ARID1A deficiency promotes progression and potentiates therapeutic antitumour immunity in hepatitis B virus-related hepatocellular carcinoma

Three datasets were explored in this study. In the CHCC-HBV cohort, patients were selected as hepatitis B core antibody (HBcAb) and hepatitis B surface antigen (HBsAg) positive. A total of 150 patients were enrolled in the analysis, among which 15 patients had ARID1A deficiency, as they carried loss-of-function mutations or copy number deletions in ARID1A. The expression data were downloaded from https://​www.​biosion.​org [7]. The mutation data were obtained from the supplementary information of the CHCC-HBV study [23]. For the TCGA-LIHC cohort, the infection status of HBV was obtained from two studies, which contained 44 and 87 HBV patients [12, 23]. A total of 108 HBV-infected patients were enrolled from the TCGA-LIHC cohort after removing duplicates and patients with mutations or missing clinical information. For the AMC cohort, only 167 patients identified to have HBV infection were included in this study [24]. All the original data can be downloaded at cBioPortal (https://​www.​cbioportal.​org). Overall, 425 HBV-related hepatocellular carcinoma patients were enrolled in the study. The clinical information of patients included in this study can be found in Supplementary file S1.

Tumour mutation burden was calculated as the number of nonsynonymous mutations per targeted sequencing length. The exon length from the TCGA cohort was estimated as 38 Mb [25], while the CHCC and AMC cohorts used the Agilent SureSelect 50 Mb system to capture the exon area [7, 24].

The immune infiltration level was estimated by single-sample GSEA using the Gene Set Variation Analysis (GSVA) program against the gene signatures representing immune cell populations (Supplementary file S2). In addition, the immune infiltration status of the tumour purity, immune components, and overall stromal status in the CHCC-HBV cohort was computed using ESTIMATE.

The human hepatocellular carcinoma cell lines Hep3B and SK-HEP-1 were purchased from ATCC. Cells were grown in Dulbecco’s modified Eagle medium (Gibco™; Thermo Fisher Scientific, MA, USA) supplemented with 10% foetal bovine serum (Gibco™; Thermo Fisher Scientific) and 1% penicillin‒streptomycin (Gibco™; Thermo Fisher Scientific) and maintained in a humidified incubator at 37 °C with 5% CO2. Hep3B and SK-HEP-1 cells stably expressing Cas9 were established by lentiviral transduction with Cas9 plasmids (Lenti-Cas9-2A-Blast, 73310; Addgene).

ARID1A knockout lentiviral plasmids were synthesized and purchased from GenePharma (Beijing, China). Plasmid psPAX2 (12260; Addgene), pmd2G (12259; Addgene), and ARID1A Knockout lentiviral plasmids were cotransfected into 293 T cells by PEI (Proteintech, #PR40001). First, Hep3B and SK-HEP-1 cells were infected with lentivirus containing polybrene (10 μg/ml) (TR-1003; Sigma‒Aldrich) for 24 h. Then, positive cells were selected with 100 μg/ml hygromycin B. The efficiency of ARID1A knockout was confirmed by Western blotting.

The sgRNA sequences were as follows:

The siRNA duplexes and negative control were synthesized and purified by GenePharma. (Beijing, China). Briefly, HepG2/2.2.15 cells were transfected with negative control siRNA using Lipofectamine™ 3000 (#L3000001; Invitrogen) according to the manufacturer’s instructions. Knockdown efficiency was tested by Western blot analysis 48 h after transfection. siRNA sequences used for ARID1A knockdown can be found in Supplementary file S3.

