FEBS Letters 584 (2010) 3035–3041

brought to you by COREView metadata, citation and similar papers at core.ac.uk

provided by Elsevier - Publisher Connector

journal homepage: www.FEBSLetters .org

MDM4 binds ligands via a mechanism in which disordered regions becomestructured

Maria C. Sanchez a, Jonathan G. Renshaw b, Gareth Davies b, Paul N. Barlow a,*, Martin Vogtherr b,**

a University of Edinburgh, Joseph Black Chemistry Building, West Mains Road, Edinburgh EH9 3JJ, United Kingdomb AstraZeneca Pharmaceuticals, Cell, Protein and Structural Sciences, Mereside, Alderley Park, Macclesfield SK10 4TG, United Kingdom

a r t i c l e i n f o

Article history:Received 28 April 2010Revised 25 May 2010Accepted 26 May 2010Available online 31 May 2010

Edited by Christian Griesinger

Keywords:MDM4MDMXMDM2p53Nuclear magnetic resonance

0014-5793/$36.00 � 2010 Federation of European Biodoi:10.1016/j.febslet.2010.05.058

Abbreviations: MDM, mouse double minute protresonance; ITC, isothermal titration calorimetry

* Corresponding author. Fax: +44 1316507055.** Correspondence to: M. Vogtherr, Merck KGaA, F

Darmstadt, Germany. Fax: +49 6151 72913535.E-mail address: Paul.Barlow@ed.ac.uk (P.N. Barlow)

a b s t r a c t

MDM2 and MDM4 are proteins involved in regulating the tumour suppressor p53. MDM2/4 and p53interact through their N-terminal domains and disrupting this interaction is a potential anticancerstrategy. The MDM2–p53 interaction is structurally and biophysically well characterised, whereasequivalent studies on MDM4 are hampered by aggregation of the protein. Here we present theNMR characterization of MDM4 (14-111) both free and in complexes with peptide and small-mole-cule ligands. MDM4 is more dynamic in its apo state than is MDM2, with parts of the protein beingunstructured. These regions become structured upon binding of a ligand. MDM4 appears to bind itsligand through conformational selection and/or an induced fit mechanism; this might influencerational design of MDM4 inhibitors.

Structured summary:MINT-7896835: p53 (uniprotkb:P04637) and MDM4 (uniprotkb:O15151) bind (MI:0407) by isothermaltitration calorimetry (MI:0065)MINT-7896820: p53 (uniprotkb:P04637) and MDM4 (uniprotkb:O15151) bind (MI:0407) by nuclear mag-netic resonance (MI:0077)

� 2010 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved.

1. Introduction [5]. X-ray crystallographic structures show nutlin bound to the

The tumour suppressor p53 plays a central role in cellularresponse to stress signals [1]. Loss of p53 function throughmutation or inactivation is often found in cancer. One inactiva-tion route involves MDM2, which acts as a specific E3 ubiquitinligase promoting degradation of p53 [2]. MDM2 and p53 interactprimarily through their N-terminal domains; in particular the100 aa MDM2 N-terminal domain binds to the p53 transactiva-tion sequence. Upon binding this otherwise unstructuredsequence within p53 adopts a helical fold [3]. Disruption of thisinteraction is a strategy to combat some types of cancer [4].Rational design based on the three-dimensional arrangement ofside chains in the MDM2-bound p53 peptide led to specificsmall-molecule inhibitors, which are exemplified by the nutlins

chemical Societies. Published by E

ein; NMR, nuclear magnetic

rankfurter Str. 250, D-64293

.

p53 binding cleft of MDM2 [5]. NMR relaxation measurements[6] suggest that ligand binding does not alter the flexibilityaround the cleft, yet they indicate release, upon p53-derivedpeptide-binding, of the extreme N-terminal MDM2 ‘‘lid”from the binding cleft, where it is weakly bound in the apo state[7,8].

A related protein, MDM4/MDMX [9], resembles MDM2 butlacks the E3 ligase activity. The two proteins form a heterodimerthat has a higher E3 ligase activity than MDM2 alone [10]. Ele-vated MDM4 levels have been reported for a variety of tumours[10,11]. Small molecules with dual MDM2/MDM4 specificity aretherefore expected to be superior to selective MDM2 binderssuch as the nutlins, which bind 40 times more strongly toMDM2 than to MDM4 [10]. Biophysical properties of MDM4are not understood at the same level of detail as they are forMDM2, yet they might explain differences between the proteinsand aid development of ligands with dual-specificity. Althoughseveral X-ray crystallographic structures are available [12–15],most studies have been hampered by poor expression and solu-bility of MDM4. Here we describe the expression and NMR char-acterization of the MDM4 N-terminal domain in its apo and twoligand-bound states.

lsevier B.V. All rights reserved.

