Cell-Free RNA Modifications as Biomarkers and Mediators of Immune Evasion

RNA carries more than 170 chemically distinct post-transcriptional modifications, but what most of them do once RNA leaves the cell remains unclear. This thesis develops sequencing methods for detecting RNA modifications in clinical and biological samples and uses them to examine how modified cell-free or extracellular RNAs can contribute biomarker information and influence innate immune sensing. Chapter 2 presents LIME-seq (Low-Input Multiple Methylation Sequencing), a method for profiling multipleWatson-Crick face modifications in ultra-low-input cfRNA from plasma. Applied to colorectal cancer cohorts, LIME-seq showed that microbiome-derived cfRNA modification features discriminated CRC from control samples more effectively than abundanceonly models (AUC = 0.92–0.98 versus 0.77). Modification stoichiometry—the fraction of molecules bearing a given mark at a given site—therefore carries diagnostic information that RNA abundance alone does not. In the cohorts examined here, these findings support cfRNA modification profiling as a proof-of-principle biomarker strategy. Chapter 3 identifies pseudouridine synthase 1 (PUS1) as a regulator of an innate immune checkpoint-like pathway in the tumor microenvironment. The data support a model in which PUS1-dependent pseudouridylation of tumor-derived extracellular mitochondrial tRNAs dampens host TLR7 sensing, weakens myeloid antigen-presentation programs, and reduces responsiveness to anti-PD-L1 across B16F10, CT26, and KPC syngeneic models. Host Tlr7 is genetically required for both the myeloid transition and the therapeutic benefit. In retrospective patient analyses, PUS1 expression is associated with myeloid antigenpresentation deficiency and poor checkpoint response, providing translational context for the model while not yet establishing prospective clinical utility. Chapter 4 introduces CRACI (Chemical Reduction Assisted Cytosine Incorporation sequencing) for quantitative, base-resolution mapping of dihydrouridine (D) across the transcriptome. CRACI enabled quantitative analysis of D sites in human, mouse, and Arabidopsis tRNAs, assignment of the four mammalian DUS enzymes to their major modification sites, and identification of a cis-regulatory relationship between adjacent D sites. The same framework also yielded a small set of candidate D sites in human mRNA, extending the method beyond its main tRNA application while leaving the broader biology of mRNA D unresolved. Taken together, these studies argue that RNA modification state is a measurable layer of information whose significance depends on context: as a biomarker feature in plasma cfRNA, as a determinant of how tumor-derived extracellular RNA is interpreted by host innate immune cells, and as a target for new sequencing chemistry