Abstract: Article15 June 1998free access How p53 binds DNA as a tetramer Kevin G. McLure Kevin G. McLure Department of Microbiology and Infectious Diseases, and Cancer Biology Research Group, University of Calgary Health Sciences Centre, Calgary, Alberta, T2N 4N1 Canada Search for more papers by this author Patrick W.K. Lee Corresponding Author Patrick W.K. Lee Department of Microbiology and Infectious Diseases, and Cancer Biology Research Group, University of Calgary Health Sciences Centre, Calgary, Alberta, T2N 4N1 Canada Search for more papers by this author Kevin G. McLure Kevin G. McLure Department of Microbiology and Infectious Diseases, and Cancer Biology Research Group, University of Calgary Health Sciences Centre, Calgary, Alberta, T2N 4N1 Canada Search for more papers by this author Patrick W.K. Lee Corresponding Author Patrick W.K. Lee Department of Microbiology and Infectious Diseases, and Cancer Biology Research Group, University of Calgary Health Sciences Centre, Calgary, Alberta, T2N 4N1 Canada Search for more papers by this author Author Information Kevin G. McLure1 and Patrick W.K. Lee 1 1Department of Microbiology and Infectious Diseases, and Cancer Biology Research Group, University of Calgary Health Sciences Centre, Calgary, Alberta, T2N 4N1 Canada *Corresponding author. E-mail: [email protected] The EMBO Journal (1998)17:3342-3350https://doi.org/10.1093/emboj/17.12.3342 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info The p53 tumor suppressor protein is a tetramer that binds sequence-specifically to a DNA consensus sequence consisting of two consecutive half-sites, with each half-site being formed by two head-to-head quarter-sites (→← →←). Each p53 subunit binds to one quarter-site, resulting in all four DNA quarter-sites being occupied by one p53 tetramer. The tetramerization domain forms a symmetric dimer of dimers, and two contrasting models have the two DNA-binding domains of each dimer bound to either consecutive or alternating quarter-sites. We show here that the two monomers within a dimer bind to a half-site (two consecutive quarter-sites), but not to separated (alternating) quarter-sites. Tetramers bind similarly, with the two dimers within each tetramer binding to pairs of half-sites. Although one dimer within the tetramer is sufficient for binding to one half-site in DNA, concurrent interaction of the second dimer with a second half-site in DNA drastically enhances binding affinity (at least 50-fold). This cooperative dimer–dimer interaction occurs independently of tetramerization and is a primary mechanism responsible for the stabilization of p53 DNA binding. Based on these findings, we present a model of p53 binding to the consensus sequence, with the tetramer binding DNA as a pair of clamps. Introduction The p53 tumor suppressor gene is the most frequently mutated cancer-associated gene yet identified, with p53 mutations occurring in over half of all human tumors (Hollstein et al., 1996). Most mutations occur in the DNA-binding domain, where point mutations may either disrupt protein–DNA interactions directly or alter the overall conformation of the DNA-binding domain (Cho et al., 1994; Friend, 1994). Mutations in the p53 DNA-binding domain probably contribute to tumorigenesis due to the requirement for DNA binding of p53 tumor suppressor activity (reviewed by Ko and Prives, 1996; Soussi and May, 1996; Levine, 1997). Wild-type, but not tumor-derived mutant, p53 binds to a double-stranded DNA consensus binding site containing two or more copies (consecutive or separated by one or two helical turns) of the 10 bp half-site 5′-PuPuPuC(A/T) (T/A)GPyPyPy-3′, where Pu and Py represent purines and pyrimidines, respectively (Kern et al., 1991; El-Deiry et al., 1992; Funk et al., 1992; Cho et al., 1994; Wang et al., 1995; Waterman et al., 1995). Thus, the consensus site comprises four inverted 5 bp quarter-sites, where the first quarter-site is underlined. X-ray crystallography has revealed how one core domain monomer (the central half of p53, residues 102–292) binds to one such quarter-site (Cho et al., 1994), and a model was constructed in which four core domains can occupy the four quarter-sites in a full