MDMX contains an autoinhibitory sequence element

Significance

The protein MDMX is an important negative regulator of the tumor suppressor p53 in normal cells and can become a powerful oncogene. Previous studies on isolated protein fragments established that the N-terminal domain of MDMX binds tightly to the N-terminal domain of p53 and inhibits it. We now find that full-length MDMX contains a regulatory element (the “WWW element”) that binds to its own N-terminal domain and prevents MDMX from binding to p53. This autoinhibition introduces a new level of regulation of MDMX, whereby accessory proteins may sequester the WWW sequence and reactivate MDMX, and may explain why a variant of MDMX that lacks the WWW sequence, which is found in some cancer cells, might be oncogenic.

Abstract

MDM2 and MDMX are homologous proteins that bind to p53 and regulate its activity. Both contain three folded domains and ∼70% intrinsically disordered regions. Previous detailed structural and biophysical studies have concentrated on the isolated folded domains. The N-terminal domains of both exhibit high affinity for the disordered N-terminal of p53 (p53TAD) and inhibit its transactivation function. Here, we have studied full-length MDMX and found a ∼100-fold weaker affinity for p53TAD than does its isolated N-terminal domain. We found from NMR spectroscopy and binding studies that MDMX (but not MDM2) contains a conserved, disordered self-inhibitory element that competes intramolecularly for binding with p53TAD. This motif, which we call the WWW element, is centered around residues Trp200 and Trp201. Deletion or mutation of the element increased binding affinity of MDMX to that of the isolated N-terminal domain level. The self-inhibition of MDMX implies a regulatory, allosteric mechanism of its activity. MDMX rests in a latent state in which its binding activity with p53TAD is masked by autoinhibition. Activation of MDMX would require binding to a regulatory protein. The inhibitory function of the WWW element may explain the oncogenic effects of an alternative splicing variant of MDMX that does not contain the WWW element and is found in some aggressive cancers.

The homologous proteins MDMX and MDM2 are major negative regulators of the level of p53 and its activity in normal cells. They are frequently up-regulated in many types of cancers, becoming oncogenes (1–3). Both proteins bind to the transactivation domain of p53 (p53TAD) and inhibit its interaction with transcriptional cofactors. In addition, MDM2 is an E3 ubiquitin ligase with a catalytically active RING domain and can target p53 for proteosomal degradation (4). MDMX has an inactive RING domain but can efficiently inhibit the p53 activity, either independently [e.g., in the vast majority of primary melanomas (5)] or in association with MDM2 by boosting its E3 ligase activity (6). Both proteins are vital for homeostasis, and knockout of either of them in mice models results in p53-dependent lethality, with survival times longer for MDMX deficiency (3).

MDM2 and MDMX are intrinsically disordered proteins, each with three folded domains (Fig. S1A). The structures of the isolated domains have been elucidated at atomic resolution (Fig. S1 B–D): the N-terminal, globular SWIB-like domain (NTD) binds to the p53 transactivation domain p53TAD (7, 8) [with W23^p53 contributing the most interaction energy (9, 10)]; the central part of the protein contains a RanBP2 type zinc finger (ref. 11 and PDB ID code 2CR8) and the C terminus is a RING-type ubiquitin ligase domain, which tends to associate and form homodimers and heterodimers and higher-order oligomers (12, 13). There are specific MDM2 inhibitors that bind to the NTD [e.g., nutlin-3 (14)], which usually have lower potency against MDMX (8). The most oncogenic form of these proteins is the MDM2–MDMX heterodimer (3, 15–17).

There are few detailed biophysical and structural studies on the full-length proteins and how the individual domains interact (18–21). Here, we analyze MDMX and find that there is a regulatory autoinhibitory sequence within MDMX that regulates its binding characteristics with p53, with biological implications.

Results

MDMX Reversibly Dimerizes.

MDMX was expressed in Escherichia coli and purified to homogeneity using standard methods (12, 21). We found by size-exclusion chromatography (SEC) that the protein eluted at volumes typical for ∼400-kDa globular proteins, much earlier than expected from a monomeric 55-kDa molecule (Fig. 1A). The elution volume was concentration dependent, suggesting that the protein was undergoing a reversible oligomerization. Multiangle light scattering (MALS) analysis yielded molecular weight of the eluted protein fraction to be ∼100–110 kDa, in agreement with a dimer (110 kDa)–monomer (55 kDa) equilibrium. We used analytical ultracentrifugation to find the dimerization K_d to be 1.12 ± 0.18 µM (Fig. S2). C-terminally truncated mutants of MDMX were exclusively monomeric, showing that the RING domain was essential for dimerization.

