Protein Binding and Functional Characterization of Plakophilin 2

Plakophilins are a subfamily of p120-related arm-repeat proteins that can be found in both desmosomes and the nucleus. Among the three known plakophilin members, plakophilin 1 has been linked to a genetic skin disorder and shown to play important roles in desmosome assembly and organization. However, little is known about the binding partners and functions of the most widely expressed member, plakophilin 2. To better understand the cellular functions of plakophilin 2, we have examined its protein interactions with other junctional molecules using co-immunoprecipitation and yeast two-hybrid assays. Here we show that plakophilin 2 can interact directly with several desmosomal components, including desmoplakin, plakoglobin, desmoglein 1 and 2, and desmocollin 1a and 2a. The head domain of plakophilin 2 is critical for most of these interactions and is sufficient to direct plakophilin 2 to cell borders. In addition, plakophilin 2 is less efficient than plakophilin 1 in localizing to the nucleus and enhancing the recruitment of excess desmoplakin to cell borders in transiently transfected COS cells. Furthermore, plakophilin 2 is able to associate with β-catenin through its head domain, and the expression of plakophilin 2 in SW480 cells up-regulates the endogenous β-catenin/T cell factor-signaling activity. This up-regulation by plakophilin 2 is abolished by ectopic expression of E-cadherin, suggesting that these proteins compete for the same pool of signaling active β-catenin. Our results demonstrate that plakophilin 2 interacts with a broader repertoire of desmosomal components than plakophilin 1 and provide new insight into the possible roles of plakophilin 2 in regulating the signaling activity of β-catenin.

related armadillo proteins found in both the desmosomal plaque and the nucleus (1)(2)(3)(4)(5). Each is composed of a basic N-terminal head domain followed by a series of 10 imperfect 42-amino acid repeats (arm repeats) and a short C-terminal tail (3). Two splice variants have been identified for PKP1 and PKP2, a shorter "a" form and a longer "b" form. Although the N-terminal head domains of PKPs exhibit relatively greater sequence diversity than the arm-repeat domains, a consensus sequence termed HR2 is shared by all the PKP head domains (3,4). PKP1 is concentrated in desmosomes of the suprabasal layers of stratified and complex epithelia (5). PKP3 can be detected in desmosomes of most simple and almost all stratified epithelia with the exception of hepatocytes and hepatocellular carcinoma cells (3,4). PKP2 has the broadest tissue distribution in desmosomes of all simple, complex, and stratified epithelia as well as non-epithelial tissues such as myocardium and lymph node follicles, in which PKP1 was not detected in desmosomes. PKP2 is concentrated in the basal layer of most stratified squamous epithelia, whereas PKP1 is mostly concentrated in desmosomes of the upper layers (2)(3)(4)6). On the other hand, PKP3 is more uniformly expressed in the living epidermal layers (3,4). PKPs have been detected in both desmosomes and nuclei in desmosome-possessing cells and only in nuclei in desmosome-lacking cells, but the mechanisms responsible for this dual location and their functions in these two different environments are still poorly understood.
Members of the armadillo family, to which PKPs belong, play critical structural and regulatory roles through their interactions with proteins in two related intercellular adhesive junctions, desmosomes and adherens junctions, which anchor intermediate filament (IF) networks and actin filaments to sites of cell-cell contact (7)(8)(9)(10). The best characterized interactions of the desmosomal armadillo proteins are those in which plakoglobin (Pg) participates. Pg associates directly with the cytoplasmic domains of the desmosomal cadherins, transmembrane glycoproteins of desmosomes that are further subdivided into the desmoglein (Dsg) and desmocollin (Dsc) subfamilies. Three isoforms exist for each of these subfamilies, which are expressed in a cell type-and differentiation-dependent manner (8,11,12). Plakoglobin links desmosomal cadherins to IF through its interactions with the plakin family member desmoplakin (DP). Functionally, loss of Pg function through mutation or genetic ablation leads to heart and skin defects, supporting the fact that this link between the cadherins and the IF-desmoplakin complex plays a key role in tissue integrity (13)(14)(15).
PKP1 also plays a critical role in tissue integrity, as patients null for PKP1 show histological evidence of aberrant desmosomes and poorly anchored IF and suffer from ectodermal dysplasia accompanied by skin fragility (16). Whereas DP in normal epidermis is concentrated in desmosomes, this plaque protein exhibits a largely diffuse, cytoplasmic distribution in PKP1 null patients. This observation raises the possibility that PKP1 is required for the efficient association of DP with the junctional plaque. Subsequent studies provided evidence supporting this hypothesis, showing that PKP1 can directly interact with the DP N terminus and enhance its recruitment to the plasma membrane. Furthermore, PKP1 collaborates with Pg and DP to promote the clustering of desmosomal plaque complexes at cell-cell borders and to enhance IF association with the plaque (17)(18)(19). Together, these results have led to the proposal that PKP1 is important in strengthening the desmosomal plaque by enhancing lateral protein-protein interactions. However, virtually nothing is known about the roles of other PKPs in desmosomes and how their participation in desmosomes affects the specific organization or functions of desmosomes in different tissues and stages of differentiation.
In addition to their function in intercellular junctions, some armadillo family members also play roles in cellular activities outside of the junctions. Both Pg (20) and p120 (21) have been found in the nucleus and implicated in various signaling pathways (22)(23)(24)(25)(26); however, the best-studied example is ␤-catenin. This close relative of Pg connects the cytoplasmic tails of classical cadherins with actin-binding proteins such as ␣-catenin to provide the linkage between actin filaments and adherens junctions (27). Besides its important roles in cell-cell adhesion, ␤-catenin is a critical downstream effector of the Wnt-signaling pathway, which mediates a large variety of developmental processes (28,29). The cellular level of ␤-catenin is tightly controlled by a complex of proteins that facilitates its rapid degradation, whereas activation of the Wnt-signaling pathway antagonizes the degradation process and leads to the accumulation of free, cytoplasmic ␤-catenin. ␤-Catenin can then translocate into the nucleus and activate target gene transcription in association with TCF/Lef1 transcription factors (30). Since the recognition of its role in Wnt signaling, an increasing number of proteins have been discovered to interact with ␤-catenin and regulate its signaling activity through various mechanisms (31). PKPs, like their armadillo family relatives, are also found in the nucleus. So far evidence for participation of PKPs in ␤-catenin-dependent signaling has not been reported. A recent study showed that nuclear PKP2 is complexed with the RNA polymerase III holoenzyme and interacts directly with its largest subunit RPC155 in vitro, opening up additional possibilities for the nuclear activity of PKP2 (32). However, because only a subset of PKP2 was observed in these complexes, other nuclear functions for this arm-repeat subfamily of proteins may exist.
PKP2 is the most ubiquitously expressed PKP family member, suggesting that it has broad cellular functions. In this study, to elucidate the biological functions of PKP2, we have characterized its protein interactions with other junctional molecules. Here we show that PKP2 interacts directly with the obligate desmosome components Pg and DP as well as the desmosomal cadherins Dsg1, Dsg2, Dsc1a, and Dsc2a and may interact indirectly with Dsg3. PKP2 is less efficient than PKP1 in enhancing the recruitment of excess DP to cell borders, which is consistent with the weaker interaction observed between PKP2 and DP and indicates distinct roles for these two PKPs in desmosome assembly. In addition, PKP2 can associate with non-cadherin-bound ␤-catenin and regulate its signaling activity in TOPFLASH reporter assays. This novel association with a component of adherens junctions and the Wnt-signaling pathway raises the possibility that PKP2 might participate in cross-talk between the two major types of intercellular junctions and signal transduction pathways connecting the cell surface and the nucleus.

