Amniotic fluid and maternal race influence responsiveness of fetal membranes to bacteria

Abstract

Spontaneous preterm birth (PTB) and preterm prelabor rupture of membranes (pPROM) occur more frequently in African-American women than in other racial groups. This may be due to an enhanced inflammatory response to pathogens associated with the condition. It is also possible that amniotic fluid (AF) has different immunomodulatory properties in African-American women that increase their risk of PTB and pPROM. To test this, we cultured fetal membranes from European-American and African-American women with sterile medium (control), Escherichia coli, Gardnerella vaginalis, Group B streptococci (GBS), Polyporphorans gingivalis, Mycoplasma hominis, Ureaplasma urealyticum or Ureaplasma parvum in the presence and absence of 50% autologous AF. Cytokine concentrations were quantified in the conditioned medium. All bacterial species increased IL-8 production. IL-1β and TNF-α production were stimulated by LPS, E. coli, and G. vaginalis compared with control, but responses to Group B streptococci and P. gingivalis were limited to IL-1β and TNF-α respectively. Genital mycoplasmas stimulated TNF-α and IL-10 but had no effect on IL-1β production. African-Americans had twice the IL-1β response to E. coli as European-Americans (P = 0.031). Conversely, European-Americans produced more IL-8 in response to LPS than African-Americans (P = 0.026). AF had both pro- and anti-inflammatory properties that varied between races and pathogens. These results suggest that the host response to fetal membrane infections is complex and not generalizable. Interventions to prevent PTB and pPROM may need to be customized based on a patient’s race, type of bacterial infection and factors in her AF.

1. Introduction

Spontaneous preterm birth (PTB < 37 weeks) is a complex syndrome characterized by etiological and pathophysiological heterogeneities. Microbial invasion of the amniotic cavity (MIAC) and intra-amniotic infection (IAI) have been associated with ~50% of all PTB (Romero et al., 2007; Goldenberg et al., 2008). Preterm prelabor rupture of the membranes (pPROM) complicates one third of all PTBs and ~75% of pPROM cases are associated with infection. The incidence of PTB and pPROM is higher in African-Americans than in other ethnic and racial groups for reasons that are unclear (Fiscella, 1996a, 1996b). Although a number of studies have identified genetic polymorphisms that are associated with increased risk of PTB, pPROM and its racial disparity (Fortunato et al., 2008; Menon et al., 2008, 2009a; Velez et al., 2008a, 2008b), the condition is not strictly a genetic disease.

Differences in the intrauterine immune responses are reflected in the pathways and likely in pregnancy outcome. Using an in vitro model of human fetal membranes, we found that IAI-associated pathogens produce distinct immune (cytokine) responses (Menon et al., 2009b). This suggests that the underlying inflammatory process that promotes labor pathways to PTB and pPROM cannot be simplified (Menon et al., 2009b). It is unclear, however, whether membranes from different races produce distinct immune responses to the various pathogens associated with PTB.

Despite its proximity, the role of amniotic fluid (AF) in the host response to ascending infection by fetal membranes is unclear. Previous studies have documented that AF contains immunomodulatory factors. AF suppressed antibody production in mice (Etlinger and Chiller, 1977), prolonged the survival of allografts in rats (Yoshimura et al., 1991), and suppressed polyclonal B-cell activation (Etlinger and Chiller, 1977), mixed lymphocyte reactions (Krco et al., 1979), mitogen-activated T-cell proliferation (Wajner et al., 1986), and IL-2 production (Lang and Searle, 1994). Recent studies have suggested that AF also altered cytokine production by unstimulated and LPS-stimulated peripheral blood mononuclear cells (Witkin et al., 2011, 2012). How AF may modulate the host response of fetal membranes to bacteria has not been studied. It is also unknown if the effects of AF differ between maternal races or pathogen associated with PTB. Therefore, the objectives of this study were:

1. To document distinct pathways of immune response to different pathogens associated with PTB and pPROM;

2. To document any racial disparities in fetal membrane immune response to these pathogens; and

3. To understand the effect of AF in regulating fetal membrane cytokine response to pathogens of IAI.

Using an in vitro model of the fetal membrane organ explant system, we tested the immune response of membranes from normal term pregnancies with no clinical evidence of infection to seven pathogens (genital mycoplasmas: Ureaplasma parvum, Ureaplasma urealyticum, Mycoplasma hominis, Escherichia coli, GBS, Polyporphorans gingivalis and Gardnerella vaginalis) by quantifying their production of five cytokines (IL-1β, IL-2, IL-8, IL-10, and TNF-α).

