Dissection of Protein Linkage between Keratins and Pinin, a Protein with Dual Location at Desmosome-Intermediate Filament Complex and in the Nucleus*

Pinin is a cell adhesion-associated and nuclear protein that has been shown to localize in the vicinity of intermediate filament (IF) convergence upon the cytoplasmic face of the desmosomal plaque as well as in the nucleus. The localization of pinin to the desmosomes has been correlated with the reinforcement of intercellular adhesion and increased IF organization. In this study, keratins 18, 8, and 19 were identified to interact with the amino end domain of pinin in a two-hybrid screening. Further truncation analyses indicated that the 2B domain of keratin contains the sequence responsible for interacting with pinin. The amino end of pinin (residues 1–98) is sufficient to bind to keratin. Point mutation analyses revealed two essential residues within the pinin fragment 1–98, leucine 8 and leucine 19, for the interaction with keratin. Finally, in vitro protein overlay binding assays confirmed the direct interaction of the amino end domain of pinin with keratins, while pinin mutant L8P GST fusion protein failed to bind to keratins in the overlay assay. Coupled with our previous morphological observations and transfection studies, these data suggest that pinin may play a role in epithelial cell adhesion and the IF complex through a direct interaction with the keratin filaments. Pinin desmosome, recruited to nascent desmosomes and pinin dec-orating filaments near the subjected to 2 HIS selection; subsequently, the colonies of surviving yeast were replicated to 2 Ade plates. Positive colonies from 2 Ade selection were subjected to liquid culture ONPG b -galactosidase assays according to the manufacturer’s procedure (CLONTECH). The interaction between p53 and SV40 large T-antigen was used as a positive control in b -galactosidase assays according to the manufacturer’s pro-cedures. The base-line level of b -galactosidase activity was determined from control yeast co-transformed with GAL4-BD-pinin (residues 1–480) and GAL4-AD. Each reported value of b -galactosidase units represented an average enzyme activity determined from three independent colonies. The “prey” plasmids were recovered from triple pos- itive (HIS, Ade, and LacZ) clones and co-transformed with the control heterologous bait: GAL4BD-p53, GAL4BD-pinin C 9 (residues 470–717), or GAL4BD alone. In addition, the “prey” plasmid was also transformed by itself into the yeast host to rule out potential false-positives. Clones growing on 2 HIS, 2 Ade, exhibiting b -galactosidase activity and exhibiting no interaction with control baits, were subjected to sequencing. To examine the ability of truncations of pinin to interact with keratin, the GAL4BD vectors containing the individual pinin truncations or point mutation constructs were co-transformed with the pGAD10 vector containing keratin 18 into PJ69-4A yeast. To examine the ability of truncations of keratin 18 to interact with the amino end of pinin, the original bait was co-transformed with individual truncations of keratin 18, fused to the activation domain of GAL4 in the pGAD10. Co-trans-formants were assayed for growth on 2 HIS and 2 Ade and b -galacto- sidase activity. of pinin and truncations interacting with pinin amino portion 1–480. Either the 2B consensus motif deletion construct K18(69–372) or the 2B domain de- letion construct K18(69–276) was cotransformed into yeast PJ69–4A with hp-(1–480). The cotransformants were selected on 2 HIS, 2 Ade medium and subjected to b -galactosidase ( b -gal ) assay. A , yeast containing K18 (69–372), as well as yeast containing full-length K18 ex- hibited growth on SD/ 2 Trp, 2 Leu, 2 Ade medium, while the yeast containing K18 (69–276) exhibited no growth. B , b -galactosidase assays indicated that K18-(69–372) is able to bind to hp-(1–480), while K18-(69–276) exhibited no binding to hp-(1–480).

Pinin was first identified to be a protein associated with the desmosome, which was recruited to the preformed desmosomes of the epithelia but absent at nascent desmosomes (1). Immunofluorescence and immuno-EM studies have shown pinin decorating keratin filaments near the cytoplasmic face of the desmosomal plaque in the vicinity of keratin filament convergence upon the desmosome. Our previous studies have revealed a correlation between the placement of pinin at the desmosome and an increase in the organization/stabilization of desmosome-IF 1 complex (1,2). Presumably, one of the functions of pinin is related to the desmosome-IF complex.
