Is there a placental microbiota? A critical review and re-analysis of published placental microbiota datasets | BMC Microbiology

Principal findings of the study

In this re-analysis of fifteen placental microbiota studies, of the ASVs which were ranked in the top five ASVs for relative abundance in any one study, Lactobacillus ASVs were clearly the most prevalent across studies. Yet, Lactobacillus ASV prevalence was explained by background DNA contamination, contamination from the birth canal during vaginal delivery, or well-to-well contamination from vaginal samples during the sequence library build process. Overall, the bacterial DNA profiles of placental samples were highly similar to those of technical controls in their respective studies. Indeed, a secondary analysis of the six studies which targeted the V4 hypervariable region of the 16S rRNA gene for sequencing, showed that the bacterial DNA signal of both placental and technical control samples clustered by study of origin rather than by sample type. In addition, the top two ASVs in placental samples from each of the six studies in the secondary analysis were also the top ranked ASVs in technical controls from the corresponding study. Considered in isolation, placental samples clustered by mode of delivery, suggesting that the process of delivery greatly affected the bacterial DNA profiles of placentas. Therefore, placental samples included in this re-analysis do not provide evidence of a consistent bacterial DNA signal in typical term pregnancy independent of mode of delivery. Instead, the modest consistency in bacterial DNA signals identified across studies was associated with general background DNA contamination or contamination introduced during vaginal delivery.

The findings of this study in the context of prior reports

Currently, the extent of bacterial presence within the placenta is under debate. There have been numerous reviews, commentaries, and editorials, which have sought to synthesize and resolve conflicting results regarding the existence of a placental microbiota [3, 30, 32, 33, 66,67,68, 77, 100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144]. Although there has been disagreement about the existence of a placental microbiota in typical human pregnancy, there is a consensus that any given body site, including the placenta, can be at least transiently infected by microorganisms. Several reviews have emphasized that microorganisms in placental tissue would not be able to survive for long durations given the structure of the placenta and the immunobiological response of the host [3, 110]. In contrast, some have proposed that microorganisms could survive intracellularly within the basal plate of the placenta and thus effectively evade the host immune system [68, 145]. Many reviews addressing prior microbiota datasets have been challenged to draw conclusions given the multiple confounding factors which could significantly influence results: the specific 16S rRNA gene hypervariable region targeted for sequencing, brand and lot number of the DNA extraction kits, gestational age at delivery and sampling, mode of delivery of the placenta, inadequate metadata for deposited sequence data, and a general lack of technical controls to account for background DNA contamination. Regardless, many have viewed the current evidence for placental and/or in utero colonization as theoretically tenuous given the existence of germ-free mammals [146, 147] and the strong potential for background bacterial DNA to influence DNA sequencing surveys of low microbial biomass samples [36, 37, 39,40,41, 91, 112]. Finally, similar to the results presented here in this re-analysis, the prevalence of Lactobacillus across placental samples in prior studies has been acknowledged, yet so too has been the high variability in the bacterial taxa reported within placental tissues across studies. Indeed, variability is high even across studies of similar cohorts from the same research groups [27, 41, 44, 63,64,65, 96].

Mode of delivery must be taken into account when investigating the existence and structure of a placental bacterial DNA signal

Eight studies [27, 38, 43, 44, 52, 54, 57, 59] concluded that the bacterial DNA signals in placentas from cesarean deliveries were not significantly different from those in placentas delivered vaginally. Yet, five other studies [36, 42, 51, 96, 98] have reported that the bacterial DNA signals in placentas from vaginal and cesarean deliveries significantly differ. The latter studies have reported increased prevalence and relative abundance of Lactobacillus and other vaginally associated taxa in placentas from vaginal deliveries. Additionally, even the rupture of membranes, a prerequisite for labor and vaginal delivery, provides microorganisms access to the amniotic cavity [148] and thus the placenta, with prolonged access leading to microbial invasion and infection [149, 150]. Notably, bacterial culture of placentas from vaginal deliveries have significantly higher positivity rates [18, 96], higher total colony counts [40], and a higher prevalence of bacterial colonies from Lactobacillus and Gardnerella, both of which are typical residents of the human vagina [88]. In contrast, placentas from cesarean deliveries consistently yield bacteria which typically predominate on the skin, such as Propionibacterium, Staphylococcus, and Streptococcus [40, 99].

