Lysosomal Integral Membrane Protein II Binds Thrombospondin-1 STRUCTURE-FUNCTION HOMOLOGY WITH THE CELL ADHESION MOLECULE CD36 DEFINES A CONSERVED RECOGNITION MOTIF*

LIMPII (lysosomal integral membrane protein II) is one of a family of proteins structurally related to the cell surface glycoprotein CD36. We recently defined a single structural domain on CD36 that mediates binding to adhesive glycoprotein thrombospondin-1 (TSP1). The CD36-TSP1 interaction is known to play a role in plate-let-tumor and platelet-monocyte adhesion, angiogenesis, and in monocyte uptake of apoptotic cells. To test whether LIMPII also binds TSP1, a LIMPII peptide corresponding to the TSP1 binding domain of CD36 was expressed as a recombinant glutathione S -transferase (GST) fusion protein. In solid phase binding assays, purified 125 I-TSP1 bound to immobilized GST/LIMPII in a time-dependent and saturable manner. Inhibition by excess unlabeled TSP1 or EDTA demonstrated specificity. LIMPII z TSP1 complex formation was specifically blocked by soluble LIMPII fusion protein, by monospecific rabbit IgG directed against the LIMPII peptide and by CD36 fusion proteins containing the TSP1 binding domain. Transfection of Bowes melanoma cells with a chimeric LIMPII cDNA that targets expression to the plasma membrane conferred the ability to bind 125 I-TSP1 and to adhere to TSP1-coated surfaces. This study defines a TSP1 binding site conserved between LIMPII and CD36 and suggests that cell surface LIMPII may function in some circumstances as an adhesion

TSP1, a 450-kDa trimeric glycoprotein component of extracellular matrix, is a well characterized ligand of CD36 (reviewed in Refs. 27 and 35). It is a major component of platelet ␣-granules secreted upon platelet activation (36). TSP1 functions in various aspects of vascular biology, including platelet aggregation (28), angiogenesis (29,30), transforming growth factor ␤ activation (31), smooth muscle proliferation (32), and plasmin generation (33,34). It also modulates adhesion and migration during embryonic development (37,38) and has been implicated in such disease processes as thrombosis, atherogenesis, and tumor metastasis (39). While CD36 is only one of multiple TSP1 receptors, we and others have demonstrated that the CD36-TSP1 interaction is involved in platelet-monocyte adhesion (40,41), platelet-tumor cell adhesion (42), macrophage uptake of apoptotic cells (17,18), and tumor cell-substratum adhesion (43). CD36 recently has been shown to mediate * This work was supported by Grants HL46403, HL18828 (Thrombosis SCOR), and DE11348 from the National Institutes of Health and by the Charles Fogarty Trust. 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.
We recently defined a single TSP1 binding domain in CD36 (47). Given the extent of conserved structure between LIMPII and CD36, we were interested in learning whether LIMPII shares CD36 TSP1 binding activity. LIMPII has been identified as a lysosomal membrane component in all cell types examined to date (9,48). It is not known whether LIMPII may function at the cell surface, nor has a ligand been identified to indicate that LIMPII could function as an authentic adhesion receptor. However, under certain circumstances, soluble lysosomal enzymes and membrane proteins are expressed at the plasma membrane. We reported previously that LAMP-1 and LAMP-2 (members of a subclass of lysosomal membrane glycoproteins not related to LIMPII) are expressed on activated platelets (49,50), while others have shown LAMP expression on tumor cells and activated leukocytes (51)(52)(53). These observations support the possibility that LIMPII might act at the cell surface and thus may share TSP1 receptor function with CD36.

