SF3A3 in Liver Hepatocellular Carcinoma: Oncogenic Role and Prognostic Significance

Abstract

Introduction

Splicing factor 3a subunit 3 (SF3A3) is a core component of the SF3A complex essential for pre-mRNA splicing. However, its role and prognostic significance in liver hepatocellular carcinoma (LIHC) remain poorly characterized.

Methods

We systematically investigated the role of SF3A3 in LIHC using RNA sequencing data, promoter methylation profiles, and proteomic data from The Cancer Genome Atlas and Genotype-Tissue Expression datasets. Associations between SF3A3 expression and clinical or pathological features were analyzed, and the prognostic value of SF3A3 was assessed using survival analysis.

Results

SF3A3 mRNA and protein levels were elevated in LIHC tumors, whereas promoter methylation was reduced compared with matched normal liver tissues. High SF3A3 expression was positively associated with advanced tumor stage and grade, as well as TP53 mutation status. Elevated SF3A3 expression was linked to significantly poorer overall survival in patients with LIHC. SF3A3 expression showed a positive correlation with tumor-associated fibroblast infiltration. Survival analyses indicated that SF3A3 expression level and tumor stage are key factors associated with LIHC overall survival. Gene set enrichment analysis revealed that tumors with high SF3A3 expression were significantly enriched for pathways related to the GABA-A receptor complex and FGFR3 mutant receptor activity.

Conclusion

The expression of SF3A3 in LIHC tumors is correlated with adverse pathological features and poor prognosis, and potential biological mechanisms underlying the role of SF3A3 in LIHC were elucidated. This study highlights SF3A3 as a promising biomarker for prognosis and disease severity in LIHC.

Introduction

According to 2020 cancer statistics, liver cancer is among the top five malignancies and Asia accounts for 72.5% of worldwide cases [1, 2]. Among all types of liver cancer, liver hepatocellular carcinoma (LIHC) accounts for the majority of morbidity and mortality. [3]. Incidence of LIHC is associated with age and gender globally, with peak incidence occurring at approximately 75 years of age, and this may be caused by a number of factors, including immune system disorder, somatic mutations, and chronic long-term risk factors [4]. It is widely reported that LIHC is several times more prevalent in males compared with females globally, which may be related to behavioral, endocrinological, and metabolic factors [5]. Apart from hepatitis B virus, hepatitis C virus, alcohol, and various metabolic factors, which are common risk factors for liver cancer [6], many genes have recently been reported to contribute to liver tumorigenesis. Among these, splicing factors have emerged as critical players in hepatocarcinogenesis. For instance, SF3B1 mutations are frequently observed in various cancers [7] and have been linked to altered splicing of genes involved in notch signaling, cell cycle, and DNA damage response (DDR) [8]. Similarly, the serine/arginine-rich splicing factor SRSF1 is known to be overexpressed in hepatocellular carcinoma (HCC) [9], where it promotes oncogenic transformation by regulating alternative splicing of key genes like BIN1 and S6K1 [10]. However, the roles of many core spliceosome components, such as the SF3A complex, remain poorly understood in LIHC. The diagnosis and treatment should be adjusted dynamically in line with different epidemiology.

Splicing factor 3a subunit 3 (SF3A3) is a protein-coding gene. Among its related pathways are processing of capped intron-containing pre-mRNA and exercise-induced circadian regulation. Gene Ontology (GO) annotations related to SF3A3 include nucleic acid binding and RNA binding. The most widely known function of SF3A3 is as a pre-mRNA splicing factor in vivo [11]. Cellular stress response 1 (CSR1) is a tumor suppressor gene that regulates cell death. CSR1 can interact with SF3A3, resulting in the translocation of SF3A3 from the nucleus to the cytoplasm. This redistribution of SF3A3 leads to decreased levels of epidermal growth factor receptor and platelet-derived growth factor receptor within cells [12].

The constitutive androstane receptor (CAR) plays an important role in the hepatic biodegradation of xenobiotics. SF3A3 can specifically interact with CAR protein to inhibit transcriptional activity in the liver [13]. YTH domain family 2 (YTHDF2) is a potential biomarker in LIHC that plays a role in carcinogenesis. Studies show that SF3A3 expression is positively correlated with YTHDF2 expression implying that SF3A3 and YTHDF2 may cooperate in LIHC carcinogenesis [14].

SF3A3 is a novel predictor gene associated with DNA repair processes and survival in hepatocellular carcinoma. [15]. SF3A3 is a newly characterized spliceosome-related RNA-binding protein involved in breast cancer [16], where it plays an important role by selectively regulating and controlling MYC-driven splicing and metabolic processes, affecting MYC-driven tumorigenesis and breast cancer oncogenesis [17, 18]. The DDR pathway relies on the proper splicing of many core component genes. Dysregulation of SF3A3, through overexpression or mutation, can cause widespread splicing defects, making DDR gene transcripts particularly vulnerable. This compromises DNA repair capacity, leading to the accumulation of DNA damage, increased mutation burden, and ultimately, cancer hallmarks such as genomic instability and oncogenic progression. Elevated expression levels of SF3A3 are related to the occurrence and development of bladder cancer. E2F6 and KDM5C can recruit KDM5C to the SF3A3 promoter and demethylate the GpC island of H3K4me2, leading to increased SF3A3 expression and bladder cancer progression [19]. SF3A3 also functions in carcinogenesis of lung cancer, with a recent study showing that circular RNA regulates SF3A3 expression to inhibit non-small cell lung cancer through activating the p53 signaling pathway [20].

