D-peptide inhibitors of the p53–MDM2 interaction for targeted molecular therapy of malignant neoplasms

Abstract

The oncoproteins MDM2 and MDMX negatively regulate the activity and stability of the tumor suppressor protein p53, conferring tumor development and survival. Antagonists targeting the p53-binding domains of MDM2 and MDMX kill tumor cells both in vitro and in vivo by reactivating the p53 pathway, promising a class of antitumor agents for cancer therapy. Aided by native chemical ligation and mirror image phage display, we recently identified a D-peptide inhibitor of the p53-MDM2 interaction termed ^DPMI-α (TNWYANLEKLLR) that competes with p53 for MDM2 binding at an affinity of 219 nM. Increased selection stringency resulted in a distinct D-peptide inhibitor termed ^DPMI-γ (DWWPLAFEALLR) that binds MDM2 at an affinity of 53 nM. Structural studies coupled with mutational analysis verified the mode of action of these D-peptides as MDM2-dependent p53 activators. Despite being resistant to proteolysis, both ^DPMI-α and ^DPMI-γ failed to actively traverse the cell membrane and, when conjugated to a cationic cell-penetrating peptide, were indiscriminately cytotoxic independently of p53 status. When encapsulated in liposomes decorated with an integrin-targeting cyclic-RGD peptide, however, ^DPMI-α exerted potent p53-dependent growth inhibitory activity against human glioblastoma in cell cultures and nude mouse xenograft models. Our findings validate D-peptide antagonists of MDM2 as a class of p53 activators for targeted molecular therapy of malignant neoplasms harboring WT p53 and elevated levels of MDM2.

The tumor suppressor protein p53 is a transcription factor that transactivates, in response to cellular stresses, the expression of various target genes that mediate cell cycle arrest, senescence, or apoptosis (1). Dubbed the “guardian of the genome” (2), p53 is critical for maintaining genetic stability and preventing tumor development (3). Not surprisingly, loss of p53 activity resulting from point mutations in the TP53 gene is responsible for approximately 50% of human tumors. Although p53 retains WT status in many other tumors, its tumor suppressor activity and in vivo stability are abrogated by regulatory molecules such as the E3 ubiquitin ligase MDM2 and its homologue MDMX (4, 5). Amplified or over-expressed in a significant fraction of cancers without concomitant TP53 mutation, MDM2 and MDMX directly contribute to p53 inactivation and tumor survival.

MDM2 itself is transcriptionally inducible by p53 in a negative feedback loop (6). MDM2 binds the N-terminal transactivation domain of p53 with high affinity to block p53 regulating responsive gene expression (7). More importantly, MDM2 controls p53 stability by targeting the tumor suppressor protein for ubiquitin-mediated constitutive degradation (8–10). Although MDMX lacks E3 ubiquitin ligase activity, the MDM2 homologue acts as an effective transcriptional antagonist of p53, and nonredundantly impedes p53-induced growth inhibitory and apoptotic responses (4, 5). In addition, MDMX forms heterodimers with MDM2 through their C-terminal RING finger domains, stimulating MDM2-mediated ubiquitination and degradation of p53 and MDMX itself (11–13). The interplay between MDM2 and MDMX confers a robust p53 inactivation in tumorigenesis (14).

Recent studies show that restoring endogenous p53 activity can halt the growth of cancerous tumors in mice through cell type–dependent multiple mechanisms, including apoptosis, senescence, and senescence-triggered innate inflammatory responses (15–17). Thus, antagonists of MDM2 and MDMX that activate the p53 pathway can potentially be developed into a class of therapeutic agents for cancer treatment (14). Much of the current efforts have been focused on combinatorial library search for and structure-based rational design of low molecular weight antagonists of MDM2 (18). Successful examples include a cis-imidazoline analogue, termed nutlin-3, and, a spiro-oxindole–derived compound, termed MI-219 (19, 20). For optimal efficacy, however, dual specific inhibitors may be needed to target both MDM2 and MDMX (14).

