Tissues were obtained from surgical resection in Shengjing Hospital of China Medical University, including 18 cases of normal gastric tissues (NGT), 17 cases of highly differentiated gastric adenocarcinoma (HDAC) tissues, and 16 cases of poorly differentiated gastric adenocarcinoma (PDAC) tissues. NGTs were obtained from normal gastric perforation or benign ulcer surgeries away from the lesion. Tumor specimens were collected from general gastric cancer radical surgeries. Fresh tissues were immediately stored in liquid nitrogen after surgical removal. All patients provided written informed consent. The experiment was approved by the ethics committee of China Medical University Shengjing Hospital.

Human gastric mucosa cell lines GEF-1, HFE-145, human gastric cancer cell lines MKN45, SGC7901, and human embryonic kidney cell line HEK-293T were obtained from the Shanghai Institutes for Biological Sciences Cell Resource Center. HFE-145, MKN45 cells, and HEK-293T cells were maintained in DMEM (Thermo Fisher Scientific Inc, MA, USA) containing 10% fetal bovine serum (FBS, Life Technologies Corporation, Paisley, UK). SGC7901 and GES-1 cells were cultured in RPMI-1640 (Sigma-Aldrich, St Louis, MO, USA) containing 10% FBS. Cells were incubated in a humidified incubator at 37°C with 5% CO₂ in an atmosphere of either normoxia (21% oxygen), or hypoxia (1% O₂, 5% CO₂, and 94% N₂ gas mixture) for 24 h.

HIF-1α or HIF-2α subunits were suppressed using siRNAs (Elvidge et al., 2006). The shRNA against human GAPLINC (18p11.31) were constructed. MKN45 and SGC7901 cell lines were transfected with GAPLINC shRNA, and named as shGAPLINC cells. For the negative control, empty vectors were used and named as NC. Opti-MEM I and Lipofectamine 3000 reagents (Invitrogen, CA, USA) were used according to the manufacturer's instructions. Stable cell lines were created through the selection by means of Geneticin (G418; Sigma-Aldrich, St Louis, MO, USA), G418-resistant clones were obtained after 3–4 weeks. The transfection efficacy was assessed by quantitative real-time PCR (qRT-PCR).

After extraction of total RNA using Trizol (Life Technologies Corporation), RNA concentration and quality were evaluated by NanoDrop spectrophotometer (ND-100, Thermo, USA). SYBR Premix Ex Taq assays of GAPLINC and GAPDH (Applied Biosystems, Foster City, CA, USA) were used for GAPLINC qRT-PCR detection. For each assay, a standard curve of threshold cycle (C_T) value vs. log input standard cDNA was constructed. Fold changes were normalized and calculated using the relative quantification (2^−ΔΔCt) method.

HIF-1α, HIF-2α content was detected by western blot. Cell homogenate protein was prepared using loading buffer, RIPA lysis buffer as well as protease inhibitor (Beyotime Institute of Biotechnology, Jiangsu, China). Cell homogenate was then boiled for 5 min. Samples were centrifuged at 17,000 rpm, 4°C. The concentrations of proteins were measured using the BCA protein assay kit (Beyotime Institute of Biotechnology).

Protein samples of interest went electrophoresis and then were transferred to polyvinylidene fluoride membrane. Membrane was blocked in TBST (Tween Tris-buffered saline) buffer containing 5% low fat milk powder, then incubated with primary antibodies respectively at 4°C overnight. Primary antibodies included HIF-1α (diluted at 1:1000, Abcam, Cambridge, UK), HIF-2α (1:500, Abcam), Bcl-2 (1:2000, Proteintech Group, Chicago, USA), Bax (1:3000, Proteintech Group), CD44 (1:1000, Proteintech Group), TGFBR2 (1:500, Abcam), GAPDH (1:5000, Cell Signaling, Beverly, MA, USA). Then, membranes were incubated with corresponding HRP-polymerization secondary antibodies diluted at 1:5000 for 2 h at room temperature. Protein bands were visualized by ECL detection system, and ChemImager5500V2.03 scanning software. The integrated light density values (IDV) were calculated with software FluorChem2.0.

