The process of postnatal angiogenesis plays a crucial role in pathogenesis of numerous diseases, including but not limited to tumor growth/metastasis, diabetic retinopathy, and in tissue remodeling upon injury. However, the molecular events underlying this complex process are not well understood and numerous issues remain controversial, including the regulatory function of integrin receptors. To analyze the role of integrin phosphorylation and signaling in angiogenesis, we generated knock-in mice that express a mutant β3 integrin unable to undergo tyrosine phosphorylation. Two distinct models of pathological angiogenesis revealed that neovascularization is impaired in mutant β3 knock-in mice. In an ex vivo angiogenesis assay, mutant β3 knock-in endothelial cells did not form complete capillaries in response to vascular endothelial growth factor (VEGF) stimulation. At the cellular level, defective tyrosine phosphorylation in mutant β3 knock-in cells resulted in impaired adhesion, spreading, and migration of endothelial cells. At the molecular level, VEGF stimulated complex formation between VEGF receptor-2 and β3 integrin in wild-type but not in mutant β3 knock-in endothelial cells. Moreover, phosphorylation of VEGF receptor-2 was significantly reduced in cells expressing mutant β3 compared to wild type, leading to impaired integrin activation in these cells. These findings provide novel mechanistic insights into the role of integrin–VEGF axis in pathological angiogenesis.
The process of angiogenesis involves coordinated endothelial cell (EC) proliferation, invasion, migration, and tube formation. This process is induced by vascular growth factors in coordination with extracellular matrix–interacting molecules such as integrins (1). Among integrins, αvβ3 heterodimer is known to be up-regulated on proliferating endothelial cells during angiogenesis and vascular remodeling. The disruption of αvβ3 integrin ligation by either blocking antibodies or cyclic peptide antagonists prevented blood vessel formation in mouse retina, rabbit cornea, chick chorioallantoic membrane, and human skin transplanted onto athymic mice (2–6). Histological examination of tumor tissue from mice treated with the αvβ3 blockers revealed reduction in the tumor cell viability and in the vascular density (7). These findings suggested that integrin αvβ3 provides cellular signals that facilitate EC proliferation and migration during angiogenesis.
However, studies using β3- and β5-null mice demonstrated enhanced tumor growth, tumor angiogenesis, and vascular endothelial growth factor (VEGF)-A–induced vascular permeability caused by elevated levels of VEGF receptor (VEGFR)-2 on ECs (8, 9). Therefore, it has been concluded that αvβ3 and αvβ5 integrins function as negative regulators of angiogenesis by restricting the VEGFR-2 expression (10). Surprisingly, neither αvβ3 blocking antibody nor cyclic peptide antagonists induced VEGFR-2 expression in any of the model systems described in previous studies. This raises the possibility that an enhanced expression of VEGFR-2 in integrin β3/β5–deficient mice could be a result of over-compensation because both integrins are pivotal for early embryonic vasculogenesis and angiogenesis (11). Another hypothesis suggested by authors of previous studies is that unligated integrins can act as negative regulators of cell survival by initiating a process referred to as “integrin-mediated cell death.” Therefore, a genetic ablation of a proapoptotic stimulus (i.e., unligated αvβ3/αvβ5) would lead to increased endothelial cell survival in vivo, contributing to increased blood vessel density (12). Overall, because of the great deal of controversy, the precise role of αvβ3 integrin in endothelial cell biology and angiogenesis remains unclear.
One of the characteristic features of integrins is the ability to transduce signals through the cellular membrane in both directions. Outside-in signaling informs the cell about the extracellular matrix environment. Inside-out signaling, known as integrin activation, is the process stimulated by agonist (i.e., thrombin or adenosine diphosphate for platelets and growth factor for EC) and results in changes in integrin functional activity (13). The cytoplasmic domains of integrins play a vital role in these bidirectional signaling processes. The cytoplasmic domain of the β3 subunit of the vitronectin receptor (αvβ3) undergoes phosphorylation in response to cell adhesion to an immobilized ligand (14). Phosphorylated β3 can recruit signaling intermediates, Shc and GRB2, suggesting that the phosphorylation of the β3 cytoplasmic domain may induce Ras activation in several cell types (15).
