3’-UTR Polymorphisms in the MiRNA Machinery Genes DROSHA, DICER1, RAN, and XPO5 Are Associated with Colorectal Cancer Risk in a Korean Population

MicroRNAs play an important role in cancer initiation and development. The aim of this study was to investigate whether polymorphisms in miRNA machinery genes are associated with the development of colorectal cancer (CRC). RAN rs14035 CT heterozygotes and T allele carriers (CT + TT) genotypes had lower risk of CRC, while the DICER1 rs3742330, DROSHA rs10719, and XPO5 rs11077 polymorphisms were not associated with CRC in the full study sample. Specifically, male RAN rs14035 CT heterozygotes and XPO5 rs11077 AA genotype (CT/AA) carriers experienced reduced CRC susceptibility (both colon and rectal). Subgroup analysis demonstrated that the combined RAN rs14035 CT + TT genotype was associated with rectal cancer, but not colon cancer. In addition, the DICER1 rs3742330 AG genotype was associated with a significantly increased risk of colon cancer. Stratified analysis revealed the RAN rs14035 combined CT+TT genotype was associated with decreased CRC risk in male patients without diabetes mellitus (DM) and in patients with rectal cancer. In addition, we found the RAN rs14035 CC genotype was related to a decreased risk of CRC with respect to tumor size and metabolism of homocysteine and folate. Furthermore, patients diagnosed with hypertension or DM who carried the DROSHA rs10719 CC genotype showed increased CRC risk, while the XPO5 rs11077 AC+CC genotype led to increased CRC risk in patients with hypertension only. Our results indicate variations in RAN rs14035, DICER1 rs3742330, XPO5 rs11077, and DROSHA rs10719 of Korean patients are significantly associated with their risk of CRC.


Study population
Between June 2005 and January 2009, 808 blood samples were collected from the study group, consisting of 408 CRC patients and 400 randomly selected non-CRC (control) subjects, following a health screening at CHA Bundang Medical Center (Seongnam, South Korea). The CRC group included 167 consecutive patients with rectal cancer and 241 consecutive patients with colon cancer. Pathological staging frequencies after curative tumor resection were as follows: tumor node metastasis (TNM) stage I, n = 43 (10.5%); stage II, n = 173 (42.4%); stage III, n = 153 (37.5%); and stage IV, n = 39 (9.6%). Patients with high baseline blood pressure ( systolic 140 mmHg or diastolic 90 mmHg) on more than one occasion or with a history of taking antihypertensive medication were classified as having hypertension (HTN). Patients with high fasting plasma glucose ( 126 mg/dl), who were taking an oral hypoglycemic agent, or who had a history of insulin treatment were classified as having DM. All study subjects were ethnic Koreans and provided written informed consent for study participation. The study protocol was approved by the Institutional Review Board of CHA Bundang Medical Center.
Genotypes determined by RFLP analysis were confirmed by two independent investigators and by sequencing 10% of the samples. For RAN rs14035 in the control group and XPO5 rs11077 in CRC patients, genotypes were verified three times to rule out possible errors attributed to violation of Hardy-Weinberg equilibrium (HWE).

Statistical analysis
To compare baseline characteristics between cases and controls, we used Chi-squared tests to analyze categorical data (sex, HTN, and DM) and the Student's t-test to analyze continuous data. Allele frequencies were calculated to identify deviations from HWE. Adjusted odds ratios (AOR) and 95% confidence intervals (CI) were used to examine the association between DICER1, DROSHA, RAN, XPO5 polymorphisms and CRC occurrence with GraphPad Prism 4.0 (La Jolla, CA, USA) and MedCalc version 12.1.4 (Panmun Education, Seoul, Korea). Gene interactions among SNP loci were analyzed using multifactor dimensionality reduction (MDR) with MDR version 2.0 (www.epistasis.org) [27][28][29]. We determined the best multilocus combinations based on MDR identification of the most significant models using maximized crossvalidation values. HAPSTAT version 3.0 (www.bios.unc.edu/~lin/hapstat/) was used to estimate haplotype frequencies for the polymorphisms determined by MDR analysis to have strong synergistic effects [30]. The false-positive discovery rate (FDR) correction was used to adjust multiple comparison tests.