EXPERIMENTAL PROCEDURES
Generation of cDNA Constructs-To generate the probe used for library screening, a partial PKP2 cDNA clone in pRSET-A vector (p673) was digested with BamHI and NcoI to isolate the insert nucleotides 1009 -2401 of the PKP2 a splicing variant (PKP2a) cDNA. Numbering of the PKP2a cDNA sequence is based on the mRNA sequence for PKP2a and 2b (GenBank TM accession number X97675) and the protein sequence of PKP2a (GenBank TM accession number CAA66265). This fragment was used to screen a normal human keratinocyte ZAPII library (provided by Dr. Masayuki Amagai). Clone (C1-1) containing PKP2a nucleotides 1-1849 was digested with KpnI and EcoRI to purify the fragment containing PKP2a nucleotides 1-1318, which was ligated into the EcoRI and KpnI sites of pBS vector to generate plasmid p793 (nt 1-1318). p793 was digested with XbaI, end-filled, and then digested with KpnI to cut out the insert, which was then ligated into a 3.4-kb fragment of p673, which had been digested with NcoI, end-filled, and digested with KpnI. The resulting plasmid p792 contains PKP2a nucleotides 1-2401 in pBS. To obtain the missing C terminus of PKP2a (nt 2402-2514), PCR was performed using the total phage DNA isolated from the above library with forward primer XC5Ј (5Ј-GGACCAATGC-CAACATC-3Ј) and reverse primer XC3Ј (5Ј-GTCGACGTCTTTAAGG-GAG-3Ј). The resulting PCR product, spanning nucleotides 2008 -2511 and containing an engineered SalI site at the 3Ј end, was directly cloned into the pGEM-T vector (Promega) to generate clone XC1-4. XC1-4 was double-digested with MscI and SphI to isolate the C terminus of PKP2a, which was then cloned into the 4.9-kb fragment of p792 digested with MscI and SphI, thus creating the complete PKP2a cDNA in pBS (p829). Complete sequencing of the PKP2a cDNA revealed three nucleotide differences compared with the originally published PKP2a sequence (2) (GenBank TM accession number X97675). The first difference is a C to T transition at nt 1097 that changes a Pro to Leu. Because this difference came from a phage clone and was also found in multiple EST clones, it was considered as a polymorphic difference in the PKP2a cDNA sequence. The other two differences are a G to T change at nt 1886 and an A to G change at nt 2164, both of which were found in PCR-generated plasmid p673 and were not present in any EST clones in the GenBank TM data base. So they were interpreted as errors introduced by the cloning process and corrected as described below.
A mammalian expression construct of PKP2a (p830) was constructed by ligating an EcoRI-SalI fragment from p829 into pFLAG-CMV-5 vector. The two random mutations at PKP2a cDNA sequence 1886 and nt 2164 were corrected using the QuikChange site-directed mutagenesis kit (Stratagene), and the resulting plasmid (p915) was fully sequenced to ensure that there were no additional mutations. Construct p915 is the expression construct for PKP2a used in this study. The N-terminal head domain of PKP2a (PKP2a-H) was generated by PCR using primers PHN (5Ј-GATGAATTCCACGATGGCAGCCCCCG-3Ј) and PHC (5Ј-CATGTCGACGTCTGCATTCCCCAGC-3Ј) with p915 as the template. The PCR product contains an EcoRI site and a Kozak consensus sequence at the 5Ј end and a SalI site at the 3Ј end. After ligation of the PCR product into pGEM-T vector and subsequent digestion with EcoRI and SalI, the PKP2a head domain (nt 1-1044) was ligated into pFLAG-CMV-5. The procedure of generating the remaining portion of PKP2a (PKP2a-A) including the arm-repeat domain and the C terminus into the pFLAG-CMV-5 vector was the same as above except for the primers used in the PCR reaction. The forward primer used was PAN (5Ј-GATGAATTCCACGATGGAGATGACTCTGGAG-3Ј), and the reverse primer was PAC (5Ј-CATGTCGACGTCTTTAAGG-GAGTGGTAGGC-3Ј). To generate the full-length PKP1a with a FLAGepitope tag at its C terminus, a full-length a form of PKP1 in pCMV script vector p724 (19) was used as the template for PCR using primers PKP1-nt-1745 (5Ј-CCTGCAATCTGGCAACTCTG-3Ј) and PKP1.FLAG (5Ј-CCTACTTGTCATCGTCGTCCTTGTAATCGAATCGGGAGGTGA-AG-3Ј). The PCR product that contains a FLAG epitope tag at its C terminus was subsequently blunt end-ligated into pSK vector digested with EcoRV. This intermediate construct was diagnostically digested to determine the orientation of the PCR insert and was double-digested with BstEII and SalI to isolate the PCR insert, which was then ligated into p724 digested with BstEII and SalI to generate full-length FLAGtagged PKP1a in pCMV script vector.
The HR2 domain of PKP2a contains amino acids 29 -60. To generate the expression construct of PKP2a with internal deletion of its HR2 domain (PKP2a ⌬HR2), a SOEing (splicing by overlap extension) procedure was utilized. PCR was performed with p915 as the template using primer A (5Ј-GTATCATATGCCAAGTC-3Ј) and primer B (5Ј-GGCGAGGGTCTGCTGGGAGCTGTCCAGTTG-3Ј) to obtain the PCR product AB. A second PCR was performed with p915 as the template using primer C (5Ј-CAGCAGACCCTCGCC-3Ј) and primer D (5Ј-CTG-CAGAAGCTTGAGG-3Ј) to obtain the PCR product CD. Then the final PCR was performed with a mixture of product AB and CD as the template using primer A and primer D to produce the PCR product AD. Product AD was digested with NdeI and HindIII and cloned into the 5.6-kb fragment from digesting p915 with NdeI and HindIII, thus creating PKP2a ⌬HR2. It was then used as the template to generate the PKP2a head domain with HR2 deletion (PKP2a-H ⌬HR2) following the same procedure described above for generating PKP2a-H. To generate full-length DP with a C-terminal Myc tag, a plasmid (p350) pCMV. DP.myc, which has a stop codon at nucleotide position 1754 due to mutation, was digested with DraIII and SalI. The 7.5-kb fragment was ligated with the 5.1-kb fragment from DraIII and SalI digestion of plasmid p612, pSK.DP.⌬C. This ligation generates full-length DP in the pCMV vector with a C-terminal Myc tag (p613).
The following constructs were described previously: full-length human Pg cDNA in a mammalian expression vector LK44 under the control of the human ␤-actin promoter (33); the Pg⌬N and Pg⌬C expression constructs under the control of the ␤-actin promoter (34); an expression construct encoding a polypeptide comprising the first 584 amino acids of DP, called DPNTP (35). Full-length ␤-catenin in mammalian expression vector pCAN was provided by Dr. Paul Polakis.
Mammalian expression constructs of full-length human Dsg1 with a C-terminal Myc tag and full-length human Dsg2 with a C-terminal Myc tag were described previously (33,36). The full-length human Dsg3 with C-terminal Myc tag in the expression vector was a gift from Dr. John Stanley.