Fig. 1.

MDMX is dimer capable of stimulating E3 activity of MDM2. (A) SEC-MALS analysis of MDMX and p53; A280 chromatograms were scaled to give the same peak heights. Masses calculated from the static light scattering are shown as thick lines. Composition of samples subjected to the analysis described in the legend. (B) Enzymatic activity of recombinant proteins; in vitro ubiquitination reactions were resolved by denaturing electrophoresis and immunoblotted with anti-ubiquitin antibodies, which recognize polyubiquitinated forms more strongly than monoubiquitinated (see Fig. S3 for detection using other antibodies).

MDMX Binds to p53.

We mixed MDMX with tetrameric p53 and subjected the complexes to SEC-MALS analysis. Their sizes were dependent on the stoichiometry, reaching maximal values of ∼400 kDa for equimolar mixtures, which corresponded to octamers built of four MDMX and four p53 subunits (Fig. 1A). Interestingly, even though interaction of flexible, multivalent receptor–ligand pairs (like MDMX and p53) could potentially lead to formation of high–molecular-weight polymers (22), we found a size cap of 400 kDa that was preserved even at high ratios of MDMX to p53 and vice versa.

MDMX Stimulates Ubiquitination.

SEC-MALS analysis confirmed that the recombinant MDMX contained functional NTD (responsible for p53 binding) and C-terminal RING domain capable of homodimerization. To confirm the biochemical activity of recombinant proteins, we performed an in vitro ubiquitination assay. MDMX did not exhibit intrinsic E3 ligase activity, and MDM2 was only weakly active. Mixing the two recombinant proteins yielded a highly active complex capable of performing polyubiquitination (Fig. 1B). The results of the in vitro assay performed with recombinant proteins were in good agreement with the previously reported activities of endogenous proteins in cellular context (6); we found that MDMX stimulated both autoubiquitination of MDM2–MDMX complex and p53 ubiquitination (Fig. S3) (23).

NMR Spectra of MDMX.

We labeled the full-length MDMX with ¹⁵N and recorded a transverse relaxation-optimized spectroscopy (TROSY) spectrum. The spectrum shown in Fig. 2A was dominated by signals from highly flexible, intrinsically disordered linkers. Such appearance, with majority of resonances occupying the central part of the spectrum is typical of intrinsically disordered proteins. In addition to these signals from disordered linkers, we could also identify a range of broader and more disperse resonances originating from the folded NTD and RanBP2 domains. The absence of RING signals from the spectrum was anticipated because in multidomain proteins with intrinsically disordered linkers, central domains are frequently more constrained rotationally than structures located distally (24, 25).

Fig. 2.

Overlays of ¹⁵N-TROSY spectra of MDMX. (A) ¹⁵N-MDMX (red) and isolated ¹⁵N-NTD (16–116; gray). (B) ¹⁵N-MDMX–¹⁴N-p53TAD (17–32) complex (blue) and ¹⁵N-NTD (16-116)–¹⁴N-p53TAD (17–32) complex (yellow). (C) ¹⁵N-MDMX–¹⁴N-p53 complex (green). Insets show fragments overlaid with spectrum (A).

There were unexpected significant differences between the spectra of the NTD in full-length protein and the isolated domains, implying that MDMX–NTD interacts with another part of the protein (Fig. 2A). Addition of the p53TAD peptide to MDMX gave an NMR spectrum that corresponded to that from the isolated p53TAD–NTD complex (Fig. 2B). Surprisingly, besides anticipated chemical shift perturbations in the canonical p53-binding pocket within the NTD, several sharp resonances characteristic for unstructured linkers appeared on addition of p53TAD to MDMX. Particularly noticeable were three downfield shifted signals that appeared at frequencies typical for HN groups of Trp side chains.

