Recognition of the N-terminal Modules of Thrombospondin-1 and Thrombospondin-2 by α6β1 Integrin*

In addition to its recognition by α3β1 and α4β1 integrins, the N-terminal pentraxin module of thrombospondin-1 is a ligand for α6β1 integrin. α6β1 integrin mediates adhesion of human microvascular endothelial and HT-1080 fibrosarcoma cells to immobilized thrombospondin-1 and recombinant N-terminal regions of thrombospondin-1 and thrombospondin-2. α6β1 also mediates chemotaxis of microvascular cells to thrombospondin-1 and thrombospondin-2. Using synthetic peptides, LALERKDHSG was identified as an α6β1-binding sequence in thrombospondin-1. This peptide inhibited α6β1-dependent cell adhesion to thrombospondin-1, thrombospondin-2, and the E8 fragment of murine laminin-1. The Glu residue in this peptide was required for activity, and the corresponding residue (Glu90) in the N-terminal module of thrombospondin-1 was required for its recognition by α6β1, but not by α4β1. α6β1 was also expressed in human umbilical vein endothelial cells; but in these cells, only certain agonists could activate the integrin to recognize thrombospondins. Selective activation of α6β1 integrin in microvascular endothelial cells by the anti-β1 antibody TS2/16 therefore accounts for their adhesion responses to thrombospondins and explains the distinct functions of α4β1 and α6β1 integrins as thrombospondin receptors in microvascular and large vessel endothelial cells.

Integrins play a major role in mediating interactions between cells and their extracellular matrix environment. In addition to providing physical anchoring of cells, engagement of integrins transmits signals into cells that regulate their survival and behavior (reviewed in Ref. 1). Conversely, cells can modulate their interactions with the extracellular matrix through regulating the expression or ligand-binding activities of specific integrins (1). Although some extracellular matrix ligands have been identified for most known integrins (reviewed in Ref. 2), our understanding of integrin ligand specificities remains incomplete.
One of the best characterized biological activities of TSP1 and TSP2 is to modulate angiogenesis. Inhibition of endothelial cell chemotaxis is mediated by the TSP1 receptor CD36 (14). Heparan sulfate proteoglycans and CD47 may also contribute to the anti-angiogenic activities of TSP1 (15)(16)(17). In certain contexts, however, TSP1 and recombinant or proteolytic fragments of TSP1 exhibit pro-angiogenic activities (18,19). We previous reported that ␣ 3 ␤ 1 mediates a pro-angiogenic activity of the N-terminal pentraxin module of TSP1 (18). This integrin is constitutively expressed on venous and microvascular endothelial cells, but its ability to bind TSP1 is regulated by cell-cell signaling involving VE-cadherin (18). Because ␤ 1 -dependent interactions of TSP1 or TSP2 with endothelial cells could not be completely inhibited by ␣ 3 ␤ 1 antagonists, we examined the role of additional ␤ 1 integrins expressed on endothelial cells as TSP receptors. 2 This effort revealed that a second ␤ 1 integrin, ␣ 4 ␤ 1 , selectively functions as a TSP1 and TSP2 receptor in large vessel endothelial cells. However, microvascular cells are refractory to ␣ 4 ␤ 1 antagonists. We have therefore examined the role of additional ␤ 1 integrins as TSP receptors in microvascular cells and report here that ␣ 6 ␤ 1 is a major integrin receptor for TSP1 and TSP2 that mediates adhesion and chemotaxis of microvascular endothelial cells. Furthermore, we identify a specific sequence in the N-terminal module of TSP1 that mediates this interaction and demonstrate that this sequence antagonizes interactions of ␣ 6 ␤ 1 with its well characterized ligand laminin-1.

EXPERIMENTAL PROCEDURES
Cell Culture-Human dermal microvascular endothelial (HDMVE) cells and human lung microvascular endothelial (HMVE-L) cells (Cambrex Bio Science Inc., Walkersville, MD) were grown under the conditions specified by the manufacturer. Adult iliac vein endothelial cells (AG10773A; NIA Repository, Coriell Institute for Medical Research, Camden, NJ) were grown on flasks coated with 0.1% gelatin in medium 199 containing 10% fetal bovine serum (FBS), 2 mM glutamine, 30 g/ml endothelial cell mitogen (Biomedical Technologies, Inc., Stoughton, MA), and 30 g/ml heparin. Human umbilical vein endothelial (HUVE) cells (Clonetics BioWhittaker Inc.) were maintained in medium 199 containing * 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.
