Differentiation therapy of myeloid leukemia: four decades of development.

Acute myeloid leukemia is characterized by arrested differentiation, and agents that overcome this block are therapeutically useful, as shown by the efficacy of all-trans retinoic acid in acute promyelocytic leukemia. However, the early promise of differentiation therapy did not translate into clinical benefit for other subtypes of acute myeloid leukemia, in which cytotoxic chemotherapeutic regimens remained the standard of care. Recent advances, including insights from sequencing of acute myeloid leukemia genomes, have led to the development of targeted therapies, comprising agents that induce differentiation of leukemic cells in preclinical models and clinical trials, thus rejuvenating interest in differentiation therapy. These agents act on various cellular processes including dysregulated metabolic programs, signaling pathways, epigenetic machinery and the cell cycle. In particular, inhibitors of mutant IDH1/2 and FLT3 have shown clinical benefit, leading to approval by regulatory bodies of their use. Besides the focus on recently approved differentiation therapies, this review also provides an overview of differentiation- inducing agents being tested in clinical trials or investigated in preclinical research. Combinatorial strategies are currently being tested for several agents (inhibitors of KDM1A, DOT1L, BET proteins, histone deacetylases), which were not effective in clinical studies as single agents, despite encouraging anti-leukemic activity observed in preclinical models. Overall, recently approved drugs and new investigational agents being developed highlight the merits of differentiation therapy; and ongoing studies promise further advances in the treatment of acute myeloid leukemia in the near future.

Introduction

Acute myeloid leukemia (AML) is a heterogeneous malignancy characterized by impaired differentiation and uncontrolled clonal expansion of myeloid precursors. A varied range of genetic and epigenetic alterations disrupt normal differentiation of myeloid precursors leading to maturation arrest coupled with self-renewal capacity. Despite extensive research and remarkable advances in our understanding of the molecular mechanisms governing AML pathogenesis, chemotherapy-based treatment regimens developed in the 1970s constitute the standard of care.[1] All-trans retinoic acid (ATRA) for acute promyelocytic leukemia (APL) is a notable exception, which not only revolutionized the clinical outcome of APL, but also demonstrated that inducing terminal differentiation is an effective mode of treating AML.[2] This therapeutic approach of using pharmacological agents to stimulate differentiation of immature cancer cells into more mature forms is termed differentiation therapy. Apart from its overt effect on APL blasts, ATRA also differentiated some AML cell lines lacking PML-RARA and the drug showed some activity in a few clinical trials;[3-6] however, in general, ATRA has been ineffective in non-APL subtypes of AML.[7-13] Initial studies with the HL-60 cell line also suggested that many compounds could induce terminal differentiation, including cytokines, vitamin D3 analogs and ligands of PPAR.[14] However, their therapeutic utility has been limited.[15, 16]

More recently, novel oncogenic pathways in AML have been identified, facilitating the development of targeted therapies, and heralding a new era in AML treatment. New differentiation-inducing agents that target mutant isocitrate dehydrogenase (IDH)1, IDH2 or FLT3 have gained regulatory approval. Other agents directed against distinct targets are being developed to induce differentiation of leukemic blasts. This review updates our previous description of AML differentiation therapy,[14,15] with a focus on new developments in the field, especially in the last decade.

Newly-approved targeted therapies for acute myeloid leukemia

Agents targeting two distinct oncogenic events, IDH1/2 mutations and activating FLT3 mutations, have been approved for AML therapy since 2017. These agents exert anti-leukemic activity by inducing differentiation of leukemic blasts and are exciting additions to differentiation therapy of AML.

Targeting mutant IDH1/2 enzymatic activity induces differentiation

IDH1 and IDH2 are NADP+-dependent enzymes that normally catalyze the oxidative decarboxylation of isocitrate to produce -ketoglutarate in the tricarboxylic acid cycle.[17] Somatic mutations in the active site of IDH1 and IDH2 are observed in about 20% cases of AML and 5% of myelodysplastic syndromes.[17] Mutations of IDH1 and IDH2, which largely occur in a mutually exclusive manner, result in accumulation of the oncometabolite D-2- hydroxy glutatrate (2-HG).[18] 2-HG disrupts activity of - ketoglutarate-dependent enzymes including members of the ten-eleven-translocation (TET) family of 5-methylcytosine hydroxylases and the jumonji-domain-containing group of histone lysine demethylases. This leads to a block of normal differentiation and promotes oncogenic transformation. [19-21]

