Prevotella bivia as a Source of Lipopolysaccharide in the Vagina

Abstract

Objectives

To compare vaginal lipopolysaccharides (LPS) concentrations between patients with and without bacterial vaginosis (BV), to evaluate the correlation between Prevotella bivia colonization density and LPS concentration, and to determine the impact of LPS on loss of dopamine neurons (DA).

Methods

Vaginal washes obtained from patients with (n=43) and without (n=59) BV were tested for quantity of P. bivia cells using quantitative PCR and for concentrations of LPS using the Limulus Amebocyte Lysate gel clot method. Prevotella bivia, Gardnerella vaginalis and Escherichia coli sonicated cell extracts were also tested for LPS production. DA neuron cells obtained from embryonic day (E) 14.5 pregnant rats were exposed to fluid from eight vaginal washes; tyrosine hydrolase immunoreactive staining was applied for visualization and cell counts.

Results

The median LPS concentrations were dramatically higher among patients who had symptoms of BV compared to those who did not have symptoms (3235.0 vs 46.4 EU/ml, respectively, P<0.001); patients who had BV also had much higher colonization densities of P. bivia (0.06±0.36 vs 5.4±2.2 log₁₀ CFU/ml, respectively, P<0.001).

Prevotella bivia cell lysates resulted in a higher LPS concentration (10,713.0±306.6 EU/ml) than either E. coli (4,679.0± 585.3 EU/ml) or G. vaginalis (0.07±0.01 EU/ml of LPS).

The loss of DA neuron was 20–27% in cultures treated with vaginal washes from BV-negative patients and 58–97% in cultures treated with vaginal washes from patients with BV.

Conclusion

P. bivia produces high LPS concentration, which may create a toxic vaginal environment that damages DA neurons.

1.1. Introduction

The gram-negative bacterial vaginosis (BV)-associated anaerobe, Prevotella bivia, inhabits the lower genital tract and has the propensity for adherence to and invasion of human cervix epithelial cells [1]. Anaerobic bacteria were isolated in 91% and 18% of BV-positive and BV- negative patients, respectively[2]; Prevotella bivia and other Prevotella spp. represented 44% of all anaerobes isolated in the BV-positive group. Evaluation of BV with gram-stained vaginal slides showed that Prevotella and Gardnerella vaginalis counts correlated strongly with the Nugent score [3].

It has been shown that an increased rate of premature labor and preterm delivery [1–3] occurrs among women with P. bivia at a concentration of >10 ⁴ bacteria/ml of vaginal fluid [4]. An experimental rabbit model also confirmed an association between P. bivia and preterm birth, increased level of TNF-α, and chronic intrauterine and fetal infection [5]. A persistent intrauterine inflammatory state that can occur in chronic conditions such as BV suggests that preterm delivery may follow chronic exposure to organisms, perhaps with resulting fetal brain damage [6, 7].

Prevotella bivia and other Prevotella species contain endotoxin, lipopolysaccharides (LPS), on their outer membranes; LPS is the most potent antigenic component of the Gram-negative bacterial cell wall. Some species of Prevotella produce LPS that is much more potent in inducing a rapid platelet response than the LPS found in Escherichia coli and Salmonella typhimurium.[8].

Exposure to a low dose of the bacteriotoxin, LPS, during a critical window of vulnerability of fetal development led to the birth of rat pups with fewer than normal dopamine (DA) neurons [9, 10].

Since BV results in a relatively high concentration of LPS-producing P. bivia in vaginal fluid, the aim of this study was to determine the correlation between the quantity of P. bivia and the level of LPS in the vagina of patients with BV. We also investigated whether LPS presented in the vaginal fluid of patients with BV is toxic for dopamine neurons.

2. 1. Material and Methods

2.1.1. Vaginal wash

A total of 102 vaginal washes (43 from patients with BV and 59 from patients without BV) were obtained from women of reproductive age visiting the Women’s Hospital of Texas, Houston, TX. The samples were studied for the level of endotoxin, lipopolysaccharides (LPS) and the quantity of P. bivia in the vagina. All patients enrolled into the study signed written informed consent. The study had approval from the Woman’s Hospital of Texas IRB committee.

