A comprehensive benchmarking study of protocols and sequencing platforms for 16S rRNA community profiling - PubMed (original) (raw)

A comprehensive benchmarking study of protocols and sequencing platforms for 16S rRNA community profiling

Rosalinda D'Amore et al. BMC Genomics. 2016.

Abstract

Background: In the last 5 years, the rapid pace of innovations and improvements in sequencing technologies has completely changed the landscape of metagenomic and metagenetic experiments. Therefore, it is critical to benchmark the various methodologies for interrogating the composition of microbial communities, so that we can assess their strengths and limitations. The most common phylogenetic marker for microbial community diversity studies is the 16S ribosomal RNA gene and in the last 10 years the field has moved from sequencing a small number of amplicons and samples to more complex studies where thousands of samples and multiple different gene regions are interrogated.

Results: We assembled 2 synthetic communities with an even (EM) and uneven (UM) distribution of archaeal and bacterial strains and species, as metagenomic control material, to assess performance of different experimental strategies. The 2 synthetic communities were used in this study, to highlight the limitations and the advantages of the leading sequencing platforms: MiSeq (Illumina), The Pacific Biosciences RSII, 454 GS-FLX/+ (Roche), and IonTorrent (Life Technologies). We describe an extensive survey based on synthetic communities using 3 experimental designs (fusion primers, universal tailed tag, ligated adaptors) across the 9 hypervariable 16S rDNA regions. We demonstrate that library preparation methodology can affect data interpretation due to different error and chimera rates generated during the procedure. The observed community composition was always biased, to a degree that depended on the platform, sequenced region and primer choice. However, crucially, our analysis suggests that 16S rRNA sequencing is still quantitative, in that relative changes in abundance of taxa between samples can be recovered, despite these biases.

Conclusion: We have assessed a range of experimental conditions across several next generation sequencing platforms using the most up-to-date configurations. We propose that the choice of sequencing platform and experimental design needs to be taken into consideration in the early stage of a project by running a small trial consisting of several hypervariable regions to quantify the discriminatory power of each region. We also suggest that the use of a synthetic community as a positive control would be beneficial to identify the potential biases and procedural drawbacks that may lead to data misinterpretation. The results of this study will serve as a guideline for making decisions on which experimental condition and sequencing platform to consider to achieve the best microbial profiling.

PubMed Disclaimer

Figures

Fig. 1

Fig. 1

Experimental design. (a) Design of single and dual-index sequencing strategy and schematic describing the 3 amplicon designs: Fusion Primer Design (A) is a one step PCR which uses a single 12-nt error-correcting Golay index sequence (blue) allowing a high multiplexing capability. Tag tailed design (B) is a 2-step PCR which uses a universal primer for the first step and a dual index barcoded primer set in the second step. Standard Illumina Nextera 8-nt index sequences were used (pink Index 5; blue Index 7). The Pac Bio Ligate Adapters design (C): Two harpin adapters (grey) were ligated to a barcoded template (BF forward barcode; BR reverse barcode) to allow multiplexing. (b) Platform Specific Amplicon Libraries: Illumina paired-end sequencing (1,2) generates 2 sequencing reads (R1 and R2) per each cluster and can have single (Standard/Golay) or dual indexes (I5, I7). Ion Torrent and 454 (3) have a single read for each bead with a single index (MID). Pacific Bioscience generate a single circular read for each molecule (SMRT bell) and can have one (BF or BR) or two indexes. The starting point and direction of sequencing reads are indicated by a solid blue line and arrows, respectively. In the case of Fusion Primer Design custom sequencing primer were used

Fig. 2

Fig. 2

Schematic representation of the combination of primers covering the 16S rRNA hypervariable regions and the sequencing platform used in this study

