Mapping AML Stem-Cell States

Evidence

Three primary studies examine acute myeloid leukemia (AML) stem-like cells from different angles: prospective identification in human samples, a self-renewal mechanism in mouse leukemia, and treatment persistence in paired pediatric bone-marrow samples. Their methods do not amount to direct replication. Together, however, they clarify why a molecular signature, a sortable surface phenotype, and survival after chemotherapy are related but non-equivalent properties.

The newest source is a July 13 bioRxiv preprint, not yet peer reviewed. Its authors combined bulk-expression cohorts with single-cell multi-omic assays to derive a prognostic signature called hrLSC2, defined by HOPX and SOCS2 co-expression. They then used that signature to identify a human AML subpopulation with the surface phenotype CD34+CD90-CLL1-CD69+CD53-. Cells isolated with this marker combination were significantly enriched for leukemia stem-cell activity in limiting-dilution xenotransplantation assays. That functional assay is important: the study did not infer stemness from expression alone. The ingested record does not provide cohort sizes, effect estimates, or transplantation frequencies, so this briefing cannot quantify the enrichment.

A separate July 16 peer-reviewed study tested a candidate regulator in murine AML. Rbm5 was highly expressed in leukemia stem cells. Removing Rbm5 impaired their self-renewal in vivo, reduced self-renewal gene expression, and promoted myeloid differentiation, while normal hematopoietic stem cells did not show the same defect in the reported model. Mechanistic experiments identified Myc as a major direct transcriptional target. The investigators also reported that RBM5 interacted with MYC and helped maintain MYC protein levels; forced Myc expression rescued the defects caused by Rbm5 depletion. These findings connect an RNA-binding protein to a Myc-centered self-renewal circuit in mice, not to a validated human AML target.

The commercially reusable full-text study examined paired pre- and post-chemotherapy marrow from 13 pediatric AML patients who had reached remission. Across 26 samples, investigators retained 227,842 high-quality single-cell transcriptomes and distinguished leukemic clusters from normal cells using transcriptomic profiles supported by mutations, chromosomal signatures, or flow-cytometry markers. Leukemia stem-cell and oxidative-phosphorylation signatures appeared in different stem- and progenitor-like populations. All seven diagnostic populations lacking either signature became undetectable after chemotherapy. By contrast, cells persisted in two of seven patients whose HSC-like populations carried a stem-cell signature and in three of six whose LMPP-like populations carried an oxidative-phosphorylation signature.

Within the seven patients whose HSC-like cells carried the stem-cell signature, two had resistant populations and five had sensitive ones. About 90% of HSC-like cells in each resistant case expressed CD69, versus a median of 39.52% across the sensitive cases, although the sensitive range extended to 88.97%. In cell lines, forced CD69 expression reduced mTOR-pathway signaling, lowered proliferation markers, increased CXCR4, and increased adhesion to mesenchymal cells and migration toward CXCL12. In a separate deconvolution analysis of public TARGET and TCGA cohorts, a high inferred CD69+ HSC-like fraction was also associated with adverse features. In TARGET, measurable residual disease after the first chemotherapy cycle was present in 61.1% of the high-fraction group and 27.5% of the low-fraction group. This was an association derived partly from inferred cell fractions, not a prospective diagnostic validation.

Analysis — Identity, State, and Function

The cross-study pattern is that AML stemness behaves less like one fixed label than a validation chain. This is analysis, not a conclusion tested jointly by the three groups. The preprint begins with a two-gene signal, converts it into a five-marker sorting phenotype, and then asks whether the isolated cells initiate leukemia in xenografts. The pediatric study starts from paired patient samples and asks which transcriptional states remain visible after chemotherapy; its stem-cell and oxidative-phosphorylation programs occupied different populations, showing that persistence need not collapse into one phenotype. The Rbm5 study adds a third layer by perturbing a candidate self-renewal circuit and rescuing the resulting defect with Myc. A useful emerging model is therefore identity plus state plus mechanism: markers locate a candidate population, longitudinal sampling tests whether it persists, and perturbation asks what sustains it. CD69 appears in both human identification efforts, but that overlap is not yet replication because the marker combinations, ages, assays, endpoints, and disease contexts differ.

Limitations

Two recent sources were ingested only as abstracts, restricting review of sample selection, statistical models, effect sizes, transplantation design, and sensitivity analyses. The newest source is a preprint and may change during peer review. Its xenotransplantation endpoint measures leukemia-initiating activity in an animal system, not clinical relapse. The Rbm5 work used murine leukemia models; apparent selectivity over normal hematopoietic stem cells in those experiments does not establish safety or therapeutic value in people.

The full-text chemotherapy study included only 13 pediatric patients, and its closest CD69 comparison involved two resistant and five sensitive HSC-like populations. Some later associations depended on computational deconvolution of bulk public datasets rather than direct single-cell measurement. CD69 expression also overlapped across the resistant and sensitive ranges, limiting any simple binary interpretation. The studies did not analyze the same samples, marker panel, treatment regimen, or perturbation. Prospective cohorts using a locked phenotype, direct post-treatment cell measurements, functional assays, and independent replication would be needed to determine whether these observations describe one reproducible AML stem-cell state.