The Function of Plakophilin 1 in Desmosome Assembly and Actin Filament Organization

Desmosomes are adhering junctions that anchor intermediate filaments to sites of cell–cell contact. Biochemically, they are distinct, but are related to the adherens junctions that anchor actin filaments. They contain two types of transmembrane proteins of the cadherin superfamily, the desmogleins (Dsgs) and desmocollins (Dscs). There are at least three different desmogleins, and three different desmocollin genes (Dsg1-3 and Dsc1-3) that are differentially expressed (Koch and Franke 1994; Schmidt et al. 1994). Whereas Dsg2 and Dsc2 are ubiquitously expressed in all cells that possess desmosomes, the expression of Dsgs1 and 3 and Dscs1 and 3 is restricted to stratified epithelia (Schmidt et al. 1994; Garrod et al. 1996). Expression patterns of isoforms of the desmosomal cadherins overlap, and individual desmosomes can contain more than one isoform (North et al. 1996). Clustering of desmosomal cadherins and desmosome formation depends on both Dsgs and Dscs (Chitaev and Troyanovsky 1997).

The intracellular domains of the desmosomal cadherins associate with a number of plaque proteins that establish the link to the intermediate filament system (Troyanovsky et al. 1993, Troyanovsky et al. 1994a, Troyanovsky et al. 1996; Mathur et al. 1994; Chitaev et al. 1996; Kowalczyk et al. 1996; Witcher et al. 1996). Plakoglobin and desmoplakin are essential components of the plaque. Plakoglobin associates with both types of desmosomal cadherins and binds to several Dsg and Dsc isoforms (Mathur et al. 1994; Troyanovsky et al. 1994a,Troyanovsky et al. 1994b, Troyanovsky et al. 1996; Chitaev et al. 1996; Wahl et al. 1996; Witcher et al. 1996). The binding of plakoglobin to E-cadherin is a prerequisite for desmosome formation, and its COOH terminus is involved in regulating desmosome size (Ruiz et al. 1996; Lewis et al. 1997; Palka and Green 1997). Moreover, a signaling role in the Wnt pathway, which is similar to that of β-catenin and the Drosophila homologue armadillo, has been reported (Karnovsky and Klymkowsky 1995; Rubenstein et al. 1997). Direct interactions between desmosomal cadherins and desmoplakin have been reported only in vitro (Smith and Fuchs 1998), and it appears that plakoglobin is necessary to link these proteins in vivo (Kowalczyk et al. 1996, Kowalczyk et al. 1997). Desmoplakin binds to intermediate filaments through its COOH-terminal domain and connects desmosomes to the cytoskeleton (Stappenbeck et al. 1993, Stappenbeck et al. 1994; Kouklis et al. 1994; Bornslaeger et al. 1996; Meng et al. 1997). Thus, intermediate filaments seem to be linked to the plasma membrane through a linear sequence of interactions between keratins, desmoplakin, plakoglobin, and the cytoplasmic tail of cadherins.

Additional components of the desmosomal plaque are plakophilins 1, 2, and 3, and p0071 (Hatzfeld et al. 1994; Heid et al. 1994; Hatzfeld and Nachtsheim 1996; Mertens et al. 1996; Bonne et al. 1999; Schmidt et al. 1999). These proteins contain a central domain that consists of a series of 45 amino acid repeats (arm repeats), and are members of the p120^ctn family of armadillo (arm) -related proteins (Reynolds et al. 1994; Hatzfeld and Nachtsheim 1996; Daniel and Reynolds 1997). Plakophilin 1 is a major component of desmosomes from stratified and complex epithelia, and it is predominantly expressed in the suprabasal layers (Kapprell et al. 1988). It binds to keratins in vitro (Kapprell et al. 1988; Hatzfeld et al. 1994; Smith and Fuchs 1998), but the significance of this interaction in vivo has not yet been established. More recently, plakophilin 1 has been described as a widespread nuclear protein that is also expressed in nondesmosome-bearing cells, where it accumulated in the nucleoplasm (Schmidt et al. 1997; Klymkowsky 1999). The function of plakophilin 1 in the nucleus remains unknown so far. An essential role in desmosome organization and stability has been suggested recently on the basis of a genetic skin disease. Patients lacking plakophilin 1 suffer from a skin fragility syndrome. Desmosomes in their skin are small and poorly formed with widening of keratinocyte intercellular spaces and perturbed desmosome/keratin filament interactions (McGrath et al. 1997). Desmoplakin was found predominantly cytoplasmic in these patients, suggesting a role for plakophilin 1 in organizing suprabasal desmosomes. These findings point to an essential role of plakophilin 1 in establishing stable cell contacts, desmosomal plaque size, and organization. In a recent paper, a direct interaction between the desmoplakin NH₂ terminus and plakophilin 1 and a role of plakophilin 1 in recruiting desmoplakin to the membrane was described (Kowalczyk et al. 1999). It was proposed that this interaction may be important for clustering of desmosomal components through lateral interactions.

