Alternative splicing of a specific cytoplasmic exon alters the binding characteristics of murine platelet/endothelial cell adhesion molecule-1 (PECAM-1).

Platelet/endothelial cell adhesion molecule-1 (PECAM-1, CD31) is a membrane glycoprotein expressed on endothelial cells, platelets, and leukocytes. Analysis of PECAM-1 expression in the developing mouse embryo has revealed the presence of multiple isoforms of murine PECAM-1 (muPECAM-1) that appeared to result from the alternative splicing of exons encoding cytoplasmic domain sequences (exons 10-16) (Baldwin, H. S., Shen, H. M., Yan, H., DeLisser, H. M., Chung, A., Mickanin, C., Trask, T., Kirschbaum, N. E. Newman, P. J., Albelda, S., and Buck, C. A.(1994) Development 120, 2539-2553). To investigate the functional consequences of alternatively spliced muPECAM-1 cytoplasmic domains, L-cells were transfected with cDNA for each variant and their ability to promote cell aggregation was compared. In this assay, full-length muPECAM-1 and all three isoforms containing exon 14 behaved like human PECAM-1 in that they mediated calcium- and heparin-dependent heterophilic aggregation. In contrast, three muPECAM-1 variants, all missing exon 14, mediated calcium- and heparin-independent homophilic aggregation. Exon 14 thus appears to modulate the ligand and adhesive interactions of the extracellular domain of PECAM-1. These findings suggest that alternative splicing may represent a mode of regulating the adhesive function of PECAM-1 in vivo and provides direct evidence that alternative splicing involving the cytoplasmic domain affects the ligand specificity and binding properties of a cell adhesion receptor.

PECAM-1 1 (CD31) has the distinctive feature of being found on platelets, leukocytes, and endothelial cells (reviewed in DeLisser et al. (1994b)). Recent observations have implicated PECAM-1 in a number of important processes. PECAM-1 fa-cilitates the diapedesis of leukocytes both in vitro and in vivo Vaporciyan et al., 1993;Bogen et al., 1994), acts as a trigger for up-regulating integrins on leukocytes (Tanaka et al., 1992;Piali et al., 1993;Berman and Muller, 1995;Leavesley et al., 1994), and thus appears to play a role in cell-cell interactions during an inflammatory response. It is also one of the first cell surface molecules to be expressed by endothelial and endocardial cells during embryonic development, suggesting that it may be involved in the establishment of the early cardiovascular system (Baldwin et al., 1994).
PECAM-1 is organized into an extracellular amino-terminal domain containing 6 immunoglobulin (Ig)-like repeats, a short hydrophobic transmembrane domain, and long cytoplasmic tail (Newman et al., 1990). The gene for human PECAM-1 (huPE-CAM-1) has recently been characterized  and is composed of 16 exons separated by introns ranging in size from 86 to more than 12,000 base pairs in length. Each of the six extracellular Ig homology domains is encoded by a single exon (exons 3-8). The transmembrane region is encoded by one exon (exon 9), while the cytoplasmic tail is encoded by a series of six short exons (exons 10 -16). This multi-exon structure of the cytoplasmic domain is quite unusual for Ig superfamily members. A number of cytoplasmic domain variants of PECAM-1, presumably arising from alternative splicing, have been identified in huPECAM-1 (Goldberger et al., 1994;Kirschbaum et al., 1994) and in murine PECAM-1 (muPE-CAM-1) (Baldwin et al., 1994). Analysis of PECAM-1 expression in the developing mouse embryo documented the presence of at least six isoforms of muPECAM-1 that appeared to result from the alternative splicing of the exons encoding the cytoplasmic domain (Baldwin et al., 1994).
Although the role of these multiple isoforms of PECAM-1 is currently unknown, studies examining the binding characteristics of huPECAM-1 suggest that alterations in the cytoplasmic domain have important functional implications. Previous experiments showed that full-length huPECAM-1 promoted heterophilic aggregation in transfected L-cells in a divalent cation-dependent, heparin-sensitive manner (Albelda et al., 1991;Muller et al., 1992;DeLisser et al., 1993). In contrast, mutants of huPECAM-1 with partially truncated cytoplasmic domains mediated aggregation that was quite different in that it was homophilic, divalent cation-independent, and heparin-insensitive (DeLisser et al., 1994a). These findings raised the possibility that naturally occurring alternatively spliced forms of PECAM-1 might also function differently (Baldwin et al., 1994).
