Cadherin binding sites of plakoglobin: localization, specificity and role in targeting to adhering junctions

Epithelial cells exhibit two major types of adhering junctions, adherens junctions anchoring actin microfilaments and desmosomes anchoring intermediate filaments (IF) to the plasma membrane. In adherens junctions intercellular adhesion is caused by classical cadherins (e.g. E-cadherin), while in desmosomes it is maintained by desmosomal cadherins, desmogleins (Dsg1-3) and desmocollins (Dsc1-3) (Green and Jones, 1990; Schwarz et al., 1990; Geiger and Ayalon, 1992; Garrod, 1993; Koch and Franke, 1994).

Intracellular plaques of both types of adhering junctions contain one common component plakoglobin (Cowin et al., 1986). It consists of three structurally distinct regions, the unique amino- and carboxyl-terminal segments, and a 560-amino-acid-long central region comprising 13 arm repeats (Franke et al., 1989; McCrea et al., 1991; Butz et al., 1992; Peifer et al., 1994). This latter region is 85% homologous to a corresponding segment of β-catenin, a protein found only in adherens junctions. Recent experiments showed that plakoglo-bin is involved in a complex pattern of interactions with other proteins. A soluble cytoplasmic form of plakoglobin (and β-catenin) associates with α-catenin and/or the tumor suppressor protein APC via the central repeat domain (Hinck et al., 1994; Hulsken et al., 1994; Rubinfeld et al., 1995). In the adherens junction, plakoglobin is associated with E-cadherin and α-catenin, which provide anchorage for F-actin and α-actinin (Ozawa et al., 1989, 1990; Knudsen and Wheelock, 1992; Nagafuchi et al., 1994; Knudsen et al., 1995; Rimm et al., 1995). Recently a tyrosine kinase substrate, p120, was also found in the E-cadherin/plakoglobin/α-catenin complex (Reynolds et al., 1994; Shibamoto et al., 1995; Staddon et al., 1995). In desmosomes plakoglobin specifically interacts with Dsg and Dsc and apparently does not bind to α-catenin or p120 (Korman et al., 1989; Troyanovsky et al., 1993, 1994a,b; Kowalczyk et al., 1994; Mathur et al., 1994). Notably, β-catenin is unable to interact with desmosomal cadherins.

Experiments with dominant negative cadherin mutants showed that the association of plakoglobin with cadherins is a critical step in the assembly of adherens junctions and desmosomes. Over-expression of the intracellular segment of classical cadherins in epithelial cells inhibits cadherin-dependent adhesion, but not desmosome assembly (Kinter, 1992; Fujimori and Takeichi, 1993; Amagai et al., 1995). Similarly, expression of the chimeric protein CoDsg, comprising the transmembrane domain of connexin32 and the carboxyl-terminal intracellular segment of the Dsg1, inhibited formation of endogenous desmosomes, but not adherens junctions (Troyanovsky et al., 1993, 1994a). Catenin-binding domains of classical cadherins and a homologous C-domain of Dsg1, both of which directly interact with plakoglobin, are responsible for these effects (Troyanovsky et al., 1994a; Mathur et al., 1994; Aberle et al., 1994; Stappert and Kemler, 1994; Jou et al., 1995). Disruption of adherens junctions and desmosomes by cadherins suggested the existence of distinct Dsg-specific and classical cadherin-specific binding sites in plakoglobin. The first direct evidence to support this hypothesis was obtained recently. It was shown that a carboxyl-terminal truncated mutant of plakoglobin, containing only three complete amino-terminal arm repeats, is able to associate with Dsg (Chitaev et al., 1996; Witcher et al., 1996), while at least six arm repeats are required for binding to classical cadherins (Sacco et al., 1995; Wahl et al., 1996). In the present work we show that distinct E-cadherin and Dsg binding sites are localized in the first five arm repeats of the plakoglobin molecule. The rest of the plakoglobin arm repeat region, containing at least two cryptic cadherin binding sites, is important for its targeting.

