The human GC cell lines SGC7901, BGC823, MGC803, MKN28 and MKN45 and the normal gastric epithelium cell line (GES1) were obtained from the Chinese Academy of Sciences Committee on Type Culture Collection Cell Bank (Shanghai, China) and Fuheng Cell Center (Shanghai, China). All cell lines were authenticated by STR profiling. Cell lines were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (FBS) in a humidified atmosphere containing 5% CO₂ at 37°C.

We obtained 60 paired GC and adjacent non-cancer tissues from patients who underwent surgery at Nanjing First Hospital of Jiangsu Province in China between 2011 and 2012 and who were diagnosed with GC based on histopathological evaluation. No local or systemic treatment was administered to these patients prior to surgery. All collected tissue samples were preserved in RNA Transport (OMEGA, Norwalk, CT, USA) and immediately frozen at −80°C until required. The clinical characteristics of all patients are listed in Table I. The present study was approved by the Ethics Committee of Nanjing Medical University.

Total RNA was extracted from cultivated cells or tissue using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA) according to the manufacturer's protocol. RNA concentrations were determined spectrophotometrically at 260 nm. Total RNA (1 µg) was reverse transcribed to cDNA in a total volume of 20 µl. Quantitative PCR was performed with the following thermocycling conditions: 95°C 30 sec for initial denaturation, 95°C 5 sec, 60°C 34 sec for PCR reaction, 40 cycles, using a 7300 Plus Real-Time PCR System (Applied Biosystems; Thermo Fisher Scientific, Inc.) and SYBR Premix Ex Taq (Takara Biotechnology Co., Ltd., Dalian, Liaoning, China) as a DNA-specific fluorescent dye. Notably, cDNA synthesis of coding gene mRNA used PrimeScript^™ RT reagent Kit (cat. no. RR037A; Takara Biotechnology Co., Ltd.) and cDNA synthesis of lncRNA used PrimeScript RT Reagent kit with gDNA Eraser (cat. no. RR047A; Takara Biotechnology Co., Ltd.). Gene expression levels were calculated relative to GAPDH and β-actin using the 2^−ΔΔCt method (22). GAPDH and β-actin were used as endogenous controls. The primer sequences are listed in Table II.

PLZF (ZBTB16) (NM_006006) Human Tagged ORF Clone (cat. no. RG206745) and pCMV6-AC-GFP Tagged Cloning Vector (cat. no. PS100010) were purchased from OriGene Technologies, Inc. (Rockville, MD, USA). Plasmid vectors for transfection were prepared using DNA Midiprep Kits (Qiagen, Duesseldorf, NRW, Germany) and transfected into GC cells using Lipofectamine 2000 (Invitrogen; Thermo Fisher Scientific, Inc.). The siRNAs were synthesized by Invitrogen and transfected into GC cells using Lipofectamine 2000 according to the manufacturer's instructions. Cells were seeded at ~2–4×10⁵ cells/well in 6-well plates and transfected with 100 nmol/l siRNAs in Opti-MEM medium using Lipofectamine 2000 reagent. The cells were then allowed to reach ~50% confluence on the day of transfection. The cells were switched into medium containing 10% FBS after transfection for ~4–6 h and incubated for another ~24–48 h until extraction for the assessment of interfering efficiencies and targeted genes. All siRNA sequences are listed in Table III.

The cell proliferation ability was examined using the CellTiter 96^® AQueous One Solution Cell Proliferation Assay (MTS; Promega Corporation, Madison, WI, USA). Absorbance values were assessed at a wavelength of 570 nm. Colony formation assays were performed to monitor the cloning capability of GC cells. The wound healing assay was used to evaluate the cell migration capabilities. The Transwell assay was performed to assess invasion. Procedural details were previously described (23).

Four-week-old female athymic BALB/c nude mice were maintained under specific pathogen-free conditions, using a 10-h light/14-h dark cycle at temperature of 25±1°C and relative humidity of 40~60% with free access to food and water, and manipulated according to protocols approved by the Ethics Committee of Animal Experiments of the Nanjing Medical University. The mice were divided into two groups of the PLZF and the empty vector with 5 mice/group. PLZF and empty vector stably transfected SGC7901 cells were harvested and 10⁷ cells were subcutaneously injected into a single side of each mouse. Tumor sizes were measured by calipers and recorded every 3 days. The tumor volumes were calculated from the length (the longest diameter across the tumor) and width (the corresponding perpendicular diameter) using the following formula: π/6 × length × width². Twenty days later, the animals were sacrificed by carbon dioxide, and tumors were harvested and preserved at −80°C and fixed in formaldehyde for qRT-PCR and immunohistochemistry (IHC) staining, respectively.

