Introduction

Prelabor rupture of membranes (PROM) refers to the spontaneous rupture of fetal membranes before the onset of labor. PROM can be divided into term PROM (TPROM, ≥ 37 gestational weeks) and preterm PROM (PPROM, < 37 gestational weeks)1. PPROM is a major feature of spontaneous preterm birth (sPTB), one of the greatest challenges facing obstetrics with complicated pathogenesis. It is generally considered a leading cause of neonatal mortality and morbidity on a global scale [2,3,4,5,6,7]. PPROM is the largest contributor to sPTB. About 40% of sPTB cases are caused by PPROM [8, 9]. Compared with preterm labor (PTL), defined as sPTB with a contact membrane before the onset of labor [10], PPROM has a stronger association with infection and/or inflammation but a significantly lower gestational age and birth weight. It often causes higher rates of chorioamnionitis, urinary tract infection, endometritis, and postpartum bacteremia [11,12,13].

The causes of PPROM are still incompletely understood while the incidence continues to rise worldwide [5]. It is believed that vaginal bacteriome homeostasis plays an important role in maternal health [14]. Bacterial vaginosis [15, 16], aerobic vaginitis [17], and pathogens colonization, such as Group B Streptococcus, Ureaplasma spp., and Chlamydia trachomatis [14, 18,19,20], have all been associated with second-trimester miscarriage, PPROM and PTB [21,22,23]. It is also found that decreased abundance of Lactobacillus can cause vaginal dysbiosis, raise the level of vaginal inflammation, and increase the risk of PPROM and sPTB [24].

The diversity of the vaginal microbiome is an important characteristic. Increased bacterial community diversity is a risk factor for sPTB and PPROM. In two studies involving Caucasians, African Americans and Asians, it was found that increased inverse Simpson index and observed features in alpha diversity of vaginal bacteriome were related to the occurrence of PPROM [25, 26]. In another study involving Caucasians and Asians, the Bray–Curtis distance [27] for vaginal microbiome beta diversity was higher in the PPROM group [27].

Ravel et al., for the first time, used hierarchical clustering methods to describe vaginal bacteriome of non-pregnant females with five distinct community state (CST) types; four of which are Lactobacillus spp. dominant, i.e., CST I (L. crispatus), CST II (L. gasseri), CST III (L. iners), and CST V (L. jensenii), and the fifth is CST IV which are characterized by a Lactobacillus spp. depleted polymicrobial community and enriched with a mixture of diverse taxa [28].

Some studies have revealed the relationship between increased PTB risk and vaginal species or CSTs, such as decreased L. crispatus/CST I [29,30,31,32,33], L. gasseri/CST II [32], and increased Mobiluncus curtisii/mulieris, Atopobium, Sneathia sanguinegens, Olsenella, and CST IV [29, 34, 35]. Meanwhile, some controversies exist on the contribution of some species, such as L. iners/CST III [36,37,38,39] and L. jensenii/CST V [31, 32, 40].

However, only a limited number of studies specifically examined the relationship between the risk of PPROM and the vaginal microbiome, especially regarding the role of specific species or CSTs. Menon et al. have shown that fetal leukocyte telomere length is reduced in PPROM compared to PTL but similar to term births [41], suggesting distinct pathophysiologic processes that differentiate these outcomes. According to You et al., in Korea, the vaginal microbiome in PPROM is significantly different from that of PTL cases [42]. Therefore, it is essential to distinguish PPROM or PTL from sPTB cases in etiological studies.

This study aimed to investigate the association between vaginal bacteriome in the early second trimester and PPROM among a cohort of Chinese pregnant females.

