Identification of PECAM-1 in Solid Tumor Cells and Its Potential Involvement in Tumor Cell Adhesion to Endothelium*

PECAM- 1 (CD3 l/EndoCAM) is an adhesion molecule in the immunoglobulin supergene family that is ex- pressed on endothelial cells, platelets, and some he-matopoietic lineage cells. In this paper, using several polyclonal and monoclonal antibodies against PECAM-1, we identified PECAM-1 molecules on human, rat, and murine solid tumor cell lines. Immunocytochemical labeling and flow cytometric analysis using either polyclonal, monoclonal, or Fab portion of the antibodies against PECAM-1 detected a distinct distribution on tumor cell surface. Immunoblotting revealed proteins ranging from 120 to 130 kDa in tumor cells derived from different species. Immunoprecipitation and sub- cellular fractionation studies indicated that PECAM- 1 is constitutively expressed on the surface of human tumor cells (Le. colon adenocarcinoma). The specificity of a major polyclonal anti-PECAM-1 used in the current study (Le. SEW-3) was confirmed by the preab- sorption studies. PECAM-1 molecules

PECAM-1 (CD3 l/EndoCAM) is an adhesion molecule in the immunoglobulin supergene family that is expressed on endothelial cells, platelets, and some hematopoietic lineage cells. In this paper, using several polyclonal and monoclonal antibodies against PECAM-1, we identified PECAM-1 molecules on human, rat, and murine solid tumor cell lines. Immunocytochemical labeling and flow cytometric analysis using either polyclonal, monoclonal, or Fab portion of the antibodies against PECAM-1 detected a distinct distribution on tumor cell surface. Immunoblotting revealed proteins ranging from 120 to 130 kDa in tumor cells derived from different species. Immunoprecipitation and subcellular fractionation studies indicated that PECAM-1 is constitutively expressed on the surface of human tumor cells (Le. colon adenocarcinoma). The specificity of a major polyclonal anti-PECAM-1 used in the current study (Le.  was confirmed by the preabsorption studies. PECAM-1 molecules on tumor cells appear to bear terminal carbohydrate moieties (i.e. sialic acid residues) different from those on platelets, since neuraminidase treatment of tumor cells, unlike platelets, did not result in a mobility shift. Polymerase chain reaction (PCR) analysis of genomic DNA derived from tumor cell lines of different species revealed the presence of PECAM-1 gene in the genome. The mRNAs of PECAM-1 in tumor cells were detected by reverse transcription-PCR followed by Southern hybridization. Screening of more than 20 human, rat, and murine solid tumor cell lines indicated that PECAM-1 is widely expressed, although the level of expression varies considerably among different cell lines. The expression of PECAM-1 message in tumor cells was confirmed by Northern blotting. DNA sequencing of the PCR fragment revealed that human tumor cell PE-CAM-1 matches 100% to the human endothelial cell counterpart. Finally, it was demonstrated that tumor cell PECAM-1 is involved in mediating tumor cell adhesion to endothelium, as evidenced by the ability of anti-PECAM-1 antibodies to decrease the adhesion of unstimulated tumor cells to microvascular endothelial cells. , CA 47115 (to K. V. H.), and CA 29997 (to K. * This work was supported by National Institutes of Health Grants V. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$4 TO whom correspondence should be addressed Dept. of Radiation Oncology, Wayne State University, 431 Chemistry, Detroit, MI 48202. Tel.: 313-577-1018;Fax: 313-577-0798. Tumor cell interactions with platelets, endothelial cells, and subendothelial matrix are considered essential intermediate steps for the completion of "metastatic cascade" (Weiss et al., 1988;Honn et al., 1992;Liotta et al., 1986). Tumor cells employ cell surface glycoprotein receptors to achieve these cell-cell and cell-matrix interactions. Organ specificity of tumor metastasis suggests that distinct tumor cell types utilize different repertoires of adhesion receptors (Nicholson, 1988;Pauli et al., 1990;Honn and Tang, 1992). Cell-cell adhesion is a complicated biological process in which various mechanisms and factors are involved. In addition to integrins which are heterodimeric membrane glycoproteins primarily involved in cell-substrate adhesion, three other major families of proteins are implicated in mediating cell-cell interactions. They are immunoglobulin (Ig)' superfamily (William and Barclay, 1988), cadherins (Takeichi, 1990), and selectins. Down-regulated cadherins (e.g. E-cadherin) have been implicated in the loss of cell-cell contact and correlated with the invasiveness and metastatic potential of tumor cells (Schipper et al., 1991). Selectins expressed on vascular endothelial cells have been shown to mediate tumor cell adhesion by recognizing carbohydrate ligands (SLe" or SLe" antigen) on the tumor cell surface Majuri et at., 1992).
The Ig supergene family are cell surface adhesion molecules that possess immunoglobulin-like folds in the extracellular domain. PECAM-1 (for platelet endothelial cell adhesion molecule-1; also called CD31 and EndoCAM) is a newly characterized adhesion molecule that belongs in the Ig superfamily and is expressed on platelets, granulocytes, monocytes, lymphocytes, macrophages, and endothelial cells, as well as certain tumor cell lines such as lymphoma and leukemic cells (Muller et al., 1989;Knapp et al., 1989;Newman et al., 1990;Stockinger et al., 1990;Simmons et al., 1990;Albelda et al., 1991;Zehnder et al., 1992). Various experiments have demonstrated that PECAM-1 is a 120-130-kDa transmembrane glycoprotein that carries a significant amount (about 40% of the molecular mass) of carbohydrate moieties (Newman et al., 1990). Molecular cloning and nucleotide sequence of human endothelial cell PECAM-1 have revealed an open reading frame of 2,114 bp that encodes a 738-amino acid protein with six extracellular C-2 type Ig-like domains (Newman et al., 1990;Albelda et aL, 1991), although some minor differences in the molecular structure have been noticed by other groups (Stockinger et al., 1990;Simmons et al., 1990;Zehnder et al., 1992). Functional studies have indicated that PECAM-1 plays an important role in establishing a contiguous endothelial cell monolayer (Albelda et al., 1990). Transfection of a full-length PECAM-1 cDNA to cells (i.e. the L cells) that do not endogenously express the molecule induced a Ca2+-dependent homotypic cell-cell aggregation and heterotypic cell-cell adhesion (Albelda et al., 1991), suggesting that PECAM-1 may be involved in a variety of intercellular adhesion processes. Based on the realization that cell-cell adhesion is an essential intermediate step in tumorigenesis and cancer metastasis, that most Ig family members are involved in tumor cell-host cell interactions, and that PECAM-1 is functional in mediating cell-cell adhesion, we hypothesized that some tumor cells may express PECAM-1 that is involved in tumor cell-tumor cell, as well as tumor cell-platelet-endothelial cell interactions. In the current paper we present biological, biochemical, and molecular evidence that PECAM-1 is expressed in cultured human and rodent solid tumor cells. Furthermore, we will show that PECAM-1 functions in supporting nonstimulated tumor cell adhesion to vascular endothelium in vitro.