consensus sequence without steric clashes. This physical model of four p53 subunits bound to the consensus site is consistent with solution studies, where four p53 core domains can bind cooperatively to a consensus DNA sequence (Balagurumoorthy et al., 1995; Wang et al., 1995). Tetramerization of p53 is a function of the C-terminal domain. This was deduced from the demonstration that C-terminal fragments of p53, spanning residues 311–367, form tetramers in solution (Pavletich et al., 1993; Wang et al., 1993). These tetramers are symmetric dimers of dimers in which all four subunits are geometrically equivalent (Clore et al., 1994, 1995a,b; Lee et al., 1994; Jeffrey et al., 1995). However, a lower order symmetry must be exhibited by DNA-bound p53 if all four subunits in the tetramer contact a consensus DNA-binding site (Waterman et al., 1995). An important missing element in the understanding of p53 DNA binding is the global orientation of the oligomerization domains relative to the DNA-binding domains. This is because the structures of the DNA-binding and tetramerization domains, which consist of amino acids 102–292 and 324–355, respectively, and are joined by a flexible linker, have been determined separately (reviewed by Arrowsmith and Morin, 1996). Since the linker is relatively long, the relative orientations of the tetramerization and DNA-binding domains cannot be deduced from available biophysical data. As a result, two different models have arisen that conceptually connect the DNA-binding domains to the oligomerization domains in an intact p53 tetramer. In one model, one dimer of a tetramer contacts the first and second quarter-sites and the other dimer binds to the third and fourth quarter-sites in the consensus sequence (Halazonetis and Kandil, 1993; Clore et al., 1994; Waterman et al., 1995). In the other model, one dimer contacts the first and third quarter-sites and the other dimer binds to the second and fourth quarter-sites (Cho et al., 1994; Lee et al., 1994; Jeffrey et al., 1995; Arrowsmith and Morin, 1996; Pennisi, 1996). In order to differentiate between the two models of how tetrameric p53 binds to DNA, we analyzed the ability of in vitro synthesized tetrameric or dimeric p53 to bind to consensus DNA sequences that had various quarter-sites mutated. Our results show that dimers bind to intact half-sites containing two consecutive quarter-sites but not to alternating quarter-sites. Furthermore, dimer–dimer interactions distinct from those involved in tetramerization drastically enhance the DNA-binding affinity of tetrameric p53. Results Wild-type 53 binds to a half-site (two consecutive quarter-sites), but not to alternating quarter-sites To determine the nature of p53 DNA binding, a series of mutants of the p53 consensus sequence were synthesized (Figure 1). The consensus sequence (CON) contained two consecutive half-sites (four quarter-sites). The non-binding (NB) mutant had all four quarter-sites mutated, with substitutions introduced at the invariant C (or base-paired G) in each quarter-site. The M34 mutant had one consensus half-site followed by mutated third and fourth quarter-sites. The M24 mutant had mutations in the second and fourth quarter-sites. The H1 mutant contained a half-site (two quarter-sites). The m2 mutant was similar to H1 except that the second quarter-site was mutated. Figure 1.Various DNA sequences. The p53 CON is shown, along with positions that are altered in variations of the consensus sequence. Quarter-sites are depicted by arrows to visualize the arrangement of quarter-sites in the various sequences. Mutated quarter-sites are depicted as dashed lines. Download figure Download PowerPoint Wild-type human p53 was translated in vitro, then assayed for binding to the various sequences. The results show that there were no novel bands with the control NB sequence (Figure 2A). There was a non-specific complex, present to varying extents with all [32P]DNA sequences, which migrated approximately one-third to half of the way down the gel. This was endogenously present in rabbit reticulocyte lysate (RRL), varied with the batch of RRL and was not supershifted by p53-specific antibodies (data not shown). Two