Loss of ARID1A activates mTOR signaling and SOX9 in gastric adenocarcinoma-Rationale for targeting ARID1A deficiency

Abstract

Background:

Gastric adenocarcinoma (GAC) is a lethal disease with limited therapeutic options. Genetic alterations in chromatin remodeling gene ARID1A and mTOR pathway activation occur frequently in GAC. Targeting the mTOR pathway in unselected patients has failed to show survival benefit. A deeper understanding of GAC might identify a subset that can benefit from mTOR pathway inhibition.

Methods:

Genomic alterations in ARID1A were analyzed in GAC. Mouse gastric epithelial cells from CK19-Cre-Arid1A^−/− and wild-type mice were used to determine the activation of oncogenic genes due to loss of Arid1A. Functional studies were performed to determine the significance of loss of ARID1A and the sensitivity of ARID1A-deficient cancer cells to mTOR inhibition in GAC.

Results:

More than 30% of cases had alterations (mutations or deletions) of ARID1A and ARID1A expression was negatively associated with phosphorylation of S6 and SOX9 in GAC tissues and patient-derived xenografts (PDXs). Activation of mTOR signaling (increased pS6) and SOX9 nuclear expression were strongly increased in Arid1A^−/− mouse gastric tissues which could be curtailed by RAD001, a mTOR inhibitor. Knockdown of ARID1A in GAC cell lines increased pS6 and nuclear SOX9 and increased sensitivity to an mTOR inhibitor which was further amplified by its combination with fluorouracil both in vitro and in vivo in PDXs.

Conclusions:

The loss of ARID1A activates pS6 and SOX9 in GAC, which can be effectively targeted by a mTOR inhibitor. Therefore, our studies suggest a new therapeutic strategy of clinically targeting the mTOR pathway in GAC patients with ARID1A deficiency.

Introduction

Despite a worldwide decrease in incidence, gastric adenocarcinoma (GAC) remains the fourth most common cancer and third leading cause of cancer-related deaths worldwide [1, 2]. In most cases, GAC is diagnosed in a late stage owing to non-specific symptoms in the early stages. Some patients with localized GAC may be cured with multimodal therapies; however, patients with metastatic GAC usually have an extremely poor prognosis, and therapeutic options for these patients are limited [3–6]. The 5-year survival rate for advanced GAC is currently less than 5% [7]. Although our understanding of the molecular underpinnings of GAC continues to improve [8–10], only 2 biomarkers (i.e., HER2 and microsatellite instability) have been shown clinical utility to guiding therapy [11], and both biomarkers are positive in only a small fraction of GAC patients. Therefore, identification of new biomarkers and effective therapeutic targets for GAC through comprehensive molecular analyses is urgently necessary.

AT-rich interactive domain 1A (ARID1A) is among the most commonly mutated genes in a broad variety of cancers. ARID1A is located on chromosome 1p36.11 and is a key component of the Switch/Sucrose Non-Fermentable chromatin-remodeling complex (SWI/SNF), which regulates diverse cellular processes, including development, differentiation, proliferation, and DNA repair [12]. Most ARID1A mutations are insertions or deletions, leading to down-regulation or loss of expression of the encoded protein [13, 14]. Moreover, recent meta-analysis indicated that the loss or low expression of ARID1A was associated with cancer-specific mortality and cancer recurrence [15]. Recent genomic profiling in GACs has reported frequent mutations or alterations in ARID1A [16]; however, the mechanisms by which its inactivation promotes tumor progression and the therapeutic significance of its loss or down-regulation in GAC need further investigation.

The activation of mTOR signaling, reflected by increased phosphor-S6 (pS6, a reliable marker for mTOR activation) [17] plays an important role in GAC cell proliferation and survival [18]. Consistently, activation of the mTOR pathway is associated with advanced stage, higher metastatic potential, and shorter overall survival. However, targeting the mTOR pathway in unselected patients with GAC has not prolonged survival [3]. Therefore, tailoring therapy with mTOR inhibitors to a selected patient group with GAC is worthy of investigation.

In this study, we found that the loss of ARID1A was significantly associated with activation of mTOR signaling and high nuclear expression of SOX9 in human GAC tumor tissues and cell lines as well as mouse gastric epithelial cells. In addition, the loss of ARID1A not only increased GAC cell migration and invasion but also sensitized GAC cells to mTOR inhibitor treatment in vitro. Moreover, inhibition of the mTOR pathway by RAD001 enhanced the cytotoxicity of fluorouracil (5-FU) in GAC cells in vitro and reduced tumor growth in vivo in ARID1A-deficient patient-derived xenograft (PDX) models. Therefore, our study suggests that the combination of classical chemotherapy such as 5-FU with mTOR pathway inhibitors is a promising strategy for the treatment of GAC with ARID1A deficiency.