The purpose of this study was to analyze the adhesive properties of each of the muPECAM-1 isoforms detected in early mouse embryos (Baldwin et al., 1994) and, if possible, identify * This work was supported by grants from the Robert Wood Johnson Foundation Minority Faculty Development Program, W. W. Smith Charitable Trust Grant, and National Institutes of Health Grants HL-46311, HL-51533, HL-39023, HL47670, CA 19144, and CA10815. 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.
** Established Investigator of the American Heart Association. ‡ ‡ To whom correspondence should be addressed: 894 Maloney Bldg., University of Pennsylvania Medical Center, 3600 Spruce St., Philadelphia, PA. 19104. Tel.: 215-349-8362;Fax: 215-349-5172. 1 The abbreviations used are: PECAM-1, platelet/endothelial cell adhesion molecule-1; hu, human; mu, murine; HBSS, Hanks' balanced salt solution; mAb, monoclonal antibody; FACS, fluorescence-activated cell sorting. specific regions of the cytoplasmic domain that determined the binding characteristics of the molecule. To accomplish this, L-cells were transfected with the cDNA for each variant and the functional properties of each isoform analyzed using the L-cell aggregation assay. These experiments documented that consistent changes in the aggregation properties of the cells were correlated with specific isoforms of muPECAM-1. Fulllength muPECAM-1 and all isoforms containing peptide sequences encoded in cytoplasmic exon 14 mediated heterophilic, calcium-dependent, heparin-sensitive aggregation. In contrast, all isoforms missing this peptide sequence in their cytoplasmic domains demonstrated homophilic aggregation that was calcium-independent and heparin-insensitive. These findings provide direct evidence for the hypothesis that naturally occurring alternatively spliced PECAM-1 isoforms have different ligand specificity and suggest that alternative splicing may be a method of regulating the adhesive function of PECAM-1 in vivo. In addition, these results pinpoint exon 14 as a key region of the cytoplasmic domain that determines the ligand and adhesive interactions of PECAM-1.

EXPERIMENTAL PROCEDURES
Naming of the Cytoplasmic Domain Variants of muPECAM-1-A total of six isoforms of muPECAM-1 were isolated and studied; a fulllength form and five alternatively spliced cytoplasmic domain variants. The full-length construct was designated muPECAM-1, while each of the five other isoforms was identified based on the exon that was deleted. For example, muPECAM-1⌬12,15 designates the isoform missing exons 12 and 15.
"Shot-gun" Cloning and Isolation of Alternatively Spliced muPE-CAM-1 Isoforms-Dated ICR mice were purchased from Harlan Sprague-Dawley (Indianapolis IN). Using previously described methods (Chomczynski et al., 1987), total RNA was isolated from the pooled tissue of early mouse embryos. Poly(A) mRNA, extracted by an oligo(dT) spin column (Micro Fast Tract, Invitrogen) was subsequently used as a template for reverse transcription (RT) in a reaction mixture of random and oligo(dT) primers and avian myeloblastosis virus reverse transcriptase as outlined in the manufacturer's instructions (cDNA Cycle Kit, Invitrogen). To polymerase chain reaction amplify the cytoplasmic domains of all possible muPECAM-1 isoforms from the reverse transcribed cDNA, the following primers were used: a sense primer (5Ј-1395 TATGAAAGCAAAGAGTGA 1412 -3Ј) flanking the BstEII restriction site within the extracellular domain and an antisense primer (5Ј-CGAATGC 2253 ATCCAGGAATCGGCTGCTCTTC 2235 -3Ј) complementary to a region 70 base pairs downstream from the stop codon of muPECAM-1 and carrying a 5Ј new NsiI recognition sequence to facilitate subsequent cloning. The polymerase chain reaction product was digested with BstEII and NsiI and ligated into a pcDNAI/Neo vector (Invitrogen) containing muPECAM-1⌬12,15 that was also cut with the same enzymes. The resulting ligation mixture was then used to transform competent cells. Antibiotic resistant clones were initially screened by size following digestion of miniprep plasmid DNA with a unique single cutting restriction enzyme. Promising clones were then screened by polymerase chain reaction with the following primer pair: sense primer, 5Ј-1852 CCAAGGCCAAACAGA 1866 -3Ј representing a region of muPECAM-1 homologous to exon 10; and antisense primer, 5Ј-2172 AAGGGAGCCTTCCGTTCT 2157 -3Ј representing a region in exon 16 of the cytoplasmic domain of the published sequence of the PECAM-1 gene. Candidate clones were then sequenced to confirm their identity using two different primer pairs from two different orientations: sense primers, 5Ј-1852 CCAAGGCCAAACAGA 1866 -3Ј and 5Ј-1572 AAGTTTTA-CAAAGAAAAGGAGGAC 1593 -3Ј; antisense primers, 5Ј 2172 AAGGGAG-CCTTCCGTTCT 2157 -3Ј and 5Ј-CGAATGC Ϫ253 ATCCAGGAATCGGC-TGCTCTTC 2235 -3Ј.
