Hypomethylation-Associated Up-Regulation of TCF3 Expression and Recurrence in Stage II and III Colorectal Cancer

Background and Objectives Transcription factor 3 (TCF3) implicates Wnt signaling pathway and regulates E-cadherin expression, which is involved in aggressiveness of tumors. This study aims to investigate the role of TCF3 in predicting prognosis of patients with stage II and III colorectal cancer (CRC). Methods Real-Time quantitative PCR was performed in 64 fresh CRC tissues and 6 cell lines to examine TCF3 mRNA expression. TCF3 protein expression dynamics were detected by immunohistochemistry of 118 paraffin-embedded specimens, and the clinical significance of TCF3 was assessed by clinical correlation and Kaplan-Meier analyses. Aberrant hypomethylation of TCF3 promoter was also investigated using bisulfite sequencing and methylation specific PCR. Results The up-regulation of TCF3 mRNA was frequently detected both in CRC tissues with recurrence and metastasis-derived cell lines. The expression level of TCF3 protein was significantly correlated with histological type (P = 0.038) and disease-free survival time (P = 0.002). Higher TCF3 expression indicated poor prognostic outcomes (P<0.05, log-rank test). Multivariate analysis also showed strong TCF3 protein expression and perineural invasion were independent adverse prognosticators in CRC (P = 0.010, 0.000). Moreover, it was showed that promoter hypomethylation of TCF3 is associated with its up-expression. Conclusions This study highlighted the prognostic value of TCF3 in stage II and III CRC. The up-regulation of TCF3, which is mainly caused by promoter hypomethylation, is one of the molecular mechanisms involved in the development and progression of CRC.


Introduction
Colorectal cancer (CRC) is one of the three leading causes of cancer-related death among worldwide [1]. Stage II and III tumors together represent approximately 70% of CRC patients [2]. Regional lymph node metastasis is one of the most powerful indicators of aggressiveness of CRCs and aids in predicting the clinical outcome. However, one third of CRC patients without histological evidence of lymph node involvement die within five years after surgery from distant metastasis or local recurrence, which suggested that the nodal status may not predict the clinical course of CRC adequately [3]. It is therefore important to identify prognostic factors predicting poor outcome and guiding therapy in stage II and III CRC.
Transcription factor 3 gene (TCF3) was located in chromosomal band 19p13.3. TCF3 is a ubiquitously expressed transcription regulator and encodes two basic helix-loop-helix (HLH) transcription factors, E12 and E47 [4]. These two proteins are characterized by broad expression pattern and ability to bind DNA [5][6][7][8]. Several studies have reported the emerging role of TCF3 in experimental tumors. Overexpression of TCF3 has been detected in CRC [9], prostate cancer [10,11], gastric cancer [12] and renal cancer [13]. As a transcriptional repressor of E-cadherin, TCF3 implicated in epithelial to mesenchymal transition and may be linked to tumor aggressiveness [14]. Increasing transcriptional activity of TCF/b-catenin complex are the initiating event in most colorectal sporadic tumors [15]. Although considerable studies had showed TCF3 was a tumor promoter, its role in cancer progression remains controversial [16,17]. Patel et al. provided three scenarios to demonstrate the potential TCF3 regulated mechanisms at the molecular level [11].
The goal of this study is to assess the clinical significance of TCF3 in human CRC, and to investigate mechanisms mediating overexpression of TCF3, which assist in early identifying a highrisk recurrence subset of the patients with stage II and III CRC.

Patients and tissue specimens
From April 2000 to November 2004, 64 fresh CRC tissues, 36 with recurrence and 28 without recurrence, were collected immediately after operation at the Fudan University Shanghai Cancer Center (Shanghai, China). Characteristics of samples were showed in Table 1. Total 118 paraffin-embedded samples, 78 with recurrence and 40 without recurrence, were collected retrospectively from archival material stored in the department of pathology at Fudan University Shanghai Cancer Center (Shanghai, China) between February 1993 and March 2004, these samples were from different patients from those used for mRNA analysis ( Table 2).
All specimens examined were taken from vital cores of histopathologically confirmed cancers at primary surgery in patients who did not undergo any local or systemic treatment before operation. Tumor samples were reviewed by at least 2 experienced pathologists, and tumor stage was assigned on the basis of system of the International Union Against Cancer. No stage II but all stage III patients received postoperative chemotherapy, but no patient received radiotherapy.
Disease-free survival was defined as the time elapsed from the date of the initial diagnosis to the appearance of local relapse or distant metastasis. Written informed consent was obtained from all patients, and the research protocol was approved by the ethics committee at Fudan University Shanghai Cancer Center.

