Improved Microbial Community Characterization of 16S rRNA via Metagenome Hybridization Capture Enrichment

Abstract

Environmental microbial diversity is often investigated from a molecular perspective using 16S ribosomal RNA (rRNA) gene amplicons and shotgun metagenomics. While amplicon methods are fast, low-cost, and have curated reference databases, they can suffer from amplification bias and are limited in genomic scope. In contrast, shotgun metagenomic methods sample more genomic regions with fewer sequence acquisition biases, but are much more expensive (even with moderate sequencing depth) and computationally challenging. Here, we develop a set of 16S rRNA sequence capture baits that offer a potential middle ground with the advantages from both approaches for investigating microbial communities. These baits cover the diversity of all 16S rRNA sequences available in the Greengenes (v. 13.5) database, with no sequence having <78% sequence identity to at least one bait for all segments of 16S. The use of our baits provide comparable results to 16S amplicon libraries and shotgun metagenomic libraries when assigning taxonomic units from 16S sequences within the metagenomic reads. We demonstrate that 16S rRNA capture baits can be used on a range of microbial samples (i.e., mock communities and rodent fecal samples) to increase the proportion of 16S rRNA sequences (average > 400-fold) and decrease analysis time to obtain consistent community assessments. Furthermore, our study reveals that bioinformatic methods used to analyze sequencing data may have a greater influence on estimates of community composition than library preparation method used, likely due in part to the extent and curation of the reference databases considered. Thus, enriching existing aliquots of shotgun metagenomic libraries and obtaining modest numbers of reads from them offers an efficient orthogonal method for assessment of bacterial community composition.

Keywords: amplicon; microbial diversity; microbiome; mock communities; next generation sequencing; shotgun libraries; target enrichment.

FIGURE 1

Overview of data analysis methods on the three library types (i.e., 16s amplicon, 16s hybridization bait capture, and metagenomic libraries).

FIGURE 2

Relative abundance of bacterial phyla in mock community controls sequenced and analyzed using different methods. Phyla listed as components of the mock communities are shown. Black vertical bar in each row represents the nominal abundance of respective phylum. Row panel strips labels identify the mock communities; colors identify library type (i.e., amplicon, enriched 16S-cap, unenriched metagenomic library) and analyzing strategy (i.e., denoising, 16Smapping, and marker gene).

FIGURE 3

Relative abundance of bacterial genera in mock community controls sequenced and analyzed using different methods. Genera listed as components of the mock communities are shown. Three families with no genus identification, Enterobacteriaceae, Listeriaceae, Bacillaceae, are plotted below the probable genus (Escherichia, Listeria, and Bacillus), respectively. Black vertical bar in each row represents the nominal abundance of respective genus. Row panel strips labels identify the mock communities; color identify library type (i.e., amplicon, enriched 16S-cap, unriched metagenomic library) and analyzing strategy (i.e., denoising, 16Smapping, and marker gene).

FIGURE 4

Fold change (i.e., upper or under) comparing the relative abundances of respective genera in each library to its nominal abundance. Duncan’s multiple range test was performed to compare each library type for each mock community. Letters indicate whether significant differences were detected.

FIGURE 5

PCoA plots were constructed using Bray-Curtis dissimilarity matrix at a family level (A) and genus level (B). Each project is represented by a colored dot (i.e., orange = BEI mock community, green = mouse samples, blue = rat samples, and purple = Zymo mock community). Each library type, sequencing read length and data analysis method is represented by a different shape (i.e., circle = amplicon library, square = 16S-cap enriched PE150 reads, diamond = unenriched PE150 analyzed with 16S mapping and triangle = unenriched PE150 analyzed with metagenome mapping). Numbers represent sample number.

FIGURE 6

A comparison of the Bray-Curtis distance metric was performed for each library type at a genus level using box plots. Bray-Curtis distance is indicated on the y-axis. Library type is indicated on the x-axis. Duncan’s multiple range test was performed to compare each library type for each mock community. Letters indicate whether significant differences were detected.