Isolation of a Novel β4 Integrin-binding Protein (p27BBP) Highly Expressed in Epithelial Cells*

The integrin β4 has a long cytodomain necessary for hemidesmosome formation. A yeast two-hybrid screen using β4 cytodomain uncovered a protein called p27BBP that represents a β4 interactor. Both in yeast and in vitro, p27BBP binds the two NH2-terminal fibronectin type III modules of β4, a region required for signaling and hemidesmosome formation. Sequence analysis of p27BBP revealed that p27BBP was not previously known and has no homology with any isolated mammalian protein, but 85% identical to a yeast gene product of unknown function. Expression studies by Northern analysis and in situ hybridization showed that, in vivo, p27BBP mRNA is highly expressed in epithelia and proliferating embryonic epithelial cells. An antibody raised against p27BBP COOH-terminal domain showed that all β4-containing epithelial cell lines expressed p27BBP. The p27BBP protein is insoluble and present in the intermediate filament pool. Furthermore, subcellular fractionation indicated the presence of p27BBP both in the cytoplasm and in the nucleus. Confocal analysis of cultured cells showed that part of p27BBP immunoreactivity was both nuclear and in the membrane closely apposed to β4. These results suggest that the p27BBP is an in vivointeractor of β4, possibly linking β4 to the intermediate filament cytoskeleton.

Interactions between basal lamina and cells are important in several phenomena including differentiation (1) and tumor progression (2). Integrins belong to a family of adhesion receptors expressed in most tissues. They are glycoprotein heterodimers formed by the noncovalent association of two subunits named ␣ and ␤ (3). Interference with integrin signaling is associated with a variety of effects, including regulation of gene expression and mitotic progression (reviewed in Ref. 4).
The integrin subunit ␤ 4 associates with ␣ 6 to form a multivalent laminin receptor (5)(6)(7). High levels of ␤ 4 are found in most epithelia, in Schwann cells, and in several tumors of epithelial origin (reviewed in Refs. 8 and 9). The amino acid sequence of ␤ 4 cytoplasmic domain is 1045 amino acids long and not homologous to any other ␤ subunits, thus suggesting that ␤ 4 has different cytoskeletal links and/or is coupled to a specific transduction pathway (10 -13). In agreement with this, in squamous and transitional epithelia, ␤ 4 is highly enriched in hemidesmosomes, specialized structures providing firm mechanical links between basal lamina and the intermediate filament cytoskeleton (14 -16). Loss of function of ␤ 4

both in human and in ␤ 4
Ϫ/Ϫ mice results in hemidesmosome disruption, blistering, and is lethal perinatally (17)(18)(19). The adhesive function of ␤ 4 in hemidesmosome-free epithelia or Schwann cells is still largely obscure, as is its function in epithelial neoplasms.
Several lines of evidence indicate that the ␤ 4 cytodomain mediates its function through the association with unknown cytoplasmic ligands. In particular, several cytodomain deletions impair the ability of ␤ 4 to translocate into hemidesmosomes, and this involves also the absence of other hemidesmosomal components (17, 20 -22). In vitro mutagenesis has restricted to a 303-amino acid stretch the region of ␤ 4 necessary for its translocation to hemidesmosomes (20), and, within this 303-amino acid region, at least one mutation that leads to loss of function, in vivo, has been mapped (17). In cells that do not form hemidesmosomes, the transfection of the wild type portion of the cytosolic region of ␤ 4 results in partial arrest in G 1 and apoptosis, an effect that is not observed when a ␤ 4 mutant deleted in its cytosolic domain is transfected (23). Finally, following antibody ligation ␤ 4 may recruit the Shc and Grb2 interactor molecules (21).
Taken together, these results strongly suggest that this part of ␤ 4 associates with unknown cytosolic ligands to signal and be targeted to hemidesmosomes. Despite this, their identification has been difficult due to the relative insolubility of ␤ 4 and of hemidesmosomal associated proteins. To bypass this problem, we have decided to isolate interactors of the ␤ 4 subunit by the yeast two-hybrid system; we have constructed a vector encompassing the critical region of ␤ 4 and screened an epithelial cDNA library. We have found that ␤ 4 interacts with a previously unknown peptide, named p27 BBP , that is segregated in two distinct pools: cytoplasmic and nuclear. Interestingly, p27 BBP is highly expressed, in vivo and in vitro, in epithelial cells containing ␤ 4 , and is associated with the intermediate filaments fraction. These data suggest a role for p27 BBP in linking ␤ 4 to the intermediate filament cytoskeleton.