Proteins and Peptides-TSP1 was purified from human platelets obtained from the National Institutes of Health Blood Bank (20). The protein was Ͼ95% intact based on analysis by SDS gel electrophoresis. The recombinant trimeric proteins consisting of the N-terminal module (N), oligomerization sequence (o), and procollagen module (C) of TSP1 residues 1-356 (NoC1) and TSP2 residues 1-359 (NoC2) were prepared as described (21). A recombinant portion of TSP1 (residues 1-175 of the mature protein) was prepared as described previously (15). Site-directed mutagenesis of Glu 90 in TSP1-  to Ala was performed as described previously (12). Residue numbers refer to the mature sequence of secreted TSP1. The forward and reverse mutation-inducing primer sequences were 5Ј-GAGTGGTCTTTCCGCGCCAGGGCCAGCA-GCGTG-3Ј and 5Ј-GAGTGGTCTTTCCGCGCCAGGGCCAGCAGCGT-G-3Ј, respectively. After growth on LB broth/ampicillin plates at 30°C, the mutated plasmid was transformed into Escherichia coli XL-1 Blue cells for isolation, transformed into Rosetta cells (Novagen) for protein expression, and grown to log phase at 37°C on LB broth plus carbenicillin (50 g/ml) and chloramphenicol (34 g/ml). Inclusion bodies were isolated, and the mutant recombinant protein was purified as described previously for the wild-type recombinant protein (15). Synthetic peptides derived from TSP1 were prepared as described previously (13). Recombinant S7D-VCAM-1 (soluble 7-domain vascular cell adhesion molecule-1, residues 1-674) was prepared as described previously (12). Murine laminin-1 was provided by Dr. Lance Liotta (Laboratory of Pathology, NCI). The E8 fragment of murine laminin-1 was prepared as described (22).
␣ 6 Integrin Expression-Semiquantitative reverse transcription-PCR was used to determine the expression levels of ␣ 6 ␤ 1 integrin on endothelial cells. Total RNA was isolated from HUVE and HDMVE cells using TRIzol reagent (Invitrogen). First strand cDNA synthesis was performed with Superscript II reverse transcriptase (Invitrogen) using 2 g of total RNA. The enzyme was inactivated at 70°C for 15 min. The cDNA was amplified using platinum Taq DNA polymerase (Invitrogen) and specific primer pairs for ␣ 6 integrin (sense, AAGTCTCAGTTTCT-TGCTTGGG; and antisense, TTCTTTGTTGACCACCCTCC). Amplification was carried out for two cycles of 1 min at 95°C and 4 min at 55°C, followed by 30 cycles of 1 min at 95°C, 2.5 min at 55°C, and 10 min at 70°C.
Flow Cytometry Analysis-HUVE and HDMVE cells were washed with DPBS containing 0.2% BSA and incubated with Puck's saline containing 0.2% EDTA and 10% FBS at 37°C for 6 min. Cells were dislodged and resuspended in a large volume of Puck's saline/EDTA solution, centrifuged, resuspended in DPBS containing 0.2% BSA at a density of 6 ϫ 10 6 cells/ml, and stored on ice. Cells (1 ϫ 10 6 /labeling reaction) were incubated with 2 g of antibody TS2/16 as a control or with 2 g of antibody GoH3 for 1 h. Cells were washed twice with DPBS and 0.2% EDTA and incubated with fluorescein isothiocyanate-conjugated anti-mouse antibody for 1 h. Labeled cells were washed again and fixed with 300 l of 1% formaldehyde in DPBS. Flow cytometry acquisition was performed using a BD Biosciences flow cytometer. Immunoprecipitation Analysis-HUVE or HDMVE cells grown in 10-cm dishes were dislodged with 2 mM EDTA, centrifuged, and resuspended (1 ϫ 10 6 cells/ml) in a 1 mg/ml solution of EZ-Link Sulfo-NHS-LC-Biotin (Pierce) at room temperature for 1 h. After washing with DPBS, the cells were lysed with radioimmune precipitation assay buffer (50 mM Tris (pH 7.5), 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 1 mM EGTA, and 1 mM NaF supplemented with 10 g/ml each antipain, pepstatin A, chymostatin, leupeptin, aprotinin, and soybean trypsin inhibitor and 1 mM phenylmethylsulfonyl fluoride), and the lysate was precleared by centrifugation. Equal volumes with equal protein concentrations were immunoprecipitated using the anti-␣ 6 antibody GoH3 prebound to anti-mouse IgG agarose (Sigma). The immune complexes were washed six times with Tris-buffered saline, diluted with sample buffer containing 10% ␤-mercaptoethanol, heated, and fractionated on precast SDS gels (Bio-Rad). After transfer to a polyvinylidene fluoride membrane, the proteins were detected using horseradish peroxidase-conjugated streptavidin and visualized using chemiluminescent substrate (Pierce).
