A novel DNA methylation panel accurately detects colorectal cancer independently of molecular pathway

Colorectal cancer (CRC) is the third most incident and the fourth leading cause of cancer-related death by cancer, worldwide [1]. The primary molecular pathways involved in CRC carcinogenesis are chromosomal instability (CIN), microsatellite instability (MSI) and CpG island methylator phenotype (CIMP), accounting for nearly 85, 15 and 10–40% of all sporadic cases, respectively [2, 3]. CIN and MSI represent two levels of genetic instability, a subtle one affecting only DNA sequences, MSI-high (MSI-H), and a gross one, affecting portions or entire chromosomes, i.e., CIN. These forms of instability are considered mutually exclusive: a CRC with CIN is most likely microsatellite stable (MSS) [4, 5]. In sporadic CRC, the most common cause of DNA mismatch repair (MMR) defects, leading to MSI-H, is aberrant bi-allelic MLH1 aberrant promoter methylation. MLH1 is also one of the markers often used to define CIMP, and, thus, an overlap between CIMP and MSI-H exists [2]. Two different panels have frequently been used to define CIMP, the classic panel proposed by An et al. (MLH1, CDKN2A (p16), MINT1, MINT2, and MINT31) and the panel designed by Weisenberger et al. (CACNA1G, IGF2, NEUROG1, RUNX3, and SOCS1) [6, 7]. Another categorization subdivides tumors into CIMP-High (CIMP-H), CIMP-Low (CIMP-L) and CIMP-negative (CIMP-0), each of them associated with different features [8]. Thus, different panels and/or a number of markers have been tested, and no consensus has been reached.

Several non-CIMP related methylation DNA targets have been found to distinguish malignant from non-malignant colorectal tissues [9]. Indeed, several studies have been conducted to identify a methylation biomarker or a panel of biomarkers with high sensitivity and specificity to be used in diagnosis and prognostication of CRC, but none has been validated [7, 10‐21]. Moreover, only a few DNA methylation biomarkers intended for CRC detection are commercially available, including ColoVantage^®, EpiproColon^® 2.0 and Abbott RealTime mS9, which are blood-based tests based on septin 9 (SEPT9) promoter methylation [12, 18]. Nonetheless, the value of SEPT9 promoter methylation as a biomarker has been questioned by several authors [22]. Because currently available screening methods are either costly or of limited specificity, impairing global implementation, DNA methylation-based tests are likely to be more cost-effective. Owing to the suboptimal performance of commercially available epigenetic tests, validation of better candidate biomarkers, which may detect CRC irrespective of molecular subtype, is warranted.

Based on an exhaustive literature review to select potentially useful gene promoters with the ability to discriminate malignant from non-malignant colorectal tissues, enabling its future testing in blood samples, we selected four genes hypermethylated in CRC [Adenomatous polyposis coli (APC) [13, 14], O-6-methylguanine-DNA methyltransferase (MGMT) [15, 23], Ras association domain family 1—isoform A (RASSF1A) [16] and Septin 9 (SEPT9) [12, 17, 18] ] and one gene hypomethylated in CRC [Insulin-like growth factor 2 (IGF2) [19‐21]], for validation in a large cohort of CRCs, in which CIMP and MSI status were also determined. Furthermore, the potential prognostic value of gene promoter methylation was also assessed.

Colorectal cancer is one of the most common and lethal malignancies, especially in developed countries [1]. Currently, CRC screening options mostly rely on fecal occult blood testing and endoscopy. Nevertheless, these strategies meet with significant limitations (e.g., low accuracy and higher cost, respectively) which impair its broader implementation [27]. Some DNA methylation-based biomarkers were developed and have been approved by FDA (Food and Drug Administration), including ColoVantage^®, Epi proColon^®, and ColoSure^®, but are still not fully implemented in clinical practice. Whereas the former two are based on SEPT9 methylation [12, 18], the latter is based on Vimentin methylation [28]. However, SEPT9 promoter methylation performance in CRC detection has been recently questioned [22]. As for Vimentin promoter methylation, this test is used in combination with colonoscopy, but sensitivity is rather variable, ranging from 38 to 88% [28, 29]. Thus, alternative screening test is demanded to increase population adherence and perfect detection accuracy.

