The Integrin Antagonist Cilengitide Activates αVβ3, Disrupts VE-Cadherin Localization at Cell Junctions and Enhances Permeability in Endothelial Cells

Cilengitide is a high-affinity cyclic pentapeptdic αV integrin antagonist previously reported to suppress angiogenesis by inducing anoikis of endothelial cells adhering through αVβ3/αVβ5 integrins. Angiogenic endothelial cells express multiple integrins, in particular those of the β1 family, and little is known on the effect of cilengitide on endothelial cells expressing αVβ3 but adhering through β1 integrins. Through morphological, biochemical, pharmacological and functional approaches we investigated the effect of cilengitide on αVβ3-expressing human umbilical vein endothelial cells (HUVEC) cultured on the β1 ligands fibronectin and collagen I. We show that cilengitide activated cell surface αVβ3, stimulated phosphorylation of FAK (Y397 and Y576/577), Src (S418) and VE-cadherin (Y658 and Y731), redistributed αVβ3 at the cell periphery, caused disappearance of VE-cadherin from cellular junctions, increased the permeability of HUVEC monolayers and detached HUVEC adhering on low-density β1 integrin ligands. Pharmacological inhibition of Src kinase activity fully prevented cilengitide-induced phosphorylation of Src, FAK and VE-cadherin, and redistribution of αVβ3 and VE-cadherin and partially prevented increased permeability, but did not prevent HUVEC detachment from low-density matrices. Taken together, these observations reveal a previously unreported effect of cilengitide on endothelial cells namely its ability to elicit signaling events disrupting VE-cadherin localization at cellular contacts and to increase endothelial monolayer permeability. These effects are potentially relevant to the clinical use of cilengitide as anticancer agent.


Introduction
Endothelial cell -matrix interactions mediated by integrin adhesion receptors play a critical role in vascular development, angiogenesis and vascular homeostasis [1]. Integrins are heterodimeric cell surface complexes formed by non-covalently associated a and b subunits, consisting of large extracellular domains, single transmembrane spanning domains and short cytoplasmic tails. A particular feature of integrins is their tight regulation of ligand binding activity. Transition from a low to a high affinity state (''affinity maturation'') can be induced by intracellular signaling events (''inside-out'' signaling) or by high-affinity ligands [2]. Ligand binding induces allosteric changes in the receptor conformation, leading to the activation of intracellular signaling pathways, including the Ras-MAPK, PI3K-PKB-mTOR and small GTPases (e.g. Rho, Rac) pathways (''outside-in'' signaling) [2]. Since integrins do not possess intrinsic enzymatic activities they require interaction with cytoplasmic adaptor molecules and kinases, including FAK and Src-family kinases, to transduce signaling events. Integrinmediated signaling is critical for the stabilization of cell adhesion and the promotion of cell migration, proliferation and survival [2].
Integrin aVb3 is expressed at low levels on quiescent endothelial cells, while it is strongly induced on angiogenic endothelial cells present in granulation tissue and cancer, and is considered as an attractive therapeutic target to inhibit pathological angiogenesis [3]. Pharmacological inhibition of aVb3 suppresses angiogenesis in many experimental models and aVb3 antagonists (i.e. antibodies, peptides and peptidomimetics) are being developed as antiangiogenic drugs [4]. Cilengitide [5] (EMD121974) is a cyclic Arg-Gly-Asp (RGD)-derived peptide binding with high affinity to aVb3 (IC 50 of 0.6 nM) and inhibiting aVb3 and aVb5dependent adhesion [6]. Cilengitide displays antiangiogenic effects in vitro [7] and in vivo [8][9][10]. It exerts antitumor effects against experimental melanoma and brain tumors [8,9,11,12], it sensitizes endothelial cells to TNF cytotoxicity in vitro [13] and enhances antitumor effects of chemotherapy [14] and radiotherapy [15] in vivo. Cilengitide is in clinical development as anticancer drug. As a single agent it is well-tolerated [16] and shows evidence of durable responses in patients with recurrent gliomas [17,18]. In combination with chemotherapy it showed evidence of activity in pancreas cancers [19] and in highly vascularized head and neck tumors [20]. Cilengitide is now in phase III clinical testing in glioblastoma in combination with radio-and chemotherapy [21].
It is generally assumed that the antiangiogenic activity of cilengitide is due to the inhibition of sprouting and differentiation and the induction of anoikis of angiogenic endothelial cells relaying on aVb3/aVb5 for adhesion and survival [7,22]. However, in addition to aV integrins, angiogenic endothelial cells express multiple integrins, including a1b1, a2b1, a4b1, a5b1, a6b1, and a6b4, which are not targeted by cilengitide [3]. Adhesion through these integrins might compromise the antiangiogenic activity of cilengitide. Little is known on the effect of cilengitide on endothelial cells expressing aVb3/aVb5 but adhering mostly through other integrins, in particular those of the b1 family.
To address this question, we examined the effect of cilengitide on HUVEC, which express aVb3, under condition of b1 integrinmediated adhesion. Here we demonstrate that HUVEC exposure to cilengitide results in the phosphorylation of Src, FAK and VEcadherin, the accumulation of aVb3 at the cell edge, the disappearance of VE-cadherin from cell-cell contacts and the increase in HUVEC monolayer permeability.

