The RNA sequencing data of The Cancer Genome Atlas (TCGA) GC tissues and the corresponding clinical data were downloaded from TCGA database. Another three public GC RNA sequencing and microarray datasets (GSE54129, GSE79973, and GSE99416) were downloaded from the Gene Expression Omnibus (GEO). In additional, the scores between RP11-138J23.1 or VAV3 mRNA and HuR protein were predicted via RNA-Protein Interaction Prediction website (http://pridb.gdcb.iastate.edu/RPISeq/).

We obtained tissue microarray (TMA) containing 74 pairs of GC and normal gastric tissues from Outdo Biotech (Shanghai, China). RNA ISH was performed to detect RP11-138J23.1 expression in TMA using probe (BOSTER Biological Technology Co. Ltd., Wuhan, China). After dewaxing and rehydration through different concentrations of ethanol solutions, the TMA were digested with proteinase K at 37°C and hybridized with the digoxin-labeled probe 3–6 h at 55°C.

Human gastric normal epithelium cell line (GES-1) and GC cells (BGC-823, SGC-7901, and MGC-803) were purchased from Shanghai Institutes for Biological Science, China. BGC-823, SGC-7901 and MGC-803 cells were cultured in RPMI 1640 medium (KeyGene, Nanjing, China). GES-1 cells were cultured in DMEM medium (KeyGene, Nanjing, China) supplemented with 1% penicillin/streptomycin, l-glutamine, and 10% fetal bovine serum (FBS, Invitrogen, Carlsbad, CA, USA). Culture plates were incubated at 37°C in a humidified atmosphere with 5% CO².

The siRNA specifically targeting RP11-138J23.1, HuR and VAV3 was designed and synthesized by Invitrogen (Invitrogen, USA). The nucleotide sequences of all siRNAs were shown in Supplementary Table S1 . The cells were transfected with siRNA using RNAiMAX (Invitrogen, USA). In addition, RP11-138J23.1 overexpression plasmid and vectors containing shRNA targeting RP11-138J23.1 was constructed and transfected with XtremeGENE HP DNA transfection reagent (Roche, Basel, Switzerland).

Total RNA was extracted from cells or tissue samples by TRIzol reagent (Invitrogen, Carlsbad). The RNA quality was evaluated by Nanodrop ND-2000 spectrophotometer. According to the manufacturer’s instructions, 1 µg RNA was reverse transcribed into cDNA with the PrimeScript RT Reagent Kit (TaKaRa, Dalian, China). For detection of lncRNA and mRNA, quantitative reverse transcription-polymerase chain reaction (qRT-PCR) was performed using the SYBR Premix Ex Taq II (Perfect Real Time) on a 7500 Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The expression of β-actin was used for normalization. Related primers are listed in Supplementary Table S1 .

MTT assay was conducted for in vitro proliferation assay in accordance with the manufacturer’s instructions. GC cells, transfected with siRNA, scrambled siRNA, pcDNA3.1, or pcDNA3.1-RP11-138J23.1 vector for 48 h, were plated at a density of 1,500 cells/well in 96-well plates. Cell viability was documented every 24 h. The experiment was performed with three replicates, and six parallel samples were performed each time.

For colony-formation assay, a total of 700 transfected cells were suspended in a fresh six-well plate and maintained in media containing 10% FBS. After the formation of the colony, the plate was gently washed with PBS and fixed with methanol for 10 min, and then stained with 0.1% crystal violet (Sigma-Aldrich, America). Visible colonies more than 50 cells were manually counted. Triplicate wells were assessed for each treatment group.

Flow-cytometric analysis was carried out to evaluate cell apoptosis and cell cycle. Forty-eight hours after transfection, cells for apoptosis analysis were stained with propidium iodide (PI) and FITC-Annexin V following the protocol of the FITC Annexin V Apoptosis Detection Kit (BD Biosciences). The cells were then analyzed by FACScan. Cells were divided into viable cells, early apoptotic cells, late apoptotic cells, and dead cells. The ratio of early apoptotic cells and late apoptotic cells was measured.

For the flow-cytometric analysis of the cell cycle, the transfected cells were fixed in 75% ice-cold ethanol at −20°C overnight. Cells were then stained with propidium oxide by the CycleTEST PLUS DNA Reagent Kit (BD Biosciences). The percentage of G0/G1, S, and G2/M phases was analyzed with a flow cytometry (FACScan^®; BD Biosciences).

