High Affinity Insertion/Deletion Lesion Binding by p53

In addition to binding DNA in a sequence-specific manner, p53 can interact with nucleic acids in a sequence-independent manner. p53 can bind short single-stranded DNA and double-stranded DNA containing nucleotide loops; these diverse associations may be critical for p53 signal transduction. In this study, we analyzed p53 binding to DNA fragments containing insertion/deletion mismatches (IDLs). p53 required an intact central domain and dimerization domain for high affinity complex formation with IDLs. In fact, the C terminus of p53 (amino acids 293–393) was functionally replaceable with a foreign dimerization domain in IDL binding assays. From saturation binding studies we determined that the K D of p53 binding to IDLs was 45 pm as compared with a K D of 31 pm for p53 binding to DNA fragments containing a consensus binding site. Consistent with these dissociation constants, p53-IDL complexes were dissociated with relatively low concentrations of competitor consensus site-containing DNA. Although p53 has a higher affinity for DNA with a consensus site as compared with IDLs, the relative number and availability of each form of DNA in a cell immediately after DNA damage may promote p53 interaction with DNA lesions. Understanding how the sequence-specific and nonspecific DNA binding activities of p53 are integrated will contribute to our knowledge of how signaling cascades are initiated after DNA damage.

In addition to binding DNA in a sequence-specific manner, p53 can interact with nucleic acids in a sequence-independent manner. p53 can bind short singlestranded DNA and double-stranded DNA containing nucleotide loops; these diverse associations may be critical for p53 signal transduction. In this study, we analyzed p53 binding to DNA fragments containing insertion/deletion mismatches (IDLs). p53 required an intact central domain and dimerization domain for high affinity complex formation with IDLs. In fact, the C terminus of p53 (amino acids 293-393) was functionally replaceable with a foreign dimerization domain in IDL binding assays. From saturation binding studies we determined that the K D of p53 binding to IDLs was 45 pM as compared with a K D of 31 pM for p53 binding to DNA fragments containing a consensus binding site. Consistent with these dissociation constants, p53-IDL complexes were dissociated with relatively low concentrations of competitor consensus site-containing DNA. Although p53 has a higher affinity for DNA with a consensus site as compared with IDLs, the relative number and availability of each form of DNA in a cell immediately after DNA damage may promote p53 interaction with DNA lesions. Understanding how the sequence-specific and nonspecific DNA binding activities of p53 are integrated will contribute to our knowledge of how signaling cascades are initiated after DNA damage.
Exposure of normal cells to agents that damage DNA initiates a p53 signal transduction cascade, resulting in either cell cycle arrest or apoptosis (1). The interaction of p53 with DNA is thought to be critical for its signaling because the majority of tumor-derived forms of p53 have mutations in the central DNA binding domain (2), abrogating the ability of p53 to bind its consensus DNA sites (3). To date, many studies have focused on the sequence-specific interaction of p53 with DNA, including those that have identified downstream transcriptional targets and studies that describe post-translational modifications that activate p53 consensus site binding.
The ability of p53 to function as a transcriptional activator is believed to be integral for its growth-suppressive properties (4,5). Sequence-specific transactivation is one of the most well understood biochemical activities of p53. After cellular stress such as DNA damage (6), hypoxia (7), viral infection (8), or activation of oncogenes such as ras (9) and myc (10), p53 becomes transcriptionally active. Once active, p53 induces, among many genes, p21 (11), an inhibitor of cyclin-dependent kinases thought to be necessary for the p53-dependent G 1 /S cell cycle arrest (12)(13)(14)(15). The increase in p53-mediated transcriptional activity may be because of elevated levels of p53 in the cell (6,16) or increased sequence-specific binding ability (17,18). Post-translational modifications of p53, including phosphorylation by S and G 2 /M phase cyclin-dependent kinase-cyclin complexes (19,20), DNA-dependent protein kinase (21), protein kinase C (22), and casein kinase II (23), as well as C-terminal acetylation (24) have been found to enhance sequence-specific DNA binding in vitro. Furthermore, p53 has been shown to be acetylated at its C terminus after exposing cells to ultraviolet or ionizing radiation (18). This acetylation may be regulated by phosphorylation of the p53 N terminus by either the ATM kinase or DNA-dependent protein kinase (18,21,25,26).
