Genetic Variants in MicroRNA Machinery Genes Are Associate with Idiopathic Recurrent Pregnancy Loss Risk

Objective Key molecules involved in microRNA (miRNA) biogenesis, such as DROSHA, XPO5, and DICER, have been identified in trophoblast cells, confirming that the miRNA biogenesis pathway is active in human placenta. In addition, miRNAs regulate uterine gene expression associated with inflammatory responses during the peri-implantation period and participate in maternal-fetal immune tolerance. The purpose of this study was to demonstrate whether genetic polymorphisms in miRNA machinery genes show an association with idiopathic recurrent pregnancy loss (RPL) in Korean women. Study design We performed a case-control study with 238 controls and 338 women who had experienced at least two consecutive pregnancy losses between 1999 and 2010. Genotypes of miRNA machinery genes, including DICER rs3742330, DROSHA rs10719, RAN GTPase (RAN) rs14035, and exportin-5 (XPO5) rs11077 were analyzed by polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) assay. The logistic odds ratios (ORs) of RPL were estimated with a 95% confidence interval (CI) in multivariate analysis after maternal age adjustment. Gene-gene interactions among the loci of the four gene polymorphisms were evaluated using the multifactor dimensionality reduction (MDR) method. Results The RAN rs14035 CC genotype and DICER rs3742330/DROSHA rs10719 GG/TC+CC, rs3742330/RAN rs14035 GG/CC, and DICER rs3742330/XPO5 rs11077 GG/AC+CC combinations were significantly associated with increased RPL risk, whereas the RAN rs14035 CT, DICER rs3742330/RAN rs14035 AA+AG/CT+TT, DROSHA rs10719/RAN rs14035 TC+CC/CT+TT, and RAN rs14035/XPO5 rs11077 CT+TT/AA combinations reduced RPL risk. The A-T-T-C and G-C-T-A allele combinations (DICER/DROSHA/RAN/XPO5) were 20 times more frequent in the RPL group than in the control group. Conclusion Our study demonstrates the relationship between RPL development and the polymorphism of the miRNA machinery gene RAN and combined genotype of DROSHA/DICER.


Introduction
The incidence of spontaneous pregnancy loss has been estimated to be 30% [1]. However, less than 5% of women will experience two or more consecutive pregnancy losses [2]. Therefore, the loss of two or more clinically recognized pregnancies is considered a distinct disorder and defined as recurrent pregnancy loss (RPL) for additional evaluation and treatment [3]. Despite the various etiologies that result in RPL, including cytogenetic abnormalities, antiphospholipid syndrome, uterine anomalies, hereditary thrombophilia, autoimmunity, sperm quality, and environmental factors, the cause of RPL still remains undetermined in the majority of cases.
MicroRNAs (miRNAs) are ,19-25-nucleotide single-strand non-coding RNA species that induce post-transcriptional gene silencing and mediate translational repression through binding to target mRNA, leading to subsequent mRNA degradation. The recent elucidation of miRNA function has provided new insight into the regulation of gene expression. Key molecules involved in miRNA biogenesis, such as DROSHA, XPO5, and DICER, have been identified in trophoblast cells, confirming that the miRNA biogenesis pathway is active in human placenta [4,5]. It has been demonstrated that the human placenta produces a large number of miRNAs and that those are involved in placental development [6]. In addition, miRNAs regulate uterine gene expression associated with inflammatory responses during the peri-implantation period and participate in maternal-fetal immune tolerance [7,8]. There are numerous reports that demonstrate the association of aberrant miRNA expression with various human diseases related to reproductive conditions [9][10][11][12][13]. Several studies have reported that genetic polymorphisms are associated with RPL development [14][15][16][17]. Furthermore, recent studies have found that nucleotide variations within pri-miRNA molecules affect miRNA processing and result in altered miRNA expression and that a single-nucleotide polymorphism (SNP) in a certain miRNA is associated with litter size in pigs [18,19].
Although SNPs have been widely implicated in RPL development, such evidence is lacking for miRNA biogenesis pathway genes. This study was performed to investigate whether polymorphisms in the miRNA machinery genes DROSHA (rs10719), DICER (rs3742330), RAN (rs14035), and XPO5 (rs11077) are associated with the prevalence of RPL in Korean women. Our previous study investigated the allele frequencies of nine SNPs in those four essential genes in primary ovarian insufficiency and control subjects [20]. The results revealed that some polymorphic sites, such as XPO5 rs3324334 and XPO5 rs11544382 that had been reported to be polymorphic markers, were monomorphic in all Korean subjects that we genotyped. Therefore, we examined the remaining four polymorphic sites in those four genes.

