Syndecan-1 regulates α_vβ₃ and α_vβ₅ integrin activation during angiogenesis and is blocked by synstatin, a novel peptide inhibitor

MDA-MB-231 human mammary carcinoma cells rely on Sdc1 for activation of the α_vβ₃ integrin during their binding and spreading on VN (17, 20). Integrin activity is abolished by silencing Sdc1 expression with human-specific small interfering RNA (siRNA) but can be rescued by expression of mouse Sdc1 (mSdc1; Fig. 1 B). Cell binding and spreading on FN, which utilizes the α₅β₁ integrin, is unaffected (Fig. 1 B) (20). Although native mSdc1 rescues integrin activity, deletion mutants (summarized in Fig. 1 A) (20) lacking a putative active site between amino acids 88 and 121 in its extracellular domain fail to rescue (see mSdc1Δ67–121; Fig. 1 B).

We showed previously (20) that α_vβ₃ integrin activity is lost in MDA-MB-231 cell adhesion/spreading assays in the presence of soluble, recombinant glutathione S-transferase–fused Sdc1 extracellular domain (GST-S1ED; ID₅₀ of ∼3 µM). To test whether shorter peptides would mimic this activity, we derived a peptide spanning amino acids 82–130 of mSdc1 (Fig. 1 A). We call the inhibitory peptide sequence SSTN. Indeed, treatment of MDA-MB-231 cells with SSTN_82-130 blocks their spreading on VN but not FN (Fig. 1 C).

We hypothesize that S1ED or SSTN_82-130 may compete for an interaction between the S1ED and the α_vβ₃ integrin. To assess this, immunoprecipitations were performed using MDA-MB-231 cells expressing either empty vector (NEO), ectopic full-length mSdc1, or the mSdc1Δ67–121 deletion mutant that lacks the active site. Immunoblotting reveals that the α_vβ₃ integrin coprecipitates with either native human Sdc1 (hSdc1) or ectopic mSdc1, yet fails to associate with the mSdc1 mutant bearing the Δ67–121 deletion (Fig. 1 D). Coprecipitation is disrupted in the presence of either 30 µM GST-S1ED protein (using either mS1ED or hS1ED to avoid reactivity with the precipitating mAb) or 3 µM SSTN_82-130 (which is not recognized by either precipitating mAb). These results suggest that these inhibitors block α_vβ₃ integrin activation by competitively displacing the integrin from an interaction with Sdc1.

To define the minimal sequence required, we tested SSTN peptides (0–100 µM) carrying sequential N- and C-terminal deletions (Fig. 2 C) for their ability to block cell spreading on VN (Fig. 2, A and B). The minimal peptide required for full activity is SSTN_92-119. Deletion of two or three residues from either end of this peptide either diminishes (approximately threefold loss in activity in SSTN_88-117) or completely abolishes (SSTN_94-119 and SSTN_88-116) activity. A peptide representing the homologous region of hSdc1, hSSTN_89-120, is equally effective as an inhibitor (Fig. 2 C), confirming that the active site is conserved between hSdc1 and mSdc1. Peptides displayed identical inhibitory effects on B82L fibroblasts, which rely solely on the α_vβ₅ integrin for adhesion and spreading on VN (unpublished data) (19).

We next questioned whether the Sdc1 regulatory mechanism is operative on vascular endothelial cells. Human aortic endothelial cells (HAECs) and human dermal microvascular endothelial cells (HMEC-1s) express modest levels of Sdc1 and higher levels of α_vβ₃ and α_vβ₅ integrins (Fig. 3 A). Mouse aortic endothelial cells (MAECs) express higher levels of Sdc1 than their human counterparts but less of the two integrins (Fig. 3 A).

To determine if Sdc1 and the integrins are in a regulatory complex, immunoprecipitates from HMEC-1s were probed for the coprecipitation of β₃ and β₅ integrin subunits with Sdc1. Both integrins coprecipitate with the syndecan and in seemingly equal proportions (quantified via comparison to a whole-cell lysate standard curve; not depicted) suggestive of a 1:1 correspondence (Fig. 3 B). The α_vβ₁ integrin, which is not regulated by Sdc1 (17, 19, 20), fails to coprecipitate (unpublished data). Immunoprecipitations were also conducted in the presence of either 30 µM GST-mS1ED or 3 µM SSTN_82-130, SSTN_92-119, or SSTN_94-119; note that none of these are recognized by the human-specific mAb used for precipitation (Fig. 3 B). As expected, both the GST-mS1ED protein (but not GST alone) and the active SSTN_82-130 and SSTN_92-119 efficiently disrupt the Sdc1–integrin interactions, whereas the inactive SSTN_94-119 is without effect.

To test whether Sdc1 interacts directly with the integrins (Fig. 3 C), GST-mS1ED immobilized on glutathione beads was mixed in solution with 1 µM purified α_vβ₃ or α_vβ₅ integrin in the presence of 10 µM competitive SSTN_92-119 or noncompetitive SSTN_94-119. The purified integrins associate with S1ED, and their binding is competed by SSTN_92-119 but not by SSTN_94-119. In addition, SSTN_92-119 prelabeled with photoactivatable Sulfo-SBED cross-linker (see Materials and methods) binds and transfers biotin label to both the α and β subunits of purified α_vβ₃ integrin, whereas no transfer is detected with SSTN_94-119 (Fig. 3 D).

