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)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

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

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

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

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

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

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

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

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

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

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

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