The plasmid BlSyPg, encoding a chimeric protein consisting of the synaptophysin and plakoglobin, has been described (Chitaev et al., 1996). For synaptophysin insert mutagenesis a plasmid BlSySM, containing the unique SalI and MluI sites at the positions of the start and stop codons of synaptophysin, respectively, was constructed. Then this plasmid was cut by MluI/XbaI and the insert was inserted into corresponding positions of the wild-type plakoglobin cDNA (Franke et al., 1989) using PCR-mediated site-directed mutagenesis.

For β-catenin/plakoglobin replacement mutagenesis a complete cDNA encoding mouse β-catenin (Butz et al., 1992) was amplified from mouse liver cDNA. The following steps in this mutagenesis, as well as deletion mutagenesis of the Dsg-binding region, were done using PCR-mediated site-directed mutagenesis.

The construction of the expression plasmids coding for the intact plakoglobin (Pg) and its fragments (Pg1/580, Pg1/147, Pg114/292, Pg305/505, Pg505/672, Pg580/744) or cytoplasmic segments of Dsg, Dsc1a and E-cadherin (HCDg, HCDc, HCUv) were as described (Chitaev et al., 1996). Some additional clones were constructed using conveniently located restriction endonuclease sites. The plasmid pQPg505/744 was constructed by ligation of the BglII/HindIII fragment of the pQPg580/744 with the appropriately digested pQPg505/672 vector. The 1,100 bp BclI/SacI fragment was excised from pQPg and inserted into the appropriate sites of the vector pQE-32 (Qiagen, Chatsworth, CA). The resulting plasmid was cut by SacI/HindIII and ligated with the corresponding fragment of the pQ580/744, to obtain a plasmid pQ305/744. For construction of the pQPg505/683, the 500 bp BglII/blunt-ended HindIII fragment of the pQPg505/744 was replaced with the BglII/blunt-ended BglI fragment taken from the same plasmid. The recombinant histidine-tagged proteins were isolated and analyzed by SDS-PAGE, essentially as described (Chitaev et al., 1996). Correct construction of all recombinant plasmids was verified by restriction endonuclease mapping and/or nucleotide sequencing.

Transfection of the A-431 cells, as well as the selection, growth, immunofluorescence microscopy and immunoprecipitation of stably transfected cell clones, have been described (Troyanovsky et al., 1993; Chitaev et al., 1996). The following primary antibodies were used: (i) polyclonal rabbit and murine monoclonal Sy38 antibodies against synaptophysin (DAKO, Hamburg, FRG); (ii) rabbit and monoclonal murine U114 antibodies against recombinant human Dsc1a (kindly provided by M. Demler and A. Schmidt, Heidelberg, FRG); (iii) rabbit pan-cadherin antibody (Sigma, St Louis, MO); (iv) rabbit antibodies recognizing α-catenin (kindly provided by Dr Wysolmerski, St Louis, MO); (v) murine mAb Dg.3.10 against bovine and human Dsg; murine mAb Dsc 210.2.9 against bovine Dsc; murine mAb PG5.1 against human plakoglobin; mixture of murine mAbs 2.15, 2.17 and 2.19 against desmoplakin (for references of these monoclonal antibodies, see Troyanovsky et al., 1994); (vi) murine mAbs against E-cadherin, α-catenin and p120 (Transduction Laboratories, Lexington, KY); monoclonal antibodies H1 against cytokeratin 8 (Troyanovsky et al., 1989).

In vitro binding assays were described previously (Chitaev et al., 1996). In brief, plakoglobin or its fragments were immobilized on a 96-well dish and incubated with increasing or fixed amounts of the different recombinant cadherin tails or fragment Pg580/744. Binding was detected by an ELISA assay with either DC210.2.9 or PG5.1 monoclonal antibodies. In competition experiments, immobilized plakoglobin or its fragments were incubated with the fixed amount of HDsg in presence of increasing amounts of the Pg580/744. For the overlay assay, plakoglobin fragments (0.8 µg per lane) were separated by SDS-PAGE and electroblotted onto nitrocellulose. Membranes after treatment with 3% bovine serum albumin were incubated with 20 µg/ml Pg580/744 for 60 minutes at room temperature. The binding was detected using the alkaline phosphatase system (Promega, Madison, WI).