We next questioned whether the Sdc1 regulatory mechanism affects integrin activity on endothelial cells. Sdc1 expression was silenced in HMEC-1s using siRNA (Fig. 3 E, left), and the cells were plated on VN. Control siRNA–transfected cells bind and spread, but the spreading is blocked by >95% if Sdc1 expression is silenced (Fig. 3 E, right). Quantification of HMEC-1 and HAEC cell spreading (Fig. 3 F, left) or attachment (Fig. 3 F, right) shows that cell spreading is completely blocked, and cell attachment is blocked by 80% for both cell types when Sdc1 expression is reduced by ∼90% (Fig. 3 F). This is compared with combined treatment with 30 µg/ml α_vβ₃ and α_vβ₅ function–blocking antibodies LM609 and P1F6 (Fig. 3 F).

We next tested whether the SSTN peptides inhibit integrin activation on the endothelial cells. HMEC-1s were plated on hSdc1 antibody (mAb B-B4), to which the cells attach and spread (Fig. 4 A). Our previous work using carcinoma cells showed that Sdc1 ligation by this method leads to integrin activation and cell spreading (17), even in the presence of EDTA to block potential ligand binding by the integrin. Indeed, the α_vβ₃ and α_vβ₅ integrins become activated when the cells attach to B-B4, detected by staining the cells with the ligand-mimetic Fab WOW-1 (Fig. 4 A) (30). However, treatment with either 0.5 µM SSTN_82-130 or 30 µg/ml α_vβ₃ and α_vβ₅ integrin function–blocking antibodies (used as a comparative control) prevents both cell spreading and recognition by WOW-1, indicating that SSTN_82-130 inhibits integrin activation. A similar experiment was conducted using suspended human umbilical vein endothelial cells (HUVECs) in which cell-surface Sdc1 is first decorated with hSdc1-specific mAb B-A38 and then clustered by addition of a secondary antibody in the presence or absence of SSTN_82-130. WOW-1 binding shows that clustering of Sdc1 activates the integrin, and this is prevented by the SSTN peptide (Fig. 4 B).

Blockade of integrin activation with the SSTN peptides is also shown by disrupted cell attachment and spreading on VN. HMEC-1s are dependent on both integrins, as partial spreading is observed if one integrin is blocked (e.g., LM609 to block α_vβ₃, or P1F6 to block α_vβ₅), but a complete block occurs in the presence of both antibodies (Fig. 4 C). Although the inactive SSTN_94-119 is without effect even at 100 µM, addition of either 5 µM mS1ED, 0.5 µM SSTN_82-130, or 0.5 µM SSTN_92-119 blocks spreading on VN as effectively as (if not better than) the combined antibody treatment (Fig. 4 B). Attachment and spreading on FN is not affected by these peptides (Fig. 4 C) or by the integrin blocking antibodies (unpublished data). Quantification of these results shows that competition with mS1ED at 1–3-µM concentrations reduces HMEC-1 cell spreading on VN by >80–90% (Fig. 4 D), with an ID₅₀ of ∼1 µM for HAECs, which are more resistant than the HMEC-1s. In comparison, SSTN_82-130 and SSTN_92-119 display an equivalent inhibitory activity when used at a 10-fold lower concentration, displaying significant inhibitory properties in the 0.1–0.3-µM range. As observed in Fig. 4 C, cell adhesion is also disrupted at the higher inhibitory concentrations and can be prevented altogether by adding increased concentrations of the peptides (unpublished data).

Lastly, the specificity of the SSTN peptides for this Sdc1-regulated mechanism was demonstrated comparing HUVECs that are either positive or negative for Sdc1 expression. Although primary HUVECs do express Sdc1 (Sdc1⁺), one isolate of these cells that we obtained failed to express the proteoglycan (Sdc1⁻), as shown by flow cytometry (Fig. 4 E, left); both the Sdc1⁺ and Sdc1⁻ cells express identical levels of α_vβ₃ and α_vβ₅ integrins. Somewhat surprisingly, both sets of cells attach and spread on VN (Fig. 4 E, right), both are inhibited by mAb LM609, and both are inhibited further by addition of mAbs LM609 and P1F6 to block both α_vβ₃ and α_vβ₅ integrins. Addition of SSTN_82-130 inhibits the spreading of the Sdc1⁺ HUVECs, similar to its effect on HMEC-1s and HAECs, but is without effect on the Sdc1⁻ HUVECs. Similarly, silencing Sdc1 expression on the Sdc1⁺ HUVECs blocks their spreading, whereas the siRNA is without effect on the Sdc1⁻ cells, as would be expected. This suggests that an alternative activation mechanism exists in the absence of Sdc1 gene expression, and that the SSTN peptide is specific only for the Sdc1-regulated α_vβ₃ and α_vβ₅ activation mechanism.

