Human liver tissue samples (5 cirrhotic liver tissues, 5 HCC tissues, and 5 normal liver tissues) were obtained from patients who underwent liver transplantation or surgical resection at the Eastern Hepatobiliary Surgery Hospital (Shanghai, China). Five normal liver tissues were collected from patients with hemangioma. Five cirrhotic liver tissues were obtained from patients with decompensated cirrhosis who underwent liver transplantation. The clinical data are provided in Supplementary Table S1. Written informed consent was obtained from all patients. The study was approved by the Medical Ethics Committee of the Eastern Hepatobiliary Surgery Hospital.

LX2 cells and HSC-T6 were obtained from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China) and cultured in Dulbecco’s modified Eagle’s medium (DMEM) with 2%–10% fetal bovine serum (FBS) and 1% penicillin–streptomycin. To confirm the reduction of circ_0098181 in activated HSCs, LX2 were stimulated with 10 ng/mL TGF-β1 (PeproTech, 100-21C) for 48 h.

To determine the decrease level of circ_0098181 in liver fibrogenensis, we also measured the expression of circ_0098181 in activated rat primary HSCs on the 14th day after HSCs isolation compared with the quiescent HSCs on the second day. Sprague-Dawley (SD) rats, weighing about 200 g, were purchased from the Experimental Animal Center of Naval Medical University. As previously described (Friedman and Roll, 1987), liver cells were isolated using a mixture of 0.5 mg/mL IV collagenase (C5138, Sigma-Aldrich), 0.5 mg/mL Pronase E (Roche), and 1% DNase. HSCs were then separated by gradient centrifugation with 40% and 60% Percoll (P4937, Sigma-Aldrich). The quiescent HSCs were cultured in DMEM with 10% FBS and 1% penicillin–streptomycin. After 14 days in culture, these quiescent rat HSCs became activated, exhibiting high expression of α-SMA. The expression of circ_0098181 was also determined in auto-activated primary rat HSCs.

Plasmids expressing has_circ_0098181 and the control plasmids (NC) were purchased from Geneseed Biotech (Guangzhou, China). Due to the high homology (91.93%) between the human and rat sequences of circ_0098181, plasmid contained has_circ_0098181 sequence was transfected into TGF-β1 (2.5 ng/mL) pre-treated LX2 cells, HSC-T6 cells, and auto-activated primary rat HSCs with lipofectamine 2000 (Lipofectamine 2000; Thermo Fisher Scientific, America). After 48 h, cells were extracted to perform qRT-PCR and Western blotting. Different primers designed for human and rat circ_0098181 sequences were used to detect the level of circ_0098181 among different species.

CCK-8 assay was performed to explore the roles of circ_0098181 on HSCs proliferation. Approximately 3,000 cells were seeded into each well of a 96-well plate, followed by transfection with circ_0098181 plasmid at the next day. Cell proliferation was then measured daily at 450 nm using the Cell Counting Kit-8 (Dojindo, Tokyo, Japan). Due to the slow proliferation of rat primary HSCs, the medium was refreshed daily to support cell growth.

Additionally, transwell assays were conducted to assess cell migration. Cells pre-transfected with circ_0098181 plasmids were seeded into transwell inserts equipped with 8.0-μm-pore filters (BD, NJ). After 72 h, cell migration was evaluated by counting the cells on the underside of the inserts using ImageJ software (ImageJ 1.x, National Institutes of Health, USA).

Total RNA was extracted from cells and tissues with Trizol Reagent (TaKaRa, Tokyo, Japan). cDNA was synthesized by reverse transcription and qRT-PCR was performed with PrimeScript RT Master Mix and TB Green Premix EX Taq (TaKaRa, Tokyo, Japan). The primers were designed by Primer3 Plus (https://www.bioinformatics.nl/cgi-bin/primer3plus). The detailed primer sequences are shown in Supplementary Table S2.

The proteins were extracted with lysis from cells or tissues, and the supernatant was centrifuged to perform further quantification. Minute (TM) Cytoplasmic and Nuclear Fractionation Kit (SC-003, Invent) was used to extract cytoplasmic protein and nuclear protein, and Western blotting assays were performed according to the manufacturer’s instructions. After SDS-PAGE electrophoresis, proteins were transferred to a PVDF membrane. Then the membrane was blocked in 5% milk and incubated with the primary antibody overnight, followed by 1 h’s incubation with the appropriate secondary antibody. Signals were scanned with the Odyssey infrared imaging scanner (LI-COR Odyssey system). The antibodies are listed in Supplementary Table S3.

