The N terminus of p53 regulates its dissociation from DNA.

It is important to gain insight into p53 DNA binding and how it is regulated. By using electrophoretic mobility shift assays and DNase I footprinting, we show that a region within the N terminus of the protein controls the dissociation of p53 from a p53-binding site. When p53 is bound by a number of N-terminal-specific monoclonal antibodies, its rate of dissociation from DNA is reduced, and its ability to protect a cognate site from DNase I digestion is increased. Moreover, greatly reduced dissociation is observed with p53 protein lacking the N-terminal 96 amino acids. By contrast, deletion of the C terminus does not affect p53 dissociation from DNA or DNase I protection. p53 protein expressed in and purified from bacterial cells displays markedly more instability on its consensus DNA-binding site than does p53 produced in insect cells, suggesting that post-translational modifications may affect the stability of the protein. Our results provide evidence that the N terminus of p53 possesses an auto-inhibitory function that is mechanistically different from the inhibitory region at the C terminus.

When a cell loses p53, through deletion or mutation of the gene, or through association of the gene product with any one of several viral or cellular proteins, it becomes vulnerable to tumorigenesis (reviewed in Refs. [1][2][3]. Wild-type p53 is a sequence-specific DNA-binding protein that can activate transcription of genes containing p53 response elements both in vitro and in vivo (reviewed in Ref. 4). Such response elements contain a repeat of the sequence 5Ј-RRRC(A/T)(T/A)GYYY-3Ј, which has been shown to be a consensus p53-binding site (5). The GADD45 (6), p21 (7), Mdm2 (8,9), cyclin G (10), Bax (11), IGF-BP3 (12), PIG3 (13), and 14-3-3 (14) genes are among a growing list of genes that contain p53 response elements in their respective promoters and are activated by induction of p53. These varied genes are likely to be points of contact between p53 and cellular processes. From the cell cycle inhibitory function of the cyclin-dependent kinase inhibitor p21 (15)(16)(17), to the apoptotic potential of Bax (11), and to the down-regulation of p53 protein expression by Mdm2 (9,18,19), it is clear that p53 protein activity affects cellular decision making.
The activity of p53 can be increased as a result of elevated protein levels, covalent modifications, association between p53 and other proteins, or combinations of these (reviewed in Refs. 2 and 3). Assembly of the transcriptional apparatus at a start site is facilitated by the presence of transcriptional activators such as p53. These activators are thought to overcome the rate-limiting step of transcription factor initiation by simultaneously binding DNA and interacting with one or more components of the general transcription machinery (20 -24). Because the ability of p53 to bind DNA sequence-specifically is tightly correlated to its tumor suppressor activity, it is biologically relevant to investigate events that may alter the affinity of p53 for DNA.
The sequence-specific DNA binding region of p53 is contained in the central conserved region of the protein (amino acids 100 -300) (25). This region is flanked on the C-terminal side by a tetramerization domain (amino acids 325-356) followed by a stretch of amino acids rich in basic residues. This C-terminal region is capable of DNA strand reassociation and DNA strand transfer activities (26,27) and demonstrates binding affinity for insertion/deletion DNA mismatches (28), Holliday junctions (29), and DNA damaged chemically or by ionizing radiation (30). Interestingly, the C terminus also contains a negative regulatory domain that may keep the tetramer poised in a state inactive for sequence-specific DNA binding. Phosphorylation of this region by protein kinase C (31,32) or casein kinase II (33), association with DnaK protein (Escherichia coli homologue of human Hsp70) (33-35), 14-3-3 protein (36), or the C-terminal specific monoclonal antibody PAb 421 (33,37) stimulates the p53 tetramer to bind DNA sequence-specifically with high affinity. Moreover, it has been demonstrated that PAb 421 is capable of restoring binding ability to some mutant p53 proteins in vitro (38) and in vivo (39). Deletion of the C terminus results in increased sequence-specific DNA binding by p53 (33), as does interaction of the C terminus nonspecifically with single-stranded DNA (40) or peptides derived from the C terminus (41,42).