BV was diagnosed based on clinical characteristics (Amsel criteria) [11] and on interpretation of a gram-stained slides, as described by Nugent et al [12]. Vaginal washes were conducted by instilling 3 ml of sterile non-pyrogenic saline into the vagina and aspirating the fluid with a sterile syringe. The process was repeated using the saline aliquot three times. The sample was maintained at −20° C and shipped to the laboratory packed in dry ice. Before testing washes were centrifuged at 11,000g × 15 min to separate cells. Supernatant was used for LPS detection and the cell pellet was used for extraction of DNA.

2.1.2. LPS determination

LPS was measured using the Limulus Amebocyte Lysate (LAL) (Cape Cod Inc., East Falmouth, MA) in pyrotell gel clot method according to manufacturer’s instruction. Briefly, supernatant from vaginal washes and bacterial cell walls was diluted two fold using endotoxin free water and tubes (Cape Cod Inc.). Then 0.1 ml (v/v) of LAL reagent was added to each tube. After 60 min of incubation in a water bath at 37°C, the tubes were removed and gently inverted. If gel had formed and remained intact at the bottom of the tube, the test was evaluated as positive. The concentration of endotoxin in the sample was calculated by multiplying the dilution factor on LAL-positive tests; the lowest concentration that can be determined using the gel clot technique is 0.03 LPS EU/ml. Escherichia coli endotoxin (Cape Cod Inc.) was used as a control.

To compare LPS production by gram-negative vaginal microorganisms, four strains of each microorganisms, P. bivia, G. vaginalis and E. coli were tested. Prevotella bivia and G. vaginalis were isolated from study patients in the Microbiology Laboratory of Women’s Hospital of Texas and four vaginal strains of E. coli were selected from our culture collection; no patients with vaginal E coli were found in the current study group.

Cells were harvested from agar plates after 24hrs of incubation and suspended in endotoxin free water (Cape Cod Inc.) to achieve 0.5 McFarland turbidity, that corresponds to 1.5 × 10⁸ CFU/ml. The suspension was then diluted 10-fold to a concentration of 10⁷ CFU/ml. Cell walls were disrupted using sonication (15 sec × 3 times). Lysate was separated from cell debris by centrifugation and then tested for LPS concentration.

2.1.3. DNA isolation

The vaginal wash pellets containing squamous epithelial and bacterial cells were used for DNA isolation. DNA was isolated using the Bactozol kit (Molecular Research Center, Inc., Cincinnati, OH). Briefly, cell suspension was treated with 0.1ml of 1X Bactozyme and incubated at 50°C for 60 min. After lysis was achieved, we used DNAzol and polyacrilamid to isolate DNA. The DNA was precipitated by adding 100% ethanol.

The total DNA concentration at A 260 and purity at ratio A 260/A280 was measured using a Beckman DU-600 spectrophotometer.

2.1.4. Quantitative PCR

For sequencing and amplification of the Prevotella bivia 16SrRNA gene, the following primers were used: forward: 5′-AGGGATAACCCACCGAAAGTTGGA-3′, reverse: 5′-TAAATCCGGATAACGCCCGAACCT-3′. Selected sequences were analyzed for homology using the GeneBank database with BLAST web site (National Center of Biotechnology Information, National Institute of Health, Bethesda, MD). All primers were obtained from IDT, Coralville, IA.

Quantitative PCR was performed using a Light Cycler 143 (Roche, Indianapolis, IN) and SYBR green, as a fluorescents dye (Takara Bio Inc., Otsu Shiga, Japan). After 95°C 30sec of denaturation, the shuttle PCR protocol involved 35 cycles of 95°C 5 sec and 60°C 20 sec. Purified genomic DNA from P. bivia ATCC 29303 was used to generate a standard curve consisting of 10³ to 10⁷ copies of DNA [13]. The bacterial counts represent the total amount in 1 ml of vaginal wash. This was obtained by determining the total DNA isolated from the wash samples times the number of bacteria in 50 ng of DNA.

