Arid1a deficiency promotes hepatocyte hyperpolyploidy and drives intrahepatic cholangiocarcinoma in mice

Abstract

Background and Aims:

Intrahepatic cholangiocarcinomas (ICCs) are aggressive liver tumors with high heterogeneity and limited therapeutic options. Although traditionally thought to arise from biliary cells, recent findings suggest that hepatocytes may also serve as a cellular origin for ICC. However, the mechanisms underlying hepatocyte malignant transformation and ICC initiation remain poorly understood.

Approach and Results:

We employed oncogene-driven and chemically induced ICC murine models, along with cellular models, to recapitulate the transformation of hepatocytes into ICC. Our findings demonstrate that mature hepatocytes undergo a significant hyperpolyploid state during ICC initiation. Hyperpolyploidy promotes aberrant cell division and chromosomal instability, accelerating hepatocyte transformation and ICC onset. Furthermore, we identified the chromatin remodeling factor Arid1a as a critical suppressor of hyperpolyploidy. Arid1a deficiency disrupts mitotic machinery at the centrosome, driving hyperpolyploidization and ICC tumorigenesis.

Conclusions:

Hepatocytes can transform into ICC through a process involving hyperpolyploidization. This study offers new insights into the pathogenesis of ICC, particularly in patients harboring frequent ARID1A mutations.

INTRODUCTION

Intrahepatic cholangiocarcinoma (ICC) is an aggressive malignancy of the liver, with an increasing global mortality rate.¹ However, the mechanisms underlying its pathogenesis and potential therapeutic targets remain limited. Traditionally, ICC was thought to arise from biliary epithelial cells. However, recent studies have shown that the development of ICC is closely associated with factors such as HBV and HCV infections, alcoholic liver injury, and fatty liver damage.² These factors primarily target hepatocytes, causing hepatocyte damage rather than bile duct cell damage, suggesting that a hepatocyte origin is more likely than a biliary epithelial origin in many ICC patients.^2,3 Bioinformatics analysis of gene mutation evolution also predicts that human ICC might originate from hepatocytes.⁴ Moreover, activation of the Notch signaling pathway has been identified as a critical driver of hepatocyte conversion into biliary lineage cells and ICC development in mouse models.^5–8 These findings indicate a potential hepatocyte origin of ICC in patients. However, the processes and the mechanisms underlying the malignant transformation of mature hepatocytes into ICC remain poorly understood.

In recent years, sequencing studies have identified hotspot mutations in cholangiocarcinoma, particularly in ICC, including BAP1 (25%), ARID1A (19%), PBRM1 (17%), and IDH1 (19%).^9,10 For instance, loss of BAP1 has been shown to induce malignant features in organoid models and enhance the proliferation of cholangiocarcinoma cell lines through activation of the ERK1/2 and JNK signaling pathways.^11,12 IDH1 mutations disrupt DNA methylation patterns, leading to the suppression of HNF4α transcriptional activity. This results in the dedifferentiation of hepatocytes into hepatoblast-like progenitor cells, driving the formation of ICC.¹³ These findings highlight the pivotal role of gene mutations in the initiation and progression of ICC.

ARID1A (AT-rich interactive domain-containing protein 1A) is a critical subunit of the SWI/SNF chromatin remodeling complex. Its primary function involves DNA binding, facilitating the SWI/SNF complex’s ability to alter nucleosome structure and regulate gene expression.^14,15 In cancer, ARID1A exhibits a high frequency of mutations across various tissue types, where it plays a significant role in tumorigenesis and cancer progression, largely mediated by its chromatin remodeling functions within the SWI/SNF complex.^16–21 In ICC, mutations in ARID1A occur in ~19% of patients. ARID1A knockout has been shown to promote proliferation and invasion in cholangiocarcinoma cell lines.²² Despite these insights, the precise roles of ARID1A in hepatocyte malignant transformation and ICC initiation, and the broader implications of its mutations in ICC pathogenesis remain unclear, hindering the development of effective therapies for ARID1A-mutated ICC.

Polyploidy—where chromosome numbers exceed diploidy—is common in organisms like yeast and plants, often enhancing resilience to environmental stress.^23,24 The liver is a major polyploid organ, with up to 90% of rodent hepatocytes and about 50% of human hepatocytes being polyploid.²⁵ These polyploid hepatocytes are characterized by one or multiple enlarged nuclei within the cell. Gene manipulation in mouse models, such as knocking out ANLN or CDK1 or overexpressing YAP, disrupts the division machinery, leading to incomplete mitosis and cytokinesis failure, which induces polyploidy.^26–28 In contrast, deletion of E2F7/8 prevents polyploidy.^29,30 Cellular polyploidy is also observed in various human malignancies, including pancreatic,³¹ renal,³² and hepatic cancers,³³ and other tumor types,³⁴ implicating its potential association with cancer development. The role of polyploidy in tumorigenesis remains elusive. Research on HCC has shown that hepatocyte polyploidy inhibits tumorigenesis by increasing the copy number of tumor suppressor genes, thereby buffering genotoxic damage in the liver.³⁵ Conversely, polyploid cells can undergo ploidy reduction or reversal, generating progeny with fewer chromosomes or even aneuploid cells through multipolar mitosis. This process contributes to chromosomal instability and tumorigenesis.^36–39 Overall, these studies affirm that the ploidy status of hepatocytes is crucial to HCC tumorigenesis, depending on the context.^40,41 However, the precise mechanisms and implications of polyploid cells in the evolution and tumor development of ICC have not been reported.

In this study, we demonstrate that mature hepatocytes exhibit a pronounced hyperpolyploidy phenotype during their malignant transformation at the initiation stage of ICC. Moreover, we identified ARID1A as a critical suppressor of hyperpolyploidy, functioning within the mitotic centrosome complex and exerting SWI/SNF-independent roles during mitosis. Loss of ARID1A induces multipolar division, drives hyperpolyploidization, and accelerates ICC initiation. Our study proposes a novel function of ARID1A, providing new insight into how ARID1A mutations contribute to ICC pathogenesis by regulating polyploidy dynamics.

