Regulation of integrin function by CD47 ligands. Differential effects on alpha vbeta 3 and alpha 4beta1 integrin-mediated adhesion.

We examined the regulation of α4β1 integrin function in melanoma cells and T cells by ligands of CD47. A CD47 antibody (B6H12) that inhibited αvβ3-mediated adhesion of melanoma cells induced by CD47-binding peptides from thrombospondin-1 directly stimulated α4β1-mediated adhesion of the same cells to vascular cell adhesion molecule-1 and N-terminal regions of thrombospondin-1 or thrombospondin-2. B6H12 also stimulated α4β1- as well as α2β1- and α5β1-mediated adhesion of CD47-expressing T cells but not of CD47-deficient T cells. α4β1 and CD47 co-purified as a detergent-stable complex on a CD47 antibody affinity column. CD47-binding peptides based on C-terminal sequences of thrombospondin-1 also specifically enhanced adhesion of melanoma cells and T cells to α4β1 ligands. Unexpectedly, activation of α4β1 function by the thrombospondin-1 CD47-binding peptides also occurred in CD47-deficient T cells. CD47-independent activation of α4β1required the Val-Val-Met (VVM) motif of the peptides and was sensitive to inhibition by pertussis toxin. These results indicate that activation of α4β1 by the CD47 antibody B6H12 and by VVM peptides occurs by different mechanisms. The antibody directly activates a CD47-α4β1 complex, whereas VVM peptides may target an unidentified Gi-linked receptor that regulates α4β1.

CD47 (integrin-associated protein) is an integral membrane protein that is required for granulocyte and T cell recruitment to sites of infection (1,2), and its absence on red blood cells leads to their rapid macrophage-mediated clearance (3). CD47 may also function as a costimulator to regulate T cell activation, survival, and Th1 versus Th2 differentiation (4,5). Endogenous ligands for the extracellular domain of CD47 include the secreted protein thrombospondin-1 (TSP1) 1 and potentially other members of the thrombospondin family, several integrins (6 -9), and some members of the signal-regulatory protein family (3, 10 -13). Engagement of CD47 by soluble ligands or sig-nal-regulatory protein counter receptors modulates several cell signaling pathways, including activation of a heterotrimeric G protein (14).
Although some signal transduction through CD47 is integrin-independent (4,15,16), association of CD47 with certain integrins was found to modulate their function to mediate cell adhesion or motility (6,8,9,17). In addition to its known functional and physical interactions with ␣ v ␤ 3 , ␣ IIb ␤ 3 , and ␣ 2 ␤ 1 integrins, several publications have suggested an association of CD47 with ␣ 4 ␤ 1 integrin. CD47 and ␣ 4 ␤ 1 were colocalized on microvilli of K562 erythroleukemia cells (18), and CD47-dependent arrest of T cells on inflammatory endothelium could be blocked by antibodies that prevent ␣ 4 ␤ 1 integrin binding to VCAM-1 (2). In the latter study, ligation of CD47 by TSP1 or signal-regulatory protein 1␣ was inferred to activate ␣ 4 ␤ 1 integrin on T cells, although this was not demonstrated either functionally or biochemically.
We recently found that CD47 expression is required for stimulation of T cell motility and expression of MMP-2 stimulated by ␣ 4 ␤ 1 integrin ligands (19). An antibody to CD47 also blocked the motility response to an ␣ 4 ␤ 1 integrin ligand (19). Therefore, at least a functional interaction occurs between CD47 and ␣ 4 ␤ 1 integrin.
We identified a binding site for ␣ 4 ␤ 1 integrin in the Nterminal domains of TSP1 and TSP2 (19), whereas two binding sites for CD47 have been localized to the C-terminal domain of TSP1 (for review, see Ref. 20). Thus, TSP1 could potentially regulate its own interaction with ␣ 4 ␤ 1 integrin through simultaneous interactions with CD47 and ␣ 4 ␤ 1 integrin on a T cell, analogous to the ability of soluble TSP1 to stimulate ␣ v ␤ 3 integrin-dependent melanoma cell adhesion to immobilized TSP1 (21).
