Systematic mapping of BCL-2 gene dependencies in cancer reveals molecular determinants of BH3 mimetic sensitivity

Ryan S. Soderquist, Lorin Crawford, Esther Liu, Min Lu, Anika Agarwal, Grace R. Anderson, Kevin Lin, Peter Winter, Merve Çakır, Kris C. WoodNature Communications 2018 · DOI

Abstract

While inhibitors of BCL-2 family proteins (BH3 mimetics) have shown promise as anti-cancer agents, the various dependencies or co-dependencies of diverse cancers on BCL-2 genes remain poorly understood. Here we develop a drug screening approach to define the sensitivity of cancer cells from ten tissue types to all possible combinations of selective BCL-2, BCL-X<sub>L</sub>, and MCL-1 inhibitors and discover that most cell lines depend on at least one combination for survival. We demonstrate that expression levels of BCL-2 genes predict single mimetic sensitivity, whereas EMT status predicts synergistic dependence on BCL-X<sub>L</sub>+MCL-1. Lastly, we use a CRISPR/Cas9 screen to discover that BFL-1 and BCL-w promote resistance to all tested combinations of BCL-2, BCL-X<sub>L</sub>, and MCL-1 inhibitors. Together, these results provide a roadmap for rationally targeting BCL-2 family dependencies in diverse human cancers and motivate the development of selective BFL-1 and BCL-w inhibitors to overcome intrinsic resistance to BH3 mimetics.

Introduction

Programmed cell death (apoptosis) is a natural safeguard against cancer, but many tumors survive by overexpressing anti-apoptotic proteins from the BCL-2 family. While BH3 mimetics—drugs designed to inhibit these proteins—have revolutionized the treatment of certain leukemias (e.g., Venetoclax), they have largely failed in solid tumors. The challenge lies in understanding the complex logic of these dependencies. This study systematically maps these dependencies across various cancer types to identify the molecular "switches" that govern sensitivity.

Methods and Datasets

The authors developed a high-throughput drug screening approach using three selective inhibitors: ABT-199 (targeting BCL-2), WEHI-539 (targeting BCL-X<sub>L</sub>), and A-1210477 (targeting MCL-1). They tested every possible single, double, and triple combination across 78 cancer cell lines from the Cancer Cell Line Encyclopedia (CCLE) and 13 primary patient-derived cultures. Functional data were integrated with genomic and transcriptomic data, and CRISPR/Cas9 loss-of-function screens were used to identify resistance mechanisms.

Key Results

A Comprehensive Map of Sensitivity

The screening revealed that 63% of cell lines were sensitive to the combination of BCL-X<sub>L</sub> and MCL-1 inhibitors, suggesting significant untapped therapeutic potential. Interestingly, a tumor's tissue of origin proved to be a better predictor of dependency than specific oncogenic mutations like TP53 or KRAS.

Figure 1: Heatmap showing % viability loss across a panel of CCLE cell lines treated with various BH3 mimetic combinations.

Molecular Determinants of Dependency

The study identified a correlation with Epithelial-Mesenchymal Transition (EMT). Epithelial cells tend to be co-dependent on BCL-X<sub>L</sub> and MCL-1, while mesenchymal cells shift their addiction specifically to BCL-X<sub>L</sub>. This shift is mediated by the PERK-NOXA axis; high NOXA levels neutralize MCL-1, leaving the cell dependent on BCL-X<sub>L</sub>. An mRNA threshold for NOXA (LOG2 > 10.33) was established to predict sensitivity to BCL-X<sub>L</sub> inhibition.

In Vivo Validation and Resistance Mechanisms

Using a mouse xenograft model, the researchers demonstrated that BCL-X<sub>L</sub> inhibition significantly reduced tumor volume in lines predicted to be sensitive.

Figure 2: Average tumor volume over time in a mouse xenograft model, comparing vehicle control to treatment with the BCL-X<sub>L</sub> inhibitor A-1331852.

In cells resistant to all combinations, CRISPR screens identified BFL-1 and BCL-w as the primary drivers of resistance. Knocking out these genes sensitized previously resistant cells to BH3 mimetics.

Discussion and Conclusion

This study provides a roadmap for rationally targeting BCL-2 family dependencies. While clinical implementation faces hurdles—such as the lack of clinical-grade BFL-1/BCL-w inhibitors and the dose-limiting toxicity of BCL-X<sub>L</sub> inhibition—the identification of biomarkers like NOXA levels offers a path toward personalized treatment strategies that maximize efficacy while minimizing toxicity.

Summary updated

Abstract

While inhibitors of BCL-2 family proteins (BH3 mimetics) have shown promise as anti-cancer agents, the various dependencies or co-dependencies of diverse cancers on BCL-2 genes remain poorly understood. Here we develop a drug screening approach to define the sensitivity of cancer cells from ten tissue types to all possible combinations of selective BCL-2, BCL-X<sub>L</sub>, and MCL-1 inhibitors and discover that most cell lines depend on at least one combination for survival. We demonstrate that expression levels of BCL-2 genes predict single mimetic sensitivity, whereas EMT status predicts synergistic dependence on BCL-X<sub>L</sub>+MCL-1. Lastly, we use a CRISPR/Cas9 screen to discover that BFL-1 and BCL-w promote resistance to all tested combinations of BCL-2, BCL-X<sub>L</sub>, and MCL-1 inhibitors. Together, these results provide a roadmap for rationally targeting BCL-2 family dependencies in diverse human cancers and motivate the development of selective BFL-1 and BCL-w inhibitors to overcome intrinsic resistance to BH3 mimetics.
BibTeX
@inproceedings{soderquist2018systematic,
  author = {Ryan S. Soderquist and Lorin Crawford and Esther Liu and Min Lu and Anika Agarwal and Grace R. Anderson and Kevin Lin and Peter Winter and Merve Çakır and Kris C. Wood},
  title = {{Systematic mapping of BCL-2 gene dependencies in cancer reveals molecular determinants of BH3 mimetic sensitivity}},
  booktitle = {Nature Communications},
  year = {2018},
  doi = {10.1038/s41467-018-05815-z},
}
Figure