MATERIALS AND METHODS
Antibodies-Five antibodies against PECAM-1 were used in the present studies. Polyclonal anti-PECAM SEW-3 and SEW-16 (IgG) were generated in rabbit using affinity-purified human platelet PE-CAM-1 as the immunogen. SEW-3 was derived from the same batch of preparation as described previously (Albelda et al., 1991). SEW-16 was raised by once-weekly subdermal injections of 100-pg doses of immunoaffinity-purified PECAM-1 antigen (Newman et al., 1992). The Fab fragment of anti-PECAM-1 was derived from cleaving pAb SEW-3 with papain according to the product manual (Pierce Chemical Co.). Monoclonal anti-PECAM-1, mAb 1.3, (IgG1) was produced by immunizing mice with purified human platelet PECAM-1 protein (Albelda et al., 1991;Newman et al., 1992). Another mAb against PECAM-1, BBA-7, was purified by affinity chromatography on protein A-Sepharose using human umbilical vein endothelial cells as the immunogen and shown to be specific for PECAM-l(R & D Systems, Minneapolis, MN).
Rabbit nonimmunized IgG or nonimmune rabbit serum (Cooper Biochemical, Malvern, PA) and mineral oil-elicited mouse ascites produced from MOPC tumor cell line (IgG1, K chain, Sigma) were used as negative (antibody) controls in immunofluorescence studies, adhesion studies, and immunoblotting. Goat whole serum (Sigma) was used as the Fc receptor blocking agent. The secondary antibodies used in the experiment were fluorescein isothiocyanate-conjugated goat anti-mouse or anti-rabbit IgG (ICN Immunologicals, Lisle, IL).
Cell Culture-Mouse microvascular endothelial cells, CD3, were isolated and characterized as described previously (Chopra et al., 1990). Large vessel endothelial cells, RAEC, were derived from rat (Sprague-Dawley) aortic rings (Diglio et aL, 1989). These endothelial cells were routinely maintained in Dulbecco's minimal essential medium supplemented with 10% fetal bovine serum (FBS, Life Technologies, Inc.) and various antibiotics (50 pg/ml gentamycin, 100 pg/ ml penicillin G, and 2.5 wg/ml amphotericin B). Cells were cultured in a humified atmosphere with 5% COS, and the culture media were changed every 48 h. Endothelial cells were passaged with a mixture of EDTA (0.1%) and trypsin (0.05%). All cells used in this experiment were free of micoplasma infection.
B16 amelanotic melanoma cell line (B16a), rat W256 carcinosarcoma (W256) cell line, and Lewis lung carcinoma (3LL) cell line were obtained from the Division of Cancer Treatment, National Institutes of Health (Frederick, MD) and adapted for cell culture as described previously (Chopra et al., 1988(Chopra et al., , 1990Grossi et al., 1988Grossi et al., , 1989Tang et al., 1993b). B16a and 3LL cells were passaged with 2 mM EDTA in syngeneic (C57BL/&J) male mice and cultured in either MEM (Life Technologies, Inc.) supplemented with 5% FBS (for B16a cells), or Dulbecco's modified Eagle's medium supplemented with 10% FBS (for 3LL cells) and antibiotics (see above). 3LL cells were cultured in a humidified atmosphere with 5% COS. W256 cells were grown in MEM supplemented with 5% FBS and antibiotics and passaged with 2 mM EDTA.
HEL (human erythroleukemia), clone A and DLD-l(human colon carcinoma; Grossi et al., 1988;Tibbetts et al., 1977), MS751 (human cervical epidermoid carcinoma; metastasis to lymph node), TCCSUP (human primary bladder transitional-cell carcinoma, grade IV; Nayak et al., 1977), ACHN (human renal carcinoma; originally derived from the malignant pleural effusion of a patient with widely metastatic renal adenocarcinoma), SK-HEP-1 (human liver carcinoma; Fogh et al., 1977), and SW900 (human lung squamous carcinoma; Fogh et al., 1977) cells were obtained from American Type Cell Culture Collection. A series of human melanoma cell lines, WM35, WM115, WM164, WM226-4, WM793, WM983-A, and WM983-B were kindly provided by Dr. M. Herlyn (The Wistar Institute of Anatomy and Biology). These cell lines have not been extensively characterized in the literature. Human prostate adenocarcinoma Du145 (Stone et al., 1978) and PPC-1 (Brothman et al., 1989), human head and neck squamous carcinoma (SSC-UM), human breast carcinoma (MCF-7; Soule et al., 1973), rat prostate adenocarcinoma (AT-3), and B16F1 and B16F10 murine melanoma cell lines were kindly provided by Drs. Institute, TX), respectively. HEL, MCF-7, and AT-3 cells were cultured in RPMI medium plus 10% FBS. SW900 cells were grown in L-15 medium with 10% FBS and human melanoma cells of the WM series were cultured in MCDB/L-15 (4:l) supplemented with 2% FBS and 5 pg/ml of insulin. All of the remaining tumor cell lines were cultured in either MEM or Dulbecco's modified Eagle's medium containing 10% FBS and passaged with 2 mM EDTA. A summary of the cell lines used in the present study is presented in Table I. Chemic& and Reagents-Protease inhibitors PMSF, leupeptin, antipain, aprotinin, chymostatin, and protein standard markers were obtained from Sigma. Immunoblotting detection kit (ECL system) was bought from Amersham Corp. Peroxidase-anti-peroxidase staining kit was purchased from Biogenex (San Ramon, CAI. Protein kinase C activator TPA and eicosanoid 12(S)-HETE (i.e. 12[S]hydroxyeicosatetraenoic acid) were purchased from Sigma and Cayman Chemical (Ann Arbor, MI), respectively. The RNA ladder (0.24-9.5 kb) and prestained protein standard SDS-7B (26.5-180 kDa) were obtained from Life Technologies, Inc.) and Sigma, respectively.
