RNA-seq information for kidney renal clear cell carcinoma (KIRC) was obtained from TCGA. The gene expression information was obtained in TPM format; data from 613 samples (479 unpaired tumor samples, 72 unpaired normal samples, and 72 paired tumor and adjacent normal samples) were included. The expression levels were log2-transformed [log2(TPM + 1)]. The Wilcoxon rank-sum test and paired t test were individually applied to compare the expression of ERO1A between unpaired ccRCC samples and corresponding normal samples or between paired ccRCC tissues and normal samples.

A total of 406 patients who were pathologically diagnosed with ccRCC from two clinical units (cohort 1, n=82; Shanghai Pudong New Area Gongli Hospital; cohort 2, n=324; Shanghai Changhai Hospital) were included in this study. Two pathologists conducted double-blinded clinicopathological data analysis on all ccRCC patients and independently scored all ccRCC samples. In addition, 68 pairs of ccRCC tissue samples were subjected to real-time PCR experiments. The present study strictly followed the Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) (21). A summary of the clinicopathological features of the enrolled ccRCC patients is presented in Supplementary Table S1. The ccRCC patients (n=406) from the two clinical centers were randomly divided into training (n=203) and validation (n=203) cohorts. The primary outcomes of ccRCC patients were overall survival (OS) and progression-free survival (PFS). All the assays were authorized by the Institutional Ethical Review Board of Shanghai Gongli Hospital and Changhai Hospital, and written informed consent was obtained from the ccRCC patients.

Real-time PCR experiments were conducted as previously described (8). Total RNA from the cell lines was extracted with TRIzol (Gibco, 15596018). cDNA was synthesized using the PrimeScript One Step RT Reagent Kit (Takara, RR064A). Quantitative RT–PCR was conducted with SYBR Green Real-Time PCR Master Mix (Toyobo, QPK201) and a StepOnePlus Real-Time PCR System (Applied Biosystems). The experimental results were normalized to the expression of glyceraldehyde 3-phosphate dehydrogenase (GAPDH). The 2^-△△Ct method was applied to calculate the expression level of each target gene relative to that of the control.

The following primer sequences were used: ERO1A (forward primer, 5’-TTGGATCTGCTGGTGGTCAT-3’; and reverse primer, 5’-CCAGTGTAGCGCTCAGGATT-3’) and GAPDH (forward primer, 5’-TGGCACCGTCAAGGCTGAGAA-3’; and reverse primer, 5’-TGGTGAAGACGCCAGTGGACTC-3’).

IHC experiments were performed as previously reported (8). In brief, the samples were fixed in 4% methanol, embedded in paraffin, and subsequently cut into 5-µm-thick sections. Deparaffinization and rehydration were performed according to routine methods, and antigen recovery was performed in heated citrate buffer (pH 6.0) or EDTA buffer (pH 8.0) for 30 minutes. The tissue microarray slides were processed according to the UltraSensitive SP (Mouse/Rabbit) IHC Kit (KIT-9710; Maixin Biotechnologies). Specifically, each slide was incubated with endogenous peroxidase blocking solution for 30 minutes, and nonspecific binding sites were blocked with normal animal nonimmune serum for 20 minutes.

The slides were incubated with the following primary antibodies at 4 °C overnight: rabbit anti-ERO1A antibody (1:100; ab177156; Abcam), rabbit IgG antibody (1:100; Cell Signaling Technology), rabbit anti-CD163 antibody (1:500; GB115709; Servicebio), mouse anti-FOXP3 (1:200; ab2004; Abcam), and rabbit anti-CD8 (1:200; ab4055; Abcam). The slides were subsequently incubated with a biotin-labeled secondary antibody and Streptomyces antibiotic protein-peroxidase for 30 minutes. Afterward, diaminobenzidine (DAB) (DAB-2031; Maixin Biotechnologies) was used for staining.

