A Novel Protein with Homology to the Junctional Adhesion Molecule

We have cloned a novel cDNA belonging to the Ig superfamily that shows 44% similarity to the junctional adhesion molecule (JAM) and maps to chromosome 21q21.2. The open reading frame of JAM2 predicts a 34-kDa type I integral membrane protein that features two Ig-like folds and three N-linked glycosylation sites in the extracellular domain. A single protein kinase C phosphorylation consensus site and a PDZ-binding motif are present in the short intracellular tail. Heterologous expression of JAM2 in Chinese hamster ovary cells defined a 48-kDa protein that localizes predominantly to the intercellular borders. Northern blot analysis showed that JAM2 is preferentially expressed in the heart. JAM2 homotypic interactions were demonstrated by the ability of JAM2-Fc to capture JAM2-expressing Chinese hamster ovary cells. We further showed that JAM2, but not JAM1, is capable of adhering to the HSB and HPB-ALL lymphocyte cell lines. Neutralizing mouse anti-JAM2 polyclonal antibodies provided evidence against homotypic interactions in this assay. Biotinylation of HSB cell membranes revealed a 43-kDa counter-receptor that precipitates specifically with JAM2-Fc. These characteristics of JAM2 led us to hypothesize a role for this novel protein in adhesion events associated with cardiac inflammatory conditions.

Intercellular contacts between endothelial cells are crucial for maintenance of vessel integrity. The various vascular segments display heterogeneity in their use of adherens, gap, and tight junctions (1). Contacts occur most apically at the tight junction; and in 1998, it was reported that a novel junctional adhesion molecule (JAM) 1 co-localizes with this dynamic regulated structure (2). JAM belongs to the Ig-like superfamily of adhesion molecules. The mouse JAM protein is expressed in endothelium and epithelium, whereas the human homologue is additionally found on peripheral blood leukocytes and platelets (2)(3)(4).
Pro-inflammatory stimuli promote tethering, rolling, and subsequent firm adhesion of leukocytes to the luminal vessel wall (5). These effects are mediated by the concerted action of endothelial cell selectins and the Ig adhesion molecules VCAM and ICAM (6). The molecular mechanisms underlying the subsequent diapedesis of leukocytes through the endothelial monolayer are not yet fully understood, although a key role for PECAM has been demonstrated using neutralizing antibodies (7,8). PECAM is also targeted to the lateral membranes of endothelia, but does not co-localize with junctional structures (9). Most recent evidence supports a role for JAM in leukocyte transmigration and inflammation. A neutralizing monoclonal antibody is effective at reducing spontaneous and chemokinemediated monocyte passage across endothelial monolayers (2). Furthermore, neutralization of JAM has proven effective in reducing not only monocyte, but also neutrophil transmigration across brain endothelium in a mouse model of meningitis (10). That JAM impinges on the inflammatory pathway is further substantiated by its redistribution to the endothelial cell surface in response to tumor necrosis factor-␣/interferon-␥ (11).
PECAM and JAM share some structural and functional similarities. Both adhesion proteins belong to the Ig superfamily; they localize at intercellular clefts; and neutralizing antibodies raised against either protein are effective at inhibiting the paracellular movement of leukocytes. PECAM engages in homotypic interactions in addition to heterotypic binding to the ␣ V ␤ 3 integrin and proteoglycans (12)(13)(14)(15)(16). To date, only homotypic binding has been described for JAM. Nevertheless, a counter-receptor on leukocytes is postulated since neutralizing antibodies do not affect JAM-JAM interactions (2,10).
JAM possesses two V-type immunoglobulin domains, a novel structure for an Ig superfamily adhesion protein. We reasoned that JAM would belong to a family of adhesion molecules possessing similar structure and perhaps intercellular location. Using the public expressed sequence tag data base, we have identified a related putative adhesion molecule that shows preferential expression in the heart. For comparative purposes, we also isolated the human homologue to mouse JAM. We propose that the originally identified mouse JAM and its human counterpart be referred to as JAM1. In this report, we describe the novel JAM2 sequence, chromosomal localization, and expression characteristics. Furthermore, we show that JAM2 shows selective adherence to immortalized T lymphocyte lines in vitro.

