The acetate-degrading anaerobic chemostat was constructed using a continuous stirred tank reactor (CSTR) with a working volume of 1.8 l that was mixed using a magnetic stirrer at 300–400 rpm (Supplementary Figure S1). The reactor was immersed in a thermostat-controlled water bath to maintain a temperature of 37°C. The seed sludge from a swine manure treatment plant (Sichuan Province, China) was used to inoculate the reactor. Artificial wastewater containing acetate as the sole carbon source was continuously supplied to the CSTR by a peristaltic pump under an atmosphere of N₂, and the effluent flowed out from a U-type tube (Supplementary Figure S1). The wastewater [total organic carbon (TOC) concentration of 8.0 g L⁻¹] contained the following components per liter of distilled water: sodium acetate, 5.46 g; acetic acid, 16.0 g; KH₂PO₄, 0.3 g; KHCO₃, 4.0 g; NH₄Cl, 1.0 g; NaCl, 0.6 g; MgCl₂·6H₂O, 0.82 g; CaCl₂·2H₂O, 0.08 g; and cysteine-HCl·H₂O, 0.1 g, supplemented with 10 ml trace element solution of DSMZ medium 318 (Deutsche Sammlung von Mikroorganismen und Zellkulturen, Germany) comprising 1.19 mg L⁻¹ of Ni²⁺ and 0.34 mg L⁻¹ of Co²⁺, and 10 ml vitamin solution of DSMZ medium 318 without vitamin B₁₂.

The reactor was operated at a dilution rate of 0.05 d⁻¹. After the system reached steady operation, additional NH₄Cl was added to gradually increase TAN concentration from 1 g L⁻¹ to 6 g L⁻¹. Biogas production, gas composition, pH, concentrations of volatile suspended solids (VSS), suspended solids (SS), TOC, as well as VFAs were measured periodically using the protocols reported previously (Chen et al., 2020b). Microbial morphology at each stage was observed using a laser-scanning confocal fluorescence microscope (Olympus FV1000, Japan). During the steady operation period at each TAN concentration (N0 ~ N6), biomass was collected from the broth and used for DNA and RNA extraction.

To investigate the microbial community, 40 ml broth from the chemostat (days 103, 138, 187, 222, 250, 301, and 390) was collected in sterile DNase-free centrifuge tubes, centrifuged at 10,000 rpm at 4°C for 10 min, and rinsed thrice with sterile phosphate buffer saline (PBS) (10 mM, pH 7.5). Total DNA was extracted via the cyltrimethyl ammonium bromide (CTAB) method (Griffiths et al., 2000). DNA extracts were subjected to 16S rRNA gene amplicon sequencing. The V4-V5 hypervariable regions of both bacterial and archaeal 16S rRNA genes were amplified using universal primers 515F (5’-GTGCCAGCMGCCGCGGTAA-3′) and 909R (5’-CCCCGYCAATTCMTTTRAGT-3′). PCR product purification, Illumina sequencing, and data processing were conducted as described previously (Zheng et al., 2019). The raw reads of the 16S rRNA gene sequencing were deposited into the NCBI Sequence Read Archive (SRA) database with the accession number PRJNA524473.

To analyze the metabolic characteristics of the community under high ammonia stress (TAN concentration of 6 g L⁻¹), sludge samples for metagenomics and metatranscriptomics sequencing were collected on days 390 and 393. Total DNA and RNA were extracted via the CTAB method (Griffiths et al., 2000). The metagenome DNA was sequenced on an Illumina HiSeq 3,000 platform (Illumina). The obtained raw data were trimmed via Trimmomatic v0.36 (Bolger et al., 2014) with a quality cutoff of 30, a sliding window of 6 bp, and a minimum length cutoff of 100 bp. The clean reads from two metagenomes corresponding were co-assembled via SPAdes v.3.5.0 (Bankevich et al., 2012), binned using MetaBAT (Kang et al., 2015), checked for completeness and contamination using CheckM (Parks et al., 2015), and calculated relative abundance using BBMap (Bushnell, 2014). Phylogenomic trees were built with PhyloPhlAn v0.99 (“-u” option) (Segata et al., 2013). Genes were then annotated using Prokka (Seemann, 2014) and manual curation was performed as described previously (Chen et al., 2020a).

