Analysis of microbial differences in amniotic fluid between advanced and normal age pregnant women

Amniotic fluid (AF) has been considered sterile for nearly a century because no microorganisms were identified by traditional culture methods [1]. However, this opinion paradigm has been challenged by recent studies based on culture-independent sequencing techniques [2‐4] and AF proteomics [5]. Recently, a conclusion that no microorganisms were present in the mid-trimester AF of healthy pregnancies was reached using culture-independent sequencing techniques [6] and seemed to settle the argument [7]. However, it could not explain why non-human proteins were identified in normal human AF supernatants [8] and why microbial exposure primes fetal immune cells in fetal tissues during fetal development [9].

Intra-amniotic infection caused by microbial invasion of the amniotic cavity (MIAC) was associated with adverse pregnancy outcomes, when the bacteria were at high concentrations [10]. This may be a reason why people think that AF is sterile and that bacteria in AF are abnormal because of their negative effects.

Meanwhile, we could identify bacterial proteins in the AF proteomics database that were consistent with the results of 16S ribosomal RNA (rRNA) gene sequencing. Nine amniotic fluid samples were collected from 9 pregnant women and the pregnancy outcomes of the participants were followed. No bacteria were found by cultivation, but a sparse microbial presence was found by proteomics analysis and 16S rRNA gene sequencing approach. The 148 microbes found in the human AF proteomics database were consistent with the microbes found in the 16S rRNA gene sequencing database. The species composition and the structure of communities in the normal age (<  35 years old) and advanced maternal age (AMA) (>  35 years old) pregnancies differed significantly. However, all of the newborns were healthy and had no allergic reactions.

Proteomics analysis has identified nearly 2000 proteins in AF during the past 20 years [11], including many non-human proteins [8]. In a recent study, 7 normal AF samples were used to generate human AF proteomes, which were divided into 4 groups: original proteins, bound proteins, flow-through proteins, and iTRAQ-labeled individual/mixed digested peptides [11]. We reanalyzed the human AF protein database and found that a part of the non-human proteins were derived from microorganisms. A total of 148 microbial-associated proteins in the normal human AF proteome, which could potentially play important roles in cellular/metabolic processes and binding/catalytic activity (Additional file 1). Notably, all of the microbial-associated proteins were present at a low level in human AF. These data suggest that there may be low concentrations of microorganisms in normal human AF samples.

To test our hypothesis, we conducted a study to investigate the presence of microorganisms in the mid-trimester AF of 9 women [4 normal age [normal] and 5 advanced maternal age [AMA]; advanced maternal age is generally defined as age above 35 years at the time of delivery [12, 13]) by cultivation and 16S rRNA gene sequencing, and to follow their pregnancy outcomes. If the AMA group differed from the normal group in the microbial-associated 16S rRNA genes, it could reflect an endogenous microbial difference between the two groups and support the hypothesis that the AF is not sterile; otherwise, the microorganisms could be generated by exogenous contamination, which could not rule out the possibility that AF is sterile.

The main findings were: (1) B-ultrasonography and karyotype findings of the fetuses were normal and no bacteria were found by cultivation; (2) all 148 microbial-associated proteins in the normal human AF proteome were found in the 16S rRNA gene sequencing database, in which Bacillus, Mycobacterium, and Pseudomonas accounted for  ~ 20%; (3) the bacterial richness of the AF samples showed no significant difference between the AMA and normal groups (Chao1 index, Welch’s t test, P  =  0.540; Additional file 2); (4) a significant difference in the species composition and structure of communities in the AF samples was found between the normal and AMA groups (Fig. 1A, B, Weighted_unifrac, OUT, Welch’s t test, P  =  0.017); (5) the newborns were healthy and had no allergic reactions up to 90 days (Table 1). Collectively, these data suggest that the normal AF is not sterile and that the species composition and structure of communities change in the AMA group, although the bacterial richness may be similar and have no effect on the babies’ health.

The differences between the AF samples from the normal and AMA groups with regard to bacterial species composition and structure of communities were as follows. Five bacterial operational taxonomic units (OTUs) were increased in the AMA AF samples: Lactobacillus helveticus, Pediococcus acidilacticii, Pasteurella multocida, Bacillus indicus, and Bacteroides vulgatus. We used the LEfSe (Linear discriminant analysis Effect Size) method to identify bacterial OTUs that were likely to explain most of the differences between the normal and AMA AF samples. The bacterial orders of OTUs differed between the two groups. Sphingomonadales were more abundant in the normal group, while Lactobacillales were more abundant in the AMA group (Fig. 1C, LDA scores  >  4, Additional file 3). The differences in OTUs mainly spanned two Orders, with the Families Lactobacillaceae (primarily Lactobacillus helveticus) and Sphingomonadaceae accounting for the majority of the differences. Collectively, there may be a low concentration of microorganisms in the normal human AF samples and significant differences between the normal and AMA AF samples.

In the present study, we found that microbial-associated proteins and 16S rRNA genes could be identified in human AF at a low concentration. Furthermore, the species composition and structure of communities differed significantly between the normal and AMA AF samples. Thus, we conclude that the microbial-associated 16S rRNA gene in human AF is real, rather than occurring through microbial contamination, and that the bacteria in AF differ between normal age and AMA pregnancies. We already know that AF neutrophils can phagocytize bacteria during intra-amniotic infection [14], but we do not know the function of bacteria in normal AF. Two possible hypotheses were proposed based on our results. First, we suggest that the AF is not sterile, but the level of microbiota may be very low and under the mother’s immune system control. Therefore, the microbial communities may be related to the establishment of fetal immune function. Immunoglobulins from the mother may help to control the number or activity of the bacteria, to precisely control and activate the fetal immune system, given that recent studies demonstrated the presence of microbes or microbial DNA in the placenta, amniotic fluid [15], and meconium. Furthermore, Florent et al. [9] found live microbes in human fetuses such as Lactobacillus, and suggested that the selective presence of live microbes in fetal organs may have broader implications toward the establishment of immune competency and priming before birth, with the microbial exposure priming fetal immune cells during early human development. Second, we suggest that the bacterial 16S rRNA gene may come from the mother, because Rodriguez and colleagues proved that bacteria are transferred to the fetus from the mother by testing the meconium of healthy babies [16]. In addition, circulating cell-free DNA fragments are able to transfer between the fetus and the mother [17], and the bacterial 16S rRNA gene can be successfully detected in cell-free plasma DNA [18]. In this way, the bacterial 16S rRNA gene could possibly enter the AF through the umbilical cord, and this would explain why we were able to identify the 16S rRNA gene and peptides, but could not cultivate the bacteria.