Targeting the MDM2/MDM4 Interaction Interface as a Promising Approach for p53 Reactivation Therapy

Discovery of new molecular targets for cancer therapy is a field in which considerable efforts are being employed, to develop new and safe therapeutic agents. The p53 pathway is an attractive target because of its well-documented oncosuppressive function (1, 2). Since the discovery of MDM2 as crucial p53 inhibitor (3, 4), there have been numerous attempts to pharmacologically disrupt the interaction between these two proteins to reactivate p53 in human wild-type TP53 tumors (2). This has led to the development of different classes of molecules that efficiently and specifically interfere with the formation of the MDM2/p53 complexes (i.e., Nutlin; refs. 5–7). Although p53 reactivation is observed in many wild-type TP53 cancer cells exposed to these molecules, only few of them die by apoptosis (7, 8).

The presence of MDM4 (or MDMX; ref. 9), the MDM2 homolog, has provided a rational explanation for the relatively modest p53-dependent apoptotic response observed in these cells (10–12). Indeed, computational and molecular studies provided evidence that the p53 binding domain of MDM2 and MDM4 are sufficiently dissimilar to explain the lack of activity of the MDM2 inhibitors towards MDM4 (13–15).

Importantly, these data confirmed previous genetic evidence that optimal p53 reactivation can only be achieved by targeting both MDM proteins simultaneously (16) and therefore highlighted the need of identifying molecules that target both the p53–MDM2 and p53–MDM4 interactions. Therefore, new efforts have been developed to find out new molecules able to dissociate MDM2 and MDM4 from p53 (2, 17).

Recent genetic evidence indicates that efficient inhibition of p53 requires the formation of MDM2/MDM4 heterodimers. The heterodimers are more efficient than the MDM2 homodimers in promoting ubiquitination and degradation of p53 (18, 19). Heterodimers also efficiently control basal p53 transcriptional activity in vivo, and thus even in the absence of MDM2 E3-ubiquitin ligase activity (20). In line with these data, MDM2 and MDM4 have been found together with p53 at the promoter of some p53 target genes to inhibit its transactivation activity (21). To further support this model, mice that express MDM4 mutants defective in MDM2 binding die during embryonic development (22, 23). These genetic experiments support the view that the heterodimer is required for proper control of p53 activity.

These data have raised the hypothesis that an alternative therapeutic approach for p53 reactivation in tumors is to target the interaction between MDM2 and MDM4 (18, 20). As the ubiquitin ligase activity of the heterodimers depends on the C-terminal RING finger domain of MDM2, and as MDM2 homodimers are still able to function as an E3 ligase (24, 25), we searched for molecules that bind MDM2 by simulating the MDM2 interaction interface of MDM4.

Lyophilized Peptide3 (KEIQLVIKVFIA), Peptide3M (KEIQLVIKVAEA), Peptide SC3A (VQEAFKLIKIVI), and SC3B (AIKIFVKVLEIQ) (synthesized from Biosynthesis) were dissolved in 100% DMSO at a concentration of 10 mmol/L. Peptides purity level was ≥ 96%. Peptides' solution was freshly prepared at a concentration of 0.5 mmol/L in sterile distilled deionized H₂O. Peptides were premixed with growth media at 37°C for 15 minutes before addition to the cells. All peptides were used at 10 μmol/L concentration unless differently specified. All peptides were chemically modified by a capping acetyl group at the NH2- and an amide group at the C-terminus.

Starting from the crystal structure of the heterodimer complex (pdb code: 2VJF; ref. 19), one complex of Pep3 with MDM2 was generated shortening the RING domain of MDM4 (chain B) from residue E428 to K478, while keeping the RING domain of MDM2 (chain A). Pep3M was obtained editing the structure of Pep3 by mutating residues F488 and I489 into alanine and glutamate, respectively. The geometry of the resulting two complexes was optimized using the Protein Preparation Wizard tool as implemented in Maestro v.9.7. (Schrödinger, LLC). Two molecular systems were then prepared inserting Pep3/MDM2 and Pep3M/MDM2 complexes into a cubic box containing TIP3P water molecules for protein solvation, and extended 10 Å away from any protein atom. Each of the systems was neutralized by adding sodium and chlorine ions at a concentration of 0.15 mol/L. Periodic boundary conditions were applied to avoid finite-size effects. Molecular dynamics (MD) simulations were performed using Desmond v.3.0 and the OPLS-2005 force field. The simulation protocol included an initial energy minimization over a maximum of 2,000 steps, with a convergence criterion of 50 kcal/mol/Å, and the presence of harmonic restraints on the solute atoms (force constant = 50.0 kcal/mol/Å2); a second energy minimization without restraints; a third stage of 12 picoseconds (ps) at 10 K with harmonic restraints on the solute heavy atoms (force constant = 50.0 kcal/mol/Å2), using the NVT ensemble and Berendsen thermostat; a fourth 12 ps at 10 K, retaining the harmonic restraints, and using the NPT ensemble and Berendsen thermostat and barostat; a fifth heating phase of 24 ps at 300 K, retaining the harmonic restraints, and using the NPT ensemble and Berendsen thermostat and barostat; a final 24 ps at 300 K without harmonic restraints, using the NPT Berendsen thermostat and barostat. The final production phase of 30 nanoseconds (ns) was run using the canonical NPT Berendsen ensemble at temperature 300 K. During MD simulations, a time step of 2 femtoseconds (fs) was used inserting constrains on the bond lengths of hydrogen atoms with the M-SHAKE algorithm (26). A total of 6252 frames of atomic coordinates were saved along the MD trajectories. The occupancy of intermolecular hydrogen bonds between Pep3, Pep3M, and MDM2 was calculated using a cut-off value of 10%.