Study subject characteristics
The 408 CRC cases included 170 males and 238 females with an overall mean age was 61.55 ±12.26 years (mean±SD; Table 1). Among the cases, 241 (59.1%) had colon cancer and 167 (40.9%) had rectal cancer. No statistically significant differences in age or sex were identified between CRC cases and controls (P = 0.772 and 0.847, respectively). However, both HTN (60.5%, P = 0.004) and DM (34.1%, P < 0.0001) were significantly more common among CRC cases than in controls.

Genotype frequencies
Genotype and allele frequencies for the four miRNA machinery genes in CRC cases and controls are shown in Table 2. Genotype distributions in both groups displayed no departure from HWE. RAN rs14035 CT heterozygotes had significantly decreased CRC risk relative to wild type homozygotes (AOR = 0.698; 95% CI, 0.511-0.952; P = 0.023). Similarly, the combined CT + TT genotype was associated with decreased CRC risk (AOR = 0.690; 95% CI, 0.510-0.934; P = 0.016). However, neither association remained statistically significant after controlling for multiple comparisons using the FDR correction. No association between CRC risk and other polymorphisms was found. In addition, we observed no significant relationship between genotype frequencies for the four miRNA machinery genes and CRC patient survival (Table A in S1 File).

Discussion
Previous studies have shown that several types of cancer are associated with alterations to miRNA machinery genes, such as DROSHA, DICER1, XPO5, and AGO2 [31][32][33]. Although modifications to these genes can significantly affect initiation and progression of cancer, the role of genetic variation in miRNA machinery genes during CRC development is not fully understood. To the best of our knowledge, this is the first report to evaluate the association between RAN polymorphisms and CRC; we found RAN rs14035 CT heterozygotes and T allele carriers (CT + TT genotypes) had a lower CRC risk than individuals with other genotypes. In addition, in a subgroup analysis targeting cancer type, we determined the combined RAN rs14035 CT + TT genotype was associated with decreased rectal cancer risk. RAN encodes a small G protein essential for the translocation of RNA and proteins through the nuclear pore complex [34]. When RAN-GTP is depleted as a result of RAN guanine nucleotide exchange factor inhibition, pre-miRNA export is greatly reduced, indicating miRNA transport is mediated by a RAN-GTP-binding export receptor [35]. Therefore, it is possible that RAN mutations play an essential role in pathology-related changes to miRNA transport and expression. The RAN protein is also a well-known downstream modulator of the PI3K signaling pathway, which mediates cancer cell invasion and metastasis [36]. Moreover, CRC tissues exhibit significantly higher levels of RAN expression than normal colorectal epithelial cells, which have been positively associated with depth of invasion, lymph node metastases, distant metastases, tumor differentiation, and tumor-node-metastasis stage [37]. Therefore, the rs14035 polymorphism in RAN's 3'-UTR may influence its function as a downstream modulator of CRC development.
Faggad et al. [38] reported that reduced DICER1 expression may contribute to tumor progression in CRC. Consistent with this report, Dewi et al. [39] determined that DICER1-deficient CRC tissues have a reduced number of alkaline phosphatase-positive reprogrammed cells relative to wild type cells. However, another report showed no significant difference in expression levels of long 3'-UTR DICER1 mRNA between CRC tumors and normal tissues [40]. Thus, the effects of up-and down regulation of DICER1 expression on CRC susceptibility are unclear. Our subgroup analysis, which target specific cancer type, revealed that DICER1 rs3742330 AG heterozygotes showed a significantly increased risk of colon cancer. Similarly, an association between this polymorphism, located in the gene's 3'-UTR, and T-cell lymphoma survival has also been reported, perhaps because its 3 0 -UTR is important for mRNA transcript stability [41]. Moreover, another SNP located in DICER1's 3'-UTR (rs1057035) may contribute to oral cancer risk by affecting miRNA binding to DICER1 [42]. Although the DICER1 rs3742330 AG genotype was associated with a significantly increased risk of colon cancer in our data, we observed no significant relationship between the genotype frequencies for the DICER1 rs3742330 and CRC patient survival. DICER1rs3742330 was reported to be associated with increased survival in T-cell lymphoma [41]. These inconsistent results