Loss of ARID1A expression promotes lung adenocarcinoma metastasis and predicts a poor prognosis

Abstract

Background

ARID1A is an essential subunit of SWI/SNF chromatin remodeling complexes. ARID1A gene mutations and loss of ARID1A expression have been observed in a variety of cancers, and to be correlated with invasion, immune escape and synthetic lethality. As yet, however, the biological effect of ARID1A expression and its role in the prognosis of lung adenocarcinoma (LUAD) patients have remained unclear. In this study we aimed to further elucidate the role of ARID1A expression in LUAD in vitro and in vivo and to assess its effect on the clinical prognosis of LUAD patients.

Methods

ARID1A expression was detected by IHC in tissue samples from LUAD patients. After regular culturing of LUAD cell lines and constructing stable ARID1A knockdown lines, wound healing and Transwell assays were used to assess the role of ARID1A in cell migration and invasion. The effect of ARID1A knockdown on metastasis was verified in vivo. Western blotting was used to examine the expression of target proteins. Univariate and multivariate analyses were performed to assess survival and to provide variables for nomogram construction. In addition, we used the “rms” package to construct a prognostic nomogram based on a Cox regression model.

Results

We found that ARID1A expression serves as an effective prognostic marker for LUAD patients. Loss of ARID1A expression correlated with a poor prognosis, as verified with a nomogram based on a Cox regression model. In addition, we found that ARID1A knockdown promoted LUAD cell proliferation, migration and invasion in vitro and enhanced LUAD metastasis in vivo by activating the Akt signaling pathway.

Conclusions

Our data indicate that loss of ARID1A expression promotes LUAD metastasis and predicts a poor prognosis in LUAD patients.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13402-021-00616-x.

Introduction

Switch/sucrose nonfermenting (SWI/SNF) chromatin remodeling complexes perform essential roles in a variety of biological processes, including DNA replication, gene expression and cell differentiation [1, 2]. In addition, they have been found to be dysregulated in various cancer types [3]. A variety of subunits of the SWI/SNF chromatin remodeling complexes, including AT-rich interactive domain 1 A (ARID1A), ARID1B and DPF2 in the canonical BRG1/BRM-associated factor (BAF) complex, ARID2 and BRD7 in the polybromo-associated BAF complex (PBAF) and GLTSCR1/GLTSCR1L and BRD9 in the newly characterized noncanonical complex (ncBAF), have been characterized in previous studies [4, 5]. BAF, PBAF and ncBAF represent the three final forms of assembled SWI/SNF chromatin remodeling complexes [6]. ARID1A is a key component of the SWI/SNF chromatin remodeling complexes that can bind DNA in a non-sequence-specific manner and are involved in the processes of DNA repair and stabilization [1, 7]. Alterations in ARID1A may be diverse and have been observed in a variety of cancer types, including urothelial carcinoma [8], gastric cancer [9] and lung cancer [3, 6]. It is well established that loss of intact BAF, especially functional loss of ARID1A, endows cells with cancerous functions and leads to a poor prognosis in multiple cancer types.

Lung cancer ranks first among all malignancies in cancer-related mortality, and the 5-year overall survival (OS) is less than 20 % in China [10]. Non-small-cell lung cancer (NSCLC) represents nearly 85 % of all lung cancer cases [11]. Loss of ARID1A has been observed in NSCLC tissues, and as a tumor suppressor, ARID1A deficiency has been found to promote proliferation and to inhibit apoptosis in NSCLC cells via the Akt signaling pathway [2]. As yet, however, it remains to be resolved whether ARID1A deficiency correlates with NSCLC invasion and/or metastasis, including its putative underlying mechanism. Moreover, the relationship between ARID1A expression and the prognosis of NSCLC patients needs further confirmation.

According to recent studies, loss of ARID1A expression may endow primary tumor cells with a metastatic tendency via a series of molecular alterations [12]. It has e.g. been reported that ARID1A, which is usually highly expressed in primary tumors, was downregulated in metastatic lesions of liver cancer, and multiple metastatic suppressors were found to be downregulated due to ARID1A deficiency [13]. We performed this study to clarify the role of ARID1A deficiency in promoting the metastasis of lung adenocarcinoma (LUAD) as well as to clarify its potential underlying molecular mechanism. In addition, we evaluated the relationship between ARID1A expression and the prognosis of LUAD patients by constructing a prognostic nomogram based on a Cox regression model.