3036 M.C. Sanchez et al. / FEBS Letters 584 (2010) 3035–3041

2. Materials and methods

2.1. Molecular biology

A pET28b (Invitrogen) vector expressing His6 – thrombin cleav-age site – MDM4 (14-111) (C17S) was synthesized (Geneart,Regensburg, Germany). The plasmid was transformed into E. coliBL21 DE3 (Invitrogen) and protein was produced in M9 minimalmedia supplemented with 0.5% w/v 15N/13C Celtone (CambridgeIsotope Laboratories Inc.). Expression was induced at OD600 = 0.6with 1 mM isopropyl 1-thio-b-D-galactopyranoside (IPTG) at18 �C overnight. Cells were harvested by centrifugation. The frozenpellet was re-suspended in buffer A (40 mM HEPES pH 7.5, 300 mMNaCl, 1 mM TCEP, 10% v/v glycerol, 10 mM imidazole) with 1 mg/ml lysozyme and 2.5 units/ml benzonase. The suspension waspassed twice through a Basic Z 2 Plus (Constant Disruption Sys-tems Ltd.). The lysates were clarified by centrifugation and the sol-uble fraction loaded onto a Ni-NTA column (Qiagen). The proteinwas eluted with 500 mM imidazole and dialyzed against buffer Ain the presence of thrombin (GE Healthcare) followed by a re-passon the Ni-NTA column. Finally, the samples were loaded onto a gelfiltration column (HiLoad 26/60 Superdex 75 GE Healthcare) in 50mM HEPES pH 7.5, 100 mM NaCl, 2 mM TCEP, 10% v/v glycerol,0.1 mM EDTA, 0.02% NaN3.

2.2. Peptide synthesis

The p53-derived peptide 17ETFSDLWKLLP27 was synthesized bysolid-phase peptide synthesis using F-moc protection of NH2

groups. Purity was checked by HPLC and MALDI.

2.3. Thermofluor stability assay

A Thermofluor fluorescence-based stability assay was used toscreen for optimal NMR buffer conditions. For this, a mixture of0.2 mg/ml protein in the buffer of choice (Table 1), and a 1:1000dilution of Sypro orange dye (Invitrogen) were added in 25 lL ali-quots into a 96-well PCR plate (Bio-Rad). The plate was heatedusing an iCycler iQ instrument (Bio-Rad), with a scan rate of 1 �C/min from 20 to 90 �C. Data were collected at 0.2 �C increments,and analysed using GraphPad Prism 4 software.

2.4. Dynamic light scattering

Dynamic laser light scattering measurements were performedusing a Wyatt DynaPro Titan plate-reader instrument. Apo-MDM4 (14-111) protein samples (50 lL) were prepared at both50 and 100 lM in a range of buffers. Samples were filtered, centri-fuged and aliquoted into Thermo (Cat. 95040000) 384 well plates.Measurements were made using a laser power of 10%, for tenacquisitions, each of 10 s duration. Data were analysed using theDYNAMICS V6 software.

2.5. NMR spectroscopy

NMR spectra were recorded on Bruker Avance 600 and Avance2800 spectrometers at 600 and 800 MHz proton resonance fre-

Table 1Buffer conditions screened for NMR by thermofluor assay. Two salts were screened,(NH4)2SO4 and NaCl at different concentrations for each pH point.

Buffer system pH range Salt concentration

MES 4.4–6.4 10 mM, 155 mMCitrate 5.8–7.8 10 mM, 155 mMPhosphate 5.8–7.8 10 mM, 155 mMMOPS 6.2–8.2 10 mM, 155 mM

quency. Typical NMR samples contained 0.1 mM single-labeled ordouble-labeled MDM4 (14-111) in 100 mM NaCl, 50 mM BisTrispH 6.5, 2 mm TCEP, 0.1 mM EDTA, 0.1% NaN3 and 1% DMSO-d6 asa lock solvent.

Ligands were added as concentrated solutions in DMSO-d6 andany resulting precipitate was removed by centrifugation. Nutlin-3(Sigma–Aldrich) was used as a racemic mixture. The nutlin-3:MDM4 (14-111) complex was prepared at low concentrationsinitially prior to concentration and addition of excess nutlin-3.Assignment followed a standard route using TROSY-type HNCA,HNCACB, HN(CO)CACB, HNCO and HN(CA)CO experiments at800 MHz. Side-chains were partially assigned using HSQC-typeH(CC)(CO)NH-TOCSY and CC(CO)NH-TOCSY experiments at 600MHz. Triple resonance data were evaluated using CARA 1.5.5 [16].

