Junctional adhesion molecule interacts with the PDZ domain-containing proteins AF-6 and ZO-1.

We have identified the PDZ domain protein AF-6 as an intracellular binding partner of the junctional adhesion molecule (JAM), an integral membrane protein located at cell contacts. Binding of AF-6 to JAM required the presence of the intact C terminus of JAM, which represents a classical type II PDZ domain-binding motif. Although JAM did not interact with the single PDZ domains of ZO-1 or of CASK, we found that a ZO-1 fragment containing PDZ domains 2 and 3 bound to JAM in vitro in a PDZ domain-dependent manner. AF-6 as well as ZO-1 could be coprecipitated with JAM from endothelial cell extracts, demonstrating the association of the endogenously expressed molecules in vivo. Targeting of JAM to sites of cell contacts could be affected by the loss of the PDZ domain-binding C terminus. Full-length mouse JAM co-distributed with endogenous AF-6 in human Caco-2 cells at sites of cell contact independent of whether adjacent cells expressed mouse JAM as an extracellular binding partner. In contrast, truncated JAM lacking the PDZ domain-binding C terminus did not co-distribute with endogenous AF-6, but was restricted to cell contacts between cells expressing mouse JAM. Our results suggest that JAM can be recruited to intercellular junctions by its interaction with the PDZ domain-containing proteins AF-6 and possibly ZO-1.


Introduction
Epithelial and endothelial cells are connected by intercellular junctions. Tight junctions, also called zonula occludens, form a regulated barrier between cells restricting the diffusion of small solutes across cellular sheets as well as the diffusion of proteins from the lateral to the apical membrane domain (1,2). Adherens junctions have evolved as signalling centers regulating tissue organization and morphogenesis (3,4). Both types of junctions are necessary to establish and maintain cellular polarity. Three integral membrane proteins or families of proteins have been identified in tight junctions. Occludin (5) and claudins (6,7) are fourtransmembrane domain proteins that are involved in the formation of tight junction strands (8). Junctional adhesion molecule (JAM) is a newly described Ig-SF member of adhesion molecules that colocalizes with tight junction molecules in both epithelial and endothelial cells (9).
The organization of tight junctions requires the association of transmembrane proteins with cytoplasmic proteins. Recently, proteins containing PDZ domains have emerged as organizers of protein complexes at the plasma membrane (2,10). PDZ domains are 80 -90 amino acids (aa) long and bind to five amino acid residue motifs at the carboxy terminus of their ligands (11). They are often found in multiple copies within a single protein and, therefore, PDZ domain-containing proteins are considered as scaffolding proteins that promote the clustering of protein complexes at specific subcellular sites (12).
In tight junctions four proteins containing one or several PDZ domains have been identified.
These include ZO-1 (13), , , and AF-6 (16). Occludin directly binds to ZO-1 and ZO-3 (15,17). Several lines of evidence suggest that the proper targeting of occludin to the tight junctions requires ZO-1 and not vice versa. For example, the absence of occludin does not change the localization of ZO-1 (18) whereas the absence of ZO-1 prevents targeting of transfected occludin to cell-cell contacts in fibroblasts (19). In addition, connexin 32-occludin chimeras containing only the ZO-1 binding domain of occludin are properly targeted to tight junctional fibrils in MDCK cells (20). And finally, in occludin-deficient epithelial cells ZO-1 is properly targeted to tight junctions (21). These findings suggest that 6 (35). Yeast expression vectors were constructed in pBTM116 (bait plasmids) and pGAD10 (prey plasmids) thereby generating fusion proteins between LexA and JAM or between the GAL4 activation domain and the prey protein, respectively. JAM bait vectors were generated by subcloning the cytoplasmic tail (aa 261-300) of murine JAM (pBTM116-JAM/cyt) or mutants of the cytoplasmic tail lacking the COOH-terminal aa V (aa 261-299, JAM/cyt∆1), The murine homologues of l-afadin and s-afadin were cloned by RT-PCR with cDNA derived from bEnd.3 endothelial cells and primers derived from the rat cDNA sequences of l-and safadin (accession numbers U83230 and U83231) (24). The 5'-half of the cDNA common to land s-afadin was generated using the two primers AFA.13S (5'-7 Heidelberg, Germany). The sequence corresponding to the first 14 aa was generated separately by annealing two complementary oligonucleotides corresponding to nucleotides 322 -367 and subsequent cloning in front of the common 5'-region of afadin into pEGFP-C1 thereby generating the complete 5'-half of l-and s-afadin comprising nucleotides 322 -3272.