For metatranscriptomics sequencing, total RNA was treated with DNase to remove the residual DNA using an RNase-free DNase set (Qiagen, Hilden, Germany). Ribosomal RNA (rRNA) was removed from the DNase-treated RNA via the Ribo-Zero rRNA Removal Kits (Illumina, San Diego, CA, United States). RNAseq libraries were created using the TruSeq RNA sample prep kit (Illumina, San Diego, CA, United States) with the standard protocol. The sample libraries were sequenced on an Illumina HiSeq 3,000 sequencer. The metatranscriptomics sequences were trimmed as the DNA-trimming step described above and mapped to metagenomic bins using the Bowtie2 aligner (Langmead and Salzberg, 2012). The expression levels of given genes from each bin were calculated separately as reads per kilobase transcript per million reads (RPKM) mapped to the bin averaged from the duplicate samples. In the heatmap illustration, the gene expression levels were further normalized to the median gene expression levels of each bin (RPKM-NM) averaged from the duplicate samples (Nobu et al., 2017). The raw reads of the metagenomics and metatranscriptomics sequencing are accessible at http://bigd.big.ac.cn/gsa, accession number: CRA008529.

The reactor operated continuously for more than 400 days and the methanogenesis kept stable during each stage with different TAN concentrations (N0 ~ N6 stages). The pH was stable at about 7.5 at all stages. From N0 to N2 stages, the biogas production did not decrease, and the acetate fed was completely degraded. From N3, the biogas production decreased slightly, but at the N6 stage, it was about 92% of that at N0 ~ N2 (Figure 1). In biogas, CH₄ accounted for 53–59%, CO₂ accounted for 41–47%, and the partial pressure of H₂ was less than 1 Pa. Acetate built up in the initial period of N5 and N6, but with the extension of the running time, it was degraded. The acetate degradation ratio was high of 99% at the N6 stage. The SS and VSS concentrations increased slightly with increased TAN which was around 1.69 and 0.98 g L⁻¹, respectively, at stage N6 (Figure 1). These results suggested that the methanogenesis was not repressed obviously even at a TAN of 6 g L⁻¹ (free ammonia nitrogen concentration of 0.23 g L⁻¹).

We performed fluorescence observation of the microbial community at each stage. F₄₂₀, the unique key coenzyme in the hydrogenotrophic methanogenesis pathway, can make cells fluoresce blue-green under UV light at 420 mm wavelength (Braks et al., 1994). During the N0 to N4 stages, the microbial morphology was mainly tubular, and almost no fluorescence cell was detected (Supplementary Figure S2). When the TAN concentration reached 5 g L⁻¹, spherical cells were clearly observed, and a strong fluorescence was detected (Supplementary Figure S2). This result indicated that with the increase of TAN concentration, cells able to produce methane through the hydrogenotrophic pathway probably increased, which also suggested that acetate oxidation probably occurred in the reactor.

The microbial community analysis was carried out using the sludge samples on the last day of each operation stage (N0 ~ N6). Based on 16S ribosomal RNA gene analysis, the structure of the microbial community varied greatly with increased TAN concentration (Supplementary Figure S3). Compared with bacteria, the archaeal populations were significantly inhibited, and the relative abundance decreased to 3.54% during the stage of N3 and N4 (Supplementary Figure S4). As the TAN concentration continued to be increased, the relative abundance of archaea gradually recovered to 18.35%. To further reveal the impact of TAN on microbial communities, the dominant bacterial and archaeal (>1% relative abundance at any stage) communities at the operational taxonomic units (OTU) level were analyzed and Spearman’s correlation coefficients were calculated to illustrate the correlation between microbial community structure and TAN concentration (Figure 2). The dominant bacterial phyla included Bacteroidetes, Synergistetes, Firmicutes, and Spirochaetae (Figure 2A; Supplementary Figure S6A). OTUs in Bacteroidetes prevailed in high TAN conditions, in which the relative abundance of Petrimonas (OTU230) and unclassified Lentimicrobiaceae (OTU281) were significantly positively correlated with TAN concentration and Proteiniphilum (OTU214) was extremely significantly correlated, suggesting the high TAN tolerance of these OTUs. In the Synergistetes phylum, the Thermovirga (OTU244) predominated in the N2 stage (61%), but the relative abundance continued to decrease in the later stages. The most abundant OTU in Firmicutes was Ruminiclostridium (OTU97), which was negatively correlated with TAN concentration, while syntrophic butyrate (Syntrophomonas [OTU233]) and acetate (Tepidanaerobacter syntrophicus [OTU176])-oxidizing bacteria were positively correlated with TAN. The main genera in Spirochaetae were unclassified Spirochaetaceae (OTU4) and Sphaerochaeta (OTU275, 202, 220, and 222), their responses to TAN stress are diametrically opposite. Unclassified Spirochaetaceae (OTU4) was strongly inhibited by TAN, whereas Sphaerochaeta was not.

