High thermostability and lack of cooperative DNA binding distinguish the p63 core domain from the homologous tumor suppressor p53.

The p53 protein is the major tumor suppressor in mammals. The discovery of the p53 homologs p63 and p73 defined a family of p53 members with distinct roles in tumor suppression, differentiation, and development. Here, we describe the biochemical characterization of the core DNA-binding domain of a human isoform of p63, p63-delta, and particularly novel features in comparison with p53. In contrast to p53, the free p63 core domain did not show specific binding to p53 DNA consensus sites. However, glutathione S-transferase-fused and thus dimerized p63 and p53 core domains had similar affinity and specificity for the p53 consensus sites p21, gadd45, cyclin G, and bax. Furthermore, the fold of p63 core was remarkably stable compared with p53 as judged by differential scanning calorimetry (T(m) = 61 degrees C versus 44 degrees C for p53) and equilibrium unfolding ([urea](50%) = 5.2 m versus 3.1 m for p53). A homology model of p63 core highlights differences at a segment near the H1 helix hypothetically involved in the formation of the dimerization interface in p53, which might reduce cooperativity of p63 core DNA binding compared with p53. The model also shows differences in the electrostatic and hydrophobic potentials of the domains relevant to folding stability.

The tumor suppressor gene p53 is the most frequent site of genetic alterations found in human tumors (1). The p53 protein functions primarily as a transcription factor regulating the expression of genes involved in cell cycle arrest, cellular senescence, anti-angiogenesis, and apoptosis (reviewed in Refs. 2 and 3).
Unlike p53, p63 is essential for embryonic development; mice lacking the p63 gene exhibit severe defects in ectodermal differentiation (20). Based on this phenotype, a role for p63 in stem cell regeneration to sustain epithelial development was suggested (22). Mutations in the p63 DNA-binding domain are the cause of the autosomal dominant EEC 1 syndrome and the EEC-like limb mammary syndrome (23).
All members of the p53 family possess a modular architecture with an N-terminal transactivation domain, an ϳ60% homologous core DNA-binding domain (DBD) that is followed by a tetramerization domain, and a regulatory C terminus (25,26). Several isotypes of p63 and p73 have a conserved C-terminal extension of ϳ100 residues that is not present in human p53 and that might be a protein-protein interaction module with a regulatory function. The structure of this region in p73 has been determined recently (27,28) and revealed structural homology to the sterile ␣-motif domain. Alternative splicing of the C-terminal region of p63 and p73 leads to the expression of a variety of splice variants (29). In addition, N-terminally deleted isoforms of p63 and p73 were identified that are not capable of transcriptional transactivation and that have an anti-apoptotic role as antagonists of their full-length counterparts (7,30).
The p53 DBD contains several hot spot regions for mutation. In addition to the loss of DNA-binding function, a gain of function for mutant p53 DBDs is described (31,32). The crystal structure of the p53 DBD in complex with DNA (33) shows that almost all mutations described target residues that either directly contact DNA or stabilize the tertiary structure. Recently, the crystal structure of the free mouse p53 DBD was solved (34). Wild-type and mutant p53 DBDs have been studied biophysically in detail (35)(36)(37)(38)(39)(40). Several studies tried to rescue the mutant or to stabilize the wild-type p53 DBD conformation based on structural information (41) or by generating secondsite suppressor mutations (42,43). These attempts and others using semirational design (44) or directed evolution (45,46) yielded protein of increased thermodynamic stability. Very recently, pharmacologically active low molecular mass com-* 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 /EBI Data Bank with accession number(s) AJ315499.
¶ To whom correspondence should be addressed. pounds were reported to stabilize the wild-type conformation of human p53 (47).
So far, little is known about the biochemical properties of the homologous DBDs, which share ϳ60% identity with the p53 DBD and are highly conserved in residues involved in Zn 2ϩ coordination and stabilization of the DNA-binding conformation (4,7,33,48) (see Fig. 1). In contrast to p53, mutational inactivation of p73 and p63 in human tumors is rare and does not seem to be important for carcinogenesis (49). In this study, the p63 DBD was biochemically characterized in comparison with the p53 DBD. In contrast to p53, the free p63 DBD is not capable of binding to specific p53 DNA consensus sites; however, in a dimerized state, GST-p63 DBD as well as GST-p53 DBD show comparable affinity and specificity for p53 consensus sites, supporting the notion that the DBDs might differ with respect to dimerization and consequently cooperative DNA binding, but not with respect to DNA specificity. Differential scanning calorimetry and urea-induced equilibrium unfolding showed that the p63 DBD is markedly stabilized in relation to p53. To understand these results on a molecular basis, a homology model of the p63 DBD was created. Some properties of the p63 DBD may rationalize the lack of cooperativity and enhanced thermostability.

EXPERIMENTAL PROCEDURES
Materials-All chemicals used were of analytical grade and obtained from major commercial suppliers. TAMRA-labeled DNA oligonucleotides were purchased from MWG-BIOTECH.
