Background

Colorectal cancer (CRC) is one of the most prevalent malignant tumors around the world, accounting for nearly 10.2% of new cancer cases and 9.2% of total cancer deaths globally [1, 2]. With increasing use of prognostic biomarkers, molecularly stratified therapy for CRC has been gradually improved [3, 4]. However, the identification of stable and easily detectable biomarkers remains a challenge.

Among these biomarkers, epigenetic alterations, especially DNA methylation, is one of the most promising candidates in clinical practice [5,6,7]. DNA methylation is a covalent modification, which always occurs on cytosine nucleotides and exclusive in the context of CpG site [8]. Aberrant methylation of CpG islands in gene promoter is associated with silencing of tumor suppressor gene, which facilitates tumorigenesis and development [9]. As a biomarker, DNA methylation has several advantages to qualify them for broad use, like high stability and repeatability [10]. Previous studies have excavated some DNA methylation biomarkers for prognosis of CRC, such as CIMP, MGMT, and SEPT9. However, reliable DNA methylation biomarkers for clinical practice of CRC are still limited [11].

To identify new DNA methylation biomarkers, we have conducted a genome-wide screen for hypermethylated genes in cancer tissue by using Infinium MethylationEPIC (EPIC) BeadChip. Among all, regulating synaptic membrane exocytosis 2 (RIMS2) is one of the top candidate genes. RIMS2, also known as RIM2, codes for a presynaptic active zone protein with multidomain, including Rim2α, Rim2β, and Rim2γ [12]. Rim2α, the full-length form of Rim2, is composed of an N-terminal Zn2+ -finger domain, a central PDZ and C2A domains, and a C-terminal C2B domain [13]. Rim2α participates in Ca2+ -dependent neurotransmitter release from synaptic vesicles by interacting with Rab3 [14, 15]. Rim2α also interacts with Munc13-1 [16], Rab8 [17], cAMP-GEFII [18], and ELKS [19]. Rim2α is expressed mainly in endocrine and neuroendocrine cells such as pancreatic β cells, pituitary, and adrenal gland [13]. Several studies have suggested that abnormal expression of RIMS2 may involve in the development of cancer [20, 21]. However, there is no relative study between RIMS2 and CRC.

In this study, we determined the methylation and expression status of RIMS2 in CRC and investigated the prognostic utility of RIMS2 promoter methylation in CRC.

Methods

Study cohort and data collection

Two cohorts were used in this study. Cohort 1 comprises 30 pairs of snap-frozen tumor tissue and adjacent normal tissue from 30 CRC patients, which was used for correlation analysis between the methylation and expression status of RIMS2 in CRC tissue. Cohort 2 comprises 254 formalin-fixed paraffin-embedded (FFPE) tumor tissue from 254 CRC patients who underwent radical surgery between 2009 and 2012 at the Sixth affiliated hospital of Sun Yat-Sen University [22, 23]. Cohort 2 was used for prognosis analysis. Only pathologically confirmed CRC cases were eligible. The following clinical materials were obtained from the Electronic Medical Record System: demographic characteristics (gender, age), tumor location, TNM cancer stage, MSI status, KRAS mutation, differentiation degree, preoperative serum CEA level, lymphovascular invasion, and perineural invasion. Polymerase chain reaction (PCR) based methods have been used to assess MSI status. The study was approved by the Medical Ethics Committee at the Sixth Affiliated Hospital of Sun Yat-sen University. Follow-up data were applied from a CRC database of the Sixth Affiliated Hospital of Sun Yat-sen University. The follow-up was censored in September 2018 with a median follow-up of 85.1 months.

Cell culture and 5-aza-2’-deoxycytidine treatment

DLD1 and HCT116 CRCcolorectal cancer cell line purchased from American Type Culture Collection (ATCC) were cultured at proper medium at 37℃ with 5% CO2 following the manufacturer’s protocol. A demethylation drug 5-aza-2’-deoxycytidine (5-Aza, sigma, A3656) was used to treat the above CRC cell lines as previous study described [24].

DNA extraction and quantitative methylation-specific PCR

Genomic DNA from snap-frozen cancer tissue and CRC FFPE was extracted by the QIAamp DNA Mini Kit (Qiagen, 51,306), and then sodium bisulfite conversion was completed by using the EZ DNA Methylation Kit (Zymo Research, D5002). RIMS2 promoter methylation was measured using quantitative methylation-specific PCR (qMSP) as the previous study described [25, 26]. CpGenome Universal Methylated DNA Kit (Millipore, S7821) was used as a fully methylated positive control (M). CpGenome Universal Unmethylated DNA Set (Millipore, S7822) was used as the negative unmethylated control (U). Aluc4 was used as an internal reference. The specific sequence of primer and probe of RIMS2 and Aluc4 used in this study were provided in Table 1. The methylation degree of the target locus was calculated by percentage methylated reference (PMR), which was calculated by dividing the RIMS2/Aluc4 ratio of a sample by the RIMS2/Aluc4 ratio of fully methylated control and multiplied by 100% [27].

