Recruitment of nectin-3 to cell-cell junctions through trans-heterophilic interaction with CD155, a vitronectin and poliovirus receptor that localizes to alpha(v)beta3 integrin-containing membrane microdomains.

Nectins present a novel class of Ig superfamily adhesion molecules that, cooperatively with cadherins, establish and maintain cell-cell adherens junctions. CD155, the cognate receptor for poliovirus, undergoes cell-matrix contacts by binding to the extracellular matrix protein vitronectin. The significant homology of nectins with CD155 prompted us to investigate the possibility of their interaction. We determined that nectin-3 binds CD155 and its putative mouse homologue Tage4 in cell-based ligand binding assays. Coculture of nectin-3- and CD155-expressing HeLa cells led to CD155-dependent recruitment of nectin-3 to cell-cell contacts. In a heterologous coculture system with CD155 expressing mouse neuroblastoma cells, HeLa cell-expressed nectin-3 was recruited to contact sites with CD155 bearing neurites. CD155 and nectin-3 colocalized to epithelial cell-cell junctions in renal proximal tubules and in the amniotic membrane. Efficient interaction depended on CD155 dimerization, which appears to be aided by cell type-specific cofactors. We furthermore found CD155 to codistribute with alpha(v) integrin microdomains on the surface of transfected mouse fibroblasts and at amniotic epithelial cell junctions. Our findings demonstrate the possible trans-interaction between the bona fide cell-cell adherens type adhesion system (cadherin/nectin) and the cell-matrix adhesion system (integrin/CD155) by virtue of their nectin-3 and CD155 components, respectively.

The CD155 gene is expressed in four isoforms. Two of the four alternatively spliced variants, CD155␣ and CD155␦, are membrane-bound and serve as poliovirus receptor, whereas the significance of two secreted versions, CD155␤ and -␥, is not clear (20). CD155 mRNA is widely expressed, although in relatively low amounts, in human and CD155tg mouse tissues, including brain, spinal cord, heart, skeletal muscle, kidney, spleen, leukocyte, liver, lung, and placenta (20 -23). In humans, CD155 is prominently expressed by enterocytes and cells of gastrointestinal lymphatic tissues (24). The ectodomain of CD155 establishes cell-matrix contacts by interaction with vitronectin (25). Although the overall structure of CD155 as an Ig superfamily protein and the similarity to nectins suggests a possible role in cell-cell adhesion, no binding partners that would support such a notion have been identified thus far. In fact, in contrast to nectins, there is no evidence for CD155mediated trans-homophilic cell-cell adhesion or for CD155 cishomodimerization (8 -10). On the other hand, the short cytoplasmic domains of CD155␣ and CD155␦ interact with Tctex-1, a light chain subunit of the dynein motor complex (26). This interaction is thought to mediate retrograde axonal transport of CD155 containing endocytic vesicles (26). In addition, the cytoplasmic domain of CD155␣ binds the mu1B subunit of the clathrin adaptor complex, an interaction that is responsible for sorting of the CD155␣ isotype protein to basolateral membranes in epithelial cells (27).
Studies of CD155 have been complicated by uncertainty over whether a functional rodent homologue of CD155 exists, a conundrum that has restricted and even misled earlier work. It is now clear that a mouse gene termed MPH, which was long held to be the mouse homologue of CD155, is, in fact, the homologue of human nectin-2 (formerly PRR2 or PVRL2) (4,5). The rodent Tage4 gene, sharing only 46% amino acid identity with CD155, has recently been proposed as the possible CD155 homologue based on conserved gene structure and syngenic chromosomal localization (19). However, no functional relationship between these two molecules could be thus far established. Mice are not a natural host for poliovirus, and cultured mouse cells cannot be infected with poliovirus. Therefore, the development of transgenic mice expressing the human CD155 gene has greatly facilitated the study of CD155 function (21,23). CD155tg mice are highly susceptible to poliovirus infection and develop disease symptoms very similar to poliomyelitis in humans, indicating that the human transgene functions similarly in mice as it does in humans.
Here we show that nectin-3 interacts with CD155, leading to unidirectional recruitment of nectin-3 to cell-cell contacts with neighboring CD155-expressing cells. This interaction was found to depend on dimerization/multimerization of CD155, which appears to be promoted by cell type-specific cofactors. Mouse Tage4 (referred to below as mCD155) showed similar affinity for nectin-3, providing the first evidence for a functional relationship between CD155 and its mouse counterpart. In vivo, nectin-3 and CD155 colocalized at lateral cell-cell junctions in epithelia of proximal kidney tubules and of the amniotic membrane. We determined that CD155 and ␣ v integrin colocalize to membrane microdomains on the cell surface of mouse fibroblasts and to cell-cell and cell-matrix junctions in the amniotic membrane. We propose that CD155 and nectin-3 are mediators of trans-heterophilic cell-cell adhesion and may provide a link between the classical cell-cell adhesion system (cadherin/nectin) and the cell-matrix adhesion system (integrin/CD155).