The sections were subsequently incubated with hematoxylin and assessed by two pathologists in a double-blinded manner. ERO1A staining was scored according to the H-score approach; the H-score was calculated by multiplying the fraction of each component observed in the sample sections (as calculated by the intensity score [range 0–3] by the percentage of positive cells [range 0–100], with a total score range of 0–300) (8). The whole slide was first observed under a low-power microscope (40× or 100× magnification). Three representative fields of view were randomly selected under a high-power microscope (200× magnification) and scored according to the previous grading method. The mean value was subsequently calculated. For TIICs, the whole slide was first observed at ×40 or ×100 magnification. Three randomly selected areas of ccRCC were subsequently evaluated at ×200 magnification to score the density of stained TIICs. Finally, the mean value was calculated. The total cell count was calculated as the number of nucleated dyed cells per field and is presented as the density (cells/mm²) (8).

The cell lines were purchased from the Cell Bank of the Type Culture Collection of the Chinese Academy of Sciences in 2023. A498 and ACHN cells were cultured in minimum essential medium (Gibco, 11095-080), and 786-O cells were cultured in RPMI-1640 medium (Gibco, 22400-089). HK-2 cells were cultured in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) (Gibco, 11995-065). The culture media of the cell lines were supplemented with fetal bovine serum (FBS, 10%, Gibco) and 1% penicillin/streptomycin (Gibco). The cell lines were incubated at 37 °C in 5% CO₂. Sunitinib- and pazopanib-resistant 786-O cells were cultured in RPMI-1640 medium supplemented with 10% (v/v) FBS and 10 µM sunitinib or 8 µM pazopanib. All cell lines were authenticated by short tandem repeat (STR) profiling and examined for mycoplasma contamination via a Mycoplasma detection kit (Selleck Chemicals), and the most recent tests were completed in April 2024. The cell lines were used within 40 passages.

RNA interference was carried out following the manufacturer’s instructions using Lipofectamine™ RNAiMAX Transfection Reagent (Invitrogen). In brief, siRNA directed against ERO1A and a non-targeting control siRNA (obtained from Sangon Biotech (Shanghai, China) with sequences: si-ERO1A−1: 5’- GAAGGCTGTTCTTCAGTGGACC-3’; si-ERO1A −2: 5’- CCCTTGTAACCAGTGTAGCGCT-3’.) were prepared in RPMI 1640 Reduced Serum Medium. The diluted siRNA was combined with Lipofectamine™ RNAiMAX reagent and allowed to incubate at room temperature for 20 minutes to allow complex formation. This mixture was subsequently added to the cell cultures in a dropwise manner. Following a 6-hour incubation period, the transfection medium was exchanged for fresh complete medium. Cells were collected 48 to 72 hours after transfection for subsequent assays, such as RNA isolation for quantitative PCR (qPCR). To ensure experimental reproducibility, all procedures were conducted in triplicate.

Cell proliferation of ccRCC cells under specified conditions was assessed using a CCK-8 assay kit (CK-04, Dojindo, Kumamoto, Japan) following the manufacturer’s protocol, as previously described. Briefly, prior to measurement, the culture medium was refreshed, and 10% (v/v) CCK-8 solution was added to each well. After incubation at 37 °C for 2 hours, absorbance was measured at 450 nm using a microplate reader (EXL800, BioTek Instruments, Winooski, VT, USA). Proliferation rates were normalized to the control group and expressed as percentages relative to control values.

Cell invasion and migration were evaluated using transwell chambers (Millipore, Billerica, MA, USA) with or without Matrigel coating (BD Biosciences, NJ, USA), as previously reported. For the assay, 1×10⁴ 786-O or 769-P cells, or 1×10⁵ U937 cells, were suspended in serum-free RPMI-1640 medium and seeded into the upper chamber. The lower chamber contained RPMI-1640 medium supplemented with 20% fetal bovine serum (FBS) and conditioned medium (CM). After 36 or 48 hours of incubation, non-invading cells on the upper surface were gently removed with a cotton swab. Cells that migrated to the lower membrane surface were fixed with 4% paraformaldehyde (E672002, Sangon Biotech, Shanghai, China), stained with crystal violet (E607309, Sangon Biotech), and imaged at 200× magnification. Data are presented as mean ± standard deviation from three independent experiments.