Methods

Study Design and Participants

The study was designed as a case cohort, with a previously described source cohort [43]. In brief, pregnant females in the early second trimester were recruited at their first prenatal visit to Anqing Municipal Hospital, Anhui Province, China, since February 22, 2018. At enrollment, a questionnaire was used to collect participants’ demographic, lifestyle, and clinical baseline data with two vaginal samples taken by skilled obstetricians. Subsequently, any changes in the data since the last visit were recorded during follow-ups. Prenatal examination information was abstracted from medical records, while pregnancy outcomes and complications were obtained from electronic medical records (n = 1298), telephone interviews (n = 85), or censored (n = 178) after the terminal event, which was delivery, abortion, or stillbirth. Up to January 22, 2020, before the COVID-19 epidemic, 1561 pregnant females were enrolled, and 178 were lost to follow-up, with an 11.4% censor proportion.

For this case-cohort study, a sub-cohort of pregnant females was selected at baseline with a ratio of 1:4 (n = 387) using a systematic sampling method. Of the 387 pregnant females, 310 with qualified vaginal bacteriome data were included in the final sub-cohort. The final sub-cohort consisted of 255 term birth without prelabor rupture of membranes (TWPROM), six PPROM, 46 TPROM, and three preterm labor. There were 235 prelabor rupture of membranes cases from the source cohort in total. Excluding those with unqualified samples (n = 18) and multiple pregnancies (n = 2), 35 and 180 pregnant females with PPROM and TPROM, respectively, were obtained (Fig. 1).

Fig. 1
figure 1

Study population. PROM, prelabor rupture of membranes; TPROM, term preterm prelabor rupture of membranes; PPROM, preterm prelabor rupture of membranes; PTL, preterm labor; TWPROM, term birth without prelabor rupture of membranes; and iPTB, iatrogenic preterm birth

Full size image

Sample Collection

At baseline, midvaginal samples were drawn by experienced obstetricians using sterile swabs. The vagina was scraped with the swab in a clockwise rotation for five spins. Paired swabs were collected for each subject. One swab was placed in a collection tube with preservation solution and stored in a low-temperature freezer at − 80 °C before genome sequencing, and the other was inspected by wet mount for clinical purposes.

16S rRNA Amplification, Sequencing, and Processing

A 600 bp region of the 16S rRNA gene (covering V3-V4) (Primer: 341F, 5′-CCTAYGGGRBGCASCAG-3′ and 806R, 5′-GGACTACNNGGGTATCTAAT-3′) for vaginal samples was amplified with the sample-specific barcodes and then sequenced on an Ion S5™ XL instrument (Novogene Co., Ltd., Beijing, China) [43].

16S amplicon sequences were analyzed using the QIIME 2 software (Quantitative Insights into Microbial Ecology 2, Core 2021.4 distribution, http://qiime2.org). Single-end reads were allotted to samples based on their specific barcodes and truncated by cutting off the barcodes and the primer sequences. The sequences were truncated to 370-bp reads determined by the quality score shown in the interactive quality plot. Subsequently, truncated reads were denoised using the DADA2 algorithm for quality control, and the representative sequences of amplicon sequence variants (ASVs) were annotated by the SILVA (version 138) database.

Similar to the previous literature [43], sequences that lacked a good resolution at the species level were queried through the National Center for Biotechnology Information (NCBI) database with Basic Local Alignment Search Tool (BLAST) + command-line applications; species items with more than one matching value were left blank, except for L. paragasseri/gasseri, which were highly identical sister taxa of the vaginal bacterial community [44].

Identification of Candidate Vaginal Bacterial Traits

The below steps were followed to identify candidate vaginal bacterial traits (VBTs). First, ASVs were collapsed to the same level and calculated based on each taxon’s relative abundance (RAB). The discriminative taxa among PPROM, TPROM, and TWPROM groups were performed on the most abundant genus or species level, which met the following criteria: (i) ≥ 5% of reads for at least one individual and (ii) ≥ 15% of pregnant females with non-zero data [45]. The taxonomic correlations were estimated using the SparCC algorithm in FastSpar 0.0.10 [46]. Higher-level taxa with a high correlation (Spearman r > 0.985) with their corresponding lower-level taxa were excluded to reduce the redundancy of phenotypic information.

