Integrin α3β1, a Novel Receptor for α3(IV) Noncollagenous Domain and a Trans-dominant Inhibitor for Integrin αvβ3*

Exogenous soluble human α3 noncollagenous (NC1) domain of collagen IV inhibits angiogenesis and tumor growth. These biological functions are attributed to the binding of α3NC1 to integrin αvβ3. However, in some tumor cells that express integrin αvβ3, the α3NC1 domain does not inhibit proliferation, suggesting that integrin αvβ3 expression is not sufficient to mediate the anti-tumorigenic activity of this domain. Therefore, in the present study, we searched for novel binding receptors for the soluble α3NC1 domain in cells lacking αvβ3 integrin. In these cells, soluble α3NC1 bound integrin α3β1; however, unlike αvβ3, α3β1 integrin did not mediate cell adhesion to immobilized α3NC1 domain. Interestingly, in cells lacking integrin α3β1, adhesion to the α3NC1 domain was enhanced due to activation of integrin αvβ3. These findings indicate that integrin α3β1 is a receptor for the α3NC1 domain and transdominantly inhibits integrin αvβ3 activation. Thus integrin α3β1, in conjunction with integrin αvβ3, modulates cellular responses to the α3NC1 domain, which may be pivotal in the mechanism underpinning its anti-angiogenic and anti-tumorigenic activities.

The NC1 4 domain of certain ␣-chains of type IV collagen display activity as inhibitors of angiogenesis and tumor growth. The capacity of the exogenous ␣1NC1 and ␣2NC1 domains to disrupt basement membrane assembly, blocking tissue development in vivo, was first described in Hydra vulgaris (1). Sub-sequent to these observations, we were the first to demonstrate that the recombinant ␣2NC1 and ␣3NC1 domains of human collagen IV potently inhibited tumor growth and angiogenesis by binding to endothelial cells in an integrin ␣v␤3-dependent manner (2). Since these initial observations, NC1 domains of different collagen IV chains have emerged as a new class of anti-angiogenic and anti-tumorigenic molecules (3,4). These domains exert their effects by direct binding to tumor and endothelial cells where they induce apoptosis and/or inhibit cell proliferation. The mechanism of action of the NC1 domains is attributed to their interactions with integrins, transmembrane receptors for extracellular matrix components (5). NC1 domains bind to distinct integrins, for example ␣1NC1 to integrin ␣1␤1 (3), ␣2NC1 to integrins ␣1␤1, ␣v␤3, and ␣v␤5 (6,7), and ␣3NC1 to integrins ␣v␤3 and ␣v␤5 (2,4,8).
The ␣3NC1 domain is the best characterized of these domains and its anti-tumorigenic effects are predominantly ascribed to its potent anti-angiogenic properties. Endothelial cells adhere to this domain in an integrin ␣v␤3-dependent manner (2,8). Furthermore, integrin ␣v␤3 is thought to mediate ␣3NC1-dependent inhibition of endothelial cell proliferation (9). In addition to its anti-angiogenic effects, the ␣3NC1 domain or peptides derived from its C-terminal third also inhibit melanoma cell growth both in vivo and in vitro in an integrin ␣v␤3-dependent manner (10 -13). Interestingly, the tumor cell inhibition is cell type specific, as the ␣3NC1 domain does not inhibit the proliferation of PC-3 prostate carcinoma cells or 786-O renal carcinoma, although these cells express functionally active integrin ␣v␤3 This dichotomy raises the issue of how integrin ␣v␤3 mediates the anti-tumorigenic activity of the ␣3NC1 domain. One possible mechanism is by transdominant inhibition and activation of ␣v␤3 by other integrins. In this context, ␣v␤3 affinity is higher in cells deficient in integrin ␣5␤1 and increased expression of ␣5␤1 reduces ␣v␤3mediated adhesion and migration (14). Similarly, integrin ␣3␤1 has been demonstrated to alter the function of other integrins and the formation of stress fibers in mouse keratinocytes (15).
We therefore undertook an unbiased approach to determine whether other integrin receptors might bind the ␣3NC1 domain and modulate integrin ␣v␤3 functions. Utilizing flow cytometry, we found that integrin ␣3␤1, a non-classical collagen binding integrin, is a novel receptor for soluble ␣3NC1 domain. Furthermore, we provide evidence that integrin ␣v␤3 affinity is negatively modulated by integrin ␣3␤1. Thus integrin ␣3␤1 may play a key role in mediating the anti-tumorigenic activity of the ␣3NC1 domain.