EXPERIMENTAL PROCEDURES
LIMPII Fusion Proteins-Sau3A fragments from base pairs 464 -704 of the LIMPII cDNA (kindly provided by Dr. I. Sandoval, Centro de Biología Molecular, Madrid, Spain (9)) encoding aa 75-155, and from base pairs 705-983 encoding aa 156 -243, were subcloned into prokaryotic expression vector pGEX-3X (Pharmacia Biotech Inc.), directly adjacent to a fragment of the recombinant glutathione S-transferase (GST) gene, downstream of an inducible LacZ promoter. Fig. 1 shows the orientation of LIMPII fusion proteins, and CD36 fusion proteins (CFP93-120, CFP67-157, and CFP298 -439 (47)), relative to the native LIMPII and CD36 sequences. A truncation mutant terminating after codon four of LFP75-155 was used as a control (GST1, LFP75-78). All plasmid constructs were restriction-mapped and sequenced across fragment junctions to confirm correct orientation and intact reading frame. Large scale production of soluble fusion protein by the method of Frangioni and Neel (54) included single step affinity purification on glutathione-Sepharose 4B (Pharmacia). Glutathione eluates were dialyzed extensively prior to storage at Ϫ70°C in 10% glycerol. For some experiments, LIMPII peptides were cleaved from the GST moiety with coagulation Factor Xa (Boehringer Mannheim) and purified by size exclusion chromatography (Centricon 10; Amicon, Beverly, MA). Fusion proteins were analyzed by 8% SDS-polyacrylamide gel electrophoresis or gel filtration (Superose-12, Pharmacia) to determine apparent molecular weights (listed in Fig. 1) and by Western blot and enzyme-linked immunosorbent assay to confirm size and document immunoreactivity.
Purified Proteins, Peptides, and Antibodies-Human platelet-rich plasma and outdated platelet concentrates were obtained from the New York Blood Center. Purified human calcium-replete TSP1 was prepared from releasate of thrombin-activated washed platelets by heparin af-finity and anion exchange chromatography on Mono Q-Sepharose (Pharmacia) (34). TSP1 peptides were from Chiron Mimotopes, Victoria, Australia. Human thrombin was from Boehringer Mannheim, and epinephrine was from Biodata Corp., Horsham, PA. CD36 was prepared from human platelet Triton X-114 extracts as reported previously (40,55,56). Murine monoclonal anti-CD51 (69.6.1) was from Immunotech, Inc., Westbrook, ME. Monoclonal anti-LAMP-1 and -2 and anti-TSP1 IgGs are described elsewhere (40,49,50,57). Anti-CD36 monoclonal antibody 8A6 was a gift of Dr. J. Barnwell, New York University Medical Center, New York. Domain-specific rabbit anti-LIMPII was generated by subcutaneous inoculation of the 10-kDa L75-155 thrombin-cleaved peptide. IgG was purified from serum by protein A chromatography (Pierce). Following depletion of GST-specific antibody on immobilized GST1 (HiTrap columns; Pharmacia), anti-LIMPII peptide IgG was purified by affinity capture on immobilized LFP75-155. These were determined to be monospecific by enzyme-linked immunosorbent assay and Western blot analyses.
Cell Lines-An EcoRI fragment encoding a chimeric LIMPII cDNA was subcloned into eukaryotic expression vector pMV7 (58) (gift of Dr. Moses Chao, Cornell Medical College). This cDNA contains the luminal and transmembrane domains of rat LIMPII coupled to the C-terminal cytoplasmic tail of human CD36 and is targeted to the plasma membrane (9). Bowes melanoma cells were transfected by liposome-mediated fusion (Lipofectin™; Life Technologies, Inc.) and stable lines selected and maintained as described for CD36-transfected Bowes cells (42). Cells with high LIMPII surface expression were enriched by adsorption with polyclonal anti-L75-155 IgG and subsequent selection with sheep anti-rabbit IgG-conjugated magnetic beads (Dynabeads M-450; Dynal Inc. Great Neck, NY). Surface expression was confirmed prior to all studies by immunofluorescence flow cytometry as described previously (42).
