A tissue microarray was purchased from Shanghai OUTDO Biotech Co., Ltd. (HkidE180Su02). It contained 144 cases of ccRCC tumor tissues and 29 cases of normal adjacent tissues available for statistical analysis after exclusion of tissue defects. The operation time range covered February 2008 to March 2010, the follow-up time ended in August 2015, and the follow-up period ranged from 5.5 to 7.5 years. Clinicopathological characteristics, including age, sex, tumor size, tumor grade, T classification, N classification, and M classification, were retrospectively collected. The T, N and M stages were determined according to the 7th edition of the American Joint Committee on Cancer (AJCC) TNM staging system.

For histological analysis, sections were stained with H&E. Briefly, sections were incubated with a hematoxylin solution for 10 min to stain nuclei, and then they were incubated with an eosin solution for 3 min and transferred into an alcohol gradient (70% ethanol for 1 min, 90% ethanol for 1 min, and 100% ethanol for 1 min). Finally, the sections were dehydrated and cleared in xylene.

Immunohistochemical staining was performed as described previously (26). Briefly, IHC was performed on the paraffin-embedded ccRCC tissues using Musashi-2 (Abcam, #ab76148), PD-L1 (GeneTech, #GT228002, Colone: ZR3), HIF1A (CST, (E1V6A), #48085), CA9 (MXB Biotechnologies, RAB-0615), KIT (MXB Biotechnologies, KIT-0029, Colone: YR145) and MKI67 (MXB Biotechnologies, MAB-0672, Colone: MX006) antibodies. IHC images were captured with a Leica Aperio Versa 200 microscope. Musashi-2 expression levels were scored by multiplying the percentage of positive cells by the staining intensity. The percentage was scored as follows: no positive cells, 0 point; 1%‐25%, 1 point; 26%‐50%, 2 points; 51%‐75%, 3 points; and 76%‐100%, 4 points. The staining intensity was scored as follows: no positive staining, 0 point; weak staining, 1 point; moderate staining, 2 points; and strong staining, 3 points. The final Musashi-2 IHC scores were classified as follows: 0‐4, low expression; and 6-12, high expression. For PD-L1 IHC evaluation, positive expression was defined as a tumor proportion score exceeding 1% (27).

The ccRCC data in the TCGA database were analyzed using the UALCAN platform (http://ualcan.path.uab.edu). UALCAN is an online platform that contains raw TCGA data, including gene expression and clinicopathological data (28). In this study, the UALCAN database was employed to analyze the mRNA and protein expression of Musashi-2 in primary ccRCC tissues and its relationships with stage and grade.

The prognostic value of Musashi-2 in ccRCC was evaluated using the TCGA datasets and the UCSC Xena Functional Genomics Explorer platform (https://xenabrowser.net/) (29). The Xena platform comprises gene expression profiles and survival information for multiple cancers, including survival and mRNA expression data for 607 ccRCC patients. The optimal cutoff value for survival was determined by the algorithms. In addition, samples lacking cancer results were excluded from the corresponding analyses.

The single-cell mRNA levels of Musashi-2 in the kidneys of normal humans were analyzed in data from The Human Protein Atlas (https://www.proteinatlas.org/). The mRNA expression in the cell type clusters identified in this study was visualized by a UMAP plot and a bar chart. Multiple transcripts of Musashi-2 were analyzed in normal kidney (GTEX) and TCGA ccRCC tissues by using the UCSC Xena platform (29).

Associations between immune cell infiltration and Musashi-2 expression were analyzed with the Tumor Immune Estimation Resource (TIMER) tool (https://cistrome.shinyapps.io/timer/). TIMER is an online tool for systematic analysis of immune infiltration in various cancers (30). In this study, the purity-corrected partial Spearman’s correlation (partial-cor) and p value provided by TIMER are shown in scatterplots, and immune infiltration estimations were calculated with immunedeconv-EPIC and CIBERSORT (31, 32). The correlations between immune inhibitors and Musashi-2 expression determined with the Xena platform are shown in heatmaps and scatterplots, and samples lacking results were excluded from the corresponding analysis.

The limma package in R software was used to study the differentially expressed mRNAs in the Musashi-2 High (n=265) and Low (n=265) groups from the TCGA ccRCC datasets. “Adjusted P < 0.05 and Fold Change >1.5 or Fold Change < −1.5” were defined as the threshold for differential mRNA expression. The Cluster Profiler package (version: 3.18.0) in R was employed to analyze the Gene Ontology (GO)-identified functions of potential targets and perform Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis (33). The genes (top 100) similar to Musashi-2 were obtained using the Gene Expression Profiling Interactive Analysis platform (GEPIA2, http://gepia.cancer-pku.cn/) (34). Similar gene visualization of the functional enrichment analysis results was achieved by using an online bioinformatic platform from China (http://www.bioinformatics.com.cn/).

