Structural Basis for High-Affinity Peptide Inhibition of P53 Interactions with MDM2 and MDMX

Structural basis for high-affinity peptide inhibition of p53 interactions with MDM2 and MDMX

Marzena Pazgiera,1, Min Liua,b,1, Guozhang Zoua, Weirong Yuana, Changqing Lia, Chong Lia, Jing Lia, Juahdi Monboa, Davide Zellaa, Sergey G. Tarasovc, and Wuyuan Lua,2

aInstitute of Human Virology, University of Maryland School of Medicine, 725 West Lombard Street, Baltimore, MD 21201; bThe First Afﬁliated Hospital, School of Medicine, Xi’an Jiaotong University, Shaanxi Province 710061, China; and cStructural Biophysics Laboratory, National Cancer Institute at Frederick, Frederick, MD 21702

Communicated by Robert C. Gallo, University of Maryland, Baltimore, MD, January 28, 2009 (received for review September 29, 2008) The oncoproteins MDM2 and MDMX negatively regulate the ac- ligase activity (10). Structurally related to MDM2, MDMX of tivity and stability of the tumor suppressor protein p53—a cellular 490-aa residues possesses domain structures arranged similarly process initiated by MDM2 and/or MDMX binding to the N- to MDM2, except that MDMX lacks ubiquitin-ligase function terminal transactivation domain of p53. MDM2 and MDMX in many (11, 12). Growing evidence supports that in unstressed cells tumors confer p53 inactivation and tumor survival, and are impor- MDM2 primarily controls p53 stability through ubiquitylation to tant molecular targets for anticancer therapy. We screened a target the tumor suppressor protein for constitutive degradation duodecimal peptide phage library against site-specifically biotin- by the proteasome (13, 14), whereas MDMX mainly functions as ylated p53-binding domains of human MDM2 and MDMX chemi- a significant p53 transcriptional antagonist independently of cally synthesized via native chemical ligation, and identified sev- MDM2 (15, 16). Under stress conditions, MDM2 and MDMX eral peptide inhibitors of the p53-MDM2/MDMX interactions. The cooperate to activate p53 through mechanisms involving both most potent inhibitor (TSFAEYWNLLSP), termed PMI, bound to MDM2 autodegradation (autoubiquitylation) and MDM2- MDM2 and MDMX at low nanomolar affinities—approximately 2 depedent degradation of MDMX (17–20). orders of magnitude stronger than the wild-type p53 peptide of In many tumors, p53 is present in its wild-type form. The the same length (ETFSDLWKLLPE). We solved the crystal structures presence of wild-type p53 strongly correlates to amplification of synthetic MDM2 and MDMX, both in complex with PMI, at 1.6 and/or over-expression of MDM2/MDMX, resulting directly in Å resolution. Comparative structural analysis identified an exten- p53 suppression and malignant progression (8, 9). Inhibition of sive, tightened intramolecular H-bonding network in bound PMI the p53-MDM2 interactions by MDM2 antagonists has been that contributed to its conformational stability, thus enhanced shown both in vitro and in vivo to reactivate the p53 pathway and binding to the 2 oncogenic proteins. Importantly, the C-terminal selectively kill tumor cells in a p53-dependent manner. Acting residue Pro of PMI induced formation of a hydrophobic cleft in synergistically in tumor cells, MDM2 and MDMX have become MDMX previously unseen in the structures of p53-bound MDM2 or 2 of the most attractive molecular targets for anticancer therapy. MDMX. Our findings deciphered the structural basis for high- Toward this end, much of the current efforts have been focused affinity peptide inhibition of p53 interactions with MDM2 and on combinatorial library search for and structure-based rational MDMX, shedding new light on structure-based rational design of design of low molecular weight inhibitors that target the N- different classes of p53 activators for potential therapeutic use. terminal p53-binding domains of MDM2 and MDMX (21). Successful examples include, but are not limited to, cis- 53 is best known as a tumor suppressor that transcriptionally imidazoline analogs termed Nutlins and, more recently, a spiro- pregulates, in response to cellular stresses such as DNA oxindole-derived compound termed MI-219 (22, 23). damage or oncogene activation, the expression of various target Peptides, because of their large interacting surfaces, offer the genes that mediate cell-cycle arrest, DNA repair, senescence or prospect of enhanced potency, high specificity and low toxicity. apoptosis—all of these cellular responses are designed to prevent However, most of the peptidic and peptidomimetic inhibitors damaged cells from proliferating and passing mutations on to the examined to date bind MDM2 at affinities ranging from high next generation (1–3). In 50% of human cancers, p53 is defective nanomolar to low micromolar concentrations, and none is nearly due usually to somatic mutations