The expression level of pinin has been correlated with the overall epithelial phenotype. HEK-293 cells, when transfected with pinin full-length cDNA, exhibited a striking phenotype change from a fibroblast-like spindle shape to cells with exten-sive cell-cell contact growing in culture as islands (2). Intriguingly, EM analysis of these transfected cells revealed that the array of epithelial cell junctions was enhanced. In addition, carcinoma-derived cells, when transfected with pinin cDNA, exhibited inhibition of anchorage-independent growth in soft agar. Furthermore, pinin's gene locus and dysregulation of pinin expression in primary tumor tissues suggest that pinin may function as a tumor suppressor in certain types of cancer (3,4).
Pinin has also been localized in the nucleus in various tissues as well as in cultured cell lines (5-7). Brandner et al. (6) has proposed an involvement of pinin in spliceosomal function. The dual location of pinin may be indicative of the involvement of pinin in multiple cellular activities, both at the desmosome and in the nucleus; however, it is not yet clear whether or not the function of pinin in cell-cell adhesion is coordinated with its function in the nucleus. As a step toward understanding the functions of pinin, we sought to identify proteins that interact with pinin. In this study, we focus on the ability of pinin to bind keratin.
Keratin filaments are anchored to the lateral plasma membrane at desmosomes. These intercellular junctions reinforce epithelial adhesion as well as integrate the IF network across the entire epithelium. Numerous structure-function studies of desmosomal proteins have revealed details pertaining to the molecular organization of desmosome-IF complex. The relationships among the desmosomal components have been extensively reviewed elsewhere (8 -11). The constitutive components of the desmosome include desmosomal cadherins (desmogleins and desmocollins) and plaque proteins, plakoglobin, desmoplakin, and plakophilin. Among these proteins, desmoplakin (12,13) and plakophilin (11,14) have been shown to bind directly to keratins. In addition, other peripherally associated desmosome proteins such as plectin (15,16), envoplakin/ periplakin (17,18), and pinin (1) are also thought to interact, directly or indirectly, with keratin. Significant questions pertaining to the molecular associations and specific roles of these accessory proteins of the desmosome remain.
To identify potential protein-protein interactions of pinin, a two-hybrid screening was performed with either the amino portion or the carboxyl portion of pinin as bait. In this study, we presented a detailed analysis on the binding of the amino end domain of pinin to one group of the identified proteins, the keratins. Keratin 18 (K18), keratin 8 (K8), and keratin 19 (K19) were shown to interact with the amino portion of pinin in the two-hybrid screen. Further truncation analyses defined the specific domain of keratin that mediates the interaction. In addition, the specific domain of pinin molecule sufficient for the interaction was characterized, and through site-directed mutagenesis, the essential residues within this particular domain were investigated. In vitro blot overlay assays were performed to confirm the interaction between the amino end domain of pinin and the keratins. Overall, our data strongly suggest that pinin is capable of binding directly to the intermediate filament proteins, specifically the keratins. These data provide important information on eventual understanding of mechanism by which pinin may affect the assembly/stabilization of epithelial cell adhesion.

MATERIALS AND METHODS
Yeast Strain and Media-The Saccharomyces cerevisiae strain PJ69 -4A (MATa trp1-901 leu2-3, 112 ura3-52 his3-200 gal4 gal80D LYS2::GAL1-HIS3 GAL2-ADE2 met::GAL7-lacZ) (19) was used in all the two-hybrid assays. The yeast was grown on synthetic medium (SD) with appropriate amino acid-omissions for plasmid selection. Tryptophan and leucine were selective markers for the co-transformed bait and prey plasmids. Histidine 3, adenine 2, and lacZ are reporter genes for interaction between GAL4-BD and GAL4-AD. In "ϪHIS" medium, histidine was omitted as well as tryptophan and leucine, while in "ϪAde" medium, adenine was omitted as well as tryptophan and leucine. In addition, 1 mM 3-aminotriazole was added in all of the media to inhibit the autoactivation of the histidine 3 reporter gene.
Bait Construct and Two-hybrid Screening-The DNA fragment encoding for pinin residues 1-480 was obtained by PCR and cloned inframe into the GAL4 DNA binding domain (GAL4BD; bait) vector pAS2-1 (CLONTECH, Matchmaker II system). The GAL4BD-pinin vector was co-transformed with a CLONTECH Matchmaker cDNA library into the yeast strain PJ69 -4A using the yeast transformation method of Gietz et al. (20). The library consisted of human fetal kidney cDNA fused to the activation domain of GAL4 (GAL4AD, prey) in the pGAD 10 vector (CLONTECH).
