Document Detail

Selection of primers for optimal taxonomic classification of environmental 16S rRNA gene sequences.
Jump to Full Text
MedLine Citation:
PMID:  22237546     Owner:  NLM     Status:  MEDLINE    
Microbial community profiling using 16S rRNA gene sequences requires accurate taxonomy assignments. 'Universal' primers target conserved sequences and amplify sequences from many taxa, but they provide variable coverage of different environments, and regions of the rRNA gene differ in taxonomic informativeness--especially when high-throughput short-read sequencing technologies (for example, 454 and Illumina) are used. We introduce a new evaluation procedure that provides an improved measure of expected taxonomic precision when classifying environmental sequence reads from a given primer. Applying this measure to thousands of combinations of primers and read lengths, simulating single-ended and paired-end sequencing, reveals that these choices greatly affect taxonomic informativeness. The most informative sequence region may differ by environment, partly due to variable coverage of different environments in reference databases. Using our Rtax method of classifying paired-end reads, we found that paired-end sequencing provides substantial benefit in some environments including human gut, but not in others. Optimal primer choice for short reads totaling 96 nt provides 82-100% of the confident genus classifications available from longer reads.
David A W Soergel; Neelendu Dey; Rob Knight; Steven E Brenner
Related Documents :
9873836 - Cloning of adriamycin-resistant related (arr) gene in an adriamycin-resistant l1210 var...
1809836 - Characterization of an oxytetracycline-resistance gene, otra, of streptomyces rimosus.
2320576 - Characterization of amplification core and esterase b1 gene responsible for insecticide...
21698286 - Integration of expressed sequence tag data flanking predicted rna secondary structures ...
11197616 - Molecular analysis of the human orosomucoid gene orm1*q0köln responsible for incompati...
10216266 - A highly conserved nematode protein folding operon in caenorhabditis elegans and caenor...
Publication Detail:
Type:  Journal Article; Research Support, N.I.H., Extramural; Research Support, Non-U.S. Gov't     Date:  2012-01-12
Journal Detail:
Title:  The ISME journal     Volume:  6     ISSN:  1751-7370     ISO Abbreviation:  ISME J     Publication Date:  2012 Jul 
Date Detail:
Created Date:  2012-06-19     Completed Date:  2013-02-26     Revised Date:  2014-03-25    
Medline Journal Info:
Nlm Unique ID:  101301086     Medline TA:  ISME J     Country:  England    
Other Details:
Languages:  eng     Pagination:  1440-4     Citation Subset:  IM    
Export Citation:
APA/MLA Format     Download EndNote     Download BibTex
MeSH Terms
Bacterial Typing Techniques / methods*
DNA Primers / genetics*
Gastrointestinal Tract / microbiology*
High-Throughput Nucleotide Sequencing
RNA, Bacterial / genetics*
RNA, Ribosomal, 16S / genetics*
Grant Support
DK00700733/DK/NIDDK NIH HHS; HG4866/HG/NHGRI NIH HHS; HG4872/HG/NHGRI NIH HHS; R01 HG004872/HG/NHGRI NIH HHS; U01 HG004866/HG/NHGRI NIH HHS; //Howard Hughes Medical Institute
Reg. No./Substance:
0/DNA Primers; 0/RNA, Bacterial; 0/RNA, Ribosomal, 16S

From MEDLINE®/PubMed®, a database of the U.S. National Library of Medicine

Full Text
Journal Information
Journal ID (nlm-ta): ISME J
Journal ID (iso-abbrev): ISME J
ISSN: 1751-7362
ISSN: 1751-7370
Publisher: Nature Publishing Group
Article Information
Download PDF
Copyright © 2012 International Society for Microbial Ecology
Received Day: 28 Month: 03 Year: 2011
Revision Received Day: 21 Month: 11 Year: 2011
Accepted Day: 03 Month: 12 Year: 2011
Print publication date: Month: 07 Year: 2012
Electronic publication date: Day: 12 Month: 01 Year: 2012
pmc-release publication date: Day: 1 Month: 7 Year: 2012
Volume: 6 Issue: 7
First Page: 1440 Last Page: 1444
ID: 3379642
Publisher Item Identifier: ismej2011208
DOI: 10.1038/ismej.2011.208
PubMed Id: 22237546