Indirect Immunofluorescence-Cultured B16a, W256, 3LL, clone A, and B16F10 cells were dissociated from the tissue culture flasks with 2 mM EDTA and washed once with MEM and then fixed with 2% paraformaldehyde in PBS containing 1 mM CaC12, 1 mM MgClt, and 5% sucrose for 20 min at room temperature. CD3 endothelial cells were used as positive control cell line. Immunofluorescent labeling was performed essentially as described previously (Tang et al., 1993a and 1993b). Briefly, for intracellular labeling, cells were permeabilized with HEPES-Triton buffer (20 mM HEPES, pH 7.6, 300 mM sucrose, 50 mM NaC1, 3 mM CaC12, and 0.5% Triton x-100) for 3 min at room temperature. For surface labeling, cells were not permeabilized. All of the coverslips were incubated with 20% goat whole serum in 4% BSA-containing PBS for 20 min at 37 "C to block nonspecific Fc-binding sites. The primary antibody reaction was performed by incubating coverslips with polyclonal (SEW-3; 30 pg/ ml), monoclonal (mAb 1.3; 8 pglml), or Fab fragment (30 pg/ml) of anti-PECAM-1, or equivalent antibody controls for 60 min at 37 "C, followed by washing (4 X, PBS). Afterward, the cells were labeled with goat anti-rabbit or goat anti-mouse IgG-fluorescein isothiocyanate (1:200), depending on the primary antibodies used. Coverslips were mounted with glycerol and PBS (9:l) containing 0.1% Npropylgallat. Phase contrast and immunofluorescence pictures were taken with a Nikon Optiphot microscope. Transmission light micrographs were taken with a Leitz Orthoplan microscope.
Immunocytochemistry-Subconfluent endothelial cells and tumor cells were cultured for 18 h before being used for peroxidase-antiperoxidase staining. Cells were fixed and permeabilized as described for immunofluorescence. After washing, the coverslips were incubated with 3% H,02 at room temperature for 5 min to eliminate endogenous peroxidase activity. Following primary antibody (SEW-3 or mAb 1.3) incubation, coverslips were sequentially incubated with anti-mouse or anti-rabbit IgG (corresponding to the primary antibodies used) and the peroxidase-anti-peroxidase complex for 60 min at 37 "C each. The staining results were revealed by incubating cells with chromagen (AEC) for 15 min at 37 ' C .
Platelet Preparation-Human and mouse blood were collected with 3.8% sodium citrate and 4.5% dextrose in 0.9% physiological saline. Platelet-rich plasma was obtained by centrifuging the collected blood at 600 X g for 15 min. This procedure was repeated twice, and the platelet-rich plasma was combined. An appropriate amount of platelet wash buffer (1.6% 0.1 M EDTA in platelet wash) was added to the pooled platelet-rich plasma to prevent platelet aggregation and release reaction, and the above mixture was centrifuged at 2,000 X g for 15 min to obtain platelets. After washing, the platelet pellet was extracted as described below.
Cell Extraction and Subcellular Fractionation-Endothelial cells (i.e. CD3 and RAEC) and tumor cells (B16a, W256, 3LL, clone A, DLD-1, HEL, and B16F10) were first washed free of media with PBS containing 5 mM of PMSF and 1% aprotinin and were then scraped off the culture flasks. The cell pellets as well as platelet pellet were lysed and extracted with the TNC lysis buffer (0.01 M Tris-acetate, pH 8.0, 0.5% Nonidet P-40, and 0.5 mM Ca2') containing a mixture of protease inhibitors (5 mM PMSF, 1 mM leupeptin, 1% aprotinin, 1 pg/ml of pepstatin and chymostatin) on ice for 45 min. The whole cell lysates were centrifuged at 14,000 X g for 30 min and the supernatants aliquoted and frozen at -70 "C until use. To prepare the membrane fraction (i.e. subcellular fractionation), platelets, clone A, or DLD-1 cells (either dissociated with EDTA or directly scraped off using a rubber policeman) were lysed in ice-cold hypotonic buffer (1 mM NaHC03, 5 mM MgCl,, 50 mM Tris-HC1, pH 7.5, 0.5 mM EGTA, 1 mM PMSF, 0.2 mM leupeptin, 1 p~ aprotinin, and 0.5 p~ pepstatin A). The cell lysates were centrifuged at 500 X g for 5 min to remove nuclei and unbroken cells. The supernatant was further centrifuged at 100,000 X g for 90 min at 4 "C. The resulting pellet, i.e. the membrane fraction (Liu et al., 1991), was washed once with the hypotonic buffer, lysed in the TNC lysis buffer, and measured for protein concentration (Bradford, 1976). In some experiments, clone A cells and DLD-1 cells treated with protein kinase C activators TPA (0.1 FM, 15 min) or 12(S)-HETE (0.1 p~, 15 min; Tang et al., 1993aTang et al., , 1993bLiu et al., 1991;Grossi et al., 1989) were used for membrane preparation as described above.
Zmmunoblotting, Immunoprecipitation, and SDS-PAGE-The procedures for Western blotting, immunoprecipitation, and gel running were detailed previously (Tang et al., 1993a(Tang et al., , 1993b. Briefly, protein samples (either whole cell lysates or membrane fractions) were dissolved into the sample buffer (0.1 M Tris, pH 6.8, 2% SDS, and 40% glycerol) in the presence or absence of 10% 2-mercaptoethanol. In some experiments the whole cell lysates from human platelet, BNa, 3LL cells or clone A cells were treated with 1.0 unit/ml of neuraminidase from Clostridium perfringem (Sigma) at pH 5.0 for 30 min, 1, 2, or 4 h. The treatment was terminated by dissolving samples in the sample buffer. In other experiments aimed at testing the specificity of pAb SEW-3 to PECAM-1, this Ab was preabsorbed with 0, 10,50, or 100 fig of purified platelet membrane before being used for Western blotting. Samples were boiled for 5 min and analyzed with 7.5% denaturing polyacrylamide gels. Gels were stained with either Coomassie or silver nitrate, or transferred to nitrocellulose membrane and proteins were detected using the ECL (Enhanced Chemiluminescence) Western blotting detection system (Tang et al., 1993a and1993b). The primary antibodies used were either pAbs (i.e. SEW-3 and SEW-16; 40 pg/ml) or mAb BBA-7 (20 pg/ml). The secondary antibody used was either goat anti-rabbit or anti-mouse IgG coupled to horseradish peroxidase. Immunodetection was performed basically according to the product directions (Amersham). Autoradiography was conducted with Hyperfilm-ECL (Amersham). For immunoprecpitation, HEL cells or clone A cells (dissociated with 2 mM EDTA) were surface iodinated as described previously (Tang et al., 1993b). The antibodies used were either SEW-3 (10 gg/ml) or mAb 1.3 (5 pg/ ml). Immunoprecipitates were dissolved in the sample buffer and separated on 7.5% SDS-PAGE under reducing conditions. Gels were stained, dried, and exposed at -80 "C using an intensifying screen.