The Cancer Genome Atlas (TCGA) is a large-scale state-of-the-art cancer data repository with multiomics datasets, which has played a crucial role on tumor analysis in recent years [21]. Up to now, there have been no published studies which have utilized TCGA to investigate the role and prognostic value of SF3A3 in LIHC. In this LIHC tumor analysis of TCGA datasets, we comprehensively investigated the expression level, clinical prognosis, genetic mutation, association with tumor mutation burden (TMB) and microsatellite instability (MSI), correlation with immune cell infiltration and immune checkpoint molecules, relation with tumor-associated fibroblast infiltration, and the most strongly correlated genes, proteins, and signaling pathways of SF3A3 in LIHC tumor types.

Materials and Methods

SF3A3 Tissue Cell and Single-Cell Level Expression Analysis

We used the Human Protein Atlas (HPA) database to construct SF3A3 mRNA expression plots of liver single cells and tissue cells. Heatmap of liver single cells and tissue cells were also used to explore the relation between SF3A3 and different types of cell markers.

SF3A3 Expression Level in Normal Liver Tissue and Liver Cancer

We obtained SF3A3 immunohistochemistry from the HPA dataset to observe SF3A3 expression level and localization in both normal liver tissue and liver tumor. We also used TCGA dataset and Genotype-Tissue Expression (GTEx) dataset to generate violin plots comparing SF3A3 expression between normal liver tissue and liver tumor.

SF3A3 Gene Expression Level in LIHC Patient Groups

The clinical and pathological information used in this study, including tumor stage, grade, age, gender, race, BMI, and TP53 mutation status, is publicly available and was directly obtained from the TCGA and GTEx databases. We screened 421 liver samples from the TCGA dataset. According to sample types, these samples were divided into 50 normal and 371 primary tumors. According to different tumor progression stages, we divided the 371 primary tumors into 4 tumor stages: stage 1, stage 2, stage 3, and stage 4 (tumor stage information is not available for 31 samples). We divided the 371 tumor patients into different group according to patients’ race: Caucasian, African American, and Asian. Patient race information is not available for 19 samples. One sample from American Indians is not considered in the plot. We divided 371 tumor patients into two genders (245 male and 117 female). Patient gender information is not available for 9 samples. We divided the 371 tumor patients into various age groups: 21–40 years, 41–60 years, 61–80 years, and 81–100 years. Patient age information is not available for 9 samples, and 4 samples from patients aged below 21 years are not considered in the plot. We divided the 371 tumor patients into different weight groups: normal weight – BMI greater than equal to 18.5 and BMI less than 25, extreme weight – BMI greater than equal to 25 and BMI less than 30, obese – BMI greater than equal to 30 and BMI less than 40, as well as extreme obese – BMI greater than 40. Patient BMI information is not available for 41 samples. Apart from different tumor stages, we also divided tumor patients into different grades according to subtypes: grade 1 for well differentiated (low grade), grade 2 for moderately differentiated (intermediate grade), grade 3 for poorly differentiated (high grade), and grade 4 for undifferentiated (high grade). We divided the 371 tumor patients into two groups according to TP53 mutation status: 105 TP53-mutant patients and 255 TP53-nonmutant patients. Whole-exome sequencing data are not available for 11 samples. We also divided the 371 tumor patients into different histological types: 361 hepatocellular carcinoma, 3 fibrolamellar carcinoma, and 7 hepatocholangial carcinoma (mixed).

SF3A3 Promoter Methylation Expression Level in LIHC

We screened 421 liver samples from the TCGA dataset. According to sample types, these samples are divided into 50 normal and 371 primary tumors. According to different tumor progression stages, we divided primary tumors into 4 tumor stages: stage 1, stage 2, stage 3, and stage 4 (tumor stage information is not available for 24 samples). We divided tumor patients into different group according to patients’ race: Caucasian, African American, and Asian. Patient race information is not available for 10 samples. Two samples from American Indians are not considered in the plot. We divided tumor patients into two genders (255 male and 122 female). We divided 371 tumor patients into various age groups: 21–40 years, 41–60 years, 61–80 years, and 81–100 years. Patient age information is not available for 1 sample in the plot. We divided tumor patients into different weight groups: normal – BMI greater than equal to 18.5 and BMI less than 25, extreme weight – BMI greater than equal to 25 and BMI less than 30, obese – BMI greater than equal to 30 and BMI less than 40, as well as extreme obese – BMI greater than 40. Patient BMI information is not available for 36 samples and 21 samples from patients with BMI <18.5 are not considered in the plot. Apart from different tumor stages, we also divide tumor patients into different grades according to subtypes: grade 1 for well differentiated (low grade), grade 2 for moderately differentiated (intermediate grade), grade 3 for poorly differentiated (high grade), and grade 4 for undifferentiated (high grade). Grade information is not available for 5 samples. We divided tumor patients into two groups according to TP53 mutation status: 109 TP53-mutant patients and 266 TP53-nonmutant patients. Whole-exome sequencing data is not available for 2 samples.

SF3A3 Proteomic Expression Level in LIHC

We screened 421 liver samples from the TCGA dataset. According to sample types, these samples are divided into 50 normal and 371 primary tumors. Z-value represents standard deviations from the median across samples for the given cancer type. Log₂ spectral count ratio values from the CPTAC were first normalized within each sample profile, then normalized across samples. We divided tumor patients into two genders (male and female). We divided tumor patients into different age groups: 21–40 years, 41–60 years, 61–80 years, and 81–100 years. SF3A3 protein expression in LIHC subtype is shown in the plot. Eleven pan-cancer subtypes (from s1 to s11) datasets were generated using mass spectrometry-based proteomic data from a compendium dataset of 2,002 primary tumors compiled from 17 studies to 14 cancer types [22]. Unsupervised clustering analysis was performed using the top 2,000 most variable proteins from this compendium TCGA dataset, using log-transformed expression values centered to standard deviations from the median within each cancer type. The s11 subtype is specific to glioblastomas and pediatric brain tumors.

