1 Introduction

Immune checkpoint inhibitors (ICIs) have generated clinical efficacy across a wide array of tumor types, including mismatch repair-deficient and microsatellite instability-high (d-MMR/MSI-H) cancers [1, 2]. However, despite the success of ICIs, the clinical responses varies among patients [3], with only 20–40% of patients benefiting from these revolutionary therapies [4, 5]. Therefore, exploring novel predictive biomarkers is critical.

AT-rich interaction domain 1A (ARID1A), a gene encoding a large nuclear protein member of the switch/sucrose non-fermentation (SWI/SNF) chromatin remodeling complex, may downregulate corresponding protein levels due to the functional mutation [6]. As a known tumor suppressor gene, ARID1A strongly regulates the DNA repair pathway, thus driving tumor formation [7]. In our previous study, ARID1A was found to serve as a novel biomarker for the prognosis and sensitivity to ICIs of advanced non-small cell lung cancer (NSCLC) [8, 9]. This finding indicates that the ARID1A mutation status has potential predictive value in immunotherapy.

In this study, we retrospectively analyzed the genomic alterations and clinical outcomes of patients harboring ARID1A mutations. Then, the predicted functions of ARID1A mutations in gastric cancer (GC) were comprehensively analyzed.

2 Materials and methods

2.1 Study population

57 tumor patients from the Affiliated Hospital of Qingdao University were retrospectively studied, including 47 patients with ARID1A mutations and 10 ARID1A wild-type GC patients undergoing immunotherapy. Each patient’s baseline information, clinical efficacy and follow-up information were collected. The clinical characteristics are shown in Supplementary Table 1. Treatment response to immunotherapy was evaluated based on the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1. Objective response rate (ORR) was calculated as complete response (CR) rate plus partial response (PR) rate under computed tomography or magnetic resonance imaging. Disease control tare (DCR) was defined as proportion of patients who had a CR, PR, and stable disease (SD) as best overall response during immunotherapy. All procedures were approved by the Ethics Committee of the Affiliated Hospital of Qingdao University. All investigations were carried out according to the rules of the Declaration of Helsinki.

2.2 Next-generation sequencing assay

Next-generation sequencing (NGS) assays from Burning Rock Co. and BGI Co. were used to analyze tissue samples. Probe hybridization and high-throughput sequencing were employed to detect the whole-exon region of 310 genes and the hot spot mutation region (including exon, intron or promoter region) of 210 genes. The assay covers single nucleotide variants (SNVs) in the target gene capture exons and short fragment insertion or deletion variants (Indels), copy number variants (CNVs), and gene rearrangements (Rearrangements/Fusions) with breakpoints within the capture range. In tissue specimen mutation detection, using 1% as the minimum threshold for determining the variant allele frequency (VAF) of a mutation. Our analysis focused on genetic mutations in SWI/SNF complex members, including ARID and SMARC family genes, DNA damage response (DDR) genes, immune-related genes, and common cancer driver genes. All variants of unknown significance (VUS) were excluded from further studies, including genes that have not yet approved by clinical practice guidelines or lack clinical evidence. We examined five microsatellite loci (BAT25, BAT26, D17S250, D2S123 and D5S346) from patients’ tumor tissue and their matched blood as controls, with 2 or more out of 5 loci mutations regarded as MSI-H, 1 and none mutated loci as microsatellite instability-low (MSI-L)/microsatellite stability (MSS).

2.3 Database and bioinformatics analysis

The cBioPortal (https://www.cbioportal.org) is a collection of multidimensional cancer genomics datasets [10, 11]. Therefore, ARID1A alterations were analyzed and visualized in GC studies and immunogenomic cohorts from the cBioPortal database [12,13,14].

MuTarget (https://www.mutarget.com/), which is a target discovery tool that connects mutation status to gene expression changes in solid tumors, was used to identify differentially expressed genes (DEGs) between ARID1A-mutant (MUT) and ARID1A-wild type (WT) GC patients [15]. Default thresholds of P < 0.01 and fold change (FC) > 1.44 were used to identify DEGs.

