A total of 18 paired human GC samples and their matched gastric normal tissues (NTs) were collected at the time of surgical resection at the Fifth People's Hospital of Shanghai, Fudan University (Shanghai, China) between February 2017 and February 2018. These samples were from 13 males and 5 females, with a median age of 64 (range 52–86). Patients were included in the study if they were initially diagnosed with GC, underwent the surgery and had complete clinicopathological information. Those who had extensively metastatic tumors, suffered from life-threatening diseases such as severe cardiovascular disease or had other types of tumors besides GC were excluded from the study. Samples were snap-frozen in liquid nitrogen and stored at −80°C. Paraffin-embedded tissues were retrieved from the Tissue Bank of the Fifth People's Hospital of Shanghai, Fudan University, and 4-µm tissue sections were prepared by the Department of Pathology of the same hospital. The present study was approved by the Institutional Ethics Committee at the Fifth People's Hospital of Shanghai, Fudan University (ethical approval form no. 2017-097) and adhered to the principles in the Declaration of Helsinki. Written informed consent was obtained from each patient prior to tissue collection for experimentation.

Microarray sections of GCs and neighboring NTs were prepared by Shanghai Outdo Biotech Co., Ltd. These sections contained 90 paired GC and NTs from patients [the tissue microarray (TMA) cohort] and the clinicopathological characteristics of these patients are summarized in Table I. Following deparaffination, rehydration in graded ethanol, antigen retrieval with citrate buffer pH 6.0 (1:300 dilution; cat. no. ZLI-9065; OriGene Technologies, Inc.) and blocking with goat serum (1:20 dilution; cat. no. C0265; Beyotime Institute of Biotechnology) at room temperature for 1 h, slides were stained with a rabbit polyclonal antibody against human TAFA5 (1:50 dilution; cat. no. 13948-1-AP; ProteinTech Group, Inc.) at 4°C overnight. Normal rat immunoglobulin G (1:50 dilution; cat. no. D110504; Sangon Biotech Co., Ltd.) instead of the primary antibody was used as a control. Subsequently, after washing with PBS, a horseradish peroxidase (HRP)-conjugated secondary antibody (1:2,000; goat anti-rabbit, cat. no. A0208 and goat anti-rat, cat. no. A0192; Beyotime Institute of Biotechnology) was added and incubated at room temperature for 1 h. Then, these sections were stained with 3,3′-diaminobenzidine (1:25 dilution; cat. no. GK500705; Gene Tech Co., Ltd.) at room temperature for 5 min and counterstained with 100% hematoxylin (cat. no. C0107; Beyotime Institute of Biotechnology) at room temperature for 2 min. A modified H-score system was used to semi-quantitate TAFA5 expression, as previously described (14). Briefly, the maximal intensity of staining (0, negative; 1, weak; 2, moderate; and 3, strong) was multiplied by the percentage of positive tumor cells (0–100%) to generate the modified H-score (range: 0-300). TAFA5 staining was categorized as high or low expression using the median H-score.

The mRNA expression of TAFA5 in GC tissues and normal mucosae was acquired from Oncomine (www.oncomine.org) and Gene Expression Omnibus (GEO; accession no. GSE79973). The Cancer Genome Atlas (TCGA) Stomach Adenocarcinoma (STAD) RNA-sequence data and corresponding phenotype data were downloaded from UCSC Xena (https://xenabrowser.net/datapages/). The results in the present study are in whole or part based upon data generated by the TCGA Research Network (cancergenome.nih.gov). The expression of TAFA5 in TCGA STAD was further validated by two online analyzing tools specifically designed for TCGA data analysis (gepia.cancer-pku.cn/detail.php?gene=FAM19A5 and ualcan.path.uab.edu/cgi-bin/TCGAExResultNew2.pl?genenam=FAM19A5&ctype=STAD). The original data for the prognostic analysis of TAFA5 were downloaded from Kaplan-Meier Plotter (www.kmplot.com) and UCSC Xena (xenabrowser.net/heatmap/).