Selection of primers for optimal taxonomic classification of environmental 16S rRNA gene sequences Alternate Title:Primer choice for 16S taxonomic classification
David A W Soergel1*
Neelendu Dey245
Rob Knight3
Steven E Brenner1
1Department of Plant and Microbial Biology, University of California, Berkeley, CA, USA
2Division of Gastroenterology, Department of Medicine, University of California, San Francisco, CA, USA
3Howard Hughes Medical Institute and Department of Chemistry and Biochemistry, University of Colorado, Boulder, CO, USA
*Current address: Department of Computer Science, University of Massachusetts, 140 Governors Drive, Amherst, MA 01003-9264, USA. E-mail:
4Current address: Center for Genome Sciences and Systems Biology, Washington University School of Medicine, Saint Louis, MO 63108, USA.
5Current address: Division of Gastroenterology, Department of Internal Medicine, Washington University School of Medicine, Saint Louis, MO 63108, USA.

Variation in 16S ribosomal gene sequences has been used since the mid-1980's to characterize microbial diversity (Stahl et al., 1984). Interest in sequence-based surveys of environmental microbes has exploded in recent years with the availability of sequencing technologies that produce ever-larger data sets at ever-decreasing cost; in particular, the Illumina platform is attractive because of throughput, despite its short reads (Sogin et al., 2006; Lazarevic et al., 2009; Claesson et al., 2010; Caporaso et al., 2011; Degnan and Ochman, 2012). Here, we examine the reliability of assignment of novel sequences to known taxa under thousands of simulated scenarios, varying primer choice, read length and environment.

Previous work on taxonomic classification of environmental 16S rRNA gene sequences has focused on whether reference sequences matching a given query share taxonomic annotations (Jonasson et al., 2002; Desantis et al., 2006; Sogin et al., 2006; Wang et al., 2007). Validations of taxonomic classifiers have typically compared a limited range of primers, read lengths and environments (Sundquist et al., 2007; Huse et al., 2008; Liu et al., 2008; Wu et al., 2008; Hamp et al., 2009). Reference databases contain many sequences derived from some environments and few associated with others, however (Supplementary Figure S1), leading to substantial variation in classification quality. In addition, the use of ‘leave-one-out' cross-validation at the sequence level (Sundquist et al., 2007; Wang et al., 2007; Liu et al., 2008; Wu et al., 2008)—where a single sequence with a known annotation is held out from a reference database and classified using the remainder—is problematic: reference sequences matching held-out query sequences are likely to originate from the same sample, because natural environments contain ‘microdiverse' clusters of closely related strains (Acinas et al., 2004).

We addressed these issues by simulating truncated reads from eight large environmental data sets of near-full-length 16S rRNA gene sequences extracted from GreenGenes (Supplementary Table S1), using pairs of 44 ‘universal' primers commonly found in the literature (Supplementary Tables S2 and S3). These were selected from an initial set of 94 primers by the criterion that each primer had to match at least 40% of the sequences in at least one of the chosen environmental samples. Single-end reads were tested from each primer with all viable amplification partners (794 combinations), and paired-end reads were tested using all 374 viable pairings of the 22 forward and 22 reverse primers. Simulations using 11 read lengths (32 nt, 48 nt, 64 nt, 80 nt, 96 nt, 112 nt, 128 nt, 260 nt, 400 nt, 800 nt and full-length), with the constraint that read length could not exceed amplicon length for each primer pair, produced 6617 single-end and 3061 paired-end datasets per environment.

Reference databases were constructed by holding out each entire study in turn from GreenGenes, clustering the remainder at 99% using UCLUST (Edgar, 2010), and selecting one representative sequence per cluster (see Supplementary Methods for details). Each query fragment was then matched against remaining representative sequences using USEARCH (Edgar, 2010), configured to penalize indels and mismatches equally. Clusters were then selected that matched within 0.5% identity of the best hit (hits <80% identity were disregarded). For paired-end query sequences, our Rtax procedure (Supplementary Methods) selected those reference clusters that matched both reads simultaneously with an average percent identity within 0.5% identity of the maximum. Taxonomic classifications were made at each level by retaining annotations agreeing among >50% of the clusters (including those with no annotation in the denominator); these generally extended at best to the genus level, because the reference database provides few species-level annotations.

