor prognosis in patients with CCA.[6-14] Recently, transcriptional profiling revealed a significant enrichment of the signature related to EGFR activation in a subclass of CCA that displays the most aggressive behavior.[7, 12]
CCA is characterized by a prominent fibrous stroma, the gene signature of which is associated with a poor prognosis. The presence of alpha-smooth muscle actin (α-SMA)-positive fibroblasts (i.e., myofibroblasts; MFs) within CCA stroma referred to as cancer-associated fibroblasts has been correlated with shorter overall and disease-free survival rate.[15-19] MFs, by secreting a variety of soluble factors (i.e., growth factors and cytokines) are considered as active promoters of tumor growth and progression in several cancers. Reciprocal interactions between tumor cells and MFs have been shown.[21, 22] Thus, tumor cells are able to secrete growth factors that act as key mediators of fibroblast activation, such as transforming growth factor beta 1 (TGF-β1).[21, 23]
Although EGFR contributes to CCA progression, the role of EGFR axis in the interaction between MF and CCA cells has not been studied. Here, we show that human liver myofibroblasts (HLMFs) increase CCA growth and progression through EGFR in a xenograft model. HLMFs and stromal MFs in human CCA tumors express HB-EGF. Conditioned media from HLMFs promote invasion of CCA cells through the HB-EGF/EGFR axis. Furthermore, activation of EGFR signaling in CCA cells enhances TGF-β1 expression that, in turn, triggers the expression of HB-EGF by HLMFs. Our data suggest that the HB-EGF/EGFR axis contributes to CCA progression through a reciprocal cross-talk between MF and CCA cells.
Materials and Methods
Isolation and Culture of HLMFs
HLMFs were isolated from liver and characterized as described previously. Liver samples were obtained from 13 patients undergoing partial hepatectomy for colon metastases. Those procedures complied with ethical guidelines stipulated by the French legislation. HLMFs at passage 1 to 3 were seeded in Dulbecco's modified Eagle's medium (DMEM; Life Technologies, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS; PAA, Les Mureaux, France). After 24 hours, cells were serum-starved for 48 hours. HLMF-conditioned media (HLMF-CM) were collected, centrifuged at 2,000×g for 5 minutes, and frozen at −80°C until use.
CCA Cell Culture
Human CCA cell lines Mz-ChA-1, SK-ChA-1, and EGI-1 were used. Mz-ChA-1 and SK-ChA-1 were provided by Dr. A. Knuth (Zurich University, Zurich, Switzerland), and EGI-1 cells were obtained from DSMZ (Braunschweig, Germany). Mz-ChA-1 and SK-ChA-1 cells were cultured in DMEM supplemented with 1 g/L of glucose, 10 mmol/L of HEPES (Life Technologies), and 10% FBS (PAA). EGI-1 cells were cultured in DMEM supplemented with essential and nonessential amino acids and 10% FBS. For starvation, Mz-ChA-1 and EGI-1 cells were incubated in serum-free medium, whereas SK-ChA-1 cells were kept in medium containing 0.5% FBS. Gefitinib or neutralizing antibodies (Abs) were added to the medium 30 minutes before treatment with HLMF-CM or HB-EGF and maintained during stimulation. CM were obtained from CCA cells grown to ≈75% confluence and serum-starved for 24 hours before medium collection.
Materials, human CCA samples and immunohistochemistry (IHC), flow cytometry, mouse xenografting experiments and positron emission tomography (PET) imaging, IHC on mouse liver tissues, liver micrometastasis detection assay, enzyme-linked immunosorbent assay, real-time polymerase chain reaction (PCR) analysis, immunoblotting and immunofluorescence (IF) experiments, β-catenin transcriptional activity assay, cell proliferation, migration and invasion assays, and the statistical analysis are described in the Supporting Information.
