Decoding RNA Biology: RNA Modification Detection and Noncoding RNA Function

RNA modifications and noncoding RNAs are emerging as key regulators of gene expression, yet their functional roles remain incompletely understood because accurate, transcriptome‑wide, quantitative, single‑nucleotide-resolution detection methods are limited, particularly for low-input samples. This dissertation develops sequencing‑based approaches for low‑input mapping of selected RNA modifications, focusing on 5‑methylcytosine (m⁵C), pseudouridine (Ψ), and N⁶‑methyladenosine (m⁶A) and, in a separate line of work, investigates how Prader-Willi syndrome (PWS)‑encoded C/D box small nucleolar RNAs (SNORDs) regulate gene expression and contribute to disease‑relevant phenotypes.

The first project establishes a technically feasible strategy for joint detection of Ψ and m⁶A in low-input RNA within the same library, providing a basis for future analyses of Ψ/m⁶A co‑occurrence and potential modification crosstalk. Initial m⁶A‑SAC‑seq/Ψ‑BID‑seq and m⁶A‑CAM‑seq/Ψ‑PUM‑seq combinations demonstrated conceptual feasibility but produced short, low‑yield libraries and relied on operationally complex or chemically damaging protocols. Replacing m⁶A‑CAM‑seq with the milder m⁶A‑GLORI 2.0 protocol and pairing it with PUM‑seq preserved kilobase‑scale RNA and enabled dual chemical treatment at nanogram‑scale input. Optimization of reverse‑transcription conditions and adaptor load further improved library quality, demonstrating that dual PUM/GLORI 2.0 libraries can be robustly generated from low-input RNA. This work provides a practical foundation for future quantitative analyses of Ψ/m⁶A co-occurrence while also improving GLORI 2.0 by producing libraries with longer fragment sizes than the original workflow.

The second project extends ultra‑mild bisulfite sequencing (UMBS‑seq) to low‑input RNA to enable transcriptome‑wide, quantitative, single‑nucleotide‑resolution mapping of m⁵C from nanogram‑scale samples, making base‑resolved m⁵C profiling accessible for scarce and compositionally challenging RNA populations. It combines ultra‑mild bisulfite chemistry with a ligation‑free, random‑primed, template‑switching library workflow and show that, across a 10‑fold input range (10 ng, 3.3 ng, 1 ng), UMBS‑treated libraries preserve relatively long inserts while maintaining high C‑to‑U conversion at non‑methylated sites and robust signal at known rRNA m⁵C positions.

The third project investigates how PWS‑encoded SNORDs, SNORD115 and SNORD116, regulate gene expression and contribute to PWS‑relevant phenotypes. It defines the first transcriptome‑wide interactomes for these SNORDs in mouse brain tissue and human cells, shows that Snord116/SNORD116 recognize RNA targets through conserved D′‑proximal seed motifs, and reveals a non‑canonical SNORD-mRNA interaction mode. It also identifies U1 snRNA as a prominent SNORD116 partner that forms stable, structurally defined duplexes and shows that SNORD116 loss globally increases nascent transcription while leaving RNA decay unchanged, with preferential up‑regulation of ribosome‑, translation‑, mitochondrial‑, and transcriptional regulator genes and SNORD116‑targeted mRNAs enriched for transcriptional and neuronal regulators. Together, these findings suggest that SNORD116 regulates gene expression in PWS through U1‑linked co‑transcriptional control, modulation of ribosome‑ and translation‑related programs, and direct regulation of selected neuronal and regulatory mRNAs.

Collectively, the methods developed in this thesis expand the toolkit for decoding RNA modification landscapes from low‑input samples and lay the groundwork for studying modification co‑occurrence and crosstalk, while the PWS work defines how disease‑linked SNORDs engage their RNA targets and modulate gene expression via co‑transcriptional and translational programs. These advances provide new opportunities to investigate RNA modifications and noncoding RNAs and highlight candidate pathways that may be particularly relevant for PWS.