Vaginal microbiota alterations in women with recurrent implantation failure and the associated metabolome

In vitro fertilization-embryo transfer (IVF-ET) is now widely applied in treating infertility. As the number of IVF cycles continues to increase, recurrent implantation failure (RIF) has become a big challenge. The cause of RIF is very complex and remains largely unrevealed, especially for those without any pathological features. It has been proved that vaginal microbiota is associated with many female reproductive diseases, such as pregnancy-related diseases, sexually transmitted diseases, tubal factor infertility, and first trimester miscarriage after in vitro fertilization (IVF) and so on. Hence, vaginal microbiota and its metabolome may also relate to RIF. In this study, we characterized the vaginal microbiota and metabolome of patients with unexplained RIF, while patients who achieved clinical pregnancy in the first IVF cycle were used as controls. Results Based on 16S rDNA sequencing of the vaginal microbiota, the RIF group presented higher microbial α-diversity than the control group (0.80±0.50 vs 0.50±0.39, P-value=0.016) and harbored more non-Lactobacillus microorganisms, including 25 significantly increased genera of both aerobic and anaerobic bacteria. The metabolomic profile showed that the relative abundances of 37 metabolites among 2,507 metabolites were significantly different between the two groups. Among them, 2',3-cyclic UMP and phosphoinositide were the top two metabolites significantly upregulated in the RIF group, while glycerophospholipids and benzopyran were important metabolites that were significantly downregulated. Lysobisphosphatidic acid (LPA) and prostaglandin (PG) metabolized from glycerophospholipids are key factors affecting implantation and decidualization. Benzopyran, as a selective estrogen receptor modulator (SERM), may affect the outcome of pregnancy. All of the metabolome outcomes may result in or from the differential microbiota the RIF patients. Conclusions In conclusion, significant differences were presented in the vaginal microbiota and metabolome between RIF patients and women who became pregnant in the first IVF cycle, which are related to embryo implantation. This study not only deeply investigates the relationship between RIF and the vaginal microbial community and its metabolites but also provides a profound understanding of the pathogenesis of RIF.


Introduction
Infertility is defined as the inability to become pregnant within a couple who have normal intercourse without contraception for one year [1,2]. The incidence of infertility is 8%-12% in reproductive-aged couples, which has become a definite global public health issue [3]. In vitro fertilization-embryo transfer (IVF-ET) is currently widely used in the treatment of infertility. In 1978, the first IVF baby was born in the UK. Over the intervening years, the success rate has steadily increased, with a cumulative live birth rate of over 52% after 3 cycles of treatment [4].
Nevertheless, IVF treatment failure still occurs frequently. Among all repeated IVF failure types, recurrent implantation failure (RIF) is a special subgroup. RIF refers to the women under the age of 40 years who received at least four good-quality embryos in a minimum of three fresh or frozen cycles that still failed to achieve a clinical pregnancy [5]. Failed IVF usually occurs in elderly patients and those with a low ovarian reserve or low ovarian response for whom there is a need to screen for high-quality embryos by repeated IVF [5]. As the number of IVF cycles continues to increase, RIF has become a new big challenge, as its occurrence rate among infertility patients has continued to increase in recent years [6]. RIF also carries a heavy financial burden and causes a deep impact on the patient's body and mind.
The investigation and management of RIF usually focuses on the quality of the embryo and endometrial receptivity. However, the importance of maternal systematical diseases, such as thyroid, thrombophilia and immunological disorders, has also been recognized recently [7][8][9]. Unfortunately, many RIF patients still cannot become pregnant after the overwhelming diagnosis and treatment.
Moreover, a recent paper reported that patients who were not of advanced age still have a very low live-birth rate (36.6%), even when euploid blastocyst transfer cycles are selected by comprehensive chromosome screening (CCS) and after confirming the window of implantation (WOI) by endometrial receptive array (ERA) [10]. Even worse, another group of unexplained RIF patients has no pathological features, whose treatment method is more intractable. Hence, the nosogenesis of RIF is very complex and requires a multidimensional explanation.
The human commensal microbiome, referred to as "the other human genome", coexisted and evolved with the human genome to help maintain human health.
Microbial dysbiosis and invasion of pathogens can lead to disease and even threaten human life. It has been proven that metabolic syndrome, diabetes [11,12], obesity [13,14], alcoholic liver disease [15], cirrhosis [16] A previous systematic review reported that an abnormal vaginal microbiota is associated with tubal factor infertility and early spontaneous abortion in IVF patients. However, that article also pointed out that the quality of evidence was very low and needs further research [30]. In addition, since a microbiota continuum exists along the female reproductive tract and is associated with pregnancy-related diseases, sexually transmitted diseases, infertility-related metabolic diseases, gynecological tumors and so on [28, 31-33], we assume that the microbiota and its metabolites might associate with every step of IVF, including gamete formation, implantation and delivery [34,35]. Moreover, the correlation between the vaginal microbiome and RIF remains totally unclear. To explore the association between vaginal microbial function and IVF/RIF, metabolomic analysis should be used as an advanced tool. Metabolomics is untargeted identification and quantification of all low molecular weight end-products of cellular biological processes [36]. The levels of metabolites ultimately reflect the integrated response of a biological system and directly influence the host.
Here, we characterized the vaginal microbiota and metabolome of patients with unexplained RIF. Additionally, we used individuals who achieved a successful pregnancy in the first IVF cycle as controls to explore the relationship of the microbial community and metabolites with RIF.

