Integrative analysis based on ATAC-seq and RNA-seq reveals a novel oncogene PRPF3 in hepatocellular carcinoma

Abstract

Background

Assay of Transposase Accessible Chromatin Sequencing (ATAC-seq) is a high-throughput sequencing technique that detects open chromatin regions across the genome. These regions are critical in facilitating transcription factor binding and subsequent gene expression. Herein, we utilized ATAC-seq to identify key molecular targets regulating the development and progression of hepatocellular carcinoma (HCC) and elucidate the underlying mechanisms.

Methods

We first compared chromatin accessibility profiles between HCC and normal tissues. Subsequently, RNA-seq data was employed to identify differentially expressed genes (DEGs). Integrating ATAC-seq and RNA-seq data allowed the identification of transcription factors and their putative target genes associated with differentially accessible regions (DARs). Finally, functional experiments were conducted to investigate the effects of the identified regulatory factors and corresponding targets on HCC cell proliferation and migration.

Results

Enrichment analysis of DARs between HCC and adjacent normal tissues revealed distinct signaling pathways and regulatory factors. Upregulated DARs in HCC were enriched in genes related to the MAPK and FoxO signaling pathways and associated with transcription factor families like ETS and AP-1. Conversely, downregulated DARs were associated with the TGF-β, cAMP, and p53 signaling pathways and the CTCF family. Integration of the datasets revealed a positive correlation between specific DARs and DEGs. Notably, PRPF3 emerged as a gene associated with DARs in HCC, and functional assays demonstrated its ability to promote HCC cell proliferation and migration. To the best of our knowledge, this is the first report highlighting the oncogenic role of PRPF3 in HCC. Furthermore, ZNF93 expression positively correlated with PRPF3, and ChIP-seq data indicated its potential role as a transcription factor regulating PRPF3 by binding to its promoter region.

Conclusion

This study provides a comprehensive analysis of the epigenetic landscape in HCC, encompassing both chromatin accessibility and the transcriptome. Our findings reveal that ZNF93 promotes the proliferation and motility of HCC cells through transcriptional regulation of a novel oncogene, PRPF3.

Introduction

Hepatocellular carcinoma (HCC) is a prevalent malignancy, ranking second among causes of global cancer deaths [1]. Despite significant advancements in treatment modalities, including surgical interventions, targeted therapies, and immunotherapies, the prognosis for HCC patients remains dismal [2]. This is primarily due to the high prevalence of late-stage diagnoses and a concerning rate of recurrence. Over the decades, systemic therapy has been the mainstay for advanced HCC [3, 4], primarily utilizing receptor tyrosine kinase inhibitors (TKIs) like sorafenib and lenvatinib. Later, treatment has transitioned from single-agent TKIs to combination therapies incorporating anti-PD-L1 monoclonal antibodies (mAbs) like atezolizumab and anti-vascular endothelial growth factor (VEGF) mAbs like bevacizumab [5, 6]. Unfortunately, a significant portion of HCC patients experience disease progression due to primary or acquired resistance to these therapies. While numerous genes associated with HCC progression have been identified, the role of epigenetic mechanisms, particularly chromatin-mediated regulation of gene expression, remains largely unexplored. Investigating these chromatin alterations and their regulatory elements is crucial for identifying novel therapeutic targets.

Eukaryotic cellular chromatin, composed of DNA and proteins, undergoes intricate folding to form chromosomes [7]. Only a small portion, roughly 1%, of the eukaryotic genome comprises accessible elements like promoters, enhancers, and other regulatory sequences [8]. Open chromatin conformation allows transcription factors (TFs) and other regulatory factors to access cis-regulatory elements and activate gene expression, whereas closed chromatin hinders this interaction, leading to gene silencing [9]. Therefore, genome-wide profiling of chromatin accessibility offers a powerful tool to identify potential regulatory elements, including specific TFs and epigenetic regulators [10]. Assay of Transposase Accessible Chromatin Sequencing (ATAC-seq) is a highly reliable method for generating high-resolution chromatin accessibility maps. It directly identifies open chromatin regions with high sensitivity [10]. This technique has been extensively applied to various developmental and disease processes to identify epigenetic landscapes and regulatory factors driving disease pathogenesis [11,12,13].

Here, we utilized ATAC-seq in conjunction with RNA-seq analysis to elucidate the relationship between chromatin accessibility and transcriptional alterations in HCC. We identified PRPF3, a crucial gene associated with differentially accessible regions (DARs), to be upregulated in tumors and correlated with a poor prognosis. Functional assays, encompassing both in vivo and in vitro models, validated the hitherto undocumented oncogenic role of PRPF3 in promoting HCC proliferation and metastasis. Furthermore, we identified and characterized ZNF93 as an upstream transcription factor that exerts its oncogenic function by upregulating PRPF3 expression.

Methods and materials

Samples

Four tumor and adjacent normal tissue samples were collected from patients with HCC at the Tianjin First Central Hospital between December 2019 and April 2020. Each sample included tumor tissue (denoted as T) and adjacent normal tissue located more than 1 cm away from the tumor edge (denoted as P). Pathological examination confirmed the tumor samples as HCC and verified the normal tissues to be tumor-free. The general information on the patients and samples is provided in Table S1. This study was approved by the Ethics Committee of Tianjin First Central Hospital (2021N066KY).