HLMFs Promote CCA Tumor Growth In Vivo
To determine the contribution of MFs on CCA biology, we performed cotransplantation experiments of CCA cells (i.e., Mz-ChA-1 cells) with primary MFs isolated from human liver (HLMFs) in a subcutaneous (SC) xenograft model into nude mice. HLMFs in primary culture were morphologically activated and expressed α-SMA and were negative for CD31 and HepPar1 (Supporting Fig. 1A,B). Mz-ChA-1 cells overexpressed EGFR, as compared to nonmalignant biliary epithelial cells (Supporting Fig. 1C).
CCA cells were injected alone or in combination with HLMFs. Eight days postinjection, only mice inoculated with CCA cells and HLMFs presented palpable tumors. HLMFs boosted CCA tumor growth at any time post–cell injection with an average 4-fold increase (Fig. 1A, gray versus white bars) and an 8-fold increase in tumor weight at time of sacrifice (48 days postinjection; Fig. 1B, gray versus white bars). We also observed that the presence of HLMFs increased tumor take rate (Fig. 1C). Tumors developed in xenografted mice were histologically similar to human CCA, because they showed a prominent stromal compartment attested by α-SMA staining. EGFR staining was exclusively detected in CCA cells (Fig. 1D).
HLMFs Promote CCA Tumor Growth Through EGFR
We next examined whether EGFR played a role in the enhancing effect of HLMFs on CCA growth by treating mice with gefitinib, a specific inhibitor of EGFR tyrosine kinase activity (Fig. 1A,B,E). From 8 days of treatment with gefitinib and until the end of the experiment (20 days of treatment), a significant growth reduction was observed in coinjected tumors, compared to vehicle-treated mice (Fig. 1A, black versus gray bars). Gefitinib decreased coinjected tumor weight with an average of 4-fold (Fig. 1B, black versus gray bars). Representative images of three tumors from each group are shown in Fig. 1E. EGFR activation, attested by its phosphorylation level status on tyrosine 1173, was detected in coinjected tumors, but not in CCA cell tumors. Gefitinib treatment abolished EGFR phosphorylation in coinjected tumors (Fig. 1F).
Tumor glucose metabolism, which reflects cell viability, was examined by monitoring 18FDG (fluorodeoxyglucose) uptake with positron emission tomography (PET) imaging. A good correlation (R = 0.95) was observed between tumor volume estimated with a caliper and PET imaging (data not shown). A significant increase of 18FDG uptake (+40%), reflected by the mean of SUV (standard uptake value), was observed in coinjected tumors, as compared with CCA cell tumors (Fig. 2A, middle and upper panels; quantifications on the right). Mice treatment with gefitinib decreased 18FDG uptake in coinjected tumors (Fig. 2A, lower panel; quantification on the right). To evaluate cell proliferation rate, tumors were immunostained with an anti-Ki67 Ab (Fig. 2B, left panels). Ki67 staining was almost exclusively observed in tumor cells. Coinjected tumors had a significantly higher number of Ki67-positive cells than CCA cell tumors (67% versus 20%). In coinjected tumors from mice treated with gefitinib, the number of Ki67-positive cells was markedly decreased (23%) (Fig. 2B, right panel).
HLMF Promote CCA Tumor Progression Through EGFR
Next, we assessed local tumor invasion by evaluating angiogenesis in SC tumors. Microvessel density evaluated with CD31 staining was increased by 2.2-fold in coinjected tumors, as compared with CCA cell tumors (Fig. 3A, upper and middle panels; quantifications on the right). The effect was lost when mice bearing coinjected tumors were treated with gefitinib (Fig. 3A, lower panel). These data were confirmed by measuring CD31 messenger RNA (mRNA) levels by quantitative reverse-transcriptase polymerase chain reaction (qRT-PCR; Fig. 3B). CCA micrometastases into the liver were evaluated by quantifying the presence of human Alu sequences. Amounts of human Alu sequences were 5-fold higher in livers from mice bearing coinjected tumors than in livers from mice bearing CCA cell tumors (Fig. 3C, gray versus white circles). Treatment of mice bearing coinjected tumors with gefitinib reduced the presence of human Alu sequences in liver (Fig. 3C, black versus gray circles). Altogether, these data suggest that in vivo HLMFs contribute to the growth and progression of CCA through EGFR-dependent signaling.