EXPERIMENTAL PROCEDURES
cDNA Cloning-JAM2 (V89915) was identified using the public expressed sequence tag (EST) data base. Sequences homologous to mouse JAM (U89915) were isolated using the tblastn program. AA406389 and AA912674 were assembled to form the 3Ј-end of human JAM2 (Fig. 1b). Rapid amplification of cDNA ends (RACE) was employed to obtain * 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  further 5Ј-sequence (Fig. 1c). For RACE-1, human placental mRNA was reverse-transcribed with avian myeloblastosis virus reverse transcriptase (CLONTECH, Palo Alto, CA) at 42°C with primer 5Ј-CCCCG-CATCACTTCTTGTCACATTTTTGATCCGG-3Ј. For RACE-2, Thermoscript (Life Technologies, Inc.) at 58°C was used with primer 5Ј-CTGCTCTGAGGAGGTCGAGGGTCCC-3Ј. Cycling was as follows: one cycle at 94°C for 30 s; five cycles at 94°C for 5 s and 72°C for 4 min; five cycles at 94°C for 5 s and 70°C for 4 min; and 25 cycles at 94°C for 5 s and 68°C for 4 min. Full-length JAM2 was constructed from two polymerase chain reaction products that encompassed an internal EcoNI restriction site. The 5Ј-product (extracellular domain) was generated with sense (5Ј-GCCGCggatccAAGATGGCGAGGAGG-3Ј) and antisense (5Ј-(GCTATTATGCCggtaccGTTGAGATCATC-3Ј) primers by amplifying with ampliTaq DNA polymerase (Applied Biosystems, Foster City, CA) as follows: one cycle at 95°C for 120 s; 35 cycles at 95°C for 20 s, 58°C for 20 s, and 72°C for 30 s; and one cycle at 72°C for 180 s. A 3Ј-product was synthesized with sense (5Ј-TAAAAATCGAGCT-GAGATGATAG-3Ј; 248 bp from ATG) and antisense (5Ј-TTAAAT-TATAAAGGATTTTGTG-3Ј) primers as follows: one cycle at 95°C for 420 s; 28 cycles at 95°C for 20 s, 56°C for 20 s, and 72°C for 30 s; and one cycle at 72°C for 300 s. To confirm each base, three independent products were sequenced (ABI sequencer, Seqwright, Houston, TX). The full open reading frame of human JAM1 was assembled from the overlapping ESTs, AA152150, T87045, T73746, and AA101561 (Fig.  1d). A polymerase chain reaction product was synthesized from human umbilical vein endothelial cells mRNA using sense (5Ј-ggatcCGCGAT-GGGGACAAAGGCGC-3Ј) and antisense (5Ј-ggtaccACCAGGAATGAC-GAGGTC-3Ј) oligonucleotides by cycling as follows: one cycle at 95°C for 7 min; 28 cycles at 95°C for 20 s, 56°C for 20 s, and 72°C for 30 s; and one cycle at 5 min for 72°C. To confirm each base, three independent products were sequenced.
Northern Analysis-An extracellular domain polymerase chain reaction product of JAM2 was random prime-labeled with [␣-32 P]dCTP and hybridized to normalized human cardiovascular specific and multiple tissue Northern blots (CLONTECH) under high stringency. Transcripts were viewed by autoradiography following exposure of the blot to Hyperfilm MP (Amersham Pharmacia Biotech) at Ϫ70°C.
Chromosomal Localization and Intron/Exon Boundaries-To identify genomic sequence, the public non-redundant data base was searched using the blastn program with JAM2 cDNA sequence. The results required minor manual modification due to dual designation of isolated bases at the end of some exon boundaries. The correct designation was based on 5Ј-and 3Ј-splice site consensus sequences. It was possible to confirm all intron/exon boundaries by retrieving identical information from more than one deposit of genomic sequence.