Human primary fibroblasts, MCF7, p53^+/+ and p53^−/−HCT116, HepG2, and H1299 cells were maintained in DMEM, GTL-16 in RPMI supplemented with 10% FBS (Gibco). MCF7, HCT116, and GTL-16 cell identity has been confirmed by PowerPlex 18D System (Promega) by BMR Genomics service (last analysis December 2013). Mycoplasma-free conditions have been routinely tested by MycoAlert kit (Lonza). MCF10A cells were maintained in MEBM supplemented with specific condiments according to the manufacturer's instructions (Lonza). p53^−/−Mdm2^−/−Mdm4^−/−MEFs (TKO-MEF) were maintained in DMEM high glucose supplemented with 10% FBS (FBS-Cambrex). Transient transfections were performed by Lipofectamine Plus reagent (Life Sciences) according to the manufacturer's instructions. All stealth siRNAs are from Life Technologies. MDM2 and MDM4 mutants were generated by QuikChange XL Site-Directed Mutagenesis Kit according to the manufacturer's instructions (Stratagene).

Imaging was performed through a confocal laser scanning microscope (Leica SP5, Leica Microsystems) equipped with four laser lines and a transmitted light detector for differential interference contrast (Nomarski) imaging. The following acquisition settings were used: objective 63× oil immersion NA 1.4; zoom factor 2; image format 1024×1024; sampled image area was x: 122 μm, y: 122 μm; pinhole 1 producing an optical section thickness of 0.8 μm; speed acquisition 10 Hz. A random set of single plane images of both fluorescence and bright field channels was taken to analyze the peptide cellular distribution throughout the coverslip. The confocal image acquisitions were performed so that all samples were imaged using the same settings for laser power and detector gain. Brightness and contrast of images were globally enhanced by using linear histogram correction and slightly oversaturated to allow a better qualitative visual evaluation of the morphologic features.

For the Annexin V–propidium iodide staining, cells were collected, washed with PBS, and stained by Annexin V–FITC and propidium iodide according to the manufacturer's protocol (Clontech). Cells were analyzed by FACScan flow cytometer and data analyzed by CellQuest Software (Becton Dickinson). Cell-cycle profiles were evaluated by FACS analyses of 70% ethanol-fixed cells stained with 0.1 mg/mL propidium iodide/PBS and 2 mg/mL RNaseA solution. Apoptotic DNA fragmentation analysis was carried out by terminal deoxynucleotidyltransferase–mediated dUTP nick end labeling (TUNEL) assays with fluorescent In Situ Cell Death Detection Kit (Roche) following the manufacturer's instructions.

For immunoprecipitation cells were lysed in Saito modified buffer (50 mmol/L Tris–HCl, pH 7.4, 0.15 mol/L NaCl, 0.5% Triton-X100, 5 mmol/L EDTA) containing protease inhibitors mix (Roche). Immunoprecipitations were performed by preclearing lysates with protein G-sepharose (Pierce) and then incubated with the indicated antibody, under gentle rocking at 4°C overnight. For Western blot analysis, cells were lysed in RIPA buffer. SDS-PAGEs were transferred onto PVDF membranes (Millipore). Membranes were developed using the enhanced chemiluminescence (ECL Amersham) by chemiluminescence imaging system, Alliance 2.7 (UVITEC Cambridge) and quantified by the software Alliance V_1607. Primary antibodies: anti-MDM4 BL1258 (Bethyl Laboratory), anti-p53 FL393 (Santa Cruz Biotechnology) and PAb421 (Calbiochem), anti-MDM2 2A10 (kindly provided by M.E. Perry) and Ab1 (Calbiochem), anti-GFP (Roche), anti-FITC (Abcam), anti-BIK/NBK (FL-160; Santa Cruz Biotechnology), anti-phospho-histone H2A.X, anti-PARP-1 (Cell Signaling Technology), anti-α-tubulin DM1A (Sigma), anti-actin C-40 (Sigma).

