mTORC1 Promotes ARID1A Degradation and Oncogenic Chromatin Remodeling in Hepatocellular Carcinoma

SWItch/sucrose non-fermentable (SWI/SNF) chromatin remodeling complex hydrolyzes ATP and uses the resulting energy to mobilize nucleosomes and alter accessibility of chromatin to transcriptional and coregulatory machineries (1). SWI/SNF complexes consist of one ATPase subunit (BRM or BRG1), core subunits (BAF47, BAF155, and BAF170) that contribute to their catalytic activity, and variable subunits that confer specificity for target promoters (2). There are two canonical SWI/SNF complexes: BRG1/BRM-associated factor (BAF) and polybromo-associated BAF (PBAF). ARID1A is a highly conserved subunit of the BAF complex. Chromatin remodeling plays important roles in normal physiology and diseases, particularly cancer. One interesting phenotype in Arid1a knockout mice is increased tissue regenerative capacity (3), suggesting that ARID1A restrains proliferative capacity that is important for tumorigenesis. Indeed, ARID1A is frequently mutated in a broad range of cancers, including liver cancer. Somatic mutations in cancer, predominantly missense mutations, occur across the entire length of the ARID1A gene, resulting in ARID1A protein loss of function (4, 5). Moreover, ARID1A mutations are significantly correlated with poor prognoses in many different types of cancers (6). Despite remarkable advances in our understanding of the functions of ARID1A in chromatin remodeling and carcinogenesis, whether and how chromatin remodeling machineries are regulated by oncogenic growth signaling pathways remains poorly understood.

Mechanistic target of rapamycin (mTOR) is a phylogenetically conserved serine/threonine kinase that forms two distinct complexes, mTORC1 and mTORC2. mTORC1 is a central controller of cell growth and metabolism in eukaryotes (7, 8). It acts downstream of PI3K and AKT. The mTORC1 pathway is a major oncogenic driver that is commonly activated during carcinogenesis (9, 10). mTORC1 promotes mRNA translation while inhibiting autophagy (7, 8). Growing evidence also documents that mTORC1 engages in nuclear signaling. mTORC1 binds to the promoters of rRNA and tRNA genes and controls their transcription by RNA polymerases I and III, in a nutrient-dependent and rapamycin-sensitive manner (11, 12). mTORC1 is also known to interact with and regulates the activity of certain Pol II target promoters (13, 14). In addition, mTORC1 targets histone acetyltransferase p300 (15), suggesting a role of mTOR in regulation of the epigenome.

Hepatocellular carcinoma (HCC) is a common malignancy and a leading cause of cancer-related deaths worldwide (16). ARID1A is mutated in 10%–15% of human HCC and is the most frequently mutated of the SWI/SNF complex genes, resulting in reduced protein expression (17). Consistently, genetic deletion of Arid1a in mice accelerates tumor progression and metastasis (18), supporting the notion that inactivation of ARID1A by somatic mutations promotes HCC development and progression. Intriguingly, somatic mutations in MTOR and SWI/SNF complex are mutually exclusive in human liver cancers, suggesting a common function in hepatocarcinogenesis. Herein, we report that mTORC1 nuclear signaling regulates ARID1A in HCC. Hyperactive mTORC1 signaling promotes SCF-dependent ubiquitination and proteasomal degradation ARID1A, which enhances chromatin remodeling-mediated activation of the YAP pathway, promoting oncogenic growth, and liver tumorigenesis. Our results reveal a common mechanism through which ARID1A protein is targeted by oncogenic growth signaling for degradation in HCC.

HA-mTOR was cloned in pcDNA3.1 for transient expression. pcDNA6-ARID1A (#39311), pRK5-HA-Ub-K48 (#17605) and pRK5-HA-Ub-K48R (#17604) were purchased from Addgene. pT3-EF1α, pT3-EF1α-HA-myr-AKT, pT2CAGGS-NRAS^V12, and pCMV-SB plasmids were kindly provided by Dr. Xin Chen (University of California San Francisco, San Francisco, CA; ref. 19). pT3-EF1α-ARID1A was constructed by subcloning ARID1A from pcDNA6-ARID1A into pT3-EF1α. shArid1a was cloned in pT3-EF1α to generate pT3-EF1α-shArid1a. siRNAs and shRNAs for HCC cells were purchased from Sigma. Sequences of siRNAs and shRNAs are shown in Supplementary Table S1.

