ADAMTS1/METH1 Inhibits Endothelial Cell Proliferation by Direct Binding and Sequestration of VEGF165*

ADAMTS1 is a metalloprotease previously shown to inhibit angiogenesis in a variety of in vitro and in vivo assays. In the present study, we demonstrate that ADAMTS1 significantly blocks VEGFR2 phosphorylation with consequent suppression of endothelial cell proliferation. The effect on VEGFR2 function was due to direct binding and sequestration of VEGF165 by ADAMTS1. Binding was confirmed by co-immunoprecipitation and cross-linking analysis. Inhibition of VEGF function was reversible, as active VEGF could be recovered from the complex. The interaction required the heparin-binding domain of the growth factor, because VEGF121 failed to bind to ADAMTS1. Structure/function analysis with independent ADAMTS1 domains indicated that binding to VEGF165 was mediated by the carboxyl-terminal (CT) region. ADAMTS1 and VEGF165 were also found in association in tumor extracts. These findings provide a mechanism for the anti-angiogenic activity of ADAMTS1 and describe a novel modulator of VEGF bioavailability.

Extracellular matrix proteins can significantly modulate growth factor signaling. This occurs to a large extent, but not exclusively, from direct non-covalent interactions that mediate selective anchorage of growth factors to the extracellular milieu (1)(2)(3). Several extracellular matrix molecules have been shown to bind and sequester growth factors, as well as to enhance signaling by altering presentation to receptor binding sites (4 -10). Angiogenesis is particularly sensitive to this type of regulation due to the critical role of paracrine growth factors in endothelial cell migration and proliferation.
The vascular endothelial growth factor (VEGF) 1 gene pro-duces several splice variants critical for capillary morphogenesis and tumor angiogenesis (11)(12)(13). Haploinsufficiency of this gene is incompatible with development due to major vascular abnormalities, as demonstrated by inactivation of the gene through homologous recombination (14,15). Isoforms of VEGF are secreted by diverse cell types including smooth muscle, fibroblasts, and epithelial cells. These proteins function by activation of two tyrosine kinase receptors, VEGFR1 and VEGFR2, as well as by binding to non-receptor tyrosine kinase coreceptors such as neuropilin 1 and 2 on endothelial cells (16,17). ADAMTS1 was the first member of a growing family of ADAMTS extracellular proteases characterized by the presence of disintegrin-like metalloprotease and a variable number of thrombospondin-like domains (18,19). Once secreted, ADAMTS1 activation requires furin cleavage and removal of the pro-domain. The active protease can undergo a secondary processing event that separates the catalytic subunit from the thrombospondin (TSP) repeats (20,21). These TSP motifs in TSP1 and TSP2 have been shown to block angiogenesis by several, and likely not independent, mechanisms (22)(23)(24).
We demonstrated previously that ADAMTS1 is able to suppress capillary growth using multiple in vivo and in vitro assays (18). Interestingly, the ability of ADAMTS1 to inhibit neovascularization in vivo was greater than that of endostatin and TSP1 at the same molar ratio. The findings are intriguing and appear in marked contrast to the current paradigm that MMPs are pro-migratory and pro-angiogenic (25)(26)(27)(28). In an effort to determine the mechanism(s) responsible for the angiostatic properties of ADAMTS1, we investigated its effects on endothelial cell growth and found that ADAMTS1 was able to drastically decrease VEGFR2 phosphorylation by a mechanism that involved direct binding and sequestration of VEGF 165 . The interaction was also verified in vivo using xenograft assays engineered to express ADAMTS1. Our results demonstrate that ADAMTS1 binds to VEGF 165 , and that this impacts the bioavailability of VEGF with consequences to receptor phosphorylation, endothelial proliferation, and angiogenesis.
Generation of Monoclonal Antibodies against ADAMTS1-Balb/c mice were immunized by intraperitoneal injections of purified ADAMTS1 (10 g) (21) and complete Freud coadjuvant (300 l). Additional immunizations were done with incomplete Freud coadjuvant at days 15 and 33 (intraperitoneal injection) and intravenously at day 45. At day 48, spleenocytes were obtained from immunized mice and fused with SP2 myeloma cells at a 4:1 ratio following established techniques (29). Conditioned media from hybridoma cultures were screened by ELISA against purified ADAMTS1 coated to plastic. Further characterization of positive clones was performed by immunoprecipitation, Western blot analysis, and immunohistochemistry. From 101 positive wells, twelve hybridomas were cloned by repeated limiting dilution based on epitopes and specific properties. Table I summarizes the selected monoclonal antibodies (mAbs).
