Physicochemical properties of cellar mud
During the long term continuous CSFL fermentation, we observed a usage duration-dependent variability of cellar mud properties. Linear regression analysis showed that the water content of cellar mud significantly increased from 28.53% in C3 to 36.82% in C100 (Table 1, Fig. S1). The average pH decreased from weakly alkaline (pH = 7.31) for C3 to weakly acidic (pH = 6.75) for C100, with the overall change being significant (p < 0.05) across cellars in the four ages (Table 1, Fig. S1). Carbon source and nitrogen source indicators including organic matter (7.78% to 16.72%), dissolved organic carbon (4.11 to 17.00 g kg-1-TS), total carbon (2.69% to 7.18%), and total nitrogen (0.39% to 0.96%) displayed significantly positive correlation with the cellar age (Table 1, Fig. S1). In terms of organic acids, the concentration of lactate (4.17 to 7.76 g kg-1-TS), acetate (2.83 to 5.05 g kg-1-TS), and butyrate (0.53 to 1.55 g kg-1-TS) considerably increased from C3 to C100 (Table 1, Fig. S1). As compared to the young cellars (C3), we found that the propionate (0.28 to 0.81 g kg-1-TS) and caproate (2.48 to 9.83 g kg-1-TS) concentrations were significantly higher in the old cellars (C10, C30, and C100). For inorganic ions, NH4+ (1.26 to 3.28 g kg-1-TS), Cl- (161.52 to 293.21 mg kg-1-TS), and PO43- (0.29 to 2.29 g kg-1-TS) significantly accumulated with the cellar age, while we rarely detected NO2− and NO3− (Table 1, Fig. S1). The accumulation of ions noticeably increased the electrical conductivity (491.33 to 1054.53 μS cm-1) in the old cellar (Table 1, Fig. S1). Pearson correlation matrix showed notable connections among these physicochemical properties of cellar mud (Fig. S2). The pH and dissolved inorganic carbon displayed a negative relationship with other properties, whereas the others were positively correlated with each other. Therefore, the results indicated that the nutrients such as carbon and nitrogen source in the cellar mud were enriched along with the long-term fermentation.
Table1 Physicochemical properties of cellar muds in different ages
Property
|
C3
|
C10
|
C30
|
C100
|
Water content (%)
|
28.53 ± 4.93a
|
37.05 ± 4.96b
|
30.80 ± 4.90a
|
36.82 ± 6.23b
|
Organic matter (%-TS)
|
7.78 ± 3.81a
|
11.17 ± 3.64b
|
13.61 ± 6.22b
|
16.72 ± 7.46c
|
pH
|
7.31 ± 0.96ab
|
7.72 ± 1.11a
|
6.53 ± 1.63c
|
6.75 ± 1.37bc
|
Electrical conductivity (μS cm-1)
|
491.33 ± 193.29a
|
786.97 ± 237.41b
|
741.33 ± 273.1b
|
1054.53 ± 381.90c
|
Total Carbon (%)
|
2.69 ± 1.72a
|
4.86 ± 1.74b
|
4.73 ± 3.19b
|
7.18 ± 4.61c
|
Total Nitrogen (%)
|
0.39 ± 0.20a
|
0.71 ± 0.19b
|
0.61 ± 0.32b
|
0.96 ± 0.50c
|
Carbon/Nitrogen ratio
|
6.65 ± 1.22a
|
6.71 ± 1.52a
|
7.63 ± 1.95b
|
7.14 ± 1.19ba
|
Dissolved organic carbon (g kg-1-TS)
|
4.11 ± 4.37a
|
7.41 ± 6.15a
|
12.06 ± 9.90b
|
17.00 ± 12.88c
|
Dissolved inorganic carbon (g kg-1-TS)
|
0.33 ± 0.20a
|
0.74 ± 0.43b
|
0.20 ± 0.20a
|
0.34 ± 0.38a
|
NH4+ (g kg-1-TS)
|
1.26 ± 0.46a
|
1.91 ± 0.67b
|
1.24 ± 0.49a
|
3.28 ± 1.24c
|
NO2- (mg kg-1-TS)
|
0 ± 0a
|
1.23 ± 3.28ba
|
0 ± 0a
|
1.66 ± 5.32b
|
NO3- (mg kg-1-TS)
|
0 ± 0a