Numerical data are presented as the means ± standard deviations (SDs). Statistical differences between variables were assessed via a two-tailed Wilcoxon test. Survival curves were plotted via the Kaplan–Meier method and compared via log-rank analysis. Variables with p values < 0.1 in the univariate analysis were included in the multivariate Cox proportional hazards analysis. Differences were considered significant at p < 0.05. Time-dependent receiver operating characteristic (ROC) curve analyses were used to calculate the cutoff values of the H-scores of ERO1A and CD163 with SPSS 26.0 (IBM Corporation) software. The prognostic accuracy of combining ERO1A expression, CD163⁺ TAM density and other indicators was examined via Harrell’s concordance index (C-index). All the analyses were conducted using GraphPad Prism 9.5 software (GraphPad Software, Inc.) and SPSS 26.0 (IBM Corporation) software.

RNA-sequencing data and clinical information for kidney renal clear cell carcinoma (TCGA-KIRC) patients were obtained from TCGA database. The ERO1A expression data were log2-transformed. Cox proportional hazards regression was conducted to evaluate the associations of ERO1A expression with OS, DSS, and the PFI. Kaplan–Meier survival curves were plotted, and the log-rank test was used to assess the statistical significance of differences between groups.

Analysis of ERO1A expression in ccRCC tissues using an online database (TCGA) revealed that ERO1A expression was higher in unpaired or paired ccRCC samples than in corresponding paracancerous samples (both p < 0.001; Figures 1A, B). Real-time PCR subsequently revealed that compared with the corresponding renal tissues, ccRCC tissues presented higher ERO1A expression (p < 0.001; Figure 1C). For validation, immunohistochemistry (IHC) was performed on ccRCC samples obtained from two hospitals (n = 406). Similar to the above findings, most ccRCC tissues presented high ERO1A expression (p < 0.001; Figure 1D), which was quantified according to the H score (refer to the Materials and Methods section for details). These findings demonstrated that ERO1A expression is consistently upregulated in ccRCC.

The relationship of ERO1A expression with malignant features of ccRCC was examined via IHC assays in ccRCC patients. As shown in Figures 2A, B, ERO1A expression was high in ccRCC samples with high TNM stages or SSIGN scores, whereas ERO1A expression was low in ccRCC samples with low TNM stages or SSIGN scores. Additionally, higher ERO1A expression was detected in ACHN cells (a metastatic ccRCC cell line) than in A498 or 786-O cells (nonmetastatic ccRCC cell lines) (Figure 2C). Elevated ERO1A expression was detected in ccRCC cell lines but not in HK-2 cells (a human renal tubular epithelial cell line) (Figure 2C). Moreover, compared with naïve cells, ccRCC cells treated with a targeted therapeutic drug (sunitinib or pazopanib) or resistant ccRCC cell lines presented higher ERO1A expression (Figures 2D, E). Therefore, these findings suggested that high ERO1A expression indicates malignant biological characteristics and strong resistance to targeted therapy in ccRCC. We next examined the biological function of ERO1A in ccRCC cells. First, siRNAs targeting ERO1A were employed in the 786-O and 769-P ccRCC cell lines (Supplementary Figure S1A). CCK-8 proliferation assay showed that compared with their respective control cells, ccRCC cells with ERO1A knockdown presented with decreased proliferation (Figures 2F, G). Moreover, ERO1A knockdown inhibited the invasion and migration abilities of ccRCC cells (Figures 2H, I).

To investigate whether ERO1A serves as a useful prognostic marker, data from ccRCC patients from two hospitals were collected (cohort 1, n = 82; cohort 2, n = 324) (Supplementary Table S1). The ccRCC patients were randomly divided into training and validation cohorts. An IHC experiment was conducted to assess ERO1A expression in samples from ccRCC patients (Figure 3A). The optimal cutoff value was utilized in the training cohort to divide ccRCC patients into subgroups on the basis of high and low ERO1A expression (Figure 3B). With 5-year OS as the end point, time-dependent ROC curve analysis revealed that the optimal cutoff value was 150 (AUC = 0.826) (Figure 3B). In the training cohort, the TNM stages (p < 0.001) and SSIGN scores (p = 0.001) were higher in the ERO1A^high subgroup than in the ERO1A^low subgroup (Table 1). Moreover, Kaplan–Meier survival assays revealed that compared with the ERO1A^low subgroup, the ERO1A^high subgroup had shorter OS (p < 0.001) and PFS (p < 0.001) (Figures 3C, D). Furthermore, the cutoff value in the training cohort was applied to the validation or combined cohort. Compared with the ERO1A^low subgroup, the ERO1A^high subgroup presented a higher TNM stage and higher SSIGN score but shorter OS and PFS (Figures 3E-H; Supplementary Tables S2, S3). To further validate the above findings, data from TCGA were analyzed, which revealed a close correlation between ERO1A upregulation and poor ccRCC prognosis in terms of OS, PFS, and disease-specific survival (DSS) (Figures 3I–K). Therefore, these results suggested that ERO1A expressions is a significant indicator of ccRCC patient survival.