The diversity and composition of the vaginal microbiota
A total of 67 samples were analyzed by 16S rDNA sequencing to investigate the vaginal microbiota, including 27 samples from the RIF group and 40 from the control group. The clinical information of the subjects is shown in Table 1a, which indicates that there was no significant difference between the RIF group and the control group (P-value > 0.05). In total, 2,824,185 reads were obtained from these 67 samples, and on average 42,152±10,415 reads per sample and 424±7 bp per read were achieved. After clustering, the rarefaction curve of the OTU number was almost a straight horizontal line, which demonstrated that the samples were sequenced with enough depth in this study (Supplementary Figure 1).
A total of 804 OTUs were obtained in the two groups. The RIF group contained 730, and the control group contained 429, among which 355 were shared between the two groups. The number of OTUs in the RIF group was much larger than that in the control group, indicating that the microbial composition of the RIF group was richer.
The α-diversity of the microbiota was calculated by the Shannon-Wiener index NLD individuals were 72.723% and 34.723%, respectively, which presented a significant difference (P-value = 0.006).

Comparative metabolomics in the vaginal environment
Twenty-five samples were subjected to metabolic analysis, including 10 RIF group samples and 15 control samples. As well as the sequenced samples, the clinical information of the subjects showed no statistical differences between the two groups (P-value>0.05) (Table 1b) We utilized OPLS-DA to observe metabolites that were differentially abundant between these two groups and classify them into group-related and groupindependent metabolites. The results elaborated that the RIF group samples were clustered together and distinct from the grouped control samples (Figure 4a). Of 2,507 annotated metabolites, 37 metabolites were found to have significant differences between the two groups, with variable importance for the projection revealed that 16 metabolites were significantly upregulated in the RIF group, among which 2',3-cyclic UMP and phosphoinositide were the top two metabolites; 21 metabolites were significantly downregulated in the RIF group, and 5 substances were downregulated by 4-fold or more: benzopyran, fatty alcohol, pyrimidine nucleoside, glycerophospholipid and naphthopyran ( Figure 4c). The network of the metabolites that significantly correlated with each other was also investigated under the proposing of Spearman's correlation coefficients (R ≥ 0.6 or R ≤ -0.6) and P-values (P 0.05).

Correlation between the microbiota and metabolome
In the RIF group, the correlation between the top 17 abundant genera and 37 differentially abundant metabolites was calculated ( Figure 5). A significant positive correlation of benzopyran and glycerophospholipids with Lactobacillus ( P-value<0.05, R=0.714) was discovered. Coincidently, the relative abundance of Lactobacillus was significantly reduced in the RIF group compared to that in the control group. This result indicated that benzopyran and glycerophospholipids might be functioned metabolites during IVF driven by Lactobacillus or related bacteria in the vaginal microbiota. Interestingly, L. iners was also positively correlated with benzopyran in the RIF group (R=0.738). For the control group, consistent changes in benzopyrans with the relative abundance of L. iners were identified (R=0.527).

