Structural Details on mdm2-p53 Interaction*

Mdm2 is a cellular antagonist of p53 that keeps a balanced cellular level of p53. The two proteins are linked by a negative regulatory feedback loop and physically bind to each other via a putative helix formed by residues 18-26 of p53 transactivation domain (TAD) and its binding pocket located within the N-terminal 100-residue domain of mdm2 (Kussie, P. H., Gorina, S., Marechal, V., Elenbaas, B., Moreau, J., Levine, A. J., and Pavletich, N. P. (1996) Science 274, 948-953). In a previous report we demonstrated that p53 TAD in the mdm2-freee state is mostly unstructured but contains two nascent turns in addition to a “preformed” helix that is the same as the putative helix mediating p53-mdm2 binding. Here, using heteronuclear multidimensional NMR methods, we show that the two nascent turn motifs in p53 TAD, turn I (residues 40-45) and turn II (residues 49-54), are also capable of binding to mdm2. In particular, the turn II motif has a higher mdm2 binding affinity (∼20 μm) than the turn I and targets the same site in mdm2 as the helix. Upon mdm2 binding this motif becomes a well defined full helix turn whose hydrophobic face formed by the side chains of Ile-50, Trp-53, and Phe-54 inserts deeply into the helix binding pocket. Our results suggest that p53-mdm2 binding is subtler than previously thought and involves global contacts such as multiple “non-contiguous” minimally structured motifs instead of being localized to one small helix mini-domain in p53 TAD.

p53 is known to be implicated in more than 50% of all human cancers and probably represents one of the proteins that is most critically associated with cancer (1,2). Understanding how this "hub" in the cancer protein network interacts with other members of the network is not only important for gaining insights into fundamental principles underlying tumorigenesis but also for efficient development of anticancer agents (3)(4)(5)(6). Ironically, establishing a structure-function relationship for p53 has been possible only for ϳ30% of its amino acid residues; namely, for those forming globular domains, a DNA binding domain (7) and an oligomerization domain (8,9). This fact can be attributed to a rather interesting finding that a large fraction (ϳ70%) of amino acid residues in p53 does not participate in forming a well defined tertiary structure, a common feature shared by many intrinsically unstructured proteins (IUPs) 2 (10 -16).
IUPs are an interesting class of proteins that maintain their function despite the lack of a well defined globular structure. Structurally, IUPs are in a similar state as folding intermediates but are distinct from the latter in that they are not in an artificially denatured state. Although some IUPs totally lack any structural elements, others have minimal secondary structural elements (12,16,(17)(18)(19). One subgroup of IUPs consists of unstructured or flexible domains consisting of more than ϳ50 amino acid residues within large mother proteins (12,20,21). Because of their flexible nature, structural features of IUPs can be characterized in a quantitative manner only by NMR spectroscopy since local structural or dynamic features are well reflected in NMR parameters such as NOEs, coupling constants, temperature coefficient of the backbone amide protons, and hydrogen exchange rate of labile protons as well as relaxation behavior (12,22,23). Recent NMR investigations on p53 TAD pointed out that the full-length transcriptional activation domain in p53 (ϳ70 residues at the N terminus of p53) is intrinsically unstructured under physiological conditions (12,13,24). When carefully analyzed, the p53 TAD was found to have three minimally structured motifs, a helix and two nascent turns, although it did not have a globular structure (12). The helix formed by residues 18 -26 can be most clearly identified and preexists even in the absence of any target protein. This helix coincides with the amphipathic helix that was reported to be induced upon mdm2 binding (25).
During the last several years extensive structural studies on the helix motif of p53 TAD have been carried out including a detailed structural investigation of the structure of the p53 helix-mdm2 complex (25)(26)(27)(28). Such studies have paved the way for designing various mdm2 inhibitors as potential anticancer therapeutics (3,4). On the other hand, keen emphasis on the significance of the helix motif has created a view that the helix must be the only specificity determinant in p53 that governs mdm2-binding or transcriptional activity. Noting a valid assumption that proteins would not form structures without a purpose, be they tertiary or secondary, we have attempted to characterize the potential role of the turn motifs in p53 TAD for p53 function. The results indicate that the helix motif is not the only determinant governing p53-mdm2 binding.

