Co-assembly of Envoplakin and Periplakin into Oligomers and Ca -dependent Vesicle Binding IMPLICATIONS FOR CORNIFIED CELL ENVELOPE FORMATION IN STRATIFIED SQUAMOUS EPITHELIA*

Plakin family members envoplakin and periplakin have been shown to be part of the cornified cell envelope in terminally differentiating stratified squamous epithelia. In the present study, purified recombinant human envoplakin and periplakin were used to investigate their properties and interactions. We found that envoplakin was insoluble at physiological conditions in vitro, and co-assembly with periplakin was required for its solubility. Envoplakin and periplakin formed soluble complexes with equimolar stoichiometry. Chemical cross-linking revealed that the major soluble form of all periplakin constructs and of envoplakin/periplakin rod domains was a dimer, although co-assembly of the fulllength proteins resulted in formation of higher order oligomers. Electron microscopy of rotary-shadowed periplakin demonstrated thin flexible molecules with an average contour length of 88 nm for the rod-plus-tail fragment, and immunolabeling EM confirmed the molecule as a parallel, in-register, dimer. Both periplakin and envoplakin/periplakin oligomers were able to bind synthetic lipid vesicles whose composition mimicked the cytoplasmic side of the plasma membrane of eukaryotic cells. This binding was dependent on anionic phospholipids and Ca . These findings raise the possibility that envoplakin and periplakin bind to the plasma membrane upon elevation of intracellular [Ca ] in differentiating keratinocytes, where they serve as a scaffold for cornified cell envelope assembly.

Terminally differentiating stratified squamous epithelial cells assemble a specialized protective barrier structure at their periphery termed the cornified cell envelope (CE) 1 (1,2). The CE replaces the plasma membrane and provides vital physical and permeability barrier properties to epithelial tissues in mammals. It consists of a 10-nm thick layer of transglutaminase cross-linked proteins. In the particular case of the epidermis, a 5-nm thick layer of ceramide lipids is covalently bound to the proteins (3). Envoplakin and periplakin have been proposed to provide a scaffolding on to which the CE is assembled (4).
As early as 1984, Simon and Green (5) identified two proteins with apparent molecular masses of 195 and 210 kDa that were incorporated into the keratinocyte CE upon activation of cellular transglutaminase. Later it was shown that the 195-kDa protein, mostly soluble in undifferentiated cells, becomes insoluble and submembranously located at the periphery of differentiated cells (6). After the sequences of these CE precursors were identified, the proteins were named periplakin and envoplakin, respectively (4,(7)(8)(9). Envoplakin-derived peptides were found among peptides released by proteolysis of isolated CEs (10,11). Moreover, both envoplakin and periplakin were identified as CE lipid-attachment sites along with involucrin (11). Their expression precedes the expression of other CE precursor proteins (4,7,12,13). Envoplakin is expressed in the suprabasal layers of stratified squamous epithelia, but not in simple epithelia or nonepithelial cells (7). Periplakin is expressed in a number of epithelial and nonepithelial tissues as well (4,14). Both proteins are up-regulated during the terminal differentiation and partially colocalize with desmosomes in addition to decorating the intermediate filaments (IF) in differentiating keratinocytes (4, 7, 14 -17). The immature CEs of cultured keratinocytes contain a higher proportion of periplakin and envoplakin than do mature CE from skin (12). Furthermore, masking of their epitopes by deposition of loricrin on the cytoplasmic face of the CE at later stages of its assembly suggested an outermost location for periplakin and envoplakin, immediately adjacent to the surrounding layer of covalently attached lipids (12). Together, these data led to the proposal that periplakin and envoplakin, along with involucrin, initiate the CE assembly, providing sites for attachment of reinforcement proteins (loricrin, small proline-rich proteins and others) from one side and CE-specific lipids from the opposite side (12).