Cloning, Expression, and Purification-The 705-base pair cDNA encoding the DNA-binding domain (amino acids 114 -349) of the human p63 protein was amplified from a human placenta cDNA library (CLONTECH) by standard polymerase chain reaction with gene-specific primers 5Ј-GCA GCG CCA TAT GGG ATC CTC CAC CTT CGA TGC TCT CTC T-3Ј and 5Ј-CGC GCT CGA GTT ATC ATG TCA TCT GGA TAC CAT GTG TGT TCT G-3Ј. The C-terminal part of the amplified sequence contained a deletion of five amino acids. This part could be retrieved as human expressed sequence tag (EST) in the EMBL Nucleotide Sequence Database (dbEST) (accession number: est_hum11: aw605859, est_hum10:aw382125) with BLAST2 (50) and was originally identified in the FAPESP/LICR Human Cancer Genome Project. The polymerase chain reaction product was subcloned into a modified pQ40 vector (QIAGEN Inc.) and into a pGEX-4T-1 vector (Amersham Pharmacia Biotech). The resulting recombinant expression vector pQ40(p63 DBD) codes for the human p63 DBD without tags, including an additional N-terminal glycine; the corresponding vector pGEX-4T(GST-p63 DBD) codes for a GST-p63 DBD fusion protein including a Gly-Ser-Gly linker that remains on the N terminus after digestion of the fusion protein with thrombin. All vector constructs were confirmed by sequencing. Proteins were expressed in Escherichia coli strains HB101 and ut5600, which were grown at 37°C in LB medium supplemented with ampicillin (100 g/ml) and/or kanamycin (50 g/ml) to an absorbance of 0.5-0.8 before overnight induction at 37°C with 1 mM (for p63 DBD) or 0.1 mM (for GST-p63 DBD) isopropyl-␤-D-thiogalactopyranoside. After induction, cells were harvested by centrifugation; resuspended in 50 mM Tris (pH 6.8), 5 mM DTT, 1 mM benzamidine, and Complete protease inhibitor mixture (EDTA-free; Roche Molecular Biochemicals); and disrupted by high-pressure dispersion using an APV-Gaulin Lab 40 Homogenisator. For purification of the p63 DBD, soluble lysate was loaded onto an SP-Sepharose Fast Flow cation-exchange column (Amersham Pharmacia Biotech) and eluted with a linear KCl gradient (0 -0.5 M). Final purification was achieved by preparative size-exclusion chromatography on a Superdex 75 HiLoad 26/60 column (Amersham Pharmacia Biotech) in 50 mM Tris (pH 7.0), 150 mM KCl, and 5 mM DTT. For the purification of GST-p63 DBD, 1% Triton X-100 (Sigma) was added to the lysate. Following centrifugation, the soluble lysate was loaded onto a GSH-Sepharose 4B column (Amersham Pharmacia Biotech). GST-p63 DBD was eluted with 10 mM glutathione in 50 mM Tris (pH 8.0), 150 mM KCl, 5 mM DTT, and 5% glycerol. Final purification was achieved by size-exclusion chromatography on a Superdex 200 HiLoad 26/60 column (Amersham Pharmacia Biotech) in 50 mM Tris (pH 7.0), 150 mM KCl, 5 mM DTT, and 5% glycerol.
Cloning, expression, and purification of the p53 DBD and GST-p53 DBD for comparison studies were performed with minor modifications of the procedure given for the p63 constructs and published elsewhere (e.g. using an additional heparin HiTrap column (Amersham Pharmacia Biotech) for the purification of the p53 DBD) (39). All proteins were concentrated using 5K Ultrafree 4 centrifugal filter devices (Millipore Corp.), flash-frozen in liquid nitrogen, and stored at Ϫ80°C.
Analytical Methods-Electrospray mass spectrometry confirmed the identity of all proteins used in this study and showed that the Nterminal methionine was cleaved off after translation. Protein concentration was measured spectrophotometrically according to the method of Bradford (51) or using extinction coefficients of ⑀ 280 nm ϭ 14,650 M Ϫ1 cm Ϫ1 for the p63 DBD and ⑀ 280 nm ϭ 15,930 M Ϫ1 cm Ϫ1 for the p53 DBD, calculated according to the method of Edelhoch (52). SDS-polyacrylamide gel electrophoresis was performed with 12.5% gels. Analytical size-exclusion chromatography was performed to determine the oligomeric state of proteins using a TSK Gel G 3000SW analytical gel filtration column (TosoHaas) on an analytical Gynkotek high-pressure liquid chromatography instrument (Dionex Corp.) equipped with Chromeleon software (Dionex Corp.) at a flow rate of 0.5 ml/min in 50 mM sodium phosphate (pH 7.0), 150 mM KCl, 5 mM DTT. Dynamic light scattering to measure hydrodynamic parameters was performed with a DynaPro molecular sizing instrument (Protein Solutions Inc.) equipped with a temperature controller. All samples (protein concentration of 1-2 mg/ml) were diluted in 40 mM sodium phosphate (pH 7.0) and 5 mM DTT and filtered through a 0.02-m membrane before analysis at 20°C.
Electrophoretic Mobility Shift Assay (EMSA)-The DNA-binding activity of the protein constructs was analyzed qualitatively by EMSA as described (53). Briefly, specific complementary oligonucleotides containing the 20-mer p53 consensus DNA-binding site PG (polygrip) (54) were end-labeled and annealed. The DNA binding buffer contained 40 mM HEPES (pH 8.0), 50 mM KCl, 20% (v/v) glycerol, 5 mM DTT, 0.1% Triton X-100, 10 mM MgCl 2 , and 1.0 mg/ml bovine serum albumin. In general, 50 ng of purified proteins in 10 l of DNA binding buffer were mixed with 10 l of DNA binding buffer containing 10 ng of doublestranded end-labeled specific p53 consensus site oligonucleotide and 5 nM supercoiled unlabeled nonspecific pBluescript II SK ϩ (pBS) competitor DNA (Stratagene) and incubated for an additional 15 min on ice. The full preincubation details are given in the figure legends. The reaction mixture was loaded onto a 4% native polyacrylamide gel and separated at 200 V for 2 h at 4°C. The running buffer consisted of 30 mM Tris-HCl (pH 7.5), 30 mM boric acid, and 1 mM EDTA with 0.01% Triton X-100. The gel was dried, and the labeled DNA was detected by autoradiography.