To elucidate the role of β3 integrin and its complexes, we characterized mice that express a mutant form of β3 integrin, but not the WT form (knock-in mice). In the mutant, the two tyrosine residues known to be involved in integrin signaling, Tyr747 and Tyr759, were substituted to phenylalanines. The mutant is unable to undergo phosphorylation of cytoplasmic domain, resulting in deficient integrin signaling. In these mice, named DiYF, the mutant β3 integrin is physically present on all cell types, which should prevent any developmental compensatory changes. Recent studies using DiYF mice demonstrated that mutation of both cytoplasmic tyrosine residues of the β3 cytoplasmic domain leads to the defective function of αIIbβ3 on platelets, unstable platelet aggregation, and clot retraction in vitro and to an enhanced tendency to rebleed in vivo (16, 17). Thus, although it is clear that β3 integrin cytoplasmic tyrosine motifs play an important role in integrin signaling in platelets, its function in endothelial and other specialized cells remains unknown.
In the specialized cells, integrins form functional complexes with other receptors on the cell surface. Most often, integrins are associated with tyrosine kinase receptors and these two types of receptors influence each others activity (18, 19). On ECs, αvβ3 integrin is intimately linked with the VEGFR-2, because VEGF is able to induce αvβ3 integrin expression and activation (20). At the same time, αvβ3 is able to enhance the phosphorylation of VEGFR-2 in response to VEGF and subsequent activation of SAPK2/p38 and FAK.
We sought to directly establish the role of β3 integrin tyrosine phosphorylation and integrin signaling in VEGF-stimulated responses of EC and in pathological angiogenesis in vivo. To this end, we combined in vivo and ex vivo angiogenesis assays with an extensive characterization of ECs derived from both WT and DiYF mice to perform a complete analysis of the role of integrin β3 and its signaling in functional responses to VEGF.
In ECs, tyrosine phosphorylation of β3 cytoplasmic domain occurs upon adhesion to integrin ligands
Previous studies using cell lines and platelets demonstrated that phosphorylation of β3 integrin cytoplasmic tyrosines is involved in outside-in integrin signaling. Therefore, we assessed whether EC adhesion to extracellular matrix proteins could induce tyrosine phosphorylation of β3 integrin. We isolated ECs from the lungs of WT and DiYF mice, plated them on various substrates, and assessed the phosphorylation of β3 integrin on Tyr747 and Tyr459 (Fig. 1, A and B).
Very low levels of phosphorylation of WT β3 integrin were observed when cells were kept in suspension or plated on poly-l-lysine. Extracellular matrix proteins that serve as ligands for αvβ3 such as vitronectin, fibronectin, fibrinogen, and gelatin strongly stimulated Tyr759 and Tyr747 phosphorylation of WT ECs. Laminin and collagen, which are recognized primarily by integrins other than αvβ3, induced lower levels of tyrosine phosphorylation. Sodium pervanadate, known to block phosphatase activity, was used as a positive control in this experiment. As anticipated, no tyrosine phosphorylation of β3 integrin was observed in DiYF EC under any conditions.
VEGF treatment of ECs induces tyrosine phosphorylation of β3 integrin
The role of the β3 subunit phosphorylation in extracellular matrix recognition was previously assessed in CHO cells and K562 cells. These systems do not provide insight into the roles of β3 integrin in angiogenesis and in the regulation of VEGF or FGF-induced EC responses. These processes can be only assessed in ECs expressing appropriate receptors and signaling intermediates. We sought to determine whether VEGF treatment affected the phosphorylation status of β3 integrin. Treatment of ECs with VEGF-A165 induced both Tyr747 and Tyr759 phosphorylation in a time-dependent manner (Fig. 1 C). Both time curves of β3 tyrosine phosphorylation followed a bell-shaped pattern, which is typical for growth factor–induced responses. Thus, not only integrin engagement, but also VEGF treatment stimulated tyrosine phosphorylation of β3 subunit in ECs, suggesting a possible regulatory role for this process in VEGF-induced angiogenesis.
Pathological angiogenesis is impaired in DiYF mice
We next evaluated whether β3 integrin cytoplasmic tyrosine motifs are crucial for a complete angiogenic response to VEGF in vivo. We implanted VEGF-A–containing Matrigel subcutaneously into WT and DiYF mice and assessed the angiogenic response based on the amount of hemoglobin extracted from Matrigel. As shown in Fig. 2 A, the hemoglobin concentration was at least fivefold lower in Matrigel plugs isolated from DiYF mice compared with WT counterparts. The vascular density in Matrigel implants was assessed by von Willebrand factor (vWF) staining. Only 50% of Matrigel plugs from DiYF mice had a distinguishable vasculature and exhibited staining for vWF (Fig. 2 B, top).
The vascular density was 4.2-fold lower in DiYF mice than in WT controls (P = 0.01) (Fig. 2 B, bottom). Thus, VEGF-induced angiogenesis was significantly impaired in DiYF mice.