DAVID (https://david.ncifcrf.gov/home.jsp), a website tool that provides a comprehensive functional annotation of a group of genes [16], was used to perform Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses and DisGeNET analysis of DEGs. P < 0.05 was considered statistically significant.

The TISIDB database (http://cis.hku.hk/TISIDB/index.php) were used to explore the association between ARID1A mutations and microenvironment components [17].

2.4 Statistical analysis

Overall survival (OS) was calculated as the time from diagnosis to death. Progressive-free survival (PFS) was referred to time from immunotherapy initiation until disease progression (PD) or mortality of any cause. OS and PFS were plotted by the Kaplan‒Meier method. The curve was compared by using the log-rank test. P-values were determined by two-tailed tests. All P-values of less than 0.05 were considered significant. Statistical analyses were performed using SPSS Statistics v.25 software.

3 Results

3.1 Clinical characteristics of ARID1A-mutant tumor patients

In our cohort, 33 patients received immunotherapy, of whom 16 had gastric cancer, 4 had endometrial cancer, 3 had non-small cell lung cancer, 3 had cholangiocarcinoma, and 2 had esophageal carcinoma. The remaining 5 patients had urothelial cancer, ovarian cancer, colon cancer, breast cancer, and pancreatic cancer, respectively. In terms of gender, 21 were males, and 12 were females. The median age was 59.5 years (range 40–78 years). 21 patients (21/33, 63.64%) at an advanced stage, 10 patients were stage III, and 2 patients were stage II. The detailed clinical features of all enrolled patients are summarized in Table 1.

Table 1 Characteristics of ARID1A-mutant patients receiving anti-PD-1/PD-L1 immunotherapy (n = 33)
Full size table

3.2 PD-L1 expression, TMB, MSI status and NGS assay

Among the entire cohort, PD-L1 expression was TPS ≥ 1% in 18 patients (18/47, 38.30%). In addition, PD-L1 expression was negative (TPS < 1%) in 9 patients. In 9 cases, the CPS ranged from 1 to 10, while in 17 cases, the CPS was ≥ 10. TMB was evaluated in all patients, of whom 23 patients (23/47, 48.94%) had TMB ≥ 10, and the median TMB was 42.36 muts/MB (range 1.08–233 muts/MB). 13 patients (13/47, 27.66%) demonstrated MSI-H tumors (Table 1).

The NGS analysis of all patients was summarized in Fig. 1. 24 of 47 patients (24/47, 51.06%) had TP53 mutations. PTEN mutations were identified in 14 patients (14/47, 29.78%). ARID1B, SMARCA4, and SMARCH1, the SWI/SNF family genes, accounted for 27.66%, 12.77%, and 14.89% of the mutations, respectively. Additionally, 15 patients (15/47, 29.79%) had at least one genetic mutation related to the MMR pathway, including MLH1, MLH3, MSH2, MSH3, MSH6, PMS1, and PMS2. The detailed genetic mutation information can be found in Supplemental Table 1.

Fig. 1
figure 1

Molecular characteristics of patients with ARID1A mutations. Blue boxes indicate missense mutations, orange boxes are truncating mutations, yellow boxes present copy number gain, purple boxes are splice mutations, black boxes are deep deletion, red boxes are fusion, and green boxes are MSI-H, brown boxes are MSS/MSI-L, and gray boxes present unknow. NA Not applicable, MSI Microsatellite instability, MSS Microsatellite stability, PD-L1 Programmed cell death ligand-1

Full size image

3.3 Efficacy of immunotherapy in patients with ARID1A mutations

Among the 33 patients receiving ICIs treatment, 27 patients had assessable lesions were evaluable for response. A Waterfall plots were used to show the best observed changes in tumor size (Fig. 2a). The other 6 patients received maintenance therapy with ICIs after surgery, 1 of these patients had disease progression (PD) due to increased cancerous ascites, and none of the other 5 patients had an assessable lesion. Among the evaluable tumors, 13 patients (13/27, 48.15%) achieved a partial response (PR) and complete response (CR), and the DCR was 92.592% (25/27). Totally 7 of 27 were MSI-H, of whom 1 patients achieved a CR, 3 patient achieved a PR, 3 patients had a SD after immunotherapy.