To learn more about the function of plakophilin 1, we have analyzed its function in desmosome assembly in more detail. Wild-type plakophilin 1 recruited endogenous desmosomal proteins into the plaque when overexpressed in keratinocytes. This function was mediated by its head domain, and we show that this domain interacts with Dsg1, desmoplakin, and keratins. In contrast, the arm repeat domain had a dominant negative phenotype: it promoted formation of filopodia and long cellular protrusions, where it colocalized with actin filaments. Deletion of a conserved motif in the center of the arm domain abolished the ability of plakophilin 1 to modulate cellular morphology and to associate with actin. Our data suggest that plakophilin 1 is involved in regulation of desmosome assembly as well as dynamics of the actin cytoskeleton.

RNA was prepared according to the LiCl/urea extraction method (Auffray and Rougeon 1980), and cDNAs were synthesized by reverse transcriptase–PCR, with expand reverse transcriptase and expand high fidelity polymerase (Roche Diagnostics). Suitable restriction sites for cloning were included in the primer sequences. PCR products were either directly cloned into the expression vectors or first ligated into the PCRII vector using the TOPO TA cloning kit (Invitrogen BV). All PCR products were sequenced completely.

Prokaryotic expression was performed in the pRSET A vector, which includes an NH₂-terminal His tag (Invitrogen Corp.). Expression in eukaryotic cells was performed with the following vectors: pCMV5 (with NH₂-terminal T7 tag; Andersson et al. 1989), pCMVscript (without tag; Stratagene), pOPRSV (without tag; Stratagene), and pEGFP (with NH₂-terminal GFP tag; CLONTECH Laboratories). Vectors for the expression of GAL4 fusion proteins in yeast were the pBD and pAD vectors (Stratagene) and the pAS2-1 and pGAD424 vectors (CLONTECH Laboratories). Fig. 1 a gives an overview of the plakophilin 1 constructs used in this study. The intracellular domains of human Dsg1, Dsg2, and Dsg3, and Dsc1a, 1b, Dsc2a, 2b, and Dsc3a, 3b were amplified by reverse transcriptase–PCR from HaCaT cell RNA and cloned into the pGAD424 vector. Dsg1 domains IA (amino acids 568–643), CS (amino acids 644–764), and the Dsg-specific domain (Dsg; amino acids 765–1,049) were amplified by PCR and cloned into the pGAD424 vector. Human keratins 8, 18, 6, and 17 and their individual domains in pAD and pBD have been described elsewhere (Schnabel et al. 1998). The complete coding sequence of human β-actin was amplified by PCR from HeLa cell RNA and cloned into pGAD424.

In vitro mutagenesis was performed to delete the 5–amino acid motif ENCMC (amino acids 452–456) in the plakophilin 1 arm repeat domain. The reaction was performed with the QuickChange site-directed mutagenesis kit (Stratagene). The deletion was verified by sequence analysis.

Plasmids were transformed into the yeast strain YRG2 (Stratagene) by electroporation. Double transformants were grown on plates lacking leucine and tryptophane. Expression of the His reporter gene was analyzed on plates lacking histidine in addition to leucine and tryptophane. lacZ reporter gene expression was analyzed in the colony lift filter assay and quantitated using the ONPG (o-nitrophenyl-β-d-galactopyranoside) substrate as described in the yeast protocols handbook (CLONTECH Laboratories).

The plakophilin 1 head and arm repeat domains in pRSET were expressed in BL21 DE3 bacteria and purified under denaturing conditions on Ni-NTA resin (Qiagen). The purified protein fragments were used for immunization of rabbits.

HeLa and HaCaT (Boukamp et al. 1988) cells were routinely cultured in DME supplemented with 10% FBS. Normal human epidermal keratinocytes cells and keratinocyte medium were obtained from Promocell. Wound healing assays were performed with HaCaT cells grown to confluency, and a wound was inserted by scraping. Cells were analyzed 24 h after wounding. For transient transfection experiments, cells were plated 12–16 h before transfection. Cells were transfected either by the calcium phosphate precipitation method (5prime→3prime, Inc.) or using the DOSPER liposomal transfection reagent (Roche Biochemicals). Cells were fixed and processed for immunofluorescence analysis after 20–44 h.

Cells grown on coverslips were rinsed in PBS and fixed in methanol at −20°C for 10 min, followed either by acetone treatment for 1 min or by treatment in 0.2% Triton X-100 in PBS for 20 min. Alternatively, cells were fixed in 3.7% formaldehyde in PBS freshly prepared from paraformaldehyde and permeabilized in 0.2% Triton in PBS. Cells were washed in PBS and incubated with 1% BSA in PBS before antibody application.

Plakophilin 1 and its fragments were detected by the polyclonal rabbit sera against the head and repeat domains. Alternatively, plakophilin 1 head and repeat domains, which were expressed in the pCMV5 vector, were detected with a T7 mAb (Novagen, Inc.). For double labeling the following antibodies were used: antidesmoplakin 1 and 2, DP1&2 2.15 or DP1&2 mix (2.15 + 2.17 + 2.20); anti-Dsg1 + 2 DG3.10, anti-Dsc3 Dsc3 U114 were obtained from Progen. Antiplakoglobin and anti–Pan-cadherin were from Sigma Chemical Co.; the keratin antibody RCK107 was from Dr. F. Ramaekers (University Hospital, Rotterdam, The Netherlands).

