The Cancer Genome Atlas (TCGA) program is one of the largest tumor databases in the world, containing clinical data and gene expression profiles of thousands of patients with 33 different cancer types [16, 17]. The RNA-seq expression files and relevant clinicopathological and survival data of TCGA-KIRC patients were acquired from the University of California, Santa Cruz (UCSC) Xena browser (https://xenabrowser.net). There were 533 ccRCC samples and 72 paired normal tissues included in this database. Oncomine database (http://www.oncomine.org/) is a gene-centric analysis platform by integrating multiple datasets, and is widely used for gene differential expression analysis [18]. Similarly, the Oncomine database was utilized to compare the EMPs family expression level, including EMP1, EMP2 and EMP3. EMP3 was chosen as our research object, and its differential expression was further confirmed in four subsets of the Oncomine website.

The ROC curve was utilized to measure the diagnostic value of EMP3 in people suffering from ccRCC. According to the median expression of EMP3, we analyzed the survival data of ccRCC patients. Then we established a nomogram model by grouping them into two groups (high versus low) and performed Cox-regression analysis to illustrate the prognostic role of EMP3 in ccRCC. The calibration curves of 1-, 3-, and 5-years were drawn to evaluate the performance of the nomogram. The ArrayExpress database, which afflicted to The European Bioinformatics Institute (EMBL-EBI), collected a large amount of chip microarray data for scientific research personnel to explore [19]. The E-MTAB-1980 cohort was used to verify the prognostic significance of EMP3 in ccRCC.

Gene set enrichment analysis (GSEA) and gene expression correlation analysis were implemented to bring us some clues to uncover the potential biological mechanism behind EMP3 in ccRCC. The false discovery rate (FDR) 0.25 was considered to have sufficient significance to eliminate inadequate signal pathways. The permutations for each phenotype were carried out 1000 times (p 0.05).

The 28 paired clinical samples were all from ccRCC patients who underwent surgical treatment at the Department of Urology, Union Hospital Affiliated to Tongji Medical College (Wuhan, China) from 2016 to 2021. They had never received preoperative chemotherapy or radiotherapy and informed consent was signed. After the tumor tissues and matched para-carcinoma renal tissues were isolated, both of them were frozen in liquid nitrogen quickly to prevent total RNA and protein degradation. Part of collected samples were immersed in 10% formalin to fix. According to the Declaration of Helsinki, the Ethics Committee of Huazhong University of Science and Technology (Wuhan, China) approved our experimental protocol.

Normal human renal epithelial cell lines 293 and RCC cell lines (A498, 786-O, CAKI-1, ACHN and OSRC-2) in our study were purchased from the cell bank of American Type Culture Collection (Manassas, VA, USA). All the cell lines except 786-O were cultivated in high-glucose DMEM supplemented with 10% fetal bovine serum (FBS, Gibco; Thermo Fisher Scientific, Waltham, MA, USA) and 1% penicillin-streptomycin (Gibco; Thermo Fisher Scientific; Waltham, MA, USA), and then cultured at the atmosphere of 5% CO${}_2$ in 37 ${}^{\circ }$C constant temperature incubator. 786-O was cultured separately with RPMI 1640 medium supplied with the above serum and penicillin-streptomycin.

According to the manufacturer’s protocol, the clinical tissue specimens were prepared by grinding while cells were collected before mixing with the TRIzol reagent (Thermo Fisher Scientific; Waltham, MA, USA) to isolate and extract total RNA from tissues and cells, respectively. The NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific; Rockford, IL, USA) was then utilized to detect its purity and concentration. Reverse transcription method was conducted to amplify the corresponding mRNA we need. All the qPCR were performed using the Step one plus (ABI; Thermo Fisher Scientific, Rockford, IL, USA) platform using the SYBR Green mix (Thermo Fisher, Massachusetts, USA). Primers for gene EMP3 were designed and purchased from TsingKe (Tsingke Biological Technology; Wuhan, China) and are listed as follows:

EMP3

Forward primer

5${}^{\prime }$-CCTGAATCTCTGGTACGACTGC-3${}^{\prime }$;

Reverse primer

5${}^{\prime }$-GCCATTCTCGCTGACATTACTG-3${}^{\prime }$.

