Transcriptional Correlates of Tolerance and Lethality in Mice Predict Ebola Virus Disease Patient Outcomes

Abstract

Host response to infection is a major determinant of disease severity in Ebola virus disease (EVD), but gene expression programs associated with clinical outcome are poorly characterized. Using the Collaborative Cross (CC) mouse model of genetic diversity, we developed a model of differential EVD severity. CC mice develop a strain-dependent spectrum of distinct EVD phenotypes, ranging from tolerance (mild, transient disease with full recovery) to lethality (severe disease that may include hemorrhagic syndrome). We performed a screen of 10 CC lines with differential phenotypes and identified clinical, virologic, and transcriptomic features that distinguish tolerant from lethal outcomes. Tolerance is associated with tightly regulated induction of immune and inflammatory responses early following infection, as well as reduced numbers of inflammatory macrophages and increased numbers of mature antigen-presenting cells, B-1 cells, and γδ T cells, allowing for control of viral replication and subsequent recovery. Lethal disease is characterized by broad suppression of early gene expression and reduced quantitiesof lymphocytes, followed by uncontrolled inflammatory signaling leading to death. Using machine learning classification, we developed and trained a transcriptomic signature that predicted outcome in CC mice at any time point post-infection with 99% accuracy. This signature predicted outcome in a cohort of EVD patients from West Africa with 75% accuracy, demonstrating its utility as a prognostic tool to guide EVD patient treatment in future outbreaks.

Keywords: Ebola, virus, mouse, Collaborative Cross, Host-pathogen interactions, host response, hemorrhagic fever, tolerance, lethality, infection, Transcriptomics, systems biology, data integration, transcriptional regulation, B-1 cells, gamma delta T cells, machine learning, classification