Endothelial Cell Proliferation-Quiescent BAEC were trypsinized and plated onto 24-well plates in Dulbecco's modified Eagle's medium supplemented with VEGF 165 (R&D Systems) (25 ng/ml), basic fibroblast growth factor (FGF-2) (2 ng/ml) (provided by Dr. Gera Neufeld) or a combination of both, in the presence or absence of recombinant ADAMTS1/METH-1 protein (5 g/ml). A pulse of 1 Ci/well of [ 3 H]thymidine (Amersham Biosciences) was applied over the last 8 h prior to harvesting. Cells were washed and fixed in 10% trichloroacetic acid. Incorporation of [ 3 H]thymidine was determined by scintillation counting, as described previously (30).
Phosphorylation Assays-Subconfluent cells were incubated overnight in serum-free medium and subsequently preincubated for 5 min with 0.1 mM Na 3 VO 4 to inhibit phosphatase activity. Cultures were then washed once and pretreated with 1.5 ml of conditioned media from 293T cells expressing ADAMTS1 or vector alone for 15 min at 37°C. The concentration of ADAMTS1 in the conditioned media ranges between 5.85 g/ml (66 nM) and 16.7 g/ml (191 nM). These values were obtained by ELISA against a standard curve of purified ADAMTS1 (data not shown). Cells were stimulated for 6 min at 37°C with specified concentration of VEGF 165 . The incubation was terminated by removal of the medium and washes with cold phosphate-buffered saline/ 0.2 mM Na 3 VO 4 . Cells were solubilized in lysis buffer (1% Triton X-100, 10 mM Tris-HCl, pH 7.6, 150 mM NaCl, 30 mM sodium pyrophosphate, 50 mM sodium fluoride, 2.1 mM sodium orthovanadate, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 2 g/ml of aprotinin) at 4°C for 15 min. Insoluble material was removed by centrifugation at 4°C for 30 min at 14,000 ϫ g. Equal amounts of cell lysate were separated by SDS-PAGE and transferred to nitrocellulose membranes. Phosphorylated proteins were detected by immunoblotting using antiphosphotyrosine antibodies (polyclonal or mAb PY20, BD Transduction Laboratories) followed by secondary antibodies coupled with horseradish peroxidase and visualized by chemiluminescence (ECL kit, Pierce). Protein-loading control was assessed by Western blot using anti-VEGFR2 (A-3, Santa Cruz Biotechnology) and or anti-enolase antibodies.
Immunoprecipitation-Cell lysates were precleared with 40 l of protein G-agarose (Roche Applied Science) for 1 h at 4°C. Beads were discarded by centrifugation, and the supernatant was incubated with 1 g/ml of the antiphosphotyrosine antibody (PY20) overnight at 4°C followed by addition of protein G-agarose for 1 h under continuous agitation. Immunoprecipitates were washed three times with lysis buffer and extracted in 2ϫ SDS-PAGE sample buffer by boiling 5 min, fractionated by one-dimensional SDS-PAGE, and further analyzed by Western blot with antiphosphotyrosine antibodies. ADAMTS1 was immunoprecipitated from tumor lysates (450 g) as described above using the mAb 5C6D5. The presence of coimmunoprecipitated VEGF was assessed by Western blot using polyclonal anti-VEGF, 375 (generous gift from Don Senger, Beth Israel Deaconess Medical Center, Boston). Levels of ADAMTS1 were determined by Western analysis (mAb 5C6D5).
Radiolabeling of VEGF-VEGF 165 was labeled with 125 I-Na using iodogen as coupling agent. Briefly, VEGF 165 (2 g) was incubated in 200 l of borate buffer (0.01 M Na 2 B 4 O 7 , 0.14 M NaCl, pH 8.2) with 0.3 mCi of 125 I-Na (Amersham Biosciences) in IODO-GEN-precoated iodination tubes (Pierce) 5 min at room temperature. The reaction was stopped by transfer to a fresh tube containing 40 l of 0.4 mg/ml tyrosine in borate buffer for 1 min at room temperature and adding 200 l of 1 mg/ml IK in phosphate-buffered saline/1% BSA. 125 I-VEGF 165 was separated from free iodine using size exclusion chromatography on Sephadex-G25 columns (Amersham Biosciences). The specific radioactivity of the purified iodinated VEGF 165 was 18,886 cpm/ng. Quality and integrity of the labeled VEGF was assessed by SDS-PAGE followed by autoradiography of the dried gel.
