Gene transcriptome profiles and related clinical data of ccRCC patients were downloaded from The Cancer Genome Atlas (TCGA; http://cancergenome.nih.gov/) website, containing 539 ccRCC samples and 72 adjacent normal tissue specimens. Subsequently, we excluded the cases that lack key clinical information and matched the ADAMTS14 gene matrix with corresponding clinical information for further analyses. All analyses were carried out by applying the R software (https://www.r-project.org/). The R package “Limma” was used to evaluate the different expression level of ADAMTS14 mRNA in TCGA ccRCC patients. Ultimately, the cutoff criteria of differently expressed genes were set as the adjusted p-value (FDR) <0.05 and |log2 fold change (FC)| ≥1.

To seek out potentially ADAMTS14-related pathways, GSEA (21) was applied with the help of the gene set “c2.cp.kegg.v7.1.symbols.gmt” of Kyoto Encyclopedia of Genes and Genomes (KEGG) (22). Moreover, it was considered statistically significant, when the |normalized enrichment score (NES)| >1.5 and the nominal p-value < 0.05. To further investigate possible molecular functions of ADAMTS14 in ccRCC, GO analysis was used for further analysis, including molecular function (MF), cellular component (CC), and biological process (BP).

To further investigate the relationship between ADAMTS14 and overall survival (OS), we performed univariate and multivariate Cox regression analyses to find whether or not ADAMTS14 and other clinical characteristics (staged M; gender; staged T; race; age; grade; staged N; stage) could serve as independent factors related to OS for ccRCC patients in TCGA database, with the threshold of p-value < 0.05 (23).

In order to make our results more credible at the histological level, we collected a total of 10 pairs of tissues from ccRCC patients who underwent radical nephrectomy from the Affiliated Hospital of Nantong University to conduct immunohistochemistry. In addition, we questioned all of these patients and all of them denied having received special treatment like medicine, radiotherapy, and immunotherapy before nephrectomy. Then, we performed routine pathology and immunohistochemistry on the cancer and adjacent normal tissue specimens of ccRCC patients to confirm pathological types. The ADAMTS14 antibody stained in immunohistochemistry was from Abcam (ab198885). After removing the paraffin, hydrating, and blocking, we added the specimen to the anti-ADAMTS14 goat polyclonal antibody (diluted at a ratio of 1:100), and incubated overnight at 4°C. By the application of microscope, we evaluated all sections by comparing the staining between each kidney cancer and adjacent specimens.

In order to obtain total RNA from ccRCC cells, we firstly added 1 ml of Trizol reagent (Life Technology, USA) to the cell culture flask of HK-2, 769-P, and CAKI-1 at ice-cold temperatures, and then applied the Thermo Scientific K1622 Revert Aid First Strand cDNA Synthesis Kit. Finally, qRT-PCR was implemented by ABI QuantStudio5 Real-Time PCR System with the measurement of Fast SYBR^® Green Master Mix and calculated by the 2^-ΔΔCt method. All relevant primers in this study were acquired from Ribobio (Guangzhou, China), containing ADAMTS14 (F 5’-TCTGAAAGCTGCACACTGCT-3’, R 5’-RGGACCAAGCACCAGAAACAT-3’) and β-actin (F 5’-ATGACTTAGTTGCGTTACACC-3’, R 5’-GACTTCCTGTAACAACGCATC-3’).

To further forecast the OS of ccRCC patients, we constructed an effective nomogram by utilizing the R package “rms”, and it consists of independent prognostic factors. After dividing points to every factor, we summed up all parameters’ points to calculate total points. Moreover, we applied the concordance index (C-index), receiver operating characteristic (ROC) curves, and calibration curves to evaluate the performance of our established nomogram (24).

As previously described (25), we marked tumors as MSI if more than two markers of all five showed the MSI to identify autosomal microsatellite tracts by MISA (http://pgrc.ipk-gatersleben.de/misa/misa.html). Moreover, based on the TCGA database, we calculated the TMB by comparing the gene mutation data between tumor tissues and adjacent normal tissue specimens (26). Moreover, we performed HLA without changing the original settings to detect the TNB based on the TCGA expression dataset (27). All these mentioned analyses were conducted by applying online Sangerbox tools (http://www.sangerbox.com/tool).