EXPERIMENTAL PROCEDURES
Two-hybrid Screen-All the procedures employed are described in detail in Golemis et al. (24). LexA-␤4 and LexA-trkB fusion proteins * This work was supported by Telethon Grants 762 (to S. B.) and 826 (to P. C. M.) and also by Consiglio Nazionale delle Ricerche target project Applicazion: Cliniche della Ricerca Oncologica (ACRO) Associazione Italiana Ricerca sul Cancro (AIRC). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  were prepared by subcloning PCR 1 fragments of ␤ 4 and of trkB in the plasmid pEG202 (25). Following subcloning, the inserts were sequenced. All the plasmids were transfected in the EGY48 yeast strain to check for their ability to self-activate the reporter genes LEU2 and lacZ, and to enter the nucleus. To perform the interaction hunt a HeLa cDNA library (obtained from Roger Brent, Harvard Medical School) was cotransfected with the bait construct encompassing the first two fibronectin domains (FNIII) and the connecting sequence. A total of 1,350,000 independent colonies were obtained from the primary transformation. The selection for interacting clones was performed in media containing galactose and lacking leucine. Surviving yeast colonies were further selected by plating them on 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside/galactose-containing medium. Criteria for the subsequent selection of the clones are described under "Results." Plasmid cDNAs were rescued in the tryptophan-deficient bacteria strain KC8 and sequenced by dye terminator technique on an automated Perkin-Elmer sequencer.
GST Fusion Protein Preparation, in Vitro Translation and in Vitro Binding Assay-The GST fusion proteins were prepared by excising from the pEG202 vector the coding regions of ␤4 and of trkB with EcoRI and NotI and subcloning them in the pGex4T1 vector (Pharmacia Biotech Inc.). All fusion proteins were prepared in the protease minus strain BL21. GST fusion proteins were purified on glutathione-Sepharose 4B beads according to the manufacturer's protocol. Before the interaction experiments, the amount of the fusion protein was determined by the Bradford protein quantitation assay (Pierce).
In vitro translated p27 BBP was prepared by transcribing the fulllength p27 BBP mRNA subcloned in the Bluescript vector with the Cap-Scribe kit (Boehringer Mannheim). The translation was subsequently performed with the wheat germ agglutinin extract from Promega, in the presence of radioactive methionine (Amersham Corp.). Conditions were as suggested by the manufacturer except for the final concentration of potassium acetate (40 mM).
The in vitro binding assay was performed as described by Swaffield and Johnston (26). In the assay the desired amount of GST fusion protein was mixed with in vitro translated p27 BBP and incubated at 4°C for 2 h. The interaction buffer included an Escherichia coli protein extract (10 mg/ml) in 50 mM potassium phosphate buffer, pH 7.5, 150 mM KCl, 1 mM MgCl 2 , 10% glycerol, 1% Triton X-100, plus protease inhibitors. GST fusion protein complexes were pooled down with glutathione beads and washed five times in the same buffer at 4°C. The reaction products were solubilized in Laemmli buffer and run in 12% denaturing SDS-acrylamide gel. The gels were dried and exposed for 5 days.
Screening of a -Zap Library, Sequence Analysis, and Cloning of Mouse Probes-To obtain full-length clones of p27 BBP , a human placental cDNA library (Stratagene) was screened using a 700-nucleotide probe spanning its COOH-terminal sequence. The screening was performed according to the manufacturer's instructions. One-million plaques were screened, and double-positive plaques were identified. Plasmid DNA was recovered by in vivo excision and analyzed by restriction analysis. The two largest clones were sequenced in both strands. Sequence analysis was performed through Wisconsin Package Version 9.0, Genetics Computer Group (GCG), Madison, WI and through the sources present in the Internet (the main servers were: "pedro," the Baylor College of Medicine, the Flybase data base, the EBI, and the NCBI).
To clone the mouse homolog, primers spanning the most conserved domain of human p27 BBP and the EST clone F15081 (from Sus scrofa) were used to amplify cDNA from mouse tissues. Primer sequences are: Scro, 5Ј TCAGGATCCGAGCTGCAACACATTCGCA and Rat, 3Ј TGG-GAATTCATTCAGCTTGAACACACTCTC. Following reverse transcription-PCR, a 400-base pair fragment was obtained, sequenced, and subcloned into the TA vector (Invitrogen). The accession number of the mouse partial sequence is Y11460.
Southern Blot Analysis-High molecular weight DNA was extracted following standard procedures (27). Ten g of human placenta DNA were digested with restriction enzymes (Promega), electrophoresed on 0.8% agarose gel, denatured in alkali, neutralized, and transferred onto nylon membrane (Hybond-N; Amersham). The membrane was hybridized with random prime-labeled p27 BBP cDNA at 42°C for 16 h in 50% formamide, 5 ϫ SSC, 5 ϫ Denhardt's solution, 0,5% SDS, and 20 g/ml denatured sonicated salmon sperm DNA. After hybridization, the mem-brane was washed once at room temperature in 2 ϫ SSC, 0.1% SDS for 20 min. For low stringency washes, the filter was treated two times at 42°C in 2 ϫ SSC, 0.1% SDS for 20 min each time. For high stringency washing the filter was treated two times at 42°C in 2 ϫ SSC, 0.1% SDS for 20 min each time and in 1 ϫ SSC, 0.1% SDS at 42, 50, 55, and 60°C for 20 min each time (28). Autoradiography was performed using Hyperfilm (Amersham) films at Ϫ70°C for 8 -16 h with intensifying screens.