2.6. Isothermal titration calorimetry

The interaction of MDM4 (14-111) with p53 (17ETFSDLWKLLP27)and nutlin-3 was quantified using a Microcal ITC 200 (GE Healthcare)fitted with an autosampler. The titration data were recorded at 25 �Cin 50 mM HEPES pH 7.5, 100 mM NaCl, 2 mM TCEP, 10% v/v glycerol,0.02% NaN3, 2% DMSO. Aliquots of 200 lM ligand stocks in bufferwere added to 20 lM MDM4 (14-111) in multiple 0.8 ll steps. Datawere analysed using Origin (Microcal).

3. Results and discussion

3.1. Optimization of construct and expression conditions

Two constructs were used for initial expression trials. NativeMDM4(1-134) was produced in high quantities but turned out tobe very unstable and only poorly soluble. We therefore utilized aconstruct originally designed for crystallographic work, theMDM4 (14-111) (C17S) mutant [13]. This construct was subse-quently used for all NMR and biophysical studies.

Since initial trials suggested low yields of recombinant MDM4(14-111), we undertook a systematic small-scale screen for opti-mal expression conditions. This utilized three strains of E. coliBL21 (DE3, Star and GOLD), two media (TB and M9), three temper-atures (37, 25 and 18 �C) and three IPTG induction concentrations(0.05, 0.1 and 1 mM). All possible combinations of these parame-ters were screened and compared for their yield of solubleMDM4 (14-111). Our results indicate better yields at lower tem-perature and high IPTG concentrations. They furthermore indicatepoor yield in minimal M9 medium, and the need for supplementa-tion of the media with Celtone for acceptable expression yield.

Relatively poor yield and stability of the MDM4 N-terminal do-main reported in previous studies highlighted the necessity forconstruct improvements. Although our construct had been de-signed for crystallography, it proved to be equally suitable for col-lecting NMR data. We did not, however, undertake a systematicstudy to find the optimal construct for an NMR-based study. Forexample, MDMX-S [17], a native, shortened construct, might havesuperior properties in this respect.

3.2. Optimisation of NMR conditions

Preliminary experiments indicated aggregation of MDM4 (14-111) under typical conditions used for NMR data collection (neu-tral pH, 300–310 K and low ionic strength). Self-association mightbe exacerbated by transient exposure of hydrophobic residues thatare predominantly buried within compactly folded apo MDM4 (14-111) molecules. We reasoned that in order to minimise aggrega-tion we should attempt to optimise conditions with respect tothe thermal stability of these folded form(s) of the protein. Wetherefore conducted a Thermofluor-based screen [18] that allows

Fig. 1. (a) Annotated TROSY spectrum of MDM4 (14-111) in complex with a p53-derived peptide. (b) TROSY spectra for apo MDM4 (black), MDM4–p53 complex (blue) andMDM4–nutlin-3 (red). (c) Extent of assignment for apo MDM4 (top), the p53–MDM4 complex (middle) and the MDM4–nutlin-3 complex (bottom).

M.C. Sanchez et al. / FEBS Letters 584 (2010) 3035–3041 3037

rapid assessment of a matrix of conditions derived from crystalli-zation studies [19] (Table 1). Thermofluor data suggested thatMDM4 (14-111) is most stable at high salt concentrations andpH above 6 (Supplementary data), and that ammonium sulfate ispreferable to sodium chloride with respect to the thermal stabilityof recombinant MDM4 (14-111). Dynamic light scattering studies

were used to more directly monitor the salt and pH-dependenciesof MDM4 (14-111) self-association. These studies, however, indi-cated a higher degree of protein aggregation at 20 �C in ammoniumsulfate buffer compared to equivalent sodium chloride buffer con-ditions (Supplementary data). All subsequent NMR experimentswere therefore carried out in 100 mM NaCl; pH 6.5 was used, as

Fig. 2. Chemical shift perturbations induced by binding of p53 peptide to MDM4 (14-111). (a) Chemical shift perturbations for amide (lower panel) and carbon (upper panel)nuclei. Asterisks denote amide signals not detectable for apo MDM4. (b) Mapping of chemical shift perturbations on the three-dimensional structure (pdb code 3FE7 [13]) ofMDM4 N-terminal domain. Unassigned residues in the apo protein are marked in white while chemical shift perturbations are mapped in red. Chemical shift perturbationsare geometric sums of all available chemical shift differences (Ca, Cb, CO, N and H) weighed by relative gyromagnetic ratios of nuclei.