To define a molecular basis for functional cross-talk between these two receptors, we have examined the effect of CD47 ligands on the function of ␣ 4 ␤ 1 integrin in T cells and melanoma cells. We confirmed that ligation of CD47 can activate ␣ 4 ␤ 1 integrin function but unexpectedly found the mechanism of this activity to be CD47 ligand-dependent. We further show that CD47 ligation differentially activates and inactivates ␣ v ␤ 3 and ␣ 4 ␤ 1 in melanoma cells and ␣ 4 ␤ 1 and ␣ 5 ␤ 1 in Jurkat cells. Unexpectedly, but consistent with a recent report demonstrating CD47-independent induction of platelet aggregation by a CD47-binding peptide from TSP1 (22), we found that effects of the CD47-binding peptide derived from TSP1 on ␣ 4 ␤ 1 -dependent T cell adhesion are CD47-independent.

EXPERIMENTAL PROCEDURES
Cell Culture and Proteins-A2058 melanoma cells and Jurkat T cells were grown in RPMI 1640 medium supplemented with 10% fetal calf serum, penicillin, and streptomycin. A CD47-deficient T cell line derived from Jurkat cells (JinB8) was graciously provided by Dr. Eric Brown (University of California, San Francisco, CA), and ␤ 1 * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. § Present address: Laboratory of Immunology, NEI, National Institutes of Health, Bldg. 10, 10B22, Bethesda, MD 20892.
Cell Adhesion Assays-Adhesion was assessed using a microscopic assay. TSP1, recombinant proteins, or S7D-VCAM-1 diluted in Dulbecco's PBS or NaHCO 3 buffer were absorbed on bacteriological polystyrene dishes overnight at 4°C. The dishes were blocked with 1% BSA in PBS and prewarmed before adding the cells diluted in RPMI containing 1 mg/ml BSA. After a 15-min incubation for T cells or a 60-min incubation for melanoma cells at 37°C, nonadherent cells were washed off gently, and the remaining attached cells were fixed in 1% glutaraldehyde in PBS and stained with Diff-Quik (Dade International). Spread and attached cells were quantified by counting using a calibrated reticle.
In some experiments, the area of spread cells were measured using ImagePro software. Approximately 100 cells were measured for each condition. The data were analyzed using a two-sided t test.
Alternatively, matrix protein-mediated cell adhesion was measured using a colorimetric assay as previously described (30). TSP1 and the NoC proteins were coated in PBS overnight at 4°C. Fresh T cell cultures (Ͻ5 ϫ 10 5 cells/ml) were resuspended at 2 ϫ 10 5 cells/ml in RPMI containing 0.1% BSA. The plates were chilled in a 4°C bath, and 100 l of cell suspension with the indicated treatments was added into each well. The plates were then incubated at 37°C for 15 min. Unbound cells were removed by washing, and adherent cells were quantified by hexosaminidase assay (30).
VCAM-1 Cell Binding Assay-S7D-VCAM-1 was labeled with 125 I using lodogen (Pierce). Jurkat cells were washed with 4°C chilled Dulbecco's PBS (without Ca 2ϩ or Mg 2ϩ ) and resuspended in chilled binding buffer (RPMI with 0.1% BSA) at 3 ϫ 10 6 cells/ml. On ice, 100 l of the cell suspension was premixed with the indicated concentrations of TS2/16 alone or in combination with different concentrations of B6H12. 125 I-VCAM-1, diluted in chilled binding buffer, was added into each tube to bring the final volume to 200 l. The tubes were mixed by vortexing and transferred to a 37°C water bath for 15 min. The cell suspensions were then transferred to plastic tubes containing 100 l of Nyosil oil (William F. Nye, Inc), centrifuged for 1 min, and washed with 200 l of cell binding buffer. The pellets were collected, and the bound radioactivity was quantified.
Western Blotting-Proteins were fractionated on SDS gels and transferred to polyvinylidene difluoride membranes. The membranes were blocked with PBS containing 3% BSA and 0.1% Tween 20. The primary antibody was added in the presence of blocking buffer and allowed to incubate while rocking. After repeated washes with PBS containing 0.1% Tween, a horseradish peroxidase-conjugated secondary antibody was added diluted in blocking buffer. The membranes were washed with PBS-Tween, and antigen antibody complex was visualized using chemiluminescent substrate (Pierce).
