Deep-sea in situ incubation of OM-decomposing bacteria
The in situ incubation experiments were separately carried out on the seafloor in the South China Sea (SCS, the largest marginal sea of the West Pacific Ocean), the deep-sea basin beside the Southwest Indian Ridge in the Indian Ocean (IO, open sea), and a flat-topped seamount in the West Pacific Ocean (PO, open sea) at 3758 m, 4434 m, and 1622 m water depths for 375, 117, and 348 days (Figs. S1-2). Several kinds of natural materials rich in polysaccharides, proteins, or lipids were selected as fast-sinking POM (Table S1), loaded into 50 mL tubes and then placed into titanium alloy incubation chambers (designated TIC for the treatment incubation chamber). Meanwhile, two incubation chambers without OM supplementation (designated CIC for the control incubation chamber) were set up for collecting in situ seawater. Finally, all of the TICs and CICs were delivered to the deep sea floor with the deep-sea in situ microbial incubator (DIMI) device. The in situ temperature and salinity of seawater at those sites were approximately 2.39–2.40°C and 34.62–34.65‰, respectively (Table S1). After being recovered from the seabed and brought to the shipboard laboratory, all incubation tubes, even those with wood chips, exhibited obvious microbial growth indicated by turbidity and colour variation (Fig. S3); most samples smelled of hydrogen sulphide, and under a microscope, the field of view appeared with bacterial cells (Table S1 and Fig. S3).
Bacterial diversity and its environmental determinants of the OM-decomposing microbial assemblages
To profile the bacterial communities colonizing different types of OM, all 39 enrichments in TICs were subjected to community composition analysis via Illumina high-throughput sequencing and compared with the in situ seawater collected by CICs. In general, the α-diversity indexes (Shannon and Chao indexes) indicated that the bacterial diversity of the enrichments within the TICs was significantly lower than that of the in situ seawater in the CICs (Fig. 1a and 1b, P < 0.01). This indicated that bacterial communities were significantly enriched in situ by the OM supplements.
Principal coordinate analysis (PCoA) revealed that the structures of the enriched communities were significantly different from those of the seawater in CICs and were quite diverse at the three sites, separated by both organic substrate and geographic location (Fig. 1c). Additionally, Adonis analyses based on the β-diversity index suggested that the enriched bacterial community structures were influenced by multiple factors, including geographic location (R2 = 0.20, P = 0.001), the type of OM (R2 = 0.13, P = 0.001), the enrichment time (R2 = 0.094, P = 0.001), and other environmental factors, including temperature, water depth and salinity (for all factors, R2 < 0.11, P < 0.0012) (Table S2). Among these factors, geographic location and OM type were the two most important determinants of the variation in microbial community structures.
Bacterial composition of deep-sea in situ enrichments with natural organic materials
Across the deep-sea in-situ enrichments, the dominant bacteria were affiliated with the classes Bacteroidia (average abundance of 28.4%), Gammaproteobacteria (20.4%), Deltaproteobacteria (13.5%), Spirochaetia (12.0%), Campylobacteria (8.4%), Clostridia (7.2%) and Fusobacteriia (6.1%) (Fig. 2a). At the family level, the dominant bacterial groups were shared across these deep-sea enrichments even though the sites were far away on the oceanic scale (Fig. 3 and Fig. S4). The bacteria were affiliated with the families Marinifilaceae, Spirochaetaceae, Psychromonadaceae, Vibrionaceae, Moritellaceae, Desulfobulbaceae, Desulfobacteraceae, Family_XII_o_Clostridiales, Fusobacteriaceae, Arcobacteraceae, Sulfurospirillaceae and Sulfurovaceae (Fig. 3 and Fig. S4). A total of 2,457 bacterial operational taxonomic units (OTUs) were retrieved from these 39 enrichments (Table S4). However, only 144 OTUs were shared among the three sites (Fig. 2b), indicating obvious variation among geographic locations at the OTU level.
Most of the dominant bacterial members listed above were anaerobes and constituted the main components of the microbial assemblages. In the in situ seawater collected by CICs, only the putative anaerobic bacteria Woeseiaceae/JTB255 and SAR202 were recovered, which have been recognized as cosmopolitan and abundant core members of deep-sea surface sediments and benthic water[21–23], respectively. However, they were not detected in the enrichments with newly input OM (Fig. 3). Thus, special groups of heterotrophic bacteria were stimulated by the newly input OM, which were faster-growing and more competitive than the oligotrophs.
