© 2011 Nature America, Inc. All rights reserved. © 2011 Nature America, Inc. All rights reserved. NATURE GENETICS VOLUME 43 | NUMBER 12 | DECEMBER 2011 1219 Gastric cancer is a heterogeneous disease with multiple environmental etiologies and alternative pathways of carcinogenesis 1,2 . Beyond mutations in TP53, alterations in other genes or pathways account for only small subsets of the disease. We performed exome sequencing of 22 gastric cancer samples and identified previously unreported mutated genes and pathway alterations; in particular, we found genes involved in chromatin modification to be commonly mutated. A downstream validation study confirmed frequent inactivating mutations or protein deficiency of ARID1A, which encodes a member of the SWI-SNF chromatin remodeling family, in 83% of gastric cancers with microsatellite instability (MSI), 73% of those with Epstein-Barr virus (EBV) infection and 11% of those that were not infected with EBV and microsatellite stable (MSS). The mutation spectrum for ARID1A differs between molecular subtypes of gastric cancer, and mutation prevalence is negatively associated with mutations in TP53. Clinically, ARID1A alterations were associated with better prognosis in a stage-independent manner. These results reveal the genomic landscape, and highlight the importance of chromatin remodeling, in the molecular taxonomy of gastric cancer. Recent studies using next-generation sequencing (NGS) have revealed an extensive repertoire of potential cancer-driving genes in several cancer types 3–17 . To further explore the genetic basis of gastric cancer, we performed whole-exome capture using the Agilent SureSelect Human All Exon kit followed by NGS on the Illumina Genome Analyzer II× or HiSeq 2000 platforms to iden- tify somatic mutations in 22 matched pairs of gastric cancer and normal tissue (Supplementary Table 1), with a mean depth of 116× and 91.4% of bases covered to at least 10× (Supplementary Table 2). Somatic mutations were predicted using algorithms described in the Supplementary Note, and an extensive subset was confirmed by Sequenom MassARRAY genotyping (Supplementary Table 3), giving a positive prediction rate of 96.8% (95% confidence interval (CI), 95.2–98.4%) for somatic single-nucleotide variation (SNV) and 97.9% (95% CI, 95.8–99.9%) for somatic insertion or deletion (indel) alterations. Overall, 7,036 somatic mutations were detected in the 22 gastric cancer samples, of which 4,653 occurred in coding regions or essen- tial splice sites (2,513 missense, 137 nonsense, 5 stop codon loss, 90 splice site, 855 indel and 1,053 synonymous mutations) (Table 1 and Supplementary Table 4). Consistent with the known consequences of mismatch repair deficiency, the gastric cancer samples with MSI had an average of 31.61 somatic mutations (including both SNVs and indels) per megabase of DNA, whereas the MSS gastric cancer samples had an average of 3.29, a difference of approximately tenfold. Likewise, MSI gastric cancer samples had an average of 620 protein-altering somatic mutations, which was tenfold higher than the the number found in MSS gastric cancer samples (mean of 62). Although there was no sig- nificant difference in the nonsynonymous-to-synonymous (NS/S) ratio between the MSI and the MSS gastric cancers, we found a signifi- cantly higher NS/S ratio in EBV-infected gastric cancer samples (4.55 ± 0.97 compared to 2.76 ± 0.72 in non-EBV–infected gastric cancer samples, P < 0.012) and in poorly differentiated gastric cancers (3.52 ± 0.94 compared to 2.56 ± 0.78 in well- and moderately differentiated gastric cancers, P < 0.014), suggestive of a higher positive selection pressure for driver mutations in these subgroups 18,19 . Of note, the NS/S ratio for the EBV-infected gastric cancer samples (range 3.64–5.56) is among the highest for solid organ cancers (range 0.97–3.5) reported to date 3,8–10,12,17 . C-to-T transitions were the most common mutation (51%) across all gastric cancers, with 68% involving CG dinucleotides. The MSI gastric cancers also had a distinctly high rate of T-to-C transi- tions (30%) (Supplementary Table 5). The observed mutation incidence for MSS gastric cancers (3.29 per megabase) is higher than our previous results from sequencing the kinome of most other cancers (0.74–1.85 per megabase for renal, colorectal and ovarian cancers) but lower than in lung cancers (4.21 per megabase) 20 . The observed mutation spectrum Exome sequencing identifies frequent mutation of ARID1A in molecular subtypes of gastric cancer Kai Wang 1,7 , Junsuo Kan 2,7 , Siu Tsan Yuen 2 , Stephanie T Shi 3 , Kent Man Chu 4 , Simon Law 4 , Tsun Leung Chan 2 , Zhengyan Kan 1 , Annie S Y Chan 2 , Wai Yin Tsui 2 , Siu Po Lee 2 , Siu Lun Ho 2 , Anthony K W Chan 2 , Grace H W Cheng 2 , Peter C Roberts 5 , Paul A Rejto 1 , Neil W Gibson 1,6 , David J Pocalyko 1 , Mao Mao 1 , Jiangchun Xu 1 & Suet Yi Leung 2 1 Oncology Research Unit, Pfizer Worldwide Research and Development, La Jolla, California, USA. 2 Department of Pathology, The University of Hong Kong, Queen Mary Hospital, Pokfulam, Hong Kong. 3 External Research Solutions, Pfizer Worldwide Research and Development, La Jolla, California, USA. 4 Department of Surgery, The University of Hong Kong, Queen Mary Hospital, Pokfulam, Hong Kong. 5 Research Embedded Business Technology, Pfizer Worldwide Research and Development, La Jolla, California, USA. 6 Present address: Regulus Therapeutics, San Diego, California, USA. 7 These authors contributed equally to this work. Correspondence should be addressed to S.Y.L. (suetyi@hku.hk) or J.X. (jiangchun.xu@pfizer.com). Received 18 May; accepted 23 September; published online 30 October 2011; doi:10.1038/ng.982 LETTERS