or deletions primarily in its as effective as Nutlins and MI-219 in tumor killing in vitro (21). DNA-binding domain and, to a lesser extent, to posttranslational Further, because the structural basis for MDMX inhibition is modifications such as phosphorylation, acetylation and methyl- much less understood than that for MDM2 inhibition, antago- ation that affect p53 function and stability. Altered p53 fails to nists designed for MDM2 are, in general, significantly less regulate growth arrest and cell death upon DNA damage, inhibitory toward MDMX. Potent peptide inhibitors against BIOCHEMISTRY directly contributing to tumor development, malignant progres- MDM2 and/or MDMX are needed as important cellular probes sion, poor prognosis and resistance to treatment (4). Conversely, of the p53 pathway in cancer biology and as useful templates for restoring endogenous p53 activity can halt the growth of can- structure-based rational design of different classes of p53 acti- cerous tumors in vivo by inducing apoptosis, senescence, and vators for potential therapeutic use. Here, we report identifica- innate inflammatory responses (5–7). tion and functional and structural characterizations of a high- As p53 mediates growth arrest and apoptosis, it is essential to keep its activity in check during normal development (2). Author contributions: M.P., M.L., and W.L. designed research; M.P., M.L., G.Z., W.Y., Multiple mechanisms exist to negatively regulate p53 activity, Changqing Li, Chong Li, J.L., J.M., D.Z., and S.G.T. performed research; M.P., D.Z., S.G.T., and among which the E3 ubiquitin ligase MDM2 and its homolog W.L. analyzed data; and M.P. and W.L. wrote the paper. MDMX (also known as MDM4) play a central regulatory role in The authors declare no conflict of interest. the developing embryo and in mature differentiated cells (8, 9). Data deposition: The atomic coordinates have been deposited in the Protein Data Bank, MDM2 consists of 491-aa residues, comprising an N-terminal www.pdb.org (PDB ID codes 3EQS and 3EQY). p53-binding domain, a central domain preceded by nuclear 1M.P. and M.L. contributed equally to this work. export and localization signals essential for nuclear-cytoplasmic 2To whom correspondence should be addressed. E-mail: [email protected]. trafficking of MDM2, a zinc finger domain, and a C-terminal This article contains supporting information online at www.pnas.org/cgi/content/full/ zinc-dependent RING finger domain that confers E3 ubiquitin 0900947106/DCSupplemental.

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900947106 PNAS ͉ March 24, 2009 ͉ vol. 106 ͉ no. 12 ͉ 4665–4670 Downloaded by guest on September 30, 2021 Fig. 1. Quantiﬁcation of the interactions of synMDM2 (50 nM) and synMDMX (100 nM) with varying concentrations of PMI, pDI, (15–29)p53, (17–28)p53, or Nutlin-3 by SPR-based competition assays. Each curve is the mean of 4 independent measurements at 25 °C in 10 mM Hepes, 150 mM NaCl, 0.005% surfactant P20, pH 7.4. Nutlin-3 was too weak for synMDMX and is not reported. Solubility of both Nutlin-3 and pDI decreased at the highest concentrations used, attributing to an upward curvature of the inhibition curves for synMDM2-Nutlin-3 and synMDMX-pDI.

affinity peptide inhibitor, termed PMI (p53-MDM2/MDMX regression analyses yielded for PMI a Kd value of 3.4 nM for inhibitor), of p53 interactions with both MDM2 and MDMX. synMDM2 and of 4.2 nM for synMDMX. By contrast, (15–29)p53 bound to synMDM2 and synMDMX at affinities of 140 nM and Results 270 nM, respectively. Although Nutlin-3 showed little binding to PMI—a Potent Inhibitor of the p53-MDM2/MDMX Interactions Se- MDMX, its Kd value of 263 nM for MDM2 was largely in line lected from a Phage Displayed Peptide Library. The p53-binding with the published result (IC50: 90 nM for Nutlin-3a and 13.6 ␮M domains of MDM2 and MDMX (25–109MDM2 and for Nutlin-3b) (23). Importantly, compared with (17–28)p53, PMI 24–108MDMX, referred to thereafter as synMDM2 and bound to synMDM2 135-fold stronger and to synMDMX 156-fold synMDMX) and their site-specifically biotinylated forms were tighter. Our data suggest that PMI, with low nanomolar affinities chemically synthesized using native chemical ligation (SI Text). for both MDM2 and MDMX, is one of the strongest peptidic Using biotin-synMDM2 and biotin-synMDMX as bait, we inhibitors of the p53-MDM2/MDMX interactions ever reported. screened a duodecimal peptide library displayed on M13 phage. Shown in Fig. S1 are the amino acid sequences from 15 binding PMI Binds to MDM2 and MDMX in a Canonical Mode Previously clones obtained after 4 rounds of selection. Two consensus Described for p53 Peptides. The overall structures of human sequences emerged for both MDM2 and MDMX: LTFEHY- synMDM2 and synMDMX are similar, as evidenced by a root WAQLTS, also termed pDI (24), and PMI–TSFAEYWNLLSP. mean square deviation (rmsd) of 1.2 Å between superimposed The 3 most critical residues involved in p53-MDM2/MDMX C␣ atoms. synMDM2 and synMDMX share the basic structural recognition, i.e., Phe-19, Trp-23 and Leu-26 (p53 numbering), elements and conserved global fold reported for human MDM2 were all present in the phage-selected consensus sequences. pDI (23, 25–28) and zebra fish MDMX (29). Both molecules are was recently identified from the same Ph.D.