Keystone bacterial species involved in OM-decomposing in different in situ enrichments
To discriminate the keystone species enriched by different types of OM, we analysed the distribution of OTUs of each dominant family. The results showed that the bacteria within a family were divergent at the OTU level in response to different OMs (Fig. 4). Evidently, bacteria of the family Marinifilaceae were the most predominant, thriving in all OM enrichments and showing the highest abundance among all bacterial communities of 26.3% on average. Interestingly, six OTUs of this family exhibited a preference for proteins, polysaccharides, and lipids. In detail, OTU6861 was mainly dominant in the consortia of substrates enriched with polysaccharides, such as wood debris, seaweed, and wheat bran, in the SCS and PO (Fig. 4). OTU4252 and OTU5909 both dominated the consortia of substrates enriched with proteins (Fig. 4), while OTU9995 occurred in all types of substrates at site PO, although it was predominant in proteinaceous enrichments. Moreover, OTU5914 not only occurred as the most predominant member in proteinaceous enrichments at the SCS site but also among all the enrichments of different OM types at the PO site.
Members of Spirochaetaceae are typical intestinal anaerobic bacteria that play an important role in the digestion of breakdown products from cellulose and hemicellulose in the termite gut. In this study, Spirochaetaceae was the second predominant group in the enrichments, and a total of 28 OTUs were retrieved. However, most of the OTUs were obviously separated from those known to be from the animal intestine in the phylogenetic tree and formed several taxonomic branches with those from marine sediment, wood-fall ecosystems, and our coastal enrichments with OM (Fig. S5). Therefore, members of Spirochaetaceae displayed obvious ecotypic differentiation. Notably, one member (OTU5914) was noteworthy for its role in degrading polysaccharides and proteins in the deep sea in situ, which occurred in most polysaccharide- or protein-rich enrichments, with relative abundances of 90.0% and 52.0% in the wheat bran enrichment at the PO site and shrimp muscle enrichment at the SCS site, respectively (Fig. 4).
In addition, the heterotrophic bacteria Moritellaceae, Psychromonadaceae, and Vibrionaceae[3, 26, 27] of Gammaproteobacteria also constituted the dominant members. The diversity within individual families varied with substrate and location at the OTU level (Fig. 4), which is described in detail in the text of the Supplementary Information and shown in Figs. S6-S8.
Members of Desulfobulbaceae are typical psychrophilic sulphate-reducing bacteria (SRBs) that can use the most common microbial fermentation products (e.g., acetate, propionate, butyrate, lactate, and hydrogen) in marine sediments as an energy source, coupled with sulphate, sulphite, and/or thiosulfate reduction. Additionally, a recent report showed that the strain KaireiS1 belonging to the Desulfobulbaceae family could utilize hydrogen as an electron donor coupled with sulphate reduction. Here, a total of 29 OTUs of Desulfobulbaceae were found in the in situ enrichments (Fig. 4). In the phylogenetic tree, they neighboured those from marine sediment, hydrothermal vents, and chemosynthetic wood-fall or whale-fall ecosystems (Fig. S9). They did not show a preference for certain OM substrates (Fig. 4). However, the dominant OTUs in this group varied with geographic location, and oddly, they were nearly absent at the IO site (Fig. 4).
Sulphur-oxidizing bacteria (SOBs) of Arcobacteraceae and Sulfurovaceae, which were enriched concomitantly with SRBs, are frequently detected in deep-sea hydrothermal chimneys, vent plumes, and vent animals as primary producers. However, phylogenetic analysis showed that the dominant members of Arcobacteraceae (represented by OTU9977, OTU4225, OTU6340, and OTU1752) were obviously separated from those with hydrothermal origins but clustered onto a branch with those from the deep-sea wood-falls (Fig. S11). The same result was observed for the most abundant bacterium of Sulfurovaceae, OTU4956 (Fig. S12). This indicates that SOBs in chemoorganoheterotrophic ecosystems display obvious ecotypic differentiation from those in chemolithoautotrophic ecosystems.
