MDM2 oligomers: antagonizers of the guardian of the genome

Abstract

Over two decades of MDM2 research has resulted in the accumulation of a wealth of knowledge of many aspects of MDM2 regulation and function, particularly with respect to its most prominent target, p53. For example, recent knock-in mouse studies have shown that MDM2 heterooligomer formation with its homolog, MDMX, is necessary and sufficient in utero to suppress p53 but is dispensable during adulthood. However, despite crucial advances such as these, several aspects regarding basic in vivo functions of MDM2 remain unknown. In one such example, although abundant evidence suggests that MDM2 forms homooligomers and heterooligomers with MDMX, the function and regulation of these homo- and heterooligomers in vivo remain incompletely understood. In this review, we discuss the current state of our knowledge of MDM2 oligomerization as well as current efforts to target the MDM2 oligomer as a broad therapeutic option for cancer treatment.

INTRODUCTION

The most commonly mutated gene in cancer is TP53,^1,2 which encodes the transcription factor p53. Involved in a multitude of stress responses, p53 upregulates the transcription of genes involved in cell cycle arrest, apoptosis, metabolism and DNA repair.^3-7 In almost all cancers, p53 signaling is disrupted, either directly through mutations that impair p53-directed transcription or indirectly through the mutation of genes involved in p53 regulation. Interestingly, many tumor-associated point mutations in p53 result in gain-of-function effects that result in worse patient outcomes.^8,9 The frequency of p53 signaling disruption in cancer illustrates the importance of normal p53 function for tumor suppression. Indeed, when considered in the context of numerous p53 mutant mouse models and the biology of human tumorigenesis, one can easily grasp why p53 has long been dubbed the ‘guardian of the genome’.¹⁰

One of the best known regulators of p53 is MDM2, which is an E3 ubiquitin ligase capable of ubiquitinating p53, thereby marking p53 for proteasomal degradation.^11-13 MDM2 has also been shown to inhibit p53 transcriptional activity directly through binding to the transactivation domain of p53.^14,15 MDM2 remains the most experimentally consistent ubiquitously expressed ubiquitin ligase that can effect the degradation of p53 in vitro and in vivo. Therefore, although p53 has been referred to as the ‘guardian of the genome’, MDM2 can be thought of as the guardian of the ‘guardian of the genome’. MDM2 is frequently overexpressed in several types of tumors (many of which harbor wild-type (WT) p53),^16-20 which has led to increased interest in developing drugs that inhibit MDM2 activity to stabilize and activate p53.²¹ However, many drugs that have shown promise in preclinical models have failed to translate into therapeutically effective drugs. One factor that could contribute to the eventual success of p53-activating drugs is a more comprehensive understanding of MDM2 function. For example, although we know that MDM2 requires the ability to oligomerize to efficiently ubiquitinate p53,^22-24 understanding the mechanics, function and regulation of MDM2 oligomerization could be useful to optimize the potency, specificity and synergy of p53-activating drugs. In this review, we discuss our current understanding of MDM2 function with an emphasis on evidence of MDM2 oligomerization and function in p53 regulation. We discuss the importance of the MDM2–MDMX (also known as MDM4) heterooligomers and the evidence of MDM2 homooligomer function. Although MDMX shares many structural features with MDM2, it is less understood and less studied than MDM2 (12-fold more MDM2 publications than MDMX accessible on PubMed). We further speculate on potential models that could help explain in vivo MDM2 behavior. Finally, we end with a discussion of efforts to target MDM2 oligomers to restore p53 activity in tumors.

BACKGROUND

MDM2 is a member of the RING (Really Interesting New Gene) domain-containing E3 ligase family and contains at least three distinct regions that are highly conserved and critical to its function as an E3 ligase for p53. These three domains include an N-terminal p53-binding domain, a central acidic/zinc finger domain, and a C-terminal RING domain (Figure 1). The p53-binding domain, which resides within the first ~ 100 amino acids, is necessary for substrate recognition and transcriptional inactivation of p53.^14,25,26 The isolated MDM2 p53 binding pocket, which minimally involves amino acids 25–108,²⁷ appears to be sufficient to bind p53, at least in vitro.²⁶ The central acidic domain (AD), which can be considered to include the central zinc finger domain (residues 300–326), has been largely characterized as a regulatory domain; the AD binds to several small proteins that can inhibit MDM2 and stabilize p53. The AD and adjacent intervening sequence can also be post-translationally modified by various kinases, including ATM (ataxia telangiectasia mutated),²⁸ ATR (ATM-related),²⁹ c-Abl,³⁰ Akt,³¹ casein kinase 1δ³² and casein kinase 2,³³ which serves as another way (or a prerequisite) to achieve MDM2 enzymatic regulation (Figure 1). Although less well studied, some of these kinases also modify residues on the MDM2 homolog MDMX, which could contribute to the regulation of MDM2 function and p53 stability.^34-36 A summary of the basic ways in which MDM2 can be inhibited through the AD is shown in Figures 1 and 2. Finally, the RING domain is responsible for the heterooligomerization of MDM2 with its homologous partner MDMX as well as for homooligomerization with other MDM2 molecules. The RING domain also contains the enzymatic activity of MDM2 and catalyzes the transfer of ubiquitin from the E2 to p53. The deletion of any of these three domains inhibits MDM2 function and stabilizes p53, suggesting that all three are required for MDM2-mediated p53 degradation.^37-40 Other aspects of MDM2 worth noting include a nuclear localization sequence (residues 181–185) and a nuclear export sequence (residues 190–200), which are responsible for shuttling MDM2 into and out of the nucleus. Moreover, a cryptic nucleolar localization sequence has been identified near the C terminus of MDM2.⁴¹

Figure 1.