2. Materials and methods

The cell culture experiments described below were performed at The Perinatal Research Center, The Centennial Women’s Hospital Nashville, TN, USA. Samples were analyzed at Emory University, Atlanta, GA, USA, where the senior author had privileges. All tissues collections, cell cultures and sample analyses were in compliance with Institutional Review Board-approved protocols from both institutions.

2.1. Subjects and collection of fetal membranes and amniotic fluid

Placentas were obtained from elective repeat Cesarean sections at term (≥37 weeks of gestation) prior to onset of labor (n = 11). As we studied racial disparity in immune response, self-reported (back three generations from the maternal and paternal side) African-Americans and European-Americans of non-Hispanic origin were recruited for this study. Maternal ages were between 20 and 32 years. Extra-placental fetal membranes (~4–5 cm from the placental disc) were isolated by sharp dissection. Adhering blot clots were removed from the membranes using saline and sterile cotton gauze. These tissues were placed in Hanks balanced salt solution (HBSS, Sigma Chemical Co., St. Louis, MO, USA) containing 1% (v/v) penicillin/streptomycin solution (Sigma) and 1% (v/v) amphotericin B solution (Sigma). AF samples were collected from the same subject using a 22-gauge needle prior to rupture of the membrane and immediately centrifuged at 2500 rpm. Clarified AF was then stored in aliquots at −80 °C until use. Women with a prior history of pPROM or PTB were excluded, as were patients with obstetrical and medical complications, cervico-vaginal Group B streptococcus colonization, infections (sub-clinical infections as indicated by high C-reactive protein levels, foul smelling vaginal discharge, fever, or those undergoing antibiotic treatment). A random sample of fetal membranes from each patient was examined by light microscopy (H&E staining of paraffin-embedded sections) to rule out sub-clinical chorioamnionitis (>5 PMNs/high power field).

2.2. Tissue-culture of normal term fetal membranes

Fetal membranes were washed twice with HBSS as described above and cross-sections measuring 6 mm in diameter were cut out using a biopsy punch and washed again with HBSS before placing them in the tissue culture system. The details of this method can be found elsewhere (Menon et al., 2006). Briefly, tissue biopsies were placed in a Falcon™ cell culture insert (Becton-Dickinson Labware, Franklin Lakes, NJ, USA) containing 200 ml Dulbecco’s modified Eagle’s medium:F12 Ham’s mixture. These inserts were placed in a Falcon™ cell culture plate with 500 ml of medium. Medium was supplemented with 15% (v/v) heat-inactivated fetal bovine serum (FBS), 1% (v/v) glutamine solution, 1% (v/v) penicillin/streptomycin solution and 1% (v/v) amphotericin B solution (all from Sigma Chemical Co., St. Louis, MO, USA). Cultures were incubated at 37 °C, 5% CO₂ for 48 h. Culture medium was changed every 24 h.