We previously reported the synthesis of the p53-binding domain of MDM2 (^25-109MDM2) and its site-specifically biotinylated form using native chemical ligation (21–23). Screening a duodecimal peptide phage library against biotinylated ^25-109MDM2 led to the identification of a potent dual specificity peptide inhibitor termed PMI (p53-MDM2/MDMX inhibitor, TSFAEYWNLLSP). PMI bound to MDM2 and MDMX in their p53-binding cavities with low nanomolar affinities (23). As L-peptide inhibitors are inherently “undruggable” as potential therapeutic agents as a result of their poor in vivo stability, we have applied mirror-image phage display (24, 25) to the discovery of a proteolysis-resistant D-peptide inhibitor of the p53-MDM2 interaction, termed ^DPMI-α (TNWYANLEKLLR), and its mutant ^DPMI-β (TAWYANFEKLLR) (26). ^DPMI-α and ^DPMI-β bound to ^25-109MDM2 at affinities of 219 and 35 nM, respectively, but failed to induce p53-dependent tumor cell death as a result of their inability to traverse the cell membrane. Here we report the identification and characterization of a distinct high-affinity D-peptide inhibitor that binds to the p53-binding pocket of MDM2, and demonstrate the therapeutic efficacy of liposome-encapsulated D-peptide activators of p53 in the treatment of malignant neoplasms harboring WT p53 and elevated levels of MDM2.

Results

Mirror-Image Phage Display.

For mirror-image phage display, we prepared N79K-biotin-D-^25-109MDM2 composed entirely of D-amino acids as described (23, 26). The D-protein was correctly folded and fully functional (SI Appendix, Figs. S1 and S2). We screened the Ph.D.-12 duodecimal peptide phage library against 100 nM N79K-biotin-D-^25-109MDM2 immobilized on streptavidin-agarose resin (as opposed to 1 μM used in the selection of ^DPMI-α). Phage particles bound to the D-protein were competitively eluted with 1 mM D-^15-29p53, and subsequently amplified in host strain Escherichia coli ER2738. After four rounds of selection, 10 plaques were picked at random and their amplified DNAs were sequenced, of which seven binding clones gave rise to the following sequence: DWWPLAFEALLR. A second independent screening experiment confirmed this result (SI Appendix, Fig. S3).

Functional Validation of ^DPMI-γ as a p53 Activator.

We quantified the binding affinity of the D-enantiomer of the phage-selected peptide, termed ^DPMI-γ (H-^DD^DW^DW^DP^DL^DA^DF^DE^DA^DL^DL^DR-amide), for ^25-109MDM2 and ^24-108MDMX using competition surface plasmon resonance (23, 27, 28) (SI Appendix, Fig. S4). ^DPMI-γ effectively competed with immobilized ^15-29p53 for MDM2 binding, yielding a K_d value of 53 ± 6 nM. Despite being 16-fold less active than PMI (K_d = 3.2 nM) (29), ^DPMI-γ is fourfold stronger than ^DPMI-α (26). Interestingly, the binding affinity of ^DPMI-γ for ^24-108MDMX (K_d = 4.9 μM) is nearly three orders weaker than that of PMI (K_d = 8.5 nM), a surprising finding in light of the fact that phage-selected peptide ligands of MDM2 are generally good binders of MDMX, albeit less potent for the latter (28).

Structural Validation of ^DPMI-γ as a p53 Activator.

To further understand the structural basis for D-peptide inhibition of the p53-MDM2 interaction (26), we determined the cocrystal structure of synthetic ^25-109MDM2 and ^DPMI-γ at 1.93 Å resolution (SI Appendix, Table S1 and Figs. S5–S8). As shown in Fig. 1A, ^DPMI-γ adopts a left-handed helical conformation: a C-terminal regular α-helix (from ^DLeu5 to ^DLeu11) preceded by an N-terminal helical turn. The docking of ^DPMI-γ into the p53-binding cavity of MDM2 is mediated primarily by extensive hydrophobic interactions involving ^DTrp2, ^DTrp3, ^DPro4, ^DPhe7, ^DLeu10, and ^DLeu11 of the D-peptide. Additional contributions to the ^DPMI-γ-MDM2 interaction come from two H-bonds between ^DTrp3 N^ε1 and Gln72 O (2.7 Å) and between ^DLeu11 O and Tyr100 O^η (2.6 Å; Fig. 1B). However, the energetic significance of water-mediated contacts between ^DTrp2 of ^DPMI-γ and Gln59 of MDM2 is less certain as the water molecule is absent from the other complex structure in the asymmetric unit. Overall, the structural findings validate the mode of action of ^DPMI-γ as a D-peptide activator of p53.