Fluorescence Correlation Spectroscopy (FCS)-FCS allows the determination of free diffusion coefficients in solution. If the difference in the diffusion coefficients of the free and bound molecules is sufficiently large, binding of the fluorescent molecule to a target can be analyzed. Spectroscopy-Fluorescence spectra were recorded for 1 M protein in 40 mM sodium phosphate (pH 7.0) and 5 mM DTT (Ϯ6 M urea for denatured spectra) in 1.0-cm thermostatted quartz cuvettes at 15°C with excitation at 280 nm and emission scanned from 300 to 400 nm using a FluoroMax-2 fluorometer (Spex, Jobin Yvon luc., Horiba) as described for the p53 DBD (39). CD measurements with 8 M (far-UV) or 80 M (near-UV) protein in 40 mM sodium phosphate (pH 7.0) (Ϯ6 M urea for denatured spectra) were acquired in 0.1-or 1-cm thermostatted quartz cuvettes using a Jasco J-715 spectropolarimeter equipped with a Peltier PTC-351S element at 15°C. Proteins were exchanged into degassed buffers prior to measurements using NAP-10 columns (Amersham Pharmacia Biotech). The program CDNN (55) was used to estimate secondary structure contents based on the advanced data base of reference spectra.
Differential Scanning Calorimetry (DSC)-DSC experiments were performed using a 5106 N-DSC Nano differential scanning calorimeter (Applied Thermodynamics, Calorimetry Sciences Corp.) with a cell volume of 867.2 l. Temperatures from 5 to 90°C were scanned at a rate of 1°C/min. All analyzed proteins were dialyzed before analysis into 40 mM sodium phosphate (pH 7.0), 150 mM KCl, and 1 mM DTT, and their concentration was adjusted to 40 M. Prior to measurements, samples were degassed. The dialysis buffer was used for base-line scans. A pressure of 3 bar was applied to the cell, and the system was allowed to equilibrate at 4°C for 15 min before scanning. Acquired thermograms were analyzed with Nano-DSC software (Calorimetry Sciences Corp.).
Equilibrium Urea Denaturation-Equilibrium urea denaturation was performed as described (39,56). Briefly, unfolding of the p63 and p53 DBDs was performed at 15°C in 40 mM sodium phosphate (pH 7.0) and 1 mM DTT to keep all cysteines reduced. For each data point of the denaturation curve, 50 l of a 160 M solution of the p63 DBD (or of a 80 M solution of the p53 DBD) in 40 mM sodium phosphate (pH 7.0) and 1 mM DTT were diluted into 950 l of the corresponding urea denaturant solution in this buffer. The samples were incubated overnight at 15°C and analyzed subsequently. The urea stock solutions were prepared gravimetrically; the corresponding urea dilutions were prepared volumetrically and confirmed by refractometry. Unfolding transitions of the p63 DBD were monitored at 15°C in 0.2-cm thermostatted quartz cuvettes by the increase in the mean CD signal at 222 nm (bandwidth of 1 nm and acquisition time of 1 min) upon unfolding. Unfolding transitions of the p53 DBD were monitored by fluorescence spectroscopy (increment of 0.1 nm, integration time of 0.1 s, and five acquisitions) at 15°C making use of the increase in the normalized fluorescence emission at 356 nm upon unfolding (39). Both unfolding transitions were analyzed on the basis of a two-state approximation (57) by nonlinear least-square analysis using the curve-fit option of the program Prism 3.0.
Molecular Modeling-The model structure of the p63 DBD was cre-ated using the crystal structure of the p53 DBD in complex with DNA (Protein Data Bank code: 1TSR, chain B) (33). The comparison with the very similar structure of the p53 DBD bonded to the ankyrin and SH3 domains of 53BP2 (58) (Protein Data Bank code: 1YCS) was helpful in the identification of flexible regions. The model was created, checked, and minimized with the Homology and Discover modules of InsightII. Sequence alignment searches were performed using Fasta. Secondary structure predictions were generated using NNPRED (59) and NNSSP (60). Lipophilic potentials were computed with an in-house program assigning a Clog P value to each atom. Figures are pictures of InsightII (MSI Inc.) screens.

