Plakophilin 1 interferes with plakoglobin binding to desmoplakin, yet together with plakoglobin promotes clustering of desmosomal plaque complexes at cell-cell borders

Adherens junctions and desmosomes are intercellular adhesive junctions specialized to provide strong but dynamic cell-cell adhesion in a variety of cell types and tissues (Cowin and Burke, 1996; Garrod et al., 1996; Green and Gaudry, 2000; Kowalczyk et al., 1999a; Schmidt et al., 1994). Members of the cadherin superfamily of adhesion molecules form the adhesive interface of these junctions in each case (Koch et al., 1999; Koch and Franke, 1994; Magee and Buxton, 1991). In addition to mediating adhesion, the cadherins also link the plasma membrane to the underlying cytoskeleton through interactions between their cytoplasmic tails and associated proteins that include catenins (Aberle et al., 1996). For desmosomes, it has been suggested that these indirect linkages may occur through both linear and lateral assemblies of proteins that form a macromolecular complex called the junctional plaque (Kowalczyk et al., 1997; Kowalczyk et al., 1999b; Smith and Fuchs, 1998). However, the mechanism of assembly of the junctional plaque is still only poorly understood.

Desmosomes arose relatively late in evolution, after the appearance of adherens junctions and as organisms increased in size and complexity. This fact is reflected by the abundance of desmosomes in tissues that experience mechanical stress. The experimental ablation or mutation of desmosomal plaque and transmembrane components results in defects in adhesion and tissue integrity in mice (Allen et al., 1996; Bierkamp et al., 1996; Gallicano et al., 1998; Koch et al., 1997; Ruiz et al., 1996). In addition, human autoimmune and inherited diseases that target the desmosomal cadherins lead to similar defects in tissue integrity (Rickman et al., 1999; Stanley, 1995). It has been shown that mutations in the desmosomal plaque protein plakophilin 1 (PKP1) result in a skin fragility syndrome (McGrath et al., 1997) and that haploinsufficiency in desmoplakin (DP) leads to a form of striate palmoplantar keratoderma (SPPK) (Armstrong et al., 1999). More recently, truncating mutations of plakoglobin and DP have been shown to cause cardiac and skin defects in humans (McKoy et al., 2000; Norgett et al., 2000).

Based on these findings, there is ample evidence to support the importance of desmosomes in tissue integrity. However, we are only now beginning to appreciate the true extent of desmosomal structural and possible functional complexity. Desmosomes are commonly expressed in all types of epithelia, but are also found in a restricted number of other specialized tissues such as the myocardium and Purkinje fiber cells of the heart, meningeal cells and the follicular dendritic cells of lymph nodes (Schwarz et al., 1990). These junctions are particularly abundant in complex epithelia such as the epidermis; however, their number, composition and structure vary depending on tissue type and during terminal differentiation (Kowalczyk et al., 1999a). Six desmosomal cadherin genes have been identified and divided into two subclasses called desmogleins and desmocollins (Buxton et al., 1994; Koch and Franke, 1994; Kowalczyk and Green, 1996). These cadherins are each expressed in a tissue- and differentiation-specific pattern. Other proteins that contribute to plaque assembly and linkage of intermediate filaments (IF) to the desmosomal cadherin tails are members of the armadillo gene family, found in intercellular junctions and the nucleus. The desmosomal arm proteins include plakoglobin and plakophilins 1-3 (Anastasiadis and Reynolds, 2000; Gelderloos et al., 1997; Hatzfeld, 1999). The plakophilins also exhibit tissue-specific and differentiation-specific distributions that may translate into differences in desmosome function in different body sites. Recent work from our lab and others has shown that two arm family members, PKP1 and plakoglobin, both interact with the obligate desmosomal component and IF-associated protein, DP (Hatzfeld et al., 2000; Hofmann et al., 2000; Kowalczyk et al., 1997; Kowalczyk et al., 1999a; Smith and Fuchs, 1998), and that this interaction is mediated by the DP N terminus, the domain that targets this IF linker to the desmosomal plaque (Hatzfeld et al., 2000; Kowalczyk et al., 1999a).