We have described a chimeric protein, SyPg, where synaptophysin is fused with the entire plakoglobin molecule. Synaptophysin is an integral transmembrane protein of the presynaptic vesicles containing four membrane-spanning regions (transmembrane domains) and cytoplasm-exposed amino and carboxyl termini (Fig. 1A,D). The presence of transmembrane domains is sufficient for accurate targeting of synaptophysin into small synaptic-like vesicles (Leube et al., 1994; Leube, 1995). Fusion of either of the synaptophysin termini with different proteins allows their exposure on the surface of these vesicles (Troyanovsky et al., 1993; Leube, 1995). We have previously found that the SyPg chimera efficiently integrates into the vesicles of transfected epithelial cells and binds to E-cadherin, Dsg, Dsc, p120 and α-catenin, similar to the wildtype plakoglobin. As a result, SyPg-containing vesicles associate with desmosomal and adherens junction plaques (Chitaev et al., 1996), while vesicles containing synaptophysin alone are randomly distributed through the cytoplasm of the transfected cells. Identical results were obtained when plakoglobin was translocated to the amino terminus upstream from synaptophysin in the PgSy construct.

Recent experiments with plakoglobin mutants (Hulsken et al., 1994; see also our results below) demonstrated strong intramolecular interactions within this molecule. The data indicate that conformational abnormalities can be induced by large deletions within plakoglobin. Our experiments with synaptophysin-plakoglobin chimeras (Chitaev et al., 1996) made it possible to develop a novel approach to study plakoglobin structure, a synaptophysin insertion mutagenesis. The insertion of synaptophysin into various positions of plakoglobin results in the formation of synaptic-like vesicles coated by plakoglobin with a functionally affected domain, leaving the unaffected domains of plakoglobin exposed on the cytoplasmic surface and involved in appropriate interactions. Here we introduce this approach, in conjunction with more conventional strategies, to characterize cadherin-binding sites of plakoglobin.

15 chimeric plakoglobins (PSyG chimeras) containing a synaptophysin insertion in different positions of the arm repeat region were constructed (Fig. 1B). Expression of these chimeras in A-431 cells resulted in a single polypeptide with a predicted molecular mass of 130 kDa (Fig. 2). A significant degradation product was observed only in the chimera PSyG470. Association of these chimeras with Dsg, Dsc, E-cadherin, p120, α-catenin and endogenous plakoglobin was examined by co-immunoprecipitation using anti-synaptophysin antibodies (see Table 1 for a summary). Western-blot analysis (Fig. 2) shows that PSyG-type proteins expressed in stably transfected cells do not associate with endogenous plakoglobin. Three PSyG-type chimeras, PSyG158, PSyG204 and PSyG211, containing a synaptophysin insertion in the first, second or beginning of the third arm repeats, respectively, interact with E-cadherin, α-catenin and p120 but not with Dsg or Dsc. Deletion of the sequences either upstream or downstream from the synaptophysin insertion in the PSyG211 chimera (chimeras PSy211 and SyG212, Fig. 1) completely abolished its binding with E-cadherin (not shown). This result demonstrates that cooperation between two fragments of plakoglobin exposed on the vesicular surface of PSyG211-expressing cells is required for the formation of the E-cadherin/plakoglobin complex.