Cloning, Expression, and Purification of the p63 DBD
The DBD of the p63 gene was amplified by polymerase chain reaction from a human placenta cDNA library (CLONTECH) using gene-specific primers. Unexpectedly, the cloned DNA sequence had a deletion of 12 base pairs encoding the exchange of five amino acids with alanine in the C-terminal region of the corresponding p63 DBD. This deletion in the DBD is also known for the ␥-isoform of murine p63, but has not been described for the human p63 isoforms so far (Fig. 1). Sequences containing this deletion can be retrieved in the EMBL Human dbEST Database (see "Experimental Procedures" for details). The cloned p63 DBD therefore probably codes for a novel isoform of human p63, preliminarily designated as p63-␦. The encoded p63 DBD (residues 114 -349) characterized in this study was chosen on the basis of the sequence alignment for the p53 family members (Fig. 1) and corresponds to the p53 DBD (residues 94 -312) used in several studies (36,37). It includes short, presumably unstructured, N-and C-terminal residues in addition to the minimal DBD (33). The degree of identity between these two domains is 55.4%. The p63 DBD and GST-p63 DBD (data not shown) plasmids yielded high-level expression of soluble protein in E. coli after induction at 37°C and showed almost no deposition in inclusion bodies ( Fig. 2A, lanes 1 and 2). The p63 DBD was purified using a combination of cationexchange and size-exclusion chromatography. This purification scheme yielded ϳ30 mg of Ͼ98% homogeneous and monomeric p63 DBD/liter of culture as judged by SDS-polyacrylamide gel electrophoresis ( Fig. 2A, lanes 1-5), analytical size-exclusion chromatography (Fig. 2B), and mass spectroscopy. Analysis of the hydrodynamic parameters by dynamic light scattering showed that the p63 DBD exhibits a hydrodynamic radius of 2.9 nm versus 2.74 nm for the p53 DBD; however, both DBDs exhibit a monomodal distribution with low polydispersity indices (data not shown). Purification of GST-p63 DBD fusion protein was performed using a combination of affinity chromatography and a final size-exclusion step. Yields were ϳ20 mg of Ͼ95% homogeneous protein ( Fig. 2A, lane 9) in a dimeric state (Fig. 2B) per liter of culture. An overview of the purified proteins used throughout this study is given in Fig. 2 (A, lanes 6 -9; and B).

DNA-binding Activity and Specificity of the p63 DBD
DNA-binding Activity of p63-Several studies showed that tetrameric p63 isotypes interact with p53 consensus sequences and that p63 can act as a transcription factor that activates promoters of several p53-responsive genes in cotransfection experiments (including p21, bax, and the artificial consensus site PG) (7,61). Therefore, the DNA-binding activity of the isolated p63 DBD for the p53-responsive consensus site PG (53, 54) was analyzed in comparison with the p53 DBD. The p53 DBD bound to a single consensus quarter-site only with very low affinity so that DNA binding could not be detected by EMSA (data not shown). Upon addition of the PG p53 consensus site, the p53 DBD tetramerized and cooperatively bound with high affinity (Fig. 3A, lane 1) (35,36,40,(62)(63)(64). In contrast, the free p63 DBD was not able to bind to the PG p53 consensus site under identical conditions (Fig. 3A, lane 2). Co-incubation of the p63 DBD with the p53 DBD did not interfere with DNA binding of the p53 DBD (data not shown). Consequently, the ability of GST-p63 DBD to bind to the PG p53 consensus site was analyzed since the GST part of the fusion protein acts as an artificial dimerization domain (Fig.  2B) and thereby should enhance the affinity for DNA. Dimerized GST-p63 DBD fusion protein was capable of binding to the PG p53 consensus sequence with an affinity similar to that of GST-p53 DBD (Fig. 3A, lanes 3 and 4), indicating that the affinity and specificity of the p63 and p53 DBDs are similar. To confirm that the p63 DBD is not differently folded in the context of the GST fusion protein, the fusion proteins were digested with thrombin, and the free DBDs were released. As expected, thrombin-cleaved GST-p53 DBD retained the DNAbinding activity of the released p53 DBD, whereas thrombincleaved GST-p63 DBD completely lost DNA-binding activity (Fig. 3A, lanes 5-8). The lack of measurable DNA-binding activity of the p63 DBD might therefore be explained by its failure to bind to DNA cooperatively. Fig. 9 depicts a schematic model for the interpretation of the experimental results (see "Discussion" for details).
Zn 2ϩ Coordination Is Required for p63 DNA Binding-The presence of Zn 2ϩ is required for the DNA-binding activity of the p53 protein (36). The crystal structure of the p53 DBD⅐DNA complex shows that the structure of the p53 DBD is stabilized by Zn 2ϩ that is tetrahedrally coordinated by three cysteines (residues 176, 238, and 242) and one histidine (residue 179) (33). Treatment of wild-type p53 with metal chelators causes the removal of Zn 2ϩ and oxidation of essential cysteine residues. This results in the disruption of the tertiary structure, with loss of DNA-binding activity, exposure of amino acids cryptic in the fully folded protein, and adoption of the monoclonal antibody PAb 240-positive phenotype identical to that of mutant forms of p53 (65,66). Primary structure data suggest that this metal-dependent structure is conserved in the p53 homolog p63 (4,7,33,48), as all residues involved in Zn 2ϩ coordination are conserved (Fig. 1). Consequently, the effect of metal-chelating agents on DNA binding was examined. Fig. 3B shows that incubation with the metal-chelating agent 1,10phenanthroline reduced the DNA-binding activities of GST-p53 DBD and GST-p63 DBD (lanes 1-10). It was surprising that the Zn 2ϩ seemed to be bound with higher affinity by GST-p53 DBD, as concentrations of 1,10-phenanthroline Ͼ20 mM were needed to abolish DNA binding, whereas GST-p63 DBD lost DNAbinding activity at lower concentrations. As the effect of Zn 2ϩ chelation and loss of DNA-binding activity varied with incubation time and temperature, the effect of 1,10-phenanthroline on the DNA-binding activity was not quantitatively evaluated.