3038 M.C. Sanchez et al. / FEBS Letters 584 (2010) 3035–3041

several peaks disappear at higher pH values. Preliminary studiesindicated slightly better sensitivity for TROSY-based than forHSQC-based experiments. This was ascribed to residual aggrega-tion and prompted us to limit concentrations to 100 lM. The con-sequent low protein concentration and high salt contentnecessitated recording of all assignment experiments with a cryo-probe and using a salt-tolerant sample tube.

3.3. NMR assignment

Only 66 out of 93 expected amide cross peaks were detected inthe resulting 1H/15N correlation spectra, despite efforts to optimiseconditions. Consequently it was impossible to fully assign nuclei ofthe apo state using a conventional assignment strategy. Addition ofthe p53-derived peptide to the labeled MDM4 (14-111) sample led

Fig. 3. Chemical shift perturbations induced by binding of nutlin-3 to MDM4. Cf. figure caption for Fig. 2 for details.

M.C. Sanchez et al. / FEBS Letters 584 (2010) 3035–3041 3039

to large changes in chemical shifts, consistent with the anticipatedbinding of peptide to protein. Additional peaks became detectablein NMR spectra of this complex, yielding all 93 expected crosspeaks. We observed similar effects upon formation of the MDM4(14-111)–nutlin-3 complex, although fewer peaks emerged (yield-ing a total of 80 cross peaks) and chemical shift perturbations weregenerally smaller. A set of NMR spectra collected on the complexwith peptide, which showed by far the best signal-to-noise in the

3D correlation experiments, enabled a complete assignment ofthe MDM4 (14-111) backbone Ca, Cb, NH, HN and CO resonancesin the MDM4–p53 peptide complex (Fig. 1a).

The assignments obtained for the MDM4 (14-111)–peptidecomplex were used to assist assignment of the apo and nutlin-3bound forms of MDM4 (14-111). Direct assignment transferencein the 2D TROSY spectrum proved to be difficult due to large chem-ical shift differences and missing peaks, as discussed above. There-

3040 M.C. Sanchez et al. / FEBS Letters 584 (2010) 3035–3041

fore, the 13C chemical shift information in 3D triple resonanceexperiments was used to infer assignments of the majority ofpeaks in the 2D TROSY spectrum. We simultaneously obtained achemical shift perturbation map of Ca, NH, HN, CO and Cb chemicalshifts, as well as identifying the missing peaks. Chemical shiftshave been uploaded to the BMRB data base with accession codes16893 (apo MDM4), 16900 (MDM4–p53 complex) and 16894(MDM4–nutlin-3 complex).

3.4. Chemical shift mapping of ligands on MDM4

Lack of amide signals is often caused by conformational transi-tions on the millisecond time scale and concomitant line broaden-ing in the intermediate-exchange regime. Indeed very broadresidual peaks were found in TROSY spectra of the apo MDM4(14-111) at positions corresponding to several cross peaks thatemerged upon complexation of MDM4 (14-111) with peptide. Thissuggests MDM4 (14-111) is a flexible protein in the apo state thatis rigidified by binding to a fragment of its natural ligand. Nutlin-3appears to be less well able to rigidify MDM4 (14-111) based onthe presence of only 13 new cross peaks in TROSY spectra of itscomplex with MDM4 (14-111).

One obvious difference between MDM2 and MDM4 that simpli-fies interpretation of chemical shift perturbation is the lack of a lidin MDM4 [20]. Therefore in MDM4 no effects on chemical shiftsdue to release of the N-terminus, which have been described forthe equivalent MDM2 complex [6], need be invoked upon bindingof peptide. Mapping onto its 3D structure the chemical shift pertur-bations induced by binding of peptide or nutlin-3 to MDM4 (14-111)highlights specific regions on the protein surface (Figs. 2 and 3).

