Precision Genomics from Gene-Regulatory Dynamics: Immunity, Prognosis, and Pharmacogenomics

Abstract

Advancements in technologies for sequencing the genome and the computational methodologies to process and extract information from the data have rapidly accelerated the understanding of the genetic basis for many diseases. The practice of precision medicine enables more precise targeting of disease, better appreciation of individual patient needs, and a higher resolution understanding of the genetic basis underlying drug responses by translating insights from genomics to clinical practice. Looking to the future, modulation of the epigenome as a therapeutic intervention promises major potential to revolutionize the approach to managing complex diseases like cancer and inflammatory or metabolic disorders. The major goals of the work presented in this dissertation are to leverage genomic data to improve our understanding of 1) immunity – particularly how dysfunctional immune cells respond differently to infectious diseases like COVID-19; 2) disease prognosis – whereby epigenetic biomarkers reveal how immune responses are primed within the first few days of infection and underlie the severity of symptoms; and 3) pharmacogenomics – in order to develop a molecular framework for the evaluation of selective glucocorticoid receptor modulators. To address the first goal, we investigated differences in innate immunity associated with mortality in COVID-19 ICU patients. Prior to decompensation, abundances of non-classical monocytes were significantly lower, and we profiled these cells following stimulation with TLR agonists using single-cell RNA-seq. I identified a transcriptional pattern of tolerance against TLR activation in the monocytes from deceased patients that explained the absence of a robust innate immune response in the agonist conditions. We concluded that secondary infections may occur more frequently in these patients, thus increasing the risk of mortality, and tested nucleic acid-scavenging MnO nanoparticles as a potential therapy to neutralize the hyperinflammatory environment and reverse the tolerance phenotype. To address the second goal, we identified biomarkers that differentiated mild from moderate Covid-19 prior to IgG seroconversion, when antibodies begin to be produced, using single-cell ATAC-seq. The IgG-negative window lasts for only the first few days following infection. Thus, our goal was to improve the understanding of COVID-19 immunity – that is, how immune responses were differentially primed and associated with disease severity – as well as to develop translational molecular targets with prognostic and therapeutic potential. The chromatin landscape was indeed remodeled significantly prior to seroconversion. Furthermore, classical monocytes had the highest enrichment of regions with differential accessibility, and I characterized prognostic biomarkers that included ~1000 domains of regulatory chromatin and differences in TF activity associated with monocyte maturation that underlie disease severity. To address the third goal, we conducted the largest comparative study of genomic responses to 10 GR ligands, measuring differential gene expression and regulatory element activity using RNA-seq and STARR-seq. 1 in 5 Americans have used short-term classical glucocorticoid therapies, such as dexamethasone, and repeated or excessive use carries a significant risk of adverse side effects. Safer ligands, proposed to selectively modulate GR activity by impairing trans-activation of some target genes, have consistently underperformed in clinical trials. I explained why by demonstrating that these ligands activate the same gene expression response as dexamethasone, just to different degrees in strength. Indeed, the ligand-specific activities at ~50,000 enhancers were dominantly explained by dexamethasone effects (R2 = 0.82). I reported the genes in the most ligand-dependent mode of these responses to be used for future genetic screens. Finally, I simulated enhancer activities using a two-component Gaussian mixture model to represent activation of a subset of dexamethasone-responsive enhancers by a hypothetical highly selective ligand, the response to which was significantly different from any of the ligands we tested (p = 0.03).