2.3. Preparation of bacteria

Stock cultures of U. urealyticum (ATCC #27816), U. parvum (ATCC #27813), M. hominis (ATCC #14027), G. vaginalis (ATCC #49145), P. gingivalis (ATCC# BAA-1703), GBS (Streptococcus agalactiae ATCC#BAA-25), and E. coli (ATCC #33908) were purchased from the American Type Culture Collection (Manassas, VA, USA). U. urealyticum and U. parvum were cultivated in beef heart infusion broth supplemented with 20% (v/v) horse serum + 10% (w/v) yeast extract + 0.2% (w/v) urea + 1000 U/ml penicillin G + 40 mg/ml phenol red (pH 6.0). M. hominis was cultivated in PPLO broth supplemented with 20% (v/v) horse serum + 10% (w/v) yeast extract + 0.2% (w/v) arginine + 1000 U/ml penicillin G. G. vaginalis was cultivated in Casman’s broth + 5% (v/v) defibrinated sheep blood. S. agalactiae was grown in brain heart infusion broth supplemented with 5% (v/v) defibrinated sheep blood. E. coli was cultured in nutrient broth and P. gingivalis was cultivated anaerobically in chopped meat broth. To prepare bacteria for the studies below, organisms were cultured to late log-phase, concentrated by centrifugation and resuspended in RPMI 1640. Quantitative cultures were then established for a color changing unit (CCU) or CFU determination by serial dilution and plating on appropriate agar. Bacteria were then heat-killed by heating at 80 °C for 1 h. Stocks of heat-killed bacteria were then stored at −80 °C until use.

2.4. Stimulation of fetal membranes with heat-inactivated bacteria and LPS

Tissue cultures were stimulated with either 107 CFU/CCU of heat-inactivated bacterial suspension or 100 ng/ml LPS (E. coli O55:B5). Dilutions of the bacterial suspensions were made in tissue culture media. A set of tissue preparations were also stimulated with bacterial suspensions and media supplemented with 50% (v/v) autologous AF. Control cultures consisted of plain media with antibiotics and media with 50% AF and no bacteria. Tissue cultures were stimulated for an additional 24 h. Culture and stimulation conditions were chosen based on our prior work (Menon et al., 2009b). Media samples from both stimulated and unstimulated cultures (hereafter referred as controls) were harvested and stored at −20 °C until assay.

2.5. Immunoassays

Cytokine concentrations for IL-1β, TNF-α, IL-10, IL-6, and IL-2, were quantified using the Bioplex™ bead array system (Bio-Rad, Hercules, CA, USA) according the manufacturer’s instructions. IL-6 concentrations were above the assay range for the multiplex assay and excluded from analysis. Concentrations found to be at or below the sensitivity of the assay were set equal to the sensitivity of the assay. All samples were analyzed in a single assay.

2.6. Statistical analysis

Cytokine concentrations were analyzed using the nonlinear and linear mixed effects models package (nmle) of R (www.r-project.org). The first analysis consisted of a mathematical model that evaluated the effect of bacterial treatment on cytokine production by membranes from both races in the absence of AF. We then performed a series of stratified analyses to evaluate the effects of maternal race on responsiveness of the membranes to bacterial stimulation. A third analysis was then performed to evaluate the effect of amniotic fluid on cultures stimulated with each pathogen for the combined and separate maternal races. Effects due to the patient were considered random and separate intercepts were fit for each patient to adjust for patient-to-patient variability in background cytokine levels. All other effects were considered fixed. Models were checked for compliance with assumptions of parametric techniques (normality, identity and independence of errors). Data were log-transformed and reanalyzed for hypothesis testing (P values) when violations were detected. Least-squares means and contrasts between individual treatments were performed using the estimable statement of the gmodels package of R. Data are shown as the least-squares means SEM for analyses conducted on untransformed data for clarity of presentation.

3. Results

We included a total of 11 subjects for this study, 6 European-American and 5 African-American subjects. Races did not differ with regard to maternal age (27.5/4.2 years) or gestational age at delivery (39.1–1.2 weeks).

3.1. Effect of bacterial stimulation on cytokine production

LPS (P < 0.001), E. coli (P < 0.001), G. vaginalis (P < 0.001) and GBS (P = 0.010) significantly increased IL-1β production compared to controls. LPS had the most proinflammatory activity, stimulating a 43-fold increase in IL-1β production (P < 0.001). Both E. coli (P < 0.001) and G. vaginalis (P < 0.001) increased IL-1β production about 20-fold. GBS induced a much smaller, but statistically significant, fourfold increase in IL-1β production (P = 0.010). Although the genital mycoplasmas increased IL-1β production by up to 2.7-fold, none of these increases was statistically significant.