$\beta$-actin

Forward primer

5${}^{\prime }$-CTCGCCTTTGCCGATCC-3${}^{\prime }$;

Reverse primer

5${}^{\prime }$-TTCTCCATGTCGTCCCAGTT-3${}^{\prime }$.

The ccRCC tumor tissues and adjacent normal renal tissues were fixed with 10% formalin and embedded in paraffin. Next, the wax blocks were cut into slices about 5 $\mu$m thick. Subsequently, after deparaffinization and hydration, the slices were incubated with primary mouse anti-EMP3 monoclonal antibody (Santa Cruz, sc-81797, 1:200) overnight in a refrigerator at 4 ${}^{\circ }$C. The tissue sections were then incubated with secondary antibody at room temperature and colored using 3,3-N-Diaminobenzidine Tetrahydrochloride (DAB). Finally, hematoxylin was used to counterstain the tissue cells, and the samples were observed under a microscope.

Collected clinical tissues and cells were lysed with a series of proteolytic agents, including RIPA lysis buffer (Beyotime Biotechnology; Haimen, China), phenylmethylsulfonyl fluoride (PMSF) and protease inhibitor cocktail (Roche Diagnostics; Indianapolis, IN, USA). After centrifugation to remove excess residue, the protein concentration was measured using the bicinchoninic acid assay (Thermo Fisher Scientific; Waltham, MA, USA) at a wavelength of 562 nm. The total protein extracts were added to 10% SDS-PAGE wells, then concentrated, separated and transferred onto the PVDF membrane (Roche Diagnostics; Indianapolis, IN, USA). The membranes were repeatedly washed three times with PBST after being blocked by defatted milk for at least one hour. According to the indicators of gel electrophoresis markers, the membranes were cut into the desired strips for further processing. The bands of interest were incubated with the primary antibody at 4 ${}^{\circ }$C for at least 12 hours, washed again and incubated with the secondary antibody at room temperature for two hours, and then photographed. Antibodies used in the study were as follows: anti-EMP3 (Santa Cruz, sc-81797, 1:200), anti-$\beta$-actin (Proteintech, 20536-1-AP, 1:1000), anti-E-cadherin (Abclonal, A18135, 1:1000), anti-N-cadherin (Abclonal, A19083, 1:1000), anti-VIM (Abclonal, A19607, 1:1000), and anti-SNAIL (Abclonal, A11794, 1:1000).

786-O and A498 cell lines were selected for in vitro experiments. The sgRNAs targeting EMP3 were designed using the synthego website (https://www.synthego.com) and utilized for knockdown experiments in the above-mentioned cell lines. The cell density in the 6-well plate was maintained at 50%–60% confluence, and Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) was mixed with the sgRNAs according to the manufacturer’s instructions to carry out the transfection procedure. The efficiency of EMP3 knockdown was verified by Western blot and qPCR after 48 h. The sgRNA sequences for EMP3 designed by the CRISPR-editing platform are as follows (Supplementary Fig. 1):

5${}^{\prime }$-CAGGGAGAGUCCACCAGGAC-3${}^{\prime }$ (1#);

5${}^{\prime }$-UCUGUCCAUCCCAGUCCUGG-3${}^{\prime }$ (2#);

5${}^{\prime }$-UUUCCCAGGGAGAGUCCACC-3${}^{\prime }$ (3#).

For cell proliferation assays, 2 $\times$ 10${}^3$ cells from each group (sg-EMP3 and CON) were seeded in a 96-well plate. Cell counting kit-8 (Bio-Rad Laboratories, Hercules, CA, USA) was diluted to 10% using DMEM and added to the wells to measure the OD value at 450 nm using a microplate reader.

