The American Society for Reproductive Medicine and the European Society of Human Reproduction and Embryology (ESHRE) defines recurrent spontaneous abortion (RSA) as two or more failed pregnancies. RSA affects approximately 1%–5% of couples trying to conceive (Practice Committee of the American Society for Reproductive Medicine, 2012; ESHRE Guideline Group on RPL et al., 2018). Existing research shows that the etiology of RSA includes genetic, anatomic, infective, thrombophilic, endocrine, and immune factors (Evaluation and treatment of recurrent pregnancy loss: a committee opinion, 2012; Rai and Regan, 2006; ESHRE Guideline Group on RPL et al., 2018; Khalife et al., 2019). Unfortunately, 50% patients do not have any of the above conditions and are hence considered idiopathic (Khalife et al., 2019; Green and O’Donoghue, 2019). This adds to patients’ financial and psychological burdens and also limits effective interventions and therapeutics. To date, international groups have not reached a consensus on the standard evaluation of RSA (Khalife et al., 2019). Various therapeutic strategies have been evaluated to decrease the rate of pregnancy loss. Low-dose aspirin, low-dose steroids, heparin, levothyroxine, progesterone, intralipid, and metformin are generally chosen to treat women with RSA through single or combined use (Rai and Regan, 2006; Vesce et al., 2014; Khalife et al., 2019; Green and O’Donoghue, 2019; Ou and Yu, 2019). The reported effects of the above-mentioned drug treatments are inconsistent. In-depth research on the etiology of RSA to determine the true causes will ensure a unified and standardized treatment in the future, thereby avoiding unreasonable treatment and reducing the financial and psychological burden of families with RSA.

With the development of high-throughput sequencing technology, an increasing number of microbiota inhabiting different parts of the human body have been associated with various diseases, which has expanded our understanding of the underlying mechanisms involved in these conditions. Although most of these studies focused on the gastrointestinal microbiome, the role of the vaginal microbiome for female reproductive health has attracted the attention of researchers over time (Chow and Mazmanian, 2010; Nikoopour and Singh, 2014; Younes et al., 2018). Two scientific conferences were held in Amsterdam to discuss the topic of women and vaginal microbiota in 2015 and 2016 to emphasize the particularity of the female microbiome and summarize the current research (Younes et al., 2018). Studies showed that the vaginal microbiota of healthy women are mainly colonized by Lactobacillus spp. including L. crispatus, L. gasseri, L. iners, and L. jensenii (Aroutcheva et al., 2001). Notably, some healthy women do not have Lactobacillus dominance, especially South African black women. The causative factors leading to ethnic and geographic differences in vaginal bacteria remain unclear (Anahtar et al., 2018). Ravel et al. analyzed the vaginal microbiota of women of childbearing age in North America and proposed five vaginal community structure types (CSTs) that have been commonly adopted in vaginal microbiome studies. The vaginal microbiota is affected by some physiological factors such as menarche, menstrual cycle, pregnancy, menopause, and other hormonal changes (Achilles et al., 2018; Greenbaum et al., 2019). Hickey et al. suggested that vaginal pH in healthy premenarche girls after menarche was usually higher than the level in healthy adult women, even with high proportions of Lactobacillus (Hickey et al., 2015). The overall diversity of the vaginal microbiome during pregnancy decreases while the stability increases. In addition, the abundance of the dominant Lactobacillus in the vagina increases significantly during pregnancy, which reduces the vaginal pH and strengthens the vagina’s ability to resist pathogenic microorganisms. After delivery, the microbiota revert to be similar to those in non-pregnant women (Neuman and Koren, 2017). Vaginal microbiota composition is also affected by exogenous factors including hygienic practices, contraceptive method, sexual behavior, stress, diet, exercise, drugs, and rectal colonization (Muzny and Schwebke, 2016; Abdelmaksoud et al., 2017; Achilles et al., 2018; Turpin et al., 2019; Song et al., 2020). According to existing literature reports, imbalanced vaginal microflora is related to the following diseases, including but not limited to, bacterial vaginosis, sexually transmitted infections, preterm birth (PTB), gynecological cancers, preterm pre-labor rupture of the fetal membranes, recurrent implantation failure, and polycystic ovary syndrome (PCOS) (Brotman, 2011; Muzny and Schwebke, 2016; Champer et al., 2017; Freitas et al., 2018; Brown et al., 2018; Elovitz et al., 2019; Fu et al., 2020; Hong et al., 2020). The relationship between PTB and the vaginal microbiome was recently described in several studies, with few reports on the vaginal microbiome of RSA (Zhang et al., 2019).