Cell Culture and Transfections-COS-7, HEK293, A431, and HaCaT cells were cultured in Dulbecco's minimal essential medium with penicillin/streptomycin and 10% fetal bovine serum. The SCC9 oral squamous cell carcinoma cells were grown in Dulbecco's minimal essential medium/F-12 medium with 10% fetal bovine serum and penicillin/ streptomycin. The SW480 human colon carcinoma cell line was cultured in Leibovitz's L-15 medium supplemented with penicillin/streptomycin and 10% fetal bovine serum. For transient transfection of COS cells and HEK293 cells, calcium phosphate transfection was performed as previously described (37), and experiments were performed 42-46 h later. SCC9 and SW480 cells were transfected using FuGENE 6 reagent (Roche Applied Science) according to the manufacturer's protocol and assayed 24 h later.
Antibodies-A rabbit polyclonal antibody poly-Myc 2026 directed against the c-Myc epitope and a rabbit polyclonal antibody 1880 against Dsg3 were kindly provided by Dr. John Stanley. A rabbit polyclonal antibody against E-cadherin was a gift from Drs. Randy Marsh and Robert Brackenbury. A mouse monoclonal antibody PC28 directed against the DP central rod domain was a gift from Dr. Pamela Cowin. A monoclonal anti-c-Myc antibody 9E10 and rabbit polyclonal antibodies against ␤-catenin (C2206) and ␣-catenin (C2081) were purchased from Sigma. A monoclonal anti-␤-catenin antibody C19220 was obtained from Transduction Laboratories. A rabbit polyclonal antibody Oct-A probe against the FLAG epitope was purchased from Santa Cruz Biotechnology, Inc. The following antibodies were described previously: rabbit polyclonal antibodies NW161 (35) and NW6 (38), directed against desmoplakin; a chicken polyclonal antibody 1407 (39), directed against plakoglobin; a mouse monoclonal antibody 6D8, directed against Dsg2 (35,40).
Co-immunoprecipitation, Immunoblot, and Sequential Detergent Extraction-For co-immunoprecipitations, 100 or 60 mm dishes of cells were washed in PBS, extracted in 1 ml or 500 l of cold co-immunoprecipitation buffer containing 1% Triton X-100, 145 mM NaCl, 10 mM Tris-HCl, pH 7.4, 5 mM EDTA, 2 mM EGTA, and 1 mM phenylmethylsulfonyl fluoride. Lysates were then processed for immunoprecipitation as described previously (41) using specific antibodies. For immunoprecipitation of FLAG-tagged proteins, anti-FLAG M2-agarose affinity gel (Sigma) was used. Immunocomplexes were separated on 7.5% polyacrylamide gels and electrophoretically transferred to nitrocellulose, which was blocked with 5% dry milk in PBS containing 0.1% Tween 20 and probed with appropriate primary and secondary antibodies (Kirkegaard and Perry Laboratories) diluted in 5% dry milk in PBS. Sequential detergent extraction was performed as described (34) except that the amount of buffer used at each step was adjusted so that the final volume of each pool was 400 l.
Immunofluorescence-Cells were plated on glass coverslips the day before transfection in 6-well tissue culture dishes. 24 h after transfection using FuGENE 6 reagent, cells were washed in PBS, fixed in methanol for 2 min at Ϫ20°C, and incubated with appropriate antibodies diluted in complete PBS. Cells were incubated with primary antibodies for 30 min at 37°C, washed, and incubated with secondary antibodies for 30 min at 37°C. The following antibodies were used at indicated dilutions: PC28 at 1:200, C2206 at 1:50, C19220 at 1:50, Oct-A-probe at 1:50, Alexa Fluor-conjugated goat anti-mouse or goat anti-rabbit secondary antibodies (Molecular Probes, Eugene, OR) were used at 1:300. Images were obtained on a Leitz DMR microscope using a Hamamatsu Orcal digital camera and Openlab imaging software (Improvision).
Yeast Two-hybrid Constructs and Assays-The plakoglobin construct lacking the N-terminal domain (Pg⌬N) in pACTII vector was described previously (p515) (42). The cDNA sequence encoding Pg⌬N was removed from p515 by digestion with NcoI and XhoI and subcloned into the same sites of pAS-CYH2 to create Pg⌬N in pAS-CYH2. Plasmid p525, which contains cDNA sequence of DPNTP in pBluescript (42), was digested with BamHI and SalI, and DPNTP was subcloned into the BamHI and XhoI sites of pACTII vector (p900). The PKP1a head domain in DNA binding domain vector pBD-GAL4 (Stratagene) was described previously (19). The intracellular domains of Dsg1, Dsg2, Dsg3, Dsc1a, and Dsc2a in pGAD424 vector were described previously (18). These inserts were removed from the pGAD424 vector by digestion with EcoRI and SalI and cloned into the EcoRI and XhoI sites of pGADT7 vectors. The mouse E-cadherin cytoplasmic domain comprises amino acids 741-884 and was generated by PCR with primers containing EcoRI and SalI sites and cloned into the EcoRI and XhoI sites of pGADT7. Full-length human ␤-catenin was cloned into the SmaI and BamHI sites of pGAD424. Both the human p120 catenin isoform 3AC and the cytoplasmic domain of protocadherin ␤ 15 were cloned into the EcoRI and SalI sites of pGBKT7. Full-length PKP2a, its head domain, and the headless portion of PKP2a were amplified by PCR and cloned into the SmaI site of pAS-CYH2 vector and confirmed by DNA sequencing.
To assay interactions between proteins, yeast strain Y189 or AH109 was co-transformed with two testing plasmids, and dual transformants were selected by growth on plates lacking both tryptophan and leucine. Transformations were performed according to methods in the Matchmaker TM two-hybrid product protocol (CLONTECH Laboratories Inc.). Briefly, competent yeast cells were transformed with 1-5 g of each plasmid DNA using a standard lithium acetate/polyethylene glycol method. For the Y189 strain, positive interactions were quantified by measuring the ␤-galactosidase activity using 4-mythylumbelliferyl ␤-Dgalactopyranoside as a substrate following the methods described before (43). For AH109 strain, dual transformants were tested for their ability to grow on SD -Leu-Trp-His-Ade plates and their ability to turn blue in the presence of X-␣-gal. Materials for base media and agar were purchased from Difco, and materials for the defined media were purchased from CLONTECH Laboratories Inc.
TOPFLASH Luciferase Assay-TOPFLASH and FOPFLASH luciferase reporter gene constructs were generously provided by Dr. Bert Vogelstein. TOPFLASH reporter gene construct contains optimized TCF-binding sites upstream of a luciferase reporter gene, whereas the FOPFLASH contains mutated sites that do not bind TCF (44). SW480 cells were seeded at 2.75 ϫ 10 5 cells/well in 6-well dishes 24 h before transfection. Each well was transfected with 0.25 g of either TOP-FLASH or FOPFLASH, 0.05 g of pRL-TK (Promega) as an internal control for transfection efficiency, and 1 g of each indicated test construct. The total amount of DNA in each transfection was kept constant by the addition of an empty expression vector plasmid (pCMV-FLAG). Transfections were performed in triplicate using FuGENE 6 reagent (Roche Applied Science), and luciferase assays were performed 24 h later following the protocol provided for the Dual-Luciferase™ reporter assay system (Promega). Luciferase activity was corrected for transfection efficiency by using the control pRL-TK activity. Each experiment was performed at least three times independently.

