The treatment of acute myeloid leukemia (AML) has significantly advanced with the introduction of BH3 mimetics, specifically ABT-199 (Venetoclax). Despite initial successes, the challenge of drug resistance and variability in patient responses necessitate a more precise approach to predicting drug efficacy. Our recent study, published in Nature Biomedical Engineering, introduces an innovative method for profiling protein-protein interactions (PPIs) to forecast the effectiveness of BH3 mimetics in AML patients.

AML is a cancer of the blood and bone marrow, characterized by the rapid growth of abnormal white blood cells that interfere with normal blood cell production. This aggressive disease requires prompt and effective treatment. BH3 mimetics, such as ABT-199, have shown promise in treating AML by inducing apoptosis, the programmed cell death crucial for eliminating cancer cells. ABT-199 targets B-cell-lymphoma-2 (BCL2), a protein that prevents apoptosis, thereby promoting the death of cancer cells. However, while ABT-199 has provided significant therapeutic benefits, its effectiveness is limited by the development of resistance and variable responses among patients.

To address this issue, we focused on the dynamic nature of protein complexes within the BCL2 family, the central to the regulation of apoptosis. By using single-molecule pull-down and co-immunoprecipitation (SMPC) techniques, we were able to quantify over 20 different types of PPIs from the millions of small cell sample. This high-resolution profiling provides a detailed map of the protein interactions that govern the cellular response to BH3 mimetics. We also collected multidimensional data from various AML patient samples and correlated these profiles with the ex vivo efficacy of ABT-199. Through this process, we identified that two specific protein complexes, BCL2-BAX and BCLxL-BAK, as critical indicators of ABT-199 response. The presence of BCL2-BAX complexes was associated with higher sensitivity to ABT-199, while BCLxL-BAK complexes were linked to resistance.

Using this data, we developed a predictive model capable of forecasting the efficacy of ABT-199 in individual AML patients. This model identifies patients who are likely to respond to the drug and those who may develop resistance, even enabling to track the intercellular PPI changes. This allows for more personalized treatment strategies. The predictive model was also applied to a cohort of AML patients, where the biomarkers demonstrated high accuracy in predicting the ex vivo efficacy with a 0.94 AUC-ROC (Area Under Curve of Receiver Operating Characteristics) accuracy. Even at the clinical level, our model successfully predicted the ABT-199 response in 9 out of 10 AML patients. The sensitivity of the model was 100%, and the specificity was 83.3%, highlighting its robustness and clinical relevance.

The ability to predict drug efficacy based on PPI profiling represents a significant advancement in precision medicine for AML. By tailoring treatments to the specific molecular characteristics of each patient's cancer, medical team can improve outcomes and reduce the likelihood of resistance. In practical terms, this approach involves obtaining a small sample of cancer cells from the patient, profiling the BCL2 family protein interactions, and applying the predictive model to determine the most effective treatment strategy. This process can be integrated into clinical workflows, providing oncologists with a powerful tool to guide therapeutic decisions.

The success of our study opens the door for similar approaches to be applied to other cancers and targeted therapies. We are exploring the application of our methodology to other BH3 mimetics and anti-apoptotic proteins, aiming to expand the predictive model's applicability. Moreover, our research underscores the importance of understanding the complex interplay of protein interactions within cancer cells. As our knowledge of the interactome grows, so too does our ability to develop more effective and personalized treatments.

Our study represents a significant leap forward in the treatment of AML. By focusing on the intricate network of protein-protein interactions within the BCL2 family, we have developed a method to predict the efficacy of BH3 mimetics with remarkable accuracy. This advancement holds the promise of improved patient outcomes and a new era of precision medicine in oncology.

Author Information: Tae-Young Yoon, Professor in the School of Biological Sciences at Seoul National Univeristy (SNU), and Youngil Koh, Professor in the Department of Internal Medicine at Seoul National University Hospital (SNUH), serve as corresponding authors. The team includes co-lead authors Changju Chun, a biological science graduate student, Ja Min Byun, Professor in the Department of Internal Medicine at SNUH, and Minkwon Cha, a postdoctoral researcher in Pohang University of Science and Technology (POSTECH) (current address). Hongwon Lee, Byungsan Choi and other scientists at PROTEINA Co., Ltd, also participate in the team.