Low Maternal Microbiota Sharing across Gut, Breast Milk and Vagina, as Revealed by 16S rRNA Gene and Reduced Metagenomic Sequencing

Abstract

The maternal microbiota plays an important role in infant gut colonization. In this work we have investigated which bacterial species are shared across the breast milk, vaginal and stool microbiotas of 109 women shortly before and after giving birth using 16S rRNA gene sequencing and a novel reduced metagenomic sequencing (RMS) approach in a subgroup of 16 women. All the species predicted by the 16S rRNA gene sequencing were also detected by RMS analysis and there was good correspondence between their relative abundances estimated by both approaches. Both approaches also demonstrate a low level of maternal microbiota sharing across the population and RMS analysis identified only two species common to most women and in all sample types (Bifidobacterium longum and Enterococcus faecalis). Breast milk was the only sample type that had significantly higher intra- than inter- individual similarity towards both vaginal and stool samples. We also searched our RMS dataset against an in silico generated reference database derived from bacterial isolates in the Human Microbiome Project. The use of this reference-based search enabled further separation of Bifidobacterium longum into Bifidobacterium longum ssp. longum and Bifidobacterium longum ssp. infantis. We also detected the Lactobacillus rhamnosus GG strain, which was used as a probiotic supplement by some women, demonstrating the potential of RMS approach for deeper taxonomic delineation and estimation.

Keywords: maternal microbiota; reduced metagenomics; sequencing.

Figure 1

Fraction of RMS reads that belong to commonly detected RMS clusters identified as B. longum ssp. longum and E. faecalis. Median values, as well as 25th and 75th percentile, are depicted in red.

Figure 2

Mapping of raw RMS reads from breast milk, stool and vaginal swab pools towards whole genome sequences of B. longum. F8 and JCM1217—B. longum ssp. longum; JCM1222—B. longum ssp. infantis. Median values, as well as 25th and 75th percentile, are depicted in red; (a) Coverage of mapped genome regions; (b) Pairwise identity of raw RMS reads towards mapped genome regions.

Figure 3

Correlation between 16S rRNA relative abundance and RMS estimates of expected target species in samples where >50% of the mapped regions corresponded to the in silico predicted RMS fragments.