Regarding the archaeal community, the OTUs were determined at the species level by phylogenetic analysis (Supplementary Figure S5). Among them, Methanothrix was the most abundant methanogen across all the stages, the genus mainly included M. soehngenii (OTU99) and M. harundinacea (OTU256), which showed totally opposite trends (Figure 2B). With the increase of TAN concentration, the dominant methanogen in the chemostat was gradually replaced from M. soehngenii (OTU99) to M. harundinacea (OTU256), suggesting that M. harundinacea was more tolerant to TAN suppression. Moreover, the diversity of methanogens increased significantly with the increase in TAN concentration. Methanosarcina (OTU 268) was only detected in the N6 stage, and its relative abundance accounted for 23.5% of total archaeal species (Figure 2B; Supplementary Figure S6B).

To profile the metabolic capability of TAN-tolerant bacteria and methanogens, metagenomics analysis was performed on microbial community samples taken from the chemostat at the N6 stage on two different dates. Overall, 34 Gbp metagenomic clean sequences were obtained. Illumina paired-end reads from the two duplicate samples were co-assembled, and binning the assembled contigs of metagenomes yielded 89 bins. Among these, 45 bacterial and 6 archaeal bins which have at least 72% genome completeness and < 6.5% contamination as estimated by CheckM (Parks et al., 2015) were analyzed (Supplementary Figures S7, S8; Supplementary Table S1). To further reveal the metabolic behavior of the microbial populations, 56.6 million metatranscriptomic reads (4.5 Gbp) were produced and mapped to the metagenomic bins. The percentages of metatranscriptomic reads mapped to dominant genomes were, respectively, analyzed at two sampling time points, and the two transcriptomics analyses displayed similar results (Figure 3). Based on mapping meta-omic reads to the obtained bins, the bacterial populations retrieved accounted for 86 and 60% of the metagenomic and metatranscriptomic reads, respectively (Supplementary Figure S9). Among all bacterial phyla, Bacteroidetes represented major fractions of the 16S rRNA (40.85%), metagenomes (49.59%), and metatranscriptomes (39.15%) at the N6 stage (Figure 3; Supplementary Figures S6A). The major genera in Bacteroidetes were Proteiniphilum (Bin76, Bin79, and Bin41), unclassified Bacteroidetes (Bin83 and Bin32), and Paludibacter (Bin 89; (Figure 3). In addition, as observed in the 16S rRNA gene analysis, populations associated with known syntrophic fatty acid-oxidizing bacteria (Tepidanaerobacter Bin10 and Syntrophomonas Bin24) were detected in Firmicutes (Figures 2A, 3).

As for the methanogen, M. harundinacea (Bin85) was the most abundant, accounting for 9.10 and 28.07% of the metagenomic and metatranscriptomic reads, respectively (Figure 3). The second dominant methanogen was Methanosarcina (Bin33) with 1.95% metagenomic abundance and 8.98% metatranscriptomic abundance. It is worth mentioning that M. soehngenii (Bin57) only accounted for 0.58 and 0.65% of the metagenomic and metatranscriptomic reads, consistent with the results of 16S rRNA gene analysis (Figures 2B, 3).