Nuclear and cytoplasmic fractions were prepared as follows: cells scraped with PBS were resuspended in hypotonic lysis buffer (10 mmol/L HEPES pH 7.9, 10 mmol/L KCl, 0.1 mmol/L EDTA, 0.1 mmol/L EGTA) added with protease inhibitors (Roche). After resuspension, NP-40 was added to a final concentration of 0.6% and the nuclei were pelleted by centrifugation at 10,000 rpm for 30 seconds at 4°C. Supernatant was collected as cytoplasmic fraction. Nuclei pellet was resuspended in nuclear extract buffer (20 mmol/L HEPES pH 7.9, 25% glycerol, 0.4 mol/L NaCl, 0.1 mmol/L EDTA, 0.1 mmol/L EGTA), rocked for 15 minutes at 4°C and then recovered by centrifugation at 14,000 rpm for 5 minutes at 4°C. The pellet containing the chromatin enriched fraction was resuspended in nuclear extract buffer and sonicated by Heat Systems UltraSonics Inc Sonicator W-375 Cell Disruptor at 20% of maximum output power for 30 seconds.

MCF7 (250 and 500 cells), and MCF10A (5,000 and 10,000 cells) were plated in complete medium in triplicate in 6, 0, and 35 mm dishes, respectively. Treatments (DMSO 0.1%, Pep3 10 μmol/L, Pep3M 10 μmol/L) were administered every 2 days. Dishes were stained by crystal violet.

CD1-Foxn1^nu 6-week-old male mice were injected subcutaneously with luciferase engineered HCT116 (5 × 10⁶/tumor), or GTL-16 (2 × 10⁶/tumor) in Matrigel 10 mg/mL (dilution 1:1; BD Biosciences) to induce tumor formation. After detection of palpable tumor (tumor volume ≥ 70 mm³), mice were treated with either Pep3 or control Pep3M injected intratumorally at a dose of 10 mg/kg and tumor growth was followed and analyzed by luciferase activity during time. Light emission was detected using the IVIS Lumina II CCD camera system, and the bioluminescent signal was analyzed with the Living Image 2.20 software package (Caliper Life Sciences). All animal experiments were conducted in accordance with institutional guidelines, in the full observation of the Directive 2010/63/UE. GTL-16 cells were infected with lentiviral vector pMA3160 (Addgene) to generate luciferase-expressing cells. HCT116 were transfected with pcDNA-Luc and pBABE Puro and selected using 1 mg/mL puromycin.

Total RNA from HCT116 cells untreated or treated with DMSO, Pep3M, or Pep3 was isolated using RNA Purification columns (Norgen, Purification Plus kit) following the manufacturer's instruction. Biologic triplicates were performed. RNA quality was assessed by Bioanalyser (RIN ≥ 9.6). After in vitro retrotranscription (Applied Biosystem), real-time PCR was performed using ABI 7900 (Applied Biosystem) and SYBR Green Master Mix (Bioline) using the following primers:(i) BIK (human, BIK, PCR product = 70 bp)fw: 5′-GAATGCATGGAGGGCAGTGAC-3′; rev: 5′-GGCTCACGTCCATCTCGTC-3′(ii) PIG3 (human, TP53I3, PCR product = 71 bp)fw: 5′-AATGCAGAGACAAGGCCAGT-3′; rev: 5′-GTCCAGATGCCTCAAGTCCC-3′(iii) GAPDH (human, GAPDH, PCR product = 184 bp)fw: 5′-GAGTCAACGGATTTGGTCGT-3′; rev: 5′-GACAAGCTTCCCGTTCTCAG-3′(iv) p21 (human, p21, PCR product = 65 bp)fw: 5′-CTGGAGACTCTCAGGGTCGAAA-3′; rev: 5′-GGCGTTTGGAGTGGTAGAAATC-3′(v) MDM2 (human, MDM2, PCR product = 88 bp)fw: 5′-ATATACCATGATCTACAGGAACTTGGTAGT-3′;rev:5′-GGTGACACCTGTTCTCACTCACA-3′(vi) TBP (human, TBP, PCR product = 169 bp)fw: 5′-GCACAGGAGCCAAGAG -3′; rev: 5′-GTGGGTGAGCACAAGG-3′. Relative mRNA levels, normalized to TBP, were calculated as follows: 2^{−[ΔC_t(Pep3 or Pep3M)−ΔC_t(DMSO)]} = 2^−ΔΔC_t, where ΔC_t equals C_t (BIK, PIG3, GAPDH, p21, or MDM2) − C_t (TBP).

