In the current study, the microarray data was used from a previous study (13). The gene expression data are deposited in the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE32323.

Probe sets from cell lines were selected according to the following criteria: i) FC >2.0 compared with that of the CpG island methylator phenotype (CIMP) RKO cell line and ii) up-regulation of gene expression in at least two CRC cell lines. For the paired clinical samples, probe sets were selected for FC of normal versus tumor tissue (N/T) >1.5 (i.e., higher expression in normal tissue than in tumor tissue) (14), and 99 genes (123 probes) that appeared to be suppressed by DNA methylation were identified (Table SI). We examined the published literature and narrowed down candidate genes in the context of genes that are hypermethylated in neoplasms, but the clinical significance of inactivation remained unclear in CRC. Finally, we selected BMP2 as the target gene of interest (Fig. 1).

Seven CRC cell lines (RKO, SW480, HT29, HCT116, COLO201, LoVo, and DLD1) were obtained from the American Type Culture Collection. These cell lines were maintained in Dulbecco's modified Eagle's medium or RPMI1640 medium (Gibco; Thermo Fisher Scientific, Inc.) containing 10% heat-inactivated fetal bovine serum, 100 units/ml of penicillin, 100 µg/ml of streptomycin, 10 mM of HEPES, and 1.0 mM of sodium pyruvate and were incubated at 37°C in 5% CO₂. Cultured cells were pelletized and used to isolate total genomic DNA for methylation assay and total RNA for mRNA expression assay.

This study included primary tumors from 498 patients (290 male and 208 female patients) who underwent curative surgical resection for CRC at Tokyo Medical and Dental University Hospital between 2008 and 2013. Of these 498 patients, 91 had stage I disease, 204 had stage II disease, and 203 had stage III disease. The median patient age was 69.0 years (range, 29–93 years). Patients did not receive any treatment prior to surgery. Postoperative adjuvant chemotherapy was administered to 14 patients with stage II disease (6.9%) and 150 patients with stage III disease (73.9%). The median follow-up period at analysis was 63 months (range, 0–122 months). Samples were included in the methylation assay.

We used methylation-specific polymerase chain reaction (MSP) to evaluate the methylation status of BMP2 (15). The phenol/chloroform method was used to isolate total genomic DNA from cell lines and surgically resected tumor samples. Bisulfite treatment was performed using the EpiTect Plus DNA Bisulfite kit (Qiagen), according to the manufacturer's instructions. Bisulfite-modified DNA was then used as template DNA for polymerase chain reaction (PCR) amplification with PCR primers corresponding to the region affected by methylation. The methylation-specific and unmethylation-specific primer sequences of BMP2 were based on the results of the study by Wen et al (16). MSP was performed using the EpiTect MSP kit (Qiagen). The PCR conditions of iCycler™ (Bio-Rad Laboratories Inc.) were as follows: 95°C for 10 min; 40 cycles at 94°C for 15 sec, 62°C for 30 sec, and 72°C for 30 sec (methylated) or 40 cycles at 94°C for 15 sec, 58°C for 30 sec, and 72°C for 30 sec (unmethylated); and finally 72°C for 10 min. EpiTect control DNA (Qiagen) was used as a positive control. After amplification, electrophoresis of PCR products was performed using 2.5% agarose gels. When positive amplification was noted with methylation-specific primers, irrespective of whether amplification was detected with unmethylation-specific primers, the tumor was considered as a ‘methylated’ tumor. When no amplification was noted with methylation-specific primers, the tumor was considered as an ‘unmethylated’ tumor.

The mRNA expression level of BMP2 in each cell line was assessed using reverse transcription-quantitative PCR (RT-qPCR). Total RNA was extracted using the RNeasy mini kit (Qiagen). cDNA was synthesized using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems; Thermo Fisher Scientific, Inc.), according to the manufacturer's instructions, with iCycler™ (Bio-Rad Laboratories, Inc.). cDNA was then amplified by PCR using a fluorescence-based real-time detection method with the ABI Prism 7300 real-time PCR system (Applied Biosystems; Thermo Fisher Scientific, Inc.), according to the manufacturer's protocol. TaqMan BMP2 and β-actin (ACTB) Gene Expression Assay-on-demand (BMP2 Assay ID, Hs00154192_m1; ACTB Assay ID, Hs99999903_m1; Applied Biosystems; Thermo Fisher Scientific, Inc.) were used for RT-qPCR. The mRNA expression level of BMP2 in each sample was normalized to that of ACTB (internal standard). The relative quantification of BMP2 mRNA expression was performed by the ΔΔCq method using SDS v1.4 with RQv1.0 software (Applied Biosystems; Thermo Fisher Scientific, Inc.) (17). Each analysis was performed in triplicate.

