Novel approaches to identify cellular vulnerabilities in TET2-dependent MDS

Abstract

Abstract MDS and related disorders manifest a tremendous diversity of molecular patterns originating from the founder lesions and depending on the stage of clonal evolution. Nevertheless, diverse lesions or mutational combinations can result in pathway convergence, leading to shared pharmacologic vulnerabilities. We proposed that related molecular patterns, if present, might be exploited to study shared cellular vulnerabilities stemming from pathway convergence or molecular phenocopy. We previously used a large molecular patient registry and machine learning tools to separate the myeloid neoplasia continuum into distinct genomic clusters by accounting for binary features (presence and absense of mutations).1 Molecular clusters were identified based on a consensus clustering approach via autoencoders coupled with a Gaussian mixture modeling strategy. However, further cluster refining can be achieved by accounting for molecular complexity. To that end we focused on TET2 mutations (TET2MT) as examples of molecular complexity given a) high frequency in MDS and AML, b) variation in the types of mutations (missense, stop codon, frameshift), c) number of mutations, d) topography and rank by variant allele frequency. More importantly, the trajectories of TET2MT follow linear or branching evolution with many other accompanying lesions and thus generate a complex molecular architecture possibly modulating drug responsiveness. Inclusion of these parameters/features might provide additional resolution in patient stratification. To date, standard analytic methods have not been successful in resolving the tremendous heterogeneity of molecular lesions co-occurring with TET2MT. From a cohort of more than 3500 patients, we compiled 679 TET2MT patients with 40% having more than one TET2 hit. Nearly all patients had another mutation in another gene. The following types of mutations were represented: frameshifts (47%), nonsense (34%), and missense (19%). Focusing on this cohort, we identified 4 distinct TET2 clusters (TC) not benchmarked on clinical outcomes but applicable in an unsupervised fashion. TC were observed in appreciable proportions of mutant patients (TC3, 9%; TC6, 35%; TC10, 17%; TC12, 15%). Investigating the phenotypic associations, TC10 was predominantly composed of low risk MDS, whereas high risk and secondary AML cases were uniformly distributed in TC3, 6, and 12. There was no difference in overall survival among the clusters containing high risk patients and the median follow-up time was 18 months. Over the years our laboratory has designed drugs active in isogenic cellular models which were devised by rational approaches to preferentially target TET2MT.2 However, TET2 mutational cooperativity might shape clinical drug responsiveness. We thus established an experimental framework to study actual drug sensitivity in primary specimens using high throughput screening to deliver case-specific pharmacotypic output. In total, 20 human specimens with variable blast count and TET2MT in combination with other gene mutations were analyzed. Each drug was quantified at an acceptable level of statistical variation (less than 15% of outlier observations). Our in house panel of FDA approved drugs was tested against bone marrow cells in a 72 hour culture. The number of metabolically active cells was determined based on ATP quantification. Nine patients carried TET2MT in the catalytic domain spanning 1129-1936 base pairs (3 missense; 4 nonsense; 2 frameshift). Drug response/refractoriness profiles were similar between nonsense and frameshift TET2MT. For some drugs, patient cells with nonsense and frameshift TET2MTwere similar in terms of resistance and sensitivity. Other drugs were effective in cells with missense TET2MT while being completely ineffective in cells with other types of mutations. For example, cells from patients with nonsense and frameshift TET2MTwere more sensitive to the menin inhibitor, revumenib. Analyzing molecular associations, two cases with frameshift and dominant TET2MTwhich were sensitive to revumenib also carried NPM1MT. In conclusion,identification of patients with distinctive genomic sub-entities will provide indications for the investigation of TET2MT patterns (single mutation vs. combinations) predicting novel drug responsiveness/resistance that can inform mechanistic studies as to the pathophysiologic pathways involved and the modeling of TET2-dependent conditions.