Correlation analyses between ADAMTS14 and six different immune cell infiltration levels were carried out by TIMER website (https://cistrome.shinyapps.io/timer/) (28). We implemented the ESTIMATE algorithm to assess the relationships between ADAMTS14 and tumor microenvironment, including the Immune, ESTIMATE, and Stromal scores (29). Furthermore, immune cells, mismatch repair proteins, and immune checkpoint molecules were separately assessed based on the expression levels of ADAMTS14 in TCGA database. Moreover, they were performed using the online Sangerbox tools as well.

We applied the Tumor Immune Dysfunction and Exclusion (TIDE; http://tide.dfci.harvard.edu/) to evaluate the potential clinical effects of immunotherapy in the groups with different ADAMTS14 expression. Moreover, the higher the TIDE score, the higher the possibility of immune exclusion, indicating a lower probability of patients benefiting from immunotherapy (30). The immunophenoscores of ccRCC patients from The Cancer Immunome Atlas (TCIA; https://tcia.at/) database were also utilized to assess the immune properties of ADAMTS14 in ccRCC (31).

To further explore the potential mechanisms of ADAMTS14 in ccRCC, we revealed several LncRNA/RBP/ADAMTS14 mRNA networks. Firstly, starBase v2.0 was utilized to predict ADAMTS14 targeted RBPs, with the threshold of strict stringency (≥5) and pan-Cancer ≥10 cancer types. After that, we used starBase database v2.0 once again to predict a selected RBP-targeted LncRNAs, with the threshold of strict stringency (≥5), pan-Cancer ≥10 cancer types, hub LncRNAs in TCGA ccRCC (p-value < 0.05, |log2 FC| ≥ 1 and FDR < 0.05), and LncRNAs positively correlated with ADAMTS14 in TCGA ccRCC (corFilter = 0.3 and pFilter = 0.001). Finally, the LncRNA/RBP/ADAMTS14 mRNA networks were identified and visualized by Cytoscape 3.6.1 software.

ADAMTS14 mRNA expression levels in pan-Cancer from TCGA are displayed in Figure 1A . Moreover, detailed ccRCC sample information from TCGA was shown in Supplementary Table 1 and the mRNA expression levels of ADAMTS14 in pan-Cancer from TCGA and GTEx datasets are shown in Supplementary Figure 1 , which contains more adjacent normal tissue specimens. As shown in Figure 1B , ADAMTS14 displayed a higher expression in ccRCC tumor tissues than in adjacent normal tissues (p-value <0.001). Pairwise boxplot and ICGC dataset boxplot suggested similar results (both p-values <0.001, Figures 1C, D ). The outcomes of qRT-PCR presented that the relative ADAMTS14 mRNA levels were highly expressed in CAKI-1 (renal cancer cell lines; p-value <0.0001), while it was not significantly downregulated in 769-P (renal cancer cell lines; p-value > 0.05), compared with HK-2 (normal kidney cell lines) by the 2^-ΔΔCt method ( Figure 1E ). In addition, we examined the expression patterns of ADAMTS14 in various cancer cell lines in the CCLE database containing renal cancer ( Supplementary Figure 2 ). According to the median expression, ccRCC patients were classified into high- and low-risk subclasses and the KM survival curves shed light on the fact that the low-ADAMTS14 groups could have a better OS than the high-ADAMTS14 groups (p-value <0.001; Figure 1F ). To further evaluate the diagnostic power of ADAMTS14, we conducted a ROC analysis and the 1-year, 3-year, and 5-year area under the curve (AUC) values were 0.728, 0.653, and 0.685, respectively, indicating a moderate diagnostic effect in diagnosing ccRCC based on the expression of ADAMTS14 ( Figure 1G ). All in all, ADAMTS14 was highly expressed in ccRCC and correlated with poor prognosis, suggesting that this gene may be a key gene in ccRCC.