Results
Cilengitide causes disappearance of aVb3 from focal adhesions and promotes its accumulation at the cell periphery Cilengitide efficiently inhibits aVb3-mediated cell adhesion and induces detachment of endothelial cells cultured on aVb3 ligands, such as vitronectin or gelatin [13]. To test the effect of cilengitide on endothelial cells expressing aVb3 but mostly using b1 integrins for their adhesion, we seeded HUVEC on fibronectin and collagen I. HUVEC use predominantly a5b1 to adhere to fibronectin (with minor contribution of aVb3) and a1b1/a2b1 to adhere to collagen I [23]. Subsequently we exposed adherent HUVEC to cilengitide at a clinically-relevant concentration (i.e. 10 mM, [17]) or EMD135981, an Arg-Ala-Asp (RAD)-based inactive cyclopeptide. First we monitored the effect of cilengitide on aVb3 localization. In HUVEC plated on fibronectin, aVb3 was present at focal adhesions while b1 integrins clustered at fibrillar adhesions, as previously observed [24]. Cilengitide, but not EMD135981, caused loss of aVb3 from paxillin-positive focal adhesions and promoted the appearance of thin, aVb3-positive and paxillin negative linings at the cell edge (Figure 1, arrows). The localization of b1 integrins at fibrillar adhesions was not perturbed by cilengitide. HUVEC cultured on collagen I showed fewer focal adhesions while had well-developed fibrillar adhesions. Cilengitide treatment induced aVb3 accumulation at the cell border without affecting b1 integrin localization ( Figure 1).

Cilengitide causes VE-cadherin disappearance from cellular junctions
VE-cadherin is a major endothelial cell junctional molecule mediating cell-cell adhesion [25]. It has been previously reported that integrin ligation through multivalent fibronectin-coated beads disrupted VE-cadherin-containing adherens junctions in bovine aortic endothelial cells [26]. We therefore tested, whether aVb3 ligation by monovalent cilengitide affected VE-cadherin localization. In confluent monolayers of HUVEC cultured for 18 hours on fibronectin or collagen I, VE-cadherin was localized at cell-cell contacts (Figure 2a and data not shown). Addition of cilengitide markedly disrupted VE-cadherin localization at cellular junctions, while the EMD135981 peptide was ineffective. Stimulation with VEGF also caused VE-cadherin disappearance from cellular junctions (Figure 2a), consistent with previous reports [27].
Next, we performed aVb3 and VE-cadherin co-staining experiments to monitor the spatial relationship between the appearance of aVb3 at the cell periphery and loss of VE cadherin from cell-cell junctions. In confluent HUVEC cultures VE-cadherin and aVb3 were localized at different locations (cell-cell contacts and focal adhesions, respectively) (Figure 2b, control). Upon stimulation with cilengitide, VE-cadherin staining became discontinuous and aVb3 appeared at cell borders, typically at sites where VE-cadherin disappeared from cellular contacts (Figure 2b, time course). Paralleling loss of VE-cadherin from cell-cell junctions, 'gaps' appeared in the monolayer (Figure 2b, asterisks), consistent with diminished cell-cell adhesion and cell retraction. Concomitant presence of VE-cadherin and aVb3 at cell-cell contacts was very rarely observed (Figure 2c, arrowheads), suggesting that co-localization of aVb3 and VE-cadherin is a rather mutually exclusive event.
Taken together, these results indicate that exposure of confluent HUVEC to cilengitide while cultured on fibronectin or collagen I, resulted in the redistribution of aVb3 from focal adhesions to the cell periphery and the concomitant disappearance of VE-cadherin from cellular junctions.