CCK8 reagent (Beyotime institute of biotechnology) and 5-ethynyl-2′-deoxyuridine (EdU) incorporation assay kit (Life Technologies Corporation) were used for cell proliferation ability detection. For CCK8 assay, MKN45 and SGC7901 cells were seeded in 96 well-plates (Corning, NY, USA) at the logarithmic growth phase, at a density of 2000/well each group. Two days later, 10 μl of CCK8 reagent was added into the culture medium and then incubated for 2 h. The absorbance at 450 nm was recorded.

For EdU incorporation assay, cells were cultured in medium with 10 μM of EdU each well and incubated for 2 h at 37°C. After fixing with 4% formaldehyde, cells were incubated with a Click-iT EdU Kit (Life Technologies Corporation), then stained with Hoechst 33342 and visualized using a fluorescence microscope (Olympus, Tokyo, Japan). The EdU incorporation rate was measured using the ratio of EdU positive cells (green cells) to total Hoechst 33342 positive cells (blue cells). Scale bar represented for 20 μm.

Cell apoptosis rate was measured using the Annexin V-PI/FITC kit (BD Biosciences, NJ, USA). Cells were resuspended into single cell suspension after washed with ice cold PBS. Then cells were collected with centrifugation and stained with Annexin V-PI and FITC for 15 min away from light. After that cells went for flow cytometry (FACScan, BD) detection, total cells and apoptotic cells which were stained with fluorescent dye were counted. Apoptosis ratio was analyzed using BD Accuri C6 software.

In cell migration and invasion assays, the 24-well plates and transwell chambers with 8 μm size pore polycarbonate membranes (Corning, NY, USA) were used. Cells were re-suspended in serum-free medium, and then seeded on the upper chamber. Matrigel solution (BD, Franklin Lakes, NJ, USA) was precoated before cell seeding for cell invasion assay. The serum-containing medium in the 24-well plates induced cells to migrate and invade. After incubated at 37°C for 48 h, the cells on the upper chamber were stripped off with a swab. Then cells on the lower membrane were fixed with the mixture of glacial acetic acid and methanol (3:1), and then stained with 20% Giemsa solution overnight. Five vision fields were randomly chosen and cells number was recorded for each well.

By screening the promoter region of GAPLINC, two binding motifs (CACGC; Ren et al., 2014 and ACGTG) of HIF-1α were found. To determine the responsive HIF-1α-binding sites in the human GAPLINC promoter, promoter activities were measured using Dual-Luciferase Reporter Assay System. Human full-length HIF-1α sequence was cloned into pEX3 vector (GenePharma). pEX3-HIF-1α and its empty vector were transfected into MKN45 and SGC7901 cells. Renilla promoters were co-transfected as an internal control. Firefly luciferase activity was normalized to renilla luciferase activity for each individual analysis.

Chromatin immunoprecipitation (ChIP) assay was carried out with Simple ChIP. Enzymatic Chromatin IP Kit (Cell Signaling Technology, Danvers, Massachusetts, USA). Cells were cross-linked with 1% paraformaldehyde, and then harvested in lysis buffer containing phenylmethane sulfonyl fluoride. Chromatin was cleaved into fragments (200–1000 bps). The 1000 bps upstream region of the putative HIF-1α binding sites was used as a negative control in ChIP assay. Lysates (2%) were used as input control. The rest lysates were immunoprecipitated with normal mouse IgG or HIF-1α antibody.

The DNA was extracted and amplified by PCR. The forward and the reverse primers were: PCR1 Forward: 5′-CAACCCTAATGGG ACGTGTTA-3′, Reverse: 5′-CAAACAGAATCA CCCCATGA-3′; PCR2 Forward: 5′-TAATCCCAG CTACTTGGGAGG-3′, Reverse: 5′-ATTCTTGCC CAGGCTGGAAT-3′, which covered a potential HRE (hypoxia response elements). The primers of the control region were: Forward: 5′-GGAAAATAA AACCATGGAGGG-3′ Reverse: 5′-CGAGAT TGAGCCACTGCACT-3′.

