Involving the microRNA Targetome in Esophageal-Cancer Development and Behavior

Esophageal cancer (EC) is the eighth most common and sixth leading cause of cancer-related mortality in the world. Despite breakthroughs in EC diagnosis and treatment, patients with complete pathologic response after being submitted to chemoradiotherapy are still submitted to surgery, despite its high morbidity. Single-nucleotide polymorphisms (SNPs) in miRNA, miRNA-binding sites, and in its biogenesis pathway genes can alter miRNA expression patterns, thereby influencing cancer risk and prognosis. In this review, we systematized the information available regarding the impact of these miR-SNPs in EC development and prognosis. We found 34 miR-SNPs that were associated with EC risk. Despite the promising applicability of these miR-SNPs as disease biomarkers, they still lack validation in non-Asian populations. Moreover, there should be more pathway-based approaches to evaluate the cumulative effect of multiple unfavorable genotypes and, consequently, identify miR-SNPs signatures capable of predicting EC therapy response and prognosis.


Introduction
Esophageal cancer (EC) is the eighth most common cancer and the sixth leading cause of cancer-related mortality in the world, with an estimated 400,000 deaths in 2012 [1,2]. The vast majority of ECs occur as either esophageal squamous cell carcinoma (ESCC) in the middle or upper third of the esophagus, or as esophageal adenomarcinomas (EAC) in the distal third or esophagogastric junction [3]. EC incidence is threefold higher in men than in women, and peak incidence rates occur in Southern and Eastern Africa and in Eastern Asia [2]. The global incidence of ESCC has remained more or less the same, whereas a rapid increase in the incidence of EAC has been observed in the United States and Western Europe [3][4][5]. In fact, EAC incidence is expected to rise dramatically in many Western countries in the coming years [6,7]. EAC typically arises from a metaplastic epithelium known as Barrett's Esophagus (BE), and established risk factors for both EAC and BE include gastroesophageal reflux disease, European ancestry, male sex, obesity, and tobacco smoking [8].
miRNA targetome polymorphisms, along with in silico insight into the pathways affected by such variations [14]. Despite the importance of these findings, validation studies are still lacking.
MiR-SNPs seem to represent an indispensable pool of novel molecular biomarkers and have recently come into focus regarding their possible role in the development of cancer. Hence, the scope of this review is to gather and systematize the information available regarding the impact of miR-SNPs in EC development and prognosis.

Evidence Acquisition
A literature search in PubMed was conducted using the search terms "miRNA", "polymorphisms", "SNPs", and "esophageal cancer". The articles were selected by relevance of their findings, namely, the significant association of miR-SNPs and esophageal cancer. Literature analysis includes scientific papers published in the last ten years (between 2008 and 2018). Obtained scientific papers were manually curated in order to determine associations between miR-SNPs and EC. Of the 47 papers found, 17 were excluded. The exclusion criteria for the collected papers were as follows: (1) no association between miR-SNPs and EC; (2) association with a benign tumor; and (3) individual papers that were already included in meta-analysis, collected for this study. For each study, information was extracted concerning the following characteristics: the name of the miRNA, SNP rs number, SNP effect on esophageal cancer (e.g., cancer risk, prognosis, therapy response), ethnicity, type of study (e.g., case control, association study, meta-analysis) and number of cases and controls used. Since BE is an established risk for, and the only known precursor of, EAC, we included the miR-SNPs that were common to both BE and EAC in our study.

Evidence Synthesis
The pooled information is synthesized in Figure 1, where we divided the EC-relevant miR-SNPs into three categories: (1) SNPs in miRNA-coding genes, (2) SNPs in miRNA-binding sites, and (3) SNPs in biogenesis machinery.

