Vascular Endothelial Junction-associated Molecule, a Novel Member of the Immunoglobulin Superfamily, Is Localized to Intercellular Boundaries of Endothelial Cells*

During the process of lymphocyte homing to secondary lymphoid organs, such as lymph nodes and tonsils, lymphocytes interact with and cross a specialized microvasculature, known as high endothelial venules. There is a great deal of information available about the first steps in the homing cascade, but molecular understanding of lymphocyte transmigration through the intercellular junctions of high endothelial venules is lack-ing. In analyzing expressed sequence tags from a cDNA library prepared from human tonsillar high endothelial cells, we have identified a cDNA encoding a novel member of the immunoglobulin superfamily. The protein, which we have termed VE-JAM (“vascular endothelial junction-associated molecule”), contains two extracellular immunoglobulin-like domains, a transmembrane domain, and a relatively short cytoplasmic tail. VE-JAM is prominently expressed on high endothelial venules but is also present on the endothelia of other vessels. Strik-ingly, it is highly localized to the intercellular boundaries of high endothelial cells. VE-JAM is most homologous to a recently identified molecule known as Junctional Adhesion Molecule, which is

The endothelial lining of blood vessels serves as a critical interface between blood and tissues, regulating processes such as vascular permeability, blood flow, thrombogenesis, hematogenous metastasis, and leukocyte extravasation (reviewed in Ref. 1). Although all vascular endothelial cells perform certain functions in common, there is a remarkable diversity of specialized functions that depend on the class of the blood vessel and the requirements of the underlying tissue. In many instances, our understanding of these diverse functions is at the phenomenological level without detailed molec-ular explanations.
One specialized microvasculature that has received a lot of attention are the postcapillary high endothelial venules (HEV) 1 of secondary lymphoid organs (reviewed in Ref. 2). These vessels, which are lined by morphologically distinctive high endothelial cells (HEC), serve as the portal of entry of blood-borne lymphocytes into secondary lymphoid organs, such as lymph nodes and tonsils, allowing lymphocytes of appropriate specificity to respond to sequestered and processed antigens within the organ. The highly selective recruitment process, termed "lymphocyte homing" involves a cascade of steps: rolling of the lymphocyte along the HEV, arrest and flattening on the endothelium, and finally transmigration across the endothelium (reviewed in Refs. 3 and 4). There has been a great deal of progress in our molecular understanding of the first two steps in the homing cascade, i.e. in the identification of primary and secondary adhesion molecules and the elucidation of signaling mechanisms involved in integrin activation (reviewed in Ref. 5). However, considerably less information is available about the molecular basis of lymphocyte transmigration across the HEV.
The capability to isolate primary HEC has opened up new approaches to explore the specialized characteristics of these endothelial cells (6 -9). We have taken a gene profile approach by constructing a cDNA library from primary human tonsillar HEC and performing EST sequencing. Several of the sequences that we identified corresponded to a novel member of the immunoglobulin superfamily (IgSF). The encoded protein, which is termed VE-JAM, is prominently expressed on HEC of tonsils but is also present on the endothelium of other vessels. It is most closely related to another member of the IgSF, i.e. JAM, which has been implicated in the recruitment of monocytes and neutrophils across endothelium (10 -14).

EXPERIMENTAL PROCEDURES
Construction and Sequencing of HEC PCR Select Library-HEC were isolated from surgical specimens of tonsils, and the cDNA library (in pcDNA1.1) was prepared as described previously (15). In parallel, an aliquot of the same HEC cDNA was employed to generate a PCR select library (16). The HEC cDNA was cut into fragments that averaged 400 bp by digestion with RsaI. Subtraction was performed with cDNA fragments (also RsaI-digested) obtained from human umbilical vein * This work was supported by the National Institutes of Health MERIT Award R37GM23547, Roche Bioscience, and the Northern California Arthritis Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF255910 and AF255911.
Cloning of VE-JAM-A pool selection technique (15) was employed. Previously mentioned primers were designed to amplify 539 bp of the insert of EST clone 1293, which contains sequence toward the 3Ј-end of the VE-JAM cDNA (Fig. 1). To favor the identification of a full-length clone, the individual pools were tested in a PCR reaction using a 5Ј-primer matching the vector and a 3Ј-primer matching VE-JAM. Southern blotting was performed on the PCR reactions to identify the pools with the longest inserts. The isolated clone was sequenced in both directions.
