Adipogenic Transdifferentiation and Regulatory Factors Promote the Progression and the Immunotherapy Response of Renal Cell Carcinoma: Insights From Integrative Analysis

Abstract

Background: Adipogenic transdifferentiation was an important carcinogenic factor in various tumors, while studies on its role in clear cell renal cell carcinoma (ccRCC) were still relatively few. This study aimed to investigate its prognostic value and mechanism of action in ccRCC.

Methods: Gene expression profiles and clinical data of ccRCC patients were obtained from The Cancer Genome Atlas database. Nonnegative matrix factorization was used for clustering. Gene set variation analysis (GSVA) and gene set enrichment analysis (GSEA) were used to analyze the pathways and biological process activities. single-sample GSEA (ssGSEA) was utilized to quantify the relative abundance of each immune cell. Tumor Immune Estimation Resource (TIMER) was used to evaluate the proportion of various immune infiltrating cells across diverse cancer types. Real-Time PCR was performed to examine the gene expression. R software was utilized to analyze the expression and prognostic role of genes in ccRCC.

Results: A total of 49 adipose-related genes (ARGs) were screened for differential expression between normal and ccRCC tissues. Based on differentially expressed ARGs, patients with ccRCC were divided into two adipose subtypes with different clinical, molecular, and pathway characteristics. Patients in cluster A exhibited more advanced pathological stages, higher expressions of RARRES2 and immune checkpoint genes, higher immune infiltration scores, and less nutrient metabolism pathways. Adipose differentiation index (ADI) was constructed according to the above ARGs and survival data, and its robustness and accuracy was validated in different cohorts. In addition, it was found that the expression of ARGs was associated with immune cell infiltration and immune checkpoint in ccRCC, among which GBP2 was thought to be the most relevant gene to the tumor immune microenvironment and play a potential role in carcinogenesis and invasion of tumor cells.

Conclusion: Our analysis revealed the consistency of higher adipogenic transdifferentiation of tumor cells with worse clinical outcomes in ccRCC. The 16-mRNA signature could predict the prognosis of ccRCC patients with high accuracy. ARGs such as GBP2 might shed light on the development of novel biomarkers and immunotherapies of ccRCC.

Keywords: GBP2; adipogenic transdifferentiation; adipose-related gene; clear cell renal cell carcinoma; immunotherapy; prognostic model.

Figure 1

Association of RARRES2 mRNA expression with clinical parameters (data source: TCGA). (A) Type (wilcoxon test, p < 0.001); (B) Clinical stage (K-W test, p < 0.001); (C) Histological grade (K-W test, p = 0.001); (D) T classification (K-W test, p = 0.005); (E) Lymph nodes (wilicoxon test, p = 0.176); (F) Distant metastasis (wilicoxon test, p = 0.002).

Figure 2

Nonnegative matrix factorization clustering identified two adipose subtypes based on differentially expressed ARGs. (A) The expression of 49 differentially expressed ARGs between normal tissues and ccRCC tissues. Tumor, red; Normal, blue. The upper and lower ends of the boxes represented interquartile range of values. The lines in the boxes represented median value. Adjusted p < 0.05 and |log2 fold changes (FC)| > 1 were used as the criteria for screening differentially expressed ARGs. The asterisks represented the statistical p value (^* p < 0.05, ^** p < 0.01, ^*** p < 0.001). (B) The cophenetic, RSS and dispersion distributions with rank = 2–10; combining these indicators results in the optimal number of clusters of 2. (C) Consensus map of NMF clustering. (D) Survival analyses for the two adipose subtypes based on 530 patients with ccRCC from TCGA cohorts including 166 cases in cluster A, and 364 cases in cluster B Kaplan-Meier curves with log-rank p  = 0.003 showed a significant survival difference among two adipose clusters. The cluster A showed significantly worse overall survival than the cluster B. (E) NMF clustering of 49 ARGs in the TCGA ccRCC cohort. The adipose subtypes, TNM stages, clinical stages, survival status and age were used as patient annotations. Red represented high expression of ARGs and blue represented low expression. (F) Proportion of cases with different stages (chi-square test, p = 0.006) and difference of RARRES2 mRNA expression (wilcoxon test, p < 0.001) in the two adipose subtypes.

