Blocking Fibroblast Growth Factor Receptor Signaling Inhibits Tumor Growth, Lymphangiogenesis, and Metastasis

Fibroblast Growth Factor receptor (FGFR) activity plays crucial roles in tumor growth and patient survival. However, FGF (Fibroblast Growth Factor) signaling as a target for cancer therapy has been under-investigated compared to other receptor tyrosine kinases. Here, we studied the effect of FGFR signaling inhibition on tumor growth, metastasis and lymphangiogenesis by expressing a dominant negative FGFR (FGFR-2DN) in an orthotopic mouse mammary 66c14 carcinoma model. We show that FGFR-2DN-expressing 66c14 cells proliferate in vitro slower than controls. 66c14 tumor outgrowth and lung metastatic foci are reduced in mice implanted with FGFR-2DN-expressing cells, which also exhibited better overall survival. We found 66c14 cells in the lumen of tumor lymphatic vessels and in lymph nodes. FGFR-2DN-expressing tumors exhibited a decrease in VEGFR-3 (Vascular Endothelial Growth Factor Receptor-3) or podoplanin-positive lymphatic vessels, an increase in isolated intratumoral lymphatic endothelial cells and a reduction in VEGF-C (Vascular Endothelial Growth Factor-C) mRNA expression. FGFs may act in an autocrine manner as the inhibition of FGFR signaling in tumor cells suppresses VEGF-C expression in a COX-2 (cyclooxygenase-2) or HIF1-α (hypoxia-inducible factor-1 α) independent manner. FGFs may also act in a paracrine manner on tumor lymphatics by inducing expression of pro-lymphangiogenic molecules such as VEGFR-3, integrin α9, prox1 and netrin-1. Finally, in vitro lymphangiogenesis is impeded in the presence of FGFR-2DN 66c14 cells. These data confirm that both FGF and VEGF signaling are necessary for the maintenance of vascular morphogenesis and provide evidence that targeting FGFR signaling may be an interesting approach to inhibit tumor lymphangiogenesis and metastatic spread.

The lymphatic system is a blind-ended network of endothelial cell-lined vessels that maintains fluid homeostasis by unidirectionally transporting tissue fluid, extravasated plasma proteins, lipids and cells from the interstitial space to the circulatory system via the thoracic duct. Several studies have demonstrated the importance of the lymphatic system as a route for tumor dissemination [6] and that metastasis is enhanced by VEGF-C via an increase in tumor lymphangiogenesis [7,8,9]. FGF-2 has also been shown to indirectly induce lymphangiogenesis, in vivo, in a mouse cornea assay by upregulating VEGF-C [10]. FGF2 also has been shown in vitro to act directly on lymphatic endothelial cell migration, proliferation and tubulogenesis [11,12].
However, no study has ever addressed the role of FGFR signaling in tumor lymphangiogenesis and metastasis via the lymphatic system. Here, we provide evidence that blockade of FGFR signaling in tumor cells using dominant negative FGFR (FGFR-2DN) approach [4,13,14], impairs mammary carcinoma growth and metastasis, leading to an improvement in overall survival. Blockade of FGFR signaling causes a decrease in tumor lymphangiogenesis, an increase in isolated lymphatic endothelial cell number and a reduction of VEGF-C expression in tumor cells. Decreased production of VEGF-C is independent of downregulation of other known regulators; COX-2, HIF-1a and PDGF-B [15,16,17]. Furthermore, we demonstrate that FGF signaling might also act directly on the tumor lymphatic endothelium by inducing the expression of lymphangiogenesis-related genes. Our results demonstrate that FGFR signaling, in addition to mediating tumor growth, regulates tumor metastasis and lymphangiogenesis via a VEGF-Cdependent mechanism.
Cobalt chloride (Sigma Aldrich, c8661) was a gift from Dr Sandra Sena, and was dissolved directly in treatment media and sterile-filtered before use.