Study participants
The study was reviewed and approved by the institutional review board (IRB) of CHA Bundang Medical Center in 1999, and written informed consent was obtained from all participants. IRB approved this consent procedure. The study population consisted of Korean participants (Asian) recruited from the Department of Obstetrics and Gynecology of CHA Bundang Medical Center, CHA University (Seongnam-si, South Korea) between March 1999 and February 2010. Participants with a history of smoking, alcohol abuse, cancer, radiation exposure, autoimmune disorder, genetic syndromes, or systemic disease affecting ovarian function, such as diabetes mellitus, were excluded from the study based on medical history and physical examination. Pregnancy loss was diagnosed with hCG testing, ultrasound, and/ or physical examination before 20 weeks of gestational age. RPL was defined as two or more consecutive pregnancy losses before 20 weeks of gestational age. All participants in the RPL group experienced recurrent miscarriage with the same partner. RPL patients with previous live births were excluded from this study.
To diagnose idiopathic RPL, we adopted a set of exclusion criteria for the study group. Patients who were diagnosed with RPL due to anatomic, chromosomal, hormonal, infectious, autoimmune, or thrombotic causes were excluded from the study group. To identify anatomic abnormalities in RPL patients, sonography, hysterosalpingogram, hysteroscopy, computerized tomography, and magnetic resonance imaging were used. Karyotyping was performed with standard protocols [21]. Hormonal causes included hyperprolactinemia, luteal insufficiency, and thyroid disease and were evaluated by measuring levels of prolactin, thyroid-stimulating hormone, free T4, follicle-stimulating hormone, luteinizing hormone, and progesterone in peripheral blood. Lupus anticoagulant and anticardiolipin antibodies were examined for autoimmune causes such as lupus and antiphospholipid syndrome. Thrombotic causes were defined as thrombophilia and were evaluated by deficiencies of protein C and protein S and by the presence of anti-b2 glycoprotein antibody. The study group consisted of 338 women diagnosed with idiopathic RPL [age range, 23-43 years; mean age 6 standard deviation (SD), 32.8164.33 years] (Table S1).
The enrollment criteria for the control group included regular menstrual cycles, normal karyotype of 46XX, a history of at least one naturally conceived pregnancy, and no history of pregnancy loss. The control group was comprised of 238 women [age range, 22-45 years; mean age 6 SD, 33.3865.79 years]. The average gestational age and number of pregnancy losses in the RPL group were 7.3261.85 weeks and 3.0461.61 losses, respectively.

Genotyping
Peripheral blood samples were collected for genotyping. Genomic DNA was extracted from collected blood in the presence of anticoagulant using a G-DEX blood extraction kit (iNtRON Biotechnology, Seongnam-si, Korea). Nucleotide changes were determined by polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) analysis. The PCR primers for this study are shown in Table S2. Restriction enzyme digestion was carried out using the following enzymes and conditions: NlaIII (DROSHA rs10719) (New England BioLabs, Ipswich, MA), BanI (DICER rs3742330), and EcoRI (XPO5 rs11077) at 37uC for 16 hours and BslI (RAN rs14035) at 55uC for 16 hours. Genotypes determined by RFLP analysis were confirmed by two independent investigators. We confirmed genotypes again by sequencing 10% of the samples by random selection.

Statistical analysis
The differences in genotype and allele combination frequencies between idiopathic RPL subjects and controls were compared using the multivariate logistic regression and Fisher's exact test, respectively. Allele frequencies were calculated to identify deviations from Hardy-Weinberg equilibrium (HWE) using P = .05 as a threshold. Odds ratios (ORs), adjusted odds ratios (AORs), and 95% confidence intervals (CIs) were used to measure the strength of association between genotypes and idiopathic RPL. Two-tailed P values ,.05 were considered statistically significant. The ORs were adjusted by the age of the participants because maternal age is a known risk factor for spontaneous abortion and RPL [22].
