Breast cancer patients (n=94) who underwent surgical treatment at several hospitals [National Hospital Organization Kyushu Cancer Center (Fukuoka, Fukuoka) Kyushu University Beppu Hospital (Beppu, Oita), Oita Prefectural Hospital (Yufu, Oita) and Takada-Chuo Hospital (Yokohama, Kanagawa), all in Japan)] between 1990 and 1999 were enrolled in the study. Prior to sample acquisition, each patient provided written informed consent at the respective hospital. The study was approved by the ethics committees of Kyushu University. Patients were excluded who had been diagnosed with ductal carcinoma in situ. Three patients who had distant metastasis at first diagnosis received no neo-adjuvant chemotherapy. Post-operative adjuvant chemotherapy and endocrine therapy were performed according to the St. Gallen Consensus Conference guidelines (17). Among the 94 patients, 63 were estrogen receptor (ER)-positive. The expression levels of the HER2 protein could not be confirmed in the cases, as the measurements used for HER2 expression were not common when the surgeries were performed. The mean observation period ranged from 1 to 124 months (median, 54 months). Among the 94 patients, only 30, 81 and 57 patients were examined for TET1, TDG and DNMT3A expression, respectively, due to the deficiency of samples for quantification of each cDNA by reverse transcription-quantitative polymerase chain reaction (RT-qPCR).

The resected tumor tissue specimens were frozen immediately in liquid nitrogen and stored at −80°C until analysis. The total RNA extraction from the primary tumors was performed according to the ISOGEN-LS (Nippon Gene Co., Ltd., Tokyo, Japan) manufacturer's instructions. The reverse transcription reactions and first-strand cDNA synthesis were performed as described previously (18).

Quantitative analysis was performed of miR-29b- and RNU6B (internal control)-specific cDNAs derived from total RNA extracted from resected tumors using gene-specific primers, according to the TaqMan MicroRNA Assay protocol (Assay IDs: 000413 for hsa-miR-29b-3p and 001093 for RNU6B; Applied Biosystems, Carslbad, California, USA). The procedures were as described previously (18). The raw miR expression levels were normalized to RNU6B expression for calculation of the relative miR expression values. To determine the relative expression levels of TET1, TDG and DNMT3A, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as the internal control. RT-qPCR was performed using the LightCycler® 480 system and the LightCycler® 480 Probes Master kit (Roche Applied Science, Penzberg, Germany). The sequences of the primers for TET1, TDG and DNMT3A were as follows: TET1 sense, 5′-TCTGTTGTTGTGCCTCTGGA-3′ and antisense, 5′-GCCTTTAAAACTTTGGGCTTC-3′; TDG sense, 5′-ATGCAGCAGTGAACCTTGTG-3′ and antisense, 5′-GTCATCCACTGCCCATTAGG-3′; and DNMT3A sense, 5′-AAGGAGGAGCGCCAAGAG-3′ and antisense, 5′-ATCACCGCAGGGTCCTTT-3′. The expression of DNMT3B was not detected in the samples.

For miR-29b analysis, differences between clinicopathological factors were analyzed using χ² tests for categorical variables. Disease-free survival (DFS) and overall survival (OS) times were measured from the time of the first surgery until the date of mortality or last follow-up. Survival curves were determined by the Kaplan-Meier method and statistical significance between groups was assessed using the Wilcoxon test. Multivariate analysis was performed to assess the relative influence of prognostic factors on OS using the Cox proportional hazards model with a forward stepwise procedure. Statistical analysis was performed by JMP® Pro version 9.0.2 for Mac OS (SAS Institute Japan, Tokyo, Japan).P<0.05 was considered to indicate a statistically significant difference.

miR-29b expression was assessed in primary tumor tissues from 94 breast cancer patients. Patients were divided into miR-29b high and low expression groups according to the median value of miR-29b expression. Clinicopathological factors were subsequently analyzed in association with miR-29b levels. The miR-29b low expression group exhibited a significantly larger tumor size and more advanced clinical stages compared to the miR-29b high expression group (Table I). In terms of DFS and OS, the miR-29b low expression group showed a significantly poorer prognosis than that of the miR-29b high expression group (Fig. 1). Among the ER-positive cases, the low miR-29b expression group had significantly poorer DFS and OS compared to the high miR-29b expression group (Fig. 2). Among the ER-negative cases, low miR-29b expression correlated with a poorer OS only (Fig. 3).

Multivariate analysis of OS showed that the low level of miR-29b expression was an independent prognostic predictor in all patients (Table II).

Additionally, the expression levels of TET1, TDG and DNMT3A were examined in breast cancer primary tumor tissues. These levels were subsequently compared between the miR-29b high and low expression groups. There were no significant differences in any of the patients between the miR-29b high and low expression groups (Fig. 4). However, in analyses of the ER-positive patients, DNMT3A showed significantly higher expression in the miR-29b low expression compared to the high expression group (P=0.027; Fig. 5).