Mdm2 is a critical negative regulator of the p53 tumor suppressor and is frequently overexpressed in human cancers. However, reports, including our own studies, suggest that Mdm2 has both p53-dependent and p53-independent functions that contribute to genomic instability and transformation when deregulated. We recently elucidated a p53-independent role for Mdm2 in the regulation of the DNA double-strand break repair response, genomic stability, and transformation through interaction with Nbs1, a member of the Mre11/Rad50/Nbs1 DNA double-strand break repair complex. In light of these findings, targeting Mdm2 in human malignancies may have effects other than activating p53. [Cancer Res 2009;69(5):1697–701]

Mdm2 is an important regulator of tumor development. Mdm2, an E3 ubiquitin ligase, regulates p53 by controlling both the stability of the p53 protein and the activity of p53 as a transcription factor (1). The importance of the p53 tumor suppressor pathway for the prevention of transformation has long been recognized, as inactivation of the pathway is a frequent step in the development of the majority of human and murine cancers (2, 3). The MDM2 gene is amplified in ∼10% of all human cancers, with a third of sarcomas harboring MDM2 amplifications (4). Overexpression of the Mdm2 protein without amplification is also very common in human cancers, including hematologic malignancies. Approximately a third of B-cell chronic lymphocytic leukemias and non–Hodgkin's lymphomas have >10-fold overexpression of Mdm2 (5). It has only recently been appreciated that Mdm2 does not have to be grossly overexpressed to contribute to tumor development. Specifically, in humans, a single nucleotide polymorphism (SNP) identified in MDM2 at position 309 of intron 1 results in 2- to 4-fold higher expression of Mdm2 protein (6). This moderate increase in Mdm2 expression was linked to accelerated tumorigenesis in spontaneous and familial cancers (6). Additionally, Mdm2 transgenic mice expressing 4-fold higher Mdm2 levels spontaneously developed a variety of cancers (7). Several lines of evidence indicate that decreased expression of Mdm2 inhibits tumor formation. Mdm2 heterozygous Eμ-myc transgenic mice had an extended life expectancy due to a significant delay in lymphoma development compared with Mdm2 wild-type Eμ-myc transgenic mice (8). Furthermore, mice with one mutant adenomatous polyposis coli allele developed fewer intestinal adenomas in direct correlation with decreasing amounts of Mdm2 expression (9). Finally, ARF−/−Mdm2+/− mice had a significantly increased tumor latency, resulting in a 6-month longer average survival compared with ARF−/−Mdm2+/+ mice (10). These studies illustrate the influence Mdm2 levels have on the regulation of tumorigenesis, yet the contribution of p53-independent pathways to this process is incompletely understood.

Although it is clear that Mdm2 promotes tumorigenesis through regulation of p53, evidence from humans and mice also supports a p53-independent role for Mdm2 in tumor development (11). For example, studies of human sarcomas and bladder cancers showed that a subset of these tumors overexpressed Mdm2 in addition to having mutated p53, and those patients whose tumors had both abnormalities had decreased survival compared with patients with tumors possessing either abnormality alone (12, 13). Moreover, one third of lymphomas that emerged in Eμ-myc transgenic mice that have mutated or deleted p53 also overexpressed Mdm2 (8). For ad tumor to overexpress Mdm2 and inactivate p53 would seem redundant, unless Mdm2 has other functions besides the regulation of p53. Studies in mice lacking p53 further cemented a p53-independent role of Mdm2 in tumorigenesis. Specifically, p53−/− mice that overexpressed Mdm2 (transgenic) or were deficient in Mdm2 (heterozygous) developed a different tumor spectrum than p53−/− mice alone (7, 14). In addition, mammary-specific Mdm2 transgenic mice exhibited defective mammary gland development that was independent of their p53 status, and these mice went on to develop mammary tumors (15). Splice variants of Mdm2 lacking the p53-binding domain were identified in human and murine tumors and subsequently shown to promote transformation in vitro and tumor development in vivo (16). Therefore, although it is becoming apparent that elevated levels of Mdm2 contribute to tumorigenesis independent of p53, how this occurs has remained a mystery. Although it has been reported that Mdm2 may influence the retinoblastoma tumor suppressor pathway (11), recent studies by our laboratory have revealed a novel, p53-independent role for Mdm2 in the maintenance of genome stability.