2.1.5. Dopamine neuron cell culture bioassay

Eight vaginal washes collected from five patients with BV and three patients with normal vaginal ecosystem were employed for the DA toxicity study. Vaginal washes collected from women with BV contained log₁₀ 6.64 ± 0.73 of P. bivia cells/ml and LPS at a level of 8072 EU/ml (2 patients) and of 5030 EU/ml (1 patient); two other washes contained log₁₀ 6.10 of P. bivia cells/ml and 2048 EU/ml of LPS. Three patients without BV had no P. bivia in the vagina and LPS at a level of 0 EU/ml (2 patients) and of 16 EU/ml (1 patient). DA neuron tissue culture was obtained from embryonic day (E) 14.5 pregnant rats. Procedures used in these studies were approved by the Institutional Animal Care and Utilization Committee (IACUC) of Rush University. The protocol for this in vitro study has been published elsewhere [14]. The number of tyrosine hydrolase (TH) immunoreactive cells assessed using procedure is described in detail elsewhere [15]. Briefly, the single cell suspension was prepared from mesencephalic tissue using sequential incubation with trypsin and DNAase/trypsin inhibitors. The tissue was triturated into single cell suspension in 3 ml complete media and plated into the 48-well plate. The plate was incubated for 24 hrs to stabilize DA neurons. The following day, 20% of the media was replaced with vaginal wash fluid for an additional 72 hrs incubation. The plate was washed with PBS, fixed with 4% formaldehyde, and prepared for immunochemistry with (TH) stain, the traditional marker for DA neurons. The plate was incubated with blocking solution containing 0.25% Triton-X and normal horse serum for one hour. Without washing, mouse anti-rat TH antibody (Immunostar, Stillwater MN; 1:20,000) was added and the plate was incubated at 4°C overnight. The endogenous peroxidase activity was eliminated using a1-h incubation with 0.1% periodate. The immunohistochemical procedure was continued by using a biotinylated horse anti-mouse IgG (0.5%; Vector Laboratories, Burlingame CA) for 1 h and peroxidase conjugated avidin–biotin complex (Vector Laboratories) for 1 h. The TH immunoreactive cells were visualized using 0.05% 3,3′-diaminobenzidine (DAB), 0.5% nickel sulfate, and 0.003% H₂O₂ in I/A solution (10 mM imidazole/50 mM sodium acetate).

2.1.6. Statistical analysis

Categorical variables were tested by performing the chi-square test. We graphically displayed the distribution of P. bivia colonization density and the LPS concentration as a scatter plot and evaluated the correlation of these two factors by calculating Spearman’s rho. For all analyses we used a P-value of 0.01 to indicate statistical significance. Analyses were performed using Stata version 9.2 (Stata Inc., College Station, TX.).

3.1. Results

3.1.1. Prevotella bivia and LPS in the vagina of patients with and without BV

As shown in table 1, patients who had BV were much more likely to have vaginal colonization by P. bivia compared to patients who did not have BV (90.7% vs 3.4%) with the mean colonization density log₁₀ 5.4 vs 0.06 CFU/ml, P<0.001. The LPS concentration also discriminated between patients with and without BV (3235.0 vs 46.4 EU/ml, P<0.001). Surprisingly, P. bivia cells were not found in the vagina of five BV-positive patients, which correlated with zero amounts of LPS in three patients and low levels of LPS (6.125 – 26.5 EU/ml) in two patients.

Table 1.