As a first step in understanding a potential role for Sdc1 in angiogenesis in vivo, we questioned whether Sdc1 was expressed either in the endothelial lining of resting blood vessels (using the mouse aorta) or in microvessels arising from the aorta in response to angiogenic stimuli. As we found when examining the MAEC line in culture (Fig. 3 A), both Sdc1 and the α_vβ₃ and α_vβ₅ integrins are present even on the resting mouse aorta (Fig. 5 A). In addition, microvessels growing out from mouse aortic explants over 7 d in response to FGF-2 stain positively for Sdc1, which appears coexpressed with the α_vβ₃ and α_vβ₅ integrins (Fig. 5 B). In addition, Sdc1 expression is observed in activated endothelial cells of mouse tumor vasculature (Fig. 5 C). Tumors derived from Wnt-1 overexpression (31) in the mammary gland show a high degree of vascularization, identified as internal networks lined by a thick cell layer that stains intensely for Sdc1 and for platelet–endothelial cell adhesion molecule–1 (PECAM-1/CD31), indicative of activated endothelium. In fact, Sdc1 expression in these cells swamps out the positive staining for Sdc1 seen in the mammary epithelial cells, which are known to be Sdc1 positive. These same cell layers are positive for α_v, β₃, and β₅ integrin subunits. A similar finding is seen in mammary tumors arising from overexpression of ΔN89β-catenin (32). Although the vascularization is not as extensive as in the Wnt-1–induced tumors, there are clear thickened layers of activated endothelial cells that are positive for PECAM-1, Sdc1, and α_vβ₃/α_vβ₅ integrin in these mammary tumors.

To test the importance of the Sdc1 regulatory mechanism during angiogenesis, we tested the activity of SSTN peptides in the in vitro mouse aortic ring outgrowth assay (33) and the in vivo mouse corneal pocket angiogenesis model (34). The aortic outgrowth assay was used to quantify microvessel outgrowth from segments of mouse aorta explanted to collagen gels in the presence of either VEGF or FGF-2, and incubated in the presence of either recombinant mS1ED protein or SSTN peptides. Microvessels are observed growing out of the aortic explants over 7 d in response to VEGF, and this is blocked by increasing concentrations of SSTN_82-130 or by an antibody (M9) that inhibits mouse α_v-containing integrins (Fig. 6 A). Single cells observed to be migrating out from the explants in the presence of SSTN or the blocking antibody are likely to be fibroblasts and pericytes. Indeed, desmin staining shows that pericytes line up on the PECAM-positive microvessels in the absence of SSTN and continue to migrate out from the explant in a disorganized fashion when SSTN prevents the outgrowth of PECAM-positive endothelial cells (Fig. 6 B).

Quantification of vessel outgrowth shows that SSTN inhibits both VEGF- and FGF-stimulated outgrowth. Although some vessel outgrowth is observed in the absence of exogenous angiogenic agents, likely caused by VEGF released by the pericytes growing out from the explant (35), the outgrowth is greatly increased by supplementation with FGF-2 or VEGF (Fig. 6, C and D). Addition of recombinant S1ED blocks VEGF-induced outgrowth by >70% at 10 µM and >95% at 30 µM (Fig. 6 C). FGF-induced outgrowth (Fig. 6 D) requires slightly higher inhibitory concentrations (∼30 µM mS1ED for 70% inhibition). In comparison, SSTN_82-130 and SSTN_92-119 are 10–30-fold more potent that mS1ED and show equivalent inhibition of VEGF- or FGF-induced outgrowth, displaying an ID₅₀ of ∼0.1 µM. These concentrations are nearly identical to the ID₅₀'s displayed in the in vitro cell attachment and spreading assays with HMEC-1s and HAECs (compare with Fig. 4) and are comparable to the effects of 10 µg/ml of the function-blocking α_v integrin mAb M9, which inhibits the activity of both the α_vβ₃ and α_vβ₅ integrins (Fig. 6).

The SSTN peptides were also tested in the in vivo corneal angiogenesis assay (Fig. 7). Poly(2-hydroxyethyl methacrylate) (Poly-HEMA) pellets containing 67 ng FGF-2 and 250 ng sucralfate (used as a slow-release agent) were implanted into the avascular mouse cornea. After 7 d, fluorescent dextran was injected retroorbitally to highlight the vascular system at the time of sacrifice. Visual inspection of either the eye or the dissected cornea shows significant vessel growth toward the implanted pellet from the margin of the cornea. To test the effects of blocking Sdc1, SSTN peptides were delivered systemically via Alzet osmotic pumps (see Materials and methods). Pumps delivering 1 µl/h of up to 100 µM SSTN_82-130 or SSTN_92-119 achieved a nearly total block of angiogenesis, with an ID₅₀ of ∼3–10 µM (Fig. 7 A; and Fig. 7 C, left). To correlate this with the active concentration in the plasma, we took advantage of rabbit polyclonal antibodies generated against mS1ED that detect active SSTN on dot blots (Fig. 7 B). This shows that SSTN accumulates to 125–150 nM in the blood after 1 wk of treatment with 10 µM SSTN. This correlates well with the ID₅₀ of ∼100 nM observed in vitro and suggests a clearance rate of 10–13 h in the plasma. In contrast, delivery of 100 µM of the inactive SSTN_88-116 or SSTN_94-119 in Alzet pumps has no effect on vessel outgrowth (Fig. 7 A; and Fig. 7 C, left).

