Lactobacillus crispatus Dominant Vaginal Microbiome Is Associated with Inhibitory Activity of Female Genital Tract Secretions against Escherichia coli

Objective Female genital tract secretions inhibit E. coli ex vivo and the activity may prevent colonization and provide a biomarker of a healthy microbiome. We hypothesized that high E. coli inhibitory activity would be associated with a Lactobacillus crispatus and/or jensenii dominant microbiome and differ from that of women with low inhibitory activity. Study Design Vaginal swab cell pellets from 20 samples previously obtained in a cross-sectional study of near-term pregnant and non-pregnant healthy women were selected based on having high (>90% inhibition) or low (<20% inhibition) anti-E. coli activity. The V6 region of the 16S ribosomal RNA gene was amplified and sequenced using the Illumina HiSeq 2000 platform. Filtered culture supernatants from Lactobacillus crispatus, Lactobacillus iners, and Gardnerella vaginalis were also assayed for E. coli inhibitory activity. Results Sixteen samples (10 with high and 6 with low activity) yielded evaluable microbiome data. There was no difference in the predominant microbiome species in pregnant compared to non-pregnant women (n = 8 each). However, there were significant differences between women with high compared to low E. coli inhibitory activity. High activity was associated with a predominance of L. crispatus (p<0.007) and culture supernatants from L. crispatus exhibited greater E. coli inhibitory activity compared to supernatants obtained from L. iners or G. vaginalis. Notably, the E. coli inhibitory activity varied among different strains of L. crispatus. Conclusion Microbiome communities with abundant L. crispatus likely contribute to the E. coli inhibitory activity of vaginal secretions and efforts to promote this environment may prevent E. coli colonization and related sequelae including preterm birth.


Introduction
Preterm birth is a major health problem affecting 12.3% of live births and is a major cause of neonatal mortality [1,2]. Escherichia coli (E. coli) has been associated with preterm birth [3], is among the leading cause of neonatal sepsis in extremely low birth weight neonates [4], and is the leading cause of early-onset neonatal sepsis and meningitis [5]. Similarly, bacterial vaginosis (BV), a dysbiotic condition characterized by the replacement of a Lactobacillus crispatus (L. crispatus) predominant microbiome with anaerobic species [6,7], has also been linked to an increased risk of preterm birth, although treatment with antibiotics in late pregnancy was ineffective [8][9][10]. These findings indicate that the vaginal microbiome plays an important role in adverse pregnancy outcomes.
Recent studies suggest that the vaginal microbiome may be linked to soluble mucosal defense. Specifically, genital tract secretions have been consistently shown to possess in vitro inhibitory activity against E. coli [11][12][13][14][15]. This activity, which may reflect contributions from the host innate immune mediators such as defensins as well as from microbiota, may be critical during pregnancy and prevent dysbiotic vaginal colonization and ascending infection. Earlier studies showed that the E. coli inhibitory activity was reduced among non-pregnant women with BV and was restored following successful treatment with metronidazole [11]. In a recent cross sectional study, genital tract secretions obtained by vaginal swabs from near term healthy pregnant women were found to have significantly higher inhibitory activity against E. coli that was inversely correlated with E. coli vaginal colonization [12]. A separate study using cervicovaginal lavage (CVL) samples from healthy non-pregnant women suggested that the inhibitory activity may be mediated, at least in part, by soluble proteins secreted by lactobacilli [15]. Using biochemical techniques including mass spectrometry, four Lactobacillus proteins (three originally described as proteins of L. crispatus and one of L. jensenii) were present exclusively in CVL samples with high (.90% inhibitory activity), but not in samples with low activity (,20%). These proteins included the S-layer protein, a bacterial surface layer protein, and a cell separation protein for L. crispatus and adhesion exoprotein for L. jensenii [15].
Building on this background, the current study was designed to further evaluate the link between E. coli inhibitory activity and the vaginal microbiome and to test the hypothesis that high E. coli inhibitory activity would be associated with healthy Lactobacillus species dominant microbiome whereas low activity would be associated with a more diverse microbiome. Utilizing samples previously obtained from healthy near term pregnant and nonpregnant women, we also tested the hypothesis that pregnancy would alter the microbiome, as suggested in a prior study [16]. A subset of samples were selected from 10 pregnant and 10 nonpregnant women and within each cohort, five were selected for high inhibitory activity (defined as .90% reduction in number of E. coli colony forming units [cfu]) and five with low activity (defined as ,20% reduction in cfu).

