Molecular subtypes and BH3 mimetic synergy with anti-leukemia agents in T-cell acute lymphoblastic leukemia

Abstract

Abstract Compared to B-cell acute lymphoblastic leukemia (ALL), treatment outcomes of T-ALL remain inferior, especially in relapsed or refractory cases, with few targeted or immunotherapeutic options. BH3 mimetics, which target anti-apoptotic BCL2 family proteins, represent a promising class of agents in hematological malignancies. However, their efficacy and potential in combination with other anti-leukemia agents in T-ALL remain unknown, especially in the context of the modern molecular taxonomy of this cancer (Pölönen et al, 2024). In this study, we evaluated the ex vivo cytotoxicity of three investigational BH3 mimetics, namely AZD4320 (BCL2/BCL-XL dual inhibitor), AZ'3202 (BCL-XL inhibitor), and AZD5991 (MCL1 inhibitor), using an imaging-based cell viability assay in a panel of 58 patient-derived xenograft (PDX) models of T-ALL. In parallel, we performed RNA-seq and/or whole-genome seq of all cases for subtype classification. Applying network-based Bayesian analyses (NetBID2) to T-ALL transcriptomic data, we also inferred gene activities in a genome-wide fashion. AZD4320 and AZD5991 showed subtype-dependent activity (P = 0.0066 and P = 3.5 × 10-4, respectively), with significantly greater sensitivity in ETP-like and resistance in TAL1 αβ-like cases. Using NetBID2-inferred gene activity, we found that BCL2, BCL2L1 (coding BCL-XL), and MCL1 activities strongly predict T-ALL sensitivity to AZD4320 (P = 0.0087, P = 4.1 × 10-4, and P = 0.0011, respectively). Interestingly, the impact of BCL2 signaling on T-ALL drug sensitivity also varied slightly by subtype. These findings highlight the interplay between molecular subtype and apoptotic signaling and their effects on BH3 mimetic sensitivity in T-ALL. Next, to investigate the therapeutic potential of BH3 mimetics in combination settings for T-ALL, we evaluated the interactions between AZD4320 and key anti-leukemic agents, i.e., asparaginase, prednisolone, nelarabine, and an LCK inhibitor, dasatinib, across 40 T-ALL PDX samples ex vivo. Using Multi-dimensional Synergy of Combinations (MuSyC) analysis to determine both synergistic efficacy (the combined maximal effect, defined by β) and synergistic potency (the extent to which one agent enhances the effective concentration of the other, defined by log(α)), we observed distinct patterns of drug-drug interactions. While there was limited synergy by efficacy, AZD4320 combined with asparaginase or dasatinib exhibited significant synergistic potency. Notably, the AZD4320–asparaginase combination resulted in bidirectional synergistic potency (i.e., both log(α1) and log(α2) >0, P = 1.8 × 10-5 and P = 1.1 × 10-5), which was consistent across T-ALL molecular subtypes. By contrast, the interaction of AZD4320 with dasatinib was uni-directional: AZD4320 potentiated dasatinib but not vice versa (log(α2) >0, P = 0.007). This synergy was also subtype-dependent, with greatest interactions in dasatinib-sensitive subtype TAL1 αβ-like cases. Finally, we sought to validate AZD4320 synergy in vivo, focusing on dasatinib and asparaginase. In three T-ALL PDX models, the combination of AZD0466 (a drug-dendrimer conjugate of AZD4320) with asparaginase consistently exhibited greater efficacy compared to monotherapy (vs AZD0466, P = 0.0007, P = 0.015, and P = 0.0006; vs asparaginase, P = 0.011, P = 0.038, and P = 0.06). Together, our study identified subtype-informed strategies to combine BH3 mimetics with chemotherapeutics in T-ALL. In particular, the synergy by potency between these drugs point to the possibility for dose reduction to mitigate toxicities while delivering greater efficacy than single agent therapy.