A comparison of rpoB and 16S rRNA as markers in pyrosequencing studies of bacterial diversity

Abstract

Background: The 16S rRNA gene is the gold standard in molecular surveys of bacterial and archaeal diversity, but it has the disadvantages that it is often multiple-copy, has little resolution below the species level and cannot be readily interpreted in an evolutionary framework. We compared the 16S rRNA marker with the single-copy, protein-coding rpoB marker by amplifying and sequencing both from a single soil sample. Because the higher genetic resolution of the rpoB gene prohibits its use as a universal marker, we employed consensus-degenerate primers targeting the Proteobacteria.

Methodology/principal findings: Pyrosequencing can be problematic because of the poor resolution of homopolymer runs. As these erroneous runs disrupt the reading frame of protein-coding sequences, removal of sequences containing nonsense mutations was found to be a valuable filter in addition to flowgram-based denoising. Although both markers gave similar estimates of total diversity, the rpoB marker revealed more species, requiring an order of magnitude fewer reads to obtain 90% of the true diversity. The application of population genetic methods was demonstrated on a particularly abundant sequence cluster.

Conclusions/significance: The rpoB marker can be a complement to the 16S rRNA marker for high throughput microbial diversity studies focusing on specific taxonomic groups. Additional error filtering is possible and tests for recombination or selection can be employed.

Figure 1. Number of OTUs as a function of fractional sequence difference (OTU cut-off) for the 16S rRNA marker gene (A) and the rpoB marker gene (B).

OTU number is plotted for filtered and trimmed sequences (undenoised), denoised sequences and denoised- and reading frame corrected sequences (automated and manual correction). The latter treatments were only applicable to rpoB.

Figure 2. Rarefaction curves showing mean expected OTU number for Proteobacteria as a function of sample size.

The 1% and 2.3% cut-offs for 16S rRNA and rpoB are chosen to reflect species definitions (see text).

Figure 3. Population-level analyses.

A: a Minimum Spanning Tree for all sequences more than 94% similar to abundant Anaeromyxobacter sequence R100. Circle size equates with the number of sequences, bar length equates with the number of nucleotide substitutions between sequences. B: a NeighbourNet phylogenetic network based on the same sequences as in A.