Targeting the IDH2 Pathway in Acute Myeloid Leukemia

Acute myeloid leukemia (AML) is diagnosed in approximately 20,000 patients each year in the United States (1). The prognosis is poor; roughly half of patients die from disease or related complications per year (1). Treatment options for AML are limited and have consisted primarily of cytarabine with anthracyclines for those able to tolerate intensive therapy (2) and, in the United States, hypomethylating agents for unfit patients (3). Cytogenetic subtypes have been the most powerful prognosticators of outcomes (4); however, given roughly half of AML patients have normal cytogenetics, mutational analysis for prognostication purposes is important (5–7). Among the initial mutations discovered via genomic sequencing was isocitrate dehydrogenase (IDH) 1 (6), which led to great interest in this family of proteins. IDH1 and IDH2 are enzymes that mediate the oxidative decarboxylation of isocitrate to α-ketoglutarate, resulting in the production of NADPH, and are an important factor in the regulation of oxidative stress (8). IDH1 is a cytosolic enzyme, whereas its homolog, IDH2, is found in the mitochondria; both perform a critical step in the Krebs cycle (8). The first association between IDH mutations and malignancy was reported in gliomas (9, 10); in AML, it has been determined that IDH1/2 mutations occur in roughly 20% of patients (7, 11, 12). In this review, we will highlight the IDH2 pathway in AML, with a specific focus on new targeted therapies.

Pathologic IDH mutations are missense mutations of catalytically active arginine residues. These mutations are located in codon 132 of exon 4 (R132) in IDH1, and codons 140 and 172 (R140 and R172) of exon 4 in IDH2 (13), with R140 being the most common mutation (14–16). IDH2 mutations are more frequent in AML (17), and IDH1 and IDH2 tend to be mutually exclusive (18, 19). Mutations in IDH catalyze the reduction of α-ketoglutarate to (R)-2-hydroxyglutarate (2HG), an “oncometabolite” (20, 21), so called because it interrupts several cell pathways that lead to oncogenesis. 2HG induces epigenetic changes by inhibiting demethylases such as JMJD2A (22, 23) and competitively inhibiting several dioxygenases including TET2, a demethylating enzyme (22). As functional proof that these are overlapping pathways, IDH2 and TET2 mutations have been found to be mutually exclusive (24). Both IDH2 mutations, R172 and R140, lead to the accumulation of extremely high levels of 2HG (20, 25), resulting in the reduction of α-ketoglutarate, leading to reduced prolyl hydroxylases and upregulation of HIF1α (26), disrupting adaptation to hypoxia. Finally, high levels of 2HG result in a differentiation blockade (27), contributing to the pathogenesis of AML. Figure 1 summarizes the mechanism of IDH2 mutations.

Several studies have tried to elucidate whether there is an association between IDH mutations and cytogenetic abnormalities, or with other mutations. It has been observed that IDH mutations are usually associated with cytogenetically normal patients and that they often co-occur with NPM1 mutations (7, 11, 13, 18, 28–31). Although IDH mutations tend to be founder mutations (28, 32), a recent study presented at the American Society of Hematology annual meeting reported that up to 16% of IDH-mutant patients in a sample size of >6,000 had “progressor mutations” (subclonal mutations rather than founder mutations). In this study, IDH1/2 mutations were mostly associated with DNMT3A, SRSF2, NPM1, ASXL1, and RUNX1 mutations, although progressor mutations were more commonly associated with TP53 and FLT-3-ITD (33). Data to support IDH as a prognostic marker have been conflicting and inconclusive. Patel and colleagues and others showed IDH2 R140Q mutations had an overall improved prognosis when associated with NPM1 mutations (7, 16). There are data to suggest that in patients with normal cytogenetics, IDH status did not affect overall survival (OS; refs. 11, 14, 15), whereas other data suggest it was associated with a lower remission rate (13) and lower OS (30, 34). In an effort to understand the impact of IDH mutations on prognosis based on all prior data, Xu and colleagues performed a meta-analysis, which showed when IDH mutations were pooled together, there was no change in prognosis in AML (31). IDH2 mutations confer a better prognosis in intermediate-risk AML patients, but not within the subcategory of AML patients with a normal karyotype. The authors also noted the association between IDH2 and NPM1 mutations conferred a benefit based on several studies previously mentioned (31). It is possible that the conflicting reports may be a result of comutational data, or in the case of IDH2, due to the analysis of IDH2 R140 and R172 together when they may prognosticate for different outcomes. This is corroborated by data showing R140 and R172 are exclusively comutated with particular genes, and they result in different levels of 2HG suppression (35). Table 1 summarizes studies that highlight the prognostic outcomes of IDH2 mutations. Although it is inconclusive whether IDH2 acts as a prognostic marker at diagnosis, 2HG levels do appear to have a prognostic significance after treatment, with higher levels of 2HG resulting in a decrease in OS (36, 37). IDH mutations have been largely studied in the adult population, but they also occur in pediatric AML cases. Although less frequent in this population (about 3%), they are also more commonly associated with normal karyotype and with high 2HG levels. These mutations confer no prognostic significance, although studies suggesting this are likely underpowered due to the low percentage of patients having this mutation (38).

