RNA sequencing data were derived from the The Cancer Genome Atlas (TCGA) repository for 374 patients with HCC and 50 normal liver samples with corresponding clinicopathological features on 4 August 2021.

A selection of 53 genes associated with pyroptosis were analysed from the available reviews (26,27). The 'limma' data package is available for screening DEGs. To construct a protein-protein interaction (PPI) network, the Search Tool for the Retrieval of Interacting Genes website (https://string-db.org/) was used. The R (v4.1.3; Robert Gentleman and Ross Ihaka) package 'glmnet' (2.0.16) can be used to narrow down the range of genes to be screened. After central normalization of TCGA dataset ('scale' function in R is applied), risk scores were calculated. The risk score formula is as follows: Risk score=∑8iXixYi (X: coefficient, Y: gene expression level). Kaplan-Meier analysis was used to analyze the overall survival (OS) times of patients with hepatocellular carcinoma obtained from the TCGA repository based on the log-rank test with a cut-off value of P<0.05. The R package 'prcomp' was used to conduct principal component analysis. Univariate and multivariate Cox regression models were performed to analyze the clinical characteristics of cases in the TCGA cohort.

Survival curves of differentially expressed CHMP3 were analyzed by GEPIA (28) to find out whether this gene's expression affects the survival of HCC patients. In addition, the staging plot was analyzed to compare CHMP3 expression in different pathological stages. The expression of CHMP3 was further compared by means of HPA between normal liver and HCC tissues with immunohistochemical (IHC) images.

Between September 2020 and March 2022, all clinical samples were collected after surgical resection and stored immediately at −80°C for subsequent experiments at the Department of General Surgery of Shengjing Hospital of China Medical University. All specimens were acquired from patients with HCC confirmed by pathologists. Written informed consent was obtained by all patients whose tissue samples were collected prior to enrolment. Human studies (approval no. 2022PS785K) were approved by the Ethics Committee of Shengjing Hospital of China Medical University (Shenyang, China). Clinicopathological characteristics of patients (including age and sex distribution) are listed in Table I.

The collected tissue was fixed in an appropriate volume of 4% paraformaldehyde at 4°C for 24 h, then procedurally dehydrated and embedded in paraffin. Tissue sections of 3-μm thickness were produced using a microtome, followed by de-paraffinization using xylene and microwave heating. Sections were placed in the configured antigen repair solution (Citrate antigen retrieval solution; Beyotime Institute of Biotechnology) and boiled for 5 min. Primary antibody against CHMP3 (1:100; cat. no. 15472-1-AP; Proteintech Group, Inc.) was added to the sections after washing with PBS and incubated at 4°C overnight. The appropriate proportion of HRP-conjugated secondary antibody (ready to use; cat. no. PR30011; Proteintech Group, Inc.) was added, and incubated at 37°C for 1 h. Antibodies were diluted with TBST, and Tween-20 was used in a dosage of 0.5 ml/l. Finally, chromogenic detection was carried out under the light microscope using DAB. The final score is the percentage of cells multiplied by the staining intensity (29).

The human liver cancer cell lines HepG2 (Procell Life Science & Technology Co., Ltd.) and Huh-7 (Nanjing KeyGen Biotech Co., Ltd.) cell lines were purchased in August 2021. Cells were expanded and stored at -80°C and only early passages (<passage 5) within 6 months of these cell lines were used in the present study. HepG2 cells were cultured in MEM (Hyclone; Cytiva). Huh-7 cells were cultured in DMEM (Hyclone; Cytiva). Ac-YVAD-CMK (cat. no. GC42721) was purchased form GLPBIO.

HepG2 and Huh-7 were tested for genotyping of the STR locus and Amelogenin locus using the 20-STR amplification protocol. The results showed that no cross-contamination of human cells was detected in the cell lines and a 100% match to their cytotyping could be found in the cell bank, named HepG2 and Huh-7, respectively.

Negative control siRNA (siRNA-NC) (UUG TCC GAA CGU CTC AAG UTT), small interfering (siRNA)-CHMP3 (UGU GAA GAU UCC AGA GAU UTT), the plasmid vector and plasmid-CHMP3 were purchased from Shanghai GenePharma Co., Ltd. Transfection was carried out at room temperature with Lipofectamine 3000 (GlpBio) following the manufacturer's instructions. Cells were transfected with 2.5 μg RNA/DNA added to 3.75 μl of reagent, and serum-free medium was replaced with complete medium 6 h later. Transfection efficiency was verified by western blotting at 24 h post-transfection. ImageJ (v1.52) (National Institutes of Health) was used to analyze the banding results.