Ivosidenib (AG-120; Tibsovo, Agios) is a first-in-class inhibitor of mutant IDH1 developed from the prototype compound AGI-5198.[22] Ivosidenib is effective against AML cells harboring mutant IDH1 by lowering their 2-HG levels and causing cellular differentiation.[22] A phase I dose escalation and expansion study of this drug in patients with relapsed/refractory AML harboring IDH1 mutations resulted in about 30% of cases achieving either complete remission (CR) or CR with partial hematologic recovery (CRh), with a median overall survival (OS) of 9 months (NCT02074839) (Table 1).[23] Ivosidenib induced myeloid differentiation of AML blasts, and differentiation syndrome (DS) occurred in about 11% of patients[23] (see description of ‘Differentiation syndrome’ below). A subset of patients, who achieved clearance of mutant IDH1, showed longer CR (11.1 vs. 6.5 months) and OS (14.5 vs. 10.2 months) compared with those without clearance of the mutant IDH1. Notably, patients who did not respond to ivosidenib had significantly higher rates of mutations of the receptor tyrosine kinase pathway.[23] In a recent study, ivosidenib monotherapy was shown to induce durable remissions in patients with newly diagnosed IDH1-mutant AML ineligible for standard chemotherapy.[24] Preliminary results from an ongoing phase Ib/II trial (NCT02677922) showed that the combination of ivosidenib with 5-azacytidine was well tolerated with CR and overall response rates exceeding those achieved with 5-azacytidine alone.[25] In 2018, the Food and Drug Administration (FDA) approved the use of ivosidenib for patients with IDH1- mutant relapsed/refractory AML; and subsequently the FDA approval was granted for newly diagnosed IDH1- mutant patients who are ≥75 years old or who have comorbidities precluding the use of intensive induction chemotherapy. Other inhibitors of mutant IDH1, which were effective in preclinical models include AG-881 (Agios Pharmaceuticals), IDH-305 (Novartis Pharmaceuticals), FT2102 (Forma Therapeutics) and BAY1436032 (Bayer); however, only FT2102 is being developed further for hematologic cancers (Table 1).[26]

Table 1.

Clinical trials of IDH1 and IDH2 inhibitors.

The first selective inhibitor of mutant IDH2 was AGI- 6780, which potently inhibited its aberrant enzyme activity in AML cells and induced differentiation of these cells in vitro.[27] Subsequently, a clinically useful inhibitor, enasidenib (AG-221; IDHIFA), was developed by Agios Pharmaceuticals/Celgene Corporation. This drug reduced 2-HG levels and induced differentiation of IDH2-mutant AML blasts ex vivo and in xenograft mouse models, conferring these mice with a survival advantage.[28] In a phase I/II study of relapsed/refractory AML with IDH2 mutation (either R140 or R172), the drug was well-tolerated and induced hematologic responses associated with maturation in 40% of patients, half of whom achieved CR (NCT01915498).[29] The median OS among patients with relapsed/refractory AML was 9.3 months, while it was 19.7 months in those who attained CR with enasidenib.[29] Moreover, older patients who were not candidates for standard cytotoxic therapy had a 30% response rate to enasidenib, including an 18% CR rate.[30] The differentiated neutrophils retained the IDH2 mutation, supporting differentiation as the mechanism of action of enasidenib.[31,32]

While enasidenib potently suppressed 2-HG in most patients, levels of 2-HG were not predictive of clinical response. Lack of clinical response despite suppression of 2-HG levels suggests the need for additional biomarkers of drug response. An absence of response to enasidenib could also occur because of other genetic or epigenetic abnormalities in the leukemic clones. For example, co-occurring mutations of genes in the RAS pathway are significantly associated with the lack of response to enasidenib.[32] In 2017, enasidenib received FDA approval for use in adult relapsed/refractory AML patients with IDH2 mutations. Ongoing clinical studies are evaluating the efficacy of combining mutant IDH1/2 inhibitors with either 5-azacytidine or other therapeutic agents (Table 1). Interestingly, an interim analysis of an ongoing study (NCT02677922) showed significantly higher response and CR rates with a combination of enasidenib and 5-azacytidine than with 5-azacytidine monotherapy in patients with newly diagnosed AML, who were not candidates for intensive chemotherapy.[33] Encouraging response rates noted in a preliminary analysis of an ongoing trial also support the feasibility of combining either ivosidenib or enasidenib with standard AML induction therapy in frontline treatment of the newly diagnosed patients harboring mutations of IDH1 and IDH2, respectively. [34]