Fig. 2
figure 2

Patients harboring ARID1A mutations have better clinical outcomes after immunotherapy. a Waterfall plot of the maximum shrinkage rate of measurable targeted lesions compared with baseline b. Patients with ARID1A mutations who received immunotherapy exhibited longer OS in our cohort (P = 0.034). c. ARID1A mutations status was not correlated with OS in pan-cancer setting by cBioPortal database analysis d. OS among patients received immunotherapy with ARID1A mutations or not from the cBioPortal database. OS Overall survival

Full size image

Until the follow-up date, patients receiving immunotherapy achieved a mOS of 51.07 months. In contrast, patients without immunotherapy had a mOS of 41.33 months. Kaplan–Meier survival analysis revealed that the non-treated ICIs group had significantly shorter survival compared to the treated group (Hazard ratio (HR) = 0.346, 95% confidence interval (CI) 0.083–1.446, P = 0.034, Fig. 2b). Additionally, analysis using the cBioPortal database showed no significant difference in OS between ARID1A-altered patients and wild-type (WT) patients (Fig. 2c). However, in a pan-cancer immunotherapy cohort, patients with ARID1A alterations had significantly longer OS, with the difference being statistically significant (P = 6.193e−3) (Fig. 2d).

3.4 Survival analysis of ARID1A-mutated malignant gastric cancer after immunotherapy

We selected GC patients for further mechanistic analysis. We retrospectively gathered data of 10 ARID1A-WT GC patients who received first-line chemotherapy plus immunotherapy (Supplemental Fig. S1). In addition, 6 GC cohorts from the cBioPortal database were analyzed for cancer genomics. ARID1A mutations were detected in 22% (179/855) of GC patients, as shown in Fig. 3a, which indicated that ARID1A mutations was one of the most frequent gene mutations in GC. In terms of the alteration types, truncation mutations were the most common, followed by missense mutations. Survival analysis showed that GC patients with ARID1A mutations had a better prognosis, as demonstrated in Fig. 3b. The disease-free survival (DFS) of ARID1A-altered GC patients was significantly longer than that of ARID1A-WT patients (84.00 months vs. 38.90 months, P = 0.0216). Moreover, compared with ARID1A-mutated patients who underwent immunotherapy combined with chemotherapy as first-line treatment, ARID1A mutations was significantly associated with increased PFS (Not reached vs 3.2 months, HR = 0.225, 95% CI 0.061–0.841, P = 0.034, Fig. 3c).

Fig. 3
figure 3

ARID1A mutations frequency and survival analysis of ARID1A-mutated GC. a. The incidence of ARID1A mutations in GC study cohorts from the cBioPortal database. b. GC patients with ARID1A mutations had longer DFS (P = 0.0216). c. Progressive-free survival of patients received immunotherapy with or without ARID1A mutations (P = 0.034). d–e. Proportion of patients responding to immunotherapy in our cohort. GC Gastric cancer, DFS Disease-free survival

Full size image

In ARID1A-WT GC from our cohort, CR were observed in 0 patient, PR in 4 patients and SD in 4 patients, 2 patients had progressive disease (PD), resulting in an ORR of 40% and a DCR of 80%. In ARID1A-mutated group, 10 advanced GC patients received chemotherapy in combination with immunotherapy as a first-line treatment, of which 3 cases achieved (30.00%) CR/PR and 7 cases (70.00%) had SD. The ORR and DCR were 30 and 100%, respectively. The difference in ORR between two group was statistically insignificant (40 vs. 30%, respectively, P = 0.1382, Fig. 3d). However, the difference in DCR showed statistical significance (80 vs. 100%, respectively, P < 0.0001, Fig. 3e).

