Fibroblast Growth Factors Are Required for Efficient Tumor Angiogenesis¹

Neoplastic growth and progression to tumor malignancy are dependent on the formation of new blood vessels (tumor angiogenesis;Ref. 1), and clinical studies have revealed that high vessel density often correlates with poor prognosis (2). Angiogenesis is regulated by a fine balance between inducers and inhibitors of this process (angiogenic and angiostatic factors,respectively; Ref. 3). VEGF-A³ and members of the FGF family, in particular, acidic FGF (FGF1) and basic FGF(FGF2), are potent inducers of angiogenesis. By binding their receptors on endothelial cells, they are able to stimulate endothelial cell proliferation, migration, and differentiation and to induce angiogenesis in vitro and in vivo(4, 5, 6). In addition, FGFs can synergize with VEGF in the induction of angiogenesis, probably by up-regulating VEGF and VEGF receptors in endothelial cells (7, 8, 9, 10).

Whereas the involvement of VEGF in tumor angiogenesis has been demonstrated repeatedly in various murine tumor models and human tumor xenograft experiments (4), experimental evidence for a direct involvement of FGFs in the regulation of tumor angiogenesis is rather circumstantial. Expression of FGF1 and FGF2 is often increased in tumors and correlates with the degree of cancer malignancy(11), and high concentrations of FGF2 have been found in both the serum and urine of cancer patients (12). However,a direct demonstration of an involvement of FGFs in tumor angiogenesis is hampered by the highly pleiotropic activities of the factors; FGFs can activate FGFRs expressed on a number of different target cells,including tumor cells, endothelial cells, and fibroblasts. For example,tumorigenicity of melanoma cells was efficiently repressed by interference with FGF2 or FGFR1 function (13). However,melanoma cells are known to depend on FGF2 for their proliferation, and it was not possible to functionally separate inhibition of tumor angiogenesis from repression of tumor cell proliferation. Similarly,s.c. tumors induced by 3T3 fibroblasts expressing a constitutively secreted form of FGF2 were repressed in their growth by neutralizing antibodies against FGF2 (14). Again, although tumor vascularization was impaired, the transforming and angiogenic activity of FGF2 could not be distinguished.

The onset of tumor angiogenesis (the angiogenic switch) has been functionally defined by Folkman et al. (15) in a transgenic mouse model of multistage tumorigenesis. In these mice,SV40 T antigen is expressed under the control of the rat insulin promoter, resulting in the development of β-cell tumors in the pancreatic islets of Langerhans (Rip1Tag2; Ref. 16). Several tumor stages can be distinguished, including normal islets,β-cell hyperplasia, adenoma, and carcinoma (17, 18). Notably, hyperplastic islets can be classified into two types:(a) nonangiogenic islets that do not affect cocultured endothelial cells; and (b) angiogenic islets that are able to induce proliferation, migration, and tube formation of endothelial cells (15). Based on these experiments, it has been proposed that soluble factors are involved in the onset of tumor angiogenesis. Normal islets of Langerhans already express VEGF-A, and its expression and secretion are moderately up-regulated in β tumor cells (19). Similarly, FGF1 is expressed in normal islets of Langerhans and throughout β-cell tumorigenesis. Notably, β tumor cells gain the capability of exporting FGF1, although it does not carry a signal sequence for secretion (20). A similar switch to the export of signal-less FGF2 has been reported during fibrosarcoma development in transgenic mice (21), and it has been speculated that the export of these angiogenic FGFs may be involved in the onset of tumor angiogenesis.

Previously, soluble receptors have been used efficiently to address the role of growth factors during tumor neovascularization. For example,inhibition of VEGF activity by soluble VEGF receptors [sFlt or soluble Flk (sFLK)] has been instrumental in demonstrating that VEGF is required for angiogenesis (22, 23, 24, 25). Similarly, experiments with soluble Tie-2 receptor have shown that angiopoietins contribute to tumor growth (26). In this report, we have used a soluble form of FGFR2 IIIb (sFGFR) to interfere with FGF function during tumor angiogenesis. Previously, this soluble receptor has been reported to impair several developmental processes, causing early embryonic lethality when constitutively expressed in transgenic mice(27). To express high levels of sFGFR in adult animals, we generated a recombinant adenovirus encoding for sFGFR. Expression of sFGFR dramatically repressed the induction and maintenance of tumor angiogenesis and tumor growth in allograft transplantation experiments and in Rip1Tag2 transgenic mice.