Despite the clinical benefit of selective inhibitors of mutant IDH1/2, acquired resistance to this targeted therapy poses new challenges in treating AML. For instance, IDH2-mutant patients developed resistance to enasidenib by acquisition of a second mutation at residues located at the interface where enasidenib binds to the IDH2 dimer.[35] Resistance to enasidenib can also occur through acquisition of additional oncogenic mutations in other genes.[36] Similarly, secondary resistance to ivosidenib was associated with acquisition of either IDH1-second site mutations or mutations of IDH2 or RTK pathway genes.[37]

Differentiation syndrome following mutant IDH1/2 therapy

DS, originally described as retinoic acid syndrome, is a common side effect observed in APL patients treated with ATRA or arsenic trioxide. It is a serious complication that develops within 1-3 weeks of initiation of differentiation therapy in APL.[38] The pathogenesis of DS involves release of pro-inflammatory cytokines from differentiating blast cells, leading to a systemic inflammatory state. This, combined with increased vascular permeability, leads to a myriad of clinical symptoms that include unexplained fevers, weight gain, hypotension, rash, hypoxia, renal failure, dyspnea with pulmonary infiltrates and pleuropericardial effusion.[38] Although potentially life-threatening, clinical symptoms of DS are routinely managed using corticosteroids, while discontinuation of differentiation therapy might be required in cases with severe DS.[38,39]

DS, similar to that observed in APL following ATRA/arsenic trioxide therapy, has also been reported in patients treated with inhibitors of mutant IDH1/2. The most common clinical features of IDH-DS were dyspnea and pulmonary infiltrates or pleuropericardial effusion, while hypotension was the least common symptom.[40] IDH-DS was effectively managed by dose interruption and treatment with glucocorticoids and hydroxyurea.[23,29,41] In contrast to a rapid occurrence of DS in APL (median of 12 days),[38,42,43] the onset of DS was delayed in patients treated with IDH inhibitors.[23,40,41] A recent comprehensive analysis concluded that the incidence of DS in IDH inhibitor-treated patients (19%)[40] is similar to the rate observed for ATRA-treated APL (∼25%).[39,42,43] This necessitates routine monitoring for DS in patients on IDH inhibitor therapy.

Further mechanistic studies are necessary to establish how inhibition of mutant IDH1 and IDH2 relieves the block in myeloid differentiation. Inhibitors of mutant IDH1/2 potently suppress 2-HG, and this is likely to restore the activity of -ketoglutarate-dependent enzymes that regulate DNA and histone methylation. Therefore, studies need to focus on investigating the function of TET enzymes and jumonji histone lysine demethylases, and whether restoration of their function correlates with clinical outcome in subjects treated with inhibitors of mutant IDH1/2. A better understanding is also required of why the clinical response to IDH inhibitors is variable despite robust inhibition of 2-HG levels. In addition, better predictive biomarkers are needed to enhance the therapeutic utility of IDH inhibitors.

FLT3 inhibitors promote differentiation of acute myeloid leukemia blasts

FMS-like tyrosine kinase 3 (FLT3) is one of the most frequently mutated genes in AML (~30%), with two distinct alterations, either an internal tandem duplication (FLT3- ITD) in the juxtamembrane domain or point mutations in the tyrosine kinase domain (TKD). These activating mutations lead to ligand-independent, constitutive FLT3 signaling which promotes proliferation of leukemic cells. Activation of FLT3 also suppresses myeloid differentiation by inhibiting the function of CEBPA via ERK1/2-mediated phosphorylation, while pharmacological inhibition of FLT3 causes granulocytic differentiation of AML cells.[44]

Early studies with FLT3 inhibitors demonstrated clinical manifestation of differentiation with evidence of neutrophilic dermatoses.[45-47] Since then, several inhibitors of FLT3 have been investigated in clinical trials and two agents, gilteritinib and midostaurin, have received regulatory approval. In clinical studies, the differentiation-inducing effect of FLT3 inhibition is clearly evident with selective FLT3 inhibitors, gilteritinib and quizartinib.