novel bands appeared with CON and M34 (Figure 2A, open arrow and a closed arrow). In contrast, the sequence in which the second and fourth quarter-sites were mutated (M24) was not bound by p53. The bands specific for CON and M34 required the presence of polyclonal antibody pAb421 to inactivate the negative regulatory domain located downstream of the oligomerization domain (i.e. close to the very end of the C-terminus), suggesting that the bound p53 was full-length (Figure 5, compare lanes 1 and 2). Figure 2.Binding of in vitro translated p53 to various sequences. (A) Direct DNA binding analysis by electrophoretic mobility shift assay. p53 or irrelevant RNA (luciferase) was translated in RRL, and aliquots were then mixed with equal counts of [32P]DNA (NB, CON, M34 or M24) as described in Materials and methods. The migration rates of two forms of p53 are indicated by arrows. (B) DNA binding competition studies. Binding of p53 to [32P]CON was carried out in increasing amounts of unlabeled competitor DNA as indicated. Ten ng of competitor was added to lanes 2, 8, 14, 20, 26 and 29; 33 ng of competitor was added to lanes 3, 9, 15 and 21; 100 ng of competitor was added to lanes 4, 10, 16, 22, 27 and 30; 333ng of competitor was added to lanes 5, 11, 17 and 23; and 1000 ng of competitor was added to lanes 6, 12, 18, 24, 28 and 31. Download figure Download PowerPoint Interestingly, the migration rates of both p53–CON and p53–M34 complexes were the same, implying that the same molecular mass and therefore oligomeric form of p53 was present in each complex. As will be shown below (Figure 4), the slower migrating band (Figure 4C, closed arrow) corresponded to tetrameric p53, and the faster migrating band (Figure 4C, open arrow) corresponded to dimeric p53. Our results (Figure 2A) suggest that the same species of full-length p53 could bind either to the one-half-site in M34 or to one or both of the two half-sites in CON, but not to alternating quarter-sites in M24. Identical results were seen with murine and human p53 (data not shown). A competition experiment was then carried out in which in vitro translated p53 was added to a 32P-labeled CON sequence in the presence of excess unlabeled DNA (Figure 2B). As expected, neither NB nor M24 competed with [32P]CON for binding to a constant amount of p53 (Figure 2B, lanes 29–31 and 14–18, respectively). CON competed the best for p53 binding, followed by M34 (Figure 2B, lanes 2–6 and 8–12, respectively). Therefore, wild-type p53 preferentially bound to CON, but also had some affinity for M34 (but not M24). Identical results were seen with murine p53 (data not shown). This confirms that the same species of p53 that could bind to CON alternatively could bind to M34. Due to the nature of the selective mutations of two out of 10 bp in the second M34 half-site (Figure 1), it was possible that the intact p53 oligomer was making other stabilizing contacts in the two mutated quarter-sites of M34. To test this possibility, p53 binding to [32P]CON was challenged with a single half-site (H1) not followed by a mutated half-site. There was virtually no difference between excess M34 and H1 (Figure 2B, compare lanes 8–12 with lanes 20–24). Therefore, p53 tetramers preferentially bind CON, and can bind single half-sites (when present in excess) but not single quarter-sites (Figure 2B, m2, lanes 26–28). Cellular p53 binds DNA similarly to p53 translated in vitro It was important to compare the DNA binding of in vitro translated p53 with that of cellular p53. To this end, subconfluent Balb/c 3T3 cells, which contain wild-type p53, were first subjected to DNA damage (actinomycin D treatment) in order to activate and stabilize p53. Extracts were then prepared from untreated (control) or DNA-damaged cells, and assayed for DNA-binding activity. In the absence of DNA damage, a very low amount of endogenous cellular p53 was activated for DNA binding by pAb421 (Figure 3, lane 14), but DNA damage-induced stabilization of p53 yielded a much greater amount of pAb421-activated DNA-binding complex (Figure 3, lane 6). The most slowly migrating form (Figure 3, lane 6, closed arrow) was the major DNA-bound species, and probably consisted of p53 