Marc Bissonnette

M.D.

University of Chicago

Associate Professor

PLCO (Learn more about this study)

2016-0057

Jan 24, 2018

Development of a sensitive and specific blood test for colon cancer by measuring base-specific hydroxymethylcytosine (5hmC) and 5-methylcytosine (5mC) in circulating cell-free DNA

Colorectal cancer (CRC) is a major cause of cancer-deaths, with poor prognosis when diagnosed late. Screening colonoscopy is the standard of care for detection and prevention, but compliance remains below 50%. Increasing efforts have focused on blood tests to diagnose cancers by detecting biomarkers in cell-free DNA (cfDNA), which could significantly improve screening compliance. In contrast to the infrequency of specific mutations in cancers, DNA epigenetic markers, 5-hydroxymethylcytosine (5hmC) and 5-methylcytosine (5mC) could serve as ideal biomarkers, given their high prevalence in the human genome, strong correlation with gene expression, and robust chemical stability. Prior technologies, however, required large amounts of DNA. Taking advantage of a highly sensitive and selective chemical labeling technology (nano-hmC-Seal), developed by the co-PI Dr. Chuan He, we have captured and profiled base-specific 5hmC and 5mC in cfDNA and paired tumor tissues collected from 47 healthy individuals and 130 patients with recently diagnosed CRC or other cancers. Consistent cancer-specific epigenetic signatures were identified in tumor tissues and cfDNA in all cases. Because of the high sensitivity (~1 ng DNA), robustness, clinical convenience and cost-efficiency, this novel technology could provide powerful biomarkers for CRC diagnosis and prognosis, and serve to monitor treatment response, predict relapse and suggest genes causal in CRC invasion. To achieve these goals we propose 3 aims:
In aim 1a, we will perform 5hmC profiling by tagging 5hmC, using nano-hmC-Seal and localized by NGS, and similarly profile 5mC after converting to 5hmC. Based on our preliminary success in our patient cohort, we will develop algorithms providing a highly specific and sensitive blood test to detect CRC using cfDNA from the PLCO biorepository taken from patients prior to CRC diagnosis. As a training set, we will study 250 patients (from 1379 CRC patients), and 500 individuals without cancer matched for age and sex. We will employ elastic net and other machine-learning algorithms such as Random Forest (RF) and Support Vector Machine (SVM) to develop discriminators of 5hmC and 5mC modified genes to diagnose CRC. We will stratify analyses by tumor location, stage, differentiation, patient age and gender that modulate 5hmC and 5mC distributions.
In aim 1b, we will study an independent 500 PLCO cfDNA samples from patients with CRC and 1000 cancer-free controls as a validation set. Differential 5hmC and 5mC will be measured. We will stratify analyses as in aim 1a and use rigorous bioinformatics and biostatistical analyses to validate signatures.
In aim 2, we will map 5hmC and 5mC epigenomes in colon cancers and advanced adenomas, matching tumors to cfDNA examined in aim 1.
In aim 3, we will use bioinformatic analyses of 5hmC and 5mC signatures to compare matching tumors and cfDNA. In aim 3a we will perform detailed analyses to uncover biomarkers in tissue and/or cfDNA that could predict metastasis, relapse, and response to treatments. In aim 3b, using pathway analysis tools, we will explore potential mechanisms contributing to differences in cfDNA and tumor signatures, including selective enrichment of genes in cfDNA predicted to drive tumor migration/invasion and metastasis.

In aim 1a, we will perform 5hmC profiling by tagging 5hmC, using nano-hmC-Seal and localized by NGS, and similarly profile 5mC after converting to 5hmC. Based on our preliminary success in our patient cohort, we will develop algorithms providing a highly specific and sensitive blood test to detect CRC using cfDNA from the PLCO biorepository taken from patients prior to CRC diagnosis. As a training set, we will study 250 patients (from 1379 CRC patients), and 500 individuals without cancer matched for age and sex. We will employ elastic net and other machine-learning algorithms such as Random Forest (RF) and Support Vector Machine (SVM) to develop discriminators of 5hmC and 5mC modified genes to diagnose CRC. We will stratify analyses by tumor location, stage, differentiation, patient age and gender that modulate 5hmC and 5mC distributions. In aim 1b, we will study an independent 500 PLCO cfDNA samples from patients with CRC and 1000 cancer-free controls as a validation set. Differential 5hmC and 5mC will be measured. We will stratify analyses as in aim 1a and use rigorous bioinformatics and biostatistical analyses to validate signatures.

In aim 2, we will map 5hmC and 5mC epigenomes in colon cancers and advanced adenomas, matching tumors to cfDNA examined in aim 1.

In aim 3, we will use bioinformatic analyses of 5hmC and 5mC signatures to compare matching tumors and cfDNA. In aim 3a we will perform detailed analyses to uncover biomarkers in tissue and/or cfDNA that could predict metastasis, relapse, and response to treatments. In aim 3b, using pathway analysis tools, we will explore potential mechanisms contributing to differences in cfDNA and tumor signatures, including selective enrichment of genes in cfDNA predicted to drive tumor migration/invasion and metastasis.