The N terminus consists of two contiguous transcriptional activation subdomains (amino acids 1-42 and 43-63) (43,44) and an adjacent proline-rich domain (amino acids 62-91) containing five copies of the sequence "PXXP" that contribute to the apoptotic function of p53 (45)(46)(47). The N-terminal region is phosphorylated at several sites, many of which are modified in response to stress signals (reviewed in Ref. 3), and a number of interesting protein kinases have been shown to be involved. These include the c-Jun N-terminal kinases (48,49), the DNAactivated protein kinase (DNA-PK 1 ; Refs. 50 and 51), casein kinase I (52,53), the cyclin-dependent kinase-activating kinase (54), ataxia-telangiectasia mutated kinase (55)(56)(57), and ataxiatelangiectasia related kinase (58). It has been reported that phosphorylation of the p53 N terminus, as opposed to its C terminus, is compartmentally restricted to the nucleus (59).
It is possible that the p53 N terminus can be involved in interdependent interaction with the C terminus to regulate defined functions of p53. Sakaguchi et al. (60) reported that C-terminal acetylation requires phosphorylation of sites within the N terminus. Furthermore, phosphorylation of N-terminal Ser-15 by DNA-PK increases the recruitment of CBP/p300 (61), which can acetylate lysine residues within the C terminus. This acetylation can then lead to increased sequence-specific binding by p53 in vitro (62). Interestingly, phosphorylation of Ser-15 also weakens the interaction of p53 with Mdm2 (51) and with the basal transcription factor TFIID (63). Results concerning phosphorylation of Ser-15 might seem contradictory with regard to transcriptional activation, as tethering to transcription factors might appear to be an indispensable link to activity. However, this attests to the intricate and complex nature of p53 regulation by cellular factors.
Work with antibodies, often seen as mimetics of cellular proteins, has invoked interesting questions with regard to the regulation of DNA binding by p53. To gain further insight into the possible roles of interacting proteins on the association of p53 with DNA, we have examined the effect of both N-and C-terminal antibodies on p53 dissociation. The ability of p53 to function as a transcriptional activator may rely not only on the sequences to which it binds but also on its ability to stay bound to DNA with kinetics that favor activated transcription. Our results show that the N terminus contains an auto-inhibitory region whose regulation of core DNA binding can occur via protein-protein interactions.

EXPERIMENTAL PROCEDURES
Purification of Proteins-p53 proteins were expressed and purified from recombinant baculovirus-infected insect cells as described previously (64). Sf-9 cells were infected with recombinant baculoviruses expressing hemagglutinin (HA)-tagged wild-type p53 or p53⌬C30 (res-idues 1-363). The proteins have affixed to their second residue the HA sequence MGYPYDVPDYA. Cells were harvested and extracted 48 h post-infection, and p53 proteins were immunopurified over a column containing protein A-Sepharose matrix cross-linked to monoclonal antibody 12CA5. Bound proteins were washed and subsequently eluted with peptide containing the HA epitope YPYDVPDYA. The p53⌬N23 and p53⌬N96 proteins used in electromobility shift competition assays were constructed and expressed by similar means. Bacterially expressed human p53 central core domain (residues 96 -312) was a gift from N. Pavletich and was prepared as described previously (65). p53 proteins expressed in E. coli were prepared as follows. The p53 cDNA was cloned into pRSETB (Invitrogen) as described in Ko et al. (54) to produce His-tagged protein at the N terminus. The construct for p53⌬N96 was cloned after polymerase chain reaction amplification of the cDNA fragment by using primers containing the restriction sites KpnI and BspMI. Expression of constructs was performed in E. coli BL21 (DE3)(pLysS) (Novagen) grown at 19°C. Proteins were purified over Ni 2ϩ -nitrilotriacetic acid matrix (Qiagen). Further purification was performed via chromatography over Hi-trap heparin (Amersham Pharmacia Biotech) and Superose 6 (Amersham Pharmacia Biotech) matrices. Monoclonal antibodies DO-2, DO-13, and DO-14 were obtained from Serotec. Monoclonal antibodies C-36, DO-1, PAb 1801, and PAb 421 were purified from hybridoma supernatants with protein A-Sepharose and dialyzed against phosphate-buffered saline.
