Genetic Variations miR-10aA>T, miR-30cA>G, miR-181aT>C, and miR-499bA>G and the Risk of Recurrent Pregnancy Loss in Korean Women

Blood samples were collected from 381 RPL patients (mean age ± standard deviation [SD], 33.00 ± 5.73 years) and 281 control participants (33.03 ± 4.36 years). Blood samples were collected prior to 20 weeks of gestation based on human chorionic gonadotropin (hCG) levels. The RPL patients were recruited from the Department of Obstetrics and Gynecology or the Fertility Center at the CHA Bundang Medical Center in Seongnam, South Korea between March 1999 and February 2010. Women in the control group were recruited from CHA Bundang Hospital and met the following criteria: history of at least one spontaneous pregnancy; current pregnancy; regular menstrual cycles; karyotype 46, XX; and no history of miscarriage. The study abided by the Declaration of Helsinki and was approved by the Institutional Review Board of CHA Bundang Medical Center (IRB approval no. BD2010-123D), and written informed consent was obtained from all participants. All RPL patients had suffered a minimum of two consecutive spontaneous miscarriages at an average gestational stage of 7.36 ± 1.93 weeks. Pregnancy loss was diagnosed based on the results of hCG tests, ultrasound, and/or physical examination before 20 weeks of gestation. None of the participants had a history of smoking or alcohol use. The following parameters were also measured: activated partial thromboplastin time (aPTT), body mass index (BMI), blood urea nitrogen (BUN), creatinine, estradiol (E2), follicle-stimulating hormone (FSH), luteinizing hormone (LH), platelet (PLT) count, and prothrombin time (PT), using participant blood samples. Patients with the following conditions were excluded from the study: RPL or implantation failure due to hormonal, genetic, anatomic, infectious, autoimmune, or thrombotic causes. Anatomic causes were evaluated using hysterosalpingogram, hysteroscopy, computed tomography, and magnetic resonance imaging to detect intrauterine adhesions, septate uterus, and uterine fibroids. Hormonal causes, including hyperprolactinemia, luteal insufficiency, and thyroid disease, were evaluated by blood analyses. Infectious causes, such as the presence of Ureaplasma urealyticum or Mycoplasma hominis, were evaluated by bacterial culture. Autoimmune causes, including antiphospholipid syndrome or lupus, were evaluated using lupus anticoagulant and anticardiolipin antibodies. Thrombotic causes, such as thrombophilia, were evaluated by identification of protein C and S deficiencies and by detection of β-2-glycoprotein 1 antibodies. A total of 150 μL of whole blood and fluorochrome-labeled monoclonal antibodies against anti-CD3-FITC (1:100, 555339), anti-CD4-PE(1:100, 357404) anti-CD8-PE-cy5 (1:20, 344769) anti-CD19-APC (1:100, 392503), anti-CD56-PE-Cy7 (1;100, 392411) NK cells were added to each tube. All antibodies were obtained from Biolegend (San Diego, CA, USA). The tubes were vortexed and incubated in the dark at room temperature for 40 min. Next, 2 mL of Lyse solution (diluted 1:10; BD Bioscience, Sunnyvale, CA, USA) was added, and the tubes were vortexed again, incubated at room temperature for 30 min, and centrifuged at 1200 rpm for 5 min. The cells were then washed three times with 2 mL of PBS each wash, and the cells were suspended in 250 μL of PBS and analyzed by flow cytometry (BD Bioscience). Chromosome analysis was conducted according to standard cytogenetic methods. Peripheral blood lymphocytes were cultured for 70 h, and then KaryoMAX Colcemid Solution (Gibco) was added when the chromosomes were at the metaphase stage. KCl (0.05 M) was added as a hypotonic agent, and the cells were fixed for harvest using a fixative formed by adding one volume of acetic acid to two volumes of methanol. Metaphase chromosome preparations obtained after cell culture were stained using the Giemsa-Trypsin-Giemsa (GTG) banding method. Genomic DNA was extracted from anticoagulant-treated peripheral blood samples using a G-DEX Genomic DNA extraction kit (iNtRON Biotechnology, Seongnam, Korea) [ 28 , 29 ]. Briefly, Proteinase K was added to a microcentrifuge tube, followed by 30 µL of blood. Next, 300 µL of Lysis solution was added, and the samples were vortexed and incubated at 55 °C for 10 min. A total of 350 µL of ethanol was then added to each sample, and the samples were bound, washed, and eluted according to the manufacturer’s protocol. Four miRNAs (SNPs) were selected using the NCBI human genome SNP database (dbSNP, http://www.ncbi.nlm.nih.gov/snp (accessed on 13 March 2019)). The SNPs miR-10a A>T (rs3809783), miR-30c A>G (rs113749278), miR-181a T>C (rs16927589), and miR-499b A>G (rs37464444) are either mature-form (rs3746444, rs-formnp8978) or pri-form (rs3809783, rs16927589). miR-10a