-12TM phage library, characterized by structural repetition of 2 assemblies of a ␤␣␤␣␤ using GST-tagged recombinant MDM2 and MDMX immobi- topology, which are related by an approximate dyad axis of lized on glutathione-agarose beads (24). However, because of pseudosymmetry (Fig. 2). Superposition of synMDM2 and low solubility and relatively weak activity of pDI, this work synMDMX points toward a large structural variation in the focused only on PMI, which is highly soluble in aqueous solution C-terminal half of the molecules, whereas the N-terminal halves and of higher affinity for MDM2 and MDMX than pDI. Two p53 are nearly identical. Except for the loop regions, the largest peptides, (15–29)p53 (SQETFSDLWKLLPEN) and (17–28)p53 conformational changes are located in the ␣2Ј helix, stemming (ETFSDLWKLLPE), were used for comparison. from a change of His-96 in MDM2 to Pro-95 in MDMX. We quantified direct interactions between synMDM2/ synMDMX and the 3 peptide inhibitors, using isothermal titra- tion calorimetry (ITC), and the results are tabulated in Table S1 (see Fig. S2 for additional data). PMI bound to synMDM2 and syn MDMX with Kd values of 3.3 and 8.9 nM, respectively, Ϸ20-fold stronger than (15–29)p53 for either protein. Compared with that of (17–28)p53—the wild type sequence of the same length, the binding affinity of PMI increased by a factor of 89 for MDM2 and of 43 for MDMX. In both cases, favorable enthalpy changes overcame unfavorable entropy changes, contributing to a dramatically enhanced binding of PMI to synMDM2 and synMDMX. To verify the ITC results, we devised a surface plasmon resonance (SPR)-based competition assay, in which (15–29)p53 was immobilized on a CM5 sensor chip for kinetic analysis of a fixed concentration of synMDM2 (50 nM) or synMDMX (100 nM) preincubated with varying concentrations of PMI, (15–29)p53, or (17–28)p53. Nutlin-3—a racemic mixture of Nutlin-3a and Nut- lin-3b (23), was used as a control. As shown in Fig. 1, all 4 inhibitors competed, in a dose-dependent manner, with immo- Fig. 2. Stereoview of superimposed structures of synMDM2-PMI (orange/ bilized (15–29)p53 for synMDM2 or synMDMX binding. Nonlinear green) and synMDMX-PMI (blue/yellow) in a ribbon diagram.

4666 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900947106 Pazgier et al. Downloaded by guest on September 30, 2021 Interestingly, a crystallographic 2-fold axis generates a dimer of the PMI-synMDM2 complex (Fig. S3), although no evidence for dimerization has been found using native gel electrophoresis and size exclusion chromatography. The N-terminal transactivation domain of p53 encompasses T18F19S20D21L22W23K24L25L26 minimally required for effective MDM2 binding (30, 31). The side chains of Phe-19, Trp-23 and Leu-26, involved in p53 transactivation (32–34), dock, in an amphipathic ␣-helix, inside a hydrophobic cavity of the oncoprotein (25). Not surprisingly, PMI retains the functionally conserved hydrophobic triad, Phe-3/Trp-7/Leu-10. Structural analysis indicates that the Phe-3-binding pockets are nearly identical in MDM2 and in MDMX. However, the Trp-7-binding pockets slightly differ in geometry (Leu-54 and Ile-99 in MDM2 versus Met-53 and Leu-98 in MDMX). Nevertheless, the Phe/Trp dyads of PMI and p53 appear indistinguishable in MDM2/MDMX binding, as indicated by calculations of buried surface area (BSA) and binding energy (Table S2). In contrast, the Leu-10- binding pocket in MDM2 (lined by Leu-54, Val-93, His-96, Ile-99, Tyr-100) differs from that in MDMX (lined by Met-53, Val-92, Pro-95, Leu-98, Tyr-99), mainly because of the shift of the ␣2Ј helix caused by the His-96 to Pro-95 change. Because Leu-10 of PMI is buried in the peptide-protein complexes to a different extent from Leu-26 of p53, the energetic contribution of that position may be context-dependent. In addition to the hydrophobic interactions, 3 intermolecular H-bonds contribute to PMI binding to MDM2 and MDMX. The Fig. 3. Structures of PMI and p53 in bound state. (A) PMI bound to MDM2 same number of intermolecular H-bonds is also present in the (Left, green) and to MDMX (Right, yellow). In PMI, Ser-2 to Ala-4 make 3 regular (i, i ϩ 4) ϾCAO⅐⅐⅐HNϽ H-bonds in the ␣-helix. Ser-2 and residues from p53-MDM2 complex between Phe-19 N, Trp-23 N␧1, and Asn-29 ␧1 ␩ Tyr-6 to Asn-8 (in the MDM2 complex) or