Single-cell atlas reveals hidden patterns of differentiation in acute myeloid leukemia microbiologystudy

Princess Margaret Cancer Center at University Health Network-led researchers constructed a detailed single-cell atlas of human blood development, improving understanding of leukemia formation and enabling a more precise classification of acute leukemia.

Therapeutic efforts in acute myeloid leukemia (AML) face persistent challenges due to diverse and unstable cell populations within and between tumors. Previous studies have investigated how specific genetic mutations disrupt the normal development of blood cells, yet a complete understanding of how these mutations shape disease behavior across different AML cases has remained elusive.

It remains unclear whether these developmental disruptions follow shared patterns or are unique to each mutation. Single-cell RNA sequencing (scRNA-seq) offers more precise insight into cancer cell identity than traditional diagnostic tools, yet limited patient sample sizes and technical variability have slowed progress.

A comprehensive, large-scale analysis has been needed to reveal how cell development in AML is altered to support more effective diagnosis and treatment.

In the study, “Single-cell Transcriptional Atlas of Human Hematopoiesis Reveals Genetic and Hierarchy-Based Determinants of Aberrant AML Differentiation,” published in Blood Cancer Discovery, researchers constructed a single-cell reference atlas of human hematopoietic differentiation to reveal how blood cell development becomes disrupted in AML.

A total of 263,159 single-cell transcriptomes were collected from bone marrow samples of 45 healthy donors to construct a reference map of normal hematopoiesis. Some 1.2 million cells from 318 acute leukemia samples were mapped against this reference to classify cellular states and genetic variations associated with AML.

Researchers identified 55 distinct hematopoietic cell states across the dataset. All leukemia cells were projected onto these reference states. Frequently co-occurring states were grouped into 13 broader differentiation states to enable comparative analysis across leukemia samples.

Mapping the 1.2 million cells onto the hematopoietic reference revealed 12 recurrent patterns of differentiation across 318 patient samples. These patterns reflected both early developmental blocks and lineage-biased differentiation toward myeloid, lymphoid, and erythroid fates.

Analysis showed that primitive stem-like differentiation patterns were more common in adult AML compared to pediatric AML, while monocytic differentiation patterns were more frequent among pediatric cases. More than 45 distinct mutations were associated with specific differentiation trajectories. Yet, samples sharing the same genetic driver could display markedly different differentiation landscapes, suggesting additional layers of disease heterogeneity.

Early lymphoid and early erythroid differentiation patterns were identified within subsets of AML traditionally classified as normal karyotype disease.

In a group of 180 adults with a type of leukemia that appears genetically normal under standard testing, patients whose cancer cells resembled early lymphoid progenitor cells faced a significantly higher risk of death. For each standard-deviation increase in the proportion of these lymphoid-like cells, the risk of dying rose by about 65% (HR ~1.65).

In contrast, patients whose leukemia showed more features of early red blood cell development had a 37% lower risk of death (HR ~0.63).

When researchers tested how these leukemia cells responded to drugs in the lab, they found that those with strong lymphoid traits were more sensitive to 23 different treatments, 18 of them receptor tyrosine-kinase inhibitors. One drug, (midostaurin), showed effectiveness unrelated to the common FLT3 mutation it is usually prescribed for, suggesting it might help a broader group of patients.

Precise classification of malignant cell states in AML may improve patient stratification, clarify diagnostic boundaries between overlapping disease categories, and inform personalized therapeutic approaches. These findings offer a framework for identifying high-risk leukemia subtypes and potential treatment vulnerabilities beyond current genomic classifications.

© 2025 Science X Network