Machine learning-fueled antibiotic discoveries, revelations in single-cell genomics' association with disease, and more.
Research Roundup: February 21, 2020
Welcome to the February 21, 2020 installment of Research Roundup, a recurring snapshot of recent studies published by scientists at the Broad Institute and their collaborators.
Ironing out ferroptosis
Accumulation of polyunsaturated phospholipid hydroperoxides in cellular membranes is a hallmark of iron-dependent cellular death, or ferroptosis. A team led by core faculty member Stuart Schreiber, postdoctoral researcher Yilong Zou in the Chemical Biology and Therapeutics Science group, and Haoxin Li (Harvard University) conducted genome-wide CRISPR-Cas9-mediated suppressor screens and identified cytochrome P450 oxidoreductase (POR) as necessary for ferroptosis in susceptible cancer cells. Reporting in Nature Chemical Biology, they show that POR contributes to ferroptosis induced by diverse mechanisms in various cancer types. The work suggests that targeting POR could help treat degenerative diseases involving ferroptosis.
Bacteria bring bad blood
A bacterial infection can lead to sepsis, a life-threatening dysregulation of the immune system. A team including Miguel Reyes, a graduate student in Paul Blainey’s lab; associate member Michael Filbin with Massachusetts General Hospital; core faculty member Paul Blainey; associate member Marcia Goldberg of the Harvard T.H. Chan School of Public Health; institute member and Cell Circuits Program co-director Nir Hacohen; associated scientist Roby Bhattacharyya in the Infectious Disease and Microbiome Program; and research associate Kianna Billman in the Blainey lab used single-cell RNA-sequencing to profile the blood of people with sepsis. In Nature Medicine, they analyzed gene-expression profiles to describe 16 immune-cell states, including a unique CD14+ monocyte state that’s expanded in people with sepsis. They characterized the monocyte state and its surface markers, proposed a model for its derivation from human bone marrow cells, and demonstrated that single-cell genomics can reveal molecular signatures associated with disease.
To fear, or not to fear?
The dorsal hippocampus region of the human brain plays a critical role in encoding contextual information and relaying it to other regions of the brain that mediate fear and stress responses. However, the neural pathways that guide these processes and ultimately lead to adaptive behavioral responses are poorly understood. Using optogenetics and electrophysiological recordings in mice brains, associate member Amar Sahay from the Stanley Center for Psychiatric Research and colleagues identified that the CA3 subregion of the hippocampus plays a critical role in routing and calibrating contextual fear discrimination. Read more in Cell Reports.
Dynamics of immune cell regulation
Memory CD4+ T cells are key immune system players, and many genetic variants associated with autoimmune disorders have been found in these cells' regulatory elements. To better understand such relationships, a team led by postdoctoral scholars Maria Gutierrez-Arcelus and Yuriy Baglaenko, with institute member Soumya Raychaudhuri, stimulated memory CD4+ T cells from healthy individuals and tracked their gene expression over time. They discovered a dynamic landscape, including conditions in which cells preferentially expressed one allele over another, and further used CRISPR-Cas9 to explore control of HLA-DQB1 gene expression. The results illustrate the shifting regulatory landscape during T cell activation and how it is affected by individual genetics. Read more in Nature Genetics.
Machine learning helps discover new antibiotics
Postdoctoral scholar Jonathan Stokes, Regina Barzilay of MIT, and institute member Jim Collins at MIT, and their colleagues used a machine learning–guided approach to identify a new antibacterial compound — which they named halicin — from the Drug Repurposing Hub. The molecule killed many disease-causing bacteria in vitro, including some strains that are resistant to all known antibiotics. It also cleared infections in two different mouse models. The deep learning model analyzed more than a hundred million molecular structures from the ZINC15 database in a matter of days and picked out several additional antibiotic candidates. Learn more in Cell.