Background

Marine microbial communities have been essential contributors to global biomass, nutrient cycling, and biodiversity since the early history of Earth, but so far their community distribution patterns remain unknown in most marine ecosystems.

Methodology/Principal Findings

The synthesis of 9.6 million bacterial V6-rRNA amplicons for 509 samples that span the global ocean's surface to the deep-sea floor shows that pelagic and benthic communities greatly differ, at all taxonomic levels, and share <10% bacterial types defined at 3% sequence similarity level. Surface and deep water, coastal and open ocean, and anoxic and oxic ecosystems host distinct communities that reflect productivity, land influences and other environmental constraints such as oxygen availability. The high variability of bacterial community composition specific to vent and coastal ecosystems reflects the heterogeneity and dynamic nature of these habitats. Both pelagic and benthic bacterial community distributions correlate with surface water productivity, reflecting the coupling between both realms by particle export. Also, differences in physical mixing may play a fundamental role in the distribution patterns of marine bacteria, as benthic communities showed a higher dissimilarity with increasing distance than pelagic communities.

Conclusions/Significance

This first synthesis of global bacterial distribution across different ecosystems of the World's oceans shows remarkable horizontal and vertical large-scale patterns in bacterial communities. This opens interesting perspectives for the definition of biogeographical biomes for bacteria of ocean waters and the seabed.

Background

The operational taxonomic unit (OTU) is widely used in microbial ecology. Reproducibility in microbial ecology research depends on the reliability of OTU-based 16S ribosomal subunit RNA (rRNA) analyses.

Results

Here, we report that many hierarchical and greedy clustering methods produce unstable OTUs, with membership that depends on the number of sequences clustered. If OTUs are regenerated with additional sequences or samples, sequences originally assigned to a given OTU can be split into different OTUs. Alternatively, sequences assigned to different OTUs can be merged into a single OTU. This OTU instability affects alpha-diversity analyses such as rarefaction curves, beta-diversity analyses such as distance-based ordination (for example, Principal Coordinate Analysis (PCoA)), and the identification of differentially represented OTUs. Our results show that the proportion of unstable OTUs varies for different clustering methods. We found that the closed-reference method is the only one that produces completely stable OTUs, with the caveat that sequences that do not match a pre-existing reference sequence collection are discarded.

Conclusions

As a compromise to the factors listed above, we propose using an open-reference method to enhance OTU stability. This type of method clusters sequences against a database and includes unmatched sequences by clustering them via a relatively stable de novo clustering method. OTU stability is an important consideration when analyzing microbial diversity and is a feature that should be taken into account during the development of novel OTU clustering methods.

[This corrects the article DOI: 10.1186/s40168-015-0081-x.].

We present a performance-optimized algorithm, subsampled open-reference OTU picking, for assigning marker gene (e.g., 16S rRNA) sequences generated on next-generation sequencing platforms to operational taxonomic units (OTUs) for microbial community analysis. This algorithm provides benefits over de novo OTU picking (clustering can be performed largely in parallel, reducing runtime) and closed-reference OTU picking (all reads are clustered, not only those that match a reference database sequence with high similarity). Because more of our algorithm can be run in parallel relative to "classic" open-reference OTU picking, it makes open-reference OTU picking tractable on massive amplicon sequence data sets (though on smaller data sets, "classic" open-reference OTU clustering is often faster). We illustrate that here by applying it to the first 15,000 samples sequenced for the Earth Microbiome Project (1.3 billion V4 16S rRNA amplicons). To the best of our knowledge, this is the largest OTU picking run ever performed, and we estimate that our new algorithm runs in less than 1/5 the time than would be required of "classic" open reference OTU picking. We show that subsampled open-reference OTU picking yields results that are highly correlated with those generated by "classic" open-reference OTU picking through comparisons on three well-studied datasets. An implementation of this algorithm is provided in the popular QIIME software package, which uses uclust for read clustering. All analyses were performed using QIIME's uclust wrappers, though we provide details (aided by the open-source code in our GitHub repository) that will allow implementation of subsampled open-reference OTU picking independently of QIIME (e.g., in a compiled programming language, where runtimes should be further reduced). Our analyses should generalize to other implementations of these OTU picking algorithms. Finally, we present a comparison of parameter settings in QIIME's OTU picking workflows and make recommendations on settings for these free parameters to optimize runtime without reducing the quality of the results. These optimized parameters can vastly decrease the runtime of uclust-based OTU picking in QIIME.