tetramer–DNA complexes. The lower band (Figure 3, lane 6, open arrow) probably represented a lower concentration of p53 dimer–DNA complexes, as was observed with in vitro translated p53. These two complexes were also detected when [32P]M34 was used as the probe, although they were present at a significantly reduced level (Figure 3, lane 7). Interestingly, there was a low level of DNA-binding activity, unique to CON, which was activated by DNA damage and detected without pAb421 (Figure 3, lane 2). This probably represented p53 DNA binding that was not inactivated by the C-terminal negative regulatory domain. Figure 3.Binding of cellular p53 to various sequences. Balb/c 3T3 total cell extract was prepared from cells with or without prior DNA damage by actinomycin D. Cell extract was added to [32P]DNA (NB, CON, M34 or M24), either with or without pAb421, as indicated. Download figure Download PowerPoint Figure 4.Binding of A344 dimers to DNA. (A) In vitro translated A344 was mixed with [32P]DNA (with pAb421 except for lane 2) as indicated. The migration rates of two forms of A344 are indicated by arrows. (B) Binding of A344 to CON and M34 at various A344 concentrations. Decreasing amounts of in vitro translated A344 were added to [32P]CON (the reactions in lanes 3 and 4 contained two-thirds and one-third the amount of A344, respectively, compared with the reaction in lane 2) or [32P]M34 (lanes 6–8). RRL alone served as a control (lanes 5 and 9). Lane 1 is an increased autoradiography film exposure time of lane 4. The migration of a single A344 dimer is indicated by an open arrow, pairs of A344 dimers by a closed arrow. (C) Comparison of migration patterns of wild-type p53–DNA and A344–DNA complexes. In vitro translated wild-type p53 or A344 was mixed with [32P]DNA as indicated and analyzed by EMSA. Download figure Download PowerPoint Figure 5.Stability of p53 bound to CON and M34. In vitro translated wild-type p53 was allowed to bind to [32P]CON or [32P]M34 for 45 min at 22°C. A 100-fold excess of unlabeled CON (100 ng per reaction) was then added (time 0) to ‘trap’ any p53 that dissociated from [32P]CON or [32P]M34 at various times thereafter. Aliquots were taken at the times indicated and loaded on a running gel. For comparison, some reactions (lanes 2, 8 and 9) did not have any unlabeled CON added (cold ‘trap’, − lanes). All reactions contained pAb421 except for lane 1. Download figure Download PowerPoint The overall pattern of cellular p53 binding was similar to that of in vitro translated p53, as CON was bound the best, with some M34 being bound, but no M24 or NB (Figure 3, lanes 6, 7, 8 and 5, respectively). Binding competition experiments were also carried out using unlabeled competitor DNA, and the results (data not shown) concurred with the direct binding studies. It is noteworthy that p53 translated in vitro (Figure 2A, lane 7) bound M34 much better than did p53 from the cell extract (Figure 3, lane 7). The reason for this is presently unknown. Another similarity to in vitro translated p53 was the non-specific band that migrated approximately one-third of the way down the gel (Figure 3, lanes 1–16). Overall, the pattern of DNA binding exhibited by cellular p53 was similar to that exhibited by in vitro translated p53. Free dimers bind to half-sites or pairs of half-sites, but not to alternate quarter-sites Binding of p53 tetramers to M34 (or H1) but not to M24 indicated that the binding was stabilized when two subunits of the tetramer interacted with the two quarter-sites within a half-site. However, these results do not differentiate between whether the dimers in a tetramer each bind to a half-site, or whether each half-site is bound by one subunit from each dimer that makes up a tetramer. To differentiate between these two possibilities, the ability of free dimers to bind to the variations of CON was compared. The p53 A344 mutant is a full-length wild-type human p53 clone containing a point mutation at residue 344 (from leucine to alanine) that disrupts the dimer–dimer interface and results in the formation of dimeric, rather than tetrameric p53 (Waterman et al., 1995). Like the tetrameric form, dimeric A344 bound neither to NB nor