OUP user menu

Besides adhesion: new perspectives of integrin functions in angiogenesis

Guido Serini, Lucia Napione, Marco Arese, Federico Bussolino
DOI: http://dx.doi.org/10.1093/cvr/cvn045 213-222 First published online: 19 February 2008


During angiogenic remodelling in embryo and adult life, endothelial cells lining blood vessel walls dynamically modify their integrin-mediated adhesive contacts with the surrounding extracellular matrix. However, besides regulating cell adhesion and migration, integrins dynamically participate in a network with soluble molecules and their receptors. Angiogenesis is characterized by opposing autocrine and paracrine loops of growth factors and semaphorins that regulate the activation of integrins on the endothelial surface through tyrosine kinase receptors (TKR) and the neuropilin/plexin system. Moreover, pro- and anti-angiogenic factors can directly bind integrins and regulate endothelial cell behaviour. This review summarizes the recent progress in understanding the reciprocal interactions between integrins, TKR, and semaphorin receptors.

  • Plexin
  • Semaphorin
  • Tyrosine kinase receptors
  • Endothelial cells

1. Introduction

The molecular mechanisms leading to cell–extracellular matrix (ECM) interactions have been crucial for the evolution from protozoans to metazoans. Integrins represent the most important family of receptors mediating cell adhesion to ECM. Each integrin is composed of non-homologous transmembrane α and β subunits and they control cell adhesion through complex molecular mechanisms. Outside-in signalling informs the cell about the ECM environment, while inside-out signalling promotes modifications in integrin functional activity.1,2

Development and remodelling of vascular systems require complex interactions of signals and physical forces orchestrating the activities of endothelial cells (ECs), pericytes, and smooth muscle cells. Besides several redundant soluble factors, which appear to have a relevant role, two classes of molecules have been identified with a high specificity for the vascular system: the family of vascular endothelium growth factors (VEGF) and their tyrosine kinase receptors (TKR), VEGFR-1, -2, and -3, and the family of angiopoietins (Ang) and Tie-2 TKR.3,4 More recently, molecules firstly characterized for their role in axon guidance (e.g. semaphorins, netrins, and slits) have been selectively involved in the remodelling and sprouting phases of angiogenesis.5

Vascular cells (i.e. ECs, pericytes, and smooth muscle cells) express a wide range of integrins including α1β1, α2β1, α4β1, α5β1, αvβ1, αvβ3, αvβ5, αvβ8, α6β1, and α6β4.6 Integrin-mediated cell-to-ECM adhesion plays a deterministic role in vascular development by contributing to cell movement, to protect cells from anoikis and to endow the vasculature with the ability to sense and respond to changes in physical forces.6,7

In the last 10 years, an increasing body of evidences has demonstrated that integrins are not mere adhesion receptors, but influence the biological activity of several other molecular systems within the cell. Here, we reviewed emerging results highlighting new roles of integrins in angiogenesis.

2. Integrins modulate the activation of vascular tyrosine kinase receptors

2.1 Integrins and vascular endothelial growth factor receptors

During angiogenesis, ECs adhere to a provisional ECM mainly through αvβ3 integrin resulting in an increased biological response to VEGF-A dependent on the formation of a complex between the integrin and VEGFR-2.811 The collagen I receptor α2β1 and the laminin receptors α6β1 and α6β4 do not exert a similar effect. The formation of the integrin-TKR complex first requires the activation of VEGFR-2 by its ligand and the ensuing binding of phosphatidylinositol 3-kinase (PI3K)9 and c-src,11 which participate in directional cell migration triggered by VEGF-A. A monoclonal antibody anti-β3 not only perturbed the complex formation, but it also markedly inhibited VEGFR-2-mediated phosphorylation, PI3K activation, focal adhesion dynamics as well as EC proliferation and migration triggered by VEGF-A.9 In contrast, αvβ3 clustering was permissive for VEGFR-2 activation and optimal response of ECs to VEGF-A.9,10,12 Formation of the VEGFR-2/αvβ3 complex requires the extracellular domains of both αv and β3 integrin subunit8 and that of VEGFR-2.13 Recently, a series of elegant in vivo and in vitro studies10,11 defined the molecular details by which the β3 cytosolic tail regulates the endothelial response to VEGF-A. Indeed, upon VEGF-A stimulation, VEGFR-2 recruits and activates c-src, which in turn phosphorylates the cytosolic tail of β3 integrin at Tyr747 and Tyr759. This c-src-dependent post-translational modification is required for the formation of the VEGFR-2/αvβ3 complex and the conformational activation of the integrin, which enhances its affinity for the ECM11 (Figure 1). Recently, coagulation factor FXIII has been reported to play a key role in the stabilization and activation of the VEGFR-2/αvβ3 complex. In such a complex, VEGFR-2 is activated in a VEGF-A-independent manner that requires both the transglutaminase and tyrosine kinase enzymatic activities of FXIII and VEGFR-2, respectively.14