The cDNA encoding sFGFR was provided by Dr. G. Merlino (NIH, Bethesda,MD) and has been described previously (27). The extracellular domain of mouse Flt-1 (amino acids 1–758) was derived from a plasmid provided by Dr. E. F. Wagner (IMP, Vienna,Austria) and fused to the same mouse immunoglobulin Fc tail used in the sFGFR construct. Recombinant E1/E3 defective adenoviruses were constructed using homologous recombination in E. coli(28). All genes of interest were under control of a cytomegalovirus immediate early promoter, followed by a rabbitβ-globin intron/polyadenylation signal. Virus cultures were initiated by transfecting the linearized genomes into 293 cells using polyethylenimine (29). After amplification of the culture,virus was purified by banding twice on CsCl gradients, transferred into HEPES-buffered saline (HBS)/40% glycerol by passage over a gel filtration column, and stored at −80°C as described previously(30). Viral quantitation was based on protein content using the conversion 1 mg of viral protein = 3.4 × 10¹² virus particles.

HUVECs and mouse microvascular endothelial cells (1G11; Ref.31) were cultured in DMEM supplemented with 20% FCS (Life Technologies, Inc., Gaithersburg, MD), 2 mm glutamine, 40μg/ml bovine brain extract, 80 units/ml heparin, and antibiotics. The medium for bovine capillary endothelial cells was supplemented with 10% FCS and FGF2 (2.5 ng/ml). The β tumor cell line βHC13T originated from a hyperplastic β tumor cell line derived from Rip1Tag2 transgenic mice (βHC13; Ref. 32) by transplantation into immunodeficient mice and subsequent culture from the allografted tumors in DMEM supplemented with 4.6 grams/liter (w/v)glucose, 10% FCS, 2 mm glutamine, and antibiotics.

HUVECs were infected with adenoviruses (2000 PPC) and starved for 48 h in serum-free medium. Lysates and conditioned media were prepared, and proteins were resolved by 7.5% SDS-PAGE and transferred to nylon membranes. Membranes were probed with 1 μg/ml anti-mouse IgG1 antibody specific for mouse immunoglobulin heavy chain (Pierce,Rockford, IL). Immunostained proteins were visualized using the enhanced chemiluminescence detection system (Amersham, Buckinghamshire,United Kingdom) according to the manufacturer’s recommendations. Serum samples from Rip1Tag2 mice (500 μl serum/mouse) were precleared by incubation with protein G-agarose beads. Unbound proteins were then adsorbed to heparin-Sepharose, and bound proteins were analyzed by immunoblotting as described above.

HUVECs were plated in triplicate in 24-well dishes (40,000 cells/well)and infected with adenoviruses (10,000 PPC) in reduced serum medium for 2–3 h. Cells were starved for 24 h in 5% FCS medium and then stimulated overnight with recombinant FGF1 (provided by Dr. T. Maciag;Maine Medical Center Research Institute, South Portland, ME) and heparin, FGF2 (Promega, Madison, WI), or VEGF (R&D Systems,Minneapolis, MN). After stimulation, 1 μCi of[³H]thymidine was added per well, and incorporation of thymidine was determined by scintillation counting as described previously (33).

Angiogenic tumor stages were isolated from Rip1Tag2 transgenic mice as described previously (19). HUVECs were infected with adenoviruses (10,000 PPC) and allowed to recover for 48 h in complete medium. Cells were then trypsinized, resuspended in 10% FCS RPMI 1640, and cocultured with tumor stages in collagen matrix as described previously (15). After 2–3 days, the response of endothelial cells to the angiogenic tumor stages was scored. The experiment was performed three times with approximately 40–50 islets each.

βHC13T cells (1 × 10⁶ cells in 200 μl of PBS) were injected s.c. together with 1 × 10¹⁰ particles of the respective adenoviruses in both flanks of MF1 nu/nu mice. After 4 days,5 × 10⁹ particles of each adenovirus in 100 μl of PBS were injected intratumorally. Tumor diameter was determined daily, and tumor volume was calculated based on the spheric shape of the tumors.

Particles (1 × 10¹⁰) of each adenovirus in 200 μl of HEPES-buffered saline (HBS) were injected i.v. in the tail vein of Rip1Tag2 mice. Mice were sacrificed,and tumor volume was determined based on the spheric shape of the tumors.