Ligand Binding Assays-Receptor-AP in situ ligand binding was performed as previously described (33,34), with some modifications. Briefly, HEK 293 cell supernatants containing AP fusion protein probes were diluted with growth medium to obtain a uniform AP activity of 2500 A 405 /ml*h for all probes. COS-1 cells transiently transfected with expression vectors for individual target proteins were incubated with AP fusion protein probes for 1 h at 37°C. The probes were then aspirated, and cells were washed five times with warm Hanks' balanced salt solution (HBSS), containing 0.5 mg/ml bovine serum albumin. The cells were then fixed with 4% formaldehyde and 60% acetone in PBS for 2 min and washed twice with HBSS/bovine serum albumin, followed by heating of the culture dishes at 65°C for 4 h in order to inactivate cell endogenous, heat-labile AP. The monolayers were washed twice with double-distilled H 2 O, and bound AP fusion probes were detected by incubation with 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium colorimetric AP substrate (Sigma). Alternatively, stable HEK 293, HeLa, L(tkϪ), and PC12 cell lines expressing individual target proteins were produced and processed as described above. To test the effect of cross-linked CD155 binding to nectin-3, CD155-AP supernatant was preincubated with anti-AP mAb 8B6 (1:500) for 1 h at room temperature before being added to the target cells. Alternatively, a fusion protein between the CD155 ectodomain and the Fc portion of mouse IgG2a (CD155-Fc) was used instead of CD155-AP (35). This protein is expressed as a dimer due to the biological activity of the Fc region (35). Binding of CD155-Fc was assessed by subsequent incubation with anti-mouse IgG secondary antibodies conjugated to AP. For antibody blocking of nectin-3-AP binding to CD155-expressing cells, the cells were preincubated with 10 g/ml of an unrelated mouse mAb or mAb D171, P286, P403, or P275 for 30 min at 37°C before the addition of nectin-3-AP.
In order to quantitate cell-bound AP fusion proteins, confluent 6-well plates of HeLa, HeLa-CD155, L(tkϪ), and L(tkϪ)-CD155 cells were incubated with AP fusion probes as above. However, instead of fixing the cells after the washing step, cells were lysed in 10 mM Tris, pH 7.4, 1% Triton, and the lysate was incubated at 68°C for 45 min. The amount of AP activity of each binding reaction was determined spectrophotometrically at 405 nm using 9 mg/ml p-nitrophenylphosphate (in 1 M diethanolamine, pH 9.8, 0.5 mM MgCl 2 ) as a colorimetric, water-soluble substrate. In parallel, 10 5 cell equivalents of total lysate were separated on 7.5% SDS-PAGE followed by Western blot analysis with CD155-specific antiserum NAEZ-8 (1:5000)/anti-rabbit horseradish peroxidase (1:10,000). Immune reactive bands were detected using Lumi-Light chemiluminescence substrate (Roche Molecular Biosciences).
lines were seeded on glass coverslips at a ratio of 3:1:1. Alternatively, HeLa and HeLa-nectin-3 cells were seeded 2 days prior to superseeding with Neuro2a-CD155 cells. Cells were incubated for an additional 2 days and then fixed with either Ϫ20°C cold methanol/acetone (1:1) for 30 min or 4% paraformaldehyde for 10 min, followed by permeabilization with 0.1% Triton X-100. Adult kidneys and amniotic membranes (embryonic day 15.5) were recovered from pregnant, CD155tg female mice (ICR-PVRTg21) (21). Human amniotic membranes used in this study were discarded clinical samples from term deliveries (kindly provided by Dr. Diana Roth, Maimonides Medical Center, Brooklyn, NY). Specimen were fixed with 4% paraformaldehyde in PBS for 4 h, followed by perfusion with 25% sucrose in PBS. Tissues were then frozen in OCT cryo-protectant (Tissue-Tek™), and 10-m-thick cryosections were collected on poly(L)-lysine-coated slides. Sections were air-dried and postfixed in cold 1:1 methanol/acetone for 90 min at Ϫ20°C before processing. For immunofluorescence detection, sections were blocked with PBS containing 5% normal horse and 2% normal goat serum (PBS/HG) for 2 h at room temperature, followed by overnight incubation at 4°C with primary antibodies in PBS/HG at the dilutions stated above. After washing five times for 10 min with PBS, the sections were incubated with secondary antibody in PBS/HG for 1 h at 37°C. The primary/secondary antibody combinations were either NAEZ-8/ mAb 103-A1 with anti-rabbit (Alexa488)/anti-rat (Cy3), mAb D171/mAb 103-A1 with anti-mouse (Alexa488)/anti-rat (Cy3), or mAb 18/anti-␣ v ␤ 3 with anti-mouse(Cy3)/anti-rabbit (Alexa488). The sections were washed five times for 20 min with PBS and mounted with Immu-Mount (Shandon, Pittsburgh, PA). Images were obtained with a Bio-Rad Radiance laser-scanning confocal microscope using the LaserSharp 2000 software package and processed with Adobe Photoshop software. Bis-(sulfosuccinimidyl)suberate (BS3) Surface Cross-linking of CD155-10-cm dishes with confluent monolayers of HeLa-CD155 and L(tkϪ)-CD155 cells were washed three times with ice-cold HBSS, followed by incubation with 1 mM of the chemical cross-linker BS3 (Pierce) in HBSS under gentle agitation for 30 min at 4°C. The monolayers were then washed twice with HBSS, and the cross-linking reaction was quenched with HBSS containing 50 mM Tris, pH 7.4. Cells were scraped off the plates and collected by centrifugation at 500 ϫ g for 5 min. The cell pellets were lysed by boiling in gel loading buffer (50 mM Tris, pH 6.8, 2% SDS, 100 mM dithiothreitol, 10% glycerol, 0.1% bromphenol blue), and 10 5 cell equivalents of lysate were subjected to CD155 Western blot as described above.

