Gastric cancer cases are often diagnosed at an advanced stage with poor prognosis. Platinum-based chemotherapy has been internationally accepted as first-line therapy for inoperable or metastatic gastric cancer. To achieve greater benefits, selection of patients eligible for this treatment is critical. Although gene expression profiling has been widely used as a genomic classifier to identify molecular subtypes of gastric cancer and to stratify patients for different chemotherapy regimens, its prediction accuracy can be improved. Adenosine-to-inosine (A-to-I) RNA editing has emerged as a new player contributing to gastric cancer development and progression, offering potential clinical utility for diagnosis and treatment. Using a systematic computational approach followed by both in vitro validations and in silico validations in The Cancer Genome Atlas (TCGA), we conducted a transcriptome-wide RNA editing analysis of a cohort of 104 patients with advanced gastric cancer and identified an RNA editing (GCRE) signature to guide gastric cancer chemotherapy. RNA editing events stood as a prognostic and predictive biomarker in advanced gastric cancer. A GCRE score based on the GCRE signature consisted of 50 editing sites associated with 29 genes, predicting response to chemotherapy with a high accuracy (84%). Of note, patients demonstrating higher editing levels of this panel of sites presented a better overall response. Consistently, gastric cancer cell lines with higher editing levels showed higher chemosensitivity. Applying the GCRE score on TCGA dataset confirmed that responders had significantly higher levels of editing in advanced gastric cancer. Overall, this newly defined GCRE signature reliably stratifies patients with advanced gastric cancer and predicts response from chemotherapy.

Significance:

This study describes a novel A-to-I RNA editing signature as a prognostic and predictive biomarker in advanced gastric cancer, providing a new tool to improve patient stratification and response to therapy.

Gastric cancer is the third leading cause of cancer-related death worldwide responsible for more than 780,000 deaths annually (1, 2). Asian population carries a higher risk for the disease in terms of incidence and mortality. The current treatment options for the advanced stage mainly employ palliative chemotherapy based on fluoropyrimidine and platinum-based compounds. However, an important question is whether we can improve the selection of patients with advanced gastric cancer for chemotherapy to achieve greater benefit from the treatment. To address this, different groups have put much effort in stratifying patients with gastric cancer into molecular subtypes (3–9); however, these studies heavily focused on gene expression and/or mutation profiling, often with the limitations of microarray-based platforms or small sample sizes; thus, novel molecular data types and approaches are needed for better guiding the chemotherapy treatment for this class of patients. Here, we demonstrate A-to-I RNA editing as a novel epigenetic classifier to reliably identify responders to chemotherapy.

RNA editing is a post- and/or cotranscriptional modification that results in specific nucleotide changes that occur on the RNA. In humans, the most frequent type of RNA editing is the conversion of adenosine to inosine (A-to-I), which is catalyzed by ADAR proteins. In vertebrates, a family of 3 ADARs, ADAR1, ADAR2, and ADAR3, has been characterized (10). ADAR1 and ADAR2 catalyze all currently known A-to-I editing sites. Inosine (I) essentially mimics guanosine (G); therefore, ADAR proteins introduce a virtual A-to-G substitution in transcripts. Such changes can lead to specific amino acid substitutions (11–16), alternative splicing (17), miRNA-mediated gene silencing (18, 19), or changes in transcript localization and stability (20–22).

As reported by us and others in the past decade, dysregulated A-to-I editing is a key driver in the pathogenesis of various cancers, such as breast cancer (23), glioma (24, 25), chronic myeloid leukemia (26), hepatocellular carcinoma (11, 27), and esophageal squamous cell carcinoma (12). Our group provided the first extensive transcriptome-wide RNA editing analysis of primary gastric tumors and highlighted a major role for RNA editing in gastric cancer disease and progression (28). This observation has been missed by previous next-generation sequencing analyses of gastric cancer focused on DNA alterations alone. We reported that gastric cancer displays a severely disrupted RNA editing balance induced by the differentially expressed ADARs (ADAR1 and ADAR2). Clinically, the differentially expressed ADARs, which are characterized by ADAR1 overexpression and ADAR2 downregulation in tumors, have great prognostic value and diagnostic potential for primary gastric cancer. However, the role of RNA editing in inoperable, locally advanced, or recurrent and/or metastatic gastric cancer and whether RNA editing signature can be used to prospectively and retrospectively stratify patients with advanced gastric cancer and predict response from chemotherapy remain largely unknown.

Despite the fact that ADARs are responsible for A-to-I RNA editing activity, there is not always a linear relationship between the expression and activity of ADARs and editing frequencies of their target RNAs (29–31), due to their differential subcellular distribution (32), cis- and trans-regulatory interactors (33–36) and posttranscriptional modifications (30, 37). On the other hand, changes in the editing level of individual sites have been shown to play a driver role in several cancer types (11, 24, 38). Therefore, RNA editing events are considered as a better proxy than ADAR expression per se to provide molecular information to be translated into clinical applications. We have previously initiated a translational “3G” (Genomic Guided therapy for Gastric cancer) trial to investigate the benefit of using a genomic classifier to guide the choice of two platinum-based chemotherapy regimens in an advanced gastric cancer setting (7). In this study, we conducted a high-throughput RNA sequencing (RNA-seq) analysis of 104 patients with advanced gastric cancer who had been enrolled into the “3G” trial and investigated the clinical utility of RNA editing events in advanced gastric cancer. To our knowledge, this is the first report that demonstrates that RNA editing alone can be employed as a prognostic and predictive factor in advanced gastric cancer independently of gene expression, DNA mutations, or other epigenetic alterations, and more importantly, a panel of 50 editing sites could be readily detected in patients with advanced gastric cancer and accurately predicts outcome of chemotherapy. Overall, our study provides insight into the role of RNA editing in gastric cancer, which may facilitate the therapeutic decision making.

Gastric cancer cohort

A total of 104 patients from the 3G Trial (7) were involved in this study. The patients were diagnosed with metastatic or recurrent gastric cancer and enrolled onto the first-line palliative chemotherapy (platinum-fluoropyrimidine doublet chemotherapy regime). For each eligible patient, fresh endoscopic biopsies of the primary tumor in situ within 3 weeks prior to treatment initiation were used for RNA-seq analysis. Treatment response to chemotherapy (PD, progressive disease; SD, stable disease; PR, partial response) was evaluated by radiologists who were blinded to the study, and assessed as the best overall response as per RECIST (version 1.1) criteria (39) by CT scans or MRI done at 6-week intervals. At the cut-off date, in this cohort of 104 patients, overall survival and tumor response data had been obtained for 54 and 55 patients, respectively, where a total of 50 patients had both overall survival and tumor response data.