Materials and Methods

Cells and reagents

The human gastric cancer cell lines AGS, MKN1, and GT5 were purchased from American Type Culture Collection and previously described [19–21]. Patient-derived cells GA0518 were each isolated from ascites or PDX tumors from a patient with metastatic GAC [22, 23]. All cell lines have been profiled and authenticated, in the Cell Line Core Facility at The University of Texas MD Anderson Cancer Center, every 6 months. The cell lines were cultured in the Roswell Park Memorial Institute Medium 1640 or Dulbecco modified Eagle medium supplied with 10% heat-inactive fetal bovine serum. The cells were maintained at 37 °C in a humidified incubator containing 5% CO₂. ARID1A knockdown cell lines (AGS ARID1A knock-down [KD] #1, AGS ARID1A KD #2, MKN1 ARID1A KD #1, and MKN1 ARID1A KD #2) were generated as previously described (Wang LL et al JCI 2020) and kindly provided by Dr. Guang Peng at MD Anderson Cancer Center. Chemotherapy agent 5-FU and mTOR inhibitor RAD001 were purchased from Sigma-Aldrich and Selleckchem, respectively. Antibodies used in this study were anti-ARID1A (HPA005456, Sigma-Aldrich) and anti–phospho-S6 (Cell Signaling Technology). Anti-SOX9 was purchased from Chemicon (EMD Millipore).

Cell proliferation assay

ARID1A-KD GAC cells and their parental/Vector control cells were treated with 0.1% dimethyl sulfoxide (as control), RAD001, 5-FU, or RAD001 combined with 5-FU at the indicated dosage for 6 days. Cell proliferation and viability were determined using the CellTiter 96 AQueous Non-Radioactive Cell Proliferation Assay (MTS) according to the instructions of the manufacturer (Promega). Briefly, cells were incubated with the MTS reaction solution for 2 h at 37 °C. The absorption at 490 nm was measured using a microplate reader. All samples were assayed in triplicate, and each assay was repeated at least 3 times. Results are presented as the percentage of control values.

CK19-Cre;Arid1a^fl/fl mouse generation and Isolation of mouse gastric epithelial cells from mouse with wild type or CK19-Cre Arid1a ^fl/fl

Mouse care and experimental protocols were approved by and conducted in accordance with the uidelines of the Institutional Animal Care and Use Committee of MD Anderson Cancer Center. We generated CK19-Cre;Arid1a1 ^fl/fl mice with targeted Arid1a^fl/fl in the gastrointestinal tract via a CK19 promoter as described previously [24].

The mouse stomach was taken and opened longitudinally. The tissue was cut into 3–5 pieces and incubated in 1.5 mM DTT-30 mM EDTA-DPBS on ice, followed by digesting in 0.8 mg/ml dispase-Hank’s balanced salt solution (HBSS) with shaking vigorously at 37C for 10 to 14 min. The tissue suspension was passed through a 70-um filter and the cells was resuspended in fresh 2 ml of complete growth medium (cGM: 5% FBS-1%-antibiotics-DMEM/F12 High glucose with supplements 0.4 μg/ml hydrocortisone, 8.4 ng/ml choleratoxin, 10 ng/ml mouse epidermal growth factor, 5 μg/ml insulin, 24 μg/ml adenine and 5 μM Y27632) in a 6-well plate with the irradiated feeder cells. The cells were cultured at 37 C with 5% CO2. The culture medium was changed every 3–5 days with 1:1 ratio of cGM versus conditional F medium.

Tumor sphere formation assay

Tumor sphere formation assay was performed as previously described [25]. Briefly, GAC cells without or with endogenous ARID1A KD (800 cells/well) were plated in triplicate onto a 6-well ultra-low attachment plate. Cells were then treated with the mTOR inhibitor, RAD001, 5-FU, or their combination as indicated. The number of tumor spheres (diameter > 100 μm) was counted under microscope after 2 weeks of culture.

Matrigel invasion assay

The invasive capability of cells was determined using Matrigel-coated invasion chambers according to a previously described protocol [26].

Colony formation assay

One day before treatment, cancer cells (AGS and MKN1 cells without or with endogenous ARID1A KD, 800 cells/well) were seeded in duplicate wells of a 6-well plate. Cells were exposed to RAD001 without or with 5-FU at various doses. Cells were then cultured for 10 to 14 days to allow colony formation. Then, cells were fixed in a 3% crystal violet/10% formalin solution. Colonies with more than 50 cells were counted, and the survival fraction was determined. All samples were assayed in duplicate.