A gastric epithelial cell line (GES-1) and GC cell lines (AGS, HGC27, NCI-N87, MGC803 and SNU-1) were obtained from the Type Culture Collection of the Chinese Academy of Sciences. All of the cells were validated by short tandem repeat DNA profiling and were confirmed negative for Mycoplasma contamination. Cells were cultured in RPMI-1640 (BBI Life Sciences Corporation) or F12K medium (Zhong Qiao Xin Zhou Biotechnology Co., Ltd.) supplemented with 10% fetal bovine serum (cat. no. E600001; Sangon Biotech Co., Ltd.), 100 µg/ml penicillin, and 100 mg/ml streptomycin at 37°C with 5% CO₂ in a humidified incubator (Thermo Fisher Scientific, Inc.).

Construction of TAFA5-targeting short hairpin (sh)-RNAs and packaging of lentiviruses. Two targeting shRNAs (shTAFA5-1 and shTAFA5-2) and a nontargeting scrambled RNA (scramble) were subcloned into the GV248 lentivirus vector by Shanghai GeneChem Co., Ltd. The shTAFA5-1 target sequence was 5′-CGCAAGAATCATCAAGACCAA-3′, and the shTAFA5-2 target sequence was 5′-CACCTGTGAGATTGTGACCTT-3′. Lentiviral stocks were prepared, as previously described (15).

Total RNA was isolated from cell cultures or from snap-frozen tissues from patients with GC using RNAiso Plus (Takara Bio, Inc.), according to the manufacturer's instructions, and reverse transcribed with HiScript Q Select RT SuperMix (cat. no. R132-01; Vazyme Biotech Co., Ltd.), according to the manufacturer's protocol. qPCR was performed with SYBR Green Master Mix (cat. no. Q131-02; Vazyme Biotech Co., Ltd.) and quantified using the 2^−ΔΔCq method (16). The thermocycling conditions were as follows: 95°C for 30 sec, followed by 40 cycles of 95°C for 10 sec, 60°C for 32 sec, 95°C for 15 sec, 60°C for 60 sec and 95°C for 15 sec. Each sample was assayed in duplicate. All PCR products were confirmed by 2.0% agarose gel electrophoresis. The sequences of the primers were as follows: TAFA5, forward 5′-GTGAGTTTGGCCACTCCGTA-3′ and reverse 5′-GAATGGACAGATGGCTGGCA-3′; and GAPDH, forward 5′-GTCAAGGCTGAGAACGGGAA-3′ and reverse 5′-AAATGAGCCCCAGCCTTCTC-3′. Experiments were repeated three times in duplicate.

Total protein was extracted from cell cultures or homogenized tissues from patients with GC using radioimmunoprecipitation assay lysis buffer (Beyotime Institute of Biotechnology) containing phenylmethylsulfonyl fluoride (Beyotime Institute of Biotechnology) and proteinase inhibitor cocktail solution (Roche Diagnostics), and quantified via the bicinchoninic acid assay method, as recommended by the manufacturer. Western blotting was performed according to standard methods as previously described (17). Briefly, proteins (30 µg) were separated by 10% SDS-PAGE and then transferred onto a polyvinylidene fluoride membrane. Subsequently, the membranes were blocked with 5% fat-free dry milk at room temperature for 1 h. Then, the blots were incubated with a rabbit polyclonal antibody against TAFA5 (1:1,000 dilution; cat. no. 13948-1-AP; ProteinTech Group, Inc.) and monoclonal antibodies against E-cadherin, N-cadherin, Vimentin and ZEB-1 (1:1,000 dilution; cat. nos. 3195, 13116, 5741 and 3396, respectively; Cell Signaling Technology, Inc.) overnight at 4°C. GAPDH (1:2,000; rabbit anti-human; cat. no. AF1186; Beyotime Institute of Biotechnology) or β-actin (1:2,000; mouse anti-human; cat. no. AF0003; Beyotime Institute of Biotechnology) were detected as the loading controls. The membranes were washed with TBST and incubated with the respective HRP-conjugated secondary antibodies (1:2,000; goat anti-mouse cat. no. A0216 and goat anti-rabbit cat. no. A0208; Beyotime Institute of Biotechnology) at room temperature for 1 h. The proteins were finally visualized using an enhanced chemiluminescence system (cat. no. P0018AS; Beyotime Institute of Biotechnology). The grayscale values of protein bands were analyzed using ImageJ software (National Institutes of Health).

