A complete list of reagents and antibodies is provided in Supplementary Material 1. All the concentrations were chosen based on previous studies or manufacturer’s instructions. The detailed screening protocol is also presented in Supplementary Material 1.

AGS (moderately differentiated GC cells), HGC27 (undifferentiated GC cells), GES-1 (healthy gastric epithelial cells), and THP-1 cells (human monocytic cells) were purchased from the cell bank of the Chinese Academy of Sciences (Shanghai, China). MKN74 and MKN45 cells (well and poorly differentiated GC cells, respectively) were purchased from the Japanese Collection of Research Bioresources Cell Bank. Human MSCs were purchased from Cyagen Biosciences (Guangzhou, China). GC and THP-1 cells were cultured in RPMI-1640 medium with 10% fetal bovine serum (FBS), and GES-1 cells and human MSCs were cultured in DMEM with 10% FBS. All cells were incubated at 37°C in 5% CO₂.

Quantitative RT-PCR was performed using the protocol provided in previous studies (Bustin et al., 2005). Total RNA was extracted from cells using the TRIzol reagent. cDNA was synthesized via reverse transcription using the manufacturer’s protocol. β-actin was chosen as the internal control. The primers were as follows: β-actin (F): 5′-GCGTGACATTAAGGAGAAGC-3′; β-actin (R): 5′-ACGTCACACTTCATGATGG-3′; NREP (F):5′ TTGAGCGAATGCTACCAGAG-3′; and NREP (R):5′-AGGCG AGGCTACGGAAAG -3′.

The protocol for western blotting was based on previous studies (Hnasko and Hnasko, 2015). Target/β-actin bands were identified with a gel image processing system (ChemiDoc XRS+). Subsequently, relative protein levels were calculated.

The study’s protocol was approved by the ethics committee of the Jiangsu Province Hospital of Chinese Medicine, and informed consent was obtained from clinicians and patients (2019NL-166-02). GC tissue and the surrounding non-tumorous tissue (margin, 5 cm) were collected during surgery from 30 previously treatment-naïve patients with GC at the Jiangsu Provincial Hospital of Traditional Chinese Medicine. Tumors were staged and graded using the 8th edition of the American Joint Committee on Cancer tumor-node-metastasis staging system (Ji et al., 2018). After extraction, tissue specimens were washed with cold phosphate-buffered saline and immediately placed in liquid nitrogen. They were then transferred and stored at −80°C until further examination.

The protocol used for Immunohistochemistry (IHC) was based on earlier studies (Nizioł et al., 2021). Images were obtained using a NIKON Eclipse Ni-E microscope (NIKON, Japan; original magnification, ×400). The H-SCORE (range 0–300, higher scores indicating stronger positive staining) was calculated as described previously (Yang et al., 2019).

We used lentiviral vectors for overexpressing and knocking down NREP. Viruses were designed, synthesized, and produced by GeneChem Corporation. Transfection was performed according to the supplier’s protocol. HGC27 and MKN74 cells were transduced with the recombinant lentiviruses using 2 μg/mL polybrene for 24 h. Subsequently, we identified stably transfected GFP-expressing cells using 1.5 μg/mL puromycin. We assessed NREP overexpression and knockdown as well as transduction efficiency using western blots.

We examined cell supernatants for TGF-β1 levels using the TGF-β1 Enzyme-linked immunosorbent assay (ELISA) Kit based on the given instruction manual. A microplate reader (BioTek Synergy HT) was used to examine optical density at 450 nm.

We assessed the clonogenic ability of cells using a clone formation assay, as described previously (Li et al., 2012). The number of colonies was counted using a compound light microscope (Olympus BX53, Japan).