Northern Blot Analysis-Total RNA from cell lines was extracted with the RNAzolB kit (Cinna/Biotecx Laboratories). Total RNA from tissues were extracted by the guanidium thiocyanate phenol-chloroform method (29). Total RNA was reconstituted in diethyl pyrocarbonatetreated water and quantified by optical density measurement and ethidium bromide staining. Ten g of total RNA were electrophoresed on denaturing formaldehyde-agarose gels (28), transferred to Hybond-N filters, and processed according to the manufacturer's instructions. Hybridization was carried with 1 ϫ 10 6 cpm of homologous random primer-labeled probe in 50% formamide at 42°C. Conditions for hybridization and washing were as described previously (30). Filters were exposed for 1 week at Ϫ70°C with intensifying screens. The filter of murine tissues mRNAs was reprobed with a glucose-6-phosphate dehydrogenase cDNA probe, and densitometric analysis was performed to normalize the results.
In Situ Hybridization-Tissues from adult mice were fixed by perfusion in 4% paraformaldehyde. Mice embryos were fixed by immersion. Tissues were dissected out, postfixed in 4% paraformaldehyde overnight, and cryoprotected in ascending sucrose solutions. Following freezing in OCT, 10-m sections were cryostat-cut and mounted on L-polylysine-coated slides. Hybridization was performed with a 35 S in vitro translated species-specific antisense riboprobes (400 nucleotides long) or with a control sense probe. High stringency hybridization and washes were carried as described previously (31). Sections were dipped to NTB-2 emulsion (Eastman Kodak Co.), exposed 2 weeks, developed with Kodak D19, and counterstained with hematoxylin.
Antibodies and Immunofluorescence-A rabbit polyclonal antiserum against the COOH-terminal peptide of p27 BBP (NH 2 -CTIATSMRD-SLIDSLT-COOH) was prepared by Eurogentec and tested for its specificity by Western blotting and immunoprecipitation both on recombinant in vitro translated p27 BBP and on cellular lysates. Integrin ␤ 4 was detected with mouse monoclonal antibody 450-11A 10 g/ml (a generous gift of Rita Falcioni, Istituto Regina Elena, Rome). Secondary antibodies were rhodamine-tagged swine anti-rabbit IgGs (1/50, Dako) and fluorescein-tagged goat anti-mouse IgGs (1/50, Antibodies, Inc.). In control experiments, primary antibodies were replaced by non-immune sera or irrelevant monoclonal antibodies. In addition, the p27 BBP antiserum was preadsorbed with the peptide used for its generation (1 M, overnight, 4°C). The immunofluorescence technique was as reported previously (16). Briefly, cell monolayers were fixed in 3% paraformaldeheyde, PBS, pH 7.6, 2% sucrose for 5 min at room temperature, permeabilized (3 min, 4°C in 20 mM Hepes, 300 mM sucrose, 50 mM NaCl, 3 mM MgCl 2 , 0.5% Triton X-100, pH 7.4), and sequentially treated with blocking solution, primary and secondary antibodies (all diluted in PBS). Following washes, staining for F-actin was performed with fluorescein-labeled phalloidin (Sigma, 200 nM for 20 min at 37°C in the dark). The coverslips were analyzed with a confocal laser scanning microscope (CLSM Bio-Rad 1024). Image files were recorded on different channels and digitally reconstructed to provide z axis views.
Fractionation, Immunoprecipitation, and Western Blot Analysis-Total extracts were prepared in boiling SDS (32). Immunoprecipitation from total extracts was performed diluting 1:10 the SDS extract in PBS followed by incubation with the p27 BBP antiserum, 1:1000, 2 h at 4°C and subsequently with protein A-Sepharose, 30 min at 4°C. Immunoprecipitates were washed five times in PBS, 1% Triton X-100 and loaded on denaturing 12% SDS-acrylamide gel. Cytosolic and nuclear extracts were run on denaturing 12% SDS-acrylamide gel (34) and transferred to Immobilon-P membranes (Millipore). The membranes were blocked with 5% bovine serum albumin in PBS and treated with p27 BBP antiserum used at 1:1000 dilution. Following washes and incubation with peroxidase-labeled anti-rabbit antibodies, the blots were developed by the chemiluminescence technique (Amersham).
Intermediate filaments/nuclear core filaments fractions were prepared exactly according to He (33). Briefly, all the soluble proteins, the non-intermediate filament cytoskeleton, DNA-associated proteins, and nuclear matrix proteins were removed by sequential washes in buffers containing Triton X-100, 250 mM ammonium sulfate, DNase I, and 2 M NaCl. At the end of this procedure, cells are composed from a cytoplasmic and nuclear intermediate filament network containing keratins, lamins, and other partially characterized intermediate filament associated proteins. The fraction was controlled by blotting with anti-keratin antibodies (mouse monoclonal from Becton Dickinson, number 7650) and electron microscopy (courtesy of A. Villa (DIBIT, Milano, Italy), not shown).