Cilengitide activates cell surface aVb3 integrin
The disappearance of VE-cadherin from cell-cell contacts suggested that cilengitide-bound integrin aVb3 might initiate intracellular signaling events by activating aVb3. To test this hypothesis we monitored the capacity of cilengitide to induce affinity maturation of aVb3 on endothelial cells using antibodies (i.e. LIBS-1 and CRC54) recognizing ligand-induced binding sites (LIBS) on b3 integrins [28,29]. Cilengitide, but not EMD135981, induced LIBS-1 and CRC54 epitope expression on HUVEC in suspension, without altering total cell surface levels of aVb3 as detected by LM609 mAb (Figure 3a). Addition of MnCl 2 , a known integrin activator, also induced LIBS-1 and CRC54 epitope expression (data not shown) as previously reported [24]. Cilengitide had no effect on b1 LIBS expression as detected by mAb HUTS-21 (data not shown). To test whether cilengitide-induced affinity maturation also occurred on adherent HUVEC, we exposed fibronectin-adherent HUVEC to cilengitide, EMD135981 or MnCl 2 and stained them with CRC54 (LIBS-1 mAb does not work on fixed cells) and LM609. In unstimulated HUVEC, focal adhesions were positive for CRC54, consistent with the ligated/active state of aVb3 ( Figure 3b). Upon cilengitide stimulation we observed CRC54-positive patches at the cell periphery, consistent with a cilengitide-ligated (activated) state (Figure 3b, arrows). In comparison, MnCl 2 treatment enhanced aVb3 clustering and expression of the CRC54 epitope at focal adhesions as already reported [24].

Cilengitide induces Src and FAK phosphorylation
Next we sought after evidence for cilengitide-induced intracellular signaling events. Src-dependent phosphorylation of focal adhesion kinase (FAK) is one of the first signaling events initiated by integrin activation [2,3]. Src, like other Src family kinases, is negatively regulated though the phosphorylation of a carboxylterminal tyrosine residue (Y 529 in human Src). This phosphorylation forces the Src C-terminal domain to interact with the SH2 and SH3 domains, thus forming a loop that masks the Src kinase domain [30]. Disruption of this loop, achieved through protein tyrosine phosphatases (i.e. PTPa, PTPIB, Shp2) -mediated dephosphorylation of Y 529 , or via integrin clustering in the absence of Y 529 dephosphorylation [31], allows Src to interact with its substrates via SH2 and SH3 domains. Cilengitide treatment of confluent HUVEC, increased Src phosphorylation of tyrosine residue Y 419 without decreasing phosphorylation of Y 529 (Figure 4a), consistent with integrin-mediated Src activation, and promoted FAK phosphorylation at tyrosine residues Y 576 and Y 577 , two well-characterized phosphoacceptor sites of Src [32], Cilengitide induces VE-cadherin phosphorylation at residues Y 658 and Y 731 Src was shown to phosphorylate the VE-cadherin cytoplasmic domain in response to VEGF stimulation [34]. We therefore tested whether cilengitide induced Src activation resulted in the phosphorylation of the VE-cadherin cytoplasmic domain. Cilengitide treatment induced VE-cadherin phosphorylation at Y 658 and Y 731 , which correspond to the binding sites for p120 catenin and b-catenin, respectively [35]. Addition of CGP77675 (2.5 mM) strongly reduced basal and cilengitide-induced phosphorylation of both residues ( Figure 5a). As reported, VEGF stimulation induced Y 658 phosphorylation and to a lesser extent Y 731 phosphorylation, which were also inhibited by CGP77675. In contrast to VEGF, however, cilengitide did not induce phosphorylation of MEK 1/2, Akt, and Ik-B, (Figure 5b and data not shown). Next we tested the effect of inhibition of Src kinase activity on the recruitment of aVb3 to the cell periphery and the disappearance of VE-cadherin from cell junctions. Indeed, CGP77675 prevented the formation of aVb3 patches at the cell edge in HUVEC plated on fibronectin or collagen I at both sub-confluent and confluent conditions, and attenuated the disappearance of VE-cadherin from cell-cell contacts induced by cilengitide ( Figure 6).
Taken together these results establish that cilengitide induces aVb3 affinity maturation, and initiates signaling events in endothelial cells leading to phosphorylation of Src, FAK and VE-cadherin. These phosphorylation events, recruitment of aVb3 at the cell periphery and disappearance of VE-cadherin from cellular contacts requires Src kinase activity.
Cilengitide enhances HUVEC monolayer permeability VE-cadherin-mediated cell-cell adhesion and integrin-mediated cell-matrix adhesion are essential for maintaining endothelial cell monolayer tightness [36,37]. Based on the above observations, we set up to test whether cilengitide treatment increased permeability of confluent HUVEC. Addition of cilengitide (10 mM) to HUVEC cultured on fibronectin or collagen-coated filter inserts, resulted in a time-dependent increase in transendothelial permeability ( Figure 7a). Microscopic examination of the filters at the end of Figure 1. Cilengitide causes loss of aVb3 from focal adhesions and promotes appearance of aVb3 patches at the cell edge. HUVEC were plated on coverslips coated with fibronectin or collagen I and were treated with 10 mM of cilengitide for 20 minutes. The localization of the aVb3 or b1 integrin (green) and paxillin (red) were monitored by immunofluorescence staining. In HUVEC plated on fibronectin aVb3 was present at focal adhesions, while b1 was present at fibrillar adhesions. Cilengitide, but not EMD 135981, caused loss of aVb3 from focal adhesions and appearance of aVb3-positive thin patches at the cell edge (arrows). b1 localization was not altered by cilengitide. A similar effect on aVb3 (arrows) was observed on cells plated on collagen I, with the difference that focal adhesions were less abundant on this matrix. Optical magnification: 4006; Bar: 10 mm. (n = 5). doi:10.1371/journal.pone.0004449.g001 the assay (240 minutes) revealed that cilengitide induced morphological changes to the cultures, in particular the appearance of dark (dense) dendritic-like cells, consistent with cells that retracted or detached from the substrate (Figure 7b, arrows). CGP77675 (2.5 mM) partially abolished cilengitide-induced increased permeability but was ineffective in preventing the appearance of retracted cells (Figure 7a and 7b). As expected treatment of HUVEC cultured on vitronectin-coated filters resulted in massive cell detachment and increased permeability, consistent with aVb3/aVb5-mediated adhesion to this substrate (data not shown).