The complete medium for culturing tumor cells in six-well plates was replaced with a serum-free medium. Then a 200 $\mu$L pipette tip was used to draw a straight wound, and the wells were washed three times to remove cell debris. The scratch images at 0 h and 24 h were taken and recorded under a microscope (Olympus; Tokyo, Japan) at 100 $\times$ magnification by selecting three random fields of view.

In the migration assays, we inoculated 2 $\times$ 10${}^4$ cells into the upper chamber of a 24-well transwell chamber (Costar, 3422; Corning Incorporated, NY, USA) and incubated them in a serum-free medium for 24 hours. At the same time, 600 $\mu$L complete medium (supplemented with 10% FBS) was added to the bottom of the chamber. The chambers were soaked in 20% methanol, fixed for 10 minutes, and then stained in 0.5% crystal violet solution for 15 minutes. After washing several times, the cells on the upper surface were gently wiped off with a cotton swab, and the bottom surface of the chambers was photographed and counted under a 200 $\times$ light microscope.

Similarly, in the invasion assays, 8 $\times$ 10${}^4$ cells were seeded into the upper chamber containing a non-serum medium pre-paved with Matrigel (Thermo Fisher Scientific; Waltham, MA, USA). Simultaneously, 600 $\mu$L complete medium supplemented with 10% FBS was added under the chamber to form a nutrient induction trend. After 24 h of co-incubation in a humid environment, the invasive cells were washed, fixed and stained, as mentioned before. Subsequently, five random fields were imaged to count the number of cells under a light microscope (Olympus; Tokyo, Japan). This experiment was independently conducted more than three times.

The cells were cultured in a six-well plate to about 30% confluence, washed with PBS and fixed with 10% formaldehyde for at least 20 minutes. After washing again, Oil Red O (Wuhan Service Biotechnology Company; Wuhan, Hubei, China) was added to each well for 30 minutes. According to the manufacturer’s instructions, a triglyceride determination kit (Nanjing Jiancheng Institute of Bioengineering; Nanjing, Jiangsu, China) was used to detect triglyceride content in cancer cells.

The “Gene Module” in the TIMER 2.0 database (http://timer.cistrome.org/) was explored to analyze the relationship between EMP3 and tumor immune cell infiltration [20]. Furthermore, “Gene_Corr Module” was a gene co-expression model constructed based on the expression data in the TCGA database. It can be used to view the Spearman’s expression correlation between target genes and immune checkpoints.

SPSS statistics version 25.0 (version 21.0, IBM Corp, Chicago, IL, USA) and GraphPad Prism (version 6.0, GraphPad Software Inc, San Diego, CA, USA) were employed for all data analysis. Unpaired or paired Student’s t-test was utilized to distinguish the difference between two groups. Furthermore, differences among multiple groups were evaluated by one-way analysis of variance (ANOVA; #). The Kaplan-Meier curve analysis and nomogram were drawn to analyze the survival probability of ccRCC patients according to the expression level of EMP3. The diagnostic value of EMP3 expression level was clarified using the receiver operator characteristic (ROC) curve. The prognostic role of EMP3 was investigated using the Cox proportional hazard regression and displayed using forest plots. Statistical data were presented as the mean $\pm$ standard deviation (SD). The data differences were considered statistically significant when the calculated p-value 0.05.

The EMP family has been reported to be related to various pathological processes; however, its role in ccRCC is not clear. To investigate the expression of EMP family members in ccRCC, the TCGA-KIRC dataset was analyzed and revealed that EMP3 mRNA expression in ccRCC was significantly elevated compared to corresponding noncancerous tissue samples (Fig. 1A,B). Furthermore, four gene expression datasets were downloaded from Oncomine for further validation; EMP3 was found to be abnormally upregulated in ccRCC (Fig. 1C). Based on these findings, we sought to further investigate the role of EMP3 in ccRCC.

Fig. 1.

