Background The transdifferentiation of hepatic stellate cells (HSCs) into myofibroblasts is a major mechanism for stroma development in hepatic metastasis, but their regulatory pathways remain unclear. Transdifferentiated HSCs from fibrotic liver express high levels of the fibrillar collagen receptor discoidin domain receptor 2 (DDR2), but it is unclear if DDR2 plays a direct profibrogenic role in the tumour microenvironment.
Aim To assess the impact of DDR2 on the prometastatic role of HSC-derived myofibroblasts.
Methods Hepatic metastases were induced in DDR2−/− and DDR2+/+ mice by intrasplenic injection of MCA38 colon carcinoma cells, and their growth and features were characterised. Stromagenic, angiogenic and cancer cell proliferation responses were quantified in metastases by immunohistochemistry. The adhesion-, migration- and proliferation-stimulating activities of supernatants from primary cultured DDR2−/− and DDR2+/+ HSCs, incubated in MCA38 cell-conditioned medium, were evaluated in primary cultured liver sinusoidal endothelium cells (LSECs) and MCA38 cells. Gene expression signatures from freshly isolated DDR2−/− and DDR2+/+ HSCs were compared and DDR2-regulated genes were studied by RT-PCR under basal conditions and after stimulation with MCA38 tumour-conditioned media.
Results Metastases were increased three fold in DDR2−/− livers, and contained a higher density of α−smooth muscle actin-expressing myofibroblasts, CD31-expressing microvessels and Ki67-expressing MCA38 cells than metastases in DDR2+/+ livers. Media conditioned by MCA38-activated DDR2−/− HSCs significantly increased adhesion, migration and proliferation of LSECs and MCA38 cells, compared with DDR2+/+ HSCs. DDR2 deficiency in HSCs led to decreased gene expression of interferon γ-inducing factor interleukin (IL)-18 and insulin-like growth factor-I; and increased gene expression of prometastatic factors IL-10, transforming growth factor (TGF)β and vascular endothelial growth factor (VEGF), bone morphogenetic protein-7 and syndecan-1. MC38 tumour-conditioned media further exacerbated expression changes in DDR2-dependent IL-10, TGFβ and VEGF genes.
Conclusion DDR2 deficiency fosters the myofibroblast transdifferentiation of tumour-activated HSCs, generating a prometastatic microenvironment in the liver via HSC-derived factors. These findings underscore the role of stromal cells in conditioning the hepatic microenvironment for metastases through altered receptor–stroma interactions.
- liver metastasis
- colon carcinoma
- extracellular matrix
- collagen receptor
- hepatic stellate cells
- tumour stroma
- hepatic metastasis
- cell matrix interaction
- hepatic metastases
- colorectal cancer
- hepatic stellate cell
- endothelial cells
Statistics from Altmetric.com
- liver metastasis
- colon carcinoma
- extracellular matrix
- collagen receptor
- hepatic stellate cells
- tumour stroma
- hepatic metastasis
- cell matrix interaction
- hepatic metastases
- colorectal cancer
- hepatic stellate cell
- endothelial cells
Significance of this study
What is already known about this subject?
Transdifferentiation of hepatic stellate cells (HSCs) into myofibroblasts is a central event in the fibrogenic responses to hepatic injury induced by non-neoplastic and neoplastic processes that operate under little-known mechanisms.
DDR2 is a non-integrin-type tyrosine kinase receptor for fibrillar collagen.
DDR2 expression increases during HSC activation in hepatic fibrosis.
DDR2 regulates the release of matrix metalloproteinases by HSCs.
What are the new findings?
This study shows for the first time that DDR2 deficiency predisposes hepatic tissue to experimental colon carcinoma metastasis.
The mechanism depended on the enhanced myofibroblast transdifferentiation response of tumour-activated HSCs in DDR2-deficient mice. This in turn generated a prometastatic microenvironment via HSC-derived factors, which enhanced tumour growth and angiogenesis during colorectal metastasis to the liver.
