Materials and Methods
Subjects
The gravidae provided consent from 2009 to 2012 under a protocol approved by the Institutional Review Boards (IRBs) of Good Samaritan Hospital, Cincinnati, Ohio (09105-09-067) and Cincinnati Children’s Hospital Medical Center (2009-0236); subsequent IRB approval for metagenomics studies was obtained from the Baylor College of Medicine IRB (H-27393). Informed consent was requested from mothers admitted for imminent preterm delivery. Inclusion criteria for the preterm cohort were mothers delivering between 32 0 and 36 6 weeks’ gestation because of preterm premature rupture of membranes, spontaneous preterm labor, or clinically diagnosed chorioamnionitis. The inclusion criteria for the term infants were women presenting with spontaneous labor at ≥38 weeks of gestation. We excluded mothers delivering infants for medical indications, such as maternal hypertensive disorders including preeclampsia and HELLP syndrome, because these disorders would be expected to have a very low incidence of chorioamnionitis and therefore skew the comparisons. Only 3 subjects (Preterm 381, Term 225, and Term 239) had findings of incidental hypertension, which was not an indication for induction of labor or delivery. Also excluded were the following: gravidae with HIV; delivered infants with acute, life-threatening illness or requiring extensive resuscitation; or fetuses with congenital malformations or infections, such as syphilis or cytomegalovirus. Maternal and neonatal demographics were collected by interview and chart review.
Sample collection
All sample collection was performed with the use of standard operating procedures and followed a strict uniform protocol established before study initiation. Nurses specifically trained in study procedures performed sample collection. After delivery of the infant and before delivery of the placenta, cord blood was collected from the umbilical vein with a sterile ViaCord collection kit (ViaCord, Waltham, MA) containing citrate phosphate dextrose anticoagulant. Placenta was delivered by standard obstetrical practice and immediately collected in sterile bags. These were kept in a refrigerator until sampling. Placental sampling was performed in a dedicated room via strict sterile precautions to prevent exogenous contamination. Placenta was kept with the fetal surface facing the operator. The amnion surface was cleaned by swabbing with 70% ethanol and drying immediately. Then, the surface of the amnion was cut with sterile instruments. A second set of sterile instruments was used to perform a blunt dissection to create a pocket on the underside of the amnion without puncturing the amnion. Swabs for microbial collection were swirled to collect samples from fetal chorion and/or villous placental membranes adjacent to the fetal side while taking care to avoid contamination from the maternal side.
The following swabs were collected and subsequently cryofrozen until analyses: (1) Dry Dacron swab for DNA collection (part 220115; BD, Franklin Lakes, NJ), (2) UP transport medium (clear media, part 220221; BD), and (3) Port-A-Cul (for anaerobic and aerobic culture; pink media, part 221607; BD). This collection method differs from our previous study involving the placental tissue microbiome. In our previous study, placental parenchyma tissue was collected 4 cm from the cord insertion site. In our current study, we examined the placental membrane microbiome in association with chorioamnionitis. The niche site is ideal for examining associations between the placental microbiome and inflammation because of the presence of decidual leukocytes.
Histology
The placenta and umbilical cord were sampled from specific tissue locations via an aseptic-standardized protocol for histologic examination. Sections of chorioamnion, umbilical cord, and placental tissues were stained with hematoxylin and eosin; all 32-36 6 week preterm and term tissue samples were scored for histologic chorioamnionitis based on Redline’s criteria by authors T.G. and S.G.K., who were blinded to the clinical history and placental pathology reports. Redline’s criteria determines the stage of chorioamnionitis on the basis of the depth of plane of neutrophil infiltration and the grade as the degree of neutrophilic infiltration. We classified chorioamnionitis as mild or severe on the basis maternal inflammation. Mild was denoted as stage 1 and grade 1 chorioamnionitis, and severe was denoted >stage 1 and/or >grade 1 chorioamnionitis. Funisitis indicating fetal inflammation was defined as neutrophilic infiltration surrounding the umbilical cord vessels or into Wharton’s jelly. Neonates with chorioamnionitis were compared with neonates born preterm with and without chorioamnionitis.
