Opening Pandora's Box: Ultra-Deep Duplex Sequencing in Long Survivors with Acute Myeloid Leukemia

Introduction: Detection of measurable residual disease (MRD) in acute myeloid leukemia (AML) is a poor prognostic indicator, but not all patients (pts) with residual disease relapse. Similarly, not all pts with undetectable residual disease are cured, irrespective of the technology used for assessment. Since 5 years in remission is frequently used to define cure in AML, we investigated the mutational landscape of AML pts in ongoing remission for 5 years using ultra-deep duplex sequencing (DS). Methods:Twenty-four bone marrow (BM) and 10 peripheral blood (PB) samples from 34 pts in remission for at least 5 years were selected for DS: 16 pts were in remission 5-11+ years after diagnosis without allogeneic stem cell transplantation (SCT); 18 patients were in remission 5-8+ years after SCT in first remission. 20/34 pts had samples sequenced at 2 separate time points. The study was approved by the institutional review board and pts provided informed consent. Technical validation samples included a dilution series containing hotspot mutations in key genes involved in AML (FLT3-ITD (x2), FLT3-TKD, NPM1, IDH1, IDH2). Cells from two primary AML diagnosis samples (with in-house NGS data) were spiked into BM cells from a mutation-free healthy donor. The accuracy of the expected percent of mutant AML cells spiked into the BM in the series was validated by flow cytometry to achieve an expected VAF of the mutations in the sample series, ranging from 0.001 to 5%. Genomic DNA from cryopreserved MNC and control samples was extracted using column-based methods and processed for sequencing using the AML-29v1 targeted panel from TwinStrandTM DS technology with enzymatic fragmentation. Libraries were generated using 1-1.5ug of DNA per the manufacturer's protocol and sequenced on a NovaSeq6000 instrument using paired 150bp reads. Alignment and variant calling were performed as per the AML-29 TwinStrandTM pipeline. Calls with a minimum of 2 alternate DS reads were included for analysis and samples were designated as MRD+ if they had 2 supporting reads for mutations present at diagnosis (0.01% sensitivity). Diagnostic mutations were identified from clinical laboratory reports with 2-5% sensitivity. Results: The average duplex-depth in the remission samples was 40,534x (19,404x-54,567x). Among the 16 pts in prolonged remission without SCT, 11/16 were ELN2022 favorable risk, 3 intermediate and 2 adverse; baseline mutations were 0 in 1 pt, 1 (isolated NPM1 or IDH2) in 5 pts, and 2-5 in 10 pts. Samples were from time-points 27 days to 10.8 years after diagnosis. None of the 7 non-transplanted pts with a baseline NPM1 mutation had detectable NPM1>3 mos after remission. However, 5/7 patients with baseline NPM1 had mutations in IDH2, DNMT3A, NRAS, KRAS and/or TP53 (VAF 0.006-9.5%) that were not reported at baseline up to ten years after initial diagnosis. One of these 5 pts developed detectable DNMTA3 and TP53 mutations at two time points while on oral azacitidine maintenance. Among non-transplanted pts, 3 had IDH2 and 1 IDH1 at baseline. One pt with IDH2 became undetectable 6.8 years after diagnosis with ongoing enasidenib. The other 3 IDH1/2 pts had persistently detectable mutations 46 days-2.2 years after diagnosis. Among the 18 transplanted pts, 8 were ELN2022 favorable, 6 intermediate and 4 adverse; pts had 2-8 mutations detected at baseline. Four pts who had MRD+ samples (3 BM, 1 PB) by DS pre-transplant also had BM samples available for DS post-transplant (30-625 days), none of which had detectable mutations. One pt had persistently detectable baseline U2AF1 (0.0042%) a year after transplant. None of the pts with baseline NPM1 mutations (8/18) had detectable NPM1 post-transplant, and from 5/6 pts with baseline IDH1/2 and an available post-transplant sample, 1 pt had persistently detectable IDH2 (0.09%) 1.9 years after transplant. Conclusions: Ultra-deepDS of AML pts in prolonged remission showed clearance of the diagnostic mutations detected in clinical sequencing. However, low-level persistence and/or appearance of mutations in known pathogenic hotspots were frequently observed after years of ongoing remission, including for pts receiving maintenance. Further investigation of the clinical relevance of low-level mutations in long-term remission is needed to inform decision-making and prevent over-treatment.