EMP3 is highly expressed in ccRCC via database analysis. (A) Heat map depicting EMPs expression in TCGA-KIRC microarray datasets (n = 605). (B) Relative EMPs mRNA expression in TCGA-KIRC. (C) EMP3 mRNA expression in four Oncomine subsets, including Jones, Gumz, Yusenko and Lenburg renal statistics. **p 0.01, ***p 0.001, and ****p 0.0001.

To explore the difference in EMP3 mRNA expression among clinical characteristics of ccRCC patients, including sex, age, tumor grade and TNM stage, data extracted from the TCGA database was mined and compared (Table 1). EMP3 expression levels in tumors and normal tissues from the TCGA-KIRC database were compared using Student’s t-test (Fig. 2A,B). The results showed that EMP3 expression was elevated in paired or unpaired tissue samples of ccRCC. Further subgroup analysis revealed that high expression of EMP3 was positively associated with male (Fig. 2C), advanced T stage (Fig. 2D), TNM stage (Fig. 2E) and histological grade (Fig. 2F). The comparison of grouped expression levels made it easier to obtain the relationship between clinicopathological features and EMP3 expression in ccRCC.

Fig. 2.

Highly expressed EMP3 is closely related to multiple clinicopathological parameters in ccRCC patients. The files of EMP3 mRNA expression level and ccRCC patient’s clinical information were downloaded from TCGA-KIRC datasets and analyzed further in the subgroups: (A) total samples, tumor vs normal. (B) Paired samples, tumor vs normal. (C) Gender. (D) Tumor stage. (E) TNM stage. (F) Pathological grade. Statistical data are presented as the means $\pm$ standard deviation. *p 0.05, **p 0.01 ***p 0.001, and ****p 0.0001, #(ANOVA).

Parameters, including metastasis, tumor stage, histological grade, TNM stage and disease recurrence, are important factors that influence treatment selection during clinical practice. All included ccRCC patients (n = 533) were stratified according to the above parameters. EMP3 expression yielded good predictive performance in differentiating between ccRCC and healthy samples with an area under the curve (AUC) of 0.9356 (95% CI: 0.9063 to 0.9649, p 0.0001) (Fig. 3A). The AUC value of paired samples was 0.9418 (95% CI: 0.9008 to 0.9829, p 0.0001) (Fig. 3B). After stratification, the AUCs for EMP3 expression in differentiating between (T1 + T2) vs. (T3 + T4), (M0 + MX) vs. (M1), (G1 + G2) vs. (G3 + G4), TNM (I + II) vs. (III + IV) and nonrecurrence vs. recurrence were 0.6288, 0.6577, 0.6602, 0.6580 and 0.6671, respectively. Statistically significant differences were observed in all groups during subgroup analysis (Fig. 3C–G, p 0.0001). Based on these observations, we concluded that EMP3 could be used as a potential diagnostic biomarker for ccRCC patients.

Fig. 3.

EMP3 is a biomarker of ccRCC with diagnostic value. (A) ROC curve analysis of EMP3 expression indicates that it could discriminate ccRCC and para-cancer tissues in an effective manner. Similar results present in other subgroups. (B) Paired samples. (C) Distant metastasis. (D) Pathological grade. (E) TNM stage. (F) Recurrence status. (G) Tumor stage.