Fig. 1.

Crystal structure of the p53-binding domain of MDM2 in complex with ^DPMI-γ. (A) ^DPMI-γ bound in the hydrophobic pocket of MDM2. The side chains colored in green, ^DTrp3, ^DPhe7, and ^DLeu11, occupy topologically equivalent positions of Phe3, Trp7, and Leu10 of PMI or Phe19, Trp23, and Leu26 of p53. The electrostatic potential displayed on the molecular surface of ^25-109MDM2 is colored red for negative, blue for positive, and white for apolar. (B) A ribbon drawing of the MDM2-^DPMI-γ complex interface. Contact residues of MDM2 and ^DPMI-γ are shown as ball-sticks, and hydrogen bonds as black dashes. (C) ^DPMI-γ (yellow) and PMI (orange) bound to superimposed MDM2. Residues forming the binding pockets of ^DTrp3/Phe3, ^DPhe7/Trp7, and ^DLeu11/Leu10 are shown as ball sticks in cyan (PMI-bound MDM2) and gray (^DPMI-γ–bound MDM2). (D) Superimposed partial structures of the MDM2-^DPMI-γ (gray/yellow) and MDM2-PMI (cyan/orange) complexes. Phe55, Gln59, and Met62 of MDM2 reorient (from cyan to gray) to form the binding pockets of ^DTrp2 and ^DLeu10 of ^DPMI-γ.

^DPMI-γ and the previously identified PMI differ not only in stereochemistry but also in amino acid sequence. In fact, the two peptides only share a low sequence similarity (25%) manifested by ^DTrp3, ^DPhe7, and ^DLeu10 in ^DPMI-γ versus Phe3, Trp7, and Leu10 in PMI. In the PMI-MDM2/MDMX complex (23), the hydrophobic triad Phe3/Trp7/Leu10 of the peptide (equivalent to Phe19/Trp23/Leu26 of p53; ref. 30) makes critical contributions to MDM2/MDMX binding (28). The residues in ^DPMI-γ topologically identical to the hydrophobic triad of PMI or p53 are ^DTrp3/^DPhe7/^DLeu11 (Fig. 1C). These three D-residues collectively contribute approximately 56% of the total buried surface area of the D-peptide in the complex, whereas ^DTrp2, ^DLeu10, and ^DPro4 contribute much of the remaining binding energy, accounting for approximately 33% of the total buried surface area (SI Appendix, Fig. S8). Significant differences exist between PMI and ^DPMI-γ in MDM2 binding despite nearly identical overall structures of peptide-bound MDM2 [rmsd(Cα) = 0.5 Å; Fig. 1D]. ^DPMI-γ shifts approximately 3.5 Å toward the α2 helix of MDM2 compared with PMI, establishing through the side chains of ^DTrp2 and ^DLeu10 hydrophobic interactions with the α2 helix. To accommodate ^DTrp2 and ^DLeu10, the side chains of Met62, Phe55, and, to a lesser extent, Gln59 of MDM2 flip away from the D-peptide ligand, forming two new binding clefts for ^DTrp2 and ^DLeu10 seen in Fig. 1A. This conformational change enables Phe55 to stack its aromatic side chain against the indole ring of ^DTrp2 and the aliphatic side chain of ^DLeu10. Aromatic-sulfur interactions are also evident between ^DTrp2 and Met62. Notably, the swing of Met62 by 2.4 Å upon ^DPMI-γ binding leaves the two binding pockets of ^DTrp2 and ^DTrp3 intimately coupled.