All pathogens, except for GBS (P = 0.107), significantly increased TNF-α production with LPS, E. coli, and G. vaginalis mounting the strongest response (P < 0.001 for each). With the exception of GBS (P = 0.446) and P. gingivalis (P = 0.123), all bacterial species also significantly increased IL-10 production. U. parvum, however, was the most effective pathogen at stimulating secretion of this anti-inflammatory cytokine, causing nearly a sevenfold increase in production (P < 0.001).

All pathogens significantly increased IL-8 production, but to an equivalent degree (Fig. 1D). None of the pathogens examined, nor LPS, had any detectable effect on IL-2 production (data not shown), which was at the sensitivity of the assay for all pathogens (2.0 pg/ml).

Fig. 1.

Racial disparity for cytokine production by fetal membranes. The least squares means ± SEM for IL-1β (A), TNF-α (B), IL-10 (C), and IL-8 (D) production for unstimulated (CTL) and E. coli (EC), Group B streptococci (GBS), G. vaginalis (GV), lipopolysaccharide (LPS), M. hominis (MH), P. gingivalis (PG), U. parvum (UP) or U. urealyticum (UU) stimulated cultures stratified by maternal race. Note that data were log-transformed for analysis and plotted on a log scale for IL-1β and TNF-α. Pairs of bars marked with an asterisk indicate statistically different effects of maternal race (P ≤ 0.05) for that bacterial stimulation. Bars marked with asterisks are significantly different from unstimulated cultures at *P < 0.05, **P < 0.01, and ***P < 0.001.

3.2. Effect of race on bacteria-stimulated cytokine production

Racial disparity in cytokine production was not evident in control (unstimulated) fetal membrane cultures derived from European-Americans and African-Americans, with each race having similar concentrations of IL-1β, TNF-α, IL-8, and IL-10 in membrane-conditioned medium (Fig. 1). Bacterial stimulation, however, revealed racial disparities in the fetal membrane cytokine response that are summarized in Table 1.

Table 1.

Effect of different bacterial species on cytokine production in fetal membranes overall (African-American, n = 5 and Caucasian, n = 6). Arrows indicate statistically significant (P <0.05) fold change in cytokines compared to unstimulated controls (↑ = 1-4 fold increase; ↑↑ = 5-9 fold increase; ↑↑↑ = 10+ fold increase).

Pathogen	IL-1β	TNF-α	IL-10	IL-8
E. coli	↑ ↑ ↑	↑ ↑ ↑	↑	↑
GBS	↑	–	–	t
P. gingivalis	–	↑ ↑	–	↑
G. vaginalis	↑ ↑ ↑	↑ ↑ ↑	↑	↑
M. hominis	–	↑	↑	↑
U. urealyticum	–	↑	↑	↑
U. parvum	–	↑	↑ ↑	↑

In European-American women, E. coli, G. vaginalis, and LPS significantly increased IL-1β production by cultured membranes (P < 0.001 relative to unstimulated controls). U. urealyticum also tended to increase IL-β by cultured membranes (3.3-fold increase), but results did not reach statistical significance (P = 0.095). Cultures from African-American women also had increased production of IL-1β in response to E. coli, LPS, and G. vaginalis (P < 0.001 each compared with controls), but also had a significant, 4.1-fold, response to GBS (P = 0.009) and no apparent response to U. urealuticum (P = 0.649). However, European-Americans produced less IL-1β in response to E. coli than African-American women (P = 0.031; Fig. 1A).

TNF-α production was significantly increased by E. coli (P < 0.001), G. vaginalis (P < 0.001), LPS (P < 0.001), M. hominis (P = 0.048), P. gingivalis (P < 0.001), U. parvum (P = 0.001), and U. urealyticum (P = 0.004) for cultures established from African-American women (Fig. 1B). European-American women had a similar trend in bacteria responsiveness, but results for M. hominis did not reach statistical significance (P = 0.151). All pathogens (P < 0.001, P < 0.001, P < 0.001, P = 0.002, P = 0.048, P < 0.001, and P < 0.001 for E. coli, G. vaginalis, LPS, M. hominis, P. gingivalis, U. parvum, and U. urealyticum respectively) except for GBS (P = 0.710) increased IL-10 production for cultures established from European-American women (Fig. 1C). A similar trend was observed for cultures from African-American women (P < 0.001 for E. coli, G. vaginalis, LPS, U. parvum, and U. urealyticum ; P = 0.025 for M. hominis ), but results failed to reach statistical significance for P. gingivalis (P = 0.132).