The gene expression profiling was performed according to the standard Agilent one-color microarray protocol (Agilent Technologies). The cy3-labeled cRNA was hybridized to Agilent 8 × 60K whole human genome chip. Images were acquired by Agilent Scanner; expression data were extracted by Feature Extraction. Data quality filtering and analysis was performed by GeneSpring GX and Microsoft Excel. Microarray functional data analysis was generated through the IPA webtool (27). Clustering analysis was done using MultiExperiment Viewer. Genes associated to specific cancer features were selected on the basis of the annotations in the Cancer Gene Index of the NCI.

HCT116 were untreated or treated with 10 μmol/L Pep3 or Pep3M for 24 hours and then incubated in culture media containing 1% formaldehyde for 10 minutes at room temperature. After formaldehyde neutralization, cells were washed and harvested by ice-cold PBS. Cell lysis was performed in lysis buffer (5 mmol/L PIPES pH 8.0, 85 mmol/L KCl, 0.5% NP-40, protease inhibitors) and nuclei were collected and suspended in nuclei lysis buffer (50 mmol/L Tris-HCl pH 8.0, 10 mmol/L EDTA, 1% SDS, protease inhibitors). Nuclei were sonicated by using a Bioruptor sonicator. Extracts were precleared with protein G (Invitrogen), and 50 μg of DNA were used for immunoprecipitation. An aliquot of precleared lysate was taken as input control. Immunoprecipitation was carried out with anti-p53 sheep polyclonal antibody (Ab-7 Calbiochem), conjugated to paramagnetic protein G beads (Invitrogen). Immune complexes were eluted in elution buffer (50 mmol/L NaHCO₃ and 1% SDS) by heating to 65°C for 15 minutes. DNA was purified by using QIAquick PCR Purification Kit (Qiagen). Purified DNA was analyzed by quantitative real-time PCR using SYBR Green ABI7900. The promoter-specific primer sequences used were as follows: PIG3 5′-GATCCCAGGACTGCGTTTTGCC-3′; 5′-GGGAACGAGACCCAACCTCTTG-3′; MDM2 5′-GCAGGTTGACTCAGCTTTTCCTCT-3′; 5′-GTGGTTACAGCCCCATCAGTAGGTA-3′.

HCT116 and MCF7 cells were plated in 96-well plates and treated with Pep3, Pep3M, or DMSO. After 48 or 72 hours, cells were washed with PBS and incubated with 25 μmol/L 2′,-7′-dichlorodihydrofluorescein diacetate (H2DCFDA, Sigma) in serum-free media at 37°C for 30 minutes. Cells were then washed once in PBS and incubated for 3 hours in complete media supplemented with or without 100 μmol/L H₂O₂. The fluorescent signal from fluorescent 2′-7′-dichlorodihydrofluorescein (DCF) was monitored by multimode plate reader Enspire (Perkin Elmer). The obtained results were normalized for the number of cells/well by Cell Titer viability assay (Promega).

MDM2 and MDM4 form heterodimers via association of their C-terminal RING finger domains (28) and the last residues of these domains play a crucial role in this interaction and in the heterodimer activity (24, 25). Analyzing the structural complex of these domains (19), three central β-sheet motifs were observed from each RING domain that are engaged in the formation of a tight β-barrel core at the interface. Particularly, the N-terminal regions of MDM2/MDM4 RING finger domains interact with each other adopting α-helix like structures, while the C-terminal regions contribute to the formation of the β-barrel core through hydrophobic interactions (Fig. 1A). Mutagenesis experiments have previously identified residues of MDM2 RING domain involved in the ubiquitin ligase activity of the complex (24, 29–31). Among these, F490, located at the C-terminus of MDM2, forms a hydrophobic patch at the interface of the complex with residues L433 and I489 of MDM4, and L430 and A434 of MDM2 (Fig. 1A). As hydrophobic patches play key roles in stabilizing protein–protein interactions (32), we investigated the role of L433, I489, L430, and A434 as hot spots of MDM2/MDM4 interaction. MDM2 and MDM4 mutants were generated, by mutating the above residues in either a negatively charged residue or a polar uncharged residue (M2-L430E or M2-L430Q, and M2-A434D or M2-A434N; M4-L433D or M4-L433N and M4-I489E or M4-I489Q, respectively; Fig. 1B).