3.5 The clinical attributes in ARID1A-mutated GC patients

We assessed the correlation between ARID1A status and clinical attributes in GC, considering the extensive influence on immunotherapy responsiveness. In our cohort, TMB value was significantly higher in GC patients with ARID1A mutations (Wilcoxon test P = 0.0088, Fig. 4a). In addition, more GC patients in the mutant group were identified with MSI-H, which suggested that these patients had more genomic instability (Fisher’s exact test P < 0.0001, Fig. 4b). The same results were corroborated in the cBioPortal database (Fig. 4c). Interestingly, the proportion of EBV-positive GC patients with ARID1A mutations was also higher than that of ARID1A-WT patients (Wilcoxon test P = 0.06, Fig. 4d).

Fig. 4
figure 4

Comparison of the clinical attributes of the ARID1A wild-type and mutated GC groups. a. ARID1A mutations is associated with TMB level in GC of our cohort. b. The associated between MSI status and ARID1A mutations c. Comparison of TMB and MSI between ARID1A wild-type and mutated groups of the cBioPortal database. d. The correlation of EBV infection with ARID1A mutations. GC Gastric cancer, TMB Tumor mutation burden, MSI Microsatellite instability, EBV Epstein-Barr virus

Full size image

3.6 Correlations between ARID1A mutations and tumor immune microenvironment (TIME)

The relationship between ARID1A and TIME was further studied. A total of 904 DEGs, including 259 upregulated and 645 downregulated genes, were subjected to functional analysis via the DAVID website. GO term analysis revealed the accumulation of DEGs in T-lymphocyte activation, immune response, and inflammatory response. Regarding molecular function, DEGs were involved in receptor binding (Fig. 5a). KEGG pathway analysis showed that the DEGs were enriched in antigen processing and presentation pathways and cell adhesion molecules (Fig. 5b). According to the DisGeNET analysis, DEGs were found to be significantly enriched in colorectal tumors, liver cancer, stomach cancer, inflammation, etc. (Fig. 5c). Furthermore, the DEGs may be involved in Epstein‒Barr virus (EBV) infection, the Wnt signaling pathway, the chemokine signaling pathway and the Hippo signaling pathway.

Fig. 5
figure 5

Functional enrichment analysis of DEGs. a–c. The results of enrichment analysis of the DEGs—namely, GO analysis, KEGG pathway analysis, and DisGeNET enrichment analysis. DEGs Differentially expressed genes

Full size image

Moreover, the correlation between the abundance of tumor-infiltrating lymphocytes (TILs) and ARID1A mutations status was analyzed in GC. Based on the TISIDB databases, we found a correlation between ARID1A mutations and the abundance of infiltrating immune cells, including activated CD8 + T cells (P = 2.52e−6), activated CD4 + T cells (P = 3.8e−11), activated dendritic cells (P = 0.000443), natural killer cells (P = 0.017), Th1 cells (P = 0.0437) and Th2 cells (P = 0.00287) (Fig. 6a). The association of ARID1A mutations with immune checkpoints was also analyzed. The results displayed that ARID1A mutations in GC was significantly associated with CD274 (P = 7e−6), PDCD1 (P = 0.0344), TIGIT (P = 0.00202), CTLA4 (P = 0.00873), HAVCR2 (P = 0.00053), and LAG3 (P = 1.82e−5) (Fig. 6b).

Fig. 6
figure 6

Correlations between ARID1A mutations and tumor immune environment in GC. a ARID1A mutations was highly associated with immune cells infiltration. b. The expression of immune checkpoint was influenced by ARID1A mutations status. GC Gastric cancer

Full size image

4 Discussion

The failure of an adequate immune response to standard immunotherapy is a major clinical challenge [18]. Some elegant studies have described ARID1A mutations may be closely related to tumor immunity and immunotherapy [19,20,21]. However, most of these studies were basic research. A retrospective clinical study from real-world is lacking, and genomic features of the ARID1A-mutated patients remains unclear.

