Abstract IA30: Using single-cell, high-dimensional approaches to unravel tumor heterogeneity in pediatric cancer

Abstract

Abstract Introduction: Despite improving responses to up-front therapies for children with cancer, relapse remains a significant barrier to long-term remissions and cure. Curing children of cancer requires effectively targeting all cancer-initiating cell populations. Understanding what cellular populations have initiating capacity and lead to poor clinical outcomes requires examining tumors at the single-cell level. Here we discuss such an approach to predict relapse in B-cell precursor acute lymphoblastic leukemia and response to common ALL therapies including glucocorticoids and chimeric antigen receptor T cells. Methods: Primary diagnostic or relapse bone marrow or peripheral blood samples or patient-derived xenografts were obtained under informed consent and IRB approval. Samples were clinically annotated for relevant prognostic features and clinical outcomes. Cryopreserved samples and healthy control bone marrows were treated in vitro with short-term stimulations to reveal signaling states. Cells were stained with a 40-antibody panel including relevant B-cell developmental proteins and intracellular signaling proteins. Samples were analyzed by mass cytometry (CyTOF). Patient samples underwent developmental classification in which each leukemia cell was assigned its most similar healthy counterpart. Further analysis was performed using various analysis methods, including machine learning modeling of relapse outcomes. Results: Studying over eighty primary patient diagnostic samples, we first organized the heterogeneous single-cell data classifying leukemic cells to 12 different subpopulations of B lymphopoiesis. We identified the transitional populations between early pro-B cells and late pre-B cells are expanded across almost all patients with BCP ALL. Extracting all measured features in these populations, we applied an elastic net model to determine cell populations associated with future relapse. This resulted in identification of early pre-B cells characterized by basally activated pCREB, pS6, pSYK, and p4EBP1 as predictive of future relapse. Further, these cells are apparent also at the time of relapse. Taking a similar approach, we also examined how these cell populations respond to treatment with glucocorticoids, a keystone of BCP ALL therapy. Similarly, we identify the same network activation in glucocorticoid-resistant patients associated also with a differentiation process. Interestingly, we can identify similar resistant phenotype in primary cells from patient treated in vivo with glucocorticoids. Finally, using these tools and methods, we describe CD19neg cells identified prior to treatment with CD19-targeted CAR T cells in patients who go on to suffer CD19neg relapse. Conclusions: Together, we present an approach to organizing single-cell tumor heterogeneity that reveals relapse-associated phenotypes. This highlights the translational potential of such an approach that could be applied to other single-cell studies in diverse tumor types. Citation Format: Kara L. Davis. Using single-cell, high-dimensional approaches to unravel tumor heterogeneity in pediatric cancer [abstract]. In: Proceedings of the AACR Special Conference on the Advances in Pediatric Cancer Research; 2019 Sep 17-20; Montreal, QC, Canada. Philadelphia (PA): AACR; Cancer Res 2020;80(14 Suppl):Abstract nr IA30.