Chemotaxis-HDMVE cells at passages 5-8 were used 2-3 days after passage. Modified Boyden chambers and 8-m pore polycarbonate membranes (NeuroProbe, Gaithersburg, MD) coated with 100 g/ml gelatin in 0.1% aqueous acetic acid were used. Cells were dislodged with 2 mM EDTA and allowed to recover for 30 min suspended in complete medium. After centrifugation, the cells were resuspended in medium 199 containing 0.1% BSA and added (56 l) to the upper chamber in the absence or presence of TSP1 (30 g/ml), NoC1 or NoC2 (20 g/ml), or blocking antibodies. Wells in the lower chamber contained 28 l of assay medium alone or with TSP1 (30 g/ml) or NoC1 or NoC2 (20 g/ml). Cells were allowed to migrate for 6 h. The membranes were fixed and stained. Migrated cells were counted microscopically.
Inhibition by the anti-␣ 6 ␤ 1 antibody was reproducibly observed using several independent isolates of HDMVE cells, but not using HUVE cells ( Fig. 2A). An ␣ 4 ␤ 1 antagonist reproducibly inhibited adhesion of antibody TS2/16-activated HUVE cells to TSP1-(1-175), but not to laminin-1 ( Fig. 2A). HUVE cell adhesion to this known ␣ 6 ␤ 1 ligand was only slightly inhibited by the anti-␣ 6 ␤ 1 antibody ( Fig. 2A), suggesting that ␣ 6 ␤ 1 integrin is either absent or not functional in these cells. To deter-mine whether the distinct behavior of HDMVE and HUVE cells was unique to endothelial cells from these anatomical sites, we tested microvascular and large vessel cells from different organs. Adhesion of iliac vein endothelial (AG10773A) cells was also insensitive to the anti-␣ 6 ␤ 1 antibody (Fig. 2B), whereas HMVE-L cell adhesion was inhibited by this antibody (Fig. 2C). Therefore, the selective expression or function of ␣ 6 ␤ 1 as a TSP1 receptor may be a general characteristic of microvascular endothelial cells.

FIG. 2.