ATAC-seq data processing and analysis

Four pairs of tumor and adjacent normal tissue samples were submitted to Shanghai Boho Biotechnology Co. for ATAC-seq sequencing. Following quality assessment, the sequencing results passed quality control and were deemed suitable for downstream analysis to extract biological information. Clean reads were obtained using FASTX and aligned to the human reference nce genome (hg38) with Bowtie2. Non-nuclear and duplicate reads were filtered out using SAMtools, and only reads with a mapping quality score ≥ 30 were retained in BAM format for subsequent analysis. MACS 2 was utilized for peak calling. Differential chromatin accessibility regions between tumor and normal tissues were identified using MANORM2. Peak annotation to the genome and nearby genes was achieved using the R package ChIPseeker. Functional enrichment analysis of the identified peaks within KEGG pathways was conducted using the R package ClusterProfiler. Transcription factor binding motif identification within these peaks was performed with the HOMER utility findMotifsGenome.pl. Publicly available ATAC-seq data for Liver Hepatocellular Carcinoma (LIHC) samples was obtained from the TCGA database. This promoter peak data was then normalized and corrected for further analysis.

RNA-seq data processing

TCGA database provided the LIHC RNA-seq data, encompassing 371 tumor samples and 50 adjacent normal samples. Differential mRNA expression analysis was performed using the R package edgeR. Differentially expressed genes (DEGs) were identified by applying a threshold of |log2(fold change)|> 1 and a p-value significance level of less than 0.05.

Correlation analysis of DARs and DEGs

Integration of the expression matrices for DARs and DEGs was performed using the R environment. Subsequently, a random correlation analysis was conducted. The Pearson correlation coefficient was employed, and a p-value threshold 0.05 was used to define statistically significant correlations.

Survival analysis

For survival analysis, we employed the R package "survival". The "surv_cutpoint" function from the survminer R package was utilized to identify the optimal cut-off point for a relevant clinical variable. Subsequently, patients were stratified into two groups based on this cut-off value, and Kaplan–Meier survival curves were generated.

Identification of upstream transcription factors (TFs)

Four sets of DARs served as input files for the MEME FIMO tool. This tool was employed to predict potential transcription factor binding motifs and their corresponding binding sites within the input sequences. The mammalian transcription factor database served as the reference for motif identification. To further refine the analysis, the promoter binding region was defined as the sequence encompassing 2000 base pairs upstream and 200 base pairs downstream of the target genes' transcription start sites. Subsequently, only motifs with predicted binding sites located within this defined promoter binding region were retained.

Immunohistochemical staining

The tissue sections were first dewaxed and rehydrated, then subjected to antigen retrieval in EDTA. Endogenous peroxidase activity was blocked using 3% hydrogen peroxide, and the sections were blocked with 5% BSA. The sections were then incubated with the primary and secondary antibodies sequentially. Next, the sections were stained with diaminobenzidine chromogen and counterstained with hematoxylin. Finally, the stained sections were dehydrated and mounted with coverslips. The staining results were examined and quantified by two pathologists independently.

Cell lines and transfection

HCC cell lines, including MHCC-97 h, Huh7, HepG2, and Jhh-7, and the normal liver cell line L02, were obtained from the Cell Bank of the Chinese Academy of Sciences. These cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) in a humidified incubator maintained at 37 °C with 5% CO₂. To silence the expression of PRPF3 and ZNF93, two independent small interfering RNAs (siRNAs) were employed. Both the siRNAs and a non-targeting control siRNA (siNC) were designed and synthesized by Genepharma (Jiangsu, China). Once the cells reached 50–70% confluence, transfection with siRNA-reagent complexes was performed. For overexpression experiments, recombinant lentiviruses harboring the target genes and empty vector control lentiviruses were synthesized and constructed by OBIO (Shanghai, China). HCC cells were infected with these lentiviruses when they reached 40–50% confluence. Following a 72-h infection period, puromycin selection was used to establish stable cell lines with overexpression of the target genes (Table 1). The sequences of the siRNAs and viruses utilized in this study are provided in Tables S1 and S2.

Table 1 The characteristics of 4 HCC and normal tissue samples

Quantitative RT-PCR and Western blotting

Total RNA was extracted from the transfected siRNA and viral-infected HCC cells using TRIzol reagent (Vazyme Biotech, Nanjing, China) according to the manufacturer's instructions. The extracted RNA was then reverse-transcribed into cDNA using the ExScript RT-PCR kit (TaKaRa, Japan) following the manufacturer's instructions. SYBR Green Master Mix (Vazyme Biotech) was used to perform qRT-PCR. The reactions were run in quadruplicate, and the values were normalized to β-actin. The relative expression of the target genes was calculated using the 2^−ΔΔCt method. The primer sequences used for PCR amplification are listed in Supplementary Table S3.

Proteins were collected using RIPA lysis buffer. Total proteins were separated using 10% SDS-PAGE and transferred to PVDF membranes under constant current. After blocking with 5% non-fat milk for 2 h, the membranes were incubated with primary and secondary antibodies sequentially. Finally, the protein bands were visualized using a chemiluminescence detection method. The protein band intensities were quantified using ImageJ software. The antibody information is provided in Supplementary Table S4.