Lipopolysaccharide-stimulated IL-8 production was greater from membranes from European-Americans than from those from African-American women (P = 0.026; Fig. 1D), but both displayed significant responses to each of the pathogens. No other statistically significant differences in bacteria-stimulated cytokine production between African-American and European-American women were detected. These data are consistent with our previously reported data using the LPS model of fetal membrane infection (Menon et al., 2006).

3.3. Racial disparity in the effects of AF on cytokine production

Racial disparity in pathogen-specific, AF-mediated immune response was evident after stimulation with pathogens associated with PTB. E. coli-stimulated IL-1β production was inhibited by AF in cultures from African-Americans (P = 0.011). Results did not reach statistical significance for European-Americans (P = 0.129; Fig. 2A), however. Although no effect of AF on E. coli-stimulated TNF-α, IL-8 or IL-2 production was detected, AF inhibited E. coli-stimulated IL-10 production for both races (P = 0.006 and 0.009 for African-American and European-American women respectively; Fig. 3A).

Fig. 2.

Racial disparity in the effect of amniotic fluid (AF) on IL-1β (A) and TNF-α (B) production by fetal membranes. Shown on a log-scale are least-squares means ± SEM for membrane cultures stimulated with medium alone (CTL), E. coli (EC), Group B streptococci (GBS), G. vaginalis (GV), lipopolysaccharide (LPS), M. hominis (MH), P. gingivalis (PG), U. parvum (UP) or U. urealyticum (UU) stratified by race. Pairs of bars marked with asterisks indicate statistically significant effects of AF at *P ≤ 0.05 or **P ≤ 0.01 for that race and bacterial stimulation.

Fig. 3.

Racial disparity for the effect of amniotic fluid (AF) on IL-10 (A) and IL-8 (B) production by fetal membranes. Shown are least-squares means ± SEM for membrane cultures stimulated with medium alone (CTL), E. coli (EC), Group B streptococci (GBS), G. vaginalis (GV), lipopolysaccharide (LPS), M. hominis (MH), P. gingivalis (PG), U. parvum (UP) or U. urealyticum (UU) stratified by race. Bars marked with asterisks indicate statistically significant effects of AF at *P 0.05 ≤ or **P ≤ 0.01 for that race and bacterial stimulation.

Lipopolysaccharide-stimulated IL-1β production was significantly inhibited by AF from African-American membranes (P = 0.008), but not European-Americans (P = 0.667; Fig. 2A). In contrast, AF increased TNF-α production by LPS-stimulated fetal membranes from African-Americans (P = 0.017), but had no effect on LPS-stimulated fetal membranes from European-Americans (P = 0.452; Fig. 2B).

Amniotic fluid significantly increased P. gingivalis-stimulated IL-1β production by cultures from African-Americans (P = 0.005), but not European-Americans (P = 0.494; Fig. 2A). G. vaginalis-stimulated IL-1β production was significantly enhanced by AF for cultures established from African-Americans (P = 0.005), but not European-Americans (P = 0.111; Fig. 2A). In contrast, AF increased IL-10 production by G. vaginalis-stimulated cultures from European-Americans (P = 0.030), but not African-Americans (P = 0.247; Fig. 3A). TNF-α production for G. vaginalis-stimulated cultures, however, was enhanced for membranes from both races (P = 0.022 and P = 0.023; African-American and European-Americans respectively; Fig. 3B).

