Regulatory and epigenetic landscapes in biological discovery, diagnostics and disease mechanisms

National Human Genome Research Institute

Investigators

Linked publications, trials & patents

Abstract

Development of methods and biomarkers for blood-based detection of ovarian cancer: The Elnitski lab previously reported the first known DNA methylation biomarker, ZNF154, with pan/multi-cancer relevance (Sanchez-Vega et al. 2013). We also showed that the marker could discriminate tumors from normal samples in simulated, dilute blood-based circulating tumor DNA (Margolin et al. 2014). We developed an assessment method to analyze DNA methylation at the ZNF154 locus, using a quasi-digital PCR and melting curve analysis to assess DNA methylation in circulating tumor DNA molecules (Miller et al. 2020). The highly sensitive and specific test effectively detects tumor DNA in plasma samples from ovarian cancer patients. As a biomarker, ZNF154 detects both serous and endometrioid subtypes of ovarian cancer, making it better than the current gold-standard biomarker, protein antigen CA-125. Along with a computational analysis method we developed, the ZNF154 assay applies to multiple solid epithelial cancers using patient plasma samples (Miller et al. 2020). We recently demonstrated the potential for this assay to detect early and late-stage pancreatic cancers, ovarian and colon cancers from blood testing (Miller et al. 2021). In comparison, blood-based testing of late-stage pancreatic cancer using KRAS mutations performed worse than ZNF154 DNA methylation, and the mutational analysis did not detect early-stage tumors. Profiling epigenetic features in up to 20 solid human epithelial cancer types: I identify attributes that subdivide tumors into homogeneous subsets to potentially improve cancer treatment efficacy using epigenetic landscapes. My group built a computational approach and demonstrated that the CpG island methylator phenotype (CIMP) is present in 14 distinct cancer types from TCGA (The Cancer Genome Atlas) and 23 cancer cell lines (Sanchez-Vega et al. 2015). My lab also published the first report of a CIMP in ovarian endometrioid tumors, and found consistency of the altered DNA methylation patterns among ovarian and uterine endometrioid tumors (Kolbe et al. 2012). We also demonstrated that CIMP tumors from different cancers have shared biological pathways and may have a common underlying etiology (Miller et al. 2016). For example, my lab identified similarities and differences among CIMP tumors from esophageal, gastric, and colorectal adenocarcinomas (Sanchez-Vega et al. 2017), which further addresses how aberrant methylation may occur in multiple tumor types and create phenotypic similarities across cancer types. In further exploration of aberrant DNA methylation patterns, my work demonstrated associations between cancer methylomes and somatic driver mutations in 18 cancer types to illustrate the relationships between the epigenome and the genome in cancers (Chen et al. 2017). Using principal component analysis of methylation-mutation associations, my group showed that genome-wide patterns of aberrant hypomethylation or hypermethylation patterns were associated with specific types of somatic mutations in specific cancer types, indicating that the methylation patterns are relevant to the tumor phenotype. Several site-specific mutations were associated with an extensive number of mutated driver genes, suggesting global epigenetic effects. This year we expanded on our previous findings by showing that DNA methylation patterns in tumor genomes are associated with distinct isoform expression patterns in these same tumors. The methylation locations are predictive of sites of promoter repression, intragenic exon inclusion, and 3'-terminal exon inclusion (Chen and Elnitski 2019). Our discovery of these relationships allows for the prediction of isoform expression based on epigenome analysis. Profiling epigenetic features in human populations: DNA methylation is typically detected using sodium-bisulfite conversion and sequencing. However, the treatment can cause interpretation errors in differential methylation analysis when sequence polymorphisms occur at CpG methylation sites. We developed the algorithm MethylToSNP, which detects characteristic patterns of DNA methylation that demonstrate confounding by polymorphisms (LaBarre et al. 2020). This method is advantageous when no genotype data are available for the methylation samples. We distinguished altered DNA methylation indicative of SNP sites from blood cell samples of YRI (Yoruba in Ibadan, Nigeria), CEPH (European descent), and KhoeSan (Southern African) populations. Furthermore, we identified uncharacterized polymorphisms in these populations and determined locations in the genome that are affected by altered methylation or sequence polymorphisms, including CTCF sites and enhancers. Similarly, my lab helped develop new analysis tools to assess the methylation enrichment of ChIP-seq data (Lichtenberg et al. 2017). My lab compared the epigenomes of KhoeSan individuals from the Kalahari Desert to individuals living in similar geographical locations but from within industrialized societies of Bantu-speaking individuals (Goncearenco et al. 2020). Our analysis identified more than 10,000 differentially methylated sites. The top 5% of differentially methylated sites distinguished the KhoeSan samples from various ethnic groups worldwide. We examined the sites of altered methylation across the KhoeSan genomes in enhancers, gene bodies, and CpG shores and shelves. We found significant underrepresentation in the expected number of methylation changes, suggesting the presence of selective pressure maintaining the integrity of the epigenetic landscape. Our discoveries about the Kalahari KhoeSan epigenomes contribute to the greater understanding of human genome diversity made possible by these individuals. Insights to mechanisms of epigenetic regulation in the human genome: Altered regulation of enhancers can occur through DNA methylation. Our recent study (Chen et al. 2018) assessed data from 14 pairs of monozygotic twins discordant for attention deficit hyperactivity disorder. We found neuroanatomic and epigenetic differences between the siblings. Despite a lack of causal gene mutations, the ADHD-affected twins had a significantly smaller volume in the striatum and thalamus and a trend toward a larger cerebellum. The affected twins showed significant differences in DNA methylation patterns associated with some enhancer regions of genes expressed in the altered brain anatomical structures. These findings are consistent with the idea that subtle changes to enhancer elements in the genome may be associated with discrete neuroanatomical anomalies. In addition to enhancer elements, noncoding regions of the genome also carry repressive elements. Our research examined characteristics of repressive elements and trained a multivariate model of silencer elements. We validated our model results with luciferase-expression assays (Petrykowska et al. 2008) and showed a statistically significant loss of gene expression attributed to our candidate silencers (Huang et al. 2019). We believe that gene repression is involved in differential expression in ovarian cancer subtypes. In this cancer type, we identified evidence for a novel regulatory complex, similar to the MegaTrans complex from breast cancer (Li et al. 2021). Finally, splicing is an important co-transcriptional regulatory process in gene expression. Using mutation data, my lab predicted mutations that affect splicing as part of the CAGI (Critical Assessment of Genome Interpretation) competition. We used a machine learning classifier and accurately identified mutations that cause aberrant mRNA splicing (Gotea et al. 2019). We also are addressing the factors involved in mRNA splicing fidelity by building a computational pipeline. This splicing surveillance mechanism is fundamental to all genes in the human genome by preventing the incorporation of