Abstract

Objective. The purpose of this study was to (i) determine the cervical microbial composition in different abortion samples and to (ii) investigate the correlation between spontaneous abortion and cervical microbes in Korean women. Methods. We collected cervical swabs from women who had never undergone abortion (), had spontaneous abortion (), and had undergone induced abortion () and subjected those samples to 16S rRNA pyrosequencing. Further, factor analysis and correlation between cervical microbiota and spontaneous abortion were evaluated by logistic regression analysis. Results. In spontaneous abortion women, 16 S rRNA gene sequences showed significant increases in Atopobium vaginae, Megasphaera spp., Gardnerella vaginalis, Leptotrichia amnionii, and Sneathia sanguinegens compared to women in nonabortion group. In multivariate logistic regression analysis, A. vaginae (OD = 11.27; 95% = 1.57–81), L. amnionii (OD = 11.47; 95% = 1.22–107.94), S. sanguinegens (OD = 6.89; 95% = 1.07–44.33), and factor 1 microbes (OD = 16.4; 95% = 1.88–42.5) were strongly associated with spontaneous abortion. Conclusions. This study showed a high prevalence of L. amnionii, A. vaginae, S. sanguinegens, and factor 1 microbes in spontaneous abortion and association with spontaneous abortion in Korean women.

1. Introduction

Cervical microbiota play a prominent role in women’s reproductive health, which is influenced by numerous factors including age, ethnicity, genetic factors, cultural and economic factors, personal hygiene, sexual activity, and environmental conditions [1, 2]. Cervical microbiota are liable to change throughout a woman’s lifetime (birth, puberty, and menopause) [3]. Such changes in vaginal microbial flora have serious consequences such as prevention of fertilization and induction of spontaneous abortion in pregnant women, as well as increased risk of preterm birth and low birth weight [48]. Spontaneous abortion prior to 20 weeks is a common adverse outcome of pregnancy. Moreover, spontaneous abortion and other adverse pregnancy outcomes have been traced to be associated with bacteria and viruses infection [9, 10]. A recent study using next-generation sequencing (NGS) techniques demonstrated a vaginal microbiota difference between preterm delivery and normal spontaneous delivery [11], and also there is evidence on cervical microbiota associations with pelvic inflammatory disease, infertility, cervical intraepithelial neoplasia, and obesity [1215]. However, the epidemiological data on the potential association between cervical microbiota and spontaneous abortion has been rarely reported and is lacking. The objective of this study was to (i) determine the cervical microbial composition in different abortion samples and to (ii) investigate the correlation between spontaneous abortion and cervical microbes in Korean women.

2. Methods

2.1. Subject Selection and Sample Collection

This study was approved by the Institutional Review Board of the National Cancer Center (IRB numbers NCCNCS-06-062 and NCCNCS 2016-0147). Written informed consent was obtained from all participants. We confirm that all experiments were performed in accordance with relevant guidelines and regulations. The study included women between 18 and 65 years of age who had participated in the Korean Prospective Study of the Transition of Human Papillomavirus into Cervical Carcinoma from 2006 to 2013 [16, 17]. These women were randomly selected from the Gynecology and Oncology clinic in six University hospitals, in South Korea. Eligible women were currently sexually active or seeking birth control, not currently pregnant, and had an intact uterus and no personal history of cervical intraepithelial neoplasia within 18 months. The exclusion criteria were women with a history of cervical cancer, incomplete questionnaire, inadequate blood sample, chronic diseases (liver cirrhosis, renal failure), cardiovascular disease, drug dependency, or psychological problems. Regarding the history of abortion, subjects were classified into 3 groups: 36 nonabortion, 23 spontaneous abortions, and 88 induced abortions among a total of 147 subjects with cervical swab samples.

2.2. Questionnaires Related to History of Abortion

Detailed interviewer-administered comprehensive health and lifestyle questionnaires, including questions on behavior related to abortion, such as history of spontaneous abortion and induced abortion and the number of each type of abortion, were completed at enrolment in the outpatient Department of Gynecology and Oncology clinic. The questionnaire included reproductive (menarche age, the number of pregnancies, the number of childbirths, gestational age, and breast feeding) and menstrual history (menopausal status); exogenous hormone use before pregnancy and after menopause, medical history, family history of cervical cancer, and sociodemographic and lifestyle characteristics were recorded. Pathological and laboratory data were collected, recorded, and entered into the epidemiological database, National Cancer Center. Medical charts and pathology reports were examined to insure that control subjects had no history of any cancer or precancerous lesions.

