Approximately half of all placentas delivered before the second trimester and 41% of those delivered by Caesarean section harbor organisms in their chorions, detectable by culture techniques (12). The high colonization rate decreases with increasing gestational age. Some of these microorganisms are typically part of the vaginal microflora that might have evolved mechanisms to colonize the placenta during early pregnancy. We are not aware of studies that evaluated to what extent these organisms induce systemic inflammatory responses in the fetus and newborn. In an effort to explore the relationships among specific groups of microorganisms recovered from the placenta and blood protein indicators of a fetal inflammatory response, we analyzed 25 protein biomarkers in dry blood spots obtained from 527 newborns born by Caesarean section in the 23rd to 27th gestation weeks. The effects of other variables, e.g., placental histology, pregnancy complications, and postnatal exposures, on the inflammatory responses in the same study population have been addressed elsewhere (13, 14; T. F. McElrath, R. N. Fichorova, E. N. Allred, J. L. Hecht, M. A. Ismail, H. Yuan, and A. Leviton, submitted for publication).

Of the 527 placentas studied, 313 were culture negative, and 214 were culture positive for one or more microorganisms of the groups shown in Table 1. Lactobacilli (detected alone or mixed with other microorganisms in 21 cases or in 9.8% of all culture-positive placentas) were associated with a prominently low risk of systemic inflammatory reaction in the preterm newborn in contrast to the groups of pathogenic microorganisms and organisms characteristic of bacterial vaginosis (BV), which were all associated with increased risks of specific sets of elevated proteins compared to placentas that did not harbor any microorganisms (Table 1). The group of BV organisms was most frequently detected, with Gardnerella vaginalis, Prevotella bivia, anaerobic Streptococcus, and Peptostreptococcus present alone or in mixed culture in 47% or 22% of all culture-positive placentas, respectively. The BV organisms were associated with a heightened proinflammatory pattern similar to those of the infectious facultative anaerobes Escherichia coli and alpha Streptococcus (each detectable in 28% or 13% of the culture-positive placentas, respectively) and distinct from those of the various genital mycoplasma species (collectively detected in 38% or 18% of the culture-positive placentas, respectively). The mixed BV microorganisms and the infectious facultative anaerobes shared an identical pattern of highest odds of elevated levels of interleukin-1β (IL-1β), E-selectin, and serum amyloid A (SAA). In addition, the organisms within these two groups variably shared odds for elevated levels of IL-6, tumor necrosis factor alpha (TNF-α), TNF receptors, IL-8, intercellular adhesion molecule 1 (ICAM-1), ICAM-3, C-reactive protein (CRP), myeloperoxidase (MPO), vascular endothelial growth factor (VEGF), and VEGF receptor 2 (VEGF-R2). In total, the proinflammatory protein profile of the cases of mixed BV culture included 9 of the 25 proteins studied and was not duplicated by any single BV microorganism, suggesting that the proinflammatory effect should be attributable to the cumulative effects of the BV community rather than that of an individual microbial species.

When BV-associated microorganisms were evaluated separately, each had its own profile of elevated concentrations of proteins in the newborn blood samples (Table 1). This analysis included all cases with a specific microorganism recovered (including monocultures and mixed cultures) versus those that did not have any microorganism recovered. While the presence of Prevotella was associated with a relatively low risk of inflammatory response, Gardnerella was marked by an increased risk of TNF-α, IL-8, ICAM-1, and VEGF-R2, anaerobic Streptococcus was associated with a pattern of elevated IL-1β, IL-6, TNF receptor type 1 (TNF-R1), TNF-R2, and E-selectin, and Peptostreptococcus was characterized by a high risk of elevated macrophage inflammatory protein 1β (MIP-1β), ICAM-3, matrix metalloproteinase 9 (MMP-9), MPO, and VEGF and undetectable insulin growth factor binding protein 1 (IGF-BP-1).

Cases with cultured genital mycoplasma species (including Ureaplasma urealyticum and other Mycoplasma spp.) showed a reproducible inflammatory protein pattern that closely resembled only that of Peptostreptococcus and was dominated by increased odds of elevated MIP-1β, I-TAC (interferon-inducible T cell alpha-chemoattractant), ICAM-3, MMP-9, MPO, and VEGF and decreased odds of elevated VEGF-R1 and IGF-BP-1. Interestingly, in contrast to the BV-mixed group and the facultative anaerobes E. coli and alpha Streptococcus, they were not associated with elevated proinflammatory cytokines or the hepatic markers of systemic inflammation CRP and SAA.

Skin-associated organisms (Staphylococcus and Propionibacterium species), detectable in 13 to 17% of the culture-positive placentas, were not associated with elevated or reduced concentrations of inflammatory proteins in the newborn blood samples.