RNA sequencing

A total of 1 μg of total RNA was used to create libraries with Illumina TruSeq Stranded Total RNA Library Prep Kit (Illumina) according to the manufacturer's instructions. Library fragment size was determined using the DNA 1000 Kit on the Agilent Bioanalyzer (Agilent Technologies). Libraries were quantified by qPCR using the KAPA Library Quantification Kit (KAPA Biosystems). Libraries were pooled in equimolar and cluster generation was performed on the Illumina cBOT system (Illumina). Sequencing (150-bp pair-end) was performed on the Illumina HiSeq 3000 system at the Duke-NUS Genome Biology Facility, according to the manufacturer's protocol (Illumina).

Identification of RNA editing events

A bioinformatics pipeline adapted from a previously published method (40) was used to identify RNA editing events from RNA-seq data by using the CSI NGS Portal (https://csibioinfo.nus.edu.sg/csingsportal; ref. 41). For each sample, raw reads with adapters were trimmed by using Trimmomatic (v0.38; ref. 42) retaining the reads with ≥ 35 bases and average read quality score >20 after trimming. Clean reads were mapped to the reference human genome (hg19) with a splicing junction database generated from transcript annotations derived from UCSC (43), RefSeq (44), Ensembl (45), and GENCODE (v19; ref. 46) by using Burrows–Wheeler aligner with default parameters (bwa mem, v0.7.17-r1188; ref. 47). To retain high quality data, PCR duplicates were removed (samtools markdup -r, v1.9; ref. 48), and the reads with mapping quality score <20 were discarded. Junction-mapped reads were then converted back to the genomic-based coordinates. An in-house perl script was utilized to call the variants from samtools pileup data and the sites with at least two supporting reads were initially retained. The candidate events were filtered by removing the SNPs reported in different cohorts (1000 Genomes Project; ref. 49), NHLBI GO Exome Sequencing Project (https://evs.gs.washington.edu/EVS/), dbSNP (v150; ref. 50) and excluding the sites within the first six bases of the reads caused by imperfect priming of random hexamer during cDNA synthesis. For the sites not located in Alu elements, that is, short repetitive DNA sequences abundant in the human genome, the candidates within the four bases of a splice junction on the intronic side, and those residing in the homopolymeric regions and in the simple repeats were all removed. Candidate variants located in the reads that map to the nonunique regions of the genome by using BLAST-like alignment tool (51) were also excluded. At last, only A-to-G editing sites based on the strand information from the strand-specific RNA-seq data were considered for all the downstream analyses. The genomic regions of the editing variants and the associated genes were annotated by using ANNOVAR (v2018; ref. 52) with the UCSC refGene table annotation. We applied the same pipeline on The Cancer Genome Atlas (TCGA) stomach adenocarcinoma (STAD) cohort (6).

To identify high confidence and common editing events, stringent filtering criteria were applied. Specifically, each editing site was required to have a coverage of at least 20 reads and editing frequency higher than 0.1 (10%) in all the samples. This resulted in 780 high confidence editing sites shared by 104 samples of our gastric cancer cohort. For TCGA STAD cohort, we did not apply these thresholds for the validation of GCRE signature to include more sites, as the number of high confidence editing sites were relatively fewer in TCGA due to the lower sequencing depth. In TCGA, we included only those samples with at least 8 of 50 sites in the GCRE signature were found to be edited. Then, for each sample, we used all these edited sites to calculate the GCRE score.

Clustering analyses and heatmaps

The clustering and heatmap analyses were performed by using R package “superheat” (53). The RNA editing levels of the corresponding sites were used as the distance matrix to perform the k-means clustering for the heatmaps in an unsupervised manner (scale = TRUE, n.clusters.rows = 2, clustering.method = “kmeans”). The side plots were drawn using “scattersmooth” parameter with the “lm” method based on the average editing value of the corresponding rows or columns calculated by “rowMeans2” and “colMeans2” functions from the “matrixStats” package, respectively.

Correlation of RNA editing with response to chemotherapy

Pearson correlation analysis was performed by using “cor.test” function in R with “Pearson” method. For each of 780 editing sites, Pearson correlation coefficient (r) and associated P value were calculated between RNA editing levels and the overall response to chemotherapy across 55 patients with available data. The editing frequencies calculated from the RNA-seq data were used as the first vector. As the second vector, the overall response data were used after transforming the original categorical variables into numerical variables (PD = 0, SD = 1, PR = 2), so that the correlation analysis can be performed. P values were used to assess the editing sites with significant correlations at P < 0.05 threshold, which resulted in 50 positively and 3 negatively correlated sites.

GCRE score calculation and GCRE signature

To predict the chemotherapy outcome based on RNA editing, we developed a GCRE score based on the “Z-score” using 50 sites that showed significant positive correlation with the overall response. First, z-transformation was performed for each site based on the RNA editing levels. Then, the samples were ranked by using the average Z-score across the sites. The samples above and below a cut-off value (0.4) were regarded as the high and low editing groups, respectively. The statistical measures were calculated based on the prediction of the responders in the cohort. Disease enrichment analysis of GCRE signature genes were performed by using Enrichr web server (54) with “PheWeb 2019” database.

Gastric cancer cell lines and validation of the GCRE signature

A total of 10 cell lines were used for the validation of the GCRE signature. AGS, SNU5, SNU16, and NCI-N87 were purchased from the ATCC between 2019 and 2020. MKN1, MKN7, MKN28, MKN45, and MKN74 were obtained from Japanese Collection of Research Bioresources Cell Bank between 2019 and 2020. YCC11 cell line was provided by Singapore Gastric Cancer Consortium (SGCC) in 2014. All the cell lines were tested negative for Mycoplasma. SNU5 cells were cultured with IMDM (Gibco BRL) supplemented with 20% FBS (Gibco BRL), and all the other cell lines were cultured in RPMI medium (Gibco BRL) supplemented with 10% FBS (Gibco BRL). Cells were grown in a humidified incubator with 5% CO2 at 37°C.

The editing levels of 26 of 50 editing sites in these 10 cell lines were quantified by Sanger sequencing. The sites were selected as the top 10 sites, which showed the highest change in editing level between high and low editing group as defined in the RNA-seq data, and additional 16 sites that are randomly picked from the remaining sites in the panel of GCRE signature. Drug response of the cell lines to oxaliplatin (Sigma-Aldrich) were assessed by IC50 values using MTT (3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay. Briefly, gastric cancer cell lines were seeded in 96-well plates with 2.5 × 103 to 10 × 103 cells/well according to their growth rate. After 72-hour oxaliplatin drug treatment, 10 μL MTT substrate (Sigma-Aldrich) was added into each well followed by 3-hour incubation. MTT substrate was then removed and cells were lysed by addition of 100 μL MTT stop solution. Absorbance at 570 nm was measured using Tecan microplate reader.

For the cell viability assay, 6 suitable cell lines were used, where the cells were seeded in 6-well plate with 3 × 104 to 15 × 104 cells/well and cultured with indicated concentration of oxaliplatin for 48 hours. Cells were stained with crystal violet (Sigma-Aldrich) for colony visualization.

