Investigating microbial eukaryotic diversity from a global census: insights from a comparison of pyrotag and full-length sequences of 18S rRNA genes

Abstract

Next-generation DNA sequencing (NGS) approaches are rapidly surpassing Sanger sequencing for characterizing the diversity of natural microbial communities. Despite this rapid transition, few comparisons exist between Sanger sequences and the generally much shorter reads of NGS. Operational taxonomic units (OTUs) derived from full-length (Sanger sequencing) and pyrotag (454 sequencing of the V9 hypervariable region) sequences of 18S rRNA genes from 10 global samples were analyzed in order to compare the resulting protistan community structures and species richness. Pyrotag OTUs called at 98% sequence similarity yielded numbers of OTUs that were similar overall to those for full-length sequences when the latter were called at 97% similarity. Singleton OTUs strongly influenced estimates of species richness but not the higher-level taxonomic composition of the community. The pyrotag and full-length sequence data sets had slightly different taxonomic compositions of rhizarians, stramenopiles, cryptophytes, and haptophytes, but the two data sets had similarly high compositions of alveolates. Pyrotag-based OTUs were often derived from sequences that mapped to multiple full-length OTUs at 100% similarity. Thus, pyrotags sequenced from a single hypervariable region might not be appropriate for establishing protistan species-level OTUs. However, nonmetric multidimensional scaling plots constructed with the two data sets yielded similar clusters, indicating that beta diversity analysis results were similar for the Sanger and NGS sequences. Short pyrotag sequences can provide holistic assessments of protistan communities, although care must be taken in interpreting the results. The longer reads (>500 bp) that are now becoming available through NGS should provide powerful tools for assessing the diversity of microbial eukaryotic assemblages.

FIG 1

Numbers of pyrotag OTUs formed at different sequence similarities compared to numbers of full-length sequence OTUs formed at 97% sequence similarity (leftmost bar). The filled section of each bar represents the number of singleton OTUs, and the exact percentage is given inside each bar. The pyrotag data set for this analysis was standardized to the full-length sequence data set by randomly subsampling pyrotag sequences in each sample to match the number of full-length sequences in the sample (Table 1).

FIG 2

Rarefaction curves generated using the full-length (FL) and pyrotag (PT) sequence data sets. OTUs were called at 97%, 98%, and 99% sequence similarities for the pyrotags. (A and B) Curves obtained using all sequences and OTUs. Panel B is an enlargement of the boxed area in panel A. (C and D) Curves obtained after the removal of singleton OTUs (i.e., OTUs with only one sequence each). Panel D is an enlargement of the boxed area in panel C.

FIG 3

Comparison of the higher-level taxonomic compositions (percentages of total sequences) of the pyrotag (PT) and full-length (FL) sequence data sets. Taxonomic information was derived from best-match results obtained by BLAST searching of representative sequences from each OTU (selected by mothur) against the SILVA database. Full-length sequence OTUs were called at 97% sequence similarity, while pyrotag OTUs were called at 97%, 98%, and 99% sequence similarities. Results with and without singleton OTUs are shown.

FIG 4

Nonmetric multidimensional scaling analysis using pairwise Bray-Curtis similarity values estimated from the pyrotag and full-length sequence data sets. (A) The number of full-length sequences was standardized to 450 sequences per sample. (B) The pyrotag data set including singletons was standardized to 4,736 sequences per sample. (C) The pyrotag data set without singletons was standardized to 4,472 sequences per sample. OTUs formed from full-length sequences were called at 97%, while OTUs formed from pyrotag sequences were called at 98%. Group average similarity values of clusters with significant differences, indicated by the CLUSTER analysis (P, <0.05 by the SIMPROF test), were overlaid on the MDS plot. Communities within the same subcluster were not significantly different from each other (P, >0.05).