Solid Phase Binding Assays-TSP1 (5-10 g/ml) or fusion proteins and peptides (10 -20 g/ml) were immobilized on detachable 96-well strips (Immulon-4 Remove-a-well, Dynatech Laboratories, Inc.) by overnight incubation at 4°C in carbonate buffer (100 mM Na 2 CO 3 , 1 mM MgCl 2 , 0.02% NaN 3 , pH 9.8). Saturable coating determined using radiolabeled proteins ranged from 200 to 280 ng/well. Washed wells were blocked with 1% BSA, then incubated in triplicate with soluble radiolabeled ligand as described previously (47). After extensive washing, bound radioactivity was quantified by a ␥ counter. Radiolabeling was performed with Na 125 I (Amersham Life Science Inc.) using immobilized chloramine T (IODO-BEADS; Pierce) (40,56). Specific activity determined for each experiment ranged from 0.06 to 1.0 Ci/g. Nonspecific binding was determined by carrying out binding in the presence of 10 mM EDTA or excess unlabeled ligand and was generally less than 10% of total.
Cell Binding Assays-Binding of 125 I-labeled TSP1 to suspensions of LIMPII-transfected Bowes cells was measured as described previously (40,42). Inhibition studies used 5-10-fold molar excess competitor protein, 80 g/ml IgG, or 1 mM EDTA. All assays were performed in triplicate. Binding of 125 I-TSP1 to activated platelets was measured using platelet monolayers prepared as described below.
Melanoma Cell Adhesion Assays-Platelet-tumor cell adhesion assays were performed as described (42). Briefly, washed rat or human platelet suspensions (4 ϫ 10 7 /ml) were aliquoted into 96-well tissue culture plates, and adherent "monolayers" formed by centrifuging plates at 1100 ϫ g for 10 min at 22°C. Platelets then were activated with 0.5 units/ml thrombin for 2 min at 37°C. Triplicate wells were washed, blocked with 20 mM HEPES, 150 mM NaCl, 0.75 mM Na 2 HPO 4 , pH 7.04, 1% BSA, then incubated with 2 ϫ 10 4 melanoma cells/well for 1 h at 37°C in 5% CO 2 . Wells were gently washed, fixed in 1% glutaraldehyde, and adherent cells scored by counting identical 1-mm 2 sectors under phase microscopy. Nonspecific adhesion was assessed as binding to BSA-coated wells. TSP1 adhesion assays were performed with identical conditions, except that instead of platelet monolayers, wells were coated with 5 g/ml TSP1 in carbonate buffer overnight at 4°C.
Immunofluorescence Flow Cytometry of Rat Platelets-Platelets were obtained by exsanguination of 1-year-old Sprague-Dawley rats. Platelet-rich plasma was harvested after low speed centrifugation of anticoagulated blood and washed platelet suspensions prepared as described (42). Surface expression of LIMPII and LAMP-1 and -2 on resting platelets, or on platelets treated with either 0.5 unit/ml human thrombin or 0.1 unit/ml epinephrine, was assessed by indirect immunofluorescence flow cytometry as described (49,50).

TSP1 Binds LIMPII
Interaction of TSP1 with LIMPII/GST Fusion Proteins-In solid phase binding assays, soluble 125 I-labeled TSP1 bound to immobilized LIMPII fusion protein LFP75-155, spanning aa 75-155 ( Fig. 2A), but not to immobilized control fusion protein GST1 (LFP75-78), nor to a downstream LIMPII fusion protein (LFP156 -243). Binding was time-dependent ( Fig. 2B), reaching equilibrium within 3 h, and upon addition of excess unlabeled TSP1 at equilibrium, was reversed by ϳ50% within 30 min and ϳ75% at 2 h. In complementary studies in which TSP1 was immobilized, soluble 125 I-LFP75-155 bound to TSP1 in a time-dependent reversible manner and reached equilibrium by 4 h (not shown), suggesting that binding was not related to artifactual influences of protein immobilization.
Saturation binding isotherms over a range of 1 nM to 1 M showed concentration-dependent binding of 125 I-TSP1 to immobilized LFP75-155 ( Fig. 2A), with an apparent affinity of 70 Ϯ 9 nM (Table I). Similar results were obtained in reverse phase assays that measured 125 I-LFP75-155 binding to immobilized TSP1, yielding an apparent affinity of 98 Ϯ 20 nM. These values are comparable with that observed for TSP1 binding to intact purified platelet CD36 (227 Ϯ 20 nM) and to CD36 fusion proteins CFP93-120 (8.6 Ϯ 4 nM) or CFP67-157 (153 Ϯ 17 nM) (47). A stoichiometry of ϳ1.6 molecules of LFP75-155 complexed with each immobilized TSP1 molecule, and ϳ2.1 of TSP1 to immobilized LIMPII was seen. This is consistent with the homotrimeric structure of intact TSP1, with one face inaccessible on the immobilized ligand. These results demonstrate that the TSP1 binding domain is structurally maintained and functionally conserved between LIMPII and CD36.
