Pan-cancer analysis of ARID1A Alterations as Biomarkers for Immunotherapy Outcomes

Abstract

ARID1A alterations would compromise mismatch repair pathway and increase the number of tumor-infiltrating lymphocytes and PD-L1 expression in some cancers, which would cooperate with immune checkpoint inhibitors (ICIs) treatment. However, a comprehensive analysis of ARID1A alteration frequency and its predictive value for ICI treatment outcome in cancers has not yet been investigated. Hence, we performed this pan-cancer analysis to evaluate the prevalence and predictive value of ARID1A alterations across >40,000 cases in multiple cancer types. We found a high frequency (6.2%) of ARID1A, which were associated with significantly higher tumor mutation burden level across various cancers. Importantly, patients with ARID1A alterations and advanced cancers had the substantially prolonged overall survival in ICI treatment cohort, suggesting it might be used to predict a survival benefit from ICI therapy across multiple cancer types. Notably, ARID1A alterations were correlated with markedly high immune infiltrates in endometrial, stomach and colon cancer. However, patients with ARID1A-mutant renal clear cell carcinoma had dramatically lower CD8⁺ T cell infiltrations than those without, indicating the association between ARID1A alterations and immune infiltrates was cancer-dependent. Collectively, our findings highlight the important value of ARID1A alterations as pan-cancer predictive biomarkers for ICI treatment.

Introduction

Although immune checkpoint inhibitors (ICIs) targeting programmed cell death protein 1 (PD-1), its ligand (PD-L1), or cytotoxic T lymphocyte antigen-4 (CTLA-4) significantly extend overall survival (OS) in patients with diverse types of cancer, it could benefit only a limited subpopulation of patients¹. Hence, there is an urgent need to investigate the rational and effective strategies to improve the efficacy of ICIs in patients with cancer. Understanding genomic correlates of response to ICIs could help the development of novel biomarkers and therapies to enhance the clinical response and expand the benefit population^{2, 3}.

The AT-rich interaction domain 1A (ARID1A) is a subunit of the chromatin remodeling complex SWI/SNF, which facilitates access of proteins to DNA. ARID1A is one of the most commonly mutated genes in cancer^{4, 5}. A recent study found that ARID1A loss and impaired ARID1A binding to the mismatch repair (MMR) protein MSH2 similarly reduced MMR and increased mutation frequency and the number of tumor-infiltrating lymphocytes and PD-L1 expression. In mice with ARID1A-deficient ovarian cancer, PD-L1 inhibitor resulted in reduced tumor volume and better survival compared with controls⁶. Then, several retrospective and exploratory studies reported the relationship between ARID1A alterations and clinical benefit to ICI in some cancers. These findings suggested that ARID1A alteration would not only cooperate with ICI treatment but also have the potential predictive value for ICI therapy. However, a comprehensive analysis of ARID1A alteration frequency and its predictive value for ICI treatment outcome in diverse cancers has not yet been investigated.

In this study, we performed this pan-cancer analysis by using online database to systematically characterize the prevalence and predictive value of ARID1A alterations across multiple cancer types. We found a relatively high frequency (6.2%) of ARID1A alterations and significant predictive value for ICIs treatment in >40,000 patients with cancers. We also investigated the association between ARID1A alterations and tumor mutation burden (TMB) level or immune cell infiltrations. The current evidence suggested that ARID1A alterations would yield promising predictive value for ICI treatment across diverse cancers.

Results and discussion

Genetic alterations of ARID1A and its association with TMB level

In this study, we defined ARID1A alterations as all kinds of nonsynonymous mutations including missense, frame-shift, splice site, nonstop, nonsense, fusions, deletions and translation start site changes. The frequency of ARID1A alterations in 40167 patients with various cancers was 6.2% (Figure 1A), with patients with endometrial cancer having the highest levels of ARID1A alterations (37.2%). Most of the alterations were missense mutations (39.6%, 979/2471; Supplemental Figure S1). The prevalence and spectrum of ARID1A alterations were slightly different in early-stage (TCGA cohort; Supplemental Figure S1A) versus advanced-stage cancers (MSK-IMPACT cohort; Supplemental Figure S1B). Co-occurring of genetic mutations in cancers with ARID1A alterations were not uncommon and some of them are popular driver genes in cancers (e.g. KMT2D/A/C, PTEN, POLE, etc.) (Figure 1B). Of note, co-occurring landscape of genetic mutations in cancers with ARID1A alterations was totally distinct in early-stage (TCGA cohort; Supplemental Figure S2A) versus advanced-stage cancers (MSK-IMPACT cohort; Supplemental Figure S2B). Considering ARID1A alterations promoting cancer mutability^{6, 7}, we then investigate the difference of TMB level between ARID1A alteration and wild type groups. We selected a subset generated from MSK-IMPACT cohort that ensure the TMB could be comparable⁸. In the MSK-IMPACT cohort⁸, 10945 samples were identified and 912 (8.3%) of them had ARID1A alterations. TMB of patients with ARID1A alterations was significantly higher than it in those without (9 vs. 4 mutations/Mb, P < 0.0001; Figure 1C). This was validated in two ICI-treated cohorts (P < 0.0001, P = 0.0012, respectively; Figure 1C and D)^{3, 9}. Notably, cancers with multiple ARID1A alterations had the highest TMB level (Supplemental Figure S2C-E). These results were consistent with a recent study, which also reported that the mutation load was significantly elevated in ARID1A-mutant tumors⁶. Together, these findings reveal a high prevalence of ARID1A alterations and its close relationship with TMB level across cancer types, suggesting that ARID1A alterations could be considered as biomarkers when conducting ICI treatment.

