Ureaplasma and Prevotella colonization with Lactobacillus abundance during pregnancy facilitates term birth

Ureaplasma and Prevotella infections are well-known bacteria associated with preterm birth. However, with the development of metagenome sequencing techniques, it has been found that not all Ureaplasma and Prevotella colonizations cause preterm birth. The purpose of this study was to determine the association between Ureaplasma and Prevotella colonization with the induction of preterm birth even in the presence of Lactobacillus. In this matched case–control study, a total of 203 pregnant Korean women were selected and their cervicovaginal fluid samples were collected during mid-pregnancy. The microbiome profiles of the cervicovaginal fluid were analyzed using 16S rRNA gene amplification. Sequencing data were processed using QIIME1.9.1. Statistical analyses were performed using R software, and microbiome analysis was performed using the MicrobiomeAnalyst and Calypso software. A positive correlation between Ureaplasma and other genera was highly related to preterm birth, but interestingly, there was a negative correlation with Lactobacillus and term birth, with the same pattern observed with Prevotella. Ureaplasma and Prevotella colonization with Lactobacillus abundance during pregnancy facilitates term birth, although Ureaplasma and Prevotella are associated with preterm birth. Balanced colonization between Lactobacillus and Ureaplasma and Prevotella is important to prevent preterm birth.

www.nature.com/scientificreports/ the microbial community, particularly in the vaginal microbiome, is related to PTB 18 . Among various bacteria, several studies have reported on the relationship between Ureaplasma and Prevotella and PTB. In addition, the effect of the microbiome of pregnant women on the prognosis during pregnancy and PTB has been studied [19][20][21][22][23] . However, according to previous studies, the vaginal microbiome shows differences according to race and place of residence 18,24 , and a study has also reported that there is no relationship between PTB and a specific microbiome in some races 17 . Ureaplasma spp. is known to be associated with PTB, however, vaginal colonization does not universally result in induction of ascending infection and pregnancy complications, even for animal models 25,26 .
Therefore, in this study, we aimed to evaluate candidates that can be a predictor of PTB among the microbiome of CVF and to determine the association between Ureaplasma and Prevotella colonization with PTB and whether it induces PTB even in the presence of Lactobacillus.

Results
Clinical characteristics. A total of 203 women participated in this age-matched case-control study, excluding nine subjects: 102 women in the PTB group and 101 women in the TB group (Fig. 1). There were no significant differences between the characteristics of the PTB and TB groups, except for the history of sPTB, white blood cell (WBC) count, cervical length (CL), gestational age at sampling (GAS), gestational age at birth (GAB), birth weight, appearance, pulse, grimace, activity, respiration (APGAR) score, and neonatal intensive care unit (NICU) admission rate (P < 0.001, Table 1).
Association between bacteria and preterm birth. Differences in microbial diversity between PTB and TB groups. Comparing the Krona chart with PTB and TB, PTB showed that Bacteroidetes, Proteobacteria, and Mollicutes were diversely distributed from the phylum level in addition to Firmicutes and Actinobacteria ( Supplementary Fig. 1). In the TB group, 88% consisted of Lactobacillus, and Gardnerella, Bifidobacterium, and Atopobium accounted for a small percentage. In the PTB group, Lactobacillus accounted for 69% and the distribution of other species such as Staphylococcus, Bacteroides, Prevotella, Ureaplasma, Sphingomonas, and Escherichia was shown ( Supplementary Fig. 1). When comparing Shannon's α-diversity index between the PTB and TB groups, the medians (interquartile ranges) were 2.20 (1.27-2.12) and 1.99 (1.60-3.03), respectively. The median was significantly higher in the PTB group (P < 0.001, Fig. 2a,b). In the β-diversity analysis, most of the microbial communities were similar, but a distance was observed between specific microorganisms (Fig. 2c,d). There was no significant difference in the CST type between the PTB and TB groups (P = 0.156, Table 1 and Supplementary  Fig. 2).