All animal experiments were approved by the ethics committee of the Jiangsu Province Hospital of Chinese Medicine (2021-5-062). Twenty-four 4-week-old male BALB/c nude mice were obtained from the Beijing Institute of Biomedicine (Beijing, China; Certificate No. SYXK2019-0010). MKN74 cells transfected with sh-NREP, oe-NREP, and NC and control cells (4 × 10⁶ cell/mouse) were injected subcutaneously into the right armpit region of the mice (n = 6 per group). 7 days later, tumor formation was observed beneath the skin. The maximum (a) and minimum tumor diameter (b) were measured twice weekly. On day 15, the mice were euthanized and all tumors were collected. Tumor volume was calculated (V = 1/2ab²), and the growth curves of the subcutaneous xenografts were drawn.

The protocol used for the wound healing assay was based on earlier studies (Pasquale et al., 2020). Cell migration toward the scratch zone was photographed using an inverted fluorescence microscope (Olympus CKX-41, Japan; ×200 magnification) every 12 h.

Cell invasion was assessed using a transwell assay based on a previously published protocol (Misra et al., 2021). The membrane in the chamber was cut and imaged using light microscopy (Olympus BX53, Japan; ×200 magnification), and cell counts were obtained using Image J software.

The TUNEL apoptosis detection kit was used to perform the TUNEL assay, as described previously (Telegina et al., 2019). TUNEL-negative (blue) and TUNEL-positive (red) cells were observed using a fluorescence microscope (Olympus CKX-41, Japan; ×200 magnification).

The protocol used for immunofluorescence staining was based on earlier studies (Donaldson, 2015). Immunofluorescence staining was observed using an epi-fluorescence microscope (Olympus, BX60-32FB2-A03) and different filters and imaged using an Olympus, DP50 camera (×400 magnification).

A non-contact co-culture unit of MSCs and GC cells was established using a co-culture transwell system (upper chamber, GC cells; lower chamber, MSCs; Long et al., 2019). The culture medium was changed every 48 h. After 4 days of non-contact co-culture, the culture in the lower chamber was terminated and cells were harvested for other experiments.

THP-1 cells (1 × 10⁵ cells/mL) were treated with phorbol 12-myristate 13-acetate (PMA; 10 ng/mL) for 48 h to allow the induction of macrophage differentiation (Genin et al., 2015). PMA-containing medium was replaced with serum-free medium, and the cells were cultured for 24 h. 2 days before the co-culture experiment, cells (1 × 10⁵ cells/mL) from the control, knock-down (sh-NREP), overexpression (oe-NREP), and negative control (NC) groups were seeded onto 0.4-μM transwell inserts. For co-culture, the culture medium in the inserts with GC cells was removed and transferred to the top of the pates with differentiated THP-1 cells. After 48 h of further co-culturing, cells were obtained, and immunofluorescence staining was performed.

Data were reported as the mean ± standard deviation. We used t-tests and one-way ANOVA to perform comparisons between two groups and among multiple groups, respectively. All data were analyzed using SPSS 26.0 (SPSS Inc., United States) and illustrated using GraphPad Prism 8.0 (GraphPad Software, Inc., United States). All experiments were carried out at least thrice. ^∗∗P < 0.01 and ^∗P < 0.05 were considered statistically significant.

The expression of NREP in GC was first investigated using the TIMER and¹ and GEPIA databases² (Pan et al., 2019; Yuan et al., 2019). The Cancer Genome Atlas (TCGA)-Stomach Adenocarcinoma (STAD) cohort and Fei et al. (2018), Li et al (2020), and Shan et al. (2021) datasets were then used to further confirm the differential expression of NREP (Rossari et al., 2018; Tian et al., 2018). The Human Protein Atlas database³, which contains data on >11,200 unique proteins, is the biggest, most comprehensive database on protein distribution in human tissues and cells (Thul and Lindskog, 2018).

Multivariate Cox regression analysis of TCGA-STAD data was used to identify whether NREP could be an independent prognosticator for GC. P values, hazard ratios (HRs), and 95% confidence intervals (CIs) were obtained using the “forest plot” R package.

