Analysis of the Binding of p53 to DNAs Containing Mismatched and Bulged Bases*

The tumor suppressor protein p53 modulates cellular response to DNA damage by a variety of mechanisms that may include direct recognition of some forms of primary DNA damage. Linear 49-base pair duplex DNAs were constructed containing all possible single-base mismatches as well as a 3-cytosine bulge. Filter binding and gel retardation assays revealed that the affinity of p53 for a number of these lesions was equal to or greater than that of the human mismatch repair complex, hMSH2-hMSH6, under the same binding conditions. However, other mismatches including G/T, which is bound strongly by hMSH2-hMSH6, were poorly recognized by p53. The general order of affinity of p53 was greatest for a 3-cytosine bulge followed by A/G and C/C mismatches, then C/T and G/T mismatches, and finally all the other mismatches.

Loss or alteration of p53 function occurs in about half of all human cancers, and the investigation of its role in prevention of tumor development has placed it in a key position within a complex network of interactions (reviewed in Ref. 1 and 2). Among the known functions of p53, one of the best characterized is the transcriptional regulation of certain genes in response to conditions threatening the integrity of the cell (reviewed in Ref. [3][4][5]. DNA damage caused by UV and X irradiation (6,7), hypoxia (8), depletion of ribonucleoside pools (9), the expression of several oncogenes (Ref. 10 and references therein), and other genotoxic conditions all lead to the arrest of cellular growth or apoptosis (11) through p53-mediated pathways (12). These events appear to be triggered by an activation and/or accumulation of p53 in the cell, but the detailed mechanisms are not fully understood. Growing evidence has revealed the importance of post-translational modifications such as phosphorylation and acetylation of p53 in this process (13,14). Modulation of p53 activity in most cases is influenced by changes in its DNA binding properties, including alterations in the regulation of binding by the C-terminal domain (reviewed in Ref. 15). Additionally, p53 binds its transcriptional response element as a tetramer, and disruption of this conformation leads to a decrease in binding activity (16).
p53 function in the cell is not limited to transactivation. Studies in vivo and in vitro have shown that p53 associates with proteins involved in DNA repair (such as XPB, XPD (17), and WRN (18)), in recombination (hRad51 (19)), and in repli-cation (DNA polymerase ␣ (20) and RPA (21)), thus pointing to the direct participation of p53 in these processes. This hypothesis has been strengthened by several studies in vivo suggesting an involvement of p53 in the control of recombination and repair (7,22,23).
Although it is generally assumed that p53 acts via mediator proteins in response to primary damage in DNA, there is growing evidence that p53 can bind directly to sites of DNA lesions. p53 can recognize and bind single-stranded (ss) 1 DNA and double-stranded (ds) DNA ends (24) in a sequence-independent manner. Both types of breaks represent intermediates of DNA damage, repair, and recombination. Additionally, p53 binds tightly in vitro to recombinational intermediates such as Holliday junctions and in doing so enhances their resolution by junction cleaving enzymes (25). p53 also binds three-stranded DNA structures (25,26). Finally, we demonstrated that p53 and its C-terminal fragment form stable complexes at sites of extra base bulges (insertion/deletion mismatches) in vitro (27). When two similar substrates that contain some nonhomologies undergo recombination, clusters of extra base bulges and single-base mismatches are left behind, and their presence must be signaled for repair. Extra base bulges and single-base mismatches may also occur as a result of replication errors or repair of DNA damage induced by irradiation or carcinogens. For example, the presence of lesions such as thymine dimers could result in the insertion of nucleotides opposite the noninformational site during translesion synthesis, thus leaving behind mispairs and bulges. Collectively, these data suggest that a direct interaction of p53 with such DNA lesions may be a relevant upstream event in the biological function of this protein. However, the role of DNA damage intermediates in activating p53 following DNA damage is not fully understood, nor is the full spectrum of lesions to which p53 may respond known.