Tissue Culture and Transfection of L-cells-L-cells were purchased from the American Type Culture Collection (ATCC, Rockville, MD) and were cultured in RPMI medium with 10% fetal bovine serum. The procedures to transfect PECAM-1 cDNA into these cells have been previously described (Albelda et al., 1991;DeLisser et al., 1994a).
Fluorescence-activated Cell Sorting (FACS) Analysis-L-cells expressing muPECAM-1 isoforms were non-enzymatically removed from T25 flasks (unlike huPECAM-1, muPECAM-1 is sensitive to trypsin), washed in medium containing 10% fetal bovine serum, and treated with various anti-muPECAM-1 monoclonal antibodies for 1 h at 4°C. The primary antibody was then removed, the cells washed twice with icecold phosphate-buffered saline, and a 1:200 dilution of fluorescein isothiocyanate-labeled goat anti-rat secondary antibody (Cappell) added for 30 min at 4°C. After washing in cold phosphate-buffered saline, flow cytometry was performed using an Ortho Cytofluorograph 50H cell sorter equipped with a 2150 data handling system (Ortho Instruments, Westwood MA).
Cell Labeling-To metabolically label cellular proteins, T25 flasks of confluent transfected L-cells were incubated in medium containing 100 Ci of [ 35 S]methionine for 18 h as described previously (Albelda et al., 1990). Cell surface biotinylation was performed as described previously (Low et al., 1991). For 32 P labeling, confluent monolayers of L-cell transfectants in six-well plates were labeled with [ 32 P]orthophosphate using methods outlined previously (DeLisser et al., 1994a).
Immunoprecipitation-Nonionic detergent cellular extracts were prepared by adding small volumes of TNC (0.01 M Tris acetate, pH 8.0, 0.5% Nonidet P-40, 0.5 mM Ca 2ϩ ) with 2 mM phenylmethylsulfonyl fluoride to confluent cells grown in tissue culture flasks or six-well trays on ice for 20 min. The resulting extracts were preadsorbed for 30 min at 4°C with protein G-conjugated Sepharose beads (Pharmacia Biotech Inc.). After removal of the beads, the appropriate antibody was added to the precleared supernatants, together with fresh protein G-Sepharose beads for 1 h at 4°C. Following immunoprecipitation, the beads were washed five times with 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 5% deoxycholate, and 0.1% SDS. The sample was then dissolved with electrophoresis loading buffer (62.5 mM Tris base, 2% SDS, 10% glycerol, pH 6.8), electrophoresed on 6% polyacrylamide gels, and processed for autoradiography using Kodak XR-5 x-ray film at Ϫ70°C.