Relationship between SF3A3 Expression and Liver Cancer Clinical Survival

We used Kaplan-Meier plotter (database: https://kmplot.com) [23] to explore the relation of SF3A3 expression with liver cancer clinical survival, such as overall survival (OS), relapse-free survival, progression-free survival, and DSS. A comprehensive list of all databases and tools utilized in this study, along with their sources and purposes, is provided in Table 1. Furthermore, we used GEPIA2 to explore the relationship between SF3A3 expression level and LIHC patient survival. A Sankey diagram with age, pTNM stage, grade, SF3A3 expression, and survival status was analyzed. We also used Cox regression to explore the association among SF3A3 expression, age, gender, and tumor stages with LIHC survival.

Table 1.

Summary of databases and tools used in this study

Database/tool	Source/URL	Purpose in this study
TCGA	https://portal.gdc.cancer.gov	Gene expression, methylation, mutation, clinical data
GTEx	https://gtexportal.org	Normal liver tissue gene expression
HPA	https://www.proteinatlas.org	Single-cell/tissue expression heatmaps, IHC images
GEPIA2	https://gepia2.cancer-pku.cn	Expression correlation, survival analysis, similar genes
Kaplan-Meier plotter	https://kmplot.com	OS, RFS, PFS, DSS survival analysis
cBioPortal	https://www.cbioportal.org	Gene alteration analysis
CAMOIP	https://www.camoip.net	Mutational landscape visualization
TIMER2	https://timer.comp-genomics.org	Immune cell and CAF infiltration analysis
STRING	https://string-db.org	Protein-protein interaction network
DepMap/CCLE	https://depmap.org	Gene effect score in cancer cell lines
Immunedeconv (R package)	https://github.com/icbi-lab/immunedeconv	Immune cell deconvolution

SF3A3 Gene Alteration Analysis

We used cBioPortal to explore SF3A3 gene alteration. Gene alteration type and mutation-related pathway were also analyzed using cBioPortal. We also used CAMOIP to explore SF3A3 gene mutational landscape in LIHC patients.

TMB and MSI Analysis of SF3A3 in LIHC Tumor

We obtained RNA sequencing expression profiles and corresponding clinical information for TCGA dataset LIHC tumor. TMB and MSI were analyzed using R version 4.0.3 and R packages maftools and MSIseq. We used Spearman correlation analysis for TMB and SF3A3 gene expression. If not stated otherwise, two-group data were performed by Wilcox test. Spearman correlation analysis was also used to correlate MSI with SF3A3 gene expression. p values <0.05 were considered statistically significant (*p < 0.05). TMB, neoantigen loads, and MANTIS score were compared between SF3A3 high-expression group and SF3A3 low-expression group in LIHC patients separately using the Mann-Whitney U test.

Gene Effect Score for SF3A3 in Different Liver Cancer Cell Lines

We used DepMap (CCLE) dataset to get SF3A3 expression level in different liver cancer cell lines. SF3A3 gene expression in selected liver cancer cell lines was explored.

Immune-Related Analysis of SF3A3 in LIHC Tumor

We obtained RNA sequencing expression profiles and corresponding clinical information for TCGA dataset LIHC tumor. To assess the reliable results of immune score evaluation, we used immunedeconv [24], which is an R software package that integrates six latest algorithms, including Tumor Immune Estimation Resource (TIMER), xCell, Microenvironment Cell Population counter (MCP-COUNTER), CIBERSORT, efficient prediction of IC manufacturing hotspots with a unified meta classification formulation (EPIC), and quanTIseq. SIGLEC15, IDO1, CD274, HAVCR2, PDCD1, CTLA4, LAG3, and PDCD1LG2 are the genes which are associated with the immune checkpoint. We extracted the expression values of these 8 genes to perform correlation analysis with SF3A3 in LIHC. All the analysis methods and R package were implemented by R version 4.0.3. If not stated otherwise, two-group data were performed by Wilcox test. p values <0.05 were considered statistically significant (*p < 0.05).

For the heatmap of SF3A3 immune score and SF3A3 gene expression in multiple TCGA dataset tumors, different colors represent correlation coefficients. Negative values indicate negative correlations, positive values indicate positive correlations, and deeper colors represent stronger correlations; asterisks (*) represent significant levels: *p < 0.05, **p < 0.01, and ***p < 0.001. The statistical difference of two groups was compared through the Wilcox test. For the heatmap of immune checkpoint-related gene expression, each box in the figure represents the correlation analysis between the expression of SF3A3 and the immune checkpoint in LIHC; asterisks (*) represent significant levels: *p < 0.05, **p < 0.01, ***p < 0.001. Different colors represent the changes in correlation coefficients.

Fibroblast Infiltration Analysis of SF3A3 in LIHC Tumor

We used the “Immune” selection in TIMER, version 2 (TIMER2) to analyze immune cells’ infiltration status. The aim of EPIC and MCP-COUNTER calculations was to explore the relationship between SF3A3 expression level and tumor-associated fibroblast infiltration.

Correlation Analysis of SF3A3-Related Genes

We used the STRING tool to plot a SF3A3 co-expression genes network of Homo sapiens with relevant parameters set as below: set “Network type” as “full STRING network”; set “meaning of network edges” as “evidence”; set “active interaction sources” as “Textmining, Experiments, Databases, Co-expression, Neighborhood, Gene Fusion, and Co-occurrence”; set “minimum required interaction score” as “low confidence (0.150)”; and set “maximum number of interactors to show” as “50.” We selected the “Most Similar Genes” part in GEPIA2 to get the top 100 most related genes that have a similar expression mode with SF3A3. GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of most similar genes were used to explore possible functions of SF3A3. Besides, pairwise correlation analysis of the most related genes with SF3A3 was done through the “Correlation Analysis” selection in GEPIA2. Prognostic results of the most correlated genes were also explored in LIHC tumor. In addition, we also explored related GO and KEGG results for SF3A3 high expression vs. low expression in LIHC patients.