Interaction of CD155 and Nectin-3 in Cell-based Ligand
Binding Assays-In order to test for interactions between nectin/CD155 family members, we employed a method developed by Flanagan and Leder (33,34) that makes use of recombinant receptor-AP fusion proteins as affinity probes. Four affinity probes were expressed in HEK 293 cells by fusing the ectodomains of CD155, mouse nectin-1, mouse nectin-2, and mouse nectin-3 to SEAP (see Fig. 1A), a genetically modified derivative of the human PLAP-1 gene (36). The fusion proteins are targeted for secretion into the medium by virtue of the signal sequence endogenous to each of the receptor moieties. Between 2500 and 10,000 A 405 /ml*h of AP activity was detected in the various culture supernatants, compared with 1-2 A 405 /ml*h in supernatants of untransfected cells. Western blot analysis with anti-PLAP-1 mAb 8B6 was performed in order to ascertain the predicted size of the expressed proteins (Fig. 1B). All fusion proteins were detected as a single immune reactive band with no visible truncation products. Although the calculated molecular masses of the recombinant proteins were between 90 and 100 kDa, the apparent molecular masses were found to be between 105-140 kDa due to varying amounts of N-linked glycosylation (Fig. 1B).
Given the recent reports of heterophilic interactions of nectin-3 with both nectin-1 and nectin-2 (1, 11, 12), we asked whether CD155, the closest known relative of nectin-2, would be capable of a similar interaction with nectin-3. For this purpose, COS-1 cells, transiently transfected with individual nectin/CD155 family members, were incubated with AP fusion proteins as indicated in Fig. 2. First, we tested whether our Cell-bound AP activity was detected by colorimetric staining with the AP substrate 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium for 2 h at room temperature. The assay confirmed the previously reported strong heterophilic trans-interactions between nectin-3 and nectin-1 (A and D), nectin-3 and nectin-2 (B), and the very weak or nonexistent homophilic trans-interaction of nectin-3 (C). Nectin-3-AP bound strongly to cells transfected with human or mouse CD155/Tage4 (E and F, respectively) but not to cells transfected with empty vector (H). Unfused AP showed no affinity to CD155 (G). Under the conditions of the assay, soluble monomeric CD155-AP did not interact with cell surface-expressed nectin-3 (I). Bars, 50 m.
experimental system can reproduce the previously established interactions between nectin family members. Indeed, intense staining, indicating a strong interaction, was observed between cell surface-expressed nectin-3 and the nectin-1-AP probe ( Fig.  2A) or the nectin-2-AP probe (Fig. 2B) as well as between nectin-1 and nectin-3-AP (Fig. 2D). Under the conditions of the assay, virtually no homophilic binding resulted from incubation of COS-nectin-3 cells with nectin-3-AP (Fig. 2C). These results support previous results of Satoh-Horikawa and colleagues, who have shown that nectin-3 and nectin-1 or nectin-3 and nectin-2 predominantly undergo trans-heterophilic rather than trans-homophilic interactions (1). When COS-1 cells were transfected with either human CD155 or mouse CD155/Tage4, significant binding of nectin-3-AP resulted from both proteins (Fig. 2, E and F). Judging from the intensity of the staining, the intraspecies interaction between mCD155/Tage4 and nectin-3-AP (containing the mouse nectin-3 ectodomain) appeared to be slightly more efficient (Fig. 2, compare F and E). To rule out the possibility that the observed interactions might be mediated by the AP moiety of our fusion proteins, we incubated CD155-transfected cells with untagged AP protein. As expected, no binding was observed (Fig. 2G). Likewise, nectin-3-AP did not bind to COS-1 cells transfected with an empty expression vector (Fig. 2H), ruling out binding to COS-1 endogenous proteins. It is worth noting that in the reverse binding assay, in which CD155-AP was used to detect nectin-3 expression, we did not observe any significant interaction (Fig. 2I). The reason for this seemingly contradictory result is based on the failure of the soluble CD155-AP protein to form homodimers, as explained below.