Notably, perturbations of amide chemical shifts probe differentphenomena than changes in carbon chemical shifts (for Ca and CO).Perturbations of amide resonances arise principally from directinteractions between the ligand and the protein [21] and thereforecan be used to infer the location of the binding site. As expected thestrongest effects on amide peaks correspond to residues lining thehydrophobic cleft. Perturbations of carbon shifts, on the otherhand, arise primarily from modulations in secondary structureand therefore indicate changes in structure and backbone dynam-ics [22]. We observed large carbon chemical shift differences forresidues in helices a2 and a20 that form the peptide-binding cleft.Furthermore, signals for the N-terminal end of helix a10 and anadjacent loop, which are located at the opposite wall of the hydro-phobic cleft, are detectable only in ligand-bound MDM4 (14-111).These observations suggest that this region is less structured in theapo protein than it is in either of the complexes. In summary, boththe sharper (i.e. more easily detected) cross peaks in the spectra ofthe complex, and the locations within the protein of residues expe-riencing carbon chemical shift perturbations upon ligand binding,suggest MDM4 (14-111) is relatively flexible in its free state butless so when complexed with ligands. P53 binds via either one ora combination of the following: an induced fit mechanism or selec-tion and stabilization of one out of multiple conformations. Wehave insufficient data to distinguish between these mechanisms.

Flexibility of apo-MDM4 N-termini might also explain the pooryield and aggregation tendency of MDM4 observed by us and othergroups. In general the effects on the NMR spectra resulting fromcomplexation of MDM4 (14-111) with nutlin-3 are less significantthan those observed for p53 (Fig. 3). This suggests that nutlin-3 isnot capable of inducing an equivalent degree of conformationaladjustment and rigidification. This is not unexpected for a smaller,less flexible molecule that lacks the adaptability of the largerpeptide. Furthermore, previous structural studies [12] indicated thatsteric hindrance between nutlin-3 and amino acid side chains (ofM53 and Y99) would be likely to prevent nutlin-3 from binding toMDM4 (14-111) in the same mode as it does to MDM2 [15].

MDM2, by contrast, appears to be more structured than MDM4in its apo form. This conclusion is based on the number of crosspeaks observable in spectra of the former and the extent of assign-ment that was possible for apo MDM2 [8]. Excluding release of theN-terminal lid, ligands have less impact on dynamics [6,8] or car-bon shifts [6] of MDM2 than they do on MDM4. However, somewidening of the cleft was observed when comparing the structuresof apo and peptide-bound MDM2 [8], with the helices a2 and a20

acting as hinges. It is reasonable to expect similar structural rear-rangements in MDM4, and local unfolding in the hinges mightfacilitate the necessary movements. Isothermal titration calorime-try studies on MDM2 [14] and MDM4 found large unfavorableentropic contributions for both proteins, which could be explainedby reduced conformational mobility in both cases. Recent crystal-lographic studies of small-molecule ligand binding to MDM4 [23]also suggest an induced fit mechanism and large structuralchanges within the a20 helix. For development of dual-specificityMDM2/MDM4 binders it will be important to understand subtledifferences in the dynamic behavior of both proteins and relativecontributions of induced fit or conformational selection in eachcase. More detailed NMR relaxation studies are therefore neededto fully understand MDM4 dynamics and how it is influenced byligand binding.

Acknowledgements

We would like to acknowledge the Centre for Translational andChemical Biology supported by Welcome Trust, SULSA and BBSRC.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.febslet.2010.05.058.

References

[1] Brown, C.J., Lain, S., Verma, C.S., Fersht, A.R. and Lane, D.P. (2009) Awakeningguardian angels: drugging the p53 pathway. Nat. Rev. Cancer 9, 862–873.

[2] Harris, S.L. and Levine, A.J. (2005) The p53 pathway: positive and negativefeedback loops. Oncogene 24, 2899–2908.

[3] Kussie, P.H., Gorina, S., Marechal, V., Elenbaas, B., Moreau, J. and Levine, A.J.(1996) Structure of the MDM2 oncoprotein bound to the p53 tumorsuppressor transactivation domain. Science 274, 948–953.

[4] Vassilev, L.T. (2005) MDM2 inhibitors for cancer therapy. Trends Mol. Med. 13,23–31.

[5] Vassilev, L.T., Vu, B.T., Graves, B., Carvajal, D., Podlaski, F., Filipovic, Z., Kong, N.,Kammlott, U., Lukacs, C., Klein, C., Fotouhi, N. and nad Liu, E.A. (2004) In vivoactivation of the p53 pathway by small-molecule antagonists of MDM2.Science 303, 844–848.

[6] Showalter, S.A., Bruschweiler-Li, L., Johnson, E., Zhang, F. andBrüschweiler., R. (2008) Quantitative lid dynamics of MDM2 revealsdifferential ligand binding modes of the p53-binding cleft. J. Am. Chem.Soc. 130, 6472–6478.