SNPs in miRNA Loci
MiRNA genes are scattered among each human chromosome. With the exception of chromosome Y, they can be encoded in independent transcription units, polycistronic clusters, or within the introns of protein-coding genes [22]. MiRNA profiling has revealed that most miRNAs are significantly downregulated in cancer [20]. Calin and colleagues mapped the chromosomal location of all known miRNA genes and discovered that many are located in regions that are frequently involved in chromosomal alterations, such as deletions or amplifications, usually found in many types of cancers [23]. Additionally, Lu and colleagues found that SNPs occur less frequently, but are more constrained in miRNAs associated with diseases when compared to the other miRNAs [24]. The majority of SNPs in miRNA-coding genes usually occur in pre-miRNAs and can be responsible for changes in stem-loop structures, consequently affecting the production of mature miRNAs [25]. Gong and colleagues performed a genome-wide scan in human pre-miRNAs, miRNA flanking regions, and target sites, and their results showed that approximately 40% of pre-miRNAs contain at least one polymorphism, and 48 SNPs were found in functionally important seed regions. The authors also observed that the SNP density of pre-miRNAs is lower than that of flanking regions, and SNP density of miRNA seed regions is significantly lower than in the pre-miRNAs and flanking regions [25]. Despite being quite rare in seed sequences, SNPs in these regions can lead to the creation or disruption of putative binding sites, consequently altering the total numbers of putative targets [26].
In this review, we found 22 studies that related a total of 13 SNPs in miRNA genes associated with an EC outcome ( Table 1). The most studied SNPs were miR-423 rs6505162, miR-196a-2 rs1161491,3 and miR-146a rs2910164, and they were all located in pre-miRNA regions with impact in the respective mature miRNA production. The miR-423 rs6505162 SNP is located in pre-miR-423 and maps to 17q11.2, with a nucleotide alteration from C to A [27]. Since both miR-423-3p and miR-423-5p are produced by the pre-miRNA of miR-423, it is possible for the polymorphism of pre-miR-423 to play different roles in cancer progression, and in different types of cancer [27][28][29]. Despite the lack of studies about mature miR-423 in esophageal cancer, miRNA-423-5p expression was associated with cell proliferation and invasion in gastric cancer cells [30]. The miR-146a rs2910164 G > C SNP is located in the precursor stem-loop region, opposite to the mature miRNA-146a sequence and involves a change from a G:U pair to a C:U mismatch, with an impact on mature miRNA-146a levels [31]. MiR-146a expression was reported as dramatically decreased in ESCC tissue and it was associated with a worse prognosis [32]. The miR-196a-2 rs11614913 C > T SNP is located in the stem-loop region opposite to the mature miR-196a-2, and the nucleotide change from C to T was suggested to alter the levels of the mature 3 passenger (3p) strand of miR-196a2 and the activity of its target mRNAs [31]. The mature miR-196a was reported as upregulated in ESCC and was suggested as a potential diagnosis biomarker and therapeutic target for this neoplasia [33,34].