Methods

Patients included in the study

Seventy-one young lung adenocarcinoma (LUAD) patients post-surgery, who were admitted to our center between March 2013 and June 2016, were included in this study, and tissue samples of patients were collected. Also, basic characteristics of the patients, including age, sex, smoking history, family history and treatment information, were collected. All patients were diagnosed with resectable invasive LUAD (TNM stage: I-III) in our department of pathology after hematoxylin and eosin (H&E) staining and immunohistochemical (IHC) evaluation. IHC slides for the evaluation of ARID1A expression were collected after obtaining informed consent. The disease status of each patient was evaluated using the standard of Response Evaluation Criteria in Solid Tumors 1.1 (RECIST 1.1), and all tumors were staged according to the 2019 American Joint Committee on Cancer (AJCC) TNM staging system for lung cancer [14]. After surgery, all patients received standard treatment according to the guidelines of the National Comprehensive Cancer Network (NCCN) for lung cancer. The Ethics Committee of the Affiliated Hospital of Qingdao University approved this study, and all investigations were carried out according to the rules of the Declaration of Helsinki. All experiments were carried out following the National Health and Family Planning Commission of the Professional Regulation Commission (PRC) guidelines.

Immunohistochemistry (IHC) assay

Five-micrometer-thick sections were cut from paraffin-embedded tissues for subsequent IHC analyses. Antigen retrieval was performed by boiling the slides in 10 mM citrate buffer (pH 6.0) for 10 min, followed by cooling at room temperature for 20 min. Each section was incubated at 4 °C with primary antibodies at appropriate concentrations [ARID1A: 1:500; pan-Akt: 1:250; phosphorylated-Akt (p-Akt): 1:100] overnight. Two investigators independently evaluated the IHC slides. Five fields of each slide were selected for the evaluation of IHC scores. Given that ARID1A is localized in the nucleus, the proportion of cells with nuclear staining and the strength of staining were evaluated. We followed a previously reported scoring method [2]. The intensity of staining was scored as 0 (no staining), 1 (weak), 2 (medium) and 3 (strong). Percentage scores were assigned as 0 (< 5 %), 1 (5–25 %), 2 (26–50 %), 3 (51–75 %) and 4 (76–100 %). Images were scanned using a NanoZoomer slide scanner (NanoZoomer-XR C12000, Hamamatsu) and viewed using NDP.view software (NDP.view2 U12388-01, Hamamatsu). The final score of each slide was calculated as the average score of the 5 fields selected randomly and ranged from 0 to 12 (intensity score x percentage score). Specifically, low expression of ARID1A was defined as a final score less than 4 (IHC score < 4). Antibody information is listed in Table S1 (supplementary file 1). As a positive control for ARID1A IHC analyses we used human kidney tissue, while as a negative control normal human lung tissue was used. The negative and positive controls used for pan-Akt and p-Akt were the same as described in the instructions for the respective primary antibodies.

Construction of a prognostic nomogram

Univariate analyses were performed with Kaplan-Meier plots (KM plots) to incorporate variables into the Cox regression model to construct a nomogram based mainly on ARID1A expression. We used the “rms” package of R software version 3.1.2 (The R Foundation for Statistical Computing, Vienna, Austria) to construct the nomogram. Harrell’s C-indexes ranging from 0.5 (no discrimination) to 1 (perfect discrimination) were used to verify the discrimination ability of the nomogram [15], and visual calibration plots were used to verify the calibration ability of the nomogram [16]. Bootstrapping with 1000 resamples was used for these analyses. The area under the curve (AUC) of the ROC curve was used to assess the efficiency of each variable in predicting the prognosis of LUAD.

Cell lines and construction of stable infectants

LUAD cell lines A549 (ATCC No.: CCL-185), HCC4006 (ATCC No.: CRL-2871), NCI-H1299 (ATCC No.: CRL-5803), HCC2279 (ATCC No.: CRL-2870) and NCI-H3255 (ATCC No.: CRL-2882) were purchased from the cell bank of the Chinese Academy of Sciences (Shanghai, China). Information for the STR Cell ID assays is depicted in supplementary file 2. All cell lines were cultured in RPMI-1640 medium supplemented with 10 % fetal bovine serum (FBS) and 1 % P/S (100 IU/ml penicillin and 100 IU/ml streptomycin) in a 37 °C humidified atmosphere under 5 % CO₂. Lentiviral vectors encoding ARID1A short hairpin RNAs (shRNAs) and a vector control were purchased from GeneChem (Shanghai, China). Using a helper solution (GeneChem, Shanghai, China) according to the manufacturer’s instructions, the cells were infected with the lentiviruses, after which the infection efficiencies were verified by fluorescence microscopy. Cell counting (as preliminary evaluation of the infection efficiency) revealed that over 90 % of the cells expressed fluorescent protein, which was considered as appropriate efficiency. Subsequently, changes in ARID1A expression were examined using Western blotting. Stably infected cell strains were selected for seven days and cultured with 1 µg/ml puromycin (Solarbio, Beijing, China). The sh-ARID1A and the vector control sequences are listed in Table S1 (supplementary file 1).