Protein extraction and Western blot analysis

Protein isolation and Western blot analyses were performed as previously described [27], and membranes were blocked in a 5% nonfat milk for 1 hour followed by incubation with the indicated primary antibodies (arid1a, 1:750 dilution; pS6, 1:750 dilution) overnight at 4 °C. After secondary antibody incubation and washing, target proteins on the membranes were detected by chemiluminescence.

Reverse-phase protein arrays (RPPA)

RPPA analysis was performed using cell lysate from GAC cells (parental AGS and MKN1 cells and their derivatives with endogenous ARID1A KD) in the Functional Proteomics RPPA core facility at MD Anderson Cancer Center. Briefly, samples with 2-fold serial dilutions were probed by 300 antibodies arrayed on nitrocellulose-coated slides. Relative protein levels were normalized to the protein loading control and determined by interpolation of each dilution curve from the standard curve as previously described [28].

Immunohistochemical (IHC) analysis

A total of 512 patients underwent total or subtotal gastrectomy with lymphadenectomy for GAC between January 2006 and December 2008 at the first affiliate hospital of China Medical University. None of the patients had received chemotherapy of radiotherapy before the surgical procedure. All patients provided written informed consent, and the study was approved by the ethics committee at the China Medical University.

IHC staining for ARID1A, pS6, and SOX9 was performed on human tissue microarrays consisting of 512 GAC tissues as described above. Coexpression analysis of Arid1A and pS6 or Arid1A and SOX9 could be performed in 512 GAC cases (Suppleemental Table 1). Sections were incubated overnight at 4 °C with antibodies against human ARID1A (rabbit polyclonal, 1:100), pS6 (rabbit polyclonal, 1:200) and SOX9 (rabbit polyclonal, 1:200). Staining results were evaluated independently by two pathologists (XD and YW) and scored for the percentage of tumor cells with nuclear staining (0: no staining; 1: 1%−10%; 2: 10%−50%; and 3: >50%) and the intensity of nuclear staining (0: negative; 1: weak; 2: moderate; and 3: strong). Final IHC scores were determined by multiplying the intensity score with the percentage score.

Real-time polymerase chain reaction

To quantify changes in the mRNA expression levels of ARID1A, mTOR, and SOX9, real-time reverse transcription–polymerase chain reaction was performed on the ABI Prism 7900 (Applied Biosystems). Briefly, 25 μL of final reaction volume containing 1× TaqMan Universal PCR Master Mix (Applied Biosystems), 1× MultiScribe Reverse Transcriptase with RNase inhibitors, and 1× gene expression assay was used to amplify 50 ng of total RNA with the following cycling conditions: 20 sec at 50 °C, 10 min at 95 °C, 40 cycles of 95 °C for 15 sec, and 60 °C for 1 min. GAPDH (glyceraldehyde-3-phosphate dehydrogenase) was used as an internal control.

In vivo CDX/PDX mouse models

AGS GAC cells with wide type of ARID1A and two patient-derived GAC ascites cells-GA0518 and GA0825 with ARID1A deficiency (loss of ARID1A expression) were subcutaneously injected into nude mice. 1×10⁶ cells per site, two sites per mouse. Ten days after inoculation, mice were randomly grouped (5 mice/group) to receive treatment with phophate-buffered saline (as control), 5-FU (intraperitoneal, 30 mg/kg/mouse), RAD001 (oral administration, 3 mg/kg/mouse), or the combination of 5-FU and RAD001, with a treatment frequency of 5 times per week for 3 weeks. The tumor size and volume were measured as described previously [25].

Statistical analysis

The unpaired Student t-test was used to compare the treatment groups, and the Fisher exact test was used to analyze IHC results. A P value of <0.05 was considered statistically significant, and all tests were 2-sided.

Results

Alterations of ARID1A in GAC tissues

ARID1A is frequently mutated in human cancers, including GAC[29]. Consistently, through The Cancer Genome Atlas (TCGA) database analysis, we found that GAC has a much higher frequency of ARID1A gene alteration than most other cancer types do (Fig. 1A). The most common alterations of ARID1A are point mutations (Fig. 1B). It has been established that genetic alteration of ARID1A leads to loss of ARID1A expression in GAC which is associated with poor survival[30]. To explore the functional importance of ARID1A alteration and identify targets for ARID1A deficiency, we genetically KD ARID1A in two GAC cell lines, AGS and MKN1, which have relatively high expression of ARID1A, and used RPPA proteomics to explore the oncogenic signaling activated by loss of ARID1A. The heatmap in Figure 1C demonstrates that many oncogenic signals are activated upon the loss of ARID1A. Among the up-regulated pathways, mTOR pathways including S6 and Raptor were the most activated and targetable molecules. Further exploration of TCGA dataset and the RNA Seq profiles of 40 cases of metastatic GAC treated at MD Anderson revealed that ARID1A is significantly inversely correlated with expression of RPS6 and the Notch signaling targets SOX9 and HES-1. These results suggest that the loss of ARID1A may activate mTOR signaling and SOX9.