2.3. HR-HPV DNA Detection and Pap Smear

Upon study entry, the participants underwent a physical and gynecological examination and had Hybrid Capture 2 testing and Papanicolaou (Pap) smears. The cervical cytological findings were classified according to the Bethesda system [18]. Cervical samples were collected using a Cervix brush (Rovers Medical Devices, Oss, the Netherlands), and the brush was immediately rinsed in a vial of PreservCyt solution (Cytyc Corporation, Marlborough, MA, USA), and the vial was placed in a Thin Prep (Cytyc Corporation, Marlborough, MA, USA) Processor. Collected samples were stored at −80°C for further analysis. The chemiluminescent HPV DNA test yielded relative light units (RLU) using a probe designed to detect 13 HR-HPV types. HPV DNA detection was performed with the Digene HC2 high-risk DNA test (Qiagen, Gaithersburg, MD, USA) with signal amplification and chemiluminescence for detection of 13 types of HR-HPV scored in relative light units (RLU). The test results were read as positive at concentrations of 1 pg/ml or levels greater than the RLU/cutoff ratio (RLU of the specimen/mean RLU of 2 positive controls).

2.4. DNA Extraction and Pyrosequencing

Genomic DNA was extracted from the cervical samples by Fast DNA SPIN extraction kits (MP Biomedicals, Santa Ana, CA, USA) following manufacturer’s instruction. Isolated DNA from cervical samples was used as template to amplify V1–V3 regions using bar-coded primers. The PCR reaction was performed in a final volume of 50 μL containing 10x Taq buffer, a dNTP mixture (Takara, Japan), 10 μM of the bar-coded fusion primers, and 2 U of Taq polymerase (ExTaq, Takara). The PCR program was as follows: initial denaturation (94°C for 5 min), product amplification, 30 cycles (30 s, 94°C), primer annealing (30 s, 55°C), and extension (30 s, 72°C), followed by a final extension for 7 min at 72°C. The amplified product was checked by 2% agarose gel electrophoresis and visualized under a Gel Doc system (Bio-Rad). The amplified products were purified with a QIAquick PCR purification kit (Qiagen, Valencia, CA, USA) and quantified using a PicoGreen dsDNA Assay kit (Invitrogen, Carlsbad, CA, USA). Equimolar concentrations of each amplicon from different samples were pooled and purified using an AMPure bead kit (Agencourt Bioscience, Beverly, MA, USA) and then amplified on sequencing beads by emulsion PCR. The beads recovered following emulsion PCR were deposited on a 454 Pico Titer Plate, and sequencing was performed using a Roche/454 GS Junior system (Roche, Branford, CT, USA). Raw MiSeq reads were demultiplexed according to the barcodes and trimmed by in-house Perl scripts for quality filtering (quality score > 25). The processed paired reads were assembled, and the assembled reads were used for operational taxonomic unit picking. Quantitative Insights into Microbial Ecology was used for the microbial community analysis. In order to confirm the richness and diversity of the bacterial types in the samples, the Chao1 and Shannon indices were calculated. The 16S rRNA gene sequences obtained from pyrosequencing have been available at the EMBL SRA database (http://www.ebi.ac.uk/ena/data/view/PRJEB5760).

2.5. Statistical Analysis

The distributional differences of the continuous and categorical variables among the groups were examined by -test or ANOVA and chi-square test, respectively. To compare the relative abundance differences of microbes among the three abortion groups, we used Wilcoxon rank sum test. Factor analysis was performed to identify the microbial patterns of 45 microbial species (filtered by a 0.1%-or-over rate in individual proportions) using the FACTOR PROCEDURE in SAS (version 9.4; SAS Institute, Chicago, IL, USA). The factors were rotated by an orthogonal transformation (Varimax rotation function in SAS) to achieve a simpler structure with greater interpretability. After the Varimax rotation, the factor scores were saved from the principal component analysis for each individual. All of the data presented here are from the Varimax rotation. Rank correlation analysis between the microbiota and epidemiological factors was performed by Somers’ multiple comparison test. Multivariate logistic regression analysis was performed after adjustment for age, BMI, menopausal status, alcohol drinking, smoking habit, and HR-HPV infection. The strength of the association of selected microbes and different abortion groups were reported as the odds ratio (OR) and 95% CI compared to a reference group.