Western blot analysis

Whole protein lysates of cells scratched from culture dishes were collected using RIPA cell lysis reagent containing proteinase and phosphatase inhibitors (Solarbio) at 4 °C for 30 min. Next, the cell lysates were centrifuged at 12,000 g for 20 min at 4 °C, and the protein concentrations of the supernatants were determined using a BCA protein assay reagent kit (Beyotime, Shanghai, China). The supernatants containing total proteins were subsequently mixed with the corresponding volume of 5× SDS loading buffer and heated at 95 °C for 5 min. Twenty milligrams of total protein from each sample was separated by SDS-PAGE and transferred to 0.22 μm nitrocellulose (NC) membranes. The NC membranes were blocked with 5 % nonfat dry milk dissolved in TBST and incubated overnight with primary antibodies at the appropriate dilutions (ARID1A: 1:1000; Akt: 1:2000; p-Akt: 1:2000; β-actin: 1:4000). After being washed with TBST solution three times for a total of 30 min, the NC membranes were incubated with HRP-conjugated secondary antibodies on a shaker for 2 h at room temperature. An ECL reagent (Pierce, Rockford, IL, USA) was used to visualize the results. Primary antibody information is listed in Table S1 (supplementary file 1).

MTT viability assay

Every cell line was seeded (5,000 cells/well) into three 96-well plates overnight to allow adherence. After 24, 48 or 72 h of incubation, 20 µl MTT solution (5 mg/ml) was added to each well, after which the cells were incubated at 37 °C for another 4 h. Next, we discarded the culture medium and added 150 µl DMSO solution (Solarbio, > 99.5 % methyl sulfoxide) to each well. Finally, we measured the absorbance at 570 nm using an ELISA plate reader to build cell proliferation curves.

Scratch wound healing assay

We performed a wound healing assay to evaluate the migration abilities of the cells. Wounds were generated in 6-well plates by scratching the surface with a 200 µl pipette tip. A light microscope (Nikon, Tokyo, Japan) was used to photograph the wounded areas after 0, 24 and 48 h incubation. The wound healing results are presented as ratios calculated using the following formula: change in the scratch area/change in the scratch area of the vector control at 24 h.

Transwell migration and invasion assays

Eight-micrometer pore Transwell compartments (Corning, NY, USA) were used for cell migration and invasion assays. For the migration assays 2 × 10⁴ cells were seeded into serum-free medium in the upper compartment and incubated for 36 h in a 37 °C humidified atmosphere under 5 % CO2. For the invasion assays Matrigel (BD Biosciences, San Jose, CA) was added to each well, after which 6 × 10⁴ cells were seeded and incubated for 72 h. After incubation, the migrated/invaded cells in the lower chamber of the Transwell were fixed with formalin solution (Solarbio, 10 % neutral buffered formalin) and stained with 0.5 % crystal violet for 20 min at room temperature. The average number of migrated/invaded cells from five random views (at 40× magnification) was calculated using light microscopy (Nikon, Tokyo, Japan).

In vivo xenograft models

Four- to six-week-old female BALB/c nude mice and CB-17 SCID mice were purchased from Beijing Vital River Laboratory Animal Technology (Beijing, China). Fourteen BALB/c nude mice and 10 CB-17 SCID mice were used to construct metastatic xenograft models. After 2 weeks of acclimatization, all mice were injected via the tail vein with either 8 × 10⁵/200 µl A549^{vector control} or A549^{sh − ARID1A} cells suspended in serum-free DMEM. After 8 weeks, all the mice were sacrificed, after which the lung tissues were embedded in paraffin for H&E or IHC staining. The animal experiments were carried out by strictly following the guidelines of the Committee on the Ethics of Animal Experiments of Qingdao University.

Bioinformatic and statistical analyses

Survival analyses were performed online using the Kaplan-Meier plotter tool [17]. In addition, we divided LUAD patients from The Cancer Genome Atlas (TCGA) into an ARID1A-high group and an ARID1A-low group and, subsequently, performed gene set enrichment analysis (GSEA) to explore potential altered signaling pathways between the two groups of patients to guide subsequent experiments. All statistical analyses were performed using GraphPad Prism 8.0 software (GraphPad, La Jolla, CA, USA), and Student’s t tests were used to determine statistical significance. P values were determined by two-tailed tests, and p < 0.05 was used to define statistical significance. The log-rank test was used for univariate analyses, and the Cox regression model was used for multivariate analyses.