Because TSP1 contains an ␣ v ␤ 3 recognition sequence in its type 3 repeats and an ␣ 4 ␤ 1 binding site in the N-terminal domain, we used recombinant N-terminal portions of TSP1 and TSP2 that contain only their ␤ 1 integrin binding sites to examine the effect of the CD47-binding peptide on ␤ 1 integrin-mediated melanoma cell adhesion to TSP1 and TSP2 (Fig. 1A). Peptide FIRVVMYEGKK but not the control peptide FIRGG-MYEGKK stimulated spreading on both NoC1 and NoC2 to a comparable extent as the ␤ 1 integrin-activating antibody TS2/ 16. Melanoma cell spreading on NoC1 and NoC2 stimulated by peptide FIRVVMYEGKK was reversed by a ␤ 1 integrin blocking antibody or the ␣ 4 integrin antagonist but was not significantly inhibited by the ␣ v integrin antagonist (Fig. 1A), suggesting that ␣ 4 ␤ 1 integrin function, like that of ␣ v ␤ 3 integrin, is regulated by CD47.
Stimulation of ␣ 4 ␤ 1 integrin-mediated spreading of melanoma cells by the TSP1 peptide was verified using the well defined ␣ 4 ␤ 1 integrin ligand VCAM-1 (Fig. 1B). The CD47binding TSP1 peptide FIRVVMYEGKK also stimulated spreading on intact TSP1 and on limiting concentrations of type I collagen, an ␣ 2 ␤ 1 -specific substrate for these cells. Therefore, FIRVVMYEGKK increases spreading mediated by several integrins on melanoma cells.
Stimulation of ␣ v ␤ 3 integrin-mediated spreading by TSP1 peptides that bind to CD47 was previously shown to be blocked by the CD47 antibody B6H12 (6). We confirmed the inhibitory activity of B6H12 for TSP1 peptide-induced melanoma cell attachment and spreading on limiting concentrations of the ␣ v ␤ 3 integrin ligand vitronectin ( Fig. 2A). In contrast, the CD47 antibody did not inhibit spreading stimulated by FIRVVMYEGKK on the ␣ 4 ␤ 1 ligands NoC1 or NoC2 and, instead, further stimulated melanoma cell attachment on NoC1 and NoC2 (Fig. 2A).
These data suggested that CD47 ligation by B6H12 may directly stimulate ␣ 4 ␤ 1 integrin-mediated adhesion. This was confirmed by examining the effect of B6H12 in the absence of other CD47 ligands (Fig. 2B). B6H12 markedly enhanced both attachment and spreading of melanoma cells on NoC2, whereas it somewhat inhibited basal attachment and spreading of the same cells on vitronectin. B6H12-induced enhancement of ␣ 4 ␤ 1 integrin-dependent adhesion was specific in that a second CD47 antibody, C1Km1, was inactive (Fig. 2B). Similar enhancement of adhesion by B6H12 was observed on the ␣ 4 ␤ 1 ligand VCAM-1 (results not shown). The same antibody moderately stimulated ␣ 2 ␤ 1 integrin-dependent spreading on type I collagen but had no effect on spreading on an ␣ 3 ␤ 1 integrin binding sequence from TSP1 (results not shown). Involvement of different integrins in the opposing responses to B6H12 on vitronectin and NoC2 was confirmed using a specific ␣ v inte-grin antagonist, which reversed FIRVVMYEGKK-enhanced spreading on vitronectin but not on NoC2 (results not shown). B6H12 was previously considered to be a function-blocking antibody for CD47 (7,17,(31)(32)(33)(34)(35), but the above results demonstrate that this CD47 antibody can both positively and negatively modulate integrin functions. Furthermore, enhancement of some TSP1 peptide responses by the antibody indicated that B6H12 does not directly inhibit binding to their cellular targets.