In order to confirm the specificity of the interaction between CD155 and nectin-3, PC12-CD155 cells were preincubated with a panel of CD155-specific mAbs, followed by a nectin-3-AP binding assay (Fig. 3). Three mAbs, D171, P286, and P403, reduced binding of nectin-3-AP binding by 54, 69, and 68%, respectively, whereas mAb 275 had no effect. Interestingly, all three inhibitory antibodies are known to bind to the N-terminal Ig-like domain and also block binding of poliovirus to CD155, an interaction that is solely mediated by this domain (30,31,37). This result suggests that, similarly to the interaction between other nectins (1-3), the binding site for nectin-3 lies within the N-terminal Ig-like domain of CD155.
The above findings present the first evidence for an interaction of CD155 with a bona fide cell-cell adhesion molecule, nectin-3, and for a functional homology between human CD155 and its mouse counterpart. We therefore propose that Tage4 be referred to as mCD155.
Interaction of Nectin-3 and CD155 Requires Dimerized CD155-To our surprise, we noticed that the reverse ligand binding assay, using soluble CD155-AP to detect nectin-3 expressed on the cell surface, did not result in significant binding (Figs. 2I and 4F) on any of the cell lines tested (HeLa, HEK 293, COS-1, L(tkϪ), PC12). During a parallel analysis of the interaction between nectin-2 and nectin-3, we made the curious observation that preincubation of HEK 293-nectin-2 cells with nectin-2 rat mAbs 17B10 and 6B3 (8), instead of blocking, immensely enhanced the binding of nectin-3-AP to nectin-2 (Fig. 4, compare A and C). The same phenomenon was observed were incubated with growth medium containing a 10 g/ml concentration of an unrelated mouse mAb (first and second bars) or 10 g/ml CD155-specific mAb D171, P286, P403, or P275. After 30 min, antibodies were removed, and cells were incubated with nectin-3-AP for 1 h. Cells were washed extensively and then lysed, followed by heat inactivation of any endogenous AP activity. Heat-stable AP activity originating from cell-bound nectin-3-AP was quantitated spectrophotometrically using the water-soluble AP substrate para-nitrophenylphosphate. The amount of nectin-3-AP binding in the absence of CD155-specific mAbs was set to 100%. Three mAbs, D171, P286, and P403, were found to inhibit nectin-3-AP binding to CD155 by 54, 69, and 68%, respectively. mAb P275 had no inhibitory effect. Results are the average of three independent determinations. on L(tkϪ)-nectin-2 cells (data not shown). This effect was specific for the interaction of nectin-2 with nectin-3-AP, since CD155-AP showed no binding to nectin-2 before or after antibody preincubation with 17B10/6B3 (Fig. 4, compare D and B). Furthermore, no cross-reactivity of mAbs 17B10 and 6B3 with nectin-3 was detected (data not shown). Since nectin-2 is already present in form of a dimer on the cell surface (9, 10), our data indicate that clustering of nectin-2 may be necessary for efficient trans-interaction with nectin-3.
The CD155-specific antibodies available to us did not cause a similar stimulation of nectin-3-AP binding to CD155-expressing cells; all of these antibodies were found to be inhibitory to this interaction or had no effect (Fig. 3). Since soluble CD155 does not form homodimers as nectins do, we speculated that the reason for the failure of CD155-AP binding to nectin-3 (Figs. 2I and 4F) is due to the monomeric nature of CD155. To address this issue, we cross-linked our CD155-AP probe by preincubation with AP-specific mAb 8B6 before adding the fusion protein to HEK 293-nectin-3 cells. We found that antibody-cross-linked CD155-AP now efficiently bound to nectin-3 (Fig. 4, compare F and G). To confirm this observation, we next tested the binding of a recombinant CD155 to which the Fc portion of mouse IgG2a was fused (CD155-Fc) (35). Due to the properties of the Fc region, this fusion protein spontaneously forms dimers. Binding of CD155-Fc to HEK 293-nectin-3 cells was detected with a secondary anti-mouse IgG antibody conjugated to AP. As can be seen in Fig. 4H, CD155-Fc bound well to nectin-3 expressed on the cell surface of HEK 293 cells, similar to the binding of the antibody-cross-linked CD155-AP (Fig. 4G). Our results indicate that CD155 dimerization/oligomerization is essential for efficient binding of nectin-3.

