Genome-Wide Transcriptional Analysis Reveals Alternative Splicing Event Profiles in Hepatocellular Carcinoma and Their Prognostic Significance

Abstract

Accumulating evidence indicates an unexpected role of aberrant splicing in hepatocellular carcinoma (HCC) that has been seriously neglected in previous studies. There is a need for a detailed analysis of alternative splicing (AS) and its underlying biological and clinical relevance in HCC. In this study, clinical information and corresponding RNA sequencing data of HCC patients were obtained from The Cancer Genome Atlas. Percent spliced in (PSI) values and transcriptional splicing patterns of genes were determined from the original RNA sequencing data using SpliceSeq. Then, based on the PSI values of AS events in different patients, a series of bioinformatics methods was used to identify differentially expressed AS events (DEAS), determine potential regulatory relationships, and investigate the correlation between DEAS and the patients' clinicopathological features. Finally, 25,934 AS events originating from 8,795 genes were screened with high reliability; 263 of these AS events were identified as DEAS. The parent genes of these DEAS formed an intricate network with roles in the regulation of cancer-related pathway and liver metabolism. In HCC, 36 splicing factors were involved in the dysregulation of part DEAS, 100 DEAS events were correlated with overall survival, and 71 DEAS events were correlated with disease-free survival. Stratifying HCC patients according to DEAS resulted in four clusters with different survival patterns. Significant variations in AS occurred during HCC initiation and maintenance; these are likely to be vital both for biological processes and in prognosis. The HCC-related AS events identified here and the splicing networks constructed will be valuable in deciphering the underlying role of AS in HCC.

Keywords: RNA-seq; alternative splicing; genome-wide; hepatocellular carcinoma; prognosis.

FIGURE 1

Landscape of alternative splicing (AS) events in hepatocellular carcinoma (HCC). (A) Diagrammatic sketch of the seven types of AS events in the present study: alternate acceptor site (AA), alternate terminator (AT), alternate promoter (AP), exon skipping (ES), mutually exclusive exons (ME), alternate donor site (AD), and retained intron (RI). (B) Number of AS events and the corresponding parent genes illustrated according to AS type (Left panel). The color bar represents AS events filtered by criteria. The black bar represents the corresponding genes involved in AS. Each AS type was divided into four groups based on the tissue source. N, normal tissue; T, tumor tissue; PT, paired tumor tissue; NPT, unpaired tumor tissue. Number of detected AS events, AS-related genes, filtered AS events, and the corresponding genes (right panel). (C) Intersection of parent genes between different AS types (n = 25,934) in HCC. One gene may incur up to four types of alternative splicing. (D) Circos plots depicting the distribution and the detailed alteration of AS events and their parent genes in chromosomes.

FIGURE 2

Identification of hepatocellular carcinoma (HCC)-related aberrant alternative splicing (AS). (A) Differences in AS events between paired HCC tissue and paracancerous tissue. Volcano plot of the differentially expressed alternative splicing (DEAS) identified in HCC; the blue and the red points represent the DEAS with statistical significance (|log(FC)| ≥ 1, adj p < 0.05). (B) Proportions of different AS types among filtered AS and DEAS. (C) Heat map of the DEAS. The horizontal axis shows the clustering information of samples divided into two major clusters: adjacent normal tissue (N = 50) and paired tumor tissue (N = 50); the left longitudinal axis shows the clustering information of DEAS. The gradual change of color from green to red represents the alteration of expression of DEAS from low to high. (D) Splice graph of some representative DEAS. The thin exon sections represent untranslated regions and the thick exon sections represent coding regions. The exons are drawn to scale, and the connecting arcs represent splice paths. (E) Differences in percent spliced in values of AS events between HCC and paired adjacent normal tissues.

FIGURE 3

Protein–protein interaction (PPI) analysis of the identified differentially expressed alternative splicing (DEAS). Interactions of the 372 parent genes affected by DEAS. These genes were used to construct an intricate PPI network comprising 249 nodes and 514 edges. The genes are denoted as nodes in the graph, and the interactions between them are represented as edges. The shape, size, and color of the nodes, respectively, represent alternative splicing type, value of log(FC), and change pattern.