Insertion of synaptophysin into the end of arm repeat 4 or the beginning of arm repeat 5 did not change binding of plakoglobin with Dsg, but completely abolished its interaction with Dsc and E-cadherin, and strongly reduced its binding with α-catenin and p120. Four chimeras (PSyG249, PSyG330, PSyG470 and PSyG504) containing a synaptophysin insert either at the end of arm repeats 3 and 5, or in two positions in arm repeat 9, respectively, show very weak or no binding to any of the above ligands. All other chimeric proteins (PSyG363, PSyG451, PSyG539, PSyG550, PSyG585 and PSyG640) show an interaction pattern similar to that of wildtype plakoglobin (Fig. 2). These data clearly indicate that different plakoglobin sequences are involved in the binding of E-cadherin, Dsg and Dsc. Insertion of synaptophysin into either of the first three arm repeats affects the interaction of plakoglobin with Dsg and Dsc, but not with E-cadherin. Insertion of synaptophysin into repeats 4 or 5 affects binding to E-cadherin and Dsc but not Dsg.

A very high level of sequence identity between β-catenin and plakoglobin and the fact that β-catenin associates only with classical but not with desmosomal cadherins provides an approach to investigate the structure of desmosomal cadherin binding sites in plakoglobin. To map a Dsg-binding site, we substituted a segment of β-catenin for a homologous segment of plakoglobin (Met¹-Arg²³³) to obtain the SyPgBC1/233 chimera (Fig. 3). The replaced plakoglobin segment includes the first two and a half arm repeats that have the highest affinity for Dsg in vitro (Chitaev et al., 1996), and its disruption by synaptophysin insertion completely abolished plakoglobinDsg binding (see above). Co-immunoprecipitation analysis showed that the SyPgBC1/233 chimera binds to E-cadherin, but not to Dsg (Fig. 3), similar to the chimera SyBC containing the entire sequence of β-catenin. The chimeric protein SyPgBC1/142, where only 142 amino acid residues (Met¹-Arg¹⁴²) of plakoglobin were replaced with the corresponding sequence of β-catenin, was able to co-precipitate Dsg, but to a lesser extent than SyPg (Fig. 3). In addition, substitution of β-catenin arm repeats 4-10 for the corresponding segment of plakoglobin in the chimera SyPgBC234/580 did not affect its ability to bind to Dsg and E-cadherin. These data suggest that the plakoglobin sequence Arg¹⁴²-Arg²³³ is essential for binding to Dsg. In agreement with this conclusion, deletion of the 21 most divergent amino acid residues (Leu¹⁸⁵-Ser²⁰⁷) within the second arm repeat of the SyPg chimera completely abolished its interaction with Dsg (see Fig. 3), but did not affect its affinity to bind E-cadherin or α-catenin. These data are in agreement with the synaptophysin insertion mutagenesis experiments and support the conclusion that a Dsg-specific binding site is located within the arm repeats 1-3 of plakoglobin. To further confirm the presence of E-cadherin-binding site in arm repeats 4 and 5, we deleted a 17-amino-acid-long sequence (Leu304-Arg320) within this region from the SyPg chimera. The resulting chimera SyPg∆(304-320) is completely unable to bind E-cadherin, but interacts normally with Dsg (Fig. 3), in agreement with results obtained using the synaptophysin insertion mutagenesis approach.

To understand the functional effect of the synaptophysin insertion on plakoglobin, the cells stably expressing PSyG-type proteins were analyzed by double immunofluorescent microscopy. As expected, chimeras unable to bind cadherins (PSyG249, PSyG330, PSyG470 and PSyG504) were localized in small vesicle-like structures randomly distributed in the cytoplasm but not in the cell-cell contacts, similar to wild-type synaptophysin (not shown). Of three chimeras unable to bind Dsg, PSyG158 was randomly distributed in the cytoplasm while the other two (PSyG204 and PSyG211) efficiently integrated into cell-cell contacts and were co-distributed with E-cadherin (Fig. 4), α-catenin and p120 (data not shown). The normal morphology of these cells indicate that the presence of the PSyG204 or PSyG211 chimeras does not detectably affect cell-cell adhesion.