M. hominis-stimulated proinflammatory IL-1β, TNF-α, and IL-8 production was unaffected by AF for membranes from either race. IL-10 production, however, was significantly enhanced by AF for M. hominis-stimulated cultures from African-Americans (P = 0.003), but not from European-Americans (P = 0.172; Fig. 3A). AF increased both IL-1β (P = 0.014) and IL-10 (P = 0.031) production for U. urealyticum-stimulated cultures from European-Americans, but not from African-Americans (P = 0.630 and 0.722 for IL-1β and IL-10; Figs. 2A and 3A). Although TNF-α production was unaffected by AF for U. parvum-stimulated membranes from African-Americans (P = 0.298), AF significantly enhanced TNF-α production by U. parvum-stimulated membranes from European-Americans (P = 0.002; Fig. 2B). AF inhibited U. parvum-stimulated IL-10 production for membranes from African-Americans (P = 0.007), but had no effect on U. parvum-stimulated membranes from European-Americans (P = 0.670; Fig. 3A). No effects of AF were detected on GBS-stimulated cytokine production for membranes from either race (Figs. 4 and 5).

Fig. 4.

Racial disparity in the effect of amniotic fluid on IL-1b (A) and TNF-α (B) production by fetal membranes. Shown are least-squares means ± SEM for membrane cultures stimulated with medium alone (CTL), or cultures stimulated with E. coli (EC), Group B Streptococci (GBS), G. vaginalis (GV)-, Lippopolysaccharide (LPS), M. hominis (MH), P. gingivalis (PG), U. parvum (UP) or U. urealyaticum (UU) stratified by race. Pairs of bars marked with asterisks indicate statistically significant effects of amniotic fluid at *P < 0.05 or **P < 0.01 for that race and bacterial stimulation.

Fig. 5.

Racial disparity for the effect of amniotic fluid on IL-8 (A) and IL-10 (B) production by fetal membranes. Shown are least-squares means ± SEM for membrane cultures stimulated with medium alone (CTL), or cultures stimulated with E. coli (EC), Group B Streptococci (GBS), G. vaginalis (GV), Lippopolysaccharide (LPS), M. hominis (MH), P. gingivalis (PG), U. parvum (UP) or U. urealyaticum (UU) stratified by race. Bars marked with asterisks indicate statistically significant effects of amniotic fluid at *P < 0.05 or **P < 0.01 for that race and bacterial stimulation.

4. Discussion

Infection, infection-associated inflammatory response and inflammatory mediator induced initiation of uterotonic activities have been well-described in PTB and pPROM (Goldenberg et al., 2008; Romero et al., 2007). Recent findings from our laboratory and others, however, have started to question the universality of the signals that initiate the preterm labor process.

Fetal membranes exhibited differential cytokine responses to seven intra-amniotic pathogens. Stimulation of cytokines by LPS validated the model and the stimulatory capabilities of the membranes in culture. All pathogens significantly increased IL-8 production, but none had any effect on IL-2 production compared with unstimulated controls. This suggests that all of the bacterial preparations used for this experiment were biologically active and that immunity to these pathogens by fetal membranes is not mediated through classical T-cells that require IL-2 for proliferation. The high levels of IL-8 production by the cultured explants in the absence of infection supports our previous finding that IL-8 is not produced by term, not-in-labor membranes (Fortunato et al., 1995). However, cell culture conditions such as removal of placental progesterone or other immunomodulators induce constitutive production of this cytokine. The constitutive nature of IL-8 explains host inflammatory response in vitro and possibly in utero during infection where this chemokine prepares the intrauterine tissues for the host response to any adverse condition and its unlikely role as a uterotonin.

Consistent with our previous study (Menon et al., 2009b), LPS, E. coli, and G. vaginalis had the most proinflammatory activity and were potent inducers of IL-1β, TNF-α and IL-10. In contrast, genital mycoplasmas U. parvum, U. urealyticum, and M. hominis provoked only marginal increases in TNF-α that were likely balanced by increased production of the anti-inflammatory cytokine IL-10. This is also consistent with a previous study where we studied the effect of U. parvum on cytokine production by fetal membranes (Menon et al., 2009b). Previous studies have demonstrated that Mycoplasma spp. can shift host immunity toward T-helper-2 type responses to favor their survival (Romero-Rojas et al., 2001; Kang et al., 2011). Genital mycoplasmas may do similar things in the reproductive tract by stimulating only slight TNF-α and no IL-1β production, but copious amounts of IL-10. This may contribute to their adaptation as natural pathogens of the reproductive tract.

