Date of Award

Spring 5-15-2022

Author's School

Graduate School of Arts and Sciences

Author's Department

Biology & Biomedical Sciences (Developmental, Regenerative, & Stem Cell Biology)

Degree Name

Doctor of Philosophy (PhD)

Degree Type



Cutaneous melanoma, a cancer of transformed melanocytes, has the highest mortality rate among skin cancers and remains difficult to treat once metastasized. Melanoma is characterized by high genetic heterogeneity, implicating epigenetic dysfunction as an additional regulator of oncogenesis. Indeed, an increasing number of epigenetic modifiers and modifications have been identified in melanoma linked to faster melanoma onset, and then targeted with therapeutics with success. Thus, identifying and mechanistically characterizing these epigenetic and transcriptional alterations in melanoma will further our understanding of their contributions to oncogenesis and enable more diverse therapeutic options.

To evaluate global epigenetic and transcriptional differences between melanocytes and melanoma cells, I performed ATAC-seq and RNA-seq on melanocytes and melanoma cells from an enhanced version of the most widely used zebrafish model of melanoma expressing BRAFV600E, a gain of function mutation in the MAPK pathway present in over half of human melanomas, in combination with a loss of function in p53. For efficient isolation of melanocytes and melanoma cells, I modified the BRAFV600E/p53-/- model with transgenic fluorescent markers for mitfa to label melanocytes combined with an existing crestin reporter to label the reactivation of aspects of an embryonic neural crest program, signifying melanoma onset. In addition to widespread epigenetic and transcriptional alterations between melanocytes and melanoma cells, I identified a complex dysregulation of multiple neural crest genes where key genes were upregulated, including sox10, tfap2a, and dlx2a, but many others were static or downregulated, including sox9b. While dysregulation of a specific transcriptional “subprogram” of neural crest was not evident (e.g. cranial versus thoracic), these data support a model in which portions of both a melanocyte and neural crest program are present in melanomas from the outset and that persist as tumors progress.

Utilizing this ATAC-seq and RNA-seq data, from among 6,760 differentially expressed candidate transcription factors, I focused on four neural crest transcription factors in the SOX and ETS families with associated differentially accessible regions. Three candidates were overexpressed in melanoma relative premalignant melanocytes, including etv4, etv5b, and sox4a, whereas sox9b was downregulated in melanoma. Consistent with its downregulation in melanoma, I found that sox9b overexpression significantly slowed melanoma onset compared to controls, and CRISPR-mediated knockdown of sox9b resulted in faster melanoma onset. This was in stark contrast to the faster onset from overexpression of its close SOXE family member sox10. Intriguingly, clinically detectable zebrafish melanoma tumors with either forced overexpression of sox9b or sox10 had similar global gene expression as evaluated by RNA-seq, despite differences in timing of melanoma onset, supporting a disparate role for these factors in melanoma onset that becomes less apparent in established tumors.

Building on the findings in my zebrafish melanoma models, I sought to better understand the different roles of SOX9/10 in melanoma using human melanoma cell lines. SOX9 and SOX10 have vastly different roles in some developmental contexts and in disease despite similar structures and DNA binding motifs, yet their interactions with each other and mechanisms of downstream regulation in melanoma remain unclear. Using SOX10-high/SOX9-low human melanoma cell lines (A375 and SKMEL5), I overexpressed HA-tagged SOX9 and performed CUT&RUN, RNA-seq, and ATAC-seq and found that in established human tumors, overall gene expression and global active promoters and chromatin accessibility did not change substantially with SOX9 overexpression. While there was a slight decrease in SOX10 expression in SOX9-overexpressing cells, SOX10 continued to bind to the same genomic sites. Indeed, SOX9 also bound to similar sites as SOX10. This suggests that while these family members can behave antagonistically during or near the time of tumor onset, in established melanomas, they can bind similar regions and activate similar pathways to maintain melanoma transcriptional identity.

Overall, I found that while the rate of melanoma onset can be altered by varying the balance of SOXE family members SOX9/10, established tumors have more rigid transcriptional programs which are difficult to perturb. Combined with the current literature, these results suggest a model in which maintaining SOX9 expression may be useful for delaying melanoma onset in melanocytes, whereas inhibiting SOX10 increases may serve as a vital factor in preventing initiation. However, SOX10’s importance diminishes once melanoma has been established, as SOX9 can serve a similar role to promote progression and metastasis. It will be vital in the future to identify binding partners contributing to these distinct processes as well as epigenetic interactions moderating the stability of the melanoma cancer state, as any therapy which attempts to decrease SOX10 activity may have limited effectiveness if SOX9 can replace its function in melanoma. As we continue to illuminate the underlying mechanisms of cancer initiation, maintenance, and progression, understanding such fine levels of transcriptional control and feedback will be essential for uncovering targetable pathways for the next line of therapeutics.


English (en)

Chair and Committee

Charles K. Kaufman

Committee Members

Kristen Kroll


Update embargo

Available for download on Saturday, April 01, 2023