Acetylation of lysine 120 of p53 endows DNA-binding ... of lysine 120 of p53 endows DNA-binding specificity at effective physiological salt concentration Eyal Arbelya, Eviatar Natana,

Acetylation of lysine 120 of p53 endowsDNA-binding specificity at effectivephysiological salt concentrationEyal Arbelya, Eviatar Natana, Tobias Brandta, Mark D. Allena, Dmitry B. Veprintseva,1, Carol V. Robinsonb, Jason W. China,Andreas C. Joergera, and Alan R. Fershta,2

aMedical Research Council Laboratory of Molecular Biology, Hills Road, Cambridge CB2 0QH, United Kingdom; and bDepartment of Chemistry, Physicaland Theoretical Chemistry Laboratory, University of Oxford, South Parks Road, Oxford OX1 3QZ, United Kingdom

Contributed by Alan R. Fersht, March 29, 2011 (sent for review February 21, 2011)

Lys120 in the DNA-binding domain (DBD) of p53 becomes acety-lated in response to DNA damage. But, the role and effects ofacetylation are obscure. We prepared p53 specifically acetylatedat Lys120, AcK120p53, by in vivo incorporation of acetylated lysineto study biophysical and structural consequences of acetylationthat may shed light on its biological role. Acetylation had no affecton the overall crystal structure of the DBD at 1.9-Å resolution, butsignificantly altered the effects of salt concentration on specificityof DNA binding. p53 binds DNA randomly in vitro at effectivephysiological salt concentration and does not bind specifically toDNA or distinguish among its different response elements untilhigher salt concentrations. But, on acetylation, AcK120p53 exhib-ited specific DNA binding and discriminated among response ele-ments at effective physiological salt concentration. AcK120p53 andp53 had the highest affinity to the same DNA sequence, althoughacetylation reduced the importance of the consensus C and G atpositions 4 and 7, respectively. Mass spectrometry of p53 andAcK120p53 DBDs bound to DNA showed they preferentially segre-gated into complexes that were either DNAðp53DBDÞ4 or DNAðAcK120DBDÞ4, indicating that the different DBDs prefer differentquaternary structures. These results are consistent with electronmicroscopy observations that p53 binds to nonspecific DNA indifferent, relaxed, quaternary states from those bound to specificsequences. Evidence is accumulating that p53 can be sequesteredby random DNA, and target search requires acetylation of Lys120and/or interaction with other factors to impose specificity ofbinding via modulating changes in quaternary structure.

The tumor suppressor p53 is a key component in the signalingnetwork activated in response to oncogenic and other cellular

stresses. It is pivotal in the prevention of cancer development, bypromoting cell cycle arrest, senescence, or apoptosis (1). As atranscription factor, p53 is active as a homotetramer of multido-main monomers, each composed of the tetramerization (Tet) andDNA-binding domains (DBD) and intrinsically disorderedregions (2). Tetrameric p53 binds to DNA response elements(REs) that consist of two decameric motifs or half sites of thegeneral form RRRCWWGYYY (R ¼ A, G; W ¼ A, T; Y ¼ C,T) separated by 0–14 base pairs, although more than two spacingbase pairs is rare (3). Although once thought essential, the C atposition 4 and the G at 7 are now known to be replaced in someREs, and there is a huge range of targets (4).

The complex activities of p53 are tightly regulated by an assort-ment of protein–protein interactions and posttranslational mod-ifications (5, 6). More than 35 different amino acids in p53 areposttranslationally modified during normal homeostasis andunder stress. In many cases, the exact role of individual posttran-slational modification is not clear. Here, we study the acetylationof Lys120 in the DBD. In response to DNA damage, Lys120 isacetylated by TIP60 and hMOF acetyltransferases of the MYSTfamily (7, 8). There is evidence that the acetylation of Lys120is important for p53-induced apoptosis, and not cell cycle arrest(7, 8). In contrast to wild-type p53, the K120R mutant is incap-

able of promoting the transcription of proapoptotic genes BAXor PUMA and inducing p53-mediated apoptosis. This discrimina-tion was explained by the selective accumulation of the Lys120-acetylated form of p53, Ac120Kp53, at proapoptotic BAX andPUMA promoters, seen by ChIP of camptothecin-treated LNCaPand MCF-7 cell lines (7). It was also suggested that Lys120 acet-ylation is important for p53-induced transcription-independentapoptosis, mainly via the mitochondria (9). The population ofp53 in the mitochondria is enriched with the acetylated Lys120variant, although the acetylation is not required for the localiza-tion of p53. Further, Lys120 acetylation is essential for effectivedisplacement of the antiapoptotic protein MCL-1 from the com-plex with proapoptotic BAK (9). Hence, it is assumed that Lys120acetylation differentiates between p53-mediated cell cycle arrestand apoptosis.

The exact mechanism by which Lys120 acetylation promotesapoptosis is not clear, especially its correlation with the acetyla-tion of six lysine residues at the C terminus. For example, homo-zygous mutant mice in which the C-terminal lysine residues aremutated to arginine show no significantly different p53 responsefrom wild-type mice, except for partially impaired transcriptionalactivation upon DNA damage (10). There are similar results forthe same mutations introduced into endogenous p53 in mouseembryonic stem cells (11). However, when all eight lysines thatare known to be acetylated are mutated to arginine (six at theC terminus and two within the DNA-binding domain, includingK120), p53 is incapable of inducing growth arrest and apoptosis,but at the same time is able to retain its DNA-binding capacityas a transcription factor and induce the p53-Mdm2 feedbackloop (12). Although Lys120 directly interacts with DNA (13, 14),the eight lysine-to-arginine mutations have no effect on DNAbinding, as measured by electrophoretic mobility shift assay.In contrast, a different study showed that the mutant K120Aexpressed at low concentration in H1299 cells has lower affinityfor DNA than does the wild-type protein (15).

