Trichomonas vaginalis Weakens Human Amniochorion in an In Vitro Model of Premature Membrane Rupture

Objective: Trichomonas vaginalis (TV) infection is associated with preterm rupture of membranes (PROM) and preterm birth. We evaluated the effects of TV growth and metabolism on preparations of human amniochorion to understand and characterize how TV may impair fetal-membrane integrity and predispose to PROM and preterm birth. Methods: Term fetal membranes were evaluated using an established in vitro fetal-membrane model. Fresh TV clinical isolates were obtained from pregnant women. The protozoa (5.0×105 to 1.5×106/ml) were incubated with fetal membranes in modified Diamond's medium for 20 h at 37°C in 5% CO2.The effects of fetal-membrane strength (bursting tension, work to rupture, and elasticity) were measured using a calibrated Wheatstone-bridge dynamometer. Tests were also performed to evaluate the effects of 1) inoculum size; 2) metronidazole (50 μg/ml); and 3) cell-free filtrate. Results: The TV-induced membrane effects were 1) isolate variable; 2) inoculum dependent; 3) incompletely protected by metronidazole; and 4) mediated by both live organisms as well as protozoan-free culture filtrates. Six of 9 isolates significantly reduced the calculated work to rupture (P ≤ 0.02); 7 of 9 reduced bursting tension; and 1 of 9 reduced elasticity. One isolate significantly increased the work to rupture and bursting tension (P ≤ 0.002). Conclusions: In vitro incubation of fetal membranes with TV can significantly impair the measures of fetal-membrane strength. This model may be used to delineate the mechanisms of TV-induced membrane damage. This study suggests that there are enzyme-specific effects as well as pH effects.

the crucial elements. 2 Supporting this theory have been findings that the microorganisms common in vaginal flora, as well as those acknowledged to be genital pathogens, produce enzymes that can weaken fetal membranes in vitro. 7'9 Further, the proteolytic enzymes liberated by both microorganisms and host inflammatory cells can damage fetal membranes, thus decreasing the measurements of bursting tension, work to rupture, and elasticity. [10][11][12] Trichomonas vaginalis (TV) is a common sexually transmitted protozoan which causes symptomatic exocervicitis and vaginitis as well as asymptomatic infection. 13 TV organisms are enzymatically well endowed, producing numerous proteases, hemolytic substances, and other factors which potentially could damage maternal-fetal tissues and predispose infected women to pPROM and preterm birth. 14 '15 In a recent large epidemiologic study correlating vaginal infection with pregnancy outcomes, TV infection identified at mid-gestation was significantly associated with PROM, preterm birth, and low birth weight. 16 '17 Previous studies have demonstrated similar findings. 18 To assess and characterize how trichomoniasis in pregnancy may impair fetal-membrane integrity and increase the risks of PROM, we investigated the ability of TV to affect measures of the biomechanical strength of human fetal membranes in a well-characterized in vitro model.

Organism
Nine isolates of TV, recovered (rom pregnant women attending the Denver General Hospital antenatal clinic, were cultured in modified Diamond's medium (Remel Media, Inc., Denver, CO). The organisms were subcultured into fresh Diamond's medium with supplemental antimicrobials (1,000 U of penicillin G/ml, 100 Ig/ml of streptomycin, and 5 Ig/ml of fungizone) until the TV cultures were free of bacterial contaminants (usually 3 passages). The bacterial contamination was assessed by subculture onto chocolate and mycoplasma A7 agar (Remel Media, Lenexa, KS) for 48 h in 5 % CO 2 at 37C.
For experiments, axenic protozoa were subsequently subcultured into Diamond's medium without antibiotics for 48 h and tested in late log phase; the protozoan densities ranged from 5 l0 s to DRAPER ET AL.
1.5 10 6 organisms/ml as determined by a hemocytometer (Neubauer Chamber, Fisher Scientific, Denver, CO). The organisms were left in their used culture media and were not subjected to centrifugation or resuspension. For isolate comparison studies, fresh medium was not added to alter the parasite concentrations due to the negative regulatory effects on virulence-factor expression (Heine and Draper, unpublished observations). Only cultures with >90% parasite motility were applied to the membranes in 0.5-ml aliquots. The dynamometer plates were incubated with 50-rpm shaking (Orbital Shaker Model 361, Fisher Scientific) for 20 h at 37C in 5% CO 2. Uninoculated Diamond's medium was the negative control.