In the present work we explore the assembly properties of proteins found in desmosomes of the differentiated layers of complex epithelia. To overcome difficulties inherent in studying these proteins in situ, where their insolubility hampers the examination of protein-protein interactions, we have employed a reconstitution approach. Using this approach we demonstrate here the existence of a complex containing DP, plakoglobin and the desmoglein 1 (Dsg1) tail. Although plakoglobin is required for the formation of triple complexes containing the first 584 amino acids of the DP N terminus, full length DP associates with the Dsg1 tail in the presence and absence of plakoglobin in HT1080 cells. This observation raises the possibility that plakoglobin-independent interactions between downstream sequences in DP and the Dsg1 tail may occur. We also examine the relative abilities of PKP1 and plakoglobin to interact with full length DP. When both PKP1 and plakoglobin are present, Triton-soluble full-length DP associates preferentially with PKP1, whether in the absence or presence of the Dsg1 tail. In spite of this difference in binding preference for DP, our results indicate that plakoglobin and PKP1 are both required for the organization of components into clustered structures appearing ultrastructurally similar to the desmosomal plaque. These results suggest that the repertoire of arm family members and associated cadherin tails expressed in a particular desmosome will have an important impact on the assembly, morphology and function of junctions in different cells and stages of differentiation.

The E-cadherin/Dsg1 chimera (E-Dsg1), comprising the extracellular domain of mouse E-cadherin and the cytoplasmic domain of human desmoglein 1, in the expression plasmid LK444 under the control of the β-actin promoter was described previously (Kowalczyk et al., 1997). As previously reported, the presence of the E-cadherin extracellular domain serves to stabilize the Dsg1 tail at cell-cell borders, which happens only inefficiently in the case of Dsg1 alone, possibly due to the absence of a desmosomal cadherin adhesion partner (Kowalczyk et al., 1997). Also described previously were constructs encoding full-length human plakoglobin in LK444 (Kowalczyk et al., 1994; Palka and Green, 1997) and a CMV construct encoding a C-terminally FLAG-epitope tagged polypeptide comprising the first 584 amino acids in DP, called DPNTP (DPNTP.FLAG) (Kowalczyk et al., 1999b). The full-length ‘a’ form of PKP-1 was assembled and cloned into pCMV script as described (Kowalczyk et al., 1999b).

Full length C-terminally FLAG-tagged DP (DP.FLAG; p806) was generated as follows. A 3.1 kb NotI-EcoRV fragment corresponding to nucleotides –332 to 3671 of DP was isolated from p611, a pSK.DPΔC construct. This fragment was ligated to NotI-EcoRV pCMV5c (SIGMA) to generate p794. An 8.6 kb SacII-PmlI insert corresponding to nucleotides 28-8616 was isolated from p613, pCMV.DP.myc, and ligated to p794 digested with SacII and EcoRV to generate p806.

HT1080 human fibrosarcoma cells, a kind gift from Dr Noel Bouck, Northwestern University, and COS-7 cells were cultured in DMEM containing 10% fetal bovine serum and penicillin/streptomycin (Kowalczyk et al., 1997). Transient transfections of HT1080 cells were performed using Fugene 6 reagent (Roche-Boeringer Mannheim) and were assayed 24 hours following transfection. Transient transfections of COS-7 cells were performed using either Fugene 6 or calcium phosphate as previously described (Stappenbeck and Green, 1992).

Cells were grown on glass coverslips, rinsed in phosphate buffered saline and fixed in methanol at −20°C for two minutes. E-Dsg1 was detected by the rat monoclonal antibody DECMA-1 directed against the extracellular domain of E-cadherin (a gift from Dr Stephen Byers, Georgetown University). Plakoglobin was detected by the mouse monoclonal antibody 11E4 (a gift from Dr M. Wheelock, University of Toledo; Kowalczyk et al., 1994) or chicken antibody 1407 (C. A. Gaudry et al., unpublished). A rabbit antibody #667 against the PKP1 head domain was used to detect PKP1 (Hatzfeld et al., 2000). DPNTP.FLAG or full length DP.FLAG were detected by the rabbit polyclonal antibody NW161 (Bornslaeger et al., 1996), directed against a bacterially expressed polypeptide comprising the first 189 amino acids of DP, by PC28, a mouse monoclonal antibody directed against the DP central rod domain (a gift from Dr Pamela Cowin), NW6, directed against the DP C terminus (Angst et al., 1990) or by M2 (Sigma Chemicals, St Louis, MO), a mouse monoclonal antibody directed against the FLAG tag. It should be noted that due to the high ratio of ectopic to endogenous protein, similar results were obtained with either M2 or DP-specific antibodies in transient transfections. Primary antibodies were detected using Alexa conjugates (Molecular Probes, Eugene, Oregon). Coverslips were examined using a Leica DMB microscope and images captured using a Hamamatsu Orca digital camera and Improvision Openlab software.

