JAM-2, a novel immunoglobulin superfamily molecule, expressed by endothelial and lymphatic cells.

Cell-cell contacts are essential for morphogenesis and tissue function and play a vital role in mediating endothelial cohesion within the vascular system during vessel growth and organization. We identified a novel junctional adhesion molecule, named JAM-2, by a selective RNA display method, which allowed identification of transcripts encoding immunoglobulin superfamily molecules regulated during coculture of endothelial cells with tumor cells. The JAM-2 transcript is highly expressed during embryogenesis and is detected in lymph node and Peyer's patches RNA of adult mice. Accordingly, antibodies specific for JAM-2 stain high endothelial venules and lymphatic vessels in lymphoid organs, and vascular structures in the kidney. Using real time video microscopy, we show that JAM-2 is localized within minutes to the newly formed cell-cell contact. The role of the protein in the sealing of cell-cell contact is further suggested by the reduced paracellular permeability of cell monolayer transfected with JAM-2 cDNA, and by the localization of JAM-2 to tight junctional complexes of polarized cells. Taken together, our results suggest that JAM-2 is a novel vascular molecule, which participates in interendothelial junctional complexes.

Cell-cell contacts are essential for morphogenesis and tissue function and play a vital role in mediating endothelial cohesion within the vascular system during vessel growth and organization. We identified a novel junctional adhesion molecule, named JAM-2, by a selective RNA display method, which allowed identification of transcripts encoding immunoglobulin superfamily molecules regulated during coculture of endothelial cells with tumor cells. The JAM-2 transcript is highly expressed during embryogenesis and is detected in lymph node and Peyer's patches RNA of adult mice. Accordingly, antibodies specific for JAM-2 stain high endothelial venules and lymphatic vessels in lymphoid organs, and vascular structures in the kidney. Using real time video microscopy, we show that JAM-2 is localized within minutes to the newly formed cell-cell contact. The role of the protein in the sealing of cell-cell contact is further suggested by the reduced paracellular permeability of cell monolayer transfected with JAM-2 cDNA, and by the localization of JAM-2 to tight junctional complexes of polarized cells. Taken together, our results suggest that JAM-2 is a novel vascular molecule, which participates in interendothelial junctional complexes.
Adhesion molecules play essential roles in the overall tissue organization and the proper physiological function of organs by establishing and organizing cell-cell contacts (1). The function of endothelial cells lining the vessel walls relies in part on their ability to express different adhesion molecules in a coordinated and regulated way. It has been demonstrated that endothelial cells regulate vascular permeability and leukocyte emigration from the blood into the surrounding tissues. Leukocyte emigration occurs in a multistep adhesion process involving different classes of adhesion molecules, namely selectins, integrins, and cell adhesion molecules of the immunoglobulin superfamily (Ig Sf) 1 (2)(3)(4). The members of the Ig Sf possess structural domains similar to the variable (V type) or constant (C type) immuno-globulin domains found in T or B cell receptor (5,6). JAM (hereafter referred to as JAM-1), a recently described immunoglobulin superfamily molecule with two V H domains, was shown to regulate monocyte transmigration across endothelial cell layers (7,8). The protein specifically localizes to tight junctions of epithelial and endothelial cells, indicating that molecules participating in intercellular junctions may regulate vascular functions (9,10). To date, the cell surface molecules known to participate specifically in tight junctions include JAM-1, occludin, and the claudins (8,11,12). JAM-1 and occludin are expressed by both endothelial and epithelial cells, while some members of the claudin family display tissue-specific distribution (7,13,14). Interestingly, claudin-5 was found in the tight junctions of endothelial cells (15), whereas claudin-11 was present in tight junctions of myelin sheaths in the brain and sertoli cells in the testis (16), reflecting some heterogeneity in the molecular composition of tight junctions found in different cellular system (17). This suggests that the function of tight junctions and their regulation may involve specific expression of junctional proteins or distinct activation events, which regulate the stability of tight junctions (18).
Hence, we started to search for novel surface molecules of the Ig Sf, which may be more specific for vascular endothelium than JAM-1. We took advantage of the fact that blood vessels in tumors are often leaky, indicating that the barrier function of angiogenic endothelial cells may be lost, due to regulation of tight junctional molecules in endothelial cells growing in the presence of tumor cells (19,20). Our experimental approach consisted in the coculture of endothelial cell lines with melanoma cells to identify regulated transcripts by differential screening (21). For this purpose, we developed a method based on the principle of RNA display that allowed the preferential identification of transcripts encoding molecules of the Ig Sf (22,23). This method led to the identification of JAM-2, which encoded for a transmembrane protein specifically incorporated in tight junctions, with a structural organization into V-C 2 domains. The protein is not found on epithelial cells and is expressed by endothelial and lymphatic cells in vivo. Our results suggest that JAM-2, together with JAM-1, may constitute the prototypes of a novel junctional adhesion molecule family. rine SV40 transformed lymph node endothelial cell line TME was provided by Dr. A. Hamann (25). The murine squamous cell carcinoma KLN 205, CHO cells, MDCK cells, and myeloma cell line Sp2/0 were obtained from the American Type Tissue Culture Collection (ATCC). All cells, except CHO, were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc., Paisley, Scotland), supplemented with 10% fetal calf serum (PAA Laboratories, Linz, Austria), 2 mM glutamine, 100 units/ml penicillin, and 100 units/ml streptomycin (all from Life Technologies, Inc.). CHO cells were grown in nutrient mixture F-12 (Ham's) medium supplemented as above. Adherent cells were detached by washing with PBS and 0.15 mM EDTA, followed by a 5-min incubation in trypsin/EDTA at 37°C.