Statistical analyses

As a rule of thumb, all the available features and the samples were included in the respective analyses as long as the data being investigated were nonmissing. Initial clustering analysis was performed in an unbiased manner based on the editing levels of all the sites identified in all the samples. The downstream analyses were performed by using a panel of 50 sites that are selected solely based on the correlation value of their editing levels obtained from RNA-seq data and the overall response to chemotherapy assessed by clinicians and radiologists. GCRE score was derived based on the editing levels of the selected genes in all the samples with response data available. No manual selection or exclusion was applied on the genes, editing sites or samples for this study, unless limited by the availability of data under investigation, for example, 55 of 104 samples had response data. These criteria were also applied to the cell lines in the same way as per patient samples. For the experimental validations, unless otherwise indicated, the data are presented as the mean ± SEM of three independent experiments. Wilcoxon rank-sum test was applied when comparing distributions, Pearson correlation coefficient was reported for paired correlation analyses, and log-rank test P value was shown for the survival analyses. A P value of less than 0.05 was considered to be statistically significant, and the type of the statistical test applied was indicated appropriately.

Ethics approval and consent to participate

Samples were collected from the cancer centers in Singapore (n = 37) and South Korea (n = 67) between 2010 and 2018. A written informed consent was obtained from all the patients prior to the enrolment to the trial, and the study was done in accordance with the Declaration of Helsinki and International Conference on Harmonisation and Good Clinical Practice guidelines. The protocol was approved by the Institutional Review Board at each study site and complied with local laws and regulations.

Availability of data and materials

RNA editing pipeline is available online at the CSI NGS Portal (https://csibioinfo.nus.edu.sg/csingsportal; ref. 41). The bioinformatics code for the downstream analyses are available upon reasonable request. The processed data are available as Supplementary Tables.

The landscape of A-to-I RNA editing in advanced gastric cancer tumors

We conducted a genome-wide A-to-I RNA editing analysis using RNA-seq data of endoscopic tumor biopsies obtained from 104 patients with advanced, metastatic, or recurrent gastric cancer at the latest stage (stage IV or inoperable at stage III) prior to their first-line palliative chemotherapy (platinum–fluoropyrimidine doublet chemotherapy regime; Supplementary Table S1). Applying our established RNA editing pipeline (34, 55) with stringent filtering criteria (Materials and Methods), we identified a total of 2,154,091 high confidence A-to-I RNA editing sites, with a median number of 17,000 editing sites per sample (Fig. 1A), predominantly located in introns and 3′ untranslated regions (Fig. 1B), consistent with the previous reports (28, 56). The number of editing sites moderately correlated with the total number of sequencing reads (Pearson r = 0.41, P = 1.96e−05) and with the overall editing activity (r = 0.39, P = 3.87e−05), where the latter was assessed by Alu editing index (57). The overall editing activity, however, was independent of the total number of sequencing reads (r = −0.03, P = 0.8) and comparable across the samples (range 0.77–1.34, average = 1.03, SD = 0.13, excluding 1 outlier). The distribution of the number of editing sites across the samples revealed an overwhelming number of sample-specific sites (n = 159,146), as well as 780 shared sites referred to as hotspots (i.e., edited in all the samples in the cohort; Fig. 1C; Supplementary Table S2). We included only these A-to-G hotspot editing sites for further analysis, and excluded the possibility of non A-to-G editing events to be a hotspot (Supplementary Fig. S1).

Figure 1.

A-to-I RNA editing landscape in advanced gastric cancer cases. A, Number of RNA-seq reads, number of editing sites, and Alu editing index (AEI) of 104 gastric cancer samples. The samples are sorted based on the decreasing value of AEI. r = Pearson correlation coefficient. B, Distribution of 2,154,091 high confidence editing sites over annotated genomic regions. Others = ncRNA_exonic, upstream, 5′UTR, exonic, upstream;downstream, splicing. C, Distribution of number of editing sites across the samples, where shared sites that are edited in all the samples in this cohort (hotspots) are highlighted.

Figure 1.

A-to-I RNA editing landscape in advanced gastric cancer cases. A, Number of RNA-seq reads, number of editing sites, and Alu editing index (AEI) of 104 gastric cancer samples. The samples are sorted based on the decreasing value of AEI. r = Pearson correlation coefficient. B, Distribution of 2,154,091 high confidence editing sites over annotated genomic regions. Others = ncRNA_exonic, upstream, 5′UTR, exonic, upstream;downstream, splicing. C, Distribution of number of editing sites across the samples, where shared sites that are edited in all the samples in this cohort (hotspots) are highlighted.

Close modal

RNA editing is a prognostic marker in advanced gastric cancer

First, we queried whether RNA editing has a prognostic value in advanced gastric cancer. To this end, we performed an unsupervised k-means clustering based on the RNA editing levels in an unbiased manner, that is, including all the hotspot editing sites (n = 780) identified from the RNA-seq data and all the patients with available survival data (n = 54). This resulted in two distinct clusters, which we defined as “high editing cluster” and “low editing cluster” referring to the relative average editing levels in the clusters (Fig. 2A). The high editing cluster demonstrated significantly better patient survival (P = 0.037, k = 2; Fig. 2B), which was also evident when using hierarchical clustering (P = 0.067; Supplementary Fig. S2A and S2B). Instead, the baseline patient characteristics we investigated did not correlate with the patient survival (Cox proportional hazards univariate analysis; Supplementary Table S3).

Figure 2.

RNA editing hotspots as a prognostic marker in advanced gastric cancer. A, Unsupervised k-means (k = 2) clustering of advanced gastric cancer samples based on RNA editing levels of 780 hotspot editing sites. The scatterplot shows average editing levels per sample. Of the cohort, 54 samples with survival data available are included in the analysis. B, Survival plot of the two editing clusters. Data with hierarchical clustering are given in Supplementary Fig. S2A and S2B.

Figure 2.

RNA editing hotspots as a prognostic marker in advanced gastric cancer. A, Unsupervised k-means (k = 2) clustering of advanced gastric cancer samples based on RNA editing levels of 780 hotspot editing sites. The scatterplot shows average editing levels per sample. Of the cohort, 54 samples with survival data available are included in the analysis. B, Survival plot of the two editing clusters. Data with hierarchical clustering are given in Supplementary Fig. S2A and S2B.

Close modal

RNA editing is a predictive marker in advanced gastric cancer

To investigate whether RNA editing also has a predictive value in gastric cancer, we focused on the overall response data of the patients to palliative platinum-based chemotherapy. For each of 780 hotspot editing sites, we applied Pearson correlation test between the RNA editing levels and overall response to chemotherapy (Materials and Methods). This led us to identify 53 key editing sites that showed significant correlation (P < 0.05; Fig. 3A; Supplementary Table S4). Interestingly, 50 of these sites had a positive correlation, implying the higher the editing level the better the response (response categories are numerically represented as PD = 0, SD = 1, PR = 2). Reapplying clustering on the patients by using the editing levels of only these 53 sites resulted in 75% accuracy of predicting the responders (i.e., patients who achieved PR; Fig. 3B), which was a significantly better prediction compared to a random selection of 53 sites (empirical P = 0.00409, N = 100,000; Supplementary Fig. S3). Randomization test thus supports the validity of our selection method and highlights the importance of these sites as predictive markers. Interestingly, disease enrichment analysis of 29 genes corresponding to these sites showed significant association with “gastric ulcer,” suggesting common etiologic factors (58), owing to the two key genes PHACTR4 and MED18 (Fig. 3C), which were reported to be involved in gastric cancer via different mechanisms (59, 60).