Using a nonbiased approach, we identified the Mre11/Rad50/Nbs1 (M/R/N) DNA repair complex in immunoprecipitations of endogenous Mdm2 from cells lacking functional p53 (17). The M/R/N complex contributes to the preservation of genome stability by participating in DNA double-strand break repair, cell cycle check point control, and telomere maintenance (18). Mre11, Rad50, and Nbs1 are essential proteins, as deletion of any of the three genes in mice is embryonic lethal (18). In humans, hypomorphic mutations in the NBS1 and MRE11 genes cause the human genome instability syndromes Nijmegen breakage syndrome (NBS) and ataxia-telangiectasia–like disorder (ATLD), respectively (18). NBS and ATLD patients exhibit genome instability, and patients with NBS are predisposed to cancer. The M/R/N complex functions in signaling the presence of DNA breaks, as well as the initial enzymatic processing of the broken DNA (18). With biochemical methods, we determined that Mdm2 associates with the M/R/N DNA repair complex by specifically binding to Nbs1 (17, 19). Using a series of Mdm2 and Nbs1 deletion mutants, we localized the Nbs1-binding domain in Mdm2 to 31 amino acids (amino acids 198–228) and the Mdm2-binding domain in Nbs1 to 39 amino acids (amino acids 474–512; Fig. 1). Mutation of conserved amino acids in the Nbs1-binding domain of Mdm2 or the Mdm2-binding domain in Nbs1 compromised the ability of the two proteins to associate (19). Showing a novel association between Mdm2 and Nbs1, a protein critical for genome stability, led us to question whether Mdm2 had a p53-independent role in the maintenance of genome stability.

Figure 1.

Schematic diagrams of Mdm2 and Nbs1. For Mdm2, the location of the binding domains for p53, Nbs1, and ARF are boxed. The E3 ubiquitin ligase ring finger (Ring) domain, nuclear localization signal (NLS), nuclear export signal (NES), and nucleolar localization signal (NoLS) are indicated. The location of the ATM phosphorylation site at Ser395 is noted. For Nbs1, the location of the binding domains for Mdm2, Mre11, and ATM are boxed. The forkhead-associated domain (FHA) and the BRCA1 COOH-terminal domains (BRCT1 and BRCT2) are indicated. The three nuclear localization signal regions are noted. The ATM phosphorylation sites at Ser278 and Ser343 are indicated.

Figure 1.

Schematic diagrams of Mdm2 and Nbs1. For Mdm2, the location of the binding domains for p53, Nbs1, and ARF are boxed. The E3 ubiquitin ligase ring finger (Ring) domain, nuclear localization signal (NLS), nuclear export signal (NES), and nucleolar localization signal (NoLS) are indicated. The location of the ATM phosphorylation site at Ser395 is noted. For Nbs1, the location of the binding domains for Mdm2, Mre11, and ATM are boxed. The forkhead-associated domain (FHA) and the BRCA1 COOH-terminal domains (BRCT1 and BRCT2) are indicated. The three nuclear localization signal regions are noted. The ATM phosphorylation sites at Ser278 and Ser343 are indicated.

Close modal

Maintaining genome integrity is essential for the prevention of transformation, and several reports provide evidence that altered Mdm2 levels affect genome stability. Specifically, Mdm2 heterozygous ARF-null mouse embryo fibroblasts (MEF) had fewer chromosomal abnormalities and were resistant to Ras-induced transformation compared with Mdm2 wild-type ARF-null MEFs (10). In vitro experiments showed that expression of Mdm2 at levels 2- to 4-fold higher than endogenous levels led to centrosome amplification and chromosome instability (20). In vivo, B cells from Mdm2 transgenic mice exhibited significantly increased numbers of chromosome/chromatid breaks and aneuploidy (21). There was increased breast epithelial cell polyploidy, a marker of genomic instability, in mammary-specific Mdm2 transgenic mice independent of their p53 status (15). These studies illustrated that decreased Mdm2 levels reduced chromosomal instability, and increased Mdm2 expression resulted in an increase in genomic instability. However, identification of an Mdm2-dependent, p53-independent component of genome stability remained elusive until we identified the Mdm2/Nbs1 interaction.