Table 1 Primer and probe used for qMSP and RT-qPCR
Full size table

RNA extraction and quantitative real-time PCR

Total RNA extraction was done by using the UNlQ-10 Column Trizol Total RNA Isolation Kit (Sangon Biotech, B511321). Reverse transcription was conducted by the QuantiTect Reverse Transcription Kit (Qiagen,205,311). Reverse transcription‑quantitative PCR (RT-qPCR) was carried out with a QuantStudio 7 flex Real-Time PCR system and FastStart Essential DNA Green Master (Roche, 06,402,712,001) following the manufacturer’s protocol. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as an internal control. Primers’ sequence of RIMS2 and GAPDH used for RT-qPCR were listed in Table1. The relative mRNA expression level of RIMS2 was calculated by the 2(−△△CT) method [28].

Statistical analysis

Comparison of the RIMS2 methylation and mRNA expression between cancer and adjacent normal tissue was done by paired Wilcoxon signed-rank test. The relationship between the DNA methylation and mRNA expression of RIMS2 in cancer tissue was tested by Spearman’s Rank Correlation analysis. The optimal cutoff point for RIMS2 methylation in prognosis analysis was determined by maximally selected rank statistic [29]. Based on this cutoff, patients in Cohort 2 were divided into methylation-low and methylation-high groups. For the comparison of baseline characteristics, Wilcox rank-sum test was applied for continuous variables, and the chi-square test was applied for discrete variables. The primary outcomes were disease-free survival (DFS). DFS was the length of time from radical resection to recurrence. Cox regression analysis and Kaplan–Meier survival curves with log-rank tests were used to compare DFS. The variables considered clinically relevant or reached the significance level of P < 0.1 in univariate Cox regression were entered into the multivariate Cox regression analysis. The comparison between predictive models was assessed using likelihood ratio (LR) and Akaike information criterion (AIC) in competing models. In general, the model with a lower AIC and a higher LR were considered a better option. The predictive nomogram was developed based on the best model. Calibration curves were plotted to explore the predictive accuracy of the nomogram. A receiver operating characteristic (ROC) curve was plotted to assess the discriminative ability of the nomogram. Discriminative ability was quantified with the area under the ROC curve (AUC). Two-tailed tests with P < 0.05 were considered statistically significant. All statistical analyses were conducted using SPSS (version 26.0) or R (version 3.6.0).

Results

RIMS2 expression is prevalently silenced by promoter methylation in CRC

By analyzing the data from Infinium MethylationEPIC (EPIC) BeadChip, RIMS2 promoter CpG island was found to be significantly hypermethylated in CRC comparing to normal tissue (Fig. 1). The methylation level of RIMS2 promoter was validated by qMSP in 30 cases of snap-frozen CRC and matched adjacent normal tissue. The methylation level of RIMS2 was significantly higher in CRC tissue (Fig. 2a). RIMS2 expression was proved to be suppressed in CRC tissue measured by quantitative real-time PCR (Fig. 2b). Spearman test and dot plot showed that RIMS2 expression was negatively correlated with RIMS2 promoter methylation (Fig. 2c). To further validate the regulation of RIMS2 expression by RIMS2 methylation, a demethylation reagent 5-Aza (5-aza-2’-deoxycytidine) was used to treat DLD1 and HCT116 cell lines. Restoration of RIMS2 expression was found in DLD1 and HCT116 after 5-Aza treatment, with the decreasing of methylation level (Fig. 2d and e). These results suggested that RIMS2 expression was silenced by promoter hypermethylation in CRC.