Subcellular fractionation for nuclei and cytoplasm was performed according to Robbins et al. (34) with slight modifications. Epithelial cells (FG2 human pancreatic or A431 carcinoma cells) cultured in standard medium were washed three times with PBS and lysed by incubation in RSB lysis buffer (10 mM Tris, pH 7.4, 10 mM NaCl, 3 mM MgCl 2 , 1 mM EGTA, 10 mM sodium phosphate ϩ protease inhibitors) at 4°C. Cells were homogenized and subjected to five cycles of centrifugation and washes with RSB (375 g). The pellet (nuclear fraction) was resuspended, sonicated in lysis buffer containing Triton X-100, and cleared by centrifugation. The technique used for fractionation, as detected with specific enzyme markers, yields pure nuclei and 95% pure cytosolic fractions.

RESULTS
Interactors of ␤ 4 Integrin-Mutational analysis has shown that a 303-amino acid segment encompassing the first two NH 2 -terminal FNIII domains and the connecting sequence is necessary for mediating (i) ␤ 4 incorporation into hemidesmosomes and (ii) signaling events (20,21). To identify polypeptides that interact with the ␤ 4 cytodomain, we made use of the yeast two-hybrid system. A 331-amino acid construct encom-passing the first two FNIII domains and the connecting sequence was joined to the DNA binding portion of the LexA repressor (␤ 4 -LexA) and transfected into the EGY48 yeast strain (Fig. 1A). The construct-encoded protein was able to enter the nucleus and bind to LexA target sequences without showing transactivating activity. This construct containing the functional region of ␤ 4 was used for the interaction hunt. Interestingly, most of the control constructs spanning the ␤ 4 FNIII domain and the connecting sequence, when transported to the nucleus, were strong transcriptional activators per se and could not be used for any subsequent analysis in the yeast two-hybrid system.
To perform the interaction hunt, ␤ 4 -LexA was cotransfected with a translational fusion cDNA library from HeLa cells. HeLa cells express the integrin ␤ 4 but do not normally assemble hemidesmosomes (not shown). Upon selection of 1-million independent yeast colonies in galactose-containing leucine-minus medium, approximately 300 colonies grew within 72 h after plating. Thus, the 331-amino acid domain of ␤ 4 , containing the FNIII domains, can interact with several proteins in the yeast two-hybrid system. Since published data indicated that a tyrosine kinase may be stably associated with ␤ 4 (21), we first investigated whether a tyrosine kinase was present among our interactors. All the plasmid cDNAs were rescued from the hunt, and PCR with degenerate primers for tyrosine kinases was performed: no evidence for tyrosine kinases associating with ␤ 4 was found (not shown). To select for specific interactors of the functional domain of ␤ 4 , we then developed a test based on the screening of the ability of all the rescued plasmids to interact specifically with this region. All cDNA interactor plasmids were cotransfected either with the original bait or with a construct coding for the last two FNIII domains of ␤4 (that despite their high homology with the first two FNIII domains have been shown to be dispensable for ␤ 4 signaling and its incorporation into hemidesmosomes; Refs. 17, 20, and 21) or with a control construct coding for trkB-LexA (trkB ϭ cytodomain of truncated tyrosine kinase trkB). Most of the interactors bound to all FNIII domains of ␤ 4 , but not trkB-LexA. Around 50 clones interacted selectively with the first two FNIII domains and the connecting sequence of ␤ 4 cytodomain, but neither with the last two FNIII domains nor with trkB. Southern blot analysis indicated that these clones were duplicates of the same cDNAs. They were sequenced and found to encode for a novel protein (named p27 BBP , BBP ϭ ␤ 4 -binding protein) that was further characterized.
p27 BBP Interacts Specifically with the First Two FNIII Domains and the Connecting Sequence of ␤ 4 Both in the Yeast System and in Vitro-p27 BBP was isolated from two independent interaction hunts with ␤ 4 , at an average frequency of 41 clones per million screened. In four independent retransfection experiments on yeast, p27 BBP was found to interact exclusively with the first two FNIII domains and the connecting sequence of ␤ 4 , but not with other part of the ␤ 4 molecule, including the last two FNIII domains or with control molecules (trkB-LexA fusion protein Fig. 1B). Sequence analysis of the shortest plas-mid coding for p27 BBP indicated that the COOH-terminal 135amino acid sequence was necessary and sufficient for the interaction with ␤ 4 .
The biochemical interaction between p27 BBP and ␤ 4 was then studied in an in vitro system. GST fusion proteins of ␤ 4 and control GST were prepared and mixed in a liquid phase assay with in vitro translated full-length p27 BBP (see below). In vitro translated p27 BBP was found to bind a ␤ 4 fusion protein containing the first two FNIII domains and the connecting sequence (i.e. the bait), but neither GST control proteins, glutathione-agarose beads (Fig. 1C), or other parts of the ␤ 4 molecule (not shown). Approximately 15% of radioactive input p27 BBP protein bound ␤ 4 . Binding between ␤ 4 and p27 BBP occurred rapidly (within minutes) and was resistant to high salt washing. The relative low levels (15%) of binding between ␤ 4 and p27 BBP in the in vitro assay were due to the strong instability of the ␤ 4 fusion protein, as short term storage at 4°C of ␤ 4 fusion protein resulted in rapid loss of binding (compare lanes 3 and 4 of Fig. 1C). This phenomenon has been recently observed for other ␤ 4 fusion proteins both in vivo and in other bacterial strains, and it has been suggested to be linked to the ability of ␤ 4 to bind calpain-like proteases (35).