Envoplakin and periplakin are members of the plakin family (18,19). Plakins are expressed in tissues that experience mechanical stress, where they play a vital role in maintaining tissue integrity. They have been shown to interconnect the microtubule, actin filament, and IF components of the cytoskeleton (20 -25), as well as cell junctional complexes (25,26). Rotary shadowing of isolated plectin and desmoplakin revealed flexible dumbbell-like shapes (20,27,28). The globular Nterminal "head" domains and C-terminal "tail" domains of several family members (desmoplakin, bullous pemphigoid antigen 1, plectin, and microtubule-actin cross-linking factor) have been shown to be responsible for the cross-linking interactions listed above (14,16,17,23,25,29,30), whereas the flexible central domains (segmented ␣-helical coiled-coil rods in most cases) are responsible for homo-oligomerization and serve as cytoskeletal spacers. Formation of dimers by desmoplakin (28) and tetramers by plectin (27) has been reported. Based on theoretical calculations, envoplakin and periplakin were predicted to be the only family members able to form heterodimers with each other, and their coimmunoprecipitation under stringent buffer conditions provided support for this idea (4). Later, in transfection studies it was found that envoplakin accumulated in nuclear and cytoplasmic aggregates but redistributed to desmosomes and the IF upon introduction of periplakin into the cells (16). Because both proteins can be found decorating the IF network, envoplakin and periplakin were proposed to connect the IF to the CE, conferring mechanical integrity to a cornified cell (4,7).
To better understand the interactions of envoplakin and periplakin, we have conducted a study of their molecular properties using purified recombinant proteins. In this paper, we report the oligomerization state, conformation, and ultrastructure of periplakin and envoplakin/periplakin molecules, and their binding properties to synthetic lipid vesicles (SLV).

EXPERIMENTAL PROCEDURES
Generation of cDNA Constructs-Full-length or fragments ( Fig. 1) of human envoplakin and periplakin cDNAs were cloned using pairs of primers (Table I) with NdeI and HindIII sites in forward and reverse primers, respectively. First strand cDNA was synthesized with the ThermoScript RT-PCR System (Invitrogen) using as a template total RNA isolated from a normal human keratinocyte cell culture, grown for 1 day after induction of terminal differentiation by 1.2 mM calcium. The reaction was carried out for 60 min at 58°C with E6 or P6 envoplakin or periplakin gene-specific primers, respectively. Target DNA was amplified by PCR using the Titan One Tube RT-PCR Kit (Roche Applied Science) in 35 cycles with specific primer pairs (Fig. 1). The products were cloned into the pET-23a(ϩ) expression vector (Novagen, Madison, WI) with vector-encoded C-terminal His 6 tag fusion, and expressed in the Epicurian Coli BL21-Codon Plus(DE3)-RIL strain of Escherichia coli (Stratagene, La Jolla, CA).
Protein Purification-Bacterial cultures were grown in LB medium containing 50 g/ml carbenicillin and 25 g/ml chloramphenicol. Recombinant protein expression was induced in mid-log phase at A 600 ϭ 0.6 -0.7 by 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h. Soluble proteins Et, Pt, and Ph were extracted from pelleted bacteria with B-PER Bacterial Protein Extraction Reagent (Pierce) supplemented with 500 mM NaCl, 5 mM imidazole, 10 mM ␤-mercaptoethanol, and 0.5 mM phenylmethylsulfonyl fluoride. The other recombinant proteins were solubilized from inclusion bodies with 500 mM NaCl, 8 M urea, 5 mM imidazole, 10 mM ␤-mercaptoethanol, and 0.5 mM phenylmethylsulfonyl fluoride in 20 mM Tris-HCl, pH 8.0. His-tagged proteins were purified using a nickel-chelated Sepharose Fast Flow column (Amersham Biosciences) followed by anion-exchange chromatography on a Mono Q column (Amersham Biosciences). Full-length envoplakin and its rod-containing fragments Er and Ert were further purified by preparative SDS-PAGE, freed of SDS by ion-pair extraction (31), and re-dissolved in assembly buffer (see below) containing 6 M urea immediately prior to use. For production of radioactively labeled proteins, bacteria were grown in M9 minimal medium (32), and [ 35 S]methionine or [ 3 H]leucine were added to 10 Ci/ml simultaneously with the addition of isopropyl-1-thio-␤-D-galactopyranoside. Protein concentrations were measured using the BCA Protein Assay Reagent (Pierce).