|
0 ± 0a
|
1.62 ± 4.47ba
|
2.4 ± 7.39b
|
Na+ (mg kg-1-TS)
|
64.60 ± 28.73a
|
52.73 ± 23.7 a
|
61.40 ± 20.97a
|
113.62 ± 143.50b
|
K+ (g kg-1-TS)
|
0.96 ±0.41a
|
1.07 ± 0.38a
|
1.23 ± 0.35a
|
2.89 ± 0.88b
|
Cl- (mg kg-1-TS)
|
161.52 ± 51.79a
|
191.03 ± 41.32b
|
197.06 ± 54.91b
|
293.21 ± 76.05c
|
PO43- (g kg-1-TS)
|
0.29 ± 0.33a
|
0.28 ± 0.20a
|
0.82 ± 0.78a
|
2.29 ± 2.23b
|
SO42- (mg kg-1-TS)
|
34.23 ± 21.99a
|
22.68 ± 20.93ab
|
11.72 ± 15.24b
|
78.49 ± 44.82c
|
Lactate (g kg-1-TS)
|
4.17 ± 3.21a
|
5.98 ± 4.91ab
|
3.52 ± 4.62a
|
7.76 ± 7.17b
|
Acetate (g kg-1-TS)
|
2.83 ± 1.98a
|
3.21 ± 2.10a
|
3.00 ± 1.56a
|
5.05 ± 3.20b
|
Propionate (g kg-1-TS)
|
0.28 ± 0.37a
|
0.51 ± 0.66ab
|
0.81 ± 0.68b
|
0.59 ± 0.56b
|
Butyrate (g kg-1-TS)
|
0.53 ± 0.69a
|
1.16 ± 1.14b
|
1.32 ± 1.06b
|
1.55 ± 1.46b
|
Valerate (g kg-1-TS)
|
0.02 ± 0.07a
|
0.36 ± 0.89b
|
0.19 ± 0.2ba
|
0.11 ± 0.19a
|
Hexanoate (g kg-1-TS)
|
2.48 ± 3.69a
|
5.64 ± 4.07b
|
9.83 ± 6.47c
|
4.78 ± 4.78ba
|
Different lowercase letters indicate a significant difference between the cellar muds from different ages.
Metagenomic reconstruction and phylogeny of cellar mud rMAGs
Metagenomics analysis of 120 cellar mud samples yielded 4082 MAGs with a CheckM-estimated completeness ≥ 70 and contamination ≤ 5, which were both higher than the proposed MAG medium quality standard (Table S2) [49]. After dereplication, we retrieved 634 rMAGs including 569 bacterial and 65 archaeal rMAGs from the obtained MAGs (Table S3). These 634 rMAGs spanned 33 phyla within the bacterial and archaeal domains, as predicted by GTDBtk (Fig. 2a, Table S3). These rMAGs predominantly belonged to the phyla Firmicutes A (175 rMAGs), Bacteroidota (86),
Firmicutes G (49), and Firmicutes B (46) for bacteria; and phyla Halobacteriota (30), Thermoplasmatota (20), and Methanobacteriota (15) for archaea (Fig. 2a, Table S3).
Out of the 634 rMAGs, we assigned only 198 (31.23%) to known genera, while the remaining belonged to various genus- (162 rMAGs, 25.55%), family- (132, 20.82%), order- (64, 10.09%), class- (72, 11.36%), and phylum-level lineages (6, 0.95%) without cultured representatives (Fig. 2b). The overlap of rMAGs in the cellar mud samples from different ages is presented as Venn diagrams (Fig. 2c). The number of rMAGs decreased from 292 in C3 to 221 in C30, while it recovered to 309 in C100, thereby indicating a falling-rising richness change with cellar age. We found 78 rMAGs (12.30%) were shared in cellars across the four ages, which accounted for 68.23%, 72.40%, 79.09%, and 54.36% of rMAGs relative abundance for C3, C10, C30, and C100, respectively (Table S4, Fig. S3a). Among the 78 core rMAGs, phyla Bacteroidota (13 rMAGs), Halobacteriota (5), Firmicutes A (25), and Firmicutes G (5) had the highest relative abundance (Fig. S3a). Contrastingly, we detected 131, 102, 32, and 122 unique rMAGs in C3, C10, C30, and C100, respectively, which accounted for 14.21%, 6.77%, 3.75%, and 19.01% of the relative abundance, respectively (Table S5, Fig. S3b). These rMAGs were mainly affiliated with phyla Chloroflexota, Proteobacteria, and Bacteroidota (Fig. S3b).
