Maturity State and MCL-1 Dependence Predetermines Response to NOTCH1 Inhibition in T-ALL

Introduction:

Acute T cell lymphoblastic leukemia (T-ALL) is an aggressive hematopoietic malignancy in children and young adults that frequently becomes treatment-refractory and relapses. The Notch1 pathway is a key oncogenic driver in T-ALL and is aberrantly activated in more than 50% of the cases. Despite promising pre-clinical data using gamma secretase inhibitors such as DBZ to target NOTCH1, resistance is rapidly occurring in vivo. As molecular heterogeneity has been linked to treatment escape, we focused our study on defining transcriptional cell states driving resistance to NOTCH inhibition and understanding their relation to mitochondrial priming.

Methods:

5 primary T-ALLs harboring NOTCH activating mutations were engrafted in NSG (NOD-scidIL2Rgnull) mice. Upon reaching ~ 10% of human CD45+ positive leukemic blasts in the peripheral blood, randomized groups of 8 mice per primary T-ALL were treated with DBZ (Dipenzazepine; 10 μM/kg every other day through tail vein) or vehicle (VEH). 3 mice per group were sacrificed after one week of treatment to assess short-term effect of DBZ, while the remaining 5 mice were weekly monitored for disease progression, leukemic blasts were collected from lymphoid organs and overall survival was determined.

Full-length transcriptome analysis of 3188 blasts present in the blood of 20 sensitive and 22 refractory mice was performed by Smart-Seq2. Based on scRNA features, 'scVelo' and 'CytoTRACE' were used to identify developmental potential and differentiation trajectories. Cell fate and transcriptional regulatory networks were defined and reconstructed using 'SCENIC'.

Assessment of mitochondrial priming as measured by BH3 profiling was used to identify anti-apoptotic vulnerabilities present in these PDX models.

Results:

Upon DBZ, short or long-term disease control was observed in two strains, while rapid resistance occurred in three strains, thus establishing two sensitive and three refractories to NOTCH inhibition PDX models. Immunohistochemical analysis showed decreased expression of active NOTCH1 in spleen biopsies of all strains, validating the efficacy of DBZ and suggesting a mechanism of resistance independent of ICN1.

Single cell transcriptional profiling showed enrichment of immature hematopoietic signatures and co-expression of lymphoid and myeloid progenitor programs in refractory models. Interestingly, pre-existing cells harboring refractory-like transcriptional circuits within the untreated sensitive population were identified. Upon treatment, despite increased differentiation in all models, lineage promiscuity was maintained in refractory strains, suggesting that cellular plasticity mediates treatment escape.

Next, we characterized cell states driving treatment refraction. RNA velocity projections identified two distinct immature states differing in cell cycle and oncogenic signaling. Clustering of untreated, sensitive leukemic cells in immature state imply that aberrant lineage commitment can predict response to NOTCH inhibition in vivo. These observations were further confirmed by differentiation state analysis, where prior to treatment, high developmental potential was correlated to treatment escape. Surprisingly, in addition to early lineage differentiation drivers such as BCL11A, state-specific regulons analysis associated immature states with BCLAF1 a transcriptional regulator of apoptosis.

We postulated that these transcriptional circuits lead to differential apoptotic priming, therefore the dependence on individual anti-apoptotic proteins was evaluated. Mitochondrial priming at baseline revealed BCL-2 dependence in sensitive strains whereas MCL1-dependence was observed in refractory ones. Upon DBZ treatment, while dependency profiles in refractory strains remained unchanged, a functional switch from BCL-2 to MCL1-dependency occurred in sensitive models.

Conclusion:

Our results suggest that response to NOTCH inhibition is predetermined by cell maturity states and their associated transcriptional circuits responsible for differential sensitivity to apoptotic priming via BCL2 and MCL1. These data suggest that combining BH3 and lineage commitment profiling may predict drug responses in vivo. Moreover, our findings highlight the importance of targeting co-existing cell states to overcome transcriptional heterogeneity as a driver of treatment escape.