Transiently transfected COS 7 cells grown on Permanox culture dishes (Electron Microscopy Sciences, Ft. Washington, PA) were processed for conventional electron microscopical analysis as previously described (Jones and Goldman, 1985) and stained sections were viewed and photographed using a JEOL 100CX electron microscope. Parallel immunofluorescence experiments were carried out for each transfection combination to ensure that ultrastructural analysis was carried out on populations exhibiting high co-transfection efficiencies.

HT1080 or COS-7 cells transiently expressing junctional components were lysed in Tris buffered saline (TBS) containing 0.5% Triton X-100; lysates were vortexed and cleared by centrifugation at 14,000 g for 30 minutes. To immunoprecipitate FLAG-tagged DP polypeptides and associated proteins, lysates were incubated with M2 agarose beads (Sigma, St Louis, MO) for 1 hour at 4°C. Beads were captured by centrifugation at 11,000 g for 2 minutes and washed 4 times by vortexing for 30 seconds each in TBS-0.5% Triton at 4°C. Proteins were released by incubation with reducing SDS-containing sample buffer and analyzed by SDS PAGE and immunoblot analysis. For fractionation of transfected cells into Triton soluble and insoluble pools, cells were lysed in 0.5% Triton X-100 containing buffer. Lysates were vortexed, centrifuged, and the cytoskeletal-associated fraction was subsequently solubilized in urea sample buffer as described (Jones and Goldman, 1985). The E-Dsg1 chimera was detected with DECMA or ECCD-2 (a gift from M. Takeichi, Kyoto University, Kyoto Japan). Plakoglobin was detected either by the rabbit polyclonal antibody VB3 (a gift from F. Watt, Imperial Cancer Research Fund, London) or a polyclonal chicken antibody 1407, developed against human plakoglobin (C. A. Gaudry et al., unpublished). β-Catenin was detected using the mAb C2206 (Sigma, St Louis, MO). PKP1 was detected with polyclonal antibody 667 (Kowalczyk et al., 1999b). DP and DPNTP were detected using the polyclonal antibody NW161 (Bornslaeger et al., 1996). After incubation with HRP conjugated goat anti-rabbit or goat anti-chicken IgG (Jackson ImmunoResearch, West Grove, PA or Aves Labs, Tigard, Oregon, respectively), immunoreactive proteins were detected by enhanced chemiluminescence (ECL; Amersham Pharmacia Biotech). Proteins present in the soluble and insoluble fractions were quantified by scanning densitometry using Molecular Analyst software.

We previously demonstrated that an N-terminal polypeptide comprising 584 amino acids of DP, DPNTP, is recruited to a chimeric cadherin with the cytoplasmic domain of human Dsg1 and the extracellular domain of mouse E-cadherin (Kowalczyk et al., 1997). Recruitment of DPNTP to the Dsg1 tail was dependent on the co-expression of plakoglobin in COS-7 epithelial cells and mouse L cell fibroblasts (Kowalczyk et al., 1997). This observation, in conjunction with data showing that plakoglobin associates directly with both DP (Kowalczyk et al., 1997; Smith and Fuchs, 1998), and desmoglein 1 (Mathur et al., 1994; Witcher et al., 1996), suggested the existence of a triple complex in which DP’s association with the Dsg1 tail is mediated indirectly by the arm protein plakoglobin. However, to date, a cellular complex containing DP and desmosomal cadherins has not been detected. This is likely due to the fact that DP, once assembled into junctions, is largely insoluble, making it difficult to assay for DP-containing protein complexes. To examine whether the N terminus of DP forms a triple complex with plakoglobin and the Dsg1 tail, we transiently expressed DPNTP.FLAG, with or without plakoglobin and E-Dsg, into COS cells, which have very few desmosomes (Kowalczyk et al., 1997) or HT1080 fibrosarcoma cells, which do not have DP and do not assemble desmosomes (Chitaev and Troyanovsky, 1997). To assess whether the previously observed recruitment of DPNTP to cell-cell borders reflected the formation of complexes containing all three proteins, analysis of the Triton-soluble lysate was carried out (Fig. 1, and data not shown). Because DP is largely insoluble (see Fig. 9), we concentrated the limited pool of soluble DP by immunoprecipitation. M2 agarose was used to precipitate DPNTP.FLAG, complexes were separated by SDS-PAGE and transferred to nitrocellulose, and the presence of plakoglobin and the E-Dsg1 chimera was assessed by immunoblotting. As shown in Fig. 1A, both plakoglobin and the chimeric cadherin were identified in a complex with DPNTP.FLAG, and this complex was only observed in the presence of plakoglobin. The E-Dsg1 chimera was not observed in a complex with DPNTP when co-expressed with the related arm protein β-catenin instead of plakoglobin, providing further evidence that plakoglobin is specifically required for this association (Fig. 1B). Similar experiments demonstrated that a triple complex containing the Dsc2 tail also forms under these conditions (not shown). These results suggest that the 584 residue N-terminal region of DP interacts indirectly with desmosomal cadherin tails by virtue of its association with plakoglobin.