Circular Dichroism (CD) Measurements-Spectra of purified proteins at 0.1 mg/ml in 50 mM potassium phosphate at pH 7.5 were measured using a Jasco J-715 spectrometer calibrated with a solution of ammonium (Ϫ)-10-camphorsulfonate. The spectra were scanned at 50 nm/min, digitized, and recorded at 0.5-nm intervals. All spectra were measured at room temperature using a 0.1-cm path length cuvette and averaged over 4 individual scans. The data were analyzed using CON-TIN-CD software (33).
Protein Assembly in Vitro-Envoplakin, periplakin, or their mixtures at 1 mg/ml in 6 M urea, 25 mM triethanolamine-HCl, pH 7.8, were diluted 20-fold with 1 mM dithiothreitol, 1 mM EDTA in 25 mM triethanolamine-HCl, pH 7.8 (Assembly buffer), and incubated at room temperature for 30 min followed by centrifugation at 175,000 ϫ g for 30 min. Supernatants and pellets were analyzed by SDS-PAGE on 4 -12 or 4 -20% Tris glycine gels (Invitrogen). Supernatants were analyzed on 4 -12% Tris glycine gels under non-denaturing conditions as well. When complexes were assembled from 35 S-labeled envoplakin and 3 H-labeled periplakin, protein bands were cut out the Coomassie Blue-stained gels and dissolved in 0.5 ml of 30% hydrogen peroxide at 60°C overnight followed by addition of 10 ml of CytoScint liquid scintillation mixture (ICN Biomedicals, Costa Mesa, CA). 35 S and 3 H radioactivity were counted separately using a Beckman LS 6500 scintillation counter (Beckman Coulter, Fullerton, CA).
Electron Microscopy-For rotary shadowing, 20-l aliquots of protein at 50 g/ml in 150 mM NaCl, 25 mM triethanolamine-HCl, pH 7.8, were mixed with 40 l of glycerol and sprayed on to freshly cleaved mica (34). The mica was placed on the rotary table of a RMC freeze-fracture unit (Boeckeler Instruments, Tucson, AZ) and pumped to ϳ2 ϫ 10 Ϫ5 pascal. The samples were rotary shadowed at room temperature with platinum/carbon at an angle of 6°and coated with carbon at 90°. The replicas were floated off on deionized water, picked up on copper grids, observed with a JEOL 1200 EX transmission electron microscope (Peabody, MA) operated at 80 kV, and photographed at a magnification of 50,000. The negatives were digitized with a Microtek ScanMaker 8700 scanner at a resolution of 1200 dpi. Digitized micrographs were processed using Adobe Photoshop version 5.5. Contour length measurements were done with Image J version 1.30. 2 Anti-periplakin Antibodies-The polyclonal chicken antibodies were produced by Aves Labs, Inc. (Tigard, OR). Purified recombinant tail domain of periplakin (periplakin amino acids 1638 -1756) without His tag was used for immunization. The antibodies were affinity purified on HiTrap NHS HP column (Amersham Biosciences) with the antigen coupled.
Lipid mixtures were dried under nitrogen followed by high vacuum for 4 h, and hydrated in 100 mM NaCl, 3 mM NaN 3 , 5 mM dithiothreitol, 200 mM sucrose, 50 mM Tris-Cl, pH 8.0, by shaking at 60°C for 30 min. Large, unilamellar vesicles with a nominal mean diameter of 250 nm were prepared by extrusion through a 200-nm polycarbonate membrane in the Avanti Mini-extruder (Avanti Polar Lipids, Alabaster, AL) at 60°C. SLV were diluted with an equal volume of the above buffer without sucrose, spun down at 175,000 ϫ g for 30 min in a Beckman airfuge (Beckman Coulter), and resuspended in the latter buffer to a final concentration of 20 mol of lipids/ml. Phospholipid concentrations were measured by phosphorus content (35).