Results

SF3A3 Expression Level in Liver Tissue

HPA is a comprehensive protein database mapping the spatial distribution of all human proteins in cells, tissues, and organs. We use it to generate the tissue cell expression level of SF3A3 in liver. Heatmap results also showed correlation between SF3A3 and different types of cell markers, including some correlation between SF3A3 and immune cell markers in tissue cell (Fig. 1a, b) expression level. Together with these findings we assumed that SF3A3 gene expression may be related to immune cell expression in liver tissue. HPA dataset immunohistochemistry result shows that SF3A3 expression level is higher in liver tumor than normal liver tissue using the same antibody HPA032055 (Fig. 1c). The antibody staining in normal liver tissue is low in both cholangiocytes and hepatocytes. However, the antibody staining intensity is stronger with staining quantity over 75% in liver tumor and SF3A3 protein expression mainly in nuclear and cytoplasmic/membranous (Fig. 1c). This is further supported by RNA sequencing data (Fig. 1d), which shows average SF3A3 expression levels across TCGA tumors is higher compared to normal liver tissue.

Fig. 1.

SF3A3 expression in liver tissue. a Tissue cell expression heatmap of SF3A3 in liver tissue, generated from the Human Protein Atlas (HPA) database, showing correlation with various cell markers including immune cell markers. b Single-cell expression heatmap of SF3A3 in liver, indicating cell type-specific expression patterns. c Immunohistochemistry (IHC) staining of SF3A3 in normal liver tissue and liver tumor (antibody HPA032055). Staining intensity is low in normal hepatocytes and cholangiocytes but strong (>75% staining) in tumor tissue, with localization in the nucleus and cytoplasm/membrane. d RNA sequencing data (TCGA + GTEx) comparing transcripts per million (TPM) normalized SF3A3 expression between normal liver vs. LIHC tumors.

SF3A3 Gene Expression across Tumor Stages, Grades, and Histological Subtypes in LIHC

The TCGA dataset analysis results showed that stage 1, stage 2, and stage 3 LIHC have significant differences in SF3A3 expression when compared with patients from the normal group (Fig. 2a). We also observed significant differences in SF3A3 expression between different LIHC tumor stages, with the stage 1 vs. stage 2 and stage 1 vs. stage 3 comparisons both showing higher SF3A3 expression in stage 2 and stage 3, respectively. Normal vs. stage 1, normal vs. stage 2, normal vs. stage 3, stage 1 vs. stage 2, and stage 1 vs. stage 3 have significant p values (p < 0.01**). No significant differences were observed between race groups (Fig. 2b), gender (Fig. 2c), age groups (Fig. 2d), or weight groups (Fig. 2e) among the 371 tumor patients. Together, these results indicate that SF3A3 expression levels are associated with tumor stage and may serve as a predictor of tumor progression.

Fig. 2.

SF3A3 gene expression across tumor stages, grades, and subtypes in liver hepatocellular carcinoma (LIHC). a SF3A3 expression in normal vs. tumor stages (stages 1–4). b SF3A3 expression across racial groups. c SF3A3 expression by gender. d SF3A3 expression across age groups. e SF3A3 expression across BMI groups. f SF3A3 expression across tumor grades (grades 1–4). g SF3A3 expression in TP53 mutant vs. nonmutant tumors. h SF3A3 expression across histological subtypes.

Next, we investigated whether SF3A3 expression is also associated with tumor grade. Significant p values were found for the difference comparison of grade 1 vs. grade 3 (p < 0.01**), grade 1 vs. grade 4 (p < 0.05*), and grade 2 vs. grade 3 (p < 0.01**). This indicates that SF3A3 expression is associated with tumor grade and may serve as a tumor grade predictor (Fig. 2f). Apart from obvious differences from normal vs. TP53-mutant group (p < 0.001***) and normal vs. TP53-nonmutant group (p < 0.001***), there is also significant difference between TP53-mutant group and TP53-nonmutant group (p < 0.001***) (Fig. 2g). The expression of SF3A3 has no association with different histological types (Fig. 2h).

SF3A3 Promoter Methylation Levels in LIHC Tumors

The comparison of SF3A3 promoter methylation expression between normal and primary tumors is significant with p < 0.001*** (Fig. 3a). The result shows that stage 1 has higher methylation than stage 4 (p < 0.05*), stage 2 has higher methylation than stage 4 (p < 0.05*), and stage 3 has higher methylation than stage 4 (p < 0.05*). These indicate that SF3A3 promoter methylation level correlated with normal and tumor stages and may serve as predictors of different tumor stages (Fig. 3b). No significant differences were observed among race groups (Fig. 3c), gender (Fig. 3d), age groups (Fig. 3e), or weight groups (Fig. 3f). Grade 1 has higher methylation than grade 2 (p < 0.05*) and grade 1 has higher methylation than grade 3 (p < 0.001***) (Fig. 3g). This indicates that SF3A3 methylation level is associated with tumor grade and may function as a predictor for different grades. There is also significant difference between TP53-mutant group and TP53-nonmutant group (p < 0.01**) (Fig. 3h).

Fig. 3.

SF3A3 promoter methylation levels in liver hepatocellular carcinoma (LIHC). a Promoter methylation in normal vs. tumor tissues. b Promoter methylation across tumor stages. c Methylation across racial groups. d Methylation by gender. e Methylation across age groups. f Methylation across BMI groups. g Methylation across tumor grades. h Methylation in TP53 mutant vs. nonmutant tumors.