The structure of the cellar mud microbial community
The taxonomic analysis showed that the bacteria (relative abundance ranged 64.90–75.38%) dominated the cellar mud community (Table S3, Fig. S4). However, the relative abundance of the archaeal community consistently increased (25.31% to 32.34%) with cellar age (Table S3, Fig. S4). The bacterial taxa in the cellars across the four ages were primarily dominated by the phyla Bacteroidota (16.30–38.50%), Firmicutes A (5.52–15.40%), Firmicutes G (2.51–18.58%), Firmicutes B (3.47–11.37%), and Chloroflexota (0.34–13.77%) (Fig. S4). The archaeal rMAGs belonged to three phyla, including Halobacteriota (18.65–24.30%), Methanobacteriota (4.09–11.48%), and Thermoplasmatota (0.35–1.88%) (Fig. S4). More specifically, the top 20 genera are shown in Fig. 3a, including Proteiniphilum, Petrimonas, Caprobacter, Syntrophaceticus, Sedimentibacter, Pseudomonas E, and uncultured -T78, -UBA4882, -UBA4871, -UBA4844, -UBA5389, -UBA4975, -Syner-03, -SR-FBR-E99, and -UBA4851 for bacteria; Methanoculleus, Methanobacterium A, Methanothrix,
Methanobacterium C, and Methanosarcina for archaea. Among these genera, Spearman correlation analysis revealed significant negative correlation between the relative abundance of Proteiniphilum (rho = -0.51, FDR < 0.001), Petrimonas (rho = -0.41, FDR < 0.001), Syner-03 (rho = -0.3, FDR < 0.001), and SR-FBR-E99 (rho = -0.64, FDR < 0.001) and the cellar age. However, the relative abundance of Methanobacterium A (rho = 0.80, FDR < 0.001), Caprobacter (rho = 0.44, FDR < 0.001), UBA4871 (rho = 0.23, FDR < 0.001), Methanosarcina (rho = 0.25, FDR < 0.001), Sedimentibacter (rho = 0.25, FDR < 0.001), and UBA4844 (rho = 0.32, FDR < 0.001) were positively correlated with the cellar age (Fig. 3b). Moreover, Spearman correlation analysis revealed considerable interactions between the physicochemical properties and the genera (Fig. S5). The genera Proteiniphilum, Petrimonas, T78, Methanothrix, Pseudomonas E, Syner−03, and SR−FBR−E99 negatively correlated with the cellar mud properties, including organic matters and inorganic ions. Contrastingly, Methanoculleus, Methanobacterium A, Caprobacter, Methanobacterium C, UBA4871, Methanosarcina, Syntrophaceticus, Sedimentibacter and UBA4844 displayed significant positive correlations with the nutrients.
Metabolic profiles of cellar mud rMAGs
To accurately reconstruct the metabolic profiles of rMAGs, we annotated the presence of the key enzymes and metabolic pathways (Table S6, Table S7) with a strict criterion that required the enzymes or pathways completeness ≥ CheckM-based rMAG completeness (Table S8, Table S9). The metabolic profile of the microbial community based on the KEGG module completeness is shown in Table S9, Fig. S6 and Fig. S7. The major functional genes and pathways identified in the abundant rMAGs (relative abundance > 0.1%) are shown in Fig. 4. Given the relative abundance of rMAGs, we defined metabolic capacity as the gene number or pathway presence × rMAG relative abundance to evaluate the metabolic potential roles these microbes played in the cellars (Table S10). We estimated the significance of the relationships between the metabolic capacity and cellar age using the Spearman correlation analysis (Table S11, Fig. S8).