Electromobility Shift Assays-EMSAs were performed as described previously (66). Synthetic oligonucleotides representing the p21 (5Ј-AAT TCT CGA GGA ACA TGT CCC AAC ATG TTG CTC GAG-3Ј) binding site were annealed, and the double-stranded species were purified on a 12% native polyacrylamide gel. The oligomer was then end-labeled with 32 P using T4 polynucleotide kinase (Promega Corp.). Each reaction mixture contained p53, monoclonal antibodies, doublestranded probe (20,000 -50,000 cpm), 4 l of 5ϫ EMSA reaction buffer (100 mM HEPES, pH 7.9, 125 mM KCl, 0.5 mM EDTA, 50% glycerol, and 10 mM MgCl 2 ), 1 l of 40 mM spermidine, 1 l of 10 mM dithiothreitol, 1 l of 0.5% Nonidet P-40, and 2 l of 1 mg/ml bovine serum albumin. Distilled water was added to a final volume of 20 l, and mixtures were incubated at room temperature for 30 min. For competition assays, 100-fold excess of unlabeled, annealed oligomers representing each relevant site were added for the appropriate time points. Mixtures were loaded onto a 4% native polyacrylamide gel containing 0.5ϫ Tris borate/ EDTA buffer, 0.05% Nonidet P-40, and 1 mM EDTA. The gels were electrophoresed in 0.5ϫX TBE buffer at 200 V for 1 h at 4°C and then dried and quantitated by PhosphorImaging using Image-Quant software (Molecular Dynamics).
DNase I Footprinting Assays-A 405-bp HindIII-ScaI fragment of WWP-luc (kindly provided by B. Vogelstein) representing a portion of the human p21 promoter and containing the p53 response element (5Ј-GAA CAT GTC CCA ACA TGT TG-3Ј), located 2.4 kilobase pairs upstream of p21 coding sequences, was labeled with 32 P via a fill-in reaction using the large fragment Klenow of E. coli DNA polymerase I. Each reaction mixture contained probe (100,000 cpm), 10 l of 5ϫ reaction buffer (162.5 mM HEPES, pH 7.9, 31.25 mM MgCl 2 , 2.5 mM dithiothreitol, 250 mM KCl, 500 g/ml bovine serum albumin, 0.25 mM EDTA, 25% glycerol, 0.125% Nonidet P-40, 2.5 mM spermidine, 25 ng/ml poly(dG⅐dC)) and 25 l of BC100 (20 mM Tris, pH 8.0, 0.1 M KCl, 0.2 mM EDTA, 20% glycerol). Along with indicated amounts of protein (baculovirus or bacterially expressed p53 and monoclonal antibodies), distilled water was added to a final volume of 50 l. Binding mixtures were incubated for 40 min at room temperature, at which time 50 l of ice-cold DNase I digestion buffer (5 mM CaCl 2 /10 mM MgCl 2 ) was added to each mixture, and the tubes were removed to ice. DNase I (45 ng, Worthington) was added to each sample, and digestions proceeded for 2.5 min on ice. Reactions were terminated by the addition of 90 l of Stop Solution (1% SDS, 20 mM EDTA, 200 mM KCl, and 250 g/ml yeast tRNA). DNA samples were then deproteinized by phenol/chloroform extraction and collected by ethanol precipitation in the presence of 7.5 M ammonium acetate. Pellets were washed once with 70% ethanol, dried under vacuum, and resuspended in loading buffer (95% formamide, 20 mM EDTA, 0.05% bromphenol blue and 0.05% xylene cyanol FF). Samples were heated to 95°C for 5 min, immediately placed on ice, and loaded onto an 8% polyacrylamide sequencing gel. Electrophoresis was performed at room temperature for 1 h at 25 mA after which the gels were dried and autoradiographed.

