Abstract
Background
Patients with ulcerative colitis (UC) are enrolled into surveillance programs for the early detection of colorectal cancer (CRC). However, most patients under surveillance are low-risk and never progress to CRC, while a significant proportion of CRCs in UC form without a preceding confirmed diagnosis of dysplasia. High resolution chromosomal copy-number alteration (CNA) analysis of unselected formalin-fixed paraffin embedded biopsies taken at surveillance colonoscopies using low pass whole genome sequencing (lpWGS) offers an appealing approach to CRC stratification.
Methods
We conducted a retrospective case-control study to compare the CNA burden in four unselected non-neoplastic left-sided colorectal biopsies from patients with E2/E3 UC derived 1–5 years prior to HGD/CRC detection (cases), with that of biopsies from patients who subsequently remained HGD/CRC-free for at least 5 years (controls). The two patient groups were matched by age, gender, duration of IBD and PSC status. lpWGS was performed using a standardised pipeline for epithelial enrichment, DNA extraction, library preparation, next generation sequencing and bioinformatic analysis.
Results
476 biopsies, derived from 42 cases and 77 controls, were analysed. Nearly 80% of patients had a detectable CNA in at least one of their biopsies, with the maximal CNA burden in a typical biopsy involving a median 1.1% of that biopsy’s genome. The CNA burden was significantly greater in the rectum compared to the sigmoid colon and descending colon. The most common CNA events were losses of between 1–30 megabases involving the sub-telomeric regions of chromosomes 5–9 and 22, which were found in similar proportion in both case and control biopsies. However, losses extending beyond sub-telomeric regions, as well as copy number gains, were found more frequently in cases biopsies (p<0.0001). The most discriminating CNA event was the presence of such a loss extending beyond subtelomeric regions in any of the patient’s four biopsies, with a high specificity exceeding 0.95 (see Kaplan-Meier plot). ROC analysis demonstrates that lpWGS output has a fair level of accuracy at predicting future HGD/CRC risk (AUC 0.73).
Conclusion
We identified multiple biopsies, predominantly in cases, with a surprisingly marked CNA burden involving over 10% of the genome, highlighting the fluid phenotype-genotype relationship. Non-dysplastic colitic epithelium can bear a significant burden of CNAs and maintain phenotypic stability for years without neoplastic transformation. Remarkably, by analysing the CNA burden of only four random biopsies, derived from less than 0.05% of the colonic surface area, we can significantly discriminate between case and control cohorts.
Whole-genome sequencing combined with a case-control study of an outbreak of staphylococcal food-poisoning
