Hadal sediment geochemistry and deep metagenomic sequencing
Sediment samples were collected using two deep-sea hadal landers from the seafloor (depth of 10840 meters) at the Challenger Deep (142° 21.7806′ E, 11° 25.8493′ N). The sediment samples were dissected into three depth segments, i.e., the surficial segment MT-1 (sediment depth 0–5 cm), the mid-segment MT-2 (5–10 cm), and the deep segment MT-3 (10–14 cm). Geochemical measurements on the samples were taken either onboard the ship ZhangJian or inland laboratories using preserved samples. The contents of total organic carbon (TOC) and total nitrogen (TN) in the sediment were measured between 0.49 and 0.55 (wt%) and between 0.05 and 0.06 (wt%), respectively (Additional file 1: Table S1). They are within the previously reported values for the sediment of the Challenger Deep [13]. The values of δ13C (− 21.41 to − 21.53‰) and δ15N (5.42 to 6.69‰) were within the ranges of the commonly observed values for marine organic matter [17]. This agrees with the results of recent studies that marine algae were the dominant source of sedimentary organic matter in the southern Mariana Trench [18, 19]. For nutrient ions, while the porewater SO42− concentrations were constant throughout the three segments, those of NH4+, PO43−, and NO2− were trending upward with increasing sediment depth (Additional file 2: Table S2). The NO3− concentrations, on the other hand, decreased with depth. The dissolved major and trace elements remained relatively homogenous with little variations among the three segments (Additional file 3: Table S3 and S4).
To explore the full microbiome structure in the Challenger Deep sediment, we performed deep metagenomic sequencing on the sediment samples which was designed with enhanced sensitivity, to capture the genetic contents of microbes with low abundances. Note that we took steps to minimize the impact of sample temperature increase by swiftly processing and freezing sediment samples, preserving the big-picture characteristics of the microbiome for metagenomic analysis. Three or more independent Illumina libraries were generated for each segment, and on average 22.6 Gb sequence data were obtained from each library, generating a total of 248.65 Gb raw sequence data (Additional file 5: Table S5). Metagenomic co-assembly was carried out using all the clean data, which resulted in a metagenome of 6.65 Gb in total length with ~ 6.21 million contigs and an N50 of ~ 1.14 kb.
To reveal the compositions of the hadal sediment communities, taxonomic profiling analysis was performed on the metagenomics sequences using the kaiju program and the NCBI-nr library [20, 21]. The relative sequence abundance for bacteria and archaea in the sediment accounted for 82.07% and 6.23%, respectively, whereas that for microeukaryotes and viruses was 0.69% and 0.12%, neither of which has been reported for the Challenger Deep sediment before. Different from previous studies based on rRNA gene PCR-tag sequencing, our approach estimated the sequence abundance values via the same workflow, generating scores directly comparable within the community scope.
At the phylum level, the most abundant components were Proteobacteria, Chloroflexi, Actinobacteria, Thaumarchaeotathat, Planctomycetes, Firmicutes, Bacteroidetes, Gemmatimonadetes, Acidobacteria, and Gemmatimonadetes, belonging to either bacteria or archaea (Fig. 1A). Our results agree with previous studies relying on 16S rRNA PCR-amplicon sequencing, in which Alphaproteobacteria, Chloroflexi, and Gemmatimonadetes were the most abundant bacteria, and Thaumarchaeota the most abundant archaea found in the Challenger Deep sediment habitat [16, 22]. However, the relative abundance of Thaumarchaeota varied largely between the different studies of the Mariana Trench habitats, ranging from 0.67 to 67% [23] and from 0.5 to 40% [22]. In a non-Mariana system, the Yap Trench, while Thaumarchaeota was similarly enriched, the abundance of Chloroflexi, Actinobacteria, Planctomycetes, and Gemmatimonadetes in its sediment microbiome was significantly lower than that of Mariana Trench sediment habitats [24]. We found that the overall relative abundance of archaea decreased with sediment depth, similar to findings in a previous study [13]. However, the relative abundance of both the microeukaryotes and marine viruses increased with sediment depth. The most abundant eukaryotes were Opisthokonta (Fungi), Alveolata, Stramenopiles, and Rhodophyta. Note the top eukaryotic phyla, Ascomycota and Basidiomycota, ranked the 16th and 17th overall in sequence abundance (Fig. 1A). We found MT-1 and MT-2 had consistent microbial compositions compared to those of MT-3, suggesting a geochemical boundary separated them at a 10-cm depth that was previously shown to limit oxygen access in hadal sediment [25]. Albeit in a relatively low abundance, the eukaryotic and viral components were unraveled for the first time in the Challenger Deep sediment, and as integral parts of the hadal biosphere, their ecological significance was further addressed below.