Selective functions of ␣ 6 ␤ 1 integrin in microvascular endothelial cells. Antibody TS2/16-stimulated HUVE cells were seeded onto substrates coated with TSP1-(1-175) or laminin-1 (LN) in the absence or presence of (4-((2-methylphenyl)aminocarbonyl)aminophenyl)acetyl-LDVP (phLDVP) or antibody GoH3 (A). Unstimulated or antibody TS2/16 (5 g/ml)-stimulated adult iliac vein endothelial cells (AG10773A) (B) or HMVE-L cells (C) (2-2.5 ϫ 10 5 cells/ml) were seeded on dishes coated overnight with TSP1 (40 g/ml) or laminin-1 (5 g/ml). ␣ 6 ␤ 1 -Mediated adhesion was evaluated using antibody GoH3 (5 g/ml). Cell attachment was quantified as described in the legend to Fig. 1. Adhesion is expressed as the number of cells/mm 2 (ϮS.D.) from at least three different experiments. major ␤ 1 integrin in HDMVE cells that recognizes TSP2. ␣ 6 ␤ 1 Integrin Expression Levels Are Similar in Large and Small Vessel Cells-The levels of ␣ 6 ␤ 1 expression were measured to examine why microvascular and large vessel endothelial cells differed in their utilization of ␣ 6 ␤ 1 as a TSP receptor. ␣ 6 mRNA was expressed at similar levels in HUVE and HD-MVE cells based on reverse transcription-PCR (Fig. 4A). Cellsurface expression of ␣ 6 and ␤ 1 subunits assessed by flow cytometry did not differ between HUVE and HDMVE cells (Fig.  4B). However, analysis of the total expression levels by immunoprecipitation using the same anti-␣ 6 antibody revealed some differences in the ␣ 6 subunit (Fig. 4C). On an underexposed blot (Fig. 4C, upper panel), more intact 120-kDa ␣ 6 subunit was reproducibly detected in HDMVE cells than in HUVE cells. By densitometry, this represented a 1.6-fold difference in ␣ 6 subunit between HDMVE and HUVE cells. After a longer exposure (Fig. 4C, lower panel), we could also detect differences in the pattern of processed ␣ 6 subunits or integrin-associated proteins. In addition to the intact ␣ 6 subunit, HUVE cell immunoprecipitates contained 70-and 43-kDa proteins. The apparent masses of these proteins are consistent with the unreduced and reduced molecular masses reported for a structural variant of the ␣ 6 integrin called ␣ 6 parvus (26). Although the gel was run under reducing conditions, it is possible that the 70-kDa band corresponds to partially unreduced ␣ 6 parvus. In the ␣ 6 immunoprecipitate from HDMVE cells, these bands were less abundant relative to the intact 120-kDa ␣ 6 subunit, but an additional unknown protein was also detected. Likewise, a slight difference in the expression levels of this integrin was also found in HMVE-L and AG10773A cells (Fig. 4D). These minor differences may not be sufficient to account for the differential function of ␣ 6 ␤ 1 in these cells.
Activation States of ␣ 6 ␤ 1 Integrin Differ in HUVE and HD-MVE Cells-The differences in expression of ␣ 6 ␤ 1 integrin in HUVE and HDMVE cells are probably insufficient to account for the differences in its function. Alternatively, the activating antibody TS2/16 may not stimulate ␣ 6 ␤ 1 integrin equally in microvascular and large vessel endothelial cells. To test this hypothesis, we compared the activity of other integrin activation agonists in these cells (Fig. 5). In contrast to antibody TS2/16, MnCl 2 and lipopolysaccharide stimulated anti-␣ 6 ␤ 1 antibody-inhibitable adhesion of HUVE cells to TSP1 and lami- Products were analyzed after 30 cycles of amplification using a 1.5% agarose gel stained with ethidium bromide. B, ␣ 6 and ␤ 1 integrin-specific antibodies were used to compare protein expression of this TSP1 receptor by flow cytometry on the surface of HUVE and HDMVE cells. C, HUVE and HDMVE cells were surface-biotinylated and immunoprecipitated using the anti-␣ 6 antibody GoH3. Precipitates were analyzed by SDS-7.5% polyacrylamide gel electrophoresis under reducing conditions, transferred to a polyvinylidene fluoride membrane, and visualized by chemiluminescence. D, ␣ 6 expression levels in HMVE-L and AG10773A cells were quantified in cell lysates prepared as described for C, immunoprecipitated with the anti-␣ 6 antibody GoH3, and visualized by chemiluminescence. FIG. 3. N-terminal regions of TSP1 and TSP2 support ␣ 6 ␤ 1 integrin binding. Adhesion of antibody TS2/16-activated HDMVE cells to NoC1 and NoC2 (30 g/ml) was assessed in the presence of the ␤ 1 -blocking monoclonal antibody 13 (5 g/ml) (A) or antibody GoH3 (5 g/ml) (B). Cell attachment was quantified, and results are expressed as the number of cells/mm 2 (ϮS.D.) from at least three different experiments. nin-1. Notably, we previously found that lipopolysaccharide does not activate ␣ 3 ␤ 1 integrin in endothelial cells (18), suggesting that this agonist is selective for activating ␣ 6 ␤ 1 integrin. We also used antibody TS2/16 to activate HUVE cells exposed to conditioned medium from HDMVE cells to determine whether the latter cells release growth factors that facilitate ␣ 6 ␤ 1 activation. However, the ␣ 6 ␤ 1 -blocking antibody had no effect under these conditions (Fig. 5). We consistently observed less inhibition of adhesion to TSP1 and laminin substrates by the ␣ 6 ␤ 1 -blocking antibody following MnCl 2 or lipopolysaccharide addition in HUVE cells compared with HDMVE cells (data not shown). These results indicate that ␣ 6 ␤ 1 is more easily activated in microvascular than in large vessel endothelial cells.