Cilengitide interferes with b1 integrin-dependent HUVEC attachment on low-density ligands
The appearance of retracted HUVEC in cilengitide-treated filter inserts during the permeability assays on fibronectin and collagen I, suggested the possibility that cilengitide might interfere with adhesion on fibronectin or collagen I. Activation of one individual integrin was previously shown to interfere with the function of other integrins though a transdominant negative effect [38]. To test whether cilengitide-induced aVb3 activation might interfere with b1 integrin-mediated adhesion, we first performed (a) Confluent HUVEC plated on fibronectin, were exposed to cilengitide or EMD135981 (10 mM each) or VEGF (100 ng/ml) for 20 minutes and stained for VE-cadherin. Cilengitide and VEGF treatments disrupted VE-cadherin localization at cellular junctions, while EMD135981 showed no effect (n = 3). Optical magnification: 4006; Bar: 10 mm. (b) Confluent HUVEC plated on fibronectin or collagen I were exposed to cilengitide (10 mM each) for the indicated time and double stained for VE-cadherin and b3 integrin. Cilengitide disrupted VE-cadherin staining and promoted appearance of b3 at VE-cadherin-depleted cell-cell borders (arrows). Paralleling loss of VE-cadherin from cell-cell junctions, 'gaps' appeared in the monolayer (asterisks). short-term adhesion assays on vitronectin (as aVb3 ligand), fibronectin (as mixed a5b1.aVb3 ligand) or collagen I (as b1 ligand). Since the transdominant negative effect is based on the competition for intracellular adaptor and signaling molecules among unclasped cytoplasmic b tails of active integrins [39], and the stoechiometry of active (i.e. aVb3) vs target (i.e. b1) integrins is critical, we tested the effect of cilengitide on HUVEC engaging decreasing levels of b1 integrins by coating decreasing concentrations of ligands. Cilengitide prevented aVb3-mediated HUVEC adhesion to vitronectin at any coating concentrations, consistent with a direct inhibition of aVb3 ligand binding activity (Figure 8a). Cilengitide showed no effect on b1-mediated HUVEC adhesion on fibronectin and collagen I coated at high concentrations, while it interfered with HUVEC adhesion to low ligand concentrations ( Figure 8a). To test the effects of cilengitide on cells already attached, we added cilengitide to HUVEC cultured for 18 hours in wells coated with graded amounts of vitronectin, fibronectin or collagen I. Cilengitide induced detachment of HUVEC cultured on vitronectin regardless of the coating concentration, while it detached HUVEC from fibronectin and collagen I only in wells coated with low protein concentrations (Figure 8b). Addition of CGP77675 did not abolish the anti-adhesive effect of cilengitide on HUVEC plated on low-density fibronectin or collagen I in a shortterm adhesion assay (Figure 8c). Sub G1 DNA content analysis of control and treated cultures revealed an increased frequency of Sub G1 DNA containing-cells in wells coated with fibronectin or collagen I at low densities and exposed to cilengitide, consistent with detachment-induced death (Table 1) [40].
Taken together, these results demonstrate that cilengitide interferes with b1-mediated adhesion under conditions of limited b1-substrate concentration and limited b1 engagement. This effect is independent of Src activity and is consistent with a b3 to b1 transdominant negative effect.