Expression of these mutants in p53^−/−Mdm2^−/−Mdm4^−/−mouse embryonic fibroblasts (TKO-MEF) and coimmunoprecipitation of MDM2/MDM4 complex revealed that mutation of both MDM4 residues abolishes its binding to MDM2 confirming that they are essential for the binding of the two proteins (Fig. 1C). Conversely, the association of MDM2 mutants to MDM4 was dependent on the replaced residue, with charged residues (L430E and A434D) and polar uncharged residues (L430Q and A434N) abolishing and reducing the binding to MDM4, respectively (Fig. 1D). Collectively, these observations suggest that the hydrophobic interactions promoted by the side chains of L433 and I489 from MDM4 are important for the stabilization of the RING domain complex. Conversely, MDM2/MDM4 complexes tolerate the replacement of L430 and A434 from MDM2 with polar uncharged residues to some extent, though not standing the presence of charged residues. Noteworthy, the ubiquitinating activity of all mutants but MDM2L430Q, is abolished (Supplementary Fig. S1A), indicating that these residues are important for MDM2/MDM4 heterodimer formation as well as for MDM2 activity. In fact, all these mutants still bind p53, indicating that loss of p53 ubiquitination is due to impairment of MDM2 activity and that the folding of the MDM2 and MDM4 proteins is not affected by the point mutations introduced, at least at the levels of the p53-binding domain (Supplementary Fig. S1B and S1C).

From previous data, different peptides that mimic the MDM2 interaction interface of MDM4 were designed. Peptide-3 (thereafter Pep3), a dodecapeptide that overlaps the COOH-terminus of MDM4 comprising the I489 residue (KEIQLVIKVFI₄₈₉A), was able to reduce p53 ubiquitination in different cell lines as compared with untreated or solvent (0,1% DMSO) treated cells (Fig. 2A and Supplementary Fig. S1E). In comparison, a control peptide in which the I489 and the F488 aminoacids were substituted by E489 and A488, respectively (Pep3M, KEIQLVIKVAEA) did not affect levels of ubiquitinated p53 (Fig. 2A and Supplementary Fig. S1D and S1E). A peptide containing only substitution of I489 retained partial inhibitory activity towards MDM2 (Supplementary Fig. S1E).

Starting from the structural data of MDM2/MDM4 RING domains, MD simulations of Pep3 and Pep3M in complex with the MDM2-RING domain were used to investigate the molecular basis of their binding modes to MDM2. Pep3 is stably anchored to MDM2, with the backbone oxygen and nitrogen atoms engaged in a pattern of stable hydrogen bonds with the backbone atoms of MDM2 (Table 1). The side chain of MDM4-I489 stably occupies a hydrophobic pocket defined by residue L458 of MDM2 (Pocket 1, Fig. 2B) along the entire MD trajectory. MDM4-F488 undergoes a conformational change docking its side chain into a second hydrophobic pocket (Pocket 2, Fig. 2B) defined by residue V477 of MDM2. Although Pep3M makes six stable backbone-mediated hydrogen bonds with MDM2 (Table 1), it does not occupy the two hydrophobic pockets (Fig. 2C). Specifically, the side chain of A488 is too short to reach Pocket 2, and the polar side chain of E489 is not suited to occupy the hydrophobic Pocket 1. These data provide a likely explanation as for why Pep3, but not Pep3M, binds to MDM2, with the former peptide matching two hydrophobic patches (Pocket 1 and 2) at the interface of the complex. These data highlight also the role of MDM4-F488 in the interaction of MDM4/Pep3 to MDM2 and explain the partial activity of the peptide-3 mutated in the I489 only. In vitro binding assays confirmed Pep3/MDM2 interaction. Pep3 conjugated to the fluorescein-isothiocyanate group (FITC-Pep3) was indeed able to coimmunoprecipitate MDM2, whereas increasing doses of FITC-Pep3M did not (Fig. 2D).

