16S rRNA gene-based profiling of the human infant gut microbiota is strongly influenced by sample processing and PCR primer choice

Abstract

Background: Characterisation of the bacterial composition of the gut microbiota is increasingly carried out with a view to establish the role of different bacterial species in causation or prevention of disease. It is thus essential that the methods used to determine the microbial composition are robust. Here, several widely used molecular techniques were compared to establish the optimal methods to assess the bacterial composition in faecal samples from babies, before weaning.

Results: The bacterial community profile detected in the faeces of infants is highly dependent on the methodology used. Bifidobacteria were the most abundant bacteria detected at 6 weeks in faeces from two initially breast-fed babies using fluorescent in situ hybridisation (FISH), in agreement with data from previous culture-based studies. Using the 16S rRNA gene sequencing approach, however, we found that the detection of bifidobacteria in particular crucially depended on the optimisation of the DNA extraction method, and the choice of primers used to amplify the V1-V3 regions of 16S rRNA genes prior to subsequent sequence analysis. Bifidobacteria were only well represented among amplified 16S rRNA gene sequences when mechanical disruption (bead-beating) procedures for DNA extraction were employed together with optimised "universal" PCR primers. These primers incorporate degenerate bases at positions where mismatches to bifidobacteria and other bacterial taxa occur. The use of a DNA extraction kit with no bead-beating step resulted in a complete absence of bifidobacteria in the sequence data, even when using the optimised primers.

Conclusions: This work emphasises the importance of sample processing methodology to downstream sequencing results and illustrates the value of employing multiple approaches for determining microbiota composition.

Keywords: 16S rRNA gene sequencing; Bifidobacteria; FISH; Infant; Intestinal microbiota.

Fig. 1

Dendrogram illustrating the microbial composition in two babies, pre-weaning. Thirty-eight sequenced samples are shown, derived from DNA extracted using the Fast DNA SPIN Kit for Soil, which contains a bead-beating step, from nine distinct samples from two babies at different time points, amplified with four primer sets (Table 2), and a further single DNA extraction of one sample using the, non-bead-beating, Qiagen QIAamp kit. N-BF indicates samples from the natural birth, solely breast-fed infant. C-MF indicates samples from the C-section birth, mixed-feeding infant. The infant age at time of sampling is shown (in weeks). The dendrogram clearly shows the difference in composition, specifically the lack of bifidobacterial sequences, between the Qiagen kit (marked with QIA and red branches in the figure) and every other sample. Different PCR primer combinations are indicated by branch colouring: yellow—27f-YM primer; green—27f-Mix combination of forward primers; the two shades of blue represent samples processed with the 27f-Bif and Bif164 control primer sets. Adjacent bar charts show the bacterial composition of the sequence data at the family level. Using the 27f-Mix PCR primers increased detection of bifidobacterial sequences compared to using the 27f-YM primer, which has two mismatches to the Bifidobacterium genus

Fig. 2

Comparison of bacterial families detected in faecal samples from two babies. Sequence data is based on 16S rRNA gene amplicons obtained using the 27f-YM (blue) or 27f-Mix (red) forward primers. a Baby N-BF: Data shows the mean percentage of sequences in each bacterial family after 15 separate DNA extractions at seven time points. b Baby C-MF: Data shows the mean percentage of sequences in each bacterial family after six separate DNA extractions at three time points. For both panels, individual data points are plotted as open circles; centre lines in the box plots show the medians; crosses represent sample means; box limits indicate the 25th and 75th percentiles as determined by R software; whiskers extend 1.5 times the interquartile range from the 25th and 75th percentiles, outliers are represented by dots. Plotted using BoxPlotR [52]

Fig. 3

Longitudinal bacterial profile of two babies (pre-weaning), comparing FISH and 16S rRNA gene sequencing data. a, b—sequencing data (27f-Mix primer set); c, d—FISH data. a, c Baby N-BF, natural birth, breast-fed only; b, d Baby C-MF, C-section, one bottle formula/day introduced from 5 weeks. FISH probes used were Eub338 (total bacterial count), Erec482 (Lachnospiraceae), Fprau645 (F. prausnitzii group of the Ruminococcaceae), Bif164 (Bifidobacterium genus), Rum730 (Rfla729 + Rbro730) (Ruminococcus flavefaciens and R. bromii subclusters of the Ruminococcaceae), Prop853 (Veillonellaceae), Bac303 (Bacteroides-Prevotella group), LAB158 (Lactobacillaceae and Enterococcaceae) and EntD (Enterobacteriaceae). The same colouring scheme has been used to illustrate overlap between bacterial taxa identified using the two methods