The levels of IL-6 receptor (IL-6R), monocyte chemotactic protein 1 (MCP-1), MCP-4, RANTES (regulated upon activation, normal T cell expressed and [presumably] secreted), and MMP-1 in the newborns’ blood samples did not appear to be significantly affected by the presence of any type of cultural bacteria in their placentas (Table 1).

Because of the distinct low inflammatory profile of the Lactobacillus-positive cases (Table 1) and because most of them (62%) represented mixed cultures, we broke down the group into monoculture and mixed-culture Lactobacillus-positive subgroups, and we further contrasted those two subgroups with cases where BV organisms were found alone or in combination with non-BV organisms (Table 2). Again, odds ratios of having top-quartile protein concentrations were calculated by comparison to those of cases with no microorganisms recovered.

Although the number of placentas colonized by Lactobacillus alone was relatively small (n = 8), resulting in lack of statistical power, it was striking that most inflammatory proteins, including those associated with individual or mixed BV microorganisms (IL-1β, IL-6, TNF-R1, TNF-R2, IL-8, MIP-1β, ICAM-1, ICAM-3, E-selectin, MMP-9, CRP, SAA, and VEGF), were completely undetectable in placentas that harbored Lactobacillus alone (Table 2). In addition, even when mixed with others (n = 13), lactobacilli continued to be associated with undetectable ICAM-1 and VEGF, and despite the wide confidence intervals due to the overall low prevalence of Lactobacillus in the ELGAN placentas, the mixed Lactobacillus cultures remained at low odds for all studied proteins.

The opposite was observed in the group positive for BV-associated species. The cases that were exclusively BV organism positive (n = 18) represented 38% of the group. When these cases were separately analyzed, they showed an even higher risk of elevated protein concentrations, including IL-1β, IL-6, TNF-α, IL-8, ICAM-1, ICAM-3, vascular cell adhesion molecule 1 (VCAM-1), VEGF-R2, and both hepatic markers of systemic inflammation, CRP and SAA (Table 2). In contrast, the group harboring BV mixed with other microorganisms (n = 29), of which a significant part (24%) was mixed with Lactobacillus, showed lower odds for top-quartile concentrations of all cytokines, chemokines, and hepatic markers of inflammation (Table 2, last column), possibly due to the modulatory effect of Lactobacillus in the mixed cultures.

The difference between the groups of Lactobacillus-positive and -negative cases is illustrated in Fig. 1 by concentration distributions for selected proteins representing the cytokines (IL-6), chemokines (IL-8), and hepatic inflammatory markers (CRP and SAA). Although the analysis of raw concentration values could not have the nominal statistical value of the percentile odds ratio analysis (Tables 1 and 2) due to the overall low prevalence of Lactobacillus species cultured from the ELGAN placentas, the data demonstrate the contrast between the observed narrow distribution and lower levels of proteins in the Lactobacillus-positive cases.

In agreement with a widely accepted belief that BV-associated bacteria are capable of inducing preterm labor, we found that mixed BV microorganisms are associated with elevated concentrations of primary proinflammatory cytokines (IL-1β, IL-6, and TNF-α), capable of inducing acute phase reactants (CRP and SAA), chemokines (IL-8), and adhesion molecules (E-selectin, ICAM-1, and ICAM-3). IL-1β, IL-6, and TNF-α, as innate immunity mediators, play protective roles at local sites of bacterial invasion; however, when simultaneously and significantly elevated in the systemic circulation, these primary cytokines can induce chemokines, adhesion molecules, and further leukocyte activation, promoting extravasation and tissue damage of noninfected tissues as shown in studies of the central nervous system (15). The mixed BV-associated blood protein patterns were similar to those of infectious facultative anaerobes, e.g., E. coli and alpha Streptococcus, confirming the pathogenic potential of the BV microorganisms. Like E. coli and alpha Streptococcus, placental colonization with mixed BV microorganisms was associated with high odds of CRP, SAA, and MPO.

In contrast to BV, genital mycoplasmas did not induce primary cytokines and acute phase reactants but were associated with an invasive pattern of cytolytic activities mediated by MPO and MIP-1β, which increase nitric oxide production (16), and extracellular matrix degradation enzymes (MMP-9), which are major mediators of inflammatory tissue destruction. MMP-9 has a well-established role in parturition and is significantly increased in the amniotic fluid in preterm delivery (17). Mycoplasmas also showed increased levels of I-TAC, a chemokine for NK and T cells, which may be induced via Nod-1 activation by mycoplasma lipopeptides (18). Nod-1 activation selectively induces immune activation while avoiding upregulation of proinflammatory cytokines (18). The lack of primary cytokine upregulation may also be a result of some active suppressive properties of these microorganisms, since decreased expression of IL-1β and IL-8 characterizes mycoplasma-infected cells (19). Ureaplasma urealyticum and Mycoplasma species were analyzed separately because numerous studies, including ours, have implicated them in preterm birth (20–27). They seemed to contribute differently to the group pattern of elevated proteins, with Ureaplasma urealyticum being the major contributor to chemokine responses, while mixed mycoplasma colonization seemed needed for increased MMP-9 levels.