CD36 Competes with LIMPII for TSP1 Binding-Because identity between the TSP1 binding domain of CD36 and the corresponding region of LIMPII is higher than overall homology (see below), we tested whether they might compete for a common binding site on TSP1. Intact purified platelet CD36 inhibited binding of 125 I-TSP1 to immobilized LFP75-155 by 92 Ϯ 2% (Fig. 3). This result was reproducible using CD36 fusion proteins containing the TSP1 binding domain. CFP67-157 inhibited by 81 Ϯ 3%, while the 27-aa minimal binding site of CFP93-120 showed 95 Ϯ 1% inhibition. In contrast, neither CFP298 -439 nor LFP156 -243 fusion proteins derived from regions downstream of the TSP1 binding domain were able to block this interaction (not shown). Results over a range of competitor concentrations (10 nM to 1 M, not shown) indicated apparent IC 50 values for CFP67-157, CFP93-120, and native CD36 of ϳ80, 55, and 46 nM, respectively (Table I). These results strongly suggest that the corresponding region of LIM-PII shares overlapping function with the CD36 TSP1 binding domain.
CD36 is known to bind to the sequence CSVTCG within the type I properdin repeat of TSP1 (43). Therefore, we tested the ability of this TSP1-derived peptide to inhibit 125 I-TSP1 binding to immobilized LIMPII fusion protein. Excess soluble CS-VTCG peptide blocked 66 Ϯ 3% of specific TSP1 binding (Fig.  3), compared with a scrambled control peptide VGSCCT (20 Ϯ 4%). An RGD-containing peptide, also a cell recognition motif  found in the C-terminal type 3 calcium-binding repeat of TSP1, showed less efficient inhibition (32 Ϯ 6%). We conclude that the TSP1 binding domain of LIMPII is functionally identical to that of CD36 and interacts directly with the CD36-binding sequence on TSP1.

Surface Expression of LIMPII Confers TSP1 Binding
Ability to Transfected Melanoma Cells LIMPII-transfected Cells Acquire the Capacity to Bind TSP1-To investigate whether LIMPII could serve as a cell surface receptor for TSP1, we employed a transfected cell system, which serves as a paradigm for studying CD36 function (42). Bowes melanoma cells were stably transfected with a chimeric LIMPII cDNA containing the luminal and transmembrane domains of LIMPII coupled to the C-terminal cytoplasmic tail (6 residues) of CD36 (9), that targets expression to the plasma membrane, such that the intraluminal domain of LIM-PII is oriented extracellularly. Surface expression of LIMPII on transfected Bowes cells was confirmed by flow cytometry using affinity-purified anti-LFP75-155 IgG (Fig. 4). As shown in Fig.  5, LIMPII-transfected Bowes cells bound soluble 125 I-TSP1, in contrast to control cells transfected with pMV7 vector alone and to a similar extent as CD36-transfected cells (not shown). Binding was inhibited in the presence of excess unlabeled TSP1 (93 Ϯ 0.5%) and was completely abolished by 1 mM EDTA (not shown), demonstrating specificity. In addition, fusion proteins LFP75-155 (95 Ϯ 0.3%) and CFP93-120 (93 Ϯ 2%), but not GST-1, were able to compete significantly for TSP1 binding, providing further evidence of a common TSP1 binding domain.