Figure 1

Effect of different ECM proteins and integrins on VEGFR-2 response. (A) Effect of αvβ3/VN pair on VEGFR-2 activation. VEGFR-2 triggering by VEGF-A activates a complex of signals involving a direct interaction of c-src to β3 integrin through the sequence YRGT762 114 and to VEGFR-2 through the adaptor T-cell-specific adaptor (TSAd).4 The figure points out the effect of longer VEGF-A isoforms (145, 165, and 189), which bind heparansulphates through a basic sequence; however, the same activity is shared by shortest VEGF-A121. This results in: (i) an association between the β3 integrin and VEGFR-2, which depends on VEGFR-2 extracellular domain, αv extracellular domain and the β3 subunit; (ii) an increased recruitment of PI3K, (iii) an increased level of VEGFR-2 phoshorylation, (iv) an increased activity of C-Src associated to β3 integrin, and (v) an enhancement of the biological properties of VEGF-A on EC. (B) Effect of α1β1 or α2β1/collagen I pair on VEGFR-2 activation. The engagement of integrins α1β1 or α2β1 by collagen I promotes the association of SHP-2 to Y1174 of VEGFR-2 promoting a fast dephosporylation of the receptor, and enhancement of its internalization and a reduction of receptor responsiveness to VEGF-A. It is possible that SHP-2 is recruited from the cytosolic tail of the integrin to the receptor (see text for details).

All the information obtained from experiments on cultured ECs require to be compared with data resulting from the analysis of β3 null mice that are alive without gross anatomical defects (reviewed by Hynes6), with the only exception of coronary defects.15 Furthermore, β3 null mice show an increased tumour-associated VEGF-A-dependent angiogenesis, which depends on VEGFR-2 over-expression (reviewed by Hynes6). Hence, in this genetic model, it seems that αvβ3 acts as a negative regulator of the VEGF-A/VEGF-R2 pathway, a role that is at odds with the anti-angiogenic effects obtained by αvβ3 antagonists. In this case, it has been proposed that antibodies and drugs interacting with αvβ3 could act as agonist of negative signals.6 Alternatively, the increased VEGFR-2 expression observed in β3 null mice could represent a molecular compensation that further put emphasis on the importance of VEGFR-2/αvβ3 integrin co-regulation in ECs. The latter hypothesis is supported by the recent finding that in knock-in mice expressing a β3 mutant unable to be phosphorylated in Tyr residues and to form a complex with VEGFR-2, tumour angiogenesis is impaired.11 Indeed, β3 integrin could play different or even opposite roles depending on different critical factors such as: (i) the phases/types of angiogenesis (e.g. sprouting, intussusception, or fusion); (ii) the ECM ligands and fragments (e.g. tumstatin or canstatin) engaged; (iii) the association/crosstalk with other receptors or extracellular proteins, such as FXIII14 or milk fat globule/EGF factor 8.16

In degranulating platelets, VEGF-A has been reported to interact with the C-terminal heparin-II domain of fibronectin (FN) and enhance its motility activity on EC.17,18 This effect results from an association between VEGFR-2 and α5β1, which is exclusively dependent on immobilized VEGF-A/fibronectin (FN) complex.

