Six human gastric carcinoma cell lines, H-111-TC, HGC-27, Kato-III, MKN-1, MKN-45 and NU-GC-4 were purchased from Riken Cell Bank (Tsukuba, Japan) and used in the present study. According to the description by the Riken Cell Bank and literature (22–27), these cell lines originated from 1 unknown origin (H-111-TC) and 5 metastatic carcinoma, as well as 2 tubular (well differentiated or adenosquamous) carcinoma (H-111-TC and MKN-1) and 4 poorly differentiated carcinoma including signet ring cell type. HGC-27 and MKN-1 cell lines were maintained in Dulbeccos modified Eagles medium (DMEM; Gibco-BRL, Rockville, MD, USA), whereas NU-GC-4, Kato-III, MKN-45 and H-111-TC cell lines were maintained in RPMI-1640 (Gibco-BRL), supplemented with 10% fetal calf serum (FCS), penicillin, and streptomycin, at 37°C in a humidified atmosphere of 5% CO₂. When they reached 80–90% confluence, the cells were washed with phosphate-buffered saline (PBS) and homogenized immediately in Isogen reagent (Nippon Gene, Osaka, Japan). Total RNA was extracted according to the manufacturers instructions.

The RNA samples were suspended in 20 µl of nuclease-free water, and miRNA-200c levels were quantified using TaqMan MicroRNA Assay technology (Applied Biosystems; Life Technologies, Carlsbad, CA, USA), as previously described (28). miRNA-200c levels were normalized against levels of RNU6B. The hsa-miR-200c oligonucleotide was a synthetic double-stranded 23-nt RNA, 5-UAAUACUGCCGGGUAAUGAUGGA-3 and 5-CGUCU UACCCAGCAGUGUUUGG-3 purchased from BONAC Corp. (Fukuoka, Japan). To assess the mRNA levels of ZEB1, ZEB2 and E-cadherin, reverse-transcription reactions were performed using M-MLV reverse transcriptase (Invitrogen, Carlsbad, CA, USA). Quantitative PCR (FastStart Universal SYBR-Green Master; Roche, Basel, Switzerland) reactions were run on an Mx3005P thermocycler (Stratagene, La Jolla, CA, USA) and analyzed using MxPro QPCR software, version 4.01 (Stratagene, Agilent Technologies, Santa Clara, CA, USA). The level of gene expression relative to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was determined. The primer sequences were as follows: ZEB1 forward, 5-TGTCA CCATGAAACCATTGC-3 and reverse, 5-AGGTAAAGTG CGCTTCCTCA-3; ZEB2 forward, 5-GGGCATTCAGTGA CCTGACA-3 and reverse, 5-GCATTGTTCCCATAGAGT TC-3; E-cadherin forward, 5-TCCTCGGATTCTCTGCT CTC-3 and reverse, 5-CTCTGACCTTTTGCCAGGAG-3; GAPDH forward, 5-ATGGGGAAGGTGAAGGTCG-3 and reverse, 5-GGGTCATTGATGGCAACAATATC-3. The experiments were performed 3 times, with the average levels and standard deviations from the mean being obtained. Correlations between average levels of miR-200c, ZEB1, ZEB2 and E-cadherin in all cell lines were estimated by Spearmans rank correlation using the SPSS software package (ver17.0; SPSS Japan Inc., Tokyo, Japan). The results were considered significant if the P<0.05.