Assaying Membrane Association of Envoplakin and Periplakin-Typically, 20-l reaction mixtures contained 1 g of protein and varying amounts of SLV (15-150 nmol of lipids) in 150 mM NaCl, 5 mM dithiothreitol, 50 mM Tris-Cl, pH 8.0, with a range of CaCl 2 concentrations. Samples were incubated at room temperature for 30 min and centri-2 Available at rsb.info.nih.gov/ij.
fuged at 175,000 ϫ g for 30 min. Supernatants (non-bound protein) and pellets (SLV-bound protein) were analyzed by SDS-PAGE on Tris glycine gels. To determine the binding capacity of liposomes, the amount of protein in the pellets was quantitated by densitometry of Coomassie Blue-stained gels.

Co-assembly with Periplakin Is Required for Proper
Folding of Envoplakin-cDNAs encoding the full-length proteins or their separate domains ( Fig. 1) were cloned into bacterial expression vectors. When expressed, most of these proteins, with the exception of Ph, Et, and Pt constructs, were insoluble in bacterial cells and accumulated in inclusion bodies. Experiments with several commercially available bacterial strains specifically adapted for recombinant protein expression and with different temperature regimens did not yield detectable levels of soluble protein. Nevertheless, all periplakin constructs and envoplakin head domain could be successfully refolded from urea solutions by step dialysis against buffers with progressively decreasing urea concentrations upon completion of purification in urea-containing buffers. Their folding is evidenced by the fairly good agreement of circular dichroism measurements of secondary structure content with predictions, both yielding a high content of ␣-helix (Table II). Whereas rod domain-containing periplakin constructs remained soluble upon diluting out the urea (Fig. 2, panels Pr, Prt, and P), full-length envoplakin or its rod-containing fragments precipitated when the urea concentration was lowered below 1 M (Fig.  2, panels Er, Ert, and E). However, if envoplakin was first mixed with periplakin in 6 M urea, it remained soluble after dilution of urea (Fig. 2, panels Er ϩ Pr, Ert ϩ Prt, and E ϩ P; Table II). The periplakin rod domain alone was sufficient to rescue full-length envoplakin from aggregation (Fig. 2, panel E ϩ Pr).
Periplakin and Envoplakin Form Hetero-oligomers at a 1:1 Ratio-When envoplakin and periplakin were refolded together in equimolar amounts and the resulting soluble fraction was analyzed by non-denaturing gels, no bands of periplakin homo-oligomers were detected, but a band of envoplakin/ periplakin hetero-oligomer appeared (Fig. 3A, compare lanes  M, MЈ, and MЉ to lanes Pr, Prt, and P). Note the absence of any soluble protein when attempts were made to refold envoplakin by itself (Fig. 3A, lanes Er, Ert, and E), and the appearance of high molecular weight soluble complexes at the bottom of gel wells in the case of full-length periplakin and its mixture with envoplakin (Fig. 3A, lanes P and MЉ). To prove that the major product of envoplakin and periplakin co-assembly was their hetero-oligomer and to determine its stoichiometry, assembly experiments were performed with 35 S-labeled envoplakin and 3 H-labeled periplakin. Protein bands were cut out of the Coomassie Blue-stained non-denaturing gels and 35 S and 3 H radioactivity were counted separately. The bands marked as Er/Pr in Fig. 3B contained both 35 S and 3 H radioactivity. When the proteins were mixed at an equimolar ratio, the major band contained equal amounts of envoplakin and periplakin (Fig.  3B). When either protein was present in up to 3-fold excess, the soluble material was near-equimolar with only a slight excess of either 35 S or 3 H radioactivity (see "Discussion"). Accordingly, these data suggest that the two proteins are present in equimolar amounts in heteromeric complexes.  FIG. 2. Periplakin rescues envoplakin from aggregation. Individual proteins at 1 mg/ml or their equimolar mixture in 6 M urea, 25 mM triethanolamine, pH 7.8, were diluted 20-fold by 25 mM triethanolamine, pH 7.8, 1 mM dithiothreitol, 1 mM EDTA in a centrifuge tube and incubated at room temperature for 30 min. Samples were centrifuged at 175,000 ϫ g for 30 min and supernatants (s) and pellets (p) were analyzed by SDS-PAGE on 4 -12% gradient gels stained with Coomassie Blue. Note that all envoplakin constructs are found in the pellet fractions when periplakin is not present (panels Er, Ert, and E), but shift to the supernatant when co-assembled with periplakin (panels Er ϩ Pr, Ert ϩ Prt, E ϩ P, and E ϩ Pr).