Animals and in vivo experiments
In vivo tumor growth and metastasis experiments were performed as previously described [4,18,19].
Orthotopic transplantation of mouse mammary carcinoma cells. 200,000 66c14 cells, containing a pool of three clones per condition (clones C1-3 or C4, C18 and C22 for control and FGFR-2DN group respectively) or parental cells were injected directly into the exposed inguinal mammary fat pad of anesthetized 6-8 week old female Balb/C mice (The Jackson Laboratory). Tumor volume was measured once a week using a caliper and calculated according to the formula V = ([major axis] x [minor axis] 2 6(p/6)). Five weeks post injection (tumor size less than 2000 mm 3 ), mice were euthanized in a CO 2 chamber, lungs inflated with 4% formalin and the number of metastatic nodules then quantified.
Xenografting of C6 glioblastoma cancer cells. Rat C6 glioma cells were injected subcutaneously into the midline of the back of 8-10 week old-immunodeficient RAG 2/cc mice (Gift of Dr J. P. Di Santo, Institut Pasteur, Paris) as previously described [4].
To study the role of FGFs in in vivo lymphangiogenesis, 66c14 tumors were harvested at the time no statistical difference of tumor size was detected, to rule out any tumor growth variation-related vascular changes, and cut in two equal pieces. The first tumor half underwent immunohistology processing (OCT-Tissue-Tek embedded) while the second half was snap frozen for RNA extraction. Half-cut C6 tumors were paraffin embedded, while the second part was snap frozen for RNA preparation. Survival time is indicated by time to the ethical endpoint, at which time the animals were humanely euthanized.
All studies were repeated twice to ensure reproducibility (with a minimum of 6 mice per group).
Animal experiments were conducted in accordance with the University of Bordeaux and University of Utah Institutional Animal Care and Use Committees.

In vitro Assays
In vitro proliferation assays were performed as previously described with minor modifications [19].
Parental 66c14 cells were sequentially reverse and forward transfected with specific mouse FGFR-1, FGFR-2 or control siRNA (Santa Cruz Biotechnology, sc-29317 and sc-29799). Transfected cells were then cultured in 0.5% serum-containing media for 72 hours; the cell number determined with a hematocytomer in quadruplicate and normalized to control siRNA condition.
In vitro lymphangiogenesis coculture assay was done using a modified version of Sakamoto's method [5,20]. Briefly, 500,000 tumor cells were mixed with 500 ml of collagen 1 (2 mg/ml, for C6 cells) or growth factor-reduced matrigel (BD Biosciences, for 66c14 cells) and plated in 24 or 48-well plates (for C6 and 66c14 respectively). HMVEC-dLys (150,000 and 15,000 cells for C6 and 66c14 experiments respectively) were seeded on the top of the matrix gel. Recombinant human VEGF-A and VEGF-C (20 and 100 ng/ml respectively) or VEGFR-2/Fc and VEGFR-3/Fc chimera (1 mg/ml) were added to the respective culture media. Lymphatic tube formation was monitored after 24 hours and lymphatic-like structures were visualized by prox-1 (Abcam, for C6 co-culture) or LEL (Lycopersicon Esculentum (Tomato) Lectin, Vector; for 66c14 co-culture) staining. Counterstaining was done with DAPI (Invitrogen). Stainings were visualized using a 100X or a 200X magnification (for respectively prox-1 and LEL staining) on a Leica confocal microscope.
For supernatant-induced lymphatic-like structure formation, 66c14 cells were plated at 10 6 cells per dish (6 cm diameter) and cultured overnight. The media was then replaced with fresh basal media without serum and collected 24 hours later. Tube formation was then performed as described just above. Phase contrast images were taken with an Olympus FSX-100 microscope at 200X magnification.
