The protein tyrosine phosphatase TCPTP controls VEGFR2 signalling.

Vascular endothelial growth factor (VEGF) is a major angiogenic factor that triggers formation of new vessels under physiological and pathological conditions. However, the mechanisms that limit the VEGF responses in target cells and hence prevent excessive and harmful angiogenesis are not well understood. Here, our objective was to study whether T-cell protein tyrosine phosphatase (TCPTP, also known as PTN2), which we found to be expressed in human endothelial cells, could alter VEGF signalling by controlling phosphorylation of VEGFR2. We show that a TCPTP substrate-trapping mutant interacts with VEGFR2. Moreover, TCPTP dephosphorylates VEGFR2 in a phosphosite-specific manner, inhibits its kinase activity and prevents its internalization from the cell surface. We found that TCPTP activity is induced upon integrin-mediated binding of endothelial cells to collagen matrix. TCPTP activation was also induced by using cell-permeable peptides from the cytoplasmic tail of the collagen-binding integrin alpha1. Controlled activation of TCPTP results in inhibition of VEGF-triggered endothelial cell proliferation, angiogenic sprouting, chemokinesis and chemotaxis. We conclude that matrix-controlled TCPTP phosphatase activity can inhibit VEGFR2 signalling, and the growth, migration and differentiation of human endothelial cells.


Introduction
VEGF is a major growth factor for endothelial cells. It regulates endothelial cell function in the formation of new blood vessels both during development and in adults (e.g. during wound healing and in proliferative diseases such as cancer) (Coultas et al., 2005). In the vascular endothelium, VEGF binds to two high-affinity tyrosinekinase receptors, VEGFR1 and VEGFR2. VEGFR2 appears to be the main mediator of VEGF-induced endothelial proliferation, survival, migration, tubular morphogenesis and sprouting in adults (Ferrara et al., 2003). The effects of VEGF are mainly controlled by its availability and by the activity of VEGFR signalling. Upon binding of VEGF, the receptors dimerize, become autophosphorylated, and recruit signal-transduction molecules such as PLC, PI3K and Ras (Olsson et al., 2006;Kerbel, 2008;Siekmann et al., 2008). By contrast, negative regulation of VEGFR2 signalling is equally important in limiting the response of VEGF in target cells. In fact, inhibition of neovascularization is a promising new way of treating diseases such as cancer and certain retinopathies (Carmeliet, 2005).
Angiogenesis is a multistep process that involves basementmembrane degradation by the endothelial cells, sprouting and migration, proliferation, structural organization, synthesis of new basement membrane, and returning to quiescence (Carmeliet, 2000;Yancopoulos et al., 2000;Avraamides et al., 2008). During the different steps of angiogenesis, the endothelial cell encounters dramatically different extracellular matrices, ranging from the mature basement membrane below the quiescent endothelial cells to the type-I-collagen-rich stroma around the sprouting cells.
Integrin-mediated adhesion to the extracellular matrix (ECM) provides permissive signals to cells by supporting migration, cell survival, and signalling by receptor tyrosine kinases (RTKs) (Giancotti, 1997;Larsen et al., 2006). Different integrins have the capacity to form complexes with growth-factor receptors on the cell membrane, and regulate their activation and internalization. For example, integrin-mediated adhesion alone can activate growthfactor-receptor phosphorylation in the absence of ligands (Moro et al., 1998) and integrins bind directly to growth-factor-receptors such as EGFR and MET (Moro et al., 1998;Trusolino et al., 2001). Adhesion of the endothelial cells to ECM molecules also profoundly affects their angiogenetic properties (Avraamides et al., 2008). For instance, binding of endothelial cells to vitronectin by αvβ3 integrin positively regulates VEGFR2 activation via direct interaction between the two molecules (Borges et al., 2000), and expression of active integrin αvβ3 induces secretion of VEGF (De et al., 2005). More recently, integrin α9β1 expressed in endothelial cells was shown to cooperate with VEGFR2 in angiogenesis (Vlahakis et al., 2007). It has also been known for a long time that collagenous matrix can inhibit endothelial growth (Form et al., 1986). However, the molecular nature of the matrix-dependent integrin-mediated signals that inhibit angiogenesis is not fully understood.
We recently reported that a collagen-binding integrin, α1β1, negatively regulates EGFR phosphorylation in malignant epithelial cells. This was the first demonstration of an inhibitory role for the ECM in signalling by RTKs. The effect of α1 integrin was mediated via coupling to the phosphatase TCPTP (Mattila et al., 2005). TCPTP is a 45-kDa protein that, despite its name, is expressed in Vascular endothelial growth factor (VEGF) is a major angiogenic factor that triggers formation of new vessels under physiological and pathological conditions. However, the mechanisms that limit the VEGF responses in target cells and hence prevent excessive and harmful angiogenesis are not well understood. Here, our objective was to study whether T-cell protein tyrosine phosphatase (TCPTP, also known as PTN2), which we found to be expressed in human endothelial cells, could alter VEGF signalling by controlling phosphorylation of VEGFR2. We show that a TCPTP substrate-trapping mutant interacts with VEGFR2. Moreover, TCPTP dephosphorylates VEGFR2 in a phosphosite-specific manner, inhibits its kinase activity and prevents its internalization from the cell surface. We found that TCPTP activity is induced upon integrinmediated binding of endothelial cells to collagen matrix. TCPTP activation was also induced by using cell-permeable peptides from the cytoplasmic tail of the collagen-binding integrin α1. Controlled activation of TCPTP results in inhibition of VEGFtriggered endothelial cell proliferation, angiogenic sprouting, chemokinesis and chemotaxis. We conclude that matrixcontrolled TCPTP phosphatase activity can inhibit VEGFR2 signalling, and the growth, migration and differentiation of human endothelial cells.
several cell types (Alonso et al., 2004). TCPTP is prominently present in the nucleus, but it is capable of translocating to the cytoplasm in response to mitogenic stimuli or α1β1-integrinmediated adhesion to collagen (Tiganis et al., 1998;Mattila et al., 2005). We studied here whether TCPTP would be present in endothelial cells and whether ligation of collagen-binding integrins or TCPTP activation could have biological effects in endothelial cells. We found that TCPTP dephosphorylates and silences VEGFR2, and inhibits biological responses to VEGF. We also devised a novel peptide from the cytoplasmic domain of integrin α1 that can be used as a tool to restrain VEGFR2 signalling via the activation of TCPTP. These data suggest that endothelial cell growth during different steps of angiogenesis can be regulated by microenvironmentally controlled phosphatase activities.

