In the last ten years, advances in deciphering genomic landscape have increasingly contributed to the refinement of prognostication in acute myeloid leukemia (AML), incorporating the mutational status of some genes (1–4) in the WHO classification of AML.

The current (2017) European Leukemia Net (ELN) classification recommends the stratification of newly-diagnosed AML based on the mutational status of five genes NPM1, CEBPA, FLT3, ASXL1, TP53, and RUNX1 (2). Indeed, mutation in both NPM1 and CEBPA genes have been associated with a favorable prognosis, in the absence of a concomitant internal tandem duplication within the FLT3 gene (FLT3-ITD) in NPM1 mutated cases. Moreover, mutational load appears to be an additional critical factor, which is why the ELN recommends a cut-off of 0.5 for the allelic ratio of mutant to wildtype FLT3. RUNX1, TP53, and ASXL1 mutations all confer an unfavorable prognosis as do FLT3-ITD mutations (at a high allelic ratio of > 0.5 in the absence of NPM1 mutations) (2).

Despite the improvements in understanding its genomic basis, AML is still a therapeutic challenge for older adults, not eligible to allogeneic stem cell transplantation, with less than 10% chance of long-term survival (2, 5). The therapeutic scenario of AML is transitioning from a standard chemotherapy regimen in all patients toward individualized therapeutic strategies (6–9).

The discovery that some AML blasts and glioma cells carry specific mutations in isocitrate dehydrogenase isoform 1 (IDH1) and 2 (IDH2), resulting in the generation of a particular metabolite — the (R)-enantiomer of 2-hydroxyglutarate ((R)-2HG), led to development of small molecular inhibitors which effectively inhibit 2HG production; effectively targeting the aberrant metabolism of AML blasts (10).

Enasidenib and ivosidenib, two small molecule inhibitors targeting IDH2 and IDH1 gene mutations, respectively, have been recently approved by FDA for their efficacy and safety. This review highlights recent advances on metabolic deregulation in AML and novel strategies for tailored therapy targeting mutations in IDH1 and IDH2.

AML arises from genetic abnormalities in hematopoietic stem or progenitor cells, responsible of uncontrolled growth and accumulation of neoplastic blasts in the bone marrow leading to organ failure and often death. Somatic mutation in IDH1 and IDH2 occur respectively in about 6%–16% and 8%–19% of AML patients (11), are frequently associated with normal karyotype and nucleophosmin (NPM1) gene mutations (12), and are without a clear prognostic relevance (13), unless their increased frequency with advanced age (2, 3, 14–16).

In AML patients, IDH1/IDH2 gene mutations are heterozygous missense mutations involving a single arginine (R) residue in the enzyme active site (9), R132 in IDH1 and R140 or R172 in IDH2, restructuring the enzyme and leading to a reduced affinity of the mutant enzymes for isocitrate while increased affinity for α-ketoglutarate (αKG) and nicotinamide adenine dinucleotide phosphate (NADPH) with production of R-2HG (17), as extensively reviewed elsewhere (18).

7Under physiological conditions the IDH3 isoform generates NADH as the canonical product of the Krebs cycle, while IDH1 (the isoform localized in the cytoplasm and peroxisomes) and IDH2 (the isoform localized in mitochondria) catalyse the oxidative decarboxylation of isocitrate to generate αKG and carbon dioxide (CO₂) and to produce reduced NADPH from NADP+, playing a critical role for the maintenance of the intracellular reduced glutathione pool and preserving cellular homeostasis (19). Mutants of IDH1/2 redirect carbon metabolites away from the Krebs cycle and oxidative catalyzing the conversion of α-ketoglutarate (α-KG) to D-2-hydroxyglutarate (D2HG), promoting glutaminolysis pathway, to provide carbon skeletons required for anaplerosis to fuel the Krebs cycle (20) and reduce the enzymatic activity of the wild-type isoforms (20).

mIDH1/2 cells still need αKG to produce D-2HG, even if on the other hand they restrict αKG synthesis as consequence of reduced glycolytic influx and impairment of Krebs cycle (20).

