Oncofetal SNRPE promotes HCC tumorigenesis by regulating FGFR4 expression through alternative splicing

Understanding the roles of spliceosome and splicing events during tumorigenesis opens new avenues for targeted therapies. Here, we identi�ed that small nuclear ribonucleoprotein polypeptide E (SNPRE) is an oncofetal splicing factor, which had a link in the poor prognosis of hepatocellular carcinoma (HCC), and was reactivated by SOX2. SNRPE knockdown effectively abolished HCC tumorigenesis and progression. Transcriptome analysis and RT-PCR results revealed that SNRPE knockdown induced intron retention (intron 4) in the �broblast growth factor receptor 4 (FGFR4) transcript. Mechanistically, SNRPE knockdown reduced FGFR4 mRNA expression by activating nonsense-mediated RNA decay. FGFR4 knockdown partially blocked the SNRPE-induced malignant progression of HCC cells. Our �ndings discovered SNRPE as a novel oncofetal splicing factor and elucidated the relationship between oncofetal splicing factors, splicing events and carcinogenesis. Therefore, SNRPE is a potential target for HCC treatment.


Introduction
Globally, HCC is the sixth most clinically documented cancer and causes the third most cancer-related deaths 1 .Lack of effective treatment targets is the major challenge in treating HCC.Scienti c research in this area aims to determine the key molecular events that cause the development of HCC 2 .Previous studies have revealed signi cant changes in spliceosome-related genes in HCC 3 .The spliceosome is a macromolecular ribonucleoprotein (RNP) complex that identi es intron/exon boundaries and regulates the reaction that results in the removal of intron and the joining of exons.In short, spliceosomes catalyze the splicing of nuclear precursor mRNAs (pre-mRNAs) into mRNAs 4 .Changes in gene expression of the spliceosome complex [5][6][7][8] produce some novel alternative splicing (AS) events in most cancers.However, the pathophysiological role of the spliceosome in HCC is still unknown and we know little about whether spliceosomes can serve as therapeutic targets for tumor treatment.
There are many similarities between tumorigenesis and embryonic/fetal development in gene expression pro le and regulation mechanisms 9 .The oncofetal genes are abnormally expressed in embryonic/fetal tissues and tumor tissue compared with normal adult tissues 10 .This distinctive expression in tumor cells makes oncofetal proteins a potential target for cancer diagnosis and prevention.Therefore, the determination of identical molecular pathways between fetal liver development and HCC will provide more scienti c information about the pathological mechanisms of HCC occurrence and development.
This research designed a method to identify an oncofetal splicing factor, small nuclear ribonucleoprotein polypeptide E (SNPRE), whose expression is silent in adult livers, high in fetal livers, and overexpressed in HCC tissues.We analyzed the effects of SNRPE on tumorigenesis and progression of HCC in vitro and in vivo.Moreover, we did high-throughput RNA sequencing and found the SNRPE-regulated oncofetal AS events.Consequently, our study provided the detail about the roles of SNRPE and alternative splicing disruption in HCC.

SNRPE is a prognosis-related oncofetal protein in Hepatocellular Carcinoma
We searched the microarray data using liver samples from the liver tissues of fetal and adult, and found a data set (GEO accession: [GSE21224]).We evaluated differentially expressed genes (fold change ≥ 2) in fetal liver using the GEO database (GSE21224) and in the TCGA-LIHC using the TCGA database, and identi ed six spliceosome genes that were upregulated in fetal liver and HCC tissues (Fig. 1A and Supplementary Table 1).Next, we analyzed the correlation between six spliceosome genes, including SNRPE, SNRPB, SF3B4, DDX39B, PRPF3, and LSM4, and the prognosis of HCC patients (Fig. S1).We found that SNRPE was one of the most differentially expressed genes in HCC and its level was strongly related with the prognosis of patients with HCC.
Publicly available databases shown that SNRPE expression was relatively high in the mouse or human fetal liver but was low in the adult liver (Fig. 1B and C).The mRNA and protein levels of SNRPE in mouse fetal liver tissue were signi cantly higher than those in adult liver tissue (Fig. 1D and E).
Then we detected SNRPE expression in human hepatocellular carcinoma tissues.The data analysis of TCGA and GEO database showed that the expression of SNRPE in HCC tissues was higher than that in normal liver tissues.(Fig. 1F and G).We also used a tissue microarray composed of normal and HCC tissues to observe the expression of SNRPE in HCC.Tissue microarrays showing SNRPE protein levels in 90 HCC tissues and 85 adjacent normal tissues.SNRPE exhibited a stronger staining density in HCC tissues comparing with adjacent normal tissues (Fig. 1H).Kaplan-Meier analysis indicated that high SNRPE protein expression in HCC tissues was correlated with decreased overall survival and recurrencefree survival (Fig. 1I and J).As shown in Figure S2J, SNRPE expression in HCC cells (SMMC7721, Bel7402, HCCLM3, SNU387, and SK-Hep-1) was higher than in non-transformed hepatic cells (HL7702, MIHA).These data elaborated that SNRPE was an oncofetal protein, whose level was linked with the poor progression of HCC.