Groups of treated cells were seeded in six-well plates at a density of 1,000 cells/well and cultured in serum-containing medium. The medium was changed every three days and culture was stopped when colony formation was observed with the naked eye. Ac-YVAN-CMK (40 μM, AYC) (GLPBIO) was added to the cells after transfection. Lastly, colonies were fixed with 4% paraformaldehyde for 2 h at room temperature as well as stained with crystal violet solution for 30 min (Beyotime Institute of Biotechnology). More than 50 cells were counted as one colony and then the total number of colonies per well was counted.

Groups of treated cells were scratched with a 200-μl pipette tip. The plates were washed with PBS wells to remove residual cells, then serum-free medium was added. Images of the wound shape were captured with a light microscope at a specific point in time (magnification, ×200). Wound healing rate=(0 h wound area-48 h wound area)/0 h wound area ×100%.

The Matrigel (Corning, Inc.) was removed from −20°C, placed at 4°C to thaw and melt into liquid form, and then the pre-cooled tip was used to spread the Matrigel evenly on the bottom of upper chamber, followed by placing it in a 37°C incubator for 2-4 h to wait for the Matrigel to completely solidify. The lower chamber of the Transwell plate was added with 700 μl of serum-containing medium, and the upper chamber was filled with groups of cells resuspended in 200-μl of serum-free medium. The upper chamber was placed in the Transwell plate and incubated for 24 h at 37°C. Cells from the upper chamber outer membrane were then fixed in 4% paraformaldehyde for 1 h and stained with crystal violet for 30 min at room temperature. Matrigel was wiped from the inner membrane of the upper chamber with a cotton sab and the upper chamber was gently rinsed with PBS to wash away excess color. The invasive cells were observed and images were captured with a light microscope.

The cell proliferation capacity was assayed according to the manufacturer's instructions of the CCK-8 assay kit (Epizyme; http://www.epizyme.cn/). HepG2 and Huh-7 (3×10³ cells/well) were inoculated into 96-well plates. Cells were transfected with si-CHMP3 and incubated at 37°C for 24 h before adding AYC (40 μM). A total of 10 μl CCK-8 was then added at each point (24, 48, 72 and 96 h) and incubated at 37°C for 4 h. The absorbance of the cells at 450 nm was measured using a microplate reader.

Tissue and cell samples were lysed in ice-cold RIPA lysis buffer and PMSF (both from Beyotime, Institute of Biotechnology). The protein concentration was calculated according to the BCA assay kit instructions (Epizyme). Gel electrophoresis separation was performed by 10%- or 15%-PAGE Gel Rapid Preparation Kit (Epizyme) and then transferred to polyvinylidene difluoride membranes (Epizyme). Each lane contained 20 μg of protein. The bands were blocked with 5% skim milk blocking solution prepared in TBST for 2 h at room temperature. After blocking was completed, incubation was carried out overnight at 4°C using the primary antibodies. The membranes were subsequently incubated for 2 h at room temperature using the HRP-conjugated Affinipure Goat Anti-Mouse/Rabbit secondary antibody (cat. nos. SA00001-1 and SA00001-2, Proteintech Group, Inc.). Protein bands were visualized using an ECL kit (Epizyme). The information of primary antibodies is provided in Table SI.

Huh-7 cells that had been treated differently were collected. The cell pellets were then pre-fixed in 2.5% glutaraldehyde overnight at 4°C. They were fixed in 1.0% osmic acid for 2 h. The samples were dehydrated and soaked with a 1:1 mixture of epoxy resin and acetone overnight at room temperature. The samples were incubated in 1% uranyl acetate staining solution for 1 h. After 4 days, the samples were sectioned with an ultrathin sectioning machine. Finally, staining was carried out with uranyl acetate and lead citrate. Examination was carried out using TEM.

