VEGF-A and FGF-2 synergistically promote neoangiogenesis through enhancement of endogenous PDGF-B–PDGFRβ signaling

Therapeutic angiogenesis, a strategy in which angiogenic growth factors improve revascularization of ischemic tissues, has been extensively investigated (Ylä-Herttuala and Alitalo, 2003). The growth factors used for therapeutic angiogenesis in previous studies were principally vascular endothelial growth factor A (VEGF-A) and fibroblast growth factor 2 (FGF-2), both of which are well-investigated pro-angiogenic molecules. VEGF-A was originally identified as a specific mitogen for endothelial cells, and the importance of VEGF-A in vascular development in vivo has been established by studies investigating VEGF-A knockout mice (Carmeliet et al., 1996; Ferrara et al., 1996). FGF-2, by contrast, has been reported to act as a mitogen for both endothelial and mural cells (D'Amore and Smith, 1993; Bikfalvi et al., 1997). The roles of FGF signaling in vascular formation have recently been reported in mouse embryo and tissue explant models using dominant-negative FGF receptor 1 (FGFR1) (Lee et al., 2000; Rousseau et al., 2003), and in embryonic stem cells lacking FGFR1 (Magnusson et al., 2004), although they have not been clarified by studies using genetically engineered mice lacking FGF ligands (Miller et al., 2000) or FGF receptors (FGFRs) (Arman et al., 1998; Deng et al., 1994; Yamaguchi et al., 1994).

Experimental studies on therapeutic angiogenesis in animals and humans using VEGF-A or FGF-2 had been carried out based on these findings. However, the results of three double-blind randomized controlled trials, one with recombinant VEGF-A and two with recombinant FGF-2, were less favorable than expected (Henry et al., 2003; Simons et al., 2002; Lederman et al., 2002). Stimulation with single angiogenic molecules thus appears to be insufficient to induce functional vessels in humans.

The combination of VEGF-A and FGF-2, however, has been reported to have potent synergistic effects on neovascular formation in experimental conditions both in vivo and in vitro (Pepper et al., 1992; Goto et al., 1993; Asahara et al., 1995). In addition to the combination of VEGF-A plus FGF-2, that of VEGF-A and platelet-derived growth factor BB (PDGF-BB) (Richardson et al., 2001), and that of FGF-2 and PDGF-BB (Cao et al., 2003) have also been recently reported to have a potent synergistic effect in inducing neovascularization in experimental animal models in vivo. However, the mechanisms underlying this synergism are not well understood.

In the present study, we examined the mechanism of angiogenic synergism between VEGF-A and FGF-2. Stimulation with VEGF-A alone directly promotes the proliferation of endothelial cells, while stimulation with FGF-2 alone directly promotes the proliferation of endothelial as well as mural cells. In addition to these direct effects, we demonstrate here that the co-stimulation with VEGF-A and FGF-2 induces efficient mural cell recruitment to nascent vessels through enhancement of endogenous PDGF-B–PDGFRβ signaling in concert: VEGF-A enhances PDGF-BB secretion by the endothelium, and FGF-2 enhances PDGFRβ expression in mural cells. Since endothelium-derived PDGF-BB protein has affinity to extracellular matrix (ECM) molecules, it is located in the vicinity of endothelial cells and may form a steep gradient in the periendothelial compartment (Abramsson et al., 2003). Mural cells with FGF-2-upregulated PDGFRβ could detect the periendothelial PDGF-BB gradient induced by VEGF-A and be recruited to the endothelium. We also demonstrate that addition of exogenous PDGF-BB, which could overwhelm the periendothelial endogenous gradient of PDGF-BB, impairs the mural cell recruitment induced by co-stimulation with VEGF-A and FGF-2 in vitro and also functional neovascular formation in vivo. These findings identify the importance of preservation of the endogenous gradient of PDGF-BB in inducing successful neovascular formation, which should be considered in designing effective strategies in therapeutic angiogenesis.