Figure 3.

GCRE signature as a predictive marker in advanced gastric cancer. A, Correlation of hotspot editing sites with overall response to chemotherapy. Categorical values of overall response are converted to numerical values (PD = 0, SD = 1, PR = 2), and then Pearson correlation test is applied to each site between editing levels and numerical values of overall response. A representative example of a positive correlation is illustrated by the inset plot. Of the cohort, 55 samples with tumor response data available are included in the analysis, n (PD) = 8, n (SD) = 29, n (PR) = 18. Gene symbols associated with the editing sites are shown for the significant cases (n = 53, P < 0.05), where genes with multiple sites are repeated. B,k-means clustering of samples based on the RNA editing levels of 53 sites that significantly correlated with overall response (k = 2). Prediction accuracy of responders is 41/55 = 75%. C, Disease enrichment analysis results of GCRE signature genes.

Figure 3.

GCRE signature as a predictive marker in advanced gastric cancer. A, Correlation of hotspot editing sites with overall response to chemotherapy. Categorical values of overall response are converted to numerical values (PD = 0, SD = 1, PR = 2), and then Pearson correlation test is applied to each site between editing levels and numerical values of overall response. A representative example of a positive correlation is illustrated by the inset plot. Of the cohort, 55 samples with tumor response data available are included in the analysis, n (PD) = 8, n (SD) = 29, n (PR) = 18. Gene symbols associated with the editing sites are shown for the significant cases (n = 53, P < 0.05), where genes with multiple sites are repeated. B,k-means clustering of samples based on the RNA editing levels of 53 sites that significantly correlated with overall response (k = 2). Prediction accuracy of responders is 41/55 = 75%. C, Disease enrichment analysis results of GCRE signature genes.

Close modal

GCRE score predicts responders with high accuracy

To predict chemotherapy outcome in a robust way, we derived a score based on the editing levels of 50 sites (GCRE signature) having positive correlation with the overall response (Materials and Methods). Briefly, we obtained the average Z-score per sample based on GCRE signature (GCRE score); then, we stratified the patients into 3 response groups (PD, SD, and PR) and predicted the responders based on this GCRE score. Overall, we observed a good performance (AUC = 0.77; Fig. 4A), and at the cut-off value of 0.4, we achieved an accuracy of 84% (sensitivity = 67%, specificity = 92%) to predict the responders (Fig. 4B), which was a better prediction compared with the clustering method (accuracy of 75%; Fig. 3B). We also confirmed that reducing the number of editing sites or including 3 negatively correlated sites compromised accuracy (Supplementary Fig. S4). Overall, these results suggest that employing the GCRE score based on the 50 sites contributing to the GCRE signature could stratify responders and nonresponders to chemotherapy with high accuracy.

Figure 4.

Utilization of a GCRE score to predict chemotherapy response. A, ROC curve showing the performance of GCRE score in prediction of responders. TPR, true positive rate; FPR, false positive rate. B, Stratification of patients with gastric cancer into chemotherapy response groups and prediction of responders based on GCRE score at the cut-off value of 0.4. GCRE score for each patient denotes average Z-score of RNA editing levels across the panel of 50 sites in the GCRE signature. The statistical measures refer to the classification of responders and nonresponders. PPV, positive predictive value; NPV, negative predictive value. Nonresponder, progressive disease + (PD) stable disease (SD). Responder, partial response (PR).

Figure 4.

Utilization of a GCRE score to predict chemotherapy response. A, ROC curve showing the performance of GCRE score in prediction of responders. TPR, true positive rate; FPR, false positive rate. B, Stratification of patients with gastric cancer into chemotherapy response groups and prediction of responders based on GCRE score at the cut-off value of 0.4. GCRE score for each patient denotes average Z-score of RNA editing levels across the panel of 50 sites in the GCRE signature. The statistical measures refer to the classification of responders and nonresponders. PPV, positive predictive value; NPV, negative predictive value. Nonresponder, progressive disease + (PD) stable disease (SD). Responder, partial response (PR).

Close modal

Validation of GCRE signature in gastric cancer cell lines

We next validated the GCRE signature in 10 commercially available gastric cancer cell lines. First, we assessed the drug response of each cell line to oxaliplatin by IC50 (the half maximal inhibitory concentration) values (Fig. 5A; Supplementary Fig. S5) and further confirmed by cell viability assay (Fig. 5B). Then, we quantified the RNA editing levels of 26 randomly selected sites from 50 sites of the GCRE signature in the same 6 cell lines by Sanger sequencing (Materials and Methods; Fig. 5C). We found a negative correlation between the average IC50 values and average RNA editing levels (r = −0.58), implying that gastric cancer cells with higher editing levels of the GCRE signature sites demonstrate higher chemosensitivity, which was consistent with our observation in the patient samples.

Figure 5.

Validation of GCRE signature in gastric cancer cell lines. A, Chemosensitivity of 10 gastric cancer cell lines to oxaliplatin. The bars show the average IC50 value of three biological replicates with SEMs. B, Cell viability assay of 6 gastric cancer cell lines cultured with indicated concentration of oxaliplatin for 48 hours. Cells were stained with crystal violet. C, Editing levels of 26 sites from the panel of GCRE signature quantified by Sanger sequencing. Numbers in cells denote the percentage of editing levels and coloring shows the relative gradient across the row (scale). Black cells denote undetectable editing level. The Pearson correlation coefficient between the average IC50 values in A and average RNA editing levels in C is r = −0.58.

Figure 5.

Validation of GCRE signature in gastric cancer cell lines. A, Chemosensitivity of 10 gastric cancer cell lines to oxaliplatin. The bars show the average IC50 value of three biological replicates with SEMs. B, Cell viability assay of 6 gastric cancer cell lines cultured with indicated concentration of oxaliplatin for 48 hours. Cells were stained with crystal violet. C, Editing levels of 26 sites from the panel of GCRE signature quantified by Sanger sequencing. Numbers in cells denote the percentage of editing levels and coloring shows the relative gradient across the row (scale). Black cells denote undetectable editing level. The Pearson correlation coefficient between the average IC50 values in A and average RNA editing levels in C is r = −0.58.