Such prometastatic behaviour was consistent with an altered gene expression profile in DDR2−/− HSCs, which included downregulation of anti-fibrogenic and anti-tumour immune response genes, and upregulation of immune suppressant, profibrogenic and prometastatic genes. More importantly, expression of several of these genes was further altered when HSCs were cultured in the presence of media conditioned by cultured MCA38 cells.
How might it impact on clinical practice in the foreseeable future?
DDR2 may act as a hepatic metastasis suppressor factor operating through tumour-activated HSCs, and patients with colorectal cancer with hepatic DDR2 expression deficiency may be at higher risk of metastasis.
However, these conclusions are based on an experimental metastasis model and in vitro studies that have limitations for a direct clinical translation.
The transdifferentiation of hepatic stellate cells (HSCs) into myofibroblasts is a central event in the fibrogenic responses to hepatic injury induced by non-neoplastic1 and neoplastic processes.2–5 Major features of fibrogenic HSCs are the expression of myofibroblastic marker α−smooth muscle cell actin (αSMA) and tyrosine kinase receptors such as platelet-derived growth factor receptorβ6 and discoidin domain receptor 2 (DDR2)7; the proliferation and migration into areas of tissue injury8; extracellular matrix production and remodelling9 and the secretion of multiple soluble factors that regulate the migration and proliferation of other cell types, including liver sinusoidal endothelium cells (LSECs),10 parenchymal cell progenitors11 and even cancer cells.10 ,12 ,13
Tumour-derived factors can efficiently activate myofibroblast generation from HSCs,4 contributing to tumour stroma and angiogenesis formation,10 which in turn supports cancer and metastasis progression in the liver.13 However, how tumorigenic and prometastatic effects of HSC-derived stromal myofibroblasts are operating is not yet clear. Moreover, whether tumorous and non-tumoral mechanisms of HSC activation and fibrogenesis share the same regulatory pathway(s) is also unknown. This might have diagnostic and therapeutic implications for patients at risk of primary and secondary hepatic cancer occurrence.
DDR2 is a non-integrin-type tyrosine kinase receptor for fibrillar collagen.14 DDR2 expression increases during HSC activation in hepatic fibrosis,7 and retroviral infection of chimeric DDR2 receptors has indicated that DDR2 regulates the release of matrix metalloproteinases (MMPs) by HSCs.7 However, it is unclear whether DDR2 expression plays a profibrogenic role in the tumour microenvironment.
In this work, we studied the prometastatic phenotype of tumour-activated HSCs during the experimental hepatic colonisation of MCA38 murine colon carcinoma in DDR2+/+ and DDR2−/− mice. Additionally, we compared the effects of DDR2+/+ and DDR2−/− tumour-activated HSCs on the adhesion, proliferation and migration of primary cultured LSEC and cancer cells. Finally, we determined the gene expression signatures of freshly isolated DDR2−/− and DDR2+/+ HSCs, and DDR2-regulated genes under basal conditions and after tumour-conditioned medium stimulation.
Our results demonstrate that DDR2 deficiency remarkably increased metastatic efficiency of MCA38 colon cancer cells in the liver. The mechanism depended on the enhanced myofibroblast transdifferentiation response of tumour-activated HSCs in the absence of DDR2 expression. This in turn generated a prometastatic microenvironment via HSC-derived factors that enhanced tumour growth and angiogenesis during colorectal metastasis to the liver. Furthermore, such prometastatic behaviour was reflected in an altered gene expression profile of DDR2−/− HSCs, including upregulation of key immune suppressant, profibrogenic and prometastatic genes such as interleukin (IL)-10, vascular endothelial growth factor (VEGF) and transforming growth factor (TGF)β.
Materials and methods
DDR2−/− mice have been described elsewhere.15 Hepatic metastases were produced by intrasplenic injection of 1.5×104 MCA38 cells/gc body weight13 into DDR2−/− and DDR2+/+ male mice. Mice were killed 14 days later. All animals received humane care according to the criteria outlined in the “Guide for the Care and Use of Laboratory Animals” published by the NIH.
