Introduction

Recurrent pregnancy loss (RPL) is defined by multiple, consecutive pregnancy loss. Some experts consider the consecutive loss of two pregnancies as RPL, whereas others restrict the definition of RPL to three or more losses. Depending on which definition is applied, it is estimated that 2–4% of women attempting to become pregnant experience1. Despite a considerable amount of recent progress identifying the causes of RPL, the underlying factors remain unknown for approximately 50% of RPL cases2. Among the possible causes are cytogenetic abnormalities, antiphospholipid syndrome, anatomical factors, and hormonal factors, but it appears that few cases are caused by a single pathogenic factor. Rather, most cases are attributed to an interaction of more than one genetic risk factor. The uncertainty about the causes of RPL has made it difficult to identify the genes and biological mechanisms involved. Using a candidate gene approach in tandem with genome-wide association studies, researchers have been able to identify low- and moderate-penetrance alleles associated with RPL risk. Although the identified risk variants cannot fully explain the RPL heritability, these variants of the genetic area were an important discovery3. Furthermore, there is accumulating evidence of differential association between specific genetic factors and distinct RPL subtypes4,5,6.

The microRNA (miRNA) is a 21- to 24-bp non-coding RNA that base pairs with its complementary sequences in an mRNA and silences the mRNA by several possible mechanisms. MiRNAs are important in many pathological conditions and aberrant miRNA expression can have serious consequences. Aberrant miRNAs have been associated with both RPL and endometriosis7 and recent studies revealed that miRNAs, which are associated with poor pregnancy outcomes, exist in bodily fluids of women and are secreted by culture-grown cells8. Furthermore, it has been reported that mature miRNA present in exosomes are associated with the development of autoimmune diseases such as Hashimoto’s disease and Graves’ disease9. In additions, miRNAs are well-known regulators of cell cycle progression, proliferation, and differentiation in the endometrium during the menstrual cycle10,11,12. All of this evidence supports a role for miRNAs in reproductive processes. The regulation of miRNA biogenesis and processing is mediated by various proteins, such as Drosha, Dicer, Exportin 5 (XPO5), DGCR8, and proteins of the Argonaute (AGO) family, along with several enzymes13. Pre-miRNAs are synthesized in the nucleus and are then transported by XPO5 and Ran-GTP into the cytoplasm where they are truncated by the RNase Dicer to form miRNA duplexes14,15. These duplexes then interact with AGO proteins present within the RNA-induced silencing complex (RISC), resulting in the formation of a functional RISC. Within the RISC, the two strands of the miRNA duplex are divided; one strand is degraded and the other acts as a template to direct binding and subsequent silencing of the target mRNA16,17.

The human AGO family is separated into four subfamilies. The only subfamily with catalytic activity is AGO2, and this subfamily serves a critical function within the RISC18. The RNA and proteins associated with AGO1 and AGO2 are present at considerable levels in many body tissues, which previously led us to focus on these two subfamilies19. AGO1 inhibits the proliferation and motility of cell through inducing apoptosis20, and regulates genes that influence growth, survival, and the cell cycle progression21. In contrast, AGO2 has been shown to be upregulated in numerous cancers and is associated with the growth of tumor cells and overall patient survival22. In a mouse model study, Ago2 regulated protein expression in mouse embryos, and this had important effects on the progression of blastocyst differentiation23. Furthermore, deletions of both Ago1 and Ago2 affect the formation and cleavage activity of RISC, and the deletion of Ago2 is associated with down-regulation of miRNAs in other tissues24. Overall, these findings reveal that miRNAs may be important for a successful pregnancy and the AGO protein is central to the functioning of miRNAs. Therefore, we hypothesized that the AGO protein is a susceptibility factor for RPL, as disruption of the AGO protein would disrupts miRNA function.

Here, we examined the associations of AGO1 and AGO2 gene polymorphisms with RPL pathogenesis and prognosis in a Korean population. Specifically, we examined two polymorphisms each for AGO1 (rs595961, rs636832) and AGO2 (rs2292779, rs4961280) because these polymorphisms have been studied previously and are already reported to be associated with other diseases. To our knowledge, this study is the first to provide evidence that AGO1 and AGO2 polymorphisms play a role in RPL of Korean women.

Results

The baseline characteristics

The baseline characteristics of the RPL patients and controls are shown in Table 1. The hematocrit, platelet count (PLT), and estradiol concentration (E2) in the RPL patients were greater than in the control group controls (P = 0.001, P = 0.003, P = 0.002, respectively). The concentration of follicle-stimulating hormone, concentration of luteinizing hormone, and prothrombin time were all significantly different between RPL patients and members of the control group (all P < 0.0001). No other significant differences were observed in the parameters mentioned.