Taken together, these data indicate that p27 BBP can directly bind the ␤ 4 cytodomain in vitro.
p27 BBP Is a Novel Evolutionary Conserved Protein Encoded by a Single Gene-The partial sequence of p27 BBP lacked an ATG start codon and was not homologous to any known nucleotide sequence. To isolate full-length clones of p27 BBP , a human placenta -Zap library was screened. From one-million clones, 50 positive plaques were rescued by in vivo excision and ana- FIG. 2. A, sequence of p27 BBP protein and alignment with the Drosophila and yeast homologs. The predicted amino acid sequence of p27 BBP derived from the cDNA isolated from the placental library is shown. Alignment of human, yeast, and Drosophila sequences is shown. Accession numbers are Y11435 for human, X97641 for Drosophila (accession number X97641), and Z49919 for yeast. The arrow points to the beginning of the shortest clone interacting with ␤ 4 . The boxed area represents the regions of homology. B, genomic Southern analysis of the human p27 BBP gene. Placental genomic DNA was digested with SacI (S), PstI (P), EcoRI (E), HindIII (H), and BamHI (B) and subjected to 0.8% agarose gel electrophoresis. The hybridization was carried out as described under "Experimental Procedures." The molecular weight markers were phage HindIII.
lyzed by restriction analysis. The size of the largest cDNA (1.1 kilobase pairs) was identical to the size of human p27 BBP mRNA (not shown and Fig. 3), contained a putative ATG initiation codon (36) and a 735-nucleotide open reading frame. The open reading frame predicts a 245-amino acid protein (expected molecular mass 27 kDa) with no signal peptides. The protein sequence was analyzed with DNA analysis programs (BLOCKS, MOTIFS, and PRODOME) without revealing any motif or homology to characterized proteins. Strikingly, FASTA and BLAST analyses identified homologies with several EST clones and with two genes encoding putative proteins in yeast and Drosophila (Fig. 2A). Alignment of these three proteins revealed more than 75% identity and up to 90% similarity.
The presence of a gene coding for p27 BBP in yeast and Drosophila suggested the possibility that duplication events might have led to homologous genes in mammals. To test this possibility, Southern blot analysis of human DNA was performed at high (Fig. 2B) and low stringency (data not shown). Human genomic DNA digested with SacI, PstI, EcoRI, HindIII, and BamHI was subjected to Southern hybridization analysis using the p27 BBP cDNA as a probe. Surprisingly, the results were consistent with a single-copy gene.
p27 BBP Is a Protein Highly Expressed in Epithelial Cells That Contain ␤ 4 -Preliminary results showed that p27 BBP mRNA is unstable, and expressed in several epithelial human tissues, with high levels in the colon (not shown). To extend the expression data to an experimental model, the mouse homolog of the human p27 BBP was cloned by reverse transcription-PCR using degenerate primers and used for analysis in adult mouse organs. Specifically, we concentrated on p27 BBP mRNA expression in epithelial tissues containing ␤ 4 (all epithelial tissues, plus ovary and tongue) as compared with tissues that do not contain ␤ 4 integrin (brain, skeletal muscle, hearth, kidney, testis, spleen). In adult mice, Northern blot analysis showed detectable levels of p27 BBP mRNA in all tissues analyzed. Among these tissues, some displayed remarkably high levels of p27 BBP mRNA. These included duodenum, colon, tongue, and lungs (Fig. 3A), that express high levels of ␤ 4 integrin in their epithelial lining (Refs. 37 and 38 and not shown). Interestingly, high levels of p27 BBP mRNA were also detected in testis that has not been reported to contain ␤ 4 integrin.
p27 BBP mRNA expression was also examined in human epithelial cell lines. Northern blot analysis of p27 BBP mRNA in human cell lines revealed high levels of expression in the epithelial cell line A431 and lower levels in HeLa and FG2 cell lines (Fig. 3B). All of these three cell lines contain high levels of ␤ 4 mRNA and protein (not shown and Fig. 5). Steady state p27 BBP mRNA level in epithelial cells was higher in subconfluent growing cells than in confluent cells (Fig. 3B). High expression levels were also seen in activated T cells (Fig. 3B).

FIG. 3. Expression of p27 BBP mRNA in mouse tissues and cell lines.
