Platelet endothelial cell adhesion molecule-1 (PECAM-1 / CD31): alternatively spliced, functionally distinct isoforms expressed during mammalian cardiovascular development

Establishment of the cardiovascular system is one of the earliest events in the developing embryo (Sabin, 1917; Sabin, 1920). The organization of emerging endothelial cells into a vascular tree that will eventually mature into the adult cardio-vascular system is the end result of a series of interactions between cells and their environment. It is a process mediated, at least in part, by a series of cell-cell and cell-matrix adhesive events involving specific receptors. In order to understand the molecular basis for the establishment of the cardiovascular system it is necessary to define the adhesion molecules mediating these interactions. This has been difficult in mammalian systems due to problems in identifying presump-tive endothelial precursors as they arise from the mesoderm. Early endothelial cell identification has been facilitated, in avian systems, by the production of monoclonal antibodies that react with presumptive endothelial cells as they appear in the developing blastodisc (Pardanaud et al., 1987; Coffin et al., 1991b; Noden, 1991). A descriptive account of vascular development in birds (Pardanaud et al., 1989; Poole et al., 1989), using the monoclonal antibody QH1, suggests that cardiovascular development commences in the embryo proper shortly after gastrulation, at about the head fold stage, as individual presumptive endothelial cells arise from the mesoderm. These cells soon connect into a network that matures into major vessels of the adult as well as the endocardium of the heart. Secondary vessels and capillary beds appear to originate from the larger vessels by angiogenesis. The molecular interactions determining the pattern of vascular formation in this system remain unknown.

A similar analysis of cardiovascular development in the mammalian embryo has been reported using lectins and antifactor VIII antibodies to identify endothelial cells (Coffin et al., 1991a). In general, the overview of vascular development provided by these studies shows a similar pattern to that seen in birds. However, the reagents used were specific for carbo-hydrate residues present on mature endothelial cells or products synthesized by endothelial cells after they have become part of an established vascular system. Therefore, certain endothelial cell populations, such as the endocardium, were not well delineated until later stages of development.

In order to evaluate early vascular development in the mammalian embryo, we initiated a search for an endothelial-specific cell adhesion molecule. We have, therefore, examined the possibility that platelet endothelial cell adhesion molecule (PECAM)-1 might be such a molecule. PECAM-1 was chosen because it is expressed by all endothelial cells in the adult (Albelda et al., 1990; Newman et al., 1990; Muller et al., 1989). It promotes cell-cell adhesion between cultured endothelial cells (Albelda et al., 1990), and between mouse L-cells transfected with PECAM-1 cDNA (Albelda et al., 1991; Muller et al., 1992). It is required for the movement of neutrophils and macrophages across the vascular endothelium (Bogen et al., 1992, 1994; Muller et al., 1993; Vaporciyan et al., 1993). Finally, ligand binding or PECAM-1 clustering results in the up-regulation of integrins in lymphocytes (Tanaka et al., 1992; Piali et al., 1993) suggesting that PECAM-1 engagement can regulate the expression of other adhesion receptors. It can also mediate adhesion by several alternative mechanisms (DeLisser et al., 1993), potentially expanding the repertoire of molecular interactions available to endothelial cells.

We report the results of an extensive analysis of PECAM-1 expression in the developing mouse embryo. The data clearly demonstrate that PECAM-1 is one of the earliest adhesion molecules expressed by presumptive endothelial cells. It is expressed as multiple isoforms that appear to be the result of alternative splicing of exons encoding portions of the cyto-plasmic domain. These isoforms differ in their sensitivity to a panel of monoclonal antibodies that block PECAM-1-mediated cell-cell aggregation, providing evidence for functional differences among naturally occurring isoforms, and supporting our previous speculation that the function of PECAM-1 may be regulated through a mechanism that involves modifications of the cytoplasmic domain (DeLisser et al., 1994).

A mouse heart cDNA library (Stratagene no. 936306, La Jolla, CA) was screened at high stringency with the full-length human PECAM-1 cDNA (Newman, 1990). A 3.2 kb fragment cloned into the EcoRI restriction site of the Bluescript SK II+ plasmid was identified and rescued from the λZapII vector according to the manufacturer’s directions. This clone was sequenced at least twice in both the forward (5′) and reverse (3′) direction by the didioxy-chain termination method using Sequenase DNA polymerase (United States Biochemical Co., Cleveland, OH). Assembly of multiple sequence contigs, as well as nucleic acid and deduced peptide comparisons, were accomplished using Version 7 of the Sequence Analysis software Package by Genetics Computer, Inc (Madison, WI).