Close modal

GCRE signature is a prevalent in late-stage gastric cancer, independent of chemotherapy regimen

To infer whether our GCRE signature is representative of different tumor stages and chemotherapy regimens, we applied the GCRE score in TCGA STAD (stomach adenocarcinoma) cohort (6). This cohort comprises of patients diagnosed with different stages of gastric cancer and treated with multiple combinations of drugs, along with multiple data points reporting for primary and follow-up treatment outcome (Supplementary Table S5). When we applied the GCRE score on these patients, we observed that responders in the latest stage (IV), but not in the previous stages (II and III), had significantly higher levels of editing in the panel of 50 sites compared with nonresponders (Fig. 6A and B), confirming our observation in our advanced gastric cancer cohort, despite the diverse drug regimens between the two cohorts. We have excluded the patients at early stage (I) from the analysis as these patients most likely had undergone tumor removal rather than chemotherapy, so the response information was ambiguous. With available data, however, we could not validate the prognostic value of the GCRE signature possibly due to the small sample size (Supplementary Fig. S6A and S6B). These results suggest that the GCRE signature is coherent to very late stage of the disease as a predictive marker, independently of different chemotherapy regimens.

Figure 6.

Validation of GCRE signature in TCGA STAD cohort. Distribution (violin plots) and performance (ROC curves) of GCRE scores in TCGA STAD cohort stratified by disease stage and response type. For score calculation, 50 sites in GCRE signature and their editing levels in TCGA patients were used wherever available and at least 8 sites were required to be edited to calculate the score. Primary response (A) refers to the “primary_therapy_outcome” and follow-up response (B) denotes the “followup_treatment_success” as reported by TCGA. Response groups are merged based on the consensus of the individual treatment outcomes: Nonresponder, stable/progressive disease; responder, partial/complete remission; STAD, stomach adenocarcinoma; TPR, true positive rate; FPR, false positive rate. P values are shown for Wilcoxon rank-sum test.

Figure 6.

Validation of GCRE signature in TCGA STAD cohort. Distribution (violin plots) and performance (ROC curves) of GCRE scores in TCGA STAD cohort stratified by disease stage and response type. For score calculation, 50 sites in GCRE signature and their editing levels in TCGA patients were used wherever available and at least 8 sites were required to be edited to calculate the score. Primary response (A) refers to the “primary_therapy_outcome” and follow-up response (B) denotes the “followup_treatment_success” as reported by TCGA. Response groups are merged based on the consensus of the individual treatment outcomes: Nonresponder, stable/progressive disease; responder, partial/complete remission; STAD, stomach adenocarcinoma; TPR, true positive rate; FPR, false positive rate. P values are shown for Wilcoxon rank-sum test.

Close modal

Patient stratification based on molecular information promises a great value for guiding diagnosis and treatment choice, particularly in genetically heterogeneous diseases such as gastric cancer. Till date, a number of studies attempted for the molecular classification of patients for this purpose, most of which are based on gene expression profiling utilizing microarray data (3–5, 7–9). With the broader availability and better quality of next-generation sequencing data, now it became possible to reliably study other molecular features than gene expression. In this study, we propose A-to-I RNA editing as a novel molecular classifier in advanced gastric cancer, and show that an RNA editing signature can be used to stratify patients with differential benefit from chemotherapy, which may not be limited to gastric cancer and can be potentially extended to other cancer types and cohorts.

In recent studies, systematic and unbiased analysis of RNA editing led to the establishment of the driver role of individual editing events in several cancers (11, 12, 23, 24, 26, 27). It also became evident that editing level of a considerable number of RNA editing sites is correlated with patient survival, indicating their utility as prognostic markers (11, 28, 38, 61, 62). Several studies have also reported that a specific RNA editing event could selectively affect the outcome of cancer therapies. For example, protein-recoding RNA editing of COG3 and GRIA2 gene increases the drug sensitivity to MEK inhibitors (38). In this study, we followed a top-down approach, starting from transcriptome-wide global editing events toward a small panel of sites with potential clinical utility. Accurate identification of RNA editing events rely on several factors, such as high sequencing quality, sufficient sequencing depth, rigorous filtering, and strict inclusion criteria. In our dataset, we achieved a median of 117M reads per sample after adapter and quality trimming, so that we could apply a very high coverage threshold (≥20 reads) to identify the editing sites. This led us to accurately quantify the differential editing levels across the samples, which was crucial for the high resolution of the clustering and correlation analyses.

Our approach demonstrates a simple yet powerful way of identification of key editing events in advanced gastric cancer, uncovering the GCRE signature consisting of 50 editing sites associated with 29 genes. This signature has two characteristics: (i) it consists of common A-to-I RNA editing events across patients with gastric cancer (i.e., hotspots); (ii) higher editing level of these A-to-I RNA editing sites is associated with favorable clinical state (i.e., good prognosis and better response to chemotherapy). Once clinically validated, this RNA editing signature can be conveniently managed in the laboratory setting for individual patients to assist in therapeutic decision-making.

In addition to the GCRE signature, we report a list of 780 editing sites shared by 104 advanced gastric cancer tumors, which presents a valuable resource for follow up studies. These sites include previously reported hotspots (frequently edited) with pathogenic role (e.g., EIF2AK2, MAVS, GATC, CTSS, METTL7A; refs. 34, 61, 63–65) as well as novel sites reported here for the first time (e.g., NEAT1, ORC2, FGD5-AS1). Although in this study, we mainly focused on a small panel of 53 sites having significant correlation with the tumor response, the remaining sites potentially carry important information for gastric cancer pathology awaiting further exploration.

The availability of TCGA data gave us a great opportunity to validate the GCRE signature independently. The heterogeneity of these patients additionally allowed us to assess its specificity in terms of disease stage and treatment regimen, although the information on the type of chemotherapy for stage IV was limited. However, additional gastric cancer datasets comprising of diverse patient characteristics with available RNA-seq and tumor response data are needed to conclude whether this signature is cohort-specific or representative of a certain patient class, and whether these 50 editing sites are definitive. In particular, the prognostic value of the GCRE signature needs investigation in cohorts with larger sample size.

Overall, in this study, we have investigated the translational potential of RNA editing process in advanced gastric cancer by using transcriptomic and clinical data of a large cohort of patients. We discovered an RNA editing signature consisting of a small panel of genes independently of expression profiling, developed an RNA editing score that can predict responders with high accuracy, and validated them in gastric cancer cell lines and an independent cohort of TCGA. These findings suggest a novel clinical utility of RNA editing events for guiding chemotherapy treatment in this deadly cancer.

O. An and L. Chen report being inventors in a patent application (10202003405Y) that is related to the work that is described in this manuscript. R. Sundar reports being an advisory board member of Bristol-Myers Squibb, Merck, Eisai, Bayer, Taiho, and reports receiving honoraria for talks from MSD, Eli Lilly, Bristol-Myers Squibb, Roche, and Taiho and travel funding from Roche, AstraZeneca, Taiho, Eisai; and research funding from Paxman Coolers, MSD outside the submitted work. S.Y. Rha reports other support from Daiichi Sankyo, Eisai, grants and other support from MSD, other support from Celltrion, BMS, Boehringer Ingelheim, and Aslan outside the submitted work. P. Tan reports receiving honoraria for travel from Illumina and research funding from Thermo Fisher, Kyowa Hakko Kirin. No disclosures were reported by the other authors.