Cell proliferation, colony formation, and EdU incorporation assays

For the CCK-8 assay, cells were seeded in 96-well plates (3,000 cells/well) and cultured for 3 days. Then, 10 μL of CCK-8 reagent was added to each well, and the plates were incubated for 2 h in the dark. The absorbance was measured at 450 nm. For the colony formation assay, cells were seeded in 6-well plates (1,000 cells/well) and cultured for 14 days. When visible colonies were formed, fix and stain the cells. The number of colonies per well was then counted. The cell proliferation was assessed using the EdU (5-ethynyl-2'-deoxyuridine) assay kit (#E-CK-A376, Elabscience, Wuhan, China). Cells were incubated with the EdU reagent according to the manufacturer's instructions. The number of EdU-positive cells was counted under a fluorescence microscope.

Cell migration and scratch assay

In the Transwell experiment, 200 μl of serum-free culture medium containing 4 × 10⁴ cells was added to the upper chamber, and 500 μl of complete culture medium was added to the lower chamber as a chemoattractant. After 24 h, the inserts were removed, the cells were fixed and stained, and the non-migrated cells on the upper surface of the membrane were gently removed with a cotton swab. The migrated cells were then imaged using an inverted microscope (× 20) and counted by ImageJ. For the invasion assay, a layer of Matrigel was added to the upper chamber to form an extracellular matrix barrier to evaluate the cell invasion ability. For the scratch wound assay, cells were seeded in a 6-well plate and cultured until they reached 100% confluence. A sterile pipette tip was used to create a vertical scratch wound in the cell monolayer. After washing with PBS to remove floating cells, a serum-free culture medium was added, and the cells were further incubated. Images of the scratch wound were captured by an inverted microscope (× 10) at 0, 24, and 48 h. The migration rate was calculated as follows: [0 h scratch width–48 h scratch width]/0 h scratch width × 100%.

In vivo proliferation assay in mice

All animal experiments were approved by the Animal Protection Committee of Tianjin First Central Hospital. Four-week-old female BALB/c nude mice were obtained from SPF (Beijing) Biotechnology Co., Ltd. and randomly divided into two groups (n = 4 per group). The experimental group received subcutaneous injections of Huh7 cells stably overexpressing PRPF3 and ZNF93 at a dose of 6 × 10⁶ cells on the right side, while the control group received an equivalent number of Huh7 cells infected with empty virus. On the 14th day, the mice were euthanized, and subcutaneous tumors were excised and photographed. Tumor weight and volume were recorded, with tumor volume calculated using the formula: Tumor volume = (length × width²)/2.

Statistical analysis

Statistical analysis of experimental data was conducted using GraphPad Prism 8.0 software. Non-paired t-test was used to compare two groups, and a one-way analysis of variance (ANOVA) was used to compare three or more groups. All quantitative experiments were repeated three times, and the results were presented as mean ± standard deviation (SD). A p-value less than 0.05 was considered statistically significant. The significance level was denoted by asterisks as follows: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Results

Identifying differential accessible chromatin regions between liver cancer tissue and normal tissues by ATAC-seq technology

We first collected liver cancer and paired normal tissue samples from four patients and performed ATAC-seq sequencing (Fig. S1A). The fragment size distribution of the ATAC-seq samples showed that most insert fragments were relatively small, representing the open chromatin regions between nucleosomes (Fig. S1B). Figure S1C, D demonstrated the enrichment of ATAC-seq signals mainly around transcription start sites (TSSs), which is consistent with previous research, indicating the reliability of sequencing quality. We then annotated these open chromatin regions, revealing that most of the peaks were in promoter and intergenic regions, with a smaller proportion in intronic regions (Fig. S1E), which suggest that transcription factors tend to bind to the promoter regions around TSSs. Subsequently, we used the MAnorm2 algorithm to identify the DARs between liver cancer and normal tissues. We categorized the peaks into those enriched in cancer, enriched in normal, and with no significant difference. (The significance of the differences was calculated via M-values) (Fig. S2A–D).

Genomic characterization and enrichment analysis of differential accessible chromatin regions