Polymers metabolism
To decipher polymers degradation in the cellar mud ecosystem, we estimated the occurrence and diversity of peptidase, lipase, and CAZymes that catalyze the hydrolysis of macromolecules (Fig. 4, Table S8, Table S11). The rMAGs encoding the
highest average number of genes contributing to degradation of polymers in the metagenome belonged to the multiple phyla, including Firmicutes E, Firmicutes H, Acidobacteriota, Armatimonadota, Planctomycetota, Bacteroidota, and Verrucomicrobiota (Table S12). Considering the relative abundance of the communities, the phyla Bacteroidota (e.g. MAG246.1, MAG 254.1), Firmicutes A (MAG554.1, MAG545.1), Firmicutes G (MAG300.1, MAG109.1), and Armatimonadota (MAG394.1, MAG397.1) predominantly encoded CAZYmes, while Bacteroidota, Firmicutes A, Firmicutes G and Halobacteriota encoded most of the peptidase and lipase genes in the metagenome (Fig. 4, Table S12), thus suggested that these phyla may encompass the major polymers degrading microorganisms in this study. Notably, polymer degradation capacity of the rMAGs containing lipase (rho = -0.38, FDR < 0.001), glycoside hydrolase (GH, rho = -0.50, FDR < 0.001), carbohydrate esterase (CE, rho = -0.37, FDR < 0.001), carbohydrate binding module (CBM, rho = -0.46, FDR < 0.001), and polysaccharide lyase (PL, rho = -0.4, FDR < 0.001) significantly decreased consistent with the cellar age, while auxiliary activities (AA, rho = 0.46, FDR < 0.001) displayed positive association (Table S11, Fig. S8).
Carbon metabolism
Glycolysis/gluconeogenesis and pentose phosphate pathway were the central carbohydrate metabolic pathways within the cellar mud communities (Fig. 4, Table S9 and Table S10). Besides these, alcohol degradation and lactate degradation through acetyl-CoA were the most abundant carbon metabolism pathways in the microbiomes of cellar mud (Fig. 4, Table S10), where the relative abundance of rMAGs containing the two pathways was 43.64% and 51.00%, respectively (Table S8, Table S10). The relative abundance of lactate degraders significantly decreased (rho = -0.32, FDR < 0.001) with the cellar age, while the alcohol metabolism capacity remained stable during the long-term CSFL fermentation (Table S11, Fig. S8). The genera Petrimonas and Proteiniphilum affiliated to the phyla Bacteroidota, displayed the highest lactate breakdown capacity that decreased with the cellar age, whereas the genera Caprobacter, Sedimentibacter, and various uncultured genus belonging to Firmicutes A increased and became the main lactate utilizer (Fig. 4, Table S10). Though the alcohol degradation remained consistent with cellar age (Table S11), the dominant genera associated with alcohol utilization shifted from Petrimonas, uncultured SR-FBR-E99, and Syner-03 in the young cellars (C3, C10) to Pseudomonas E, Caprobacter, and uncultured UBA4871 in the old cellars (C30, C100) (Fig. 4, Table S10).
