Germline mutations in mitochondrial complex I reveal genetic and targetable vulnerability in IDH1-mutant acute myeloid leukaemia

Abstract

The interaction of germline variation and somatic cancer driver mutations is under-investigated. Here we describe the genomic mitochondrial landscape in adult acute myeloid leukaemia (AML) and show that rare variants affecting the nuclear- and mitochondrially-encoded complex I genes show near-mutual exclusivity with somatic driver mutations affecting isocitrate dehydrogenase 1 (IDH1), but not IDH2 suggesting a unique epistatic relationship. Whereas AML cells with rare complex I variants or mutations in IDH1 or IDH2 all display attenuated mitochondrial respiration, heightened sensitivity to complex I inhibitors including the clinical-grade inhibitor, IACS-010759, is observed only for IDH1-mutant AML. Furthermore, IDH1 mutant blasts that are resistant to the IDH1-mutant inhibitor, ivosidenib, retain sensitivity to complex I inhibition. We propose that the IDH1 mutation limits the flexibility for citrate utilization in the presence of impaired complex I activity to a degree that is not apparent in IDH2 mutant cells, exposing a mutation-specific metabolic vulnerability. This reduced metabolic plasticity explains the epistatic relationship between the germline complex I variants and oncogenic IDH1 mutation underscoring the utility of genomic data in revealing metabolic vulnerabilities with implications for therapy.

Fig. 1. Mitochondrial respiratory chain variants and expression in AML.

a Schematic showing all genes encoding components of respiratory chain complexes I–V. Complex I comprises three subunits α, β and γ. Nuclear-encoded genes with rare variants identified are coloured red. Black indicates nuclear-encoded genes with no variant identified. Grey circles denote mitochondrial-encoded genes; these were not sequenced in the Australian cohort. b Co-mutation plot for the Australian patient cohort (n = 213). Mutation groups are shown in rows with each individual patient represented by a column. The presence of a mutation is indicated by coloured bar. Near mutual exclusivity was observed between rare nuclear-encoded complex I mutations and somatic IDH1 R132 mutations (FDR = 3.5 × 10⁻⁰⁸, weighted exclusivity test). c Co-mutation plot for the Stanford patient cohort (n = 124) showing segregation of rare mitochondrial-encoded complex I variants and IDH1 mutations (FDR 3.26 × 10⁻⁵, weighted exclusivity test). d Variant allele frequency (VAF) analysis of complex I variants in patients with co-occurring mutations in either IDH1 (n = 10) or IDH2 (n = 34) in the Australian cohort, Stanford cohort and the Beat AML study (P = 0.0088, two-sided Wilcoxon rank sum test). Box and whisker plots indicate median, 25th and 75th percentile and ±1.5 interquartile range. Each dot represents a different sample. Source data are provided as a Source Data file. e Positive enrichment of nuclear-encoded complex I genes, including NDUFS8, with genes up-regulated in IDH1- versus- IDH2-mutant AML samples (Beat AML dataset). Statistical testing performed by gene set enrichment analysis, FDR < 0.0001. NES normalized enrichment score.

Fig. 2. Mitochondrial respiration in primary AML samples.

a Relative mitochondrial copy number (MCN) determined by measuring the DNA abundance of mitochondria-encoded cytochrome B (CYB) relative to nuclear-encoded glucuronidase beta (GUSB) for human bone marrow mononuclear cells from healthy donors (hBM, n = 8), complex I (C-I) mutated (n = 9), IDH1- (n = 8), IDH2-mutated (n = 12), and IDH1/2 wild type (WT, n = 7) AML. Samples used to determine oxygen consumption rates in (b–d) are shown in filled symbols. Significance determined by one-way ANOVA (Dunnett’s multiple correction): P = 0.016 (hBM-IDH1), 0.013 (hBM-IDH2). b Basal oxygen consumption rate (OCR) and (c) maximal OCR normalized to basal OCR as measured after addition of uncoupling agent for hBM (n = 5), C-I mutated (n = 4), IDH1- (n = 4), IDH2-mutated (n = 5) and WT (n = 8) AML samples. For (b) and (c), significance determined by one-way ANOVA (Dunnett’s multiple correction): significance values for (b): P = 0.0015 (hBM vs. C-I) and 0.026 (hBM vs. WT); significance values for (c): P < 0.0001 (hBM vs. C-I), P = 0.0005 (hBM vs. IDH1) and 0.0035 (hBM vs. IDH2). Box and whisker plots indicate median, 25th and 75th percentile and range of data. Each dot indicates a different patient sample. d Normalized OCR for the individual samples summarized in (b) and (c) showing reduced OXPHOS capacity in AML with rare complex I variants (n = 4) and IDH1- (n = 4) and IDH2- mutated (n = 5) AML samples, compared to wild type (WT) samples (n = 8) and hBM controls (n = 5). WT samples are FLT3-ITD negative and WT for IDH1, IDH2, DNMT3A, NPM1, and complex-I. Each line represents an independent patient sample tested in triplicate and each value is the mean of a given measurement timepoint. Error bars represent SD. Data represented as fold change relative to basal OCR. Details of patient samples are provided in Supplementary Data 5. Source data are provided as a Source Data file.