mRNA analysis by qPCR
Total RNA was isolated with an RNeasy Mini Kit (Qiagen GmbH, Hilden, Germany) and treated with Dnase. According to the manufacturer's instructions, cDNAs were synthesized with Oligo-dT primers (Promega, Madison, WI). The TCF3-specific primers used were 59-CTCGAGAAGAACAGGCCAAG-39 (forword) and 59-GGGGCAGGTACTGAACACAT-39 (reverse). Glyceraldehyde-3phosphate dehydrogenase (GAPDH) served as a control for normalization of gene expression and was amplified using primers 59-GAAGGTGAAGGTCGGAGTC-39 (forward) and 59-GAAGA-TGGTGATGGGATTTC-39 (reverse). The qPCR were carried out using the default PCR cycle on a sequence detection system ABI Prism 7900HT (Applied Biosystems), and amplified cDNA was detected by SYBR Green I dye (Qiagen GmbH, Germany). Data were analyzed using ABI Prism 7900 SDS software (Sequence Detection System 2.0; Applied Biosystems). Ratios of the intensities of the target gene and GAPDH signals were used as a relative measure of the expression level of target genes. All experiments were repeated three times.

Immunohistochemistry
Archival hematoxylin and eosin-stained slides were reviewed by 2 experienced pathologists in accordence with the 2000 World Health Organization classification. A 3-tiered histologic grading system was applied. The TNM stage was evaluated according to the 2002 International Union Against Cancer classification.
A polyclonal rabbit antihuman TCF3 antibody (dilution 1:400) was obtained from Abcam (Cambridge, UK). In the study, 4-mm sections from archival paraffin blocks were deparaffinized and heated twice for 10 minutes each in a microwave oven (500 W) before exposure to the first antibody. Immunoperoxidase staining was carried out using the 2-step EnVision method (DAKO, Glostrup, Denmark) according to the manufacturer's instructions and visualized with 3,39-diaminobenzidine tetrachloride (Sigma, St Louis, MO).
Cytoplasm and nuclear staining were measured for this antibody. Positive cells were counted by 2 pathologists who were blind to clinical outcome. For clinicopathological correlation, we used a 4-tiered scoring system (negative to 3+), which took into account the percentage of positive cells and staining intensity as described previously [18]. The detailed approach was used to generate a score for each tissue core as follows: no staining or staining in ,10% of tumour cells (score 0), faint/barely perceptible partial staining in .10% of tumour cells (score 1+), weak-to-moderate staining in .10% of tumour cells (score 2+), and strong staining in .10% of tumour cells (score 3+). We separately interpreted TCF3 -and 1+ as 'low expression' and TCF3 2+ and 3+ as 'strong expression'.

Bisulfite genomic sequencing (BGS)
Genomic DNA was isolated from tissues using the DNeasy Kit (Qiagen, Hilden, Germany) and stored at 220uC before use. DNA was modified with EZ DNA Methylation-Gold KitTM (ZYMO, Orange, CA) according to the manufacturer's instruction. Generally, 1 ml modified DNA was used in subsequent PCRs. The primers for BGS are 59-GTATAAGGTTGAAAATTTGGGT-39 (forward) and 59-CTCCCTAAAATCCTAAAAATCTTA-39 (re-

Methylation specific PCR (MSP)
The primers for the methylated sequence of TCF3 gene were 59-AATTTTATAGGAAAAAGGCGC-39 (forward) and 59-AAC-CTCGAACGCACATACTA-39 (reverse). For unmethylated sequence were 59-TGGAATTTTATAGGAAAAAGCTGT-39 (forward) and 59-AAAAACCTCAAACACACATACTA-39 (reverse). Qiagen HotStarTaq DNA polymerase was used for PCR. Thermocycling conditions used were 1 cycle at 95uC for 15 min; 30 cycles at 94uC for 30 s, 53uC (for M primer set) or 51uC (for U primer set) for 30 s, and 72uC for 30 s; and final extension at 72uC for 7 min. PCR products were loaded 2% agarose gels followed by staining with ethidium bromide and directly visualized under UV illumination. A sample was classified as hypermethylated when the methylation amplification product alone was observed, partially methylated when both methylated and unmethylated amplification products were seen, and unmethylated when it showed unmethylated amplification products alone, or neither amplification products were found. For the statistical analysis, the hypermethylated and partially methylated samples were considered as the methylated (M) group and compared with the unmethylated (U) group [19].