The 53 taxa generated from the above processing were transformed to binary traits due to either the presence/absence (P/A) or dominance/non-dominance (D/ND) of a retained taxon given its distribution across the study samples. Specifically, the taxon with zero counts in more than 5% of the study samples was transformed to the P/A trait, and those with a dominant threshold level (90%) of RAB in more than 5% of the study samples were transformed to the D/ND trait [43].

The alpha diversity metrics, including (i) ASV richness, (ii) the Shannon index, (iii) Faith’s phylogenetic diversity, and (iv) Pielou’s evenness, and the beta diversity metrics, including (i) Bray–Curtis dissimilarity, (ii) Jaccard distance, (iii) weighted UniFrac, and (iv) unweighted UniFrac were detected using the QIIME 2 diversity plugin. The CSTs of vaginal bacteriome were clustered based on a Bray–Curtis distance matrix by the partitioning around medoids algorithm [47]. Finally, CST I-IV was derived based on the gap statistic.

Statistical Analysis

The continuous variables in each group were illustrated with mean ± standard deviation or median (interquartile range). The categorical variables were illustrated in number and proportion [n (%)]. A χ2 test was performed for screening single VBT, including general demographic and health information, alpha diversity, CSTs, and binary taxa within two pairs of groups, PPROM vs. TWPROM and PPROM vs. TPROM, separately. The regression model considered variables statistically significant (2 test, P < 0.15) afterward.

According to the literature review [33, 43, 48,49,50] and the actual data-generating process in our study, three variables, including age, pre-pregnancy body mass index (BMI), and passive smoking, were considered potential confounders and included as adjustment variables. The missing values in passive smoking (14%) and pre-pregnancy BMI (1%) were imputed with the R mice package based on a random forest model reported previously [43, 51].

A binary unconditional logistic regression model followed for screening single VBT, including alpha diversity, CSTs, and binary taxa, to obtain the adjusted odds ratio (aOR) and their 95% confidence interval (CI). Differences in beta diversity metrics between PPROM and the control groups (180 TPROM and 255 TWPROM cases separately) were estimated using principal coordinates analysis and Adonis testing.

Furthermore, a Kaplan–Meier survival analysis on the risk of PPROM was performed for significant VBTs in the logistic model [34]. A weighted Cox model with PPROM was analyzed using Lin-Ying’s weighted Cox model in the R package (R 4.0.3) [52, 53] to obtain the adjusted hazard ratio (aHR) and their 95% CI.

The absence and non-dominance state of the binary variables were used as the reference in the models. The level of statistical significance was set at two-sided P = 0.05, and the false discovery rate (FDR) correction level was set at P = 0.10. Generally, the variables with PFDR < 0.10 in preliminary screening models were included and analyzed in the multivariable models. The taxa with P(FDR) < 0.10 in preliminary screening models or P < 0.05 in the multivariable model were further included in the subgroup analyses. The Kaplan–Meier survival analysis on the risk of PPROM with different species was also performed [34].

Results

Data available for our final analysis (n = 472) included 35, 180, and 255 pregnant females in the PPROM, TPROM, and TWPROM groups, respectively. Most (95%) of them at baseline did not have vaginitis according to white blood cell counts, Trichomonas and yeast tests. The three groups’ characteristics, including age, pre-pregnancy BMI, ethnicity, and education level, were not significantly different (Table 1).

Table 1 Baseline characteristics of study participants
Full size table

Of 357 pregnant females, 44(12.3%) were censored in the sub-cohort, all of whom were excluded in this study. Baseline characteristics were not significantly different between censored females and those with the tracked pregnancy outcome, including age, pre-pregnancy BMI, ethnicity, and education level (Table S1).