A, Northern blot of 10 g of total RNA extracted from mouse tissues. pla ϭ placenta; ova ϭ ovaries; bra ϭ brain; ton ϭ tongue; ski ϭ skin; mus ϭ muscle; lun ϭ lungs; hea ϭ heart; kid ϭ kidney; tes ϭ testis; spl ϭ spleen; col ϭ colon; duo ϭ duodenum. This experiment is representative of four independent analyses. p27 BBP mRNA levels in the different tissues have been normalized reprobing the filter with a glucose-6-phosphate dehydrogenase cDNA probe (graphic). p27 BBP mRNA has an apparent size of 1.1 kilobase pairs (arrow). B, Northern blot of 10 g of total RNA extracted from nonconfluent epithelial cell lines A431, FG2, and HeLa and from the activated T lymphocytes. Northern blot of 10 g of total RNA extracted from nonconfluent epithelial cell lines A431 (A431 subconf) and confluent overgrown A431 cells (A431 conf). The ethidium bromide staining of the nylon sheet is shown for comparison.

FIG. 4. p27 BBP mRNA is expressed in embryonic epithelial cells and in the proliferative compartment of the adult colon.
In situ hybridization of p27 BBP mRNA in the developing embryo (A, B) and in the adult mouse (C, D). In the embryonic mouse p27 BBP mRNA is highly expressed in the skin (Fig. 5A, upper panel, dark field and 5B, bright field) and in the gut (Fig. 5A, lower panel, dark field). In the adult mouse p27 BBP mRNA is observed at low levels in the crypts of the gut (Fig. 5C, bright field). No labeling is seen using a sense probe (Fig. 5D). The scale bar is 40 m for B, C, and D and 110 m for A.
To detect the cellular localization of p27 BBP mRNA, in situ hybridization studies were carried out, with particular attention to embryonic epithelial tissues expressing ␤ 4 . In the developing mouse, p27 BBP mRNA was readily detected in developing epithelia, notably in the skin (Fig. 4A, upper) and in the primitive gut (Fig. 4A, lower). No labeling was seen using a control sense probe (not shown and Fig. 4D). In the embryos, p27 BBP mRNA was never restricted to the basal layer. By in situ hybridization, most murine adult tissues showed undetectable levels of p27 BBP mRNA. In the adult, p27 BBP mRNA was detected in skin keratinocytes and in gut epithelium (Fig. 4, C  and D, and not shown). In these adult tissues, p27 BBP mRNA was either mainly basal, such as in the epidermis, or concentrated in the cells lining intestinal crypts.
Taken together the expression data show that p27 BBP mRNA is always expressed in all ␤ 4 -containing epithelial tissues and cell lines and seems enriched in proliferating epithelia, such as in the embryonic skin. In addition, p27 BBP mRNA is also present at lower, but detectable, levels in tissues that have not been reported to contain ␤ 4 .
The p27 BBP Protein Is Highly Insoluble, Associated with the Intermediate Filament Cytoskeleton Both in the Cytoplasm and in the Nucleus-To analyze the biochemical properties of p27 BBP , several approaches were tried. First, we tried to express the recombinant protein in heterologous systems such as E. coli; so far we have been unable to produce soluble p27 BBP . Second, a rabbit polyclonal antiserum was raised against the COOH-terminal peptide of p27 BBP and used to study the distribution of endogenous p27 BBP . In agreement with the predicted size of the protein based on the cDNA sequence, by Western blot analysis this antiserum recognized a 27-kDa band in total lysates of wild type epithelial cells. Interestingly, in epithelial cells, a strong band was observed with as little as 10 g of total extracts prepared in boiling SDS. This antiserum could also quantitatively immunoprecipitate p27 BBP from SDSsolubilized epithelial cells extracts (Fig. 5A). However, no immunoprecipitation of p27 BBP was seen using 1 milligram of protein extracted with mild detergents such as Brij 96, and only partial recovery was seen with 1 milligram of extracts prepared with 1% Triton X-100 (not shown). We then tried to establish whether p27 BBP could associate, in vivo, with ␤ 4 . Western blot analysis of total extracts of A431 or FG2 epithelial cell lines analyzed with either the anti-p27 BBP or an anti-␤4 indicated that they expressed high levels of both p27 BBP and of ␤ 4 (Fig. 5, B and C). Coimmunoprecipitation of ␤ 4 and p27 BBP from FG2 and A431 epithelial cell lines was attempted by immunoprecipitating with a mouse monoclonal antibody directed against the extracellular domain of ␤ 4 , in detergent conditions that partially solubilize ␤ 4 and/or p27 BBP , but may preserve protein-protein interaction (low SDS concentrations, Triton X-100, and Brij 96) followed by Western blot analysis with anti p27 BBP . Under these conditions no association could be detected.
The high insolubility profile of p27 BBP suggested that this protein could be part of the intermediate filament cytoskeleton, a relevant hypothesis, since ␤ 4 is ultrastructurally linked to intermediate filaments (17,18,19,35), but so far no molecule that directly binds ␤ 4 and is associated with intermediate filaments has been found. We then performed fractionation studies on epithelial cells using a protocol that strips cells of all cellular components and leaves only an intermediate filament fraction that comprises both cytoplasmic intermediate filaments and the nuclear core filaments (33). Strikingly, p27 BBP was found to be highly enriched in the intermediate filament fraction (Fig. 5D, lane 5) and to be absent from all the other insoluble fractions (Fig. 5D, lanes 2-4). In agreement with immunoprecipitation studies, some of p27 BBP was present in the Triton X-100-soluble fraction (Fig. 5D, lane 1), thus indicating that this protein exist in an equilibrium between a soluble and a highly insoluble form.