FIGURE 4

Multi-omics analysis of the 71 splicing factors (SFs) in hepatocellular carcinoma (HCC). (A) cBioPortal analysis of the 71 SFs in The Cancer Genome Atlas HCC patients. OncoPrint was used to produce a landscape of genomic alterations (legend) in SFs (rows) at the individual level (columns). In-frame deletions, truncated mutations, and missense mutations with known or unknown significance are shown in orange, blue, green, and gray, respectively, with one-third height. The copy number variations are annotated with the full height; amplification is shown in red and copy number loss is in blue. Heat map matrix shows the variation in SFs at expression level. The expression abundance from high to low is represented by color gradient from red to blue. (B) Expression of the 71 SFs in 33 tumor types. Heat map color gradient depicts the normalized expression of SFs between different tumor types. (C) Differential expression analysis of representative SF TIA1 in HCC. The expression of TIA1 in HCC tissue was significantly higher than that in normal liver tissue.

FIGURE 5

Specific regulatory network of hepatocellular carcinoma-related alternative splicing (AS) events. (A) Correlation network of splicing factors (SFs) and differentially expressed alternative splicing. The shape, size, and color of nodes, respectively, represent type (AS event or SF), value of log(FC), and change pattern (upregulated or downregulated). The breadth of the line represents the interaction strength. (B–G) Representative dot plots of the correlations between the expression of SFs and percent spliced in values of AS events.

FIGURE 6

Prognostic value of differentially expressed alternative splicing (DEAS) in hepatocellular carcinoma. Part DEAS events simultaneously associated with overall survival (OS) and disease-free survival (DFS). Univariate analysis of DEAS for OS and DFS, respectively. Unadjusted hazard ratios (boxes) and 95% confidence intervals (horizontal lines) limited to alternative splicing events, with p < 0.01. The box size is inversely proportional to the width of the confidence interval.

FIGURE 7

Kaplan–Meier curves for overall survival (OS) and disease-free survival (DFS) according to the percent spliced in (PSI) value of alternate terminator (AT) in NEK2 (left panel) and AT in TROPT (right panel). The PSI value distribution for each hepatocellular carcinoma patient and their corresponding survival time (OS and DFS) and survival status is shown in bar and point graphs. Kaplan–Meier curves and log-rank tests were used to compare the survival outcomes between patients with high and low PSI (median cut).

FIGURE 8

Alternative splicing clusters associated with prognosis. (A) Elbow and gap statistical analyses for different numbers of clusters (k = 2–8). (B) Consensus matrix heat map defining four clusters of samples for which consensus values ranged from 0 (white, samples are never clustered together) to 1 (dark blue, samples are always clustered together). (C) Heat map of the 263 differentially expressed alternative splicing in 371 HCC patients ordered by identified cluster showing distinct chromatism. (D,E) Kaplan–Meier survival analysis of patients in different clusters for both overall survival (D) and disease-free survival (E).

FIGURE 9

Functional diversity of CXCL12 splicing variants in hepatocellular carcinoma (HCC). (A) The amino acid sequences of human CXCL12 splice variants. From gene to functional protein – transcription and translation of human CXCL12 variants. (B) The expression level of CXCL12 at exon 5 in tumor and paired adjacent normal samples. (C) The expression levels of CXCL12 isoforms, NP_001029058.1, and NP_954637.1 in HCC and normal tissues. (D) Sequence alignment of NP_001029058.1 and NP_954637.1. (E) The predicted protein structure of NP_001029058.1 and NP_954637.1. (F) The secreted His-NP_001029058.1 protein revealed by the anti-His mAb accumulates massively at the cell surface of HepG2 cells. (G–J) MTT assay (G), colony formation assay (H), wound healing (I), and Transwell assays (J).