O. An: Conceptualization, data curation, software, formal analysis, visualization, methodology, writing–original draft, writing–review and editing. Y. Song: Formal analysis, validation. X. Ke: Formal analysis, validation. J.B.-Y. So: Conceptualization, resources, data curation, supervision, funding acquisition, investigation, project administration, writing–review and editing. R. Sundar: Conceptualization, resources, data curation, funding acquisition, methodology, writing–review and editing. H. Yang: Resources, supervision, funding acquisition. S.Y. Rha: Data curation, project administration. M.H. Lee: Data curation. S.T. Tay: Data curation. X. Ong: Data curation. A.L.K. Tan: Data curation. M.C.H. Ng: Data curation. E. Tantoso: Data curation. L. Chen: Conceptualization, resources, supervision, funding acquisition, validation, investigation, methodology, project administration, writing–review and editing. P. Tan: Resources, funding acquisition, investigation, project administration. W.P. Yong: Resources, data curation, funding acquisition, investigation, project administration, writing–review and editing. Singapore Gastric Cancer Consortium: Resources, data curation, funding acquisition, project administration.

The authors greatly thank the patients and their families for participating in the clinical trial and for contributing tissue magnanimously to the biomarker study. This work was supported by the National Research Foundation Singapore under its Translational and Clinical Research (TCR) Flagship Programme grant, administered by the Singapore Ministry of Health's National Medical Research Council and awarded to the Singapore Gastric Cancer Consortium (SGCC). This research was also supported by National Research Foundation Singapore; Singapore Ministry of Education under its Research Centres of Excellence initiative; the RNA Biology Centre at Cancer Science Institute of Singapore, NUS, under the National Research Foundation Singapore's and the Singapore Ministry of Education's Research Centres of Excellence initiative Tier 2 grants MOE2018-T2-1-005, MOE2019-T2-2-008; Tier 3 grant MOE2014-T3-1-006 and National Medical Research Council (NMRC) Clinician Scientist-Individual Research Grant (CSIRG) MOH-CIRG18nov-0007, Project ID: MOH-000214 awarded to L. Chen, as well as a grant from the National R&D Programme for Cancer Control, Ministry of Health and Welfare, Republic of Korea (1520190). R. Sundar is supported by a National Medical Research Council (NMRC) Fellowship (NMRC/Fellowship/0059/2018), Singapore. P. Tan is supported by Duke-NUS Medical School and the Biomedical Research Council, Agency for Science, Technology and Research. This work was also supported by National Medical Research Council grants TCR/009-NUHS/2013, NR13NMR111OM, and NMRC/STaR/0026/2015.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Bray
F
,
Ferlay
J
,
Soerjomataram
I
,
Siegel
RL
,
Torre
LA
,
Jemal
A
. 
Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries
.
CA Cancer J Clin
2018
;
68
:
394
424
.
2.
Global Burden of Disease Cancer Collaboration
,
Fitzmaurice
C
,
Abate
D
,
Abbasi
N
,
Abbastabar
H
,
Abd-Allah
F
, et al
Global, regional, and national cancer incidence, mortality, years of life lost, years lived with disability, and disability-adjusted life-years for 29 cancer groups, 1990 to 2017: a systematic analysis for the global burden of disease study
.
JAMA Oncol
2019
;
5
:
1749
68
.
3.
Lei
Z
,
Tan
IB
,
Das
K
,
Deng
N
,
Zouridis
H
,
Pattison
S
, et al
Identification of molecular subtypes of gastric cancer with different responses to PI3-kinase inhibitors and 5-fluorouracil
.
Gastroenterology
2013
;
145
:
554
65
.
4.
Tan
IB
,
Ivanova
T
,
Lim
KH
,
Ong
CW
,
Deng
N
,
Lee
J
, et al
Intrinsic subtypes of gastric cancer, based on gene expression pattern, predict survival and respond differently to chemotherapy
.
Gastroenterology
2011
;
141
:
476
85
.
5.
Cristescu
R
,
Lee
J
,
Nebozhyn
M
,
Kim
KM
,
Ting
JC
,
Wong
SS
, et al
Molecular analysis of gastric cancer identifies subtypes associated with distinct clinical outcomes
.
Nat Med
2015
;
21
:
449
56
.
6.
The Cancer Genome Atlas Research Network
. 
Comprehensive molecular characterization of gastric adenocarcinoma
.
Nature
2014
;
513
:
202
9
.
7.
Yong
WP
,
Rha
SY
,
Tan
IB
,
Choo
SP
,
Syn
NL
,
Koh
V
, et al
Real-time tumor gene expression profiling to direct gastric cancer chemotherapy: proof-of-concept “3G” trial
.
Clin Cancer Res
2018
;
24
:
5272
81
.
8.
Li
Z
,
Gao
X
,
Peng
X
,
May Chen
MJ
,
Li
Z
,
Wei
B
, et al
Multi-omics characterization of molecular features of gastric cancer correlated with response to neoadjuvant chemotherapy
.
Sci Adv
2020
;
6
:
eaay4211
.
9.
Heo
YJ
,
Park
C
,
Yu
D
,
Lee
J
,
Kim
KM
. 
Reproduction of molecular subtypes of gastric adenocarcinoma by transcriptome sequencing of archival tissue
.
Sci Rep
2019
;
9
:
9675
.
10.
Keegan
LP
,
Leroy
A
,
Sproul
D
,
O'Connell
MA
. 
Adenosine deaminases acting on RNA (ADARs): RNA-editing enzymes
.
Genome Biol