Tyr-6 to Trp-7 (in the MDMX complex) OXT of the peptide and Gln-72 O , Leu-54 O, and Tyr-100 O also make 4 or 3 (i, i ϩ 3) ϾCAO⅐⅐⅐HNϽ H-bonds. In addition, Ser-2 also forms of the protein, respectively (25). However, the H-bonding pat- 2 main chain-side chain, and 1 side chain-side chain H-bonds with Glu-5. syn terns involving Tyr-100 in PMI- MDM2 and Tyr-99 in PMI- However, only the (energetically signiﬁcant) H-bonds unique to PMI are ␩ synMDMX differ. Tyr-100 O donates an H-bond to Leu-10 O of shown in dashes. (B) Superposition of PMI (green) and p53 bound to MDM2 PMI, whereas Tyr-99 O␩ forms an H-bond with Ser-11 N of PMI, (Left), and of PMI (yellow) and p53 bound to MDMX (Right). Residues 17–29 reflecting structural differences in the C-terminal region of PMI of p53 (gray) bound to human MDM2 (PDB entry 1YCR) and 17–27 of p53 and in the vicinity of Tyr-100/Tyr-99 between different protein- (pink) bound to zebra ﬁsh MDMX (PDB entry 2Z5T) are shown. peptide complexes. 24 and 36, Tyr is strongly selected by phage display over a Leu PMI Differs from p53. A secondary-structure analysis, using the residue at the same position. This selection probably can be Kabsch–Sander algorithm (35), reveals that the ␣-helix of PMI rationalized by the tyrosyl side chain capable of making addi- is more extended than that of p53. In PMI, the regular ␣-helix tional hydrophobic, ␲-cation, and electrostatic interactions with starts at Phe-3 and ends at Asn-8 (6 residues), followed by a ␲ helical turn comprising Leu-9 and Leu-10 in MDM2 and only residues inside the PMI-binding pocket. Tyr-6 forms cation- Leu-9 in MDMX. By contrast, the regular ␣-helix in p53, starting interactions with Lys-94 of MDM2 or possibly Lys-93 of MDMX. at Phe-19 and ending at Trp-23, is shortened by 1 residue. Two Further, the buried surface area of Tyr-6 of PMI is significantly (i, i ϩ 3) ϾCAO⅐⅐⅐HNϽ H-bonds involving Trp-7-Leu-10 (3.0 Å) larger than that of Leu-22 of p53 (Table S2). Finally, the hydroxyl and Asn-8-Ser-11 (3.3 Å) in PMI-synMDM2, or the Trp-7-Leu-10 group of Tyr-6 participates in an elaborate, water-mediated H-bond (3.1 Å) in PMI-synMDMX, are missing in the helical turn H-bonding network comprising the side chain(s) of Gln-72 and region of p53 (Fig. 3). Further, Ser-2 of PMI participates in a Lys-94 of MDM2 or Gln-71 of MDMX (Fig. S4). more extensive and stronger H-bonding network than does synMDM2-PMI Differs from synMDMX-PMI. The most profound struc-

Thr-18 of p53. In addition to Ser-2 O-Glu-5 N and Ser-2 O-Tyr-6 BIOCHEMISTRY N—two backbone H-bonds also found for Thr-18 of p53, PMI tural difference between the 2 complexes centers on the 2 possesses 2 additional main chain-side chain H-bonds, i.e., Ser-2 C-terminal residues (Ser-11 and Pro-12) of PMI in the vicinity syn N-Glu-5 O␧1 (2.8 Å in MDM2 and 2.9 Å in MDMX) and Ser-2 of Tyr-100/Tyr-99 (Fig. 4). In the MDM2-PMI complex, O␥-Glu-5 N (3.1 Å in both MDM2 and MDMX) (Fig. 3). Pro-12 of PMI is fully disordered (refer to the electron density Optimally aligned, these two H-bonds likely contribute to the map, Fig. S5), and the side chain of Ser-11 does not contribute conformational stability of PMI in the MDM2/MDMX complex. to MDM2 binding. This finding is consistent with the previous By contrast, the topologically equivalent main chain-side chain observation that C-terminal residues flanking Leu-26 of p53 H-bonds involving Thr-18 of p53, i.e., Thr-18 N-Asp-21 O␦2 (3.5 (equivalent to Leu-10 of PMI) do not make direct contact with Å) and Thr-18 O␥1-Asp-21 N (3.6 Å) (25), appear too long to be MDM2 (25). The aromatic ring of a protruding Tyr-100, H- energetically significant. It is worth noting that a side chain-side bonded to the carbonyl O of Leu-10 of PMI, appears to ‘‘squeeze chain H-bond was thought to exist between Thr-18 O␥1 and out’’ Ser-11 and Pro-12 to point away from the protein. By Asp-21 O␦2 as part of the Thr-18-Asp-21 H-bonding network contrast, structurally ordered Ser-11 and Pro-12 of PMI fit important for p53 stability (25). Similar interactions also exist snugly in the synMDMX-PMI complex, where Tyr-99 recesses to between Ser-2 O␥ and Glu-5 O␧1 in PMI-MDM2 and PMI- form an H-bond with Ser-11 N of the peptide. MDMX. The ␣2Ј helix of MDMX moves outward in relation to the ␣2Ј Tyr-6 of PMI or Leu-22 of p53 makes van der Waals contacts helix of MDM2, coinciding with the C␣ atoms of His-96 of MDM2 with Val-93 of MDM2 or Val-92 of MDMX. As reported in refs. and Pro-95 of MDMX moving apart by 2.4 Å. This change

Pazgier et al. PNAS ͉ March 24, 2009 ͉ vol. 106 ͉ no. 12 ͉ 4667 Downloaded by guest on September 30, 2021 Fig. 5. Conformational changes of Tyr-100/Tyr-99 seen in different peptide- protein complexes: PMI-synMDM2, green; p53-MDM2, gray (PDB entry 1YCR); PMI-synMDMX, yellow; and p53-MDMX, pink (PDB entry 1Z5T).