Function prediction of the deep-sea bacterial communities
To understand the relationships between bacterial communities and OM types, two different methods, PICRUSt and FAPROTAX, were applied to predict the functions of the enriched bacterial consortia. The PICRUSt analysis based on KEGG module clustering suggested that the properties of OM contributed more to microbial functional diversity (R2 = 0.185, P = 0.01) than other factors, such as geographic location, enrichment time, temperature, depth and salinity (Fig. 5 and Table S3). Protein and polysaccharide enrichments formed separate clusters, while lipid enrichments did not form any independent clusters (Fig. 5).
Furthermore, the enrichment patterns of key metabolic pathways of the bacterial communities varied among OM substrates. For example, several key modules for carbohydrate metabolism were obviously enriched in polysaccharide enrichment, such as glycolysis (KEGG modules: M00001 and M00002), pentose phosphate pathway (M00007), Entner-Doudoroff pathway (M00008), pectin degradation (M00081), D-galacturonate degradation (M00631), and galactose degradation (M00632) (Table S5). Intriguingly, the module for nitrogen fixation was selectively enriched in the nitrogen-deficient enrichments of polysaccharides and lipids. The relative abundance of the nitrogen fixation module (M00175) in polysaccharide enrichments was 2.7 times higher than that in protein enrichments (Table S5). In lipid enrichments, the modules thought to be responsible for fatty acid metabolism were significantly enriched (Table S5), such as beta-oxidation (M00087), acyl-CoA synthesis (M00086), the ethylmalonyl-CoA pathway (M00373 and M00774), and acylglycerol degradation (M00098).
In addition, the FAPROTAX predictions revealed that the top ten functional groups were chemoheterotrophy, fermentation, respiration of sulphur compounds, sulphate respiration, hydrogen oxidation, sulphur respiration, nitrate reduction, thiosulfate respiration, oxidation of sulphur compounds, and iron respiration. These results indicated that the bacterial decomposition of newly input OM during the “hot time” in the deep sea was an anaerobic process achieved via fermentation and diverse forms of anaerobic respiration, and these processes were tightly coupled with the cycles of nitrogen, sulphur, and hydrogen (Fig. S16).
Metagenomic and metatranscriptomic analyses of the enriched bacterial communities
To confirm the response of the bacterial assemblages to different OM inputs, metagenomic and metatranscriptomic investigations were further conducted to retrieve the abundance and expression of the functional genes potentially involved in carbon, nitrogen, and sulphur metabolism. To this end, five enrichments of P-MX (wood chips), P-FP (wheat bran), P-YR (fish muscle), P-YL (fish scales), and P-YY (fish oil) were selected for metagenomic and metatranscriptomic sequencing based on the analysis results using MicroPITA software (Table S6). A large number of genes responsible for hydrolysing polysaccharides and proteins and fatty acid oxidization were found in these metagenomes (Fig. 6 and Table S7). Approximately 1,407 to 3,034 genes that matched hits in the Carbohydrate-Active enZYmes (CAZy) database were detected in these metagenomes. The CAZy number was the highest in polysaccharide incubations but was the lowest in protein incubations.
In contrast, the genes involved in protein metabolism ranked highest (5406 genes) in the enrichment with collagen-rich fish scales (P-YL). These results indicated that the corresponding genes involved in the degradation of polysaccharides and proteins were selectively enriched in different OM substrates. In these metagenomes, genes encoding peptidases, oligopeptides, amino acid transporters, aminotransferases, and amino acid dehydrogenases for protein hydrolysis and their further metabolism were obviously enriched (Fig. 6 and Table S7).
Moreover, the genes involved in fermentation (e.g., production of formate (pflD), ethanol (aldH and adhE), acetate (acdAB and ackA), propionate (mmdAC and pccB), lactate (lldF and dlD), and hydrogen (hydABC, hydS, and hydM)) were enriched in all OM enrichments (Fig. 6 and Table S9). These simple compounds are the major end products of OM fermentation, which subsequently drive sulphate reduction by SRBs.
Similar to those involved in carbon metabolism, the genes involved in sulphur metabolism were also selectively enriched. The following processes of sulphur metabolism were detected in all five metagenomes: oxidation genes of hydrogen sulphide (sqr), thiosulfate (Sox complex system), sulphite (soeABC and suoX), and various dissimilatory reductions of sulphite, tetrathionate, and thiosulfate, in addition to the respiration of dimethyl sulfoxide (Fig. 6 and Table S9).