Northern Blot Analysis-The probe consisted of the entire insert of EST clone 1293 (Fig. 1). The probe was labeled with [ 32 P]ATP (Amersham Pharmacia Biotech) using the Strip-EZ DNA kit (Ambion, Austin, TX). Multiple tissue and immune system tissue Northern blots (CLON-TECH, Palo Alto, CA) containing human poly(A)ϩ mRNA were hybridized at 60°C overnight after prehybidization for 2 h in ExpressHyb hybridization solution (CLONTECH). The blots were then washed twice in 2ϫ SSC, 0.1% SDS for 15 min at room temperature before autoradiography. These blots were then stripped and rehybridized with a ␤-actin probe that was supplied with the blots.
Flow Cytometry-Chinese hamster ovary (CHO) cells were grown to 80% confluency in 100-cm 2 flasks and transfected with the empty vector or the VE-JAM plasmid (8 g/flask) using LipofectAMINE (Life Technologies, Inc.). 48 h after transfection, the cells were harvested with 4 mM EDTA in PBS without calcium or magnesium and washed once in 0.2% bovine serum albumin in PBS, and 5 ϫ 10 5 cells/well were placed in a 96-well plate. Cells were incubated with the appropriate antibody for 30 min on ice and washed three times with cold PBS containing 0.2% bovine serum albumin. After incubation with fluorescein isothiocyanate goat-anti-rabbit secondary antibodies (Zymed Laboratories Inc., South San Francisco, CA) for 30 min on ice and washing, flow cytometry was performed. Peripheral blood was obtained by venipuncture and stained as above. Leukocyte populations were identified by their forward and side scatter characteristics. Platelets were identified with an antibody to ␣IIb␤3 (Plt-1, Immunotech, Westbrooke, ME). Staining was analyzed using CellQuest (Becton-Dickinson, Franklin Lakes, NJ).
Protein Purification-A cDNA for the extracellular domain of VE-JAM (1-717 bp of the coding region) was inserted into Ig-pEFBos vector, which encodes the Fc region of human IgG1 (18). The expression plasmid containing the VE-JAM/Ig chimera was transfected into COS cells using serum-free media. After 5 days, the medium was collected. 25 mM Tris-HCl, pH 8.0 (Bio-Rad), was added to the supernatant, which was agitated overnight at 4°C with protein A beads (Repligen, Needham, MA). The following day, the beads were poured into a column equilibrated in PBS. The protein was eluted with 100 mM triethylamine (pH 11.0) into a tube containing 1/10 the final volume of 3 M Tris-HCl, pH 6.8. The eluate was concentrated with a Centricon30 (Amicon, Beverly, MA) and equilibrated with PBS.
A glutathione S-transferase (GST) fusion protein was prepared by inserting a cDNA of the extracellular domain of VE-JAM (1-717 bp) into a modified PEAK 10 vector (Edge Biosystems, Gaithersburg, MD), which encodes a portion of GST (bases 258 -917) (Amersham Pharmacia Biotech) (18). The VE-JAM⅐GST fusion protein was produced in PEAK Rapid cells (Edge Biosystems) as recommended by the manufacturer. The fusion protein in the conditioned medium was purified with glutathione-agarose (Sigma) by overnight incubation at 4°C. The beads were placed into a column, washed with PBS, and eluted with 10 mM reduced glutathione, pH 8.0 (Sigma), in PBS. The eluate was concentrated with a Centricon30 (Amicon) and equilibrated with PBS.
Antibodies-Polyclonal antibodies were produced by Research Genetics (Huntsville, AL) where 800 g of VE-JAM/GST fusion protein was used to immunize two rabbits. Antibody titer was determined by enzyme-linked immunosorbent assay using VE-JAM⅐Ig fusion protein as the antigen (19). Antibodies were affinity purified by chromatogra-phy on a column of VE-JAM⅐Ig fusion protein coupled to cyanogen bromide-activated Sepharose (Sigma) according to standard procedures (19). Anti-VE-JAM antibody was biotinylated using EZ-link-Sulfo Link (Pierce) according to the manufacturer's instructions.
Peptide Microsequencing-10 g of purified VE-JAM/GST fusion protein was subjected to SDS-polyacrylamide gel electrophoresis on a 7.5% gel electroblotted onto Problott (Applied Biosystems, Foster City, CA). The appropriate band, identified by staining with Coomassie Brilliant Blue R-250 (BDH Chemicals Ltd., Poole, United Kingdom), was excised and subjected to Edman degradation analysis.