We next tested the cell-penetrating activity of FITC-Pep3 in vivo. Strikingly, internalization of the peptide in a time and dose-dependent manner was observed upon its addition to the culture medium of the human cancer cell line MCF7 (Fig. 3A and B). Confocal imaging showed an intense FITC-Pep3 staining confined to vesicular structures frequently surrounding the nucleus. In addition, several FITC-Pep3–positive vesicles were observed in proximity to the plasma membrane indicative of an active internalization process of FITC-Pep3. Similar results were obtained for Pep3M (Supplementary Fig. S2A). FACS analysis allowed us to quantify the high proportion of FITC-positive cells (Fig. 3B and Supplementary Fig. S2B). Similar results were obtained with the colorectal cancer cell line HCT116. Interestingly, the FITC-Pep3–positive HCT116 cells showed a significant increase of apoptotic cell death compared with FITC-Pep3M–treated cells (Fig. 3C and Supplementary Fig. S2C). In contrast, Pep3 and Pep3M peptides caused only a limited increase of cell death in HCT116 cells lacking functional p53 (p53^−/−HCT116), indicating that Pep3-induced apoptosis is dependent on the presence of p53 (Fig. 3C). Similar results were obtained when using untagged peptides (Supplementary Fig. S2D). In contrast, scramble peptides (SC3A and SC3B) containing the same aminoacids present in Pep3 but assembled in a different order did not induce cell death (Supplementary Fig. S2E), confirming the specificity of Pep3 activity.

Treatment of HCT116 cells with increasing doses of Pep3 (2, 5, 10, 20, 30 μmol/L) showed dose-dependent cell death reaching a plateau at 20 μmol/L (Fig. 3D). Pep3 also induced apoptosis in wt-p53 GTL-16, MCF7 (Fig. 3E), and HEPG2 (Supplementary Fig. S2F) cell lines. Conversely, it was ineffective in p53-null H1299 (Supplementary Fig. S2G) or in p53-knockdown MCF7 cells (shp53-MCF7) (Fig. 3F). Importantly, Pep3 did not significantly increase cell death in MDM2-knockdown GTL-16 cells (siMDM2-GTL-16), indicating that its activity requires the presence of MDM2 (Fig. 3G and Supplementary Fig. S2H). Pep3 also significantly inhibited colony formation in MCF7 cells in comparison with control cells, suggesting its ability to affect the tumorigenic potential of cancer cells in vitro (Fig. 3H). Importantly, Pep3 did not increase apoptotic cell death nor affected cell-cycle profile or colony forming ability in untransformed MCF10A cells (Fig. 3I and J and Supplementary Fig. S3A). Similarly, it did not increase cell death (by FACS and PARP-1 cleavage analysis) nor affected cell-cycle profile in primary human fibroblasts (Supplementary Fig. S3B–S3D), suggesting that this peptide selectively affects the viability of tumor cells but not of untransformed ones.

To assess the therapeutic potential of Pep3 in vivo, we tested the antitumor effects of Pep3 in xenograft models. CD1-Foxn1^nu mice were injected in both flanks with HCT116, or GTL-16 cells expressing a stable luciferase reporter gene. Once luciferase-positive tumor masses were clearly detected (corresponding to tumor volume ≥ 70 mm³), the mice were randomly divided into two groups and Pep3 or Pep3M injected (10 mg/kg) every other day in one of the two flanks for at least 15 days. Tumor growth was monitored by measuring bioluminescence and quantified as percentage of the bioluminescence of the treated versus untreated contralateral tumor. Whereas Pep3 significantly reduced tumor growth, Pep3M had only a marginal effect (Fig. 3K and L and Supplementary Fig. S4A). The p53 protein showed comparable levels in all samples, suggesting that the gene remained wild-type during the course of the experiment (see Fig. 4C). No significant host toxicity was observed as evaluated by animal weight (Supplementary Fig. S4B). Accordingly, FITC-Pep3 injection in the hind limb muscle of mice showed no diffusion of the fluorescence signal within 24 hours (Supplementay Fig. S4C). Taken together, these data demonstrate that Pep3 effectively suppresses tumor growth in vivo, and confirm that the MDM4/MDM2 interaction interface is a potential molecular target for anticancer therapy.

To understand Pep3 activity, we analyzed the effects of Pep3 treatment at the molecular level. Pep3 exposure caused a marked decrease of full-length PARP-1 (fl-PARP-1), a marker of cell apoptosis, in both MCF7 and HCT116 cells (Fig. 4A and B). Caspase-3 cleavage was not detected, suggesting the involvement of caspase-7 in PARP-1 cleavage (33). Importantly, fl-PARP-1 was significantly decreased also in Pep3-treated but not in Pep3M-treated xenograft tumor samples (Fig. 4C and D), indicating that Pep3-induced suppression of tumor growth is likely caused by in vivo cell apoptosis. Unexpectedly, Pep3-induced apoptosis did not correlate with an increase in p53 steady-state protein levels (Fig. 4A–C). However, analysis of p53 subcellular distribution revealed a significant Pep3-mediated increase in the pool of p53 bound to chromatin at the detriment of its soluble/nucleoplasmic fraction (Fig. 4E). The MDM4 protein was redistributed similarly to p53, whereas MDM2 subcellular distribution was not affected by Pep3 (Fig. 4E). These data are consistent with an alteration of the MDM2/MDM4 nuclear complexes upon Pep3 addition. In fact, association between MDM4 and MDM2 was reduced by Pep3 in the nuclear fraction but not in cytoplasmic extract (Fig. 4F and Supplementary Fig. S5A). Treatment with the proteasome inhibitor MG132 did not exacerbate the Pep3-dependent increase in p53 protein levels, including the chromatin bound fraction (Fig. 4G), confirming that Pep3 interferes with p53 degradation.

