Effect of ARID1A mutation and expression on prognosis of hepatocellular carcinoma and cholangiocarcinoma: a meta-analysis

Background: Hepatocellular carcinoma (HCC) and cholangiocarcinoma are the most common liver malignant tumors worldwide. Investigating the molecular basis of these malignancies is vital to the development of new treatments. Recent studies have found that mutation and abnormal expression of ARID1A, are frequently shown in HCC and cholangiocarcinoma. Herein, we aimed to estimate the effects of ARID1A mutation and expression on prognosis of HCC and cholangiocarcinoma. Methods: We searched PubMed, EMBASE, Web of Science, the Cochrane Library for studies evaluating the relationship between ARID1A mutations or expression and the survival of HCC or cholangiocarcinoma patients. Meta-analysis was performed to generate a combined HR with a 95% condence interval (CI) for overall survival (OS). Results: We identied 12 articles that evaluated the impact of ARID1A mutation or expression on the prognosis of patients with HCC or cholangiocarcinoma. Six studies provided data on HCC survival and 6 studies examined cholangiocarcinoma survival. For HCC, ARID1A mutation carriers or patients with low ARID1A expression had worse OS (HR = 1.75; 95% CI = 1.22–2.51) than non-carriers or patients with high ARID1A expression. For cholangiocarcinoma, ARID1A mutation carriers or patients with low ARID1A expression also had signicantly shorter OS (HR = 3.70, 95% CI = 2.88–4.76). Conclusions: Our study suggests that ARIDIA mutation or low expression is strongly associated with poor prognosis of patients with HCC or cholangiocarcinoma, and may be considered as a potential prognostic biomarker for these patients.


Background
Liver cancer is one of the most common malignancies in the world and the fourth leading cause of cancer death [1,2]. Cholangiocarcinoma, also known as bile duct cancer, originating from the biliary epithelium, is a group of heterogeneous malignant tumors with poor prognosis which ranked the second most common cancer of the liver and gallbladder worldwide [3].
ATP-dependent chromatin remodelers utilize the energy of ATP to regulate DNA transcription [4]. Next-generation gene sequencing shows that subunits of the SWI/SNF complex, encoding the ATP-dependent chromatin remodelers, are frequently mutated in a wide variety of cancers, such as lung cancer [5], rhabdoid tumors, gastric adenocarcinoma, malignant melanoma [6]. ARID1A mutations are frequently shown in SWI/SNF genes. Previous study had demonstrated that repeated mutation of ARID1A related to the development of many types of cancer. Most ARID1A mutations are inactivated mutations that lead to loss of ARID1A expression [7]. Thus, ARID1A was considered as a broad tumor suppressor [8]. Mutations and abnormal expression of ARID1A have been found in a variety of cancers, such as urothelial bladder tumors [9], colorectal cancer [10], nonfunctional pancreatic neuroendocrine tumors [11], invasive micropapillary carcinomas of the breast [12], endometriosis and uterine endometrioid carcinoma [13,14].
However, inconsistent conclusions were demonstrated that ARID1A mutation or expression has no signi cant in uence on OS [26][27][28]. In HCC, Zhao, Jie et al. reported that 87.8% tumors ARID1A mRNA and BAF250a protein expression levels were increased compared with adjacent noncancerous tissues [27]. Animal experiment showed that ARID1A had both oncogenic and tumor suppressor functions in liver cancer [29]. Therefore, to further understand the association between ARID1A and liver cancer, we perform a meta-analysis to assess the effectiveness of ARID1A mutation and expression on the prognosis of HCC and cholangiocarcinoma.

Methods
This meta-analysis followed the preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement [30].

Literature search
All relevant studies were selected from PubMed, Embase, Web of Science and Cochrane Library from the inception of the databases through July 15, 2020.
Searching terms included subject terms("ARID1A") AND ("hepatocellular carcinoma" OR "cholangiocarcinoma") and their related entry terms. We also retrieved the references of the included publications and related reviews manually.

Eligibility Criteria
The selection criteria of this study were: (a) patients recruited with proven HCC or cholangiocarcinoma according to histological results; (b) investigated the association between ARID1A mutation or expression and prognosis of HCC or cholangiocarcinoma; (c) reported the hazard ratio or 95% con dence interval or provided data for calculation. We excluded the studies as follow: (a) non-English paper; (b) animal or in vitro experiments; (c) review articles, case reports, editorials, letters, commentaries; (d) not HCC or cholangiocarcinoma; (e) studies without liver cancer overall survival data or without su cient data or Kaplan-Meier plots to calculate risk.