Immunohistochemistry-Human tonsil, foreskin, lung, placenta, and heart samples were obtained from surgical or autopsy specimens at the University of California, San Francisco. Rat heart was isolated from a Harlan Sprague-Dawley female by standard procedures. Tissues were frozen in OCT embedding medium (Miles Inc, Elkhart, IN). Frozen sections were cut (10 m) and fixed in 1:1 methanol:acetone for 5 min. Endogenous peroxidase activity was quenched with 0.3% hydrogen peroxide in 100% methanol for 20 min. Sections were blocked with PBS containing 5% normal goat serum (or 0.2% bovine serum albumin for placenta and rat heart only). Rabbit polyclonal VE-JAM and anti-CD31 (monoclonal antibody 2148, Chemicon, Temecula, CA), rabbit polyclonal von Willebrand factor (DAKO, Denmark), and MECA-79 were used at 0.3, 1, and 1 g/ml, respectively. After staining for 1 h at room temperature, sections were washed two times in PBS. For immunofluorescence, bound antibodies were detected with Cy 3-conjugated goat anti-rabbit IgG or Cy 2-conjugated goat anti-mouse IgG (1:1000) (Jackson Immunoresearch, Westgrove, PA). For immunocytochemistry, biotinylated goat anti-rabbit IgG (Vector, Burlingame, CA) was added to detect bound rabbit antibodies. This was followed by addition of ELITE ABC (Vector) mixture, using Nova Red (Vector) as the substrate. Normal rabbit IgG (Caltag, Burlingame, CA) or mouse IgG1 (Zymed Laboratories Inc.) was used as control.
Human Tonsillar Stroma Preparation-Tonsillar stroma was prepared as described by Baekkevold et al. (7) with a few modifications. Briefly, surgical specimens were stored for less than 4 h in sterile saline at 4°C. Disinfection was achieved with Cliniscrub solution (Clinipad Corp., Rocky Hill, CT) diluted 1/50 in PBS. Tonsils were finely chopped with a razor blade, and the resulting mince was depleted of lymphocytes by washing with PBS using a Falcon 70-m cell strainer (Becton Dickinson, Franklin Lakes, NJ). The remaining tissue was digested for 15 min at 37°C in RPMI 1640 medium (Fischer, Pittsburgh, PA) containing 1 mg/ml collagenase A (Roche Molecular Biochemicals), 1 mg/ml dispase (Roche Molecular Biochemicals), and 0.04 mg/ml DNase I (Roche Molecular Biochemicals). After free lymphocytes were removed by washing over a cell strainer, the residual stromal components were digested for an additional 60 min at 37°C with fresh enzyme solution. The digested stroma was washed once with RPMI 1640 and plated onto Vitrogen (Celtrix Laboratories, Palo Alto, CA) coated glass slides. Nonadherent cells were washed away after 90 min, and adherent cells were cultured in EBM-MV medium (Clonetics, Walkersville, MD).
Immunoprecipitation from Human Tonsil Lysates-A frozen human tonsil obtained from the Cooperative Human Tissue Network, Western Division (Case Western University, Cleveland, OH) was homogenized in 25 ml of PBS plus 2% Triton X-100 (Roche Molecular Biochemicals), 5 mM EDTA plus 1X Complete™ protease inhibitors (Roche Molecular Biochemicals). The lysates were spun at 20,000 ϫ g for 20 min at 4°C. The supernatants were collected and frozen in 1-ml aliquots. 1 g of VE-JAM antibody was added to 1 ml of tonsillar lysate and rocked overnight at 4°C. 10 l of protein A beads were added and incubated for 1 h at 4°C. The protein A beads (Repligen) were then spun down and washed five times with PBS. The beads were boiled in 20 l of 6ϫ SDS loading buffer and run on 10% SDS-polyacrylamide gel electrophoresis. The separated proteins were transferred to a Pro-Blott membrane and probed with 1 g/ml biotinylated anti-VE-JAM in 3% bovine serum albumin in PBS for 1 h at room temperature. The blot was incubated with horseradish peroxidase-conjugated streptavidin (Caltag) for 1/2 h with detection of specific bands by the ECL method (Amersham Pharmacia Biotech).

RESULTS
To identify genes that are selectively expressed in HEC, we prepared a PCR select library (16) from HEC cDNA, which was subtracted with HUVEC and PBL cDNAs. This approach utilizes relatively small cDNA fragments (400-600 bp) that are generated with the RsaI restriction enzyme. We subjected the library to an EST analysis. A full description of this library and our analysis of ESTs will be the subject of a future communication.