To study the gene regulatory differences between tumor and normal tissues, we performed genomic characterization and enrichment analysis of the DARs. We defined peaks with M-value > 0.5 or < − 0.5 as significantly differential, and obtained 605 peaks significantly enriched in the cancer samples (termed "up") and 400 peaks significantly enriched in the normal samples (termed "down"). Genomic annotation showed that regardless of whether the peaks were enriched in cancer or normal tissues, they were mainly distributed in promoter and distal intergenic regions, with fewer in intronic regions (Fig. 1A), which suggests that the proximal regulatory elements, such as promoters, have higher transcriptional activity in both liver cancer and normal tissues. To further analyze the functions of these DARs, we annotated them to their regional genes and performed enrichment analysis. KEGG pathway enrichment analysis showed that the peaks enriched in cancer samples were mainly associated with genes involved in cell–cell adhesion, HCC, gastric cancer, colorectal cancer, MAPK signaling, and FoxO signaling pathways, while the peaks enriched in normal samples were mainly associated with genes involved in TGF-β, cAMP, and p53 signaling pathways (Fig. 1B). To identify the key transcription factors (TFs) driving the expression differences between cancer and normal tissues, we used HOMER to scan the ± 200 bp flanking regions of the ATAC-seq peaks in the DARs for TF motifs and TF binding sites (TFBS). We found that the DARs enriched in tumor tissues were primarily associated with transcription factor families such as the ETS family, including ETS1, PU.1, Etv2, and Elf4. Additionally, we found that members of the activator protein-1 (AP-1) family, including Jun (JUN-AP1) and Fos (Fosl2), were among the top predicted motifs in the cancer-specific hyper-accessible regions. In contrast, the DARs enriched in normal tissues were mainly associated with CTCF and its paralog BORIS (Fig. 1C). This indicated that changes in chromatin accessibility affected the gene expression of many components of signaling cascades and regulators, and played a critical role in HCC progression.

Fig. 1

Identification and analysis of differential chromatin accessible regions. A Pie charts showing the proportion of up-regulated (left) and down-regulated (right) chromatin accessible regions within the indicated genomic regions in hepatocellular carcinoma(HCC), compared with adjacent normal tissue. B The KEGG pathways significantly enriched by genes in or nearby the up-regulated (left) and down-regulated (right) chromatin accessible regions in HCC. C The top 15 known TF motifs enriched in the up-regulated (left) and down-regulated (right) chromatin accessible regions in HCC

The DARs were positively correlated with the nearest DEGs

To identify DARs potentially driving DEGs, we retrieved matched ATAC-seq and transcriptomic data for liver cancer from TCGA. Leveraging genomic annotation files, we observed that the open chromatin regions within the TCGA-LIHC-ATAC-seq data primarily resided in promoter regions (Fig. S3A). This finding indicated high consistency between the database's genomic distribution characteristics and our sequencing data. Differential analysis of the TCGA LIHC transcriptomic data revealed a total of 1853 DEGs, with 1219 upregulated and 634 downregulated genes (Fig. S3B). By comparing our previously generated DARs with the TCGA-LIHC-ATAC-seq peaks, we identified four sets of overlapping regions: O1 (731 peaks), O2 (559 peaks), O3 (691 peaks), and O4 (1138 peaks) (Fig. S3C). To further investigate potential correlations between DARs and DEGs, we assigned DARs in O1–O4 to their nearest genes. Integration with RNA-seq data revealed 432 and 296 DEGs associated with hyper-accessible regions in tumor and normal tissues, respectively. Notably, most DEGs associated with tumor-specific hyper-accessible regions were upregulated in tumors, while most DEGs associated with normal-specific hyper-accessible regions displayed downregulation (Fig. S3D, E). These findings indicate that the alterations in chromatin accessibility contribute to the differential expression of relevant genes in hepatocellular carcinoma.

DEGs associated with changes in chromatin accessibility are highly expressed in liver cancer

To further elucidate gene expression changes driven by chromatin accessibility alterations, we integrated the expression matrices of the four DAR groups and DEGs. Following correlation analysis and annotation of peaks to their nearest genes, we identified 22 DEGs significantly associated with changes in chromatin accessibility across the four groups (Table 2). Figure 2A depicts a correlation heatmap between chromatin accessibility and associated DEGs across the four sample groups. Notably, PRPF3 and PLIN2 emerged as consistently present across all four groups (Fig. 2B). The correlation analysis between the PRPF3 gene and its corresponding peak revealed a Pearson correlation coefficient of 0.57 (Fig. 2C). Visualization of the ATAC-seq data demonstrated that the PRPF3 transcription start site exhibited a higher peak intensity in tumor samples compared to normal tissues, signifying increased chromatin accessibility (Fig. 2D). Consistent with this finding, RNA-seq data revealed elevated PRPF3 expression in liver cancer (Fig. 2E). We further validated this observation by demonstrating elevated PRPF3 expression in liver cancer cell lines (Fig. 2F, G). Given the high PRPF3 expression in HCC, we investigated its potential as a prognostic biomarker. Analysis of TCGA data revealed a significant association between PRPF3 upregulation and poor patient prognosis (Fig. 2H). Additionally, immunohistochemical staining of four paired tumor and normal tissue samples demonstrated higher PRPF3 expression level in tumor tissues compared to normal tissues (Fig. 2I). These findings collectively suggest that PRPF3 may play a crucial role in HCC development and progression.

