Identification and Characterization of LAMP-l as an Activation-dependent Platelet Surface Glycoprotein*

Platelets normally circulate in a quiescent state. When activated, they undergo biochemical and morphological changes which greatly alter their function and contribute to their role in thrombosis and hemostasis. We have identified, cloned, and sequenced a cDNA from a human unbilical vein endothelial cell library that encodes a 110-kDa integral membrane protein. This protein is present on the surface of activated but not resting platelets and has previously been identified as lysosomal-associated membrane protein 1 (LAMP-1). Half-maximal surface expression of platelet LAMP-1 was induced by concentrations of thrombin that resulted in lysosome enzyme release, not alpha-, or dense granule release. Also consistent with lysosome enzyme studies, there was little surface expression of LAMP-1 in response to the weak agonists ADP and epinephrine. In addition, sucrose density gradient fractionation of platelet granules showed colocalization of LAMP-1 with the lysosomal enzyme, beta-galactosidase, and not with markers of alpha- or dense granules. While we found virtually no LAMP-1 on the resting platelet surface (0-90 molecules/cell), we estimated a mean of 1175 LAMP-1 molecules on the thrombin-activated platelet surface. The translocation of this heavily glycosylated protein to the platelet surface upon stimulation may play a role in the adhesive, prothrombic nature of these cells.

The translocation of this heavily glycosylated protein to the platelet surface upon stimulation may play a role in the adhesive, prothrombic nature of these cells.
Upon contact with any of a variety of stimuli, platelets change shape, release their granular components, and aggregate with one another to form the primary thrombus. As part of this change from a resting to a prothrombotic, adhesive phenotype, the protein profile of the cell surface undergoes a transformation. This is accomplished by several incompletely understood mechanisms. For example, a calcium-dependent conformational change occurs in the glycoprotein IIb/IIIa integrin complex, resulting in the acquisition of functional capacity as a fibrinogen receptor, which then mediates platelet aggregation (Bennet and Vilaire, 1979). PADGEM (GMP140), an a-granule integral membrane protein, translocates to the plasma membrane (Hsu-Lin et al., 1984;Berman et al., 1986) where it functions as a leucocyte "receptor"  (Larsen et al., 1989). Thrombospondin, another m-granule protein, is secreted and binds to receptors on the platelet surface. There it stabilizes the nascent aggregate and may act as an adhesive bridge between activated platelets and other cells (e.g. monocytes, macrophages), or extracellular matrix (Silverstein and Nachman, 1987). Fibrinogen (Bennet and Vilaire, 1979), fibronectin (Plow and Ginsberg, 1981), vitronectin, and von Willebrand factor (Koutts et al., 1978, McEver andMartin, 1984) are also preferentially expressed on activated platelets.
Characterization of the molecular and cellular details of the activated platelet surface is important in understanding normal platelet function, as well as the pathophysiology of thrombosis and atherosclerosis.
We now report a novel molecular approach to identify other membrane proteins that may play a role in regulating platelet function. In this paper we detail the cloning of one such protein, whose expression is agonistdependent. This protein has also been identified in other cells (Chen et al., 1988; as human lysosomalassociated membrane protein 1 (LAMP-l).' In nucleated cells, LAMP-l has been characterized as an integral membrane protein of the lysosome. The role of LAMP-l on the activated platelet surface is unknown, but activation-dependent translocation of this heavily glycosylated lysosomal protein to the plasma membrane may be part of the conversion of the platelet membrane to an adhesive, prothrombotic surface.  Isopropyl-P-D-thiogalactopyranoside,  ADP, and indomethacin  were  purchased  from Sigma. D-Phenylalanyl-L-propyl-L-arginine chloromethyl ketone was from Calbiochem.

Library Screening
Two million XGTll recombinants were plated in the primary library screen. After 3.5 h of infection, the bacteria were overlaid with nitocellulose previously soaked in 10 mM isopropyl-/3-D-thiogalactopyranoside (Young and Davis, 1983). The filters were screened immunologically with the anti-platelet IgG, and positive clones were identified using an alkaline phosphatase calorimetric system (Blake et al., 1984 The platelets were then layered over silicone oil and bound radioactivity separated from free by centrifugation, as previously described (Silverstein and Nachman, 1987). For some studies, platelets were treated for 10 min prior to addition of antibody with specific agonists (ADP, epinephrine, or thrombin), as above. Specific binding was quantified as that inhibited by a 20-fold excess of unlabeled antibody and was generally >90%.

