Uncovering Epigenetic Drivers of Tumorgenity in CD133+/- Melanoma Cells

Benshel Bright

Mentor: Dr. Markus Hoffmann, Assistant Professor, Department of Biochemistry and Molecular & Cellular Biology, Georgetown University Medical Center.

Date/Time: August 22nd, 2025 at 12:00 PM.

Abstract: Melanoma is one of the top four cancers among young adults, with an estimated 104,960 cases and 8,430 deaths this year [1]. Melanoma, while not being the most prevalent skin cancer, is responsible for a large proportion of skin cancer deaths [2]. This is mostly due to its ability to metastasize readily. Previous studies have determined that CD133 is associated with stemness and tumorigenic capability, and that CD133 overexpression is correlated with increased chemoresistance and decreased overall survival [3,4]. Prior functional studies show that knockdown of CD133 leads to increased apoptosis, evident by reduced levels of the anti-apoptotic genes AKT and BAD and increased expression of the pro-apoptotic gene BAX [5]. Despite these findings, the epigenetic mechanisms responsible for these phenotypic differences remain largely unexplored. This study aims to investigate the epigenetic regulatory differences underlying the potential to form tumors in melanoma.

This is done through isolating the melanoma-initiating cells based on CD133 expression using magnetic-activated cell sorting (MACS). Melanoma-initiating cells are cancer stem cells that have the ability to divide into more stem cells and differentiated cancer cells [6,7]. Then, the three biological replicates of CD133+ and CD133- samples are sequenced for ATAC-seq and RNA-seq. The open chromatin regions from ATAC-seq are used to create candidate transcription factor (TF) regions [8]. Candidate regions are mapped to nearby genes and scored for transcription factor binding affinity [8]. These affinities are scored by their predicted regulatory impact and then ranked using the differential expression from RNA-seq [8]. This results in a list of transcription factors and the genes they regulate.

The RNA-seq data were processed using TrimGalore! to trim the reads, STAR to map the reads to the reference genome (UCSC_hg38_refGene_10-Jan-2020), and Salmon to quantify the expression (gencode v48) [9]. These steps were done on a private Galaxy instance [9]. All the following steps were run on default parameters from the nf-core/tfactivity pipeline [10]. Differential expression was calculated from the Salmon counts using DESeq2 [8]. The ATAC-seq data were used as open chromatin regions and processed using HINT for footprinting to predict transcription factor binding sites [8]. TEPIC was used to quantify peak-transcription factor affinity using the TRAP model and STARE was used to transform it into gene-level transcription factor affinity scores by linking regulatory regions to nearby genes [8]. A regression model (DYNAMITE) is used to link gene-level transcription factor affinity scores with the observed differential expression from DESeq2 [8]. Per-gene TF scores were calculated by combining affinity, regression coefficient, and differential expression [8]. Finally, a ranking was established with the most significant transcription factors and the genes they are predicted to target [8].

​​TF-Activity prioritized the transcription factor SOX4 (Discounted Cumulative Gain = 0.750, a score ranging from 0 to 1 where 1 is the most relevant items), which plays a crucial role in tissue and organ development, specifically in stem cell activation [11,12]. SOX4 is known to normally promote early differentiation while inhibiting terminal differentiation [12]. TF-Activity showed that SOX4 regulates PROM1, CD133’s gene (normalized gene score = 1.000; a score ranging from 0 to 1, this is the Per-gene TF scores mentioned above). This is consistent with the differential expression, as PROM1 had a log2 fold change of 3.618 (padj = 1.027e-157) and SOX4 had a log2 fold change of 0.482 (padj = 1.169E-06). In the volcano plot for differential expression, PROM1’s p value was unusually low when compared to other significant genes (padj = 1.027e-157). PROM1’s highly significant overexpression is expected, as the original samples were sorted by CD133. We also computationally found SOX4 binding sites in PROM1’s gene region, with a binding sequence of 5’ AACAAAG 3’, further validating SOX4’s function as a regulator of PROM1.

With this data and, hence, our hypotheses, we conclude that CD133’s overexpression in melanoma-initiating cells was caused by the transcription factor SOX4’s increase in expression. SOX4 has also been abnormally expressed in other cancers [13]. More importantly, overexpression of SOX4 causes overexpression of the NF-κB signaling pathway [13]. This pathway impacts stress responses, immune responses, inflammation, cell growth and survival, creating an environment favorable to the cancer [13,14]. Together, these findings indicate that melanoma stemness may be epigenetically sustained through SOX4-mediated transcriptional changes.

TF-Activity also prioritized STAT1 (DCG = 0.179), a core JAK–STAT pathway transcription factor that drives antigen presentation in cells [15,16]. From our samples, STAT1 was slightly downregulated in CD133⁺ cells (log2 FC = -0.199, padj = 0.201), and the pipeline did not strongly link STAT1 to CD133. TF-Activity also prioritized STAT3 (DCG = 0.500), a transcription factor that supports survival, proliferation, and wound-healing phenotypes [17]. STAT3 was likewise downregulated in CD133⁺ cells (log2 FC = -0.280, padj = 0.056), and the model did not identify it as a primary driver of CD133 expression. Both STAT1 and STAT3 regulated many of the same genes, which ranged from other transcription factors to long noncoding RNAs.

There were several challenges that I encountered during this project. First, hardware limitations in cores and memory made it difficult to efficiently process large-scale sequencing data. Because I ran most of the analysis on my computer, large data files required careful file management. I also faced chromosome naming mismatches between reference genomes and input files. which created formatting issues that required troubleshooting. Finally, because nf-core/tfactivity is still under active development, small bugs existed between steps, making it hard to differentiate pipeline errors and formatting errors.