Structural Comparison of ^DPMI-γ with ^DPMI-α.

Different selection stringencies resulted in two distinct D-peptide ligands for MDM2, ^DPMI-γ (this work), and ^DPMI-α (26). As shown in Fig. 2, ^DPro4 in ^DPMI-γ creates a helix-breaking kink, allowing the selection of ^DTrp2, which makes extensive hydrophobic contacts with the depression lined by Phe55, Leu57, Gln59, Ile61, and Met62 of MDM2. By contrast, the selection of ^DTyr4 in place of ^DPro4 in ^DPMI-α necessarily restricts the bending of the N terminus, leaving the same depression on MDM2 unoccupied. However, as the binding affinity of ^DPMI-γ for MDM2 is nearly identical to that of L7F-^DPMI-α (K_d = 60 nM) (26), the energetic advantage of the selection of ^DPro4 and ^DTrp2 in ^DPMI-γ appears rather limited. This is caused at least in part by the loss in the complex of ^DPMI-γ-MDM2 of electrostatic interactions involving the side chains of ^DGlu8 and Lys94 and His96 of MDM2, which are present in ^DPMI-α-MDM2, counteracting the energetic gain of ^DPro4 and ^DTrp2 in ^DPMI-γ (Fig. 2).

Fig. 2.

Comparison between ^DPMI-γ (present study) and ^DPMI-α (26). (Upper) Close-up view of ^DPMI-γ (yellow) and ^DPMI-α (green) in a Cα ribbon diagram bound to the p53-binding pocket of MDM2. Only the hydrophobic side chains of ^DPMI-γ and ^DPMI-α are shown on the molecular surface of MDM2, of which the three residues docking in topologically equivalent positions of Phe3, Trp7, and Leu10 of PMI are colored red. (Lower) Stereo drawing of a superposition of binding interfaces of MDM2-^DPMI-γ (gray/yellow) and MDM2-^DPMI-α (orange/green). Only contact residues are shown and depicted as sticks. The MDM2 structures can be superimposed with an average rmsd of 1.1 Å between equivalent Cα atoms.

Mutational Data in Support of Structural Findings.

To further verify the structural findings, we prepared the following D-Ala-scan analogues of ^DPMI-γ: W2A, W3A, W2A/W3A, P4A, F7A, and L11A, and quantified their binding affinities for ^25-109MDM2 using the SPR-based competition assay as described (23, 29) (SI Appendix, Fig. S9). The Pro4 kink in ^DPMI-γ was functionally important as the P4A mutation caused a decrease in ^DPMI-γ binding to MDM2 by almost two orders of magnitude. Phe7 of ^DPMI-γ, like its topologically equivalent Trp7 of PMI or Trp23 of p53, contributed the greatest energy to MDM2 binding as the F7A mutation weakened the affinity of ^DPMI-γ for MDM2 by 500-fold. The mutational effect of L11A on ^DPMI-γ—an increase in K_d by nearly two logs—was largely in line with that of L10A on PMI or L26A on p53 with respect to MDM2 binding (28). The most interesting finding, however, was the nonadditive effect displayed by the N-terminal Trp pair of ^DPMI-γ. Individually, each ^DTrp residue contributed modestly to MDM2 binding as evidenced by an approximate 10-fold increase in K_d upon Ala substitution for ^DTrp2 or ^DTrp3. In contrast, the W2A/W3A double mutation caused a dramatic reduction in binding affinity of ^DPMI-γ for MDM2 by almost four orders of magnitude. Interacting side chains and conformational changes are known sources of nonadditivity of mutational effects in proteins (31, 32). The very large nonadditive effect displayed by ^DTrp2 and ^DTrp3 likely stemmed from the coupling of their respective binding packets as indicated by the structural analysis.

Cytotoxicity of ^DPMI-γ or Lack Thereof.