In contrast to the αvβ3/vitronectin (VN) pair, collagen I, the ligand of α1β1 and α2β1 integrins, exerts an inhibitory action on this TKR.19 EC adhesion to collagen I reduces VEGF-A-induced VEGFR-2 autophosphorylation by recruiting the tyrosine phosphatase SHP2 to the phosphorylated Tyr1117 of the receptor cytosolic tail. The interaction of SHP2 with VEGFR-2 is strictly dependent on EC adhesion to collagen I. The highest VEGFR-2 de-phosphorylation correlates with the highest degree of its internalization. We speculate that the pro-endocytic and inhibiting activity exerted by SHP2 on VEGFR-2 could be crucial to allow an accurate response of ECs migrating along VEGF-A gradients, as revealed by studies in Drosophila melanogaster.20 The effect of collagen I on VEGFR-2 parallels the effect of tissue inhibitor of metalloprotease (TIMP) -2, which negatively regulates VEGFR-2 by activating SHP1 phosphatase.21 TIMP-2/α3β1 integrin signalling, via SHP-1 activation, inhibits cell-cycle and enhances the expression of the anti-migratory membrane protease inhibitor RECK (reversion-inducing-cysteine-rich protein with kazal motifs), finally resulting in angiogenesis inhibition.22,23 Upon ECs stimulation with TIMP-2, SHP1 shifts from α3β1 integrin to VEGFR-2, which is de-phosphorylated and unable to trigger proliferation. Similarly, α1β1 integrin engaged by collagen I activates the T-cell protein tyrosine phosphatase function that inhibits EGF receptor signalling.24,25 It has been reported that in vascular smooth muscle cells αvβ3 engagement by VN results in tyrosine phosphorylation of β3 cytosolic domain and recruitment of SHP2, which modulate the activity of insulin growth factor I receptor.26 Thus, we hypothesize a protective role on VEGFR-2 signalling by VN-engaged αvβ3, which recruits SHP2 and preserves the receptor from phosphatase activity. In contrast EC adhesion on collagen I, which is mediated by α1β1 but not by αvβ3, could allow SHP2 interaction with VEGFR-2 (Figure 1).

The modulatory role of integrins on VEGFRs is not restricted to the type 2. Stimulation of VEGFR-3 by VEGF-C or VEGF-D plays a major role in lymphangiogenesis. It has been demonstrated that VEGF-C induces VEGFR-3 to associate with α5β1. Furthermore, integrin α5β1 ligation by FN is required for the optimal activation of VEGFR-3 signalling, not only at the receptor level, but also at its downstream PI3K/Akt pathway.27 Similarly, it has been demonstrated that β1 integrin engaged by FN or collagen transactivates VEGFR-3 by promoting the physical association between the integrin and the TKR.28

2.2 α5β1 integrin and the angiopoietin receptor Tie2

The Ang family and its Tie-2 TKR are required for correct organization and maturation of the newly formed vessels. In vitro, Tie-2 activation elicits EC adhesion, motility, and survival, while Ang-2 has an inhibitory effect.3 We have recently reported in ECs that Tie2 may be present both in free-form and associated with α5β1 integrin.29 The activation of α5β1 integrin by FN increases its interaction with Tie2 and modulates the time and concentration window of the receptor activation. When α5β1 is activated, Tie2 is phosphorylated at lower Ang-1 concentrations than those required on other ECM proteins. Furthermore, α5β1/Tie-2 complex allows the prolonged stimulation of Tie2 tyrosine kinase activity up to 1 hour, while free Tie-2 activation is shorter and more transient. Therefore, it seems that α5β1 activation could influence Tie2 signal duration and strength. Ang-1 triggers biochemical signals that recruit to the complex the p85 regulatory subunit of PI3K and focal adhesion kinase (FAK). It is known that p85 binds activated Tie2,3 whereas FAK is recruited to the cytosolic tail of clusterized integrins at focal adhesions.7 Thus, Ang-1 stimulation triggers both Tie2 and α5β1 signalling, and allows a cross-talk between these pathways by modulating Tie2/α5β1 complex. Our observation that Ang-1/Tie2 promotes the α5β1-dependent activation of PI3K signalling, which is known to depend on FAK,30 suggests that FAK recruitment to Tie2/α5β1 complex could be dependent on activated Tie2 inside-out signalling (Figure 2).

Figure 2

Effect of α5β1/FN pair on Tie-2 response to Ang-1. (A) Free-Tie-2 responds to high Ang-1 concentrations with transient activation. (B) Integrin α5β1 activated by fibronectin forms a complex with Tie-2, which in turn responds to a low Ang-1 concentration for long-lasting time; Tie-2 stimulation supports an integrin inside-out signal. The signal coming from α5β1/Tie-2 complex is further reinforced by a direct α5ß1 integrin stimulation by Ang-1. The stimulation results in a combined activity of inside-out and outside-in signals.