Recombinant human VEGF-A (VEGF165, R&D systems, Inc.), FGF-2 (R&D systems), PDGF-BB (PeproTech, EC, Ltd), monoclonal antibody anti-PDGFRβ (APB5, eBioscience) and non-specific IgG (eBioscience) were mixed by pipetting, in combination or separately, with regular Matrigel (BD Biosciences). FGF-2 was used with 0.1 mg/ml of heparin (Aventis Pharma, Japan) in all Matrigel plug assays in this study. The Matrigels (500 μl each) were injected subcutaneously into the abdominal region of male ICR mice (CLEA Japan Inc.) after anesthesia with Avertin. Two plugs with different ligands were injected into each mouse to avoid differences between individuals. Each Matrigel plug was harvested on day 7 and divided into two blocks, one fixed overnight in 4% paraformaldehyde (PFA) and then paraffin embedded for Hematoxylin and Eosin (H & E) staining, and another directly frozen in dry-iced acetone for immunohistochemistry. For the in vivo permeability assay, FITC-conjugated dextran (M_r 2×10⁶; Invitrogen Molecular Probes) was administered intravenously via lateral tail veins 10 minutes before harvesting the Matrigel plugs. Some samples were also fixed for electron microscopy in 2.5% glutaraldehyde. All experimental protocols were conducted in accordance with the policies of the Animal Ethics Committee at the University of Tokyo.

Frozen Matrigel plugs were sectioned at 10 μm thickness in a cryostat, and then briefly fixed with 4% PFA. The fixed sections were reacted with monoclonal antibodies, anti-murine PECAM1 (Mec13.3, BD Pharmingen 553370), CD34 (FITC-conjugated RAM34, BD Pharmingen 553733) or F4/80 (CI:A3-1, Serotec Inc. MCA497), or rabbit polyclonal antibodies against PDGFRβ (Upstate Group, Inc. 06-498), NG2 (Chemicon International, Inc. AB5320) or collagen IV (Chemicon AB756P). Specimens were subsequently stained with secondary antibodies conjugated with Alexa Fluor^® 488, 594 or 647 anti-rat/rabbit IgG (Invitrogen Molecular Probes). Because anti-mouse secondary antibodies stain the whole Matrigel plug section, we used Cy3-conjugated murine monoclonal antibody for SMA (1A4, Sigma-Aldrich Corporate C6198) and murine monoclonal antibody for desmin (D33, Dako Cytomation M0760) prestained with the Zenon labeling kit (Invitrogen Molecular Probes).

We quantified the formation of vessels and vessel-like structures in Matrigel plugs by measuring lengths in H & E-stained samples in 0.3 mm² microscopic fields. First, we classified vasculatures into three categories as follows: (1) vessels with red blood cells (RBCs), (2) vasculatures without RBCs, and (3) lined cells without cavities. When PDGF-BB or the anti-PDGFRβ neutralizing antibody was mixed with VEGF-A and FGF-2, an additional category was included, i.e. (4) vessels with extravascular RBCs, which indicate hemorrhage. We drew lines manually along each category of vasculature using Adobe PhotoShop (Adobe systems, Inc). Then we calculated the total lengths of the lines in each category using ImageJ software (National Institute of Health). Microsoft Excel (Microsoft Corporation) software was used for statistical analysis. The data are average of nine samples from three independent Matrigel plugs.

Maintenance, differentiation, culture and magnetic-bead cell sorting of MGZ5 ES cells (gift from H. Niwa) using an anti-VEGFR2 monoclonal antibody (phycoerythrin-conjugated Avas12a1; eBioscience), were performed as described previously (Yamashita et al., 2000). We plated 2.5×10⁴ VEGFR2-positive (VEGFR2⁺) cells per well on 8-well CultureSlides (BD Falcon) for immunostaining or 2.0×10⁵ cells per well on 1-well CultureSlides for total RNA preparation. CultureSlides were coated with 30 μg/ml of collagen IV (Nitta Gelatin) prior to use. The cells were incubated in serum-free conditions for 1.5 days as described previously (Hirashima et al., 2003), in the presence of various ligands. FGF-2 was used with 1 μg/ml of heparin in all in vitro experiments in the present study. For staining of acetylated low-density lipoprotein (AcLDL) in endothelium, we used Alexa Fluor^® 594-conjugated AcLDL (Invitrogen Molecular Probes) following the manufacturer's protocol.