To examine whether full length DP behaves similarly to DPNTP, recruitment and immunoprecipitation assays were carried out using DP.FLAG. In the absence of a co-expressed cadherin, full length DP colocalized mostly with IF networks, as expected from its known association with IF polypeptides (Kouklis et al., 1994; Meng et al., 1997; Stappenbeck et al., 1993; Stappenbeck and Green, 1992) (Fig. 2B). Plakoglobin appeared largely diffuse in the cytoplasm in methanol-recruit plakoglobin and DP.FLAG to the cell-cell borders, these proteins were co-expressed with the E-Dsg chimera. In these cells, both plakoglobin (not shown) and DP colocalized at cell-cell borders (Fig. 2C-D, arrows).

To examine whether, like DPNTP, full length DP forms a complex with the Dsg1 tail that is dependent on the presence of plakoglobin, M2 agarose was used to precipitate DP.FLAG and the presence of protein binding partners was assessed as described above. As shown in Fig. 1C, full length DP was observed in a complex with plakoglobin and E-Dsg1. Interestingly, E-Dsg1 was also detected in a complex with DP.FLAG in the absence of plakoglobin (Fig. 1B), raising the possibility that sequences downstream in full length DP that are not present in DPNTP may contribute to interactions with the E-Dsg1 tail.

The results described above support the idea that DP is linked fixed cells (Fig. 2A), although some co-localization with IF networks can be seen in detergent-extracted cells (not shown). It is important to note that both plakoglobin and full length DP are capable of incorporating into desmosomes on their own (Palka and Green, 1997; Smith and Fuchs, 1998; Stappenbeck et al., 1993); this localization is more apparent in cells with abundant junctions and in cells expressing lower levels of protein. To test whether a desmosomal cadherin will to the plasma membrane through plakoglobin and the desmosomal cadherin tails. However, it has also been proposed that lateral interactions play an important role in plaque assembly. We reported that PKP1 may be a candidate for mediating lateral interactions by binding to DP and enhancing its recruitment to cell borders in the most differentiated layers of the epidermis (Kowalczyk et al., 1999b). However, PKP1 is expressed in these same layers along with plakoglobin; furthermore, both PKP1 and plakoglobin have been shown to interact with the same 584 amino acid region of the DP N terminus (Kowalczyk et al., 1997; Kowalczyk et al., 1999b). Thus the question arises as to whether PKP1 binding influences binding of plakoglobin to this domain, and if so, what the impact will be for desmosome plaque assembly and function.

We previously showed that PKP1 recruits DPNTP to cell-cell borders in transient transfection assays (Kowalczyk et al., 1999b). To address whether full length DP is also recruited to borders by PKP1, DP.FLAG was expressed with or without PKP1 in COS cells. Immunofluorescence analysis showed that like DPNTP, full length DP is recruited to plasma membranes in the presence of PKP1 (Fig. 3E,F). Likewise, an obvious increase in continuous staining for endogenous DP was observed in COS cells singly transfected with PKP1 (Fig. 3C,D).

