Loss of ARID1A expression is associated with systemic inflammation markers and has important prognostic significance in gastric cancer

The tumor suppressor gene AT-rich interactive domain 1A (ARID1A) and systemic inflammatory response (SIR) have been reported to be related to the sensitivity to immunotherapy. This study intended to explore the relationship between ARID1A expression and SIR, and to further elucidate the prognostic value of ARID1A expression in gastric cancer (GC). The mRNA and protein expression of ARID1A were detected in 272 formalin-fixed paraffin-embedded (FFPE) tumor tissues. The data of nine systemic inflammation markers were collected 1 week before gastrectomy. Univariate and multivariate COX analysis were used to screen out independent predictors of GC. Negative expression of ARID1A protein was related to GC with deficient mismatch repair (dMMR) (p = 0.033), positive programmed cell death-ligand 1 (PD-L1) (p = 0.005) and lower albumin level (p = 0.0064). Low expression of ARID1A mRNA was common in GC with abnormal E-cadherin (p = 0.020) and a higher platelet/lymphocyte ratio (PLR) (p = 0.0391). Multivariate COX analysis showed that the expression of ARID1A protein (p = 0.023), age (p = 0.004), T stage (p = 0.009) and N stage (p = 0.009) were independent predictors of GC. The nomogram established by independent predictors can accurately evaluate the survival risk of patients with GC. The loss of ARID1A protein expression was associated with the dMMR subtype and high expression of PD-L1 in GC. Negative ARID1A protein and low expression of mRNA were associated with aberrant systemic inflammatory markers. The expression of ARID1A protein had important prognostic significance in GC.