The quantity of P. bivia and the level of LPS in the vagina of patients with and without bacterial vaginosis

No-BV (59 patients)	P. bivia cells (in 10 log)	LPS (EU/ml)
21	0	0
16	0	6.125±3.23
16	0	26.5±4.26
4	0	54.0±13.58
1	1.9	64.0
1	2.10	128.0
BV (43 patients)
3	0	0
1	0	32.0
1	2.11	128.0
4	3.32±0.12	128.0
4	4.32±0.17	490.0
2	5.62±0.03	520
15	6.31±0.08	3210.0±950.29
2	7.14±0.007	5030.0
9	7.52±0.13	7570.0±496.1

There was strong correlation between the P. bivia colonization density and the LPS concentration (Spearmen’s rho = 0.9750, figure 1).

Figure 1.

Scatterplot displaying the association between the lipopolysaccharide concentration and the quantity of Prevotella bivia in vaginal flora.

Setting a cutpoint of log₁₀ 2 of P. bivia and a level of LPS of 128 EU/ml, we showed that patients who had BV were much more likely to have a high density of P. bivia and LPS concentration than those without BV (Tables 2 and 3).

Table 2.

Association of BV with quantity of Prevotella bivia in the vagina

Characteristics	BV	No BV	P value
P. bivia < log 2 cfu/ml*	4	57	0.01
P. bivia ≥ log 2 cfu/ml	39	2

^*Cut point log 2 of P. bivia were used to classify the observed samples

Table 3.

Association of BV with level of LPS in the vagina

Characteristics	BV	No BV	P value
LPS < 128 EU/ml*	9	34	0.01
LPS ≥128 EU/ml	58	1

^*Cut point 128 EU/ml of LPS were used to classify the observed samples

3.1.2. Comparison of level of LPS produced by tested gram-negative microorganisms

Production of LPS by the mean count of cells log₁₀ 7.1/ml for P. bivia, G. vaginalis and E. coli were different. Prevotella bivia produced 10,700±307.0, whereas the same numbers of E. coli cells produced less than half that amount of LPS (4,700± 585.1 EU/ml). A trace amount of LPS (0.07±0.01 EU/ml) was observed from the G. vaginalis lysates.

3.1.3. Effect of vaginal washes with different level of LPS on loss of DA neurons in cell culture

Endotoxin levels correlated positively with DA neuron loss (P=0.006). The loss of DA neurons in cultures to which washes from normal vaginal environments were added was 20% (two patients with no LPS found) and 27% (one patient with LPS level 16 EU/ml), while DA neuron losses in the cultures to which vaginal washes from BV patients were added were 58% (one patient with LPS level 2048 EU/ml), 84% (two patients with LPS level 5030 EU/ml), and 97% (two patients with LPS level 8072 EU/ml).

4.1. Discussion

The current study showed that altered vaginal microecology creates a toxic environment in the lower genital tract of women. Patients with E coli as a source of LPS were excluded from this study. Elevated numbers of P. bivia in vaginal fluid showed that this Gram negative anaerobe might be a possible source of LPS in the vagina. Recently published data showed that vaginal fluid of healthy microecology contains twenty different aerobic and anaerobic bacterial flora but no P. bivia [16].

Among BV-associated microorganisms, G. vaginalis is the most frequently found in the vagina of patients with bacterial vaginosis [17]. However, compared to P. bivia, G. vaginalis produces an extremely minimal amount of LPS. These microorganisms have a symbiotic relationship [18][16] and were found growing together in 59.1% of women with BV compared to 3.9% of women without BV [19]. When a mixture of facultative anaerobes and P. bivia were grown together, the pathogenicity of P. bivia increased [20]. Prevotella bivia and G. vaginalis were detected in a high rate in patients with cervical cancer [21].

The views about LPS production by G. vaginalis vary. Sadhu et al. showed that G. vaginalis has a unique cell wall structure, resembling that of gram-positive bacteria, but with an unusually thin structure [22]. Using an LAL test these researchers showed that LPS was produced by G. vaginalis, albeit in low amounts. Greenwood and Picket described G. vaginalis as more closely resembling gram-negative bacteria. They also found cell wall material of G. vaginalis positive for endotoxin, but only very high numbers of cells were studied [23].

