Nitrate input simplified interactive network
Ten networks based on Spearman’s correlation coefficient were constructed to identify possible interactions. Key network topologies were listed in Table S1, from which we selected network size, modularity and connectivity to correlate three community metrics (Figure 2A, 2B, 2C). Sizes of control networks remained time-steady (428.8 ± 23.7 nodes,P >> 0.05) while nitrate networks kept shrinking, from 308 (Day 0) to 94 nodes (Day 32), meaning that fewer OTU pairs were generating significant abundance covariations. Such changing patterns applied to all metrics in Figure 2 that elevated nitrate was simplifying community compositions (with respect to richness and α-diversity) and microbial interactions. To visualize this simplification, all networks were plotted (Figure 3). The Fruchterman Reingold layout (Figure 3A) well exhibited the shrinking sizes of nitrate networks, while the Yifan Hu layout (Figure 3B) emphasized the decreasing modularity and node-node interactions.
Putative keystone nodes referred to those of great network connectivity, either within a module (i.e. module hub) or among modules (i.e. connector). Their taxa were therefore putative keystone taxa. We plotted the distribution of keystone nodes in Figure S2A. Of all the 3352 nodes, most were insignificant peripherals (97.67%), 44 were module hubs and 34 were connectors (Table S2). No network hub was detected. When correlating node significance to OTU relative abundances (Figure S2B), we found that abundance did not necessarily determine its network significance, although keystone nodes turned to be sparse OTUs. Of the 78 keystone nodes, 31 were rare (< 0.1% relative abundance), 27 were moderately rare (0.1 - 0.5%), 13 were moderately abundant (0.5 - 1%), and only 7 were abundant (>1%). In addition, the relative abundances of module hubs and connectors indicated no significant difference (P > 0.05).
To investigate the putative keystone taxa, we prepared a phylogenic tree that indicated (i) all the putative keystone taxa at the phylum or genus level; (ii) their phylogenic distance based on neighbor-joining method; and (iii) their keystone type (Figure S3). Most keystone nodes (85.5%) could be assigned to a phylum but only 28.9% could be assigned to a genus. Members of phylum Proteobacteria (e.g. genusSmithella and Syntrophorhabdus ) and Chloroflexi(genus Bellilinea , Leptolinea and Pelolinea ) constituted the majority of keystone taxa. Genus Smithella in particular, was found highly interactive in networks since 5 OTUs belonging to Smithella were keystone nodes. In addition, four keystone taxa were methanogenic archaea (Lee et al. 2015) and were members of genus Methanolinea , Methanothrix , and Methanobacterium.
Generally, the number of keystone nodes decreased as the nitrate amendment proceeded longer (Figure S4). That is, the network simplification occurred alongside the loss of keystone taxa. Recall thatThiobacillus was barely detectable in the control communities (< 0.02% relative abundance) but dominated the late nitrate communities (~60.88%). However, it was never a keystone genus, showing no significant covariations in any network (Table S3).