The immune responses to GBS and the periodontal pathogen, P. gingivalis, were limited to one cytokine each (IL-1β and TNF-α respectively). Previous studies have demonstrated that systemic exposure to P. gingivalis results in vertical transmission of the organism to the placental tissues (Boggess et al., 2005a; Belanger et al., 2008). In rabbits infected with P. gingivalis, no effect of infection on placental production of IL-1β was detected, consistent with our results (Boggess et al., 2005b).

Racial disparities in responses to pathogens associated with PTB are summarized in Table 1. African-Americans produced significantly more IL-1β in response to E. coli than European-Americans. This is consistent with previous studies by our lab where we found that AF IL-1β is higher in African-Americans with PTB, compared with European-Americans (Menon et al., 2007). Although results did not reach statistical significance, LPS-stimulated TNF-α production was greater for membrane cultures prepared from European-Americans than African-Americans and corresponds with a previous study by our group (Fortunato et al., 2004). Although the slight increase in IL-8 production by LPS-stimulated membranes from European-Americans compared with African-Americans is unlikely to be biologically significant, this result is similar to our previous report where we found increased production of IL-8 in AF from PTB cases in European-Americans, but not African-Americans (Menon et al., 2007).

We found that AF has very dramatic effects on cytokine production by the fetal membranes, but that it varied between races and pathogens and cannot be considered globally pro- or anti-inflammatory (Table 2). For example, AF tended to inhibit the cytokine production by E. coli-stimulated membranes, but promoted G. vaginalis-stimulated cytokine responses. This conflicts with previous studies that reported immunosuppressive effects of AF (Etlinger and Chiller, 1977; Krco et al., 1979; Wajner et al., 1986; Yoshimura et al., 1991). Those studies, however, focused on transplantation immunity and T-cell-type immune responses. Recently Witkin et al. (2012) reported that mid-trimester AF increases production of IL-6, IL-10, TNF-α, and MCP-1 by cultured peripheral blood mononuclear cells (PBML). However, we detected no effects of AF on cytokine production by unstimulated membranes, although we did find that AF significantly suppressed LPS-stimulated IL-1β production and enhanced TNF-α production by cultures established from African-Americans, but not European-Americans. This suggests that the immunomodulatory effect of AF is likely dependent on race/ethnicity of the individual, tissue/cell type, and stimulant.

Table 2.

Effect of autologous AF on bacteria-stimulated cytokine production by membranes from African-Americans (AA) and Caucasians (C). Arrows indicate a significant (P<0.05-one arrow or P<0.01-two arrows) effect of amniotic fluid on cytokine production for a given race-pathogen combination.

Pathogen	IL-1β		TNF-α		IL-10		IL-8
	AA	C	AA	C	AA	C	AA	C
AF + Medium	–	–	–	–	–	–	–	–
E. coli	↓	–	–	–	↓ ↓	↓ ↓	–	–
GBS	–	–	–	–	–	–	–	–
G. vaginalis	↑ ↑	–	↑	↑	↑	↑	–	–
LPS		–	↑	–	–	–	–	↑
M. hominis	–	–	–	–	↑ ↑	–	–	–
P. gingivalis	↑ ↑	–	–	–	–		–	–
U. urealyticum	–	↑ ↑	–	–	–	↑	–	–
U. parvum	–	–	–	↑	↓ ↓	–	–	–

Further studies are needed to identify the immunomodulatory factors in AF. Soluble factors such as sCD14, which have been reported (Gardella et al., 2001; Espinoza et al., 2002) may enhance immunity to Gram-negative pathogens by enhancing the formation of TLR-4-MD2/MyD88 complexes on the surface of immune and epithelial cells. Other factors such as H2A and H2B proteins that are also in AF may antagonize Gram-negative bacteria by binding to the lipid A component of LPS (Kim et al., 2002). Another mediator, surfactant protein A, is also present in high concentrations in AF (Shimizu et al., 1989; Miyamura et al., 1994) and can enhance LPS-stimulated cytokine production by monocytes (Song and Phelps, 2000a, 2000b). Gelsolin and IL-23 binding protein may help to suppress immune responses by binding to LPS and inhibiting neutrophil responses to bacteria respectively (Witkin et al., 2011). Although the specific factors in AF are unknown and may vary based on maternal exposure to a multitude of environmental factors, these results suggest that supplementing culture medium with autologous AF may be needed in order to adequately model immune responses at the maternal–fetal interface. It is also possible that racial differences in the content of these immunomodulators contribute to the racial disparity in infection-mediated preterm birth and pPROM.