Platelet Release Assays
Serotonin release was measured by loading platelets with ["C]5hydroxytryptamine for 10 min. After activation, platelets were pelleted by centrifugation and total, released, and cell-associated radioactivity were measured by scintillation counting. Released thrombospondin was measured by double sandwich enzyme linked immunosorbent assay (Voller et al., 1976).

Sucrose Density Gradient Fractionation
Platelets were collected from fresh platelet-rich plasma, washed three times in Tris-citrate buffer (63 mM Tris, pH 6.5, 95 mM NaCl, 5 mM KCl, 12 mM citric acid), and sonicated five times for 10 s at 4 "C (Micro-ultrasonic cell disrupter, Kontes; power setting 8). After brief centrifugation (500 x g, 5 min) the supernatant was layered over a gradient of 30-60% sucrose and centrifuged for 90 min, 217,000 x g, 4 "C (Broekman, 1990). In one experiment, platelets were loaded with ["Clhydroxtryptamine prior to sonication. Fractions were collected, diluted with 10% sucrose, and centrifuged for 15 min at 40,000 x g, 4 "C. sample buffer was added to the supernatants. The samples were boiled, and 1 ~1 of the lysates was electrophoresed on a gradient (lo-15%) polyacrylamide gel using the Phastgel system (Pharmacia). Three ng of purified human LAMP-l were run as a control. Proteins were transferred to nitrocellulose by capillary action (Smith, 1989), and the filters were incubated for 18 h in a solution of TBS containing 0.05% Tween, and 5% Carnation instant nonfat dry milk. Blots were then incubated in TBS/Tween containing rabbit anti-LAMP-l IgG or non-immune IgG for 1 h at 22 "C. After washing in TBS/Tween, they were incubated with alkaline phosphatase-conjugated goat antirabbit IgG for 20 min. The blots were washed and developed using the 5-bromo-4-chloro-3-indolyl phosphate p-toluidine salt/p-nitro blue tetrazolium chloride color development reagent system (Blake et al., 1984).
tin, and von Willebrand factor. To identify full-length cDNAs, a size-selected human umbilical vein endothelial cell cDNA library in XGTlO was screened with this insert DNA. Several overlapping clones were identified, mapped, and sequenced. The largest was 2.2 kilobases and included the 3' end, identified by a consensus polyadenylation signal (AAUAAA) 19 bp 5' to a poly(A) tail (As). An analysis of the nucleotide and deduced protein sequence revealed identity (98% homology) to another recently cloned protein, human lysosomal-associated membrane protein 1 (Chen et al., 1988. A comparison of the nucleotide sequence we obtained with that of LAMP-1 revealed a single difference in the coding region (nucleotide 367), but the amino acid is conserved. In the untranslated 3' region, there were a number of differences, including a 25-base pair insertion at nucleotide 1850, and the presence of a poly(A) tail. LAMP-l was reported to be an integral membrane protein of the lysosome (Chen et al., 1988). Nucleotide sequence predicted that LAMP-l is transcribed as a 416-amino acid protein. There is a 27-amino acid sequence consistent with a classic signal peptide at the amino-terminal. LAMP-l consists of two homologous domains, each containing 4 evenly spaced (36-38 residues) cysteines, separated by a Pro-Ser-rich IgA-like hinge region. There is a 24-amino acid 5' hydrophobic region, consistent with a transmembrane domain and a short ll-amino acid cytoplasmic tail. The predicted extracytoplasmic (intralysosomal) region contains 18 consensus sites for N-linked glycosylation.
The 3'-untranslated region is remarkable for a series of repeats that are extremely guanosine-rich.

Identification
and Analysis of the Activation-dependent Clone-Approximately 2 x lo6 XGTll recombinants from a human umbilical vein endothelial cell cDNA library were screened with a broadly reactive polyclonal anti-platelet antibody. This library was chosen for screening because platelets have little mRNA and because endothelial cells have many proteins in common with platelets. The initial screen yielded 40 positives. At the end of the tertiary screen, there were five highly positive clones: three cross-hybridized and were pursued further. The other two have not been completely characterized.
Flow cytometric analysis of platelets using IgGs affinity purified on immobilized recombinant fusion proteins identified a cDNA encoding an antigen expressed preferentially on thrombin-activated platelets.
As shown in the flow cytographs in Fig. 1, while the screening IgG reacted with both resting and activated platelets (panel A), the affinity purified IgG bound only to activated platelets (panel B). Control (nonimmune) rabbit IgG did not react with fusion protein or with platelet surfaces. Thus, this cDNA encodes a peptide epitope unique to the surface of activated platelets.