Gene-gene interactions among the SNP loci were analyzed using the log-linear model-based multifactor dimensionality reduction (LM-MDR) and MDR software version 2.0 (available at www.epistasis.org) [23][24][25]. The allele combination frequencies for the selected models by MDR analysis were estimated with the HAPSTAT program version 3.0 (www.bios.unc.edu/,lin/ hapstat). The false-positive discovery rate (FDR) correction was used to adjust multiple comparison tests, and associations with FDR-adjusted P value ,.05 were considered to be significant data. Statistical analyses were performed with GraphPad Prism 4.0 (GraphPad Software Inc., San Diego, CA) and StatsDirect software version 2.4.4 (StatsDirect Ltd., Altrincham, UK). The statistical power was calculated using G*POWER 3.9.1 (Institut für Psychologie, Christian-Albrechts-Universitä t Kiel, Kiel, Germany).

Results
The genotypic distribution and allele frequencies of DROSHA rs10719, DICER1 rs3742330, RAN rs14035, and XPO5 rs11077 are shown in Table 1. All of the genes that were analyzed showed polymorphisms and occurred in HWE in both groups. The DICER rs3742330, DROSHA rs10719, and XPO5 rs11077 polymorphisms were not associated with the prevalence of RPL. However, the RAN rs14035 polymorphism was associated with the prevalence of RPL. The frequency of the CT genotype of RAN rs14035 was significantly higher in control subjects than in RPL patients. This association remained statistically significant after FDR correction. Thus, RAN rs14035 CT+TT was associated with the decreased prevalence of RPL (AOR = 0.657, 95% CI = 0.469-0.921).
The genotype frequencies were stratified according to the number of consecutive pregnancy losses (Table S3). For DROSHA rs10719 and XPO5 rs11077, there were no statistically significant differences between the two groups. The DICER rs3742330 recessive model (GG compared with AA+AG) was associated with subjects who had history of three or more and four or more pregnancy losses (AOR = 1.718, 95% CI = 1.032-2.861 and AOR = 2.170, 95% CI = 1.181-3.986, respectively), For RAN rs14035, the CT genotype was less frequently observed in the subgroups of RPL participants with two or more and four or more consecutive pregnancy losses (AOR = 0.624, 95% CI = 0.412-0.947 and AOR = 0.502, 95% CI = 0.287-0.879, respectively).
We performed combination analyses for miRNA machinery gene polymorphisms that we tested for the study (Table 2). These analyses revealed that the DICER rs3742330 GG/DROSHA rs10719 TC+CC genotype was significantly more frequent in RPL patients than in control subjects (AOR = 1.990, 95% CI = 1.048-3.778). The frequency of the DICER rs3742330 AA+ AG/RAN rs14035 CT+TT genotype was significantly higher in the control group (AOR = 0.688, 95% CI = 0.475-0.997). In addition, DROSHA rs10719 TC+CC/RAN rs14035 CT+TT and RAN rs14035 CT+TT/XPO5 rs11077 AA genotypes were more frequently observed in control participants (AOR = 0.621, 95% CI = 0.389-0.991, AOR = 0.631, 95% CI = 0.435-0.914, respectively). Combination analysis according to the number of pregnancy losses was also performed ( Table 3). Although these associations with RPL were not significant after FDR correction in combination analysis (Table 2), those associations became more evident when we performed combination analysis according to the number of pregnancy losses even after FDR correction (Table 3). In this combination analysis, there was a significant association between RPL development and the DICER/DROSHA GG/TC+ CC, DICER/RAN GG/CC, and DICER/XPO5 GG/AC+CC genotypes. The RAN CT+TT genotype showed a significant association with RPL when it was combined with the XPO5 AA genotype. These significances were maintained when we performed multiple comparison correction using FDR, for all except the RAN/XPO5 CT+TT/AA genotype. Thus, both individual and combined genotyping analysis suggested that both the DICER rs3742330 GG and RAN rs14035 CC genotypes are associated with RPL prevalence. To evaluate whether gene-gene interaction models exert synergistic effects on RPL risk, we conducted MDR-based allele combinations of four polymorphisms of miRNA machinery genes ( Table 4 and Table S4). We selected two interaction models  Table S5 represents the statistical power of the significant genotype. In addition, we conducted power analysis to detect the associations in the case-control study at the 5% significance level. The power varies for an OR by the proportion of exposure (i.e., the target genotype) in the control. A total sample size of 576 (338 RPL cases vs. 238 controls) could have reasonable power to detect an OR of 1.7 when the proportions of exposure in the control were 20% (power = 76.3%), 30% (power = 84.9%), and 40% (power = 87.7%). Additionally, this sample size has reasonable power to detect an OR of 0.55 for the proportions of 20% (power = 73.1%), 30% (power = 85.6%), and 40% (power = 91.3%) in the control. Furthermore, this sample size might have sufficient power to detect an OR of 1.7 or greater and an OR  of 0.55 or less. However, the power will be smaller when it is compared with multiple genotypes.