In order to study the effect of Lys120 acetylation on p53structure and DNA-binding properties, we used an evolvedMethanosarcina barkeri pyrrolysyl-tRNA synthetase∕tRNACUApair to encode genetically the incorporation of acetylated lysinein response to an amber stop codon in Escherichia coli (16). This

Author contributions: E.A., E.N., A.C.J., and A.R.F. designed research; E.A., E.N., T.B.,M.D.A., and A.C.J. performed research; D.B.V., C.V.R., and J.W.C. contributed newreagents/analytic tools; E.A., E.N., T.B., A.C.J., and A.R.F. analyzed data; and E.A. andA.R.F. wrote the paper.

The authors declare no conflict of interest.

Data deposition: The atomic coordinates and structure factors of the Lys120-acetylatedDBD have been deposited in the Protein Data Bank, www.pdb.org (PDB ID code 2YBG).1Present address: Biomolecular Research Laboratory, OFLC/103 Paul Scherrer Institut, 5232Villigen PSI, Switzerland.

2To whom correspondence should be addressed. E-mail: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1105028108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1105028108 PNAS ∣ May 17, 2011 ∣ vol. 108 ∣ no. 20 ∣ 8251–8256

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

CHEM

ISTR

Y

system allowed us to express and purify a superstable quadruplemutant p53 (17) and the wild-type DBD, acetylated at positionLys120 (AcK120p53 and AcK120DBD, respectively). We wereable to measure the binding affinity of AcK120p53 to differentDNA REs, solve the crystal structure of the AcK120DBD, andperform mass spectrometry studies to search for biophysicaland structural effects of acetylation that may be relevant to itsbiological role.

ResultsExpression and Purification of Site-Specifically Acetylated p53. Weexpressed site-specifically acetylated p53 in E. coli cells by geneti-cally encoding the incorporation of Nε-acetyl-lysine in responseto the amber stop codon. The quadruple-mutant version of full-length p53 (p53) (17, 18) or the wild-type DBD (residues 94–293)carrying a TAG stop codon at position 120, were coexpressed withan orthogonal Nε-acetyllysyl-tRNA synthetase∕tRNACUA pair, inthe presence of Nε-acetyl-lysine. Incorporation of acetylatedlysine was verified by Western blot, using an antibody againstacetylated lysine (Fig. 1, lanes 3–6), and by electrospray ioniza-tion mass spectroscopy. Typical yields of pure protein were about20% to 30% of that obtained for the nonacetylated version due,mainly, to the expression of the truncated version (residues1–119, Fig. 1, lane 9). When protein was expressed withoutNϵ-acetyl-lysine in the medium (Fig. 1, lane 10), only the trun-cated version could be detected byWestern blot using an antibodyagainst residues 44–56 of p53. We found that in vitro, Lys120acetylation has no effect on the thermodynamic and kinetic sta-bility of the DBD (Fig. S1). In addition, our data suggest that invitro, AcK120DBD does not bind to Mcl-1 (residues 183–330)with a Kd in the micromolar range (Fig. S2).

DNA-Binding Affinity and Specificity. We used fluorescence aniso-tropy in order to study the effect of Lys120 acetylation onDNA-binding affinity and specificity (Table 1) (19, 20). We com-pared the binding of p53 and AcK120p53 to REs with differentaffinities: (i) high affinity sequences—a 20 base-pair consensus-sequence response element (20 bp RE) (21) and the p21 RE; (ii)moderate affinity sequence, PUMA2; (iii) low affinity sequences,BAX and PUMA1 REs; and (iv) as controls for nonspecific bind-ing, a 20-bp random sequence and the 30-bp NFκB RE (Table 1).

We used three major buffers containing 90, 150, or 225 mMNaCl as salts plus 25 mM phosphate buffer pH 7.2. These havecalculated ionic strengths, I, of 150 mM, 211 mM, and 286 mM,respectively. Strictly speaking, the classical concept of ionicstrength breaks down because of ion condensation effects, butfor convenience we refer to these buffers in terms of ionicstrength. The middle buffer 211 mM I is approximately equalto the effective salt concentration in vivo. At 150 mM I, p53bound all REs within an affinity range of 22–59 nM. Ac120Kp53had similar affinities to p53, apart from weaker binding to the20-bp random sequence. In effective physiological salt concentra-tion buffer (211 mM I), p53 began to discriminate weakly againstthe 20-bp random sequence, but not among the others. In con-trast, AcK120p53 showed strong discrimination against the 20-bprandom sequence and significant discrimination among the REs.In 286 mM I buffer, p53 showed strong discrimination against the20-bp random sequence and the 30-bp NFκB, as well as BAX andPUMA REs. AcK120p53 showed even stronger discrimination.Previous experiments with a construct of p53 containing justthe DBD and the tetramerization domain (CT) showed a similarloss of specificity at 150 mM I, but specificity at 225 mM I(19, 20). For comparison in Table 1, CT has a similar specificityat 225 mM I buffer to AcK120p53 at 211 mM I buffer.

The addition of 2 mM Mg2þ, which is above mammaliancellular levels of 0.2–0.6 mM (22–24), to the effective physiolo-gical salt concentration buffer (211 mM I) did not affect the bind-ing of strongly binding REs but did weaken the binding of a weak

binder (Table 1) so that AcK120p53 reached its full selectivity,which was still much higher than that for nonacetylated p53, mea-sured under the same Mg2þ concentrations. Substitution of KClfor NaCl had a negligible effect on binding (Table 1).

We scanned the more general specificity of AcK120p53 bysynthesizing 30 different competitor DNA sequences, systemati-cally mutating each position within one half-site of the referencesequence (positions 1–10 in Fig. 2) and measuring their affinityby competition with a fluorescently labeled standard sequence(Table S1) (21). AcK120p53 had the highest affinity to the sameDNA sequence as p53, measured at 286 mM I (Fig. 2). But, theacetylation of Lys120 lowers the importance of positions 4, thecanonical C, and 7, the canonical G (equivalent to positions 14and 17 in our longer sequences) in defining the preferred DNA-binding site. Compared with nonacetylated p53, the importanceof other positions in the consensus sequence was not affected bythe acetylation of Lys120 (Table S1), including positions 3 and 8(equivalent to positions 13 and 18), which are known to formhydrogen bonds with the Lys120 side chain (13, 14).