Table 1 Baseline characteristics between RPL patients and controls.
Full size table

Genotype frequency analyses of the AGO1 and AGO2 gene polymorphisms between RPL patients, subgroups of RPL patients, and controls

To confirm that depending on the increasing number of pregnancy losses was associated with AGO1 and AGO2 gene polymorphisms, the patient subgroup was further divided into two groups based on a number of pregnancy losses. The first group is that had three or more pregnancy loss (PL) (subgroup PL ≥ 3), and the second group is that had four or more PL (subgroup PL ≥ 4). We investigated the AGO1 polymorphisms rs595961G>A and rs636832A>G, and the AGO2 polymorphisms rs2292779C>G and rs4961280C>A, in all groups. The results are shown in Table 2. The genotype frequencies of the polymorphisms were satisfied in Hardy-Weinberg equilibrium (P > 0.05). The AGO1 polymorphisms rs595961G>A and rs636832A>G was associated with prevalence of RPL prevalence in the subgroup PL ≥ 4 (Table 2). These two polymorphisms were significantly associated with RPL under the recessive model (AGO1 rs595961 GG + GA vs. AA: AOR = 4.008, 95% CI = 1.047–15.349, P = 0.043; rs636832 AA + AG vs. GG: AOR = 2.821, 95% CI = 1.210–6.577, P = 0.016). However, we did not detect a significant association between RPL and the polymorphisms in total RPL patients (PL ≥ 2) or and the subgroup (PL ≥ 3).

Table 2 Genotype frequency analyses of polymorphisms of genes for the Argonaute proteins AGO1 and AGO2 in RPL patients, subgroups of patients with RPL and controls.
Full size table

Allele combinations analysis of AGO1 and AGO2 polymorphisms in RPL patients and control women

Table 3 shows allele combination models and the frequencies in which they were observed in the RPL and control groups. We analyzed allele combinations for all four polymorphism and observed an association between seven allele combinations (G-A-C-A, G-A-G-C, G-G-G-C, A-A-C-C, A-A-G-C, A-G-C-C, A-G-G-C) and RPL risk (Table 3). Among them, the combinations G-A-C-A (AOR = 3.705), G-A-G-C (AOR = 1.347), A-G-C-C (AOR = 4.137), and A-G-G-C (AOR = 5.736) had an increased association with RPL prevalence compared to the control group, while the other allele combination models had a decreased association with RPL compared to the control group. Furthermore, this tendency held for allele combination analysis of two and three polymorphisms. Particularly, when the allele combination included the minor allele of AGO1 rs595961G>A and rs636832A>G, we observed increased association with RPL (Table 3). For example, all of the combinations A-G-C (AGO1 rs595961/AGO1 rs636832/AGO2 rs2292779; AOR = 3.903, 95% CI = 2.122–7.179, P < 0.0001), A-G-C (AGO1 rs595961/AGO1 rs636832/AGO2 rs4961280; AOR = 3.796, 95% CI = 2.260–6.375, P < 0.0001), and A-G (AGO1 rs595961/AGO1 rs636832; AOR = 2.961, 95% CI = 1.839–4.767, P < 0.0001) had increased odds ratios for RPL risk compared to the control subjects. In addition, we performed a combination analysis that compared genotype combination frequencies in RPL patients and control subjects (Supplementary Table S2).

Table 3 Analyses of allele combinations of AGO1 and AGO2 polymorphisms in RPL patients and controls.
Full size table

Differences in clinical traits with respect to AGO1 and AGO2 polymorphisms

We measured the platelet count of peripheral blood, fasting blood sugar, the hematocrit, body mass index, plasma levels of fasting total homocysteine, and triglyceride levels (Table 4). The platelet counts and fasting blood sugar level were elevated in RPL patients with the AGO1 rs636832 GG genotype compared to the AGO1 rs636832 AA and AG genotypes (Fig. 1). The hematocrit was lower in control individuals with the AGO1 rs636832 GG and AGO2 rs2292779 GG genotypes compared to the other genotypes. Also, the plasma triglyceride levels were elevated in patients with the AGO2 rs4961280 AA genotype (Table 4). Other factors were not associated with the polymorphisms examined here.

