Molecular Detection of Minimal Residual Disease Provides the Most Powerful Independent Prognostic Factor Irrespective of Clonal Architecture in Nucleophosmin (NPM1) Mutant Acute Myeloid Leukemia

NPM1 mutation (NPM1c) defines the commonest molecular subtype of AML, which is molecularly heterogeneous, with outcome influenced by the pattern of cooperating mutations. This has generated interest in molecular profiling to guide treatment approach, particularly concerning allogeneic transplantation (SCT) in first complete remission (CR1). Patients with NPM1c AML with FLT3-ITD and/or DNMT3A mutations have been associated with poorer outcome and are widely considered candidates for SCT. Conversely, those with the NPM1c/FLT3-ITDneg genotype are no longer routinely transplanted in CR1, due to their generally favorable outcome. Given the molecular heterogeneity, use of real time quantitative PCR (RT-qPCR) to detect residual leukemia (MRD) could improve outcome prediction. However, its value has been questioned by recent studies providing evidence that relapse can arise from preleukemic clones.

To address these issues we analysed sequential samples from 346 patients (median age 50y, 6-81y) with NPM1c AML treated in the UK NCRI AML17 trial (median follow-up 35mo). An established RT-qPCR assay with mutation-specific primers was used, allowing MRD detection in all patients, covering 27 NPM1 mutations. Assays were confirmed to be mutant specific, with a median sensitivity of 1 in 10⁵ (1 in 10^3.7-7.1). Overall 2,569 follow-up samples (902 BM, 1667 PB) were analysed (median 6 samples/pt). To determine if MRD assessment provided independent prognostic information, targeted sequencing (Haloplex, Agilent Technologies) of 51 genes was undertaken in 223 cases with available diagnostic DNA. The panel included established recurrent mutation targets in NPM1c AML and 14 genes found to be mutated in whole exome sequencing of 27 cases with differing kinetics of relapse or sustained molecular remission (CRm).

For patients in documented morphological CR, early MRD assessment distinguished those at markedly differing risk of relapse and overall survival (OS). For patients with NPM1 mutant transcripts still detectable in peripheral blood (PB) following chemotherapy course 2 (30 of 194, 15%), cumulative incidence of relapse was 77% at 3 years, compared to 28% in those testing PCR negative (p<0.0001), associated with poorer OS (25% vs 77%, p<0.0001). While PCR positive patients had a higher rate of FLT3-ITD (66% vs 34%, p=0.002) and were more likely to have poor risk features (47% vs 17% p=0.002) considering age, cytogenetics, WBC and 2° AML, multivariable analysis found MRD status was the most significant prognostic factor (HR 4.51 (2.43-8.35) p<0.0001), identifying patients with poorer outcome within the NPM1c/FLT3-ITDneg group. Conversely, achievement of PCR negativity in PB by post-course 2 distinguished a major cohort (93 of 117, 79%) with relatively favorable outcome (OS 76%) despite an adverse genotype (FLT3-ITD &/or DNMT3A mutant). After adjustment for MRD status neither FLT3-ITD (p=0.8), DNMT3A (p=0.7) nor risk group (n=0.8) provided additional discrimination.

Serial monitoring of paired BM and PB samples showed that analysis of marrow increases MRD detection rate, affording a median 1-log increment in sensitivity, associated with a longer time from molecular conversion to relapse (BM median 144d vs 88d for PB, p=0.3). Detection of MRD beyond consolidation predicted progression in 53 of 64 patients, preceded by a median 0.7-log rise in transcripts (range 0.2-3.6) per month; 8 patients remain in CR following pre-emptive therapy including SCT or epigenetic therapy and only 3 have not yet relapsed at 2-10 months follow-up. Targeted sequencing analysis at molecular or clinical relapse, showed mutant allele frequencies of genes including DNMT3A and TET2 exceeding the NPM1 mutation in 21 of 51 (41%) cases, consistent with presence of preleukemic clones. Moreover, amongst patients achieving CRm with long-term follow-up BM samples (NPM1c neg, sensitivity >10^-4), DNMT3A-R882 remained readily detectable by digital PCR at 27-40 months (allele frequency 7-49%) in all 8 patients treated with chemotherapy alone; whereas the DNMT3A mutant clone was eliminated in the 3 patients after SCT. Irrespective of the clonal architecture, the NPM1mutation was found to be a stable marker of AML status, detected at relapse in 69 of 70 cases analysed.

Therefore the NPM1c RT-qPCR assays track the leukemic clone with high fidelity and merit adoption into routine practice to assess response and inform patient management.