Immunohistochemistry and image analysis
Liver sections (10 μm thick) fixed in cold acetone were processed for haematoxylin/eosin stain or allowed to react with 1:50 dilution of anti-human αSMA monoclonal antibody (Sigma Chemicals, St Louis, Missouri, USA), rat anti-mouse CD31 polyclonal antibody (1:100, Dako, Denmark) and rabbit anti-mouse desmin polyclonal antibody (1:200, Dako, Denmark), followed by the appropriate fluorescent secondary antibodies.13 Some sections were incubated with anti-human Ki67 monoclonal antibody (1:200, Sigma Chemicals). Cell nuclei were counterstained with 4',6-diamidino-2-phenylindole (DAPI). Irrelevant appropriate immunoglobulins were used as negative controls. αSMA, desmin and CD31 triple immunostained (n=72) and ki67/DAPI stained (n=36) tumour micrometastasis of 0.5–2 mm2 average diameter were microphotographed under same microscopy conditions to determine myofibroblast and angiogenic cell recruitment and cell proliferation indices, respectively. Specific immunostaining was quantified using the Analysis V3.2 program (Olympus Soft Imagine Solutions GMBH, Munster, Germany). Results were expressed as percentage of specifically coloured tissue area relative to the whole micrometastatic foci area.
Isolation and primary culture of HSCs and sinusoidal endothelium cells
Hepatic sinusoidal cells were isolated as described.16 In brief, enzymatic perfusion was followed by isopycnic gradient centrifugation of dispersed liver cells placed on top of a triple layer of Percoll solutions (50%, 31% and 24%). HSCs were collected from the 24/31% interphase and LSECs from the 31/50% interphase. HSCs were cultured on plastic at 1.5×105 cells/ml of RPMI-1640 media. LSECs were shed at 1.5×106 cells/ml onto 0.001% collagen type I matrix, unless otherwise advised. HSC and LSEC primary cultures were recognised as pure lineages by vitamin A autofluorescence (for HSCs) and specific expression of CD31 in the absence of CD45 (for LSECs).
Generation of media conditioned by untreated and tumour-activated HSCs
To obtain MCA38-conditioned medium (MC38A-CM), supernatants from subconfluent MCA38 colon carcinoma cell cultures, maintained for 24 h in serum-free RPMI media, were diluted 1:1 with fresh media. HSCs isolated 2 days before were maintained for 24 h in basal serum-free media or in MC38-CM to generate untreated HSC-CM and tumour-activated HSC-CM, respectively.
MCA38 proliferation assay
MCA38 cells were treated for 48 h with RPMI media or with untreated HSC-CM and tumour-activated HSC-CM. Proliferation was measured by the MTT method, as previously described.17
Chemotactic cell migration assay
Twenty-five hundred thousand cells were placed on top of 0.001% type I collagen-coated inserts with 8 μm pores in RPMI supplemented with 0.5% fetal calf serum, as previously described.4 Inserts were placed on top of 2 cm2 wells containing basal media, untreated HSC-CM or tumour-activated-HSC-CM plus 0.5% fetal calf serum. Cells which had migrated were counted after 20 h (for LSECs) or 6 h (for MCA38) in 10 20×-fields per insert.
MCA38 cell adhesion to LSECs
An adhesion assay was performed as previously described18 with minor modifications. In brief, monolayer LSEC cultures were treated for 2 h with RPMI media or conditioned media from tumour-activated or untreated HSCs. Then, MCA38 cancer cells labelled with BCEF-AM (Molecular Probes, Denmark) were added to the LSEC cultures. After 30 min, co-cultures were washed and the level of MCA38-LSEC adhesion was measured as fluorescent emission at 630 nm. Results were expressed as percentage of MCA38 cells adhering to untreated LSEC monolayers.