Next, to examine the relative ability of DP to associate with PKP1 and plakoglobin, full length DP.FLAG was transfected along with PKP1, plakoglobin or both PKP1 and plakoglobin in COS cells. Triton soluble lysates were isolated, DP.FLAG was precipitated with M2 agarose and complexes were analyzed by SDS-PAGE and immunoblotting. As shown in Fig. 4 when expressed separately, both PKP1 and plakoglobin interacted with full length DP. When DP, plakoglobin and PKP1 were expressed together, only PKP1 was clearly seen in a complex with DP (Fig. 4). Plakoglobin was undetectable under these conditions, and only barely detectable at higher exposures (Fig. 4, right lane).

To further explore the contribution of PKP1 and plakoglobin to desmosome plaque assembly, we carried out a series of co-transfections, followed by localization of ectopically expressed proteins in COS cells. As shown in Fig. 5, when E-Dsg1 was co-expressed with PKP1 and DP.FLAG, the protein distributions shifted relative to those observed when PKP1 and DP (Fig. 3) were co-expressed. Some cell border staining was observed in cells co-expressing PKP1, DP and E-Dsg (Fig. 5A-D, arrows), but the majority of staining appeared in cytoplasmic linear arrays co-localizing with keratin IF (data not shown). The appearance of this particular pattern was dependent on the co-expression of all three components, as cells within the same population, expressing only one or two of the proteins, displayed different patterns of localization. (See, for instance, the cell in 5C marked by arrowheads which exhibits a more diffuse, cell surface pattern and expresses E-Dsg1 but not PKP1). However, when all four proteins, PKP1, plakoglobin, DP.FLAG and E-Dsg1, were co-expressed in COS cells, many of the cells present in the population exhibited a pattern of dots in the cytoplasm (arrowheads in Fig. 7) and at cell borders (arrows in Figs 6 and 7). The extent of cell-cell border staining varied depending on expression levels within individual cells in the population, ranging from more widely-spaced dots to more heavily stained but punctuated borders. In high expressers the dots frequently appeared stitched together into a more continuous but bumpy pattern that was distinct in character from the smoother swathlike borders observed in cells expressing plakoglobin, DP and E-Dsg1 (compare Figs 6A-D and 7A-B with Fig. 2C,D) or PKP1 and DP only (compare Figs 6A-D and 7A-B with Fig. 3E,F). The borders seen within the population of cells ectopically expressing PKP1, DP and the E-Dsg1 chimera also tended to be more continuous in nature than those in quadruply transfected cells (compare Figs 6A-D and Fig. 7A-B with Fig. 5A-D). Thus, in this reconstitution system, ectopic PKP1 and plakoglobin appeared to be required to efficiently cluster desmosomal plaque components into punctate structures.

In order to more closely examine differences in cytoplasmic and border staining among transfected populations and to compare the appearance of assembled structures to desmosomal plaques, ultrastructural analysis was performed. Due to the presence of some endogenous plakoglobin in COS cells, particular focus was placed on a comparison of triple transfectants expressing E-Dsg1, plakoglobin and DP with quadruple transfectants expressing E-Dsg1, plakoglobin, PKP1 and DP. Cells in triply transfected cell populations exhibited extended plaque structures at cell-cell borders not found in untransfected or vector only controls (compare Fig. 8A to D). These plaques had a fine filamentous substructure which appeared similar to meshworks previously seen in the cytoplasm of COS-7 cells transiently transfected with DP constructs lacking the membrane-associated, multi-layered, paired and unpaired plaques grouped together in the cytoplasm. Nothing resembling these structures was seen in the triply transfected cell populations lacking PKP1 or in control cells. Thus it seems likely that these membrane-associated structures corresponded to the dots seen in the cytoplasm of quadruply transfected cells (Figs 6 and 7) which co-stained in triple label immunofluorescence experiments (Fig. 7, arrowheads). The plaques were often semicircular or arc-shaped, apparently associated with membrane vesicles (Fig. 8B,C,E,F). A thinner plasma membrane targeting domain (Stappenbeck and Green, 1992). The plaques lacked an electron dense region subjacent to the plasma membrane (compare Fig. 8A to D).