Display, Cloning, and Sequence Analysis-For co-culture experiments, 5 ϫ 10 5 tEnd.1 cells were grown together with 2.5 ϫ 10 4 B16 F10 melanoma cells for 64 h in 10-cm tissue culture dishes. As control, 5 ϫ 10 5 tEnd.1 and 2.5 ϫ 10 5 B16 F10 cells were grown separately under the same conditions, resulting in confluent monolayers after 64 h. Total RNA was directly extracted in Petri dishes with Trizol reagent following the manufacturer's instructions (Life Technologies, Inc.). The cDNA was prepared from 5 g of total RNA, employing oligo(dT) (16-mer) primer and Superscript reverse transcriptase (Life Technologies, Inc.). The quality and the quantity of cDNA were checked by running 27 cycles of PCR on 1 l of cDNA diluted 1:5, using primers specific for the housekeeping hypoxanthine phosphoribosyltransferase cDNA. We then performed the differential PCR with the following degenerated primers: 5Ј-TAYAGNTGYNNNGCYTCYAA-3Ј, 5Ј-TAYCRGTGYNNNGCYTCY-AA-3Ј, and 5Ј-TAYTAYTGYNNNGCYTCYAA-3Ј encoding for the most frequent amino acid sequences encountered in C 2 domains: YRCXAS, YQCXAS, and YYCXAS. The PCR conditions were as follows: 2 l of diluted cDNA, 2.5 l of 10ϫ Goldstar PCR buffer, 2 l of MgCl 2 , 2 l of degenerated primers (0.3 mM), 0.5 l of dNTP (0.1 mM), 0.1 l of [␣-33 P]dATP (10 mCi/ml; Amersham Pharmacia Biotech, Dü bendorf, Switzerland), 15.65 l of H 2 O, 0.25 l of Goldstar Taq polymerase (Eurogentech, Seraing, Belgium). The parameters for the PCR were as follows: 45 s at 94°C, 90 s at 50°C, and 45 s at 72°C for 40 cycles. Formamide/EDTA loading buffer was added, and samples were denatured for 2 min at 94°C. The PCR products were then separated on a 6% polyacrylamide gel and autoradiographed using Kodak OM-Mat. Band intensity was compared, differentially expressed bands were cut from the dried polyacrylamide gel, and fragments were retrieved by boiling and ethanol precipitation as described previously (22). The PCR products were then reamplified using increased concentrations of dNTPs (0.2 mM instead of 2 M) without [ 33 P]ATP. The products of re-amplification were cloned into pGem-T Easy Vector (Promega Corp., Wallisellen, Switzerland). Nucleic acid sequences of two independent clones were determined using the Thermo Sequence fluorescence-labeled primer cycle sequencing kit (Amersham Pharmacia Biotech) and the LI-COR DNA analysis system (MWG-Biotech GmbH, Ebersberg, Germany).
The QD10 PCR product was identical to three murine ESTs encoding for a new putative Ig superfamily molecule (accession no. AA726206, AA052463, and AA175925). All ESTs were incomplete, and the remaining 5Ј-coding sequence was obtained by rapid amplification of cDNA ends using three primers designed on the basis of the EST sequences (5Ј-rapid amplification of cDNA ends PCR system, Life Technologies, Inc.). The full-length coding sequence for JAM-2 was assembled in the pGemt vector (Promega Corp.) from the cloned 5Ј-rapid amplification of cDNA ends PCR product and the EST (accession no. AA726206) using the internal HpaI restriction site. The cDNA encoding JAM-1 was kindly provided by Dr. P. Naquet (CIML, Marseille-Luminy, France). The nucleic acid sequences were determined using the Thermo Sequence fluorescent labeled primer cycle sequencing kit (Amersham Pharmacia Biotech) and the LI-COR DNA analysis system (MWG-Biotech). Further sequence analyses were performed via the applications available on the ExPASy Molecular Biology Server (Blast, Prosite, Swiss-Prot, and signal peptide prediction).
Northern Blot-Total mRNA from cells or murine tissues was extracted using Trizol (Life Technologies AG, Basel, Switzerland) according to manufacturer's instructions. Poly(A) mRNA was extracted from 250 g of total RNA with the Oligotex mRNA purification kit (Qiagen, Zurich, Switzerland). Embryonic poly(A) Northern blot was purchased from CLONTECH (P. H Stehelin and Cie AG, Basel, Switzerland). The riboprobes were prepared from pcDNA3 vector (Invitrogen, Leek, Netherlands) and comprised the sequences encoding for the immunoglobulin domains of JAM-1 and JAM-2, or the full-length coding sequence for ␤-actin. Hybridization was performed at 62°C in buffer containing 50% formamide. The blots were washed twice (0.5ϫ SSC, 0.1% SDS, 67°C) and autoradiographed on Kodak X-Omat at Ϫ80°C.