In our study, immunotherapy significantly prolonged the survival time of patients with ARID1A mutations. Among these patients, 2 patient achieved CR, and 11 patients achieved PR. All but PD patients had varying degrees of tumor shrinkage. Moreover, we validated our results on survival benefits in an independent immune cohort from a database. Thus, identifying potential immune therapeutic options based on ARID1A alterations is promising.

In line with other reports [22], the majority of ARID1A mutations in our study were inactivating mutations that could result in the loss of ARID1A expression. Moreover, in our cohort, ARID1A mutations are often accompanied by co-mutations in genes involved in DDR pathways, including ATM, ATR, BRCA2, etc. We found MMR pathway genes were frequently mutated, with MSH6 mutations being the most common. These critical genes alterations decrease DNA repair capacity, therapy increasing tumor burden and activating immunity. Interestingly, in all patients, 48.94% patients had high TMB, and 27.66% patients were microsatellite instable. Consistent with the cBioPortal database analysis results, patients harboring ARID1A mutations were more likely to have increased genomic instability, while TMB levels were also significantly elevated. Currently, MSI-H/dMMR and TMB-H have already been recognized as positive indications for immunotherapy [23]. Thus, ARID1A mutations have potential application value for the prediction of response to ICIs. Shen et al. revealed that ARID1A inactivation hampered the recruitment of MSH2 to chromatin during DNA replication and induced dMMR in a proteomic screen [20]. In addition, deletions in ARID1A expression were associated with methylation of the promoter of the MLH1 gene [24]. These could be underlying mechanism. Furthermore, the DCR in ICIs-treated patients with MSS was 88.24% (15/17). This finding suggests that ARID1A mutations are still an effective predictor of ICI efficacy regardless of MSI status.

Previous researchers have revealed that EBV-positive tumors are related to checkpoint blockade responses [25, 26]. A possible factor is that EBV-positive cancers often exhibit amplification of the 9p24.1 locus linked to the overexpression of JAK2, CD274, and PDCD1LG2 [27]. Our findings showed that ARID1A mutations was associated with EBV infection in GC. This association implied potential benefits for ICIs therapy in some patients with ARID1A mutations in GC. Unfortunately, GC patients in our cohort were not tested for EBV status.

In pan-cancer patients, we investigated the correlation between ARID1A mutations and the immune cell microenvironment in GC. The latest literature [28] showed that, compared to HER2-positive GC patients, ARID1A mutations was significantly enriched in HER2-negative GC patients. HER2-negative GC with ARID1A mutations may be sensitive to ICIs, due to increased T-cell lymphocytosis. TILs play a vital role in the improved survival of cancers, and both the quantity and quality of TILs are possible factors in determining immune therapeutic benefits, especially T lymphocytes [29]. Moreover, NK cells were proven to be critical for the therapeutic effects of PD1 blockade [30]. Our findings confirmed that ARID1A mutations are associated with immune infiltration and immunosuppressive receptors in the TIME. GC patients harboring ARID1A alterations had a greater abundance of activated CD8 + /CD4 + T lymphocytes, DCs, and NK cells, which improved the sensitivity to ICIs. Furthermore, some immunosuppressive targets, such as PD-L1, CTLA-4, HAVCR2, LAG3, and TIGIT, had a significant expression difference in ARID1A-mutant GC. At present, accumulating studies have shown that HAVCR2 and LAG3 are valuable as potential targets for immunotherapy, and preclinical tumor models have shown that the use of immunoinhibitors to block HAVCR2, LAG3, and TIGIT restricts the growth of cancer masses [31,32,33]. Admittedly, our study has some limitations. This work is a single-center retrospective study, and multicenter prospective studies are needed to further verify the conclusions.

5 Conclusions

Based on a single-center retrospective analysis, we concluded that ARID1A mutations were often co-mutated with DNA damage response genes and could be associated with better immunotherapy outcomes in solid tumors. Bioinformatics analysis confirmed the elevated expression of immunosuppressive receptors and increased immune cell infiltration in ARID1A-mutant tumors, both of which are beneficial to the use of ICIs and affect patient outcomes. Therefore, ARID1A can serve as a novel biomarker for immunotherapy for malignant tumors, especially GC.