Transfected cells in populations expressing E-Dsg1, plakoglobin, PKP1 and DP were readily identified due to the abundance of approximately 20 nm electron dense region resembling the outer dense plaque of desmosomes, was subjacent to a less electron dense, fibrous region measuring ∼75 nm. This fibrous region of the plaque was thicker than that associated with the smaller desmosomes in control COS cells (compare control in Fig. 8D with E). In many areas, the plaque structures were associated with, and even connected by, surrounding 10 nm IF (Fig. 8B′,C,F, arrow). A proportion of these intracellular structures in each transfected cell exhibited paired mirror-image plaques (Fig. 8B′ and C). Electron dense plaques were also present on plasma membranes, both in the presence (Fig. 8H) and absence (Fig. 8G) of neighboring cells. Plaques at cell-cell borders occupied varying lengths of plasma membrane, and frequently appeared to be composed of shorter structures fused in linear arrays and interrupted by less electron dense regions (e.g. arrow in Fig. 8H).

During desmosome assembly precursor proteins are recruited from a Triton-soluble to a Triton-insoluble pool, which is thought to reflect assembly into intercellular junctions and

association with the IF cytoskeleton (Pasdar and Nelson, 1988; Pasdar and Nelson, 1989; Penn et al., 1987). To assess the impact of co-expression of specific desmosomal proteins on their solubility, a series of co-transfections was carried out, followed by fractionation into Triton-soluble and Triton insoluble pools. As shown in Fig. 9, the E-Dsg1 chimera is largely soluble when expressed on its own, or co-expressed with plakoglobin, PKP1 or both arm proteins. Co-expression of E-Dsg1 and full length DP shifts the cadherin from a soluble:insoluble ratio of ∼75:15 to ∼50:50. Co-expression of E-Dsg1 with both arm proteins and DP leads to an additional decrease in E-Dsg1 solubility to ∼30:70. The majority of protein in the quadruply tranfected cells is found in the Triton-insoluble pool, likely representing the immunofluorescent structures seen in Figs 6 and 7 and plaque-containing structures seen in Fig. 8.

In this study we have investigated the contribution of components found in desmosomes of the differentiated layers of complex tissues to the assembly of the desmosomal plaque. In particular our results shed light on the relative roles played by two members of the desmosomal arm family, plakoglobin and PKP1.

It is now possible to make a comparison between the protein complexes in which classic versus desmosomal cadherins participate. In the case of classic cadherins, it has been shown that the E-cadherin tail binds directly to β-catenin which in turn binds to α-catenin (Jou et al., 1995). Here we have demonstrated that there exists a triple complex containing DP, plakoglobin, and a desmosomal cadherin tail. The presence of E-Dsg1 in a complex with the N-terminal DPNTP domain is dependent on the presence of plakoglobin. This observation is consistent with previous data showing that the Dsg1 tail did not recruit the N-terminal DPNTP region of DP to COS cell borders in the absence of ectopically expressed plakoglobin (Kowalczyk et al., 1997). In addition, E-Dsg1 is not observed in the complex when plakoglobin is replaced with β-catenin, consistent with previously reported yeast two hybrid data showing that DP is unable to bind directly to β-catenin (Kowalczyk et al., 1998).

In addition, in contrast to α-catenin, which interacts only indirectly with E-cadherin through β-catenin (Jou et al., 1995), full length DP was associated with the Dsg1 tail in the absence of ectopically expressed plakoglobin, albeit at a somewhat reduced level. This experiment was done in HT1080 cells which express extremely low levels of plakoglobin that are undetectable in this experiment, suggesting the possibility that sequences in full length DP, downstream of the first 584 amino acids, can interact under certain conditions with the Dsg1 tail in a plakoglobin-independent fashion. Although it seems likely that in desmosome-containing cells desmosomal cadherins are usually associated with an arm family member, interactions outside the arm binding domain could provide another mechanism for stabilizing the complex and promoting lateral interactions in the plaque. We cannot rule out the presence of endogenous HT1080 proteins in the observed FL DP: E-Dsg1 complex; however, it seems likely that in this transient overexpression system such proteins would be present in minor amounts.