Of the novel cDNAs present within the library, one appeared five times as partial sequences. A full-length clone was obtained from an unsubtracted HEC cDNA expression library (15) using a pool selection technique in conjunction with PCR (15). The full-length cDNA (Fig. 1) contains a single open reading frame, which is predicted to encode a protein of 298 amino acids followed by a 3Ј-untranslated region of 72 bp and a poly(A) tail. The start codon satisfies the criteria for a strong context for translation initiation (20). A cleavable signal peptide of 28 amino acids was predicted using SignalP (21), which was verified by amino-terminal sequencing of a GST fusion protein expressed in PEAK Rapid cells. Overall the cDNA is predicted to encode a type I transmembrane protein with an extracellular portion of 237 amino acids, a transmembrane domain of 23 amino acids, and a cytoplasmic tail of 38 amino acids (GenBank TM accession number AF255910). The protein is hereafter referred to as VE-JAM based on the characterization provided below.
The extracellular region of VE-JAM is comprised of two IgSF domains, a membrane distal V-set domain, and a membrane proximal C2-type domain. It resembles an extensive list of cell surface antigens with two tandem IgSF domains (see "Discussion" below). A subset of these molecules (i.e. ChT1, HCAR, and A33 antigen) contains two additional, similarly spaced cysteines in their C2 type domains, in addition to the conserved pair of cysteines that is characteristic of most IgSF domains (22)(23)(24). VE-JAM exhibits these extra cysteines with the identical spacing found in these other proteins (Fig. 2).
Using the human sequence VE-JAM cDNA sequence as a probe, we identified a homologous mouse sequence in the NCBI dbEST data base. When the clone containing the EST was retrieved and fully sequenced, we obtained a full-length cDNA (GenBank TM mEST AI154320) (Fig. 1). The sequence predicts a protein of 298 amino acids, which is 80% identical to human VE-JAM and exhibits the same general features; therefore, this protein is very likely to be the murine ortholog of VE-JAM (GenBank TM accession number AF25911).
Sequence comparisons with the GenBank TM NR data base revealed that VE-JAM is most homologous to JAM (Fig. 2), a recently identified protein that is localized to intercellular boundaries of epithelial and endothelial cells and that is strongly implicated in transendothelial migration of myeloid cells (10 -14). Homology between VE-JAM and JAM is seen throughout the predicted proteins with an overall identity of 34%. Two stretches of 9 and 11 amino acids in the second IgSF domains are identical. In addition, the potential glycosylation sites in the second IgSF domain of VE-JAM are conserved in mouse and human JAM. However, the extra pair of cysteines is not present in either mouse or human JAM (Fig. 2B). The next most homologous protein to VE-JAM is A33 antigen, which exhibits 22% identity in amino acid sequence, including conservation of the extra cysteine pair (22).
A Northern analysis on a multitissue blot (human) was carried out with a 509-bp probe corresponding to the 3Ј-end of the VE-JAM cDNA (Fig. 3A). Three major species of Ϸ1.2, 1.4, and 4.4 kilobases were detected in several tissues, the smallest of which is consistent with the minimum expected message size. Variable expression in different tissues was observed with highest expression in the heart followed by the placenta and lymph node.
Because cDNAs corresponding to VE-JAM were initially detected in a library derived from HEC mRNA, we wished to confirm the presence of transcripts in HEC. As shown in Fig.  3B, VE-JAM-specific primers amplified a band in a PCR reaction using cDNA made from an independent preparation of HEC. There was no specific band when reverse transcriptase was not included in the preparation of template, proving that RNA, and not genomic DNA, was responsible for the amplification. There was no detectable signal from PBL cDNA.
To study VE-JAM at the protein level, we prepared a rabbit polyclonal antibody to the ectodomain of the protein. A GST fusion protein containing the entire extracellular domain of the human VE-JAM was expressed in mammalian cells, purified on a glutathione-agarose column, and used to immunize rabbits. Specific antibodies were affinity purified on a column of an immunoglobulin chimera of VE-JAM, which contained the extracellular domain of VE-JAM fused to the Fc portion of IgG1. The purified antibodies stained CHO cells that were transfected with the VE-JAM cDNA but not the parental cells (Fig.  4A). Normal IgG was negative on both populations of cells. CHO cells transfected with a cDNA encoding human JAM were not stained by the VE-JAM antibody, establishing that the antibody was not cross-reactive to the related protein (data not shown).