Gene array analysis and quantitative reverse transcription-PCR in DDR2−/− and DDR2+/+ HSCs
Total RNA from primary HSCs was extracted using RNeasy columns (Qiagen GmbH, Hilden, Germany). RNA quantity and quality were analysed with a capillary 2100 Bioanalyzer (Agilent Technologies, Palo Alto, California, USA). Total RNA from DDR2−/− HSCs and DDR2+/+ HSCs were labelled with Cy3-CTP and Cy5-CTP fluorochromes (LRILAK plus, two-colour; Agilent). A custom microarray GE 4×44K designed upon a commercially available Whole Mouse Genome Microarray Kit, a Genechip Scanner 3000 7G and the analysis program “GeneSpring GX v9.5” were used (Agilent Technologies). DDR2−/− HSCs were compared with DDR2+/+ HSCs under basal conditions and after tumour-conditioned stimuli. To validate GeneChip expression data, real-time quantitative reverse transcriptase-PCR was done with the same RNA samples used in the GeneChip array experiment, and the genes having significant expression differences between DDR2−/− and DDR2+/+ HSCs were targeted. Suitable Assays-on-Demand primers and probe sets (Applied Biosystems, Foster City, California, USA) for 18S RNA and the genes of interest were obtained. These consisted of two, unlabelled PCR primers and an FAM dye-labelled TaqMan Minor Groove Binder probe. Two-step real-time reverse transcriptase-PCR with 5_ nuclease chemistry was carried out by adding an aliquot of cDNA to TaqMan Universal PCR Master Mix, and quantitative gene expression data were acquired on an Applied Biosystems ABI PRISM 7900HT Sequence Detection System (Applied Biosystems).
Statistical results refer to mean±SD. Statistical analysis was performed by SPPS statistical software for Microsoft Windows (Professional Statistics). Individual comparisons were made with a Student two-tailed unpaired t test. Significant changes in gene expression were identified by unpaired t test and subsequent Benjamin–Hochberg correction using a p value of 0.05 as criterion of statistical significance.
Hepatic MCA38 colon carcinoma metastasis development in DDR2−/− mice
Both the hepatic volume occupied by MCA38 colon carcinoma metastases and the number of metastatic foci per area unit were significantly increased, three- and 2.5fold, respectively, in DDR2−/− compared with DDR2+/+ mice (figure 1A,B, columns). Based on size distribution of metastasis, DDR2−/− livers contained a higher number of foci than DDR2+/+ for all size groups, indicating that more favourable conditions were operating at both implantation and growth stages of the MCA38 colon carcinoma metastasis in the liver of DDR2-deficient mice (figure 1A,B, doted lines).
Confocal microscopy on triple immunohistochemically stained tissue sections plus immunofluorimetric image analysis on cells labelled with anti-CD31, anti-αSMA and anti-desmin antibodies was used to study the stromagenic and neoangiogenic status of hepatic MCA38 metastasis in DDR2−/− compared with DDR2+/+ mice. As shown in figures 2A,B, MCA38 micrometastases at a panlobular stage (0.5–2 mm2 in diameter) were mainly of replacement growth-type,19 and led to a sinusoidal-type stromal pattern built by desmin/αSMA-coexpressing HSC-derived myofibroblasts.20 This occurred irrespective of hepatic DDR2 gene expression. However, figure 2A,C and figure 3A,C, demonstrate that myofibroblast density in DDR2−/− hepatic metastases was significantly higher than their density in metastases from DDR2+/+ mice, which reflects an enhanced recruitment of HSC-derived myofibroblasts into the DDR2−/− metastatic foci.
Previously, analysis of CD31 and desmin/αSMA expression demonstrated that tumour-activated HSCs play a proangiogenic role during liver metastasis.13 In the current model, DDR2−/− hepatic metastases contained more CD31-expressing cells per unit area than metastases from DDR2+/+ mice (figures 2C and 3C), indicating higher recruitment of angiogenic hepatic sinusoidal endothelial cells into metastases from DDR2−/− mice. Moreover, the Ki67-expressing cell fraction also increased in DDR2−/− hepatic metastases compared with DDR2+/+ ones (figures 2D and 3D), indicating a higher cancer cell proliferation rate in highly angiogenic hepatic metastases from DDR2−/− mice.