To further verify the ADAMTS14 protein expression level of tissues in ccRCC, we selected a total of 10 patients undergoing radical nephrectomy from the Affiliated Hospital of Nantong University in the past 2 years. Then, we performed routine pathology and immunohistochemistry on the cancer and para-cancerous tissue samples of ccRCC patients. As detailed in Figure 2 , ADAMTS14 protein was positively expressed in ccRCC cancer tissues than in adjacent normal tissue specimens, and the numbers of patients included in each of the low-, medium-, and high-ADAMTS14 groups were 3, 3, and 4, respectively. Consistent with the mRNA expression levels of ADAMTS14, its protein expression levels were also elevated in ccRCC tumor tissues.

As shown in Figure 3 , elevated ADAMTS14 expression was significantly linked to gender (p-value = 0.0084), grade (p-value < 0.001), stage (p-value < 0.001), staged T (p-value < 0.001), staged M (p-value < 0.001), and staged N (p-value = 0.0063), while it was not related to age (p-value = 0.81). All of the abovementioned results suggested that ccRCC patients with elevated ADAMTS14 expression might be more likely to progress to advanced stages and develop metastasis, compared with patients with low ADAMTS14 expression.

We performed univariate and multivariate Cox regression analyses to identify independent factors highly related to the OS for ccRCC from ADAMTS14 and other clinical characteristics (staged M, gender, staged T, race, age, grade, staged N, and stage). Our outcomes showed that stage (HR = 1.914, p-value < 0.001), grade (HR = 1.338, p-value = 0.013), age (HR = 1.038, p-value < 0.001), and ADAMTS14 (HR = 1.185, p-value < 0.001) were finally screened out ( Supplementary Figure 3 and Table 1 ). Hence, four independent prognostic factors related to ccRCC were identified, consisting of stage, grade, age, and ADAMTS14.

To forecast the OS prognosis of ccRCC patients, we constructed an effective four-factor nomogram, based on the results of multivariate Cox analysis ( Figure 4A ). C-index and ROC curves associated with AUC values of 1, 3, and 5 years were 0.780, 0.858, 0.795, and 0.76, respectively, showing a moderate performance ( Figures 4B–D and Table 2 ). Calibration plots of 1, 3, and 5 years also suggested a consistency of actual and predicted results ( Figures 4E–G ). In summary, we successfully constructed a well-performing nomogram plot and predicted the prognosis of ccRCC patients well.

To seek out potentially ADAMTS14-related pathways, GSEA [21] was applied with the help of the gene set “c2.cp.kegg.v7.1.symbols.gmt”. According to the threshold of the |NES| >1.5 and nominal p-value < 0.05, we identified five signaling pathways that showed significantly different enrichment in the ADAMTS14 expression phenotype, including INSULIN, MTOR, PPAR, renal cell carcinoma, and renin–angiotensin system pathways ( Figure 5 and Table 3 ). We further performed Gene Ontology (GO) analysis including molecular function (MF), cellular component (CC), and biological process (BP) in Supplementary Figure 4 . As shown in Supplementary Figure 4 , ADAMTS14 was also proved to be significantly associated with immunity. Overall, biological functions and possibly signaling pathways associated with ADAMTS14 are revealed by our results, which were instructive for subsequent functional studies of ADAMTS14 in ccRCC.

Results of the PPI network suggested that a total of ten genes (COL5A3, COL11A1, COL1A2, COL5A1, COL1A1, THBS1, ADAMTS12, ADAMTS13, ADAMTS15, and ADAMTS20) were significantly related to ADAMTS14 ( Figure 6A ). Our outcomes also shed light on the fact that ADAMTS14 was significantly linked to TMB (p-value <0.001), while it was not related to MSI (p-value = 0.62) or TNB (p-value = 0.2) ( Figures 6B–D ). All in all, TMB might be a vital factor involved in the roles of ADAMTS14 in ccRCC.