MDM2 and MDMX share significant homology. MDM2 and MDMX harbor similar structural domains, including an N-terminal p53-binding domain, a central acidic domain and C4 zinc finger domain, and a C-terminal RING domain. The central AD and Zn regions serve as binding sites for several small proteins, including ribosomal proteins and ARF, any of which results in the inhibition of MDM2 E3 ligase function. MDM2 can bind to other MDM2 molecules of to MDMX through their respective C-terminal RING domains. In contrast to MDMX, MDM2 possesses a nuclear localization sequence (NLS), a nuclear export sequence (NES) and a cryptic nucleolar localization sequence (NoLS). Moreover, whereas MDM2 exerts E3 ligase activity toward p53, MDMX does not possess appreciable E3 ligase activity. Red circles indicate phosphorylation sites. AD, acidic domain; CK, casein kinase; p53 BD, p53 binding domain; RING, really interesting new gene domain; Zn, C4 zinc finger domain.

Figure 2.

Many different p53 regulatory mechanisms are channeled through MDM2. MDM2 is modified post-translationally (phosphorylation by DNA damage kinases) or by direct protein binding (RPs and ARF) in response to various stresses (cues). When affected by any of the MDM2 inhibitors, MDM2 loses its ability to bind to and regulate p53, resulting in stabilized p53 levels and increased p53 transcriptional activity. Ablation of any of the three arms of MDM2 regulation severely impairs the activation of p53 in response to the corresponding stress cue. Red circles denote phosphorylation sites, the blue lines denote the binding sites of the indicated RP, and the black line represents the binding site of ARF. The inhibitors boxed in red have been shown to affect MDM2 homo- or heterooligomer formation. ARF, alternative reading frame.

In its functional configuration as an E3 ligase, MDM2 forms homooligomers and heterooligomers with MDMX.^22,42 MDMX is structurally similar to MDM2 (Figure 1), which reflects its evolution through a gene duplication event from MDM2 ~ 440 million years ago.⁴³ Despite extensive homology between these two genes (especially in the RING- and p53-binding domains), MDMX itself does not possess appreciable intrinsic E3 ligase activity toward p53.^44,45 Recent mutational analyses offered insight into the basis for this difference in catalytic proficiency. Interestingly, only two point mutations (N448C and K478R) are sufficient to restore E3 ligase activity to the MDMX RING domain in vitro.⁴⁶ However, it appears as though additional regions of MDMX that deviate from MDM2, including the AD and the analogous nucleolar localization sequence region of MDMX, may be equally necessary to convert MDMX into a functional p53 E3 ligase in cells.⁴⁷ MDMX also does not contain a nuclear localization sequence, which accounts for its predominantly cytoplasmic localization. Despite not possessing a nuclear localization sequence, MDMX can be transported to the nucleus by piggybacking onto MDM2 in a RING domain-dependent manner.⁴⁸ Purified MDMX RING domains have recently been shown to form oligomers suggesting that MDMX may also form homooligomers in cells.⁴⁹ Interestingly, MDM2 transcription is upregulated by p53, forming an autoregulatory inhibitory feedback loop.^50,51 For many years following its discovery, MDMX expression was thought to occur independently of p53 control;^44,52 however, recent reports have identified a functionally active p53 response element within intron 1 of the MDMX locus.^53,54 This response element can be induced in a p53-dependent manner in response to at least some stresses. Thus, MDM2 and MDMX appear to participate in negative feedback loops to control the p53 response, which could have a role in the rapid attenuation of the p53 response when an apoptotic outcome is not warranted. MDM2 and MDMX also share the ability to inhibit p53 transcriptional activity through direct binding and masking of the p53 transactivation domain.^15,16 Another common thread between MDM2 and MDMX is their ability to be ubiquitinated by MDM2. At least under overexpressed conditions, MDM2 can auto-ubiquitinate itself, and MDM2 can ubiquitinate MDMX in trans resulting in their respective degradation. Although MDM2 may be subject to degradation through other E3 ligases under physiological conditions,²⁴ the ubiquitination of MDM2/MDMX by MDM2 offers an autoregulatory mechanism through which MDM2 can reduce its own activity. One of the major questions that remains to be determined is how MDM2 directs its ligase activity toward p53, MDMX or itself, such as in the context of the MDM2–MDMX–p53 ternary complex.^55-57 Recent studies have suggested that binding of small proteins such as p14 alternative reading frame (ARF) to the MDM2 AD could direct the substrate specificity of MDM2 toward MDMX.^58,59 Under non-stressed conditions, MDMX is thought to stabilize MDM2, which could contribute to the increased potency of MDM2–MDMX heterooligomers for p53 ubiquitination relative to MDM2 homooligomers.^60-63 On the other hand, the absence of MDMX effectively stabilizes p53, as the deletion of the MdmX gene in mice (like the deletion of the Mdm2 gene in mice) results in an embryonic lethal phenotype that can be rescued by the concomitant deletion of p53.⁶⁴ Nonetheless, several studies have shown that MDM2 alone is capable of ubiquitinating and degrading p53.²² Moreover, MDM2 may be sufficient for p53 degradation, as the conditional deletion of MdmX in adult tissue does not cause extensive p53 stabilization and apoptosis when compared with Mdm2 deletion.⁶⁵