Tumors and pancreata were fixed overnight in 4% paraformaldehyde in PBS and immersed in 30% sucrose/PBS for 12 h. Finally, they were embedded in OCT (Tissue Tek, Torrance, CA) and snap frozen in liquid nitrogen. For BrdUrd labeling, 2 h before sacrifice, mice were injected i.p. with 100 μg of BrdUrd (Sigma, St. Louis, MO) per gram of body weight. Sections (10-μm thick) were cut, mounted on silane-coated slides, and immunostained as described previously(34). For detection of apoptosis, the TUNEL technique was used as described previously (34). Vessel density was determined by CD31 (PECAM-1) immunostaining. Specific antibody staining was visualized using the ABC-Vector horseradish peroxidase kit according to the manufacturer’s recommendations (Vector Laboratories,Burlington, CA). Antibodies were antimouse CD31 antibody (1:50;PharMingen, San Diego, CA) and anti-BrdUrd antibody (1:50; Zymed, San Francisco, CA). Proliferating and apoptotic cells were counted in at least 10 comparable fields per section (magnification, ×200), and blood vessels were counted in at least 10 comparable fields per section(magnification, ×100).

Using soluble receptor constructs, we wished to interfere with the functions of FGFs and VEGF during tumor angiogenesis. Stabilized soluble forms of FGFR2 IIIb (sFGFR) and VEGF receptor 1 (sFlt) were generated by fusing the extracellular domain of each receptor to a mouse immunoglobulin heavy chain (Fc). To express high levels of sFGFR or sFlt in endothelial cells, replication-deficient, recombinant adenoviruses were generated, in which sFGFR (AdsFGFR) or sFlt (AdsFlt)were expressed under the control of the cytomegalovirus enhancer/promoter. Adenoviral expression of functional sFGFR and sFlt-1 was confirmed in cultured endothelial cells in vitro. When HUVECs were infected with 2,000 PPC of AdsFGFR and AdsFlt, high levels of sFGFR and sFlt were synthesized and secreted efficiently into the culture medium (Fig. 1,A). To assess the capability of sFGFR to block FGF-induced endothelial cell proliferation, HUVECs were infected with AdsFGFR, AdFlt, or a control virus expressing eGFP (AdeGFP) at 10,000 PPC, starved in 5% FCS serum medium, and then stimulated with FGF1 (10 ng/ml) and heparin (10μg/ml), FGF2 (5 ng/ml), or VEGF (5 ng/ml). The incorporation of radiolabeled thymidine was determined as a measurement for the induction of DNA synthesis and thus proliferation. FGF1/heparin-induced DNA synthesis was completely inhibited by AdsFGFR but not by AdeGFP(Fig. 1,B). In contrast, only a mild inhibitory effect of AdsFGFR was observed with cells that were treated with FGF2, indicating that sFGFR interfered predominantly with FGF1 function, interfered to a lesser extent with FGF2 function, and did not interfere at all with VEGF function (Fig. 1,B) The strong specificity of sFGFR for FGF1 reflects the higher binding affinity of FGFR2 IIIb for FGF1 in comparison to FGF2 (27). Conversely, AdsFlt was able to block endothelial cell proliferation induced by VEGF but not that induced by FGF1/heparin or FGF2 (Fig. 1 B). Similar results were obtained with mouse microvascular endothelial cells (1G11) and bovine capillary endothelial cells (data not shown).

Endothelial cells differentiate into net-like structures when plated on a Matrigel substrate in the presence of growth factors, such as FGF1,FGF2, or VEGF (35). Expression of sFGFR also effectively inhibited FGF1-induced endothelial cell differentiation in this assay(data not shown). Together, these results indicate that sFGFR is able to interfere with FGF1-induced endothelial cell proliferation and differentiation.

