Mechanistic dissection of venetoclax sensitivity in AML through single-molecule BCL2 protein complex profiling and mutation-specific apoptotic rewiring

Abstract

Abstract Venetoclax (VEN) combined with hypomethylating agents (HMA) has improved response rates in elderly or unfit acute myeloid leukemia (AML) patients, yet 30–40% of treated individuals fail to achieve durable remissions. Traditional genomic risk stratification alone (e.g., ELN 2022) does not capture the dynamic interplay of pro- and anti-apoptotic signals that drive treatment sensitivity or resistance. To better resolve the apoptotic signaling mechanisms driving the VEN + HMA response, we employed the single-molecule protein interaction detection (SPID) platform, which enables high-resolution imaging of endogenous protein–protein interactions (PPIs) (Chun et al., Nat Biomed Eng, 2024). Using this platform, we quantified 12 BCL2-family PPIs in pre-treatment AML samples from 47 patients at Winship Cancer Institute. In a subset of 28 patients, paired pre- and on-treatment samples were analyzed to track VEN+HMA-induced PPI remodeling. Each assay measured complex occupancy between anti-apoptotic proteins (BCL2, BCL-xL, MCL1) and their pro-apoptotic binding partners (e.g., BAX, BIM, BAK), normalized to total protein levels. Somatic mutations were defined by targeted sequencing. Baseline PPI analysis revealed mutation-specific rewiring of apoptotic complexes that clarify differential VEN + HMA responses. AMLs with isolated FLT3-ITD mutations (without NPM1 or DNMT3A co-mutations) exhibited marked alterations in BCL2-family PPIs compared to FLT3-ITD–wildtype AML. These included elevated BCL2–BIM complexes, increased levels of unoccupied BCL-xL and MCL1, and higher total BCL-xL and MCL1 protein levels—suggesting enhanced buffering capacity for BIM and BAX, which may be displaced by VEN. In contrast, AMLs with isolated NPM1 mutations showed minimal or modest PPI shifts, indicating a more apoptotically primed state at baseline. Notably, co-mutations of FLT3-ITD with either DNMT3A or NPM1 synergistically increased anti-apoptotic complex formation beyond levels seen with single mutations, establishing FLT3-ITD as the dominant driver of baseline apoptotic rewiring. These findings suggest a mutational hierarchy in regulating anti-apoptotic dependency, with FLT3-ITD exerting the strongest effect, followed by DNMT3A and NPM1. Paired-sample analysis revealed dynamic, response-associated shifts in BCL2 family PPIs during VEN + HMA therapy. In responders, VEN-induced displacement of BAX from BCL2 was accompanied by high BCL2 occupancy (low levels of unoccupied BCL2), consistent with effective apoptotic engagement. In contrast, non-responders exhibited persistent or increased BCL2–BAX complexes with low BCL2 occupancy (high levels of unoccupied BCL2), suggesting inefficient VEN binding and impaired apoptotic activation. Notably, all TP53-mutant AMLs showed stable or increased BCL2–BAX complexes and elevated unoccupied BCL2 on-treatment relative to pre-treatment profiles, indicative of persistent apoptotic resistance. In contrast, NPM1- and IDH1/2-mutant AMLs demonstrated the opposite pattern—reduced BCL2–BAX occupancy and increased BCL2 occupancy following VEN-HMA treatment—consistent with effective apoptosis induction. Additionally, VEN + HMA treatment led to marked displacement of BAK from BCL-xL and MCL1 in IDH1/2-mutant AMLs, further highlighting distinct, mutation-specific apoptotic rewiring that underlies VEN sensitivity or resistance. This integrative PPI–genomic analysis establishes a mechanistically grounded framework for predicting response to VEN + HMA therapy in AML. By linking mutation-specific apoptotic rewiring to quantitative single-molecule measurements of BCL2 family PPIs, we uncovered complex interaction patterns that are not detectable by conventional protein assays. Together, this work supports the development of a next-generation, biomarker-driven strategy for VEN + HMA therapy in AML, with the potential to guide personalized treatment decisions and improve clinical outcomes.