Discussion
The antiangiogenic activity of cilengitide is largely attributed to its ability to directly interfere with aVb3/aVb5-mediated adhesion of angiogenic endothelial cells, thereby inducing cell detachmentmediated death (anoikis) of cells relying on these integrins for adhesion and survival [5,40]. Angiogenic endothelial cells, however, in addition to aVb3 and aVb5, express other integrins, in particular a1b1, a2b1, a4b1, a5b1, a6b1, and a6b4, which are not targeted by cilengitide [3]. Based on the current model of action of cilengitide, endothelial cells expressing and using these integrins would be insensitive to cilengitide effects. The present work was initiated to test whether endothelial cells expressing aVb3, but predominantly using b1 integrins for adhesion, are insensitive to cilengitide, or whether they may indeed show some effects. Here we report five effects of cilengitide under such conditions: i) affinity maturation of aVb3 and its accumulation to the cell periphery; ii) phosphorylation of Src (Y 419 ), FAK (Y 397 and Y 576/577 ) and VEcadherin (Y 658 and Y 731 ); iii) disappearance of VE-cadherin from cell-cell contacts; iv) detachment of HUVEC cultured on lowdensity b1 substrates; v) increased HUVEC monolayer permeabil-ity. These findings unravel a more complex picture of the mechanistic effects of cilengitide on endothelial cells and, more generally, highlight the role of integrins and integrin-induced signaling events in the regulation of endothelial cell functions (See Figure 9 for a working model).
There is structural evidence that high affinity RGD-based cyclopeptides, including cilengitide, can induce large-scale conformational changes of soluble truncated integrins consistent with ligand-induced activation [41,42]. RGD-based soluble ligands were shown to induce some signaling events, such as intracellular calcium mobilization in neurons, smooth muscle cells and T lymphocytes [43,44], or protein kinase C activation in oocytes [45]. Our work extends these findings by demonstrating that a monomeric, high-affinity RGD-based ligand induces affinity maturation (e.g. activation) of cell surface aVb3 leading to Src, FAK and VE-cadherin phosphorylation. The mechanism by which cilengitide elicits these signaling events remain to be determined. Current knowledge implies integrin clustering induced by immobilized or soluble-multivalent ligands as an essential step to recruit adaptor proteins or kinases and initiate signaling events. Our results demonstrate that a monovalent highaffinity ligand is nevertheless sufficient to elicit some signaling events. A plausible explanation is that since Src is constitutively associated with the cytoplasmic domain of the b3 subunit, cilengitide-induced aVb3 'outside-in' activation and unclasping of the aVb3 cytoplasmic domains is sufficient to activate Src. Active Src can then complex with FAK resulting in mutual Src-FAK phosphorylation promoting lateral association of cilengitideoccupied aVb3 through its SH3 domain, resulting in patches formation at the cell periphery [46]. aVb3 activation by cilengitide, however, appears insufficient to fully activate downstream signaling pathways, such as ERK1/2, NF-kB or PI3K/Akt probably due to the lack of additional adaptor and signaling proteins normally present at focal adhesions [3]. The mutually exclusive localization of aVb3 and VE-cadherin at cellular junctions in confluent HUVEC suggests that a causal link between cilengitide-induced aVb3 and Src activation and the disappearance of VE-cadherin from cell-cell contacts. In our model activated aVb3 recruiting at cell-cell junctions brings active Src to VE-cadherin-catenin complexes, thereby promoting VEcadherin phosphorylation at residues Y 658 and Y 731 , dissociation from band p120 catenins and disappearance from cell-cell contacts, consistent with published results [26,35].
The observed increased permeability induced by cilengitide is consistent with phosphorylation of VE-cadherin and disappearance from cell-cell contacts. However, in contrast to phosphorylation and displacement of VE-cadherin, which could be effectively prevented by pharmacological inhibition of Src, cilengitide-induced permeability was only partially prevented by Src inhibition. This partial effect of CGP77675 is likely due to the fact that cilengitide-ligated and activated aVb3 exerts a transdominant negative on b1 integrins insensitive to Src inhibition. On low matrix densities the transdominant negative effect results in decreased cell adhesion and increased cell detachment. On highmatrix densities the same effect is insufficient to detach cells, but (a) HUVEC in suspension were exposed to 10 mM of cilengitide or EMD135981 for 10 minutes, stained by immunofluorescence for b3 LIBS and total aVb3 expression (with LIBS-1 and LM609 mAbs, respectively) and