Total RNA was extracted from staged CD1 mouse embryos by a modification of the method of Chomcsynski and Sacchi (1987) and Poly(A)⁺ mRNA was isolated using an oligo-dT spin column (Micro Fast Tract, Invitrogen). Complementary cDNA was obtained by reverse transcription using a mixture of random and oligo-dT primers, mRNA, and AMV Reverse Transcriptase (cDNA Cycle Kit, Invitro-gen). PCR reactions were performed in a Perkin-Elmer-Cetus automatic thermal cycler. Each 100 μl reaction mixture included 10 mM Tris-HCl, pH 8.0, 1.5 mM MgCl₂, 50 mM KCl, 0.1% gelatin, 10% DMSO, 0.06% 2-mercaptoethanol, 200 μM each dNTP and 2.5 units of Taq DNA polymerase (Perkin-Elmer-Cetus). Reactions were cycled (25 cycles for the extracellular domain primers and 35 cycles for the cytoplasmic domain primers) through a program that included incubations at 94°C for 1 minute, 56°C for 1 minute and 72°C for 2 minutes in the presence of 200 ng of each primer. For the nested PCR reactions, the initial (outside) reaction mixtures were filtered through an Ultrafree-MC Polysufone filter (Millipore), redissolved in 20 μl of TE, and 5 μl of this solution was used as a source of template for the second (inside) reaction under the conditions described above. Primers used to amplify a portion of the cDNA corresponding to the extracellular domain are delineated in Fig. 3. Primers designed to amplify the cDNA coding for a portion of the cytoplasmic domain of murine PECAM-1 were: 5′ primer = ccagctgctccacttctgaa and the 3′ primer = gcactgccttgactgtctta. The cytoplasmic primers were selected with the use of the Oligo 4.0 Primer Analysis Software (National Bio-sciences) to minimize false priming and maximize primer efficiency. Reaction products utilizing the extracellular primer set were resolved on a 0.8% agarose gel. PCR products utilizing the cytoplasmic primer mixture were resolved on a 4% Nusieve GTG agarose gel (FMC Bio-products) and visualized with ethidium bromide.

Agarose gels were transferred to nitrocellulose for Southern blotting according to standard protocols (Ausubel et al., 1993). In the case of the extracellular amplifications, ³²P-labeled probe was prepared by random priming. For the cytoplasmic amplifications, end labeled oligonucleotide cDNA probe corresponding to exon 13 was prepared using γ³²P[ATP] and T4 polynucleotide kinase (GIBCO-BRL).

To PCR amplify the cytoplasmic domains of all possible muPECAM-1 isoforms from the reverse transcribed cDNA, the following primers were used: sense primer (5′-¹³⁹⁵TATGAAAGCAAAGAGTGA¹⁴¹²-3′), flanking the BsteII restriction site within the extracellular immunoglobulin-like loop 5 and antisense primer (5′-CGAATGC²²⁵³ATCCAGGAATCGGCTGCTCT-TC²²³⁵-3′), com-plementary to a region 70 bps downstream of the stop codon of muPECAM-1, and carrying a 5′ mutagenic NsiI recognition sequence. The PCR product was digested with BsteII and NsiI and ligated into muPECAM-1Δ12,15 in the pcDNAI/Neo vector (Invitrogen). The resulting ligation mixture was then used to transform competent cells. DNA isolated from antibiotic resistant colonies was initially screened for endonuclease sensitivity. Selected clones were then characterized by PCR analysis using the following primer pair: sense primer, 5′-¹⁸⁵²CCAAGGCCAAACAGA¹⁸⁶⁶-3′, representing a region of muPECAM-1 homologous to exon 10 and antisense primer, 5′-²¹⁷²AAGGGAGC-CTTCCGTTCT²¹⁵⁷-3′, representing sequences homologous to exon 16 of the huPECAM-1 gene (Kirschbaum et al., 1994; Kirschbaum and Newman, 1993). DNA from colonies carrying isoforms of muPECAM-1 was sequenced using two different primer pairs in two different orientations: (sense primers, 5′-¹⁸⁵²CCAAGGC-CAAACAGA¹⁸⁶⁶-3′ and 5′-¹⁵⁷²AAGTTTTACAAAGAAAAGGAG-GAC¹⁵⁹³-3′; antisense primers, 5′-²¹⁷²AAGGGAGCCTTC-CGTTCT²¹⁵⁷-3′ and 5′-CGAATGC-²²⁵³ATCCAGGAATCGGCTG-CTCTTC²²³⁵-3′).

To obtain L-cell transfectants expressing high levels of protein, the muPECAM-1 cDNAs that had been subcloned into the pcDNAI/Neo (Invitrogen, San Diego, CA) expression vector were transfected into L-cells using calcium phosphate precipitation as previously described (Albelda et al., 1991) and muPECAM-1-expressing clones selected using G418.