Because integrin recognition sites often require an Asp or Glu residue (2), our second approach to identify a sequence recognized by ␣ 6 ␤ 1 integrin in TSP1 and TSP2 was to search for conserved Asp or Glu residues in the N-terminal modules of the known TSP1 and TSP2 sequences. Such conserved acidic residues are present at positions 14, 35, 90, 111, 126, 127, 145, and 162 of TSP1. We tested synthetic peptides containing six of the conserved acidic residues that were identified on this basis. Control peptides for each had Ala substituted for the conserved Asp or Glu residues. The conserved Asp/Glu residue at position 162 was previously identified as an ␣ 4 ␤ 1 -binding site in TSP1 and TSP2 (12) and did not affect ␣ 6 ␤ 1 -mediated adhesion. Synthetic peptides containing acidic residues 35 (ELT-GAARKGSGRRLVKGPD), 126 and 127 (VSVEEA), 111 (SNG-KAGTLDLS), and 145 (LFVQEDRAQLYI) were inactive (Fig.  6). However, a peptide containing Glu 90 (LALERKDHSG) strongly inhibited HDMVE cell adhesion to NoC2 (Fig. 7A), TSP1-(1-175) (Fig. 7B), and the E8 fragment of laminin-1 (Fig.  6). This inhibition was dose-dependent and specific in that control peptide LALARKDHSG was inactive (Fig. 7, A and B).
Glu 90 of TSP1 Is Required for ␣ 6 ␤ 1 Integrin Recognition-Several residues in the inhibitory sequence LALERKDHSG are highly conserved among TSP1 and TSP2 sequences from different species (Fig. 9A). Notably, the Glu residue is completely conserved in all known TSP1 and TSP2 sequences. To test the role of this residue, we mutated Glu 90 in TSP1-(1-175). Mutation of Glu 90 to Ala completely inhibited HDMVE cell adhesion to this portion of TSP1 (Fig. 9B). This loss of activity was not due to a global alteration in folding of the protein because the same mutation did not affect adhesion of Jurkat T cells to TSP1-(1-175) mediated by ␣ 4 ␤ 1 integrin (Fig. 9C) (12). Furthermore, the mutant protein retained the heparin-binding activity of the native sequence (34) (data not shown). These results indicate that Glu 90 plays an important role in ␣ 6 ␤ 1mediated adhesion to this portion of TSP1. The high degree of homology for this sequence among TSPs from different species and the total inactivation following mutation of the conserved Glu 90 residue strongly suggest that this constitutes the ␣ 6 ␤ 1binding site.