Lysis buffer (R0010–100 ml, Solarbio) was used to extract proteins from Hep3B and SK-HEP-1 cells. Cells were washed twice with ice-cold PBS, and then protein lysates were separated by 12.5% SDS‒PAGE at 120 V for 70 min (Omni-Easy™ One-Step PAGE Gel Fast Preparation Kit, PG213; EpiZyme, Shanghai, China) and transferred to PVDF membranes (IPVH00010; Merck Millipore) with transfer buffer plus 10% methanol on ice at 300 mA for 2.5 h. (Transfer Buffer (10x), PS109; EpiZyme, Shanghai, China). Subsequently, the membranes were blocked in 5% skim milk for 1 h and incubated overnight with antibodies against ARID1A (rabbit monoclonal antibody, 1:1000, HPA005456; Sigma‒Aldrich) followed by secondary antibodies (goat anti-rabbit IgG H&L (HRP), ab6721; goat anti-mouse IgG H&L (HRP), ab6789; Abcam). β-Actin (mouse monoclonal antibody, 1:2000; ab8226; Abcam) was used as the loading control. Images were visualized using chemiluminescent horseradish peroxidase substrate (WBKLS0100; Merck Millipore) on Bio-Rad Gel Doc 2000 system analysis software (Bio-Rad Laboratories, Inc., Hercules, CA).

Total RNA was isolated from harvested cells using TRIzol reagent (Invitrogen, Thermo Fisher Scientific). cDNA synthesis was performed with 1 μg of total RNA using PrimeScript RT Master Mix (RR036;ATakara). In addition, quantitative reverse transcription-PCR (qRT‒PCR) was performed using TB Green® Premix Ex Taq™ II (RR820A; Takara) on a quantitative PCR machine (qTOWER3G, Analytik Jena, Jena, Germany). The primers used for qPCR are listed in Supplementary file S3.

The effects of ARID1A deficiency on HCC proliferation were assessed by using a Cell Counting Kit-8 (CCK-8) (HY-K0301; MedChemExpress) and clone formation assays. Hep3B, SK-HEP-1 wild-type cells, and knockout cells were collected and seeded in a 96-well plate (3,000 cells per well). In addition, 10 μL of CCK-8 solution was added to each well at 24 h, 48 h, 72 h, and 120 h. The optical density (OD) of each well was measured (EnSpire, PerkinElmer, CA, USA) at 450 nm 2 h after incubation. For clone formation, 2000 cells were seeded into a six-well plate and cultured for 2 weeks or more. The cells were then fixed with Paraformaldehyde Fix Solution for 30 min and stained with 0.1% crystal violet for 15 min. At least three replicate samples were analysed for all assays.

Transwell migration and invasion assays were performed using 8-μm-pore inserts (Corning Incorporated Costar, Tewksbury, USA). For the migration assay, cells (5 × 104 cells per well) were seeded into the upper chamber directly with serum-free medium. For the invasion assay, the inserts were coated with 50 µL of Matrigel (Solarbio, Beijing, China), and 600 µL culture medium containing 15% FBS was added to both lower chambers. Migrated or invaded cells were stained with 0.1% crystal violet 24 h after culture and counted in three random fields under light microscopy in a 100 × scope.

We retrospectively reviewed the data of 19 consecutive HCC patients admitted to the oncology department of Peking University International Hospital between 1 April and 30 September 2020 and performed next-generation sequencing of tumour tissues. Three patients had ARID1A mutations. Paraffin sections from HCC patients were baked at 60 °C for 2 h, deparaffinized, and hydrated with xylene and graded alcohol. Heat-based antigen retrieval was performed using citrate or EDTA-containing buffer. Hydrogen peroxide (3%) was used to block endogenous peroxidase in tissues, which were then incubated with primary antibodies against ARID1A (rabbit monoclonal antibody, 1:100, HPA005456; Sigma‒Aldrich), TIM-3 (rabbit monoclonal antibody, 1:200, ab241332; Abcam), CD8 (rabbit monoclonal antibody, 1:200, ZA-0508; Zhongshan Goldenbridge) and CD56 (rabbit monoclonal antibody, 1:200, ZM-0057; Zhongshan Goldenbridge) overnight. Furthermore, the slides were incubated with polyperoxidase-anti-mouse/rabbit IgG (PV-9000; Zhongshan Goldenbridge) for 30 min. Finally, the slides were stained with DAB and haematoxylin. For each section, three random fields were captured under light microscopy (Carl Zeiss, Primovert, NY) (100 × scope; 200 × scope) and analysed by ImageJ software.

Flow cytometric analysis was performed to determine the effects of ARID1A deficiency on galectin-9 expression. Cells were collected and washed twice with ice-cold PBS and then incubated with Alexa Fluor® 488 anti-human Galectin-9 Antibody (348918; Biolegend) for 15 min. Finally, Galectin-9 expression was detected using FCM (FACS Calibur flow cytometer, BD Biosciences) and analysed using FlowJo software.