Periplakin Forms Heterodimers and Higher Order Complexes with Envoplakin That Are as Stable as Periplakin Dimers-To
determine the numbers of polypeptide chains in periplakin homo-oligomers and periplakin/envoplakin hetero-oligomers, chemical cross-linking was performed with the bifunctional reagent bis[2-sulfosuccinimidooxycarbonyloxy)ethyl]sulfone. Cross-linking of full-length periplakin and envoplakin resulted in two major products: a band with apparent mobility of about 850 kDa and a large complex that was unable to enter the gel and stayed in the loading well (Fig. 4A, left panel). Both crosslinking products were soluble as judged by ultracentrifugation (Fig. 4A, right panel). Absence of any intermediate between monomers and the 850-kDa band even at the shortest time of cross-linking led us to believe that this band is a dimer.
The major cross-linking product of all periplakin constructs containing the rod domain was a dimer (Fig. 4B, panels Pr, Prt, and P). Co-assembly of periplakin and envoplakin rod domains stopped at the dimer level as well (Fig. 4B, panel Er ϩ Pr). However, when the tail domains were included in the constructs, multimers were the predominant assembly product at urea concentrations below 1 M (Fig. 4B, panel Ert ϩ Prt). The stability of the multimers increased further when the head domains were also present (Fig. 4B, panel E ϩ P). Moreover, whereas attempts to cross-link the separate head or tail domains of either protein did not reveal formation of any oligomers, dimers of periplakin tail and both dimers and tetramers of periplakin head domains were observed on non-dena-turing gels or on SDS gels under non-reducing conditions (data not shown). These data support the view that with envoplakin and periplakin, as with other plakins (16,36), the rod domain plays an important role in oligomerization; however, they also serve notice that both the N-and C-terminal domains have roles as well. Interestingly, it has been recently reported that removal of the entire rod domains from envoplakin and periplakin did not completely inhibit their interaction (17).
Despite the theoretical prediction that a parallel homodimer of periplakin should be more stable than its heterodimer with envoplakin (0.79 ionic interactions per pair of heptads versus 0.57, respectively) (4), we found that envoplakin/periplakin hetero-oligomers were at least as resistant to urea denaturation as periplakin dimers (compare panels Pr versus Er/Pr, Prt versus Ert/Prt, and P versus E/P in Fig. 4B). However, both complexes are less stable than desmoplakin dimers, whose disruption required urea concentrations in excess of 6 M (28).
To obtain further evidence of envoplakin/periplakin heterodimer formation, assembly experiments were carried out with inclusion of either protein radioactively labeled, with the other being "cold" and therefore invisible on radioautographs. Very similar band patterns were observed in the dimer region on the gel radioautograph when either radioactive envoplakin or periplakin were present during cross-linking with the corresponding unlabeled partner (Fig. 4C, lanes 2 and 4, respectively). The presence of multiple bands in the dimer region could be explained by different amounts of cross-links per protein chain in the cross-linking products.
