p63α Mutations Lead to Aberrant Splicing of Keratinocyte Growth Factor Receptor in the Hay-Wells Syndrome*

p63, a p53 family member, is required for craniofacial and limb development as well as proper skin differentiation. However, p63 mutations associated with the ankyloblepharon-ectodermal dysplasia-clefting (AEC) syndrome (Hay-Wells syndrome) were found in the p63 carboxyl-terminal region with a sterile α-motif. By two-hybrid screen we identified several proteins that interact with the p63α carboxyl terminus and its sterile α-motif, including the apobec-1-binding protein-1 (ABBP1). AEC-associated mutations completely abolished the physical interaction between ABBP1 and p63α. Moreover the physical association of p63α and ABBP1 led to a specific shift of FGFR-2 alternative splicing toward the K-SAM isoform essential for epithelial differentiation. We thus propose that a p63α-ABBP1 complex differentially regulates FGFR-2 expression by supporting alternative splicing of the K-SAM isoform of FGFR-2. The inability of mutated p63α to support this splicing likely leads to the inhibition of epithelial differentiation and, in turn, accounts for the AEC phenotype.

The recently discovered p53 homologue, p63, utilizes two promoters and an alternative splicing mechanism to encode six distinct protein isotypes (1)(2)(3). These isotypes (TAp63␣, TAp63␤, TAp63␥, ⌬Np63␣, ⌬Np63␤, and ⌬Np63␥) differ in their ability to transactivate responsive genes, control the cell cycle, and induce apoptosis (1,2,4). p63␣ is the longest isotype that contains a transactivation, DNA binding, and oligomerization domain and a sterile ␣-motif at the extreme carboxyl terminus (2). p63 functions as a transcriptional regulator involved in epidermal-mesenchymal interactions during embryonic development where it is required for regenerative proliferation of limb, for craniofacial and epithelial development, and for skin renewal (5,6). In addition, p63 expression has been closely identified with keratinocyte stem cell proliferation (7). The proliferating cells of stratified squamous epithelia con-sist of stem cells and transient amplifying cells producing holoclones and paraclones, respectively. While p63 is abundantly expressed by epidermal and limbal holoclones, it is undetectable in paraclones (7).
Several developmental craniofacial abnormalities (ectodermal dysplasias) are caused by p63 mutations in humans (8 -10). Mutations in the p63 DNA binding domain are found in patients with ectrodactyly-ectodermal dysplasia-cleft lip/palate, split hand/foot malformation, and limb-mammary syndrome (8). Moreover a number of p63 mutations associated with the ankyloblepharon-ectodermal dysplasia-clefting (AEC) 1 syndrome (Hay-Wells syndrome) have been mapped inside the sterile ␣-motif (9). Thus, p63 mutations found in the AEC syndrome represent a perfect model to study unique p63 protein-protein interactions (9).
The sterile ␣-motif is a 65-70-amino acid residue sequence found in many diverse proteins from yeast to humans whose functions range from signal transduction to transcriptional repression. The sterile ␣-motifs have been implicated in mediating protein-protein interactions via the formation of homoand heterotypic oligomers (11)(12)(13)(14). We sought to identify protein candidates that interact specifically with p63 via its sterile ␣-motif and to evaluate the effect of known AEC genetic alterations on this molecular interaction. We found that p63␣ binds apobec-1-binding protein-1 (ABBP1), a member of the RNA processing machinery, resulting in a shift of FGFR-2 alternative splicing toward the epithelial specific K-SAM isoform. However, mutant p63 harboring AEC-derived mutations failed to bind ABBP1. Thus, mutations mapped to p63 sterile ␣-motif may result in the AEC syndrome by abolishing the interaction with ABBP1 and modulating the FGFR-2 pathway leading to aberrant differentiation.