3. Results

Table 1 lists the study participants’ general characteristics, which include the following epidemiological factors: age, body-mass index (BMI), marital status, menopausal status, number of children, education level, family income, oral contraceptive use, smoking status, alcohol-drinking status, and oncogenic high-risk human papillomavirus (HR-HPV) infection. The participants’ mean age was 44 years, and 65% were premenopausal. Significant differences in age, education level, and HR-HPV were observed in the spontaneous abortion group relative to the nonabortion group.

After quality control, a total of 1 431 278 valid reads were obtained, and an average of 92.2 operational taxonomic units (OTUs) per sample were observed by 16S rRNA pyrosequencing analysis. The sequence reads were assigned to 101 OTUs in the nonabortion group, 93 OTUs in the spontaneous abortion group, and 90 OTUs in the induced abortion group. The diversity (Shannon) and richness estimation (Chao 1) were used for measuring alpha diversity. The observed mean values of the Shannon indices and Chao 1 were 1.9 and 118.2 for nonabortion, 2.0 and 111.2 for spontaneous abortion, and 1.9 and 113.2 for induced abortion (Supplementary Figure  1 in Supplementary Material, available online at https://doi.org/10.1155/2017/5435089). Among abortion samples, the spontaneous abortion women showed high diversity and low richness compared to nonabortion women. As for the bacterial communities, taxonomic classification revealed that Firmicutes (73.5%, 54.0%, and 66.5%) was the most dominant phylum followed by Actinobacteria (8.72, 29.21, and 10.31%), Bacteroidetes (4.21, 9.03, and 8.51%), Proteobacteria (8.06, 3.28, and 5.30%), Fusobacteria (0.11, 2.48, and 3.49%), and Tenericutes (2.73, 1.02, and 4.00%) in nonabortion, spontaneous abortion, and induced abortion, respectively. The abundances of Firmicutes, Proteobacteria, and Tenericutes were lower in spontaneous abortion than in nonabortion.

Next, we investigated whether the relative abundances (>0.1%) of the cervical microbiota differed between abortion and nonabortion at the species level by 16S rRNA sequencing. We observed that women who had spontaneous abortion showed significant increases () in Atopobium vaginae, Megasphaera spp., Gardnerella vaginalis, Leptotrichia amnionii, and Sneathia sanguinegens compared to nonabortion group (Table 2). Also, Lactobacillus crispatus were higher in nonabortion, but not statistically significant (Supplementary Table  1). Further, the microbes which showed significant results are selected for logistic regression analysis (univariate and multivariate).

In factor analysis, a total of seven factors showing an eigenvalue greater than 1.5 were identified, and, for each, the factor-loading values of 45 species were provided (Table 3). Among the factors, factor 1 scored a high eigenvalue and was selected for logistic regression analysis. The high-scoring factor 1 microbes included Megasphaera sp. (0.711), P. amnii (0.528), P. timonensis (0.414), L. amnionii (0.400), A. vaginae (0.259), S. sanguinegens (0.208), and D. microaerophilic (0.146); the lowest-scoring factor 1 microbes were L. gasseri (−0.039), L. acidophilus (−0.025), L. crispatus (−0.298), L. vaginalis (−0.191), L. fornicalis (−0.133), L. jensenii (−0.115), and L. psittaci (−0.13). High eigenvalues also were observed for factors 4, 5, and 7, but these were not included in the subsequent logistic regression analysis; rather, they were used for comparison with the epidemiological factors in estimating the associations with abortion.

Next, rank correlation between epidemiological factors and the relative abundances of 45 microbial species (>0.1%) were conducted (Table 4). We noticed high positive associations between spontaneous abortion and A. vaginae (; ), S. sanguinegens (; ), G. vaginalis (; ), and L. amnionii (; ). On the other hand, negative correlations were shown with education, first parity age, and L. acidophilus. However, income and education level were positively associated with P. bivia and Lactobacillus sp. The other epidemiological factors showed only rare correlations with microbial species.