Two-Hybrid Screen
Two-hybrid screening experiments were performed essentially as described (36). Briefly, the Saccharomyces cerevisiae reporter strain L40 expressing a fusion protein between LexA and the cytoplasmic tail of JAM (aa 261-300) was transformed with 250 µg of DNA derived from a day 9.5/10.5 mouse embryo cDNA library (37) according to the method of Schiestl and Gietz (38). The transformants were grown for 16 h in liquid selective medium lacking tryptophan, leucine, and uracil (SD-TLU) to maintain selection for the bait and the library plasmid. The transformants were then plated onto synthetic medium lacking tryptophan, histidine, uracil, leucine, and lysine (SD-THULL) in the presence of 1 mM 3-aminotriazole   Table 1). The same results were obtained with X-gal assays (data not shown).
Together with the observation that the COOH-terminal aa residues of JAM match the binding motif predicted for group II PDZ domains (11) these results indicate that JAM binds to AF-6 in a PDZ domain-dependent manner.

Specificity of the interaction of JAM with the PDZ domain of AF-6
Both JAM and AF-6 have been described to be located at tight junctions. Since additional . Surprisingly, none of these recombinant forms of ZO-1 bound to JAM (Fig. 3). We conclude that JAM-binding to ZO-1 is not dependent on PDZ domain 1, but requires the two PDZ domains 2 and 3 as well as the connecting sequence segment between these two domains.

Interaction of GST-JAM with native, endogeneous AF-6 from epithelial and endothelial cells
Since JAM is expressed in epithelial and endothelial cells (9)  JAM and AF-6. The JAM mAb also co-precipitated ZO-1 and, similar to AF-6, the association between JAM and ZO-1 was sensitive to detergent treatment. Longer exposure of the film revealed that the presence of 0.02% SDS in the lysis buffer reduced but did not abolish the association between JAM and ZO-1 whereas the presence of 0.5% deoxycholate in addition to 0.02% SDS disrupted the interaction completely (data not shown). Together, these findings indicate that JAM associates with AF-6 and ZO-1 in vivo.

JAM is colocated with AF-6 in endothelial and epithelial cell contacts
We in epithelial cells (9,16). It was, however, also reported that AF-6 can be found at both tight and adherens junctions (31,43) and even exclusively at adherens junctions (24). In order to Thus, these cells do not allow to decide whether AF-6 is predominantly localized in tight junctions or adherens junctions or both.

JAM can be recruited to cell junctions by homophilic JAM interactions
We analyzed whether the COOH-terminal 9 amino acids of JAM are necessary for its localization to sites of cell-cell contact. To this end CHO cells were microinjected with cDNA constructs coding either for full length JAM or for truncated JAM lacking the last 9 amino acids at its COOH-terminus (JAM/∆9). When CHO cells were microinjected with full length JAM, staining of JAM was observed only in areas of cell contacts where neighbouring cells were also microinjected, but not at cell borders to non-injected cells (Fig. 7A). Surprisingly, truncated JAM also appeared at cell contact sites between cells that both expressed truncated JAM (Fig. 7B). Staining for endogeneous AF-6 revealed that AF-6 was diffusely localized in the cytoplasm around the nuclei, more weakly within the nuclei, but not at cell-cell contact sites. We conclude that JAM can be recruited to and maintained at cell-cell contact sites in an AF-6-and PDZ-domain-independent manner. This is most likely due to the interaction in trans mediated through homophilic interactions between JAM molecules expressed by neighbouring cells.