We also tested SSTN peptide on male BALB/c Sdc1 knockout mice, as it has been previously reported that angiogenesis is not blocked in this animal (36). Indeed, we found that Sdc1^−/− mice display corneal vessel outgrowth in response to FGF-2 (Fig. 7 C, right, blue bars). In addition, this vessel outgrowth is reduced >60% by circulating cRGDfV, a cyclic RGD inhibitory peptide specific for α_v integrins, compared with an inactive cRADfV peptide. This duplicates the extent of inhibition observed for this RGD peptide in other mouse models (15, 37). However, the angiogenesis is not blocked by as much as 100 µM SSTN_82-130; this concentration of SSTN_82-130 inhibits angiogenesis by 90% in a parallel experiment with wild-type BALB/c mice (Fig. 7 C, right, red bars). Thus, a compensatory mechanism exists for the activation of the α_vβ₃ and α_vβ₅ integrin in the Sdc1^−/− mouse that is insensitive to the inhibitory SSTN peptide.

Finally, SSTN_82-130 was tested in an in vivo orthotopic tumor model (Fig. 8). MDA-MB-231 human mammary carcinoma cells were injected subcutaneously into female, athymic BALB/c nude mice. Palpable tumors were allowed to form for 1 wk after injection, at which time Alzet pumps delivering 0.25 µl/h of either PBS alone (control) or SSTN_82-130 were implanted. Tumor growth was monitored for an additional 4 wk before sacrifice. Treatment with 120 µM (0.37 ± 0.02 µM in the serum; not depicted) or 400 µM (1.87 ± 0.06 µM in the serum; not depicted) SSTN_82-130 results in a significant reduction in tumor size relative to control mice (Fig. 8, A and B). Moreover, staining the tumor sections for PECAM-1 or desmin shows a significant reduction in the number of tumor vessels in the SSTN-treated tumors; both endothelial cells and pericytes are observed, but in reduced numbers of vessels (Fig. 8 C). Quantification of angiogenesis either by measuring total vessel length in the sections or total PECAM-1 positive staining shows an 11-fold reduction in animals bearing pumps containing 400 µM SSTN (Fig. 8 D).

Previous work described a dependence on Sdc1 for activation of the α_vβ₃ and α_vβ₅ integrins. This raises the question of whether this regulatory mechanism exists on the vascular endothelium during angiogenesis. Although expression of Sdc1 in the vascular endothelium has not been studied extensively, several reports indicate that it is expressed in the endothelium of newly formed vessels in wound granulation tissue (38–41) and the endothelial lining of intramyocardial blood vessels in the rat heart (42); however, no expression is seen in blood vessels in unwounded skin (39) nor in the normal rabbit aorta (43), suggesting that its expression may be regulated, perhaps by ephrin signaling (44). A role in angiogenesis, however, is seemingly cast in doubt by the finding that angiogenesis is not impeded in the Sdc1-null mouse (36).

Our findings suggest that Sdc1 is indeed expressed in the vascular endothelium and that its core protein has an important regulatory role during the angiogenic process. It is detected, along with the α_vβ₃ and α_vβ₅ integrins, in cultured MAECs, in resting mouse aorta, and in angiogenic aortic microvessels in vitro and tumor vasculature in vivo. Although we have not investigated directly whether or not Sdc1 expression is up-regulated during angiogenesis, it does display intense staining in the endothelial cells undergoing angiogenesis in the aortic ring outgrowth assay and in mammary tumors. Up-regulated expression of Sdc1 would mimic that of the α_vβ₃ integrin, which displays low or nonexistent expression in most adult tissues and is dramatically up-regulated in the vascular endothelium during angiogenesis (45–47). Most importantly, blocking Sdc1 expression or competitively blocking its interaction with these integrins using SSTN serves to inactivate the integrins and block endothelial cell attachment, spreading, and invasion/outgrowth in vitro, as well as angiogenesis and tumor growth in vivo. The regulatory mechanism remains unknown, but it appears to rely on a direct interaction between the integrins and Sdc1.

Interestingly, our findings do confirm the ability of the Sdc1-null mouse to mount a seemingly normal angiogenic response, a response that appears dependent on the α_vβ₃ and α_vβ₅ integrins despite the absence of Sdc1. This finding is matched by an isolate of HUVECs that are also Sdc1 negative. They display active α_vβ₃ and α_vβ₅ integrins, but unlike their Sdc1-positive counterparts, they are not dependent on Sdc1 and are unaffected by SSTN. In these Sdc1-negative cells (and in the Sdc1-null mice), integrin activation presumably occurs via a compensatory mechanism that is activated in the absence of Sdc1 gene expression. This not only explains why angiogenesis is observed in the Sdc1-null mouse, but it also provides a strong argument for the specificity of the SSTN peptide. That is, the peptide appears to compete for a specific interaction of Sdc1 with these integrins that is essential for their activation on Sdc1-expressing cells. If this mechanism is absent, the peptide has no apparent off-target effects even at concentrations up to 100-fold greater than its ID₅₀.