Secondary antibodies were donkey anti–rabbit or anti–mouse coupled to Cy2 or Cy3 (Jackson ImmunoResearch Laboratories, Inc., through Dianova) or Alexa 488 goat anti–mouse or goat anti–rabbit IgG (Molecular Probes, Inc.). Actin filaments were visualized by incubation with FITC- or TRITC-labeled phalloidin (Sigma Chemical Co.).

Microscopy was carried out with a Nikon Eclipse E600 microscope with narrow band filters.

Cells processed for immunofluorescence microscopy were analyzed using a Zeiss LSM 510 laser scanning microscope equipped with a helium-neon and an argon laser and a Plan-Apochromat 63× objective. Excitation wavelengths were 488 nm for Alexa 488 and 543 nm for Cy3. The used detection filters were BP505-530 for Alexa 488 and LP560 for Cy3. Fluorescence was recorded using the multitracking procedure to get complete separation of the fluorescence signals.

Total protein extracts were prepared by adding SDS sample buffer heated to 100°C to the cell culture dishes. Yeast protein extracts were prepared according to the SDS-urea method in the presence of the complete protease inhibitor cocktail tablets (Roche Diagnostics), as described in the Yeast Protocols Handbook. Samples were separated on 8 or 10% acrylamide gels and transferred to nitrocellulose. Filters were blocked in 5% nonfat dry milk in TBS with 0.05% Tween 20. Primary antibodies were applied for 2 h at room temperature or overnight at 4°C. Filters were washed and incubated with alkaline phosphatase–coupled secondary antibodies, and bound antibodies were visualized either with the CDP-Star chemiluminescence reagent (Tropix) or with NBT/BCIP (Boehringer Ingelheim Bioproducts). In some experiments, the ECL detection system (Amersham Pharmacia Biotech) was used.

To address the function of plakophilin 1 in desmosome assembly and structure, we studied targeting of its domains in epithelial cells and analyzed its direct binding partners in the yeast two-hybrid system. Fig. 1 a summarizes the plakophilin 1 constructs tested in transfection assays and in the yeast two-hybrid system. The GFP constructs of all domains were analyzed in parallel with nontagged or T7-tagged constructs to verify that the GFP tag did not interfere with intracellular sorting.

Rabbit polyclonal antibodies against the plakophilin 1 NH₂-terminal domain and the arm repeat domain were generated and tested for their specificity by Western blotting on total cellular extracts. Fig. 1 b shows that both antibodies reacted with a single band of 80 kD, demonstrating that they did not cross-react with related proteins, such as plakophilin 2 (96 kD) and 3 (86 kD), p120^ctn (various isoforms of 96–115 kD), or p0071 (130 kD). The majority of the protein was detected in the insoluble protein fraction.

Since plakophilin 1 has been described as a protein with dual localization in desmosomes and in the nucleus (Schmidt et al. 1997), we analyzed intracellular targeting of the protein after overexpression. Attempts to obtain clonal cell lines that strongly overexpress plakophilin 1, or its head, or repeat domain, thus far, have been unsuccessful. This may be due to the phenotype that is caused by strong overexpression of plakophilin 1 or its fragments (see below). Therefore, we have used transient transfection studies to analyze the function of plakophilin 1 and its domains in a cellular context. Wild-type plakophilin 1, which was overexpressed in HaCaT keratinocytes, localized predominantly to the nucleus and to cell borders in confluent monolayers (Fig. 2 a), which is in agreement with the intracellular localization of the endogenous protein (Schmidt et al. 1997). The balance between nuclear localization and plasma membrane association appeared similar in transfected and nontransfected cells. Double labeling with desmoplakin antibodies revealed a strong increase of endogenous desmoplakin at the plasma membrane in the transfected cells compared with nontransfected cells (Fig. 2 a′).

To identify the domains that target plakophilin 1 to desmosomes and to the nucleus, the head and the arm repeat domains of plakophilin 1 were expressed separately. Whereas the arm repeat domain colocalized with the actin cytoskeleton (see below), the head domain, like the full-length protein, was detected in the nucleus and along the cell periphery (Fig. 3, a–e) and strongly enhanced recruitment of desmoplakin to the plasma membrane (Fig. 3 a′). Costaining for other desmosomal proteins revealed that recruitment of Dsg (Fig. 3 b′), Dsc (Fig. 3 c′) and, to a lesser extent, plakoglobin (Fig. 3 d′) was also enhanced. The amount of recruited protein roughly correlated with the size of the membrane pool of plakophilin 1. In cells with a large membrane pool of plakophilin 1, recruited proteins were detected continuously along the plasma membrane (Fig. 3, a–d'). In other cells, the typical punctate pattern of individual desmosomes was retained (Fig. 4). Costaining for keratins showed colocalization of a small pool of these proteins to plasma membrane patches enriched for plakophilin 1 (Fig. 3e and Fig. e′). These data demonstrate that plakophilin 1 is able to recruit various desmosomal plaque proteins to the plasma membrane, and that this effect is mediated by its head domain.