Drs. Tomas Leanderson and Alaitz Aranburu (Lund University, Lund, Sweden). His-tagged p40 protein (residues 1-356, smallest p63 isotype) was obtained from Dr. Keiho Yamaguchi (The Johns Hopkins University, Baltimore, MD) and purified as described previously (34). The human skin keratinocyte cell line (HaCaT, spontaneously immortalized) and human embryonic retina cell line (911, immortalized by adenovirus E1 expression) were obtained from American Tissue Culture Collection. Both HaCaT and 911 cells were grown in Dulbecco's minimal essential medium with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin. All cultures were incubated in a humidified atmosphere of 5% CO 2 at 37°C. Primary head and neck squamous cell carcinoma and lung adenocarcinoma cell lines were isolated by us during 1985-1992 as described previously (33). Skin tissue samples were obtained from p63-null mice (Ref. 5, Jackson Laboratories). Both p63ϩ/Ϫ and Ϫ/Ϫ mice were bred at The Johns Hopkins University Animal Core Facility. Normal C57Bl6 mice were purchased from Charles River Breeding Laboratory.
Preparation of Adenoviral Expression Vectors-Recombinant adenoviruses driving expression of wild type or mutant p63␣ proteins under control of a cytomegalovirus (CMV) promoter were prepared as described elsewhere (33). cDNA for wild type or mutant p63␣ was subcloned into a pAdTrack-CMV shuttle vector and then introduced in viral backbone vector pAdEasy1 by homologous recombination in BJ5183 Escherichia coli cells (QBiogene). Resulting constructs were screened by PacI restriction digestion. PacI-linearized constructs for empty pAd, pAd-p40, pAd-TAp63␣ (wt), pAd-TAp63␣ (mut, L518F), pAd-⌬Np63␣ (wt), and pAd-⌬Np63␣ (mut, L518F) were introduced into 911 cells to generate adenoviruses. Recombinant adenoviruses were amplified by propagation in 911 cells. The CsCl-purified final viral preparations displayed a multiplicity of infection of ϳ10 12 plaque-forming units/ml monitored by titration and expression levels of the green fluorescent protein (GFP).
Transfections-Cells (in a six-well plate) were transiently transfected for 24 h with the desired plasmids (5 g) using LipofectAMINE 2000 (20 l, Invitrogen) according to the manufacturer's protocol or by a modified calcium phosphate-mediated protocol (1). 5 g of plasmid DNA was mixed with 2ϫ MES buffer, pH 7.05, and 2.5 M CaCl 2 for 20 min and then added to 911 cells (in a six-well plate) for 24 h. For transfection of cells grown in a T-25 or T-75 flask, the amount of DNA and reagents was scaled up. The efficiency of transfection was monitored by co-transformation with a pAdTrack-GFP vector. The average transfection efficiency was ϳ80 -90%.
Protein Expression-Two isoforms, ABBP1 and ABBP1⌬ex1/6, were expressed in bacteria as GST fusion recombinant polypeptides. The full-length cDNA for A/B heteronuclear ribonucleoprotein (hnRNP) (MGC 10739) was purchased from ATCC. The A/B hnRNP protein differs from the ABBP1 protein in that it lacks the carboxyl-terminal 47-amino acid residue insert (residues 263-311, see Fig. 3). The ABBP1 cDNA as a GST fusion construct was obtained from Dr. Paul P. Lau (Baylor College of Medicine, Houston, TX). Sequence analysis revealed that this cDNA has a BamHI deletion in the RNA binding domain at the amino terminus of ABBP1 (residues 101-186). Therefore, this plasmid was designated as pGST-ABBP1⌬BamHI. To restore the full-length ABBP1 sequence, we amplified this DNA fragment by PCR using A/B hnRNP cDNA as a template and the following primers: sense, 5Ј-GCGGATCCCAACACTGGACGGTCAAG-3Ј; antisense, 5Ј-GCGGATC-CATTGGCAATTCAATGGCCT-3Ј. Both primers contained putative BamHI restriction sites allowing us to subclone this