Genetic mouse models have shown that the reactivation of p53 signaling is sufficient for the ablation of tumors.^66-68 Moreover, the activation of p53 signaling appears to be specifically damaging enough to cancer cells to spare normal tissue from severe side effects from therapeutics that stabilize or restore the activity of p53. To fully capitalize on the potential of drugs that activate p53, such as MDM2 inhibitors, understanding how p53 is controlled by MDM2 oligomers is important. Moreover, to understand how to manipulate MDM2 oligomers to stabilize p53, we must understand MDM2 oligomers in detail. In the following few sections, we discuss our current knowledge on MDM2 oligomer formation.

MDM2 OLIGOMERIZATION IN VITRO STUDIES

The initial pieces of evidence for MDM2–MDMX synergy have come from in vitro studies. In 1999, Tanimura et al.,²² and subsequently Sharp et al.,⁴² revealed that MDM2 binds to MDMX through their respective RING domains. By 2003, Linares et al.⁶⁰ showed that MDMX enhances MDM2 E3 ligase activity. Subsequent studies revealed additional mechanistic details regarding the method of binding between MDM2 and MDMX. For example, although binding between the RING domains of MDM2 and MDMX was known, multiple studies showed that the extreme C-terminal tails (479–491 and 478–490 in human MDM2 and MDMX, respectively) of both proteins must be present for MDM2 oligomerization and E3 ubiquitin ligase activity.^69,70 Uldrijan et al.⁷⁰ also showed that point mutations within the MDM2 extreme C terminus can be introduced that do not disrupt MDM2 oligomerization but inhibit p53 ubiquitination, suggesting that the extreme C terminus may participate directly in the ubiquitin transfer reaction. Moreover, the lengths of the extreme C-terminal tails of MDM2 and MDMX are highly conserved, and the addition of residues that extend the length of either C-terminal tail inhibits MDM2 E3 function but not homo- or heterooligomerization.⁷¹

Another milestone in understanding MDM2–MDMX heterooligomerization is the publication of an x-ray crystal structure of a heterooligomer between the MDM2 and MDMX RING domains.⁷² This structure was preceded by the nuclear magnetic resonance-based resolution of the MDM2 RING domain homooligomer and the MDM2–MDMX RING domain heterooligomer in solution.⁷³ These studies were largely consistent in describing the structure of the MDM2–MDMX RING heterooligomer; however, due to unstructured regions in MDM2, we lack full-length MDM2 protein oligomer structures. Without a full-length crystal structure of MDM2, determining whether other domains participate in homo-and heterooligomerization must be accomplished by other techniques. Participation of other residues in addition to the RING domain and extreme C terminus of MDM2 in oligomer formation seems likely, especially concerning MDM2 homooligomers. In describing the purification of the MDM2 RING homo-oligomer, Linke et al.⁷² declared that the relative instability of MDM2s (residues 432–491) homooligomers could be improved by including additional N-terminal residues (MDM2l, residues 417–491). Moreover, a recent study performed by Dolezelova et al.⁷¹ suggested that MDM2 homooligomers and MDM2–MDMX heterooligomers likely form through different mechanisms. Another recent study using overexpression co-immunoprecipitation experiments provided evidence supporting this theory, suggesting that MDM2 homooligomers and MDM2–MDMX heterooligomers require different domains of MDM2.⁷⁴ Thus, despite extensive similarities in the reported oligomeric RING structures of MDM2 RING homooligomers and MDM2–MDMX RING heterooligomers, MDM2 appears to bind to MDMX and other MDM2 molecules through different mechanisms, which could imply differences in MDM2 oligomer function or efficiency. Whether these differences in binding have implications on the cellular level and whether these differences have implications for the development of effective MDM2 inhibitors remains to be determined.