Unlike its L-enantiomer, ^DPMI-γ is fully resistant to proteolysis (SI Appendix, Fig. S10). Despite its significantly higher binding affinity for MDM2 compared with nutlin-3 (23), ^DPMI-γ failed to show any cytotoxicity against both HCT116 p53^+/+ and HCT116 p53^−/− cell lines at up to 50 μM (SI Appendix, Figs. S11 and S12), presumably because of its inability to traverse the cell membrane. By contrast, progressive killing of HCT116 p53^+/+ cells by nutlin-3—a hallmark of p53-induced, time-dependent apoptosis—became evident 2 d after the treatment and leveled off at d 4, yielding an IC₅₀ value of approximately 2 μM, in excellent agreement with various published results (19, 20). As expected, nutlin-3 killing of HCT116-p53^−/− cells was time-independent and required significantly higher concentrations for efficacy (IC₅₀ of approximately 50 μM) (19, 20). When conjugated via its C terminus to a cluster of nine ^DArg residues (^DPMI-γ-^DR9), the D-peptide rapidly (within a few hours) and quantitatively killed both cell types at 6.25 μM (IC₅₀ of approximately 3–4 μM) in a p53- and time-independent fashion. Flow cytometric analysis suggests that ^DPMI-γ-^DR9 killed HCT116 cells primarily through the mechanism of necrosis rather than p53-dependent apoptosis (SI Appendix, Fig. S13).

Liposome-Encapsulated ^DPMI-α Kills Human Glioblastoma Cells in a p53-Dependent Manner.

To achieve intracellular delivery of D-peptides, we used liposomes as carriers to encapsulate the cargo (SI Appendix, Fig. S14). ^DPMI-α was used for this study because of its efficient encapsulation despite being less potent than ^DPMI-γ in MDM2 binding. As integrin αvβ3 is highly expressed on the surface of glioblastoma tumor cells (33, 34), liposomes were also decorated via a PEG spacer with a cyclic RGD peptide (RGD^DYK) to facilitate tumor-specific targeting. As a negative control, RGD-coated liposomes were loaded with ^LPMI-α—the L-enantiomer of ^DPMI-α—whereas doxorubicin, a DNA-intercalating molecule widely used in cancer chemotherapy, was encapsulated as a positive control.

RGD-liposome-^DPMI-α induced a dose-dependent growth inhibition of human glioma cell line U87 (WT p53) with an IC₅₀ value of 1.9 μM as determined by the MTT cell viability assay (Fig. 3A)—more effective than nutlin-3 (IC₅₀ of 3.8 μM). As expected, liposome-^DPMI-α without RGD coating (IC₅₀ of 8.3 μM) was significantly less active than RGD-liposome-^DPMI-α. Both free ^DPMI-α and RGD-liposome-^LPMI-α showed little inhibitory activity at 100 μM and 25 μM, respectively—the highest concentrations used—suggesting that the observed growth inhibition of U87 resulted solely from intracellular ^DPMI-α. In sharp contrast, human glioma U251 cells (mutant p53) were largely resistant to the treatment of RGD-liposome-^DPMI-α (SI Appendix, Fig. S15), suggesting that ^DPMI-α functioned in a p53-dependent manner.

Fig. 3.

(Upper) Growth inhibition of U87 cells by Nutlin-3, free ^DPMI-α, liposome-^DPMI-α, RGD-liposome-^DPMI-α, and RGD-liposome-^LPMI-α as determined by the standard MTT cell viability assay. The inhibition curves are averages of three independently measurements. (Lower) Western blot analysis of p53, MDM2, p21, and β-actin expression in U87 and U251 cells 12 h after treatment with indicated concentrations of RGD-liposome-^DPMI-α.

MDM2 itself is transcriptionally inducible by p53 in a negative feedback loop essential for maintaining low levels of p53 during normal development (35, 36), whereas the p53 responsive gene p21 encodes a cyclin-dependent kinase inhibitor that regulates cell cycle progression (37–39). To further investigate the mechanisms of action of intracellular ^DPMI-α, we analyzed the expression of MDM2 and p21 in U87 and U251 cells by Western blotting. As shown in Fig. 3B, 12 h after treatment with RGD-liposome-^DPMI-α, dose-dependent induction of p53, MDM2, and p21 became evident in U87 cells. By contrast, basal levels of (mutant) p53 were maintained and no expression of MDM2 and p21 was detected in U251 cells. These findings strongly support that intracellular ^DPMI-α inhibited U87 cell growth by reactivating the p53 pathway.