Targeted sequencing and gene panels

Alterations in ARID1A, epidermal growth factor receptor (EGFR) or anaplastic lymphoma kinase (ALK) expression were confirmed by next-generation sequencing (NGS). NGS details have been reported before [18]. Here, tissue samples from 71 young LUAD patients were analyzed by NGS using three capture-based targeted sequencing panels, as listed in Table S2 (supplementary file 1). The detailed procedure of targeted sequencing based on the three panels is depicted in supplementary file 1 (supplementary material and methods for targeted sequencing). ARID1A (GeneID: 8289), EGFR (GeneID: 1956) and ALK (GeneID: 238) were the primary genes studied.

Results

Prevalence of ARID1A mutations in LUAD and relationship with ARID1A expression

ARID1A is the most frequently mutated subunit of the SWI/SNF chromatin remodeling complexes [19]. To explore the correlation between ARID1A mutations and loss of ARID1A expression, NGS was performed to identify ARID1A mutations among our cohort of patients. According to datasets from the cBioPortal for Cancer Genomics [20, 21], we found that approximately 6 % of LUAD patients (94/1709) harbored ARID1A mutations (Fig. 1A), and that ARID1A mutations are likely associated with a poor LUAD differentiation (p = 0.0002; Fig. 1B). Young LUAD patients in our cohorts exhibited a significant difference compared to the whole LUAD population, i.e., we found that 11.3 % of the young patients harbored ARID1A mutations (Fig. 1C. Among the mutation types, missense mutations were most common, accounting for 55.6 % of them (Fig. 1D). Other types of ARID1A mutations, such as coding sequence deletions (CDS-del; in 22.2 % of patients), CDS insertions (CDS-ins; in 11.1 % of patients) and frameshift mutations (in 11.1 % of patients), were also detected among the young LUAD population. ARID1A mutations were found to be associated with a poor OS (38.37 months versus 54.34 months, p = 0.0147) in LUAD patients (Fig. 1E). By comparing ARID1A expression assessed by IHC and confirming ARID1A mutation status by NGS, we found that patients harboring ARID1A mutations were more likely to lose ARID1A expression or to maintain a lower expression level than patients without ARID1A mutations (p = 0.04; Fig. 1F). Detailed information on the young LUAD patients harboring ARID1A mutations and the corresponding expression levels evaluated through IHC are listed in Table 1. Interestingly, 62.5 % of young LUAD patients harboring ARID1A mutations exhibited concomitant EGFR mutations (Table 1). In addition, we found that ARID1A deficiency may be associated with an increased incidence of EML4-ALK fusion alterations in young LUAD patients (14.0 % versus 9.52 %).

Fig. 1.

Prevalence of ARID1A mutations and correlation between ARID1A mutations and loss of ARID1A expression. A: Mutation status of ARID1A among lung adenocarcinoma patients from the cBioPortal for Cancer Genomics; B: Comparison of histology grade between 2 groups of patients; C: Proportion of ARID1A mutations in our cohort of young lung adenocarcinoma patients; D: ARID1A mutation types in young lung adenocarcinoma patients; E: Relationship between overall survival of lung adenocarcinoma patients with ARID1A mutation status; F: Correlation between ARID1A mutations and loss of ARID1A expression in our LUAD patient cohort. (LUAD: lung adenocarcinoma)

Table 1.

The Correlations between ARID1A mutations and expression

Patient ID	ARID1A mutation type	Protein change	ARID1A expression (IHC score)	EGFR status	ALK status
1	missense	p.V1982I	low (3)	Wild type	Wild type
2	deletion	p.P16[6 > 5]	low (1)	Exon 19 deletion	Wild type
3	deletion	p.P16[6 > 5]	low (0)	Exon 19 deletion	Wild type
3	insertion	p.Q1327[8 > 9]	low (0)	Exon 19 deletion	Wild type
4	missense	p.G1254S	low (2)	Exon 21 L858R	Wild type
5	frameshift	p.K1059Gfs*45	low (3)	Wild type	Wild type
6	missense	p.R1383Q	high (4)	Exon 21 L861Q	Wild type
7	missense	p.M50V	low (2)	Wild type	Wild type
8	missense	p.M1141V	low (2)	Exon 18 G719X; Exon 20 S768I	Wild type

