EIF4A3 serves as a prognostic and immunosuppressive microenvironment factor and inhibits cell apoptosis in bladder cancer

Gene expression profile

The pan-cancer expression of EIF4A3 was obtained from TIMER2 (http://timer.cistrome.org/) (Li et al., 2020), and the TCGA database was used to collect mRNA expression data and clinical information of BLCA. The differential expression of EIF4A3 in BLCA patients exhibiting different clinical features was analyzed using the R statistical environment (version 3.6.3).

Survival analysis

The GSE32894, GSE32548, and GSE31684 datasets, including the mRNA expression data and clinical records, were acquired from the GEO (https://www.ncbi.nlm.nih.gov/geo/) database. The R packages of “survival” and “survminer” were used to analyze and visualize the OS(overall survival) difference in the three GEO datasets and different sub-groups based on the TCGA dataset. Additionally, PrognoScan (http://dna00.bio.kyutech.ac.jp/PrognoScan/index.html) (Mizuno et al., 2009) tool to was applied to perform the OS and DSS (disease special survival) analysis of the patients in the GSE13507 dataset. Furthermore, the Kaplan–Meier plotter (http://kmplot.com/analysis/) tool was used to analyze the OS difference in patients with enriched and decreased CD8⁺ T cells, based on the TCGA dataset.

A nomogram constructed to predict BLCA survival and its evaluation

Based on the TCGA cohort, EIF4A3 expression and all statistically significant clinical factors in univariate analysis were included to construct the nomogram using the “rms” package. The “timeROC” package was utilized to get the 1, 3, and 5-year curves. The calibration curve was also performed via “rms” package.

Immune infiltration, gene expression correlation analysis, and anti-PD-L1 therapy response prediction

The correlation between EIF4A3 expression and the infiltration of different immune cells, including CD8⁺ T cells, CD4⁺ T cells, B cells, NK (natural killer) cells, macrophage cells, MDSCs (myeloid-derived suppressor cells), CAFs (cancer-associated fibroblasts), macrophage M2, and Treg (regulatory T) cells, was determined using the TIMER2 tool, Tumor purity was applied to make adjustments in immune infiltration analysis.

The correlation analysis between EIF4A3 expression and the immune checkpoint, including PD1 (programmed cell death 1, also known as PDCD1), PD-L1 (programmed cell death 1-ligand 1, also known as CD274), and CTLA4 (CTL-associated antigen 4), was also explored using the Gene_Corr module on TIMER2 (with tumor purity adjustment). Two external GEO (GSE32548, GSE32894) datasets were used to validate the correlation between EIF4A3 and PD-L1 expression using Spearman’s rank correlation coefficient, and the results were visualized via “ggplot2” packages.

Furthermore, we downloaded the “IMvigor210CoreBiologies” (http://research-pub.gene.com/IMvigor210CoreBiologies/packageVersions/) package containing anti-PD-L1 therapy clinical trial outcome for uroepithelial carcinoma (Mariathasan et al., 2018). According to the therapy response, the differences in the EIF4A3 expression of the patients were compared and the “ggplot2” package was used to visualize the results.

Gene set enrichment analysis

GSEA (gene set enrichment analysis) (Subramanian et al., 2005) was performed between high- and low-EIF4A3 expression groups using gene set from MSigDB Collections and the potential signaling pathways that EIF4A3 regulates were identified using the “clusterProfiler” package.

Cell culture and transfection

The human uroepithelial cancer cell lines T24, 5,637, and UM-UC-3, as well as human normal uroepithelial cell line SV-HUC-1, were bought from the cell bank of Type Culture Collection of Chinese Academy of Science, Shanghai Institute of Cell Biology (Shanghai, China). T24, 5,637, UM-UC-3, and SV-HUC-1 cells were cultured in DMEM, MEM, RPIM-1640, and F12K media, respectively, supplemented with 10% fetal bovine serum (FBS, Hyclone) and 100 U/mL penicillin/streptomycin and grown at 37 °C, 5% CO₂. The T24 and 5,637 cells were seeded into six-well plates (2 × 10⁵ cells/well), and small interference RNA (siRNA) and control siRNA (si-NC) were transferred to the cells using lipofectamine 2000 (Thermo Fisher Scientific, Waltham, MA, USA), as per the manufacturer’s instructions The siRNA sequence of EIF4A3(RiboBio, Guangzhou, China) was as follows: siRNA-1: 5′-CGAGCAATCAAGCAGATCA-3′, siRNA-2: 5′-GCTGGATTACGGACAGCAT-3′.