Binding Assays-Subconfluent cells were incubated at low serum (2% serum for HAEC and 0.1% for PAE and PAE-VEGFR2 cells) for 5 h at 37°C, rinsed in binding buffer (media containing 20 mM HEPES, 0.2% gelatin) and equilibrated at 4°C for 15 min. Indicated concentrations of 125 I-VEGF 165 were added to the wells in a final volume of 350 l of binding buffer in the presence or absence of ADAMTS1. After 1 h of incubation at 4°C, cells were washed four times in binding buffer and solubilized in 500 l of lysis buffer (2% Triton X-100, 10% glycerol, 1 mg/ml BSA). Bound protein was assessed on a ␥-counter. Nonspecific binding was calculated in the presence of 20-fold excess of cold VEGF 165 . Specific binding was determined by subtracting nonspecific binding from total binding.
Purification of ADAMTS1 and Evaluation of VEGF Levels-Recombinant ADAMTS1 protein was purified from conditioned media of stable 293T cells expressing ADAMTS1 by heparin and Zn 2ϩ -chelate affinity chromatography (21). Conditioned media from 293T cells transfected with the vector alone was used as control. Presence of VEGF in the ADAMTS1 fractions was assessed by Western blot (Chemicon). Levels of ADAMTS1 were tested using the mAb 5D4E11B5 after stripping of the same membrane.
Gel Permeation Chromatography-ADAMTS-VEGF complexes purified by heparin chromatography were concentrated to 1 ml and dialyzed into 10 mM HEPES, 150 mM NaCl, 25 mM 1-decanesulfonic acid sodium TABLE I Characterization of mAbs against ADAMTS ELISA data show the reactivity of the hybridoma culture supernatant to purified ADAMTS1 immobilized on microassay plates. Numbers represent measured absorbance at 405 nm after the color reaction was developed using a secondary antibody coupled to alkaline phosphatase and p-nitrophenyl phosphate as substrate. 110/87/65 correspond to the ADAMTS1 forms recognized by the mAbs tested on immunoprecipitation (I.P.) and Western blot (W.B.) under reducing (R.C.) and nonreducing (N.R.C.) conditions. Also evaluated was the ability of the antibodies to recognize the protein after fixation on 3% formaldehyde or 2% methanol by immunofluorescence (I.F.). ND, Not Done; NEG, Negative. Binding of VEGF to Immobilized ADAMTS1-ADAMTS1 was immunoprecipitated from conditioned media of stably transfected 293T cells using the mAb 5C6D5. Protein A beads were washed three times with phosphate-buffered saline, 1% Triton X-100, 1% BSA, 1 mM CaCl 2 , 1 mM MgCl 2 , and equilibrated in binding buffer (50 mM HEPES, 150 mM NaCl, 0.5% Nonidet P-40, 0.5% BSA, 1 mM CaCl 2 , 1 mM MgCl 2 ). Soluble ligands: VEGF 165 (50 ng) (R&D Systems) and/or heparin (50 ng) (Sigma), were added to the pellets in a final volume of 500 l of binding buffer, and incubated for 30 min at 4°C. The beads were subsequently washed three times with binding buffer. Proteins bound to the immunoconjugates were subjected to SDS-PAGE, and the presence of coimmunoprecipitated VEGF was analyzed by Western blots. Immunoprecipitated ADAMTS1 was tested reproving the same membrane with the mAb 5D4E11B5.
Cross-linking Experiments Using Purified Proteins-Purified ADAMTS1 (1.25 g) was incubated with VEGF 165 (25 ng) in the presence or absence of heparin (50 ng) in 75 l of binding buffer (50 mM HEPES, 100 mM NaCl, pH 7.2) for 2 h at room temperature. Bound proteins were cross-linked with disuccinimydil suberate (1.5 mM) (Pierce) for 30 min at room temperature. The reaction was stopped by adding 20 mM Tris, 50 mM glycine, 2 mM EDTA (pH7.4) for 15 min at room temperature. The samples were processed by SDS-PAGE, and protein complexes, indicating protein-protein interaction, were visualized by Western blots.
Cross-linking Experiments Using Conditioned Media-Subconfluent 293T or T47D cells expressing full-length or different deletion constructs of ADAMTS1 or pCDNA (negative control) grown on 24-well plates were washed twice with serum-free Dulbecco's modified Eagle's medium and incubated in 300 l of serum-free media for 18 h in the presence or absence of VEGF 165 (150 ng) or VEGF 121 (125 ng) (provided by Dr. Gera Neufeld) with or without heparin (1.5 g). Conditioned media was collected, and the interacting proteins were cross-linked and analyzed as described above.