Figure 1. Alterations of ARID1A in gastric cancer tissue.

A. Genetic alteration of ARID1A in gastric malignant tumors compared with other tumor types from The Cancer Genome Atlas (TCGA). B. ARID1A mutation profile in gastric cancer from TCGA database. C. Reverse-phase protein array proteomics profile and heatmap in GAC cell lines AGS and MKN1 with ARID1A KD compared with vector-treated or parental cells to explore mTOR and other oncogenic signaling that may be activated due to loss of ARID1A. D&E. The negative association of ARID1A expression with RPS6 expression and expression of Notch signaling targets SOX9 and HES-1 from exploration of TCGA dataset and our own 40 metastatic GAC RNAseq profiles.

Gastric deletion of Arid1a results in abnormal gastric structure, activation of mTOR, and increased SOX9 in gastric tissues and epithelial cells

To investigate the loss of function of Arid1a in mouse stomachs, we generated CK19-Cre;Arid1a^fl/fl mice as described in Materials and Methods and locally induced deletion of Arid1a expression by oral gavage with activated 4-OH-tamoxifen for 5 days continuously. We observed abnormality in 100% of stomachs of CK19-Cre;Arid1a ^fl/fl mice after 1 year, and no such abnormality was noticed in wide type of mice macroscopically and microscopically (Figs. 2A and 2B). Histologic analysis revealed morphologic changes in CK19-Cre;Arid1a ^fl/fl mice, including loss of normal stomach column structure, dysplastic proliferation, metaplasia, and cystic structures (Fig. 2B, middle and right) after 6 months. These abnormal stomach structures were more pronounced after 1 year in CK19-Cre;Arid1a ^fl/fl mice (100%) but remained absent in wild-type mice.

Figure 2. Gastric deletion of ARID1A results in abnormal gastric structure, activation of mTOR, and increased SOX9 in CK19-Cre; Arid1a^fl/fl gastric tissues and gastric epithelial cells.

A. Representative gross image of mouse stomach in wild-type (WT) and ARID1A knockout mice (CK19-Cre;Arid1a^fi/fi). B. Representative hematoxylin and eosin (HE) histology of mouse stomachs from WT and CK19-Cre;Arid1a^fl/fl ARID1A-knockout mice. Scale bar 50μm. C. Dual immunofluorescence staining of ARID1A/phosphor-S6 (pS6) and SOX9/Ki67 were performed in mouse stomach tissues from WT and CK19-Cre;Arid1a^fl/fl ARID1A-knockout mice. Scale bar 20μm. D&E. Mouse gastric epithelial cells from WT and CK19-Cre;Arid1a^fl/fl mice were isolated and grown in 2D culture treated with or without RAD001 at 10nM for 48 hours and then stained with dual immunofluorescence (D) and Western blotting (E), respectively. Scale bar 10μm.

To investigate whether the activation of mTOR and increased SOX9 were due to loss of Arid1A, we performed dual immunofluorescence staining of ARID1A/phosphor-S6 (pS6) and SOX9/Ki67 in gastric tissues of CK19-Cre;Arid1a ^fl/fl and wild-type mice using their specific antibodies. As shown in Figure 2C (top), pS6 was significantly increased in gastric tissues of CK19-Cre;Arid1a ^fl/fl mice compared with those of wild-type mice, while ARID1A expression was lost in most gastric columnar cells of CK19-Cre;Arid1a ^fl/fl mice, indicating the successful deletion of ARID1A by tamofixen. Similarly, SOX9, a reported gastrointestinal stem cell marker[31], accumulated in the nuclei of gastric cells in CK19-Cre;Arid1a ^fl/fl mice compared with wild-type mice.

To further confirm these observations in mice, mouse gastric epithelial cells from wild-type and CK19-Cre;Arid1a ^fl/fl mice were isolated and grown in 2D culture. Immunofluorescence and Western blotting showed greatly increased expression of both pS6 and SOX9 in gastric cells from CK19-Cre;Arid1a ^fl/fl mice compared with that in normal gastric cells from wild-type mice. Furthermore, the activation of mTOR shown by p-S6 and expression of SOX9 in gastric cells from CK19-Cre;Arid1a ^fl/fl mice were dramatically blocked more by RAD001 compared with that from wild type mice (Figs. 2D and 2E). These results demonstrate that Arid1a deficiency results in activation of mTOR (increased pS6) and increased SOX9, which can both be blocked by an mTOR inhibitor.