The aggregation assay used in these studies has been described in detail previously (DeLisser et al., 1993). Briefly, stable L-cell trans-fectants, which had been plated (8–10×10⁶ cells/75-cm² flask) and grown overnight, were non-enzymatically suspended. The cells were washed twice with 10 mM EDTA in PBS, pH 7.2, and twice with HBSS without divalent cations. Cells were finally resuspended to a concentration of 8×10⁵/ml in HBSS with or without 1 mM calcium. Antibodies were added to a final concentration of 50 μg/ml. One ml aliquots of suspended cells were transferred to wells in a 24-well non-tissue culture plastic tray (Costar Corp., Cambridge, MA) that had been previously incubated with 2% BSA in HBSS for at least 1 hour and washed thoroughly with HBSS immediately before use to prevent nonspecific binding to the tissue culture dish. The non-tissue culture trays containing the suspended L-cells were rotated on a gyratory platform (100 rpm) at 37°C for 30 minutes.

Aggregation was quantified by examining representative aliquots from each sample on a hemocytometer grid using phase contrast optics. The number of single cells or cells in aggregates of 3 or less 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 percentage of the total cells present in aggregates.

L-cells transfected with full-length muPECAM-1, muPECAM-1Δ12,15 or muPECAM-1Δ14,15 cDNAs were non-enzymatically removed from T25 flasks, washed in medium containing 10% FBS, and treated with various anti-muPECAM-1 monoclonal antibodies for 1 hour at 4°C. The primary antibody was then removed, the cells washed twice with ice-cold PBS, and a 1:200 dilution of FITC-labeled goat anti-rat secondary antibody (Cappell) added for 30 minutes at 4°C. After washing in cold PBS, flow cytometry was performed using an Ortho Cytofluorograph 50H cell sorter equipped with a 2150 data handling system (Ortho Instruments, Westwood MA).

The following anti-muPECAM-1 monoclonal antibodies were used for immunohistochemical staining, FACS analysis, and to block the aggregation of transfected L-cells: mAb Mec13.3, provided by Dr Elizabetta Dejana (Institute Marion Negri, Milan), mAb EA3 (Piali et al., 1993) provided by Drs Luca Piali and Beat Imhof, and mAb 390, a monoclonal antibody generated in the rat following immunization with a mouse 32D leukocyte cell line and screened against muPECAM-1Δ12,15 (Edelman et al., unpublished data). Each of these antibodies immunoprecipitated the characteristic 120–130×10—?3 protein from L-cells transfected with muPECAM-1Δ12,15 and selectively stained murine vessels in acetone-fixed tissue sections.

Timed, pregnant CD1 mice were purchased from Harlan Sprague Dawley, Indianapolis, IN. The day of the appearance of the vaginal plug was considered day 0. Embryos of various gestational ages were removed by Cesarean section, dissected free of the desidual mass and staged according to the method of Kaufman (1992). Embryos used for whole-mount in situ hybridization were fixed in 4% paraformalde-hyde, washed with phosphate buffer containing Tween-20, dehy-drated stepwise in methanol, and stored at —20°C in 70% methanol until used. Embryos used for cross sectional immunohistochemistry were processed through a sucrose gradient in phosphate-buffered saline, embedded in OCT, and frozen in liquid nitrogen-cooled isopentane.

In situ hybridization was performed using the procedure of Conlon and Herman (1993) with slight modifications. Briefly, non-isotopic riboprobe was prepared from a 1.8 kb cDNA fragment consisting of 1.6 kb of 3′ coding sequence plus 0.2 kb of the 3′ untranslated sequence of murine PECAM-1 using RNA polymerase and a standard nucleotide mix including digoxigenin-labeled UTP (Boehringer Mannheim Genius 4 System). Embryos were rehydrated, washed and briefly digested with proteinase K. Following washing, the embryos were re-fixed with glutaraldehyde, washed and prehybridized at 63°C in 50% formamide, 0.75 M NaCl, 1× PE (10 mM Pipes, pH 6.8, 1 mM EDTA, pH 8.0), 0.1% BSA, 0.01% heparin, and 100 μg/ml yeast RNA with 1% SDS for 2-4 hours. The embryos were then hybridized in a similar buffer containing approximately 0.5–2 μg/ml of digoxi-genin-labeled probe overnight. After extensive washing, the embryos were exposed to pre-immune serum followed by sheep anti-digoxi-genin antibodies conjugated with alkaline phosphatase. Antibody was detected using nitroblue tetrazolium salt/5-bromo-4-chloro-3-indolyl phosphate.