Expression in CHO Cells-The full-length clone of JAM2 with a C-terminal hemagglutinin tag was subcloned into pcDNA6 (Invitrogen, Carlsbad, CA). CHO-K1 cells were transfected with 10 g of vector possessing either no insert or JAM2 using FuGENE TM 6 reagent (Roche Molecular Biochemicals). Stable cells lines (control and JAM2) were selected with 5-10 g/ml blasticidin. For Western blot analysis, cells were lysed in 1% Triton X-100 buffer in the presence of protease inhibitors (protease inhibitor mixture set III, Calbiochem). Approximately 36 g of protein was electrophoresed through 10% polyacrylamide gels and probed with a 1:2000 times dilution of preimmune or anti-JAM2 polyclonal serum. Specific bands were viewed using enhanced chemiluminescence with a 1:30,000 dilution of goat anti-mouse horseradish peroxidase (Fisher).
Expression in Insect Cells-For JAM2, the extracellular domain polymerase chain reaction product was digested with BamHI/KpnI and subcloned into a pFastBac1 vector (Life Technologies, Inc.) possessing the constant region of mouse IgG2a (17). For JAM1, the ectodomain was synthesized with sense (5Ј-ggatccGCGATGGGGACAAAGGCGC-3Ј) and antisense (5Ј-GATggtaccCACATTCCGCTCC-3Ј) primers with cycling as follows: one cycle at 95°C for 120 s; 23 cycles at 95°C for 20 s, 56°C for 10 s, and 72°C for 10 s; and one cycle at 72°C for 300 s. Secreted recombinant JAM2 and JAM1 fusion proteins were purified from the media of infected Sf21 cells using HiTrap protein A columns (Amersham Pharmacia Biotech).
Antibodies-For preparation of JAM2 immunogen, thrombin cleavage was used to release JAM2 from the Fc portion of the secreted recombinant fusion protein. The digest was reapplied to the protein A column to remove Fc. Female BALB/c mice (8 week olds; Harlan Sprague Dawley, Inc., Indianapolis, IN) were immunized and then boosted three times, 28 days apart, by intraperitoneal and subcutaneous injections of 100 g of purified JAM2 extracellular domain emulsified with an equal volume of Freund's adjuvant. Complete Freund's adjuvant was used for the first immunization, and incomplete Freund's adjuvant for subsequent injections. Serum was collected 10 days following each boost.
Immunofluorescence-CHO-K1 (control or JAM2-expressing) cells grown on glass slides to confluence were fixed with 1% paraformaldehyde and stained with a 1:100 dilution of either preimmune or mouse anti-JAM2 polyclonal serum. goat anti-mouse-fluorescein isothiocyanate at 1:100 was used as a secondary antibody. Fluorescence was viewed using a Noran TM confocal laser-scanning microscope (Noran Instruments Inc., Middleton, WI) equipped with an argon laser and appropriate optics and filter module for fluorescein isothiocyanate detection. Digital images were obtained at ϫ400 using a 0.75N/A Nikon ϫ20 lens. A z axis motor attached to the inverted microscope stage was calibrated to move the plane of focus at 0.4-m steps through the sample. Collected 12-bit gray scale images at 512 ϫ 480 resolution, stored on a rewritable optical hard disc, were volumetrically reconstructed using the Image-1/Metamorph TM three-dimensional module (Universal Imaging Corp., West Chester, PA).
Adhesion Assay-Ectodomain-Fc adhesion assays were performed in 96-well plates essentially as described previously (18). Briefly, 50 l of goat anti-mouse IgG2a was coated at 5 g/ml in phosphate-buffered saline and used to capture 4.8 pmol of JAM2-Fc, JAM1-Fc, VCAM-Fc, or mouse IgG2a (control). CHO and various leukocyte cell lines (i.e. T lymphocytes, HSB and HPB-ALL; B lymphocytes, RAMOS; monocytic cells, HL-60 and THP-1; and the erythroleukemic K562 line) were labeled with calcein (Molecular Probes, Inc., Eugene, OR) at 50 g/ml for 25 min at 37°C. Cell binding was performed for 90 min at 37°C with 250,000 cells/well in binding buffer consisting of Tris-buffered saline plus 1 mM each of CaCl 2 , MgCl 2 , and MnCl 2 . Wells were washed three times and lysed with 50 mM Tris (pH 7.5), 5 mM EDTA, and 1% Nonidet P-40, and fluorescence was read in a Cytofluor with excitation at 485/20 nm and emission at 530/25 nm. Specific binding was calculated by removal of values obtained by capture of mouse IgG2a. When the various cell lines were compared, curves for calcein uptake were generated for each experiment to convert arbitrary fluorescence units into cell number.