Secretion of endothelial cell adhesion- and migration-stimulating factors by basal and tumour-activated HSC from DDR2−/− and DDR2+/+ mice
As shown in figure 1, hepatic MCA38 metastasis number was higher in DDR2−/− mice, suggesting that the hepatic microenvironment provided by DDR2−/− mice promoted cancer cell implantation in the liver. Adhesion of circulating cancer cells to LSECs significantly contributes to hepatic metastasis implantation,21 and activated HSCs release soluble factor(s) that enhance(s) adhesiveness of LSECs for cancer cells.13 Although highly purified rat and mouse LSEC fractions did not express DDR2 (data not shown), we hypothesised that DDR2 deficiency in HSCs might abrogate inhibitory effects of resting HSC-derived factors on LSEC adherence for circulating cancer cells. As shown in figure 4A, addition of conditioned media from untreated DDR2+/+ HSC to primary cultured LSECs did not significantly alter the endothelial adherence rate of MCA38 cells, compared with the adherence of untreated LSECs. However, MCA38 cancer cell adhesion to LSECs incubated with conditioned medium from tumour-activated DDR2+/+ HSCs increased by 120% compared with MCA38 adhesion to untreated LSECs. More importantly, when LSECs received untreated DDR2−/− HSC-CM the adhesion of MCA38 cells increased by 160%, and, when LSECs received tumour-activated DDR2−/− HSC-CM MCA38 cell adhesion increased by 360% with respect to their adhesion to untreated LSECs. Therefore, the endothelial cell adhesion-stimulating activity of untreated DDR2−/− HSCs was as potent as the activity of tumour-activated DDR2+/+ HSCs, and this capability was further induced by tumour-derived factors in DDR2-deficient HSCs, which is consistent with the higher hepatic metastasis implantation rate seen in DDR2−/− mice.
Tumour-activated HSCs also release migration factors for LSECs that promote angiogenesis in the hepatic metastasis microenvironment.13 As shown in figure 4B, LSEC migration-stimulating activity of DDR2−/− HSC cultured under basal conditions was once again as potent as the activity of tumour-activated DDR2+/+ HSC, and this capability of DDR2−/− HSC was further induced by tumour factors as well. These results are consistent with the high CD31-expressing cell recruitment detected in hepatic metastases from DDR2−/− mice (figures 2C and 3C).
DDR2 deficiency leads to increased MCA38 cancer cell migration and enhanced proliferative factors by basal and tumour-activated HSCs
Figure 2 indicates that the main mechanism of hepatic MCA38 colon cancer metastasis development was of the hepatic tissue replacement-type. Given the efficient invasive behaviour of cancer cells during this growth mechanism in the liver, the secretion of cancer cell migration-and proliferation-stimulating factors by HSC from DDR2−/− and DDR2+/+ mice was evaluated. Conditioned medium from basal conditions-cultured DDR2+/+ HSCs did not significantly alter MCA38 cell migration rate, compared with untreated MCA38 cells (figure 4C). Conditioned media from tumour-activated DDR2+/+ HSC slightly stimulated migration MCA38 cells compared with media from untreated DDR2+/+ HSCs, while conditioned media from untreated DDR2−/− HSC increased MCA38 cell migration by 60% and from tumour-activated DDR2−/− HSCs it enhanced migration by 90%. Therefore, the secretion of cancer cell migration-stimulating factor that occurred in DDR2−/− HSCs was even further stimulated by tumour-derived factors. This is consistent with cancer cell migration activity performed by MCA38 cancer cells colonising the liver through a replacement-type metastasis mechanism.
As previously reported, tumour-activated HSCs also release proliferative factors for cancer cells that promote growth of metastases.4 As shown in figure 4D, MCA38 cell proliferation-stimulating activity also significantly increased in the supernatant from basal condition-cultured DDR2−/− HSCs more than in the supernatant from tumour-activated DDR2+/+ HSCs. Again, the highest proliferation of MCA38 cells was achieved when they were treated with supernatants from tumour-activated DDR2−/− HSCs. These results are also in agreement with the higher density of Ki67-expressing cells in the hepatic metastases from DDR2−/− mice (figures 2D and 3D).