Further, we assessed tumor-induced pathological angiogenesis in DiYF and WT mice. To this end, mouse melanoma cells (B16F10) were implanted subcutaneously into mice, and after 10 d tumors were excised. The average weight of tumors formed in DiYF mice was at least twofold lower than average tumor weight in WT mice (0.14 g vs. 0.29 g) (Fig. 2 C). Tissue section analysis revealed the presence of well-developed blood vessels that stained positively for CD31, and laminin in tumors from WT mice. In contrast, in tumors formed in DiYF mice, blood vessels were sparse and thin walled based on laminin staining (Fig. 2 E, top). The vascular density in tumors grown in DiYF mice was sixfold lower than that in WT mice (P = 0.009) (Fig. 2 E, bottom). Similar defects in tumor growth and neovascularization were observed using alternative model of tumor growth in vivo. Mouse prostate cancer cells (RM-1) were implanted into WT and DiYF mice, and tumor was excised after 10 d. Tumors formed in DiYF mice were smaller than in WT mice (1.17 g vs. 0.8 g) (Fig. 2 D). Tissue section analysis indicated the presence of substantially larger number of vWF-positive blood vessels in tumor developed in WT mice compared with DiYF mice (Fig. 2 F). Collectively, our in vivo studies demonstrate that tumor angiogenesis is impaired in DiYF mice, resulting in reduced tumor growth. Thus, it appears that β3 integrin cytoplasmic tyrosine motifs play crucial role in the regulation of pathological angiogenesis.
Knowing that monocytes/macrophages use β3 integrin for their arrest and extravasation at sites of inflammation, we assessed whether DiYF mutation affects macrophage recruitment into tumors using anti-F4/80 antibody staining. As shown in Fig. 2 G, considerable amounts of macrophages were found to be present in tumors implanted in both WT and DiYF mice. No substantial reductions in macrophage recruitment were found in DiYF mice. Similar results were observed in Matrigel model of VEGF-induced angiogenesis (Fig. 2 H). Thus, DiYF mutations within β3 integrin do not impair the process of macrophage recruitment in tumors.
Normal embryogenesis and adult vasculature development in DiYF mice
Homozygous WT and DiYF mice were bred to obtain homozygous littermates. The average number of embryos per litter was the same. Size and weight of pups were not substantially different between WT and DiYF mice. In contrast to β3 integrin knockout mice, no embryonic mortality caused by placental defects was observed in DiYF mice. DiYF mutation did not affected survival of embryos at early stages of development. DiYF embryos did not exhibit any broad vascular defects (Fig. 3 A).
Prenatal DiYF mutant mothers did not suffer from intrauterine hemorrhage and placental defects (Fig. 3 B) as observed in β3 integrin knockout mice. To examine the survival of DiYF mutant mice after birth, the numbers of pups per litter from homozygous crosses were monitored. Even after 7 d of postpartum there was no significant difference in mortality of pups between these two groups. The DiYF mutant embryos (P = 12) or pups did not show any vasculature defects and hemorrhage in skin or in gastrointestinal tract as observed in β3 integrin knockout mice (Fig. 3 C). Development of large blood vessels exemplified by aorta also appeared to be normal because no significant differences (P = 0.445) in thickness and general histological structures of aorta were observed in DiYF mice compared with WT counterparts (Fig. 3 D).
To further investigate the effect of DiYF mutation on adult vasculature, tissue sections of liver (Fig. 3 E) and kidney (Fig. 3 F) were stained for vWF and laminin. No substantial differences in blood vessel density of vascular maturation were observed in DiYF mice. Vascular density was assessed in skin sections from WT and DiYF mice using staining for vWF, and there were no significant differences between these two groups (P = 0.725, n = 16; Fig. 3 G). Since vascular maturation and pericyte recruitment are known to be essential for pathological angiogenesis (21), staining for smooth muscle actin was performed to reveal mature blood vessels (Fig. 3 H). Quantitative results demonstrate the similar density of smooth muscle actin–positive blood vessels in skins of WT and DiYF mice (P = 0.127; n = 11). Since β3 integrin is known to be expressed on circulating white blood cells and involved in immune response, we demonstrated that the numbers of circulating immune cells do not differ among DiYF and WT mice. The numbers of circulating lymphocytes were 86.5 ± 0.93 and 80.1 ± 2.7 × 103 cells/μl for WT and DiYF mice, respectively. Monocyte counts were also similar: 2.34 ± 0.5 and 2.32 ± 0.6 × 103 cells/μl for WT and DiYF mice, respectively. Collectively, these data clearly indicate that DiYF mutation did not affect the embryogenesis, organogenesis, postpartum developments, normal vascular development and maturation, and levels of circulating white blood cells.