Introduction
Gastric cancer (GC) is the sixth most common type of malignancy and the fourth leading cause of global cancerrelated death (Sung et al. 2021). In recent years, various clinical trials have shown that the immune checkpoint blockade (ICB) therapy using anti-programmed cell death 1 (PD-1)/programmed cell death-ligand 1 (PD-L1) antibodies has achieved remarkable treatment outcomes in GC (Fuchs et al. 2018;Janjigian et al. 2021;Kang et al. 2017). However, the objective response rates in clinical tiral ATT RAC TION-2 (Kang et al. 2017) and KEYNOTE-059 (Fuchs et al. 2018) were 11.2% and 11.6%, respectively, which means that only a small part of patients can respond to anti-PD-1/PD-L1 antibodies. Biomarkers must be found to select patients who might benefit from immune checkpoint inhibitors (ICIs). KETNOTE-059 trial showed that pembrolizumab had a better therapeutic effect in gastric or esophago-gastric junction (EGJ) adenocarcinoma with PD-L1 combined positive score (CPS) ≥ 1 than in those with CPS < 1 [objective response rate (ORR) 15.5% and 6.4%, respectively] (Fuchs et al. 2018) and pembrolizumab was approved by the Food and Drug Administration (FDA) for the treatment of PD-L1 positive (CPS ≥ 1) recurrent gastric or EGJ cancer in 2017. KETNOTE-062 demonstrated that pembrolizumab can significantly improve overall survival (OS) in patients whose tumors express PD-L1 (CPS ≥ 10) compared with chemotherapy (17.4 months vs 10.8 months) (Shitara et al. 2018). CHECKMATE-649 has revealed that nivolumab combined with chemotherapy has superior OS than chemotherapy alone in advanced GC and EGJ cancer (13.8 months vs 11.6 months), especially when PD-L1 CPS ≥ 5 (14.4 months vs 11.1 months); therefore, nivolumab plus chemotherapy was approved by the FDA as a new standard first-line treatment for advanced GC and EGJ cancer in 2021 (Janjigian et al. 2021). These remarkable results illustrated the value of PD-L1 expression level in guiding the use of ICIs. Another research showed that patients with microsatellite instability-high (MSI-H) or Epstein-Barr virus (EBV)-positive metastatic GCs had a significantly higher overall response rate (ORR) to pembrolizumab (85.7% and 100%, respectively) (Kim et al. 2018). In KEYNOTE-061 and KEYNOTE-062, the estimated 12-month OS rates of patients with MSI-H GCs receiving pembrolizumab monotherapy were significantly higher than those of MSI-H patients receiving chemotherapy (73% vs 25% in KEYNOTE-061 and 79% vs 47% in KEYNOTE-062) (Chao et al. 2021). In a recent phase Ib/ II clinical trial (NCT02915432), high tumor mutational burden (TMB-H, TMB ≥ 12 muts/Mb) was significantly associated with better OS than low tumor mutational burden (TMB-L, TMB < 12 muts/Mb) group in advanced GC patients receiving immunotherapy (14.6 months vs 4.0 months) (Wang et al. 2019). These studies provided reliable basis for identifying biomarkers of ICB therapy. Systemic inflammatory response (SIR) was reported to have prognostic value for patients receiving ICIs (Formica et al. 2020;Jiang et al. 2018). A recent meta-analysis concluded that elevated blood neutrophil/lymphocyte ratio (NLR) before treatment was a promising prognostic biomarker for advanced cancer patients receiving immunotherapy (Jiang et al. 2018). Gastric Inflammatory Prognostic Index (GIPI) combining neutrophil/lymphocyte ratio (NLR), C-reactive protein (CRP), and albumin established by Formica et al. had significant prognostic value in metastatic EGJ/GC patients receiving ICIs (Formica et al. 2020).
AT-rich interactive domain 1A (ARID1A) is an important subunit of the Switch/Sucrose Non-fermentable (SWI/ SNF) chromatin remodeling complex. ARID1A is frequently mutated among a wide variety of tumors, such as ovarian clear cell carcinoma (40%-57%) (Jones et al. 2010;Wiegand et al. 2010), GC (18.7-23%) (Cho et al. 2019;Jones et al. 2012), hepatocellular carcinoma (10-17%) (Fujimoto et al. 2012;Guichard et al. 2012), and breast cancer (5-15%) (Liang et al. 2018;Mamo et al. 2012), and its mutations usually lead to loss of expression (Wu et al. 2014). ARID1A mutations are frequently present in EBVpositive GC (Comprehensive molecular characterization of gastric adenocarcinoma 2014), and ARID1A-negative expression was tightly correlated with PD-L1 positivity and can increase the expression of PD-L1 via activating PI3K/AKT signaling pathway (Kim et al. 2019). ARID1A deficiency has been reported to cause mismatch repair (MMR) protein dysfunction and was associated with MSI subtype and TMB-H (Shen et al. 2018). Although loss of ARID1A was considered a marker of poor prognosis (Yang et al. 2016), ARID1A alteration was an independent predictor of longer progression-free survival (PFS) after ICB therapy in many kinds of cancers (Okamura et al. 2020). Tumor infiltrating lymphocytes (TILs) are also an important factor affecting the efficacy of immunotherapy (Chen et al. 2016;Tumeh et al. 2014), and loss of ARID1A protein in both tumors and paired-adjacent normal peritumor mucosa tissues was correlated with increased TILs (Zou et al. 2020).
In this study, we aimed to figure out the relationship between the expression pattern of ARID1A and clinicopathological characteristics and other markers associated with the efficacy of ICB in gastric cancer patients. Furthermore, we evaluated the prognostic significance of ARID1A expression pattern in GC patients, and established a nomogram based on ARID1A protein expression and clinical variables to predict the survival risk of patients with GC.

Patients
We retrieved clinicopathological characteristics from patients with gastric adenocarcinoma who underwent D2 gastrectomy between August 2017 and January 2019 in Nanjing Drum Tower Hospital. Patient characteristics of age, gender, TNM stage [according to the 8th edition of the American Joint Committee on Cancer (AJCC)] (Zhao et al. 2021), lymph-node metastasis, vascular invasion, and neural invasion were included in the study. Nine systemic inflammatory markers including platelet, monocyte, and lymphocyte count; neutrophil/lymphocyte ratio (NLR), platelet/lymphocyte ratio (PLR), lymphocyte/monocyte ratio (LMR); and albumin, lactate dehydrogenase (LDH) and C-reactive protein (CRP) within 1 week prior to gastrectomy were collected.