In addition to its plasma membrane association, the head domain showed very strong nuclear localization. Surprisingly, some desmoplakin, Dsg, and Dsc were also detected in the nucleus, suggesting that plakophilin 1 coimported a fraction of these proteins into the nucleus.

To analyze the recruitment of desmosomal plaque proteins to the plasma membrane in more detail, we used laser scanning microscopy on HaCaT cells expressing the head domain of plakophilin 1. Whereas plakophilin 1 and E-cadherin staining overlapped only very little (Fig. 4 a), a high degree of overlap was found between plakophilin 1 and desmoplakin staining (Fig. 4 b), demonstrating that the major portion of overexpressed plakophilin 1 head domain does not localize to adherens junctions. These data indicate that plakophilin 1–mediated recruitment of proteins occurs primarily in desmosomes. To investigate the effect of the recruitment on desmosome size and number, we quantitated desmoplakin staining at cell borders by scanning along plasma membrane stretches (Fig. 4b′ and b′′, arrows). The data are displayed as fluorescence intensity profiles below the corresponding image. The number and size of the peaks within these profiles were significantly higher when recorded along cell borders of two transfected cells (Fig. 4 b′), compared with the cell border between transfected and nontransfected cells (Fig. 4 b′′). Assuming that each peak represents a desmosome or a group of desmosomes, these data indicate that plakophilin 1–mediated recruitment of plaque proteins might result in the generation and enlargement of desmosomes.

To determine the region within the plakophilin 1 head domain responsible for desmosome association, several fragments were constructed (Fig. 1 a). Whereas all of them were still able to associate with desmosomes in HaCaT cells (Fig. 5), only the ΔN1, ΔN2, and ΔC1 fragments were capable of significantly enhancing the recruitment of endogenous desmoplakin (Fig. 5) and other desmosomal plaque proteins (data not shown) to the cell membrane. The ΔC2 fragment did not recruit endogenous desmosomal proteins, although it associated with desmosomes. A major portion of the ΔN2 and ΔC2 fragments remained cytoplasmic (Fig. 5). All fragments were still able to enter the nucleus, but nuclear targeting was more efficient with the ΔN2 and ΔC2 constructs. These experiments show that at least one region mediating plasma membrane targeting of plakophilin 1 as well as a signal directing nuclear localization is retained in all head deletion constructs.

When overexpressed in simple epithelial HeLa cells, plakophilin 1 accumulated in the nucleus, but was not recruited to the plasma membrane (Fig. 6 a), suggesting that the nuclear function of plakophilin 1 is conserved among all cells, whereas its function in stabilizing intercellular junctions is restricted to certain cell types. The lack of desmosome association of plakophilin 1 in HeLa cells may be due either to the lack of an appropriate binding partner such as cell type–specific Dsg and/or Dsc isoforms, or to different regulatory mechanisms that control modification and/or assembly of desmosomal proteins in HeLa cells. In addition to its nuclear localization, plakophilin 1 was found along actin filaments, as demonstrated by double labeling with phalloidin (Fig. 6, a and a′).

Transfection studies with the plakophilin 1 head domain in HeLa cells showed almost exclusive nuclear localization of the fragment (Fig. 6 b). Decoration of actin filaments was not observed, suggesting that the binding site for direct or indirect actin filament association is located in the arm repeat domain (see below). Desmoplakin staining was strong in the transfected cells, but it appeared in a punctate pattern in the cytoplasm rather than in membranes (Fig. 6 b′). A similar distribution of desmoplakin was seen in mitotic cells (Fig. 6 b′, arrowheads), where desmosomal proteins have been internalized in vesicles. Nontransfected, nonmitotic cells revealed the punctate staining pattern along the plasma membrane, which is typical of desmosomes (Fig. 6 b′, arrows). The extent of cytoplasmic staining of desmoplakin seemed to correlate with plakophilin 1 expression levels. The cytoplasmic staining could be due to internalization of desmosomes and/or enhanced synthesis and assembly of desmosomal proteins in the cytoplasm (Demlehner et al. 1995). The ΔC1 (Fig. 6 c), ΔC2, ΔN1, and ΔN2 (not shown) constructs showed almost exclusive nuclear localization with the same effect on desmoplakin distribution as described above.

Since wild-type plakophilin 1 decorated actin filaments in transfected HeLa cells, we analyzed plakophilin 1 localization more carefully in nontransfected cells to distinguish whether this was an artifact due to heavy overexpression that disturbed the intracellular sorting mechanisms, or whether it was connected to a novel function of plakophilin 1. In a wound healing experiment with HaCaT cells, colocalization of actin filaments and plakophilin 1 was observed at the tips of cellular protrusions (Fig. 6d and Fig. d′), suggesting a role for plakophilin 1 in regulating actin filament organization. Association with stress fibers was not observed.