Two chimeras, PSyG288 and PSyG304, able to bind Dsg but not E-cadherin and Dsc, were evenly distributed between the cytoplasm and cell-cell contacts. A significant portion of Dsg in these cells was co-distributed with the chimeric proteins along cell-cell contacts (Fig. 5a,a’). Double immunostaining of these cells with desmoplakin/Dsg antibodies revealed multiple desmosome-like structures containing desmoplakin, which in some cases have very fine Dsg staining (Fig. 5b,b’). These structures were smaller then most desmosomes of the parental A-431 cells and anchor only fine keratin bundles compared to normal desmosomes (Fig. 5c-e). In this respect, desmosomes in the PSyG288-or PSyG304-expressing cells are similar to a population of small-size desmosomes found in A-431 cells that interact only with thin bundles of IF (Fig. 7d,d’ and f,f). This observation suggests the interesting possibility that PSyG288 or PSyG304 chimeras inhibit growth of desmosomes. The role of the arm repeats 4 and 5 of plakoglobin in desmosome assembly needs more detailed investigation.

Although most of the chimeric proteins containing a synapto-physin insert in arm repeats 6-13 (PSyG363, PSyG451, PSyG539, PSyG550, PSyG585, PSyG640) were able to associate with all tested plakoglobin binding proteins, they failed to integrate efficiently into desmosomes. Only two chimeras of this group (PSyG539 and PSyG550), containing the synaptophysin insert either at the very end of arm repeat 10 or in the unique sequence separating repeats 10 from 11, faithfully accumulated in cell-cell contacts and co-localized with E-cadherin (Fig. 6b,b’), α-catenin and p120 (not shown). In contrast to SyPg (or PgSy), both these chimeras induced detectable defects in desmosomes. Desmosomes in cells expressing these proteins were less abundant and Dsg was no longer associated with desmosomes exclusively, but a detectable pool of this protein was distributed randomly throughout cel-cell contact regions (Fig. 6a,a’).

Among six chimeric proteins that bind to all the tested cadherins, only PSyG363 and PSyG451 were localized exclusively to the cytoplasm, similar to wild-type synaptophysin (not shown). The majority of chimeric proteins PSyG585 and PSyG640, containing the synaptophysin insertion in arm repeats 11 and 12, were also found in the cytosol and only occasionally seen in desmosomes or adherens junctions (Fig. 6c,c’). No abnormalities in the localization of the desmoglein or desmoplakin in these cells were detected. These observations indicate that binding to cadherins is not sufficient to target plakoglobin to the adherens junctions or desmosomes.

Recently we have showed (Chitaev et al., 1996; Fig. 1) that intact recombinant plakoglobin may contain at least three independent cadherin binding sites. One of these sites corresponds to the Dsg-specific site described above. Two other sites, exhibiting a similar affinity to the C-domains of Dsg1, Dsc1a and E-cadherin, were mapped to a region spanning arm repeats 6-12. However, as reported above, disruption of this region by synaptophysin insertion did not change plakoglobin binding to cadherins in vivo. To solve this discrepancy we studied the binding of cadherins to the C-terminal or N-terminal truncated plakoglobin mutants using a solid-phase binding assay described previously (Chitaev et al., 1996). The C-terminal truncation of plakoglobin (mutant Pg1/580 lacking repeats 1113 and a unique carboxyl-end region) did not affect binding to Dsc and Dsg, but slightly increased binding to E-cadherin compared to the entire plakoglobin molecule (Fig. 7). Cadherin binding was abolished by amino-terminal deletion of arm repeats 1-4 (Fig. 7, mutant Pg305/744, Met¹-Leu³⁰⁴ deletion). The latter result is in agreement with our in vivo data that mapped Dsg and E-cadherin binding sites within arm repeats 1 to 5, but contradict our previous results showing that mutants Pg305/505 and Pg505/672 have strong cadherin binding properties (Chitaev et al., 1996, see also Fig. 7). This apparent discrepancy can be explained if the plakoglobin sequence Lys⁶⁷³-Ala⁷⁴⁴, which is absent in these two mutants, inhibits cadherin binding activity of mutant Pg305/744. To test this possibility, we inserted the Lys⁶⁷³-Ala⁷⁴⁴ fragment into the Pg505/672 mutant and demonstrated that the insertion eliminated the cadherin-binding activity of Pg505/672 protein (mutant Pg505/744, Fig. 7). Further deletion-mapping experiments showed that a similar inhibitory effect on cadherin binding can be achieved by insertion of ten amino acid residues K⁶⁷³-Q⁶⁸³ (mutant Pg505/683, Fig. 7). Based on these data we hypothesized that the inhibitory effect of the ten amino acid residues K⁶⁷³-Q⁶⁸³ from arm repeat 13 on cadherin binding is due to its interaction with the upstream cadherin binding sites and masking their activity. The results presented in Fig. 8 support this conclusion and demonstrate that the plakoglobin fragment Pg580/744 interacts strongly with the upstream fragments Pg305/505 and Pg505/672. Its binding to the fragment Pg114/292, containing 1-4 arm repeats, was significantly weaker. No binding was detected with the Pg1/147 fragment containing the unique amino-terminal region and the first arm repeat. Similar results were obtained using both solid-phase and overlay assays (Fig. 8A,B).

