The classical cytoplasmic partner of VE-cadherin is β-catenin, a very versatile adaptor molecule that binds and recruits several regulatory and scaffold molecules to cell junctions (for a provisional list of molecules engaged to adherens junctions through β-catenin see Lampugnani [46]). Therefore, β-catenin provides a fundamental contribution to the molecular architecture and function of endothelial junctions. Indeed, truncated VE-cadherin, lacking the carboxy-terminal half of the cytoplasmic domain and unable to bind β-catenin, induces defects in vivo and in vitro superimposable to those observed after null mutation of VE-cadherin [11, 70].

It has been proposed that the phosphorylation of tyrosine residues of β-catenin can decrease its binding affinity for VE-cadherin. This has been reported in response to vascular endothelial growth factor (VEGF) [67]. In addition, focal adhesion kinase (FAK) could contribute to the weakening of endothelial junctions in response to VEGF, phosphorylating the tyrosine residue 142 of β-catenin and decreasing its association with VE-cadherin in vivo [13].

However, other reports are in contrast to the model in which VE-cadherin/β-catenin complex is modulated by tyrosine phosphorylation of β-catenin (or VE-cadherin itself; see below). Tyrosine phosphorylation of β-catenin could contribute to loss of barrier function, without requiring detachment from VE-cadherin. Indeed, Timmerman et al. [102] report tyrosine phosphorylation of VE-cadherin-associated β-catenin in response to thrombin, and no dissociation of the complex. Nonetheless, tyrosine phosphorylation of β-catenin is critical to sustain the permeability increase in response to thrombin. Indeed, depletion of the phosphatase SHP2, which dephosphorylates β-catenin, inhibits recovery of the barrier after thrombin challenge [102]. This may suggest that tyrosine phosphorylated β-catenin could change its association with partners other than VE-cadherin, and in this way could convey different signals. In summary, no general agreement on this issue has so far been reached. The specific mode of modulation of the VE-cadherin/β-catenin complex appears dependent on the stimulus and, possibly, on the origin of the endothelium. These variables could determine phosphorylation of specific tyrosine residues of β-catenin [56], resulting in apparently contrasting results (for example, see the opposite results obtained either with human umbilical vein endothelial cells [HUVEC] cultured on fibronectin, or human pulmonary microvascular endothelial cells cultured on collagen, reported by Timmerman et al. [102] and Monaghan-Benson and Burridge [67], respectively).

In some conditions of severely and constantly disorganized junctions, β-catenin can partially dissociate from VE-cadherin. For example, this has been reported in CCM1-silenced endothelial cells [31], and we have observed a similar event in CCM3-knockout endothelial cells [8]. A reasonable hypothesis is that, in endothelial cells, even a limited decrease in the association between VE-cadherin and β-catenin is sufficient to create subtle local heterogeneity in the molecular composition of cell-to-cell junctions that weaken the continuity of the monolayer and impair its barrier function.

In addition to its adaptor role at junction, β-catenin is a crucial regulator of transcription, particularly in association with transcription factors of the Tcf family [16]. This aspect is discussed below (see also comments in the legend for Fig. 6.1).

A particular example of indirect control of transcription by β-catenin localized to junctions is the regulation of nuclear localization of the transcription factor FoxO1 (see below).

Among the several molecules that associate with VE-cadherin at cell-to-cell junctions and control vascular permeability and vascular structure, we report here on the small GTPase Rap1 [32] and the CCM proteins (CCM1, 2, and 3) [85] as our laboratory also contributed to the definition of their function in the endothelium.

These molecules are discussed together as CCM1 has been demonstrated to recruit and localize Rap1 to cell-to-cell junctions [31]. CCM1 associates with adherens junctions, directly binding β-catenin through its FERM domain [31]. In addition, co-immunoprecipitation of CCM1 with the adherens junction components VE-cadherin , α-catenin, and afadin has been reported [31]. However, interaction with β-catenin appears to be nonexclusive for junctional localization of CCM1. Indeed, CCM1 can be recruited to cell junctions through interaction of its FERM domain (the groove between subdomain 1 and 3) with the C-terminal region of the transmembrane protein heart of glass (HEG1) [29]. In addition, CCM1 appears to sustain functions downstream of HEG1. Indeed HEG1, CCM1, or CCM2 murine and zebrafish mutants show convergent traits in their abnormal phenotypes [40].