Role of ␣ 6 ␤ 1 Integrin in Endothelial Cell Chemotaxis-TSP1 inhibits microvascular endothelial cell motility induced by fi-  (1-175), or the E8 fragment at the same concentrations given in the legend to Fig. 6. Cells were incubated for 1 h in the presence of the ␤ 1 -activating antibody TS2/16 (10 g/ml). Jurkat T cell adhesion to TSP1 (10 g/ml) was determined using cells activated with antibody TS2/16 (4 g/ml) in the presence of the indicated TSP1 peptides (E). MDA-MB-231 breast carcinoma cell adhesion to NoC1 (15 g/ml) or type I collagen (5 g/ml) was determined using antibody TS2/16-activated cells in the presence or absence of LALERKDHSG (F). Cell attachment was quantified after 1 h by fixation with 1% glutaraldehyde and staining with Diff-Quick solution II. Results are expressed as the number of cells/mm 2 (ϮS.D.). broblast growth factor-2 through binding to its receptor CD36 (14), but TSP1 also stimulates chemotaxis of murine lung capillary cells (LE-II) and bovine aortic endothelial cells (35). The stimulatory activity of TSP1 was inhibited by an antibody that recognizes its N-terminal module (35), suggesting that this domain of TSP1 stimulates endothelial cell chemotaxis. Consistent with the function of ␣ 3 ␤ 1 integrin in aortic endothelial wound repair (18), the anti-␣ 3 ␤ 1 antibody P1B5 partially blocked HDMVE cell chemotaxis stimulated by TSP1 or NoC1 (Fig. 10, A and B). The anti-␣ 6 ␤ 1 antibody GoH3 partially blocked migration stimulated by TSP1, NoC1, or NoC2 (Fig. 10, A-C), demonstrating that ␣ 6 ␤ 1 integrin also mediates chemotaxis of endothelial cells to TSP1 and TSP2. The effects of the ␣ 3 -and ␣ 6 -blocking antibodies on TSP1-or NoC1-stimulated endothelial cell chemotaxis were additive (Fig. 10, A and B), indicating that both ␣ 3 ␤ 1 and ␣ 6 ␤ 1 integrins are necessary for migration to TSP1. Only the ␣ 6 -blocking antibody inhibited NoC2-stimulated migration, and no additivity was observed (Fig. 10C). Given that NoC2 lacks the binding site for ␣ 3 ␤ 1 integrin, ␣ 6 ␤ 1 integrin may be the primary receptor mediating endothelial cell chemotaxis to TSP2.
Motility responses of endothelial cells to TSP1 involve both random (chemokinesis) and directional (chemotaxis) components (35). To define the contribution of each to HDMVE cell migration induced by the ␣ 6 ␤ 1 -binding domains of TSP1 and TSP2, we exposed the cells to each with or without a gradient (Table I). TSP1, NoC1, or NoC2 was added to the upper chamber, the lower chamber, or both. Our results show that TSP1, NoC1, or NoC2 in the lower chamber gave the strongest stimulation of migration (Table I). However, in the absence of gradient, when the proteins were added to the upper and lower chambers, we also observed endothelial cell motility. These results indicate that as reported for native TSP1, the N-terminal domains of TSP1 and TSP2 have both directional and random effects on cell migration.
Although linear peptides are sufficient for recognition by many integrins, previous attempts to define recognition sites for ␣ 6 ␤ 1 in its protein ligands have yielded conflicting results. The mechanism by which laminins bind ␣ 6 ␤ 1 has been the most extensively studied. An ␣ 6 ␤ 1 -binding site in laminin-1 is clearly located in the C-terminal proteolytic E8 fragment (28,29). Two peptides derived from this region of the laminin ␣1 subunit, NRWHSIYITRFG (AG-10) and TWYKIAFQRNRK (AG-32), show some specificity for antagonizing ␣ 6 ␤ 1 (30), but other studies using recombinant fragments containing the same sequences argue against this hypothesis (43,44). Consistent with the latter, we detected no activity in TSP1 peptides paralogous to the laminin peptide AG-32.
Phage display screening for ␣ 6 ␤ 1 integrin ligands also failed to identify any known laminin sequences, but identified three peptides that inhibit laminin-1 binding to ␣ 6 ␤ 1 : VSWFS-RHRYSPFAVS, HRWMPHVFAVRQGAS, and FGRIPSPLAY-TYSFR (45). These sequences also bear no relationship to our TSP1 peptide. Finally, an ␣ 6 ␤ 1 -binding consensus sequence was identified in ADAM-2: X(D/E)ECD (46). The active ADAM-2 sequences resemble our TSP1 peptide in that they contain one or two Glu residues, the first having the same spacing from a conserved Asp residue as seen in the TSP1 sequences (Fig. 9A). However, based on differential regulation by phorbol esters and divalent cations, the activation states of ␣ 6 ␤ 1 that recognize laminin and ADAM-2 may be different (47), implying that the recognition mechanisms also differ. Therefore, it remains to be determined whether TSPs and ADAM-2 share a common binding mechanism for ␣ 6 ␤ 1 .