fragment into the BamHI site in the pGST-ABBP1⌬BamHI plasmid. The resulting construct was verified by sequencing and designated as ABBP1. Using the following PCR primers, sense, 5Ј-GCGCGAATTCATGTCGGAAGCGG-GCGAGGAGCAG-3Ј, and antisense, 5Ј-GCGGCCGCTCAGTATGGCT-TGTAGTTATTCTG-3Ј (containing EcoRI and NotI sites, respectively), we amplified the open reading frames for ABBP1, ABBP1⌬ex1/6, and ABBP1⌬BamHI and then subcloned them into a pCMV-Sport6 vector for mammalian expression and into a pGEX-4T-1 for E. coli expression of GST fusion proteins. For mammalian expression each pCMV-Sport6 construct was transiently transfected into 911 or HaCaT cells. For bacterial expression, each pGEX-4T-1 construct was transformed into E. coli BL21 (Novagen). Overnight cultures of E. coli transformed with recombinant pGEX plasmids were diluted 1:10 in L broth with 50 g/ml ampicillin and incubated at 37°C to an A 600 ϭ 0.5. Isopropyl-␤-Dthiogalactopyranoside was then added to a final concentration of 1 mM at 37°C. After a further 2 h of growth, cells were pelleted at 5000 ϫ g for 10 min at 4°C and resuspended in a 1:5 (v/v) dilution of the original culture volume of NETN (0.5% Nonidet P-40, 1 mM EDTA, 20 mM Tris, pH 8.0, 100 mM NaCl) containing proteases inhibitors (Roche Applied Science). Cells were then sonicated and centrifuged at 10,000 ϫ g for 5 min at 4°C. GST fusion proteins were purified from 1-liter cultures using glutathione-agarose column chromatography (Sigma) according to the manufacturer's protocol. Fractions eluted from glutathione-agarose column were additionally purified by fast performance liquid chromatography on a 5-ml MonoQ HR5/5 column (Amersham Biosciences) using a 30-ml, 50 -800 mM KCl gradient (1 ml/min) containing 10 mM Tris-HCl (pH 7.4), 1 mM dithiothreitol, 1 mM EDTA, 5 mM MgCl 2 , and 1% phenylmethylsulfonyl fluoride. Approximately 1.0 -2.0 mg of fusion protein was purified from 3-4 g of bacterial cells.
RT-PCR Assay-Total RNA was isolated from cells using Trizol reagent (Invitrogen). 1 g of total RNA was used to generate cDNA from each sample using a one-step RT-PCR kit (Qiagen) and custom primers for the FGFR-2 common region or specific primers for either the K-SAM or BEK exons: FGFR-2 com , sense, 5Ј-TGTTGAAAGATGCCGCCGTG-3Ј; FGFR-2 com , antisense, 5Ј-CGTGTGATTGATGGACCCGTATTC-3Ј; FGFR-2 Bek , antisense, 5Ј-GCGTCCTCAAAAGTTACATTCCG-3Ј; KGFR K-Sam , antisense, 5Ј-CGGTCACATTGAACAGAGCCAG-3Ј. As an internal loading control we amplified the GAPDH region using the following primers: sense, 5Ј-GAGAAGGCTGGGGCTCATTT-3Ј; antisense, 5Ј-CAGTGGGGACACGGAAGG-3Ј. PCR products were resolved by 2% agarose electrophoresis or 4 -20% gradient non-denatured PAGE (Bio-Rad) and visualized with ethidium bromide. For each primer set the number of amplification cycles was predetermined to remain in the exponential phase. To provide a high degree of standardization all experiments including HaCaT and 911 were performed simultaneously using the same reaction mixture, and GAPDH was co-amplified to confirm equal amounts of starting cDNA. RT-PCR amplification results were analyzed digitally by Kodak 1D 3.5 software (Eastman Kodak Co.). The net intensity of PCR bands for the full-length FGFR-2-BEK and K-SAM (KGFR) were measured and normalized to net intensity of the control GAPDH bands.

Two-hybrid Screening of p63-interacting Proteins-Previous
reports clearly identified germ line mutations of p63 sterile ␣-motif in the AEC syndrome (9). We hypothesized that protein interactions mediated by the sterile ␣-motif of p63␣ would contribute to the AEC phenotype. We thus sought to identify protein candidates that interact with the p63 sterile ␣-motif by means of two-hybrid screens.