SF3A3 Protein Expression Level in LIHC Tumor

SF3A3 protein expression is higher in primary tumors compared with normal tissues with p < 0.0001*** (online suppl. 1A; for all online suppl. material, see https://doi.org/10.1159/000552986). The comparison result shows no obvious difference between male and female groups (online suppl. 1B). When we divided tumor patients into different age groups, apart from the difference between normal and different age groups, significant difference is seen in comparisons between 21–40 years and 61–80 years (p < 0.01**), also between 41–60 years and 61–80 years (p < 0.05*) (online suppl. 1C). For tumor stages, significant differences were found for S2 vs. S10 (p < 0.05*), S3 vs. S4 (p < 0.01**), S3 vs. S5 (p < 0.01**), S3 vs. S6 (p < 0.05*), S4 vs. S7 (p < 0.001***), S4 vs. S9 (p < 0.01**), S4 vs. S10 (p < 0.001***), S5 vs. S7 (p < 0.001***), S5 vs. S9 (p < 0.001***), S5 vs. S10 (p < 0.001***), S6 vs. S7 (p < 0.05*), S6 vs. S10 (p < 0.001***), S7 vs. S10 (p < 0.001***), and S9 vs. S10 (p < 0.001***) (online suppl. 1D).

Association of SF3A3 Expression with Patient Survival

Kaplan-Meier analyses revealed that higher SF3A3 expression in LIHC patients is significantly associated with worse prognosis across multiple survival outcomes, including OS (p < 0.01, HR = 1.95, 95% CI: 1.38–2.76), relapse-free survival (p = 0.0073, HR = 1.57, 95% CI: 1.13–2.18), progression-free survival (p = 0.016, HR = 1.44, 95% CI: 1.07–1.94), and disease-specific survival (p = 0.0054, HR = 1.86, 95% CI: 1.19–2.90) (Fig. 4a). Cox regression result showed that LIHC OS is associated with SF3A3 expression level and tumor stages (Fig. 4b). A Sankey diagram showed specific relation among age, pTNM stage, grade, SF3A3 expression, and survival status (Fig. 4c). The Sankey diagram visually integrates the flow of patients across multiple clinical categories and SF3A3 expression levels. It demonstrates a clear enrichment of high SF3A3 expression among patients with aggressive clinicopathological features. Specifically, the majority of patients with advanced-stage (stage III) and high-grade (G3/G4) tumors, as well as those who succumbed to the disease (status: dead), were clustered within the high SF3A3 expression group. Furthermore, high SF3A3 expressions appeared more prevalent in patients aged ≤60 years. This multidimensional visualization strongly reinforces that SF3A3 overexpression defines a clinically aggressive HCC subtype characterized by advanced disease stage, poor differentiation, and unfavorable outcome.

Fig. 4.

Association of SF3A3 expression with patient survival. a Kaplan-Meier survival curves for overall survival (OS), relapse-free survival (RFS), progression-free survival (PFS), and disease-specific survival (DSS) in liver hepatocellular carcinoma (LIHC) patients stratified by SF3A3 expression (high vs. low). b Cox regression forest plot showing association of SF3A3 expression, age, gender, and tumor stage with OS. c Sankey diagram illustrating relationships among age, pathological stage (pTNM stage), tumor grade, SF3A3 expression, and survival status.

TMB and MSI of SF3A3

SF3A3 expression in LIHC tumor has positive correlation with TMB and MSI with a p value of 0.02 and 0.02, respectively (Fig. 5a, b). We observed higher TMB (Fig. 5c), neoantigen loads (Fig. 5d), and MANTIS score (Fig. 5e) in SF3A3 high-expression group than in the SF3A3 low-expression group with all p < 0.05*. Analysis of the mutational landscape in LIHC shows TP53 has a significantly higher number of mutations in the SF3A3 high-expression group (Fig. 5f).

Fig. 5.

Correlation of tumor mutation burden (TMB) and microsatellite instability (MSI) with SF3A3 expression in liver hepatocellular carcinoma (LIHC). a Correlation between SF3A3 expression and TMB. b Correlation between SF3A3 expression and MSI. c TMB in SF3A3 high- vs. low-expression groups. d Neoantigen load in SF3A3 high- vs. low-expression groups. e MANTIS score in SF3A3 high- vs. low-expression groups. f Mutational landscape in LIHC, highlighting TP53 mutation frequency in SF3A3-high group.

Gene Effect Score of SF3A3 in Liver Tumor Cell Lines

SF3A3 gene expression in SNU423 cell line has the most positive gene effect score of 0.3183, and SF3A3 gene expression in JHH1 cell line has the most negative gene effect score of 0.8648 (online suppl. 1E). This may reflect SF3A3 tumorigenic effect and its role as oncogene in LIHC.

Immune Correlations and Immune Checkpoint of SF3A3

Immune cell correlation with SF3A3 gene expressions is shown in Figure 6a (p < 0.05). The correlation between SF3A3 and the expression of immune checkpoint genes (CD274, CTLA4, HAVCR2, LAG3, PDCD1, PDCD1, PDCD1LG2, SIGLEC15, and TIGHT) is shown in Figure 6b. Immune cells with a strong positive correlation with SF3A3 were: M2-type macrophages, Tregs, myeloid-derived dendritic cells, etc. Immune cells with a strong negative correlation with SF3A3 were: CD8⁺ T cells, NK cells, M1-type macrophages, etc. The expression of SF3A3 was significantly positively correlated with multiple immune checkpoint genes (such as PD-L1, CTLA-4, LAG-3, TIGIT, etc.), especially in hepatocellular carcinoma. This indicates that the high expression of SF3A3 may be closely related to the immunosuppressive microenvironment.

Fig. 6.