Regulation of p53 by Mdm2 is one mechanism by which Mdm2 can affect genome stability, but the identification of a p53-independent association of Mdm2 with the M/R/N DNA repair complex suggested an additional pathway through which Mdm2 could regulate the stability of chromosomes. In support of this concept, we observed that Mdm2 overexpression doubled the frequency of spontaneous chromosome/chromatid breaks in primary fibroblasts, including MEFs that lacked p53 (19). The increase in chromosome/chromatid breaks that occurred when Mdm2 was overexpressed was dependent on Nbs1 binding to Mdm2, as mutation of the Nbs1-binding domain in Mdm2 abrogated the ability of Mdm2 to induce DNA breaks. This increase in genomic instability caused by Mdm2 overexpression promoted the transformation of p53-null MEFs, and again, an intact Nbs1-binding domain in Mdm2 was required for the increased transformation potential of Mdm2 (19). These studies showed that elevated levels of Mdm2 compromise genomic stability, leading to transformation independent of p53 and dependent on interaction with Nbs1.

The M/R/N complex functions in the initial sensing, signaling, and processing of DNA double-strand breaks (18, 22). Defective DNA repair proteins are a common cause of genomic instability, which can lead to transformation (22). Our data revealed that Mdm2 overexpression inhibited DNA double-strand break repair independent of the presence of p53 (17, 19). Surprisingly, the ubiquitin ligase activity that has defined Mdm2 for 10 years was dispensable for its effects on DNA break repair. We did show that Mdm2 inhibited DNA double-strand break repair through interaction with Nbs1 because Mdm2 mutants containing the Nbs1-binding domain could inhibit repair, whereas those lacking their Nbs1-binding domain could not. Notably, Mdm2 with alanine point mutations in the Nbs1-binding domain that compromised Mdm2/Nbs1 association, but not binding to p53 or ARF, a negative regulator of Mdm2, was unable to inhibit repair (19). Moreover, Mdm2 did not inhibit DNA repair in cells expressing a mutant form of Nbs1 that was unable to bind Mdm2. Although elevated levels of Mdm2 inhibited DNA break repair through interaction with Nbs1, we did not observe an effect on the cell cycle checkpoint functions of Nbs1 (19). How Mdm2 interferes with the DNA repair function of Nbs1, but not the checkpoint function of Nbs1, remains unclear. However, cells with mutations in the M/R/N complex only exhibit partial defects in the intra-S and G2-M checkpoints (18), and consequently, current assays may not be sensitive enough to detect modest differences caused by increased Mdm2 expression. Therefore, our studies have revealed a novel p53-independent, Nbs1-dependent function of Mdm2 in the regulation of DNA break repair that seems independent of Nbs1-mediated cell cycle checkpoint functions.

Following DNA damage, the M/R/N complex promotes autophosphorylation of ataxia-telangiectasia mutated (ATM) on Ser1981 and activation of ATM (Fig. 2; ref. 18). Activated ATM then phosphorylates numerous proteins involved in the DNA repair response, including histone H2AX, Smc1, Chk2, Nbs1, p53, and many others (22). The M/R/N complex is involved in ATM activation, as cells with hypomorphic mutations in Nbs1 or Mre11 have attenuated phosphorylation of ATM and ATM targets following DNA damage. ATM phosphorylates histone H2AX, and phosphorylated H2AX, referred to as γ-H2AX, is important for the long-term retention of repair factors at sites of DNA breaks, which is necessary for efficient repair of DNA breaks (22). When a DNA repair response occurs, proteins involved in DNA repair, including ATM and the M/R/N complex, relocalize to sites of DNA breaks and form nuclear foci (a punctate localization pattern). On completion of DNA repair, the nuclear foci resolve. Because the M/R/N complex is involved in early DNA damage signaling events, resulting in the activation of ATM, phosphorylation of H2AX, and the formation of nuclear foci, we postulated that the delay in DNA break repair by Mdm2 may be due to interference with the function of Nbs1 in these early DNA repair response events. In support of this notion, our studies showed that elevated levels of Mdm2 delayed the phosphorylation of H2AX, the formation of γ-H2AX foci, and the appearance of phospho-S1981-ATM/phospho-ATM target protein foci (19). Importantly, the delay in foci formation was dependent on the Nbs1-binding domain in Mdm2, indicating that Mdm2/Nbs1 association was necessary for Mdm2 to inhibit the initial phosphorylation events immediately following DNA damage. Additionally, elevated levels of Mdm2 prolonged the presence of Nbs1 at foci, delayed the resolution of γ-H2AX and phospho-S1981-ATM/phospho-ATM target protein foci, and extended the time it took for DNA breaks to be repaired (19). These effects of Mdm2 also seemed to require Mdm2/Nbs1 interaction. Therefore, our studies indicate that through interaction with Nbs1, increased levels of Mdm2 inhibit early DNA damage signaling events resulting in a delay in double-strand DNA break repair. These results add valuable new insight into Mdm2 by revealing a p53-independent function of Mdm2 that contributes to genome stability and tumorigenesis.

