Identification of Novel (cid:1) 1 Integrin Binding Sites in the Type 1 and Type 2 Repeats of Thrombospondin-1*

In addition to the three known (cid:1) 1 integrin recognition sites in the N-module of thrombospondin-1 (TSP1), we found that (cid:1) 1 integrins mediate cell adhesion to the type 1 and type 2 repeats. The type 1 repeats of TSP1 differ from typical integrin ligands in that recognition is pan- (cid:1) 1 -specific. Adhesion of cells that express one dominant (cid:1) 1 integrin on immobilized type 1 repeats is specifically inhibited by antagonists of that integrin, whereas adhesion of cells that express several (cid:1) 1 integrins is partially inhibited by each (cid:2) -subunit-specific antagonist and completely inhibited by combining the antagonists. (cid:1) 1 integrins recognize both the second and third type 1 repeats, and each type 1 repeat shows pan- (cid:1) 1 specificity and divalent cation dependence for promoting cell adhesion. Adhesion to the A TSR123 affinity column was prepared using a CNBr-activated agarose bead matrix (following the manufacturer’s protocol). Surface-biotinylated Jurkat cells (2–5 (cid:1) 10 8 ) were lysed in the presence of 5–10 ml of buffer containing 100 m M n -octylglucoside, 50 m M Tris-HCl, pH 7.4, 150 m M NaCl, 1 m M MgCl 2 , 1 m M CaCl 2 , and 1 m M phenylmethylsulfonyl fluoride. After removal of insoluble material by centrifugation at 3000 (cid:1) g for 20 min, the extract was concentrated to 1 ml using Centripep 10 (Millipore). The column was previously washed with equilibration buffer containing 25 m M n -octylglucoside, 50 m M Tris-HCl, pH 7.4, 150 m M NaCl, 2 m M MgCl 2 , 0.1 m M CaCl 2 , and 0.5 m M MnCl 2 . The cell extract was applied to the column and incubated fo r 1 h atroom temperature or overnight at 4 °C. The column was washed with equilibration buffer until protein disappeared from the eluate, then washed again with washing buffer (equil- ibration buffer with 300 m M NaCl) until protein disappeared from the eluate. Finally, the integrin was eluted from the column by adding washing buffer plus 15 m M EDTA. Fractions of 200–300 (cid:3) l were collected in 1.5-ml tubes with 6 (cid:3) l of 1 M MgCl 2 . The fractions were resolved by SDS-polyacrylamide gel electrophoresis, transferred into polyvinylidene difluoride membrane, and incubated with streptavi- din-horseradish peroxidase. Protein bands were detected by chemiluminescence (Pierce). The fractions containing the integrins were also immunoprecipitated with anti- (cid:2) 1 integrin antibody. As a control for this experiment and to assure that the interaction of TSRs with (cid:2)

Thrombospondin-1 (TSP1) 1 is an extracellular matrix glycoprotein that modulates cell adhesion, growth, motility, differentiation, and survival. TSP1 interacts with cells via a number of receptors, including several integrins. Locations of four integrin binding sites have been mapped in TSP1. The RGD sequence in the last type three repeat of TSP1 is a ligand for ␣ v ␤ 3 and ␣ 5 ␤ 1 integrins (1-3), although the exposure of this site in native TSP1 may be limited (4 -7). At least three ␤ 1 integrins recognize distinct sites in the N-terminal pentraxin-like domain of TSP1. ␣ 3 ␤ 1 recognizes a sequence near the carboxyl end of the N-module containing the motif NVR (8), but a truncated recombinant N-module lacking this site retains binding to ␣ 6 ␤ 1 and ␣ 4 ␤ 1 (3,9). Mutation of Glu (90) abolishes ␣ 6 ␤ 1 but not ␣ 4 ␤ 1 binding to this region of TSP1 (9). Binding to the latter integrin is at least partially mediated by an LDVP sequence (3).
Although three ␤ 1 integrin binding sites in TSP1 have now been described, at least two publications have noted ␤ 1 integrindependent activities of TSP1 that cannot be explained by these known sites. DeFreitas et al. (11) report that a TSP1 antibody (A4.1), which was believed to recognize the type 1 repeats (TSR, also known as properdin repeats) (10), inhibited ␣ 3 ␤ 1 -dependent neurite outgrowth on TSP1 (11). They further noted that a 50/70-kDa chymotryptic fragment of TSP1 containing the procollagen domain, the three TSRs, and part of the EGF-like type 2 repeats supported neurite outgrowth, which was completely inhibited by a ␤ 1 antibody. Approximately 80% of this response was inhibited by ␣ 3 ␤ 1 -specific antibodies. Furthermore, adhesion of osteosarcoma cells to a similar 70-kDa core fragment of TSP1 produced by limited proteolysis with chymotrypsin in the absence of calcium was ␣ 4 ␤ 1 -dependent (12). Notably, none of these proteolytic fragments contain the N-module or the RGD sequence of TSP1.