Glycosylation Inhibition-For deglycosylation experiments, L-cell transfectants were cultured in normal media with deoxymannojirimycin (40 g/ml; Calbiochem Corp.; Diamond et al., 1991) or swainsonine (0.1 mM; Sigma; Nguyen et al., 1992) for 48 h. The cells were then detached from the cell culture plates with trypsin-EDTA and replated with fresh media with inhibitor for an additional 24 h. To remove sialic acid residues, L-cell transfectants were detached from cell culture plates with Enzyme Free Cell Dissociation Solution (Specialty Media, Lavallette, NJ), washed three times with phosphate-buffered saline with 1 mM MgCl 2 and 0.3 mM CaCl 2 , and resuspended in the same buffer containing 0.2 unit/ml Vibrio cholera neuraminidase (Calbiochem Corp.) for 75 min at 37°C with gentle rocking (Diamond et al., 1991). Following treatment, cells were then used in aggregation experiments or biotinylated detergent extracts were prepared and immunoprecipitated as described above.
Aggregation of L-cell Transfectants-The aggregation assay used in these studies has been described in detail previously . Briefly, stable L-cell transfectants, which had been plated (8 -10 ϫ 10 6 cells/75-cm 2 flask) and grown overnight, were non-enzymatically removed. The cells were washed twice with 10 mM EDTA in phosphatebuffered saline, pH 7.2, and twice with HBSS without divalent cations. Cells were finally resuspended to a concentration of approximately 5-7 ϫ 10 5 /ml in HBSS with or without 1 mM calcium. Heparin (sodium salt, from porcine intestinal mucosa; Sigma), and antibodies were added as indicated to give the desired concentration. After the cells had been dispersed to a single cell suspension, 1-ml aliquots were transferred to wells in a 24-well non-tissue culture plastic tray (Costar Corp., Cambridge, MA) that had been previously incubated with 2% bovine serum albumin in HBSS for at least 1 h and washed thoroughly with HBSS immediately before use to prevent nonspecific binding to the plastic of the tray. The non-tissue culture-treated trays containing the suspended L-cells were rotated on a gyratory platform (100 rpm) at 37°C for 30 min. Aggregation was quantified by examining representative aliquots from each sample on a hemocytometer grid using phase-contrast optics. The number of single cells (cells in aggregates Յ3) remaining versus those present in aggregates of greater than three cells were counted from four 1-mm squares. At least 400 cells were counted from each sample. Data were expressed as the percent of total cells present in aggregates.
Mixed Aggregation Assay-To determine whether the muPECAM-1-dependent L-cell aggregation was mediated by homophilic or heterophilic mechanisms, "mixed aggregation" assays were performed (Muller et al., 1992;DeLisser et al., 1993). In these experiments, L-cell aggregation was performed by mixing non-transfected and transfected cells, with one of the two cell types fluorescently labeled prior to mixing. After the cell line designated for labeling had been washed once with EDTA, it was resuspended in HBSS to a total volume of 1 ml. One ml of rhodamine-conjugated dye solution at a final concentration of 1 mM (Sigma), in buffer provided by the manufacturer, was added, followed by incubation at room temperature for 5-10 min. Labeling was terminated by adding an equal volume fetal bovine serum and by washing the cells with HBSS. The second EDTA wash and the two HBSS washes were then performed as described above. Each set of cells, one labeled, and the other unlabeled, were resuspended at 5-7 ϫ 10 5 cells/ml. Aliquots of 0.5 ml of each were combined in the wells of a 24-well non-tissue culture treated plate and allowed to aggregate as described above. After the aggregation was completed, the cells were viewed under epifluorescence. The number of fluorescent cells in each aggregate of a given size was counted. Quantitative analysis of the aggregating cell populations was performed as described by Sieber and Roseman (1981).

