Role of Fibulin-5 in Metastatic Organ Colonization

This article is featured in Highlights of This Issue, p. 529

The interaction of tumor cells with the surrounding stroma has attracted much attention in recent years. Stimulation of the stroma by tumor cells results in the production of numerous factors that influence carcinoma cells and stimulate tumor progression, invasion, and eventually metastatic spread (1–4). The tumor-associated stroma is composed of various cell types, one of the most abundant and important of which is activated fibroblast. The role of carcinoma-associated fibroblasts in tumor progression and metastasis has been shown in many studies but the precise mechanism and the factors involved have yet to be identified (1).

The metastatic spread of tumors involves multiple steps, during which tumor cells escape from the primary tumor, migrate and invade surrounding tissues, enter the vasculature, circulate and reach secondary sites, extravasate, and finally establish metastatic foci (5). Spontaneous metastasis assays do not allow these steps to be distinguished from one another. However the colonization of organs by circulating tumor cells can be explored specifically by administering tumor cells directly into the blood circulation after intravenous injection in an experimental metastasis assay.

Previously we have shown that the fibroblast-produced S100A4 protein fosters metastatic spread of tumor cells at the level of the primary tumor by affecting the development of a functionalized tumor-associated stroma. Stroma development in primary tumors was suppressed in S100A4^−/− tumor-bearing mice (6, 7). This suppression could be overcome by coinjection of tumor cells with S100A4-expressing activated fibroblasts [mouse embryonic fibroblasts (MEF)] in spontaneous metastasis assays.

Here we show that by responding to soluble factors produced by tumors, fibroblasts are also instrumental in stimulating metastasis formation at the level of organ colonization. This process is S100A4-independent. Gene expression profiling of S100A4-deficient fibroblasts identified genes whose expression is modulated by soluble factors derived from tumor cells. Fibulin-5 (FBLN5), a matricellular protein that modulates elastogenesis, extracellular matrix (ECM), cell motility, cell adhesion, and binds integrins (12), was downregulated in response to tumor-derived factors. FBLN5 was found to inhibit metastatic colonization of the liver and lung and to suppress the activity of matrix metalloproteinase 9 (MMP9). These data reveal a tumor-driven Fbln5-mediated autocrine regulation of the ability of fibroblasts to invade the ECM and generate a microenvironment favorable for metastasis formation.

VMR, CSML0, and CSML100 mouse mammary adenocarcinoma cell lines were derived from 3 independent spontaneous tumors in A/Sn mice (8). Wild-type MEF/S100A4^+/+ and S100A4 knockout MEF/S100A4^−/− MEFs were isolated as described (6). The cell lines were maintained in Dulbecco's minimum essential medium (DMEM) GlutaMax plus I supplemented with 10% FBS (Gibco BRL), penicillin (100 units/mL), and streptomycin (100 units/mL) in a humidified 5% CO₂ atmosphere.

MEF/S100A4^−/− were grown in conditioned media (CM) from VMR cells or DMEM containing 10% FBS for 48 hours. The cells were harvested and total RNA was isolated as described previously (9). The integrity of the RNA was analyzed by gel electrophoresis. The MicroPoly(A) Purist Kit (Ambion Inc.) was used according to manufacturer's instructions to isolate poly(A) mRNA. Customized in-house production of microarray chips as well as probe labeling and chip hybridization has been described previously (10). The arrays were scanned by using an Axon GenePix 4000B microarray scanner (Axon Instruments Inc.) and the GenePix Pro program. Genes were considered as differentially expressed if they showed at least a 2.5-fold difference in signal intensity.

Total cellular RNA was isolated as described (9). RNA was extracted by using the Nucleospin Triprep Kit (Macherey-Nagel). First-strand cDNA synthesis was carried out by using SuperScript II RT (Invitrogen) with random primers. Quantitative real-time PCR (qRT-PCR) was carried out by using primers (Supplementary Table S1) and Fast SYBR Green Master Mix kit (Roche Applied Science) in a LightCycler 2.0 instrument. All analyses were carried out at least in duplicate reactions.

