There is controversy about whether the amniotic fluid contains bacteria. With the use of sequencing-based methods, recent studies report that the amniotic fluid is colonized by microorganisms. However, background-contaminating DNA might lead to false-positive findings when such a low microbial biomass sample is examined.
The purpose of this study was to determine whether the midtrimester amniotic fluid of patients who subsequently had normal pregnancy outcomes contains a microbial signature.
In this prospective cohort study, 42 amniotic fluid samples were collected from 37 pregnancies (5 twin and 32 singletons) during genetic amniocentesis in the midtrimester. The subsequent pregnancy outcomes of all the participants were followed. Multiple methods were used to detect the presence of microorganisms in this study, which included cultivation, quantitative real-time polymerase chain reaction, and 16S ribosomal RNA gene sequencing. Multiple positive control samples (n=16) served as quality control samples and included 3 adult fecal samples, 4 vaginal swabs, and 9 artificial bacterial communities that were run in parallel with negative control samples (n=12) that included 4 samples from the hospital operating room and 8 samples from the laboratory, to account for background-contaminating DNA during each step of the experiments.
No bacteria under anaerobic or aerobic conditions or genital mycoplasmas were cultured from any of the amniotic fluid samples. Quantitative polymerase chain reaction did not reveal greater copy numbers of 16S ribosomal RNA gene in amniotic fluid samples than in negative control samples. 16S Ribosomal RNA gene sequencing did not indicate a significant difference in the microbial richness or community structures between amniotic fluid and negative control samples.
With multiple methods of microbiologic inquiry, no microorganisms were identified in the midtrimester amniotic fluid of healthy pregnancies with a normal pregnancy outcome.
Based on culture method for >100 years, the intraamniotic cavity traditionally has been viewed as sterile. However, this “sterile womb” paradigm has been challenged by recent studies that have been based on culture-independent sequencing techniques. The molecular evidence of unique bacterial communities could be detected in the amniotic fluid (AF), placenta, endometrium, , and meconium , of term healthy pregnancies.
Why was this study conducted?
The purpose of this study was to determine whether bacteria are present in the midtrimester amniotic fluid of patients who subsequently had a normal pregnancy outcome.
Amniotic fluid did not contain any cultivable bacteria or genital mycoplasmas. With the use of molecular microbiologic techniques (quantitative real-time polymerase chain reaction and 16S ribosomal RNA gene sequencing), there was no evidence of bacterial DNA.
What does this add to what is known?
This study demonstrates that, with the use of multiple methods of microbiologic inquiry, the midtrimester amniotic fluid of normal pregnancies does not contain bacteria.
Microbial invasion of the amniotic cavity (MIAC) that results in intraamniotic infection has been associated with adverse pregnancy outcomes, such as spontaneous preterm birth, , , preterm premature rupture of membranes, , and histological/clinical chorioamnionitis. , An ascending pathway is considered the most common route of MIAC, given the direct evidence that microorganisms that were isolated from AF of women with intraamniotic infection are present in the lower genital tract.
Considering the unavoidable occurrence of environmental or reagent contamination during the entire experimental process, , multiple rigorous experimental control samples are urgently required when such a low microbial biomass sample is studied. Several subsequent studies found the microbial profiles of AF , and placenta to be indistinguishable from negative control samples, which has reestablished the controversy regarding the “in utero colonization hypothesis.”
As pregnancy progresses, the uterus changes from an early proinflammatory condition to an antiinflammatory condition in the second trimester and then back to a proinflammatory condition before the onset of labor. The intraamniotic microbial profile might change as the cervix shortens and dilates. Nevertheless, most studies collected the samples at the time of delivery.
With the use of multiple inquiry methods that included cultivation, quantitative real-time polymerase chain reaction (qPCR), and 16S ribosomal RNA (rRNA) gene sequencing, we conducted a prospective cohort study to investigate the presence of microorganisms in the midtrimester AF of 37 women in parallel with multiple negative and positive control samples and followed their subsequent pregnancy outcomes.
Materials and Methods
This prospective cohort study was conducted in Peking University First Hospital from May 2018 to March 2019. Women who underwent amniocentesis for prenatal diagnosis between 19 and 22 weeks of gestation were recruited for the study. The exclusion criteria were (1) fetal malformation, (2) clinical infection and antibiotic treatment within 2 weeks, and (3) refusal to participate. This study was reviewed and approved by the institutional ethics committee of Peking University First Hospital (2015), and all the participants provided the written informed consent.