For immunofluorescence staining, cells were first fixed with 4% PFA for 10 min, then permeabilized with 0.1% Triton X-100 and blocked with 5% BSA. The cells were incubated with the primary antibody overnight, followed by incubation with the secondary antibody for 1 h the next day. Nuclei were stained with DAPI, and the signals were observed using a confocal microscopy.

Fluorescence in situ hybridization (FISH) was performed to explore the location of circ_0098181 in HSCs according to the manufacturer’s recommended methods. The FISH probe for circ_0098181 and the positive control of cytoplasm (18S probe) were synthesized by Ribobio Company (Guangzhou, China). In brief, the cells were fixed with 4% Paraformaldehyde (PFA) and perforated with 0.5% Triton X-100. Then cells were incubated with a pre-hybridization solution for 30 min, and hybridized with the hybridization solution (including probes, 20 μM) at 37°C overnight. On the next day, cells were washed with 4 × SSC, 2 × SSC, and 1 × SSC three times each in turn. Finally, DAPI was used to stain the nucleus and the signals were detected by a confocal microscopy.

To explore the roles of circ_0098181 in vivo, a rat fibrotic model was constructed. SD rats, weighing about 150–160 g, were intraperitoneally injected with CCl₄ (0.5 mL/kg, CCl₄ = 1:3) twice a week for total 6 weeks to induce liver fibrosis, while the controls were intraperitoneally injected with Oil. Given that adeno-associated virus-8 (AAV8), a hepatotropic virus, could specifically target hepatocytes, we selected adenovirus to overexpress circ_0098181, enabling effective infection of hepatic stellate cells (HSCs). An adenovirus carrying the rat circ_0098181 sequence (Ad-circ_0098181) and a control virus (Ad-NC) were constructed by Genechem Co. Ltd (Shanghai, China). Twenty SD rats were randomly assigned to four groups: Oil + Ad-NC, Oil + Ad-circ_0098181, CCl₄ + Ad-NC, and CCl₄ + Ad-circ_0098181. At the fourth week after CCl₄ or oil administration, Ad-circ_0098181 (1 × 10¹⁰ pfu) or Ad-NC (1 × 10¹⁰ pfu) was injected into the tail vein. The rats were sacrificed 2 weeks after adenovirus delivery.

Serum was collected for ALT and AST measurements, and liver tissues were harvested for further analysis, including HE, Sirius Red, and immunohistochemistry (IHC) staining for Col1a1 and α-SMA. The remaining liver tissues were used for RNA and protein extraction for additional validation.

Liver paraffin sections (4 μm) were deparaffinized and hydrated to remove the paraffin and allow staining reagents to penetrate. HE, Sirius Red, and immunohistochemistry staining were performed as previously described (Huang et al., 2023). Images of HE, Sirius Red, and immunohistochemistry staining were captured using an Olympus BX43F microscope.

To investigate the downstream genes and pathways of circ_0098181, HSC-T6 cells treated with either circ_0098181 (n = 3) or NC (n = 3) plasmids were used to perform RNA sequencing by Shanghai Biotechnology Corporation (China). Briefly, total RNA was extracted using Trizol reagent (Invitrogen). RNA concentration and purity were measured using a NanoDrop 2000 system (Thermo Fisher Scientific, Wilmington, DE, USA). Sequencing was performed on NovaSeq6000 (Illumina) using PE150 mode. All the downstream analyses were conducted on high-quality clean data. HISAT2 tools software (version:2.0.4) was used to map to the reference genome. Finally, differentially expressed genes (DEGs) were identified using the edgeR R package. Multiple testing correction was applied using the Benjamini–Hochberg method to control the false discovery rate. Genes with an adjusted p value < 0.05 and an absolute value of log2 (Fold change) > 1 were considered to be differentially expressed. The ClusterProfiler R package was used to identify significantly enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways compared to the entire genome background. Gene Set Enrichment Analysis (GSEA) was performed using GSEA software 3.0. Additionally, protein-protein interaction (PPI) network analysis was performed to highlight relevant and potential pathways.

