Linker, S. | Search bioRxiv

bioRxiv2025

KDM5 demethylases suppress R-loop-mediated viral mimicry and DNA damage in breast cancer cells

Tumors with low expression of Interferon-Stimulated Genes (ISG) and Antigen Presentation (AP) genes respond relatively poorly to current immunotherapies. One of the early hallmarks of cancer is DNA hypomethylation in genomic repeat regions, resulting in the expression of normally silenced endogenous "viral" elements. Such epigenetic changes have the potential to augment anti-tumor immune responses as well as reduce tumor cell fitness through the generation of aberrant nucleic acid species (NAS) and consequent activation of NAS-sensing pathways. Therefore, tumor evolution should favor additional selective events that suppress NAS generation, possibly yielding specific therapeutic vulnerabilities. Here, we show that the Lysine Demethylase 5 (KDM5) family of epigenetic regulatory enzymes suppress R-loop formation in genomic repeat regions in cancer cells. We find that KDM5 inhibition in luminal breast cancer cells results in R-loop-mediated DNA damage, reduced cell fitness and an increase in ISG and AP signatures as well as cell surface Major Histocompatibility Complex (MHC) class I, mediated by RNA:DNA hybrid activation of the CGAS/STING pathway. KDM5 inhibition does not result in DNA damage or activation of the CGAS/STING pathway in normal breast epithelial cells, suggesting that KDM5 inhibitors may enable a wide therapeutic window in this setting, as compared to STING agonists or Type I Interferons. These findings provide new insights into the interplay between epigenetic regulation of genomic repeats, R-loop formation, innate immunity, and cell fitness in the context of cancer evolution and therapeutic vulnerability.

cancer biology↗

bioRxiv2017

BrainImageR: Spatiotemporal gene set analysis referencing the human brain

RATIONALENeurological molecular analyses such as transcriptomics, epigenetics, and genome-wide association studies must be assessed in the context of the human brain in order to generate biologically meaningful inferences. It is often difficult to access primary human brain tissue; therefore, approximations are made using in vitro modeling or by identifying disease-associated genes from DNA extracted from blood. Gene sets from these studies are then compared to the post-mortem human brain to provide an assessment of the brain region and the developmental time point that a gene set is most closely associated with. However, most analyses of postmortem datasets are achieved by building new computational tools each time in-house, which can cause discrepancies from study to study, indicating that the field is in need of a user-friendly suite of tools to examine spatiotemporal expression with respect to the postmortem brain. Such a tool will be of use to the molecular interrogation of neurological and psychiatric disorders, with direct advantages for the disease-modeling and human genetics communities.\n\nRESULTSWe have developed brainImageR, an R package that calculates both the spatial and temporal association of a dataset with post-mortem gene expression data from the Allen Brain Atlas. BrainImageR performs a robust analysis of gene set enrichment to identify enriched anatomical regions, provides a global high-resolution visualization of these enrichments across the human brain, and predicts when in developmental time the sample is most closely matched to, a task that has become increasingly important in the field of in vitro neuronal modeling. These functionalities of brainImageR enable a quick and efficient characterization of a given dataset across normal human brain development.\n\nAVAILABILITY AND IMPLEMENTATIONBrainImageR is released under the Creative Commons CC BY-SA 4.0 license and the source code can be downloaded through github at https://github.com/saralinker/brainImageR.\n\nCONTACTgage@salk.edu

bioinformatics↗