Analysis, Optimization and Verification of Illumina-Generated 16S rRNA Gene Amplicon Surveys

Abstract

The exploration of microbial communities by sequencing 16S rRNA genes has expanded with low-cost, high-throughput sequencing instruments. Illumina-based 16S rRNA gene sequencing has recently gained popularity over 454 pyrosequencing due to its lower costs, higher accuracy and greater throughput. Although recent reports suggest that Illumina and 454 pyrosequencing provide similar beta diversity measures, it remains to be demonstrated that pre-existing 454 pyrosequencing workflows can transfer directly from 454 to Illumina MiSeq sequencing by simply changing the sequencing adapters of the primers. In this study, we modified 454 pyrosequencing primers targeting the V4-V5 hyper-variable regions of the 16S rRNA gene to be compatible with Illumina sequencers. Microbial communities from cows, humans, leeches, mice, sewage, and termites and a mock community were analyzed by 454 and MiSeq sequencing of the V4-V5 region and MiSeq sequencing of the V4 region. Our analysis revealed that reference-based OTU clustering alone introduced biases compared to de novo clustering, preventing certain taxa from being observed in some samples. Based on this we devised and recommend an analysis pipeline that includes read merging, contaminant filtering, and reference-based clustering followed by de novo OTU clustering, which produces diversity measures consistent with de novo OTU clustering analysis. Low levels of dataset contamination with Illumina sequencing were discovered that could affect analyses that require highly sensitive approaches. While moving to Illumina-based sequencing platforms promises to provide deeper insights into the breadth and function of microbial diversity, our results show that care must be taken to ensure that sequencing and processing artifacts do not obscure true microbial diversity.

Methane-oxidizing seawater microbial communities from an Arctic shelf

“…The V4-V5 region of the 16S rRNA gene was amplified with forward primer 518F (5 -xx-CCAGCAGCYGCGGTAAN-3 ), and an 8 : 1 : 1 mix of the reverse primers 926R1 (5 -yy-CCGTCAATTCNTTTRAGT-3 ), R2 (5 -yy-CCGTCAATTTCTTTGAGT-3 ), and R3 (5 -yy-CCGTCTATTCCTTTGANT-3 ) (Nelson et al, 2014). Primers included 33 base pair (bp) adapters (xx, yy) at the 5 end.…”

Assessing myxozoan presence and diversity using environmental DNA

“…In fact, with the very high coverage and relatively low complexity in this study, a rarely mentioned problem of "leakage" 435 between libraries was observed. This is also referred to as index misassignment and may interfere with high sensitivity amplicon sequencing approaches (Nelson et al, 2014). Some OTUs, very abundant in only one library (e.g.…”

“…An obvious example was the leakage of 1-3 reads from the single marine library into the freshwater libraries and we used this to set the level of the threshold (0.1%) for 440 accepting the presence of an OTU in a library. Such technical artifacts during the MiSeq run are often not accounted for and may result from sequencing and image analysis errors during the index sequencing phase of the run (Nelson et al, 2014). Carry-over of amplicons from previous runs was not possible as the two runs (first containing the cormorant library and second containing all other libraries) with the myxosporean barcodes and primers were run several months apart.…”

Fast and simple analysis of MiSeq amplicon sequencing data with MetaAmp

“…This problem, resulting from sequencing and/or image analysis errors during the index sequencing phase of the MiSeq run (a separate step in the sequencing process), causes a small fraction of amplicons from one library to be incorrectly assigned to an index of another library (Nelson et al, 2014). In the mock community analysis we clearly showed the relevance of this problem by identification of 2-7 OTUs that were contaminants from DNA extraction kits and cross-talk.…”

“…We strongly recommend the use of this procedure because our analysis showed that the contamination rate can be more than 2% in a MiSeq run. Within-run cross-talk can artificially inflate OTU numbers and diversity measurements if not properly addressed, leading to incorrect interpretation of results when investigating low abundance OTUs (Nelson et al, 2014).…”