2004
;
5
:
209
.
11.
Chen
L
,
Li
Y
,
Lin
CH
,
Chan
TH
,
Chow
RK
,
Song
Y
, et al
Recoding RNA editing of AZIN1 predisposes to hepatocellular carcinoma
.
Nat Med
2013
;
19
:
209
16
.
12.
Qin
YR
,
Qiao
JJ
,
Chan
TH
,
Zhu
YH
,
Li
FF
,
Liu
H
, et al
Adenosine-to-inosine RNA editing mediated by ADARs in esophageal squamous cell carcinoma
.
Cancer Res
2014
;
74
:
840
51
.
13.
Shigeyasu
K
,
Okugawa
Y
,
Toden
S
,
Miyoshi
J
,
Toiyama
Y
,
Nagasaka
T
, et al
AZIN1 RNA editing confers cancer stemness and enhances oncogenic potential in colorectal cancer
.
JCI Insight
2018
;
3
:
e99976
.
14.
Lazzari
E
,
Mondala
PK
,
Santos
ND
,
Miller
AC
,
Pineda
G
,
Jiang
Q
, et al
Alu-dependent RNA editing of GLI1 promotes malignant regeneration in multiple myeloma
.
Nat Commun
2017
;
8
:
1922
.
15.
Teoh
PJ
,
An
O
,
Chung
TH
,
Chooi
JY
,
Toh
SHM
,
Fan
S
, et al
Aberrant hyperediting of the myeloma transcriptome by ADAR1 confers oncogenicity and is a marker of poor prognosis
.
Blood
2018
;
132
:
1304
17
.
16.
Hu
X
,
Wan
S
,
Ou
Y
,
Zhou
B
,
Zhu
J
,
Yi
X
, et al
RNA over-editing of BLCAP contributes to hepatocarcinogenesis identified by whole-genome and transcriptome sequencing
.
Cancer Lett
2015
;
357
:
510
9
.
17.
Rueter
SM
,
Dawson
TR
,
Emeson
RB
. 
Regulation of alternative splicing by RNA editing
.
Nature
1999
;
399
:
75
80
.
18.
Borchert
GM
,
Gilmore
BL
,
Spengler
RM
,
Xing
Y
,
Lanier
W
,
Bhattacharya
D
, et al
Adenosine deamination in human transcripts generates novel microRNA binding sites
.
Hum Mol Genet
2009
;
18
:
4801
7
.
19.
Kawahara
Y
,
Zinshteyn
B
,
Chendrimada
TP
,
Shiekhattar
R
,
Nishikura
K
. 
RNA editing of the microRNA-151 precursor blocks cleavage by the Dicer-TRBP complex
.
EMBO Rep
2007
;
8
:
763
9
.
20.
Morita
Y
,
Shibutani
T
,
Nakanishi
N
,
Nishikura
K
,
Iwai
S
,
Kuraoka
I
. 
Human endonuclease V is a ribonuclease specific for inosine-containing RNA
.
Nat Commun
2013
;
4
:
2273
.
21.
Scadden
AD
. 
The RISC subunit Tudor-SN binds to hyper-edited double-stranded RNA and promotes its cleavage
.
Nat Struct Mol Biol
2005
;
12
:
489
96
.
22.
Zhang
Z
,
Carmichael
GG
. 
The fate of dsRNA in the nucleus: a p54(nrb)-containing complex mediates the nuclear retention of promiscuously A-to-I edited RNAs
.
Cell
2001
;
106
:
465
75
.
23.
Fumagalli
D
,
Gacquer
D
,
Rothe
F
,
Lefort
A
,
Libert
F
,
Brown
D
, et al
Principles governing A-to-I RNA editing in the breast cancer transcriptome
.
Cell Rep
2015
;
13
:
277
89
.
24.
Maas
S
,
Patt
S
,
Schrey
M
,
Rich
A
. 
Underediting of glutamate receptor GluR-B mRNA in malignant gliomas
.
Proc Natl Acad Sci U S A
2001
;
98
:
14687
92
.
25.
Paz
N
,
Levanon
EY
,
Amariglio
N
,
Heimberger
AB
,
Ram
Z
,
Constantini
S
, et al
Altered adenosine-to-inosine RNA editing in human cancer
.
Genome Res
2007
;
17
:
1586
95
.
26.
Jiang
Q
,
Crews
LA
,
Barrett
CL
,
Chun
HJ
,
Court
AC
,
Isquith
JM
, et al
ADAR1 promotes malignant progenitor reprogramming in chronic myeloid leukemia
.
Proc Natl Acad Sci U S A
2013
;
110
:
1041
6
.
27.
Chan
TH
,
Lin
CH
,
Qi
L
,
Fei
J
,
Li
Y
,
Yong
KJ
, et al
A disrupted RNA editing balance mediated by ADARs (Adenosine DeAminases that act on RNA) in human hepatocellular carcinoma
.
Gut
2014
;
63
:
832
43
.
28.
Chan
TH
,
Qamra
A
,
Tan
KT
,
Guo
J
,
Yang
H
,
Qi
L
, et al
ADAR-mediated RNA editing predicts progression and prognosis of gastric cancer
.
Gastroenterology
2016
;
151
:
637
50
.
29.
Sapiro
AL
,
Shmueli
A
,
Henry
GL
,
Li
Q
,
Shalit
T
,
Yaron
O
, et al
Illuminating spatial A-to-I RNA editing signatures within the Drosophila brain
.
Proc Natl Acad Sci U S A
2019
;
116
:
2318
27
.
30.
Tan
MH
,
Li
Q
,
Shanmugam
R
,
Piskol
R
,
Kohler
J
,
Young
AN
, et al
Dynamic landscape and regulation of RNA editing in mammals
.
Nature
2017
;
550
:
249
54
.
31.
Wahlstedt
H
,
Daniel
C
,
Enstero
M
,
Ohman
M
. 
Large-scale mRNA sequencing determines global regulation of RNA editing during brain development
.
Genome Res
2009
;
19
:
978
86
.
32.
Sansam
CL
,
Wells
KS
,
Emeson
RB
. 
Modulation of RNA editing by functional nucleolar sequestration of ADAR2
.
Proc Natl Acad Sci U S A
2003
;
100
:
14018
23
.
33.
Aktas
T
,
Avsar Ilik
I
,
Maticzka
D
,
Bhardwaj
V
,
Pessoa Rodrigues
C
,
Mittler
G
, et al
DHX9 suppresses RNA processing defects originating from the Alu invasion of the human genome
.
Nature
2017
;
544
:
115
9
.
34.
Hong
H
,
An
O
,
Chan
THM
,
Ng
VHE
,
Kwok
HS
,
Lin
JS
, et al
Bidirectional regulation of adenosine-to-inosine (A-to-I) RNA editing by DEAH box helicase 9 (DHX9) in cancer
.
Nucleic Acids Res
2018
;
46
:
7953
69
.
35.
Huang
H
,
Kapeli
K
,
Jin
W
,
Wong
YP
,
Arumugam
TV
,
Koh
JH
, et al
Tissue-selective restriction of RNA editing of CaV1.3 by splicing factor SRSF9
.
Nucleic Acids Res
2018
;
46
:
7323
38
.
36.
Quinones-Valdez
G
,
Tran
SS
,
Jun
HI
,
Bahn
JH
,
Yang
EW
,
Zhan
L
, et al
Regulation of RNA editing by RNA-binding proteins in human cells
.
Commun Biol
2019
;
2
:
19
.
37.
Behm
M
,
Wahlstedt
H
,
Widmark
A
,
Eriksson
M
,
Ohman
M
. 