Identification and validation of DDR2-dependent genes in HSCs
HSCs from DDR2−/− and DDR2+/+ mice were isolated and cultured for 3 days and their total RNAs were isolated and hybridised onto DNA microarrays for gene expression analysis. From the 41 406 analysed genes, 1712 were twofold or more upregulated in the DDR2−/− cells compared with DDR2+/+ cells, while 671 genes were twofold or more downregulated (figure 5A). Next, genes that had previously been related to key mechanisms in the hepatic metastasis process were analysed by quantitative real time RT-PCR comparing HSCs from DDR2−/− and DDR2+/+ under both basal culture conditions and after tumour-conditioned media stimulation (figure 5B). Under basal culture conditions DDR2 deficiency already involved altered HSC expression of key genes associated with immune response regulation: it decreased immune-stimulating factor IL-18,22 while it increased immune-suppressant factors IL-1023 and TGFβ.24 It also involved decreased expression of the insulin-like growth factor (IGF)-I gene,25 while expression level of pro-metastatic genes VEGF,26 bone morphogenetic protein 727 and syndecan-128 significantly increased. More importantly, IL-10, VEGF and TGFβ gene expression further increased in HSCs from DDR2−/− mice, compared with those from DDR2+/+ mice, when they were treated with MCA38 cancer cell-conditioned medium (figure 5B).
We have shown that host DDR2 deficiency remarkably increases the hepatic colonisation efficiency of MCA38 colon cancer cells. Experimental metastases that developed in the DDR2-deficient hepatic microenvironment preferentially grew through a replacement-type mechanism, and contained a sinusoidal-type angiogenesis, supported by HSC-derived myofibroblasts. Compared with those from wild-type mice, hepatic metastases from DDR2-deficient mice contained a higher density of αSMA-expressing cells, suggesting that the HSC transdifferentiation-stimulating activity elicited by tumour-derived factors4 was fostered in DDR2-deficient mice. Second, DDR2-deficient mice also contained a higher density of CD31-expressing cells, which supports previous findings on the proangiogenic role of tumour-activated HSCs in hepatic metastases.13 And third, they had a higher density of Ki67-expressing cells, which suggests a higher cancer cell proliferation rate.
Consistent with these in vivo findings, the supernatant from primary cultured tumour-activated DDR2−/− HSCs was enriched by soluble factors promoting adhesion, migration and proliferation of both sinusoidal endothelial cells and MCA38 cells, compared with supernatants from DDR2+/+ HSCs. Interestingly, the gene expression profile of primary cultured DDR2−/− HSCs showed constitutive downregulation of key genes contributing to anti-tumour defence, and upregulation of other genes supporting tumour growth and angiogenesis. Moreover, this altered gene expression pattern was further stimulated in DDR2−/− HSCs by tumour-derived factors. Therefore, the results of this study suggest that DDR2 deficiency altered functional properties of tumour-activated HSCs, which in turn generated a prometastatic microenvironment fostering tumour growth and angiogenesis during the hepatic MCA38 colorectal metastasis process.
Previous evidence of DDR2 augmentation during HSC activation in fibrosis,7 and on the association of DDR2-dependent molecules, such as MMPs,29 with increased metastasis are at first glance in contradiction with conclusions from this study. However, several prometastatic factors originating in tumour-activated HSCs from DDR2-deficient mice were identified in this study that predisposed hepatic tissue to metastasis.