To determine the ability of migration and invasion, Transwell assays were conducted. For cell migration, 5 × 10⁴ cells were seeded into the upper chamber of Transwell assay (Millipore, Billerica, MA, USA) containing 200 μl of serum-free medium with a membrane (5 μm pores). Medium containing 10% FBS was filled into the lower chamber. For invasion assays, 1 × 10⁵ transfected cells were seeded into the upper chamber with Matrigel and the lower chamber were filled with medium containing 10% FBS. After 24 h, the cells on the lower membrane surface were fixed with 4% paraformaldehyde, stained with crystal violet, and photographed with a digital microscope. Finally, the cell numbers were imaged and counted in five random fields for each chamber.

Male BALB/c nude mice (4-week-old) were used for the tumor formation assay. Animal experiments were approved by the Institutional Animal Care and Use Committee of Zhejiang University (Hangzhou, China). BGC-823 cells transfected with shRNA targeting RP11-138J23.1 or empty vector were harvested and resuspended at a concentration of 1 × 10⁷ cells/ml. A total of 100 µl of suspended cells was subcutaneously injected into a single flank of each mouse (n = 6). Tumor growth was measured using calipers every week, and growth curves were plotted for each group. Tumor volumes were calculated using the equation: V = 0.5 × D × d ² (V, volume; D, longitudinal diameter; d, latitudinal diameter). Tumors were removed from all animals after 4 weeks, and tumor weights were measured. Parts of the tumors were further detected by the expression of Ki-67 via immunohistochemistry and RP11-138J23.1 by qRT-PCR.

The GC cells was lysed using lysis buffer containing the mammalian protein extraction reagents RIPA (Beyotime China), protease inhibitor cocktail (Roche, Basel, Switzerland), and PMSF (Roche). Followed by centrifugation at 13,000 rpm for 20 min, the concentration of total protein was determined using the Pierce™ BCA Protein Assay Kit (Thermo Scientific™, USA). Fifty micrograms of the protein lysates were electrophoresed on 10% SDS-polyacrylamide gel electrophoresis (SDS-PAGE), transferred to 0.22 μm NC membranes (Sigma), and incubated with specific antibodies. The specific bands were visualized using an ECL detection reagent (BioRad, USA) and quantified by densitometry (Quantity One software, BioRad). VAV3 antibody was purchased from Abcam (Cambridge, USA), and β-actin (Abcam) was used as a control.

Nuclear and cytosolic fractions were separated and purified using the PARIS Kit (Life Technologies, Carlsbad, CA, USA) according to the manufacturer’s instructions.

A Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, Billerica, MA, USA) was used for RNA immunoprecipitation (RIP) experiments following the manufacturer’s instructions. BGC-823 cells were lysed, and the cell extract was incubated with magnetic beads conjugated with anti-HuR and anti-IgG at 4°C. Immunoprecipitated RNA was purified and analyzed via qRT-PCR. Antibodies for RIP assays against HuR and IgG were purchased from Millipore.

GC cells were transfected with siRNAs or overexpression vectors for 48 h, and then treated with actinomycin D (1 μg/ml). For further determination of RNA and protein, cells were collected at different time points.

SPSS19.0 software (IBM, Chicago, IL, USA) was used for all statistical analyses, and differences were considered to be statistically significant with p < 0.05. The Student’s t-test (two-tailed) and Chi-square test were used to analyze differences between groups. All data were presented as the mean ± SD.

In order to identify dysregulated lncRNAs in human GC, we firstly downloaded and analyzed GC microarray gene-profile datasets from GEO (GSE99416, GSE54129, and GSE79973) and RNA sequencing datasets from TCGA ( Figures 1A–D ). As presented in Figures 1E, F , four datasets shared 2 and 1 lncRNAs that were upregulated and downregulated, with a ≥2.0-fold change relative to their corresponding normal counterparts. To elucidate the diagnostic, therapeutic, and prognostic roles of lncRNAs in GC, RP11-138J23.1, significantly increased in all four datasets, was chosen for further analysis. To confirm the reliability of the above bioinformatics analysis results, the level of RP11-138J23.1 was ascertained by in situ hybridization (ISH) in 74 paired clinical GC tissues and corresponding normal gastric tissues. We found that RP11-138J23.1 is mainly located in the cytoplasm and the positive lesions are manifested as red. The result confirmed the aberrant upregulation of RP11-138J23.1 in GC tissue compared with adjacent normal tissues ( Figures 2A, B ). Subsequently, survival analysis was performed in 285 stomach adenocarcinoma samples by using the bioinformatics tool “TANRIC”. As a result, high expression level of RP11-138J23.1 indicated a poorer prognosis in stomach adenocarcinoma patients (Cox p-value = 0.041) ( Figure 2F ). Overall, these findings revealed that RP11-138J23.1 might exert a crucial role in GC development.