The differences in NREP levels were analyzed based on various classification parameters, such as the T/N/M stage, pathologic stage, and histologic grade.

Differences in survival based on NREP expression were examined using Kaplan–Meier survival analysis and log-rank tests. P values, HRs, and 95% CIs were obtained using log-rank tests and univariate Cox proportional hazards regression. Time receiver operating characteristic (ROC) analysis was applied to calculate the accuracy of prognostication based on NREP levels.

The GeneMANIA database was used to identify genes that showed NREP-lined expression and to explore their potential functions (Franz et al., 2018). Genes co-expressed with NREP were identified based on TCGA-STAD data (criteria: |logFC > 3| and P < |0.05|). Subsequently, we conducted functional enrichment analysis of NREP and the identified genes using the Enrichr database (Kuleshov et al., 2016).

To analyze the connections among proteins, genes co-expressed with NREP were assessed using the Search Tool for the Retrieval of Interacting Genes database (STRING)⁴ (Szklarczyk et al., 2021); we visualized the results obtained after setting the minimum interaction score to 0.4 in Cytoscape (Doncheva et al., 2019). In addition, based on the constructed protein–protein interaction network (PPI), important modules were screened out using the Molecular Complex Detection (MCODE) tool. Hub genes were obtained by setting the following cutoffs: degree cutoff value = 2, node density cutoff value = 0.1, node score cutoff value = 0.2, k-core = 2, and maximum depth = 100. Finally, the relationship between the levels of these hub genes and GC prognosis was examined using TCGA-STAD data.

Gene set enrichment analysis (GSEA) was performed using the Broad Institute GSEA software 3.0. The geneset “subset of GO (Gene Ontology)” (Molecular Signatures Databases; http://www.gsea-msigdb.org/gsea/msigdb/index.jsp) was used for GO enrichment analysis (Powers et al., 2018). Statistical significance was defined at a normal P value < 0.05. The GSCALite online tool was used to explore the relationship between the 10 hub genes and EMT, and we calculated the co-expression relationship between NREP and 6 classical factors of EMT using TCGA-STAD data (Liu et al., 2018). In addition, single-cell analysis was conducted based on GSE134520 to find more evidence on the potential functions of NREP.

The correlation between NREP expression levels and fibroblast levels was first calculated based on the Explicitly Parallel Instruction Code (EPIC) and Mixed Complementarity Problem (MCP)-counter algorithm from the TIMER web tool (Sturm et al., 2019). GC patients were grouped into high- and low-expression cohorts based on the median NREP expression, and differential analysis was performed to identify the differentially abundant cells. Furthermore, CIBERSORT — a high-performance computational method used for quantifying cellular components from bulk-tissue gene expression profiles — was used to accurately estimate immune infiltration (Chen et al., 2018). Spearman’s rank correlation coefficients were calculated for pairwise correlation comparisons; P < 0.05 was defined as statistically significant. All findings were illustrated using “ggplot2” and “pheatmap.”

The Cancer Genome Atlas-Stomach Adenocarcinoma data showed that NREP expression was higher in GC tissues than in normal tissues (Figure 1A; P < 0.05). Moreover, the analysis of TIMER and TCGA data also showed high NREP expression in GC (Figure 1B).

We further explored the expression of NREP in GC tissues using data from the Gene Expression Omnibus (GEO) database. Data from the Fei et al. (2018), Li et al (2020), and Shan et al. (2021) datasets indicated a significant difference in NREP expression between GC tissues and adjacent tissues (Figures 1C–E). Western blot, RT-PCR, and IHC staining also revealed that NREP was over-expressed in GC cells and tissue. The mean H-SCOREs for NREP expression in GC and paracancerous tissue were 93.15 ± 18.21 and 10.49 ± 3.94, respectively (Figures 1F–J; P < 0.01, ANOVA). NREP protein expression in GC was further verified using IHC data from The Human Protein Atlas. NREP was found to be primarily expressed in the cell membrane and cytoplasm (Figure 1H).