Immune correlations and fibroblast infiltration associated with SF3A3 expression in liver hepatocellular carcinoma (LIHC). a Heatmap showing correlation between SF3A3 expression and immune cell infiltration scores. b Correlation between SF3A3 expression and immune checkpoint gene expression. c Correlation between SF3A3 expression and cancer-associated fibroblast (CAF) infiltration estimated by EPIC and MCP-COUNTER algorithms.

Analysis of Tumor-Related Fibroblast Infiltration

A recent published study indicated that tumor-related fibroblasts of stroma play an important role in regulating various tumor immune cells infiltration [25]. Therefore, we used EPIC and MCP-COUNTER calculations to explore the relation of tumor-related cell infiltration with SF3A3 expression level in LIHC tumor. SF3A3 expression level was positively associated with tumor-related fibroblast infiltration in LIHC (Fig. 6c). Analysis of tumor microenvironment composition revealed a significant positive correlation between SF3A3 expression and the infiltration level of cancer-associated fibroblasts (CAFs) (Rho = 0.253, p < 0.001***), as estimated by the EPIC algorithm. In contrast, no significant association was observed between SF3A3 and tumor purity (Rho = 0.048, p > 0.05). This suggests a specific link between SF3A3 and the fibroblastic component of the tumor stroma.

SF3A3-Related Genes and Proteins Analysis

We identified 50 proteins that interacted with SF3A3 by using STRING (Fig. 7a). According to the BioGRID 4.3 database, SF3A3 physically interacts with SNRPE, SF3A2, SF3B5, SNRPA1, SNRPB2, SNRPC, SNRPD2, TRA2B, SART1, and EP300 proteins (Fig. 7b). To further explore the function and mechanism of SF3A3 play during tumorigenesis/tumor progression, we extracted the top 100 genes most highly correlated with SF3A3 expression mode across TCGA database LIHC tumors. The top 100 genes have well-characterized roles in processes such as mRNA splicing via spliceosome, DNA repair, mRNA processing, DNA replication, cell division, RNA binding and protein binding, and so on (Fig. 7c). KEGG signaling pathway enrichment analysis results showed that these genes are related to spliceosome, nucleocytoplasmic transport, and cell cycle (Fig. 7d). Gene expression correlation results showed that SF3A3 has the most correlation with UTP11L (R = 0.89), SFPQ (R = 0.86), GNL2 (R = 0.83), PRPF38A (R = 0.83), SNRNP40 (R = 0.83), RNF220 (R = 0.81), DPH2 (R = 0.8), CDCA8 (R = 0.8), and YBX1 (R = 0.8) (online suppl. 1F). PRPF38A and SNRNP40 are genes encoding core spliceosome components. SFPQ and RNF220 are involved in spliceosome regulation and assembly. UTP11L, GNL2, and DPH2 are involved in ribosome biogenesis. CDCA8 and YBX1 have primary functions outside the spliceosome/ribosome but are critically linked to cell proliferation. Prognostic analysis results showed that the high expression of these most correlated genes also led to poor prognosis of LIHC tumor (Supplement 1G). According to our results, SF3A3 may mainly manipulate spliceosome pathway to promote tumorigenesis/tumor progression. Figure 7e shows gene set enrichment analysis enrichment score plot. We observed that GABA-A receptor complex and FGFR3 mutant receptor activity were enriched in the SF3A3 high-expression group.

Fig. 7.

SF3A3-related gene and protein analysis. a Protein-protein interaction network of SF3A3 from the STRING database. b Physical interactors of SF3A3 from BioGRID 4.3. c Top 100 genes co-expressed with SF3A3 in liver hepatocellular carcinoma (LIHC), with Gene Ontology (GO) functional annotation. d Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of SF3A3 co-expressed genes. e Gene set enrichment analysis (GSEA) of SF3A3 high-expression group. Enrichment score plots showing top enriched pathways (GABA-A receptor complex and FGFR3 signaling) in SF3A3 high-expression LIHC tumors.

Discussion

The TCGA database contains high-throughputs sequencing and clinical data for 33 tumor types making it a valuable resource for exploring the molecular and genetic mechanisms underlying different cancers [26–29]. Due to the general development of bioinformatics analysis and wide application of the TCGA database [30–32], the discovery of tumor-related molecular markers to provide clinical therapeutic strategies has been widely valued [33–35]. Liver cancer is one of the most lethal cancer diseases in the world. The pathogenic factors vary in different environments, and there are also individual differences in patients of different ages/genders. The aim of this study was to investigate the role of SF3A3 and related genes in the progression of liver tumors and to assess its prognostic value. Unlike other splicing factors studied in HCC (e.g., SF3B1, SRSF1), SF3A3 is a core component of the SF3A complex, which is essential for the early stages of spliceosome assembly. Its unique position in the splicing machinery, coupled with its specific interactions with cellular stress response proteins (CSR1) and nuclear receptors (CAR), suggests that SF3A3 may have distinct, nonredundant functions in tumorigenesis, particularly in coordinating splicing with metabolic and stress signaling pathways. SF3A3 has been previously studied in multiple TCGA database tumors and has been shown to be associated with the occurrence and progression of some tumor types [17, 36]. However, there are few comprehensive studies based on SF3A3 and its role in liver cancer progression.

SF3A3 gene encodes subunit 3 of the splicing factor 3a protein complex. The splicing factor 3a heterotrimer includes subunits 1, 2, and 3 and is necessary for the in vitro conversion of 15S U2 snRNP into an active 17S particle that performs pre-mRNA splicing. Subunit 3 interacts with subunit 1 through its amino terminus while the zinc finger domain of subunit 3 plays a role in its binding to the 15S U2 snRNP. This gene has a pseudogene on chromosome 20. Alternative splicing results in multiple transcript variants. One well-known molecular function of SF3A3 is that it enables RNA binding and protein binding [37, 38]. Many core proteins in the DDR pathway are encoded by genes that are themselves highly susceptible to splicing errors. This creates vulnerability. When SF3A3 is dysregulated (overexpressed or mutated), it causes widespread splicing defects, and the transcripts of DNA repair genes are among the most vulnerable. With DNA repair systems compromised, the cell becomes vulnerable to accumulating DNA damage. This leads directly to the hallmarks of cancer genomes: increased mutation burden, specific mutational signatures, and chromosomal instability.

