Cell-free DNA 5-hydroxymethylcytosine profiling for the assessment of colorectal cancer biology and treatment response in blood.

Abstract

3585 Background: Colorectal cancer is the third most common cancer worldwide, accounting for about 10% of all cancer cases, and is expected to claim more than 50,000 lives in 2025. Approximately 33% of CRC patients will develop metastases throughout their cancer continuum, and their 5-year survival rate is about 15%. The majority of patients with metastatic colorectal cancer (mCRC) cannot be cured. However, a subset of mCRC patients with localized recurrence or isolated metastases in the liver and/or lungs may achieve a cure through surgical intervention. Yet, current methods for identifying patients who are candidates for more favorable responseremain inadequate. Therefore, there is a critical need for predictive biomarkers to accurately identify patients who are likely to experience better outcomes following surgery. 5-hydroxymethylcytosine (5hmC) is an epigenetic modification that is associated with active genes and regulatory regions that are cell type- and disease-specific. Here, we developed a model using cell-free DNA (cfDNA) 5hmC profiles to detect CRC and identified pathways distinguishing mCRC patient outcomes following treatment. Methods: Plasma was collected from 294 CRC patients and 588 non-cancer individuals to obtain cfDNA. cfDNA was enriched for 5hmC-containing DNA fragments. Input and 5hmC-enriched cfDNA were subsequently used to generate sequencing libraries to obtain WGS and 5hmC profiles, respectively. Machine learning operating on 5hmC and WGS data was used to develop a CRC detection model which was subsequently tested on an independent set of mCRC (n = 69) and non-cancer samples (n = 70). Differential 5hmC analysis was performed using edgeR and Gene Set Enrichment Analysis (GSEA). Results: The performance of the CRC prediction model was evaluated through 10-fold cross-validation producing an auROC curve of 0.86. An independent validation set of mCRC and non-cancer patients displayed an auROC of 0.94. Comparative GSEA using gene body 5hmC levels revealed biological pathways associated with CRC biology such as Myc signaling. cfDNA 5hmC profiling of plasma obtained from mCRC patients before surgery revealed quantitative differences in patients who show recurrence of disease within 2 years post-surgery from the patients who remain recurrence-free for at least 2 years after surgery. These differences between relapsed and non-relapsed groups included 5hmC changes over genes involved in pathways known in mCRC, such as the Wnt/β-catenin signaling (p < 0.05). Lastly, quantitative changes in 5hmC profiles measured in pre-surgery plasma samples enabled prediction of disease recurrence in patients within 2 years post-surgery. Conclusions: 5hmC analysis of cell free DNA) offers a novel, non-invasive approach for identification of colorectal cancer biology and assessment of treatment response in blood samples.