Globally, the distribution of promoter methylation levels for the five gene promoters tested among CRC and CRN tissues paralleled those of previous reports. Indeed, IGF2 methylation levels were significantly lower in CRC than in CRN, in accordance a previous study [19], whereas the remainder genes showed significantly higher methylation levels in CRC, as reported by others [14, 15, 17, 30, 31]. Interestingly, SEPT9 and MGMT methylation levels were significantly lower in rectal cancer patients that underwent neoadjuvant treatment, which is in line with studies in which ionizing radiation exposure induced global hypomethylation [32, 33], including for colon cancer [34].

Regarding the use of gene promoter methylation as CRC biomarker, the best detection performance was disclosed by SEPT9 promoter methylation, followed by MGMT. Remarkably, SEPT9 validity estimates were similar to those of the trademark assays using SEPT9 methylation [12, 35], further confirming its biomarker potential. Nonetheless, MGMT displayed better performance (77.2% sensitivity, 84.0% specificity) than previously reported (46–53% sensitivity, 74–100% specificity), especially concerning sensitivity [23, 36]. This might be due not only to differences in the population under study, but also differences in the methodological approaches. The same might apply to RASSF1A performance, which disclosed lower sensitivity (33.1% vs. 81.0%) but higher specificity (90.0% vs. 51.0%) [16], although the definition of the cutoff value might also significantly impact on performance. Nonetheless, a combination of these three gene promoters significantly improved diagnostic performance when compared with previously published panels (Table 5). In our experience, the use of gene promoter panels, usually comprising two to four genes, improves sensitivity, without compromising specificity, as previously demonstrated for prostate and breast cancers, especially when tested in liquid biopsies [37, 38]. Thus, the use of this gene panel might augment the performance of approved DNA methylation-based assays.

A potential downside of the use of DNA promoter methylation-based biomarkers for CRC detection is its potential association with a specific molecular subtype, namely MSI-H, which is due to defects in MMR pathway [3]. This system includes genes like MLH1, MSH2, MSH6, and PMS2 whose expression was screened by immunohistochemistry. Therefore, the loss of expression of any of these genes is considered indicative of defective MMR and, consequently, of MSI-H, whereas the remaining most likely represent MSI-L or MSS tumors. Using this strategy, only 3.7% of cases were disclosed as MSI-H, which is lower than reported in other series, especially those using direct MSI analysis [2, 3]. It is noteworthy, however, that in our series, almost 50% of tumors were localized in the rectum, whereas in most series rectal carcinoma represents less than 30% of CRC. Because rectal carcinoma usually displays a lower frequency of MSI-H cases, as verified in our series, the “over-representation” of tumors localized in the rectum might have decisively influenced the overall frequency. Curiously, CIMP cases were more frequent in the rectum (11.5% vs. 5.6%), with an overall frequency of 8.5%, which is close to the lower end of published series [2, 3]. Owing to the lack of agreement on the best strategy to define a CRC case as CIMP, it is not possible, however, to perform direct comparisons. Nonetheless, it should be emphasized that excepting for APC, no significant differences in gene promoter methylation levels were found among MSI-H, CIMP+ or CIN CRC molecular subtypes, nor among different pathological stages or tumor location, in our series. Thus, it may be concluded that the performance of the three-gene promoter panel is likely to be homogeneous across molecular subtypes, and the same applies for primary tumor localization and pathological stage.

We also tested the prognostic value of the five genes promoter methylation status, as this might convey valuable information for clinical decision making. Remarkably, higher SEPT9 promoter methylation was independently associated with increased DSS in CRC, whereas no prognostic value was depicted for the remainder gene promoters. A meta-analysis that included 14 studies showed that MGMT methylation status was not associated with CRC prognosis [47], whereas RASSF1A promoter methylation has been associated with poor prognosis, although when assessed in plasma samples [30]. Furthermore, CIMP status also did not disclose prognostic value in our study, which is in accordance with a recent meta-analysis that found no significant effect of CIMP status in DFS or OS in 8 out of 11 and in 13 out of 19 studies previously published, respectively [48]. The same was reported regarding DSS, in 3 of 4 studies considering the classical CIMP panel [8]. Thus, globally, our results are by most of the published literature on the subject.