Discussion
We evaluated four SNPs in DICER, DROSHA, XPO5, and RAN, which are essential for miRNA biogenesis [26,27]. Our results comparing differences in genotype frequencies in miRNA machinery genes revealed that DICER rs3742330 GG genotype had significant associations with RPL in Korean women when it was combined with the DROSHA rs10719 TC+CC genotype. In addition, the ORs of RPL prevalence were higher because patients who possessed the DICER rs3742330 GG/DROSHA rs10719 TC+ CC genotype experienced more pregnancy loss. Defective DICER function due to genetic polymorphism might be intensified by DROSHA polymorphisms. These two molecules are considered major regulators of miRNA biogenesis. However, there is no evidence that these molecules interact with each other. It remains unclear how genetic polymorphisms in these genes influence each other. To elucidate the mechanisms by which different alleles of DICER or DROSHA affect RPL development, additional functional studies should be performed. Although the exact mechanism could not be elucidated in the present study, our finding might be used in the future as a marker to predict RPL development in women with a spontaneous abortion.
Dicer is essential in miRNA biogenesis, and its function has been reported to be involved reproduction. Dicer expression is detected in female reproductive organs, including the ovary, oviduct, and uterus [28][29][30]. Female mice with general hypomorphic mutation of Dicer were infertile because of luteal deficiency [30]. The conditional knockout of Dicer within ovarian granulosa cells affects ovulation rate by influencing the number of preovulatory follicles that achieve proper development [28,29]. On the other hand, Dicer is reported to be involved in regulating placental growth by influencing pro-mitogenic signaling molecules through miRNA action [31].
Although there is limited evidence that polymorphisms of DICER alter biologic functions depending on alleles, a number of studies have reported that DICER polymorphisms, including rs3742330, affect disease development and patient survival in various cancers [31][32][33]. In addition, the functional analysis of the DICER rs1057035 SNP, which resides in the 39UTR of DICER, has revealed that targeting of has-miR-574-3p to DICER was differentially affected by the type of DICER rs1057035 allele and that the DICER rs1057035 variant C allele led to reduced expression of Dicer [34]. In the DICER gene, the polymorphic site DICER rs3742330A.G, found in the 39 UTR of DICER, is considered to be important for mRNA transcript stability. In terms of human reproduction, DICER rs12323635, located in promoter region of DICER, has been reported to decrease the risk of Table 4. Allele combination analysis of miRNA machinery genes in RPL patients and control participants using the multifactor dimensionality reduction method.  oligozoospermia [35]. Therefore, Dicer is likely to be a crucial molecule for pregnancy. There are several potential mechanisms that explain why DICER polymorphisms are associated with miscarriage. DICER polymorphism may affect decidualization of the endometrium and induce implantation defects in embryos, resulting in recurrent miscarriage. Decidualization of endometrial stroma is one of the crucial steps for trophoblast invasion and placenta development [36]. A recent study using human embryonic stem cells demonstrated that DICER expression increased during in vitro decidualization. DICER inhibition affected the expression of decidual markers and transcription factors-such as PRL and HOXA10-which play a role during decidualization [37]. Second, DICER polymorphisms may have an influence on placental development by affecting miRNA expression involving cytotrophoblast proliferation. Cytotrophoblasts are trophoblastic stem cells that play multiple roles [6]. Recent research has demonstrated enhanced proliferation of cytotrophoblasts without exogenous growth factors following Dicer knock-down in first-trimester placenta. Although the miRNAs affecting cell proliferation were not identified, the authors suggested that Dicer-dependent miRNA affected the proliferation of cytotrophoblasts [5]. Dicer polymorphisms induce aberrant expression of miRNAs regulating cytotrophoblast proliferation, possibly resulting in abnormal placental development and abortion. Finally, the miRNA expression affected by the DICER polymorphism may induce chromosome mis-segregation, leading to aneuploidy, which is the most common cause of miscarriage. The centrosome is known to be a source of aneuploidy. The extra centrosome forms a multipolar spindle during mitosis in a mother cell, resulting in aneuploid daughter cells. This pathogenesis has been shown in cancer development [38]. miR-210 overexpression is reported to induce centrosome amplification in renal cancer [39]. Therefore, aberrant miRNA expression may cause chromosomal aneuploidy in cancer. A similar mechanism exists by which abnormal miRNA expression may cause aneuploid embryos and recurrent miscarriage during early development. A significant association between the RAN rs14035 genotype and RPL prevalence was observed when it was combined with the XPO5 rs11077 AA genotype. Ran, which is a member of the Ras superfamily of GTPases, is essential for the translocation of pre-miRNAs from the nucleus to the cytoplasm through the nuclear pore complex [40]. Xpo5 mediates nuclear export of pre-miRNA in a RAN GTP-dependent manner by binding to pre-miRNA and RAN GTPase in the nucleus [40]. Although several reports have shown the association of these molecules with disease development in various cancers, no report has elucidated the role of these molecules in human reproduction [41,42].