For VEGF-C western blotting, 66c14 tumor cells were incubated in serum-free basal media, for 24 hours. The following day, cell supernatants were concentrated using a 10 kDa cut-off amicon concentrator column (Millipore), and the total protein concentration determined by BCA (Pierce). Equal amounts of supernatant proteins underwent SDS-PAGE, transferred on PVDF membrane and incubated with a Goat anti VEGF-C antibody (Santa Cruz Biotechnology, clone C20, sc-1881). Coomassie blue staining of the corresponding membrane was utilized as loading control.
For phospho-protein western blotting, 66c14 tumor cells were serum-starved overnight. The following day, cell treatment with the FGFR inhibitor PD-173074 (30 mM, for 10 or 60 minutes) was performed, in the presence or absence of FGF-2 (20 ng/ml, 10 minutes). Cells were lysed in RIPA buffer and the total protein concentration determined by Bradford (Biorad). Equal protein amounts underwent SDS-PAGE, transferred on nitrocellulose membrane and incubated with the PathScan Multiplex Western Cocktail I to detect phospho-p90RSK, phospho-Akt, phospho-p44/42 MAPK and phospho-S6 Ribosomal Protein (Cell Signaling Technology, #5301). Rab11 expression level, detected using the same antibody cocktail, was utilized as loading control, and kinase activities were normalized to their respective controls. Representative western blot and densitometry is shown.

Standard and Quantitative RT-PCR
Total RNA was extracted from cells using the RNAEasy mini kit (Qiagen) under conditions recommended by manufacturer. RNA concentration was determined using a nanodrop spectrophotometer (Nanodrop-ThermoScientific) and 1 mg of total RNA was reverse transcribed using the SuperScript III First-Strand Synthesis kit (Invitrogen). Standard PCRs were performed using mouse FGFs and FGFRs-specific primers (Table S1), GoTaq DNA polymerase (Promega) on an eppendorff thermocycler for 35cycles. Mouse brain cDNA and water were used as positive and negative control, respectively.
Quantitative PCRs were performed using gene-specific primers (SABiosciences Corporation), SYBR Green mix (ABgene) on an ABI Prism 7900 HT Real-Time PCR System (Applied Biosystems) using a 384-well plate. Quantification was performed by the standard curve method using GAPDH gene as normalizer.

Histology and Immunostaining
For C6 tumors, 10 mm sections were obtained using a microtome and stored at room temperature until use. Deparaffinized sections were treated with 10 mM Sodium Citrate in a microwave oven prior to immunostaining. For 66c14 tumors, 6 mm frozen tumor sections were air-dried for 30 minutes at room temperature and washed 3 times in PBS before immunostaining.
The following primary antibodies were then incubated overnight at +4uC: Goat anti mouse VEGFR-3 (R&D Systems, AF743), Syrian hamster monoclonal anti mouse podoplanin (Developmental Studies Hybridoma Bank, University of Iowa, 8.1.1 clone) and Rabbit anti cytokeratin (Dako, Z0622 ). Staining was then visualized by incubation with adapted secondary antibodies (Molecular Probes for fluorescent antibodies, DAKO for HRP-conjugated antibodies) and, if necessary, with DAB (DAKO). Pictures from at least five tumor areas were taken in the individual corresponding fluorescent channel using an Olympus IX71 inverted microscope at 400 X magnification. Peroxydase staining pictures were taken as previously described (Auguste et al, 2001). For each picture, vessel density was manually determined using the Image J software. For each experiment, specificity of the labeling was controlled by omitting the primary antibody ( Figure S10).

Statistical Analysis
Data are shown as mean 6 Standard Error of the Mean (SEM) of 6 to 9 samples from 2 to 3 independent experiments. Statistical analyses were carried out using Statview (SAS Institute, Inc.) or GraphPad Prism (for the Kaplan-Meier survival curves). A P value less than 0.05 (*, #) was defined as statistically significant.