Approximately 10 6 transformants were screened. They were initially subjected to ϪHIS selection; subsequently, the colonies of surviving yeast were replicated to ϪAde plates. Positive colonies from ϪAde selection were subjected to liquid culture ONPG ␤-galactosidase assays according to the manufacturer's procedure (CLONTECH). The interaction between p53 and SV40 large T-antigen was used as a positive control in ␤-galactosidase assays according to the manufacturer's procedures. The base-line level of ␤-galactosidase activity was determined from control yeast co-transformed with GAL4-BD-pinin (residues 1-480) and GAL4-AD. Each reported value of ␤-galactosidase units represented an average enzyme activity determined from three independent colonies. The "prey" plasmids were recovered from triple positive (HIS, Ade, and LacZ) clones and co-transformed with the control heterologous bait: GAL4BD-p53, GAL4BD-pinin CЈ (residues 470 -717), or GAL4BD alone. In addition, the "prey" plasmid was also transformed by itself into the yeast host to rule out potential false-positives. Clones growing on ϪHIS, ϪAde, exhibiting ␤-galactosidase activity and exhibiting no interaction with control baits, were subjected to sequencing.
To examine the ability of truncations of pinin to interact with keratin, the GAL4BD vectors containing the individual pinin truncations or point mutation constructs were co-transformed with the pGAD10 vector containing keratin 18 into PJ69-4A yeast. To examine the ability of truncations of keratin 18 to interact with the amino end of pinin, the original bait was co-transformed with individual truncations of keratin 18, fused to the activation domain of GAL4 in the pGAD10. Co-transformants were assayed for growth on ϪHIS and ϪAde and ␤-galactosidase activity.
Generation of Pinin/Keratin Truncations and Pinin Point Mutations-Truncations of pinin and truncations of keratin 18 were generated by PCR using the primer sets listed in Tables I and II. PCR products of human pinin were fused in frame to the GAL4BD in the vector pAS2-1 at NdeI/SalI sites. PCR products of human keratin 18 were fused in frame to the GAL4AD in the vector pGAD10 at XhoI/ EcoRI sites. Point mutations of the pinin amino end 1-480, fused in frame to GAL4BD in the pAS2-1 vector, were generated using the Quick Change site-directed mutagenesis kit (Stratagene, La Jolla, CA) with the primer sets listed in Table III. As above, co-transformants were assayed for growth on ϪHIS and ϪAde and ␤-galactosidase activity.

Expression of Pinin Fusion Protein in Escherichia coli and Generation of the Polyclonal Antibody against the Pinin GST-Fusion Protein-
Pinin residues 1-165 were obtained by PCR with primers STS 65 (5Ј-CCG AAT TCC CGC TTC AGA GAG AAG ATG-3Ј) and STS 61 (5Ј-CGC TCG AGG GCC TTT CAG TAG CAA CAG-3Ј). This PCR fragment was cloned in frame to vector pGEX-4T-3 (Amersham Pharmacia Biotech) at XhoI/EcoRI sites. The glutathione S-transferase (GST) fusion protein GST-cp-(1-165) was expressed in E. coli strain BL21 (Novagen) and purified with glutathione-Sepharose 4B (Amersham Pharmacia Biotech) according to the manufacturer's instructions. Similarly, a mutant GST fusion protein of the residues 1-165, GST-cp-(1-165) L8P, with a substitution of leucine 8 by a proline, was generated with the Quick Change site-directed mutagenesis kit (Stratagene, La Jolla, CA), expressed and purified as described above.
The pinin DNA encoding for 5Ј-end residues 1-165 was also cloned into pET 28(ϩ)b (pET system; Novagen) and expressed as a T7-tagged and His 6 fusion protein in E. coli strain BL21 (Novagen). The fusion protein was affinity-purified using the charged HIS⅐Bind metal chelation resin (Ni 2ϩ beads) following the instructions of the manufacturer (Novagen, pET System Manual).
A rabbit polyclonal antibody (UF215) was generated using the GSTcp-(1-165) as antigen (Cocalico Biologicals, Inc.). The specific immunoactivity of UF 215 to pinin's amino domain was verified by Western blot on pET System expressed His 6 fusion protein described above (data not shown).