2.3 Integrins and the hepatocyte growth factor receptor Met

Besides these relatively specific angiogenic regulators, vascularization is under the control of pleiotropic molecules. One of them is hepatocyte growth factor (HGF), which activates its TKR Met on EC31 and promotes angiogenesis in a large variety of models. Moreover, HGF has been found to promote integrin-mediated adhesion.32 When HGF is released by platelets forms a complex with FN or VN. These hetero-complexes, but not HGF alone, trigger the association between Met and integrins. In particular HGF/FN and HGF/VN, respectively, induce Met to associate with α5β1 and αvβ3 integrins, with subsequent sustained level of auto-phosphorylation when compared with the non-associated receptor.33 Since it has been found that in cancer cells Met can complex with α6β4 integrin,34,35 known to be a regulator of new blood vessel formation in cancers,36 it is tempting to speculate about a possible regulatory role of HGF/Met activity via α6β4 during tumour angiogenesis.

2.4 Integrins and the fibroblast growth factor receptors

By using specific neutralizing antibody anti-αvβ3, Cheresh’s group suggested that this integrin cooperates with the TKRs of fibroblast growth factor (FGF)-2 to promote signalling events necessary for vascular survival and endothelial cell motility, thereby facilitating angiogenesis.37 Specific peptide antagonists of αvβ3 integrin inhibit the second and late wave of FGF-2-mediated activation of mitogen-activated protein kinase (MAPK) resulting in an inhibition of angiogenesis.38 Further studies allowed a better definition of the signals upstream MAPK and regulated by αvβ3 integrin. The sustained activation of MAPK by FGF-2/αvβ3 depends on p21-activated kinase, which phosphorylates c-Raf.39 However, the molecular mechanisms of the described cooperation between FGF-2 and αvβ3 integrin are far to be elucidated. A first hint is the demonstration in ECs that activated FGR receptor-1 binds this integrin engaged by fibrionogen.40 This result parallels other data showing an association between FGF receptor-3 and a wide number of integrins.41

3. Integrins bind angiogenic modulators

3.1 Integrins bind vascular endothelial growth factors

ECM can display a significant avidity for soluble molecules with the consequent variation of their diffusivity and affinity for their cognate receptors. Emerging evidences demonstrate that both angiogenic inducers and inhibitors may be entrapped by ECM and engage with integrins. Actually, α9β1 integrin forms a complex with both immobilized VEFG-A165 and VEGF-A121.42 EC adhere to and migrate on both isoforms using α9ß1. In response to immobilized VEGF-A, VEGF-R2 and α9ß1 assemble together and signal in an additive manner through phosphorylation of the downstream intermediates ERK and paxillin. Importantly, this complex seems to be operative in vivo because an antibody anti-α9β1 integrin partially reduces the angiogenic effect of VEGF-A.42 The same integrin has been reported to interact with VEGF-C, VEGF-D, and HGF, a finding that may help explain the abnormal lymphatic phenotype of mice expressing a null mutation of the α9 subunit.43,44 Similar observations have been reported for α3β1 and αvβ3;45 these integrins bind to VEFG-A165 and VEGF-A189, but not to the shorter isoform VEGF-A121 allowing EC migration and survival in a VEGF receptor-independent way (Figure 3). It is possible that the integrin/VEGF interaction could account at least in part for the diverse biological activities of VEGF-A isoforms and their ability to differently interact with ECM and co-receptors.4,46,47 Genetically modified mice show that the different VEGF-A isoforms can influence vascular patterning.46,47 These differences in the activity of VEGF-A isoforms could be likely due to their different ability to bind and interact with components of ECM.

Figure 3

Interactions between inducers (A) and inhibitors (B) of angiogenesis and integrins (see text for details).

3.2 Integrins bind angiopoietins

Integrins are non-endothelial-specific receptors for Ang and partially mediate the biological activities of this protein family. The molecular determinants of Ang involved in integrin binding are presumably localized in the fibrinogen-like domain; here it localizes the sequence QHREDGS, which resembles the integrin motifs KRLDGS or REDV of fibrinogen and FN, respectively.29,4850 EC, as well as fibroblasts, adhere to Ang-1 and -2, but only the former is able to induce cell spreading, haptotaxis, and a complete activation of cytoskeleton dynamics. β1 integrin heterodimers, and in particular α5β1, seem to be the most efficient adhesive receptors for Ang-1 independently from the presence of Tie-2.29,50 This observation, together with the observed cooperation between Tie-2 and α5β1 (see above), indicates that Ang-1 is capable of triggering both inside-out and outside-in signals. Indeed different events may occur in vascular ECs: (i) when high amounts of Ang-1 are available both high affinity Tie-2 receptor and low affinity/high avidity integrins may be independently activated; (ii) low Ang-1 concentrations, which are unable to sustain prolonged Tie-2 and integrin activation, can instead efficiently signal only through the fraction of Tie-2 that is constitutively associated with α5ß1 and make ECs more sensitive to low amounts of Ang1.