Gilteritinib (Xospata, Astellas Pharma) is a type I FLT3 inhibitor active against both ITD and TKD mutations. In a murine xenograft model, gilteritinib induced regression of tumors expressing mutant FLT3.[48] In an initial clinical study, 49% of FLT3-mutant AML patients responded to gilteritinib. Of note, 37% of the patients who had received prior treatment with another FLT3 inhibitor responded to gilteritinib [NCT02014558 (CHRYSALIS study)] (Table 2).[49] Significantly, gilteritinib stimulated differentiation towards granulocytic and monocytic lineages in about half of the FLT3-mutant patients, who exhibited significant reduction in marrow blast percentage, even though the mutation burden of FLT3 was unchanged.[50] In this subgroup of patients, DS was reported in two patients, in whom it was resolved with glucocorticoid therapy.[50] Intriguingly, the authors noted no apparent signs of myeloid differentiation in the other half of patients treated with gilteritinib although reductions in AML blasts and bone marrow cellularity were observed. This alternate pattern of response to gilteritinib was associated with a reduction in allele frequency of mutant FLT3 and indicated that mechanisms other than induction of myeloid differentiation, might contribute to the anti-leukemic effect of gilteritinib. Promising data emerging from an interim analysis of a phase III clinical study (NCT02421939 [ADMIRAL study]) comparing gilteritinib monotherapy to salvage chemotherapy in FLT3-mutant relapsed/refractory AML patients led to the FDA (November 2018) and European Medicines Agency (EMA) (September 2019) approving single-agent gilteritinib therapy in adult patients with FLT3-mutated relapsed/refractory AML. This trial demonstrated significantly longer OS (9.3 months vs. 5.6 months) and higher response rates (CR/CRh rates of 34% vs. 15%) in gilteritinib- treated patients than in those treated with chemotherapy.[51] The efficacy of gilteritinib is also being explored in a variety of other clinical settings, including in combination with either 5-azacytidine or other chemotherapeutic agents, as well as for maintenance therapy in patients with FLT3-mutated AML (Table 2).

Table 2.

An overview of acute myeloid leukemia clinical trials of FLT3 inhibitors.

Quizartinib is a type II inhibitor that binds adjacent to the ATP-binding domain when the FLT3 protein is in its inactive conformation and, therefore, lacks activity against TKD-mutant FLT3.[52] It is a highly potent and selective inhibitor of FLT3 and showed comparatively greater efficacy as a single agent than chemotherapy in a larger phase III trial (QuANTUM-R study; NCT02039726). In this study of relapsed/refractory AML with FLT3-ITD, quizartinib monotherapy prolonged OS compared with salvage chemotherapy.[53] Quizartinib caused terminal myeloid differentiation of leukemic blasts with a surge in the number of FLT3-ITD-retaining neutrophils, and development of clinical DS.[54] Superior response rates were obtained in relapsed/refractory AML harboring FLT3-ITD when quizartinib was combined with either 5-azacytidine or low-dose cytarabine.[55] A larger, placebo-controlled trial (QuANTUM-First; NCT02668653) combining quizartinib with standard induction and consolidation chemotherapy in newly-diagnosed FLT3-ITD-positive AML patients is ongoing (Table 2).

Although gilteritinib and quizartinib act as differentiating agents, differentiation is not universally observed in all patients treated with FLT3 inhibitors, suggesting alternative mechanisms of clearing AML blasts. Midostaurin (Rydapt, Novartis) is a first-generation FLT3 inhibitor that received FDA and EMA approval in 2017 for use in adults with FLT3-mutated newly diagnosed AML when combined with intensive chemotherapy; the EMA also approved midostaurin for single-agent maintenance therapy. As a single-agent, midostaurin causes a reduction in leukemic blasts in FLT3-mutant relapsed/refractory AML patients, without achieving CR.[56] Combining midostaurin with chemotherapy significantly improved its efficacy, with a 92% CR rate being observed in FLT3-mutated patients (Table 2).[57] Moreover, in a large, phase III, double-blind, placebo-controlled trial in newly diagnosed FLT3- mutated AML, addition of midostaurin to chemotherapy resulted in significant improvements in OS and event-free survival.[58] However, none of these studies demonstrated that midostaurin exerted its anti-leukemic activity via differentiation of AML blasts. Midostaurin is a broad-range multikinase inhibitor, which targets KIT, PDGFR/, VEGFR2, and members of the serine/threonine kinase PKC family, apart from inhibiting FLT3. Thus, it is possible that its clinical efficacy results from blocking additional pro-survival and proliferation signals. Indeed, midostaurin also caused clinical responses in FLT3-negative AML and further trials are carefully investigating its clinical activity in FLT3wild-type AML (Table 2). Crenolanib is another selective FLT3 inhibitor, which produced high response rates in newly-diagnosed and relapsed FLT3-mutant AML patients, as well as those refractory to other tyrosine kinase inhibitors.[52,59] It is being investigated in several late-stage clinical studies, primarily in combination with chemotherapy (Table 2). Several preclinical studies have demonstrated that FLT3 inhibitors broadly induce apoptosis of AML cells.[60,61] Hence, FLT3 inhibitors possibly function through multiple pathways, among which induction of apoptotic cell death and terminal differentiation of leukemic blasts are prominent.