Cell Type-specific Cofactors Are Required for the Proper Display of CD155 on the Cell Surface and for the Interaction with
Nectin-3-When the binding of nectin-3-AP to different CD155expressing cell lines was compared, a considerable and reproducible difference was noticed. This was particularly striking in the case of CD155-expressing mouse L(tkϪ) versus human HeLa cells (Fig. 5). Although L(tkϪ)-CD155 and HeLa-CD155 cells both expressed high levels of CD155 (Fig. 5B, third and fourth lanes; compare with HeLa cell endogenous CD155 in first lane), very little nectin-3-AP binding above background (Fig. 5A, second bar) was observed on L(tkϪ)-CD155 cells (compare Fig. 5A, third and fourth bars). We do not believe this phenomenon to be a species-specific effect of mouse versus human cells, since CD155 expressed in other rodent cells, like mouse Neuro2a and rat PC12 cells, was efficiently recognized by nectin-3-AP ( Fig. 3 and data not shown). We deemed it more likely that CD155 expressed on L(tkϪ) cells is not presented to nectin-3 in a proper higher order structure (dimer, oligomer, or complex with other proteins). To address this point, we carried out surface cross-linking of L(tkϪ)-CD155 and HeLa-CD155 cells with the chemical cross-linker BS3 (Fig. 6). Four high molecular weight complexes containing CD155 were obtained from HeLa-CD155 cells (Fig. 6, third lane). The smallest, and most prominent, of these complexes was 155-160 kDa in size, which corresponds to the expected size of a CD155 dimer (Fig.  6, third lane, asterisk). Three additional higher molecular mass CD155 containing complexes ranging from 250 to ϳ400 -450 kDa were observed (Fig. 6, third lane, arrowheads). The components of these complexes, besides CD155, are not known at present. In contrast, CD155 expressed on L(tkϪ) cells appeared to be exclusively present as the typical 75-80-kDa monomer (Fig. 6, fourth lane). We suggest that a correlation exists between the absence of these high molecular weight CD155 complexes on L(tkϪ) cells and the inability of nectin-3-AP to recognize CD155. This is most likely due to missing, as yet unidentified, cofactors on L(tkϪ) cells.  (1,2,8,9). Furthermore, we and others have shown that nectin-3 is also present at specialized heterotypic cell-cell junctions such as Sertoli cellspermatid junctions in testis (11) 1 or at synaptic puncta adherentia (12). Therefore, we asked whether CD155 and nectin-3 colocalize at sites of cell-cell contact. For this purpose, we devised an in vitro coculture system that allowed us to assay for trans-interactions between cells. When nectin-3 expressing HeLa cells (HeLa-nectin-3) were cocultured with parental HeLa cells until confluence (Fig. 7, A-D), nectin-3 distributed evenly to all junctions between neighboring cells (Fig. 7, B and  D). There was no noticeable preference of nectin-3 localization to junctions between HeLa-nectin-3 cells over junctions between HeLa-nectin-3 and parental HeLa cells (Fig. 7, B and D). This distribution is similar to previously reported nectin-3 expression in L cells (1). We furthermore detected significant nectin-3 immunoreactivity on the cell membranes not involved in cell-cell contacts (Fig. 7B). To test for interaction of nectin-3 with CD155, HeLa-nectin-3 cells were cocultured together with HeLa-CD155 cells and parental HeLa cells (as an internal control) (Fig. 7, E-H). The cells were then stained simultaneously for nectin-3 and CD155 by indirect immunofluorescence with rat mAb 103-A1 and mouse mAb D171 and analyzed by fluorescence microscopy (see "Materials and Methods"). We found that wherever HeLa-nectin-3 cells came into contact with HeLa-CD155 cells, virtually all nectin-3 was unidirectionally attracted to cell contacts with the CD155-expressing neighboring cells (Fig. 7F). Fig. 7, E-H, shows a typical example of four HeLa-nectin-3 cells (labeled 3 in Fig. 7G) adjacent to three HeLa-CD155 cells (labeled 5 in Fig. 7G). CD155 protein remained evenly distributed over the whole surface of the cell regardless of the identity of its neighbor. In contrast, nectin-3 protein was redistributed to the contact sites with the HeLa-CD155 cells and excluded from the junction with other HeLanectin-3 cells or the junction with parental nontransfected  F; red), or nectin-2 (J; red). When cocultured with parental HeLa cells, nectin-3 expressed by HeLa-nectin-3 cells distributed evenly to all cell junctions. In addition, significant nectin-3 was found associated with free cell membranes (B). When HeLa-nectin-3 cells were cocultured with HeLa-CD155 cells, nectin-3 was found exclusively at junctions with HeLa-CD155 cells, indicating a strong preference for trans-heterophilic interaction between CD155 and nectin-3 (F and H). In contrast, when HeLa-nectin-2 cells were cocultured with HeLa-CD155, nectin-2 preferentially localized at junctions between HeLa-nectin-2 cells but not between HeLa-CD155 and HeLa-nectin-2 cells, suggesting a strong preference for trans-homophilic interaction of nectin-2 (J and L). 5, HeLa-CD155 cell; 3, HeLa-nectin-3 cell; 2, HeLa-nectin-2 cell. The white lines in G and K indicate the approximate location of cell junctions. Bars, 20 m.