We assessed association between the selected microbes and spontaneous abortion using logistic regression model (Table 5). Based on the univariate analysis there was strong association between spontaneous abortion and high tertile A. vaginae, 5.38 (95% CI 1.68–17.29), S. sanguinegens, 5.14 (95% CI 1.35–19.54), L. amnionii, 4.26 (95% CI 1.11–16.42), and factor 1 microbes, 4.51 (95% CI 1.41–14.45). As for the multivariate analysis, the ORs were calculated after adjustment for age, menopause, BMI, smoking, alcohol, and HPV as categorical variables: L. amnionii 11.47 (95% CI 1.22–107.94), A. vaginae 11.27 (95% 1.57–81.0), and S. sanguinegens 6.89 (95% CI 1.07–44.33) were found to be associated with spontaneous abortion, and factor 1 microbes also showed a high score 16.4 (95% CI 1.88–42.5). The associations between Megasphaera spp. and G. vaginalis and spontaneous abortion were found to be nonsignificant. In the present study, the number of women using hormone replacement therapy (HRT) was small: 6 in nonabortion, 4 in spontaneous abortion, and 14 in induced abortion group. After excluding women with using HRT, the logistic result was almost the same (data not shown).

4. Discussion

By comparing the cervical microbiota profiles of different abortion groups, we found higher prevalence of cervical microbes such as L. amnionii, A. vaginae, S. sanguinegens, and factor 1 microbes in spontaneous abortion women and showed strong association with spontaneous abortion. In the general characteristics of our study subjects among the three abortion groups, we found significant differences in age, education, and oncogenic HPV infection rate between women who had spontaneous abortion and those who had undergone induced abortion (Table 1). With respect to age, we found, similarly to the result of Gracia et al. [19], that the highest percentage of spontaneous abortion cases (43.5%) was that of individuals within the 45–54 age range. Also, among those whose education ended at high school, the prevalences of spontaneous abortion and induced abortion were relatively high compared with that of the nonabortion group (Table 1). Our results agree with previous reports, which mention that spontaneous abortion women were more likely to be unmarried and to have had less education and a history of abortion and pelvic inflammatory disease [20]. Bacteria such as Lactobacillus fornicalis, L. jensenii, and L. vaginalis showed positive correlations with education groups (Table 4). A study conducted by Conde-Ferraez et al. reported that HPV infection was not significantly associated with spontaneous abortion [21]. But, in our study, we found that oncogenic HR-HPV infection was significantly associated with spontaneous abortion (). The possible mechanisms of spontaneous abortion by HPV infection are not yet clear, but there is already some evidence that HPV induces apoptosis of infected trophoblasts, thereby negatively affecting implantation and placental physiology [2225].

At the phylum level, the predominant cervical microbiota of each group was Firmicutes, Actinobacteria, Bacteroides, Proteobacteria, Tenericutes, and Fusobacteria. These observed relative abundances are similar to those observed in our earlier analysis of cervical microbiota in CIN samples [15]. Also, the OTUs were higher in the nonabortion group than in the spontaneous and induced abortion samples. We hypothesize that women in the spontaneous abortion group, relative to those in the nonabortion group, might have had cervical microbial dysbiosis, or that there were other factors such as the sample collection method, sequencing method, hygiene, glycogen level, menstrual cycle, or host-genetic factors [2628]. It is well known that healthy women harboring high numbers of Lactobacillus sp. produce lactic acid and hydrogen peroxide, thus preventing or suppressing the entry of G. vaginalis, Mobiluncus sp., Prevotella, and Bacteroides that are associated with BV [29, 30]. In this study, we found low relative abundances of L. inners, L. crispatus, and L. johnsonii and higher abundances of A. vaginae, A. christensenii, L. amnionii, P. amnii, L. fornicalis, U. parvum, M. hominis, and S. sanguinegens in spontaneous abortion women when compared with women without abortion history. These pathogens are reported to be associated with adverse pregnancy outcomes such as preterm delivery, abortion, chorioamnionitis, and BV [3133]. Similarly, Dasari et al. reported that low Lactobacillus population is associated with reduced vaginal-secretion leukocyte protease inhibitor and increased abnormal flora [34]. Additionally, the relative abundances of unclassified bacteria (AY958888, AY959109, and AY958940) significantly increased in spontaneous abortion compared with nonabortion women (Supplementary Table  1). To better characterize these unknown bacteria and their role in abortion, further studies are required.