METHODS

Cell lines

HEK293T (Source: human) cells were grown in the DMEM (L110KJ) containing 10% (v/v) fetal bovine serum (FBS, S660JY), 100 U/mL penicillin, and 100 U/mL streptomycin (S110JV). THLE2 (CRL-2706, Source: human) cells were grown in Endothelial Cell Medium (ECM) (Sciencell, #1001). AML12 (CRL-2254, Source: mouse) were grown in RPMI 1640 Medium (BasalMedia, L220KJ) containing 10% FBS (S660JY), 1% Insulin-Transferrin-Selenium Media Supplement (ITS, C0343). HuCCT1 (Source: human) and RBE (Source: human) cells were grown in DMEM medium (L110KJ) containing 10% (v/v) FBS (S660JY), 100 U/mL penicillin, and 100 U/mL streptomycin (S110JV). All cell lines were maintained at 37 °C in a humidified 5% CO₂ atmosphere.

Human samples

In Figure 6, 6 tumor-adjacent liver samples were obtained from ICC patients, and 3 adjacent normal liver samples were from hepatic hemangioma, hepatobiliary stones, and hepatic focal nodular hyperplasia patients. All research was conducted in accordance with both the Declarations of Helsinki and Istanbul. Prior informed consent was obtained from each patient, and the research was approved by the ethics committee of Sun Yat-sen University and Shanghai Jiao Tong University School of Medicine. The Ethics Committee of First Affiliated Hospital, Sun Yat-Sen University, approved the collection and use of human samples.

FIGURE 6.

ARID1A low expression is highly correlated with SOX9 and hepatocyte polyploidy in human ICC tumor-adjacent tissues. (A) Schema for ARID1A and SOX9 staining on normal liver samples adjacent to non-ICC hepatic disease tissues and ICC tumor-adjacent tissues. (B) Representative fluorescent images of ARID1A in the indicated liver sections. Scale bar, 50 μm. (C) Quantification of nuclear size from (B). Cell number (n) = 145–150 cells for each group. (D) Representative IHC images of ARID1A in the indicated liver sections. Scale bar, 50 μm. The higher magnification revealed varying levels of ARID1A expression in cells with different nucleus sizes. Black arrowheads point to cells with low expression of ARID1A. Scale bar, 20 μm. Relevance statistics for ARID1A expression levels and nuclear size are quantified on the right. N=3 and 6 patients/group. Cell number (n) = 205 cells for each group. (E) Representative fluorescent images of ARID1A and SOX9 in the indicated liver sections. Scale bar, 50 μm. (F) Working model: During ICC induction, hepatocytes transform into progenitor-like cells coexpressing SOX9 and HNF4α, displaying a hyperpolyploidy phenotype. ARID1A acts as a key suppressor of hyperpolyploidy, and the loss of ARID1A promotes polyploidization by impairing its function within the mitotic centrosome complex, thereby contributing to chromosome instability and ICC formation. All data in this figure are represented as mean ± SEM, ****p<0.0001. Data were analyzed by a 2-tailed Student t test. Abbreviations: ICC, intrahepatic cholangiocarcinoma; IHC, immunohistochemistry; SOX9, SRY-box transcription factor 9.

Mice

C57BL/6 mice and BALB/c nude mice were purchased from Shanghai SLAC Laboratory Animal Co., Ltd. Six to eight-week-old mice were used. C57BL/6 strain background Arid1a ^Flox/Flox ; Albumin-Cre mice were used for liver-specific knockout as previously reported.⁴² C57BL/6 strain background P53 ^Flox/Flox mice (JAX: 008462) were kindly provided by Dr Qi Wang [Shanghai Jiao Tong University School of Medicine (SHSMU), China]. Rosa26-mTmG mice (Jax stock No. 007676) were kindly provided by Dr Fubin Li (SHSMU, China). KRT19-creERT2 (JAX: 026925) mice were kindly provided by Dr Jing Zhang (SHSMU, China). PCR-based genotyping was performed using the primer pairs described in Supplemental Table S1, http://links.lww.com/HEP/J834. All experiments were done in a sex-controlled and age-controlled fashion unless otherwise noted in the figure legends. All experiments related to animals have received approval from the Institutional Animal Care and Use Committee of Shanghai Jiao Tong University School of Medicine (Permit number: A-2021-012).

HDT-induced ICC cancer model

The hydrodynamic transfection (HDT) ICC model was performed as previously reported.⁴³ The plasmid ratio is pT2-shmArid1a/pT2-shmAurka/pT2-shscramble:pT3-NICD/pT3-YAP:pT3-AKT:SB100 (5 μg:10 μg:5 μg:4 μg). Serum TBA (total bile acid) was measured using TBA kits (E003-2-1, Nanjing Jian Cheng Bioengineering Institute) at a fixed time after HDT. Livers were harvested to determine tumor burden.

Images quantification

The classic 3-zone system for the hepatic lobule divides the CV–PV axis into 3 parts. The centrilobular zone is for 1–5, the mid-lobular zone for 6–10, and the periportal zone for 11–15, as previously reported.⁴⁴ For in vivo and in vitro hepatocyte quantification, only hepatocytes with round nuclei were analyzed. Nuclear and cellular sizes were measured using DAPI and β-catenin signals by Microscopy Image Analysis Software (Imaris), respectively. For the nuclear and cellular size measurement along the CV–PV axis, more than 5 different fields of the microscope were selected randomly from each liver, 3–5 mice for each group and ~50 CV–PV axes were analyzed. Cells with a nuclear size of 100 and 200 μm² or greater are classified as polyploid cells and hyperpolyploid cells. For the quantification of multipolar division, centrosomes of synchronized cells were measured using AURKA or γ-Tubulin signals by Microscopy Image Analysis Software. Cell division with 3 or more centrosomes was referred to as multipolar division. Each group was analyzed from over 100 dividing cells from 3 independent experiments. All images were taken with a fixed exposure time and under the same conditions in 1 experiment. Fluorescence intensities were quantified from raw 16-bit images, and statistical analysis was performed using GraphPad Prism.

RNA-sequencing accession number

The sequencing data are available under the GEO accession: RNA-seq (GSE276342).

