Signature of gene aberrant alternative splicing events in pancreatic adenocarcinoma prognosis

Abstract

Alternative splicing (AS), as an effective and universal mechanism of transcriptional regulation, is involved in the development and progression of cancer. Therefore, systematic analysis of alternative splicing in pancreatic adenocarcinoma (PAAD) is warranted. The corresponding clinical information of the RNA-Seq data and PAAD cohort was downloaded from the TCGA data portal. Then, a java application, SpliceSeq, was used to evaluate the RNA splicing pattern and calculate the splicing percentage index (PSI). Differentially expressed AS events (DEAS) were identified based on PSI values between PAAD cancer samples and normal samples of adjacent tissues. Kaplan-Meier and Cox regression analyses were used to assess the association between DEAS and patient clinical characteristics. Unsupervised cluster analysis used to reveal four clusters with different survival patterns. At the same time, GEO and TCGA combined with GTEx to verify the differential expression of AS gene and splicing factor. After rigorous filtering, a total of 45,313 AS events were identified, 1,546 of which were differentially expressed AS events. Nineteen DEAS were found to be associated with OS with a five-year overall survival rate of 0.946. And the subtype clusters results indicate that there are differences in the nature of individual AS that affect clinical outcomes. Results also identified 15 splicing factors associated with the prognosis of PAAD. And the splicing factors ESRP1 and RBM5 played an important role in the PAAD-associated AS events. The PAAD-associated AS events, splicing networks, and clusters identified in this study are valuable for deciphering the underlying mechanisms of AS in PAAD and may facilitate the establishment of therapeutic goals for further validation.

Keywords: alternative splicing; alternative splicing factors.; genomic analysis; pancreatic adenocarcinoma; risk model.

Figure 1

Overview of seven types of alternative splicing and differentially spliced AS events analysis in pancreatic adenocarcinoma. (A) Schematic diagram of the alternative splicing type. (B) Number of alternative splicing events of each type and distribution of genes involved in alternative splicing events. Five alternative splicing types are marked in blue, reaching 6-7 alternative splicing types marked in red. (C) Differentially alternative splicing events involve Venn plots of genes and differentially expressed genes, showing the number of their isomorphisms. (D) Volcanic maps of differentially alternative splicing events, where blue represents a downregulated differential alternative splicing event, red represents an upregulated alternative splicing event, and grey represents a non-differential alternative splicing event.

Figure 2

Alternative splicing events statistics and GO functional enrichment analysis. (A) The Circos diagram of the survival-related alternative splicing events and its related genes, the Circos panel from the outside to the inside, is expressed as follows: chromosome number, genomic axis, survival-related alternative splicing event-related genes, names of related genes, the number of related genes occurring in the overall alternative splicing events (1-10 (>10), and showing 1-10 different heights, over 10 calculations on time), the number of alternative splicing types of related genes in the overall events, the p value of the relevant gene in the difference analysis (expressed by the conversion value of -log10 (p-values), and the higher the height, the more significant the p value), the fold change value of the relevant gene in the difference analysis (where red represents upregulation and black represents downregulation), correlation between genes. (B) A forest map of the most important TOP5 survival-related alternative splicing events in each of the alternative splicing types after single factor Cox regression analysis. (C) GO and KEGG Enrichment analysis results of survival-related alternative splicing event-related genes after one-way Cox regression analysis.

Figure 3

Kaplan-Meier plots and ROC curves of predictive Kaplan-Meier plots and ROC curves of predictive factors in the TCGA pancreatic adenocarcinoma cohort. (A)-(E) Kaplan-Meier curves plotted for prognostic models of each type of alternative splicing event. (F) Kaplan-Meier curves drawn from the prognostic model after integration of each type. The red line represents the high-risk group, and the blue line represents the low-risk group. (G) ROC curves for each type and post-integration alternative splicing event. (H) 3, 5, and 7 year ROC curves for alternative splicing events after integration.

Figure 4

Prognosis-associated molecular subtype cluster analysis. (A)-(B) Statistical analysis of Elbow for different numbers of clusters (k = 2 to 8) and PCA analysis for K=4. (C) The consensus matrix heat map defines four sample clusters with consensus values ranging from 0 (white, samples never gathered together) to 1 (dark blue, samples are always clustered together). (D) Survival analysis in the identified four sample clusters. (E) The distribution of each clinical information in four sample clusters.

Figure 5

Survival analysis and detail distribution of splicing factors. (A) Forest plots visualizing the p-value of 15 splicing factors identified by survival analysis of TCGA. (B) The details of the 15 splicing factors in the TCGA cohort.

Figure 6

Correlation analysis of splicing factors and prognosis-related AS predictors. (A) Correlation analysis between splicing factors and AS prognostic predictors. The upper panel shows the correlation between correlation coefficient and splicing factor expression and PSI values of prognostic-related AS events. The size and colour of the circle represent the weight of the correlation coefficient, * p <0.05, ** p <0.01, *** p <0.001, and the scatter plot shows the correlation between the expression of the splicing factor and the PSI value of the survival-related AS event. (B) Alternative splicing network: the square node is the splicing factor, the circular node is the gene involved in the prognosis-related alternative splicing event, the blue node is the protective factor, and the red node is the risk factor. (C) Significantly enriched GO term, significantly enriched KEGG or Reactome pathway.

Figure 7

Specific analysis of splicing factors ESRP1and RBM5. (A) ESRP1 is negatively correlated with FAM72A, and RBM5 is positively correlated with C11orf31. (B) GESA analysis of splicing factors ESRP1and RBM5. (C) Survival curves of the identified survival-associated splicing factors ESRP1 and RBM5, with the red line representing the high expression group and the blue line representing the low expression group.

Figure 8

Representative images for immunostaining of splicing factors ESRP1and RBM5. (A) Pie chart reports the numbers of PC tissues versus normal tissues with 'positive' or 'negative' ESRP1 staining. (B) Pie chart reports the numbers of PC tissues versus normal tissues with 'positive' or 'negative' RBM5 staining. (C) Immunostaining for ESRP1 of PC tissues and normal tissues (x200). (D) Immunostaining for RBM5 of PC tissues and normal tissues (x200).