Based on the cutoff values of correlation coefficient greater than 0.3 and p-value less than or equal to 0.001, we investigated the potential relationships between ADAMTS14 and six immune cells’ infiltration levels and found that CD4+ T cells and neutrophil cells were significantly correlated with ADAMTS14 expression, while other immune cells were not ( Figure 7A ). As for the tumor microenvironment, significant results were also shown between ADAMTS14 and ESTIMATE, Stromal, and Immune scores ( Figure 7B ). Moreover, the ADAMTS14 expression was highly correlated to methyltransferases including DNA methyltransferase 2 (DNMT2), DNA methyltransferase 3A (DNMT3A), DNA methyltransferase 1 (DNMT1), and DNA methyltransferase 3B (DNMT3B) in the diverse tumors especially in ccRCC (all p-values < 0.01, Figure 7C ). Thus, ADAMTS14 was highly related to immunity and methyltransferases, providing potential guidance for immunotherapy.

To further analyze the correlations between ADAMTS14 and immune checkpoint molecules, our results showed that ADAMTS14 were markedly related to immune checkpoint molecules in ccRCC such as CD28, CD44, CD276, CD80, and CTLA4 (all p-values < 0.05, Figure 8A ). Moreover, ADAMTS14 was highly associated with several immune cells in ccRCC, such as activated dendritic cell, central memory CD8 T cell, central memory CD4 T cell, and activated CD4 T cell (all p-values < 0.05, Figure 8B ). As for mismatch repair proteins, ADAMTS14 was potentially associated with PMS2 and EPCAM mismatch repair proteins in ccRCC (both p-values < 0.01, Figure 8C ). All in all, ADAMTS14 was discovered to be remarkably related to immune checkpoint molecules, mismatch repair proteins, and immune cells.

As for pan-cancer immune subtypes, our results indicated that ADAMTS14 was differently expressed in pan-cancer immune subtypes of C1, C2, C3, C4, and C6 (p-value = 0.001, Figure 9A ). The immunophenoscores of ccRCC patients from TCIA database (https://tcia.at/) were also utilized to assess the immune properties of ADAMTS14 in ccRCC ( Figure 9B and Supplementary Figure 5 ). Our results indicated that the low- and high-ADAMTS14 groups had significant immunogenicity for CTLA4 immunotherapy (p-value = 0.026), but not for PD1 immunotherapy (p-value = 0.075). Moreover, we applied the TIDE database to evaluate the potential clinical effects of immunotherapy in groups with different ADAMTS14 expression. As shown in Figures 9C–E , high expression of ADAMTS14 had a higher TIDE score, a higher T-cell dysfunction score, and a lower MSI score, indicating that these patients might have a lower efficacy and worse outcome after accepting the immunotherapy than ccRCC patients with low expression of ADAMTS14. In conclusion, our results demonstrated the predicted response of ccRCC patients when receiving immunotherapy. On the one hand, ccRCC patients with high ADAMTS14 expression shall have a poor prognosis when receiving immunotherapy. On the other hand, CTLA4 might be an effective target for ccRCC patients with low ADAMTS14 expression.

To reveal potentially ADAMTS14-related mechanisms in ccRCC, we presented several LncRNA/RBP/ADAMTS14 mRNA networks. As detailed in Figure 10A , starBase v2.0 was firstly utilized to predict ADAMTS14-targeted RBPs, with the threshold of strict stringency (≥5) and pan-Cancer ≥ 10 cancer types. After identifying the potential RBPs, starBase v2.0 was utilized once again to predict selected RBP-targeted LncRNAs, with the threshold of strict stringency (≥5), pan-Cancer ≥ 10 cancer types, hub LncRNAs in TCGA ccRCC (p-value < 0.05, |log2 FC| ≥ 1, and FDR < 0.05), and LncRNAs positively correlated with ADAMTS14 in TCGA ccRCC (corFilter = 0.3 and pFilter = 0.001). Figures 10B–D were utilized to identify selected RBP-targeted LncRNAs by Venn diagrams. Three Venn diagrams represent three RBP-involved axes (the LncRNAs/RBFOX2/ADAMTS14 axis, the LncRNAs/TAF15/ADAMTS14 axis, and the LncRNAs/TARDBP/ADAMTS14 axis) to find RBP-targeted LncRNAs. Finally, the potential LncRNA/RBP/ADAMTS14 mRNA networks were identified and visualized by Cytoscape 3.6.1 software in Figure 10E .