After grouping the patients, Kaplan-Meier (KM) analysis was conducted to assess the prognostic value of EMP3 in terms of overall survival. Higher EMP3 expression correlated with shorter OS in ccRCC patients (Fig. 4A, Supplementary Fig. 2). In order to verify the prognostic significance of EMP3 in ccRCC, another database ArrayExpress (E-MTAB-1980) was explored and a consistent conclusion was obtained (p = 0.0115, Supplementary Fig. 3). Univariate and multivariate analyses showed that dysregulated EMP3 expression could predict poor prognosis of ccRCC (Fig. 4B). Finally, we established a clinical prediction model based on EMP3 expression and other prognostic parameters screened by the multivariate Cox analysis. As shown in Fig. 4C, all independent predictors of OS including age (HR, 1.672; p = 0.001), T stage (HR, 1.568; p = 0.014), N stage (HR, 1.938; p = 0.049), M stage (HR, 2.749; p = 0.000), tumor grade (HR, 1.653; p = 0.007) and EMP3 expression level (HR, 1.511; p = 0.013), were integrated in a prognostic model to individually predict survival of ccRCC patients. The AUCs of overall survival of ccRCC patients at 1-, 3- and 5-years were respectively estimated as 0.6516, 0.6324 and 0.6229. To assess the accuracy of the prediction model, the 1-, 3- and 5-years calibration plots were created to display the degree of concordance between observed and predicted survival outcomes. The calibration plots showed optimal agreement with a concordance index of 0.760 (95% CI: 0.689–0.831), exhibiting the good predictive power of our model (Fig. 4E). In conclusion, our research findings suggested that EMP3 is a robust biological marker for predicting prognosis in ccRCC patients.

Fig. 4.

EMP3 is a biomarker of ccRCC with prognostic significance. (A) Kaplan-Meier curves showed that different expression levels of EMP3 were related to overall survival (OS) of ccRCC patients. (B) Univariate and multivariate analyses between EMP3 mRNA level and overall survival of ccRCC patients. (C) The nomogram for predicting the probability of 1-, 3-, and 5-year overall survival in ccRCC patients. (D) The time-dependent ROC curves of 1-, 3-, and 5-year overall survival probabilities for ccRCC patients were predicted by nomogram. (E–G) The calibration curves of overall survival probabilities of 1-, 3-, and 5-year were predicted by nomogram in ccRCC patients.

To further validate high EMP3 expression in ccRCC, we analyzed its expression in ccRCC cell lines and tissues. Consistent with our in silico analysis findings, EMP3 expression was significantly increased in ccRCC cell lines compared to normal human renal epithelium cell line 293 (Fig. 5A,B). Consistent results were obtained in ccRCC cell lines; both EMP3 mRNA and protein expression in tumor tissues were abnormally upregulated compared with neighboring normal tissues (Fig. 5C,D). In addition, immunohistochemistry confirmed high EMP3 protein levels in ccRCC tissue specimens (Fig. 5E). Taken together, by integrating bioinformatics and experimental analyses, we further substantiated that EMP3 was upregulated in ccRCC.

Fig. 5.

EMP3 is significantly upregulated in ccRCC cell lines and tissues. The expression of EMP3 in ccRCC clinical samples and cell lines. (A,B) Analysis of EMP3 expression in ccRCC cell lines using Western blot and qRT-PCR. (C–E) Verification of EMP3 expression in clinical tissues using qRT-PCR, Western blot and IHC. *p 0.05; **p 0.01; ***p 0.001; ****p 0.0001.

Studies have shown that the dysregulation of EMP3 expression played an oncogenic role in various malignant tumors. To substantiate that the biological functions of EMP3 in ccRCC, we successfully knocked down EMP3 in both ccRCC cell lines (A498 and 786-O) by plasmid transfection and then quantified EMP3 protein and mRNA levels (Fig. 6A,B). Subsequently, we explored the functional impact of EMP3 knockdown on the two ccRCC cell lines. As shown in Fig. 6C,D, the cell proliferation rate was significantly decreased in the sg-EMP3 group compared with the NC groups. Moreover, silencing EMP3 significantly reduced the number of cancer cells stained on the bottom of the transwell (Fig. 6E,F). Similarly, after 24 hours of serum-free culture, the migration ability of tumor cells in the knockdown EMP3 group was significantly weakened compared with the control group (Fig. 6G,H). Taken together, it can be inferred that EMP3 acts as an oncogene in ccRCC by promoting the proliferation, migration and invasion of cancer cells.

Fig. 6.