Although the mechanisms of integrin activation have received considerable scrutiny, especially during processes such as platelet activation and in angiogenesis, the reliance of the α_vβ₃ and α_vβ₅ integrins on Sdc1 during angiogenesis has not been appreciated until now. The α_vβ₃ integrin has been shown previously to associate with other membrane proteins, with some leading to enhanced activation of the integrin (48). Integrin-associated protein (also known as CD47) presents an intriguing comparison to Sdc1 in that it associates with the α_IIbβ₃ and α_vβ₃ integrins via its extracellular domain, and this interaction alone is sufficient to activate these integrins (49, 50).

The activation of an integrin typically refers to a conformation change that dramatically increases ligand binding affinity, and the α_IIbβ₃ and α_vβ₃ integrins have been used as models to begin to understand the activation mechanism. It proceeds through two or more activation states, each resulting in a conformational change in the integrin's extracellular domain in response to intracellular signals and extracellular ligand binding (51–55). In a current model for α_vβ₃ activation, the inactive integrin is in a bent configuration and opens like a switchblade when activated (56). This occurs in response to talin binding to the β₃ integrin cytoplasmic domain and breaking of the salt bridge that maintains the inactive conformation (57, 58). However, it is also enhanced by oligomerization of the integrin subunits within the plane of the plasma membrane (59), and by extracellular interactions with matrix ligands and/or other membrane proteins (48).

Our data now suggest that the active conformation of the α_vβ₃ and α_vβ₅ integrins is favored by their direct association with Sdc1. The integrins are activated by plating endothelial cells on Sdc1 antibody, or by antibody-induced clustering of Sdc1 on suspended endothelial cells, as shown in this study by the binding of the ligand-mimetic Fab WOW-1; this has also been shown with carcinoma cells (20). The activation that occurs when Sdc1 is engaged by antibody is not blocked by depletion of divalent cations that are otherwise necessary for ligand binding by the integrin, indicating that Sdc1 is engaged as a membrane coreceptor, not as a ligand. The fact that WOW-1 is a monovalent Fab indicates that the activation is a true increase in integrin affinity rather than increased avidity. As we now show that the syndecan and α_vβ₃ and/or α_vβ₅ integrins form a complex, likely via lateral interactions of their extracellular domains, it seems apparent that clustering Sdc1 leads to clustering of the integrins as well. It is possible that the clustering brings together kinases associated either with the integrin or the Sdc1, initiating kinase transphosphorylation and activation of signaling pathways that activate the integrin through talin. In vivo, clustering would be envisioned to take place when Sdc1 engages the heparin-binding domains of multivalent matrix ligands. SSTN appears to inhibit tumorigenesis and angiogenesis by displacing Sdc1 from the integrins and, thus, blocking their activation during this clustering event.

Sdc1-specific antibodies used were mouse mAbs B-B4 and B-A38 (hSdc1; AbD Serotec) and rat mAb 281.2 (mSdc1) (60). Rabbit polyclonal antibodies against recombinant mS1ED were affinity purified as previously described (20) and used to detect SSTN. Mouse mAb 3G10, which specifically recognizes terminal unsaturated uronic acid residues that remain on the core protein after treatment with heparin lyases (61), was used for simultaneous detection of mSdc1 and hSdc1 on Western blots. Integrin antibodies include the polyclonal antibodies AB1923, AB1932, and AB1926 (against α_v, β₃, and β₅, respectively; Millipore) and mAb N29 (used for Western blotting for β₁; Millipore); blocking mAbs P1F6 and M9 (for α_vβ₅ and α_v, respectively; Millipore), LM609 (for α_vβ₃; courtesy of D. Cheresh, University of California, San Diego, La Jolla, CA, and the Scripps Research Institute, La Jolla, CA); and “Fire & Ice” (rabbit anti-β₃; courtesy of P. Newman, Blood Research Institute, Blood Center of Southeastern Wisconsin, Milwaukee, WI). Mouse α_vβ₅ ALULA function-blocking mAb was provided courtesy of D. Sheppard (University of California, San Francisco, San Francisco, CA). WOW-1, a ligand-mimetic Fab that serves as an “activation sensor” for α_vβ₃, was provided courtesy of S. Shattil (University of California, San Diego, La Jolla, CA) (30). Rat mAbs MEC 13.3 (BD) or 390 (Millipore) against mouse CD31/PECAM-1 were used to identify endothelial cells. Mouse mAb D33 (Dako) against desmin was used to identify pericytes.

Mouse mammary tumors were provided by C. Alexander (University of Wisconsin-Madison, Madison, WI); tumors were from mice overexpressing Wnt-1 under control of the MMTV-LTR (31, 62) or from FVB MMTV–ΔN89β-catenin transgenic mice (63). Tumors were embedded in optimum cutting temperature (OCT) compound (VWR) and frozen sections were prepared for staining.