analyzed by flow cytometry. Cilengitide, but not EMD135981, induced LIBS expression (left histograms, thick lines), without affecting total aVb3 expression (right histograms, thick lines). Dotted lines: cellular fluorescence in the absence of primary antibody. (n = 3). (b) Fibronectin-adherent HUVEC were exposed to 10 mM cilengitide, EMD135981, or 1 mM MnCl 2 for 10 minutes, stained for b3 LIBS and total aVb3 expression (with CRC54 or LM609 mAbs, respectively) and analyzed by immunofluorescence microscopy. Total aVb3 and b3 LIBS were present at focal adhesions in unstimulated HUVEC and at tiny patches at the cell edge in cilengitide-exposed HUVEC, thus confirming that aVb3-positive patches contain active aVb3. MnCl 2 stimulated recruitment and activation of aVb3 at focal adhesions. (n = 2). Optical magnification: 4006; Bars: 10 mm. doi:10.1371/journal.pone.0004449.g003 might cause cellular retraction as observed at early time points (5-30 minutes after addition of cilengitide) in HUVEC cultured on plastic wells or partial cell detachment as observed for HUVEC cultured on filters at the end of the permeability assays. Although a role for integrins in controlling vascular permeability has been proposed before [47,48], the molecular mechanisms involved remained elusive. Of interest, in a recent report the extracellular matrix protein big-h3/TGFBI was shown to increase vascular permeability in a Src-dependent manner by binding to aVb5 and causing dissociation of VE-cadherin from endothelial junctions [49]. Taken together our report extends these observations, by demonstrating cilengitide-induced increased vascular permeability of HUVEC monolayers though combined Src-dependent disruption of VE-cadherin localization at cell-cell contacts, and Srcindependent cell retraction consistent with a transdominant negative effect on b1 integrins.
These observations raise a number of questions that need to be addressed in future studies. One question relates to the potential role of aVb5 (the second integrin targeted by cilengitide) in these effects. On HUVEC, aVb5 is likely not to play a significant role since it is Figure 4. Cilengitide induces Src and FAK phosphorylation. (a) Western blotting analysis of Src phosphorylation at Y 529 and Y 419 and total Src in HUVEC grown on fibronectin and exposed for 10 minutes to EMD135981, cilengitide (10 mM each) and CGP77675 (2.5 mM) as indicated. Cilengitide increased Src phosphorylation at Y 419 but did not alter Y 529 phosphorylation. CGP77675 prevented Y 419 phosphorylation. (b) Western blotting analysis of the same cells as in panel a, but for phosphorylation of FAK at Y 397 and Y 576 and total FAK. Cilengitide increased FAK phosphorylation at both tyrosine residues and this was inhibited by CGP77675. EMD135981 had no effect on Src or FAK phosphorylation. Actin was detected to demonstrate equal loading of the lanes. The bar graph represents the relative level of phospho Src/FAK over total Src/FAK as determined by band density analysis. (n = 3). doi:10.1371/journal.pone.0004449.g004  expressed at much lower levels compared to aVb3 [37] (and data not shown). In vivo, however, the situation might be different as aVb5 is also highly expressed on angiogenic endothelia and its ligation was shown to promote increased vascular permeability in response to angiogenic growth factors [50]. A second question relates to the mechanism of Src activation by cilengitide-bound aVb3, and the contribution of the a and b subunit cytoplasmic domains since both domains can bind Src [3]. Another question relates to the role that soluble high-affinity ligands binding to non matrix-ligated integrins might exert on endothelial cell functions. In this perspective, this work extends previous observations demonstrating that soluble integrins ligand can induce COX-2 mRNA and protein expression [23]. This is of particular interest, since many circulating plasma proteins are natural integrin ligands (e.g. fibronectin, vitronectin, fibrinogen) and their binding to luminal integrins (i.e. not engaged in cell-matrix adhesion) may elicit important, yet largely uncharacterized, regulatory signals. A last important question is whether the permeability-promoting effect of cilengitide observed in this study may have therapeutic implications. Cilengitide is currently in clinical development in oncology. Phase I and II clinical studies have demonstrated that it is well tolerated (no dose-limiting toxicities were observed) and provided initial evidence of activity. Phase III studies in combination with chemotherapy and radiotherapy are underway in glioblastoma multiforme [21]. Cilengitide might be a particularly well-suited drug to combine with chemotherapeutic agents with the purpose to improve chemotherapy delivery to tumors, which is one of the limiting events in cancer therapies.