Construction of Expression Vectors-The sequence encoding EGFP was subcloned from pEGFP-1 vector (CLONTECH, P. H. Stehelin and Cie AG) into pcDNA3 using HindIII and NotI sites, therefore named pcDNA3/EGFP. The 3Ј restriction sites, HpaI and ScaI, found in the sequence encoding, respectively, the cytoplasmic domain of JAM-2 and JAM-1, were used to fuse the two sequences at the N terminus of the EGFP in pcDNA3 vector (Invitrogen). The inserts encoding JAM-2 or JAM-1 were excised from pGemt or pRc/CMV using SacII/HpaI or HindIII/ScaI digestions, respectively. The coding sequences were then cloned in pcDNA3/EGFP vector digested with AgeI, blunted by fill-in, and further digested with HindIII or SacII enzymes. This resulted in fusion sites at amino acid positions DGV 291 for JAM-2 and QPS 285 for JAM-1. The transfection of CHO cells was performed as described previously (26). Stable transfectants used for permeability assays were selected by growing transfected CHO cells for 2 weeks in medium containing 1 mg/ml G418. Resistant colonies were isolated and checked for EGFP fluorescence intensity by flow cytometry (FACScalibur apparatus; Becton Dickinson, Mountain View, CA) and fluorescence localization by microscopy (Axiovert; Zeiss, Oberkochen, Germany). Timelapse video microscopy was performed using an Axiovert fluorescence microscope and Openlab software for image acquisition. To produce soluble molecules, the sequence encoding extracellular domains of JAM-1 or JAM-2 were amplified by PCR from plasmids containing the full-length coding sequences using, respectively, T7 and 5Ј-gctctagacacagcatccatgtgtgcagcctc-3Ј for JAM-1, or T7 and 5Ј-gctctagaatagacttccatgtcctgcc-3Ј for JAM-2. The reverse primers were modified by XbaI site to clone the PCR products in frame with Flag Tag sequence in pcDNA3 vector, as described previously (27). The absence of PCR-induced errors was confirmed by sequencing the constructs on both strands. 293T cells were then transfected using calcium phosphate precipitation method, and supernatants were collected every 2 days over a period of 10 days. Isolation of the recombinant soluble molecules (namely sJAM-1 or sJAM-2) was achieved on a M2 column (Sigma-Aldrich Co., FlukaChemie AG, Buchs, Switzerland), and purity was checked by SDS-PAGE and Coomassie Blue staining (data not shown).
Reagents, Enzyme-linked Immunosorbent Assay, and Immunofluorescence Analysis-The following monoclonal antibodies were used: anti-PECAM (GC51, rat IgG 2a ; EA-3, rat IgG 1 ), anti-JAM-1 (H2O2.106.7.4, rat IgG 1 ; (7, 28), anti-occludin (Zymed Laboratories Inc., Gebr Mä chler AG, Basel, Switzerland), anti-ZO-1 (R40/76; Ref. 29), and anti-E-cadherin (Arc-1; Ref. 30). The panel of CRAM antibodies against murine JAM-2 was generated in the laboratory using standard techniques, and recombinant soluble molecule as immunogen in rats (31). Briefly, 100 g of purified recombinant soluble JAM-2 molecule (sJAM-2) mixed with Titermax adjuvant (Sigma) was used to immunize male Fischer rats intraperiteonally. Two days after a final intravenous injection of 50 g of sJAM-2, splenocytes were fused to Sp2/0 cells, and hybridoma were selected in HAT-containing medium. Resistant clones were screened by ELISA for the production of mAbs recognizing specifically sJAM-2. For this purpose, Maxisorb Immunoplates (Nunc) were coated overnight at 4°C with M2 antibody diluted at 2 g/ml in 150 mM Nacl, 50 mM borate buffer, pH ϭ 9. Wells were washed, blocked for 1 h with serum-containing medium, and incubated for 1 h with supernatants of transfected 293T cells. After three washes with PBS plus 0.2% bovine serum albumin, hybridoma supernatants were added to the wells and incubated for 1 h at 4°C. After washing, bound antibodies were detected using mouse anti-rat peroxidase (Jackson Immunoresearch, Milan AG, La Roche, Switzerland), and ABTS (Sigma). Optical densities at 405 nm were read using a kinetic microplate reader and SoftMAXPro software (Molecular Devices Corp). Positive clones were subcloned twice, rescreened, and tested by cytofluorimetry on CHO cells stably transfected with plasmids encoding EGFP alone (mock), JAM-1-EGFP, or JAM-2-EGFP. All CRAM antibodies were of IgG 1 or IgG 2a isotype except CRAM-25F24, which is of the IgG 2b subclass. Antibodies were purified on protein G-Sepharose columns (Amersham Pharmacia Biotech) according to the manufacturer instructions. CRAM-19H36 mAb was used for immunoprecipitation and CRAM-18F26 for immunohistochemistry. Immunofluorescence analysis was performed using secondary reagents coupled to FITC or Texas Red (Jackson Immunoresearch) for cytofluorimetry and immunohistochemistry, respectively. For immunohistochemistry, samples were fixed 5 min with cooled (Ϫ20°C) methanol. Samples were rehydrated in PBS, 0.2% gelatin, 0.05% Tween 20, incubated overnight with the primary antibodies before washing, and revealed with the appropriate secondary reagent coupled to Texas Red. For the analysis of fresh endothelial cells, dissociation of freshly dissected murine tissues was performed using collagenase/dispase digestion, according to established procedures (32). The dissociated cells were stained for 2 h at 37°C with diI-acetylated LDL (Molecular Probe Europe BV, Leiden, Netherlands) before staining with mAbs to JAM-1 or JAM-2, and goat anti-rat FITC probe. After this step, normal rat serum diluted 1/50 was added for washes and incubations. After wash, cells were stained with biotinylated anti-CD31 (PharMingen) and streptavidin red 670 (Life Technologies AG). JAM-1 or JAM-2 expression was analyzed on cells positive for the two endothelial cell markers: acetylated LDL (FL-2) and CD31 (FL-3). Negative controls were obtained by omitting primary antibody.