We integrated the whole-exome sequencing data from three cohorts, including TCGA-LIHC, AMC, and CHCC-HBV. In total, 425 HBV-related hepatocellular carcinoma patients with a median age of 53 years (range 20–83) were included in the study (Supplementary file S1). Approximately 10% of patients carried mutations in ARID1A, and the most frequently mutated gene was TP53 (Fig. 1a). We primarily focused on ARID1A mutations; only samples with frameshift mutations, splice site mutations, nonsense mutations, or copy number (CN) deletions were exclusively considered for the ARID1A deficient group. In contrast, the rest of the samples were defined as the ARID1A normal functioning group. Eventually, a total of 29 (6.8%) samples were classified as the ARID1A deficient group.

To verify the accuracy of the methods, we first analysed the mRNA and protein levels of ARID1A in the CHCC-HBV cohort between the deficient and normal functioning groups. A total of 150 patients, with 15 (10%) from the ARID1A deficient group, were divided into two groups following the previously described standards. There was no significant difference in mRNA expression between the two groups (Fig. 1b). In contrast, the protein expression of ARID1A was significantly higher in the normal functioning group (Fig. 1c). The findings confirmed that frameshift mutation, splice site mutation, nonsense mutation, and copy number deletions resulted in ARID1A deficiency. In addition, the mRNA expression of ARID1A did not correlate with the protein expression, and according to the Human Protein Atlas database, low expression of ARID1A mRNA was associated with better survival because the low expression of mRNA does not necessarily represent ARID1A deficiency at the protein level. Therefore, we confirmed that frameshift mutations, splice site mutations, nonsense mutations, and copy number deletions led to decreased protein levels and ARID1A deficiency in HBV-related HCC. In addition, the clinical implication of ARID1A deficiency in HBV-related HCC was assessed using a Kaplan–Meier analysis of overall survival (OS) from the integrated dataset. The OS of the ARID1A deficient group was significantly worse than that of the normal functioning group (Fig. 1d). This observation concurs with previous studies [26], indicating that ARID1A may serve as a tumour suppressor gene. Further analysis revealed that the ARID1A deficient group was associated with cirrhosis (Fig. 1e), which is the main risk factor for HCC initiation.

To gain biological insight into pathways affected by ARID1A deficiency, we performed gene set enrichment analysis (GSEA). The cell cycle-related E2F pathway, consisting of genes necessary for progression through the S phase of the cell cycle, was highly enriched (Fig. 2a, b). Another pathway enriched in the ARID1A deficient group was the epithelial-mesenchymal transition (EMT) pathway, which is relevant to metastasis, invasion, and more aggressive performance (Fig. 2c). We further explored the gene markers related to cells in epithelial or mesenchymal states. Several mesenchymal markers had increased expression of matrix metalloproteinase 3 (MMP3), matrix metalloproteinase 9 (MMP9), and vimentin (VIM) in the ARID1A deficient group (Fig. 2d).

Upregulation of mesenchymal markers indicated that HCC cells with ARID1A deficiency were stimulated to undergo the EMT process, which eventually led to metastasis. The results of bioinformatics analysis were supported by in vitro experiments. We performed ARID1A CRISPR knockout in the human HCC cell line SK-HEP-1 and the Hep3B cell line to investigate the function of ARID1A in vitro. Western blotting confirmed the efficacy of CRISPR knockout (Fig. 2e). The decreased expression of ARID1A promoted proliferation (Fig. 2f, g), migration, and invasion in vitro (Fig. 2h).

To investigate the immunomodulatory role of ARID1A in HBV-HCC, we calculated the TMB, which was measured by the quantity of somatic mutations per Mb in the tumour genome. Then, the data from the three cohorts were normalized based on the length of the sequenced target region. Compared with the ARID1A normal function groups, the ARID1A deficient group exhibited a higher TMB level (Fig. 3a). This observation indicates that ARID1A deficiency was associated with the genomic instability of HBV-HCC. Moreover, the finding was consistent with our previous study on HCC and studies of other tumours.