Cells and culture media, antibodies and reagents
Cells and culture media: HUVEC cells were prepared and cultured as described previously [37] and were used between passage 3 and 5. Antibodies

Immunofluorescence microscopy
HUVEC cells were cultured subconfluent or at confluency on glass coverslips pre-coated with fibronectin (3 mg/ml) or collagen I (1 mg/ml) placed in 12 well plates in complete M199 medium. After 16 hours, cultures were stimulated with cilengitide or EMD135981 (10 mM each) or VEGF (100 ng/ml) for 20 minutes, or otherwise at the indicated times. To inhibit Src, CGP77675 (2.5 mM) was added 15 minutes before addition of cilengitide. Cultures were then immediately fixed in 4% PFA for 10 minutes at room temperature, permeabilized with 0.1% Triton X-100, blocked with 1% BSA and incubated for 1 hour with the relevant primary antibodies (5 mg/ml). After washing, cells were incubated with a Cy5 or FITC-conjugated secondary antibody. DAPI was used to counterstain nuclei. Stained cells were mounted in Prolong Antifade medium (Molecular Probes, Invitrogen) and viewed by epifluorescence microscopy (Axioplan with objective EC Plan Neofluar 406/1.3 oil ph 3, Carl Zeiss AG, Zürich). Images were acquired with an Axiocam camera (Carl Zeiss AG) and the Axiovision program (release 4.7, Carl Zeiss AG) and processed (zooming, gamma and contrast adjustments) with Adobe Photoshop CS3 (Adobe Systems Inc. San Jose, CA)

Quantification of VE-cadherin immunostaining
Quantification of fluorescence of VE-cadherin staining was performed with the program Metamorph 7.5 (Molecular Devices, Downingtown, PA). Briefly, a program script was defined to draw a line region along the plasma membrane of each cell on the images (magnification 406). Then, segment regions (i.e. squares of 2.4 mm66.4 mm -length6width) were created along the line and, in each segment region, the fluorescence was measured according to a threshold defined from a negative control. The measured fluorescence average intensities were then normalized within each segment by multiplying the fluorescence average intensity by the ratio of the threshold area divided by the total area. These normalized intensities (NI) were then arbitrarily divided in two groups: group 0 for NI,5; group 1 for 5,NI, corresponding to absent or faint vs intense labeling, respectively. On each image Figure 8. Cilengitide interferes with HUVEC adhesion on low-density b1 integrin substrates. (a) HUVEC short-term adhesion assays performed on vitronectin (aVb3 ligand), fibronectin (a5b1.aVb3 ligand) and collagen I (a1b1/a2b1 ligand) coated at the indicated concentrations, in medium only (black bars) or in the presence of EMD135981 (gray bars) or cilengitide (white bars). On vitronectin, cilengitide inhibited adhesion at all coating concentrations while on fibronectin and collagen I it blocked adhesion only at low coating concentrations. (n = 5). (b) HUVEC detachment assays. HUVEC were cultured for 18 hours on vitronectin, fibronectin or collagen I coated at the indicated concentrations, to allow for full attachment, before exposure for 4 hours to medium only (black bars), EMD135981 (gray bars) or cilengitide (white bars). Cilengitide detached HUVEC cultured on vitronectin at all coating concentrations, while it induced HUVEC detachment from fibronectin and collagen only at low coating concentrations (Triplicate wells/condition, n = 3). (c) HUVEC short-term adhesion assays performed on fibronectin and collagen I coated at the indicated concentrations, in medium only (black bars), in the presence of CGP77675 (white bars), cilengitide (gray bars), cilengitide+CGP77675 (hatched bars). Src inhibition did not prevent cilengitide-induced inhibition of cell adhesion on low matrix concentrations. (Triplicate wells/condition, n = 2). Attached cells were quantified by Crystal Violet staining and OD determination at 540 nm wavelength. Asterisks indicate statistical significant differences of the values relative to untreated controls (p,0.05). doi:10.1371/journal.pone.0004449.g008 Table 1. Relative cell death of HUVEC cultures exposed to cilengitide or EMD135981. these groups were counted and plotted as shown in the Figure 6b. Quantification was performed on 30 cells per conditions.

