Alternative splicing (AS) allows generation of cell type-specific
mRNA transcripts and contributes to hallmarks of
cancer. Genome-wide analysis for AS in human
hepatocellular carcinoma (HCC), however, is limited. We sought to obtain a comprehensive AS landscape in HCC and define
tumor-associated variants. Single-molecule real-time long-read
RNA sequencing was performed on patient-derived HCC cells, and presence of splice junctions was defined by SpliceMap-LSC-
IDP algorithm. We obtained an all-inclusive map of annotated AS variants and further discovered 362 alternative spliced variants that are not previously reported in any database (neither RefSeq nor GENCODE). They were mostly derived from intron retention and early
termination codon with an in-frame open reading frame in 81.5%. We corroborated many of these predicted unannotated and annotated variants to be
tumor specific in an independent cohort of primary HCC
tumors and matching nontumoral liver. Using the combined Sanger sequencing and TaqMan junction assays, unique and common expressions of spliced variants including
enzyme regulators (ARHGEF2, SERPINH1),
chromatin modifiers (DEK, CDK9, RBBP7),
RNA-binding proteins (SRSF3, RBM27, MATR3, YBX1), and receptors (ADRM1, CD44v8-10,
vitamin D receptor, ROR1) were determined in HCC
tumors. We further focused functional investigations on ARHGEF2 variants (v1 and v3) that arise from the common amplified site chr.1q22 of HCC. Their biological significance underscores two major
cancer hallmarks, namely
cancer stemness and epithelial-to-mesenchymal transition-mediated cell invasion and migration, although v3 is consistently more potent than v1. Conclusion: Alternative
isoforms and
tumor-specific
isoforms that arise from aberrant splicing are common during the liver
tumorigenesis. Our results highlight insights gained from the analysis of AS in HCC.