Total RNAs from ESC-derived VEGFR2⁺ cells treated with various ligands were extracted using the RNeasy Mini Kit (QIAGEN). RNAs were reverse transcribed by random hexamer priming using SuperScript III Reverse Transcriptase (Invitrogen). Quantitative RT-PCR analysis was performed using the GeneAmp 5700 system (Applied Bioscience) and SYBR^® Green (Applied Bioscience). All expression data were normalized to GAPDH. Primer sequences were: PDGF-B (NM_011057) 5′-agcagagcctgctgtaatcg-3′, 5′-ggcttctttcgcacaatctc-3′; PDGFRβ (NM_008809) 5′-tgccagttccaccttgaatgaa-3′, 5′-agttgtgcctcaggctctgctt-3′; PECAM1/CD31 (NM_008816) 5′-ccaaacagaaacccgtggagat-3′, 5′-gtaatggctgttggcttccaca-3′; SMAα/Acta2 (NM_007392) 5′-agcgtgagattgtccgtgacat-3′, 5′-gcgttcgtttccaatggtga-3′; and GAPDH (NM_008084) 5′-tgcagtggcaaagtggagatt-3′, 5′-tgccgttgaatttgccgt-3′.

Cells were fixed in 1:1 acetone-methanol solution and incubated with antibodies against PECAM1 (Mec13.3), SMA (1A4, Sigma-Aldrich Corporate A2547) and PDGFRβ (APB5, eBioscience). The cells were then incubated with secondary antibodies and Sytox Green nuclear counterstain (Invitrogen Molecular Probes). For staining of PDGF-B, we used a rabbit polyclonal antibody (Abcam ab15499) at 1:100 dilution with specimens fixed in 10% formalin.

Distances between the edge of each SMA⁺ mural cell and that of its nearest PECAM1⁺ endothelial sheet were measured using Adobe PhotoShop. We analyzed the data statistically using Microsoft Excel. Experiments were performed in triplicate, and the data for statistical analyses were taken from six fields of microscopic views from three independent sets of experiments.

A long-running video of the live cells on the CultureSlides (BD Falcon) was made using a Leica DM IRB microscope equipped with a hardware-controlled motor stage. The video images were analyzed using ImageJ software (National Institutes of Health, USA).

We used the Matrigel plug assay in vivo to investigate synergism by co-stimulation with VEGF-A and FGF-2 (referred to as the V/F treatment; Fig. 1A). We examined the effect of stimulation with VEGF-A alone (V, Fig. 1B) or FGF-2 alone (F, Fig. 1C), and compared them with the V/F treatment. Newly formed vessels and vessel-like structures in Matrigel plugs were quantified by measuring their lengths. We classified them into the following three categories: (1) red blood cell (RBC)-containing vessels (red bars in Fig. 1D; scattergram shown in Fig. 1E), (2) cavities surrounded by cells but without RBCs (light blue bars in Fig. 1D), and (3) cells arrayed in line but without cavities (blue bars in Fig. 1D). The function of the neovasculature was also investigated by a permeability assay using intravenously administrated FITC-conjugated dextran (M_r 2×10⁶).

The V/F treatment was most successful in inducing functional blood vessels, which contained RBCs and dextran in the vessel lumen with no leakage (Fig. 1A,D,E). Significantly more RBC-containing vessels were formed with the V/F treatment (**P<0.01) than with stimulation with VEGF-A or FGF-2 alone. Immunohistochemical analysis revealed that PECAM1⁺ endothelial cells in the V/F-induced blood vessels were surrounded by and attached tightly by SMA⁺ mural cells. Injected dextran was observed strictly within the vessel lumen (Fig. 1A, white arrowheads). PECAM1⁺ cells in newly formed blood vessels were also positive for an endothelial marker, CD34 (data not shown) (Simmons et al., 1992). We investigated perivascular SMA⁺ mural cells with other molecules reported to be markers of less mature pericytes, NG2 and desmin (Gerhardt and Betsholtz, 2003) (Fig. S1 in supplementary material). The SMA⁺ mural cells in Matrigel plugs were all positive for NG2, whereas only main cellular pericytic process, which firmly attached to endothelium, was positive for desmin as reported.