TCPTP is present in human endothelial cells
We hypothesized that human endothelial cells might express TCPTP and that the collagen-binding integrin α1 could regulate the activity of TCPTP in these cells. We found α1β1 integrin in early-passage human umbilical vein endothelial cells (HUVECs) by immunoblotting (Fig. 1A). FACS analyses further showed that α1 integrin is expressed on the surface of HUVECs (supplementary material Fig. S1A). Immunoblotting experiments clearly revealed that TCPTP protein is present in HUVECs (Fig. 1A). In fact, the amount of TCPTP protein in these endothelial cells was almost comparable to that seen in T lymphocytes in peripheral blood (Fig.  1B). Immunocytochemistry further showed that TCPTP is present both at the membrane and in the cytoplasm in HUVECs. TCPTP staining was specific, because it could be competed out with excess of recombinant soluble TCPTP (Fig. 1C).
TCPTP was also found in intact human endothelium in vivo. In frozen sections of human umbilical cords, TCPTP, α1 integrin and VEGFR2 were all found to be expressed in the endothelial layer of the veins (Fig. 1D). Also, in inflamed tonsils, specific staining with anti-TCPTP antibody was evident in endothelial cells (supplementary material Fig. S1B). In tonsils, the intensity of TCPTP staining in endothelial cells was lower than that in many resident leukocytes, but it clearly exceeded that of some leukocytes and many other cell types. To our knowledge, this is the first indication of TCPTP expression in the vascular endothelium and suggests that TCPTP plays a role in vascular biology.

TCPTP activity regulates responsiveness of endothelial cells to VEGF
To study the possible biological effects of TCPTP activity in endothelial cells, we silenced or induced TCPTP activity and studied the behaviour of HUVECs in two different models. For cellproliferation experiments, we silenced endogenous TCPTP in HUVECs using siRNA. Silencing of TCPTP was efficient and it led to a modest, but statistically significant, increase (28±6%, P<0.05) in VEGF-induced proliferation when compared with control cells (Fig. 2A). No difference in proliferation was seen between TCPTP-silenced and control cells in the absence of exogenous VEGF (not shown).
Wild-type 45-kDa TCPTP protein (TC45) exists in a partially auto-inhibited state involving the C terminus of the protein, and mitogenic stimulus or interaction with α1β1 integrin is needed for full activity (Hao et al., 1997;Mattila et al., 2005). To circumvent the need for TCPTP activation, we used constitutively active 37-kDa TCPTP (TC37), which is a truncation mutant lacking the C-terminal part of the 45-kDa molecule (Hao et al., 1997). The effect of enhanced TCPTP activity on motility of HUVECs was then analyzed in cell-migration assays. We found that chemotaxis of vector-transfected HUVECs was stimulated by VEGF by more than twofold. Strikingly, expression of constitutively active TCPTP rendered the cells completely unable to migrate towards VEGF (Fig.  2B). Together, these functional studies showed that decreased TCPTP activity enhances and increased TCPTP activity inhibits (A) Immunoblotting of TCPTP, α1 and β1 integrins, and VEGFRs from HUVEC lysates. HeLa cell lysate is a positive control for expression of integrins and TCPTP, and tubulin blots serve as loading controls. Shown is a representative blot out of three, which each had similar results. (Β) Immunoblotting of HUVECs and peripheral blood mononuclear cells (PBMCs, mainly consisting of T lymphocytes) isolated from different donors (numbers above the lanes) for TCPTP. (C) HUVECs grown on coverslips were stained for TCPTP in the absence (control) or presence of recombinant TCPTP (TC45). Note the specific TCPTP expression in the cytoplasm and membrane of the control HUVECs, which is competed out by pre-incubation of the antibody with the soluble target antigen. Scale bars: 10 μm. (D) Two-colour immunofluorescence stainings of human umbilical cords for TCPTP, α1 integrin and VEGFR2, and control IgGs. Individual channels and merged images with the indicated antibodies from an umbilical vein are shown. DAPI stainings reveal the nuclei of the endothelial lining. Scale bars: 20 μm. E, endothelial lining; L, lumen. VEGF-driven responses in primary endothelial cells. This led us to consider the possibility that TCPTP could regulate endothelial cell physiology by dephosphorylating VEGF receptors.