DICER/DROSHA
Our previous study revealed that the XPO5 rs2257082 allele was more frequently observed in premature ovarian insufficiency (POI) patients than in control subjects [20]. However, we could not identify any association between RAN polymorphism and POI development. Additionally, RAN rs14035, but not XPO5 rs11077, exerted an effect on RPL development. Regarding both molecules, few reports have evaluated the clinical significance of SNPs not only in cancer but also in pregnancy-related disease. In addition, no study has elucidated the functions of those SNPs in miRNA biogenesis. Therefore, additional functional studies are required to elucidate how RAN rs14035 affects disease promotion and prevention depending on its genotype.
We obtained contrasting results regarding RPL development in the combination analysis for the DROSHA rs10719 TC+CC genotype. When it was combined with DICER rs3724330 GG genotype, RPL development was more frequently observed in RPL group. When the DROSHA rs10719 TC+CC genotype was combined with the RAN rs14035 CT+TT genotype, RPL development was rare in the RPL group. One potential explanation for this observation is that idiopathic RPL has various unknown etiologies. Many reports in the literature have demonstrated that several causes are associated with recurrent miscarriage, such as angiogenesis, immunologic factors, and miRNAs [15,18,43]. This may explain why the same polymorphism affects disease development differently. The result of our combination analysis of polymorphisms according to the number of pregnancy losses supports this explanation ( Table 3). The AORs increased as the number of previous pregnancy losses increased in patients who have the DROSHA rs10719 TC+CC genotype combined with the DICER rs3742330 GG genotype (1.596 vs. 2.460 vs. 4.084). Conversely, no such tendency was noted with the DROSHA TC+ CC/RAN CT+TT genotype. Instead, the effect of polymorphisms was attenuated as the number of miscarriages increased. This suggests that genetic polymorphisms with respect to miRNA machinery genes were associated with disease development not through a common pathway but through various pathogeneses.
There are several potential limitations in this study. First, this case-control study only shows an association between SNPs in miRNA machinery genes and RPL development, not a causal relationship. Therefore, it is difficult to conclude that miRNA machinery genes affect recurrent miscarriage. Second, the manner in which polymorphisms in miRNA biogenesis pathway genes affect RPL development is still unclear, and a functional study of those SNPs to elucidate the pathogenesis related to RPL was not conducted. Third, the lack of information regarding placental pathology, immunologic profiles, and miRNA expression could have contributed to the investigation of potential roles of miRNA machinery genes during the peri-implantation and early pregnancy periods. Lastly, the sample size of the control group is relatively small, compared with that of the study group. This is because we included all eligible cases during the study period to avoid selection bias. This difference in sample size will increase type 2 errors in the genetic association study. To reduce the bias, we compared the RAN rs14035 and XPO5 rs11077 genotype frequencies in the control group with those from other studies that were performed in Korea. The genotype frequencies of RAN rs14035 and XPO5 rs11077 in the control group of our study showed similar patterns to the results in the control groups of previous reports regarding lung cancer and breast cancer in Korea [26,27].
Recent groundbreaking discoveries that small, non-coding RNAs can regulate gene expression transcriptionally or posttranscriptionally and that these RNAs are involved in placental development through trophoblast differentiation, proliferation, and angiogenesis provide us with insight into the pathogenesis of idiopathic RPL [4,7,44]. To the best of our knowledge, this is the first report to investigate the association between polymorphisms in miRNA biogenesis machinery genes and the prevalence of RPL in a Korean population and suggests that genetic polymorphisms in miRNA biogenesis machinery genes show a relationship with the risk of RPL. The present study may contribute to assessments of individual risk of RPL. However, further epidemiologic studies with larger subject numbers should be performed to confirm and expand our results. In addition, the results of our study warrant further functional studies to elucidate the mechanisms by which polymorphisms of miRNA machinery genes affect RPL development.