Participants and Sample Collection
The parent study was described previously in detail [12]. Briefly, healthy pregnant women were recruited between 35 and 37 weeks of gestation and healthy non-pregnant were recruited during a routine gynecologic visit. Vaginal swabs were collected to measure soluble immune mediators and endogenous E. coli inhibitory activity. Following approval from the Montefiore Medical Center Internal Review Board, all women gave written informed consent. For this sub-study, vaginal swab pellets from 10 pregnant women between 35 and 37 weeks of gestation and 10 healthy non-pregnant women (5 with E. coli inhibitory activity . 90% and 5 with activity ,20% within each group) were selected and analyzed in a blinded manner for the vaginal microbiome. None of the participants had clinical BV using Amsel's criteria [17].
Vaginal swabs were placed in a 1.5 mL sterile eppendorf tube that was pre-filled with 0.5 mL of sterile normal saline, placed on ice, and processed within 6 hours of collection. The sample was vortexed and clarified by centrifugation at 2000 rpm for 7 minutes at 4uC and the cell pellets were re-suspended in 200 mL of phosphate buffered solution (PBS) and stored at 280uC until used for the microbiome analysis. The supernatants were divided into aliquots and used to measure E. coli inhibitory activity as well as concentrations of immune mediators [12]. The Internal Review Board of Albert Einstein College of Medicine approved the parent study; all participants signed informed consent and only subjects who agreed that samples could be used for future studies were included.

DNA Extraction and Amplification
Relatively low-cycle amplification and next generation sequencing (NGS) on the Illumina HiSeq2000 platform, in conjunction with pplacer software was used to analyze the bacterial composition of vaginal samples. DNA for sequencing was extracted from pelleted cells by incubating 150 ml of sample in 250 ml of proteinase K digestion cocktail containing 1% sodium laureth-12 sulfate at 55uC for 2 hours heated to 95u610 minutes and then the DNA was precipitated in a 0.825 M ammonium acetate/ethanol (AAE) solution, pelleted by centrifugation and re-suspended in TE (10 mM Tris, pH 8.0, 0.1 mM EDTA), as described previously [18][19][20]. The V6 region of 16S rRNA genes were amplified in an ABI 9700 Thermal Cycler (Life, Carlsbad, CA) using HotStart-IT FideliTaq DNA Polymerase (Affymetrix, Santa Clara, CA) and 50 ng of template DNA in a total volume of 50 ml. Reaction parameters included initial denaturation at 94uC for 2 min, followed with 30 cycles of denaturation at 94uC for 30 seconds, annealing at 60uC for 30 seconds, and extension at 68uC for 30 seconds; and a final extension at 68uC for 5 min. Sense target primers contained unique 8-bp Hamming DNA barcodes [21], which allowed for the identification of reads from each sample. Successfully amplified DNA from all samples was pooled, purified and isolated using gel electrophoresis and electroelution. Following quality control and library preparation, DNA was sequenced on an Illumina HiSeq 2000 using a paired-end protocol yielding 100 base pairs of sequence in each direction. Raw sequences were joined (using eautils' fastq-join), processed to remove chimeras (using uchime), and low quality reads and nucleotides, and assigned to their sample of origin by demultiplexing (using mubiomics). Demultiplexed reads were aligned to the bacterial reference library -a vaginal microbiome reference library available from http://microbiome.fhcrc.org/apps/refpkg/, containing 633 sequences, representing 138 bacterial taxa, using pyNAST. The aligned sequences were then mapped to their bacteria of origin using the classification algorithm pplacer [22] and visualized using R-scripts that were developed at Einstein [18] (see Statistical Analyses section).

E. coli Inhibitory Activity
Spent culture supernatants (SCS) from L. crispatus (ATCC 33197, M35, SJ-3C, and 60), L. iners (DSM 13335, UPII 60B-BEI HM-131, and 143D-BEI HM-126), and G. vaginalis (ATCC 49145, 14018) were prepared by centrifuging overnight cultures of bacteria at 2000 g for 15 min at room temperature and then filtering the supernatant to remove bacteria and particulate matter (0.22 mm syringe filter). L. crispatus strains were grown in MRS broth; L. iners in 1% proteose peptone, 1% beef extract, 0.5% yeast extract, 86 mM NaCl, 0.8 mM MgSO 4 , 0.3 mM MnSO 4 , 11.5 mM K 2 HPO 4 , 10% fetal bovine serum, and 2% glucose; and G. vaginalis in brain heart infusion broth supplemented with 10% FBS and 5% Fildes enrichment. The SCS from each of these cultures and their respective growth media were serially diluted in normal saline and then incubated with E. coli (ATCC strain 43827) (,10 9 cfu/mL) for 2 h at 37uC. The mixtures were then further diluted in saline (to yield 800-1000 E. coli colonies on control plates) and plated on agar enriched with trypticase soy broth [15]. Colonies were counted using ImageQuant TL v2005 after an overnight incubation at 37uC. To control for differences in cfu of the L. crispatus, L. iners or G. vaginalis cultures, the inhibitory activity was normalized to lowest bacterial yield in a post-hoc analysis or, alternatively, the SCS were diluted based on different yields prior to incubating them with E. coli. All samples were tested in duplicate and the percentage inhibition was determined relative to the colonies formed on culture media control plates.