Ultimately, 37 women (5 with dichorionic diamniotic twin pregnancies and 32 with singleton pregnancy) were enrolled; 42 AF samples were collected by amniocentesis ( Figure 1 ). The clinical characteristics were obtained via electronic medical records, which included the history of spontaneous or in vitro fertilization (IVF) conception and pregnancy outcomes ( Supplemental Table 1 ).
To reflect the potential DNA contamination during the experimental procedure, 12 negative control samples were designed that included (1) sterile saline solution (9 mg/mL NaCl) that was collected in the hospital operating room (n=4) and served as sampling control samples; (2) DNA extraction kit buffers (n=4) that were collected in the laboratory, without AF samples but processed exactly as AF samples, that served as extraction control samples; and (3) PCR amplification reagents (n=2) and DNA-free water (n=2) without an extraction protocol that served as amplification control samples and blank control samples, respectively. Moreover, 16 positive control samples that consisted of adult stool specimens (n=3), vaginal swabs (n=4), and artificial bacterial communities (n=9) spiked into AF were used as experimental quality control samples. Artificial bacterial communities 1–7 contained various Gram-positive and Gram-negative bacteria with known numbers of colony-forming units. Artificial bacterial community 7 was serially diluted to a 10 –1 dilution (artificial bacterial community 8) and to a 10 –2 dilution (artificial bacterial community 9) for a lowest limit of detection ( Supplemental Table 2 ). Five gradients of plasmids from Escherichia coli were set up to generate the qPCR standard curve. The presence of microorganisms was determined using (1) cultivation, (2) qPCR of the 16S rRNA gene, and (3) 16S rRNA gene sequencing. In summary, 46 samples were subjected to cultivation; 64 samples were subjected to qPCR; 68 samples were subjected to 16S rRNA gene sequencing, and 39 AF samples were subjected to cytokines detection ( Figure 1 ). The detailed information is presented in Supplemental Table 3 .
Clinical definition and pregnancy outcomes
Gestational age was determined by the date of the last menstrual period and confirmed by ultrasound examination. Gravidity and parity were recorded according to the current admission status. Preterm birth is defined as birth that occurred at <37 gestational weeks and includes spontaneous and iatrogenic preterm birth. Premature rupture of membranes is defined as spontaneous rupture of membranes before the onset of labor. Preterm premature rupture of membranes occurs at <37 gestational weeks. The diagnosis of histologic chorioamnionitis is based on the presence of acute inflammatory changes in the chorionic plate and/or in the chorioamniotic membrane, as previously described. ,
Sample collection and preparation
Amniocentesis was performed in a sterile operating room, and sample preparation was carried out by the same researcher wearing sterile mask and gloves. After centrifugation at 1300 g (10 minutes, room temperature), AF supernatants were collected and divided into 5 aliquots in a biosafety cabinet (BSC-1500IIB2-X; Biobase, Jinan, China) for further analysis. Three aliquots were placed in sterile tubes and transported to the laboratory for anaerobic, aerobic, and genital mycoplasma cultures; the remaining 2 aliquots were immediately stored at –80°C for molecular sequencing and cytokine detection. Four negative control samples were collected in the same operating room according to the same sampling and aliquoting procedures.
Each aliquot of sample (2–3 mL) was injected into a BD Bactec Lytic/10 Anaerobic/F Culture Vial and a BD Bactec Plus Aerobic/F Culture Vial (Becton Dickinson and Company, Sparks, MD) with the use of a sterile syringe. The entire procedure was performed in a biological safety cabinet, and the vials were incubated in a Bactec FX Instrument (Becton Dickinson and Company) at 35°C for 5 days, according to the manufacturer’s protocol.
The genital mycoplasma cultivation assay (Mycoplasma IES; Autobio, Zhengzhou, China) was also performed under the manufacturer’s instructions. In brief, 300 μL of the AF were transferred into the reconstituted medium. Afterward, 100 μL of the suspension were inoculated into the wells of the strip. All the strips were incubated at 35°C–37°C for 24 hours; the appearance of a red color indicated a positive reaction and microbial growth.
DNA extraction was performed in a biological safety cabinet under the standard protocol for microbial analysis with a QIAamp DNA Stool Mini kit (Qiagen, Hilden, Germany). During the entire experimental process, the study personnel wore sterile laboratory coats, hairnets, face masks, and sleeves. DNA concentrations of the AF samples and control samples were measured with a Qubit 3.0 Fluorometer (Q32866; Life Technologies, Carlsbad, CA), and purified DNA was stored at –20°C.
