Lack of plakophilin 1 increases keratinocyte migration and reduces desmosome stability

Desmosomes (DMs) are specialised cadherin-mediated adhesion complexes that impart stability and rigidity to tissues under mechanical stress(Garrod et al., 1999). These structures are dynamic and continual renewal of stratifying epidermis involves both rapid synthesis and degradation of cellular junctions. Such characteristics are important in epidermal homeostasis and wound healing(Moll et al., 1999; Wallis et al., 2000). Desmosomal adhesion may also be compromised in invading and metastasising transitional cell carcinomas, leading to speculation regarding possible tumour suppressor function of desmosomal attachment(Tselepis et al., 1998). DMs are formed later than adherens junctions during embryogenesis(Fleming et al., 1991) but their biological importance is underlined by the embryonic lethal mouse knockouts of the desmosomal components plakoglobin(Bierkamp et al., 1996; Ruiz et al., 1996) and desmoplakin (Dp) (Gallicano et al.,1998). Recently evidence has emerged demonstrating the importance of DMs in epithelial morphogenesis and cell positioning(Runswick et al., 2001; Vasioukhin et al., 2001).

Desmosomal cell-cell attachment is mediated through two types of transmembrane glycoproteins of the cadherin superfamily, desmoglein and desmocollin (Collins et al.,1991; Schafer et al.,1994). There are currently three known isoforms of both these molecules and each display differentiation-specific and tissue-specific expression patterns (Koch et al.,1992). Indirect attachment of desmosomal cadherins to intermediate filaments (IFs) is facilitated by plaque and armadillo proteins including Dp,plakoglobin and plakophilins (for review, see Green and Gaudry, 2001). Dp and plakoglobin are ubiquitous components of DMs whereas the plakophilins have multiple members displaying tissue-specific expression patterns that vary with epithelial differentiation (Hatzfeld et al., 1994; Heid et al.,1994; Mertens et al.,1996; Schmidt et al.,1997; Schmidt et al.,1999).

Understanding the precise role of DMs in cell biology has been difficult owing to the heterogeneity of their molecular composition. However,considerable insight has been gained from human and mouse model mutants. Specifically, naturally occurring monogenic disorders involving dysfunction or absence of desmosomal components have resulted in loss of cell-cell adhesion and striking clinical phenotypes (reviewed by Green and Gowdry, 2000; Norgett et al., 2000). These observations are supported by knockout and transgenic mice studies (reviewed by Green and Gaudry, 2000; Chidgey et al., 2001; Vasioukhin et al., 2001).

The first monogenic human disorder involving a desmosomal component to be described was the skin fragility-ectodermal dysplasia syndrome (OMIM 604536)characterised by complete ablation of plakophilin 1 (PKP1)(McGrath et al., 1997). Although PKP1 had been identified previously as a desmosomal component(Kapprell et al., 1988; Hatzfeld et al., 1994; Heid et al., 1994), it was not until the correlation of its absence with a human skin fragility disease that the importance of plakophilins in desmosomal adhesion was recognised. To date,three patients displaying complete ablation of PKP1(McGrath et al., 1997; McGrath et al., 1999; Whittock et al., 2000) have been reported. These naturally occurring mutations provide valuable resources for the investigation of human PKP1 function.

PKP1 is a major component of DMs in stratifying and complex epithelia but also is expressed widely in the nuclei of cells devoid of DMs(Schmidt et al., 1997),leading to speculation of an, as yet, unidentified nuclear signalling function. Other plakophilins, including plakophilin 2(Mertens et al., 1996) and plakophilin 3 (Schmidt et al.,1999), display dual localisation to DMs and nuclei but have differentiation-specific distributions.

Models of DM structure and macromolecular interactions have been derived from in vitro domain mapping, reconstitution studies (for review, see Green and Gaudry, 2001) and immunoelectron microscopy(North et al., 1999). Keratin IFs have been shown to bind to Dp(Stappenbeck and Green, 1992; Kouklis et al., 1994; Bornslaeger et al., 1996; Gallicano et al., 1998) as well as to PKP1 (Kapprell et al.,1988; Smith and Fuchs,1998; Hofmann et al.,2000). Dp has been shown to bind to plakoglobin (Steppenbeck et al., 1993; Smith and Fuchs,1998) and PKP1 (Kowalczyk et al., 1999; Hatzfeld et al.,2000; Hofmann et al.,2000), while plakoglobin binds to desmosomal cadherins(Kowalczyk et al., 1994; Troyanovsky et al., 1994). Plakophilin has also been found to bind desmosomal cadherins in vitro(Smith and Fuchs, 1998; Kowalczyk et al., 1999; Hatzfeld et al., 2000) but the significance of this interaction, as well as the interaction between PKP1 and keratin IFs in vivo, is unclear and a model of PKP1 providing lateral interactions for Dp has been suggested(Kowalczyk et al., 1999) which is supported by other immunoelectron microscopy data(North et al., 1999).