Table 2 Positively correlated DARs and DEGs in 4 sets of samples

Fig. 2

Identification and expression analysis of differential genes correlated with chromatin accessibility changes positively. A The heatmap showing the differentially accessible regions (DARs) in chromatin and differentially expressed genes(DEGs) across the four pairs of samples. B The Venn diagram shows that PRPF3 and PLIN2 are the common DARs-associated DEGs across the four sample groups. C The correlation analysis between PRPF3 and its corresponding DARs. The Pearson’s correlation coefficient (R) and the corresponding P value are shown. D The panel showing the chromatin accessibility at the PRPF3 gene locus in 4 pairs of tumor and adjacent normal tissue samples, with red and blue boxes highlighting regions with differential accessibility between them. E PRPF3 expression level between HCC and normal samples in TCGA LIHC database. F, G The expression of PRPF3 in the HCC cell lines, including MHCC-97 h, Huh-7, HepG2, Jhh-7, and the normal liver cell line LO2 were investigated using qRT-PCR(F) and western blotting(G). H The Kaplan–Meier survival curves stratified by PRPF3 expression level. I Representative immunohistochemical staining of PRPF3 protein in 4 pairs of tumor and adjacent normal tissues (scale bar, 100 μm)

Knockdown of PRPF3 inhibits HCC cell proliferation, migration, and invasion in vitro

To investigate the functional role of PRPF3 in HCC, MHCC-97 h, and Huh7 cell lines were selected for further experimentation. Western blot and RT-qPCR analyses confirmed the efficacy of two independent siRNAs (siPRPF3-1 and siPRPF3-2) in downregulating PRPF3 expression at both the mRNA and protein levels (Fig. 3A, B). The CCK8 assay revealed a significant decrease in absorbance in the siPRPF3-1 and siPRPF3-2 groups compared to the siNC group at 24, 48, and 72 h, indicating reduced cell viability (Fig. 3C). Colony formation assays further demonstrated a decline in the colony formation ability of PRPF3-knockdown HCC cells (Fig. 3D). The EdU incorporation assay corroborated these findings, showing a reduction in the number of proliferating HCC cells following PRPF3 knockdown (Fig. 3E, F). Collectively, these results suggest that PRPF3 knockdown suppresses HCC cell proliferation. Next, Transwell assays were performed to assess migration. As shown in Fig. 3G and H, the number of cells migrating through the membrane was significantly reduced in the PRPF3 knockdown groups. Similarly, wound healing assays revealed a marked decrease in wound closure rate in the knockdown groups (Fig. 3I, J), demonstrating that PRPF3 knockdown also impairs the migration and invasion abilities of HCC cells.

Fig. 3

Knockdown of PRPF3 inhibits the proliferation, migration, and invasion of liver cancer cells in vivo. A, B The knockdown efficiency of PRPF3 was confirmed in the MHCC-97 h and Huh7 cell lines using qRT-PCR(A) and western blotting(B). C CCK8 analysis of cell viability in MHCC-97 h and Huh7 cells transfected with siNC, siPRPF3-1, or siPRPF3-2. D The colony formation images of MHCC-97 h and Huh7 cells transfected with siNC, siPRPF3-1, or siPRPF3-2 on day 14. E, F EdU proliferation assay in MHCC-97 h and Huh7 cells transfected with siNC, siPRPF3-1, or siPRPF3-2. G, H Transwell assay in MHCC-97 h and Huh7 cells transfected with siNC, siPRPF3-1, or siPRPF3-2. I, J Wound healing assay in MHCC-97 h and Huh7 cells transfected with siNC, siPRPF3-1, or siPRPF3-2

PRPF3 enhances HCC cell proliferation in vitro and in vivo and promotes migration and invasion in vitro

Having established that PRPF3 knockdown suppressed proliferation, migration, and invasion of HCC cells, we investigated whether its overexpression could conversely promote these processes. Stable PRPF3-overexpressing cell lines were established in MHCC-97 h and Huh7 cells via lentiviral transfection (Fig. S4A, B). CCK8, colony formation, and EdU incorporation assays consistently demonstrated that PRPF3 overexpression enhanced the proliferative capacity of HCC cells (Fig. S4C–E). Similarly, Transwell and wound healing assays revealed that PRPF3 overexpression promoted the migration and invasion abilities of HCC cells (Fig. S4F, G).

To assess the in vivo effects of PRPF3 on tumor progression, stable PRPF3-overexpressing Huh7 cells were subcutaneously injected into the right flank of nude mice. A control group received cells transfected with the empty pWPXL vector. Tumor volume and weight were significantly lower in the control group compared to the PRPF3-overexpressing group (Fig. S4H). These findings collectively demonstrate that PRPF3 can promote HCC cell proliferation both in vitro and in vivo.

The TF ZNF93 regulates the expression of PRPF3

To identify potential upstream transcription factors regulating PRPF3 expression, we employed MEME FIMO to scan the DARs for transcription factor binding motifs. Given that transcription factors typically exert their regulatory effects by binding to gene promoter regions, we subsequently screened for factors with binding sites located within 2 kb upstream to 100 bp downstream of the PRPF3 transcription start site. Table 3 summarizes the top five transcription factors identified from each set of scanning results that satisfied these criteria. Notably, ZNF93 emerged as a potential transcription factor binding to the PRPF3 promoter region in all four DAR sets, exhibiting consistently high scores. To preliminarily validate the ZNF93-PRPF3 interaction, we utilized the Cistrome Data Browser database. The ChIP-seq results revealed a distinct enrichment peak for ZNF93 at the PRPF3 promoter region (Fig. 4A), suggesting potential binding of ZNF93 to the PRPF3 promoter, which displays increased chromatin accessibility in tumor tissues (Fig. 4B). Furthermore, western blot, RT-qPCR, and immunohistochemistry analyses demonstrated that ZNF93 expression is elevated in HCC cells and tissues (Fig. 4C–E). To functionally validate the regulatory role of ZNF93 on PRPF3 expression, we assessed PRPF3 levels in HCC cells with ZNF93 knockdown or overexpression. As expected, a positive correlation between PRPF3 and ZNF93 expression was observed (Fig. 4F–I). These findings collectively suggest that ZNF93 acts as an upstream transcription factor directly targeting and promoting PRPF3 expression.

