Pharmacological activation of wild-type p53 in the therapy of leukemia

Abstract

The tumor suppressor p53 is inactivated by mutations in a majority of human solid tumors. Conversely, in leukemias p53 mutations are rare since they are only observed in a small fraction of the patient population, predominately in patients with complex karyotype acute myeloid leukemia or hypodiploid acute lymphoblastic leukemia. However, the loss of p53 function in leukemic cells is often caused by abnormalities in p53-regulatory proteins, including overexpression of MDM2/MDMX, deletion of CDKN2A/ARF, and ATM alterations. For example, MDM2 inhibits p53-mediated transcription, promotes its nuclear export, and induces proteasome-dependent degradation. The MDM2 homolog MDMX is another direct regulator of p53, which inhibits p53-mediated transcription. Several small molecule inhibitors and stapled peptides targeting MDM2 and MDMX have been developed and have recently entered clinical trials. The clinical trial results of the first clinically used MDM2 inhibitor, RG7112, illustrated promising p53 activation and apoptosis induction in leukemia cells as proof of concept. Side effects of RG7112 were most prominent in suppression of thrombopoiesis and gastrointestinal symptoms in the leukemia patients. Predictive biomarkers for response to MDM2 inhibitors have been proposed, but they require further validation both in vitro and in vivo such that the accumulated knowledge concerning pathological p53 dysregulation in leukemia, and novel molecular targeted strategies to overcome this dysregulation, can be translated safely and efficiently in to novel clinical therapeutics.

Introduction

The tumor suppressor p53 is activated in response to oncogenic signaling, DNA damage, and other forms of cellular stress [1–5]. p53 prevents the propagation of transformed cells through induction of cell cycle arrest, senescence, and/or apoptosis, and it also regulates extracellular communications between cells within the tumor microenvironment. Although the underlying mechanisms vary with respect to the origin and progression of various cancers in humans, the inactivation of p53 can be viewed as a universal feature of the tumorigenic process [6–9]. For example, TP53 is mutated in the majority of human solid tumors including triple-negative breast cancer, head and neck squamous cell carcinoma, lung cancer, prostate cancer, high-grade serous ovarian tumors, and nonmelanoma skin cancers [10]. TP53-mutant cells exert non-functional p53 effects, including dominant-negative suppression of wild-type p53 function, gain-of-function activities, and/or prion-like protein aggregation all of which can promote aberrant cell growth [11, 12]. Conversely, TP53 mutations are relatively infrequent (i.e., < 10%) in human leukemias. Yet, normal p53 function in leukemic cells is thought to be frequently abnormal as well [1–3, 13]. This may occur via regulatory protein defects like MDM2/MDMX overexpression and/or CDKN2A/ARF/ATM alterations (Fig. 1) [14–24].

Figure 1.

Impaired p53 response in leukemia. p53 transcriptional activity is suppressed by p53-regulatory proteins upstream of p53. Red ovals indicate overexpressed or activated proteins and blue ovals indicate inactivated proteins in leukemia.

The major protein regulator of p53 is MDM2, which directly binds to the protein and acts as an E3-ubiquitin ligase. MDM2 inhibits p53-mediated transcription, promotes its nuclear export, and induces proteasome-dependent degradation. MDMX (also known as MDM4 or HDM4) is a MDM2 homolog and another direct regulator of p53. MDMX lacks ligase activity, but it is able to inhibit p53-mediated transcription through its binding to the transactivation domain of the protein.

Recent advances have led to many different approaches to p53-targeted cancer therapy including TP53 gene therapy, p53 vaccines, and rescue of mutant p53 function by small molecule inhibitors. TP53 gene therapy and p53 vaccines have been extensively studied in patients with solid cancers [25, 26]. Some small molecules have also been described to restore wild-type p53 function in p53-mutant cells. The most widely investigated small molecules have been PRIMA-1 (p53 activation and induction of massive apoptosis-1)/APR-017 and its derivative PRIMA-1MET/APR-246, which are postulated to promote an active protein conformation of mutant p53, thus enhancing its DNA binding and p53-mediated apoptosis. APR-246 has shown a favorable safety profile and some clinical effects in a Phase I/II clinical study in hematological malignancies and prostate cancer [27]. A novel approach for the restoration of wild-type p53 function in p53-mutant cells uses a cell-permeable peptide that inhibits p53 aggregation [28]. The lead compound, ReACp53, has halted aggregation of mutated p53 in cancer cells, thereby restoring some of its wild-type function and anti-tumor effects. For human cancers with wild-type p53, therapy with MDM2 and/or MDMX inhibitors has been an attractive strategy to activate the protein. Several compounds and peptides have been described that block the interaction of p53 with MDM2 and/or MDMX [3, 29–37]. We will review p53 pathway abnormalities in leukemia cells and the development/use of MDM2/MDMX inhibitors to activate wild-type p53, in a nongenotoxic manner, focusing especially on those inhibitors that have entered clinical trial in patients with hematological malignancies. We will also describe some predictive biomarkers to gauge response and toxicities in patients receiving these inhibitors.