Cryostat sections (7 μm) of sucrose-embedded embryos were fixed in methanol at 4°C and stored at —20°C. They were rehydrated in PBS for 30 minutes, soaked in a solution containing 4% BSA and 4% goat serum in PBS for 30 minutes and exposed to the appropriate primary monoclonal antibody for 1 hour in a humidified chamber at room temperature. The slides were washed three times in PBS and then coun-terstained with a FITC-conjugated goat antirat secondary antibody in a humidified chamber at room temperature in the dark for 1 hour. The specimens were washed three times in PBS, mounted in anti-quench mixture containing DABCO and photographed using a Leica fluores-cent microscope. mAb 390, a rat anti-mouse PECAM-1 antibody was used as the primary antibody for all PECAM-1 staining and MF20, a mouse anti-muscle myosin antibody (Bader et al., 1982) provided by the Developmental Studies Hybridoma Bank, University of Iowa, was used as a cardiac muscle marker.

For purposes of maximum specificity throughout these studies, mouse PECAM-1 (muPECAM-1) cDNA was isolated from an adult heart library screened with human PECAM-1 (huPECAM-1) cDNA. The complete nucleotide and amino acid sequence of the coding region of a 3 kb fragment of muPECAM cDNA is shown in Fig. 1. The full-length muPECAM-1 cDNA encodes a protein consisting of 727 amino acids. It exhibits a 60% identity (74% similarity) to human PECAM-1 at the amino acid level. Structurally, muPECAM-1, like huPECAM-1, is organized into an extra-cellular amino-terminal domain containing 6 immunoglobulin (Ig)-like repeats, followed by a short, hydrophobic transmem-brane domain and finally terminating in a cytoplasmic domain. The positions of all cysteines required for folding of the Ig repeats are conserved, as would be expected for a member of the Ig superfamily. In this respect, all known PECAM-1 cDNAs were found to be identical. Differences were however noted between the cDNA sequence reported previously (Xie and Muller, 1993), and those reported here. Three of four of the differences noted within the cytoplasmic domain (Fig. 1B) have been resolved in agreement with those presented here (Muller, personal communications). Differences within the extracellular domain have not been resolved and may well indicate the existence of other isomorphic forms of muPECAM-1. In contrast, we isolated cDNAs in which segments of the cytoplasmic domain sequence were absent (Fig. 1B). The size and location of the missing oligonucleotides corresponded precisely to exons within the human PECAM-1 gene (Kirschbaum and Newman, 1993; Kirschbaum et al., 1994) suggesting that they were the result of alternative splicing. These isoforms will be designated muPECAM-1Δ12,15 or muPECAM-1Δ14,15 to indicate the absence of oligonucleotides corresponding to exons 12 and 15 or 14 and 15 respectively.

Based upon the predicted exon arrangements in the human PECAM-1 gene (Kirschbaum et al., 1994), alternatively spliced isoforms would have differences in addition to the deleted polypeptide. Splicing out exon 12 would theoretically result in a change of the amino acid at the splice junction, as exon 12 is bracketed by type 1 intron inserts. Removal of exon 15 would result in a truncated terminal peptide with an altered amino sequence beginning at the splice junction joining exon 14 to exon 16 due to a type I intron insert at the 5′ end of exon 15 and a type 0 insert at the 3′ end. Thus, the deletion of exons 14 and 15 would result in a frame shift and early termination of the cytoplasmic domain (see Fig. 1B).

To examine the possible functional consequences of alternative splicing and premature termination, muPECAM-1 cDNAs encoding either full-length or alternatively spliced isoforms were cloned into expression vectors and transfected into L-cells. Transfected cells were examined for muPECAM-1 expression by FACS analysis using the monoclonal antibody 390 (Fig. 2). Control, mock transfected L-cells carrying the vector alone failed to react with this antibody. In contrast,strong expression of muPECAM-1, at equivalent levels, was detected on the surface of cells transfected with each isoform. The ability of each isoform to bind to a panel of monoclonal antibodies was similarly compared by FACS analysis (Table 1). By these criteria, the cDNAs encoding each of the isoforms were equally active. That is, all transfected cells expressed muPECAM-1 or one of its isoforms at similar levels and all appeared to bind each of the mAbs to a similar extent (Table 1).