EBV-encoded small RNA in situ hybridization
EBV-encoded RNA (EBER) was detected using automated EBER staining method with Ventana Benchmark in situ hybridization system (Ventana Medical Systems) according to the manufacturer's protocol. When > 20% of the tumor cells showed EBER stained, the case was defined as EBER positive (Jin et al. 2017).

mRNA extraction and quantitative real-time polymerase chain reaction (qRT-PCR)
Total RNA was extracted from FFPE tumor tissues according to a proprietary procedure (European patent number EP1945764-B1). After extraction and purification, the total RNA was used to generate cDNA with M-MLV Reverse Transcriptase Kit (Invitrogen, Carlsbad, CA, USA). The expression of ARID1A and β-actin (used as endogenous control) was detected by qRT-PCR using QuantStudio™ 7 Flex (Applied Biosystems) under the following conditions: one cycle at 95 °C for 20 s, 40 cycles of 95 °C for 1 s and 60 °C for 20 s. Taqman™ Fast Advanced Master Mix (Thermo Fisher Scientific, #4444556) and probes of ARID1A (Thermo Fisher Scientific, Hs00195664_m1) and β-actin (Thermo Fisher Scientific, Hs03023943_g1) were used for amplification of the cDNA.
Relative expression level of mRNA was calculated according to the formula 2 −ΔΔCt . Until now, no study has yet clearly established the grading standard for the relative expression level of ARID1A mRNA extracted from FFPE, and the loss of expression caused by ARID1A mutation has special value for research. Considering that the mutation rate of ARID1A in GC is 18.7-23%, in this study, we arranged the relative expression level of ARID1A mRNA in ascending order, and the first 20% of cases were defined as low expression of ARID1A mRNA (n = 47, 20%), and the remaining 80% of cases were defined as high expression of ARID1A mRNA (n = 193, 80%).

Classification of molecular subtypes
We divided GC into five molecular subtypes according to previous research (Setia et al. 2016): EBV-positive GC, MSI-H GC, GC with aberrant expression of E-cadherin, GC with aberrant expression of p53, and GC with p53 expression in normal. EBV status was determined via in situ hybridization as described above. Cases with aberrant MLH1 or PMS2 were defined as deficient mismatch repair (dMMR). The remaining cases were divided into E-cadherin aberrant expression subtype, p53 aberrant expression subtype, and p53 normal expression subtype according to the expression of E-cadherin and p53.

Statistical analysis
ARID1A protein expression and clinical variables including gender, age, stage, T stage, N stage, and tumor size were involved in univariate Cox analysis to select OS related variables. Significant variables in the univariate Cox analysis were further incorporated into the multivariate Cox analysis to determine independent prognostic predictors. The nomograms were established with independent prognostic predictors selected by multivariate Cox analysis to predict individual probabilities of clinical events. The Harrell's concordance index(C-index) was used to quantify the discrimination of the nomogram, and the calibration curve was used to measure the discrepancy between the predicted probabilities and the actual survival (Hanley and McNeil 1983). All statistical analyses were carried out using IBM SPSS version 25.0 (IBM, SPSS Statistics) and R version 3.6.2.

Expression of ARID1A protein was associated with mRNA level
All the patients were divided into ARID1A-negative group (n = 58, 21.3%) and ARID1A positive group (n = 214, 78.7%) by IHC (Fig. 1a, b). The mRNA expression of ARID1A was detected in 236 samples by qRT-PCR and relative expression level of mRNA was calculated according to the formula 2 −ΔΔCt . T test analysis showed that mRNA Fig. 1 Representative immunohistochemical staining images of ARID1A in GC. a Negative ARID1A staining in tumor tissues. b Positive ARID1A staining in tumor tissues. Scale bars = 100 μm. c Relative mRNA level of ARID1A in protein negative and positive group expression level of ARID1A protein negative group was significantly lower than that of protein positive group (Fig. 1c, p = 0.017).