Importantly, through robust analysis of the entire bacterial DNA signal from hundreds of placental samples, this re-analysis clearly highlights the influence of mode of delivery on the bacterial DNA signal from placental samples by demonstrating mode of delivery-associated clustering across six studies. Furthermore, it is apparent that removing the exterior layers (i.e., amnion, chorion, and basal plate) of a placenta delivered vaginally is not sufficient to eliminate delivery associated DNA contamination of the sample since the diversity and structure of bacterial DNA profiles from the inner layers (i.e., subchorion, villous tree) of the placenta differed significantly between cesarean and vaginal deliveries. Evidence in the literature combined with this re-analysis warrants careful consideration of mode of delivery and even time since rupture of membranes [52, 149, 150] when investigating the bacterial DNA signal from placental samples.

Background bacterial DNA limits analysis of bacterial 16S rRNA gene signal from the placenta

Theoretically, a low bacterial biomass community is detectable using 16S rRNA gene sequencing when its concentration is at least 10–100 CFU/mL [151]. However, discerning a true tissue-derived low bacterial DNA signal from other potential sources is exceedingly difficult. This re-analysis, along with eight other studies [36, 37, 39,40,41, 91, 96], found that placental samples and technical controls share highly abundant bacterial taxa when 16S rRNA gene sequencing is used. Since technical controls represent the environment and reagents to which the placenta is exposed post-delivery, it follows that a majority of the bacterial DNA signal from placental samples is also acquired from those environments and reagents. While a placental tissue limit of bacterial detection through DNA sequencing is yet to be determined, other low-bacterial-biomass sample types such as oral rinse, bronchoalveolar lavage fluid, and exhaled breath condensate were predominated by stochastic noise below 104 16S rRNA gene copies per sample [152]. Even the bacterial DNA signal from a pure culture of Salmonella bongori serially diluted to a final concentration of 103 CFU/mL was mostly contamination [78]. If these limits are comparable to those in placental tissue, then stochastic noise and background DNA contamination would predominate the bacterial DNA signal from placental tissue leaving any true DNA signal well beyond the detection limits of 16S rRNA gene sequencing. Therefore, it follows that 16S rRNA gene sequencing by itself without additional verification is inadequate to make a clear assessment of the existence of a placental microbiota.

Prior reports of 16S rRNA gene sequencing on placentas from term pregnancies

With the prior considerations in mind, out of the 40 studies which performed 16S rRNA gene sequencing on placental samples, 32 included at least some term deliveries. However, only 16 focused exclusively on placentas from term deliveries [28, 37, 39,40,41, 43, 49, 53, 54, 56,57,58, 62,63,64,65]. Additionally, only nine of these studies focused exclusively on placentas from cesarean deliveries [28, 39, 41, 49, 56, 58, 62, 64, 65] and only three included technical controls and their DNA sequencing data thus accounting for gestational age, mode of delivery, and background DNA contamination [39, 41, 49]. Two of three concluded that there was no evidence for a placental microbiota in the context of term cesarean delivery [39, 41].

Many studies have reported evidence for a low biomass placental microbiota [27, 29, 30, 43,44,45,46,47, 49, 50, 52,53,54, 57, 58, 60, 61, 63,64,65, 92, 93, 145, 153] but only nine of these studies exclusively sampled placentas from term deliveries [43, 49, 53, 54, 57, 58, 63,64,65]. Predominant bacterial taxa reported in these studies included Pseudomonas [54, 64, 65], Lactobacillus [49, 54], Bacteroidales [64], Enterococcus [63], Mesorhizobium [43], Prevotella [58], unclassified Proteobacteria [57], Ralstonia [43], and Streptococcus [54]. Two studies from this term delivery subset, which sampled multiple regions of the placenta, observed gradients of Lactobacillus relative abundance across levels of the placenta, but in opposite directions [43, 49].