SDS-PAGE and Western blotting
HUVEC cells were treated with the different compounds as described in the text or figure legends. Total cell lysates (20 ml per lane) were resolved by 7.5% SDS-PAGE and blotted onto Immobilon-P membranes (Millipore, Volketswil, Switzerland). Membranes were blocked with 5% BSA prepared in 16 TBS-T. Primary antibodies were prepared as 1:2000 dilutions in 16 TBS-T with 3% BSA and added to the membrane overnight at 4uC. Phosphoproteins were always detected first prior stripping the membranes to detect total proteins. Membranes were then washed and processed for HRP-coupled secondary antibody according to standard laboratory protocols and the ECL system was used for detection (Amersham-Pharmacia Biotech, Dübendorf, Switzerland). Membranes were reprobed for actin to assess equal loading of the samples. To compare expression levels of the different proteins, revealed Western blots were scanned and band pixelization for total and phosphorylated proteins were analyzed with the AIDA bioimaging software (raytest Isotopenmessgeraete GmbH, Straubenhardt, Germany). The number of pixels of each individual phosphoprotein bands was normalized to the corresponding total protein band pixel numbers.
Adhesion and detachment assays HUVEC were seeded in triplicate in 96-well plates (2610 4 cells/ well) pre-coated with graded amounts of vitronectin (1-0.1 mg/ml), fibronectin (3-0.1 mg/ml), collagen I (1-0.01 mg/ml). After coating, the wells were saturated with 1% BSA. For short-term adhesion assays, cells were seeded in serum-free medium in the absence or presence of 10 mM cilengitide or EMD135981 and adhesion quantified after 2 hours. For survival assays, the cells were seeded in complete media and left to adhere for 16 hours followed by a 3 hours starvation in serum-free medium before addition of 10 mM cilengitide or EMD135981 for another 2 hours. To inhibit Src, CGP77675 (2.5 mM) was added 15 minutes before addition of cilengitide. At the end of the assay period, cultures were rinsed with two gentle washes with PBS (with Ca/Mg), fixed for 15 minutes with 4% paraformaldehyde and stained for 15 minutes with 0.5% Crystal Violet. Stained wells were washed and CV was extracted with 100 ml CV distain solution (29.4 g/l Na 3 -citrate in 50% ethanol) and absorbance measured at 540 nm wavelength.

Cell viability assay
Sub-G1 DNA content assay was performed as described [51]. Briefly, HUVEC were collected as above, resuspended in 70% ice cold ethanol under vortex and incubated for 2 hours at 220uC. Cells were recovered by centrifugation and resuspended in PBS. 50 mg/ml RNase A (Roche, Basel) was added and samples were incubated at room temperature for 5 minutes before staining with propidium iodide (PI, 50 mg/ml) for 30 minutes at 37uC. Stained cells were analyzed with a FACScan IIH and Cell QuestH software (Becton Dickinson, Mountain View, CA).

In vitro permeability assay
The assay was adapted from [52]. Briefly, HUVEC were seeded, in triplicate, at a density of 40610 3 cells on polystyrene filter inserts (3 mm pore size, BD Biosciences, Basel, Switzerland, catalogue number 353096) pre-coated with fibronectin (3 mg/ml) or collagen I (1 mg/ml) or vitronectin (1 mg/ml), in 12-well plates in a total volume of 200 ml and 1 ml of complete M199 for the upper and lower chambers, respectively. After 20 hours, the medium in the upper chamber was gently exchanged with fresh one containing the paracellular permeability tracer molecule FITC-dextran [53] (av. M r 40610 3 , Sigma-Aldrich, Basel, Switzerland, catalogue number FD40S, 0.5 mg/ml final concentration) and either nothing else (control), CGP77675 (final concentration 2.5 mM), cilengitide (final concentration 10 mM) or both compounds. CGP77675 (2.5 mM) was added 15 minutes before addition of cilengitide. At given time points, 50 ml aliquots from the lower chamber were removed for measurement and replaced with 50 ml of fresh medium in order to maintain the hydrostatic equilibrium. The fluorescence of each sample diluted (1:20) in PBS, was measured at 485/530 nm excitation/emission wavelengths. The zero time point (t = 0) was defined by diluting 50 ml (1:20) in PBS. After the last time point, wells were fixed with 100 ml 4% PFA, stained by Crystal Violet and photographed (Axiovert 40 CFL, Carl Zeiss AG, Zürich, Switzerland).

Statistical analysis
Results are expressed as mean6s.d.. Data were analyzed by Student's t-test for. P values,0.05 were considered significant.