A novel multi-target NGS application for molecular monitoring of measurable residual disease in Acute Myeloid Leukemia

Abstract Background . Measurable residual disease (MRD) in Acute Myeloid Leukemia (AML) is a prognostic indicator used in clinical practice for risk stratification and treatment planning. There is no uniform approach for detecting MRD in AML. Multiparametric flow cytometry-MRD remains challenging to standardize, while quantitative PCR (qPCR) approach is limited to 40-60% of all AML cases with a targetable alteration. The ELN guidelines proposed Next Generation Sequencing for MRD quantification (NGS-MRD), although the optimal markers to monitor and threshold levels have not been defined. NGS-MRD is potentially applicable to the large majority of AML patients and allows following each patient-specific mutation with a single assay. It also provides information on clonal evolution. The main limitation of NGS-MRD remains the high error rate intrinsic to conventional NGS that requires the use of an error suppression approach and the need of standardization. Methods. We performed a single-center NGS-MRD-based analysis on a cohort of 40 AML patients treated with allogeneic hematopoietic stem cell transplantation (allo-HSCT) in complete remission between 2010-2024 when an NGS study was available at diagnosis. We applied a novel error suppression NGS-MRD method designed by Sophia Genetics starting from 800 ng of DNA that warranted a high level of sensitivity (10-4). The gene panel included 23 commonly mutated genes in AML (CALR, CEPBA, DDX41, ETV6, EZH2, FLT3, IDH1, IDH2, JAK2, KIT, KRAS, MPL, NPM1, NRAS, PTPN11, RAD21, RUNX1, SF3B1, SRSF2, STAG2, TP53, U2AF1, WT1), whileDNMT3A, TET2, ASXL1were excluded from the panel because non-informative for MRD. The NGS method was refined by the use of UMI CUMIN adapters and bioinformatic tool to minimize PCR error propagation.Libraries obtained by the applied panel were sequenced on an Illumina NovaSeq instrument to achieve a very high depth. Results. The median age at transplant of the 40 AML patients was 58 years (range 19-70). According to ELN 2022, the risk stratification was favourable, intermediate, and adverse in 20%, 37% and 42.5% of the patients, respectively. A total of 138 somatic mutations were identified at diagnosis, and 57 mutations were still detectable at the time of conditioning. The median number of gene mutations at diagnosis was 3 (range 1-9). Only 4 patients of the cohort had only one mutation at disease onset. To assess NGS-MRD evaluation, we obtained a high number of good-quality sequences that allowed us to reach a uniform coverage greater than 500000X, a key premise for a high sensitivity level. Potential germline and CHIP mutations were excluded from MRD analysis as non-informative for MRD. 62.5% had at least one mutation detectable with a VAF≥0.01% (NGS-MRDPOS), and 15% of patients had more than one mutation in CR (NGS-MRDPOS≥2). To validate the NGS-MRD approach, 16 NPM1-mutated AML patients were evaluated in parallel by the standardized qPCR method, while 16 IDH1/2, 1 SRSF2, and 2 c-KIT mutated patients by digital droplet PCR (ddPCR). MRD results by NGS, qPCR, and ddPCR were highly concordant (K Cohen 0.68). In a univariate analysis of RFS and OS, the persistence of two or more mutations in remission using the threshold of 0.01% (MRDPOS≥2) was significantly associated with shorter OS (HR=5.2, p=0.009) and RFS (HR=8.4, p=0.0002). When we looked into single mutations, only FLT3-ITD and NPM1 persistence in CR (VAF ≥0.01%) significantly impact on RFS (HR=36.3, p=0.003; HR=5.3, p=0.005) and CIR (HR=36.3, p=0.0003; HR=7.5, p=0.02) but not on OS. When considering the predictive value of MRD according to the different function of the analysed genes (“Signalling”, “Epigenetic”, “Transcription” and “Splicing”), only the persistence of mutations in “Signalling” genes (FLT3, c-KIT, PTPN11, NRAS, KRAS, MYC) showed a significant impact on RFS (HR=3.0, p=0.03). The statistical significance of MRDPOS≥2 was confirmed in a multivariate analysis including age, ELN 2022 categories, and transplant-related factors (conditioning regimen, donor).Conclusions. Our data suggest that the persistence of at least two mutations in CR at pre-conditioning is associated with unfavourable outcomes following allo-HSCT. NGS-MRD can be a reliable tool to detect MRD in the large majority of AML patients and can identify patients who can benefit from a post-transplant treatment. An extended analysis of a larger cohort of AML patients receiving an allo-HSCT is ongoing