across different cancers suggest that the abnormal expression patterns of miRNA pathway genes might be associated with tissue-specific effects. The SNP is located in the 3'-UTR, a region that is important for DICER1 mRNA stability, polyadenylation, translation efficiency, and localization. The 3'-UTR contains binding sites for microRNAs. Many studies have reported the regulatory roles of miRNAs in genetic networks underlying various cellular pathways, indicating that oncogenic miRNAs might be involved in the genetic networks regulating the functional pathway deregulated in different cancer cells [7,43]. DICER1 rs3742330 has been identified as the target site of two miRNAs-miR-3622a-5p [44] and miR-5582-5p [45]. However, the regulation of these miRNAs of DICER1 rs3742330 in CRC and T-cell lymphoma has not been experimentally validated. The expression levels of DICER1 have global effects on the biogenesis of miRNA. For instance, a pattern of down-regulation of DICER1 is associated with poor prognosis in skin, lung, breast, and ovarian cancers, among other cancers [46]. However, the analysis of prostate cancer and CRC shows overexpression of DICER1 and other miRNA biogenesis genes in metastatic lesions [47]. Mucoepidermoid cancers arising in the throat or upper esophagus exhibit both over-and under-expression of DICER1 compared with normal tissues from the throat and esophagus [48]. The miRNA-SNP of rs3742330 of DICER1 has been identified for its association with the cancer outcome of CRC, T-cell lymphoma, oral premalignant lesions and renal cell carcinoma [49][50][51]42].The AA allele of rs3742330 located in the DICER1 gene exhibited a significantly increased risk of CRC. However, in T-cell lymphoma, Patients carrying the GG genotype had a significantly increased overall survival (OS) compared with those carrying the GA and AA genotypes. In oral premalignant lesions, patients carrying the GA and AA genotypes had a significantly increased risk of OPL, and there was no significant relationship between the genotype frequencies for the DICER1 rs3742330 and renal cell carcinoma. These studies, together with our study, suggest genetic polymorphisms in cancer may play a different role in the regulation of miRNA machinery genes, including DICER1, and therefore influence the prognosis of cancer.
In the miRNA processing system, the XPO5/RAN-GTP complex mediates the nuclear transport of pre-miRNAs. XPO5 mutants display reduced miRNA processing levels and target inhibition, while restored XPO5 acts as a tumor suppressor that reverses the impaired export of pre-miRNA in colon cancer [52]. Previous studies have reported associations between the XPO5 rs11077 SNP (located in the gene's 3'UTR) and esophageal cancer, non-small cell lung cancer, and multiple myeloma [53,54]. In addition, associations between polymorphisms in the XPO5 and AGO1 genes and renal cell carcinoma risk have been reported [55]. In particular, XPO5 rs11077 was linked with increased risk of renal cell carcinoma in a recessive model [55]. In support of these earlier reports, we found that RAN rs14035 CT heterozygotes and XPO5 rs11077 AA genotype carriers experienced reduced susceptibility to CRC, a link that was especially apparent in male patients. Gender differences have been documented among patients diagnosed with CRC. A higher CRC age-adjusted incidence among men than among women has persisted over the past 30 years, but the underlying cause remains unclear [56]. In our study, RAN rs14035 CT/XPO5 rs11077 AA and RAN rs14035 CT/DROSHA rs10719 AA genotypes were associated with reduced susceptibility to CRC in males. In addition, RAN rs14035 CT heterozygotes and DROSHA rs10719 AA genotype carriers reduced susceptibility to CRC subtype. Polymorphisms within the 3'-UTRs of miRNA machinery genes may be responsible for locally altered mRNA secondary structures. For example, 3'-UTR polymorphisms can result in different secondary mRNA structures and distinct allele-dependent differences in mRNA stability [57]. Similarly, several reports have suggested that SNPs outside of the miRNA binding site result in altered miRNA binding due to allele-dependent changes in secondary mRNA structure [58]. In addition, alterations in secondary structure can interfere with RNAbinding proteins, which can lead to altered mRNA stability [59]. Thus, the SNPs located in the 3'-UTRs of RAN, DICER1 and XPO5 may affect mRNA stability and subsequent expression. However, the results from this study require validation by another CRC case-control study and by laboratory-based expression methods.