The next step will be experimental validation of SOX4’s binding site in PROM1’s regulatory region using CUT&RUN. A preliminary Western Blot comparing CD133+/- groups confirms an increase of SOX4 protein in the CD133+ cells. Validating this interaction will provide substantial evidence that SOX4 regulates CD133 expression in melanoma-initiating cells. In the long term, targeting SOX4 or its downstream regulatory network could represent a therapeutic strategy to reduce melanoma stemness and tumorigenic potential.

In this study, we combined ATAC-seq, RNA-seq, and the nf-core/tfactivity pipeline to investigate transcriptional regulators of tumorigenicity in CD133+/- melanoma cells. SOX4 was determined to be a key regulator of CD133 expression, supported by both computational predictions and the identification of SOX4 binding motifs within the PROM1 gene region. STAT1 and STAT3 were also prioritized, though they regulated broader immune and survival-related functions instead of PROM1. Despite the challenges, this project demonstrates the usefulness of multiomic approaches to uncover potential epigenetic drivers of cancer stemness. Moving forward, CUT&RUN will be critical to experimentally confirm SOX4’s regulatory role. Overall, our findings predict SOX4 to be a promising upstream regulator CD133 in melanoma-initiating cells and provide a potential target for future therapeutic exploration.

References: “Melanoma of the Skin – Cancer Stat Facts.” n.d. SEER. Accessed August 16, 2025. https://seer.cancer.gov/statfacts/html/melan.html. Girouard, Sasha D., and George F. Murphy. 2011. “Melanoma Stem Cells: Not Rare, but Well Done.” Laboratory Investigation; a Journal of Technical Methods and Pathology 91 (5): 647–64. Li, Zhong. 2013. “CD133: A Stem Cell Biomarker and Beyond.” Experimental Hematology & Oncology 2 (1): 17. Simbulan-Rosenthal, Cynthia M., Ryan Dougherty, Sahar Vakili, Alexandra M. Ferraro, Li-Wei Kuo, Ryyan Alobaidi, Leala Aljehane, et al. 2019. “CRISPR-Cas9 Knockdown and Induced Expression of CD133 Reveal Essential Roles in Melanoma Invasion and Metastasis.” Cancers 11 (10): 1490. Simbulan-Rosenthal, Cynthia M., Yogameenakshi Haribabu, Sahar Vakili, Li-Wei Kuo, Havens Clark, Ryan Dougherty, Ryyan Alobaidi, Bonnie Carney, Peter Sykora, and Dean S. Rosenthal. 2022. “Employing CRISPR-Cas9 to Generate CD133 Synthetic Lethal Melanoma Stem Cells.” International Journal of Molecular Sciences 23 (4): 2333. Girouard, Sasha D., and George F. Murphy. 2011. “Melanoma Stem Cells: Not Rare, but Well Done.” Laboratory Investigation; a Journal of Technical Methods and Pathology 91 (5): 647–64. Villani, Vincenzo, Francesco Sabbatino, Cristina R. Ferrone, and Soldano Ferrone. 2015. “Melanoma Initiating Cells: Where Do We Stand?” Melanoma Management 2 (2): 109–14. Hoffmann, Markus, Nico Trummer, Leon Schwartz, Jakub Jankowski, Hye Kyung Lee, Lina-Liv Willruth, Olga Lazareva, et al. 2022. “TF-Prioritizer: A Java Pipeline to Prioritize Condition-Specific Transcription Factors.” GigaScience 12 (December). https://doi.org/10.1093/gigascience/giad026. Afgan, Enis, Clare Sloggett, Nuwan Goonasekera, Igor Makunin, Derek Benson, Mark Crowe, Simon Gladman, et al. 2015. “Genomics Virtual Laboratory: A Practical Bioinformatics Workbench for the Cloud.” PloS One 10 (10): e0140829. “Tfactivity: Introduction.” n.d. Accessed August 16, 2025. https://nf-co.re/tfactivity/dev/. Hanieh, Hamza, Emad A. Ahmed, Radhakrishnan Vishnubalaji, and Nehad M. Alajez. 2020. “SOX4: Epigenetic Regulation and Role in Tumorigenesis.” Seminars in Cancer Biology 67 (Pt 1): 91–104. Moreno, Carlos S. 2020. “SOX4: The Unappreciated Oncogene.” Seminars in Cancer Biology 67 (Pt 1): 57–64. Cheng, Qiong, Jinfeng Wu, Yaohua Zhang, Xiao Liu, Nan Xu, Fuguo Zuo, and Jinhua Xu. 2017. “SOX4 Promotes Melanoma Cell Migration and Invasion Though the Activation of the NF-κB Signaling Pathway.” International Journal of Molecular Medicine 40 (2): 447–53. Mao, Hongmei, Xiaocui Zhao, and Shao-Cong Sun. 2025. “NF-κB in Inflammation and Cancer.” Cellular & Molecular Immunology 22 (8): 811–39. Hu, Xiaoyi, Jing Li, Maorong Fu, Xia Zhao, and Wei Wang. 2021. “The JAK/STAT Signaling Pathway: From Bench to Clinic.” Signal Transduction and Targeted Therapy 6 (1): 402. Tolomeo, Manlio, Andrea Cavalli, and Antonio Cascio. 2022. “STAT1 and Its Crucial Role in the Control of Viral Infections.” International Journal of Molecular Sciences 23 (8): 4095. “Role of Stat3 in Skin Carcinogenesis: Insights Gained from Relevant Mouse Models.” n.d. Accessed August 16, 2025. https://onlinelibrary.wiley.com/doi/10.1155/2013/684050.