Analysis of the interactions between BP180, BP230, plectin and the integrin α6β4 important for hemidesmosome assembly

Hemidesmosomes (HDs) are multi-protein complexes that promote epithelial-stromal cohesion in stratified and complex epithelia, and connect the intermediate filament system of basal epithelial cells to proteins of the extracellular matrix. These complexes, which ultrastructurally appear as tripartite structures along the plasma membrane of basal cells, are composed of at least five different proteins: the laminin-5 receptor α6β4, the bullous pemphigoid antigens 180 (BP180, BPAG2 or type XVII collagen) and 230 (BP230 or BPAG1-e), CD151 and plectin (Jones et al., 1998; Borradori and Sonnenberg, 1999; Sterk et al., 2000). In certain tissues, such as intestinal epithelia, and cultured epithelial cells, a second type of HD has been identified, which is composed of α6β4 and plectin (Uematsu et al., 1994; Orian-Rousseau et al., 1996). These type II HDs, in contrast to the classical or type I HDs, do not exhibit the typical tripartite structure.

BP230 and plectin are cytoplasmic proteins that belong to the plakin protein family, which also includes desmoplakin, envoplakin and periplakin. These proteins are crucially involved in the organization of the cytoskeleton (Ruhrberg and Watt, 1997; Leung et al., 2001). They are composed of domains that have considerable sequence homology. Their N-terminus consists of a plakin domain containing a number of subdomains of high α-helical content, designated NN, Z, Y, X, W and V, whereas the central coiled-coil rod domain is composed of heptad repeats thought to be involved in the dimerization of the plakin (Green et al., 1992). Finally, their C-terminal and exhibits one or more homologous repeat sequences designated A, B or C. In plectin as well as in neuronal and muscular isoforms of BP230 (BPAG1-a and BPAG1-b, respectively), a calponin-type actin-binding domain (ABD) precedes the plakin domain (Brown et al., 1995; McLean et al., 1996; Leung et al., 2001). The C-terminal end of plakins has intermediate filament binding properties (Meng et al., 1997; Wiche et al., 1993; Yang et al., 1996; Leung et al., 1999), whereas the N-terminal end harbors specific sequences that target the proteins to distinct membrane sites, such as HDs or desmosomes, cell-cell adhesion complexes in a variety of epithelia (Kowalczyk et al., 1997; Rezniczek et al., 1998; Geerts et al., 1999; Hopkinson and Jones, 2000).

The α6β4 integrin plays a central role in the assembly of HDs. Loss of α6β4 due to mutations in the genes for either the α6 or β4 subunit causes a distinct form of pyloric atresia associated with junctional epidermolysis bullosa (PA-JEB), and is characterized by fragility and extensive blistering of the skin. In affected patients HDs are rudimentary or completely absent (Vidal et al., 1995; Niessen et al., 1996; Ruzzi et al., 1997). A similar phenotype is observed in α6 or β4 null mutant mice (van der Neut et al., 1996; Georges-Labouesse et al., 1996; Dowling et al., 1996). The large cytoplasmic domain of the integrin β4 subunit is essential for the formation of HDs (Nievers et al., 1998; Murgia et al., 1998). It is over 1000 amino acids long and harbors two pairs of fibronectin type III (FNIII) repeats, separated by a connecting segment (CS) (Hogervorst et al., 1990; Suzuki and Naitoh, 1990). The second FNIII repeat and the first 35 amino acid residues of the CS of the β4 integrin are required for the recruitment of plectin into HDs (Niessen et al., 1997a; Nievers et al., 2000; Niessen et al., 1997b). The CS and the third FNIII repeat have been implicated in the binding to BP180 (Borradori et al., 1997; Aho and Uitto, 1998; Schaapveld et al., 1998). Furthermore, the cytoplasmic domain of β4 has been reported to interact with BP230 (Hopkinson and Jones, 2000).

BP180 is a type II transmembrane protein with a 466-amino acid cytoplasmic domain and a large collagenous extracellular domain (Giudice et al., 1992). There is evidence that, like α6β4, BP180 binds laminin-5 (Reddy et al., 1998). In cultured epithelial cell lines, the localization of BP180 in HDs is dependent on its interaction with the cytoplasmic domain of β4 (Borradori et al., 1997; Schaapveld et al., 1998). Furthermore, the extracellular domain of BP180 is also able to interact with the α6 integrin subunit (Hopkinson et al., 1995; Hopkinson et al., 1998). Finally, BP180 is involved in the recruitment of BP230 into HDs (Borradori et al., 1998; Hopkinson and Jones, 2000).

