Altered Oxidative Phosphorylation Confers Vulnerability on IDH1-Mutant Leukemia Cells: Is This Therapeutically Tractable?

Summary:

Isocitrate dehydrogenase (IDH)–mutant acute myeloid leukemia (AML) is treatable with inhibitors of mutant IDH and also responds well to combination therapies including venetoclax, but most patients with IDH-mutant AML either never achieve complete remission or relapse because mutant hematopoietic stem cells persist despite treatment. An interesting new study in Blood Cancer Discovery characterizes a specific vulnerability in the mitochondrial oxidative phosphorylation system in preleukemic hematopoietic stem cells from patients with IDH1 mutations that is not present in those with IDH2 mutations; will this susceptibility prove amenable to therapy?

See related article by Landberg et al., p. 114 (10).

Recurrent somatic mutations at hotspots in the IDH1 and IDH2 genes were first found in solid tumors—initially in colorectal cancer, and then a high proportion of malignant gliomas—before their initial description in 2009 in malignant blast cells from patients with acute myeloid leukemia (AML; ref. 1). Mutations at IDH1^R132, IDH2^R140, or IDH2^R172 are relatively common in AML, present in 10% to 20% of newly diagnosed patients, with IDH2 mutated somewhat more often than IDH1. This class of mutation is also detectable in hematopoietic stem cells (HSC) and their more differentiated progeny from patients with the preleukemic myeloid neoplasms myelodysplastic syndromes (MDS) and myeloproliferative neoplasms (MPN), albeit less frequently than in AML. In AML, IDH mutations occur more often in patients with normal cytogenetics—25% to 30% of normal chromosome AML cases have an IDH mutation, compared to <5% with a complex monosomal karyotype—and their influence on prognosis is modest, allele-specific, and context-dependent (2).

Occasionally, IDH mutations are also found in blood and marrow cells from people with premalignant clonal hematopoiesis, although almost always in patients who also have cytopenias, defined as clonal cytopenias of undetermined significance (CCUS). In contrast, even though IDH mutations are thought to be disease-initiating events and are sometimes the only mutations detectable in purified leukemic stem cell populations, IDH-mutant clonal hematopoiesis of indeterminate potential (CHIP) represents <1% of CHIP cases (3). This seeming paradox of an early disease-initiating mutation being rare in a precursor condition may be a consequence of an unusual degree of suppression healthy hematopoietic clones by IDH-mutant cells.

The dehydrogenase enzymes encoded by the IDH1 and IDH2 genes normally catalyze oxidative decarboxylation of isocitrate to 2-oxoglutarate (also known as α-ketoglutarate) in the Krebs/tricarboxylic acid cycle, that mnemonic nemesis of undergraduate biology students everywhere. But the hotspot mutations in IDH1/IDH2 arginine residues result in neomorphic enzymatic activity that instead generates a potent oncometabolite, (D)-2-hydroxyglutarate (2-HG), which in turn inhibits the activity of α-ketoglutarate–dependent dioxygenases. Critical for neoplasia pathology is the epigenetic modulator TET2, a dioxygenase/hydroxylase that catalyzes conversion of 5-methylcytosine to 5-hydroxymethlcytosine (5hmC) as a first step in demethylation of cytosines in regulatory elements across the genome. TET2 is mutated in AML as well as in other myeloid neoplasms, and in contrast to IDH mutations, TET2 is one of the two most common mutations in CHIP and CCUS. TET2 and IDH1/IDH2 mutations are almost entirely mutually exclusive, indicating redundant pathologic function in myeloid oncogenesis (4). 2-HG inhibition of other enzymes including histone lysine demethylases may also play a role in disease pathology.

While no modulators of TET2 function are available yet for clinical use, the development of targeted inhibitors of mutant IDH1/IDH2 has been a major therapeutic advance. I can still remember the joy experienced by when one of my former patients with AML, a middle-aged man who achieved a complete response (CR) during the first-in-human clinical trial of the first mutant IDH1 inhibitor, ivosidenib, which at that time was known as AG-120. He had been failed by five previous treatment approaches including allogeneic hematopoietic cell transplant, but with ivosidenib, he achieved a CR with no evidence of measurable residual disease (MRD) by both flow cytometry and a clinical next-generation sequencing assay with a sensitivity of <1% variant allele frequency (VAF). He lived more than a year with an excellent quality of life before succumbing to an unrelated illness while still in remission.