The Head Domain of PKP2a Is Sufficient for Its Cell Border
Localization-Because PKP1a associates with desmosomal components and is targeted to desmosomes through its head domain (18), we tested whether the PKP2 head domain is similarly responsible for its localization to cell borders. PKP2a, PKP2a-H, or PKP2a-A ( Fig. 1) were transiently transfected into the SCC9 keratinocytes, and their subcellular distribution was examined with respect to the endogenous desmosomal marker DP. Ectopic full-length PKP2a was distributed primarily in a continuous fashion along cell borders (Fig. 2a), which overlapped, but was not completely coincident, with the punctate discontinuous pattern exhibited by endogenous DP (Fig.  2b). This distribution pattern of PKP2a suggests that overexpressed PKP2a might be associated with non-desmosomal components at the plasma membrane.
The head domain of PKP2a (PKP2a-H) colocalized more extensively with endogenous DP and appeared more punctate compared with full-length PKP2a (Fig. 2c). In addition, the PKP2a head domain was detected in the cytoplasm as discrete dots. In contrast to full-length ectopic PKP2a, which was rarely seen concentrated in the nucleus, in about 50% of the transfected cells the PKP2a head domain was detected in the nucleus. The PKP2a head domain was also localized to both cell borders and the nucleus in Madin-Darby canine kidney, HeLa, and COS cells (data not shown), reminiscent of the intracellular localization reported for the PKP1a head domain (17,18). In contrast, PKP2a-A exhibited a largely diffuse distribution throughout the cytoplasm without any specific pattern or cell border localization. Cells overexpressing PKP2a-A were also frequently found to form many filopodia-like membrane protrusions and to have altered morphology. These results suggest that the PKP2a head domain, which also has the ability to accumulate in the nucleus, is sufficient for the membrane targeting of PKP2a.
PKP2a Co-immunoprecipitates with DP and the N-terminal 584 Amino Acids Polypeptide DPNTP-It was shown previously that PKP1a interacts directly with the N-terminal domain of DP and enhances the recruitment of ectopic and endogenous DP to cell-cell borders (18,19). To compare the ability of PKP2a and PKP1a to interact with DP, full-length DP or DPNTP was transiently expressed in HEK293 cells with or without one of the PKP constructs, followed by immunoprecipitation using M2-agarose to precipitate FLAG-tagged PKP1a or PKP2a and the associated proteins. As shown in Fig. 3, like PKP1a, PKP2a co-immunoprecipitated DP and DPNTP. Neither DP nor DPNTP was nonspecifically detected in the immunocomplexes with M2-agarose when expressed alone. Although similar amounts of PKP1a and PKP2a were immunoprecipitated by M2-agarose, substantially more full-length DP or DPNTP was present in the immunocomplex with PKP1a, suggesting that DP may interact more robustly with PKP1a than with PKP2a. This possible difference in the strength of the interactions between DP and two plakophilins was further supported by yeast two-hybrid results (Fig. 6).
PKP2a Co-immunoprecipitates Pg and the Arm-repeat Domain of Pg Is Sufficient for This Association-In a previous study, it was reported that PKP1a was unable to interact with Pg in an in vitro overlay assay (45). Here, the ability of PKP1 and PKP2 to interact with Pg was compared by co-immunoprecipitation from transiently transfected COS cells. In agreement with previous findings, we detected little or no Pg association with PKP1a (Fig.  4A). In contrast, when Pg was co-expressed with FLAG-tagged PKP2a in COS cells followed by immunoprecipitation using M2agarose, Pg was efficiently co-precipitated together with PKP2a (Fig. 4B). Deletion of either the N-terminal head or C-terminal tail of Pg did not adversely affect its association with PKP2a, indicating that the Pg arm-repeat domain primarily mediates the interaction. A similar amount of PKP2a co-immunoprecipitated considerably more Pg⌬C than either Pg⌬N or full-length Pg. Next, the ability of the PKP2a head domain (PKP2a-H) and PKP2a arm-repeats with the C-terminal tail (PKP2a-A) to coimmunoprecipitate Pg was tested (Fig. 4C). PKP2a-H co-  (a and b), the deletion constructs PKP2a-H (c and d), or PKP2a-A (e and f). 24 h after transfection, cells were processed for double-label immunofluorescence using the polyclonal antibody Oct-A-probe to detect the FLAG epitope of ectopic polypeptides (a, c, and e) and the monoclonal antibody PC28 to detect endogenous DP (b, d, and f). Both PKP2a and its head domain are distributed in the cytoplasm and along cell borders, where they partially colocalize with endogenous DP (a-d). PKP2a-A is diffuse throughout the cytoplasm without specific border localization (e). PKP2a head domain is often found to concentrate in the nucleus (c). Scale bar, 10 m. precipitated less Pg compared with full-length PKP2a, and PKP2a-A was even less efficient in doing so, considering the much higher amount of PKP2a-A present in the cell lysates and the immunocomplex.
PKP2a Co-immunoprecipitates Dsg1 and Dsg2 but Not Dsg3 When Co-expressed in COS Cells, and the HR2 Domain of PKP2a Is Not Necessary for Its Complex Formation with Other Desmosomal Components-To investigate possible interactions between PKP2a and the three desmoglein isoforms Dsg1, -2, and -3, co-immunoprecipitation experiments were performed using lysates from transiently transfected COS cells. PKP2a specifically co-immunoprecipitated both Dsg1 and Dsg2 but not Dsg3 (Fig. 5A). The PKP2a head domain, but not its arm-repeat domain, mediated the interaction between PKP2a and Dsg2 (data not shown). The parallel experiment could not be successfully performed with Dsg1 due to quenched expression of the Dsg1 construct when co-transfected with PKP2a-H. Because COS cells express endogenous Pg, which might be able to interact with both PKP2a and Dsg1 or Dsg2 in the same complex, we attempted to determine if the association of Dsg2 with PKP2a was mediated by a limiting amount of free endogenous Pg. If this were the case, it should be possible to elevate the amount of Dsg2 associated with PKP2a by co-expression of ectopic Pg. However, the presence of ectopic Pg did not influence the level of Dsg2 co-precipitated with PKP2a (Fig. 5B).
The conserved HR2 domains in the head domains of PKPs have been postulated to be important for certain common functions (3, 4). However, deletion of HR2 from PKP2a did not affect its ability to co-immunoprecipitate other desmosomal components (Fig. 5C). In addition, both PKP2a ⌬HR2 and PKP2a-H ⌬HR2 exhibited the same distribution pattern as their respective wild-type counterparts PKP2a and PKP2a-H when expressed in SCC9 cells (data not shown). These results suggest that the HR2 domain of PKP2a is not critical for either its association with other desmosomal components or its proper localization in cells.
Association of Ectopic PKP2a with Endogenous Desmosomal Components-To determine whether PKP2a exhibits the same repertoire of interactions with endogenous desmosomal proteins, we analyzed endogenous desmosomal proteins that coprecipitated with FLAG-tagged PKP2a in SCC9 epithelial cells. When transiently expressed FLAG-tagged PKP2a was immu-FIG. 3. PKP2a co-immunoprecipitates both full-length DP and its N-terminal polypeptide DPNTP. DP or DPNTP was transiently expressed in HEK293 cells either alone or together with FLAG-tagged PKPs, which were then immunoprecipitated by M2-agarose, which binds the FLAG epitope of ectopic PKPs. The cell lysates and immunocomplexes were analyzed by Western blot to detect DP and DPNTP using the anti-DP antibody NW161 and to detect ectopic PKPs using the anti-FLAG antibody Oct-A-probe. Both PKP1a and PKP2a specifically co-precipitate DP and DPNTP, but more DP and DPNTP are found to be co-immunoprecipitated by PKP1a than by PKP2a. The variable levels of DP and DPNTP observed in cell lysates are likely due to promoter competition that results from transfection of multiple plasmids.

