Cell lines used in this study were obtained from the Cell Resource Center for Biomedical Research, Tohoku University (Miyagi, Japan) and the American Type Culture Collection (Manassas, VA, USA). Fifteen CRC cell lines (CCK81, CoCM1, Colo205, Colo320, ColoTC, DLD1, HCT116, HCT15, HT29, LoVo, RCM1, RKO, SW48, SW480 and WiDr) were cultured in appropriate medium and under conditions described by the providers, with media obtained from Gibco (Grand Island, NY, USA) or Sigma (St. Louis, MO, USA), supplemented with 10% heat-inactivated fetal bovine serum (FBS; Nichirei, Tokyo, Japan), 100 μg/ml streptomycin, and 100 U/ml penicillin (Invitrogen, Carlsbad, CA, USA).

HCT116, RKO, Colo320, SW480 and HT29 CRC cell lines were treated with 5-aza-dC, each experiment was performed in triplicate. The five cell lines were cultured at 37°C in 5% CO₂. Cells (5×10³ per well) were plated in their respective culture media on day 1, on day 3 cultured cells were treated with PBS dissolved and filter-sterilized 0.5 μM 5-aza-dC (Sigma) for 72 h with media changed every 24 h, treatments were performed in parallel with drug-free phosphate-buffered saline (PBS) as a control. The dose of 5-aza-dC was administered on the basis of its pharmacological dose and the results of our preliminary experiments (12). On day 6, cells were harvested following incubation with trypsin-EDTA then stained with trypan blue and counted. For the analysis of mRNA expression total RNA was extracted using QIAshreder with RNeasy minikit (Qiagen, Hilden, Germany). Genomic DNA (gDNA) was extracted by proteinase K (Invitrogen) digestion, phenol/chloroform extraction and ethanol precipitation method for methylation analysis.

Paired samples from 23 patients, who underwent surgical treatment for CRC in 2007 at Tokyo Medical and Dental University Hospital were included in this study. All specimens (macroscopically resected) were stored at −80°C until further use. Mean age was 66.3±10.9 years (median 67.0 years; range 38–81 years) and male-to-female ratio was 1.3:1. The number of cases for stages I, II, III and IV was six, six, five and six, respectively. Written informed consent was obtained from all patients and the Institutional Review Board at Tokyo Medical and Dental University approved the study. All the samples were used in methylation analysis, of them 14 assigned to gene expression study.

RNA was extracted from cell lines before and after treatment with 5-aza-dC. Contaminant DNA was removed by digestion with RNase-free DNase (Qiagen). The integrity of the total RNA obtained was assessed using an Agilent 2100 BioAnalyzer (Agilent Technologies, Palo Alto, CA, USA). All samples had an RNA integrity number of ≥5.0 prior to gene expression analysis. Complementary RNA (cRNA) was prepared from 2 μg total RNA using one-cycle target labeling and a control reagents kit (Affymetrix, Santa Clara, CA, USA). Hybridization and signal detection of the Human Genome (HG) U133 Plus 2.0 array (Affymetrix) were performed according to the manufacturer’s protocol.

The gene expression data sets were submitted to Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/, accession no. GSE32323). To investigate the difference in expression patterns, the raw gene expression data were derived from each probe signal intensity (CEL file) using the model-based robust multi-array average method algorithm as implemented in the Affymetrix Expression Console™ software (version 1.1). Expression levels were log₂-transformed, and 62 control probe sets were removed from subsequent analyses. R 2.11.1 statistical software together with a Bioconductor package (R Foundation for Statistical Computing, Vienna, Austria) was used to calculate fold-change (FC) values for each of the 54,613 probes on the HG-U133 Plus 2.0 array. FC ≥1.2 in all cell lines was selected as the threshold for up-regulation of gene expression. Genes with an FC difference greater than threshold in all of the five cell lines with FC >1.5 in at least one cell line were selected.

gDNA was extracted using phenol/chloroform method. Bisulfite treatment was performed using EpiTect Bisulfite kit (Qiagen) according to the manufacturer’s instructions. Sodium bisulfite treatment of gDNA converts unmethylated cytosine residues (but not methylated cytosine) to uracil, which is then converted to thymidine during the subsequent PCR step, giving sequence differences between methylated and unmethylated DNA.

The methylation status of the cyclin-dependent kinase inhibitor 2A (CDKN2A/p14^ARF), cyclin-dependent kinase inhibitor 2A (CDKN2A/p16^INK4A), growth arrest and DNA-damage-inducible beta (GADD45B), protein-tyrosine phosphatase receptor type O gene (PTPRO) and thrombospondin, type I, domain containing 1 (THSD1) genes was determined by methylation-specific polymerase chain reaction (MSP) with 1 μl of bisulfite-treated DNA as template and AmpliTaq Gold 360 Master Mix (Applied Biosystems, Foster City, CA, USA) for amplification, as previously described (13). Bisulfite-converted gDNA, methylated and unmethylated human control DNA (EpiTect control DNA, Qiagen) was used as a positive control for the methylated and unmethylated experiments, respectively.

Methylated and unmethylated primer sequences (Life Technologies Inc., Rockville, MD, USA), amplicon lengths and PCR conditions have been previously reported (12,14–16). None of the primer sets amplified non-bisulfite-treated gDNA (data not shown), this eliminate the possibility of PCR product amplification from unconverted DNA. All reactions were performed in duplicate and PCR products were loaded onto a 2.0% agarose gel, stained with 0.5 μg/ml ethidium bromide and visualized under ultraviolet illumination.

PTPRO gene was included as a positive control for target gene methylation analysis and verified the integrity of gDNA used in this study (i.e., showed unmethylated bands in both normal and tumor tissues) (17). Methylation status of p14^ARF and p16^INK4A gene was used to confirm the efficient demethylation of CpG-dinucleotides in the treated cell lines (17).

