The Genomic Landscape and Its Clinical Implication in NPM1-Mutated AML Patients: A Study within the AMLSG 09-09 Clinical Trial

Background: Mutations in the Nucleophosmin 1 gene (NPM1^mut) represent one of the most common genetic lesions in acute myeloid leukemia (AML). Based on its characteristic clinico-pathologic features, NPM1^mut AML has been recognized as a distinct entity among the category “AML with recurrent genetic abnormalities”. According to the ELN 2022 genetic risk-stratification, NPM1^mut AML, in the absence of FLT3-ITD mutation, is associated with a favorable prognosis. However, a significant proportion of these patients (pts) relapse after intensive therapy suggesting that other co-mutations may have an impact on outcome.

Aims: To comprehensively characterize the genomic landscape and leukemogenic trajectories in a large cohort of NPM1^mut AML pts and to investigate its prognostic and predictive impact on outcome.

Methods: Targeted DNA sequencing (mean read depth: 1817) on the entire coding region of 263 genes was performed in 568 NPM1^mut AML pts (median age: 58.7 years; 18-60 years, n=317; >60 years, n=251). All pts were enrolled in the randomized open-label Phase 3 AMLSG 09-09 trial [NCT00893399; Döhner H et al. Lancet Haematol 2023]. In this trial, pts were assigned to intensive chemotherapy plus all-trans retinoic acid with or without gemtuzumab ozogamicin; none of the pts received midostaurin.

Results: In total n=3,058 variants (variant allele frequency of ≥1%) were identified in 195/263 genes. The median number of co-mutations was 3 (range 0-11). The most common co-mutated genes were DNMT3A (49.5%), FLT3-TKD (42.8%) PTPN11 (24.8%), NRAS (22.7%) TET2 (21.7%), IDH2 (21.3%), IDH1 (18%), and FLT3-ITD (17.3%). DNMT3A^R882 hotspot mutations occurred more frequently in younger pts (36.6% vs 17.1%), while there was no difference for DNMT3A^nonR882 mutations between the two age groups (21.8% and 21.9%). An age-dependent difference was also identified for mutations in myelodysplasia-related genes (ASXL1, BCOR, EZH2, RUNX1, SF3B1, SRSF2, STAG2, U2AF1, and ZRSR2) as defined by the ICC, occurring with a higher incidence in older pts (30.3% vs 12.3% in younger pts).

Analyzing the mutational pattern of co-mutations, we found statistically significant tertiary gene-gene interactions: e.g., NPM1-NFE2-STAG2 (p<.001), NPM1-IDH2-SRSF2 (p<.001), NPM1-CEBPA-TET2 (p<.001), NPM1-DNMT3A^R882-NRAS (p=.002), and NPM1-ASXL1-SRSF2 (p=.004); mutual exclusivities were identified for NPM1-DNMT3A^R882-DNMT3A^nonR882 (p<.001), NPM1-IDH2-TET2 (p<.001), NPM1-DNMT3A^R882-SRSF2 (p<.001), NPM1-IDH1-TET2 (p<.001), and NPM1-FLT3-ITD-KRAS (p<.002). Correlating co-mutation data with outcome, we found that DNMT3A^R882 hotspot mutations confer inferior event-free (EFS) and overall survival (OS) only in younger pts (EFS, p<.001 vs p=.11, Figure 1a; OS, p=.003 vs p=.8), whereas DNMT3A^nonR882 mutations did not impact prognosis within the two age groups. We also found a negative prognostic impact of IDH1 mutations which was restricted to younger pts (EFS, p=.05), whereas IDH2 mutations were associated with superior EFS in older pts (p=.04) and superior OS in both groups (p=.05 and p=.03). Of note, co-mutations occurring in one or more of the myelodysplasia-related genes did not impact EFS or OS (Figure 1b). In multivariable analysis (all pts) including age, WBC, LDH, allogeneic transplantation in CR1, and mutations with an incidence of at least 3% as covariables, age (HR,1.03; p<.001), DNMT3A^R882 (HR, 1.86; p<.001), FLT3-ITD (HR, 1.54; p=.012), IDH1 (HR, 1.48; p=.009), MYC (HR, 1.83; p=.032), and WT1 (HR, 1.73; p=.012), were associated with an inferior EFS, while SMC3 mutation showed favorable EFS (HR, 0.44; p=.019). To further study the leukemogenic trajectories, we used an oncogenetic tree modeling algorithm, which yielded a tree with several main branches including DNMT3A^R882, DNMT3A^nonR882, FLT3-TKD, IDH1, IDH2, PTPN11, and TET2. These mutations might represent initiating events which predispose to additional events with further distinct branches.

Conclusions: Our study provides comprehensive data on the genomic landscape and its clinical impact in pts with NPM1^mut AML fit for intensive chemotherapy. The co-mutational pattern clearly differs between younger and older NPM1^mut AML pts. Using this large dataset allowed the identification of secondary and tertiary gene-gene interactions with significant impact on outcome. Further data analysis is ongoing and will be presented at the meeting.