Loss of ARID1A expression promotes migration and invasion of LUAD cells

As we found that ARID1A mutations and ARID1A deficiency are frequently present in LUAD tissues and genomic alterations in ARID1A are related to a low expression of ARID1A and a poor prognosis of LUAD patients, we next set out to investigate altered cell properties induced by ARID1A deficiency in vitro. We examined the expression levels of ARID1A in five LUAD cell lines (A549, HCC4006, NCI-H1299, HCC2279 and NCI-H3255) and selected A549 and HCC4006 for subsequent experiments due to their high expression of ARID1A and EGFR mutation status (Fig. 2A). Among these two selected cell lines, the reported EGFR mutation status was confirmed (Fig. S1; supplementary file 1). A549 is a cell line with wild-type EGFR (wt-EGFR), while the HCC4006 cell line harbors an EGFR mutation [22]. Next, stable ARID1A knockdown cell strains were generated by lentivirus infection. Fluorescent images of the infection process are shown in Fig. 2B, while the structure of the lentiviral construct is displayed in Fig. S2. ARID1A knockdown was confirmed by Western blotting (Fig. 2A). To better understand the correlation between ARID1A knockdown and changes in biological behavior, a series of experiments was performed. Using proliferation curves, we found that loss of ARID1A function significantly enhanced LUAD cell proliferation (Fig. 2C). Significant differences were observed between the two groups after 24 h (A549: 0.94 versus 0.79, p < 0.0001; HCC4006: 0.78 versus 0.64, p < 0.0001), 48 h (A549: 1.47 versus 1.02, p < 0.0001; HCC4006: 1.10 versus 0.83, p < 0.0001) and 72 h (A549: 1.56 versus 1.17, p < 0.0001; HCC4006: 1.18 versus 0.82, p < 0.0001) incubation of the two cell lines as evaluated through the OD-570 nm value. Regardless of cell proliferation ability, we found that both the invasive and metastatic abilities of the cancer cells were markedly enhanced after ARID1A knockdown in vitro (Fig. 2D and E). Using wound healing assays, significant changes in cell migration were noted. After 24 h of incubation with HCC4006-sh-ARID1A, a significant reduction in the scratched areas was observed compared with those of the vector control (3 versus 1, p < 0.0001), while both A549-sh-ARID1A (6 versus 3, p < 0.0001) and HCC4006-sh-ARID1A (6.5 versus 2, p < 0.0001) showed markedly enhanced migration abilities after 48 h incubation (the area change in the vector control group after 24 h culture was set at 1). Both A549-sh-ARID1A (invasion assay: 9.33 cells versus 2.00 cells, p = 0.0117; migration assay: 8.67 cells versus 0.33 cells, p = 0.0009) and HCC4006-sh-ARID1A (invasion assay: 11.33 cells versus 0.67 cells, p = 0.002; migration assay: 13.00 cells versus 1.33 cells, p = 0.0006) migrated to the lower chamber more efficiently than the corresponding vector controls with or without Matrigel. These results indicate that loss of ARID1A expression enhances the degree of malignancy of the LUAD cells tested and changes the biological behavior of these cells. Moreover, changes in biological behavior were observed in cell lines harboring either wt-EGFR or mt-EGFR. Thus, loss of ARID1A may trigger LUAD progression regardless of the EGFR mutation status.

Fig. 2.

Loss of ARID1A expression promotes invasion and migration of LUAD cells. A: ARID1A expression levels in different human LUAD cell lines and after ARID1A lentiviral knockdown in A549 and HCC4006 cells; B: Fluorescence images of lentivirus infection; C: Proliferation curves of A549 and HCC4006 cell lines; D: Scratch wound healing assays for A549 and HCC4006 cell lines; E: Migration and invasion assessment of A549 and HCC4006 cell lines using Transwell assays

ARID1A knockdown enhances LUAD metastasis in vivo via the Akt signaling pathway

To further clarify the role of ARID1A knockdown in promoting LUAD metastasis, we performed in vivo experiments with the stably infected cell lines. Xenograft metastatic models were constructed in two types of mice. The results obtained from the in vivo experiments suggest that loss of ARID1A expression significantly enhances the metastatic ability of the LUAD cells tested. Representative images of gross tissues and the preliminary pathological results of the two groups are displayed in Fig. 3A1 and A2, respectively. Differences in the numbers of metastatic nodules in both BALB/c-nu mice (p = 0.0278) and CB-17 SCID mice (p = 0.0075) (evaluated by a specialist in pathology) were statistically significant (Fig. 3B, C). H&E staining was used to diagnose metastatic nodules, and IHC staining was used to measure the expression of target proteins (Fig. 3G). We found that the expression of ARID1A in the tumor tissues was significantly downregulated in the lentivirus-treated group compared with the vector control group. In addition to the differences in the number of metastatic nodules, both the size of the metastatic nodules (diameter > 300 μm: 51 versus 6; diameter > 1,000 μm: 4 versus 0) assessed with NDP.view software (NDP.view2 U12388-01, Hamamatsu) and the percentage of bilateral lung metastases (57.1 % versus 14.3 %) were increased in the sh-ARID1A groups in both types of mice (Fig. 3D, F). We further examined the expression of pan-Akt and p-Akt. After IHC staining of metastatic tissues we found that the phosphorylation level of Akt was markedly upregulated in the sh-ARID1A group. Through GSEA based on LUAD patients derived from the TCGA database (Fig. 4A) we found activation of the potential Akt signaling pathway (nominal p value = 0.0040). To verify activation of this signaling pathway by ARID1A knockdown, we performed in vitro experiments to estimate the expression levels of these two proteins (Fig. 4B). We found that ARID1A knockdown significantly upregulated the expression of p-Akt in A549 and HCC4006 cells. Quantitative Western blot results are displayed in Fig. 4C.