We generated yeast expression constructs containing the CT of p63␣ (residues 411-642) or the SAM (residues 499 -568). Both sequences (CT or SAM) were subcloned in-frame to the Gal4 DNA binding domain (BD), and resulting bait plasmids were designated as pGal4-BD-p63CT or pGal4-BD-p63-SAM, respectively. Using pGal4-BD-p63CT as bait, we screened four two-hybrid cDNA libraries from human fetal kidney, human fetal liver, human keratinocytes, and HeLa cells. After screening a total 6.9 ϫ 10 6 yeast transformants, we identified 48 positive clones. All clones were verified to grow on different selective media, SD/Trp Ϫ Leu Ϫ , SD/His Ϫ Trp Ϫ Leu Ϫ , and SD/ Ade Ϫ His Ϫ Trp Ϫ Leu Ϫ in the presence of X-gal to screen for ADE2, HIS3, and MEL1 expression. This high stringency screen virtually eliminates false positive interactions but preserves low affinity interactions. As controls, we used yeast expression constructs with a p53/SV40 pair (positive) or a p53/ lamin C pair (negative) (data not shown). In addition, the intermolecular interactions between bait and prey proteins were verified by subsequent transformation of both plasmids into the AH109 yeast strain followed by ␣and ␤-galactosidase assays. These individual clones were sequenced and found to encode the following human proteins: RACK1 (G-binding protein, protein kinase C receptor, 16 clones), Ral guanine nucleotide exchange factor (four clones), ABBP1 (seven clones), Era guanine trinucleotide phosphohydrolase A (five clones), Scaf4/ rA4 RNA-splicing protein (nine clones), ␤-catenin (four clones), and ␤-microglobulin (three clones). All pGal4-AD plasmids isolated from identified positive yeast colonies were also found to interact with pGal4-BD-SAM (data not shown). Qualitative and quantitative assays showed that the interactions between p63 and RACK1, ABBP1, or Scaf4/rA4 proteins are specific (Fig. 1).
To establish whether mutations found in the AEC syndrome affect any of these protein interactions, we generated a bait p63 construct with a mutation in the sterile ␣-motif at position 518 (Leu to Phe) representing the most common genetic alteration in AEC. Pairs of wild type and mutant pGal4-BD-p63-SAM baits and appropriate prey plasmids were retransformed into the AH109 yeast strain, and protein interactions were monitored by selective growth and ␣and ␤-galactosidase assays. As shown, introduction of the L518F mutation dramatically decreased the association of p63 with ABBP1 and Scaf4/rA4, while interactions with the RACK1 prey plasmid were unaffected (Fig. 1). The interaction of p63 with other prey plasmids (Era, Ral, ␤-catenin, and ␤-microglobulin) was also unaffected by this mutation (data not shown).
p63-SAM Specifically Associates with ABBP1-We decided to focus our efforts on further study of the ABBP1-p63 interaction. Sequence analysis of the prey cDNA that interacted with the p63 sterile ␣-motif identified predominantly the carboxyl-terminal domain of human ABBP1 (Fig. 2). Two isoforms of the A/B hnRNP family were identified previously in mammalian cells including humans (ABBP1 and A/B hnRNP, Refs. 15 and 16). They share the same sequence except for a 47residue region, which is present in ABBP1 but missing in the alternatively spliced isoform A/B hnRNP (17)(18)(19). To evaluate the specific region of ABBP1 that mediates its association with the p63 sterile ␣-motif, we carried out two-hybrid yeast expression studies between pGal4-BD-p63-SAM and either pGal4-AD-ABBP1, ABBP1⌬BamHI, or pGal4-AD-ABBP1⌬ex1/6 ( Fig.  2). We found that only ABBP1 was capable of interacting with p63-SAM, while ABBP1⌬ex1/6 failed to show binding to the p63 sterile ␣-motif in yeast (data not shown). Moreover the mutated p63-SAM baits (L518F, L518V, C526G, G534V, Q540L, and I541T) also failed to maintain binding between p63 and ABBP1 (data not shown). These observations suggest that these residues are critical for maintaining the interaction between the p63 sterile ␣-motif and ABBP1.
To confirm the results of two-hybrid screens by independent techniques, we generated pCMV-Sport6 expression cassettes for ABBP1, ABBP1⌬BamHI, and ABBP1⌬ex1/6 driven by CMV promoter. The resulting constructs were transiently transfected into 911 cells together with the p63 expression constructs (⌬Np63␣, wild type or mutant), and expression was monitored by RT-PCR (Fig. 3, A and B, respectively). Protein levels of recombinant polypeptides were evaluated by immunoblotting with the antibodies indicated (Fig. 3, C and D). We observed that p63␣ exclusively associated with the full-length ABBP1 and ABBP1⌬BamHI, while it failed to associate with ABBP1⌬ex1/6 (Fig. 3E). However, all ABBP1 isoforms failed to associate with the p63␣ mutant (L518F) emphasizing the importance of this residue in forming physical complexes between p63␣ and ABBP1.