In vivo tumor formation was investigated by constructing a xenograft nude mouse model. A total of 18 male BALB/c nude mice (4 weeks old, weighing ~20 g) were used in the animal experiments and were purchased from Beijing HFK Bioscience co. Ltd. All mice were housed under specific pathogen-free conditions (temperature 22°C; humidity 50%; light/dark cycle 12/12 h), with free access to food and water, and padding changed twice every three days. Huh-7 cells at a density of 1×10⁶ were resuspended in 100 μl PBS to form a cell suspension. Each mouse received a gentle, slow subcutaneous inoculation of 100 μl of cell suspension in the right axilla. When the tumors grew to an optimal volume, the tumor-bearing mice were randomly divided into three groups. siRNA-CHMP3 in vivo (10 μg/μl) was incubated with Lipofectamine 3000 for 30 min at room temperature to form siRNA-lipofectamine complexes, which were injected intratumorally every 3 days for 4 weeks. AYC (0.1 mg/kg/day) was injected intraperitoneally every 24 h. Tumor volume (mm³) was calculated using the following formula: Volume=(width)² × length/2. Tumor size was measured every three days. Animal experiments (approval no. 2022PS779K) were approved by the Ethics Committee of Shengjing Hospital, China Medical University (Shenyang, China).

The experimental values are shown as the mean ± SEM of at least three independent experiments. One-way analysis of variance (ANOVA) followed by Tukey's multiple comparison test was used to analyze the data with GraphPad Prism (version 9.0.0; Dotmatics). The chi-square test was used to statistically analyze the relationship between different clinical characteristics and CHMP3 expression. Differences between groups were considered statistically significant when P<0.05. All experiments were repeated three times.

The workflow was summarized in a flowchart (Fig. S1). 39 DEGs (all P<0.05) were recognized through comparing 53 pyroptosis-associated genes' expression in TCGA downloadable data from patients with HCC. From these, 33 genes were upregulated in tumor tissues, while the six other genes exhibited low expression (Fig. S2A). PPI network and an association network containing 20 pyroptosis-associated genes are demonstrated in Fig. S2B and C (red: positive association). When the clustering variable k=2, the intra-group relevance is higher and the inter-group relevance is lower (Fig. S2D). The heat map showed little difference in clinical characteristics between the two clusters except for the degree of tumor differentiation (P<0.001) and OS (P<0.05) (Fig. S2E and F).

Univariate Cox regression was used to find 8 genes that met the criteria of P<0.5 and hazard ratios (HRs) >1 for the next analysis (Fig. S3A). A seven-gene signature was constructed on the basis of the optimal λ values derived from least absolute shrinkage and selection operator Cox regression analysis (Fig. S3B and C). Risk scores were calculated and 424 patients with HCC were divided into low and high-risk groups based on median scores (Fig. S3D). As demonstrated in Fig. S3E-G, it could be concluded that patients in the divided high-risk group clearly had higher mortality and shorter survival compared with the low-risk group. The area under the receiver operating characteristic curve varied from 0.7-0.85, indicating that the prediction model was effective (Fig. S3H).

Univariate (HR=7.222, 95% Confidence Interval (CI): 4.322-12.066; Fig. S4A) and multifactorial (HR=6.315, 95% CI: 3.638-10.964; Fig. S4B) Cox regression analyses indicated that risk score could be used as an independent predictor of poor survival. Finally, a heat map of clinical features (Fig. S4C) revealed a distinct patient distribution by sex and tumor differentiation in different subgroups (P<0.01). The relationship between CHMP3 and HCC among these seven genes (BAK1, BAX, CHMP3, GSDME, CASP8, GSDMC and SCAF11) remains unstudied. Therefore, the role of CHMP3 in the progression of liver cancer was investigated in the present study.

As demonstrated in Fig. 1A, CHMP3 was highly expressed in LIHC. The curves of OS and disease-free survival indicated that patients with high expression of CHMP3 presented lower survival rate (Fig. 1B and C). It was also identified that CHMP3 expression varies across pathological stages (Pr (>F) i.e., P<0.05) (Fig. 1D). IHC results obtained from HPA revealed that the expression of CHMP3 was markedly higher in HCC compared with normal tissue (Fig. 1E and F). The database results illustrated two indications: i) CHMP3 participates in the advancement of HCC and, ii) it might be related to pyroptosis.