RESULTS

Hepatocytes dedifferentiate and transform into ICC under chemical toxins and genetic stress

To elucidate the hepatocyte origin of ICC, we established 2 mouse models based on the previously reported methods: a hydrodynamic transfection (HDT) method induced an ICC model and a chemical carcinogen (DEN+CCl₄) induced ICC model in p53 knockout mice (Figures 1A, B).⁴⁵ The result showed that in the HDT-induced ICC model, gross liver morphology and histological examinations revealed ductular reactions and tumor nodules characterized by CK19-positive cells, demonstrating that ICC was successfully induced after 3 weeks under the activation of oncogenes (overexpression of NICD+AKT or YAP+AKT) (Figures 1C, D and Supplemental Figure S1A, http://links.lww.com/HEP/J834). Additionally, we observed solitary and clustered CK19-positive cells, indicating the early cellular origin of ICC (Figure 1D). To further eliminate the possibility that oncogene expression within hepatocytes non-autonomously leads to ductular cell proliferation, we established a lineage tracing model (Supplemental Figure S1B, http://links.lww.com/HEP/J834). The results demonstrated that after HDT, CK19-positive ICC tumor cells originated directly from hepatocytes, rather than through non-autonomous clonal expansion from ductular cells (Supplemental Figure S1C, http://links.lww.com/HEP/J834). Additionally, we observed HNF4α (hepatocyte marker) and SOX9 (liver progenitor marker) double-positive hepatocytes during the very early stage of HDT-induced ICC (Figure 1E). This suggests that hepatocytes dedifferentiate into hepatic progenitor-like cells at the initial phase of ICC. These progenitor-like cells are likely the origin of ICC formation. Similarly, in hepatocyte-specific p53 knockout mice treated with the hepatocarcinogen DEN plus CCl₄, immunostaining for SOX9 and HNF4α revealed that hepatocytes undergo dedifferentiation during the early phase (Figure 1E). These transformed cells subsequently progressed to ICC, and with lung metastasis at later stages (Figures 1F, G and Supplemental Figure S1D, http://links.lww.com/HEP/J834). Overall, these 2 models effectively demonstrate the potential of hepatocytes to transform into ICC tumor cells, with SOX9-positive hepatocytes serving as the source of ICC tumor cells in the early stages of induction.

FIGURE 1.

Hepatocytes transform into ICC cells, undergoing hyperpolyploidy during the process. (A, B) Schema to generate HDT-induced ICC model and p53Fl/Fl; Alb-Cre + DEN and CCl₄-induced ICC model. (C) Liver gross images from mice at 3 weeks after HDT. Scale bar, 500 μm. TBA levels on the right. (D) Representative fluorescent images of CK19 in liver sections from mice 3 weeks post-HDT (PT3-NICD+AKT). The circled area highlights ICC tissues. The magnified image and white arrows indicate CK19-positive hepatocytes. Scale bars, 50 μm. (E) Representative fluorescent images of HNF4α and SOX9 in liver samples. HDT mice at 2 weeks, p53Fl/Fl; Alb-Cre (DEN+CCl₄) mice at 2 months. Arrows indicate HNF4α and SOX9 double-positive hepatocytes. Scale bar, 50 μm. (F) Liver and lung gross images from p53Fl/Fl and p53Fl/Fl; Alb-Cre mice at 4 months after DEN+CCl4. Scale bar, 500 μm. (G) Representative fluorescent images of liver tissues from p53Fl/Fl; Alb-Cre mice following 4 months of DEN+CCl₄ treatment. The circled area showed ICC tissues, and white arrows indicated CK19-positive cells. Scale bars, 50 μm. (H) Representative H&E images of the indicated liver samples. HDT mice at 2 weeks, p53Fl/Fl; Alb-Cre mice at 2 months. Scale bar, 50 μm. The magnified images reveal enlarged nucleus sizes in the livers of 2 ICC models. Scale bar, 20 μm. (I) Zonal distribution of SOX9-positive (SOX9+) hepatocytes along the CV–PV axis. Scale bar, 50 μm. (J) The nuclear size and frequency of SOX9+ hepatocytes in the indicated liver samples. Grouped analyses of SOX9+ hepatocytes with nuclear size ≥100 μm² (highlighted in the shadow area). N=3 mice/group from 3 independent experiments. n=500 cells/group. (K) Schema for primary hepatocyte isolation by 2-step collagenase perfusion and then flow cytometry. (L) The DNA content of hepatocytes in mice 2 weeks post-HDT was analyzed by flow cytometry. Flow cytometry analysis was performed to assess SOX9 and HNF4α expression in 2c, 4c, and ≥8c cells, respectively. All data in this figure are represented as mean±SEM, ***p<0.001, and ****p<0.0001. Data were analyzed by a 2-tailed Student t test. All experiments were performed in triplicate. Abbreviations: AKT, AKT serine/threonine kinase; CCl₄, carbon tetrachloride; DEN, diethylnitrosamine; HDT, hydrodynamic transfection; H&E, hematoxylin and eosin; ICC, intrahepatic cholangiocarcinoma; NICD, Notch intracellular domain; SOX9, SRY-box transcription factor 9; TBA, total bile acid; WT, wild type.

During the initiation of ICC, SOX9-positive hepatocytes are predominantly hyperpolyploid cells

Next, we analyzed the pathology and histology of liver samples at the initial phase of ICC. Interestingly, HNF4α immunofluorescent and H&E staining showed that some hepatocytes in both models were larger in size compared to those in normal liver, suggesting the possibility of polyploidy (Figures 1E, H). We quantified nuclear area and the nucleus-to-cytoplasm ratio in immunofluorescence images (Supplemental Figures S1E, F, http://links.lww.com/HEP/J834). Comparative results indicated that in both ICC models, the nuclei were significantly enlarged, with a reduced nucleus-to-cytoplasm ratio compared to normal hepatocytes. Then, we isolated hepatocytes and measured their DNA content using flow cytometry. The results showed an increased proportion of polyploid hepatocytes, exhibiting 8c and higher DNA content, across both models than normal liver and PT3-Vector controls (Supplemental Figure S1G, http://links.lww.com/HEP/J834). This indicates that hepatocytes altered their chromosomal ploidy in response to toxic or genetic stress, with some transitioning into polyploid cells. Notably, there was a significant increase in cells with chromosome copy numbers ≥8 (Supplemental Figure S1G, http://links.lww.com/HEP/J834). We refer to these cells as hyperpolyploid cells. We then examined the spatial distribution of these cells using glutamine synthetase (GS) as a marker for zone 3 and discovered that hyperpolyploid hepatocytes were predominantly located in zone 2, between the central vein and the portal area (Supplemental Figure S1H, http://links.lww.com/HEP/J834).