To further examine the association of p63␣ and ABBP1, we generated GST fusion polypeptides for ABBP1, ABBP1⌬ex1/6, and ABBP1⌬BamHI. The BamHI deletion of ABBP1 produces an isoform lacking a major part of the RNA binding domain, and deletions of exons 1 and 6 produce isoforms lacking the p63␣ binding domain. We also expressed the wild type and mutant p63␣ using expression constructs and recombinant adenoviruses. Total cell lysates expressing p63 isotypes were mixed with purified GST fusion ABBP1 isoforms, and protein complexes were precipitated with an antibody to CBF-A and blotted with an antibody to p63 (Fig. 4). As a control, we used the His-tagged p40 protein (the smallest p63 isotype lacking the extreme carboxyl-terminal of p63␣). Our data demonstrate that p63␣ forms a physical complex with ABBP1 or ABBP1⌬BamHI but not with ABBP1⌬ex1/6 ( Fig. 4C). As expected, His-p40 failed to bind either of the ABBP1 polypeptides (Fig. 4C). Mixing of total lysates of 911 cells expressing p63 polypeptides with GST fusion ABBP1 protein showed that both TAp63␣ and ⌬Np63␣ associate with ABBP1, whereas p40 and all tested p63␣ proteins bearing AEC-derived mutations failed to form complexes with GST-ABBP1 (Fig. 4D). Thus, we propose that the 47-residue insert of ABBP1 (residues 263-311) likely functions as a binding site for the p63 sterile ␣-motif.
The primary structures of the p63 and p73 sterile ␣-motifs are very similar, and based on p73 (12), a homology model for the p63 sterile ␣-motif has been proposed (3). According to this model all AEC-associated mutations can be separated into two different subgroups (9). The first subgroup contains mutations affecting residues that are predicted to be buried inside the protein molecule and have small solvent-accessible surfaces (e.g. Leu-518, Ile-541, and Cys-526). The second subgroup contains surface-located mutations that generally affect residues (e.g. Gly-534 and Gln-540). Mutations of the first subgroup of amino acids are likely to affect overall structure and stability of the protein, while the second subgroup of mutations is less likely to cause gross conformational changes. We found that AEC-associated mutations from both groups (518, Leu to Phe or Val; 526, Cys to Gly; 534, Gly to Val; 540, Gln to Leu; 541, Ile to Thr) abolished the interaction between p63␣ and ABBP1 under these experimental conditions (Fig. 4).

p63-ABBP1 Specifically Regulates Splicing of FGFR-2-K-SAM (KGFR)-From the panel of two-hybrid interacting clones
we identified only two proteins (ABBP1 and Scaf4/rA4) whose interaction with the carboxyl-terminal domain of p63␣ affected the sterile ␣-motif mutations derived from AEC. Both ABBP1 and Scaf4/rA4 are RNA-binding proteins that function in RNA processing and play a critical role in mRNA splicing (15, 16, 20 -25). These proteins are involved in several RNA-related biological processes such as transcription, pre-mRNA processing, mRNA export from the nucleus to the cytoplasm, and mRNA translation. However, the major role of the hnRNP proteins is regulation of mRNA splicing.