The anaerobic breakdown of organic materials could produce SCFA byproducts like acetate, propionate, and butyrate, whose breakdown is highly thermodynamically challenging [50]. SCFAs are important intermediates and the oxidation of syntrophic SCFAs with methanogenic partners was proposed as a key step in the anaerobic process [50]. Consistent with the elevated butyrate and acetate concentrations (Table 1, Fig. S1), the cellar mud microbiomes exhibited increased potential of syntrophic butyrate (rho = 0.20, FDR < 0.05) and acetate (rho = 0.28, FDR < 0.01) oxidation with cellar age, whereas propionate degradation potential decreased (rho = -0.24, FDR < 0.05) (Table S11, Fig. S8). We found 99 rMAGs that spanned across 67 genera, encoded genes involved in the butyrate oxidation pathway. The relative abundance of rMAGs encoding butyrate oxidation pathway components increased from 16.77% in C3 to 23.63% in C100 (Table S10). Petrimonas, uncultured UBA4871, UBA4844, and UBA5389 members dominated the butyrate oxidation group in all the cellars, whereas species affiliated within Pseudomonas E and uncultured T78 considerably increased in C100 (Fig. 4, Table S10). Propionate degradation group comprised 72 rMAGs classified into 50 genera (Table S8). The relative abundance of rMAGs encoding propionate degradation pathway components decreased from 23.38% in C3 to 16.46% in C100 (Table S10). Petrimonas, Proteiniphilum, and uncultured UBA4882 mainly contributed to the metabolic capacity of propionate oxidation (Fig. 4, Table S10). For syntrophic acetate oxidation, we found that 27 rMAGs spanning across 20 genera encoded all the genes for the Wood-Ljungdahl pathway (WLP), whereas the glycine cleavage pathway (GCP) encoded by 111 rMAGs spanned across 74 genera (Table S8). Moreover, the average relative abundances of rMAGs encoding the WLP and GCP genes were 3.01% and 10.61% within the cellar mud microbiome, respectively, thus suggesting that GCP may be the main pathway for syntrophic acetate oxidation (Table S10). Syntrophaceticus members, including the well-characterized syntrophic acetate-oxidizing bacteria Syntrophaceticus schinkii (MAG106.1) dominated the WLP group of cellar mud (Fig. 4, Table S8). With respect to GCP, we found that Sedimentibacter, uncultured UBA4844, and UBA4975 were abundant genera (Table S10). Additionally, we predicted multiple uncultured species (25 out of 27 rMAGs for WLP and 104 out of 111 rMAGs for GCP) to have acetate oxidation ability, thereby implying the abundance of novel syntrophic acetate oxidation bacteria within the cellar mud microbiome (Table S8).
We ubiquitously identified methanogenesis that generated methane as the final product in the archaeal rMAGs of this study (Fig. 4, Table S8). The H2-utilizing pathway we found in 37 out of 65 archaeal rMAGs, was the main methanogenesis pathway within the cellar microbiome, with the relative abundance of rMAGs encoding this pathway significantly increasing (rho = 0.35, FDR < 0.001) from 17.38% in C3 to 29.03% in C100 (Fig. S8 and Table S8). Methanoculleus (9 rMAGs), Methanobacterium C (5 rMAGs), and mixotrophic methanogen Methanosarcina members (9 rMAGs) mainly contributed to the hydrogenotrophic methanogenesis across the cellars of the four ages (Fig. 4, Table S8, Table S10). The relative abundance of Methanoculleus first increased from 8.78% in C3 to 17.79% in C10, and then decreased to 14.54% in C100, while the Methanobacterium C members were stable (2.61~2.30%). Notably, Methanobacterium A (two rMAGs) significantly increased from 0% to 7.82% with cellar age (Fig. 4, Table S8, Table S10). Since the increase of Methanobacterium A was the most significant change within not only the hydrogenotrophic methanogens but also the whole microbiome with increasing cellar age, we clustered the KEGG module completeness of archael rMAGs to explore the adaption mechanism of Methanobacterium A in the cellar environment (Fig. S6). We found that Methanobacterium A and Methanobacterium distinctively encoded the complete ectoine biosynthesis pathway (M00033), which is proven to be the protection system against environmental stress [51]. However, the role of this pathway behind Methanobacterium A adaption remains unknown. Although a previous study also reported the presence of ectoine biosynthesis genes in Methanobacterium using metagenomics and genomic comparison, no evidence was found supporting the existence of the ectoine synthesis in archaea [52]. The population of acetotrophic methanogens fluctuated from 9.32% in C3 to 5.07% in C100 with no significant correlation with cellar age. Methanothrix (2 rMAGs) and Methanosarcina (9 rMAGs) were the abundant acetotrophic methanogens within the cellar mud. We also identified methylated compounds (i.e., methanol, dimethylamine, and monomethylamine) metabolism producing methane in the communities (Table S8, Table S10). The methylotrophic methanogenesis capacity increased from 4.86% to 7.61% with the cellar age, with Methanobacterium C, Methanosarcina, and Methanobacterium A dominating this group (Table S8, Table S10). Additionally, we defined 20 rMAGs affiliated with phylum Thermoplasmatota as methylotrophic methanogens (Table S8, Table S10). Out of the 20 rMAGs, 17 belonging to the order Methanomassiliicoccales were uncultured.