Inhibition of Fibroblast Growth Factor Activity Blocks Proliferation, Metastasis and Extends Survival in Tumorbearing Mice
To investigate the role of FGFR signaling in tumorigenesis, metastasis and tumor lymphangiogenesis, we first inhibited FGFR signaling in mouse mammary carcinoma cells (66c14) using the dominant-negative FGF receptor strategy [4,13,14]. 66c14 cells are known to metastasize to the lungs mainly via the intratumor lymphatic vessels, in a VEGF-C-dependent manner, when injected directly into the mammary fat pad [18,21]. Expression of multiple FGF ligands and receptors (including FGF-2, FGFR-1 and both FGFR-2 IIIb and IIIc isoforms) was also detected by RT-PCR in 66c14 tumor cells ( Figure 1A and data not show for FGFR-2 isoforms). FGF-2, the FGF ligand prototype, was shown to bind with a high affinity to the IIIc splice variant of both FGFR-1 and R-2 [1], and the expression of a dominant negative form of this receptors splicing variant, truncated for its intracellular tyrosine kinase domains, was capable of disrupting FGF-2-mediated biological functions [4,13,14].
PD-173074-induced inhibition of 66c14 cell proliferation correlated with the respective up-and down-regulation of the cyclin inhibitor, p21, and the cyclin D1 mRNA expression ( Figure  S2A). C-Myc expression was increased in 66c14 cells treated with PD-173074. A decrease in both basal and FGF-2-stimulated Erk and S6 ribosomal protein phosphorylation was observed in PD-173074-treated cells ( Figure S3A and B). A similar reduction in basal Erk activity was detected in FGFR-2DN-expressing 66c14 cells as compared to controls ( Figure S4A and B). Interestingly, these mitogenic alterations might not be coupled to any modification in migration/invasiveness as no significant variation in epithelial-to-mesenchymal transition (EMT)-related or in metalloproteinase genes was detected, except for MMP-14, in FGFR inhibitor-treated 66c14 cells as compared to controls ( Figure S2B).
Taken together, this data indicate that FGFR-2DN expression, FGFRs siRNA or treatment with the kinase inhibitor PD-173074 inhibits cancer cell proliferation, to a similar degree.
We next implanted dominant-negative or control 66c14 cells into the mammary fat pad of Balb/c mice to monitor in vivo growth of orthotopic mouse mammary tumors. As shown in Figure 2A, tumor development was slower in mice bearing FGFR-2DNexpressing tumors (left panel), leading to smaller tumors compared to mock-transfected and parental controls (middle and right panel).
66c14 carcinoma cells aggressively metastasize to the lungs ( Figure 2B left panel, white arrows and [18]). We therefore evaluated metastasis in mice bearing FGFR-2DN-expressing tumors in comparison to control. As shown in Figure 2B, lungs of mice bearing FGFR-2DN-expressing 66c14 tumors exhibited a significant reduction in size and number of metastatic nodules (left panel). Quantification of foci number revealed a two-fold decrease in mice implanted with FGFR-2DN expressing cells compared to control ( Figure 2B, right panel). As 66c14 tumors develop and disseminate in a tight and similar time lapse, mice bearing control tumors died in a narrow time period (50% mortality at 53 days post implantation, Figure 2C black line). Conversely, 100% of FGFR-2DN tumor-bearing mice survived at 53 days post implantation ( Figure 2C, red line). Together, these results provide evidence that FGFR signaling modulates 66c14 tumor growth, dissemination of tumor cells to the lungs, and overall survival of mice.

Inhibition of FGFR Signaling Suppresses Lymphangiogenesis in Primary Tumors by Reducing VEGF-C Expression in Tumor Cells
We next sought to explain the observed decrease in tumor metastasis. Development of a tumor lymphatic system has been shown to contribute to tumor dissemination [6]. Cancer cells intravasate into lymphatic vessel lumens, migrate through the lymphatic system, and successively invade lymph nodes and distal organs such as lungs [8,9]. Thus, we first ascertained that 66c14 tumor cells could be detected in the lumen of VEGFR-3-positive tumor lymphatic vessels ( Figure 3A left panel, white arrows) and that they invaded distal axillary lymph nodes ( Figure 3A right panel, white arrows). These data are in agreement with previous findings by Caunt et al., who reported that inhibition of functional intratumoral lymphatic vessels, via the blockade of VEGF-Cinduced cell functions, decreased metastasis of 66c14 cells to lymph nodes and lungs [18].