Secondary resistance frequently develops in AML patients treated with FLT3 inhibitors, and complex patterns of clonal selection and evolution have been reported.[52,59,62-64] Acquired resistance to type II FLT3 inhibitors often involves gain of secondary mutations either in the TKD or the gatekeeper residues of the FLT3 gene,[65] and type I inhibitors, such as gilteritinib and crenolanib, which have activity against both ITD and TKD mutations are effective in these patients. Overall, rational combination therapies with agents targeting known resistance pathways are needed to eradicate leukemic clones.

Other differentiation-inducing agents evaluated in clinical trials

The approval of new targeted drugs bodes well for AML differentiation therapy, but improving therapeutic options for all subgroups remains a major challenge. Epigenetic enzymes are attractive targets for AML therapy and inhibitors of histone deacetylases (HDAC), KDM1A, DOT1L and BET proteins have been extensively studied in AML. We discuss below several of these drugs, which stimulated differentiation of leukemic blasts in preclinical models, but their single-agent activity in clinical trials has been disappointing.

Inhibition of KDM1A (LSD1) promotes myeloid differentiation

KDM1A is a demethylase enzyme with specificity for monomethyl- and dimethyl-histone H3 lysine-4 (H3K4) and lysine-9 (H3K9). It is highly expressed in AML cells and is required to maintain the oncogenic program downstream of MLL-AF9.[66] Pharmacological inhibition of this enzyme potentiates ATRA-induced differentiation of AML cells and diminishes their engraftment in immune-deficient mice.[67] Combined treatment with ATRA and tranylcypromine (an inhibitor of KDM1A) caused a gene-specific increase in H3K4me2 levels and upregulation of myeloiddifferentiation associated gene expression.[67] MLL-AF9- expressing cells treated with KDM1A inhibitors also exhibit gain in chromatin accessibility at sites bound by PU.1 and CEBPA; and depletion of either of these transcription factors confers resistance of MLL-AF9-expressing cells to KDM1A inhibition.[68] Preclinical results have also demonstrated sensitivity of other AML subtypes (e.g., AML with RUNX1-RUNX1T1) to KDM1A inhibitors.[69] The efficacy of tranylcypromine combined with ATRA is being evaluated in clinical trials (Table 3).

Table 3.

Clinical trials on drugs targeting epigenetic regulators.

ORY-1001 (Iadademstat), a more potent and selective derivative of tranylcypromine, effectively reduced growth of AML cell lines, induced expression of differentiation markers and increased survival of murine xenograft models. [70] The drug is synergistic with ATRA as well as with other epigenetic inhibitors targeting DOT1L, DNMT1 or HDAC.[70] Because of favorable pharmacological characteristics and tolerability in a phase I/IIa study in relapsed/refractory AML,[71] ORY-1001 is now being investigated in combination with 5-azacytidine in elderly AML individuals who are not eligible for intensive chemotherapy (Table 3). Other KDM1A inhibitors are currently being investigated as either monotherapy or in combination with ATRA in early phase clinical trials (Table 3).

DOT1L inhibition in MLL-rearranged acute myeloid leukemia

Rearrangements of the MLL gene occur in 5-10% of AML cases and are associated with an adverse prognosis. DOT1L, a H3K79 histone methyltransferase, is recruited by MLL-fusion proteins resulting in increased H3K79 methylation at target gene loci.[72,73] This mis-targeting of DOT1L activity is essential for the transforming ability of MLL fusion proteins, and enhances expression of HOXA9 and MEIS1.[74-76] Depletion of DOT1L in MLL-AF9 leukemic cells induces differentiation and apoptosis.[75]

EPZ004777 was the first selective inhibitor of DOT1L, which inhibited H3K79 methylation and reversed the gene expression signature of MLL fusion leukemias, including downregulation of HOXA cluster genes.[74,75,77] This drug stimulated myeloid differentiation in murine and human cells harboring MLL fusions.[76-78] Pinometostat (EPZ-5676) is closely related to EPZ004777 but is more potent and has better pharmacokinetic properties.[79] However, in a phase I trial of pinometostat in advanced stage MLL-rearranged leukemia, only two of 51 patients achieved a CR, while seven non-responding patients also exhibited morphological changes consistent with myeloid differentiation (NCT01684150).[80] Despite apparent biological activity (inhibition of H3K79 methylation) in patients, pinometostat as a stand-alone therapy was not sufficient to produce a clinical benefit. This lack of therapeutic activity may be attributed to heterogeneity of MLL fusion proteins, which might result in differential sensitivity to DOT1L inhibitors.