ARID1A expression pattern and clinicopathological characteristics of patients
Clinicopathological characteristics of 272 GC patients are summarized in Table 1. The protein or mRNA expression of ARID1A was not associated with gender, age, Body Mass Index (BMI), stage, depth of invasion, lymph-node metastasis, distant metastasis, perineural invasion, or vascular invasion. However, loss of ARID1A protein expression was significantly correlated with dMMR, PD-L1 CPS ≥ 1, and CPS ≥ 10 (p = 0.033, p = 0.005 and p = 0.012, respectively). In addition, among 47 patients with low expression of ARID1A mRNA, there were 20 (42.6%) patients with PD-L1 CPS ≥ 10 and only 29.6% (56/189) patients with high expression of ARID1A mRNA were PD-L1 CPS ≥ 10, although the difference was not statistically significant.
The low expression of ARID1A mRNA was correlated with abnormal E-cadherin (p = 0.003). There was no correlation between the expression of p53, AFP, or VEFGR2 and the expression of ARID1A protein or mRNA.

Relationship between ARID1A expression pattern and systemic inflammatory markers of patients
Both SIR and ARID1A were reported to have predictive value in the prognosis of immunotherapy (Formica et al. 2020;Okamura et al. 2020), so we investigated whether there was any association between ARID1A expression and systemic inflammatory markers. ARID1A-negative protein group had a lower level of albumin (p = 0.0064, Fig. 3e), while the low mRNA group had a higher level of PLR (p = 0.0391, Fig. 3i). In addition, the LMR level appeared to be lower in the ARID1A-negative protein group, but the difference was not statistically significant (p = 0.0689, Fig. 3b). We also analyzed the relationship between systemic inflammatory markers and ARID1A expression by the Chi-square test (Table S1), and found ARID1A protein negativity was associated with high level of PLR and low level of albumin (p = 0.038 and p = 0.008, respectively, Table S1). Moreover, ARID1A mRNA was low in 17 (17.7%) of the 96 cases with low LMR, and in 30 (32.3%) of the 93 patients with high LMR, but the difference was not statistically significant (p = 0.073, Table S1).

ARID1A expression pattern was significantly associated with overall survival
Follow-up of 172 patients was carried out up to the date of death from any cause or last contact, and the median survival time was 23 months. Both negative expression of ARID1A protein and low expression of ARID1A mRNA were significantly associated with poor prognosis (p = 0.002 and p = 0.013, respectively)   (Fig. 4a, b). Univariate and multivariate Cox proportional hazard models were performed to assess the value of ARID1A protein expression pattern and clinical variables on survival (Table 2). Univariate analysis demonstrated that negative ARID1A protein expression (p = 0.002), older age (p = 0.003), higher T stage (p = 0.002), and higher N stage (p = 0.003) were associated with decreased OS. Multivariate Cox analysis indicated that negative ARID1A protein expression (p = 0.023), older age (p = 0.004), higher T stage (p = 0.009), or higher N stage (p = 0.009) were independent risk factors for poor prognosis.

Development and validation of the nomogram
To predict the survival risk of patients with GC, a nomogram was established based on all independent risk factors selected by multivariate Cox analysis (Fig. 5a). The C-index of the nomogram was 0.719 (95% CI: 0.683-0.755), which was higher than 8th edition of the AJCC staging system (C-index = 0.657, 95% CI: 0.621-0.692). In addition, the calibration plots of our nomogram showed high consistency between the nomogram-predicted outcomes and the observational outcomes (Fig. 5b, c).