DNA isolation and whole-genome shotgun (WGS) sequencing
Using manufacturer’s instructions, we extracted DNA from the Dacron swab with the Power Soil DNA isolation Kit (cat. no. G-3246-100; MO BIO Laboratories Inc, Carlsbad, CA). This DNA was subsequently used for microbiome analyses that used WGS sequencing as previously described. Resulting sequences were quality filtered, and contaminating host genomic DNA (>99%) was removed with the use of Best Match Tagger (BMTagger; ftp://ftp.ncbi.nlm.nih.gov/pub/agarwala/bmtagger/ ) via a heuristic approach. After the removal of contaminating host genomic DNA, sequences were uploaded to MG-RAST (ie, Metagenomic Rapid Annotations using Subsystems Technology; http://metagenomics.anl.gov/ ) for taxonomic and pathway identification using the M5NR database with a maximum e -value of 1, minimum percent identity cutoff of 50%, and minimum length alignment cutoff of 15. Taxonomy and pathway abundance were normalized by the number of reads per sample. QIIME (ie, Quantitative Insights Into Microbial Ecology; http://qiime.org/ ) was used for the generation of principal coordinates analysis (PCoA) plots using Canberra and Bray-Curtis distance. LEfSe and Wilcoxon rank sum test were used to identify the pathway with differential abundance between cohorts. Reconstructed metabolic pathways were correlated to taxonomic abundance via the use of R. Those with significant differences in association with chorioamnionitis were examined further.
Bacterial cultures
The presence of Ureaplasma or Mycoplasma species was determined by both traditional culture methods and by polymerase chain reaction optimized to detect all the serovars.
Cytokine/chemokine concentration
Cytokine/chemokine concentrations from the plasma separated from umbilical cord blood were determined by Luminex using MILLIPLEX MAP Human Cytokine/Chemokine Magnetic Bead Panel (Millipore, Billerica, MA). Concentrations were calculated from standard curves using recombinant proteins and expressed in pg/mL.
Statistical analysis
The statistical significance of beta diversity displayed by PCoA plots was determined by permutational multivariate analysis of variance (PERMANOVA) and permutational analysis of multivariate dispersion (PERMDISP) using QIIME. Both PERMAVONA and PERMDISP were used to determine clustering between cohorts (PERMANOVA) and to determine the clustering within a cohort (PERMDISP). Statistical analysis determined by t test or 1-way analysis of variance (ANOVA) was generated using GraphPad Prism (La Jolla, CA) software.
Results
Chorioamnionitis diagnosis and subject demographics
Term subjects were enrolled on the basis of a spontaneous birth ≥38 weeks of gestation, and preterm subjects were enrolled if they spontaneously delivered between 32 0 and 36 6 weeks of gestation. Late preterm subjects were the focus of this study in an attempt to further separate inflammation from infectious, pathogenic agents. By design, our study allowed for a week between term and preterm subjects, resulting in significant differences in the gestational age ( P < .0001) and neonatal birthweight ( P < .0001) between term and preterm cohorts ( t test; Table 1 ). After enrollment and sample collection, placental tissue underwent histologic analysis with hematoxylin and eosin staining. Placentas were scored blindly on the basis of Redline’s criteria. On the basis of the histologic scores, we had the following 2 term cohorts: spontaneous term birth (cohort 1) and spontaneous term birth with chorioamnionitis (cohort 2). We also had the following spontaneous PTB cohorts: PTB (cohort 3), PTB with mild chorioamnionitis (cohort 4), PTB with severe chorioamnionitis (cohort 5), and PTB with severe chorioamnionitis and funisitis (cohort 6). In subjects with histologic chorioamnionitis, we were able to detect significant leukocyte infiltration into the chorion ( Figure 1 , A). In addition, we found that inflammatory cytokines were increased in the cord blood of subjects with the most severe chorioamnionitis ( Figure 1 , B), which is consistent with previous studies examining inflammation and chorioamnionitis. Overall, these data are consistent with long-standing findings by others demonstrating pathologic inflammation occurring in the placenta of subjects with severe histologic chorioamnionitis.