CD47 Ligands Stimulate ␣ 4 ␤ 1 Integrin-mediated Adhesion of T Cells-The preceding data demonstrated that ligation of CD47 by the B6H12 antibody can have opposing effects on the activities of different integrins, inhibiting ␣ v ␤ 3 but stimulating ␣ 4 ␤ 1 and ␣ 2 ␤ 1 integrins. Because melanoma cells express both ␣ 4 ␤ 1 and ␣ v ␤ 3 integrins, however, it was possible that the activation of ␣ 4 ␤ 1 integrin was mediated by cross-talk with ␣ v ␤ 3 integrin (see Ref. 36), which in turn was modulated by engaging CD47 (20). We therefore used Jurkat T cells, which highly express ␣ 4 ␤ 1 integrin but lack significant ␣ v ␤ 3 integrin expression (results not shown), to examine the ability of CD47 ligation to modulate ␣ 4 ␤ 1 integrin activity independent of ␣ v ␤ 3 integrin. B6H12 induced a similar enhancement of Jurkat T cell adhesion on ␣ 4 ␤ 1 integrin ligands (Fig. 4). Although Jurkat cells attached somewhat on immobilized VCAM-1 without activation (Fig. 4A), the CD47 antibody B6H12 antibody significantly increased the number of Jurkat cells attached on immobilized VCAM-1 3-fold and increased their mean spread area by 26% (p Ͻ 0.01, Fig 4A). This compared with a 51% increase in cell area in the presence of the ␤ 1 integrin-activating antibody (p Ͻ 0.001, Fig. 4A). Similar enhancements of Jurkat cell adhesion by B6H12 were observed on NoC1 and NoC2 (Fig. 4,  A-B). As observed using melanoma cells, the CD47 antibody C1Km1 was much less active (Fig. 4B). B6H12-induced adhesion of Jurkat cells to NoC1 and NoC2 was inhibited by a ␤ 1 -blocking antibody and by the ␣ 4 ␤ 1 antagonist (Fig. 4C).
B6H12 also stimulated Jurkat cell adhesion to native TSP1 and to a recombinant N-module of TSP1 containing the ␣ 4 ␤ 1 binding site, TSP1 (1-175) (Fig. 4D). The activity of B6H12 required both CD47 and ␤ 1 integrin expression, because the stimulatory activity of B6H12 was absent in Jurkat mutants lacking either receptor (Fig. 4D).
Surprisingly, although B6H12 stimulated ␣ 4 ␤ 1 -mediated adhesion of Jurkat cells, binding of the soluble ␣ 4 ␤ 1 ligand, VCAM-1, to the same cells was diminished in a dose-dependent manner by B6H12 (Fig. 5). Therefore, the increased adhesion probably results from increased integrin avidity rather than from an enhancement of ␣ 4 ␤ 1 integrin affinity by this antibody.
Consistent with the activity of the B6H12 on Jurkat cells, FIRVVMYEGKK stimulated ␣ 4 ␤ 1 integrin-mediated adhesion on ␣ 4 ␤ 1 -dependent substrates for either unstimulated or TS2/ 16-stimulated Jurkat cells (Fig. 6A). The response to FIRVV-MYEGKK on a TSP1 substrate was stronger than on the NoC2 fragment containing only the ␣ 4 ␤ 1 integrin binding site (Fig.  6A), suggesting that interaction of the TSP1 type 3 repeats with ␣ 5 ␤ 1 integrin may also be stimulated. To detect changes in ␣ 5 ␤ 1 -mediated adhesion specifically, we used a 33-kDa recombinant cell binding portion of fibronectin containing only this integrin recognition site (25). Adhesion on FN33 was stimulated by either FIRVVMYEGKK (Fig. 6A) or B6H12 (Fig. 6B). In contrast, FIRVVMYEGKK only weakly stimulated T cell adhesion on type I collagen. However, after activation of the cells using the ␤ 1 integrin antibody, the peptide further stimulated adhesion on type I collagen (Fig. 6A). In contrast, the CD47 antibody B6H12 stimulated spreading of unstimulated T cells on collagen (Fig. 6B). As was found in melanoma cells, the B6H12 antibody did not inhibit stimulation by the TSP1 peptide, although significant additivity was not detected for any of the ligands tested.
T Cell Responses to Some CD47 Ligands Are CD47-independent-CD47 binding antibodies have been shown to act as both agonists and antagonists of specific CD47 responses (4,38). Therefore, the differences in responses to integrin ligands induced by B6H12 and FIRVVMYEGKK in Fig. 6B could be explained by their acting as selective agonists of CD47. However, the recent evidence that a related CD47-binding peptide from TSP1 modulates platelet aggregation independent of CD47 (22) suggested an alternate explanation for these results.