Immunoprecipitations-Immunoprecipitations were performed as described previously (33) using 10 mM Tris-HCl buffer, pH 7.4, 150 mM NaCl, 0.5% Triton X-100, 0.5% Nonidet P-40, protease inhibitor mixture (Roche Diagnostics Ltd, Rotkreuz, Switzerland) for lysis. After immunoprecipitation, SDS-PAGE, and transfer to nitrocellulose membrane, the biotinylated proteins were revealed using streptavidin coupled to peroxidase (Jackson Immunoresearch) and ECL (Amersham Pharmacia Biotech). Deglycosylations were performed after four washes of the beads in lysis buffer. Immunoprecipitates were incubated with 10 l of PBS, 0.5% SDS and boiled for 3 min to ensure denaturation of the proteins before deglycosylation. Thereafter, 90 l of PBS, 10 mM EDTA, 0.5% Triton X-100, protease inhibitors (Complete, Roche), and 10 units of N-glycosidase F (Roche) were added and incubated overnight at 37°C. Reaction was stopped by adding loading buffer, samples were submitted to SDS-PAGE, blotted, and visualized with streptavidin peroxidase.
Permeability Assays-Permeability was measured using Transwell chambers (6.5-mm diameter; PC filters, 0.4-m pore size, Costar Corp). In brief, 1 ϫ 10 4 transfected or nontransfected CHO cells were cultured to confluence on filters previously coated for 30 min with 0.2% gelatin. After 5 days, the medium was changed for prewarmed nutrient/F-12 medium without fetal calf serum (500 l in the lower chamber and 200 l in the upper chamber). FITC-dextran (M r 38,900, Sigma) was added in the upper chamber at 1 mg/ml final concentration. After 1 h, chambers were removed and fluorescence was read directly in the lower chamber using Cytofluor II. The mean fluorescence intensity of five independent chambers was calculated and compared using Statview software and t test unpaired comparisons. To normalize experiments, the value of mean fluorescence intensity obtained with wild type CHO cells was taken as 100%.

RESULTS
Identification of JAM-2 by Selective RNA Display-During the coculture of endothelial cells with tumor cells, endothelial properties are modified by transcriptional regulations of adhesion molecules (21). Such a model is therefore ideal to search for novel regulated adhesion molecules by differential screening method. Due to the lack of specificity of RNA display, we modified the technique to selectively identify transcripts encoding adhesion molecules of the Ig Sf. For this purpose, we used degenerated primers targeting the sequences encoding molecules with C 2 domains. This was achieved by the alignment of C 2 domains of several Ig Sf adhesion molecules (data not shown) and the identification of a linear amino acid consensus, surrounding the cysteine residue participating to the C 2 domain structure: Y-(RQYS)-C-X-A-S-N-X 2 -G. Therefore, we used the reverse translation of the most frequent consensus sequences (YRCXAS, YQCXAS, YYCXAS) to design the degenerated primers used for differential display. The level of degeneracy, between 2048 and 4096 different forms, was sufficient to obtain discrete PCR products without the use of a second reverse primer. The display of one representative radioactive reverse transcription-PCR with three different degenerated primers is shown on Fig. 1A. The comparison of the radioactive PCR products obtained from cDNA prepared from the endothelial cell line alone (lane 1), from the mix of endothelial and tumor cell lines (lane 2), or from the tumor cell line alone (lane 3), allowed the selection of four differentially expressed transcripts. After cloning and sequencing the four PCR products, it appeared that RU9 and QU11 encoded respectively for PAI-1 and RalGDS-like factors. The sequences of the two remaining products, QU14 and QD10, were unknown, but were good candidates as coding for partial sequences of novel Ig Sf transcripts since they presented a partial open reading frame containing the canonical sequence of C 2 domain, YYCXAS. Furthermore, the QD10 sequence showed the C 2 domain canonical residues Asn and Gly at positions ϩ4 and ϩ7 relative to the targeted cysteine.