Further complicating the issue of MDM2 oligomerization, MDM2 and MDMX can form intramolecular interactions (Figure 3). The MDM2 RING domain can fold back and bind the AD, thereby generating additional tertiary structure that could play a role in oligomerization and MDM2 activity.⁷⁵ Similarly, in MDMX, the p53-binding domain and the RING domain can each fold back and interact intramolecularly with the AD, although whether the AD can interact simultaneously with both the p53-binding domain and the RING finger remains unknown.^49,76 Although these interdomain interactions may have direct implications for previously described intermolecular oligomerization (for example, RING domain of one MDM2 molecule interacts with the AD of another MDM2 molecule⁷⁷), intramolecular domain interactions likely have a more significant role based on their kinetic favorability due to covalent attachment. Interestingly, these intramolecular interactions have been implicated in MDM2 E3 ligase activity⁷⁵ as well as in MDMX nuclear localization.⁷⁶ Disruption of the MDMX RING–AD interaction increases nuclear localization, suggesting that its ability to bind MDM2 for nuclear import is impeded when present in the RING-AD intramolecular configuration. These types of studies offer hints to the supramolecular assembly of the MDM2 oligomers for which we do not have direct structural data. Collectively, these studies suggest that MDM2 ternary complexes involving p53 require more than simple RING–RING and p53–p53-binding domain interactions.

Figure 3.

MDM2 and MDMX form intramolecular interactions. Recent studies have shown that MDM2 forms intramolecular interactions that involve RING–AD interactions. MDMX is also capable of RING–AD and p53 BD–AD intramolecular interactions. The implications of these intramolecular interactions remain unclear particularly in terms of function and oligomerization. However, these intramolecular interactions have clear effects on p53 regulation and likely effects on MDM2–MDMX oligomerization. AD, acidic domain; p53, p53 binding domain; RING, really interesting new gene domain.

REGULATION OF MDM2 OLIGOMERS

Based on the importance of MDM2 oligomerization for regulating p53, understanding how oligomer formation is regulated is of crucial importance. MDM2 and MDMX levels and activities are regulated in many different ways, including at the transcriptional level (p53–MDM2/MDMX autoregulatory loops), direct binding by other proteins (ARF,⁷⁸ RPL5,^79,80 RPL11,^81-83 CK1α⁸⁴ CK1δ^32,85), subcellular localization⁸⁶ and post-translational modifications (phosphorylation, ubiquitination and neddylation, reviewed by Wade et al.²¹). The respective ADs of MDM2 and MDMX appear to feature prominently in the regulation of intermolecular interactions, including oligomerization. In addition to the newly discovered intramolecular interactions involving the MDM2 and MDMX ADs, binding of the AD by small proteins and phosphorylation of the AD likely also contribute to MDM2 oligomer regulation. The AD–RING intervening sequence also appears to be a key regulatory point for MDM2 oligomerization. A study by Cheng et al.²³ showed that phosphorylation of sequence upstream of the RING domain of MDM2 results in the inhibition of MDM2 RING domain oligomerization, which correlates with reduced E3 ligase activity toward p53. Moreover, ATM-mediated phosphorylation of these residues in MDM2 inhibits the ability of MDM2 to oligomerize.⁸⁷ Interestingly, enforced oligomerization using a cross-linkable FKBP domain fused to the N terminus of the MDM2 RING domain significantly enhances MDM2 E3 activity, which provides direct evidence that MDM2 oligomerization is a critical point of regulation for MDM2 activity. As one might expect based on homology, MDMX phosphorylation at a similar region could result in decreased MDM2–MDMX heterooligomer formation and stabilized p53 levels as well; however, this remains to be tested.

MDM2 E3 ligase activity may also be regulated indirectly by manipulating MDMX. Because MDM2–MDMX heterooligomers are more efficient E3 ligases, it is thought that directing MDM2 E3 ligase activity toward MDMX could effectively reduce MDM2 E3 activity and thus stabilize p53. In a recent study, overexpression of the small protein ARF could decrease MDMX levels and stabilize p53 by directing MDM2 E3 ligase activity toward MDMX instead of p53.⁵⁸ Consistently, in the absence of ARF (as in many tumors that express WT p53), MDM2 E3 ligase activity appears to be less readily diverted towards MDMX degradation, which could explain why tumors that retain WT p53 are pressured for the mutation of the ARF locus. Because ARF binds to the MDM2 AD and possibly not the MDMX AD,⁸⁸ it is tempting to speculate that the positively charged ARF is required to stabilize the quaternary structure between the two MDM ADs. Similar observations were reported for RPL11.⁸⁹ In another study analyzing RPS27l knockout mouse cells, the induction of ribosomal stress by the absence of RPS27l resulted in MDM2-mediated MDMX degradation, which reduced MDM2–MDMX heterooligomer formation and stabilized p53.⁹⁰ The effect of RP binding on MDM2 oligomer formation has not been tested directly; however, based on the reported effects of ARF–MDM2 binding,⁵⁸ we suspect that RP–MDM2 binding may behave in a similar manner in terms of reducing MDM2–MDMX heterooligomer formation. On the basis of these studies, it would be interesting to determine whether ARF- or RP-mediated stabilization of MDM2 AD–MDMX AD binding could offer a viable method through which to generate x-ray crystallographs using the full-length proteins. Notably, Linke et al.⁷² reported that the presence of l-arginine, a positively charged amino acid, was necessary to obtain crystals of MDM2–MDMX RING domains. If the conditions for full-length crystal structures of MDM2 can be resolved, then it would be interesting to determine how the binding of small proteins, such as ARF or RPL11, affects the overall structure and substrate selection of MDM2.