Structure of the Acetylated Core Domain. We solved the crystalstructure of AcK120 DBD (residues 94–293) at 1.9-Å resolution(Table 2). The crystals used for structure solution belonged tospace group P21 with four molecules in the asymmetric unitand were isomorphous to those reported for the nonacetylatedwild-type DBD (25) and a cancer-related mutant (26). The over-all structure of the β-sandwich and the DNA-binding surface ofAcK120 DBD was virtually identical to that of the nonacetylatedDBD [Fig. 3, PDB ID code 2OCJ (25)]. The different moleculesof AcK120 DBD and nonacetylated DBD can be superimposedwith a rmsd of 0.3–0.8 Å, which is in the same range as the rmsdwhen superimposing the molecules within each asymmetric unit.As with all structures of unbound DBD, the side chain of Lys120showed a high degree of flexibility, and the lack of defined elec-tron density prevented unambiguous modeling of the acetylatedlysine side chain. The backbone conformation of the L1 loop,however, was well defined and similar to that of the unmodifiedprotein, indicating that K120 acetylation does not alter theconformation of this loop in its DNA-free state.

Composition of Acetylated and Nonacetylated p53DBDs in Complexwith DNA. We studied the composition of complexes formedby competition of excess concentrations of p53DBD and

Fig. 1. Expression of site-specifically acetylated p53. Coomassie blue stainingof site-specifically acetylated and nonacetylated p53 (lanes 1, 2). Only pro-teins expressed in the presence of acetylated lysine were detected (lanes3–6) on analysis by Western blot using an antibody against Nε-acetyl-lysine.Using an antibody against p53 (residues 44–56), we detected both acetylatedand nonacetylated proteins (lanes 7–10). The presence of the amber stopcodon in position 120 gives rise to the expression of a truncated version (lane9, middle band, lipoyl domain and residues 1–119) together with the full-length protein (lane 9, upper band, lipoyl domain and full-length p53).Specificity of the Nε-acetyllysyl-tRNA synthetase∕tRNACUA pair is demon-strated by only the truncated version being expressed when cells were grownin media not supplemented with Nε-acetyl-lysine (lane 10).

8252 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1105028108 Arbely et al.

AcK120DBD for the 20-bp RE, for which they have equalaffinity. We incubated different ratios of 13C-15N-labeled nona-cetylated DBD and AcK120DBD with the consensus 20-bp REand used soft ionization electrospray mass spectrometry (ESI-MS) to analyze the composition of the resultant complexes(Fig. 4) (27–29). The relative abundance of a particular specieswas calculated from the ratio between the sum of intensities as-sociated with that species and the sum of all intensities. Comple-tely random formation could theoretically result in five differentspecies, with DBD∶AcK120DBD stoichiometries of 0∶4, 1∶3,

2∶2, 3∶1, or 4∶0. A binomial distribution of equal amounts ofboth species should yield a ratio of 1∶4∶6∶4∶1.

The spectra in Fig. 4 were obtained when 13C-15N-labelednonacetylated DBD and AcK120DBD were mixed at differentmolar ratios with DNA. As expected, when only 13C-15N-labeled

Table 1. DNA-binding affinities determined by fluorescence anisotropy titration

I ¼ 150 mM I ¼ 211 mM I ¼ 286 mM I ¼ 225 mM

DNA*p53 WTKd∕nM (h)

AcK120Kd∕nM (h)

p53 wtKd∕nM (h)

AcK120Kd∕nM (h)

p53Kd∕nM (h)

AcK120Kd∕nM (h)

CT†

Kd∕nM (h)

20-bp random 59 (1.7) 140 (1.2) 290 (1) 1,100 (1) ≫1;000 ≫1;00020-bp random + 2 mM MgCl2 380 (1)20-bp random KCl buffer 310 (1)30-bp NFκB 25 (2.1) 35 (2.0) 41 (1.4) 85 (1.2) 500 (1) ≫1;00020-bp RE 22 (1.5) 17 (1.4) 18 (1.3) 22 (1.3) 51 (1.3) 70 (1)20-bp RE + 2 mM MgCl2 18 (1.6) 25 (1.4)P21 40 (1.8) 29 (1.8) 26 (1.9) 33 (1.3) 55 (1.2) 180 (1) 10P21 KCl buffer 38 (1.7)PUMA2 39 (1.5) 30 (1.5) 24 (1.9) 66 (1) 69 (1) 1,000 (1) 14PUMA2 + 2 mM MgCl2 36 (1.4) 120 (1)BAX 46 (1.6) 45 (1.3) 33 (1.4) 390 (1) 570 (1) ≫1;000 150PUMA1 35 (1.6) 43 (1.4) 32 (1.5) 350 (1) ≫1;000 ≫1;000 520PUMA1 + 2 mM MgCl2 80 (1.1) ≫1;000PUMA1 KCl buffer 45 (1.5)

Data fitted to Hill equation, with variable Hill constant (h) (20). Tabulated values of Kd are concentrations of p53 in terms of monomer required for50%binding of labeled DNA. The apparent cooperativity, h, of binding results from the association of p53 dimers into tetramers with a Kd of 20 nM (45,46). Thus, the predominant form of p53 during the measurement of weak binders (that is around the concentration for 50% binding) is the tetramer,and the binding is not cooperative (20).*20-bp random ¼ 50-GGAAATTTCCGGAAATTTCC-30.30-bp NFκB ¼ 50-TCGACAGAGGGGACTTTCCGAGAGGCTCGA-30.20-bp RE ¼ 50-GGACATGTCCGGACATGTCC-30.p21 ¼ 50-ATCAGGAACATGTCCCAACATGTTGAGCTC-30.PUMA2 ¼ 50-CGCGCCTGCAAGTCCTGACTTGTCCGCGGC-30.Bax ¼ 50-TGGGCTCACAAGTTAGAGACAAGCCTGGGCG-30.PUMA1 ¼ 50-GGGTCCTCCTTGCCTTGGGCTAGGCCCTGCC-30.Labeled DNA sequences were 5′ fluorescein- or Alexa488-labelled (see Materials and Methods).†Data for the core plus Tet domain construct (CT) are taken from Weinberg et al. (19).