In addition to binding DNA containing consensus sites, p53 can interact with nucleic acids in a sequence-independent manner. p53 can bind RNA (27), short single-stranded DNA (ssDNA) 1 (28 -30), and double-stranded DNA containing nucleotide loops (31); these diverse associations may be critical to p53 signal transduction. The ability of p53 to bind ssDNA is of interest because this form of DNA is an intermediate of both DNA damage and repair. Studies have correlated p53 signaling activation with both the timing and amount of DNA strand breaks. Nelson and Kastan (32) have shown that p53 levels increased after electroporation of enzymatically active restriction endonucleases into cells. Microinjection of single-stranded circular phagemid or circular DNA with a large gap into nuclei of normal human fibroblasts induced a p53-dependent G 1 arrest (33). Jayaraman and Prives (34) reported stimulation of p53 consensus site binding in vitro in the presence of short ssDNA fragments. Further proof that p53 may directly interact with damaged DNA was provided in a study reporting p53 binding to DNA fragments containing insertion/deletion lesions (31). Also, confocal microscopy studies have shown colocalization of p53 protein with sites of damaged DNA in histological sections of human skin exposed to UV light (35). Collectively, these reports suggest that p53 may be directly or indirectly regulated by DNA damage intermediates. A direct interaction of p53 with either DNA lesions or with proteins that bind damage intermediates may be a relevant upstream event in the biochemical engagement of the protein.
In this study, we analyzed the interaction of p53 with DNA fragments containing insertion/deletion lesions. In contrast to many studies that have shown p53 C-terminal binding to ssDNA or DNA lesions, we demonstrate that an intact central domain and dimerization capability are required for wild-type (wt) human p53 binding to IDLs. The results of our binding analyses demonstrate that the affinity of p53 for DNA fragments containing either an IDL or a consensus site is in the pM range. Competition binding assays revealed that p53-IDL complexes were dissociated with relatively low concentrations of consensus site-containing DNA. However, the number and availability of each DNA site immediately after DNA damage may promote p53 binding to DNA lesions in lieu of sequencespecific DNA binding.
p53 expression vectors (4) were transformed into Saccharomyces cerevisiae strain YPH681. Expression was induced by substituting galactose for dextrose in liquid cultures. Protein extracts were obtained by adding DBB and glass beads to the yeast pellets. Yeast were lysed using a bead beater, and extracts were clarified by centrifugation at 12,000 ϫ g for 10 min at 4°C.
DNA Binding Assay-To study the p53-IDL interaction, we used an in vitro protein-DNA binding assay developed by McKay (36) that allows for quantitative analysis. A monoclonal antibody that recognizes the N terminus of p53, PAb1801, was chemically cross-linked to protein A-Sepharose (PAS) with 52 mM dimethyl pimelimidate (Pierce); this antibody does not interfere with the oligomerization or DNA binding ability of p53. For assays using murine p53 (N-terminally tagged with six histidine residues), a monoclonal Penta-His antibody (Qiagen) was cross-linked to Protein G-Sepharose (PGS). The antibody-PAS/PGS complex was added to yeast or baculoviral protein extracts and mixed end-over-end for 1.5 h at 4°C in 250 l of DBB. Immunoprecipitated p53 was washed once with DBB, followed by a 5 min end-over-end wash with 0.5 M NaCl in Buffer B (5X Buffer B contains: 100 mM Tris-HCl pH 8.0, 5 mM EDTA, 50% glycerol), and finally by a 5 min end-over-end wash in DBB. Subsequent analysis of these samples by SDS-PAGE and silver staining showed immunopurification of p53 to ϳ95-98% homogeneity. The immunopurified protein was then mixed end-over-end for one h with 3.5 fmoles of [ 32 P]-end-labeled DNA fragments in 250 l of DBB for 1 h at room temperature. After three washes with DBB, protein components of the complex were digested with SDS/Proteinase K (VWR Scientific) in TE8 (20 mM Tris pH 8.0, 10 mM EDTA). DNA was phenolchloroform extracted, ethanol precipitated, and electrophoresed in 1ϫ Tris-acetate/EDTA on a 10% PAGE at 100 V. Gels were fixed in 5% methanol, 5% acetic acid before drying and exposure to film. DNA was quantified using a 445 SI PhosphorImager (Molecular Dynamics) and an Instant Imager (Packard Instruments). Alternative processing of the protein-DNA complexes for protein detection involved adding Laemmli sample buffer to the final complex and subjecting samples to 10% SDS-PAGE electrophoresis. Gels were stained with GelCode blue stain reagent (Pierce).
Saturation Binding Assays-DNA binding assays were performed as outlined above using 0.1 pmol of wt p53 protein immunopurified from Sf9 cells and the indicated amounts of consensus or C 3 DNA fragment. For each DNA input, the nonspecific binding component was determined by performing side-by-side assays with the PAb1801-protein A-Sepharose complex incubated with protein from mock-infected Sf9 cells. After final washes, the tubes were counted in a scintillation counter.