The chi-square test was used to estimate differences between the groups. Relapse-free survival (RFS) was calculated from the date of surgery to recurrence or death, whichever occurred first. Overall survival (OS) was calculated from the date of surgery to death from any cause. RFS and OS curves were estimated using the Kaplan-Meier method, and log-rank tests were used to assess treatment differences overall. Factors affecting RFS and OS were examined with univariate and multivariate analyses using the Cox proportional hazards model, and Cox models were used to estimate hazard ratios (HRs) and their confidence intervals (CIs). All statistical analyses were performed using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), which is a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria). More precisely, it is a modified version of R commander designed to add statistical functions frequently used in biostatistics (18). P<0.05 was considered to indicate a statistically significant difference.

This study was conducted in accordance with the Declaration of Helsinki and its later amendments or comparable ethical standards. The study protocol was approved by the Institutional Review Board of Tokyo Medical and Dental University, and written informed consent was obtained from all patients before enrollment.

As shown in Fig. 1, the microarray data was used from a previous study (13). We defined ‘candidate genes’ according to the following criteria: i) FC >2.0 compared with that of the CIMP RKO cell line and ii) up-regulation of gene expression in at least two CRC cell lines. For the paired clinical samples, probe sets were selected for FC of normal versus tumor tissue >1.5. 99 candidate genes were identified among the 54,613 probes (Table SI). We examined the published literature for additional analysis of these genes. In order to identify epigenetically affected genes with methylation, we first excluded genes in neoplasm not affected by methylation. In addition, we removed genes with high or unknown expression in neoplasm, genes with hypomethylation or unknown methylation status in neoplasm. We selected BMP2 reported to be a tumor-suppressor gene in CRC and DNA aberrant hypermethylation in neoplasms, but the clinical significance remain unclear as a target gene for further investigation.

MSP was performed in seven CRC cell lines, and BMP2 methylation was detected in the following three cell lines: RKO, HCT116, and COLO201 (Fig. 2A). With regard to mRNA expression detected by RT-qPCR, the expressions in these three cell lines with BMP2 methylation was quite low compared with the expressions in the other four cell lines without BMP2 methylation (Fig. 2B), suggesting that the mRNA expression of BMP2 was down-regulated by DNA hypermethylation.

The relationships between BMP2 methylation status and important clinicopathological factors are shown in Table I. BMP2 methylation was observed in 302 of the 498 patients (60.6%). BMP2 methylation was associated with positive lymph nodes (P=0.012), venous invasion (P=0.027), and stage III disease (P=0.010). There were no associations of BMP2 methylation with sex, tumor location, histological type, tumor invasion depth, and lymphatic invasion.

The RFS and OS curves of all 498 patients are presented in Fig. 3. The 5-year RFS rates in the methylated BMP2 and unmethylated BMP2 groups were 70.0 and 75.5%, respectively (Fig. 3A). The 5-year OS rates in the methylated BMP2 and unmethylated BMP2 groups were 82.8 and 90.3%, respectively (Fig. 3B). OS tended to be worse in the methylated BMP2 group than in the unmethylated BMP2 group (RFS, P=0.142; OS, P=0.067).

The RFS and OS curves stratified by TNM-stage are presented in Fig. 4. In the stage I and II subgroups, there were no differences in both RFS and OS between the methylated BMP2 and unmethylated BMP2 groups (Fig. 4A-D). On the other hand, in the stage III subgroup, OS was significantly worse and RFS was marginally worse in the methylated BMP2 group than in the unmethylated BMP2 group (P=0.012 and P=0.061, respectively) (Fig. 4E and F).

We focused on stage III patients and investigated the impact of BMP2 methylation status on prognosis (Table II). With regard to poor RFS, in the univariate analysis, left-sided colon (P=0.004), high tumor invasion depth (P=0.022), and lymph node metastasis (P=0.003) were identified as risk factors. In the multivariate analysis involving variables with P-values <0.10 in the univariate analysis, high age (HR=1.67; 95% CI, 1.05–2.64; P=0.029), left-sided colon (HR=2.28; 95% CI, 1.34–3.87; P=0.002), high tumor invasion depth (HR=1.73; 95% CI, 1.08–2.77; P=0.022), and lymph node metastasis (HR=1.75; 95% CI, 1.10–2.79; P=0.019) were identified as independent factors for poor RFS. With regard to poor OS, in the univariate analysis, high age (P<0.001), lymph node metastasis (P=0.026), and BMP2 methylation (P=0.016) were risk factors. In the multivariate analysis, high age (HR=2.97; 95% CI, 1.58–5.59; P<0.001), lymph node metastasis (HR=1.96; 95% CI, 1.05–3.66; P=0.035), and BMP2 methylation (HR=2.36; 95% CI, 1.04–5.39; P=0.041) were identified as independent factors for poor OS.

As the prognostic impact of sidedness has attracted attention in recent years, we investigated the relationship between BMP2 methylation status and prognosis according to the sidedness of the primary tumor (19). In right-sided colon cancer patients, there were no differences in both RFS and OS between the methylated and unmethylated BMP2 groups (Fig. 5A and B). On the other hand, in left-sided colon cancer patients, RFS and OS were significantly worse in the methylated BMP2 group than in the unmethylated group (RFS, P=0.048; OS, P=0.031; Fig. 5C and D).