Cell Culture-Human umbilical vein endothelial cells (HUVECs), obtained from BioWhittaker, were grown in EGM-2 MV medium (BioWhittaker) and used between passages 4 and 8. Human melanoma HT-144 and lung carcinoma A549 were from ATCC and maintained in McCoy modified medium or F12K medium supplemented with 10% fetal bovine serum.
Recombinant Proteins-In all studies where soluble human ␣3NC1 domain has shown anti-tumorigenic activity in vivo, the domain contains a 12-residue collagenous sequence at the N terminus containing an RGD motif in addition to the 232-amino acid noncollagenous region. This domain was originally produced by collagenase digestion of native basement membranes, to ensure the preservation of epitopes for Goodpasture auto-antibodies (21,22). This recombinant protein is equivalent to tumstatin (NCBI accession number AAF72632 (Gen-Bank TM )) in other reports (4,23). All the assays reported in this manuscript, except where indicated, were performed with this recombinant protein. Recombinant human NC1 domains that carried the FLAG sequence on the N terminus were stably expressed in human embryonic kidney 293 cells and purified from conditioned medium by affinity chromatography on anti-FLAG-agarose, as described previously (8,24). The chimeric ␣3/␣1 proteins were purified as described (24).
Cell Adhesion Assay-Proteins at different concentrations were coated on 96-well plates in Na 2 CO 3 /NaHCO 3 , pH 9.5, at 4°C. After 12 h nonspecific binding sites were blocked with 1% bovine serum albumin (BSA) at 37°C for 2 h. Cells were harvested, washed, and suspended in RPMI supplemented with 0.25% BSA and 1 mM MgCl 2 . Where indicated adhesion buffer was supplemented with 0.2 mM MnCl2. 10 5 cells in 100 l were added to each well and incubated at 37°C. After 1 h nonadherent cells were removed by washing with phosphate buffer. The attached cells were fixed with 4% paraformaldehyde, stained with 0.1% crystal violet, and lysed with 10% acetic acid. Cell adhesion was quantified by reading the plates at 595 nm with a microtiter plate reader. The absorbance of the wells coated with 1% BSA was subtracted from each well. For adhesion blocking experiments antibodies to indicated integrins (10 g/ml) were added to 96-well plates prior to addition of cells.
Flow Cytometry-In all the flow cytometry assays performed, the ␣3NC1 domain was preincubated with mAb EB3 for 30 min at a 2:1 molar ratio. This ensures detection of ligand receptor interactions that rely not only on integrin affinity but also on avidity (25). This mixture was then added to 2 ϫ 10 6 cells suspended in RPMI media with 1 mM MgCl 2 . After 1 h, cells were washed twice, incubated with FITC-conjugated anti-mouse IgG1 antibodies, and analyzed using a FACScan (BD Biosciences). For divalent cation chelating experiments, the RPMI was supplemented with 5 mM EDTA. For integrin-blocking experiments, cells were incubated with anti-integrin antibodies (0.5 mg/ml) prior to addition of the ␣3NC1 ϩ mAb EB3 mixture. ␣3NC1 binding to cells was detected with a FITC-conjugated monoclonal rat anti-mouse IgG1 which does not react with any of the anti-integrin antibodies. For WOW-1 binding, cells were incubated sequentially with the ␣3NC1 ϩ mAb EB3 mixture, WOW-1, and a FITC-conjugated anti-His Tag mouse mAb that recognizes WOW-1. Data collected in flow cytometry experiments were analyzed using Cell Quest software (BD Biosciences).
To evaluate integrin expression levels cells were incubated with specific integrin antibodies for 45 min, washed twice, and then incubated with phycoerythrin-conjugated anti-mouse or anti-hamster antibodies and analyzed with a FACScan.
Proliferation Assays-CT26 cells (5 ϫ 10 3 cells/well) were seeded into 96-well plates and treated with ␣3NC1 at various concentrations in Dulbecco's modified Eagle's medium supplemented with 2% fetal bovine serum and proliferation was measured using [ 3 H]thymidine incorporation as described (26).