We have previously shown that one of the characteristics of huPECAM-1 is its ability to mediate specific, mAb-sensitive aggregation of transfected L-cells (Albelda et al., 1991; DeLisser et al., 1994). Therefore, the ability of L-cells, transfected with each isoform, to aggregate in a muPECAM-1 dependent, mAb-sensitive manner was evaluated. The data are summarized in Fig. 3. When compared to mock transfected controls, each muPECAM-1 isoform was capable of mediating L-cell aggregation to the same extent as full-length muPECAM-1. Thus, alternative splicing did not compromise the ability of muPECAM-1 to promote L-cell aggregation. However, interesting differences were noted in the sensitivity of the transfectants to three anti-muPECAM-1 mAbs. Aggre-gation of cells expressing either the full-length muPECAM-1 or the Δ12,15 isoform was equally sensitive to inhibition by all three mAbs. In contrast, cells transfected with the Δ14,15 isoform, while remaining sensitive to mAbs Mec 13.3 and EA3, were completely resistant to inhibition by mAb 390. This resistance was not due to loss of reactivity, as FACS analysis revealed that the muPECAM-1Δ14,15 transfected cells continued to bind mAb 390 (Fig. 2, Table 1). This differential sensitivity to anti-muPECAM-1 mAbs indicates that changes in the cytoplasmic domain result in a selective change in the configuration of the extracellular, ligand-binding domain of the molecule. In addition, these data demonstrate that this muPECAM-1 isoform is mediating L-cell aggregation by a mechanism different from that of full-length or the Δ12,15 isoform. Thus, alternative splicing may provide a mechanism by which the function of muPECAM-1 may be regulated.

Three sets of oligonucleotide primers were designed to determine (1) the earliest time at which muPECAM-1 mRNA could be detected in the developing embryo and (2) the possible existence of alternatively spliced forms. The expression of muPECAM-1 in staged embryos was determined by nested RT-PCR analysis of mRNA. To avoid possible complications due to the presence of alternatively spliced isoforms, primers were designed to amplify a sequence encoding a portion of the extracellular domain (Fig. 4). The RT-PCR products were analyzed by agarose gel electrophoresis (Fig. 4A). Extracts from all embryos yielded fragments identical in size to those amplified from the control plasmid vector (P) carrying muPECAM-1 cDNA. Each nested product carried an expected DraI endonuclease site (Fig. 4; DraI). Their identities were confirmed by Southern blot analysis (Fig. 4B). These products were taken as specific indicators of the presence of muPECAM-1 mRNA. They were of the expected size; they hybridized with muPECAM-1 cDNA of the appropriate sequence; and no contaminating DNA was detected when the reverse transcription step was omitted or mRNA from 3T3 fibroblasts was used as a source of template. It should be noted that the fragment size would be considerably larger if contam-inated genomic DNA were amplified, as these primers spanned multiple exons (Kirschbaum et al., 1994). Thus, mouse PECAM-1 mRNA was present in the embryo immediately after implantation and prior to somitogenesis as would be expected for a molecule expressed by endothelial or hematopoietic progenitors.

In order to test the possibility that alternatively spliced isoforms of muPECAM-1 were expressed in the embryo, primers were designed, as shown in Fig. 5, to amplify sequences that included the alternatively spliced regions of the cytoplasmic domain. Messenger RNA was isolated from 8- and 12-day embryos, as well as from hearts of 8-, 12-, 16-day embryos and adults. The RT-PCR products from each preparation were surprisingly complex (Fig. 5), consisting of a mixture of oligonucleotides ranging in size from approxi-mately 317 to 239 bp. The top and bottom bands seen by ethidium bromide staining corresponded in size and sequence to the PCR products amplified from plasmids carrying muPECAM-1 or muPECAM-1Δ12,15 cDNA respectively. Approximately five different bands could be resolved by this method (Fig. 5A). These RT-PCR products fell within a size range that would be predicted for a mixture of muPECAM-1 cytoplasmic domain fragments missing various combinations of predicted exons. They were too small to represent products of genomic DNA amplified across several exons, and too large to represent a single exon (Kirschbaum et al., 1994). Failure to detect RT-PCR products of this size upon omission of the reverse transcription step, and the absence of any detectable products from mouse fibroblasts further confirm the specificity of these primers and the authenticity of the RT-PCR products.

Southern blot analysis, using probes homologous to the cytoplasmic domain, confirmed the existence of multiple muPECAM-1 isoforms (Fig. 5B). None of these bands hybridized with probes homologous to the extracellular domain of muPECAM-1 (data not shown). The presence of mRNA encoding multiple isoforms of muPECAM-1 was confirmed by sequencing ‘shot gun’ clones of RT-PCR products generated from these mRNA preparations using primers designed to span the entire cytoplasmic domain, beginning within the 3′-untranslated region of the message and extending into the extracellular domain (see Materials and Methods). These products, about 0.9 kb in size, each carried a BsteII endonuclease site on the 5′ end and a NsiI site on the 3′ end to facilitate site-specific cloning into a plasmid carrying muPECAM-1 cDNA digested with each of these enzymes. Random clones were sequenced as described in Materials and Methods. While an analysis of all clones has not been completed, the two isoforms whose sequences are shown in Fig. 1 were among the first isolated. Multiple isoforms have been consistently detected at different stages of embryonic development. So far, however, no reproducible differences in relative isoform abundance have been correlated with specific developmental stages.