To determine whether the decrease in tumor metastasis was associated with modification of the lymphatic vessel density, tumor sections were stained with anti-VEGFR-3 or podoplanin antibodies. FGFR-2DN-expressing tumors displayed a decrease in lymphatic vessel density, compared to mock transfected control or parental groups, 6 weeks after the injection of tumor cells into the mammary fat pads (data not shown). To rule out the possibility that the vascular changes were caused by a difference in tumor size, we performed staining for lymphatic vessel markers on tumor tissue isolated 3 weeks after implantation, a time point where no significant growth difference was detected (Figure 2A). Once again, we detected a reduction in the density of VEGFR-3-positive lymphatic vessels in FGFR-2DNexpressing tumors as compared to mock-transfected or parental tumors ( Figure 3B). As FGF signaling has been shown to regulate endothelial VEGFR-3 expression [10], to avoid any underestimation of lymphatic vessel in the FGFR-DN group, podoplanin staining was also performed on tumor sections 3 weeks after implantation. As expected, the density of tumor lymphatic vessel was reduced in FGFR-2DN tumors versus control groups ( Figure 3C). Interestingly, a switch from lumenized vessels to an isolated endothelial cell (EC) phenotype was observed in FGFR-2DN-expressing tumors compared to control ( Figure 3B and C, white arrows).
VEGF-C and PDGF-B are the most important factors implicated in tumor lymphangiogenesis [8,22]. In our 66c14 tumor model, the vascular phenotype correlated only with an inhibition of VEGF-C mRNA expression in FGFR-2DN-expressing 66c14 compared to control tumors ( Figure 3D). This indicates that VEGF-C but not PDGF-B is an intermediate of FGF-induced tumor lymphangiogenesis. Caunt and coworkers also demonstrated that subcutaneous rat C6 glioblastoma tumors develop a VEGF-C-dependent tumoral lymphatic vessel network [18], while no lymphangiogenesis is normally observed in situ. To validate our results in an independent tumor model, FGFR-2DN or mock transfected C6 tumor cells were subcutaneously injected in immunodeficient mice as previously described [4]. A decrease in the density of VEGFR-3 and podoplanin-positive lymphatic vessels, an increase in the ratio of isolated ECs to structured vessels (Figure S5A red arrows and data not shown), and a reduction in VEGF-C mRNA expression ( Figure S5B) were also detected in FGFR-2DN-expressing C6 tumors (2A7 and C18) versus the control group (BH2). These data provide evidence that FGFR signaling promotes tumor lymphangiogenesis in vivo by regulating VEGF-C expression and EC organization in functional lumenized vessels.

Blockade of Fibroblast Growth Factor Signaling Suppresses VEGF-C Expression in Tumor Cells
We next investigated whether FGFR-DN expression, FGFRs siRNA or PD 173074 modifies VEGF-C expression in the cancer cells. VEGF-C mRNA expression was measured in FGFR2-DNexpressing, FGFRs siRNA or PD 173074-treated cells. In 66c14 cells ( Figure 4A left panel and S1B), expression of FGFR-2DN led to a strong reduction in VEGF-C mRNA level, when compared to mock-transfected cells. This decrease was confirmed at the protein level, in the concentrated supernatant of FGFR-2DN-expressing 66c14 cells compared to control cells (Fig 4A right panel and S1C). PD-173074 (5-30 mM) and FGFR-2, but not FGFR-1, siRNA also down-regulated VEGF-C mRNA expression ( Figure 4B and C; respectively). Conversely, FGF-2-treated 66c14 cells displayed an increased VEGF-C mRNA expression ( Figure S6). In C6 cells, a decrease in VEGF-C was also detected in FGFR-2DNtransfected cells (clones 2A7, C18) in comparison to control (BH2, Figure S7A).