Solid Phase Ligand Binding Assays-Purified integrin ␣3␤1 (Chemicon) was coated on 96-well plates at 1 g/ml in TBS overnight at 4°C. The plates were blocked with TBS with 0.1% BSA and 0.3% Tween 20. The NC1 domains preincubated with mAb EB3 were added to integrin coated wells in binding buffer (TBS, 0.1% BSA, 1 mM MgCl 2 , 5 mM octyl glucoside) and incubated for 90 min at room temperature. After extensive washes (TBS, 1 mM MgCl 2 , 0.05% Tween), the bound proteins were detected with alkaline phosphatase-conjugated anti-mouse IgG antibodies. p-Nitrophenyl phosphate substrate (Sigma) was added to wells, and the absorbance was measured at 405 nm.
Statistical Analysis-The Student's t test was used for comparisons between two groups, and analysis of variance using Sigma-Stat software for statistical differences between multiple groups. p Ͻ 0.05 was considered statistically significant.

The Binding of Soluble ␣3NC1 Domain to HUVECs Is
Integrin-dependent-The ␣3NC1 domain is thought to mediate its effects primarily by binding to ␣v integrins. These receptors were identified using cell adhesion assays or affinity chromatography where the NC1 domain was immobilized on a fixed substratum. However, soluble, and not the immobilized NC1 domain, is the exogenous physical form used in cell culture and animal experiments. Hence, we developed a flow cytometrybased cell binding assay to determine whether soluble ␣3NC1 domain interacts with other membrane receptors.
As integrins ␣v␤3 and ␣v␤5, the major ␣3NC1 domain-binding receptors (8), are highly expressed on endothelial cells, we initially performed our flow cytometry assay on HUVECs. As shown in Fig. 1A, the ␣3NC1 domain bound to HUVECs, and this binding was significantly inhibited by EDTA, suggesting that the principal receptors for this ligand were integrins. As HUVECs express ␣v␤3, the principal integrin to which the ␣3NC1 domain binds, we investigated whether this binding was dependent on the RGD sequence located at the N terminus of the ␣3NCI domain. Thus, the ␣3NC1 domain, with or without the RGD motif, was utilized for flow cytometric assays. Both forms of the domain bound to HUVECs with equal efficiency, indicating that the binding is independent of the RGD sequence (Fig. 1B). The binding was specific for the ␣3NC1 domain as C6, a recombinant protein derived from the ␣1NC1 domain, which contains the epitope for mAb EB3 (22), did not bind to HUVEC (Fig. 1C). To determine whether binding of the ␣3NC1 domain was primarily dependent on ␣v integrins, flow cytometry was performed in the presence of cyclo-RGD peptides, which block the ligand binding site of ␣v integrins. Surprisingly, these peptides did not affect ␣3NC1 domain binding (Fig. 1D), suggesting that non-RGD binding receptors for the ␣3NC1 domain are present on HUVEC. To determine whether ␣3NC1 domain binding was dependent on a ␤1 integrin, similar flow cytometry assays to those described above were performed on mouse endothelial cells. Human cells cannot be used for antibody-dependent inhibition experiments because functional blocking antibodies, like EB3, are of the IgG1 isotype and interfere with detection by flow cytometry. A small decrease in ␣3NC1 binding to mouse endothelial cells was seen in the presence of an antibody to ␤1 integrin (Fig. 1E), suggesting that ␣s␤1 integrins could potentially be receptors for the ␣3NC1 domain in the absence of integrin ␣v␤3.
␣3NC1 Domain Binds to Integrin ␣3␤1-To identify these putative ␣s␤1 receptors, we used mouse colon carcinoma CT26 cells, as these cells do not express integrin ␤3 and only very low levels of integrin ␣v (Ref. 27 and data not shown). As seen in Fig.  2A, despite the lack of integrin ␣v␤3 expression, the ␣3NC1 domain bound to CT26 cells. To confirm whether a ␤1 integrin was the receptor, we preincubated the cells with anti-␤1 integrin antibody before addition of the ␣3NC1 domain. This anti-  body markedly decreased the binding, indicating that the ␣3NC1 domain binds to a ␤1 containing integrin (Fig. 2B).
To identify the ␣ subunit, we analyzed integrin expression of CT26 cells by flow cytometry and Western blot analysis and found high levels of integrin ␣2, ␣3, ␣5, and ␣6 (data not shown). Blocking antibodies to mouse ␣2, ␣5, ␣6, and ␣v integrins had no significant effect on ␣3NC1 binding (data not shown), suggesting that these integrins were not involved. Thus, integrin ␣3␤1 was the best candidate receptor for the ␣3NC1 domain. Since neutralizing antibodies to mouse integrin ␣3␤1 are not available, integrin ␣3-null cells were used for subsequent experiments.