In the past, based on traditional serology and culture methods, Mycoplasma hominis, Ureaplasma urealyticum, Listeria monocytogenes, Gardnerella vaginalis, and other less frequent pathogens were identified more often in patients experiencing spontaneous abortion (Penta et al., 2003; Kuon et al., 2017). However, Contini et al. proposed that Mycoplasma hominis and Ureaplasma urealyticum may not cause embryo loss after analyzing bacterial DNA in the aborted tissues of women with early pregnancy loss and women underwent voluntary interruption of pregnancy via quantitative real-time PCR methods (Contini et al., 2018). With the development of technology, we will have a better understanding of the pathogenesis of abortion. Recent application of 16S rRNA gene-based metataxonomics for exploring the association between vaginal bacterial composition and abortion suggested that vaginal bacterial composition of abortion in the first trimester was related to decreased Lactobacillus spp. (Al-Memar et al., 2019). However, the authors could not conclusively show when the loss of Lactobacillus spp. occurs. In addition, a recent study has found that Atopobium, Streptococcus, and Prevotella were significantly more abundant in patients with unexplained recurrent miscarriage, while two (Lactobacillus and Gardnerella) taxa were overrepresented in controls (Zhang et al., 2019). As this study was limited by a small sample size, a definitive conclusion could not be drawn.

In our study, 108 patients with RSA were enrolled to characterize the vaginal microbiota composition in non-pregnant patients with RSA and assess the effects of drugs on bacterial composition using 16S rRNA gene-based metataxonomics.

The pathogenesis of RSA is complex, and at least half of the cases do not seem to have any concrete underlying cause. The use of next-generation sequencing technology to determine the characteristics of vaginal microbes in patients with RSA may provide new clues for exploring the etiology. A comprehensive understanding of the characteristics of vaginal microbiota in RSA and its response to drug treatment is essential to improve diagnosis and treatment strategies. Our study showed that patients with RSA have decreased richness of the vaginal microbiome, along with unchanged dominant Lactobacillus spp. and increased abundances of Megasphaera and Sneathia sanguinegens. Additionally, the vaginal microecology composition in the patients was different from that of healthy women. The difference was attenuated to a certain extent after drug treatment. Treatment with metformin combined with aspirin might significantly increase the abundance of vaginal Lactobacillus spp. in patients.

According to previous reports, factors including age, BMI, hygiene habits, diet, exercise intensity, smoking, and perceived stress can affect the vaginal microbiome (ESHRE Guideline Group on RPL et al., 2018; He et al., 2019; Khalife et al., 2019; Turpin et al., 2019; Song et al., 2020). In this study, all factors except for age were similar among the experimental groups ( Table 1 ). This is likely because healthy women in the control group had successfully conceived twice and were older than those in the case group. Based on the results of statistical analysis, the difference in age between the experimental groups has no effect on the comparison of differences in microbiota between groups.

As the dominant organism of the vaginal community, Lactobacilli play a key role in maintaining vaginal health. Lactobacillus species inhibit the growth of other microorganisms by producing lactic acid through metabolizing extracellular glycogen to lower the pH of the vaginal environment, producing bacteriocins and hydrogen peroxide (H₂O₂), and competing for nutrients and space, thereby protecting the vaginal ecosystem from harmful microbial communities. (Aroutcheva et al., 2001; Caselli et al., 2020). Therefore, vaginal pH (greater than 4.5) is one of the important indicators for the diagnosis of bacterial vaginosis in clinical practice (Verstraelen and Verhelst, 2009). H₂O₂ is generally considered to prevent bacteria associated with bacterial vaginosis (Eschenbach et al., 1989; Beigi et al., 2005). However, the ability of different types of Lactobacillus to produce organic acids, bacteriocins, and H₂O₂ is not completely the same (Eschenbach et al., 1989; Aroutcheva et al., 2001). Herein, there was no statistical difference in the measurement of vaginal pH and H₂O₂ in all participants between each group. This might be due to the similar distribution of vaginal Lactobacillus species, which can produce organic acids and H₂O₂ in each group. Sialidases are enzymes widely present in bacteria, fungi, viruses, mycoplasma, and animals that catalyze the removal of sialic acid from various glycoconjugates (Taylor, 1996). Many of these pathogens use sialidase to assist their pathogenesis and/or nutritional requirements. The detection of vaginal sialidase activity has also been used in the auxiliary diagnosis of bacterial vaginosis (Myziuk et al., 2003; Paladine and Desai, 2018). In the present study, vaginal sialidase activity was negative in all subjects. LE, an indicator of inflammatory cells present in tissues or samples, is used to evaluate urogenital tract infections and periprosthetic joint infection (Abbasi et al., 1985; Mårdh et al., 2003; Lee et al., 2017). It was found that LE activity of RSA group was higher than that of the control group, and the difference was statistically significant. No significant difference was found between no-medication group and drug treatment group in the test of LE activity. LE tests were weakly positive in 46.2% of patients with RSA. In the control group, there was only one positive case of LE test, accounting for 5.5%, and no weak positive case was found. This finding suggested that nearly half of the patients in case group may have mild inflammation in the vagina. The reason of this mild inflammation might have to do with changes in their vaginal microbiome composition.

