Figures
Figure 1. Metrics concerning community successions. In the PCoA plot based on Bray-Curtis distance (A ), control communities were rather time-steady but those of nitrate group were progressively drifting from previous states. Samples falling out of the circle were too divergent from other replicates and were excluded from futher analysis. Communities of all subgroups except the pairs of 4-C/16-C and 8-C/16-C were significantly different (P < 0.05) in the significance matrix of one-way ANOSIM (B ), especially those in the N-N and C-N areas (P < 0.001). Most control communities had comparable dissimilarities within and among subgroups, while dissimilarities among nitrate subgroups were much greater than dissimilarities within nitrate subgroup (C ). The categorized taxonomic composition at the phylum and class levels suggested a stepwise enrichment of beta-proteobacteria occuring to nitrate sediment communities (D ). In particular, genera Thiobacillus andLuteimonas outcompeted other genera (E , unclassified genera were left blank) and became dominant (F ).
Figure 2. Comparison of community and network metrics.(A ) Number of OTUs present in more than 8 out of 12 sediment replicates for each subgroup, (B ) Shannon’s diversity, (C ) Simpson’s evenness, (D ) Number of nodes of each constructed network per subgroup, (E ) Number of modules and (F ) Number of links. When referring to network modularity and node connectivity, we focused on modules consisting of more than 5 interconnected nodes. (A ), (B ) and (C ) were calculated based on bootstrap method (N=1000). R2 and F-significance (P ) were calculated by regression analyses.
Figure 3. Networks based on 12 replicates were constructed by the RMT-MENA program. Fruchterman Reingold layout (A ) exhibited the shrinking nitrate networks. Node size was proportional to its degree. Modules consisting of over 5 nodes were randomly colored, the rest left grey. Positive covariations were colored green and negative covariations were colored red. For convenience, we unified the node density so that network sizes could be directly read based on the diameter. Yifan Hu layout (B ) emphasized the network connectivity. Nodes and links of the same module shared the same color, and all nodes were in a uniform size. In summary, Nitrate networks were getting smaller (A) and simpler (B).
Figure 4. Predictive metabolic profiling, based on KEGG orthology (ko) Level 2. The late nitrate communities (16-N and 32-N) were significantly different from the others, as seen from the enhanced metabolisms of rare amino acids, glycan, nucleotide, cofactors and vitamins, and seen from the suppressed expression in terpenoids, polyketides, xenobiotics, lipid and carbohydrate.
Figure 5. Thiobacillus became the best competitor in diluted communities. (A) community compositions of diluted, enriched communities at the genus level. CK – control communities; NS – the medium, followed by the dilution factors. (B) ,(C) and (D) – linear fitting for Thiobacillusrelative abundance versus nitrate consumption, nitrite production and sulfate production.