Nectin-3 Is Recruited to Cell-Cell Contacts through Interaction with CD155-The members of the nectin family have been shown to localize to homotypic cadherin-based cell-cell junctions both in cultured cells and in vivo
HeLa cells (Fig. 7, F and H). Furthermore, any nectin-3 previously found associated with the free cell surface was now incorporated into junctions with HeLa-CD155 cells. We conclude that the efficient recruitment of nectin-3 is a direct result of the interaction with CD155 on HeLa-CD155 cells. An opposite picture emerged when HeLa-CD155 cells were cocultured with HeLa-nectin-2 and parental HeLa cells (Fig. 7, I-L). Here, nectin-2, which does not interact with CD155 (data not shown), was found to be excluded from junctions with HeLa-CD155 or parental HeLa cells, in favor of homophilic interaction with nectin-2 on neighboring HeLa-nectin-2 cells (Fig. 7, J and L). Together, these data indicate that nectin-3 preferentially undergoes trans-heterophilic interactions with CD155 over transhomophilic interactions, whereas nectin-2, under the experimental conditions of the assay, engages only in the transhomophilic mode of interaction.
We extended our coculture system to the analysis of heterotypic cell-cell junctions by using a mouse neuroblastoma cell line, Neuro2a, stably transfected with CD155 (Neuro2a-CD155) together with HeLa-nectin-3 and parental HeLa cells. In contrast to the "HeLa-only" coculture described above, here HeLa-nectin-3 and HeLa cells were seeded first and allowed to grow for 2 days to a 70% cell confluence, during which time nectin-3 distributed evenly to all cell-cell junctions (similarly to Fig. 7B). Neuro2a-CD155 cells were then superseeded, followed by an additional 2 days of coculture. Cells were processed as above and analyzed by laser-scanning confocal microscopy (Fig.  8). Under the given culture conditions, a portion of Neuro2a-CD155 cells were found to extend long neurites. Like the rest of the cell surface, these neurites were richly decorated with CD155 protein (Fig. 8A, a1 and a2). Wherever CD155-positive neurites extended over a HeLa-nectin-3 cell, almost the entire nectin-3 pool of that cell was found to be associated with the crossing section of the CD155-bearing neurite (Fig. 8, B (b1 and   b2) and D (d1 and d2)). Fig. 8C (c1 and c2) depicts two CD155expressing neurites that traverse the middle of several HeLanectin-3 cells and in some cases cross over the region of the underlying cell's nucleus (arrowheads). As an interesting implication of this result, we conclude that the junction assembled by CD155 and nectin-3 can form anywhere on the cell surface. Our coculture data suggest to us that nectin-3 is being actively recruited away from HeLa-nectin-3 homotypic cell junction and incorporated into HeLa-nectin-3/Neuro2a-CD155 heterotypic cell junctions. Thus, nectin-3 protein at HeLa-HeLa cell junctions undergoes only weak and transient interactions, and upon encountering CD155 on Neuro2a-CD155 cells, it will be incorporated into a more stable adhesion complex. This leads to a concentration of nectin-3 along CD155-expressing surfaces until all nectin-3 has been has "mopped up" through interaction with CD155.