To investigate further the inhibitory effect of the C-terminal arm repeat on the activity of upstream cadherin binding sites we examined the ability of the Pg580/744 fragment, which does not bind to cadherins (Fig. 7), to compete with the binding of plakoglobin fragments for the C-domain of Dsg. Experiments presented in Fig. 8C show that the Pg580/744 fragment competed with binding of Dsg to the plakoglobin fragments Pg305/505 or Pg505/672. The Pg580/744 fragment had little or no effect on Dsg binding to Pg114/292 or to the entire plako-globin molecule. These experiments demonstrate that the 11/13 arm repeats segment, containing the inhibitory fragment K⁶⁷³-Q⁶⁸³, directly interacts with upstream arm repeats 5-10 to negatively affect the activity of cadherin binding sites located in this region.

Using deletion mapping analysis of SyPg-type chimeras, we have shown (Chitaev et al., 1996) that the first three arm repeats of plakoglobin (Arg¹⁴¹-Met²³⁴) contain a Dsg binding site. Synaptophysin insertion mutagenesis and homologous plakoglobin/β-catenin replacement experiments presented here confirm this observation. In addition, we show that this region is not involved in plakoglobin binding with E-cadherin and α-catenin, since synaptophysin insertions in positions 158, 204 and 211 specifically abolished only Dsg and Dsc binding. Elimination of plakoglobin binding to Dsg upon deletion of 21 residues (Leu¹⁸⁵-Ser²⁰⁷) demonstrate that the end of the second arm repeat is an essential element of the Dsg binding site. These conclusions are in agreement with our data demonstrating that deletion of this sequence in myc-tagged plakoglobin also abolished its binding to desmosomal cadherins (S.M. Troyanovsky, unpublished results). Reduction in the Dsg binding by the chimeric protein SyPgBC1/142 compared to SyPg indicates that the plakoglobin sequence upstream from Thr¹⁴² can also participate in binding to Dsg. This result raises the possibility that the Dsg binding site partially overlaps with the α-catenin binding sequence in the Gln¹⁰⁹-Ala¹³⁷ region (Aberle et al., 1996). An overlap between α-catenin and Dsg-binding sites was also suggested recently by Witcher et al. (1996). These data taken together suggest that a relatively long sequence (about 100 amino acids), including most of the first three arm repeats, is involved in Dsg-binding (Fig. 1E,F).