SOX2 regulates SNRPE expression
To evaluate the mechanisms monitoring SNRPE level in HCC, we studied SRY-box 2 (SOX2), which is famous as stem cell transcription factors holding stem cell-like properties and tumor-initiating funciton of HCC cells 11 .HCC cells were transfected with SOX2 small interfering RNA (siSOX2) (Fig. 1K, Fig. S2A-S2C).Knockdown of SOX2 decreased the levels of SNRPE mRNA and protein (Fig. 1L, Fig. S2D-S2J).Bioinformatics study of the binding sites of SOX2 in the SNRPE promoter region proposed eight SOX2binding sites (Additional File 1).These data elaborated that SOX2 might increase SNRPE level through direct binding to the SNRPE promoter.

SNRPE promotes tumorigenicity of HCC cells
To determine the effects of SNRPE in hepatocarcinogenesis, SNRPE-overexpressed HCC, MIHA, and HL7702 cells were used (Fig. 2A, Fig. S3A, Fig. S4A).We observed the activity of the clonal populations in soft agar, a widely used substitute for tumor formation.SNRPE overexpression increased soft agar colony formation (Fig. 2B, Fig. S3B).The colony formation assays and growth curve assays draw the conclusion that ectopic expression of SNRPE also stimulated cell growth (Fig. 2C and D, Fig. S3C).
Similar results were obtained using other HCC cell lines (Fig. S4B and S4C).
Next, we investigated the role of SNRPE in HCC carcinogenesis in vivo.HL7702 and SMMC7721 cells that stably overexpressing SNRPE or vector were injected subcutaneously into nude mice.SNRPE overexpression promoted the growth of tumors.SNRPE overexpression also induced tumorigenesis of HL7702 cells in vivo (Fig. 2E-G).Immunohistochemical staining of xenografts by Ki67 further con rmed that SNRPE promoted HCC cells proliferation (Fig. 2H and I).

SNRPE is required for tumor progression of HCC cells, but not for non-malignant MIHA cells
To further clarify the role of SNRPE in the occurrence of HCC, we used two siRNAs to knockdown SNRPE in HCCLM cells (Fig. 3A).SNRPE knockdown signi cantly inhibited HCCLM3 cell growth and migration (Fig. 3B-F).To predict the SNRPE-associated biological functions in HCC, we performed GSEA using the TCGA dataset.Biological processes, including spliceosome, cell cycle, and DNA replication, exhibited the strongest association with high SNRPE levels.Apoptosis was negatively correlated with high levels of SNRPE (Fig. 3G).Cell cycle progression was arrested during the G2/M phases in SNRPE knockdown cells (Fig. 3H).An apoptosis assay also showed that SNRPE knockdown induced apoptosis of HCCLM3 cells (Fig. 3I).Next, HCCLM3 cells knocked down by SNRPE or negative control were injected subcutaneously into nude mice.Knockdown of SNRPE almost completely inhibited the carcinogenesis of HCCLM3 cells (Fig. 3J-K).By using the tail-vein metastasis model, we clari ed that SNRPE knockdown also led to a remarkable decrease in lung metastasis (Fig. 3L).
To evaluate the effects of SNRPE on human normal hepatocytes, we silenced SNRPE in MIHA cells (Fig. S6A).The growth curve assay showed that SNRPE knockdown did not affect the proliferation of MIHA cells (Fig. S6B).In addition, it did not induce G2/M arrest or apoptosis of MIHA cells (Fig. S6C and S6D).

