When to suspect contamination rather than colonization - lessons from a putative fetal sheep microbiome

Abstract

There is an ongoing controversy around the existence of a prenatal, fetal microbiome in humans, livestock, and other animals. The 'in utero microbial colonization' hypothesis challenges the clinical paradigm of the 'sterile womb' but has been criticized for its reliance on DNA-based evidence to detect microbiomes and the failure to conciliate the routine experimental derivation of germ-free animals from surgically resected embryos with a thriving fetal microbiome. In order to avoid the propagation of misinformation in the scientific literature, a critical assessment and careful review of newly published studies, particularly those that challenge the convincing current clinical dogma of the sterile womb, is of critical importance.We read with interest a recent publication that postulated the presence of a fetal microbiome in sheep, but questioned the plausibility of the reported findings and their meaningfulness to prove "microbial colonisation of the fetal gut […] in utero". We reanalyzed the published metagenomic and metatranscriptomic sequence data from the original publication and identified evidence for different types of contamination that affected all samples alike and could explain the reported findings without requiring the existence of a fetal microbiome.Our reanalysis challenges the reported findings as supportive of a prenatal fetal lamb microbiome. The shortcomings of the original analysis and data interpretation highlight common problems of low-biomass microbiome projects. We propose genomic independence of separate biological samples, i.e. distinctive profiles at the microbial strain level, as a potential new microbiome marker to increase confidence in metagenomics analyses of controversial low-biomass microbiomes.

Keywords: Fetal microbiome; contamination; metagenomics; metatranscriptomics; microbial strains; phiX; sheep.

Figure 1.

Comparable relative abundance profiles in different DNA and RNA samples and controls. Relative abundances were calculated based on the number of reads mapped to reference genomes with BWA, either relative to the total number of quality-filtered reads (for O. aries) or relative to the number of qualify-filtered reads after removal of sheep sequence data (all others). Reads were mapped iteratively using a filtering approach, i.e. only those reads that did not map to one genome were used as input for the mapping to the next genome, in the order O. aries -> H. sapiens -> E. coli -> C. marimammalium -> phiX.

Figure 2.

Shared single-nucleotide variants in fetal lamb metagenomes and positive control identify the same strain. Two screenshots show trimmed and filtered metagenomic reads mapped with BWA to E. coli K12 DH10B (GCF_000019425.1). Alignments were visualized with the Integrative Genomic Viewer (IGV; https://software.broadinstitute.org/software/igv/). Based on unique and shared single nucleotide variation (SNV) profiles the positive control (p) contains two strains. With respect to the depicted genome region, one strain is identical to DH10B whereas the other strain is also found in three fetal lamb samples (C1, C4, C5)

Figure 3.

Sample and control metagenomes and metatranscriptomes contain the same phiX strain. A screenshot shows trimmed and filtered metagenomic and metatranscriptomic reads mapped with BWA to phiX174(GCF_000819615.1). Sequencing coverage in all datasets extends over the entire phage genome (100%), including non-transcribed regions around the origin of replication