Intrinsic and Microenvironmental Determinants of AML Response to HMA-Venetoclax Therapy at the Single-Cell Transcriptomic Level

The combination of hypomethylating agents (HMA) with Venetoclax has become standard of care in patients with Acute Myeloid Leukemia (AML) unfit for intensive chemotherapy. Genetic and cellular resistance mechanisms, such as RAS and TP53 mutations as well as monocytic differentiation, were recently unveiled (Bataller et al. Blood Adv 2024). Nonetheless, a comprehensive understanding of the resistance mechanisms to HMA-Venetoclax is lacking. To gain further insights into cell-intrinsic and –extrinsic resistance mechanisms, we analyzed single-cell (sc) RNA-sequencing (RNA-seq) data coupled with TCR repertoire profiling from 10 longitudinal sample pairs from AML patients treated with HMA-Venetoclax.

Bone marrow (BM) mononuclear cells from 10 AML patients were collected at diagnosis (TP1) and after the first treatment cycle (TP2). Median age at diagnosis was 70 years, 5 patients were considered at intermediate risk and 5 adverse as per the European LeukemiaNet 2022 classification. 5’ sc-RNA-seq samples coupled with TCR profiling (10X Genomics) were analyzed with a bioinformatic pipeline based on Seurat, ScRepertoire and Bone Marrow Map packages in the R software. Mutational status was obtained by Next Generation Sequencing of bulk DNA at TP1.

Six patients achieved Complete Response with or without complete count recovery (CR/CRi), 2 (20%) were refractory and 2 (20%) experienced early relapse. At TP1, 3 patients (30%) had TP53 abnormalities/complex karyotype; recurrently mutated genes included ASXL1 (30%), DNMT3A (30%) IDH1 (30%) NRAS (30%) and TET2 (30%). Non-responders were enriched in TP53 and RAS mutations.

A total of 126.075 cells were analyzed to assess the role of BM architecture in treatment resistance. Of these, 39.697 cells were unbiasedly classified as Hematopoietic Stem Cells (HSC), Multipotent Progenitors (MPP), Granulocyte Monocyte Progenitors (GMP) and cells of monocytic derivation. Longitudinal differential expression analysis of leukemic cells, comparing non-responders to responders, unveiled the upregulation of: i) TNFa signaling via NF-KB in non-responders at TP2, shaped by the overexpression of JUNB and BCL2A1 in immature and monocytic cells respectively; ii) epithelial to mesenchymal transition (EMT), marked by PRSS2, LGALS1 and BASP1 upregulation in immature myeloid cells of non-responders at both timepoints. Furthermore, higher expression of an established leukemic stem cell score (LSC-17, Ng et al. Nature 2016) in the HSCs of non-responders was observed (p<0.001).

Given the biological relevance of T-cells in AML, T cell phenotype and TCR repertoire were investigated to assess dynamic changes among T cells. A total of 45.539 T cells with paired TCR sequences were recovered. TCR clonotypes were labeled according to their relative frequency into small (<0.1%) medium (<1%), large (<10%) or hyperexpanded groups (>10%), identifying clonal diversity. From TP1 to TP2, responders showed a significant increase in CD4 naïve, CD8 central memory (CM) and effector memory (EM) subsets, combined with a significant increase in small TCR clonotypes (p<0.001), suggesting dynamic reshaping of TCR repertoire. Non-responders were also characterized by an increase of CD4 naïve cells, but decreased CM and EM cells from TP1 to TP2 (p<0.001). In this setting, no significant variations of TCR diversity were highlighted. Importantly, terminally exhausted CD8 T cells were observed only in patients who progressed during treatment. Finally, patients achieving long term response showed a persistent clonally expanded CD8 Terminally differentiated Effector Memory cell subset (TEMRA), characterized by GZMA/B, ZNF863 overexpression and FOS and GZMK downregulation, suggesting high cytotoxic activity involved in active immunosurveillance.

This study provides novel findings associated with response in clonal and immune BM cells after treatment with HMA-Venetoclax in AML. In the context of TP53 and RAS mutation, EMT and inflammatory pathways shape the transcriptional profile of Venetoclax resistance: EMT emerges as a negative predictive factor while inflammatory pathways modulate the expression of Venetoclax targets at different cellular levels according to cell of origin. Increase in TCR diversity and persistence of highly cytotoxic TEMRA clones characterized long term responders while restricted TCR diversity and CD8 T cell exhaustion associated with early relapse.