Identification of DEGs and DELs between gastric cancer tissues and normal tissues and enrichment analysis of DEGs
Having obtained the gene expression profiles of 19069 coding genes and 14448 lncRNAs in 407 gastric cancer tissues and adjacent non-tumorous gastric tissues, we identified 2386 DEGs and 985 DELs based on the gene/lncRNA selection criteria. Cluster analysis (Fig. 1a) suggested that the expression profiles of identified DEGs and DELs can be distinguished between the gastric cancer tissues and the normal gastric tissues.
GO and KEGG pathway analysis was performed to predict the potential roles of identified DEGs (Table S1–S4). As shown in Fig. 1b, KEGG analysis showed these DEGs were mostly enriched in pathways including cell cycle, calcium signaling pathway, cell adhesion molecules (CAMs) and DNA replication. Biological processes of GO analysis (Fig. 1c) showed the enrichment of DEGs in mitotic nuclear division, DNA replication, cell cycle G1/S phase transition, G1/S transition of mitotic cell cycle, meiotic cell cycle, DNA-dependent DNA replication, and DNA replication initiation, etc. It is noteworthy that both analytic methods have indicated the enrichment of these DEGs in tumor-related pathways such as cell cycle regulation and DNA replication, supporting the potential participation of these DEGs in regulating tumor development and progression.
Construction of co-expression network between DEGs and DELs
WGCNA was exploited to cluster closely co-expressed DEGs and DELs into co-expression networks. We clustered these DEGs and DELs by average-linkage hierarchical clustering analysis by transforming adjacency matrix into TOM, and set each network module with a minimum of 30 genes/lncRNAs based on Dynamic Tree Cut standard (Fig. 2a, b). The eigengenes for each module were then calculated, and 11 new modules were generated on the basis of correlation efficiency (Fig. 2c). Genes in yellow, black, brown, magenta, green, blue and red modules were associated with gastric cancer tissues, while genes in green, yellow, purple, pink and turquoise modules were associated with normal tissues. Notably, the gray module was unable to be clustered into other modules. A total of 2374 DEGs and 979 DELs were allocated into 11 modules and the information of these DEGs and DELs were listed in Table S5.
The association between each module and clinical traits was calculated by the Pearson’s correlation coefficient between ME and sample traits (Fig. 2d). DEGs and DELs clustered in blue module showed the strongest correlation with gastric cancer, suggested blue module was gastric cancer highly correlated module.
Screening for key modules and hub genes
Pathway enrichment analysis was performed on gene sets in each module (Table S6), and 7 modules were enriched in 64 KEGG pathways (Fig. S1). Distinct enriched pathways were displayed between different modules, inferring the independent functional notes of each module. It is notable that 6 of 11 pathways (54.5%) enriched in the blue module, such as cell cycle, DNA replication, homologous recombination, and p53 signaling pathway, were the same as we have identified for all DEGs.
We selected blue module as the key module depends upon the analytic results from the enrichment analysis and the module–cancer interactions. 15 hub genes were identified by way of calculating the correlation coefficients between genes and MEs in the blue module (Table S7). It was interesting again to find most of these genes were enriched in the process of cell replication and cell cycle regulation (Table S8).
Construction of gene–gene–lncRNA network
Aiming for observing the correlation between lncRNA and hub genes in co-expression modules, we first determined the hub genes related lncRNAs by calculating the correlation coefficients between individual lncRNA and the hub genes. 5 lncRNAs of CTD-2510F5.4, RP11-120D5.1, RP5-991G20.1, DLEU2 and AC015849.16 were ultimately screened out.
We then generated the interaction network between these hub genes using STRING online database, and constructed gene–gene–lncRNA network after recruiting identified lncRNAs. As shown in Fig. 3a, CTD-2510F5.4 was closely related to 10 hub genes: E2F3, DTL, RBL1, NUSAP1, ATAD2, KIF18B, MCM10, RAD54L, BUB1 and KPNA2. Figure 3b indicates CTD-2510F5.4 possibly regulated these 10 hub genes in trans.