Down-regulation of ARID1A increases pS6 and SOX9 expression in human GAC cells and tumor tissues

To characterize the role of ARID1A in regulating SOX9 expression and pS6, a reliable marker for the mTOR pathway activation, in human GAC cells, we used siRNA to KD ARID1A in the AGS and MKN1 GAC cell lines. KD of ARID1A in both AGS and MKN1 cells increased pS6 and SOX9 expression not only at protein levels (Fig. 3A) but also increased SOX9 at mRNA levels (Fig. 3B). In addition, significant increases in pS6 and SOX9 expression due to KD of ARID1A were confirmed using immunofluorescence in MKN1 and AGS cells respectively (Fig. 3C and Supplemental Fig.2).

Figure 3. Down-regulation of ARID1A increased pS6 and SOX9 expression in GAC cell lines and GAC tissues.

A. pS6 and SOX9 expression were determined in ARID1A knockdown (KD) GAC cell lines (AGS and MKN1) compared with the parental cells using Western blotting. B. SOX9 mRNA level was detected using quantitative real-time polymerase chain reaction (Q-PCR) in ARID1A-knockdown cell lines (AGS and MKN1) compared with their parental cells. *P<0.05. C. Levels of pS6, ARID1A, and SOX9 expression were determined in ARID1A-knockdown cell lines (MKN1) compared with control cells using immunofluorescence staining. D. pS6, ARID1A, and SOX9 expression were determined in TMA of gastric cancer tissues containing more than 400 GAC cases by immunohistochemical analysis (IHC) using their specific antibodies. Scale bar 10μm. E. Expression of ARID1A, SOX9 and phosphor-S6 was stained using dual immunofluroscent staining in PDX tumors with or without Arid1A deficiency. Scale bar 20μm.

To further determine the significance of ARID1A expression in GAC TMA and the correlation with pS6 and SOX9, a TMA containing a total of 512 patients underwent total or subtotal gastrectomy with lymphadenectomy for GAC were stained using IHC for antibodies against ARID1A, pS6 and SOX9. We found that 47 of 245 (19.2%) cases were found no ARID1A expression in normal tissues, and 373 of 512 (72.9%) tumor tissues were negative for ARID1A expression. Loss of ARID1A in tumor was found correlated with poor tumor differentiation (P=0.021) and diffused type by Lauren’s classification (P=0.022) (Supplemental Table 1). Loss of ARID1A had trend toward the short survival (Supplemental Figre 1A). In the correlation study, among 421 cases were stained both ARID1Aand SOX9, there was significantly negative association between the expression of ARID1AArid1a and SOX9 (P=0.019) (Supplemental Figure 1B, and Supplemental Table 2), while loss of ARID1A did not show negative correlation with pS6 in this GAC cohort (Supplemental Table 3). These results suggest that ARID1A plays a role in regulation of pS6 and SOX9 expression in human GACs.

Knockdown ARID1A promotes GAC cell proliferation, invasion and colony formation

To investigate the functional impact of ARID1A loss in GAC, we performed assays for cell proliferation, invasion, and migration. KD of ARID1A significantly increased invasion (Fig. 4A and 4B) and cell proliferation at 3 days and 6 days (Fig. 4C and Supplemental Fig. 3A) in both AGS and MKN1 cell lines. The colony formation and tumor sphere formation capabilities of GAC cells were also remarkably enhanced by ARID1A KD in both MNK1 and AGS cells (Figs. 4D and 4E and Supplemental Fig. 3B). These data suggest that ARID1A plays a tumor suppressor role in GAC cells and that loss of ARID1A confers aggressive phenotypes in these cells.

Figure 4. Down-regulation of ARID1A promotes GAC cell proliferation, invasion, and colony formation.

A& B. Downregulation of ARID1A promoted invasion in ARID1AKD GAC cell lines (AGS and MKN1) compared with parental cells. Invasion in different groups is quantified (low panel). **P<0.01, ***P<0.001. C. Cell growth was detected at Day 3 using MTS assay in two ARID1AKD clones in both cell lines (AGS and MKN1) compared with their parental cells as described in Materials&Methods. *P<0.05;**p<0.01. D. Colony formation assay was performed in ARID1AKD cell lines (AGS and MKN1) and parental cells. Colony numbers were calculated using Image J. **P<0.01, ***P<0.001. E. Representative image (left) and the quantification of tumor sphere numbers (right) are shown in ARID1AKD cell lines (MKN1) versus parental cells (WT). **P<0.01. Experiments were repeated 3 times. Scale bar 50μm.