The TROSY spectrum of a ¹⁵N-MDMX–¹⁴N-p53 complex (Fig. 2C) was also dominated by signals of flexible linkers, although few residual signals of RanBP2 domain (but not NTD) were also identifiable. Notably, the detectable resonances aligned well with the MDMX–p53TAD complex, and similar patterns were observed in the random coil and Trp regions of the spectrum. The absence of NTD signals is most likely explained by exchange between binding of TAD (p53 19–26) and a further transactivation fragment (p53 49–54) (26) and possibly also rotational restriction of the domain in the complex.

NTD Interacts Intramolecularly with a Conserved “WWW Element.”

The spectral changes in the NTD and flexible linkers suggested that the p53TAD displaces an intramolecular interaction of the NTD with a segment of MDMX, releasing Trp residues. Basing on observing that three tryptophans were involved and primary sequence analysis, we hypothesized that the interaction might involve a tryptophan-rich segment within residues 190–210 (¹⁹⁰FEEWDVAGLPWWFLGNLRSNY²¹⁰, which we call the WWW element). Notably, the element is specific to MDMX, as the corresponding region of MDM2 contains an unrelated sequence (Fig. 3A).

Fig. 3.

WWW element interacts with NTD. (A) Multiple sequence alignment of WWW sequences from different species; corresponding sequence of MDM2 lacks similarity. (B) ¹⁵N-HSQC, assigned spectrum of isolated WWW (181–209). (C) ¹⁵N-WWW (181–209) bound to ¹⁴N-NTD (1–111).

To verify binding of the segment experimentally, we ¹⁵N-labeled the WWW-containing peptide (181–209) and titrated it with the isolated NTD. The NTD–WWW interaction was strong enough to occur between isolated domains that were not covalently linked. Numerous resonances of the WWW were broadened on binding. We assigned the heteronuclear single-quantum coherence (HSQC) spectrum of the WWW construct and mapped the interaction site to amino acids 194–206 (Fig. 3 B and C). Interestingly, the WWW element remained in the intermediate timescale of chemical exchange in excess of MDMX–NTD. This suggested that, even in the bound state, the WWW was not stabilized in a single conformation. The unbound, isolated WWW seemed to lack stable structure, as its backbone chemical shifts were close to the random coil values, ¹H-¹⁵N heteronuclear nuclear Overhauser effect (NOE) enhancements were low and nuclear Overhauser effect spectroscopy (NOESY) spectrum revealed little interresidual contacts (Fig. S4). Notably, we observed that p53TAD and to a lesser extent also nutlin-3 were capable of dissociating the preformed NTD–WWW complex (Fig. S5 A–E).

Addition of ¹⁴N-WWW to ¹⁵N-NTD (Fig. 4A) caused numerous NTD cross-peaks to become broadened or perturbed. The chemical shift perturbations in the NTD were analogous to those induced by p53TAD; using the NTD backbone assignment, we confirmed that the WWW and p53 TAD binding sites coincided (Fig. 4C). The resulting spectrum of the NTD–WWW complex reproduced that of full-length MDMX (Fig. 4B).

Fig. 4.

WWW binding resembles p53TAD in binding. (A) Titration of 15N-NTD (16–116) with ¹⁴N-WWW (181–209). Reference spectrum in red and NTD–WWW complex in blue. (B) Overlay of full-length MDMX spectrum (yellow) with spectra of NTD–WWW complex (blue) and isolated zinc finger domain (pink). (C) WWW binding site mapping. Areas around amides with chemical shift perturbations (Δδ_1H²+0.04Δδ_15N²)^1/2 > 0.2 ppm colored red.

The TROSY spectrum of a monomeric MDMX variant (1–340) matched both the spectrum of the NTD–WWW complex and that of full-length MDMX, suggesting that the NTD domain in MDMX interacts with the WWW element in its own chain (Fig. S5F).

WWW Element Inhibits p53 Binding.

We used isothermal titration calorimetry to measure the thermodynamics of MDMX–p53 interactions and the contribution of the WWW element (Table 1). p53TAD was tightly bound by isolated NTD (K_d = 30 nM). In contrast, MDMX bound p53TAD 100-fold less tightly (K_d = 2900 nM). Titrations with truncated MDMX variants (Table 1, with p53TAD) revealed that as long as the WWW element was present, p53 binding was inhibited. The K_d of NTD and WWW was 1,300 nM.

Table 1.