Expression of the Carboxyl-terminal (CT) Region of ADAMTS1 in T47D Cells-The construct for expression of the CT region of ADAMTS1 corresponding to Pro 549 -Ser 951 was obtained by PCR of the full-length FIG. 1. ADAMTS1 inhibits the mitogenic effect of VEGF 165 and dampens VEGFR2 phosphorylation. A, ADAMTS1 inhibits VEGF 165stimulated endothelial cell proliferation. BAEC were synchronized in G 0 by serum starvation (48 h) post-confluence. Cells were seeded and stimulated with VEGF 165 (25 ng/ml), FGF-2 (2 ng/ml), or a combination of both, in the presence or absence of recombinant ADAMTS1 (5 g/ml). Treatment was performed for 24 h. During the last 8 h, a pulse of 1 Ci/well of [ 3 H]thymidine was added. Incorporated counts were determined by liquid scintillation. All counts were normalized to maximum stimulation (VEGF 165 and FGF-2). B, ADAMTS1 partially blocks VEGFR2 phosphorylation in VEGF 165 activated PAE-VEGFR2. PAE-VEGFR2 cells were stimulated with VEGF 165 (25 ng/ml) for 6 min in the presence of conditioned media from 293T cells expressing ADAMTS1 or vector alone. Phosphorylated proteins were visualized using the anti-phosphotyrosine polyclonal Ab. The membrane was stripped and reprobed with anti-VEGFR2. C, ADAMTS1 decreases VEGFR2 phosphorylation in VEGF 165stimulated BAEC. BAEC were stimulated with VEGF 165 (50 ng/ml) as described in B. Anti-enolase was used as loading control. Inset shows VEGFR2 levels of the same samples resolved in a parallel Western blot. D, ADAMTS1 regulates increasing VEGFR2 phosphorylation stimulated by increasing concentrations of VEGF 165 . BAEC were incubated with increasing concentration of VEGF 165 (0 -50 ng/ml) in the presence or absence of ADAMTS1-conditioned media. Phosphoproteins were immunoprecipitated and later analyzed by Western blot using the same antibody. In each case, arrows point to phosphorylated VEGFR2 (P-VEGFR2), VEGFR2, and enolase. cDNA. A KpnI site was introduced at the 5Ј-end by site-directed mutagenesis using the following forward oligo: 5ЈTTTTCATGGTACCT-GGGGAATGTGGG-3Ј. The reverse oligo used was 5Ј-ACTGCATTCT-GCCTTTGTGCAAAAGTC-3Ј. The resulting PCR product was cloned into the pSecTag2/Hygro B vector (Invitrogen) using KpnI/EcoRV. Stable transfectant clones were generated using this plasmid in T47D cells by calcium phosphate transfection and selected with hygromycin-B (300 g/ml). 165 Affecting VEGFR2 Phosphorylation-In a previous report we showed that ADAMTS1 antagonizes endothelial cell proliferation induced by a combination of FGF-2 and VEGF 165 (18). To further dissect the mechanism of action of ADAMTS1, we repeated these experiments using each growth factor independently and in combination. Addition of recombinant ADAMTS1 decreased cell proliferation induced by FGF2 (49% Ϯ 16%) and FGF-2 ϩ VEGF 165 (78% Ϯ 18%). Proliferative signals mediated by VEGF 165 were completely blocked in the presence of ADAMTS1 (Fig. 1A).

ADAMTS1 Inhibits the Mitogenic Effect of VEGF
The mitogenic action of all VEGF isoforms is mediated through binding to the 205-kDa receptor tyrosine kinase VEGFR2 (31). Therefore, we tested whether activation of VEGFR2 was affected by ADAMTS1. VEGFR2 phosphorylation was strongly inhibited by ADAMTS1 in several endothelial cell cultures including PAE-VEGFR2 (Fig. 1B), BAEC (Fig. 1C) and human aortic endothelial cells (data not shown).
Stimulation of BAEC with increasing concentrations of VEGF 165 resulted in a dose-dependent phosphorylation of VEGFR2 and ADAMTS1 reduced the levels of phospho-VEGFR2 in all cases (Fig. 1D). Together these results imply a direct link between VEGF 165 signaling and the anti-angiogenic effects mediated by ADAMTS1.