Having established that sFGFR is a valid tool to interfere with FGF function in vitro, we next wanted to determine whether sFGFR inhibits tumor angiogenesis in vivo. For this purpose, we used the Rip1Tag2 transgenic mouse model of β-cell carcinogenesis(see “Introduction”). In these mice, the switch to the angiogenic state can be visualized by coculturing hyperplastic islets of Langerhans with endothelial cells in a three-dimensional collagen gel matrix (15). In the presence of an angiogenic hyperplastic islet, endothelial cells proliferate and migrate toward the tumor biopsy. To assess whether sFGFR is able to interfere with this angiogenic response, HUVECs were infected with either AdeGFP or AdsFGFR before coculturing them with angiogenic islets isolated from Rip1Tag2 transgenic mice. Three classes of endothelial cell response could be distinguished: (a) a strong response characterized by chemotactic migration of endothelial cells toward the islet and survival of many endothelial cells in the vicinity of the islet;(b) a weak response characterized by the migration and survival of only a subset of endothelial cells; and (c) no response characterized by failure of the endothelial cells to migrate toward the angiogenic islet and by their subsequent death (Fig. 2,A). With noninfected or AdeGFP-infected endothelial cells, 60–80% of the samples showed a strong response toward the islets, and only a minority of cases exhibited a weak response or no response (Fig. 2,B), also indicating that most of the selected islets were indeed angiogenic. In contrast, with endothelial cells expressing sFGFR, 60–70% of the samples did not exhibit any reaction toward the angiogenic islets (Fig. 2 B). In control experiments, infected endothelial cells were cultured in a collagen gel in the absence of any tumor biopsy. These experiments did not reveal any significant difference between cells infected with AdsFGFR or AdeGFP, indicating that the inhibitory effect of AdsFGFR was due to the inhibition of FGFs released by the tumor biopsies (data not shown). These results indicate that FGFs are required for endothelial cell migration, proliferation, and survival,which are hallmarks of ongoing tumor angiogenesis, as visualized in this collagen gel assay.

We next assessed whether interference with FGF function by sFGFR would affect tumor growth in allograft tumor transplantation experiments. Tumor cell lines that have been derived from β-cell tumors of Rip1Tag2 mice (βHC13T) were transplanted under the skin of immunodeficient mice. Previously, we have reported that β tumor cells express FGF1 and efficiently export FGF1 into the culture medium(20). However, although β tumor cell lines express several FGF receptors on their surface, in particular, FGFR4, their proliferation is not stimulated by recombinant FGFs (Ref.36; data not shown). Consistent with these results, the growth rate of βHC13T was not significantly changed on infection with AdeGFP, AdsFlt, or AdsFGFR (Fig. 3). These observations allowed us to assess the contribution of FGF1 to tumor growth in a setting where FGFs did not directly affect tumor cell proliferation but rather stimulated cells of the stromal compartment,for example, endothelial cells.

Immunodeficient (MF1 nu/nu) mice (five mice/group) were injected s.c. in both flanks with 1 × 10⁶ tumor cells together with 1 × 10¹⁰particles of AdeGFP, AdsFGFR or AdsFlt. One group was injected with a combination of AdsFGFR (5 × 10⁹ particles) and AdsFlt (5 × 10⁹ particles). In the first few days after implantation, mice infected with AdsFGFR did not exhibit any significant delay in tumor development as compared with AdeGFP-infected controls. In contrast, infection with AdsFlt resulted in a significant inhibition of nodule formation, indicating a critical role of VEGF in the initial phases of tumor transplantation (Fig. 4,A). Consistent with the inability of sFGFR to interfere with this process, an intermediate delay in tumor onset was observed with the combination of the two adenoviruses where half the dose of AdsFlt was applied together with AdsFGFR (Fig. 4,A). Four days after transplantation of the tumor cells, 5 × 10⁹ particles of the respective adenoviruses were injected intratumorally, based on a previous report demonstrating that intratumoral injection of AdsFlt maintained high levels of sFlt expression (23). One mouse per group was sacrificed to monitor tumor state, and, subsequently,tumor size was determined daily. AdeGFP-injected tumors started to grow exponentially, whereas AdsFGFR treatment resulted in a significant repression of tumor outgrowth over the following 6 days (Fig. 4,B). At this later time point, AdsFlt-injected tumors started to grow at rates that were comparable with those of AdeGFP-injected controls. In contrast, combined expression of sFGFR and sFlt exhibited a strong synergistic effect in blocking tumor growth(Fig. 4,B), which was also demonstrated by the dramatic decrease in final tumor weight at the end of the experiment (Fig. 4,C; P < 0.005). Histopathological analyses of the latter tumors revealed a high degree of necrosis, and the residual tumor cells were found to cuff around major vessels (Fig. 4,D), a situation that is reminiscent of tumors that have been treated with antiangiogenic compounds (37). Quantitation of tumor cell proliferation (BrdUrd incorporation),apoptosis (TUNEL staining), and blood vessel density (CD31 staining)revealed that the mitotic index of tumor cells was not affected by the expression of sFGFR or sFlt, whereas the extent of apoptosis was marginally increased in tumors of AdsFGFR-infected mice (Table 1). In contrast, vessel density, a measure of the extent of tumor vascularization, was decreased by approximately 50% in tumors of AdsFGFR- and AdsFlt-infected mice(Table 1). Tumors of mice that had been treated with both AdsFGFR and AdsFlt were too small to be included in this statistical analysis.

