SEPT9 and SHOX2 DNA methylation status and its utility in the diagnosis of colonic adenomas and colorectal adenocarcinomas

Background Colorectal cancer (CRC) appear to arise from precursor lesions in a well-characterized adenoma-carcinoma sequence. Significant efforts have been invested to develop biomarkers that identify early adenocarcinomas and adenomas with high-grade dysplasia, since these are believed to harbor a particularly high risk for malignant transition and thus require resection. Promoter methylation of SEPT9 and SHOX2 has been suggested as a biomarker for various solid malignant tumors. Hence, the present study aimed to test their biomarker potential in CRC and precursor lesions. Results Assessment of promoter methylation of SEPT9 distinguished adenomas and CRC from controls as well as advanced from non-advanced adenomas (all p < 0.001). Correspondingly, SHOX2 methylation levels in adenomas and colorectal carcinomas were significantly higher compared to those in normal control tissues (p < 0.001). Histologic transition from adenomas to CRC was paralleled by amplification of the SEPT9 gene locus. Conclusions SEPT9/SHOX2 methylation assays may help to distinguish colorectal cancer and adenomas from normal and inflammatory colonic tissue, as well as advanced from non-advanced adenomas. Further studies need to validate these findings before introduction in clinical routine.


Background
Colorectal cancer (CRC) is one of the most common and intensively studied cancer entities worldwide [1]. Although its molecular pattern is to a certain extent heterogeneous, more than 80 % of sporadic CRC appear to arise from precursor lesions [2]. The "adenoma-carcinoma sequence" [3] reflects this transition with a genetic characteristic, e.g., mutations in TP53, KRAS, and APC, [4] and histopathologic alterations [5,6]. The evolving knowledge about precancerous lesions of CRC, the oftentimes slow progression towards malignant transformation, and a tremendously better prognosis for early detected and treated CRCs make this cancer entity particularly attractive for screening strategies [7]. Owing to multimodal therapy regimes, improved surgical techniques, and screening programs, incidence of CRC has steadily decreased over the last decades [8,9]. Despite this significant success, population's enrollment in recommended screening programs is difficult [10] and significant efforts have been invested in the development of non-invasive diagnostic tests. Nonetheless, colonoscopy remains the non-replaceable gold standard in every CRC screening program [11,12].
Epigenetic changes to genomic tumor DNA are biostable and often cancer-specific alterations that are therefore issue of numerous ongoing research projects worldwide [13]. Our and other groups established differential methylation-specific qPCR assays of the stature homeobox 2 (SHOX2) and septin 9 (SEPT9) in various cancer entities with possible biomarker properties for early detection and response prediction strategies [14]. For example, our team previously used a combined assay to measure SHOX2 and SEPT9 promoter methylation for the discrimination between benign and malignant pleural effusions [15]. Promoter methylation of SEPT9, a gene encoding for a GTP-binding protein with various functions in formation and control of the cytoskeleton, proved to be present in >90 % of CRC specimens (mean methylation level 26 %, range 0-89 %) [16][17][18]. Furthermore, SEPT9 DNA methylation in blood plasma was successfully validated in a large prospective trial including 8000 asymptomatic subjects, undergoing routine colonoscopy (ClinicalTrials.gov Identifier: NCT00855348) [19]. Consequently, SEPT9-based diagnostic tests for colorectal cancer screening are available to patients in Europe and the USA as CE-marked In Vitro Diagnostics (CE-IVDs) and Laboratory Developed Tests (LDTs). Very recently, the American Food and Drug Administration (FDA) approved the commercial SEPT9 methylation assay, "Epi proColon®," as a blood-based in vitro diagnostic (IVD) test for screening of CRC. However, the SEPT9 methylation pattern of adenomas has only been studied in a small amount of studies [20][21][22][23][24]. It is known today that certain types of adenomas harbor an increased risk for malignant transformation [25] and that adenomatous subtypes carry different epigenetic profiles [26]. Up-to-date detailed information about the SEPT9 methylation status in different types of adenomatous polyps is still fragmentary.
SHOX2 harbors two large CpG islands, located at the 3′ and the 5′ end of the gene, and is involved in limb formation and cardiac development [27,28]. So far, the SHOX2 DNA methylation status has mainly been evaluated for its value in the detection of lung cancer with reasonable success [29][30][31][32][33][34]. More than 90 % of histologically confirmed lung cancer patients showed a hypermethylation of this gene locus in comparison to normal tissue [35]. Additionally, the SHOX2 methylation status showed promising results in other cancer types [15,36]. In order to probe the apparently broad utility of SHOX2 as a biomarker in CRC, this target was included in the present study.

