Single-Cell Multiomics Unveils Venetoclax-Resistant Monocytic Differentiation and Immune Evasion in TP53 Mutant AML Clones

Introduction:

Acute myeloid leukemia (AML) is a heterogeneous disease characterized by the accumulation of abnormal hematopoietic stem and progenitor cells. TP53 mutations, present in 8-30% of AML cases, are associated with high risk of relapse and poor prognosis. To investigate the resistance mechanisms, we analyzed 52 longitudinal single-cell RNA sequencing (scRNA-seq) samples from 26 TP53 mutant (TP53-mut) patients. In parallel, we employed single-cell DNA and protein sequencing (scDNA+protein seq) and single-cell mass cytometry (CyTOF) in 10 longitudinal samples from 5 TP53-mut AML patients. This approach allowed us to establish a multiomic (genomic, transcriptomic, proteomic) link in the context of TP53-mut AML to identify potential resistance mechanisms.

Methods:

• For scRNA-seq, we sequenced a total of 97,980 cells from 26 paired samples before and after treatment, 19 of which were TP53-mut.
• For single-cell DNA+protein analysis, we sequenced a total of 44,550 cells from these 10 samples. Samples with at least 70% viability were stained with the Total-Seq™-D Human Heme Oncology Cocktail, containing 45 oligo-conjugated antibodies for surface markers relevant to AML. For mutation analysis, we utilized a validated AML custom panel (Morita et al., 2020) covering recurrent mutations in 37 genes.
• For CyTOF, we multiplexed samples and stained them with 50 antibodies characterizing lineage, apoptosis, and signaling pathways. We successfully constructed a machine learning model for integrating scDNA+protein and CyTOF using our gradient-boosting algorithm established pipelines and were able to predict DNA mutation status in CyTOF data with over 95% accuracy.

Results:

To elucidate the mechanisms underlying TP53-mut AML persistence, we conducted a comprehensive analysis integrating single-cell transcriptional, DNA, and proteomic profiling. Firstly, we developed an advanced pipeline for blast calling in AML, which combined (i) CNV inference using Numbat, (ii) automated cell type annotation (CellTypist), and (iii) surface marker validation using clinical flow cytometry lab gating strategies. This pipeline demonstrated a high correlation between inferred and clinically reported blasts in diagnostic samples (r²= 0.745) exceeding previously reported correlations in scRNA AML studies. However, it yielded poor correlation in post-treatment samples (r²= 0.533).

Clonal evolution analysis using scDNA+protein revealed that TP53-mut clones undergo monocytic differentiation as an escape mechanism. More specifically, the resistant clone shifted from CD34+ blasts pre-treatment to CD16+ monocytic cells post-treatment. Pseudotime analysis of CyTOF data showed consistent findings: post-treatment AML cells lose stemness and express monocytic differentiation markers, suggesting monocytic differentiation as a mechanism of resistance.

Proteomic profiling of resistant cells in TP53-mut AML blasts, annotated in the CyTOF dataset using machine learning models from scDNA+protein data, revealed increased expression of Ki67, BCL-XL, CD68, and CD11b, and decreased expression of HLA-DR and CD34, suggesting an upregulation of anti-apoptotic mechanisms and escape from immune surveillance. TP53 mutation prediction correlated with accumulation of p53 protein, further validating our model performance. Unexpectedly, clonal architecture analysis showed TP53 mutations in lymphocyte subpopulations. For instance, Patient 1 had TP53 mutations in 16.5% of CD4+ T cells, 35.9% of CD8+ T cells, and 19.2% of NK cells. The presence of TP53 mutations in T- and NK cells has not previously been observed. We demonstrate that TP53 mutations are causative of T-cell exhaustion (L Li, et al., M Andreeff, ASH 2024). Additionally, we have identified and validated MYC (Y Nishida et al., M Andreeff, ASH 2024) and PLK4 (E Ayoub et al., M Andreeff, ASH 2022) as potential targets for TP53-mut AML, and we are currently investigating their translational potential in TP53-mut AML.

Conclusion:

This study establishes a genotype-phenotype connection through single-cell multiomic profiling of TP53-mutated AML and maps the clonal evolution and immunophenotypic dynamics during treatment. We propose monocytic differentiation and T-cell exhaustion due to the presence of TP53 mutations in T- and NK cells as a potential mechanism of resistance, providing new insights into AML persistence and therapy escape.