HepG2, SNU449, HEK293T, and MEF cells were from Zheng Laboratory cell line stocks. Huh7 was a kind gift from Dr. Xin Chen (University of California San Francisco). These cell lines were authenticated by the ATCC Cell STR Testing Service (last tested on May 20, 2021). Cell lines were typically maintained for up to 5 passages after reviving in DMEM (Invitrogen) plus 10% FBS (Biological Industries). SNU449 human HCC cells were cultured in RPMI-1640 Medium (Invitrogen) plus 10% FBS. Cells were incubated at 37°C in a humidified chamber containing 5% CO₂. Plasmid and siRNA transfections were performed with Lipofectamine 3000 (Invitrogen) according to the manufacturer's protocol. Lentivirus was produced following the manufacturer's instruction. 2 days after transfection into HEK293T cells, viral supernatants were harvested. Huh7 cells stably expressing shControl and shARID1A were established by infection with shControl and shARID1A lentivirus, followed by selection in 2 μg/mL puromycin. Cell cultures were regularly tested for Mycoplasma using LookOut Mycoplasma PCR Detection Kit (MP0035; Sigma; last tested on January 13, 2021).

Rapamycin, AZD8055, MHY1485, MG132, MLN4924, verteporfin, and peptide 17 were purchased from Selleck Chemicals. Cycloheximide was purchased from Sigma. Cells were treated with 100 nmol/L rapamycin, 100 nmol/L AZD8055, 50 μmol/L cycloheximide, 10 μmol/L MHY1485, 20 μmol/L MG132, 1 μmol/L MLN4924, 1 μmol/L verteporfin, and 1 μmol/L peptide 17, respectively. Antibodies against β-tubulin (#2128), β-actin (#3700), p-AKT(S473; #4060), AKT (#4691), ARID1A (for IP and ChIP, #12354), ARID2 (#82342), BAF155 (#11956), BRD9 (#71232), BRG1 (#49360), GAPDH (#5174), HA-tag (#3724), lamin B1 (#17416), p-LATS1(Ser909; #9157), LATS1 (#3477), mTOR (for IP and Western blot, #2972), PBRM1 (#89123), PCNA (#13110), Raptor (#2280), p-S6 (S235/236; #4858), p-S6K (T398; #9234), S6K (#2708), V5-tag (#13202), and YAP (#14074) were purchased from Cell Signaling Technology. Antibodies against ARID1A [for Western blot, proximity ligation assay (PLA), immunofluorescence (IF), and IHC, HPA005456], CTGF (HPA031075), and Rictor (SAB4200141) were purchased from Sigma. mTOR antibody for PLA (MA5–31505) was purchased from Invitrogen. DPF2 antibody (ab134942) was purchased from Abcam.

Cells were lysed in immunoprecipitation (IP) lysis buffer (50 mmol/L Tris-HCl pH 7.5, 150 mmol/L NaCl, 5 mmol/L EDTA, 0.5% NP-40, phosphatase and protease inhibitors). After clarification by centrifugation, 500 μg cell lysates were used to incubate with IP antibodies overnight at 4°C, and then incubated with Proteinase A Sepharose beads (Millipore) for 2 hours. After washing with cold IP lysis buffer 6 times, SDS loading buffer was added. Antibody-bound materials were eluted by boiling for 5–10 minutes, separated by SDS polyacrylamide gel electrophoresis and analyzed by Western blot. Cell lysis buffer is composed of 20 mmol/L Tris-HCl pH 7.5, 150 mmol/L NaCl, 2.5 mmol/L sodium pyrophaosphate, 1% Triton, 1 mmol/L Na₂EDTA, and a phosphatase and protease inhibitor cocktail. For Western blot, polyvinylidene difluoride membrane was blocked with 5% milk in Tris buffer (10 mmol/L Tris-HCl pH 7.4, 0.1% Tween-20) and then incubated with primary antibodies with dilutions as recommended by Manufacturers (1:1,000 dilution) at 4°C overnight. After washing with Tris buffer for 3 times, the membrane was incubated with a peroxidase-conjugated secondary antibody (Bentyl) for 1 hour at room temperature. Blots were developed with Super-Signal West Pico Chemiluminescent Substrate.

PLA was performed with Duolink in situ reagents (Sigma) according to the manufacturer's instruction. Briefly, cells cultured in 35-mm dishes with 10-mm glass bottom slides (MatTek) were fixed in 4% formaldehyde for 15 minutes at room temperature. Samples were then subjected to PLA analysis. Nuclei were stained using 4′, 6-diamidino-2-phenylindole dihydrochloride (DAPI; Life Technologies). Images were acquired in multitracking mode using an Olympus scanning confocal microscope (FV1000, Olympus).