In addition to group-specific responses, this study also unveiled unique features of some individual BV-associated microorganisms. Among the BV-associated species, Prevotella is frequently found in preterm placentas with high-grade chorionic plate inflammation (28), and it was among the most prevalent BV-associated microorganisms colonizing the ELGAN placenta in our study. In vitro work has shown that some Prevotella strains can exhibit a strong invasive but weak cytokine-inducing capacity, and this combination of properties may lead to higher colonization rates and survival in the host (29). In agreement with these findings, our study showed relatively weak inflammatory responses to placental colonization with Prevotella alone. Gardnerella, the most common BV organism (30, 31), appeared to be the most inflammation-provoking organism and was associated with the highest odds of elevated TNF-α, IL-8, and ICAM-1. Peptostreptococcus, also frequently associated with high-grade chorionic plate inflammation (28) and linked to tissue destruction in the gut and oral mucosa (32, 33), showed a mycoplasma-like inflammatory protein pattern and was associated with the highest odds of tissue destructive enzymes, e.g., MMP-9 and MPO as well as MIP-1β and ICAM-3. Thus, each of the individual BV-associated microorganisms contributed uniquely to the pattern of mixed BV colonization, which demonstrated the biggest variety of elevated inflammatory proteins compared to any other group pattern. The mixed BV colonization conferred a risk of newborn inflammatory response higher than that of any of the individual BV-associated microorganisms alone, thus supporting the concept of BV as a polymicrobial pathogenic state.

This study suggests that Lactobacillus alone or in mixed placental colonization may suppress inflammatory responses in extremely preterm newborns. Our observations add to a growing body of experimental evidence indicating that lactobacilli have anti-inflammatory properties. Lactobacilli have been found to suppress pathogen attachment and proliferation in the vaginal epithelium (9). Findings from in vitro studies with nonvaginal Lactobacillus strains have suggested that lactobacilli are active modulators of signaling pathways, leading to a strain-specific repertoire of cytokine release and immune tolerance as well as downregulation of proinflammatory signaling (34–37). Intestinal lactobacilli also inhibit epithelial inflammatory responses to Salmonella lipopolysaccharide (38).

In combination or independently of placental microbial colonization, a number of other prenatal or postnatal factors may influence the inflammatory protein levels in a newborn’s blood. Since most of the blood specimens used for the analysis reported here were obtained on the day of birth, the blood protein concentrations in our analysis were not adjusted for postnatal phenomena and exposures (e.g., nutrition, therapeutic interventions, postnatal onset of anemia, sepsis, etc.); however, postnatal exposures and clinical outcomes have been analyzed in longitudinal samples from the same babies collected within the first two postnatal weeks and published elsewhere (13, 14; McElrath et al., submitted). The wide variance of protein concentrations in ELGANs with culture-negative placentas (illustrated for some proteins in Fig. 1) that exceed by logs those found in ELGANs with Lactobacillus-positive placentas and are sometimes close to those of other microorganism-positive cases raises the possibilities that part of the inflammation evident in preterm newborns may be attributable to noncultured microorganisms or noninfectious factors not addressed in this paper.

Previous publications from our group have shown that the organisms recovered from the placenta are unlikely to be contaminants (28, 39) and that recovery of these organisms provides information about the risk of preterm delivery (40) as well as structural and functional consequences of brain damage (23, 41). We have also reported that inflammation of the placental tissue, often a consequence of bacteria in the placenta, provides information about the probability of elevated concentrations of inflammation-related proteins in the newborn’s blood (13, 14). Moreover, elevated concentrations of the same inflammation-related proteins provide information about the risk of sonographically defined brain damage (42). Thus, the present study provides glue to the existing construct of clinical evidence linking the organisms in the placenta to the risk of brain damage attributable to elevated systemic inflammatory proteins in the neonate.

Evidence that intrauterine inflammatory phenomena contribute to neonatal disorders (43) prompts us to consider what we found to be biologically and not just statistically significant. Although we have found some evidence for the specificity of organism-cytokine relationships, we prefer to emphasize that no inflammatory stimulus affects only a small set of proteins. Indeed, an inflammatory stimulus is capable of influencing the expression of more than a thousand genes (44). Thus, what we measured is likely only the tip of the iceberg. Our neonatal blood specimens were collected in the intensive care nursery, almost invariably within 24 h of delivery. Because many inflammatory mediators, e.g., interleukins (45, 46) and TNF-α (47), have relatively short half-lives, we are reluctant to view what we measured after birth to be limited to intrauterine phenomena. How much of what we measured should be considered evidence of neonatal rather than fetal inflammation remains to be determined. Nevertheless, our data clearly demonstrate that maternal microorganisms are associated with systemic inflammatory patterns detectable after birth.