Crystal Structure of a Multidomain Human p53 Tetramer Bound to the Natural CDKN1A (p21) p53-Response Element

The p53 tumor suppressor gene regulates the response of cells to DNA damage (1, 2). It is the most frequently mutated gene in human cancer (3, 4), possibly to allow cancer cells to survive and proliferate in the presence of oncogene-induced DNA replication stress (5). The protein encoded by p53 is a transcription factor containing 2 independently folding domains: a sequence-specific DNA-binding domain at the center of the protein and a homotetramerization domain toward its C-terminus (6–8).

Three-dimensional structures of p53 have been determined both in its free and DNA-bound forms (9–15). From the protein side, these structures encompass either only the sequence-specific DNA-binding domain (9–14) or a larger polypeptide that includes both the DNA-binding and homo-oligomerization domains (15). The first group of structures shows no conformational change in the p53 DNA-binding domain upon interaction with DNA, which is atypical of sequence-specific DNA-binding proteins (16–18). In contrast, the structures of the multidomain p53 tetramer bound to DNA show a conformational switch involving loops L1 of the DNA-binding domains of 2 p53 subunits (15). We have argued that this conformational switch allows p53–DNA binding on-rates and off-rates to be regulated independently of dissociation rates. Thus, complexes of p53 with specific DNA have much longer half-lives than complexes with nonspecific DNA, even though the affinities of p53 for specific and nonspecific DNA differ by less than 10-fold (15, 19). Such a model to explain DNA-binding specificity could be generally applicable, as conformational switching is a feature of all characterized sequence-specific protein–DNA interactions (20).

Despite the great progress made towards structural characterization of p53–DNA complexes, none of the structures determined so far contain a natural p53-response element. To address this gap, we crystallized a multidomain p53 polypeptide bound to the well-characterized p53-response element present upstream of the promoter of the CDKN1A (also known as p21, WAF1, and CIP1) gene (21–26).

The p53 polypeptide p53CR2 (15) was expressed in E. coli. The cells were lysed in buffer consisting of 25 mmol/L Bis-Tris propane (pH 6.8), 250 mmol/L NaCl, 5 mmol/L dithiothreitol, and protease inhibitors, and the polypeptides were purified by cation exchange (Sepharose SP column; Pharmacia Biotech) and gel filtration (Superdex 200 column; Pharmacia Biotech) chromatography. After purification, the p53 protein was incubated with DNA at a 1:1.1 protein:DNA molar ratio at 4°C for 1 hour in 25 mmol/L Bis-Tris propane (pH 6.8), 150 mmol/L NaCl, and 5 mmol/L dithiothreitol buffer. The complex was then purified by gel filtration chromatography (Superdex 200 column; Pharmacia Biotech), concentrated to about 8 mg/mL protein concentration and crystallized at 4°C under standard hanging drop vapor diffusion conditions in 48-well plates (Hampton Research). The precipitant solution used for crystallization consisted of 0.2 mol/L potassium sodium tartrate tetrahydrate (pH 7.0) and 20% polyethylene glycol 3350. The sequence of the DNA used for crystallization is shown in Fig. 1A.

X-ray diffraction data sets were collected at the ID23-2 microfocus beamline of the European Synchrotron Radiation Facility. Reflection data were indexed, integrated, and scaled with the CCP4 software package (27). Initial solutions were determined by molecular replacement and refined under noncrystallographic symmetry restraints for the protein chains with the programs CNS and O (28, 29). Figures were prepared with the programs MOLSCRIPT, BOBSCRIPT, and RASTER3D (30–32).

The p53 protein used in this study has been previously characterized and described as protein p53CR2 (15). It spans residues 94 to 358 of human p53 and encompasses the sequence-specific DNA-binding (residues: 94–291) and homo-oligomerization (residues: 324–355) domains, but contains a deletion of residues 293 to 321, which correspond to the flexible linker between these domains. The sequence-specific DNA-binding domain in p53CR2 has been engineered to prevent aggregation and unfolding in vitro. This was accomplished by introducing 13 amino acid substitutions targeting nonconserved residues that are far away from the DNA-binding surface. The goal of these substitutions was to enhance the contribution of the hydrophobic effect to protein folding. Analysis of the stabilized p53 DNA-binding protein both in vitro and in cell-based assays showed that the engineered amino acid substitutions had no effect on DNA-binding specificity and tumor suppressor function (15). In addition to the amino acid substitutions in the DNA-binding domain, the p53CR2 protein contains in its oligomerization domain 2 amino acid substitutions that only allow homodimer formation. However, when incubated with sequence-specific DNA sites, 2 p53CR2 dimers form a stable complex with DNA, thereby recapitulating the binding of p53 homotetramers to DNA (15).