to M24 (Figure 4A, lanes 1 and 5); however, it formed a single complex with M34 (Figure 4A, lane 4, open arrow). Interaction of A344 with CON resulted in the formation of two complexes, one migrating identically to the A344–M34 complex, and one migrating more slowly (Figure 4A, lane 3, open and closed arrows, respectively). As was the case with tetrameric p53, pAb421 was required to activate (or stabilize) the binding of the A344 dimer to DNA (Figure 4A, compare lanes 2 and 3). Based on the migration rates of the A344–M34 and A344–CON complexes, and in view of the dimeric nature of A344, it seemed logical to suggest that the lower band (Figure 4A, open arrow) that was common to both CON and M34 represented one dimer, and the upper band (Figure 4A, closed arrow) found only on CON represented two dimers (a tetramer). This interpretation was assessed further by varying the protein–DNA ratio in the reaction mixture (Figure 4B). Decreasing the concentration of A344 while keeping the concentration of [32P]CON constant resulted in a clear bias against the formation of the upper band compared with the lower band. This was precisely what one would expect if the upper band and the lower band corresponded to double and single A344 dimers, respectively. Decreasing the concentration of dimer relative to DNA would give an unbound dimer a greater chance of binding one of two half-sites on an unoccupied CON molecule compared with one half-site on CON that already had one A344 dimer bound. Accordingly, the faster migrating band of A344–CON (Figure 4B, open arrow) most likely represented one A344 dimer bound to one half-site in CON, whereas the more slowly migrating A344–CON band (Figure 4B, closed arrow) represented two dimers of A344 bound to the two half-sites in CON. Does wild-type p53 bind DNA similar to A344 dimers? The less abundant, faster migrating species of p53–CON and p53–M34 migrated at a position identical to that of A344–M34, probably representing one dimer bound to DNA (Figure 4C, lanes 2 and 3 compared with 7, open arrow). Likewise, the slower migrating species (Figure 4C, closed arrow) which migrated at the same position as the upper band of A344–CON probably contained one p53 tetramer. The observation that the predominant species bound to one half-site (M34) was tetrameric p53 suggested that most wild-type p53 had already formed tetramers before binding to DNA. Tetramers bind CON (two half-sites), but not a single half-site, with high affinity We noted that p53 tetramers (but not free dimers) invariably bound better to CON than to M34 (Figure 2A, compare lanes 6 and 7). To assess whether this reflected a difference in affinity, p53–[32P]DNA complexes were first allowed to form, and excess unlabeled CON was then added to ‘trap’ any p53 that dissociated from the 32P-labeled sequence over time. The results showed that the binding of p53 to CON was very stable, with a calculated half-life (t1/2) of 25 min (Figure 5, lanes 2–8; Table I). In contrast, most p53–M34 had a t1/2 much shorter than 5 min, estimated at ∼30 s (Figure 5, lanes 9–14). The dissociation of the majority of p53 from DNA probably follows first order kinetics and, therefore, the overall dissociation constant of p53–CON is 4.6×10−4/s (Table I). It is interesting to note that although most p53 on M34 dissociated within 5 min (Figure 5, compare lanes 9 and 10), the p53 that remained bound after 5 min persisted with an estimated t1/2 of ∼5 min (Figure 5, lanes 10–14). This could represent a minor population of p53 that was bound to M34 more stably than the majority of p53. Table 1. Half-lives (t1/2) and dissociation constants (KD) of p53–DNA complexes Protein–DNA complex t1/2 (min) KD (s−1) p53–CON 25 4.6×10−4 p53–M34 0.5 1.4 (A344)2–CON 15 7.7×10−4 A3441–CON ∼1 s ∼0.012 The half-lives of various p53–DNA complexes were calculated based on time courses similar to those depicted in Figures 5 and 6. The assay was not sensitive enough to determine accurately the very short half-life of a single dimer bound to CON (A3441–CON), so an estimate was made (based on a very faint A3441–CON band evident at the 0.5 min time point when gels similar to the one depicted in Figure 6B were