Since the intermediate/nuclear core filament fractionation method leads to the enrichment of both cytoplasmic filaments, composed in epithelial cells mainly of keratins, and of nuclear core filaments, whose biochemical composition is largely unknown, we then performed fractionation studies between the nucleus and cytoplasm. Strikingly, p27 BBP was found present both in the nuclear and in the cytosolic insoluble fractions (Fig.  5E). The double localization of p27 BBP in the nucleus and in the cytoplasm was not due to cross-contamination of the two fractions, as shown by analysis with specific markers (not shown) and immunofluorescence studies ( Fig. 6 and not shown).
To see the fine localization of p27 BBP , we analyzed by immunofluorescence and CLSM its expression pattern in epithelial cell lines. p27 BBP immunoreactivity was found in the cytoplasm and in the nucleus and could be abolished completely by preadsorbing the antiserum with the peptide (Fig. 6B and not  shown). Cytoplasmic associated p27 BBP immunoreactivity ap- Above, the same blot was reacted with an anti-keratin antibody (see "Experimental Procedures") recognizing cytokeratins (8, 18, and 19 from the Moll's Catalog, molecular masses ranging from 39,000 to 50,000 daltons) as a control for the extraction procedure (E), p27 BBP is present both in the cytoplasmic fraction (SDS-solubilized extracts, lane 1) and in the nuclear fraction (lane 2). Fractions were prepared and controlled for their purity as described under "Experimental Procedures." The blot was decorated with p27 BBP antiserum. peared weaker in interphase cells than during mitosis (not shown). In addition, in human epithelial cells CLSM reconstruction of digitized images showed that p27 BBP was present at the immediate submembranous surface in a distribution consistent with ␤ 4 (Fig. 6, A and B) and just adjacent to the subcortical cytoskeleton.

DISCUSSION
Through an extensive yeast two-hybrid screening, we have isolated p27 BBP , an interactor of the ␤ 4 integrin. By means of two assays, i.e.the yeast two-hybrid system and in vitro binding, we have shown that p27 BBP binds to the region of the ␤ 4 molecule necessary for signaling (17,20,21,23). In addition, both in vivo and in vitro, p27 BBP has been found to be highly expressed in epithelial cells that contain ␤ 4 . In epithelial cells, p27 BBP is localized in the cytoplasm, mainly in the intermediate filament fraction, and at the submembrane level in close apposition with ␤ 4 . Taken together these data suggest that p27 BBP may be a novel cytoplasmic interactor of ␤ 4 integrin, possibly linking it to the intermediate filament cytoskeleton.
In addition, various data indicate that p27 BBP has also a ␤ 4 independent function: (i) part of p27 BBP is found in the nucleus, where to our knowledge ␤ 4 has not been detected; (ii) the p27 BBP protein is highly conserved throughout evolution being expressed also in yeast, where no ␤ 4 homolog has been yet described; (iii) expression data show that p27 BBP is also expressed in tissues where ␤ 4 has not yet been detected.
Cytoplasmic Interactors of ␤ 4 and the Possible Role of p27 BBP -Our paper provides the first evidence of a cytoplasmic molecule binding directly ␤ 4 integrin. The possibility that p27 BBP and ␤ 4 interact in vivo is supported by four lines of evidence: the specific association in the yeast two-hybrid assay, the direct association in the in vitro GST pulldown assay, the coexpression of p27 BBP and ␤ 4 in epithelial cells in vivo, and the presence of p27 BBP immunoreactivity in epithelial cells at the submembrane level. It is important to stress that the binding of p27 BBP to ␤ 4 occurs exclusively in the domain of ␤ 4 that has been shown in several contexts, in vitro and in vivo, to be necessary for its function (7,17,20,21,23). However, so far, we were unable to coprecipitate ␤ 4 and p27 BBP , and although preliminary experiments on cells transfected with p27 BBP seem to indicate a redistribution of ␤ 4 , 2 the relevance of this finding for the in vivo situation is still to be pondered. It is likely that the inability to coprecipitate ␤ 4 and p27 BBP is due to the detergents needed to solubilize the cytoskeletal associated fraction of p27 BBP and that, under these conditions, the interaction between the two proteins is lost.
