Sub-phenotypes of pneumonia defined by pulmonary histopathological features
Many host-directed therapies are available for severe pneumonia, including anti-inflammatories, immunostimulants, fibrinolytics, and more. Multiple clinical studies show that these therapies are effective in subsets of patients. Incomplete understanding of the varying types of lung biology that cause severe pneumonia prevents the rational and effective provision of host-directed therapies to appropriate patients. A team from the Pulmonary Center used microscopy to examine lung tissues of people who died with pneumonia, to better decipher the underlying biology in these failing lungs. They scored multiple different histopathological features across hundreds of subjects, and then used machine learning to define patterns and associations amongst these features. Many different pathologies were observed, with no single feature consistently found across all pneumonias, confirming that pneumonia includes heterogeneous pulmonary responses. Using machine learning, the microscopic data segregated these pneumonia cases into seven different types of pulmonary pathology. These sub-phenotypes involved different types of immune cells in the lungs, and they could be modeled experimentally. These pulmonary pathology sub-phenotypes of pneumonia now become the focus of new research designed to associate discriminatory biomarkers, define underlying mechanisms, and test responses to therapies.
The studies were led by bioinformatics PhD student Amulya Shastry and postdoctoral fellow Dr. Brad Hiller, with mentorship from Dr. Jay Mizgerd. The multidisciplinary team involved other members of the Pulmonary Center and throughout BU (including Computational Medicine, Pathology and Laboratory Medicine, Virology, Immunology, and Microbiology, and the NEIDL) as well as collaborators from the Banner Sun Health Research Institute in AZ and the Mass General-Brigham. The study was published in the American Journal of Respiratory and Critical Care Medicine.