Taxonomic profile of microbial communities on the Deception Island volcano
Through the annotation of reads, we observed that the taxonomic composition in the 98 oC fumarole was distinct in comparison with other fumaroles and glaciers. Archaea were dominant in samples from the 98 oC fumarole (relative abundance between 31.5 and 87.3%), with the most abundant archaeal phyla classified as Crenarchaeota (23.8-79.3%), followed by Euryarchaeota (2.5-7.5%) and Korarchaeota (0.1-0.4%). Firmicutes (3.1-22.4%), Bacteroidetes (0.6-15.3%), Aquificae (0.3-4.6%), and Thermotogae (0.3-1.0%) were also detected in minor proportions in the 98oC fumarole. Looking at the class level, Thermoprotei, Thermococci, Methanococci, Archaeoglobi, Methanobacteria, Methanopyri, and Methanomicrobia represented the most abundant archaeal classes (>0.1%) in the 98 oC site, and Bacilli, Gammaproteobacteria, Betaproteobacteria, Fusobacteria, Flavobacteria, Aquificae (order Aquificales) and Thermotogae (order Thermotogales) were the dominant classes within Bacteria (Figure 2a).
Archaea were less dominant in the other samples, with a relative abundance of 0.7-2% in fumaroles <80 oC and 0.4-0.6% in glaciers. Although some dominant phyla were common between <80 oC fumaroles and glaciers (e.g. Bacteroidetes, Proteobacteria, and Firmicutes), less dominant phyla were uniquely distributed according to temperature. For example, Thaumarchaeota was predominantly found in <80oC fumaroles (0.8-1% for Whalers Bay and 0.2-0.3% in Fumarole Bay). Verrucomicrobia and Acidobacteria were only detected in glaciers (1.2-3.1% and 1-1.6%, respectively) (Figure 2a). The main classes affiliated within the Bacteroidetes phylum were Cytophagia, Flavobacteria and Sphingobacteria, whereas Gamma- and Alphaproteobacteria were the most represented classes within Proteobacteria, followed by Beta-, Delta- and Epsilonbacteria (Figure 2b). Solibacteres was the abundant class within Acidobacteria, and Verrucumicrobiaea within Verrucomicrobia. Thaumarchaeota assignments were not classified at the class level using reads annotation in MG-RAST. The taxonomic annotation of contigs through the IMG/M system showed similar patterns when compared to reads annotation (Supplementary Figure 1).
We then used co-occurrence network analysis to explore the complexity of interactions within the microbial communities in each treatment (Figure 2c). For this, we calculated SparCC correlations between microbial taxa at the genus level based on metagenome reads annotated in MG-RAST. In general, the complexity of the community increased with the temperature. We also noted that communities of Fumarole Bay were more complex than Whalers Bay. The FBA (98 ºC) site showed the highest level of complexity and a modular structure, whereas the WBC (0 ºC) site had the least complex network. Interestingly, the proportion of positive/negative correlations also changed according to the temperature; at higher temperatures, the proportion is even, while in lower temperatures there was an increase in the number of positive correlations.