The recurrence and metastasis of cancers are attributed to signal transduction pathways in malignant tumors and the regulatory network of interplay between cancer cells and tumor-infiltrating immune cells (TIICs) (22). To determine whether ERO1A expression is related to the infiltration of TIICs in ccRCC, IHC assays were performed to examine TIIC markers, such as CD163 (a TAM marker), CD8 (a marker for cytotoxic T lymphocytes, CTLs), and FOXP3 (a marker for regulatory T cells, Tregs), in ccRCC (23, 24) (Figure 4A; Supplementary Figures S1B, C). Correlation analysis revealed a positive correlation between ERO1A expression and CD163⁺ TAM density in ccRCC, but there was no correlation between ERO1A expression and the density of CD8⁺ CTLs or FOXP3⁺ Tregs (Figure 4A; Supplementary Figures S1B, C).

Because recent studies have demonstrated that combining intratumoral and TIIC markers increases the accuracy of survival prediction for ccRCC patients, we assessed whether the combination of ERO1A expression and CD163⁺ TAM density is beneficial for assessing ccRCC prognosis. Time-dependent ROC curves were constructed to determine the optimal cutoff values of ERO1A expression and CD163⁺ TAM density for classifying ccRCC patients in the training cohort (Figures 3B, 4B). The ccRCC patients were classified into the following four subgroups: ERO1A^highCD163^high, ERO1A^lowCD163^low, ERO1A^highCD163^low, and ERO1A^lowCD163^high (Figures 4C-H; Supplementary Table S4). As expected, the ERO1A^highCD163^high subgroup had a higher TNM stage (p = 0.002) and SSIGN score (p = 0.002) but also shorter OS and PFS (both p < 0.001) (Figures 4C, D; Supplementary Table S4). Furthermore, the above results were confirmed in ccRCC patients from the validation and combined cohorts (Figures 4E–H; Supplementary Table S5, S6). These results indicated that combining ERO1A expression and CD163⁺ TAM density is helpful for predicting ccRCC patient disease development and survival rates.

To conduct a more in-depth study of the ability of ERO1A expression and CD163⁺ TAM density to predict the survival rate of ccRCC patients, we assessed whether ERO1A expression and CD163⁺ TAM density are independent risk factors for assessing the survival rate of ccRCC patients via univariate and multivariate Cox regression analyses. As expected, ERO1A expression, CD163⁺ TAM density, TNM stage, and SSIGN score were confirmed to be independent risk factors for survival in ccRCC patients in the randomized training, validation, and combined cohorts, even after multivariable adjustment (Table 2; Supplementary Tables S7, S8). Therefore, these findings demonstrated that ERO1A expression and CD163⁺ TAM density serve as independent risk factors for predicting the survival rate of ccRCC patients.

Many studies, including our previous study, have indicated that integrating cancer markers, TIIC biomarkers, and TNM stage or the SSIGN score increases the accuracy of survival rate prediction for ccRCC patients (8). We next showed that the combination of ERO1A expression, CD163⁺ TAM density, and TNM stage or SSIGN score had a greater ability to predict the survival rate of ccRCC patients than any single marker alone. Time-dependent C-index assays were performed in the training cohort, which demonstrated that combining ERO1A expression and CD163⁺ TAM density increased the accuracy of the survival rate of ccRCC patients compared with any of the markers alone (Table 3). Furthermore, combining ERO1A expression, CD163⁺ TAM density, and TNM stage or SSIGN score resulted in the highest accuracy in evaluating ccRCC prognosis (Table 3), which was also validated in the validation and combined cohorts (Table 3). Taken together, these results indicated that combining ERO1A expression and CD163⁺ TAM density with TNM stage or SSIGN score increases the accuracy of predicting the prognosis of patients with ccRCC.