Previous studies have shown that ARID1A deficiency might impair the DNA mismatch repair (MMR) pathway [19], implying the underlying mechanism of increased mutagenesis. Therefore, we investigated the mutational rate of the eight DNA damage repair (DDR) pathways. We found that ARID1A deficiency was more likely to display dysregulation in these pathways, out of which the base excision repair (BER) pathway exhibited a significantly higher mutational frequency in the ARID1A deficient group (Fig. 3b, c). Although not statistically significant, the same trend was observed in the MMR pathway. In addition, mutations in the MMR pathway occurred only in certain genes, including MSH6, MLH1, and PMS1 (Fig. 3d). Overall, an alternative mechanism for the high mutational rate in DDR pathways might impair the DNA repair function and result in genomic instability in HBV-HCC with ARID1A deficiency.

Gene set enrichment analysis revealed that ARID1A deficiency was positively correlated with active immune signatures, including the interleukin 6 (IL6) and IFN γ pathways (Fig. 4a). Notably, there was a significantly higher mRNA expression level of IL6 in the ARID1A-deficient group for the IL6 pathway. Based on the mRNA transcription data, we next estimated the activity of stromal and immune cells present in tumour tissues. The ARID1A-deficient group had higher ESTIMATE scores for stromal and immune cells (Fig. 4b). Further investigation of the immune cells infiltrating tumour tissues revealed multiple enriched immune signatures in the ARID1A-deficient group (Fig. 4c). CD8 + T cells, which are important for the immune surveillance of tumours, were also significantly enriched along with other types of cells, including neutrophils, M1 macrophages, and cytotoxic cells. We further confirmed elevated immune activity in ARID1A-deficient HCC. HCC specimens from 6 patients were stained for ARID1A, CD8, and CD56 and examined by immunohistochemistry (IHC). The patient-1, 2, and 3 had p.G1500W mutation, p.G108Efs*7 mutation, and p.S1985Y mutation in ARID1A, respectively, and can be classified as the ARID1A deficiency (Fig. 4d), wide type ARID1A (patient-4,5,6) in three HCC patients as control. IHC staining of these tumours additionally revealed that the loss of ARID1A significantly upregulated CD8 and CD56 infiltrating levels in these tumours (Fig. 4e, f). These findings were consistent with the bioinformatic analysis. Overall, ARID1A deficiency was associated with increased immune activity in tumour tissues.

ARID1A deficiency is associated with elevated PD-L1 expression in various cancers [27‐29]. Thus, we performed an analysis to explore the relationship between ARID1A deficiency and immune checkpoint protein expression. Our analysis revealed no significant difference in PD-L1 expression between the ARID1A deficiency and normal functioning groups. Nevertheless, the expression of TIM-3, which is another inhibitory immune checkpoint, was significantly higher in the deficient group (p = 0.004) (Fig. 5a). TIM-3, a coinhibitory receptor in the immune response, interacts with its ligand galectin-9. This interaction results in CD8-positive T-cell exhaustion and suppressed immune responses, especially in chronic viral infections [19]. Moreover, recent studies have shown that IFN-γ could stimulate galectin-9 expression in tumour and immune cells via EZH2 [30, 31]. Our bioinformatic analysis revealed an active IFN-γ signalling pathway in the ARID1A-deficient group. Therefore, we assessed whether the expression of galectin-9 could be upregulated in ARID1A knockout HCC cells. Surprisingly, increased galectin-9 expression was observed in the ARID1A knockout Hep-3B cell line and ARID1A knockdown HepG2/2.2.15 cell line (Fig. 5b-e), whereas the increase was not observed in the SK-HEP-1 cell line. This observation implied that the activity of the TIM-3/Galectin-9 pathway was upregulated only in HBV-HCC. Furthermore, IHC staining for TIM-3 and ARID1A in human HCC samples confirmed a strong negative correlation between the levels of these two proteins (Fig. 5f). Thus, the TIM-3-Galectin-9 interaction might play an essential role in immune suppression in HBV-HCC with ARID1A deficiency. Furthermore, this interaction might be an underlying mechanism for primary and secondary drug resistance to ICB treatment in HBV-HCC. In contrast, the combined inhibition of TIM-3 with current therapeutic approaches might achieve a superior response in HBV-HCC with ARID1A deficiency.