Acute myeloid leukemia (AML) remains a significant therapeutic challenge, with 5-year survival rates around 30%. Treatment failure can occur early (primary refractory disease or early relapse) or years after treatment. Early failure is attributed to resistance in leukemia-regenerating subclones, while relapse often stems from minimal residual disease (MRD). Retrospective studies suggest therapy-resistant cells pre-exist at diagnosis, but prospective identification remains difficult. Relapse is often linked to leukemia stem cells (LSCs), but the classical LSC model, where chemotherapy-resistant LSCs accumulate, is challenged by evidence that cytarabine efficiently depletes LSCs in vivo. A regenerative response following chemotherapy and a reversible state of cellular senescence contributing to disease persistence are also proposed. The understanding of LSC plasticity and population dynamics during chemotherapy is incomplete, necessitating longitudinal single-cell studies combined with robust LSC identification methods and clonal tracking. This study uses a previously developed microRNA (miRNA) score, which includes miR-126, a miRNA strongly linked to LSC and HSC function, quiescence, and chemotherapy resistance, to investigate the longitudinal response of LSCs to chemotherapy.

The literature extensively discusses the role of leukemia stem cells (LSCs) in AML relapse and the challenges in treating this disease. Several studies explore the concept of minimal residual disease (MRD) and its contribution to relapse, while others focus on the heterogeneity of AML cells and their plasticity. The classical LSC model, suggesting the accumulation of chemotherapy-resistant LSCs during treatment cycles, has been questioned due to findings that cytarabine effectively depletes LSCs in vivo. Studies highlight the importance of a regenerative response after chemotherapy and the potential role of cellular senescence in disease persistence. However, direct evidence of LSC plasticity in response to chemotherapy has been lacking until this study.

This study employed a multi-faceted approach to investigate the longitudinal response of AML cells to chemotherapy at the single-cell level. The researchers used droplet-based single-cell RNA sequencing (scRNAseq) on leukemia-enriched bone marrow (BM) cells from patients with NPM1-mutated AML and del(7) AML at multiple time points (diagnosis, day 14, and day 30 post-chemotherapy, and relapse). To distinguish malignant from non-malignant cells, they determined the NPM1 mutational status or the expression of genes located on chromosome 7 in single cells. Xenograft models were established to validate the scRNAseq-based classification algorithm and to investigate LSC dynamics upon chemotherapy. A previously described microRNA (miR-126) reporter, enabling prospective flow cytometry-based identification of LSC-enriched AML subpopulations, was utilized in the xenograft models. This reporter vector contains miR-126 target sites within the GFP 3'UTR and an untagged control mCherry, thus allowing the quantification of miR-126 activity by analyzing the GFP/mCherry ratio. Limiting-dilution serial transplantation was used to assess leukemia-initiating cell (LIC) frequency. In vivo chemotherapy experiments were performed in mice using cytarabine and daunorubicin. Gene expression profiling was performed on sorted miR-126(high) and miR-126(low) blasts from xenografts. Gene Set Enrichment Analysis (GSEA) was used to identify pathways and processes enriched in different AML subpopulations before and after chemotherapy. Finally, the miR-126(high) gene signature was applied to publicly available AML datasets to assess its association with patient outcome. In vitro experiments were done to investigate the role of oxidative phosphorylation in LSCs.

The scRNAseq analysis revealed 15 distinct clusters in NPM1-mutated AML and 12 clusters in del(7) AML, representing different cell states and differentiation stages. Chemotherapy induced a generalized inflammatory and senescence-associated response. Progenitor AML cells showed heterogeneity: some proliferated and differentiated, expressing oxidative phosphorylation (OxPhos) signatures, while others were OxPhos(low), miR-126(high), and exhibited stemness and quiescence features. The miR-126(high) LSC subpopulation was enriched at diagnosis in chemotherapy-refractory AML and at relapse. This miR-126(high) signature strongly predicted patient survival in large AML cohorts. Xenograft studies using the miR-126 reporter showed that chemotherapy reduced the total blast population, but a small subpopulation with very high miR-126 activity persisted. This miR-126(high) population was enriched in the G0 cell cycle phase and displayed increased LIC frequency. Gene expression profiling revealed enrichment for HSC/LSC and lymphoid gene signatures in miR-126(high) blasts. Chemotherapy induced inflammatory and senescence-associated responses in both miR-126(high) and miR-126(low) blasts. ScRNAseq of xenografts confirmed the persistence of miR-126(high) LSCs after chemotherapy, with enhanced stemness features. Chemotherapy led to the expansion of differentiated myeloid and erythroid blasts, while the proportion of HSC-like blasts remained unchanged. Further high-resolution scRNAseq revealed distinct LSC subclusters within the HSC-like compartment, with some exhibiting increased dormancy and others showing increased proliferation after chemotherapy. In vitro experiments with OxPhos inhibition showed increased miR-126 activity and enhanced repopulation activity, highlighting the importance of OxPhos in miR-126(low) progenitors. Analysis of longitudinal patient samples revealed that miR-126(high) LSCs were enriched during chemotherapy and at relapse, often exhibiting a quiescent state at diagnosis that transitioned to a more proliferative state at relapse. Relapse samples were characterized by increased oxidative phosphorylation in all clusters. In patients who did not respond to reinduction chemotherapy, the persistent disease was enriched for miR-126(high) cells which showed quiescence and enrichment for hypoxia, stress responses and KRAS signaling.

This study provides strong evidence for the persistence of LSCs during early chemotherapy, emphasizing the limitations of cytotoxic chemotherapy alone in curing AML. The findings support a refined LSC model where the presence of a high proportion of quiescent miR-126(high) LSCs at diagnosis predicts treatment failure and relapse. The contrasting responses of committed progenitors (proliferation and OxPhos upregulation) and LSCs (quiescence and inflammatory response) upon chemotherapy are reminiscent of similar findings in solid tumors. The study suggests that chemotherapy induces senescence-like responses in LSCs, possibly contributing to enhanced stemness, warranting further investigation into causal relationships between inflammation and stemness. The observed lymphoid transcriptional bias in miR-126(high) LSCs extends the rationale for exploring lymphoid-marker directed immunotherapies. The data suggest that therapeutic targeting of oxidative phosphorylation, at least at the level of ETC complex I inhibition, may not be effective in NPM1-mutated AML patients with induction failure. Combining targeted agents active against dormant LSCs with chemotherapy may improve outcomes. This personalized approach needs careful patient selection.

This study demonstrates the persistence of leukemia stem cells (LSCs) with a miR-126(high) signature during early chemotherapy in acute myeloid leukemia (AML), highlighting their crucial role in treatment failure and relapse. The identification of this LSC subpopulation, which exhibits a unique transcriptional profile, provides a potential marker for patient risk stratification and suggests new therapeutic targets for improving AML treatment outcomes. Future studies should focus on elucidating the precise molecular mechanisms driving LSC persistence and exploring the development of targeted therapies.

The study has a relatively small sample size of patients, limiting the generalizability of the findings. Longitudinal single-cell tracking of individual LSCs was not performed, preventing a direct demonstration that relapse unequivocally originates from a quiescent LSC that persists through induction chemotherapy. Further investigations are needed to fully elucidate the complex interplay between different AML subpopulations and the precise molecular mechanisms underlying therapy resistance.