We used a 3D in-vitro model of angiogenesis to investigate the effects of different growth factors on vessel formation and stabilization in vitro. Vascular endothelial growth factor (VEGF) was the only factor that induced the formation, elongation and sprouting of capillary-like structures (CLS) by bovine retinal capillary endothelial cells (BREC), an effect that was dose-dependent and saturable. Basic fibroblast growth factor 2 (FGF2) enhanced capillary formation in the presence of VEGF, leading to a more complex network of CLS and a higher rate of BrdU incorporation than VEGF alone, indicating that whereas VEGF acts as a morphogen, FGF2 is primarily a mitogen. Addition of transforming growth factor β1 (TGFβ1) to the 3D assay along with VEGF and FGF2, reduced tube formation in a dose-dependent manner. When added at the time of cell plating TGFβ1 completely suppressed formation of VEGF/FGF2-stimulated CLS. Angiopoietin 1 (Ang1) prevented regression of the TGFβ1-induced CLS, an effect that was blocked by angiopoietin 2 (Ang2), but required the continuous presence of VEGF.
Vessel assembly, patterning and maintenance is a complex and highly regulated process. Targeted gene disruption and overexpression studies have revealed critical, rate-limiting roles for a number of molecules, including growth factors, receptors and transcription factors. Embryonic or perinatal lethality results from deficiency of vascular endothelial growth factor (VEGF), platelet-derived growth factor beta (PDGFB), transforming growth factor β1(TGFβ1) and the angiopoietins, as well as their receptors, and analyses of vessel structure in animals deficient in these proteins have provided some insights into their roles. Mice deficient in VEGF (Carmeliet et al., 1996; Ferrara et al., 1996) or its receptors VEGFR1 (also known as Flt1) (Fong et al., 1995), VEGFR2 (also known as Flk1 and Kdr) (Shalaby et al., 1995) and neuropilin (Kawasaki et al., 1999), are embryogenic lethal due to disruption of endothelial cell (EC) differentiation or vessel assembly. Absence of PDGF-B (Leveen et al., 1994) or its receptor PDGF-β-receptor (Soriano, 1994) results in a hemorrhaging phenotype and a reduced number of perivascular cells. Mice deficient in the endothelial-specific receptor Tie2 or its ligand Ang1 are embryonic lethal as a result of defects in vessel remodeling (Sato et al., 1995; Suri et al., 1996), similar to mice over-expressing Ang2 (Maisonpierre et al., 1997). In addition, absence of TGFβ1 or its receptors TGFβR2 or endoglin, leads to defects in vascular wall structures (Dickson et al., 1995; Li et al., 1999; Oshima et al., 1996). A number of studies using cultured ECs or angiogenesis models such as the aortic ring assay (Nicosia et al., 1997) and the metatarsal assay (Leenders et al., 2002) have attempted to elucidate the relative contribution of the individual factors. Although some insight has been provided by these studies, they are limited by the fact that they, for the most part, examine a single growth factor at a time. Thus, little is known about the manner in which the growth factors interact to govern vessel formation.
An additional level of vessel regulation appears to lie with the heterotypic interactions between the endothelium and pericytes (reviewed by Armulik et al., 2005). A growing body of in vivo observations clearly demonstrates that the association of pericytes with forming vessels marks the maturation of the vessel (Benjamin et al., 1998; Darland and D'Amore, 2001). Analysis of experimental tumor models reveals that vessels that lack pericytes are targets of regression whereas the more mature microvessels, characterized by pericyte association, do not regress (Benjamin et al., 1999). In fact, there is increasing realization that the efficacy of anti-angiogenic therapy may be limited by the maturation state of the vasculature (Bergers et al., 2003; Gee et al., 2003). However, the molecular mechanism that underlies the pericyte-mediated vessel stabilization is not known and the complexity of the in situ environment makes it difficult to elucidate the basis of pericyte stabilization.