Fig. 2. Sequence logo representation of DNA binding. Fluorescently labeledDNA was allowed to form a complex with AcK120p53 (A) or p53 (21) (B) andthe complex was titrated with unlabeled competitor DNA. Bit values (48) areplotted against the position in sequence. Bar heights indicate the relativeimportance of different nucleotides at each position. Both proteins hadthe highest affinity to the same consensus sequence written at the bottom.

Table 2. X-ray data collection and refinement statistics

A. Data CollectionSpace group P21Cell, a, b, c, Å 68.93, 69.58, 83.49Cell, α, β, γ, ° 90.00, 90.12, 90.00Molecules∕AU 4Resolution, Å* 49.0–1.9 (2.0–1.9)Unique reflections 60,047Completeness, %* 96.6 (98.6)Multiplicity* 2.0 (2.0)Rmerge, %*,† 8.0 (34.1)hI∕σIi* 7.5 (2.4)Wilson B value, Å2 17.9B. RefinementNumber of atoms

Protein‡ 6,077Water 644Zinc 4

Rcryst, %§ 17.4Rfree, %§ 22.6rmsd bonds, Å 0.006rmsd angles, ° 0.97Mean B value, Å2 20.9Ramachandran favored, %¶ 98.2Ramachandran outliers, %¶ 0.0

*Values in parentheses are for the highest resolution shell.†Rmerge ¼ ∑ðIh;i − hIhiÞ∕∑ Ih;i‡Number includes alternative conformations.§Rcryst and Rfree ¼ ∑ jjFobsj − jFcalcjj∕∑ jFobsj, where Rfree was calculated over5% of the amplitudes chosen at random and not used in the refinement.

¶Determined using MOLPROBITY (47).

Arbely et al. PNAS ∣ May 17, 2011 ∣ vol. 108 ∣ no. 20 ∣ 8253

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

CHEM

ISTR

Y

nonacetylated DBD was mixed with DNA at a molar ratio of 8∶1(monomer to double stranded DNA), only homotetramers ofnonacetylated DBD and DNAwere observed (bottom spectrum).When equal amounts of 13C-15N-labeled nonacetylated DBD andAcK120DBD were mixed with DNA at molar ratios of 8∶8∶1,both homotetramers and heterotetramers formed in solution(top spectrum). The hetero-tetramers were composed of differ-ent ratios between acetylated (red dots) and nonacetylated p53DBDs (blue dots), as marked in Fig. 4 (Top) according to thecomposition of the tetramer associated with each peak.

All theoretically possible species were observed (Fig. 5), butimportantly the ratio of these complexes significantly deviatedfrom the values expected for a binomial distribution. Of the totalpopulation of complexes monitored, 20% were composed ofDNA and four nonacetylated DBDs, whereas 17% were com-posed of DNA and four AcK120DBDs (Fig. 5), three times high-er than expected from a random binomial distribution, of about6%. The deviation from binomial distribution was even moresignificant for the 2∶2 heterotetramers: 7% observed versus anexpected 38%. AcK120DBDs and p53DBDs prefer to formhomocomplexes with DNA and have the least preference forDNAðAcK120DBDÞ2ðp53DBDÞ2 heterocomplexes.

DiscussionUsing an orthogonal pyrrolysyl-tRNA synthetase/tRNA pairevolved to incorporate Nε-acetyllysine in response to the amberstop codon (16), we were able to express in vivo and purify p53and its DBD that were site-specifically acetylated at positionLys120 and then study in vitro the effect of Lys120 acetylationon the structure and activity of p53. Acetylation did not signifi-cantly alter the crystal structure of the DBD, although the acety-lated side chain of Lys120 was too mobile to be located. Thethermodynamic and kinetic stability of the DBD were unaffected(Fig. S1). But, there were unexpected consequences on thespecificity of binding to DNA.

Acetylation of Lys120 and DNA Binding. Salt concentration. The mostsignificant effect of acetylation of Lys120 on the activity of p53 isto modulate its sensitivity to salt concentration. Previous studiesonmeasuring the specificity of binding of p53 in vitro (using a sim-plified construct, CT, lacking the N and C termini) note that p53binds indiscriminately to DNA at lower salt concentrations, notdistinguishing between different response elements or even ran-dom sequences. So, measurements were made at an ionic strengthof 225 mM (19) where discrimination is observed. Subsequentsystematic surveys of full-length p53 in vitro have needed touse higher salt concentrations still, equivalent to 286 mM I (21).

The value of intracellular salt concentration is often consid-ered to be equivalent to a calculated ionic strength of 250 mM.Allowance for formation of complexes between ions reduces thevalue to 200 mM (23). An effective physiological ionic strengthof 215 mM I is also calculated (30). However, the classicalDebye–Hückel theory of electrostatic effects of ions breaks down

Zn

L1 loop

L2 loop

L3 loopDNA-binding surface

Lys120

N

Fig. 3. Structural effect of Lys120 acetylation. Superposition of the crystalstructures of AcK120DBD (gray; PDB ID code 2YBG, this work) and nonace-tylated DBD (blue; PDB ID code 2OCJ) in their DNA-free form shows thatacetylation of Lys120 has no significant effect on the conformation of theL1 loop in the unbound DBD. Both structures are shown as cartoon represen-tation, with the position of residue 120 highlighted in red. The figure wasgenerated using PyMOL (www.pymol.org).

