The identification of RNA editing sites deeply depends on sequencing technology and bioinformatics approaches. We developed a pipeline for identifying RNA editing events in primary gastric cancer and normal tissues by screening RNA differences from reference genome followed by successive and rigorous filtering criteria. Most of previous studies have used coupled RNA and DNA sequences to identify editing events [28, 81], by the contrary, we identified RNA editing sites using RNA sequencing data alone. Our analyses found significant number of editing sites, vast majority of them harbored in 3´UTR regions, which has been reported in previous studies [80, 82]. Also a few novel editing sites were found, which were reported for the first time in the current study. Although the number of identified RNA editing sites was huge, most of the sites exhibited low editing levels and approximately half of the identified sites were edited in less than 27% of their related transcripts.
Our analyses found that the RNA editing sites were highly associated with both number of protein coding genes and Alu elements distribution in the genome. Also, frequency of editing sites were correlated with size of chromosomes. These results are in a good agreement with Chigaev et al. study, who reported that correlation of editing frequency with protein coding genes is stronger than lincRNA density [80]. However, these correlation could result from the bias of the library preparation step of RNA sequencing projects. Since oligo-dT primers apply to capture the RNA through the poly-A tail, most of the reads will be related to protein coding genes.
To date, no specific sequence has been found that characterize editing sites of any of the ADAR enzymes. However, in the neighborhood of edited adenosine, there are preferred and opposed preferences. Consistent with previous studies, there was an over-representation of guanosine in the neighboring position downstream, while guanosine was depleted in the upstream neighboring position [26, 82]. Since some of adenine bases in the right context do not edit, other features proposed to be involved in determination of editing. Daniel et al. described editing inducer elements distance from the edited adenine, which increase the editing efficiency and specificity of a highly edited site [20]. Wong et al. reported that editing efficiency is strongly influenced by the base opposing the edited adenosine. They found that when there is an A:C mismatch at the editing site, editing by ADAR enzyme was enhanced compared to when A:A or A:G mismatches or A:U base pairs occurred at the same site [21]. Due to the contradictory results, it is difficult to make definitive conclusions about potential editing sites.
We wonder whether RNA editing could function as an additional mechanism contributing to tumorigenesis by generating specific RNA editing sites that are unique to cancer samples. In the search of the answer to this question we found that 28.4% and 32.2% of the identified editing sites were specific to cancer and normal tissues, respectively. These tissue specific editing sites could contribute to cancer initiation and progression, if they located in important gene. Some of cancer-specific editing sites and their role in pathogenesis of cancer have been identified in previous studies. RNA editing of transcription factor PROX1, a candidate tumor suppressor, leads to several missense substitutions including E328G, R334G, and H536R and loses tumor suppressive functions. These editing events have been seen in a number of esophageal, pancreatic, and colon cancer samples, but no such editing is seen in a number of cDNA libraries of many normal tissues [17].
We also found a remarkable number of common editing events between cancer and normal tissues, which their editing levels were significantly different in cancer and normal tissue. Deregulated editing level in cancer and normal common editing sites could be an important contributor in tumorigenesis. Chen et al. reported that RNA editing level of AZIN1 increases by at least 10% in hepatocellular carcinoma compared to adjacent normal liver. The edited isoform compared with wild-type AZIN1 has increased affinity to antizyme, which leads to neutralization of antizyme-mediated degradation of ornithine decarboxylase and cyclin D1 and promotes cell proliferation [83]. In this regard, Han et al. reported a higher level of editing on RHOQ in tumor compared with normal tissue in colorectal cancer, which results in N136S amino acid substitution. This RNA mutation increases RHOQ protein activity, actin cytoskeletal reorganization and invasion potential [84]. On the contrary, hypo-editing of several genes are associated with cancer phenotypes. The pre-mRNA transcript encoding the GluR-B has two functionally important editing sites (Q/R and R/G sites) and the Q/R site almost entirely edited, which is necessity for normal function of receptor. It has been proved, in malignant tissue of human brain tumors, this editing site of GluR-B considerably under-edited compared with control tissues [85]. Our results corroborate that the RNA editing frequency can be regulated in a tissue specific manner, which is consistent with observations reported previously.