RAD001 suppresses GAC cell growth and colony formation more efficiently in ARID1A-deficient cells than ARID1A wild-type cells

Because ARID1A loss increased pS6 expression, a biomarker for mTOR pathway activation in GAC, we hypothesized that cells with ARID1A KD are more sensitive to mTOR inhibitors than are cells with ARID1A intact. To test this hypothesis, we determined the effect of RAD001 in ARID1A-KD cells compared with their parential cells. We found that both pS6 and expression of SOX9 increased in ARID1A-KD MKN1 and AGS cells, while RAD001 greatly reduced pS6 and the expression of SOX9 in ARID1A-deficient MKN1 and AGS cells (Fig. 5A). Interestingly, RAD001 inhibited cell growth and colony formation more efficiently in ARID1A-KD MKN1 and AGS cells than in ARID1A wild-type cells (Figs. 5B and 5C). RAD001 effectively suppressed phosphorylation of S6 in GT5 cells and a patient-derived metastatic GA0518 cells (Fig. 5D) and their proliferation (Supplemental Fig. 4A). Furthermore, we noticed that GA0518 patient-derived cells have high pS6, nuclear SOX9 expression but loss of ARID1A expression (Fig. 5E). The combination of RAD001 and 5-FU decreased the activation of mTOR (increase in pS6) and dramatically suppressed the expression of SOX9 (Fig. 5F). These results suggest that loss of ARID1A expression sensitizes patient-derived GAC cells to treatment with RAD001 especially when in combination with 5-FU. Therefore, ARID1A deficiency should be considered for GAC patients selection in future clinical trials of mTOR inhibitors.

Figure 5. RAD001 suppresses GAC cell growth and colony formation more efficiently in ARID1A-deficienct cells than that in those ARID1A wild-type cells.

A. AGS and MKN1 cell lines with or without ARID1A knockdown (KD) were treated with RAD001 at different dosages (1 μM, 5μM) for 24 hours, and pS6 and SOX9 expression were detected using Western blotting. B. Cell growth inhibition was measured using MTS assay in AGS, MKN1, and their corresponding ARID1A-knockdown clones treated with RAD001 (RAD) at different dosages (0.5 μM, 1 μM) for 6 days, with cell growth inhibition calculated as percentage of control. Con, control treatment. **P<0.01. C. Colony formation assay was performed in AGS, MKN1, and their corresponding ARID1A-knockdown clones treated with RAD001 at 0.5 μM for 14 days, and colony numbers were calculated using imageJ. *P<0.05, **P<0.01. D. pS6 was detected using immune blotting in another GAC cell line (GT-5) and a patient-derived cell line (GA0518) from an advanced-stage gastric cancer ascites treated with RAD001 at different dosages for 24 hours (left). E. Dual immunofluorescence staining of ARID1A/ SOX9/ and ARID1A/phosphor-S6 (pS6) were performed in GA051816 cells respectively. Loss of ARID1A was observed in GA0518 cells. Scale bar: 25μm. F. pS6 and SOX9 expression were determined in GA0518 cells treated with control, RAD001 at 0.5μM, 5-FU and in combination using Western blotting according to the Materias&Methods. ARID1A KD GAC cell lines (AGS and MKN1) compared with the parental cells using Western blotting.

RAD001 synergizes with 5-FU in inhibiting growth of ARID1A-deficient GAC cells in vitro and PDX growth in vivo

To determine the effects of treatment with RAD001 alone or in combination with 5-FU on inhibition of the growth of GAC cells with or without ARID1A deficiency, we first seeded AGS and MKN1 GAC cell lines, including two ARID1A-KD clones of each and their parental cells, in 96-well plates and treated them with RAD001 alone, 5-FU alone, or combination of RAD001 and 5-FU at the indicated concentrations. Although RAD001 alone more effectively inhibited ARID1A-KD AGS and MNK1 than parental cells, the combination of RAD001 and 5-FU was the most effective in suppressing tumor cell growth, especially in ARID1A-KD clones in both AGS and MKN45 cells (Supplemental Figs. 4B&4C). Correspondingly, expression of SOX9 and pS6 were dramatically suppressed by the combination of RAD001 and 5-FU compared with either agent alone in GA0518 (Fig.5F). These results provided us the rationale for further investigating the combination treatment in vivo.