To further investigate the relationship between hepatocyte malignant transformation and hyperpolyploidy, we quantified the nuclear area of SOX9-positive (SOX9+) hepatocytes using image analysis and flow cytometry. The results demonstrated a positive correlation between SOX9+ hepatocytes and hyperpolyploid nuclei in both ICC models (Figures 1I–L and Supplemental Figure S1I, http://links.lww.com/HEP/J834). This suggests that in the context of ICC tumorigenesis, the SOX9+ hepatocytes are predominantly hyperpolyploid. To determine whether SOX9+ hyperpolyploid hepatocytes ultimately contribute to cancer, we induced ICC models using HDT and sorted hepatocytes at early stages before nodule formation. Using flow cytometry, we separated hepatocytes into hyperpolyploid (≥8c) and non-hyperpolyploid (2c+4c) populations based on DNA content (Supplemental Figures S1J, K, http://links.lww.com/HEP/J834). In vitro, non-hyperpolyploid hepatocytes failed to sustain long-term proliferation, whereas hyperpolyploid hepatocytes exhibited prolonged proliferation and eventually became CK19-positive cells (Supplemental Figures S1L, M, http://links.lww.com/HEP/J834). Furthermore, isolated hyperpolyploid cells formed ICC tumors in vivo (Supplemental Figure S1N, http://links.lww.com/HEP/J834), indicating that these hyperpolyploid hepatocytes could be the source of tumorigenesis. In summary, these findings demonstrate that during the initiation phase of ICC, SOX9+ hepatocytes are predominantly hyperpolyploid cells, likely serving as the origin of ICC.

Hyperpolyploid hepatocytes undergo malignant transformation into ICC cells in vitro

Next, we utilized a lentiviral method to transduce NICD and AKT plasmids into the immortalized human (THLE2) and mouse (AML12) hepatocyte cell lines, simulating HDT-induced ICC in vitro (Supplemental Figure S2A, http://links.lww.com/HEP/J834). Following in vitro induction for 21 days, gene expression analysis revealed increased levels of SOX9, as well as other biliary cell markers (Supplemental Figure S2B, http://links.lww.com/HEP/J834), indicating that this method effectively induces stepwise conversion of hepatocytes over time in vitro. During this process, nuclear staining revealed a marked increase in nuclear size (Supplemental Figures S2C, D, http://links.lww.com/HEP/J834). Flow cytometry sorting followed by chromosome analysis confirmed that the cells had acquired a hyperpolyploid state (Supplemental Figures S2E, F, http://links.lww.com/HEP/J834). Fluorescent and flow cytometry staining for SOX9 and HNF4α demonstrated that SOX9+ cells were hyperpolyploid, whereas SOX9-negative or low-intensity SOX9 cells were non-hyperpolyploid after NICD+AKT induction (Supplemental Figures S2G–K, http://links.lww.com/HEP/J834). These results suggest that polyploidization occurs concurrently with hepatocyte conversion in vitro.

Cells that experience mitotic or cytokinesis failure become hyperpolyploid and may subsequently enter senescence or undergo cell death if they attempt further division, triggered by checkpoint activation. To assess this, we used flow cytometry to analyze cell death. The results showed that overexpression of NICD alone induced polyploidization and significant cell death (Supplemental Figure S2L, http://links.lww.com/HEP/J834). However, when NICD and AKT were overexpressed together, the number of hyperpolyploid cells increased, while cell death dramatically decreased compared to the NICD-only group (Supplemental Figure S2L, http://links.lww.com/HEP/J834). We hypothesize that AKT overexpression inhibits cell death, enabling hyperpolyploid cells to survive. These hyperpolyploid cells might then undergo additional aberrant divisions, leading to chromosomal instability, which may contribute to the development of ICC (Supplemental Figures S2M, N, http://links.lww.com/HEP/J834). These findings suggest that hepatocytes become hyperpolyploid cells during the malignant transformation process. Genes such as AKT inhibit cell death, enabling these hyperpolyploid cells to transform.

Intervention in polyploidization affects hepatocyte lineage conversion and ICC tumorigenesis

Next, we examined the features of hyperpolyploid cells during the ICC transformation. We transfected THLE2 cells with NICD and AKT, then separated hyperpolyploid cells (≥8c) and non-hyperpolyploid cells (2c+4c) for bulk RNA sequencing (Figure 2A). Differential gene expression analysis revealed that compared to 2c+4c cells, hyperpolyploid cells showed enrichment of signaling pathways such as TCA cycle, cysteine and methionine metabolism, and TNF signaling pathway (Supplemental Figure S3A, http://links.lww.com/HEP/J834). This suggests that hyperpolyploid cells exhibit altered metabolism and immunogenicity, which might contribute to lineage conversion and tumor initiation. Additionally, pathways related to the mitotic spindle were significantly suppressed in hyperpolyploid cells (Figures 2B–D). These pathways are unsurprising, as previous reports have shown that the alteration of these genes can lead to replication stress or mitotic errors, resulting in failed cell division and polyploid cell formation.²³

FIGURE 2.

Intervention in polyploidization affects hepatocytes' malignant transformation and ICC tumorigenesis. (A) THLE2 cells induced by NICD+AKT were sorted by FACS. The sorted hyperpolyploid (≥8c) and non-hyperpolyploid (2c+4c) populations were subjected to RNA-seq. (B) The bubble chart visualizes the upregulated and downregulated gene-enriched pathways in hyperpolyploid versus non-hyperpolyploid THLE2 cells from the RNA-seq analysis. (C) The volcano map of upregulated and downregulated genes in hyperpolyploid versus non-hyperpolyploid THLE2 cells from the RNA-seq analysis. (D) The heatmap shows the CPM values of related genes in the mitosis pathway. (E) Schema to generate AURKA knockdown and NICD+AKT overexpression THLE2 cells. (F) The DNA content of shNC and shAURKA THLE2 cells with or without NICD+AKT overexpression was analyzed by flow cytometry, and the percentages of 8c cells were quantified as shown. (G) Nuclear morphology of shNC and shAURKA THLE2 cells with or without NICD+AKT overexpression. Scale bar, 50 μm. (H) Representative fluorescent images of HNF4α and SOX9 in shNC and shAURKA THLE2 cells. Scale bar, 20 μm. Arrows indicate HNF4α and SOX9 double-positive hepatocytes. Arrowheads indicate HNFα+/SOX9− cells. The proportions of SOX9+ cells and SOX9− cells in shNC and shAURKA THLE2 cells with NICD+AKT activation are shown on the right. (I) Schema for generating ICC model using the NICD+AKT HDT method with or without Aurka knockdown. (J) Representative H&E images of in vivo HDT livers from (I). Mice were collected at 2.5 weeks. Scale bar, 50 μm. (K) The DNA contents of hepatocytes in PT2-shNC/shAurka HDT mice were analyzed by flow cytometry. Mice were collected at 2.5 weeks. (L) Representative fluorescent images of HNF4α and SOX9 in PT2-shNC/shAurka HDT mice. Mice were collected at 2.5 weeks. Scale bar, 50 μm. (M) The proportion of SOX9+ hyperpolyploid, SOX9− hyperpolyploid, SOX9+ non-hyperpolyploid, and SOX9− non-hyperpolyploid cells in PT2-shNC/shAurka HDT mice, measured by Microscopy Image Analysis Software (Imaris). Cells with a nuclear size of 200 μm² or greater were categorized as hyperpolyploid cells. N=5 mice/group. (N) Representative H&E and CK19 staining of liver sections from PT2-shNC/shAurka mice at 3 weeks post-HDT (NICD+AKT). Scale bar, 100 μm. All data in this figure are represented as mean±SEM, *p<0.05, and ***p<0.001. Data were analyzed by 2-way ANOVA. All experiments were performed in triplicate. Abbreviations: AKT, AKT serine/threonine kinase; AURKA, aurora kinase A; HDT, hydrodynamic transfection; ICC, intrahepatic cholangiocarcinoma; NC, negative control; NICD, Notch intracellular domain; RNA-Seq, RNA sequencing; SOX9, SRY-box transcription factor 9; WT, wild type.