5,500 7,5006,500 m/z

::8 0 1

DNA::::8 8 1

WT DBD AcK120 DBD

WT AcK

Fig. 4. ESI-MS of complexes formed between AcK120-DBD and/or 13C-15N-labeled nonacetylated DBD and DNA. Nonacetylated DBD is marked by bluedots, whereas AcK120DBD is marked by red dots. Molar ratio of13C-15N-DBD∶AcK120DBD∶DNA gradually changed from 8∶0∶1 (Bottom) to8∶8∶1 (Top). The composition of the complex associated with each peak ismarked at the top (i.e., the ratio between 13C-15N-DBD and AcK120DBD).The intensity of each peak is proportional to the relative abundance ofeach species. Dashed lines represent the expected m∕z value for theDNAðp53DBDÞ2ðAcK120DBDÞ2 complex (m∕z ¼ 6;461, 6,865, and 7,322). Onlythe peak at m∕z ¼ 6;865 was detected, due to low occurrence of theDNAðp53DBDÞ2ðAcK120DBDÞ2 complex.

Binomial distribution Experimental

Per

cent

of c

ompl

exes

Number of AcK120 monomers

0

10

20

30

40

0 1 2 3 4

6

25

37

25

6

20

32

7

24

17

Fig. 5. Distribution of AcK120DBD and nonacetylated DBD in tetramericcomplex with DNA. Equal amounts of 13C-15N-nonacetylated DBD andAcK120DBDwere mixed with DNA at molar ratios of 8∶8∶1 and samples wereanalyzed by ESI-MS. Results are presented as function of number ofAcK120DBDs in each complex: 0—a homotetramer composed of four nona-cetylated DBDs; 4—a homotetramer composed of four AcK120DBDs. Barsrepresent the distribution of complexes relative to the total amount offormed complexes. For example, 17% of the complexes were in the formof DNAðAcK120DBDÞ4 complex, whereas 7% were in the form of DNAðAcK120DBDÞ2ðp53DBDÞ2 complex. Experimental data are compared to re-sults expected in a nonbiased experiment, assuming binomial distribution.

8254 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1105028108 Arbely et al.

at these high concentrations, and it is better to use a term such aseffective physiological salt concentration (31). Here, at 150 mM I,both p53 and AcK120p53 bound DNA near indiscriminately,with equal affinity (Table 1). But, at 211 mM effective salt con-centration, close to physiological effective salt concentrations,AcK120p53 achieved >10-fold discrimination between p21 andBax response elements, whereas p53 still did not discriminatebetween different DNA response elements and bound well toa random 30-mer. Nonacetylated p53 exhibited specificity at286 mM I, as expected (21).

Mg2þ ions can affect DNA–protein interactions by competingfor DNA sites (31, 32). The concentration of Mg2þ in mammaliancells is about 0.2–0.6 mM (22–24). The addition of 2 mMMg2þ tothe effective physiological salt buffer (Table 1) had little effect onthe binding of either AcK120p53 or nonacetylated p53 to strong-binding REs, but it weakened the binding to weak-binding REs.The affinity of nonacetylated p53 to Puma1 was reduced by a fac-tor of approximately 2 (from 32 to 80 nM) and the affinity ofAcK120p53 was reduced from 350 nM to >1 μM. The specificityof AcK120p53 for Puma1 was thus increased to the level found at286 mM I. Perhaps, other divalent ions could also enhance selec-tivity. A major effect of acetylation of Lys120 is to endow p53 withselectivity at salt concentrations close to the effective physiologi-cal values where nonacetylated p53 does not exhibit selectivity.

General specificity of AcK120p53. The affinity of p53 for variousREs is mostly affected by mutations of the C at position 4 or Gat position 7 (Fig. 2), but acetylation of Lys120 relaxed somewhatthe stringent requirements for C and G at those positions(Table S1). It implies that acetylation of Lys120 shifts the equili-brium toward binding to REs with lower binding affinities, inaddition to reducing nonspecific DNA binding. Competitionbetween binding to specific and nonspecific DNA sequence haspreviously suggested as part of the lactose (lac) operon regulatorysystem (33). According to this model, the binding of lac repressorto specific DNA relative to nonspecific DNA is a crucial para-meter in the regulatory mechanism of the lac operon. Accordingto in vitro measurements presented here, at effective physiologi-cal salt concentration the ratio between specific (p21 RE) to non-specific DNA-binding affinities increases from 11 to 33 uponacetylation (Table 1). Assuming similar in vivo binding affinities,these data suggest a potential regulatory role for the acetylationof Lys120 by controlling the ratio between specific to nonspecificbinding of p53 to DNA and hence its activity as a transcriptionfactor.

Structural Consequence of Acetylation. AcK120DBD had the samecrystal structure as p53DBD, with the acetylated side chain toomobile to be located. We have been unable to crystallize com-plexes of AcK120DBD with DNA, but obtained useful datafrom mass spectrometry of complexes formed between a RE anda mixture of excess p53DBD and AcK120DBD. According to thespectra in Fig. 4, the homocomplexes DNAðp53DBDÞ4 or DNAðAcK120DBDÞ4 were preferentially formed relative to DNAðp53DBDÞ2ðAcK120DBDÞ2 (Fig. 5). Lys120 is at the DNA–

protein interface of DNAðp53Þ4 complexes and not the protein–protein interfaces (13, 14). Accordingly, DNAðp53DBDÞ4 andDNAðAcK120DBDÞ4 must have different protein–protein inter-faces, induced by geometrical changes at the protein–DNA inter-faces. These results are entirely consistent with recent electronmicroscopy observations that p53 binds well to nonspecific DNAat 211 mM I, but in different, relaxed, quaternary states from p53bound to specific sequences (34).