Ultrastructure and Dimensions of Periplakin-The purified recombinant proteins were studied in the electron microscope using the rotary shadowing technique. The most uniform distributions of regularly spread molecules were obtained for the "headless" periplakin rod plus tail fragment (Fig. 5, A and C). They revealed thin flexible filaments of quite uniform length, averaging 88 nm (S.D. 13 nm, n ϭ 121) (Fig. 5E).
To test for polarity, molecules were labeled with antibodies against the periplakin tail domain and examined by rotary shadowing. Single end labeling would imply a parallel mode of association, and double end labeling, an antiparallel mode of association. Only the former (single end) labeling was observed (Fig. 5, B and D). On average, the contour length between the center of a labeling antibody molecule and the opposite end of the periplakin fragment was 89 nm (S.D. 11 nm, n ϭ 33), indicating that the antibodies labeled the end of the molecule. Taken together, these observations show that this periplakin construct forms a parallel, in-register, dimer.
We also tried shadowing experiments with other constructs but the data, although indicative of filamentous structures, did not yield a predominant morphological type: rather, the high flexibility and inter-crossing of molecules, presumably in higher order oligomers, resulted in images that were too complicated and too variable to be readily interpretable. A possible exception may be full-length periplakin for which we observed a few linear, i.e. non-branched although variably curved, molecular images of nearly double the dimer length (data not shown). Such molecules could represent an end-to-end association of dimers. However, further data will be required to substantiate this proposition. In no case did we observe large globular end domains in full-length periplakin samples, such as have been reported for plectin (27).
Both Periplakin and Envoplakin/Periplakin Oligomers Bind to Model Membranes in a Calcium-and Anionic Phospholipiddependent Manner-As periplakin and envoplakin are recruited to the cell periphery during the earliest stages of epithelial differentiation, we explored the possibility of their direct binding to phospholipid bilayers in vitro. Binding to SLV mim- icking the phospholipid composition of the inner side of the eukaryotic plasma membrane has been demonstrated for another CE precursor protein, involucrin (37). Using a similar assay, we found that both periplakin and the envoplakinperiplakin complex were able to bind to SLV. Efficient binding required the presence of Ca 2ϩ ions in the reaction mixture (Fig.  6, A, D, and E) with anionic phospholipids in the SLV, phosphatidylserine and phosphatidylglycerol providing the most efficient binding (Fig. 6C). It should be noted that significant binding of the envoplakin-periplakin complex was observed without Ca 2ϩ ions present, and that addition of Ca 2ϩ to 1 mM increased its binding by about 50%, whereas binding of periplakin increased 4 -5-fold under the same conditions (Fig.  6, D and E). On the other hand, envoplakin/periplakin showed severalfold higher affinity for SLV, compared with periplakin, independent of SLV composition (Fig. 6, C-E).
Next, we tried to determine which domain(s) of the proteins were responsible for the interaction with phospholipid bilayers. However, none of the separate domains of envoplakin or periplakin bound to SLV significantly (data not shown), although they were properly folded according to CD measurements (Table II). DISCUSSION In this study, we demonstrated that envoplakin and periplakin form dimers and multimers thereof in vitro. Periplakin is capable of forming homodimers but if envoplakin is available, it preferentially forms heterodimers. Envoplakin, on the other hand, requires co-assembly with periplakin for proper folding and dimerization. The central rod domain of periplakin is sufficient for this interaction. All these observations are in full agreement with the in vivo findings of Di-Colandrea and co-workers (16) who studied the interaction of envoplakin and periplakin by a transfection assay. However, their approach did not allow detailed characterization of their oligomeric state. Whereas the elongated flexible shape of plakins complicates the use of approaches such as gel filtration to determine their oligomerization states, chemical cross-linking may be used for this purpose. Previous cross-linking studies showed the formation of a dimer by desmoplakin (28) and a tetramer by plectin (27).