To explore the clinical significance of RP11-138J23.1, we evaluated the correlation of RP11-138J23.1 expression level with clinicopathological parameters (i.e., tumor size, lymphatic metastasis, or TNM stage). According to the median value of RP11-138J23.1, 74 patients were further classified into high- and low-level groups. Highly expressed RP11-138J23.1 was found to exhibit a significant correlation with larger tumors, lymph node metastasis, and advanced stage (p < 0.05, Figures 2C–E and Table 1 ).

Additionally, using corresponding normal gastric tissues as control, we plotted a receiver operating characteristic (ROC) curve to investigate the diagnostic value of RP11-138J23.1 diagnosing GC. The cutoff value to differentiate the benign and malignant gastric tissues was 4 (ISH score). The specificity and sensitivity were 97.3% and 36.5%, respectively, and the AUC of the ROC curve was 0.68 ( Figure 2G ). These data demonstrated that RP11-138J23.1 may act as a diagnostic biomarker for human GC because of the high specificity.

The RNA expression levels of RP11-138J23.1 in the human GC cell lines (BGC-823, SGC-7901, MGC-803) and the normal gastric mucosa cell line (GES-1) were examined by RT-qPCR. Compared with GES-1, RP11-138J23.1 is shown as obviously more abundant in GC cell lines ( Figure 3A ). To explore the biological functions of RP11-138J23.1 in GS, we used two siRNAs to downregulate endogenous RP11-138J23.1 level in BGC-823 and SGC-7901 cells. Forty-eight hours posttransfection, RP11-138J23.1 expression levels were knocked down by more than 60% ( Figure 3B ). Meanwhile, RP11-138J23.1 was ectopically expressed with pcDNA3.1-RP11-138J23.1 vector in MGC-803 cells and the efficiency of transfection was 58.9-fold ( Figure 3C ).

When RP11-138J23.1 was silenced, cell viability assessed by MTT assays in BGC-823 and SGC-7901 cells decreased ( Figure 3D ). Consistently, the cell viability of MGC-803 in the pcDNA3.1-RP11-138J23.1 group exceeded that of the pcDNA3.1 group ( Figure 3E ). Colony-formation assay was conducted to evaluate the clonogenic survival potential of cells, and we obtained similar results ( Figures 3F, G ).

To further determine the impact of RP11-138J23.1 on GC cell growth via affecting cell cycle or apoptosis, FACS analysis was conducted. Suppression of RP11-138J23.1 led to an increase in the number of cells in the G0/G1-phase ( Figure 4B ). In addition, we also found that depletion of RP11-138J23.1 elevated seemingly higher levels of apoptosis in GC cells ( Figure 4A ). Unfortunately, it was not statistically significant.

In cancer malignant progression process, metastasis was an important biological feature. The results of Transwell assay demonstrated that RP11-138J23.1 knockdown inhibited the migratory and invasion capability of BGC-823 and SGC-7901 cells, while the capability of migratory and invasion observably increased once RP11-138J23.1 was overexpressed ( Figures 4C, D ). Together, these data indicated that RP11-138J23.1 plays an important role in GC metastasis.

To further estimate the oncogenic role of RP11-138J23.1 in vivo, we employed a mouse xenograft model. We established stable knockdown BGC-823 cells through a lentiviral vector harboring a short hairpin RNA targeting RP11-138J23.1 (sh-RP11-138J23.1-1 and sh-RP11-138J23.1-2). Cells stably transfected with either sh-RP11-138J23.1 or control empty vector were subcutaneously injected into nude mice. Indeed, tumors derived from sh-RP11-138J23.1-1 and sh-RP11-138J23.1-2 cells had significantly smaller tumor volumes and lower tumor weights compared with control empty vector mice ( Figures 5A–C ). The expression of RP11-138J23.1 was confirmed down-regulated in sh-RP11-138J23.1-1 and sh-RP11-138J23.1-2 group by qPCR analysis ( Figure 5D ). Meanwhile, compared with the control tumors, sh-RP11-138J23.1-1 and sh-RP11-138J23.1-2- derived tumors exhibited lower levels of the proliferation marker Ki-67 ( Figure 5E ).

