Revolutionizing AML MRD: Gene-level kinetics to uncover chemoresistant clones and predict relapse risk via NGS.

252

Background: Minimal residual disease (MRD) assessment in AML remains binary in practice, despite heterogeneity across genomic subtypes. We evaluated gene-specific MRD kinetics via serial NGS to define mutation-level clearance patterns, relapse risk, and resistance mechanisms. Methods: Sixty-nine newly diagnosed AML patients (median age 47, follow-up 25.7 months) were prospectively followed. Seventy percent harbored actionable mutations. Induction was predominantly azacitidine-venetoclax based; 71% underwent allogeneic transplant. Serial NGS tracked 189 DNA mutations, 12 RNA fusions, and 1 CNV. Among 60 patients with known regimens, 147 evaluable DNA mutations (57 patients) were analyzed for post-induction clearance and longitudinal kinetics. Results: Combined remission rate was 87%; relapse rate 27.5% (19/69). NGS detected MRD in 65% versus 22% by flow cytometry (concordance 47.8%, κ = 0.11). 67.6% of morphologic CR patients harbored persistent molecular disease. At mutation level, 65.3% achieved post-induction clearance; only 42.1% of patients achieved complete MRD negativity, reflecting resistant co-clones. Clearance varied sharply by gene. Signaling mutations showed high chemosensitivity: FLT3-ITD 87.5%, FLT3-TKD 75%, NRAS 80%, KIT 85.7%, RUNX1 100%, NPM1 78.9%. Epigenetic/metabolic mutations were resistant: IDH1/2 25.0% (OR 0.14, p = 0.0006), TP53 0% (p = 0.0014), and DTA genes 50%. Combined as “epigenetic modifiers” (DNMT3A/TET2/ASXL1/IDH1/2), clearance was 38.2% versus 73.5% for non-epigenetic mutations (OR 0.22, p = 0.0004), the strongest predictor of MRD persistence. Gene-specific relapse patterns refined risk. FLT3, IDH1/2, TP53, ASXL1, and BCOR relapsed largely in MRD-positive patients; no MRD-negative FLT3 or IDH1 patients relapsed. Conversely, WT1 (67% overall relapse) and KIT showed relapse even in MRD-negative states, highlighting genotype-dependent MRD limitations. NPM1 MRD negativity conferred excellent outcomes (0% relapse), while TP53 exhibited “smoldering persistence” with oscillatory VAF kinetics despite therapy. Functional hierarchy emerged: signaling/proliferative mutations cleared in 75% versus 35% for epigenetic/tumor suppressor lesions (p = 0.031), suggesting distinct resistance biology. Intensive regimens improved early clearance, whereas HMA+VEN+targeted combinations showed strongest late clearance and lowest relapse (16.7%). Conclusions: AML MRD is mutation-context dependent. Epigenetic modifier mutations define a chemoresistant MRD phenotype independent of ELN risk, while signaling mutations display paradoxical chemosensitivity. NGS-MRD outperforms flow cytometry , revealing actionable kinetic patterns informing transplant timing, targeted maintenance, and genotype-adapted surveillance. These data support a shift from binary MRD assessment to gene-specific kinetic modeling in AML.