To visualize the protein encoded by the VE-JAM cDNA, we carried out a Western blot analysis on VE-JAM-transfected CHO cells. A single band of Ϸ40 K d was detected in lysates of transfected cells but not in the parental cells (Fig. 4B). Normal rabbit IgG did not react with any bands. The same molecular weight was observed when VE-JAM was labeled at the cell surface by surface biotinylation (not shown). The unmodified mature protein is predicted to have a molecular mass of 33 kDa. The extra mass is likely attributable to glycosylation at the potential sites for N-linked modifications.
To determine the cellular expression pattern of VE-JAM, we carried out immunocytochemistry on cryostat sections of sev-  (38). The phylogenetic tree was plotted using NJPlot (39). eral human tissues with the polyclonal VE-JAM antibody.
Staining of human tonsil sections demonstrated the VE-JAM was strongly expressed in HEV (Fig. 5, A and B), consistent with the occurrence of ESTs in the HEC cDNA library and the reverse transcriptase-PCR experiments. Specificity was established by demonstrating that all reactivity was blocked when the antibody was mixed with purified VE-JAM⅐IgG (not shown). In human and rat heart, staining with the VE-JAM antibody was observed on the endothelium of endocardium, arteries, venules, and capillaries (Fig. 5, C and D). Co-staining for PECAM-1 or von Willebrand factor, widely used markers for vascular endothelium (reviewed in Refs. 25 and 26), verified that VE-JAM was present on the endothelial lining of endocardium and the vessels (not shown). A survey of other human tissues (lung, placenta, and foreskin) established staining on the endothelia of both large and small vessels. Interestingly, VE-JAM was not detected in oral epithelium, epidermis of skin, respiratory epithelium of bronchus (not shown), or in mesothelium of the pericardium (Fig. 5C). This result is in contrast to the expression pattern of JAM, which is broadly distributed on both simple and stratified epithelia, as well as vascular endothelium (10,11). Additionally in flow cytometry experiments, VE-JAM was not detected on the surface of platelets or leukocytes isolated from blood (Table I). This latter finding was consistent with the failure to detect VE-JAM mRNA in RNA of PBL (Fig. 3A) and the absence of staining in T-and B-cell zones of tonsils (Fig. 5, A and B).
In many of the HEV, VE-JAM appeared to be concentrated at lateral boundaries of HEC (Fig. 5B). To study the distribution of VE-JAM on these cells in more detail, we isolated tonsillar stroma and placed it in tissue culture. During the first several days of culture, distinct clusters of endothelial cells were evident, as indicated by staining with PECAM-1 (Fig. 5, E and F). Co-staining with PECAM-1 and VE-JAM antibodies indicated a co-distribution of the two proteins at intercellular boundaries, whereas staining was much weaker at the free surfaces of the cells (Fig. 5, E and F). When the clusters were dual stained with VE-JAM antibody and MECA-79 (an HEV-specific monoclonal antibody (27)), the former antibody highlighted intercellular boundaries as above, whereas the latter reagent stained cell bodies diffusely (not shown).
To characterize the naturally occurring form of VE-JAM, we took advantage of its strong expression in the HEV of tonsils. We immunoprecipitated a detergent lysate with the VE-JAM antibody and then performed Western blotting on the precipitate with a biotinylated version of the same antibody (Fig. 6). A single band was visualized, which ran at Ϸ40 kDa under both reducing and nonreducing conditions. This result argues that VE-JAM is unlikely to use the extra pair of cysteines in its C2-type IgSF domain for interchain disulfide bonds.
The chromosomal location of the VE-JAM gene was determined by searching for human genomic clones in the NCBI GenBank TM NR data base. The VE-JAM cDNA sequence was distributed across several genomic clones (AP000223, AP000225, and AP000226), which are tandemly arrayed along the long arm of chromosome 21 (21q21.2). Only two differences were noted between the genomic sequences and our cDNA, neither of which affects the amino acid sequence of the mature protein. It should be noted that mutations, deletions, and du-plications in the long arm of chromosome 21 are associated with a number of disorders including Usher syndrome, Down syndrome, and Alzheimer's disease. (28 -30). DISCUSSION Taking advantage of procedures for the isolation of primary HEC from secondary lymphoid organs, a number of recent efforts have been directed at the identification of genes that are differentially expressed in HEC (6,8,9). Our approach has been to apply EST sequencing to a subtracted tonsillar cDNA library. We identified sequences corresponding to a novel member of IgSF, for which we obtained a full-length cDNA in both mouse and human. Because of its highly restricted expression in vascular endothelium and its association with intercellular boundaries, we have termed the molecule encoded by this cDNA as VE-JAM for vascular endothelial junction-associated molecule.