Discussion
In previous studies, the mutation rate of tumor suppressor gene ARID1A was 18.7-23% in GC (Cho et al. 2019;Jones et al. 2012), and its mutations often led to loss of protein expression, which made ARID1A a poor therapeutic target.
In this study, we evaluated the protein and mRNA expression of ARID1A by IHC staining and qRT-PCR, respectively. 58 (21.3%) of 272 patients were assessed as negative ARID1A protein. We found that the expression level of ARID1A mRNA was also lower in the negative protein group. GC is a malignant tumor with complex pathogenic factors and high heterogeneity. The Cancer Genome Atlas (TCGA) divided GC into four molecular subtypes: EBVpositive subtype, MSI subtype, chromosomal instability (CIN) subtype, and genomically stable (GS) subtype (Comprehensive molecular characterization of gastric adenocarcinoma 2014). This study found that EBV-positive GC had a higher frequency of ARID1A mutation (55%), extreme DNA hypermethylation, and amplification of JAK2 and PD-L1. MSI-H GC showed elevated mutation rates and MLH1 gene silencing. GS subtype was mostly histological diffuse type with a high frequency of CDH1 and RHOA mutations. CIN subtype showed obvious aneuploidy and was mostly intestinal type or gastroesophageal borderline tumor (Cancer Genome Atlas Research Network 2014). In 2016, Setia et al. proposed a more convenient classification of GC subtypes using IHC and ISH: EBV-positive subtype, MSI-H subtype, E-cadherin abnormal subtype, p53 abnormal subtype, and p53 normal subtype GC (Setia et al. 2016). It was also found that EBV-positive GC had higher PD-L1 expression and GC with abnormal E-cadherin expression was associated with histological diffuse type. GC with abnormal p53 expression was associated with histological intestinal type. EBV-positive and MSI-H type GC had better prognosis. In our study, 272 cases with GC were divided into the above-mentioned 5 subtypes according to the detection of EBER and IHC of MLH1, PMS2, E-cadherin, and p53. We found that the negative rates of ARID1A protein in the MSI-H subtype (5/9, 55.6%) were significantly higher than those in the p53 abnormal subtype (33/164, 20.1%, p = 0.037) and the p53 normal subtype (12/63, 19.0%, p = 0.046). Among the three EBV-positive cases, one case was negative for ARID1A protein. Due to the small sample size of EBV-positive cases, we could not find a difference in ARID1A protein negative rates between EBV-positive GC and other subtypes. According to previous studies (Kim et al. 2019), loss of ARID1A protein expression was significantly associated with higher PD-L1 expression in GC, which was consistent with our findings. Considering that GC with MSI-H and positive PD-L1 expression had a better response to ICIs and that negative ARID1A protein expression was significantly related to MSI-H and PD-L1 positivity (Fuchs et al. 2018;Kim et al. 2018Kim et al. , 2019Shen et al. 2018), expression of ARID1A may become a biomarker for immunotherapy. Reduced expression of ARID1A was reported to down-regulate the transcription of CDH1 and enhance tumor invasion (Yan et al. 2014), and our study found that lower ARID1A mRNA level was related to aberrant E-cadherin expression.
Cancer-related inflammation is a key determinant of tumor progression and metastasis in most cancers (Hanahan and Weinberg 2011). With the rapid development of immunotherapy, some systemic inflammatory markers (such as NLR, PLR, CRP, albumin, etc.) have been found to be closely related to the prognosis of ICIs in different kinds of tumors (Dharmapuri et al. 2020;Formica et al. 2020;Rossi et al. 2020). ARID1A mutation has been reported as an independent predictor of longer PFS during tumor immunotherapy (Okamura et al. 2020), and is closely related to increased expression of gene signatures of immune checkpoint, cytotoxic T-cell function, and antigen presentation (Mehrvarz Sarshekeh et al. 2021). Increased expression of key genes (HAVCR2,IDO1,IL4I1,LAG3,PDCD1,PDCD1LG,and THFRSF4) which are known to be related to immune response, is also associated with ARID1A mutation (Mehrvarz Sarshekeh et al. 2021). In our study, we found that loss of ARID1A protein was associated with lower albumin (p = 0.0064) and low ARID1A mRNA was associated with higher PLR (p = 0.0689). Since both ARID1A and SIR have prognostic value in tumor immunotherapy, the mechanism of their interrelation deserves further study to provide more information for the research of ICIs' treatment biomarkers.
In clinical practice, the TNM staging system is usually used to assess the prognosis of patients with malignant tumors. However, the current staging system is not sufficient to predict prognosis. It is essential to establish more accurate clinical prognostic tools for GC. In our research, we established a nomogram based on four independent predictors (including expression of ARID1A protein, age, T stage, and N stage) for OS as a practical prognostic tool. The reasons why we chose the expression of ARID1A protein rather than mRNA expression to establish nomogram are as follows. First, IHC staining is performed more frequently than mRNA detection in clinical practice. Second, 20 of 172 cases with survival data could not be detected for RNA expression due to the degradation of RNA, which would result in a reduction in the number of modeling samples. Third, we tried to carry out univariate and multivariate cox analysis to assess the prognostic value of ARID1A mRNA expression and clinical variables of these 152 cases, and found that the expression level of mRNA was not an independent prognostic factor (p = 0.184). Therefore, ARID1A mRNA expression could not be included in the nomogram. Nomogram quantifies the risk by including all clinical and pathological variables related to the prognosis, which can be used for individualized prognosis prediction (Graesslin et al. 2010;Han et al. 2012). Our research was the first to combine ARID1A and clinicopathological characteristics to establish a nomogram for survival prediction in GC. Furthermore, we compared the predictive accuracy between our nomogram and the 8th edition of AJCC TNM staging system, and found that our nomogram was superior to the TNM staging system in predicting the prognosis of OS in GC patients. These results showed that our nomogram was relatively good at identifying high-risk populations and predicting prognosis.
There were some limitations in our study. First, in 36 of 272 samples, we could not detect the expression of the target genes (including ARID1A and β-actin) by qRT-PCR using total RNA extracted from FFPE samples due to the degradation of RNA. Second, our samples lacked the detection data of GS and CIN, so the association between ARID1A expression and TCGA subtypes could not be analyzed. Third, we were unable to obtain information on postoperative adjuvant treatments, such as chemotherapy, radiotherapy, immunotherapy, or targeted therapy, so we could not compare the relationship between ARID1A expression and prognosis in different kinds of treatments.