Competition Binding Assays-DNA binding assays were performed as outlined above. The equilibrium DNA binding condition of C 3 was used (0.1 pmol of p53, 22 pM C 3 DNA). After counting the initial bound cpm, radiolabeled C 3 fragment bound to p53 was competed by the addition of the indicated amounts of unlabeled DNA in 250 l of DBB. After 1 h of incubation at room temperature, complexes were washed three times with DBB, and the tubes were counted in a scintillation counter.
Prism software (GraphPad) was used for analyses of binding data. The percent incorporation of each radiolabeled DNA fragment, the nonspecific binding component, and radioactive decay rate of the DNA fragments were corrected for in the analyses.

RESULTS
To study p53-IDL interactions, we used an in vitro protein-DNA binding assay developed by McKay (36) that allows for quantitative analysis. The DNA lesions in this study were the same that Lee et al. (31) used to demonstrate p53 binding to IDLs through gel shift and electron microscope analyses. Random sequence, double-stranded DNA fragments with either one (C 3 ) or three (3C 3 ) sets of triplet cytosines in one of the strands were used; the extra bases caused a nucleotide loop in the DNA duplex.
Full-Length p53 Binds to DNA Fragments Containing IDLs-p53 protein produced in either baculoviral or yeast overexpression systems bound to DNA fragments containing IDLs at a level proportional to the number of triple cytosine loops in the DNA (compare middle and right panels of Fig. 1). Comparable p53 binding activity was also seen using DNA fragments of different sequence context that contained a triple cytosine loop (data not shown). p53 did not bind to a control DNA duplex without triplet cytosines (see the first panel of Fig. 1). There was minimal DNA binding activity detected with control immunoprecipitates from yeast and Sf9 extracts lacking p53 protein (Fig. 1, see Protein Con lanes). The p53 273 tumor-derived mutant form of p53 was also tested in the assay and did not exhibit significant binding to DNA fragments containing IDLs.
Tumor-derived Mutant Forms of p53 Lack C 3 Binding Activity-The lack of IDL binding by the p53 273 protein prompted us to screen other p53 proteins with point mutations in the central domain. Four mutant p53 proteins representative of tumorderived forms were produced in yeast and immunopurified, and equivalent amounts of protein were analyzed in the binding assay (Fig. 2B). We found that all four mutant p53 proteins had Ͻ5% of the wt p53 binding activity to the C 3 IDL (Fig. 2A). These data suggest that full-length p53 requires an intact central domain to bind DNA fragments containing IDLs. However, these results do not rule out the possibility that the p53 C terminus, as a separate entity, can also bind IDLs, as previously reported by Lee et al. (31).
Deletion of the C-Terminal 40 Amino Acids of p53 Abrogates C 3 Binding-To study the role of the p53 C terminus in IDL binding, engineered proteins with successive deletions of the C terminus were overexpressed in yeast, and equivalent amounts of proteins were tested for C 3 DNA fragment binding (Fig. 3B). Up to 40 amino acids could be deleted from the C terminus of p53 without loss of protein binding to IDLs ( Fig. 3A; see 1-353).
In fact, deletion of 20 amino acids from the C terminus increased binding activity by more than 2-fold over wt p53 ( The loss of IDL binding observed with deletions of 50 or more amino acids from the C terminus suggested that either the intrinsic C-terminal sequence was required for IDL binding or the oligomerization domain, which is encompassed within residues 312-365 of human p53 protein (37), must be intact for this p53 activity. To determine which of these two properties was necessary for p53-IDL binding, we used chimeric proteins containing various C-terminal-truncated p53 fused to the coilcoil (CC) dimerization domain of the yeast transcription factor GCN4 (residues 249 -281) (38) for further analyses (Fig. 4A). p53-CC fusion proteins with deletions up to 100 amino acids from the p53 C terminus were able to bind IDLs (Fig. 4B). In fact, the p53-CC fusion proteins 293CC, 323CC, 333CC, 343CC displayed binding activities ϳ2-4-fold greater than wt protein. However, despite the enhanced binding seen with these fusion proteins, a single point mutation in the central domain of the p53 portion of the chimeric protein resulted in 90% loss of IDL binding (Fig. 4A, compare 343CC to 347CC 175 ). Fusion proteins with p53 C-terminal deletions greater than 100 amino acids (273CC and 283CC) had less than 10% wt binding activity; these p53 deletions disrupted the central DNA binding domain, which is encompassed by residues 100 -300 (39). Thus, the data suggest that full-length p53 binding to IDLs requires an intact central domain and that the entire C terminus of p53 is functionally replaceable with a foreign dimerization domain in these assays C 3 Binding Activity of Full-length p53 Versus the C-terminal Fragment-Previous studies have reported that the C terminus alone can bind IDLs (31). To determine the relative binding activities of full-length p53 and the C-terminal fragment, we assayed the C 3 binding properties of wt human p53 and three forms of murine p53 produced in Sf9 cells: full-length p53, amino acids 1-360, and amino acids 315-390 (40). Binding assays were performed with equimolar amounts of p53 protein incubated with the C 3 DNA fragment (Fig. 5). Murine and human wt p53 had equal C 3 binding activity. Similar to our results with human p53 C-terminal truncation mutants (Fig.  3A), partial deletion of the C terminus of murine p53 significantly lowered C 3 binding ability. Consistent with the results of Lee et al. (31), we found that the C-terminal fragment of p53 could bind C 3 , albeit with less than 10% full-length p53 IDL binding activity.