PAb 1801 but Not PAb 421 Stimulates the Ability of p53 to
Protect a p53-binding Site from Cleavage by DNase I-The monoclonal antibody PAb 421, which recognizes amino acids 372-382 at the C terminus of p53, has been previously shown to stimulate dramatically the quantity of p53-DNA complexes that can be detected by the electromobility shift assay (EMSA) (33,37,(67)(68)(69). We sought to confirm this result via an alternative DNA binding assay, the DNase I footprint. A 32 P-labeled, 405-bp fragment of the p21 promoter containing one p53 response element was used as a binding substrate. Wild-type HA-tagged human p53 protein, which had been expressed and affinity purified from insect cells, was used as a source of protein. Increasing amounts of p53 protein conferred a dosedependent increase in protection of the consensus DNA-binding site from DNase I digestion (Fig. 1B). To examine the possible stimulatory effect of antibodies, we chose from preliminary binding curves a concentration of p53 yielding relatively weak protection. Surprisingly, upon titration of PAb 421 (amino acids 372-382) at ratios of antibody to p53 that were sufficient to shift completely and stimulate p53-DNA complexes in the EMSA (Fig. 1C), we observed no increase the amount of protection provided by p53 alone (Fig. 2A). The interaction of the p53 protein with antibody was also confirmed by Western blot analysis (data not shown). Furthermore, if immune complexes between p53 and PAb 421 were allowed to form prior to introduction into the footprinting reactions, no stimulation was apparent (data not shown). By contrast, when an N-terminalspecific monoclonal antibody, PAb 1801 (amino acids 46 -55) (70), was added to the binding reaction mixtures with the same amount of p53, a significant increase in the amount of protection provided by p53 alone was observed (Fig. 2B). This was striking in light of the fact that PAb 1801, as shown previously (71), conferred little or no effect on p53 binding in the EMSA (Fig. 1C). It should be noted that neither PAb 421 nor PAb 1801 caused any changes in the footprinting pattern of p53 binding to its consensus sequence (i.e. identical nucleotides in the sequence were protected with and without antibody).
N-terminal Antibodies Increase the Ability of p53 to Protect Its Site from DNase I Cleavage-To extend the observation that PAb 1801 stimulates p53 protection of DNA from DNase I cleavage, we examined a series of antibodies that bind different N-terminal epitopes for their possible effects on sequence-specific binding by p53 (Fig. 3). By using PAb 1801 as a control for stimulation of binding, we titrated antibodies DO-2 (amino acids 10 -16) (72) 72, 73) into DNase I footprinting reaction mixtures that contained a fixed amount of p53 (Fig. 4). Each of the antibodies with epitopes N-terminal to the PAb 1801 epitope (amino acids 46 -55) stimulated p53 protection from DNase I digestion (Fig.  4, A-C), whereas DO-14, recognizing an antigenic determinant downstream of the PAb 1801 epitope, had no effect on binding by p53 (Fig. 4D). The abilities of all antibodies, including DO-14, to interact with p53 were confirmed in EMSA experiments that showed that p53-DNA complexes were completely shifted at similar ratios of antibody:p53 as those used in DNase I footprints (see Fig. 6A and data not shown). Our data therefore suggest that a region within the N terminus, extending to within the vicinity of residue 55, negatively regulates sequence-specific binding by the core domain.
PAb 1801 and PAb 421 Reduce the Dissociation Rate of Baculovirus p53 from DNA in Competitive Electromobility Shift Assays-DNase I footprinting experiments showed that N-terminally directed antibodies stabilize sequence-specific DNA binding activity of p53. To investigate further this observation, we performed EMSAs under conditions that allow measurement of the dissociation rate of protein-DNA complexes. After pre-equilibration of protein with labeled DNA, 100-fold excess of cold, annealed oligomers containing the p21binding site were added to the binding mixtures. After a certain time interval, as specified in Fig. 5A, the mixtures were loaded onto a 4% native gel, and p53-labeled DNA complexes remaining were quantitated. Under these conditions, the remaining p53-labeled DNA complexes were fitted with a single exponential decay profile assuming first-order kinetics of dissociation. Re-association of p53 with labeled DNA is negligible due to a 100-fold excess of cold DNA used in competition. Fig. 5A is a representative plot of the values generated following competition of p53 alone on DNA as compared with p53 bound by PAb 1801 or PAb 421. The k off rate of p53 from DNA is estimated to be 1.9 Ϯ 0.2 min Ϫ1 , and in the presence of PAb 1801 or PAb 421 antibody the k off rate is reduced to 0.61 Ϯ 0.19 and 0.91 Ϯ 0.02 min Ϫ1 , respectively. Fig. 5B is an average of five independent competitive EMSA experiments reflecting the rates of dissociation of p53 alone from DNA as compared with p53 bound by PAb 1801 and PAb 421. From these experiments it is clear that PAb 1801 has a dramatic ability to reduce the dissociation rate of p53 from DNA. PAb 421 also causes a reduction in the dissociation rate but to a lesser extent than PAb 1801.