We observed that AURKA was downregulated in this system (Figure 2D). AURKA (Aurora Kinase A) is reported to play a critical role in cell division, and its dysfunction can lead to multipolar spindle formation, cytokinesis failure, and the generation of polyploid cells. We then investigated whether manipulating hyperpolyploidy by knocking down AURKA could affect ICC formation. We performed AURKA knockdown along with NICD and AKT induction, both in vitro and in vivo, and examined the formation of hyperpolyploidy and cell malignant transformation (Figures 2E, I). Flow cytometry and image results revealed that AURKA knockdown led to a further increase in hyperpolyploid cells, accompanied by an elevated number of SOX9-positive staining cells (Figures 2F–H). In the mouse ICC model, histological and fluorescence image quantification showed a higher number of hyperpolyploid cells in the liver tissue of the shAurka group (Figures 2J, K and Supplemental Figure S3B, http://links.lww.com/HEP/J834). Consistently, SOX9 and HNF4α double-positive hepatocytes increased in the early stages of malignant transformation when Aurka knockdown (Figures 2L, M). Furthermore, Aurka knockdown increased ICC tumorigenesis (Figure 2N). These findings suggest that AURKA is essential for proper hepatocyte division and that modulating polyploidy formation through AURKA knockdown enhances hepatocyte-to-ICC malignant transformation.

Arid1a deficiency induces hyperpolyploidy and facilitates ICC tumorigenesis in mouse models

The cellular origin of ICC in humans, whether it arises from hepatocytes and undergoes transformation and whether the involvement of polyploidy in this process remains uncertain. We selected the differentially expressed genes in hyperpolyploid cells (≥8c) compared to non-hyperpolyploid cells (2c+4c), and compared them alongside the genes mutated in human ICC to evaluate human relevance. Notably, only ARID1A gene expression was found to be reduced in hyperpolyploid cells, aligning with the loss-of-function mutations observed in ICC patient data (Figure 3A). The role of ARID1A and its mutations in the development of ICC is not well understood. To investigate whether ARID1A regulates hepatocyte polyploidy and ICC formation, we utilized liver-specific Arid1a knockout (Arid1a LKO) mice and induced ICC using the HDT method (Figure 3B). At the initiation stage of ICC, Arid1a LKO mice exhibited significantly enlarged nuclei and an increased number of polyploid cells (Figures 3C–F). Compared to wild-type (WT) controls, Arid1a LKO livers showed an increased presence of SOX9 and HNF4α double-positive hepatocytes, indicating a higher rate of tumor cell initiation (Figures 3G, H and Supplemental Figures S4A–C, http://links.lww.com/HEP/J834). These SOX9+ hepatocytes are primarily located in the intermediate zone of the liver tissue (Supplemental Figure S4D, http://links.lww.com/HEP/J834). At later stages, Arid1a LKO mice demonstrated a more severe tumor burden compared to WT controls, evidenced by larger and more numerous CK19-positive nodules (Figures 3I, J). We also conducted hepatocyte and bile duct cell-specific knockout to confirm that Arid1a deficiency in hepatocytes does not affect cholangiocyte responses or clonal expansion non-autonomously (Supplemental Figures S5A–J, http://links.lww.com/HEP/J834). Moreover, we knocked down Arid1a alongside NICD and AKT induction. The results revealed that Arid1a knockdown increased cell hyperpolyploidy at the early phase and ICC formation at the late phase (Supplemental Figures S5K–R, http://links.lww.com/HEP/J834). Furthermore, we generated Arid1a and p53 double-knockout mice. Compared to p53 single knockout, the double gene knockout mice exhibited an increased number of hyperpolyploid hepatocytes expressing SOX9 at the early stage of ICC (Figures 3K, L). These late-stage cancers often presented with a more malignant phenotype and lung metastasis (Figures 3M, N). In summary, our findings indicate that Arid1a is crucial for maintaining cell ploidy. Arid1a deficiency promotes hepatocyte polyploidy and increases the risk of ICC formation and progression.

FIGURE 3.