Pattern search using correlation analysis of the vaginal microbiome. When the examination was performed at the representative genus level to identify significant differences in pattern using correlation analysis, Staphylococcus showed a positive correlation of 0.3 with Gemmiger and Ruminococcus and showed a negative correlation of 0.3 with Lactobacillus, and in this case, the association with TB was high (Fig. 4a). Bacteroides showed a positive correlation of ≥ 0.8 with 43 genera and a negative correlation of 0.6 with Lactobacillus (Fig. 4b). Sphingomonas showed a positive correlation of ≥ 0.8 with 46 genera and a negative correlation of 0.6 with Lactobacillus (Fig. 4c). Ureaplasma had a correlation of 0.3 with Escherichia, and Prevotella had a positive correlation of 0.5 with Fusobacterium (Fig. 4d,e). When Ureaplasma showed a negative correlation with Lactobacillus, the subjects associated with TB were marked in red on the right box line (Fig. 4d). Similarly, in the case of Prevotella, negative correlations with Lactobacillus were highly correlated with TB. In addition, positive correlations with Ruminococcus were also associated with TB (Fig. 4e). Escherichia showed a positive correlation with Gemmiger and Ruminococcus and a negative correlation with Lactobacillus, and in this case, the association with TB was high (Fig. 4f).

Discussion
This is the first study to conduct correlation analysis with bacteria of the vaginal microbiome, and we suggest that the relevance to the PTB should be evaluated by the community, not just the presence of specific bacteria. Although Ureaplasma and Prevotella are related to PTB, if these bacteria coexist with Lactobacillus, pregnancy outcome will be followed by TB.
These results were obtained by comparing the CVF microbiome of the PTB and TB groups using 16s rRNA metagenome sequencing. Through this method, we were able to understand the entire ecosystem of the vagina and identified 20 candidate bacteria related to PTB and three bacteria related to TB among all colonizing microbiomes. The Lactobacillus spp. and Gardnerella vaginalis were dominant in the TB group, whereas Bacterioides spp., Ureaplasma spp., Staphylococcus spp., Escherichia fergusonii, Prevotella spp., and Sphingomonas spp. etc. www.nature.com/scientificreports/ were increased in the PTB group. However, although bacteria related to PTB were present, there were cases where delivery became TB, which coexisted with Lactobacillus, Ruminococcus, and Gemmiger.

Scientific Reports
Our study results were similar to those of previous studies that used 16s rRNA metagenome sequencing. When comparing between the PTB with TB groups, Lactobacillus spp., including L. crispartus, L. jensenii, and L. gasseri, were associated with TB 18 . In pregnant women, the vaginal microbiome is dominated by Lactobacillus from the second trimester, reducing bacterial diversity 27 . Vaginal Lactobacillus spp. is known to protect the vaginal environment from harmful bacteria as beneficial bacteria, which is thought to be because lactic acid or hydrogen peroxide secreted by Lactobacillus plays a bacteriostatic role. Therefore, in the group with a lower dominance of Lactobacillus, the risk of PTB increases with the result of increasing pathogens 15,19,28,29 . www.nature.com/scientificreports/ In contrast to the findings of previous studies 30, 31 , we found that G. vaginalis was associated with TB. However, controversy exists because there were positive findings in healthy women 31 . Thus, we realized that G. vaginalis needed to be identified up to the strain level 32 .
Among the increased bacteria in PTB cases, B. fragilis, U. parvum, and P. bivia have been associated with PTB in previous studies 23,33,34 . Staphylococcus spp., Escherichia spp., and Sphingomonas spp. are commonly known as opportunistic bacteria [35][36][37] . In our study, B. thetaiotaomicron and B. fragilis were found to increase in PTB cases. Prevotella spp. is one of the genera known to be associated with PTB and BV 33 . The Mollicutes class, which includes Ureaplasma and Mycoplasma, has also been studied for its potential association with PTB 22,34 , and this is also associated with the cause of a shorter cervical length 38 . However, there was controversy about colonization in healthy women, and this study revealed that colonization itself was not associated with all infection and pregnancy complications 21 .