N-terminal Antibodies Reduce the Dissociation of Full-length p53 from Its Cognate Binding
Site-Our observation that Nterminal antibodies can stimulate wild-type p53 sequence-specific binding using DNase I footprints led us to hypothesize that perhaps other regions in the N terminus also lead to destabilization of p53 on DNA. Thus, interaction with these antibodies, although not necessarily facilitating a conversion of p53 from a latent to an active form for binding, somehow affects the kinetics of binding leading to a more stable complex of p53 on DNA. N-terminal antibodies were therefore examined for their effects on the dissociation of p53 from its cognate binding site in the p21 promoter. These  Fig. 6 (A and B), antibodies DO-1, C-36, and PAb 1801 were compared with p53 alone and the C-terminal antibody PAb 421 (amino acids 372-382) in competitive dissociation assays. Of the N-terminal antibodies tested that were able to reduce the dissociation of p53, PAb 1801 was consistently the most effective. It is important to mention here that the dissociation of different preparations of p53 protein from DNA varied somewhat. Nevertheless, relative effects of the different antibodies were internally consistent with different preparations of p53.
In separate experiments, DO-13 and DO-14 were also compared with p53 alone or with PAb 1801 in competitive dissociation assays (Fig. 6, C and D). DO-13 was able to reduce dissociation by 50% over p53 alone, comparable to the effect conferred by PAb 1801. In contrast, DO-14 was repeatedly unable to stabilize p53 on DNA, a result that corresponds to that obtained with DNase I footprints. These data demonstrate that the interaction with several N-terminally directed antibodies results in reduced dissociation of p53 from DNA and suggests this as a basis for the increased protection from DNase I cleavage in footprinting reactions. Table I shows the relative dissociation rates of free p53 and p53 bound by different antibodies.
Full-length p53 Expressed in Baculovirus and E. coli Exhibit Differences in Their Abilities to Bind DNA Sequence-specifically-Previous studies (33,38) in this and other laboratories have suggested that the ability of p53 to bind sequence-specifically might depend on the source of the expressed protein.
Post-translational modifications have been implicated in the observation that p53 from bacterial cells is able to bind its consensus sequence less well than p53 from insect cells (33). Phosphorylation of p53 by several kinases (reviewed in Ref. 67) or acetylation by P/CAF (60,75) or p300 (62,75) have been shown to stimulate further an otherwise partially latent pool of p53. These phosphorylation and acetylation events were shown to stimulate baculovirus-expressed p53, despite its innate binding ability, presumably by increasing the amount of protein competent to bind DNA. Activation of the specific DNA binding function of bacterially derived p53 by PAb 421 has been demonstrated (33) which implies that post-translational modification is not a prerequisite. However, the extent of activation of bacterial p53 by these in vitro methods exceeds that of baculovirus-expressed p53 due to an extremely low level of initial binding activity by the bacterial protein. Since the activation of bacterial p53 was assessed by EMSA, we were interested in comparing the binding activity of these two types of p53 protein using the DNase I footprint assay.
A range of 150 ng to 1.5 g of both baculovirus-expressed and bacterially expressed p53 was added into DNase I footprinting reaction mixtures (Fig. 7A). At a relatively low concentration (370 ng) of baculovirus-expressed p53, DNase I hypercutting was detected (an indication of protein binding), and complete protection of the consensus sequence was observed upon the addition of higher amounts of protein. By contrast, at all concentrations of bacterially expressed p53 tested, no hypercutting was observed nor any protection. Fig. 7B shows the p53 preparations used on a silver-stained SDS-polyacrylamide gel. Concentrations of each protein were adjusted so that equivalent amounts of each preparation were used in Fig. 7A.
Given that bacterial p53 was apparently inert in DNase I footprint assays, we next wanted to examine if PAb 1801 would also stimulate its DNA binding activity (Fig. 7C). In fact, PAb 1801 dramatically increased protection by bacterial p53 in DNase I footprints and verified its competency for binding. Our data showing that binding by bacterial p53 is markedly more sensitive to negative regulation by the N terminus than is baculovirus-derived p53 imply that, in animal cells, p53 may be modified in a manner resulting in reduced dissociation from DNA.