Saturation Binding of p53 to DNA Fragments Containing Consensus Sites or IDLs-The significance of p53 binding to its consensus DNA site is supported by many biological studies (41) and the report of nM affinities for p53 binding to derivatives of its consensus site (42). The lack of significant C 3 binding by tumorderived mutant forms of p53 suggests this activity, like the p53 consensus binding, may be biologically relevant. To determine whether IDL binding occurs in a physiologic range, we performed saturation binding assays to compare the affinity of full-length, wt human p53 for its consensus site with that for C 3 . The p53 binding site in the p21 promoter (nucleotides 2293-2332) was used for consensus binding analyses. Using 0.1 pmol of p53 protein and increasing amounts (0.5 to 750 pM) of labeled DNA, we determined that the K D for p53 binding to its consensus site was 31 pM (Fig. 6A), and the K D for p53 binding to the C 3 DNA was 45 pM (Fig. 6B). Transformation of the saturation binding isotherms into linear Scatchard plots facilitates the comparison of p53 binding to the two DNA fragments (Fig. 6C). Although the binding affinity and capacity of p53 for its consensus site are higher than those for C 3 , both K D values represent physiologically significant binding.
Competition of p53 DNA Binding-To confirm the relative affinities determined using saturation binding curves and to study the DNA binding property of p53 when both a consensus DNA site and IDL DNA site were available, competition binding assays were performed. Under equilibrium conditions, p53 was first bound to the radiolabeled C 3 DNA fragment, and increasing concentrations of either unlabeled consensus site, C 3 , or a random sequence dsDNA fragment were added. p53 was efficiently competed off of C 3 after addition of consensus site DNA with an EC 50 of 0.14 nM (Fig. 7). The EC 50 for the C 3 DNA fragment competition of prebound C 3 DNA was 0.50 nM (Fig. 7). In contrast, a random sequence dsDNA fragment was an ineffective competitor for C 3 binding. The EC 50 for this latter competition assay could not be accurately assessed because 50% competition was difficult to attain (Fig. 7). Also, in these assays, ssDNA was not an effective competitor as compared with C 3 (data not shown). These data demonstrate that p53 has a higher affinity for its consensus site as compared with an IDL site and will likely bind to the former if both DNA sites are present, with all other variables remaining constant.

DISCUSSION
In this study, we analyzed p53 binding to DNA fragments containing IDLs. We found that full-length p53 requires an intact central domain and dimerization capability to bind IDLs. Tumor-derived mutant forms of p53 lost IDL binding, and C-terminal truncation mutants still bound IDLs if the protein maintained dimerization capability. The pM dissociation constant that we observed for p53 binding to IDLs provides evidence for potential in vivo relevance.
Previously, the C terminus of p53 (amino acids 311-393) was shown to bind IDLs (35). Reed et al. (43) also find that the C terminus of p53 binds nonspecifically to dsDNA (43). These data contribute to a model suggesting that the p53 C terminus negatively regulates the sequence-specific binding activity of the protein. When the p53 C-terminal end is deleted, phosphorylated, acetylated, or bound to antibody, peptide, or singlestranded DNA, p53 binding to its consensus site is stimulated (18,24,34,44). Although we demonstrated that the C terminus of murine p53 could bind IDLs, an equimolar amount of murine p53 lacking the C terminus had greater affinity for IDLs than the C-terminal fragment alone.