A>T, miR-30c A>G, miR-181a T>C, and miR-499b A>G were genotyped according to TaqMan ® SNP Genotyping Assays system (Applied Biosystems, Foster City, CA, USA). Based on the intensity of fluorescence signals of FAM and VIC, samples were automatically classified into one of three groups corresponding to the genotypes AA, AG, or TT of miR-10a A>T; AA, AG, or GG of miR-30c A>G; TT, TC, or CC of miR-181a T>C; and AA, AG, or GG of miR-499b A>G. The basic principle of the assay is as follows: when the allele-specific probe is fully hybridized to the template DNA, Taq polymerase cleaves the reporter dye, leading to fluorescence emission. However, if a single base mismatch exists between the probe and template DNA, hybridization is inefficient, and reporter dye fluorescence is thus reduced. The sequences of the SNPs were as follows: miR-10a A>T, CTCTT ATTTTTCCAG AAGAAAAAAA[A/T]ATATATATAT GTATATGTAG TATTT; miR-30c A>G, TACTTTCCACAGCTG AGAGTGTAGG[A/G]DTGTTTACAGT ATCTGTCGCT CAGTG; miR-181a T>C, AAAAT AGCACAAAAT TATCCAATTG[T/C] GACAGTTCTT ATCACATTTC ACTTT; and miR-499b A>G, ATGTTTAACT CCTCTCCACG TGAAC[A/G]TCACAGCAAG TCTGTGCTGC TTCCC. Information regarding the miRNA probes was as follows: miR-10a A>T, wild type homozygous AA (VIC reaction & FAM no reaction), heterozygous AT (VIC reaction & FAM reaction), mutant homozygous TT (VIC no reaction & FAM reaction); miR-30c A>G, wild homozygous AA (VIC reaction & FAM no reaction), heterozygous AG (VIC reaction & FAM reaction), mutant homozygous GG (VIC no reaction & FAM reaction); miR-181a T>C, wild homozygous TT (VIC reaction & FAM no reaction), heterozygous TC (VIC reaction & FAM reaction), mutant homozygous CC (VIC no reaction & FAM reaction); miR-499b A>G, wild homozygous AA (VIC reaction & FAM no reaction), heterozygous AG (VIC reaction & FAM reaction), mutant homozygous GG (VIC no reaction & FAM reaction). Plasma PAI-1, total cholesterol, uric acid, and homocysteine levels were measured in participant blood samples. Plasma was separated by centrifugation of whole blood at 1000× g for 15 min. PAI-1 levels were determined using a human serpin E1/PAI-1 immunoassay (R&D Systems, Minneapolis, MN, USA). Uric acid and total cholesterol levels were measured using enzymatic colorimetric tests (Roche Diagnostics, GmbH, Mannheim, Germany). Homocysteine levels were measured using a fluorescence polarization immunoassay with an Abbott IMx analyzer (Abbott Laboratories, Abbott Park, IL, USA). The significance of differences in the frequencies of the miR-10a A>T, miR-30c A>G, miR-181a T>C, and miR-499b A>G SNPs between the control and patient groups were assessed using Fisher’s exact test and a logistic regression model. p -values were calculated using two-sided t -tests for continuous variables and chi-square tests for categorical variables. Allele frequencies were calculated to investigate the deviation from Hardy–Weinberg equilibrium. The genotype distribution of RPL patients and controls with ≥h or ≥o pregnancy loss was investigated. Odds ratios (ORs), adjusted odds ratios (AORs), and 95% confidence intervals (CIs) were used to examine the associations between various miRNA polymorphisms and RPL risk. Data are presented as the mean ± SD for continuous variables or a percentage for categorical variables. The results of the allele and genotype combination analysis were consistent with those derived from Fisher’s exact test during regression analysis. Statistical analyses were carried out using MedCalc software, version 12.1.4 (MedCalc Software bvba, Mariakerke, Belgium) or GraphPad Prism 4.0 software (GraphPad Software, Inc., San Diego, CA, USA). Logistic regression analysis was applied to data regarding baseline characteristics, genotype frequencies, genotype combinations, and allele combinations for quantitative traits shown in Table 2, Table 3, Table 4 and Table 5. The HAPSTAT program (v.3.0, www.bios.unc.edu/~lin/hapstat/ (accessed on 10 April 2018)), which exhibits a strong synergistic effect, was used to estimate the frequencies of polymorphic haplotypes. A p -value < 0.05 indicated statistical significance. HAPSTAT allows testing of haplotype (or allele combination) effects by maximizing the likelihood (from the observed data) that properly accounts for phase uncertainty and study design. False-positive discovery rate (FDR) correction was used to adjust multiple comparison tests and associations with FDR-adjusted p -values < 0.05 were considered statistically significant [ 30 ]. FDR calculation is also used for multiple hypotheses testing to correct for multiple comparisons. Multifactor dimensionality reduction (MDR) analysis was used to determine the best-model gene-gene interaction for RPL risk. The advantage of using MDR is that it overcomes the sample size limitations often encountered during logistic