Fusion-specific co-mutational architecture and treatment-response associations in pediatric core-binding factor acute myeloid leukemia: a multicenter exome-profiled cohort.

Introduction: Core-binding factor acute myeloid leukemia (CBF-AML) is defined by one of two founding fusions, but cooperating lesions may differ between fusion subtypes. We characterized the broader diagnostic co-mutational architecture of pediatric CBF-AML and examined its relationship with post-induction measurable residual disease (MRD) and outcome. Methods: This retrospective analysis included 170 children with diagnostic whole-exome sequencing (WES) data (2019-2023), a subset of a reported 289-patient cohort. Candidate leukemia-associated variants were curated conservatively because matched normal samples were unavailable. Variant frequencies were compared between RUNX1::RUNX1T1 and CBFB::MYH11 AML with false-discovery-rate control. Associations with MRD were tested using Fisher's exact test. Primary feature-specific Cox models for event-free survival (EFS) were adjusted only for fusion subtype and induction regimen; sensitivity analyses additionally addressed age and presenting WBC count, MRD, and transplantation. Results: In a limited selection audit, fusion-subtype and FLT3-ITD distributions were similar between WES-included and non-included patients, whereas clinical-source KIT positivity was more frequent in the WES subset. The analytic cohort included 138 RUNX1::RUNX1T1 and 32 CBFB::MYH11 cases; 137 received FLAG-IDA, and 33 received DAE induction. At least one curated candidate variant was present in 135 patients (79.4%), and 69 (40.6%) had alterations in two or more genes. Tyrosine-kinase signaling alterations occurred in 115 patients (67.6%). RAS-pathway alterations were enriched in CBFB::MYH11 AML (53.1% vs. 19.6%; P < 0.001, q=0.002), driven principally by NRAS (50.0% vs. 15.2%; P < 0.001, q < 0.001). Chromatin-modifier alterations were numerically enriched in RUNX1::RUNX1T1 AML (24.6% vs. 6.2%; P = 0.028) but did not remain significant after multiplicity correction (q=0.114). After false-discovery-rate correction, no tested genomic feature was associated with MFC-MRD or molecular MRD positivity after the second induction. With a median follow-up of 33.2 months, 3-year EFS and overall survival were 84.2% and 89.8%, respectively. No baseline genomic feature showed a significant association with EFS in either the primary or expanded baseline-covariate models. Conclusions: Among WES-profiled patients, pediatric CBF-AML showed fusion-specific co-mutational architecture, especially RAS-pathway enrichment in CBFB::MYH11 AML. In this WES-defined cohort, baseline genomic features were not established as response or outcome predictors. These findings are descriptive and hypothesis-generating; they do not support treatment-specific causal inference.