The SSTN peptides (80–90% pure) were purchased from GenScript Corporation. They are based on the mSdc1 sequence. Recombinant mS1ED fusion protein was prepared as previously described (20).

Type-I collagen was purified from rat tail tendons (64), solubilized in 0.1 M acetic acid, and dialyzed against 0.1× DMEM (no serum) at pH 4 to remove the acetic acid. VN was purified from human plasma using a previously described method (65).

All cell lines were cultured at 37°C in 92.5% air/7.5% CO₂. MDA-MB-231 human mammary carcinoma cells and MAECs were grown in DMEM (Thermo Fisher Scientific) supplemented with 10% FCS (Thermo Fisher Scientific), 4 mM l-glutamine (Sigma-Aldrich), and 100 U/ml penicillin and 100 µg/ml streptomycin (Invitrogen). MAECs and HAECs were provided by R. Auerbach (University of Wisconsin-Madison, Madison, WI). These are spontaneously immortalized cell lines identified as endothelial by phenotypic markers (66). SV40-T immortalized HMEC-1s were developed and provided by E.W. Ades and F.J. Candal (Center for Disease Control, Atlanta, GA) and T.J. Lawley (Emory University, Atlanta, GA) (67). HUVECs were obtained from Cambrex Life Sciences Corp. Human endothelial cells were cultured in MCDB-131 (Invitrogen) medium supplemented with 5 mM l-glutamine, 20 mM NaHCO₃ (Sigma-Aldrich), 10 ng/ml epidermal growth factor, 1 µg/ml hydrocortisone, bovine brain extract with heparin and antibiotics (SingleQuot kit; Cambrex Life Sciences Corp.), and 15% FBS (Gemini Bio-Products).

Cell attachment and spreading assays on Sdc1-specific antibody or VN and silencing of Sdc1 expression by siRNA were conducted as previously described (19, 20). Silencer control siRNAs (Applera Corporation) were use for control siRNA transfections. Images were acquired using an objective (PlanFluor 10× [0.5 NA], PlanApo 20× [0.75 NA], or PlanApo 63× [1.4 NA]; Nikon) on a microscope (Microphot-FX; Nikon) equipped with a cooled charge-coupled device camera system (Image-Point; Nikon).

The mSdc1^Δ67-121 mutant was constructed by digesting full-length mSdc1 cDNA (pcDNA3; provided by R. Sanderson, University of Alabama at Birmingham, Birmingham, AL) with BmgBI and BstEII. The BstEII overhang was then filled in using the Klenow DNA polymerase fragment (Promega) and the vector then blunt-end ligated using T4 DNA ligase (New England Biolabs, Inc.). The mutant was transfected into MDA-MB-231 cells using LipofectAMINE PLUS (Invitrogen) and 10 µg of plasmid, in accordance with the manufacturer's instructions. A stable population (100–300 colonies) was selected in 1.5 mg/ml G418 (EMD) and comparatively sorted against cells expressing full-length mSdc1 for equal expression levels by flow cytometry using mAb 281.2 under sterile conditions on a triple-laser FACSVantage SE equipped with a FACSDiva digital electronics analyzer (BD).

Cells were scanned at the University of Wisconsin Comprehensive Cancer Center Flow Cytometry Facility using a benchtop cytometer (FACSCalibur; BD Biosciences). For cell-surface staining of Sdc1 and/or integrins, cells were lifted in Tris-buffered saline (TBS) containing 5mM EDTA (TES), washed in HEPES-buffered DMEM (Hb-DME) containing 10% serum, incubated for 1 h on ice with 1 µg of primary antibody per 5 × 10⁵ cells, washed, and counterstained with Alexa Fluor 488–conjugated secondary antibodies (Invitrogen). Cell scatter and propidium iodide (1 µg per sample; Sigma-Aldrich) staining profiles were used to gate live, single-cell events for data analysis. For WOW1 Fab staining (i.e., detection of activated α_vβ₃ integrin), HUVECs were lifted in TES and washed in serum-free medium (Hb-DME containing 0.1% heat-denatured BSA [hdBSA]); this medium was used for all washes and incubations that follow unless otherwise specified. To cluster Sdc1, suspended cells were incubated with 1 µg/ml of mAb B-A38 per 5 × 10⁵ cells (or mIgG1 as an isotype control) for 15 min at 37°C, washed, and incubated with 5 µg/ml goat anti–mouse IgG secondary antibody (Jackson ImmunoResearch Laboratories) for an additional 15 min. Cells were washed and blocked for 15 min in serum-free medium containing 1% hdBSA and a 1:100 dilution of monovalent Fab fragment of goat anti–mouse IgG. Cells were washed with a stoichiometric excess of serum-free medium and incubated with WOW-1 Fab (1:4 dilution) for 30 min at 37°C in the presence or absence of 1 µM SSTN_82-130. Cells were washed with serum-free medium, fixed in 4% paraformaldehyde (PFA) for 30 min, and quenched in 0.1 M glycine (in PBS) for 15 min. Cells were then counterstained with 10 µg/ml Alexa Fluor 488–conjugated goat anti–mouse IgG secondary antibody (Invitrogen) before scanning. Mean fluorescent intensity of triplicate samples was used as a measure of activated α_vβ₃ integrin.