SEPT9 and SHOX2 methylation status in advanced adenomas
For three patients with colorectal adenoma, no information regarding size and grade of dysplasia could be retrieved, narrowing the amount of individuals in this analysis to 72.
In correlation to the stepwise increase of SEPT9 and SHOX2 methylation level, the adenoma group was evaluated for other characteristics with known potential for supporting a malignant transition.

SEPT9 and SHOX2 methylation levels in adenoma subtypes
Regarding the methylation status of the SEPT9 gene locus, (tubulo-)villous adenomas showed the highest methylation levels of all adenomas, irrespective of assay type. Consequently, villous adenomas showed a significant hypermethylation of SEPT9 (Triplex assay) in comparison to serrated adenomas (164.9 vs. 70.3 %, p = 0.022, Fig. 4). All other types of adenomas showed no difference in methylation using the Triplex assay. According to results from the Triplex assay, QM assay Comparison of different SHOX2 methylation levels obtained from the a Triplex assay (control n = 62, adenomas n = 75, CRC n = 25) and the b QM assay (control n = 63, adenomas n = 75, CRC n = 28) for the three subgroups. **p < 0.001 marks significant differences in methylation levels between the indicated subgroups, and *p < 0.05 marks significant differences in methylation levels between the indicated subgroups analysis also showed a  (Fig. 4).