Our finding that the C terminus of p53 is not required for IDL binding is in agreement with the study of Parks et al. (45) demonstrating that p53 binding to IDLs was unaffected by PAb421, a monoclonal antibody that binds the C-terminal amino acids 371-380. Also, Bakalkin et al. (28) suggest that the domain of p53 responsible for nonspecific DNA binding depends on the DNA substrate; the C terminus of p53 binds single-stranded ends of DNA, whereas the central DNA binding domain of p53 binds internal ssDNA segments (28). The identity of the p53 binding domain(s) for various forms of DNA damage may suggest distinct roles for p53 in DNA damagesignaling pathways or DNA repair.
FIG. 6. p53 affinity for DNA fragments containing IDLs. Binding assays were performed by incubating 0.1 pmol of p53 with increasing amounts of either A, p53 consensus site representing nucleotides 2293-2332 of the p21 promoter; or B, C 3 DNA fragment. K D represents the concentration of p53 required to reach half-maximal binding of the DNA. B max is the maximal binding. The equation used to plot the data represents a rectangular hyperbolic function, indicative of binding that follows the mass action law: Y ϭ B max ϫ X/(K D ϩ X). Scatchard analysis was perfomed on the saturation-binding isotherms, and the linear representation of the data is seen in panel C. Results are representative of three independent experiments. Competition assays were performed at equilibrium binding conditions of p53 binding to the C 3 fragment. After binding the 32 P-end-labeled C 3 DNA fragment, increasing concentrations of either unlabeled control, consensus site, or C 3 DNA fragments were incubated with the complex. EC 50 represents the concentration of competitor that reduced p53-DNA binding by 50%. Competition assays were analyzed using GraphPad Prism software. The equation used to plot the data is indicative of competition of binding to a single site: Y ϭ Bottom ϩ (Top-Bottom)/(1 ϩ 10 (X Ϫ EC50) ).
Several reports suggest that p53 is an important determinant in nucleotide excision repair (NER). Using cells derived from patients with Li-Fraumeni syndrome, Ford and Hanawalt (46) show that the efficiency of global NER was dependent on p53 status. Compared with cells heterozygous for p53, homozygous mutant p53 cells exhibited global NER deficiency; however, transcription-coupled repair was unaffected by the p53 status. The ability of p53 to bind RPA (47) and subunits of TFIIH (48), both of which are essential components of NER, suggests that p53 may play a direct role in NER. These protein associations are among those that define similarities between p53 and XPA, the damage recognition and binding component of NER (49). XPA binds DNA cooperatively with RPA (50) and once complexed with DNA, recruits TFIIH to the site of damage (51). p53 binding to IDLs may also be stimulated by RPA. XPA can bind to various damage lesions (49), and we have observed p53 binding to DNA fragments containing a cholesterol adduct (data not shown), an artificial DNA lesion used in in vitro DNA repair assays (52).
Although we show that p53 has higher affinity for its consensus site as compared with a DNA lesion, the temporal availability of p53 consensus binding sites must be considered. Wu and Levine (53) have reported that the p21 gene is the first measurable target of p53 transactivation after high dose UV irradiation. However, p53-mediated induction of p21 gene expression does not occur until 2 to 5 h post-irradiation. In agreement with this result, using in vivo footprinting, Chin et al. (54) did not observe significant DNaseI cleavage protection of the p21 promoter until 2 h after exposure of cells to 20 gray of ionizing radiation. After exposure of cells to 1 gray of ionizing radiation, 2 to 8 double-strand breaks/genome have been shown to occur (55). In separate studies, Ji et al. reported that treatment of cells with malondialdehyde (an endogenous product of lipid peroxidation) resulted in the formation of 300 M 1 G-DNA adducts/genome and subsequent elevation of p53 activity several hours later (56). Thus, immediately after exposure of cells to genotoxic agents, the number of DNA lesions in a cell would likely exceed accessible p53 consensus sites, and p53 binding to DNA lesions may occur. The pM dissociation constant that we report for p53 binding to IDLs is consistent with the physiological concentrations of both p53 and DNA lesions in the cell after DNA damage. However, once DNA repair is initiated, the availability of consensus sites would likely increase, and the higher affinity of p53 for these sites would shift the binding equilibrium.
Our understanding of the biochemical activities that are required for p53 tumor-suppressive activities has increased enormously in recent years. However, the role of DNA damage intermediates and the network of signaling pathways by which cells activate p53 in the overall response to DNA damage is not well defined. Determining if p53 is directly activated by DNA damage intermediates and understanding how the sequencespecific and sequence-nonspecific DNA binding activities of p53 are integrated will contribute to our knowledge of how human cells respond to a wide range of DNA lesions.