The aim of our study was: (1) to further assess the potential of the α6β4, BP180, BP230 and plectin to associate with each other; (2) to map the involved binding sites and, finally, (3) to define the importance of these interactions for the recruitment of these proteins into HD in combined yeast two-hybrid assays and cell-transfection studies. Our findings show that interactions between these hemidesmosomal components are more complex than previously anticipated, uncover the potential of BP180 to interact with plectin and reveal that the recruitment of these proteins into HDs is regulated by a hierarchy of interactions that each appear to have a different impact on HD assembly.

Immortalized β4-deficient PA-JEB keratinocyte cell line has been described previously (Schaapveld et al., 1998). BP180-deficient keratinocytes, immortalized by transfection with human papilloma virus-18 E6 and E7 genes were kindly provided by P. Marinkovich (University of California San Francisco, San Francisco, CA). The PA-JEB and GABEB (generalized atrophic benign epidermolysis bullosa) keratinocytes were grown in keratinocyte serum-free medium (SFM) (Gibco-BRL) supplemented with bovine pituitary extract, 5 ng/ml epidermal growth factor, 100 U/ml penicillin and 100 U/ml streptomycin. Cells were grown at 37°C in a humidified, 5% CO₂ atmosphere.

Mouse monoclonal antibody (mAb) 233 against BP180 (Nishizawa et al., 1993) and mouse mAb 121 against plectin/HD1 (Hieda et al., 1992) were kind gifts from K. Owaribe (University of Nagoya, Nagoya, Japan). Rat mAb 439-9B recognizes an extracellular epitope on the integrin β4 subunit and mouse mAb 450-11A directed against the cytoplasmic domain of β4 were purchased from Pharmingen (San Diego, CA). The rabbit polyclonal antiserum against BP230 (Tanaka et al., 1990) was a kind gift from J. R. Stanley (University of Pennsylvania, Philadelphia, PA). The human mAbs 5E and 10D against BP230 were generously provided by T. Hashimoto (Keio University, Tokyo, Japan) (Hashimoto et al., 1993). Rabbit polyclonal antibody against laminin-5 was kindly provided by P. Rouselle (Lyon, France). The rabbit polyclonal antibodies against the extracellular domain of β4 (sc-9090) and against hemagglutinin (HA) epitope tag (sc-805) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Secondary antibodies were purchased from Rockland (Gilbertsville, PA) (FITC-conjugated goat anti-mouse immunoglobulin (Ig) G) and Molecular Probes (Eugene, OR) (Alexa-488 conjugated goat anti-human IgG and Texas-Red-conjugated goat anti-rat and anti-rabbit IgG).

PA-JEB or GABEB cells were grown to 40% confluence in 12-well tissue-culture plates (Falcon; Becton Dickinson, Lincoln Park, NJ). Transient transfections were performed with 0.8 μg cDNA using Lipofectin, according to the manufacturer's instructions (Gibco-BRL). Transfection mixtures were replaced by SFM medium after 6 hours and incubated in this medium for 12 hours. Subsequently, the SFM medium was replaced by Nutrient Mixture Ham's F12/Dulbecco's MEM (1:3) for an additional 24 hours after which the cells were assayed for gene expression.

PA-JEB and GABEB cells grown on glass coverslips were fixed with 1% paraformaldehyde in PBS for 10 minutes and permeabilized with 0.5% Triton X-100 in PBS for 5 minutes at room temperature. After rinsing in PBS and blocking with 2% BSA in PBS for 30 minutes at room temperature, the cells were incubated with primary antibodies for 60 minutes at room temperature and then washed three times with PBS. Cells were subsequently incubated with Alexa-488, FITC- and Texas Red-conjugated secondary antibodies directed against mouse, rat, rabbit or human IgG for 45 minutes at room temperature. Coverslips were washed three times, mounted in Mowiol/DABCO and viewed under a Leica confocal scanning laser microscope.

The yeast strain Saccharomyces cerevisiae PJ69-4A (a kind gift from P. James, University of Wisconsin, Madison, WI), which contains the following genetic markers: trp1-901, leu2-3, his3-200, gal4Δ, gal80Δ, LYS2::GAL1-HIS3 and GAL2-ADE2 (James et al., 1996) was used as host for the two-hybrid assays. This strain contains two tightly regulated selectable Gal4-driven reporter genes, his and ade, allowing sensitive detection of protein-protein interactions between Gal4 fusion proteins. The Gal4 activation domain (AD)- and Gal4 binding domain (BD)-fusion plasmids were co-transformed into PJ69-4A, as described previously (Schaapveld et al., 1998), and equal aliquots of transformed cells were spread out on plates containing yeast synthetic complete medium lacking leu and trp (vector markers) or leu, trp, his and ade (vector and interaction markers). Plates were incubated at 30°C and growth of colonies was scored after 6 and 10 days. The plating efficiencies on -leu,-trp,-his,-ade (SC-LTHA) plates, as compared with the plating efficiency on -leu,-trp (SC-LT) was used as a measure for the strength of the two-hybrid protein interaction. Autonomous activation of the reporter genes was suppressed by the addition of 2 mM 3-amino-1,2,4-triazole (a His antagonist) (A8506; Sigma Chemical Co.). Expression of the Gal4-fusion proteins was analyzed by immunoblotting with antibodies against the Gal-activation or -DNA binding domain (sc-1663 and sc-510, respectively; Santa Cruz).