In other cell types lacking Tie-2, Ang interact only with integrins. In cardiomyocytes, a wider range of integrins (α2β1, α5β1, α6β1, α6β4, αvβ1, and αvβ3) bind Ang-1 and Ang-2, resulting in cell survival due to Akt activation and caspase-3 inhibition.48 Glioma cells adhere to Ang-2 via α3, α5, αv, β1, and β3, but only αvβ1 engagement by Ang-2 sustains efficient signalling that results in matrix metalloprotease-2 production, cell migration, and invasion.49 On breast cancer cells instead adhere and migrate towards Ang-2, by using α5β1.51 Collectively these data suggest that each cell type binds Ang through a distinct set of integrins, which only partially overlaps among different cell types (Figure 3).

3.3 Integrins bind angiogenesis inhibitors

A significant number of angiogenesis inhibitors derive from the proteolytic cleavage of ECM proteins52 and therefore it is not surprising that they bind integrins (Figure 3). The interaction of integrins with their canonical and intact ECM ligands results in EC adhesion and activation of anti-apoptotic signals.7 On the contrary, integrins engaged by these ECM-protein fragments promote apoptosis and reduce proliferation and motility. Altogether these data suggest that different and opposing outside-in signals are triggered by integrins depending on whether they are engaged by either intact ECM protein or fragments deriving from its cleavage.

Endostatin, the C-terminal non-collagenous domain of type XVIII collagen, exerts its inhibitory activity on EC migration mainly by interacting with α5β1. An Arg-rich peptide at the N-terminus of endostatin seems to be important for its interaction with β1,53 supporting previous data excluding the role of an Arg-Gly-Asp sequence.54 However, it has been also reported that Arg-Gly-Asp cyclic peptides inhibit EC binding to immobilized endostatin,55 implying that soluble or immobilized endostatin differently interacts with α5β1. It is conceivable that α5β1 conformation is different when it binds an intact ECM ligand or endostatin, thus leading in the latter case to aberrant and perturbed signalling events, such as sustained activation of src, inhibition of FAK and MAPK.55,56

The non-collagenous domain present in the α chains of type IV collagen generates three different angiogenic inhibitors: tumstatin, arresten, and canstatin. Tumstatin is the non-collagenous domain of the α3 chain of type IV collagen that induces apoptosis and inhibits EC proliferation through its binding to αvβ3 integrin, leading to suppression of cap-dependent protein translation. The tumstatin/αvβ3 interaction is independent from the Arg-Gly-Asp binding site and therefore it may explain the inhibitory signals triggered by this interaction. Actually tumstatin inhibits the activation of FAK, Akt, and mTOR (target of rapamycin) -mediated phosphorylation of the eukaryotic initiation factor 4E-binding protein involved in the control of protein synthesis.55,57 Arresten corresponds to the non-collagenous domain of the α1 chain of type IV collagen and blocks EC functions by competing with collagen IV binding to α1ß1 integrin and inhibiting MAPK-mediated signals.58 Finally, canstatin corresponds to the α2 chain of type IV collagen, binds αvβ3 and αvβ5 leading to activation in ECs of an apoptotic program involving both caspase-8 and -9.59 The activation of both caspase-8 and -9 results in the amplification of mitochondrial-dependent apoptotic events and in the activation of caspase-3, the central executioner of the apoptotic process. A similar process is activated by angiostatin, which corresponds to the N-terminal four kringles of plasminogen and by binding αvβ3 triggers the non-mitochondrial caspase-8-dependent apoptotic pathway in ECs.59,60

4. Semaphorins regulate endothelial integrin function and angiogenic remodelling

Semaphorins (Sema) are a family of secreted and membrane-bound repulsive cues, which have been originally identified for their ability to affect axon behaviour in the developing nervous system.61 Semaphorins signal through four classes of plexins, named type A-D, a family of membrane receptors characterized by the presence in their cytosolic tail of two domains with homology to the R-Ras GTPase activating proteins (GAPs), separated by a linker region that can bind other small GTPases, such as Rnd-1 and Rac1.61 In vertebrates, members of the secreted class 3 Sema employ Neuropilin (Nrp)-1 or -2 as co-receptors in association with type A or type D plexins.61