Quantitative Analysis of the DNA-binding Activity and Specificity of p63-FCS was applied to measure the apparent binding constants of fluorescently labeled DNA consensus oligonucleotides for p63 and p53 DBDs. In the first experiments, the effect of supercoiled unlabeled nonspecific pBS competitor DNA on the DNA-binding activities of p63 and p53 DBDs for a specific TAMRA-labeled PG oligonucleotide was determined (Fig. 4, A and B). Whereas binding of the p53 DBD and GST-p53 DBD (Fig. 4A) as well as GST-p63 DBD (Fig. 4B) to PG was specific and almost unaffected over a wide concentration range, DNA binding of the p63 DBD was nonspecific and could be competed at 5 nM nonspecific pBS DNA (Fig. 4B). This concentration of nonspecific pBS DNA was comparable to that used in the EMSA (Fig. 3A), which accounts for the lack of DNA binding of the p63 DBD. In the following experiments, the specific binding affinity of the GST-DBDs for 1 nM TAMRA-labeled p53 consensus oligonucleotides was therefore determined in the presence of 10 nM pBS. For comparative reasons, the question of DNA specificity was addressed using different p53 consensus oligonucleotides. An exemplary binding curve for binding of GST-p53 DBD and GST-p63 DBD to the TAMRA-labeled p21 oligonucleotide is shown in Fig. 4C. The apparent K d values for binding to the different p53 consensus oligonucleotides are given in Table I. Dimerized GST-p63 DBD and GST-p53 DBD showed affinities in a comparable range and an almost identical specificity pattern. With the exception of the bax consensus oligonucleotide, GST-p63 DBD displayed an affinity for the p53 consensus oligonucleotides ϳ3-5-fold higher compared with GST-p53 DBD, so the relative affinities with regard to the selected consensus sites are comparable. Both GST-DBDs showed the highest affinity for the artificial PG site, followed (lanes 7 and 8) were analyzed for their DNA-binding activity before (Ϫ) and after (ϩ) cleavage with 1 g of thrombin. BG probably indicates background binding of a small portion of monomeric GST-p53 that is in equilibrium with dimeric GST-p53 and not detectable by size-exclusion chromatography; FP indicates free probe. B, influence of 1,10-phenanthroline on the DNA-binding activity. 50 ng of GST-p53 DBD (lanes 1-5) and GST-p63 DBD (lanes 6 -10) were incubated for 10 min with the indicated concentrations of 1,10-phenanthroline at 4°C and subsequently subjected to an EMSA at 4°C.  5 and 6) and GST-p63 DBD by the p21, gadd45, and cyclin G consensus sites. The affinity for the bax consensus site was lowest for both DBDs.

Spectroscopy of the p63 DBD
CD Spectroscopy-CD spectra of the native and denatured p63 DBD in comparison with the spectra of the p53 DBD are shown in Fig. 5(A and B). The far-UV spectrum of the p63 DBD indicates a high content of ␤-sheets as secondary structure, which was lost upon denaturation with 6 M urea. The p53 DBD shows a less typical spectrum. The results of secondary structure estimation with the program CDNN (55) (␣-helical content: 8.1% for p53 and 9.3% for p63; ␤-sheet content: 46.4% for p53 and 45.7% for p63) are comparable for both DBDs. Within the limits of secondary structure prediction from CD spectra, these values indicate a similar overall fold of the DBDs and are in accordance with the secondary structure content as deduced by the p53 DBD crystal (33) and the p63 model structure (see below). As the three-dimensional structure and dynamics of the p63 DBD are unknown, the differences in the far-UV CD spectra of the p53 and p63 DBDs cannot be unambiguously explained. Differences in the unique amino acid sequence and composition of the two domains might contribute to the observed differences. However, as the ␣-helical content of the p53 and p63 DBDs is small, even slight differences in the ratio and stability of ␤-sheet and coil segments as well as slight changes in the helical composition of the domains can well account for the observed differences. On the one hand, the increased signal intensity of the p63 DBD spectrum may reflect the significantly higher thermostability of the p63 DBD in comparison with p53 (see Fig. 7). An increased conformational stability of p63 DBD secondary structure elements over time may be the major reason for the observed higher signal intensities of the far-UV CD spectra. On the other hand, the higher signal intensity observed may also be indicative of the existence of slight differences in the structures of the two DBDs regardless of the assumed similar overall fold of the two DBDs (see Fig. 8). The near-UV CD spectra of the DBDs are shown in Fig. 5B. They are indicative of stable tertiary structures of both DBDs. In distinction to p53, the p63 DBD showed a band in the positive range.
Fluorescence Spectroscopy-The characteristic fluorescence spectra of the native and denatured p63 DBD in comparison with the spectra of the p53 DBD are shown in Fig. 5C. Upon excitation at 280 nm, the p53 DBD spectrum is dominated by the contribution of several tyrosine residues at 304 nm, whereas the single tryptophan residue is quenched and shows low fluorescence in the native state; upon urea denaturation, tryptophan fluorescence at 354 nm is strongly enhanced (37). On the other hand, the p63 DBD displays a fluorescence spectrum that is dominated by the single nonconserved tryptophan residue present in the p63 DBD. Upon denaturation, the maximum of emission is red-shifted from 342 to 354 nm.