MDM2–MDMX HETEROOLIGOMERS ARE REQUIRED IN VIVO

In addition to the abundance of in vitro evidence dissecting MDM2 oligomerization mechanics, several mouse models have been developed that offer complementary in vivo platforms to investigate MDM2 oligomerization. When the results of these studies are analyzed in aggregate, they offer mechanistic insight as well as a clearer picture of the importance of MDM2 oligomerization in p53 regulation. The earliest mouse models based on the knockout of Mdm2 or MdmX have revealed p53-dependent embryonic lethal phenotypes for either gene.^64,91,92 These studies were critical to show the importance of each of these two genes in p53 regulation. Moreover, these models show that neither MDM2 nor MDMX can completely compensate for the deficiency of the other. Of note, overexpressing an Mdm2 transgene in mice can compensate for MDMX deficiency.⁹³ Nonetheless, although MDM2 and MDMX have been attributed with the ability to physically mask the p53 transactivation domain, it appears as though this activity is insufficient for the control of p53 during mouse development. This idea was supported by a study that showed that mice homozygous for the MDM2 RING domain structural mutation C462A, which abrogates MDM2 E3 activity and MDM2–MDMX heterooligomerization, die in utero in a p53-dependent manner.²⁴ As mentioned in the preceding section, the corresponding human MDM2 RING mutant C464A retains the ability to form homooligomers under overexpressed conditions,⁷⁴ which may contribute to residual p53 inhibitory activity, as MDM2 likely requires some type of oligomerization to function (discussed further below).

Subsequent knock-in mouse studies have hinted at the importance of MDM2–MDMX heterooligomer formation for the control of p53 activity during embryonic development. Two similar studies by Pant et al.⁹⁴ and Huang et al.⁹⁵ showed that mice expressing mutant versions of MDMX lacking the RING domain or lacking proper RING domain structure display an embryonic lethal phenotype. In these mice, although the p53-binding domain of MDMX remains intact and MDM2 is WT, the mice die in utero in a p53-dependent manner (approximately day E9.5). These studies suggest that heterooligomerization is the key to p53 control, at least during the embryonic stages of development. Interestingly, heterooligomerization is dispensable for p53 regulation during adulthood.⁹⁴ Consistently, the death of these mice, as well as all embryonic lethal Mdm2/MdmX knock-in mice, approximately coincides with the stage in which p53 expression is dramatically and ubiquitously upregulated, suggesting the crucial need for effective p53 regulation during this stage of embryonic development (~ E8.5, Schmid et al.⁹⁶). Expounding on the C462A and RING MDMX knock-in mouse studies, Tollini et al. showed that mice expressing an MDM2 knock-in mutant (Y487A) that is E3-dead but retains the ability to heterooligomerize with MDMX are viable with no phenotypic differences under unstressed conditions.⁹⁷ Because the Mdm2^Y487A/Y487A mouse lacks the ability to ubiquitinate p53, in the context of the MdmX knock-in models, it appears as though MDM2–MDMX heteroligomerization is sufficient to control p53 during embryonic development and is dispensable during adulthood, whereas MDM2 E3 ligase activity is sufficient (and necessary) to control p53 in adult tissue, especially in the presence of stress, and is dispensable during embryonic development.