Preabsorption Studies-To confirm the specificity of SEW-3, the major polyclonal antibody used throughout the current study, we preabsorbed this antiserum with PECAM-1 and then employed the preabsorbed antibody in the immunoblotting as well as immunostaining of tumor cells. L cells that do not express endogenous PECAM-1 were transfected with complete PECAM-1 cDNA sequence or the vector alone (Albelda et al., 1991). One mg of the SEW-3 IgG was incubated with 5.6 X lo7 of L cells transfected with either PECAM-1 or vector for 1 h at room temperature with mixing. The cells were then removed by centrifugation. This absorption step was repeated two more times. The final supernatant was filtered through a 0.2-pm filter and spun at 100,000 X g at 4 "C for 1 h. The antibody was then tested by ELISA to determine the activity remaining using purified PECAM-1 plated in the microtiter wells. Dilutions of l:lO, 1:50, 1:250, and 1:1250 were used for the PECAM-1-absorbed antibody and the control-absorbed SEW-3 and normal rabbit IgG were used at 100, 20, 4, and 0.8 pg/ml in the ELISA experiments. The absorbed and control-absorbed SEW-3 IgG were then used in immunofluorescent labeling and immunoblotting of tumor cells, as described above.
Tumor Cell Adhesion to EndotheCurn-In uitro cell adhesion assay was run to determine the potential functions of tumor cell PECAM-1 molecules. B16a or 3LL cells metabolically labeled with 0.1 mCi/ ml of [32P]orthophosphate (37 "C for 5 h in P-free MEM) were dissociated (with 2 mM EDTA) and washed twice with MEM. Then tumor cell adhesion to confluent CD3 cells in 24-well culture plates (Falcon) was performed according to the following three protocols: (a) tumor cells were first incubated with 40 pg/ml of polyclonal anti-PECAM-I in MEM (containing 4% BSA) for 30 min at 15 "C and then added (100,000 cells/well) to CD3 monolayer; (b) CD3 monolayer was first treated with Ab (the same amount as in a) for 30 min at 15 "C and then untreated tumor cells were added and (c) tumor cells were suspended in the Ab solution and immediately added onto EC monolayer. Time course was run as described in c) for 10, 30, and 60 min following addition of tumor cells. Dose studies were performed using different concentrations of polyclonal anti-PECAM-1 (40, 20, and 10 pg/ml) or an equivalent amount of nonimmune rabbit whole serum. In some experiments, tumor cell adhesion to CD3 monolayers was performed by preincubating either tumor cells or endothelial cells with anti-PECAM-1 followed by washing to remove residual antibodies. The adhesion was terminated by aspirating media and nonadherred cells. The culture wells were rinsed with PBS (4 x) and the contents harvested with a mixture of 0.1 N NaOH and 1% SDS. The number of adherent tumor cells was determined by counting the radioactivity. Triplicates for each experimental condition were performed, and the experiment was repeated three times with comparable results.
Polymerase Chain Reaction (PCR) Analysis of Genomic DNA and Reverse Transcription-Polymerase Chain Reaction (RT-PCR)-Genomic DNA of cultured solid tumor cells was isolated using SDS/ proteinase K method (Sambrook et al., 1987). PCR of genomic DNA was performed using 0.25 pg of DNA as the template in 50 pl of the following reaction buffer: 10 X PCR buffer (200 mM Tris-HC1, pH 8.3, 750 mM KCl, 1 mg/ml BSA, 25 mM MgC12, 0.25 pg/pl primers, and 0.1 unit/pl Taq polymerase). PCR reaction without the template was used as the experimental control (i.e. negative control). To prevent carry-over (i.e. the product) contamination, 0.5 unit/pl of restriction enzyme AuaII, which cuts three times in the PCR fragment specified by the nested pair of primers (see below), was used to treat the reaction buffer (37 "C, 60 min) before adding template and the enzyme. The cycling conditions were the same as used for RT-PCR (see below). Total RNA was extracted from whole cell lysates using guanidium thiocyanate-CsC1 method (Chang et al., 1992;Tang et al., 1993b). RT-PCR was performed basically as described previously Tang et al., 1993a and1993b). Briefly, 1 pg of total RNA was reverse transcribed in a 20-p1 transcription buffer made up of 50 mM Tris, pH 8.3,75 mM KC1,3 mM MgC12, 10 mM dithiothreitol, 0.5 mM dNTPs, 20 units of RNase inhibitor (RNasin; Promega), 1 p~ of antisense (see primer B described below), and 200 units of M-MLV reverse transcriptase (Life Technologies, Inc.). RT reactions without RNA template was used as the negative control. To prevent DNA contamination, 1 unitlpl of DNase was used to pretreat the mixture prior to the addition of primers and reverse transcriptase (37 "C, 60 rnin). The mixture was heated at 95 "C for 10 min (to deactivate DNase). Then primers and reverse transcriptase were added, and the mixture was incubated at 42 "C for 1 h, followed by heating at 100 "C for 10 min and immediate cooling on ice. In another set of experiments, the complete RT mixture was treated with 0.5 unit/pl RNase A and then used for PCR in order to confirm the absence of DNA contamination. Two pl of the above cDNA was used as the template in PCR. Nested PCR was performed to detect PECAM-1 message. Since the genomic sequence of PECAM-1 is unknown, we designed two pairs of primers, presented as follows (also see Fig. 11), on the basis of published human endothelial cell PECAM-1 cDNA sequence (Albelda et al., 1991): primer A, 5'-CAA AGA CAA CCC CAC TGA AG-3' (sense); primer B, 5'-CAC TCC GAT GAT AAC CAC TG-3' (antisense); primer C, 5'-CTG AGG GTG AAG GTG ATA GC-3' (nested sense); primer D, 5'-AGT ATT TTG CTT CTG GGG AC-3' (nested antisense).
Primer A and primer B covers the nucleotide sequence region from 1533 to 1975 (443 bp). The nested (or internal) pair of primers (i.e. primer C and D) encompass nucleotides 1615-1910 (296 bp). The whole region amplified includes a segment of the transmembrane domain and a section of the adjacent extracellular Ig domain (Newman et al., 1990). Two pl of the reverse transcription mixture was amplified in a total of 100 p1 of the PCR buffer (20 mM Tris, pH 8.3, 50 mM KCl, 2.5 mM MgCl,, 0.25 mM dNTPs, and 0.1 mg/ml BSA) containing 1 unit of AmpliTaq DNA polymerase (Perkin Elmer Cetus). The first round of PCR was run in the presence of primers A and B using GeneAmp PCR System 9600 (Perkin Elmer Cetus) at 94 "C X 1 min, 51 "C x 1 min, and 72 "C X 2 min for 30 cycles. TWO pl of the first round PCR product was used for the second round of PCR using primers C and D at 94 "C X 30 s, 50 "C X 30 s, and 72 "C X 1 min for 30 cycles. All PCR buffers used in RT-PCR were treated with AvaII (37 "C, 60 min) to prevent carry-over contamination. PCR reaction without cDNA template was used as the negative control.