Recent studies have found that the diversity and richness of vaginal microbiota in women who miscarried in the first and second trimesters were significantly higher than those of normal controls (Al-Memar et al., 2019). Our results showed that compared with healthy women, non-pregnant patients with RSA have reduced vaginal microbiome richness and unchanged vaginal flora diversity. In contrast, Zhang et al. showed that there was no significant difference in flora richness and diversity compared with the control group (Zhang et al., 2019). This could be attributed to the significant difference in the number of patients in the two articles, and the fact that their study focused on 10 patients with unexplained miscarriage. Flora diversity was decreased when patients received aspirin combined with metformin treatment or other medications including immunomodulators, low-dose steroids, and/or Chinese medicine, except metformin.

At the genus level, predominant microbiota, including Lactobacillus, Gardnerella, Prevotella, Streptococcus, Atopobium, Bifidobacterium, Escherichia/Shigella, and Pseudomonas were not significantly different between the no-medication patient and control groups. However, Zhang et al. reported that Atopobium, Prevotella, and Streptococcus were more abundant in 10 patients and Lactobacillus and Gardnerella were overrepresented in 10 controls (Zhang et al., 2019). Kuon et al. indicated that patients with RSA had more colonization by Gardnerella vaginalis and Gram-negative anaerobes including Prevotella, Bacteroides, and Veillonella species, based on culture-dependent methods (Kuon et al., 2017). We speculate that the reason for the inconsistent conclusions of these two studies is because of differences in sample size and research methods.

According to the prevalence of Lactobacillus and other bacteria in the vaginal flora, CSTs are divided into five types. Among these five CSTs, four of them are predominated by species of Lactobacillus (CST I, L. crispatus; CST II, L. gasseri; CST III, L. iners; and CST V, L. jensenii). CST IV is composed of diverse bacteria dominated by anaerobic bacteria (Ravel et al., 2011). According to the analysis of the vaginal microbiome characterization of all subjects, it was found that there were statistical differences in the classification of the vaginal flora between the control group and the case group including no-medication and drug-treatment groups. No significant difference was found between no-medication patients and drug-treatment patients in the frequencies of CSTs. This may be caused by a different vaginal microbiota structure in patients than in controls. It is in line with the conclusion that RSA is associated with the changes in vaginal microbiome constituents as analyzed by PCoA. The prevalence of each CST within the three groups was compared and no statistical differences was found except in CST I. The prevalence of CST I was higher in the patient group than in the control group. The protective role of Lactobacillus species, including L. crispatus, was demonstrated in many previous studies. (Aroutcheva et al., 2001; Hočevar et al., 2019; Caselli et al., 2020). This difference may be attributed to the small sample size of control subjects and the effect of drug therapy on patients’ vaginal microbiome profile. Prevalence of CST IV in the first trimester stages was correlated with gestational age at delivery (DiGiulio et al., 2015). By contrast, Elovitz et al. suggested that the prevalence of Lactobacillus spp. was not associated with spontaneous preterm birth after conducting a study involving a prospective cohort of 2,000 single pregnancy women (Elovitz et al., 2019). In future, a larger sample size is needed to analyze the characteristics of vaginal CSTs between patients with RSA and healthy controls.