Stably infected AGS and HGC-27 cells (2×10³ cells/well) were seeded in 96-well plates and cultured for 24, 48, 72 or 96 h. Then, 10 µl Cell Counting Kit-8 (CCK-8) reagent (10% v/v in serum-free RPMI-1640 medium; Beyotime Institute of Biotechnology) was added to each well and incubated at 37°C for 1 h. Absorbance was measured at 450 nm using a microplate reader (BioTek Synergy 2; BioTek Instruments, Inc.).

A Transwell migration assay was performed as previously reported (18). Briefly, cells (4×10⁵ cells/ml) were seeded in serum-free RPMI-1640 or F12K medium in the top chamber of the Transwell insert. Medium containing 20% FBS in the lower chamber served as chemoattractant. Following incubation for 24 h at 37°C, the cells on the top side of the membrane were removed with a cotton swab, and the migrated cells on the bottom side were fixed with methanol for 20 min and then stained with crystal violet (0.1% in PBS) for 15 min. A total of 6 randomly selected fields per well were imaged under an inverted microscope (Leica Microsystems GmbH) using the ImageScope software (Leica Microsystems GmbH) with a magnification of ×200. Then cells were manually counted and analyzed by Excel 2010 (Microsoft Corporation) and the number of migrated cells was counted.

A monolayer scratch wound assay was employed to evaluate cell migration, as previously described (19). Briefly, cells (4×10⁵ cells/well) were seeded in 12-well plates and grown to nearly 100% confluence in 10% FBS-containing medium, before being washed with PBS and transferred to serum-free medium for 24 h. Then, a scratch wound was generated with a 200 µl pipette tip. Wound closure was imaged at 0, 24 and 48 h under an inverted microscope (Leica Microsystems GmbH) using the ImageScope software (Leica Biosystems GmbH) with a magnification of ×40. Then the wound closure rates were analyzed by ImageJ (v1.8.0; National Institutes of Health).

The Gene set enrichment analysis (GSEA) is developed by the Broad Institute website (http://www.broadinstitute.org/gsea/index.jsp). All GSEA analyses were performed using the GSEA 2.2.2 software. The TCGA STAD RNA-seq data were loaded into GSEA and run with the latest version of Hallmarks gene sets. The number of permutations was set at 1,000, and patient data were divided into TAFA5 high and low groups by the median of TAFA5 expression. Then GESA was performed according to the default weighted enrichment statistics and genes were ranked by the Pearson method. Gene sets were considered significantly enriched if P<0.05 and FDR < 0.25. Those gene sets that were significantly enriched in the TAFA5 high group were designated as TAFA5-positively related gene sets, while those that were significantly enriched in the TAFA5 low group were designated as TAFA5-negatively correlated gene sets.

Analyses were performed using Microsoft Excel 2010 (Microsoft Corporation), GraphPad Prism7 (GraphPad Software, Inc.), and SPSS version 22 (IBM Corp.). Student's t-test was used for statistical comparisons between two groups, while statistical comparisons among multiple groups were performed by one-way analysis of variance followed by Bonferroni's post hoc test. Pearson's χ² test and Fisher's exact test were used for categorical comparisons. Survival analyses were conducted using the Kaplan-Meier method. Univariate and multivariate survival analyses were conducted with the Cox proportional hazards regression model. All statistical tests were two-sided. P<0.05 was considered to indicate a statistically significant difference.

By analyzing the TAFA5 expression profile data from a GEO GC dataset (GSE79973), the present study revealed that TAFA5 was upregulated in GC tissues compared with NTs (n=10 pairs; P<0.001; Fig. 1A). RNA sequencing data from TCGA STAD also revealed an increase in TAFA5 expression in GC tissues (n=415) compared with in NTs (n=35; P<0.001; Fig. 1B), which was further validated by online analysis tools Gepia and Ualcan (data not shown). Two additional GC datasets [Derrico gastric(GEO: GSE13911), n=69; and Cho gastric(GEO: GSE13861), n=90](20,21) with Lauren's classification in the Oncomine database indicated that TAFA5 was generally overexpressed in each cancer group when compared with the control group, although the differences were not significant in three cases with smaller samples (Fig. 1C and D). To further validate these observations, the present study measured TAFA5 expression in 18 GC tissues and their matched NTs by RT-qPCR and western blotting. The results indicated that TAFA5 mRNA and protein expression levels were increased in GC tissues compared with NTs (Fig. 1E-H). Taken together, these results indicated that TAFA5 was upregulated in GC.