4.1 Semaphorins and vascular development

The first evidences for a role of Sema/Nrp/plexin system in vascular biology were provided by the groups Klagsbrun and Fujisawa, which respectively demonstrated that in ECs Nrp-1 acts as VEGFR-2 co-receptor62 and found that Nrp-1 is required for mouse cardiovascular development.63 Afterwards, several reports confirmed and extended these observations. In zebrafish, knockdown of sema3aa affects the migration of Nrp-1+ angioblasts, finally impairing dorsal aorta formation and normal circulation.64 Additionally, single morpholino knockdown of either sema3aa or sema3ab in TG(fli1:EGFP)y1 embryos results in a less dramatic phenotype with patterning defects of intersomitic vessels.65 Knockdown of Sema3a gene in outbred CD-1 mouse strain and over-expression of dominant negative Sema3 receptor mutants66 or delivery of anti-Sema3A antibodies67 in chick embryos were found to cause angiogenic remodelling defects. In a different colony of outbred CD-1 mice, Sema3a knock down did not result in vascular defects.68 These discrepancies could be due to differences in the genetic background. Indeed, outbred stocks undergo genetic heterogeneity depending on colony maintenance; moreover, even within the same outbred colony the genetic background can change over time.69

The observation that both in vitro and in vivo66 angiogenic ECs display autocrine loops of several Sema3 other than Sema3A7073 together with the partial penetrance of the vascular phenotype in Sema3a−/− mice suggest that multiple Sema3 could cooperate to regulate angiogenesis. Notably, during angiogenesis and in cultured ECs, opposing autocrine loops of Sema3A66,70,71,73 and VEGF-A72,7477 have been found and the observed loss of autocrine Sema3A in favour of VEGF-A in ECs during malignant tumour progression73 could account at least in part for the structural and functional abnormalities of tumour vasculature.

In ECs plexinD178 and, albeit to lesser extent, plexinA279 are the most abundant plexins. Both Sema3A and Sema3C bind with a significantly higher affinity to a receptor complex formed by the association of Nrp-1 and/or -2 with plexinD1 than to a complex in which Nrps associates with plexinA1.78 Therefore, the Nrp/plexinD1 complex could represent the most efficient transducer of the chemorepulsive effect of Sema3A.72,73,8082 Different from other Sema3, Sema3E can directly bind to plexinD1.83 Mainly based on defects in the intersomitic vessel patterning of Sema3E and plexinD1, Sema3E/plexinD1 has been proposed to be the major signalling pathway regulating vascular development.83 However, while Sema3e null mice are viable and do not show any gross abnormality,83 all plexinD1−/− pups become cyanotic shortly after birth and succumb within 24 h because of severe cardiovascular defects.78 Therefore, it is likely that in ECs, plexinD1 transduces signals not only from Sema3E, but also from other Sema3 likely employing Nrp as co-receptors, as originally proposed by Gitler and colleagues.78 In this respect, it is worth noting that in neurons the simultaneous presence of Nrp1 and plexinD1 on the cell surface converts Sema3E/plexinD1 signalling from repulsive to attractive.84 Based on the fact that ECs express high levels of both Nrp1 and plexinD1 and on observations that Sema3E promotes tumour angiogenesis,85,86 at present these findings cannot be easily reconciled with the proposed chemorepulsive effect played by Sema3E via plexinD1 on ECs of developing mouse embryos.83 Finally, the recent observation that Sema4A upon binding to plexinD1 inhibits EC migration and in vivo angiogenesis further indicates that plexinD1 conveys to ECs signals from multiple Sema.87

The complexity of Sema system in vasculature is further supported by the recent data showing that Sema3A is a powerful vasopermeabilizing molecule through a mechanism that is independent from its effects on integrin functions and involves VE-cadherin phosphorylation.88

4.2 Semaphorins regulate integrin function

By investigating the cellular and molecular events by which Sema can regulate vascular development, we found that Sema3A impairs EC adhesion and migration by inhibiting integrin activation.66 Accordingly, after few minutes of stimulation with Sema3A, adherent ECs lose their focal adhesions and then collapse,82 and Sema3F causes EC retraction as well.89 Moreover, Sema3A and Sema3F have been shown to inhibit integrin activation and adhesion to the ECM in several cell types.9094 Interestingly, Sema3C, which contains an Arg-Gly-Asp motif, has been instead reported to promote EC adhesion and migration.95

