Synthetic Statistical Approach Reveals a High Degree of Richness of Microbial Eukaryotes in an Anoxic Water Column

Abstract

Molecular surveys suggest that communities of microbial eukaryotes are remarkably rich, because even large clone libraries seem to capture only a minority of species. This provides a qualitative picture of protistan richness but does not measure its real extent either locally or globally. Statistical analysis can estimate a community's richness, but the specific methods used to date are not always well grounded in statistical theory. Here we study a large protistan molecular survey from an anoxic water column in the Cariaco Basin (Caribbean Sea). We group individual 18S rRNA gene sequences into operational taxonomic units (OTUs) using different cutoff values for sequence similarity (99 to 50%) and systematically apply parametric models and nonparametric estimators to the OTU frequency data to estimate the total protistan diversity. The parametric models provided statistically sound estimates of protistan richness, with biologically meaningful standard errors, maximal data usage, and extensive model diagnostics and were preferable to the available nonparametric tools. Our clone library exceeded 700 clones but still covered only a minority of species and less than half of the larger protistan clades. Our estimates of total protistan richness portray the target community as very rich at all OTU levels, with hundreds of different populations apparently co-occurring in the small (3-liter) volume of our sample, as well as dozens of clades of the highest taxonomic order. These estimates are among the first for microbial eukaryotes that are obtained using state-of-the-art statistical methods and can serve as benchmark numbers for the local diversity of protists.

Microbial community structure in the North Pacific ocean

“…Here, we extend this observation to the microbial Eucarya. Even at the conservative 95% tag sequence identity (B97% SSU rRNA similarity), our OTU observations and richness estimates are almost one order of magnitude higher than any previously reported (Supplementary Table S2) (Jeon et al, 2006;Zuendorf et al, 2006;Countway et al, 2007). The difference between our Eucarya richness estimates for unique and clustered sequences indicates that Eucarya display equivalent levels of microdiversity to the Bacteria community (Acinas et al, 2004;Brown and Fuhrman, 2005).…”

Comparison of Species Richness Estimates Obtained Using Nearly Complete Fragments and Simulated Pyrosequencing-Generated Fragments in 16S rRNA Gene-Based Environmental Surveys

“…We also used six different parametric distributions to fit the frequency data by the method of maximum likelihood, using a downloadable program (http: //www.stat.cornell.edu/ϳbunge/software.html) according to the criteria described previously (55). As previously observed (23,29,49,55), the parametric model of mixture of two exponentials-mixed Poisson was the model that best fit the frequency data in all the datasets (the eight short pyrosequencing-simulating datasets, as well as the nearly full-length data set) and, hence, was used to estimate species richness. The choice of the most appropriate approach for species richness estimation is an issue hotly debated by microbial ecologists (29,46), as well as macro ecologists (22,42).…”

“…However, our goal is not to accurately determine the true species richness in the habitats examined but to provide a comparative tool for datasets that contain the same numbers of taxa. Under this scenario, we applied both parametric and nonparametric methods for estimating species richness as a comparative tool, an approach that is widely accepted in ecological studies (23,26,29,49). OTU assignments and species richness determinations for other environments.…”

Protistan community patterns within the brine and halocline of deep hypersaline anoxic basins in the eastern Mediterranean Sea

“…We estimated the total number of phylotypes (operational taxonomic units) in each sampled community using statistical procedures described previously Jeon et al 2006;Zuendorf et al 2006). In brief, we fit seven candidate parametric abundance models to the observed phylotype frequency Table 1.…”