Enhanced Thermodynamic Stability of the p63 DBD
Thermostability of the p63 DNA-binding Activity-The effect of temperature on the DNA-binding activity of GST-p63 DBD and GST-p53 DBD was examined after 5 min of incubation at 30, 34, 37, 40, 45, 50, 55, 60, and 70°C. Proteins were preincubated at the indicated temperatures and subsequently kept on ice until analysis of the DNA-binding activity by EMSA. Fig.  6 shows that GST-p53 DBD activity was reduced following preincubation at 37°C and abolished after preincubation at 40°C. In contrast, GST-p63 DBD lost DNA-binding activity after preincubation at 52°C for 5 min.
p63 DBD Is Stabilized against Thermal Denaturation-DSC experiments were performed with the p63 and p53 DBDs. As  for the p53 DBD (37), thermal denaturation of the p63 DBD was irreversible and occurred during a single DSC run. However, the melting point T m can be used as a semiquantitative indicator of thermostability. For both proteins, thermal unfolding could be measured, and the melting points were highly reproducible. The p63 DBD showed significantly enhanced thermostability with an apparent T m of 61.0 Ϯ 0.1°C versus 44.0 Ϯ 0.1°C for the p53 DBD (Fig. 7). Thus, the DNA-binding activities of GST-p63 DBD and GST-p53 DBD are lost before the DBDs show obvious signs of thermal denaturation. Due to the irreversibility of the denaturation, no quantitative determination of ⌬H cal and ⌬G(T) was performed. It should be noted that the isolated GST fusion protein unfolded at an apparent T m of 57.3°C and completely lost enzymatic activity after incubation for 5 min at 58°C (data not shown), whereas GST-p63 DBD lost DNA-binding activity after incubation at 52°C.
p63 DBD Is Stabilized against Urea-induced Equilibrium Unfolding-The observation of enhanced thermodynamic stability of the p63 DBD was confirmed using urea-induced equilibrium unfolding. The unfolding of the p63 DBD was measured at 15°C based on the decrease in the CD signal at 222 nm upon unfolding (Fig. 5A). Unfolding transitions are found to be reversible under these conditions, so the data can be fitted according to a two-state transition model (57). As the fluorescence spectra of the p53 DBD are better suited for a quantitative analysis than the CD spectra, equilibrium unfolding of the p53 DBD was measured for comparative reasons based on the increase in the normalized fluorescence signal at 356 nm (Fig.  5C) upon unfolding. Unfolding transitions were performed under the stated conditions at 15°C and were found to be unaffected by the protein concentrations used. Despite the inferior quality of CD unfolding data, however, unfolding transitions yielded comparable results to those monitored by fluorescence spectroscopy. Table II (37,39,43,44). Table III gives a statistical comparison of the p63 and p53 DBDs. Most parameters are comparable; however, the p63 DBD displays a higher hydropathicity (GRAVY) (67) and a higher aliphatic index (68) as well as a lower instability index (69), which can be indicative of elevated thermostability. To investigate the molecular mechanisms of thermostabilization of the p63 DBD and to see whether the p63 and p53 DBDs might differ in their dimerization and thermostability properties, a model of the p63 DBD was created based on the crystal structure of the p53 DBD bound to its DNA consensus sequence (33). Due to the high level of identity and homology between p63 and its parent molecule, p53 (Fig. 1), the sequence alignment is unambiguous, and a similar overall fold for the homologous DBDs can be assumed. Among the important residues, e.g. residues chelating Zn 2ϩ , binding to DNA, and frequently being reported as mutated in p53, all are identical with the exception of an arginine residue in contact with DNA that is replaced by a lysine (Fig. 8A). Concerning the residues near these residues, four non-homologous substitutions are found between p63 and p53. First, Gln 165 (p53) becomes Lys 194 (p63). In the model, this lysine is exposed to the solvent with a charge further compensated by a glutamic acid located two amino acids away. Second, the substitution Gln 192 (p53) with Ser 223 (p63) is also exposed to the solvent. The third substitution Thr 284 (p53) with Ala 315 (p63) is located in the major groove and may play a role in the recognition of DNA consensus sequences. The fourth change concerns Ser 183 (p53) to Arg 212 (p63). This arginine residue is part of a region of 10 amino acid residues (LSREFNEGQI) that is located just after the H1 helix with the histidine coordinating the Zn 2ϩ . This short stretch of residues is not strictly conserved and shows an insertion of two amino acids.

Modeling of the p63 DBD
A search into the Protein Data Bank for this short sequence has not revealed a particular homology for this segment. How-ever, using a data base of C-␣ three-dimensional coordinate matrices extracted from the Protein Data Bank, three backbone loops are suggested for the nonconserved region. The first shows an extended hairpin motif on the surface of the protein.