Fig. 4. Structural differences in PMI binding between MDM2 and MDMX. (A) Electrostatic potential distribution (negative in red, positive in blue, and PMI binding to the oncoproteins. Central to the N-terminal H- neutral in white) at the molecular surfaces of MDM2 (Left) and MDMX (Right). bonding network in PMI is Ser-2, which donates and accepts up to PMI is shown in a ribbon and stick diagram. (B) Close-up view of the binding 5 H-bonds as judged by their geometry. Not surprisingly, mutation pockets for Leu-10 and Pro-12 of PMI, where only the residues lining the binding pockets are shown in sticks. The side chain of Tyr-100 of MDM2 is of Ser-2 to Ala reduced the binding affinity of PMI for MDM2 by H-bonded to Leu-10 O of PMI (Left), whereas Tyr-99 O␩ of MDMX forms an 1 order of magnitude, underscoring the importance of intramolec- H-bond with Ser-11 N of PMI (Right). ular hydrogen bonding in PMI-MDM2/MDMX interactions. It has been shown that Thr-18, highly conserved in p53 and important for MDM2 binding, can be substituted only by Ser (31). Consistent with propagates to Tyr-99 of MDMX. A resultant conformational the structural findings, the results from ITC measurements showed adjustment surrounding Tyr-99, characterized by a shift of its C␣ that the free energy of association gained by PMI over (17–28)p53 for atom by 1.5 Å, creates for Pro-12 a new binding cleft lined by Val-49, MDM2/MDMX came entirely from a favorable enthalpy change Met-53, Tyr-99 and Leu-102 (a BSA of Pro-12: 99 Å2). Notably, counteracted by an unfavorable entropy change. A greater entropy Leu-10 of PMI is largely shielded from the bulk solvent by His-96 loss for PMI (Table S1) was indicative of a more stable peptide and Tyr-100 of MDM2 (a BSA of Leu-10: 122 Å2). However, it conformation in the complex than (17–28)p53. becomes less buried in the complex with MDMX (a BSA of Leu-10: An enhanced ␣-helicity of PMI and the tightening of its intramo- 84 Å2) as a result of the significant conformational change in the ␣2Ј lecular H-bonding network seen in the MDM2 and MDMX helix region. Pro-12 interacting with its newly formed hydrophobic complexes likely result, at least in part, from the selection of Ser-11 cleft in synMDMX-PMI likely compensates, at least in part, for the (as opposed to Pro-27 in p53). Zondlo et al. recently showed that lost binding energy by a partially exposed Leu-10. mutation of the highly conserved Pro-27 in (12–30)p53 to Ser increased its binding affinity for MDM2 by 50-fold (a decrease in Discussion Kd from 229 to 4.7 nM) (40). The authors attributed the improve- Inhibition of p53 interactions with the oncogenic proteins MDM2 ment in MDM2 binding to increased ␣-helicity of the peptide in the and MDMX is of important therapeutic value in cancer treatment. complex—a thesis largely confirmed in a molecular dynamic sim- However, potent inhibitors active at low nanomolar concentrations ulation study by Dastidar et al. (41). against both MDM2 and MDMX are nonexistent. Using phage One of the most interesting structural findings entails the C- display coupled with chemical protein synthesis via native chemical terminal residue Pro-12 of PMI interacting with a hydrophobic cleft ligation (37, 38), we identified PMI—the strongest peptide ligand unique to MDMX. Pro-12 is the third most buried residue at the ever reported for MDM2/MDMX. Further, high-resolution crystal PMI-MDMX interface with the second highest solvation energy structures of synMDM2 and synMDMX in complex with PMI have effect (Table S2). By contrast, residues flanking Leu-26 of p53 do been determined, unveiling not only the structural differences not make energetically meaningful contact with either human between MDM2 and MDMX but also the molecular determinants MDM2 or zebra fish MDMX (25, 29). In fact, deletion of Ser-11 and for high-affinity peptide inhibition of the 2 oncogenic proteins. Pro-12 in PMI, while exhibiting little effect on MDM2 binding, PMI binds MDM2 and MDMX similarly to p53 peptides. How- weakened MDMX binding by 1 order of magnitude (Kd jumped ever, the improvement in binding affinity of PMI over (17–28)p53 by from 3.6 to 29 nM), indicative of the importance of Pro-12 in 2 orders of magnitude was surprising, partly because both peptides binding MDMX but not MDM2. As shown in the known structures contained the same hydrophobic triad, Phe/Trp/Leu, known to be of MDM2 and zebra fish MDMX (23, 25–29), the geometry of the most critical for p53 transactivation and for MDM2 interactions binding pocket for the C-terminal residues of p53 hinges on the (32–34, 39). Structural analyses of the PMI complexes and of conformation of Tyr-100 of MDM2 or Tyr-99 of MDMX. The p53-bound MDM2/MDMX structures suggest that an extensive, orientation of Tyr-100 or Tyr-99 seen in different MDM2 and tightened intramolecular H-bonding network found in PMI likely MDMX complex structures (Fig. 5) is clearly determined not only plays an important role in stabilizing its more extended ␣-helical by the intrinsic properties of the ␣2Ј helices but also by the chemical conformation in the complex, thus contributing to high-affinity nature of the C-terminal residues of the ligand, supporting the