DiYF mutations within the β3 integrin cytoplasmic domain impair angiogenic properties of ECs
Previous studies demonstrated that αvβ3 integrin controls migration, invasive potential, and angiogenic phenotype of ECs (20). Therefore, we assessed whether impaired β3 integrin tyrosine phosphorylation affected EC capillary and tube formation ex vivo. WT but not DiYF ECs were able to form well-assembled and complete capillary cord-like structures in the presence of VEGF-A (Fig. 4 A).
A major phenotypic characteristic of ECs is the ability to assemble into the interconnected network of tube-like structures when grown in a three-dimensional matrix (22). WT but not DiYF ECs formed clearly defined and well-connected networks of EC tubes (Fig. 4 C). These structures were relatively stable and remained well organized for at least 18 h. As evident from Fig. 4 C, DiYF ECs were not able to complete tube formation, and no obvious patterns were formed even in the presence of VEGF. The extent of tube formation was quantified by measuring the length of tubes; the mean values of three independent experiments are represented graphically in Fig. 4 D. VEGF treatment induced approximately threefold increase in the length of the tubes formed by WT ECs but had a little effect on DiYF ECs.
β3 integrin cytoplasmic tyrosine motifs controls ex vivo angiogenesis in response to VEGF
We next assessed whether DiYF mutations affected an outgrowth of vascular sprouts from aortic segments isolated from mice. First, an ex vivo angiogenic assay was performed in Matrigel enriched with growth factors. The aortic rings from WT mice produced an extensive network of vascular sprouts, whereas DiYF aortic rings failed to do so (Fig. 4 E). To elucidate the role of β3 integrin in VEGF-induced responses, the aortic ring assay was performed in the presence or absence of VEGF using growth factor–reduced Matrigel. Under these conditions, aortic rings from WT mice produced a significantly higher number of vascular sprouts both in the absence and presence of VEGF than did aortic rings from DiYF mice (Fig. 4 F). Quantification of the number of aortic ring sprouts indicated that DiYF cells formed fourfold fewer vascular sprouts ex vivo, regardless of stimulation, than did WT aortic rings (Fig. 4 G). VEGF produced a small increase in capillary formation of DiYF rings; however, the number of sprouts was only 20–25% of that observed in WT aortic rings (Fig. 4 G).
To further analyze the capillary growth from aortic rings, a detailed kinetic study was undertaken. The time curves of vascular growth are presented in Fig. 4 H. In the absence of stimulation, very few microvessels were detected in either WT or DiYF implants even after a prolonged incubation, whereas serum-induced neovascularization was considerably higher in WT implants than in DiYF implants (Fig. 4 H). The peak values of capillary growth were observed 8 d after implantation and were 10 and 45 microvessels per ring for DiYF and WT, respectively. VEGF with endothelial growth supplement was a stronger stimulus than endothelial growth supplement alone, but both produced extensive formation of capillaries in WT but not in DiYF aortic rings (Fig. 4 H). Collectively, these results indicate that the impaired pathological angiogenesis in DiYF mice was caused by the defective functional responses of endothelial cells.
β3 integrin cytoplasmic tyrosine motifs regulates EC adhesion, spreading, and migration
To further define the nature of the angiogenic defect observed in DiYF mice, we compared angiogenesis-relevant functions of ECs isolated from WT and DiYF mice. We first assessed whether the mutation in β3 integrin cytoplasmic domain had any effect on EC adhesion and subsequent cell spreading on extracellular matrix substrates. WT and DiYF ECs were plated on various integrin ligands and numbers of attached and spread cells per field were counted. WT and DiYF ECs adhered and spread equally well on fibronectin, laminin-1, and collagen-coated plates (Fig. 5, A and B).
In contrast, a significant difference (P < 0.001) in the behavior of WT and DiYF ECs was found using the αvβ3 ligands vitronectin and entactin. On these substrates, DiYF ECs showed a twofold reduction in adhesion and a fourfold decrease in the number of spread cells on vitronectin and threefolds on entactin (Fig. 5, A and B). To further examine the role of β3 integrin cytoplasmic tyrosine motifs in regulation of receptor ligand-binding strength, we examined cell adhesion under conditions of shear stress. WT and excessive DiYF endothelial cells were plated on vitronectin-coated glass coverslips to achieve equal number of adherent cells. Then, adherent cells were subjected to perfusion with shear rates of 5–30 dyn/cm2 for 1 min. As shown in Fig. 5 C, shear stress revealed a dramatic difference in strength of adhesion between WT and DiYF endothelial cells. Shear stress of 10 to 15 dyne/cm2 almost completely abolishes DiYF endothelial cell adhesion to vitronectin, whereas >60% of WT endothelial cells remained tightly adhered to the matrix. Shear rate sufficient to wash away 50% of adherent cells was 23 and 8 dyne/cm2 for WT and DiYF endothelial cells, respectively (Fig. 5 C). This finding demonstrates that β3 integrin cytoplasmic tyrosine motifs regulates ligand-binding strength.