In VEGF-A-treated samples, neither RBC- nor dextran-containing vessels were observed (Fig. 1B,D, and Fig. S2 in supplementary material); only some linear cavities surrounded by cells were found. From electron microscopic examination cells did not appear to form continuous layers (data not shown). Although cells were all negative for PECAM1, major parts of them were either positive for CD34 or SMA. Moreover, some of these cells were also positive for F4/80, a murine macrophage marker. The findings suggest that there could be inflammatory hematopoietic cells in the Matrigel plug under VEGF-A stimulation (Fig. S2 in supplementary material). SMA⁺ cells in gels under VEGF-A stimulation were positive for NG2 but negative for desmin (Fig. S1 in supplementary material). These findings suggest that stimulation with VEGF-A alone was not sufficient for inducing mature blood vessel formation in the present experimental conditions.

FGF-2 stimulation induced significantly less RBC-containing blood vessels than did V/F treatment (yellow arrowheads in Fig. 1C and the red bar in Fig. 1D). A permeability assay revealed that the vessels treated with FGF-2 alone are leaky, and an immunohistochemical study showed less extensive SMA⁺ mural coverage of endothelium than in V/F treatment (Fig. 1C). SMA⁺ cells treated with FGF-2 were NG2⁺ and desmin^– as were the VEGF-A-treated cells. The total length of RBC-containing blood vessels induced by VEGF-A alone and FGF-2 alone was significantly less than that induced by V/F treatment (red bars in Fig. 1D). Thus the V/F stimulation had a synergistic effect on mature neovascular formation.

Since mural coverage of endothelial tubes is considered important for stabilization of newly formed blood vessels (Carmeliet, 2004), we focused on this cell-cell interaction in the following experiments.

To model the mural-endothelial communication induced by V/F treatment in vitro, we used embryonic stem cell (ESC)-derived VEGFR2⁺ cells cultured under serum-free condition on collagen IV-coated dishes (Yamashita et al., 2000; Watabe et al., 2003). PECAM1⁺ endothelial sheets, which were also positive for uptake of acetylated low-density lipoprotein (data not shown), appeared exclusively in the presence of VEGF-A, whereas SMA⁺ mural cells appeared in the presence of PDGF-BB as described previously (Yamashita et al., 2000) (Fig. 2B,D).

Under the V/F treatment both endothelial and mural lineages appeared (Fig. 2A,E). In addition, mural cells were recruited to the endothelial sheets (Fig. 2A, and corresponding long-running video-microscopic records available on-line: see Movie 1 in supplementary material), reflecting the findings in the in vivo Matrigel plug assay under V/F treatment. Video-microscopic observation for more than 12 hours revealed that, although the mural cells exhibited high motility, they still move toward the endothelium.

Upon treatment with FGF-2 alone, various lineages, including cells negative for PECAM1 and SMA, were observed (Fig. 2C). Notably, apparent endothelial-mural cell interaction was not observed in cells treated with FGF-2, although both endothelial and mural lineages were observed.

To examine the mechanism of mural-endothelial interaction during V/F treatment, we determined the levels of expression of PDGF-B and PDGFRβ, because these two molecules constitute one of the key signaling pathways in endothelial-mural communication, promoting mural cell migration to endothelium (Lindahl et al., 1997). PECAM1 and SMA, which are marker genes for endothelium and mural cells, respectively, were assessed as well (Fig. 3A). All data were normalized to GAPDH.

High expression of PECAM1 was observed in VEGF-A-treated and V/F-treated cells, reflecting the formation of endothelial sheets. SMA was mainly expressed in PDGF-BB-treated cells. These results are consistent with those obtained by immunohistochemical staining (Fig. 2). Expression of PDGF-B was high in VEGF-A-treated as well as in V/F-treated cells (colored bars), suggesting that expression of PDGF-B was induced by VEGF signaling. Immunostaining of V/F-treated cells (Fig. 3B) confirmed high expression of PDGF-BB in endothelial sheets compared to that in mural cells (red arrow).