Currently, two models for how p53 binding to sites of damage might alter events in the cell have been suggested. In one, p53 binding to lesions causes transcriptional activation of a cascade of genes leading to growth arrest or programmed cell death (28), but the mechanism of p53 transactivation of genes located at a great distance when tethered to sites of DNA damage was unclear (29). Recently, it has been shown that once bound to an extra base bulge in DNA, p53 can dissociate in the presence of ATP (30). It is possible that the protein dissociates in an altered form (for example, a structural alteration or a modification such as phosphorylation) and provides the appropriate signals at a distance from damage sites.
The second model proposes a more direct participation of p53 in the DNA repair and recombination pathways (26,29). The population of p53 in cells may not be homogeneous and may contain a mixture of molecules with different post-translational modifications (29). Thus, some forms of p53 could preferentially bind to response elements, whereas others may show a higher affinity for lesions in DNA. p53 molecules that bind to DNA lesions could, in turn, signal their presence to loci containing p53 response elements as well as recruit one or more proteins with which it interacts. It is also possible that in vivo, p53 exists primarily within different multiprotein complexes. Some of these could be specific for the recognition and repair of DNA damage and might include members of the mismatch repair protein family.
Given the growing evidence that p53 is a potential damage recognition protein, it has become important to survey a large number of lesions in DNA to determine the extent to which each is bound by p53. Some lesions may directly induce p53 binding, whereas other lesions may be recognized by different proteins that act as a signal for the p53 response. Determination of the spectrum of lesions to which p53 does or does not bind directly is thus critical to further development of models that explain the role of this protein in the cellular response to DNA damage. Although binding of p53 to extra base bulges has been examined, its ability to directly recognize single-base mismatches has not. Dudenhoffer et al. (26) designed DNA substrates consisting of a recombinational intermediate with single-base mismatches and observed p53 binding to these structures when they contained A/G but not G/T mismatches. However, the complexity of these structures made it difficult to determine the influence of the mismatch component alone. Therefore, analysis of p53 binding to single-base mismatches in otherwise simple linear DNA is crucial. To address this need, we generated duplex DNA substrates containing all possible single-base mismatches and have examined their interactions with p53.
Single-base mismatches and extra base bulges are repaired by the mismatch repair system. A number of these mismatches are recognized by the Escherichia coli mismatch repair protein, MutS and its eukaryotic homologs (MSH2, MSH3, and MSH6) with differing affinities found for the various mismatches in vitro (Ref. 31 and references therein). p53 and the mismatch repair proteins share many similar functions such as the ability to recognize extra base bulges (27,32,33) and Holliday junctions in vitro (25,34) as well as inhibition of DNA recombination upon encountering mismatches in vivo and in vitro (26,35,36). To place the in vitro p53 binding studies within the context of repair events in the cell, the affinity of p53 and hMSH2-hMSH6 for the various mismatches was compared.

EXPERIMENTAL PROCEDURES
DNA Substrates and Proteins-Double-stranded DNA substrates (49 bp) were made by annealing complementary single strand oligonucleotides (high pressure liquid chromatography purified from Life Technologies, Inc.). The duplex DNA substrates were end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase (New England Biolabs Inc.). Benzoylated naphtoylated DEAE-cellulose (Sigma) was used to remove any contaminating ssDNA (37). The substrates were desalted using Sephadex G50 spin columns (Amersham Pharmacia Biotech) and then purified on 15% nondenaturing polyacrylamide gels. Human p53 protein was overexpressed in SF9 cells using a vector provided by Dr. Arnold Levine and purified as described previously (38). The hMSH2-hMSH6 heterodimer was purified as described previously (39) using a vector provided by Dr. Richard Fishel.
Filter Binding Assays-For filter binding studies, reactions (20 l) were carried out in a DNA binding buffer containing 20 mM Hepes (pH 7.5), 10% glycerol, 50 mM KCl, 0.1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride) for 20 min at room temperature. The amount of DNA substrate, competitor DNA, and protein are specified in the figure legends. The samples were applied to presoaked nitrocellulose proteinbinding filters (Protran; Schleicher & Schuell) at a flow rate of 1 ml/min by vacuum suction. Each filter was washed twice with the DNA binding buffer and air-dried for 30 min. Radioactivity was measured by scintillation counting. Binding of the DNA alone to the filters in the reaction mixtures (protein-free background binding) was less than 5%. The amount of p53-bound DNA was calculated as the ratio of the cpm/filter (minus the protein free background) to the total cpm of the reaction mixture.