Plakophilin 1 has been shown to bind to Dsg1, Dsc1a, and desmoplakin in in vitro overlay assays (Smith and Fuchs 1998), and to Dsg1 and desmoplakin in the two-hybrid system (Kowalczyk et al. 1999). To localize the binding sites of these proteins in plakophilin 1, the cytoplasmic domains of Dsgs1-3 and Dscs1a,b-3a,b and the NH₂ terminus of desmoplakin were tested in the yeast two-hybrid system. From all the desmosomal cadherins, only Dsg1 interacted with the plakophilin 1 head domain (Fig. 7 a) and with all head domain deletion constructs (Fig. 7b and Fig. c), although the ΔN2 and ΔC2 constructs appeared somewhat less efficient in reporter gene activation, suggesting that the Dsg1 binding site was not completely retained in these constructs. Desmoplakin binding was retained in the ΔC1 and ΔC2 fragments (Fig. 7 b), but not in the ΔN1 and ΔN2 fragments (Fig. 7 c), demonstrating that desmoplakin binds close to the NH₂ terminus of plakophilin 1. These results suggest that desmoplakin and Dsg1 do not compete for the same binding site in the plakophilin 1 head.

Since plakophilin 1 and plakoglobin (Troyanovsky et al. 1993; Chitaev and Troyanovsky 1997) both bind to desmoplakin and Dsg1, we wanted to analyze if these two proteins provide alternative links between the cadherins and the cytoskeleton, or if plakophilin 1 stabilizes the Dsg-plakoglobin-desmoplakin interaction through additional protein interactions. Therefore, we determined the plakophilin 1 binding site in the Dsg1 cytoplasmic domain. The plakophilin 1 head domain interacted with the intact Dsg1 cytoplasmic domain, the Dsg + CS domain and, although to a somewhat lesser extent, with the Dsg domain alone (Fig. 7 d), indicating that the plakophilin 1 binding site differs from the plakoglobin binding site in the CS domain. The requirement of the CS domain for strong binding suggests that the plakophilin 1 binding site is close to the plakoglobin binding site, and that simultaneous binding might be prevented because of steric hindrance.

Since plakophilin 1 has been shown to bind keratins in vitro (Kapprell et al. 1988; Hatzfeld et al. 1994; Smith and Fuchs 1998), we also examined several keratin constructs for their interaction with the plakophilin 1 head domain. As shown in Fig. 7 e, the type I keratins K17 and K18 strongly interacted with the plakophilin 1 head, whereas of the type II keratins tested, only K8 showed a weak interaction. Interaction studies with the head fragments revealed binding of K17 and K18 to the ΔC1 and ΔC2 constructs, but not to ΔN1 and ΔN2. The K8 binding site appeared to differ since binding was retained in the ΔN1, ΔC1, and ΔC2 constructs, but was lost in the ΔN2 fragment (not shown).

Interactions between the plakophilin 1 head fragments and desmoplakin and Dsg1 were quantitated by measuring LacZ reporter gene activation with the ONPG substrate. As shown in Fig. 7 g, the desmoplakin–plakophilin 1 and the Dsg1–plakophilin 1 interactions were much stronger with the plakophilin 1 head in the pBD vector compared with the pAS2-1 vector, although this vector allows high protein expression levels as verified by Western blotting with Gal4 and plakophilin 1–specific antibodies (Fig. 8 a). The high protein expression could either interfere with correct folding of plakophilin 1, or the desmoplakin and Dsg1 binding sites are masked by inter- or intramolecular interactions after expression in the pAS vector. Dsg1 interacted most strongly with the ΔN1 construct, which lacks the desmoplakin binding site. Interaction with the ΔC1 construct was somewhat weaker. In contrast, the ΔC2 and ΔN2 constructs showed considerably reduced reporter gene activation, suggesting that these constructs did not contain the entire Dsg1 binding site. Desmoplakin interacted most strongly with the ΔC2 construct, which lacks part of the Dsg1 binding site. The interaction was lost with the ΔN1 and ΔN2 constructs, indicating that the binding site is in the NH₂-terminal region.

Since all previous experiments had shown interactions between cadherins and the arm repeat domains, but not the end domains of arm proteins, we also analyzed whether the headless plakophilin 1 or the repeat domain interacted with any of the desmosomal cadherins. As shown in Fig. 7 f, none of the desmosomal cadherins interacted with headless plakophilin 1. The same result was obtained with the arm domain construct. Expression of the headless fragment in yeast cells was verified by Western blotting with anti-Gal4 antibodies (Fig. 8 a, lane 7′). Here, the headless fragment gave the strongest signal, indicating that a lack of protein expression did not account for the lack of binding.