Supporting the idea that plakoglobin-independent interactions between DP and desmosomal cadherins can occur, an interaction between DP and the desmocollin 1 tail was reported (Troyanovsky et al., 1994). More recent studies have suggested the Dsc1 tail can interact directly with DP, and that the first 176 amino acids of DP is sufficient for this interaction (Smith and Fuchs, 1998). It is not known whether the ability to bind to cadherins through both plakoglobin-dependent and independent interactions is a unique property of Dsg1 and Dsc1, or whether it is a common trait held by other desmosomal cadherins. However, it is interesting to note that both Dsc1 and Dsg1 are differentiation-specific cadherins, expressed together in the superficial layers of complex epithelia (King et al., 1996; Nuber et al., 1996). It seems plausible that the existence of multiple sites for DP on the Dsc1/Dsg1-catenin complex could confer additional mechanical strength to the superficial cells of the epidermis, enabling these tissues to better withstand mechanical stress.

The continued growth of the plakophilin family of desmosomal molecules highlights the fact that the desmosomal plaque is much more complex than once appreciated (Bonne et al., 1999; Hatzfeld et al., 1994; Heid et al., 1994; Mertens et al., 1996; Schmidt et al., 1999; Schmidt et al., 1997). Previous reports describing a skin fragility syndrome in patients lacking PKP1, suggested that in desmosomes of the superficial epidermis, this arm protein plays a critical role in maintaining tissue integrity (McGrath et al., 1999; McGrath et al., 1997). The work described here which showed that PKP1 efficiently concentrates endogenous and ectopically expressed full length DP at the plasma membrane is consistent with previous studies (Hatzfeld et al., 2000; Kowalczyk et al., 1999b; Smith and Fuchs, 1998). In addition, the abundance of 10 nm IF seen associated with the plaque structures in quadruply transfected cells (Fig. 8) highlights the role of PKP in recruiting IF to the desmosomal plaques. Together, these observations support our proposed model that PKP1 increases the number of IF binding sites in desmosomes of the superficial epidermis by recruiting its preferred binding partner, DP, to the membrane (Kowalczyk et al., 1999b). A contribution for PKP1 in directly anchoring IF to the desmosomal plaque, supported by in vitro binding studies (Kapprell et al., 1988; Hofman et al., 2000; Hatzfeld et al., 1994), cannot be ruled out in our study, but has recently been questioned in light of PKP1’s position immediately proximal to the plasma membrane (North et al., 1999). A structural role for PKP1 in this membrane-proximal region is also suggested by the increased electron density of the outermost plaque in quadruply transfected cells compared with triply transfected cells lacking PKP1 (Fig. 8).

Most plakoglobin null mice die early in embryogenesis, probably due to the inability of the heart to withstand mechanical stress, and those that survive exhibit epidermal defects. Thus, like PKP1, plakoglobin also plays an important role in tissue integrity (Bierkamp et al., 1996; Ruiz et al., 1996). However, it must do so in the context of plakophilins, as it is now clear that all desmosomes are likely to have one or more plakophilins in addition to plakoglobin in the cytoplasmic plaque. Here we investigated the relative roles played by plakoglobin and PKP1 in the formation of desmosome-like structures.

In quadruple transfections when both plakoglobin and PKP1 were introduced into cells with E-Dsg1 and DP a pattern of dots in the cytoplasm and at cell-cell borders was observed. These cytoplasmic dots and punctuated borders differed from the continuous staining visible at cell-cell borders in cells transfected with all other combinations of plasmids. Furthermore, the solubility profile of components in the quadruple transfectants was what might be predicted based on studies of normal epithelial cells. That is, the majority of DP (∼80%), PKP1 (∼75%) and E-Dsg1 (70%) were present in a Triton-insoluble pool of proteins when all four proteins were co-expressed. These light microscope observations were corroborated by ultrastructural analysis of transfected cells which revealed structures in the cytoplasm and on cell surfaces with multi-layered, membrane-associated electron dense plaques tethered to IF. These plaques appeared more robust than the smaller, more rarely seen desmosomes in untransfected or vector transfected COS cells (compare Fig. 8D with B′ and C), and more closely resembled desmosomes in stratified epithelia.

Plasma membrane-associated plaques were frequently seen in the absence of cell-cell contact. A previous report from Demlehner et al., suggested that in normal cultured epithelial cells desmosomal half plaques are continuously synthesized and subsequently endocytosed in the absence of cell-cell contact to stabilize the structure (Demlehner et al., 1995). Although these authors could not exclude the possibility that the vesicles they observed formed in the cytoplasm without first assembling at the plasma membrane, they suggested that most formed by endocytosis. Whether the intracellular vesicles we observed in transfected cells formed first in the cytoplasm and were then delivered to the plasma membrane, or formed first at the plasma membrane and were subsequently endocytosed is not known. However, in these highly expressing transfected cells, it seems possible that both could occur.