Another thing that is apparent from analyzing the various MDM2/MDMX mouse models is that MDM2 is a more effective regulator of p53 in vivo than MDMX.⁹⁸ Indeed, the loss of MDM2 is consistently more detrimental to mice than the equivalent loss of MDMX (reviewed in Wade et al.⁹⁸). For example, in a study of mice expressing a brain-specific Cre-inducible p53 allele, Francoz et al.⁶⁵ showed that p53^LSL/−;MdmX^−/−;Nes-Cre mice survived birth, whereas p53^LSL/−;Mdm2^−/−;Nes-Cre mice died during development. In this study, assuming Cre-mediated gene excision occurred at similar efficiencies in the two mice, the presence of MDM2 alone resulted in a better outcome than the presence of MDMX alone. This trend is noticeable in other mouse models as well. Knockout mice homozygous for Mdm2 deletion die at a slightly yet consistently earlier embryonic stage when compared with homozygous MdmX knockout mice (E5.5 for MDM2-null vs E7.5 for MDMX-null mice).^64,91,92 Consistent with a more severe phenotype, Mdm2 knockout mice display extensive apoptosis, whereas MdmX knockout mice display extensive cell cycle arrest.⁹⁹ In another study, brain-specific loss of Mdm2 in mice resulted in death at E12.5 compared with brain-specific loss of MdmX, which resulted in death at E17.5.¹⁰⁰ Moreover, using a tamoxifen-inducible p53 fusion protein (p53ER) expressed in mice, the Evan lab showed in separate studies that the activation of p53ER in adult Mdm2^−/− mice results in lethality around 5 days, whereas p53ER activation in adult MdmX^−/− mice results in lethality in ~29 days.^101,102 Interestingly, mice harboring an E3-dead, heterooligomerization-deficient MDM2^C462A/C462A mutant die around E7.5, which is closer to the MDMX-null mutant mouse, suggesting that MDM2^C462A might possess residual ability to inhibit p53 that is not present in the MDM2-null mice.²⁴ What could explain the ability of the MDM2-C462A mutant to inhibit p53 in vivo? One possibility is the p53-binding domain, which can still bind to p53 in vivo.²⁴ The MDM2 p53-binding domain actually displays a three- to four-fold greater affinity for p53 than that of MDMX, which suggests that the presence of the MDM2 p53-binding domain in the Mdm2^C462A/C462A mice offers an advantage in transcriptional suppression over that of the MDMX p53-binding domain, which is also present in these mice. Physical MDM2^C462A–p53 interaction could contribute to the delay in death of mice that express the MDM2 p53-binding domain compared with mice that do not (for example, Mdm2^C462A/C462A mice vs MdmX knockout mice^103,104), although further experiments will be required for confirmation. Moreover, although it is possible that MDM2^C462A protein retains the ability to suppress p53 through direct binding of the p53 transactivation domain, whether the homooligomerization capacity of MDM2 is required remains to be determined in vivo.

The MDM2–MDMX heterooligomer is a clearly more efficient E3 ligase platform than the MDM2 homooligomer.^39,62 However, the secret behind the increased ligase efficiency of the heterooligomer (as well as the basic requirement for MDMX, which lacks E3 activity) remains a mystery. Structural data imply that MDMX binding to MDM2 creates a unique platform capable of more efficiently transferring the ubiquitin molecule from the E2 to p53.^72,73 However, other models cannot be excluded based on the current data. For example, MDM2 is commonly thought to bind directly to the E2. However, because direct evidence for E2 binding to MDM2 remains elusive, it is conceivable that MDMX could bind to and transport the cognate E2 to MDM2, which in turn catalyzes the transfer of ubiquitin to p53. Circumstantial evidence in support of this possibility includes the currently inexplicable observation that overexpression of MDMX in the presence of many different types of E3-dead MDM2 constructs can rescue MDM2 E3 ligase activity.^70,97,105 If MDMX is able or necessary to transport the E2-Ub to MDM2, then overexpression of MDMX could increase the intracellular concentration of MDM2–MDMX–Ub complexes in the cell, forcing the transfer of Ub to p53 and resulting in E3 ligase rescue. Observations in our lab suggest that MDMX shows specific binding to the E2 UbcH5 when compared with MDM2 (unpublished observations). Another consideration is that MDMX could bind to a different E2 than MDM2 does resulting in the targeting of different lysine residues or in the conjugation of differently linked Ub chains. The endogenous E2 associated with MDM2-dependent ubiquitination of p53 has not yet been confirmed, thus speculation persists regarding MDM2–MDMX-mediated ubiquitin transfer to p53. Perhaps the greatest advances in predicting and understanding MDMX enhancement of MDM2 E3 ligase activity could be achieved through the analysis of complete full-length crystal structures of the heterooligomer compared with the homooligomer.

MDM2–MDMX HETEROOLIGOMERIZATION IS IMPORTANT, WHAT ABOUT MDM2 HOMOOLIGOMERIZATION?

A uniting factor in all of the MDM2/MDMX mutant mouse models is that the presence of MDM2 (even structural mutant MDM2^C462A) is sufficient to delay lethality. Notably, prior to the evolution of MDMX, MDM2 was tasked with the regulation of p53 for organismal survival. Thus, although MDM2 likely retains the ability to inhibit p53 by itself, the predominant mechanism of MDM2-mediated p53 inhibition in vivo remains to be determined. Early evidence has suggested that the direct binding of MDM2 to p53 is sufficient to suppress p53 activity.^15,16 Although direct binding and inhibition of p53 could account for this delay in lethality, whether MDM2 homooligomerization also contributes to p53 degradation/inhibition in vivo is unknown. Studies confirming the in vivo existence and the role of MDM2 homooligomers need to be conducted; however, inherent difficulties are associated with analyzing MDM2 homooligomers in vivo because of the identical nature of the alleles of MDM2. With the increasing popularity and standardization of modern-era genome editing tools, a useful mouse or genome-edited cell model that could address these questions is one that expresses different epitope tags on each of the two MDM2 alleles. Cells expressing endogenous levels of multiple epitope-tagged MDM2 alleles could be used to confirm the existence of MDM2 homooligomers under physiological conditions. Another MDM2 knock-in model of interest is an MDM2 mutant that selectively prevents homooligomer formation while leaving heterooligomer formation intact. This model could be used to determine how the abrogation of MDM2 homo-oligomers under physiological conditions affects MDM2 function and organismal fitness. Although point mutations may not be sufficient to achieve the selective inhibition of MDM2 homo-oligomers, the deletion of AD residues appears to confer a selective defect in MDM2 homooligomer formation while leaving MDM2–MDMX heterooligomerization and E3 ligase activity largely intact, at least under overexpressed conditions.⁷⁴ Because E3 ligase activity appears to remain intact in the presence of MDM2 AD deletion constructs,^106,107 one would expect viable offspring. However, the construction of a conditional knock-in mutant mouse harboring an inducible deletion of the Mdm2 AD could be a worthwhile endeavor to determine the effect of Mdm2 AD deletion on survival and p53 stability during the adult stages. A better understanding of how MDM2 oligomers form and function in vivo could provide valuable insight into the design of MDM2-targeting drugs.