SNPs in miRNA-Binding Sites
SNPs occurring in noncoding regions can affect transcriptional regulation or post-transcriptional gene expression, thereby affecting mRNA half life and resulting in altered protein levels trough the deregulation of miRNA-mRNA binding [56]. The most widely studied SNPs in noncoding regions are the SNPs located in the 3 -UTR of mRNAs, also known as poly-miRTSs [57]. Roughly 180,000 SNPs in the human genome that are located in the 3 -UTR region were identified, along with about 2600 mature miRNA sequences that are deposited in the mirBase (v.21), which suggests that these SNPs may introduce miRNA-binding changes [20]. Additionally, by performing a genome-wide scan, Gong and colleagues found a total of 98,008 possible functional SNPs in 3 -UTRs [25]. Despite the fact that the majority of the studies about polymorphisms in miRNA targets focus on the SNPs in 3 UTRs, it is important to note that some studies revealed that miRNAs could also bind to 5 UTRs or coding sequences of target mRNAs, suggesting that these variants could also affect miRNA regulation [58,59]. Functional SNPs in the miRNA target regions are likely to alter gene expression via affecting miRNA targeting, namely, through the creation or disruption of miRNA-binding sites. In this review, we found eight studies relating nine different SNPs in miRNA targets (8 in 3 -UTR regions and 1 in a coding sequence) with association with EC risk and outcome ( Table 2). In a study that evaluates the impact of SNP regulation of miRNA expression in colon-cancer risk, KIAA023 rs1053667 was found to be associated with differential expression of one of its target miRNAs, miR-19b-3p, in normal colon tissue when compared to tumor tissue [65]. Histone lysine methyltransferase (SET8) rs16917496 results in a C to T transition that might destroy the G:C bond in the miR-502 and SET8 binding site, therefore modulating SET8 expression. The C allele is associated with a perfect complementarity with miR-502, which will lead to efficient mRNA degradation and lower SET8 protein levels. In EC, lower SET8 expression was associated with a better prognosis and survival [17]. The Basigin (BSG) rs11473 consists of a C to T transition that destroys the binding site of miR-483-5p at the 3 -UTR of BSG, resulting in higher mRNA levels of this gene. In a study involving EC patients, those that carried the TT genotype expressed higher levels of BSG mRNA and protein, and consequent presents higher EC risk, compared with patients with the CC genotype carriers [61]. All-trans-retinol dehydrogenase 8 (RDH8) encodes for a short-chain dehydrogenase/reductase enzyme involved in rhodopsin regeneration in the vision pathway and its enzymatic activity has also been linked to estrogen biosynthesis. The rs1644730 is located in the 3 -UTR of RDH8 and its predicted miR-630 binding site [8]. Given the significantly higher incidence of EAC among males versus females, a potential protective effect of estrogen has been proposed, which is in agreement with the fact that patients carrying the rs1644730 A allele presented decreased EAC risk [8]. PTPRT is a tumor suppressor that plays a crucial role in regulating tumorigenesis mechanisms [62]. Two 3 -UTR SNPs, rs2866943 C>T and rs6029959 C>A, located in the binding sites of miR-218 and miR-142-5p, respectively, were studied in EC patients, but only rs2866943 was able to disrupt the inhibitory role of miR-218 on PRPT expression and act as a protective factor in ESCC risk. Patients carrying rs2866943 CT and TT genotypes presented a small tumor size as well as the low probability of metastasis [62]. The Erb-b2 receptor tyrosine kinase 4 (ErbB4) gene, also known as human epidermal growth factor receptor 4 (HER4), has been reported as overexpressed in EC tissue and is also associated with TMN stage and lymph-node metastasis [66]. ErbB4 rs1595066 creates a binding site for miR-200*, a member of miR-200 tumor-suppressor miRNAs, and is associated with a lower EC risk, probably through the downregulation of ErbB4 [42]. BRCA1 is a widely studied tumor suppressor gene and it is deregulated in several cancers. The rs799917 T>C polymorphism located in the BRCA1 coding sequence influences miR-638-mediated regulation of BRCA1 expression [58]. BRCA1 mRNA expression analyses showed that the rs799917 C allele carriers significantly decreased BRCA1 expression in both normal and cancer esophagus tissue compared with T allele carriers, suggesting that lower BRCA1 expression may lead to a higher risk of malignant transformation of esophagus cells [58]. MDM4 is an oncoprotein that negatively regulates p53 function. The rs4245739 A>C SNP, located in the MDM4 3 -UTR, creates a binding site for miR-191, resulting in decreased MDM4 expression [63]. Rs4245739 AC and CC genotype carriers significantly decreased MDM4 expression in normal esophagus tissue compared with AA genotype carriers, indicating consistent genotype-phenotype correlation [63]. The rs6573 SNP is a substitution from A to C, and disrupts the binding of miR-196a to RAP1A 3 -UTR, resulting in a higher constitutive expression of RAP1A, which is a member of the RAS oncogene family [64]. Wang and colleagues observed that RAP1A was overexpressed in ESCC tissue, and correlated with RAP1A rs6573 CC genotype and lymph-node metastasis. The authors also performed an in vitro study where they concluded that RAP1A might function as a promoter for esophageal cancer-cell migration and invasion through matrix metalloproteinase 2 [64].