For antibody inhibition, either JAM2-Fc protein captured on wells or HSB cells were incubated for 30 min at room temperature in binding buffer with a 1:100 dilution of preimmune (normal mouse serum) or mouse anti-JAM2 polyclonal serum. Following incubation, excess antibody was removed by washing three times prior to continuation of the assay.
Cell-surface Biotinylation/Counter-receptor Precipitation-HSB or K562 cells were surface-biotinylated using EZ-Link sulfosuccinimidobiotin (Pierce) according to the manufacturer's instructions. Cells (2.5 ϫ 10 7 /ml) were washed three times following incubation with 0.5 mg/ml sulfosuccinimidobiotin for 30 min at room temperature. Cell lysis was achieved in Tris-buffered saline (pH 7.5), 1% Triton X-100, and 1 mM each MnCl 2 , MgCl 2 , and CaCl 2 with the inclusion of protease inhibitor mixture set III. Approximately 5 g of JAM-Fc fusion protein was added to ϳ1 mg of lysate and incubated at 4°C overnight. Proteins bound to JAM were precipitated with protein A-Sepharose (30 l), boiled for 5 min with 10 mM dithiothreitol in SDS sample buffer, and separated on 9% SDS gels. Following transfer to polyvinylidene difluoride membrane, biotinylated proteins were detected using streptavidin-horseradish peroxidase (1:4000) and enhanced chemiluminescence (ECL, Amersham Pharmacia Biotech). In the extracellular region, there are three potential N-linked glycosylation sites (NX(S/T)) at amino acids 98, 187, and 236. In the short intracellular tail, at amino acid 279, a protein kinase C phosphorylation consensus site ((S/T)X(R/K)) exists. At the extreme C terminus, a potential binding motif for PDZ domains is apparent (SFII). We were unable to identify an in-frame and upstream stop codon for human JAM2 in the 5Ј-untranslated sequence. That the designated ATG is the true translational initiation signal is supported by its context within a Kozak consensus sequence and alignment with JAM1 ATG (Fig. 3).

Fig
To compare sequences of human JAM family members, we isolated human JAM1 cDNA. Prior to our report of this sequence, human JAM1 was isolated using a different approach by Ozaki et al. (11) (GenBank TM /EBI Data Bank accession number AF111713). Alignment of human JAM2 with JAM1 revealed 44% similarity and 35% identity at the amino acid level (Fig. 3). Similarly, alignment with mouse JAM1 amino acid sequence showed 43% similarity and 35% identity (data not shown). The glycosylation motif at amino acid 187 is conserved between family members. However, JAM1 possesses a cAMP/cGMP consensus site in the intracellular domain that is absent in JAM2.  JAM2 was mapped to chromosome 21 at position q21.2 using the public data base. Sequence was retrieved at 100% identity from two contiguous non-overlapping sequences of 100,000 bp each (GenBank TM /EBI Data Bank accession numbers AP000087.1 and AP000086.1). The coding region of JAM2, which constitutes 897 bp, is distributed over 10 exons (Table I).
The limits of the JAM2 cDNA sequence shown in Fig. 2 span ϳ74,853 bp of genomic DNA. Various exons were also found in AP000223 (coding exon 1), AP000225 (coding exons 2-6), and AP000226 (coding exons 6 -10). Since the complete JAM2 transcript(s) is considerably larger than 897 bp (Fig. 4), further exons in the untranslated regions remain to be identified either up and/or downstream. All intron/exon boundaries conform to the consensus GT/AG rule (19).