These publications suggested that ␣ 4 ␤ 1 and ␣ 3 ␤ 1 recognize additional sites in the central stalk region of TSP1, possibly in the TSRs. Using recombinant regions of TSP1, we have now verified that activated ␣ 4 ␤ 1 and ␣ 3 ␤ 1 recognize secondary sites in this region of TSP1. We report here evidence for three such sites in the second and third TSRs and in the type 2 repeats of TSP1. However, we find that the specificity of these sites differs from those of most previously described integrin ligands in that they are pan-␤ 1 -specific. We further show that a widely used TSP1 function blocking antibody, A4.1, inhibits ␤ 1 integrin recognition of both repeats in TSP1.

EXPERIMENTAL PROCEDURES
Proteins and Peptides-TSP1 was purified from human platelets (13). Recombinant proteins containing various domains of TSP1 or TSP2 (summarized in Fig. 1A) were prepared as described (14 -16). Vitrogen type I collagen was from Cohesion Technologies (Palo Alto, CA).
Antibodies and Antagonists-A ␤ 1 integrin function-blocking antibody (mAb13) and a fibronectin antibody (13G12) were provided by Dr. Ken Yamada (NIDCR, National Institutes of Health, Bethesda, MD) (17,18). The ␤ 1 integrin-activating antibody TS2/16 (19) and the anti-TSP1 antibody HB8432 were produced from hybridoma cell lines obtained from the American Type Culture Collection (Manassas, VA). The anti-␣ 2 antibody 6D7 was provided by Dr. Harvey Gralnick (20). TSP1 * 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.
Integrin Purification by TSR Affinity-A TSR123 affinity column was prepared using a CNBr-activated agarose bead matrix (following the manufacturer's protocol). Surface-biotinylated Jurkat cells (2-5 ϫ 10 8 ) were lysed in the presence of 5-10 ml of buffer containing 100 mM n-octylglucoside, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 , and 1 mM phenylmethylsulfonyl fluoride. After removal of insoluble material by centrifugation at 3000 ϫ g for 20 min, the extract was concentrated to 1 ml using Centripep 10 (Millipore). The column was previously washed with equilibration buffer containing 25 mM n-octylglucoside, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2 mM MgCl 2 , 0.1 mM CaCl 2 , and 0.5 mM MnCl 2 . The cell extract was applied to the column and incubated for 1 h at room temperature or overnight at 4°C. The column was washed with equilibration buffer until protein disappeared from the eluate, then washed again with washing buffer (equilibration buffer with 300 mM NaCl) until protein disappeared from the eluate. Finally, the integrin was eluted from the column by adding washing buffer plus 15 mM EDTA. Fractions of 200 -300 l were collected in 1.5-ml tubes with 6 l of 1 M MgCl 2 . The fractions were resolved by SDS-polyacrylamide gel electrophoresis, transferred into polyvinylidene difluoride membrane, and incubated with streptavidin-horseradish peroxidase. Protein bands were detected by chemiluminescence (Pierce). The fractions containing the integrins were also immunoprecipitated with anti-␤ 1 integrin antibody. As a control for this experiment and to assure that the interaction of TSRs with ␤ 1 integrins was not mediated by fibronectin, surface-biotinylated Jurkat cells lysates were also immunoprecipitated with a fibronectin antibody (13G12) to determine the presence of this protein in the cell lysates.
