Single-Cell Proteogenomics Analysis of AML Samples Treated with Decitabine and Venetoclax Reveals Divergent Treatment Escape Mechanisms

Background: The combination therapy of hypomethylating agents and venetoclax (HMA+VEN) has emerged as a standard of care for acute myeloid leukemia (AML) patients unfit for intensive chemotherapy. Nonetheless, significant clinical challenges persist, including relapse and treatment resistance. Previous studies have revealed genetic (RAS/RTK pathway mutations and TP53 mutations) and phenotypic (monocytic differentiation) resistance mechanisms in the context of HMA+VEN therapy. Despite these advancements, our comprehensive understanding of the HMA+VEN resistance mechanisms remains incomplete. To gain further insights into the resistance mechanism in HMA+VEN therapy, we implemented single-cell proteogenomics analysis on longitudinally collected samples from AML patients treated with decitabine+venetoclax (DAC+VEN) trial.

Methods: We performed a high-throughput, single-cell targeted DNA-antibody sequencing (DAb-seq) using Tapestri platform from Mission Bio Inc. Sixty-three AML samples sequentially collected from 33 unique patients who participated in DAC+VEN clinical trial (NCT03404193) were analyzed. Our custom DNA panel covered 37 recurrently mutated genes in AML, whereas the custom antibody panel included either 15 cell surface proteins (N = 20 patients) or 48 proteins (N = 13 patients). Cellular phenotypes were inferred based on the patterns of cell surface protein expression. Median 2,767 cells were sequenced per sample, with total 196,719 cells sequenced for this study.

Results: The median age of the cohort was 70 years (range 21-84). Twenty-four (73%) patients had primary AML and 9 (27%) had secondary or therapy-related AML. Based on the FAB classification, 13 patients had M0/M1/M2, 6 had M4/M5, one had M7, and the rest had unknown subtypes. The most frequent gene mutation at baseline was DNMT3A (45%), followed by NPM1 (39%), TET2 (30%), and TP53 (24%). Among the 33 patients, 32 had evaluable treatment response; 27 (82%) were defined as responders by achieving CR or CRi, whereas 5 (15%) were non-responders. Six patients experienced relapse after responding to the therapy, and their baseline-relapse pairs were analyzed by the single-cell platform. Using the composition of cellular phenotype data, AML samples were classified into either stem-like (N = 5), progenitor-like (N = 1), predominantly monocytic (N = 4), predominantly erythroid (N = 2) and mixed population AML (mixed population of stem, progenitor, monocyte or erythroid-like population, N = 21). Genotype-phenotype analysis showed a significant correlation between the KRAS mutation and the mixed phenotype (p= 0.0403), as well as between the U2AF1 mutation and the mixed phenotype (p= 0.0471). There was no significant association between treatment response and AML phenotype. Analysis of six baseline-relapse pairs showed that relapse was associated with phenotypic shift of AML cells. Stem-like cells were largely eradicated at relapse and replaced by more differentiated cells. Among the six baseline-relapse pairs, three showed monocytic shift, one showed erythroid shift and two showed mixed monocytic/erythroid shift. Phenotypic shifts at relapse were not always associated with expansion of new genetic mutations, except for in two cases: one had monocytic shift accompanied with expansion of KRAS mutation, and another had erythroid shift accompanied with expansion of NRAS mutation.

Conclusions: Phenotypic transition from stem-like cells to more differentiated cells (either monocytic or erythroid) was frequently observed in AML relapse following DAC+VEN therapy. These findings are consistent with the known dependency of these cells on non-BCL-2 anti-apoptotic proteins such as MCL-1 and BCL-xL. Close monitoring of these phenotypic alterations throughout the course of HMA+VEN treatment may potentially facilitate the precise identification of cellular populations predisposed to relapse.