In contrast, five studies did not find evidence for a microbiota in placentas from term deliveries since neither the placental bacterial DNA signal from 16S rRNA gene sequencing [37, 39,40,41] nor the bacterial load as determined by quantitative real-time PCR [37, 39,40,41, 56] were significantly different from technical controls. One study even noted that no operational taxonomic units greater than 1% relative abundance in placental samples, were less than 1% in technical control samples, emphasizing the overlap between the two sample types [37]. Three of these studies [40, 41, 56] also attempted to culture viable bacteria from placental tissue, but were rarely successful. In cases where culture was successful, viable bacteria often conflicted with the DNA sequencing results suggesting that cultured bacteria were likely contaminants [40, 41].

Clinical significance

Non-viable or viable bacterial DNA could feasibly be filtered from maternal blood by the placenta leading to a placental bacterial DNA signal

The placenta is a transient internal organ with functions that include promotion of gas exchange, nutrient and waste transport, maternal immunoglobulin transport, and secretion of hormones critical for fetal growth and development [154]. These exchanges and transfers occur due to diffusion gradients between fetal and maternal blood, the latter of which bathes the chorionic villi in the intervillous space of the placenta [108]. This maternal blood, which cannot be fully drained from the placenta before biopsy or sampling, can undoubtedly contain bacterial particles or even the remnants of a low-grade bacterial infection [56, 112, 155]. Because of its structure, the placenta functions as a filter and retains these particles or bacteria, dead or alive. A bacterial DNA signal due to this filtering process would be extremely weak and transient. In addition, the bacterial taxa identified would be highly variable since they do not correspond to a specific niche, which is consistent with the findings of this re-analysis.

Infection is a potential source for the placental bacterial DNA signal

Instead of in utero colonization, it is more likely that the bacterial DNA signal coming from a subset of placental samples is caused by infection. It is curious to note that specific bacteria are associated with stronger bacterial DNA signals and inflammation in placental tissue resulting in adverse pregnancy outcomes including preterm birth and/or preterm prelabor rupture of membranes (PPROM) [52, 55, 98]. Spontaneous preterm birth has been shown to increase bacterial load [55] and the relative abundances of several taxa in placental samples including but not limited to Ureaplasma [26, 36, 38, 42, 51, 52, 156, 157], Fusobacterium [51, 52], Mycoplasma [42, 51, 52], Streptococcus [36, 51], Burkholderia [27], Escherichia/Shigella [55], Gardnerella [51], Gemella [52], and Pseudomonas [50]. Ureaplasma urealyticum, Mycoplasma hominis, Bacteroides spp., Gardenerella spp., Mobiluncus spp., various enterococci, and Streptococcus agalactiae (also known as Group B Streptococcus or GBS) are frequently associated with histologic acute chorioamnionitis as well as uterine infection [16, 26, 108, 157]. GBS is also a major cause of early onset neonatal sepsis and has been commonly isolated at autopsy in addition to E. coli, and Enterococcus [16, 158]. While metagenomics sequencing could identify genes with pathogenic potential to determine pathogenicity of bacteria (bacterial DNA) detected in the placenta, it is difficult to infer pathogenicity exclusively from 16S rRNA gene sequencing data. Nevertheless, the DNA of the notoriously pathogenic bacterial genera detailed above were all found in placental tissue, suggesting an invasive phenotype rather than commensal colonization.