Fig. 1
figure 1

Schematic diagram of the differential methylation region in 48 pairs of cancer and normal tissue. The methylation values of each EPIC probe in CRC and normal tissue was shown as a heatmap. The differential methylation region with the most significant difference was marked by a red box and used for following qMSP, which was located at CpG island of RIMS2. RIMS2: regulating synaptic membrane exocytosis 2; qMSP: quantitative methylation-specific PCR

Full size image
Fig. 2
figure 2

RIMS2 expression is silenced by promoter hypermethylation in CRC. a. RIMS2 promoter methylation level in paired fresh CRC and adjacent normal tissue. Y-axis showed the percentage methylated reference (PMR), which was used to calculate the methylation level. b. RIMS2 mRNA expression level in paired fresh CRC and adjacent normal tissue. c. Correlation between RIMS2 promoter methylation and mRNA expression in CRC tissue. d. RIMS2 promoter methylation level in HCT116 and DLD1 CRC cell line before and after 5-Aza treatment. e. RIMS2 mRNA expression level in HCT116 and DLD1 CRC cell line before and after 5Aza treatment. *P < 0.05; f. Maximally selected rank statistic was used to select the optimal cutoff point for RIMS2 promoter methylation

Full size image

Association of RIMS2 methylation with clinicopathological, molecular and features in patients with CRC

254 formalin-fixed paraffin-embedded (FFPE) tumor tissue from CRC patients who underwent radical surgery were obtained for prognosis analysis. RIMS2 methylation level in these patients’ samples was measured by qMSP as well. By using the maximally selected rank statistics in an R package called “Maxstat”, PMR of 16.22% was set as the optimal cutoff point for RIMS2 methylation (Fig. 2f). Based on this cutoff, these 254 patients were split into methylation-low (n = 113) and methylation-high (n = 141) group. As shown in Table 2, RIMS2 hypermethylation was correlated with KRAS mutation and differentiation. No significant association was found between RIMS2 methylation and age, gender, tumor location, TNM stage, MSI, lymphovascular invasion, perineural invasion and preoperative serum CEA.

Table 2 Association of the RIMS2 methylation with the baseline characteristics
Full size table

Prognostic value of RIMS2 methylation in CRC

Kaplan–Meier curve revealed a significantly poor DFS (P = 0.01) in the RIMS2 hypermethylation group (Fig. 3a). Univariate analysis showed that older age, advanced TNM stage, lymphovascular invasion, preoperative higher CEA and RIMS2 hypermethylation were associated with poor DFS (Table 3). To eliminate the influence of potential confounders, a multivariate Cox analysis was conducted. According to the multivariate analysis, older age, advanced TNM stage, KRAS mutation, lymphovascular invasion and RIMS2 hypermethylation were associated with poor DFS (Table 3). Taken together, RIMS2 methylation was independently associated with DFS (HR: 1.992 (1.140–3.479), P = 0.015) in CRC. Furthermore, we compared the prognostic value of RIMS2 methylation in different subgroups. Kaplan–Meier curves showed that DFS was not significantly different in the TNM stage I-II subgroup (Fig. 3d) and MSS subgroup (Fig. 3j), only significantly different in the TNM stage III-IV subgroup (Fig. 3g).

Fig. 3
figure 3

Kaplan–Meier curves of DFS in all patients and different subgroups. (a d g j). Kaplan-Meier survival curves of DFS of low RIMS2 methylation group and high RIMS2 methylation group in all patients, TNM stage I-II subgroup, TNM stage III-IV subgroup and MSS subgroup, respectively; (b e h k). Kaplan-Meier survival curves of DFS of Low + KRASWT group and no Low + KRASWT group in all patients, TNM stage I-II subgroup, TNM stage III-IV subgroup and MSS subgroup, respectively; (c f i l). Kaplan-Meier survival curves of DFS of KRAS mutant group and KRAS wild group in all patients, TNM stage I-II subgroup, TNM stage III-IV subgroup and MSS subgroup, respectively

Full size image
Table 3 Cox proportional hazard analyses on DFS in patients with CRC
Full size table

Next, we explored whether RIMS2 methylation could further stratify KRAS wild and KRAS mutant patients in marker prediction assays. In survival analysis, patients were divided into different groups according to whether with low methylation and KRAS wild type at the same time (Patients with unknown KRAS status were excluded). As expected, patients with low methylation and KRAS wild type had a significantly better DFS not only in all patients (Fig. 3b) but also in the 3 subgroups (Fig. 3e, h, k). As for KRAS status, DFS was significantly different only in the TNM stage I-II subgroup (Fig. 3f), but not in all patients (Fig. 3c) and other subgroups (Fig. 3i, l). So Low + KRASWT (Patients with low RIMS2 methylation and KRAS wild type) may be a better prognosis biomarker. Another multivariate COX analysis, including age, Low + KRASWT, TNM stage, lymphovascular invasion and CEA, showed that Low + KRASWT was independently associated with DFS (Table 3).