The cDNA was estimated to be approximately 400 bp based on agarose gel analysis and by the size of the fl-galactosidase fusion protein, determined by Western blot (data not shown). Nucleotide sequence analysis showed it to be different from other platelet activation-specific antigens, including PADGEM (GMP140), GPIIb/IIIa, fibrinogen, TSP, fibronec-Northern blot analysis of HUVEC total RNA, after stringent washing conditions, showed a single species of message of approximately 3 kilobases (data not shown), consistent with previous findings . Immunoblot analysis of platelet and endothelial cell lysates (Fig. 2) showed that LAMP-l in these cells, similar to purified human hepatic LAMP-l, migrated as a broad band of about 110 kDa. Some heterogeneity in the size of the protein from these different cell types was observed and is probably a result of differential glycosylation.
There was no reactivity with control antibody at this molecular weight (not shown Gel-filtered human platelets were treated with thrombin (0.5 unit/ml) or buffer and then incubated with either the rabbit antiplatelet IgG used to screen the cDNA library (panel A) or an IgG fraction affinity purified on the p-galactosidase fusion protein (panel B). After washing and fixation, the platelets were incubated with FITC-conjugated goat antirabbit IgG and bound fluorescence analyzed by flow cytometry. Data is expressed as cell number (y axis) as a function of fluorescence on a log scale. The screening IgG (A) bound to the surface of both resting and activated platelets, while the affinity purified antibody (B) bound only to the surface of activated platelets.

LAMP-I-Murine
monoclonal antibody to human LAMP-l was used to analyze platelet LAMP-l localization and expression. Flow cytometry studies demonstrated binding to the surface of activated but not resting platelets (Fig. 3A), confirming the results with fusion protein affinity purified polyclonal IgG. Immunofluorescent microscopy of intact and saponin-permeabilized resting platelets showed no binding to intact platelets, but intracellular granular fluorescence of permeabilized platelets (Fig. 3B). These data suggest that LAMP-l is translocated from an intracellular granular compartment to the surface upon stimulation.
To quantitate LAMP-l expression, binding of ""I-conjugated monoclonal anti-LAMP-l IgG to activated and resting platelets was measured. Binding was time-and concentration- Purified human LAMP-l was run as a control in lane 3. After transfer and blocking, the blot was incubated with rabbit anti-LAMP-1 IgG, followed by alkaline phosphatase-conjugated goat anti-rabbit IgG. There is a broad band of specific reactivity in the 110-120 kDa range, although heterogeneity in the different cell types is apparent. dependent, and as seen in a representative equilibrium binding isotherm (Fig. 4), reached saturation between 1.5 and 2.0 pg/ml. In this study, nonspecific binding at maximum input antibody concentration was 0.9% (901 cpm). Scatchard analysis (inset) of these data revealed 2190 binding sites/activated platelet. The number of LAMP-l molecules expressed on activated platelets varied among individual donors as well as from day-to-day with the same donor. In six separate studies we found a mean + SD. of 1175 f 590 (range 525-2190) binding sites/platelet. Resting platelets expressed little or no LAMP-l on their surfaces. As shown in Fig. 5, when compared to platelets maximally stimulated by thrombin, an average of 5.4% of binding was seen on resting platelets (range O-90 molecules/platelet, n = 9). To correlate LAMP-l surface expression with release of specific platelet granules, expression was compared with specific markers of (Y-, dense, and lysosomal granules as a function of agonist (thrombin) concentration (Fig. 6). LAMP-l expression was half-maximal at 0.05 unit/ml thrombin. This clearly differed from both LY-and dense granule release. Surface expression of PADGEM (GMP140), a marker for (Ygranule release, was half-maximal at a lo-fold lower thrombin concentration (0.005 unit/ml), while the half-maximal release of serotonin (["'CIHT), a marker of dense granules occurred at a 5-fold higher thrombin concentration (0.25 unit/ml). In these studies, maximum total binding of '?-anti-PADGEM was l,lOOcpm/aliquot and of ""I-anti-LAMP-1 was 11,800 cpm/aliquot. Nonspecific binding was 14 and 4% of the total, respectively. Total ['4C]hydroxytryptamine uptake was 12,100 dpm, nonspecific release was lo%, while maximum specific release was 87%. These data suggest that LAMP-l translocated from a different compartment than (Y-or dense granules.
Platelet lysosomal acid hydrolase release has been studied FIG. 3. Immunofluorescence demonstration of platelet LAMP-l expression.
