Impact of library preparation protocols and template quantity on the metagenomic reconstruction of a mock microbial community

Abstract

Background: The rapid development of sequencing technologies has provided access to environments that were either once thought inhospitable to life altogether or that contain too few cells to be analyzed using genomics approaches. While 16S rRNA gene microbial community sequencing has revolutionized our understanding of community composition and diversity over time and space, it only provides a crude estimate of microbial functional and metabolic potential. Alternatively, shotgun metagenomics allows comprehensive sampling of all genetic material in an environment, without any underlying primer biases. Until recently, one of the major bottlenecks of shotgun metagenomics has been the requirement for large initial DNA template quantities during library preparation.

Results: Here, we investigate the effects of varying template concentrations across three low biomass library preparation protocols on their ability to accurately reconstruct a mock microbial community of known composition. We analyze the effects of input DNA quantity and library preparation method on library insert size, GC content, community composition, assembly quality and metagenomic binning. We found that library preparation method and the amount of starting material had significant impacts on the mock community metagenomes. In particular, GC content shifted towards more GC rich sequences at the lower input quantities regardless of library prep method, the number of low quality reads that could not be mapped to the reference genomes increased with decreasing input quantities, and the different library preparation methods had an impact on overall metagenomic community composition.

Conclusions: This benchmark study provides recommendations for library creation of representative and minimally biased metagenome shotgun sequencing, enabling insights into functional attributes of low biomass ecosystem microbial communities.

Fig. 1

Sample overview. Each tube on this plot represents a mock metagenomic library preparation. The control library is an unamplified TruSeq library of the same mock community sample generated from 200 ng input DNA

Fig. 2

a Insert size and (b) GC profiles of Illumina sequence data from each of the three different library preparation methods. Unamplified control library is represented by the red dashed line

Fig. 3

Relative abundances of each mock community member and the unmapped reads with the exception of Nocardiopsis dassonvillei (too few sequences mapped to reference). Individual GC plots corresponding to the high and low GC references are displayed in Additional file 1: Figure S1, which further illustrates the shift in relative abundance from low to high GC organisms across the dilution series. Unamplified control library is represented by the red diamond

Fig. 4

a Mock community relative abundances across each library prep kit and across each dilution including the TruSeq 200 ng Control library (top). b Principal coordinates analyses of Euclidean distances derived from mapping reads to the mock community reference genomes (left column) and of k-mer frequencies (right column, averaged sample k-mer frequencies from k2-k10 sample by sample k-mer distance matrices). Individual samples are colored by either library preparation (top) or starting input quantity (bottom). The unamplified 200 ng control library is represented by the red point in each ordination

Fig. 5

Heatmap noting the completeness of genomic bins extracted from each of the low input metagenomes. The color bar on top of the figure refers to each of the three tested library types and control library (Control = red, Nextera XT = green, Mondrian = blue and MALBAC = yellow). Samples are also arranged where the highest input quantity is located on the left side of each library type