Constructs for the yeast two-hybrid studies were generated using standard cloning techniques. All nucleotide and amino acid positions are numbered with the ATG initiation codon at position one. The cDNA sequences used for alignment and designation of primers are available from GenBank under accession number m77830 (desmoplakin); mn_000494 (BP180); m69225 (BP230); u53204 (plectin) and x53587 (β4). Plasmid inserts were generated by restriction digestion or PCR using the proofreading Pwo DNA polymerase (Boehringer Mannheim) and gene-specific sense and antisense primers containing restriction site tags. Numbers in superscript correspond to the amino acid residues of subclones encoded within the Gal4 (AD) or -(BD) fusion proteins. Vectors used were the yeast Gal4 AD or BD expression vectors pACT2 or pAS2.1 (Clontech).

For the quantitative analysis of β-galactosidase activity, five yeast colonies were combined and grown to an OD₆₆₀ of approximately 1.0 in selective medium lacking Leu and Trp. β-Galactosidase activity was determined at 37°C using the Pierce yeast β-galactosidase assay kit (75768) with O-nitrophenyl β-D-galactopyranoside as substrate. The A₄₀₅ was measured in an ELISA-reader and the time at which the reaction reached a value of 0.2 was taken to calculate the β-galactosidase activity using the equation: 1,000×A₄₀₅/(cell volume (ml) × time of reaction (min) × OD₆₆₀). Samples that did not reach this value within 4 hours were left overnight and measured the next morning. The final values are the results from three independent determinations.

To specify the molecular interactions important for the formation of HDs, we first analyzed, by immunofluorescence microscopy, the distribution of various hemidesmosomal components in keratinocyte cell lines lacking either β4 or BP180.

In β4-deficient PA-JEB cells, the transient expression of the β4 subunit was shown to restore the cells' ability to form HD-like structures (Schaapveld et al., 1998). In extension to these studies, we found that, in many cells of a PA-JEB keratinocyte cell line that stably express β4 (PA-JEB/β4 cells), the subcellular localization of laminin-5 and plectin was largely identical to that of α6β4 (Fig. 1A,B). Furthermore, the distribution of BP180 and BP230 was the same, i.e. in structures appearing as dots and patches, which are typical for HD-like structures or stable anchoring contacts (Fig. 1C). By contrast, co-localization of either β4 with BP180 or of plectin with BP230 was mostly only partial (Fig. 1D-I). These results indicate that certain, but not all, HD-like structures contain, in addition to α6β4 and plectin, BP180 and BP230 as shown by immunofluorescence microscopy studies, i.e. they represent both type I and type II HDs (Uematsu et al., 1994).

When BP180-deficient keratinocytes obtained from a patient with GABEB were analyzed by immunofluoresence microscopy, the distribution of β4 appeared to be normal; it was concentrated in patches at sites of cell-substrate contacts (Fig. 2A-C). However, although in nearly all PA-JEB/β4 cells the staining of β4 and plectin largely overlapped, in many of the BP180-deficient GABEB cells there were several patches containing β4 but not plectin (compare Fig. 2A with Fig. 1B). Furthermore, consistent with previous observations (Borradori et al., 1998), BP230 was not present in these HD-like structures (Fig. 2C). Nevertheless, on transient transfection with cDNA for BP180, we found that BP230 was recruited into HD-like structures together with BP180 (Fig. 2D-I). Together, these findings obtained in PA-JEB and GABEB keratinocytes indicate that the recruitment of BP230 into HD-like structures is regulated by distinct — either direct or indirect — interactions, involving mainly β4 and BP180. Furthermore, for the localization of plectin in HD-like structures to be efficient, it seems that BP180 is also required.

Previous studies have shown that the recruitment of BP180 into HDs depends on a direct association of BP180 with β4 (Borradori et al., 1997). To identify the β4 binding region on the cytoplasmic domain of BP180, several deletion mutants were generated and expressed in yeast cells together with a β4 construct containing all four FNIII repeats of the β4 cytoplasmic domain, β4^1115-1666 (Fig. 3). Interactions were detected by growth of yeast cells on plates lacking histidine and adenine (see Materials and Methods). Although constructs BP180^1-231 and BP180^145-401 bound to β4^1115-1666, there was no interaction with either BP180^1-147 or BP180^145-230. Notably, the latter contains the stretch of amino acids 145-230 shared by the above two BP180 constructs, which had binding activity. Because proper expression of the BP180^145-230 mutant was ascertained by immunoblotting of yeast cell lysates, the lack of binding could not be due to defective protein expression (data not shown). Furthermore, deletion of the stretch of amino acids 145-230 from the cytoplasmic domain of BP180, BP180^{1-401,Δ145-230} had no impact on its interaction with β4^1115-1666. Together, these results strongly suggest that there are at least two distinct binding sites for β4 on BP180: one located in the first 230-amino-acid stretch of BP180 and a second in a more C-terminally located region encompassing amino acids 231-401.