TCPTP binds to VEGFR2
To test whether TCPTP and VEGFR2 could interact, we used a substrate-trapping technique. Binding of a substrate to the TCPTP-D182A (TC45-D182A) mutant results in a stable interaction between the enzyme and the substrate (Tiganis et al., 1998). Because substrate-enzyme complexes are transient by nature and thereby hard to detect, the substrate-trapping approach has been widely used to identify substrates of protein tyrosine phosphatases Tiganis et al., 1998). When TCPTP protein was immunoprecipitated from the TC45-D182A-and VEGFR2-transfected cells, VEGFR2 was readily co-precipitated (Fig. 3A). We detected the VEGFR2-TCPTP complex both in VEGF-treated and non-treated cells (Fig. 3A,B). This suggests that the molecules interact constitutively or that VEGFR phosphorylation is stimulated by endogenous VEGF that, after the transfection, is present in the culturing medium, thereby triggering the interaction with the substratetrapping mutant prior to the final, intentional VEGF stimulation. Immunoprecipitation with VEGFR2 also resulted in clear coprecipitation of TCPTP (Fig. 3B). Importantly, when sodium orthovanadate was present during the immunoprecipitation, TCPTP no longer co-immunoprecipitated with VEGFR2 (Fig.  3B). The fact that VEGFR2 was competed off from TCPTP-D182A by sodium orthovanadate suggests that the interaction occurs via the catalytic-site active cysteine. Together, these reciprocal immunoprecipitations show that TCPTP phosphatase can bind to VEGFR2 kinase.

TCPTP dephosphorylates VEGFR2 in a site-specific manner
Having demonstrated that VEGFR2 and TCPTP can interact, we investigated the ability of TCPTP to dephosphorylate VEGFR2. Isolated VEGFR2 receptor was treated with purified recombinant TCPTP (wild-type TC45 in all these assays) and total phosphotyrosine levels were detected. TCPTP decreased VEGFR2 phosphorylation by 68±11% (Fig. 4A). This suggests that TCPTP Journal of Cell Science 121 (21) either partially dephosphorylates many of the phosphotyrosines on VEGFR2 or that TCPTP targets VEGFR2 in a phosphorylationsite-specific manner.
VEGFR2 contains several critical tyrosine residues that are autophosphorylated or that bind distinct effector molecules. Some of the most important ones are: tyrosines 1054 and 1059 (Tyr1054/1059), which are located in the activation loop of the tyrosine-kinase domain and are required for maximal activation of the kinase activity of the receptor (Claesson-Welsh, 2003) and for VEGF-A-induced sprouting angiogenesis in vitro (Kawamura et al., 2008); tyrosine residue 1175 (Tyr1175), which binds to PLCγ and Shb; and Tyr1214, which triggers the p38 cascade via unknown intermediates (Olsson et al., 2006). Our phosphorylation assays showed that TCPTP was able to target specific tyrosines in VEGFR2. The autophosphorylation sites Tyr1054/1059 and Tyr1214 were dephosphorylated by TCPTP (Fig. 4B). Tyr996, the functional significance of which is currently uncertain (Olsson et al., 2006), was a TCPTP target as well. Tyr1175, by contrast, remained phosphorylated and was not a TCPTP target (Fig. 4B).
Phosphorylation of VEGFR2 controls the internalization of the receptor (Dougher and Terman, 1999;Lampugnani et al., 2006). This prompted us to use the internalization as a biological read-out for the activation status of VEGFR2. In microscopic assays, transfection of TC37 into subconfluent HUVECs strongly inhibited the internalization of VEGFR2 to endosomeresembling vesicles (Fig. 4C). In biochemical analyses, we used cell-surface labelling with cleavable biotin, which allows the quantification of the internalized, biotinylated VEGFR2 that is protected when any biotinylated receptors remaining on the cell surface are cleaved (Ivaska et al., 2002). We found that VEGF stimulation increased VEGFR2 internalization in sparse mocktransfected HUVECs. This was evident from the increased amount of biotinylated VEGFR2 protected from cleavage ( Fig.  4D). In cells expressing TC37 or wild-type TCPTP (TC45), the internalization of VEGFR2 was clearly inhibited (Fig. 4D). Together, these experiments revealed that TCPTP binds to VEGFR2, dephosphorylates it in a phosphosite-specific manner and alters its biological activity. HUVECs were transfected with no oligo ('-' on the blot), siRNA against TCPTP or a control siRNA (Scr), and TCPTP expression was measured using immunoblotting (inset). Cells were treated with 80 ng/ml VEGF and proliferation of sparse cells that were plated on gelatin was assayed after 16 hours using a BrdU-based assay (mean ± s.d.; n=4-6; *P<0.05). (B) Activation of TCPTP decreases chemotaxis of HUVECs. Subconfluent HUVECs were transfected with the constitutively active TCPTP (TC37) or vector control (pCG) and chemotaxis towards VEGF was measured using a Transwell assay for 4 hours. Cells adhering to the bottom of the filter were fixed and stained with crystal violet (micrographs). The number of migrated cells was counted from four randomly selected fields of view in three parallel wells (mean ± s.e.m.; n=3; **P<0.01). The western blot shows expression of TC37 in the transfected cells (also detected is the 45-kDa endogenous TCPTP). TCPTP inhibits VEGFR2 signalling