To test the hypothesis that TAFA5 protein abundance in GC contributes to increased malignancy, the present study examined the TAFA5 expression levels in primary GC tissues of a TMA cohort by IHC. As shown in Fig. 2A, TAFA5 was mainly expressed in the cytoplasm of cancer cells, and its expression was stronger in GC tissues than in NTs. The H-score was then used to quantitate TAFA5 staining in these tissues, which revealed that TAFA5 expression was significantly increased in GC compared with NTs (Fig. 2B). The associations between TAFA5 expression and clinicopathological variables are summarized in Table II. High TAFA5 expression was significantly associated with poor differentiation, and worse tumor, nodal and metastasis stages.

Next, the present study determined the association between TAFA5 expression and patient outcomes. Kaplan-Meier survival analysis of the TMA cohort revealed that patients with high TAFA5 expression exhibited a shorter overall survival (OS) compared with those with low expression [estimated mean OS 30.5 months with 95% confidence interval (CI) 22.9–38.1 months vs. OS 53.0 months with 95% CI 44.5–61.6 months; log-rank test, P=0.001; Fig. 2C]. To further confirm the adverse prognostic role of TAFA5 in patients with GC, the present study downloaded and analyzed TAFA5 transcription data from TCGA and Kaplan-Meier Plotter datasets. The results demonstrated that high TAFA5 expression was correlated with worse OS and recurrence-free survival (RFS) in both patient cohorts (Fig. 2D-G). Multivariate analysis using a Cox proportional hazards model revealed that TAFA5 overexpression was significantly associated with shorter OS [hazard ratio (HR), 1.90; 95% CI, 1.04–3.44; P=0.036], following adjustments for tumor size, tumor stage and nodal stage (Table III). Taken together, these results indicated that TAFA5 was a prognostic marker in GC that was associated with unfavorable prognoses.

To gain insight into the potential role of TAFA5 in GC progression, the present study first examined the intrinsic expression levels of TAFA5 in a normal gastric epithelial cell line, GES-1, and in five GC cell lines, via RT-qPCR and western blotting. The results indicated that TAFA5 expression was markedly increased in GC cells when compared with GES-1 cells (Fig. 3A and B). TAFA5 expression was then silenced in HGC-27 and AGS cells with lentiviruses carrying two TAFA5-specific shRNAs (Fig. 3C and D). In cultured GC cells, CCK-8 assays revealed that knockdown of TAFA5 expression significantly inhibited proliferation (Fig. 4A and B). In addition, knockdown of TAFA5 expression decreased cell migration in the scratch wound healing assay (Fig. 4C and D) and in the Transwell migration assay (Fig. 4E and F). Collectively, these results indicated that TAFA5 promoted the proliferation and migration of GC cells.

To characterize the potential mechanism of TAFA5 in promoting tumor progression, the present study used the RNA-seq data for GC samples from TCGA to conduct gene set enrichment analysis (GSEA) (22,23). The results revealed that TAFA5 expression was positively correlated with genes involved in EMT, but not with those in the Wnt/β-catenin signaling pathway (Fig. 5A). Furthermore, as shown in Fig. 5B, TAFA5 expression was moderately positively correlated with snail family transcriptional repressor (SNAI) 1, SNAI2, twist family bHLH transcription factor (TWIST) 1, TWIST2, zinc finger E-box binding homeobox (ZEB) 1, ZEB2, Vimentin, N-cadherin, Wnt9a and Wnt9b, and weakly correlated with SNAI3, E-cadherin and β-catenin (data not shown). In addition, protein expression levels of the epithelial marker E-cadherin were increased, while the mesenchymal markers N-cadherin and Vimentin, and the EMT-associated transcription factor ZEB1 were decreased, following TAFA5 silencing in GC cells (Fig. 5C). Taken together, these results suggested an active role for TAFA5 in promoting tumor aggressiveness via EMT in GC.