Selective inactivation of E-cadherin binding activity in the chimeras PSyG288, PSyG304 and SyPg∆(304-320) suggests that arm repeats 4 and 5 (the sequence Leu²⁵⁰-Val³³⁴) are involved in E-cadherin binding. Interestingly, the PSyG-type chimeras containing the insert in Dsg-or E-cadherin-binding regions are not able to interact with Dsc. Additional mutagenesis experiments will be needed to determine whether the Dsc-binding site overlaps with both E-cadherin and Dsg binding sites. Alternatively, binding of Dsc to plakoglobin could be dependent on the initial binding of plakoglobin to other cadherins or could be mediated through unidentified protein(s). Moreover, the question remains as to whether Dsg and E-cadherin binding sites are totally distinct. Simultaneous inactivation of the Dsg and E-cadherin binding activities in the chimera PSyG249 suggests that the end of the third arm repeat may constitute a structural component of both sites. On the other hand the negative effect of the synaptophysin insertion in this position could be due to steric hinderance. A similar mechanism may be responsible for the simultaneous inactivation of the Dsg and E-cadherin binding sites in the other PSyG-type chimeras (PSyG330, PSyG470 and PSyG504). Synapto-physin inserts in the chimeras PSyG288 and PSyG304 do not affect the known α-catenin-binding sequence (Aberle et al., 1996). At the same time they both co-immunoprecipitate an unexpectedly low amount of α-catenin. Very weak binding to α-catenin was also found for SyPg∆(304-320) (R.B. Troy-anovsky, unpublished). Both observations support the hypothesis that α-catenin- and Dsg-binding to plakoglobin are mutually exclusive. Whether the complexes consisting of α-catenin and plakoglobin lacking E-cadherin binding are unstable or simply do not form remains to be determined.

Localization of the Dsg binding site of plakoglobin to the arm repeats 1-3 is in good agreement with our previous experiments showing high affinity binding of the plakoglobin fragment Pg114/292, encompassing the sequence of the first three arm repeats, to Dsg in vitro (Chitaev et al., 1996). In addition, in vitro binding experiments show that two non-overlapping plakoglobin fragments, Pg305/505 and Pg505/672, both derived from arm repeats 6-13, exhibit strong binding to Dsc, Dsg and E-cadherin C-domains. Since deletion of arm repeats 1-3 completely abolished Dsg binding, we hypothesized that in intact plakoglobin, cadherin binding sites localized in arm repeats 6-13 are hidden (Chitaev et al., 1996). The data presented here strongly support this possibility. We found that the plakoglobin fragment Pg580/744 (containing arm repeats 11 to 13 and the carboxyl terminus of plakoglo-bin) is unable to bind to cadherins but binds to fragments Pg305/505 and Pg505/672, and efficiently competes for their interaction with cadherins. These observations suggest that in intact plakoglobin, arm repeats 11-13 interact with upstream repeats 6-10, masking cadherin binding sites present in this segment of plakoglobin (Fig. 1F). Although it is not clear whether activation of these two cryptic sites is required for desmosome assembly, it is possible that their activation during desmosome or adherens junction assembly leads to cross linking of several cadherins into one supramolecular complex.

Plakoglobin/cadherin complexes can associate with the protein kinase substrate p120 (Reynolds et al., 1994; Shibamoto et al., 1995; Staddon et al., 1995). Synaptophysin insertion mutagenesis did not reveal a specific binding site for p120 in plakoglobin. We have found that most of the PSyG chimeras that are able to form complexes with E-cadherin coprecipitate a large amount of p120. This is in agreement with the observation of Daniel and Reynolds (1995), demonstrating a direct interaction between p120 and catenin-binding region of E-cadherin. At the same time, two observations suggest that plakoglobin may also have a binding site for p120. First, chimera PSyG288, unable to bind E-cadherin, co-precipitates p120. Second, chimeras PSyG363 and PSyG451, able to bind E-cadherin, precipitate only a residual amount of p120. This binding site can also contribute to the association of p120 with plakoglobin/E-cadherin complexes.