TARGETING THE MDM2 OLIGOMER

Considering their crucial role in p53 regulation, particularly in the context of DNA damage, targeting MDM2/MDMX is a promising avenue of pharmacological research. MDM2/MDMX dual inhibitors could be particularly effective in tumors that overexpress these targets and harbor WT p53 (Figure 4). Nonetheless, most compounds developed to date offer relatively specific MDM2 inhibition by targeting the p53-binding domain on MDM2.^108,109 These inhibitors have been carefully optimized, and some have shown strong efficacy in preclinical studies; however, their efficacy in patients has been underwhelming. The lack of clinical success with MDM2 inhibitors can be attributed to multiple factors, the most prominent of which is on-target toxicity.¹¹⁰ Thus, although highly potent MDM2 inhibitors are currently available (RG7112, K_d = 11 nm; MI-888, K_d = 0.44 nm; RO-5963, IC₅₀ = 33 nm), better targeting mechanisms or dosing optimization, perhaps by combining MDM2 inhibitors with other therapies, may be necessary to achieve the best patient outcomes. However, even if side effects could be minimized, it is possible that MDM2 inhibition alone could be ineffective for several types of cancers.¹¹¹ For example, inhibitors that target only MDM2 and not MDMX, such as nutlin-3a (400-fold less effective against MDMX than against MDM2), are ineffective against cancer cells that over-express MDMX presumably due to MDMX-mediated inhibition of p53.^112-114 Moreover, long-term treatment with MDM2 inhibitors introduces a selective pressure that could promote the development of p53 mutations and multi-drug resistance.^115,116 Another issue unique to MDM2-specific inhibitors is the requirement for high doses of the drug to counter the p53-regulated increase in MDM2 expression as a result of inhibiting MDM2 in the first place. Although some of these issues also apply to MDM2/MDMX dual inhibitors, current efforts in MDM2 inhibition have paid more attention to the effects of MDMX as well.

Figure 4.

Reactivation, inactivation and stabilization of p53 through direct means or through MDM2/MDMX inhibition has shown promise. The MDM2/MDMX-p53 pathway is currently being targeted through two broad mechanisms: MDM2/MDMX inhibition through small molecules (RO-5963, MEL23/24, HLI98) and stapled peptides (ATSP-7041) and mutant p53 reactivation through compounds such as APR-246 and inactivation through compounds such as aptamers. MDM2/MDMX inhibition could be particularly useful for p53 WT tumors, whereas mutant p53 reactivation/inactivation could be useful for mutant p53 tumors and for the expansion of the tumor spectrum for which MDM2 inhibitors may be used. Mutp53: mutant p53, WTp53: wild-type p53.

Differences in the structures of the p53 binding pockets of MDM2 and MDMX present a challenge to the development of drugs that effectively target both molecules.¹¹⁷ Nonetheless, recent drug candidates have emerged that simultaneously target MDM2 and MDMX^118-120 (see Khoo et al.¹²¹ and Burgess et al.¹²² for detailed reviews of MDM2 drugs). In a study by Graves et al.,¹²⁰ the authors show that RO-5963, an indolyl hydantoin derivative, binds the p53 binding pocket of MDM2 and MDMX with similar affinities in vitro. Moreover, RO-5963 binding simultaneously engages MDM2 and MDMX, resulting in the enforced oligomerization of the two molecules. This results in p53 stabilization and cell death in several cancer cell lines, especially in cancer cells expressing high levels of MDMX, which is a promising proof-of-principle example of dual MDM2–MDMX inhibition. Although the clinical utility of this compound as a single agent may not be likely possibly due to the compensatory induction of MDM2, the use of compounds such as RO-5963 in conjunction with other p53-activating compounds may prove effective for certain types of tumors.