To analyse the role of PKP1 in keratinocyte cell biology we have characterised cell lines derived from the unique resource of PKP1-deficient keratinocytes. Specifically, we have retrovirally delivered recombinant PKP1 to cells derived from two separate PKP1-deficient patients and then compared these modified keratinocytes with appropriate vector controlled cells. We also have compared these cells with unaffected keratinocytes from two separate control individuals.

We demonstrate that PKP1 increases desmosomal components within cells but does not enhance expression of key desmosomal components at the transcriptional level. We also demonstrate that the level of the adherens junction proteins E-cadherin and β-catenin, and that of keratin 14 and keratins identified with a pan-keratin antibody, are not altered significantly. Levels of cellular plakophilin 2 and plakophilin 3 are not altered significantly relative to the other desmosomal proteins studied,indicating that in cultured PKP1-deficient cells there is no compensatory up-regulation of plakophilin 2 or plakophilin 3. Furthermore, PKP1 does not affect keratinocyte cell growth but does influence cell migration. We demonstrate that PKP1 has a role in the transition of DMs from a calcium-dependent state to a calcium-independent state and that PKP1 also affects the size and number of keratinocyte DMs.

After gaining appropriate consent, biopsy samples were obtained from thigh skin of one 3-month-old PKP1-deficient male patient (patient 1), buttock skin from one five-year-old PKP1-deficient male patient (patient 2), breast skin from one 31-year-old healthy female (control 1), and leg skin from one 45-year-old healthy female (control 2). Primary keratinocytes from these sources were isolated and grown in the presence of an irradiated 3T3 feeder layer (Rheinwald and Green,1975). Cells were immortalised with HPV16 (E6^E7) as described previously (Storey et al.,1988). All experiments were carried out between 15-25 passages after immortalisation. Cell lines with the prefix null indicate immortalised PKP1-deficient keratinocytes, while cell lines with the prefix norm indicate immortalised breast- or leg-derived keratinocytes derived from the controls.

Cells under normal Ca²⁺ conditions (approx. 1.2 mM) were cultured in DMEM:Ham's F12 medium (Invitrogen BV) at a ratio 3:1 and 10%foetal calf serum (FCS; ICN Biomedicals), supplemented with 1% penicillin and streptomycin (Sigma-Aldrich), 0.4 μg/ml hydrocortisone (Sigma-Aldrich),10^–10 M cholera toxin (Sigma-Aldrich) and 5 μg/ml insulin(Sigma-Aldrich) at 37°C in 10% CO₂. For low Ca²⁺experiments, cells were cultured in keratinocyte SFM (serum free; Invitrogen BV) at a Ca²⁺ concentration of 0.09 mM. For Ca²⁺ switch experiments (see below) controls were first performed using keratinocyte SFM supplemented with 1.2 mM CaCl₂. Cell numbers were determined using a CASY® 1 (Scharfe System GmbH) cell counter.

Full-length PKP1 was cloned into the pBabe puro retroviral vector(Morgenstern and Land, 1990)using standard molecular biology techniques. PKP1 constructs were fully sequenced to confirm retention of the wild-type open reading frame. The Phoenix amphotrophic retroviral packaging cell line, ΦNX-Ampho Cells(obtained from G.P. Nolan, Stanford, CA), was used to generate viral particles as described previously (Kinsella and Nolan, 1996). Cells were transduced with retroviral vectors at a low multiplicity of infection to prevent multiple integrations. Cells were selected using 2 μg/ml puromycin 48-72 hours post-transduction for a maximum of 14 days prior to expansion. Recombinant PKP1 expression in selected populations of transduced cells was verified to >99% using immunofluorescence microscopy.

Cells grown on glass coverslips were fixed using the following methods.

• Cells membranes were made permeable by treatment with 0.2% Triton X-100 in PBS for 5 minutes before fixation for 20 minutes in methanol at–20°C or 4% formaldehyde at room temperature, followed by three 5-minute washes in PBS at room temperature.

• Cells were fixed for 20 minutes at –20°C in 1:1 methanol:acetone followed by three 5-minute washes in PBS at room temperature.

