MODULATION OF UV-INDUCED DNA DAMAGE FORMATION BY CYTOSINE METHYLATION AND TRANSCRIPTION FACTOR BINDING

Ultraviolet (UV) light is one of the most ubiquitous exogenous DNA damaging agents and is the primary etiological agent in skin cancers such as melanoma. UV light from the sun can be directly absorbed by DNA to induce helix-distorting lesions between neighboring pyrimidine bases, including cyclobutane pyrimidine dimers (CPDs) and pyrimidine-pyrimidone (6-4) photoproducts (6-4PPs) along with other atypical photoproducts. Mutagenic bypass of these lesions results in mutation accrual that can contribute to skin carcinogenesis. Previous studies have indicated that DNA sequence context, chemical modifications, and protein-DNA interactions play significant roles in modulating the formation of UV-induced DNA damage. However, the extent to which the base modification 5-methylcytosine (5mC) contributes to UV-induced lesion formation has been widely contested. In Chapter 2 of this dissertation, we elucidate the impact of 5mC on canonical CPD formation following UVB and UVC exposure on an unprecedented genome-wide scale. We found that CPDs are formed on average ~2-fold more frequently at sites of 5mC following UVB exposure. We also report a much more modest effect on CPD formation following UVC exposure. The magnitude of induction in the presence of methylation following each type of UV exposure is highly dependent upon the propensity of the surrounding tetranucleotide sequence to form damage in the absence of methylation. This work serves to advance our understanding of the factors that sensitize DNA to damage formation, particularly within CpG sequences implicated in mutational hotspots of human skin cancers. In Chapter 3 of this dissertation, we explore the effects of protein-DNA interactions on UV-induced lesion formation. We demonstrate that binding of sequence-specific transcription factors (ssTFs) to DNA induces characteristic UV damage patterns known as damage ‘fingerprints’. These fingerprints can be exploited by Random Forest machine learning algorithms to identify both known and novel transcription factor binding sites (TFBS) in yeast and human cells. This technique represents a marked improvement upon current gold-standard chromatin immunoprecipitation (ChIP)-based techniques used to identify TFBS, which require formaldehyde crosslinking and protein-specific antibodies.