Our study has a number of strengths. First, we used membranes from patients undergoing elective Cesarean sections who were not in labor. This ensures that the samples were collected under sterile conditions in the operating room and excludes the possibility that the vaginal flora or cytokines produced in response to labor are affecting our results. Another strength of the model system is that by culturing discs cut out from the membranes we are able to avoid artifacts of processing tissues that are problematic with cultures of isolated amniotic or chorionic cells. We are also able to keep the three-dimensional architecture of the tissue intact and preserve the cell-to-cell contacts that comprise the paracrine environment. Since only cross-sections of membranes measuring 6 mm in diameter are used for each culture, each patient provides sufficient material to apply all of the treatments to cultures derived from her donated tissues. This enables us to use each patient as her own control in the statistical analyses, greatly reducing the patient-to-patient variability that can confound these types of experiments. We also quantified an array of cytokines with definitive roles in infection-mediated preterm birth. This permits us to derive a cytokine signature for each pathogen.

Our findings are limited by a number of factors that are inherent to cell culture experiments. Membranes were exposed to relatively high quantities of heat-killed pathogens for only a limited period of time (24 h) and the model system only permits study of local immune responses. In pregnancies complicated by ascending infections, there may be continual exposure to lower quantities of pathogens over a longer period of time and there may be recruitment of immune cells from distant sites that could contribute to the cytokine milieu. We are also limited by the use of heat-killed pathogens because live bacteria would quickly overgrow the cultures and greatly disrupt the tissue culture system. Some of these organisms may contain thermal-labile proinflammatory components and heating may have reduced their ability to stimulate cytokine production. It is possible to address this limitation by using detergent extracts of the organisms; however, the host response to a growing population of bacteria may still differ from simple exposure to their biochemical components. It is also unclear what impact maternal hormones such as progesterone and estradiol may have had on the immune response by the fetal membranes to bacterial infections. The use of human tissues, also limits us to term pregnancies when the tissues would normally be discarded. Gestational tissues from pregnancies at earlier ages may respond differently to pathogens associated with preterm birth. Tissues from pregnancies ending in preterm birth, however, would be likely overrun by infection and inflammation and many other factors that are downstream (such as stress, drugs administered, and pPROM) from the host–pathogen interactions that were studied in this experiment.

Within-race variability may have made it more difficult to detect differences between races. Power analyses based on data acquired from an earlier study (Fortunato et al., 1995), however, suggested that only 5 women would be needed in each group to detect the previously observed difference of 183 pg/ml IL-1β production between races as statistically significant with 80% power. Therefore, we used 5–6 women for each group in the present study.

Another limitation is that we used immunoassays to measure chemical amounts of cytokine as an index of the host response. Biological activity of cytokines such as TNF-α can be modulated by the presence of soluble receptors, which are present in AF and may vary between races (Fortunato et al., 2004). Although they suffer from reproducibility and sensitivity, further work with bioassays to quantify cytokine levels and general inflammation may help to overcome this issue in future experiments. Animal model studies may also be useful for quantifying how pathogens may differentially regulate cytokine production in vivo. However, few of the pathogens explored above are natural pathogens of rodents that are commonly used to model preterm birth. As most PTB are associated with polymicrobial etiology, the immune response tested and demonstrated here may not mimic the ideal intraamniotic infection associated with PTB. Our ongoing work will explain immune mechanisms associated with such conditions.

In summary, this study demonstrates that each pathogen associated with PTB and pPROM is marked by its own biochemical signature. AF has potent immunomodulatory properties, but the activity differs between races and pathogens. Supplementation of culture medium may be needed to adequately model immune responses to pathogens by fetal membranes with in vitro studies.