It is well known that integrins undergo conformational modifications that regulate their affinity for ECM ligands.2 In the low affinity state, the integrin extracellular domain is bent over the plasmamembrane and the α and β cytoplasmic tails tightly interact. In ECs chemokines, growth factors, or fluid shear stress activate signalling pathways that finally favour the transition of integrins towards the high affinity state.2,7 Direct interaction of talin trefoil FERM (Protein 4.1, the ezrin, radixin, moesin) domain with the cytodomain of integrin β subunits results in the unclasping of the cytoplasmic tails of integrin α and β subunits together with the extension of the extracellular domain that, by exposing the ECM binding site, accomplish the transition of integrins to the high affinity state.96 In talin, the integrin binding site within the trefoil FERM domain is masked by an intramolecular interaction with the rod domain. Interaction of membrane phosphatidylinositol 4, 5 bisphosphate (PIP2) with talin rod domain lessen this inhibition, thus allowing talin head-to-tail dimerization and integrin binding to the FERM domain. Notably, once activated talin can in turn bind and activate the enzyme phosphatidylinositol-4-phosphate 5-kinase (PIPKIγ 661) that, producing further PIP2, gives rise to a positive feed-back loop with talin that stabilizes cell adhesion to the ECM.97 In addition, integrin function is also regulated by Rap 1 and R-Ras small-GTPases.98 Specifically, activated R-Ras-GTP localizes at adhesive sites through its C-terminal tail;99 here R-Ras is thought to promote cell adhesion by favouring the activation of other small GTPases, such as Rap1 and Rac1.100 In this regard, it has been recently shown that binding of activated R-Ras to RLIP (Ral interacting protein) 76 leads to Arf (ADP-ribosylation factor) 6 activation, which promotes adhesion-induced GTP loading of Rac1.101 In the presence of Nrp-1, the juxtamembrane basic sequence of class A plexins directly binds to FARP2, a Rac (ras-related C3 botulinum toxin substrate 1) guanosine exchange factor (GEF).94 Sema3A binding to the Nrp-1/plexinA1 receptor complex triggers the dissociation of FARP2 from PlexinA1. Afterwards, FARP2 (FERM, RhoGEF, and pleckstrin domain protein 2) is free to exert its GEF activity leading to a rapid increase of active Rac1-GTP that in turn favours the binding of the constitutively active small GTPase Rnd1 (Rho family GTPase 1) to the linker region of plexinA1 cytoplasmic (Figure 4). This event finally activates PlexinA1 latent R-Ras GAP activity that then switches-off R-Ras, thus inhibiting integrin function. Importantly, Toyofuku and colleagues also found that, similarly to talin, FARP2 contains FERM domain by means which it binds plexinA1. Once detached from plexinA1, the FERM domain of FARP2 competes with talin for binding to PIPKIγ661, finally impairing the PIP2-based positive feedback required for the formation of focal adhesions.

Figure 4

Effect of Sema3A on the inhibition of integrin function. (A) On the cell surface, αβ integrin heterodimers exist in a high affinity state, stabilized by the interaction of talin head (FERM domain) with the β subunit tail. The interaction between talin and the actin cytoskeleton occurs. Talin-activated PIPKIγ661 generates a PIP2-based positive feedback loop, which amplifies talin activation. The small GTPase R-Ras participates in integrin activation through a still poorly characterized mechanism. In the absence of SEMA3A, plexinA1 associates with Nrp-1 and the FERM domain-containing guanosine exchange factor FARP2. (B) Upon SEMA3A binding, FARP2 is released from plexinA1. Free FARP2 then activates Rac and favours Rnd-1 association with and activation of plexinA1 cytoplasmic GAP domain, which in turn inhibits R-Ras and integrin function. Moreover, released FARP2 inhibits PIPKIγ661 activity.

EC migration and angiogenesis are also regulated by transmembrane Sema4D, which, depending on the cellular context, can however exert either chemoattractive or chemorepulsive activities. Negishi’s laboratory provided the first evidence that plexins are endowed with an R-Ras GAP enzymatic activity by showing that Sema4D-mediated plexinB1 stimulation suppresses R-Ras activation102 that in turn inhibits β1 integrin activation.103 Sema4D-driven R-Ras inhibition depends on the RasGAP activity triggered by Rnd1 binding to plexinB1.102,103 However, the inhibitory activity of Sema4D is not a general phenomenon. Indeed, as first reported by Gutkind and colleagues104 and then by us,105 Sema4D is pro-angiogenic and elicits EC migration.106 Sema4D stimulatory activity on ECs requires the formation of a complex between plexinB1 and Met TKR.105 In addition, upon treatment with Sema4D plexinB1 receptor associates with a Rho-GEF capable of activating Rho GTPase and its downstream effector Rho kinase, which by phosphorylating myosin light chain could control the assembly and the contraction of actin stress fibres.106 Altogether these data indicate that Sema4D/plexin B1 mediate different and sometimes opposite effects depending on the cellular context. As recently suggested, this may be caused by plexinB1 association with different TKR receptors; indeed, in carcinoma cells Sema4D can have pro- and anti-migratory effects depending on the interaction, respectively, with either Met or ErbB-2 TKRs.107

