Cell-free DNA 5-hydroxymethylcytosine profiles of long non-coding RNA genes enable early detection and progression monitoring of human cancers.

background5-Hydroxymethylcytosine (5hmC) is a significant DNA epigenetic modification. However, the 5hmC modification alterations in genomic regions encoding long non-coding RNA (lncRNA) and their clinical significance remain poorly characterized.

resultsA three-phase discovery-modeling-validation study was conducted to explore the potential of the plasma-derived 5hmC modification level in genomic regions encoding lncRNAs as a superior alternative biomarker for cancer diagnosis and surveillance. Genome-wide 5hmC profiles in the plasma circulating cell-free DNA of 1632 cancer and 1379 non-cancerous control samples from different cancer types and multiple centers were repurposed and characterized. A large number of altered 5hmC modifications were distributed at genomic regions encoding lncRNAs in cancerous compared with healthy subjects. Furthermore, most 5hmC-modified lncRNA genes were cancer-specific, with only a relatively small number of 5hmC-modified lncRNA genes shared by various cancer types. A 5hmC-LncRNA diagnostic score (5hLD-score) comprising 39 tissue-shared 5hmC-modified lncRNA gene markers was developed using elastic net regularization. The 5hLD-score was able to accurately distinguish tumors from healthy controls with an area under the curve (AUC) of 0.963 [95% confidence interval (CI) 0.940-0.985] and 0.912 (95% CI 0.837-0.987) in the training and internal validation cohorts, respectively. Results from three independent validations confirmed the robustness and stability of the 5hLD-score with an AUC of 0.851 (95% CI 0.786-0.916) in Zhang's non-small cell lung cancer cohort, AUC of 0.887 (95% CI 0.852-0.922) in Tian's esophageal cancer cohort, and AUC of 0.768 (95% CI 0.746-0.790) in Cai's hepatocellular carcinoma cohort. In addition, a significant association was identified between the 5hLD-score and the progression from hepatitis to liver cancer. Finally, lncRNA genes modified by tissue-specific 5hmC alteration were again found to be capable of identifying the origin and location of tumors.

conclusionThe present study will contribute to the ongoing effort to understand the transcriptional programs of lncRNA genes, as well as facilitate the development of novel invasive genomic tools for early cancer detection and surveillance.