Gel Retardation Assays-The standard assay for p53 and hMSH2-hMSH6 binding was performed in a buffer optimized for hMSH2-hMSH6 studies (39) containing 25 mM Hepes (pH 7.5), 15% glycerol, 50 mM NaCl, 1 mM dithiothreitol, 0.01 mM EDTA in a 20-l reaction volume. The concentration of DNA substrates, protein, and unlabeled competitor DNA are indicated in the figure legends. Unlabeled lesionfree dsDNA ( Fig. 1) was used as nonspecific competitor DNA. The reactions were incubated at room temperature for 20 min unless otherwise indicated. Upon completion, the samples were separated on 5% polyacrylamide gels in 0.5ϫ TBE buffer (44.5 mM Tris, 44.5 mM boric acid, 2 mM EDTA) at 200 V at 4°C for 2 h. Gels were dried and quantified using a Storm 840 PhosphorImager (Molecular Dynamics).

Filter Binding Analysis Reveals p53
Binding to Some Mismatches in Duplex DNA-The affinity of p53 for various forms of damage in linear duplex DNA typical of those generated during replication or recombination needs to be determined to more fully understand how p53 responds to genetic insult of this form. To carry out these experiments, we designed a series of linear 49-bp DNA templates (Fig. 1). The lesion-free DNA contains no damage and was based on a template described previously (27). Variants were constructed containing all mismatches located centrally: p53 Binding to DNAs Containing Mismatched and Bulged Bases clusters of mismatches are frequently left behind following recombination between two similar but not identical DNAs. As a positive control for p53 lesion binding, a template containing a single 3-cytosine bulge was synthesized. Two additional controls for testing the DNA binding properties of p53 were also designed. The first template contained the full 20-bp human GADD45 p53 response element (41) at the center, and the second contained the same response element with 4 nucleotides shown to be particularly important for p53 binding substituted with other bases. Because ssDNA alters p53 binding (28,42), the duplex templates were first incubated with benzoylated naphtoylated DEAE-cellulose and then purified from nondenaturing 15% polyacrylamide gels to remove any contaminating ssDNA. For all the studies below, a single buffer previously optimized by Gradia et al. (39) for hMSH2-hMSH6 binding studies was employed. Comparative studies showed no significant difference (data not shown) when this buffer was used for p53 as contrasted to buffers containing magnesium (43), whose inclusion raised the potential complication of the activation of the putative exonuclease activity in the p53 (44).
Filter binding assays were employed to evaluate p53 binding to these substrates. The human p53 used here was produced in SF9 insect cells, and analysis of this protein indicated that there was at least one phosphorylated site. 2 End-labeled DNA (1 nM) was incubated with p53 (15 nM of tetramers) for 20 min in DNA binding buffer ("Experimental Procedures") at room temperature in the presence of a 15-fold excess of unlabeled nonspecific competitor DNA. The results (Fig. 2) showed that p53 exhibits different affinities for substrates with different mismatches. p53 showed low binding to the negative controls: 4.2% of the lesion free and 3.8% of modified GADD45 templates were protein bound. The 3-cytosine bulge substrate, which serves as a positive control for lesions, showed the high binding (14% of the probe was bound by p53). Similarly, p53 bound extremely well to the full GADD45 response element (38% of binding). For the DNA containing A/G mismatches, 5.8% of the probe was bound by p53, whereas binding to DNA with C/C was 7.1%. Binding of p53 to DNA containing the other mismatches was lower than 4%. When higher concentrations of p53 were used, a small increase in binding was observed (except for C/C); however, the relative order of affinity remained the same: (Table  I). It should be noted that a titration of p53 on these probes in the absence of nonspecific competitor showed higher levels of binding (data not shown).