Study Selection And Data Extraction
Two investigators (G.-X. M. and L.-J. Y.) performed a preliminary screening on titles and abstracts of all retrieved literatures independently. Then, full texts were reviewed for eligibility. When two investigators had different opinions, disagreements were resolved by thorough discussion with another reviewer (Z.-R.D.). We extracted the following information from each article: authors' information, year of publication, country, test methods, number of patients (altered and non-altered), follow-up period, endpoints (OS) and clinical features of HCC patients included sex, cirrhosis, tumor size, histological differentiation, HBV infection. If HR and 95%CI were not provided, Get Data software was used to extract the survival information from the Kaplan-Meier curve.

Quality Assessment
The quality of each article was independently assessed by two investigators (G.-X. M. and L.-J. Y.) according to the Newcastle-Ottawa Scale (NOS). Each article can be rated with a maximum of 9 points: selection has 4, comparability has 2 and outcome has 3.

Statistical analysis
We extracted any reported relative risks (RRs), hazard ratios (HRs), odds ratios (ORs) with 95% con dence interval (CI) for OS. When adjusted and unadjusted effect sizes were both available, priority was given to the former one. The heterogeneity between the groups was examined by Q test. I 2 > 50% accompanied with p value < 0.1 was considered as a cutoff for statistically signi cant heterogeneity. The xed-effects model was initially used to obtain the aggregated HRs and 95% CI, and if substantial heterogeneity existed, the aggregated HRs were estimated using random-effects models. The publication bias was assessed through visually inspecting the symmetry of the funnel plot and evaluated with Begg's funnel plot and Egger's test. Sensitivity analysis is used to detect the stability of the results. This meta-analysis was performed using the STATA 12.0 software (Stata Corporation, College Station, TX, USA).

Study selection
We retrieved 563 papers according to search strategy. There left 384 articles after the exclusion of duplicates. Finally, 12 articles meeting the eligibility criteria were included in this study. One of them was a meeting abstract. Figure 1 summarizes the process of literature screening. The average NOS score of these studies was 8.5(Additional le 1: Supplementary Table 1), which indicated that included studies were of relatively high quality.

2.Study characteristics
The meta-analysis for HCC included six studies. The characteristics of the patient cohort are shown in Table 1. The relationship between the clinical characteristics of HCC patients and the expression level of ARID1A is shown in Table 2. The studies published between 2015 and 2020. A total of 950 patients were assimilated into our study including 195 patients with ARID1A mutation or low expression, and 755 patients with no ARID1A mutation or high expression.
Two studies focused on ARID1A gene mutations [16,17], and four focused on ARID1A expression [22,26,27,31]. In these studies, the most used methods to detect mutations and expression of ARID1A were PCR, next-generation sequencing, and immunohistochemistry.   The meta-analysis for cholangiocarcinoma included six studies. The characteristics of the patient cohort are shown in Table 3. The studies published between 2016 and 2020. We included 620 patients composed of 70 patients with ARID1A mutation or low expression and 550 patients with no ARID1A mutation or high expression. Four studies focused on ARID1A gene mutations [23,24,28,32], and two focused on ARID1A expression [25,33]. In these studies, the most used methods to detect mutations and expression of ARID1A were generation sequencing, PCR, western blotting, and immunohistochemistry.  Fig. 2). Because of the substantial heterogeneity (I² = 61%, p = 0.025), the random-effects model was used. Due to different states of altered ARID1A, six articles were divided into two groups. Two articles focused on the effect of ARID1A gene mutation on OS of HCC, combined HR was 2.39 (95% CI, 1.50-3.79). Heterogeneity was not found in two studies (I² = 0%, p = 0.687). Four articles focused on the effect of ARID1A expression on OS of HCC, combined HR was 1.51 (95% CI, 1.01-2.37). Signi cant heterogeneity was found in four studies (I² = 63.6%, p = 0.041). Besides, the results showed that the expression of ARID1A in HCC was not related to sex, cirrhosis, histological differentiation, or HBV infection, but was negatively correlated with tumor size (RR = 1.53, 95% CI, 1.18-1.99; Additional le 2: Supplementary Fig. 1).