Therefore, our research is meaningful because, in the analysis of the vaginal microbiome, we revealed the importance of interpretation through an understanding of the microbiome relationship. Ureaplasma and Prevotella, previously considered pathogenic, did not cause complications when coexisting with Lactobacillus in our study. Furthermore, while a previous study showed cervical epithelial damage increased PTB, the protective role of Lactobacillus may explain why experimental Ureaplasma infection only resulted in a maximum of 28% induced PTB 26 . This was consistent with the bacterial risk score model, a study that analyzed the microbiome using machine learning, developed by this research team to predict PTB 21 . In this study, Ruminococcus and Gemmiger were suggested to play a protective role against pathogenic bacteria. They have been studied as part of a healthy gut microbiome, so we suggest the possibility of an association with the healthy intestinal microbiome and TB 39,40 .
The results of this study using correlation analysis suggest that Lactobacillus is thought to show a protective effect against Ureaplasma and Prevotella as well as most pathogenic bacteria. However, in this study, only the results using bioinformatics were shown, and no experiments were conducted to reveal the causal relationship. Therefore, for future research, in order to confirm the biological mechanism, it may be necessary to study the analysis of proteomics and metabolomics in addition to genomics. Furthermore, studies such as changes in Table 1. Clinical characteristics of the study subjects. Categorical variables were expressed as frequencies (percentage) and analyzed using the chi-square test and Fisher's exact test. Continuous variables were expressed as the mean ± standard deviation (SD) or median (interquartile range) and were compared using the t-test or Mann-Whitney U test. BMI body mass index, PTB preterm birth, WBC white blood cell, GAS gestational age at sampling, CST community-state type, ND normal delivery, CS cesarean section, GAB gestational age at birth, APGAR appearance, pulse, grimace, activity, respiration, NICU neonatal intensive care unit. *Statistical significance was defined as P < 0.05.

Characteristics
Preterm birth (n = 102) Term birth (n = 101) P-value www.nature.com/scientificreports/ cytokines or immune activation to determine how this microbiome acts with the host should be conducted together.
To our knowledge, this is the first study to examine the microbiome relationship using pattern search through correlation analysis. Through this study, it was suggested that the relationship through the quantitative analysis of the microbiome, not simply the presence or absence of pathogenic bacteria, would be more important in predicting PTB. Furthermore, after target selection through metagenome analysis, multiplex quantitative PCR (qPCR), which can save relatively cost and time, can be used in clinical practice for predicting PTB 21 . This study has strengths as a large-scale, multicenter study targeting pregnant Korean women. The limitations of this study are that it was not analyzed, including strain level measurements for U. parvum and G. vaginalis, despite recent studies showing that the pathogenicity of Ureaplasma and Gardnerella differs depending on the serovar or clade levels. In addition, this study may be limited in that it did not analyze the microbiome in amniotic fluid or placenta to confirm intra-amniotic infection. As a limitation of the method itself, 16s rRNA metagenome sequencing can analyze all colonized microbiome of the vagina with high sensitivity, but it is difficult to identify the actual activity and pathogenicity of the microbiome. However, despite these limitations, candidates related to PTB were discovered through various bioinformatics analyses to understand their relationship.
Ureaplasma and Prevotella colonization with Lactobacillus abundance during pregnancy facilitates TB, although Ureaplasma and Prevotella are associated with PTB. Balanced colonization between Lactobacillus and Ureaplasma and Prevotella is important for preventing PTB. The experiments were conducted in accordance with the approved guidelines, and informed consent was obtained from all the subjects. Subjects included women with a singleton pregnancy and at a gestational age between 15 and 36 weeks, asymptomatic pregnant women who visited the outpatient clinic department regularly, and hospitalized pregnant women with symptoms of PTL or preterm premature rupture of membranes. The CVF sample was collected from the posterior vaginal fornix using sterile cotton swab through the speculum exam before any vaginal examination or clinical treatment, such as antibiotics, steroids, progesterone, and tocolytics.