HB-EGF Expression in MF of Human CCA
HB-EGF is an EGFR ligand that has been involved in the cross-talk between MF and tumor cells in uterine cervical cancers. We first analyzed the expression of HB-EGF in MF in paraffin-embedded serial sections from 10 human CCA samples (Fig. 4). The presence of MF in tumor stroma was attested by an α-SMA-positive staining. HB-EGF was expressed both in MF and in tumor cells (Fig. 4; arrowheads indicate MF). EGFR staining revealed that this receptor was expressed by tumor cells mainly at the plasma membrane and was not detected in MF. In primary cultures of HLMF obtained from 13 independent liver preparations, HB-EGF was secreted into cell supernatants (22.61 ± 4.91 pg/mL).
HLMF Promote EGFR Activation in CCA Cells
In vitro analyses were performed to further study the interplay between HLMF and CCA cells through EGFR. We investigated whether HLMF activated EGFR in Mz-ChA-1, SK-ChA-1, and EGI-1 cells, which overexpress EGFR (Supporting Fig. 1C). HLMF-CM incubated with CCA cell lines increased the phosphorylation of EGFR, signal transducer and activator of transcription 3 (STAT3), and extracellular signal-regulated kinase 1/2 (ERK1/2; Fig. 5A). EGFR activation was decreased in the three CCA cell lines when HLMF-CM were mixed with neutralizing Abs against either EGFR or HB-EGF (Fig. 5B, and Supporting Figs. 3A and 4A). All these data suggest that HLMF activate EGFR in CCA cells through the paracrine secretion of HB-EGF.
HLMF Cause CCA Cell Scattering, Migration, and Invasion Through EGFR Signaling
Data obtained from cotransplantation experiments in nude mice showed increased proliferation of CCA cells in the presence of HLMF (Figs. 1A and 2A). Consistently, stimulation with HLMF-CM increased Ki67 immunostaining in CCA cells. This effect was abolished in the presence of gefitinib (Supporting Fig. 2A). However, no significant effect on CCA cell proliferation evaluated by the Ki67 index was observed upon stimulation with HB-EGF per se in CCA cell lines (Supporting Fig. 2B).
Next, effects of HLMF-CM on CCA cell migration and invasion were investigated. Upon incubation with HLMF-CM for 24 hours, the three CCA cell lines displayed a fibroblast-like phenotype and scattered (Fig. 5C and Supporting Figs. 3B and 4B). This effect was abrogated with a neutralizing Ab against HB-EGF or EGFR as well as with gefitinib (Fig. 5C and Supporting Figs. 3B and 4B). CCA cell dispersion induced by HLMF-CM was secondary to the disruption of cell-cell junctions, as evidenced by E-cadherin internalization from the plasma membrane to the cytoplasm (Fig. 6A and Supporting Figs. 3C and 4C) and by β-catenin translocation from the plasma membrane to cytoplasm and nucleus (Fig. 6B). The effects of HLMF-CM were mimicked by exogenously added HB-EGF (Fig. 6A,B). Nuclear localization of β-catenin upon treatment with HLMF-CM or HB-EGF was attested by its increased transcriptional activity in optimal Tcf-binding site/far-from-optimal Tcf-binding site (TOP/FOP) luciferase assays (Fig. 6C). We observed that HLMF-CM also increased the migratory (Fig. 6D, left panel) and invasive (Fig. 6D, right panel) properties of CCA cells. All these effects were significantly abolished by gefitinib. Altogether, these results support our in vivo data and suggest that HLMF promote the acquisition of an invasive phenotype by CCA cells through EGFR activation.