FIG. 4. PKP2a co-immunoprecipitates Pg and its terminal deletion mutants.
A, PKP1a is unable to co-immunoprecipitate Pg efficiently. Pg was expressed either alone or together with FLAG-tagged PKP1a followed by immunoprecipitation with M2-agarose and Western blot analysis with 1407 and Oct-A-probe. B, terminal domains of Pg are not essential for its association with PKP2a. Pg and its deletion mutants were transiently expressed in COS cells either alone or together with FLAG-tagged PKP2a followed by immunoprecipitation with M2agarose and Western blot analysis. An anti-Pg polyclonal antibody 1407 was used to detect Pg, and anti-FLAG antibody Oct-A-probe was used to detect FLAG-tagged PKP2a. PKP2a co-immunoprecipitates Pg and its deletion mutants lacking either the N-terminal or C-terminal domain. C, deletion of the head or arm-repeat domain of PKP2a abrogates its association with Pg. Pg was transfected either alone or together with different FLAG-tagged PKP2a constructs followed by immunoprecipitation with M2-agarose and Western blot analysis. The ability of PKP2a to co-precipitate Pg is severely compromised by the deletion of PKP2a arm-repeat domain plus C-terminal tail and almost completely abolished by the deletion of PKP2a head domain. noprecipitated from the 1% Triton X-100-soluble cell lysates of SCC9 cells, endogenous Pg was consistently detected in the immunocomplex (Fig. 5D). DP, which is mostly insoluble, was variably detected in the complex with PKP2a (not shown), suggesting a weak and/or transient association. Surprisingly, whereas Dsg3 was present at low levels in the immunocomplex with ectopic PKP2a, we could not detect endogenous Dsg2 co-precipitated by ectopic PKP2a from SCC9 cells.
Yeast Two-hybrid Analysis of Protein Interactions between PKP2a and Other Desmosome Components-To extend the results obtained using co-immunoprecipitation assays, the ability of PKP2a to interact directly with DPNTP, Pg⌬N, and the desmosomal cadherin tails was tested using the yeast twohybrid system. The relative strength of interactions between DPNTP and PKP1a or PKP2a was compared using yeast strain Y187 because it is more suitable for quantitative ␤-galactosidase assays. Consistent with previous reports (19), DPNTP interacted strongly with the head domain of PKP1a (Fig. 6). DPNTP directly interacted with full-length PKP2a and its head domain at a much weaker level but not with the arm-repeat domain of PKP2a. PKP2a and its head domain, but not PKP2a-A, interacted directly with Pg⌬N at a similar level.
Yeast two-hybrid assays employing growth and blue/white selection demonstrated that PKP2a directly interacted with the cytoplasmic domains of Dsg1, Dsc1a, and Dsc2a but not with Dsg2 or Dsg3 (Table I). The PKP2a head domain interacted with both Dsg1 and Dsg2 directly. However, no interactions were detected between PKP2a-A and any of the cytoplasmic domains, nor was an interaction detected between PKP2a and E-cadherin. As a positive control, the E-cadherin cytoplas-mic domain interacted with p120 catenin in the same assay (data not shown).

PKP1a and PKP2a Exhibit Different Localization Patterns and Abilities to Enhance the Border Recruitment of Excess DP
When Overexpressed in COS Cells-The protein interaction studies described above suggest that PKP1a and PKP2a have overlapping but distinct repertoires of desmosomal binding partners and that they may contribute differently to desmosome assembly. It has been suggested that a critical function of PKP1 is its ability to enhance DP recruitment to desmosomes (19). Therefore we compared the abilities of PKP1a and PKP2a to recruit DP to cell-cell borders in transiently transfected COS cells. As reported previously (17), PKP1a localized to both the nucleus and cell borders in COS cells when it was expressed alone (Fig. 7A, aЈ), and its expression resulted in increased border staining for endogenous DP compared with adjacent, non-transfected cells (Fig. 7A, a, arrow). Staining for endogenous DP in the population of non-transfected cells is weak to undetectable (Fig. 7A, a and b, arrowhead). In contrast to the intense nuclear accumulation of PKP1a observed in more than 95% of transfected COS cells, less than 10% of cells transfected with PKP2a showed weak nuclear staining. Like PKP1a, PKP2a enhanced the recruitment of endogenous DP to cell borders (Fig. 7A, b and bЈ).
Interestingly, the difference between PKP1a and PKP2a in their ability to enhance DP recruitment was revealed when both PKP and DP were co-transfected into COS cells. When expressed alone, ectopic DP predominantly colocalized with the IF network in the cytoplasm and sometimes localized to cell borders in transfected COS cells as described before (Fig. 7B, a,   FIG. 5. PKP2a co-

immunoprecipitates Dsg1, Dsg2 but not Dsg3 from co-transfected COS cells, and endogenous Pg and Dsg3 from singly transfected SCC9 cells.
A, each of the three Dsg isoforms was transiently expressed in COS cells either alone or together with FLAG-tagged PKP2a followed by immunoprecipitation with M2-agarose and Western blot analysis of the immunocomplex and cell lysates. Dsg1 was detected with the poly-Myc antibody, Dsg2 was detected with the anti-Dsg2 antibody 6D8, and the anti-Dsg3 antibody 1880 was used to detect Dsg3. B, co-expression of Pg does not enhance the association between PKP2a and Dsg2. Dsg2, Pg, and FLAGtagged PKP2a were transiently expressed in COS cells in combinations as indicated followed by immunoprecipitation with M2-agarose and Western blot analysis. The poly-Myc antibody was used to detect the Myc-tagged ectopic Pg. C, the HR2 domain of PKP2a is not required for its association with other desmosomal components. Different desmosomal molecules (DM) were transiently expressed in HEK293 cells either alone or together with PKP2a or PKP2a⌬HR2, both of which are FLAG-tagged and were immunoprecipitated with M2-agarose. Deletion of the HR2 domain in PKP2a does not affect its co-immunoprecipitation with other desmosomal components. D, SCC9 cells were transiently transfected with either empty vector or FLAG-tagged PKP2a construct, and the Triton-soluble cell lysates were subjected to co-immunoprecipitation using M2-agarose to isolate ectopic PKP2a and its associated endogenous proteins. Pg, Dsg3 but not Dsg2 were detected in the complexes with ectopic PKP2a.
inset) (17). When PKP2a was co-expressed with DP, both PKP2a and DP colocalized extensively along intermediate filaments (Fig. 7B, a and aЈ) and could also be detected along cell borders in certain cases (Fig. 7B, b and bЈ). In comparison, co-expression of PKP1a and DP resulted in a dramatic recruitment of DP to cell-cell borders together with PKP1a and a concomitant disappearance of DP from IF networks in co-transfected cells (Fig. 7B, c and cЈ). Co-expression with DP did not affect the accumulation of PKP1a in the nucleus (Fig. 7B, cЈ). These results showed that even though both PKPs can enhance the recruitment of endogenous DP to cell borders, PKP1a is much more efficient than PKP2a in promoting the border localization of excess DP.