VBTs

Among the 472 pregnant females, the most abundant species of VBTs was L. iners (8,531,139 out of 19,435,766 reads). The vaginal CSTs were clustered into four groups, CST-I including two subtypes—CST-Ia (L. crispatus) and CST-Ib (a mixture of dominant L. iners and L. crispatus), CST-II (L. paragasseri/gasseri), CST-III (dominated by L. iners), and CST-IV without any dominant flora but a mixture of diverse species, including Gardnerella vaginalis, Atopobium vaginae and others (Fig. S1).

Among the 35 pregnant females with PPROM, the most abundant species of VBTs was L. iners (623,444 out of 1,263,957 reads, 49.32%), followed by L. paragasseri/gasseri (28.72%), Atopobium vaginae (3.20%), and Gardnerella vaginalis (1.83%) (Table S2). The highest proportion of CSTs of PPROM was CST-III (42.86%), followed by CST-Ia (28.57%), CST-IV (22.86%), CST-Ib (2.86%), and CST-II (2.86%). There were no significant differences in the composition of CSTs between the above three groups (Fig. S2).

Alpha diversity (ASV richness, P = 0.022; Faith’s phylogenetic diversity [PD] index, P = 0.024) and beta diversity (Jaccard index, P = 0.008; unweighted UniFrac, P = 0.010) of vaginal bacteriome metrics significantly differed between PPROM and TWPROM (Fig. 2a, c, 3b, c). Meanwhile, the P values of alpha and beta diversity (Faith’s PD index, P = 0.030; unweighted UniFrac, P = 0.023) between PPROM and TPROM were also significant (Fig. 2c, Fig. 3c). There were no significant differences in the vaginal bacteriome metrics of Shannon index, Pielou’s evenness, Bray–Curtis, and weighted UniFrac between the above three groups (Fig. 2b, d, 3a, d).

Fig. 2
figure 2

Comparisons of alpha diversities of the vaginal microbiota among the PPROM group, the TPROM group, and the TWPROM group. Boxplots display the comparisons of alpha diversity metrics of a ASV richness, b Shannon index, c Faith’s PD, and d evenness among groups. Boxes show the interquartile ranges, lines inside the boxes show medians, and circles show outliers. TPROM, term prelabor rupture of membranes; PPROM, preterm prelabor rupture of membranes; TWPROM, term birth without prelabor rupture of membranes; Faith’s PD, Faith’s phylogenetic diversity

Full size image
Fig. 3
figure 3

Comparisons of beta diversities vaginal microbiota among the PPROM group, the TPROM group, and the TWPROM group. PCoA was used to visualize compositional relationships among groups according to beta diversity metrics of a, e Bray–Curtis dissimilarity, b, f Jaccard dissimilarity, c, g unweighted UniFrac distance, and d, h weighted UniFrac distance. The percentage on axis labels is the proportion of variance explained by that axis. P values between PPROM and the other two groups are calculated using permutational multivariate analysis of variance analysis. TPROM, term prelabor rupture of membranes; PPROM, preterm prelabor rupture of membranes; TWPROM, term birth without prelabor rupture of membranes; PCoA, principal coordinates analysis

Full size image

Association of Binary Vaginal Bacterial Taxa with PPROM Among Pregnant Females

Fifty-three taxa passed the pre-screening, and their binary traits were included in the χ2 analysis. Thirteen taxa were explored using a logistic regression model (Table S3). The results showed that the presence of vaginal Megasphaera genus P/A (aOR = 2.27; 95% CI: 1.09–4.70), Faecalibacterium genus P/A (aOR = 3.29; 95% CI: 1.52–7.13), Bifidobacterium genus P/A (aOR = 3.26; 95% CI: 1.47–7.24), Xanthomonadales genus P/A (aOR = 2.76; 95% CI: 1.27–6.01), Gammaproteobacteria class P/A (aOR = 2.36; 95% CI: 1.09–5.14), or Alphaproteobacteria class P/A (aOR = 2.45; 95% CI: 1.14–5.26) was associated with an increased risk and the presence of L. mulieris species P/A (aOR = 0.35; 95% CI: 0.17–0.72) was associated with a decreased risk of PPROM compared with TWPROM. Compared with TPROM, the presence of the vaginal Mycoplasmataceae family P/A (aOR = 3.30; 95% CI: 1.47–7.42) was associated with an increased risk of PPROM. While applying the logistic model to the sample that included TPROM and TWPROM cases, the presence of L. mulieris P/A (aOR = 0.35; 95% CI: 0.17–0.72) was also associated with a decreased risk of TPROM (Table 2).