Next, since fatty acids are the expected metabolites for CSFL fermentation, we identified the microorganisms contributing to fatty acids biosynthesis within the cellar mud. The metabolic potential of propionate biosynthesis significantly decreased from 21.36% in C3 to 9.74% in C100 within the cellar mud microbiome with increasing cellar age (Fig. S8 and Table S10). The genera Proteiniphilum and Petrimonas were the main contributors of propionate biosynthesis (Fig. 4 and Table S10). We detected the butyrate biosynthesis pathway in 230 rMAGs, thereby accounting for 34.21–36.92% of the mapped metagenomic reads obtained from cellar mud across the four ages (Table S8 and Table S10). Though the relative abundance of butyrate biosynthesizer was stable, the composition of this population significantly changed with the cellar age. Petrimonas, uncultured T78, and UBA4882 members dominated the butyrate biosynthesis group in C3, but decreased in C100, whereas Caprobacter, UBA4871, Sedimentibacter and Pseudomonas E increased in C100 (Table S8 and Table S10). Microbial fatty acid chain elongation occurs via two routes, the fatty acid biosynthesis (FAB) pathway and the reverse beta oxidization (RBO) pathway [53]. We observed a significantly increased abundance (8.70–25.59%) of genomes (176 rMAGs) encoding the FAB pathway within the cellar mud microbiome with increasing cellar age (Fig. S8 and Table S10). Particularly, we taxonomically assigned most of the rMAGs to uncultured UBA4882, UBA4871, Caprobacter, and Pseudomonas E (Fig. 4 and Table S8). However, for the RBO pathway, 67 rMAGs accounted for 4.40–7.50% of the whole community across the cellars of four ages, with most of them being assigned to Pseudomonas E, uncultured UBA4851, and UBA5314 (Table S8 and Table S10). Therefore, these findings imply that the FAB is vital in the fatty acid chain elongation in the cellar mud microbiome.
Energy conservation capacity
Metabolism under anaerobic conditions is complemented by substrate oxidation with electron balance. The interspecies transfer of electrons from formate or hydrogen to the methanogens is critical for maintaining a low H2 concentration for microbial growth [54]. With a focus on electron transfers, we used a manually curated energy-conserving marker gene database to annotate the rMAGs (Fig. 4, Table S13). We found 1781 hydrogenases in 592 out of the 634 rMAGs, which indicated that the cellar mud microbiome has a high potential for hydrogen metabolism (Table S13). We classified these hydrogenases were classified into 12 types against the HydDB database, including nine [NiFe] and three [FeFe] classes (Table S13). The [FeFe] hydrogenases we detected only in the bacterial rMAGs were putatively involved in fermentative H2-evolution or electron-bifurcation (group A). We predicted multiple [NiFe] hydrogenases distributed in the archaeal rMAGs and their activities as bidirectional H2 consumption/production (group 3) and H2 uptake (group 1 and group 4). Consistent with the change of relative abundance for bacteria and archaea, the metabolic capacity of the [FeFe] hydrogenases linked to H2 production significantly decreased (rho = -0.4, p < 0.001), while those of the [NiFe] hydrogenase associated with H2 utilization was quite the opposite (Fig. S8, Table S11). Therefore, the result demonstrated the potential of the community level microbial auto-regulation of H2 balance in the long-term fermentation of CSFL.
Over two-thirds of the bacterial rMAGs (497 out of 569) contained [FeFe] hydrogenase linked to H2 production (Table S13). Particularly, most of the rMAGs affiliated to the fermentative phyla, including Bacteroidota, Firmicutes A, Firmicutes B, and Firmicutes G encoded H2 producing hydrogenases (Hyd, HydABC, and HndABCD). The rMAGs of the hydrogenotrophic methanogens (phyla Halobacteriota and Methanobacteriota) and fatty acids metabolizers (phyla Firmicutes A and Firmicutes B) encoded the [NiFe] hydrogenase associated with H2 utilization (Fig. 4 Table S13). In the archaeal community, H2-dependent methanogenic rMAGs possessed genes for H2 oxidation via reverse electron transport (EhaA-T, EhbA-T, and EchA-F), electron bifurcation (F420-non-reducing [NiFe] hydrogenase MvhADG and heterodisulfide reductase HdrABC), and F420 reduction (FrhABG) (Table S13). Fatty acids metabolic rMAGs encoded EchA-F, cytochrome b-linked NiFe hydrogenase (HyaABC or HybABCO), and a cytosolic NiFe hydrogenase (HoxEFUHY) (Table S13).