p53 regulatory abnormalities in leukemia

Acute leukemia (AML and ALL)

TP53 mutations are rare (i.e., approximately 5%) in de novo acute myeloid leukemia (AML) (Table 1) but if present, they are associated with a very poor prognosis (< 1% overall survival at 3 years) [38–40]. p53 mutations have been frequently detected in patients with complex karyotype (60 to 80%) or therapy-related AML (30%) [41–43]. TP53 mutations have not occurred in association with specific AML-related genetic abnormalities [39], but the strong association with complex karyotype attests to TP53's role as the guardian of the genome. TP53 mutations are also uncommon in acute lymphoblastic leukemia (ALL), except for cases with a low hypodiploid karyotype or MYC-translocations [44, 45]. p53 mutations have been found in most (> 90%) cases of hypodiploid ALL and approximately 60% of MYC-translocated ALL. As is the case in AML, p53 mutations have been associated with poor prognosis in ALL.

Table 1.

TP53 mutations in hematological malignancies

Acute myeloid leukemia	~ 5%
Acute lymphoblastic leukemia	~ 5%
Chronic myelogenous leukemia in chronic phase	< 5%
Chronic myelogenous leukemia in accelerated or blast crisis	20%
Ph-negative myeloproliferative neoplasms in chronic phase	< 5%
Ph-negative myeloproliferative neoplasms, transformed	40%
Chronic lymphocytic leukemia, newly diagnosed	10%
Chronic lymphocytic leukemia, refractory	50%
Adult T-cell leukemia/lymphoma, chronic	10%
Adult T-cell leukemia/ lymphoma, acute	40%
Follicular lymphoma, new	10%
Follicular lymphoma, transformed	80%

As we mentioned previously, while the majority of acute leukemias retain wild-type p53, they often acquire defects in the p53 pathway rendering p53 incapable of exerting its tumor suppressing functions. Most abnormalities occur in the p53 regulatory proteins upstream of p53, including MDM2 overexpression, MDMX overexpression, CDKN2A/ARF deletion and ATM inactivation. Importantly, most of these abnormalities eventually lead to MDM2 activation. MDM2/MDMX overexpression has been reported in AML and ALL [14–20, 46]. Leukemias exhibit high MDM2 and MDMX protein expression without increased copy numbers, as has been described in melanoma, Ewing's sarcoma, colon carcinoma and retinoblastoma. It has been suggested that leukemia-specific abnormalities increase MDM2/MDMX levels via transcriptional or post-translational stabilization [18]. MDM2/MDMX overexpression is leukemogenic, as overexpressed MDM2 or MDMX have caused hematologic malignancies in mouse models [47,48]. The CDKN2A gene encodes two tumor suppressor genes p16INK4A and p14ARF (p19ARF in the mouse). p14ARF stabilizes p53 by antagonizing MDM2, it binds to MDM2, sequesters MDM2 in the nucleolus and thereby stabilizes p53. CDKN2A/ARF deletions are common (i.e., occurring in roughly 50%) of ALL patients, with homozygous deletions as the most frequent mechanism of inactivation [22, 23]. XPO1 is involved in the nuclear export of p53, and cytoplasmic p53 is not able to act as a transcription factor. In AML, XPO1 may play some role in suppressing p53 function by nuclear exclusion of p53 [49]. Importantly, MDM2 inhibition may induce autophagy in AML through activation of AMP kinase [50]. FLT3-ITD and CBFβ-SMMHC [inv(16)(p13q22)] have shown to respectively induce the p53-deacetylating proteins SIRT1 and HDAC8 and suppress p53 function [51].