BP180 has recently been shown to associate with BP230 via a region of 280 amino acids (residues 180-460) spanning half the cytoplasmic domain (Hopkinson and Jones, 2000). To define the region on BP180 that binds to BP230, we carried out additional yeast two-hybrid assays. As shown in Fig. 3, the BP180 mutants that contain the region of 85 amino acids 145-230 bound to BP230^1-555 and BP230^1-1156, whereas those lacking it did not. Furthermore, the BP180^145-230 mutant that only contains this stretch of amino acids was also able to bind, albeit less efficiently. Thus, although additional sequences might further strengthen their interaction, this 85-amino-acid stretch is necessary and sufficient for the binding of BP180 to BP230. Finally, two mutants carrying a deletion of the first 37 N-terminal amino acids (BP180^38-422 and BP180^{38-422,Δ229-324}), a region previously identified as being crucial for recruiting BP230 into HDs in transfection studies (Hopkinson et al., 1995; Borradori et al., 1998), were both able to interact with BP230. These results suggest that BP180 interacts with BP230 through a region of 85 amino acids (BP180^145-230) and that the first 37 N-terminal amino acids are dispensable for binding in yeast.

Next, we analyzed in transfection studies the importance of the stretch of 85 amino acids of BP180, which contains a binding site for BP230, for the recruitment of the latter into HDs. When introduced in BP180-deficient keratinocytes, BP180^Δ145-230 is correctly recruited into HDs and, like wild-type BP180, is co-localized with β4 (Fig. 4A). The same was observed in cells expressing BP180^Δ1-36 (Fig. 4C). However, at variance with cells expressing either wild-type BP180 (Fig. 2I) or the BP180^Δ1-36 mutant (Fig. 4D), in cells expressing BP180^Δ145-230, BP230 was found only rarely in HD-like structures together with the BP180 mutant (Fig. 4B). Thus, consistent with the results in yeast two-hybrid interaction assays, in which a region encompassing amino acids 145-230 of BP180 is required and sufficient for its interaction with BP230, the deletion of this sequence from BP180 dramatically reduced recruitment of BP230 into HDs. Deletion of the 36 most N-terminal amino acid residues of BP180 had no effect on the recruitment of either BP230 or β4 into HD-like structures.

We next investigated which region of BP230 is involved in binding to BP180. For this purpose, a series of mutant forms of BP230 were generated and tested together with BP180^1-401 in yeast two-hybrid assays. The results show that BP230 interacted with BP180 by the Z-Y subdomains (Fig. 5A,B) and, although the Y domain alone was able to bind to BP180^1-401, albeit slightly less efficiently (Fig. 5B), no binding of the Z domain alone could be detected. In turn, BP180^145-230, which interacted weakly with BP230^1-555 (Fig. 3), did not bind to the isolated Z-Y or Y subdomains of BP230 (data not shown), possibly because interaction of this short stretch of BP180 with BP230 requires additional sequences flanking the Z-Y subdomains of BP230. Indeed, when two larger fragments of BP230 containing the Z-Y subdomains (BP230^1-555 and BP230^1-887) were tested for binding to BP180^1-401 using a β-galactosidase assay, they did bind more strongly than the isolated Z-Y or Y subdomains (Fig. 5C).

The observations that in BP180-deficient keratinocytes the localization of plectin in HD-like structures appeared to be somewhat impaired and that the Y domain of plectin is homologous to that of BP230 (Fig. 5E) prompted us to investigate whether plectin can directly bind to BP180. We therefore generated two constructs encoding the Z-Y or Y domains of plectin and tested their binding ability in yeast. As shown in Fig. 5B, both constructs interacted with BP180^1-401, although less strongly than the corresponding constructs of BP230 as confirmed by a quantitative β-galactosidase assay (Fig. 5C). Hence, BP180 is capable of binding to the Y domain of not only BP230 but also of plectin. Surprisingly, we found that the Z-Y and Y domains of desmoplakin, which is a component of desmosomes, also interacted with BP180. As in the case of BP230, the isolated Z-Y and Y subdomains of plectin or desmoplakin did not interact with BP180^145-230 (data not shown).