Purification of Keratin Filament Protein from Madin-Darby Canine Kidney Cells-Madin-Darby canine kidney cells were grown to confluence in Dulbecco's modified Eagle's medium (Life Technologies, Inc.), supplemented with 10% fetal calf serum (Life Technologies, Inc.), 2 mM glutamine and 200 units/ml each of penicillin and streptomycin. Keratin proteins were then prepared from these cells according to a procedure described elsewhere (21,22) with slight modifications. Cells were lysed in PBS (containing 1% Triton X-100, 0.6 M KCl, 1 mM MgCl 2 , 5 mM EDTA, 5 mM EGTA, and the following protease inhibitors: 1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol, 1 mg/ml leupeptin, 1 mg/ml pepstatin A, 1 mg/ml aprotinin (Sigma)). The extract was treated with DNase (0.5 g/ml) at 37°C for 20 min and then centrifuged at 2000 ϫ g at 4°C for 10 min to pellet the IF-enriched cytoskeleton. The IF-enriched cytoskeletal preparation was first extracted with PBS containing 5 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, and 1 mM dithiothreitol. The pellet was then sequentially extracted with low salt buffer (60 mM KCl, 1 mM EDTA, 1 mM cysteine, 10 mM ATP, 40 mM imidazole, pH 7.1), high salt buffer (0.6 M KCl, 1 mM EDTA, 2 mM ATP, 1 mM cysteine, 40 mM imidazole), and low salt buffer again. This KCl-extracted pellet was dissolved in 8 M urea in 10 mM Tris⅐HCl buffer supplied with protease inhibitors and subjected to ultracentrifugation at 125,000 ϫ g for 1 h at 4°C. The supernatant was dialyzed into 10 mM Tris⅐HCl and frozen at Ϫ80°C.
In Vitro Blot Overlay Binding Assays-In vitro overlay protein binding assays were performed as described elsewhere with slight modification (11). 2 g of purified keratin, bovine serum albumin, pinin amino end fragment 1-165, and mutant pinin 1-165 L8P were separated on a 10% SDS-PAGE. The proteins were then transferred to nitrocellulose membranes. Membranes were blocked by incubation in reaction buffer (10 mM Tris⅐HCl, 150 mM NaCl, 1 mM MgCl 2 , pH 7.4) with the addition of 0.1% (v/v) Tween 20 and 5% (w/v) nonfat milk powder at 4°C overnight. Blots were incubated 4 h at room temperature with the bacterially expressed pinin amino domain, either wild type GST-cp-(1-165) or mutant GST-cp-(1-165) L8P (3 g/ml in the reaction buffer with the addition of protease inhibitor mixture (Boehringer Mannheim) and 0.1% Tween 20, 1% bovine serum albumin, and 0.5% Triton X-100). After incubations, the blots were washed thoroughly with several fresh changes of the reaction buffer and subjected to routine Western blotting with anti-pinin antiserum and ECL (Amersham Pharmacia Biotech). Specifically, UF215 diluted 1:1000 in TBST (10 mM Tris⅐HCl, 150 mM NaCl, 1 mM MgCl 2 , 0.1% Tween 20, pH 7.4) was used as the primary antibody. Blots were incubated in 5% normal goat serum in TBS prior to secondary antibody (goat anti-rabbit IgG, Amersham Pharmacia Biotech; 1:10,000) incubation. As a control for the protein overlays, GST was used instead of wild type pinin fusion protein and subsequently probed with anti-GST antibody (Amersham Pharmacia Biotech) via Western blot.

RESULTS
K18, K8, and K19 Were Identified in a Yeast Two-hybrid Screening by the Amino Portion Fragment of Pinin-In an effort to identify proteins that bind to the amino-terminal domain of pinin, a yeast two-hybrid screening on a human fetal kidney cDNA library (CLONTECH) using pinin (residues 1-480) as bait was performed. Of the approximately 10 6 transformants screened, 21 independent cDNA clones were isolated. The recovered prey plasmids were verified by co-transforming each of the plasmids with either GAL4BD-pinin NЈ (residues 1-480) or control heterologous baits, including GAL4BD-p53, GAL4BD-pinin CЈ (residues 470 -717), and GAL4BD. All of the negative controls displayed no growth on the selective media, indicative that the interaction of the prey plasmid with the specific pinin bait resulted in the activation of the reporter gene requisite for growth. The most prevalent protein that exhibited binding to the amino end of pinin was keratin. Five of the identified clones encoded full-length keratin 18 (residues 1-430), one encoded the rod domain of keratin 8 (residues 90 -387), and another encoded the rod and the tail domain of keratin 19 (residues 69 -400) (Fig. 1).