Table 2 Association of specific binary taxa of vaginal bacteriome with preterm prelabor rupture of membranes, term prelabor rupture of membranes, and term birth without prelabor rupture of membranes
Full size table

Comparison of Cumulative Hazard of PPROM with Specific Vaginal Bacteriome Among Pregnant Females

The survival curve was used to explore the difference in cumulative hazard of PPROM in pregnant females with significant vaginal bacteriome, namely, L. mulieris and Megasphaera genomosp type 1. In the weighted Cox model that considered the P/A of L. mulieris on the outcome of PPROM, the result showed that L. mulieris might have a protective effect in PPROM (aHR = 0.46; 95% CI: 0.23 to 0.93, Fig. 4a), which was similar with the result of the logistic model. In contrast, no significant relevance was found in pregnant females with the presence of Megasphaera genomosp type 1 in the weighted Cox model (Fig. 4b).

Fig. 4
figure 4

Cumulative hazard of preterm prelabor rupture of membranes (PPROM) with vaginal specific species in pregnant women. a Comparisons of cumulative hazard of PPROM between pregnant women with the presence (L. mulieris = 1) versus the absence (L. mulieris = 0) of L. mulieris were performed using survival analysis. b Comparisons of cumulative hazard of PPROM between pregnant women with the presence (Megasphaera genomosp type 1 = 1) versus the absence (Megasphaera genomosp type 1 = 0) of Megasphaera genomosp type 1 were performed using survival analysis

Full size image

Discussion

This study examined the emergence during pregnancy of an association between VBTs and increased risk of PPROM compared separately with TWPROM and TPROM. When compared with TWPROM, the PPROM risk-associated VBTs were (i) high alpha diversity and beta diversity (Fig. 2, Fig. 3), (ii) the absence of L. mulieris (Table 2, Fig. 4a), and (iii) the presence of Megasphaera genus, Faecalibacterium genus, Bifidobacterium genus, Xanthomonadales genus, Gammaproteobacteria class, or Alphaproteobacteria class (Table 2). With TPROM as the control group, the risk-associated VBTs included (i) high alpha diversity and beta diversity (Fig. 2, Fig. 3) and (ii) the presence of the Mycoplasmataceae family (Table 2). We also discovered that the presence of L. mulieris (aOR = 0.35; 95% CI: 0.17 to 0.72) was associated with a decreased risk of TPROM compared with TWPROM (Table 2).

Previous studies identified high diverse vaginal microbiome as a risk factor for subsequent PPROM [25,26,27]. Jayaprakash et al. [27] found that the vaginal microbiome in PPROM cases had a higher Bray–Curtis distance than those with full-term pregnancy (P < 0.001). Brown et al. [25, 26] also reported higher alpha diversity represented as observed species and inverse Simpson index in the PPROM group in 2018 and 2019, respectively. Four alpha diversity indicators and four beta diversity indicators were calculated in this study to describe the diversity of samples from multiple dimensions, such as richness, evenness, and evolutionary distance. The results showed that PPROM had higher alpha and beta diversity than TPROM or TWPROM, which was consistent with previous studies.