Our data clearly demonstrate that FGFR signaling positively regulates VEGF-C expression in 66c14 and C6 cancer cell types and that FGFR blockade inhibits this.

FGFR Signaling Stimulates VEGF-C Expression Independently of Either COX-2, HIF-1a or NF-kB
Both cyclooxygenase-2 (COX-2) and hypoxia inducible factor-1a (HIF-1a) have been reported to modulate VEGF-C expression in different tumor models [15,16,17]. FGFR-2DN-expresssing 66c14 cells displayed a decrease in COX-2 and HIF-1a mRNA expression levels, when compared to mock-transfected cells ( Figure  S8A). Decreased mRNA levels were also seen when cells were treated with PD-173074 ( Figure S8B). However, inhibition of COX-2 or HIF-1a in 66c14 cells using specific inhibitors (5-100 mM of NS-398 and 30-100 mM of #400083, inhibitors of Cox2 and HIF-1a, respectively) was not sufficient to modify VEGF-C mRNA expression level in either normoxia or cobalt chloride-induced hypoxia ( Figure S8C and D, respectively). However, changes in VEGF-A expression were detected, validating the cobalt-induced hypoxia and inhibitor efficacy. Similarly, chemical inhibition of NF-kB, another known regulator of VEGFs, did also not alter VEGF-C expression (0.1 to 5 mM, Cayman chemical #CAY10512, data not shown). As expected, NS-398 had no effect on VEGF-C mRNA expression in C6 cells either ( Figure   S7B). These data clearly show that VEGF-C expression is regulated by FGFs through a mechanism other than simple regulation of COX-2, HIF-1a and NF-kB.

FGFR Signaling Stimulates Expression of Lymphangiogenic Factors in Endothelial Cells
We next investigated whether tumor-derived FGFs could act in a paracrine manner on lymphatic endothelial cells. We examined the expression of selected lymphangiogenic genes in HMVEC-dLys treated with FGF-2 for 48 hours. We found that activation of FGFR signaling in HMVEC-dLys led to an increase in the expression level of VEGF receptors (VEGFR-2 and VEGFR-3), netrin-1, prox1 and integrin a9 ( Figure 5A-B). These data further demonstrate that FGFR signaling acts on lymphatic endothelial cells to enhance the lymphangiogenic response.

In vitro Lymphangiogenesis is Inhibited by Expression of FGFR-2DN in Tumor Cells
We next examined whether inhibition of FGF signaling in tumor cells is sufficient to explain the reduction in tumor lymphangiogenesis using an in vitro lymphangiogenesis co-culture assay [5,20]. Briefly, FGFR-2DN expressing 66c14 and C6 cells or empty vector transfected control cells, were mixed with growth factor-reduced matrigel or collagen-1, respectively, and added onto tissue-culture plates. After polymerization, HMVEC-dLys were seeded on the top of the tumor cell-containing gels. Lymphatic tube formation was assessed by LEL or prox1 staining (for 66c14 and C6 cocultures, respectively) and confocal microscopy. As expected, lymphatic tubes were observed when gels contained mock-transfected cells ( Figure 6A1, B and S9A), while isolated unorganized lymphatic EC clusters were observed in the presence of FGFR-2DN-expressing cells ( Figure 6A2;3, B and S9B-C). The same result was obtained when control cells were treated with PD-173074 ( Figure 6A6 and B).
To prove that lymphatic tubulogenesis depended on endothelial soluble factors, rather than a cell-cell contact mechanism, lymphatic endothelial cells, plated on matrigel, were incubated with the supernatants from FGFR-2DN C4-, C22-expressing, control or parental 66c14 cells ( Figure 6C). Lymphatic endothelial cells formed vascular tubes upon treatment with parental and control cells supernatants, while isolated lymphatic ECs were observed in the FGFR-2DN-conditioned media groups ( Figure 6C).