CLL

p53 mutations have been found in 5 to 15% of B-cell chronic lymphocytic leukemias (CLL), and are associated with aggressive disease that does not respond to alkylating agents or purine analogue-based therapy [52, 53]. In general, p53 mutant clones expand as disease progresses, and approximately 40% of fludarabine-refractory patients have been reported to carry TP53 mutations or 17p deletion (TP53 is located at 17p13.1). MDM2 protein overexpression has been reported in CLL [14, 16, 17]. MDM2 gene is located on chromosome 12q15. Although trisomy 12 is the most common cytogenetic change in CLL, trisomy 12 has not been associated with overexpression of MDM2. MDMX overexpression has also been reported in CLL [19]. ATM, which is activated in response to DNA double-strand breaks, synchronizes DNA repair with the induction of p53-dependent apoptosis. A deletion of the long arm of chromosome 11 (where ATM gene located) is observed in 10 to 20% of CLL patients and has been associated with poor prognosis [24]. ATM mutations in the remaining allele have been detected in 30 to 40% of CLL cases with 11q deletion. Unexpectedly, CLL with mutations of TP53 and/or deletions of 17p [54] showed high response rates of approximately 80% to the BCL-2 inhibitor ABT-199 (Venetoclax), resulting in FDA approval in May 2016 [55].

CML

p53 mutations are rare (i.e., < 2%) in the chronic phase of chronic myelogenous leukemia (CML-CP) [56], but the incidence increases to 20% when CML progresses to blast crisis (CML-BC). Isochromosome 17q, which is one of the major secondary chromosomal abnormalities closely related to blast crisis, leads to a loss of the short arm of chromosome 17 and TP53 deletion. In addition, CDKN2A/ARF deletions have been associated with lymphoid blast crisis in CML [23]. In contrast to the high incidence (20 to 40%) of CDKN2A/ARF deletions in cases of lymphoid blast crisis, the deletion has not been found in CML-CP or in myeloid blast crisis, underscoring its lineage-specific occurrence. BCR-ABL protein has been shown to induce SIRT1 which functions as a negative regulator of p53 by deacetylating several lysine sites [51].

Development of MDM2 inhibitors

Given that the majority of leukemias express transcriptionally competent, wild-type p53 that is sequestered in some way by regulatory proteins, MDM2/MDMX inhibitors have been considered for some time as an attractive therapeutic strategy for the disease. A break-through was the development of nutlin by Vassilev and colleagues, the first small-molecule inhibitor of the MDM2–p53 interaction [29], which explains why the nutlins are the most studied MDM2 inhibitors are the nutlins (e.g., nutlin-3a, RG7112, and RG7388) [29–35]. Preclinical studies have shown that p53 activation by nutlin-3a induced apoptosis in leukemia cells while inducing transient cell cycle arrest in their normal counterparts [29, 30]. It is of interest to note that higher MDM2 protein expression predicts a higher apoptotic response to nutlin-3a in AML [30]. Cell-killing efficacy of small molecule inhibitors of MDM2, such as the cis-imidazoline nutlins and the spiro-oxindoleMI compounds (e.g. MI-219, MI-773), has been described in hematological malignancies including AML [30, 57], ALL [18, 58], CLL [59–61], myeloma [62], mantle cell lymphoma [63], and some solid cancers including neuroblastoma [64] and melanoma [65] that often have wild-type p53.

RG7112 binds to MDM2 with stronger affinity (Kd, approximately 11 nmol/L) than nutlin-3a [31, 66]. Results of the Phase I Trial of RG7112 have been recently reported in liposarcoma (characterized by MDM2 gene amplification) and hematological malignancies [32, 33]. The primary end points of these small studies were to assess tumor biomarkers of p53 pathway activation and short-term toxicity. For liposarcomas, RG7112 treatment promoted an increase in intratumoral p53 and p21 protein levels and in MDM2 mRNA levels. A small decrease in Ki-67 positive cells was also observed, which could reflect reduced tumor cell proliferation. Blood concentrations of macrophage-inhibitory cytokine 1 (MIC-1), a secreted protein induced by p53, were correlated with drug exposure. Only one of the 20 treated patients exhibited a partial response and 14 had stable disease. The remaining 5 patients had disease progression.