Molecular
Cloning and Identification of muPECAM-1 from Early Mouse Embryos-To isolate muPECAM-1 cytoplasmic domain isoforms for sequencing and functional analysis, we employed a shot-gun cloning strategy (detailed under "Experi-mental Procedures") in which a mixture of murine PECAM-1 isoforms isolated from an early mouse embryo was directly cloned into an expression vector. Clones were screened by polymerase chain reaction with primer pairs designed to amplify the entire cytoplasmic domain. The identities of candidate clones were then determined by dideoxy DNA sequencing. Comparison with the sequence of the human gene encoding PECAM-1 confirmed that these six isoforms, a full-length form (muPECAM-1) and five cytoplasmic domain variants (muPE-CAM-1⌬12, muPECAM-1⌬14, muPECAM-1⌬15, muPECAM-1⌬14,15, and muPECAM-1⌬12,14,15) ( Fig. 1), were the result of alternative splicing. Four isoforms (muPECAM-1⌬15, mu-PECAM-1⌬12,15, muPECAM-1⌬14,15, and muPECAM-1⌬12,14,15) were found to have an alternative exon 16, resulting from a frameshift caused by the insertion of an extra A or G residue before the first base of exon 16. These isoforms are unlikely to be the result of errors introduced during the reverse transcription, as these extra bases were found in 14 clones sequenced from four different isoforms. A seventh isoform missing exons 12 and 15 was previously isolated from an adult heart library (Baldwin et al., 1994).
Expression of muPECAM-1 Isoforms in L-cell Fibroblasts-Each isoform was expressed in L-cells. After selection in G418, FIG. 1. Comparison of nucleotide sequences of muPECAM-1 isoforms and huPECAM-1. The nucleotide sequences encoding the cytoplasmic domains of the seven muPECAM-1 (Mu) isoforms are compared to the sequence for huPECAM-1 (Hu) as previously reported by Newman et al. (1990). Identical residues are indicated by hyphens (Ϫ), while the solid bars designate deleted nucleotide sequences. A space appears between predicted exon sequences . Note the change in the reading frame of exon 16 and the resulting truncation of the cytoplasmic domain of muPECAM-1⌬15, muPECAM-1⌬12,15, muPECAM-1⌬14,15, and muPECAM-1⌬12,14,15. clones were isolated, and levels of PECAM-1 expression on the cell surface were determined by FACS analysis. Each clone studied expressed PECAM-1 on its surface on greater than 90% of the cells (Fig. 2) at similar fluorescence intensities (Table I).
Multiple independent clones of each isoform expressing equivalent levels of protein were chosen for further study. To confirm the molecular mass of each variant, surface biotinylated protein extracts from L-cell transfectants were immunoprecipitated with mAb 390 and resolved on SDS-polyacrylamide gel electrophoresis. As illustrated in Fig. 3, each isoform, relative to full-length muPECAM-1 was of the appropriate molecular mass.
Deletion of Exon 14 Changes Monoclonal Antibody Sensitivity-To confirm that the interactions studied were in fact PE-CAM-1-dependent, three monoclonal antibodies (390, EA-3, and Mec 13.3; 50 g/ml) were tested for their ability to inhibit aggregation. Each antibody recognized all isoforms by FACS analysis (Table I) and by immunoprecipitation (data not shown). All three antibodies inhibited aggregation mediated by muPECAM-1, muPECAM-1⌬12, muPECAM-1⌬15, and muPE-CAM-1⌬12,15 (Fig. 5A). A different pattern, however, was noted for muPECAM-1⌬14, muPECAM-1⌬14,15, and muPE-CAM-1⌬12,14,15. Although mAb EA-3 and mAb Mec 13.3 blocked aggregation, no inhibition was seen with mAb 390 in the presence (or absence; data not shown) of calcium (Fig. 5B). For muPECAM-1⌬14,15, the control antibody appeared to increase aggregation; however, mAb 390 did not inhibit aggregation below the condition with no antibody. Failure of mAb 390 to inhibit aggregation was not due to a loss in ability to bind to these isoforms since each displayed roughly equal binding on FACS analysis (Table I)  isoforms The ability of an irrelevant control antibody and three ant-muPE-CAM-1 monoclonal antibodies (390, EA-3, and Mec 13.3) to bind to L-cells expressing muPECAM-1 isoforms was assessed by fluorescenceactivated cell sorting analysis. The mean (log) fluorescence intensity for each antibody was determined for each cell type. The three monoclonal antibodies bound equally well to each isoform.

FIG. 3. Immunoprecipitation of muPECAM-1 isoforms expressed in L-cell.