The sites of muPECAM-1 expression in the developing embryo were determined by in situ hybridization using digox-igenin-labeled anti-sense and sense riboprobes. Anti-digoxi-genin immunoglobulin conjugated with alkaline phosphatase was used to detect hybridized complexes. muPECAM-1 mRNA was first detected in the region of the blood islands in the yolk sac of the pre-somite (day 7.0–7.5) mouse embryo (Fig. 6A, arrow). As development progressed, the muPECAM-1 mRNA was found in patterns consistent with it being expressed by the endothelium of the forming vasculature (Fig. 6B,C). In the 5-somite embryo (day 7.75–8.0), the message was found concentrated rostral to the foregut (fg) in the region of the developing heart (ht), in the forming dorsal aortae (da), as well as in the neural folds (nf) of the developing head process (Fig. 6B). By day 9.0–9.5 (Fig. 6C), the pattern of PECAM-1 mRNA expression clearly reflected that of the developing vas-culature including the second and third branchial arches (2,3), the ventricles (v) and atria (a) of the developing heart and inter-segmental arteries (isa). It continued to be evident in the maturing vasculature of the head and trunk including the dorsal aortae. An example of a control embryo exposed to the ‘sense’ oligonucleotide is shown in Fig. 6D. There was, at best, only faint background staining of control embryos.

In order to monitor message translation and to identify the cells synthesizing muPECAM-1, cross sections of an embryo (Fig. 7), including the heart, at the straight heart tube stage (Fig. 7A,B; day 8.5) were immunostained using mAb 390 specific for muPECAM-1. At this stage, the heart consists of cells organized into two concentric tubes. The outer layer of cells constitutes the developing myocardium (m) and the inner layer the developing endocardium (e). The dorsal aortae (da) are visible just below and lateral to the neural fold (nf). This section also included the extraembryonic membranes and the yolk sac (arrowheads). The muPECAM-1 mAb reacted only with the endocardial cells in the developing heart. No staining was detected in the outer myocardial layer. Immunoreactivity was also noted on cells organized into two major vessels on either side of the neural tube just above the developing foregut in the position of the dorsal aortae (da). At higher magnification (Fig. 7B), it was clear that the staining involved a single layer of cells surrounding the lumen of the dorsal aortae as well as the endocardium. In both cases, the staining pattern indicated a concentration of the antigen at cell-cell borders.

The anti-muPECAM-1 mAb also reacted with a single layer of cells beneath the extraembryonic endoderm of the surrounding yolk sac (Fig. 7A,B, arrowheads). At higher magnifications, it is clear that muPECAM-1 was expressed only on the cells forming the ‘wall’ of the blood islands and not the inner, presumably hematopoietic cells (Fig. 8). The extraem-bryonic mesoderm and endoderm were both negative for muPECAM-1. The position and distribution of this staining was consistent with the observation that muPECAM-1 mRNA was expressed in the yolk sac in the region of the forming extraembryonic vasculature and not on cells that might be considered extraembryonic hematopoietic stem cells.

During cardiac morphogenesis, the separation of the compartments of the heart and the partitioning of the conotruncal region into two major vessels, the aorta and the pulmonary artery, commence with the formation of the endocardial cushions (Markwald et al., 1984; Markwald et al., 1985). This involves a morphological transition of endocardial cells to mesenchymal cells (epithelial-mesenchymal transformation) accompanied by their migration into the underlying extracel-lular matrix to form the stroma of the endocardial cushions. A section through the conotruncus of an 11.5-day embryonic heart, in the region of endocardial cushion formation (Fig. 7C), illustrates the specific cellular components of this process. After staining with an anti-PECAM-1 mAb, the endocardium (e) lining the ventricle (v), the right and left atria (ra; la) and overlying the forming endocardial cushion (edc) demonstrated strong immunoreactivity. In contrast, the stromal cells of the endocardial cushion that originated from the overlying endo-cardium were completely unreactive with the antibody. This is particularly evident when viewed at higher magnification (Fig. 7D). The absence of immunoreactive muPECAM-1 was not due to masking of the antibody reactive site, as no reactivity was noted using three independently isolated mAbs and no muPECAM-1 mRNA expression was detected by in situ hybridization (Baldwin, unpublished observations). Fig. 7E shows an adjacent section of the same heart stained with MF 20, a mAb specific for muscle myosin (Bader et al., 1982). This antibody reacts only with muscle cells. Its absence from the region labeled endocardial cushion (edc) verifies that this is, in fact, the septating conotruncus and outflow tract.