PKP2a Co-immunoprecipitates the Adherens Junction Component ␤-Catenin, Which Forms Mutually Exclusive Complexes with either E-cadherin or PKP2a-In contrast to plakoglobin, which can be found in both desmosomes and adherens junctions, ␤-catenin is found exclusively in adherens junctions through its direct interaction with classical cadherins such as E-cadherin. Because PKP2a interacts with the arm-repeat do-main of plakoglobin, which is highly homologous to that of ␤-catenin, we tested whether ␤-catenin also binds to PKP2a. In co-immunoprecipitation experiments performed from transiently co-transfected COS cells, ␤-catenin specifically co-precipitated with both full-length PKP2a and its head domain but only very weakly with PKP2a-A, suggesting that the head domain of PKP2a is mainly responsible for its interaction with ␤-catenin (Fig. 8A). Full-length PKP2a did not co-precipitate two other adherens junction components ␣-catenin and E-cadherin in co-transfected COS cells, showing the specificity of its association with ␤-catenin in this assay (data not shown). Fulllength PKP2a interacted with ␤-catenin in yeast two-hybrid assays ( Table I), suggesting that they directly bind to each other. The negative control interaction between ␤-catenin and protocadherin ␤ 15 was negative (data not shown).
To determine if ␤-catenin forms distinct complexes with PKP2a and E-cadherin, COS cells were co-transfected with constructs expressing E-cadherin, ␤-catenin, and PKP2a. After immunoprecipitation of either ␤-catenin or PKP2a, the immunocomplex was analyzed for the presence of E-cadherin. As shown in Fig. 8B, E-cadherin efficiently associated with ␤-catenin when ␤-catenin was immunoprecipitated with an anti-␤catenin antibody; however, E-cadherin was not detected in the complex with PKP2a and ␤-catenin when PKP2a was immunoprecipitated by M2-agarose beads. This observation suggests that the pool of ␤-catenin associated with PKP2a is unable to interact with E-cadherin and that ␤-catenin forms mutually exclusive complexes with E-cadherin and PKP2a.
To examine the complex formation between PKP2a and other adherens junction components in a more physiological environment in keratinocytes, SCC9 cells transfected with only FLAGtagged PKP2a were used for M2-agarose immunoprecipitation experiments. Both endogenous E-cadherin and ␤-catenin, but not ␣-catenin, co-immunoprecipitated with PKP2a (Fig. 8C). The ability of ectopic PKP2a to associate with endogenous ␤-catenin is consistent with the protein interaction data. The presence of endogenous E-cadherin in a complex with ectopic PKP2a raises the possibility that in certain cell types, PKP2a might associate indirectly with E-cadherin through other protein partners such as Pg. Indeed, when Pg was overexpressed in COS cells with ectopic E-cadherin and PKP2a, a complex containing all three proteins could be detected (not shown), which is in contrast to the co-immunoprecipitation experiment described above for ␤-catenin (Fig. 8B).
Because endogenous PKP2a and ␤-catenin are localized to different junctional complexes and do not usually colocalize with each other at cell borders (Ref. 2; data not shown), we sought to determine whether overexpressed PKP2a could colocalize with endogenous ␤-catenin at the membrane. Both PKP2a and its head domain colocalized with endogenous ␤-catenin at cell borders in transiently transfected SCC9 cells (Fig. 9). The expression of PKP2a or PKP2a-H did not alter the distribution of endogenous ␤-catenin or affect its accumulation along cell borders. These data are consistent with PKP2a having the ability to associate with ␤-catenin in cells.
The Effects of PKP2a on ␤-Catenin/TCF Signaling-Free, cytoplasmic ␤-catenin that is not bound to cadherins can participate in the Wnt-signaling pathway through its interaction with TCF/Lef1 transcription factors (28). PKP2a co-precipitates with a non-cadherin-bound pool of ␤-catenin, raising the possibility that PKP2a might be involved in regulating ␤-catenin/TCF signaling. To better understand the physiological consequence of the PKP2a/␤-catenin interaction, we investigated the effect of PKP2a on ␤-catenin/TCF-regulated transcription in human colon carcinoma line SW480 cells, which have been shown to have constitutive ␤-catenin/TCF transcription activ-FIG. 6. PKP2a interacts directly with DPNTP and Pg⌬N in yeast two-hybrid assays. Yeast strain Y187 cells were co-transformed with DPNTP or Pg⌬N in pACTII vector and various PKP constructs as indicated. Co-transformants were selected on plates lacking tryptophan and leucine (ϪTrp/ϪLeu), and individual clones were used for quantitative ␤-galactosidase assay. Each bar represents the average of triplicate measurements for an independent clone. Interaction between p53 and large T antigen (LgT) is shown as a positive control. The measurements for the interactions between PKP2a-A and either DPNTP or Pg ⌬N are negligible. DPNTP interacts most strongly with PKP1a head domain and at a much weaker level with both PKP2a and its head domain. Pg⌬N directly interacts with PKP2a and its head domain but not with PKP2a-A. No significant measurements were obtained from clones co-transformed with PKP constructs and pACTII empty vector (data not shown). MUG, 4-mythylumbelliferyl ␤-D-galactopyranoside.
ity (44,46). TOPFLASH luciferase assays were performed to measure the activation of ␤-catenin/TCF signaling. 24 h after transfection, TOPFLASH reporter was strongly activated by endogenous ␤-catenin/TCF signaling in SW480 cells compared with negligible FOPFLASH activity (0.012 Ϯ 0.002) (Fig. 10). Overexpression of PKP2a significantly elevated the activation of the TOPFLASH reporter by ϳ 33% compared with vector  and bЈ). Double-label immunofluorescence was carried out to detect the localization of ectopic PKPs using the anti-FLAG antibody Oct-A-probe (aЈ and bЈ) and the localization of endogenous DP using the anti-DP antibody PC28 (a and b). Both PKP1a and PKP2a are localized to cell borders, where they colocalize with and enhance the staining of endogenous DP. PKP1a is also found intensely in the nucleus (aЈ). The arrows (a and b) indicate enhanced border staining of endogenous DP. The arrowheads (a and b) point to cell borders between untransfected cells. B, COS cells were cotransfected with DP together with FLAG-tagged PKP2a (a, b, aЈ, and bЈ) or FLAG-tagged PKP1a (c and cЈ) followed by double-label immunofluorescence using antibodies against the FLAG epitope (aЈ, bЈ, and cЈ) and DP (a-c). Ectopic PKP2a colocalizes extensively with overexpressed DP along intermediate filaments and cell borders (a, b, aЈ,  and bЈ). Overexpression of DP alone in COS cells results in its colocalization mainly along intermediate filaments (a, inset). Ectopic PKP1a is localized predominantly to the nucleus and along cell borders, where DP is strongly recruited. The IF pattern of distribution of overexpressed DP is dramatically reduced when PKP1a is coexpressed (c and cЈ). Scale bars, 10 m.