Implications for Searching for Target Response Elements. p53 bindstightly to nonspecific DNA. p53 has to search for its targetsequences among the nonspecific DNA, which is in vast excessso that, even if it binds less tightly, would sequester p53. Electron

microscopy shows that the binding to nonspecific sequencesoccurs from p53 adopting different conformational states, leadingto speculation that accessory proteins could force p53 tetramersto adopt the “specific binding conformation” (34). Acetylation ofLys120 is now seen to increase the specificity of p53 and changeits conformational preferences. If the data in vitro extend to thesituation in vivo, then p53 prior to acetylation of Lys120 wouldnot be in a search process for its targets, but would bind toany exposed nonspecific sequences, possibly sliding along them(35). Acetylation would then start the search process by weaken-ing the binding to nonspecific sequences.

Materials and MethodsProtein Expression and Purification. Nonacetylated versions of full-lengthp53 and DBD (residues 94–293) were expressed in BL21 cells as previouslydescribed (36). Briefly, cells induced with 1 mM IPTG and incubated at22 °C overnight in 2×TY medium supplemented with 0.1 mM ZnCl2. Cellswere lysed in 50 mM phosphate buffer pH 8, 300 mM NaCl, protease inhibi-tors (Roche), and 10 mM β-mercaptoethanol, and clear lysate was loadedonto a Ni column. Protein was eluted (50 mM phosphate buffer pH 8,300 mM NaCl, 10 mM β-mercaptoethanol, and 250 mM Imidazole) and dia-lyzed overnight (25 mM Tris buffer pH 7.5, 300 mM NaCl, 10% glycerol, 5 mMDTT) with tobacco etch virus (TEV) protease (p53) or thrombin (p53DBD).Dialyzed solution was diluted (1∶10, ice-cold 25 mM Tris buffer pH 7.5,10% glycerol, 5 mM DTT), and cleaved p53 was separated from the His-tagged lipoyl domain using a heparin column. Protein was eluted fromthe heparin columnwith a gradient over 20 column volumes of elution buffer(25 mM Tris buffer pH 7.5, 1 M NaCl, 10% glycerol, and 5 mMDTT). Combinedfractions were concentrated using Centricon concentrators (Millipore) andfurther purified by gel filtration. DBD was purified with a HiLoad 26/60Superdex 75 column (GE Healthcare) using citrate buffer (20 mM citratebuffer pH 6.1, 150 mM NaCl, and 10 mM DTT). Full-length p53 was purifiedwith a HiLoad 26/60 Superdex 200 column (GE Healthcare) using phosphatebuffer (25 mM phosphate buffer pH 7.2, 300 mM NaCl, 5 mM DTT, and 10%glycerol). All purification steps were carried out at 4 °C.

Acetylated proteins were expressed in a similar manner, except for thefollowing modifications. A DNA fragment of K120TAG mutant p53 wascloned as N-terminal fusion of His6, lipoyl domain, and TEV protease cleavagesite (p53) or thrombin cleavage site (p53DBD) between XhoI and NcoI sites onpCDF-Duet vector encoding the pyrrolysine tRNA gene from M. barkeri withan lpp promoter and rrnC terminator (16). BL21 cells were cotransformedwith the above plasmid and a pAcKRS-3 plasmid (16) (coding for M. barkeripyrrolysine tRNA synthetase with the mutations L266M, L270I, Y271F, L274A,and C313F) and grown in 2×TY medium supplied with 10 mM acetylatedlysine. Before induction with 1 mM IPTG, cells were supplemented with20 mM Nicotinamide. Purification of acetylated proteins was carried out asdescribed above for nonacetylated proteins.

DNA-Binding Affinity and Specificity. DNA-binding affinities to specificREs were measured by fluorescence anisotropy, using a Cary Eclipse (Varian)spectrometer equipped with a Microlab M dispenser (Hamilton). Excitationand emission wavelength were 480 nm and 530 nm, respectively. All mea-surements were done at 20 °C. Buffers of varying salt concentrations wereused (90 or 150 or 225 mM NaCl plus 25 mM phosphate buffer pH 7.2,10% (vol∕vol) glycerol, 5 mM DTT, 0.2 mg∕mL bovine serum albumin). The150 mM KCl buffer was identical to the NaCl buffer, except that potassiumsalts were used instead of sodium salts. Measurements in the presence ofMg2þ ions were made in 145 mM NaCl, 2 mM MgCl2 plus 25 mM phosphatebuffer pH 7.2, 10% (vol∕vol) glycerol, 5 mM DTT, 0.2 mg∕mL bovine serumalbumin. In a typical experiment, 250 μL of 1–5 μM protein solution weretitrated into 900 μL of 20 nM labeled DNA solution. Data were correctedfor the resulting dilution effect. Fluorescence intensities were measured60 s after each titration step to allow equilibration. All data analysis was doneusing laboratory software (results are described in Table 1). All sequenceswere 5′ fluorescein-labeled, except for 20-bp RE, which was 5′ Alexa488-labeled.

General DNA binding was measured by fluorescence anisotropy in 96-wellplates using a Pherastar plate reader (BMG Labtech) equipped with a Bravo96-channel pipetting robot (Velocity 11) as previously described (21). In atypical titration experiment, 120 nM (final monomer concentration) of full-length p53 was allowed to form a complex with 20 nM of the 20-base-pairreporter sequence GGACATGTCCGGACATGTCC, labeled at the 5′ end withAlexa488 fluorophore, and the complex was then titrated with unlabeledDNA (50 μM) at 22 °C. The reporter sequence consists of two identical copies

Arbely et al. PNAS ∣ May 17, 2011 ∣ vol. 108 ∣ no. 20 ∣ 8255

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

CHEM

ISTR

Y

of the half-site GGACATGTCC, known to be one of the tightest-bindingsequences for human p53. First, binding affinity was measured betweenAcK120p53 and a reference DNA—a nonlabeled version of the reporterDNA. This binding affinity was then compared to that of the other 31 com-petitor DNA sequences, in which a single mutation was introduced system-atically at each position within one half-site of the palindromic referencesequence. Buffer conditions for all experiments were 25 mM sodium phos-phate pH 7.2, 225 mM NaCl (286 mM total I), 10% vol∕vol glycerol, 5 mMDTT and 0.2 mg∕mL bovine serum albumin. Data were analyzed accordingto cooperative binding and competition models using laboratory-developedsoftware (21).