Integrin ␣3␤1 Binding to ␣3NC1 Domain Requires Residues 177-232-Three distinct binding sites for integrin ␣v␤3 have been mapped in the ␣3NC1 domain: a RGD site at the N terminus, a second site between amino acids 56 -75 and a third site between amino acids 185-203 (8) (Fig. 4A). To identify the binding sites for integrin ␣3␤1, we used ␣3/␣1NC1 chimeric proteins in which fragments of ␣3NC1 were replaced with homologous fragments of ␣1NC1 (1-3-1, 3-3-1, 1-3-3) (Fig.   FIGURE 3. Integrin ␣3␤1 is a receptor for ␣3NC1 domain. A, flow cytometry of mouse integrin ␣3-null (B10) or reconstituted with the human integrin ␣3 subunit (R10) cells incubated with increasing amounts of soluble ␣3NC1. ␣3NC1 binding is expressed as fold increase in fluorescence and represents the ratio between the mean fluorescence intensity of samples incubated with ␣3NC1 and controls incubated with mAb EB3 alone. Shown is a representative experiment of three independent experiments performed. B, analysis of ␣3NC1 and C6 chimera binding to immobilized integrin ␣3␤1. Immobilized integrin ␣3␤1 (1 g/ml) was incubated with ␣3NC1 domain or C6 at increasing concentrations and bound proteins were detected with mAb EB3. Specific binding was calculated as the difference between the absorbance of sample with and without integrin. The data represent the mean Ϯ S.D. of triplicate wells. The experiment was repeated twice with similar results. *, difference between ␣3NC1 and C6 were significant with p Ͻ 0.01. FIGURE 4. Integrin ␣3␤1 binding to ␣3NC1 requires the C-terminal third of ␣3NC1. A, schematic representation of ␣3NC1 and ␣1NC1 domains indicating the amino acid positions where fragments of ␣3NC1 were replaced with homologous fragments from ␣1NC1 in the ␣3/␣1NC1 chimeras. The black boxes indicate the binding sites for integrin ␣v␤3, 56 -75 and 185-203, and the gray box indicates the mAb EB3 epitope. B, analysis of ␣3/␣1NC1 chimeras binding to integrin ␣3␤1. Immobilized integrin ␣3␤1(1g/ml) was incubated with ␣3/␣1NC1 chimeras (10 g/ml) and binding measured as described in the legend to Fig. 3. The data represent the mean Ϯ S.D. of triplicate wells. The experiment was repeated twice with similar results. *, difference between 133 and C6 were significant with p Ͻ 0.05 and between 333 and C6 with p Ͻ 0.01 (**). C, flow cytometry of CT26 cells incubated with 20 g/ml ␣3/␣1 chimeras indicated. Shown is a representative experiment of two independent experiments performed.
CT26 Cells Proliferation Is Decreased by ␣3NC1-As HUVEC proliferation is decreased by the ␣3NC1 domain via interactions with ␣v integrins and cell adhesion to the ␣3NC1 domain is primarily mediated by the same integrins (8,28), we determined whether either of these cell functions could be induced by ␣3NC1 in the absence of integrin ␣v␤3. As shown in Fig. 5A, CT26 cell proliferation was inhibited in a dose-dependent manner with increasing concentrations of soluble ␣3NC1 domain. In contrast, there was minimal adhesion of CT26 cells to immobilized ␣3NC1 domain (Fig. 5B) at concentrations where endothelial cells adhered well (data not shown). CT26 cells, however, adhered well to their natural ligand fibronectin (Fig. 5B). Thus, the binding of soluble ␣3NC1 domain to CT26 cells expressing integrin ␣3␤1 correlates with the inhibition of cell proliferation. However, in the absence of integrin ␣v␤3, CT26 cell adhesion to immobilized ␣3NC1 domain is minimal.
Integrin ␣3␤1 Is a Trans-dominant Inhibitor of Integrin ␣v-Full-length and fragments of the ␣3NC1 domain do not decrease cell proliferation in all tumor cells that express integrin ␣v␤3 (4,9,23,28). However, as shown above, integrin ␣3␤1 binds to the ␣3NC1 domain in solution. Furthermore, it is well known that the affinity of ␣v integrins can be altered by the transdominant effect of other integrins (14,15). To test whether integrin ␣3␤1 can indeed alter the affinity of ␣v integrins, we measured the activation status of ␣v integrins in B10 and R10 cells. WOW-1, an antibody that recognizes only the active form of ␣v integrins (17), was used to measure ␣v integrin affinity. R10 and B10 cells express similar but low levels of ␣v integrin (18), and only 1.63% of R10 cells expressed activated ␣v integrins. In contrast, the percentage of B10 cells (4.25%) that bound WOW-1 in the absence of the ␣3NC1 domain was more than double that of R10 cells (Fig. 6A). This difference was enhanced in B10 cells (8.15%) by incubation with the ␣3NC1  domain suggesting that ␣3NC1 domain per se can activate integrin ␣v␤3. Together these data suggest that ␣v integrins are transdominantly inhibited by the expression of integrin ␣3␤1.