Accumulation of nuclear ADAR2 regulates adenosine-to-inosine RNA editing during neuronal development
.
J Cell Sci
2017
;
130
:
745
53
.
38.
Han
L
,
Diao
L
,
Yu
S
,
Xu
X
,
Li
J
,
Zhang
R
, et al
The genomic landscape and clinical relevance of A-to-I RNA editing in human cancers
.
Cancer Cell
2015
;
28
:
515
28
.
39.
Eisenhauer
EA
,
Therasse
P
,
Bogaerts
J
,
Schwartz
LH
,
Sargent
D
,
Ford
R
, et al
New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1)
.
Eur J Cancer
2009
;
45
:
228
47
.
40.
Ramaswami
G
,
Zhang
R
,
Piskol
R
,
Keegan
LP
,
Deng
P
,
O'Connell
MA
, et al
Identifying RNA editing sites using RNA sequencing data alone
.
Nat Methods
2013
;
10
:
128
32
.
41.
An
O
,
Tan
KT
,
Li
Y
,
Li
J
,
Wu
CS
,
Zhang
B
, et al
CSI NGS portal: an online platform for automated NGS data analysis and sharing
.
Int J Mol Sci
2020
;
21
.
42.
Bolger
AM
,
Lohse
M
,
Usadel
B
. 
Trimmomatic: a flexible trimmer for Illumina sequence data
.
Bioinformatics
2014
;
30
:
2114
20
.
43.
Haeussler
M
,
Zweig
AS
,
Tyner
C
,
Speir
ML
,
Rosenbloom
KR
,
Raney
BJ
, et al
The UCSC Genome Browser database: 2019 update
.
Nucleic Acids Res
2019
;
47
:
D853
D8
.
44.
O'Leary
NA
,
Wright
MW
,
Brister
JR
,
Ciufo
S
,
Haddad
D
,
McVeigh
R
, et al
Reference sequence (RefSeq) database at NCBI: current status, taxonomic expansion, and functional annotation
.
Nucleic Acids Res
2016
;
44
:
D733
45
.
45.
Cunningham
F
,
Achuthan
P
,
Akanni
W
,
Allen
J
,
Amode
MR
,
Armean
IM
, et al
Ensembl 2019
.
Nucleic Acids Res
2019
;
47
:
D745
D51
.
46.
Frankish
A
,
Diekhans
M
,
Ferreira
AM
,
Johnson
R
,
Jungreis
I
,
Loveland
J
, et al
GENCODE reference annotation for the human and mouse genomes
.
Nucleic Acids Res
2019
;
47
:
D766
D73
.
47.
Li
H
,
Durbin
R
. 
Fast and accurate short read alignment with Burrows-Wheeler transform
.
Bioinformatics
2009
;
25
:
1754
60
.
48.
Li
H
,
Handsaker
B
,
Wysoker
A
,
Fennell
T
,
Ruan
J
,
Homer
N
, et al
The sequence Alignment/Map format and SAMtools
.
Bioinformatics
2009
;
25
:
2078
9
.
49.
The 1000 Genomes Project Consortium
,
Auton
A
,
Brooks
LD
,
Durbin
RM
,
Garrison
EP
,
Kang
HM
, et al
A global reference for human genetic variation
.
Nature
2015
;
526
:
68
74
.
50.
Sherry
ST
,
Ward
MH
,
Kholodov
M
,
Baker
J
,
Phan
L
,
Smigielski
EM
, et al
dbSNP: the NCBI database of genetic variation
.
Nucleic Acids Res
2001
;
29
:
308
11
.
51.
Kent
WJ
. 
BLAT–the BLAST-like alignment tool
.
Genome Res
2002
;
12
:
656
64
.
52.
Wang
K
,
Li
M
,
Hakonarson
H
. 
ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data
.
Nucleic Acids Res
2010
;
38
:
e164
.
53.
Barter
RL
,
Yu
B
. 
Superheat: An R package for creating beautiful and extendable heatmaps for visualizing complex data
.
J Comput Graph Stat
2018
;
27
:
910
22
.
54.
Kuleshov
MV
,
Jones
MR
,
Rouillard
AD
,
Fernandez
NF
,
Duan
Q
,
Wang
Z
, et al
Enrichr: a comprehensive gene set enrichment analysis web server 2016 update
.
Nucleic Acids Res
2016
;
44
:
W90
7
.
55.
Tang
SJ
,
Shen
H
,
An
O
,
Hong
H
,
Li
J
,
Song
Y
, et al
Cis- and trans-regulations of pre-mRNA splicing by RNA editing enzymes influence cancer development
.
Nat Commun
2020
;
11
:
799
.
56.
Chigaev
M
,
Yu
H
,
Samuels
DC
,
Sheng
Q
,
Oyebamiji
O
,
Ness
S
, et al
Genomic positional dissection of RNA editomes in tumor and normal samples
.
Front Genet
2019
;
10
:
211
.
57.
Roth
SH
,
Levanon
EY
,
Eisenberg
E
. 
Genome-wide quantification of ADAR adenosine-to-inosine RNA editing activity
.
Nat Methods
2019
;
16
:
1131
8
.
58.
Hansson
LE
,
Nyren
O
,
Hsing
AW
,
Bergstrom
R
,
Josefsson
S
,
Chow
WH
, et al
The risk of stomach cancer in patients with gastric or duodenal ulcer disease
.
N Engl J Med
1996
;
335
:
242
9
.
59.
Cho
CJ
,
Jung
J
,
Jiang
L
,
Lee
EJ
,
Kim
DS
,
Kim
BS
, et al
Combinatory RNA-sequencing analyses reveal a dual mode of gene regulation by ADAR1 in gastric cancer
.
Dig Dis Sci
2018
;
63
:
1835
50
.
60.
Xuan
Y
,
Wang
Y
. 
Long non-coding RNA SNHG3 promotes progression of gastric cancer by regulating neighboring MED18 gene methylation
.
Cell Death Dis
2019
;
10
:
694
.
61.
Paz-Yaacov
N
,
Bazak
L
,
Buchumenski
I
,
Porath
HT
,
Danan-Gotthold
M
,
Knisbacher
BA
, et al
Elevated RNA editing activity is a major contributor to transcriptomic diversity in tumors
.
Cell Rep
2015
;
13
:
267
76
.
62.
Wang
Y
,
Xu
X
,
Yu
S
,
Jeong
KJ
,
Zhou
Z
,
Han
L
, et al
Systematic characterization of A-to-I RNA editing hotspots in microRNAs across human cancers
.
Genome Res
2017
;
27
:
1112
25
.
63.
Blow
M
,
Futreal
PA
,
Wooster
R
,
Stratton
MR
. 
A survey of RNA editing in human brain
.
Genome Res
2004
;
14
:
2379
87
.
64.
Sharpnack
MF
,
Chen
B
,
Aran
D
,
Kosti
I
,
Sharpnack
DD
,
Carbone
DP
, et al
Global transcriptome analysis of RNA abundance regulation by ADAR in lung adenocarcinoma
.
EBioMedicine
2018
;
27
:
167
75
.
65.
Stellos
K
,
Gatsiou
A
,
Stamatelopoulos
K
,
Perisic Matic
L
,
John
D
,
Lunella
FF
, et al
Adenosine-to-inosine RNA editing controls cathepsin S expression in atherosclerosis by enabling HuR-mediated post-transcriptional regulation
.
Nat Med
2016
;
22
:
1140
50
.