Together, the observations presented here are consistent with the idea that a desmosomal cadherin tail, a plakophilin, plakoglobin and DP together assemble a clustered plaque structure with cytoplasmic features similar to that of a desmosome. Although plaque lengths in quadruply transfected cells varied at cell-cell borders, there was a general trend that expression of both arm proteins together led to more clustered borders (Figs 6, 7, arrows). A role for plakoglobin in limiting the size of desmosomes is suggested by the increased size of desmosomes in the hearts of plakoglobin null mice (Ruiz et al., 1996) and in A431 cells expressing C-terminally truncated plakoglobin molecules (Palka and Green, 1997). These observations are consistent with the transformation of the continuous border staining described above into interrupted cell borders when both PKP1 and plakoglobin are present. Based on the sequestration of E-Dsg1/PKP1/DP complexes along filament arrays in the absence of co-expressed plakoglobin it is also possible that plakoglobin ensures the proper behavior and transport of the desmosomal cadherins to the plasma membrane. Such a role was recently suggested for β-catenin in the transport of E-cadherin to basolateral membranes of MDCK cells (Chen et al., 1999). Interestingly, plakoglobin by itself does not appear to be sufficient to limit the size of the plaque, as cells expressing E-Dsg1, DP and plakoglobin exhibited more continuous border staining as shown in Fig. 2. It was only when plakoglobin and PKP1 were both expressed that plaques of desmosome-like appearance and dimension were generated, suggesting that this event requires the cooperative activity of both proteins.

These observations raise an apparent paradox. That is, how does this cooperation occur when PKP1 binds to DP at the expense of plakoglobin? The DP-PKP1 complexes analyzed in Fig. 5 represent interactions in the soluble pool only. It seems likely that DP complexes containing both PKP1 and plakoglobin do exist in the insoluble pool representing the cytoplasmic dots and puncta containing all three proteins shown in Figs 6 and 7 and plaque structures in Fig. 8. The data suggest furthermore that the presence of PKP1-DP and E-Dsg1-Pg-DP complexes favors the assembly of multiple plaques of more limited size. Precisely how this occurs at a molecular level is unclear. One contributing factor is likely to be the repertoire and affinity of protein-protein interactions that can occur. So, for instance, PKP1 and plakoglobin do not interact (X.-T. Chen and K. Green, unpublished). Thus, a possible scenario could be that insertion of a particular component, such as PKP1, might inhibit further lateral outgrowth of plaques via plakoglobin when both are present. The final size of desmosomes and extent of clustering would be dependent on the subcellular localization and availability of plakoglobin and PKP1 for assembly and incorporation into the growing plaque. In addition, the stoichiometry of components, something that is difficult to regulate in the transient approach used here, is also likely to play an important role in regulating the size and dimensions of the plaque. Thus it is not surprising that there is substantial variation in border lengths observed in quadruply transfected cells, which could be explained by varying ratios of plakoglobin and PKP1 expressed in these cells. Finally, desmocollins are normally a major desmosomal constituent. In the current work, we have bypassed the requirement for desmocollins at the adhesive interface by replacing the Dsg1 extracellular domain with the E-cadherin extracellular domain. However, both the longer ‘a’ and shorter ‘b’ form of the desmocollins most likely make important contributions to the normal plaque structure of desmosomes.

Our results provide several critical insights into the rules that govern desmosome assembly. They suggest that the repertoire of arm family members, in concert with the cadherin tails, plays an important role in determining the structure and dimensions of the plaque, as well as the number of IF binding sites along the length of the plaque. The overlapping expression patterns of desmosomal components in stratified tissues would thus provide an opportunity for tailoring desmosome structure and function in response to different functional requirements that occur during epithelial differentiation.

The authors thank Cheryl Olson for her generous assistance with preparation of materials for EM. Thanks go also to those who provided antibody reagents including Drs Peggy Wheelock and Keith Johnson, Pam Cowin, Stephen Byers, Masatoshi Takeichi, and Fiona Watt. Members of the Green lab provided helpful discussion and critical reading of the manuscript. This work was supported primarily by grants RO1 AR43380 and project #4 of PO1 DE12328 to K. Green, with additional support from AR41836.