Adhesion of endothelial cells to collagen activates TCPTP
In malignant epithelial cells, α1 integrin and TCPTP function as a signalosome in focal-adhesion-like structures at the membrane (Mattila et al., 2005). We therefore analyzed whether α1-integrinmediated adhesion of endothelial cells to collagen would regulate the localization of TCPTP. Immunofluorescence stainings showed that adhesion of HUVECs to collagen (via α1 and α2 integrins) had striking effects on the localization of TCPTP in HUVECs. α1β1 integrin and TCPTP were both visible at the membrane in HUVECs adhering to collagen (Fig. 5A). Interestingly, the α1 integrin and TCPTP signals overlapped with vinculin staining in HUVECs on collagen (Fig. 5A). However, these structures did not resemble focal adhesions. By contrast, in HUVECs adhering to gelatin (mainly via αvβ3 integrin), TCPTP was not visible at the membrane and α1β1 integrin did not localize to membrane protrusions or to prominent vinculin-positive focal adhesions (Fig. 5A).
We then measured the phosphatase activity of TCPTP from HUVECs adhering to collagen or to gelatin. Immunoprecipitated TCPTP from HUVECs grown on gelatin showed basal phosphatase activity. Notably, adhesion of HUVECs to collagen promoted activation of TCPTP by 47±1% (P<0.05) (Fig. 5B). Binding to collagen did not significantly increase the activity of a ubiquitous phosphatase, SHP2 (also known as PTPN11). The phosphatase activities of both TCPTP and SHP2 were normalized to the amounts of immunoprecipitated molecules. Interestingly, unlike matrix adhesion, stimulation of HUVECs with human recombinant VEGF165 had no significant effect on TCPTP activity (not shown). Thus, binding of HUVECs to collagen matrix is able to induce TCPTP activity.

The cytoplasmic domain of α1 integrin activates TCPTP and inhibits VEGFR2 activity
We have demonstrated that, in malignant epithelial cells, a cellpermeable TAT-fusion peptide from the cytoplasmic tail of integrin α1, but not from that of integrins α2, α5, α10 or α11, specifically activates TCPTP (Mattila et al., 2005). Therefore, we studied whether this peptide could be used as a tool to trigger TCPTPdependent VEGFR2 dephosphorylation in normal endothelial cells.
We observed that the FITC-labelled α1-TAT fusion peptide and scrambled control peptide (ScrTAT) entered HUVECs with the same efficiency within 30 minutes, and, for both, the fluorescence signal remained high for over 16 hours (supplementary material Fig. S2C). The peptides had no obvious effects on cell morphology or survival (not shown). VEGF-induced phosphorylation on VEGFR2 Tyr1214 was markedly inhibited by the α1-TAT peptide (Fig. 5C). The inhibition was specific, because the peptide was not able to induce dephosphorylation of VEGFR2 at Tyr1175, which is in line with the in vitro phosphatase assay (Fig. 4B). Remarkably, α1-TAT peptide also abolished the kinase activity of VEGFR2 in VEGFtreated HUVECs (Fig. 5D). Thus, the α1-TAT peptide can be used to dephosphorylate and inactivate VEGFR2 in HUVECs.
To further increase the stability of TAT-fusion peptides in cells, we designed novel TAT peptides with D amino acids instead of normal L isomers. In the in vitro phosphatase assay, α1-D-TAT highly significantly activated TCPTP (P<0.01) (Fig. 5E). Again, the specificity of the TCPTP activation was evident, inasmuch as the ScrTAT fusion made of D isomers had no effect.
TCPTP activation by the cytoplasmic tail of integrin α1 inhibits endothelial proliferation, migration and morphological changes Capillary morphogenesis requires the adhesion, migration, proliferation and differentiation of endothelial cells, and VEGFR2 plays a major role in all of these steps (Ferrara et al., 2003;Avraamides et al., 2008;Kerbel, 2008). The α1-TAT peptide inhibited VEGF-induced morphological changes by 60% compared with non-treated cells and by 40% compared with control peptidetreated cells in a fibrin-gel assay (Fig. 6A). These results show that activation of TCPTP leads to inhibition of VEGF-driven changes in endothelial morphology.
We next determined the chemokinesis and chemotaxis of HUVECs in response to the peptide treatment. When the motility of HUVECs on gelatin substrate was measured for 3 hours using time-lapse imaging, chemokinesis was induced by VEGF by 84±21% (Fig. 6B). In this assay, α1-TAT treatment inhibited VEGF-induced chemokinesis by 61±4%, whereas ScrTAT had no significant effect. It was also clear from the images that neither TAT peptide was harmful to the cells, because all cells were viable and mitosis was entered occasionally (not shown). VEGF-induced chemotaxis of HUVECs was also inhibited by treatment with α1-TAT peptide, but not with the ScrTAT peptide (Fig. 6C).
siRNA silencing was finally used to further confirm that the effect of α1-TAT on VEGF-induced HUVEC migration is mediated via TCPTP. As shown in Fig. 6D, α1-TAT had no inhibitory effects on the VEGF-induced chemotaxis of cells treated with TCPTP siRNA, whereas α1-TAT effectively inhibited VEGF-induced migration of cells treated with scrambled siRNA. Together, these data indicate that TCPTP is the target of the cytoplasmic peptide of α1 integrin in HUVECs and that decreased TCPTP activity results in augmented VEGFR2-mediated proliferation and migration of endothelial cells. were treated with VEGF (100 ng/ml, 5 minutes). Cell lysates were immunoprecipitated (IP) with anti-TCPTP or control (IgG) antibodies and blotted for VEGFR2 and TCPTP as indicated. Aliquots of the lysates were also analyzed to control for equal protein expression. (B) HEK293 cells transfected as in A, or an untransfected control, were treated with VEGF or left untreated. Immunoprecipitations were performed with anti-VEGFR2 or control (IgG) antibodies and blotted as indicated. The co-immunoprecipitations were done in the presence or absence of 100 μM sodium orthovanadate (Na 3 VO 4 ) as indicated. Note the disappearance of the co-immunoprecipitated TCPTP when sodium orthovanadate is present.  (pCG), or left untransfected, and surface-labelled with cleavable biotin. VEGFR2 was allowed to internalize for 15 minutes in the presence or absence of VEGF (100 ng/ml). Biotin remaining on the cell surface was cleaved and VEGFR2 was immunoprecipitated using anti-VEGFR2 antibody. Internalized VEGFR2 was detected by blotting for biotin. Because biotin is cleavable by reducing reagents, a non-reducing gel was used, and this alters the mobility of the proteins compared to a reducing gel. The numbers below the blot show levels of internalized VEGFR2 (mean ± s.d., two individual experiments). TCPTP inhibits VEGFR2 signalling Controlled TCPTP activation can be exploited to inhibit sprouting angiogenesis in vitro VEGF promotes angiogenic sprouting and the responsiveness of endothelial cells to VEGF must be tightly controlled for normal development of the vasculature to occur (Siekmann et al., 2008). To test the effect of TCPTP activation on sprouting angiogenesis, we used three-dimensional cultures of HUVEC spheroids (Korff and Augustin, 1999). VEGF induced both the cumulative length and the number of sprouts in control and ScrTAT-treated cells (Fig.  7A,B). The α1-TAT peptide significantly reduced the length of the VEGF-induced sprouts (Fig. 7A). It also reduced the number of sprouts (Fig. 7B). Similar results were obtained also with TATpeptide-treated spheroids formed by porcine aortic endothelial (PAE) cells, which we also found expressed TCPTP (Fig. 7C and not shown). These results show that activation of TCPTP through the cytoplasmic domain of integrin α1 leads to inhibition of VEGFdriven sprouting angiogenesis in vitro.