Tissue expression of JAM2 was examined on normalized human multiple tissue Northern blots (Fig. 4A). Two transcripts of ϳ1.5 and Ͻ4.4 kilobases were apparent under high stringency. These two species likely represent alternatively spliced products. Of the 12 tissues examined, human JAM2 appears to be preferably expressed in the heart. However, the transcript could also be detected to some degree in several other tissues, most notably the placenta. Fig. 4B shows a more detailed examination of the JAM2 transcript in the heart. A clear chamber-specific expression was not apparent. Relative to glyceraldehyde-3-phosphate dehydrogenase, there was somewhat lower expression in fetal heart. However, major differences in the aorta, atrium, and ventricles were not observed.
To facilitate study of this protein, we generated a JAM2overexpressing CHO cell line and a mouse anti-JAM2 polyclonal serum raised against the ectodomain. Western blotting of CHO cell lysate estimated a 48-kDa protein (Fig. 5). This is ϳ14 kDa larger than the size predicted from the peptide sequence. Glycosylation of JAM2 on at least one of its three N-linked glycosylation consensus sites could explain this phenomenon.
For cellular localization, we stained confluent monolayers of both control and JAM2-expressing CHO cells with the anti-JAM2 antiserum. Fig. 6 shows that JAM2 mainly partitioned to sites of cell-cell contact, although some surface membrane fluorescence was observed. The border pattern of staining was identical to those shown by mouse JAM1 expressed in CHO cells and by endogenous human JAM1 expressed in human umbilical vein endothelial cells. For JAM1, this phenomenon contributed to the hypothesis that the extracellular domain forms homotypic interactions between cells.
To address the possibility of JAM2 homotypic binding, we utilized the recombinant ectodomain and tested its ability to capture the JAM2-expressing CHO cell line. Adhesion of calcein-loaded cells was performed under static conditions according to a previously established and well documented in vitro binding assay (18). In the same series of experiments, we examined JAM1 homotypic interactions and asked whether JAM2 could form heterotypic interactions with JAM1. Fig. 7 shows that JAM2 was clearly capable of capturing CHO cells expressing JAM2, but not control cells or those expressing JAM1. Furthermore, JAM1 was unable to maintain homotypic interactions under these conditions.
It has been demonstrated that JAM1 participates in leukocyte migration. Reasoning that JAM2 may also impinge on leukocyte adhesion and/or migration, we examined the capacity of the JAM2 extracellular domain to adhere to various leukocyte cell lines. Calcein-loaded cells were allowed to interact with JAM2-Fc captured in 96-well plates. Nonspecific binding of cells to captured mouse IgG2a was determined simultaneously and subtracted. JAM1-Fc was incorporated into the assay for comparative purposes. VCAM-Fc was included as a positive control for all cell lines with the exception of K562, which does not express VCAM counter-receptors. Fig. 8 shows that JAM2-Fc was able to capture the T lymphocyte cell lines HSB and HPB-ALL quite efficiently. The binding of these cells to IgG2a was always Ͻ2% of the binding to JAM2. In contrast, interactions with B lymphocytes (RAMOS), monocytic cells (HL-60 and THP-1), and the erythroleukemic K562 cell line were negligible. Most interestingly, JAM1-Fc was unable to adhere to any of these cell lines.
As demonstrated, JAM2 is capable of homotypic interactions. Thus, we next addressed whether the JAM2 ectodomain bound HSB cells through this mechanism. Fig. 4A shows that, unlike JAM1, JAM2 did not show expression in peripheral blood leukocytes. Nevertheless, to verify lack of expression in HSB cells, we used the mouse polyclonal serum to probe for JAM2 protein expression by Western blotting. No protein was detected (Fig.  5). As further proof, we compared the surface JAM2 expression levels using the following more sensitive test. The HSB and control and JAM2-expressing CHO cells were loaded with calcein and incubated with either normal mouse serum or anti-JAM2 serum. Cell surface-bound anti-JAM2 antibody was detected by cell capture in 96-well plates coated with goat antimouse secondary antibodies. Table II shows that although the anti-JAM2 serum was very effective at capturing CHO cells expressing the JAM2 protein, no HSB cell binding was apparent.