To calculate signaling pathway scores, ccRCC RNA-sequencing (RNA-seq; FPKM) data were downloaded from the TCGA. The R software GSVA package (35) was used to perform single-sample gene set enrichment analysis (ssGSEA), and the correlations between Musashi-2 and pathway scores were analyzed by Spearman correlation analysis. For analysis of drug IC50 values, we predicted the therapeutic response for each sample based on the largest publicly available pharmacogenomic database [the Genomics of Drug Sensitivity in Cancer (GDSC), https://www.cancerrxgene.org/]. The prediction process involved the use of the R package “pRRophetic” (31), in which sample IC50 values were estimated by ridge regression and the prediction accuracy. All parameters were set to the default values with removal of the batch effect of combat and tissue type of all solid tumors, and duplicate gene expression data are summarized as the mean value. The potential therapeutic response to immune checkpoint blockade (ICB) was predicted with the TIDE algorithm as described previously (36). All the above analysis methods and R packages used the R foundation for statistical computing (2020) version 4.0.3.

Clinical data analysis was conducted using GraphPad Prism 7.5 software. Differences between two groups were compared by using the Mann–Whitney test. Correlations were determined using Pearson correlation coefficient tests. Survival curve were plotted by the log-rank test with the hazard ratio (HR) and 95% confidence interval. Chi-squared tests were applied to analyze the relationships between Musashi-2 expression and clinicopathological features. A p value <0.05 was considered to indicate a significant difference.

As an oncogene, Musashi-2 has been found to exhibit increased expression in a variety of tumors. However, there are currently no studies on Musashi-2 in ccRCC. Therefore, we first analyzed the expression of Musashi-2 across cancers, including colon adenocarcinoma (COAD), breast cancer (BRCA), esophageal cancer (ESCA), and lung adenocarcinoma (LUAD), etc ( Figure 1A ). Unlike the findings for other types of tumors, the mRNA and protein expression levels of Musashi-2 in normal kidney tissues were significantly higher than those in cancer tissues ( Figures 1B, C ). This finding was further confirmed with the GSE53757 and GSE66272 datasets (37, 38)( Figures 1D, E ). The associations of Musashi-2 mRNA expression with the stage and grade of ccRCC were also analyzed. As illustrated in Figures 1F, G , Musashi-2 mRNA expression was closely associated with the cancer stage and grade of ccRCC patients. We further assessed the prognostic value of Musashi-2 mRNA expression in patients with ccRCC. Surprisingly, lower mRNA expression levels of Musashi-2 were significantly related to poorer overall survival (OS), progression-free survival (PFS) and disease-specific survival (DSS) in patients with ccRCC ( Figures 1H , S1A, B ). Collectively, these results indicate that Musashi-2 is significantly downregulated in ccRCC and the lower expression predicts a poorer prognosis in ccRCC patients.

To explore why abnormal expression of Musashi-2 appears in the kidneys, we obtained a single-cell map of the kidneys from The Human Protein Atlas. Normal kidney tissue cells were divided into 14 single-cell subpopulations, including 9 proximal tubular cell clusters, 1 distal tubular cell cluster, 1 collecting duct cell cluster, and 3 immune cell clusters. The mRNA expression levels of Musashi-2 in the distal tubular cell cluster-c11, collecting duct cell cluster-c12 and proximal tubular cell cluster-c5 were higher than those in the other cell clusters ( Figures 2A, B ). We further analyzed all transcripts of Musashi-2 in normal kidney and ccRCC tissues. It was found that 14 of the 19 transcripts of Musashi-2 were mainly expressed in kidney, with the transcripts ENST00000284073.6 and ENST00000581523.5 accounting for a relatively large proportion (total proportion exceeded 80%) ( Figures 2C, D ). In addition, some transcripts including ENST00000581776.5, ENST00000579590.5 and ENST00000584476.1were only expressed in ccRCC; but they were scarcely expressed in the normal kidneys. Most importantly, the expression abundance (TPM) of the transcript ENST00000581523.5, accounting for a large proportion in the isoform percentage, was significantly reduced in ccRCC, whereas the expression abundance (TPM) of the transcripts ENST00000284073.6 was similar in the normal kidney and ccRCC ( Figures 2C, D ). We further assessed the expression of Musashi-2 by immunohistochemical staining of a TMA including 29 normal kidney tissues and 144 primary ccRCC tissues. Our immunohistochemical results were consistent with the above finding that the expression of Musashi-2 in normal kidney tissues was mainly distributed in the distal renal tubules and collecting ducts and was higher than that in cancer tissues ( Figures 2E, F ). In conclusion, the expression characteristics of Musashi-2 in normal kidney and ccRCC tissues are abnormal.