Liposome-Encapsulated ^DPMI-α Inhibits Glioblastoma Growth and Prolongs Animal Survival in Nude Mouse Xenograft Models.

To examine therapeutic efficacy of RGD-liposome-^DPMI-α on U87 xenografts in vivo, we used two different brain tumor models based on sites of inoculation—s.c. and intracranial. In the first animal model, primary tumors were established s.c. (in 2 wk) in 24 BALB/c nude mice, followed by a 2-wk treatment regimen involving i.v. injections via tail vein into four randomly divided groups of mice (n = 6) of multiple doses of saline solution, RGD-liposome-doxorubicin at 2.5 mg/kg, or RGD-liposome-^DPMI-α at 7.5 mg/kg (high) and 3 mg/kg (low). The volumes of tumors were measured every other day. As shown in Fig. 4A, the tumor size in animals of the untreated group increased by fivefold over a period of 2 wk. Doxorubicin treatment was effective as a positive control. RGD-liposome-^DPMI-α at low dose was effective in slowing tumor growth compared with mock treatment (P < 0.05). Remarkably, the high-dosage group showed tumor growth inhibition and no signs of weight loss were observed (P < 0.01).

Fig. 4.

(A) Growth inhibition of U87 s.c. xenografts in nude mice by ^DPMI-α and doxorubicin. An independent-samples t test (α = 0.05, two-tailed) was used to calculate the P values for statistical significance: P = 0.014 for low-dose ^DPMI-α/saline solution and P = 0.001 for high-dose ^DPMI-α/saline solution. (B) Kaplan-Meier survival curves for intracranial glioblastoma-implanted nude mice treated with ^DPMI-α and doxorubicin. The P values for statistical significance are P = 0.000 for high-dose ^DPMI-α/saline solution, P = 0.007 for low-dose ^DPMI-α/saline solution, and P = 0.738 for high-dose ^DPMI-α/doxorubicin. (Inset) In vivo near-infrared fluorescence imaging of U87 intracranial xenografts in nude mice 2 h after injection of 200 μL of DiR encapsulated cyclic-RGD liposome via tail vein (DiR, 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindotricarbocyanine iodide).

In the second animal model, four groups of intracranial glioblastoma-bearing mice (n = 6) were treated via tail vein with multiple doses of saline solution, RGD-liposome-doxorubicin at 5 mg/kg, or RGD-liposome-^DPMI-α at 10 mg/kg (high) and 4 mg/kg (low), with death as an endpoint. Mice treated with RGD-liposome-^DPMI-α survived significantly longer than the untreated mice (P < 0.05 for low-dosage group; P < 0.001 for high-dosage group; Fig. 4B). The average survival time of untreated mice was 22.5 d, and prolonged to 25.5 d upon ^DPMI-α treatment at low dose and to 28 d at high dose. Of note, the doxorubicin-treated positive control mice survived an average of 28.5 d. To evaluate tumor-targeting and delivery efficiency, mice were mock-treated with fluorophore-encapsulating RGD-liposomes 2 wk after intracranial inoculation of U87 cells. In vivo imaging clearly established the accumulation, within 2 h of treatment, of the fluorescent cargo at the site of the brain tumor (Fig. 4B Inset).