Southern Hybridization and Northern Blotting-Twenty pl of PCR-amplified product was separated on a 1.5% agarose gel and transferred to GeneScreen plus membrane (Du Pont-New England Nuclear) using a PosiBlot Pressure Blotter (Stratagene, La Jolla, CA). DNA was then UV-linked to the membrane. Radioactive cDNA probe was prepared with [3ZP]dCTP using Prime-it random primer labeling kit (Stratagene) and a 1.8-kb cDNA insert encoding PECAM-1 (this fragment was originally cloned in the plasmid vector ptZl8R and released by EcoR I and HindIII). The membrane was prehybridized in the solution containing 50% formamide, 5 X Denhardt's, 10% dextran sulfate, 5 X SSPE, 0.1% SDS, and 100 pg/ml salmon sperm DNA (42 'C, 4 h). Hybridization was performed with the addition of labeled PECAM-1 probe (42 "C, overnight). The membrane was washed under high stringency conditions (0.1 X SSC and 0.1% SDS for 15 min at room temperature and several washes at 65 "C) and exposed to Kodak XOMAT x-ray film at -80 "C using an enhancing screen. For Northern blotting, total cellular RNA was obtained from cultured HEL, 3LL, B16F1, B16F10, clone A, and W256 cells as described above for RT-PCR. Poly(A+)-RNA was isolated using PolyATtract mRNA Isolation Systems (Promega) and 4 pg of denatured (glyoxal/dimethyl sulfoxide, 50 "C, 1 h) mRNA was loaded onto a 1.0% agarose gel. A 0.24-9.5-kb RNA ladder was used as the bridized and hybridized as described above. After hybridization, the molecular mass standard. Gel was transferred and membrane prehy-cDNA probe. membrane was deprobed and then reprobed with radiolabeled p-actin DNA Sequencing-The PCR amplified fragment migrating at the predicted size (as defined by the primers) on agarose gels was cut from the gel and purified by electroelution. The sequence of the amplified double strand cDNA fragment was determined by the Sanger dideoxynucleotide termination method using the AmpliTaq sequencing kit (Perkin Elmer Cetus) with some modifications . Purified fragment (template) and primers (C or D) were denatured at 100 "C for 5 min and annealed on dry ice for 2 min. Extending DNA chain was labeled with "S-dATP. Samples were loaded onto a 6% polyacrylamide sequencing gel. Autoradiography and data processing were performed as described .

RESULTS
Immunological Identification of PECAM-1 on Human, Rat, and Murine Solid Tumor Cells-Cultured tumor cells from different species were surface labeled with either pAb (Fig.  l), mAb (data not shown), or Fab fragment (Fig. 2) against PECAM-1. Immunodetection was conducted using immunofluorescence, peroxidase-anti-peroxidase staining, and flow cytometry. As shown in Fig. 1, tumor cells, like endothelial cells (Fig. la), expressed PECAM-1 molecules on their cell surface. The distribution pattern of the positive labels varied among different tumor cell lines. B16a (Fig. lb) cells demonstrated a homogeneous surface labeling although heterogeneity existed among individual cells (i.e. some tumor cells expressed a much lower amount of PECAM-1 than others). B16F10 melanoma cells exhibited a similar staining pattern (Fig. le). In contrast, larger aggregates of positive label were detected on 3LL cells (Fig. 1, c and h). On the other hand, clone A cells (human colon carcinoma) appeared to be enriched for PECAM-1 molecules at the cell periphery in subconfluent cultures (Fig. Id) and cell borders in confluent cultures (data not shown). Peroxidase-anti-peroxidase staining revealed brownish granules on the cell surface of B16a cells (Fig. lg) and also in the perinuclear region of 3LL cells (Fig. lh). When cells were permeabilized with Triton-HEPES buffer, immunostaining with SEW-3 detected an intracellular pool of PECAM-1 molecules (data not shown). Staining with the Fab fragment of SEW-3 also detected cell surface labeling on B16a, 3LL, W256, clone A, and B16F10 cells (Fig. 2, a-e, respectively), excluding the possibility of nonspecific binding of intact antibody to Fc receptors. In addition, staining of 3LL cells with control antibodies, i.e. nonimmune rabbit IgG (Fig. 1, f and  Biochemical Identification and Partial Characterization of PECAM-1 on Tumor Cells-Several PAbs and mAbs were used to detect PECAM-1 in solid tumor cell lines. PAb SEW-3 detected a 130-kDa protein, together with multiple lower bands, in human platelet lysates (Fig. 3, A and B). The lower bands represent either the degraded species or nonspecific staining since in other preparations of human platelets only the 130 kDa band protein was detected (e.g. see Fig. 3, C and

cells (data not shown). Rat aortic endothelial cells (RAEC)
revealed a protein band of identical size to that of the W256 cells, i.e. 125 kDa (Fig. 3A). Rb I& only stained some low molecular mass nonspecific protein bands (Fig. 3A). Two bands right below the 128 kDa band in clone A cells are probably the degradation products (see Fig. 30 and Fig. 5, C and D, for comparison). When used to stain murine cells, SEW-3 detected a 130-kDa protein in CD3 microvascular endothelial cells and 3LL tumor cells and a -125-kDa protein in two melanoma cell lines, B16a and B16 F10 (Fig. 3 R ) . 3LL cells also demonstrated two lower molecular mass bands which are probably the degradation products. Again, RbIgC only detected some low molecular mass nonspecific bands (data not shown). The reactivity of SEW-3 to PECAM-1 was. in a dose-dependent manner, blocked by preincubating this Ab with purified platelet membrane (Fig. 31)). thus providing indirect evidence for the specificity of SEW-3. Another pAb, SEW-16, also detected a -128-kDa protein in clone A and DLD-1 cells (Fig 3E). Interestingly, this pAb does not recognize PECAM-1 in rodent tumor cells (data not shown). A mAb, BBA-7, detected similar protein bands in clone A and DLD-1 cells (Fig. 3F).