Our cross-linking data revealed that periplakin formed homodimers, and envoplakin co-assembled with periplakin yielded heterodimers as well as larger soluble structures. Other lines of evidence indicate that the basic building block is a parallel, in-register, dimer in which the two polypeptide chains associate primarily via the formation of a segmented coiled-coil by their rod domains. Thus the CD data showed very high content of ␣-helix, particularly for rod domain constructs (Table II). The periplakin rod domain was predicted to contain 13 heptad-rich subdomains, accounting for about 560 amino acid residues (4). With an axial rise of 0.15 nm per residue, these subdomains should form a segmented coiled-coil of total length of 84 nm. This value is in good agreement with the EM-based measurement, 88 nm, of the contour length of the periplakin rod-plus-tail construct. These data substantiate the widely held assumption that the rod domains of plakin proteins form parallel two-stranded coiled-coils (36). A key observation for the conclusion that periplakin-envoplakin complexes are heterodimers was the double radiolabeling experiment in which the dimers were found to contain nearly equimolar amounts of both proteins, regardless of the starting ratio. However, with a 3-fold excess of envoplakin, the ratio measured for soluble E:P was 1.6, slightly higher than expected. Such excess may reflect, in some oligomerization events, more than one envoplakin rod combining with (different sections of) the periplakin rod.
Our data indicate that dimers have a propensity to form higher order structures, provided that the head domains are present (Fig. 4). Our attempts to deduce a mode of association by rotary shadowing did not have a conclusive outcome, the images being complicated by the flexibility and length of the FIG. 5. Electron microscopy of shadowed periplakin molecules. The samples were rotary shadowed with platinum/carbon as described under "Experimental Procedures" and photographed at a magnification of 50,000. A and C, purified recombinant periplakin rod-tail fragment (Prt) alone. B and D, the fragment was incubated with 2-fold molar excess of anti-periplakin tail antibodies for 1 h at room temperature before processing. E, contour length measurements of periplakin rod-tail fragment (Prt). A and B, field views. C and D, representative molecules. Scale bar of 100 nm applies to all images. component dimers. In this context, we note that a symmetric tetramer with two dimers aligned in the antiparallel mode was proposed for plectin by Foisner and co-workers (20,27), who also observed stellate higher order complexes. Such a tetramer affords a bifunctional cross-linker suitable for connecting like cytoskeletal elements and may also apply to envoplakin/periplakin.
As both envoplakin and periplakin localize at the periphery of differentiating cells (4,7,16), they should have a binding partner(s) in the submembrane region. Interaction of periplakin via its N-terminal head domain with the cortical actin cytoskeleton in primary human keratinocytes was demonstrated (16). Here we explored the possibility of a direct interaction of both periplakin and envoplakin with a phospholipid bilayer. Both full-length periplakin and envoplakin/ periplakin oligomers were able to bind to SLV containing physiological amounts of anionic phospholipids. In contrast to involucrin, which exhibited a preference for phosphatidylserine-containing SLV (37), envoplakin and periplakin did not distinguish between phosphatidylserine-and phosphatidylglycerol-containing SLV. Notably, all SLV, independent of their composition, exhibited severalfold higher binding capacity for the envoplakin-periplakin complex compared with periplakin. Considering the wider tissue distribution of periplakin, envoplakin being a more "CE-specialized" protein, their heterooligomerization could be what brings periplakin to the CE, the role of periplakin in CE assembly being to serve as a chaperone for envoplakin. That periplakin is present in CEs in much lesser amounts than envoplakin (12) is consistent with this notion. The significance of envoplakin for CE assembly is evidenced by the reported increase in the proportion of fragile CE in envoplakin-null mice (38), accompanied by a delay in epidermal barrier acquisition. On the other hand, the fact that neither of the separate domains of envoplakin or periplakin was able to bind SLV suggests that interaction of different parts of the molecules is required to provide an SLV binding-competent conformation. It is worth mentioning that some reported interactions of periplakin required a combination of rod and tail sequences, whereas the individual domains were not sufficient to support the interaction (39 -41).