MEF Matrigel invasion assays were carried out as described previously (6). MEFs in three-dimensional (3D) culture conditions were incubated for 24 hours either in 10% FBS-DMEM or CM from tumor cells supplemented with 10% FBS. Blocking of invasion was carried out with 5 μmol/L of the MMP2/9 inhibitor SB-3CT (Biomol International) for 24 hours. For the 3D coculture, 1 × 10⁴ tumor cells were suspended in Matrigel (BD Biosciences) before polymerization. Fibroblast invasion was documented by taking photos with a phase-contrast microscope every day over a 5-day period. The extent of invasion of fibroblasts in the presence of VMR CM was quantified as in (9) and measured after 48 hours. Experiments were repeated 3 times, each with 6 independent samples.

Monolayer wound healing assays were carried out as described previously (6). Scratched monolayers were incubated with either 2% FBS-DMEM or 2% FBS in serum-free CM from VMR cells. Closure of the wound was monitored for 24 hours by measuring the width of the residual wound by using Multi Gauge software (Fujifilm). The plotted data represent the mean of 3 experiments, each including triplicate measurements. The relative motility is normalized to the size of the wound postscratching at 0 hour.

Eight- to 10-week-old wild-type and S100A4^−/− A/Sn mice (6) were used for this study. All animals were maintained according to the guidelines of the Federation of European Laboratory Animal Science Association (FELASA). For Matrigel plug assays, 1 × 10⁶ of VMR and MEF cells were harvested, mixed with 200 μL of Matrigel and implanted into the left and right flanks of the mouse. After 24 hours, Matrigel plugs were excised and RNA was isolated as described before (9). Experiments were carried out by using 5 mice per group. For experimental metastasis assays, 2.5 × 10⁵ VMR or CSML100 tumor cells or a mixture of tumor cells with MEFs (ratio 1:1) were injected into the tail vein of experimental animals. For survival experiments, mice were sacrificed when clinical signs of illness were documented. To compare the metastatic burden in different groups, the experiment was terminated when the first animals manifested signs of illness. Lungs and livers were isolated and subjected to standard histological analysis. The metastatic burden in lungs and livers in 4% PFA fixed tissue sections was determined as described earlier (7). Experimental groups comprised 5 mice per group. All experiments were repeated 2–3 times.

Formalin-fixed paraffin-embedded tissue sections were subjected to antigen retrieval by treatment with 0.01 M Tris-HCl (pH 9.0), followed by incubation with primary anti-MMP9 (Biorbyt; diluted 1:100) or anti-FBLN5 (BSYN 1923; ref. 11; diluted 1:600), rabbit polyclonal antibodies and the EnVision system (DAKO A/S, Glostrup). Positively stained cells were visualized by enhanced 3,3V-diamino-benzidine chromogen (brown; Pierce). Sections were counterstained with Mayers hematoxylin followed by microscopic analysis.

Proteins were detected by using a standard Western blot procedure following SDS-PAGE Primary antibodies against MMP3 (Nordic Biosite), MMP9 (Biorbyt), and FBLN5 (BSYN 1923) were diluted according to the suppliers’ instructions. Horseradish peroxidase-conjugated rabbit anti-mouse or goat anti-rabbit antibodies (DAKO) and ECL-plus chemiluminescent substrate (Amersham) were used for visualization.

For zymography, 3D cultures were grown in 10% FBS-DMEM or 10% FBS CM-VMR and replaced after 48 hours with serum-free DMEM for 24 hours, collected and concentrated 10–20 times in concentrator columns (Millipore; 10-kDa cutoff) and assayed for protease activity in 1 mg/mL gelatin or β-casein gels as described before (9). A LAS-1000 analyzer and MultiGauge software from Fujifilm were used to process the images.

All the statistical analyses were carried out by using GraphPad Prism 5.0 statistical software (GraphPad Software Inc.). The Wilcoxon log-rank test was used for the comparison of survival curves. Metastatic burden comparisons were carried out by using the Mann-Whitney nonparametric test. P values for the rest of the analyses were carried out by using the 2-tailed unpaired t-test. P values less than 0.05 were regarded as significant and were marked with an asterisk (*), P < 0.01 (**) and < 0.001 (***).