LIMPII Is Expressed on the Surface of Activated Platelets and May Participate in TSP1 Binding-We previously established that lysosome-associated membrane proteins are expressed preferentially on the surface of activated platelets (49,50). Evidence of lysosomal membrane flow and fusion predicts that LIMPII also would appear as a surface membrane component upon stimulation of platelet secretion. To test this hypothesis, we used immunofluorescence flow cytometric analysis to assess the presence of native LIMPII. As shown in Fig. 7, anti-L75-155 IgG bound to the surface of thrombin-activated platelets (mean channel fluorescence (mcf) ϭ 0.256), but not to resting platelets (mcf ϭ 0.165). The level of anti-LIMPII IgG fluorescence appeared similar to that of anti-LAMP-1 IgG (mcf ϭ 0.242), while nonimmune IgG reactivity with activated or resting platelets was indistinguishable from that of anti-LIMPII IgG reactivity to resting platelets (mcf ϭ 0.155). LIM-PII expression did not follow stimulation with epinephrine (mcf ϭ 0.174), consistent with previous data showing that weak agonists did not induce surface appearance of platelet lysosomal proteins (49,50). By inference from anti-LAMP-1 relative  Table I) were calculated as the mean of all data sets. fluorescence intensity, LIMPII expression probably also is in the range of 1500 molecules/platelet (50). This is in contrast to 20,000 -60,000 TSP1 binding sites on activated platelets (27), suggesting that LIMPII accounts for a minor fraction of potential TSP1 binding sites.
In an attempt to dissect out the contribution of LIMPII to total platelet TSP1 binding, inhibition studies were performed comparing the effect of LFP75-155 to that of blocking reagents with specificity toward the full array of TSP1 receptors. In the absence of competitor, 21 Ϯ 0.6 nmol of 125 I-TSP1 bound per mg of thrombin-stimulated platelets (Fig. 8). Binding to resting platelets was significantly less than to activated platelets. Surprisingly, TSP1-specific binding to activated platelets was inhibited 55 Ϯ 8% in the presence of 10-fold molar excess LFP75-155, far greater than that predicted from estimates of total LIMPII expression and ϳ20% greater than inhibition by a CD36 fusion protein. Control fusion proteins or GST1 alone were ineffective inhibitors. Similarly, the peptide-specific anti-L75-155 IgG and a monoclonal anti-CD36 IgG (FA6) each blocked ϳ40% of TSP1 binding, while neither nonimmune rabbit or mouse IgG nor an anti-LIMPII IgG raised against a different peptide showed inhibitory activity. To characterize further the extent of CD36/LIMPII-mediated binding in this system, inhibitory TSP1 peptides were tested. A CSVTCGcontaining peptide from the type I properdin domain, known to block CD36-TSP1 binding, inhibited 51 Ϯ 2% of total TSP1 binding to activated platelets, while no significant inhibition was seen with a scrambled control peptide VGSCCT (not shown). TSP also can bind to ␤ 3 and ␤ 1 integrins on activated platelets and to fibrinogen and fibronectin bound in turn to integrins. Hence, RGD-containing peptides that block TSP1integrin interactions and fibrinogen/fibronectin-platelet binding also should block TSP1-platelet binding. We found in this assay that monoclonal antibodies against the ␤ 3 integrins blocked 42 Ϯ 2% of 125 I-TSP1 binding, while the synthetic peptide GRGDS inhibited 49 Ϯ 1%. These results suggest that the LIMPII fusion protein LFP75-155 and antipeptide antibody blocked binding of TSP1 to related domains on both CD36 and LIMPII (and perhaps other proteins) and that approximately 50% of total TSP1 binding was mediated by these domains.

