The present study investigated ARID1A gene mutation in kidney cancers on the Genome Data Commons (GDC). The results showed that ARID1A mutation was one of the top-ten mutated genes most common found in kidney cancers together with PBRM1 (Polybromo 1), VHL (von Hippel-Lindau), TP53 (tumor suppressor 53), SETD2 (SET domain-containing protein 2), etc. Among these, mutations of PBRM1 and VHL are the majority to cause ccRCC subtype [13, 14]. Mutation of MET proto-oncogene and inactivation of SWI/SNF chromatin/histone modifiers are common in pRCC type I, whereas mutations of FH (fumarate hydratase), SETD2, and CDKN2A (cyclin dependent kinase inhibitor 2A) are more common in pRCC type II [13, 14]. Chromophobe RCC is associated with mutations of PTEN (phosphatase and tensin homolog), TP53, and TERT (telomerase reverse transcriptase) mitochondrial genes [13, 14]. Consistent with our present study, a recent study by Ricketts and colleagues  has shown that mutations of SWI/SNF complex genes, including PBRM1, ARID1A and SMARCA4, were found notably in ccRCC (47.1%), pRCC (53.0%) and chRCC (14.9%). It was thus plausible that ARID1A mutation might get involved in development and/or progression of RCC.
We further investigated mutation profiles of ARID1A gene in patients with kidney cancers. From a total of 32 mutations found, the majority came from frameshift (14/32) and missense (14/32), whereas only 4 had stop gained mutations. The mutation profiles occurred along the protein sequence of ARID1A, including the two conserved domains (ARID DNA-binding domain and SWI/SNF-like complex subunit BAF250/Osa domain on the C-terminus) of the Pfam structure. Human ARID1A, also known as BAF250a, comprises 2,285 amino acids with a molecular mass of approximately 250 kDa and is a component of the SWI/SNF chromatin remodeling complex that regulates gene transcription . The ARID DNA-binding domain (at residues 1019-1104) has been reported to play an important role in selective binding to AT-rich sites of DNA and assembly of polymorphic BRG-/BRM-associated factor (BAF) and polybromo-associated BAF (PBAF) complexes . The C-terminal region domain of ARID1A (at residues 1976-2231) is essential for stimulating glucocorticoid receptor (GR)-mediated transcriptional activation . Structural modeling has shown that ARID1A mediates GR interaction via the SWI/SNF complex that facilitates the transcriptional activation by GR . Recurrent mutation and protein loss of ARID1A gene are most frequently found in ovarian clear cell carcinoma (mutation = 46-57 % and protein loss = 41-59 %) and endometrial endometrioid carcinoma (mutation = 39-44 % and protein loss = 19-34 %) [19-21]. Due to these mutations found in ARID1A, the structural and functional complexes of ARID1A protein and SWI/SNF would be expected to be affected in cancer patients.
Although three common predictive and prognostic markers, including VHL, VEGF (vascular endothelial growth factor) and CAIX (carbonic anhydrase IX) have been validated in RCC, their uses in clinical routine are still regarded as controversial . Recently, P4HB (Prolyl 4-hydroxylase, beta polypeptide)  and five-proteins panel (APC; adenomatous polyposis coli, NOTCH1; neurogenic locus notch homolog protein 1, EYS; protein eyes shut homolog, filamin A and ARID1A)  have been proposed as novel, potential diagnostic and prognostic biomarkers for kidney renal clear cell carcinoma. In this study, we found that the patients with mutated ARID1A had a shorter overall survival than those without ARID1A mutations. The results from OSkirc web tool, which is a reliable and user-friendly analysis tool , also confirmed that the patients with low ARID1A expression had a significantly worse prognosis than those with high ARID1A expression. Consistently, several other studies have also reported that the patients with mutations or loss of ARID1A had a shorter disease-free survival and/or poorer overall outcome/survival [25-27]. Our results were in concordance with those reported previously indicating that ARID1A is an important tumor suppressor gene in RCC and may also serve as a prognostic biomarker in RCC patients.
Globally, most of the RCC cases are affected by three main RCC subtypes, including ccRCC (75%), pRCC (15%), and chRCC (5%) . Sarcomatoid differentiation is distinct in a high-grade and poorly differentiated components and is found in only 5% of RCC . In this present study, we recruited a variety of RCC subtypes and grades, including ccRCC (grade I), ccRCC (grade II), chRCC, pRCC and sRCC from patients with age of > 45 years following the recent classification system and histopathological descriptions . To our knowledge, this is the first evidence demonstrating ARID1A protein expression in sRCC.
