MAC-Attack! Unpacking Responses to Venetoclax with Azacitidine

The addition of venetoclax, a potent BCL-2 inhibitor, to the hypomethylating agent 5-azacitidine (Aza) is transforming treatment options for patients with acute myeloid leukemia (AML) who are not candidates for fully intensive therapy, and is rapidly replacing single-agent treatment with Aza as the standard of care.^1,2  BCL-2 family proteins regulate cellular apoptosis and include both pro-survival mediators and triggers of cell death. The discovery of BCL-2 as an important pro-survival factor in B-cell lymphomas led to the development of venetoclax (Ven) — a highly selective, oral inhibitor of BCL-2 that spares BCL-X_L, a related family member necessary for platelet survival³  — creating a therapeutic that induces apoptotic cell death with reduced thrombocytopenia compared to agents (such as navitoclax) that inhibit both BCL-2 and BCL-X_L.⁴ 

Compared to single-agent Aza therapy, Ven/Aza combination therapy improves rates of complete remission from 28% to 66% and lengthens overall survival by seven months.¹  However, a proportion of leukemias will be resistant to treatment. Predicting which cases are most likely to respond is crucial to spare patients from the potential toxicity of a treatment that is unlikely to be efficacious. Certain molecular subtypes of AML (e.g., those involving NPM1, IDH1/2, TET2, and RUNX1 mutations) have been associated with enhanced sensitivity to Ven, while mutations in FLT3, TP53, SF3B1, and RAS have been associated with a decreased likelihood of response given their reduced dependence on BCL-2.^5,6  Some studies have also suggested that patients with more differentiated, monocytic leukemias tend to have relatively lower expression of the venetoclax target BCL-2, resulting in greater resistance to venetoclax-based therapy.⁷ 

In their recent Cancer Discovery paper, Dr. Alexander Waclawiczek and colleagues highlight a new approach to predicting clinical responses to Ven/Aza based on the expression of BCL-2 family proteins in leukemic stem cells using a simple intracellular flow cytometry-based assay, which yielded a positive predictive value of 97%.⁸ 

The authors first screened 19 AML cell lines for sensitivity to Ven/Aza, stratifying the cell lines according to monocyte-like phenotype (CD64 expression, Mono-AML) vs. more primitive-like AML (Prim-AML) and observing that the Mono-AML lines were more than 150-fold more resistant to Ven/Aza than Prim-AML. However, when similar analyses were conducted in a cohort of 54 patients receiving firstline Ven/Aza therapy, AML remission could not be predicted based on the frequency of CD64⁺ monocyte-like blasts, and no outgrowth of monocytic cells was observed in resistant samples, suggesting that monocytic differentiation was not a reliable predictor of therapy resistance.

The authors next looked in more detail at cellular heterogeneity within the AML tumors, using GPR56 expression to identify “leukemic stem cells” (LSCs), i.e., those with leukemia-initiating capacity. RNA-seq analyses indicated that LSCs exhibited a more than fourfold increase in BCL-2 expression, whereas mature blasts exhibited a more than twofold increase in MCL-1 expression. This pattern was also confirmed at the protein level, using an optimized intracellular flow cytometry staining protocol. Subsequently, Dr. Waclawiczek and colleagues investigated the functional consequences of these differences in expression of BCL-2 family proteins, verifying that LSCs were relatively more dependent on BCL-2 for survival, while mature cells were relatively more dependent on MCL-1. Importantly, the authors observed that LSCs from both Prim-AML and Mono-AML subtypes were highly sensitive to Ven/Aza, suggesting that any differences seen at the bulk level were due to differences in the proportion of the disease-propagating LSCs, rather than intrinsic differences in the mature cells.

They next asked whether responses to Ven/Aza ex vivo could be used to predict clinical responses in a cohort of 53 diagnostic samples from patients with AML. Samples with higher BCL-2 levels were more sensitive to Ven/Aza. Based on the rational assumption that MCL-1 and BCL-X_L both promote survival via pathways independent of BCL-2, they hypothesized that predicting the degree of dependence on BCL-2 would be helpful to predict responses to therapy. Indeed, calculating the “MAC ratio” (BCL-2 / (MCL-1 + BCL-X_L) for LSCs (GPR56+) further improved the predictive value when compared with the values observed for individual protein levels in two independent cohorts, in addition to predicting the response duration.

Of note, the MAC-Score also correlated with the differential responses observed among certain molecular subgroups: Patients with AML who had mutations in TET2, JAK2, and CALR had exclusively low MAC-Scores, in line with clinical observations that prior MDS and MPN are associated with clinical refractoriness to Ven/Aza therapy. Conversely, patients with IDH1/2, DNMT3A, and splicing factor mutations had predominantly high MAC scores.