Competition Enzyme-linked Immunosorbent Assay -TSP1 was diluted to 0.005 M on a subunit basis in Tris-buffered saline (TBS) plus 2 mM calcium, and Probind 96-wellmicrotiter plates were coated for 16 h at 4°C with 50 l/well of the diluted solution. The plates were washed once with TBS plus 0.05% Tween 20 (TBST) and blocked with 5% nonfat dried milk and 1% BSA for 1 h. The recombinant TSP fragments (competitors) and A4.1 were diluted so that when combined the final concentrations for each competitor was 3.0, 1.0, 0.3, 0.1, 0.03, 0.01, or 0 M, and the final concentration of the A4.1 was 3.0 g/ml (ϳ0.003 M IgM pentamer or 0.03 M IgM Fab segment). The competitors were added to A4.1 in microcentrifuge tubes, and the mixtures were incubated for 1 h before being added to the blocked and washed TSP1-coated plates. After 2 h in the wells, the mixtures were removed, and the wells were washed 4 times with TBST. Alkaline phosphatase-conjugated goat anti-mouse IgM ( chain-specific) from Sigma was diluted 1:3000 in TBST plus 0.1% nonfat dried milk and 0.02% BSA and incubated for 1 h with the plate. After washing 4 times with TBST, Sigma 104 alkaline phosphatase substrate at 1 mg/ml in TBS, pH 9.0, was added to each well. Color development was monitored at 405 nm.

RESULTS
Three Domains of TSP1 Are Recognized by ␤ 1 Integrins-In addition to the three known ␤ 1 integrin binding sites in the N-module of TSP1 (3,8,9), a survey of cell adhesion to other recombinant regions of TSP1 in the absence or presence of a ␤ 1 integrin-activating antibody revealed that the type 1 and type 2 repeats contain ␤ 1 integrin-dependent adhesion sites (Fig. 1). Of several cell types examined, mesangial cells attached and spread most avidly to both the type1 and type 2 repeats, which was enhanced after activation of ␤ 1 integrins using the antibody TS2/16 ( Fig. 1B). At equimolar concentrations, NoC1 supported slightly less adhesion than intact TSP1, and three constructs containing all or only the third TSR supported adhesion at ϳ50% of this level. The type 2 repeats alone were weakly active at this concentration, whereas the C-terminal regions of TSP1 were inactive. Results were qualitatively similar using TS2/16-activated Jurkat T cells (Fig. 1C), although these cells do not spread due to their limited cytoplasmic volume. Doseresponse curves showed that NoC1 is ϳ3-fold less active than intact TSP1 for mediating cell attachment, and CTSR123 is 13-fold less active. Endothelial cells also recognized the TSRs but did not adhere at significant levels on the type 2 repeats (Fig. 1, D and E, and results not shown). As noted for the other cell types, adhesion was stimulated by the ␤ 1 antibody TS2/16 (results not shown). Adhesion of endothelial cells on the third TSR (TSR3) was completely inhibited by a ␤ 1 -blocking antibody (mAb13), but 30 -50% of adhesion on the second TSR was resistant to this antibody.
To verify that ␤ 1 integrins are required for cell adhesion to the TSRs, a somatic mutant of Jurkat cells deficient in all ␤ 1 integrins was examined (Fig. 2). Adhesion of wild type Jurkat cells to TSR123 containing all three TSRs was stimulated by TS2/16 and to a lesser extent by phorbol ester. The latter response was completely blocked by the ␤ 1 -specific blocking antibody mAb13 ( Fig. 2A), indicating that the type 1 repeats are recognized by ␤ 1 but not by ␤2 or ␤7 integrins, which are also expressed on Jurkat cells (26). Activation-dependent adhesion was absent in the A1 somatic mutant lacking ␤ 1 (26), further indicating that ␤ 2 and ␤ 7 integrins do not recognize the TSRs. Therefore, ␤ 1 integrins are necessary for T cell adhesion on the TSRs.
Because the TSP1 TSRs are known to interact with CD36, which associates with at least two ␤ 1 integrins (27), and with the ␤ 1 integrin ligand fibronectin (28), ␤ 1 integrin-dependent adhesion to the TSRs could be indirect. However, detergentsolubilized ␤ 1 integrins bound to an affinity column of TSR123 and were specifically eluted by chelating divalent cations (Fig.  2B). Based on immunoprecipitation with the antibody TS2/16, the major proteins eluted under these conditions corresponded to ␤ 1 integrin with a molecular mass around 130 kDa, which in Jurkat cells associates primarily with ␣ 4 and ␣ 5 subunits (140 and 135 kDa, respectively, under reducing conditions). Indirect interaction of integrins with the TSRs mediated by CD36 can be excluded because Jurkat cells do not express detectable levels of this receptor. Neither was fibronectin involved, be-cause immunoprecipitation of biotinylated Jurkat cells using a specific fibronectin antibody did not detect this protein in the cell lysate (data not shown).