On the basis of such activity, ivosidenib was approved by the FDA for relapsed/refractory (R/R) AML in 2018 and for newly diagnosed AML the following year, and in 2023 an MDS indication was added to the label. A targeted mutant IDH2 inhibitor, enasidenib (formerly AG-221), was FDA approved for R/R AML in 2017 and has also since added other indications. Additional IDH inhibitors have been developed, such as IDH1 inhibitor FT-2102 (olutasidenib), which was approved for AML in 2022 and offers better brain penetrance so may be more useful for gliomas.

Unfortunately, most patients with AML do not obtain the degree of benefit that my patient did from currently available IDH inhibitors, despite robust target inhibition. MRD is often detectable in patients with AML who achieve CR during IDH inhibitor therapy, which greatly increases recurrence risk, and even among those who achieve MRD negativity, relapse is still common. With enasidenib monotherapy, patients with IDH2-mutant R/R AML can expect a CR rate of 20% to 30% and a median response duration of 8 to 9 months; with ivosidenib monotherapy in IDH1 mutant R/R AML, CR rate and duration are similar. Outcomes are better for newly diagnosed IDH-mutant AML, but IDH inhibitors combined with older chemotherapeutic agents or DNA hypomethylating agents in initial therapy still fail many patients.

When relapse occurs after an MRD-negative CR, this implies that while the bulk of the malignant blasts were sensitive to the chosen therapy, a few leukemic HSCs remained to serve as the nidus of relapse. Therapeutic resistance to IDH inhibitors can be caused by many factors, including concomitant receptor tyrosine kinase and MAPK pathway mutations, IDH second site mutations, isoform switching, immune evasion, and pharmacologic limitations. New approaches to IDH-mutant AML and to other IDH-mutant neoplasms are needed.

Patients with IDH-mutant AML do respond to other treatment approaches besides the mutation-specific inhibitors and older cytotoxics. For example, IDH mutations induce BCL2 dependence in myeloblasts via inhibition of cytochrome C oxidase in the mitochondrial electron transport chain, and this can be exploited clinically (5). Venetoclax, an orally administered BCL2 inhibitor, received accelerated FDA approval in 2018 and full approval in 2020 for use in combination with injectable azacitidine, decitabine, or low-dose cytarabine to treat AML in older patients (≥75 years of age) or those with comorbidities preventing intensive induction chemotherapy. When combined with DNA hypomethylating agents, venetoclax induces a high rate of CR in IDH-mutant AML: a composite CR rate (inclusive of those with persistent cytopenias) of nearly 80%, better than with IDH–wild-type (composite CR ∼65%; ref. 6).

Altered fatty acid metabolism and perturbed mitochondrial function resulting from IDH mutations may also be potential vulnerabilities in IDH-mutant AML. With respect to the latter, previous work showed heightened sensitivity of IDH1-mutant AML cells to mitochondrial complex I oxidative phosphorylation (oxphos) inhibitors, including the clinical-grade inhibitor IACS-010759 (7). Unfortunately, two phase I clinical trials of IACS-0107059 in R/R AML (NCT02882321, n = 17) and advanced solid tumors (NCT03291938, n = 23) were both recently terminated due to adverse events including high blood lactate and neurotoxicity, without observation of clinical responses (8). Rotenone, another electron transport chain/complex I inhibitor useful as a tool compound and sometimes used topically to treat scabies and head lice in animals, is a known neurotoxin that has been linked to Parkinsonism in farm workers who used it as a pesticide or piscicide without wearing adequate protective gear.

Germline mutations in mitochondrial or nuclear genes encoding components of the cytochrome c/oxphos pathway are associated with peripheral neuropathy, ataxia, encephalopathy, myopathy, and optic atrophy, sometimes in syndromic form accompanied by abnormalities in other tissues or by lactic acidosis (9). This clinical observation coupled with the sobering IACS-010759 trial experience highlight the difficulty in targeting fundamental metabolic processes in cancer, and underscores how challenging it is likely to be to achieve a therapeutic index with agents targeting oxphos, especially with respect to the nervous system which seems to be especially sensitive to alterations in mitochondrial function.