In relapsed/refractory leukemias, such as AML, ALL, CML-BC, and CLL [33], clinical activity of RG7112 was observed, particularly in the 30 patients with AML assessed at the maximum tolerated dose. Five out of 30 evaluable patients achieved either a complete (2 patients) or partial (3 patients) response, and 9 patients had stable disease. RG7112 induced p53 target gene expression (i.e., CDKN1A, BBC3, FDXR, MDM2, PERP, ZMAT3, FAS, BAX, TFNRSF10B, TP53INP1) in circulating leukemic blasts with wild-type p53. The best response in AML patients showed a trend with high baseline mRNA levels of MDM2 [33], supporting the preclinical data that high baseline MDM2 levels could predict response to MDM2 inhibition [30].

Surprisingly, two patients with p53 mutations (R175H and S240G) responded to RG7112. The patient with the R175H mutation had 43% peripheral blasts on day 1, and suffered from tumor lysis syndrome after treatment. The peripheral blasts decreased to 0% on day 22. The patient with the S240G mutation showed a decrease in peripheral blasts from 42% on day 1 to 14% on day 28. This led to the speculation that patients with certain p53 mutations could also benefit from MDM2 inhibition [33]. However, since the presence of p53 mutation does not definitively indicate that the p53-mutated clone accounts for the majority of leukemia cells, the hematological response might simply reflect intra-tumor heterogeneity with a majority of p53 wild-type clones and a small number of mutant clones.

The clinical trial of RG7388 in relapsed/refractory AML assessed at the recommended dose revealed clinical activity in two out of nine patients [34, 67]. One patient achieved a complete response and another partial response, and a response rate of 30% in combination with AraC. Several new-and-improved MDM2 inhibitors have been developed and are currently undergoing clinical evaluation for the therapy of hematological malignancies (Table 2). The results of these clinical trials are not yet available.

Table 2.

Small molecule MDM2 antagonists in clinical trials for hematological malignancies

RG7112 (RO5045337)
RG7112	Phase I	AML, ALL, CML-BC, CLL	NCT00623870
RG7112 + Cytarabine	Phase I	AML	NCT01635296
RG7388 (RO5503781, Idasanutlin)
RG7388 + cytarabine	Phase I	AML	NCT01773408
Cytarabine ± RG7338	Phase III	AML	NCT02545283
RG7388 + Venetoclax	Phase I	AML	NCT02670044
RG7388	Phase I	PV, ET	NCT02407080
RG7388 + Ixazomib + Dexamethasone	Phase I/II	MM	NCT02633059
RG7388 + Obinutuzumab	Phase I/II	Lymphoma	NCT02624986
RG7775 (RO6839921)
RG7775	Phase I	AML	NCT02098967
DS-3032b
DS-3032b	Phase I	AML, ALL CML, MDS	NCT02319369
	Phase I	MM	NCT02579824
	Phase I	Lymphoma	NCT01877382
HDM201		AML, ALL MDS	NCT02143635

AML: acute myeloid leukemia, ALL: acute lymphoblastic leukemia, CML: chronic myelogenous leukemia, CML-BC: chronic myelogenous leukemia blast crisis, CLL: chronic lymphocytic leukemia, ET: essential thrombocythemia, MDS: myelodysplastic syndromes, MM: multiple myeloma, PV: polycythemia vera. Sourced from the ClinicalTrials.gov database.

Toxicity

Typically, MDM2 inhibitors mimic key residues (e.g., Phe19, Trp23, and Leu26) in the transactivation domain of p53 essential for MDM2 binding, and they bind to MDM2 in the p53-binding pocket to disrupt MDM2–p53 interaction in cells. Some MDM2 inhibitors show higher binding affinities to MDM2 through additional interactions with Val14, Thr 16, or His96 [68, 69]. A major concern of p53-activating MDM2 inhibitors is their on-target toxicity to normal cells. For example, p53 has long been known to play an essential role in the killing of DNA-damaged cells after exposure to ionizing radiation. For example, total body irradiation has been used in conditioning regimens prior to allogeneic stem cell transplantation to eliminate the recipient's native hematopoietic compartment. In addition to the eradication of hematopoiesis in these patients, this procedure is often associated with severe gastrointestinal toxicity due to the excessive loss of enterocytes. Therefore, cells constituting the hematopoietic and gastrointestinal systems are considered to be particularly sensitive to MDM2 inhibitor-induced p53 activation.