Sequence comparison of the QD10 PCR product with sequence data bases revealed several identical sequences in the murine EST data base. Unfortunately, none of these comprised the initiating methionine, and so the complete coding 5Ј region was obtained by using a rapid amplification of cDNA ends approach. As shown in Fig. 1B, the cDNA sequence contained an open reading frame encoding a protein of 310 amino acids of a predicted molecular mass of 34.8 kDa for the core protein.
The initiating methionine was followed by a signal peptide (underlined) that may be cleaved between Ala 31 and Val 32 (34). The protein structure was typical of a type I protein comprising one membrane-distal V H and one proximal C 2 immunoglobulin domain. As expected, the sequence of the initial QD10 PCR product encoded for the end of the C 2 domain (italic nucleic acid sequence). Two putative N-glycosylation (Asn 104 and Asn 192 ) and four putative cytoplasmic phosphorylation sites (Ser 274 , Ser 281 , Tyr 293 , and Ser 297 ) were identified. Comparison of the protein sequence with nonredundant sequence data bases identified the tight junctional protein, JAM-1 (8), as the closest related sequence with 31% identity at the protein level (Fig.  1C). The transcript corresponding to the QD10 PCR product was therefore named JAM-2. The similarities in the extracellular part of the molecules were largely confined to the residues participating in the structure of the Ig domains (5), and to sequences surrounding the cysteines, the tryptophan residues of c strands, or the N-glycosylation site within the C 2 domain (NSSF 195 ). Interestingly, two additional cysteines are found within the C 2 domain of JAM-2, while they are absent from the JAM-1 sequence. In the cytoplasmic domain, three stretches of homology are identified: the AYRRGYF motif, the SPGK sequence, and the C-terminal parts of the proteins.

JAM-2 Is Highly Expressed during Embryogenesis, and Restricted to Endothelial Cell Subpopulations in Adult Tissues-
The tissue distribution of JAM-2 transcript was explored by Northern blotting and compared with JAM-1 (Fig. 2). The JAM-2 transcript was 2 kilobases long, highly expressed in embryonic tissue and in Peyer's patches, lymph nodes, kidney, and testis of adult animals. A putative splice variant of 1.8 kilobases was detected in testis. The expression of the JAM-2 transcript was low in lung, liver, spleen, and thymus. The relative abundance of JAM-1 and JAM-2 were compared during embryogenesis; the mRNA encoding JAM-2 was detectable as early as day 7.5 post coitum, whereas JAM-1 mRNA was not detected during embryogenesis. This suggested that the JAM-2 transcript might be abundantly expressed in tissues undergoing remodeling with restricted expression in specialized compartments in adult tissues.
To analyze the expression of JAM-2 in more details, we produced a panel of monoclonal antibodies (CRAM) against a soluble form of JAM-2 (sJAM-2). The specificity of CRAM mAbs was addressed by performing ELISA on soluble forms of JAM-1 and JAM-2. All CRAM mAbs recognized specifically sJAM-2 (Fig. 3A). To further confirm the specificity of the selected mAbs, cytofluorimetric analysis was performed on CHO cells stably transfected with EGFP or chimeric EGFP fusion proteins of JAM-1 or JAM-2 (Fig. 3B). These results confirmed the specificity of the mAbs for JAM-2 (CRAM panel), and the absence of cross-reactivity on JAM-1 transfected cells, which were specifically recognized by the already described H2O2-106 mAb (7). Therefore, we performed immunohistological analysis of JAM-2 expression on kidney and mesenteric lymph node sections (Fig. 4), using the CRAM-18F26 mAb. In the cortical region of the kidney, a weak staining of intertubular structures was observed with CRAM-18F26 in areas brightly stained with anti-PECAM (GC51), whereas anti-ZO-1 or anti-JAM-1 stained predominantly the tubular epithelial cells (data not shown). We therefore focused our attention on the medulla of the kidney and found a staining of sparse linear structures with CRAM-18F26 mAb against JAM-2. To ask whether this staining concerned vascular structures, we performed serial sections and identified the vessels by their PECAM expression (Fig. 4d). On the identical region of serial sections, linear interendothelial staining was detected using antibodies to JAM-2, JAM-1, or ZO-1 (Fig. 4, a, b, and c, respectively). On mesenteric lymph node sections, the CRAM-18F26 mAb stained high endothelial venules (HEV) (Fig. 4e). The HEVs also expressed JAM-1, ZO-1, and PECAM (Fig. 4, f, g, and h), although with differences in the subcellular localization of the stainings (Fig. 4,  e-h, insets). In the cortical area of the mesenteric lymph nodes, all antibodies labeled cells in subcapsular sinuses (Fig. 4, i-l), which corresponded to cells of afferent lymphatic vessels. Thus, the staining with anti-JAM-2 mAb seemed to be restricted to lymphatic vessels and a subset of endothelial cells.