4.3 Is FAK a converging node for signals triggered by VEGF-A, integrins, and semaphorins?

Engagement of integrin receptors with their extracellular ligands leads to the formation of well-defined structures, termed focal adhesions, linking the ECM, and cytoplasmic actin cytoskeleton. These adhesive structures, which are dynamic and composed by a wide array of transmembrane and cytosolic proteins, serve as sites of force transmission required for cellular movements. Indeed, the coordinated regulation of formation and turnover of focal adhesions is central to cell responses to chemotactic and chemokinetic stimuli.108 A primary element of focal adhesion is FAK, a kinase that is primarily recruited to sites of integrin clustering via interactions between its C-terminal region and integrin-associated proteins such as talin and paxillin. The cytosolic tail of β integrins facilitates FAK activation probably involving FAK clustering and autophosphorylation of Tyr 379 (reviewed by Mitra and Schlaepfer30). Furthermore, FAK is a downstream effector of several TKRs, including VEGFR-2 and FGF receptors in ECs.39,109,110 VEGF-A via Src induces the site-specific tyrosine phosphorylation of FAK on Tyr 861, leading to the formation of a complex between FAK and αvß5, which is essential for the vascular permeability induced by VEGF-A.111

Recent evidences indicate that the repulsive or attractive functions of semamaphorins can involve FAK as well. Indeed, Sema3B can attract neurons by inducing membrane re-localization of phosphorylated FAK, which in turns activates the cytosolic tyrosine kinase Fyn.112 Similarly, Sema4D elicits EC motility and angiogenesis104,105 by activating proline-rich tyrosine kinase-2, a non-receptor tyrosine kinase closely related to FAK.106 In contrast, both in cultured ECs and in chick chorionallantoic membrane, chemorepulsive Sema3A reduces the basal phosphorylation of Src and FAK, and induces a rapid disappearance of focal contacts followed by the collapse of the actin cytoskeleton,82,88 thus supporting its role in inhibiting integrin function.66

In addition to FAK and integrins, many other proteins localize to focal adhesion, including VEGFR-2.113 Even though the membrane topology of semaphorin receptors is still unknown, it is tempting to speculate that ECM adhesive structures could represent ‘rendez-vous‘ points, where multiple cross-talks between positive and negative regulators of integrin function take place.

5. Conclusions

Angiogenesis is a relevant target for the treatment of many diseases. Inhibition of angiogenesis is the aim of protocols developed for tumours, chronic inflammatory diseases, and retinopathies, while vascular regeneration inspires therapeutic angiogenesis in the treatment of ischaemic injuries. Within the molecules investigated as exploitable targets for angiogenic therapies, VEGF-A and its receptors are considered the most promising. However, there are tremendous differences between pre-clinical and clinical results. The data reviewed here clearly indicate that the activity of every drug has to be considered in connection with the robust network of regulatory signals that controls angiogenesis. Indeed, the vast accumulation of experimental and clinical reports on tumour angiogenesis can be revisited in light of the mechanisms by which cancer modifies the robustness of regulatory networks leading to the observed abnormal vascularization, the final aim being to pharmacologically restore the robustness of the angiogenic regulatory networks in the everyday clinical practice. This interpretation might explain why some tumours display lack of sensitivity and others develop resistance to anti-VEGF-A therapy.


This work has been partially supported by Telethon (grant: GGP04127), Associazione italiana per la ricerca sul cancro (AIRC), European Community (LSHM-CT-2003-503254), Regione Piemonte (Ricerca Finalizzata 2006, Ricerca Scientifica applicata 2004 grants D10 and A150, Ricerca industriale e competitiva 2006: grant PRESTO), Fondazione Cassa di Risparmio di Torino, Torino, and ‘Ministero della Salute’ (Programma Ricerca Oncologica 2006, Ricerca Finalizzata 2006).

Conflict of interest: none declared.


View Abstract