PROM directly results from the strength degradation of membranes. Spontaneous rupture occurs when membrane tension is abnormally reduced until it reaches the critical point [54]. The spontaneous rupture point is the critical point where the membrane can be maintained under the tension of the uterine contents; in other words, the membrane tension is equal to the resistance tension of the uterine contents. Some scholars believe that the strength degradation of membranes results from local inflammation of the membranes. Richadson et al. found that inflammatory factors, fetal membrane, and leukocyte telomere lengths induced by oxidative stress in the amniotic fluid of PPROM patients were shorter, with obvious aging and stress characteristics, suggesting that chronic oxidative stress and the resulting aging may make membranes more prone to rupture [55, 56]. High diversity of vaginal microbiome was significantly associated with inflammation [57, 58]. The results of this study conform to the mechanism by which inflammation leads to reduced membrane strength.

Studies using 16S gene sequencing technology in PPROM rarely observed species-level correlations. Megasphaera spp. type I and Prevotella spp. were found to exist in all samples of PPROM cases and thought to be associated with the occurrence of PPROM by Jayaprakash et al. [27]. However, the correlation was not statistically significant due to the small sample size. In addition, Brown et al. [25] found that reduced Lactobacillus spp. may cause PPROM, but the species level of Lactobacillus was not subdivided. Before the sequencing technology development, many studies used isolation culture methods to analyze the characteristics of the vaginal microbiome in PPROM cases. Of the risk taxa of PPROM in this study, Megasphaera, Bifidobacterium, and Mycoplasmataceae were associated with an increased risk of PPROM/PROM in the separation culture method studies. Besides, Gammaproteobacteria or Alphaproteobacteria presence was reported to reflect the dysbiosis or an unstable microbial community structure, tending to become colitogenic microbes that can trigger inflammatory responses, leading to the strength degradation of membranes [59, 60].

The time of spontaneous rupture of membranes can reflect the level of strength degradation of the membranes. Spontaneous rupture of membranes in cases of PPROM occurred before 37 gestation weeks, with the least volume of the fetal and amniotic fluid, requiring the least anti-tension. Therefore, the strength of membranes was lowest in PPROM cases. The membranes of TPROM also reached the point of spontaneous rupture. However, the fetal and amniotic fluid volume was larger at term, so the tension of spontaneous rupture was higher than that of PPROM. TWPROM cases did not reach the spontaneous rupture point with the highest membrane strength. As a result, strength degradation of membranes in PPROM was worst among the three groups, followed by TPROM. The result of L. mulieris and Megasphaera genus proved this hypothesis.

This study’s findings support the possible protective effect of L. mulieris, discovered in 2020, and genetically related species,[43, 61] in PPROM discovered in the logistic model and proved in the Cox model. In addition, it was found that L. mulieris also has a protective effect on TPROM, with a larger OR value, indicating that its protective effect on TPROM is weaker than that in PPROM. This Lactobacillus spp. was once hard to distinguish from its sister species L. jensenii, [61, 62] the dominant species in CST-V. A decreased risk of PROM with L. jensenii has been reported by Payne et al. [63]. In this study, we could not duplicate the findings of the associations mentioned above. It implied that the real species linked to PROM might be L. mulieris instead of L. jensenii. Some studies showed that L. mulieris can produce some antimicrobial substances (such as hydrogen peroxide, antimicrobial peptides, bacteriocin, and biosurfactants) and promotes local immunity to reduce the adhesion and colonization of pathogenic microorganisms [64,65,66], which can cause inflammation [57, 58] and often lead to the strength degradation of membranes. In other words, L. mulieris can delay the strength degradation of membranes by reducing the level of local inflammation. To the best of our knowledge, this study is the first to report the protective role of L. mulieris in PPROM.

Besides L. mulieris and Megasphaera genus, the presence of Faecalibacterium genus, Bifidobacterium genus, Xanthomonadales genus, and Gammaproteobacteria class also has a promoting effect on TPROM with a lower intension than that of PPROM. However, there is no statistical significance, which may be due to the insufficient sample size of this study. These vaginal traits need to be further explored in future studies.