Long-term toxicities of MDM2 inhibitors are currently unknown. Although the lack of acute adverse effects in preclinical models of various MDM2 inhibitors have suggested that fine-tuned p53 activation may minimize toxicity to normal tissues, results of early phase clinical trials have indicated that hematological and gastrointestinal toxicities are indeed dose-limiting. In the clinical trial of RG7112 in liposarcoma, adverse events included nausea, vomiting, asthenia, diarrhea and thrombocytopenia [32]. Importantly, grade 3 or 4 adverse events were observed in 40% of the patients, the majority of which were hematological in nature. In relapsed/refractory leukemia, the most common toxicities of RG7112 treatment were again gastrointestinal and hematological, with 22% of patients experiencing grade 3 and 4 febrile neutropenia and in particular prolonged thrombocytopenia [33]. The similar toxicology profile was observed in the clinical trial of RG7338 in relapsed/refractory leukemia, in which the most common adverse events were gastrointestinal (diarrhea reported by > 85% of patients) or infection-related (> 70% of patients) [67]. Results of clinical trials of other MDM2 inhibitors will increase our knowledge about the off-target toxicity associated with MDM2 inhibition.

MDM2/X inhibitors

The small molecule RO-5963 has been reported to block both MDM2-p53 and MDMX-p53 interaction, thereby exerting high in vitro and in vivo efficacy against cancer cell lines and mouse xenografts that express high levels of MDMX. Recent studies have established that stapled peptides targeting both MDM2 and MDMX effectively enter cells in a variety of cell lines examined with wild-type TP53 to activate a TP53 response. Stapled peptides sMTide-02 and ATSP-7041 have shown highly selective and potent on-target effects in these cells [36, 37]. The stapled peptides have shown excellent pharmaceutical properties with good pharmacological distribution throughout the body, persistence in the plasma, and a slow dissociation rate from MDM2 compared to nutlin. ALRN-6924 is the newest stapled peptide that binds equally to MDM2 and MDMX and disrupts both MDM2–p53 and MDMX–p53 interactions. ALRN-6924 has recently entered Phase 1 clinical trials for patients with advanced solid tumors or lymphomas with wild-type TP53 (NCT02264613).

Predictive biomarkers for response to MDM2 inhibitors

The three major biomarkers that have been proposed to determine responses to MDM2/X inhibitors are p53 status, MDM2 levels, and MDMX levels [32–34, 70]. Interestingly, MDM2 overexpression, MDMX overexpression, or p53 mutation are often mutually exclusive in various solid tumors including liposarcoma, glioblastoma, melanoma, bladder, prostate, lung, and breast cancers [71]. It has been reported that increased expression of MDM2 predicts higher cell sensitivity to MDM2 inhibitors [30, 58]. On the other hand, cells with MDMX overexpression or TP53 mutations generally show poor inhibitor sensitivity [19, 29–35]. By interrogating high-throughput cell line and patient-derived tumor xenograft sensitivity data with genomic profiling data, a gene expression signature consisting of 13 p53 transcriptional target genes has been identified that predicts sensitivity to the MDM2 inhibitor NVP-CGM097 [72]. However, the study contained a significant number of miss-classified p53-mutant cell lines, and removal of these lines unfortunately abolished the predicative power of the screen [73].

Zhong et al. [74] established an in vitro MDM2 antagonist therapy-predictive mRNA signature score, by assessing genome-wide associations between growth inhibitory effects of RG7112 among 287 human cancer cell lines and pretreatment RNAseq-derived transcript levels. The signature score reflects transcript levels of four p53 target genes MDM2, XPC, CDKN2A, and BBC3. The score has been successfully validated in AML clinical studies with RG7112 and RG7388. Interestingly, although none of the four genes alone showed significant association with patient outcomes, an association with response was identified when a multi-gene algorithm was applied.