To address this issue and to gain insight into how the various polypeptide factors work together, we have developed a tissue culture model in which ECs cultured alone are induced to form CLS by the addition of VEGF. We have used this system to examine the combinatorial effects of the relevant growth factors, while at the same time eliminating the complicating issues of heterotypic interactions that would result from co-cultures. Thus, we have been able to assess the effect of `context' on the actions of the growth factors.
We used a 3D in-vitro model of angiogenesis to investigate the effects of different growth factors on vessel formation and stabilization. The functions and sequential interactions of the factors were examined by evaluating formation of CLS, proliferation and apoptosis.
VEGF induces capillary formation
BREC grown in 2-D cultures were trypsinized and embedded in collagen gels. During gelation the cells settled out and attached to the underlying collagen layer. Cells plated into the collagen matrix in the absence of any growth factors remained round (Fig. 1A). Among all the growth factors tested (VEGF, FGF2, TGFβ1, Ang1, and Ang2), VEGF was the only factor that induced the formation, elongation and sprouting of CLS by the capillary ECs. The formation of the CLS was independent of plating cell density. The effect of VEGF on tube formation was dose dependent and saturable. In the presence of 1.5% serum, CLS first appeared at a concentration of 10 ng/ml VEGF and the maximal length of CLS was observed at 25-50 ng/ml. Initiation of CLS was also serum dependent; formation of CLS was seen in concentrations as low as 0.5% serum (5 ng/ml VEGF) and was completely suppressed at 10% serum (data not shown), suggesting the presence of an inhibitory factor in serum.
A serum concentration of 1.5% was used in all the experiments reported below as it supported significant CLS and provided sufficient nutrients so that the medium could be replaced after 2 days without any additional serum. VEGF was used at 25 ng/ml in all the reported studies. Addition of VEGF every other day was necessary to maintain the network of CLS. The formation of CLS by large vessel (bovine aortic) ECs occurred in the absence of added VEGF (not shown), probably because they produce higher levels of endogenous growth factors in vitro than capillary ECs (P.A.D., unpublished observations).
FGF2, a potent endothelial mitogen, enhanced capillary formation by BREC in the presence of VEGF (25 ng/ml). The optimal concentration of FGF2 varied as a function of the VEGF concentration. Therefore, we used 2.5 ng/ml FGF2 and 25 ng/ml VEGF in our system. The additive effect of VEGF and FGF2 was evident after 4 days (Fig. 1A,B). VEGF combined with FGF2 led to a more complex network of CLS and a higher rate of BrdU incorporation than VEGF alone (141 mm versus 104 mm total length of CLS; 1247 vs 610 BrdU-positive cells; VEGF plus FGF2 vs VEGF alone; Fig. 1B). When formed in the presence of VEGF plus FGF2, CLS were composed of numerous endothelial cells, as visualized by nuclear staining (Fig. 1C). Many nuclei, visible by DAPI staining, were identified by TUNEL staining as apoptotic figures. Because of the high density of cells in the CLS and the 3D nature of the structures, it was not technically possible to quantify the apoptotic nuclei. Dividing cells, detected by BrdU incorporation, were evenly distributed (Fig. 1C). These observations indicate that whereas VEGF has morphogenic activity, FGF2 acts primarily as a mitogen.
TGFβ1 inhibits capillary formation and induces apoptosis
TGFβ1 has been reported to both stimulate and inhibit angiogenesis (Pepper et al., 1993). When added to capillary ECs in the collagen gel along with VEGF plus FGF2, TGFβ1 reduced tube formation in a dose-dependent manner with effects observed at concentrations as low as 0.25 ng/ml. Addition of TGFβ1 (1 ng/ml) at the time the cells were plated completely suppressed formation of CLS by VEGF (25 ng/ml) plus FGF2 (2.5 ng/ml). If CLS were permitted to form for 2 days and then TGFβ1 (along with fresh VEGF plus FGF2) was added, the CLS completely regressed within the following 2 days (Fig. 2A,B; compare Fig. 2A and Fig. 1A). BrdU incorporation decreased by 30% within the first 13 hours and nearly all of the cells underwent apoptosis (Fig. 2B,C). The inhibitory effect of TGFβ1 was serum-dependent, with increasing inhibition noted at higher serum levels (data not shown).