6, but not with ZO-1 staining
We next performed microinjection experiments with human Caco-2 cells. These cells express endogeneous JAM, (detected with mAb BV16, not shown), that is not recognized by the mAb BV11 directed against mouse JAM (not shown). As described above for CMT cells, staining with the polyclonal AF-6 antibody resulted in diffuse cytoplasmic staining pattern in postconfluent cells (see Fig. 8C). In grow-out zones and recently confluent cells, AF-6 was localized to the nucleus as well as in areas of cell-cell contacts with a circumferential rim staining typical for junctional components such as ZO-1 ( Fig. 8A and B). When full length JAM was microinjected into recently confluent Caco-2 cells, JAM co-distributed with AF-6 and was recruited to cell-cell contact sites irrespective of JAM expression in the neighbouring cell (Fig. 8A). In contrast, when truncated JAM (JAM/c∆9) was microinjected into recently confluent Caco-2 cells a clustering of JAM was only observed at cell contacts between injected cells, despite the fact that the injected cells showed a typical circumferential rim staining for AF-6 (Fig. 8B). These findings suggest that JAM can be recruited by AF-6 and point to a role of AF-6 for the recruitment of JAM during the generation of cell contacts. This was further supported by experiments where full length JAM was injected in postconfluent Caco-2 cells that showed a typical circumferential rim staining for ZO-1, but not for AF-6 ( Fig. 8C). Injected JAM did not colocalize with ZO-1 but only clustered at contact zones between injected cells. This indicates that the presence of ZO-1 is not sufficient to recruit JAM to cell-cell contact sites. Taken together, these findings suggest that JAM can be recruited to sites of cell-cell contacts by two mechanisms: a) by homophilic interaction of two

JAM molecules on neighbouring cells, and b) by a PDZ domain-dependent interaction
possibly requiring junction-associated AF-6 in the same cell. They also suggest that endogeneous ZO-1 in post-confluent cells is not sufficient to recruit JAM. Reports on the subcellular distribution of JAM and AF-6 have localized JAM at tight junctions while AF-6 has been described as an adherens junction-as well as a tight junctionassociated component (16,24). This dual localization of AF-6 has been verified in more recent studies. In wild-type ectoderm of mouse embryos, AF-6 colocalizes with both ZO-1 and Ecadherin (31). In addition, the subcellular localization seems to depend on the cell type used. Unfortunately, a clear assignment of AF-6 to junctions was not possible in the cells used here.