To determine whether modulation of T cell adhesion by the CD47-binding peptides required CD47 expression, we compared responses in wild type and CD47-deficient Jurkat cells (Fig. 8). In contrast to B6H12 (see Fig. 4D), peptides containing either of the known CD47-binding sequences from TSP1 had equivalent stimulatory activities for adhesion of Jurkat cells expressing or lacking CD47 (Fig. 8A). However, activities of peptides derived from both VVM motif regions in TSP1 in the CD47-deficient mutant were specific because control peptides with amino acid substitutions shown previously to ablate CD47 binding were inactive.
Notably, FIRVVMYEGKK containing the second VVM motif of TSP1 activated T cell adhesion to TSP1 to a greater extent than an optimal concentration of the ␤ 1 integrin-activating antibody TS2/16 (Fig. 8A), suggesting that this stimulatory activity might also be ␤ 1 integrin-independent. However, stimulation of adhesion by FIRVVMYEGKK was markedly diminished in ␤ 1 -deficient Jurkat cells (Fig. 8B). Similarly reduced stimulation of adhesion was observed using NoC2 (results not shown), but the peptide did not increase adhesion on fibronectin (Fig. 8B). The ␤ 1 -deficient clone lacks any detectable ␤ 1 integrin by flow cytometry and is unresponsive to a ␤ 1 -function stimulating antibody but has normal levels of cell surface CD47 (19). Therefore, stimulation of T cell adhesion on fibronectin by the peptide is entirely ␤ 1 integrin-dependent, but an alternate FIG. 3. Stimulation of ␣ 4 ␤ 1 integrin-mediated melanoma cell spreading by the CD47 antibody B6H12 and a CD47 binding TSP1 peptide is synergistic with ␤ 1 integrin activation. Melanoma cells were allowed to adhere on substrates coated with suboptimal concentrations of TSP1 (15 g/ml) or NoC2 (15 g/ml). Melanoma cells were either untreated (control) or treated with anti-CD47 antibody (B6H12, 5 g/ml) or the ␤ 1 integrin-activating antibody (TS2/ 16, 5 g/ml) alone or in combination with B6H12 (5 g/ml) or a CD47binding peptide from TSP1 (p7N3, 5 M). The cells were also treated with TSP1 peptide FIRVVMYEGKK (p7N3, 5 M) alone or control peptide FIRGGMYEGKK (p604, 5 M) alone. Spread cells are presented as mean Ϯ S.D., n ϭ 3.
adhesion receptor may also contribute to the TSP1 peptide enhancement of adhesion on TSP1 and NoC2. ␤ 1 integrins are required, however, for maximal responses on TSP1 and NoC2 substrates.
CD47-mediated responses to the TSP1 peptides were shown previously to require pertussis toxin-sensitive G proteins (6,9,14,17), whereas the CD47-independent activity of these peptides reported in platelets was sensitive to Src inhibitors (22). Remarkably, the stimulation by FIRVVMYEGKK of adhesion on limiting concentrations of TSP1 or NoC2 was equally sensitive to pertussis toxin in both the wild type and CD47-deficient Jurkat cells (Fig. 9), but the Src inhibitor PP2 had no effect on either response. Therefore, the adhesion responses to VVM peptides in both T cell lines differ from the aggregation response of platelets to the same peptides and involve G protein signaling that is independent of CD47. DISCUSSION We have identified two additional integrins that are functionally regulated by CD47, ␣ 4 ␤ 1 and ␣ 5 ␤ 1 . Two CD47 ligands, VVM-containing peptides and the anti-CD47 antibody B6H12, stimulated cell attachment or spreading mediated by these integrins. Although activity of the CD47 antibody is clearly dependent on CD47 and we show that CD47 physically associates with ␣ 4 ␤ 1 integrin, at least part of the activities of the VVM-containing peptides to stimulate ␣ 4 ␤ 1 -dependent adhesion are independent of CD47. The stimulatory activity of B6H12 for ␤ 1 integrin responses contrasts with the inhibitory activities of B6H12 for ␣ v ␤ 3 (6) and ␣ IIb ␤ 3 integrin functions (9). Ligation of CD47 by B6H12, therefore, can selectively enhance and inhibit the function of different integrins. This dif- FIG. 4. Regulation of ␣ 4 ␤ 1 integrin-mediated adhesion in Jurkat T cells. A, ␣ 4 ␤ 1 integrin activity is induced by B6H12 in T cells. Jurkat