Statistical analysis
Statistical analyses were conducted with Stata (version SE/10; StataCorp, College Station, TX). The association among categorical data was analyzed by using the x2 test. Survival curves were generated by the Kaplan-Meier method, and univariate survival distributions were compared with the use of the log-rank test. The multivariate Cox proportional hazards model was used for detection of independent prognosticator. The 2-tailed P value for significance was established at 0.05.

TCF3 expression was up-regulated in recurrent CRC tissues and CRC-metastasis-derived cell lines
In 64 fresh samples, TCF3 mRNA expression was significantly higher in recurrent CRC tissues than in those without recurrence (P = 0.026; Figure 1a). Receiver-operator curve (ROC) analysis showed that the best cut-off value to distinguish between recurrent and non-recurrent CRCs was 0.0041. The the areas under the ROC curves was 0.668 (95% CI 0.534-0.803) (Figure 1b). Relative quantities of TCF3 mRNA in CRC cell lines were expressed as N-fold difference in relation to Caco-2 and normalized to the GAPDH as a reference gene. The TCF3 mRNA expression in SW620, LoVo, and Colo205 were increased 3.4-, 1.8-and 6.9-fold, respectively, compared with that of Caco-2. Whereas TCF3 mRNA levels of SW480 and HCT116 were decreased 0.5-and 0.4-fold, respectively (Figure 1c). For 118 immunostaining samples, TCF3 protein expression was significantly associated with recurrence (Table 3).

Correlation between TCF3 protein expression and clinicopathological features
Positive staining was observed primarily in cytoplasm and nuclear of the cancer cells. Analysis of TCF3 expression in the 118 CRC tissues revealed that 37% of samples demonstrating strong (2+ and 3+) intensities and 63% low (2 and +) intensities. Immunostaining of TCF3 protein is illustrated in Figure 2.
No significant differences were observed regarding age, sex, size, lymphvascular and perineural invasion. However, the disease-free survival rate was significantly lower in patients with high TCF3 protein expression than in those with low expression (P = 0.002), and TCF3 expression was significantly associated with histological type of cancer (P = 0.038). Table 2 showed the relations between clinicopathological features and TCF3 protein expression in118 samples.

TCF3 protein expression and disease-free survival time
Univariate analysis by Kaplan-Meier plots revealed that strong TCF3 expression was significantly associated with unfavorable disease-free survival results (P = 0.0001). Kaplan-Meier curves did not demonstrate any survival difference in CRC according to other parameters, including sex, age, tumor location, and lymphvascular invasion (P,0.05). However, survival distributions were significantly different in CRC with and without perineural invasion. Characteristic plots of TCF3 expression and perineural invasion are shown in Figure 3. Strong TCF3 expression was associated with recurrence (Table 3). Multivariate Cox analysis indicated that perineural invasion, TCF3 protein expression, and  Table 4).

Up-expression of TCF3 is associated with its promoter CpG island hypomethylation
A CpG island encompassing about 2.3 kb was found in the human TCF3 gene promoter region (Figure 4c). To understand the mechanism of TCF3 regulation in CRC, the methylation status in the promoter region of TCF3 was tested. We amplified and sequenced the promoter region of TCF3 using sodium bisulfite-treated genomic DNA prepared from CRC tissues. More densely methylated CpG sites (90.7% vs. 44.3%) were detected in non-recurrent CRC tissues in this study (Figure 4d, 4e, and 4f). To further validate the sequencing results, the promoter region was characterised by MSP using methylation or unmethylation-specific primers in 47 CRC specimens. Methylation of TCF3 was detected in 23 of (95.8%) 24 non-recurrent CRC specimens. In contrast, the frequency of methylation was noticeably lower in recurrent CRC specimens (16/23, 69.6%; P = 0.023, Figure 5a, 5b). Furthermore, it was shown that the TCF3 expression is significantly associated with the methylation status, indicating epigenetical inactivation might play an important role in regulatory of TCF3 expression (P = 0.001, Figure 5c).