Membrane Preparation
Our in vitro model of membrane rupture has been described in detail elsewhere. 10,11 For our studies, 17 human fetal membranes were collected aseptically from normal, term placentas delivered by elective cesarean from women who had no evidence of PROM or chorioamnionitis. The membranes were transported to the laboratory for immediate processing. After being washed twice in pseudoamniotic fluid (PAF: 10 mM of urea, 2 mM of glucose, 20 mM of HEPES buffer with 125 mM of NaC1, 7 mM of KC1, 4 mM of calcium lactate, 1.4 mM of MgSO 4, and 0.4 mM of KHzPO4, pH 7.0) to remove blood and debris, the membranes were mounted between 2 sterile Plexiglas plates. These plates have circular perforations which, when aligned and bolted together as a unit, provide exposed surfaces of either amnion or chorion. From membrane, 3 plates were prepared yielding 60 wells/membrane for testing. In this way, the membrane both proximal and distal to the placenta was used. The chorionic side was inoculated with the TV culture or control medium, incubated for 20 h, and rinsed. After incubation, yet prior to dynamometer testing, cultures for bacterial contamination were performed on each well. If it was contaminated, the data for that well were eliminated. The amniotic side of each well was tested with a dynamometer to assess the bursting tension, work to rupture, and elasticity. In general, for each test variable, [10][11][12][13][14][15][16][17][18][19][20] replicates were tested on each membrane and 3 membranes were tested for each type of experiment. Strength Measurement The dynamometer is constructed from a 3-mm metal probe mounted on a pressure plate which is slowly advanced toward the membrane by a 3-V motor (2.7 mm/min). The pressure plate contains a strain gauge connected to a Wheatstone bridge. The Wheatstone-bridge signal is amplified and sent to the 1-mV strip chart recorder. The probe is advanced against the amniotic side of the membrane until rupture is achieved, and the membrane deformation curve is used to calculate the measures of membrane strength ( Fig. 1).

Metronidazole Protection
For our test of the protective effect of metronidazole, a highly cidal concentration of the drug, which should kill >90% of the isolates, was added to the inocula of parasites and incubated on the membranes. Four strains were chosen for study on 2 different membranes. The inocula were prepared from young, log-phase (40 h old) TV cultures at eliminate pH effect, then split into 2 parts. One portion was filtered and inoculated onto fetal membranes. The effect of this inoculum was compared with the parasite-containing culture. To prepare parasite-free filtrates, 10 ml of a 48-h culture in log phase was centrifuged at 750g for 10 min. The supernatant was removed and filtered through a 0.45-1xm filter (E-D Scientific Specialists, Intermountain Scientific, Salt Lake City, UT).