Ang1 stabilizes and Ang2 destabilizes capillaries in the presence of TGFβ1
Ang1, acting via the Tie2 receptor on ECs, is essential for the remodeling and stabilization of the embryonic vasculature (Uemura et al., 2002). Addition of Ang1 prevented the regression of TGFβ1-induced CLS (Fig. 3A). The effect was dose dependent; significant effects were detected at 100 ng/ml Ang1 and maximal effects were seen at 500-1000 ng/ml. At 500 ng/ml Ang1 rescued 60% of the pre-established endothelial network from TGFβ1-induced regression (Fig. 3A,B). Ang1 did not have a significant effect on EC proliferation. Increased concentrations of VEGF up to 100 ng/ml were not sufficient to prevent the TGFβ1-induced regression. Ang2 at 500 ng/ml had no effect on TGFβ1-induced regression (Fig. 3A,B). Consistent with the concept of Ang2 as an antagonist, simultaneous addition of Ang1 and Ang2 significantly reduced the Ang1-mediated `rescue'. In the absence of TGFβ1, neither Ang1 nor Ang2 had a detectable effect on the formation of CLS or EC proliferation.
VEGF is necessary for stability of CLS in the presence of TGFβ1
A network of CLS was established for 2 days in the presence of VEGF plus FGF2. With the subsequent addition of TGFβ1 and Ang1, the network remained stable, as described above. To determine if continuous VEGF was needed for vessel maintenance, the action of VEGF was blocked by the addition of a VEGF-Trap (1 μg/ml) after tubes had been forming in the presence of FGF2 plus VEGF for 48 hours. The neutralization of VEGF led to a dramatic increase in EC death (Fig. 4A). Although EC proliferation was not affected, the network completely regressed (Fig. 4B). These observations are consistent with the concept that the major action of VEGF is not as a mitogen but rather as an EC survival factor.
Ang1 stabilizes capillaries in the absence of VEGF
Reports using both in-vivo and in-vitro models have suggested that Ang1 can substitute for the stabilizing effect of VEGF. However, our observations (above) indicate that in the presence of TGFβ1, Ang1 is not sufficient to stabilize CLS in the absence of VEGF. Rather, the continued presence of VEGF is necessary for Ang1 to counter the pro-apoptotic effects of TGFβ1. In the absence of TGFβ1, Ang1 was indeed able to rescue capillary ECs from VEGF withdrawal (Fig. 5A). Furthermore, in the absence of TGFβ1, Ang2 did not antagonize the Ang1 rescue effect. As noted above, proliferation was not significantly altered by the angiopoietins (Fig. 5B).
We were also interested in determining the importance of FGF2 in this system. Thus, parallel experiments were conducted in the absence of FGF2. Addition of VEGF-Trap to cells incubated only in the presence of VEGF for 2 days led to dissolution of the CLS network, and this effect was only partially rescued by Ang1 (Fig. 5C,D). In the presence of VEGF-Trap, re-addition of FGF2 had no significant effect on CLS length and proliferation (Fig. 5D). From these results we conclude that FGF2 is additive not only with VEGF during formation of CLS but also with Ang1 during stabilization of these structures.
Vessel assembly, patterning and maintenance are complex and highly regulated processes. Gene knockout and overexpression studies have revealed critical, limiting roles for a number of growth factors and receptors, including VEGF, PDGF, TGFβ1, Ang1 and their receptors. However, little is understood about the manner in which various cells and their secreted factors interact to govern vessel formation. To gain further insight into mechanisms that underlie the control of vessel formation and stability, we have studied the combinatorial effects of a number of growth factors in an in-vitro model of vessel assembly. To avoid the complicating issues of heterotypic interactions between ECs and pericytes (or smooth muscle cells), and to isolate the effects to the factors, we used pure primary cultures of retinal capillary ECs.