FIG. 8. PKP2a associates with ␤-catenin that is not bound to E-cadherin. A, the head domain of PKP2a is required for its association with ␤-catenin (cat). ␤-Catenin was transiently transfected into COS cells either alone or with different FLAG-tagged PKP2a constructs as indicated, which were then immunoprecipitated with M2-agarose. Western blots of cell lysates and immunocomplexes were performed using the polyclonal anti-␤-catenin antibody C2206 and anti-FLAG antibody Oct-A-probe. PKP2a and its head domain efficiently co-immunoprecipitate ␤-catenin, whereas the headless fragment PKP2a-A almost completely loses this ability. B, PKP2a forms complexes with ␤-catenin that is not bound to E-cadherin. ␤-Catenin, E-cadherin, and FLAG-tagged PKP2a were co-expressed in COS cells. Specific antibodies against ␤-catenin (C19220) and the FLAG epitope (M2) were used to immunoprecipitate ␤-catenin and PKP2a, respectively. ␤-Catenin coimmunoprecipitates both PKP2a and E-cadherin, but E-cadherin is not detected in the immunocomplex associated with PKP2a, indicating distinct pools of ␤-catenin in association with either PKP2a or Ecadherin but not both of them. The upper two bands recognized by the anti-E-cadherin antibody represent the precursor and mature form, and the lower band is likely a breakdown product lacking part of the cytoplasmic ␤-catenin binding domain. C, SCC9 cells were transiently transfected with either empty vector or FLAG-tagged PKP2a construct and the Triton-soluble cell lysates were subjected to co-immunoprecipitation using M2-agarose to isolate ectopic PKP2a and its associated endogenous proteins. E-cadherin and ␤-catenin but not ␣-catenin were detected in the complexes with ectopic PKP2a.

TABLE I
Yeast two-hybrid analysis of protein-protein interactions AH109 yeast cells were co-transformed with PKP2a or its subdomains in pAS-CYH2 and with the cytoplasmic domains (cyto) of desmosomal cadherins in pGADT7, full-length ␤-catenin in pGAD424 (␤-cat), or the cytoplasmic domain of E-cadherin in pGADT7 (E-cad cyto). Positive interactions were determined through analyzing the abilities of co-transformants to grow and turn blue on selection plates lacking adenine, histidine, leucine and tryptophan and containing X-␣-gal. Three separate clones from each co-transformation were assessed, and the same results were obtained for all three clones in all the assays. Clones that could grow as blue colonies on selection plates were considered as positive and are designated with plus symbols, and clones that did not show growth on the selection plates were considered as negative and are designated with minus symbols. White colonies were not observed. ND, not determined.
control, and the FOPFLASH activity was not significantly affected (0.011 Ϯ 0.001). The level of ␤-catenin in the cell lysates was not affected by expression of PKP2a. Double-label immunofluorescence showed that overexpressed PKP2a was largely diffuse in the cytoplasm and excluded from the nucleus, and no changes in the subcellular localization of ␤-catenin were detected (data not shown). E-cadherin expression dramatically reduced the TOPFLASH activity as has been reported previously (47,48). This inhibition of ␤-catenin/TCF signaling by E-cadherin requires its ␤-catenin binding site and was shown to involve the binding and sequestering of ␤-catenin from the signaling pool by E-cadherin (48). Co-expression of E-cadherin with PKP2a in SW480 cells resulted in a similar level of inhibition of TOPFLASH activity as by E-cadherin alone. This inability of PKP2a to up-regulate TOPFLASH activity in the presence of E-cadherin suggests that PKP2a may exert its effect on ␤-catenin/TCF signaling through interaction with the same pool of signaling active ␤-catenin that E-cadherin sequesters.

DISCUSSION
PKPs represent a family of arm-repeat proteins present in both desmosomes and the nucleus. Commonly found in stratified and complex epithelia, PKPs exhibit cell type-and differentiation-dependent patterns of desmosomal localization, suggesting distinct roles in the regulation and function of desmosomes. Their constitutive nuclear localization independent of their presence in desmosomes points to possible novel functions in the nucleus. The identification of human patients bearing PKP1 gene mutations has provided compelling evidence for its critical involvement in organizing and stabilizing desmosomes and further catalyzed the study on its role in desmosomes. However, little is known about the function of its most widely expressed relative, PKP2. In the present study, we investigated the ability of PKP2a to interact with other junctional molecules and conducted functional analyses on its role in desmosome formation and regulation of ␤-catenin-signaling activity.
PKP1a has been shown to interact directly with DP, Dsg1, and keratin (18,45,49) through its head domain (18,19). We showed here by co-immunoprecipitation from co-transfected cells that PKP2a associates with DP, Pg, Dsg1, and Dsg2. Yeast two-hybrid analysis demonstrated furthermore that interactions with DP, Pg, Dsg1, and possibly Dsg2 are direct and further uncovered Dsc1a and 2a as PKP2a interaction part-ners. Pg co-precipitated with PKP2a and the deletion of either the N-terminal head or C-terminal tail of Pg failed to abrogate this interaction, suggesting that it is primarily mediated by the arm-repeat domain of Pg. Intriguingly, the association between Pg and PKP2a was further enhanced by deletion of the Pg C-terminal end domain. Previous work showed that Pg⌬C ex- FIG. 10. Effect of PKP2a on ␤-catenin/TCF signaling in SW480 cells. SW480 cells were transiently transfected with TOPFLASH luciferase reporter gene and a control luciferase reporter gene pRL-TK to control for transfection efficiency. The results are expressed as relative activity, which is derived from dividing the TOPFLASH luciferase activity by the activity of pRL-TK. The results are the means and S.D. of triplicate values obtained in one representative experiment of three. Negligible activity was observed using FOPFLASH reporter gene, which contains mutated TCF-binding sites (data not shown). PKP2a expression significantly elevated the TOPFLASH activity by ϳ33% (p ϭ 0.011), whereas E-cadherin expression inhibited its activity and prevented its up-regulation by PKP2a. Mean relative activities were not significantly different between cells expressing E-cadherin and cells expressing both E-cadherin and PKP2a (p ϭ 0.405 Ͼ 0.05). Cell lysates prepared for the reporter gene assay were quantitated, and the same amount of total protein was analyzed by SDS-PAGE and Western blot using the anti-␤-catenin (cat) antibody C2206. The total level of ␤-catenin is not altered by the expression of either PKP2a or E-cadherin. The asterisk designates values significantly different from control by Student t test at p Ͻ 0.05.
FIG. 9. PKP2a and its head domain colocalize with ␤-catenin (cat) at the cell membrane. SCC9 cells were transiently transfected with FLAG-tagged PKP2a (a, b, and c) or its head domain PKP2a-H (d, e, and f), and double-label immunofluorescence was carried out to detect endogenous ␤-catenin (b and e) and FLAG-tagged ectopic proteins (a and d).
Composite pictures are shown to demonstrate the colocalization of ␤-catenin with PKP2a and PKP2a-H along the cell borders, as indicated in yellow (c and f). Some of the overlapped signals do not appear as yellow after the overlay because the intensity of the individual signals is unequal. pression in A431 cells caused formation of longer desmosomes or groups of tandemly linked desmosomes that were still attached to keratin intermediate filaments (34). It was proposed that deletion of the Pg C terminus might enhance interactions with other desmosomal components and lead to generation of longer desmosomes. Together with our data, it thus seems possible that enhanced interaction between Pg⌬C and PKP2a could contribute to the effect on desmosome morphology previously attributed to Pg⌬C expression.