The second one extends the helix containing the above-mentioned histidine and adds a subsequent shortened loop. The third loop proposed has no particular secondary structure. Of the three models, the one containing the hairpin motif is probably the most stable (Fig. 8A). Moreover, this strand following the H1 helix is the one that differs in conformation between the two known structures of p53 and was suggested to form the protein-protein interface in the tetrameric p53⅐DNA complex (33,58). Fig. 8B shows the lipophilic surface potentials of the p53 DBD structure and the p63 model structure in comparison. In general, a structure-based comparison between the p63 DBD and p53 parent structure reveals no major differences with regard to surface area, molecular volume, or packing density. However, the p63 DBD is apparently less polar than the parent p53 DBD, particularly on the surface opposite to the DNAbinding region (lower panels). DISCUSSION Both, the p53 tumor suppressor protein and its homolog p63 contain a tetramerization domain and form homotetramers in solution. Tetrameric p53 binds specifically to a DNA consensus sequence consisting of two consecutive 10-base pair half-sites, where each half-site is formed by two head-to-head quartersites (33,54,63,70,71). The isolated tetramerization domain forms a symmetric dimer of dimers (72)(73)(74), and two contrasting models have been proposed to describe how the DBDs of each dimer are bound to the quarter-sites, namely with either consecutive or alternating arrangements (75). Several studies showed that four p53 DBDs bind cooperatively to a DNA consensus sequence (35,36,39,40,62,63). The crystal structure of the p53 DBD⅐DNA complex is compatible with a model whereby four p53 DBDs bind to the DNA consensus sequence (33). We compared the DNA-binding properties of a highly purified, monomeric p63 DBD with those of the p53 DBD. In contrast to the p53 DBD, the isolated p63 DBD cannot bind to a specific DNA consensus oligonucleotide in EMSA. There are two general explanations for this result: the p63 DBD either binds  In this case, the resultant binding of a single DBD would not be detected by EMSA due to its low affinity. To distinguish between these possibilities and to enhance affinity, p63 DBDs were attached to GST as an artificial dimerization site. Dimerized GST-p63 DBD binds to several p53 consensus sites with comparable affinity and specificity, as does GST-p53 DBD. Upon digestion of GST-p63 DBD with thrombin, it loses DNAbinding activity. These results support the conclusion that the p63 DBD is sufficient and properly folded to bind to DNA; however, under the experimental settings, it is not capable of binding to DNA cooperatively. This discrepancy can be attributed to differences in the putative dimerization interface made up by the H1 helix in p53. Fig. 9 depicts a schematic model illustrating the experimental results. Whereas the p53 DBD contains a dimerization interface, p63 DBD lacks a functional interface and can just bind to DNA upon GST-mediated dimerization (see below). Recent studies have examined the ability of the p53 family members to transactivate various p53-responsive promoters in reporter gene assays. p73-␣ and p73-␤ were reported to activate some, but not all, previously identified p53 target genes (17,76). For p63, only minor differences in the transcriptional activation of several cellular p53 target genes mediated by p63 FIG. 8. A, superposition of the backbone representation of the p53 DBD structure (shown in cyan) (33) and the p63 DBD homology model (shown in red). The four residues coordinating the Zn 2ϩ are indicated in violet; the six hot spot point mutations in p53 are displayed in green; and the amino acids that are engaged in DNA interactions are highlighted in orange. In gray are the residues located at 2.5 Å of the three first sets. The bound schematic DNA consensus double helix is displayed in yellow. The black oval highlights the region that differs between the two domains. B, lipophilic surface potentials for the p53 DBD (upper panels) and p63 DBD (lower panels) calculated for each atom using Clog P. The coloring is indicated, from blue for negative polar values to red for positive hydrophobic values. For each DBD, the right panel is rotated by 180°r egarding the left panel. and p53 could be shown (61). In that study, it was also found that p63 activates the bax promoter more efficiently than p53 and at levels comparable to those of p21. We examined the question of DNA specificity and selectivity by comparing the affinity of the GST-DBDs for different natural p53 consensus oligonucleotides. In general, the affinities for the consensus sites differ according to the same pattern, so GST-p53 DBD and GST-p63 DBD show almost identical specificity with regard to the selected representative p53 consensus sites. The affinity of GST-p53 DBD in comparison with GST-p63 DBD is relatively 3-5-fold lower for the PG, p21, gadd45, and cyclin G consensus sites, with exception of the bax consensus site, to which both GST-DBDs bind with similar, but lowest affinity. These results show that GST-p63 DBD binds to known p53 consensus sites with similar specificity and at least the same affinity as GST-p53 DBD. However, as could be seen for bax, in vitro binding affinities of the DBDs for p53 consensus site oligonucleotides and the degree of reporter gene activation in cellular assays using the natural promoter sequences cannot be directly compared, as the isolated DBDs and the full-length proteins might differ in their specificity and as additional sequence elements are involved in reporter gene activation (77).
The observed differential recognition of p53 promoters might account for the distinctive role of the p53 family members in tumor surveillance and development (14). It was therefore suggested that the differential responses of the target genes might be due to slightly different specificities or conformational states of the DBDs (61). However, our results with the isolated p63 DBD are not in favor of this hypothesis. Rather, the distinctions in target specificity seem to be based upon a differing specificity of the full-length proteins in the cellular context. Part of these differences might be mediated by the C-terminal regions of the proteins (61). Despite the limitations stated above, quantitative in vitro studies on the binding to larger sets of DNA target sequences might help to explain the different responses of p53-regulated genes in response to p53 activation (18,78) and to resolve the question of how target discrimination occurs within the p53 family (25).
The p63 DBD can be expressed at 37°C in a soluble form, whereas the p53 DBD is highly temperature-sensitive (39,79) and is deposited almost exclusively in inclusion bodies at this temperature. This observation supports the notion that the p63 DBD shows enhanced thermostability. In addition, the p63 DBD is highly stable in contrast to the p53 DBD, as it has almost no tendency to aggregate and precipitate in solution even at elevated temperatures. The enhanced thermodynamic stability of the p63 DBD could be confirmed by DSC and ureainduced equilibrium unfolding. The observed gain in stability of the p63 DBD of a calculated ⌬⌬G([urea] 50% ) ϭ Ϫ4.57 kcal mol Ϫ1 (T m ϭ 61°C versus 44°C and [urea] 50% ϭ 5.15 M versus 3.06 M for the p53 DBD) is remarkable. It is, however, surprising that Zn 2ϩ seems to be more tightly coordinated by the thermodynamically less stable p53 DBD, in particular as all residues directly involved in coordination are highly conserved in the p63 DBD. This slight destabilization might be the consequence of the unique region of 10 amino acid residues that is located just after the H1 helix in p53 with His 208 coordinating the Zn 2ϩ . Several attempts to stabilize the p53 DBD have been undertaken so far. Using DNA shuffling, 20 amino acid residues (positions 101-120) could be identified as responsible for a thermostable phenotype (46 (44) was achieved by semirational design of a quadruple mutant (M133L/V203A/N239Y/N268D), two homologous residues (Leu 133 and Ala 203 ) of which can be found in the p63 DBD as well. Matsumura and Ellington (45) created thermostable p53 variants by in vitro evolution and found that two of three stabilizing mutations reside in the p53 DBD (N239Y and N268D). Both of these mutations were also found by Nikolova et al. (44), but they are not present in the p63 DBD.