sFGFR and sFlt, respectively, could be detected by immunoblotting in the blood of injected mice, suggesting that the soluble receptors were systemically present in the infected mice (data not shown). Histological examination revealed that the majority of tumor cells in AdeGFP-infected mice expressed eGFP 4 days after the initial coinjection and at lower levels at the end of the experiment (data not shown), suggesting that tumor cells were the main producers of sFGFR and sFlt in AdsFGFR- and AdsFlt-infected mice.

To assess the contribution of FGFs to tumor progression and tumor angiogenesis in vivo, Rip1Tag2 mice were injected i.v. once a week with 1 × 10¹⁰ particles of AdeGFP, AdsFlt, AdsFGFR, or a combination of AdsFGFR and AdsFlt. Titration experiments in unrelated control mice revealed that in this experimental setting, serum levels of soluble receptors were highest 6 days after i.v. injection, and approximately 70% were still detectable 10 days after injection (data not shown). Tumor progression in Rip1Tag2 transgenic mice is highly reproducible, i.e., all mice develop β-cell tumors with similar incidence and kinetics and die around 14 weeks. Hence, injections were applied between 8 and 12 weeks of age, a time period during which angiogenesis was known to be highly active in β-cell tumorigenesis. No apparent effects on the general health of the animals was observed during the course of the experiments. At 12 weeks of age, the mice were sacrificed, and tumor volumes were determined. In AdsFGFR-infected mice, tumor volumes were dramatically reduced as compared with AdeGFP-infected control mice(P < 0.005; Fig. 5,A). Tumor volumes were also significantly reduced in AdsFlt-infected mice, however, they were reduced to a lesser degree as compared with AdsFGFR-infected mice(P < 0.01; Fig. 5,A). A synergistic effect between AdsFGFR and AdsFlt was not apparent in these experiments, rather the reduction of the AdsFGFR dose by half in the combination treatment resulted in an intermediate response (Fig. 5 A). These results indicate that in this experimental setting, sFGFR is more efficient than sFlt in repressing tumor growth.

Analysis of tumor cell proliferation, apoptosis, and vessel density revealed a modest decrease in vessel number of AdsFGFR- and AdsFlt-infected mice as compared with control mice infected with AdeGFP(Fig. 6 ; Table 2). Concomitantly, a modest increase in the apoptotic index was observed. Proliferation rates remained unchanged, indicating that interfering with FGF and VEGF function did not directly affect β tumor cell proliferation (Table 2). A comparable decrease in vessel density has been reported in Rip1Tag2 mice treated with antiangiogenic agents (38, 39),suggesting that sFGFR and sFlt repressed the formation of new blood vessels (angiogenesis) rather than eliminating preexisting vessels. Histological examination of AdeGFP-infected mice showed that when these viruses were applied i.v., they predominantly infected the parenchymal cells of the liver and, to a lesser degree, the spleen (data not shown). Immunoblotting analysis of blood serum samples from adenovirus-infected mice revealed detectable levels of sFGFR and sFlt(Fig. 5 B), suggesting that the soluble receptors were produced mainly by hepatic and splenic cells and delivered via the blood stream to the tumor sites in the pancreas. It should be noted that serum levels of the soluble receptors as determined by immunoblotting did not significantly correlate with the degree of tumor repression, indicating that the levels of soluble receptors were not rate-limiting in these experiments.

In the present study, we have used a recombinant adenovirus that encodes a soluble form of FGFR 2 IIIb (AdsFGFR) to assess the functional contribution of FGFs to tumor angiogenesis. This form of FGF receptor binds FGF1, FGF3, FGF7, and FGF10 with high affinities and consequently inhibits their function (27). When cultured endothelial cells were infected with AdsFGFR, sFGFR specifically repressed proliferation and differentiation induced by FGF1, indicating that AdsFGFR is an appropriate tool to study the involvement of FGFs in tumor angiogenesis in vivo.