16S rRNA qPCR
The bacterial DNA copy number was detected by TaqMan qPCR of the V3–V4 region of the 16S rRNA gene using the following: primer-F, 5’-ACTCCTAYGGGRBGCASCAGT-3’; primer-R, 5’-CCTAGCTATTACCGCGGCTGCT-3’; probe, 5’-6FAMCGGCTAACTMCGTGCCAGBHQX-3’. A 20-μL reaction contained 10 μL of Premix Taq (2x) Mix (Takara, Shiga, Japan), 1 μL of DNA, and 5 μmol of primer and probe. Amplifications were performed with an initial denaturation at 95°C for 2 minutes, followed by 40 cycles of denaturation at 94°C for 5 seconds and annealing at 60°C for 30 seconds. All reactions were performed with the use of a Roche 480 Real Time PCR instrument (Roche, Basel, Switzerland). A plasmid that contained the 16S PCR amplicon from E coli was diluted serially from 10 5 copies to 10 copies to generate a standard curve. Each sample was amplified in duplicate.
16S rRNA gene sequencing
The 16S rRNA gene V3–V4 region was chosen for Illumina sequencing (Illumina Inc, San Diego, CA) to identify the bacterial taxonomic composition by a 2-step PCR. Extracted DNA was first amplified by digital droplet PCR. Droplet generation, droplet transfer, and plate sealing were performed according to the protocol. DNA was amplified with 1x KAPA HiFi Master Mix (16SAFP02; Coyote, Beijing, China), 0.2 μmol of each primer (primer-F: 5’-CCTAYGGGRBGCASCAG-3’; primer-R: 5’-GGACTACNNGGGTATCTAAT-3’), and 9 μL of DNA. The conditions were as follows: 95°C for 3 minutes, followed by 30 cycles of denaturation at 98°C for 15 seconds, annealing at 50°C for 50 seconds, and extension at 72°C for 30 seconds, with 1 cycle at 72°C for 10 minutes. All reactions were performed with the use of a 96-well PCR instrument (Coyote), and amplification products were purified with VAHTS clean beads (Na44-02; Coyote). After attachment of barcode adapters (16SAFP03; Coyote), the second PCR was performed under the same conditions as mentioned earlier, with only 8 cycles and an increased annealing temperature of 58°C for 30 seconds. Amplicon libraries were purified with VAHTS clean beads (Na44-02) and quantified with a Qubit dsDNA HS Assay Kit (Q32851; Life Technologies). The final library was sequenced with the use of the Illumina HiSeq 2500 platform (Illumina Inc).
Multiplex bead array assay for cytokines
The AF concentrations of the following 21 cytokines were measured with an EMD Millipore Milliplex Kit (HCYTO-60K, 21X-Milliplex; Merck Millipore, Burlington, MA), according to the manufacturer’s instructions. Thirteen proinflammatory cytokines (transforming growth factor alpha, granulocyte colony-stimulating factor, interferon gamma, interleukin [IL]-12P40, IL-15, sCD40L, IL-17A, IL-1α, IL-1β, IL-2, IL-6, IL-7, and vascular endothelial growth factor), 3 antiinflammatory cytokines (fibroblast growth factor-2, IL-10, and IL-4), and 5 chemokines (fractalkine, macrophage-derived chemokine, IL-8, interferon gamma-induced protein 10, and macrophage inflammatory protein-1α) were quantified. Briefly, each well of 96-well plates was loaded in duplicate with 25 μL of assay buffer and 25 μL of standard, control, or AF supernatant. Next, 25 μL of magnetic beads were added into each well and incubated for 2 hours at room temperature; the wells were washed twice with 200 μL of wash buffer, followed by the addition of 25 μL of detection antibody. The plates were then incubated for 1 hour at room temperature, after which 25 μL of streptavidin-phycoerythrin was added to each well and incubated for 30 minutes at room temperature. The plates were washed twice more with wash buffer, and 150 μL of the drive fluid were added into each well for measurement with the use of a Luminex Magpix instrument (Thermo Scientific, Waltham, MA). Standard curves were generated, and the values of samples were calculated from the curve.