ARID1A deficiency induces hepatocyte hyperpolyploidy and facilitates ICC formation. (A) The Venn diagram represents the differential genes of hyperpolyploid versus non-hyperpolyploid THLE2 cells and the genes mutated in human ICC. The histogram shows the mutation rates in ICC patients from TCGA data. (B) Schema for generating the HDT-induced ICC model in Arid1aFl/Fl and Arid1aFl/Fl; Alb-Cre mice. (C) Representative H&E images of livers from Arid1aFl/Fl and Arid1aFl/Fl; Alb-Cre mice at 2.5 weeks after HDT. Scale bar, 50 μm. The magnified images at the bottom reveal variations in hepatocyte nucleus sizes. Scale bar, 20 μm. (D) Nuclear morphology of livers from Arid1aFl/Fl and Arid1aFl/Fl; Alb-Cre mice at 2.5 weeks post-HDT. Scale bar, 20 μm. (E) Quantification of hepatocyte nuclear size in the indicated liver samples from (D). N = 3 mice/group from Arid1aFl/Fl and Arid1aFl/Fl; Alb-Cre mice. n = 334 and 338 cells/group. (F) The DNA contents of hepatocytes in Arid1aFl/Fl and Arid1aFl/Fl; Alb-Cre at 2.5 weeks post-HDT were analyzed by flow cytometry. (G) Representative fluorescent images of HNF4α and SOX9 of liver sections from Arid1aFl/Fl and Arid1aFl/Fl; Alb-Cre at 2.5 weeks post-HDT. Scale bar, 50 μm. (H) Quantification of SOX9 and HNF4α double-positive cells in Arid1aFl/Fl mice and Arid1aFl/Fl; Alb-Cre mice at 2.5 weeks post-HDT. N=3 mice/group. (I) Liver gross images from Arid1aFl/Fl and Arid1aFl/Fl; Alb-Cre at 3.5 weeks post-HDT. Scale bar, 500 μm. (J) Representative H&E images and CK19 staining of the indicated livers at 3.5 weeks post-HDT. Scale bar, 100 μm. (K) Schema for generating a DEN+CCl4-induced ICC model. WT, p53Fl/Fl; Alb-Cre and Arid1aFl/Fl; p53Fl/Fl; Alb-Cre mice were injected with DEN (20 mg/kg) at p14 and 10% CCl4 biweekly for 2 months or 4 months. (L) Representative fluorescent images of SOX9 and CTNNB1 of liver sections from the indicated mice. Mice were collected at 2 months. Scale bar, 20 μm. (M) Representative gross liver and lung images of mice. Mice were collected at 4 months. Scale bar, 500 μm. (N) Representative H&E and SOX9 images of mice from (M). Scale bar, 50 μm. All data in this figure are represented as mean ± SEM, **p<0.01 and ***p<0.001. Data were analyzed by a 2-tailed Student t test. All experiments were performed in triplicate. Abbreviations: CCl₄, carbon tetrachloride; H&E, hematoxylin and eosin; HDT, hydrodynamic transfection; ICC, intrahepatic cholangiocarcinoma; SOX9, SRY-box transcription factor 9; WT, wild type.

ARID1A interacts with the centrosome complex via its C-terminal domain during mitosis, regulating polyploidy through SWI/SNF-independent functions

ARID1A is a typical chromatin remodeling factor that constitutes the SWI/SNF complex, restructuring chromatin and regulating gene expression.^17,42 The mechanism by which ARID1A influences hepatocyte polyploidization remains unclear. To investigate the molecular mechanism, we first knocked down ARID1A in THLE2 and AML12 cells, then transfected them with NICD+AKT in vitro (Figures 4A, B and Supplemental Figure S6A, http://links.lww.com/HEP/J834). Nuclear staining and flow cytometry for DNA content revealed that following ARID1A knockdown, cells exhibited a significant increase in hyperpolyploidy (Figures 4C, D and Supplemental Figures S6B, C, http://links.lww.com/HEP/J834). SOX9 immunofluorescence staining showed that the number of SOX9+ cells was increased following ARID1A knockdown, and SOX9+ cells were primarily larger cells, while SOX9-negative were smaller cells (Figures 4E, F and Supplemental Figure S6D, http://links.lww.com/HEP/J834). These data indicate that ARID1A deficiency enhances hepatocyte polyploidization and malignant transformation in vitro. We also utilized 3’UTR shRNA to knock down endogenous ARID1A and then reintroduced Myc-ARID1A wild-type to examine the rescue effect. The results showed that most of the cells transfected with Myc-ARID1A exhibited smaller nuclear sizes compared to the control (Figures 4G–I). Furthermore, when ARID1A was reintroduced, the number of SOX9+ cells decreased compared to the control (Figures 4J, K). Similar findings of reduced hyperpolyploidy, decreased SOX9 transformation and diminished tumor burden were observed in the in vivo model upon restoration of ARID1A in Arid1a LKO mice (Supplemental Figure S6E–S6G, http://links.lww.com/HEP/J834). These results demonstrate that ARID1A plays a crucial role in inhibiting polyploidization and malignant transformation of hepatocytes.

FIGURE 4.

ARID1A deficiency induces hyperpolyploidy during ICC transformation in vitro. (A) Schema to generate ARID1A knockdown and NICD+AKT overexpression in THLE2 or AML12 cells. (B) SOX9 and EpCAM mRNA levels in NICD+AKT-induced THLE2 cells with shNC or shARID1A, as measured by qPCR. (C) Nuclear morphology of THLE2 cells 14 days post-NICD+AKT overexpression, with or without ARID1A knockdown. Scale bar, 50 μm. (D) The DNA content of THLE2 cells 14 days post-NICD+AKT overexpression, with or without ARID1A knockdown, was analyzed by flow cytometry. (E) Representative fluorescent images of HNF4α and SOX9 in NICD+AKT-induced THLE2 cells, with or without ARID1A knockdown. Scale bar, 20 μm. Arrows indicate HNF4α and SOX9 double-positive hepatocytes. Arrowheads indicate HNFα+/SOX9− cells. (F) Proportions of SOX9+ and SOX9− cells in THLE2 cells with NC or NICD+AKT overexpression, and with or without ARID1A knockdown. n=1500 cells/group. (G) THLE2 cells were knocked down for ARID1A using ARID1A-3’UTR shRNA, and then wild-type Myc-ARID1A plasmid was reintroduced. NICD+AKT were overexpressed. The DNA contents of THLE2 cells, as measured by flow cytometry. (H, I) Representative fluorescent images of Myc-ARID1A in THLE2 cells from (G). White arrows point to cells with representative nuclei size and ARID1A expression level. The magnified image showed that nuclei overexpressing ARID1A became smaller. Scale bar, 50 μm. Quantification of nuclear size is shown in (I). (J) Representative fluorescent images of HNF4α and SOX9 in NICD+AKT-induced THLE2 cells with or without shARID1A-3’UTR and ARID1A plasmid rescue. Scale bar, 20 μm. White arrows indicate cells with larger nuclei and higher SOX9 expression. Arrows indicate HNF4α and SOX9 double-positive hepatocytes. Arrowheads indicate HNFα+/SOX9− cells. (K) Proportions of SOX9+ cells and SOX9− cells in NICD+AKT-induced THLE2 cells with or without shARID1A-3’UTR and ARID1A plasmid rescue. n=1200 cells/group. All data in this figure are represented as mean±SEM, **p<0.01 and ****p<0.0001. Data were analyzed by a 2-tailed Student t test and 2-way ANOVA. All experiments were performed in triplicate. Abbreviations: AKT, AKT serine/threonine kinase; ICC, intrahepatic cholangiocarcinoma; NC, negative control; NICD, Notch intracellular domain; SOX9, SRY-box transcription factor 9.