Regarding the formate metabolism, most of the bacterial rMAGs (524 out of 569) harbored formate dehydrogenase, including NADH-dependent formate dehydrogenase (FdhAB), putative NADPH-dependent formate dehydrogenase (FdpAB), ferredoxin-dependent formate dehydrogenase (FdhH), Hyc-linked formate dehydrogenase (FdhA-HycB-G), and putative NAD+-dependent electron-bifurcating complex (FdhA-hydBC: formate dehydrogenase organized with HydBC-related subunits) (Table S13). Multiple rMAGs affiliated to the phyla, including Firmicutes A (24 rMAGs), Firmicutes B (29), Firmicutes G (15) and Synergistota (9), encoded at least 4 types of formate dehydrogenase, thereby suggesting that they may have high capability to generate formate (Table S13). The generated formate served as an electron acceptor in the methanogenesis process, with the hydrogenotrophic methanogens (35 out of 65 archaeal rMAGs) encoding FdhAB for formate oxidation (Table S13). Similar to the hydrogenases, the metabolic capacity of the bacterial formate dehydrogenase significantly decreased (rho = -0.31, p < 0.001), whereas that of the archael formate dehydrogenase increased (Fig. S8 and Table S10). However, the average metabolic capacity of the bacterial formate dehydrogenase (191.10% ± 26.50%) was significantly higher than that of the bacterial [FeFe] hydrogenases (129.01% ± 25.96%) across the cellars in four ages, thus suggesting that we need the formate for interspecies electron transfer to complete with H2 (Fig. S8 and Table S10).
We also identified other electron transfer modules (e.g., RnfA-G: Rnf electron transport complex and NfnAB: NADH dependent NADP: Fd oxidoreductase) and direct interspecies electrons transfer (putative e-pili and putative cytochrome c). Diverse phyla and niches encoded enzymes involved in supporting the thermodynamically challenging reactions and metabolism (Fig. 4, Table S13). Coexistence of various electron transfer modules in the pit mud community highlighted the importance of having multiple energy conservation routes for CSFL fermentation.
Nitrogen metabolism and sulfur metabolism
We investigated the nitrogen and sulfur metabolism pathways among the cellar mud microbial community. We observed a stable high potential (28.33–31.18%) of nitrogen fixation, with the abundant rMAGs being assigned to methanogens Methanoculleus, Methanothrix and fatty acid metabolizer UBA4871, Caprobacter, and UBA4844 (Fig. 4, Fig. S8, Table S8 and Table S10). We identified and assigned two rMAGs encoding the nitrification pathway to uncultured JAAYCB01 and Methylocystis, with an increased abundance from 0 in C3 to 1.86% in C100. Additionally, we detected the low abundance of the denitrification pathway (0–0.21%) across the cellar mud samples (Fig. 4, Fig. S8, Table S8 and Table S10). The relative abundance of the nitrate reduction pathway fluctuated from 0.55% to 4.51%, with most rMAGs being assigned to Proteiniphilum, uncultured JAAYCB01, and UBA5261 (Fig. 4, Fig. S8, Table S8 and Table S10). For sulfur metabolism, we detected the complete bi-directional sulfate reduction/oxidation pathway and the thiosulfate oxidation pathway (SOX system). Furthermore, we observed a significantly higher sulfate reduction/oxidation potential (8.02%) in C100 as compared to the other cellars (0.54–0.75%), with Pseudomonas E, uncultured JAAYCB01, Thermoguttaceae, and UBA7640 members contributing to this increase (Fig. 4, Fig. S8, Table S8, and Table S10). We also observed a reduced thiosulfate oxidation potential (0.58%–0) from C3 to C100. Genera Thiobacillus and Allorhizobium members were abundant for thiosulfate oxidation (Fig. 4, Fig. S8, Table S8, and Table S10).