SiRNAs targeting ADAR1 and QKI within rat genomes were synthesized by General Biology (Anhui, China), with the sequences provided in Supplementary Table S4. Based on the silencing efficiency of these siRNAs, shRNA targeting ADAR1 (shRNA-ADAR1) was designed. The plasmids containing the rat circ_0098181 sequence along with its flanking regions were supplied by Shanghai Calm Biotechnology Co., Ltd., with detailed information on the flanking sequences available in Supplementary Table S5.

To determine the binding region of PKM2, the PKM2 gene was divided into three equal segments. Plasmids containing these three segments, along with the full-length PKM2 (details in Supplementary Table S6), were constructed and transfected into HSC-T6 cells for RNA immunoprecipitation (RIP) assays. Similarly, to identify the binding sites of circ_0098181, the sequence of circ_0098181 was divided into three segments, and primers were designed based on each segment for RIP experiments. The primers for these segments are listed in Supplementary Table S2.

For RNA pull-down, the circ_0098181 positive probe (5′-biotin CTG​GTC​GCT​TGG​AAG​ACA​TCC​TTC​CGG​CTC​GTT​CTT​GAT​C-3′) and negative probe (5′-biotin GAT​CAA​GAA​CGA​GCC​GGA​AGG​ATG​TCT​TCC​AAG​CGA​CCA​G-3′) were provided by Cloud Biotech (Shanghai, China). After proteins were pulled down with the RNA probes in HSC-T6 cells, the eluted proteins were analyzed with mass spectrum. Easy nLC 1000 system (ThermoFisher Scientific) was used for liquid chromatography separation of peptide segments, and mass spectrometry analysis was performed with a Q-Exactive master spectrometer (ThermoFisher Scientific). RIP experiments were taken in HSC-T6 cells (about 1 × 107) with RNA Immunoprecipitation Kit (Geneseed Biotech, Guangzhou, China). After digestion, the cells were dissociated with lysis buffer. The supernatant was hybridized with magnetic beads that had been bound to PKM2 or rabbit IgG antibodies. Total RNA which bound proteins were eluted for further qRT-PCR.

Continuous parameters were presented as the mean ± standard deviation. Categorical variables were expressed as numbers and percentages or frequencies. For the statistical analysis, one-way analysis of variance (ANOVA) was determined for the analysis of multiple groups involving a single variable. Differences in categorical variables between groups were assessed using the Chi-square test. All statistical analyses were performed using R software version 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria).

Experimental data were collected and analyzed by SPSS V23.0 or Prism8 software (GraphPad Software, La Jolla, CA). The transwell, IHC, and Sirus red staining were analyzed by ImageJ software. Data were presented as the mean ± standard deviation. For datasets that conformed to a normal distribution, a Student’s t-test was utilized to compare two datasets. In cases where the data did not follow a normal distribution, the Mann-Whitney U test was applied. When analyzing multiple groups with a single variable, one-way ANOVA was employed. For comparing more than two groups with multiple variables, two-way ANOVA was used. p-value <0.05 was reckoned statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001).

The specific circRNAs profiles were screened by RNA sequencing in five paired human cirrhotic liver tissues and normal samples. Based on fold-change, there were 26 upregulated and 41 downregulated circRNAs in cirrhotic tissues, including circ_0000131, circ_0098181, circ_0036353, circ_0008520, circ_0004777, and so on (Figure 1A). Among the identified circRNAs, circ_0098181 exhibited a significant reduction in cirrhotic samples compared with that in normal tissues (p < 0.05). Further analysis demonstrated that circ_0098181 levels were highest in normal tissues, decreased in cirrhotic tissues, and were lowest in HCC tissues (Figure 1B), indicating a strong correlation between declining circ_0098181 expression and the progression of liver damage. Consistent with the data in tissues, our observations also revealed that circ_0098181 was most reduced in TGF-β1 treated LX2 cells compared with circ_0008520, circ_0004777, and circ_0001573, as well as in auto-activated primary rat HSCs, accompanied by increased expression of Col1a1 and α-SMA. (the marker of activation of HSCs) (Figures 1C–E; Supplementary Figure S1A). Furthermore, the FISH probe illustrated the enrichment of circ_0098181 in the cytoplasm of primary rat HSCs and activated LX2 cells (Figure 1F; Supplementary Figure S1B).

