Single-cell developmental classification of B cell precursor acute lymphoblastic leukemia at diagnosis reveals predictors of relapse

Abstract

Improved insight into cancer cell populations responsible for relapsed disease will lead to better outcomes for patients. Here, we report a single-cell study of B-cell precursor acute lymphoblastic leukemia at diagnosis that revealed hidden developmentally dependent cell signaling states uniquely associated with relapse. With mass cytometry, we simultaneously quantified 35 B-cell developmental proteins in 60 primary diagnostic samples. Each leukemia cell was then matched to it’s nearest healthy B-cell population by a developmental classifier that operated at the single-cell level. Machine learning identified 6 features of expanded leukemic populations sufficient to predict patient relapse at diagnosis. These features implicated pro-BII cells with activated mTOR signaling, and pre-BI cells with activated and unresponsive pre-B-cell receptor signaling, to be associated with relapse. This model, termed Developmentally Dependent Predictor of Relapse (DDPR), significantly improves currently established risk stratification methods. DDPR features exist at diagnosis and persist at relapse. Leveraging a data-driven approach, we demonstrate the predictive value of single-cell ‘omics’ for patient stratification in a translational setting and provide a framework for application in human cancers.

How Clinical Flow Cytometry Rebooted Sepsis Immunology

“…The general concept of utilizing mass cytometry to characterize immune states associated with disease progression or response to therapy has been demonstrated in multiple clinical contexts , including malignancies , rheumatological diseases , aging , traumatic injury , and pregnancy . While the use of mass cytometry in sepsis is still in its infancy , observational studies in the context of traumatic injury provide the groundwork for the deep immune profiling of complex inflammatory states such as sepsis.…”

High risk glioblastoma cells revealed by machine learning and single cell signaling profiles

“…For example, Levine et al used Phenograph 4 and an understanding of the coupling of surface markers and signaling status in healthy bone marrow to classify negative prognostic leukemia cells. Similarly, a map of the healthy developmental lineage was instrumental in using DDPR to identify features of negative prognostic leukemia cells 6 . Supervised methods, including CITRUS and Cytofast, require that samples to be grouped at the beginning of the analysis before generating an overview of cell cluster phenotypes in cytometry data 54 .…”

“…CellCNN is another supervised analysis tool that requires prospective assignment of samples to categories and uses convolutional neural networks to learn a filter that predicts whether new cells match one of the groups 15 . Other cell subset discovery approaches do not supervise the analysis with knowledge of clinical outcomes but do use prior biological knowledge to identify cell subpopulations and then test whether differential outcomes are associated these cell subsets 5,6,16 . In mass cytometry analysis, another common approach is to use tools for automated, unsupervised cell discovery and characterization, including SPADE 17 , t-SNE 18 , UMAP 19 , FlowSOM 20 , and Marker Enrichment Modeling (MEM 21 ).…”

“…However, bench and computational tools that have driven our understanding of altered phospho-protein signaling networks in cancer have historically been under-used in solid tumor research due to a lack of technology and samples. In blood cancers, single cell profiles of signaling networks have revealed cancer cells present at diagnosis whose abundance is closely linked to subsequent clinical outcomes, including patient survival [3][4][5][6][7] . This single cell snapshot proteomics approach focuses on a select set of key proteins that govern cell functional identity and can robustly measure targets, such as phospho-proteins, that are inaccessible to sequencing modalities.…”

From Bivariate to Multivariate Analysis of Cytometric Data: Overview of Computational Methods and Their Application in Vaccination Studies

“…Automated analysis was applied to a dataset derived from a large retrospective study of individuals at the early stage of HIV infection, and allowed to identify three T-cell subsets whose frequency during early infection had a statistically significant relationship with clinical progression to AIDS [20]. In the field of hematologic malignancies, successful application of computational methodologies have also been reported for acute lymphoblastic leukemia (AML) aimed at improving the discrimination between leukemic and normal cells [73,74], identifying B cell precursor as predictors of disease relapse [75], monitoring the minimal residual disease [76,77], evaluating the disease progression [78], or characterizing the immune alterations in AML patients [79]. Moreover, computational methods have been used on existing clinical flow cytometric data to improve diagnostic accuracy to distinguish mantle cell lymphoma from small lymphocytic lymphoma [80], or discriminate various subpopulations of blood cells in the context of B-chronic lymphocytic leukemia [81].…”