Our previous work has shown that activated fibroblasts in the stroma of primary tumors promote metastasis in an S100A4-dependent manner (6). However, these studies did not address whether activated fibroblasts only exert their metastasis-promoting effect locally, at the site of primary tumor or whether fibroblasts in distant organs can also influence metastasis during the process of organ colonization. To address this issue, we carried out tail vein i.v. administration of VMR mammary adenocarcinoma cells that are metastatic to the lungs and liver. Following i.v. injection of VMR cells into wild type and S100A4^−/− mice, the survival of the S100A4^−/− mice was substantially improved compared with the wild type control (Fig. 1A). However, when a mixture of VMR cells and MEFs were injected intravenously into S100A4^−/− mice, a substantial decrease in the survival of the mice was observed, regardless of whether the MEFs were derived from wild type or S100A4^−/− mice (Fig. 1A). Similar data were obtained by using a second model, namely CSML100 mammary adenocarcinoma cells that are metastatic to the lungs (Fig. 1B). Again, survival of S100A4^−/− mice was substantially reduced when CSML100 cells were mixed with either wild type or S100A4^−/− MEFs. Thus in 2 different models, MEFs were able to promote metastatic organ colonization independently of the S100A4 expression. These data indicate that fibroblast-derived factors other than S100A4 stimulate metastatic organ colonization.

To identify fibroblast-derived factors other than S100A4 that might affect the metastatic potential of tumor cells, we used expression profiling to identify genes that are specifically regulated in S100A4^−/− MEFs in response to CM from VMR mammary adenocarcinoma cells (array A). We also compared the expression profiles of cocultures of S100A4^−/− MEFs and VMR cells versus S100A4^−/− MEF and VMR monocultures (array B). As our goal was to identify genes with a stable and sustained response to tumor cell CM, S100A4^−/− MEFs were cultivated in VMR CM or cocultured for 48 hours. To focus subsequent analysis on potentially drugable targets, we selected genes encoding transmembrane or secreted proteins. Thirteen genes were selected based on these criteria and are presented in Table 1. Differential expression of some of these genes was also documented in array B.

To confirm the results of the microarray analysis, we carried out qRT-PCR analysis of 7 of the genes identified as being differentially expressed in fibroblasts treated with VMR CM, and could confirm differential expression of 6 of them (Fig. 2A). Additional verification was obtained in vivo by implanting Matrigel plugs containing VMR cells and/or MEFs into mice, then by using qRT-PCR to analyze RNA extracted from the plugs. These data confirmed that expression of Fbln5, Cdh16, and Ccn2 [connective tissue growth factor (Ctgf)] in MEFs was downregulated in the presence of VMR tumor cells (Fig. 2B). However, we could not confirm the upregulation of Ly6c and Mmp3. Consistent with this idea, Table 1 shows that in the array B the upregulation of Ly6c and Mmp3 were not observed. Moreover, these genes were not upregulated in 3D coculture of VMR tumor cells and S100A4^−/− fibroblasts (data not shown). To test the possibility that the expression of these genes is regulated by soluble factors whose production from VMR cells is suppressed on coculture, we cultured fibroblasts in 3D Matrigel in the presence of VMR CM. We also pretreated fibroblasts with VMR CM before injection into mice in the Matrigel plugs. Indeed, VMR CM induced expression of Ly6c and Mmp3 both in vitro in 3D Matrigel culture and in vivo in Matrigel plugs (Fig. 2C).

Taken together our analysis revealed 5 genes whose expression is regulated in fibroblasts by factors derived from tumor cells, and which have the potential to modulate metastasis development. We suggest that Ly6C and MMP3 could act as stimulators of organ colonization, whereas FBLN5, CCN2, and Cadherin-16 could suppress the ability of tumor cells to colonize organs.

To provide a proof of principle we focused next on fibroblast-derived FBLN5, a matricellular protein (12) to determine whether it has the ability to suppress metastatic organ colonization. Initial experiments showed that Fbln5 is expressed in fibroblasts but is hardly detectable in carcinoma cell lines (Supplementary Fig S1A). Furthermore, CM from different mammary carcinoma cells inhibited transcription of fibroblast Fbln5 independently of the metastatic potential of the tumor cells (Supplementary Fig S1B).