Discussion

In mirror-image phage display (24, 25) phage-expressed peptide libraries are screened against the D-enantiomer of a native protein of interest, yielding an L-peptide ligand that binds specifically to the D-protein. After enantiomeric inversion, the resultant D-peptide ligand, for reasons of symmetry, binds specifically to the native L-protein with the same affinity. Mirror-image phage display takes advantage of vast biodiversity presented by a phage library and affords an elegant and powerful tool for the discovery of potent and proteolysis-resistant D-peptides for therapeutic applications (24, 25, 40, 41). As total chemical synthesis is the only method to generate D-proteins for phage library screening, mirror image phage display is limited to the protein targets that are chemically accessible, and, thus, has rarely been practiced in drug discovery. Kent and colleagues pioneered native chemical ligation, a revolutionary synthetic methodology that enables two or more fully unprotected synthetic peptides to react chemoselectively in aqueous solution, yielding a longer polypeptide chain linked in native peptide bonds (21, 22). As its debut, native chemical ligation has been widely used to chemically synthesize domain-sized proteins with high efficiency (42). Our work reported here further showcases the power of a combination of native chemical ligation and mirror image phage display through the identification of high-affinity D-peptide antagonists of MDM2 as p53 activators.

The D-peptide antagonists of MDM2 are resistant to proteolysis, and their therapeutic value hinges on the efficiency of peptide cellular uptake by tumor cells. Arg-rich cell penetrating peptides such as the protein transduction domain sequence from the nuclear transcription activator protein (Tat) encoded by HIV-1 are capable of promoting efficient cellular uptake of covalently attached proteins and peptides to the cytoplasm and nucleus of many cell types (43–45). However, ^DPMI-γ-^DR9 killed both HCT116 cell types independently of p53 status. This finding is not entirely surprising as p53-like peptides conjugated to cationic cell-penetrating peptides have previously been shown to induce necrosis of tumor cells without p53 activation (29, 46–48). Conjugation of the hydrophobic peptide ^DPMI-γ to a cluster of Arg residues likely generated a detergent-like molecule that would be indiscriminately toxic against all cell types.

The apparent failure of ^DPMI-γ-^DR9 to preferentially kill HCT116 p53^+/+ cells in a p53-dependent manner demonstrates the limited value of cationic cell-penetrating peptides in delivering hydrophobic cargo, and underscores the need to develop alternative clinically viable cellular delivery vehicles for therapeutic peptides and proteins. Particularly noteworthy in this regard is the hydrocarbon stapling technique developed by Verdine and colleagues, which enables side-chain cross-linked α-helical peptides to actively traverse the cell membrane with enhanced biological activity and proteolytic stability (49). Several hydrocarbon-stapled peptides that antagonize intracellular proteins have been designed to inhibit tumor growth in cell cultures and/or experimental animals (50–52). It is conceivable that a hydrocarbon-stapled D-peptide antagonist of MDM2, when properly designed, may exert robust p53-dependent antitumor activity in vivo as a result of its full resistance to proteolytic degradation. Our work reported here validates liposomes as therapeutically viable D-peptide carriers for targeted molecular therapy.

Glioblastoma multiforme (GBM) is the most common and most malignant type of brain cancer and has an extremely poor prognosis (53). GBM remains very difficult to treat in part because of severe side effects associated with and resistance to conventional genotoxic chemotherapies compounded by the blood–brain barrier to many drugs. New therapies are needed that exploit aberrant activation or suppression of cellular signal transduction pathways impaired by underlying genetic abnormalities in GBM. A majority of GBM harbors WT p53 concurrent with amplification or overexpression of MDM2 and MDMX or deletion of the MDM2 inhibitor ARF (54). Antisense anti-MDM2 oligonucleotides have been shown to inhibit glioblastoma growth in vitro and in vivo in a p53-dependent manner (55). We have demonstrated that D-peptide antagonists of MDM2, when encapsulated in liposomes decorated with an integrin-targeting cyclic-RGD peptide, exert potent growth inhibitory activity against human glioblastoma in experimental animals. In vitro mechanistic studies suggest that ^DPMI-α inhibits tumor growth by reactivating the p53 pathway. Our work validates D-peptide antagonists of MDM2 as a unique class of p53 activators with potential therapeutic value in the treatment of not only GBM, but many other malignant neoplasms harboring WT p53 and elevated levels of MDM2 as well.