To assess whether an interaction of BP230 with BP180 is required for the localization of BP230 in HDs, we generated various HA-tagged BP230 mutants in which the Z-Y domains had been deleted. Because human keratinocytes lacking BP230 expression have not yet been identified, we expressed these mutants in PA-JEB/β4 cells, whereas immunofluorescence microscopy analyses were carried out with polyclonal antibodies against the HA-epitope tag to distinguish the localization of the ectopically expressed BP230 mutant proteins from that of endogenous BP230. Although, in most cells, a BP230 construct containing the first 1-887 N-terminal amino acids (BP230^1-887) was found diffusely distributed over the cytoplasm (Fig. 6A), in very few of the transfected cells it was concentrated in HD-like structures, together with BP180 (Fig. 6B). By contrast, a mutant form of BP230, BP230^1-2161, which encompasses the rod domain, was found not only in HD-like structures in a larger proportion of the cells, but also in aggregates in the cytoplasm, particularly in transfected cells that strongly expressed the cDNA (Fig. 6C). Other than BP230^1-2161, BP230^1-2161,ΔZY from which the Z-Y domains had been deleted was not colocalized with BP180 in HD-like structures, but was found in small aggregates (Fig. 6D).

Because both BP230^1-2161 and BP230^1-2161,ΔZY contain the central rod domain of BP230 and thus have the potential to form dimers with endogenous BP230, transfected cells were also stained with an anti-BP230 antiserum raised against a C-terminal fragment of BP230 (amino acids 1722-2203). Because a large portion of this fragment is part of the two BP230 mutants, the anti-BP230 antiserum will stain not only endogenous BP230, but also the two BP230 mutants. Consistent with the localization of transfected BP230^1-2161 in HDs, the staining pattern with the anti-BP230 and anti-HA antibodies overlapped in transfected cells and was comparable to that seen with anti-BP230 in untransfected cells, i.e. in HDs (Fig. 6E). By contrast, BP230^1-2161,ΔZY was again found in small aggregates and it was not obviously co-localized with endogenous BP230 in HDs (Fig. 6F). Thus, even if dimerization of BP230^1-2161,ΔZY mutants with endogenous BP230 occurs, this did not result in the recruitment of the mutated molecule into HDs.

Collectively, these results suggest that: (1) the Z-Y domain contains sequences important for the localization of BP230 into HDs, most likely via binding to BP180, and (2) the presence of the rod domain of BP230 thought to be implicated in the formation of dimers (Ruhrberg and Watt, 1997; Leung et al., 2001) increases the correct targeting of the protein into HDs.

In BP180-deficient keratinocytes, the recruitment of BP230 into HD-like structures is crucially dependent on the re-expression of BP180 (Borradori et al., 1998). However, in PA-JEB keratinocytes, i.e. in the absence of α6β4, no HDs are formed, despite the fact that these cells express BP180 and BP230. We therefore wondered whether the β4 subunit might directly bind to BP230 and thus also affects its localization. We tested the binding activity of several β4 constructs, in which one or more FNIII repeats were deleted or in which the CS was progressively shortened, with a BP230 construct containing the first 1156 amino acids of the N-terminus in yeast two-hybrid assays. As depicted in Fig. 7, removal of part of the fourth FNIII repeat resulted in loss of interaction. The deletion of the first pair of FNIII repeats from the cytoplasmic domain of β4 had no effect on binding, but the removal of also the CS abolished the interaction. Specifically, an interaction between BP230 and β4 still occurred with progressive deletions up to amino acid 1436, but not beyond. These data show that the extreme C-terminal portion of the CS in combination with the third and the fourth FNIII repeats of β4 are required for its binding to BP230.

Finally, as a control, the various β4 constructs were also tested against the cytoplasmic domain of BP180, BP180^1-401. In extension to previous studies (Geerts et al., 1999), we found that construct β4^1457-1552 encompassing the third FNIII contains the minimal sequences necessary for the interaction with BP180.

We next investigated which sequences within BP230 are involved in its binding to β4 in yeast. Various BP230 constructs were tested with a β4 construct containing the first FNIII repeat up to the fourth repeat, β4^1115-1666 (Fig. 5A). The results indicate that the first stretch of 92 amino acids at the N-terminus of BP230 is sufficient for its interaction with β4.

To ascertain that a binding site for β4 was not accidentally introduced by fusing the Gal4 domain to the N-terminal extremity of BP230, we generated additional mutants as depicted in Fig. 5D. All these constructs were able to interact with β4, providing evidence that the interaction was indeed dependent on BP230-specific sequences and not on the Gal4 moiety. Furthermore, the results show that the first three amino acids of BP230 are dispensable for this interaction.