regression analysis in studies of high-level interactions. The MDR method consists of two main steps. First, the best combination of multi-factors is selected, and second, genotype combinations are classified into high- and low-risk groups [ 31 ]. We constructed all possible allelic combinations by MDR analysis to identify combinations with strong synergy. Allelic combinations for multiple loci were estimated using the expectation-maximization algorithm with SNPAlyze (v. 5.1; DYNACOM Co, Ltd., Yokohama, Japan), and those having frequencies < 1% were excluded from statistical analysis. We also applied multiple regression models to further explain the results of the allelic combination analysis. Genetic interaction analyses were performed using the open-source MDR software package (v.2.0), which is available at www.epistasis.org (accessed on 15 March 2018). The pre-miRs (miR-10a , miR-30c , and miR-181a ) and their flanking regions were amplified from human genomic DNA and cloned into the vector pcDNA3.1(−) (Invitrogen, Carlsbad, CA, USA). The primers used in the study included F: 5′-TGC GAA CTG GCT ACT TGA AA-3′, R: 5′-TTC CAA TAA AGC CTC CCT GA-3′ ( miR-10a ); F: 5′-GCA CCA TGT GTC ACA CAG GT-3′, R: 5′-CAA GTG TTG GGA AGA TGC TAT-3′ ( miR-30c ); and F: 5′-ACA TTT TCT CAG ACA TTC AT-3′, R: 5′-ATG TGA GAA AAC TGA GAC AC -3′ ( miR-181a ). For single-point mutations, we used an Intron Muta-direct kit (Intron, Seoul, Korea). The sequences of these three vectors were confirmed by direct sequencing, and the SNPs were the only differences detected. To generate the miRNAs target gene::luciferase reporter constructs, similar to the cloning vectors, fragments of the PAI-1 gene corresponding to the 3′-UTR region clone (OriGene, Rockville, MD, USA) were amplified and cloned into the pGL4.13-luciferase vector (Promega, Madison, WI, USA). The resulting cDNAs were PCR amplified using the following primers: forward 5′-CCC TGG GGA AAG ACG CCT T-3′ and reverse 5′-TTC GTA TTT ATT TAT TTT ATT TTT T-3′ with Xba I (TCTAGA)and Fse I (GGCCGGCC) linker (New England Biolabs, Ipswich, MA, USA), and all constructs were verified by sequencing. Cells from a human endometrial cell line (Ishikawa) were plated at 1 × 10 6 cells per well in 6-well plates and transfected 24 h later using JetPRIME transfection reagent (Polyplus, France). Transfection reactions for miR-10a contained 500 ng of miR10a-A (in pcDNA3.1-) or 500 ng of miR-10a-T (in pcDNA3.1-) with 500 ng of 3′-UTR-PAI-1 in pGL4.13 and 200 ng of pGL4.75 (Renilla-normalization control); for miR-30c, reactions contained 500 ng of miR-30c-A (in pcDNA3.1-) or 500 ng of miR-30c-G (in pcDNA3.1-) with 500 ng of 3′-UTR-PAI-1 in pGL4.13 and 200 ng of pGL4.75 (Renilla-normalization control), for miR-181a2, reactions contained 500 ng of miR-181a-T (in pcDNA3.1-) or 500 ng of miR-181a-G (in pcDNA3.1-) with 500 ng of 3′-UTR-PAI-1 in pGL4.13 and 200 ng of pGL4.75 (Renilla-normalization control). TRIzol reagent (Invitrogen, Waltham, MA, USA) was used to isolate total RNA from Ishikawa cells that were transfected with 2.5 μg of vector after 16 h. Total RNA was then reverse transcribed using an M-MLV reverse transcriptase PCR kit (Biofact, Co., Ltd., Daejeon, Korea) and random or oligo dT20 primers (Invitrogen, Waltham, Massachusetts, USA) in addition to specific primers for PAI-1 and glyceraldehyde 3-phosphate dehydrogenase ( GAPDH) . Quantitative real-time PCR (qPCR) was performed as 20 μL reactions, containing each sequence-specific primer and quantitative PCR master mix (Solgent, Co., Ltd., Daejeon, Korea), using a Rotor-Gene 6000 real-time PCR system (Qiagen, Co., Ltd., Hilden, Germany). Expression levels were calculated according to the comparative threshold cycle (Ct) method using the formula 2 −ΔΔCt . Primer sequences for amplification were as follows: has-miR-10a-pre forward: 5′-CCG AAT TTG TGT AAG GAA TTT TG-3′ and reverse 5′-AAG AGC GGA GTG TTT ATG TCA A-3′; has-miR-10a-mature forward: 5′-TAC CCT GTAG ATC CGA ATT T and reverse: universal primer (Qiagen Cat# 218193); has-miR-30c-pre forward: 5′-TGT GTA AAC ATC CTA CAC TCT CAG C-3′ and reverse: 5′-CCA TGG CAG AAG GAG TAA ACA-3′; has-miR-30c-mature forward: 5′-AAA CAT CCT ACA CTC TCA GC-3′ and reverse universal primer (Qiagen Cat# 218193); has- miR-181a -pre forward:5′-TAT CAG GCC AGC CTT CAG AG-3′ and reverse: 5′-AAT CCC AAA CTC ACC GAC AG-3′; miR-181a -mature forward:5′- TTC AAC GCT GTC GGT GAG TT-3′ and reverse: universal primer (Qiagen Cat# 218193); Human RNU6B (RNU6-2) forward:5′-ACG CAA ATT CGT GAA GCG TT-3′ and reverse universal primer (Qiagen Cat# 218193). An online search was conducted to identify targets for miR-10a , miR-30c , miR-181a , and miR-499b using the TargetScan ( http://www.targetscan.org (accessed on 21 May 2018)) and miRIAD databases ( http://bmi.ana.med.uni-muenchen.de/miriad/ (accessed on 16 May 2018)). We used these databases to predict miRNAs that target overlapping regions of PAI-1 mRNA transcripts. Target mRNA sequences, particularly within the 3′-UTR, are often obtained from the National Center for Biotechnology Information ( www.ncbi.nlm.nih.gov/ (accessed on 26 September 2018)). We found that miR-30c, miR-10a , and miR-181a were predicted to be targets of the PAI-1 3′-UTR. Therefore, a luciferase reporter assay was used to evaluate the roles of miR-30c, miR-10a, and miR-181a in regulating the expression of target genes, as previously described. Briefly, wild-type pGL4.13-luciferase vector (Promega, Madison, WI, USA). constructs containing the 3′-UTRs of the PAI-1 gene were generated by amplifying the 3′-UTR region clone (OriGene, Rockville, MD, USA) and cloning the amplification products into the downstream region of the pGL4.13 vector (Promega, Madison, WI, USA) using the Xba I and Fse I endonucleases (New England BioLabs, Ipswich, MA, USA). Positive clones were selected by sequence-specific PCR, restriction enzyme digestion, and DNA sequencing. Ishikawa cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Thermo Fisher Scientific, Inc. Waltham, Massachusetts, USA). All medium was supplemented with 10% fetal bovine serum (FBS) (Thermo Fisher Scientific, Inc. Waltham, Massachusetts, USA) and 1% penicillin/streptomycin (Thermo Fisher Scientific, Inc. Waltham, Massachusetts, USA). All cell lines were maintained in a CO 2 incubator (5% CO 2 ) at 37 °C. The Ishikawa cells used in this study were endometrial and are commonly used in RPL studies. Next, miR-10a, miR-30c, and miR-181a mimics (50 nM) were co-transfected into Ishikawa cells with 200 ng of the 3′-UTR of PAI-1 in pGL4.13 constructs using lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). After 16 h of incubation, the luciferase activity was measured using a dual-luciferase reporter assay system (Promega, Madison, WI, USA). Each transfection was performed as triplicates.

Recurrent pregnancy loss (RPL) is generally defined as three or more consecutive losses of pregnancy before 20 weeks of gestation. However, the American Society for Reproductive Medicine recently redefined RPL as more than two consecutive pregnancy losses [ 1 ]. Worldwide, RPL is a serious health problem that is significantly associated with morbidity and mortality. Factors contributing to the etiology of RPL include advanced maternal age, maternal anatomic anomalies, placental anomalies, chromosomal abnormalities, endocrine dysfunction, antiphospholipid syndrome, hereditary thrombophilia, psychological trauma, and environmental factors, such as smoking, excessive alcohol consumption, and stress [ 2 ]. Additionally, women who miscarry during their first pregnancy are 5% more likely to develop RPL than healthy women [ 3 ]. Although many relevant factors have been identified, the root cause of most cases of RPL remains unknown. RPL is also associated with blood clotting angiogenesis and immune disorders. MicroRNAs (miRNAs) are small (approximately 23 nucleotides), noncoding, single-stranded RNA molecules that form base pairs with complementary target messenger RNAs (mRNAs) [ 4 ]. It has been demonstrated that miRNAs modulate gene expression via destabilization or translational repression of target mRNAs [ 5 , 6 ]. Furthermore, miRNAs have been implicated in the regulation of several biochemical pathways in various eukaryotic organisms [ 7 , 8 ]. RNA polymerase II transcribes miRNAs into long precursor transcripts known as primary (pri)-miRNAs, which are subsequently converted into pre-miRNAs by DROSHA, which is a ribonuclease type III enzyme that forms a functional complex with DiGeorge syndrome critical region 8 [ 9 , 10 ]. The pre-miRNA is then exported to the cytoplasm by the exportin5 (XPO5)-RAS–related nuclear protein (RAN)-guanosine-5′-triphosphate (GTP) complex [ 11 ]. RAN is a small GTP-binding protein, and the RAN GTPase-XPO5 complex forms a heterotrimer with the pre-miRNA [ 12 ]. The pre-miRNA is processed by RNase III DICER to release the miRNA duplex, which is a double-stranded RNA approximately 23 nucleotides in length. DICER also initiates the formation of the RNA-induced silencing complex (RISC) [ 13 ], which is responsible for miRNA-mediated gene silencing and RNA interference. The biological function of the miRNA is initiated by binding to the 3′-untranslated region (UTR) of the target mRNA, thereby repressing its expression. A single miRNA can regulate the expression of multiple target mRNAs, thus serving as a master controller of gene expression. Multiple studies have recently demonstrated the roles of miRNAs in the pathophysiology of several ovarian diseases, including polycystic ovary syndrome (PCOS) and primary ovarian insufficiency (POI) [ 