MATERIALS AND METHODS
Protein Preparation-The full-length 15 N-labeled p53 TAD (1-73) was expressed and purified as has been reported previously (12). A recombinant mdm2 construct corresponding to residues 3-109 was expressed in pLM1 vector (29). Transformed Escherchia coli BL21(DE3) cells were grown at 37°C to an A 600 of 0.6, and the culture was induced with 0.4 mM isopropyl thio-␤-D-thiogalactopyranoside. Then the cells were further cultivated at 20°C for 16 h. The harvested cell suspension was sonicated in 50 mM Tris-HCl (pH 7.5), 0.4 M NaCl, 1 mM phenylmethylsulfonyl fluoride, 10 mM ␤-mercaptoethanol and centrifuged for 30 min at 30,000 ϫ g. Proteins in the supernatant were precipitated with ammonium sulfate. For purification the same method was used to purify non-labeled, 15 N-labeled, or 13 C, 15 N-labeled mdm2-(3-109) using an SP-Sepharose column, a Q-Sepharose column, and a Hiprep 26/60 Sephacryl S-200 FPLC column (Amersham Biosciences). The molecular weights of the purified proteins were confirmed by MALDI-TOF mass spectrometry.
NMR Spectroscopy-NMR spectra were acquired using a Varian Unity INOVA 600 spectrometer, a Bruker DRX 600 MHz spectrometer equipped with a cryoprobe, and a Bruker DRX 800 MHz spectrometer. Aliquots of a concentrated mdm2 stock solution were added in a stepwise fashion to the 15 N-labeled p53 TAD during titration. NMR samples containing 0.4 mM 15 N-labeled p53 TAD alone or with mdm2 were prepared in 90% H 2 O, 10% 2 H 2 O, 50 mM tritiated sodium acetate (pH 6.3) and 50 mM NaCl and remained stable for more than 6 months. At each titration point an 1 H, 15 N heteronuclear single quantum coherence spectroscopy (HSQC) spectrum was collected at 5°C. The molar ratios of p53 TAD to mdm2 were 1:0.3, 1:0.6, 1:1, and 1:2. Next, 15 N-labeled mdm2-(3-109) was titrated with a series of non-labeled p53 peptides including the helix peptide (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29), the linker peptide (28 -37), the long turn peptide (39 -57), the turn I peptide (39 -48), the turn II peptide (49 -54), and the Trp analog of the turn II. During titration, 1 H, 15 15 N HSQC spectra of the full-length p53 TAD in the absence (A) and in the presence (B) of mdm2-(3-109) is shown. For the bound state, the molar ratio of p53 TAD to mdm2-(3-109) is 1:1. A ϩ symbol is used to indicate the disappeared resonances, and a color scheme is used for visual clarity; red, the helix-forming residues, green, the turn residues; blue, others that are affected by mdm2 binding. Shown in panel C is the amino acid sequence of p53 TAD along with locations of secondary structural elements previously assigned (12) and observed spectral changes due to mdm2 binding. Open circles indicate weakened peaks, whereas the filled circle is used for disappeared peaks. Gray squares denote the residues whose chemical shifts have been changed upon mdm2 binding. Color coding scheme is the same as that used in panel B.
obtained by standard two-dimensional NMR experiments such as TOCSY, double-quantum-filtered correlation spectroscopy, and NOESY. Mixing times of 100 ms for NOESY and 70 ms for TOCSY experiments were used. The three-bond coupling constants, 3 J HNH␣ , for backbone torsion angles were measured using phase-sensitive doublequantum-filtered correlation spectroscopy experiments. The two-dimensional NMR data consist of 2048 complex points in the t 2 dimension with 256 complex t 1 increments. Spectral widths were 8 kHz in both dimensions.