Cells cultured on 10-mm glass bottom slides (MatTek) in 35-mm dishes were fixed in 4% formaldehyde at room temperature for 40 minutes. After blocking with PBS containing 4% BSA (Sigma) for 90 minutes, cells were incubated with ARID1A antibody, followed by a secondary antibody conjugated to Alexa Fluor 488 (Life Technologies). Nuclei were stained using DAPI (Life Technologies). Images were acquired in multitracking mode using an Olympus scanning confocal microscope (FV1000, Olympus).

Total RNA was extracted from cells using TRIzol (Invitrogen) according to the manufacturer's instruction. cDNA was synthesized by reverse transcription reaction using the GoScript Reverse Transcription System kit (Promega). qRT-PCR was performed using GoTaq RT-PCR Master Mix (Promega). All samples were analyzed in triplicate and underwent denaturation at 95°C for 10 minutes and then 40 amplification cycles at 95°C for 15 seconds, 60°C for 1 minute, 55°C for 30 seconds and 95°C 30 seconds using a Roche Lightcycler 96 RT-PCR system (Roche Diagnostics). For relative quantification, ΔΔC_t was calculated using GAPDH as an internal standard control for each sample. For graphical representation, ΔΔC_t values were normalized to controls and expressed as the difference of fold change. Primer sequences are shown in Supplementary Table S1.

HCC cells were treated without or with 100 nmol/L rapamycin in the presence of 50 μmol/L cycloheximide and analyzed for ARID1A protein stability by Western blot.

After mice were euthanized, liver tissues were collected and fixed in formalin for 24 hours. For IHC staining, tissues were dried at 65°C for 2 hours and then deparaffinized in xylenes and rehydrated in graded ethanol. Endogenous peroxidase activity was blocked by incubation in 3% hydrogen peroxide (H₂O₂) in methanol for 10 minutes. Tissue sections were subjected to autoclave antigen retrieval in 1 mmol/L ethylenediaminetetraacetic acid (EDTA) buffer (pH 8.0) for 3 minutes at the maximum pressure. After cooling to the room temperature, tissue sections were incubated with anti-HA (1:500), anti–p-S6 (1:100), anti-ARID1A (1:300), anti-PCNA (1:2,000) anti-CTGF (1:50) or anti-YAP (1:100) antibodies overnight at 4°C, followed by incubation with peroxidase-linked second antibody (DakoCytomation) for 30 minutes at room temperature. Tissue sections were developed with 3,5-diaminobenzidine (DAB) substrate and counterstained with Mayer's hematoxylin. IHC quantification was performed using the IHC profile plugin in ImageJ software.

ATAC-seq was performed as described previously (20). Nuclei isolated from approximately 50,000 cells were used for transposition reaction with transposase (Illumina). Column-purified DNA was amplified in 20 μL reactions with high-fidelity 2×PCR Master Mix (NEB) using primers with unique barcodes (IDT) for setup library. For RNA-seq, total RNA was extracted from cells using TRIzol (Invitrogen) according to the manufacturer's instructions. Preparation of RNA-seq library and sequencing of ATAC-seq and RNA-seq libraries were performed by Romics. All sequencing reads were processed for quality control by FastQC. Reads of each sample were aligned to the GRCh38/hg38 reference genome using BWA (21). Peaks of ATAC-seq were identified using MACS v2.0 with cutoff q < 0.05 (22). RPKM (reads per Kb per million reads) was used to calculate gene expression from RNA-seq data. Differential accessible peaks and differential expressed genes were defined using significant analysis and FDR analysis (|log₂FC|>0, P < 0.05) by DESeq2 (23). For Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis, we used the Fisher exact test to select the significant pathway, and the threshold of significance was defined by P value and FDR (24). ATAC-seq signal tracks were presented by Integrative Genomics Viewer software (25). Gene set enrichment analysis (GSEA) was performed as described previously (26). The GEO access number for ATAC and RNA-seq data is GSE159165 (private token: kjyrmkgapryjjep).

YAP chromatin immunoprecipitation (ChIP) assay was performed using standard protocol of SimpleChIP Enzymatic Chromatin IP Kit with Magnetic Beads (CST). ChIP-enriched chromatin was analyzed by qPCR with GoTaq RT-PCR Master Mix (Promega) and normalized to the input. ChIP-qPCR primers are listed in Supplementary Table S2.