cells were allowed to adhere on substrates coated with 15 g/ml NoC2 or 3 g/ml VCAM-1 in the absence (control) or presence of the ␤ 1 -activating antibody (TS2/16, 5 g/ml) or anti-CD47 antibody (B6H12, 5 g/ml). B, two CD47 antibodies differentially regulate ␣ 4 ␤ 1 integrin-dependent adhesion of T cells. Jurkat cells were allowed to adhere to uncoated polystyrene wells (no coating) or wells coated with NoC1 (10 g/ml) or NoC2 (10 g/ml). The cells were untreated (Control) or treated with B6H12 or C1Km1 at the indicated concentrations in g/ml. Adhesion was determined by assay of hexosaminidase activity and is presented as mean Ϯ S.D., n ϭ 3. C, adhesion of Jurkat cells on wells coated with BSA, 10 g/ml NoC1, or 10 g/ml NoC2 was determined using the colorimetric assay for unstimulated cells (control) or cells stimulated with 8.8 g/ml B6H12 alone or in the presence of 6.7 g/ml of the ␤ 1 blocking antibody mAb13 or 1 M of the ␣ 4 antagonist. D, adhesion of wild type Jurkat cells, the CD47-deficient mutant, or the ␤ 1 integrin-deficient mutant on substrates coated using 7 g/ml TSP1 (solid bars) or 20 g/ml TSP1 (1-175) (striped bars) was determined for unstimulated cells (control) and cells treated with 13 g/ml B6H12. ferential effect on integrin activities may be important for understanding the biological functions of CD47.
Although ligation of CD47 is now known to stimulate functions of three ␤ 1 integrins, this response is not universal in that ␣ 3 ␤ 1 integrin function was not stimulated by a CD47-binding peptide (39). As previously demonstrated for ␤ 3 integrins and ␣ 2 ␤ 1 , CD47 is physically associated with ␣ 4 ␤ 1 in cells, where it regulates function of this ␤ 1 integrin. Both CD47 (40,41) and ␣ 4 ␤ 1 integrin (42) associate with lipid rafts in Jurkat cells, suggesting that their interaction may involve these membrane microdomains. Notably, ␣ v ␤ 3 activity is increased when CD47 translocates out of raft domains (43).
Function of the ␣ v ␤ 3 integrin is modulated by CD47, and this integrin can in turn regulate ␣ 4 ␤ 1 integrin (36). In this study, however, we show that regulation of ␣ 4 ␤ 1 integrin by CD47 is independent of ␣ v ␤ 3 integrin expression. Instead CD47 may modulate ␣ 4 ␤ 1 integrin function by physically associating with the integrin. Association of integrins with other extracellular proteins has been generally mapped to the extracellular domains of both the integrin and the associating membrane protein (44 -46). Consistent with this finding, regulation of ␣ 4 ␤ 1dependent adhesion by CD47 required the presence of only the extracellular domain and a single transmembrane domain (2).
Although some activities of the CD47-binding peptide from TSP1 are clearly mediated by CD47, the peptides must also interact with a different receptor. Activity of these TSP1 peptides in CD47-deficient platelets was partially dependent on FcR ␥ expression (22), but Jurkat cells do not express this receptor (47) and do not show the same sensitivity to Src inhibition as was reported in platelets (22). Understanding the role of CD47 in response to the TSP1 peptide is further complicated by our observation that the CD47-independent pathway shares the sensitivity to pertussis toxin that characterizes CD47 signaling. Until the second receptor for the TSP1 peptides is identified and its ability to recognize intact TSP1 is assessed, activities of these peptides should be interpreted with caution. Furthermore, recognizing that the CD47-binding peptides from TSP1 can, in at least two cell types, act independent of CD47 may necessitate a reexamination of the conclusion that CD47 "function-blocking" antibodies block a CD47-signaling pathway. In almost all such cases, the reported blocking of CD47 responses represents antagonism of TSP1 peptide responses (6,17,32). An equally plausible hypothesis is that the CD47 antibody indirectly antagonizes signaling stimulated by binding of the TSP1 peptides to a receptor other than CD47. In this case, the observed antagonism could be a negative crosstalk between two agonist pathways.