The community hosted diverse methanogens, including M. harundinacea (Bin85), M. soehngenii (Bin57), Methanosarcina (Bin33), Methanoculleus chikugoensis (Bin50), Methanobacterium arcticum (Bin58), and Methanomassiliicoccus luminyensis (Bin45) (Supplementary Figure S8; Supplementary Table S1). These methanogens have distinct substrate spectrums for methane production, including the acetate cleavage pathway, CO₂ reduction pathway, and methyl cleavage pathway (Figure 4A; Supplementary Table S2). During methanogenesis, methanogens synthesize ATP via a transmembrane Na⁺/H⁺ gradient formed by an electron transfer chain (Figure 4B; Supplementary Table S3). To further understand the metabolic profile of methanogens, we analyzed the methanogenic and energy metabolic pathways of target bins by comparison with known pathways in the UniProt database, then mapped the metatranscriptomic reads to bins and calculate RPKM-NM to analyze the expression level of target pathways (Figure 4C).

Specifically, M. harundinacea (Bin85) and Methanosarcina (Bin33) expressed acetate utilization pathway highly (top octile of expressed genes in the corresponding bin) and M. soehngenii (Bin57) expressed at a lower level (Figures 4A,C; Supplementary Table S2). The M. harundinacea and Methanosarcina correspondingly highly expressed (top quartile or octile) genes for acetoclastic methanogenesis mediated by Fpo (no FpoF) and H₂-cycling (via EchA-F and VhoACG; Figures 4B,C; Supplementary Table S3; Welte and Deppenmeier, 2014).

Bins of hydrogenotrophic methanogens, such as M. chikugoensis (Bin50), M. arcticum (Bin58), and Methanosarcina (Bin33), encode and express genes for CO₂-reducing methanogenesis (Figures 4A,C; Supplementary Table S2). The Methanoculleus highly expressed (top quartile or octile) genes for the oxidation of H₂ (MvhADG-HdrABC and FrhABG) and formate (FdhAB; Figures 4B,D; Supplementary Table S3). Another H₂-oxidizing population, M. luminyensis (Bin45), highly expressed genes for coupling of H₂ oxidation with the reduction of methanol and dimethylamine to methane (Figure 4; Supplementary Tables S2, S3). Interestingly, Methanothrix is a well-known obligate acetotrophic methanogen (Hattori, 2008), but we found the expression of the CO₂-reducing methanogenesis pathway was high in Methanothrix, especially in Bin85 (Figures 4A,C; Supplementary Table S2; Holmes et al., 2017). However, Methanothrix did not express the key genes for utilizing H₂ (Figures 4B,C), it may directly use extracellular electrons for reducing CO₂ to CH₄ (Rotaru et al., 2014; Zhao et al., 2020). Thus, the acetate-fed chemostat was likely to produce methane not only via acetoclastic pathway but also possible via the CO₂ reduction pathway under high TAN-suppressed conditions.

As discussed above, in the acetate-fed and TAN-suppressed methanogenic system, besides directly decomposing acetate into methane, methanogens can also use the products (electron, H₂, and CO₂) of syntrophic acetate oxidizers to generate methane. There are currently two acetate oxidation pathways reported (Zhu et al., 2020), the reversed Wood–Ljungdahl pathway and the glycine cleavage pathway, which are different in the process from CH₃-CO-S-CoA to CH₂ = THF (Figure 5A). To systematically analyze the potential acetate-oxidizing bacteria in the system, 45 bacterial bins were annotated and then compared with the known acetate-oxidizing pathways in the UniProt database (Supplementary Figure S10; Supplementary Table S4). We screened out 20 bacterial bins with complete acetate oxidation pathway or metagenomic abundance greater than 1%, then mapped the metatranscriptomic reads to bins and calculated RPKM-NM to analyze the expression levels of genes related to acetate oxidation, H₂/formate metabolism, and electron transport (Figure 5).

Within the acetate-degrading community, unclassified Thermoanaerobacteraceae (Bin68) encode all genes for both reversed Wood–Ljungdahl and glycine cleavage pathways with high expression of GlyA and GcvP (top quartile of expressed genes in the corresponding bin; Figures 5A,C; Rui et al., 2011). A population (Bin10) related to a known acetate-degrading genus, Tepidanaerobacter, expressed conversion of acetate to formate through the reversed Wood–Ljungdahl pathway (Figures 5A,C; Westerholm et al., 2011). Another population (Bin54) belongs to Clostridiales expressed genes for acetate degradation (Grd, GlyA, GcvP, GcvT, and GcvH) at high levels (top octile; Figure 5C; Westerholm et al., 2018). Notably, Proteiniphilum (Bin76, Bin79, and Bin41) had the highest abundance in both metagenomic and metatranscriptomic reads (Figure 3) and highly expressed genes of glycine cleavage pathway (top quartile or octile; Figures 5A,C); thus, these populations are likely novel acetate degraders.