Fig. 3.

Loss of ARID1A expression promotes metastasis of LUAD cells in vivo. A1-A2: Representative images of lung tissues from two groups of mice; B: Number of metastatic nodules in the lungs from two groups of BALB/c-nu mice; C: Number of metastatic nodules in the lungs from two groups of CB-17 SCID mice; D: Number of metastatic nodules with a diameter > 300 μm in the lungs from two groups of mice; E: Number of metastatic nodules with a diameter > 1000 μm in the lungs from two groups of mice; F: Percentage of bilateral lung metastases in two groups of mice. G: H&E staining and IHC images of ARID1A, pan-Akt and phosphorylated Akt in two groups of mice (50× and 200×)

Fig. 4.

Activation of the Akt signaling pathway is induced by loss of ARID1A expression. A: GSEA based on LUAD patients derived from the TCGA database; B: Western blotting verification of pan-Akt and phosphorylated-Akt expression after ARID1A knockdown; C: Quantitative Western blotting results

Role of ARID1A expression in the prognosis of LUAD patients

Genomic ARID1A alterations may be associated with changes in its expression and a poor LUAD prognosis. Loss of ARID1A expression was found to be associated with proliferation, migration, invasion and metastasis through in vitro and in vivo experiments. Next, clinical cases were employed to further clarify the role of ARID1A deficiency in the prognosis of LUAD patients. Seventy-one young LUAD patients were enrolled in this study, with ages ranging from 27 to 45 years. Genomic alterations in EGFR and ALK were determined using NGS, as detailed in the supplementary materials and methods section (supplementary file 1). We found that 43 young LUAD patients (60.56 %) harbored EGFR mutations, and that 9 patients (12.68 %) harbored ALK rearrangements. Two patients (2.82 %) harbored concomitant EGFR mutations and ALK rearrangements. However, the EGFR or ALK status did not demonstrate compelling efficiency in predicting the prognosis of young LUAD patients (p = 0.6070). Next, based on expression of the ARID1A protein, we divided our cohort of young LUAD patients into 2 groups, i.e., positive or negative. IHC images of positive and negative ARID1A staining are shown in Fig.  5A1 and A2, respectively. Univariate analyses were subsequently performed to determine useful variables in predicting patient prognosis. TNM stage (p < 0.0001) and ARID1A expression (p = 0.0036) were found to possess significance in predicting the prognosis of young LUAD patients (listed in Table 2). KM plots of ARID1A among our cohort of patients are shown in Fig. 5B. We also investigated the difference in ARID1A expression between patients with progressive disease or stable disease and found that ARID1A exhibited significant expression loss in patients with progressive disease (p = 0.0003; Fig. 5C). To evaluate the efficacy in predicting the prognosis of LUAD, the AUC of ARID1A for disease-free survival (DFS) of young LUAD patients was calculated. We found that ARID1A expression (AUC = 0.81, p < 0.001) served as an efficient variable in predicting the prognosis of our cohort of LUAD patients. Given the particularity of our cohort of young LUAD patients, OS data were not available.

Fig. 5.

Loss of ARID1A function predicts a poor prognosis of LUAD patients. A1-A2: IHC images of ARID1A (50× and 200×); B: Survival analyses in our LUAD patient cohort according to ARID1A expression; C: Difference in ARID1A expression between progressive and stable groups; B: Progression-free survival of lung adenocarcinoma patients predicted by ARID1A; C: Overall survival of LUAD patients predicted by ARID1A. (LUAD: lung adenocarcinoma)

Table 2.

Basic information on patients and survival analyses

Variable	Total (n = 71) patients (%)	Stable disease (n = 53)	Progressive disease (n = 18)	Univariate analysis	Multivariate analysis	Months of DFS (95% CI)
age						NA
> 45	0 (0)	0	0	NA	NA
≤ 45	71 (100)	53	18
sex						NA
male	17 (23.94)	12	5	NA	NA
female	54 (76.06)	41	13
TNM stage
I-II	46 (64.79)	41	5	< 0.0001	0.003	41.70 (37.71 to 45.68)
III	25 (35.21)	12	13			32.00 (23.17 to 40.83)
EGFR or ALK mutation						NA
positive	52 (73.24)	38	14	0.6070	0.0550
negative	19 (26.76)	15	4
pathology type						NA
adenocarcinoma	71 (100)	53	18	NA	NA
other	0 (0)	0	0
smoking history						NA
smoker	7 (9.86)	4	3	0.3540	0.4520
nonsmoker	64 (90.14)	49	15
ARID1A status
low expression	40 (56.34)	34	16	0.0036	0.0020	33.80 (28.82 to 38.78)
high expression	31 (43.66)	29	2			48.95 (44.21 to 53.70)