Conclusion
In conclusion, we found that ARID1A protein negative group had a lower level of mRNA and was significantly associated with PD-L1 positive and aberrant systemic inflammatory markers. The established nomogram based on ARID1A, age, T stage, and N stage was a practical tool for evaluating the prognosis of GC.

Acknowledgements
The authors would like to thank all patients who participated in this study.
Author contributions XW: drafting the article; KC: revising it critically for important intellectual content; TS: the acquisition of data; QL: Visualization; XX: analysis and interpretation of data; HW: analysis and interpretation of data; LY: the acquisition of data; BL: supervision and final approval of the version to be submitted; JW: conceptualization and funding acquisition.
Funding This work was funded by grants from National Natural Science Foundation of China (82073382) and the Distinguished Young Scholars of Jiangsu Province (BK20190001). The funding sources had no role in the study design, data collection, data analysis, data interpretation, or writing of this review. The funding sources have no involvement in study design; the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the paper for publication.

Data availability
The data described in this manuscript are contained in published articles or available from the corresponding author upon reasonable request.

Conflict of interest
The authors have no conflicts of interest to declare.
Ethical approval This study was conducted in accordance with the code of ethics of the World Medical Association (Declaration of Helsinki) and approved by the Ethics Committee of Nanjing Drum Tower Hospital (No. 2016-196-01).
Consent to participate Informed consent was obtained from all individual participants included in the study.
Consent to publish Not applicable.