Robust Detection of DNA Hypermethylation of ZNF154 as a Pan-Cancer Locus with in Silico Modeling for Blood-Based Diagnostic Development

Sites that display recurrent, aberrant DNA methylation in cancer represent potential biomarkers for screening and diagnostics. Previously, we identified hypermethylation at the ZNF154 CpG island in 15 solid epithelial tumor types from 13 different organs. In this study, we measure the magnitude and pattern of differential methylation of this region across colon, lung, breast, stomach, and endometrial tumor samples using next-generation bisulfite amplicon sequencing. We found that all tumor types and subtypes are hypermethylated at this locus compared with normal tissue. To evaluate this site as a possible pan-cancer marker, we compare the ability of several sequence analysis methods to distinguish the five tumor types (184 tumor samples) from normal tissue samples (n = 34). The classification performance for the strongest method, measured by the area under (the receiver operating characteristic) curve (AUC), is 0.96, close to a perfect value of 1. Furthermore, in a computational simulation of circulating tumor DNA, we were able to detect limited amounts of tumor DNA diluted with normal DNA: 1% tumor DNA in 99% normal DNA yields AUCs of up to 0.79. Our findings suggest that hypermethylation of the ZNF154 CpG island is a relevant biomarker for identifying solid tumor DNA and may have utility as a generalizable biomarker for circulating tumor DNA.CME Accreditation Statement: This activity (“JMD 2016 CME Program in Molecular Diagnostics”) has been planned and implemented in accordance with the Essential Areas and policies of the Accreditation Council for Continuing Medical Education (ACCME) through the joint providership of the American Society for Clinical Pathology (ASCP) and the American Society for Investigative Pathology (ASIP). ASCP is accredited by the ACCME to provide continuing medical education for physicians.The ASCP designates this journal-based CME activity (“JMD 2016 CME Program in Molecular Diagnostics”) for a maximum of 36 AMA PRA Category 1 Credit(s)™. Physicians should only claim credit commensurate with the extent of their participation in the activity.CME Disclosures: The authors of this article and the planning committee members and staff have no relevant financial relationships with commercial interests to disclose.One in four deaths in the United States is due to cancer, despite an emphasis on prevention, early detection, and treatment that has lowered cancer death rates by 20% in the past two decades.^{1, 2} Further improvements in survival rates are likely to come from improving the limits of detection sensitivity at earlier stages of cancer. Currently, a diagnosis results from a cadre of screening and diagnostic tools that may include physical examination, radiographic imaging, sputum cytologic testing, blood tests, endoscopy, and/or biopsies. However, new approaches that rely heavily on genomic information may change future testing strategies.The future looks bright because minimally invasive sampling techniques coupled with genomic features that distinguish tumor cells from normal cells have the potential to detect cancer at earlier stages. For example, circulating tumor cells or cell-free plasma DNA can be detected in venous peripheral blood and tested for the presence of common mutations.³ Cell-free tumor DNA can also be detected in buccal epithelium, saliva, urine, stools, and bronchial aspirates.⁴ Such DNA has been used to detect mutations in patients with both localized and metastatic cancers.⁵ Moreover, somatic mutations in ovarian and endometrial cancers can potentially be detected using Papanicolaou specimens.⁶In addition to genetic mutations, epigenetic markers are emerging as tools with discriminatory power for disease detection. For example, DNA methylation is a robust epigenetic marker for which a number of commercially available tests have been developed. These tests detect tissue-specific DNA methylation using clinical specimens and are used in colorectal cancer (SEPT9, blood; VIM, stool), lung cancer (SHOX2, bronchial fluid), and brain cancer (MGMT, tumor).⁷ One advantage of this approach is marker stability under common storage conditions.⁴ However, despite DNA methylation''s potential as a diagnostic marker, a general lack of consensus on the methods remains. This is the principal reason for its slow implementation in clinical diagnostics.^{4, 7}Previously, our laboratory reported a pan-cancer hypermethylation signal around a CpG island near human ZNF154.⁸ This signal was initially detected by us in ovarian and endometrial cancers and replicated by us in multiple, independent cohorts from The Cancer Genome Atlas (TCGA), incorporating a total of 15 distinct tumor collections from 13 different organs with almost 6000 samples.⁸ These previous analyses relied on data generated from Illumina Infinium methylation arrays (Illumina Inc, San Diego, CA) to detect the methylation levels at select CpG sites. In this study, we measure the ZNF154 methylation signal across five tumor types using bisulfite amplicon sequencing. With this method, individual sequence reads are used to quantitate methylation levels of all CpGs within the amplicon while providing quantitative data for each DNA molecule in the pooled sample. Furthermore, the approach provides an intrinsic measure of quality control by tracking bisulfite conversion efficiency at cytosines in the non-CpG context wherein extensive amounts of unconverted cytosines signal an incomplete conversion reaction. This procedure is both time efficient and cost-effective because multiple samples can be sequenced in parallel using a 96-well plate and, as we report, generate reproducible measurements when assayed in independent experiments. The amplicon sequencing provides greater resolution of a target region than a methylation array by covering all amplified CpGs, revealing patterns of DNA methylation useful for distinguishing tumor from normal samples.We report that the magnitude and reproducibility of the ZNF154 hypermethylation signal across five solid tumor types reinforces the potential of this site as a biomarker for circulating tumor DNA (ctDNA). Next, we assess the potential application of various computational data classification methods toward cancer screening. By investigating a variety of technical approaches to characterize methylated bases within the sequenced samples, we identify features useful for distinguishing tumor samples from normal samples. Finally, we use a computational simulation to demonstrate the utility of these features in classifying samples as tumor or normal tissue at various abundance levels; here, tumor DNA methylation patterns are compiled into a background of normal DNA methylation patterns, at limiting dilution levels, mimicking the fractions at which ctDNA is recovered from blood.