Nectin-3 and CD155 Colocalize at Lateral Cell-Cell Junctions of Kidney and Amniotic Epithelia-In order to confirm the physiological significance of the interaction between nectin-3 and CD155, we searched for sites of colocalization of both proteins in mouse tissues. Cryosections of CD155tg mouse kidney and amniotic membranes were stained simultaneously with antibodies 103-A1 and NAEZ-8 and analyzed by laserscanning confocal microscopy (Fig. 9). Basolateral expression of CD155 was seen in epithelial cells lining the proximal tubules of the kidney, decreasing in intensity toward the most apical portions of the lateral cell-cell junctions (Fig. 9A, a1). Nectin-3 was most concentrated in the anterior aspects of lateral cell-cell junctions (Fig. 9B, b1), consistent with its localization at cadherin-based adherens junctions that has been described in other epithelia (1). The expression level of nectin-3 decreased toward the basal aspect of the lateral membranes (Fig. 9B, b1). Significant colocalization of CD155 staining and nectin-3 staining could be detected in the medial region of the lateral cell contacts (Fig. 9C, c1). A very similar observation was made in the epithelium of the amniotic membrane, except that here CD155 was found mainly along the lateral cell-cell junctions and less at the basal side of the cell (Fig. 9, D-F, d1-f1, and  d2-f2). We consider it possible that nectin-3 and CD155 may be involved in maintaining the distinct compartmentalization and polarity of the epithelial cell membrane in the transition zone between the apical and basolateral membrane region of epithelial cells.
cis-Colocalization of CD155 with ␣ v ␤ 3 Integrin-The basolateral distribution of CD155 in epithelia (24,27,38), reminiscent of an integrin expression pattern, and the fact that both CD155 and ␣ v integrins serve as vitronectin receptor led us to speculate on a functional connection between CD155 and integrins. We therefore set out to analyze their subcellular expression and possible colocalization in cultured cells. A mouse mAb specific for the ␣ v ␤ 3 integrin heterodimer (clone LM609; Chemicon) and a rabbit polyclonal antibody against ␣ v ␤ 3 were found to intensely and specifically stain L(tkϪ) cells cotransfected with human ␣ v and ␤ 3 integrin expression vectors (data not shown). When the cells were cotransfected to express ␣ v , ␤ 3 , and CD155, the observed punctate surface staining for ␣ v ␤ 3 integrin (Fig. 10, A and D) coincided almost completely with that of CD155 (Fig. 10, B and E). Although we presently do not know whether they physically interact, the fact that both CD155 and ␣ v ␤ 3 are targeted to identical subcellular membrane microdomains suggests a functional relationship, possibly their presence in a common supermolecular adhesion complex. Moreover, no significant colocalization was observed between ␣ v ␤ 3 and nectin-1, nectin-2, or nectin-3, attesting to the specificity of the observed colocalization with CD155 (data not shown).
Finally, we determined that CD155 and ␣ v integrin colocalize at the basolateral membranes in human amniotic epithelium (Fig. 10, G-I and insets). Unlike the colocalization with nectin-3 that only occurred at lateral cell-cell contacts (Fig. 9F), CD155 and ␣ v colocalized in the entire basolateral region. Adequate staining in human amniotic epithelium was only obtained with a polyclonal antibody against ␣ v ␤ 3 and not with the ␣ v ␤ 3 -specific mouse mAb LM609. Thus, it is not clear whether the observed staining reflects expression of ␣ v ␤ 3 or that of another ␣ v -containing heterodimer cross-reacting with this polyclonal antibody. For instance, this polyclonal antibody very efficiently stained focal contacts in HeLa (data not shown), a cell line that reportedly does not express ␣ v ␤ 3 but does express ␣ v ␤ 5 (39,40). Unfortunately, we were unable to test colocalization of nectin-3 and CD155 in human amnion due to the unavailability of suitable nectin-3 antibodies specific for the human protein.  , a1), whereas nectin-3 was found most concentrated in apical aspects of the lateral cell-cell junctions (B and b1). A significant area of colocalization was observed at the lateral membranes (C, c1; yellow). Similarly, CD155 (D, d1, and d2) and nectin-3 (E, e1, and e2) colocalized at lateral cell-cell junctions of the amniotic epithelium (F, f1, and f2; yellow). A tangential midlevel section through an amniotic epithelial cell (due to twisting of the specimen) shows nearly complete overlap of CD155 and nectin-3 expression (d2-f2). Images represent single 0.2-m optical z-sections taken with a Bio-Rad Radiance laser-scanning confocal microscope. Bars, 25 m (C) (insets magnified ϫ 3) and 50 m (F) (insets magnified ϫ 3).

DISCUSSION
In this study, we have identified and characterized a novel interaction between nectin-3 and CD155, the latter being a receptor for poliovirus and for the extracellular matrix protein vitronectin. Our evidence suggests that CD155, which appears to be part of an ␣ v ␤ 3 integrin-containing complex, may function as a bona fide cell-cell adhesion molecule by interacting with nectin-3, an adhesion receptor present at cadherin-based adherens junctions. We propose that this interaction may play a role in mediating heterophilic adhesion between the integrin and the cadherin/nectin adhesive systems. Direct interactions such as between the ␣ 2 ␤ 1 or ␣ E ␤ 7 integrins and E-cadherin have been described previously (41,42). Our results provide a first indication that an interaction between the cadherin and integrin system may also occur through their associated proteins.