Total RNA was extracted from primary CRC and adjacent normal tissues (14 pairs) using RNeasy minikit (Qiagen) and from 14 colon cancer cell lines. Total RNA (10 μg) was reverse-transcribed into complementary DNA (cDNA) samples using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) according to the manufacturer’s protocol. TaqMan gene expression assays (Applied Bioystems: THSD1, Hs00938785_m1, and β-actin, Hs99999903_m1) were used to determine the expression of THSD1. β-actin was used as an internal control. The PCR reaction was carried out with the TaqMan Universal PCR Master Mix (Applied Biosystems). Thermal cycling conditions (7300 ABI PRISM, TaqMan; Applied Biosystems) were as follows: 50°C for 2 min, 95°C for 10 min, 40 cycles of 15 sec denaturation at 95°C and 1 min annealing at 60°C. All calculated concentrations of target genes were normalized by the amount of the endogenous reference with the comparative Ct method for relative quantification (ΔΔCt method) using Relative Quantification Study software (7300 Sequence Detection System version 1.2.1, Applied Biosystems). Each assay was performed in 20 μl including 1 μl of cDNA and all assays were done in duplicate.

Differences between groups were estimated using; the χ² test, Fisher’s exact test, Mann-Whitney U-test and Wilcoxon signed ranks test, where appropriate. A probability level of 0.05 was used for statistical significance. Statistical analyses of gene expression were performed with SPSS (version 17.0, SPSS Inc., Chicago, IL, USA) for Windows software.

To find hypermethylated genes associated with CRC, microarray analysis comparison of the mock and 5-aza-dC treatment expression patterns in five colon cancer cell lines (HCT116, RKO, Colo320, SW480 and HT29) was done, this identified the reactivated genes expression profiles (method summarized in Fig. 1). Comparison of the resultant gene expression profiles revealed 139 genes that had FC >1.5 in signal intensities in at least one cell line with ≥1.2-fold change in the remaining four 5-aza-dC-treated cell lines collectively. Of note, this group of up-regulated transcripts contains several genes of cancer-germline antigen families (i.e., GAGE, MAGE, PAGE and XAGE), which is consistent with previous reports (18). Also the functional analysis by DAVID bioinformatics resource [(19); http://david.niaid.nih.gov] showed the significant enrichment (FDR <0.05%) of genes located on X chromosome (20 genes). In addition, the reactivation of gene transcripts known to be hypermethylated in CRC was also detected including DDB1 and CUL4 associated factor 4-like 1 (DCAF4L1), DEAD (Asp-Glu-Ala-Asp) box polypeptide 43 (DDX43), intercellular adhesion molecule 1 (ICAM1), msh homeobox 1 (MSX1), placental growth factor (PGF), PTPRO and zinc finger protein 42 homolog (ZFP42) (11,16,20–23).

In order to identify candidate genes potentially affected by methylation with a putative tumor suppressor activity, first we removed duplicated probes (16 genes), then we excluded genes with poor annotation (20 genes) and genes with a chromosomal location on X chromosome (20 genes). Genes with no CpG islands in the promoter (two genes), genes with unknown function (8 genes) and genes with no relevant function in biological mechanisms of tumorigenesis (23 genes), were also excluded. Oncogenes and genes with potential oncogenic activity (22 genes) were also removed. And lastly genes for which methylation status of their promoter CpG island loci has already been reported for CRC (8 genes) were eliminated. We identified 20 candidates (Fig. 2) that have not been reported to be affected by epigenetic mechanism in CRC (Table I) of which GADD45B and THSD1 were selected for further analysis. Both genes have been recently reported to be frequently methylated in cancer types other than CRC (14,15). Also they were identified as having CpG-rich sequences fulfilling the criteria of CpG island (GC content ≥50%, CpG:GpC ratio ≥0.6, and minimum length 200 bp) in their 5′ regions (24).

CpG island searcher (25) and UCSC Table Browser (26) using the current genome assembly (GRCh37/hg19), were used for the potential CpG islands screening. In addition, we searched the published literature for additional analysis of the up-regulated genes.

To identify novel targets for aberrant methylation in CRC, we further studied methylation status of two candidate genes, GADD45B and THSD1, in colon cancer cell lines. By using MSP, we determined their methylation status in 15 colon cancer cells. GADD45B showed no methylation in colon cancer cell lines and was excluded from further analysis, while THSD1 was methylated in 27% (Table II; Fig. 3A).

Methylation status of THSD1 in five CRC cell lines before and after 5-aza-dC treatment was analyzed by MSP. 5-aza-dC treatment caused complete demethylation in the RKO cell line, partial demethylation in the Colo320 cell line and no changes in methylation were seen in SW480, HCT116 and HT29 cell lines (Fig. 3B). This verified that THSD1 promoter is affected by demethylation effect of 5-aza-dC and confirms the data obtained by oligonucleotide microarray analysis.

To test whether the aberrant methylation identified in colon cancer cell lines was also present in primary colon cancer specimens, we studied the methylation status of THSD1 gene in 23 matched CRC samples. Aberrant methylation in primary colon cancers was identified in 9% for THSD1. Distinctively, THSD1 was completely unmethylated in all the samples from normal tissues (Table II; Fig. 3C). There was no significant difference in the methylation of THSD1 between tumor and normal tissues (P=0.49; Table II).

To investigate the relationship between gene expression and promoter hypermethylation, we analyzed THSD1 mRNA expression in 14 CRC cell lines and 14 paired CRC tissues. The level of THSD1 mRNA was reduced (Fig. 4) in all tumor tissues relative to cancer cell lines and adjacent non-tumor tissues, however it did not reach statistical significance (P=0.613 and 0.158, respectively).