Sequences from novel taxa (or sequences that appear novel due to sequencing error or chimerism) clearly cannot be correctly classified; however, such sequences may constitute a substantial proportion of a given sample (Supplementary Figure S1 and Supplementary Table S1). The version of GreenGenes that we used excluded taxa (defined by 97% identity) that were unique to a single sample, as one of the several strategies to remove chimeras. These unique sequences were therefore excluded from our query sets. Thus the classification rates we report represent the proportion of non-chimeric, non-unique sequences that can be classified to each rank. If an environmental sample is not similarly filtered before classification, then the classifiable proportion (that is, taken with respect to the total sample) will be correspondingly lower.

Classification rate and accuracy vary widely among environments and sequence regions, for several reasons: (1) the reference database provides different levels of coverage of each environment, (2) no primer is truly ‘universal' and different primers (and pairs) hit different proportions of sequences in each environment and (3) the targeted regions are variably informative. Figure 1 shows proportions of sequences from each environment classified to each rank, for all 9678 single-ended and paired-end primer and read length combinations. Horizontal panels compare unfiltered results to classifications passing 80% and 95% estimated accuracy filters (see Supplementary Methods), showing that most classifications can be made with high accuracy when optimal primers are chosen. Remarkably, only 96 nt of sequence (taken as a single read or as a pair of 48 nt reads) can provide 82–100% of the 80% accurate genus classifications available from any read length (Supplementary Table S4). Paired-end sequencing can provide substantial gains in classification rate for some—but not all—environments and read lengths. Paired-end classifications are typically more accurate than those made from single reads, and so are more likely to pass the 95% estimated accuracy filter (Supplementary Table S5). Another surprise is that hypervariable regions need not be specifically targeted, as there is no obvious relationship between taxonomic informativeness of a region and the extent to which it overlaps any of the classical ‘V-regions'.

No one combination of primers and read length works best in all environments, but near-optimal performance in six out of the eight environments is available using paired-end 80 nt reads from primers such as E517F, U515F or E341F paired with E1406R or closely related primers (Table 1). However, practical considerations such as ability to amplify low-biomass samples will sometimes influence which primers are used. For instance, short amplicons may be preferred because these are less subject to length heterogeneity biases and chimera formation. Similarly, short single-ended sequences are less subject to errors due to chimeras, simply because they are less likely to span a breakpoint. Classification performance for experimental choices matching such constraints can be found in the supplementary data.

The choice of reference database and taxonomy can have a dramatic impact on the resulting classification accuracy. In this study, we used the current GreenGenes taxonomy, which has been filtered to remove chimeras and where the taxonomic annotations are comprehensive and consistent with the phylogenetic tree (McDonald et al., 2011). Experiments using a previous version of the GreenGenes taxonomy lacking these features yielded far poorer accuracy (data not shown). In addition, bolstering areas of low coverage in reference databases will substantially improve classifier performance. For instance, taxa in the hypersaline mat, coral and grassland soil samples were underrepresented in the reference database (Supplementary Figure S1), and—presumably as a consequence—classifications of sequences from those samples were less likely to prove correct (Figure 1). Additional data sets from poorly sampled environments will also help to distinguish chimeric from legitimate but novel sequences.

In combination, these results indicate that taxonomic classifications of short reads—especially genus-level classifications—should be treated with skepticism, unless the specific combination of primer, read length, environmental source, reference database and assignment method has been thoroughly validated. At the same time, optimal choices of these parameters allow high classification rates and high accuracy. Thus, large-scale projects such as the Earth Microbiome Project (Gilbert et al., 2010), which aims to collect and analyze samples from tens of thousands of microbial habitats around the globe, may reasonably proceed with standardized primer choices and short reads.


Supplementary Information accompanies the paper on The ISME Journal website (

This research used ShaRCS, UC Shared Research Computing Services Cluster, which is technically supported by multiple UC IT divisions and managed by the University of California, Office of the President. This work was supported in part by the National Institutes of Health (HG4872; HG4866; T32 institutional training grant DK00700733) and the Howard Hughes Medical Institute.