Although the staining with anti-JAM-2 antibody appeared to be endothelial, we wished to confirm this endothelial specific expression of JAM-2. For this purpose, we performed cytofluorimetric analysis of JAM-2 expression on various cell lines or on freshly isolated endothelial cells from dissociated tissues using CRAM-4H31 mAb. Identical results were obtained with the different anti-JAM-2 mAbs of the CRAM panel described in Fig. 3. Endothelial cell lines (tEnd.1, eEnd.2, and TME) expressed low levels of JAM-2 on the cell surface and variable levels of JAM-1 (Fig. 5A). We were unable to detect JAM-2 on any murine epithelial cell lines (data not shown), whereas the majority of these expressed JAM-1 with similar levels to the carcinoma cell line, KLN 205. This is consistent with the expression of JAM-1 by various epithelial cells such as entero-cytes, (8), and the restricted expression of JAM-2 to HEVs and lymphendothelium in lymphoid organs or vascular structures of the kidney. Therefore, flow cytometric analysis of freshly isolated endothelial cells was performed following collagenase/ dispase organ dissociation of Peyer's patches, lymph node, and kidneys. Endothelial cells were identified by their staining with both PECAM/CD31 and acetylated LDL (36). The double positive cell population was analyzed for JAM-1 or JAM-2 expression, and the results are shown in Fig. 4B. In kidney and Peyer's patches, all cells that were positive for CD31 and acetylated LDL also expressed JAM-2 or JAM-1, demonstrating that, at least in these organs, endothelial cells express JAM-2 and JAM-1 in vivo. When the staining was performed on cells obtained from lymph node, JAM-2 and JAM-1 expression were weaker and only found on a subpopulation of PECAM/LDL double-positive cells, reflecting a possible heterogeneity of endothelial cell phenotypes within this tissue. Taken together, the results of cytometric and immunohistochemical analysis show that JAM-2 expression is restricted to subpopulations of endothelial cells found in kidney, Peyer's patches, and lymph nodes.
JAM-2, a 45-kDa Protein, Interacting Homophilically-Since the TME cell line derived from high endothelial venules expressed the highest level of JAM-2, we used this murine cell line to further study the subcellular localization of JAM-2, and compare it with that of JAM-1. The localization of the JAM-2 protein on the surface of the endothelial cells was restricted to cell-cell contacts (Fig. 6A, a). The staining for JAM-2 was weaker than that observed for JAM-1 and less prominent in the membrane extensions between cells. Next, we wished to inves- tigate whether JAM-2 present at cell-cell contacts interacted homophilically with JAM-2 or whether it interacted heterophilically with another molecule on the neighboring cell. For this purpose we fused the JAM-2 protein to green fluorescent protein (JAM-2-EGFP), and transfected the construct into CHO cells. When CHO cells transfected with JAM-2-EGFP cDNA reached confluence, JAM-2 was observed in cell-cell contacts between two cells, exclusively when both cells expressed the protein (Fig. 6B). The contacts between expressing and nonexpressing cells were devoid of JAM-2 (Fig. 6B, a, indicated  by arrowhead). The same result was obtained when cells were transfected with the chimeric molecule JAM-1-EGFP (Fig. 6B,  b). This indicated that either JAM-2 or JAM-1 needed homophilic interactions to be localized at cell-cell contacts. Furthermore, experiments with mixed cells transfected with both constructs indicated that JAM-1 was unable to interact with JAM-2 (data not shown).
For a biochemical characterization of JAM-2, we performed immunoprecipitation with CRAM-19H36 mAb. First, we wanted to confirm the reactivity of CRAM-19H36 mAb and the predicted molecular weight of JAM-2 using lysates of MDCK cells transfected with plasmid encoding the full-length sequence of JAM-2. As shown in Fig. 6C, the CRAM-19H36 mAb immunoprecipitated a single band of Ϸ45 kDa from lysate of transfected cells (lane 1), and no signal was detectable when lysate of nontransfected cells were used (lane 2). After treatment of the immunoprecipitated material with N-glycosidase F, we observed a reduction of the apparent molecular mass from 45 to 34 -36 kDa (lane 3), the faint upper band resulting probably from a partial digestion with the enzyme. This was in agreement with the predicted molecular weight of the core JAM-2 protein. When immunoprecipitations were performed on TME cell lysate, the molecular mass of 45 kDa for JAM-2 protein was confirmed (Fig. 6D, lane 3), and JAM-1 protein was resolved as a single band of lower molecular mass (Fig. 6D, lane  2). This difference in size between JAM-1 and JAM-2 was further confirmed by immunoprecipitation of the EGFP recombinant fusion proteins after surface biotinylation of transfected cells. As shown in Fig. 6E, single broad bands of 70 and 73 kDa were, respectively, obtained for JAM-1-EGFP (lane 2) and JAM-2-EGFP (lane 4). This was expected since EGFP has a molecular mass of 28 kDa.