Understanding the microbiome associated with PPROM is critical to creating strategies to prevent this reproductive outcome and determine when to initiate the delivery. We found the role of vaginal bacteriome, including L. mulieris, Megasphaera genus, Faecalibacterium genus, Bifidobacterium genus, Xanthomonadales genus, Gammaproteobacteria class, Alphaproteobacteria class, and Mycoplasmataceae family in PPROM. Their steady roles in the maintenance of vaginal microecology require further exploration. New questions were opened for future studies, e.g., the dynamics of different species at different gestation weeks, the exact species linked to PPROM, vaginal bacterial biology, and their interactive effect in PPROM. Besides, there are very few vaginal microbiological studies in PPROM using 16 s gene sequencing technology, especially in East Asia. This research largely fills this gap and plays a positive role in the more precise exploration and prevention of PTB.

Distinguishing PPROM from PTL is significant for researching etiology and prevention strategies for sPTB, a combination of PPROM and PTL with higher clinical variability. The difference in the pathogenesis of PPROM and PTL lies in the strength degradation of membranes and the degree of the labor onset. Menon et al. reported more microfractures in PPROM membranes than in PTL membranes [67], reflecting different etiologies for PTL and PPROM. We hypothesized that the factors which can affect both PPROM and TPROM and have a diminishing effect could be considered as definite factors that can affect the strength of membranes. In this study, L. mulieris and Megasphaera genus conformed to this rule and was an independent protective factor for reduced membrane strength. Although the difference in Faecalibacterium genus, Bifidobacterium genus, Xanthomonadales genus, and Gammaproteobacteria class between TPROM and TWPROM groups had no statistical significance, they still conformed to this rule. Using this method to conduct studies with a larger sample size, more meaningful information can be obtained about the mechanisms of membrane strength reduction.

The presence of Faecalibacterium genus, Bifidobacterium genus, Xanthomonadales genus, Gammaproteobacteria class, Alphaproteobacteria class, and Mycoplasmataceae family may be used as a screening index to prevent PPROM. Presently, clinical studies on treating bacterial vaginosis by transplanting vaginal flora have been conducted, and the therapeutic effect is good [68]. L. mulieris, as a strain that maintains membrane stability, can be considered a target species for transplanting vaginal flora in the future to prevent PROM and sPTB.

This study has substantial strengths for investigating the role of vaginal microbiome composition and PPROM among Asian females. (i) The prospective nature of this case-cohort design reduced information bias. (ii) A refined classification scheme of vaginal bacteriome taxa was implemented: the indiscernible L. paragasseri and L. gasseri were combined into one taxon; the newly discovered L. mulieris was separated from its closest kin, L. jensenii; those unclassified sequences were further searched and classified on local BLAST with NCBI database. (iii) The microbial characteristics of TPROM were analyzed, and a method to determine the independent factors of strength degradation of membranes was proposed using TPROM as a mediator. (iv) We presented a relatively complete picture of the association between VBTs and PPROM.

This study is not without limitations. (i) A detailed evaluation of vaginal swabs, such as pH value, was not done. (ii) Due to the limited sample size, we did not include PTL cases in the control group to compare differences in vaginal microbiome profiles between PPROM and PTL. (iii) Urinary tract coinfection was not evaluated.

Conclusions

This study showed that the risk of PPROM can decrease with vaginal L. mulieris but increase with high alpha or beta diversity and several vaginal bacteria in pregnant females in the early second trimester.

What’s already known about this topic?

• A high diversity of vaginal microbiome is associated with PPROM occurrence

• Lactobacillus depletion may increase the risk of PPROM

What does this study add?

• Vaginal L. mulieris in pregnant females in the early second trimester could decrease the risk of PPROM by delaying the strength degradation of membranes

• Vaginal Megasphaera genus, Faecalibacterium genus, Bifidobacterium genus, Xanthomonadales genus, Gammaproteobacteria class, Alphaproteobacteria class or Mycoplasmataceae family was associated with an increased risk of PPROM