14 , 15 ]. POI, which is also known as premature ovarian failure, is characterized by insufficient or premature depletion of ovarian reserves, which leads to infertility [ 16 ]. The findings of the present study suggest that miRNAs play an essential role in the normal function and regulation of reproductive organs. The expression of a given gene may be affected or regulated by its genetic variations, and single-nucleotide polymorphisms (SNPs) are the most common genetic variation affecting DNA [ 17 ]. SNPs or mutations in genes encoding miRNAs can affect miRNA properties, resulting in their altered expression and/or maturation [ 18 ]. Sequence variations around the processing sites of miRNAs or in the mature miRNA itself, particularly in the seed sequence, can profoundly affect miRNA biogenesis and function [ 19 ]. Polymorphisms in pre-miRNAs were first reported in 2005 [ 20 ], and several studies on the associations of these polymorphisms have since been reported [ 21 , 22 ]. Aberrant miRNA expression has been implicated in numerous diseases; therefore, considerable research efforts are currently being made for miRNA-based therapies [ 23 ]. In the present study, we performed a database search and identified four SNPs in pre-form miRNAs: miR-10a A>T (rs3809783), miR-30c A>G (rs113749278), miR-181a T>C (rs16927589), and miR-499b A>G (rs3746444). All of these miRNAs are reportedly associated with various reproductive diseases [ 24 , 25 , 26 , 27 ]. Therefore, we hypothesized that the SNPs miR-10a , miR-30c , miR-181a , and miR-499b play a role in the development of RPL. The minor allele frequency of these SNPs is >5% in the Asian population; however, whether they are genetically associated with RPL or whether miRNA expression varies as a function of these pre-form polymorphisms remains unclear. We, therefore, investigated the correlation between RPL and these miRNA polymorphisms.

The characteristics of RPL patients and control subjects are summarized in Table 1 . The mean age was approximately 33 years for both groups, and both groups were 100% female. PLT count, aPTT, and concentrations of E2 and LH were greater in RPL patients than in controls ( p = 0.0007, p = 0.005, p = 0.001, and p = 0.011, respectively). There were no significant differences in age, BMI, uric acid level, or FSH level between the two groups. Table 2 shows the distribution of genotypes in RPL patients with ≥3 or ≥4 pregnancy losses and control subjects. Significant differences in the miR-10a SNP were observed between the RPL and control groups and were significantly correlated with RPL prevalence. Consistently, the absence of these miRNA polymorphisms showed a negative correlation with RPL. The associations of these polymorphisms were very interesting in RPL patients because the miRNA polymorphisms were related to decreased RPL, but they were not associated with RPL risk ( Table 2 ). In addition, the number of RPL patients with risk factors was very small. Therefore, the associations with RPL occurrence will require further investigation. miR-10a A>T (chr17:48579816, rs3809783), miR-30c A>G (chr6:71377017, rs113749278), miR-181a T>C (chr9:124692981, rs16927589), and miR-499b A>G (chr20: 34990400, rs37464444) were all in the miRNA mature-form (rs3746444, rs-formnp8978) or pri-form (rs3809783, rs16927589). The SNPs in miRNA genes, including pri-miRNAs, pre-miRNAs, and mature miRNAs, could potentially influence the processing and/or target selection of miRNAs. Since we selected four SNPs in pri-form or mature-form, we wanted to determine whether all these miRNAs could influence the expression and regulation of target genes. Based on the intensity of FAM and VIC fluorescence, samples were automatically classified into one of three groups corresponding to genotypes AA, AT, or TT of miR-10a A>T; AA, AT, or GG of miR-30c A>G; TT, TC, or CC of miR-181a T>C; and AA, AG, or GG of miR-499b A>G. The miR-30c AG+GG genotype was associated with decreased risk of RPL for age < 33 years (odds ratio [OR] = 0.583; 95% confidence interval [CI] = 0.371–0.918; p = 0.022) ( Table 3 ). However, the miR-181a TC+CC genotype was associated with increased risk of RPL for age < 33 years (OR = 1.677; 95% CI = 1.038–2.709; p = 0.035), and the miR-499b AG + GG genotype was associated with increased risk of RPL for age ≥ 33 years (OR = 1.631; 95% CI = 1.028–2.588; p = 0.038). The miR-10a AT+TT genotype was associated with increased risk of RPL for BMI ≥ 25 kg/m 2 (OR = 2.840; 95% CI = 1.544–5.223; p = 0.001). The miR-499b AG + GG genotype was associated with increased risk of RPL for BMI <25 kg/m 2 (OR = 1.456; 95% CI = 1.029–2.059; p = 0.034) and with increased risk of RPL for BMI ≥25 kg/m 2 (OR = 2.284; 95% CI = 1.377–3.789; p = 0.001). The miR-181a TC + CC genotype was associated with increased risk