Human colon adenocarcinoma cells (i.e. clone A and DLD-1) were further studied by immunoprecipitation and subcellular fractionation (Fig. 5). It appears that PECAM-1 is constitutively expressed on the surface of clone A cells, since immunoprecipitation of radioiodinated clone A cells with both SEW-3 (Fig. 5A) and mAb 1.3 (Fig. 5B) resulted in the expected -128 kDa protein band. Immunoblotting using membrane fractions readily detected PECAM-1 in both clone A and DLD-1 cells (Fig. 5C). In addition, it appears that the method of harvesting tumor cells (i.e. scraping uersuS EDTA dissociation) does not affect the detectability of PECAM-1 by SEW-3 (Fig. 5C). Treatment of clone A cells with TPA or 12(S)-HETE did not significantly alter the level of PECAM-1 associated with plasma membrane (Fig. 5 0 ) , although these two agents have been shown to increase the surface expression of integrin receptor avB3 (Tang et al., 1993a and1993b).
PECAM-1 is a heavily glycosylated molecule where about 40% of the molecular mass is composed of carbohydrates. It is primarily N-glycosylated and has been demonstrated to possess terminal sialic acid residues (Newman et ol., 1990). A preliminary characterization of tumor'cell PECAM-1 molecules was performed using neuraminidase treatment. As presented in Fig. 6, 30-min treatment of human platelets with 1.0 unit/ml of neuraminidase resulted in a molecular mass reduction of approximately 5 kDa. In contrast, this decrease in the molecular weight, following neuraminidase treatment, was not observed with all of the tumor cell lines tested (i.e. B16a, W256, and 3LL; Fig. 6), This differential sensitivity of PECAM-1 to C. perfringens neuraminidase in platelets and tumor cells was observed following treatment of samples for up to 4 h (data not shown). Preliminary experiments with clone A cells revealed similar insensitivity of PECAM-1 to neuraminidase treatment (data not shown).
PCR of Genomic DNA, RT-PCR, Southern Hybridization. and Northern Blotting-PCR analysis of genomic DNA revealed the PECAM-1 genes in the genomes of several human, rat, and murine tumor cells (Fig. 7). The size (296 bp) of PECAM-1 fragments (which were confirmed by hybridization) in all cell lines examined was precisely the same as predicted from the size covered by the nested pair of primers which are based on the cDNA sequence, suggesting that this fragment represents a gene encoding segment (i.e. no intron is included in this fragment). All cell lines demonstrated the same size of PECAM-1 fragment (Fig. 7), suggesting the  lanes 3 and 3 ' ) , and W256 cells (lone 4 and 4 ' ) using SEW-3 preincubated with either control L cells (lanes 1-4) or L cells transfected with PECAM -1 (lanes 1'-4'). The detected PECAM-1 proteins migrating at 130 kDa (for human platelet, lone I ) , -128 kDa (for clone A and DLD-1 cells, lanes 2 and 3 ) , and -125 kDa (for W256 cells, lane 4 ) were indicated by three arrowheads. c-e, immunofluorescent labeling of clone A (c and e) and 3LL cells ( d ) with SEW-3 preabsorbed with either control L cells (c and d) or L cells expressing transfected PECAM-1 (e). X 600. genomic composition of PECAM-1 may be similar. PECAM-1 mRNA was examined by RT-PCR using PECAM-specific primers and DNA hybridization using PECAM-1 cDNA probes. The results of these experiments are depicted in Fig.  8. Interspecies and intraspecies differences in the amount, pattern, and sequence homology of PECAM-1 molecules were noted. For example, the PECAM-1 message of the predicted size (0.3 kb) was observed in two human tumor cell lines, i.e. HEL cells (Fig. 8, lane 3) and clone A cells (Fig. 8, lane 6 ) . But HEL cells also expressed two larger forms of the message as confirmed by hybridization. On the other hand, murine endothelial cells (CD3; Fig. 8, lane 1 ), murine fibroblasts (Fig.  8, lane 2) and B16F10 melanoma cells (Fig. 8, lane 7) expressed abundant PECAM-1 message, while B16a cells (lane 4 ) and 3LL cells (lane 5) expressed lower PECAM-1 mRNA which was barely visible on RT-PCR by ethidium bromide staining but whose presence was confirmed by hybridization. This low amount may also result from the possibility that the PECAM-1 sequence of B16a and 3LL is divergent from both 1 from surface iodinated clone A cells using mAh 1.3. HEL cells were used as the positive control cell line. The expected PECAM-1 band was indicated by arrows. C, cultured clone A or DLD-1 cells were either scraped using a rubber policeman or dissociated using 2 mM EDTA and then lysed in the hypotonic buffer for membrane preparation (see "Materials and Methods"). An equal amount of membrane protein was loaded onto 7.5% SDS-PAGE and the membrane was blotted using SEW-3. D. clone A cells were treated with either TPA or 12(S)-HETE (0.1 PM, 1 B min). scraped off using a rubher policeman, and used for membrane preparation. Fqual amounts of membrane proteins were separated on a 7.5% SDS-PAGE and the blotted membrane was stained using SEW-3. H P was included as a control. 6. N e u r a m i n i d a s e does not affect the mobility of murine and rat t u m o r cell PECAM-1 molecules. Nonidet P-40extracted protein samples prepared from human platelets or from tumor cells were treated with 1.0 unit/ml neuraminidase derived from C. perfringens (marked by +) or with buffer alone (lanes marked hy -) for 30 min on ice and then analyzed by immunoblotting using polyclonal anti-PECAM. Shown on the left are molecular mass in kDa. Note, the lower bands in 3LL lane are degradation products. The molecular mass standards are EcoRV-restricted X-phage fragments (see Fig. 8 for values). Note: the black spots below the PECAM-1 fragment in the hybridization panel are nonspecific binding.

FIG.
human and other murine cell counterparts. Interestingly, B16a and 3LL also had an upper band that comigrated with one of the upper bands in HEL cells but it did not hybridize to PECAM-1 probes, suggesting that they might be coamplified by-products. In contrast, all rat cell lines, including rat aortic endothelial cells (lane 8), rat carcinosarcoma cells (W256, lane 9), and AT 3.0 rat prostate carcinosarcoma cells (lane IO), expressed PECAM-1 mRNAs which were not amplified well by the human sequence-based PECAM-1 primers used in RT-PCR. However, the presence of the PECAM-1 message was corroborated by hybridization.