RNA extraction and real-time quantitative RT-PCR (qRT-PCR) analysis

The TRIzol reagent (Invitrogen, Waltham, MA, USA) was used to extract the total RNA of the clinical specimen and transfected cells. All specimens were obtained with the patient’s consent and signed an informed consent form. The study has been approved by the Ethics Committee of the First Affiliated Hospital of Nanchang University. cDNA was synthesized using the FastKing RT Kit (with gDNase; TIANGEN Biotech, Beijing, China), and PCR was performed in the ABI PRISM 7500 real-time PCR machine (Applied Biosystems, Waltham, MA, USA) following the manufacturer’s instructions and cycling conditions were as follows: denaturation at 95 °C for 1 min, amplification at 95 °C for 15s and 60 °C for 30s for 40 cycles. The primer sequences used in this analysis were as follows: EIF4A3: forward primer: 5′-CAACGAGCAATCAAGCAG-3′ and reverse primer: 5′-GTGGGAGCCAAGATCAAA-3′; ACTB: forward primer: 5′-TCTCCCAAGTCCACACAGG-3′ and reverse primer: 5′-GGCACGAAGGCTCATCA-3′.

Western blot

The transfected cells and tissue lysates were prepared, fractionated using sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and followed transferred to polyvinylidene fluoride (PVDF) membranes. Subsequently, the membranes were blotted in 5% nonfat milk for 1 h and incubated overnight at 4 °C with the following primary antibodies: GAPDH (ab8245, 1:1000; Abcam, Cambridge, UK), EIF4A3 (sc-365549, 1:1000), PARP (9532, 1:1000; CST, Danvers, MA, USA), Caspase3 (9662, 1:1000; CST, Danvers, MA, USA), Bcl-2 (15071, 1:1000; CST, Danvers, MA, USA). Thereafter, the membranes were washed three times with tris-buffered saline with 0.1% Tween 20 and exposed to secondary antibodies (anti-mouse for GAPDH, EIF4A3, and anti-rabbit for PARP, Caspase3, Bcl-2) for 1 h at room temperature. Lastly, the chemiluminescence detection reagent (WBKLS0100, Millipore, Burlington, MA, USA) was applied to scan and visualize the protein bands.

Cell proliferation and cloning assay

The effects of EIF4A3 on the vitality of BLCA cells were assessed using the CCK-8 detection kit (Keygen, Nanjing, China). The si-RNA-treated T24 and 5,637 cells were seeded into a 96-well plate, (3,000 cells/well) after 24 h, 48 h, 72 h, and 96 h, of incubation, the cells were incubated with the CCK-8 solution for 2 h at 37 °C and their absorbance were measured at 450 nm. The cells were also seeded in a 6-well plate (400 cells/well) to detect their proliferation ability, after 7-14 d of incubation, the cells were fixed and stained by 6% glutaraldehyde and 0.5% crystal violet. Image J was used to count clone numbers.

Cell apoptosis analysis

To investigate cell apoptosis events, the Annexin V-FITC/PI Cell Apoptosis Detection Kit (Transgen, Beijing, China) was used. Firstly, cells were digested with EDTA-free trypsin, washed twice with cold PBS, and then resuspended with binding buffer. Subsequently, cells were incubated in the dark with 5 µl Annexin V-FITC and/or 5 µl PI for 15 min. The detection of apoptosis events was carried out using a flow cytometer (BriCyte E6, Mindray, Shenzen, China) and analyzed using the FlowJo software.

Statistical analysis

R (version 3.6.3) and GraphPad Prism 7.01 were used for statistical analysis. The association between EIF4A3 expression and clinical characteristics was analyzed by the Wilcoxon rank-sum test, Kruskal-Wallis test, and Fisher’s exact test. The survival rate of the patients was determined by the Kaplan–Meier method. The other continuous data were analyzed by student’s t-test. Univariate and multivariate Cox regression analysis was used to screen out the independent prognosis factors. P < 0.05 was considered statistically significant.