Table 3 Transcription factors with binding sites located within the PRPF3 promoter regions in 4 sets of DARs

Fig. 4

ZNF93 positively regulates the expression of PRPF3 and is highly expressed in HCC. A The ChIP-seq data of ZNF93, the red boxes show a significant binding peak in the promoter region of PRPF3. B The panel shows the chromatin accessibility at the ZNF93 gene locus in 4 pairs of tumor and adjacent normal tissue samples, with red and blue boxes highlighting regions with differential accessibility. C, D The expression of ZNF93 in the HCC cell lines MHCC-97 h, Huh-7, HepG2, Jhh-7, and the normal liver cell line LO2 were investigated through qRT-PCR(C) and western blotting(D). E Representative immunohistochemical staining of ZNF93 protein in 4 pairs of tumor and adjacent normal tissues (scale bar, 100 μm). F, G The expression of PRPF3 was decreased in HCC cells after knocking down ZNF93. H, I The expression of PRPF3 was increased in HCC cells with ZNF93 overexpression

Knockdown of ZNF93 inhibits HCC cell proliferation, migration, and invasion in vitro

To investigate the functional role of ZNF93 in HCC cells, MHCC-97 h, and Huh7 cell lines were employed. Two independent siRNAs (siZNF93-1 and siZNF93-2) efficiently downregulated ZNF93 expression (Fig. 5A, B). Subsequent CCK8, colony formation, and EdU incorporation assays consistently demonstrated that ZNF93 silencing inhibited HCC cell proliferation (Fig. 5C–F). Similarly, Transwell and wound healing experiments revealed that ZNF93 knockdown suppressed both migration and invasion abilities of HCC cells (Fig. 5G–J).

Fig. 5

Knockdown of ZNF93 inhibits the proliferation, migration, and invasion of liver cancer cells in vivo. A, B The knockdown efficiency of ZNF93 was confirmed in the MHCC-97 h and Huh7 cell lines using qRT-PCR(A) and western blotting(B). C CCK8 analysis of cell viability in MHCC-97 h and Huh7 cells transfected with siNC, siZNF93-1, or siZNF93-2. D The colony formation ability of MHCC-97 h and Huh7 cells transfected with siNC, siZNF93-1, or siZNF93-2. E, F EdU assay in MHCC-97 h and Huh7 cells transfected with siNC, siZNF93-1, or siZNF93-2. G, H Transwell assay in MHCC-97 h and Huh7 cells transfected with siNC, siZNF93-1, or siZNF93-2. I, J Wound healing assay in MHCC-97 h and Huh7 cells transfected with siNC, siZNF93-1, or siZNF93-2

ZNF93 promotes HCC cell proliferation in vivo and in vitro, and enhances migration and invasion in vitro

To further investigate the functional consequences of ZNF93 overexpression on HCC cells, we established stable ZNF93-overexpressing cell lines in MHCC-97 h and Huh7 cells using lentiviral infection (Fig. S5A, B). CCK8, colony formation, and EdU incorporation assays consistently revealed that ZNF93 overexpression promoted the proliferation of HCC cells (Fig. S5C–E). Similarly, Transwell and wound healing experiments demonstrated that ZNF93 overexpression enhanced the migratory and invasive abilities of HCC cells (Fig. S5F–G). To assess the in vivo effects of ZNF93 on tumor growth, we employed a xenograft tumor model in immunodeficient mice. The results revealed that both tumor volumes and weights were significantly higher in the ZNF93-overexpressing group compared to the control group (Fig. S5H). These findings collectively indicate that ZNF93 can promote HCC cell proliferation both in vitro and in vivo.

PRPF3 is a target oncogene of ZNF93 in HCC cells

To functionally validate whether PRPF3 functions as a downstream target of ZNF93 and mediates its oncogenic effects, we performed rescue experiments. Stable ZNF93-overexpressing Huh7 cells were subjected to PRPF3 knockdown using siPRPF3-1 and siPRPF3-2 (Fig. 6A). CCK8 and EdU incorporation assays demonstrated that PRPF3 downregulation effectively abrogated the ZNF93-induced proliferative effect (Fig. 6B, C). Transwell and wound healing assays further corroborated these findings, revealing that siPRPF3 impedes ZNF93-mediated enhancement of proliferation and migration in HCC cells (Fig. 6D, E). Collectively, these experimental results suggest that PRPF3 likely functions as a functional effector of ZNF93, mediating its tumor-promoting effects.