VEGF Co-purifies with ADAMTS1-It has been reported that TSP repeats in thrombospondin 1 and in connective tissue growth factor bind to VEGF 165 and modulate its activity on endothelial cells (33,34). The CT of ADAMTS1 contains three domains that share significant homology to the TSP repeats of the TSP1 molecule (18). Therefore, we investigated the possibility of an interaction between ADAMTS1 and VEGF 165 as a regulatory mechanism to explain the effect of ADAMTS1 on VEGFR2 phosphorylation and the binding to the endothelial cell surface. We found that indeed VEGF (Fig. 3A, lanes 1-3) was present in the samples containing ADAMTS1 (Fig. 3A,  lanes 5-7) after purification from heparin affinity chromatography (Fig. 3A, lanes 1 and 5) and from chromatography on a Zn 2ϩ -chelate affinity column (Fig. 3A, lanes 2 and 6, 3 and 7). Although detection of VEGF in the first purification step was not surprising, since VEGF interacts avidly with heparin (34,36); its presence after Zn 2ϩ chromatography was unexpected since this growth factor does not bind to Zn 2ϩ . To ascertain the degree of purification of VEGF from these two chromatography procedures in the absence of ADAMTS1, conditioned media from parental cell lines transfected with vector alone was sub- jected to the same chromatography purification. As expected, VEGF was eluted from heparin column at 1 M (Fig. 3B), whereas no binding was detected on Zn 2ϩ -chelate affinity chromatography in the absence of ADAMTS1 (Fig. 3C). 293T cells have been shown previously to secrete VEGF; our results concur with those findings (37,38). These results provide evidence that VEGF binds to ADAMTS1.
The two molecules can be dissociated and purified from one another by gel filtration chromatography (Fig. 4A). When bound to ADAMTS1, VEGF 165 was unable to phosphorylate VEGFR2 (Fig. 4B, sample A). However, VEGF 165 regained its activity when dissociated from the protease (Fig. 4B, sample 13). ADAMTS1 Binds to VEGF 165 -The interaction of ADAMTS1 with VEGF 165 was also evaluated by co-immunoprecipitation and cross-linking assays. Conditioned media was collected from cells that were stably transfected with either vector alone (pCDNA) or ADAMTS1 and immunoprecipitated with anti-ADAMTS1 antibodies. The immunocomplexes were subsequently incubated with VEGF 165 , and binding was evaluated by Western blots. VEGF 165 was detected in the ADAMTS1 immunocomplexes and addition of exogenous heparin increased binding of ADAMTS1 to VEGF 165 (Fig. 5A). No VEGF 165 was bound to immunoprecipitated complexes when media conditioned from cells expressing vector alone was used (Fig. 5A).
Cross-linking experiments were conducted with disuccinimydil suberate to evaluate physical proximity of ADAMTS1 and VEGF 165 . Heterodimeric complexes of about 250 and 130 kDa were detected with VEGF antibodies in the presence of ADAMTS1 (Fig. 5B). VEGF antibodies also recognized 130-kDa species in the ADAMTS1 preparations, indicating presence of VEGF in ADAMTS1 preparations as demonstrated previously (Fig. 3). We predict that the 130-kDa form corresponds to the sum of ADAMTS1 (87 kDa) and VEGF (43 kDa). Addition of heparin enhanced the formation of the high molecular weight complexes; however, the presence of heparin was not a requirement (Fig. 5B).
To ascertain whether ADAMTS1-VEGF 165 complexes were formed in vivo, we incubated 293T cells transfected with ADAMTS1 or empty vector with purified VEGF 165 and heparin. Evaluation of VEGF after cross-linking revealed a shifted band of about 130 kDa only in the ADAMTS1-conditioned media (Fig. 5C). Under these experimental conditions, formation of the complex ADAMTS1-VEGF 165 required heparin. To ensure the presence of ADAMTS1 in the 130-kDa band, the same blot was re-probed with anti-ADAMTS1 antibodies (Fig.  5C, blot to the right). Indeed a 130-kDa band was found only in the presence of VEGF 165 and overlapped with the band identified by the VEGF antisera.
Cross-linking experiments were also performed with VEGF 121 ; however, no interaction was found (Fig. 5D). These results indicated that the carboxyl-terminal heparin-binding domain of VEGF participates in an interaction with ADAMTS1.