Deletion of the N-terminal 96 Amino Acids of p53 Stabilizes Its Sequence-specific DNA Binding Activity-Since our data with antibodies suggested that sequences in the N terminus destabilize p53 DNA binding, it was of interest to assess the impact of deletion of this region of p53 in our assays. Therefore, we performed competitive dissociation experiments utilizing a series of p53 proteins with N-and C-terminal truncations.
After 5 min of exposure to competing DNA, approximately 80% of the wild-type full-length p53 became dissociated from the probe (Fig. 8A). As expected, the p53⌬C30 mutant (residues 1-363) bound DNA approximately 2.5-fold better than wildtype protein. However, this stimulation in the quantity of p53⌬C30-DNA complexes was not due to increased stability of sequence-specific DNA binding because after 5 min with competitor, approximately 90% of the protein-probe complexes had dissociated, resulting in similar levels of bound probe as wildtype. The bacterially expressed central core protein (residues 96 -312) (65), also capable of sequence-specific binding as determined by EMSA, was extremely unstable on DNA (Fig. 8B). Even at the "zero" time point, when samples were loaded immediately onto the gel upon addition of competitor, the core protein had almost completely dissociated from the probe. This held true over a range of protein concentrations (data not shown). This may explain why we have not been able to observe core protection of DNA from DNase I digestion. 2 The p53⌬N23 mutant displayed a dissociation pattern similar to wild-type protein. However, a more extensively truncated protein, p53⌬N96, exhibited a dramatically reduced rate of dissociation. After 5 min, the maximum time allowed in the presence of competing DNA, only 45% of this protein had dissociated from the probe. The relative dissociation rates (calculated as in Fig.  5) of these versions of p53 are shown in Fig. 8C.
To examine further the effect of deletion of the N terminus on p53 binding, we compared the bacterially expressed p53⌬N96 mutant to the bacterially expressed wild-type protein (Fig. 8D). As above, 10 -600 ng of the wild-type bacterial protein did not protect its consensus sequence in the absence of PAb 1801. Strikingly, however, as little as 30 ng of the p53⌬N96 mutant conferred strong protection of the binding site, and by 150 ng there was virtually complete protection of the site. Taken together, our results show that DNA binding by both bacterialand baculovirus-expressed p53 proteins is negatively regulated by sequences within their N-terminal domains. DISCUSSION The vast majority of p53 mutations that occur in human cancers are located within the central sequence-specific DNA binding domain of the protein. This suggests strongly that DNA binding is critical to the p53 tumor suppressor function and, in turn, implies that this activity is likely to be tightly and extensively regulated. Indeed, there is a large array of literature documenting such regulation through effects on residues and regions within the C terminus. As mentioned above, phosphorylation, acetylation, deletion, antibody binding, and short basic peptides have all been shown to augment DNA-protein complexes as measured by electrophoretic mobility shift assay.
A more limited series of studies has described an alternative mode by which the N terminus of p53 also regulates core domain binding. Hansen et al. (34,76) and our group (71) have identified the N terminus as the source of thermally sensitive specific DNA binding. Interaction of N-terminal-specific antibodies, including PAb 1801, have been shown to stabilize temperature-sensitive DNA binding by wild-type (34) and tumorderived mutant (71) forms of p53. Moreover, both groups have also provided data suggesting that the effect of the C-terminal antibody PAb 421 is independent of the interactions and properties of the N terminus in that N-terminally deleted forms of p53 can still be stimulated by C-terminal antibody using the EMSA (64,76).
In this study we have shown that antibodies that recognize a set of distinct epitopes spanning residues 10 -55 in the p53 N terminus augment the ability of p53 to protect its binding site in the p21 promoter from DNase I cleavage. Interaction of p53 with N-terminal antibodies also reduces its rate of dissociation from DNA containing this binding site in the competitive EMSA and may contribute to the stimulation of binding observed in DNase I protection.