EIF4A3 expression is upregulated in BLCA

A TIMER2 analysis of EIF4A3 expression in tumors and normal tissues indicated that cancer expressed EIF4A3 notably more than normal tissues, such as BLCA, breast cancer, Cholangiocarcinoma, Colon adenocarcinoma, Esophageal carcinoma, etc. compared to the normal tissues. In contrast, EIF4A3 expression was lower in kidney chromophobe, kidney renal clear cell carcinoma, and thyroid carcinoma compared to the normal tissues (Fig. 1A). Furthermore, combined analysis of TCGA and GTEx (the genotype-tissue expression) data revealed that EIF4A3 expression was significantly upregulated in BLCA (p < 0.001) (Fig. 1B). Moreover, paired bladder cancer samples (from the same patient) mRNA expression data from TCGA also revealed a significantly high EIF4A3 expression in tumor tissues (p < 0.001) (Fig. 1C). Subsequently, a correlation between clinical characteristics and EIF4A3 expression levels was determined by grouping patients based on their clinical characteristics. These results showed that the higher expression of EIF4A3 was significantly correlated with histologic grade (p < 0.001), pathological subtype (p < 0.001), pathologic stage (p < 0.05), race (p < 0.01), primary therapy outcome (p < 0.01), OS (p < 0.01), and DSS (p < 0.05) (Figs. 1D–1J). A median cut-off was considered, according to the EIF4A3 expression, and differential analysis was performed between the low and high EIF4A3 expression groups based on different clinical features (Table 1).

Furthermore, qRT-PCR assay using BLCA cell lines and 22 paired BLCA samples, to verify the mRNA expression levels of EIF4A3 in BLCA, revealed that EIF4A3 expression was significantly elevated (p < 0.01) in 5,637, T24, and UM-UC-3 cells compared to the SV-HUC-1 cell line (Fig. 2A). Similarly, a significant increase in EIF4A3 mRNA expression was found in cancer tissues compared with paracancerous tissues in the BLCA tissue samples (p < 0.01; Fig. 2B). And western blot results also revealed that EIF4A3 expression was increased in both cancer cell lines and tumor tissues (Figs. 2C–2D). These results reveal that EIF4A3 expression was upregulated in BLCA at both transcription and translation levels.

EIF4A3 is an independent prognostic factor in BLCA and validated in the external databases

To identify the prognostic value of EIF4A3 in BLCA, the patients in the TCGA dataset were divided into low and high EIF4A3 expression groups, based on the median EIF4A3 expression, to perform survival analysis. As shown in Figs. 3A–3B, high EIF4A3 expression was related to poor OS (HR =1.57, p = 0.003) and DSS (HR =1.56, p = 0.017) in BLCA. To further verify the prognostic value of EIF4A3 expression, we screened four external independent GEO datasets. Among these, GSE32894, GSE32548, and GSE31684 showed a lower OS for BLCA patients with higher EIF4A3 expression (HR =4.72, p < 0.001, HR =2.72, p = 0.017, HR =1.77, p = 0.033 respectively) (Figs. 3C–3E). Likewise, Figs. 3F–3G suggested that high expression group patients had a worse OS (HR =1.59, p = 0.016) and DSS (HR =2.70, p = 0.0002) in the GSE13507 cohort, which was obtained from the PrognoScan database.

In addition, the subgroup analysis showed that high EIF4A3 expression was correlated with low survival in the following BLCA cases: patients with a smoking history (HR = 1.55, p = 0.013), male patients (HR =1.59, p = 0.012), age < = 70 (HR =1.73, p = 0.013), white race (HR =1.49, p = 0.017), high grade (HR =1.55, p = 0.004), N0 stage (HR =1.76, p = 0.018), non-papillary subtype (HR =1.62, p = 0.006), T3 stage (HR =1.72, p = 0.009), pathologic stage III (HR =2.04, p = 0.009) (Figs. 3H–3P). BLCA patients’ EIF4A3 expression was negatively correlated with the OS in univariate and multivariate cox regression analyses (HR =1.568, p = 0.003, HR =1.698, p = 0.033, respectively; Table 2). All of these results demonstrate that EIF4A3 could be an independent prognostic marker for BLCA.

Nomogram construction based on EIF4A3 expression in BLCA

We included T, N, M, age, and pathologic stage clinical features based on univariate analysis and EIF4A3 expression to construct a nomogram for prognostic prediction in TCGA. The predicted 1, 3, and 5-year OS rates of BLCA by this nomogram were shown in Fig. S1A. In addition, the performance of the model was evaluated using ROC (receiver operating characteristic) curves and calibrations. The C-index for the nomogram was 0.665. Using the nomogram, the estimated 1, 3, and 5-year overall survival AUC values were 0.724, 0.711, and 0.706, respectively (Fig. S1B). The calibration plots showed that the real OS result had mostly complied with the nomogram-based predicted prognosis (Figs. S1C–S1E). These results demonstrated this model’s potential clinical prognostic value.