Statistical Analyses
E. coli inhibitory activity was dichotomized as .90% (high) and ,20% (low). Categorical variables were compared between groups by Chi-square or Fisher's exact test. Continuous variables were compared by the Student t test or the Mann-Whitney U test, depending on the distribution of the data. Clinical data were analyzed using STATA (v11.0; StataCorp, Inc., College Station, TX). All plotting and statistical comparisons for NGS were performed in R v2.12.2 using a script developed in-house (available upon request). The pairwise Kantorovich-Rubinstein (KR) distances (equivalent to the weighted UniFrac distance [23] between samples were calculated with p = 1 and normalized with respect to the diameter of the reference tree. Principal component analysis was performed to determine differences between pregnancy and the microbiome and dichotomized E. coli inhibitory activity and the microbiome. Using a conservative approach, the PERMANOVA analysis was performed on the Kantorovich-Rubinstein (KR) distances between sample microbiomes, with pregnancy and E. coli inhibitory activity as the factors, respectively. Fisher's exact tests were performed with pregnancy status and E. coli inhibitory activity as binary factors against the dominant bacterial taxon. Shannon and Simpson diversity indices were calculated to determine significant differences in diversity between pregnant and non-pregnant samples and samples with high and low E. coli inhibitory activity. To ensure that sufficient sampling of the microbiome had occurred, a rarefaction analysis was performed. Bacterial taxa were grouped to reflect healthy lactobacilli (group 1), intermediate bacteria (at times associated with BV diagnosed by Nugent's score [24], group 2) and pathogenic bacteria (groups 3 and 4): Group 1: L. crispatus, L. jensenii, and L. gasseri, Group 2: L. iners, Group 3: G. vaginalis, Sneathia Sanguinegens, BVAB1, BVAB2, Megasphaera, Prevotella bivia, Prevotella melaninogenica, Prevotella genogroup 1 and genogroup 2, and Atopobium vaginae, and Group 4: Streptococcus anginosus and Staphylococcus hominis. Principal component analyses were performed using Group 1, Group 3, and Group 4 to reduce the dimensions of the data. Spearman correlation coefficients (SCC) were used to examine correlations between bacterial groups and soluble immune mediators measured in the parent study [12]. The following mediators were included in the analysis: secretory leukocyte protease inhibitor (SLPI), human neutrophil peptide 1-3 (HNP1-3), human beta defensin (HBD)-1, HBD-2, and HBD-3, and cytokines/chemokines including interleukin (IL)-1a, IL-1b, IL-1 receptor antagonist (IL-1ra), IL-6, IL-8, macrophage inhibitory protein (MIP)-1a, MIP-1b, and regulated on activation, normal T-cell expressed and secreted (RANTES). All tests were two-sided with p value of ,0.05 considered statistically significant.

Description of Participants
Sixteen of the 20 subjects had paired-end reads that were sufficient to characterize the microbiome. The characteristics of these subjects are summarized in Table 1 grouped according to E. coli inhibitory activity (10 with high activity and 6 with low activity). The two groups did not differ with respect to age, race, number of pregnant women, current smoking status, history of prior sexually transmitted infections (STI), contraceptive use,  (Table 1).