For all study subjects, baseline demographic data and health-related characteristics including age, pre-pregnancy body mass index, education level, and maternal PTB history were collected. A routine blood test was performed to examine the WBC CL at the time of CVF collection. After delivery, the outcomes of the pregnancy were evaluated, including delivery mode; GAB; neonatal birth weight; APGAR score; and NICU admission. Among the enrolled subjects, those diagnosed with gestational diabetes mellitus, preeclampsia, and insufficient medical records were excluded. The PTB group was categorized as subjects who delivered at < 37 weeks of gestation, whereas the TB group was characterized by subjects who delivered after 37 weeks of gestation.
Metagenome analysis using 16s rRNA gene sequencing. The collected samples were subjected to bacterial DNA extraction for microbiome analysis using the NucleoSpin Tissue Kit (Macherey-Nagel, Düren, Germany) following the manufacturer's instructions. A 16S rRNA sequencing library was constructed according to the 16S metagenomic sequencing library preparation protocol targeting the V3 and V4 hypervariable regions of the 16S rRNA gene [41][42][43] . The KAPA HiFi HotStart ReadyMix (KAPA Biosystems, Wilmington, USA) and Agencourt AMPure XP system (Beckman Coulter Genomics, Brea, USA) were used for PCR and purification of the PCR product, respectively. The initial PCR was performed with 12 ng of the template DNA using region-specific primers that were compatible with the Illumina index and sequencing adapters (Supplementary Table 1). After magnetic bead-based purification of PCR products, a second PCR was performed using primers from the Nextera XT Index Kit (Illumina, San Diego, USA) with a limited cycle. Subsequently, purified PCR products were visualized using gel electrophoresis and quantified with a DropSense96 (Trinean, Gentbrugge, Belgium). The pooled samples were run on the Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA) for quality analysis prior to sequencing. Libraries were quantified by qPCR using the CFX96 Real-Time System. www.nature.com/scientificreports/ After normalization, sequencing of the prepared library was conducted using the MiSeq system (Illumina, San Diego, USA) with 300 bp paired-end reads. Sequencing data were processed using QIIME1.9.1 to assemble paired-end reads into tags according to their overlapping relationships. In the pre-processing step, the primer was removed, demultiplexed, and filtered for quality (Phred ≥ 20). USEARCH7 was used to perform denoising and chimera detection/filtering in the operational taxonomic unit (OTU) group. Then, the Silva132 and NCBI databases were used to determine the OTUs with 97% similarity using UCLUST and the close-reference analysis method and to determine the OTU identifiers. Comparative OTU assignment was performed with the database in terms of phylum, class, order, family, genus, and species separately using RDP classifiers. Using QIIME, the α-diversity was analyzed with the Shannon index to understand the local population of the microbiome and the β-diversity was analyzed for estimating the correlation among other factors and microbes by Bray-Curtis.
Basic statistical analyses, such as t-test, Mann-Whitney U test, Chi-square test, Fisher's exact test, heat map, Krona chart, and the linear discriminant analysis (LDA) effect size (LEfSe) were performed. Multivariate analyses, such as principal coordinate analysis and non-metric multidimensional scaling, were performed. The adjusted P value was calculated by adjusting the false-positive rate using the false discovery rate. Correlations between the taxa and sample groups were analyzed using the Pearson correlation coefficient r as the distance measure. Statistical analyses were performed using R software (version 3.6.2), and microbiome analysis was performed using the MicrobiomeAnalyst (https:// www. micro biome analy st. ca/) and the Calypso (http:// cgeno me. net/ Calyp so/) software.