Expression of PDGFRβ (which mediates PDGF-BB signaling) was induced in PDGF-BB-treated cells, a finding attributable to the expression of PDGFRβ in mural cells. In FGF-2-treated cells, levels of expression of PECAM1 and SMA correlated well with the results of immunostaining. Notably, expression of PDGFRβ was enhanced while that of SMA remained low in FGF-2-treated cells (blue bars, Fig. 3A; PDGF-BB-treated cells are represented by red bars, for comparison). These findings suggest the possibility that FGF-2 enhances PDGFRβ expression not only in mural cells but also in PECAM1/SMA double-negative cells. The level of expression of PDGFRβ was also higher in V/F-treated cells than in VEGF-A-treated cells. We then immunohistochemically determined the presence of PDGFRβ-expressing cells (Fig. 3C). With VEGF-A treatment, no cells expressed PDGFRβ. With V/F treatment, PDGFRβ was detected only in SMA⁺ cells (yellow arrows). With FGF-2 treatment, PDGFRβ was detected not only in SMA⁺ cells, but also in PECAM1/SMA double-negative cells (arrowheads). We thus concluded that FGF-2 enhances expression of PDGFRβ in cells of non-endothelial lineage, including the double-negative cells.

These findings explain well the mural cell recruitment observed with V/F treatment: VEGF-A makes the endothelial sheet more `attractive' to mural cells by enhancing PDGF-BB secretion, while FGF-2 makes mural cells more `sensitive' by enhancing mural PDGFRβ expression. Endothelium-derived PDGF-BB may be located in the vicinity of endothelial sheets by binding to the ECM and form a gradient of PDGF-BB, resulting in effective recruitment of mural cells. VEGF-A and FGF-2 thus appear to synergistically enhance endogenous PDGF-B–PDGFRβ signaling to promote migration of mural cells and their attachment to endothelial cells.

We added exogenous PDGF-BB (V/F/P treatment; Fig. 4A) or APB5, a neutralizing antibody against PDGFRβ (Sano et al., 2001) (V/F/Ab treatment; Fig. 4C) to V/F-treated cells (Fig. 4B) in order to examine the possibility that the enhancement of endogenous PDGF-B–PDGFRβ signaling is a key mechanism of V/F co-stimulation. Excessive exogenous PDGF-BB may distribute homogeneously and could disrupt mural cell migration by overwhelming the endogenous ECM-bound PDGF-BB gradient around the endothelial sheets.

Endothelial sheets and mural cells were significantly more distant from each other in both the V/F/P and V/F/Ab treatments than in the control V/F treatment (Fig. 4D). A total of 82% of mural cells were within 10 μm (one-tenth of the scale bars in Fig. 4A-C) of the endothelial sheets in the V/F treatment, whereas only 20% of the mural cells were within 10 μm in V/F/P and V/F/Ab treatments. Non-specific immunoglobulin had no effect when added to the V/F treatment (data not shown). With regard to cell populations, more SMA⁺ cells were induced by V/F/P (Fig. 4E), possibly because of mitogenic function of exogenous PDGF-BB. Both PECAM1⁺ cells and SMA⁺ cells appeared in all treatments tested, including the V/F/Ab treatment.

We could thus speculate that the mural cell recruitment induced by V/F treatment is mediated through enhancement of an endogenous gradient of periendothelial PDGF-BB, as well as mural PDGFRβ expression.