P. bivia has been isolated from the blood of patients delivered by caesarian section [24] as well as those diagnosed with acute pelvic inflammatory diseases [25]. It has been shown to have the capability to invade the human cervix [1] and cause intrauterine infection [20]. These reasons make this organism a risk factor for pregnancy outcome and fetus development. Our study findings suggest that the concentrations of P. bivia and LPS in the vaginal fluid may contribute to the severity of BV and to the development of complications in pregnancy, such as preterm labor, premature rupture of membranes, postpartum endometritis, and postoperative pelvic infection [26, 27].

Microbial attack of the fetus takes place in approximately 10% of pregnancies with intra-amniotic infection, and the human fetus is capable of deploying an inflammatory response (cellular and humoral) in the mid-trimester of pregnancy [28].

LPS is the most potent antigenic component of the gram-negative bacterial cell wall and modulates the expression of various pro-inflammatory cytokines, activates production of lysosomal enzymes, and increases phagocytotosis [29]. Purified endotoxin from P. bivia is clearly able to induce a cytokine response in monocytic cells [30].

The elevated concentration of LPS in the vaginal fluid and cervical mucus of pregnant women with BV was reported by Platz-Cristensen et al. [31]. High amounts of LPS in the vaginal environment can have numerous consequences. It was demonstrated that the toll-like receptor-4 (TLR-4) plays a role in lipopolysaccharide-induced preterm birth in mice [32]. Antenatal exposure to intra-amniotic endotoxin initiates a complex series of events including an inflammatory cascade that can alter lung structure [33]. An animal study showed that LPS causes histopathological changes in the reproductive organs of pregnant animals [34].

While BV has been considered a relatively benign problem, it can be a risk factor for the offspring and contribute to the risk of long-term adverse neurological outcomes, such as hyperactivity, academic difficulties in school, and severe handicaps, including cerebral palsy and preventricular leukomalacia.[35–38].

Maternal LPS triggers cytokine production by fetal tissue and even in low doses can dramatically sensitize the immature brain to injury [39–41].

Periventricular leukomalacia (PVL), the dominant form of brain injury in premature infants, is characterized by diffuse white matter injury and is associated with cerebral palsy. Clinically, cerebral palsy and periventricular leukomalacia are associated with intrauterine infection and inflammation[37, 42]. Animal models help to understand mechanism of the brain injury [43–45]. Induction of maternal bacterial infection in rabbits can reproduce a white matter injury in the fetal rabbit brain that has some similarities to PVL [46]. It was shown that the presence of intrauterine LPS can produce diffuse glial cell death and cavitation in fetal white matter [47]. Sherwin and Fern [48] showed that LPS can bind directly to developing white matter astrocytes and microglia to evoke rapid cell death in neighboring oligodendroglia. Astrocyte binding was particularly intense around blood vessels. In addition to direct toxicity, LPS increased the degree of acute cell death evoked by ischemia.

One of the author of the present study, Z. Ling and his colleagues (Neuropharmacology Laboratory, Department of Pharmacology, Rush University, Chairman and Project Director P. Carvey, PhD) experimentally showed that fetal brain injury by LPS also can be a risk factor for the future development of Parkinson’s diseases [10, 49]. The loss of DA neurons in these animals increased with age thereby mimicking the progressive pattern of cell loss seen in human Parkinson’s disease [9, 50]. Prenatal LPS exposure may cause disturbance in offspring brain and render DA neurons susceptible to the secondary neurotoxin insult [51].

Limitation of this study is that in vitro model of DA neurons loss caused by LPS found in vaginal fluid patients with BV can not give a full picture of all inflammatory cascade that take place in women with BV.

Our nearest goal is the study of pregnant patients’ vaginal environment, tracking pregnancy outcome and newborn health.

Conclusion

Production of LPS by P bivia can be an important factor in the pathogenesis of BV and subsequence complications. Further study is required to determine the range of LPS levels that can be used as an indicator for the development of adverse events for pregnant patients and for increased risk of fetal brain injury.

Level of LPS can be a marker for BV complications in reproductive age women.