First, HSCs from DDR2−/− mice showed a higher myofibroblast transdifferentiation and recruitment rate in response tumour-derived factors than those from wild-type mice. Second, prometastatic effects of DDR2−/− HSCs did not depend solely on their increased recruitment and activation at metastastic sites. Under basal culture conditions, the supernatants from DDR2−/− HSCs already contained soluble factors increasing endothelial and cancer cell adhesion, migration and proliferation. Such stimulating factors could not be detected under basal conditions in the supernatants from cultured DDR2+/+ HSCs, yet their activity significantly increased in response to tumour-derived factors. These results suggest that DDR2 and their dependent factors may provide tumour-activated HSCs with factors that inhibit metastasis, while DDR2 deficiency may allow other HSC factors supporting metastasis to act exclusively. Such HSC-derived factors may have implications in the earliest stages of the metastasis process. For example, HSC release of endothelial cell adhesion-stimulating factors may create proadhesive hepatic sinusoids promoting the arrest of circulating cancer cells in the hepatic microcirculation from DDR2-deficient mice. Moreover, HSC-stimulating factors released by primary gastrointestinal tumours located in portal vein-draining areas may further induce the ability of DDR2-deficient HSCs to activate the adherence status of LSECs for circulating cancer cells. This remote organ preconditioning could foster premetastatic niche formation30 in the livers of patients with colorectal cancer. DDR2−/− HSCs also secreted cancer cell migration-stimulating factors and this was also further stimulated by tumour-derived factors, supporting the replacement-type metastasis growth pattern19 of MCA38 cancer cells. Moreover, tumour-activated HSCs also released hepatic endothelial cell migration-stimulating factors promoting angiogenesis in the hepatic metastasis microenvironment.20 Herein, the endothelial cell migration-stimulating activity of basal condition-cultured DDR2−/− HSCs was once again as potent as the activity of tumour-activated DDR2+/+ HSCs, and this capability of DDR2−/− HSCs was further induced by tumour-derived factors as well. These results are in agreement with the higher CD31-expressing cell recruitment into hepatic metastases from DDR2−/− mice.
Third, compared with HSCs from wild-type mice, those from DDR2-deficient mice had overexpression of key prometastatic genes (IL-10, TGFβ and VEGF) with known effects on immune suppression,31 angiogenesis32 and cancer cell survival and growth,29 and expression level of these genes further increased when HSCs were stimulated with tumour-conditioned media. IL-10 is a Th2-type cytokine that attenuates immune stimulation of hepatic fibrogenesis33 and exerts its antifibrotic activities by inhibiting HSC expression of TGFβ.34 Carcinoembryonic antigen-expressing colorectal tumours can prevent cancer cell death by inducing hepatic IL-10, thereby inhibiting hepatic NO-dependent cancer cell death.35 Production of IL-10 by tumour-activated DDR2−/− HSCs may be an additional source of this prometastatic molecule. TGFβ is the most important growth factor involved in liver fibrosis,31 but at the same time, it is a strong immune suppressor,24 a proangiogenic factor32 and promotes metastasis of colorectal cancer cells, thereby acting as an oncogene.36 It has been reported that carbon tetrachloride (CCl4)-induced hepatic cirrhosis increased hepatic metastasis.37 Therefore, overexpression of TGFβ may account for growth-stimulating properties of tumour-activated DDR2−/− HSCs on colorectal cancer cells.
DDR2−/− HSCs also had altered expression of anti- and prometastatic genes. The IGF-I gene expression was decreased by threefold, while prometastatic genes such as VEGF, increased twofold in basal conditions and threefold after tumour-conditioned medium stimulation. Expression of IGF-I by activated HSCs reduces fibrogenesis and enhances regeneration after liver injury.25 However, liver production of IGF-I decreases with ageing, which leads to hepatic injury via oxidative stress,38 while exogenous IGF-I has hepatoprotective effects.39 Because prometastatic effects of tumour-induced inflammation22 are oxidative stress-dependent,40 decrease of IGF-I gene expression from DDR2−/− HSCs may also contribute to metastasis. VEGF expression has long been associated with metastasis in patients with colorectal cancer.41 Activation of HSCs by tumour-derived factors increases VEGF production,13 which in turn induces LSEC adherence for cancer cells,42 and angiogenesis.13 Therefore, VEGF gene expression by DDR2−/− HSCs may also contribute to the prometastatic microenvironment of the liver.