When the migratory activity of WT and DiYF ECs toward various extracellular matrix proteins known to be recognized by integrins was compared, results were similar to those of the adhesion assays: WT and DiYF ECs migrated equally well toward fibronectin, laminin, and collagen, but not toward vitronectin and entactin, where a 2.5-fold reduction in migration of DiYF ECs compared with WT was observed (Fig. 5 D).
We then compared the VEGF-induced migratory activity of WT and DiYF ECs. Stimulation of WT ECs with VEGF at 5, 10, and 20 ng/ml induced 1.5-, 2.5-, and 2.9-fold increases of migration compared with untreated ECs (Fig. 5 E). DiYF ECs also responded to VEGF stimulation; however, the rate of migration was substantially reduced when compared with WT cells (Fig. 5 E). Thus, tyrosine phosphorylation of αvβ3 integrin appears to play an important role in VEGF-induced EC migration to extracellular matrix. To further confirm these results, we used an alternative and more physiologically relevant method to assess EC migration. WT and DiYF ECs were plated on various integrin ligands and were allowed to form a confluent monolayer. Then, a wound in the monolayer was created and the healing process was monitored at different time points. The quantitative aspects of wound recovery and representative images of ECs are presented in Fig. 5, F and G, respectively. Whereas WT and DiYF EC migrated equally well on fibronectin, laminin, and collagen, a threefold reduction in migration on vitronectin was observed in DiYF ECs compared with WT (Fig. 5 F). The DiYF mutation impairs αvβ3 integrin-dependent and VEGF-stimulated responses of ECs, underscoring the role of phosphorylation of tyrosines in the cytoplasmic domain in the regulation of a cross-talk between αvβ3 and VEGFRs.
β3 integrin cytoplasmic tyrosine motifs are required for sustained activation of VEGFR-2
Next, we sought to identify a molecular mechanism responsible for abnormalities observed in DiYF ECs. Previous studies using β3-null mice demonstrated that the absence of β3 leads to up-regulation of VEGFR-2 and consequently to augmentation of angiogenic responses of ECs (8). However, no differences in VEGFR-2 levels were observed between DiYF and WT ECs from lung or aortic origin (unpublished data). It was previously shown that integrin β3 forms a complex with VEGFR-2 immediately upon stimulation with VEGF, and this association was proposed to be necessary for the activation of angiogenic program in ECs (23). Therefore, we sought to determine whether the DiYF mutations impaired the interaction of β3 integrin with VEGFR-2. Low levels of β3–VEGFR-2 interaction were observed in unstimulated WT ECs in suspension or upon adhesion to extracellular matrix. VEGF stimulated a dramatic increase in formation of the complex between β3 and VEGFR-2 in WT ECs plated on vitronectin, but not in suspension or on laminin, demonstrating the ligand specificity of this phenomenon (Fig. 6 A).
In contrast, no interaction between β3 and VEGR-2 was observed in DiYF ECs under any conditions (Fig. 6 B). Thus, β3 tyrosine phosphorylation was essential for an interaction between VEGFR-2 and αvβ3 integrin.
To investigate the mechanism of β3 integrin-dependent VEGF signaling, we analyzed the VEGFR-2 phosphorylation status in WT and DiYF ECs after VEGF stimulation. The time courses of VEGR-2 phosphorylation are shown in Fig. 6 C. In WT ECs, VEGF induced a bell- shaped response with a maximum sixfold increase in VEGFR-2 phosphorylation relative to unstimulated cells. After 45 min, VEGFR-2 phosphorylation returned to the control level. In contrast, the maximum VEGFR-2 phosphorylation in DiYF ECs upon VEGF stimulation was only 2.5-fold over control. Importantly, VEGFR-2 remained phosphorylated three times longer in WT ECs than in DiYF ECs (Fig. 6 C). Thus, the lack of integrin phosphorylation in DiYF ECs resulted in reduced phosphorylation, and therefore activation, of VEGFR-2 in response to VEGF, which in turn affected all the signaling events downstream of VEGFR-2. To further examine this phenomenon WT and DiYF cells were induced with VEGF for various time periods and cell lysates were analyzed for phosphorylation of extracellular signal–regulated kinase (ERK)-1/2 as downstream target molecule of VEGFR-2. In WT ECs, VEGF induced strong bell-shaped response with a 3.5-fold increase in ERK-1/2 phosphorylation relative to unstimulated cells within 2.5 to 15 min. In contrast, the maximum ERK-1/2 phosphorylation in DiYF ECs upon VEGF stimulation was only 2.5-fold over control and appeared only after 15 min of VEGF stimulation. Importantly, VEGFR-2–mediated ERK-1/2 phosphorylation was immediate and longer in WT ECs than in DiYF ECs (Fig. 6, D and E).