To determine the antitumor effects of RAD001 in combination with 5-FU in vivo, we us the PDX model. We implanted nude mice with GA0518 cells with ARID1A loss (Fig.3E), which easily form large tumors. We then randomized the PDX-bearing mice into four groups and treated them with phosphate-buffered saline (as control), RAD001 alone, 5-FU alone, or RAD001 combined with 5-FU. Tumor volumes were measured over the course of 3 weeks, and PDX tumor weights and mouse body weights were measured at the end of treatment (Fig. 6A). Tumor weights and tumor volumes were significantly reduced in the combination group compared with those in the single-agent groups (Figs. 6B–6D), but the mouse body weights did not differ significantly between groups (Fig. 6E). Consistent with the tumor growth inhibition, IHC staining in mouse tumor tissues indicated that the combination of RAD001 and 5-FU resulted in much lower expression of pS6, SOX9, and Ki67, a cell proliferation marker, than either treatment alone did, corroborating the inhibitory effects of the combination at the molecular level (Fig. 6F). To further determine whether inhibition of mTOR and in combination of 5-FU is more effective in the ARID1A deficient PDXs vs wide type tumors, we treated another PDX tumor (GA0825) with loss of ARID1A expression (Fig. 3E and Supplemental Fig.5A) and CDX tumors from AGS cells with wide type and high ARID1A expression. As shown in Supplemental Fig.5, the combination of RAD001 and 5-FU significantly block aggressive GA0825 tumor growth and dramatically reduced tumor volume and tumor weight, while the combination treatment in AGS cells did not show that benefit with either treatment. Thus, treatment with RAD001 in combination with 5-FU has synergistic antitumor effects in vivo in the PDX models of ARID1A-deficient GAC.

Figure 6. RAD001 synergizes with 5-FU in inhibiting tumor growth in PDX tumors in vivo.

A. The combination treatment strategy was demonstrated in a PDX model. B. 1×10⁶ patient-derived tumor cells (GA051816) were injected subcutaneously in nude mice. Mice were treated with RAD001 alone, 5-FU alone, or the combination (5 mice/group) as described in Materials and Methods. Tumor weight (B), tumor volume (C and D), and mouse body weight (E) were calculated as described in Materials and Methods. *P<0.05; **P<0.01. F. pS6 and expression of ARID1A, SOX9, and Ki67 were determined using IHC in GA0518 mouse PDX tumors in each treatment group. Scale bar, 20 μm. G. Proposed model for GAC tumors with ARID1A deficiency (loss/mutation).

Discussion

ARID1A is frequently mutated, deleted or loss expression in GAC, but the effective targeting of GAC tumors with ARID1A deficiency in the clinic has remained a challenge. In this study, by evaluating the mechanisms of ARID1A deficiency in GAC using TCGA, RNA Seq of our clinical sample database [32], and mouse and human GAC cell lines with genetic modulation of ARID1A, we demonstrated that loss of ARID1A expression in GAC was associated with mTOR activation reflected by increased phosphorylation of S6 (pS6) and increased nuclear SOX9 expression in human GAC tissues, GAC cell lines, and CK19-Cre;Arid1a^fl/fl mice. Loss of ARID1A not only increases aggressive phenotypes in GAC cells but also sensitizes GAC cells to mTOR inhibition in vivo in a PDX model with ARID1A deficiency. These data indicate that mTOR inhibitors are likely to be very effective in ARID1A-deficient GAC tumors.

ARID1A mutation/deletion, which usually causes reduction or loss of its protein expression, and Akt-mTOR pathway activation have been frequently reported in many human cancer types, including GAC[33–35]. Both ARID1A loss of function and mTOR pathway activation are associated with poor prognosis of GAC [36] [37]. In addition, increased sensitivity of GAC to AKT pathway inhibition as a result of ARID1A loss has recently been investigated in other tumor types [38, 39]. However, further validation of ARID1A loss as a cause of increased sensitivity to mTOR inhibition in GAC using a PDX model and proof of this concept in an ARID1A-deficient genetic mouse model are still of interest.

Most importantly, we confirmed the activation of mTOR and increased nuclear SOX9 expression in gastric epithelial cells of ARID1A deficient mice and GAC cell lines with ARID1A KD as well as PDX tumors with ARID1A mutation/loss (Fig.3E)but also provided strong evidence that mTOR inhibitor abolished phosphorylation of S6 and activation of mTOR in these cells. These findings provided a strong rationale for indirectly targeting ARID1A deficientcyin GACs. Interestingly, we also observed that SOX9, a gastrointestinal stem cell marker and a reported target of Notch and Hippo signaling [25, 31], was up-regulated in a conditional CK19-Cre;Arid1a^fl/fl mouse model compared with the wild type. Upregulation of SOX9 due to loss of ARID1A was confirmed in our ARID1A-KD GAC cell lines and PDX tumors with ARID1A loss or mutations (Fig.3E). The significant negative correlation of the expression of SOX9 and ARID1A was also demonstrated in the RNA sequencing data and IHC staining of a large cohort of TMA of GAC (Fig. 1E, Supplemental Figure 1B and Supplemental table 2). We therefore propose that the up-regulation of SOX9 plus activation of the mTOR pathway due to ARID1A deficiency (deletion/loss) in mice and in human GAC cell lines was responsible for abnormalities including gastric dysplasia and proliferation. Our finding on upregulation of SOX9 upon ARID1A KD is consistent with the report of Hiramatsu Y et al that deletion of ARID1A led to concomitant with Sox9 overexpression in Lgr5(+) intestinal stem cells (ISCs) restores self-renewal in ARID1A -deleted Lgr5(+) ISCs[40]. However, our observation that ARID1A negatively regulated SOX9 was not consistent with the report by Kimura Y et al. in pancreatic ductal cells that they found that ARID1A-deficient pancreatic ductal cells (PDCs) reduced expression of SOX9 and had reduced activity of the mTOR pathway[41], the incosistency among the studies may be due to possibly tissue-specific functions and different cell states (epigenomics). The detail mechanisms of SOX9 status upon loss of ARID1A warrants further validation and investigation.