Given the high homology (91.93%) between the human and rat sequences of circ_0098181, plasmids carrying has_circ_0098181 were transfected into primary rat HSCs and activated LX2 to further explore the biological function of circ_0098181 (Figures 2A, B). qRT-PCR and Western blotting demonstrated that circ_0098181 remarkably suppressed the mRNA and protein levels of Col1a1 and α-SMA (Figures 2A–D). CCK-8 assay verified the suppression of circ_0098181 on HSCs proliferation in both primary HSCs and LX2 (Figures 2E, F). Furthermore, transwell experiments documented that circ_0098181 obviously inhibited the migration of HSCs (Figures 2G–J). Additionally, circ_0098181 also reduced the expression of Col1a1 and α-SMA, and suppressed the proliferation and migration of HSC-T6 (Supplementary Figures S2A-D). All these findings highlight the evident inhibitory effect of circ_0098181 on HSCs activation and proliferation in vitro, suggesting its potential as a therapeutic target for liver fibrosis.

Adenovirus carrying circ_0098181 was injected into CCl₄-induced rat fibrosis models to explore the effect of circ_0098181 on fibrogenesis. In the sixth week after intraperitoneal injection of CCl₄, the livers were enlargement, and the surface became rough and granular, indicating the significant manifestations of liver fibrosis induced by CCl₄. In contrast, rats treated with Ad-circ_0098181 (CCl₄+Ad-circ_0098181 group) displayed more normal liver morphology, with a smoother surface, compared to those treated with the control virus (CCl₄+Ad-NC group) (Figure 3A). In addition, the liver-to-body weight ratio and serum ALT levels were significantly lower in the CCl₄+Ad-circ_0098181 group compared to the model group (CCl₄+Ad-NC) (both p < 0.05) (Figures 3B–D). However, there were no significant differences in serum AST levels between the two groups. Exogenous administration of circ_0098181 notably reduced the severity of fibrosis and collagen deposition, as evidenced by HE and Sirius Red staining (Figures 3E, F). Additionally, immunofluorescence staining, qRT-PCR, and Western blotting revealed decreased expression of Col1a1 and α-SMA in the circ_0098181-treated rats (CCl₄+Ad-circ_0098181 group) compared to the CCl₄+Ad-NC group (Figures 3G, H). These results demonstrate the strong anti-fibrotic effects of circ_0098181 in vivo.

Previous studies have documented that circRNAs are produced from the pre-mRNAs of gene exons by back-splicing. According to the data from circBase (http://www.circbase.org/), circ_0098181 is derived from exons 2 and 3 of the SOX5 gene. Hereby, we elucidated the synthesis regulation of circ_0098181 in liver fibrogenesis. RNA binding protein-driven cyclization was one of the important approaches to generate circRNAs. ADAR1 and QKI, two major RNA binding proteins involved in circRNA cyclization, could competitively bind to ICSs in the flanking intron of circRNAs and regulate the biogenesis of circRNAs (Ivanov et al., 2015; Xi et al., 2022). To identify the key RNA-binding protein involved in circ_0098181 cyclization, we synthesized siRNAs targeting ADAR1 and QKI, and transfected them into HSC-T6 cells. The expression of ADAR1 and QKI was successfully knocked down by their respective siRNAs (Figure 4A). However, an increase in circ_0098181 levels was only observed following si-ADAR1 transfection. Additionally, sequence alignment of intron 1 and intron 3 of the SOX5 gene revealed highly reverse complementary sequences (approximately 80%) in the flanking introns of circ_0098181(Supplementary Figures S3A, B), rather than paired Alu sequences (Zhang et al., 2014). Thus, we hypothesized that ADAR1 might bind to the ICSs in the flanking regions, thereby regulating the synthesis of circ_0098181. We termed the complementary sequences in intron 1 and intron 3 of the SOX5 gene as ICS1 (reverse complementary sequences in intron 1) and ICS3 (reverse complementary sequences in intron 3), respectively. Primers for ICS1 and ICS3 were then designed to conduct RIP experiments. The results demonstrated that ADAR1 could bind to both ICS1 and ICS3 (Figure 4B). Furthermore, plasmids containing the full circ_0098181 sequence, along with ICS1 and ICS3 (P1), and plasmids with missing ICS1 or ICS3 (P2-P4), were constructed and transfected into HSC-T6 cells (Figure 4C; Supplementary Table S5). As shown in Figure 4D, circ_0098181 was significantly upregulated in HSC-T6 cells transfected with the P1 plasmid, while plasmids P2-P4 had no effect on circ_0098181 levels. Moreover, to confirm the ADAR1-dependent cyclization, a plasmid overexpressing ADAR1 was introduced into HSC-T6 cells transfected with shRNA-ADAR1. As expected, the inhibition of ADAR1 prevented the reduction of circ_0098181 induced by ADAR1 overexpression (Figure 4E). These findings suggest that ADAR1 suppresses the biogenesis of circ_0098181 by disrupting the intronic ISC pairs.

