ӳ����ý

Welcome to the May 7, 2021 installment of Research Roundup, a recurring snapshot of recent studies published by scientists at the ӳ����ý and their collaborators.

A protein that helps tumors hide 

Immune checkpoint inhibitors boost the immune system’s ability to find and destroy tumor cells, but they don’t work for most cancer patients. Gabriel Griffin, Jingyi Wu, Arvin Iracheta-Vellve, Tumor Immunotherapy Discovery Engine co-director Robert Manguso, institute member and Gene Regulation Observatory and Epigenomics Program director Brad Bernstein, and colleagues have identified a protein called SETDB1 that helps cancer cells evade the immune system. SETDB1 prevents many transposable elements from producing RNA and proteins. Blocking SETDB1 in tumor cells increased expression of proteins encoded by these elements, which then attracted immune cells to the tumor. The team suggests that inhibiting SETDB1 could make immunotherapy more effective in more patients. Read more in and a ӳ����ý story.

Uncovering the proteomic signature of severe COVID-19

The mechanisms underlying severe COVID-19 remain poorly understood. Infectious Disease and Microbiome Program associate members Michael Filbin and Marcia Goldberg of Massachusetts General Hospital (MGH), Arnav Mehta, institute member and Cell Circuits Program director Nir Hacohen, and colleagues used proteomics to analyze blood samples from 306 COVID-19 patients and 78 symptomatic controls. Published in , their results showed that the plasma proteomic profiles of COVID-19 patients were dramatically different than those of the controls. The team identified 250 proteins associated with severe COVID-19, as well as exocrine pancreas proteases and several immune proteins associated with survival. The findings reveal pathways that may serve as therapeutic targets and help stratify high-risk patients for early interventions. Read more in an MGH story.

A balanced approach to methylation profiling

Bisulfite sequencing is the gold standard for profiling DNA methylation. Whole genome bisulfite sequencing (WGBS) offers the greatest coverage, but is inefficient; reduced representation bisulfite sequencing (RRBS) is more targeted, but often misses enhancers and other elements outside regions known as CpG islands. Sarah Shareef, associate member Volker Hovestadt of Dana-Farber Cancer Institute (DFCI), Bernstein, and colleagues have developed a new extended representation bisulfite sequencing (XRBS) method that uses pooled, barcoded DNA fragments and efficiently measures DNA methylation across promoters, enhancers, and insulators. Writing in , they show that XRBS can assess methylation levels and regulatory states of these elements in small samples and even single cells. 

New tool for interpreting single-cell RNA-seq data

Single-cell RNA-seq (scRNA-seq) datasets are complex and high in dimensionality, which can complicate their representation and interpretation. Common methods for reducing dimensionality are limited by confounding biological and technical variability, cell crowding, and poor representation of temporal relationships between cells. Jiarui Ding and core member (on leave) Aviv Regev developed scPhere, a deep-learning method for modeling these datasets that embeds cells' data into a hyperspherical or hyperbolic space while addressing biological and technical batch effects. scPhere may help solve challenges such as mapping patients’ cells to a reference atlas, pinpointing cells impacted by disease, and mapping cells to spatial locations in tissue. Learn more in .

GWAS gets GUIDANCE

A team led by Josep Mercader of the Metabolism Program and collaborators from Barcelona Supercomputing Center developed a strategy called GUIDANCE that 1) allows analysis of existing and newly-generated genome-wide association study data with better efficiency, 2) tests association under other models beyond the commonly used additive model, and 3) leverages imputation from several reference panels, thus improving the chances of variant discovery. By applying GUIDANCE to genomic data from 62,281 individuals across 22 age-related diseases, the researchers identified 94 genome-wide associated loci, 26 of which were previously unreported. The study, reported in , highlights the benefits of applying comprehensive analytical methods and innovative strategies to better uncover the genetic architecture of complex diseases.

Right off the Bat

The adaptor molecule Bat3’s role in regulating immune responses is clearer, thanks to new findings from Chen Zhu (HMS/BWH), Karen Dixon, Cell Circuits associate member Meromit Singer of DFCI, Regev, and institute member Vijay Kuchroo. The scientists previously showed that Bat3 binds to and suppresses Tim-3, a checkpoint inhibitor that reduces immune responses in autoimmunity, cancer, and chronic viral infection. Using mouse models, they identified a new mechanism by which Bat3 directly suppresses Tim-3 signaling and helps control the mTORC2-Akt-Blimp-1 pathway, preventing terminal differentiation and dysfunction in T cells. Bat3 deficiency may trigger T cell dysfunction in proinflammatory autoimmune conditions. Read more in .

Dynamics of delayed defiance

Most melanoma patients who respond to immune checkpoint blockade eventually develop resistant tumors. Cancer Program associate members David Liu of DFCI and Genevieve Boland of MGH, Jia-Ren Lin (HMS), Emily Robitschek, Epigenomics Program associate member Manolis Kellis of MIT, Alvin Shi, and others studied 37 tumor samples over nine years from a patient with metastatic melanoma, who had complete clinical response but eventually suffered recurrence and died. Seven tumor lineages co-evolved with multiple convergent resistance-associated alterations, and all recurrent tumors emerged from a single lineage lacking chromosome 15q. Appearing in , the work yielded a high-resolution map of evolutionary dynamics of resistance to immune checkpoint blockade, characterized a de-differentiated neural-crest tumor population, and described site-specific differences in tumor-immune interactions.

Predicting cancer's origins

The treatment course for a cancer patient typically depends on the details of their primary tumor. However, some metastatic cancers are first diagnosed as “cancer of unknown primary origin,” posing challenges for decision-making in patients' clinical care. A team led by Ming Lu and  associate member Faisal Mahmood (BWH) has now developed a deep learning algorithm, called Tumour Origin Assessment via Deep Learning (TOAD), that can identify a tumor as primary or metastatic and predict its site of origin using routine histology slides — providing valuable information to physicians. Learn more in and a from Brigham and Women’s Hospital.

To learn more about research conducted at the ӳ����ý, visit broadinstitute.org/publications, and keep an eye on broadinstitute.org/news.