SNRPE regulates the AS of FGFR4 and CREB3L4 mRNA
To determine the potential mechanisms elaborating the tumor-promoting effects of SNRPE, we performed transcriptome sequencing of SMMC7721 cells with control or SNRPE siRNA and identi ed signi cant changes in AS events based on the percentage-spliced-in (PSI) value.We performed an analysis of SNRPE-regulated AS events and identi ed 3750 AS events with signi cant changes in PSI values (Fig. 4A, Additional File 2).KEGG enrichment analysis showed that these genes were closely related to the cell cycle and spliceosome (Fig. 4B).Cyclic AMP-responsive element-binding protein 3-like protein 4 mRNA (CREB3L4) intron retention (IR) and broblast growth factor receptor 4 (FGFR4) IR were the two main AS events.Consistent with transcriptome sequencing results, SNRPE knockdown stimulated FGFR4 and CREB3L4 intron retention (Fig. 4C, and Fig. S7A and S7B).Next, we identi ed a positive correlation between SNRPE expression and CREB3L4, FGFR4 component gene expression in TCGA data (Fig. 4D).

SNRPE negatively regulates FGFR4 and CREB3L4 expression via activating NMD
To con rm whether SNRPE regulates FGFR4 level through the nonsense-mediated RNA decay (NMD) pathway, we used siRNA to knockdown the basic core protein UPF1 RNA helicase and ATPase (UPF1) of the NMD machinery 12 in SMMC7721 and HCCLM3 cells.As shown in Fig. 5A and 5B, compared to control SNRPE-silenced HCCLM3 and SMMC7721 cells, UPF1 knockdown signi cantly increased FGFR4 mRNA level.To further con rm that NMD is associated with downregulation of FGFR4 mRNA in SNRPEsilenced HCC cells, we used the translation inhibitor cycloheximide (CHX, 100 µg/mL) to impair the function of NMD in SNRPE-silenced SMMC7721 and HCCLM3 cells, and found that CHX increased the expression of FGFR4 mRNA in SNRPE-silenced SMMC7721 and HCCLM3 cells (Fig. 5C and D).These data indicated that SNRPE depletion resulted in the splicing of FGFR4 precursor mRNA into non-coding transcript variants that were degraded by NMD.
To con rm that NMD is associated with down-regulation of CREB3L4 mRNA in SNRPE-silenced HCC cells, CHX (100 µg/mL) was used to inhibit the effect of NMD in SNRPE-knockdown SMMC7721 cells.The results showed that CHX upregulated the expression of CREB3L4 mRNA in SNRPE-silenced SMMC7721 cells (Fig. S8).These data suggest that SNRPE knockdown induced CREB3L4 pre-mRNA splicing and that CREB3L4 pre-mRNA is partially degraded by NMD.

Effect of FGFR4 on HCC tumorigenesis, and rescue of the biological effects of SNRPE-overexpressed HCC cells by FGFR4 knockdown
To investigate the effects of FGFR4 and CREB3L4 on HCC tumorigenesis, FGFR4 and CREB3L4 were knocked down in SMMC7721 and HCCLM3 cells (Fig. 6A and B, and Fig. S9B and S9C).In addition, in TCGA-LIHC, the expressions of FGFR4 and CREB3L4 in HCC tissues were higher than that in normal tissues (Fig. 6C, Fig. S9A).As shown in Fig. 6D-J, FGFR4 knockdown remarkably suppressed cell proliferation and caused G2/M arrest and apoptosis.In addition, it remarkably suppressed cell migration compared to control cells (Fig. 6K and L).However, CREB3L4 knockdown did not affect the cell proliferation of SMMC7721 and HCCLM3 cells (Fig. S9D and S9E), as well as cell cycle and apoptosis of SMMC7721 cells (Fig. S9F and S9G).CREB3L4 knockdown also did not affect the cell migration of SMMC7721 cells (Fig. S9H).
To con rm whether FGFR4 is involved in SNRPE-mediated HCC cell proliferation and migration, we transfected FGFR4 or negative control siRNA into SNRPE-overexpressed SMMC7721 and HCCLM3 cells.FGFR4 knockdown partially inhibited the cell proliferation and colony formation in SNRPE-overexpressed SMMC7721 and HCCLM3 cells (Fig. 7A-D).In addition, FGFR4 knockdown partially blocked cell migration in SNRPE-overexpressed HCC cells (Fig. 7E and F).