As expected, VEGF was necessary for assembly of ECs into CLS, an observation consistent with other reports of cells in culture (Goto et al., 1993) and with the lack of vessels observed in VEGF-deficient mice (Carmeliet et al., 1996; Ferrara et al., 1996). Consistent also with the observed modest mitogenic effect of VEGF in vitro (Yoshida et al., 1996), VEGF led to only a small increase in the proportion of BrdU-positive cells in CLS whereas FGF2 combined with VEGF caused a threefold increase in BrdU incorporation over controls and this was reflected in measurably longer CLS.
The role of FGF2 in vivo is not clear. Targeted disruption of FGF2 does not cause a dramatic vascular phenotype, but its absence causes a modest defect in wound healing (Ortega et al., 1998) and a decrease in vascular tone (Zhou et al., 1998). Thus, FGF2 may act in vivo to titrate vessel density as opposed to directly controlling assembly. The effects of FGF2 were additive not only with VEGF but also with Ang1. This effect may be due to its mitogenic effect or to its general effects on survival (Garcia et al., 2005; Miyamoto et al., 1998).
Mice deficient in TGFβ1 or its receptor TGFβR2 display approximately 50% embryonic lethality, with defective endothelial-mesenchymal interactions (Dickson et al., 1995; Oshima et al., 1996). Mice lacking endoglin exhibit a 70% reduction in smooth muscle cells (Li et al., 1999) and although mice heterozygous for endoglin display normal vascular development, ECs isolated from these mice have reduced ability to form tube-like structures in vitro (Jerkic et al., 2006). TGFβ1 has been shown to inhibit endothelial growth and to induce apoptosis (Ramsauer et al., 2002; Yan and Sage, 1998). These observations in monolayer cultures were reproduced in our 3D model where application of TGFβ1 either blocked initial capillary formation or caused a collapse of a pre-established capillary network.
However, these in vitro observations do not elucidate the action of TGFβ1 in the developing vessel where we suspect the TGFβ1 is supplied as a result of EC-mesenchymal cell (pericyte and/or smooth muscle cell) interactions. Recently, a model has been proposed in which TGFβ, by binding to TGFβR2, can activate two distinct type I receptors in ECs, i.e. the EC-restricted activin receptor-like kinase-1 (ALK1) and the broadly expressed ALK5, which have opposite effects on EC behavior (reviewed by Bobik, 2006; Lebrin et al., 2005). ALK1 via Smad1/5 transcription factors stimulates EC proliferation and migration, whereas ALK5 via Smad2/3 inhibits these processes. These receptor-regulated Smads interact with the common mediator Smad, Smad4, and translocate to the nucleus. It is suggested that TGFβ regulates the activation of the endothelium via a fine balance between ALK1 and ALK5 signaling (Goumans et al., 2002).
Using co-culture models of ECs and pericyte/smooth muscle cell precursors, we and others have shown that TGFβ1 that is locally activated upon contact between the cells inhibits endothelial proliferation (Antonelli-Orlidge et al., 1989) and migration (Sato and Rifkin, 1989), induces mesenchymal differentiation into pericytes (Hirschi et al., 1998) and stimulates pericyte production of VEGF (Darland et al., 2003). In addition, it has been reported that pericytes/smooth muscle cells are a local source of Ang1 (Sundberg et al., 2002). Our observations of the combinatorial effects of growth factors on ECs alone parallel these reports. Addition of TGFβ1 to the CLS caused regression of the tubes. By contrast, similar to co-cultures where contact leads to the generation of activated TGFβ1 and the pericytes secrete VEGF, FGF2 and Ang1 (Davis et al., 1996), addition of Ang1 can rescue CLS from the pro-apoptotic effects of TGFβ1.