Molecular Modeling-A structural model for the complex of mdm2 with p53 turn II motif was generated using the molecular modeling package Insight II on a Silicon Graphics O 2 work station. On the basis of transferred NOE data, the turn II peptide of p53 was modeled into an ideal ␣-helix using the builder and biopolymer modules of Insight II.
Initially, the turn II peptide was manually docked on the p53 helix binding groove of the crystal structure of mdm2 (25-109) (Protein Data Bank code 1YCR). The modeled complex structure was based on the contact points from the NMR chemical shift perturbations. The structure was energy-minimized until convergence to 0.01 kcal/mol/Å to remove steric clashes. The DISCOVER module of Insight II and cvff forcefield were used to carry out energy minimization and molecular dynamics. The molecular dynamics simulation was performed for 200 ps at 300 K with the protein backbones fixed. The integration time step was set up to 1 fs. At the end of the dynamics run, energy minimization was performed again until convergence to 0.01 kcal/mol/Å. Figures were drawn by using the programs GRASP and RIBBONS.

RESULTS
To test the hypothesis that the turn motifs or any other segments in p53 TAD may bind to mdm2, we first acquired and compared two 1 H, 15 N HSQC spectra of 15 N-labeled p53 TAD, one in the absence and the other in the presence of mdm2-(3-109). The results summarized in Fig. 1 indicate that mdm2 binding affects a large number of resonances in p53 TAD, not just those from the helix; ϳ60% (43 of 73) of p53 TAD resonances either completely disappear, broaden, or experience chemical shift perturbation upon mdm2 binding. No further changes in the NMR spectra were observed beyond the titration point of (mdm2/p53 TAD) ϭ 1. The fact that the resonances from the helix-forming residues disappear due to mdm2 binding concurs well with the previous obser-  (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29). B, the linker peptide (28 -37). C, the long turn peptide (39 -57). D, the turn I peptide (39 -48). E, the turn II peptide (49 -54). F, the Trp analog of turn II. The ⌬␦ c 1 H, 15 N value is calculated as described earlier (27) when the molar ratio of mdm2 to p53 is 1:3. Note the different scales of ⌬␦ c 1 H, 15 N depending on the panels.
vation that the helix motif binds to mdm2 with an affinity of ϳ1 M (25). On the other hand our data suggest that the residues from the linker region (residues 28 -37) between the helix and the turns and those involved in formation of nascent turns may be involved in mdm2 binding as the resonances from these residues also experience spectral change. Binding of p53 TAD (ϳ8 kDa) to mdm2 (ϳ12 kDa) would slow down the overall tumbling motion of p53 TAD and consequently shorten its transverse relaxation time (T 2 ), resulting in broader resonances. Further resonance broadening is expected for the helix-forming residues since these residues may exist in an mdm2-bound or an mdm2unbound state as the mdm2 binding affinity of the helix motif is only of intermediate strength (25,26). Resonance broadening due to these two factors seems to be sufficient to make the signals from the helix-forming residues in p53 TAD disappear completely upon titration with mdm2. Similar resonance broadening has been noticed in binding of p53 TAD with hRPA 1-168 (23). Immediate neighbors of the helix such as the residues Ser-15-Glu-17 also experience non-negligible chemical shift perturbations.
In contrast to the helix-forming resonances, the turn-forming resonances experience only chemical shift change because the turn motifs have weaker mdm2 binding affinities of 20 -100 M (Fig. 5). However, intensities of the turn-forming resonances are also weakened to a noticeable degree due to exchange broadening, as can be seen in Fig. 1B. The addition of a helix peptide to the mixture of mdm2 and p53 TAD makes the disappeared helix-forming resonances of p53 TAD nearly fully reappear, whereas adding a long turn peptide encompassing both turns (residues 39 -57) leads to only partial reappearance (data not shown). Interestingly, the resonances from the "linker" residues, i.e. between the helix and the turns, also disappear even though they do not directly participate in mdm2 binding (Figs. 2B and 3B). This linker region was previously shown to have some structural order rather than being disordered. Because their T 2 values are as short as those of the helix-forming residues in the mdm2-free state of p53 TAD (12), additional reduction in T 2 due to formation of p53 TAD-mdm2 complex seems to cause their resonances to disappear.