The levels of phosphorylated histone H2AX (γ-H2AX), a marker of DNA damage did not increase upon Pep3 exposure (Fig. 4E), indicating that recruitment of p53 to the chromatin is not associated with activation of the DNA damage response. Accordingly, phosphorylation of p53 at Ser46 or Ser15, two features of stress-mediated p53 activation were not observed (Supplementary Fig. S5B and S5C).

Transcriptionally active p53 can induce apoptosis through different target genes (34). Genome-wide transcriptomic analysis of Pep3-treated HCT116 cells compared with untreated, DMSO or Pep3M-treated cells, identified 347 and 198 genes that were significantly deregulated 24 and 48 hours after treatment respectively (Supplementary Tables S1 and S2). Overlapping of the two groups of genes deregulated at 24 and 48 hours revealed a significant alteration of a common subgroup of 49 genes (Fig. 5A and Supplementary Fig. S6). Clusters corresponding to the different treatments were clearly apparent on a principal component analysis (PCA; Fig. 5B).

Ingenuity Pathways Analysis revealed that most of the 347 mRNAs deregulated by Pep3 encode for proteins involved in cell death and survival (Fig. 5C and Supplementary Fig. S6E). Of note, p53 was identified as one of the top putative upstream regulators of these genes confirming that Pep3 induces p53 activity (Fig. 5D).

Among the well-established p53 target genes induced by Pep3 are two genes encoding for proapoptotic and pro-oxidant factors: BCL2 interacting killer, BIK/BP4/NBK/BIP1 (35), and tumor protein p53-inducible protein-3, PIG3/TP53I3 (Fig. 5D; ref. 36). qRT-PCR confirmed the significant induction of these mRNAs at 24 (Fig. 5E) and 48 hours (Supplementary Fig. S6F), whereas, in comparison, p21/WAF1 and MDM2, two p53 target genes, involved mainly in growth arrest, were not induced by Pep3. Chromatin immunoprecipitation (ChIP) confirmed the increased recruitment of p53 at the promoter of PIG3 in cells exposed to Pep3 (Fig. 5F). In contrast, no increased recruitment of p53 at the MDM2 promoter in the same experimental conditions was observed (Supplementary Fig. S7A). Consistent with these findings, BIK protein levels were highly increased upon Pep3 exposure both in MCF7 and HCT116 cells (Fig. 6A and B and Supplementary Fig. S7B). Of importance, siBIK by single oligos or their mix (Supplementary Fig. S8A) resulted in decreased PARP-1 cleavage (Fig. 6B) and Pep3-induced cell apoptosis (Supplementary Fig. S9A), confirming the role played by this gene in Pep3-mediated activity. Given the involvement of BIK and PIG3 in p53-mediated oxidative stress response (36, 37), we quantified reactive oxygen species (ROS) levels using the fluorogenic probe DCF upon H₂O₂ stimulation. This analysis revealed a significant increase of intracellular ROS levels following Pep3 exposure both in HCT116 and MCF7 cells (Fig. 6C and Supplementary Fig. S9B), confirming that Pep3 counteracts the antioxidant potential of the cell. The Pep3-dependent increase of ROS levels in HCT116 cells was detectable even in the absence of H₂O₂ stimulation (Fig. 6C). Importantly, ROS levels were not induced by Pep3 in p53^−/−HCT116 cells (Fig. 6C), confirming the crucial role played by p53 in this Pep3-induced response. Finally, siBIK or siPIG3 (Supplementary Fig. S8) rescued the increase of ROS levels mediated by Pep3 treatment (Fig. 6D), confirming the role played by these genes in this response.

Most of the pharmacologic approaches aiming at reactivating p53 in cancer cells have focused to date on the interaction interface between p53 and MDM2 or MDM4. Here, we developed an alternative strategy that aims at inhibiting the activity of the MDM2/MDM4 complexes by interfering with their heterodimerization. Our study demonstrates that the binding of a peptide mimicking the MDM4 C-terminus tail to MDM2, impairs MDM2-mediated p53 ubiquitination and activates p53-dependent transcription and oncosuppressive activities.