For the in vivo study, xenograft experiments were performed in balb/c nude mice. Mice were purchased from BEIJING HFK BIOSCIENCE Co. LTD and raised under specific pathogen-free condition, under the guidance of the laboratory animal center guideline. All process involving animals was subjected to approval by the Research Animal Care and Use Committee of Shengjing Hospital. Cells were subcutaneously implanted into the right flanks of mice, at density of 1 × 10⁶ cells/mouse. Every 5 days, tumor volumes were measured. Once tumors reached 150 mm³, either bevacizumab (10 mg/kg every 3 days) or vehicle control PBS (Phosphate Buffered Saline) were intraperitoneal injected to the mice. Tumor volume (mm³) = length × width²/2.

All data were presented as means ± standard deviation (SD) and derived from at least three independent experiments. All statistical analyses were performed using SPSS 18.0 statistical software with the Student's t-test. Probability P < 0.05 was considered as statistically significant. The spearman rank correlation test was used for the association between different variables analyzation. For survival analysis, Mantel-cox test was used.

A recent study reported that the lncRNA GAPLINC is upregulated in human GC (Hu et al., 2014). Assessment of HIF-1α expression showed that it was markedly upregulated under conditions of hypoxia in GC specimens (Figure 1). Immunoblot analysis showed that HIF-1α was highly expressed in gastric adenocarcinoma, especially in poorly-differentiated tumors, compared with NGT (P < 0.05, Figure 1A). qRT-PCR results showed that the pattern of expression of GAPLINC was similar to that of HIF-1α (P < 0.05, Figure 1B). Spearman correlation analysis showed a positive correlation between GAPLINC and HIF-1α expression (R = 0.6041, P < 0.05, Figure 1C). We therefore examined the association between HIF-1α and GAPLINC in further detail. GAPLINC was expressed at higher levels in the GC cell lines MKN45 and SGC7901 than in the gastric mucosa cell lines GES-1 and HFE-145 (Figure 1D). Overall survival analysis indicated that GAPLINC overexpression was associated with poor prognosis in GC patients (P < 0.05, Figure 1E).

Accumulating evidence indicates that the hypoxic microenvironment provides an important niche for drug-resistant cells (Brown, 1990; Trédan et al., 2007; Lin et al., 2011). HIF, as a finely regulated transcriptional factor, could stably exist in the hypoxic environment in MKN45 and SGC7901 cells (Figure 2A). To investigate the effect of hypoxia on GAPLINC expression, GAPLINC levels were monitored under conditions of normoxia and hypoxia. GAPLINC expression was gradually upregulated in both cell lines in a hypoxic atmosphere, and after 24 h the difference became significant (Figure 2B, P < 0.01). Knockdown of the two active subunits of HIF (HIF-1α and HIF-2α) did not affect GAPLINC expression under normoxia, whereas GAPLINC was markedly downregulated under hypoxia. The reduction was more obvious in the siHIF-1α group (Figures 2C,D).

The results above prompted us to examine the effect of the HIF-1α-GAPLINC axis on tumor growth. GAPLINC was knocked down in MKN45 and SGC7901 cells using two shRNAs (Figure 3). We next assessed cell proliferation, migration, invasion, and apoptosis rate. The CCK8 assay was performed out to examine the functional role of GAPLINC in cell proliferation. Cells were transfected with GAPLINC shRNA or empty vector. Under normoxic conditions (Figure 4A), cell proliferation was decreased in the shGAPLINC group compared with that in the shGAPLINC-NC group in the MKN45 and SGC7901 cell lines (P < 0.05). This suggested that GAPLINC is a carcinogenic factor involved in regulating GC viability. Under hypoxia, the viability of GC cells was inhibited; however, the shGAPLINC group showed a lower survival rate than the shGAPLINC-NC group (P < 0.01), suggesting that GAPLINC plays an important role in sustaining cell survival under hypoxia. EdU incorporation assay results supported this phenomenon (Figure 4B). The proportion of positive cells (green) among total cells (blue) represented the mitotic cell population. Positive cell numbers were lower in shGAPLINC groups than shGAPLINC-NC groups (P < 0.05).