Phorbol ester treatment and TS2/16 similarly induced Jurkat T cell adhesion to E123, containing the three type 2 repeats of TSP1 (Fig. 2C). PMA-stimulated adhesion to the type 2 repeats was completely inhibited by the ␤ 1 antibody mAb13, indicating specificity for ␤ 1 integrins. The Jurkat A1 mutant was inactive, demonstrating that ␤ 1 integrins are also necessary for adhesion of T cells on the EGF-like repeats of TSP1.
TSP1 Type 1 Repeats Are pan-␤ 1 Integrin Ligands-Several ␣␤ 1 dimer-specific blocking antibodies and antagonists were tested to determine which ␤ 1 integrins recognize the TSRs (Fig.  3). Remarkably, the ␤ 1 integrin specificity for adhesion on TSP1 type 1 repeats was cell type-specific. Jurkat cell adhesion on the type 1 repeats was partially inhibited by an ␣ 4 ␤ 1 integrin antagonist, but ␣ 3 , ␣ 2 , and ␣ v inhibitors were inactive (Fig.  3A). The potent ␣ 4 antagonist phLDVP inhibited adhesion on the type 1 repeats by ϳ40%, whereas it inhibited adhesion on the N-terminal region of TSP1 by more than 80%, suggesting that additional ␤ 1 integrins may contribute to T cell adhesion on the TSRs, although ␣ 4 ␤ 1 is the most abundant ␤ 1 integrin expressed by Jurkat T cells (26). Notably, none of the specific ␤ 1 integrin blocking antibodies or antagonists except mAb13 significantly inhibited activation-dependent Jurkat cell adhesion on the type 2 repeats (Fig. 3A).
In contrast to T cells, MDA-MB-231 breast carcinoma cell attachment on the TSRs was inhibited by ␣ 2 ␤ 1 and ␣ 3 ␤ 1 function-blocking antibodies but not by the ␣ 4 ␤ 1 antagonist (Fig.  3B). Neither ␣ 3 ␤ 1 nor ␣ 2 ␤ 1 antibodies completely inhibited MDA-MB-231 cell attachment. In the case of the ␣ 2 ␤ 1 antibody, this was not due to a limiting antibody concentration because cell attachment on type I collagen was inhibited to a much greater extent. Combining the ␣ 2 and ␣ 3 antibodies, however, produced a specific additive effect for adhesion on TSR123 but not on NoC1 or type I collagen, suggesting that both integrins recognize the TSRs. However, the ␣ 4 antagonist, which inhibited T cell adhesion to the same protein (Fig. 3A), was inactive and showed no additivity with the ␣ 2 ␤ 1 antibody. Lack of activity of the ␣ 4 antagonist presumably is due to absence of ␣ 4 expression, because the ␣ 4 antagonist also had no effect on the known ␣ 4 ␤ 1 ligand NoC1, which was recognized exclusively by ␣ 3 ␤ 1 in the breast carcinoma cells (Fig. 3B). Based on these results, ␣ 2 ␤ 1 and ␣ 3 ␤ 1 , but not ␣ 4 ␤ 1 , are adhesion receptors in MDA-MB-231 cells for the TSP1 TSRs. Notably, ␣ 2 ␤ 1 and ␣ 3 ␤ 1 are the most abundant integrins expressed by MDA-MB-231 cells (29). Mesangial cells exhibited a third phenotype (Fig. 3C). Antibodies to several ␤ 1 integrins, including ␣ 1 ␤ 1 , ␣ 2 ␤ 1 , ␣ 3 ␤ 1 , and ␣ 6 ␤ 1 , partially inhibited mesangial cell spreading on the type 1 repeats of TSP1. Small molecule antagonists of ␣ 4 ␤ 1 and ␣ v integrins were also partially inhibitory. However, the ␤ 1 antibody completely inhibited spreading (Fig. 3C) and inhibited overall cell attachment by ϳ50% (results not shown). Consistent with this result and the partial inhibition by antagonizing individual integrins, additive effects were observed by combining the ␣ 2 ␤ 1 antibody and the ␣ v antagonist. Therefore, grins all contribute to mesangial cell adhesion on the TSP1 TSRs.