Based on our observation that DO-14 did not affect DNase I protection or dissociation, we speculate that general antibody binding, thus creating a mass effect, is not responsible for the observed increase in binding by p53, that the N terminus, extending to residue 55, contains a negative regulatory function for binding, and that the proline-rich domain is not likely to be involved in destabilizing p53 DNA binding. Further insight into the regions on p53 that affect its stability on DNA came from experiments using truncated forms of p53. Our data show that the core alone binds extremely unstably to DNA when compared with full-length protein or protein lacking the C-terminal 30 amino acids. Thus, the tetrameric core binds with greater stability than does the monomeric core. Counter-2 C. Cain, S. Miller, J. Ahn, and C. Prives, unpublished data. acting the stabilizing effect of the tetramerization domain is the region identified herein at the N terminus. This is evidenced by the fact that N-terminal antibodies exert highly stabilizing effects, and a form of p53 lacking the entire N terminus binds with markedly greater stability than does fulllength p53. Taken together these data add further support to the realization that the core domain is subject to complex regulation by regions elsewhere in p53.
It is interesting to consider why such different results are obtained with DNase I footprints and simple gel shift assays when used to study regulation of p53 DNA binding. In particular, PAb 421 dramatically increases the amount of p53-DNA complexes in mobility shift gels but has no effect on DNase I protection. By contrast, PAb 1801 stimulation of p53-DNA complex formation as determined by EMSA is minimal, whereas this antibody exerts a dramatic effect in the footprinting assay. The "caging" effect (77) apparent in electromobility shift experiments may have influenced the interpretation of results obtained in studies pertaining to negative regulation of core sequence-specific DNA binding by the p53 C terminus. The DNase I protection assay is not affected by this complication due to the fact that any complexes formed, having reached equilibrium during the binding reaction, are essentially trapped on or off DNA at a precise point in time during DNase I cleavage. We suggest that DNase I footprinting provides a sort of kinetic "snapshot" of the pool of p53 protein at equilibrium, with both k on and k off reflected. In contrast, the competitive EMSA solely reflects the dissociation of protein from DNA. Based on our data, however, the discrepancy between PAb 421

TABLE I A comparison of dissociation rates of free and antibody-bound p53
Competitive dissociation EMSAs were performed and quantitated as described under "Experimental Procedures." All values generated for k off were then normalized relative to free p53, which is taken as 1.
Another interesting result of our work is that in the presence of antibody, there is a discrepancy between curve fitting and data (see Fig. 5A). This discrepancy may be interpreted as the existence of two classes of complexes, one a fast dissociating form (Form A) and one a more slowly dissociating form (Form B). In the absence of antibody, however, our purified protein fits a curve depicting first-order exponential decay, suggesting the existence of only one form of complex. It is possible that the interaction between antibody and p53 induces a conformational change from Form A to B, where upon an equilibrium between the two forms is reached; this is evidenced by the fact that the all data points do not fit a first-order exponential decay profile. We can hypothesize that activated transcription requires a certain stability of p53-DNA complexes at the promoter site. This stability could be provided when there is interaction between p53 and N-terminal binding proteins.
The p53 N terminus is a prime candidate for potential regulation due to its various molecular interactions. Cellular proteins shown to bind to the p53 N terminus include Mdm2 (8), CBP/p300 (79,80), the DNA replication factor RP-A (81,82), and several transcription factors such as TBP and TFIID (20,83), TBP-associated factors (23,24), and THIIH (21). At least one viral protein, the adenovirus E1B 55-kDa polypeptide, has also been shown to interact with the p53 N terminus (84). Although the contribution of these interactions to the dissociation of p53 from DNA awaits experimentation, it is tempting to postulate that the interaction of cellular proteins with p53 at the N terminus might affect the kinetics of sequence-specific DNA binding. It is a particularly attractive model when considering the molecular interactions of p53 with the general transcription machinery. In fact, we previously demonstrated by EMSA and in vitro DNase I footprint analysis that TFIID and TBP stimulate p53 binding to its consensus sequence even in the absence of the TATA box (66). TFIID and TBP have been shown to bind to p53 within the N terminus between residues 20 and 57 (20), the region that we have shown can be targeted for stabilization. Cooperativity of transcriptional activation is demonstrated when p53 and TBP proteins are co-transfected and overexpressed in Drosophilia Schneider cells (85).
Stabilization of p53 on DNA might not only facilitate transactivation by p53 but in some cases repression as well. Martin and Berk (86) observed a 10-fold stimulation of p53 sequencespecific binding upon the addition of adenovirus E1B 55K to fixed amounts of p53 using in vitro DNase I footprint analysis.