Cholangiocarcinoma prognosis
Six studies were conducted on the effect of ARID1A mutations and expression on the OS of cholangiocarcinoma. Patients with ARID1A mutation or low expression had a signi cantly shorter OS, with a combined HR of 3.70 (95% CI, 2.88-4.76; Fig. 2). On account of slight heterogeneity without statistical signi cance (I² = 43.6%, p = 0.115), the xed effect model was used for analysis. Due to different states of altered ARID1A, six articles were divided into two groups. Four articles focused on the effect of ARID1A gene mutation on OS of cholangiocarcinoma, combined HR was 3.84 (95% CI, 2.92-5.05). Signi cant heterogeneity was found (I² = 63.2%, p = 0.043). Two articles focused on the effect of ARID1A expression on OS of cholangiocarcinoma and combined HR was 3.07 (95% CI, 1.65-5.71). Heterogeneity was not found (I² = 0%, p = 0.586).

Publication bias analysis
Slight publication bias was detected by the Egger's test in HCC studies, but there was no publication bias according to Begg's test in this meta-analysis (Fig. 3).

Sensitivity analysis
To understand how each study on the in uence of this meta-analysis, we conducted a sensitivity analysis. The result showed that no studies that have an impact on the pooled HR (Fig. 4). This con rmed the robustness of the meta-analysis.

Discussion
In our study, we estimated the effects of ARID1A mutation and expression on the prognosis of HCC and cholangiocarcinoma. The results showed that both HCC and cholangiocarcinoma patients with ARID1A mutation or low expression had signi cantly shorter OS. Our ndings suggest that ARID1A may be considered as a potential prognostic biomarker for HCC and cholangiocarcinoma patients.
Genome sequencing nds there are many common mutated genes in HCC and cholangiocarcinoma, such as TP53, CTNNB1, ARID1A [15,[17][18][19]. ARID1A is considered to be a broad tumor suppressor. However, the exact mechanism by which ARID1A abnormality leads to increased risk of HCC and cholangiocarcinoma and decreases overall survival remains largely unknown. Existing studies suggest that subunits of the SWI/SNF complex, encoding the ATP-dependent chromatin remodelers, are frequently mutated in a wide variety of cancers and ARID1A mutations are frequently shown in SWI/SNF genes. Most ARID1A mutations are inactivated mutations, resulting in the loss of ARID1A expression [7]. Because of the absence of ARID1A, abnormal SWI/SNF complex leads to a defect in enhancer activity, which impairs differentiation programs and causes widespread dysregulation of gene expression, thus promoting tumor formation [34]. Furthermore, mutations and deletions of ARID1A contribute to the proliferation and migration of HCC [29,35], and this situation also exists in the cholangiocarcinoma [25], which may lead to poor prognosis of patients suffering from HCC and cholangiocarcinoma. It is worth noting that the inactivation of ARID1A is signi cantly correlated with the activation of the PI3K/mTOR pathway, which is considered as a carcinogenic pathway by virtue of its role in cell proliferation and cell cycle control [36,37]. This is consistent with our ndings.
Our meta-analysis also integrated the clinicopathologic features of HCC patients. The results showed that expression of ARID1A was not related to sex, cirrhosis, histological differentiation, or HBV infection. However, low ARID1A expression in HCC was associated with tumor size. This further indicated that the low expression of ARID1A promoted cancer proliferation. Since only four studies included these clinical features, it is necessary to conduct more accurate detection of ARID1A expression and clinicopathological features in the future.
This study also has some limitations that cannot be neglected. Firstly, due to the number of included studies, there are not enough cases to adequately evaluate the relationship between ARID1A changes and OS with HCC and cholangiocarcinoma. Secondly, the included studies are retrospective cohort studies rather than randomized controlled studies (RCT), inevitable recall bias and selection bias still exist, which may lead to a less convincing conclusion. Owing to limited OS data, we only conducted a group of subgroup analyses. Finally, slight publication bias was detected by the egger's test in HCC studies, but there was no publication bias according to Begg's test, this may be caused by the low number of studies.

Conclusions
In summary, this meta-analysis provides evidence for the hypothesis that ARID1A mutation or low expression is associated with poor prognosis in patients with HCC or cholangiocarcinoma. Thus, ARID1A may be considered as a potential prognostic biomarker for HCC and cholangiocarcinoma patients. However, more detailed studies with more dependable data and larger sample sizes are needed to further clarify the prognostic role of ARID1A mutation and expression.

Declarations
Ethics approval and consent to participate Not applicable.