The cycle threshold values of the qPCR assay, defined as the number of thermal cycles required for the detection threshold, were converted to copy numbers according to the standard curve. Sequences of the 16S rRNA gene were clustered with the use of QIIME with 97% nucleotide similarity and taxonomic classification with the use of the Greengenes database. Paired reads were merged with flash software with a maximum of 10% allowed between the number of mismatched base pairs and the overlap length. Sequencing of DNA extracts yielded 3,884,287 sequences for AF samples; 2,149,237 sequences for positive control samples, and 3366 sequences for negative control samples. Based on 97% nucleotide similarity, the sequences clustered into 400 operational taxonomic units (OTUs) for AF samples, 2010 OTUs for positive control samples, and 13 OTUs for negative control samples. The Good’s coverage values of all but 1 AF sample exceeded 99.8%. The exception was 99.7% (S18). Good’s coverage values of all positive control samples and negative control samples exceeded 99.9%. For analyses of alpha diversity, individual sample libraries were subsampled with the use of the cutoff value of 30,000; samples with <30,000 sequences were not subsampled.
Bacterial compositions were visualized with a heat map, which was generated via Seaborn, a Python data visualization library. Alpha diversity was evaluated with the Chao1 and Simpson indexes. Beta diversity was assessed by unweighted UniFrac distance matrices and visualized by principal coordinates analysis, with 1000 permutations, and statistically calculated by the nonparametric multivariate analysis of variance methods with the use of the Adonis function in the R package vegan. The metric variable was shown as the mean ± standard deviation or median (interquartile range) and compared by Student’s t test or Mann-Whitney U test according to the normality of the data distribution. Chi-square and Fisher’s exact tests were used to compare of proportions of analytes. A probability value of <.05 was considered significant. GraphPad Prism (version 7.0; GraphPad Software, San Diego, CA) was used for the statistical and graphic analyses.
The demographic and clinical characteristics of all the patients are shown in Table 1 . Among the 37 women, 14 conceived with IVF (5 with dichorionic diamniotic twins), and 23 conceived spontaneously. One woman with twin gestation (subject ID 3) had an intrauterine fetal death of 1 twin at 22 weeks. Two (subject IDs 4 and 7) of 4 cases of preterm birth had preterm premature rupture of membranes and delivered at 36 and 30 gestational weeks, respectively; the histologic acute chorioamnionitis (stage 2, grade 2) was identified in subject ID 4. One woman (subject ID 10) delivered electively at 36 gestational weeks for gestational hypertension, and another woman (subject ID 13) delivered electively at 33 gestational weeks because of preeclampsia ( Supplemental Table 1 ).
|Age, y a||35.4±4.4|
|Gravidity b||2 (1–2)|
|Parity b||0 (0–1)|
|Gestational age at amniocentesis, wk b||21 (20–21)|
|Gestational age at delivery, wk b||39 (38–39)|
|Birthweight, g b||3235 (2743–4150)|
|Clinical indication, n (%)|
|In vitro fertilization||14 (37.8)|
|Cesarean delivery||24 (64.9)|
|Preterm birth||4 (10.8)|
|Preterm premature rupture of membranes||2 (5.4)|
|Histologic chorioamnionitis||1 (2.7)|
All the AF samples and control samples did not yield any bacteria under aerobic or anaerobic conditions, nor the growth of genital mycoplasmas.
Real-time qPCR for 16S rRNA gene copy number
To quantify the microbial biomass in AF samples, we used the qPCR assay to measure the copy number of the 16S rRNA gene. A standard curve over a range of 10–10 5 gene copies was generated by the linear regression analysis of an E coli plasmid ( Figure 2 , A; slope=–3.7; R 2 =0.99). Compared with stool samples (median, 2.4×10 8 [minimum/maximum, 2.1×10 8 /2.6×10 8 ] copies/μL) and vaginal swabs (4.5×10 7 [minimum/maximum, 1.3×10 7 /7.6×10 7 ] copies/μL), AF samples contained very low bacterial biomass (553 [minimum/maximum, 32/24106] copies/μL).
Considering the potential contamination, we assessed the 16S rRNA gene copy number in the negative control samples. Although the AF samples contained higher numbers of 16S gene copies than the extraction (22 [minimum/maximum, 4/287] copies/μL, Mann-Whitney U test; U=9; P <.01) and blank (14 [minimum/maximum, 4/22] copies/μL; U=0; P <.001) control samples, no significant difference was observed between the AF samples and the sampling control samples (566 [minimum/maximum, 453/659] copies/μL; U=79.5; P =.87; Figure 2 , B).