Figure 1.

Genetic alterations of ARID1A and its association with TMB level. A. Prevalence of ARID1A alterations in different cancer types; B. Co-occurring of genetic mutations in cancers with ARID1A alterations; C. The association between TMB and ARID1A mutations in MSK-IMPACT cohort; D. The association between TMB and ARID1A alterations in immune checkpoint inhibitors treatment cohort; E. The association between TMB and ARID1A alterations in patients with microsatellite-stable solid tumors received immune checkpoint inhibitors treatment. TMB, tumor mutation burden; Mut, mutation; WT, wild type.

Predictive and prognostic value of ARID1A alterations

Next, we evaluated the association between ARID1A alterations and clinical outcome in both whole group and ICI-treated cohort. We firstly found that patients with ARID1A alterations showed a significantly longer disease-free survival or progression-free survival [DFS/PFS, not reached vs 142 months, hazard ratio (HR) = 0.74, 95% confidence interval (CI) 0.64-0.91, P = 0.0026; Figure 2A] but markedly shorter OS (68 vs 109 months, HR = 1.30, 95% CI 1.22-1.47, P < 0.0001; Figure 2B) than those without in whole group. In early-stage cancers, ARID1A alterations were also correlated with longer DFS/PFS (P = 0.0005; Supplemental Figure S3A). However, the prognostic value of ARID1A alterations was only found in advanced-stage cancers (P = 0.0094; Supplemental Figure S3C) instead of early-stage cancers (P = 0.2086; Supplemental Figure S3B). In the ICI treatment cohort⁹, we firstly identified 1661 patients with different cancers receiving ICI therapy and 193 of them with ARID1A alterations (Supplemental Table S1). Interestingly, patients with ARID1A alterations had a substantially prolonged OS of 28 months vs 18 months in the wild-type group (HR = 0.73, 95% CI 0.61-0.93, P = 0.0092; Figure 2C). Subgroup analysis revealed that patients with multiple ARID1A alterations had the longest OS than those with single ARID1A alterations or without (not reached vs. 27 vs. 18 months, P = 0.0225; Figure 2D). Of note, we did not observe the association between ARID1A alterations and OS in patients with microsatellite-stable (MSS) solid tumors (P = 0.4075; Supplemental Figure S4A). Even in patients with multiple ARID1A alterations and MSS solid tumors, we also did not found the close association (P = 0.4899; Supplemental Figure S4B). These findings were consistent with previous publications that ARID1A alterations contributes to impaired MMR and mutator phenotype in cancer, and may cooperate with ICI treatment⁶.

Figure 2.

Predictive and prognostic value of ARID1A alterations. A. Predictive value of ARID1A alterations in all cancers; B. Prognostic value of ARID1A alterations in all cancers; C. Predictive value of ARID1A alterations in patients received ICI therapy; D. Subgroup analysis the predictive value of ARID1A alterations subtypes in patients received ICI treatment. Mut, mutation; WT, wild type.

Immune landscape of cancer with ARID1A alterations

To unravel the potential mechanism of the predictive value of ARID1A alterations for ICI treatment, we then surveyed the relationship between ARID1A alterations and six common immune infiltrates (B cells, CD4⁺ T cells, CD8⁺ T cells, macrophages, neutrophils and dendritic cells) across different cancer types¹⁰. Interestingly, ARID1A deficiency was correlated with significantly higher six immune infiltrates in most of the cancer types, especially CD8⁺ T cells, including endometrial cancer (Figure 3A), stomach carcinoma (Figure 3B) and colon adenocarcinoma (Figure 3C). However, patients with ARID1A-mutant renal clear cell carcinoma had dramatically lower CD8⁺ T cell infiltrations than those without (Figure 3D). Of note, copy number variations (either deletion or amplification) of ARID1A were associated with substantially lower six immune infiltrates in most of the cancer types including endometrial cancer, gastric cancer, colon adenocarcinoma, head and neck cancer, lung squamous cell carcinoma and breast invasive carcinoma (Supplemental Figure S5).

Figure 3.

Immune landscape of cancer with ARID1A alterations. The association between ARID1A alterations and six immune infiltrates in A. Endometrial cancer; B. Stomach carcinoma; C. Colon adenocarcinoma; D. Renal clear cell carcinoma.

Conclusions

To our knowledge, this study firstly reported a high frequency of ARID1A alterations and the predictive significance of ARID1A alterations for ICI treatment in multiple cancer types. We also observed that ARID1A alterations were associated with significantly higher TMB level. Although ARID1A alterations were correlated with significantly inferior OS in total populations, they were associated with significantly prolonged OS in ICI treatment cohort, suggesting it might be used to predict a survival benefit from ICI therapy across multiple cancer types. Notably, patients with ARID1A alterations were correlated with markedly high immune infiltrates in endometrial, stomach and colon cancer but dramatically lower CD8⁺ T cell infiltrations in ARID1A-mutant renal clear cell carcinoma, indicating the association between ARID1A alterations and immune infiltrates was cancer-dependent. Collectively, our findings highlight the important value of ARID1A alterations as pan-cancer predictive biomarkers for ICI treatment.

Supplementary Material

Supplementary figures and tables.

Abbreviations

ARID1A

AT-rich interaction domain 1A

CI

confidence interval

CTLA-4

cytotoxic T lymphocyte antigen-4

HR

hazard ratio

ICIs

immune checkpoint inhibitors

MMR

mismatch repair

MSS

microsatellite-stable

OS

overall survival

PD-1

programmed cell death protein 1

PD-L1

PD-1 ligand

TMB

tumor mutation burden