PKP1 was detected with the monoclonal antibody PP1-5C2 (Progen) using method (i) (for nuclear and membrane localisation) or (ii) (for membrane only localisation) (Schmidt et al.,1997). Method (ii) was used with all other primary antibodies/antisera, which were as follows: 11-5F [Dps I and II(Parrish et al., 1987)], PG5.1(plakoglobin; Cymbus Biotechnology Ltd); Dsc-3-U114 (Desmocollin 3; Cymbus Biotechnology Ltd), AHP319 (desmoglein 3; Serotec). Secondary antibodies used were Alexa 488 goat antimouse (Molecular Probes, Inc.), Alexa 568 goat anti-rabbit (Molecular Probes, Inc.), FITC goat anti-mouse (DAKO) or FITC swine anti-rabbit (DAKO) IgG. Actin filaments were visualised by TRITC-coupled phalloidin (Molecular Probes, Inc.) after fixation using method (i). Counterstaining was performed using DAPI (Molecular Probes, Inc.) and microscopy was carried out using a Nikon Optiphot microscope (Nikon) with Kodak Microscopy Document System 290 (Kodak). Antibodies not used for immunfluorescence but for immunoblotting were as follows: sc-7298(β-actin; Autogen Bioclear), ab6276 (HSC-70; Abcam Ltd), AHP322(pan-desmocollin; Serotec), AHP321 (pan-desmoglein; Serotec), 33-3D[desmoglein 2 (Vilela et al.,1995)], PP2/62/86/150 multi-epitope cocktail (plakophilin 2;Progen), 310.9.1 (plakophilin 3; Progen), HECD-1 (E-cadherin), 6F9(β-catenin; Sigma-Aldrich), and LL001 [Keratin 14(Purkis et al., 1990)], LP34[reactive with Keratins 1, 5, 6 and 18(Lane and Alexander, 1990)]. Antibodies used for flow cytometry were as follows: FB12 (integrin α1;Chemicon International), P1E6 (integrin α2; Chemicon International),P1B5 (integrin α3; Chemicon International), P1D6 (integrin α5;Chemicon International), G0H3 (integrin α6; Serotec), Y9A2 (integrinα9; Chemicon International), L230 (integrin α_V;prepared from hybridoma cells obtained from ATCC), LM609 (integrinα _Vβ3; Chemicon International), P1F6 (integrinα _Vβ5; Chemicon International), 10D5 (integrinα _Vβ6; Chemicon International), 3E1 (integrin β4;Chemicon International), and IgG1 (mouse control; DAKO).

Total protein extracts were obtained from cells seeded at high density(5×10⁵ cells/35 mm Petri dish) and grown for 72 hours under normal Ca²⁺ conditions as previously described(Wan et al., 2003). Internal controls of heat-shock protein HSC-70 or β-actin were used to reprobe each blot to demonstrate equivalent protein loading. All blots were subject to densitometry analysis using the NIH Image 1.61 software Gel Plotting macro.

Cells grown in 35 mm plastic Petri dishes were fixed in 2% formaldehyde with 2.5% glutaraldehyde in 0.1 M Sorrenson's phosphate buffer pH 7.4 followed by post-fixation in aqueous 1.3% osmium tetroxide and en-bloc staining with 2%uranyl acetate in 50% ethanol. Cells were not detached from the plastic Petri dishes. The samples were then processed using standard techniques and embedded in TAAB 812 (medium hardness) epoxy resin(Eady, 1985). Ultrathin sections (60-90 nm) were stained with 2% uranyl acetate in 50% ethanol and Reynold's lead citrate before observation in a JEOL 100CX transmission electron microscope (JEOL). To measure DM density, electron micrographs were taken at 5,000× magnification and 6-8 arbitrary fields per sample were examined. The negatives were scanned into a personal computer using Adobe Photoshop image software. Only DMs displaying recognisable opposing plaques were scored. To estimate the DM density relative to cell cytoplasm we applied a coherent single square lattice over each EM negative to produce a total of 252 test points (at a distance of 0.5 cm, corresponding to 10 μm on the section). Estimation of DM density was achieved by counting the total number of DMs per negative relative to the total number of test points falling on keratinocyte cytoplasm only. From these counts the DM density was calculated and expressed as a ratio of DM per cytoplasm test points. To measure DM size,electron micrographs were taken at 13,000× magnification of several arbitrary fields that contained DMs in the ultrathin sections used for DM counting. The length of the extracellular space between two clear, opposing plaques was measured using the Photoshop measurement tool.

