The mutational and transcriptional landscape of pediatric KMT2Ar AML
While overall prognosis for pediatric acute myeloid leukemia (pAML) has improved, relapse remains a critical challenge, affecting one-third of patients and carrying a significantly reduced chance of cure. This highlights an urgent need to delineate the mechanisms driving treatment resistance and disease recurrence. Our study focuses on KMT2A-rearranged (KMT2Ar) pAML, a high-risk subgroup with distinct clinical and genetic features. We usedsingle-cell transcriptomics to map the cellular and transcriptional heterogeneity of KMT2ArpAML across different disease stages from diagnosis over treatment to relapse. The goal is to identify specific relapse-driving cell populations and transcriptional programs that can be targeted for more effective, personalized therapies.
We analyzed bone marrow aspirates from 18 pediatric KMT2Ar AML patients (age 2–13 years), including seven who relapsed and eleven who achieved continuous complete remission (CCR). The cohort included various fusions: KMT2A::MLLT1 (n=3), KMT2A::MLLT3 (n=6), KMT2A::MLLT10 (n=7), and KMT2A::MYO1F (n=1). Samples were collected at key time points: diagnosis (Dx), post-induction (F1), post-consolidation (F2), and at relapse (Re). Single-cell RNA sequencing (scRNA-seq) (10x Genomics) was performed to quantify gene expression profiles. To accurately discriminate malignant from normal cells and track clonal evolution, we implemented a combined strategy involving targeted long-read scRNA-seq to detect patient-specific KMT2A fusion transcripts, which was integrated with the high-resolution transcriptomic data. This methodology enabled a direct link between the genotype (fusion transcript presence) and the malignant cell phenotype across disease evolution.
Integration of scRNA-seq data with mutation calls revealed a highly diverse landscape of malignant cell states, most of which displayed progenitor-like transcriptional signatures. We documented substantial phenotypic shifts in malignant populations during treatment and at relapse indicating clonal plasticity. Analysis of relapsed samples identified distinct transcriptional programs and cell states that were either expanded or newly emerged, suggesting their role in treatment escape.For instance, in one patient, the primary diagnosis population was predominantly myelocyte-like but transitioned to a novel prog-DC-like phenotype at relapse, while also exhibiting minor monocyte-like and T-cell-like malignant components. Interestingly, a small, yet consistent, subset of malignant pro-B-like cells was detected across three distinct patients. This finding suggests that lymphoid-affiliated programs may unexpectedly contribute to leukemogenesis and/or treatment resistance in a subset of KMT2Ar AML, which warrants further investigation.
This integrative single-cell study provides a comprehensive view of KMT2Ar pAML heterogeneity, revealing significant transcriptional plasticity and clonal evolution across the course of treatment and relapse. By linking genotype (the KMT2A fusion) to the malignant phenotype at single-cell resolution, we have successfully identified potential relapse-associated subpopulations and their underlying transcriptional programs. The unexpected discovery of KMT2Ar cells with lymphoid-affiliated programs suggests a previously underappreciated layer of complexity. Ultimately, these findings provide a crucial framework for designing precision therapies that specifically target drug-resistant clones, paving the way for more durable cures for children with high-risk KMT2Ar AML.