An important remaining question is the mechanism of vessel stabilization. Ang1 has been shown to be involved in vascular remodeling and maturation (Suri et al., 1996), including vessel stabilization (Uemura et al., 2002) and control of permeability (Thurston, 2002). Consistent with these in vivo observations, we have shown that Ang1 rescues ECs from TGFβ1-induced apoptosis, an action likely mediated by the phosphatidylinositol 3-kinase (PI 3-kinase)-Akt pathway (Kim et al., 2002). Furthermore, only in the presence of TGFβ1 can Ang2 overcome Ang1 action. Therefore, the functions of Ang2 appear to be more complex and Ang2 probably acts in a context-dependent manner as an antagonist of Tie2 signaling to loosen up the vessel structure. Ang2, the natural antagonist of Ang1 (Maisonpierre et al., 1997), is produced by the endothelium (Fiedler et al., 2004). The expression of Ang2 by ECs indicates that it can act in an autocrine fashion to control endothelial quiescence and responsiveness. Studies of cultured cells indicate that VEGF induces Ang2 synthesis in ECs (Oh et al., 1999), thus leading to vessel destabilization that precedes new vessel growth.
Our results also address the role of VEGF in vessel stabilization and indicate that in the presence of TGFβ1, both VEGF and Ang1 are necessary for survival. These observations are consistent with our recent findings that virtually all adult vascularized tissues not only express VEGF but also display constitutive VEGFR2 phosphorylation (Maharaj et al., 2006). We conclude that vessels, which are induced to form by VEGF and/or FGF2, are subsequently stabilized by the coordinated action of TGFβ1 and Ang1, and the continued action of VEGF. New vessel growth or vessel regression results from a disruption in this balance. Excess VEGF can be provided by tumor cells, ischemic tissues or inflammatory cells, to name a few sources. The in vivo correlate of this phenomenon has been reported in a mouse model where dormant metastases were found to have high BrdU incorporation by both tumor cells and endothelial cells, and high apoptosis, whereas growing tumors displayed equally high BrdU and low apoptosis (Holmgren et al., 1995). The authors speculate that the difference in apoptosis was due to absence of trophic factors such as VEGF and FGF2.
The local production of activated TGFβ1, which we postulate to be generated locally by EC-pericyte interactions, can be interrupted by the loss of pericytes, such as in diabetic retinopathy (Hammes et al., 2002) or when EC-pericyte interactions are abnormal, such as in tumor vessels (Morikawa et al., 2002). Additionally, production of Ang2 by the endothelium (in the presence of TGFβ1) will act to antagonize the effects of Ang1, thus leading to vessel destabilization.
Thus, vessel development and maturation is a locally controlled process that involves a number of growth factors acting in an autocrine and paracrine fashion. The triad of pro-angiogenic growth factors (VEGF, FGF2 and Ang1) interacts to influence the proliferation, apoptosis, and differentiation of microvascular endothelial cells in respect to their anti-angiogenic counterparts (TGFβ1 and Ang2). Although solitaire functions of angiogens can be studied using isolated single factors, their full potential is displayed in a well-balanced combination, and withdrawal or addition of only one factor can cause deleterious effects.
Materials and Methods
Recombinant human VEGF 165, FGF2, TGFβ1, and Ang2 were obtained from R&D Systems (Minneapolis, MN). Bow Ang1-Fc and VEGF-Trap (VEGFR1R2-Fc) were kindly provided by Regeneron Pharmaceuticals, Inc. (Tarrytown, NY). Other sources of material are given below.
Isolation and cultivation of bovine retinal endothelial cells (BREC)
Bovine eyes were obtained from a local slaughterhouse. The retinas were removed under sterile conditions and homogenized using a handheld pistil. The homogenized retinas were filtered through an 88-μm mesh and the capillaries trapped on top of the mesh were collected. This mix was subjected twice to a digestion with 0.5% collagenase Type II (Gibco, Carlsbad, CA) for 45 minutes at 37°C and the resulting cells [bovine retinal endothelial cells (BREC)] were plated on plastic dishes coated with fibronectin (Sigma, St Louis, MO). BREC were maintained in culture in endothelial basal medium (EBM; Clonetics, Walkersville, MD) supplemented with 10% horse serum (Sigma, St Louis, MO) and bovine brain extract with heparin (Clonetics, Walkersville, MD) and equilibrated with 95% air and 5% CO2 at 37°C. Medium was changed every other day and cells were passed when reaching confluence. Passage five cells were used in these experiments. Homogeneity was assessed by the expression of von Willebrand factor VIII-related antigen. Absence of pericytes or smooth muscle cells was confirmed by absence of α-smooth muscle actin.