Discussion
Our study investigates the correlation between TCF3 expression and the clinicopathological features in patients with stage II and III CRC. The TCF3 expression in recurrent CRC tissues was found significantly higher than in those without recurrence. Survival analysis showed that strong TCF3 protein expression and perineural invasion were independent adversely prognostic factors, and chemotherapy was an independent protection factor. To eliminate treatment bias caused by adjuvant chemotherapy, we analyzed the correlation between TCF3 expression and survival time in stage II and III patients, respectively. Chemotherapy did not influence the association between TCF3 expression and prognosis ( Figure S1). Hypomethylation of promoter CpG islands of oncogene is capable of activating these genes. To determine whether the upregulation of TCF3 was associated with aberrant methylation, we investigated the methylation frequency in 47 CRC tissues by MSP. The methylated allele was detected in 39 of 47 (83%) tumors tested. The frequency of methylation was significantly lower in recurrent CRC specimens compared with tumors without recurrence (P = 0.023). The TCF3 expression in 39 tumors with promoter methylation was significantly lower than that in 8 cases without promoter methylation (P = 0.001), indicating that promoter hypomethylation was the major mechanism of the upregulation of TCF3. Increased TCF/b-catenin signaling is one of the hallmarks of CRC [20]. We, herein, provide the evidence showing methylation of TCF3 is a common event in colorectal tumors. These data suggest that TCF3 methylation inactivation is necessary for the inhibitory oncogenic potential of the TCF/bcatenin signaling. On the contrary, aberrant hypomethylation increases TCF3 expression, which is associated with recurrence of stage II and III CRC.
The members of TCF family bind to unphosphorylated bcatenin [21][22][23][24] and regulate transcription of target genes, such as TCF itself and the oncogenes cyclin D1 and c-myc [25,26]. Misregulation of this signaling pathway is an important event in the development of several malignancies such as colon cancer, melanoma and prostate cancer. Besides misregulation of TCF/bcatenin pathway in tumor cells, it has been known that malfunction of E-cadherin allows tumor cells to invade the surrounding tissues [27]. E-cadherin expression is also reduced or absent in many epithelial cancers, including gastric and breast cancer [28][29][30]. As a transcriptional repressor, the role of TCF3 is down-regulating E-cadherin during tumor progression [31,32]. TCF3 binds the E-box elements at the proximal promoter site of E-cadherin leading to transciption inactivation of E-cadherin, and the E-cadherin downregulation plays an important role in reducing cell-cell adhesion systems and promotes metastasis [33].
In view of the comprehensive data presented in this study, we propose that over-expression of TCF3 in stage II and III CRC patients was significantly associated with poor prognosis and low disease-free survival rate (Figure 3), which contributes to identify those high-risk CRC patients. Some stage II and all stage III CRC cases could be considered for adjuvant therapy [34]. However, the role of adjuvant chemotherapy in those patients with stage II tumors is still unclear [35,36]. To select high-risk stage II disease and maximize the benefits of adjuvant therapy, an independent prognostic marker could be helpful in identifying aggressive phenotypes within stage II CRC. We have identified GAS1 as a factor to select patients with high risk of recurrence in our previous work [18], and some interesting candidate genes as prognostic factors for these samples are also validating in our laboratory (unpublished data). In this study, we showed that strong TCF3 expression can identify a subset of patients at high-risk of recurrence and these patients may therefore benefit from more aggressive treatment.
In conclusion, our data highlights the pivotal role of TCF3 hypomethylation in CRC recurrent development. These results underscore the potential prognostic value of TCF3 as a biomarker for choosing adjuvant therapy for patients at high-risk of a poor outcome. Figure S1 Kaplan-Meier estimated survival rates according to TCF3 expression. Patients with high TCF3 expression showed significantly poorer prognisis than those with low TCF3 expression in stage II (a) and III patients (b) (P,0.05, log-rank test).