Total RNA was isolated from cultured cells grown for 72 hours post-confluency. RNA antisense probes were prepared as follows. Fragments of PKP1, plakoglobin, Dp, desmoglein 2, desmoglein and GAPDHwere PCR amplified from keratinocyte cDNA and cloned using the TOPO TA Cloning® Kit Dual Promoter (Invitrogen BV). RNA probes generated using[³²P]dUTP (NEN Life Science Products, Inc.) and T7 RNA polymerase(Promega Corporation) were then purified from unincorporated nucleotides and DNase treated to remove plasmid DNA using RNeasy® Mini Kit columns(Qiagen). Between 4 and 8 μg of total RNA was used for RNase protection assays using an RNase Protection Kit (Roche) according to the manufacturer's specifications. Each RNA probe representing a desmosomal gene was mixed with GAPDH probe as an internal control in each separate assay. Protected products were resolved using 5% TBE-urea ready gels and a mini-PROTEAN® 3 cell(Bio-Rad), then dried, exposed with a phosphorimager and analysed using ImageQuant™ software (Amersham Pharmacia Biotech).

Cells were seeded at equal density (7×10⁵/35 mm Petri dish) and grown for 48 hours. Mitomycin C (Sigma-Aldrich) was used to arrest mitosis before confluent cultures were washed with medium and wounded by scraping the blunt end of a 1 ml micro-Gilson pipette tip across the centre of the cell sheet, removing cells in a linear fashion to a width between 3 and 6 mm. The cultures were then washed three times and transferred to a specially constructed two-piece circular aluminium housing that had a glass lid and an epicentric hole in the base through which the cells were observed. The chamber was gassed with 10% CO₂ in humidified air and was placed onto the stage of an Olympus IMT-2 inverted phase contrast microscope (Olympus Microscopes), fitted with a Fujitsu TC2-336P charge coupled device camera (EOS Electronics AV Ltd) and surrounded by a Perspex temperature-regulated jacket that was adjusted to 37°C. Images were collected every 6 minutes to a Power Mac 7100 computer by use of Adobe Premiere software. After calibration using Optilab Pro 2.6.1 software the area of cells occupying each image at the given time points was calculated and expressed in μm moved/hour.

Cells were seeded at equal density (7×10⁵/35 mm Petri dish) and grown for 72 hours in 1.2 mM Ca²⁺ medium. Cells seeded at this density were 100% confluent after 16 hours. 1.2 mM Ca²⁺ medium was replaced with 0.09 mM Ca²⁺ medium after washing three times with 0.09 mM Ca²⁺ medium (low Ca²⁺ switch) and the cells were transferred to the time-lapse apparatus detailed above. When cells change from a flattened morphology towards a rounded one they become more refractile. We monitored changes in adhesiveness using image analysis. Images were recorded every 6 minutes. At given time points, 256 grey scale (0=black,255=white) images were collected and analysed. The average grey for image 1(time=0 minutes) of all samples was calculated and used as a background threshold. For each image analysed the number of pixels above this threshold was calculated as a percentage of total pixels. Control samples were grown in calcium-supplemented low Ca²⁺ medium (1.2 mM) for 72 hours and monitored in this fashion. In control experiments no difference was noted in the percentage pixels above the given grey threshold for each cell line tested(data not shown).

Integrin complement of the cell lines was assessed using flow cytometry. Confluent sheets of cells were harvested, washed and incubated with primary antibody for 40 minutes at 4°C. Secondary antibody was applied for 30 minutes at 4°C after washing 4 times. Cells were then washed twice with PBS and resuspended in 0.4 ml PBS. Labelled cells were scanned on a FACSCalibur cytometer (Becton Dickinson) and analysed using Cellquest software, acquiring 1×10⁵ events. Flow cytometry was performed in triplicate on two separate occasions. For each antibody the geometric mean was noted and the value of IgG1 control was subtracted to give a value for level of expression.

Cells were seeded at equal high density (7×10⁵/35 mm Petri dish) and grown for 72 hours in 1.2 mm Ca²⁺ before being either extracted, subject to 1 hour of a low Ca²⁺ switch and extracted, or subject to 24 hours of a low Ca²⁺ switch followed by extraction. Extractions were performed as described(Palka and Green, 1997). Equal volumes of each extracted pool were resolved and immunoblotted as described above.

Keratinocytes isolated from two PKP1-deficient patients and two healthy control individuals were immortalised as described in Materials and Methods. Cells were cultured after crisis for over 10 passages with no obvious gross differences in morphology or behaviour (differentiation, proliferation, cell turnover).

The pBabe-puro vector and Phoenix amphotrophic packaging cell line were used to introduce recombinant PKP1 into both normal and PKP1-deficient cell lines as described. A vector-only construct was also used in order to provide control populations. Viral particles were collected from transiently transfected packaging cells as described and target cells were selected for puromycin resistance, transmitted by the puromycin resistance gene within the vector. Viral titre was not calculated or optimised and infection rates were low, less than 5% as judged by the initial growth of puromycin-resistant cells, indicating that multiple integrations (more than two) were unlikely(Garlick et al., 1991). PKP1 was detected in >99% of selected null cells expressing recombinant PKP1 2 weeks post-infection, as judged by immunofluorescence(Fig. 1A-C).