To further examine the expression of PECAM-1 in tumor cells, about 20 human tumor cell lines derived from different histological and pathological origins were screened for the expression of PECAM-1 message utilizing RT-PCR technique combined with Southern hybridization. The result of the screening (Fig. 9) indicated that all cell lines tested contained PECAM-1 message, although some tumor cell lines (e.g. MS751 human cervical carcinoma; lane 16) expressed little PECAM-1 mRNA. Again, heterogeneity was observed among these PECAM-1 messages, even among the tumor cells of the same histological and pathological source. A typical example was the WM series of melanoma cell lines (Fig. 9, lanes 1-7), some of which demonstrated an upper band. Surprisingly, none of these upper bands hybridized to the PECAM-1 cDNA probe. In contrast, the upper band in some other tumor cell lines (e.g. Du 145; lane 9 ) was positive by Southern blot.
The presence of PECAM-1 messages in tumor cells were further confirmed by Northern blotting analysis of human (HEL and clone A), rat (W256), and murine (3LL, B16F1, and B16F10) tumor cells (Fig. 10). The results revealed three bands, i.e. 3.7, 3.4, and 3.0 kb, for HEL cells, as reported by others (Zehnder et al., 1992). Hybridization revealed a single message of -4.1 kb in clone A, W256, B16F10, and, to a lesser extent, in B16F1 cells (Fig. 10). 3LLfcells demonstrated a mRNA band of 3.3 kb (Fig. 10). The loading of samples was confirmed by rehybridization to 8-actin cDNA probe.
Human Tumor Cell PECAM-1 Sequence Matches 100% to Human Endothelial Cell PECAM-1 Sequence-For final analysis, we obtained a partial cDNA sequence of PECAM-1 from a human tumor cell line, i.e. clone A colon carcinoma cells. Clone A cells were chosen because these cells are highly invasive, easily cultured, and express abundant amount of PECAM-1 protein (see Figs. 1-3). The DNA sequence ob-  belda et al., 1991), the tumor cell sequence was found to be identical (Fig. ll), therefore providing the conclusive evidence that solid tumor cells express PECAM-1.
PECAM-1 Is Involved in Tumor Cell Adhesion to Vascular Endothelium-Homologous in vitro cell adhesion assay was performed to examine the function of tumor cell PECAM-1 molecules. Radiolabeled tumor cells (i.e. B16a and 3LL) were coincubated with confluent murine microvascular endothelial cells (CD3) in the presence of anti-PECAM (SEW-3, mAb 1.3, or Fab Ab) or non-immune rabbit IgG or MOPC ascites (as the Ab control). Fig. 12a demonstrated that all three Abs could inhibit B16a cell adhesion to endothelium, although in general the pAb SEW-3 demonstrated the strongest inhibitory effect. An inhibition of approximately 40% was obtained with all of the antibodies. The adhesion-blocking effect was observed 10 min following addition of the antibody and persisted throughout the experimental period (up to 60 min). The inhibition by the pAb of 3LL cell adhesion to CD3 appeared to be greater than the inhibition with B16a (Fig. 12b). Dose studies indicated that SEW-3 exhibited a dose-dependent inhibition of B16a adhesion to endothelium (Fig. 12c). Incubation with antibody for a shorter time period (i.e. 10 min) appeared to give rise to a greater inhibition than observed with a longer time period (i.e. 20 min; Fig. 12c). When either tumor cells or endothelial cells were individually treated with antibodies (after which Abs were washed off) and then used in the adhesion assay (see "Materials and Methods" for details), inhibition of adhesion was also observed (data not shown). Collectively, these data suggest that tumor cells express functional surface PECAM-1 molecules which are involved in tumor cell adhesion to endothelium.

DISCUSSION
PECAM-1 is a member in the Ig family of adhesion molecules. A large array of Ig family adhesion molecules have been implicated in tumorigenesis and cancer metastasis. For example, both N-CAM and Ng-CAM have been detected in neuroblastoma and phaeochromocytoma and found to be related to tumor cell invasion (Brackenbury, 1985). Vascular cell adhesion molecule-1 expressed on cytokine-activated endothelial cells have been demonstrated to mediate tumor cell adhesion to the vascular monolayer via binding to VLA-4 (Martin-Padura et al., Taichman et al., 1991). Carcinoembryonic antigen gene family are expressed on solid tumor cell lines such as colon carcinoma and breast cancers and mediate either Ca2+-dependent (Turbide et al., 1991) or Ca2+independent homotypic tumor cell aggregation or heterotypic cell-cell adhesion. Another Ig family member, ICAM-1 (intercellular cell adhesion molecule-1), which is normally expressed on activated endothelium, has been observed to be expressed on solid tumor cells and its expression is correlated with metastatic potential (Johnson et al., 1989). Based on the above observations, we hypothesized that some solid tumor cells may express PECAM-1 and that PECAM-1 may be involved in tumor cell-endothelial cell interactions. Hence we undertook the studies presented in this paper.
Tumor cell PECAM-1 molecules were shown to be expressed on the cell surface, as indicated by immunocytochemical surface labeling, flow cytometry, immunoprecipitation of surface iodinated cells, as well as by subcellular fractionation studies using diverse antibodies. Tumor cells from different histological origins appear to exhibit different topological distribution patterns of PECAM-1 on the surface, since some tumor cells demonstrate homogeneous labeling (e.g. B16a melanoma), while others (e.g. 3LL lung carcinoma) demonstrate larger surface "granules" (or aggregates), and still others (e.g. clone A colon carcinoma) appear to be enriched for PECAM-1 molecules at the cell periphery ( i e . cell-cell contact zones; see Fig. 1). The heterogeneity in PECAM-1 expression is also observed within a specific tumor cell type, i.e. the level of expression is not homogeneous among all cells in a population. Labeling of permeabilized tumor cells also reveals an intracellular pool.
