Myelodysplasia-Related Gene Mutations for NGS-Based MRD Assessment in Acute Myeloid Leukemia

Introduction: Measurable residual disease (MRD) plays an increasing role for treatment management of AML patients. While next-generation-sequencing (NGS) based MRD assessment can be well standardized, there is uncertainty about the type of gene mutations that associate with prognosis and are useful for MRD assessment. The 2022 ELN recommendations assign patients with at least one mutation in the genes ASXL1, BCOR, EZH2, RUNX1, SF3B1, SRSF2, STAG2, U2AF1 or ZRSR2 to the subgroup of myelodysplasia-related gene mutations (MRGM), which is associated with poor prognosis. We aimed to define the prognostic effect of MRD after induction chemotherapy in AML patients with MRGM using MRG mutations for MRD assessment by NGS.

Methods: We retrospectively collected bone marrow and peripheral blood samples from AML patients in complete remission (CR) or CR with incomplete hematologic recovery (CRi) after one to two cycles of 7+3-based standard induction chemotherapy (IC) with at least one MRGM detected at diagnosis by myeloid panel sequencing covering 48 AML associated genes. Patients treated with one course of IC were included if they proceeded to allogeneic hematopoietic cell transplantation (alloHCT) already after one cycle. Molecular MRD status was assessed using targeted NGS and bioinformatics error-correction with 0.01% sensitivity as described before using only MRGM as MRD markers.

Results: MRD was assessed on 104 adult newly-diagnosed AML patients (median age 58.6) with a total of 166 MRGM, resulting in a median of one MRGM per patient (range 1-4). 43 (41.3%) patients had secondary or therapy-related AML and 20 (19.2%) had myelodysplasia-related cytogenetic abnormalities. 68 (65.4%) patients were MRD positive for at least one MRGM after one (18 patients (17.3%)) or two (86 patients (82.7%)) cycles of intensive chemotherapy, while 36 (34.6%) became MRD negative. Measured variant allele frequencies (VAF) of positive samples ranged from 0.02-43.1% (median VAF 1.34%). MRD positive patients were older and more likely to be male compared to MRD negative patients, while cytogenetic risk, WBC count, Hgb, platelet count and ECOG status at diagnosis were similar to MRD negative patients. 53 (77.9%) MRD positive and 23 (63.9%) MRD negative patients underwent alloHCT in first CR/CRi. Median follow-up was 3.35 years. OS (P = 0.92), RFS (P = 0.99) and CIR (P = 0.98) were similar in MRD positive (at least one MRGM detectable) and negative (no MRGM detectable) patients (Figure 1). OS, RFS and CIR were also similar between MRD positive and negative patients, when MRD negativity was defined by conversion of at least one MRGM to negativity (55 (52.9%) MRD positive, 49 (47.1%) MRD negative), and when ASXL1 was excluded from the analysis. As variants with VAF >5% in CR/CRi samples may indicate clonal hematopoiesis, we excluded all variants with VAF >5% and repeated the analysis with the 88 (84.6%) remaining patients. Again, OS, RFS and CIR were similar for MRD positive and negative patients. We next reasoned that reduction of VAF correlates with the sensitivity of leukemic cells and possibly with depth of remission and long-term outcome. The fold-reduction of VAFs from diagnosis to the time of MRD analysis was calculated for all variants. For patients with several MRGMs, the variant with the largest fold-reduction was considered including undetectable variants. Patients with more or less than a 2-log, 3-log or 4-log reduction (log 10) had similar OS, RFS and CIR, respectively. Evaluating each MRGM individually, only RUNX1-MRD associated with RFS and by trend CIR (comparing MRD positive to negative patients for OS, HR 3.41, 95% CI 0.43-26.82, P = 0.24; RFS, HR 4.37, 95% CI 1.07-17.79, P = 0.04; CIR, HR 3.30, 95% CI 0.79-13.82, P = 0.1).

Conclusion: 65.4% of AML-MRGM patients in CR/CRi remained MRD positive when MRGMs were used for MRD assessment by NGS. MRD positivity was associated with age above 60 years and male sex. Using all MRG mutations for MRD assessment at a threshold of 0.01% or by various degrees of log-reduction was not predictive of outcome after intensive induction chemotherapy. This limits the number of potential MRD markers for NGS analysis before alloHCT, although individual markers like RUNX1 may be of prognostic value.