Phylogenetic diversity of the hadal sediment microbiome
Metagenomic binning on the assembled metagenome and filtering resulted in 178 quality draft genomes, i.e., MAGs (metagenome-assembled genomes) (Additional file 6: Table S6). These MAGs were at or above the medium-quality standards for draft genome, which required > 50% completeness and < 10% contamination (“Materials and method”) [26]. Furthermore, among them, 101 draft genomes were > 70% completeness, and 22 were > 90% completeness. The MAGs represented the most diverse metagenome reconstructed from the hadal sediment biosphere, comprising members from 26 phyla or candidate phyla. Sixteen of them had MAGs found from the Changer Deep sediment for the first time, i.e., Gemmatimonadetes (8), Verrucomicrobia (8), Nitrospinae (5), Elusimicrobia (1), Ignavibacteriae (1), Calditrichaeota (4), candidate phyla including Zixibacteria (4), Ca. Hydrogenedentes (2), Ca. Dependentiae (1), Ca. Diapherotrites (1), Ca. Doudnabacteria (2), Ca. Latescibacteria (2), Ca. Omnitrophica (3), Ca. Marinimicrobia (2), and unclassified bacteria including Bacterium SM23_31 (2), Ca. Sumeriaeota (2) and a newly defined phylum UBP7_A (1) [27]. In comparison, previous studies reported eleven and thirty MAGs that were co-assembled by combing water and sediment microbial sequences from the Mariana Trench and were affiliated with three and twelve phyla, respectively [28, 29]. Besides, using the single-cell sequencing approach, twelve single-cell amplified genomes (SAGs) were generated for Parcubacteria from the Mariana Trench sediment [30]. Note that these earlier works had fewer MAGs despite co-assembly by mixing sample data from multiple sources. Thus, the deep metagenomics approach has significantly enhanced the coverage and sensitivity of the hadal microbiome.
To illustrate the taxonomic diversity of the hadal sediment microbes, a phylogenetic tree for the 178 MAGs was constructed (Fig. 1B), using 43 conserved single-copy, protein-coding marker genes [31, 32]. A substantial number of uncultured microbial lineages were uncovered and classified with the phylogenetic analysis. The main bacterial groups included Alphaproteobacteria (22 MAGs), Gammaproteobacteria (19), Chloroflexi (19), Planctomycetota (14), Bacteroidota (10), Actinobacteriota (10), Gemmatimonadetes (8), Verrucomicrobia (6), Nitrospinae (4), as well as Elusimicrobia (1), Ignavibacteriae (1), Nitrospirae (1), and Calditrichaeota (4). Notably, bin.97 formed a monophyletic clade in the phylogenetic tree and was distant from the PVC groups that include Planctomycetes (Fig. 1B). It was closely related to Planctomycetales bacterium 4484_113 based on analysis using GTDB-Tk [33], for which the Average Nucleotide Identity (ANI) to bacterium 4484_113 stood at 63.86%. Bacterium 4484_113 was originally defined as unclassified Planctomycetes, but a recent study separated it from Planctomycetes to form a new phylum UBP7_A [27]. Thus, our study uncovered a likely second member of UBP7_A from the Challenger Deep sediment habitat.
The main archaea groups included Ca. Woesearchaeota (3 MAGs), Thaumarchaeota (4), and Ca. Diapherotrites (1) (Fig. 1B). Note that Ca. Diapherotrites, represented by bin.150, was found for the first time in the Challenger Deep sediment. Intriguingly, bin.150 was situated as an outgroup to both Thaumarchaeota (TACK group) and Woesearchaeota, and had a closer relationship with bacteria in the phylogenetic tree (Fig. 1B). Based on analysis using GTDB-Tk, bin.150 was close to the unclassified Diapherotrites archaeon UBA493 that is affiliated to DPANN group. Its ANI to UBA493, the closest relative, was 72.19%. We placed bin.150 with additional members of Ca. Diapherotrites for phylogenetic analysis and found bin.150 formed a long-branch that represents a specialized Diapherotrites clade in the Challenger Deep sediment (Additional file 17: Fig. S1). A previous study reported that as a member of Ca. Diapherotrites, Ca. Iainarchaeum acquired anabolic genes from bacteria via horizontal gene transfer [34]. The same reason may explain why bin.150 had a close relationship with bacteria in our phylogenetic results. So by reconstructing the largest metagenome of the hadal sediment biosphere, we recovered representative genomes of all major prokaryotic lineages previously identified by 16S rRNA gene amplicon-based surveys for the Challenger Deep sediment habitat [35], providing valuable references for us to further look into details of hadal microbiome regarding genetic diversity, metabolic functions, and symbiotic relationship. While our analyses were based on constructed MAGs, we acknowledged the missing pathway components that are likely attributed in part by the incomplete assembly, and the many assembled sequences of unknown molecular functions that are knowledge gaps to be bridged with new means [36].