ADAMTS1 and VEGF Form Complexes in Vivo-The in vitro data implied that one possible mechanism for ADAMTS1 activity is binding and inactivation of VEGF 165 function. Therefore, we investigated whether ADAMTS1 and VEGF formed complexes in vivo. The experiments were performed using xenograft tumor lysates from T47D mammary carcinoma cells expressing either ADAMTS1 or vector alone (pCDNA). As previously described, we found that biochemical evaluation of VEGF from tissues/tumors reveals multiple bands immunoreactive with several independent VEGF antibodies (Fig. 6A) (39,40), representing different VEGF forms (17,41) as well as possible complexes formed with other proteins (42)(43)(44)(45). Comparison between control and ADAMTS1 expressing tumors revealed a novel VEGF immunoreactive species of ϳ130 kDa present exclusively in the ADAMTS1 tumor lysates (Fig. 6A). The 130-kDa band co-migrated and overlapped with a species FIG. 3. Endogenous VEGF 165 co-purifies with ADAMTS1. A, VEGF 165 copurifies with ADAMTS1. ADAMTS1 was purified by heparin (H) or Zn 2ϩ -chelate affinity chromatography from conditioned media of 293T cells expressing ADAMTS1. 7 g of purified recombinant protein was resolved by SDS-PAGE under non-reducing conditions and evaluated for the presence of VEGF 165 by Western blot. Arrows indicate different forms of VEGF 165 . After stripping of the membrane, levels of purified ADAMTS1 were assessed by Western blot (mAb 5D4E11B5). B, VEGF 165 was purified by heparin affinity chromatography in the absence of ADAMTS1. Conditioned media from 293T cells transfected with vector alone in presence of soluble heparin (5 g/ml) was purified by heparin affinity chromatography. Peaks corresponding to different elution steps ([NaCl]) were resolved by SDS-PAGE and evaluated by Western blot for presence of VEGF. C, VEGF 165 was not purified by Zn 2ϩ chromatography in the absence of ADAMTS1. Conditioned media from 293T cells transfected with vector alone was subjected to Zn 2ϩ -chelate affinity chromatography. Peaks corresponding to samples eluted with different concentration of EDTA were resolved by SDS-PAGE and analyzed by immunoblot for VEGF. In all cases purified VEGF 165 was used as control (C). recognized by ADAMTS1 antibodies (Fig. 6A). Furthermore, VEGF was detected in immunoprecipitates of ADAMTS1 from tumor lysates, whereas no VEGF was detected in lysates from pCDNA tumor tissue treated in an identical manner (Fig. 6B). These in vivo results corroborate the in vitro finding that ADAMTS1 and VEGF form a stable complex.
Defining Domains of ADAMTS1 Involved in VEGF 165 Binding-To determine the domain(s) of ADAMTS1 that participate in the interaction with VEGF 165 , we used several deletion constructs (Fig. 7A). In these experiments, only the full-length P87-ADAMTS1 was able to bind to VEGF 165 (Fig. 7B). Levels of each deletion protein were determined by Western blot analysis in a parallel experiment (Fig. 7C). VEGF appears as a rather diffused and wide band. It is pertinent to the interpretation of these experiments to clarify that the VEGF antibodies have a higher titer than the ADAMTS1 antibodies, with a limit of detection of ϳ10 pg (VEGF) versus 500 pg in the case of ADAMTS1 antibodies (data not shown).
VEGF 165 Binds to the CT Region of ADAMTS1-Previous results indicated that the interaction of VEGF 165 to ADAMTS1 requires the two last TSP repeats and at least part of the spacer region. To test this possibility, we generated a deletion construct consisting of the CT of ADAMTS1 from the 1st TSP repeat to the last residue (including the spacer region) (Fig.  8A). Fig. 8B shows the relative expression levels of several clones, which showed a truncated protein of ϳ50 kDa. We found that the protein bound avidly to the cell surface of expressing cells and was only released when in the presence of heparin (Fig. 8B). As anticipated, the CT region of ADAMTS1 binds to VEGF 165 , resulting in a 90-kDa cross-linked complex (Fig. 8C). Purification of significant levels of the CT region for evaluation on angiogenesis has been hindered due to its avidity to bind matrix proteins. However we have evaluated the function of this domain in phosphorylation assays and found significant decrease of VEGFR2 phosphorylation on VEGF 165stimulated PAE-VEGFR2 in the presence of the CT-truncated protein as well as the full-length of ADAMTS1 (Fig. 8D). DISCUSSION The results presented in this study support our previous finding that ADAMTS1 is angio-inhibitory and demonstrate that this activity is due, at least in part, to its CT domain. In that region, ADAMTS1 contains three motifs that share significant homology to the antiangiogenic domain of TSP1 and a cysteine-rich region (18). It has been reported that these motifs in TSP1 and in connective tissue growth factor bind to VEGF 165 and negatively modulate VEGF function on endothelial cells (33,34).
VEGF 165 is one of the most specific mediators of tumor angiogenesis (46 -48). The importance of the growth factor and its receptor VEGFR2 to pathophysiological states has been exemplified by studies using dominant-negative VEGFR2 (48), VEGFR2 kinase inhibitors (49), and neutralizing VEGF and VEGFR2 antibodies (47,50,51). In all cases, suppression of VEGF signaling resulted in inhibition of angiogenesis and concomitant reduction of tumor burden. Moreover, overexpression of VEGF and VEGFR2 is associated with invasion and metastasis of human cancers (52,53), further demonstrating the impact of this growth factor in tumor biology.