Ivosidenib and enasidenib are tolerated because they bind specifically to mutant IDH and do not inhibit wild-type IDH or α-ketoglutarate production, only generation of 2-HG. Targeting upregulated or downregulated but qualitatively normal/wild-type enzymes in these fundamental energy-generation pathways is likely to have a much lower therapeutic index. That said, IM156 is a novel biguanide mitochondrial protein complex 1 inhibitor that was recently tested in a trial of 22 patients with advanced solid tumors (NCT03272256), and although no clinical responses better than stable disease were seen, it was reasonably well tolerated without neuropathy – although lactic acid increase was seen in several patients. A study of this agent in combination with chemotherapy for patients with pancreatic cancer is ongoing (NCT05497778) and hopefully will meet with success.

Now, in a new report in Blood Cancer Discovery, Landberg and colleagues in the Majeti lab at Stanford University and an international group of collaborators further elaborate a specific oxphos vulnerability conferred by IDH1 mutations in preleukemic HSCs (pHSC; ref. 10). The investigators studied both primary pHSCs from patients with AML (which are rare and can be difficult to isolate) and genetically engineered human HSCs that express the IDH1^R132H mutation. One finding from the team's transcriptomic analysis was that pHSCs downregulate MHC class II–related genes, which likely contributes to immune evasion and persistence of these cells, even though their short-term proliferation and colony-forming capacity is reduced. In addition, 5hmC and RNA sequencing as well as direct metabolic assays confirmed that IDH1-mutant pHSCs are metabolically vulnerable to targeting with inhibitors of oxidative phosphorylation (oxphos): IACS-0107059, rotenone, and IM156 mentioned above. These differences from wild-type cells were not seen in IDH2-mutant samples and are specific to IDH1.

One puzzle is why IDH1-mutant pHSCs cause neoplasia when they exhibit reduced proliferation and have a short-term competitive disadvantage, similar to TP53 mutant HSCs but in contrast to other mutation subtypes that confer a more aggressive growth trajectory. But a strong short-term competitive advantage of mutant HSCs is not strictly necessary for accumulation of malignant cells, especially because most of the malignant cell proliferation occurs in progeny. Immune evasion is important, and an analogy might also be made with chronic lymphocytic leukemia, where proliferation rates are low but apoptosis is reduced and longer-term survival enhanced, leading to an eventual accumulation of mature clonal lymphocytes.

The prospect of early clinical intervention in states of clonal hematopoiesis driven by a single initiating mutation—before additional mutations are acquired, and clonal evolution results in a more complex, less tractable condition such as AML—is highly attractive. Elimination of residual pHSCs in patients with AML in clinical remission is a related and important goal, so better characterizing their vulnerabilities, as Landberg and colleagues have done, is helpful. Experience has shown that exploiting these vulnerabilities in the clinic can be difficult, and risk-benefit balance must be carefully considered.

While CCUS has a natural history similar to lower-risk MDS, most people with CHIP will never develop a malignancy; statistically, cardiovascular events are a greater risk. Identification of those with CHIP at highest malignancy risk is a priority, and some progress has been made on this front, including development of the Clonal Hematopoiesis Risk Score (http://www.chrsapp.com/). An agent used in the CHIP setting would need to have a greater therapeutic index than mitochondrial complex I inhibitors (or in fact most drugs) have exhibited thus far. Clinically relevant endpoints in CHIP also require larger, longer studies to demonstrate than trials in an overt malignancy like AML, although therapy might be able to be administered for only a short time if the treatment quickly eradicates mutant pHSCs, reducing chronic adverse effects.

IDH inhibitors are occasionally associated with differentiation syndrome and other adverse events, but now there are enough safety and efficacy data in overt neoplasia that there are an ongoing trials of ivosidenib (NCT05030441) and enasidenib (NCT05102370) in IDH1- and IDH2-mutant CCUS, respectively. These trials have an innovative “remote enrollment” design to overcome geographic barriers to accrual. Regulatory agencies have understandably taken a conservative approach requiring an extensive safety database before allowing trials in CCUS or CHIP. But early detection and precision intervention exploiting genotype-driven metabolic and other vulnerabilities, potentially including oxphos, are the way of the future in cancer control and prevention.

Authors’ Disclosures

D.P. Steensma reports other support from Ajax Therapeutics, Novartis, Bluebird, Gamida Cell, and Arrowhead, and other support from Gilead outside the submitted work.