Dynamic Localization of JAM-2 to Cell-Cell Contacts-To understand the mechanism by which JAM-2 was specifically localized to cell-cell contacts, time-lapse video microscopy was performed. CHO cells, stably transfected with the fluorescent chimeric molecule, were trypsinized and plated into chamber slides for imaging. After cell spreading, surface expression of JAM-2-EGFP was not uniform, but was rather clusterized at existing cell-cell contacts (Fig. 7A, cells depicted by asterisks). During the formation of new cell-cell contacts, localization of JAM-2-EGFP to cell junctions was observed and an intense fluorescence signal was detected at the novel contact point between the cells forming the new cell-cell contact (arrows). The chimeric protein was enriched in the membrane protrusions between contacting cells, leading to the "zipper like" pictures seen after 12 or 18 min. Interestingly, the localization of JAM-2 at the primary cell-cell contacts was not lost during the formation of the new membrane contact (see upper left corner cell contacts). This finding indicated that, once JAM-2-EGFP was specifically localized to the new cell contact, its localization was stable. To further address the requirements for JAM-2 localization, time-lapse video microscopy was performed after wounding the cell monolayer (Fig. 7B). Cells at the wounded edge maintained JAM-2 at their intact contact sites (arrowhead), but lost JAM-2 localization at the wounded side (arrows), indicating that JAM-2 engagement was necessary to maintain its membrane localization. Over a period of 90 min following wounding, cells bordering the wound began to migrate into the wounded area. Interestingly, these cells maintained contacts with neighboring cells via membrane protrusions that were brightly fluorescent, i.e. JAM-2-positive (arrowhead). These results support the hypothesis that JAM-2 homophilic interactions may play a role in the establishment or maintenance of cell-cell contacts.
JAM-2 Increases Monolayer Tightness and Participates in Tight Junctional Complexes-Since a number of molecules par-ticipating in cell-cell contacts have been shown to regulate the paracellular permeability of cell monolayers, we tested whether JAM-2 might also affect this function. Transfection of JAM-2-EGFP reduced the paracellular permeability to FITCdextran and improved sealing of CHO cell monolayers by 42.5%, whereas transfection of the unrelated molecule Tac (interleukin-2 receptor ␣) did not significantly reduce the paracellular permeability (Fig. 8). The transfection of JAM-1-EGFP also reduced the paracellular permeability of the cells monolayer as described previously for wild type JAM-1 (8). Since CHO cells do not present tight junctions, and because paracellular permeability has been shown to be regulated by molecules participating in tight and adherens junctions, we wished to explore the participation of JAM-2 to such subcellular junctional complexes. To address this issue, we transfected the JAM-2-EGFP chimeric protein in MDCK cells, known to possess well defined tight and adherens junctions. As shown in Fig. 9A, when serial pictures taken at 0.9-m intervals were analyzed for EGFP fluorescence and compared with occludin staining, JAM-2-EGFP was specifically enriched in cell-cell contacts at the level of the tight junction (Fig. 9A, fourth and fifth pictures to the right). At the basal level, we observed intracellular dots of EGFP fluorescence, probably due to the overexpression of the protein. The restricted localization of JAM-2-EGFP to tight junctions was more striking when 40 pictures taken at 0.3-m intervals were stacked, and rotated along the x-y axis (Fig. 9B). Colocalization with ZO-1 was observed (Fig. 9B, a), whereas JAM-2-EGFP was not found in the adherens junctions stained with Arc-1 anti-E-cadherin antibody (Fig. 9B, b). Similar results were obtained when nonmodified JAM-2 was transfected (data not shown), indicating that EGFP does not affect JAM-2 localization to tight junctions. These results demonstrate that JAM-2 presents the properties of a novel junctional adhesion molecule, targeted to tight junctions. DISCUSSION We herein describe the cloning and characterization of JAM-2, a novel junctional adhesion molecule of endothelial cells. In adult murine tissues, JAM-2 expression is restricted to lymphendothelial cells, endothelial cells in the kidney, and HEVs of lymphoid organs.
The identification of JAM-2 is achieved by a selective method of RNA display, which uses degenerated primers to selectively amplify transcripts of interest. The sequence of the degenerated primers used in this technique is based on the amino acid sequence participating to the structure of C 2 immunoglobulin domains. Therefore, JAM-2 is a type I protein with two Ig domains: one V domain and one C 2 domain. The molecule is homologous to the recently described JAM-1 (7,8). In its membrane-proximal domains, JAM-2 contains two cysteine residues more than JAM-1, which leads us to conclude that it is organized as a V H /C 2 molecule unlike JAM-1, which was classified as a V H /V H molecule (8,37). Several stretches of sequence conservation between JAM-1 and JAM-2 are also identified within their cytoplasmic domains. Since the C terminus PDZ binding domain of JAM-1 was recently shown to bind and recruit PDZ-containing proteins, it will be interesting to ad-dress whether these shared sequences may control the specific localization of both molecules to the tight junctions of polarized cells (38,39). The specific localization of the EGFP chimeric proteins, in which the conserved C-terminal sequence SSFVI was removed, excluded a role for this conserved sequence in the targeting of the proteins (38). Indeed, the protein sequence motifs identified as putative phosphorylation sites for PKCs (Ser 274 , Ser 281 ) may be sufficient to guide JAM-2 to cell-cell contacts, since a role for PKCs in the establishment of cell-cell contacts was recently suggested (40,41). However, we cannot exclude a role of the extracellular domains in the specific localization of the proteins since JAM-2 and JAM-1 were solely enriched in the membrane contacts between transfected cells.