mSdc1 and hSdc1 were immunoprecipitated using mAbs 281.2 and B-B4, respectively. 3–5 × 10⁶ cells per milliliter were washed and lysed in buffer containing 50 mM HEPES (pH 7.4), 150 mM NaCl, 10 mM EDTA, 1% Triton X-100, 0.5% sodium deoxycholate, and a 1:100 dilution of protease inhibitor cocktail set III (EMD) for 20 min on ice. Insoluble cell debris was removed by centrifugation at 20,000 g for 15 min at 4°C. Cell lysates (2-mg input determined by BCA assay per reaction; Thermo Fisher Scientific) were precleared using 50 µg/ml of isotype-matched IgG (rat IgG2A for 281.2 and mouse IgG1 for B-B4) and 100 µl of GammaBind Sepharose (50:50 bead/lysis buffer slurry; GE Healthcare). Precleared lysates were incubated at 4°C overnight with 10 µg/ml of anti-syndecan antibody. For competition experiments, 30 µM GST (negative control), 30 µM GST-S1ED, or 3 µM SSTN was added to the precleared lysates in conjunction with anti-Sdc1 antibody. Immune complexes were precipitated with GammaBind Sepharose, pelleted, washed, and incubated in 50 µl of heparitinase buffer (50 mM Hepes, 50 mM NaOAc, 150 mM NaCl, and 5 mM CaCl₂ [pH 6.5]) containing 2.4 × 10⁻³ IU/ml heparitinase (IBEX Technologies, Inc.) and 0.1 conventional U/ml of chondroitin ABC lysase (ICN Biochemicals) for 4 h at 37°C (with fresh enzymes added after 2 h) to remove glycosaminoglycan side chains from the immunoprecipitated syndecan. For precipitations using purified components, GST or GST-S1ED protein was immobilized to glutathione agarose beads (GE Healthcare) and incubated with either purified α_vβ₃ or α_vβ₅ integrin (U.S. Biological) in the presence of either competitive SSTN_92-119 or noncompetitive SSTN_94-119 peptide in sterile PBS containing the protease inhibitor cocktail for 4 h at 4°C. The samples were pelleted and washed in sterile PBS. In biotin label transfer assays, SSTN “bait” (active peptide 92–119 or inactive peptide 94–119) was labeled with Sulfo-SBED Biotin (Thermo Fisher Scientific) in accordance with the manufacturer's instructions and incubated with purified α_vβ₃ integrin “prey” protein in solution. The SBED conjugate was photoactivated with UV light (365 nm at 5 cm for 15 min) to cross-link the bait–prey complexes. After reversal of the cross-links with dithiothreitol, biotin-labeled prey protein was captured using monomeric avidin agarose beads (Thermo Fisher Scientific). All samples were extracted in Laemmli sample buffer (68) and resolved by electrophoresis on a 7.5% Laemmli gel, transferred to ImmobilonP, and probed with alkaline phosphatase (AP)–conjugated streptavidin alone, or 10 µg/ml rabbit anti-β₃ Fire & Ice, rabbit polyclonal anti-β₅ integrin (1:1,000), 10 µg/ml mouse anti–human α_v mAb 3F12 (courtesy of S. Blystone, State University of New York Upstate Medical University, Syracuse, NY), or 1 µg/ml polyclonal anti–mouse S1ED or mAb B-A38 antibody followed by an AP-conjugated secondary. Visualization of immunoreactive bands was performed using ECF reagent (GE Healthcare) and scanned on a Storm PhosphoImager (GE Healthcare).

All animal experiments were approved by the University of Wisconsin Institutional Animal Care and Use Committee. Mouse aortic ring outgrowth assays were conducted as in Masson et al. (33). The thoracic aortae from female BALB/c mice were cut into 1-mm segments and placed in 1.6-mg/ml collagen gels containing MCDB-131 endothelial cell growth medium (Invitrogen) and fresh 2.5% mouse serum, 100 U/ml penicillin, and 100 µg/ml streptomycin. Media containing 30 ng/ml FGF-2 (PeproTech) or 50 ng/ml VEGF (PeproTech) and recombinant mS1ED or SSTN peptide was changed every 2 d for the duration of the experiment (7 d). Total vessel outgrowth was summed for each ring using Metamorph software (version 6.1; MDS Analytical Technologies). Immunofluorescent images were acquired using a PlanApo 20× (0.75 NA) objective and a camera (CoolSnap ES; Roper Scientific) on a microscopy system (Eclipse TE2000U; Nikon). Digital images were acquired using a dissecting microscope (SMZ1500; Nikon) and a digital camera (CoolPix E5000; Nikon).