To demonstrate the VEGF dependency, VEGF-A and VEGF-C recombinant proteins were supplemented into the gel with FGFR2-DN-expressing 66c14 cells. VEGF addition rescued the loss of lymphatic tube formation ( Figure 6A4 and B). On the other hand, mock-transfected 66c14 cells mixed into the gel with VEGFR-2/Fc and VEGFR-3/Fc chimeras ( Figure 6A5 and B), led to an inhibition of tubulogenesis (as compared to Figure 6A1;5 and B). Finally, blood vascular endothelial cells (human dermal microvascular endothelial cells, HMVEC-ds), which endogenously secrete VEGF-A and VEGF-C, were seeded instead of FGFR-2DN-expressing cells. This stimulated the formation of capillarylike structures ( Figure S9D).
Together, these data indicate that FGFR signaling stimulates, in both tumor cell lines in vitro lymphatic vessel formation, through VEGF-C.

Discussion
Fibroblast growth factors are expressed in a number of tumors and have regulatory functions in tumor development. Thus, targeting FGFR signaling may hold promise for therapy [1]. However, attention has been mostly focused on other receptor tyrosine kinase signaling pathways, such as VEGF receptors, so far, resulting in an incomplete study of the role of FGFR signaling in tumor progression and dissemination.
We demonstrate herein that inhibition of FGFR signaling not only impairs tumor growth in vitro and in vivo, but also affects lymphangiogenesis and metastasis. As previously reported, we confirm the existence of an autocrine FGF-loop in tumorigenesis [4,23]. The intracellular effector Erk is activated by FGFR signaling, and its inhibition correlated with an impairment of cancer growth [1], an increased p21 and a decreased cyclin D1 expression [24]. Similarly, up-regulation of the proto-oncogene cmyc, upon inhibition of FGFR signaling, might reflect an arrest in cell growth and an induction of cell apoptosis [25].
A reduction in lung metastatic foci size and number was also observed in mice implanted with FGFR-2DN cells. Cancer cells need to leave the primary tumor site and to invade distant organs. Taeger et al. reported that inhibition of FGFR signaling results in a decrease in tumor cell motility due to reduction in integrins, extracellular matrix proteins, mediators of epithelial-mesenchymal transition, and MMPs [23,26]. However, we observed that FGFR signaling blockade, in 66c14 carcinoma cells, did not alter significantly expression of EMT and invasion markers, except for MMP-14, confirming the role of FGF family molecules in its regulation [27].
Other evidence over the past 15 years also clearly reveals the crucial role of the tumor lymphatic system, and its main inducer, VEGF-C, in metastasis [8]. However it is not yet known whether FGFR signaling has a role in tumor lymphangiogenesis. We demonstrate herein that FGFR signaling indeed plays a role in tumor lymphangiogenesis. Inhibition of FGF activity in tumor cells abrogated tumor lymphangiogenesis in vitro and in vivo by downregulating VEGF-C expression, in two independent tumor models. To our knowledge, this is the first demonstration that, in two different tumor settings, FGFR signaling modulates tumor lymphangiogenesis.
In addition to this autocrine mechanism on tumor cells, FGF acts directly on lymphatics by modulating lymphangiogenesis genes. We reported that loss of FGF signaling led, in vitro and in vivo, to a disruption of lymphatic vessel morphogenesis and an increase in the number of isolated endothelial cells. This was rescued in vitro by exogenous VEGFs. Interestingly, a similar phenotype was seen in FGFR-DN-expressing (blood) vascular endothelial cells that down-regulate both VEGFR-2 expression and response to VEGF-A stimulation [14]. Finally, it has been demonstrated that both FGFR and VEGFR signaling is necessary, to promote EC migration and tubular morphogenesis [28]. Taken together, these results emphasize the idea of an interconnection or synergy between the FGF and VEGF pathways. Onimaru et al. described a similar lymphangiogenic connection between FGF and PDGF-B [29]. However, inhibition of FGFR signaling in 66c14 tumor cells did not alter PDGF-B expression level in vitro or in vivo.