1. DFS: disease-free survival;

2. CI: confidence interval;

3. NA: not applicable

4. EGFR or ALK mutation: EGFR 19 del, EGFR 21 L858R or EML4-ALK rearrangement

Since our cohort consisted of only young LUAD patients, we decided to extend the survival analyses to all LUAD patients listed in the TCGA database. By doing so, we found that ARID1A expression [progression-free survival (PFS): p < 0.0001; OS: p < 0.0001] still showed a robust efficacy in predicting the prognosis of LUAD patients (Fig.  5D1-D2). In summary, our data indicate that ARID1A may serve as a novel biomarker for the prognosis and treatment of LUAD.

Role of ARID1A expression in LUAD prognosis based on hazards models

Multivariate analyses, of which the variables included ARID1A expression, TNM stage, genomic EGFR or ALK status and smoking history (based on the Cox regression model), were used to introduce variables into the nomograms. Eventually, two variables, ARID1A expression (p = 0.0020) and TNM stage (p = 0.0030), were used to construct the nomograms, as shown in the multivariate analyses in Fig. 6A and C. We constructed two types of nomograms based on ARID1A expression to predict DFS for young LUAD patients. Nomogram 1 provided an efficient method for determining the possibility of a relatively short DFS (DFS < 40 months) after surgery. The total score could be easily determined through the sum of the scores of two variables, after which the corresponding possibility could be calculated from the total score. Through nomogram 2, physicians could accurately calculate the possibilities of DFS shorter than one year or three years. The C-index of our prognostic model was 0.88, which suggests that our prognostic model has a robust ability to predict DFS for young LUAD patients. The calibration plots of nomogram 1 (Fig. 6B), the one-year DFS plots of nomogram 1 (Fig. 6D) and the three-year DFS plots of nomogram 2 (Fig. 6E) indicated that the probabilities of our prognostic models agreed with the accuracy probabilities on acceptable scales (the dashed lines in the calibration plots correspond to a 10 % margin of error). The basic information of the patients involved and the survival analyses are summarized in Table 2.

Fig. 6.

Nomograms based on ARID1A expression for predicting disease-free survival of young LUAD patients. A: Nomogram 1 for the disease-free survival of our cohort; B: Calibration plot of nomogram 1 (the curves above the nomogram represent the density distribution of the patient cohort for each subgroup of variables); C: Nomogram 2 for the disease-free survival of our cohort; D: Calibration plot for 1-year disease-free survival of our cohort; E: Calibration plot for 3-year disease-free survival of our cohort. (DFS: disease-free survival)

Discussion

Survival analyses based on the TCGA database and our cohort of patients suggest that ARID1A may serve as a tumor suppressor in LUAD. Loss of ARID1A predicts a poor prognosis in LUAD patients, especially in young LUAD patients. As such, ARID1A may function as a novel biomarker for LUAD treatment. The newly developed nomograms, with robust efficiency, could predict DFS in young LUAD patients after surgery and provide physicians with an accurate method for establishing a follow-up plan or a post-surgery treatment regimen. We confirmed the relationship between ARID1A expression and the prognosis of LUAD patients and, thereby, provide a direction for further research on SWI/SNF chromatin remodeling complexes in lung cancer.

ARID1A mutations, which are diverse, frequently result in loss of ARID1A function, and they have been detected in a variety of cancers. In this study, we found that ARID1A mutations may result in a low expression of ARID1A protein regardless its mutation type. Furthermore, we found that ARID1A was more frequently mutated in young LUAD patients than in the overall LUAD population, which may explain the different biological behaviors between the two groups of patients. Our results strongly suggest that ARID1A mutations are associated with loss of ARID1A expression. Previous studies have convincingly demonstrated a relationship between functional loss of ARID1A and activation of the PI3K/Akt signaling pathway in gastric cancer [9], ovarian cancer [23, 24] and endometrial cancer [25]. Here, we confirmed an increase in the phosphorylation level of Akt induced by loss of ARID1A expression in LUAD in vitro and in vivo. Concordantly, the Akt signaling pathway was found to be significantly activated after ARID1A downregulation. By combining previous results and our data, we conclude that ARID1A serves as a regulator of the Akt signaling pathway and that loss of ARID1A expression may lead to activation of the Akt signaling pathway, thereby contributing to changes in the biological behavior of LUAD tumors. The role of ARID1A loss in promoting cell proliferation and migration has been confirmed in multiple types of cancer, including osteosarcoma [26], endometrial cancer [27], nasopharyngeal carcinoma [28], liver cancer [29] and clear cell renal cell carcinoma [30]. The latter study suggested that ARID1A mutations and the resultant expression loss are associated with clinicopathologic parameters [31]. As yet, however, few studies have focused on the mechanism of ARID1A loss in lung cancer. Here, we clarified the role of ARID1A loss in the promotion of LUAD cell metastasis in vitro and in vivo, regardless of the EGFR mutational status.