To explore the potential effects of Musashi-2 in ccRCC, we first analyzed the significance of Musashi-2 in our clinical TMA. Musashi-2 expression was significantly downregulated in tumor tissues compared with adjacent tissues ( Figure 3A , Table 1 ). Furthermore, as shown in Table 2 , Musashi-2 expression was closely associated with tumor grade (p=0.006) in ccRCC patients, while the associations with other clinical features were not significant. In our cohort of 144 cases, we found that the patients with higher expression of Musashi-2 also had a better prognosis (p=0.009) ( Figure 3B ). The results of univariate and multivariate Cox analyses indicated that Musashi-2 expression could act as an independent prognostic factor for ccRCC ( Table 3 ). Moreover, we performed immunohistochemical staining analysis of PD-L1 in the TMA. However, the patients with higher expression of PD-L1 had a poorer prognosis (p=0.0491) ( Figure 3C ). Notably, the patients with higher Musashi-2 expression had lower PD-L1 expression, and Musashi-2 expression was slightly negatively correlated with PD-L1 expression (p=0.0208, r=-0.1886) ( Figures 3D, E ). In conclusion, the expression of Musashi-2 in ccRCC is associated with clinicopathological characteristics and PD-L1, and can act as an independent prognostic factor for ccRCC.

In view of the fact that Musashi-2 expression predicted a better prognosis and was negatively correlated with PD-L1 expression, we speculated whether Musashi-2 affected the tumor immune microenvironment. We first compared the immune infiltration in TCGA ccRCC patients with high and low expression of Musashi-2 and found that CD4⁺ and CD8⁺ T cells were significantly increased in the high Musashi-2 expression group ( Figure 4A ). Then, we further analyzed the relationships between Musashi-2 expression and the infiltration of various types of cells in the tumor microenvironment. The results showed that the expression level of Musashi-2 was significantly positively correlated with the infiltration of CD8⁺ T cells (Rho=0.364) and CD4⁺ T cells (Rho=0.603) ( Figures 4B, C ) but negatively correlated with that of regulatory T (Treg) cells (Rho=-0.332) and natural killer (NK) cells (Rho=-0.523) ( Figures 4D, E ). Cancer-associated fibroblast (Rho=-0.121) and endothelial cell (Rho=0.189) infiltration was also evaluated ( Figures 4F, G ). Collectively, these results indicate that Musashi-2 expression is closely associated with immune infiltration, especially CD4⁺ T-cell, CD8⁺ T-cell and Treg-cell infiltration.

Treg cells play an important inhibitory role in antitumor immunity, and Muashi-2 expression was inversely correlated with PD-L1 expression and Treg-cell infiltration in ccRCC ( Figures 3D , 4D ). To explore how Musashi-2 affects tumor immune infiltration, we focused on the profiles of key immune inhibitors in ccRCC, and the relationships between Musashi-2 and immune inhibitors in 607 cases of ccRCC were evaluated. In cases with high expression of Musashi-2, the levels of inhibitory immunoregulators were significantly downregulated ( Figure 5A ). Further correlation analysis revealed that the expression of Musashi-2 was negatively correlated with the expression of TGFB1 (r=-0.6932), IL10RA (r=-0.5663), IL10RB (r=-0.6856), NT5E (r=-0.2887), ENTPD1 (r=-0.5296), IL2RA (r=-0.4217), FOXP3 (r=-0.4519), PDCD1 (r=-0.4836), CTLA4 (r=-0.4313), IDO1 (r=-0.5099), LAG3 (r=-0.5158), TIGIT (r=-0.5208), and HAVCR2 (r=-0.4046) ( Figures 5B–N ). Additionally, the roles of Musashi-2 and immune inhibitors in the prognosis of 602 ccRCC cases were also assessed, and high expression of Musashi-2 in ccRCC was associated with a better prognosis (p=0.02164), while high expression of immune inhibitors, including IL10RA (p=0.0723), IL10RB (p=0.01748), FOXP3 (p=0.00009978), PDCD1 (p=0.02439), CTLA4 (p=0.000272), LAG3 (p=0.001281) and TIGIT (p=0.01299), was associated with a poor prognosis ( Figures S2A–N ). In summary, these results indicate that Musashi-2 expression is negatively correlated with the expression of immune inhibitors.