Given the incidence of IDH mutations in AML, there was interest in developing a specific inhibitor of cells with mutant copies of IDH (25). The first IDH2 inhibitor, AGI-6780, targeted the IDH2 R140 mutation and induced differentiation of primary human IDH2-mutated AML cells, while decreasing 2HG (39). Another IDH2-mutant small-molecule inhibitor, enasidenib (previously known as AG-221), was developed and was shown to be highly selective with improved solubility and oral availability (40). Initial laboratory studies showed that enasidenib reduced serum 2HG by >90% and allowed primary myeloid cells to differentiate (40). This led to the clinical development of enasidenib, in which reduction of 2HG levels was seen in treated patients (35, 41).

In a phase I/II clinical trial, 239 relapsed/refractory AML patients 18 years or older were treated with enasidenib in a dose-escalation (doses ranging from 60 to 650 mg QD) and dose-expansion phase. Enasidenib had an overall response rate of 40%, with a median response duration of 5.8 months and a median OS of 9.3 months. Nineteen percent of patients achieved complete remission; these patients had a median OS of 19.7 months. The most common adverse event was indirect hyperbilirubinemia, which occurred in 35% of patients (41) and may be due to off-target inhibition of the UGT1A1 enzyme and had no clinical sequelae (41).

Significant toxicity included differentiation syndrome, tumor lysis syndrome, and leukocytosis, which all occurred in <7% of the patients (41). Differentiation syndrome is a unique complication that can occur with effective therapy in patients with AML. As described above, high levels of 2HG from IDH2-mutated cells can result in a differentiation blockade, and treatment with enasidenib can induce rapid differentiation of leukemia cells, resulting in differentiation syndrome, which has disparate clinical features (42). In addition, treatment-related leukocytosis, in concert with differentiation, can make it difficult to discern whether patients are progressing or responding. The treatment of differentiation syndrome involves corticosteroids, hydroxyurea for leukocytosis, and occasionally, a treatment interruption (43, 44). Based on promising results compared with historical controls in this nonrandomized clinical trial, enasidenib received FDA approval for relapsed and refractory (RR) IDH2-positive AML in 2017.