of RPL for PLT count <255.62×10 3 /μL) (OR = 1.779; 95% CI = 1.038–3.048; p = 0.036). Finally, the miR-30c AG+GG was associated with decreased risk of RPL for aPTT < 32.83 s (OR = 0.364; 95% CI = 0.185–0.717; p = 0.004). The results of combined gene-genotype analyses are shown in Table 4 . The miR-10a/miR-30c combined genotype AT/AG was associated with increased RPL risk (OR = 2.156; 95% CI = 1.120–4.151; p = 0.022). The miR-10a A>T/ miR-181a T>C combined genotype AT/TT was associated with increased RPL risk (OR = 1.974; 95% CI = 1.065–3.658; p = 0.031). The miR-10aA>T/miR-499 A>G combined genotype AT/AG was associated with increased RPL risk (OR = 2.195; 95% CI = 1.156–4.169; p = 0.016). The miR-30c A>G /miR-181a T>C combined genotype AG/TT was also associated with increased RPL risk (OR = 1.839; 95% CI = 1.054–3.210; p = 0.032). Similarly, increased RPL risk was associated with the miR-30c A>G /miR-499 A>G combined genotypes AA/GG (OR = 4.324; 95% CI = 1.423–13.141; p = 0.010) and AG/AG (OR = 1.921; 95% CI = 1.145–3.224; p = 0.013). The miR-181a T>C/ miR-499 A>G combined genotype TT/GG was also associated with increased RPL risk (OR = 8.320; 95% CI = 1.043–66.384; p = 0.046). However, after false-discovery rate (FDR)- p correction, there were no significant differences between RPL patients and controls in the ORs for the combined genotypes, except for the miR-30c A>G/ miR-499 A>G combined genotypes AA/GG and AG/AG. The results of allele combination analyses of miRNA polymorphisms in RPL patients and control subjects are shown in Table 5 and Supplementary Tables S2–S4 . The allele combinations miR-10a/miR-30c/miR-181a/miR-499b A-T-G-G (OR = 1.952; 95% CI = 1.120–3.149; p = 0.006), A-C-A-G (OR = 2.343; 95% CI = 1.111–4.942; p = 0.026), A-C-G-A (OR = 2.136; 95% CI = 1.095–4.165; p = 0.028), T-T-G-A (OR = 0.455; 95% CI = 0.215–0.962; p = 0.044), and T-C-G-A (OR = 13.020; 95% CI = 0.739–229.300; p = 0.017) were associated with an increased risk of RPL. However, after FDR- p correction, there were no significant differences between RPL patients and controls in the ORs of the allele combinations, except for the A-T-G-G and T-C-G-A allele combinations. The impact of SNPs on the interaction of miR-10a A>T, miR-30c A>G, miR-181a T>C, and miR-499b A>G on their targets was investigated by constructing various expression plasmids ( pri-miR-10aA , pri-miR-10aG , pri-miR-30cA , pri-miR-30cG , pre-miR-181a T, pre-miR-181a G, pri-miR-499b A, and pri-miR-499b G) under control of the cytomegalovirus (CMV) promoter with either the major or minor allele. These plasmids were used in a dual luciferase assay performed with the 3′UTR of PAI-1, one of the predicted targets of miR-10a, miR-30c and miR181a, in Ishikawa human endometrial cells. A schematic diagram of a gene with a 3′-UTR of PAI-1 containing possible miR-10a and miR-30c binding sites in a conserved region is shown in Figure 1 A,B. The luciferase activity of the 3′UTR of PAI-1 was significantly lower in pre-miR-10a having the A allele as compare to pre-miR-10a having the T allele ( p < 0.05) ( Figure 1 C). Similarly, the luciferase activity of the 3′UTR of PAI-1 was significantly lower in the pre-miR-30c with the A allele as compared to pre-miR-30c with the G allele ( p < 0.05) ( Figure 1 D). Associations between miRNA polymorphisms and the levels of homocysteine, folate, total cholesterol, uric acid, blood urea nitrogen (BUN), estradiol (E2), thyroid-stimulating hormone (TSH), FSH, LH, prolactin, creatinine, platelets (PLT), as well as CD3 + , CD4 + , CD8 + , CD19 + , and CD56 + NK cells, in addition to the PT and aPTT were assessed by ordinal logistic regression analyses. We divided the risk factors into 10 grades and performed ordinal logistic regression using a proportional odds model. We found that the genotype frequency of miR-30c A>G was significantly associated with aPTT (AA: 32.46 ± 4.71, GG: 27.56 ± 3.59, p = 0.001), creatinine (AA: 1.19 ± 1.94, GG: 6.26 ± 3.71, p = 0.001), and E2 (AA: 1.19 ± 1.94, GG: 6.26 ± 3.71, p = 0.001). Levels of FSH differed significantly ( p G AA (mean ± SD, 32.36 ± 4.30 and 6.96 ± 4.29, respectively) and GG genotypes (30.49 ± 3.02 and 33.82 ± 55.85, respectively) ( Table 6 , Figure 2 A,C,D). Additionally, levels of hematocrit (Hct) and total cholesterol (T. chol) differed significantly ( p C genotype frequency was significantly associated with levels of creatinine (TT: 2.38±3.24, TC: 1.17±1.76, p = 0.011), Hcy (TT: 6.76 ± 2.01, CC: 9.98 ± 4.50, p = 0.001), LH (TT: 4.81 ± 2.74, CC: 4.20 ± 0.71, p = 0.038), PT (TT: 11.43 ± 1.14, CC: 10.20 ± 0.28, p = 0.048), and T. chol (TT: 136.96 ± 86.18, TC: 185.65 ± 76.23, p = 0.001). The miR- 499b A>G genotype frequency was significantly associated with aPTT (TT: 31.20 ± 4.29, GG: 32.10 ± 4.18, p = 0.026) ( Table 6 , Figure 2 B).