Gel Retardation Analysis Confirms p53 Binding to Some Mismatch Containing DNAs-Although filter binding assays provide a comparison of the total amount of bound DNA with the different substrates, this method does not allow direct visualization of protein-DNA complexes. A number of authors have documented nonspecific binding of p53 to DNA (Refs. 24 and 45; reviewed in Ref. 29). To evaluate specific mismatch-dependent interactions of p53 to these substrates, gel retardation assays were performed.
The multi-staged process of substrate purification and the necessity of using different single-stranded oligonucleotides to produce the double-stranded substrates excluded the possibility of having both the same amount of radioactivity and the same concentration of DNA for the different substrates. Therefore, binding of p53 to mismatches was tested in two different ways. First, probes were added to the reactions so that they contained the same amount of cpm for each substrate. As mentioned above, this required the addition of different concentrations of DNA. Therefore, different amounts of p53 were added to each probe to ensure a 1:10 molar ratio of DNA to protein. The DNA substrates and p53 were incubated for 20 min at room temperature in the presence of a 20-fold excess of nonspecific unlabeled competitor DNA followed by gel retardation assays ("Experimental Procedures"). p53 formed complexes with some of the mismatched substrates as seen by the retarded mobility of the probes, and complex formation varied depending on the type of the mismatch. As shown in Fig. 3, binding was greatest for DNA with a 3-cytosine bulge (Fig. 3A,  lane13) and progressively less for the modified GADD45 sequence (lane 14), followed by the C/C (lane 10) and C/T containing substrates (lane 11). Finally, binding to the G/T (lane 12) and A/C (lane 9) containing DNAs was weak but greater than the lesion-free DNA, which showed no binding (lane 8). p53 also showed modest binding to the modified GADD45 probe. Quantitative phosphorimaging analysis shows that p53 binding to the 3-cytosine bulge was 48-fold greater than lesionfree DNA, whereas binding to DNA with the modified GADD45 sequence was 11-fold higher than lesion-free DNA (Fig. 3B). The fold increase in binding for the C/C, C/T and G/T containing substrates as compared with the lesion-free DNA was 8.7, 6.0, and 4.6, respectively. The A/C containing substrate behaved similar to the lesion-free DNA.
Using a second approach, experiments were set up such that the concentration of DNA in each of the reactions was the same, but because of the reasons described above, the amount of radioactivity for each substrate was not the same. Each  Filter binding assays were carried out using mismatch containing substrates A/C, A/G, G/T, C/C, 3-cytosine bulge (BC), and lesion-free (WT) DNA (1 nM (Fig. 4) shows a series of distinct slow migrating bands corresponding to the position of the p53-DNA complex (Fig. 4A, lane 11, arrows a, aЈ, and aЉ). In addition, there was a smear of faster migrating material (Fig. 4A, bracket c) that probably represents nonspecific complexes of monomeric p53 with the substrates. These gel retardation assays were carried out multiple times (n ϭ 3), and protein-DNA complexes were quantified. The level of p53 binding to mismatches was compared with the binding to the wild type substrate. The results showing the highest binding were seen with the substrate containing a C/C mispair, followed by A/G, and finally A/C and T/T. DNA containing C/T, G/T, A/A, and G/G mismatches showed binding that was lower than the wild type DNA. As expected, the highest binding was seen with the 3-cytosine bulge. Over the course of these experiments several different preparations of p53 were utilized. Even though the differences between binding to the various mismatches were not great in some cases, the hierarchy of binding to the different substrates was the same among the different p53 and DNA preparations.