| Cohort | 1 (n = 15) | 2 (n = 12) | 3 (n = 13) | 4 (n = 11) | 5 (n = 9) | 6 (n = 11) |
|---|---|---|---|---|---|---|
| Gestational age, weeks (days) | Term | Term | Preterm | Preterm | Preterm | Preterm |
| Average | 39 (5) | 39 (6) | 35 (1) | 35 (5) | 35 (1) | 34 (6) |
| Range | 38 (1)−41 (3) | 38 (4)−41 (1) | 32 (3)−36 (3) | 32 (4)−36 (3) | 32 (6)−36 (6) | 32 (0)−36 (6) |
| Birthweight, g | 3283.9 ± 351.6 | 3359.3 ± 376.8 | 2725.0 ± 362.7 | 2950.5 ± 322.1 | 2552.5 ± 568.1 | 2425.5 ± 425.9 |
| Betamethasone | 0% (0) | 0% (0) | 23.1% (3) | 18.2% (2) | 55.6% (5) | 27.3% (3) |
| Antibiotics (72 hours predelivery) | 20% (3) | 25% (3) | 76.9% (10) | 54.5% (6) | 55.6% (5) | 72.7% (8) |
| Chorioamnionitis (histologic) | No | Yes | No | Mild | Severe | Severe |
| Funisitis (histologic) | No | No | No | No | No | Yes |
| Ureaplasma positive | 0% (0) | 16.7% (2) | 15.4% (2) | 9.1% (1) | 22.2% (2) | 27.3% (3) |
| Mode of delivery | ||||||
| Vaginal | 73.3% (11) | 75% (9) | 84.6% (11) | 100% (11) | 77.8% (7) | 72.7% (8) |
| Cesarean | 26.7% (4) | 25% (3) | 15.4% (2) | 0% (0) | 22.2% (2) | 27.3% (3) |
| Race/ethnicity | ||||||
| Non-Hispanic white | 80% (12) | 75% (9) | 69.2% (9) | 63.6% (7) | 55.6% (5) | 36.4% (4) |
| African-American/black | 20% (3) | 8.3% (1) | 15.4% (2) | 36.4% (4) | 33.3% (3) | 63.6% (7) |
| Asian | − | 8.3% (1) | − | − | − | − |
| Other | − | 8.3% (1) | 15.4% (2) | − | 11.1% (1) | − |
Subject characteristics and treatment summaries are provided in Table 1 . Few neonatal comorbidities were seen in each cohort, with the most frequent comorbidities seen in subjects with severe chorioamnionitis and funisitis (cohort 6). The most prevalent neonatal comorbidity was respiratory distress with 15.4% (n = 2) for preterm subjects without chorioamnionitis (cohort 3), 27.3% (n = 3) for preterm subjects with mild chorioamnionitis (cohort 4), 22.2% (n = 2) for preterm subjects with severe chorioamnionitis (cohort 5), and 54.5% (n = 6) for preterm subjects with severe chorioamnionitis and funisitis (cohort 6). High-frequency ventilation was used on 22.2% (n = 2) of neonates born to preterm subjects with severe chorioamnionitis (cohort 5). Preterm subjects with severe chorioamnionitis and funisitis (cohort 6) was the only cohort to have neonates with pneumonia (9.1%; n = 1), early-onset sepsis (9.1%; n = 1), or malformations (9.1%; n = 1). Intriguingly, preterm subjects without histologic chorioamnionitis (cohort 3) had 2 subjects that test positive for Ureaplasma species using culture-based and polymerase chain reaction methodologies. Thus, all cohorts with histologic chorioamnionitis had subjects with cultures positive for Ureaplasma species; however, not all subjects were positive for Ureaplasma . This finding is consistent with the findings in an experimental ovine model of intra-amniotic infection. These results bolster the idea of a causative agent other than the vaginal microflora and thus warranted subsequent investigation into the placental membrane microbiome.