Our data indicated that cytotoxic effects of RAD001 are much stronger in ARID1A-deficient than in ARID1A wild-type GAC cells and that these effects were enhanced by co-treatment with 5-FU. Moreover, the combination of RAD001 and 5-FU synergistically and significantly suppressed PDX growth more than either agent alone (Fig. 6) and had better antitumor activity in PDX tumors (GA0518 and GA0825) with Arid1A deficiency than that of AGS CDX tumors with Arid1A wide type (Fig.6&Supplemental Fig.5).. To our knowledge, our study is the first to show that RAD001 (also called everolimus) exerts strong synergistic antitumor effects when combined with 5-FU in the GAC PDX models via targeting of both mTOR activation and increased SOX9. Most importantly, our data suggest that selection of patients with ARID1A deficiency (deletion/mutation or loss of expression) may be critical for increasing the rate of response and benefit from RAD001 therapy. The phase III trial GRANITE-1 showed that the use of everolimus to treat advanced GAC did not significantly improve overall survival [42], and this lack of effect was probably due to a lack of enrichment of patients for the trial. It would be interesting to review the GRANITE-1 trial data to see whether cancer patients with ARID1A deficiency showed a significant improvement in overall survival after the RAD001 therapy.

Further study of the mechanisms of activation of Notch signaling, including SOX9 and Hes-1, due to loss of ARID1A and on the effect of a combination of Notch pathway inhibitor with an mTOR inhibitor in ARID1A-deficient GAC tumors is under way in our laboratory.

In summary, we demonstrated that ARID1A deficiency (deletion/mutation/loss of expression) in GAC tumor tissues, GAC cell lines, mouse gastric epithlieal cells as well as PDX tumors induces mTOR pathway activation and accumulation of nuclear SOX9 and thereby sensitizes tumor cells to mTOR inhibitor. Our data suggest that genetic deletion/mutation and loss of expression of ARID1A can function as a useful biomarker for selecting patients whose GAC is more likely to respond to mTOR inhibition.

What is already known on this subject?

Gastric adenocarcinoma (GAC) is a lethal disease often diagnosed at advanced stages with limited therapeutic options. Genetic alterations in chromatin remodeling gene ARID1A and mTOR pathway activation occur frequently in GAC, however targeting the mTOR pathway in unselected patients has failed to show survival benefit in the clinic.

What are the new findings?

In this study, we uncovered that the loss of ARID1A was significantly associated with activation of mTOR signaling and high nuclear expression of SOX9 in human GAC tumor tissues and cell lines as well as in gastric epithelial cells of a CK19-Cre;Arid1a^f/f mouse model which can be curtailed by RAD001, an mTOR inhibitor. Further, loss of ARID1A not only increased GAC cell migration and invasion but also sensitized GAC cells to mTOR inhibition in vitro. Moreover, inhibition of the mTOR pathway by RAD001 enhanced the cytotoxicity of fluorouracil (5-FU) in GAC cells in vitro and reduced tumor growth in vivo in an ARID1A-deficient patient-derived xenograft (PDX) model.

How might it impact on clinical practice in the foreseeable future?

Our data suggest that genetic loss or mutation of ARID1A can function as a useful biomarker for selecting patients whose GAC is more likely to respond to an mTOR inhibitor. The combination of classical chemotherapy such as 5-FU with mTOR pathway inhibitors is a promising strategy for the treatment of GAC with ARID1A deficiency.

Abbreviations:

GAC

Gastric Adnocarcinoma

PDXs

patient-derived xenografts

5-FU

5-fluorouracil

cGM

complete growth medium

CDX

Cell line derived xenograft

KD

knockdown

TCGA

The Cancer Genome Atlas

GAPDH

glyceraldehyde-3-phosphate dehydrogenase

IHC

Immunohistochemical staining

Data and materials availability:

All dataset and materials generated from this study are available for scientific community upon request.