The colocalization of wild-type plakophilin 1 with stress fibers as well as actin-rich structures at the tips of filopodia (Fig. 6) pointed to a possible role of plakophilin 1 in regulating the actin cytoskeleton. Therefore, we expressed the plakophilin 1 arm repeat and the headless domain in HeLa and HaCaT cells and verified expression by Western blotting (Fig. 8 b). Cells with high levels of these plakophilin 1 fragments displayed a highly unusual morphology with formation of filopodia, lamellipodia, or long protrusions (Fig. 9, a–d), which interfered with normal monolayer formation. The transfected cells often sat on top of the other cells. Cells with lower expression levels still displayed a normal cell morphology (Fig. 10, a–e). However, double labeling with desmoplakin antibodies showed that desmoplakin had been internalized, and disintegration of junctions had already begun (Fig. 10e and Fig. e′). The plakophilin 1 arm repeat domain colocalized with actin in lamellipodia (Fig. 10, a, c, and d) and sometimes stress fibers (Fig. 10c and Fig. c′), suggesting a role in regulating actin polymerization and filopodia formation. This phenotype was observed in HeLa and HaCaT cells.

A similar phenotype, the formation of long dendritelike cellular protrusions, had been observed in transfection studies with full-length p120^ctn (Reynolds et al. 1996) and δ-catenin (Lu et al. 1999). This suggested that the phenotype is conserved among p120^ctn family members, and might depend on the interaction with a common binding partner that is involved in regulating actin filament organization. To characterize this binding site in plakophilin 1, we constructed a deletion mutant that lacks a central pentapeptide motif conserved among all p120^ctn family members. The motif (ENCM/VC) is specific for this family and not detected in other arm related proteins. Transfection studies with this mutant construct (plakophilin 1 arm ΔENCMC) showed that the mutant had lost its capacity to induce changes in cell morphology and no longer associated with actin filaments in filopodia (Fig. 10f and Fig. f′). Instead, it accumulated in the cytoplasm, sometimes in an aggregated form (Fig. 10 f).

To analyze if the interaction between plakophilin 1 and actin is direct, we used the two-hybrid system. These experiments revealed no direct interaction between β-actin and the plakophilin 1 repeat domain (data not shown), suggesting that the interaction either depends on an intact microfilament or is mediated through an actin-associated protein in vivo.

Plakophilin 1 was shown to localize to desmosomes and to the nucleus, raising the possibility for a dual function in cell adhesion and signal transduction (Schmidt et al. 1997). In the present study, we have determined the regions in plakophilin 1 responsible for binding of desmosomal proteins and provide a functional analysis of the plakophilin 1 domains.

Plakophilin 1 is a desmosome-associated protein and has been shown to bind Dsg1, Dsc1a, and desmoplakin in vitro (Smith and Fuchs 1998) and desmoplakin in vivo (Kowalczyk et al. 1999). Using the yeast two-hybrid assay, we have mapped the binding sites of desmosomal proteins and keratins within plakophilin 1. We show that it is the head domain that mediates the interactions between plakophilin 1 and Dsg1, desmoplakin, as well as keratins. Whereas desmoplakin binds close to the NH₂ terminus between amino acids 1–70 of the head domain, the Dsg1 binding site is located between amino acids 70 and 213 (Fig. 11). Our data suggest that these two sites do not act independently. Dsg1 bound most strongly to the deletion construct lacking the desmoplakin binding site and vice versa. This could be due to reduced accessibility of the binding sites because of intramolecular interactions, which are similar to those described for vinculin and ERM family members (Winkler et al. 1996; Tsukita et al. 1997), or interactions with other proteins. A similar observation was made for plakoglobin, certain internal fragments of which bound better to E-cadherin than the entire molecule (Chitaev et al. 1996). Alternatively, high expression of plakophilin 1 could interfere with correct folding and thereby prevent the interaction in an unspecified manner.

The localization of the plakophilin 1 binding site within the Dsg1 cytoplasmic tail showed that it is distinct from the reported plakoglobin binding site (Mathur et al. 1994; Chitaev et al. 1996). However, the close proximity of the two binding sites could prevent simultaneous binding. In contrast to plakophilin 1, other arm family members including β-catenin, plakoglobin, and p120^ctn associate with classical or desmosomal cadherins through their arm repeat region (Hinck et al. 1994; Mathur et al. 1994; Aghib and McCrea 1995; Daniel and Reynolds 1995; Sacco et al. 1995; Shibamoto et al. 1995; Aberle et al. 1996; Chitaev et al. 1996; Reynolds et al. 1996; Troyanovsky et al. 1996; Wahl et al. 1996; Witcher et al. 1996). Moreover, mutational analysis has revealed that the arm repeat domains of β-catenin and plakoglobin were sufficient to direct nuclear localization (Funayama et al. 1995; Karnovsky and Klymkowsky 1995). It is interesting that both characteristics, cell contact association as well as nuclear localization, are conserved between β-catenin, plakoglobin, and plakophilin 1, but the domains responsible for these functions are not in the conserved sequence region.