Crystallization and Structure Solution. Crystals of acetylated DBD were grownusing the sitting drop vapor diffusion technique with 26% (wt∕vol) PEG 3350,43 mM sodium acetate, and 100 mM Hepes, pH 7.5 as the crystallization solu-tion. Drops composed of 200 nL of protein at 200 μM (20 mM citratebuffer pH 6.1, 150 mM NaCl, and 10 mM DTT) and 200 nL of crystallizationsolution were equilibrated for a minimum of 3 d at 17 °C above a reservoirsolution of 100 μL. Crystals were transferred tomother liquor containing 20%glycerol and flash-frozen in liquid nitrogen. The protein crystals belongedto space group P21 with a ¼ 68.926 Å, b ¼ 69.581 Å, c ¼ 83.494 Å, andβ ¼ 90.12°. An X-ray dataset was collected on beamline ID14-2 at theEuropean Synchrotron Radiation Facility. Data were indexed and integratedusing MOSFLM (37) and were further processed using the CCP4 package (38)(Table 2). The structure was solved by molecular replacement using PHASER(39) with the chain A of PDB ID code 1TSR (40) as a search model. Model

building and structure refinement were performed using COOT (41) andPHENIX (42). Data collection and refinement statistics are summarized inTable 2.

Mass Spectrometry. Acetylated and nonacetylated p53 DBDs were dialyzedagainst 500 mM ammonium acetate pH 6.9 at 4 °C. Acetylated DBD wasmixed with the nonacetylated variant at increased ratios up to a final ratioof 1∶1. DNA was added to each of the mixtures, and the mass of the formedcomplex was immediately measured using nanoflow ESI-MS. Data wererecorded on a Synapt HDMS system (Waters Corp.) optimized for the trans-mission of noncovalent complexes (43). In a typical experiment, 3 μL of themixture were introduced by electrospray from gold-coated borosilicatecapillaries prepared in-house, as described (44). The following experimentalparameters were applied: capillary voltage ¼ 1.3–1.5 kV, sample cone ¼100 V, trap and transfer collision energy ¼ 100 V; backing pressure ¼5 mbar, source pressure ¼ 0.06–0.07 mbar; trap pressure ¼ 0.05 mbar;IMS pressure ¼ 0.5 mbar; time-of-flight analyzer pressure ¼ 1.2 × 10−6 mbar.The mass spectrometer was calibrated with cesium iodide solution(100 mg∕mL). Data were processed with MassLynx 4.0 software (Waters/Micromass) and are shown with minimal smoothing and without back-ground subtraction.

ACKNOWLEDGMENTS. This work was partly supported by European MolecularBiology Grant ALTF-650-2006 (to E.A.) and MRC Programme Grant G0901534(to A.R.F.).

1. Vousden KH, Prives C (2009) Blinded by the light: The growing complexity of p53. Cell137:413–431.

2. Joerger AC, Fersht AR (2008) Structural biology of the tumor suppressor p53.Annu RevBiochem 77:557–582.

3. Riley T, Sontag E, Chen P, Levine A (2008) Transcriptional control of human p53-regulated genes. Nat Rev Mol Cell Biol 9:402–412.

4. Menendez D, Inga A, Resnick MA (2009) The expanding universe of p53 targets. NatRev Cancer 9:724–737.

5. Kruse JP, Gu W (2009) Modes of p53 regulation. Cell 137:609–622.6. Toledo F, Wahl GM (2006) Regulating the p53 pathway: In vitro hypotheses, in vivo

veritas. Nat Rev Cancer 6:909–923.7. Sykes SM, et al. (2006) Acetylation of the p53 DNA-binding domain regulates

apoptosis induction. Mol Cell 24:841–851.8. Tang Y, Luo J, Zhang W, Gu W (2006) Tip60-dependent acetylation of p53 modulates

the decision between cell-cycle arrest and apoptosis. Mol Cell 24:827–839.9. Sykes SM, Stanek TJ, Frank A, MurphyME, McMahon SB (2009) Acetylation of the DNA

binding domain regulates transcription-independent apoptosis by p53. J Biol Chem284:20197–20205.

10. Krummel KA, Lee CJ, Toledo F, Wahl GM (2005) The C-terminal lysines fine-tuneP53 stress responses in a mouse model but are not required for stability control ortransactivation. Proc Natl Acad Sci USA 102:10188–10193.

11. Feng L, Lin T, Uranishi H, Gu W, Xu Y (2005) Functional analysis of the roles ofposttranslational modifications at the p53 C terminus in regulating p53 stabilityand activity. Mol Cell Biol 25:5389–5395.

12. Tang Y, Zhao W, Chen Y, Zhao Y, Gu W (2008) Acetylation is indispensable for p53activation. Cell 133:612–626.

13. Chen Y, Dey R, Chen L (2010) Crystal structure of the p53 core domain bound to a fullconsensus site as a self-assembled tetramer. Structure 18:246–256.

14. KitaynerM, et al. (2006) Structural basis of DNA recognition by p53 tetramers.Mol Cell22:741–753.

15. Zupnick A, Prives C (2006) Mutational analysis of the p53 core domain L1 loop. J BiolChem 281:20464–20473.

16. Neumann H, Peak-Chew SY, Chin JW (2008) Genetically encoding Nϵ-acetyllysine inrecombinant proteins. Nat Chem Biol 4:232–234.