4668 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900947106 Pazgier et al. Downloaded by guest on September 30, 2021 notion that mutual conformational modulation of a protein-peptide with 400 ␮L of 10 nM biotin-synMDM2/biotin-synMDMX for 60 min before adding complex leads to higher affinity (41). The hydrophobic binding cleft phage-target solution to 50 ␮L of streptavidin-agarose resin (Pierce) for affinity in MDMX induced by Pro-12 of PMI obviously affords an excellent capturing; (ii) wash unbound phage; (iii) elute bound phage with 1 mM (15–29)p53; opportunity to fine-tune binding affinity and specificity for MDM2/ (iv) amplify the eluate and collect phage for the next round of panning; (v) repeat MDMX inhibitors. steps 1–4; (vi) sequence selected binding clones according to the procedures recommended by the manufacturer. Finally, it is worth noting that PMI, despite its high binding affinity for MDM2/MDMX, is much less active than Nutlin-3 in the ϩ/ϩ Surface Plasmon Resonance (SPR) Spectroscopy. Competition binding kinetics killing of p53 HCT116 cells (Fig. S6). The attenuated cytotoxic was carried out at 25 °C on a Biacore T100 SPR instrument, using a (15–29)p53- activity in vitro may reflect a low intracellular concentration of immobilized CM5 sensor chip (17 RUs for MDM2 and 36 RUs for MDMX). The PMI, likely resulting from a combination of proteolytic degrada- buffer was 10 mM Hepes, 150 mM NaCl, 0.005% surfactant P20, pH 7.4. 50 nM tion, inefficient cellular uptake, and endosomal sequestration (42). synMDM2 or 100 nM synMDMX was incubated at room temperature for 30 min For these reasons, peptide inhibitors as potential therapeutic agents with varying concentrations of inhibitor, and injected at a flow rate of 20 ␮L/min are generally considered ‘‘undruggable’’ compared with traditional for 2 min, followed by 4 min dissociation. The concentration of unbound syn- low molecular weight compounds. Importantly, however, the dis- MDM2 or synMDMX in solution was deduced, based on p53-association RU values, covery of small molecule drugs depends on high-resolution crystal from a calibration curve established by RU measurements of different structures of MDM2/MDMX complexed with high-affinity peptide concentrations of synMDM2 or synMDMX injected alone. Nonlinear regres- ligands (22, 23). The information obtained from high-affinity sion analysis was performed using GraphPad Prism 4 to give rise to Kd peptide inhibition of MDM2/MDMX is equally valuable for the values. Protein and peptide solutions were quantified by UV absorbance measurements at 280 nm, using molar extinction coefficients calculated design of miniature proteins that potently activate the p53 pathway from an algorithm published in ref. 46. and effectively kill p53ϩ/ϩ tumor cells in vitro (24, 43, 44). Our structural work reported here should aid in silico library screening Structure Determination and Refinement. Conditions for crystallization and data and structure-based rational design of different classes of p53 collection are described in SI Text. The structures of both complexes were solved activators for anticancer therapy. by molecular replacement, using Phaser (47) and search models based on the Shortly before we submitted this manuscript, the crystal structure previously solved and refined structures of (17–125)MDM2- (15–29)p53 (PDB entry of human (23–111)MDMX in complex with (15–29)p53 at 1.9 Å 1YCR) and zebra fish (15–129)MDMX- (15–29)p53 (PDB entry 2Z5T) (25, 29). The syn resolution was reported (45). The complex structure is very similar structures were refined to 1.65 Å for MDM2-PMI (R ϭ 0.155; Rfree ϭ 0.194) and (15–29) syn ϭ ϭ to zebra fish MDMX- p53 used for comparison in our article. 1.63 Å for MDMX-PMI (R 0.154; Rfree 0.168) with the program Refmac (48), Superposition of human MDMX- (15–29)p53 and synMDMX-PMI and rebuilt using the program COOT (49). Parameters for data collection and yielded a RMSD of 0.8 Å. As observed in zebra fish MDMX- results of refinement are summarized in Table S3. The atomic coordinates of syn syn (15–29)p53 (29), the C-terminal residues of p53 (Pro-27 and Glu-28) MDM2-PMI (3EQS) and MDMX-PMI (3EQY) have been deposited in the are loosely bound to human MDMX with a single stabilizing Protein Data Bank. H-bond between Tyr-99 O␩ and Pro-27 O. The hydrophobic cleft syn Note Added in Proof. Several new MDMX structures have been determined in for Pro-12 of PMI seen in MDMX-PMI is not formed in the complexes with a single-domain antibody (PDB ID code 2VYR) (50), peptido- (15–29) human MDMX- p53 complex. mimetic inhibitors (PDB ID codes 3FE7 and 3FEA) (51), and pDI (PDB ID code 3FDO). Materials and Methods Phage Display. Ph.D.-12—a combinatorial library of random peptide 12-mers ACKNOWLEDGMENTS. This work was supported by American Cancer Society fused, via a short spacer GlyGlyGlySer, to the N terminus of a minor coat protein Research Scholar Grant CDD112858, National Institutes of Health Grants (pIII) of M13 phage—was purchased from New England Biolabs, Inc. The basic AI056264 and AI061482 (to W.L.), and the Intramural Research Program of the procedures for library screening are as follows: (i) incubate input phage (10 ␮L) National Institutes of Health (S.G.T.).