For EdU proliferation assay, Cells were cultured on 10-mm glass bottom slides (MatTek) in 35-mm dishes. EdU staining was performed using the Click-iT EdU Cell Proliferation Kit for Imaging, Alexa Fluor 594 (Invitrogen) according to the manufacturer's protocol. For CCK-8 cell proliferation assay, HCC cells were seeded at 1,000 cells per well using 96-well plate. CCK-8 reagent (Dojindo) was added to the cell culture medium at a ratio of 1:10, and the absorbance was measured at 450 nm after incubation for 2 hours at 37°C.

For hydrodynamic transfection, 8-week-old FVB/N mice were used in the study. Animals were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. Hydrodynamic injections were performed as described previously with randomized mice (27). To determine the functions of ARID1A in AKT/NRAS-driven liver tumor development, a high ARID1A to AKT/NRAS plasmid ratio was used [pT3-EF1α (20 μg), pT3-ARID1A (20 μg) or pT3-shArid1a (20 μg) along with AKT (4 μg) and/or NRAS (4 μg)]. Animal weight was monitored weekly throughout the study. With regard of sex as a biological variable, because liver cancer is a male dominant cancer, only male mice were used. Power analysis indicates that 8–12 mice/group is necessary to detect 1.5-fold differences (0.83 power, standard deviation of 30% and two-tailed t test— G*Power software). Mice were housed, fed, and monitored in accordance with protocols approved by the Institutional Animal Care and Use Committee of Sun Yat-Sen University.

Statistical data analyses were performed GraphPad Prism software (version 8.0). Experimental data were expressed as mean ± SD. The unpaired two-tail Student t test was used to evaluate statistical significance to detect difference between two groups. The Fisher exact test was used to analyze mutually exclusive of somatic mutations in chromatin remodeling and mTOR pathways in HCC. Spearman correlation analysis was used to compare the correlation between AKT–mTOR signaling and ARID1A protein expression. Repeated-measures ANOVA was performed to analyze stability and proliferation of cultured cells. Kaplan--Meier plots and log-rank test were performed to compare survival rates between mice in different hydrodynamic injected cohorts. The z score of log(AUC) was used to analyze drug sensitivity of HCC cell lines. For all statistical analyses, P < 0.05 was considered as statistically significant.

All data generated or analyzed in this study are included in this article and Supplementary Information). Materials generated in this study can be made available upon reasonable request. The GEO access number for ATAC and RNA-seq data is GSE159165.

The PI3K–mTOR pathway is one of the most mutated and activated pathways in human cancer (28). To better understand how mTOR signaling interacts with other major cancer pathways in HCC, we analyzed co-occurrence of somatic mutations in the mTOR pathway with other frequently mutated pathways (28) in human HCC tumors using The Cancer Genome Atlas (TCGA) database. Strikingly, somatic mutations of PI3K–mTOR and SWI/SNF chromatin remodeling pathways were mutually exclusive (Fig. 1A), suggesting that these two pathways are functionally linked. To further explore their relationship, we analyzed mTOR signaling and ARID1A protein expression in 184 human primary HCC tumors available at The National Cancer Institute (NCI) Clinical Proteomic Tumor Analysis Consortium (CPTAC; ref. 29). The results showed a negative correlation between ARID1A protein expression with p-AKT(T308), p-AKT(S473), and p-S6K(T389) markers for mTOR signaling (Fig. 1B and C). These observations raised an interesting possibility that mTOR signaling negatively regulates ARID1A protein level in human HCC.

To test whether mTOR regulates ARID1A, we treated Huh7 and HepG2 cells with rapamycin that resulted in blockage of mTORC1 signaling as measured by p-S6K (Fig. 1D and E). Rapamycin treatment led to an increase in ARID1A protein level (Fig. 1D and E). In contrast, rapamycin did not significantly affect protein expression of DPF2/BAF45D (a core component of BAF complex), ARID2 and PRBM1 (components of PBAF), and BRD9 (component of ncBAF; Fig. 1D and E). Conversely, treatment of HCC cells with MHY1485, a small-molecule activator of mTORC1 (30), led to decreased ARID1A protein expression (Fig. 1F and G). Similarly, IF staining showed that nuclear ARID1A protein level was enhanced by rapamycin in Huh7 and HepG2 cells (Supplementary Fig. S1A). In contrast, ARID1A mRNA expression was not affected by drug treatment as judged by qRT-PCR (Supplementary Fig. S1B). Collectively, these data indicated that mTORC1 selectively regulates ARID1A at the post-transcriptional level.