ITC results

Protein	KD, nM	ΔH, kcal/mol	ΔS, cal⋅mol−1⋅K−1
With p53TAD*
MDMX	2,870 ± 558	−8.1 ± 0.8	−2.3
MDMX 1–340	2,700 ± 531	−8.3 ± 0.6	−2.8
MDMX 1–303†	8,200 ± 826	−10.3 ± 0.3	−11.9
MDMX 1–238†	5,560 ± 491	−16.3 ± 0.9	−31.5
MDMX 1–111	30.3 ± 3.9	−13.9 ± 0.1	−13.2
MDMX ΔWWW‡	92.6 ± 11.7	−17.6 ± 0.2	−27.7
MDMX W200D/201D	385 ± 43.5	−15.3 ± 0.3	−22.7
With p53
MDMX + p53§	1,110 ± 101	−8.3 ± 0.2	−1.2
With WWW element
MDMX 1–111¶	1,290 ± 641	−3.4 ± 0.4	15.3
Mutants with p4||
MDMX	508 ± 36.8	−14.4 ± 0.1	−20.3
MDMX W200A	209 ± 27.6	−15.8 ± 0.2	−23.4
MDMX W201A	485 ± 50	−14.3 ± 0.2	−19.9
Splicing variant**
MDMX-S + p53TAD	40.3 ± 6.3	−15.2 ± 0.2	−18.0

^*p53TAD, residues 17–32.

^†Protein has C-term His-tag, which minimizes degradation.

^‡Residues 193–210 deleted.

^§Independent association of p53 and MDMX monomers assumed; values calculated for monomers concentrations.

^¶Sequence 181–211 of MDMX.

^||P4, high-affinity p53-derived peptide [sequence (5,6-FAM)-LTFEHYWAQLTS] (37).

^**MDMX alternative splicing variant (27).

The K_d of MDMX with full-length p53 (1,110 nM) was similar to that with p53TAD (2,870 nM), implying that association of the NTDs is the main energetic component of binding.

Deletion or Mutation of the WWW Element Enhances Binding.

We further corroborated the WWW–NTD interaction in the full-length protein, by introducing mutations in WWW aiming to disrupt the WWW–NTD interaction. As a key residue for p53–NTD interaction is tryptophan (9, 10) and the NMR spectrum of the NTD–WWW complex resembles that of the NTD–p53TAD complex, we focused on mutating tryptophans that were implied by NMR to be directly involved in the interaction (W200 and W201). Although mutation of a single Trp to Ala did not increase the p53 affinity markedly (Table 1), double mutation to Asp had more pronounced effect and complete removal of the WWW element resulted in MDMX binding p53 almost as tightly as the isolated NTD (90 nM).

MDMX-S Splicing Variant.

MDMX is regulated at the posttranscriptional level by alternative splicing. One of the main splicing forms, MDMX-S, is a product of exon 6 skipping, producing a 127-aa protein sequence containing the NTD and 17 additional amino acids (27). Recombinant MDMX-S binds p53 with high affinity similar to the isolated NTD and by two orders of magnitude stronger than full-length MDMX. Owing to their large difference in p53 binding affinities, even low levels of MDMX-S relative to MDMX may thus have a significant effect on p53 activity.

Discussion

We noted that the NMR spectrum of the N-terminal domain of MDMX resembled that of the isolated domain when bound to a peptide rather than the unligated domain. Furthermore, we noted that three tryptophan side chains in a disordered region of the protein had perturbed signals. We surmised that an internal sequence of the protein was binding to its N-terminal domain. We located the relevant sequence element by deletion and point mutation analysis, and showed that the isolated sequence element bound tightly to MDMX–NTD. Furthermore, we found that MDMX lacking the sequence element bound 32-fold more tightly to p53.

Consequently, there is an important sequence element in the central region of MDMX (but not MDM2), the existence of which implies a mechanism for its regulation. Residues 190–210 in that intrinsically disordered region constitute an inhibitory module, which we term the WWW element. It competes with the p53TAD for binding to the NTD of MDMX, thus lowering the affinity of MDMX for p53. Efficient binding of MDMX to p53 requires a mechanism to relieve the allosteric self-inhibition.

The WWW element does not contain known posttranslational modifications sites, implying that activation of MDMX for p53 binding must be achieved by another mechanism. There is some evidence for two processes: alternative splicing; or by accessory proteins that can bind the WWW element (Fig. 5).