The oligonucleotides used for crystallization anneal to form DNA that is 24 bp long with overhangs that are 2 nucleotides long at each 3′ end (Fig. 1A). The overhangs were designed to allow formation of pseudocontinuous DNA molecules through the crystal, as was, indeed, the case. The 24-bp sequence corresponds to the natural p53-response element present in the human CDKN1A (p21) gene (20 bp) plus the natural flanking sequences (2 bp on each side). The p53-response element consists of 4 contiguous pentamer repeats, which we have labeled a, b, c, and d, reflecting their 5′ to 3′ order in the DNA sequence (Fig. 1A). Repeats b and c, which are at the center of the p53–DNA-binding site, are referred to as the inner repeats, whereas repeats a and d are the outer repeats. Of note, the CDKN1A–p53-response element has only 60% identity to the artificial consensus site (oligonucleotide CONS26) that we previously crystallized bound to p53CR2 (Fig. 1A; ref. 15).

The 3-dimensional structure of 2 p53CR2 dimers bound to the CDKN1A–p53-response element was determined at a resolution of 2.8 Å (Fig. 1B and Table 1). The overall structure is very similar to the previously determined structure of p53CR2 bound to oligonucleotide CONS26 (Fig. 1C). Thus, the CDKN1A–p53-response element exhibits the same helical axis shift at its center, as previously described for CONS26, and all the base pairing is of the Watson–Crick type. The intersubunit contacts between the 4 p53–DNA-binding domains are also identical to those previously described for the p53CR2–CONS26 DNA complex (15). Nevertheless, despite the high level of overall similarity between the p53CR2–CDKN1A and p53CR2–CONS26 structures, there are also interesting differences, as will be described later. We will be referring to the p53 subunits as A, B, C, and D, matching the label of the pentamer repeat, whose major groove they contact (Fig. 1A and B).

As in the previously described p53CR2–CONS26 DNA structure, there are similarities and differences in the way the p53 subunits recognize the inner and outer repeats. The sequence-specific contacts involve predominantly bp 2 and 3 of each pentamer repeat and are similar for all p53 subunits. Using subunit B as an example, one observes the previously described hydrogen bonds between Arg280 and the O6 and N7 atoms of the guanine base at position 2 of repeat b (Fig. 2A and B). The sugar backbone of the same nucleotide and the sugar backbone of the thymidine at position 3 are in contact with the side chain of Arg248 of subunit A (Fig. 2B). From the side of the major groove, the thymine at position 3 of repeat b forms a hydrogen bond with the side chain of Cys277 of subunit B (Fig. 2A and C). In addition, van der Waals and hydrophobic interactions stabilize the C5A atom of the thymine at an equidistant position between the side chains of Ala276 and Cys277. In the CDKN1A–p53-response element, a thymine is present at position 3 in all 4 pentamer repeats (Fig. 1A), perhaps helping in explaining the high affinity of the CDKN1A element for p53.

Unlike the contacts of p53 with bp 2 and 3, which are essentially identical in all 4 subunits, the interactions with bp 4 and 5 vary depending on whether the p53 subunit contacts an inner or an outer repeat. This is because the conformation of loop L1 is different in the inner and outer p53 subunits. In the inner subunits, loop L1 adopts an extended conformation, the same conformation that is also observed in the absence of DNA. This extended conformation is accommodated thanks to the shift in the DNA helical axis, which moves the DNA backbone away from loop L1, relative to B-form DNA. In contrast, in the outer subunits, loop L1 adopts a recessed conformation to avoid steric clashes with the DNA, which here adopts the standard B-form DNA (15).

Using p53 subunit B as an example of the contacts of subunits B and C to the inner repeats b and c, respectively, the side chain of Lys120 makes a hydrogen bond with atom N7 of the adenine base at position 4′ of repeat b (Fig. 2A and D). Furthermore, the side chain of Ser121 makes hydrogen bonds with the sugar–phosphate backbone of the cytidine at position 5′ of repeat b (Fig. 2A and E).

In p53 subunit A, representing the contacts of subunits A and D to the outer repeats a and d, respectively, Lys120 does not contact DNA, whereas, the side chain of Ser121 contacts the phosphate backbone of the nucleotide 3 positions 5′ to the p53-response element (Fig. 3A). In addition, Arg283 contacts the phosphate backbone of the cytidine at position 5′ of repeat a (Fig. 3A). The corresponding cytidine in repeat b is contacted by the side chain of Ser121 of subunit B (compare with Fig. 2A).