Except for the first 56 amino acids, the N-terminal region of BP230 is almost identical to that of two isoforms, dystonin-1 and -2, encoded by the BPAG1 gene that also encodes BP230. These two isoforms, previously thought to be specifically expressed in the nervous system (Yang et al., 1996), differ from the BP230 isoform by the presence of an ABD at their extreme N-terminus that is absent from BP230. To determine whether or not the interaction between BP230 and β4 is mediated by the stretch of amino acids 56-92 common to BP230 and the two dystonin isoforms, we generated constructs comprising amino acids 1-92, 1-56 or 56-190 of BP230. When these constructs were assayed for an interaction with β4^1115-1666, it was found that amino acids 1-56 of BP230 interacted, although less efficiently than amino acids 1-92, with β4, whereas amino acids 56-190 did not (Fig. 5D). These data indicate that the interaction of BP230 with β4 is mediated by sequences specific for the epidermal isoform, although obviously the common part also influences binding.

The finding that BP230 can interact with both β4 and BP180 prompted us to investigate whether BP230 can replace plectin in supporting the formation of HDs. For this, we made use of a mutant β4 subunit (β4^R1281W) that is unable to interact with plectin, but can bind to BP180 and BP230 (Geerts et al., 1999; Koster et al., 2001). Stable expression of β4^R1281W in PA-JEB cells by using retroviral transduction revealed that this mutant can support the formation of HD-like structures containing BP180 and BP230, consistent with previous observations using transient transfection protocols (Geerts et al., 1999). However, these HD-like structures appear to be less conspicuous and dense than those formed in the presence of wild-type β4 (compare Fig. 1 and Fig. 8). Furthermore, in the vast majority of cells, the β4 mutant was not co-localized with BP180 or BP230. From these results, we conclude that: (1) when plectin is not bound to β4, β4 binding to BP180 is strongly reduced, and (2) BP230 cannot replace plectin in supporting proper localization of BP180 into HDs.

In this study we have further characterized the interactions between different hemidesmosomal components involved in the assembly of HDs. In extension to recent studies, we show that the β4 cytoplasmic domain interacts with both BP180 and BP230 (Schaapveld et al., 1998; Hopkinson and Jones, 2000) via sequences located in its C-terminal half, encompassing the third FNIII repeat (for BP180) and the C-terminal sequences of the CS up to the fourth FNIII repeat (for BP230). These sites are different from those implicated in the binding of plectin and its localization into HDs, which are located in the first pair of FNIII repeats and the N-terminal portion of the CS (Geerts et al., 1999; Koster et al., 2001). Furthermore, we show that the binding sites on BP180 for plectin and BP230 are different from those involved in the binding to β4, and encompass the same stretch of 85 amino acids in the N-terminal half of the cytoplasmic domain of BP180.

The different interactions between the various hemidesmosomal components are depicted in Fig. 9, along with a model that shows how these interactions are likely to contribute to the formation of stable type I HDs.

The localization of BP180 in HDs containing α6β4 and plectin has previously been shown to depend on an interaction of BP180 with the cytoplasmic domain of β4 (Borradori et al., 1997; Schaapveld et al., 1998). Interactions between the α6 subunit and the extracellular NC16a domain of BP180 are probably also implicated (Hopkinson et al., 1998). Here, we identified a third interaction partner of BP180, i.e. plectin.

The interaction between BP180 and plectin is not sufficiently strong to induce the formation of HDs in the absence α6β4, because in β4-deficient PA-JEB cells, which contain BP180 and plectin, no HDs are formed. However, it may strengthen the interaction between α6β4, BP180 and plectin in a three-molecular complex, thereby ensuring their proper incorporation in HDs. Consistent with this idea, we found less plectin in the β4-positive adhesion structures of the BP180-deficient GABEB keratinocytes than in those of PA-JEB/β4 keratinocytes, at least as assessed by immunofluorescence. Moreover, studies with PA-JEB keratinocytes stably expressing a mutant β4 subunit (β4^R1281W) that is unable to interact with plectin (Geerts et al., 1999; Koster et al., 2001) revealed that the incorporation of BP180 into cell-substrate structures is severely compromised. Only few cells formed HD-like adhesion structures containing α6β4, BP180 and BP230. Thus, although α6β4 can bind to BP180, proper incorporation of this protein into HDs only occurs when plectin is also available.

The results with the PA-JEB/β4^R1281W cells also suggest that the role of plectin in supporting efficient localization of BP180 in HDs cannot be replaced by BP230. This is surprising considering the fact that BP230 can interact with both β4 and BP180. Both BP230 and plectin probably bind via sequences contained in their Y subdomain to the same region (amino acids 145-230) on BP180 (see below). Thus, it is possible that, when plectin is not bound to α6β4, it competes with BP230 for binding to BP180. However, given the fact that the binding activity of BP180 for BP230 is greater than that for plectin, this is not very likely. An alternative possibility, which we favor, is that binding of plectin to β4 is required for activating the β4 cytoplasmic domain, by unfolding it and rendering it accessible for interaction with BP230 (see also Fig. 9). Evidence for such an intramolecular folding of the β4 cytoplasmic domain has been presented in both biochemical and yeast two-hybrid assays (Rezniczek et al., 1998; Geerts et al., 1999). Regardless of what the mechanism might be, it is clear that plectin and BP230 are not exchangeable in their ability to support the formation of proper HDs.