Fig. 6

Knockdown of PRPF3 in HCC cells overexpressing ZNF93 can suppress the oncogenic effects of ZNF93. A The protein expression profile of ZNF93 and PRPF3 in Huh7 cells overexpressing ZNF93 and with PRPF3 knockdown (siPRPF3-1, -2). B CCK-8 assay, C EdU assay, D Wound healing assay, and E Transwell assay were used to detect the effects of PRPF3 knockdown on the proliferation, migration, and invasion abilities of Huh7 cells stably overexpressing ZNF93

Discussion

Epigenetic regulation refers to the control mechanisms that modulate gene expression without altering the DNA sequence, such as DNA methylation, histone modifications, and non-coding RNAs [14]. These regulatory mechanisms thus influence transcription and expression levels, impacting cellular functions and behaviors. Tumors often exhibit altered epigenetic patterns, and changes in epigenetic regulatory factors contribute significantly to the initiation and progression of tumors [15]. Over the past decades, extensive research has been conducted on the epigenetic landscape of hepatocytes. With the development of sequencing technologies, the methodologies and findings of epigenetic studies in adult HCC can be leveraged to explore the most common aberrant molecular pathways, guide tumor early diagnosis and subtype classification [16], and even inform clinical treatment [17,18,19]. However, an accurate understanding of the chromatin state-defined transcriptional regulatory networks in HCC remains elusive. ATAC-seq is an innovative technique for epigenetic research, identifying open chromatin regions by transposase-mediated fragmentation [10]. Compared to existing methods like FAIRE-seq, DNA-seq, and MNase-seq, ATAC-seq offers significant advantages in its low input cell requirement, simplicity, speed, and accuracy, making it the most widely used approach for genome-wide chromatin accessibility analysis [10]. In this study, we utilized ATAC-seq to identify accessible chromatin regions in HCC and normal tissues, and compared the DARs between them. This allowed us to determine the unique chromatin accessibility signatures, biological pathways, and transcription factors in both HCC tumor and paired normal tissues. In HCC, genes near accessible chromatin regions are primarily involved in cancer-related pathways, including the MAPK and FoxO signaling pathways. Conversely, normal tissues exhibit higher activity in the TGF-β, cAMP, and p53 signaling pathways. These pathways directly influence tumor formation and cellular behavior. The MAPK/ERK signaling pathway is activated by signals from cell surface receptors such as receptor tyrosine kinases (RTKs) or G protein-coupled receptors (GPCRs). Dysregulation of this pathway leads to aberrant cellular behaviors, including increased cell growth and proliferation, dedifferentiation, and survival, all of which exert oncogenic effects [20]. Among various oncogenic signaling pathways, the MAPK/ERK pathway is activated in about 50% of early-stage HCC patients and most late-stage HCC patients [21]. The FOXO proteins are transcription factors regulated by growth factors and stress, which control a wide range of biological processes, including differentiation, cell cycle progression, DNA damage repair, and apoptosis [22, 23]. The FOXO transcription factors have been shown to exhibit both oncogenic and tumor-suppressive properties [24]. For instance, FOXO3a knockdown in MDA-MB-231 breast cancer cells via xenograft models led to smaller tumor sizes and reduced invasion by modulating matrix metalloproteinases while also activating calcitonin-induced proliferation in serum-deprived HCC cells [25, 26]. For HCC, previous studies have confirmed that the FOXO transcription factors, FOXO1 and FOXO3, play critical roles in the malignant progression of HCC [27]. Our analysis aligns with the pathway characteristics reported in prior studies, providing a solid foundation for further bioinformatics analyses. The transcription factor E26 Transformation-Specific Sequence 1 (ETS1) belongs to the ETS gene family, characterized by the ETS DNA-binding domain. ETS1 binds to ETS-binding sites (EBS motifs; 5'-GGAA/T-3') in the promoters or enhancers of its downstream target genes, thereby mediating the regulation of various cellular processes in human cancers, including proliferation, development, metastasis, and angiogenesis [28,29,30]. Previous relevant studies have confirmed that ETS1 can promote the proliferative and migratory capabilities of HCC cells [31]. In the current study, the researchers found that the ETS family transcription factors exhibited heightened activity in the tumor tissues compared to the normal tissues. This suggests that in addition to ETS1, other ETS family members may also be involved in the development and progression of HCC.