Cell lines were lysed in RIPA buffer. After boiling with sodium dodecyl sulfate loading buffer, equal amounts (10 µg) of the proteins were electrophoresed on 10% sodium dodecyl sulfate-polyacrylamide gels (SDS-PAGE) and transferred to Immobilon membranes (Millipore Inc., Bedford, MA, USA) by semidry blotting. Five percent milk powder dissolved in PBS buffer containing 0.1% Tween-20 was used to block non-specific binding. Then, using standard techniques, the membranes were probed with the following antibodies: anti-E-cadherin antibody (mouse monoclonal antibody to human E-cadherin; clone: HECD-1; dilution: 1:200; Takara, Tokyo, Japan), anti-ZEB1 antibody (rabbit monoclonal antibody to ZEB1; clone: D80D3; dilution: 1:100; Cell Signaling Technology, Danvers, MA, USA), and anti-β-actin antibody (mouse monoclonal antibody to β-actin; clone: C4; dilution: 1:1,000; Santa Cruz Biotechnology, Dallas, TX, USA). Antibody-antigen complexes were detected using Immobilon Western chemiluminescent horseradish peroxidase substrate (Millipore), and signals were recorded on a LAS-3000 mini system (Fujifilm, Tokyo, Japan).

Human tissue specimens were obtained from 97 patients who underwent gastric cancer surgery at Tokyo Medical University Hospital between 2003 and 2010. The Tokyo Medical University institutional review board approved the present study, and all patients provided written informed consent. The patient cohort included 74 males and 23 females, ranging in age from 40 to 92 years (average 66.3 years). Other clinicopathological characteristics are listed in Table I. The tumors were cut into ~5-mm sections after fixation in 10% formalin and embedded in paraffin. From these paraffin blocks, tissue microarrays (TMAs) were constructed, each consisting of 12 specimens of 6-mm diameter on one slide. As previously described (29), TMAs allow the relocation of multiple tissue samples from conventional histologic paraffin blocks so that tissues from multiple patients can be analyzed on the same slide.

In situ hybridization (ISH) was performed on TMAs containing 3-µm consecutive sections of formalin-fixed, paraffin-embedded (FFPE) samples. Probes utilized were based on 5 digoxigenin (DIG)-labeled ISH locked nucleic acid (LNA) technology (miRCURY-LNA detection probe; Exiqon A/S, Vedbaek, Denmark). Slides were prepared using the Ventana HX System BenchMark (Ventana Medical Systems, Inc., Tucson, AZ, USA). We used the following 5 DIG-labeled LNA probes for ISH: hsa-miR-200c, 5-TCCATCATTACCCGGCAGTATTA-3; and non-target negative control sequence, 5-GTGTAACA CGTCTATACGCCCA-3.

Immunohistochemistry (IHC) was performed on TMAs containing 4-µm consecutive sections of FFPE gastric tissue samples. After deparaffinization and rehydration, endogenous peroxidases were blocked by incubation in 0.3% hydrogen peroxide solution for 20 min. To expose antigens, the sections were autoclaved in EDTA buffer (pH 9.0) for 10 min and cooled for 30 min. After rinsing in 0.05 M Tris-buffered saline containing 0.1% Tween-20 (pH 7.6), the sections were incubated with an affinity-purified anti-ZEB1 antibody (clone: D80D3; dilution: 1:200; Cell Signaling Technology) and anti-E-cadherin antibody (clone: HECD-1; dilution: 1:1,000; Takara) for 3 h at RT. Samples were then washed three times in PBS and incubated with Dako secondary antibody for 15 min at RT. Horseradish peroxidase-labeled polymer conjugated to a mixture of goat anti-mouse and anti-rabbit immunoglobulin antibodies (Code: K5007; prediluted; Dako, Glostrup, Denmark) was used as the secondary antibody. 3,3-Diaminobenzidine tetrachloride (DAB) was used for color development and sections were counterstained with hematoxylin.