Discussion
Large screening programs have decreased the incidence of CRC [8,9], but low compliance rates especially regarding preventive colonoscopy hamper a maximum success of modern primary prevention strategies [11]. The principle of "tailored screening" may improve screening effectiveness by stratification into risk tiers [38]. Therefore, entirely non-invasive early detection tests for CRC are urgently needed.
Epigenetic alternations of the SEPT9 gene locus were previously described in minor series of tissue from colorectal adenomas [20][21][22][23] or blood samples [24,39]. The presented comprehensive data confirmed previous findings and highlight the relevance of SEPT9 methylation in these lesions. Moreover, it was revealed that the assessment of promoter methylation of SEPT9 as well as SHOX2 may actually help distinguish CRC and adenomas from normal epithelium without dysplasia. These results are in line with Tänzer et al. who found a higher  . No significance level p < 0.05 has been reached in these assays frequency of SEPT9 and ALX4 DNA methylations in plasma of AA compared to healthy controls [23]. On top of that, the methylation status of SEPT9 as quantified with our Triplex assay was able to distinguish between adenomas vs. CRC and AA vs. N-AA. Despite their statistical significance, the methylation level of both genes within the subgroup showed a scattered and partly overlapping distribution, elevating the risk of false positive or negative results. These observations are in line with the finding that epigenetic changes resemble an early event in the carcinogenesis of CRC [40,41]. In accordance, genome-wide comprehensive methylation analysis of adenomas and CRC tissue revealed accumulation of epigenetic alternations in the progression from early adenomas towards invasive adenocarcinomas [42,43]. Although, advanced adenomas are believed to harbor an increased potential for malignant transformation (see [44] and [25]), a direct proof of malignant transformation of advanced adenomas or of a certain histologic subtype remains somewhat elusive.
Interestingly, a stepwise increase of the SEPT9 methylation level from controls over not-advanced adenomas to advanced adenomas and invasive adenocarcinoma could be observed. This may explain false positive test results of the Epi proColon® test indicating methylated DNA segments in the SEPT9 region in the absence of an invasive CRC. SEPT9 may therefore potentially be used as an ancillary marker in the identification of advanced adenomas in case of difficulties with histologic diagnosis of a colorectal biopsy (i.e., due to cautery artifacts). Additionally, the presented data support a future inclusion of SEPT9 and SHOX2 in methylation biomarker panels to support the diagnosis of CRC, though both markers have to be validated in a large prospective trial. By combining these and other markers, overlapping results especially between AA and CRC may be minimized.
Surprisingly, >50 % of SEPT9 Triplex methylation levels in CRC and adenomas showed a value exceeding 100 %, technically reasoned by a distant location of the reference gene (ACTB at chromosome 7) and the analyzed gene loci (SEPT9 gene at chromosome 17 and SHOX2 at chromosome 3q) [15]. Hence, higher methylation values in comparison to QM assay may be due to focal amplification of the SEPT9 gene locus and/or a deletion of the reference gene (ACTB). A SEPT9 amplification later during carcinogenesis is supported by the observation that especially advanced adenomas and carcinomas showed a level >100 % in the Triplex assay. These findings are in congruence with the results from Ben-David et al. who found that genes, which have been up-regulated in early adenomas, showed a tendency for amplification in later stages of colorectal carcinogenesis [45]. Taken together, one may carefully speculate that promoter hypermethylation of SEPT9 triggers a focal amplification that may foster a malignant transition from adenoma to invasive carcinoma. This mirrors the assumption of Saha et al. and Bardelli et al. that gene amplification is attributed to play a role in stage transition of CRC [46,47]. Concordantly, SEPT9 amplification has only been identified in advanced breast cancer patients [48] and in vitro [49], while early localized tumors showed no SEPT9 amplification. Additionally, the same group was able to identify a worse survival in endometrial adenocarcinoma, which shows a copy number variation (CNV) of SEPT9 [50]. In contrast to this hypothesis, SEPT9 amplification was only found in 0.4 % of CRC in a cohort of 257 CRC patients and in 2/615 CRC cases of the TCGA data repository (0.3 %) [4]. If the above assumptions are proven true, SEPT9 may be an excellent plasmatic biomarker. Because, due to DNA amplification, the total number of tumor DNA alleles in plasma including their specific epigenetic SEPT9 alteration may be easier to detect among background DNA. This phenomenon has already been evidenced for SHOX2 in squamous cell lung cancer [34]. Another possible explanation for this phenomenon was hypothesized to be a MSI, which is known for causing hyper-mutated and epigenetically altered genomes [4]. However, no correlation between MSI and amplifications of SEPT9 and SHOX2 or clinical parameters were found in the present study.
Overall, SEPT9 methylation showed a better performance than SHOX2 in distinguishing the different study groups. This finding is in line with results about the potential of plasmatic free-circulating tumor DNA and tissue SEPT9 methylation [17][18][19][51][52][53], although sensitivity (52-94 %) and specificity (88-95 %) vary greatly between the studies [17,18,51,54]. This variability may be attributed to different testing methods or common pre-existing conditions like high age, gender, and comorbidities and have to be interpreted with caution [53,55].
The present study has several limitations. For example, only a small series was included and analyzed in a retrospective fashion hampering statistical power and lowering the evidence level. Furthermore, methylation assays have only a limited suitability to detect CNV, and more sensitive assays (e.g., FISH, CGH, or next-generation sequencing) have to validate the data.

Conclusions
In conclusion, SEPT9 or SHOX2 methylation may be auxiliary biomarkers for the differentiation of CRC and advanced adenomas to non-advanced adenomas and normal tissue. Ideally, both markers should be integrated into a marker panel for CRC screening and validated in a large prospective trial. The investigations suggest that CNV of SEPT9 may contribute to a malignant transition from adenomas into advanced adenomas and adenocarcinomas. Further studies with CNV sensible assays are needed to elucidate the distinct role of these gene loci in the carcinogenesis of CRC and their potential as biomarker.