Several studies have already established the correlation of high-load A. vaginae and G. vaginalis with preterm birth [35]. In this study, we found that L. amnionii, A. vaginae, S. sanguinegens, and factor 1 microbes are highly associated with spontaneous abortion (Table 5). The bacteria L. amnionii can be characterized as anaerobic, gram-negative, and pleomorphic coccobacillus found in the oral cavity and genital tract [36]. A recent Norwegian case study of L. amnionii was the first to isolate L. amnionii from renal abscess in spontaneous abortion patients with chorioamnionitis [37]. Although L. amnionii is associated with spontaneous abortion, the epidemiological evidence of an association of high-risk cervical microbiota with spontaneous abortion is lacking. S. sanguinegens is a gram-negative, anaerobic, nonmotile, and non-spore forming bacteria found in the gastrointestinal and female genital tracts. This bacteria is reported to be associated with bacterial vaginosis, a vaginal disorder in women of reproductive age worldwide, and is the most common genera detected in amniotic fluid; its presence can lead to inflammation, histological chorioamnionitis, and/or amnionitis [38, 39]. Some studies have demonstrated the phylogenetic relationship between L. amnionii and Sneathia, where the former has been assigned to the genus Sneathia [40]. The presence of A. vaginae is even associated with spontaneous abortion. A. vaginae is a species of gram-positive, rod-shaped anaerobic bacteria that can lead to maternal sepsis and spontaneous abortion [41]. Our hypothesis, therefore, is that high proportions of L. amnionii, S. sanguinegens, A. vaginae, and factor 1 microbes in the vagina can increase the likelihood of uterine, fetal-membrane, or fallopian-tube infection leading to spontaneous abortion. Though other notable organisms like Candida species and Trichomonas vaginalis are also associated with spontaneous abortion [42], the targeted 16S rRNA in the current experiment could not reveal their significant presence.

Strengths of our study are as follows. (1) To our knowledge, this is the first large-cohort study to explore the relative abundances of species associated with the cervical microbiota of women who had never undergone abortion, had spontaneous abortion, or had undergone induced abortion. (2) The use of the pyrosequencing method for identification of fastidious or uncultivable microbes reduced the bias/error as compared with cultivation-based microbiological methods.

However, we recognize certain limitations in this study. (1) Since the abortion status was determined using questionnaires, and the microbiome analysis of the cervical swabs was performed at enrolment, the timings of the event and swab sampling were not the same. This might have biased the findings of the present study. However, further studies are required to understand the importance of association between cervical microbial community and HPV persistence.

(2) 35% of the participants in the current experiment were postmenopausal; the correlation of microbiome profiles of pre- and postmenopausal participants with the abortion might vary significantly. (3) The small sample size might have limited the significance of the obtained results; however, it was certainly adequate for this study’s preliminary determination: that L. amnionii, A. vaginae, S. sanguinegens, and factor 1 microbe prevalence are highly associated with spontaneous abortion.

5. Conclusion

This study compared the cervical bacterial communities of different abortion women and revealed the association of high L. amnionii, A. vaginae, S. sanguinegens, and factor 1 microbes with spontaneous abortion.

Conflicts of Interest

The authors declare no competing financial interests.

Authors’ Contributions

Selvaraj Arokiyaraj, Mi Kyung Kim, and Sang Soo Seo conceived and designed the study. Selvaraj Arokiyaraj and Mi Kyung Kim wrote the manuscript. Sang Soo Seo, Jae Kwan Lee, and Ye Lee Yu provided patient samples and clinical information. Hea Young Oh and Ji Sook Kong collected the data and managed subjects biosample and DNA sequencing. Minji Kwon, Ji Sook Kong, and Moon Kyung Shin performed statistical analysis. Sang Soo Seo and Selvaraj Arokiyaraj contributed equally.

Acknowledgments

This work was supported by grants from the Korea National Cancer Center (Grants nos. 1610210 and 1310360).

Supplementary Materials

Supplementary table 1: Distribution of vaginal bacteria from non-abortion, spontaneous abortion and induced abortion women in Korea.

Supplementary figure 1: Shannon index (a) and Chao 1 (b) index across non-abortion, spontaneous abortion and induced abortion women.

  1. Supplementary Material