The onset of tumor angiogenesis can be nicely observed in a collagen gel assay in which hyperplastic islets of Langerhans isolated from Rip1Tag2 mice are cocultured with endothelial cells (15). Angiogenic islets readily induce endothelial cell proliferation,chemotactic migration, and survival. Besides VEGF-A, angiogenic islets express FGF1, FGF7, and FGF8, among which FGF1 is the only FGF with angiogenic activity (36). The fact that sFGFR has high specific affinity for FGF1 (Fig. 1) and the fact that expression of sFGFR efficiently inhibits the angiogenic response in the collagen gel assay indicate that FGF1 is one of the factors required for the transition from a nonangiogenic to an angiogenic tumor stage. Moreover,despite the lack of a signal sequence for secretion, FGF1 is efficiently exported by β tumor cell lines (20), raising the possibility that FGF1 secretion by β tumor cells contributes to the onset of tumor angiogenesis. The requirement for VEGF in this experimental setting has been demonstrated recently (42). Because VEGF-A exhibits a synergistic effect with FGF1 and FGF2 under these experimental conditions, it is conceivable that both FGF1 and VEGF-A are required for the switch to the angiogenic phenotype during Rip1Tag2 tumorigenesis (7, 8, 19, 20).

In allograft transplantation experiments, coinjection of AdsFlt withβ tumor cells resulted in an initial delay of tumor formation;however, at later time points, tumor growth was only moderately affected. In contrast, expression of sFGFR did not inhibit the early phases of tumor formation but rather repressed tumor growth at later time points. Similar to sFlt, sFGFR mediated its inhibitory function by interfering with tumor angiogenesis, as demonstrated by a 2-fold reduction in vessel density as compared with control tumors. Finally, a combination of sFGFR and sFlt revealed a synergistic effect in the repression of tumor growth, indicating that FGF1 and VEGF-A cooperate in inducing tumor neovascularization. Tumor cell proliferation was not affected by any of the soluble receptors in vitro or in vivo, excluding the possibility that tumor repression was due to a direct inhibition of tumor cell proliferation.

In Rip1Tag2 mice, i.v. injection of AdsFGFR dramatically repressed tumor growth. Whereas tumor cell proliferation was not affected, vessel density was modestly but significantly decreased (Table 2). In particular, the number of small microcapillaries was reduced, whereas larger vessels were less affected, suggesting that sFGFR repressed tumor growth by inhibition of tumor angiogenesis (Fig. 6). Several lines of evidence support the notion that β tumor cell proliferation or survival is not directly affected by interfering with FGF function:(a) proliferation of β tumor cells was not affected by the expression of sFGFR in Rip1Tag2 transgenic mice (Table 2);(b) tumor cell lines established from Rip1Tag2 transgenic mice did not respond to exogenous FGFs and were not affected by the expression of dominant-negative or soluble forms of FGFR (Fig. 3; data not shown); and (c) tumorigenesis was not affected in Rip1Tag2 mice deficient for FGFR4, the FGF receptor that is predominantly expressed by β tumor cells (36).

In our experiments, sFGFR was more effective in the inhibition of tumor growth than sFlt, a treatment that has been demonstrated to efficiently repress physiological and tumor angiogenesis (22, 23, 25, 40, 41). However, although both soluble receptors were detectable in the serum of infected mice, protein levels and half-life might vary significantly; therefore, a quantitative comparison between the two reagents is not possible. The differences observed between the allograft transplantation experiments and the experiments with Rip1Tag2 mice may be due to the fundamental differences between fast tumor formation of cultured tumor cell lines and the slow genesis of endogenously growing tumors.

From our results, we conclude that FGFs, together with VEGF, contribute to tumor angiogenesis and that inhibition of their activities represses tumor growth. During the multiple stages of tumor development, FGFs, in particular the angiogenic factors FGF1 and FGF2, exert highly pleiotropic functions. They can promote tumor cell proliferation in an autocrine fashion; they can stimulate the growth, survival, and migration of stromal cells, including fibroblasts and smooth muscle cells; and, as demonstrated in this study, they participate in the induction of tumor angiogenesis. Thus, therapeutic approaches that are based on the inhibition of FGF function may allow the simultaneous targeting of different cell types. Moreover, such treatment may potentiate therapeutic inhibition of VEGF function in cases in which both growth factors are expressed and act in a synergistic manner.

We thank Mediyha Saltik and Karl Mechtler for technical support,W. Jochum for expertise in histopathology, and B. LaRochelle, G. Merlino, T. Maciag, E. Dejana, and A. Vecchi for reagents. We are grateful to M. Busslinger and A. Neubüser for critical comments on the manuscript and Hannes Tkadletz for art work. Animal care was in accordance with institutional guidelines.