Western blotting using two pAbs and a mAb (Fig. 4) demonstrated that tumor cell PECAM-1 migrates in the range of 120-130 kDa, a molecular mass similar to PECAM-1 expressed in platelets and endothelial cells (Muller et al., 1989;Albelda et al., 1990;Newman et al., 1990;Albelda et al., 1991;this study). The specificity of a major polyclonal antibody used in the current study, i.e. SEW-3, was confirmed indirectly by preabsorbing this antibody with purified platelet membrane as well as directly by preincubating the antibody with PECAM-1-transfected cells. The molecular weight of tumor cell (i.e. clone A) PECAM-1 is not affected by reduction (data not shown), suggesting that this protein, like endothelial cell PECAM-1, is made up of a single polypeptide. The Western blotting results were confirmed by immunoprecipitation and subcellular fraction studies. PECAM-1 appears to be constitutively expressed on the surface of some tumor cells, e.g. clone A cells. Several lines of experimental data support this conclusion. First, immunofluorescence, peroxidase-antiperoxidase staining, and flow cytometric analysis all detected the surface labeling. Second, immunoprecipitation with surface-radiolabeled cells resulted in the protein band. Third, immunoblotting using membrane fraction revealed the PE-CAM-1 band. Finally, non-treated clone A cells and clone A cells treated with TPA or 12(S)-HETE do not demonstrate a difference in terms of the amount of membrane-associated PECAM-1. TPA and lB(S)-HETE, by activating protein kinase C, have previously been shown to increase the surface expression of integrin avP3 in endothelial cells (Tang et al., 1993a(Tang et al., , 1993b. Therefore, the observations in the present were resuspended in different concentrations of various antibody solutions and immediately added to confluent microvascular endothelial cell monolayer (CD3 cells). The adhesion was performed for different periods of time. Shown is the percentage of tumor cell adhesion to CD3 relative to the control (ie. control antibody-treated cells). Each condition was performed in quadriplicate and standard deviations were between 1-5% (not shown). a, B16a adhesion to endothelial cells was inhibited by antibodies (pAb, mAb, and Fab) against PECAM-1 (time course); b, pAb to PECAM-1 inhibited adhesion of both B16a and 3LL to endothelium (time course); c, dose study of pAb inhibition of B16a adhesion to endothelial cells. Two time points (10 and 20 min) of antibody treatment were shown. study suggest that PECAM-1 is constitutively expressed on the surface of tumor cells.
The form of PECAM-1 present on some tumor cells may have different biochemical properties than its platelet counterpart. Neuraminidase treatment does not result in any alterations in the molecular mass of tumor cell PECAM-1, although an approximate 5 kDa decrease is observed with the platelet PECAM-1 following neuraminidase treatment. Two possibilities arise from this observation. Tumor cell PECAM-1 molecules may not be significantly sialylated in the termini of their carbohydrate chains. Alternatively, tumor cells may possess aberrant or abnormal terminal sialyl residues which are not cleaved by neuraminidase from C. perfringens. This possibility is especially tempting and more plausible when considering that tumor cells are reported to possess aberrant glycosylation of surface glycoproteins.
The presence of PECAM-1 on solid tumor cells is confirmed by detection of PECAM-1 gene in tumor cell genomes using the PCR technique. The presence of PECAM-1 message in tumor cells was investigated by RT-PCR followed by hybridization of PCR-amplified fragments. Both PCR analysis of genomic DNA and RT-PCR of cellular RNA revealed a PCR fragment of the same size which was confirmed to be PECAM-1 fragment by subsequent hybridization. These observations suggest that this segment of PECAM-1 molecule as defined by the nested pair of primers does not encompass an intron. Several lines of experimental data exclude the possibility that the PECAM-1 fragment detected by RT-PCR is due to contaminated genomic DNA. First, the RT reaction buffer was treated by DNase prior to the initiation of RT reaction (see "Materials and Methods"). Second, in another set of experiments, the RT buffer was pretreated with RNase prior to RT and the subsequent PCR reaction did not reveal any product.2 Third, the PCR and hybridization patterns of genomic PCR and RT-PCR are significantly different. For instance, in RT-PCR B16a and 3LL cells and all rat cell lines do not demonstrate a well-defined PECAM-1 band (see Fig. 8). However, these cell lines demonstrate a strong PECAM-1 band in genomic PCR (Fig. 7). Another example is that hybridization of some RT-PCR-derived fragments gives two or more bands (e.g. in Du 145 cell line, see Fig. 9) while hybridization of genomic PCR-derived fragments all result in a single predicted band (Fig. 7).
Previously, Simmons et al. (1990) detected PECAM-1 transcripts in a metastatic colon carcinoma and the authors concluded that the transcripts might come from tumor-infiltrated macrophages. These authors did not detect PECAM-1 mRNA on other solid tumor cells. From our own experiments, we suspect that the negative results obtained by these authors are due to insensitivity of Northern blotting using total cellular RNA. Using purified mRNA for Northern blotting, we detected the PECAM-1 message in solid tumor cells. This provides substantive corroboration for the RT-PCR data. Three messages are detected in HEL cells. This observation is consistent with our RT-PCR data revealing the presence of multiple PECAM-1 mRNAs and with the Northern blotting results of others (Zehnder et al., 1992). In clone A, W256, 3LL, and B16F10 cells, only a single species of message is observed. The size of PECAM-1 mRNA in 3LL cells appears to be smaller than that in other cells and identical to the second species of mRNA in HEL cells ( i e . 3.3 kb). This difference in the size of PECAM-1 message in different tumor cell lines may represent cell type-specific alternative splicing. The molecular biology experiments we performed allow us to conclude: (a) PECAM-1 appears to be expressed on many human solid tumor cells, although the level of expression varies greatly among different cell lines; ( b ) human tumor cell PECAM-1 sequence is identical or highly homologous to that of endothelial cell PECAM-1; and ( c ) rodent solid tumor cells also express PECAM-1.
The significance of the detection of PECAM-1 on solid tumor cells is reinforced by the fact that it is widely expressed and involved in mediating tumor cell adhesion to vascular endothelium, one of the most important steps leading to organ preference of metastasis (Pauli et al., 1990;Honn and Tang, 1992). Adhesion assays either in the presence or absence (but with pretreatment of tumor cells with anti-PECAM-1) of Abs has consistently demonstrated that PECAM-1 is functional in mediating tumor cell adhesion to unstimulated endothe-D. G. Tang and K. V. Honn, unpublished observations. lium. PECAM-1 has been shown to be enriched at the cellcell borders of confluent endothelial cells in culture and of the vessel wall in situ (Muller et al., 1989;Albelda et al., 1990;Newman et al., 1990). Of interest, morphological studies on tumor cell-endothelial cell adhesion frequently indicate that adhesion preferentially occurs at the apposition zone between neighboring endothelial cells (Pauli and Lee, 1988). Thus, it is tempting to speculate that PECAM-1, among other adhesion molecules, may mediate early phase (ie. "docking"; see Honn and Tang, 1992) tumor cell adhesion to unstimulated endothelium, when activation-dependent adhesion molecules such as ICAM-1, E-selectin, P-selectin, and vascular cell adhesion molecule-1 are not available. Interestingly, Lee et al. (1992) recently reported that adhesion of melanoma cells to cultured human microvascular endothelial cells is independent of vascular cell adhesion molecule-1, E-selectin, and ICAM-1. It was hypothesized that some novel microvesselrelated adhesion proteins are involved. Our results suggest that PECAM-1 may be one of these "novel" adhesion molecules.
It is worthwhile to point out that we did not examine whether all available tumor cell lines express PECAM-1 protein although RT-PCR results indicated that most of these cells express readily detectable message. In situ hybridization experiments are underway to determine the expression of PECAM-1 mRNA in tumor cells in uiuo.