Identification of an Evolutionarily Conserved Structural Domain from Analysis of the TSP1 Binding Motif in CD36 Gene Family Members
Sequence comparison within the CD36 gene family was performed to identify evolutionarily conserved amino acid residues likely important in interactions with TSP1. Using huCD36 aa 76 -163 (corresponding to LIMPII aa 75-155) as query, a BLAST search (59) successfully identified additional C. elegans homologs. These included F11C1.3 and CO3F11 (531 and 562 residues, chromosome 3), R07B1.3 (536 residues, chromosome X), and F07A5.3 (590 residues, chromosome I), bringing the total number of CD36 gene family members to 14. Two Saccharomyces cerevisiae genes, Nes24p and HDF1, may represent candidate yeast homologs. Multiple sequence alignments (ClustalW (60), MAP (61), PIMA (62, 63)) of deduced amino acid sequences corresponding to LIMPII aa 75-155 (Fig. 9) revealed a highly homologous stretch from aa 89 -123, with 59% identity and 68% similarity between LIMPII and CD36, suggesting the FIG. 4. Immunofluorescence flow cytometric demonstration of melanoma cell surface LIMPII expression. Bowes melanoma cells were trypsinized, incubated with affinity-purified anti-L75-155 IgG for 1 h at 4°C, followed by fluorescein isothiocyanate-conjugated goat anti-rabbit IgG for 45 min. Cells were fixed briefly with 2% paraformaldehyde and flow cytometry performed as described previously (42). Histogram shows antibody binding to LIMPII-transfected Bowes cells (peak B), but not to pMV7-transfected control cells (peak A). existence of a conserved TSP1 recognition motif. Using Block-Maker (combined Gibbs/MOTIF analysis (64)), which identifies short runs of ungapped homology within related sequences indicative of common motifs, three highly conserved segments were identified. Block A (Gibbs only) represents a protein kinase C consensus phosphorylation site (GPYTYR). Block B (MOTIF only) begins at the 5Ј boundary of CD36 Exon 5 (aa 95-143) (65). Block C appeared identical by either analysis. This provides strong evidence of actual motif structure, since each algorithm is based on different principles with complementary strengths and weaknesses. The average interblock distance of 10 residues between Blocks B and C predicts that this motif could appear in other proteins as a bipartite domain. DISCUSSION The studies reported here demonstrate that a TSP1 binding domain originally identified in CD36 is conserved in the genetically related protein LIMPII. In vitro binding of TSP1 to a LIMPII peptide and to LIMPII cDNA-transfected cells was similar in apparent affinity to CD36 and CD36-derived peptides. Close structural similarity was confirmed further by the ability of the CD36 fusion protein CFP93-120 (minimal TSP1-binding peptide) and TSP1-derived peptide CSVTCG to block TSP1 binding to LFP75-155 and by the ability of LIMPII and CD36 peptides to cross-compete for TSP1 binding to LIMPIItransfected cells. These data indicate that both CD36 and LIM-PII recognize the same binding site on TSP1.
The biological significance of the TSP1-LIMPII interaction is not known. Recently, Cuervo and Dice (66) described LAMP-2 as a receptor for selective uptake of intracellular substrates for lysosomal proteolysis in a reconstituted in vitro system. This raises the possibility that native LIMPII also may function in a specialized import pathway and might serve as a lysosomal receptor for TSP1. Rapid internalization and degradation of fluorescently-labeled TSP1 by LIMPII-transfected melanoma cells similar to CD36 transfectants further supports such a role. 4 Lysosomal CD68 (LAMP-related macrosialin) also has been implicated as a surface receptor for oxidized low density lipoproteins on activated macrophages (67). The possibility that LIMPII shares scavenger receptor function with CD36 and SR-BI is supported by recent studies from our laboratory indicating that LFP75-155 specifically inhibited binding of oxi-4 R. Crombie and S. F. A. Pearce, unpublished data. were fixed briefly in 2% paraformaldehyde. Platelet suspensions were incubated with primary antibody for 1 h at 22°C, followed by fluorescein isothiocyanateconjugated goat anti-rabbit IgG for 45 min, and analyzed as in Fig. 5. A-C show platelet reactivity with polyclonal anti-LIMPII IgG. D and E show platelet reactivity with nonimmune, anti-LAMP, and anti-TSP1 polyclonal IgGs, respectively. x ϭ mean channel fluorescence (n ϭ 2). dized low density lipoproteins to CD36. 5 In addition to providing insight into the structural basis of TSP1 binding by CD36-related proteins, demonstration of activation-dependent expression of LIMPII on the surface of platelets presents the possibility of a direct role as a cell adhesion molecule. In certain physiological settings, surface expression of lysosomal membrane proteins is a common feature. For example, we have shown previously that LAMP-1 and -2 translocate to the cell surface during platelet activation (49,50). Others have confirmed this phenomenon in neutrophil and monocyte activation (69,70). LIMPII might participate in normal function of platelets and leukocytes by contributing to pro-adhesive changes characteristic during coagulation, inflammation, and immune responses. Another resident lysosomal protein recognized as a platelet activation marker, the four-transmembrane superfamily prototype CD63 (LIMPI, LAMP-3), recently was shown to interact directly with integrin cell adhesion receptors (71).