Discussion
The signalling pathways that downregulate and terminate VEGF action are incompletely understood, although they are of paramount importance in controlling physiological and pathological angiogenesis. In this study we describe a novel phosphatasemediated inhibition of endothelial cell proliferation, migration and sprouting. We show that TCPTP binds to VEGFR2 and dephosphorylates its tyrosines in a phosphosite-specific manner. TCPTP activation diminishes the kinase activity of VEGFR2 and blocks its internalization from the cell surface. We also demonstrate that, in endothelial cells, the cytoplasmic domain of α1 integrin activates TCPTP and that a cell-membrane-permeable form of the peptide efficiently attenuates VEGFR2-driven functions such as proliferation, migration, sprouting angiogenesis and tube formation in actively growing cells. Together, these data suggest TCPTP as a crucial regulator of the growth and differentiation of endothelial cells. The immunoprecipitates were also resolved on SDS-PAGE gels and probed for SHP2 and TCPTP. The phosphatase activities of TCPTP and SHP2 were normalized to the amounts of immunoprecipitated (IP) molecules. A representative experiment out of three with similar results is shown (mean ± s.e.m.; n=9; *P<0.05; ***P<0.005; n.s., not significant). (C) Serum-starved subconfluent HUVECs were treated for 1 hour with the integrin-α1 cytoplasmic-tail fusion peptide (α1-TAT, 200 nM) followed by VEGF-induction (100 ng/ml, 15 minutes) as indicated. Cell lysates were resolved on SDS-PAGE and immunoblotted for phosphorylated VEGFR2 (or tubulin as a control). (D) HUVECs were treated as in C and VEGFR2 kinase activity (mean ± s.d.; n=3) was measured from the cell lysates. **P<0.01. (E) Recombinant, purified TCPTP was incubated with no peptide (-), α1 or α2 cytoplasmic tail peptides (L amino acids), or with α1 and scrambled TAT peptides (D amino acids) and analyzed for phosphatase activity (mean ± s.d., n=3). **P<0.01.
We show here that TCPTP dephosphorylates specific phosphorylation sites of VEGFR2, such as Tyr1054/1059. Phosphorylation of these residues has been shown to be required for maximal kinase activity and for sprouting angiogenesis, and VEGF-induced internalization of VEGFR2 is delayed when these two sites have been mutated to phenylalanine (Dougher and Terman, 1999;Kawamura et al., 2008). This correlates well with our observation that TCPTP activation impairs sprouting angiogenesis and VEGFR2 internalization in HUVECs. Tyr1175 is an important autophosphorylation site, which is not influenced by TCPTP. This site mediates activation of the PI3K and PLCγ pathways, and plays a role in cell proliferation (Takahashi et al., 2001), although VEGF does not seem to be a powerful mitogen for endothelial cells (Olsson et al., 2006). The lack of TCPTP-mediated dephosphorylation of Tyr1175 is most probably the reason why activation of TCPTP has only modest effects on VEGF-induced proliferation. TCPTP also targeted Tyr1214 of VEGFR2. Phosphorylation of this site has been implicated in VEGF-induced actin remodelling and could thus play a role in endothelial cell Journal of Cell Science 121 (21) migration (Lamalice et al., 2004). Our findings that TCPTP controls VEGF-driven chemotaxis and chemokinesis of HUVECs might thus be due to dephosphorylation of the Tyr1214 site [although Tyr1175 can also influence motility (Olsson et al., 2006)].
TCPTP-deficient mice show normal intrauterine development, but they die soon after birth because of an overwhelming inflammatory phenotype (You-Ten et al., 1997). This phenotype suggests that TCPTP is not absolutely necessary for vasculogenesis. However, the postnatal angiogenetic response of these mice has not been studied to our knowledge. There are many examples of molecules that are not needed for vasculogenesis but can still modulate angiogenesis. For instance, deficiency of placental growth factor or its inhibition with an antibody does not affect embryonic blood-vessel development, but reduces pathological angiogenesis during ischemia, wound healing and cancer Fischer et al., 2007). Unfortunately, the role of the TCPTP targets of VEGFR2, Tyr1054/1059, has not been studied in transgenic mice. Notably, there are major differences in TCPTP between mice and humans, which makes it relevant to study the role of TCPTP in primary endothelial cells of human origin. The TCPTP transcript is alternatively spliced resulting in the expression of different isoforms in human and mouse. In human cells, TCPTP is expressed in two isoforms (48 kDa and 45 kDa), each with a distinct subcellular localization, whereas in murine cells only mRNA coding for the 45-kDa isoform is detected (Ibarra-Sanchez et al., 2000). Thus, the fact that TCPTP dephosphorylates Tyr1054/1059, which control the kinase activity of VEGFR2, suggests that TCPTP could modulate angiogenesis also in vivo.
VEGF upregulates expression of the α1β1 and α2β1 integrins in microvascular endothelial cells. In addition, a combination of function-blocking anti-integrin-α1 and -α2 antibodies inhibit VEGF-driven angiogenesis in mouse skin (Senger et al., 1997) and in tumor xenografts (Senger et al., 2002). We have previously shown that cross-linking of α1 integrin with the exactly same function-blocking monoclonal antibody leads to activation of TCPTP in malignant epithelial cells (Mattila et al., 2005). Therefore, the inhibitory effects of anti-integrin-α1 antibodies on vascular growth in vivo might be due to antibody-triggered activation of TCPTP in addition to blocking the interaction of integrin α1β1 with collagen.
Anti-VEGF antibodies are in clinical use and VEGFR2 kinase inhibitors are being developed for blocking neovascularization in several diseases. Our data suggest that inhibition of VEGFR2 activity by inducing TCPTP could be a novel strategy for antagonizing VEGF activity. Because TCPTP-dependent VEGFR2 inactivation was only seen in subconfluent but not in confluent (supplementary material Fig. S2A,B) cells, its modulation might spare the normal vessels. Because TCPTP activation also directly inhibits other RTKs that control cell growth (Tiganis et al., 1999;Galic et al., 2003;Persson et al., 2004;Simoncic et al., 2006), it might simultaneously target both the malignant cells and the vessels nourishing them. To obtain suitable tools to test this hypothesis in vivo, we have started to screen for novel small-molecule TCPTP activators.