β3 integrin cytoplasmic tyrosine motifs are critical for VEGF-induced αvβ3 integrin activation
An intrinsic property of integrin is an increase in soluble ligand binding in response to stimulation, a process referred to as integrin activation. We and others previously reported that VEGF activates αvβ3 integrin on ECs in response to VEGF (24). We sought to determine whether impaired activation of VEGFR-2 in DiYF ECs resulted in defective αvβ3 activation by VEGF. VEGF induced a sixfold increase of fibrinogen binding to WT ECs, but only a threefold increase in binding to DiYF ECs (Fig. 7 A).
MnCl2, an agonist known to activate integrins and stimulate β3 integrin tyrosine phosphorylation (25), produced at least a 12-fold increase in fibrinogen binding to WT ECs compared with a 5-fold increase observed in DiYF ECs (Fig. 7 B). The specificity of ligand binding was confirmed by addition of 10-fold excess of unlabeled fibrinogen. Similar results were observed when integrin activation was monitored using a monovalent activation-dependent ligand WOW-1 Fab. VEGF and MnCl2 stimulated 9- and 11-fold increases, respectively, in WOW-1 binding to WT ECs and 3.5- and 4-fold increases, respectively, to DiYF ECs (Fig. 7 C and Fig. 6 D). Thus, it is apparent that the mutations within the cytoplasmic domain of β3 integrin in the DiYF mice significantly impair the process of integrin activation, resulting in defective cell adhesion and migration.
To investigate the role played by β3 integrin in pathological angiogenesis, we replaced the WT β3 integrin with a mutated form unable to undergo phosphorylation, creating the knock-in mouse DiYF. The mutant β3 integrin is expressed and is present on cell surfaces at the normal levels, but is defective in signaling. The major advantage of this model is that it allows monitoring of αVβ3 function in primary endothelial cells as opposed to model cell lines lacking appropriate receptors and signaling intermediates. The ex vivo and in vitro results from this model provided novel information that is different from the previously published results from in vitro studies and β3 knockout mice experiments (8). A unique aspect of our experimental system is that compensatory or over-compensatory responses were not observed; these responses often occur in knockout models when a protein of interest is absent during development (26). Using DiYF knock-in mice, we assessed pathological neovascularization in vivo and performed an extensive analysis of the underlying mechanisms of angiogenesis using ex vivo models. The major findings of this paper are the following. (a) Phosphorylation of β3 integrin in WT ECs occurred in response to integrin ligation and VEGF stimulation. (b) Lack of β3 integrin phosphorylation in DiYF knock-in mice resulted in impaired angiogenic response in three distinct in vivo models. As a result of reduced vascularization, tumor growth was significantly inhibited in DiYF mice compared with WT. (c) At the same time, DiYF mutation did not affect embryogenesis, organogenesis, and overall vascular development. In contrast to the phenotype of β3 integrin knockout mice, DiYF pups were broadly normal and did not show any hemorrhage in any organ. Vascular density and maturation in adult DiYF animal was also normal, indicating that integrin phosphorylation is crucial for pathological but not normal vascularization. (d) In ECs from DiYF mice, VEGF-induced functional responses (cell adhesion, spreading, migration, and capillary tube formation) were defective compared with WT. (e) Lack of β3 integrin phosphorylation in DiYF ECs lead to disruption of the VEGFR-2–β3 integrin complex and lack of VEGFR-2 phosphorylation in response to VEGF. (f) VEGF-induced integrin activation (inside-out signaling) was suppressed in DiYF ECs compared with WT ECs.
Our data show that, in WT ECs, phosphorylation of β3 integrin occurs in response to integrin ligation and VEGF stimulation. The finding that β3 phosphorylation occurs not only in response to adhesion to extracellular matrix (i.e., as a result of outside-in integrin signaling), but also upon treatment with VEGF, indicates that β3 might play an important regulatory role in outside-in integrin signaling, also known as integrin activation. In DiYF ECs, β3 cannot be phosphorylated, and integrin activation, as measured by soluble ligand binding to αVβ3, was considerably reduced compared with WT. Previous studies demonstrated that ligation or the treatment with the agonist Mn2+ induces β3 integrin phosphorylation in cells other than ECs (25). At the molecular level, it was also shown that β3 integrin phosphorylation controls the strength of receptor–ligand interaction (27), and might contribute to the overall functional activity of the integrin.