SNPs in miRNA Processing Machinery
Although the role of miRNA's biogenesis pathway genes in cancer development and its progression has been well established, the association between genetic variants of these pathway genes has been less studied. The occurrence of SNPs in the components of the miRNA biogenesis pathway can affect transcription, processing, transport and target gene identification, consequently affecting the overall expression of miRNAs. In this review, we found three studies relating SNPs in XPO5, Gem-associated protein 3 (GEMIN3), and Gem-associated protein 4 (GEMIN4) genes with EC risk outcome, and all of the SNPs reported were associated with a better prognosis (Table 3).
XPO5 is a key factor in this process, as it is responsible for the nuclear export of the pre-miRNA to the cytoplasm, where it is further processed to its final miRNA conformation in order to be loaded to RNA-induced silencing complex to exert its regulatory effect [67]. It has been postulated that XPO5 miRNA regulation can be a limiting step for miRNA development since its impairment can lead to pre-miRNA trapping in the nucleolus and therefore influence cancer risk [68]. The XPO5 rs11077 consists of an A to C transition that leads to the disruption of the miR-617 binding site and the creation of a new binding site for miR-4763-5p in XPO5 3 -UTR, with an impact on XPO5 Cancers 2018, 10, 381 8 of 13 mRNA levels [69]. In fact, patients' carriers of rs11077 AA genotype displayed a trend for high XPO5 expression in ESCC tissues, and these high XPO5 expression levels were also associated with high survival rates [17]. GEMIN3 and GEMIN4 are members of the GEMIN protein family, and are part of the RNA-induced silencing complex (RISC) that participates in the target RNA recognition and repression by mature miRNAs. With the exception of GEMIN4 rs910924 that is located in a 5 -UTR, all the other studied GEMIN SNPs were missense variants, meaning that they resulted in different amino acid sequences that could impact GEMIN3 and GEMIN4 protein structure and consequent miRNA regulatory function [55,70].

Discussion
Major breakthroughs in the diagnosis and treatment of EC have been achieved during the past few decades. However, the selection criteria for operative management after chemoradiotherapy are still lacking, and patients that present a complete pathologic response are still submitted to surgery. Since miRNAs are important in carcinogenesis and are capable of regulating the expression of hundreds of target mRNAs, miR-SNPs may produce more significant functional consequences and represent an ideal candidate for disease prediction.
SNPs in miRNA and miRNA-binding sites can potentially modulate miRNA-mRNA interaction and potentially create or destroy miRNA-binding sites, while those in biogenesis pathway genes can influence miRNA transcription either through altering transcription, processing, or maturation. Due to their impact in cancer development, this new class of SNPs has been widely studied, mainly due to their potential applicability as disease biomarkers. Indeed, in this review we found a total of 34 miR-SNPs that were associated, in their majority, with EC risk and were mostly studied in Asian populations where the incidence of EC is higher. However, despite promise, the impact of these miR-SNPs requires further validation, especially in non-Asian populations where these types of studies are lacking, and the incidence of EC, especially EAC, will rise dramatically in the following years. Additionally, since EC is a type of cancer that involves multiple miR-SNPs with impact on the miRNA targetome, the candidate gene approach that considers one or few genes/SNPs at a time can give us the functional impact of that genetic variant in cancer risk or survival, but fails to relate that information with the molecular pathways involved. As such, more pathway-based approaches that evaluate the cumulative effect of multiple unfavorable genotypes on the miRNA targetome are needed to identify signatures of genetic variations capable of predicting EC therapy response and prognosis. Furthermore, it is also important to take into account that sometimes changes in a gene's mRNA levels are not reflected in its protein levels. Hence, when studying SNPs that alter miRNA:mRNA binding capacity and, consequently, mRNA regulation, they should be accompanied by the monitorization of the protein levels. This type of approaches would shed some light on the fine regulatory mechanisms by which these variations contribute to EC pathogenesis instead of only focusing in cancer risk.

Conclusions
Despite the complexity of the functional effects of SNPs that occur in noncoding regions, which is the case of miR-SNPs, more attention has been paid to this recent class of genetic polymorphisms mainly due to their potential impact in cancer. In EC, the majority of the studies focuses on the association cancer risk and overall survival, but are lacking in terms of therapy response and prognosis. Despite the potential of miR-SNPs as biomarkers, more studies are needed, especially in non-Asian populations, in order to validate their application in the clinical practice.