Discussion
We identi ed SNRPE as a novel oncofetal protein associated with HCC occurrence and development.We conducted a series of experiments to understand the mechanism linked with the effects of SNRPE on HCC.
Oncofetal genes have drawn garnered great attention in the diagnosis and treatment of cancers, due to their unique expression pattern.Oncofetal protein AFP has been applied for the diagnosis of HCC 13 .Glypican 3 (GPC3) and spalt-like transcription factor 4 (SALL4) are potential diagnostic markers of HCC 14, 15 .However, their effects on tumorigenesis and cancer progression, and more importantly, the possibility of these genes becoming therapeutic targets, are both unclear.Our discovery about SNRPE as an oncofetal protein also demonstrates the important role of SNRPE in HCC occurrence and progression, suggesting that it is a key oncogenic driver of HCC.Sm E/E' (or SNRPE) is one of seven Sm proteins/SNRP proteins 16 .Sm protein and uridine rich small nuclear RNA (snRNA) form uridine rich small nuclear RNP (U1, U2, U4/U6, and U5).Five snRNP subunits form the spliceosome machinery complex 17 .An imbalance in SNRPE proteins disrupts pre-mRNA splicing which produces unexpected mRNA variants from one gene 6, 18-20 .SNRPE is involved in the regulation of lung, breast, and prostatic cancer 19,21,22 .
We also found that knockdown of SNRPE inhibited the proliferation of HCC cells.SNRPE knockdown arrested HCC cells in the G2/M phase and induced apoptosis.However, Quidville et al. revealed that SNPRE knockdown induced autophagy, but had no signi cant effect on cell cycle and apoptosis in lung or breast cancers 19 .This may be due to the different frequencies and types of alternative splicing that occur in cancers of different tissue origins, resulting in functional differences in SNRPE.Moreover, we also found that SNRPE enhanced the migration of HCC cells.
The current research mainly focuses on the development of small molecular compounds for SF3B1 23 .
These compounds have shown promising effects on cancer in preclinical studies 24 .However, the early clinical trials were terminated due to its toxic effects.Current study has reported that silencing SF3B1 has cytotoxicity to normal cells 25 .However, knockdown of Sm protein does not affect the proliferation and apoptosis of lung broblasts cells IMR-90, but can induce the apoptosis of NSCLC cells 26 .Similarly, SNRPE expression was silenced in normal hepatocytes, and inhibition of SNRPE did not affect the proliferation and survival of normal hepatocytes.Collectively, these data demonstrated the potential of SNRPE to serve as a therapeutic target for HCC.
Recent advances in HCC genomic research have brought new insights into hepatocarcinogenesis and identi ed many oncogenes with DNA changes 27 .Using whole-genome sequencing, Jia et al. revealed that SNRPE was ampli ed in HCC 28 .We also demonstrated that aberrant expression of SNRPE in HCC is regulated by the stemness-related genes SOX2, which also play critical roles in fetal liver development.mRNA expression was in accordance with protein levels, implying that SNRPE is regulated primarily at the transcriptional level.Our ndings broaden knowledge of the mechanism of hepatocarcinogenesis.
Among different types of AS events, IR is generally considered to be the result of incorrect splicing, and IR is least studied.Moreover, an increasing number of studies have shown that IR regulation is a physiological mechanism regulating transcriptome function [29][30][31][32] .Intron retention is often associated with NMD down-regulation of gene expression, mainly because the IR sequence usually leads to the introduction of the premature termination codon (PTC) 33 .Our study found that knockdown of SNRPE resulted in increased IR of FGFR4 and CREB3L4 transcripts and subsequent degradation of FGFR4 and CREB3L4 mRNA by NMD.
CREB3L4 is a transcription factor containing an ER membrane-bound bZIP domain 34 .Previous studies have reported that CREB3L4 can promote prostate cancer proliferation 35 , gastric cancer angiogenesis 36 and malignant progression of breast cancer 37 .In addition, Inagaki et al. also reported that CREB3L4 was ampli ed and highly expressed in HCC 38 .Our results also found that CREB3L4 is highly expressed in HCC relative to normal tissues.However, CREB3L4 had no affect proliferation and metastasis in HCC, indicating that CREB3L4 plays different roles in human tumors of very different histologic origins.SNRPE may not play a role in promoting the proliferation and metastasis of HCC by regulating the IR of CREB3L4.
FGFR4-overexpression has been detected in multiple cancers including rhabdomyosarcoma and breast, liver, colon, and prostate cancer.FGFR4-overexpression is also associated with decreased survival time 39 .However, the reason why FGFR4 is highly expressed in HCC remains unclear.A previous study showed that elevated FGFR4 expression was due to increased transcriptional activity that was regulated by hepatocyte nuclear factor-1 alpha 40 or forkhead box C1 41 .In this study, we found that depletion of SNRPE resulted in decreased FGFR4 expression.Among the FGFRs, FGFR4 promotes cell proliferation, anti-apoptosis, and metastasis during HCC progression 42,43 .Consistently, our results demonstrate that FGFR4 knockdown suppress cell proliferation and migration, and induces cell apoptosis in HCC.We also observed that FGFR4 knockdown partly blocked the migration and proliferation of SNRPE-overexpressed HCC cells.These ndings not only propose a new mechanism of abnormal expression of FGFR4, but also con rm the important role of the SNRPE/FGFR4 axis in HCC.Thus, targeting the SNRPE/FGFR4 axis could be a novel therapeutic strategy for HCC.