[7] McCoy, M.A., Gesell, J.J., Senior, M.M. and Wyss, D.F. (2003) Flexible lid to thep53-binding domain of human MDM2: implications for p53 regulation. PNAS100, 1645–1648.

[8] Uhrinova, S., Uhrin, D., Powers, H., Watt, K., Zhelava, D., Fischer, P., McInnes, C.and Barlow, P.N. (2005) Structure of free MDM2 N-terminal domain revealsconformational adjustments that accompany p53-binding. J. Mol. Biol. 350,587–598.

[9] Shvarts, A., Steegenga, W.T., Riteco, N., van Laar, T., Dekker, P., Bazuine, M., vanHam, R.C., van Oordt, W., Hateboer, G., van der Eb, A.J. and Jochemsen, A.G.(1996) MDMX: a novel p53-binding protein with some functional propertiesof MDM2. EMBO J. 15, 5349–5357.

[10] Wade, M. and Wahl, G.M. (2009) Targeting MDM2 and MDMX in cancertherapy: Better living through medicinal chemistry? Mol. Cancer Res. 7, 1–11.

[11] Marine, J.-C.W., Dyer, M.A. and Jochemsen, A.G. (2007) MDMX: from bench tobedside. J. Cell Sci. 120, 371–378.

[12] Popowicz, G.M., Czarna, A., Rothweiler, U., Szwagierczak, A., Krajewski, M.,Weber, L. and Holak, T.A. (2007) Molecular basis for the inhibition of p53 byMDMx. Cell Cycle 6, 2386–2392.

[13] Kallen, J., Goepfert, A., Blechschmidt, A., Izaac, A., Geiser, M., Tavares, G.,Ramage, P., Furet, P., Masya, K. and Lisztwan, J. (2009) Crystal structure ofhuman MDMX (HdmX) in complex with p53 peptide-analogues revealsurprising conformational changes. J. Biol. Chem. 284, 8812–8821.

M.C. Sanchez et al. / FEBS Letters 584 (2010) 3035–3041 3041

[14] Pazgier, M., Liu, M., Zou, G., Li, C., Li, C., Li, J., Monbo, J., Zella, D., Tarasov, S.G.and Lu, W. (2009) Structural basis for high-affinity peptide inhibition of p53interactions with MDM2 and MDMX. PNAS 106, 4665–4670.

[15] Phan, J., Li, Z., Kasprzak, A., Li, B., Sebti, S., Guida, W., Schönbrunn, E. andChen, J. (2010) Structure-based design of high affinity peptides inhibiting theinteraction of p53 with MDM2 and MDMX. J. Biol. Chem. 285, 2174–2183.

[16] Keller R. (2004) Optimizing the process of nuclear magnetic resonancespectrum analysis aided resonance assignment. Diss ETH No 15947.

[17] Rallapalli, R., Strachan, G., Cho, B., Mercer, W.E. and Hall, D.J. (1999) A novelMDMX transcript expressed in a variety of transformed cell lines encodes atruncated protein with potent p53 repressive activity. J. Biol. Chem. 274,8299–8308.

[18] Ericsson, U.B., Hallberg, B.M., DeTitta, G.T., Dekker, N. and Nordlund, P. (2006)Thermofluor-based high-throughput stability optimization of proteins forstructural studies. Anal. Biochem. 357, 289–298.

[19] Segelke, B.W. (2001) Efficiency analysis of sampling protocols used in proteincrystallization screening. J. Cryst. Growth 232, 553–562.

[20] McCoy, M.A., Gesell, J.J., Senior, M.M. and Wyss, D.F. (2003) Flexible lid to thep53-binding domain of human Mdm2: implications for p53 regulation. PNAS100, 1645–1648.

[21] Medek, A., Hajduk, P.J., Mack, J. and Fesik, S.W. (2000) The use of differentialchemical shifts for determining the binding site location and orientation ofprotein-bound ligands. J. Am. Chem. Soc. 122, 1241–1242.

[22] Berjanskii, M.V. and Wishart, D.S. (2005) A simple method to predict proteinflexibility using secondary chemical shifts. J. Am. Chem. Soc. 127, 14970–14971.

[23] Popowicz, G., Czarna, A., Wolf, S., Wang, K., Wang, W, Dömling, A. and Holak,T.A (2010) Structures of low molecular weight inhibitors bound to MDMX andMDM2 reveal new approaches for p53–MDMX/MDM2 antagonist drugdiscovery. Cell Cycle 9, 1104–1111.