It has been shown recently that CD155 exhibits affinity for the extracellular matrix protein vitronectin, an interaction that may be involved in establishing cell-matrix contacts (25). The major receptors for vitronectin are members of the ␣ v integrin family, in particular ␣ v ␤ 3 and ␣ v ␤ 5 . Integrins are considered the quintessential mediators of adhering epithelial cells to the underlying extracellular matrix (43). Their presence has been shown in at least two very distinct adhesion complexes. These are focal adhesions, in which they bridge the extracellular matrix and the actin cytoskeleton (44), and hemidesmosomes (45,46) that tie into the intermediate filament network of the epithelial cell (see Ref. 47 and references therein). Our finding of cis colocalization of CD155 and ␣ v ␤ 3 integrin on transfected mouse fibroblasts together with the fact that both proteins serve as vitronectin receptors seems to indicate their close association, possibly in a multiprotein adhesion complex. In addition, the CD155-specific antibody NAEZ-8 blocks the infection of rhabdomyosarcoma cells by Coxsackie virus A9, 4 a virus that uses ␣ v ␤ 3 as its cognate cellular receptor (48). Such a blocking effect could best be explained by steric hindrance, due to the close association of CD155 and ␣ v ␤ 3 integrin. However, it is presently unknown whether CD155 and ␣ v ␤ 3 integrin interact physically. Interestingly, ␣ v ␤ 3 , ␣ v ␤ 5 , CD155, and their mutual ligand vitronectin are all up-regulated in glioma cells, and a positive correlation between metastatic/invasive properties of gliomas and ␣ v ␤ 3 expression is well documented (49 -53). In the future, it will be interesting to address the possible role of CD155 in such malignancies.
We have also begun to determine the mechanisms of the interaction that we have identified between nectin-3 and CD155. This heterophilic interaction is not a simple bimolecular reaction but rather requires the presence of CD155 in a higher order complex that facilitates dimer formation and proper display of CD155 on the cell surface. The correct presentation of CD155 needed for interaction with nectin-3 appears to be cell type-specific (Figs. 4 -6). L(tkϪ) cells were found to be unable to assemble such a CD155 complex, possibly due to the lack of an as yet unidentified cellular cofactor(s). As a result, no functional CD155-mediated nectin-3 binding sites were formed on L(tkϪ)-CD155 cells (Figs. 4 -6). Dimerization of binding partners alone, however, may not explain the complex interactions between nectin/CD155 family members. Although nectin-2 spontaneously forms homodimers in all cell lines that were tested (9, 10), cross-linking with anti-nectin-2 greatly facilitated binding of nectin-3-AP to HEK 293-nectin-2 cells (Fig. 4). Thus, cis-dimer formation may not be the sole requirement for efficient binding. Rather, the formation of higher order structures, such as nectin clusters or "lattices," may be necessary. For example, the requirement of integrins to form supermolecular clusters at focal contacts has been well documented (see Refs. 44 and 54 and references therein). We suggest that this may also be the case for trans-interactions mediated by nectin/CD155 family members.
It is attractive to speculate that cofactors associated in cis with CD155 such as ␣ v ␤ 3 integrin may aid in the formation of clusters of CD155 binding sites on the cell surface. Furthermore, the identification of CD155 as a vitronectin receptor (25) and its colocalization with ␣ v ␤ 3 that we report here point toward the possibility of a trimolecular vitronectin receptor (␣ v ␤ 3 /CD155).
Our preliminary results indicate that nectin-3-AP binding to HeLa-CD155 cells does not block poliovirus binding or infection (data not shown). Poliovirus virions avidly bind monomeric CD155 in vitro (29), an observation suggesting that monomeric CD155 is an efficient cellular receptor for the virus. This is supported by the fact that L(tkϪ)-CD155 cells, which only express CD155 monomers (Fig. 6), present an excellent sub-4 M. Roivainen, personal communication. strate for poliovirus binding and infection (17) 5 while not supporting nectin-3-AP binding (Fig. 5). Our data indicate that several distinct CD155 populations exist on HeLa cells (Fig. 6). The majority of CD155 is present as monomers that may serve as poliovirus receptors, whereas CD155 pools that contain dimerized CD155, possibly in a complex with other proteins, may constitute the binding sites for nectin-3.
Finally, we have shown that mouse CD155/Tage4 interacts with nectin-3 much in the same manner as human CD155 (Fig.  2). This result provides the first functional evidence that mouse CD155/Tage4 is, indeed, the authentic rodent homologue of human CD155. In this context, it is interesting to note that both CD155 and Tage4 are overexpressed in colon carcinoma (13,18).