Of note, much of the current efforts have been focused on the development of small-molecule antagonists of MDM2 and/or MDMX for p53-dependent anticancer therapy, and several of them are in early clinical trials (14, 19, 20). D-peptides as therapeutic agents are attractive in view of their ability to bind and antagonize target proteins with high affinity and unsurpassed specificity; their resistance to proteolytic degradation in vivo should translate into excellent bioavailability. To achieve optimal therapeutic efficacy of ^DPMI-α, ^DPMI-β, or ^DPMI-γ, it may be necessary to (i) improve their binding affinity for MDM2 and (ii) design advanced peptide drug carriers to further augment cellular delivery (and release) efficiency and to enhance tumor-targeting specificity.

Materials and Methods

Crystallization, Data Collection, Structure Determination, and Refinement.

Crystallization screening of the ^DPMI-γ-^25-109MDM2 complex was conducted at room temperature using the hanging-drop vapor diffusion method and commercial crystallization matrices (Hampton). The best crystals were grown upon mixing 1 μL of ^DPMI-γ-^25-109MDM2 at approximately 10 mg/mL in 10 mM Hepes buffer, 0.1 mM TCEP, pH 7.5, with 1 μL of 0.05 M potassium phosphate monobasic solution containing 20% PEG 8,000. X-ray diffraction data were collected with a Raxis-4++ image plate detector mounted on a Rigaku-MSC Micromax 7 generator (at the x-ray crystallography core facility at University of Maryland, Baltimore, MD). Data integration and scaling, and structure solution and refinement were performed as described (23). The previously determined structure of ^25-109MDM2 in complex with PMI (3EQS) was used as a search model for molecular replacement. Data collection and refinement statistics are summarized in Table S1 in the SI Appendix. The coordinates and structure factors have been deposited in the PDB with accession code 3IWY. Molecular graphics were generated using Pymol (http://pymol.org).

MTT Cell Viability Assay.

Human glioblastoma cancer cell lines U87 and U251 were cultured in DMEM supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin at 37 °C with 5% CO₂ under fully humidified conditions. Cells were seeded at 3,000 cells/well in 96-well plates and allowed to grow for 24 h before treatment. Spectroscopic readings were taken 3 d after treatment, and percent cell viability was calculated on the basis of optical density values of sample wells versus reference wells.

Western Blot Analysis.

The protein fraction of cell lysates was resolved by 10% SDS/PAGE before membrane transfer. Primary antibodies were from Bioworld Technology (p53 and MDM2), Santa Cruz Biotechnology (p21), and Sigma-Aldrich (β-actin); secondary antibodies conjugated with horseradish peroxidase were from Bio-Rad Laboratories.

Growth Inhibition of Glioblastoma in Xenograft Models.

Male BALB/c nude mice, 6 to 8 wk old, were inoculated with 6× 10⁶ U87 cells (in 200 μL of media) in the s.c. tissue of the right shoulder blade. Tumor volumes were calculated on the basis of caliper measurements using the following formula:

Treatment of four groups of mice (six mice per group) was initiated (i.e., time 0) after the tumor had been established in 2 wk as a palpable mass (50–120 mm³ in size). RGD-liposome-doxorubicin was administered at a dose of 2.5 mg/kg on d 1, 6, and 11. RGD-liposome-^DPMI-α was administered at high (7.5 mg/kg) and low (3 mg/kg) doses on d 1, 3, 5, 8, 10, and 12. Saline solution was used as a negative control. For intracranial tumor establishment, male BALB/c nude mice, 6 to 8 wk old, were anesthetized and slowly injected with U87 cells (5× 10⁵ cells suspended in 5 μL PBS solution) into the right striatum (1.8 mm lateral and 0.6 mm anterior to bregma at 3 mm depth). Four groups of intracranial U87-bearing mice (six mice per group) were treated with saline solution; RGD-liposome-^DPMI-α at high (10 mg/kg) and low (4 mg/kg) doses on d 6, 8, 10, 12, 14, 16, 18, and 20 after inoculation; and RGD-liposome-doxorubicin at 5 mg/kg on d 6, 10, 14, and 18 after inoculation.