SF3A3 is closely related to the development and prognosis of many tumors, such as breast cancer, bladder cancer, and non-small cell lung cancer, to name but a few. In HCC, YTHDF2 plays an important role in tumorigenesis, which is related to poor prognostic results of liver tumor patients. A previous study found that SF3A3 gene expression positively correlated with YTHDF2 gene expression, which indicates that they may cooperate to lead to liver tumorigenesis [14]. SF3A3 function is also related to DNA repair that is valuable in distinguishing liver tumor tissues and normal liver tissues [39]. In this study, we comprehensively explored the role of SF3A3 in liver cancer through analyzing indexes such as tissue cell and single-cell expression, comparison between normal liver tissues and liver tumor tissues expression, survival of tumor patients, gene expression, promoter methylation expression, proteomic expression, genetic alterations, tumor mutation burden, MSI, immune cell correlations, immune checkpoint, tumor-related fibroblast infiltration, and SF3A3-related molecular mechanism.

Our analysis found the expression of SF3A3 in liver tumor tissues was significantly higher than in normal liver tissues. Further analysis of SF3A3 expression and clinical prognosis indicates that high SF3A3 expression was related to worse OS and DFS prognosis of liver cancer. Upregulated SF3A3 gene expression level and SF3A3 promoter methylation expression are all associated with different tumor stages, different grades, and TP53 mutation factors. Upregulated SF3A3 proteomic expression level is related to age factors.

Many studies have shown that gene mutations have an important effect on tumor progression, drug resistance, and treatment [40, 41]. Our study concluded that SF3A3 gene missense mutation accounts for most gene alterations in LIHC tumor and is in close connection with TP53 mutation pathway. SF3A3 expression also has correlation with TMB and MSI.

The interaction between tumor cells and immune cells plays an important role in tumor progression [42], and the expression level of immune checkpoint genes can affect the tumor immune microenvironment [43, 44]. CAFs are the main constitutes of stromal cells, which are reported to have close connection with treatment resistance, worse tumor prognostic results, and disease reappearance [45–48].

Our study found SF3A3 expression level was associated with immune cell expression and immune checkpoint genes’ expression, as well as CAFs infiltration in LIHC tumors. As a splicing factor, SF3A3 may affect the recruitment and function of immune cells in the tumor microenvironment by regulating the alternative splicing of immune-related genes. SF3A3 is positively correlated with the expression of multiple immune checkpoint genes, suggesting that high expression of SF3A3 may identify a type of immunosuppressive tumor, which may be more sensitive to immune checkpoint inhibitors. SF3A3 may serve as a potential biomarker for combined immunotherapy to screen patients who may benefit from anti-PD-1/CTLA-4 therapy [49]. From a mechanistic perspective, SF3A3 may indirectly promote tumor immune escape by regulating the splicing or expression of immune checkpoint molecules. These findings underscore several potential clinical applications. First, given its significant upregulation in tumor versus normal tissue, SF3A3 could be explored as an auxiliary diagnostic marker, particularly for distinguishing early-stage HCC from benign liver lesions. Second, the strong positive correlation with multiple immune checkpoints (PD-L1, CTLA-4, LAG3) suggests that high SF3A3 expression might identify an “immune-desert” or “exhausted” phenotype, potentially serving as a predictive biomarker for patient response to immune checkpoint inhibitors. Patients with high SF3A3 expression could be prioritized for combination immunotherapy strategies. Third, as a core spliceosome component, SF3A3 is a druggable target. Small-molecule inhibitors of the spliceosome, such as pladienolide B and its derivatives (e.g., H3B-8800), are under clinical investigation. Our data provides a rationale for evaluating such spliceosome inhibitors specifically in SF3A3-high LIHC patient subsets. The significant association between SF3A3 and CAF infiltration, independent of tumor purity, uncovers a potential non-cell autonomous role for SF3A3 in shaping a pro-tumorigenic stroma. We hypothesize that SF3A3 may regulate the alternative splicing of key secreted factors that recruit or activate CAFs, thereby fostering an immunosuppressive and pro-fibrotic microenvironment conducive to tumor progression and therapy resistance. This link positions SF3A3 not only as a cell-intrinsic oncogene but also as a modulator of the tumor ecosystem, suggesting that its inhibition could normalize the tumor stroma and enhance the efficacy of conventional therapy and immunotherapy [25]. Recent reviews have highlighted the critical roles of various molecules in modulating the tumor immune microenvironment and cancer progression. For instance, ZG16B has been identified as a key regulator in these processes across multiple cancer types, further underscoring the importance of exploring novel molecular players like SF3A3 in the context of tumor-stroma-immune cross talk [50]. Our findings add SF3A3 to this growing list of pivotal regulators. Overall, our analysis results elucidate the potential value of SF3A3 expression in tumor immunology and clinical prognostic results of LIHC tumor.

Additionally, we discovered genes related to SF3A3 using GEPIA2. These similar genes are characteristic genes that involve spliceosome pathways, which have well-characterized functions in tumorigenesis, such as mRNA splicing via spliceosome, DNA repair, mRNA processing, DNA replication, cell cycle, cell division, nucleocytoplasmic transport, RNA binding and protein binding, and so on. Besides, SF3A3 physically interacts with SNRPE, SF3A2, SF3B5, SNRPA1, SNRPB2, SNRPC, SNRPD2, TRA2B, SART1, and EP300 proteins. These findings lay the foundation for investigation of SF3A3 gene function and signal pathway mechanism.