VE-JAM is predicted to be a type I integral membrane protein with two tandem extracellular IgSF domains of the V-set (membrane distal) and C2-set (membrane proximal). This organization is characteristic of the CD2 subgroup (CD2, CD58, CD48, 2B4, and CDw150) and as well as other members of the IgSF (CD80, CD86, OX-2, A-33, HCAR, and ChT1) (22)(23)(24)31). Interestingly, many of these have been implicated as mediators of cell-cell interactions, generally through heterophilic interactions with other IgSF members (31,32). A further feature of VE-JAM, shared by only three of these proteins (A-33, HCAR, and ChT1), is an extra pair of cysteines separated by the same spacing in the C2 domain (22)(23)(24).
Consistent with the occurrence of VE-JAM ESTs in the HEC cDNA library, reverse transcriptase-PCR analysis detected VE-JAM transcripts in RNA prepared from independently isolated HEC. Moreover, our immunohistochemical and immunoprecipitation experiments confirmed the expression of VE-JAM at the protein level in HEC. Staining of primary cultures of HEC and CHO cell transfectants established that VE-JAM was expressed at the cell surface.
Although VE-JAM was highly expressed in HEC, it was not restricted to this site, as first indicated by the presence of transcripts in a variety of nonlymphoid organs, but most conspicuously in heart and placenta. Immunocytochemistry indicated the presence of VE-JAM in endothelium of blood vessels in human and rat heart, human placenta, lung, and foreskin.   6. Visualization of VE-JAM from human tonsils. A detergent lysate of tonsils was reacted with VE-JAM antibody or normal rabbit IgG. The precipitates were electrophoresed on a 10% SDS gel, with or without reduction, and Western blotted with biotinylated VE-JAM antibody as the probe. Without reduction, the apparent molecular weight of VE-JAM was less than with reduction, consistent with the presence of disulfide bonds.
containing two V-set domains. The proteins, which are virtually the same length (298 or 299 amino acids), show the same basic organization with homology from the extracellular domains through the cytoplasmic tails. Like VE-JAM, JAM is expressed on endothelial cells of many blood vessels (10). However, JAM is also distributed on a wide variety of epithelia. Although our limited survey of human tissues cannot exclude the possibility that VE-JAM may be present on some epithelia, we failed to detect VE-JAM on several examples of epithelial cells (simple and stratified) where strong JAM expression has been documented (10). Furthermore, whereas JAM is on platelets (10, 11), we did not detect VE-JAM on these cells or any other blood cells.
A striking feature of JAM is that it is concentrated at intercellular boundaries of endothelial monolayers (9,10). A number of other molecules, including PECAM-1, and various proteins associated with tight and adherens junctions also exhibit this distribution (reviewed in Ref. 33). Dejana and co-workers (10,13) have shown that an antibody to mouse JAM can block transendothelial migration of monocytes both in vitro and in vivo. Both spontaneous and chemokine-induced transendothelial migration were found to be inhibited. Importantly, the antibody did not block adhesion of monocytes to cultured endothelium but prevented their transit across the endothelium, indicating the involvement of JAM in the transmigration process (10). A more recent study showed that administration of the antibody in a mouse experimental meningitis model attenuated monocyte and neutrophil recruitment into both cerebrospinal fluid and the brain parenchyma (13). The inhibitory effects of the JAM antibody on transmigration were incomplete, suggesting the involvement of other endothelial molecules in the process. PECAM-1, also concentrated at intercellular boundaries and implicated in the transmigration of myeloid cells across endothelia (34,35), is an obvious candidate for collaboration with the JAM system.
As reviewed in the Introduction, HEV are specialized for recruitment of lymphocytes into secondary lymphoid organs. Although many of the early events associated with lymphocyte rolling and arrest on HEV have been elucidated at the molecular level, the process of transmigration is poorly understood. JAM is expressed on these vessels but at a relatively low level (10). The prominent expression of VE-JAM in HEV and its distribution at interendothelial boundaries are, therefore, very intriguing findings. Given the many parallels between VE-JAM and JAM, future experiments should be directed at the investigation of the potential role of VE-JAM in lymphocyte transmigration across HEV.