Although apparently contradictory, the disparate data for these three cell lines can be rationalized by considering the relative expression levels of the various ␤ 1 integrin in each cell line. Adhesion to the TSRs was inhibited by blocking the most FIG. 2. ␤ 1 integrins are necessary for activation-dependent T cell adhesion to the type 1 and type 2 repeats and bind to type 1 repeats. A, somatic T cell mutants verify ␤ 1 is necessary for adhesion on the type 1 repeats. Jurkat and Jurkat A1 ␤ 1 -deficient cells were plated on substrates coated with 8 g/ml NoC1 (solid bars) or 20 g/ml TSR123 (striped bars) and incubated for 20 min at 37°C. Cells were used without stimulation (control), activated with TS2/16 antibody, or with 20 ng/ml PMA alone or in the presence of 5 g/ml ␤ 1 blocking antibody mAb13. B, isolation of ␤ 1 integrins on a TSR123 affinity column. Surface-biotinylated Jurkat cells were extracted with 100 mM octyl glucoside buffer containing 1 mM Mg 2ϩ and 1 mM Ca 2ϩ and applied to a TSR123 affinity column. After washing to remove unbound proteins, bound integrins were eluted using buffer containing 15 mM EDTA, resolved on a SDS gel, and detected using horseradish peroxidase-streptavidin and ECL. Surface-biotinylated Jurkat cells lysate (lane 1) and a pool of EDTA-eluted fractions (lane 2) were immunoprecipitated (IP) with TS2/16 antibody, resolved on a SDS gel, and detected using horseradish peroxidase-streptavidin and ECL C, Jurkat mutants verify ␤ 1 is necessary for adhesion on the type 2 repeats. Adhesion of wild type and A1 ␤ 1 -deficient Jurkat T cells was assessed on surfaces coated with 20 g/ml E123.
Both TSR2 and TSR3 Are Recognized by ␤ 1 Integrins-The above results could be explained either by a single promiscuous ␤ 1 integrin binding site in the TSRs or by the presence of distinct binding sites for each integrin. To further localize the integrin binding site(s) in the TSRs of TSP1, recombinant forms of the second and third TSRs were tested for adhesive activity (Fig. 4). TS2/16 stimulated mesangial cell adhesion on both the second and third TSRs, although the third TSR was ϳ2-fold more active (Fig. 4A). However, these two TSRs alone were less active than a construct containing all three TSRs (TSR123) or intact TSP1. In the absence of activation, mesangial cells also showed significant dose-dependent adhesion to TSR2, TSR123, and intact TSP1. The absence of adhesion of unstimulated cells to TSR3 may be an artifact of the limited concentration of this fragment that could be tested or its relatively inefficient adsorption on polystyrene. Adhesion to the two individual TSRs exhibited the divalent cation dependence typical of integrin mediated adhesion (Fig.  4B). The addition of 0.4 mM Mn 2ϩ -stimulated adhesion and the addition of 10 mM EDTA abolished mesangial cell spreading and cell attachment on both TSR2 and TSR3. Spreading on TSR123 was also abolished by EDTA, but attachment was only partially inhibited by EDTA. Attachment on type I collagen that was even stronger than to TSR123 was completely inhibited by EDTA, suggesting that some mesangial cell attachment on the full-length type 1 repeats is divalent cation-independent.
To determine whether the apparent pan ␤ 1 integrin specificity is intrinsic to each repeat or results from each TSR recognizing different ␤ 1 integrins, we compared their activities to promote adhesion of T cells and breast carcinoma cells (Fig. 4, C-D). Adhesion of Jurkat T cells to TSR2 or TSR3 was inhibited strongly by blocking ␣ 4 ␤ 1 and weakly by blocking ␣ 5 ␤ 1 (Fig. 4C). Adhesion of MDA-MB-231 cells to TSR2 or TSR3 was inhibited partially by blocking ␣ 2 ␤ 1 or ␣ 3 ␤ 1 and completely by blocking both integrins (Fig. 4C). This additivity was not observed for adhesion of the same cells on NoC1, which MDA-MB-231 cells recognize exclusively via ␣ 3 ␤ 1 (33). Based on the ability of both TSR2 and TSR3 to be recognized by at least three different ␤ 1 integrins in a cell-specific manner, we conclude that at least the second and third TSRs of TSP1 have pan-specific ␤ 1 integrin binding sites.