The 2017 American Society of Hematology meeting highlighted several studies using enasidenib in the upfront setting; none have been published in manuscript form. A phase I study used enasidenib in combination with induction chemotherapy in 56 adult patients with IDH2 mutations who had not previously been treated. Patients with IDH2 mutations were treated with either enasidenib plus cytarabine and daunorubicin or idarubicin, followed by consolidation with cytarabine plus enasidenib for up to four cycles. They were then maintained with single-agent enasidenib (100 mg) for up to 2 years. Ninety-one percent of patients had >1 grade 3 or higher TEAE with febrile neutropenia and elevated bilirubin being the most common adverse events. They had a total response rate of 62% with a complete remission (CR) of 50% (45). In another analysis, 38 untreated patients from the dose-escalation/-expansion study of predominantly relapsed AML patients who received single-agent enasidenib were analyzed separately; none were deemed candidates for standard treatment. The adverse event profile was similar to that of the relapsed population. The overall response was 32% with a CR of 18%. The median OS in these patients was 11.3 months (46). This regimen is also being explored in the Beat AML master trial, in which newly diagnosed older IDH2-mutated patients receive single-agent enasidenib, with the option to add azacitidine after the first cycle of treatment (NCT03013998, BAML-16-001-S3). The combination of enasidenib and azacitidine was also reported in the upfront setting; 6 patients with IDH2 mutations received enasidenib (100 or 200 mg) in combination with azacitidine. The majority of the patients had an R140 mutation, and they had a median treatment of 9 cycles. The study had an overall response rate of 4 out of 6 patients, and 2 patients in the 100-mg enasidenib arm achieved a complete remission (47).

Note that 2017 was a year of several FDA approvals for new agents in the treatment of AML. Among these agents is midostaurin, a multitargeted kinase inhibitor, which acts as an inhibitor against FLT3-ITD/FLT-3-TKD. Midostaurin in combination with induction and consolidation chemotherapy, followed by maintenance therapy, showed an increased OS from 25.6 versus 74.7 months; however, unlike enasidenib, midostaurin has not been shown to be effective as a single agent in the treatment of AML (48). Among other therapies with single-agent activity is sorafenib, a multikinase inhibitor that has been used in the treatment of relapsed and refractory (RR) AML. In a phase I study including 50 patients, an overall response was achieved in 10% of patients with FLT3-ITD mutations, with 34% of patients showing reduction of blasts, and duration of response lasting for at least two cycles (49). It has also been shown to improve duration of remission in FLT3-ITD patients treated with sorafenib after allogeneic transplantation (50). Another agent to gain reapproval from the FDA is gemtuzumab ozogamicin (GO), an anti-CD33 antibody used for the treatment of CD33⁺ newly diagnosed AML patients. It was approved for administration in combination with induction chemotherapy, as it showed improvement in event-free survival (51). As a single agent, it can be used in newly diagnosed AML patients not fit for chemotherapy, with an improved median survival from 3.6 to 4.9 months (52). Treatment of RR AML with GO resulted in 26% of the patients achieving a CR with median duration of 11 months and OS of 8.4 months (53). Single-agent azacitidine has been shown to be beneficial in the elderly population with untreated AML, as presented in a subgroup analysis of the AZA-AML-001 study. In this study, elderly patients were randomized to azacitidine versus conventional care regimens (best supportive care, low-dose cytarabine, or intensive chemotherapy). Azacitidine resulted in a CR rate of 18%, with OS of 19.1 versus 13.4 months with conventional care regimens (54). Among experimental agents that appear to have promising single-agent activity is ivosidenib, an IDH1 inhibitor. Preliminary results from a phase I clinical trial on IDH1-mutated relapsed/refractory AML patients showed an overall response rate [CR/CR with incomplete count recovery (CRi)] of 30%, with a median duration of 8.2 months, which appears comparable with enasidenib efficacy in the IDH2-mutated population (55). Although impossible to truly compare efficacy across trials, based on response rates of above-mentioned agents, enasidenib has comparable, if not better, response rates to other single-agent therapies and appears to be a good option in a selected population of patients.

AG-881 is an inhibitor of both IDH1 and IDH2 mutations and has the added benefit of penetrating the blood–brain barrier. There is currently an ongoing phase I, multicenter study with AG-881 to evaluate safety profile in advanced hematologic malignancies with IDH1/2 mutations (NCT02492737); no results have been reported to date.

