Fostered by advances in and accessibility of DNA sequencing technologies, there have been dozens of molecular investigations into the vaginal microbiota’s potential role in sPTB over the last decade 29–55,57−60. Yet, despite this high volume of investigations, there has not been a resolute conclusion as to whether the structure of the vaginal microbiota in pregnancy is associated with or factors into the likelihood of sPTB, or if such information has clinical utility for the prediction of women at risk for delivering preterm. Indeed, one recent systematic review concluded that the available molecular-based data were limited and contradictory71, whereas according to another, only women with vaginal microbiotas characterized by a low relative abundance of Lactobacillus crispatus have a modest increased risk for delivering preterm56. Therefore, to date, no positive association has been consistently identified between a specific bacterial taxon residing in the vagina and an increased risk of sPTB.
There are several potential explanations for the inconsistency and lack of consensus in results among existing studies. First, there is high variation in sample sizes which can translate into variation in statistical power and the capacity to detect significant results among studies. Second, there is further variation in the ethnicity profiles of the cohorts under investigation among studies. Investigations have been conducted in populations of American (Asian, African, Caucasian, Hispanic)29–34, 37,42,43,50,59, Canadian (Arabic, Asian, African, Caucasian)39,40, European (Asian, African, Caucasian)35,36,38,44,51,55, Brazilian41,45, Peruvian47, Nigerian54, Indian53, Chinese57,60, Korean46,48, Japanese58, and Thai (Karen, Burman)52 women. Although populations of women worldwide have the same bacterial taxa in their vaginal microbiota61–64, 75, the relative abundances of these taxa can vary between and within populations, including by ethnicity30,42,43,51,59,61–64,76−78. It is therefore likely that associations between the structure of the vaginal microbiota and sPTB will be observed in some populations and/or ethnicities, but not in others34,36,42,43,51,59, and that this will contribute to the inconsistencies in results among studies. Third, there has been variation among the longitudinal studies in the distributions of gestational age at sampling between cases and controls 29,31,34,43,44,52,59. Given that gestational age at sampling has a pronounced effect on vaginal microbiota profiles in pregnancy65–67, 78, it is critical that the distributions of gestational age at sampling between women delivering preterm versus at term are similar. Specifically, if the vaginal microbiota profiles of women delivering preterm or at term are being compared, then samples collected after 37 weeks of gestation should not be included in the analysis, and the distribution of gestational age at sampling of samples collected prior to 37 weeks for the two groups of women should be matched. Otherwise, bacterial taxa that decrease (e.g. members of CST IV) or increase (e.g., Lactobacillus crispatus) in relative abundance with advancing gestational age may be erroneously identified as being associated with delivery preterm or at term, respectively. Fourth, there has been variation among studies in how sPTB has been defined. It has been variably defined as delivery < 37 weeks37,45,52−54, non-indicated delivery < 3341, <3429,55, <3533, or < 3730,32,39,40,43,44,47,48,50,51,55,59 weeks, non-indicated delivery < 34 or < 37 weeks excluding PPROM35, PTL with delivery < 37 weeks36,58, or sPTB with cervical dilation and/or PPROM < 37 weeks42. Furthermore, most prior studies have considered cases of sPTB uniformly29–32, 34,37, 39–44,47,48, 50–55,59, without subclassification into those deriving from PTL and those due to PPROM. As the relationship between the vaginal microbiota and underlying etiologies of PTL and PPROM may differ79–82, it is important that these two conditions leading to preterm delivery be considered separately. In summary, variation in sample size, cohort ethnicity, gestational age at sampling, and definitions of preterm birth potentially explain the inconsistency in results among prior studies evaluating associations between the vaginal microbiota and sPTB.
Herein, we report the largest investigation yet conducted to evaluate associations between the composition and structure of the vaginal microbiota and sPTB. This study therefore had comparatively higher power to detect significant associations if they indeed exist, and the repeated measurements from each pregnancy allowed for a higher effective sample size compared to a cross-sectional study design. The study cohort is primarily African American (94.3%), a population which experiences disproportionately negative perinatal health outcomes83–93. Given that the cohort is primarily African American, there is not a risk of diversity in ethnicity within the cohort muting associations that exist in one or more ethnicities in the cohort but not in others. Furthermore, the distributions of the gestational ages at sampling between women ultimately delivering preterm or at term were very similar. All vaginal fluid samples included in the analyses were collected before 37 weeks (i.e., preterm), and all analyses included gestational age at sampling as a co-variate. Lastly, we considered sPTB as a uniform condition yet further subclassified it as PTL or PPROM; analyses were conducted for each of these pregnancy outcome categories. This enabled us to identify phenotype-specific associations in the structure of the vaginal microbiota throughout gestation and the occurrence of PTL or PPROM.
The alpha diversity of the vaginal microbiota was highest among women ultimately experiencing early PPROM; these were the only women who exhibited vaginal microbiota alpha diversity values higher than those of controls. Notably, this pattern was evident throughout gestation. Increased alpha diversity of the vaginal microbiota among women ultimately delivering preterm compared to those delivering at term has been previously reported30,39,43,44,52,54, however, this pattern has not been universally observed32,53,59. Yet, like we report here, two prior studies did identify an association between high vaginal microbiota alpha diversity and the incidence of PPROM specifically38,46. Therefore, although high vaginal microbiota alpha diversity is not consistently associated with sPTB, it does appear to be associated with PPROM, and especially early PPROM.