Table 4 Clinical parameters of individuals in the study with different genotypes with respect to AGO1 and AGO2 polymorphisms.
Full size table
Figure 1
figure 1

Association between platelet counts, fasting blood sugar levels, and AGO1 rs636832A>G polymorphisms in patients with RPL. (A) Patients with the AGO1 rs636832 GG genotype had significantly higher platelet levels than patients with the AGO1 rs636832 AA genotype or AG genotype. (B) Patients with the AGO1 rs636832 AG genotype had significantly flower fasting blood sugar levels than patients with the AGO1 rs636832 GG genotype. Abbreviation: RPL, recurrent pregnancy loss.

Full size image

Discussion

RPL is defined repeated consecutive spontaneous abortions. Important factors involved in recurrent early pregnancy loss are genetic, endocrine, anatomical, and immunological in origin25. At present, the correlation between RPL and genetic polymorphisms that have been identified with next-generation sequencing is unclear, although genetic polymorphisms have been noted to confer susceptibility to RPL26. In this study, the association between four polymorphisms of AGO1 and AGO2 genes and the susceptibility for RPL in a Korean population was examined, specifically AGO1 rs595961G>A, AGO1 rs636832A>G, AGO2 rs2292779C>G, and AGO2 rs4961280C>A. These four polymorphisms were significantly associated with the RPL patient group and the association between the polymorphisms and RPL can be explained by mechanisms other than protein expression: the polymorphism may affect the splicing of precursor mRNA and the conformation and function of proteins associated with this process27; or the polymorphisms may affect the structure of AGO proteins rather than their quantity. We believe the present study is the first to reveal a relationship between susceptibility to RPL and AGO1 and AGO2 polymorphisms.

The miRNAs have been implicated in numerous pathological conditions13 and in the regulation of biochemical pathways, including cell proliferation, differentiation, metabolism, apoptosis, development, inflammation, and immunity in many eukaryotic organisms28,29. The mature miRNA in exosomes has been associated with immune regulation30. Previous studies identified an association between AGO1 and AGO2 and an angiogenesis defect model associated with the inflammation31. Furthermore, AGO1 was associated with an angiogenic pathway involving hypoxia-responsive miRNAs that could be a potentially suitable target for anti- or pro-angiogenesis32. Importantly, the regulation of AGO2 is reportedly the safety mechanism that limits the range of the anti-inflammatory activity of miR-146a33. Our previous study also demonstrated an association between miR-146a and risk of recurrent implantation failure34. As noted, AGO2 is the catalytic core of mammalian RISC involved in miRNA expression, and is reportedly essential to mammalian gastrulation and mesoderm formation35,36,37,38. Since, it is well-known that different miRNAs can bind to distinct sequences caused by polymorphisms39. Interestingly, the expression of vascular endothelial growth factor (VEGF) was significantly correlated with the expression of AGO2 protein40. Therefore, there is a well-established link between AGO1 and AGO2 and RPL, based in part on the relationship between AGO proteins, immune regulation, and autoimmune diseases30,41.

In accord with our hypothesis, our analyses revealed that the AGO1 rs595961G>A and AGO1 rs636832A>G genotypes were significantly associated with the prevalence of RPL in this study. Several factors associated with RPL, specifically, a patient’s hematocrit, platelet count, body mass index, homocysteine, triglyceride levels, and fasting blood sugar have been shown to be significantly associated with AGO1 and AGO2 gene polymorphisms in this study. Interestingly, in a recent survey, it was concluded that that there are no differences in effective factors between two and three or more consecutive pregnancy losses42. However, we found associations between AGO1 and AGO2 polymorphisms and number of RPLs. Specifically, the occurrence of four or more PLs was associated with AGO1 rs595961G>A and AGO1 rs636832A>G polymorphisms.

In the present study, we found that the AGO1 rs595961G>A genotype was associated with decreased body mass index in RPL patients. Moreover, the AGO2 rs4961280C>A genotype was associated with the decreased hematocrit in RPL patients. The previous studies were demonstrated that Ago2 had mediated the functions in the control of hematopoiesis in the mouse model and hematopoietic stem cell model, and was reported that had functioned as a critical regulator of erythropoiesis in the mouse model43,44. These result and previous studies were showed the possibility that this polymorphism of AGO2 might affect the process of hematopoiesis or erythropoiesis. Also, the AGO1rs636832A>G and AGO2rs2292779C>G genotypes were associated with the lower homocysteine levels in RPL patients. Furthermore, patients with the AGO1 rs636832GG genotype had significantly higher platelet counts than patients with the AGO1 rs636832AA or AG genotypes (Fig. 1A). In addition, patients with the AGO1rs636832AG genotype had significantly lower fasting blood sugar levels than patients with the AGO1rs636832GG genotype (Fig. 1B). We conclude that the AGO1 polymorphisms studied here are associated with RPL.

