The Composition of the Microbiota in the Full-Term Fetal Gut and Amniotic Fluid: A Bovine Cesarean Section Study

The development of a healthy intestinal immune system requires early microbial exposure. However, it remains unclear whether microbial exposure already begins at the prenatal stage. Analysis of such low microbial biomass environments are challenging due to contamination issues. The aims of the current study were to assess the bacterial load and characterize the bacterial composition of the amniotic fluid and meconium of full-term calves, leading to a better knowledge of prenatal bacterial seeding of the fetal intestine. Amniotic fluid and rectal meconium samples were collected during and immediately after elective cesarean section, performed in 25 Belgian Blue cow-calf couples. The samples were analyzed by qPCR, bacterial culture using GAM agar and 16S rRNA gene amplicon sequencing. To minimize the effects of contaminants, we included multiple technical controls and stringently filtered the 16S rRNA gene sequencing data to exclude putative contaminant sequences. The meconium samples contained a significantly higher amount of bacterial DNA than the negative controls and 5 of 24 samples contained culturable bacteria. In the amniotic fluid, the amount of bacterial DNA was not significantly different from the negative controls and all samples were culture negative. Bacterial sequences were identified in both sample types and were primarily of phyla Proteobacteria, Firmicutes, Bacteroidetes, and Actinobacteria, with some individual variation. We conclude that most calves encounter in utero maternal-fetal transmission of bacterial DNA, but the amount of bacterial DNA is low and viable bacteria are rare.


Quality check
FastQC 0.11.8 was ran for all the files and the resulting reports were checked manually 1 . FastQC reports were also compiled and assessed with MultiQC 1.7 2 .

Trimming
All leftover primers and spacers were removed with Cutadapt v.1. The trimmed sequences were again checked with FastQC 0.11.8 and MultiQC 1.7 1,2 .

Mapping file
A QIIME 2 compatible mapping file containing the sample metadata was created with Google Sheets and validated with Keemei 4,5 .

QIIME2
Fastq.gz files were first imported to QIIME2 v2019.4 with command qiime tools import. The sequences were checked again using command qiime demux summarize 5 . The mean number of raw sequences was 114380 (max 187955, min 31, total 7892241) with read quality starting to drop below 20 in forward direction at 280 and reverse at 220. DADA2 was ran using the command dada2 denoise-paired with truncating option F280 and R220 10 . This results in an amplicon sequence variant (ASV) table.
Stats from the denoising and merging steps were checked with qiime metadata tabulate. Resulting data was explored with visual summaries created with qiime feature-table summarize and qiime feature-table tabulate-seqs.
The phylogenetic tree was created using qiime alignment mafft 6 . Highly variable positions adding noise were masked with qiime alignment mask. FastTree was used to create a phylogenetic tree from the masked alignment: qiime phylogeny fasttree and qiime phylogeny midpoint-root 7 .
Preliminary diversity analysis was created with qiime diversity core-metrics-phylogenetic using sampling depth 15988.
The taxonomy was assigned according to the SILVA v132 QIIME release 99 % 8 . Representative set of sequences and taxonomy were imported to QIIME2 with qiime tools import. Reference reads were first extracted from the sequence directory with qiime feature-classifier extract-reads 9 . Then a Naïve Bayes classifier was trained to the curated taxonomy with qiime feature-classifier fit-classifier-naïvebayes 10 . Finally, the actual classification was called with qiime feature-classifier classify-sklearn 11 .
The taxonomy was visualised again with qiime metadata tabulate and qiime taxa barplot. After this the ASV table was exported from QIIME2 for handling in spreadsheet programs. First the table and taxonomy data was exported in biom-format with qiime tools export. The two were combined with biom package: biom add-metadata 13 . Finally the ASV table and table with taxonomy were converted to .tsv format with biom convert.

Data decontamination
Mitochondria, chloroplasts and singletons were filtered out, leaving only bacteria with at least phylum-level identification.
The processed data was filtered further to remove ASVs which represented probable contaminants 12 .
An ASV was removed if its prevalence in actual samples was ≤2× its prevalence in field controls (empty sampling instruments exposed to the surgery room environment), and if its mean relative abundance in actual samples was ≤10× its mean abundance in field controls (Supplemental Figure 1). The filtering was performed separately for meconium and amnion samples. If less than 500 reads remained after the decontamination, as in six meconium samples, the sample was removed from further analyses.
Supplementary Figure 1. Schematic illustration of the data decontamination strategy. ASV1 is accepted as the mean relative abundance and prevalence were both sufficiently high in meconium samples vs. negative controls. ASV1 was accepted as both abundance and prevalence were above threshold. ASV2 and ASV3 are accepted because either abundance or prevalence were sufficient. ASV4 and ASV5 are rejected as neither abundance or prevalence are above the threshold.