With a few exceptions (4), the CD47 antibody B6H12 has been generally found to block responses mediated by other CD47 ligands (7,17,(31)(32)(33)(34)(35). Our data confirm that B6H12 reverses the activation of ␣ v ␤ 3 integrin function stimulated by a CD47-binding peptide from TSP1, but we also show that the antibody has a direct stimulating function for at least two ␤ 1 integrins in the same cells. This stimulating function was confirmed in both melanoma cells and T cells. Therefore, it may be proper to reclassify B6H12 as a function-modifying antibody for CD47. A function blocking antibody recognizing ␤ 1 integrins was similarly shown to allosterically modulate ligand binding (48). Because B6H12 has opposing effects on ␤ 1 and ␤ 3 integrin function in melanoma cells, both the positive and negative effects of B6H12 on integrin function may result from direct or allosteric regulation of CD47 association with or signaling to integrins rather than direct inhibition of ligand binding to CD47.
In T cells, peptide FIRVVMYEGKK and a ␤ 1 integrin-activating antibody had additive effects on adhesion to TSP1 and FIG. 6. CD47 ligands differentially regulate function of several integrins in T cells. A, A CD47-binding peptide activates ␣ 5 ␤ 1 integrin in Jurkat cells. Jurkat cells were allowed to adhere onto TSP1 (16 g/ml), a recombinant portion of TSP2 containing its ␣ 4 ␤ 1 binding site (NoC2, 12 g/ml), a recombinant portion of fibronectin containing its ␣ 5 ␤ 1 binding site (FN33 25 g/ml), or the ␣ 2 ␤ 1 ligand type I collagen (25 g/ml). Cells were either untreated (control) or treated with CD47binding peptide FIRVVMYEGKK (p7N3, 10 M), control peptide FIR-GGMYEGKK (p604, 10 M), or the ␤ 1 -activating antibody (TS2/16, 4 g/ml) alone or in combination with p7N3 or p604. Attached cells are presented as mean Ϯ S.D., n ϭ 3. B, anti-CD47 antibody B6H12 activates ␣ 5 ␤ 1 integrin in Jurkat cells. Jurkat cells were allowed to adhere onto TSP1 (16 g/ml), NoC2 (12 g/ml), FN33 (25 g/ml), and type I collagen (25 g/ml). Cells were either untreated (control) or treated with CD47-binding peptide FIRVVMYEGKK (p7N3, 10 M) or anti-CD47 antibody (B6H12, 20 g/ml) alone or in combination with p7N3. Attached cells are presented as mean Ϯ S.D., n ϭ 3. collagen. However, the peptide had minimal activity in the absence of the ␤ 1 integrin-activating antibody, suggesting that TSP1 signaling through CD47 is insufficient to activate ␣ 2 ␤ 1 integrin in these cells. The additivity also suggests that the effect of this signal is to increase integrin avidity rather than affinity. Likewise, the maximal dose of the TSP1 peptide only stimulated ϳ50% of T cells to attach on an ␣ 5 ␤ 1 integrin ligand, whereas the addition of the ␤ 1 -activating antibody further stimulated ␣ 5 ␤ 1 -dependent adhesion at saturating doses of the TSP1 peptide. This differs from the strong stimulation of ␣ 4 ␤ 1mediated adhesion by the CD47-binding peptide. Thus, the TSP1 peptide differentially affects three ␤ 1 integrins in the T cells. The mechanisms for differential cross-talk between this signal and each integrin remain to be defined.
B6H12 was previously reported to delay neutrophil transmigration toward formylmethionylleucylphenylalanine (49). B6H12 also inhibited chemotaxis of endothelial cells stimulated by TSP1 and its CD47-binding peptide (32). We found that B6H12 inhibited T cell chemotaxis to TSP1, NoC1, and NoC2 (19). Because NoC1 and NoC2 do not contain the CD47 binding sites, we concluded that CD47 ligation must indirectly inhibit migration (19). This inhibition was interpreted as indicating a positive requirement for CD47 in ␣ 4 ␤ 1 -mediated chemotaxis, consistent with the absence of motility in CD47-deficient cells (19), but the same results could also be obtained if B6H12 binding to CD47 increased the avidity of ␣ 4 ␤ 1 . Inhibition of cell motility has been previously observed as a result of activation as well as inhibition of integrins (50,51). B6H12 may, therefore, inhibit migration by activating integrins.
