Changju Chun, Ja Min Byun, Minkwon Cha, Hongwon Lee, Byungsan Choi, Hyunwoo Kim, Saem Hong, Yunseo Lee, Hayoung Park, Youngil Koh, and Tae-Young Yoon
BH3 mimetics are protein-protein interaction (PPI) inhibitors that saturate anti-apoptotic proteins in the BCL2 family to lift the resistance of cancer cells to apoptosis, a prominent example of protein complex-targeting therapies1–4. Despite the remarkable success of the BH3 mimetic ABT-199 in treating hematological malignancies5–7, only a fraction of patients respond to ABT-1998–10and eventually develop resistance11, 12, necessitating the usage of predictive biomarkers for both initial responses and resistance development. We here used the single-molecule pull-down13, 14and co-immunoprecipitation (co-IP)15, 16platform to quantify more than 20 different types of PPI complexes using ∼1.2x106cells in total, revealing the latest rewired status of the BCL2-family PPI network. By comparing the obtained multi-dimensional data with BH3 mimetic efficacies determinedex vivo, we constructed an analysis model for ABT-199 efficacy. Our model designates the BCL2-BAX and BCLxL-BAK complexes to contrasting roles, with the former complex driving ABT- 199 efficacy and the latter contributing to resistance development. We then applied this model to assist in therapeutic decision-making for acute myeloid leukemia patients in a prospective manner. Our work demonstrates a capability for the extensive characterization of PPI complexes in clinical specimens, potentially opening up a new avenue of precision medicine for protein complex-targeting therapies.