Experiments with dominant negative mutants of Dsg and classical cadherins (Kinter, 1992; Troyanovsky et al., 1993, 1994a; Fujimori and Takeichi, 1993) demonstrate an important role of β-catenin and plakoglobin in the assembly of adherens junctions and desmosomes. One of the functions of β-catenin and plakoglobin is to couple classical cadherins with α-catenin. That, in turn, mediates the attachment of the adherens junctions to the cortical actin cytoskeleton (Nagafuchi et al., 1994; Knudsen et al., 1995; Rimm et al., 1995). The function of plakoglobin in desmosome assembly is not well understood.

Synaptophysin insertion mutagenesis divides the arm repeat region of plakoglobin into two functionally distinct regions, encompassing arm repeats 1 to 5 and 6 to 13 (regions 1/5 and 6/13, respectively). Five of seven mutants containing the synaptophysin insert within the 1/5 region were unable to associate either with E-cadherin or with Dsg, while six out of eight mutants containing the insert in 6/13 region bind equally to all cadherins studied. As discussed above, the 1/5 region of plakoglobin includes distinct binding sites that appear to overlap or be closely juxtaposed. This region mediates direct linkage of plakoglobin with Dsg or with E-cadherin and α-catenin. The role of the 6/13 region is probably more complex. Despite their strong binding with cadherins, chimeras PSyG363, PSyG451, PSyG585 and PSyG640 containing synaptophysin in the 6/13 region, were unable to associate with desmosomes and adherens junctions. Vesicles formed by these chimeras were localized exclusively in the cytoplasm and had no influence on desmosome or adherens junction assembly. Deletion of the arm repeats 11-13 was also found to affect correct destination of the SyPg chimera into cell-cell junctions (Chitaev et al., 1996). Therefore, the 6/13 region may function in the targeting of plakoglobin/cadherin complexes to cell-cell junctions. Whether cryptic cadherin binding sites, which are hidden or disrupted in plakoglobins containing synaptophysin in the 6/13 region, are involved in the targeting mechanism remains to be determined.

In addition, our experiments demonstrate clear differences in the assembly of desmosomes and adherens junctions. First, PSyG-type chimeras, regardless of the site of synaptophysin insertion or cadherin binding, are unable to integrate efficiently into desmosomes, whereas several of them can be integrated into adherens junctions. Second, typical epithelial morphology and normal distribution of the adherens junction proteins in the A-431 clones stably expressing any of the PSyG-type chimeras, suggest that these chimeras are not able to affect this type of junction. In contrast, immunofluorescence microscopy shows interesting changes in desmosomes of cells transfected with PSyG288, PSyG304, PSyG539 and PSyG550. In the first two chimeras, PSyG288 and PSyG304, synaptophysin was inserted into the E-cadherin binding site of plakoglobin. Upon expression in A-431 cells, these two chimeric proteins bind to Dsg but not Dsc and produce desmosomes that are smaller in size. These desmosomes are unable to bind normal IF bundles. Chimeras PSyG539 and PSyG550 carry a synaptophysin insertion in the unique 22-amino-acid-long region, separating arm repeats 10 and 11. These proteins bind both the desmosomal cadherins, Dsg and Dsc. Their expression in A-431 cells results in a dramatic reduction in the number of desmosomes. All four chimeric proteins were efficiently sorted into cell-cell contact regions but did not incorporate into desmosomes. The mechanism of the dominant negative effect of these two groups of mutants is not clear since other similar chimeric proteins that bind desmosomal cadherins do not negatively affect desmosome formation. To explain these findings an attractive hypothesis is that plakoglobin contains two functionally distinct domains, a binding 1/5 region and a targeting 6/13 region. Occupation of the Dsg or E-cadherin specific sites in the binding domain induces two different 6/13 domaindependent mechanisms responsible for correct and efficient targeting of cadherins to either desmosomes or adherens junctions.

We would like to thank Dr G. Goldberg, Dr A. Eisen (St Louis, MO) and Dr W.W. Franke (Heidelberg, FRG) for valuable discussion. The work has been supported in part by a Pfizer Pharmaceutical Career Development Award from Dermatology Foundation.