The concept of using molecular tests to detect epigenetic methylation changes in circulating cell-free DNA has gained much enthusiasm as a simple and non-invasive method for CRC and adenoma population-based screening. Our results demonstrate the potential for using epigenetic DNA methylation signatures to identify patients with colorectal cancer. We focused on ten highly prospective target genes and the majority of these displayed high differential methylation between CRC tissue and healthy NAT (Fig. 1). Although each of the target genes performed well individually a combined marker panel was observed to have an overall higher sensitivity and specificity. This could be due to the inherent genetic variability among colorectal cancers which means that testing for a single target gene is likely to lead to more false negative results than testing for multiple targets at once.
Importantly, we found that the sensitivity and specificity of the both the target genes and combined gene panels were high in early stage I and II disease. This quality is imperative for any diagnostic test in CRC since the patient outcomes of treatment for early-stage disease are significantly better than late-stage disease. Previous studies have found a variable association between stage of CRC and levels of CpG island methylation. For instance, the same research group has found various levels of association across multiple publications investigating BCAT1 and IKZF1 [8, 15–16]. Whilst most of the research suggests an increasing level of methylation with stage the results are inconsistent. Perhaps the important feature to highlight is that the diagnostic accuracy of most genes seen in this study in stage I and II disease is equal to or above the expected results of the currently used FIT (73% stage I; 80% stage II) [17]. Johnson et al. and Symonds et al. similarly found no significant difference in diagnostic accuracy when they directly compared epigenetic methylation biomarkers to faecal immunochemical tests [15, 18].
There are several factors other than cancer that have been shown to alter CpG island methylation patterns. Smokers has been found to have a significantly altered genome-wide methylation pattern when compared to non-smokers [19]. Furthermore, complex age-related DNA methylation changes have been shown to occur throughout life. In early life, there is methylation gain globally but this is more focussed at the CpG islands and intergenic regions. However, in later life there is overall DNA methylation loss, but the CpG islands continue to gain methylation [20]. In this study the potential association of DNA methylation in the target genes to confounding variables such as age, smoking, metastatic disease, and co-morbidities is important because of the effects this could have on the utility of these genes as biomarkers. For instance, the cut-off values for a positive result may have to be altered based on age or smoking status. Similarly, these markers may not be as accurate in the presence of significant co-morbidities. The Charlson Co-morbidity Index was utilised for analysis however the individual co-morbidities that are associated with higher methylation levels is of more importance in clinical diagnostic tests. Although there is a complex relationship between CpG island methylation and potential confounding factors this does not discount the significant differences seen in this study between CRC tissue and normal colonic tissue. Although there were significant associations found between certain individual genes and the age and CCI, this study is not designed to look at these factors specifically and there is a need for more clarification on their effect on epigenetic based biomarkers.
There has been a limited number of blood-based circulating tumour DNA assays approved for clinical use. The most notable of these are Epi proColon 2.0 which detects methylated SEPT9 and Colvera which detects methylated BCAT1 and IKZF1. Both tests have had large cohort studies performed to assess their efficacy. Epi proColon 2.0 is the most studied marker and exhibits a large variation in the sensitivity (48–95%) and specificity (80–99%) between studies [21–22]. However, this range is in part due to the variability with which the results are analysed. Colvera was found to have a lower variation in sensitivity (62–77%) and specificity (89–94%) when compared to the Epi proColon 2.0 test [8, 15, 23–24]. Additionally, in a direct comparison to FIT the Colvera test was found to have a comparable sensitivity with slightly better specificity. However, the sensitivity for the detection of advanced adenomas was significantly higher for FIT [15]. For these reasons the Colvera test is currently only used for monitoring for disease recurrence rather than primary diagnosis or CRC screening. Despite the approval for use of these tests in the clinical setting their role has been limited due to their high cost, limited potential benefit when compared to currently used methods of detection and poor ability to detect pre-cancerous polyps. In fact, a cost-effectiveness analysis of SEPT9 methylation concluded that FIT is less costly and more effective [25].
There are several limitations in this study. Even though it was small, the sample size was adequate as a pilot study to provide preliminary data on the methodology of detection and overall statistical efficacy. The samples used here are from tissue only and although there is sufficient evidence of plasma ctDNA detection in CRC among other studies, the ability of the biomarker panels from this study to be translated into a liquid biopsy platform remains unknown at this time. Furthermore, there were no pre-cancerous adenoma tissues used in this study and therefore we cannot predict how these markers may be altered in these lesions.