overexposed). High-affinity DNA binding results primarily from dimer–dimer interaction While it seemed probable that the greater stability of the p53–CON complex compared with the p53–M34 complex was due to interaction of both dimers of a tetramer with the two half-sites in CON, it was important to determine whether dimer–dimer interaction played any role in stabilizing this binding. To this end, the above finding (Figure 4) that two A344 dimers could bind side-by-side to CON was exploited to determine whether dimers stabilized one another by interacting after binding to DNA. Two complexes of A344–CON were again observed: a faster migrating A344 dimer bound to one of the half-sites in CON and a slower migrating pair of A344 dimers bound to the two half-sites in CON (Figure 6A, lanes 1 and 7, open and closed arrows, respectively). When excess unlabeled CON was added to pre-formed A344–[32P]CON complexes, the two A344 dimers bound side-by-side on CON had a t1/2 of ∼15 min (Figure 6A, lanes 1–7, closed arrow; Table I). This t1/2 corresponds to a dissociation constant of 7.7×10−4/s, assuming first order kinetics of dissociation (Table I). Strikingly, in contrast to the pair of A344 dimers, the single A344 dimer on DNA had a t1/2 much less than 5 min (Figure 6A, compare lanes 1 and 2, open arrow). As expected, the single A344 dimer on M34 also had a short t1/2 of <5 min (lanes 8–13). Shorter time courses revealed that the single A344 dimer on CON had a t1/2 much less than 30 s (Figure 6B). Consequently, the t1/2 of two dimers bound to CON (15 min) was much greater than double the t1/2 of a single dimer bound to one half-site (estimated to be ∼1 s; Table I footnotes). Figure 6.Stability of A344 bound to CON and M34. (A) In vitro translated A344 was allowed to bind to [32P]CON or [32P]M34 for 45 min. A 100-fold excess of unlabeled CON (100 ng per reaction) was then added (time 0) to ‘trap’ any A344 that dissociated from [32P]CON or [32P]M34 at various times thereafter. Aliquots were taken at the times indicated and loaded on a running gel. For comparison, some reactions (lanes 1, 7 and 8) did not have any unlabeled CON added (cold ‘trap’, −). The migration of one A344 dimer is marked by an open arrow; pairs of A344 dimers are marked by a closed arrow. All reactions contained pAb421. (B) Same as (A), except that aliquots were taken at shorter times after addition of unlabeled ‘trap’. Download figure Download PowerPoint The above experiment demonstrated that one A344 dimer cooperatively stabilized the binding of the second dimer, which would also presumably occur in dimers of a wild-type p53 tetramer bound to CON. Based on the t1/2, p53–CON appeared to be more stable than (A344)2–CON, suggesting that the C-terminal tetramerization domain may contribute further to tetramer DNA binding. Discussion p53 binds DNA with each dimer of the tetramer contacting its own half-site Although the individual structures of the major functional domains of p53 have been determined (reviewed by Pennisi, 1996), a major unresolved issue has been how the four DNA-binding domains are related to the oligomerization domains in a DNA-bound p53 tetramer (reviewed by Arrowsmith and Morin, 1996). Based primarily on NMR and crystallographic analysis of the C-terminal tetramerization domain of p53, it is now generally accepted that p53 is a dimer of dimers (Clore et al., 1994, 1995a,b; Lee et al., 1994; Jeffrey et al., 1995). Since each monomeric subunit of a dimer binds to a quarter-site, a dimer theoretically could bind to two contiguous quarter-sites that comprise a half-site, or to alternating quarter-sites within a full consensus sequence. Additionally, considering the head-to-head nature of the two quarter-sites within a half-site, binding of the two monomeric subunits within a dimer to contiguous quarter-sites would suggest that the dimer is rotationally symmetric (i.e. head-to-head). Conversely, binding of these monomers to alternating quarter-sites would suggest that the dimer is translationally symmetric (i.e. head-to-tail). These considerations are important since they have major implications for how the two dimers within a tetramer cooperatively interact upon DNA binding. We demonstrate