TGF-β1 Is Synthesized by CCA Cells and Induces HB-EGF Expression in HLMF
We further investigated whether CCA cells could affect HLMF functions. CM were prepared from CCA cells (i.e., Mz-ChA-1 cells; CCA cell-CM) and added to primary cultures of HLMF (Fig. 7A-E). HLMF proliferation was evaluated by real-time monitoring of cell index (Fig. 7A). CCA cell-CM had no effect on HLMF cell index, whereas platelet-derived growth factor (PDGF), a well-known inducer of MF proliferation, increased this index (Fig. 7A). Evidence indicates that cancer-cell–secreted factors, such as TGF-β1, modulate MF activation in the tumor microenvironment. In CCA, TGF-β1 was expressed by CCA cells and its receptor, TGF-β RII, was detected both in carcinoma cells and stromal MF (Supporting Fig. 5A). Addition of exogenous TGF-β1 increased α-SMA mRNA level (Supporting Fig. 5B, left panel) and induced HLMF activation (Supporting Fig. 5B, right panel). Effect of TGF-β1 was mimicked by CCA cell-CM that also up-regulated α-SMA mRNA level (Fig. 7B). This effect was abolished by the addition of a neutralizing Ab against TGF-β1 in CCA cell-CM (Fig. 7B). Consistently, CCA cell-CM led HLMF to spread out with more elongations and thus acquire a more myofibroblastic appearance (Fig. 7C). Modification of HLMF morphology was inhibited by TGF-β1 neutralizing Ab (Fig. 7C).
Furthermore, TGF-β1 markedly enhanced HB-EGF mRNA level in HLMF with an average of 22-fold (Fig. 7D). CCA cell-CM also increased HB-EGF mRNA level with an average of 8-fold in HLMF that was significantly reduced by TGF-β1 neutralizing Ab (Fig. 7D). Interestingly, TGF-β1 expression in CCA cells was enhanced upon HB-EGF stimulation (Fig. 7E). These data suggest that TGF-β1 produced by CCA cells may favor HLMF activation that, in turn, expressed increased level of HB-EGF.
The importance of the local stroma in tumor growth and progression has been recognized in several cancers. However, little is known about the contribution of the MFs to CCA progression. This is particularly unfortunate because CCA is characterized by a prominent desmoplastic stroma enriched in α-SMA-positive MF, of which the presence and gene signature have been associated with poor pronosis.[12, 18, 19] Here, we provide evidence that HLMFs contribute to CCA growth and progression, and that EGFR-dependent reciprocal exchanges occur between the two cellular compartments. All these findings are recapitulated in Fig. 8.
Stromal components, such as MF, participate toward tumor growth and progression by feeding cancer cells with multiple growth factors. In CCA, only a few studies have explored the signaling pathways involved in the exchanges between MF and cancer cells in CCA progression. The stromal-derived factor-1 (SDF-1)/CXR4 axis has been recently identified as one of these pathways.[29-31] Findings from Fingas et al. also emphasized the role of MF-derived PDGF-BB in CCA cell protection from TRAIL cytotoxicity through a Hedgehog-dependent signaling pathway. Recently, Cadamuro et al. have demonstrated that PDGF-D secreted by CCA cells promoted recruitment of MF through its cognate receptor, PDGF-Rβ, in human CCA.
To demonstrate the contribution of the EGFR-dependent signaling pathway in the interplay between MF and cancer cells, tumor xenograft experiments were performed in immunodeficient mice. HLMF promote a marked increased of CCA tumor growth and progression. A specific inhibitor of EGFR kinase activity, gefitinib, abrogated this effect. In vitro, we used CM from HLMF to highlight the role of MF on proliferation and invasion of CCA cells through EGFR. To our knowledge, this is the first report demonstrating the contribution of EGFR in the promotion of carcinoma tumor development by MFs and, more specifically, in CCA. Beyond EGFR, other members of the EGFR family, such as HER-3 and its ligand, heregulin-1, have been involved in the cross-communication between stromal and tumoral cells in several cancers, including colorectal, gastric, and pancreatic cancers. Thus, the results of earlier studies, together with the present findings, suggest that the HER/ErbB family, to which EGFR belongs, is likely to play a significant role in the establishment of cellular interactions between stromal MF and malignant cells in several cancers, including CCA.