Cell adhesion assays were then performed to determine whether this transdominant effect changed cell adhesion on immobilized ␣3NC1 domains. As shown in Fig. 6B, R10 cells adhered significantly less than B10 cells. When Mn 2ϩ was added, R10 and B10 cell adhesion to ␣3NC1 increased, but the enhancement was greater with the R10 cells (5.5-fold versus 1.8-fold for 1.6 g/ml ␣3NC1). Thus ␣3␤1-dependent transdominant inhibition of ␣v integrin affinity decreases cell adhesion to the ␣3NC1 domain.
Since most cells expressing ␣v integrins also express ␣3␤1 integrin, we examined whether the expression levels of ␣3␤1 influences integrin ␣v affinity. We screened for human cell lines that expressed varying levels of integrins ␣3␤1 and ␣v␤3, as functional blocking antibodies to human ␣3 integrin are available. We found that melanoma HT-144 cells express high levels of ␣v␤3 and ␣3␤1, HUVECs express high levels of integrin ␣v␤3 but slightly lower levels of integrin ␣3␤1 than HT-144 cells, and the lung carcinoma cells A549 express similar levels of integrin ␣3␤1 but lower levels of integrin ␣v␤3 compared with HT-144 cells (Fig. 7A). As expected, the adhesion of these cells to the ␣3NC1 domain correlated with the levels of ␣v␤3 integrin with HT-144 and HUVEC cells adhering significantly more than A549 cells (Fig. 7B). In the presence of a ␣3 integrin blocking antibody, adhesion of A549 cells increased significantly. Interestingly, the adhesion of HT-144 cells to the ␣3NC1 domain also increased when integrin ␣3␤1 was blocked, although not to the same extent as for A549 cells. In contrast, this antibody had a minimal effect on HUVECs (which express the lowest levels of integrin ␣3␤1). Adhesion of all three cell lines to the ␣3NC1 domain is integrin ␣v␤3-dependent as antibodies to this integrin completely inhibited their adhesion (Fig.  7B). These data suggest that modulating both the expression and occupancy of integrin ␣3␤1 alters the affinity and adhesive functions of integrin ␣v␤3.

DISCUSSION
The efficacy of soluble ␣3NC1 domain as an anti-tumorigenic agent has been ascribed to its binding to ␣v␤3 integrin (2, 9, 11, 23, 28 -32). This binding was defined by either integrin ␣v-dependent cell adhesion or affinity chromatography with immobilized NC1 domain. To identify whether the physical immobilization influenced integrin binding, a flow cytometry assay was devised to explore the cellular receptors for soluble ␣3NC1 domain. This assay is similar to a novel flow cytometry method recently described to quantify integrin affinity and avidity changes at the single cell level (25). Unexpectedly, the nonclassical collagen receptor, integrin ␣3␤1, was shown to bind ␣3NC1 domain. CT26 cell proliferation was inhibited by soluble ␣3NC1 domain; however, these cells only minimally adhere to immobilized ␣3NC1 domain. These results reveal the limitation of assays that rely on immobilized ligands, when screening for receptors for soluble NC1 domain. In addition, we demonstrate that functional blocking of integrin ␣3␤1 transdominantly increases ␣v-integrin affinity for the ␣3NC1 domain. Taken together, these results raise the possibility that, in addition to ␣v integrins, integrin ␣3␤1 might play a role in the anti-tumorigenic effects of the ␣3NC1 domain by either directly affecting cell proliferation or altering the affinity of ␣v integrins on either tumor or endothelial cells. The identification of receptors that bind to the soluble form of NC1 domain is important as this is the physical state in which they are administered and exert their biological effects. Our result that integrin ␣3␤1 binds soluble ␣3NC1 domain, but cell adhesion is not mediated upon mobilization of the same ligand, highlights this point. Thus only performing assays with NC1 domain attached to fixed substrata might not identify critical receptors that are expressed by tumor cells and may mediate anti-tumoral effects.
