Publication Date




Embargo Period


Degree Type


Degree Name

Doctor of Philosophy (PHD)


Cancer Biology (Medicine)

Date of Defense


First Committee Member

J. William Harbour

Second Committee Member

Nanette H. Bishopric

Third Committee Member

Robert B. Levy

Fourth Committee Member

Noula Shembade

Fifth Committee Member

Ivonne H. Schulman


Uveal melanoma and retinoblastoma are the two most common primary intraocular malignancies, with uveal melanoma being the most common in adults and retinoblastoma in children. For both tumors, a critical need exists to find better biomarkers for disease progression and metastasis and therapeutic targets to help treat these deadly and debilitating diseases. For uveal melanoma, our laboratory has previously developed a gene expression profile test (UM-GEP) that is performed on a needle biopsy sample from the primary tumor and renders a result of Class 1 (low risk) or Class 2 (highrisk). While the UM-GEP provides the most accurate prognostication currently available for UM, there was a critical need to further improve its prognostic accuracy, since ~10% of Class 1 UMs metastasize. Thus, we conducted global transcriptomic analyses of Class 1 metastatic samples compared to Class 1 samples that did not metastasize with at least three years of follow-up. We found that the most significantly upregulated gene in Class 1 metastatic samples was Preferentially Expressed Antigen in Melanoma (PRAME), and we validated PRAME expression was predictive of metastasis in an additional three independent datasets. Most recently, we analyzed PRAME on more of our samples and over 500 de-identified samples provided by Castle Biosciences (the company that commercially provides the UM-GEP test). We found that PRAME+ is in 15% of Class 1 and 30% of Class 2. These retrospective results showed that PRAME also predicts increased risk of metastasis in Class 2 tumors, but it is a more significant predictor in Class 1 patients. This biomarker is now being prospectively validated in a multi-center study as part of the Collaborative Ocular Oncology Group 2 (COOG2), a cohort of 70 ocular oncologists throughout the United States. In parallel with biomarker discovery, we sought to better understand when the capacity for metastasis emerges during uveal melanoma. As mutations in BAP1, SF3B1, and EIF1AX are associated with high, intermediate, and low risk of metastasis, respectively, we focused on studying the evolutionary emergence of these and other characteristic genomic alterations. In order to do this, we conducted a global genetic and epigenetic analysis of uveal melanoma on 138 exomes. As part of this analysis, we developed an optimized pipeline to identify the key driver mutations in uveal melanoma, particularly large deletions in the tumor suppressor BAP1, which were missed using existing bioinformatics pipelines. Additionally, new key driver mutations of spliceosome components were identified that were previously unrecognized in this disease. Strikingly, using a combination of subclonality algorithms, we found that BAP1 mutations and all other canonical genomic aberrations arise in an early punctuated burst, followed by neutral evolution extending to the time of clinical detection. These provocative findings suggest that the metastatic proclivity of UM is “set in stone” early in tumor evolution, which may explain why decades of advances in primary tumor treatment have not led to a measurable improvement in survival. For retinoblastoma, while biallelic inactivation of the RB1 tumor suppressor gene is the common initiating event for virtually all retinoblastomas, the genetic and genomic landscape underlying tumor progression and varying patient outcomes remains largely unknown. A highly aggressive phenotype in retinoblastoma that is associated with poor outcome and increased chemoresistance is tumor seeding into the vitreous and subretinal space. Through exome sequencing of advanced retinoblastoma tumors, while no single mutation was common to all patients, we identified mutational profiles clustering around estrogen-related receptor gamma (ESRRG), affecting many proteins known to bind and regulate its function. In addition, the most common copy number alteration was a gain on chromosome 1q, where ESRRG is located. Subsequent RNA-sequencing analyses revealed high expression of ESRRG in retinoblastoma samples and in vitro testing further confirmed that primarycultured and Rb cell lines decrease proliferation, reduce cell aggregation, and undergo cell death in response to drug inhibition of ESRRG, as well as shESRRG-mediated knockdown. Further analyses, demonstrated that dysregulation of ESRRG may be due to RB1 loss, which was found to directly bind to and regulate ESRRG. ChIP-Seq analyses found that ESRRG binds to genes essential for neurogenesis and metabolism, including many genes that are involved in oxidative phosphorylation and glycolysis, and ESRRG was found to localize specifically at the tumor edge along the hypoxic vitreous. These findings suggest that ESRRG may promote a metabolic shift in retinoblastoma, adapting the tumor for survival in the hypoxic retina. Discoveries laid out in this thesis provide new biomarkers and drug targets for both uveal melanoma and retinoblastoma and shed light on the evolutionary mechanisms of both tumors. Ultimately, our hope and main goal is to utilize these and future findings to help save patients’ eyes and lives.


uveal melanoma; retinoblastoma; PRAME; BAP1; ESRRG; cancer

Available for download on Wednesday, May 06, 2020