Our discovery that the interaction of Mdm2 with Nbs1 results in the inhibition of DNA double-strand break repair and an increase in chromosome/chromatid breaks provides an explanation as to how elevated levels of Mdm2 promote genomic instability and transformation independent of p53 (Fig. 2). Our data also implicate Mdm2 in the disruption of early signaling events necessary for mounting a timely response to DNA damage. Specifically, the delay in phosphorylation of ATM targets, including H2AX, at sites of DNA breaks detected when Mdm2 was overexpressed suggests that ATM activity is inhibited by increased Mdm2 expression. The fact that mutation of the Nbs1-binding domain in Mdm2 abrogated this effect indicates that Mdm2/Nbs1 association is necessary for Mdm2 to exert its inhibitory effects. However, we cannot exclude the possibility that Mdm2 also may directly alter ATM activity, because Mdm2 has been shown to associate with and is phosphorylated by ATM after DNA damage (23). Nevertheless, a delay in the phosphorylation of ATM target proteins that localize to sites of DNA damage should negatively affect subsequent DNA repair signaling and possibly the recruitment of other proteins involved in DNA repair, resulting in a slower processing of DNA breaks, as we observed. Moreover, because little is known about how a cell determines which DNA double-strand break repair pathway to use, delaying the initial signaling events and altering the kinetics of protein recruitment and/or activation may affect pathway choice, which should affect mutation rates. The choice between homologous recombination (HR) and nonhomologous end-joining (NHEJ) is partially influenced by the phase of the cell cycle, but there is also competition between the two pathways (24). If repair that would normally occur by an error-free pathway of HR is instead repaired by the more error-prone NHEJ pathway, an increased rate of mutations would be expected, potentially increasing the chance of a transforming mutation. Recently, a role for ATM in the suppression of NHEJ-related degradation of damaged DNA ends was shown (25). If Mdm2 overexpression delays ATM activation through its interaction with Nbs1, Mdm2 could alter this ATM function, resulting in a more mutagenic DNA repair process. Therefore, it will be important to determine precisely how Mdm2 is inhibiting the early DNA repair response signal and how this may affect the recruitment of proteins to sites of DNA breaks, the activation of proteins that have already localized to foci, and the repair pathway that corrects the breaks.

Figure 2.

Model of inhibition of DNA break repair by Mdm2. A DNA double-strand break induced by γ-irradiation is detected by the M/R/N complex, which localizes to the DNA break. The M/R/N complex recruits ATM to the DNA break and facilitates ATM dimer dissociation and activation (autophosphorylation). Activated ATM phosphorylates Nbs1, histone H2AX, p53, Mdm2, and numerous other proteins that are not pictured that are involved in the DNA break repair response. The presence of elevated levels of Mdm2 delays early phosphorylation events mediated by ATM that are necessary for a rapid DNA double-strand break repair response, resulting in inefficient repair of DNA breaks, genomic instability, and ultimately tumorigenesis.

Figure 2.

Model of inhibition of DNA break repair by Mdm2. A DNA double-strand break induced by γ-irradiation is detected by the M/R/N complex, which localizes to the DNA break. The M/R/N complex recruits ATM to the DNA break and facilitates ATM dimer dissociation and activation (autophosphorylation). Activated ATM phosphorylates Nbs1, histone H2AX, p53, Mdm2, and numerous other proteins that are not pictured that are involved in the DNA break repair response. The presence of elevated levels of Mdm2 delays early phosphorylation events mediated by ATM that are necessary for a rapid DNA double-strand break repair response, resulting in inefficient repair of DNA breaks, genomic instability, and ultimately tumorigenesis.