The 2B Domain of Keratin Contains the Binding Site for Pinin-K18, K8, and K19 are three keratins expressed in the simple epithelial cells. These keratins share common structural properties. Each possesses an amino end nonhelical head domain, a central coiled-coil ␣-helical domain, and a nonhelical tail domain in various lengths (23,24). Because pinin (residues 1-480) bound equally well to each of these keratin clones and the common domain shared by all of the clones was the rod domain, we surmised that the rod domain might contain the sufficient sequence for the interaction with pinin. To further map the binding site within keratin, truncation constructs coding either coil 1 or coil 2 of K18/K8/K19 were generated and examined for their ability to bind pinin in a two-hybrid assay (Fig. 1). While constructs containing the coil 1 domain of K18 (residues 69 -240), K19 (residues 81-235), and K8 (residues 91-235) exhibited no significant binding to pinin (residues 1-480), the coil 2-containing constructs of K18 (residues 234 -391), K19 (residues 244 -390), and K8 (residues 260 -381) all exhibited interaction with pinin. It was, however, noticed that the coil 2-pinin interactions were approximately 10-fold weaker than the interaction of the intact keratin rod domain as indicated by the ␤-galactosidase assay. While reporter gene activity, such as ␤-galactosidase, does not correspond linearly with the strength of interaction, these assays can be useful in estimating relative strength of interactions between similar molecules or domains. The data suggest that either some sequence outside the coil 2 domain may contribute to the interaction or that the longer constructs may present the binding domain of keratin in a more advantageous conformation for pinin binding.
The carboxyl terminus of the 2B domain within coil 2 contains a highly conserved consensus motif, suggested to be significant for assembly/stabilization of the intermediate filaments in cells (25)(26)(27)(28). K18 (residues 69 -276), which excluded the entire 2B domain, failed to interact with pinin (residues 1-480). However, K18 (residues 69 -372), which contained the majority of the rod domain but not the consensus motif, retained the ability to bind to pinin (residues 1-480) (Fig. 2). Considering this, together with the results shown in Fig. 1   tins-The amino end of pinin (residues 1-480) contains a short domain with heptad repeats, a few glycine loops (29), and a rather extensive glutamate-rich ␣-helix domain (2). To more precisely map the domain of pinin that is sufficient for the interaction with keratin, five pinin truncation constructs were generated for two-hybrid analyses (Fig. 3). Constructs lacking the amino terminus of pinin (residues 85-480, 250 -480, and 85-252) exhibited no significant interaction with keratin, while constructs (residues 1-252 and residues 1-98) containing amino end heptad repeats and glycine loops exhibited binding to keratin. Leucine 8 and Leucine 19 within Pinin Are Essential for Binding to Keratin-To further define the specific region within the amino end of pinin that is essential for binding to keratin, site-directed mutagenesis was employed. Leucine residues at positions 8, 19, and 29, which were predicted to locate at either the "a" or "d" position of the heptad repeats within pinin (30 -32), were substituted with proline (NЈ L8P, NЈ L19P, and NЈ L29P). Interestingly, both NЈ L8P and NЈ L19P resulted in no growth at all on ϪAde medium (Fig. 4A) and a base line level of ␤-galactosidase activity (Fig. 4B), indicating the interaction between pinin and K18 was abolished with a single mutation. On the contrary, NЈ L29P retained the ability to grow on ϪAde medium, but the ␤-galactosidase activity was somewhat reduced. One glycine within the predicted first glycine loop of pinin was replaced by glutamate (NЈ G53Q). This substitution, similar to NЈ L29P, did not affect the growth of transformed yeast under selection conditions, but it resulted in a somewhat weaker interaction, as indicated by reduction in ␤-galactosidase activity. Charged residues have been speculated to stabilize coiled-coil conformations. However, changes of arginine 6 and lysine 28 to aspartate and glutamate, respectively (NЈ R6D, NЈ K28E), resulted in only a slight dampening in the ␤-galactosidase activity (Fig. 4). In summary, leucine 8 and 19 were shown to be essential for pinin-keratin interaction, whereas leucine 29, glycine 53, arginine 6, and lysine 28 were not essential but may somehow be involved in the optimal pinin-keratin interaction. Whether or not multiple (additive) substitutions of the residues would result in a more obvious effect on the pinin-keratin interaction is currently under investigation.