16S rRNA gene sequencing
As the positive control samples, the taxonomic composition of each artificial bacterial community was consistent with expectations ( Figure 3 ). The relative abundances of bacterial compositions were highly similar among artificial bacterial communities 7, 8, and 9, except for Micrococcus , which could not be measured accurately in artificial bacterial community 9 ( Figure 3 , G), which indicated the lowest detection limit. Therefore, the feasibility and reliability of the sequencing data were ensured. Except for 1 extraction control and 2 amplification control samples, we identified sequenced reads in 95.6% of the samples (65/68). Five AF samples and 3 extraction control samples were removed from further analyses because no bacterial OTU was annotated in these samples.
The bacterial richness of AF samples was significantly lower than that of stool ( Figure 4 , A; Mann-Whitney U test; Chao1 index; U=0; P <.001; 4, B, Simpson index; U=2; P <.001). However, no significant difference was found between AF samples and control samples ( Figure 4 , A, Chao1 index; U=35; P =.37; 4, B, Simpson index; U=36.5; P =.42). To assess bacterial community structure, principal coordinates analysis that was based on the unweighted UniFrac distance was performed. In general, the samples clustered according to sample types ( Figure 4 , C). Stool and vaginal swabs were distinct from AF samples (nonparametric multivariate analysis of variance, stool vs AF: F=5.547; P <.001; vaginal swabs vs AF: F=3.055; P =.001) and control samples (stool vs control samples: F=6.308; P =.022; vaginal swabs vs control samples: F=2.683; P =.027). Similarly, no significant difference in bacterial community structure was observed between the AF samples and control samples (F=1.166; P =.243).
To identify the potential bacterial OTUs that were unique to AF samples, a secondary analysis was performed to eliminate the background signals in the negative control samples. The sequenced data were filtered more stringently (<10 identical sequences were filtered) compared with the primary analyses (<5 identical sequences were filtered). Eventually, 3 stool samples and 4 vaginal swabs and 13 AF samples met this threshold and possessed at least 1 OTU.
Given the absence of bacterial OTUs at the genus level, 3 AF samples (S21, S2, S10) were excluded from further analysis. In the remaining 17 cases, 22 predominant OTUs at the genus level were identified based on a relative abundance >1% ( Figure 5 ). Among the 22 bacterial OTUs, 14 were found in AF samples, and 10 of them ( Pseudonocardia , Adhaeribacter , Dialister , Roseburia , Delftia , Sutterella , Bifidobacterium , Corynebacterium , Bdellovibrio , and Iamia ) were identified in only 1 AF sample, which suggests that they were likely contaminants rather than genuine signals. In addition, Bdellovibrio and Iamia are usually found in soil and plants, and Pseudonocardia and Dialister are usually found in soil, which are unexpected findings for the human amniotic cavity from an ecologic perspective. In contrast, each of the remaining 4 bacterial OTUs, Bacteroides , Propionibacterium , Faecalibacterium , and Ruminococcus , was found in >2 of the AF samples, which was considered ecologically plausible, because the origin might be the human vagina, gut, or skin. The 4 bacterial OTUs were identified in 9 AF samples, which included 8 (S3, S4, S7, S13, S14, S16, S18, and S19) from the IVF conception and 1 (S20) from the spontaneous conception ( Figure 5 ).
Cytokine concentrations in AF
To investigate any inflammatory response to the aforementioned bacterial signals, the cytokine profile of AF samples was further assessed. In general, all the cytokines concentrations of AF samples either with (n=7) or without (n=32) bacterial signals were both considerably low, when compared with these of patients with microbial-associated intraamniotic inflammation ( Supplemental Table 4 ). With IL-6 as an example, the AF concentration ≥2.6 ng/mL is used to define the intraamniotic inflammation. However, the median concentration of IL-6 of 39 AF samples was 105.5 pg/mL and ranged from 12.3–736.9 pg/mL, which was much lower than the cutoff value of intraamniotic inflammation.