Microbiome Analysis
NGS produced 9,111,237 high quality, non-chimeric, joined, paired-end reads, of which 8,946,862 were successfully assigned a bacterial identity. Those that could not be aligned (and therefore assigned) were of non-bacterial origin. Less than 2% of the reads were determined to be chimeric and were therefore removed. The short reads were submitted to the NCBI Short-Read Archive (Submission ID: SRP034665).
Community compositions and proportional abundances of bacteria are shown in the heat map of Figure 1. All but one of the samples (JG46) was dominated by Lactobacillus species including L. crispatus and L. iners, which is consistent with the absence of clinical BV in the cohorts. Most strikingly, the majority of samples with high E. coli inhibitory activity were dominated by L. crispatus, while those with low activity were dominated by L. iners. The two exceptions were JG46 and JG50, which had a microbiome dominated by Atopobium vaginae and Lactobacillus gasseri, respectively; both samples were in the low E. coli inhibitory activity cohort. The microbiome communities found in the samples clustered into three groups: L. crispatus, L. iners, and L. gasseri and are consistent with the community types found by Ravel et al. [6] A rarefaction plot is a standard analysis used to evaluate whether sufficient reads have been tested to detect species richness. If the gradient of the rarefaction curve decreases sharply after some level of species richness, it is a good indication that increasing the read depth would discover only the very rarest of species. Due to the low gradient of the curves of the rarefaction analysis at the read depth at which the clinical samples were sequenced, only the rarest of bacteria are likely to have been missed from our analyses (data not shown).
High E. coli inhibitory activity was predictive of an L. crispatus dominated microbiome (Fisher's exact test, p value = 0.007), while low activity was not significantly associated with any dominant bacterial taxon. Furthermore, the vaginal microbiome in women with high E. coli inhibitory activity were significantly different than women with low activity by principal component analysis (p = 0.001, Figure 2A). However, there were no differences between the pregnant and non-pregnant women ( Figure 2B) and no association was found between pregnancy and any of the dominant bacterial taxa. Furthermore, we found no significant association between age, race, vaginal wall pH and L. crispatus predominance. There was also no difference in bacterial diversity between pregnant and non-pregnant women or between those with high or low E. coli inhibitory activity using Shannon and Simpson diversity indices (not shown).
In the parent study, the concentrations of protein and a subset of cytokines, chemokines, defensins and antimicrobial peptides were measured in the vaginal swab supernatants and SLPI, HBD-1, HBD-2, IL-1b, IL-6, and IL-8 correlated modestly with E. coli inhibitory activity among non-pregnant, but not pregnant women [12]. To assess whether any of these mediators were associated with the bacterial groups, Spearman correlation coefficients were measured. None of the bacterial groups correlated with the soluble immune mediators.

Spent Culture Supernatants from L. crispatus have Significant E. coli Inhibitory Activity
To gain further insight into the link between the microbiome composition and E. coli inhibitory activity, SCS from L. crispatus, L. iners, and G. vaginalis were tested for the ability to inhibit E. coli relative to the growth media for each bacteria. The bacterial yields of the cultures used to produce the SCS were as follows: L. crispatus 33197: 2.5610 9 cfu/mL; L. iners ARL1 (13335): 4.2610 8 cfu/mL; L. iners 60B: 1.0610 9 cfu/mL; L. iners 143D: 6.5610 8 cfu/mL; G. vaginalis 49145: 4.5610 8 cfu/mL; G. vaginalis 14018:1.0610 9 cfu/ mL. When the SCS were diluted 1:10 prior to mixing with E. coli, L. crispatus 33197 showed the greatest magnitude of inhibitory activity and reduced the E. coli cfu by greater than 1 log after adjusting for differences in bacterial yields, although SCS from L. iners 60B and G. vaginalis 14018 also showed significant inhibitory activity (p,0.05) (Figure 3a). To further explore the inhibitory activity of L. crispatus, we tested serial dilutions of 3 additional strains (after diluting the SCS to adjust for differences in bacterial yields as follows: L. crispatus M35: 4.1610 8 cfu/mL; L. crispatus SJ-3C: 1.3610 8 cfu/mL; and L. crispatus 60: 2610 7 cfu/mL). L. crispatus 60 displayed the most potent activity and retained significant activity even after dilution of 1:25 (Figure 3b).