Based on our preliminary evidence, we highlighted an important role of RP11-138J23.1 in human GC; however, the mechanism(s) governing the oncogenic role of RP11-138J23.1 in GC have yet to be elucidated. Knowledge of subcellular localization of lncRNAs might provide a clue about lncRNA mechanisms. RP11-138J23.1 was proved to be predominantly located in the cytoplasm via the prediction website (http://www.csbio.sjtu.edu.cn/bioinf/lncLocator/) ( Figure 6A ) and confirmed in GC tissues using ISH. Coincidentally, we performed a subcellular fractionation assay and determined that RP11-138J23.1 transcript is enriched in the cytoplasmic fraction in GC cells ( Figure 6B ). These data strongly support that RP11-138J23.1 has distinct cytoplasm localization and likely participates in posttranscriptional regulation.

To further investigate the underlying mechanisms of RP11-138J23.1, we predicted the interaction probabilities of RP11-138J23.1 and RNA-binding protein via the RNA-Protein Interaction Prediction website (http://pridb.gdcb.iastate.edu/RPISeq/). As presented in Figure 6C , the interaction probabilities of RP11-138J23.1 with FUS, HuR, and YBX1 were 0.7, 0.75, and 0.85 using RF classifier; meanwhile, these were 0.794, 0.877, and 0.818 using SVM classifier. Additionally, we also confirmed the combination of RP11-138J23.1 and HuR in CLIP-seq data. To validate the physical interaction between RP11-138J23.1 and HuR, we carried out the RIP analysis and confirmed that RP11-138J23.1 could bind with HuR in BGC-823 cells ( Figure 6D ). Further analysis showed that HuR and RP11-138J23.1 cannot regulate each other’s expression ( Figures 6E, F ). Based on these observations, we propose that RP11-138J23.1 may function as an oncogene through binding HuR and exert regulation role at post-transcriptional level.

In order to analyze the “RP11-138J23.1-HuR”-associated transcripts, we applied RNA sequencing and identified 852 genes being differentially expressed (fold-change >2, p < 0.05) after depletion of RP11-138J23.1, including 575 downregulated genes and 277 upregulated genes ( Figure 6G ). Gene Ontology (GO) and KEGG pathway analyses indicated that cell proliferation, cell growth, cell junction, and cell motility were involved in the affected functional processes in RP11-138J23.1-depeleted cells, which was consistent with the result of function assay ( Figure 6H ).

In a previous study, Xie et al. conducted high-solution transcriptome microarray in BGC-823 cells with HuR knockdown (13). We then explored potential targets that are regulated by “RP11-138J23.1-HuR” and found 139 genes ( Figure 6I ). Subsequently, qRT-PCR and Western blot assay validated a significant decrease of VAV3 observed after RP11-138J23.1 or HuR knockdown in BGC-823 and SGC-7901 cells ( Figures 6K, L ).

HuR is the ubiquitous member of a family of RBPs and involved in the regulation of mRNA stability (14). To better understand the impact of “RP11-138J23.1-HuR” on VAV3 mRNA stability, we treated BGC-823 and SGC-7901 cells with actinomycin D. Notably, qRT-PCR analysis showed that the half-life of VAV3 mRNA was significantly decreased in RP11-138J23.1- or HuR-downregulated GC cells ( Figure 7A ). Whereas, in RP11-138J23.1-overexpressed GC cells, VAV3 mRNA exhibited longer half-life ( Figure 7A ). Notably, the interaction probabilities of VAV3 mRNA with HuR were also predicted by RPISeq. The result showed 0.75 and 0.95 using RF classifier and SVM classifier, respectively ( Figure 7B ). The result of rescue assays also confirmed that RP11-138J23.1 exerts the mRNA stability-promoting effect depended on HuR protein. We found that the phenomenon of RP11-138J23.1-increased VAV3 mRNA half-life could eliminated by HuR knockdown.

The biological functions of VAV3 in GC cells were explored in the study of Xie et al., but whether VAV3 is involved in RP11-138J23.1-induced biology remains unclear. To determine whether RP11-138J23.1 exerts function via the HuR/VAV3 pathway, we conducted a rescue assay by manipulating RP11-138J23.1 and VAV3 expression levels ( Figures 7C, D ). The result from MTT, colony-formation, and Transwell assays showed that MGC-803 cells co-transfected with pcDNA3.1-RP11-138J23.1 and VAV3 siRNAs could partially reverse RP11-138J23.1-induced GC cell proliferation and metastasis ( Figures 7E–G ). Thus, we come to a conclusion that oncogenic activity of RP11-138J23.1 is partially attributable to its facilitation of VAV3 mRNA stability through association with HuR.