Several functional peptide sequences have been identified in the TSRs that interact with other known TSP1 ligands or receptors. To determine whether any of these known functional sequences contribute to integrin-mediated adhesion on the TSRs, we examined the effects of synthetic peptides on the adhesion of mesangial cells (Fig. 5). VTCGGGVQKRSRL (peptide 245) is a CD36-binding peptide from the third TSP1 TSR that inhibits angiogenesis (34). VTCGDGVITR (peptide 205) and SPWSSCSVTCGDGVITR (peptide 616) are CD36-binding peptides from the second TSR (34). KRFKQDGGWSHWSP-WSS (peptide 246) is a heparin-, transforming growth factor-␤-, and fibronectin-binding peptide from the border of the first and second TSRs (28,35,36). GPWSPWDICSVT (pep- tide 186) is a heparin-binding peptide from the third TSR (35). A related WSXW peptide from SCO-spondin was recently implicated in mediating ␣ 1 ␤ 1 integrin-dependent neurite outgrowth (37). However, none of the TSP1 peptides containing this sequence or the CD36 binding sequences inhibited adhesion of TS2/16-activated cell adhesion on TSR123 or type I collagen (Fig. 5).
The Epitope for TSP1 Antibody A4.1 Is in the Third Type 2 Repeat-The TSP1 antibody A4.1 blocks several cellular responses to TSP1, and its epitope was initially mapped to the TSRs (10). However, A4.1 did not recognize CTSR123-1, which contains all three TSRs, on a Western blot (Fig. 6A, lane b). Rather, A4.1 recognized the EGF-like repeats and bound to all such constructs that contain the E3 module (lanes c-f). A4.1 did not bind to full-length recombinant TSP2 but bound to fulllength murine TSP1 (lanes h and i).
Specificity of A4.1 for E3 of TSP1 was confirmed using a competitive enzyme-linked immunosorbent assay (Fig. 6, B and  C). All constructs containing the E3 module equally inhibited A4.1 binding to human platelet TSP1, whereas the N-terminal region of TSP1, NoC1, and a region of TSP2 containing its EGF-like repeats (E123CaG2) were inactive. Therefore, A4.1 specifically recognizes an epitope in E3 of TSP1.
Inhibition of Integrin Recognition by TSP1 Antibody A4.1-Antibody A4.1 inhibits CD36-dependent motility responses of endothelial cells to TSP1, and this activity has been attributed to inhibiting recognition of the TSRs by CD36 (38 -41). To determine whether A4.1 also inhibits recognition of the TSRs by ␤ 1 integrins, we examined its effects on mesangial cell adhesion (Fig. 7). A4.1 inhibited mesangial cell adhesion only on constructs containing the type 2 repeats (Fig. 7A). A4.1 inhibited cell attachment on delNo and E123 and completely inhibited spreading on delNo and on E123 but not on TSR123 or NoC1. HB8432, a second TSP1 antibody that binds to the two most N-terminal type 2 repeats, 2 had a similar profile but was less inhibitory than A4.1 (Fig. 7A). The ability of A4.1 to completely block spreading on delNo implies that the antibody can sterically block integrin binding to the TSRs as well as proximally inhibit recognition of the type 2 repeats. To test this hypothesis we compared adhesion on the third TSR expressed alone or fused to the type 2 repeats (Fig. 7B). A4.1 dose dependently inhibited mesangial cell adhesion on TSR3E123 but not on TSR3 or TSR123. The TSP1 antibody 5G11, which recognizes TSR3, 2 blocked adhesion on TSR3 and TSR3E123 but not E123, indicating that TSR3 contributes substantially to the adhesive activity of TSR3E123 (results not shown). The same conclusion is supported by activity of the ␣ 4 antagonist to inhibit Jurkat cell adhesion on TSR3E123 but not E123 (results not shown). Considering these results, the data presented in Fig. 7B indicate that A4.1 proximally blocks ␤ 1 -dependent adhesion to the type 2 repeats and also indirectly inhibits ␤ 1 -dependent adhesion to the TSRs.
Disintegrins Inhibit Adhesion on TSP1 Type 2 Repeats-Although adhesion on the type 2 repeats is resistant to most specific integrin antibodies and antagonists, it is sensitive to some disintegrins (Fig. 8). Mesangial cell adhesion to type 2 repeats was more resistant to individual ␤ 1 integrin antagonists than was adhesion of the same cells on the type 1 repeats, but the disintegrins VLO-5 and obtustatin significantly inhibited spreading on E123 (Fig. 8A). At lower concentrations of obtustatin, additive effects were observed with an ␣ 2 ␤ 1 antibody (Fig. 8B). The activity of VLO-5 may not be due to its specific antagonism of ␣ 4 ␤ 1 in mesangial cells because the dose response was much broader than for inhibiting T cell adhesion to the known ␣ 4 ␤ 1 ligand NoC1 (Fig. 8C). Furthermore, Jurkat T cell attachment on E123 was completely resistant to VLO-5 inhibition, although attachment of the same cells on NoC1 was inhibited as expected. This suggests that ␣ 4 ␤ 1 does not mediate adhesion on E123 and that activity of the disintegrin VLO-5 for mesangial cells may be due to antagonism of other integrins, possibly including ␣ 9 ␤ 1 (24).