To confirm the molecular mass of each isoform, surface biotinylated protein extracts from L-cells expressing each variant were immunoprecipitated with mAb 390. FL designates (fulllength) muPECAM-1; ⌬15, muPECAM-1⌬15; ⌬12, muPECAM-1⌬12, etc. To facilitate comparison, constructs were ordered on the gel based on the number of amino acids in their cytoplsmic domain. The molecular mass of the different muPECAM-1 constructs varied from ϳ110 to 130 kDa. monoclonal antibody used as a control was not inhibitory. These findings confirm the involvement of PECAM-1-dependent interactions and support the conclusion that removal of exon 14 leads to differences in the binding characteristics of the different isoforms.
Deletion of Exon 14 Results in Homophilic Aggregation-The experiments described above suggest that the aggregation mediated by the isoforms missing exon 14 differed significantly from that of the isoforms of Group 1. To determine if the ligand binding specificity was heterophilic or homophilic for each of the isoforms, a mixed aggregation assay was used (Muller et al., 1992;DeLisser et al., 1993). In these studies, equal mixtures of non-transfected L-cells and L-cells transfected with muPECAM-1 were allowed to aggregate together in the presence of calcium, after fluorescent labeling of one of the cell lines. After 30 min, the composition of the aggregates of specific sizes was determined and a frequency distribution tabulated. Histographic analysis of the resulting aggregates can easily distinguish between a homophilic mechanism (PECAM-1 interacting with PECAM-1) and a heterophilic process (PECAM-1 interacting with a non-PECAM-1 ligand) (Albelda et al., 1991;Muller et al., 1992;DeLisser et al., 1993). In mixed aggregation assays with the isoforms of Group 1, the majority of the aggregates were composed of mixtures of transfected and non-transfected L-cell (heterophilic adhesion) (Fig. 6A). This pattern was consistent with a heterophilic interaction as noted for huPE-CAM-1 (Muller et al., 1992;DeLisser et al., 1993). Strikingly different results were obtained with variants missing exon 14 (Group 2) where aggregates consisted primarily of transfected cells (Fig. 6B). This is characteristic of a homophilic process and has been noted previously for cadherins (Jaffe et al., 1990;Muller et al., 1992) and for cytoplasmic domain mutants of huPECAM-1 in which exons 12-16 or 14 -16 were deleted (DeLisser et al., 1994a). Similar results were obtained for 4and 6-cell aggregates and when experiments were conducted in the absence of calcium (data not shown).
FIG. 6. Mixed aggregation studies. Mixed aggregation assays were performed in which equal numbers of non-transfected and transfected L-cells were mixed together after fluorescent labeling of one of the cell lines. After incubation, the number of labeled cells within each 5 cell aggregate was counted. The data is representative of at least three experiments. A, when muPECAM-1, muPECAM-1⌬12, muPECAM-1⌬15, and muPECAM-1⌬12,15 L-cell transfectants (Group 1) were mixed with non-transfected L-cells the majority of aggregates were made up of mixtures of transfected and non-transfected cells. A "normal" distribution is noted, indicative of a heterophilic interaction (see DeLisser et al., 1993). B, similar mixed aggregation experiments with muPECAM-1⌬14, muPECAM-1⌬14,15, and muPECAM-1⌬12,14,15 (Group 2) yielded aggregates that were composed primarily of transfected cells. The frequency distribution is shifted toward the right, reflecting a homophilic interaction. tracted with nonionic detergents and the extracts immunoprecipitated with mAb 390. Like huPECAM-1, full-length muPECAM-1 was constitutively phosphorylated (Fig. 7B, lane 1) (DeLisser et al., 1994a). Interestingly, although exon 14 contains more serine and threonine residues (three threonine and one serine) than exon 12 (one serine) or exon 15 (one serine), the level of phosphorylation for muPECAM-1⌬14 was comparable to that of full-length muPECAM-1 (Fig. 7B, Lane  4), while the loss of exon 12 or 15 was associated with diminished incorporation of 32 P label (Fig. 8B, lanes 2, 3, and 5-7). FACS analysis (see Fig. 2), as well as immunoprecipitation of parallel samples of transfected L-cells labeled with [ 35 S]methionine (Fig. 7A), showed that the reduced 32 P incorporation was not due to differences in the amount of PECAM-1 expressed by the transfected cells or to differences in the ability of the monoclonal antibody to immunoprecipitate each isoform. These data demonstrate that there is no simple correlation between gross changes in levels of phosphorylation and loss of exon 14.