Close modal

Although the majority of our data define conditions when Mdm2 is overexpressed, which mirrors Mdm2 expression in many cancers, we did test whether Mdm2 regulated Nbs1 function when present at normal physiologic levels. Interestingly, MEFs with a mutant of Nbs1 that was unable to bind to endogenous Mdm2 seemed to have an accelerated rate of DNA repair compared with MEFs with wild-type Nbs1 (19). These observations allude to a physiologic role for Mdm2 in the regulation of the Nbs1-mediated DNA repair response, but further investigation will be needed to elucidate the precise mechanism and whether this is altered in tumor cells. In addition, a recent study identified a negative feedback loop between p53 and ATM, which is important for the pulses of p53 and Mdm2 observed during the response of a cell to DNA damage (26). Our observation that ATM activation may be blunted by increased levels of Mdm2 associating with Nbs1 adds greater complexity to this ATM-p53 feedback loop. It is interesting to speculate that the role of Mdm2 that may have evolved is one in which as DNA damage is sensed, ATM is activated, which then activates p53 leading to an up-regulation of Mdm2, which then suppresses both ATM and p53 activity. If DNA breaks persist, this cycle would begin again. It will be important to test this model and determine whether Mdm2 has a transient effect or whether it has a sustained effect that affects the rounds of DNA damage signaling that occur when DNA damage persists.

It is now apparent that Mdm2 can influence genome stability and, therefore, transformation through a p53-independent pathway. These findings have implications for the many cancers that overexpress Mdm2, as Mdm2 overexpression would promote chromosome instability and facilitate tumor development through two pathways: one involving p53 and one involving Nbs1. Additionally, our findings imply that humans who have elevated levels of Mdm2, such as those that carry the SNP309 in MDM2, may be expected to have elevated rates of genomic instability, which would contribute to an increased cancer incidence. Furthermore, multiple small-molecule inhibitors are currently being developed to combat the effects of Mdm2 by interfering with Mdm2 regulation of p53 (27). However, our recent findings would indicate that the p53-independent function of Mdm2 in its regulation of Nbs1 and DNA break repair merits consideration when designing drugs to target Mdm2 in cancers. For example, small-molecule inhibitors of Mdm2 that enhance Mdm2/Nbs1 association could be used to increase the therapeutic benefits of radiation and DNA damaging chemotherapy. Therefore, with the identification of the expanding functions of Mdm2 within a cell, a better understanding of tumorigenesis is revealed, which opens new avenues of investigation and increases hopes of improved therapeutics to overcome the negative effects of Mdm2 overexpression in human malignancies.

We apologize to colleagues we were unable to reference due to space constraints.

Grant support: National Cancer Institute grants CA098139 and CA117935 (C.M. Eischen) and University of Nebraska Program of Excellence Graduate Assistantship (A. Bouska). C.M. Eischen is a Leukemia & Lymphoma Society Scholar.

We thank the members of the Eischen laboratory and Dr. Tim McKeithan for thoughtful discussions.