We also observed that the down-regulation of VEGF-C was associated with a decrease in both COX-2 and HIF-1a mRNA expression. COX-2 has been shown to induce lymphangiogenesis in human breast cancer via an upregulation of VEGF-C [15]. However, in our models, COX-2 inhibitors did not modulate VEGF-C expression in tumor cells; in vitro, similar to what has been previously reported [30]. It is likely that COX-2 activates lymphangiogenesis indirectly through other mechanisms, such as the upregulation of VEGF-C in macrophages [31]. Alternatively, VEGF-C might function as an upstream regulator of COX-2, leading to their successive inhibition upon FGFR inhibition [32]. HIF-1a is a well-studied inducer of VEGF-A ligand and receptor expression [33]. Several publications have also reported that VEGF-C expression correlates with HIF-1a levels [16,17]. In our hands, inhibition of HIF-1a did not suppress VEGF-C expression, although a decrease in VEGF-A expression was detected. Further investigations are needed to completely elucidate the mechanisms controlling the induction of VEGF-C by FGFs.
It has been shown that the lymphangiogenic activity of FGF-2 is mediated, in vitro by the Akt/mTOR/p70S6 kinase pathway [11]. Furthermore, inhibition of pancreatic tumor lymphangiogenesis and metastasis has been achieved by rapamycin, a specific inhibitor of mTOR, which down-regulates VEGF-C [34]. As we reported that blockade of FGFR signaling reduces mTOR/p70S6 kinase pathway in 66c14, the potential link between FGFR signaling and mTOR in the regulation of VEGF-C expression needs further studies.
Our findings that FGFR signaling induces lymphangiogenic molecules in lymphatic endothelial cells, including VEGFR-2, VEGFR-3, netrin-1 and integrin a9, also validate the idea of FGF acting in a paracrine manner on tumor lymphatic endothelial cells. This strengthens the positive role of FGFs in tumor progression and dissemination.
In agreement with our results, Murakani and coworkers have recently demonstrated that dominant negative FGFR-expressing blood endothelial cells down-regulate VEGFR-2 expression, leading to a loss of response to VEGF stimulation in vitro and in vivo [14].
We also demonstrated that FGF signaling induces netrin-1 in lymphatic endothelium. Despite controversial data in the vasculature [35,36,37], compelling evidences support that netrins are pro-lymphangiogenic factors in vitro and in vivo [19,38] and data not shown. Moreover, it has been shown that netrin-1, through its canonical receptors stimulates tumor growth and metastasis [39,40]. Netrin-1 is highly expressed in metastatic human breast and pancreatic tumors and its inhibition by siRNA or soluble receptor strategies suppresses metastasis formation [39,40].
Additional evidence indicates an important role for integrins in lymphangiogenesis and tumor metastasis [41,42,43,44]. Integrin a9 expression is regulated in lymphatic endothelium by prox1. This integrin is required for lymphatic endothelial cell migration in vitro and development of a murine lymphatic system in vivo [45,46]. Finally, direct interactions between lymphatic growth factors and integrins have been described, further supporting their role in lymphangiogenesis [47].
Taken together, these results confirm an important role for FGFR signaling in promoting tumorigenesis (Figure 7). FGFR signaling has pleiotropic effects on tumor development including (1) control of tumor growth, (2) promotion of tumor metastasis, (3) stimulation of tumor lymphangiogenesis, and (4) induction of VEGF-C and expression of VEGFRs or other pro-lymphangiogenic/survival factors in tumor cells and the lymphatic endothelium. Thus, antagonizing FGFR activity may be an interesting approach for anti-cancer therapy and not only limit primary tumor growth but also tumor lymphangiogenesis and metastatic spread.