Cells expressing pBabe-puro vector only are referred to with the suffix pB while cells expressing pBabe-puro containing and expressing PKP1 are referred to with the suffix PKP. Null indicates null cells and norm indicates normal cells. Where needed, 1 indicates cells derived from patient 1 or control 1 while 2 indicates cells derived from patient 2 or control 2 (see Materials and Methods). Therefore nullpB1 refers to vector control null cells derived from patient 1 while nullPKP2 refers to null cells derived from patient 2 expressing recombinant PKP1.

PKP1 was detected in nullPKP cells but not in nullpB or null only populations as determined by immunoblotting(Fig. 1G). PKP1 was detected in immortalised normal keratinocytes as well as in derived populations expressing vector controls. All experiments and observations of transduced cell populations were made using passages 5-10 after introduction of retroviruses.

Electron micrographs at 13,000× magnification from nullpB1,nullPKP1 and normpB1 cell lines fixed after 8 days growth were analysed blind and a representative DM field from each sample set (5 negatives for each line) is presented in Fig. 1D-F. No noticeable difference is seen in DM ultrastructure from these fields.

Using western blotting we determined whether de novo expression of PKP1 affected expression of other desmosomal proteins, adherens junctional proteins, or keratins. Fig. 2Ashows that, in comparison with the retroviral control lines (nullpB),re-expression of PKP1 in the nullPKP lines resulted in significant increases in desmoplakin, desmocollin and plakoglobin, but not E-cadherin. Increases were also observed in desmoglein 3, while levels of desmoglein 1,plakophilin 2 and plakophilin 3 were not significantly different when comparing nullpB with nullPKP western blots(Fig. 2A). No discernible difference was observed in the level of β-catenin expression between cell lines tested. Using antibodies against keratin 14 and a pan-keratin antibody recognising keratins 1, 5, 6 and 18, we could not identify any differences between nullpB, nullPKP and normpB cell lines(Fig. 2A).

In order to assess whether the different levels of protein identified in Fig. 2A were a result of increased mRNA expression, RNA probes for GAPDH, PKP1, Dp, plakoglobin,desmoglein 2 and desmoglein 3 were used in RNase protection assays of total RNA isolated from nullpB1, nullPKP1 and normpB1. No PKP1 mRNA was detected in nullpB1 cells but it was evident in normpB1 and nullPKP1 cells(Fig. 2B). No significant difference in Dp, desmoglein 3 or plakoglobin expression(Fig. 2B) was seen. A slight decrease was seen in desmoglein 2 mRNA levels between null and norm cells, but there were no observed differences between nullpB1 and nullPKP1 cells. Overall we demonstrated no differences in levels of mRNA expression of desmosomal components between cells lacking PKP1 (nullpB1) and matched cells expressing PKP1(nullPKP1).

PKP1-deficient skin demonstrates hyperkeratosis and thickening compared with normal skin (McGrath et al.,1997; McGrath et al.,1999). We investigated whether PKP1 expression influences in vitro cellular growth rates. Cells were seeded at equal density in identical 6-well plates, harvested at indicated time intervals, and counted as described. No significant differences were observed in the growth rate of null and norm populations expressing recombinant PKP1 or vector controls in either 1.2 mM Ca²⁺ or 0.09 mM Ca²⁺ (data not shown).

Patients lacking PKP1 are characterised by skin fragility and an altered pattern of wound healing (chronic erosions and excessive scale-crust after trauma). We compared in vitro wound healing between nullpB, nullPKPand normpB cells. We noted repeatedly that cells lacking PKP1 migrate at a faster rate than cells expressing PKP1 in response to wounding(Fig. 3). Two sets of experimental data were generated on two different sets of time-lapse apparatus(A and B). Between 8 and 16 hours post-wounding in data set A, nullpB1 cells migrated at speeds ranging between 161.8 μm/hour and 189.85 μm/hour, while nullPKP1 cells migrated at speeds ranging between 106.49 μm/hour and 139.09 μm/hour (n=3 in both cases,see Fig. 3). Between 8 and 16 hours post-wounding in data set B, nullpB2 cells migrated at speeds ranging between 120 μm/hour and 193.3 μm/hour, while nullPKP2cells migrated at speeds ranging between 73.6 μm/hour and 105.5 μm/hour(n=3 in both cases, data not shown). We considered that a reorganisation of the actin cytoskeleton into a more migratory phenotype (less stress fibres, more ruffling membranes) might also characterise PKP1 deficiency. However, examination of actin filament organisation revealed no obvious differences between nullpB and nullPKP cells at the wound edge or in a monolayer (data not shown). We also considered that the integrin complement could be altered between nullpB and nullPKP cells but could not show any significant difference in the levels of integrins α1, 2, 3, 5, 6, 9, αV, αVβ5,αVβ6 and β4 (data not shown) between nullpB, nullPKPand normpB cells. No αVβ3 was detected in any of the cell lines (data not shown).