In addition to preventing interaction with p53, other methods of inhibiting MDM2 and MDMX have involved the inhibition of E3 ligase activity and MDM2/MDMX heterooligomerization. Early studies suggested that the MDM2–MDMX RING–RING oligomer was not an attractive drug target due to the lack of a defined catalytic site and a predominantly hydrophobic interaction-based binding mechanism.^72,73 Nonetheless, experimental evidence has suggested that the inhibition of MDM2–MDMX heterooligomerization could be effective, as MDMX RING domain overexpression competes for MDM2 binding, inhibits endogenous MDM2–MDMX oligomer formation, and results in the stabilization and activation of p53.^62,123 Small molecules that inhibit MDM2 enzymatic activity have been reported, including HLI98 and MDM2 E3 Ligase (MEL) inhibitors. Although HLI98 showed specificity for MDM2 inhibition in vitro, it also displayed non-specific effects and p53-independent effects at higher concentrations in cells as well as poor pharmacological characteristics overall.¹²⁴ A more promising candidate of the MDM2 ligase inhibitor class, the MEL series of inhibitors (MEL23 and MEL-24) showed specific inhibition of MDM2–MDMX E3 ligase activity, stabilization of p53 and the induction of apoptosis in a p53-dependent manner.¹²⁵ Although more work using in vivo models is required to validate the MEL inhibitors, targeting MDM2–MDMX enzymatic activity may not be as difficult as once thought. In an even more recent study using computational and rational methods, a peptide-based inhibitor of MDM2–MDMX oligomerization was reported.¹²⁶ This inhibitor, named Peptide3, showed a p53-dependent apoptotic response; however, its reported mechanism of action is peculiar in that only nuclear MDM2–MDMX heterooligomerization is affected. Perhaps using a more thorough and unbiased screening method, such as phage display, coupled with emerging peptide technologies, such as peptide stapling, could yield even more effective peptide-based MDM2–MDMX oligomerization inhibitors.

Another area of intense research includes the development of p53 mimetic stapled peptides. In a study by Chang et al., the authors describe a stapled α-helical peptide that binds to the p53-binding domains of MDM2 and MDMX with low-nanomolar affinities.¹¹⁸ Although controversial, stapled peptide technology offers the unique opportunity to specifically mimic the p53-binding site in MDM2 and/or MDMX with strong affinity. It will be interesting to see how far stapled peptide drug candidates, such as ATSP-7041, can progress through clinical trials.

EXPANDING THE TUMOR SPECTRUM OF MDM2 INHIBITORS

What is even more exciting is how the spectrum of tumors that respond to drugs that inhibit MDM2 oligomerization and function (WT p53 and/or overexpressing MDM2/MDMX) could be increased by combining them with complementary drugs that also target the MDM2/MDMX–p53 pathway. For example, the compound APR-246 (also known as PRIMA-1^MET), a mutant p53 reactivator, has shown some promise in patients harboring mutant p53 tumors in an early-stage clinical trial.¹²⁷ The combination of APR-246 with various standard chemotherapeutics has also shown synergistic effects in the treatment of cancer cells and warrants further investigation.^128-130 Moreover, considering the reduced efficacy of MDM2 inhibitors in many p53 mutant tumors as well as the selective pressure that MDM2 inhibitors, such as nutlin-3a, exert on tumors to develop p53 mutations,^115,116 p53 mutant reactivators could be effective adjuvants to increase the scope of MDM2/MDMX inhibitors. In fact, because many p53 mutations result in gain-of-function phenotypes,⁸ the treatment of p53 mutant tumors with MDM2 inhibitors may actually require p53 reactivation.

Another therapeutic modality being investigated is the use of aptamers, short single-stranded DNA or RNA constructs that can be engineered to bind specific proteins. Although delivery of nucleic acids remains an obstacle, aptamers offer the ability to specifically target deleterious point mutations characteristic of tumors, such as the p53 mutation R175H.¹³¹ Moreover, delivery of p53R175H aptamers via nanoparticles showed impressive efficacy in a mouse xenograft tumor model. Similarly to the potential for PRIMA-1^MET synergism with MDM2 inhibitors, aptamers could also be useful adjuvants to expand the effective tumor profile of MDM2 inhibitors.

Because cancer is a heterogenous disease with multiple subclonal populations, future curative cancer treatment attempts may focus on targeting all of the tumor clone subsets. Because aberrant p53 signaling is an early event in many types of tumors, combining direct p53-targeting treatments with drugs that target upstream factors like MDM2 inhibitors could result in synergistic effects that could be effective in many types of tumors. Although no p53-activating drugs have been approved for clinical use, the future of p53-targeting drugs offers a promising, relatively unexplored avenue through which a broad array of tumors may be treated.

CONCLUDING REMARKS

Although we have made great progress in understanding MDM2 oligomerization and how MDM2 can affect tumor development, basic concepts remain to be determined. Some crucial aspects about MDM2 oligomerization that remain to be discovered include the following: How does MDMX enhance MDM2 E3 ligase activity toward p53? What role do MDM2 homooligomers play in vivo? What mechanisms control the direction of MDM2 E3 activity toward p53 vs other substrates (for example, MDMX)? What residues of MDM2 are necessary for homo- vs heterooligomerization? Are other domains of MDM2 worthy targets for the development of small molecules that can activate p53? As we ponder new methods to address these questions, we eagerly anticipate new insights from the brilliant minds of the MDM2/p53 field.