There had a few limitations, which should be considered when interpreting the results, in this study. If we were progressed the functional study of other polymorphisms of Argonaute in the future, we will gain more powerful evidence. However, we have focused on the study that would be finding the biomarkers of RPL, and we have progressed that associated the study between various polymorphism of Argonaute located in intron variants and the prevalence of RPL. Therefore, the functional study of these polymorphisms of Argonaute is not possible to the progressing because of these polymorphisms were located in the intron variants. Moreover, the study groups were small because we relied on a single medical center and selected participants based on strict criteria. Despite these limitations, we detected statistically significant differences between the patient and control groups. If our results are confirmed in larger, multicenter studies, the regulation of AGO1 and AGO2 may serve as a biomarker to the diagnosis of RPL.

In conclusion, we have identified associations between polymorphisms in the AGO1 gene and prevalence of RPL in a Korean population, as well as the significant clinical effects. These polymorphisms of AGO1 gene were located to intron variant but showed that have the possibility of biomarkers to diagnose and estimate the risk of RPL. Furthermore, we suggest that these polymorphisms may be associated with the pathogenesis of RPL.

Methods

Study population

The population in this study comprised 631 participants. The control and patient groups met strict criteria. The 246 women in the control group were recruited at the CHA Bundang Medical Center. Each control subject had a normal (46, XX) karyotype, regularly occurring menstrual cycles, a history of one or more pregnancies that were conceived naturally, and no prior history of pregnancy losses. The 385 women in the patient group were diagnosed with RPL between March 1999 and December 2012 at the Department of Obstetrics and Gynecology of CHA Bundang Medical Center, CHA University (Seongnam, Republic of Korea)45. RPL was defined as a minimum of two consecutive losses of pregnancy, and pregnancy loss was diagnosed by various clinical tests, namely, ultrasound, human chorionic gonadotropin testing, or physical examination prior to 20 weeks gestation2. RPL subgroups were assigned according to the number of pregnancy losses. The study excluded patients with pregnancy losses caused by infectious, hormonal, anatomical, autoimmune, chromosomal, or thrombotic factors. Anatomical abnormalities were determined using sonography, hysterosalpingogram, hysteroscopy, magnetic resonance imaging, or computed tomography. Hormonal causes of RPL, such as thyroid disease, luteal insufficiency, and hyperprolactinemia, were identified by measuring peripheral blood levels of follicle-stimulating hormone, luteinizing hormone, prolactin, thyroid-stimulating hormone, progesterone, and free T4. The autoimmune factors lupus and antiphospholipid syndrome were assessed using lupus anticoagulant and anti-cardiolipin antibodies, respectively. Thrombophilia, a thrombotic factor for RPL, was evaluated from levels of anti-β2 glycoprotein antibody and deficiencies in proteins C and S. A total of the 481 patients were initially evaluated for the study, but 96 patients exhibited trisomy, hypothyroidism, intrauterine adhesion, antiphospholipid syndrome, or chromosomal translocation (patients or spouses), and were therefore excluded from the study, leaving 385 subjects in the patient group5. This study was approved by the CHA Bundang Medical Center Institutional Review Board (IRB) and the Hospital Ethics Committee (IRB No. BD2010-123D). All study protocols abided by the recommendations of the Declaration of Helsinki and written informed consent was obtained from all study participants.

Evaluation of clinical factors in control and patient samples

Blood samples of RPL patients were obtained for a period of pregnancy and were used to measure blood coagulation factors, urate concentration, total cholesterol, and levels of folate, and homocysteine after fasting for 12 hours. Homocysteine levels were measured on an Abbott IMx analyzer (Abbott Laboratories, Abbott Park, IL, USA) using a fluorescence polarization immunoassay46. Folate, creatinine, and blood urea nitrogen (BUN) levels were measured by competitive immunoassay using ACS:180 (Bayer Diagnostics, Tarrytown, NY, USA)46. Urate and total cholesterol levels were measured with enzymatic colorimetric tests (Roche Diagnostics, Mannheim, Germany). An automated photo-optical coagulometer (ACL TOP; Mitsubishi Chemical Medicine, Tokyo, Japan) was used to measure prothrombin time (PT), and activated partial thromboplastin time (aPTT), and a Sysmex XE2100 automated hematology analyzer (Sysmex, Kobe, Japan) was used to determine platelet counts46. Hormone levels were measured in one of two ways, depending on the hormone; radioimmunoassays were used to measure estradiol, thyroid-stimulating hormone and prolactin (Beckman Coulter, Inc., Brea, CA, USA), and enzyme immunoassays were used to measure follicle-stimulating hormone and luteinizing hormone (Siemens AG, Munich, Germany)47. The proportion of the CD56+ natural killer cells proportion in peripheral blood was determined by flow cytometry (FACS Calibur; BD Biosciences, USA)48.