We found an interesting clue as to how the interactions between p63␣ and ABBP1 (or Scaf4/rA4) could result in the AEC phenotype. Our initial data with the Affymetrix DNA chip array analysis of p63 downstream targets (confirmed by Northern blot hybridization) showed that p63␣ induces FGFR-2 (ϳ2.5-4-fold), especially its alternatively spliced K-SAM isoform (data not shown). We hypothesized that mRNA accumulation might not be related to activation of gene transcription but rather to promotion of an alternative splicing mechanism. The FGFR-2 gene encodes several alternatively spliced isoforms with mutually exclusive exons including K-SAM (or III␤) and BEK (or IIIc) exons (17,18,26,27). Numerous reports have demonstrated that epithelial cells predominantly express the K-SAM isoform, whereas mesenchymal cells produce mainly the BEK isoform of FGFR-2 (9, 28 -30). Moreover changes in the regulation of the alternative splicing of the two major FGFR-2 isoforms were reported to play an important part in
To test our hypothesis, we designed an experimental model to evaluate FGFR-2 splicing in the presence or absence of ABBP1 isoforms and p63␣. Several cell lines were checked by RT-PCR for the presence or absence of the BEK and K-SAM isoforms of FGFR-2 (data not shown). We found tissue-specific isoforms of FGFR-2 (823 bp for BEK, 807 bp for K-SAM, and 542 bp for the common region) in 911 and HaCaT cells (Fig.  5A). To verify the identity of the near migrating PCR bands, the PCR fragments were digested with the appropriate enzymes to identify unique sites for AvaI (779 bp) or HincII (754 bp), respectively (Fig. 5B). We observed that 911 cells express both the BEK (ϳ90%) and K-SAM (ϳ10%) isoforms, while HaCaT cells exclusively produce the K-SAM isoform of FGFR-2 (Fig. 5). These differences are likely explained by the origin of these cell lines since 911 cells are derived from human embryonic retina, whereas HaCaT cells are derived from human adult skin keratinocytes.
We then took advantage of the low level of K-SAM expression in 911 cells to modify its regulation through expression of p63␣-ABBP1. 911 cells were transiently transfected with various expression constructs for ⌬Np63␣ (wild type), ⌬Np63␣ (mutant, L518F), or ABBP1. Total RNA was used as a template for RT-PCR with FGFR-2 primers corresponding to the common region (Fig. 5, ComD-R), or BEK-specific (Fig. 5, ComD-B/BEK), and K-SAM-specific (Fig. 5, ComD-K/K-SAMIII) regions. Gel electrophoresis separation of PCR products demonstrated that ectopic expression of either ⌬Np63␣ (wild type) or ⌬Np63␣ (mutant, L518F) had no effect on the expression of the BEK or FGFR-2-K-SAM isoforms compared with cells transfected with an empty vector (data not shown). However, we detected a dramatic decrease of the K-SAM isoform when ABBP1 was ectopically expressed (Fig. 5C, lane 16 versus  lane 15). Co-expression of ABBP1 with ⌬Np63␣ (wild type), but not with ⌬Np63␣ (mutant, L518F), restored the expression of the K-SAM isoform in 911 cells (Fig. 5C, lanes 17 and 18). These striking observations support the notion that direct physical association of ⌬Np63␣ with ABBP1 has a dramatic effect on expression of the K-SAM isoform (Fig. 5).

DISCUSSION
Members of the p53 family (p53, p73, and p63) possess a similar modular structure with high homology; however, their functional attributes contribute quite differently to the regulation of cell proliferation, differentiation, tumorigenesis, and development (4 -7, 33-38). Despite the structural similarities of the p53 family members, only certain p63 and p73 isotypes contain an extended carboxyl terminus with the sterile ␣-motif (4). We demonstrated here that the p63 sterile ␣-motif mediates interactions with critical regulatory proteins, including ABBP1 and Scaf4. Both ABBP1 and Scaf4 are involved in RNA transcription, pre-mRNA processing, RNA editing, and splicing (21,24,25,39,40). We further observed that AEC-derived mutations predominantly affect interactions of p63␣ with these proteins. This physical association was confirmed by co-immunoprecipitation of either bacterially expressed purified GST fusion proteins or proteins ectopically expressed in mammalian cells. We also found that ABBP1 strongly interacted with p63␣, whereas the alternatively spliced variants that lack the carboxyl-terminal 47-residue insert failed to associate. Moreover a broad range of mutations found in the p63 sterile ␣-motif of AEC patients abrogated the binding of p63␣ to ABBP1 or Scaf4 under a variety of experimental conditions.

FIG. 7.