The isolation of ␤ 4 -binding proteins has been a difficult task, despite the overwhelming evidence that its long cytoplasmic tail is involved in signal transduction and in hemidesmosome formation by interacting with unknown cytosolic proteins (17,20,21,23). This is due both to the difficulty to solubilize ␤ 4 and cytoskeletal proteins and to the different biological models employed. A further step in complexity is that, although ␤ 4 is necessary for hemidesmosome formation, it is mostly expressed by epithelial cells that do not form hemidesmosomes, such as in the gut. At the present time, it is unknown whether ␤ 4 binds the same cytoskeletal linkers in hemidesmosome-forming epithelial cells and in other epithelial cells, although the fact that mutations in the cytoplasmic domain of ␤ 4 result in epithelial detachment both in skin and in the pylorus suggests the existence of a common cytoplasmic linker (17). Due to the fact that p27 BBP seems to be tightly associated with the intermediate filament cytoskeleton, its interaction with ␤ 4 may provide the long sougth link between this integrin and the intermediate filaments both in epithelial cells that form hemidesmosomes as well in epithelial cells that do not form hemidesmosomes.
Circumstantial evidence, based on the available literature, would predict two types of molecules binding ␤ 4 : a tyrosine kinase (21) and an hemidesmosome/intermediate filament associated protein (13,20). In neither case, identification of these partners and evidence of direct association was so far obtained. The evidence of a tyrosine kinase associated with ␤ 4 derived from immunoprecipitation with mild detergents followed by kinase assay and led to the model that a tyrosine kinase, not yet identified is intimately associated with ␤ 4 (21). In this context, it is of interest to note that although our bait contained the motif of ␤ 4 phosphorylated by the kinase, no tyrosine kinase could be identified in our two-hybrid screening. The possible explanations are that either ␤ 4 does not associate directly with a kinase, that other parts of ␤ 4 molecule are involved in the binding of the kinase, or that a posttranslational modification of ␤ 4 is required to observe kinase association. The association of ␤ 4 with cytoskeletal proteins is suggested by the fact that a deletion in the cytoplasmic first two FNIII domains causes the lack of ␤ 4 association with hemidesmosomes and perturbs intermediate filaments organization (17,20). Four other major proteins of the hemidesmosome have been described: BPAG1, BPAG2, HD-1, and ␣ 6 . ␣ 6 directly associates with ␤ 4 to form a laminin receptor through a juxtamembrane domain not overlapping with our bait and thus explaining why no ␣ 6 was isolated in our screening. Interestingly, using the cytoplasmic domain of ␤ 4 , containing the first two FNIII domains as a bait, we did not observe any known hemidesmosomal proteins interacting with ␤ 4 . This is not surprising, as no molecule binding ␤ 4 directly has ever been identified, but it is of some interest, since it has been recently reported that recombinant HD-1 may precipitate ␤ 4 integrin from cell lysates (35). In this work (35), no evidence of direct binding of ␤ 4 with HD-1 was produced, thus leaving open the possibility that this association occurs through intermediate molecules. The failure to isolate HD-1 in our two-hybrid screening clearly supports this latter possibility. Finally, the possibility that p27 BBP is a component of the hemidesmosome will be clarified only by the generation of antibodies suitable for immunoelectron microscopy analysis or by the discovery of p27 BBP mutants that target hemidesmosome assembly.
p27 BBP , a Novel Intermediate Filament-associated Protein with a Broad Function-The difficulty encountered to solubilize p27 BBP could have reflected the possibility that this protein was associated with cytoskeletal elements. In agreement with this observation, most of the p27 BBP molecule was found to be present in the intermediate filament fraction. This finding 2 S. Biffo, unpublished observations. FIG. 6. p27 BBP is located both in the nucleus and at the basal aspect of the membrane of FG2 cells. CLSM z axis digital reconstructions of two FG2 cells stained for ␤ 4 (red) and F-actin (green, A) and for p27 BBP (red) and F-actin (green, B) show that a fraction of p27 BBP is colocalized with ␤ 4 in coherence with the submembranous F-actin cytoskeleton (indicated by triangles) as well as with nuclei (large red dots).
raises several interesting questions that will require extensive investigations. The most important of these relates to the general function of p27 BBP in the absence of ␤ 4 . Two facts are particularly intriguing: p27 BBP is already present in yeast cells and is highly identical to human p27 BBP . Strikingly, the identity is not limited to a specific domain, as it is scattered throughout the molecule and includes conservation of its length (245 amino acids in all species). This suggests that a strong evolutionary pressure is exerted on maintaining a tertiary molecular structure, possibly involved in multiple interactions with other proteins or in polymerization. At this stage sequence analysis does not allow any speculation on the mechanism of action of p27 BBP , since no motifs homologous to functional domains present in other proteins have been detected.
A second fact is the presence of p27 BBP in the nucleus. Recent extensive analysis has shown that also in the nucleus p27 BBP is present in the highly insoluble core filament fraction and is characterized by a striking redistribution during mitosis. 3 In the past, ultrastructural data based on nonconventional electron microscopy techniques have shown that the nucleus contains a system of intermediate filament cytoskeleton that is not based on classical intermediate filaments such as keratins (39). One may speculate that p27 BBP is part of this intermediate filament cytoskeleton, whose biochemical composition is completely unknown. Interestingly, this core filament cytoskeleton is conserved from yeast to humans (39 -41). The generation and analysis of yeast mutants will possibly clarify the function of p27 BBP .