Protease Assay
The protease activity in culture supernatants was assessed by a fluorescent substrate cleavage assay. Fifty microliters of 0.4% resorufin-labeled casein (Sigma Chemical Co., St. Louis, MO) was mixed with a 100-1xl aliquot of culture supernatant and 50 Ixl of incubation buffer (0.2 M Tris with mM of CaCI2, pH 7.8) and incubated for 3 h at 37C. The reaction was stopped with 500 Ixl of 5% trichloroacetic acid. The mixture was incubated for 10 min at 37C and centrifuged at 2,600 rpm in a Beckman table-top centrifuge for 10 min. Four hundred microliters of supernatant was added to 600 txl of assay buffer (0.5 M Tris, pH 8.8) and the mixture read on a Perkin Elmer (Norwalk, CT) M2600 fluorescence spectrophotometer at an excitation wavelength of 578 nm and emission wavelength of 592 nm. Purified trypsin (Sigma Chemical Co.) was used as a standard to construct the activity curves.
Data Analysis A Gateway 2000 PC computer and Sigma Scan program (Jandel Scientific, Corte Madera, CA) were used for the calculation of work to rupture (from area under the curve), bursting tension (from peak height), and elasticity/plasticity (from slope of the peak). 0,11 Statistical tests were performed on averaged data from 10 to 20 replicates for each test inoculum or medium controls/membrane. Additionally, the location of the test replicates was randomized across all test plates to normalize intramembrane variability. Experiments were performed on 2-3 different membranes for each isolate. The data from wells containing control medium were compared with data from TV-inoculated wells. The data are presented as relative percents of effect rather than grams of work. This normalization is necessary so that the effects can be compared between membrane experiments. The statistics were  0.93. Across a membrane surface (or within a single membrane), the variability could range from as little as 10% to as much as 40%.
The effects of metronidazole treatment on TVinduced membrane weakening were observed. Metronidazole killed trichomonads as evidenced by a 10to 100-fold drop in hemocytometer counts (1.0 x 106 to 1.5 x 104) for all strains after 20 h incubation. Metronidazole treatment only partially protected the membranes from parasite attack. The effect of strain T 6 was significantly reduced (P < 0.05). There was a tendency toward protection with the 3 remaining isolates (P 0.07). This finding suggested that live parasites were required for maximal membrane damage to occur and that virulence factors were preformed and secreted in the culture supernatant (Fig. 3). Garber  Trichomonas isolate (T) Fig. 5. Protease activity in cell-culture supernatants of TV. Four protozoan isolates from pregnant women were grown in modified Diamond's medium and centrifuged, and the supernatant was filtered through a 0.45 im filter. The filtrates were assayed in triplicate and expressed as a mean of protease activity relative to a trypsin standard.
Tests also were performed to assess the role of extracellular roducts of TV metabolism on membrane damage. Cell-free filtrates from 48-h logphase parasites significantly decreased the fetalmembrane bursting tension and work to rupture, suggesting that extracellular factors produced by TV can impair fetal-membrane strength (Fig. 4).
Preliminary studies aimed at identifying the virulence factor demonstrated the presence of proteases in the culture supernatants. Protease activities ranged from 10 mU activity/ml of supernatant to 44 mU/ml. The level of protease activity did not entirely correlate with the amount of membrane damage (Fig. 5).
DISCUSSION TV can significantly impair human fetal-membrane strength in an established in vitro model by reducing the measures of bursting tension, work to rupture, and rarely elasticity. At lower test inocula, TV isolates from pregnant women damaged the fetal membranes in a strain-variable manner. These effects can be attributed to several factors including ) numbers of viable parasites; 2) secretion of membrane-damaging molecules; and 3) pH effects.
The inoculum comparison study ( 104 to 105 and occasionally to 106 organisms/ml, which approximate the test inocula in this study, z For our studies, we used clinically relevant test inocula from actively growing organisms. Inocula were not adjusted to equivalent densities for each isolate due to the negative regulatory effects of fresh media on virulence-factor expression (Heine and Draper, unpublished observations). It is important to realize that the parasites were applied to membranes in their used (spent) culture media, which contained live organisms, as well as secreted factors, and that both of these factors appear to contribute to trichomonal virulence.
In Table 1, the appearance of one isolate that increased bursting tension and work to rupture is difficult to explain. It may be that this strain produces a large amount of denaturation of membrane protein and actually causes the membrane to toughen. The basis of this is not clear. Our membrane studies of strength and pI-I sensitivity suggest that membranes weaken as pI--I drops. However, we did not test pI-I values below 4.0. In laboratory studies of parasite growth, the pH decrease usually ranges from 6.5 to 5.0, although we have seen as low as 4.0. The organism does not survive for long below pH 5.0 and rapidly dies and lyses. However, the lysing organism liberates hydrogenosomes which are rich in acidic metabolic products. 21 This observation may explain the phenomenon of membrane toughening for this isolate.
In the metronidazole studies, protection was not significant in 3/4 tests, although a trend was seen toward protection (P 0.07). The failure to provide full protection may in part be explained by inoculum age, density, and the presence of preformed virulence enzymes that are not sensitive to the antibiotic. We chose a test inoculum of 5 10 s organisms/ml and an inoculum age of 40 h. This strategy was necessary for several reasons. This culture density was chosen so that the inoculum allowed additional growth to the maximum density achievable in Diamond's medium (1-2 106 organisms/ml). The culture age provided parasites that were still in log-phase growth and presumably sensitive to the antibiotic. However, a culture age of 40 h means that there has been time to release some secreted factors. This timing was necessary because protozoa placed into fresh medium do not damage membranes within 20 h and do not produce virulence enzymes in Diamond's medium for at least 24 h until presumably all of their nutritional needs are met (Draper, unpublished observations). Finally, strain variability in the production of secreted virulence factors may explain failures of significant metronidazole protection.
Our data indicate that parasite-free culture filtrates are capable of damaging membranes even when corrected for pH. The fact that TV produces extracellular cytotoxic factors suggests that TV need not be directly present in order to damage amniochorion or other tissues. Honigberg et al. 22 and others 23'24 have demonstrated ultrastructural changes at a distance from trophs in tissue biopsies of vaginal epithelium during trichomoniasis, suggesting the presence of diffusible, extracellular virulence factors. Conversely, Alderate and Pearlman 25 and other workers 24'26 using tissue-culture systems have suggested that parasite contact is required for cytotoxicity. Garber et al. 27 have identified a cell detaching factor (CDF), which is a secreted, high-molecular-weight protease, and have shown that purified CDF is capable of disturbing cell monolayers in the absence of parasites. Additionally, our metronidazole studies show that a por-tion of the effect is due to viable growing parasites. The observations of Garber et al. 27 and Honigberg et al. 22 coupled with our findings of membrane damage from trichomonal supernatants suggest that diffusible, membrane-damaging factors are produced and could damage fetal membranes overlying the cervical os.
Our preliminary analysis of the factors in culture supernatant indicates that proteases are present.
Presumably, these virulence factors attack the components of membranes that engender strength and resist rupture. These components are probably collagen fibers and fiber bundles. It has been suggested that the dense matrix of collagen fibers underlying the basement membrane of the amnion is the main "load-bearing" structure. 28 Proteases which are collagenases or gelatinases (attack denatured collagen) would be obvious enzymes of virulence, and it is known that this protozoan is protease rich. 14' 5 It is also known that trichomonal proteases are expressed in vivo and that antibodies are produced in response to protease expression during infection. 29 Finally, the observation of parasite variability in impacting fetal-membrane strength has clinical relevance. It suggests that not all TV strains are alike and may explain why all pregnant women with undetected trichomoniasis do not suffer from PROM. In vivo, the membrane damage required for PROM may be an interaction of the number of protozoa in the vaginal canal, liberated virulence factors in the upper genital tract, virulence of the specific parasite, and host response factors.
In this in vitro model, clinical isolates of TV impaired the measures of fetal-membrane strength in a strain-variable and inoculum-dependent manner. Membrane damage was partially prevented by metronidazole treatment. Further basic research and clinical investigation are required to better evaluate the mechanisms of TV-associated effects that can increase the risks of pPROM and preterm birth.