Given their insufficient clinical efficacy as single agents, combinatorial strategies incorporating DOT1L inhibitors are being tested. Preclinical studies have shown synergistic anti-leukemic effects of combining DOT1L inhibition with either cytotoxic chemotherapy, hypomethylating agents, or inhibition of KDM1A or menin (a MLL cofactor), [81-84] suggesting that dual targeting of the MLL-fusion complex may prove superior to DOT1L inhibitor monotherapy. Phase Ib/II trials of pinometostat plus either standard chemotherapy or 5-azacytidine are ongoing in AML with MLL gene rearrangements (Table 3).

Myeloid differentiation induced by inhibition of BET proteins

Bromodomain and extra terminal (BET) family of proteins bind to acetyl-modified lysine residues of histone tails and activate transcription. A screen of RNA interference in a mouse AML model expressing the MLL-AF9 fusion and oncogenic NRAS identified BRD4, a BET family member, as indispensable for disease progression.[85]Genetic and pharmacological inactivation (JQ1 and I-BET151) of BRD4 led to downregulation of MYC, BCL2 and CDK6 in AML blasts, inhibited growth of MLL-rearranged AML cell lines and induced differentiation of primary leukemic samples and murine hematopoietic cells transformed with MLL fusions.[85-87] Another BET bromodomain inhibitor, AZD5153, which ligates two bromodomains of BRD4 simultaneously, has prominent anti-proliferative activity against AML cell lines in vitro and in xenograft models.[88] In other preclinical studies, BRD4 inhibitors exhibited anti-leukemic effects in AML driven by mutations of either IDH2, NPM1 or FLT3-ITD.[89-91] Several BET inhibitors are currently under clinical investigation (Table 3); however, emerging results from these early phase trials are disappointing. In a phase I trial of OTX015 (MK-8628), this BET inhibitor elicited a clinical response in only five of 36 AML patients, with the response lasting 2–5 months.[92] The clinical response to mivebresib (ABBV-075) was similarly inadequate and frequently associated with adverse effects including cytopenia.[93] A number of trials of BET inhibitors have been terminated because of a combination of issues including low response rates, dose-limiting toxicities and inter-patient variability in pharmacokinetic properties (Table 3).

Improved drug design and combination strategies might establish the therapeutic utility of BET inhibitors. For instance, hetero-bifunctional compounds, BET-PROTAC (proteolysis-targeting chimera, ARV-825 and ARV-771), in which a BET inhibitor (OTX015/JQ1) is linked to a ligand for an E3 ubiquitin ligase (either cereblon or von Hippel– Lindau),[94,95] have exhibited high potency in attenuating MYC and inducing apoptosis of AML cells.[96] Moreover, concomitant inhibition of both BRD4 and DOT1L had significant synergistic activity against MLL-rearranged leukemia cell lines, primary human AML cells and murine leukemia models.[97] Clearly, additional studies are needed to establish the clinical efficacy of drug combinations involving BET inhibitors.

Histone deacetylase inhibitors

Recruitment of HDAC is a major pathogenic event in AML subtypes driven by PML-RARA, inv(16) (CBFAMYH11) and RUNX1-RUNX1T1 chimeric proteins, indicating that inhibiting the activity of HDAC might be therapeutic. In preclinical studies, a combination of a HDAC inhibitor (trichostatin A) with ATRA relieved the differentiation block in ATRA-resistant APL cells. Panobinostat, another HDAC inhibitor, caused proteosomal degradation of the RUNX1-RUNX1T1 oncoprotein, triggering myeloid differentiation of leukemic cells in a murine model of human t(8;21) AML.[98] Although several classes of HDAC inhibitors showed promise in preclinical models, their clinical efficacy as monotherapy was negligible in early phase clinical trials.[99-102] Combinations of HDAC inhibitors with hypomethylating agents, chemotherapeutic agents or immunotherapy and targeted therapies are currently in different stages of clinical testing. Despite evidence of synergism between HDAC inhibitors and hypomethylating agents in preclinical and early clinical studies, large multi-arm clinical studies showed increased toxicity and decreased OS in patients treated with a HDAC inhibitor plus 5-azacytidine compared with 5-azacytidine alone.[103-105] Collectively, the data suggest that HDAC inhibitors have limited clinical benefit, illustrating the potential problem of off-target effects arising from a lack of specificity of this approach.