here for the first time how the DNA-binding domains are connected to the oligomerization domains in DNA-bound p53 tetramers. Tetramers could bind to a DNA sequence that contained quarter-sites one and two (M34), but not to a sequence that contained only quarter-sites one and three (M24). This suggests that stable DNA binding occurs only when the two adjacent quarter-sites in one half-site are occupied. However, this finding did not discriminate between which subunits of a tetramer were interacting with a half-site. The subsequent use of dimeric p53 (A344) resolved this issue. Like tetrameric p53, dimeric p53 bound only M34 but not M24, indicating that the two subunits of one dimer could bind to two adjacent quarter-sites within a half-site. It is easy to envisage how this binding would be more stable than one involving a single subunit and a single quarter-site. Furthermore, there may well be cooperative interaction between the two subunits that would enhance the overall stability of the complex further. In the case of the tetramer, binding to M34 is probably achieved via the interaction of a dimer in the tetramer with the one half-site in M34, leaving the other dimer either ‘hanging off’ the DNA or perhaps interacting non-specifically with DNA. Binding of a tetramer to CON would involve the binding of the two dimers to the two respective half-sites in CON. Based on our present findings and other previous suggestions for various aspects of p53 DNA binding (Halazonetis and Kandil, 1993; Cho et al., 1994; Clore et al., 1994; Lee et al., 1994; Waterman et al., 1995), a model is depicted schematically in Figure 7A. In this model, p53 can be visualized as a pair of clamps, with each dimer being one clamp. The two arms of each clamp represent the two core domains that interact with the two adjacent quarter-sites within a half-site. As initially suggested by Cho et al. (1994), the bulk of the four core domains would lie on one face of the DNA helix. Flexible linkers extend to the other face of the helix and may have no fixed structure, thereby allowing the tetrahedrally symmetric tetramerization domain to retain its symmetry while the attached core domains are bound to DNA with a lower order of 2-fold cyclic symmetry (Waterman et al., 1995). The oligomerization domain is a tetrahedrally symmetric dimer of dimers, consisting of two pairs of interlocking subunits (Clore et al., 1994, 1995a,b; Lee et al., 1994; Jeffrey et al., 1995). Each dimer is formed by two monomers that dimerize via an antiparallel β-sheet followed by a turn and an α-helix, the latter of which forms the dimer–dimer interface (Clore et al., 1994, 1995a,b; Lee et al., 1994; Jeffrey et al., 1995). The tetramerization domain probably resides on the opposite side of the helix relative to the core domains (Lee et al., 1994; Jeffrey et al., 1995). Figure 7.Model of a p53 tetramer binding to DNA. (A) Overall structure of the p53–CON complex. One dimer (green and yellow) binds to one half-site on the DNA. The second dimer (blue and red) of the tetramer binds to the adjacent half-site in a consensus DNA sequence. After binding, the dimers interact independently of the tetramerization domain to stabilize binding. (B) In search of CON by p53 in an environment of excess, random DNA. Initially, the p53 tetramer interacts transiently with non-specific DNA or with a single quarter-site (left; f1/4 site = 64−1). Detectable, but unstable, binding occurs when a less frequently occurring half-site is encountered (middle; f1/2 site = 4096−1). The tetramer becomes stably bound to DNA only when it encounters an infrequent full consensus sequence (right; ffull−site = ∼1.7×10−7). Download figure Download PowerPoint A model that has the DNA-binding domains of one dimer in a tetramer binding to one half-site of a consensus DNA sequence has been postulated previously, but not demonstrated. One group demonstrated that free dimers could bind to half-sites, but it was not ascertained whether dimers in a tetramer could bind to quarter-sites one and two or one and three in the consensus DNA (Halazonetis and Kandil, 1993; Waterman et al., 1995). Another group showed that tetramers (lacking the C-terminal nega