CD47 and ␣ 4 ␤ 1 integrin are two signaling receptors that sense the extracellular environment. Because TSP1 is a ligand for both receptors, we examined the roles of each in mediating responses of T cells to TSP1 (19). For mediating adhesion on TSP1, ␣ 4 ␤ 1 was necessary and sufficient (19). However, the present results show that ligation of CD47 can modulate this response by activating ␣ 4 ␤ 1 . For stimulating chemotaxis to TSP1, the integrin and CD47 are both necessary, but direct interaction of TSP1 with the integrin is sufficient to stimulate chemotaxis. The same mechanism applies to regulation of matrix metalloproteinase gene expression; CD47 and ␣ 4 ␤ 1 are both necessary, but the role of CD47 is indirect. Conversely, for inhibiting T cell receptor signaling or T cell proliferation, CD47 is necessary, but ␣ 4 ␤ 1 integrin is not. Therefore, we concluded that CD47 and ␣ 4 ␤ 1 integrin are each functional signaling receptors for TSP1 in T cells that elicit distinct signaling pathways (19). In some cases, these pathways engage in cross-talk. In addition to potential cross-talk between their downstream effectors, the present data demonstrate that CD47 may modulate ␣ 4 ␤ 1 integrin function through lateral interactions in the plasma membrane. We found that ligation of CD47 can increase ␣ 4 ␤ 1 integrin activity to mediate cell adhesion, but these lateral interactions may also modulate outside-in signaling through ␣ 4 ␤ 1 integrin. The mechanism of this regulation and the possible requirement for other membrane proteins to mediate CD47-␣ 4 ␤ 1 cross-talk remains to be defined. FIG. 8. CD47-binding peptides from TSP1 stimulate T cell adhesion independent of CD47. A, wild type (solid bars) and CD47 deficient Jurkat cell adhesion on TSP1 (striped bars) is stimulated by CD47-binding peptides. Unstimulated Jurkat and JinB8 cells were allowed to adhere on plates coated with 20 g/ml TSP1. The cells were either not treated (control) or treated with CD47-binding TSP1 peptide FIRVVMYEGKK (p7N3, 10 M), the p7N3 control peptide FIRGGMY-EGKK (p604, 10 M), another CD47 binding TSP1 peptide RFYV-VMWK (p4N1, 20 M), the p4N1 control peptide RFYGGMWK (p4N1GG, 20 M), or 5 g/ml ␤ 1 integrin-activating antibody (TS2/16). Attached cells are presented as the mean Ϯ S.D., n ϭ 3. B, ␤ 1 integrin expression is required for maximal stimulation of adhesion by TSP1 peptide 7N3. Wild type or ␤ 1 integrin-deficient cells were allowed to adhere to TSP1 (20 g/ml) or fibronectin (FN, 10 g/ml). Cells adhering to TSP1 were either untreated (control) or treated with CD47-binding peptide FIRVVMYEGKK (p7N3, 10 M), p7N3 control peptide FIRGG-MYEGKK (p604, 10 M), the CD47-binding TSP1 peptide RFYVVMWK (p4N1, 20 M), the p4N1 control peptide RFYGGMWK (4NGG, 20 M), or Lys-modified p4N1 KRFYVVMWKK (p4N1K, 5 M). Attached cells are presented as mean Ϯ S.D., n ϭ 3.
FIG. 9. CD47-independent stimulation of adhesion is pertussis toxin-sensitive. Wild type Jurkat cells or CD47-deficient cells were allowed to adhere to TSP1 (20 g/ml)-coated plates. The cells were either not treated (control) or pretreated for 1 h with 5 M PP2 or 1 g/ml pertussis toxin (PT). Attached cells in the presence or absence of the CD47-binding peptide FIRVVMYEGKK (p7N3, 10 M) are presented as mean Ϯ S.D., n ϭ 3.