To address the modulatory role of CTD-2510F5.4 on these 10 hub genes, we knocked down CTD-2510F5.4 by siRNA transfection strategy. As shown in fig. S2, siRNA1 effectively reduced more than 60% of CTD-2510F5.4 expression after 48 h of transfection (p < 0.0001), whereas siRNA2 only reduced about 30% of gene expression (p < 0.0001). The mRNA expression level of 10 CTD-2510F5.4-related hub genes was then measured by qRT-PCR. In results, reduction in CTD-2510F5.4 expression significantly decreased expression of 9/10 hub genes (Fig. 3c–m), implying CTD-2510F5.4 could regulated these hub genes in gastric cancer cells.
CTD-2510F5.4 knock down significantly reduced cell viability of gastric cancer cells
The impact of CTD-2510F5.4 on the cell viability of gastric cancer cells was detected by the CCK-8 assay (Fig. 4a). Compared to the mock cells, CTD-2510F5.4 knock down significantly reduced the cell viability after 24 h of transfection, and such effect was constantly seen after 48 and 72 h (p < 0.001), suggesting CTD-2510F5.4 knock down could cause cell death in the gastric cancer cells.
Regulation of cell cycle and apoptosis by CTD-2510F5.4 knock down
Downregulation of CTD-2510F5.4 caused reduced expression of cell cycle related genes, which prompted us to investigate if CTD-2510F5.4 was functionally related to cell cycle distribution. As shown in Fig. 4b, the percentage of cells in the G0/G1 phase increased from 27.3% (mock group) to 37.5% (CTD-2510F5.4 silencing group), suggesting induction of G0/G1 cell cycle arrest in the absence of CTD-2510F5.4 (p < 0.001).
The relationship between CTD-2510F5.4 and apoptosis was also investigated by flow cytometry (Fig. 4c). In results, there was a 2.5-fold change of increase in the late apoptotic cells in the CTD-2510F5.4 knock down cells (14.5%) when compared with mock cells (5.9%) (p < 0.001). No significant difference of necrotic cells or early apoptotic cells was observed between the two groups. These results implied the impact of CTD-2510F5.4 on the late apoptosis in gastric cancer cells.
CTD-2510F5.4 expression in paired gastric cancer tissues and adjacent-normal gastric tissues
We then detected the expression level of CTD-2510F5.4 in 90 paired gastric cancer tissues and adjacent-normal gastric tissues by ISH. CTD-2510F5.4 expression was observed in nuclei and cytoplasm (Fig. S3a). Significantly higher expression level of CTD-2510F5.4 was detected in the gastric cancer tissues (159.6%, 119.6–174%) than in the normal tissues (98.2%, 62–132.2%) (p < 0.01), suggesting CTD-2510F5.4 may be a potential biomarker for gastric cancer (Fig. S3b).
Receiver operating characteristic (ROC) curve analysis determined the cut-off value for CTD-2510F5.4 expression
The clinicopathological parameters of 90 gastric cancers were listed in table S9. ROC curve analysis revealed CTD-2510F5.4 expression could be significantly distinguished by clinicopathological parameters of pathological grade, vascular or nerve invasion, AJCC TNM stage and overall survival (OS) (Fig. 5). The area under the curve (AUC) and p value for these parameters were 68.1% and 0.005 for OS, 64.7% and 0.034 for vascular or nerve invasion, 66.3% and 0.008 for pathological grade, and 66.9% and 0.006 for AJCC TNM stage. Cut-off value of 148.5% for CTD-2510F5.4 expression was determined, at which maximum Youden index was obtained by comparing AUC and p value for each parameter. CTD-2510F5.4 staining with H-score > 148.5% were considered as high CTD-2510F5.4 expression, and ≤ 148.5% were considered as low CTD-2510F5.4 expression. Accordingly, 36 gastric cancer tissues (40.0%) and 54 gastric cancer tissues (60.0%) showed low and high expression levels of CTD-2510F5.4, respectively.