Based on the above results, we inferred that Musashi-2 may regulate immune infiltration in ccRCC, but the biological function of this regulation remained unknown, leaving us to wonder whether it could affect signal transduction in related pathways. First, the associations between Musashi-2 expression and major signaling pathway scores were analyzed. The expression of Musashi-2 was negatively correlated with anti-inflammatory IL-10 (p =1.74e-10, cor =-0.27), the cellular response to hypoxia (p =2.22e-23, cor =-0.41), P53 signaling (p =7.75e-15, cor =-0.33), tumor proliferation (p =4.24e-17, cor =-0.35), angiogenesis (p =2.62e-12, cor =-0.30) and apoptosis (p =8.51e-11, cor =-0.28) ( Figures S3A–F ). Then, we further analyzed the association of Musashi-2 with metabolism-related eigengenes in ccRCC ( Figure S4A ). It was found that Musashi-2 expression was positively correlated with VHL (p =3.495e-23, r =0.3875), HIF1A (p =3.490e-24, r =0.3957) and KIT (CD117) (p =2.771e-58, r =0.5904) but negatively correlated with EPAS1 (p =6.674e-18, r =-0.3401), VEGFA (p =3.589e-57, r =-0.5857), SLC2A1 (p =8.422e-58, r =-0.5884), CA9 (p =4.302e-123, r =-0.7759), MKI67 (p =6.728e-35, r =-0.4714) and TP53 (p =3.270e-22, r =-0.3794) ( Figures S4B–J ). Among these genes, VHL, HIF1A, VEGFA, SLC2A1 and CA9 play critical functions in ccRCC development, tumorigenesis, energy metabolism and angiogenesis (2, 39). The immunohistochemical features of HIF1A, CA9, MKI67 and CD117 were further verified in our clinical specimens, and Musashi-2 expression was positively correlated with HIF1A, but negatively correlated with CA9 ( Figure S4K ).

More importantly, the differentially expressed genes between the Musashi-2 high and low groups in the TCGA ccRCC datasets showed that there were 574 upregulated genes and 197 downregulated genes ( Figures 6A, B ). The differentially expressed genes were then analyzed by KEGG and GO enrichment analyses. The upregulated KEGG pathways and GO terms mainly included valine, leucine and isoleucine degradation; carbon metabolism; and small molecule catabolic process. In contrast, the downregulated KEGG pathways and GO terms mainly included cytokine–cytokine receptor interaction, coronavirus disease-COVID-19 and regulation of leukocyte migration ( Figures 6C–F ). Additionally, the top 100 genes similar to Musashi-2 were obtained with GEPIA2 ( Supplementary Table 1 ). The biological functions of these genes were also explored by GO annotation and KEGG pathway analyses. As shown in Figure 6G , the top 10 KEGG pathways with p < 0.05 for the genes similar to Musashi-2 were determined. KEGG pathways such as hsa00020 (citrate cycle (TCA cycle)), hsa00190 (oxidative phosphorylation), hsa00010 (glycolysis/gluconeogenesis) and hsa00620 (pyruvate metabolism) were correlated with the functions of the genes similar to Musashi-2 in ccRCC. In addition, the GO annotation results are shown in Figure 6H . Biological process (BP) terms such as GO:0006099 (tricarboxylic acid cycle) and GO:0006086 (acetyl-CoA biosynthetic process from pyruvate), cellular component (CC) terms such as GO:0033176 (proton-transporting V-type ATPase complex) and GO:1990204 (oxidoreductase complex), and molecular factor (MF) terms such as GO:0044769 (ATPase activity, coupled to transmembrane movement of ions, rotational mechanism) and GO:0046961 (proton-transporting ATPase activity, rotational mechanism) were prominently related to the genes similar to Musashi-2. Taken together, these results indicate that Musashi-2 may regulate multiple signaling pathways, especially tumor metabolic processes and immune regulation in ccRCC.

At present, the first-line targeted TKIs for advanced or metastatic renal cancer mainly include sorafenib, sunitinib, temsirolimus, gefitinib, pazopanib and axitinib. We wondered whether the expression level of Musashi-2 is related to the clinical response to targeted drug therapy. We further predicted the IC50 values for first-line treatments for ccRCC in the high and low Musashi-2 expression groups. It was found that in ccRCC patients with high Musashi-2 expression, sorafenib, sunitinib, temsirolimus and gefitinib had lower IC50 values ( Figures 7A, B, E, F ), reflecting higher sensitivity to targeted drug treatment. In contrast, pazopanib and axitinib did not show significant differences ( Figures 7C, D ). In addition, consistent with the above conclusion that Musashi-2 expression was negatively correlated with immunosuppressive factors, TIDE scores were used to predict the tumor response to ICB treatment, and patients with lower Musashi-2 expression were found to be more responsive to ICB treatment ( Figure 7G ). Based on the above findings, we concluded that the expression level of Musashi-2 may guide responsiveness to clinical therapeutic options, such as TKI or ICB treatment.