Description of Materials and Methods

Samples and RNA preparation
Frozen tumor and normal gastric mucosa were collected from gastrectomy specimens from Queen Mary Hospital, The University of Hong Kong. 90 primary gastric adenocarcinomas, lymph node metastases from 14 of the 90 primary gastric cancers, and 22 samples of non-neoplastic gastric mucosa were analyzed using DNA microarrays; details of this analysis are presented in another publication (Chen et al, manuscript in preparation). Another 59 adenocarcinomas were later tested independently by quantitative RT-PCR as described below. This study was approved by the Ethics Committee of the University of Hong Kong and the IRB from Stanford University.

Tissues were frozen in liquid nitrogen within half an hour after they were resected. Samples of non-neoplastic mucosa from stomach were dissected free of muscle when fresh, and histologically confirmed to be tumor-free by frozen section. The tumors were unselected with respect to grade, but there was, in both cases, selection for specimens judged to have at least 50% tumor cells in the block. The clinical records were surveyed and the clinical parameters (for this study, overall survival) recovered after the biological experiments for each dataset were complete.

Tumors were classified using the Lauren classification into intestinal, diffuse, mixed and indeterminate types. The tumor stage was defined by the General Rules for Gastric Cancer Study of the Japanese Research Society for Gastric Cancer.

Total RNA was extracted using Trizol (Gibcon BRL) and mRNA was isolated from total RNA by the FastTrack mRNA isolation kit (Invitrogen).

Microarray procedure and data analysis
We used a cDNA microarray containing 44,500 cDNA clones, representing about 30,300 unique genes. The methods for microarray production, hybridization and data analysis were as previously described and are described in detail elsewhere (Chen et al, manuscript in preparation; website: http://genome-www.stanford.edu/Gastric_Cancer2).

Quantitative RT-PCR
Quantitative RT-PCR was performed as described. In brief, total RNA was further purified with an RNAqueous kit (Ambion), including DNAse I digestion to remove any genomic DNA contamination. Human GAPDH primer and probe reagents (Applied Biosystems) were used as the normalization control in subsequent quantitative analysis. Quantification was performed using the ABI Prism® 7900HT Sequence Detection System (Applied Biosystems) via a 2-step non-multiplexed Taqman® 5''3' exonuclease assay (Taqman® Reverse Transcription Reagents Kit and Taqman® PCR Core Reagents Kit, Applied Biosystems) according to the relative standard method. Calibration curves were generated for each transcript and validated using linear regression analysis (r2 > 0.99). Transcript quantification was performed in triplicate for every sample and reported relative to GAPDH. The primes and probe used for PLA2G2A in this study are: PLA2F: CCGCACTCAGTTATGGCTTCT; PLA2R: AGCGATCCGTTGCATCCTT; PLA2probe: CACTGTGGCGTGGGTGGCAGA

In-situ hybridization
In-situ hybridization for PLA2G2A was performed using 35S-labeled anti-sense and sense ribo-probes containing nucleotides 14-940 of PLA2G2A as described.

Statistical analysis
Prior to the statistical analysis, the missing values in the dataset were estimated with KNNimpute algorithm using 12 neighbors. A non-parametric t test with p-value cutoff of 0.001 based on 10,000 random column permutations was used to identify the differentially expressed genes. The False Discovery Rate (FDR) was estimated based on the number of genes that passed the p value cutoff of 0.001 in five datasets with randomized columns. The chi-square test with Yates correction was used to analyze relationships between categorical subgroups. Kaplan-Meier survival analysis was carried out by using WINSTAT EXCEL plug-in software (http://www.winstat.com).