We showed that the integrin ␣3␤1 binding site encompasses residues 177-232 of the ␣3NC1 domain. This region overlaps with residues 185-203 that constitutes a peptide that promotes adhesion of human melanoma cells and inhibits their proliferation in vitro (12). This peptide is proposed to mediate its effects by interacting with the integrin ␣v␤3 CD47-integrinassociated protein complex (31). However, our data suggest the mechanism of action of this peptide on tumor cell proliferation might be mediated through integrin ␣3␤1 as CT26 cells do not express integrin ␣v␤3, and residues 185-203 are encompassed in the integrin ␣3␤1 binding site on the ␣3NC1 domain.
Overexpression of the C-terminal fragment of ␣3NC1 (residues 183-232) inhibits tumor growth of B16F1 melanoma cells in vivo (12). Furthermore mice treated with a plasmid DNA encoding the ␣3NC1 domain develop smaller CT26 cell-derived tumors than control animals (27). Our results suggest that these in vivo anti-tumorigenic effects of the ␣3NC1 domain might be mediated by its binding to integrin ␣3␤1, rather than via proposed ␣v␤3-mediated effects on endothelial cells.
The mechanisms whereby the ␣3NC1 domain induces its anti-tumorigenic effects are becoming increasingly complex. Some groups proposed that its action is mediated by anti-angiogenic activity residing in residues 54 -132, while the C-terminal derivatives do not inhibit angiogenesis (4,9,23,28,30). Furthermore the same group proposed that these effects are mediated by both endogenous and exogenously administered ␣3NC1 domain. In contrast, others found that polypeptides encompassing amino acids 179 -208 within the C terminus have both anti-angiogenic (33) and anti-tumorigenic effects (11,13,29,(31)(32)(33). These effects have only been observed with exogenously administered ␣3NC1 domain or when the ␣3NC1 domain is produced by tumor cells. We now demonstrate that the C-terminal third of the ␣3NC1 domain interacts with integrin ␣3␤1 where it decreases tumor cell proliferation in vitro. Taken together the data suggest that the anti-angiogenic affects of the ␣3NC1 domain in vivo might be mediated by both the C and N terminus of the domain. In contrast, the anti-tumorigenic effects are mediated only by the C terminus of the domain, via interactions with either integrin ␣v␤3 or ␣3␤1. The relative expression levels of these integrins on tumor cells in vivo might determine their response to exogenous ␣3NC1 domain exposure.
We demonstrated that both expression levels and occupancy of integrin ␣3␤1 by ligand alters integrin ␣v␤3 affinity. Endothelial cells and most carcinomas express both of these integrins in varying amounts. Thus, soluble NC1 domain interactions with integrin ␣3␤1 might alter ␣v␤3-dependent cell functions by transdominant activation or inhibition. This mechanism of modulating integrin function is well described in ␣v integrins. Expression of integrin ␣5␤1 reduces ␣v␤3-mediated adhesion (14), and ␣v␤3-mediated endothelial cell migration is altered by the ligation state of integrin ␣5␤1 (14). Furthermore, antibodies against ␣v␤3 can inhibit both integrin ␣5␤1-mediated phagocytosis (34) as well as ␣3␤1/␣6␤1-mediated cell adhesion to ␣4-laminin (35). Moreover, the observation that cells lacking integrin ␣3␤1 adhere to fibronectin and collagen better than wild type cells (15) shows that integrin ␣3␤1 is also a trans-dominant inhibitor of integrin activation. Thus, our data that the ligation state and expression levels of integrin ␣3␤1 can exert significant alterations on integrin ␣v␤3 function might explain why certain tumor cells are not responsive to the anti-proliferative effects of the ␣3NC1 domain despite their expression of integrin ␣v␤3.
In conclusion, we demonstrate that the ␣3NC1 domain binds integrin ␣3␤1, and cell proliferation is decreased in cells that lack integrin ␣v␤3 but express integrin ␣3␤1. In addition, the ligation state of integrin ␣3␤1 can modulate the affinity of ␣v␤3 integrin, a key integrin required for tumor angiogenesis. Thus, integrin ␣3␤1 might be a direct mediator of the anti-tumorigenic actions of the ␣3NC1 domain. In addition, interactions between the ␣3NC1 domain and ␣3␤1 integrin might play a key role in the inhibition of tumor angiogenesis by altering ␣v␤3 integrin affinity on endothelial cells.