Multi-site Gastrointestinal Cancer Detection by Stool DNA Methylation

Gastrointestinal (GI) malignancies lead cancer deaths worldwide, killing ~3 million annually. In the U.S., only colorectal cancers (CRC) are screened. Other GI cancers are not screened due to lack of accurate tests or because prevalence is deemed too low for cost-effective screening. Consequently, most patients with GI cancers present at late-stage and cure rates remain abysmally low. Effective early detection is desperately needed to improve outcomes. Our group was central in development and validation of the FDA-approved multi- target stool DNA test for CRC screening. We have begun to expand this approach, showing feasibility to detect supra-colonic GI cancers by stool DNA testing. However, it is critical for a non-invasive molecular test to localize the site of a primary cancer (“site-prediction”). Key preliminary data suggest that this may now be possible. First, we have completed rigorous next-generation sequencing to identify differentially methylated regions (DMRs) which appear highly discriminant for universal and site-specific detection of GI cancers. Second, we prioritized these DMRs by strict filtering criteria and performed a confirmatory study with statistical cross-validation on an independent set of CRC, gastroesophageal and pancreatico-biliary cancer, and normal control tissues. This showed that a panel of 8 selected DMRs could distinguish cancer from normal (95% accuracy) and assign organ site (94-95% accuracy for each category) with overall site-prediction accuracy of 88%. Third, these findings have been confirmed with novel DMRs assayed from stool specimens obtained from CRC and pancreatic cancer patients and normal controls (30, each). Using a 2-stage analysis, cancers were distinguished from controls at 90% specificity in the first stage. At stage 2, the markers accurately classified CRCs from pancreatic cancers with 90% accuracy. It is now our central hypothesis that luminal and ductal adenocarcinomas can be detected and localized by stool assay of universal and site-specific DMRs. This raises 3 key questions: 1) will stool assay of our novel DMRs show the high overall cancer sensitivity and the site-prediction we have seen in preliminary data; 2) will the DMRs be specific for cancer across a wide patient demographic spectrum and in the setting of non-malignant GI diseases; and 3) can sensitivity and specificity be improved by novel assay technology? These will be addressed in the following parallel, integrated, but independent specific aims: 1) Assess panel sensitivity and site-prediction accuracy in stool specimens for adenocarcinoma at esophageal, pancreatic and colorectal sites; 2) Confirm and evaluate DMR specificity in stool; and 3) Optimize novel assay conditions and marker selection for cancer detection and site-prediction at esophageal, pancreatic and colorectal sites. With our team’s strong track record, extensive stool archive, and unique access to a state-of-the art assay platform, we expect to demonstrate in a cost-efficient manner the feasibility of a novel DMR panel for the detection and site prediction of specific GI adenocarcinomas. Results will inform designs of future studies ranging from large- scale case-control studies (phase 2) on early-stage cancer and pre-cursors to pivotal cohort validation (phase 4) of a non-invasive multi-GI cancer screening test. The potential impacts on cancer control are far-reaching.