Experiment design and sample process.
in depth) from the Pearl River Delta (22°14’50” N, 113°11’50” E) were sampled in December 2017. Plant residues, macrofauna and large particles were removed by hand grubbing. The resulting sediment was vigorously homogenized and distributed to 40 g (wet weight) per share. Sterilized pure water with or without ~7.74 mM calcium nitrate was added to the sediments, making each tube to the volume of 45 ml. In sum, we prepared 60 tubes with nitrate (as treatment) and 60 without nitrate (as control). All tubes were anaerobically cultured at room temperature. 12 treatment and 12 control tubes were collected on Day 0, 4, 8, 16, 32 respectively for physiochemical characterizations and DNA extraction. We quantified the levels of AVS, ferrous iron, total iron, pH, nitrate, nitrate, sulfate, and TOC (total organic carbon). See supplementary information for measurements. For convenience, we assigned all 120 samples into two groups: nitrate and control, 60 samples each. Each group had five subgroups based on sampling time, i.e. 0-N, 4-N, 8-N, 16-N, and 32-N for nitrate group, while 0-C, 4-C, 8-C, 16-C and 32-C for control. Each subgroup consisted of 12 replicates.
To analyze the composition of nitrate-reducing sulfur-oxidizing bacteria (NR-SOB) in the sediment, a groups of serial dilution enrichments (Lagier et al. 2018) with 12 series repetitions were prepared in medium NS (nitrate as the sole electron acceptor and sulfide as the sole electron donor). See supplementary information for medium chemical compositions. 1g wet sediment (~80% w/w water content) was added to 9 ml medium to get the 101 dilution factor. The serial dilution was conducted every ten-fold until 10-8. Enrichments at each dilution factor had 12 parallels. Note that parallels here were not equivalent to replicates since every dilution was random so that the resulting community compositions might differ. See supplementary information for sample cultivation. All samples were quantified of remaining nitrate and sulfate to estimate the metabolic intensity, which was roughly represented by the amount of nitrate consumption and sulfide/thiosulfate consumption. The ultimate NR-SOB strains were isolated by plate-streaking on medium NTS (thiosulfate as the sole electron donor) plates supplemented with 1.5% agar respectively. After one-month anaerobic cultivation, different colonies were isolated, purified and identified by colony PCR of full length 16S rRNA gene.
Illumina Miseq sequencing and predictive functional profiling.
Sequencing of 16S ribosome RNA gene V4 region was used to track community successions. Qualified DNA libraries were sequenced and clustered into de novo OTUs by the Usearch (v9.2.64) pipeline (Edgar 2010). See supplementary information for detail. Briefly, each sample was rarefied to 10,000 clean reads. Sequencing of 16S rRNA gene gives no direct metabolic information so we used FAPROTAX (Loucaet al. 2016) and PICRUSt (Langille et al. 2013) just to roughly estimate the functional changes under nitrate stimulation. Clean 16S rRNA gene tags were re-clustered against the GreenGenes database (v13.8, 97_otus.fasta) to be compatible with FAPROTAX (v1.1) and PICRUSt (v1.1.1). FAPROTAX was used to identify putative functional microbial members, and PICRUSt was to compare the gene abundance before and after nitrate stimulation, especially those involved in N or S metabolisms. Both predictions were based on the same normalized closed-reference OTU table.
Interactive network construction and classification.
Each network was constructed based on Spearman’s correlation of the 12 replicates by the MENA pipeline (http://ieg4.rccc.ou.edu/MENA/) (Zhou et al. 2010; Zhou et al. 2011; Deng et al.2012) yielding 10 networks. In network construction, the threshold was automatically determined in the RMT-based modeling. Only OTUs present in more than 8 out of 12 replicates were involved. Output networks were visualized using Gephi (v0.9.2). A visualized interactive network consisted of nodes and links (or edges). A node represented an OTU that had significant Spearman’s correlations (above the network threshold) with other nodes. A link between two nodes represented their negative or positive covariation (Montoya et al. 2006; Bascompte 2007) depend on the calculated correlation coefficient (-1 to +1). A positive correlation meant the two nodes were sharing a same changing pattern. A module referred to a group of nodes that were highly interconnected within the group and had fewer or no connections outside the group. MENA has four built-in methods for module separation, and we selected the greedy modularity optimization because it generated the highest Modularity index (M ). Generally, the number of nodes defines the network size, while the numbers of modules and links per node define the network complexity. See supplementary information for visualization settings.
Statistical analysis.
Shannon’s index (H) and Simpson’s evenness (J) were calculated to represent alpha-diversity of each community. Community dissimilarity between control and nitrate group was compared using PCoA and one-way ANOSIM, both were based on Brey-Curtis distance. Genera that significantly changed in relative abundance over time or over treatment were selected. We confirmed the compliance of normal distribution of genus frequencies and then quantified the dissimilarities by one-way T-test. Alpha-diversity, PCoA and ANOSIM were performed with R (Vegan package, v2.4-6) (Oksanen et al. 2010).