Comparative metagenomic and rRNA microbial diversity characterization using archaeal and bacterial synthetic communities

Abstract

Next-generation sequencing has dramatically changed the landscape of microbial ecology, large-scale and in-depth diversity studies being now widely accessible. However, determining the accuracy of taxonomic and quantitative inferences and comparing results obtained with different approaches are complicated by incongruence of experimental and computational data types and also by lack of knowledge of the true ecological diversity. Here we used highly diverse bacterial and archaeal synthetic communities assembled from pure genomic DNAs to compare inferences from metagenomic and SSU rRNA amplicon sequencing. Both Illumina and 454 metagenomic data outperformed amplicon sequencing in quantifying the community composition, but the outcome was dependent on analysis parameters and platform. New approaches in processing and classifying amplicons can reconstruct the taxonomic composition of the community with high reproducibility within primer sets, but all tested primers sets lead to significant taxon-specific biases. Controlled synthetic communities assembled to broadly mimic the phylogenetic richness in target environments can provide important validation for fine-tuning experimental and computational parameters used to characterize natural communities.

Figure 1

Characterization of the Archaea-Bacteria community by 454-FLX-T (A) and Illumina-HighSeq (B) metagenomic sequencing. The accuracy of retrieving the known composition of the metagenome is indicated for each organism as a ratio of the observed genomic coverage to the known genome abundance in the community and is plotted against its known abundance in the community. Shading zones indicate a low level of bias (dark: <1.5 fold; light: 1.5–2 fold) from the perfect value of 1.

Figure 2

Effect of sequence processing parameters on OTUs. Sequenced amplicons from V13 region of SSU rRNA of both Archaea (A) and Bacteria (B) were filtered, trimmed, and clustered using the parameters specified in a table form (A–H). Sequences were trimmed to the same coordinates after alignment against the SILVA database and clustered using either complete linkage clustering (CLC) or average linkage clustering (AL) with distances based on Infernal alignment or SILVA based alignment in mothur, respectively. The numbers of OTUs at 97% sequence similarity (distance 0.03) are shown with distinguished contribution from OTUs consisting of one, two or more sequences.

Figure 3

OTU-based diversity estimation as a function of genetic distance and analytical approach relative to the reference genomic SSU rRNA sequences. Bacterial V13, V4, archaeal V13 and the combined archaeal-bacterial V48 amplicon datasets are shown. The results for the other amplicons are shown online in the Supplementary Figure 8. Silva–SLP–AL (red) and Silva–SLP–AL-HQ (black): single-linkage pre-clustering 2% and average linkage clustering of SILVA alignment of sequences not purged of errors and on sequences with the chimeras removed (parameters B and G in Figure 2, respectively). PW–SLP–AL (green): single-linkage pre-clustering 2% and average linkage clustering of Needleman-Wunsch (NW) pairwise alignment of sequences not purged of errors. PW-AN-AL (orange): average linkage clustering of pairwise alignment of sequences after denoising and chimera removal using AmpliconNoise/Perseus. For comparison, OTUs obtained by clustering the reference sequences using Silva–SLP–AL (blue) are shown. Note that the y-axis in (A) is scaled logarithmically.

Figure 4

Taxonomic diversity and abundance inferences based on shotgun metagenomic and amplicon sequencing. The accuracy ratio (observed abundance/expected abundance) is represented as a heat map diagram with values for each organism and data set. Bias values of >1.5-fold are represented as a heat map of increasing color intensity (red for underestimated and green for overestimated abundance). A value of 0.0 indicates >10 fold underestimated abundance, but detection at low levels. ND indicates that no sequence for that organism was identified in that amplicon dataset. Values are averages of three independent amplification and sequencing replicates.

Figure 5

(A) Principal Coordinate Analysis (Bray-Curtis similarity) of Bacteria and Archaea community composition inferred using metagenomics (454-M and ILM-M) and SSU rDNA amplicon sequencing relative to the known composition based on community assembly (REF). Replicates for each amplicon are presented, with closer grouping indicating less variability. The V48 data is presented separately for Archaea and Bacteria in those respective panels but was obtained using the combined AB community. (B) Hierarchical clustering (Bray-Curtis similarity) of community composition accuracy indexes for each amplicon region and sequencing strategy.