Surprising prokaryotic and eukaryotic diversity, community structure and biogeography of Ethiopian soda lakes

Abstract

Soda lakes are intriguing ecosystems harboring extremely productive microbial communities in spite of their extreme environmental conditions. This makes them valuable model systems for studying the connection between community structure and abiotic parameters such as pH and salinity. For the first time, we apply high-throughput sequencing to accurately estimate phylogenetic richness and composition in five soda lakes, located in the Ethiopian Rift Valley. The lakes were selected for their contrasting pH, salinities and stratification and several depths or spatial positions were covered in each lake. DNA was extracted and analyzed from all lakes at various depths and RNA extracted from two of the lakes, analyzed using both amplicon- and shotgun sequencing. We reveal a surprisingly high biodiversity in all of the studied lakes, similar to that of freshwater lakes. Interestingly, diversity appeared uncorrelated or positively correlated to pH and salinity, with the most "extreme" lakes showing the highest richness. Together, pH, dissolved oxygen, sodium- and potassium concentration explained approximately 30% of the compositional variation between samples. A diversity of prokaryotic and eukaryotic taxa could be identified, including several putatively involved in carbon-, sulfur- or nitrogen cycling. Key processes like methane oxidation, ammonia oxidation and 'nitrifier denitrification' were also confirmed by mRNA transcript analyses.

Figure 1. Parametric richness estimates.

Box-plots cover 95% Bayesian confidence intervals of total OTU richness for each sample. Grey boxes indicate DNA amplicon datasets, white boxes cDNA amplicons and black boxes DNA amplicon datasets derived from prefilters. Solid lines below the box plots indicate rarified OTU richness. Arithmetic means of medians for DNA amplicon datasets (excluding prefilter-derived) are shown below lake names.

Figure 2. Venn diagram showing the distribution of shared OTUs across lakes.

White numbers indicate the number of OTUs in each possible subset, adjusted for differences in sequencing depth.

Figure 3. Non-metric multidimensional scaling (NMDS) based on Bray-Curtis dissimilarities between OTU compositions of individual datasets.

Sequence datasets OTUs and fitted physicochemical parameters are plotted on the first two NMDS axes. The colors and shapes of individual OTUs and sequence datasets represent their taxonomical classification or dataset type, according to the legends.

Figure 4. Distribution matrix with, DNA/RNA ratio, number of OTUs and rRNA contigs for the five most abundant taxa in each habitat.

Abundances are based on DNA amplicons from collection filters except those indicated with a star (*), instead based on prefilter-derived datasets. Taxa were defined at family level except for RF3 and MG I where information was not available at this resolution. DNA/RNA ratios are based on the dataset with highest RNA abundance and number of rRNA contigs include only those >750 bp. The dendogram indicate average linkage clustering of habitats based on OTU distribution (BC-dissimilarity).