A, flow cytometry. Gel-filtered platelets prepared as in Fig. 1  by several investigators and found to differ from 01-and dense granule release in several respects in addition to thrombin sensitivity. Unlike (Y-and dense granule secretion, lysosome enzyme release did not follow stimulation by the weak agonists ADP or epinephrine (Kaplan et al., 1979). As shown in Fig. 5, we found that only 10% of maximal thrombin-induced LAMP-l expression was detected in response to platelet stimulation by ADP or epinephrine. These platelets secreted thrombospondin normally (measured by ELBA) and expressed surface thrombospondin (measured by flow cytometry) indicating that both agonists induced typical activation (data now shown). Also, consistent with published observations of lysosomal enzyme release, we found that LAMP-l surface expression was partially dependent on the presence of extracellular calcium (Fig. 5) and was insensitive to cyclooxygenase inhibition by indomethacin (not shown). Thrombin concentration in the indomethacin studies was low (lo-20 milliunits/ml).
At these concentrations, a-granule release (measured as TSP secretion) was inhibited by 90%, while in parallel tubes (m). Bound irnI or released 14C at 1 unit/ ml of thrombin was considered to be lOO%, and the data were expressed as a percent of this value.
LAMP-l surface expression was not inhibited. These data suggest that LAMP-l translocated from the lysosomal membrane, not 01-or dense granule membrane.
A series of sucrose density gradient platelet fractionation studies were done to investigate further the localization of LAMP-l.
Platelets were sonicated in such a way as to only disrupt their outer plasma membrane and centrifuged through a 30-60% continuous sucrose gradient. A typical banding pattern was observed. Two narrow, low density "membrane" bands (fractions 2 and 4) were clearly distinguished from an intermediate density band (fraction 6) and two high density bands (fractions 8 and 10). These bands were collected along with the intervening regions (fractions 1, 3, 5, 7, 9, ll-13), and subjected to protein and ELBA analyses. In one experiment, platelets were loaded with ['"Cl hydroxytryptamine and radioactivity in the fractions detected by scintillation counting. Fig. 7, panel A, demonstrates colocalization of P-galactosidase, a lysosomal enzyme, with LAMP-l, in fraction 6. The M indicates the peak fraction in which GPlb, a marker of the external platelet plasma membrane, was found. Panel B shows the distribution of ["Clhydroxytryptamine, a dense granule marker. Greater than 47% of the total counts recovered were found in one high density fraction, fraction 10. Panel C shows the localization of the a-granule marker, TSP. Greater than 90% of the total TSP was found in fractions 9 and 10. This biochemical pattern of 01-and dense granule contents is similar to that reported by others (Van Nostrand et al., 1990). From these data, colocalization of LAMP-l with a lysosomal enzyme and clear differentiation from the two other granule pools, we conclude that LAMP-l is lysosomal in platelets. We have also studied LAMP-l expression on platelets from a patient with Hermansky-Pudlak syndrome, an inherited platelet storage pool disease. Platelets from these patients have been described as having diminished or absent ability to secrete dense granule and lysosomal contents (Rendu et al., 1987). As shown in Fig. 8 shown). These data further support the conclusion that LAMP-l is a platelet lysosomal membrane protein.

DISCUSSION
Using a novel molecular approach, we have identified and cloned a platelet protein expressed on the cell surface only after activation.
This 110-120 kDa protein is identical to a previously identified lysosomal membrane protein, LAMP-l. Our data suggest that in platelets this protein is also lysosomal and becomes incorporated into the external plasma membrane during secretion. These data include immunofluorescence studies localizing the antigen to an interior granular compartment in permeabilized, unstimulated platelets and several sets of studies correlating surface expression of LAMP-l best with lysosomal secretion, not with secretion of either of the other two platelet granular compartments (01-and dense). Previous studies (Kaplan et al., 1979) have shown that platelet lysosomal acid hydrolase release was clearly discernible from (Y-and dense granule release by virtue of 1) different dose-response to thrombin stimulation, 2) insensitivity to stimulation by the weak agonists ADP and epinephrine, 3) partial sensitivity to the presence of extracellular calcium, and 4) insensitivity to cycle-oxygenase inhibition when stimulated by low concentrations of thrombin.
Our studies have shown that, similar to platelet lysosomal acid hydrolase secretion, LAMP-l surface expression was maximal at thrombin concentrations intermediate between those necessary to obtain maximal release of a-or dense granules (Fig. 4) minimally induced by ADP or epinephrine (Fig. 5). In addition, we have shown colocalization of LAMP-l with a lysosomal enzyme, P-galactosidase (Fig. 7), after fractionation and sucrose density centrifugation.
In this study, there was clear separation of LAMP-l-containing granules from (Y-and dense granules. Finally, we demonstrated impaired expression of LAMP-1 in Hermansky-Pudlak platelets, which are defective in lysosomal and dense granule release, but showed normal expression of the a-granule membrane protein PADGEM (GMP140).