According to our in vitro experiments, the proliferation, migration and invasion abilities were markedly enhanced by loss of ARID1A expression in LUAD cells. In the subsequent in vivo experiments, we found that the quantity and size of pulmonary metastatic tumors and the percentage of bilateral lung metastases of xenografted metastatic cell models were significantly elevated in the sh-ARID1A group, which strongly supports the idea that ARID1A plays a role in governing the metastasis of LUAD. Loss of ARID1A expression in LUAD upregulates the phosphorylation level of Akt and ultimately promotes metastasis by activating the Akt signaling pathway. All of these changes enable cancer cells to distantly metastasize and adapt to a new target organ. All our experimental results were verified in cell lines with either wt-EGFR or mt-EGFR. As the Akt signaling pathway is one of the downstream signaling pathways of EGFR, we conclude that loss of ARID1A expression enhances oncogenic functions by directly activating its downstream signaling pathway. As a result, ARID1A deficiency may lead to failure of targeted therapy, which was also concluded in a previous study [30]. Overall, our research provides insight into the mechanism of LUAD metastasis, which may begin with ARID1A mutations followed by loss of ARID1A expression. Subsequent activation of the Akt signaling pathway may change the biological behavior of LUAD cells, including metastasis to a new site.

The Akt signaling pathway plays an important regulatory role in multiple cancer-related features. The phenotype induced by loss of ARID1A expression may vary in different cell types. Next to promotion of the proliferation and metastasis of cancer cells, loss of ARID1A expression is thought to be associated with sensitivity to immune checkpoint inhibitors (ICIs). Recent studies indicate that functional loss of ARID1A may lead to upregulation of the expression of programmed death-ligand 1 (PD-L1), and to an elevated tumor mutational burden (TMB) by increasing somatic mutations and interacting with mismatch repair (MMR) proteins [32], ultimately contributing to sensitivity to immunotherapy in gastric cancer [33] and lung cancer [6]. Further studies are needed to identify the phenotypic changes induced by ARID1A deficiency in LUAD cells.

Several limitations to our research exist, including the small sample size of young LUAD patients and the lack of in-depth information on the molecular mechanism underlying activation of the Akt signaling pathway induced by loss of ARID1A expression. Also, the role of ARID1A in the OS of young LUAD patients remains to be elucidated. Besides, whether ARID1A expression has an effect on altering the genomic status of EGFR or ALK remains unclear. Although further studies are needed to fully explore the underlying mechanism activated by loss of ARID1A expression, ARID1A may act as a tumor suppressor in LUAD and may serve as a novel biomarker for evaluating the prognosis and treatment of lung cancer.

Abbreviations

SWI/SNF

Switch/sucrose nonfermenting

ARID1A

AT-rich interactive domain 1A

ARID1B

AT-rich interactive domain 1B

BAF

BRM-associated factor

PBAF

Polybromo-associated BAF complex

ncBAF

Newly characterized noncanonical complex

NSCLC

Non-small cell lung cancer

EGFR

Epidermal growth factor receptor

ALK

Anaplastic lymphoma kinase

NGS

Next-generation sequencing

IHC

Immunohistochemistry

AUC

Area under the curve

DFS

Disease-free survival

FBS

Fetal bovine serum

shRNA

Short hairpin RNA

ICI

Immune checkpoint inhibitor

PD-L1

Programmed death-ligand 1

TMB

Tumor mutational burden

MMR

Mismatch repair

Author contributions

Conception/Design: Helei Hou and Hong Li;

Provision of study material or patients: Helei Hou, Dantong Sun and Han Zhao;

Collection and/or assembly of data: Dantong Sun, Tiantian Bian, Tianjun Li and Kewei Liu;

Data analysis and interpretation: Dantong Sun, Lizong Feng;

Manuscript writing: Helei Hou and Dantong Sun and Yan Zhu;

Final approval of manuscript: All authors.

Funding

This work was funded by the Qilu Sanitation and Health Leading Talents Cultivation Project (to Helei Hou), the Chinese Postdoctoral Science Foundation (2017M622143 to Helei Hou) and a Qingdao Postdoctoral Application Research Funded Project (2016052 to Helei Hou).

Declarations

Conflicts of interest

The authors declare no conflicts of interest.