In vitro angiogenesis assay
This assay was performed in 48-well plates. Collagen gels were prepared by mixing eight parts Vitrogen (Cohesion, Palo Alto, CA) with one part 10× Dulbecco's modified Eagle's medium and one part 0.1 M NaOH. Aliquots (175 μl) of collagen were allowed to polymerize for 45 minutes at 37°C (5% CO2). Later, an additional layer of 75-μl collagen containing BREC (1×106 cells per ml) was added and allowed to solidify for another 45 minutes. An equal amount of 250 μl EBM supplemented with 1.5% horse serum was added on top of the collagen sandwich. Growth factors and soluble receptors were included in the medium as desired at initial plating. After 2 days the medium was replaced with serum-free EBM and growth factors. For proliferation studies, BrdU (5-bromo-2′-deoxyuridine; Sigma, St Louis, MO) at a final concentration of 0.2 mM was added 1 hour after the addition of growth factors and left for 12 hours. Proliferation was therefore assessed 13 hours after addition of growth factors, on the third day.
Immunohistochemistry of cells embedded in collagen gels
For staining, collagen gels containing cells were fixed with 4% formaldehyde solution for 16 hours at 4°C. After washing with phosphate-buffered saline (PBS) the specimens were permeabilized with 1% Triton X-100 in PBS overnight at 4°C and processed as below.
To identify apoptotic cells, TUNEL (TdT-mediated dUTP nick end labeling) technique was performed by modifying the DeadEnd™ Colorimetric TUNEL System (Promega, Madison, WI). Specimens were pre-equilibrated with equilibration buffer for 1 hour. Subsequently gels were immersed in TUNEL reaction mix containing biotinylated nucleotides and TdT (terminal deoxynucleotidyl transferase) enzyme and incubated at 37°C. After 4 hours the reaction was terminated with 2× SSC solution for 30 minutes at room temperature. After several washes with PBS for 24 hours, non-specific binding sites were saturated by incubation of specimens in PBS containing 0.1% BSA and 10% goat serum for 12 hours. Biotinylated nucleotides were detected by incubation with Streptavidin Alexa Fluor® 568 conjugate (1:100; Molecular Probes, Eugene, OR) in the dark for 24 hours at 4°C.
Cell nuclei were visualized by staining with DAPI (4′,6-diamidino-2-phenylindol-dihydrochloride, 0.5 μg/ml in PBS; Sigma, St Louis, MO) for 12 hours. After thorough rinsing with PBS for 2 days, specimens were visualized using immunofluorescence microscopy.
Cell proliferation was measured by BrdU incorporation and subsequent immunostaining. DNA was denatured with 1 M HCl for 1 hour at 42°C. After neutralization with 0.2 mM Borax (sodium tetraborate) solution for 15 minutes, cells were stained with an anti-BrdU Alexa Fluor® 488 conjugate (1:10; Molecular Probes, Eugene, OR) for 24 hours at 4°C.
Fixed and stained specimens were viewed with an Eclipse TE2000-S microscope (Nikon, Melville, NY). Images were captured with a SPOT camera (Diagnostic Instruments, Inc., Sterling Heights, MI). For each experimental condition, representative 2× low-magnification fields were photographed. Each field corresponded to an area of 25 mm2. Quantification was performed with the IPLab software for Macintosh Version 3.6 (Scanalytics, Fairfax, VA). To determine total length of CLS, capillaries were manually traced. Nuclear staining of proliferating cells was recognized automatically and confirmed by eye.
Experiments were performed three times in triplicate. Data are expressed as mean ± standard deviation (s.d.). Statistical significance was tested using ANOVA and set at P<0.05.
The authors acknowledge the helpful discussions with members of the D'Amore laboratory particularly Magali Saint-Geniez and Eric Finkelstein. The work was supported by the Juvenile Diabetes Research Foundation Center Grant, NIH EY05318 and NIH EY015435. P. D'Amore was a Senior Scholar of Research to Prevent Blindness at the time the work was conducted.