Upon VEGF stimulation, VEGFR2 is rapidly internalized and the rate of internalization is governed by phosphorylation (Dougher and Terman, 1999;Labrecque et al., 2003). Contrary to the earlier dogma, internalized cell-surface receptors maintain their signalling activity in the endosomes and may even recruit specific subset of downstream effectors (Di Fiore and De Camilli, 2001;Kermorgant et al., 2004;Lampugnani et al., 2006). Endothelial cells in nascent vessels have elaborate cell-cell contacts under homeostatic conditions and their turnover is very slow. Under these conditions, VEGFR2 is kept inactive by the VE-cadherin-β-catenin complex, which targets it to the vicinity of DEP1 (also known as PTPRJ) phosphatase, which dephosphorylates the receptor and prevents its internalization (Kroll and Waltenberger, 1997;Mitola et al., 2006). Once internalized, the receptor is actively signalling and is phosphorylated at least on tyrosines 1175 and 1214. SHP2 has also been shown to interact with VEGFR2, presumably via its SH2 domain, but there are contradictory reports about its ability to dephosphorylate the receptor (Mitola et al., 2006). Nevertheless, SHP2 has been reported to enhance internalization of VEGFR2 (Bhattacharya et al., 2005;Ikeda et al., 2005). By contrast, we show here that, in subconfluent cells, TCPTP activation dephosphorylates VEGFR2 and inhibits its internalization.
In the different steps of angiogenesis there are endothelial cells with distinct phenotypes. These are, for example, the tip cells of sprouts (which are migratory, display filopodia and lead the growing sprout through the connective tissue surrounding the vessel) and the non-sprouting stalk cells (which form the vessel lumen) (Siekmann et al., 2008). The tip cells have been shown to upregulate the expression of VEGFR2 and PDGF-B (Gerhardt et al., 2003). We show here that TCPTP dephosphorylates VEGFR2 and that activation of TCPTP inhibits sprouting angiogenesis. Furthermore, PDGF-receptor signalling is upregulated in fibroblasts lacking TCPTP (Persson et al., 2004). Taken together, it is possible that TCPTP plays an important role in regulating sprouting angiogenesis because of its ability to modulate phosphorylation of several Fig. 7. Controlled TCPTP activation inhibits VEGF-driven sprouting angiogenesis. (A,B) HUVEC spheroids embedded in three-dimensional collagen gel were treated with VEGF (20 ng/ml) and 400 nM TAT peptides as indicated. Spheroids were analyzed after 24 hours at 10ϫ magnification (A) and the cumulative sprout length (A, bottom) and the number of sprouts per spheroid (B) were quantified using image analysis. Shown is a representative of two similar experiments (mean ± s.e.m., n=5 spheroids per treatment). (C) PAE spheroids embedded in three-dimensional collagen gel were treated with VEGF (20 ng/ml) and 400 nM TAT peptides as indicated. Spheroids were analyzed after 24 hours at 10ϫ magnification and the cumulative sprout length was quantified using image analysis. Shown is a representative of two similar experiments (mean ± s.e.m.; n=5 spheroids per treatment). signalling molecules that are relevant for the function of the tip cells.
ECM proteins have been shown to positively regulate VEGFR2 activity in a few cases. For instance, ligation of integrin αvβ3 supports VEGF-induced biological responses in endothelial cells in vitro and integrin αvβ3 associates with VEGFR2 in response to VEGF treatment (Borges et al., 2000). Here, we define the first molecular pathway that leads to matrix-guided inhibition of endothelial cell responsiveness to VEGF. This pathway is triggered by binding of endothelial cells to collagenous matrix through integrin α1, which leads to activation of TCPTP and inhibition of VEGFR2 signalling. Our findings, together with other reports on the role of two other phosphatases in controlling VEGFR2, allow us to propose a model of sequential phosphatase-regulated steps in angiogenesis.
(1) In intact endothelium, cells form elaborate cellcell contacts and are unable to respond to VEGF. This is regulated by the VE-cadherin-β-catenin complex, which targets VEGFR2 to cell-cell contacts and results in VEGFR2 dephosphorylation by DEP1, and possibly by other junctional phosphatases. Ligation of integrin α1β1 to basement-membrane collagen IV and subsequent TCPTP activation would not have any additional effect in the endothelial cells in nascent vessels because VEGFR2 is already functionally inactivated through DEP1 in cells with cell-cell contacts (supplementary material Fig. S2A,B). (2) Upon induction of angiogenesis, basement membrane is degraded and the sprouting endothelial cells will become surrounded by stroma ECM and be ligated, most likely via integrins αvβ3, α5β1 and α2β1 (which bind with high affinity to vitronectin, fibronectin and type-I collagen, respectively). The cells are sensitive to VEGF, and VEGFR2 is fully phosphorylated and actively signalling to drive vessel sprouting.
(3) Upon formation of a new basement membrane, integrin α1β1 (binding to collagens IV and I) will activate TCPTP, and possibly SHP2, which begin to attenuate proliferation and migration of the endothelial cells until they return to quiescence upon maturation of the new vessel (Carmeliet, 2000;Conway et al., 2001). Based on this model, different phosphatases would inhibit VEGF signalling at different steps of angiogenesis.
siRNA against TCPTP has previously been shown to be specific to TCPTP and to have identical effects to another TCPTP-targeting oligo (Mattila et al., 2005). TCPTP siRNA and a control siRNA were from Ambion. Peptides containing the cytoplasmic tails of the α1 or α2 integrins, the integrin-α1 tail fused to the 11-aminoacid-long TAT peptide, and scrambled TAT peptides were synthesized by Innovagen. Human recombinant VEGF165 was purchased from Peprotech. Recombinant TCPTP was produced and purified as a GST-fusion protein in Escherichia coli and cleaved from the GST using PreScission according to the manufacturer's instructions (BD Biosciences).