DISCUSSION
Our data confirms prior indirect evidence for ␤ 1 integrinmediated adhesion to the type 1 repeats of TSP1 (11,12). However, we could not reproduce the reported specificity of these sites for individual ␤ 1 integrins (11,12). Rather, we find that pan-␤ 1 integrin recognition sites are present in the second and third TSRs and in the type 2 repeats of TSP1. Adhesion mediated by the TSRs is inhibitable by function-blocking antibodies specific for the major ␣␤ 1 dimers present in any given cell type as well as by small molecule antagonists of the respective integrins. Although we could not test all of the known ␤ 1 integrins, at least seven of these recognize the TSRs. This suggests that the TSP1 TSRs are pan-specific ligands for ␤ 1 integrins. In contrast, no ␣-specific antagonists have been identified that strongly block adhesion to the type 2 repeats, although ␤ 1 is necessary, and the latter activity is sensitive to a ␤ 1 blocking antibody and some disintegrins. Thus, the type 1 and type 2 repeats of TSP1 contains additional cell adhesion sites, and their distinct interactions with ␤ 1 integrins may play significant but previously unrecognized roles in the many biological activities that have been mapped to these domains of TSP1.
Although we cannot strictly exclude binding to ␤ 1 integrinassociated molecules, our data strongly support a direct inter- action between ␤ 1 integrins and the TSRs. ␤ 1 integrins specifically bound to a TSR affinity column and were eluted under nondenaturing conditions by EDTA, consistent with the known cation dependence for most integrin ligand binding. Second, some of the integrin ␣-subunit-specific antagonists we used are small molecules that are highly unlikely to sterically prevent binding of the TSRs to an integrin-associated protein.
Although ␤ 1 integrins are necessary for adhesion of T cells on the TSRs, we do not know whether recognition of the TSP1 TSRs or type 2 repeats is limited to ␤ 1 integrins. ␤ 1 -deficient Jurkat cells express ␤ 2 and ␤ 7 integrins, and these are activated by phorbol esters, but we found no PMA-stimulated adhesion of these cells to TSRs. Therefore, ␤ 2 and ␤ 7 integrins do not recognize the TSRs. Recognition of the TSRs by integrins containing other ␤-subunits has not been examined.
Although our data indicate that ␤ 1 integrins directly recognize the TSP1 TSRs, several indirect mechanisms were also considered. CD36 also binds to the type 1 repeats (42,43). Furthermore, CD36 associates with two of the integrins that mediate adhesion to the type 1 repeats, ␣ 3 ␤ 1 and ␣ 6 ␤ 1 (27). However, binding of the type 1 repeats to CD36 seems unlikely to explain our observed pan-␤ 1 specificity because CD36 does not associate with several of the integrins implicated by our data, including ␣ 2 ␤ 1 , ␣ 4 ␤ 1 , and ␣ 5 ␤ 1 (27). Some of the cells that adhere on the type 1 repeats also lack CD36 expression. In addition, several CD36-binding peptides from TSP1 did not inhibit adhesion of cells on immobilized TSRs. In some cases, biological activities of A4.1 were inferred to result from its blocking CD36 binding to the TSRs (38 -41). However, the relevance of A4.1 blocking to CD36 function may need to be reevaluated based on our evidence that A4.1 also blocks recognition of the TSRs by ␤ 1 integrins and our new data mapping the A4.1 epitope to the third type 2 repeat rather than the TSRs.
We also considered heparan sulfate proteoglycans and fibronectin as potential indirect ligands. However, fibronectin was not present at detectable concentrations in our adhesion assays or the affinity purification experiment. Furthermore, fibronectin should preferentially be recognized by ␣ 5 ␤ 1 , antagonists of which were only minimally effective to block adhesion to the TSRs. Although a heparin binding sequence from the TSRs stimulates ␣ v ␤ 3 -dependent melanoma cell adhesion (2), this peptide had no effect on ␤ 1 -dependent adhesion to immobilized TSRs.