Light and electron microscopy of PKP1-deficient skin sections revealed widening of the intercellular spaces between keratinocytes in the suprabasal layers but not in the basal or upper granular layers(McGrath et al., 1997; McGrath et al., 1999). Keratinocytes lacking PKP1 growing in culture under high (1.2 mM)Ca²⁺ concentrations showed no obvious morphological differences from keratinocytes expressing PKP1 (Fig. 4, upper panel). We examined the response of nullkeratinocytes to a low Ca²⁺ switch. In low Ca²⁺concentrations, cell-cell zonula occludens and zonula adherens junctions are disassembled and inhibited from forming whereas DMs are only disassembled in culture if they have not reached a state of maturity known as calcium-independence (Garrod,1996) (discussed below). It was observed that, after a low Ca²⁺ switch, nullpB cells appeared to round up more and become more refractile under phase contrast light microscopy than nullPKP or normpB cells. We quantified this finding using phase contrast time-lapse microscopy in two separate experimental data sets using six cell lines derived from four different individuals. Cells seeded at high density and grown for 72 hours in high Ca²⁺ medium were then switched to low Ca²⁺ medium and photographed every 6 minutes for 24 hours. Using image analysis software it was determined that nullpBcell lines became more refractile than nullPKP and normpBcell lines (Fig. 4). This difference was not observed if the cells were grown for less than 48 hours(nullPKP and normpB became as refractile as nullpBcells, data not shown).

Using immunofluorescence microscopy, we examined the distribution of the desmosomal components PKP1 (Fig. 5A), Dp (Fig. 5B),plakoglobin (data not shown), desmoglein 3 (data not shown) and desmocollin 3(Fig. 5C) in cells grown on coverslips in high Ca²⁺ medium(Fig. 5A-C, left hand panels),cells that were grown in high Ca²⁺ medium and subject to a low Ca²⁺ switch for 1 hour (Fig. 5A-C, middle panels), and cells grown in high Ca²⁺medium and subject to a low Ca²⁺ switch for 24 hours(Fig. 5A-C, right hand panels)prior to fixation. No differences in desmosomal protein distribution could be seen between nullpB, nullPKP and normpB cell lines grown in high Ca²⁺ medium (examples of Dp and desmocollin 3 shown in the left hand panels of Fig. 5B,C). However, 1 hour after a low Ca²⁺ switch the distribution of Dp markedly changed in nullpB cell lines compared with nullPKPand normpB cell lines (Fig. 5B, right hand panels). Far less Dp was localised to the cell membrane in nullpB cells after 1 hour of a low Ca²⁺switch, although a small proportion of cells did retain membrane bound Dp,compared to nullPKP and normpB cells. This difference was not obvious for plakoglobin (data not shown), desmoglein 3 (data not shown) or desmocollin 3 (Fig. 5C, middle panels) at the one-hour time point. However, 24 hours after a low Ca²⁺ switch, we noted a large reduction in membrane bound Dp,plakoglobin (data not shown), desmocollin 3 and desmoglein 3 (data not shown)in nullpB cells compared to nullPKP and normpBcells for all desmosomal components examined (examples of Dp and desmocollin 3 shown in Fig. 5B and 5C, right hand panels).

The observation that Dp seems to move from the cell membrane in response to low Ca²⁺ earlier than other desmosomal components analysed is interesting in light of the fact that, of the desmosomal components examined in this study, only Dp showed gross differences in distribution in PKP1-deficient skin compared with control skin(Fig. 6A,B). In order to assess whether Dp staining moved from membrane to cytoplasm before that of the other desmosomal components examined, nullpB cells were double stained for Dp and desmoglein 3 after 1 hour of a low Ca²⁺ switch. Cells that contained only membrane-bound desmoglein 3, rather than colocalised Dp and desmoglein 3, were readily detected (Fig. 6C, upper panels). Co-localised membrane staining of Dp and desmoglein 3 was also readily detectable as expected, but membrane bound Dp in the absence of desmoglein 3 expression was not detected. This was not the case with double staining for Dp and desmoglein 3 in the nullPKP cells,where co-localisation was readily seen in the majority of cells(Fig. 6C, lower panels).