Acinas SG,Klepac-Ceraj V,Hunt DE,Pharino C,Ceraj I,Distel DL,et al. Year: 2004Fine-scale phylogenetic architecture of a complex bacterial communityNature43055155415282603
Caporaso JG,Lauber CL,Walters WA,Berg-Lyons D,Lozupone CA,Turnbaugh PJ,et al. Year: 2011Global patterns of 16S rRNA diversity at a depth of millions of sequences per sampleProc Natl Acad Sci USA108(Suppl 14516452220534432
Claesson MJ,Wang Q,O'Sullivan O,Greene-Diniz R,Cole JR,Ross RP,et al. Year: 2010Comparison of two next-generation sequencing technologies for resolving highly complex microbiota composition using tandem variable 16S rRNA gene regionsNucleic Acids Res38e20020880993
Desantis TZ,Hugenholtz P,Larsen N,Rojas M,Brodie EL,Keller K,et al. Year: 2006Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARBAppl Environ Microbiol725069507216820507
Degnan PH,Ochman H. Year: 2012Illumina-based analysis of microbial community diversityThe ISME J6183194
Edgar RC. Year: 2010Search and clustering orders of magnitude faster than BLASTBioinformatics262460246120709691
Gilbert JA,Meyer F,Antonopoulos D,Balaji P,Brown CT,Brown CT,et al. Year: 2010Meeting report: the terabase metagenomics workshop and the vision of an Earth Microbiome ProjectStand Genomic Sci324324821304727
Hamp TJ,Jones WJ,Fodor AA. Year: 2009Effects of experimental choices and analysis noise on surveys of the ″rare biosphere″Appl Environ Microbiol753263327019270149
Huse SM,Dethlefsen L,Huber JA,Welch DM,Relman DA,Sogin ML. Year: 2008Exploring microbial diversity and taxonomy using SSU rRNA hypervariable tag sequencingPLoS Genet4e100025519023400
Jonasson J,Olofsson M,Monstein HJ. Year: 2002Classification, identification and subtyping of bacteria based on pyrosequencing and signature matching of 16S rDNA fragmentsAPMIS11026327212076280
Lazarevic V,Whiteson K,Huse S,Hernandez D,Farinelli L,Osterås M,et al. Year: 2009Metagenomic study of the oral microbiota by Illumina high-throughput sequencingJ Microbiol Methods7926627119796657
Liu CH,Lee SM,Vanlare JM,Kasper DL,Mazmanian SK. Year: 2008Regulation of surface architecture by symbiotic bacteria mediates host colonizationProc Natl Acad Sci USA1053951395618319345
McDonald D,Price MN,Goodrich J,Nawrocki EP,DeSantis TZ,Probst A,et al. Year: 2011An improved Greengenes taxonomy with explicit ranks for ecological and evolutionary analyses of bacteria and archaeaISME Je-pub ahead of print 1 December 2011; doi:10.1038/ismej.2011.139
Sogin ML,Morrison HG,Huber JA,Welch DM,Huse SM,Neal PR,et al. Year: 2006Microbial diversity in the deep sea and the underexplored ″rare biosphere″Proc Natl Acad Sci USA103121151212016880384
Stahl DA,Lane DJ,Olsen GJ,Pace NR. Year: 1984Analysis of hydrothermal vent-associated symbionts by ribosomal RNA sequencesScience22440941117741220
Sundquist A,Bigdeli S,Jalili R,Druzin ML,Waller S,Pullen KM,et al. Year: 2007Bacterial flora-typing with targeted, chip-based pyrosequencingBMC Microbiol710818047683
Wang Q,Garrity GM,Tiedje JM,Cole JR. Year: 2007Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomyAppl Environ Microbiol735261526717586664
Wu D,Hartman A,Ward N,Eisen JA. Year: 2008An automated phylogenetic tree-based small subunit rRNA taxonomy and alignment pipeline (STAP)PLoS One3e256618596968

Article Categories:
  • Short Communication

Keywords: 16S ribosomal RNA, taxonomy, phylogeny, classification, bacteria, sequencing.

Previous Document:  Dinitrogen fixation in a unicellular chlorophyll d-containing cyanobacterium.
Next Document:  In situ proteo-metabolomics reveals metabolite secretion by the acid mine drainage bio-indicator, Eu...