Knockdown of EMP3 inhibits the malignant phenotypes of ccRCC in vitro. (A,B) The expressions of EMP3 mRNA and protein were successfully knockdown in A498 and 786-O cells. (C,D) CCK-8 assays showing cell growth curves of A498 and 786-O cells. (E,F) Representative images of migration and invasion assays performed using A498 and 786-O cells; the number of cells was counted in five randomly selected images from each group. (G,H) Wound healing of EMP3-knockdown A498 and 786-O cells at 0 hour and 24 hours (scale bar =200 $\mu$m). *p 0.05; **p 0.01; ***p 0.001; ****p 0.0001.

To determine the regulatory mechanisms of EMP3 in ccRCC, GSEA was used to explore the enrichment signaling pathways and possible biological functions of EMP3. As shown in Fig. 7A, fatty acid metabolism was inhibited. Previously, we documented that lipid consumption could inhibit the malignant phenotypes of ccRCC cells [21, 22]. Accordingly, oil red staining and triglyceride determination were performed in the present study to validate that knockdown of EMP3 could reduce intracellular lipid accumulation (Fig. 7B,C). GSEA revealed that EMP3 was significantly enriched in Epithelial-mesenchymal transition (EMT) and Myc Targets V2 signaling pathways. The correlation plot of gene expression demonstrated that EMP3 was negatively correlated with E-cadherin (CDH1) and positively correlated with VIM, SNAIL and N-cadherin (CDH2), which were confirmed by Western blot (Fig. 7D,E). The above findings indicated that EMP3 could increase tumor malignancy by promoting EMT and lipid accumulation in ccRCC.

Fig. 7.

EMP3 affects the malignant phenotypes of ccRCC by promoting EMT and lipid accumulation. (A) GSEA was performed to investigate the biological pathways that may be involved in the regulation of EMP3 based on the TCGA database. (B,C) Photomicrographs of Oil Red O staining of knockdown EMP3 cell lines and its negative control (Magnification: 200 $\times$ , scale bar = 50 $\mu$m). Relative TG (mmol/gprot) detected by triglyceride assay kit. (D) The correlations of E-cadherin, VIM, SNAIL and N-cadherin and EMP3 were performed using the TCGA-KIRC data. (E) These related EMT markers were tested by Western blot after knockdown of EMP3. *p 0.05; **p 0.01.

The TIMER 2.0 online tool was used to explore the relationship between EMP3 expression pattern and tumor immune infiltration in ccRCC. The results in Supplementary Fig. 4 showed that the expression of EMP3 had a significant correlation with a variety of immune cells, including CD8${}^{+}$ T cells (p = 3.04 $\times$ 10${}^{-6}$, Rho = 0.215), macrophages (p = 1.86 $\times$ 10${}^{-3}$, Rho = –0.145), neutrophils (p = 1.49 $\times$ 10${}^{-2}$, Rho = 0.113) and myeloid dendritic cells (p = 2.10 $\times$ 10${}^{-11}$, Rho = 0.305), which also had a significant negative correlation with tumor purity in KIRC (p = 4.67 $\times$ 10${}^{-6}$, Rho = –0.211). However, there was no significant correlation observed in CD4${}^{+}$ T cells and B cells (p 0.05).

Given that EMP3 expression was closely related to several immune cells, we focused on several characteristic immune checkpoints, including CD274 (p = 5.37 $\times$ 10${}^{-5}$, Rho = –0.174), PDCD1 (p = 1.69 $\times$ 10${}^{-16}$, Rho = 0.347), CTLA4 (p = 8.83 $\times$ 10${}^{-11}$, Rho = 0.276), LAG3 (p = 2.54 $\times$ 10${}^{-21}$, Rho = 0.395), CD160 (p = 1.21 $\times$ 10${}^{-1}$, Rho = 0.067) and IDO2 (p = 1.24 $\times$ 10${}^{-2}$, Rho = 0.108). The results suggested that the up-regulated EMP3 in ccRCC may be closely related to T cell exhaustion, thereby promoting tumor immune escape.