We have recently reported a novel method to generate tumor-type-specific gene signatures based only on a dataset of primary AML samples using the random forest method [75]. The majority of genes participating in predictive gene signatures for MDM2 inhibition were not direct targets of p53. Combining those genes and p53-related genes empowered the predictive performance, compared to that of each gene set, indicating that the usage of a p53-related gene signature combined with p53-independent algorithms may increase sensitivity and specificity. An alternative approach would be to select for tumors with MDM2 amplification given the mutual exclusivity of p53 mutations and MDM2 amplification. However, MDM2-amplified liposarcomas did not show high sensitivity to MDM2 inhibition [32]. At present, the above-mentioned biomarkers, while logical in their choice, may not sufficiently predict which patients may benefit from MDM2 inhibitor therapy, but testing of novel biomarker algorithms is under way in several ongoing clinical trials.

Resistance mechanisms to MDM2 inhibitors

TP53 mutant clones may be present in patients with hematological malignancies before treatment [50, 76]. Thus, a selection of p53 mutant clones during therapy could be a potential mechanism of resistance to MDM2 inhibitors. There are a few reports of an increased incidence of p53 mutations following prolonged exposure to MDM2 inhibitors [77–79]. Point mutations in the p53-binding pocket of MDM2 may also confer resistance to MDM2 inhibitors, and a selection of MDM2 mutants that are able to interact with p53 in the presence of nutlin has been reported [80]. High MDMX levels have been reported to confer resistance to MDM2 inhibitors [81–83], since MDM2 inhibitors are not able to block the MDMX-p53 interaction.

The major route of p53-induced apoptosis is via induction of apoptotic BH3 proteins and disruption of the balance between anti-apoptotic and pro-apoptotic BCL-2 family proteins. It is therefore anticipated that acquired changes in cellular levels of pro- and anti-apoptotic BCL-2 family proteins may lead to resistance to MDM2 inhibitors. However, a recent BH3 profiling study has shown that nutlin-induced apoptosis was minimally affected by specific properties of BCL-2 dependency in leukemia cells [84]. At present, no resistance mechanisms to MDM2 inhibitors have been established in the clinical setting. It is however conceivable that the elimination/reduction of wild-type TP53 cells will reveal TP53-mutant clones that were initially present at low, undetectable frequencies when assayed with standard molecular techniques. A combined strategy targeting both wild-type and mutant TP53 concomitantly may be advisable.

Numerous combinatorial therapies have been developed pre-clinically and some are now undergoing clinical testing (see Table 2). The two main apoptosis regulators p53 and BCL-2 were targeted with nutlin-3a and ABT-737, resulting in major synergism [85]. The combination of ABT-199 (BCL-2 specific) and RG7388 (MDM2 specific) is presently being investigated in AML. MDM2 inhibitors have also reported to work well in combination with other drugs, such as the mitotic inhibitor vincristine, CDK inhibitors such as roscovitine, Aurora kinase inhibitors, FLT3 inhibitors, XPO1 inhibitors and DNA-damaging agents such as doxorubicin [3].

Concluding remarks

Pharmacological activation of wild-type p53 is a logical therapeutic strategy for leukemia where p53 is inactivated by abnormalities in p53-regulatory proteins. Several small molecule MDM2 inhibitors and stapled peptides targeting both MDM2 and MDMX have entered clinical trials, and the results have demonstrated their clinical efficacy and toxicological properties. As advances in p53-targeted therapy start to translate successfully into novel p53-based therapeutics, the clinical practice of treating leukemias will certainly incorporate such strategies.

Figure 2.

Negative correlation of MDM2 and p53 in AML. Three-dimensional surface plot of p53 against MDM2 and XPO1 in 511 patient-derived AML samples is shown [48]. High levels of MDM2 are associated with low levels of p53, implying a pathological role of MDM2 in suppressing p53 function in AML. There is a significant relationship (P = 0.0002) between p53, MDM2 and XPO1, raising the possibility that the p53 nuclear exporter XPO1 also regulates p53 in AML. The color scale refers to the expression (on the log 2 scale like the rest of the data) of the z-axis component (i.e., p53).

Highlights.

• MDM2 and MDMX are direct, negative regulators of the tumor suppressor p53

• MDM2 and MDMX represent important therapeutic targets in leukemias

• MDM2 inhibitors and stapled peptides targeting MDM2/MDMX have been developed

• Clinical trials have demonstrated on-target activities and toxicities of MDM2 inhibitors

• Biomarkers are needed to predict responsiveness to MDM2/MDMX inhibitors