1
Marine JC, Francoz S, Maetens M, Wahl G, Toledo F, Lozano G. Keeping p53 in check: essential and synergistic functions of Mdm2 and Mdm4.
Cell Death Differ
2006
;
13
:
927
–34.
2
Hollstein M, Rice K, Greenblatt MS, et al. Database of p53 gene somatic mutations in human tumors and cell lines.
Nucleic Acids Res
1994
;
22
:
3551
–5.
3
Sherr CJ. Divorcing ARF and p53: an unsettled case.
Nat Rev Cancer
2006
;
6
:
663
–73.
4
Momand J, Jung D, Wilczynski S, Niland J. The MDM2 gene amplification database.
Nucleic Acids Res
1998
;
26
:
3453
–59.
5
Watanabe T, Ichikawa A, Saito H, Hotta T. Overexpression of the MDM2 oncogene in leukemia and lymphoma.
Leuk Lymphoma
1996
;
21
:
391
–7.
6
Bond GL, Levine AJ. A single nucleotide polymorphism in the p53 pathway interacts with gender, environmental stresses and tumor genetics to influence cancer in humans.
Oncogene
2007
;
26
:
1317
–23.
7
Jones SN, Hancock AR, Vogel H, Donehower LA, Bradley A. Overexpression of Mdm2 in mice reveals a p53-independent role for Mdm2 in tumorigenesis.
Proc Natl Acad Sci U S A
1998
;
95
:
15608
–12.
8
Alt JR, Greiner TC, Cleveland JL, Eischen CM. Mdm2 haplo-insufficiency profoundly inhibits Myc-induced lymphomagenesis.
EMBO J
2003
;
22
:
1442
–50.
9
Mendrysa SM, O'Leary KA, McElwee MK, et al. Tumor suppression and normal aging in mice with constitutively high p53 activity.
Genes Dev
2006
;
20
:
16
–21.
10
Wang P, Greiner TC, Lushnikova T, Eischen CM. Decreased Mdm2 expression inhibits tumor development induced by loss of ARF.
Oncogene
2006
;
25
:
3708
–18.
11
Ganguli G, Wasylyk B. p53-independent functions of MDM2.
Mol Cancer Res
2003
;
1
:
1027
–35.
12
Cordon-Cardo C, Latres E, Drobnjak M, et al. Molecular abnormalities of mdm2 and p53 genes in adult soft tissue sarcomas.
Cancer Res
1994
;
54
:
794
–9.
13
Lu ML, Wikman F, Orntoft TF, et al. Impact of alterations affecting the p53 pathway in bladder cancer on clinical outcome, assessed by conventional and array-based methods.
Clin Cancer Res
2002
;
8
:
171
–9.
14
McDonnell TJ, Montes de Oca Luna R, Cho S, Amelse LL, Chavez-Reyes A, Lozano G. Loss of one but not two mdm2 null alleles alters the tumour spectrum in p53 null mice.
J Pathol
1999
;
188
:
322
–8.
15
Lundgren K, Montes de Oca Luna R, McNeill YB, et al. Targeted expression of MDM2 uncouples S phase from mitosis and inhibits mammary gland development independent of p53.
Genes Dev
1997
;
11
:
714
–25.
16
Harris LC. MDM2 splice variants and their therapeutic implications.
Curr Cancer Drug Targets
2005
;
5
:
21
–6.
17
Alt JR, Bouska A, Fernandez MR, Cerny RL, Xiao H, Eischen CM. Mdm2 binds to Nbs1 at sites of DNA damage and regulates double strand break repair.
J Biol Chem
2005
;
280
:
18771
–81.
18
Stracker TH, Theunissen JW, Morales M, Petrini JH. The Mre11 complex and the metabolism of chromosome breaks: the importance of communicating and holding things together.
DNA Repair (Amst)
2004
;
3
:
845
–54.
19
Bouska A, Lushnikova T, Plaza S, Eischen CM. Mdm2 promotes genetic instability and transformation independent of p53.
Mol Cell Biol
2008
;
28
:
4862
–74.
20
Carroll PE, Okuda M, Horn HF, et al. Centrosome hyperamplification in human cancer: chromosome instability induced by p53 mutation and/or Mdm2 overexpression.
Oncogene
1999
;
18
:
1935
–44.
21
Wang P, Lushnikova T, Odvody J, Greiner TC, Jones SN, Eischen CM. Elevated Mdm2 expression induces chromosomal instability and confers a survival and growth advantage to B cells.
Oncogene
2008
;
27
:
1590
–8.
22
Shiloh Y. The ATM-mediated DNA-damage response: taking shape.
Trends Biochem Sci
2006
;
31
:
402
–10.
23
de Toledo SM, Azzam EI, Dahlberg WK, Gooding TB, Little JB. ATM complexes with HDM2 and promotes its rapid phosphorylation in a p53-independent manner in normal and tumor human cells exposed to ionizing radiation.
Oncogene
2000
;
19
:
6185
–93.
24
Shrivastav M, De Haro LP, Nickoloff JA. Regulation of DNA double-strand break repair pathway choice.
Cell Res
2008
;
18
:
134
–47.
25
Rahal EA, Henricksen LA, Li Y, Turchi JJ, Pawelczak KS, Dixon K. ATM mediates repression of DNA end-degradation in an ATP-dependent manner.
DNA Repair (Amst)
2008
;
7
:
464
–75.
26
Batchelor E, Mock CS, Bhan I, Loewer A, Lahav G. Recurrent initiation: a mechanism for triggering p53 pulses in response to DNA damage.
Mol Cell
2008
;
30
:
277
–89.
27
Vassilev LT. MDM2 inhibitors for cancer therapy.
Trends Mol Med
2007
;
13
:
23
–31.