Multivariate analysis of TCGA-STAD data revealed that NREP overexpression, age, and tumor stage were related with a poor prognosis in GC (P < 0.001; Figure 2A). In particular, NREP levels were identified as an independent predictor of survival in GC patients. The relationship of NREP levels with the clinicopathological characteristics of GC patients — including T/N/M stage, histologic grade, pathologic stage, gender, and race — is illustrated in Figure 2B. The mRNA levels of NREP only showed a correlation with T stage (P < 0.01, P < 0.001), and no such relationship was observed for other clinical features. Further analysis of the prognostic value of NREP using TCGA-STAD data showed that the low-risk group had a longer duration of survival than did the high-risk group (Figure 2C; P < 0.05). Additionally, according to TCGA-STAD data, the areas under the ROC curves of NREP expression for 1-, 3-, and 5-year OS were 0.589, 0.651, and 0.708, respectively, (Figure 2D). In summary, the results showed that NREP overexpression could be used as an indicator for OS in GC.

The functional network of NREP and its neighboring genes obtained using GeneMANIA is displayed in Figure 3A. We identified genes showing expression levels positively or negatively correlated with NREP expression using TCGA-STAD data and the “DESeq” R package (Figure 3B). The 73 differentially expressed genes (DEGs) were imported into the DEG PPI (Figure 3C). After applying Cytotype MCODE, we identified 10 hub genes among which 8 were up-regulated and 2 were down-regulated (Figure 3D and Supplementary Figures 1A–J). The GSCALite online tool was used to elucidate the relationship between the 10 hub genes and EMT (Figure 3E). Further enrichment analysis suggested that NREP may be associated with “extracellular structure organization,” “external encapsulating structure organization,” “extracellular matrix organization,” “collagen-containing extracellular matrix,” “platelet-derived growth factor binding,” “focal adhesion,” and “ECM-receptor interaction” (Figures 3F–I).

Additional survival analysis revealed that the levels of 5 up-regulated genes were associated with the prognosis of GC patients (Supplementary Figures 1K–T).

In vitro NREP silencing in HGC27 and MKN74 cells using shRNA constructs significantly down-regulated NREP expression (Supplementary Figure 2; P < 0.01). NREP silencing also decreased tumor cell clone formation (Figure 4A). In contrast, the opposite trend was observed in cells overexpressing NREP. Moreover, stable NREP overexpression in MKN74 cells promoted the formation of subcutaneous xenograft tumors in vivo (Figures 4B–D; P < 0.01). GSEA for NREP revealed the potential role of NREP in “epithelial–mesenchymal transition” and “TGF-beta signaling.” Analyses using TIMER data revealed a positive correlation between NREP and TGF-β1 expression (R = 0.520, P < 0.001; Figure 4E). Subsequent in vitro experiments using ELISA revealed a reduction in TGF-β1 levels in culture medium after NREP silencing (P < 0.05; Figure 4F), and the opposite trend was observed when NREP was overexpressed. NREP overexpression was also found to increase EMT-related phenotypes such as the migration and invasion of GC cells and the expression of EMT-associated proteins. However, this effect was attenuated after treatment with the TGF-β signaling kinase inhibitor LY364947 (Figures 4G–J). In addition, cell viability assays showed that the selected concentration of LY364947 did not affect cell proliferation (Supplementary Material 1). Further analyses based on TIMER data also revealed a positive correlation between the expression of NREP and that of CDH2 (R = 0.64, P < 0.001), MMP2 (R = 0.65, P < 0.001), MMP9 (R = 0.13 P = 0.01), and VIM (R = 0.69, P < 0.001). Further, NREP expression was found to show a negative correlation with CDH1 expression (R = −0.130, P = 0.01; Figure 4K). Therefore, NREP silencing and overexpression altered the levels of EMT-related proteins (Figures 4L,M).