Species-level resolution of the preterm placental microbiome with and without chorioamnionitis
DNA was isolated from placental membrane swabs and subjected to WGS sequencing. For each cohort, DNA from placental membranes of subjects in each was sequenced (n≥8 subjects per group). The resultant sequences were quality filtered and analyzed, and only those with high read yield and without concern for contamination were included in further analysis. On the basis of placental histology and the increase in inflammatory cytokines between preterm subjects with severe chorioamnionitis without and with funisitis ( Figure 1 ), we combined these groups for analysis because funisitis, which was determined by umbilical cord histology, did not appear to differentiate these 2 cohorts. Upon analyzing beta diversity (diversity between cohorts) at the species level using Canberra distance, we found significant differences in clustering by PERMANOVA ( P = .005) of our PCoA plots ( Figure 2 , A). Furthermore, on close examination of the PC1 and PC2 axis of the term cohorts, we found significant differences in both the PC1 ( P = .012 by ANOVA) and PC2 axis ( P = .007 by ANOVA; Figure 2 , A). In addition, our analysis of beta diversity on the PC1 and PC2 axis demonstrates significant differences in our term cohorts without and with chorioamnionitis ( Figure 2 , A, P = .05 by t test on PC1, P = .01 by t test on PC2).
Given this significance of difference of the placental membrane microbiome community by both preterm and chorioamnionitis, we next examined Bray-Curtis dissimilarity (beta diversity) within our term and preterm cohorts by severity of chorioamnionitis ( Figure 2 , B). We found that the term cohort with chorioamnionitis was most similar to our preterm cohort without chorioamnionitis or with mild chorioamnionitis ( Figure 2 , B). Interestingly, our term cohort without chorioamnionitis had a greater dissimilarity within subjects compared with term subjects with chorioamnionitis, preterm subjects without chorioamnionitis, and preterm subjects with mild chorioamnionitis ( Figure 2 , B). Again, we found that preterm cohorts with severe chorioamnionitis had the greatest dissimilarity within the cohort ( Figure 2 , B). When we alternately examined alpha diversity (within sample diversity), we found that preterm subjects with severe chorioamnionitis manifest as diminished species diversity (Shannon diversity index; Figure 2 , C). These data indicate that preterm subjects with chorioamnionitis have fewer bacterial constituents of their placental membrane microbiome ( Figure 2 , C). Our results are in agreement with previous gut microbiome studies that indicate that alpha diversity is decreased in association with clinical infection and histologically significant inflammation.
One of the distinct advantages of WGS metagenomics is the high resolution of bacterial taxa at the species level between cohorts. As shown in Figure 3 , A, we observed significant differences in overall taxonomy that was not individual subject dependent ( Figure 3 , B). Specifically, preterm subjects with severe chorioamnionitis had high abundances of Ureaplasma parvum , Fusobacterium nucleatum , and Streptococcus agalactiae ( Figure 3 ). This finding is in agreement with our previous study in which we found enrichment of Streptococcus species in the placental parenchyma microbiome of subjects with a remote history of antenatal infection. In addition, these data bolster the results obtained from examining Bray-Curtis dissimilarity ( Figure 2 , B) and alpha diversity ( Figure 2 , C) and are in concordance with previous studies of PTB and chorioamnionitis.