Discussion
Although well polarized CMT cells (Fig. 6B) or Caco-2 cells (not shown) on filters allowed to co-localize ZO-1 and JAM at the apical border of lateral cell contacts, AF-6 was not detectable at cell contacts in such post-confluent cell cultures (Fig. 8C).  (44), but with adherens junction components such as α-catenin in fibroblasts (45). During the formation of epithelial cell contacts after wounding ZO-1 co-localizes with E-cadherin/β-catenin at the initial stage of junction formation, but with occludin in established junctions (46). It is conceivable that JAM associates with AF-6 in a transient way during the formation of intercellular junctions. Whether AF-6 and JAM still associate in fully polarized epithelium in tissues still needs to be tested.
AF-6 and ZO-1 are both directly linked to the actin cytoskeleton (24,47). This allows JAM to be anchored to the actin filament system via its binding to AF-6 and ZO-1. Of the three integral membrane proteins described at tight junctions, occludin (5), claudins (6,8) and JAM (9), occludin has been reported to be linked to the actin cytoskeleton via ZO-1 (47) and claudins were recently described to associate with ZO-1, ZO-2, and ZO-3 by which they might be linked to actin filaments (22). Besides anchoring to the cytoskeleton, binding to PDZ domain proteins would enable junction proteins to interact with regulatory molecules. AF-6 is a target for the small GTPase ras (25,(48)(49)(50). AF-6 can directly bind to ZO-1 (16,26) and the ras associating domain in AF-6 is identical to the binding site for ZO-1 (16). Activated ras has been described to inhibit the interaction between AF-6 and ZO-1 and overexpression of ras was shown to perturb cell-cell contacts and decrease the amount of both AF-6 and ZO-1 at the cell surface (16). Together with these findings our results suggest that the ras-induced disruption of cell contacts could be mediated in part through the JAM/AF-6 or JAM/ZO-1 complex. Although in post-confluent Caco-2 and CMT cells AF-6 is not available at junctions for such regulatory mechanisms, it is present at epithelial and endothelial junctions in tissue (16,24).
Depending on the cell type and the state of confluency that was analyzed we have found that microinjected JAM can be recruited to sites of cell contact by two mechanisms: the first is independent and the second is dependent on the PDZ domain targeting motif of JAM. The PDZ domain targeting motif-independent pathway was observed with full length as well as with truncated JAM in CHO cells that lack expression of junctional AF-6 and in post- In two epithelial cell lines, CMT and Caco-2 cells, junctional localization of AF-6 was only found in areas of recently confluent cells or grow out zones, but not in densely confluent cell layers. This suggests that in these cells AF-6 may have a dynamic regulatory function in generating cell polarity and junction formation rather than a structural function in maintaining cellular polarity. This hypothesis is supported by the phenotype of AF-6-deficient mice developed independently by two different laboratories (31,32). Polarity of the embryonic ectoderm was strongly disturbed and the mice died at day 10.5 p.c. The analysis of AF-6 deficient mice revealed that AF-6 expression is restricted to a subset of epithelia during early embryogenesis with strong AF-6 expression only in areas where dynamic tissue rearrangements are ready to occur such as the primitive streak regions, the neural groove, and somites (31,32). This indicates that AF-6 is dynamically regulated and suggests a role for AF-6 in the formation, rather than in the maintenance of polarized cell layers. In the light of these results, it is interesting that we observed a cell type-specific difference in AF-6 localization.
In contrast to epithelial cells, AF-6 was constitutively present in cell-cell contacts of endothelial cells independent of their state of confluency, reflecting differences in the stability of junctions of the two cell types. In endothelial cells, tight and/or adherens junctions are expected to be subject to dynamic regulation in order to allow the migration of circulating leukocytes from the blood to subendothelial compartments. In the light of this special function of endothelial cells one might speculate that AF-6 could be involved in the reversible process of opening and closing interendothelial cell junctions. In this context it is intriguing that antibodies against JAM can block monocyte migration through endothelial cell layers in vitro and in vivo (9). It may be worthwhile to analyze whether the binding of JAM to AF-6 and the binding of the small GTPase ras to AF-6 are involved in this process.
We were surprised to find that no single PDZ domain of ZO-1 was able to bind to JAM while recombinant ZO-1 with all three PDZ domains as well as native ZO-1 bound efficiently.
Binding required the presence of PDZ domains 2 and 3 and the interconnecting flanking  GST-JAM fusion proteins were used to precipitate fusion proteins containing various regions of the ZO-1 protein synthesized in vitro as described in Fig. 2. A, Schematic organization of the ZO-1 fusion proteins that were used in these experiments. Each construct was fused to a 10 amino acid myc-tag as indicated. The single PDZ-domain constructs that were used in the experiments shown in Fig. 2         Growth of yeast colonies was scored as -(no growth), +/-(colony diameter below 0.5 mm), + (colony diameter between 0.5 and 2 mm), ++ (colony diameter larger than 2 mm). The JAM constructs comprised the full length cytoplasmic tail of JAM (aa 261-300, JAM/cyt), or truncation mutants thereof lacking carboxy-terminal aa residues (aa 261-299, JAM/cyt∆1; aa 261-297, JAM/cyt∆3; aa 261-291, JAM/cyt∆9). Constructs containing only carboxy-terminal aa residues of JAM comprised aa residues 290-300 (JAM/cyt11), aa 295-300 (JAM/cyt6), and aa 296-300 (JAM/cyt4). Yeast strains coexpressing VE-cadherin and p120 cas or α-catenin and VE-cadherin were used as positive and negative controls, respectively.
by guest on April 27, 2019