Gene set enrichment analysis also revealed functional enrichment for NREP under the “ACTIN FILAMENT ORGANIZATION” and “ACTIN_CYTOSKELETON” domains (Figure 5A). Subsequent in vitro experiments revealed that NREP overexpression caused the up-regulation of F-actin (Figures 5B,C; P < 0.01). Moreover, after TUNEL staining, no TUNEL-positive cells were detected in cells overexpressing NREP. In contrast, NREP knockdown and LY364947 treatment increased the number of TUNEL-positive cells (Figure 5D).

The abundance of cancer-associated fibroblasts (CAFs) was found to be a potential prognostic factor in GC (Table 1). Using the EPIC and MCP-counter algorithms and TCGA-STAD data (Figures 6A,B), we found that NREP expression was positively correlated with CAF abundance (EPIC: R = 0.715, P = 1.13e-60 and MCP-counter: R = 0.761, P = 5.74e-73). Subsequent single-cell-level analyses revealed that NREP was mainly expressed in fibroblasts, which are important players in EMT (Figures 6C–F).

Local and recruited MSCs are known to transform into CAFs at close proximity to tumor cells. To test whether NREP overexpression in GC facilitates the conversion of MSCs into CAFs, we co-cultured MSCs with GC cells (Figure 6G). After the co-culture of MSCs with GC cells, CAF markers were remarkably up-regulated in MSCs (Figure 6H; P < 0.01, P < 0.001). Furthermore, we co-cultured MSCs with GC cells showing different levels of NREP expression for 4 days. Immunofluorescence staining revealed that α-SMA and Vimentin levels were up-regulated in the NREP overexpression group, whereas they were down-regulated in the sh-NREP group (Figures 6I,J; P < 0.05, P < 0.01). Interestingly, ELISA revealed a 4-fold increase in the levels of TGF-β1 in the cell supernatant after MSCs were co-cultured with NREP-overexpressing cells (Figure 6K; P < 0.05, P < 0.05, and P < 0.001). Consistent with our previous results, these findings also showed that a TGF-β inhibitor can decrease the ability of MSCs to differentiate into CAFs in a co-culture unit.

Recently, two modes of cancer cell invasion have been defined: collective cancer cell invasion and fibroblast-led collective invasion. Next, we examined the effect of NREP overexpression on CAF recruitment through a transwell experiment. We observed that with an increase in NREP expression, the ability of GC cells to recruit CAFs was significantly enhanced (Figures 6L,M).

Together, these results indicated that NREP may contribute to GC progression by recruiting and activating fibroblasts.

Using CIBERSORT, we found that the proportion of tumor-infiltrating cells was positively correlated with NREP levels and the presence of M2 macrophages in patients with GC (P < 0.001; Figure 7A). Subsequently, the relationship of NREP expression with macrophage polarization was assessed using the Li et al, 2020 and Fei et al., 2018 datasets. The correlation between NREP expression and immune cell subpopulations is shown in Figure 7B. Interestingly, NREP expression levels were found to be positively correlated with M2 macrophage abundance (Figure 7C).

Therefore, we calculated the correlation of NREP with M2 surface markers using the TIMER database and observed a positive correlation between NREP expression and MRC1 (CD206; R = 0.34, P < 0.001) and CD163 (R = 0.37, P < 0.001) expression (Figure 7D). This series of results suggested the presence of a positive association between NREP expression and M2 macrophage infiltration. To further investigate the influence of NREP overexpression on M2 macrophage abundance in GC, we established a tumor–macrophage cell co-culture model using a transwell non-contact co-culture unit (Figure 7E). We observed that NREP overexpression significantly up-regulated the surface markers of M2 TAMs (CD206 and CD163; Figures 7F,G).

Hence, our findings confirmed the positive correlation between NREP levels and the abundance of M2 macrophage infiltration.