In view of our finding that binding to BP180 is mediated by the Y domain of plectin, it is interesting to note that Pulkkinen et al. (Pulkkinen et al., 1996) have described a deletion of three amino acids, QEA, in this region in a patient with epidermolysis bullosa simplex associated with muscular dystrophy (EBS-MD). A pathological consequence of this deletion, therefore, could be that plectin is unable to bind to BP180, and that the ternary complex of α6β4, BP180 and plectin, which serves as a platform for the incorporation of BP230 into HDs (see below), cannot be (properly) formed, leading to deficient assembly of HDs. Future experiments, however, should reveal whether indeed the deletion of these three amino acids in the Y domain prevents the binding of plectin to BP180.

In previous studies (Hopkinson et al., 1995; Borradori et al., 1997), it was found that the 36 most N-terminal amino acids of the cytoplasmic domain of BP180 were implicated in the recruitment of BP180 into HDs. However, in this study, no evidence was found for the involvement of this region in the localization of BP180 in HDs. An important difference between our study and the study by Borradori et al. (Borradori et al., 1997) is that the latter used a chimeric protein, which consisted of the membrane targeting sequence of K-ras fused to the cytoplasmic region of BP180, whereas in the present study full-length BP180 has been used. Full-length BP180 forms trimers (Hirako et al., 1996) and, therefore, it is predicted that it interacts more strongly with β4 than the chimeric monomeric molecule lacking the collageneous extracellular domain. In that case, binding of full-length BP180 to β4 may have become less dependent on additional sites of interaction that reside on the cytoplasmic domain of BP180. However, it remains unexplained why we have not been able to detect an interaction between β4 and a BP180 mutant containing the first 36 amino acids (BP180^1-147) in yeast two-hybrid interaction assays. Perhaps the fusion of the N-terminus of BP180 to the Gal4 (BD) has destroyed the β4-binding site in the N-terminus of BP180. Although the importance of the first 36 amino acids of BP180 for the binding to β4 could not be confirmed in the yeast two-hybrid system, the finding that both BP180^1-231 and BP180^145-401, but not the amino acids 145-230 that are shared by these two constructs, interact with β4 suggests that there are two β4 binding sites on BP180. One of them is located in the first 230 N-terminal amino acids of the BP180 cytoplasmic domain and the other is located C-terminally to this region. Although the β4-binding site in the first 230 N-terminal amino acids of BP180 may overlap with that for BP230 (see below), the two sites are clearly not identical. Binding of β4 still occurs when the region of amino acids 145-230, which is essential and sufficient for the interaction with BP230, has been deleted from the cytoplasmic domain of BP180 (BP180^{1-401,Δ145-230}). In line with the results obtained in yeast, a full-length BP180 molecule carrying this deletion became co-aligned with β4 in HD-like structures.

The observation that in BP180-deficient keratinocytes, BP230 was not co-localized with α6β4 and plectin in HD-like structures but diffusely distributed over the cytoplasm indicates that β4 and plectin are not sufficient for the proper localization of BP230 into HDs. Efficient recruitment of BP230 only occurs when BP180 is also expressed in these cells (Schaapveld et al., 1998). However, it is likely that the additional binding site for BP230 that is provided on β4 also contributes to the recruitment of BP230 in HDs. Thus, a model emerges in which α6β4 interacts with plectin, after which they bind BP180 in a ternary complex. Ultimately, BP230 binds to two components in the complex, i.e. α6β4 and BP180, and stable type I HDs are then formed (Fig. 9).

Our results show that sequences in the C-terminal domain of β4 (the second pair of FNIII repeats) interact with residues 1-56 of BP230, which are contained within the N-terminal domain of BP230 (residues 1-980) that has been shown to be involved in the interaction with β4 (Hopkinson and Jones, 2000). In the same study, a second β4 binding site on BP230 has been identified in its C-terminal domain, residues 1812-2649. By contrast, we found that a fragment of BP230 encompassing residues 2077-2649 did not associate with β4, but bound to intermediate filaments in yeast and was able to decorate intermediate filaments in transfected cells (J. Koster, unpublished observations). It is possible that the binding site for β4 is located in the region of amino acids 1812-2077 of BP230 and that this region contributes to the recruitment of BP230 into HDs. In support of this, in transfection experiments with PA-JEB/β4 cells, BP230^1-887 was localized in only very few of the HDs, whereas a BP230 construct, which also includes the rod domain and the region 1812-2077 (BP230^1-2161), was detected in a higher percentage of the HDs. The rod domain, which is an integral part of all members of the plakin family and is responsible for the dimerization of these molecules, may also contribute to the localization of BP230 in HDs by its ability to dimerize the binding sites in the N- and C-terminal domains.