Cells
HEK293 and HeLa cells (ATCC) were maintained in DMEM, 10% FBS, 2 mM Lglutamine, 5% CO 2 . Primary HUVECs were freshly isolated as described (Koskinen et al., 2004) and cultured in endothelial cell growth medium (PromoCell). Only HUVECs passaged ≤ three times were used in this study. They were transfected with DNA plasmids or siRNA duplexes (100 nM) by electroporation using Bio-Rad Gene Pulser Xcell (Ambion).

Immunofluorescence stainings and FACS analyses
The cell-surface expression of integrins in HUVECs was analyzed using a previously described protocol (Pellinen et al., 2006). In brief, detached cells were incubated with the indicated antibodies followed by an Alexa-Fluor-488/555/647-labelled secondstep reagent. The cells were analyzed using FACScalibur.
For immunocytochemistry, HUVECs were plated on type-IV collagen and fixed with 4% paraformaldehyde. VEGFR2, TCPTP and α1 integrin were visualized using indirect immunofluorescence stainings of 0.1% Triton-X-100-permeabilized cells in PBS containing 10% horse serum, 1% BSA and 0.01% azide. The samples were mounted in Vectashield with DAPI (Vector Laboratories, CA) and analyzed using a confocal microscope (Axioplan 2 with LSM 510; Carl Zeiss MicroImaging). Frozen sections from human umbilical cord were fixed and stained as above. The samples were analyzed using an Axiovert 200M inverted microscope and 20ϫ objective.
Immunoprecipitations TCPTP-D182A-and VEGFR2-encoding plasmids were co-transfected into HEK293 cells using Lipofectamine 2000. After 1 day, the cells were serum-starved (overnight), stimulated with VEGF (50 ng/ml, 5 minutes), washed with ice-cold PBS and lysed in the presence or absence of 100 μM sodium orthovanadate. Pre-cleared lysates were subjected to immunoprecipitation with anti-TCPTP, anti-VEGFR2 or control mouse IgG in the presence or absence of 100 μM sodium orthovanadate, resolved on SDS-PAGE and subjected to immunoblotting. Aliquots of the original lysates (5% from input of immunoprecipitations) were analyzed in parallel to control for protein expression.