ISH- or IHC-stained slides were examined by light microscopy using an Olympus BX50 microscope. Three different fields (×200) in each tissue were randomly selected. Digital images of ISH and IHC were captured by the NY-D5000 super system (Microscope Network; Nikon, Tokyo, Japan) and were printed by a true-color printer (IPSiO SP C420; Ricoh, Tokyo, Japan). Positive ISH results caused the cytoplasm to appear blue, whereas positive IHC results appeared brown. Nuclear immunostaining was assessed for ZEB1, while cytomembranous immunostaining was assessed for E-cadherin. Images were interpreted semi-quantitatively by assessing the extent and intensity of staining on the entirety of each tissue section present on the slides, as previously described (30). Briefly, the total percentage of positively-stained tumor cells was first determined. Then, the percentage of weakly-, moderately- and strongly-stained cells was determined, so that the sum of these categories equated with the overall percentage of positivity. A final staining score was then calculated as the sum of 1 × percentage of weak, 2 × percentage of moderate, and 3 × percentage of strong staining (maximum score=300).

Prior to statistical analyses, clinicopathlogical data as well as tissue image results were dichotomized into two groups of ‘high’ and ‘low’ levels, according to the median values. Thus, the final score for staining of miR-200c was used to define high-level expression (final score ≥201) and low-level expression (final score <201). Likewise, the final score for staining of ZEB1 was used to define high-level expression (final score ≥1) and low-level expression (final score <1), as well as that of E-cadherin used to define high-level expression (final score ≥111) and low-level expression (final score <111). Tumor diameter was dichotomized as either high-level (diameter ≥40 mm) or low-level (diameter <40 mm). High-age (≥65 years old) and low-age (<65 years old) groups were established according to convention, as were advanced (depth ≥mp) and early (depth <mp) cancer groups. Histologically, both a well/moderately differentiated tubular adenocarcinoma group and a poorly differentiated adenocarcinoma group were established. Then, statistical analyses including Chi-square test and multivariate analysis using logistic regression models were performed using the SPSS software package (ver17.0; SPSS). Results were considered significant if the P<0.05.

We investigated miR-200c expression in 6 gastric carcinoma cell lines by qRT-PCR analysis. As shown in Fig. 1A, the expression level of miR-200c descended in the following order (from high to low): Kato-III > MKN-45 > NU-GC-4 > H-111-TC > HGC-27 > MKN-1. Of note, the latter 2 cell lines HGC-27 and MKN-1 expressed very low levels of miR-200c. We also investigated the expression of particular target mRNAs, namely ZEB1 and ZEB2, along with E-cadherin, in the same cell lines. The expression level of ZEB1 descended in the following order (from high to low): HGC-27 > MKN-1 > NU-GC-4 > H-111-TC > MKN-45 > Kato-III (Fig. 1B). Of note, the 2 cell lines in which miR-200c expression level was very low exhibited the greatest expression level of ZEB1. Similarly, the expression level of ZEB2 descended in the following order (from high to low): HGC-27 > H-111-TC > NU-GC-4 > MKN-45 > MKN-1 > Kato-III (Fig. 1C). Here, one of the 2 cell lines in which miR-200c expression level was very low exhibited the greatest expression level of ZEB2. On the other hand, the expression level of E-cadherin descended in the following order (from high to low): H-111-TC > MKN-45 > NU-GC-4 > Kato-III > MKN-1 > HGC-27 (Fig. 1D). Of note, the 2 cell lines in which the expression level of miR-200c was very low exhibited the lowest expression level of E-cadherin. As shown in Table II, an inverse correlation was observed between miR-200c and ZEB1 (P<0.05; by Spearmans rank correlation).

The expression of ZEB1 and E-cadherin proteins was examined in the 6 human gastric carcinoma cell lines mentioned thus far by western blot analysis. As shown in Fig. 2, expression of ZEB1 was only detected in HGC-27 and MKN-1 cells, which exhibited the lowest expression level of miR-200c determined by qRT-PCR analysis. Conversely, the expression of E-cadherin was not detected in these 2 cell lines; instead, E-cadherin expression was detected in NU-GC-4 and H-111-TC cell lines, although very weak expression was also detected in Kato-III cells.