Multiple genes/gene clusters encoding formate dehydrogenase and hydrogenase are detected in the aforementioned syntrophs for formate and H₂ generation. These enzymes are necessary for the reoxidation of reducing equivalents (i.e., menaquinone, NADH, NADPH, and reduced ferredoxin [Fd]) generated during acetate oxidation (Figures 5B; Supplementary Table S5). For formate metabolism, all syntrophic metabolizers in Firmicutes expressed Fd-dependent formate dehydrogenases (FdhH), NADH-dependent formate dehydrogenases (FdhAB), and putative electron-confurcating formate dehydrogenases (FdhA-HydBC) except for Tepidanaerobacter (Bin10) and unclassified Peptococcaceae (Bin87; Figures 5B,C; Supplementary Figure S11). Among them, unclassified Clostridia Bin8 highly expressed FdhAB (top octile; Figure 5C). For H₂ generation, all syntrophic metabolizers in Firmicutes (except for Syntrophomonas) expressed cytoplasmic [FeFe]-type electron-confurcating hydrogenases (HydABC) (Figures 5B,C; Supplementary Figure S11). These hydrogenases catalyze the thermodynamically favorable production of H₂ from Fd_red to drive the unfavorable production of H₂ from NADH (Buckel and Thauer, 2013). We detected high expression of [FeFe]-type NADP⁺-dependent hydrogenases (HndABCD) in unclassified Clostridiales (Bin22), Tepidanaerobacter (Bin10), Sphaerochaeta (Bin29), Proteiniphilum (Bin79), and unclassified Bacteroidia (Bin42) (top octile or quartile; Figure 5C; de Luca et al., 1998).

To support the thermodynamically challenging H₂ and formate generation, the syntrophs also expressed energy-conserving electron transfer enzymes. The Proteiniphilum populations (Bin79 and Bin41) highly expressed (top octile) the membrane-bound Rhodobacter nitrogen fixation complex (RnfA-G), which can either extrude cation and transfer an electron from Fd_red to NAD⁺ to gain energy or transport cation inward and transfer electron from NADH to Fd_ox to drive electron-confurcating hydrogen or formate generation (Figures 5B,C; Biegel et al., 2011). In addition, Proteiniphilum (Bin79) also highly expressed (top quartile) the NADP⁺-Fd_red oxidoreductase (NfnAB), which transfers electrons from NADPH to NAD⁺ and Fd_ox via electron disproportionation (Figures 5B,C; Wang et al., 2010).

Since the ROS-detoxification mechanism supported the adaptation of Proteiniphilum species (Wu et al., 2021) and ammonia caused oxidative stress in methanogens (Kato et al., 2014; Li et al., 2022), we also examined the occurrence of antioxidant genes in target bins (Figure 6; Supplementary Table S6). All the populations of Proteiniphilum highly expressed genes (top octile of expressed genes in the corresponding bin) encoding superoxide dismutase (Sod) and peroxiredoxins (Prx), respectively, which belong to energy-free ROS scavenger (Supplementary Table S6; Martins et al., 2019). For methanogens, in addition to detecting highly expressed Sod and Prx, we also identified the expression of energy-dependent ROS scavengers (superoxide reductase, Sor; rubrerythrin, Rbr) (Figure 6; Supplementary Table S6), which require additional energy sources with F₄₂₀H₂ as the ultimate electron donor (Martins et al., 2019). Methanothrix harundinacea (Bin85) and Methanosarcina (Bin33) which represented major fractions of the metagenomes and metatranscriptomes in methanogens highly expressed Sor and Rbr (top octile; Figures 3, 6). Thus, the predominant bacterial genus Proteiniphilum and methanogens in the TAN-suppressed system possessed mechanisms of anti-oxidative stress thereby providing a selective advantage under suppressive conditions.