Our results showed that the vast majority of editing sites in gastric cancer were located in 3´UTR and up/down stream regions as well as a large number of editing sites were observed in coding regions. According to their genomic location, these RNA editing events could lead to various functional impacts and apply their effects through several dominant mechanisms. First and most important, RNA editing events in exonic region can cause amino acid change and imitate cancer-associated missense mutations. Our pipeline identified 81 editing events with non-synonymous effect, including 12 novel editing events. Notably, we found four missense RNA mutations in mucin family (MUC3A, MUC4 and MUC6). Normal gastric epithelial cells transcribe MUCs, which have several functions including; protection against mechanical and infectious lesions, lubrication and acid resistance [86]. Several studies have been reported that transcription profile of mucins are changed in gastrointestinal cancers, which overall suggests an important role for MUCs in gastric cancer [87–89]. Our results reinforced the hypothesis that inappropriate RNA editing can be involved in gastric cancer development.
Second, RNA editing could affect microRNAs target recognition and subsequently affect the expression profile of the genes. Previous computational analyses suggested that RNA editing tends to avoid microRNA target sites in general, even though RNA editing events have a potential to block the microRNA target recognition. Dysregulation of microRNA target recognition has been linked to cancers [90, 91]. In this context, 44 editing events were found in the present study, where at least one microRNA binding was disrupted. In consistent with our research, Soundararajan et al. identified 652 editing events in lung cancer, which were located in the 3´UTR of 205 target genes and mapped to 932 potential microRNA target binding sites [92]. All together these findings are inconsistent with Liang and Landweber previous computational analyses, where they suggested that RNA editing tends to avoid microRNA target sites in general, even though RNA editing events have a potential to block the microRNA target recognition [93]. It is worth to remind, RNA editing events in addition to disrupting existing microRNA binding sites, could generate novel microRNA regulatory networks. In a completely separate mechanism from what has been mentioned, RNA editing could affect microRNA biosynthesis. miR-142 is highly expressed in hematopoietic tissues, conversely it is not expressed in non-hematopoietic tissues. Also, its expression in patients with acute myeloid leukemia is significantly lower than that in controls. Yang et al. showed that editing of pri-miR-142, leads to suppression of its processing by Drosha and subsequently it degradation [94].
Third, editing of microRNA sequences could alter their binding affinity or target recognition properties. Since microRNAs play a role in nearly all cellular pathways and pathological processes, including cancer initiation and progression, fluctuations of their targeting are an important contributor to cancer [95]. Our analysis revealed 42 editing sites in 17 cancer-associated microRNAs, some of them exclusively edited in cancerous tissue. Consistent with our results, Nigita et al. identified 40 and 18 potential editing sites in Lung Adenocarcinoma and Lung Squamous Cell Carcinoma, respectively [96]. Indeed, our results showed miR-34a, a cancer-specific edited microRNA, was edited in 10 position. Previous studies have been identified this microRNA as a tumor suppressor in gastric cancer cell lines [58]. On the other hand, it was shown miR-34a epigenetically down-regulated or silenced in gastric cancer tissues and cell lines [97]. We therefore speculate that editing in some positions could terminate the function of miR-34a, but further studies are required to confirm this possibility.
To our knowledge, this is the first time to comprehensively characterize editome of normal and cancerous tissue of gastric. Findings of the current study uncovered relatively large number of RNA editing sites, which were unevenly distributed across genome. Editing level of these sites and editing rate of different genes had diverse distribution. We also found a significant number of exclusively edited genes in cancer and normal tissue, which are likely to contribute to cancer initiation and progression.