With respect to nitrogen metabolism, we detected the enrichment of genes involved in both nitrogen fixation (nifDKH) and nitrogen reduction, such as dissimilatory nitrate reduction (napAB), nitrite reduction (NirA and narB), nitric oxide reduction (norBC), and nitrous oxide reduction (nosZ), with variation in abundance in these enrichments (Fig. 6 and Table S9). More intriguingly, the metagenome-assembled genomes (MAGs), possessing nitrogen fixation-related genes, were mostly binned from polysaccharide and lipid metagenomes. These MAGs were mainly affiliated with Marinifilaceae, Spirochaetaceae, Psychromonadaceae, Desulfobacteraceae, and Desulfobulbaceae (data not shown).
To confirm that these metabolic pathways were active in situ, metatranscriptomic sequencing analyses were also performed with the five aforementioned enrichments. Approximately 87.0–97.0% of filtered paired reads were mapped to their own metagenomic scaffold dataset, indicating a high utilization rate of metatranscriptomic reads (Table S6). Metatranscriptomic data showed that in the polysaccharide enrichments, such as that with wood chips, the total transcript per million reads (TPM) values of genes encoding glycoside hydrolase (GH) were at least 4 times than those in other enrichments (Fig. 6 and Table S7-S8). In contrast, the total TPM values of peptidase-encoding genes were obviously higher in both fish scale and fish muscle enrichments. The difference in the transcriptional pattern of genes for polymer hydrolysis between polysaccharide and protein enrichments reconfirmed the active response to corresponding organic substrates in situ (Fig. S17). For example, glycoside hydrolase (families GH13, GH3, GH32, GH57, and GH43) was enriched and expressed in the wood chip and wheat bran enrichments (Table S8), while protein metabolism-related genes, such as those encoding peptidase and aminotransferase, were enriched and expressed in the fish scale and fish muscle enrichments.
In addition, the genes associated with nitrogen and sulphur metabolism, as well as organic fermentation, were also actively transcribed in situ. In particular, the genes involved in dissimilatory sulphite reduction in protein (P-YR) and lipid (P-YY) enrichments had higher transcriptional levels than those in other enrichments. The genes encoding Sqr, NapA, NorB, and NosZ also showed high transcriptional activity in enrichments with wood chips (P-MX) and fish scales (P-YL). Ethanol and acetate production genes were more actively transcribed in enrichments with wheat bran (P-FP) and fish scales (P-YL). Furthermore, genes involved in hydrogen production via fermentation were actively transcribed in wood chip (P-MX), wheat bran (P-FP), and fish scale (P-YL) enrichments (Fig. 6 and Table S9).
Microbial co-occurrence networks within the bacterial assemblages of different OMs
Since the functional differences of enriched microbial communities were mainly affected by the properties of the organic substrates (Fig. 5 and Table S3), to further understand the co-occurrence relationships among bacteria during the process of OM degradation, we constructed three co-occurrence networks for polysaccharide-, protein-, and lipid-enriched communities, respectively (Fig. 7). The complexity of the three network diagrams was high, and there were two modules in each network diagram (Fig. 7). Each network diagram consisted of 58 to 75 OTUs, mainly belonging to the dominant families mentioned above (Fig. 7 and Table S10). The network results reconfirmed that these dominant microorganisms were the key members and played a pivotal role in the establishment of OM-degrading consortia in the deep sea.
The co-occurrence relationships of the bacterial OTUs were closely related to the enrichment substrates. The relationships among bacteria within the communities of proteinaceous substrates were negative in 90.7% of cases (the relationship lines in red) (Fig. 7b and Table S11). However, cooperative relationships among bacteria dominated in the communities in the polysaccharide and lipid incubations (the cross lines are mainly green) (Fig. 7a, Fig. 7c, and Table S11). This may be because nutrients in proteinaceous substrates are replete and full of not only nitrogen, carbon, and phosphate sources but also other nutrients, such as vitamins; in contrast, many nutrients are quite limited in cellulose, lignin, and lipids. Therefore, microorganisms in proteinaceous enrichments showed competitive relationships and were more independent from each other. In contrast, the other two kinds of consortia displayed cooperative relationships to maintain nutrient balance and sustain the whole community. For example, in the polysaccharide substrates, the potential OM degraders Marinifilaceae, Spirochaetaceae, Vibrionaceae, and Moritellaceae were generally positively correlated with SRBs, SOBs, and Fusobacteriaceae. At the OTU level, the co-occurrence of the keystone species is described in detail in the Supplementary Materials.