Deletion of Exon 14 Alters the Dependence of PECAM-1mediated Aggregation on Glycosylation but Does Not Change Its Binding Characteristics-Glycosylation is known to influence the function of a number of Ig-like cell adhesion molecules. Since PECAM-1 is heavily glycosylated (Newman et al., 1990), we postulated that changes in glycosylation might affect one or both types of PECAM-1 mediated binding or might convert one type of binding to the other type. We therefore investigated the effect of inhibitors of carbohydrate processing on each type of muPECAM-1-dependent aggregation. The inhibitors used were 1-deoxymannojirimycin, an early inhibitor of the synthesis of hybrid and complex-type N-linked oligosaccharides), swainsonine (an agent which disrupts N-linked oligosaccharides processing at a later step in the biosynthetic pathway), and neuraminidase (an agent which removes sialic acid residues). One representative variant from each group (muPECAM-1 from Group 1 and muPECAM-1⌬14 from Group 2) was exposed to each inhibitor. The activity of each agent was confirmed by an increase in mobility on SDS-polyacrylamide gel electrophoresis of extracts of cells treated with each agent (Fig. 8A). All three compounds were found to partially inhibit the aggregation of muPECAM-1 (Fig. 8B) which remained calcium-dependent and heterophilic in nature (data not shown). In contrast, none of the inhibitors affected the binding demonstrated by muPECAM-1⌬14 (Fig. 8B). Thus, the lack of sensitivity to removal of carbohydrates on PECAM-1 represents another difference between the binding characteristics of these two groups of mu-PECAM-1 isoforms.

DISCUSSION
To determine the functional consequences of alternative splicing of muPECAM-1, the behavior of L-cell transfectants expressing various isoforms was compared in an established adhesion assay (Albelda et al., 1991;Muller et al., 1992;DeLisser et al., 1993). Full-length muPECAM-1 and the three isoforms containing exon 14 (muPECAM-1⌬12, muPECAM-1⌬15, FIG. 8. The effect of deglycosylation on muPECAM-1-dependent aggregation. A, to confirm the activity of these deglycosylating compounds, extracts of full-length muPECAM-1 (FL) and muPECAM-1⌬14 (⌬14) L-cell transfectants, treated with each agent, were immunoprecipitated with mAb 390. Increased mobility was noted for both isoforms following deglycosylation by each agent. B, deoxymannojirimycin (dMM), swainsonine, and neuraminidase were studied for their effect on the aggregation process in the presence of calcium (1 mM). All three compounds were found to inhibit aggregation mediated by mu-PECAM-1 but not that of muPECAM-1⌬14. The data presented are representative of at least two experiments done in duplicate or triplicate. Standard deviation is shown. and muPECAM-1⌬12,15) mediated calcium-dependent, heparin-sensitive, heterophilic aggregation that was inhibited by the anti-murine PECAM-1 monoclonal antibody, mAb 390, and was sensitive to deglycosylation. In contrast, all three of the muPECAM-1 variants missing exon 14 (muPECAM-1⌬14, mu-PECAM-1⌬14,15, and muPECAM-1⌬12,14,15) mediated homophilic aggregation that was calcium-independent, heparininsensitive and not inhibited by mAb 390 or affected by deglycosylation (see Table II). These findings are consistent with experiments showing that deletion of regions of the cytoplasmic domain of huPECAM-1 containing exon 14 resulted in a switch in the aggregation properties of transfected cells (DeLisser et al., 1994a) and give a possible biological context to these observations. These results provide further evidence that naturally occurring isoforms produced by alternative splicing lead to functionally different cytoplasmic domain variants of muPECAM-1. Given its wide distribution among vascular-associated cells and its putative roles in inflammation (Tanaka et al., 1992;Muller et al., 1993;Piali et al., 1993;Vaporciyan et al., 1993;Bogen et al., 1994;Berman and Muller, 1995), hematopoietic development (Leavesley et al., 1994), angiogenesis (Albelda et al., 1990), and vascular development (Baldwin et al., 1994), changes in the cytoplasmic domain of PECAM-1 generated by alternative splicing may play a role in modulating the ligand interactions of PECAM-1 and thus allow for a diversity of function.