Correlation between CTD-2510F5.4 expression and clinicopathological parameters of gastric cancer
Having determined the cut-off value for CTD-2510F5.4 expression, we statistically analyzed the correlation between CTD-2510F5.4 expression and clinicopathological parameters of gastric cancers (Fig. 5; Table 1). (1) In regard to pathological grade, we found that high CTD-2510F5.4 expression was present in significantly more gastric cancers at pathological grade = III (73%) than those at pathological grade < III (47.9%) [adjusted p = 0.011, adjusted OR (95% CI) = 0.303 (0.120–0.760)]. (2) For tumor location, high CTD-2510F5.4 expression was detected in 76.2% of patients with middle third located gastric cancers, whereas 41.2% and 59.6% in patients with upper third or lower third located gastric cancers, respectively [adjusted p = 0.04, adjusted OR (95% CI) = 0.229 (0.056–0.935)]. (3) For depth of invasion, 68.3% of gastric tissues with serous membrane invasion and 43.3% of gastric tissues without serous membrane invasion showed high CTD-2510F5.4 expression [adjusted p = 0.01, adjusted OR (95% CI) = 0.271 (0.099–0.744)], respectively. (4) For vascular or nerve invasion, 79.2% of gastric tissues with vascular or nerve invasion showed high CTD-2510F5.4 expression, while 53.0% gastric tissues without vascular or nerve invasion showed high CTD-2510F5.4 expression with significant difference [adjusted p value = 0.028, adjusted OR (95% CI) = 0.285(0.093–0.871)]. (5) As for AJCC TNM stage, we observed more gastric cancers at III/IV stage (70.8%) obtained high CTD-2510F5.4 expression than those at I/II stage (47.6%) [adjusted p = 0.007, adjusted OR (95% CI) = 0.201 (0.063–0.644)]. (6) In analyzing the OS, high CTD-2510F5.4 expression was present in 68.3% deaths and 43.3% survived patients [adjusted p = 0.007, adjusted OR (95% CI) = 0.230 (0.078–0.673)], respectively.
Table 1 Correlation of CTD-2510F5.4 expression with clinicopathological parameters in patients with gastric cancer
Based on these findings, we concluded that high CTD-2510F5.4 expression potentially correlated with pathological grade, tumor location, serous membrane invasion, vascular or nerve invasion and AJCC TNM stage, suggesting high CTD-2510F5.4 expression has high propensity to be present in gastric cancer patients with poor prognosis. In addition, no statistical differences were identified in regard to other parameters including age, pathological type, tumor size, tumor location, pathological morphology and lymph node metastasis.
Correlation between CTD-2510F5.4 expression and prognosis of gastric cancer
Next, we analyzed CTD-2510F5.4 expression in all 90 cases of gastric cancers (Table 2). Median survival time (MST) was significantly shorter in cases with high CTD-2510F5.4 expression (32.849 months, 95% CI = 25.194–40.503 months) than in those with low CTD-2510F5.4 expression (49.083 months, 95% CI = 39.613–58.554 months) (p = 0.012, Fig. 5e). Multivariate cox analysis also revealed high CTD-2510F5.4 expression was a risk factor for the prognosis of gastric cancer patients [hazard ratio (HR, 95%CI) = 2.303 (1.316–4.028), p = 0.003].
Table 2 Association of CTD-2510F5.4 expression and clinicopathological features with OS in gastric cancer patients
We further analyzed the correlation of CTD-2510F5.4 expression and OS stratified patient groups in regard to clinicopathological parameters. Of 48 patients with gastric cancer at pathological grade < III, there was a significant correlation between CTD-2510F5.4 expression and the patients’ OS (Fig. 5f). The MST in patients with high CTD-2510F5.4 expression (n = 23) was significantly shorter (37.696 months, 95% CI = 26.215–49.177 months) than in patients with low CTD-2510F5.4 expression (n = 25) (55.200 months, 95% CI = 44.343–66.057 months) (p = 0.028). Such correlation was also found in 66 patients without vascular or nerve invasion (Fig. 5g). The MST was 33.036 months and 50.742 months for patients with high CTD-2510F5.4 expression (n = 35) (95% CI = 23.638–42.435 months) and low CTD-2510F5.4 expression (n = 31) (95% CI = 40.817–60.667 months), respectively (p = 0.013). High CTD-2510F5.4 expression was also proved to be an independent risk factor for gastric patients with cancer at pathological grade < III (HR (95% CI) = 2.362(1.069–5.219), p = 0.034), or without vascular or nerve invasion [HR (95% CI) = 2.349(1.245–4.433), p = 0.008]. These data supported a role of high CTD-2510F5.4 expression as a risk factor in predicting poor prognosis for gastric cancers at pathological grade < III or without vascular or nerve invasion.