To ask whether mTORC1 is closely involved in regulation of ARID1A, we investigated their possible interaction. Transiently expressed V5-ARID1A and HA-mTOR in HEK293T cells were found to be associated with each other as judged by coimmunoprecipitation (co-IP) with anti-HA and anti-V5 antibodies (Fig. 2A). Endogenous mTOR was also found to be bound to ARID1A in a co-IP experiment in Huh7 cells using an anti-ARID1A antibody, but not a control IgG (Fig. 2B). The interaction between ARID1A and mTOR was further detected in the nucleus of Huh7 and HepG2 cells by co-IP (Fig. 2C and D), and PLA (Supplementary Fig. S1C and S1D), a method for detecting protein–protein interaction in situ in intact cells (31). Moreover, ARID1A was found to be present in the anti-Raptor IP, but not anti-Rictor IP (Fig. 2E), and interaction between Raptor and ARID1A was strengthened after treatment with mTOR kinase inhibitor in HCC cells (Supplementary Fig. S1E and S1F), indicating that ARID1A specifically interacts with mTORC1 (not mTORC2). ARID1A is a component of BAF complex that contains BRG1 and BAF155 as foundational subunits (32, 33). Interestingly, knockdown of ARID1A impaired Raptor binding to BAF core complex (BRG1 and BAF155) in HCC cells (Fig. 2F and Supplementary Fig. S1G). In contrast, knockdown of BRG1 disrupted BAF core complex (loss of both BRG1 and BAF155) but did not affect raptor binding to ARID1A (Fig. 2G and Supplementary Fig. S1H), suggesting that mTORC1 interacts with BAF complex in an ARID1A-dependent manner.

To ask how mTORC1 controls ARID1A protein level, we examined ARID1A stability under conditions of mTORC1 inhibition. Huh7 and HepG2 cells were treated with rapamycin and AZD8055 in the presence of cycloheximide (CHX) to block new protein synthesis, allowing measurement of ARID1A protein half-life. The result showed that ARID1A protein degradation was attenuated when HCC cells were treated with mTOR inhibitors (Fig. 3A and B, and Supplementary Fig. S2A). Proteasome inhibitor MG132 also stabilized ARID1A protein (Fig. 3C), suggesting that proteasome-dependent degradation is involved. Ubiquitin monomers form polyubiquitin chains through different lysine residues, among which, K48-linked chains are common for targeting proteins to proteasome for degradation (34). We ectopically expressed in Huh7 and HepG2 cells an HA-Ub-K48 mutant in which all lysine residues were substituted with arginine except K48. Anti-ARID1A IP showed that ARID1A was polyubiquitinated by HA-Ub-K48, which was inhibited by rapamycin (Fig. 3D and Supplementary Fig. S2B). In contrast, no ubiquitination of ARID1A was detected by HA-Ub-K48R, a mutant that carries the K48R mutation with all other lysine residues intact (Fig. 3E and Supplementary Fig. S2B). Hence, mTORC1 promotes ARID1A polyubiquitination through K48.

The SKP1-CUL1-F-box (SCF) E3 ubiquitin ligase plays an important role in K48-mediated ubiquitination and substrate degradation (35). Interestingly, ARID1A was reported to be a substrate of SCF complex in gastric cancer cells induced by DNA damage (36, 37). We, therefore, asked whether or not mTORC1 regulates ARID1A ubiquitination via the SCF E3 ubiquitin ligase in HCC cells. To test this conjecture, we treated Huh7 and HepG2 cells with MHY1485 to activate mTORC1 in the absence or presence of MLN4924, a selective inhibitor of NEDD8-activating enzyme necessary for SCF ubiquitin ligase activity. Indeed, MLN4924 efficiently blocked mTORC1-induced ARID1A degradation (Fig. 3F and Supplementary Fig. S2C). Moreover, SKP1, a key component of SCF complex was found to bind to ARID1A, and this interaction was inhibited by rapamycin (Fig. 3G). These results indicate that mTORC1 promotes targeting of ARID1A by SCF for ubiquitin-dependent degradation in HCC cells.