Table 2 summarizes all the p53–DNA contacts observed in the p53CR2–CDKN1A structure.

The structure of p53CR2 bound to the CDKN1A–p53-response element shows 2 distinct loop L1 conformations; an extended conformation for the inner subunits (Fig. 2A) and a recessed conformation for the outer subunits (Fig. 3A). The p53CR2–CONS26 structure also displayed extended and recessed conformations (15). The extended loop L1 conformations of the p53CR2–CDKN1A and p53CR2–CONS26 structures are very similar to each other and to the conformation of loop L1 in free p53 (15 and data not shown). However, the recessed conformations in the p53CR2–CDKN1A and p53–CONS26 structures differ. In p53CR2–CDKN1A, the side chain of Val122 faces towards the protein interior, as expected for a hydrophobic residue, whereas the side chain of Ser121 faces away from the protein interior and contacts the DNA backbone (Fig. 3A and B). In contrast, in p53CR2–CONS26, the side chain of Val122 faces towards solvent, whereas the side chain of Ser121 faces somewhat towards the protein interior and fails to contact DNA (Fig. 3C and ref. 15).

Superimposition of the 3 loop L1 conformations (extended, recessed CONS26, and recessed CDKN1A) reveals that the recessed CONS26 conformation is intermediate between the extended and recessed CDKN1A conformations (Fig. 3D). Specifically, the conformational switch between the extended and recessed CONS26 conformations does not affect the position of the α-carbon atom of Val122; whereas, in the recessed CDKN1A conformation the α-carbon atom of Val122 has moved 6.5 Å away (Fig. 3D). It is important to note that in both the p53CR2–CONS26 and p53CR2–CDKN1A structures, loops L1 do not participate in crystal packing contacts and, therefore, their conformation cannot be a crystal packing artifact.

We report here the first, to our knowledge, 3-dimensional structure of p53 bound to a natural response element. The structure recapitulates well the features observed previously in the complex of a multidomain p53 tetramer bound to a consensus DNA site. Thus, it provides additional support for our conclusion that binding of p53 to specific DNA sites is accompanied by conformational changes in the p53–DNA-binding domain, which, in turn, regulate the half-life of the p53–DNA complex (15).

Perhaps the most interesting feature of the p53CR2–CDKN1A structure is the novel conformation of loop L1 of the outer subunits. This novel conformation is satisfying in the sense that the hydrophobic side chain of Val122 points towards the protein interior and the side chain of Ser121 contacts the DNA phosphate backbone (Fig. 3A). In the extended loop L1 conformation, the side chain of Val122 also faces towards the protein interior and Ser121 contacts the DNA (Fig. 2A). However, the α-carbon atoms of Val122 and Ser121 have moved 6.5 and 10.4 Å, respectively, between the extended and p53CR2–CDKN1A recessed conformations (Fig. 3D).

The recessed loop L1 conformation in the previously determined p53CR2–CONS26 structure can be seen to adopt a conformation that is intermediate between the extended and recessed p53CR2–CDKN1A conformations (Fig. 3D). One could argue that loop L1 adopts many different conformations according to the DNA sequence being contacted. Formally, we cannot exclude this possibility. In fact, different loop L1 conformations may allow p53 to exhibit different DNA-binding kinetics for different DNA sites, which would translate in different half-lives of the p53–DNA complexes (15). However, it is also possible that the p53CR2–CONS26 structure has trapped a transition intermediate between the extended and recessed p53CR2–CDKN1A conformations. This latter possibility is attractive because the solvent-exposed hydrophobic side chain of Val122 observed in the p53CR2–CONS26-recessed conformation (Fig. 3C) could serve to enhance the kinetic barrier for the conformational switch. Once, this barrier is bypassed, the side chain of Val122 would again face towards the protein interior (Fig. 3B), stabilizing the DNA-bound recessed p53–CDKN1A conformation. In support of this possibility, we have already shown that substitutions of Val122 with Gly or Ser decrease the half-life of sequence-specific p53–DNA complexes without an accompanying decrease in affinity (15). Determination of structures of p53 bound to a variety of natural and consensus p53-response elements will help elucidate the spectrum of loop L1 conformations and their role in regulating the DNA-binding properties of p53.

No potential conflicts of interest were disclosed.

The authors thank the staff of the ESRF, Grenoble, France for help in collecting the X-ray diffraction data. Three-dimensional structure coordinates and structure factors have been deposited at the Protein Data Bank under the accession number 3TS8.

This work was supported by a Swiss National Foundation grant to T. D. Halazonetis.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.