Our histopathological data showed that ARID1A is localized mainly in the nuclei of lymphocytes, fibroblasts, intraglomerular cells and renal tubular epithelial cells in the non-neoplastic areas. By contrast, nuclear ARID1A expression was diminished in the cancer cells, especially in ccRCC and chRCC, whereas no significant changes were observed in pRCC and sRCC. Recent studies have demonstrated that nuclear ARID1A protein was generally found in lymphocytes, fibroblasts and endothelial cells in various normal tissues [30-32]. Therefore, expression of ARID1A in these cells is frequently used as a positive control to compare with that of cancer cells [30-32]. However, the precise mechanisms of the defective ARID1A expression in these cancer cells have not been thoroughly investigated. A variety of extracellular matrices, stromal cells (such as fibroblasts, mesenchymal cells, pericytes, endothelial cells and lymphatic vascular networks) and immune cells (such as lymphocytes, natural killer cells and tumor-associated macrophages) participate in enhancing cancer cell survival, growth, invasion and metastasis [33, 34]. A recent evidence has suggested that cooperation of ARID1A and PIK3CA mutations enhanced proliferation of ovarian surface epithelial cells isolated from mice carrying mutant alleles by overproduction of interleukin-6 . It is thus possible that ARID1A may help in the prevention of an inflammation-driven tumorigenesis in mouse ovarian clear-cell carcinoma . Therefore, the interactions between the tumor microenvironment (defined as the interplay between the nascent cancer cells and their surrounding environment) and ARID1A expression in these cancer cells and their effects on cancer initiation, progression and metastasis should be further investigated.
In our present study, the decrease/loss of ARID1A expression was commonly found in the nuclei of the cancer cells, especially in ccRCC and chRCC, when compared with the paired adjacent non-cancer tissues. Such loss or decrease could be explained by mutations and molecular/epigenetic variations because the majority of the ARID1A mutations were the inactivation by nonsense and frameshift mutations throughout the gene that cloud lead to the loss of ARID1A protein expression [1, 21, 36]. Alternatively, in-frame indel (insertion-deletion) mutations of ARID1A affecting the nuclear export signal could disturb the stability of ARID1A protein expression, like nonsense and frameshift mutations. Degradation of abnormal nuclear ARID1A is regulated by the nuclear ubiquitin-proteasome system . Furthermore, epigenetic silencing might be also involved [4, 5]. ARID1A gene promoter hypermethylation is recognized as the main etiologic factor for the decreasing ARID1A mRNA expression in many cancers. It has been shown that decreasing ARID1A mRNA expression is controlled by a repressive histone modification (H3K27Me3) in invasive breast cancers . Promoter hypermethylation has been demonstrated to be related with low ARID1A mRNA expression in squamous cell carcinoma cell lines . In kidney cancers, unmethylated CpG promoter of ARID1A gene has been reported . However, the epigenetic analysis of ARID1A gene in renal cell carcinoma is still rarely conducted and further studies are needed to address the precise role of epigenetics in regulation of ARID1A expression.
Currently, there is no clearly specific method for investigating ARID1A protein expression although some interpretation systems, such as the three staining grades  and Histo-score , had been used. In this study, we employed the Allred scoring and grading system to quantify ARID1A protein expression in the renal tissues. This strategy was also applied in a recent study evaluating ARID1A expression in breast cancer tissues . The Allred scores of ARID1A protein expression in cancer tissues, particularly ccRCC (grade I), ccRCC (grade II) and chRCC, were significantly lower than in the paired non-cancer tissues, whereas those of pRCC and sRCC subtypes had no significant changes. Interestingly, a previous report has shown a similar results indicating that ARID1A protein expression was significantly decreased mostly in ccRCC . In addition, the four-grade system allowed us to discriminate the RCC patients with negative to weak expression from those with moderate to strong expression. Previous studies have reported associations of the decrease/loss of ARID1A expression with clinical factors, such as the higher nuclear grade, larger tumor sizes, higher pTNM stage, and presence of metastatic of cancer in ccRCC patients [7, 8]. In this study, we showed that the decrease/loss of ARID1A expression was significantly associated with ccRCC (grade II) and chRCC subtypes, presence of comorbidity, and low eGFR levels. As a result, the RCC patients with negative to weak ARID1A expression tended to have lower progression-free survival, consistent with another study in cervical cancer demonstrating that the patients with ARID1A loss had significantly lower survival .
In this study, statistical analysis by a binary logistic regression model and the Gaussian model was also performed to assess the prognostic variable to a progressive-free survival of patients. Univariated analysis revealed that negative ARID1A expression, high AJCC stage (II-IV), and presence of chronic kidney disease were significantly associated with the progressive-free survival, but negative ARID1A expression was not an independent prognostic factor for patients with RCCs in multivariate analysis. However, due to the small sample size, the prognostic significance of ARID1A expression in RCCs requires clarification in further studies, using the Cox proportional hazard model with adequate sample size.
In summary, we have shown herein that ARID1A was one among the most common mutated genes found in kidney cancers. Immunohistochemistry demonstrated that ARID1A protein expression was markedly decreased in the RCC cancer tissues (particularly ccRCC and chRCC) as compared to the adjacent non-cancer area. Additionally, level of ARID1A protein expression in the adjacent non-cancer renal tissues was significantly lower in ccRCC (grade II) than in ccRCC (grade I). Moreover, all of the RCC cases with negative ARID1A protein expression (3/11 cases with ccRCC (grade II) and 2/6 chRCC cases) had metastasis 1−50 months after the surgical removal. Finally, progression-free survival tended to be shorter in RCC patients with negative to weak ARID1A expression. Taken together, these data implicate that the defect/loss of ARID1A expression is associated with poor prognosis and metastasis of RCC and thus may serve as the prognostic marker of RCC, particularly ccRCC and chRCC subtypes.