We also investigated the levels of cytosolic, membrane or junctional/cytoskeletal associated plakoglobin and desmoglein 3 using a sequential detergent extraction method described by Palka and Green(Palka and Green, 1997)(Fig. 5D). The overall level of plakoglobin and desmoglein 3 is less in nullpB2 cells than in nullPKP2 and normpB2 cells over all three fractions. This is in agreement with the corresponding protein levels in the total cell lysates presented in Fig. 2A. Comparing the Triton X-100 soluble pool (membrane associated) at 24 hours after a low Ca²⁺ switch, the level of plakoglobin and desmoglein 3 in the nullpB2 pool is virtually undetectable compared with nullPKP2 and normpB2 pool, supporting the immunofluorescence data which showed less desmosomal membrane-bound retaining cells in the nullpB2 population after 24 hours of a low Ca²⁺ switch(Fig. 5A-C).

To investigate the effect of re-expressing PKP1 in PKP1-deficient cells on DM number and length, we compared EM micrographs of nullpB1 and nullPKP1 lines. Results are shown for cells in culture after 3 days in 1.2 mM Ca²⁺ as well as cells grown for the same time period and subjected to a low Ca²⁺ switch for 1 hour(Fig. 7). In total, we counted 130 DMs from 31 EM micrographs for nullpB1 and 354 DMs from 32 EM micrographs of nullPKP1. We measured the length of 74 DMs from 33 EM micrographs of nullpB1 and 91 DMs from 27 EM micrographs of nullPKP1. Fig. 7 shows that PKP1 influences number and size of DMs in culture after 3 days of confluent growth (Fig. 7A,B).

The localisation of PKP1 to nuclei of cells indicates a possible signalling function, as demonstrated for other members of the armadillo gene family(Peifer et al., 1994). Plakophilin 2 has been shown to complex with nuclear particles, specifically RNA polymerase III, suggesting an important nuclear role for this family member (Mertens et al., 2001). It has also been shown that plakophilin 2 can associate with β-catenin, a critical downstream effector of the Wnt signalling pathway(Chen et al., 2002). One possible nuclear role of PKP1 then could be enhanced transcriptional levels of desmosomal proteins at sites where more adhesive strength is needed. We have demonstrated that this is not the case for Dp, plakoglobin, desmoglein 2 or desmoglein 3 (Fig. 2B).

Patients lacking PKP1 show a marked thickening of the skin, even in non-lesional sites, indicating a possible role for PKP1 in regulating keratinocyte proliferation. In our in vitro model, we have shown that re-expression of PKP1 in PKP1-deficient keratinocyte lines or over-expression of PKP1 in control keratinocyte lines does not affect cellular growth rates(data not shown).

Previous work has demonstrated that PKP1 concentrates endogenously and ectopically expressed desmosomal components, in particular Dp, to the plasma membrane (Smith and Fuchs,1998; Kowalcyzk et al., 1999; Hatzfeld et al., 2000). In this study, we identified an increase in levels of desmosomal proteins within cells expressing PKP1 compared with PKP1-deficient cells(Fig. 2A), although this difference was not a result of transcriptional regulation(Fig. 2B). We were unable to demonstrate a notable difference in the amount of membrane bound desmosomal proteins in cells lacking PKP1 compared with those expressing PKP1 using immunofluorescence analysis (for example Fig. 5B,C, left panels). When we looked biochemically at membrane bound desmosomal protein levels for plakoglobin and desmoglein 3 (Fig. 5D, Triton X-100 soluble, lanes 1, 4 and 7) we noted less protein in cells lacking PKP1 (in agreement with the data seen from total cell lysate levels, Fig. 2A) but that the proportion of cytosol-membrane-cytoskeletal pool protein level was similar. After 24 hours of a switch to low Ca²⁺ we noted that membrane bound plakoglobin and desmoglein 3 were virtually undetectable in nullpB2cells compared with nullPKP2 and normpB2 cells. Turnover of desmosomal proteins is high (Penn et al.,1987; Pasdar and Nelson, 1989) and the data presented here are consistent with the concept that PKP1 stabilises desmosomal proteins within the cell (Kowalcyzk et al., 1999), although we cannot rule out post-translational modification as a mode of action for increasing the levels of desmosomal proteins in cells expressing PKP1 compared to those lacking PKP1. We did not detect a significant difference in levels of E-cadherin,β-catenin or keratins between cell lines tested.

Calcium-dependence of DMs has been shown to correlate with wounding of confluent cell sheets (Wallis et al.,2000). Specifically, calcium-independence can be reversed upon wounding and this reversal is propagated to cells hundreds of micrometers from the wound edge (Wallis et al.,2000). We have shown that PKP1 influences not only the Ca²⁺ responsiveness of DMs but also keratinocyte migration in our culture system. We cannot rule out that PKP1 also acts on cell motility through other mechanisms, as it has been reported that PKP1 co-localises withβ-actin at the edge of wounded HaCat cells(Hatzfeld et al., 2000), but we saw no differences in β-actin organisation (data not shown) or integrin complement (data not shown). The data presented here demonstrate for the first time, that keratinocyte motility is affected by the lack of a specific desmosomal protein.