Co-transfection experiments demonstrated that adenovirus E1B 55K represses p53 transcriptional activation without apparent disruption of its association with hTAF31 or TBP. Moreover, E1b possesses a transcriptional repression domain, and it was suggested that the stabilized p53-DNA complex might be more effectively repressed than one with a more ephemeral association with DNA (87). Given that Mdm2 has been postu-  lanes 2 and 5), 1 (lanes 3 and 6), and 2 l p53 (lanes 4 and 7). As a protein concentration standard, bovine serum albumin was loaded in lated to possess a transcriptional repression function (88) and may interact with the same region on p53 as E1b, we might speculate that Mdm2-mediated repression of p53 is further assisted by reducing the dissociation of p53 from DNA. The effects of Mdm2 on p53 DNA binding, however, have not been fully clarified. Zauberman et al. (89) reported that Mdm2 might inhibit the interaction of p53 with DNA, whereas Bottger et al. (90) did not observe such an inhibition. Experiments to determine whether Mdm2 affects p53-DNA complexes are currently in progress.
Numerous residues within the p53 N terminus including Ser-15, Ser-20, Ser-33, and Ser-37 have been shown to be phosphorylated in response to various forms of DNA damage (reviewed in Refs. 3 and 67). Kinases that have been shown to be able to phosphorylate these sites in vitro have been identified. The outcomes of phosphorylation of p53 at these sites are not fully understood, and mutational analysis has, for the most part, led to confusing and contradictory results. Phosphorylation of Ser-15 and Ser-37 on p53 by DNA-PK leads to a conformational change occurring at the N terminus and results in reduced association with Mdm2 in vitro (51). Although the amounts of p53-DNA complexes are not significantly increased when p53 is phosphorylated by DNA-PK (51), it will be interesting to test the effects of such phosphorylation on p53 dissociation and in DNase I protection assays.
Our observation that unmodified (bacterially expressed) p53 is extremely unstable on DNA, displaying no protection from DNase I cleavage in footprint analysis, suggests the possibility that post-translational modification of p53, such as phosphorylation, might serve to counteract this instability. However, since in the presence of monoclonal antibody PAb 1801, bacterial p53 demonstrates its competency for binding and is stabilized on its consensus DNA-binding site, such modifications are not absolutely essential for such stabilization. Indeed, when the N-terminal 96 amino acids are removed, the bacterial protein exhibits a remarkable reduction in dissociation from DNA, exceeding that of post-translationally modified, baculovirusexpressed p53 bound by antibody. Experiments by Perez-Howard et al. (91) showed that the k on for TBP and DNA is reduced 45-fold upon deletion of the TBP N terminus, thus providing another example of a protein containing a region that regulates its DNA binding kinetics. The results of our study provide further insight into negative regulation of the p53 central DNA binding domain. Implicit in our work is the possibility of interactions not yet defined between the N terminus of p53 and cellular proteins, perhaps members of the general transcription machinery, which regulate p53 DNA binding. Because the N terminus is responsible for transcriptional activation, it is interesting that through this portion of the molecule, the sequence-specific DNA binding ability of p53 can be stabilized. In vivo, this could result in reduced dissociation and increased association of p53 under FIG. 8. Deletion of N-terminal 96 amino acids of p53 stabilizes sequence-specific DNA binding activity. A, the p53⌬N96 mutant exhibits reduced dissociation from sequence-specific DNA binding. 25 ng of full-length and truncated forms of p53 were used in competitive dissociation EMSA using a 115-bp labeled fragment containing the p21-binding site as probe. 100 ng of competitor (unlabeled, annealed p21 oligomers) was added to each reaction mixture as described. B, the p53 core (residues 96 -312) exhibits extreme instability on DNA. C, a graph representing the relative k off values generated for fulllength p53 and p53 truncation mutants in A and B. All values were normalized to the k off of p53⌬N96. D, DNase I footprinting analysis using equivalent amounts of E. coli-expressed full-length p53 and E. coli-expressed p53⌬N96. Binding mixtures contained the following: no protein (lanes 1, 9 and 16 conditions requiring the activation of specific genes for specific functions.