We identified differentially expressed genes between HSC-T6 cells treated with circ_0098181 plasmids (n = 3) and control plasmids (n = 3). Among these differential genes, 255 genes showed a downward trend, while 116 genes were significantly upregulated compared to the NC group (Figures 5A, B). Heatmap and PPI analysis revealed a marked reduction in several pro-inflammatory cytokines, including TNF-α, Fas, Bcl3, and Cxcl11, following circ_0098181 overexpression in HSC-T6 cells (Figure 5C). The decreased levels of TNF-α, Fas, Bcl3, and Cxcl11 were further confirmed by qRT-PCR (Figure 5G). GSEA enrichment analysis revealed that circ_0098181-regulated genes were primarily enriched in pathways related to inflammatory response and leukocyte activation (Figures 5D, E). KEGG enrichment analysis was also conducted, and results highlighted the involvement of inflammation-related genes after circ_0098181 treatment (Figure 5F; Supplementary Figure S4). Additionally, MPO staining was performed to assess neutrophil infiltration across the four groups. Compared to the CCl₄+Ad-NC group, MPO level of CCl₄+Ad-circ_0098181 group was significantly reduced, indicating decreased neutrophil infiltration (Figure 5H).

In our previous study, hsa_circ_0098181 sponged miR-18a-3p to regulate PPARA in HCC (Luo et al., 2022), but this axis was ineffective in HSCs (Supplementary Figure S5A). To investigate alternative mechanisms of circ_0098181 in fibrosis, RNA pull-down and mass spectrometry were conducted in HSC-T6 cells to identify the direct binding proteins of this circRNA. The potential interacting proteins are presented in the heatmap (Figure 6A). Among the identified target proteins, we focused on PKM2, a key regulator of inflammation and fibrosis (Rao et al., 2022), and YBX1, a well-known RNA-binding protein (Xu J. et al., 2020). Further RIP experiments vindicated that circ_0098181 was able to interact with PKM2, but not with YBX1 (Figure 6B, Supplementary Figure S5B). Given that the interaction between circRNAs and proteins is location-dependent, co-localization of PKM2 and circ_0098181 was assessed using fluorescence staining. As expected, PKM2 and circ_0098181 showed strong co-localization in HSC-T6, LX2, and primary rat HSCs (Figure 6C, Supplementary Figure S6A). To further identify the specific binding regions of circ_0098181, the PKM2 sequence was divided into three equal segments, and RIP assays located the binding site within the second segment of PKM2 (Figure 6D). Additionally, the 443 bp sequence of circ_0098181 was split into three fragments, with RIP assays indicating that PKM2 binds to the third segment of circ_0098181 (Figure 6E). Given that the biological function of PKM2 depends on its localization and form (Wong et al., 2015), we examined PKM2 protein levels in the whole cell, nucleus, and cytoplasm after circ_0098181 overexpression in HSC-T6 cells. The results showed a significant reduction of PKM2 in the nucleus of circ_0098181-treated HSC-T6 cells, with no notable changes in overall cellular or cytoplasmic PKM2 levels (Figure 6F). Additionally, the level of phosphorylated PKM2 (p-PKM2, Tyr105) was significantly decreased following circ_0098181 treatment (Figure 6G).