Increasing evidence suggests that miRNAs play critical roles in the pathophysiology of various reproductive disorders [ 14 , 15 , 32 ]. Here, we investigated whether four pre-miRNA SNPs ( miR - 10a, miR-30c, miR-181a , and miR-499b ) were associated with the risk of RPL in a cohort of Korean women. Specifically, we focused on the genotypes and allele combination of the selected miRNA polymorphisms and aimed to determine how they affected the risk of RPL. Using a genotype-based analysis method, we found that the GG and dominant (AA vs. AG + GG) miR-499b genotypes were significantly more common in RPL patients (PL ≥ 3 and PL ≥ 4, p G/ miR-499 A>G were significantly more common in RPL patients than in controls. As the activities of many genes are interconnected in complex conditions such as RPL, gene-gene interactions may affect gene-disease associations. The MDR method enables the detection of gene-gene interactions, regardless of the chromosomal locations of the genes [ 33 ]. We used a novel genotype-based MDR approach to examine the effects of potential interactions between different miRNAs on RPL risk. These results of these analyses, which examined the effects of four miRNA polymorphisms associated with RPL, suggested that gene-gene interactions involving these four miRNA polymorphisms also play roles in determining the risk of RPL. Allele combination MDR analyses indicated that the two combination conferred by the miR-10a A>T/ miR-181a T>C/ miR-30c A>G/ miR-499 A>G (A-T-G-G and T-C-G-A), the two combination conferred by the miR-10a A>T/ miR-181a T>C/ miR-30c A>G (T-T-A, T-C-G), the two combination conferred by the miR-10a A>T/ miR-30c A>G/ miR-499 A>G allele combination (C-A-G, C-G-A), and the genotype conferred by the miR-10a A>T/ miR-30c A>G allele combination (T-A) occur more frequently in patients with RPL than control subjects, suggesting a significant association with increased risk of RPL (all p T/ miR-181a T>C/ miR-30c A>G allele combination T-T-G and the miR-10a A>T/ miR-30c A>G/ miR-499 A allele combination C-G-G were found to be less frequent in RPL patients than controls, suggesting these combinations exert a protective effect (all p < 0.05). SNPs that occur in miRNA genes, miRNA machinery genes, or miRNAs that target genes involved in miRNA synthesis or function could adversely affect downstream gene expression [ 34 ]. Several studies have provided evidence supporting the critical role of miRNAs in RPL [ 35 ]. A previous study demonstrated that miR-499 was associated with the transforming growth factor (TGF)-β signaling pathway [ 24 ]. Furthermore, the 3′-UTR of the TGF-β3 gene has been shown to contain a putative binding site for miR-30c (rs928508) ( http://www.targetscan.org (accessed on 21 May 2018)), which targets the drug metabolism gene SULT1A1 [ 25 ]. Several TGF-β superfamily members perform critical functions in the female reproductive system. Specifically, these proteins regulate all processes of ovarian follicle development, including granulosa and theca cell proliferation, primordial follicle recruitment, gonadotropin receptor expression, ovulation, oocyte maturation, luteinization, and corpus luteum formation [ 36 ]. Additionally, the 3′-UTR of the prostaglandin F2 receptor inhibitor gene has been shown to contain a predicted binding target for miR-604 ( http://www.targetscan.org (accessed on 21 May 2018)), and prostaglandin F2 is required for placenta retention [ 37 ]. Furthermore, the miR-10a A>T polymorphism has been associated with regulation of IL-6 expression [ 26 ], and a previous study reported abnormal IL-6 expression in both animal models and patients with recurrent spontaneous abortions [ 38 ]. An online search for miR-10a, miR-30c, miR-181a, and miR-499b targets using the Target Scan and miRIAD databases ( http://bmi.ana.med.uni-muenchen.de/miriad/ (accessed on 21 May 2018)) returned many putative mRNA targets. Among these targets, we focused on PAI-1 for further functional analyses of miR-10a, miR-30c, and miR-181a because this gene has been shown to play several important roles in pregnancy and infertility [ 27 ]. PAI-1 is the primary inhibitor of plasminogen activators, including tPA and uPA. In the human placenta, PAI-1 is expressed in the extravillous interstitial and vascular trophoblasts. During implantation and placentation, PAI-1 inhibits extracellular matrix degradation, which thereby inhibits trophoblast invasion. We reviewed the literature regarding various reproductive diseases in which PAI-1 plays a role. Elevated PAI-1 levels have been detected in patients with RPL, preeclampsia, intrauterine growth restriction, gestational diabetes mellitus (GDM), endometriosis, and PCOS. Furthermore, both GDM and PCOS development have been reported to be related to the genetic role of the 4G/5G polymorphism in PAI-1. In general, elevated blood levels of PAI-1 are associated with an increased risk of infertility and poor pregnancy outcomes. In contrast, deficiency of PAI-1 results in transiently impaired placentation in mice [ 39 ], and deficiency of the PAI-1 gene is associated with abnormal bleeding after trauma or surgery in humans [ 40 ]. PAI-1 functions as a major inhibitor of fibrinolysis, and its overexpression leads to fibrin accumulation and placental insufficiency during pregnancy. PAI-1 acts as a major inhibitor of fibrinolysis, resulting in fibrin accumulation and