Discussion
This study is an attempt to explain the influence of the vaginal microbiota and its metabolites on RIF patients while using the single-procedure successful IVF cases as controls. The vaginal microbiota of the RIF group presented a higher diversity and harbored more potential pathogens than that of the control group. Moreover, the abundance of Lactobacillus was also significantly decreased in the RIF group. The metabolite profile delineated that benzopyran and glycerophospholipids were the significantly differentially abundant metabolites between the two groups and may result in or from changes in the microbiota composition.
An increase in vaginal microbial diversity in RIF patients was observed in this study, which is consistent with the study of Richard et al. in 2012. That report was the first study to suggest that the species diversity index of the vaginal microbiota distinguished women who had live births from those who did not [35]. Another study also showed that the clinical pregnancy rate and live-birth rate were significantly reduced (P-value = 0.01) in a cohort with a vaginal Shannon-Wiener index higher than 0.93 [37]. To date, few vaginal microbiota studies have investigated in RIF patients, but association studies on an abnormal vaginal microbiota with IVF have been discovered. A review concluded that culture-dependent studies indicated that an abnormal vaginal microbiota was not linked to the outcome of IVF. However, this conclusion was overturned by high-throughput sequencing results, which elaborated that microbiota disorders negatively influenced the IVF outcome [38]. Another review demonstrated that the incidence of bacterial vaginosis (BV) was associated with early abortion of IVF pregnancies and infertility due to tubal factors and was not associated with pregnancy rate and live-birth rate [30]. However, the association between the vaginal microbiota and preterm birth has been well  [40]. Similarly, the abundance of Lactobacillus was negatively related to the abundance of potential BV/AV pathogens in this study, including Gardnerella, Prevotella, Atopobium, and Streptococcus, which indicated the protective function of Lactobacillus. Our results also indicated that Lactobacillus was losing its dominant position in the RIF group which may be associated with the occurrence of repeated pregnancy failures. What's more, the pregnancy rate of NLD subjects was significantly decreased compared with that of LD subjects. Although the relative abundance of Lactobacillus was significantly lower in the RIF group than in the control group, the LDA value was still too low to be identified as a biomarker to predict the occurrence of RIF. However, this conclusion may also be manipulated by the research sample size of the study.

Conclusion
In this study, the vaginal microbiota and metabolome of patients with unexplained RIF (RIF group) and women who became pregnant in the first IVF cycle (control group) were investigated. In general, significant differences were discovered in the vaginal microbiota and metabolome between the two groups, which are related to embryo implantation. The RIF group presented lower abundance of Lactobacillus and higher abundance of other aerobic or anaerobic microorganisms than the control group and showed a significantly higher α-diversity. The metabolome profile indicated that LPA and PG metabolized from glycerophospholipids, a metabolite significantly downregulated in the RIF group, are key factors affecting implantation and decidualization, while benzopyran, another metabolite significantly downregulated in the RIF group that is a SERM and pathogen inhibitor, may contribute to the outcome of pregnancy.

Ethics approval and consent to participate
The study was approved by the Ethics Committee of Peking University Shenzhen Hospital at May 10, 2018. The study is sponsored by the Peking University Shenzhen Hospital [2018]No.017 and was conducted as a single-center study without any investigational product. All the enrolled subjects provided written consent and gave permission to access medical records to obtain their related clinal information and vaginal specimens.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author (Shangrong Fan, fanshangrong@163.com) on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
The study was funded by the Shenzhen Municipal Government (JCYJ20180228162311024 and SZSM201612065), Shenzhen, China.

Authors' contributions
This paper has 6 authors in total. SF is the correspondence author. MF and XZ are the co-first -authors who contribute equally to the study. SF and MF contributed to the conception and design of the study. MF and WQ contributed to the achieving of ethical approval. MF, WQ and YL conducted the collection of samples. MF, SF and XZ analyzed the data. XZ and MF drafted and SF revised the manuscript. All authors read and approved the final manuscript. Table   Table 1a. Clinical characteristics of the participants in the two groups whose samples were submitted for 16S rDNA sequencing of the vaginal microbiota.  Supplementary Table 1. Genera in the vaginal microbiota that were significantly differentially abundant between the RIF and control groups.
Supplementary Table 2. Relative abundance of Lactobacillus species discovered in the samples that were significantly different between the RIF and control groups.
Supplementary Table 3. Quantities of 37 metabolites that were found to have significant differences between the two groups via the determination of variable importance for the projection (VIP) values and P-values.

Figures
28 Figure 1 The α-diversity of the vaginal microbiota in the two groups was calculated and is shown by th   Correlation analysis between the differentially abundant metabolites and differential genera