Our transfected melanoma cell model provides evidence that cell surface LIMPII has the capacity to act as an authentic TSP1 receptor on tumor cells. Of particular interest is the observation that surface expression of lysosomal membrane proteins correlates with increased metastatic potential (72)(73)(74)(75)(76)(77). Expression of LAMP-1 and -2 and lysosomal CD63 on tumor cells is well documented (52,53,78), and monoclonal anti-LAMP-1 antibody was shown to inhibit tumor cell migration (76). It has been proposed that TSP1-mediated platelet-tumor cell thrombi facilitate extravasation by trapping circulating metastatic cells in the microvasculature (57). LIMPII-dependent modulation of tumor cell adhesion to matrix or to platelet TSP1 could promote tumor progression and metastasis.
Computer-assisted comparison of the TSP1 binding region in CD36-related genes revealed three blocks of homology (Fig. 9). FIG. 8. Competitive inhibition of 125 I-TSP1 binding to activated platelets. Thrombin-activated platelet monolayers were formed on removable 96-well strips and blocked with 1% BSA. A fixed concentration of 125 I-TSP1 (100 nM) was added in the absence or presence of 1 M competitor fusion proteins or peptides, or with ϳ100 g/ml IgG, and assayed as for solid phase binding (Fig. 3). Inhibition is expressed relative to maximum specific binding of TSP1 in the absence of competitor or antibody (n ϭ 2 for fusion protein and LIMPII antibodies; n ϭ 1 for peptides and monoclonal IgGs; error as S.D.).
The first is a protein kinase C consensus sequence (GPYTYR). Asch et al. (13) have shown that protein kinase C-dependent phosphorylation of CD36 Thr 92 blocked TSP1 binding to CD36, and may regulate CD36 function in vivo. Consistent with this model, we have shown that while Thr 92 is not necessary for TSP1 binding activity, phosphorylation of peptides containing Thr 92 partially inhibited TSP1 binding (15). Thr 93 of some family members, including Cla-1/SR-BI, is replaced by Val; therefore, one might predict constitutive rather than regulated binding. Definition of the protein kinase C phosphorylation consensus site as an optional regulatory domain, separate from the TSP1 binding motif, would be consistent with CD36 genomic organization. Exon 5 encodes aa 95-143 of human CD36, corresponding to aa 93-143 of LIMPII. The 48 -50 residue peptide contains the other two TSP1 binding motif blocks and thus is an excellent candidate to fit a model of evolutionary exon shuffling of a functional "module." As the first structural domain delineated within the CD36 gene family, we propose to designate this motif CLESH-1 (CD36, LIMPII, emp, SR-BI Homology sequence 1).
Finally, using a pattern-based BEAUTY homology search (68), we have identified CLESH-1 related sequences in several classes of proteins outside the CD36 family. These include several proteins known to bind TSP1. Some of the matched sequences conform to a bipartite organization, as predicted by separate motif blocks (Fig. 9). Our studies provide direct evidence that this motif can function as a site for protein-protein interaction, showing TSP1 binding by either the CD36 or LIMPII domain and blockade of TSP1 binding by fusion proteins containing the CD36/LIMPII TSP1 binding motif. While the role of CLESH-1 sequences in other proteins remains to be determined, we have shown that a protein not previously known to interact with TSP1 (human immunodeficiency virus type 1 envelope gp120) contains a split CLESH-1 motif and complexes with TSP1 (5). Thus, the CLESH-1 motif clearly is indicative of a potential TSP1 recognition module. Further identification and functional characterization of CLESH-1-related domains may facilitate elucidation of novel TSP1 interactions.