Pregnancy outcomes of women who harbored bacterial signals
Subsequently, we investigated the pregnancy outcomes of the women who harbored the bacterial signals ( Table 2 ). Among them, 7 women conceived through IVF, and 1 woman conceived spontaneously. One woman (subject ID 3) had a single intrauterine fetal death at 22 weeks gestation. Two women underwent preterm birth: 1 woman (subject ID 10) delivered electively at 36 gestational weeks because of gestational hypertension, and the other woman (subject ID 13) delivered electively at 33 gestational weeks because of preeclampsia without any signs of inflammatory response during the histopathologic examination. With the use of criteria previously described, 1 woman (subject ID 15) was diagnosed with clinical chorioamnionitis at the time of delivery because of the presence of fever (temperature, 38°C) that was accompanied by the symptoms of tachycardia (heart rate, 106 beats/min) and leukocytosis (leukocyte count, 23,000 cells/mm 3 ), but no acute inflammatory responses were found from histopathologic examination. No adverse pregnancy outcomes were reported in remaining women.
|Subject identification||Predominant bacterial operational taxonomic units a||16S ribosomalRNA gene copies||Interleukin-6, pg/mL||IVF/spontaneous conception||Gestational age, wk||Pregnancy complications||Histologic chorioamnionitis|
|At amniocentesis||At delivery|
|3||Bacteroides (60.8%) Faecalibacterium (8.2%)||3016||47.82||IVF||20||Fetal death at 22 weeks gestationalage||No||N/A|
|3||Propionibacterium (30.1%) |
|5||Bacteroides (61.9%)||420||143.28||IVF||20||39||Gestational diabetes mellitus||N/A|
|10||Bacteroides (30.8%)||833||58.43||IVF||21||36 (Iatrogenic)||Gestational hypertension||N/A|
|13||Ruminococcus (32.8%)||1531||132.57||IVF||21||33 (Iatrogenic)||Preeclampsia||No chorioamnionitis|
|15||Bacteroides (100%)||4439||144.19||Spontaneous||21||41||Clinical chorioamnionitis b||No chorioamnionitis|
Principal findings of the study
The main findings included (1) cultivation did not yield viable bacteria or genital mycoplasmas in any AF samples, (2) qPCR did not distinguish the AF samples from the negative control samples based on 16S rRNA gene copy number, and (3) 16S rRNA gene sequencing did not reveal a difference in microbial composition or community structure between AF samples and negative control samples. The identification of Bacteroides , Propionibacterium , Faecalibacterium , and Ruminococcus against negative control samples, although intriguing, was not supported by the absence of intraamniotic inflammation in the cytokines detection. Overall, we did not find the consistent evidence that the midtrimester AF of normal pregnancy contains microorganisms.
Controversial views regarding intraamniotic microorganisms
A groundbreaking publication in 2014 claimed that the placentae from uncomplicated pregnancies harbor a unique microbiome that is similar to the microbiota of the human oral cavity according to 16S rRNA gene sequencing and metagenomics sequencing in a subset of samples. This finding stimulated a wave of research that explored the microbiota in placenta, , , , , AF, and uterus tissue , , that challenged the long-standing “sterile womb” dogma. However, gradually accumulating studies have argued that the unavoidable background DNA contamination during the experimental processes might lead to false-positive findings. Recently, Rowlands et al, with the use of species-specific and broad-range PCR techniques, found that the midtrimester AF samples were negative for the presence of bacteria. Furthermore, Lim et al failed to identify a unique microbiota in AF samples from healthy pregnancies that differs from negative control samples. In a recent study of 10 uncomplicated pregnancies with intact amniotic membranes, no significant difference was observed in the bacterial loads between AF samples and negative control samples.
In addition, using multiple inquiry methods and adequate technical control samples, Theis et al claimed no consistent evidence to support the existence of a unique placental microbiota in patients who delivered at term without labor. More recently, de Goffau et al also provided evidence that the placenta is not colonized by microorganisms in healthy pregnancies and that any bacterial signals are associated with contamination DNA and/or batch effects. Consistent with previous studies by Rowlands et al, Rehbinder et al, and Lim et al, our study included multiple technical control samples and multiple complementary methods of inquiry (bacterial culture, 16S rRNA gene qPCR, and 16S rRNA gene sequencing) and provided the robust evidence of the “sterile womb.”
Consistent with recent studies, no viable bacteria were cultured from any of AF samples. Given that a positive culture of AF samples is the sign of MIAC , and that no intraamniotic inflammatory response was identified during the cytokines detection, our study supported a sterile intraamniotic environment at midtrimester.
In line with Lim et al, the 16S rRNA gene copy numbers in AF were much lower than those in stool samples and vaginal swabs. When compared with the negative control samples, the microbial abundance in AF was similar to that of the sampling control samples but higher than that of the laboratory control samples, which indicated extra contributions of contamination during the sampling process.