GABA-A receptor activity and FGFR3 mutant receptor activity were both significantly enriched in the SF3A3 high-expression group. But how might SF3A3, a spliceosome component, mechanistically intersect with these pathways? We propose a splicing-dependent model. For the FGFR3 pathway, SF3A3 overexpression could alter the alternative splicing of FGFR3 itself. This is plausible as FGFR3 undergoes alternative splicing, generating isoforms with different ligand-binding affinities and constitutive activities. An SF3A3-driven shift toward isoforms that resemble the oncogenic FGFR3Δ7-9 mutant (which lacks the acid-box domain) would lead to ligand-independent activation of downstream AKT and MAPK signaling, promoting proliferation and metastasis. For the GABA-A receptor complex, which is composed of multiple subunits (e.g., α, β, γ), the specific subunit composition dictates receptor function. SF3A3 could regulate the alternative splicing of genes encoding these subunits (e.g., GABRA2, GABRG2), potentially biasing assembly toward oncogenic isoforms that exhibit constitutive activity or altered chloride conductance. This would not only promote cancer cell proliferation but also suppress T-cell activity within the tumor microenvironment, contributing to immune evasion. Our findings of positive correlations with Treg infiltration and negative correlations with CD8⁺ T cells are consistent with this model. Therefore, in SF3A3-high LIHC, dysregulated splicing rewires both cell-intrinsic (FGFR3) and cell-extrinsic (GABA-A) signaling to drive tumor progression and immune suppression. Overexpression of genes encoding key functional components of these receptor-activated pathways may contribute to enhanced tumor growth and cancer progression. Elevated FGFR3 activity caused by activating mutations significantly enhances HCC cell proliferation through the TET2-PTEN-AKT pathway [51] and has been associated with reduced infiltration and activation of inflammatory CAFs in the tumor microenvironment [52]. Elevated FGFR3 activity is known to promote the spread of HCC through metastasis via facilitating MCP-1-mediated angiogenesis [53]. Additionally, elevated FGFR3 promotes an immunosuppressive tumor microenvironment by reducing T-cell cytotoxicity and immune cell infiltration and increasing the number of immunosuppressive TREM2+ macrophages [52, 54]. Therefore, by creating a “cold” tumor microenvironment characterized by sparse immune cell infiltration, the tumor becomes less vulnerable to immune attack thereby facilitating cancer progression and metastasis [54]. GABA-A receptor is also implicated in cancer progression. It is overexpressed in HCC where it promotes the proliferation of cancer cells [55] and can suppress T-cell proliferation and immune activity, thereby contributing to tumor immune evasion [56]. Therefore, the elevated GABA-A and FGFR3 mutant receptor activity in the SF3A3 high-expression group, may suggest SF3A3 contributes to tumor progression through dysregulation of these receptor complexes and their associated pathways.

Limitation

Of course, there are some limitations in this dataset-based analysis study. Our analysis suggests that SF3A3 is a potential biomarker for the diagnosis and prognosis of liver cancer and a promising therapeutic target for LIHC. Several important limitations must be acknowledged. First, the results are primarily based on RNA sequencing data from TCGA, which, while comprehensive, is retrospective in nature. This design is subject to selection bias and cannot establish causality. Second, the lack of an independent, prospective clinical cohort for external validation limits the generalizability of our prognostic conclusions. Third, the TCGA LIHC dataset predominantly includes samples from Caucasian and Asian populations, with underrepresentation of other ethnic groups; thus, the findings may not be universally applicable. Fourth, and most critically, the study is purely computational and lacks any in vitro or in vivo experimental verification (e.g., knockdown/overexpression models, xenograft studies) to directly assess the oncogenic function and mechanistic hypotheses we have proposed. Limitations of this study lie in that the results of this study were mainly based on the RNA sequencing data of LIHC tissues in the TCGA database, and the activity of SF3A3 downstream signaling pathway and related protein levels in LIHC tissues could not be directly assessed.

Conclusion

In summary, SF3A3 may play a promoting role in LIHC tumor and lead to poor clinical prognosis of liver tumor patients. Tumor mutation burden, MSI, immune cell, immune checkpoint, tumor-related fibroblasts infiltration correlation analysis of SF3A3 could provide possible molecular mechanisms in LIHC tumorigenesis. The analysis of SF3A3-correlated gene indicates that SF3A3 may function cooperatively with some genes in LIHC tumor. Therefore, further experiments and clinical studies are required to verify the precise molecular mechanism of SF3A3 in liver tumor pathogenesis and progression.

Statement of Ethics

This study utilized de-identified and publicly available data from the TCGA and GTEx databases. The original data collection for TCGA and GTEx was approved by the relevant institutional review boards and ethics committees, and written informed consent was obtained from all participants in accordance with the Declaration of Helsinki. As this study involved the secondary analysis of anonymized data, no additional ethical approval or written informed consent was required, as confirmed by the local legislation.

Conflict of Interest Statement

The authors declare that they have no known competing financial interests or personal relationships that could have influenced the work reported in this study.

Funding Sources

No funding was received.

Author Contributions

Conception, design, collection and assembly of data, and manuscript writing: Fangran Liu. Supervision: Paul David Blakeley. Provision of study materials, data analysis and interpretation, review and editing, and final approval of the manuscript: All authors.

Funding Statement

No funding was received.

Data Availability Statement

The data that support the findings of this study are not publicly available due to privacy/ethical restrictions but are available from the corresponding author upon reasonable request. The raw sequencing data used in this study is publicly available from TCGA and GTEx repositories.