This report documents three significant new observations con-cerning the immunoglobulin superfamily molecule PECAM-1. They are: (1) muPECAM-1 is expressed in multiple alternatively spliced forms in the developing mouse embryo; (2) alternative splicing of exons within the cytoplasmic domain is accompanied by changes in reactivity with function-blocking mAbs; (3) the temporal and spatial patterns of expression are consistent with a role for muPECAM-1 in the initial organization of the mammalian cardiovascular system.

Evidence for the existence of alternatively spliced mRNA encoding multiple muPECAM-1 isoforms is based upon three observations. First, multiple RT-PCR products were generated using mRNA preparations from staged embryos and primers bracketing the cytoplasmic domain (Figs 1–3). Second, comparative sequence analysis revealed that the isoforms were all missing segments corresponding to specific exons described for the huPECAM-1 gene (Kirschbaum and Newman, 1993; Kirschbaum et al., 1994). Third, the changes in oligonucleotide and polypeptide sequence at the splice sites were limited to those that would be predicted from the exon-intron organization within the homologous region of the huPECAM-1 gene. These observations have been consistent for over 30 different muPECAM-1 cDNAs sequenced to date (Yan, unpublished data). The distribution of cytoplasmic domain codons over several exons distinguishes PECAM-1 from other members of the immunoglobulin superfamily such as VCAM-1 and the ICAMs normally expressed by endothelial cells. The cyto-plasmic domains of these latter molecules are encoded entirely within a single exon (Voraberger et al., 1991; Xu et al., 1992; Cybulsky et al., 1991, 1993). In this respect, PECAM-1 is more like N-CAM (Owens et al., 1987) or isoforms of carcinoem-bryonic antigen (Barnett et al., 1989). The cytoplasmic domains of both of these Ig-superfamily molecules are encoded by more than a single exon and alternatively spliced isoforms with altered cytoplasmic domains have been identified. The functional significance of these isoforms is not known, but in both cases, larger regions of the cytoplasmic domain are lost by alternative splicing than is found for muPECAM-1. Inter-estingly, alternative splicing of the cytoplasmic domains of two members of the CEA family result in frame shifts and and premature termination as is reported for the two muPECAM-1 isoforms described here.

The possibility that the abundance of muPECAM-1 isoforms portends functional significance assumes, of course, that alternatively spliced isoforms retain their ability to mediate cell adhesion. While this has yet to be proven in situ, alternatively spliced isoforms were found to be as effective in mediating the aggregation of transfected L-cells as full-length human or mouse PECAM-1. However, variations in sensitivity to mAbs were noted among the isoforms tested here. L-cell aggregation mediated by either full-length muPECAM-1 or the Δ12,15 isoform was equally sensitive to a panel of three mAbs. In contrast, aggregation mediated by muPECAM-1Δ14,15 was unaffected by mAb 390 while remaining sensitive to two other adhesion-blocking mAbs (Fig. 3). This difference was reproducible and cannot be explained on the basis of changes in the level of muPECAM-1 expressed on the surface of the transfected cells, as they all showed similar levels of expression as measured by FACS analysis (Fig. 2; Table 1). In addition, all isoforms continued to localize to cell-cell borders precisely as noted for cultured endothelial cells (Albelda et al., 1991) when transfected cells are allowed to grow as monolayers (Baldwin, unpublished observations), supporting the contention that alternatively spliced forms retain their ability to participate in intercellular adhesive events.

The change in mAb sensitivity associated with the muPECAM-1Δ14,15 isoform and the absence of such a change in the Δ12,15 isoform indicate that the deletion of a specific portion of the cytoplasmic domain can result in secondary alterations within the extracellular, ligand binding domain of muPECAM-1. That a conformational change in the extracel-lular domain resulted from the cytoplasmic domain modifica-tion is also indicated by the change in mAb sensitivity. It is well known that antibodies recognize molecular conformation rather than peptide sequence and that most epitopes are dis-continuous and structurally distinct (Amit et al., 1986; Graeme et al., 1990). This characteristic has been exploited in several systems to demonstrate structural alterations accompanying changes in receptor function. For example, integrins on the surface of activated platelets undergo a structural change that is characterized by the acquisition of ligand binding ability and accompanied by the reproducible presentation of new epitopes or a change in the affinity of mAb binding (Du et al., 1993; reviewed by Ginsberg et al., 1992). Such changes in molecular conformation are common among the integrins, and are indica-tive of changes in ligand binding specificity as well as receptor activation. They are, however, less common among members of the immunoglobulin superfamily that have, up until this time, not been found to undergo transmembrane modifications of receptor function.