Fig. 3. IDH1 mutation confers sensitivity to complex I inhibition.

a (R)-2-hydroxyglutarate (2HG) abundance in supernatants from THP-1 cells transduced with doxycycline-inducible IDH1-wild type (WT), IDH1 R132H, IDH2-WT, or IDH2 R140Q. Bars represent mean of technical duplicates from a single experiment. Data points for IDH1-WT were below background and thus, are not represented in the graph. b and c Cell growth of GFP+ doxycycline-inducible IDH1-WT, IDH1 R132H, IDH2-WT or IDH2 R140Q expressing THP1 cells at 72 h after treatment with (b) 1 µM rotenone or (c) 5 µM IACS-010759, relative to DMSO vehicle control. Bars represent mean ± SEM from three independent experiments. Each dot represents the mean of an independent experiment. Statistical significance determined by two-tailed unpaired t-test. Significance values for (b) P = 0.0146 (IDH1-WT vs. IDH1 R132H), and (c) P = 0.0028 (IDH1-WT vs. IDH1 R132H) and 0.0246 (IDH2-WT vs. IDH2 R140Q). d Ratio of live (propidium iodide-negative) cells for healthy CD34+ samples (n = 2), IDH1/2 WT (n = 7), IDH2- (n = 4) or IDH1-mutant (n = 12) AML samples treated with 5 µM IACS-010759 (IACS), relative to DMSO vehicle control over 72 h. Samples tested for response to ivosidenib in Supplementary Fig. 5 are shown as S, sensitive or R, resistant to ivosidenib. Values represent mean for each sample in a single experiment and each dot is a data point. Significance determined for patient groups using one-way ANOVA (Tukey’s multiple correction), P = 0.0033 (IDH1/2 WT vs. IDH1 mutant), 0.0396 (IDH1 mutant vs. IDH2 mutant). e Variable response of IDH1-mutant samples treated for 72 h with 5 µM IACS-010759 (IACS) following 24 h pre-treatment with 10 µM ivosidenib (IVO) compared to IACS alone. Data presented relative to DMSO control. Data show mean of a single experiment and each dot represents a data point. Details of patient samples used in d-e are provided in Supplementary Data 5. Source data are provided as a Source Data file.

Fig. 4. Metabolic plasticity of IDH1-mutant AML.

a Total ATP production rate at basal and following induction with 5 µM IACS-010759 (IACS) for isogenic THP1 cells expressing IDH1 wildtype (WT), IDH1 R132H, IDH2-WT or IDH2 R140Q, measured by ATP rate assay. Contribution to ATP production rate from glycolysis (grey) and OXPHOS (white) are shown as mean ± SD of six technical replicates. Statistical significance determined by one-way ANOVA (Tukey’s multiple correction), ****P < 0.0001 and *P = 0.029. b Mitochondrial ATP as a percentage of total ATP in untreated isogenic THP1 cells shown as mean ± SD of six technical replicates. Significance determined by two-tailed unpaired t-test, P = 0.0005 (IDH1-WT vs. IDH1 R132H) and P < 0.0001 (IDH2-WT vs. IDH2 R140Q). c Change in glycolytic ATP following addition of 5 µM IACS-10759 (IACS) relative to basal glycolytic ATP production rate, measured by ATP rate assay for isogenic THP1 cells. Data shown as mean ± SD of six technical replicates. Significance determined by two-tailed unpaired t-test, P = 0.0001 (IDH1-WT vs. IDH1 R132H) and P < 0.0001 (IDH2-WT vs. IDH2 R140Q). Raw oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) data related to (a–c) are presented in Supplementary Fig. 6a, b. d Change in glycolytic ATP following addition of 5 µM IACS-10759 relative to the basal glycolytic ATP production rate, for healthy CD34+ samples (n = 3), IDH1/2 WT (n = 3), IDH2- (n = 1) or IDH1-mutant (n = 2) AML. Data presented as mean ± SD of four or more technical replicates per sample, no significance was observed between groups. Individual sample results in Supplementary Fig. 7. e NADPH levels in CD34+ samples (n = 4), IDH1/2 WT (n = 4), IDH2- (n = 5) or IDH1-mutant (n = 5) AML. Data presented as mean of the group ± SEM. Each dot is the mean of an independent sample. Significance determined by one-way ANOVA (Dunnett’s multiple correction), significance P = 0.0091 (CD34+ vs. IDH1 mutant). Details of patient samples used in (d) and (e) are provided in Supplementary Data 5. Source data for (a–e) are provided in Source Data file. Schematic representation of metabolic rewiring of (f) IDH1 wild type AML and (g) IDH1 mutant AML following complex I inhibition. See text for explanation. Mitochondria shaded in blue. AcCoA acetyl CoA, FA fatty acid, IACS IACS-010759, IDH isocitrate dehydrogenase, ME1 malic enzyme 1, OAA oxaloacetic acid, OXPHOS oxidative phosphorylation, PPP pentose phosphate pathway, αKG alpha ketoglutarate, 2HG 2-hydroxyglutarate.