insufficient placental formation due to overexpression. Previous reports also suggested that elevation of PAI-1 levels is the most frequent hemostasis-related abnormality associated with unexplained RPL [ 41 ]. Thus, increased expression of PAI-1 leading to inhibition of fibrinolysis is believed to be the main cause of RPL. To determine whether polymorphisms in miR-10a, miR-30c, and miR-181a affect target gene expression, we compared the expression levels of the 3′-UTR of PAI-1 harboring the different polymorphisms of miRNAs in Ishikawa human endometrial cells. Aberrant PAI-1 expression resulting from the expression of miR-10a with the A allele was significantly lower ( p < 0.05) than aberrant PAI-1 expression resulting from the expression of miR-10a with the T allele. In addition, the expression of miR-30c with the A allele was significantly lower ( p < 0.05) than expression of premature and mature miR-30c with the G allele. Expression of genotypes of miR-30c G as well as those of miR-10a T led to reduced expression of PAI-1 mRNA. These results suggest that SNPs in miR-30c and miR-10a regulate the expression of the PAI-1 gene. PAI-1-mediated inhibition of fibrinolysis and fibrin accumulation is currently believed to be the principal culprits for RPL; however, further studies are required to fully elucidate the underlying mechanisms. FSH is the primary gonadotropin responsible for regulating the progression of pregnancy [ 42 ]. Optimal levels of FSH, especially during the first few months of pregnancy, are critical for proper formation of the placenta [ 43 ]. Our clinical data indicated significant changes in FSH levels in RPL patients harboring the miR-30c A>G polymorphism. We, therefore, hypothesized that abnormal regulation of PAI-1 expression mediated by mutant miR-30c SNP results in aberrant FSH expression or disruption of the normal response to FSH. Imbalances in homocysteine and folate levels in particular are thought to contribute to low birth weight [ 44 ]. Specifically, higher homocysteine and lower folate concentrations during early pregnancy have been reported to be associated with lower placental weight and birth weight. However, we did not observe any associations between folate and homocysteine concentrations and placental weight. We found that the dominant miR-499b AG genotype (AA vs. AG + GG) was significantly more frequent in RPL patients ( p < 0.05). Earlier studies used a global approach to identify and profile miRNA expression at important stages during the estrous cycle and found a role of miRNAs in ovulation. Additionally, one-way ANOVA analysis of variance of data from RPL patients ( Table 6 ) revealed that in comparison with miR-30c AA, the miR-30c GG genotype was associated with significantly lower aPTT, E2 (pg/mL), Hct, and T. chol (mg/dL) and significantly higher creatinine (mg/dL) and FSH (mIU/mL). Compared with miR-181a TT, the miR-181a CC genotype was associated with significantly higher homocysteine levels, suggesting this genotype is associated with increased risk of RPL ( p < 0.05). Compared with miR-181a TT, the miR-181a TC genotype was associated with significantly higher T. chol levels, suggesting this genotype is associated with increased risk of RPL ( p < 0.05). However, in the case of creatinine levels, the miR-181a TC genotype was associated with significantly lower levels than the miR-181a TT genotype, indicating a protective effect, although the results were inconsistent with OR and therefore, the difference was not significant.

We investigated the relationship between various miRNA polymorphisms and the occurrence and risk of RPL. Several genotypes and allele combinations were positively correlated with RPL occurrence and unfavorable prognoses according to reproductive disease risk factors, including FSH, LH, and E2 levels. However, this study has several limitations. First, how the miRNA polymorphisms in the PAI-1 gene affect the development of RPL remains unclear. In addition to studies of PAI-1 , future follow-up studies of other RPL-related genes and the miR-10a and miR-30c targets are planned, particularly studies of the role of genes related to the TGF-β signaling pathway. As TGF-β regulates cell proliferation, apoptosis, and homeostasis, it plays a critical role in regulating the progression of pregnancy. Second, the control subjects in our study were not completely healthy because some of them had sought medical attention for other issues. Our experience shows that recruiting healthy participants through imaging and laboratory testing results in significantly reduced enrollment rates. However, enrollment of participants without imaging and laboratory testing can introduce another challenge to risk factor assessment. Lastly, the study population was restricted to Korean patients. Although the results of our study provide the first evidence suggesting that miRNA polymorphisms in the PAI-1 gene may serve as diagnostic and prognostic biomarkers for RPL, a prospective study involving a larger cohort of patients is warranted to validate these findings. A genome-wide analysis (using transcriptome-seq and miRNA-seq) is needed to identify the primary target genes, particularly the common genes regulated by these miRNAs. Determining the expression of these genes in the relevant gene-miRNA networks would provide stronger evidence in support of the results of the present research.