Differentiation therapies in early clinical development

Differentiation of acute myeloid leukemia blasts using Aurora kinase inhibitors

Aurora kinases are serine/threonine kinases that regulate chromosomal alignment and segregation during mitosis and meiosis. Expression of Aurora kinase A and B is higher in leukemic blasts than in normal CD34+ cells.[106] Inhibition of Aurora kinase A in acute megakaryocytic leukemia (AMKL) increases polyploidization of blasts and induces their differentiation.[107] Furthermore, these inhibitors significantly prolonged survival of mice in an AMKL model, as well as mice engrafted with primary human AMKL cells.[107,108] Alisertib, an Aurora kinase A inhibitor, is currently under clinical evaluation for use in the treatment of AMKL (NCT02530619). Several trials of either specific or dual inhibitors of Aurora kinase A and B showed moderate efficacy for treating other types of AML. In a phase I study, a combination of alisertib and conventional induction therapy (cytarabine and idarubicin) resulted in an 86% remission rate in newly diagnosed AML patients (NCT01779843),[109] suggesting enhanced activity of this drug combination. Thus, Aurora kinase inhibitors may hold promise in AML therapy.

Restoration of TET activity with vitamin C promotes differentiation

TET proteins are dioxygenase enzymes that catalyze oxidation of 5-methylcytosine to 5-hydroxymethylcytosine, leading to DNA demethylation. The TET2 gene is frequently mutated in AML and other hematological malignancies, as well as in clonal hematopoiesis of indeterminate potential. [110] Stimulation of the remaining wild-type allele of TET2 is a plausible therapeutic approach.

In a murine model of reversible silencing of TET2, reexpression of Tet2 in hematopoietic cells primed the cells towards myeloid differentiation and promoted cell death.[111] Vitamin C is a co-factor that activates TET proteins; and its depletion impairs the function of TET2 and accelerates tumorigenesis in a murine model of AML.[112] Importantly, vitamin C treatment mimics pharmacological restoration of TET2 activity as evidenced by increased 5-hydroxymethylcytosine levels in AML cells and Tet2-deficient mice.[111] In AML with biallelic TET2mutations or copy number neutral loss of heterozygosity of the TET2 locus, vitamin C might still enhance 5-hydroxymethylcytosine levels by potentiating TET3 activity, which can potentially compensate sufficiently for lack of TET2.[111] Serum ascorbic acid (vitamin C) levels are significantly reduced in patients with diverse hematologic malignancies,[113,114] and high doses of oral vitamin C might improve their outcome. The clinical efficacy of vitamin C combined with 5-azacytidine in patients with AML or myelodysplastic syndrome with or without TET2 mutations is currently being investigated (Table 3). Conceivably, dietary supplements of vitamin C may also help to augment TET2 activity and impede clonal hematopoietic progression in individuals with clonal hematopoiesis of indeterminate potential, especially those harboring mutant TET2.

Inhibition of DHODH relieves the block in myeloid differentiation in acute myeloid leukemia

HOXA9 is suppressed during myeloid differentiation, but its upregulation occurs in a large proportion of AML driven by different oncogenic events. In a murine model of inducible Hoxa9 expression, inhibition of dihydroorotate dehydrogenase (DHODH), a key enzyme in uridine synthesis that catalyzes conversion of dihydroorotate to orotate, induced terminal differentiation of leukemic cells.[115] In an independent CRISPR/Cas9 screen of AML cell lines, loss of DHODH also decreased growth of AML cells.[116] This suggests that targeting the de novo pyrimidine synthesis pathway may induce AML differentiation.

DHODH is neither mutated nor overexpressed in cancers; but cancer cells that are metabolically reprogrammed may have an increased dependency on de novo pyrimidine production,[117] thus creating a potential therapeutic window. Earlier attempts to target DHODH with inhibitors (e.g., teriflunomide, brequinar, and leflunomide) in several cancers were unsuccessful because of either low potency or off-target effects.[118] Subsequently, more potent and selective DHODH inhibitors (ASLAN003, BAY2402234 and PTC299) (Table 4) have been developed, which show antileukemic activity in various preclinical models;[117,119-121] and are in early phase clinical trials.

Table 4.

Clinical trials of DHODH inhibitors in acute myeloid leukemia.

The mechanism of action of DHODH inhibitors is unclear; but these drugs appear to have potential therapeutic value for the treatment of AML. However, by analogy with ATRA and the differentiating agents discussed above, it is unlikely that monotherapy will achieve durable responses.

Potential new targets for acute myeloid leukemia differentiation therapy

Recent studies have brought to light previously uncharacterized modulators of myeloid differentiation which can be targeted to overcome the differentiation block in AML. Pharmacological agents targeting these candidate modulators need to be developed to explore their clinical utility.