We then examined whether each of the mini-domains in p53 TAD, the helix, the linker, the turn I, and turn II that experience spectral change due to p53 TAD-mdm2 binding is able to bind to mdm2. This was accomplished by recording a series of 1 H, 15 N HSQC spectra of 15 N-labeled mdm2 in the presence of various p53 TAD fragments corresponding to these mini-domains. Shown in Fig. 2 are six HSQC spectra for the helix, the linker, the long turn peptide (residues 39 -57), the turn I, the turn II, and a Trp analog of the turn II, respectively. Fig. 3 is a summary of chemical shift perturbation for individual residues in mdm2 due to binding of various p53 TAD peptides. Chemical shift perturbation due to the turn I is small (⌬␦ c Ͻ 0.1), indicating that its binding is minimal. Binding of the linker peptide is negligible with ⌬␦ c Ͻ 0.03, confirming that disappearance of the linker resonances ( Fig. 1) is not due to mdm2 binding. These results are depicted in the color-coded structures of mdm2 shown in Fig. 4. In the backside of structures (data not shown) several residues outside of the helix binding pocket are affected due to ligand binding, indicating that there is a long range conformational change within mdm2 upon ligand binding as previously noted (27). Fast hydrogen exchange rates (disappearing within less than an hour after addition of D 2 O) of most mdm2 backbone amide protons (data not shown) suggest that the N-terminal domain (3-109) of mdm2 undergoes breathing rather than being rigid despite its globular shape.
Several analogs of the turn II with reasonable hydrophobicity have severely reduced mdm2 binding affinities, suggesting that the hexameric sequence integrity DIEWQF of turn II is important for mdm2 binding. One exception is the Trp analog, which shows a tighter binding to mdm2 than the native peptide (Fig. 3). When tested with a larger mdm2 N-terminal domain (residues 10 -154), the mdm2 residues affected by binding of the turn II peptide remain the same, indicating that there is no other binding site for the turn motifs beyond residue 109 in mdm2.  (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29). B, the linker peptide (28 -37). C, the long turn peptide (39 -57). D, the turn I peptide (39 -48). E, the turn II peptide (49 -54). F, the Trp analog of turn II. The chemical shift perturbations shown in Fig. 3 are mapped onto the molecular surface of mdm2 from the crystal structure of mdm2 bound with the p53 helix peptide (24). The value of ⌬␦ c is calculated when the molar ratio of mdm2 to p53 is 1:3. Color-coding is based upon the degree of chemical shift perturbation: gray, ⌬␦ c Ͻ 0.1 ppm; yellow, 0.1 ppm Ͻ ⌬␦ c Ͻ 1 ppm; red, ⌬␦ c Ͼ 1 ppm.  The peptide was docked as a helical turn as obtained by transferred NOE experiments onto the binding cleft of mdm2 formed by residues 25-109, and molecular dynamics simulation was performed by DISCOVER. Positive and negative potentials are colored blue and red, respectively. The backbone of the turn II peptide is shown as a violet worm. Bottom, a schematic ribbon diagram of mdm2 bound with the p53 turn II (49 -54) peptide. The residues of p53 peptide are presented in a ball and stick mode with nitrogen atoms in blue and oxygen atoms in red. The hydrophobic residues in p53 that complement the hydrophobic surface in mdm2 are labeled. The helical axis of the turn II motif runs in an opposite direction to that of the helix motif (25).