Next, we tested the ability of Fbln5 to suppress metastases in vivo. VMR and CSML100 mammary adenocarcinoma cells were engineered to ectopically express Fbln5 (VMR/Fbln5 and CSML100/Fbln5, respectively; see Supplementary Fig. S2 and Supplementary Materials for details of generation of these cell lines). These cells were introduced i.v. into mice in experimental metastasis assays. Assessment of metastases in lungs and livers of mice revealed that overexpressing Fbln5 tumor cell lines had a substantial reduced percentage of metastatic burden (Fig. 3A and B). Immunostaining of tumor-containing tissue sections with FBLN5 antibodies revealed that the tumor cells continue to ectopically express Fbln5, indicating that inhibition of metastatic colony formation is associated with FBLN5 production (Fig. 3C). Ectopic expression of Fbln5 in CSML100 and VMR adenocarcinoma cells did not affect their proliferation rate (data not shown), indicating that suppressed proliferation in response to FBLN5 was not the reason for the reduced metastatic burden in the mice.

We further tested whether constitutive expression of Fbln5 in fibroblasts could also promote the ability of tumor cells to colonize distant organs. To this end, MEF/Fbln5 cells constitutively expressing Fbln5 (see Supplementary Fig. S3 and Supplementary Materials for the details) were intravenously coinjected together with VMR cells into mice in experimental metastatic assays. Assessment of the metastatic burden in the lungs and livers showed that MEF/Fbln5 also substantially decreased the formation of pulmonary and liver metastases by VMR cells (Fig. 3D). We therefore conclude that FBLN5 is a stromal cell-derived factor that reduces the ability of tumor cells to form metastases in distant organs.

We reasoned that the ability of fibroblasts, to interact productively with disseminating tumor cells that extravasate into distant organs could be pivotal in the formation of a tumor microenvironment that is supportive of metastasis formation. By using S100A4^−/− fibroblasts that constitutively express Fbln5 (MEF/Fbln5) or that has a stable knockdown of Fbln5 (MEF/shFbln5; see Supplementary Fig. S4 and Supplementary Materials), we therefore analyzed a number of parameters that could be important for such an interaction.

The proliferation rate of MEF/Fbln5 and MEF/shFbln5 was similar (data not shown).

The motility of S100A4^−/− MEFs with overexpressed or silenced Fbln5 in the monolayer wound-healing assay showed that MEF/Fbln5 cells closed the wound substantially faster than the control vector-infected cells. Consistently, wound closure was delayed with MEF/shFbln5 cells. However, the presence of VMR CM did not influence the speed of wound closure, suggesting that Fbln5-associated changes in MEF motility are not associated with the effect of VMR CM (Fig. 4A).

Next, we tested whether expression of Fbln5 in fibroblasts affects their ability to invade into the ECM. When VMR adenocarcinoma cells were cocultivated with fibroblasts in 3D Matrigel, we found that in contrast to control cells the MEF/Fbln5 cells were not able to invade the Matrigel (Fig. 4B). This suggested that suppression of Fbln5 expression in fibroblasts might be needed to induce MEF invasion. We therefore compared the invasion of MEF/Fbln5 and MEF/shFbln5 cells in 3D Matrigel in the presence of VMR CM. As shown in Figure 4C, CM from VMR cells induces invasion of MEF/shFbln5 and MEF/vector control cells but not MEF/Fbln5 cells. Quantification of the extent of fibroblast invasion confirmed these observations (Fig. 4C and D, bottom).

Cell invasion involves protease-mediated remodeling of the ECM. It has been recently reported that FBLN5 suppresses lung cancer invasion by inhibiting MMP7 expression (13). To determine whether constitutive expression of Fbln5 in MEFs also modulates expression of proteases, we used β-casein and gelatin zymography to examine protease activity in CM, collected from the 3D Matrigel MEF cultures. We found that CM from VMR cells stimulated MEFs, independently of Fbln5 expression, to produce a protein of approximately 50 kDa with a strong caseinolytic activity (Fig. 5A). We propose that this activity is attributable to MMP3, which is known to exhibit caseinolytic but not gelatinolytic activity (14). Results obtained by using qRT-PCR analysis of Mmp3 expression in 3D MEF cultures, as well as Western blot analysis of the CM from these cultures by using MMP3 antibodies are consistent with this notion (Fig. 5B).