Excessive mTORC1 signaling is a driver of hepatocarcinogenesis (38). Hepatic activation of mTORC1 signaling (e.g., Pten knockout or activated Akt) is sufficient to induce HCC in genetically engineered mouse models (39, 40). Stable expression of constitutively active, myristoylated AKT together with NRAS by hydrodynamic transfection (HDT) through tail vein injection, causes HCC development by activation of mTORC1 signaling in mice (27, 41). AKT-induce tumorigenesis can be further enhanced by NRAS (27, 41). To interrogate the role of ARID1A in mTORC1-driven hepatocarcinogenesis in vivo, we generated liver tumors by HDT-mediated stable expression of activated AKT/NRAS together with ARID1A overexpression or knockdown (Fig. 4A; Supplementary Fig. S3A and S3B). As expected, AKT/NRAS mice developed large liver tumor masses by week 8 after injection (Fig. 4B). Overexpression of ARID1A attenuated tumor development, as judged by reduced tumor burden and the liver to body weight ratio compared with the AKT/NRAS group (Fig. 4B and C). No apparent tumor lesions were observed in this group by week 11 after injection (Fig. 4B and C). In contrast, co-expression of shArid1a accelerated tumor development with tumors already developed by week 4 after injection (Fig. 4B and C). Consistently, the mean survival time for the AKT/NRAS/ARID1A and AKT/NRAS/shArid1a groups was 14 and 8 weeks, respectively, compared with 11 weeks for the AKT/NRAS group (Fig. 4D). These results indicate that ARID1A antagonizes AKT/NRAS-driven hepatocarcinogenesis.

AKT-expressing cells formed preneoplastic lesions in the AKT/NRAS group by week 4 after HDT and small tumors began to emerge by week 8 after HDT (Fig. 4E and F). In contrast, livers in the AKT/NRAS/ARID1A group were normal by week 4 and only showed small preneoplastic lesions by week 8 (Fig. 4E and F). On the other hand, in the AKT/NRAS/shArid1a group, large tumors already developed by week 4 (Fig. 4E and F). Consecutive liver sections showed that AKT/NRAS activated mTORC1 as indicated by elevated p-S6 staining, which was not affected by ARID1A nor shArid1a (Fig. 4E). Consistent with our in vitro observation that mTORC1 is a negative regulator of ARID1A, ARID1A was downregulated in AKT/NRAS-driven liver tumors compared with adjacent normal tissues (Fig. 4F and Supplementary Fig. S3C). Ectopic expression of ARID1A was associated with lower tumor cell proliferation as judged by PCNA staining, whereas shArid1a enhanced tumor cell proliferation (Fig. 4F and Supplementary Fig. S3C). Together, these results indicate that ARID1A suppresses AKT–mTOR pathway-driven hepatocarcinogenesis.

To evaluate whether mTORC1 regulates chromatin remodeling and the role of ARID1A, we performed ATAC-seq in Huh7 cells treated with rapamycin in the absence or presence of ARID1A knockdown. Rapamycin treatment affected nearly 11,000 chromatin accessible sites compared with the untreated control, and the number of rapamycin-sensitive sites was reduced after ARID1A knockdown (Fig. 5A and Supplementary Fig. S4A), suggesting that mTORC1 regulates chromatin remodeling in an ARID1A-dependent manner in HCC cells. KEGG pathway enrichment analysis of mTORC1–ARID1A-specific sites showed enrichment of several oncogenic pathways with Hippo pathway as the top-ranked (Fig. 5B and Supplementary Table S2). Hippo-YAP signaling is a major growth regulator of hepatocytes and an oncogene for HCC (42). This observation suggests that mTORC1 controls YAP-dependent transcription through ARID1A-dependent chromatin remodeling. Consistently, chromatin accessibility at YAP/TEAD consensus-binding sites at the transcription start-site and intergenic regions of YAP target genes (CTGF, CYR61, ARKND1, ASAP1, BICC1, and CAT) was decreased by rapamycin, and the rapamycin effect was abrogated by ARID1A knockdown (Fig. 5C and Supplementary Fig. S4B–S4H). In contrast, ARID1A knockdown did not affect rapamycin inhibition of chromatin accessibility at the promoters of COX5a and PGC1α (Fig. 5D and E), two known mTORC1 target genes (13), suggesting that ARID1A selectively mediates mTORC1 signaling to regulate YAP target genes. To corroborate the role of the mTORC1–ARID1A axis on the regulation of gene expression, we conducted RNA-seq of Huh7 cells treated with rapamycin in the absence or presence of ARID1A knockdown. The results showed that expression of YAP target genes was repressed by rapamycin in an ARID1A-dependent manner (Fig. 5C and Supplementary Fig. S4B–S4H). GSEA showed that YAP signature genes were suppressed by rapamycin, and ARID1A knockdown alleviated rapamycin's inhibitory effect (Fig. 5F and G). Moreover, there was a strong correlation between mTORC1–ARID1A-specific chromatin-remodeling sites and alterations of mRNA expression (Fig. 5H). These observations suggest that mTORC1 regulates chromatin remodeling and YAP-dependent transcriptome in an ARID1A-dependent manner.

