Abstract
Background: Venetoclax (VEN), a selective BCL-2 inhibitor, has improved outcomes in AML when combined with hypomethylating agents (HMAs), particularly in older or unfit patients. However, response to VEN-based therapy in secondary AML (s-AML), especially in patients previously treated for antecedent myelodysplastic neoplasm (MDS), remains variable. The mechanisms underlying treatment resistance in this high-risk group are not well characterized. Given the cellular heterogeneity of s-AML and the limitations of bulk profiling, single-cell RNA sequencing (scRNA-seq) offers a powerful tool to dissect resistance mechanisms in specific cellular compartments such as hematopoietic stem/progenitor cells (HSPCs) and immune populations.
Methods: We performed scRNA-seq on 279,211 bone marrow (BM)-derived cells from 25 samples collected from 14 s-AML patients treated with HMA/VEN at Seoul St. Mary's Hospital between February 2020 and November 2023. Samples were collected longitudinally before and after therapy. Differential transcriptomic signatures were analyzed across HSPCs, T/NK cells, and myeloid compartments. Findings were validated using flow cytometry, RT-qPCR, and a VEN-resistant SKM-1 AML cell line model.
Results: scRNA-seq revealed that pre-treatment responders harbored an expanded HSPC subcluster enriched for MYC target genes, which was selectively depleted post-treatment, whereas non-responders exhibited expansion of this subcluster during therapy. Metabolic analysis showed upregulation of the serine biosynthesis pathway—particularly PHGDH and PSAT1—in non-responders. Functional validation in VEN-resistant AML cell lines confirmed that PHGDH inhibition (NCT-503) restored drug sensitivity.
Immunologically, naïve CCR7⁺CD45RA⁺ CD4⁺ T cells were markedly depleted in non-responders, while immunosuppressive NEAT1⁺ CD4⁺ T cells were enriched. Similarly, pro-inflammatory monocyte/macrophage subclusters (e.g., Mono-S100A12, Mac-FCGR3A) were diminished in non-responders, who instead exhibited expansion of MALAT1⁺ monocytes and MHC-II–high immunosuppressive myeloid populations. Immune interaction analysis revealed disrupted LAIR-1 and MHC-I signaling in non-responders, suggesting impaired immune surveillance within the BM niche.
Conclusion: Our findings identify HSPC-intrinsic mechanisms—specifically MYC activation and PHGDH-driven serine biosynthesis—as candidate drivers of VEN resistance in s-AML evolved from MDS. In parallel, immune dysregulation within the bone marrow microenvironment, including depletion of naïve CD4⁺ T cells and impaired LAIR-1 signaling, was associated with non-responsiveness. While these findings are based on patients with s-AML arising from MDS without direct comparison to de novo AML, they offer insight into potential mechanisms of resistance in this clinically relevant subset. These results support further investigation into metabolic and immunologic vulnerabilities and provide a rationale for developing combinatorial therapeutic strategies targeting both metabolic reprogramming and immune dysfunction to improve outcomes in VEN-treated s-AML.