p53 Binding to Lesion-containing DNA Is Concentration-dependent-To ensure that the relative affinities of p53 for the different substrates described above were not due to fortuitous choices of DNA or p53 concentrations, binding of the protein in the gel retardation assays was followed as a function of DNA concentration. The concentration of DNA substrates was increased from 0.75 to 7.7 nM, whereas the concentration of p53 was kept constant at 12 nM. Fig. 5 shows representative binding curves from some of the mismatches tested. In all cases, the lesion-free substrate was included as a control. Binding was modeled by PRISM software (GraphPad Inc.), assuming only one binding site (a cluster of mismatches) per DNA template. p53 binding to DNA containing the GADD45 site increased by 5-fold between the lowest and highest concentration of substrate tested (data not shown). At equal concentration of DNA ligands, the amount of bound DNA was higher for the substrates containing the A/G (Fig. 5A), the 3-cytosine bulge (Fig.  5B), and C/C (data not shown) lesions than for the lesion-free DNA. Substrates containing A/C, C/T, G/G, and T/T mismatches showed binding curves similar to the lesion-free DNA (curves for A/C and C/T shown in Fig. 5A). On the other hand, substrates containing G/T and A/A mismatches showed binding curves lower than lesion-free DNA (Fig. 5B). Although we used a mathematical model describing one-site binding to generate the binding curves, this model does not completely fit the experimental conditions. Specific p53-DNA complexes could be affected by factors such as the presence of DNA ends, which could provide additional sites that might compete for binding. Thus, efforts to calculate association/dissociation constants for p53 based on one-site binding models may not be applicable in this case.
A similar series of gel retardation assays were carried out in which the DNA concentration was held constant while the p53 concentration was varied. The data shown in Fig. 6 indicate that the rate of increase in p53 binding for the four substrates examined was different. For example, changing the p53 concentration from 12 to 40 nM led to a 4.3-fold increase in binding to the lesion-free substrate and a 3.2-fold increase in binding of the substrate containing the A/C mismatch. In contrast, 8.9and 7.2-fold increases were observed for the A/G and C/C substrates, respectively. The binding curves appear to be bi-phasic with little increase between 3 and 20 nM followed by a marked rise when the concentration was increased from 20 to 40 nM. Indeed, at 28 and 40 nM p53, very significant differences were seen in binding to the A/C and lesion-free substrates in contrast to the substrate containing the C/C mismatch. Consistent with the results above, the affinity of p53 for the lesion-free DNA was not the lowest observed; rather the substrate with A/C mismatches bound p53 even less effectively. Such effects have been observed in other studies (see "Discussion").
The Stability of p53-DNA Complexes Varies Depending on the Lesion in the DNA Substrate-The stability of the p53-DNA complexes can be probed by competition analyses using increasing concentrations of nonspecific competitor DNA. Here, equal amounts of radiolabeled substrates were incubated with p53 for 20 min at room temperature, followed by addition of increasing amounts of unlabeled competitor DNA for a further 15 min. As shown in Fig. 7, when a 200 fold excess (500 ng) of competitor DNA was added, p53 binding to the GADD45 and 3-cytosine bulge containing substrates was reduced by less than 2-fold (from 83 to 58% and from 84 to 45% of the DNA bound, respectively). In contrast, under the same conditions p53-DNA complexes were reduced to ϳ3% when competitor was added to DNAs containing A/C or G/T mismatches or the lesion-free DNA, arguing that these complexes are much less stable. When p53 was incubated with DNA substrates containing A/G or C/C mismatches followed by addition of a 200-fold excess of competitor, 8 and 13% of the DNA remained in complex with p53. Thus, the hierarchy of complex stability was as follows: the highest for the GADD45 and 3-cytosine bulge containing substrates, then DNAs with A/G or C/C mismatches, and lowest for A/C and G/T containing substrates, which is in agreement with the results above.
