Pancreatic adenocarcinoma is the most common malignancy of the pancreas, accounting for about 85% of all pancreatic cancer cases. Although it is only the 11th most common cancer globally, pancreatic adenocarcinoma comprises a disproportionate number of total cancer deaths. It ranks 7th worldwide for cancer mortality, responsible for over 400,000 annual deaths globally. Though advances in cancer research and clinical oncology have significantly improved patient outcomes for many human cancers, the prognosis of pancreatic adenocarcinoma remains dismal: ~5% would live for five years after diagnosis. Early detection in high risk individuals or population screening appears to be the only viable option for increased chances of patient survival. In this study, we sought to develop a robust diagnostic model for pancreatic adenocarcinoma based on genome-wide mapping of 5-hydroxymethylcytosines (5hmC), a novel epigenetic marker that has been implicated in cancer pathobiology, in circulating cell-free DNA (cfDNA). Specifically, we prospectively recruited 158 patients with pancreatic adenocarcinoma (age: 60.9 ± 11.5; males: 57.0%), 91 patients with non-cancer pancreatic diseases (age: 48.3 ± 15.6, males: 41.8%), and 159 healthy individuals (age: 47.5 ± 12.0, males: 62.2%) from Peking Union Medical College Hospital and Zhongshan Hospital of Fudan University in China. This cohort was split into a training set and two validation sets. The patient diagnosis was confirmed by a study pathologist. We excluded patients treated with chemotherapy, radiation therapy or immunotherapy and then obtained baseline clinical, laboratory, and treatment data from medical records using a standard protocol. We reviewed pathology and medical records to verify disease progression or recurrence. We then performed genome-wide 5hmC mapping in cfDNA using the 5hmC-Seal technique, a highly sensitive method utilizing only 1-2 ng of cfDNA from <5 mL of plasma. A diagnostic model was then developed using a case-control design and elastic net regularization on a multivariate logistic regression model. Notably, the 5hmC-Seal data in cfDNA were enriched in genomic areas of pancreas-derived cis-regulatory markers (e.g., H3K4me1, a mark for tissue-specific enhancers). A weighted diagnostic score (wd-score) based on 10 genes (e.g., HOXD3, CSF1, and PLEKHG ), some of which have been found to be dysregulated in pancreatic adenocarcinoma (e.g., HOXD3), accurately distinguished cancer patients from healthy individuals in the training (AUC = 91.2%; 95% CI, 86.4-96.0%) and two validation sets (AUC = 91.0%; 95% CI, 83.4-98.5%; and AUC = 85.7%; 95% CI, 77.0-94.5%). This model markedly outperformed conventional cancer biomarkers (e.g., CEA, carcinoembryonic antigen) and shows that 5hmC-Seal mapping in cfDNA can identify a robust diagnostic model for detection of pancreatic adenocarcinoma.

Citation Format: Zhou Zhang, Lei You, Xingyu Lu, Huanyu Wang, Jinshou Yang, Taiping Zhang, Emily Kunce Stroup, Wenhui Lou, Chuan He, Wei Zhang, Yupei Zhao. Genome-wide 5-hydroxymethylcytosine mapping in circulating cell-free DNA Identifies robust diagnostic markers for pancreatic adenocarcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 433.