Secretion in response to a stimulus from cellular storage pools is a general characteristic of many cells, including platelets. In his widely accepted membrane flow hypothesis, Palade suggested that during secretion, granule integral membrane proteins are incorporated into the plasma membrane. PADGEM (GMP140), a platelet a-granule integral membrane protein, provided evidence that a membrane fusion event does indeed occur during platelet activation and secretion (Berman et al., 1986). We report here, for the first time, that LAMP-l, LAMP-I on Platelet Surfaces an integral lysosomal membrane protein, also translocates to the plasma membrane during an activation and secretion event, suggesting that lysosomal membrane fusion also occurs during platelet secretion. Nieuwenhuis et al. (1987) have recently characterized a monoclonal anti-platelet antibody (designated 2.28) that reacts with activated platelets and immunolocalizes to lysosomes. The 53-kDa antigen recognized by that antibody has not yet been molecularly characterized but is thought to be a secreted protein, not an integral membrane protein.
LAMP-l is one of a family of lysosomal membrane proteins that have been identified and characterized in several species. These include rat lgp 110 and 120 (Himeno et al., 1989;Howe et al., 1988), chicken LEPlOO (Fambrough et al., 1988), and mouse (Chen et al., 1985) and human LAMP 1 and 2 (Fukuda et al., 1988;. All are type 1 membrane proteins of 380-396 amino acids with short (lo-11 amino acids) cytosolic tails. They are heavily N-glycosylated on the intralysosomic domains, with the glycan coat more than half the total protein weight. The proteins have two homologous intralysosomic domains, each with 4 conserved cysteines that form two 36-38 amino acid loops . These two repeated domains are separated by a 25-30 amino acid segment rich in proline and serine or threonine, homologous to the IgA hinge region . Granger et al. (1989) have grouped these proteins into two categories: 1gpA and B, corresponding to human LAMP-l and -2, respectively. Intraspecies homology within each group is high (e.g. human and mouse LAMP-l share -66% homology). Sequence homologies between the two groups, however, are less (human LAMP-l is -35% homologous to LAMP-2). These differences are felt to have functional importance, since lysosomes contain proteins from both groups. Recently, it has been shown than LAMP-l and LAMP-2 are encoded by genes localized to different chromosomes. LAMP-l was found on chromosome 13q34, and LAMP-2 on Xq24-25. A putative pseudogene for LAMP-l was also identified on chromosome 12~133 (Mattei et al., 1990). These results, in addition to the rather low sequence homology between LAMP-l and LAMP-2, support the hypothesis that these two genes diverged early in evolution, and probably have distinct functions. There has been speculation that the heavy glycosylation protects the polypeptide backbones of these proteins, and thus, the lysosomal membrane from hydrolytic enzymes (Carlsson et al., 1988). While there is no known function for any of these proteins, the amino acid sequence of LAMP-l is homologous to two leukemia cell glycoproteins, GP130/P2B from the highly metastatic tumor cell line MDAY-DX (Chen et al., 1985) and a differentiation marker in hematopoietc cells (Fukuda, 1985). In addition, the N-linked polylactosaminoglycans carried on the lysosomal proteins have been correlated with increased metastatic potential (Dennis et al., 1987). Although a minor fraction (<2%) of LAMP-l is associated with the plasma membrane of most nucleated cells (Lippincott-Schwartz and Fambrough, 1987), presumably as a result of selective exchange of lysosomal and plasma membranes, the present studies are the first to demonstrate significant and regulated cell surface expression of LAMP-l on normal adult cells. As previously noted, increased surface expression of LAMP-l has been observed on transformed cells of high metastatic potential, and interestingly, on embryonic cells (Lippincott-Schwartz and Fambrough, 1986). Activated platelet surfaces share certain functional characteristics with these cells, such as enhanced adhesiveness and protease activity. It is thus possible that the regulated expression of LAMP-l on the surface of activated platelets may play a role in these functions. A deglycosylated form of LAMP has been shown to bind collagen and the fibronectin adhesion peptide RGD (Laferte and Dennis, 1988), further suggesting a possible role in adhesion.
In summary, we report the identification and characterization of a previously unknown platelet activation marker, by affinity purification of a broadly reactive antibody directly from recombinant phage fusion protein. This protein translocates from the lysosomal membrane and has characteristics of an adhesive molecule.
Identification of proteins, like LAMP-l, that are preferentially expressed on activated cells may aid in the design of novel diagnostic and therapeutic approaches to diseases such as atherosclerosis and thrombosis.