17. Nikolova PV, Henckel J, Lane DP, Fersht AR (1998) Semirational design of active tumorsuppressor p53 DNA binding domain with enhanced stability. Proc Natl Acad Sci USA95:14675–14680.

18. Joerger AC, Allen MD, Fersht AR (2004) Crystal structure of a superstable mutantof human p53 core domain. Insights into the mechanism of rescuing oncogenicmutations. J Biol Chem 279:1291–1296.

19. Weinberg RL, Veprintsev DB, Bycroft M, Fersht AR (2005) Comparative binding of p53to its promoter and DNA recognition elements. J Mol Biol 348:589–596.

20. Weinberg RL, Veprintsev DB, Fersht AR (2004) Cooperative binding of tetrameric p53to DNA. J Mol Biol 341:1145–1159.

21. Veprintsev DB, Fersht AR (2008) Algorithm for prediction of tumour suppressor p53affinity for binding sites in DNA. Nucleic Acids Res 36:1589–1598.

22. Levy LA, Murphy E, Raju B, London RE (1988) Measurement of cytosolic freemagnesium ion concentration by 19F NMR. Biochemistry 27:4041–4048.

23. Mouat MF, Manchester KL (1998) The intracellular ionic strength of red cells and theinfluence of complex formation. Comp Hematol Int 8:58–60.

24. Rotevatn S, et al. (1989) Cytosolic free magnesium concentration in cultured chickheart cells. Am J Physiol 257:C141–C146.

25. Wang Y, Rosengarth A, Luecke H (2007) Structure of the human p53 core domain inthe absence of DNA. Acta Crystallogr D Biol Crystallogr 63:276–281.

26. Joerger AC, Ang HC, Veprintsev DB, Blair CM, Fersht AR (2005) Structures of p53 cancermutants and mechanism of rescue by second-site suppressor mutations. J Biol Chem280:16030–16037.

27. Fenn JB, Mann M, Meng CK, Wong SF, Whitehouse CM (1989) Electrospray ionizationfor mass spectrometry of large biomolecules. Science 246:64–71.

28. Sharon M, Robinson CV (2007) The role of mass spectrometry in structure elucidationof dynamic protein complexes. Annu Rev Biochem 76:167–193.

29. Wendt S, et al. (2003) Quantitative evaluation of noncovalent chorismate mutase-inhibitor binding by ESI-MS. J Am Soc Mass Spectrom 14:1470–1476.

30. Freund J, Kalbitzer HR (1995) Physiological buffers for NMR spectroscopy. J BiomolNMR 5:321–322.

31. Record MT, Jr, deHaseth PL, Lohman TM (1977) Interpretation of monovalent anddivalent cation effects on the lac repressor-operator interaction. Biochemistry16:4791–4796.

32. Kao-Huang Y, et al. (1977) Nonspecific DNA binding of genome-regulating proteinsas a biological control mechanism: Measurement of DNA-bound Escherichia coli lacrepressor in vivo. Proc Natl Acad Sci USA 74:4228–4232.

33. von Hippel PH, Revzin A, Gross CA, Wang AC (1974) Non-specific DNA binding ofgenome regulating proteins as a biological control mechanism: 1. The lac operon:Equilibrium aspects. Proc Natl Acad Sci USA 71:4808–4812.

34. Melero R, et al. (2011) Electron microscopy studies on the quaternary structure ofp53 reveal different binding modes for p53 tetramers in complex with DNA. Proc NatlAcad Sci USA 108:557–562.

35. Tafvizi A, Huang F, Fersht AR, Mirny LA, van Oijen AM (2011) A single-moleculecharacterization of p53 search on DNA. Proc Natl Acad Sci USA 108:563–568.

36. Bullock AN, et al. (1997) Thermodynamic stability of wild-type and mutant p53 coredomain. Proc Natl Acad Sci USA 94:14338–14342.

37. Evans P (2006) Scaling and assessment of data quality. Acta Crystallogr D Biol Crystal-logr 62:72–82.

38. CCP4 (1994) The CCP4 suite: Programs for protein crystallography. Acta Crystallogr DBiol Crystallogr 50:760–763.

39. McCoy AJ (2007) Solving structures of protein complexes by molecular replacementwith Phaser. Acta Crystallogr D Biol Crystallogr 63:32–41.

40. Cho Y, Gorina S, Jeffrey PD, Pavletich NP (1994) Crystal structure of a p53 tumor sup-pressor-DNA complex: Understanding tumorigenic mutations. Science 265:346–355.

41. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot.Acta Crystallogr D Biol Crystallogr 66:486–501.

42. Adams PD, et al. (2002) PHENIX: Building new software for automated crystallographicstructure determination. Acta Crystallogr D Biol Crystallogr 58:1948–1954.

43. Hernandez H, Robinson CV (2007) Determining the stoichiometry and interactions ofmacromolecular assemblies from mass spectrometry. Nat Protoc 2:715–726.

44. Nettleton EJ, et al. (1998) Protein subunit interactions and structural integrity ofamyloidogenic transthyretins: Evidence from electrospray mass spectrometry. J MolBiol 281:553–564.

45. Brandt T, Petrovich M, Joerger AC, Veprintsev DB (2009) Conservation of DNA-binding specificity and oligomerisation properties within the p53 family. BMCGenomics 10:628.

46. Rajagopalan S, Huang F, Fersht AR (2010) Single-Molecule characterization ofoligomerization kinetics and equilibria of the tumor suppressor p53. Nucleic AcidsRes 39:2294–2303.

47. Davis IW, et al. (2007) MolProbity: All-atom contacts and structure validation forproteins and nucleic acids. Nucleic Acids Res 35:W375–383.

48. Schneider TD, Stephens RM (1990) Sequence logos: A new way to display consensussequences. Nucleic Acids Res 18:6097–6100.

8256 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1105028108 Arbely et al.