The CAF1 complex as a novel target for differentiation therapy of acute myeloid leukemia

CHAF1B is a subunit of the chromatin assembly factor 1 complex (CAF1) that facilitates assembly of H3-H4 tetramers at replication forks during the S phase. High CHAF1B levels are observed in leukemia and are associated with a poor prognosis.[122] Overexpression of CHAF1B enhances the leukemic potential of the MLL-AF9 oncoprotein, and its inhibition promotes differentiation of leukemic cells.[122] CHAF1B protein occupies discrete genomic loci and interferes with the CEBPA-mediated differentiation of leukemic cells;[122] and is a potential therapeutic target.

Inhibiting transcriptional repressors to re-activate the myeloid differentiation program

A common feature of AML driven by different oncogenic events is increased recruitment of transcriptional co-repressors, which affects transcription of differentiation-promoting genes. Two co-repressors (SMARCA5 and CHD4) are enriched in the PU.1/RUNX1/CEBPA master transcription factor hub specifically in AML cells.[123] An inhibitor of SMARCA5/CHD4 suppresses growth of AML cell lines via their terminal differentiation while not affecting growth of normal hematopoietic progenitors.[123]

Cell cycle regulators as therapeutic targets in acute myeloid leukemia

CDK6 is required for the growth and maintenance of an immature state of MLL-rearranged AML cells, and its suppression (by PD-0332991) induced myeloid differentiation in human and murine AML cells expressing the MLL fusion.[124] Another cell cycle regulator, CDK1 phosphorylates CEBPA, thereby inhibiting its differentiation-inducing activity. In preclinical studies, pharmacological and genetic inhibition of CDK1 relieved the differentiation block in AML cells harboring mutations of FLT3.[125] Inhibition of CDC25 using IRC-083864 can also overcome the differentiation block in FLT3-ITD AML cells..[126] Overall, cell cycle regulators may be potential therapeutic targets in AML.

Combination therapies using differentiation-inducing agents

Combinatorial therapeutic approaches may be needed to increase efficacy and reduce development of resistance in AML. In APL, ATRA is effective in achieving remission by inducing terminal differentiation, but does not eradicate the leukemic clone; however, high cure rates are attained when ATRA is combined with arsenic or anthracycline. It is postulated that ATRA differentiation therapy in APL is unable to eliminate the leukemic stem cell, and this may also be relevant to differentiation therapy of AML more widely. In fact, a recent study revealed that mature APL cells become leukemogenic upon termination of ATRA treatment, suggesting that withdrawal of differentiation therapy can lead to reacquisition of clonogenic and leukemogenic properties through de-differentiation.[127] Hence, combination regimens involving newly approved differentiation therapies (mutant IDH1/2 and FLT3 inhibitors) should be optimized for efficient clearing of the leukemic clones. Similarly, rational combination strategies might enhance the clinical utility of epigenetic inhibitors that have been largely ineffective as single agents.

Concluding remarks

Since 2017, four drugs (inhibitors of mutant IDH1/2 and FLT3) that induce differentiation of AML cells have been approved by the FDA. These novel therapeutic agents mark significant strides forward in differentiation therapy of AML. However, for other differentiation-inducing agents (e.g., inhibitors of HDAC, BET proteins, DOT1L), which showed promise in preclinical studies, early reports from clinical trials have not been encouraging. This highlights the limitations of preclinical efficacy models usually performed in immunocompromised mice lacking the immune component, as well as species-specific differences in drug pharmacokinetics and metabolism. Animal models are also limited in their ability to adequately replicate the molecular heterogeneity of AML. In addition, there is a pressing need to identify molecular subgroups which might be responsive to the new agents, as illustrated by the success of the RATIFY trial in assessing the benefit of midostaurin in FLT3-mutant AML. Another concern is that the new drugs may have been tested on limited cohorts of patients. For example, DOT1L inhibitors have been clinically evaluated in AML patients with rearranged MLL, but their differentiation-promoting activity has also been demonstrated in DNMT3Amutant AML cells in preclinical studies,[128,129] suggesting that investigation of DOT1L inhibitors could be expanded to other cohorts of patients. In addition, initial trials usually involve patients with relapsed/refractory disease which can be biologically and clinically distinct from newly diagnosed AML. Overall, this illustrates that numerous difficulties need to be overcome for translation of preclinical candidates into clinically useful differentiation therapy. APL is arguably the least molecularly complex type of AML, which may be relevant to the early success of differentiation therapy in APL.

In summary, new insights into disease pathogenesis acquired from mutational and gene expression analyses have directed AML differentiation therapy to several novel targets. For many targets, pre-clinical potential has not translated into clinical success; nonetheless, additional promising targets (DHODH, Aurora kinases) have emerged and are being actively explored. Overall, we are in an exciting era of new therapeutic targets and drugs not imagined 10 years ago.