Fig. 5.

Regulation of MDMX activity by autoinhibition. Full-length MDMX rests in a latent state in which its binding activity with p53TAD is masked. Activation of MDMX can achieved either by alternative splicing, yielding a truncated MDMX variant (MDMX-S), or by binding of the inhibitory module by an accessory factor.

Alternative splicing is a very common regulatory mechanism in proteins containing inhibitory modules (28), and indeed, a splicing variant MDMX-S, containing only the NTD is overexpressed in some cancers and strongly associated with a negative outcome (29, 30). The affinity of MDMX-S toward p53 was two orders of magnitude higher than for the standard variant, which may be a possible explanation of the high oncogenic potential of the short protein.

The sequence of the WWW has been strongly preserved during evolution (Fig. 3A) (31). High abundance of hydrophobic amino acids and its intrinsically unstructured character, suggested that the motif is likely to be a promiscuous binding site for other partners besides the NTD of MDMX. Indeed, Chen and coworkers (32, 33) have recently found that the WWW element provides a docking site for casein kinase 1α, which in turn activates MDMX for p53 binding and phosphorylates it on S289. DNA damage relieves the CK1α–MDMX interaction, resulting in release of p53 from the complex. Accordingly, it is likely that other p53TAD binding partners and other proteins might activate MDMX by binding to the WWW segment.

Materials and Methods

Protein Production.

Full-length MDMX, MDM2, as well as truncated constructs [MDMX 1–111, MDMX-S, MDMX 16–116, MDMX 1–238 C-terminal 6×His, MDMX 1–303 C-terminal His, MDMX 1–340, MDMX W200,201D, MDMX ΔWWW (lacking residues 193–210), MDMX-WWW 181–209, MDMX zinc finger 292–340] were cloned into pETM plasmids (34) (obtained from the Protein Expression and Purification Core Facility at the European Molecular Biology Laboratory Heidelberg) and purified to homogeneity using a combination of affinity and size-exclusion chromatographies, using a protocol that essentially has been described in refs. 12 and 21. The final purification step for all constructs except for MDMX 181–211 was SEC in a buffer containing 20 mM Tris, pH 7.2, 150 mM NaCl, and 0.5 mM tris(2-carboxyethyl)phosphine (TCEP). MDMX WWW 181–209 was additionally purified by RP-HPLC using Discovery BIO Wide Pore C5-5 column (Supelco Analytical) and 9–90% (vol/vol) gradient of AcCN in water, 0.1% TFA, and subsequently lyophilized. Isotope labeling with ¹⁵N or ¹⁵N, ¹³C was achieved by growing the bacteria in M9 medium supplemented with ¹⁵N-NH₄Cl or ¹⁵N-NH₄Cl and ¹³C-glucose, respectively.

p53 stabilized quadruple mutant was produced and purified as described (25, 35).

Analytical SEC.

Analytical gel filtrations were performed using a GE Healthcare Superose 6 10/300GL column with a flow of 0.5 mL/min. One hundred microliters of protein sample at different concentrations were injected. The running buffer contained 20 mM Tris, pH 7.2, 150 mM NaCl, and 0.5 mM TCEP. The chromatography system was coupled to a DAWN HELIOS MALS instrument (Wyatt Technology) and an Optilab rEX refractometer (Wyatt Technology). The measurements of the intensity of the Rayleigh scattering as a function of the angle and differential refractive index were performed simultaneously on-line and used to determine the weight average molar mass of eluted oligomers and protein complexes (36). Data analysis was performed using the ASTRA5 (Wyatt Technology) software.

Isothermal Titration Calorimetry.

The isothermal titration calorimetry (ITC) experiments were performed using a Microcal ITC₂₀₀ instrument (Microcal). The sample cell of the calorimeter was loaded with 207 µL of MDMX solution. The syringe was loaded with 40 µL of p53TAD peptide (14–33), 5,6-FAM-p4 peptide (37) (sequence LTFEHYWAQLTS), MDMX WWW peptide (181–211), or full-length, stabilized quadruple mutant of p53. Titrant and titrand solutions were prepared in a matching buffer (20 mM Tris, pH 7.2, 150 mM NaCl, 0.5 mM TCEP), and titrant concentration was kept 10 times higher than the titrand. All solutions were degassed for 10 min before measurement. Titrations were performed at 20 °C with stirring speed 800 rpm, injection volumes of 2 µL, and a spacing of 120 s. The data were fit using a one-site binding model available in the Origin ITC data analysis software (version 7.0).