Overall, our ATAC-seq results largely align with previous studies on regulatory factors and pathway analysis in HCC. However, to our knowledge, most of the identified DARs have not been studied in sufficient detail to determine how they influence on HCC progression through modulating transcriptional changes. To enable integrated analysis with RNA-sequencing (RNA-seq) data, we overlapped our ATAC-seq data with publicly available HCC data from TCGA database. Integrating RNA-seq data revealed a significant positive correlation between a subset of DARs and DEGs, supporting our hypothesis that increased chromatin accessibility leads to upregulation of essential HCC-associated genes such as FOXQ1, SOX4, and PRPF3. This integrated analysis identified 22 key genes associated with the DARs. Further analysis identified two core regulatory genes, PRPF3 and PLN2, by overlapping the genes identified in all four sample groups. Existing studies have reported the oncogenic role of PLN2 [32]. Precursor mRNA (pre-mRNA) splicing is critical for generating protein diversity. Aberrant splicing is a hallmark of many diseases, particularly cancers [33]. In biological process, splicing is a critical step in gene expression, allowing a single gene to encode multiple protein isoforms and is emerging as a major driver of phenotypic heterogeneity. Understanding the connection between tumor heterogeneity and splicing regulation is crucial for elucidating disease pathogenesis and improving therapeutic development. Splicing represents a promising avenue for identifying novel molecular targets in precision oncology and addressing cancer disparities [34, 35]. Among the splicing factors, pre-mRNA processing factor 3 (PRPF3) is a component of the U4/U6 di-snRNP complex, essential for U4/U6·U5 tri-snRNP formation and recruitment to the active spliceosome, which is critical for pre-mRNA splicing [36, 37]. Relevant studies have shown that PRPF3 plays an oncogenic role in most solid tumors. In gastric cancer, the KCNE2-PRPF3 gene pair effectively predicts recurrence in patients receiving 5-FU-based chemotherapy [38]. Hepatocyte nuclear factor 4α (HNF4α), an essential liver transcriptional regulator, plays a crucial role in liver development and hepatocyte differentiation. One study has shown that PRPF3 is regulated by HNF4α and exhibits increased expression in mouse and human HCC [39]. Li et al. demonstrated that TMEM43 promotes pancreatic cancer progression by stabilizing PRPF3 and regulating the RAP2B/ERK axis [40]. However, the biological function of PRPF3 in HCC remains largely unexplored. Only Liu et al. reported PRPF3 upregulation in various HCC cases, associating it with poor prognosis. They proposed E2F1 as an important PRPF3 regulator, potentially influencing the HCC cell cycle and proliferation [41]. Notably, no prior studies have investigated the effects of PRPF3 on HCC biological functions or the underlying regulatory mechanisms. In our study, we observed a significant increase in PRPF3 transcriptional activity in HCC tissues. Immunohistochemistry and Western blot analyses confirmed abnormally high PRPF3 expression in liver cancer tissues. Finally, functional experiments verified that PRPF3 promotes HCC cell proliferation and metastasis both in vivo and in vitro.

Since transcriptional activity alterations drive changes in PRPF3 expression, and transcription factor binding to promoters/enhancers is the prevalent mode of transcriptional regulation, identifying PRPF3's regulatory transcription factors can offer valuable insights into its physiological role in HCC. Leveraging bioinformatics tools, we screened for transcription factors with binding sites around the PRPF3 promoter region. We found that the transcription factor ZNF93 was repeatedly enriched in four ATAC-seq datasets, suggesting that PRPF3 may be a target gene of ZNF93. This hypothesis was further validated by the positive correlation observed between ZNF93 ChIP-seq data and expression. Zinc finger proteins (ZNFs) constitute the largest human genome transcription factor family, accounting for nearly 5% of the genome, and are characterized by a distinct zinc finger DNA-binding domain [42]. Due to the diversity of the zinc finger structure, ZNF proteins exhibit varying gene regulatory roles in different cellular contexts and under various stimuli. For instance, ZNF306 promotes colorectal cancer development by transcriptionally activating integrin β4 and vascular endothelial growth factor [43]. ZNF384 directly upregulates cyclin D1 expression to promote HCC cell proliferation [44]. In thyroid cancer, ZNF677 suppression enhances the proliferation and colony formation of thyroid cancer cells, and its low expression correlates with poor patient prognosis [45]. ZNF259 activates the ERK/GSK3β/Snail signaling pathway to promote breast cancer cell invasion and migration [46]. These studies suggest that ZNF genes may function as oncogenes and contribute to tumor development. ZNF93, a member of the zinc finger protein family, exhibits aberrant expression in various cancers. A previous study by Cui et al. revealed that ZNF93, the transcription factor, may promote the proliferation and migration of serous ovarian cancer cells through the MYC and G2/M signaling pathways and holds promise as a prognostic biomarker for ovarian cancer [47]. However, the biological function of ZNF93 in HCC remains unclear. Our ATAC-seq visualization analysis suggested high ZNF93 transcriptional activity in tumor tissues, and Western blotting and immunohistochemical staining further confirmed abnormally high ZNF93 expression in tumors. These results indicate that ZNF93 may play a role in HCC biology. Finally, our experiments demonstrated that ZNF93 can promote HCC cell proliferation in vivo and in vitro. Conversely, the functional rescue experiment showed that PRPF3 knockdown significantly reduces the pro-cancer effect of ZNF93 overexpression. These findings confirm that ZNF93 targets and regulates PRPF3, exerting a pro-cancer effect in HCC. This result provides a novel perspective and potential molecular target for HCC-targeted therapy. However, our study has limitations. First, the specific ZNF93 binding site within the PRPF3 promoter remains to be elucidated. Second, the detailed mechanism by which PRPF3 mediates HCC proliferation and migration promotion remains unclear. Further multi-omics research and combined prediction experiments are warranted, which will also be the focus of our future endeavors.

Conclusion

As we know, this study presents the first comprehensive investigation of chromatin accessibility differences between HCC and normal tissues. We identified core genes, including PRPF3 and its upstream regulator ZNF93, associated with these accessibility changes in HCC. Notably, our findings reveal that ZNF93 promotes HCC progression by enhancing PRPF3 transcriptional expression.