Figure 3

Biological and immune-related characteristics in different clusters. (A) GSVA showing the activation states of biological pathways in two clusters. The heatmap was used to visualize these biological processes, and red represented activated pathways and blue represented inhibited pathways. (B) The abundance of each TME infiltrating cell in cases in two clusters. The upper and lower ends of the boxes represented interquartile range of values. The lines in the boxes represented median value, and colorful dots showed outliers. The asterisks represented the statistical p value (^* p < 0.05, ^** p < 0.01, ^*** p < 0.001; ns, Not Statistically Significant). (C) The expression of ICP mRNAs in two clusters. The upper and lower ends of the boxes represented interquartile range of values. The lines in the boxes represented median value. The asterisks represented the statistical p value (^* p < 0.05, ^** p < 0.01, ^*** p < 0.001; NS, Not Statistically Significant).

Figure 4

Construction of ADI model. (A) 18 genes associated with OS of ccRCC patients were obtained through univariate Cox regression analysis. (B) A 16-mRNA signature was constructed by LASSO Cox regression. (C) Prognostic analysis of 16-gene signature in the training set. The dotted line represented the median risk score and divided the patients into low- and high-risk group. More dead patients corresponded to the higher risk score according to the curve of risk score and survival status of the patients. (D) Heatmap of the expression profiles of the 16 prognostic genes in low- and high-risk group.

Figure 5

Validation of the robustness of the model in both internal and external cohorts. (A–C, G) Kaplan–Meier survival analysis of the training ccRCC set, testing ccRCC set, whole ccRCC set, and validation set (NIHMS1611472 from NCBI). The survival rate of the patients in the high-risk group was significantly lower than those in the low-risk group respectively. (D–F, H) Time-dependent ROC analysis of the training ccRCC set, testing ccRCC set, whole ccRCC set, and validation set (NIHMS1611472 from NCBI). The AUC suggested that the prognostic accuracy of the 16-mRNA signatures in the discovery set was robust and accurate. (I) The nomogram for predicting proportion of patients with 3- or 5-year OS. The asterisks represented the statistical p value (*p < 0.05, **p < 0.01, ***p < 0.001).

Figure 6

Validation of the prognostic efficacy of our model under the stratifications of different clinical parameters. (A) age > 65 and age <=65, (B) male and female, (C) histological grade 1/2 and 3/4, (D) clinical stage I/II and III/IV, (E) T 1/2 and 3/4, (F) N 0 and 1, (G) M 0 and 1.

Figure 7

GBP2 was associated with the tumor microenvironment of ccRCC. (A) The correlation matrix of each immune checkpoint in modeling genes. (^* p < 0.05, ^** p < 0.01, ^*** p < 0.001). QuanTIseq analysis of RNA-seq data from 19 TCGA solid cancers: (B) Correlation of GBP2 expression with CD8⁺ T cells, (C) Correlation of GBP2 expression with immune checkpoint PD-1 and PD-L1. (D) Higher GBP2 expression was associated with more severe clinical parameters in ccRCC patients such as histological grade, clinical stage, T classification., lymph nodes, distant metastasis according to data from TCGA.

Figure 8

GSEA analysis revealed several activated oncogenic pathways associated with the expression of GBP2. (A) TNF-α signaling via NF-κB. (B) IL-6-JAK-STAT3 signaling. (C) KRAS signaling. (D) PI3K/AKT/mTOR signaling. (E) Apoptosis. (F) P53 pathway.

Figure 9

Validation of the role of GBP2 at the translational and transcriptional levels. (A) Immunohistochemical images from the HPA database show GBP2 protein expression in normal kidney (Normal) and KIRC (Tumor) tissues by different antibodies. (B) The mRNA expression of GBP2 was significantly different between normal kidney and ccRCC tissues according to the PCR results, as well as non-renal cancer cell line and several RCC cell lines. (C) The protein GBP2 expression of GBP2 was significantly different between normal kidney and ccRCC tissues according to the data from CPTAC. The asterisks represented the statistical p value (**p < 0.01, ***p < 0.001).