To further validate the results of the public database, the expression of CHMP3 was determined by IHC based on previous predictions. It was found that CHMP3 was apparently overexpressed in the HCC samples compared with the corresponding para-cancer tissues (Fig. 2A-D). The high/low expression is based on tissue type. Another five pairs of carcinoma and para-cancerous tissues were collected to verify the CHMP3 expression and pyroptosis-related proteins by western blot assay, for detecting the level of pyroptosis in HCC. A previous study demonstrated reduced caspase-1 activation and IL-1β expression for some cancers (12). It was then observed that caspase-1, IL-1, IL-18 and GSDMD were significantly downregulated, suggesting the same trend (Fig. 2E and F). Analysis of the clinical characteristics of 65 patients with HCC revealed a significant association between CHMP3 expression and tumor size using the chi-square test (P<0.05).

According to the aforementioned bioinformatics analysis and the results of IHC and western blotting, CHMP3 was highly expressed in liver cancer. Therefore, CHMP3 was knocked down and overexpressed to observe the effect on the proliferative capacity of liver cancer. The results of western blot analysis showed a significant decrease in CHMP3 expression after transfection with si-CHMP3 in HepG2 and Huh-7 cells (Fig. 3A and B). Colony formation assays revealed that knocking down CHMP3 reduced the number of colonies in liver cancer cell lines, which suggests that suppression of CHMP3 impaired the proliferation ability in liver cancer (Fig. 3C-E). On the contrary, cells transfected with plasmid CHMP3 had significantly enhanced colony formation capacity compared with cells transfected with vector (Fig. 3F-J). The transfection efficiency of CHMP3 overexpression was detected by western blotting.

To explore whether CHMP3 leads to pyroptosis in Huh-7 cells, cellular alterations were observed after knockdown using transmission electron microscopy. It was observed that the transfected cells exhibited typical membrane rupture and cytoplasmic leakage (red arrows) as demonstrated in Fig. 3K. The expression of several proteins was apparently increased after CHMP3 knockdown, including cleaved caspase-1, IL-1β, IL-18 and N-terminal GSDMD (Fig. 3L and M). However, no changes were observed regarding their corresponding precursor forms of expression. Thus, it can be assumed that CHMP3 might promote liver cancer progression through pyroptosis mediated via the caspase-1/IL-1β pathway.

To explore whether CHMP3 inhibition-induced pyroptosis is modulated by caspase-1, Ac-YVAD-CMK (AYC, an inhibitor of caspase-1) was used in the present study. The changes in proliferative capacity of liver cancer was examined after the administration of AYC by colony formation and CCK-8 assays. Knocking down CHMP3 with si-CHMP3 significantly decreased the proliferation of HepG2 and Huh-7 cells, while the application of AYC reversed the ability of si-CHMP3 to reduce cell proliferation (Fig. 4A-F). By carefully examining the data, it was identified that AYC diminished CHMP3 knockdown-induced caspase-1 activation and reduced IL-1β, IL-18 and GSDMD cleavage in liver cancer cells (Fig. 4G and H). Briefly, these data supported the notion that CHMP3 inhibits pyroptosis by caspase-1, and this effect can be reversed by caspase-1 inhibitor.

The effect of CHMP3 in tumor formation was then examined in vivo. Knockdown of CHMP3 significantly inhibited the growth of subcutaneous tumor, and the volume and weight of the tumor were lower than those of the control group. The addition of AYC reversed this inhibition, indicating that low-expression CHMP3 also inhibited the growth of HCC in vivo (Fig. 5A-D).

There is a well-known fact that HCC is a highly aggressive and metastatic cancer. In the wound healing assay, the capacity of cell migration was impaired after CHMP3 silencing (Fig. 6A-C) and enhanced after overexpression (Fig. 7A-C). Transwell invasion assay revealed that si-CHMP3 reduced the number of invasive cells (Fig. 6D-F), while CHMP3 overexpression in the plasmid-CHMP3 group increased the number of HepG2 and Huh-7 cells crossing Matrigel (Fig. 7D-F). To provide further evidence of changes in invasion and migration capacity, expression of EMT-related proteins was next investigated in both cells by western blotting. Knocking down CHMP3 caused an significant decrease of N-cadherin, matrix metalloproteinase 9 (MMP9) and vimentin expression, and an increase of E-cadherin expression in liver cancer cells (Fig. 6G and H).