Recent evidence shows that components of metabolic syndrome (MetS), including aging, HTN, and DM, may also be associated with the risk of developing CRC [60,61]. The relationship between individual components of MetS and CRC risk has been analyzed by several studies [62,63]. But an inconsistent link between MetS and its components on CRC mortality was observed in another study [64]. Diabetes mellitus type 2 is associated with a 20%-60% increased risk of CRC [65], as insulin resistance may promote carcinogenesis directly by stimulating colonic cell growth [66]. Our data show that the RAN rs14035 CT + TT genotype may result in lower CRC risk in patients without DM, but not in patients with DM. However, a recent metaanalysis demonstrated that DM was an independent, increased risk factor for CRC in both men and women, even after controlling for smoking, obesity, and physical exercise of the patients [65]. However, we found the RAN rs14035 CT + TT genotype was associated with a lower CRC risk in male patients and in all patients with rectal cancer. Although numerous studies have shown that tumor size is of no prognostic significance in CRC [67,68], our data showed that DROSHA rs10719 CC was associated with an increased CRC risk regarding tumor size (<5 cm) and age (62 years). We also found that the DICER1 rs3742330 and XPO5 rs11077 genotypes were not associated with the risk of age, gender, hypertension, DM, tumor site, tumor size, lymph node invasion, HTN and folate, suggesting that these SNPs might not modulate the susceptibility to CRC in the Korean population.
Several studies have reported the associations between homocysteine levels and cancer development. For example, cohort studies found elevated homocysteine was correlated with increased adenoma recurrence [69,70]. Similarly, we observed that DROSHA rs10719 and RAN rs14035 CC genotypes were associated with CRC with respect to the subjects' tumor sizes. We also found that the RAN rs14035 CC genotype was associated with decreased risk of CRC in patients with < 12.97 μmol/l homocysteine. However, to our knowledge, no other significant associations have been found between homocysteine and risk of CRC [71].
Folate provides 1-carbon units for DNA synthesis and methylation, and neoplasia results from the disruption in both of these processes. Modifications to DNA methylation occur on a genomic-and gene-specific level in colorectal neoplasia [72], and some studies have shown that low plasma folate concentrations were associated with reduced CRC risk and that plasma folate concentrations were positively related to CRC risk [73,74]. However, our data show that the RAN rs14035 CC genotype was associated with decreased risk of CRC in subjects with > 3.72 ng/ml folate.
We also examined the interplay between genetic factors and patient characteristics with respect to CRC occurrence. For example, we determined the DROSHA rs10719 CC genotype was associated with increased CRC risk in patients with HTN and DM, while the XPO5 rs11077 AC+CC genotype correlated to increased CRC risk in patients with HTN. Recent studies have observed increased risk of CRC associated with HTN [60,62], and the presence of HTN, obesity, and hyperglycemia [75]. However, a study of Finnish male smokers did not support these associations [76]. In addition, our data showed that the CC genotype of rs10719 located in the DROSHA gene had an increased CRC risk in patients with age (62 years) compared with those carrying the TT and CC genotypes. An accumulation of evidence indicates MetS, with its systemic and hormonal effects, may affect a patient's susceptibility to carcinoma and the prognosis of patients diagnosed with various cancers [77,78]. Our study provides new information regarding the association between CRC and MetS in the context of polymorphisms in miRNA machinery genes. However, additional studies of other miRNA machinery genes will be needed to clarify the association between these polymorphisms and CRC.
In summary, we investigated the relationship between CRC susceptibility and the miRNA machinery gene DROSHA rs10719, DICER1 rs3742330, RAN rs14035, and XPO5 rs11077 polymorphisms. To date, few studies have investigated the association between polymorphisms in RAN gene and CRC risk and there were no associations between the RAN gene polymorphisms and CRC risk [24][25][26]. Though our results provide the first evidence for a significant association between RAN polymorphisms and CRC in Korean patients, our findings will benefit from additional data regarding the polymorphism's effect upon mRNA stability, binding efficiency, and selectivity. Therefore, our study, while novel in its findings, requires validation by functional studies investigating the polymorphism's effect upon miRNA machinery and downstream cellular activities.
Supporting Information S1 File. Table A in S1 File. miRNA biogenesis gene genotype frequencies and CRC patient survival. Table B in S1 File. The combination of miRNA biogenesis genes polymorphisms based on MDR and CRC patients. Table C in S1 File. The combination of the polymorphisms of microRNA machinery genes in CRC patients and controls: male subgroup. Table D in S1 File. The combination of the polymorphisms of microRNA machinery genes in CRC patients and controls: female subgroup. Table E in S1 File. The combination of the polymorphisms of microRNA machinery genes between the subgroup for colorectal cancer and control subjects. Table F in S1 File. Genotype frequencies of miRNA biogenesis gene polymorphisms and CRC patients survival in TNM stage I + II. Table G in S1 File. Genotype frequencies of miRNA biogenesis genes polymorphisms and CRC patients survival in TNM stage III + IV. Table H in S1 File. Allele combinations of miRNA processing genes in CRC patients and controls using multifactor dimensionality reduction. Table I