In our two-hybrid assay, we could not confirm the interaction between plakophilin 1 and Dsc1 reported by Smith and Fuchs 1998 using an overlay assay. Since most of the two-hybrid vectors allow only low protein expression we were unable to detect expression of the Dsg, Dsc and keratin fragments by Western blotting. Therefore, we cannot unequivocally rule out the possibility that we might have missed the plakophilin 1–Dsc1 interaction because of a lack of protein expression or a lack of nuclear import of Dsc 1. Alternatively, both proteins could associate in vitro after their denaturation, but not under physiological conditions.

We also detected interactions between plakophilin 1 and keratins. We found a weak binding of K8 and strong binding of K17 and K18, suggesting a preference for type I keratins that had also been proposed on the basis of in vitro overlay assays (Kapprell et al. 1988). In contrast, Smith and Fuchs 1998 have reported that plakophilin 1 binds preferentially to type II keratins. The controversial data may be either due to the analysis of different keratins (K5 and K14 versus K8, K18, K6, and K17) or the use of different assay systems (in vitro overlay versus in vivo assays). Since plakophilin 1 is expressed in suprabasal cells of stratified epithelia, its in vivo interaction partner is probably one of the keratins specifically expressed in differentiated keratinocytes such as K10.

We have analyzed intracellular targeting of plakophilin 1 after overexpression. We chose two different cell lines for our studies, HaCaT keratinocytes, which express endogenous plakophilin 1 and consequently all its essential interaction partners, and simple epithelial HeLa cells. These cells possess desmosomes and express the ubiquitous desmosomal proteins, but lack certain cell type–specific desmosomal proteins including Dsg1 and 3, Dsc1 and 3, and plakophilin 1 (Schmidt et al. 1994). Moreover, desmosomes are less abundant and smaller in HeLa cells.

In HaCaT cells, overexpressed plakophilin 1 was found in the nucleus as well as plasma membrane associated, in agreement with the intracellular localization of the endogenous protein (Schmidt et al. 1997). Using deletion clones of plakophilin 1, we have determined which domains target plakophilin 1 to desmosomes (Table). Whereas cell contact association of other arm proteins including β-catenin, plakoglobin, and p120^ctn is mediated by their arm repeat domain (Hinck et al. 1994; Mathur et al. 1994; Aghib and McCrea 1995; Daniel and Reynolds 1995; Shibamoto et al. 1995; Aberle et al. 1996; Chitaev et al. 1996; Reynolds et al. 1996; Troyanovsky et al. 1996; Wahl et al. 1996; Witcher et al. 1996), we found that it is the head domain of plakophilin 1 that directs its localization to desmosomes as well as to the nucleus. This is consistent with the localization of the binding sites for desmosomal proteins determined in the two-hybrid system. The nuclear localization was observed in all cell types examined, indicating that the nuclear function is conserved among different cell types. In contrast, desmosome association was restricted to HaCaT cells, suggesting that binding to a cell type–specific desmosomal protein might be essential for targeting, or that regulatory mechanisms prevent the cell contact association of plakophilin 1 in simple epithelial HeLa cells. All the head domain fragments were still able to associate with desmosomes and to enter the nucleus. With the ΔC1 and ΔN1 fragments, desmosome association was preferred over the nuclear localization. This may be due to better accessibility of desmosomal binding sites in these constructs (see above).

In HeLa cells, full-length plakophilin 1 also decorated actin filaments. The head domain alone localized only to the nucleus, indicating that it is the arm repeat domain that mediates the association with the actin cytoskeleton. In HaCaT cells, we also found colocalization of the plakophilin 1 head domain with keratins along membrane patches. This could be due either to the direct interaction between plakophilin 1 and keratins, as shown in the two-hybrid system, or to recruitment via the keratin-binding protein desmoplakin. Nevertheless, these data, together with the results of the two-hybrid assay, suggest that plakophilin 1 interacts with keratins in vivo. Recruitment of desmoplakin and keratins to the plasma membrane has also been described in cells overexpressing a plakoglobin-synaptophysin chimera (Chitaev et al. 1996).

In a recent report, Kowalczyk et al. 1999 showed that the desmoplakin NH₂ terminus was recruited to the membrane when overexpressed together with plakophilin 1 in COS cells. We extended these experiments and analyzed the recruitment of various endogenous desmosomal proteins in HaCaT cells overexpressing plakophilin 1. As judged by immunofluorescence recruitment of desmoplakin, Dsg and Dsc were strongly enhanced, and that of plakoglobin was slightly enhanced, suggesting a major role for plakophilin 1 in desmosome assembly. In cells with high plakophilin 1 expression, we observed nuclear localization of other desmosomal proteins including desmoplakin, probably due to coimport mediated by plakophilin 1. This conclusion is supported by the fact that the ΔN1 and ΔN2 constructs, which lack the desmoplakin binding site, never coimported desmoplakin into the nucleus (Fig. 5).