We next examined the effect of rapamycin on YAP recruitment to its target promoters. Rapamycin treatment inhibited binding of YAP to the promoters of CTGF and CYR61 in ARID1A wild-type cells Huh7 and HepG2, and that rapamycin effect was abrogated by ARID1A knockdown (Fig. 6A and C; Supplementary Fig. S5A and S5B). In contrast, rapamycin did not affect binding of YAP to its target promoters in ARID1A-mutant cells SNU449 (Fig. 6B and D). The binding of ARID1A and BRG1 to YAP target promoters were further analyzed by ChIP-qPCR. ARID1A displayed increased binding to YAP target promoters by rapamycin treatment (Supplementary Fig. S5C–S5F). Rapamycin also stimulated BRG1 association with YAP target promoters, which was abrogated by ARID1A knockdown (Supplementary Fig. S5C–S5F), indicating that mTORC1 regulates the binding of BAF complex to YAP target promoters in an ARID1A-dependent manner. Consistently, rapamycin inhibited mRNA expression of YAP target genes in Huh7 and HepG2 cells in an ARID1A-dependent manner, but not in ARID1A-mutant SNU449 cells (Fig. 6E and F and Supplementary Fig. S5G). Moreover, protein expression of CTGF and CYR61 was inhibited by rapamycin in ARID1A wild-type HCC cells in an ARID1A-dependent manner, but not in ARID1A-mutant HCC cells (Fig. 6G and Supplementary Fig. S5H).

Interestingly, the total protein and phosphorylation of LATS1, an upstream tumor suppressor of YAP, and YAP protein were not affected by rapamycin or ARID1A knockdown, suggesting that mTORC1 does not regulate YAP signaling per se (Fig. 6G and Supplementary Fig. S5H). Furthermore, IHC staining showed that CTGF expression was elevated in AKT/NRAS-driven liver tumors compared with the adjacent normal liver tissues (Fig. 6H and I). Co-expression of ARID1A suppressed AKT/NRAS-induced CTGF overexpression, whereas ARID1A knockdown enhanced CTGF overexpression in liver tumors (Fig. 6H and I). Consistently with the in vitro results, YAP protein was not affected by AKT–mTORC1 signaling or ARID1A status (Fig. 6H and I). Expression of YAP target genes remained to be inhibited by rapamycin after activation of YAP signaling by LAST1 knockdown (Supplementary Fig. S5I). Furthermore, analysis of TCGA database showed that YAP signature genes were more highly expressed in ARID1A-mutant HCC tumors than ARID1A wide-type HCC tumors (Supplementary Fig. S5J–S5O). Collectively, these observations show that mTORC1 promoting YAP-dependent transcription by suppressing ARID1A-mediated chromatin remodeling.

Because ARID1A mediates rapamycin inhibition of mTORC1-dependent chromatin remodeling and YAP target gene expression, we sought to understand the impact of somatic mutations in chromatin remodeling on therapeutic response to mTORC1-targeted therapy. To this end, we analyzed the Cancer Cell Line Encyclopedia and the Genomics of Drug Sensitivity in Cancer drug sensitivity datasets. Interestingly, HCC cell lines carrying somatic mutations in the chromatin remodeling pathway tended to be resistant to the rapalog temsirolimus, an FDA-approved oncology drug, whereas HCC cells wild-type in the chromatin-remodeling pathway showed propensity to be temsirolimus sensitive (Fig. 7A and Supplementary Fig. S6A). To verify this observation, we assayed for inhibition of HCC cell proliferation by rapamycin using the EdU incorporation assay. The result showed that rapamycin potently inhibited proliferation of ARID1A wild-type HCC (Huh7 and HepG2) cells, but not ARID1A-mutant HCC (SNU449) cells (Fig. 7B,–G). Knockdown of ARID1A in Huh7 and HepG2 cells not only enhanced cell proliferation, but also rendered them rapamycin resistant (Fig. 7B, D, and F, and Supplementary Fig. S6B–S6G). Because loss of ARID1A led to activation of the YAP pathway, we asked whether or not HCC cells deficient of ARID1A were responsive to YAP inhibition and found that ARID1A knockdown in Huh7 and HepG2 cells or ARID1A-mutant SNU449 cells are exquisitely sensitive to verteporfin and peptide 17, both of which are small-molecule YAP/TEAD inhibitors (Fig. 7B,–G and Supplementary Fig. S6B–S6G). Consistently, ectopic overexpression of ARID1A in ARID1A-mutant SNU449 cells restored sensitivity to rapamycin (Fig. 7G and Supplementary Fig. S6H–S6I). These findings indicate that ARID1A mutation status affects anticancer response to mTORC1 inhibition in HCC.