Moreover, 2DHG acts as an oncometabolite through a variety of mechanisms.

First, D2HG directly inhibits α-KG-dependent enzymes such as the ten-eleven translocation (TET) family enzymes (21).

Second, D2HG accumulation additionally causes genetic instability and contributes to tumorigenesis, either directly by inhibiting α-KG-dependent alkB homolog (ALKBH) DNA repair enzymes or indirectly by altering the expression of DNA repair genes (22). Uncontrolled increase of D2HG levels inhibits αKG-dependent lysine demethylases in a competitive manner, thus to increase histone methylation in a variety of cell line models (22–24). Further D2HG-mediated inhibition of cellular differentiation promotes in turn the pathological self-renewal of stem-like progenitor cells, favoring the malignant transformation (19, 25). Additional changes in the epigenetic machinery of hematopoietic progenitors can occur via several other molecular mechanisms, that could explain why IDH1 and IDH2 mutations are mutually exclusive with Tet methylcytosine dioxygenase 2 (TET2) mutations (26).

IDH1 mutations are found in almost all FAB (French-American-British classification) AML subtypes, associated to elderly age, diploid or intermediate cytogenetics, increased platelet counts in mutated cases (27). Experimental observations collected in vitro, where IDH2 mutants can block the differentiation of HSPCs (26, 28, 29) and in vivo suggest that IDH1/IDH2 mutations act as canonical class II mutations, required to cooperate with class I mutations to promote AML.

Conditional knock-in mice carrying IDH1^R132H mutation have a characteristic hypermethylation signature, similar to that one observed in human IDH1- or IDH2-mutant AML. Mutant mice do not develop leukemia and had a similar survival as wild-type mice, but they display a myeloproliferative phenotype, characterized by increased number of early hematopoietic progenitors, splenomegaly and anemia with extramedullary hematopoiesis, along with a partial blockage in myeloid differentiation (28).

In hematopoietic stem and progenitor cells (HSPCs) obtained from two mouse models incorporating two common class I mutations observed in human AML, FLT3-ITD and Nras^G12D, IDH2^R140Q or IDH2^R172K mutations conferred inferior overall survival and drive aggressive AML. mIDH2^R172 murine AMLs displayed the histopathological and molecular features of the human disease, including chemoresistance phenotype associated to poor patient survival (29).

In mice, mIDH2^R140Q in hematopoietic tissues was not sufficient for the development pf AML phenoype, suggesting the requirement of additional driver mutations, like overexpression of HoxA9 and Meis1a or mutations in FMS-like tyrosine kinase 3 (FLT3) to drive acute leukemia in vivo (30).

Combining the loss-of-function due to miRNA Mir142 with IDH2^R140Q mutation in vivo, only recipients of double mutant cells developed fatal leukemia, since HOXA cluster genes in myeloid progenitors were alleviated by Mir142 loss-of-function (31).

Thus, IDH1/IDH2 mutations play a double role in leukemogenesis: first, mutants induce alterations in the pattern of histone modifications and aberrant DNA methylation, with consequent accumulation of epigenetic aberrancies, associated with reduced TET2 activity (26); second, mutants can rewire metabolism of AML blasts (25), with sustained inhibition of the activity of cytochrome c oxidase (COX) in the mitochondrial electron transport chain (20) and the glutamine addiction of AML cells for survival (32). Small molecule inhibitors of mIDH1 and mIDH2, and vitamin C supplementation [cofactor of TET2 proteins (33, 34)] can abrogate the production of 2HG or enhance TET2 activity, respectively, restoring DNA methylation patterns and myeloid cell differentiation (35).

The same pathways can be elicited by metabolic derangements in IDH1/IDH2 wild-type blasts. For example, overexpression of branched chain amino acid in leukemia cells decreased intracellular αKG levels and caused DNA hypermethylation through altered TET activity, leading to a DNA hypermethylation phenotype similar to cases carrying an IDH mutant (36).