There are two ways in which additional desmosomal proteins could be recruited to the plasma membrane. First, plakophilin 1 could bind to endogenous desmosomal proteins, target them to the plasma membrane, and thereby dramatically increase their stability. This is consistent with the finding that desmosomal proteins are usually synthesized in excess, and their cytoplasmic pool is rapidly degraded (Pasdar and Nelson 1988, Pasdar and Nelson 1989). In this model, plakophilin 1 plays a structural role in desmosome assembly. Second, plakophilin 1 might be directly involved in regulating the synthesis of desmosomal proteins in the nucleus. In this model, the induction of desmosomes would depend on a putative signaling function of plakophilin 1. Our experiments do not allow us to distinguish between these two models, since all fragments that were able to induce desmosome formation also revealed nuclear localization and, therefore, might combine the signaling and structural functions. Further, both mechanisms might contribute to the recruitment of endogenous desmosomal proteins. Plasma membrane association of desmoplakin and plakophilin 1 after combined overexpression in COS cells (Kowalczyk et al. 1999) argues for a contribution of the recruitment mechanism.

Using laser scanning microscopy, we demonstrate that plakophilin 1 preferentially associates with and recruits desmosomal proteins, and that the recruitment of desmosomal components might result in the generation and the enlargement of desmosomes. The possibility that expression of plakophilin 1 enhances desmosome formation in keratinocytes is consistent with the observation that desmosomes of suprabasal cells are larger than basal cell desmosomes, and that desmosomes are more numerous in suprabasal cells. Moreover, this finding explains why desmosomes were small and rare in a patient lacking plakophilin 1 (McGrath et al. 1997). Therefore, we propose that plakophilin 1 plays an essential role in regulating desmosome organization and size during keratinocyte differentiation.

In HeLa cells, overexpressed plakophilin 1 associated with actin filament, suggesting that it might be involved in regulating the actin cytoskeleton. This association was also observed in nontransfected cells, where plakophilin 1 colocalized with actin in normal cells at the tips of plasma membrane protrusions. A similar localization has been described for β-catenin, which interacts with the actin filament bundling protein fascin through its arm repeat domain (Tao et al. 1996). Actin filament association of plakophilin 1 also appeared to be mediated by its arm repeats. When overexpressed at high levels in HeLa and HaCaT cells, the arm repeat domain induced the formation of long cellular protrusions, supporting a possible role in the regulation of cell motility. This phenotype interfered with intercellular adhesion, and transfected cells were separated from the monolayer. Since full-length plakophilin 1 had no such effect, we conclude that desmosome association of the head domain is preferred. A similar phenotype has been described for p120^ctn (Reynolds et al. 1996) and δ-catenin (Lu et al. 1999) after expressing the full-length protein. In the case of p120^ctn, the arm domain was required for this effect (Reynolds et al. 1996), suggesting that this function is conserved in the arm domain of p120^ctn family members. We identified a 5–amino acid motif (ENCMC) that is conserved among p120^ctn family members. Deletion of this motif in the plakophilin 1 arm repeat domain abolished the ability of the mutant to associate with actin filaments and to induce the phenotype. This suggests that a protein–protein interaction mediated by this motif is responsible for this effect. Since we were unable to detect a direct interaction between the plakophilin 1 arm repeats and actin in the two-hybrid system, the interaction either requires an intact microfilament, as opposed to an actin monomer, or it is mediated by an actin-binding protein.

The phenotype in patients lacking plakophilin 1 suggested an important role for plakophilin 1 in stabilizing intercellular adhesion, although the lack of hair follicles and sweat glands suggests an additional role in certain differentiation processes (McGrath et al. 1997). Our results support the conclusion that plakophilin 1 has an important structural function and explain the role of plakophilin 1 in desmosome assembly at a molecular level. The localization of desmosomal binding sites to the head domain correlates with the finding that this domain recruits endogenous desmosomal proteins to sites of cell contact, whereas the arm repeat domain reduced cell contacts and induced the formation of motility-associated structures. In confluent keratinocytes, localization of plakophilin 1 to desmosomes is preferred over association with adherens junctions and actin filaments. This is consistent with strong intercellular adhesion in these cells. However, in cells that lack contact to adjacent cells, plakophilin 1 localizes to filopodia. Here, it may have a function in inducing junction formation as soon as the tip of the cell contacts an opposing cell. This idea is consistent with the finding that formation of actin-associated cell contacts precedes desmosome formation and is a prerequisite for desmosome formation (Lewis et al. 1994, Lewis et al. 1997). Plakophilin 1 could play a role in recruiting desmosomal proteins from the cytoplasm to the plasma membrane at sites of newly formed cell contacts.

We are grateful to K. Green for providing the desmoplakin construct for the two-hybrid studies and to F. Ramaekers for keratin antibodies. We thank K. Green for many helpful suggestions and discussions and E. Bornslaeger, K. Green (both from Northwestern University, Chicago, IL), and M. Osborn (MPI-Biophysical Chemistry, Göttingen, Germany) for critical reading of the manuscript. We would also like to thank C. Horn for technical help, C. Nachtsheim for her contribution in the initial phase of this study. We are also grateful to M. Iwig and the Zeiss company for help with the confocal microscopy.

This work was supported by grants from the Deutsche Forschungsgemeinschaft (Ha 1791/3-1 and 3-2, Ha 1791/5-1) and the BMBF.