In three distinct models, one that measured capillary formation in Matrigel and two others for vascularity of implanted tumors, we observed that pathological angiogenesis was severely impaired in DiYF mice. Our findings are consistent with the results of studies using αVβ3 blocking antibodies and demonstrate a key role for αVβ3 integrin in angiogenesis, suggesting that β3 integrin is a positive regulator of angiogenesis and its phosphorylation is a critical step in EC responses during VEGF-induced neovascularization. Further, our findings provide an additional argument that increased angiogenesis observed in β3 knockout mice (8) is a result of molecular compensation via VEGR-2; this emphasizes the physiological importance of VEGR-2–αVβ3 cross-regulation in ECs, but does not reflect the true function of αVβ3 integrin in angiogenesis in an organism expressing normal amounts of this receptor. Since our experimental approach does not involve a complete deletion of β3 integrin from the cell surface, it does not trigger compensatory responses (i.e., up-regulation of VEGR-2). The phenomenon of molecular compensation—in many cases, over-compensation—in knockout mice has been described in numerous reports and represents a major drawback of the knockout approach. Frequently, the loss of a particular protein results in increased expression of other members of the same family, revealing the functional redundancy (28). It has been shown, for example, that αVβ5 integrin can compensate for the loss of α5β1 integrin (26). Analysis of the substitute molecule, which is structurally and functionally distinct from the targeted protein (29, 30), might reveal new connections within the molecular network. It is intriguing that VEGR-2, a tyrosine kinase receptor, can compensate for the loss of a member of integrin family of cell adhesion molecules, as VEGR-2 is distinct in regard to ligand repertoire and other functional aspects.
In the present study, impaired angiogenesis in DiYF knock-in mice resulted from the decreased adhesion, spreading, and migration of ECs in response to VEGF. Interestingly, impaired cell adhesion and migration was observed only on vitronectin as a substrate, but not on fibronectin or collagen. Nevertheless, ex vivo angiogenesis in Matrigel, and in vivo angiogenesis, was defective, emphasizing a role for αVβ3 in this process. At the molecular level, this study demonstrated that impaired tyrosine phosphorylation of β3 integrin in DiYF ECs abrogated αVβ3–VEGFR-2 complex formation and resulted in dramatically reduced phosphorylation of VEGFR-2 upon VEGF stimulation. Since VEGF (via VEGFR-2 as its major functional receptor on ECs) induces phosphorylation of αVβ3 and phosphorylation of αVβ3, in turn, is required for complete and sustained phosphorylation of VEGFR-2, it can be concluded that these two receptors are able to cross-activate each other in ECs, therefore forming a functional partnership that is essential for successful angiogenesis.
Collectively, our findings demonstrate that αVβ3 and its inside-out and outside-in signaling is essential for pathological angiogenesis and undoubtedly represents a promising target for pharmaceutical approaches. However, when developing new inhibitors for angiogenesis, one should take into consideration the complexity of αVβ3 regulation, its interactions with other receptors, and possible compensatory changes resulting from its suppression.
Materials And Methods
DiYF mice were generated in the laboratory of Dr. David R. Phillips and maintained on a C57/Bl6 background (seven generations of backcrossing). 6–8-wk-old WT and DiYF mice were used in our study. We performed all procedures according to protocols approved by Cleveland Clinic Foundation Institutional Animal Care and Use Committee.
In vivo Matrigel plug assay.
WT and DiYF animals received an abdominal subcutaneous injection of 500 μl Matrigel mixed with VEGF (60 ng/ml) and heparin (60 units/ml). After 7 d, the animals were killed and dissected. Matrigel plugs were removed and digested using 5 ml Drabkin reagent, and quantification of neovascularization was assessed using a hemoglobin assay as per the manufacturer's protocol.
WT and DiYF mice were subcutaneously injected with freshly harvested 106 B16F10 mouse melanoma or RM-1 mouse prostate cancer cells. Tumors were collected 10 d after injection and tumor weights were measured. Tumors were also photographed and processed for immunohistochemical staining.
Immunohistochemistry and image analysis.
Immunohistochemistry and image analysis was performed as described previously (21). Matrigel plugs or tumor tissues sections were stained with polyclonal rabbit anti-vWF antibody (Abcam, Inc.) or rat anti-F4/80 (Serotec, Inc). Sections were counterstained with hematoxylin. WT and DiYF skin, liver, and kidney tissue samples were processed and stained for vWF or laminin. B16F10 tumor sections were stained with polyclonal anti-CD31 (Fitzgerald Industries Intl.) and anti-laminin (Sigma-Aldrich) antibody.