By contrast, among term subjects we found a greater abundance of Enterobacter species (gammaproteobacteria) and Lactobacillus crispatus (prevalent vaginal species) among term subjects inclusive of those with a cesarean delivery (n = 3, term cohorts). Although we have previously detected Lactobacillus species as a minor constituent of the placental microbiome, here we detected that L. crispatus has an increased relative abundance in term placentas without and with chorioamnionitis ( Figure 3 ). This result may be attributable to niche specification within the placenta. We further examined this possibility by comparing our previous placental parenchyma study with our current placental membrane study ( Supplemental Figure 1 ). We found that Escherichia coli was still prevalent within the placental membrane microbiome ( Figure 3 ), but other bacterial taxa, such as Xanthomonas campestris , Propionibacterium acnes , and Lactobacillus species, were also present in high abundance ( Supplemental Figure 1 ). However, despite some detected differences in bacterial taxa, inferred metabolic function was similar between the placental parenchyma and the placental membrane microbiomes ( Supplemental Figure 1 ).
We next wanted to examine clinical confounders to determine their impact on the placental membrane microbiome. Differences in beta diversity between cohorts were not biased by mode of delivery ( Figure 4 , P = .82 and Supplemental Figure 2 , P = .82) nor antenatal betamethasone ( Figure 4 , P = .26). In addition, term cohorts were unaltered by antibiotic use ( P = .37), as were genus level of bacterial taxa prevalent in our preterm cohorts that have been previously associated with intrauterine infection ( Table 2 ).
| Genus | No antibiotics, mean abundance (±SD) | Antibiotics, mean abundance (±SD) | P value ( t test) |
|---|---|---|---|
| Mycobacterium | 0.0084 (± 0.0049) | 0.0090 (± 0.0065) | .76 |
| Burkholderia | 0.0086 (± 0.0046) | 0.0077 (± 0.0052) | .59 |
| Mycoplasma | 0.0004 (± 0.0006) | 0.0005 (± 0.0015) | .89 |
| Ureaplasma | 0.0242 (± 0.0585) | 0.0126 (± 0.0319) | .52 |
| Fusobacterium | 0.0610 (± 0.210) | 0.0419 (± 0.191) | .79 |
| Streptococcus | 0.0451 (± 0.0903) | 0.0663 (± 0.168) | .62 |
Distinct bacterial metabolic pathways
A second advantage to WGS metagenomics is the capacity to reassemble and annotate (map) bacterial metabolic pathways by gene count. Thus, we next determined functional differences in metabolic pathways between our term and preterm subjects with and without chorioamnionitis by using the Kyoto Encyclopedia of Genes and Genomes. Because our previous analysis demonstrated that chorioamnionitis resulted in different alterations in the placental membrane microbiome between term and preterm subjects ( Figures 2 and 3 ), we first separated the examination of bacterial metabolic pathways on the basis of gestational age (term and preterm). Overall, term subjects without chorioamnionitis had Acinetobacter spp. and Streptococcus thermophilus significantly positively correlated to the functional pathways of metabolism of cofactors and vitamins ( Figure 5 ). Similarly, these pathways were correlated with L. crispatus , A. johnsonii , Fusobacterium species, and Enterobacter species in preterm subjects without chorioamnionitis ( Figure 5 ). Conversely, in term and preterm subjects with chorioamnionitis that have an increased abundance in S. thermophilus and Fusobacterium species ( Figure 3 ), we detected alterations in the biosynthesis of secondary metabolites and in lipid metabolism associated with these oral commensal bacteria ( Figure 5 ).
When examining term cohorts without and with chorioamnionitis, we found significant decreases ( P < .05) in the following metabolic pathways: amino sugar and nucleotide sugar metabolism, butanoate metabolism, riboflavin metabolism, and aminobenzoate degradation ( Figure 6 , A). The butanoate metabolism is intriguing because butyrate has been shown to reduce inflammation in the intestine. In addition, a decrease in riboflavin metabolism also has been associated with inflammation. Thus, decreases in butyrate and riboflavin may explain the inflammation present in the placenta of term subjects with chorioamnionitis.