16S rRNA gene sequencing
Based on 16S rRNA gene sequencing, the bacterial richness and community structure were similar between the negative control samples and AF samples, which was consistent with previous studies that revealed that the amniotic cavity is sterile in uncomplicated pregnancies. , , In the secondary analysis of the same data, 4 bacterial signals, Bacteroides , Propionibacterium , Faecalibacterium , and Ruminococcus , were identified against the negative control samples. As typical inhabitants in the human vagina, gut, or skin niches, these bacterial signals are considered to be ecologically plausible. The Bacteroides , Faecalibacterium , and Ruminococcus have been identified in endometrium , ; Propionibacterium has been identified in umbilical cord blood of a healthy newborn infant and chorioamnion samples of healthy pregnancies, which suggests the possibility that these human commensal bacteria may enter the amniotic cavity through ascending pathway or hematogenous dissemination. ,
It is noteworthy that 87.5% of the women (7/8) who harbored the 4 bacterial signals conceived through IVF ( Table 2 ), which may be associated with the different profiles of physiology, immunology, and endocrinology in the IVF pregnancies. The process of IVF treatment is complicated. Various cytokines are involved in the balance between the immunogenic resistance and tolerance. The exogenous estrogen and progesterone in IVF treatment are associated with the risk of microbial invasion, because microbes can alter and use host hormones to facilitate growth and survival. In addition, IVF women are exposed to more invasive operations, such as cervical excision procedures, hysteroscopic resection, induced abortion, and the IVF treatment itself, which may be a route for MIAC. In current study, it is hard to untangle the key contributors to the identification of 4 bacterial signals.
MIAC frequently is accompanied by the presence of high AF concentration of cytokines and chemokines, such as IL-1, IL-6, , IL-8, , interferon gamma–induced protein-10, , , and other inflammation-related proteins. IL-6 plays a key role in the diagnosis of intraamniotic inflammation, given the strong association with preterm labor and an increased rate of neonatal morbidity and death. , , IL-1 has been implicated as a signal for the onset of human parturition in the setting of infection, and the IL-1α serves as an alarmin and plays a core role in the microbial-associated intraamniotic inflammation.
Although there is a wide range of cutoffs in AF IL-6 levels for intraamniotic inflammation, , , any AF samples from our study did not reach the threshold described in previous studies, which indicates the absence of intraamniotic inflammatory response. Besides, no spontaneous preterm delivery or histopathologic chorioamnionitis was observed in women with 4 bacterial signals. Therefore, no further comparison was conducted. In our current study, we did not find sufficient evidence to support the presence of bacterial signals identified in 16S rRNA gene sequencing.
Criteria for a genuine signal
In combination with all previous reports, we listed the criteria of microorganisms in AF: (1) absence in negative control samples, (2) 16S rRNA gene copies and microbial profiles of AF are distinct from those of negative control samples, (3) the presence in at least 2 AF samples to exclude the accidental signals, (4) ecologic plausibility, (5) confirmation of cytokines detection, and (6) the use of positive control samples to clarify the lowest limit of detection.
MIAC that results in intraamniotic inflammation or infection is a risk factor for spontaneous preterm birth. , , A recent study found that the bacteria cultured from AF of women with intraamniotic infection were identified in the vagina, which provided the direct evidence of ascending infection as the primary cause of intraamniotic infection. In our study, the MIAC is absent at midtrimester in uncomplicated pregnancies. It is noteworthy that 4 ecologically plausible bacterial signals are identified in parts of our AF samples; however, the AF cytokines concentrations are far below the cutoff value for the identification of intraamniotic inflammation. Because the intraamniotic inflammation remains a key contributor to spontaneous preterm delivery, our study emphasizes the necessity of AF cytokines detection, especially when a possibly positive result appears.
The oral cavity, vagina, and gut are regarded as the high biomass sites. The amniotic cavity and placenta, in contrast, are sterile or contain low microbial biomass. Thus, the experimental conditions for testing the presence of bacteria in these tissues or biologic fluids need to be different from the approaches that have been used in high biomass niches. In this study, not only negative and positive control samples, but also artificial bacterial communities that consisted of both Gram-positive and Gram-negative bacteria were included. Gram-positive bacteria served as quality control samples, considering that the high mechanical strength of the cell wall can be affected by different extraction methods. , The consistency between the expected and detected results ensured the feasibility of the extraction method. To determine the lowest limit of detection, the serial dilutions of artificial bacterial communities with known concentrations were included. The detected bacterial compositions were consistent with expectation, although the relative abundance at the genus level was not exactly the same as that expected. Given the technical limitations of 16S rRNA gene sequencing, metagenomic sequencing is warranted for further investigations if any bacterial signals are found.