PKP2a co-immunoprecipitated Dsg1 and Dsg2, but not Dsg3 in co-transfected COS cells, and its interaction with Dsg1 was confirmed by yeast two-hybrid assays as being directly mediated by the PKP2a head domain. It is unclear why we detected a direct interaction between the Dsg2 cytoplasmic domain and the PKP2a head domain but not full-length PKP2a in yeast two-hybrid assays, but it seems likely that a direct binding site for Dsg2 is contained in the head domain of PKP2a in light of the abilities of both PKP2a and PKP2a-H to co-precipitate Dsg2 in transfected cells. The fact that increasing the level of cellular Pg does not enhance the association between PKP2a and Dsg2 ( Fig. 5B) is consistent with the hypothesis that Pg does not interact with PKP2a and Dsg2 simultaneously. The binding sites in Pg for Dsg1 and Dsc1a reside in its arm-repeat domain (40,50,51), as does the binding site for PKP2a as we show here. Thus, it is possible that desmosomal cadherins and PKP2a compete with each other for binding to Pg. How the observed direct interactions discussed here translate into the formation of the physiological complexes in cells may depend on other factors, including the availability of specific protein partners in specific cell types. For instance, although direct interactions with Dsg3 were not observed in co-transfection or yeast twohybrid assays, Dsg3 was present in an endogenous complex with ectopic PKP2, whereas Dsg2 was not (Fig. 5D). Together these data suggest that Pg or other unknown proteins may mediate an indirect association between PKP2a and Dsg3 but not Dsg2. The absence of an identifiable complex between ectopic PKP2a and endogenous Dsg2 by co-immunoprecipitation may also indicate that in SCC9 cells, which express both Dsg3 and Dsg2, Dsg3 may compete more effectively for PKP2 due to its relative abundance compared with Dsg2 in the detergentsoluble pool or other factors favoring this interaction.
The presence of PKP2 in desmosomes is widespread as it is found in all simple epithelia and in complex epithelia where it is localized throughout the layers and concentrated in desmosomes of the basal layer. These localization data are consistent with our observations, and together they suggest that PKP2a can associate with all Dsg isoforms, either directly or indirectly. Thus PKP2 may be involved in desmosome function on a much broader scale than PKP1, and it is intriguing to consider that it may serve different functions when it is associated with different desmosomal cadherins.
In COS cells, ectopic PKP1a efficiently enhances the recruitment of DPNTP to cell borders, and this ability is conferred by the PKP1a head domain (19). Because DPNTP contains only the first 584 amino acids of DP and does not bind to IF, we compared the ability of PKP1a and PKP2a to enhance the cell border recruitment of full-length DP that is sequestered along the IF network. Co-expression of PKP1a with DP in COS cells resulted in a pattern of DP distribution vastly different from that when DP was expressed alone. Instead of exhibiting extensive co-alignment with the IF network and occasional border localization, DP was mostly concentrated along the borders, colocalizing with PKP1a. In contrast, the majority of DP remained sequestered along IF in cells co-expressing PKP2a and DP. The fact that both PKPs enhanced the border localization of endogenous DP to a similar level (Fig. 7A, a and b) could be due to a limiting amount of endogenous DP available for recruitment and that to reveal the higher capacity of PKP1a to recruit DP requires that excess DP be expressed.
The different abilities of PKP1 and PKP2 to enhance cell border recruitment of ectopic DP in COS cells might be explained by several mechanisms. Biochemical evidence suggested that DP interacts with PKP2a less robustly than with PKP1a, and an interaction with a certain strength could be required to directly recruit DP to cell borders by PKPs. Secondly, the two PKPs might enhance DP recruitment indirectly through their differential ability to recruit other desmosomal components or as yet unidentified proteins. Furthermore, PKP2a could differ from PKP1a in its propensity to incorporate into desmosomes, and PKP1a may preferentially assemble into desmosomes in COS cells. Because PKP2 has been found mostly concentrated in desmosomes in the basal layer of various stratified epithelial tissues (2), perhaps the participation of PKP2 in desmosomes of the basal layer contributes to the formation of smaller (52) and possibly more dynamic desmosomes that are suitable for the amplification and generation of new cells from the basal layer. This may partly be achieved through the reduced capacity of PKP2 to recruit and stabilize other desmosomal components such as DP.
Another striking difference between PKP1a and PKP2a is their ability to accumulate in the nucleus in cells overexpressing these proteins. PKP1a and its head domain efficiently concentrated in the nuclei when expressed in a variety of cell lines (17)(18)(19). Here we show that PKP2a is mostly excluded from the nucleus in transiently overexpressing cells. The PKP2a head domain can localize to the nucleus more efficiently than fulllength PKP2a, indicating that the arm-repeat domain of PKP2a either contains nuclear export signals or has the ability to retain PKP2a in the cytoplasm through interactions with other proteins.
Besides interacting with other desmosome plaque proteins, PKP2a is also capable of associating with ␤-catenin mainly through the PKP2a head domain. Co-expression of E-cadherin, ␤-catenin, and PKP2a led to the formation of mutually exclusive complexes that contained ␤-catenin with either E-cadherin or PKP2a, suggesting that PKP2a is unlikely to interact with ␤-catenin in the adherens junction. However, ectopic PKP2a formed complexes with endogenous E-cadherin and ␤-catenin but not with ␣-catenin in SCC9 cells. This result suggests that other proteins such as Pg could mediate the association between E-cadherin and PKP2a. Because this complex lacks ␣-catenin, it is unlikely to be found in mature adherens junctions, but it perhaps represents a junction intermediate involved in sorting and segregation of components during the early stages of junction formation or during the dynamic regulation of junctional structures. Even though we were unable to detect colocalization of endogenous PKP2a and ␤-catenin in SCC9 cells, we did observe extensive colocalization of ectopic PKP2a and its head domain with endogenous ␤-catenin along the cell borders. It has been shown previously that there is a pool of ␤-catenin localized in the lateral membrane of Madin-Darby canine kidney cells that does not contain E-cadherin (53), so possibly ectopic PKP2a could interact with this pool of ␤-catenin at the membrane but not the E-cadherin bound pool of ␤-catenin.
Despite a lack of apparent effect on the total level of ␤-catenin, PKP2a overexpression resulted in a modest, but reproducible increase in the level of endogenous ␤-catenin/TCF signaling in SW480 cells. Because we could only achieve about 15% transfection efficiency in SW480 cells, some moderate effect on ␤-catenin protein level by expression of PKP2a may not have been detectable in the total cell population. PKP2a expression did not seem to change the subcellular localization of ␤-catenin in transfected cells, and experiments using COS cells and HEK293 cells, which could be transfected at efficiencies greater than 50%, did not reveal any significant effect by PKP2a expression on either the total level of ␤-catenin or its partitioning into different detergent soluble pools (data not shown). These results suggest that PKP2a modulates ␤-catenin signaling through mechanisms other than regulating its stability, as have been suggested for several regulators of ␤-catenin/TCF signaling (54 -56). Expression of ectopic E-cadherin inhibited endogenous ␤-catenin/TCF signaling and suppressed PKP2adependent increases in ␤-catenin/TCF signaling in SW480 cells. A recent paper presented evidence for the existence of a small pool of transcriptionally active ␤-catenin among the much larger pool of cytosolic but transcriptionally inactive ␤-catenin in SW480 cells (47). Our result is consistent with the model in which E-cadherin sequesters the transcriptionally active pool of ␤-catenin and prevents its binding to PKP2a.
The data presented in this study show that PKP2 and PKP1 exhibit overlapping but distinct properties in their repertoires of binding partners and functional relationships with DP. These results, coupled with the observed partially overlapping yet different expression profiles, suggest specific participation of individual or combinations of PKPs in creating functionally distinguishable desmosomes that are tailored for the needs of different tissues or differentiation stages. The ability to interact with ␤-catenin and modulate its signaling activity indicates that PKP2a might be involved in other cellular processes such as the cross-talk between desmosomes and adherens junctions and the transduction of signals from the cell surface to the nucleus.