Search

Single-cell methylation studies are essential where cell-to-cell methylation variation play a key role. It is also critical when the cell samples subject to analysis are precious, rare or in minute amounts. Current techniques available to perform mammalian single-cell methylation analysis rely on bisulfite conversion. The DNA damage and loss caused by bisulfite conversion hinders the methylome coverage, and brings great difficulties in data analyzes, rendering high accuracy differential methylation comparison extremely difficult to perform between individual single cells. In this thesis, the development of a new high coverage methylation sequencing technique is presented, sc-TAPS, for mammalian CpG analysis. sc-TAPS significantly reduces DNA loss and DNA damage, while maintaining a high methyl-conversion rate. Unlike bisulfite conversion, only methylated cytosines will be converted to uracil during sc-TAPS, therefore the DNA complexity remain intact, resulting in a more uniform amplification, higher mapping rate, and less PCR bias. A highly efficient DNA methylation amplification has been developed to compensate for DNA loss during sc-TAPS. Methylome Replication Loops with Methyl-Transferase (MERLOT) utilizes methyl-transferase DNMT1 for maintenance of the methylation pattern. MERLOT amplifies the DNA template 1.7-fold per cycle. Over 95% methylation maintenance efficiency could be achieved by MERLOT while having a low de novo methylated rate of 1.5%. MERLOT combined with sc-TAPS (MERLOT-TAPS), gives by far the highest coverage of all mammalian single-cell whole methylome sequencing techniques known thus far. The low DNA loss and damage property of sc-TAPS makes it possible to combine with many other techniques previously unable by bisulfite conversion. scATAC-TAPS can simultaneously profile nucleosome occupancy and DNA methylation in single cells. Simultaneous measurement of three-dimensional genome structure and DNA methylation could be achieved by combining dip-C with TAPS for mammalian diploid cells. Finally, scTAPS followed by multiplex PCR enables deep methylation sequencing for over 50 specific gene targets simultaneously. Together, sc-TAPS is proved to become a new way for single-cell mammalian methylation analysis with high coverage and high accuracy. Many techniques can be applied along with sc-TAPS to build a toolbox, that could overcome a variety of difficulties previously unsolvable by bisulfite conversion.

DNA methylation is an epigenetic mechanism known to play roles in regulating gene expression in various developmental and disease contexts. However, little is known about its function during eye development. Two types of methylation marks, 5- methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC), are thought to serve as silencing and activating signals for gene regulation, respectively. De novo methyltransferases (dnmt3 family) are responsible for the establishment of 5mC, while cytosine dioxygenases (tet family) convert 5mC into 5hmC, a stable epigenetic mark that can either remain on the genome or undergo subsequent demethylation. Here I performed gene expression and functional tests to elucidate the roles for both of these cytosine-modifying enzyme families during development, with an emphasis on the eye. All dnmt3-family and tet-family genes are expressed tissue-specifically in relevant domains during eye development. Single and double mutants for genes within dnmt3 family develop normally without any overt eye phenotype, indicating that these genes possess redundant functions during eye development. In contrast, in tet2-/-;tet3-/- mutants, retinal neurons are specified but most fail to terminally differentiate. Retinal ganglion cells lack a proper retino-tectal projection, and photoreceptors fail to generate outer segments. Mechanistically, mosaic analyses revealed a surprising cell non-autonomous requirement for tet activity during retinal neurogenesis. Through a combination of candidate gene analysis, transcriptomics and pharmacological manipulations, I identified candidate cell-extrinsic pathways regulated by tet2 and tet3. Additionally, genome-wide 5mC and 5hmC distribution profiles for retinal progenitor cells (RPCs) and differentiated retinal neurons are still currently unknown. To this end, I performed parallel bisulfite and oxidative bisulfite reactions followed by next-generation sequencing (BS/OXBS-seq) to generate the first nucleotide-resolution combined methylome/hydroxymethylome map of retinal cells during development and correlated these with gene expression. This genome- wide approach revealed expected 5mC/5hmC profiles of candidate retinal developmental genes, and identified several novel, uncharacterized genes with potential roles during RPC differentiation. These genes are candidates for further investigation to determine their functions during retinal neurogenesis. Data presented in this Dissertation uncover the role of DNA methylation and hydroxymethylation during eye development and provide the first epigenomic maps of 5mC/5hmC dynamics during retina formation.

Cancer research is rapidly advancing toward more personalized treatments and diagnostics. The major driving force of this advancement are the technological developments of sequencing which has resulted in greatly reduced costs. The democratization of sequencing assays has produced a unique set of challenges that must be addressed in order to successfully conduct clinical research. DNA methylation (DNAm) has increasingly been assayed by sequencing due to its role in human disease and cancers, including breast cancer. DNAm is an epigenetic modification and primarily functions to regulate gene expression. In high-risk breast cancer subtypes, the focus of our most recent research, DNAm is able to predict survival for subtypes with no molecular markers of risk. There are many sequencing assays for DNAm, but the two most common approaches are methylation capture (MethylCap-seq) and bisulfite conversion (BS-seq). As with any sequencing analyte, choosing the appropriate approach for a study is made more difficult by the continual improvements made to the methods and the sequencing technologies. Additionally, each DNAm assay requires a unique data analysis workflow in order to derive valid conclusions. In this work, we will describe a novel computational quantification method for MethylCap-seq and the study design and analysis of a BS-seq experiment as applied to a preventative therapy in breast cancer. Together, these results demonstrate the development and planning needed for the successful implementation of a sequencing assay in clinical research. MethylCap-seq is an approach that uses a protein that binds to DNAm to capture methylated DNA fragments. A region with a large number of captured fragments is interpreted as having high DNAm. Therefore, quantification is based on the number of fragments. Although MethylCap-seq is an inexpensive genome-wide assay, it does have a few limitations. The one we sought to address was the limited resolution of the data. Biologically, DNAm occurs at a single nucleotide; however, the resolution of MethylCap-seq is the fragment which is several hundred nucleotides. To improve the quantification of MethylCap-seq data, we developed a novel computational method. Our method, PrEMeR-CG, imputes a more accurate DNAm quantification at single nucleotide resolution. Since development, PrEMeR-CG has been used to make discoveries in multiple clinical studies.Unlike MethylCap-seq, BS-seq provides DNAm quantification at nucleotide resolution and is considered the gold-standard of DNAm sequencing assays. We utilized BS-seq to assess DNAm for an on-going clinical trial that aims to investigate the efficacy of high-dose polyunsaturated fatty acids (PUFA) treatment to prevent breast cancer in subtypes that have no preventative therapies. This work is motivated by the anti-inflammatory effect observed from PUFA treatment and by pro-cancer effects of inflammation in breast cancer. Our hypothesis is that anti-inflammatory effects of PUFA are mediated by DNAm changes in the breast tissue. In this work, we present two achievements that advance the research of PUFA treatment. First, we describe the design and the approach to investigate DNAm in the breast tissue in an on-going phase II clinical trial of PUFA treatment. One important result from this process was that we developed a novel inflammation related breast cancer specific candidate gene list for interrogating DNAm. Second, we report DNAm changes due to PUFA treatment in the PBMCs of peripheral blood. Our data showed that PUFA treatment resulted in specific DNAm changes in inflammation related pathways. Additionally, we identified potential inflammation related genes that we will follow up as treatment response biomarkers.