The stabilizing activity of Rap1 on endothelial cell-to cell junctions has long been identified and analyzed in several models of endothelial cells cultured in vitro [32, 42]. The cyclic adenosine monophosphate (cAMP) analog 8-pCPT-2ʹ-Me-cAMP (007), a synthetic cell-permeable activator of Epac1 (an Rap1 GEF), is a potent stabilizer of VE-cadherin-based endothelial cell-to-cell junctions [32, 42]. In addition, Rap1b isoform knockout mice show hemorrhagic phenotype due to unstable blood vessels [14]. When ablated in mice, Rap1a, the other isoform of Rap1, does not determine a vascular phenotype. Similar effects of vessel fragility that result predominantly in cerebral hemorrhages have been observed in zebrafish larvae treated with Rap1b morpholino [34]. Local signaling activated by Rap1 would determine polymerization of cortical actin through the activation of Rac [6, 32]. Rap1 can recruit an Raf1/Rok-α complex to junctions for local modulation of actomyosin contraction [111], as discussed above). However, the molecular mechanism of junction stabilization in response to Rap1 activation still remains largely unknown. Association with junctions of GEFs and GTPase-activating proteins (GAPs) can regulate the localization and duration of Rap signaling [32]. PDZ-GEF1, an Rap1 GEF, can be found to be associated with VE-cadherin through β-catenin and MAGI1 [90, 32]. Another Rap1 GEF, Epac1, can associate with VE-cadherin through the phosphodiesterase PDE4D [83]. Afadin (AF6) [60], which can recruit both Rap-GTP and RapGAPs, can associate with adherens junctions [6, 32].

While constitutive activation of Rap1 is required for maintaining stable junctions, and its activation further reinforces them, it is debated whether Rap1 deactivation or delocalization can destabilize established cell-to-cell junctions [32].

Interestingly, besides CCM1, which binds directly to Rap1 [94], CCM2 and CCM3, which can form a complex with CCM1 [97, 107, 114], also strongly contribute to the maintenance of stable endothelial cell-to-cell junctions in both in vitro and in vivo models [7, 8, 51, 58]. Indeed, as soon as they are ablated, adherens junctions (and, as a consequence, tight junctions; see below) become disorganized and permeability is increased.

Local activation of Rap1 at adherens junctions, as promoted by CCM1, is critical for stabilization of endothelial junctions. Indeed, stimulation of Epac1 with 8-pCPT-2ʹ-Me-cAMP (007) in CCM1-ablated cells does not restore either endothelial barrier or localization of Rap1 to cell-to-cell junctions [31]. Consistently, activation of Epac1 does not restore junction and permeability in Rap1-depleted endothelial cells [77]. While it is established that CCM2 binds directly to CCM1, and CCM3 binds to CCM2 [97], it is not known how the absence of either CCM2 or CCM3 could influence the localization or activity of Rap1 bound to CCM1. In addition to affecting Rap1 GTPase, ablation of CCM1, 2, or 3 stimulate the activity of Rho, with consequent phosphorylation of MLC2, actomyosin contraction, and disorganization of cell-to-cell junctions [99, 109] (for Rho activation after CCM1 and 2 ablation); [8] (for Rho activation after CCM3 ablation). Inhibition of Rho activity can restore the structure and function of junctions in CCM-ablated endothelial cells, and inhibit formation of vascular lesions in vivo (see below). Rap1 could mediate the downregulation of Rho by CCM proteins recruiting Rac and inhibiting Rho activation at cell-to-cell junctions [6, 99]. Figure 6.2 summarizes the molecular interaction at cell-to-cell junctions between VE-cadherin, β-catenin, CCM proteins, and small GTPases.

The permanence of clustered VE-cadherin at cell-to-cell junctions can be modulated by post-transcriptional modification of the cytoplasmic domain of VE-cadherin, particularly phosphorylation on tyrosine (Y658 and Y685 [75]) and serine (S665 [26]) residues. Such post-transcriptional modifications can regulate the internalization rate of VE-cadherin and, as a consequence, the stability, composition, and signaling activity of adherens junctions.

VE-cadherin posses several tyrosine residues (eight in the human protein and nine in the mouse protein) that could be potential targets of tyrosine kinases [81]. Of these residues, only tyrosine residues at positions 658 and 685 have, to date, been reported to be phosphorylated in vivo in the mouse 45, 75]. In a current model, tyrosine phosphorylation of VE-cadherin in response to angiogenic or inflammatory stimuli has been associated with weak and permeable cell-to-cell junctions [81, 103]. However, tyrosine phosphorylation of VE-cadherin may not be sufficient, and other signaling pathways possibly need to concur to determine the loss-of-barrier function [1]. This concept has been further reinforced by the work of Orsenigo et al. [75], who observed that VE-cadherin is indeed phosphorylated on tyrosine residues 658 and 685 in vivo, specifically in the veins under physiological conditions. Such phosphorylated VE-cadherin is regularly distributed to cell-to-cell junctions, and the rate of its internalization is not increased in comparison to the nonphosphorylated form. However, such phosphorylation sensitizes the cells to respond to permeability-increasing agents such as bradykinin and histamine, with consequent enhanced ubiquitination and internalization of VE-cadherin. Endothelial cells expressing the nonphosphorylatable form of VE-cadherin, such as Y658F or Y685F, with tyrosine (Y) mutated in phenylalanine (F), do not reduce their barrier function in response to either bradykinin or histamine, nor internalize the mutated VE-cadherin. Specific phosphorylation of VE-cadherin in veins is determined by the low shear stress characterizing veins and mediating the constitutive activation of Src that directly or indirectly targets VE-cadherin [75].