The adhesive properties of DMs within confluent sheets of epithelial cells differ from those within sub-confluent sheets of epithelial cells(Watt et al., 1984; Mattey and Garrod, 1986). At sub-confluency, DMs are said to be calcium dependent as their formation and internalisation is influenced by increasing, or decreasing (<0.1 mM),Ca²⁺ concentration. Conversely, cells at confluency possess DMs that are mostly calcium independent since neither depleting the Ca²⁺ concentration of the medium nor adding chelating agents promote DM disruption. However, calcium-independent DMs are not permanent structures, since they retain the capacity to revert to calcium-dependence in certain situations, such as following wounding(Wallis et al., 2000). Using phase-contrast microscopy and time-lapse image capture, we found that PKP1-deficient cells became significantly more refractile in response to a low Ca²⁺ switch. This implies that cells are less adherent to neighbouring cells in response to a low Ca²⁺ switch over 24 hours(Fig. 4).

Calcium-independent DMs have been identified by immunofluorescent staining of desmosomal components after Ca²⁺ depletion(Watt et al., 1984; Mattey and Garrod, 1986; Wallis et al., 2000). Using this approach we have shown that, after 24 hours of a low Ca²⁺switch, far fewer calcium-independent DMs are present in nullpB cells than in nullPKP and normpB cells. These differences can be identified using an antibody to Dp after only 1 hour of a low Ca²⁺switch (Fig. 6C). We also demonstrated, using biochemical methods, that the proportion of membrane bound plakoglobin and desmoglein 3 was far less after a low Ca²⁺ switch for 24 hours in the nullpB population than in nullPKP and normpB populations.

Collectively these data suggest that, firstly, PKP1 has a role in the transition of DMs from a calcium-dependent to a calcium-independent state and,secondly, in cells lacking PKP1, Dp is displaced from the membrane in response to low Ca²⁺ before other desmosomal proteins studied here. This indicates a key function for PKP1 in stabilising Dp within DMs and is consistent with Dp immunostaining in patient skin samples(Fig. 6A,B).

Ultrastructural comparison of PKP1-deficient cells expressing recombinant PKP1 or vector controls shows that the reexpression of PKP1 increases the size of DMs in our culture system (Fig. 7B). This observation supports other models that predict PKP1 enhances desmosomal cohesiveness through lateral interactions with Dp,providing additional links with the DM and the IF network (Kowalcyzk et al.,1999). Ultrastructural analysis of cells under Ca²⁺ switch conditions demonstrates that after 3 days of confluent culture, nullpB cells retained fewer DMs than nullPKP(Fig. 7A), supporting the immunofluorescence data identifying a lower proportion of calcium-independent DMs. These data, when assessed with those showing a reduction of membrane bound desmosomal components in nullpB cells compared to nullPKP and normpB after Ca²⁺ switching, suggest that calcium-independence may take longer to form in cells lacking PKP1.

We have examined cellular and biochemical characteristics of keratinocytes derived from two patients with skin-fragility ectodermal dysplasia syndrome. For ethical reasons our two control populations of keratinocytes were derived from adult, rather than aged matched, donors. Nonetheless these `older' cells behaved no differently than did PKP1-deficient keratinocytes to which we restored PKP1 expression. The most direct comparisons though were with vector alone controls and our results suggest that PKP1 influences DM stability and organisation. Specifically, restoration of PKP1 expression stabilises the aggregation/assembly of other DM components, especially Dp, although this process does not involve changes in DM gene transcription. Lack of PKP1 leads to epidermal thickening and hyperkeratosis but our study shows that PKP1 does not affect in vitro cell growth. PKP1 does have a role in regulating cell migration since lack of PKP1 increases keratinocyte migration after wounding. PKP1 also influences the transition of DM from a calcium-dependent state to a calcium-independent state, which may be important in restoration and maintenance of an intact epithelial barrier during wound healing, a function that is compromised in patients with inherited ablation of this DM component.

We thank Fiona Keane for providing immortalised normal breast keratinocytes, David R. Garrod for providing the antibody 11-5f, Gary P. Nolan and ATCC for providing the ΦNX-Ampho Cells and Mark Morgan for assistance with flow cytometry. This work was supported by DEBRA UK, the Wellcome Trust,the FH Muirhead Charitable Trust, the National Foundation for Ectodermal Dysplasia (NFED), and Action Research.