Absence of p63 modulates FGFR-2 splicing in p63 knockout mice. A, expression of p63 and ABBP1 variants as indicated. B, GAPDH control. C, FGFR-2 common and tissue-specific isoforms. D, distribution of tissue-specific isoforms of FGFR-2 in the skin of p63 knockout and wild type mice. Semiquantitative RT-PCR was performed using RNA templates (2 g) obtained from total skin extracts of p63(ϩ/ϩ) and p63(Ϫ/Ϫ) frozen mouse embryos using the following primers: sense, 5Ј-GGCCCGGGCCACCTCGTGCAGTCTCAGTC-3Ј; antisense, 5Ј-GGTCGATGGGCGCTCATTCTCCTTCCTCTTTGA-3Ј. skin, p63 is predominantly expressed in proliferated basal keratinocytes, and its expression is rapidly diminished when cells start to differentiate. However, expression of mutated p63 in AEC is no longer restricted to the proliferative cell layer and persists in cells that normally undergo epithelial differentiation. We propose that p63 sterile ␣-motif mutations affect the RNA processing (e.g. splicing) of genes implicated in epithelial differentiation. We further suggest that FGFR-2 differential splicing may play a critical role in the AEC syndrome since FGFR-2 mutations and aberrant splicing were implicated previously in pathogenesis of some craniofacial defects (31,32,41).
Alternative splicing of FGFR-2 is mostly tissue-specific, producing epithelial variants (␤ isoforms) and mesenchymal variants (c isoforms) (26,27,42,43). A number of reports demonstrated that epithelial cells support splicing of the K-SAMIII␣ exon leading to FGFR-2-SAM (KGFR) expression, whereas mesenchymal cells preferentially support splicing of the BEK III␤ exon leading to expression of the FGFR-2-BEK isoform (28, 29, 44 -46). We identified alternative splicing of the FGFR-2 receptor as a likely consequence of the p63 sterile ␣-motif and ABBP1 interaction. We clearly showed that ABBP1 induces preferential splicing of the BEK isoform, while physical association with p63␣ leads to inhibition of ABBP1 activity followed by an increase in the keratinocyte-specific K-SAM isoform.
Abrogation of either FGFR-2 spliced isoforms in mice was reported to cause severe developmental defects supporting a critical role for FGFR-2 splicing in differentiation and morphogenesis (30,32,43,45). Moreover the FGF10 mesenchymal derived ligand of the KGFR was shown to play an important role in limb development by promoting expression of p63 and ␤-catenin suppressing expression of sonic hedgehog (47). Similarly preliminary evidence suggests that FGF10 induces ⌬Np63␣ transcription, while the ectopically expressed p63␣ protein leads to accumulation of FGFR-2 RNA in Saos-2 cells (data not shown). Furthermore there is a strong correlation of endogenous p63␣ protein levels with FGFR-2-K-SAM isoform levels in head and neck cancer cell lines and primary tumor specimens (data not shown). Interestingly we also found that wild type p63␣ expression led to activation of several genes involved in RNA processing (splicing factor-1, Ran-binding protein-2, FUSE-binding protein-2, and NAPOR/ELAV), while mutated p63␣ had no effect on their gene transcription (data not shown). Thus, the interaction between the p63 sterile ␣-motif and ABBP1 could promote a feedback mechanism of preferential expression of KGFR isoform in cells through regulation of alternative splicing of FGFR-2, eventually leading to epithelial type differentiation after induction of cell proliferation.
The post-transcriptional regulation of gene expression can specifically affect the on/off regulation of particular gene products in a temporal and spatial fashion (48). This regulation allows cells of different types or developmental stages to finely tune the gene expression patterns. On/off regulation of RNA processing allows a cell to respond to environmental cues more quickly than that permitted by de novo transcription. Many critical events in development (pattern formation and terminal differentiation) are regulated by an array of post-transcriptional mechanisms controlling mRNA stability, localization, and translation. The post-transcriptional regulation of gene expression can also generate an enormous range of protein products to generate RNA sequence diversity. They include alternative splicing of pre-mRNAs, alternative polyadenylation site selection, and RNA editing (48). Aberrant RNA processing can result in the synthesis of deleterious proteins causing a disease as a result of loss or gain of function by generating a dominant-negative inhibitor or by making a correct protein at the inappropriate time. Our data suggest that AEC-derived mutations in p63 mediate aberrant RNA processing that targets the FGFR (among others) resulting in the characteristic phenotypic epithelial defects seen in the Hay-Wells syndrome (49).