The serine residue 665 (S665) appears to be a critical target in the action of VEGF as an agent that decreases the endothelial barrier [26]. Although the effects of such phospho-residue have been studied in detail in in vitro models of cultured endothelial cells , its presence has been observed as a short-term response to VEGF also in vivo in VE-cadherin from mouse skin vessels [27]. Once more, a key factor would be Src, which, activated by VEGF receptor 2 (VEGFR2), upon VEGF binding, would phosphorylate Vav2. This GEF starts a chain of activating events: activation of the small-GTPase Rac, which stimulates its downstream target p21-activated-kinase (PAK) to phosphorylate the residue S665 in VE-cadherin. Β2-arrestin recognizes such phosphorylated residue and, upon binding to VE-cadherin, promotes clathrin-dependent internalization of VE-cadherin. The consequence of this process is weakening of cell-to-cell junctions. Interestingly, this sequence of events can be blocked by angiopoietin-1 (Ang1). Ang1, promoting the sequestering of Src through mDia, inhibits the phosphorylation of VE-cadherin on S665, its internalization, and the ensuing weakening of junctions.

Phosphatases are expected to modulate the phosphorylation of VE-cadherin. Until now, the tyrosine phosphatase SHP2, which is associated with VE-cadherin through β-catenin, has been reported to dephosphorylate β-catenin, γ-catenin, and p120 associated with VE-cadherin , but not VE-cadherin itself [102, 105].

The endothelial-specific vascular endothelial phosphatase (VE-PTP) is associated with VE-cadherin at junctions through interaction of the extracellular domains of the two proteins. VE-PTP can indeed dephosphorylate VE-cadherin, besides Tie2 and plakoglobin [24, 89]. It has recently been reported that VEGFR2, when associated with junctions, can also be a target of VE-PTP [35].

In addition, clathrin-dependent internalization of VE-cadherin can be inhibited by p120 catenin masking an endocytic signal [69].

VE-cadherin clustered at junctions can also modulate the activity of receptors that act as crucial regulators of vessel organization, such as VEGFR2 [11, 50] and transforming growth factor (TGF)-βR2 [88]. Activated by VEGF, VEGFR2 becomes phosphorylated on several tyrosine residues [15]. These phosphorylated tyrosines become docking sites for different mediators that can activate locally-specific signaling pathways [15], depending on the stability of junctions. In5 particular, we have observed that when VE-cadherin is engaged in stable contacts, as in confluent endothelial monolayers, activated VEGFR2 associates with VE-cadherin through β-catenin and activates phosphatidylinositol3-kinase (PI3K) [also associated with VE-cadherin through β-catenin] to phosphorylate Akt and to signal cell survival and resistance to apoptosis [11, 100]. When VE-cadherin is engaged in dynamic junctions, as in cells undergoing angiogenic responses, VEGF-activated VEGFR2 preferentially stimulates the phospholipase C (PLC) and mitogen-activated protein kinase (MAPK) pathway to signal proliferation [49]. Such signaling can take place from endocytic compartments in which VEGFR2 is internalized in association with VE-cadherin through a clathrin-dependent process [50].

VE-cadherin clustered in stable contacts can modulate endothelial behavior, forming a complex with another regulator of the vascular phenotype, the TGFβ receptor TGFβR2 (as well as with Alk1 and Alk5 [88]). The functional consequence of such local regulation is the activation, by TGFβ, of antiproliferative and antimigratory responses. Such signaling would contribute to the stability of the mature endothelial layer. On the contrary, in endothelial cells with weakened junctions, as is the case after treatment with a VE-cadherin-blocking antibody [88] or with a reduced level of VE-cadherin [58], TGFβR2 is not associated, or is less associated, with VE-cadherin [88]. This deactivates the antiproliferative and antimigratory response elicited by TGFβ, and can even redirect TGFβ signaling towards dedifferentiation (endothelial-mesenchymal transition [EndMT]) of the endothelial cells, as described in more detail in the following paragraphs in CCM pathology.

VE-cadherin has also recently been shown to be associated with fibroblast growth factor receptor 1 (FGFR1) and to reduce its phosphorylation in response to FGF2 engaging the phosphatase Dep1 in a multiprotein complex [28].

In addition to regulating endothelial functions through the local organization of specific signaling complexes to adherens junctions, as described in the previous paragraphs, the ‘degree of tightness’ of VE-cadherin association with junctions appears to regulate the transcriptional profile of endothelial cells [100]. ‘Degree of tightness’ means distinct and specific molecular complexes recruited by VE-cadherin to adherens junctions as a function of the duration and stability of its engagement in an adhesive interaction. In general, the molecular details of the dynamic remodeling of such complexes are still poorly known. Some examples have been presented in the previous paragraphs. Among the interactors of VE-cadherin, β-catenin plays an important role in coordinating the state of the junction compartment to the quality and type of transcription in the nuclear compartment. Although such regulation is evident from a functional point of view, details of the underlying molecular mechanism are still virtually unknown.

Examples of loss and acquisition of differentiated phenotype through the cross-talk between adherens junctions and the nucleus mediated by β-catenin in endothelial cells will be discussed in more detail below; however, before then, we have to face the still unsolved problem of cell biology.