The prevalence of mutant ARID1 and its role in the prognosis of NSCLC
According to the datasets acquired from the cBioPortal for Cancer Genomics, ARID1 mutation is common among NSCLC patients. As shown in Fig. 1A, the mutation frequencies of the subunits of ARID1 including ARID1A and ARID1B were 7% and 4% in NSCLC patients, respectively. As far as we are concerned, gene mutations or hypermethylation lead to low ARID1 protein expression [24–25]; therefore, we further investigated the relationship between the survival of NSCLC patients and the expression of the ARID1 protein. Figure 1B and Fig. 1C describe the relationships between the disease-free survival (DFS) or OS of NSCLC patients and the expression of ARID1A or ARID1B. As shown in the figure, ARID1A and ARID1B both are convincing biomarkers for NSCLC prognosis with compelling efficiency, and ARID1A or ARID1B deficiency was significantly related to the poor prognosis of NSCLC (ARID1A [DFS: P < 0.0001; OS: P < 0.0001]; ARID1B [DFS: P = 0.0045; OS: P < 0.0001]).
ARID1A or ARID1B mutation correlates with an improved outcome for ICI treatment
The relationship between ARID1A or ARID1B mutation and the outcome of ICI treatment was then studied. Through systematic analyses of the datasets from the cBioPortal for Cancer Genomics, we found that both ARID1A and ARID1B mutations were associated with an improved outcome for ICI treatment in advanced NSCLC patients. As shown in Fig. 2A, more responders (CR + PR + SD) were confirmed in the mutant-type (MT) group than in the wild-type (WT) group for patients harboring ARID1A (50% versus 19%, P = 0.045) or ARID1B (50% versus 16%, P = 0.034) mutations. Figure 2B displays the median PFS (mPFS) with ICI treatment for the two groups. Patients harboring ARID1A (6.8 months versus 5.5 months, P = 0.313) or ARID1B (10.0 months versus 5.4 months, P = 0.032) mutations benefited more from treatment and achieved a longer PFS time than those in the WT group. Survival analyses for ICI treatment were then performed as shown in Fig. 2C. Compared with those in the WT group, patients harboring mutant ARID1B achieved significant survival benefits with treatment (P = 0.031). However, although a trend toward a difference existed for the survival curve of patients harboring mutant ARID1A, no statistical significance was found.
Establishment of a prognostic nomogram for the prognosis of ICI treatment in NSCLC
Eighty-six patients of the selected NSCLC cohort with integrated information on clinical features, targeted sequencing and PD-L1 expression evaluated by immunohistochemistry (IHC) were involved in the construction of the novel nomogram. First, univariate analyses were performed to identify variables to include in nomogram construction. As shown in Fig. 3A to Fig. 3F, multiple variables were confirmed to be significantly associated with the prognosis of ICI treatment, including EGFR mutation (P = 0.021), ARID1B mutation (P = 0.024), PD-L1 expression (P = 0.010), TMB (P = 0.012), treatment lines (P = 0.003) and smoking history (P = 0.007). Through the univariate analyses, we found that patients with mutant ARID1B, elevated PD-L1 expression (≥ 50% percentage positive staining), a high TMB value (≥ 75th percentage) or a history of smoking could benefit from ICI treatment, while patients harboring mutant EGFR might not derive survival benefits from ICI treatment. In addition, first-line administration of ICIs in advanced NSCLC patients might be a better choice than later administration. The nomogram based on these variables was then established as shown in Fig. 3G. In total, 3 types of patient information, clinical information, pathological information and genomic signatures, were included in the nomogram. Through this novel nomogram, physicians could easily obtain a score based on the Cox regression model for each variable listed in the graph, and then the total number would be assessed as the sum of all variable scores. Therefore, the survival risks of ICI treatment for advanced NSCLC patients could be quantified before treatment. The C-index for this prognostic model was 0.71, which suggests that the model has a relatively robust ability to predict the PFS of advanced NSCLC patients treated with ICIs. The calibration plots shown in Fig. 3H indicated that the probabilities of our prognostic model agreed with the accuracy probabilities on acceptable scales (the dashed lines in the calibration plots correspond to a 10% margin of error).
The mutation status of ARID1A or ARID1B is associated with the TMB, PD-L1 expression and TIME of NSCLC
Based on the research above, we reasonably deduced that ARID1A or ARID1B mutation serves as a novel biomarker for ICI treatment and could have a connection with characteristics that are proven to be associated with sensitivity to cancer immunotherapy. As shown in Fig. 4A, ARID1A or ARID1B mutations were associated with a higher TMB value (ARID1A: 16.2 versus 9.3, P = 0.001; ARID1B: 17.1 versus 9.4, P = 0.020) and a higher proportion of PD-L1-positive cells (ARID1A: 38.9% versus 12.9%, P = 0.040; ARID1B: 41.3% versus 12.4%, P = 0.020) in advanced NSCLC patients. Figure 4B reveals the relationships between the TIME and the expression of ARID1A or ARID1B. As shown in the figure, ARID1A or ARID1B expression was related to immunosuppression in 517 lung adenocarcinoma (LUAD) samples and 501 lung squamous cell carcinoma (LUSC) samples, which is mainly characterized by significant reductions in the abundances of activated CD8 + T cells and activated dendritic cells (DCs). This result suggested that ARID1A or ARID1B deficiency might change the TIME via the activation of the antigen presentation process and cellular immunity and thus contribute to the change in the sensitivity to ICI treatment.