Interestingly, our data indicate that Pep3 may be mainly effective towards a specific subcellular pool of MDM2/MDM4/p53 as evidenced by the alteration of the nuclear complexes but not of the cytoplasmic fraction. Consistently, the p53 transcriptional program activated by Pep3 is restricted to some p53 target genes involved in the oxidative stress response, suggesting a specific control of this function by the MDM2/MDM4 complexes. In fact, MDM2 and MDM4 act as cotranscriptional inhibitors at the p53-responsive elements on the promoter of PIG3, but not of MDM2 or RCHY1/PIRH2 genes (21). In addition, MDM2 mediates histone ubiquitylation and transcriptional repression (38). At present, it is not known whether subcellular MDM2/MDM4 complexes present different features. The conformation assumed by the complexes consequent to DNA binding might contribute to the selective sensitivity of the nuclear heterodimers to Pep3. Alternatively, additional partners and/or post-translational modifications may distinguish the subcellular complexes.

In contrast to the majority of the approaches that target the interaction between p53 and MDM2, our method causes p53-dependent cell death of various cancer cells, supporting the requirement of the heterodimer for efficient control of p53 function. The observation that the preferred response elicited by Pep3 is apoptosis is also in agreement with recent data from our and other labs that reported impairment of MDM4/MDM2 heterodimer function during apoptosis (22, 39–41). Importantly, this observation highlights the potential therapeutic value of targeting the MDM2/MDM4 heterodimers.

Cancer cells are subjected to increased and persistent oxidative stress and accumulating evidence indicates that increasing ROS levels represents an effective approach to induce cancer cell–specific apoptosis (42–44). Accordingly, the p53-reactivating molecule RITA appears to function by ROS-activating mechanism as Pep3 (45). Our data, indicating that Pep3 reduces the antioxidative potential of the cell and enhances ROS levels, further support the targeting of MDM2/MDM4 complexes as a cancer cell–specific efficient therapeutic strategy. The absence of Pep3 activity in untransformed cell line MCF10A and in human fibroblasts is consistent with this hypothesis.

Compared with the current p53-reactivating approaches, the ability of Pep3 to activate the chromatin-enriched fraction of p53 might endow this strategy of reduced toxicity. In fact, in a recent exploratory clinical trial, the p53-activating molecule RG7112, a member of Nutlin family, has shown treatment-related serious adverse events (46), all of which related to haematologic toxicity (47) caused probably by the treatment-derived high levels of p53. Similarly, highly levels of p53 have been observed following the treatment with the unique double MDM2/MDM4 inhibitor developed so far (17).

In addition, given the increased specificity of anticancer peptides in the recognition of their targets (48), Pep3 may overcome side effects and toxicities of other cancer therapies. In fact, small-molecule inhibitors of MDM2-mediated ubiquitinating activity, have shown unintended activities towards other RING finger E3 ligases (49, 50).

On the basis of these data, targeting the MDM2/MDM4 interaction interface appears a promising therapeutic strategy for developing new anticancer lead compounds with high specificity and effectiveness.

No potential conflicts of interest were disclosed.

Conception and design: M. Pellegrino, A. Macchiarulo, F. Moretti

Development of methodology: M. Pellegrino, F. Mancini, R. Lucà, N. Giacchè, I. Manni, E. Teveroni, M. Buttarelli, L. Fici, T. Bruno

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Florenzano, R. Brandi, M. Fanciulli, M. D'Onofrio, G. Piaggio, R. Pellicciari

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Pellegrino, R. Lucà, A. Coletti, I. Arisi, F. Florenzano, M. Buttarelli, J.C. Marine, A. Macchiarulo, F. Moretti

Writing, review, and/or revision of the manuscript: M. Pellegrino, R. Lucà, A. Pontecorvi, J.C. Marine, A. Macchiarulo, F. Moretti

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Pellegrino, R. Lucà

Study supervision: F. Moretti

The authors thank Dr. F. Maina for GTL-16 cells, Dr. Cinzia Rinaldo for the pSuper.retro-ip53 plasmid, and Ada Sacchi for her support.

This work was supported by grants from Associazione Italiana Ricerca sul Cancro (AIRC; IG-8825, IG-12767; F. Moretti), Project “FaReBio di Qualita” from Italian Ministry of Economy and Finance to the CNR, FIRB RBAP1153LS_007 (A. Pontecorvi), and FIRB B81J10001870008_RBAP10L8TY_003-MD, FP7 European-Grant PAINCAGE, grant: 603191 (M. D'Onofrio).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.