Oncofetal splicing factor screening
The data set was download from the Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/)and the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/)databases.Differential gene expression was analyzed with limma and edgeR.The change in gene expression was considered signi cant with q-value less than 0.05 and fold change ≥ 2. Next, the splicerosome genes were obtained from the KEGG database and were presented in Supplementary Table 2.We selected differential spliceosome genes according to the spliceosome genes list and presented in Supplementary Table 3.Then we sorted out the common differential spliceosome genes.

Real-time qPCR analysis
Total RNA was extracted from cells by Trizol reagent (vazyme biotech, Nanjing, China).Reverse transcription of RNA to cDNA was completed with hiscript QRT supermix (qPCR; vazyme biotech).
StepOne Real-Time PCR System (Applied Biosystems) was used for PCR analysis.Primer sequences for PCR ampli cation are listed in Supplementary Table 4. 18S served as an internal control.Relative expression of mRNAs was calculated from comparative Ct formula.

Western blot analysis
Extraction of total protein was done using RIPA lysis buffer (NCM Biotech, China), followed by denaturation by adding 5× loading buffer.Proteins were separated by SDS-PAGE.Membranes was blocked by 5% BSA for 1 hour at room temperature.The membranes were placed in the primary antibodies: SNRPE (Thermo, PA5-96342, 1:1000), FGFR4 (Abcam, ab178396, 1:1000).Next, the membranes were placed in secondary antibodies after washing with TBST.Finally, membranes were observed in a Gel imaging System (TANON, Beijing, China).The images were quanti ed by Image J software.GAPDH or β-Actin was used as an internal reference.

Gene expression analysis
The expression plot and survival plot of SNRPE were collected from GEPIA (Gene Expression Pro ling Interactive Analysis) 44 .The correlation among the expression of SNRPE, FGFR4, and CREB3L4 was generated from cBioProtal (https://www.cbioportal.org/).Tissue microarray and immunohistochemistry TMA (HLivH180Su16) containing HCC samples were obtained from Outdo Bioth (Shanghai, China).The TMA specimens were used for IHC analysis.Multiplicative staining index (0-3) and the average staining score (0-4) were employed for the quanti cation of SNRPE protein expression levels.We marked the tumor with IHC score ≥ 8 as SNRPE-high expression and tumor with IHC score < 8 as SNRPE-low expression.
Immunohistochemistry (IHC) was performed as previously described.In brief, tumor tissues were xed in paraformaldehyde and blocked with FBS.Sections were incubated with Ki67 antibody (1:50) after permeabilization with Triton X-100.Next, sections were washed with PBS and placed in the secondary antibody.After counterstaining with hematoxyline, sections were visualized under a microscope (BX53, Olympus, Japan).

Vector construction and siRNA transfections
Cell growth to 50-60% con uency was followed by the transfection with siRNAs targeting SNRPE, FGFR4, UPF1, SOX2, or negative control (NC) for 8 h.Lipofectamine 3000 (Invitrogen, USA) was used in all transfection experiments.To generate HCCLM3 cell lines stably knocking down SNRPE, shRNA or negative control-shRNA were constructed using a lentiviral shRNA technique.The shRNA and siRNA sequences are listed in Supplementary Table 4.
We transfected pEGFPN1-SNRPE (NM_003094.4) or pEGFPN1-empty vector using lipofectamine 3000 to detect the effect of overexpression of SNRPE on HCC cells.To stably express SNRPE in HCC cells, we also used lentiviral vectors.After lentivirus infection, hepatoma cells were exposed to 5 µg/mL puromycin for one week.Before further analysis, the stable expression of SNRPE in cell lines was detected by Western blot.

Cell proliferation and colony formation assay
Cell growth curve was detected through CCK8 (Vazyme biotech, Nanjing, China) assays.Brie y, an equal number of the cells were dispersed into 96-well plates at day 0, and after treatment absorbance was measured at 450 nm with the help of a multi-detection micro plate reader (thermo, USA).EdU Kit (RiboBio, Guangzhou, China) was used for the EdU immuno uorescence staining.The results were analyzed with a light microscope (Olympus Corporation, Japan) and ImageJ software.
For the analysis of colony formation, Cells (500 cells/well) were dispersed into 6-well plates and contained for 14 days.Then methanol was used to x, crystal violet to stain, and Image J software for counting the number of colonies.