Discussion
The current study demonstrates that a L. crispatus dominant vaginal microbiome community is associated with E. coli inhibitory activity and supports the contention that the microbiome contributes to host defense. This notion is further supported by the finding that culture supernatants from L. crispatus had the most potent inhibitory activity, although the activity varied between different strains. The results are consistent with prior observations that E. coli inhibitory activity is reduced in the setting of BV, a dysbiotic condition associated with loss of L. crispatus [6,7] and with a small proteomic study showing that Lactobacillus proteins (three originally described as proteins of L. crispatus and one of L. jensenii) were present exclusively in CVL samples with high but not in samples with low E. coli inhibitory activity [15]. While the microbiome from women with high E. coli inhibitory activity was dominated by L. crispatus, women with low activity (even in the absence of clinical BV) had a trend towards L. iners dominance. However, we did not observe any significant increase in bacterial diversity in samples from women with low E. coli inhibitory activity. It is possible that the other common healthy Lactobacillus species such as L. jensenii, also contribute to E. coli inhibitory activity. L. jensenii was one of the more prevalent bacteria in the women with high activity (Figure 1).
There were no differences detected in the vaginal microbiome between healthy near term pregnant and non-pregnant women. These findings are consistent with one other study in which the microbiome exhibited less diversity and richness in pregnant women sampled between 18-32 weeks gestation, but returned to the non-pregnant community structure in late gestation (.32 weeks) [16]. Both studies focused on healthy women, the majority of whom delivered at term. Larger studies in women at risk for preterm birth are needed to determine whether changes in the The spent culture supernatants obtained from three additional strains of L. crispatus (M35, SJ-3C, and 60) were normalized by diluting in culture media to that of the lowest growth (2610 7 cfu/ml) and then serial dilutions (neat, 1:5, 1:10 and 1:25) were mixed with E. coli and tested for inhibitory activity. Results are mean 6 SD from duplicate plates. The number symbol indicates that no bacterial colonies were observed on plates after incubation with undiluted L. crispatus 60 SCS. The asterisks represent a significant reduction in E. coli cfu relative to its control growth media (p,0.01). doi:10.1371/journal.pone.0096659.g003 vaginal microbiome and/or E. coli inhibitory activity will provide a biomarker of risk for adverse outcomes including E. coli colonization and associated sequelae (e.g. chorioamnionitis and neonatal sepsis).
The potential utility of E. coli inhibitory activity as a biomarker of mucosal health, however, may differ in populations where lactobacilli species are not the dominant microflora. This notion is supported by two small sub-studies of African women who were at high-risk for HIV acquisition [25]. In the latter studies, having higher E. coli inhibitory activity was associated with an increased risk of HIV acquisition and, in one study, with a higher viral set point [26]. However, the participants in these studies had relatively high Nugent scores and lower median E. coli inhibitory activity (50% inhibition) compared to that observed in the healthy U.S cohorts (.70% inhibition). Moreover, in the HIV seroconverters, E. coli inhibitory activity correlated with the concentrations of several pro-inflammatory cytokines and chemokines, suggesting that the activity may be a biomarker of inflammation in high-risk women. In contrast, in the current study, there were no significant correlations between concentrations of mucosal immune mediators and the bacterial groups further indicating that perhaps host immune factors contribute little to this antimicrobial activity in populations where the microbiome is typically dominated by protective Lactobacillus species. Thus, in populations where there is a relative paucity of protective lactobacilli (e.g. high risk African cohorts), E. coli inhibitory activity may be more influenced by inflammatory molecules and serve as a biomarker of HIV risk, whereas in populations where L. crispatus is common, high activity may be representative of a healthy vaginal microbiome.
A limitation of this exploratory study is the small sample size. Thus, as noted above, we cannot preclude a broader association with other healthy Lactobacillus species. Furthermore, conclusions about the effect of race, ethnicity, age, pH, contraceptive use and host immune mediators on the vaginal microbiome should be made with caution. Another study limitation is that we did not confirm sexual abstinence. Finally, differences in the inhibitory activity of bacterial SCS may be impacted by differences in the bacterial yields from which supernatants were obtained. In an effort to account for this difference, we adjusted the E. coli cfu by the bacterial cfu/mL used to produce the SCS in a post hoc analysis (Figure 3a) or adjusted the SCS by dilution prior to incubating the samples with E. coli (Figure 3b).
The current study highlights the possibility that E. coli inhibitory activity may be a functional feature of a Lactobacillus crispatus dominant healthy vaginal microbiome. In addition to studies with larger more diverse cohorts, future studies should also include proteomic and metabolomic analyses on the sample to further define the nature and origin (host and/or microbiome) of the molecules that contribute to the E. coli inhibitory activity. A longitudinal study in pregnancy is ongoing and will help to elucidate whether high inhibitory activity against E. coli translates to a reduction in sub-clinical or overt genital tract infection and preterm birth. These findings could promote the identification of novel strategies to enhance the antimicrobial activity of genital tract secretions. Further studies are needed to determine if introducing lactobacilli derived molecules or sustaining a L. crispatus dominant microbiome may promote a healthy vaginal environment. These approaches could lead to reduced risk of bacterial vaginosis, chorioamnionitis, preterm birth and perinatal infection and be used as a safer alternative to traditional antibiotics [27].