NMR Spectrometry.

NMR spectra were recorded at 293 K using Bruker NMR spectrometers (Avance II+ 700 MHz, Avance III 600 MHz, DRX 600 MHz, or Avance 500 MHz) equipped with cryogenically cooled probes. The standard NMR buffer contained 20 mM Tris, pH 7.2, 150 mM NaCl, 5% D₂O, and buffer-exchange before measurement was achieved either by SEC or using Zeba Spin 7K MWCO, 2-mL desalting kit. The spectra were recorded using TROSY (38) or fast HSQC (39) pulse sequences, typically as 160 × 1,024 complex points matrices. Typical sample concentrations were in the range of 20–50 µM for full-length protein constructs and 100 µM for isolated domains constructs. Typical measurement time for TROSY spectrum of full-length MDMX was 20 h. Topspin 3.1 was used for data processing; before Fourier transformation, the free induction decays were zero-filled and apodized using sine-bell squared (QSINE) function. Processed spectra were plotted and visualized using Sparky (University of California, San Francisco). Tryptophan indole resonances were identified from a ¹H-¹⁵N-HSQC spectrum obtained omitting the ¹H refocusing pulse during t1 evolution and measuring the ¹H-¹⁵N coupling constants, which were higher for indole side chains (¹J = ∼99 Hz).

Resonance assignments were obtained manually using ¹⁵N-¹³C–labeled samples, using a standard set of triple resonance (HNCA, CBCACONH, HNCACB, HNCO, HNCACO) (40) and ¹⁵N-NOESY-HSQC experiments (41). ¹H-¹⁵N NOE enhancements were calculated from intensities of cross-peaks in presaturated and nonsaturated spectra acquired at 600-MHz ¹H Larmor frequency, in an interleaved mode, according to ref. 42. The 5-s interscan relaxation delay was set to avoid water saturation.

In Vitro Ubiquitination Assay.

Ubiquitination assays were carried out by a modification of protocol from ref. 43. Reaction buffer contained 40 mM Tris, pH 7.4, 10 mM MgCl₂, 0.6 mM DTT, and 10 mM ATP. Two hundred microliters of reaction mixes were prepared by mixing the following: 80 ng of E1 (UbE1), 4 µg of E2 (UbE2D2), 20 µg of ubiquitin, 6 µg of p53, and (i) 1.4 µg of MDMX, (ii) 1.6 µg of MDM2, and (iii) 1.6 µg of MDM2 plus 1.4 µg of MDMX. All of the proteins were recombinant and produced in-house (except E1 ligase which was purchased from Boston Biochemicals). Reactions were preincubated at 37 °C and started simultaneously by addition of ATP. Aliquots were mixed with 4× LDS sample buffer (Invitrogen) after 15 min. The samples were resolved by denaturing electrophoresis on 4–12% precast Bis-Tris gels (Invitrogen) and subjected to immunoblotting analysis monoclonal anti-ubiquitin MCA1398g (AbD Serotec, Bio-Rad), anti-MDM2 SMP14 (Millipore), anti-MDMX 8C6 (Millipore), and anti-p53 DO-7 (Dako) antibodies.

Analytical Ultracentrifugation.

Sedimentation equilibrium experiments on MDMX were performed using Beckman XL-I analytical ultracentrifuge equipped with an absorbance and interference detection system with an An50Ti rotor and six-sector 12-mm pathlength cells (Beckman Coulter, USA). Data were collected at 10 °C following absorbance at 280 nm at 18,000, 20,000, and 24,000 rpm. The protein concentration was 1.5 µM, and the buffer contained 20 mM Tris, pH 7.4, 500 mM NaCl, and 4 mM TCEP. Scans were collected at 8-h intervals until the equilibrium was reached by repetitive scans. Data were analyzed using the UltraSpin software (http://www.mrc-lmb.cam.ac.uk/dbv/ultraspin2/). Data were fitted to a double-exponential model indicating equilibrium of two components with masses corresponding to that of monomer and dimer, respectively.