At the community level, the majority (~ 60%) of variation in the composition and structure of the vaginal microbiota in pregnancy was due to individual subject identity. Even though women worldwide harbor the same general bacterial taxa in the vagina61–64, 75, their microbiota is nonetheless individual-specific. The second largest effect on the composition and structure of the vaginal microbiota was gestational age at sampling (~ 2%). Only one-percent of the variation in the vaginal microbiota throughout pregnancy (i.e., from 8 to 36+ 6 weeks) was explained by clinical outcome (i.e., PTL vs PPROM vs term delivery). Indeed, in this study, there was only one significant association between vaginal CST and clinical outcome. Specifically, there was an association between vaginal CST IV-B, comprised of L. iners, Gardnerella vaginalis, Megasphaera sp., and Fannyhessea (Atopobium) vaginae, and early PPROM. As CST IV-B is a species rich and diverse CST (see Fig. 2C), this finding is consistent with the noted positive association between vaginal microbiota alpha diversity and the likelihood of early PPROM. The unique feature of CST IV-B is a high relative abundance of Fannyhessea (Atopobium) vaginae. This bacterium has been associated with bacterial vaginosis94–97 and with sPTB54,98,99. Indeed, in a recent study of the vaginal microbiota in midtrimester pregnant women in Nigeria, F. vaginae was the best predictor of preterm birth54, and in a study of first trimester pregnant women in China, it was the best predictor of spontaneous abortion100. Aside from its ability to contribute to biofilm formation in women with bacterial vaginosis101, the virulence factors of F. vaginae remain largely unknown102. This warrants further investigation.
In addition to the association between the vaginal microbiota considered at the community level (i.e., CST IV-B) and the likelihood of PPROM, there were significant associations between individual bacterial taxa and early PPROM as well. Specifically, ASVs in the genera Anaerococcus, Finegoldia, Mobiluncus, Peptoniphilus, Prevotella, and Sneathia were positively associated with early PPROM, and some of these associations (i.e., Anaerococcus, Mobiluncus, Prevotella, Sneathia) were already evident within the first trimester. Mobiluncus42,51, Prevotella39,43,51,52,54, and Sneathia42–44, 53 have been identified as being positively associated with sPTB in general in multiple recent molecular investigations. Notably, bacteria within these genera have documented virulence factors that could compromise the integrity of the cervicovaginal epithelium as well as the fetal membranes. The genus Mobiluncus includes two species, M. curtisii and M. mulieris. Both inhabit the human vagina and are associated with bacterial vaginosis103,104. They are motile by means of multiple flagella103,104, can adhere to vaginal epithelial cells105, and produce extracellular cytotoxins that are destructive to epithelial cells106. The genus Prevotella has many species107. In the current study, P. amnii, P. bivia, and P. buccalis were the most relatively abundant Prevotella species within the vagina. Prevotella spp. are associated with both bacterial vaginosis108,109 and intra-amniotic infection110–112. They are key producers of the enzyme sialidase, which can degrade the mucosa of the vaginal epithelium, resulting in tissue damage108,109,113,114. The genus Sneathia has two species, S. sanguinegens and S. vaginalis (formerly S. amnii115). Both reside principally in the vagina61,116,117. Sneathia spp. have been strongly implicated in bacterial vaginosis118–127, and are among the most frequently detected bacteria in the amniotic fluid of women with intra-amniotic infection with intact128 or ruptured112,129 membranes. Both Sneathia species elicit pro-inflammatory responses from vaginal epithelial cells130. The virulence factors of S. sanguinegens have not been investigated in depth, but studies of S. vaginalis revealed that S. vaginalis can adhere to cervicovaginal epithelial cells116 and jeopardize the integrity of the fetal membranes131. Mechanisms of action for injury of host epithelial cells include cytotoxins131, invasins116, internalins116, and sialidase activity132. Given the potential role of Mobiluncus, Prevotella, and Sneathia in gynecologic and obstetric disease, including early PPROM, further investigation is needed to develop a detailed understanding of the collective molecular mechanisms underlying their pathogenesis.
In this study, the presence of only one bacterium was negatively associated with the incidence of early PPROM – Lactobacillus crispatus. In prior molecular investigations of the vaginal microbiota and sPTB, Lactobacillus has been consistently associated with delivery at term31, 34–36,41,43,44,48,49, 51–54,59, and L. crispatus is hypothesized to play a protective role in the vagina both in and outside of pregnancy34–36,43,48,66,67,69,70,133−137. The mechanisms underlying the hypothesized protective effect of L. crispatus within the vagina include its production of lactic acid (especially D-lactic acid)138–141 and antimicrobial compounds (e.g. bacteriocins)139,142, both of which can inhibit the growth of potentially pathogenic bacteria. Given the evident benefits of L. crispatus in the vagina, its potential use as a prophylactic probiotic in women’s reproductive health is under consideration143–146.
The impetuses for molecular investigations of the vaginal microbiota in the context of sPTB are twofold. First, identifying the specific bacterial taxa most associated with sPTB can facilitate elucidation of the candidate mechanisms of pathogenesis, thereby pointing the way toward novel therapeutics. Second, establishing microbial biomarkers that identify early in pregnancy the women at greater risk for sPTB, and thus providing opportunities for proactive clinical management. Herein, using machine learning models evaluated via cross-validation, we show that the structure of the vaginal microbiota in the first and second trimester does not predict sPTB or PTL. However, the structure of the vaginal microbiota prior to 28 weeks modestly predicts the occurrence of PPROM cases delivered between 28 and 34 weeks, and the structure of the vaginal microbiota prior to 24 weeks predicts PPROM with delivery by 30 weeks. Therefore, molecular investigation of the vaginal microbiota in pregnancy may have clinical utility in a subset of sPTB cases – those culminating in early PPROM. The contribution of the vaginal microbiota to the underlying etiologies of early PPROM should be a focus of perinatal microbiome investigation moving forward.