Aortic ring assay and isolation of endothelial cells.
The mouse aortic ring assay was performed as described previously (21). WT and DiYF mouse lungs were excised, minced, and digested using collagenase-dispase reagent (3 mg/ml). Digests were strained and the resulting cell suspension was plated on flasks coated with 1 mg/ml fibronectin. Endothelial cells were isolated and characterized as described previously (31).
Cell adhesion and cell spreading assay.
The cell adhesion assay was performed as described previously (32). Mouse lung endothelial cell suspensions were added to ligand-coated wells and placed in humidified incubator for 45 min. The wells were gently washed three times with DMEM and photographs were taken. The numbers of attached and spread cells per field were counted. Adhesion assay under shear stress conditions was performed as described previously using flow chamber (33). Prior to the beginning of the experiment WT and 2.25-fold excess DiYF endothelial cells were added on vitronectin-coated coverslips to achieve approximately equal number of adherent cell population and allowed to interact with the matrix for 15 min. These coverslips were gently washed and introduced into the bottom of the flow chamber. Shear stress of 5–30 dyne/cm2 was applied using syringe pump for 1 min. Number of WT endothelial cells adhered to vitronectin under static condition per field was assigned the value of 100% and relative adhesion was calculated.
Endothelial cell migration and wound healing assay.
Endothelial cell migration assay was performed as described previously (24). WT and DiYF mouse lung endothelial cells were grown to confluence in 12-well plates precoated with various integrin ligands. Cells were serum starved and then a wound was created. Images were recorded immediately after wounding (time zero) and 12 h later. Cell migration was quantified using image analysis of five randomly selected fields of denuded area. The mean wound area is expressed as percent of recovery (% R) from three identically treated plates using the equation % R = [1− (Tt/T0)] × 100, where T0 is the wounded area at 0 h and Tt is the wounded area after 12 h.
WOW-1 and fibrinogen binding assay.
WOW-1 Fab binding assay was performed as described previously (34). To assess fibrinogen binding, WT and DiYF mouse lung endothelial cells were serum starved for 4 h and further induced with 20 ng/ml VEGF-165 or 1 mm MnCl2. FITC-labeled fibrinogen was added at a final concentration of 200 nM for 45 min. Cells were fixed and washed twice with ice-cold PBS. FACS was performed using a FACS Calibur (Becton Dickinson), and data were analyzed using CellQuest software program.
Endothelial tube and precapillary cord formation assay.
The formation of vascular tube-like structures by WT and DiYF mouse lung endothelial cells were assessed on the basement membrane matrix preparation. WT and DiYF mouse lung endothelial cells were seeded on Matrigel-coated plates. Medium with or without 20 ng/ml VEGF was added, and cells were further incubated at 37°C for 8 h. The tube formation was observed using an inverted phase–contrast microscope, and photographs were taken from each well. Using ImagePro software, the degree of tube formation was quantified by measuring the length of tubes in random fields. The precapillary cord formation assay was performed as described previously (22).
Immunoprecipitation and Western blot analysis.
Immunoprecipitation was performed as described previously (23). To analyze the β3 integrin tyrosine phosphorylation of WT and DiYF mouse lung microvascular endothelial cells, ECs were added to various integrin ligand-coated plates and incubated at 37°C for 60 min. Cell lysates containing equal amount of proteins were subjected to Western blot analysis using rabbit anti–integrin β3 [pY747] and [pY759] antibody (Biosource International, Inc). Cell lysates were also analyzed for β3 integrin expression as a loading control. Serum-starved WT and DiYF ECs were also treated with 20 ng/ml VEGF for 0–60 min. Cell lysates were analyzed by Western blot using anti–integrin β3 [pY747], anti–integrin β3 [pY759], anti–ERK-1/2 and anti–p-ERK-1/2 (Cell Signaling Technology) antibodies.
Values were expressed as mean ± SDs. p values were based on the paired t test. Results were considered statically significant with P < 0.05.
We acknowledge financial support from the National Institutes of Health (HL071625 and HL073311 to T.V. Byzovat) and the American Heart Association (0625271B to G.H. Mahabeleshwar).
The authors have no conflicting financial interests.
Abbreviations used: EC, endothelial cell; ERK, extracellular signal–regulated kinase; VEGF, vascular endothelial growth factor; VEGFR, vascular endothelial growth factor receptor; von Willebrand factor, vWF.