3.1 Clinicopathological characteristics of all 49 sample

According to the data format, the clinicopathological characteristics of all the samples are listed in Table 1.

Table 1 Baseline characteristics of 49 patients in the study3.2 Overall average and site-specific DNA methylation levels of CDH13 exon 1 in CRCs.

Using the BSAS method, we quantified the DNA methylation status of 13 individual CpG sites located in the CDH13 exon 1. The aim of this analysis was to define the methylation levels of the CRC patient’s tumor and adjacent normal tissue and to assess whether any of the tested sites could be considered as a potential tumor-tissue marker. As shown in Fig. 2B, methylation level in all samples is presented, and percent methylation at 13 CpG sites, as well as overall average percent methylation, are significantly higher in CRCs than in comparable normal tissues, as shown in the following Table 2 and Fig. 2C.

Table 2 Methylation level in all samples

ROC curve analysis was performed to evaluate the diagnostic value of differentially methylated CpG sites. As shown in Fig. 3, 13 CpG sites showed a moderate AUC of the ROC curves, ranging from 0.729 to 0.776. Since an AUC of ≥ 0.75 generally indicates a marker that could potentially have clinical utility [23], we combined these 13 CpG sites to improve the diagnostic power. The result indicated that a combination of these indicators increased sensitivity and specificity (AUC = 0.844; p < 0.0001). Age-dependent DNA methylation has been recognized as an important physiological process [24, 25]; In this study, we included age in the diagnostic model, and the result revealed no significant difference in diagnostic performance compared with the combined diagnosis of the 13 CpG sites. This suggests that the DNA methylation status of specific CpG sites in CDH13 are not promising diagnostic biomarker candidates for the differentiation of CRC tissues from surrounding non-malignant tissue.

Fig. 3

ROC curve analysis was performed to evaluate the diagnostic value of differentially methylated CpG sites

3.3 Association between DNA methylation level and clinicopathological characteristics in CRCs.

Although the average DNA methylation level in adjacent normal tissue was low in exon 1, it could be far beyond the average in some cases. Even in a few samples, the methylation levels in non-malignant tissues were higher than the cutoff values derived from the ROC curve analysis. This non-negligible methylation pattern suggests that a criterion must first be set up to define hypermethylation at CpG sites in order to analyze the association of CDH13 hypermethylation with clinicopathological characteristics. We set a threshold value of 25% DNA methylation difference between tumor and non-malignant to ensure and only aberrant hypermethylation cases were assigned as positive. Next, the relationship between the methylation status of exon 1 and clinicopathological features was assessed by univariate logistic analysis. The findings are shown in Table 3. A positive correlation was observed between hypermethylation at CpG site 1 and the presence of distant metastasis, which was statistically significant (OR: 7.60 [95% CI 1.19–48.44], p = 0.032). Considering that both site 1 and site 5 indicate poor prognosis, which is shown in Fig. 4M and N, we combined them to integrally analyze their relationship between the DNA methylation status and clinicopathological features. A significant correlation between the co-hypermethylation of sites 1 and 5, and distant metastasis was observed (OR: 9.75 [95% CI 1.46 − 65.36]; p = 0.019) in Table 3.

Table 3 Univariate associations of site-specific and overall methylation status of exon 1 in colorectal cancer with clinicopathological characteristics (odds ratio and 95% CI)Fig. 4

Kaplan–Meier survival curves for overall survival according to the threshold value for prognostic DNA methylation sites. N non-hypermethylation, H hypermethylation

3.4 Effect of CDH13 hypermethylation on poor survival of CRC patients

The last follow-up date for this study was April 19, 2020, with an average follow-up time of 116 months. During the follow-up period, survival data of 49 patients were obtained, of whom 27 (55.1%) died. The time from patient’s surgery to death for various reasons or the last follow-up visit was defined as the overall survival (OS) time. With the cutoff values described above, we used Kaplan–Meier survival curves and a log-rank test to determine the effect of CDH13 hypermethylation on the survival of CRC patients. In accordance with the findings shown in Fig. 4, the hypermethylation versus non-hypermethylation status of site 1 (OS, 52.0 ± 12.6 months vs 94.7 ± 5.6 months on average, p = 0.003) and site 5 (OS, 79.9 ± 7.3 months vs 97.0 ± 8.2 months on average, p = 0.032) was associated with poor prognosis (Fig. 4M and N), but no significant correlation was found in the other sites (Fig. 4A–L). Considering the statistical significance between tumorous hypermethylation and OS was only found in these two sites, we combined them to integrally analyze their contribution to patients’ outcomes. When we combined patients with hypermethylation at both sites, they achieved significantly worse survival than the patients without hypermethylation at both sites (57.0 ± 13.4 months vs 92.8 ± 5.8 months on average, p = 0.020) (Fig. 4O).

The prognostic value of clinicopathological factors and hypermethylation were analyzed by univariate and multivariate Cox regression analysis. The multivariate Cox analysis included CEA, TNM stage, lymph node metastasis and distant metastasis. To answer the question of whether the co-hypermethylation status of site 1 and site 5 contributes independently prognosis evaluation or is associated with disease factors such as CEA, TNM stage, lymph node metastasis and distant metastasis, the relationship of site-specific hypermethylation with OS was studied using multivariate Cox regression analysis. Results indicated that co-hypermethylation of sites 1 and 5 is a predictor of poor prognosis for CRC patients both in univariate analysis (HR:2.82 [95%CI 1.13–7.04]; p = 0.027) and multivariate Cox regression analysis (HR: 4.43 [95% CI 1.27–15.46]; p = 0.019) (Table 4).

Table 4 Prognostic value of clinicopathological factors and co-hypermethylation of site 1 and site 5 in univariate and multivariate Cox regression analysis (n = 49)