GAPLINC played a vital role in inhibiting cell apoptosis according to the cell apoptosis assay (Figure 4C). The shGAPLINC groups showed higher apoptosis rates than the shGAPLINC-NC groups under normoxia or hypoxia. Under hypoxic conditions, GAPLINC knockdown resulted in more cell death in MKN45 cells (P < 0.01).

The cell migration and invasion assays showed similar results (Figures 5A–D and Figure S1). GAPLINC knockdown attenuated cell mobility in 12 and 48 h after seeding. According to the results, cell migration and invasion abilities relied more on GAPLINC under hypoxia than normoxia. These results clearly revealed that GAPLINC could sustain cell survival especially under poor aerobic conditions.

We speculated that there may be a direct interaction between GAPLINC and HIF-1α, the most important factor induced by hypoxia. HIF-1α binds the “ACGTG” motif according to JASPAR prediction (Figure 6A) and the “CACGC” sequence (Ren et al., 2014), which is characteristic of hypoxic response elements. Screening of the promoter region of GAPLINC identified two potential binding regions for HIF-1α (−320/−324 and −578/−582). To further investigate the mechanism underlying the HIF-1α mediated regulation of GAPLINC expression, a luciferase assay was performed (Figure 6B). The position of the transcription start site (TSS) was predicted by DBTSS. Putative and wild-type HIF-1α binding sites were constructed. Cotransfection with pEX3-HIF-1α resulted in the activation of the GAPLINC promoter containing two HREs by 4.3- and 3.7-fold in MKN45 and SGC7901 cells, respectively. For the GAPLINC promoter containing one HRE, the activation was 2.5- and 2.8-fold in MKN45 and SGC7901 cells, respectively.

To confirm that HIF-1α binds to the HREs of the GAPLINC promoter, a ChIP assay was performed (Figure 6C). Pull down with an anti-HIF-1α antibody identified the PCR fragments of two HREs in the GAPLINC promoter. No PCR fragment was detected in samples pulled down by control IgG antibody. No interaction with the negative control region was observed. Taken together, the results of luciferase and ChIP assays demonstrated that HIF-1α could bind to the GAPLINC promoter and upregulate the promoter activities in human MKN45 and SGC7901 cells.

Despite reports that GAPLINC increases GC invasiveness by rescuing CD44 from miR-211, we next examined functional proteins related to tumor growth and invasiveness. Since GAPLINC can absorb miR-211, which increases Bcl-2, and Bcl-2 abrogation upregulates the apoptotic protein Bax, Bcl-2, and Bax expressions were detected. As shown in Figure 7A, GAPLINC ablation upregulated Bcl-2, increased the CD44 content, and downregulated Bax in MKN45 and SGC7901 cells under hypoxia (P < 0.05). This could partly account for the phenomenon that GAPLINC rescues cell viability and invasiveness under poor oxygen conditions.

To further confirm the hypoxic induction of GAPLINC, an in vivo xenograft model was used. Nude mice were raised under pathogen free condition, and then MKN45 and SGC7901 cells were inoculated at a density of 10⁶ subcutaneously. Cells then grew as xenografts. Mice were treated with either bevacizumab to increase tumor hypoxia or PBS as vehicle control (Choudhry et al., 2015). As shown in Figure 7, bevacizumab administration upregulated GAPLINC RNA levels in xenografts derived from both MKN45 and SGC7901 cells. As a positive control, HIF-1α levels were measured in xenograft tumors from both cell lines. HIF-1α protein levels were increased after bevacizumab treatment along with GAPLINC. Taken together, these results indicated that GAPLINC was induced by hypoxia both in vivo and in vitro.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer TL and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review.