B-cell Lymphoma-2 (Bcl-2) is an antiapoptotic protein that is overexpressed in several malignancies (56, 57), including AML (58), and high levels of Bcl-2 have oncogenic properties via impaired apoptosis (59). Venetoclax is a small molecule that inhibits Bcl-2, and it gained FDA approval for chronic lymphocytic leukemia in 2016. Although not a direct inhibitor of IDH, IDH-mutated cells may be sensitive to venetoclax. Chan and colleagues first reported the dependence of IDH-mutant cells on the Bcl-2 pathway (60) by using a synthetic lethal approach, taking advantage of the fact that cells that express an oncogenic mutation often depend on a particular pathway and/or other genes for survival. Using this screen, IDH-mutant AML cells were noted to depend on Bcl-2 for survival and were sensitive to venetoclax (60). In a study of relapsed/refractory AML patients that used single-agent venetoclax, the overall response rate for 32 patients was 19%. However, patients with IDH1/2 mutations had a 33% CR/CRi rate (61), suggesting venetoclax may be a promising agent for AML patients with IDH mutations. Off-label venetoclax has also been used in combination with low-intensity chemotherapy in patients with RR myeloid malignancies. A recent study looked at with RR AML patients receiving venetoclax in combination with hypomethylating agents or low-dose cytarabine. Patients received a median of 2 cycles of treatment. Of 11 patients with IDH mutations, 27% responded to treatment, with 1 CR, 1 CRi, and 1 morphologic leukemia-free state. One patient quickly cleared peripheral blasts but had infectious complications leading to hospice care, and another patient had greater than 50% blasts reduction, without full hematologic recovery (62).

The outcome for patients with AML has remained poor, in part due to a lack of new effective treatments. In a very short time from the discovery of IDH as a pathogenic mutation, there is now an FDA-approved targeted therapy for patients with this mutation, a very impressive accomplishment for a field that has had few victories over many years. IDH2 has been shown to be an early and stable mutation, seen in about 8% to 19% of patients with AML. It is unclear whether and how IDH2 mutations affect prognosis, but it is suspected that oncogenesis relies on the overproduction of the oncometabolite 2HG. The specific and potent IDH2 inhibitor enasidenib decreases 2HG and can result in differentiation of AML cells, leading to clinical responses. However, although nearly all patients treated with enasidenib have rapid and profound reductions in 2HG, not all patients respond to treatment, suggesting we do not fully understand the mechanism of action of this agent. Furthermore, not all patients achieving a CR after enasidenib treatment have a complete molecular response (35). While trying to elucidate why some patients are nonresponders, recent data suggest patients with R172 mutation have less suppression of 2HG when compared with R140, and patients with comutated genes had lower responses to enasidenib. Particularly, patients with comutation of RAS pathways achieved less CR (35). Enasidenib has been primarily studied in the relapsed/refractory setting, with overall response rates around 40%. Although this represents an improvement over alternative treatments with respect to activity and tolerability, given that IDH2 is an early, stable mutation and enasidenib is a highly effective IDH2 inhibitor, one has to consider whether improvements will be possible from a strategy of targeting a single pathway or mutation in this very heterogeneous disease. It is conceivable that multiple mutations could be simultaneously targeted. However, considering relatively few mutations are druggable, the extreme heterogeneity of this disease, and the potential toxicity of multiple concomitant therapies, this is unlikely to be feasible. The role of targeted therapies therefore must be more clearly defined. Focusing on targeting IDH2 in the upfront treatment setting, likely in combination with conventional intensive and nonintensive treatment approaches, may result in longer remission durations, but requires upfront knowledge of the patient's mutational profile, which can be difficult for patients who present in extremis. Alternatively, targeting IDH2 may be best used as a tool to achieve a remission that would allow a patient to proceed to a potentially curative therapy, such as a stem cell transplantation, with chemorefractory disease that otherwise prevents such a maneuver. Otherwise, it may remain a medium-term and relatively nontoxic strategy for a minority of patients with RR disease. Given the clinical activity in this area, the coming years will better elucidate the ideal scenario for targeting IDH2.

D.A. Pollyea reports receiving commercial research grants from Agios and is a consultant/advisory board member for Agios and Celgene. No potential conflicts of interest were disclosed by the other author.