We further examined the same set of conditions in vivo in the Matrigel plug assay (Fig. 5). VEGF-A and FGF-2 (Fig. 5B) were mixed with PDGF-BB (V/F/P treatment; Fig. 5A) or APB5 (V/F/Ab treatment; Fig. 5C). Lengths of vessels and vessel-like structures in Matrigel plugs were quantified as in Fig. 1, but an additional category was included, i.e. vessels with extravascular RBCs, which indicate hemorrhage. Significantly more vessels with extracellular RBCs were observed in the V/F/P or V/F/Ab treatments than in the V/F treatment (orange arrowheads in Fig. 5A,C and orange bars in Fig. 5D; *P<0.05, **P<0.01). Significantly fewer RBC-containing vessels were formed in these conditions (Fig. 5D,E). The immunohistochemical study revealed defective attachment of SMA⁺ cells to PECAM1⁺ cells in the V/F/P and V/F/Ab treatments (Fig. 5A,C, white arrowheads). Some SMA⁺ cells did not attach to PECAM1⁺ cells under these conditions (yellow arrows), and leakage of dextran was observed (arrows). All perivascular SMA⁺ cells were NG2-positive, but desmin-negative, in V/F/P and V/F/Ab treatments (Fig. S3 in supplementary material). Electron microscopic study revealed that the cells in the V/F/P or V/F/Ab treatments failed to attach to each other and formed neither continuous layers nor tight junctions in such vessel-like structures. Non-specific IgG had no effect on V/F-induced blood vessel formation (data not shown). These findings suggest the importance of preservation of the endogenous periendothelial gradient of PDGF-BB, in ligand-stimulated neovascularization in vivo as well.

Combined stimulation by PDGF-BB with VEGF-A (V/P treatment) or FGF-2 (F/P treatment) was also compared with the V/F treatment in our experimental system (Fig. 6). Significantly fewer mature vessels were formed in V/P or F/P treatments than in the V/F treatment. Although SMA⁺ cells were observed near the PECAM1⁺ cells (especially in the F/P treatment), their attachments were defective. Leakage of dextran was also observed with F/P treatment. SMA⁺ cells were NG2-positive and desmin-negative as in other non-V/F treatments (Fig. S4 in supplementary material). Addition of exogenous PDGF-BB thus appeared to disrupt the perivascular gradient of PDGF-BB in vivo, leading to formation of less mature vessels.

Therapeutic angiogenesis, a strategy to induce neovascularization using angiogenic molecules, has been expected to be useful as an alternative method of treatment in patients with ischemic cardiovascular diseases. However, three large-scale clinical trials using recombinant VEGF-A or FGF-2 alone yielded results less significant than anticipated (Henry et al., 2003; Simons et al., 2002; Lederman et al., 2002). By contrast, combined stimulation with two angiogenic molecules from among VEGF-A, FGF-2 and PDGF-BB has been reported to be potent in inducing neovascular formation at least in experimental conditions. However, the mechanisms of synergistic effect are still not well enough understood, to be able to design optimal combinations of such agents. In the present study, we addressed the mechanism of the synergism of VEGF-A and FGF-2. We found that these factors play unique roles in synergistic enhancement of endogenous PDGF-B–PDGFRβ signaling, to promote mature blood vessel formation, in addition to their known mitogenic effects (summarized schematically in Fig. 7).

During development, mural cells are considered to be required for establishment of the integrity of new vessels (Sims, 1986). Mural cell deficiency induces aberrant microvascular formation (Betsholtz, 2004). One of the key signaling pathways in mural cell recruitment has been reported to be the PDGF-B–PDGFRβ pathway, as demonstrated in embryonic blood vessel formation of the brain capillaries and the kidney glomerular capillary tuft (Lindahl et al., 1997) and in postnatal renal and retinal function (Lindblom et al., 2003; Uemura et al., 2002). Recently, the PDGF-B–PDGFRβ pathway was also shown to support postnatal tumor vessel formation (Abramsson et al., 2003). We here demonstrate that the PDGF-B signaling pathway is also indispensable for adult neovascular formation induced by exogenous ligands, VEGF-A and FGF-2.

One of the major roles of FGF-2 in the co-stimulation system appears to be enhancement of expression of PDGFRβ in non-endothelial cells, in addition to stimulating the proliferation of both mural and endothelial cells. Enhancement of expression of PDGFRα and PDGFRβ by FGF-2 was recently reported in newly formed vessels in the rabbit ischemic hind limb model, as determined by in situ hybridization (Cao et al., 2003), although the types of cells expressing PDGFRs were not specified. Another report noted that expression of PDGFRα, but not of PDGFRβ, was enhanced in vascular smooth muscle cells by FGF-2 (Schöllmann et al., 1992). We found that FGF-2-induced enhancement of expression of PDGFRβ, but not of PDGFRα, is functionally important during blood vessel formation, as revealed by the effects of neutralizing antibody against PDGFRβ in the present study. In Matrigels treated with FGF-2, however, PECAM1⁺ cells and SMA⁺ cells appeared to communicate less with each other than when treated with V/EGF-A plus FGF-2. This decrease in intercellular communication might be explained by the low level of expression of PDGF-B with administration of FGF-2 alone, in contrast to the high level of expression of PDGF-B in the endothelium in the V/F treatment, as suggested by in vitro experiments (Fig. 3).