The dynamics of JAM-2 localization to cell-cell contacts was explored using JAM-2 protein fused to green fluorescent protein and time-lapse video microscopy. Time-lapse imaging of JAM-2-EGFP during the formation of cell-cell contacts showed that the protein was located within minutes to the contacts between cells. Once partial cell-cell contacts were formed, the JAM-2 enriched membranes closed in a "zipper-like" fashion, indicating that JAM-2 homophilic interaction is an early event in the establishment of cell connections and may play a role in cell-cell contact organization. The latter hypothesis gains further support from paracellular permeability experiments in CHO cells transfected with JAM-1-EGFP or JAM-2-EGFP. A similar reduced paracellular permeability was described following transfection of VE-cadherin and JAM-1, two molecules participating in adherens and tight junctions, respectively (8,42). Accordingly, when expressed in MDCK cells, JAM-2 was colocalized with ZO-1 and occludin, two components of tight junctions (11,29), indicating that JAM-2 is specifically associated with subcellular junctional complexes in epithelial cells. Together, these results show that JAM-1 and JAM-2 not only have sequence similarities, but also have similar properties in terms of subcellular localization, permeability regulation, and homophilic interactions.
The major difference between JAM-2 and JAM-1 appears to be in their tissue distribution. Although the morphogenesis of tight junctions during vasculogenesis or angiogenesis is not known, the absence of the JAM-1 transcript and the prominence of the JAM-2 transcript during embryogenesis may reflect a sequentially ordered role of the proteins in the biogen- esis of tight junctions. More interestingly, the expression of JAM-2 in adult murine tissue is restricted to endothelial cell subpopulations such as HEVs or lymphatic cells, whereas JAM-1 is expressed by platelets, antigen-presenting cells, endothelial cells, and epithelial cells (7,8). However, due to the lack of tight junctions in cultured endothelial cells (43), the participation of JAM-2 to tight junctional complexes was examined in epithelial cells upon transfection. Therefore, its participation to epithelial tight junctions may indicate its participation to similar compartments in endothelial cells expressing it. Nevertheless, this raises the question of the presence of endothelial tight junctional complexes in HEVs or lymphatic vessels, two endothelial compartments specialized in leukocyte recirculation. According to the specific function of HEVs in the recirculation of leukocytes, many studies demonstrated the absence of tight junctions in HEVs, whereas others suggested that partial or normal tight junctions may be present (44 -46).
In the present study we demonstrated that HEVs express ZO-1, JAM-1, and JAM-2, three molecules that incorporate into tight junctions. Since the presence of such structures in HEVs is still a matter of debate, it remains to be seen whether ZO-1, JAM-1, and JAM-2 participate in "tight junctional-like" domains in HEVs. More interestingly, JAM-2 was not expressed by adult brain vasculature (data not shown), which is known to express ZO-1, occludin, and JAM-1, and which forms one of the tightest blood tissue barrier of the body (43,47,48). It is therefore tempting to speculate that JAM-2 may be present in tight junctional-like structures, which undergo continuous remodeling to allow the passage of transmigrating leukocytes. An alternative hypothesis is that JAM-2 expression is induced by microenvironmental factors, resulting in loosening of the existing endothelial tight junctions. The negative regulation of endothelial tight junctions may result from the competition of JAM-2 and JAM-1 for intracellular proteins involved in the stabilization of tight junctional complexes. Interestingly, the JAM-1 staining on HEVs, where JAM-2 was highly expressed, seemed to be diffuse on the cell surface (Fig. 4, insets). We therefore believe that the relative participation of JAM molecules to subcellular compartment such as tight junctions may regulate the properties of the vascular bed (49 -51). A future challenge will be to identify the molecular architecture of junctional complexes in situations such as chronic inflammation or angiogenesis known to affect the function of the vascular barrier (49 -51).
Addendum-In conclusion, the characterization of JAM-2, its structural relation to JAM-1, and the overlapping but distinct tissue distribution of both molecules suggest the existence of a novel junctional adhesion molecule family. This is further supported by the fact that, while the present report was under review, a related molecule named VE-JAM was characterized by Palmeri and colleagues (35). This molecule is different from the presently described JAM-2 molecule, and is not recognized by our CRAM panel of anti-JAM-2 mAbs (data not shown). Therefore, it is tempting to speculate that more members of the junctional adhesion molecule family have to be expected, and that the specific expression of each member is of central importance for studying the interendothelial junctions in respect to permeability control and leukocyte trafficking.