Ivosidenib (Tibsovo; Agios Pharmaceuticals, Inc.; formerly AG-120) is the first-in-class, selective, allosteric IDH1^R132 inhibitor (37), derived from AGI-5198, which could not enter clinical studies due to its poor pharmaceutical properties (37). Ex vivo, AG-120 can induce differentiation of primary mIDH1 blasts obtained from AML patients, as shown by enhanced ability to form differentiated colonies in methylcellulose assays, increased levels of cell-surface markers of differentiation, and increases in the proportion of mature myeloid cells (37). In vivo, in a human IDH1-mutated tumor xenograft model it showed potent activity in lowering tumor levels of R-2-HG, good pharmacokinetics properties and was well tolerated (37).

Enasidenib (Idhifa; Celgene Corp.; formerly AG-221) is the first-in-class, orally available, small molecule, selective inhibitor of mIDH2. AG-221, developed by Agios Pharmaceuticals in partnership with Celgene, is specific for mutants IDH2^R140Q and IDH2^R172K, by binding to its allosteric site and stabilizing the homodimer conformation preventing the conformation change required for its catalytic action and the production of the R2HG (46).

Preclinical studies conducted in vitro and in vivo showed that AG-221 could induce differentiation due to increased granulation, phagocytic activity and expression of mature myeloid markers with concomitant changes in global DNA hypermethylation, although retention of mIDH2 in differentiated cells. Survival benefit in primary human IDH2-mutated AML xenograft mice compared to mice treated with cytarabine was dose-dependent, associated with reduction in R-2-HG, providing strong rationale for a phase I/II clinical trials (46).

Evaluation of the IDH inhibitors in rational combinations are ongoing, with a goal to more fully eradicate the leukemic clones. In this perspective, the clinical benefit of adding ivosidenib or enasidenib to induction, consolidation and maintenance therapy for patients with newly diagnosed IDH1/IDH2-mutated AML is being evaluated in a phase 3, double-blind, randomized, placebo-controlled study presently recruiting (ClinicalTrials.gov NCT03839771).

In the setting of either newly diagnosed or relapsed/refractory AML patients, elderly or not candidate to intensive chemotherapy, mIDH1/mIDH2 inhibitors could be used in combination with either the hypomethylating agent 5-azacitidine or the bcl-2 inhibitor venetoclax. Indeed, excess D2HG produced by IDH mutant AML blasts confers susceptibility to venetoclax due to ETC dysregulation, suggesting why patients carrying IDH1/IDH2 mutations have more robust responses to the combination treatment of venetoclax and azacitidine. While IDH inhibitors appear to reduce this sensitivity in vitro, PDX models and initial clinical trial data of the combination of venetoclax and ivosidenib are encouraging (58, 59).

Results on the phase I multicenter study (NCT02632708), evaluating the safety of enasidenib (100 mg once daily) or ivosidenib (500mg daily) in combination with standard induction (daunorubicin 60 mg/m2/day or idarubicin 12 mg/m2/day for 3 days with cytarabine 200 mg/m2/day for 7 days) and consolidation therapy in 154 newly diagnosed patients with mIDH1 or mIDH2 AML, have been recently published. The frequency of IDH-related DS was low, as expected given the concurrent administration of cytotoxic chemotherapy. There were no significant differences QT interval prolongation among patients treated with ivosidenib or enasidenib, while enasidenib was most frequently associated to increased total bilirubin, consistent with this inhibitor’s known potential to inhibit UGT1A1, without significant clinical consequences.

At the end of induction, In patients receiving ivosidenib (n = 60) or enasidenib (n = CR was achieved respectively in 55% patients treated with ivosidenib and 47% of patients treated with enasidenib; CR/CR with incomplete neutrophil or platelet recovery (CR/CRi/CRp) rates were 72% and 63%, respectively. In patients with a best overall response of CR/CRi/CRp, 16/41 (39%) receiving ivosidenib had IDH1 mutation clearance and 15/64 (23%) receiving enasidenib had IDH2 mutation clearance by digital polymerase chain reaction (50).