p53 Has a Higher Affinity for Some Mismatches than Does hMSH2-hMSH6 -The work above and previous studies (27,26) provide evidence that p53 can directly recognize certain single-base mismatches and bulges in DNA. hMSH2-hMSH6 is a heterodimeric protein involved in the primary recognition steps of mismatch repair in eukaryotic cells (46,47). The observations made here would be of particular importance were it found that the relative affinities of p53 and hMSH2-hMSH6 under similar experimental conditions were comparable for at least some of the substrates examined. The incubation buffer used here and in the experiments above for p53 is one previously optimized for hMSH2-hMSH6 binding studies by Gradia et al. (39). The mismatched substrates were incubated with p53 or hMSH2-hMSH6 for 20 min at room temperature at the same DNA:protein molar ratio (1 DNA molecule/35 p53 tetramers or hMSH2-hMSH6 heterodimers) in the presence of nonspecific unlabeled competitor DNA. Using equal DNA concentrations of each probe, the proteins were incubated under the same binding conditions ("Experimental Procedures"). The affinity of hMSH2-hMSH6 for the 3-cytosine bulge was similar to that of p53 (Fig. 8, A, compare lane 13 with lane 6, and B). hMSH2-hMSH6 recognizes G/T containing substrates more strongly than p53 (Fig. 8, A, lanes 5 and 12, and B) as expected because hMSH2-hMSH6 is known to have the strongest affinity for G/T mismatches as compared with all other mismatches (summarized in Ref. 31). Indeed, as seen in Fig. 8, hMSH2-hMSH6   FIG. 5. Affinity of p53 for different mismatched substrates. Increasing amounts of mismatch containing DNAs were incubated with a constant amount of p53 in the absence of nonspecific competitor DNA. The protein-DNA complexes were separated on nondenaturing polyacrylamide gels, and the amounts of free and bound probe were quantified using a PhosphorImager. The curves were drawn using Prism software as described in the text. A shows the binding curves for substrates containing A/C, A/G, and C/T mismatches. B shows the binding curves for substrates containing G/T, 3-cytosine bulge, and A/A lesions. WT, wild type.

p53 Binding to DNAs Containing Mismatched and Bulged Bases
showed much lower binding to the other mismatches. Among the other mismatches, hMSH2-hMSH6 only showed higher binding to the A/C mismatch as compared with p53. All the remaining mismatches tested were bound with higher affinity by p53 than hMSH2-hMSH6 (Fig. 8B).

DISCUSSION
Determination of those lesions and unusual structures in DNA that generate signals for p53 binding is central to further understanding the response of this protein to cellular DNA damage. This is of particular importance because we previously showed that p53 strongly recognizes Holliday junctions and insertion/deletion mismatches in the form of extra base bulges (27,25). In this study, we examined the ability of purified human p53 protein to bind to single-base mismatches in linear DNA. Additionally, the affinity of p53 for the various mismatches was compared with the human mismatch repair protein, hMSH2-hMSH6, under the same conditions.
The mismatched substrates we used contain a cluster of three mismatches separated by 2-4 bp. Our rationale for using a cluster of mismatches was severalfold. First, such clusters were found by Alani et al. (40) to provide stronger signals for yMSH2 binding than single-base mismatches. Second, we assumed that if no binding was seen to a 3-mismatch cluster, then there would be little reason to probe binding to a single mismatch alone. Third, such clusters are commonly found following homologous recombination through similar but not identical DNAs, making them an inherently interesting lesion in DNA. It is possible that for some of the mismatches, the close proximity of three mismatches could destabilize the helix and create a transient single stranded bubble to which p53 would bind. However, were this the case, these bubbles should create strong binding sites for hMSH2-hMSH6 and just the opposite was observed; the mismatches bound best by p53 were not recognized by hMSH2-hMSH6.
Filter binding and gel retardation assays were carried out in several different ways to examine the affinity of p53 for linear DNAs containing single-base mismatches. Although the results varied slightly among the approaches, they were in good agreement and revealed the same pattern of p53 binding to the different substrates. As a positive control and to generate a link with the hMSH2-hMSH6 results, binding was compared with a substrate containing a single 3-cytosine bulge. p53 and hMSH2-hMSH6 showed strong and roughly equal affinities for this lesion as seen by gel retardation and filter binding assays and matched previous reports (summarized in Ref. 31). In all of the assays, p53 reproducibly showed the highest affinity for DNAs containing A/G and C/C mismatches, and this was always higher than binding to the lesion-free DNA. Competitor titration (Fig. 7) also showed that the p53 complexes formed on substrates with A/G and C/C mismatches were more stable than ones formed on templates containing other mismatches or with the lesion-free DNA. Two of the substrates containing G/T and C/T mismatches did not show significant p53 binding above the lesion-free DNA level (but were never lower). It is interesting to note that Dudenhoffer et al. (26) demonstrated that p53 had a higher affinity for a recombinational intermediate substrate containing an A/G mismatch than ones lacking a mismatch or with A/C and G/T mismatches. Although they did not examine all possible mismatches as described here and their substrates were relatively complex, our results are nonetheless in good agreement with their findings.