Western blot analysis was performed according to standard protocols as previously described (24). The primary antibodies of H3K27me3 (cat. no. 9733), EZH2 (cat. no. 5246), PLZF (cat. no. 39784) and GAPDH (cat. no. 5174) and the secondary antibody of HRP-linked anti-rabbit IgG (cat. no. 7074) were purchased from Cell Signaling Technology, Inc. (Danvers, MA, USA). The primary antibodies were diluted at ~1:500-1,000 and incubated at 4°C overnight and the secondary antibody was diluted at 1:2,000 at room temperature for 2 h.

DNA is modified by sodium bisulfite treatment converting unmethylated, but not methylated, cytosines to uracil. Following the removal of bisulfite and completion of the chemical conversion, this modified DNA is used as a template for PCR. Primers were designed to discriminate between methylated and unmethylated alleles following bisulfite treatment and to discriminate between DNA modified (vs. unmodified) by bisulfite. The primer sequences of methylated and unmethylated PLZF are listed in Table II.

RNA-Protein Interaction Prediction (RPISeq) (http://pridb.gdcb.iastate.edu/RPISeq/) was used to analyze the interaction of ANRIL and PLZF. In the RPISeq website, the sequences of ANRIL and PLZF were copied into the designated spaces and submitted to the system, and then the interaction probability was revealed. Interaction probabilities generated by RPISeq range from 0 to 1. In performance evaluation experiments, predictions with probabilities >0.5 were considered ‘positive’, indicating that the corresponding RNA and protein are likely to interact.

The EZMagna RNA immunoprecipitation (RIP) Kit (EMD Millipore, Billerica, MA, USA) was performed based on the manufacturer's protocol. Cells were lysed in complete RIP lysis buffer (containing cOmplete™ Protease Inhibitor Cocktail, Roche, Branford, CT, USA). The lysis mixture was incubated with magnetic beads conjugated with specific antibodies or control IgG overnight at 4°C. Beads were washed and incubated with proteinase K to remove unconjugated proteins. Finally, RNAs were purified and subjected to qRT-PCR assay to analyze possible RNAs.

5-Aza (cat. no. S1200) and suberoylanilide hydroxamic acid (SAHA) (cat. no. S1047) were purchased from Selleck Chemicals (Houston, TX, USA) and dissolved in dimethylsulphoxide (DMSO), and used at 10 and 2 µM, respectively. Subconfluent cell cultures were treated with 5-Aza and SAHA for 48 h and then extracted for MPCR and western blot analysis.

All results were represented as the mean ± standard deviation (SD) from at least 3 independent experiments. Differences were determined by the two-tailed Student's t-test (comparison for two groups) and the one-way ANOVA (comparison for >2 groups) and the Tukey's post hoc test for multiple comparisons. Pearson's Chi-squared test was used to analyze the pathological features of PLZF expression in GC. Survival curves were drawn using a Kaplan-Meier survival plot and assessed using log-rank tests. Cox regression analysis was used to perform univariate and multivariate analyses of the clinicopathological factors for disease-free survival. All statistical analyses were performed using SPSS 19.0 software (IBM Corp., Armonk, NY, USA).

We evaluated PLZF expression in the normal human gastric epithelial cell line GES1 and in a panel of GC cell lines, and we found that PLZF had low expression in all the examined GC cell lines even after 40 cycles of amplification and western blot analysis, whereas it was expressed in the normal epithelial cell line at both the mRNA and protein levels (Fig. 1).

Overexpression of PLZF inhibits the proliferation, migration and invasion of GC cells. To determine the role of PLZF in GC, we monitored the effect of PLZF overexpression on GC cell lines. Fig. 2A revealed the efficiencies of PLZF overexpression. PLZF overexpression inhibited the viability of BGC823 and SGC7901 cells (Fig. 2B). Similarly, PLZF overexpression alleviated clone formation in BGC823 and SGC7901 cells (Fig. 2C). Transwell assay revealed that the number of invaded cells in the control group were higher than those in the PLZF overexpression group (Fig. 2D). In the wound healing assay, PLZF overexpression suppressed wound healing in both BGC823 and SGC7901 cells (Fig. 2E). Collectively, these findings indicated that increased PLZF expression in GC inhibited GC cell proliferation, migration and invasion.

In nude mouse xenografts, SGC7901 cells stably transfected with the PLZF plasmid were inoculated into the flanks of mice, and SGC7901 cells stably transfected with the empty vector were also inoculated into the flanks of mice as a control. The efficiencies of PLZF overexpression were assessed by qRT-PCR and presented in Fig. 3A. PLZF overexpression significantly inhibited tumorigenesis in vivo given that the tumor weight and size were significantly decreased compared with those of the controls (Fig. 3B and C). Furthermore, we detected stronger Ki-67 expression in tumors derived from control nude mice than in those derived from PLZF overexpression (Fig. 3D).