Together, these results exclude the possibility that binding of the turn motifs to mdm2 is due to nonspecific interactions. Surface plasma resonance experiments have yielded a K d for the turn II motif of ϳ20 M (Fig. 5), whereas the turn I motif binds only minimally to mdm2. The binding of turn I is so weak that its K d is estimated to be Ͼ100 M, which is more or less a measurement limit by the surface plasma resonance technique. Because a reasonable correlation exists between the K d and the ⌬␦ c values, one may use the ⌬␦ c value of an mdm2 ligand as an indicator for mdm2 binding affinity. The long turn peptide is likely to exhibit a slightly higher affinity than the turn II peptide based on the chemical shift perturbation (compare Fig. 3, C and E).
The result that the significantly affected mdm2 residues due to binding of the turn II and the long turn peptide mostly overlap with those affected by helix binding suggests that this turn motif binds to the same helix binding pocket in mdm2. However, the turn-bound mdm2 fraction would be small since the mdm2 binding affinity of the turn region is weaker than that of the helix. One should view the results shown in Fig.  1C as reflecting the presence of a mixture between the helix-bound form of mdm2 and an mdm2 complexed with the turn region (40 -55) of p53 TAD. Interestingly, binding of the helix and turn II differentially influences the residues within the same helix binding pocket. The turn II affects resonances in the order Phe-55 Ͼ Glu-25 Ͼ His-73 Ͼ Lys-94, whereas the helix affects resonances in the order His-73 Ͼ Val-93 Ͼ Tyr-100 Ͼ Ser-22. Therefore, the binding modes of the helix and the turn II must be different from each other. To characterize the mdm2 binding mode of the turn II motif in detail, we have applied transferred NOE methods to the long turn peptide as shown in Fig. 6A. Upon mdm2 binding the turn II motif becomes more ordered from a nascent turn to a well defined one consisting of a complete turn of an amphipathic helix as shown in Fig. 6B. This structure was generated by a molecular docking process where the helical axis of the turn II motif runs in an opposite direction to that of the helix motif. One face of this one-turn helix contains hydrophobic residues such as Ile, Trp, and Phe, favorable for binding to the hydrophobic helix binding pocket in mdm2. The docked structure of the turn II motif consistent with the chemical shift perturbation data can be generated only with the shown orientation of the turn II peptide. The turn II does not fully fill up the helix binding pocket as does the helix. Differences in the binding mode between the helix and the turn II are reflected in their mdm2 binding affinities; the K d for the helix is ϳ1 M, whereas the K d of the turn II is ϳ20 M.

DISCUSSION
We have shown that in addition to the well known helix located within the N-terminal subdomain, there are two additional mini-domains or motifs within the C-terminal subdomain of p53 TAD that are able to bind to mdm2. As found in many protein-protein interactions involving weak but specific binding motifs (21,30), p53-mdm2 binding apparently involves only weak interactions that rely mostly on hydrophobicity. The mdm2 binding affinity of the helix motif, which binds most tightly among the tested p53 peptide tested, is only ϳ1 M (3,25). The affinities of the turn motifs described here are even weaker, 20 M or larger. Even though their mdm2 affinities are weak, a specific sequence requirement for the turns, especially for the turn II motif, suggests that binding of this turn motif to mdm2 is not opportunistic. Our results suggest we now need to view the p53-mdm2 binding with a new perspective considering that there are multiple, not one, functional motifs in p53 TAD that mediate binding of p53 to mdm2. A recent study indicates that the turn region is involved in binding to yet another protein (23), supporting the notion that the minimally structured motifs we have originally identified (12) have a broad implication for p53 function.