Next, we conducted rescue experiments using ARID1A truncated constructs. We ectopically expressed truncated ARID1A plasmids in cells with endogenous ARID1A knocked down (Figure 5A). The results showed that the C-terminal domain of ARID1A could significantly rescue the hyperpolyploid and SOX9 expression caused by the ARID1A deficiency, while the N-terminal and ARID domains were not as effective in reversing the phenotypes in the context of NICD+AKT overexpression (Figures 5B–D and Supplemental Figures S7A, B, http://links.lww.com/HEP/J834). These findings suggest that ARID1A plays a role in regulating polyploidy, primarily through its C-terminal domain. Given that the C-terminal ARID1A does not regulate transcription, we investigated whether loss of ARID1A directly regulates cytokinesis failure to induce polyploidization. We synchronized the cells and examined the localization of ARID1A during mitosis. The result showed that ARID1A localized to the centrosomes during mitosis (Figure 5E and Supplemental Figure S7C, http://links.lww.com/HEP/J834). This suggests that ARID1A performs distinct functions at different stages of the cell cycle, indicating a novel role in mitosis beyond its epigenetic function during interphase. Subsequently, we examined the interacting proteins of ARID1A using immunoprecipitation (IP)–mass spectrometry ²¹and found that ARID1A interacted with several cell division-related factors, such as AURKA, AURKB, TPX2, and others (Figure 5F). Co-immunoprecipitation and co-immunofluorescent assays confirmed the interaction of ARID1A with AURKA, TPX2, and KIF14 (Figure 5G and Supplemental Figures S7D–E, http://links.lww.com/HEP/J834). Further analysis revealed that the C-terminus of ARID1A interacts with AURKA (Figure 5H). Additionally, ARID1A deficiency led to the upregulation of phospho-AURKA in mouse livers, potentially contributing to aberrant cell division (Figure 5I). This suggests that, in addition to its chromatin remodeling function during interphase, ARID1A may directly engage in mitotic mechanisms by interacting with cell division-related molecules. Interestingly, ARID1A knockdown increased the number of cells with AURKA misdistribution (Figure 5J). However, reintroducing either wild-type ARID1A or the C-terminal truncated form restored the localization of AURKA (Figure 5J). Similarly, knocking down ARID1A also affected the distribution of TPX2 in the context of NICD plus AKT (Supplemental Figure S7F, http://links.lww.com/HEP/J834). Together, these results highlight the pivotal role of ARID1A as a scaffold protein in maintaining the correct localization and function of mitotic-associated molecules within centrosomes. ARID1A deficiency leads to dysfunction of centrosome complex molecules, including AURKA and TPX2, resulting in abnormal mitosis and hyperpolyploidization during ICC initiation, which eventually contributes to chromatin instability and ICC formation.

FIGURE 5.

ARID1A interacts with the centrosome complex via its C-terminal domain during mitosis, regulating polyploidy through SWI/SNF-independent functions. (A) Schema for ARID1A truncates overexpression in ARID1A-3’UTR knockdown THLE2 cells. (B) The DNA content of THLE2 cells 14 days post-NICD+AKT overexpression, with ARID1A-3’UTR knockdown and ARID1A truncates rescue, was analyzed by flow cytometry. (C) Representative fluorescent images of HNF4α and SOX9 in NICD+AKT-induced THLE2 cells with shARID1A-3’UTR and ARID1A truncates rescue. Scale bar, 20 μm. Arrows indicate HNF4α and SOX9 double-positive hepatocytes. Arrowheads indicate HNFα+/SOX9− cells. Double-stained nuclei images of SOX9 and HNF4α revealed variations in their expression levels. Scale bar, 50 μm. (D) Proportions of SOX9+ cells and SOX9− cells in NICD+AKT-induced THLE2 cells with shARID1A-3’UTR and ARID1A truncates rescue. n=1200 cells/group. (E) Confocal images of THLE2 cells expressing AURKA-EGFP, α-Tubulin-EGFP, and ARID1A-mCherry. Scale bar, 50 μm. (F) Schema for the immunoprecipitation–mass spectrometry (IP–MS) on HEK293T cells with overexpressed Flag-ARID1A. The higher magnification showed the ARID1A interacting partners related to cell cycle and cell division. (G) HEK293T cells were co-transfected with Flag-V5-ARID1A and AURKA-HA as indicated for 48 hours. Co-IP was performed using Flag beads; HA, Vinculin antibodies were used for IB. (H) HEK293T cells co-transfected with AURKA-HA and Myc-ARID1A truncates as indicated for 48 hours. Co-IP was performed using Myc beads; HA, Myc, and Vinculin antibodies were used for IB. (I) Western blot shows phosphor-AURKA expression in mouse liver tissue after ARID1A deletion. P-AURKA, Vinculin antibodies were used for IB. (J) Representative fluorescent images of AURKA in NICD+AKT-induced THLE2 cells with shARID1A-3’UTR and ARID1A truncation rescue. The magnified images and arrows showed multipolar division and bipolar division in cells. Scale bar, 50 μm. The pie chart showed the proportion of bipolar and multipolar cells. N=3-well cells/6-well plate. All data in this figure are represented as mean±SEM, **p<0.01, ***p<0.001, and ****p<0.0001. Data were analyzed by a 2-tailed Student t test and a 2-way ANOVA. All experiments were performed in triplicate. Abbreviations: AKT, AKT serine/threonine kinase; AURKA, aurora kinase A; HA, hemagglutinin; IP, immunoprecipitation; IB, immunoblotting; NICD, Notch intracellular domain; SOX9, SRY-box transcription factor 9.