Our results, however, differ from those reported by Xie and Muller (1993) who described a full-length muPECAM-1 that mediated calcium-independent, homophilic aggregation. The reasons for these differences are unclear, although they may be partially explained by somewhat different assay conditions in that a greater concentration of cells, higher divalent cation concentrations, and longer incubation time were used. Since muPECAM-1 and huPECAM-1 have a high degree of homology and resemble each other in all other respects studied, it seems reasonable that the basic aggregation characteristics of each form would also be similar.
A key question is how small changes in cytoplasmic domain can dramatically affect the binding characteristics of PE-CAM-1. The localization of exon 14, a small exon composed of 18 amino acids (LGTRATETVYSEIRKVDP), as the key region may be especially useful. A portion of this exon is very highly conserved between mouse and human PECAM-1. Interestingly, the 10-amino acid region, TETVYSEIRK, is almost identical to the cytoplasmic terminus (TETVYSEVKK) of the related Igsuperfamily member biliary glycoprotein (Rojas et al., 1990). Like PECAM-1 biliary glycoprotein mediates calcium-dependent adhesion. Loss of exon 14 could result in a number of possible effects including changes in phosphorylation or glycosylation, alterations in associations with cytoplasmic partners, and direct conformational changes.
Since there are 5 amino acids in exon 14 that could potentially serve as phosphorylation sites, one potential mechanism responsible for the changes in function resulting from loss of exon 14 could involve loss of a key phosphorylation site. PE-CAM-1 is constitutively phosphorylated and there is evidence to suggest that cellular activation is accompanied by changes in phosphorylation at serine and threonine residues Zehnder et al., 1992). Our data (Fig. 8), however, suggest that changes in the overall level of phosphorylation do not correlate with the deletion of exon 14. However, these data do not rule out the possibility that the absence of one or more of the specific phosphorable residues in exon 14 may be important.
Since glycosylation of the extracellular domain has been shown to play a crucial role in regulating the binding properties of a number of Ig superfamily members, now including PE-CAM-1, it is possible that loss of exon 14 could alter posttranslational processing and thus lead to a change in adhesive function of PECAM-1. Although there may be changes in specific carbohydrate residues induced by loss of exon 14, we have been unable to observe any gross changes in the overall level of glycosylation, as determined by the change in molecular size after deglycosylation. 2 Another possibility, currently being examined, is that association of the cytoplasmic domain of PECAM-1 with another membrane or cytoplasmic protein may be critical for its function. Exon 14 may be important in regulating this association. Preliminary experiments using differential detergent extractions have not revealed differences in the ability of the various isoforms to associate with the actin cytoskeleton. 3 However, this PECAM-1-associated molecule may be another PECAM-1 molecule, a molecule with enzymatic activity (such as the association of the cytoplasmic domain of CD4 with the tyrosine kinase p56 lck (Turner et al., 1990) or a novel protein.
Finally, it is theoretically possible that loss of exon 14 might directly lead to a change in the overall conformation of the molecule. A proline residue is the terminal amino acid of exon 14. Loss of this amino acid may induce an important conformational change that alters the ability of the cytoplasmic domain to fold normally. Clearly, this change must be rather subtle, since all of the mAbs tested bind equally well to all isoforms. Regardless of the mechanism(s) involved in regulating PE-CAM-1 ligand interactions, the implications of these findings extend beyond their significance to the structure-function relationships of PECAM-1. First, it is possible that the sequences found within exon 14 may have importance in the function of other Ig superfamily members (Rojas et al., 1990). Second, our findings indicate that observations made with artificial mutations of huPECAM-1 (DeLisser et al., 1994a) are likely to have physiological relevance and thus advance the emerging concept of the cytoplasmic domain as a regulator of not only the strength but the mechanism of adhesion. Finally, these data also suggest that alternative splicing of cytoplasmic domain regions may represent a novel way a cell can alter its interactions with the environment during development, inflammation, and wound healing.