It remains to be determined why there should be such multiple weak mdm2 binding motifs that are positionally conserved or in tandem in p53 TAD. One may envisage that the presence of multiple motifs capable of competing for the same binding pocket facilitates the initial recruitment of co-activators and targets into the transcriptional machinery. A strategy of "facilitated recruitment" is utilized in some nuclear receptor superfamily transcriptional activators containing multiple hydrophobic LXXLL motifs, or "NR boxes," which compete for the same hydrophobic site in a target protein. These activators use a "temporal binding" scheme to avoid competition among multiple motifs (14,31). Such multiple mini-activation domains in nuclear hormone receptors are known to be responsible for long range intramolecular hydrophobic interactions and synergism in transcriptional activity (14,31,32). In the case of p53, the presence of the turns in p53 TAD could increase the initial mdm2-bound fraction of p53 compared with the case where only the helix motifs exists in p53 TAD. This is because there should be a small fraction of a turn-bound mdm2 that may eventually convert to a more stable helix-bound complex. More generally, multiple motifs may be useful in recruiting more than one target proteins or domains within a large transcriptional initiation complex (30). Another possibility is that multiple motifs with differing affinities toward different contact points of one target protein may interact with yet different affinities to different contact points in other target proteins. For example, it has been shown recently that the turn region of p53 TAD can bind to hRPA  , whereas the helix does not bind (23). The well known fact that several other activation domains such as VP16 TAD of herpes simplex virus (33) and activation domains in E2F1 (34) and HIF-1␣ (35) also contain similar multiple hydrophobic mini-domains (Fig. 7) capable of forming local minimally structured motifs (29) suggests that what has been found in p53 TAD may prove to be a general structural or functional feature in other transcriptional activation domains or factors.
A rather fascinating finding is that such minimally structured functional motifs in p53 (including the helix) with differing mdm2 binding affinities are all present within the transactivation domain of p53, where most of the post-translational modification sites are also located. The whole p53 has 16 posttranslationally modifiable sites, 13 phosphorylation sites and 3 acetylation sites (12,36). Nine (2 threonines and 7 serines) of 13 phosphorylation sites in p53 are found within its TAD. Phosphorylation of two threonines has a profound effect on p53 function; Thr-18, which flanks the helix motif at the N terminus, disables the binding of mdm2 to p53 TAD when phosphorylated (37). Phosphorylation of Thr-55 that flanks the turn II motif at the C terminus also destabilizes p53 (38). Phosphorylation of serines in the p53 TAD is also associated with p53-mediated transcription. For example, modification of Ser-15, which is located near the helix, influences p53 TAD-mdm2 binding (39) and enhances the activity of p53 via cAMP-response element-binding protein (CREB)-binding protein/p300 binding (40). Phosphorylation of Ser-20 that is in the middle of the helix prevents degradation of p53 by inhibiting binding of mdm2 to p53 (41). Also, phosphorylation of Ser-33 and Ser-37 belonging to the linker region of p53 TAD results in non-negligible changes in p53 function (42). Recent studies point out that the sites of structural disorder and phosphorylation in many other unstructured proteins are highly correlated (43). In this regard, placing local structural motifs along with various posttranslational modification sites within a relatively small transactivation domain of p53 may well represent an elegant way of producing highly diverse functional states of p53, enabling the hub in the cancer protein network to promiscuously interact with many targets (1).
The helix and the turns placed in tandem within the loosely folded p53 TAD "preexist" in the target-unbound state (12) and act as low affinity hooks on a "fishing line of an unstructured protein" (30) allowing rapid reversible modulation of intracellular signals (15). IUPs are predicted to occupy a good fraction of the whole protein kingdom (11,44), and several of them are found to have such pre-formed motifs in their target-unbound state (16). Technically, minimally structured motifs such as the turn motifs in p53 TAD have only minimal structural ordering in the absence of target protein and can be detected only by meticulous NMR analysis (12). Our NMR strategy used in this investigation might be useful for characterizing similar minimal structural features of other IUPs. Mdm2 continues to be the prime target of anticancer agent development (3)(4)(5)(6), and several ␣-helix based peptides are being designed as mdm2 antagonists. Because the turn II motif is able to form one ␣-helical turn in the mdm2-bound state, it may turn out to be another useful structural template for designing novel mdm2 antagonists.