Phosphatase and kinase assays
HUVECs were plated onto collagen-or gelatin-coated plates and allowed to adhere for 1 hour. TCPTP, SHP2 and control immunocomplexes were isolated from cell lysates, resuspended in the phosphatase-assay buffer and assayed for phosphatase activity using 6,8-difluoro-4-methylumbelliferyl phosphate (diFMUP, Molecular Probes) as a substrate in the presence of a serine/threonine-phosphatase-inhibitor cocktail (Sigma). Thereafter, the antigens from the beads were used for immunoblotting. The phosphatase activities were always normalized to immunoblotted proteins. The phosphatase assays with purified recombinant TCPTP protein were performed using diFMUP as a substrate.
VEGFR2 kinase activity was studied from HUVECs, which were plated on gelatin, serum-starved overnight, treated with α1-TAT or ScrTAT peptides (400 nM) and stimulated with VEGF (100 ng/ml, 15 minutes) using Kinase-Glo Plus kinase assay kit according to the manufacturer's protocol (Promega).
For the in vitro dephosphorylation experiments, HEK293 cells transfected with VEGFR2 were treated with VEGF as described above. VEGFR2 was immunoprecipitated from the lysates (50 mM HEPES pH 7.5, 150 mM NaCl, 10% glycerol, 2.5 mM EDTA, 1% NP-40, Complete protease-inhibitor cocktail, Sigma PPASE inhibitor cocktails I and II). The beads were washed with the phosphataseassay buffer (10 mM HEPES, 150 mM NaCl, 1 mM DTT) and exposed to recombinant TCPTP (10 μg/ml) or buffer only for 10 minutes at +37°C. Immunoblotting was used to detect total VEGFR2 and specific phosphotyrosines.

Cell-proliferation and survival assays
The number of viable HUVECs was analyzed with WST-1 reagent (Roche). Cell proliferation was measured with BrdU assay (Cell Proliferation ELISA Biotrak system, Amersham) from HUVECs grown in RPMI1640 + 1% FBS + 2 mM L-glutamine supplemented with 80 ng/ml VEGF.

Tube-formation assay
In the fibrin-gel in vitro angiogenesis assay (Chemicon), HUVECs were plated onto a fibrin gel and overlaid with a second fibrin gel after an overnight culture. The endothelial cell growth medium with or without VEGF (50 ng/ml) was added in the presence or absence of α1-TAT or scrambled TAT peptides (400 nM). At the 24-hour end-point, the formation of cellular networks was quantified from three randomly selected fields from each well. Scores 0-5 were assigned by a naïve observer to each field based on cellular aligning, capillary-tube formation, sprouting and formation of polygons according to the manufacturer's instructions.

Sprouting assay
Sprouting assays from spheroids were based on the method described earlier (Korff and Augustin, 1999). Briefly, HUVECs or PAE cells were divided into round-bottomed non-treated 96-well plates (Greiner Bio-One), 3000 cells per well, in EBM2 or Ham's F12 GlutaMAX (Gibco) medium with 0.25% methyl cellulose (Sigma). After overnight incubation at +37°C, five spheroids per treatment were pooled and resuspended in spheroid medium (40% FBS, 0.5% methyl cellulose, plain medium). Collagen gel [20 mM HEPES, 1ϫ DMEM (Sigma), PureCol collagen (Vitrogen)] was added 1:1 to cell suspension and the mix transferred to 48-well plates. After polymerization, medium with or without 25 ng/ml VEGF, and 400 nM ScrTAT or α1-TAT-peptide were added. Cumulative length of the sprouts around spheroids or the number of sprouts formed from spheroids was quantified using 10ϫ magnification after 24 hours of incubation.

VEGFR2-internalization assays
For immunofluorescence analysis, HUVECs were transfected with control plasmid or with the constitutively active truncated 37-kDa TCPTP mutant (TC37), grown on gelatin and stimulated with VEGF. VEGFR2 was stained as above and VEGFR2positive endosome-like structures were scored from microscopic images. For biochemical analysis, 48 hours after transfection, vector (pCG empty vector), TC45 or TC37-transfected semiconfluent HUVECs were surface-biotinylated with EZ-Link Sulfo-NHS-SS-Biotin (Pierce) on ice for 30 minutes. After the biotinylation, an aliquot of the cells was immediately lysed after cleaving the biotin on the cell surface with a reducing agent, sodium 2-mercaptoethanesulfonate (MesNa, Fluka), to verify the efficacy of the cleavage reaction (not shown). For the remaining cells, biotinylated proteins were allowed to internalize at +37°C in the presence or absence of VEGF for 15 minutes. Thereafter, the plates were transferred on ice and any remaining biotin on the surface was cleaved with MesNa (leaving only the internalized proteins biotinylated). The cells were then lysed, after which VEGFR2 was immunoprecipitated and run on non-reducing SDS-PAGE. Internalized VEGFR2 was blotted with antibiotin-HRP and the total immunoprecipitated VEGFR2 with anti-VEGFR2 antibody. Lysates were also blotted for TCPTP to verify TCPTP transfection efficacy.

Time-lapse chemokinesis assays
HUVECs were plated on gelatin-coated wells and allowed to attach for 1 hour in the presence or absence of 200 nM α1-TAT or ScrTAT peptides and 20 ng/ml VEGF. Migration of individual cells was recorded at 5-minute intervals with an Axiovert 200M inverted microscope with atmosphere control unit for 4 hours using a 20ϫ objective and digital video recording. The migration pathway of at least ten randomly picked cells was tracked using the Metamorph program. The cumulative migration distances were also plotted.

Chemotaxis assays
HUVECs were treated with 200 nM α1-TAT or ScrTAT, or left untreated for 1 hour. Transwell filters (8-μm pore size, Costar, Cambridge, MA) were coated with gelatin, and medium (RPMI1640, 4 mM L-glutamine, 10% FBS) and VEGF (100 ng/ml) were added to the lower chamber. HUVECs were added into Transwell inserts (4ϫ10 5 cells/well) in RPMI1640 with 0.2% soybean trypsin inhibitor. Chemotaxis was allowed to proceed for 4 hours (+37°C, 5% CO 2 ). The inserts were removed, and the nonmigrated cells from the top of the insert were removed by scraping. The inserts were fixed in 4% paraformaldehyde, cells stained with 0.5% crystal violet and the number of migrated cells was counted using a microscope.