Comparison between different models

Model 3 had a lower AIC and a higher LR compared with model 1 (AIC: 764.44 s. 768.59; LR: 17.34 vs. 11.81, P = 0.013; Table 4). In the comparison between model 4 and 5, after RIMS2 methylation was added, a lower AIC and a higher LR were observed (AIC: 530.90 vs. 526.91; LR: 24.61 vs. 30.60, P = 0.014; Table 4). These results showed that RIMS2 methylation could increase the prognostic values of current prognostic panels. In the comparison between model 5 and model 6, after replacing KRAS status and RIMS2 methylation with Low + KRASWT, a lower AIC and a higher LR were observed (AIC: 526.91 vs. 520.83; LR: 30.60 vs. 34.68, P = 0.043; Table 4). The result showed that the Low + KRASWT could predict prognosis better, and model 6 was the best predictive model in the study.

Table 4 Comparison between different models
Full size table

A nomogram for predicting DFS in CRC patients

A nomogram for predicting 3-year and 5-year DFS was generated using the variables from model 6, including older age, TNM stage, lymphovascular invasion and Low + KRASWT(Fig. 4a). The calibration curves for the nomogram were shown. The C-indexes of the nomogram for predicting 3-year and 5-year DFS were 0.712 and 0.713, respectively (Fig. 4b, c). ROC curve displays the performance of the nomogram. The AUCs of 3-years DFS and 5-years DFS were 0.728 and 0.750, respectively (Fig. 4d).

Fig. 4
figure 4

A nomogram, calibration curves and ROC curves for predicting DFS in CRC patients. a. A nomogram to predict 3-year and 5-year DFS based on model 6; (b, c). The calibration curves for predicting 3-year and 5-year DFS; d. ROC curves for 3-year and 5-year DFS

Full size image

Discussion

In this study, we found that RIMS2 promoter showed a pattern of hypermethylation and low expression in snap-frozen CRC tissue comparing to normal tissue, which was consistent with the mechanism of tumor suppressor gene inactivation [30]. The findings that RIMS2 expression may be subject to promoter methylation was further validated by 5-Aza-treatment. Then, we explored the prognosis value of the RIMS2 methylation level in CRC. Patients with a high level of RIMS2 methylation tended to have less frequent KRAS mutation and high differentiation. These patients also showed significantly worse postoperative outcome. Furthermore, a combination of RIMS2 promoter methylation and KRAS status could predict the DFS better.

As a presynaptic protein, RIMS2 plays an important role in normal neurotransmitter release [31]. Previous studies showed that RIMS2 was mainly involved in some nervous system diseases, such as schizophrenia, heroin addiction, and Autism Spectrum Disorder [32,33,34]. Only limited evidence indicated that RIMS2 may also contribute to the development of cancer [21]. However, RIMS1, another member of the RAS gene superfamily with similar structure and function to RIMS2, has shown a promising prognostic value in several kinds of cancer, like gastric cancer [35,36,37,38]. In addition, some studies found that dysregulation of Rab3, the interacting protein of RIMS2, may have a significant role in carcinogenesis and progression of CRC [39, 40]. As far as we know, this is the first study investigating the role of RIMS2 in CRC. This study found that RIMS2 methylation level was significantly higher in tumor compared to normal tissue, which results in low RIMS2 expression in CRC. Furthermore, patients with higher methylation levels of RIMS2 tended to have a worse outcome, indicating that RIMS2 played an important role in the occurrence and progression of CRC. In this study, the association between RIMS2 promoter methylation and prognosis of CRC remained significant after adjustment for some clinicopathological predictors, which showed a promising prognostic value of RIMS2 promoter methylation in CRC. The expression of RIMS2 was silenced by promoter methylation in CRC, indicating that RIMS2 may be a potential epigenetic therapeutic target. As a biomarker, RIMS2 promoter methylation can be easily detected by a PCR-based method, which allows it to be used broadly in clinical practice.

Still, there are some limitations in this study. The current study aimed to investigate new biomarkers and mainly focused on the prognostic value of RIMS2 in CRC. However, the specific mechanism of RIMS2 methylation in tumorigenesis and development remained unknown, which needed further confirmation by a series of experimental evidence. The influence of RIMS2 methylation on phenotypes of CRC cell lines will be valuable in studying the mechanism. Besides, patients were only from Chinese, and all the clinical data was obtained from the Sixth affiliated hospital of Sun Yat-Sen University. Multicenter, larger-scale studies would be helpful for confirming this result. In addition, some patients in cohort 2 whose molecular characteristics were incomplete may influence the study results. At last, as a retrospective study, the evidence level is insufficient and a prospective cohort is needed in future studies.

Conclusion

RIMS2 is frequently hypermethylated and lowly expressed in CRC, and patients with hypermethylated RIMS2 tend to have poor survival outcomes. RIMS2 promoter methylation is an independent prognostic biomarker for DFS in CRC.