Based on sequence alignment of TSP N-terminal modules with laminin G modules (27), the ␣ 6 ␤ 1 -binding sequence in TSP1 should span the C-terminal end of strand D and the loop between strands D and E of its predicted secondary structure (Fig. 11). Therefore, we predict that ␣ 6 ␤ 1 -binding sites in laminins may also reside in the D-E loops of their G modules. Based on a previous alignment of G modules (44), Glu residues are common in the D-E loops (Fig. 11). The positions of the Glu residues are variable, however; so further work is needed to determine whether any of these are part of the ␣ 6 ␤ 1 -binding sites in laminins. The D-E loop is surface-exposed on laminin G modules (44), suggesting that the inhibitory sequence we identified may also be exposed on the TSP N-terminal modules.
Does ␣ 6 ␤ 1 play a role in angiogenesis? In vivo expression studies demonstrated that ␣ 6 integrins are highly expressed in capillary endothelial cells (48), although ␣ 6 ␤ 1 and ␣ 6 ␤ 4 could not be distinguished by this method. The ␣ 6 null mouse is not informative for this issue because ␣ 6 ␤ 4 plays a critical role in FIG. 11. Alignment of laminin G module sequences with the ␣ 6 ␤ 1 -binding sequences of thrombospondins. Alignment of laminin ␣3-5 sequences and strand assignments are according to Ref. 44. Only the N-terminal portion of the extended D-E loop of laminin ␣1G3 is shown. The underlined sequence indicates the ␣ 6 ␤ 1 -binding peptide identified in human (h) TSP1.
FIG. 10. ␣ 6 ␤ 1 -dependent endothelial cell chemotaxis to TSP1, NoC1, and NoC2. Chemotaxis was assessed in modified Boyden chambers. HDMVE cell chemotaxis was induced by 30 g/ml TSP1 (A) or 20 g/ml recombinant NoC1 (B) or NoC2 (C) in medium 199 with 0.1% BSA added to the lower chamber. Cells (2-3 ϫ 10 5 /well) added to the upper chamber were allowed to migrate for 6 h at 37°C in 5% CO 2 . HDMVE cell migration induced by the indicated attractants was assayed in the absence or presence of the ␣ 3 ␤ 1 -blocking antibody (2 g/ml), the ␣ 6 ␤ 1 -blocking antibody (10 g/ml), or both combined. Migrated cells were counted microscopically after fixation. Results are presented as the number of cells migrated per field (ϮS.D.). epithelial integrity that results in perinatal lethality (49). ␣ 6 ␤ 4 also recognizes laminin-8 and laminin-1 E8 fragments, but we do not know whether ␣ 6 ␤ 4 can recognize TSP1 or TSP2. However, an ␣ 6 ␤ 1 -blocking antibody prevents cord formation by endothelial cells plated on Matrigel, implicating ␣ 6 ␤ 1 integrin in capillary morphogenesis (50). The ␣ 6 ␤ 1 ligand Cyr61 is also a stimulator of angiogenesis (51) and promotes vascular smooth muscle chemotaxis mediated by ␣ 6 ␤ 1 integrin (52). Deletion of Cyr61 disrupts developmental angiogenesis (53), and ␣ 6 ␤ 1 mediates pro-angiogenic activities of this molecule for HUVE cells (54). Our results demonstrate that ␣ 6 ␤ 1 integrin also mediates TSP1-and TSP2-stimulated chemotaxis in microvascular endothelial cells. Thus, the ␣ 6 ␤ 1 ligands TSP1 and Cyr61 share at least two target cell types and some biological activities. However, a report published during the review of this manuscript identified a recognition sequence in Cyr61 for ␣ 6 ␤ 1 that bears no similarity to the sequence we identified in TSP1 (25).
With the addition of ␣ 6 ␤ 1 , microvascular and large vessel endothelial cells are now known to differ in their expression or regulation of three TSP1 receptors. CD36 is expressed selectively in capillary endothelium (55) and is clearly required for some activities of TSP1 (14). Both ␣ 4 ␤ 1 and ␣ 6 ␤ 1 are generally expressed in endothelial cells in vitro, but we have shown that ␣ 4 ␤ 1 functions selectively in large vessel endothelial cells, 2 whereas ␣ 6 ␤ 1 is preferentially activated in microvascular cells. We are currently investigating the mechanism for this differential integrin activation and its functional role in mediating the effects of thrombospondins on angiogenesis.