To extend these studies, we tested the ability of the mouse anti-JAM2 serum to neutralize HSB binding to recombinant JAM2. Antibody was used to block epitopes on recombinant JAM2 captured on 96-well plates. Table III shows that although preimmune serum was ineffective, anti-JAM2 serum successfully prevented HSB binding. Since relatively high levels of JAM2 were coated on these wells, we were confident that if low levels were expressed on HSB cells, the antibody should be capable of producing inhibition when incubated directly with HSB cells. As predicted, under this experimental setup, the anti-JAM2 antibody was unable to inhibit HSB interactions with recombinant JAM2.
The studies thus far led us to postulate that HSB cells express a counter-receptor for JAM2. To strengthen this hypothesis and to gain a preliminary characterization of the protein, we performed precipitation experiments using JAM2-Fc. HSB cells were surface-biotinylated, washed, lysed, and incubated with JAM2-Fc. Bound proteins were precipitated using protein A and viewed on Western blots with avidin-horseradish peroxidase. Fig. 9 reveals that JAM2 could indeed specifically capture a surface protein from HSB cells of ϳ43 kDa. This band was not apparent in surface-biotinylated K562 cells, in agreement with the cell adhesion studies described above. Furthermore, JAM1-Fc, which was unable to bind calcein-loaded HSB cells, did not precipitate this protein. DISCUSSION We describe a novel adhesion molecule that is most similar in primary sequence to the junctional adhesion molecule (2,11). It possesses two immunoglobulin-like domains, and its amino acid sequence displays 35% identity to human JAM1. JAM2 transcripts show preferential expression in the heart. Although JAM1 is also found in the heart, expression is equally high for this protein in the lung, kidney (2), and liver. 2 At the cellular level, JAM1 is found in epithelia, endothelia, leukocytes, and platelets (2)(3)(4)20). Using our current antiserum, we were unable to characterize exactly which cells in the heart expressed JAM2. However, by Northern blotting, we were able to demonstrate lack of expression in peripheral blood leukocytes. While this manuscript was under review, a sequence identical to JAM2 was cloned, and its expression was localized exclusively to endothelial cells (21). Whether JAM1 and JAM2 show differing or overlapping patterns of expression within endothelial cells of the heart remains to be determined. JAM2 localizes to the long arm of chromosome 21 at 21q21.2. This band also contains the amyloid precursor protein gene. The latter is responsible for the accumulation of ␤-amyloid in the brains of Alzheimer's and Down's syndrome patients (22,23). It will be interesting to determine whether a "gene dosage effect" of JAM2, as may occur in trisomy and monosomy 21, impinges on the pathophysiology of affected individuals. Although Down's syndrome is associated with congenital cardiac defects, the genes likely responsible have been mapped to 21q22 (24,25).
We have raised a mouse polyclonal serum against the JAM2 extracellular domain. It has proven useful for immunolocalization of JAM2 when overexpressed in CHO cells. We found that JAM2 partitions to the intercellular membranes of these cells, a pattern identical to that shown by JAM1 (2). Further analysis is required to determine whether JAM2 is also confined to tight junctions like endothelial/epithelial JAM1 or localizes below the apical junction like PECAM (9).
Proteins containing PDZ domains are predominantly localized to the plasma membrane and are recruited to specialized sites of cell-cell contact (26). Many PDZ domains mediate protein-protein interactions by interacting with short consensus motifs found at the free carboxyl terminus of transmembrane proteins (27). The intracellular tails of both JAM2 and JAM1 possess an SFII or SFLV sequence, respectively, which can be predicted to interact with PDZ domains (26). Most recently, it has been reported that the intracellular domain of JAM1 binds to the tight junction-associated proteins ZO-1 and AF-6 via their PDZ domains (28,29). Thus, it is highly likely that JAM2 will display similar binding activities.