1. Vogelstein B, Lane D, Levine AJ (2000) Surfing the p53 network. Nature 408:307–310. 18. Stommel JM, Wahl GM (2004) Accelerated MDM2 auto-degradation induced by DNA- 2. Vousden KH, Lane DP (2007) p53 in health and disease. Nat Rev Mol Cell Biol 8:275–283. damage kinases is required for p53 activation. EMBO J 23:1547–1556. 3. Vousden KH, Lu X (2002) Live or let die: The cell’s response to p53. Nat Rev Cancer 19. de Graaf P, et al. (2003) Hdmx protein stability is regulated by the ubiquitin ligase 2:594–604. activity of Mdm2 J Biol Chem 278:38315–38324. 4. Kirsch DG, Kastan MB (1998) Tumor-suppressor p53: Implications for tumor develop- 20. Kawai H, et al. (2003) DNA damage-induced MDMX degradation is mediated by ment and prognosis. J Clin Oncol 16:3158–3168. MDM2. J Biol Chem 278:45946–45953. 5. Xue W, et al. (2007) Senescence and tumour clearance is triggered by p53 restoration 21. Murray JK, Gellman SH (2007) Targeting protein–protein interactions: Lessons from in murine liver carcinomas. Nature 445:656–660. p53/MDM2. Biopolymers 88:657–686. 6. Ventura A, et al. (2007) Restoration of p53 function leads to tumour regression in vivo. 22. Shangary S, et al. (2008) Temporal activation of p53 by a specific MDM2 inhibitor is Nature 445:661–665. selectively toxic to tumors and leads to complete tumor growth inhibition. Proc Natl 7. Martins CP, Brown-Swigart L, Evan GI (2006) Modeling the therapeutic efficacy of p53 Acad Sci USA 105:3933–3938. restoration in tumors. Cell 127:1323–1334. 23. Vassilev LT, et al. (2004) In vivo activation of the p53 pathway by small-molecule BIOCHEMISTRY 8. Toledo F, Wahl GM (2006) Regulating the p53 pathway: In vitro hypotheses, in vivo antagonists of MDM2 Science 303:844–848. veritas. Nat Rev Cancer 6:909–923. 24. Hu B, Gilkes DM, Chen J (2007) Efficient p53 activation and apoptosis by simultaneous 9. Marine JC, Dyer MA, Jochemsen AG (2007) MDMX: From bench to bedside. J Cell Sci disruption of binding to MDM2 and MDMX. Cancer Res 67:8810–8817. 120:371–378. 25. Kussie PH, et al. (1996) Structure of the MDM2 oncoprotein bound to the p53 tumor 10. Honda R, Tanaka H, Yasuda H (1997) Oncoprotein MDM2 is a ubiquitin ligase E3 for suppressor transactivation domain. Science 274:948–953. tumor suppressor p53. FEBS Lett 420:25–27. 26. Sakurai K, Schubert C, Kahne D (2006) Crystallographic analysis of an 8-mer p53 peptide 11. Jackson MW, Berberich SJ (2000) MdmX protects p53 from Mdm2-mediated degrada- analogue complexed with MDM2 J Am Chem Soc 128:11000–11001. tion. Mol Cell Biol 20:1001–1007. 27. Grasberger BL, et al. (2005) Discovery and cocrystal structure of benzodiazepinedione 12. Shvarts A, et al. (1996) MDMX: A novel p53-binding protein with some functional HDM2 antagonists that activate p53 in cells. J Med Chem 48:909–912. properties of MDM2 EMBO J 15:5349–5357. 28. Fasan R, et al. (2006) Structure-activity studies in a family of beta-hairpin protein 13. Kubbutat MH, Jones SN, Vousden KH (1997) Regulation of p53 stability by Mdm2. epitope mimetic inhibitors of the p53-HDM2 protein–protein interaction. Chembio- Nature 387:299–303. chem 7:515–526. 14. Haupt Y, Maya R, Kazaz A, Oren M (1997) Mdm2 promotes the rapid degradation of 29. Popowicz GM, et al. (2007) Molecular Basis for the Inhibition of p53 by Mdmx Cell Cycle p53. Nature 387:296–299. 6. 15. Toledo F, et al. (2006) A mouse p53 mutant lacking the proline-rich domain rescues 30. Lai Z, Auger KR, Manubay CM, Copeland RA (2000) Thermodynamics of p53 binding to Mdm4 deficiency and provides insight into the Mdm2-Mdm4–p53 regulatory network. hdm2(1–126): Effects of phosphorylation and p53 peptide length. Arch Biochem Cancer Cell 9:273–285. Biophys 381:278–284. 16. Francoz S, et al. (2006) Mdm4 and Mdm2 cooperate to inhibit p53 activity in prolifer- 31. Schon O, et al. (2002) Molecular mechanism of the interaction between MDM2 and p53 ating and quiescent cells in vivo. Proc Natl Acad Sci USA 103:3232–3237. J Mol Biol 323:491–501. 17. Pan Y, Chen J (2003) MDM2 promotes ubiquitination and degradation of MDMX. Mol 32. Bottger A, et al. (1997) Molecular characterization of the hdm2–p53 interaction. JMol Cell Biol 23:5113–5121. Biol 269:744–756.