Most integrin ligands show relatively high specificity for one or a few integrins, but at least three ligands were previously known to mediate general ␤ 1 -dependent adhesion. Invasin mediates internalization of Yersinia pseudotuberculosis and is a well characterized promiscuous ␤ 1 integrin ligand (54). Specific residues in the ␤ 1 subunit are required for invasin binding (55). The disintegrin domains of ADAMs-2 and -3 were also shown to have broad binding specificity for ␤ 1 integrins (56). Finally, cell adhesion to several fibronectin type 3 repeats that lack the RGD sequence could be stimulated by TS2/16 or by PMA and was sensitive to ␤ 1 blocking antibodies but not to any ␣-specific blocking antibodies tested (57). These results resemble adhesion on the type 2 repeats of TSP1, which can be blocked only by disintegrins or ␤ 1 blocking antibody but contrasts with adhesion to the TSP1 TSRs, which is sensitive to inhibition by FIG. 8. Adhesion on the type 2 repeats is resistant to most specific integrin antagonists but sensitive to some disintegrins. A, mesangial cell spreading on TSP1 type 2 repeats is more resistant to individual ␣␤ 1 integrin antagonists (antag.) that spreading on the type 1 repeats but is inhibited by two disintegrins. The indicated integrin function-blocking antibodies were used at 5 g/ml. ␣ v and ␣ 4 antagonists, VLO-5, and obtustatin were used at 1 M. Results are expressed as percentage of TS2/16 control (39 Ϯ 5 and 121 Ϯ 7 cells on TSR123 and E123, respectively). B, additive effects of combined obtustatin (330 nM) and anti-␣ 2 integrin antibody 6D7 (5 g/ml). Results are expressed as percentage of TS2/16 control (33 Ϯ 1 and 58 Ϯ 2 cells on TSR123 and E123, respectively). C, dose-dependent activity of the ␣ 4 /␣ 9 disintegrin VLO-5 for inhibiting attachment of T cells on NoC1 (q, mediated by ␣ 4 ␤ 1 ) or E123 (E) or mesangial cells on E123 (OE).
␣-subunit-specific antibodies and small molecule antagonists. These results suggest that the fibronectin type 3 repeats and the TSP1 type 2 repeats may be recognized autonomously by the activated ␤ 1 subunit, whereas the TSRs must interact with sites that are close to the known ligand binding sites at the ␣␤ interface that interact with small molecule integrin antagonists (58). The mechanisms by which ␤ 1 integrins recognize the TSRs and type 2 repeats of TSP1 remain to be determined.
We have shown that both the second and third TSP1 TSRs are ␤ 1 integrin ligands. We have not determined whether the first repeat is also active. Preliminary evidence indicates that TSRs from TSP2 have similar integrin recognition sites. 3 Some evidence suggests that integrin recognition is a more general property of TSRs. SCO-spondin contains TSRs, which mediate neurite outgrowth in a ␤ 1 integrin-dependent manner (37). Notably, these authors reported specificity of a peptide containing their SCO-spondin motif for ␣ 1 ␤ 1 and found no significant inhibition using ␣ 2 ␤ 1 , ␣ 3 ␤ 1 , ␣ 4 ␤ 1 , or ␣ 6 ␤ 1 antibodies. However, the observed integrin specificity should be verified using an intact recombinant TSR rather than the synthetic peptide. Notably, we could not inhibit cell adhesion to TSP1 TSRs using a similar peptide sequence from TSP1. Because TSRs are found in many important proteins that regulate cell behavior and embryonic development (reviewed in Ref. 59), integrin recognition could have broad significance to their biology.
EGF-like repeats are also found in many proteins, but their role as integrin ligands is poorly understood. A 39-amino acid ␤ 1 integrin binding sequence was identified in an EGF-like repeat of entactin/nidogen (60). Milk fat globule-EGF-factor 8 contains an RGD motif in its second EGF repeat that is recognized by ␣ v ␤ 3 (61) as does developmental endothelial locus-1 (62). The EGF-like type 2 repeats of TSP1 do not contain any known integrin binding motifs.
In summary, we have identified three additional ␤ 1 integrin recognition sites on TSP1. These are less active than the three specific sites previously defined in the N-module of TSP1 but recognize a broader range of ␤ 1 integrins. Therefore, cells lacking the three integrins that bind to the N-module may still interact with TSP1 via ␤ 1 integrins. The mechanisms for integrin recognition of the TSRs and EGF-like repeats and the potential for the unique glycosylation of Trp and conserved Ser/Thr residues in TSRs (63) to modulate this integrin binding remain to be explored.