Elucidating Mechanisms of Acquired Resistance to IDH Inhibition By Saturation Variant Screening of Base-Edited Leukemia Cells

Somatic mutations in catalytic arginine residues of isocitrate dehydrogenases IDH1 and IDH2 occur frequently in myeloid malignancies. Heterozygous mutation of cytosolic IDH1 or mitochondrial IDH2 contributes to leukemia development by accumulating the oncometabolite (R)-2-hydroxyglutarate (R-2-HG), causing alterations of histone and DNA methylation through impaired α-ketoglutarate-dependent dioxygenases. Small molecule inhibitors ivosidenib and enasidenib, which specifically target mutant IDH1 and IDH2 respectively, are now approved for the treatment of newly diagnosed and/or relapsed/refractory acute myeloid leukemia (AML) patients with IDH1/2 mutations. However, multiple cases of therapeutic resistance to IDH inhibitors were reported. Distinct mechanisms were described for clinical resistance of IDH inhibition, such as the acquisition of secondary mutations in cis or trans on the allosteric binding site of IDH inhibitors or isoform switching. Although a few amino acid substitutions including IDH1-S280F, IDH2-Q316E and IDH2-I319M were identified to be associated with acquired resistance, the extent to which the mutational repertoire at IDH1/2 proteins and the associated mechanisms contribute to acquired resistance to IDH inhibition remains unknown.

In this study, we systematically examined all potential amino acid substitution mutations on IDH1 and IDH2 by multiple orthogonal saturating mutagenesis screens in base-edited myeloid leukemia cells. We first engineered independent myeloid leukemia cell lines containing endogenous patient-derived IDH1/2 mutations by CRISPR/Cas9-mediated base-editing. We created clonal leukemia cell lines containing heterozygous IDH1-R132H (IDH1^R132H/WT) or IDH2-R140Q (IDH2^R140Q/WT) mutations, which are the most frequent IDH1/2 mutations in AML. The base-edited IDH1 or IDH2-mutant leukemia cells were characterized by the accumulation of the oncometabolite R-2-HG, alterations of histone and DNA methylation, and aberrant cytokine-independent cell growth. To identify all possible amino acid substitution mutations with resistance to IDH inhibition, we performed MITE-seq-based saturation variant screens, in which each amino acid was systematically substituted by all other possible amino acids, in base-edited leukemia cells. We not only validated known mutations such as IDH1-S280F, IDH2-Q316E and IDH2-I319M, but also identified a list of de novo amino acid substitutions associated with acquired resistance to IDH inhibition. Furthermore, the base-edited leukemia cells were treated continuously with ivosidenib and enasidenib up to 16 weeks to identify acquired resistance-associated IDH1/2 mutations by next-generation sequencing. Finally, we performed ultra-deep targeted sequencing of IDH1/2 genes in 11 paired diagnosis and relapsed AML samples in patients treated with ivosidenib or enasidenib as monotherapy, and identified significantly enriched IDH1/2 secondary mutations in the relapsed samples as compared to AML samples at diagnosis. Integrative analyses of multiple orthogonal mutagenesis screens uncovered a compendium of high-confidence and clinically relevant IDH1/2 mutations with resistance to IDH inhibition. Structural modelling of IDH1/2 mutation-associated protein domains provided new insights into the mechanisms for acquired resistance due to dysregulated protein 3D structure, co-factor binding, and/or steric hindrance with the binding of small molecular inhibitors. Follow-up functional studies established new IDH mutations as candidate drivers of acquired resistance to IDH inhibition through distinct mechanisms in AML.

Taken together, we generated isogenic, base-edited IDH1/2 mutant myeloid leukemia cells and identified candidate amino acid substitutions conferring resistance to IDH inhibition through orthogonal saturating variant screens. We integrated the results with targeted sequencing of IDH1/2 mutations in AML patients, structural modeling, and functional studies. Our findings not only uncover de novo pathogenic mutations associated with resistance to IDH inhibition, but also provide new insights into the molecular mechanisms for acquired resistance to a targeted therapy in AML.