As a paracrine mediator, one of the rate-limiting steps in VEGF signaling relates to its ability to reach the target cell. The rate of diffusion of VEGF is greatly impacted by the presence of a heparin-binding domain in its carboxyl-terminal. Splice variants of VEGF that lack this domain, such as VEGF 121 , have been shown to interact poorly with the extracellular environment and have a greater diffusion rate (54). Thus, the balance between the selective anchorage of VEGF by extracellular matrix proteins and its programmed release by matrix metalloproteases regulates the level of growth factor bioavailability and function. Consequently, alterations in either matrix-binding proteins or secretion of proteases have been shown to directly impact VEGF function (23,55).
Our results demonstrate that ADAMTS1 binds to VEGF 165 , as indicated by cross-linking and co-immunoprecipitation studies and that this binding prevents the association of this growth factor with its receptor, affecting VEGFR2 phosphorylation and endothelial cell proliferation. The interaction of ADAMTS1 with VEGF 165 occurs under culture conditions and is maintained through purification procedures. VEGF 165 was detected in purified samples of ADAMTS1 after Zn 2ϩ -chelation chromatography, but not in control samples subjected to the same purification scheme. ADAMTS1 protein purified in this manner was still quite effective in the suppression of angiogenesis, as previously demonstrated (18,21). Our data indicate that VEGF is present in purified ADAMTS1 from 293T-conditioned media. The levels of VEGF are not equimolar to ADAMTS1, and addition of exogenous VEGF results in more binding and inactivation of the growth factor. Also, we were unable to detect VEGF using silver staining of purified FIG. 4. VEGF phosphorylates VEGFR2 after dissociation from ADAMTS1. A, dissociation of interacting ADAMTS1-VEGF 165 . Purified ADAMTS1 complexed to VEGF 165 (sample 3 or 7 from Fig. 3A) was subjected to denaturing conditions, and dissociated proteins were separated by gel permeation chromatography. Eluted samples were analyzed for the presence of VEGF 165 under reducing conditions by Western blot techniques. After stripping the membrane, levels of ADAMTS1 were assessed using the antibody 5D4E11B5. B, dissociated VEGF 165 phosphorylates VEGFR2. PAE-VEGFR2 cells were stimulated with VEGF (50 ng/ml), purified ADAMTS1 (ADAMTS1 complexed to VEGF 165 ), samples 6 (ADAMTS1) and 13 (VEGF 165 ) from size exclusion chromatography (A). Proteins were resolved by SDS-PAGE, and phosphorylated proteins were visualized using the antiphosphotyrosine sera. The membrane was stripped and reprobed with anti-enolase to assess protein levels. SM, starting material; C, control; V, VEGF; A, purified ADAMTS1. ADAMTS1 preparations. Consequently, ADAMTS1 is likely more potent than previously anticipated (18). Evaluation of the ADAMTS1-VEGF 165 complex in VEGFR2 phosphorylation assays indicated no activity, demonstrating that once bound, VEGF 165 is unable to activate its receptor. Using gel-filtration chromatography under denaturing conditions we have been able to separate both proteins and determine that the resulting VEGF 165 is active, demonstrating that the reversibility of this interaction has the potential to impact biological processes.
Interestingly, we found that ADAMTS1 did not bind to VEGF 121 , indicating that the interaction requires the heparinbinding domain. Kuno et al. (20,56) have found that ADAMTS1 binds to the extracellular matrix through the spacer/cysteinerich region. We also found that the association of ADAMTS1 with cell surfaces and matrix can be blocked/competed by heparin (21). The cross-linking results suggest that heparin might be acting as a bridge between ADAMTS1 and VEGF. The function of heparin as a chaperone has been previously reported in similar protein-protein interactions (4,57). In vivo, however, the interaction of ADAMTS1 to VEGF is not dependent on exogenous addition of heparin indicating that is either not necessary or that another heparin-like molecule, such a heparan-sulfate proteoglycan i.e. syndecan, could be the physiological mediator of this interaction.