Due to its intercellular location and ability to confer reduced paracellular permeability in CHO cells, mouse JAM1 is postulated to form homotypic interactions (2). However, although JAM1 homotypic binding may occur between ectodomains expressed on opposing epithelial/endothelial cells at tight junctions, this characteristic is lost when cells are detached (2,3). In this work, we present direct evidence for JAM2 homotypic binding by capture of JAM2-expressing CHO cells with JAM2-Fc. On the other hand, the lack of adhesion between the JAM1 ectodomain and JAM1-expressing CHO cells is in keeping with the results of former studies (2,3). It has been suggested that the JAM1 homotypic interaction is of low affinity, with avidity being increased upon JAM1 clustering at tight junctions. Additionally, JAM1-JAM1 adhesion may be susceptible to changes in complex association that could occur between the  cytoplasmic tail and other intracellular proteins following cell detachment. A similar phenomenon occurs for VE-cadherin, which is unable to promote aggregation of detached cells (30). Thus, either JAM2 ectodomains engage in higher affinity homotypic interactions, or they are less dependent upon cell attachment. Although our current data suggest that JAM2 does not form heterotypic interactions with JAM1, it is possible that the adhesion may be too weak for detection in this assay or reliant upon the factors discussed above.
Many adhesion proteins belonging to the immunoglobulin superfamily are able to engage with leukocyte counter-receptors. Furthermore, interactions are often studied using ectodomain-Fc fusions. Indeed, the extracellular domains of some adhesion proteins, e.g. selectins, VCAM, and ICAM, are released under certain conditions in vivo and correlate with disease (31)(32)(33). Using JAM2-Fc, we demonstrated that it is capable of adhering to the HSB and HPB-ALL T-cell lines. Although we were not able to detect interactions with other cell types, e.g. monocytic, adhesion specificity may not necessarily be specific to T-cells. A complete characterization of JAM2 binding to peripheral blood leukocytes under control and activated conditions is required to address this question. The adhesion of JAM1 to leukocytes has not yet been reported, and we were unable to demonstrate an interaction in this study.
Despite the ability of recombinant JAM2 to capture HSB cells, we were unable to adhere HSB cells to CHO cell monolayers expressing JAM2. This is likely due to the intercellular location of JAM2. In an attempt to render JAM2-binding epitopes more accessible, we tried alternative methods employing immobilized HSB cells with detached CHO cells. These efforts were also unsuccessful. The most likely explanation for these data is that the heterotypic interaction requires the clustering of JAM2, which would occur at cell-cell junctions, to increase the avidity of the adhesion. This would be lost in detached cells, but mimicked by capture of the JAM2 ectodomain to a high density on 96-well plates.
To further verify that JAM2 specifically interacts with HSB cells, we sought to identify a binding partner for JAM2 on HSB cell membranes. Our data demonstrate that a protein of ϳ43 kDa specifically coprecipitates with the JAM2 ectodomain. The immunoglobulin folds of certain adhesion proteins are able to recognize epitopes on integrins (reviewed in Ref. 34). However, the JAM2 counter-receptor appears to constitute a single species that is considerably smaller than either of the heterodimeric ␣/␤ integrin subunits. Adhesive interactions between immunoglobulin-like folds are well documented, examples being CD28/CD152 recognition of CD80/CD86 and CD2 binding to CD48/CD58 (35,36). It is possible that the 43-kDa protein may encode another member of the immunoglobulin superfamily. We are actively pursing the identity of this counter-receptor to determine its primary amino acid sequence. This will facilitate study of its expression characteristics and allow the physiological role of the JAM2 interaction to be more easily addressed.
There is now convincing evidence showing the involvement of mouse JAM1 in leukocyte transmigration during inflammation in vivo. Our current results, along with the sequence similarities of the two proteins, predict some role for JAM2 in the inflammatory process. We would speculate that JAM1 and JAM2 have distinct rather than redundant roles in the heart since the JAM1 ectodomain appears unable to share the JAM2 counter-receptor. It is noteworthy that PECAM expression is also preponderant in the heart in addition to the lung and kidney (37). These molecules may collaborate during the inflammatory process by contributing discrete functions.
JAMs may play a role in cardiac inflammatory conditions such as occur during myocarditis and allograft rejection. They may also contribute to the pathology of chronic ischemic heart disease and dilated cardiomyopathy. Whether JAM2 supports leukocyte migration and the identity of the HSB counter-receptor are key questions that will dictate our future studies.