Cell migration assay
Cells dispersed in 250 µL of serum-free medium were seeded into the upper compartment of the transwell.Migrated cells were immobilized and stained using crystal violet.Cell number was counted by getting average from 5 random elds under a light microscope (Olympus Corporation, Japan).

Cell Apoptosis and Cell Cycle Assays
For the determination of cellular apoptosis, cells were collected after treatment and suspended in binding buffer.Cells were then allowed to interact with PI and Annexin V-FITC (Vazyme Biotech, Nanjing, China) for 15 min and analyzed by ow cytometry (BD, Biosciences, USA).
Cells were treated for cell cycle assay, xed in ethanol and stained by Cell Cycle staining kit (Keygen Biotech, Nanjing, China).Flow cytometer and FlowJo 10.6.2 were used for analysis.

Xenograft assay
All animal experiments were performed following the instructions of the Animal Care Committee of China Pharmaceutical University (Approval No.2110748).Five-week-old female mice (BABL/c nude) were obtained from the Hangzhou Ziyuan Experimental Animal Technology Co., Ltd.HL7702 (4 ×10 6 cells in 100 µL PBS) or SMMC772 (2.5 ×10 6 cells in 100 µL PBS) or HCCLM3 (2×10 6 cells in 100 µL PBS) were administered subcutaneously into the right ank of mice.Time period for tumor growth was 4 weeks.Tumor growth and mice weight were recorded every other day and were calculated.The formula is as follows: tumor volume= (length × width^2)/ 2 .At the end, tumors were removed and photographed.
For in vivo metastasis assays, HCCLM3 cells (2×10 6 cells in 200 µL PBS) with SNRPE shRNA (HCCLM3-shSNRPE) or negative control HCCLM3 cells (HCCLM3-shNC) was used.The cells were injected into mice via the tail vein.Then mice were killed after 5 weeks of injections.Para n-embedded nude mice lungs were stained with H&E to observe metastatic nodules.

RNA sequencing and data analysis
Extraction of total RNA was done using TRIZOL Reagent (Vazyme).After quanti cation with Agilent 2100 Bioanalyzer and Nano Drop, samples were used for library preparation.Libraries having divergent indices were multiplied and loaded on an Illumina HiSeq instrument.2×150bp paired-end con guration was carried out for sequencing; base calling and image analysis were carried out on the HiSeq instrument by the HiSeq Control Software + OLB + GAPipeline-1.6(Illumina).Further analysis of sequences was completed by GENEWIZ.The RNA-Seq data were saved in GEO Database under accession number GSE197092, available at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE197092.
Alternative splicing events are determined by the human transcriptome.CASH was used to analyze differential splicing of SNRPE knockdown samples (compared to the negative control) 45 .Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis was accomplished with DAVID (https://david.ncifcrf.gov/).Differentially spliced genes (PSI ≥ 0.2, Bayes-factor ≥ 10) were used for KEGG enrichment analysis.

Gene set enrichment analysis (GSEA)
Gene Set Enrichment Analysis (GSEA) was performed according to the instructions of the Broad Institute.The RNA-HiSeq data of HCC were divided into SNRPE high expression group and SNRPE low expression group according to the FPKM value of SNRPE.P value was obtained by comparing the enrichment score with the enrichment result produced by the random arrangement of 1000 genes, and its statistical signi cance was evaluated.

Splicing assay using RT-PCR
Total RNA isolation and single-stranded cDNA were obtained according to the above-mentioned method.PCR was executed with Rapid Taq Master Mix (Vazyme).The PCR program was set as denaturation at 95°C for 3 min, ampli cation by 35 cycles at 95°C for 15 s, 60°C for 25 s, 72°C for 15 s, nally 72°C for 5 min.Final PCR products were separated by agarose gel electrophoresis followed by the scanning with gel imaging system (TANON, Beijing, China).The primer sequences are presented in Supplementary Table 4.

Data analysis
Graphpad Prism 6.02 was used for statistical analyses.For comparisons, Student's t-test (two-sided) or the one-way analysis of variance was applied between the data pairs.P 0.05 was statistically signi cant.Kaplan-Meier method was applied for survival curve.We used log-rank test to compare survival time between two groups.