TF-RBP-AS Triplet Analysis Reveals the Mechanisms of Aberrant Alternative Splicing Events in Kidney Cancer: Implications for Their Possible Clinical Use as Prognostic and Therapeutic Biomarkers

Abstract

Aberrant alternative splicing (AS) is increasingly linked to cancer; however, how AS contributes to cancer development still remains largely unknown. AS events (ASEs) are largely regulated by RNA-binding proteins (RBPs) whose ability can be modulated by a variety of genetic and epigenetic mechanisms. In this study, we used a computational framework to investigate the roles of transcription factors (TFs) on regulating RBP-AS interactions. A total of 6519 TF-RBP-AS triplets were identified, including 290 TFs, 175 RBPs, and 16 ASEs from TCGA-KIRC RNA sequencing data. TF function categories were defined according to correlation changes between RBP expression and their targeted ASEs. The results suggested that most TFs affected multiple targets, and six different classes of TF-mediated transcriptional dysregulations were identified. Then, regulatory networks were constructed for TF-RBP-AS triplets. Further pathway-enrichment analysis showed that these TFs and RBPs involved in triplets were enriched in a variety of pathways that were associated with cancer development and progression. Survival analysis showed that some triplets were highly associated with survival rates. These findings demonstrated that the integration of TFs into alternative splicing regulatory networks can help us in understanding the roles of alternative splicing in cancer.

Keywords: RNA-binding protein; TF–RBP–AS triplets; alternative splicing; regulation mechanism; transcription factor.

Figure 1

Overview of AS events profiling in KIRC. (A) Seven types of AS events: exon skip (ES), mutually exclusive exons (ME), retained intron (RI), alternate promoter (AP), alternate terminator (AT), alternate donor site (AD), and alternate acceptor site (AA). (B) UpSet plot of interactions between the seven types of detected AS events (n = 46,415) in KIRC. (C) Heat map of significant ASEs (n = 33). Horizontal axis shows clustering information of samples (normal or tumor); left longitudinal axis shows clustering information of ASEs. Gradual change in color from green to red represents PSI value of ASEs altered from low to high. (D) Splice graphs of some representative ASEs. Exons were drawn to scale, and connecting arcs represent splice paths. (E) Boxplots of four differentially expressed ASEs showing different expressions of AS events between KIRC and normal samples. Wilcoxon test was used for data comparison.

Figure 2

Identified triplets in KIRC. (A) Six mode subcategories of TFs according to correlation between expression of RBP and PSI value of ASE. Number in the pie chart means the percentage of each subcategory. (B) Clusters by the regulation patterns of TFs that are involved in the triplets containing the RBP MAP3K9. Six clusters were grouped according to TF subcategories. (C) Information of some representative triplets involved MAP3K9. (D) Correlation heat map of TFs involved in 6519 triplets. (E) Correlation heat map of RBPs involved in 6519 triplets.

Figure 3

Construction of protein–protein interaction network of triplets. (A) Protein–protein interaction network of 6519 triplets. Pink triangles represent TFs, green rhombuses represent RBPs, purple circles represent ASEs. Purple lines represent relationships between TF and RBP genes, green lines represent relationships between TF genes and ASEs, and blue lines represent the relationships between RBP genes and ASEs. (B) Triplet network involved in known PPI network in STRING.

Figure 4

Functional analysis of TF–RBP–AS triplets. (A) Survival analysis of three representative triplets. Red line represents the high-risk subgroup, and green line represents the low-risk subgroup. (B) Enrichment analysis of TF genes involved in 6519 triplets. (C) Enrichment analysis of RBP genes involved in 6519 triplets. (D) Cancer-relevant modulators identified according to the Network of Cancer Genes (NCG) and Tumor Suppressor Gene (TSGene) databases (* using Metascape for gene-enrichment analysis).

Figure 5

PCNA_58648_AP influenced by TFs and RBPs involved in triplets in KIRC. (A) Examples of some TFs and RBPs involved in triplets. (B) Four triplets influencing splicing of PCNA_58648_AP: four TFs (ETV7, HOXA7, ARNT2, BHLHE41) and four RBPs (IGF2BP2, PPARGC1A, RBM47, SAMD14). Red, samples in high TF expression group (top 40%); green, samples in low TF expression group (bottom 40%). X axis is the expression level of RBP, and Y axis is the PSI value of PCNA_58648_AP.

Figure 6

Research methodology. (A) Key ASEs and differentially expressed RBPs and TFs were identified, expression data of RBP and TF and PSI value of ASE were integrated, relevant clinical information of the patient from the data of TCGA-KIRC was extracted, and all data were integrated. (B) The linear mixed-effects model was used to predict triplets. Only triplets with significant β₁, β₂, and β₃ p values were considered and selected for the following analysis. Each triplet contains three objects: the expressions of TF and RBP and the PSI value of ASE. (C) For each triplet, we grouped samples into “low” and “high” groups on the basis of the expression level of TF (bottom/top 40% samples) in the specific triplet, and we compared Pearson’s correlation coefficient values of RBP expression and PSI value of ASE in two groups, identifying TF function categories. (D) In order to explore the prognostic value of these triplets, clustering analysis was performed, and a PPI network was constructed; enrichment and survival analyses were carried out.