These observations are consistent with our previous results showing that the deletion of increasingly larger portions of the cytoplasmic domain of huPECAM-1 results in a molecule that exhibits different adhesion characteristics when compared to full-length PECAM-1 (DeLisser et al., 1994). In this case, L-cells transfected with full-length huPECAM-1 aggregate in a divalent cation dependent, heterophilic manner that can be blocked with sulfated glycosaminoglycans. In contrast, aggregation of L-cells expressing specific truncated forms of PECAM-1 occurs by a very different mechanism. It is divalent cation independent, homophilic, cannot be inhibited by sulfated glycosaminoglycans and exhibits different mAb sensitivities from full-length PECAM-1 (DeLisser et al., 1994). Preliminary evidence suggests that similar differences exist in the aggregation characteristics of L-cells expressing isoforms represented by muPECAM-1Δ14,15 (Yan et al., unpublished observations). Given the diversity of adhesive mechanisms utilized by muPECAM-1, and the ability to modify the adhesive characteristics of the molecule by manipulating the cytoplasmic domain, it is possible that the establishment of the embryonic vasculature, and perhaps later angiogenesis, depends upon the ability to regulate the adhesive properties of muPECAM-1 through its cytoplasmic domain either by changing the polypeptide sequence, as shown here, or perhaps by phosphorylation or interaction with cytoplasmic protein kinases or other factors known to promote signal transduction.

The temporal and spatial patterns of muPECAM-1 expression are consistent with it being one of the early adhesion receptors utilized by vasculogenic cells. It is expressed in the yolk sac immediately after implantation, and subsequently in the embryo itself. Its pattern of expression resembles that reported in the quail embryo, using the endothelial cell-specific marker QH-1 (Pardanaud et al., 1987; Coffin et al., 1991b), except that muPECAM-1 was never detected on extraembryonic hematopoietic cells within the forming blood islands (Fig. 8). However, like QH-1, muPECAM-1 was detected in the early post implantation embryo across the anterior intestinal portal in the region of the heart primordia (Baldwin et al., 1991) and extending into the head process on cells with thin, thread-like processes suggestive of vascular primordia (Fig. 7). Also, as was found for QH-1, muPECAM-1 expression proceeds in a rostral to caudal fashion in a pattern consistent with it being expressed on endothelial cells as they emerge from the mesoderm and are organized into the forming vasculature and heart. Secondary vessels appear to arise from the primary vascular tree. This is particularly evident along the dorsal aortae where the intersegmental arteries arise as regularly spaced lateral extensions of the main vessel (Fig. 7). The fact that muPECAM-1 was expressed by early vascular endothelial cells was further supported by the correlation of the pattern of muPECAM-1 expression seen by immunohistochemistry and that noted by in situ hybridization. This expression pattern is similar to that reported for members of the receptor protein tyrosine kinase (RTK) superfamily. These include flk-1, tek and tie-2 in the mouse (Yamaguchi et al., 1993; Oelrichs et al, 1993; Dumont et al., 1992; Millauer et al., 1993; Schurch et al., 1993) and quek 1 and quek 2 in the quail (Eichmann et al., 1993). muPECAM-1 is seen at the same time and place as flk-1 (Yamaguchi et al., 1993), and is co-expressed with tie 2 (Schurch et al., 1993) suggesting that the muPECAM-1 gene might be activated as a result of RTK/ligand interactions.

Transplantation experiments in which endothelial cells from the quail were placed in chick embryos suggest that much of the information required for the organization of endothelial cells into vessels is intrinsic to the cells themselves, and relies on cues from the extracellular environment (Noden, 1990). This, more than likely, is mediated through cell surface receptors capable of responding to the molecular composition of the extracellular matrix as well as to neighboring cells. muPECAM-1 may well be such a molecule. The ability of PECAM-1 to mediate adhesion via several different mecha-nisms (DeLisser et al., 1993, 1994) raises the possibility that it may participate in a broad spectrum of ligand interactions, some resulting in adhesion and others triggering intracellular signals necessary to stabilize other adhesive events (Tanaka et al., 1992; Piali et al., 1993). Thus, once present, PECAM-1, like other adhesion receptors, might, in addition to functioning as an adhesion molecule, initiate a signal transduction cascade required for vascular organization in response to a variety of extracellular cues.

We wish to thank Ms. Irene Crichton and Ms. Catherine Buck for excellent technical assistance and Ms. Marie Lennon for invaluable help in preparation of this manuscript. This work was funded by NIH HL 51533; HL 39023; HL 47670; HL 2917; CA 10815; CA 19144; W. W. Smith Charitable Trust Grant H12901 and the Robert Wood Johnson Foundation Minority Faculty Development Program.