Identification and verification of a novel anoikis-related gene signature with prognostic significance in clear cell renal cell carcinoma

Abstract

Purpose: Clear cell renal cell carcinomas (ccRCCs) are the most common form of renal cancer in the world. The loss of extracellular matrix (ECM) stimulates cell apoptosis, known as anoikis. A resistance to anoikis in cancer cells is believed to contribute to tumor malignancy, particularly metastasis; however, the potential influence of anoikis on the prognosis of ccRCC patients is not fully understood.

Methods: In this study, anoikis-related genes (ARGs) with discrepant expression were selected from the Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. The anoikis-related gene signature (ARS) was built using a combination of the univariate Cox and least absolute shrinkage and selection operator (LASSO) analyses. ARS was also evaluated for their prognostic value. We explored the tumor microenvironment and enrichment pathways between different clusters of ccRCC. We also examined differences in clinical characteristics, immune cell infiltration and drug sensitivity between the high- and low-risk sets. In addition, we utilized three external databases and quantitative real-time polymerase chain reaction (qRT-PCR) to validate the expression and prognosis of ARGs.

Results: Eight ARGs (PLAUR, HMCN1, CDKN2A, BID, GLI2, PLG, PRKCQ and IRF6) were identified as anoikis-related prognostic factors. According to Kaplan-Meier (KM) analysis, ccRCC patients with high-risk ARGs have a worse prognosis. The risk score was found to be a significant independent prognostic indicator. According to tumor microenvironment (TME) scores, stromal score, immune score, and estimated score of the high-risk group were superior to those of the low-risk group. There were significant differences between the two groups regarding the amount of infiltrated immune cells, immune checkpoint expression as well as drug sensitivity. A nomogram was constructed using ccRCC clinical features and risk scores. The signature and the nomogram both performed well in predicting overall survival (OS) for ccRCC patients. According to a decision curve analysis (DCA), clinical treatment options for patients with ccRCC could be improved using this model.

Conclusion: The results of validation from external databases and qRT-PCR were basically agreement with findings in TCGA and GEO databases. The ARS serving as biomarkers may provide an important reference for individual therapy of ccRCC patients.

Keywords: Anoikis-related gene; Clear cell renal cell carcinoma; Drug sensitivity; Prognostic markers; Tumor microenvironment.

Fig. 1

Analysis of gene expression data for anoikis-related genes (ARGs) in ccRCC. A The heatmap showing the differentially expressed ARGs in TCGA-KIRC. B A volcano plot illustrating the differential expression of ARGs between tumors and normal tissue samples. Green dots indicate gene expression that is down-regulated in ccRCC and red dots denote gene expression that is up-regulated in ccRCC. C A forest map of prognostic ARGs obtained by uni-COX regression. D Copy number variations (CNVs) of ARGs. E An analysis of ARGs with CNVs located at different chromosomal locations. The red dot represents “gain” and the blue dot indicates “loss”

Fig. 2

Subgroups of ccRCC linked to ARGs. A Consensus matrixes were obtained for k = 2, 3, and 4. When k = 2, the CDF curve has the lowest slope. B–D Two subtypes were identified by PCA, tSNE and UAMP based on the expression of ARGs. E Kaplan-Meier (KM) survival analysis of two subgroups. F A heatmap showing clinical and pathological features of two subtypes according to the expression of ARGs. G The gene set variation analysis (GSVA) of the differences in KEGG pathways within subgroups A and B. H The top five pathways enriched in cluster B shown by gene set enrichment analysis (GSEA)

Fig. 3

Immune infiltration and gene expression analysis of differentially expressed genes (DEGs) in the two subgroups. A A ssGSEA analysis showing the patterns of immune infiltration in two subgroups. *$p<0.05$; **$p<0.01$; ***$p<0.001$. B Prognostic ARGs expression in two subgroups. *$p<0.05$; **$p<0.01$; ***$p<0.001$

Fig. 4

Establishment of a risk score signature based on ARGs. A Plots of the coefficient profiles of eight prognostic ARGs. B Eight prognostic genes were identified by least absolute shrinkage and selection operator (LASSO) analysis with 10-fold cross-validation. C The risk score of clusters A and B. D Survival plots illustrating each sample’s survival status in the training set. E Survival analysis of high and low-risk patients in the training set. F The heat map of 8 ARGs expression in the training set. G 1-, 3-, and 5-year ROC curves of the training set. H Survival plots illustrating each sample’s survival status in the testing set. I Survival analysis of high and low-risk patients in the testing set. J The heat map of 8 ARGs expression in the testing set. K 1-, 3-, and 5-year ROC curves of the testing set. L Survival plots illustrating each sample’s survival status in the all set. M Survival analysis of high and low-risk patients in the all set. N The heat map of 8 ARGs expression in the all set. O 1-, 3-, and 5-year ROC curves of the all set

Fig. 5

An analysis of the prognostic value of risk scores in ccRCC patients from the TCGA and GEO datasets. A Assessment of clinical features and risk scores using a multivariate Cox analysis. B Predictive nomogram for survival over 1, 3, and 5 years based on risk groupings and clinical characteristics. ***$p<0.001$. C Comparison of actual and predicted outcomes at 1, 3, and 5 years based on calibration curves. The decision curve analysis (DCA) of risk scores and clinical features at 1 (D), 3 (E), and 5 (F) years

Fig. 6

A comparison of the high- and low-risk tumor microenvironments (TME). A An illustration of the fractions of 22 types of infiltrating immune cells for low- and high-risk patients. B A relation between infiltrating immune cells and 8 prognostic ARGs, along with the risk score. C TME scores between the low- and high-risk patients, including stromal score, immune score, and estimate score. *$p<0.05$; ***$p<0.001$. D Relation between immune checkpoints and risk scores. *$p<0.05$

Fig. 7

Drug sensitivity analysis of patients with ccRCC. A Drugs showing more sensitivity to the high-risk patients with ccRCC. B Drugs showing more sensitivity to the low-risk patients with ccRCC

Fig. 8

ARGs in single-cell RNA sequencing. A Annotation of all cell types in GSE139555 and the percentage of each cell type. B, C Expression of PLAUR, HMCN1, CDKN2A, BID, PLG, PRKCQ and IRF6 in each cell type

Fig. 9

External datasets validation of the protein expression and prognosis of ARGs in ccRCC. A The protein expression of ARGs in ccRCC from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) dataset. ***$p<0.001$. B The expression of five proteins between normal and tumor was analyzed by immunohistochemistry (IHC) from the Human Protein Atlas (HPA) dataset. Scale bar = 100 $\mu m$. C Overall survival (OS) analysis of ARGs in ccRCC from the Kaplan–Meier (KM) Plotter dataset

Fig. 10

qRT-PCR validation of the mRNA expression of ARGs in ccRCC. Gene expression differences between ccRCC cell line 786-O and normal renal cell line HK-2 were confirmed by qRT-PCR. *$p<0.05$; **$p<0.01$ according to unpaired t test with Welch’s correction