The hypomethylating agent decitabine and the BCL-2 inhibitor venetoclax combination (DEC-VEN) is now standard of care in AML, yet a subset of patients remains refractory, and factors that reduce efficacy remain to be fully elucidated. This study aimed to systematically identify molecular determinants of resistance via single-cell (sc)RNA-seq and scATAC-seq analysis of bone marrow (BM) samples from patients participating in a prospective clinical trial of DEC-VEN (NCT03404193). For scRNA-seq, 17 BM samples (total 39,607 cells) from 12 AML patients (median age 67; 8 responders; 4 non-responders; 5 pre- and post-treatment pairs) and 3 healthy BM controls were integrated and annotated into 23 transcriptional cell clusters. In the Uniform Manifold Approximation and Projection (UMAP) plot, C1 overlapped with healthy hematopoietic stem cells (HSCs), C2 appeared only in AML patients, and C3 overlapped with progenitor cells. C1 was characteristic of refractory AML and persisted after DEC-VEN treatment. (Figure 1) C1 cells showed significantly higher expression of FLT3, BCL2, MCL1 and transcription factors MYB, ETV6, RUNX1, LMO2, JUND compared to HSCs from healthy donors. DEC-VEN treatment markedly increased ribosomal protein ( RP ) genes and mitochondrial electron transfer COX family genes and increased expression of the leukemia stem cell (LSC) marker CD74 in C1 of non-responders. In C2 cells, high expression of LSC marker genes FLT3, PIM1, MYC and transforming master regulator genes MYB, ETV6, RUNX2, LMO2, STAT3, NOTCH1, BCL11A was observed. DEC-VEN significantly upregulated RP genes and CD74 in C2 in non-responders, as in C1. In responders, hemoglobin family genes were significantly increased in C2. C3 cells were characterized by highly expressed iron metabolism genes, which were further increased by DEC-VEN in responders. GSEA analysis demonstrated that DEC-VEN upregulated ATP metabolic processes and ribosomal biosynthesis and downregulated stem cell differentiation genes in refractory C1 and C2. In responders' AML clusters, DEC-VEN upregulated cellular oxidant detoxification and iron ion homeostasis and downregulated myeloid leukocyte activation. Next, we extended the AML transcriptome analysis by profiling accessible chromatin by scATAC-seq using BM from pre- and post- DEC -VEN treatment of 2 AML patients (1 refractory, 1 relapsed after remission with incomplete count recovery [CRi]), and 2 healthy controls (total 41,062 cells). Leukemia-like cluster (LLC) cells showed significantly higher chromatin accessibility of CD74, RUNX1, RUNX2, NOTCH1, LMO2, ETV6, JUND, STAT3, BCL2, BCL11A and enriched binding motifs of RUNX2, ETV6, MYB, JUN, JUND, STAT3 compared to other clusters, which was consistent with gene expression findings of C1 and C2 by scRNA-seq. DEC-VEN treatment induced further increased motif enrichment of STAT3 in LLC in CRi AML, and STAT3 RUNX2, IKZF1 in refractory AML. To interpret T cell kinetics in response to DEC-VEN in AML, T cells from scRNA-seq profiling were subclassified based on canonical gene expression. The results showed that the CD3+ and CD3+CD8+ cell signatures were significantly lower in non-responders compared to responders post-treatment (p<0.005). Notably, the ratio of exhausted CD8+ cell clusters was significantly higher in non-responders than in responders in DEC-VEN post-treatment (p<0.05), with downregulation of CD8+ effector marker GZMK, and regulator of CD8+ cytotoxic T cell development TOX. In summary, we identified a subpopulation of AML cells that overlap with HSCs in UMAP plots and are associated with DEC-VEN resistance. LSC-related genes were enriched in resistant AML clusters, while hemoglobin-related genes were increased in sensitive clusters. DEC-VEN induced mitochondrial metabolism genes in resistant cells and further enhanced erythrocytic differentiation in sensitive cells. In addition, scATAC-seq revealed LSC-specific chromatin accessibility patterns, and depletion of CD8+ T cells was observed in non-responders. These findings shed light on the molecular mechanisms of DEC-VEN resistance.
Read full abstract