The gelatin zymography revealed that CM from VMR adenocarcinoma cells substantially boosted production of MMP9 in MEF/vector but not in MEF/Fbln5 cells, whereas MMP2 activity remained unchanged (Fig. 5C). VMR CM-mediated stimulation of Mmp9 but not Mmp2 transcription in MEF/vector cells was confirmed by qRT-PCR analysis of RNA isolated from 3D MEF cultures. Western blot analysis of the CM from these cultures with MMP9 and MMP2 specific antibodies confirmed the qRT-PCR data (Fig. 5D). Importantly, treatment of MEF/vector cells with an MMP2/9 selective inhibitor (SB-3CT) in 3D Matrigel culture blocked the ability of VMR CM to induce invasion of these cells (Fig. 4D). We therefore conclude that Fbln5-mediated suppression of Mmp9 expression is instrumental in inhibiting the ability of MEF/Fbln5 cells to invade the 3D matrix.

To provide in vivo corroboration of the above data, immunohistochemical analysis was carried out on lung tissue sections from VMR/vector and VMR/Fbln5 tumor-bearing animals. Immunostaining with MMP9 antibodies showed reduced expression of Mmp9 in VMR/Fbln5 pulmonary lesions compared with the controls (Fig. 6A). Consistently, we found that although transcription of Mmp3 was upregulated in VMR/Fbln5 cells compared with controls, Mmp9 transcription was reduced (Fig. 6B). Together, these data substantiate the notion that fibroblast-derived FBLN5 blocks the ability of tumor cells to colonize secondary organs by suppressing transcription of certain Mmps, in particular Mmp9.

Both tumor cells and host-derived stromal cells contribute to the formation of a microenvironment that is conducive for tumor growth and metastasis formation. We and others have showed the role of activated fibroblasts at the site of the primary tumor in stimulating the metastatic spread of tumor cells. Molecules such as inflammatory cytokine CCL-2, hepatocyte growth factor, stroma-cell derived factor, SDF-1, and S100A4 have been shown to promote tumor growth, angiogenesis, invasion into the ECM, and eventually metastases (1,3,6,15, 16). The interaction of disseminating tumor cells with the stroma in distant organs is also important during metastasis formation (17, 18) but is much less well understood. Here, we show that fibroblast-derived factors, different from those produced by fibroblasts at the site of primary tumor, could play a key role during metastatic organ colonization. In this study, we have identified 5 genes (Mmp3, Ly6c, Cdh16, Ccn2, and Fbln5) whose expression in fibroblasts is modulated by tumor-derived soluble factors and which could potentially play a role in metastatic organ colonization. Expression of 2 genes (Mmp3 and Ly6c) was upregulated. MMP activation, in particular Stromelysin-1 (MMP3), has been shown to have a strong protumorigenic effect (19). Ly6c, a leukocyte differentiation antigen, is specifically expressed in tumor-associated macrophages infiltrating mammary tumors (Ly6C^hi TAM). Ly6C^hi bone marrow-derived myeloid precursors are attracted to the growing tumor and differentiate into TAMs. This allows us to speculate that increased Ly6C in the tumor microenvironment may play a protumorigenic role (20).

Three genes were downregulated in fibroblasts in response to treatment with tumor-derived factors (Cdh16, Ccn2, and Fbln5). Expression of Cdh16, a member of the cadherin superfamily, is thought to have a tumor suppressor function in renal carcinomas (21). Cadherin-16 may also suppress metastasis formation, as E-cadherin, a related member of the cadherin cell–cell adhesion family that is a well-recognized metastasis-suppressor gene (22, 23). CCN2 and FBLN5 both belong to the group of modulators of cell-ECM interactions known as matricellular proteins. CCN2, a member of the CCN protein family, plays a critical role in cardiovascular and skeletal development, injury repair, fibrotic diseases, and cancer (24). As both tumor-promoting and tumor-suppressing activities have been attributed to CCN2, it is thought that the role of CCN2 in cancer is probably context dependent (25).