Patient samples
The study was approved by the Institutional Review Board (IRB) of the University Hospital of Bonn (Number: 222/13). Formalin-fixed, paraffin-embedded (FFPE) tissue specimens from 166 patients, treated or diagnosed at the University Hospital of Bonn academic hospitals between 2002 and 2013, were included. The study group included various non-invasive epithelial lesions including tubular adenomas, tubulovillous adenomas, villous adenomas, and serrated adenomas. Furthermore, sporadic colorectal cancer were included. Patients with a history of familial adenomatous polyposis (FAP), Lynch syndrome, or other second primary malignancy were not included. Additionally, patients with Crohn's disease showing high-grade dysplasia in the resected specimen or in any biopsy taken at time of admission were excluded from the study. All specimens were histologically diagnosed by an experienced pathologist blinded to the patient's history. Histologic classification was performed according to the most recent recommendations by the World Health Organization [56] and the latest TNM classification [57]. Advanced adenomas were defined as adenomatous polyps with a size ≥10 mm, ≥25 % of villous features, or a high-grade dysplasia [58][59][60][61].

SHOX2/SEPT9/ACTB Triplex assay
As previously described, a methylation-specific Triplex qPCR assay was used [15]. It quantifies the number of methylated SHOX2 and SEPT9 alleles, referred to total DNA copy numbers. The total DNA copy number was quantified using a qPCR assay targeting the β-actin (ACTB) gene locus comprising no CpG sites [15]. A calibrator sample (bisulfite-converted artificially methylated DNA) was used in order to allow for an accurate quantification as previously described [15].

Calculation of methylation levels
The calibrator and all the samples were analyzed in triplicate (Triplex assay) and in duplicate (QM assay), respectively. The mean average of the CT values was calculated and used for further analysis. Previously described adapted ΔΔCT methods were applied to determine a relative methylation value from the QM assay [68] and the Triplex assay [15] measurements. Invalid PCR results indicated by high cycle threshold (CT) values (CT ACTB > 33 (Triplex assay); CT methylated allele > 33 and CT unmethylated allele > 33 (QM assay)) were omitted from the analysis.

Immunohistochemical staining for MSI
Tumors with MSI are known for their hyper-mutated genome and epigenetic alternations [4]. To analyze SHOX2/SEPT9 in association with MSI, immunohistochemical staining (IHC) was used to estimate the protein expression levels of the four major mismatch repair enzymes (MLH1, PMS2, MSH2, and MSH6). For this purpose, tissue microarrays (TMA) were assembled from formalin-fixed, paraffin-embedded tissue. For each case (adenoma, carcinoma, control group), representative areas were marked on the hematoxylin-eosin-stained section. Subsequently, at least 1-mm cores were punched and arrayed in a paraffin block. IHC stainings were performed with a Ventana Benchmark-automated staining system (Ventana Medical Systems, Tucson, AZ, USA), following the manufactures' protocol. The following antibodies from BD Pharmingen were used: MLH-1 #550838 clon G168-15 (1:50), MSH-2 #556349 mouse anti-human, MSH6 BD Pharmingen #610918 mouse anti-human, and PMS2 BD Pharmingen #556415 mouse monoclonal. The slides were then counterstained with hematoxylin, dehydrated, and mounted. Presence of mismatch repair enzymes was then evaluated by an experienced pathologist.

Availability of data and materials
The TCGA dataset supporting the conclusions of this article is available in the TCGA data repository (http:// cancergenome.nih.gov/). The dataset supporting the conclusions of this article is included within the article and its figures.

Data evaluation and statistical analysis
Values are expressed as mean or median, unless otherwise stated. Differences between groups were tested using t test, Mann-Whitney U test, or ANOVA, where appropriate. Correlation between values was tested using Spearman's rank correlation. The Bonferroni correction was used in case of multiple pairwise comparisons. Receiver operating characteristic (ROC) curves were calculated to observe the ability of the methylation level of SHOX2 and SEPT9 gene loci to differentiate between the subgroups. Areas under the curve (AUCs) were reported. p values of <0.05 were considered to be statistically significant. Statistical analyses were performed with SPSS (IBM, Armonk, NY, USA, version 23) and Graph-Pad Prism 4 (GraphPad Software Inc., San Diego, CA, USA).