Pazgier et al. PNAS ͉ March 24, 2009 ͉ vol. 106 ͉ no. 12 ͉ 4669 Downloaded by guest on September 30, 2021 33. Lin J, Chen J, Elenbaas B, Levine AJ (1994) Several hydrophobic amino acids in the p53 42. Vives E, Schmidt J, Pelegrin A (2008) Cell-penetrating and cell-targeting peptides in amino-terminal domain are required for transcriptional activation, binding to mdm-2 drug delivery. Biochim Biophys Acta 1786(2):126–138. and the adenovirus 5 E1B 55-kD protein. Genes Dev 8:1235–1246. 43. Li C, et al. (2008) Turning a scorpion toxin into an antitumor miniprotein. J Am Chem 34. Picksley SM, Vojtesek B, Sparks A, Lane DP (1994) Immunochemical analysis of the Soc 130:13546–13548. interaction of p53 with MDM2—fine mapping of the MDM2 binding site on p53 using 44. Kritzer JA, et al. (2006) Miniature protein inhibitors of the p53-hDM2 interaction. synthetic peptides. Oncogene 9:2523–2529. Chembiochem 7:29–31. 35. Kabsch W, Sander C (1983) Dictionary of protein secondary structure: Pattern recog- 45. Popowicz GM, Czarna A, Holak TA (2008) Structure of the human Mdmx protein bound nition of hydrogen-bonded and geometrical features. Biopolymers 22:2577–2637. to the p53 tumor suppressor transactivation domain. Cell Cycle 7:2441–2443. 36. Bottger V, et al. (1996) Identification of novel mdm2 binding peptides by phage 46. Pace CN, et al. (1995) How to measure and predict the molar absorption coefficient of display. Oncogene 13:2141–2147. a protein. Protein Sci 4:2411–2423. 37. Dawson PE, Kent SB (2000) Synthesis of native proteins by chemical ligation. Annu Rev 47. Storoni LC, McCoy AJ, Read RJ (2004) Likelihood-enhanced fast rotation functions. Acta Biochem 69:923–960. Crystallogr D 60:432–438. 38. Dawson PE, Muir TW, Clark-Lewis I, Kent SB (1994) Synthesis of proteins by native 48. Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular structures chemical ligation. Science 266:776–779. by the maximum-likelihood method. Acta Crystallogr D 53:240–255. 39. Massova I, Kollman PA (1999) Computational alanine scanning to probe protein– protein interactions: A novel approach to evaluate binding free energies. J Am Chem 49. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. Acta Soc 121:8133–8143. Crystallogr D 60:2126–2132. 40. Zondlo SC, Lee AE, Zondlo NJ (2006) Determinants of specificity of MDM2 for the 50. Yu GW, Vaysburd M, Allen MD, Settanni G, Fersht AR (2009) Structure of human activation domains of p53 and p65: proline27 disrupts the MDM2-binding motif of p53 Mdm4 N-terminal domain bound to a single-domain antibody. J Mol Biol 385:1578– Biochemistry 45:11945–11957. 1589. 41. Dastidar SG, Lane DP, Verma CS (2008) Multiple peptide conformations give rise to 51. Kallen J, et al. (2009) Crystal structures of human MdmX (HdmX) in complex with p53 similar binding affinities: Molecular simulations of p53-MDM2 J Am Chem Soc peptide-analogues reveal surprising conformational changes. J Biol Chem, 10.1074/ 130:13514–13515. jbc.M809096200.

4670 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900947106 Pazgier et al. Downloaded by guest on September 30, 2021