Predicting Response to IDH1/IDH2 Inhibitors Beyond IDH Mutations Using a Computational Model and Its Validation: A Beat AML Project Study

Background: Isocitrate dehydrogenase (IDH) is an essential enzyme in the TCA cycle. Recurrent mutations in IDH1 or IDH2 are prevalent in several cancers, including glioma and acute myeloid leukemia (AML). IDH mutations reduce cellular levels of α-ketoglutarate (α-KG), resulting in accumulation of 2-hydroxyglutarate (2-HG), an α-KG antagonist. This leads to the inhibition of both α-KG-dependent histone lysine demethylases and the TET family of DNA hydroxylases. Even though current IDH inhibitors have high specificity towards the IDH1 and IDH2 mutant forms, the overall clinical response rate to these drugs remains ~19%, suggesting the presence of IDH mutations alone does not reflect positive clinical outcome. However, defining biomarkers and developing methods to predict response to IDH inhibitors will expedite the drug development process and provide patient selection criteria for future clinical trials.

Aim: To create a digital drug model for IDH1/2 inhibitors and validate the predicted response in AML patient samples using ex vivo drug sensitivity data.

Methods: The Beat AML project (supported by the Leukemia & Lymphoma Society) collects clinical data and bone marrow specimens from AML patients (n=720). Bone marrow samples were analyzed by conventional cytogenetics, whole-exome sequencing, RNA-seq, and an ex vivo drug sensitivity assay. For 98 of randomly chosen Beat AML patients, every available genomic abnormality was entered into a computational biological modeling (CBM) program (Cellworks Group) that uses PubMed and other online resources to generate patient-specific protein network maps of activated and inactivated protein pathways. Digital drug simulations with IDH1 (AGI-5198) and IDH2 (AGI-6780) inhibitors were conducted by quantitatively measuring drug effects on a composite AML disease inhibition score (i.e., cell proliferation, viability, and apoptosis). Computational predictions of drug response were compared to AGI-5198 and AGI-6780 IC₅₀ values derived from the Beat AML ex vivo drug sensitivity assay.

Results: Ninety-eight profiles with adequate genomic and drug sensitivity data were selected and analyzed by CBM to predict response to IDH inhibitors. CBM correctly predicted the sensitivity of 95/98 (97%) profiles to AGI-5198, and 94/98 (96%) to AGI-6780.

Ten patients harboring IDH1 variants (R132H (n=3), R132C (n=1), R132S (n=3), and R132G (n=1)) were identified, and all available genomic information was entered into the CBM program. CBM predicted that all 10 patients would be non-responders to AGI-5198, and AGI-5198 IC₅₀ values derived from the ex vivo sensitivity assay were consistent with CBM predictions in all 10 cases (100%).

Nine patients harboring IDH2 variants (R140Q (n=8), R172K (n=1)) were identified, and all available genomic information was entered into the CBM program. Three patient samples showed sensitivity to AGI-6780 (IC₅₀<0.09), while 6 samples showed no response (IC₅₀>1). CBM correctly matched 8/9 (88.8%) response predictions to drug sensitivity data.

This analysis indicated that not all patients harboring IDH1 / IDH2 activating mutations respond to IDH inhibitors, likely due to co-occurrence of additional mutations that negate drug response. For instance, one patient with an IDH1 R132C mutation was predicted to not respond to AGI-5198 because of the co-occurrence of a TET2 loss-of-function (LOF) mutation. In this case, the 2-hydroxyglutarate (2-HG) mediated downstream activity would have been impaired, resulting in hypermethylated DNA.

Conclusion: We developed a digital drug model of IDH1 and IDH2 inhibitors, and validated the CBM predictions by comparing IC₅₀ values measured in an ex vivo drug sensitivity assay. This computational platform can accurately identify patients who may or may not respond to IDH inhibitors by analyzing genomic and cytogenetic data. This technique may identify patient populations and inclusion criteria for upcoming clinical trials of IDH inhibitors, and identify sensitivity and resistance mechanisms associated with response.