By acting as oncogenes or tumor suppressors, many lncRNAs contribute to cancer occurrence and development by epigenetically regulating gene expression (25). We screened a panel of lncRNAs to search for the associated lncRNA of PLZF expression and found that both ANRIL and HOTAIR knockdown recovered the expression of PLZF in BGC823 cells (Fig. 4A and B). Since ANRIL knockdown increased PLZF expression more than HOTAIR knockdown, we chose ANRIL to further study PLZF regulation. Furthermore, we confirmed the negative association between PLZF and ANRIL expression in BGC823 and SGC7901 cells; ANRIL knockdown increased the expression of PLZF in both BGC823 and SGC7901 cells (Fig. 4C and D).

To verify crosstalk between ANRIL and PLZF, we performed bioinformatics analysis and found that PLZF could possibly interact with ANRIL. In our analysis, the RF classifier was 0.7, and the SVM classifier was 0.99; both were larger than 0.5. In our study, efficiencies of PLZF overexpressed cell lines were presented in Fig. 5A and B. RIP experiment confirmed the interaction between ANRIL and PLZF in GC cells (Fig. 5C). Recent discoveries have marked lncRNAs as new important players in DNA methylation regulation (26). Therefore, we inferred that ANRIL may regulate PLZF methylation by binding to PRC2. MSP is a method for analyzing DNA methylation patterns in CpG islands, and it revealed that PLZF was methylated in BGC823 and SGC7901 cells, not in the normal cell lines GES1 (Fig. 5D). BGC823 and SGC7901 cells were treated with the methylation inhibitor 5-Aza, DNA methylation of PLZF was decreased (Fig. 5E), but PLZF expression was increased (Fig. 5F). In addition, the histone deacetylase (HDAC) inhibitor SAHA did not have the same effect, which suggested that PLZF was methylated, not acetylated (Fig. 5F).

It has been reported that ANRIL strongly enriches PRC2, which was confirmed (Fig. 6A) (27). To further investigate the relationship between PLZF and PRC2, we assessed the expression levels of EZH2 and H3K27me3. We found that EZH2 and H3K27me3 were highly expressed in GC cell lines compared to that in the normal cell line GES1, which suggested that EZH2 and H3K27me3 were negatively associated with PLZF (Fig. 6B and C). Moreover, upon EZH2 knockdown by siRNA transfection, PLZF expression was markedly increased (Fig. 6D and E), confirming that PLZF was negatively associated with EZH2 and H3K27me3. Therefore, we concluded that ANRIL regulated PLZF level through collaboration between DNA methylation and H3K27me3.

To investigate the effect of PLZF on EMT, we identified EMT markers in the PLZF-overexpressed GC cell lines BGC823/PLZF and SGC7901/PLZF. The results revealed that the epithelial marker E-cadherin was enhanced and that the mesenchymal markers N-cadherin, vimentin, Snail, Slug and ZEB1 were reduced compared to those in GC cells transfected with the empty vector (Fig. 7A and B). In GC patient tissue samples, we found that PLZF and E-cadherin expression levels in carcinoma tissues had parallel trends and were less than those in adjacent non-cancer tissues, while EZH2, SUZ12, vimentin and N-cadherin expression levels in carcinoma tissues were higher than those in adjacent non-cancer tissues (Fig. 7C and D). These results indicated that the level of PLZF expression contributed to mesenchymal-epithelial transformation (MET).

To validate the differential expression of PLZF in GC patients, we assessed the PLZF expression levels in tissue samples from 60 GC patients by qRT-PCR assay. Relative PLZF mRNA expression was assessed and normal adjacent non-cancer tissue was presented as 1. PLZF expression of cancer tissue was normalized to the adjacent non-cancer tissue, and a ratio <1 represented a low PLZF level, while a ratio of >1 represented a high PLZF level. The results revealed that PLZF expression was significantly lower in cancer tissues than in adjacent normal tissues (Fig. 7E). By analyzing clinicopathological factors, we found that the low PLZF expression was correlated with poor prognosis in GC patients (Fig. 7F). The PLZF levels were also correlated with histological differentiation, tumor invasion depth, tumor-node-metastasis (TNM) stage and lymphatic metastasis. No relation was found between PLZF expression and other factors, e.g., sex, age and distant metastasis (Table I). Multivariate analysis further revealed that PLZF expression could be regarded as a potential diagnostic biomarker for disease-free survival in patients with GC, as well as distant metastasis (Table IV).