The results of qRT-PCR analyses and western blot analysis were in agreement. Cell lines with a high expression level of miR-200c exhibited low ZEB1/2 and high E-cadherin expression, whereas those with a low expression level of miR-200c exhibited high ZEB1/2 and low E-cadherin expression. Therefore, we investigated characteristics that illustrated differences between the group of cell lines with high miR-200c expression and those with low miR-200c expression. As shown in Fig. 3, we first compared the tumor origin (primary or metastatic tumor) of the cell lines. However, the tumor origin did not clearly correspond to the levels of miR-200c expression, as cell lines derived from metastatic tumors exhibited both low and high miR-200c expression levels. Secondly, we compared the histological differentiation state of the tumors from which the cell lines were derived was compared. Although cell lines from poorly differentiated tumors exhibited high miR-200c expression, the results were inconclusive as one of the cell lines with low miR-200c expression (HGC-27) was originally from a poorly differentiated tumor. Thirdly, we revealed that the morphology of the cell lines was well correlated with the expression level of miR-200c. Cell lines with high miR-200c expression (Kato-III, MKN-45 and NU-GC-4) were round-shaped, those with low miR-200c expression (HGC-27 and MKN-1) were spindle-shaped, and those with intermediate miR-200c expression (H-111-TC) exhibited a phenotype which was a mid-way between round- and spindle-shaped (polygonal or short spindle-shaped).

In human tissue samples, positive miR-200c expression, as determined by ISH, was denoted by cytoplasmic granular blue staining. Most of the normal tubules stained positive for miR-200c, whereas stromal tissue was not stained. As shown in Fig. 4, miR-200c expression patterns in gastric adenocarcinoma differed substantially between tubular adenocarcinoma and poorly differentiated adenocarcinoma. In tubular adenocarcinoma, miR-200c expression was observed with an intensity level either stronger than or the same as normal tubules. Conversely, in poorly differentiated adenocarcinoma, miR-200c expression was either not identified or observed with weaker staining patterns than normal tubules. The average staining score for miR-200c was 265±81.6 in tubular adenocarcinomas, while it was 104±75.2 in poorly differentiated adenocarcinomas.

Positive ZEB1 protein expression was denoted by nuclear staining. Overall positivity of ZEB1 was relatively low in normal tubules and gastric carcinomas. ZEB1 was almost negative in tubular adenocarcinomas; however, positive expression of ZEB1, although generally weak, was sporadically observed in poorly differentiated adenocarcinomas. The average staining score for ZEB1 was 0.84±2.93 in tubular adenocarcinomas, while it was 5.8±8.6 in poorly differentiated adenocarcinomas. Positive E-cadherin protein expression was indicated by cytomembranous staining. Normal tubules and tubular adenocarcinomas were strongly stained, while poorly differentiated adenocarcinomas were weakly stained (Fig. 4). The average staining score for E-cadherin was 146±52 in tubular adenocarcinomas, while it was 39±38.6 in poorly differentiated adenocarcinomas.

Use of the Chi-square test revealed a correlation between miR-200c and E-cadherin expression levels (P<0.001) and an inverse correlation between miR-200c and ZEB1 expression levels (P<0.001) as well as between ZEB1 and E-cadherin expression levels (P<0.001). We further compared the expression levels of miR-200c, ZEB1 and E-cadherin with clinicopathological factors (Table III). High expression levels of miR-200c were correlated with older age (≥65 years old) (P=0.002) and tubular histology (P<0.001). High expression levels of ZEB1 were correlated with poorly differentiated histology (P<0.001) and advanced stage (P=0.040). High expression levels of E-cadherin were correlated with older age (≥65 years old) (P=0.039) and tubular histology (P<0.001). In the multivariate analysis (Table IV), among the clinicopathological factors significantly correlated with the expression levels of miR-200c, ZEB1 and E-cadherin, there was a significant association between the expression level of ZEB1 and advanced stage (P=0.031), as well as a significant correlation between the expression level of ZEB1 and histology (P<0.001).