Strengths and limitations
There are several strengths of our study. First, a prospective cohort study was conducted to collect AF samples at midtrimester, hence the potential impacts of maternal-fetal complications and labor onset were avoided. Second, multiple complementary modes of inquiry and 2-step analysis of 16S rRNA sequence data were performed to provide a more robust conclusion. Finally, the thorough negative control samples were included in this study, covering each step of the experiment. To ensure the qualification accuracy, the artificial bacterial communities that contained various bacterial species with known abundances, instead of a single bacterium, were also included.
Nevertheless, the sample size was small, and it was difficult to determine the origin of bacterial signals found in AF samples. Furthermore, although the cytokine profile is investigated in our study, the results of 16S rRNA gene sequencing were not validated by metagenome sequencing.
With multiple methods of inquiry, we did not identify the presence of microorganisms in midtrimester AF from the pregnancies with a normal pregnancy outcome.
We acknowledge Mr Yichen Liu and Ms Lanying Zhang (COYOTE Medical Laboratory, Beijing, China) for contributing to the 16S rRNA gene sequencing data analysis; Dr Lingzhen Meng (Peking University First Hospital, Beijing, China) for technical support on cultivation assays; Mr Jiming Yin (Capital Medical University Affiliated Beijing You An Hospital, Beijing, China) for technical support with the multiplex bead array assay.
|Subject identification||Sample no.||Age, y||Gestation age at amniocentesis, wk||Single/twin pregnancy||IVF/spontaneous conception||Delivery mode||Cesarean indication||Gestational age at delivery, wk||Gravidity||Parity||Complication||Preterm premature rupture of membranes||Histologic condition||History of uterine surgery||Gender 1||Birthweight 1, g||Gender 2||Birthweight 2, g|
|3||S3||40||20||Twin||IVF||Vaginal||—||38||2||0||Single intrauterine fetal death||No||1||Male||2490||—||—|
|4||S5||35||23||Twin||IVF||Cesarean||PPROM||36||1||1||Gestational diabetes mellitus||Yes||Acute chorioamnionitis||0||Male||2730||Female||2540|
|6||S8||35||21||Single||IVF||Cesarean||Prolonged labor||37||1||0||Gestational diabetes mellitus||No||0||Female||2780|
|7||S9||28||23||Twin||IVF||Cesarean||PPROM||30||2||0||Short cervix, epilepsy, intrauterine infection||Yes||0||Male||1680||Male||1120|
|8||S11||40||21||Single||IVF||Cesarean||Scarred uterus||38||1||0||Gestational diabetes mellitus||No||0||Male||3590|
|10||S14||29||21||Twin||IVF||Cesarean||History of cesarean delivery||36||4||1||Gestational hypertension||No||2||Female||2800||Male||2400|
|13||S18||35||21||Single||IVF||Cesarean||Severe preeclampsia||33||1||0||Severe preeclampsia||No||No chorioamnionitis||0||Male||1460|
|15||S20||38||21||Single||Spontaneous||Cesarean||Prolonged labor||41||1||0||Clinical chorioamnionitis||No||No chorioamnionitis||0||Male||3680|
|16||S21||37||21||Single||Spontaneous||Cesarean||History of cesarean delivry||39||2||1||No||0||Male||3550|
|18||S23||38||22||Single||Spontaneous||Vaginal||—||39||4||2||Myoma of uterus||No||1||Female||2910|
|20||S25||38||21||Single||Spontaneous||Cesarean||History of cesarean delivery||38||2||1||Gestational diabetes mellitus||No||0||Female||3500|
|22||S27||40||21||Single||Spontaneous||Cesarean||History of cesarean delivery||39||4||1||Gestational diabetes mellitus||No||1||Female||3590|
|23||S28||37||22||Single||Spontaneous||Cesarean||History of cesarean delivery||39||2||1||No||0||Male||3270|
|24||S29||35||21||Single||Spontaneous||Cesarean||Prolonged labor||40||1||0||Fetal distress||No||0||Male||4150|
|32||S37||33||18||Single||Spontaneous||Cesarean||History of cesarean delivery||39||2||1||No||0||Male||3600|
|34||S39||39||19||Single||Spontaneous||Cesarean||History of cesarean delivery||39||2||1||Postpartum hemorrhage||No||0||Male||3160|
|35||S40||35||21||Single||Spontaneous||Vaginal||—||38||1||0||Preeclampsia, cervical incompetence||No||1||Female||2330|