EIF4A3 expression is correlated with immune infiltration, PD-L1expression, and can potentially predict anti-PD-L1 therapy response in BLCA

TIMER2 analysis of the EIF4A3 expression and immune filtration coefficients revealed that immune-active cells, including CD8⁺ T cells (Rho = −0.239, p < 0.001) and CD4⁺ T cells (Rho = −0.198, p < 0.001) were negatively associated with EIF4A3 expression, while immune-suppressive cells, including MDSCs (Rho =0.408, p < 0.001), CAFs (Rho =0.12, p < 0.05), Treg cells (Rho = 0.165, p < 0.01) and M2 macrophage (Rho =0.137, p < 0.01) were positively correlated with EIF4A3 expression (Fig. 4A).

Checkpoint receptors such as PD1 /PD-L1 and CTLA4 can be blocked to relieve CD8⁺ T cells exhaustion and restart prime, respectively, thereby eradicating tumor cells that express antigens (Farhood, Najafi & Mortezaee, 2019). Thus, we analyzed the association between EIF4A3 and PD1/PD-L1 and CTLA4 expression using the TIMER2 tool. As was shown in Fig. 4B, the expression of EIF4A3 was positively correlated with PD-L1 expression (Rho = 0.337, p < 0.001), but not with PD1 and CTLA4 (p > 0.05) expression. Additionally, two independent external GEO datasets (GEO32548 r = 0.237, p = 0.006, GEO32894 r = 0.200, p < 0.001 respectively) confirmed the correlation between EIF4A3 and PD-L1 expression (Figs. 4C–4D). These results show that EIF4A3 is negatively correlated with CD8⁺ T cell infiltration. Subsequently, we analyzed whether there was a difference between the OS and EIF4A3 expression in patients with enhanced or reduced CD8⁺ T cells using the Kaplan–Meier plotter tool. Interestingly, we found that in patients with enriched CD8⁺ T cells, high EIF4A3 expression indicated the worst OS (HR =1.55, p = 0.024) (Fig. 4E), whereas a low infiltration of CD8+ T cells did not result in any differences. To evaluate the importance of EIF4A3 for clinical application in immunotherapy, we acquired the expression and therapeutic effect data of an open-source clinical study on anti-PD-L1 therapy for uroepithelial carcinoma, published in 2018 (Mariathasan et al., 2018). The expression of EIF4A3 was higher in patients who responded to immunotherapy compared to the non-responsive patients (p < 0.01) (Fig. 4F). Therefore, EIF4A3 expression was associated with the immunosuppressive microenvironment and PD-L1 expression and can thus act as a potential biomarker for immunotherapy response in BLCA patients. however, this needs to be validated by further studies.

Gene set enrichment analysis results based on EIF4A3 expression

To identify the signaling pathways regulated by EIF4A3 overexpression, GSEA was performed by using hallmark gene sets to compare groups with high and low expression of EIF4A3 based on the TCGA dataset. The results revealed that several signaling pathways, including Hedgehog, Wnt-β-catenin, Angiogenesis, IL6-JAK-STAT3, Interferon-α response, PI3K-AKT-mTOR, TGF-β, and Apoptosis were regulated by EIF4A3 overexpression (Figs. S2A–S2H).

EIF4A3 knockdown inhibited cell proliferation and promoted apoptosis in vitro

The above results showed that EIF4A3 is overexpressed in BLCA, implying its crucial role in BLCA tumorigenesis. GSEA analysis also suggested the elevated EIF4A3 expression was negatively correlated with the enrichment of apoptosis gene signatures. Thus, we used siRNA to knock down EIF4A3 expression to observe its biological impact on 5,637 and T24 cells, and CCK8 and cell clone formation assays were used to evaluate the effect of EIF4A3 on cell proliferation. Figs. 5A–5B suggested that the OD values and clone numbers were significantly decreased in both 5,637 and T24 cells with EIF4A3 knockdown, suggesting its proliferative role. Furthermore, cell apoptosis level analysis using flow cytometry and Annexin V and PI double staining, revealed that the percentage of apoptosis cells was significantly increased after in the EIF4A3-knockdown cell lines, compared to the control 5,637 and T24 cell lines (Figs. 5C–5D). Meanwhile, analysis of apoptosis-related biomarkers using western blots revealed that the anti-apoptotic protein, Bcl2 was significantly decreased, whereas the pro-apoptotic proteins cleaved-PARP and cleaved-caspase3 were upregulated in the EIF4A3 knockdown cells (Fig. 5E). Altogether these results demonstrate that EIF4A3 is involved in BLCA progression by promoting proliferation and inhibiting apoptosis in BLCA cell lines.