The stability of globular proteins in solution depends on several factors. Generally, the stabilization of proteins is due to only a few molecular interactions (80). A statistical comparison of the p63 and p53 DBDs revealed no major differences. However, the p63 DBD has a higher portion of hydrophobic residues that might stabilize the hydrophobic core and shows a lower instability index (69). The p53 and p63 DBDs probably have a similar overall fold with minor structural differences. By building a model of the p63 DBD, it could be shown that the p63 and p53 DBDs differ mainly in the above-mentioned region of 10 amino acid residues that is located after the H1 helix in p53. We have been able to highlight a particular conformation of an extended hairpin loop on the surface of the p63 DBD. Moreover, the significance of the H1 helix region for the putative dimerization interface in the tetrameric p53⅐DNA complex has been previously suggested (see Fig. 7B in Ref. 33). Studies on the overall structure of the p53 DBD⅐DNA complex proposed that DNA bends to avoid steric clashes in this interface when four FIG. 9. Schematic model for the DNA-binding properties of the p63 and p53 DBDs and the respective GST fusion proteins. The p53 DBD (black circles) contains a functional dimerization interface putatively formed by the H1 helix (represented by gray rectangles) and is capable of cooperative binding to its consensus DNA. The p63 DBD (white circles), in contrast, does not bind to DNA cooperatively due to a different or lacking dimerization interface (represented by gray rectangles). Artificial dimerization mediated via fusion with GST (gray circles) restores the DNA-binding activity of GST-p63 DBD. GST-p53 DBD and GST-p63 DBD bind to DNA with similar affinity and specificity. The arrows indicate a further putative intermolecular oligomerization interface.
p53 DBDs are bound to DNA (64). The particular constitution and orientation of that hairpin motif in the p63 DBD may reduce cooperativity of DNA binding (see Fig. 9), e.g. by interfering with the H1 helix and/or preventing conformational rearrangements necessary for dimerization. As our model does not integrate the first amino acids and the C-terminal domain of the full-length p63 protein, the thermostable character and the dimerization behavior are perhaps not fully reflected by a partial three-dimensional structure of the DBD. Nevertheless, the hydrophobic character of the protein in conjunction with the packed conformation of the extended hairpin motif may provide a way to explain the enhanced (thermo) stability and the lack of cooperative DNA binding of the p63 DBD. It is tempting to speculate how this distinction in the potential dimerization interface in the DBDs influences the DNA-binding properties of the tetrameric full-length p63. Based on the sequence alignment and the modeled structure, it will be possible to design chimeric p53 DBDs possibly lacking dimerization properties or, vice versa, p63 DBDs displaying dimerization properties of p53 for biochemical and cellular studies.
Knowledge of the molecular basis of thermostabilization of the p63 DBD might allow the identification of essential residues and make it feasible to engineer more thermostable p53 DBDs, e.g. for the application of highly stabilized p53 variants in gene therapy. The p63 protein is possibly the most ancient member of the p53 family (7,16). The enhanced in vitro thermostability of the p63 DBD might reflect the evolutionary development and the different in vivo functions of p53 and p63 ("regulation of function through protein stability" (13)). Little cellular data on the cellular stability of the p63 isoforms are available (81); however, p63 probably shows a comparable long half-life due to its more constitutive physiological function. In contrast, the tumor suppressor p53 is present only in low concentrations and shows a short half-life under physiological conditions. Following several stress signals, p53 is activated through post-translational mechanisms and rapidly accumulates (82,83). Consequently, the p53 protein has to be tightly regulated due to its potential "toxicity" for undamaged cells. One mean to achieve this might be the low thermodynamic stability of the p53 DBD that leads to a rapid turnover of inactive unfolded p53 protein mediated by ubiquitination and proteolysis. It is in this line of evidence that MDM2 binds to the N-terminal part of p53, thereby inhibiting the transcriptional activity and targeting p53 for degradation by the ubiquitinproteasome pathway (84 -87), whereas p63 levels are independent of MDM2, and neither MDM2 nor MDMX is capable of binding to or targeting p63 for degradation (88).
Very recently, it was demonstrated that certain p63 isotypes form complexes with wild-type p53 and that these interactions are mediated by both DBDs (81). Other experiments support a direct interaction of tumor-derived mutant p53 DBDs with p63 and p73 (89,90). The availability of highly purified and characterized p63 and p53 DBDs makes it now possible to study the basis of these reported interactions.
In summary, the characterization of a new isoform of the DBD of the p53 homolog p63, p63-␦, revealed several novel features. The free p63 DBD does not bind to DNA cooperatively. However, dimerized GST-p63 DBD shows comparable affinity and specificity for representative p53 consensus sites as GST-p53 DBD. In comparison with p53, the p63 DBD is thermodynamically remarkably stabilized. These results might support biochemical and structural studies to elucidate the molecular basis for the lack of cooperative DNA binding and the high thermostability of the p63 DBD in order to gain further insight into the function of the p53 family members in tumorigenesis and development.