Among EGFR ligand secreted by MF in cancer, HB-EGF has emerged as a paracrine factor that contributes to intercellular communications between MF and tumor cells in uterine cervical and breast carcinoma. In human CCA specimens, HB-EGF immunoreactivity was detected in MF. In addition to MF, we also detected an expression of HB-EGF in tumor cells. Therefore, we can assume that HB-EGF participates in the autocrine and paracrine activation of EGFR. HB-EGF produced by MF is likely to act only on tumor cells in CCA because EGFR was only detected in these cells. The expression of HB-EGF by MF prompted us to hypothesize that MF may constitute an additional source of ligands required to activate EGFR on the cancer cell surface. EGFR heterodimerizes with other receptors of the ErbB/HER family (i.e., ErbB/HER2 and ErbB/HER3). Upon stimulation of CCA cells with HB-EGF, EGFR and, to a lesser extent, ErbB/HER2 and ErbB/HER3 are activated (data not shown). Thus, a potential contribution of ErbB/HER2 and/or ErbB/HER3 through EGFR heterodimerization cannot be excluded in the cross-talk between MF and tumor cells in CCA.
To date, the role of HB-EGF in CCA has not been explored. In vitro, an HB-EGF-neutralizing Ab inhibited activation of EGFR and dispersion of CCA cells in response to HLMF-CM. Consistently, exogenous addition of HB-EGF to CCA cells caused cell migration and invasion, as previously described in many cancers.[37, 38] Although HB-EGF activated EGFR and downstream pathways, including ERK1/2, we were unable to show an effect of HB-EGF on CCA cell proliferation. Thus, stimulation of EGFR by HB-EGF in CCA cells is likely to play a role in tumor invasion and metastasis, which is consistent with the IHC and genomic profiling studies that demonstrated high EGFR expression in patients with aggressive phenotype and poor prognosis CCA.[7, 11, 12, 14, 39] In addition to EGFR overexpression, Sia et al. have recently showed an enrichment of EGFR activation in a subgroup of CCA. From our studies, we may hypothesize that activation of EGFR is related to EGFR ligand produced by stroma cells. It would be worthwhile to explore the gene expression profiling of stroma in this subgroup of CCA tumors.
Through the production of soluble factors, cancer cells have the ability to communicate with stromal myofibroblasts located arround them. This point has been stressed in several cancers, including HCC, colorectal, uterine cervical cancers, and in CCA. Our results showed that EGFR activation in CCA cells promotes the expression of TGF-β1. TGF-β1 is expressed in a vast majority of CCA. As previously reported,[41-43] we detected TGF-β1 in carcinoma cells and its receptor, TGF-β RII, both in carcinoma and stromal MF. Recently, Andersen et al. have reported on an enrichment of TGF-β receptor in the CCA stromal compartment. TGF-β1 derived from cancer cells has been shown to promote MF activation and secretion of growth factors (i.e., hepatocyte growth factor, heregulin) in colon and squamous carcinoma. Here, we showed that in HLMF, TGF-β1 stimulates HB-EGF synthesis. Thus, CCA cell-derived TGF-β1 may sustain MF in an activated state to produce HB-EGF and thereby maintain the HB-EGF/EGFR pathway active in the tumor cells.
In conclusion, the present study provides evidence of the existence of a cross-communication between cancer cells and MF in CCA tumor based on the HB-EGF/EGFR axis. These data reinforce the notion that the EGFR system plays a crucial role in CCA progression. Therapies targeting EGFR (erlotinib and/or cetuximab) in combination with GEMOX have been tested in clinical trials for CCA treatment.[46-48] Despite encouraging results, EGFR therapies have shown restricted efficiency in patients with CCA. Our study suggests that EGFR-targeted therapies could be more effective in CCA in subgroups of patients showing marked EGFR expression and prominent stroma.
The authors thank the Tumeur-Est tissue bank for cholangiocarcinoma human samples. The authors also thank Colette Rey (INSERM, UMRS 938, Centre de Recherche Saint-Antoine) for assistance in animal experiments, Dr. Françoise Praz (INSERM, UMRS 938, Centre de Recherche Saint-Antoine) for technical assistance for Alu sequence detection, and Dr. Bruno Saubaméa (Cellular and Molecular Imaging facility of the IFR71-IMTCE, Paris Descartes University) for confocal imaging.