The results presented in this study argue that the CT region of ADAMTS1 is responsible, at least in part, for the antiangiogenic properties displayed by this molecule. Interestingly, we have previously reported that this protein is processed extracellularly resulting in the release of the CT end (21); this processing does not affect the catalytic function of the protease. Thus, ADAMTS1 processing might release two fragments with distinct/independent functions in the extracellular milieu and add to the list of extracellular matrix molecules whose functional features are multiplied by processing events. The contribution of the catalytic domain to the angio-inhibitory properties of ADAMTS1 is the subject of continuing work and is likely to be additive to the vascular inhibitory properties displayed by the carboxyl-terminal domain that we described here. FIG. 5. ADAMTS1 binds to VEGF 165 . A, ADAMTS1 immunocomplexes bind to VEGF 165 . Immunoconjugates generated by immunoprecipitation from conditioned media of 293T cells expressing ADAMTS1 or the pCDNA vector alone were incubated with soluble ligands: VEGF 165 (50 ng) in presence or absence of heparin (50 ng). The conjugates were evaluated for the presence of bound VEGF 165 by immunoblotting under non-reducing conditions. Arrows indicate immunoreactive VEGF 165 oligomers. Presence and relative levels of P87-ADAMTS1 were assessed by Western blot (inset). B, cross-linking analysis of ADAMTS1 and VEGF 165 . Purified ADAMTS1 and VEGF 165 were incubated in the presence or absence of heparin. Bound proteins were cross-linked, and complexes were analyzed under non-reducing conditions. Arrows indicate shifted VEGF 165 species. Arrowheads indicate the migration of VEGF 165 in the absence of ADAMTS1. C, ADAMTS1 from conditioned media supernatants binds to VEGF 165 . Conditioned media from 293T cells expressing ADAMTS1 or pCDNA were incubated in presence or absence of VEGF 165 and cross-linked. Complexes were evaluated under reducing conditions and probed with anti-VEGF antibodies. Levels of ADAMTS1 were evaluated by reprobing the same membrane with MAb 5D4E11B5. Arrow indicates the species recognized by both antibodies. D, ADAMTS1 does not bind to VEGF 121 . Complexes were analyzed under non-reducing conditions. Arrow indicates the shifted VEGF 165 band. Levels of ADAMTS1 were evaluated with the mAb 5D4E11B5. A1, ADAMTS1; Ctrl, Control.
At this point, it is unclear whether the CT region in ADAMTS1 binds and modulate other growth factors. Throm-bospondin1 has been shown to bind to several growth factors contributing to the regulation of many biological processes (22, 58 -60). Although the murine ADAMTS1 protein has the RFK tripeptide known to activate TGF-␤ in TSP1, the human ADAMTS1 homolog does not show conservation of this activating motif. However, based on extensive flanking sequence FIG. 6. ADAMTS1 and VEGF form complexes in vivo. A, ADAMTS1 binds to VEGF 165 in vivo. Tumor lysates (150 g) from T47D xenografts expressing either ADAMTS1 or empty vector (pCDNA) were separated by SDS-PAGE under non-reducing. VEGF 165 was visualized by immunoblot techniques using a carboxyl-terminal-specific anti-VEGF antibody. Expression of ADAMTS1 was evaluated reprobing the same membrane with a mixture of mAbs (3B12F6, 3C8F8, 4C2C4, and 5C6D5). Arrows indicate overlapping proteins, VEGF 165 and ADAMTS1, in ADAMTS1 xenografts samples not present in control tumors. Anti-enolase sera was used to assess loading levels. B, VEGF 165 co-immunoprecipitates with ADAMTS1 in xenograft tumors. ADAMTS1 was immunoprecipitated from control and ADAMTS1 xenograft tumor lysates using the mAb 5C6D5. Presence of co-immunoprecipitated VEGF was tested by Western blot using a carboxyl-terminal anti-VEGF antibody. Immunoprecipitated ADAMTS1 was visualized with 5C6D5 and 3C8F8 antibodies. PC, control tumor; A1, ADAMTS1 tumor. homology it is likely ADAMTS1 binds to TGF-␤, a possibility that deserves examination.
The present study has extended and provided clarification to the mechanism by which ADAMTS1 acts to suppress angiogenesis. However, the overall contribution of ADAMTS1 as a physiological endogenous inhibitor remains to be fully clarified. Our previous results have indicated that added exogenously, ADAMTS1 is capable of blocking neovascularization in vivo and suppressing endothelial cell proliferation in vitro more effectively than recognized inhibitors, such as TSP1 and endostatin. These findings have to be reconciled, however, with the fact that loss of function mutations of ADAMTS1 in mice results in pathological states not directly linked to the vascular compartment (61). Although an extensive evaluation of the null animal with focused attention in angiogenic states is currently missing, it may not be surprising that ADAMTS1 functions in a multiplicity of biological scenarios with either partially overlapping or entirely independent functions. This has been the case for other extracellular matrix proteins and proteases, including TSP1 and TSP2, both recognized physiological inhibitors of neovascularization (62,63). Studies that aim to determine the mechanism of action of ADAMTS1 and/or the TSP repeats are likely to enhance our understanding of the biology of angiogenesis and provide a platform for the development of drugs for pharmacological intervention in situations of unwanted neovascular growth.