ARID1A low expression is highly correlated with hepatocyte SOX9 expression and polyploidy in human ICC tumor-adjacent tissues

Next, we examined hepatocyte hyperpolyploidy and ICC initiation in human samples. Immunohistochemistry and immunofluorescence were performed on tumor-adjacent normal tissues from 6 ICC patients and adjacent normal liver tissues from 3 non-ICC hepatic disease patients to investigate the correlation between ARID1A and SOX9 expression, as well as hepatocyte polyploidy (Figure 6A). Due to the limited number of samples, we did not observe a significant difference in ARID1A staining patterns between HBV-positive (HBV+) and HBV-negative (HBV−) ICC patients. However, nuclear size was significantly larger in ICC tumor-adjacent tissues compared to normal liver tissues from non-ICC hepatic disease (Figures 6B, C). Notably, hepatocytes with low ARID1A signal exhibit larger nuclei (Figure 6D). Furthermore, we observed that SOX9-positive hepatocytes consistently displayed faint ARID1A staining and mildly enlarged nuclei in the tumor-adjacent tissues of ICC patients (Figure 6E). Additionally, based on nuclear size quantification, human ICC tissues and ICC cell lines (HUCCT1 and RBE) did not exhibit hyperpolyploidy (Supplemental Figures S8A, B, http://links.lww.com/HEP/J834). These findings in human tissue support our mouse model results, suggesting that hyperpolyploidy serves as a critical transient state for ICC initiation. Reduced ARID1A expression in patients may promote hepatocyte hyperpolyploidy and drive malignant transformation, potentially contributing to the initiation of ICC in humans (Figure 6F).

DISCUSSION

In this study, we elucidate a critical mechanism in the development of ICC. We discovered that hepatocytes transform into progenitor-like cells coexpressing SOX9 and HNF4α during ICC induction. These cells enter a hyperpolyploid state, which promotes chromosomal instability and drives malignant transformation. Furthermore, we identified epigenetic factor ARID1A as a key suppressor of hyperpolyploidy. The loss of ARID1A promotes polyploidization by impairing its function within the mitotic centrosome complex, thereby contributing to ICC formation (Figure 6F). Our findings highlight a novel SWI/SNF-independent role for ARID1A, providing new insights into how ARID1A mutations contribute to the pathogenesis of ICC.

In our system, we observed that after the HDT injection, the proportion of 2c cells increased, while 4c cells decreased compared to normal livers (without HDT) (Supplemental Figure S1G, http://links.lww.com/HEP/J834), which contrasts with previous literature suggesting that 4c cells are often the most abundant population in the adult mouse liver. We hypothesize that this shift may be attributed to the HDT insult, as the injection volume (10% of body weight) represents a significant physiological stressor.

Our results demonstrate that hepatocytes undergo hyperpolyploidy and cell death during NICD-induced transformation. However, co-transfection with NICD and AKT reduced apoptosis, enabling the survival of hyperpolyploid cells. We hypothesize that hyperpolyploidy triggers the checkpoint of mitotic errors and induces apoptosis, but oncogenic factors such as AKT activation or p53 loss allow checkpoint bypass. This bypass promotes abnormal cell division characterized by asymmetric division, micronuclei formation, and chromosomal instability, ultimately driving tumorigenesis. Therefore, hyperpolyploidy represents an essential intermediate state in hepatocyte malignant transformation during ICC initiation. However, we found that at the early stages of ICC, very few SOX9+ cells were Ki-67 or p-H3 positive, whereas the ratio of double-positive cells (SOX9+/Ki-67+ or SOX9+/p-H3+) increased in the later stages of ICC, particularly at the tumor site (Supplemental Figure S8C, http://links.lww.com/HEP/J834). These results indicate that upon transformation into SOX9+ hepatocytes, the cells are not actively proliferating at the ICC cancer-origin stage, but acquire proliferative capacity during the later stages. Further investigation is required to determine whether these hyperpolyploid cells are undergoing stress-induced senescence, adopting a state analogous to resting stem cells, or undergoing lineage conversion prior to acquiring malignant properties. Additionally, the mechanisms by which these cells re-enter the cell cycle, undergo ploidy reversal, and contribute to the development of ICC warrant further exploration.

ARID1A, a chromatin remodeling factor, is traditionally characterized by its role in gene expression. However, we identified that ARID1A localizes to the centrosome during cell division. Through knockdown and truncation rescue experiments, we demonstrated that ARID1A regulates mitotic machinery and modulates AURKA activity via its C-terminal domain, independently of its SWI/SNF complex function. Human data also indicate that approximately one-third of ARID1A mutations in ICC samples are located in the C-terminal region (Supplemental Figure S8D, http://links.lww.com/HEP/J834). However, we also observed that knockdown of ARID1A alters the expression of numerous genes. Pathway analysis of these differentially expressed genes revealed enrichment in splicing-related pathways and others, which could also contribute to hepatocyte-to-ICC malignant transformation (Supplemental Figure S8E, http://links.lww.com/HEP/J834). Furthermore, as shown in Supplemental Figure S4A, http://links.lww.com/HEP/J834, Arid1a knockout also promotes the expression of SOX9 in 2c and 4c cells. We hypothesize that this may result from direct transcriptional upregulation of SOX9 by Arid1a or other distinct mechanisms, which are separate from its role in hyperpolyploidy. Collectively, these findings suggest that ARID1A may regulate polyploidy and ICC through dual mechanisms, involving both mitotic regulation and transcriptional control.

AURKA, a serine/threonine kinase, plays a critical role in cell mitosis and has been reported as a promising target in cancer therapy when using AURKA inhibitors. We tested the AURKA inhibitor Alisertib in our system and found that it significantly induced mitotic catastrophe and increased cell death (Supplemental Figures S8F, G, http://links.lww.com/HEP/J834). We hypothesize that cells with mild cell cycle dysfunction may develop hyperpolyploidy, which could favor malignant progression. In contrast, severe mitotic catastrophe, which leads to extensive cell death, is not conducive to malignant transformation. Furthermore, we observed that ARID1A-knockdown human ICC cancer cell lines (HUCCT1 and RBE) were more sensitive to Alisertib treatment compared to vehicle controls (Supplemental Figure S8H, http://links.lww.com/HEP/J834). We interpret this as the inhibition of AURKA, potentially affecting cell proliferation in ARID1A-deficient ICC tumor cells through mechanisms other than hyperpolyploidy. These findings suggest that AURKA inhibitors could represent a promising therapeutic strategy for treating ARID1A-deficient ICC cancers.

Supplementary Material

hep-83-1416-s001.pdf ^{(2.5MB, pdf)}