Substrates with A/C, A/A, G/G, and T/T mismatches showed binding that generally did not exceed that of the lesion-free DNA, and in several experiments, these substrates showed levels of binding that were reproducibly lower than the control . Arrows indicate the products of the reaction: arrows a, aЈ, and aЉ, p53-DNA complexes; arrow b, hMSH2-hMSH6-DNA complexes; arrow c, free dsDNA substrates. The hMSH2-hMSH6-DNA complexes have a lower molecular mass than the p53-DNA complexes. B, specific DNA-protein complexes were quantified as described under "Experimental Procedures." The y axis represents the percentage of DNA bound. WT, wild type.
DNA. The observation that certain base/base mismatches present a particular sequence context can lead to reduced protein binding was noted previously in our studies with the recA protein (48). Binding of recA to linear duplexes containing C/C, G/G, T/T, and C/T mismatches was depressed relative to the lesion-free DNA (48). Further, Dudenhoffer et al. (26) noted that the binding of p53 to complex recombinational intermediates containing mismatches was lower when the DNA contained a C/T mismatch in contrast to homoduplex DNA. The explanation remains unclear but points to the importance of sequence context in the recognition of both transcriptional activator sites as well as lesions.
The potential biological significance of p53 binding to the different mismatches was examined by comparing its affinity for these lesions to that of the human mismatch repair complex, hMSH2-hMSH6 (Fig. 8). The substrate containing the 3-cytosine bulge was used as a positive control for both proteins, and our results show that the affinity of p53 for this lesion equals that of hMSH2-hMSH6. In agreement with previous reports (31), hMSH2-hMSH6 binds G/T mismatches with a markedly higher affinity than any other single-base mismatch, and this was found to be equal to its affinity for the 3-cytosine bulge. All other single-base mismatches are recognized by hMSH2-hMSH6 within a range that is severalfold lower than the G/T mismatch (Fig. 8). Nevertheless, these mismatches are still signaled for repair by the mismatch repair proteins. Therefore, the lower binding of hMSH2-hMSH6 to the mismatches must be within a biologically significant range. p53 binding to all the single-base mismatches fell in the lower range exhibited by hMSH2-hMSH6. However, p53 had higher affinity for several mismatches than hMHS2-hMSH6. These data suggest that the binding of p53 for several of the mismatches (A/G and C/C) is potentially biologically relevant because they lie within the range of hMSH2-hMSH6 binding. Clearly, in vivo, these affinities will be altered by modifications to the proteins (below) and by the binding of other proteins. Nonetheless, the affinity of p53 alone for these lesions, as measured here, provides a needed base line for understanding whether or not these proteins have significant potential to respond to such damage in vivo.
The DNA binding properties of p53 are influenced by posttranslational modifications such as phosphorylation and acetylation (12). The p53 and hMSH2-hMSH6 proteins we used were generated in insect cells, a common source of p53 or hMSH2-hMSH6, and both have been used in previous published work on mismatch and Holliday junction recognition (27,25,39). It has been documented in several recent papers that the posttranslational status of p53 may influence its binding properties (13,14). Analysis of the phosphorylation state of the baculovirus expressed p53 used here by two-dimensional gel electrophoresis 2 indicates that it has at least one phosphorylated site. It was essential that we first carry out studies using proteins purified from insect cells as we and others have done before to establish a foundation of understanding how p53 recognizes different lesions in DNA and to compare it with similar studies with hMSH2-hMSH6. With this study completed, it will be important in the future to determine how individual site-specific modifications of p53 alters its affinity for different lesions in DNA. Additionally, the effect of other proteins on p53 binding to the various lesions also needs to be examined.