Since the synergism in inhibiting differentiation and apoptosis of blast shown in preclinical evaluation of azacitidine combined to ivosidenib, the combination was investigated in an open-label, multicenter, phase Ib trial comprising dose-finding and expansion stages to evaluate the safety and efficacy of combining oral ivosidenib 500 mg once daily continuously with subcutaneous azacitidine 75 mg/m2 on days 1–7 in 28-day cycles in patients with newly diagnosed mIDH1 AML ineligible for intensive induction chemotherapy (ClinicalTrials.gov identifier: NCT02677922). Twenty-three patients received ivosidenib plus azacitidine (median age, 76 years; range, 61–88 years). Treatment-related grade ≥ 3 adverse events occurring in > 10% of patients were neutropenia (22%), anemia (13%), thrombocytopenia (13%), and electrocardiogram QT prolongation (13%). Adverse events of special interest included all-grade IDH differentiation syndrome (17%), all-grade electrocardiogram QT prolongation (26%), and grade ≥ 3 leukocytosis (9%). Median treatment duration was 15.1 months (range, 0.3–32.2 months). The ORR was 78.3%, including 60.9% CR. With a median follow-up of 16 months, median duration of response in responders had not been reached. Median OS has not been reached with a 12-month survival estimate of 82.0% (95% CI, 58.8% to 92.8%). mIDH1 clearance in bone marrow mononuclear cells by BEAMing (beads, emulsion, amplification, magnetics) digital polymerase chain reaction with a sensitivity of 10^-4 (60) was seen in 10/14 patients (71.4%) achieving CR (43).

These findings led to the development of this combination in a phase 3 double-blind placebo-controlled study of azacitidine with or without ivosidenib (AGILE, ClinicalTrials.gov NCT03173248)⁶¹ that is currently recruiting in a total of 166 study centers in North America, South America, Asia, and Europe which will provide additional data on efficacy and safety.

IDH1 and IDH2 mutations are noted to co-occur with FLT3-ITD mutations in 15%–27% and 8%–30% of AML, respectively (40, 61). Preliminary reports indicate that IDH1/2m and FLT3-ITD mutations respond less well to enasidenib or ivosidenib as monotherapy (42, 49).

However, there is no impact of IDH mutation status in newly diagnosed FLT3-ITD+ mutated AML patients. In the first large retrospective study, there was a possible trend of inferior response and survival in dual mutants treated with lower intensity FLT3 inhibitor based regimens, particularly in the older AML population (61). To better understand this unmet clinical need, a retrospective cohort of 91 FLT3-ITD and IDH1 or IDH2 “double-mutated” AML patients at s MD Anderson Cancer Center was evaluated (62). FLT3 and/or IDH inhibitors (FLT3Is and/or IDHIs) were given as a single agent or in combination with cytotoxic chemotherapy or low-intensity therapy (62).

While single-agent FLT3Is and IDHIs demonstrated limited activity, the combination of cytotoxic chemotherapy and FLT3Is resulted in 100% CR+Cri in frontline and 64% in R/R AML patients. Combining FLT3Is with cytotoxic chemotherapy or low-intensity therapy may be effective induction therapy in patients with moderate to high FLT3-ITD allele frequency and/or those with lower IDH allele frequency in the presence of a concomitant FLT3-ITD, because these are likely instances of FLT3-driven disease. However, larger, prospective studies are required to design the optimal personalized approach for FLT3-ITD and IDH1 or IDH2 “double-mutated” AML patients.

Ivosidenib and enasidenib are promising differentiating agents in the AML therapeutic landscape, that are overall well tolerated with reasonable efficacy as monotherapy agents.

Ongoing and future clinical trials are required to identify the best patient setting and optimize the sequential approach of effective agents to further improve clinical outcome using tailored treatment in AML.

All authors: writing—original draft preparation. CC, ND, CD, and HK: supervision. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.