Single-cell tracing of myeloid leukemia evolution in down syndrome reveals a latent progenitor bridging disease onset, progression and relapse

Abstract Children with Down syndrome (DS), or trisomy 21 (T21), have a markedly elevated risk of acute myeloid leukemia (ML-DS) in early childhood, stemming from aberrant hematopoietic stem/progenitor cell development in utero. ML-DS is often preceded by Transient Abnormal Myelopoiesis (TAM), a neonatal pre-leukemic condition driven by truncating GATA1 mutations acquired before birth. Although TAM typically regresses after birth, it carries significant mortality and ~20–30% of cases progress to ML-DS by age four, indicating that a subset of TAM cells persists beyond the perceived remission. While primary ML-DS is highly chemo-responsive, the treatment related morbidity is high, and prognosis upon relapse is dismal. This underscores the need to understand the developmental origins of TAM and ML-DS, and to identify (pre)-leukemic progenitors that endure TAM remission, subsequently seed ML-DS with up to 4-year latency and have the potential to drive relapse. To trace the evolutionary trajectory of ML-DS from fetal development through relapse, we performed 5′ short- and long-read single-cell RNA sequencing (scRNA-seq) and immunophenotyping on 41 longitudinal samples from 6 TAM patients (intrauterine, pre- and term birth, treated or incompletely resolved TAM), and 15 ML-DS patients, including 4 relapsed patients, collected at diagnosis, post-therapy and relapse. Integrating single cell transcriptomes with GATA1 and ML-DS stage mutation calls enabled us to track mutant blasts and their clonal behavior across disease stages. We defined distinct molecular and immunophenotypic signatures distinguishing TAM from ML-DS blasts, revealing both shared (e.g. IRX1) and disease stage-specific features (CLCA1 vs PRAME respectively) across patients. While both intrauterine and neonatal TAM blasts showed signatures of high proliferation and hypoxia-adaption, ML-DS blasts exhibited a robust type I interferon response signature largely absent in TAM blasts, highlighting disease stage specific properties and their environmental context. Upon unbiased clustering of ~700,000 single-cell transcriptomes from longitudinal samples, we identified a subpopulation of GATA1s-mutated cells present as a minor subset at TAM diagnosis at birth and in utero, that directly linked to the predominant blast cluster, but lacked high expression of the classical TAM/ML-DS blast markers CD33 and CD7. We confirmed that this unique blast subset can persist beyond presumed TAM remission and expand before and during ML-DS months to years later, indicating unique mechanisms that allows its survival. Moreover, this “bridge” subpopulation of blasts persisted following low dose cytarabine therapy for TAM, while the bulk blasts were eradicated. Notably, this (pre)-malignant population was present in all disease stages in every patient analyzed. Transcriptomically, the bridge subcluster of blast cells co-expressed many stem cell-associated transcription factors (MLLT3, ERG, MSI2), exhibited a more quiescent signature and displayed an adhesion molecule profile suggesting unique niche-interactions distinct from bulk TAM and ML-DS blasts. By projecting these data to our new 11- to 20-week T21 and euploid fetal liver and bone marrow scRNA-seq atlas, we mapped the (pre)-leukemic states onto the continuum of hematopoietic development that is influenced by unique intrinsic and microenvironmental alterations in T21. While most TAM and ML-DS-blasts resembled fetal megakaryocyte-erythroid progenitors (MEPs) with similar immunophenotype, the bridge subcluster retained an earlier, stem-like signature observed in immature fetal HSCs/MPPs. The subcluster persisted upon relapse often representing the ancestral leukemic subclone that links ML-DS to relapse, harboring the founding GATA1s mutation, and acquired cooperative abnormalities such as trisomy 8 or deletions in chromosome 7. By resolving disease trajectory at single-cell resolution, we propose a model of DS myeloid leukemogenesis connecting altered trisomy 21 fetal hematopoiesis to the emergence of a leukemia prone, disease initiating progenitor subset distinct from bulk blasts that spans all disease stages and can endure therapy and contribute to relapse. These novel insights have direct implications to early risk stratification, optimized disease monitoring and design of targeted interventions, and may extend to other infant leukemias initiated in utero.