DDR2 is an unusual receptor tyrosine kinase, because its ligand is fibrillar collagen rather than a growth factor-like peptide. DDR2−/− fibroblasts exhibit impaired migration through Matrigel in response to a chemotactic stimulus, which is also correlated with diminished activity of MMPs29 and altered gene expression pattern for extracellular matrix (ECM)-related molecules43 We previously reported7 that DDR2 signalling blockade reduced MMP2 expression and activity in the HSC-T6 cell line cultured in the presence of fibrillar collagen type I. Gelatin zymography analysis indicated that primary cultures of DDR2−/− HSCs also fail to express and secrete MMP-2 in response to collagen type I (Olaso E, unpublished data). Therefore, our current results suggest that enhancement of metastasis occurs in a liver microenvironment where the main tumour-associated stromal cell (ie, HSC-derived myofibroblasts) cannot secrete MMP-2 via a DDR2-dependent mechanism. This does not exclude the contribution of other metalloproteinases to metastasis apart from MMP-2, which are preserved—for example, MMP-9.
Furthermore, connective ECM, especially collagen, appears to limit the penetration of malignant cells through multiple means, acting as a physical barrier, by anchoring via ECM receptors, or directly via a growth arrest in the G0/G1 phase of the cell cycle. Hence, DDR2-deficiency may be an alternative means, apart from MMPs, to escape from growth-limiting signals, of hepatic ECM in tumour-activated HSCs. This would be consistent with the enhanced recruitment of HSC-derived myofibroblasts into the DDR2−/− metastatic foci. Moreover, increased αSMA staining and collagen type I expression occurs in CCl4-treated DDR2−/− mice compared with DDR2+/+ mice (Olaso E, unpublished data), indicating that DDR2 deficiency also leads to a stronger fibrogenic response to a hepatotoxin. Finally, liver regeneration after hepatectomy, which is relevant to surgery for resectable liver metastases, is known to stimulate metastasis,44 and our preliminary data suggest that DDR2 expression is decreased in whole-liver samples after partial hepatectomy in an experimental regeneration model (Badiola I, unpublished data). Therefore, reduced DDR2 expression may foster the transdifferentiation of HSCs into myofibroblasts in response to altered microenvironments created by tumour-derived factors,4 hepatotoxicity and hepatectomy. Moreover, HSC-derived myofibroblasts are known for early infiltration that precedes hypoxia-induced angiogenesis in hepatic metastasis.13 Although we have not directly tested DDR2 expression in HSCs in response to hypoxia, Chen et al 45 found that hypoxia induces vascular smooth muscle cell migration via DDR2 expression, suggesting a role for hypoxia as regulator of DDR2 in HSCs.
Finally, DDR2 expression was not consistently expressed within the MCA38 colon carcinoma cells of the experimental hepatic metastasis, suggesting that these cells can, at same time, degrade the ECM and escape from DDR2-dependent migratory and growth-restrictive signals induced by the ECM.
In summary, this study describes for first time the ability of HSCs to regulate the hepatic metastasis microenvironment via DDR2-dependent factors. The mechanism of hepatic colon carcinoma metastasis diathesis in DDR2-deficient mice may depend on a unique prometastatic microenvironment operating in the absence of certain DDR2-dependent factors, preventing the expression of key factors fostering tumour-induced hepatic fibrogenesis, tumour cell adhesion and proliferation and endothelial cell migration. DDR2 expression in the microenvironment of hepatic metastases from patients with colorectal cancer was low. DDR2 may therefore act as a hepatic metastasis suppressor factor operating in tumour-activated HSCs, and patients with colorectal cancer with hepatic DDR2 expression deficiency may be at higher risk of metastasis. However, these conclusions are based on an experimental metastasis model and in vitro studies that have limitations for a direct clinical translation, such as time scale of the hepatic metastasis process, use of intrasplenic route for experimental delivery of cancer cells to the hepatic microcirculation, and the use of established cancer cell lines and not spontaneous tumours.
IB and EO contributed equally to this work.
Funding This work was supported in part by grants from the Spanish Ministry of Health (FIS04/2785), International Association for Cancer Research (04-274) and Basque Government Department of Industry (SAIOTEK) to EO; grants from NIH (DK37340 and DK56621) to SLF; Spanish Ministry of Innovation and Science (SAF2009-12376) and Basque Government Department of Education (IT-487-07) to FVV and EO, and The Spanish coslos III Health Institute (ADE09/90041 and The Burdinola profesorship on molecular medicine to FVV.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.