Fig. 3

Fig. 3

a Error rates across four different platforms. Platform had a significant impact on error rate (Kruskal-Wallis comparing MS, 454, IT and PB ROI non-parametric ANOVA _p_=0.015) as did number of CCS cycles for PB (_p_=0.016). b Percentage of reads not matching across the four different platforms. Platform had a significant impact on percentage matching (Kruskal-Wallis non-parametric ANOVA _p_=0.0001) as did number of CCS cycles for PB (_p_=0.016)

Fig. 4

Fig. 4

a Impact of overlapping reads on MS error rates for the DI library preparation method. Overlapping reads significantly reduced error rates for the DI library preparation method (t-test comparing forward [mean 1.38 %] and overlapped error rates [0.13 %] p_=0.00016). b Impact of overlapping reads on MS error rates for the FG library preparation method. Overlapping reads did not significantly reduce error rate for the FG library preparation method (t-test comparing forward [mean 0.50 %] and overlapped error rates [0.42 %] p_=0.36). It is also worth mentioning here that not all the reads overlapped, for example, for the MS platform, and with the given settings in PANDAseq (as discussed in the main text), the statistics for the percentage of reads that were assembled successfully are: 80.93 % (1_s t quantile); 89.02 % (median); 81.07 % (mean); and 95.67 % (3_r d quantile)

Fig. 5

Fig. 5

a Impact of no. of PCR cycles on the forward MS error rate. Increasing number of cycles did increase forward error rate with marginal significance for the FG library preparation method with Q5 Taq (t-test 15 cycles [mean 0.58 %] vs 25 cycles [mean 0.64 %] _p_=0.11). b Impact of PCR starting amount on percentage of chimeric reads. Decreased starting amount reduced percentage of chimeras for the FG library preparation method with HiFi Taq but not significantly (t-test comparing 1 ng [mean 0.08 %] and 10 ng [mean 0.2 %] _p_=0.20). c Impact of no. of PCR cycles on the percentage of chimeric reads. Increasing cycle number increased the percentage of chimeric reads for the FG library preparation method with Q5 Taq (t-test 15 cycles [mean 0.00 %] vs 25 cycles [mean 0.66 %] _p_=0.0245)

Fig. 6

Fig. 6

a Heatmap for EM communities (showing the bacterial species) reconstructed from different platforms using a range of experimental designs for amplicons. The design parameters are shown on top (b) NMDS plot based on Bray-Curtis distance comparing the samples showns in (a)

Fig. 7

Fig. 7

Quantitative results for two EM-UM pairs (among a total of 22) for MS and PB are shown. The fitted line through the points is represented by a blue line with R-squared shown on top. The red line is the ground-truth with the slope difference from the blue line also shown on top

References

    1. Lim K, Furuta Y, Kobayashi I. Large variations in bacterial ribosomal RNA genes. Mol Biol Evol. 2012;29(10):2937–48. doi: 10.1093/molbev/mss101. -DOI -PMC -PubMed
    1. Rinke C, Schwientek P, Sczybra A, Ivanova NN, Anderson IJ, Cheng JF, et al. Insight into the phylogeny and coding potential of microbial dark matter. Nature. 2013;499:431–7. doi: 10.1038/nature12352. -DOI -PubMed
    1. Van de Peer Y, Chapelle S, De Wachter R A quantitative map of nucleotide substitution rates in bacterial rRNA. Nucl Acids Res. 1996;24:3381–91. doi: 10.1093/nar/24.17.3381. -DOI -PMC -PubMed
    1. Klindworth A, Pruesse E, Schweer T, Pelies J, Quast C, Horn M, et al. Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next-generation sequencing-based diversity studies. Nucleic Acids Res. 2012;20:1–11. -PMC -PubMed
    1. Ghyselinck J, Pfeiffer S, Heylen K, Sessitsch A, De Vos P. The effect of primer choice and short read sequences on the outcom of 16S rRNA gene based diversity studies. PLoS One. 2013;8(8):e71360. doi: 10.1371/journal.pone.0071360. -DOI -PMC -PubMed

Publication types

MeSH terms

Substances

Grants and funding

LinkOut - more resources