The abundances of 15 genera were significantly different between the no-medication and control groups (p < 0.05). However, in untreated patients, Megasphaera abundance was significantly increased. The abundance of Sneathia sanguinegens in untreated patients was also higher than in controls, and this difference showed a trend toward statistical significance (p = 0.051). Therefore, Sneathia sanguinegens, Megasphaera, and the reduced 14 genera might be related to RSA. Megasphaera was also previously shown to have a positive association with PTB in studies by of Hočevar et al. and Nelson et al. (Nelson et al., 2014; Hočevar et al., 2019). Sneathia spp., which can adhere to cervical epithelial cells and have a high cytotoxic potential, were previously associated with serious pregnancy complications including spontaneous abortions, preterm labor, and preeclampsia resulting from invasion into the uterine cavity and amniotic sac (Harwich et al., 2012; Seo et al., 2017; Fettweis et al., 2019). Gentile and colleagues recently discovered a novel virulence-related CptA produced by S. amnii that can permeabilize chorionic trophoblast cells and lyse human red blood cells (Gentile et al., 2020). Our findings showed that Sneathia sanguinegens might have a specific role in RSA. Of the 14 reduced bacterial genera, 5 (Corynebacterium_1, Rhodococcus, Sphingomonas, Burkholderia-Caballeronia-Paraburkholderia, and Pseudomonas) contributed the most to the difference between controls and cases on in-depth analysis by LDA, meaning that they were most closely related to RSA. All five genera are composed of aerobic bacteria, and further studies are needed to explore the underlying mechanism of this phenomenon.

Metformin is an effective antidiabetic drug that is also used to treat endocrine diseases caused by PCOS (e.g., polycystic ovary-related recurrent miscarriage) and normalize endocrine, metabolic, and reproductive functions (Rai and Regan, 2006; Practice Committee of the American Society for Reproductive Medicine, 2012; Khalife et al., 2019; Wu et al., 2020). Metformin has attracted much attention in recent years because of its anticancer effects on several human solid tumors (Yu et al., 2019). Some scholars speculated that the benefits of metformin in cancer prevention and treatment might be mediated by the intestinal flora (Wu et al., 2020). Given its antiplatelet and anti-inflammatory properties, aspirin is often used to prevent and treat cardiovascular diseases, as well as for the treatment of unexplained recurrent miscarriage and autoimmune-related recurrent miscarriage. Aspirin has also been proven to have antitumor effects, especially in colorectal cancer; this might be due to the relatively favorable migration of the intestinal microbiome caused by aspirin (Wu et al., 2020). In our study, the vaginal microbiota composition of patients significantly recovered when they received metformin alone. After administration of metformin plus aspirin, patients’ vaginal microbial composition was partially restored. Specifically, at the genus level, the difference between the cases and controls was reduced from the original difference of 15 genera to 6 genera. It is worth noting that when patients were treated with metformin and aspirin, the abundance of vaginal Lactobacillus spp. increased significantly. To our knowledge, this has not been reported in past studies and provides new ideas for the future treatment of diseases caused by the decrease or absence of Lactobacillus spp. Ideally, further in-depth research studies should be performed on larger cohorts of subjects with RSA to verify our results and explore related mechanisms.

Patients experiencing RSA presented a less rich vaginal microbiome with decreased abundance of Pseudomonas, Burkholderia-Caballeronia-Paraburkholderia, Corynebacterium_1, Rhodococcus, and Sphingomonas, along with increased abundance of Megasphaera and Sneathia sanguinegens. Drug treatment affects vaginal microbiota composition. Metformin alone or in combination with aspirin might normalize microbiota composition. Metformin plus aspirin might significantly increase the abundance of vaginal Lactobacillus spp.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: NCBI SRA; PRJNA683172.

The studies involving human participants were reviewed and approved by the Ethics Committee of the Obstetrics and Gynecology Hospital of Fudan University. The patients/participants provided their written informed consent to participate in this study.

CY, MD, and FZ conceived and designed the study. JG and ZW contributed to obtaining ethical approval. YC was responsible for financial management. Participant recruitment and sample collection were conducted by MD and FZ. Experiments and data collection were performed by ZW, CL, YC, NH, and FZ. Data were analyzed by JG, MW, MD, FZ, CY, and LC. Figures and tables were generated by FZ and LC. The manuscript was written by FZ and reviewed by CY and MD. All authors contributed to the article and approved the submitted version.

Financial support was provided by the National Nature Science Foundation of China (grant number: 81873970).

NH and LC were employed by the company Sinotech Genome Technology Co., Ltd.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.