In a recent study (Hopkinson and Jones, 2000) it was shown that the first 170 amino acids, as well as amino acids 555-700 of an N-terminal 700 amino acid fragment of BP230, are required for its binding to BP180^1-517. Our finding that BP230^1-555 interacted with BP180^1-401 shows that the amino acids 555-700 of BP230 are not essential. Furthermore, no binding was detected when a fragment containing the first 172 amino acids of BP230 was used, whereas this fragment did bind to β4. Binding of BP180 to BP230 appeared to be mediated by the Y subdomain of BP230. The importance of the Y subdomain for binding is further supported by the fact that in transient transfection experiments using PA-JEB/β4 cells, BP230^1-2161, but not the construct from which the Z-Y domain had been deleted (BP230^1-2161,ΔZY) was recruited into HDs. However, the latter construct had a tendency, when expressed in transfected cells, to form aggregates in the cytoplasm, which might prevent its recruitment into HDs. The fact that Hopkinson and Jones (Hopkinson and Jones, 2000) found that sequences additional to the Z-Y subdomains of BP230 are required for interaction with BP180^1-517, in contrast to our results with BP180^1-401, may be explained by their use of a less-sensitive assay system. This is supported by the finding that BP180^1-401 binds more strongly to BP230^1-555 and BP230^1-887 than to the isolated Z-Y and Y subdomains of BP230 and, furthermore, that BP180^1-401, but not BP180^145-230, can bind to the Z-Y and Y subdomains of BP230. Thus, it seems that additional sequences stabilize the interaction of the Y subdomain with BP230.

We found, unexpectedly, that the Y subdomain of desmoplakin was also able to interact with BP180. In fact, desmoplakin is not found in HDs (Green and Jones, 1996), but in desmosomes, in which BP180 is not present. It is possible that desmoplakin does not recruit BP180 into desmosomes (despite its ability to directly interact with BP180) because α6β4 is absent from these adhesion complexes. This assumption is indirectly supported by the observation that α6β4 is necessary for the efficient recruitment of BP230 and plectin into HDs, despite the ability of these two molecules to bind to BP180 (Borradori et al., 1997; Schaapveld et al., 1998). Alternatively, the interaction between plakins and cell-surface receptors might be subject to regulation. Clearly, these results further stress the important role of α6β4 in the assembly of HDs.

The crucial role of plectin in the formation of HDs as observed in cultured keratinocytes does not seem to be supported by findings in vivo. Indeed, it has been shown that HDs can be formed in plectin-deficient mice, although the HDs appear to be rudimentary and their number is reduced (Andrä et al., 1997). We think that this discrepancy is partly due to a different role of BP180 in vitro and in vivo. As we have shown in vitro, the clustering and polarization of BP180 at the cell basis is entirely dependent on the presence of α6β4 and plectin, probably because the specific ligand for BP180 is not produced by keratinocytes in culture. No polarized expression of BP180 was observed in β4-deficient PA-JEB keratinocytes (Schaapveld et al., 1998). By contrast, in vivo, where the ligand for BP180 is almost certainly present and deposited into the basement membrane (it may be produced by the mesenchymal cells), clustering of BP180 is probably less dependent on plectin and α6β4, and thus can also occur in the absence of plectin. BP230 may then subsequently interact with the BP180-α6β4 complexes, resulting in the formation of rudimentary HDs. In support of this, we have found that when α6β4 is absent, as in patients with PA-JEB, some HDs can be observed, the formation of which may have been initiated by interaction of BP180 with an unidentified ligand in the epidermal basement membrane (Niessen et al., 1996).

In conclusion, we have characterized interactions between several molecules that are crucially involved in the assembly of HDs. We show that the different hemidesmosomal components can interact with multiple other components and that often interactions with more than one component are required for their efficient recruitment into HDs. Our findings that the cytoplasmic domain of β4 bears a scaffold of FNIII repeats responsible for interactions with the three major hemidesmosomal components plectin, BP180 and BP230, underscores its essential role in the formation of HDs.

We thank our colleagues for their generous gifts of antibodies and P. Marinkovich for providing the GABEB keratinocyte cell line. We are grateful to C.P.E. Engelfriet for comments on the mansuript. This work was supported by a grant from the Dystrophic Epidermolysis Bullosa Research Association (DEBRA Foundation, Crowthorne, UK) to A.S. Part of this work supported by the Swiss National Foundation of Scientific Research (32-51083.97 and 32-56727.99) to L.B.