In Vitro Overlay Binding Assays Verified the Direct Interaction between Pinin Amino End Domain and Keratins-Keratin, purified from Madin-Darby canine kidney cells and bacterially expressed pinin fragments, both wild type GST-cp-(1-165) and mutant GST-cp-(1-165) L8P, were utilized in the blot overlay binding analyses. Blots containing keratin preparations were overlaid with either wild type pinin GST fusion protein GSTcp-(1-165) or mutated pinin GST fusion protein GST-cp-(1-165) L8P and subsequently reacted with UF 215 (Fig. 5B). Only the wild type pinin construct exhibited binding to keratin, as visualized by its immunoreactivity with anti-pinin antiserum UF215. The fact that the mutation L8P, which eliminated pinin-keratin binding in the two-hybrid assay, showed no binding in the overlay assays provides strong support for the specificity of the in vitro binding assay and confirmed the observations from the two-hybrid assays. We conclude that the amino end domain of pinin is capable of directly binding to keratin. DISCUSSION In this study, we present data demonstrating the direct interaction of the amino end domain of pinin with the 2B domain of keratin from simple epithelial cells. These data are not only consistent with our previous morphological observations but provide biochemical support of pinin-IF association.
There are four distinct coiled-coil stretches, 1A, 1B, 2A, and 2B in the central rod domain of a keratin molecule. Our data indicate that pinin binds to the sequence within the 2B domain of keratin. Coil 1 of keratin exhibited no binding to pinin, strongly supporting the conclusion that the interaction between the 2B domain of keratin and pinin amino-terminal domain is indeed specific and not due to nonspecific interaction with coiled-coil-containing proteins. Direct binding to the rod 2B domain of keratin 18 has been reported for BPAG 2, a hemidesmosome-associated protein (33). While desmoplakin has been shown to bind to the head domain of epidermal keratins, such as keratin 1/keratin 10 and keratin 5/keratin 14 (11), it has also been shown to be capable of binding to the rod domain of simple epithelial keratin K8/K18 heterodimer (13).
The truncation analyses suggested the amino end of pinin (residues 1-98) contained the sequence responsible for binding to keratin. Although short coiled-coils composed of four to five heptad repeats have been reported (31), it is not clear whether the four and a half heptad repeats at the amino end of pinin are actually sufficient to form a coiled-coil structure in vivo. The amino end of pinin does not contain a "trigger sequence" (34), so it may not participate in the formation of a coiled-coil. However, data derived from point mutation analyses of the aminoterminal domain of pinin suggest the sequence within the heptad repeats is indeed essential for the interaction with keratin. NЈ L8P and NЈ L19P completely abolished the binding of pinin to K18, whereas NЈ L29P did not, suggesting that the heptad repeats located nearer the amino end of pinin may play a more significant role in pinin-keratin interaction.
We have suggested that pinin may function as a tumor suppressor based on chromosomal location of pinin and tumor biological analyses (4). It has been shown that the expression of pinin was absent or greatly reduced in certain carcinomas, including renal cell carcinoma and transitional cell carcinoma. On the other hand, pinin expression was up-regulated in a subset of melanoma tissues (3) and a subset of renal cell carcinoma (4). Decreased expression of pinin was correlated with loss of epithelial cell-cell adhesion, while increasing pinin expression by transfection of pinin cDNA was shown to enhance cell-cell adhesion (4,35). Interestingly, K18 and K8 have long been considered as cytological markers for carcinomas due to their persistent expression in tumor cells derived from simple epithelia and their aberrant expression in malignant progression of nonepithelial cells (36 -39). In addition, several studies suggested that K18/K8 filaments have a role in the tumorigenicity. For example, in K8-deficient mice, adult animals developed pronounced colorectal hyperplasia (40), and the expression of K8 and K18 in human melanoma cell lines resulted in increased invasive and metastatic properties of the cells (37,41). It is tempting to speculate that the tumor suppressor function of pinin is related to the interaction of pinin with keratin.
This study did not address the important issue regarding the relationship between the desmosome and pinin. Our initial two-hybrid screens identified other, as of yet uncharacterized, proteins interacting with pinin NЈ bait 1-480. 2 One of these clones coded for a protein containing motifs highly homologous to periplakin (18), a desmosome-IF-associated protein forming cornified envelope in the stratified epithelial cells. The possibility of pinin connecting to desmosome through this periplakin-like protein is currently being addressed.
In summary, we have demonstrated that pinin can bind to keratin 18, keratin 8, and keratin 19. The 2B domain of keratin contains the sequence mediating the interaction with pinin, and the amino end (residues 1-98) of pinin was sufficient to bind keratin. Identification of specific binding sites within pinin for keratin and for other proteins will be an integral step for future studies. We believe that investigation of the function(s) of pinin in cell adhesion and IF organization will greatly contribute to our current knowledge of epithelial cell-cell adhesion.