Genotyping

A Genomic DNA Extraction Kit (iNtRON Biotechnology, Seongnam, Republic of Korea) was used to extract genomic DNA from the peripheral blood collected from patients and controls. We examined the following four polymorphisms in this study: AGO1 rs595961G>A, AGO1 rs636832A>G, AGO2 rs2292779C>G, and AGO2 rs4961280C>A. The classification of alleles of the genetic polymorphism was confirmed to the East Asian population on the 1000Genomes study. The polymorphisms were assessed using polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) with the following primers: AGO1 rs595961G>A, forward 5′-CCC TAC ATC CAG GAA TTT GGG-3′ and reverse 5′-TCG ACA CTG TTT TTG GGG TG-3′; AGO1 rs636832G>A, forward 5′-CTG ATT CCA GAA CAT ATC ACT CAT-3′ and reverse 5′-GGT ATA CCC AGA GAC TGA AAG TAA A-3′; AGO2 rs2292779C>G, forward 5′-CGG AAC AAG CAG TTC CAC AC-3′ and reverse 5′-TGA CAG GGA AAG GCT GAT GA-3′; and AGO2 rs4961280C>A, forward 5′-TGC CCC TGT CTC CTT CAC ATG TCC-3′ and reverse 5′-GTT CCC CAA CAC AGC GCT CAA AGG-3′. The four polymorphisms were confirmed by digestion with restriction enzymes NlaIII AciI, BfaI, and Hpy166II (New England Bio Laboratories, MA, USA) for 16 hours at 37 °C. For the AGO1 rs595961G>A polymorphism, the wild-type genotype GG was confirmed by the presence of one 349 bp band after enzymatic digestion, the heterozygous genotype GA was confirmed by the detection of three bands at 349, 203, and 146 bp after enzymatic digestion, and the mutant genotype AA was confirmed by the detection of two bands at 203 and 146 bp after enzymatic digestion. For the AGO2 rs2292779C>G polymorphism, the wild-type genotype CC was confirmed by the presence of two bands at 142 and 99 bp, the heterozygous genotype CG was confirmed by the presence of three bands at 241, 142, and 99 bp, and the mutant genotype GG was confirmed by the presence of a single band at 241 bp. For the AGO1 rs636832G>A polymorphism, the presence of two bands at 96 bp and 25 bp were indicative of the wild-type genotype GG, three bands at 121 bp, 96 bp, and 25 bp were indicative of the heterozygous genotype GA, and a single band at 121 bp were indicative of the mutant genotype AA. The digestion fragments for the AGO2 rs4961280C>A polymorphism differed only slightly between genotypes; the wild-type genotype CC produced only one band at 169 bp, and the heterozygous genotype CA produced three band at 169 bp, 146 bp, and 23 bp, and the mutant genotype AA produced two band at 146 bp and 23 bp. Information regarding the PCR primers, restriction enzymes, and genotype-specific fragment sizes for each polymorphism is provided in Supplementary Table S1. Genotypes were determined by resolving the digestion fragments on 4% agarose gels using electrophoretic separation. A random 30% of the PCR product for each AGO polymorphism was used in a duplicate PCR assay, the product of which was subjected to DNA sequencing on an ABI 3730xl DNA Analyzer (Applied Biosystems, Foster City, CA, USA) to verify the RFLP results. There was 100% concurrence between the RFLP and sequencing results for each sample.

Statistical analysis

The frequencies of the AGO1 and AGO2 polymorphisms were compared between the patient and control groups using logistic regression analysis and Fisher’s exact test. Adjusted odds ratio (AOR) and 95% confidence intervals (CIs) were determined from age-adjusted logistic regression and were used to evaluate the association between each polymorphism and RPL incidence. Additionally, the 95% CI and AOR were used to evaluate the association of each polymorphism and allele combination. A P-value < 0.05 was deemed to indicate statistical significance. None of the polymorphisms examined in this study deviated from Hardy-Weinberg equilibrium (P > 0.05). Statistical analyses were conducted with following software: GraphPad Prism 4.0 (GraphPad Software, Inc., San Diego, CA, USA); HAPSTAT 3.0 (University of N. Carolina, Chapel Hill, NC, USA)5; and StatsDirect statistics software version 2.4.4 (StatsDirect Ltd., Altrincham, UK).