Recommendations for future studies

In order to establish the existence of a viable placental microbiota several criteria need to be met, which have been detailed previously [36, 41]. Studies which aim to assess the viability of a bacterial DNA signal in a purported low biomass sample type should start with the null hypothesis that the entire DNA signal results from contamination and subsequently attempt to reject it with experimental evidence [159]. Therefore, any study evaluating a potential microbiota of the placenta should attempt to demonstrate viability through both culture and DNA sequencing. Placentas should come from term cesarean deliveries without labor to obviate contamination during vaginal delivery and subjects should be screened to ensure that only healthy women are sampled (i.e., no history of antenatal infection, pre-eclampsia, recent antibiotic use, signs of infection or inflammation). Additionally, future studies need to include ample sequenced technical controls in order to identify and account for sources of contamination, which will inevitably exist no matter how rigorous and/or sterile the protocol [75]. Universal sources of contamination include environmental DNA, reagents used to process samples and build sequence libraries, and even the sequencing instruments themselves. Further, biological replicates from the same placenta should also be included to ascertain the consistency of any bacterial DNA signal. Moreover, it should not be assumed that all remaining sequences are legitimate after post-hoc contaminant removal by algorithms such as DECONTAM, especially in low microbial biomass environments such as the placenta.

Since 16S rRNA gene sequencing limits of detection have not yet been thoroughly interrogated in placental tissue, serial dilutions of spiked-in live bacteria or cell-free DNA should be included in a portion of tissue samples to demonstrate the feasibility of recovering the bacterial DNA signal from placental tissue. When multiplexing samples, unique dual index primer sets should be used to eliminate the possibility of barcode hopping which is another source of sample “contamination” [160, 161], and before sequencing, low biomass samples should be segregated from higher biomass samples to avoid well-to-well contamination [39, 162]. Furthermore, 16S rRNA gene sequencing results should be complemented with shotgun metagenomics sequencing to allow for strain level microbiome analyses that can more effectively link DNA detected in placental tissues to its source. With strain-level resolution, bacterial DNA signals can be identified as being unique to the placenta in an individually-specific manner within a study and thus suggestive of a placental microbiota, or shared across the placental samples in a study indicating universal sample contamination. If multiple sequencing methods or other investigative methods such as culture are utilized, concordance should be sought among the data from these multiple methods to determine the legitimacy of potentially credible microbial signals. Finally, in conjunction with publishing, all sequence data and accompanying detailed metadata should be submitted to a public database and code for any data analyses or manipulations should be made publicly available so that others can replicate and verify the results.

Strengths of this study

Broad searches of the available literature were utilized to ensure that all publicly available 16S rRNA gene sequencing data of placental samples (with associated metadata to partition pooled sample data) were incorporated into the re-analysis, which re-examined the data with in-depth comparisons of term placental samples to technical controls. This allowed for the detection of background DNA contamination in the bacterial DNA signal from placental tissue. In addition, potential confounding variables such as mode of delivery, gestational age at delivery, and 16S rRNA gene target hypervariable region were controlled for whenever possible. By utilizing DADA2 to process the sequence data, variation and biases due to post-sequencing processing were eliminated. This enabled ASV-to-ASV comparisons for six studies which targeted the same 16S rRNA gene hypervariable region using the same PCR primers, a first in the placental microbiota field.

Limitations of this study

The quality and public availability of data and metadata were the primary limiting factors of this re-analysis. Unfortunately, the availability of metadata or even the data themselves is a pervasive issue in the microbiome field [163,164,165]. While study cohort statistics were well reported overall, detailed metadata for each subject are required in order to perform a proper re-analysis. Ideally, any study investigating the existence of a viable placental microbiota would, at a minimum, include associated metadata by subject for potential confounders (e.g., gestational age at delivery, and mode of delivery).

Additionally, the impacts of individual low abundance ASVs (i.e., less than 1% mean relative abundance) were not evaluated. While some of these ASVs could potentially represent DNA signals from viable bacteria, most were likely stochastic environmental DNA contamination. Finally, while the R package DECONTAM was used to remove likely contaminants by comparing the prevalence of ASVs in biological samples and technical controls, this tool would be unable to identify contaminants introduced during sampling, delivery, or to identify taxa which were truly present in a sample but also happened to be present in most control samples. In addition, the contaminant identification accuracy of DECONTAM also diminishes when used on low biomass samples such as placental samples where the majority of the sequences are likely contaminants [75, 166].

Read more here: Source link