For corneal angiogenesis, pellets were created using 3 vol Poly-HEMA (Sigma-Aldrich) in 95% ethanol, and 1 vol each of sucralfate (100 µg/ml in PBS; Sigma-Aldrich), FGF-2 (3.2 mg/ml in PBS; PeproTech), and PBS. The solution was rapidly mixed, and 0.5-µl drops were allowed to dry for 1 h at 4°C before being cut into four equal 0.125-µl sections for implantation into corneal pockets created in anesthetized mice using a 1.5-mm microdissecting knife (Roboz Surgical Instrument Company, Inc.). Osmotic pumps (model 2004; Alzet Osmotic Pumps) delivering 1 µl/h containing PBS or PBS with a range of SSTN concentrations from 0–100 µM, or PBS containing 22.7 mM cRGDfV or cRADfV peptide (Enzo Life Sciences, Inc.), reconstituted in a 50:50 DMSO/double-distilled H₂O solution, were implanted immediately after pellet implantation. After 7 d, 0.1 ml FITC-dextran (2,000,000 mol wt; Sigma-Aldrich) was retroorbitally injected into each mouse before sacrifice, and corneas were removed and fixed in 4% PFA for imaging. Total vessel outgrowth from the limbic vessel at the cornea margin was summed for each mouse using Metamorph software. Immunofluorescent images were acquired using a PlanFluor 4× (0.45 NA) objective and a CoolSnap ES camera on an Eclipse TE2000U microscopy system. Digital images were acquired using a SMZ1500 dissecting microscope and a CoolPix E5000 digital camera.

Under asceptic conditions, 5 × 10⁶ MDA-MB-231 human mammary carcinoma cells (0.2 ml vol) were injected subcutaneously in the right rear flank of 6–8-wk-old female, athymic BALB/c nude mice (Taconic) using a 3-ml sterile syringe and a 26-gauge needle. Palpable tumors were allowed to form for 1 wk after injection, at which time each animal was surgically implanted with an Alzet osmotic pump (0.25 µl/h) containing either PBS (control) or PBS plus SSTN_82-130 peptide (either 120 or 400 µM). Tumor growth was monitored for an additional 4 wk by weekly examination and caliper measurements. At 5 wk after injection, the mice were humanely sacrificed and tumors/serum were harvested.

5–8-µm serial sections cut from frozen, OCT-embedded tumors were periodate-lysine-paraformaldehyde fixed, quenched in 0.05% NaBH₄ followed by 0.1M glycine, and permeabilized/blocked in IHC buffer (0.02% NaN3, 0.1% BSA, 0.2% Triton X-100, and 0.05% Tween 20 in PBS) containing 10% goat serum. Sections were stained with rat anti–mouse PECAM-1/CD31 mAb MEC 13.3 (1:25 dilution in IHC buffer) and/or mouse anti-desmin (used neat) followed by Alexa Fluor 488–conjugated goat anti–rat/mouse secondary antibody. Images were acquired using a PlanFluor 40× objective (1.3 NA; Nikon) and a CoolSnap ES camera on an Eclipse TE2000U microscopy system. Mean vessel length and PECAM-1 staining intensity were calculated across six representative fields for each cohort (i.e., control vs. SSTN-treated mice) using Metamorph software.

SSTN serum levels were calculated by dot blotting using affinity-purified, rabbit polyclonal anti-S1ED antibody. In brief, mouse serum collected from SSTN-treated and control mice was pooled and heated at 65°C for 20–30 min (to inactivate endogenous AP activity), and 200 µl was spotted in triplicate on 0.2 µm nitrocellulose (Bio-Rad Laboratories) using a vacuum (Minifold Dot-Blot System; Schleicher & Schuell). After washing, the membrane was fixed in 0.25% glutaraldehyde for 30 min, blocked in TBS containing 6% BSA for 2 h at 4°C, and then probed using 1 µg/ml anti-S1ED antibody followed by AP-conjugated donkey anti–rabbit secondary antibody (1:5,000 dilution). Immunoreactive wells were visualized using ECF reagent and scanned on a Storm PhosphoImager. SSTN levels were calculated using Image Quant software (GE Healthcare) by comparison against a standard curve (i.e., 0–18 µM SSTN-spiked, heat-inactivated serum collected from control mice) spotted on the same blot.

Statistical analyses using a two-tailed Student's t test were performed using Prism software (version 5; GraphPad Software, Inc.). Data that satisfy confidence levels of P < 0.05 or 0.01 are noted. Data are presented as means ± SEM, unless otherwise noted.

Drs. D. Cheresh and S. Shattil, C. Alexander and R. Auerbach, P. Newman, D. Sheppard, S. Blystone, and R. Sanderson are thanked for generously providing cells, tissues, and/or reagents used in this work. G. Thomas is thanked for technical assistance, and Dr. R. Sanderson is thanked for critically reading the manuscript.

This work was supported by funds to A.C. Rapraeger from the National Institutes of Health (grants R01-CA109010 and CA118839) and the American Heart Association (grant AHA0655734Z), and the Robert Draper Technology Innovation Fund of the University of Wisconsin. D.M. Beauvais is supported by a Susan G. Komen Foundation postdoctoral fellowship.

The authors have no conflicting financial interests.