Chromatin remodeling enables dynamic changes in chromatin architecture, which is used to modulate gene expression programs (43). Despite recent advances in dissecting the chromatin remodeling machineries and their molecular mechanisms of action, whether and how growth signaling controls the activity of chromatin remodeling complexes, particularly in the context of oncogenesis, remain poorly understood. In this study, we show that oncogenic mTORC1 signaling negatively regulates ARID1A protein stability in HCC. Mechanistically, mTORC1 promotes binding of ARID1A to the SCF E3 ubiquitin ligase, and polyubiquitination and proteasomal degradation of ARID1A (Fig. 7H). We further show that this chromatin remodeling mechanism mediates mTORC1 promotion of YAP-dependent transcriptome.

ARID1A has the highest somatic mutation rate among known epigenetic regulators in cancer, suggesting that ARID1A is a major target of carcinogenesis (44). Cancer-causing substitutions are distributed throughout the length of ARID1A protein, which are correlated with loss of ARID1A protein expression and are associated with large tumor size and late stages of disease (45–47). Genetic ablation of Arid1a in mice attenuates tumor initiation but accelerates tumor progression and metastasis (18). These observations indicate that loss of ARID1A is almost certainly a late tumorigenic event that promotes tumor growth and progression. Chromatin remodeling and mTOR signaling are two of the most frequently targeted pathways for genetic alterations in HCC (28). Remarkably, the two groups of somatic mutations are mutually exclusive in HCC (Fig. 1A), suggesting that they act within the same oncogenic process. Although somatic mutations frequently activate mTORC1 signaling as an oncogenic driving event in HCC (38), they inactivate ARID1A chromatin remodeling pathway (45–47). Our results suggest that oncogenic mTORC1-regulated degradation of ARID1A protein represents another common mechanism to inactivate ARID1A during carcinogenesis.

YAP is a major regulator of hepatocyte proliferation, controlling liver organ size and regenerative capacity. Activation of the YAP pathway occurs frequently in human cancers (42). Hepatic overexpression of YAP or knockout of negative upstream regulators (e.g., NF2/Merlin, MST1/2, and LATS) leads to development of HCC in mice (48). These observations demonstrate that the –YAP pathway is a major oncogenic driver of liver cancer. Herein, we have shown that inhibition of mTORC1 by rapamycin blocks promoter chromatin accessibility and expression of YAP target genes in an ARID1A-dependent manner. Conversely, knockdown or genomic mutation of ARID1A increases YAP binding to its target promoters and expression of YAP-dependent transcriptome in a rapamycin-insensitive manner. Thus, mTORC1 promotes oncogenic proliferation and hepatocarcinogenesis through activation of YAP transcription by suppressing ARID1A-dependent chromatin remodeling. HCC cells with ARID1A mutation or knockout are resistant to rapamycin, suggesting that ARID1A status is a useful biomarker for mTORC1-targeted HCC therapy. The effect of ARID1A mutations on rapamycin sensitivity appears to be HCC-specific. Moreover, liver cancer cells with ARID1A deficiency are exquisitely responsive to YAP blockage, providing a potentially effective approach to targeting ARID1A-mutant HCC tumors.

No disclosures were reported.

S. Zhang: Conceptualization, data curation, formal analysis, validation, visualization, methodology, writing–original draft, writing–review and editing. Y.-F. Zhou: Data curation, investigation. J. Cao: Resources, formal analysis, writing–original draft. S.K. Burley: Formal analysis, writing–review and editing. H.-Y. Wang: Resources, formal analysis, supervision, funding acquisition, investigation, writing–review and editing. X.F.S. Zheng: Conceptualization, formal analysis, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

This work was supported by NJCCR DFHS18CRF007 (to X.F.S. Zheng). RCINJ advanced microscopy, biomedical informatics, biospecimen repository, flow cytometry/cell sorting, and metabolomics shared resources were supported by P30 CA072720. The RCSB Protein Data Bank is jointly funded by the National Science Foundation (DBI-1832184), the US Department of Energy (DE-SC0019749), and the National Cancer Institute, National Institute of Allergy and Infectious Diseases, and National Institute of General Medical Sciences of the National Institutes of Health under grant R01GM133198.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.