Induction of expression of PDGF-B by VEGF-A contributes to effective enhancement of mural cell recruitment. However, VEGF-A itself failed to induce mature vessels in Matrigel plugs (Fig. 1B) but it did induce sheets of PECAM1⁺ cells in the in vitro ESC-derived system on collagen IV-coated dishes (Fig. 2B). At present, the reasons for this discrepancy in findings are unclear. One possible explanation is that formation of endothelial sheets requires layers of collagen IV (Hangai et al., 2002; Carmeliet, 2004). Consistent with this possibility, VEGF-treated gels were negative for collagen IV staining, whereas V/F-treated gels were positive for collagen IV around mature vessels with surrounding mural cells (our unpublished observation).

Therapeutic angiogenic strategies using exogenous PDGF-BB as a component have emerged recently, in addition to the combination of VEGF-A and FGF-2. However, the endothelial–mural interaction did not occur efficiently when Matrigel plugs or ESC-derived VEGFR2⁺ cells were treated with a combination of three growth factors, VEGF-A, FGF-2 and PDGF-BB, although both endothelial and mural cells were observed in these systems (Fig. 4A, Fig. 5A). V/F/P treatment induced mural cells more effectively than the treatment without PDGF-BB, probably due to the mitogenic effect of exogenous PDGF-BB. By contrast, exogenous PDGF-BB appeared to have an inhibitory effect on local mural cell migration, possibly by disruption of the local gradient of PDGF-BB derived from endothelial cells by excessive exogenous PDGF-BB. It remains unclear, however, to what extent mural cell migration and neoangiogenesis could be affected by differences of ECM molecules bound by PDGF-BB molecules, or differences in local concentration of exogenous PDGF-BB.

The effect of exogenous PDGF-BB on disruption of neoangiogenesis was further confirmed in the Matrigel plug assays. In a recent study using a variety of combinations of growth factors, including VEGF-A and FGF-2 (V/F treatment), FGF-2 and PDGF-BB (F/P treatment), were found to yield the best results in inducing stable vessels in an in vivo corneal assay (Cao et al., 2003). The combination of VEGF-A and PDGF-BB (V/P treatment) was also reported to be effective in inducing new vessels in a polymer gel system for stepwise and gradual release of VEGF-A followed by PDGF-BB. By contrast, in our Matrigel plug assay, V/F treatment induced more mature vessels than did the V/P or F/P treatments (Fig. 6). We thus concluded that exogenous PDGF-BB, which may distribute homogenously and overwhelm the endogenous distribution of the molecule, could in fact impair blood vessel stabilization by mural cells. Preservation of the endothelial cell-derived local PDGF-BB gradient thus appears important for appropriate migration of mural cells towards endothelial cells.

Use of the combination of VEGF-A and FGF-2 thus has the outstanding advantage in therapeutic angiogenesis that the ligand and the receptor are properly induced in a cell-type-specific manner: PDGF-BB is induced in endothelial cells by VEGF-A, and PDGFRβ is induced in mural cells by FGF-2, resulting in enhancement of the cell-derived endogenous gradient of PDGF-BB and appropriate targeting of migrating mural cells. Understanding of the endogenous gradients of signaling molecules, as shown in this study, will enable rational design of effective strategies for induction of mature and functional blood vessels in vivo.

We thank Hisako Hirano for secretarial assistance, and Kyoko Kawasaki and Koichi Mishima for useful discussions. We are also grateful to Kimiko Yamamoto and Masahiro Shibata (Department of Biomedical Engineering, Graduate School of Medicine, University of Tokyo) for valuable suggestions on technical issues in the permeability assay. This research was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science, and Technology of Japan.