FBLN5, a 66 kDa glycoprotein, is produced by various cell types and is essential for the formation of elastic fibers. Apart from its elastogenic function, FBLN5 plays multiple roles in diverse cellular processes (12). A tumor-suppressor function associated with the antiangiogenic activity of FBLN5 has been proposed (26–30). Additionally, FBLN5 has been implicated in the suppression of metastasis, as FBLN5 was found to reduce lung cancer invasion and metastasis by inhibiting MMP7 expression (13). However, a stimulatory effect of FBLN5 on tumor progression has also been reported. TGF-β stimulation in normal mammary epithelial cells led to upregulation of FBLN5, production of MMPs, and enhanced growth of FBLN5-positive tumor cells (27). Studies by using Fbln5-deficient mice showed that loss of FBLN5 binding to β-integrins inhibits tumor growth by increasing the level of reactive oxygen species (31). Given that FBLN5 regulates a variety of processes such as structuring and ECM remodeling, angiogenesis, and epithelial-mesenchymal transition, all of which are controlled by a complex network of different proteins, the balance of factors that control the activity of FBLN5 in the tumor microenvironment will determine the exact effect that FBLN5 will exert (12). Accordingly, the effect of FBLN5 has been shown to be context-dependent (28).

Tumor cells coinjected with fibroblasts that constitutively express Fbln5 were unable to form tumor colonies in peripheral organs. At the same time, these fibroblasts failed to invade 3D Matrigel. As the invasion of host-derived cells into developing secondary tumors to create a supportive stroma is a rate-limiting process in metastasis (2), we suggest that fibroblast-derived FBLN5 might block the ability of tumor cells to colonize secondary organs by modulating the structure of ECM and suppressing the invasive activity of fibroblasts. Invasion is typified by MMP-dependent ECM remodeling and is orchestrated by a complex network of proteases and their inhibitors that are mainly produced by stromal cells such as fibroblasts. Accordingly, we found that Fbln5 expression in fibroblasts suppressed Mmp9 expression, inhibited fibroblast invasion, and correlated with reduced metastasis formation.

MMP9 has been attributed a number of roles in tumor growth and metastasis, including triggering the tumor-induced angiogenic switch (32, 33). Our data suggest that factors produced by disseminating tumor cells trapped in the peripheral organs, or even released from the primary tumor itself, serve to downregulate Fbln5 in fibroblasts within the peripheral organs. In turn, this results in upregulation of Mmps such as Mmp7 or Mmp9, ECM remodeling, stimulation of angiogenesis, and invasion of fibroblasts. All these events will create a fertile milieu within which tumor cells can develop metastases.

This scenario is reminiscent of recently described tumor-induced premetastatic changes in organs where metastases will ultimately develop (34). In this context, Mmp9 is specifically expressed in endothelial cells and macrophages of the premetastatic lung and was found to significantly promote metastasis (35). Furthermore, lysyl oxidase is an elastin and collagen cross-linking enzyme whose essential role in the formation of premetastatic niches has recently been discovered (36). Formation of elastic fibers is a multistep process that involves complex interaction of tropoelastin, fibrillins, fibulins, and lysyl oxidases (37). Given that FBLN5 also exerts an elastogenic organizing activity (12, 38), it is intriguing to speculate that FBLN5 might also act to suppress metastasis formation at this level.

In summary, our data indicate that stromal fibroblasts in organs where metastases develop are addressed by tumor-derived factors, and thereafter play an active role in supporting metastatic organ colonization. Understanding the complexity of the regulatory networks involved holds the promise of identifying ways to block the ability of tumor cells to settle in peripheral organs and form metastases.

No potential conflicts of interest were disclosed.

We thank Birgitte Kaas, Hanne Nors, and Lene Bregnholt Larsen for careful technical assistance.

The authors acknowledge funding from the FP7 European Union collaborative project TuMIC, contract no. HEALTH-F2-2008-201662 (E. Lukanidin and J. Sleeman) and the Danish Cancer Society (E. Lukanidin).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.