Versatile metabolic function of the hadal sediment microbiome
Heterotrophy vs. autotrophy
To investigate the metabolic potential of each component, the 178 MAGs reconstructed from the Challenger Deep sediment microbiome were assigned with metabolic functions based on KEGG annotation. To determine microbes’ lifestyle, the MAGs were first investigated for heterotrophic potential and were found to contain genes involved in various pathways for degradation of carbohydrates (all MAGs), hydrocarbons (34 MAGs affiliated with 9 phyla), and aromatic compounds (146 MAGs affiliated with 24 phyla) (Fig. 2 and Additional file 7: Table S7). It is likely that sinking particulates from the upper ocean or terrestrial inputs, partly due to the funneling effect and earthquake-inducing landslides, were the source of the organic matter in the deepest habitat [37].
On the other hand, we found 69 out of 178 MAGs (39%; 16 phyla) contained pathways for inorganic carbon fixation. Six different CO2 fixation mechanisms were found to involve various microbes (Fig. 2). Among them, 3-hydroxypropionate bi-cycle (3-HP) was the most prevalent in 43 MAGs, compared to the Calvin-Benson-Bassham Cycle (CBB or Calvin cycle) (11 MAGs), Wood-Ljungdahl pathway (WL pathway) (6 MAGs), the reverse TCA cycle (rTCA) (2 MAGs), the 3 hydroxypropionate-4 hydroxybutyrate cycle (3-HP/4-HB cycle) (6 MAGs), and methanogenesis (16 MAGs). Previously, 3-HP was found in Chloroflexaceae, Alpha- and Gammaproteobacteria, and SAR202 [38, 39]. In our study, the 43 MAGs detected to have the 3-HP pathway were taxonomically assigned to eight phyla (31%) (Fig. 2 and Additional file 7: Table S7). They include the unclassified Chloroflexi bacterium RBG_16_64_32, Alpha-, Delta-, and Gammaproteobacteria, Nitrospinae, Acidobacteria, Actinobacteria, Calditrichaeota, Ca. Hydrogenedentes, and Gemmatimonadetes.
For other CO2 fixation pathways, the key enzymes in the Calvin Cycle, ribulose-bisphosphate carboxylase (rbcS), and phosphoribulokinase (prkB) were detected in five phyla (Fig. 2 and Additional file 7: Table S7). The rTCA cycle biomarker Acl was present in one MAG for Nitrospira and one MAG for Ca. Woesearchaeota. Further, rTCA cycle was the only CO2 fixation pathway employed by Nitrospira and Ca. Woesearchaeota. While nitrite-oxidizing Nitrospira was previously known to use rTCA cycle for CO2 fixation [40], this is the first reported case that Ca. Woesearchaeota may be capable of rTCA reaction. 3-HP/4-HB cycle was the most energy-efficient aerobic autotrophic pathway for CO2 fixation [41]. The biomarker enoyl-CoA hydratase/3-hydroxyacyl-CoA dehydrogenase involved in 3-HP/4-HB or DC/4-HB cycle was found in three bacterial groups, Nitrospinae, Acidobacteria, and Planctomycetes, and one archaea, Thaumarchaeota. The WL pathway biomarkers, i.e., acetyl-CoA decarbonylase/synthase complex subunit delta (cdhD), acetyl-CoA synthase (acsB), anaerobic carbon monoxide dehydrogenase (cooS and cooF), and 5-methyltetrahydrofolate corrinoid/iron-sulfur protein methyltransferase (acsE) were found in Bacteroidetes, Alphaproteobacteria, Ca. Sumerlaeota, Chloroflexi, Gammaproteobacteria, and Planctomycetes. Moreover, we found 16 MAGs associated with 7 phyla, i.e., Proteobacteria, Bacteroidetes, Chloroflexi, Gemmatimonadetes, Nitrospinae, Planctomycetes, and Thaumarchaeota, contained the Coenzyme M biosynthesis and methanogenesis genes (Fig. 2 and Additional file 7: Table S7), indicating the hadal microorganisms had the potential to produce methane from fixing CO2 in the hadal zone.
Overall, among the 178 MAGs reconstructed for the hadal sediment microbiome, 69 MAGs (~ 39%) were affiliated with 16 phyla and mixotrophic based on their capacity of inorganic carbon fixation (Fig. 2 and Additional file 7: Table S7). The enrichment and significant taxonomic expansion of mixotrophic microbes we revealed for the deepest hadal microbiome may indicate an adaptive niche that mixotrophy confers to microbes living in an oligotrophic habitat like the Challenger Deep sediment.
Aerobic vs anaerobic respiration
We investigated the hadal microbiome for its potential to carry out aerobic or anaerobic respiration. In total, 155 MAGs (~ 87%) were found to contain aerobic respiratory genes, such as Cytochrome c oxidases (Cox/Cyd/Qox/cco/Cyo). These MAGs are associated with 24 phyla, such as Acidobacteria, Bacteroidetes, Gemmatimonadetes, and Proteobacteria (Fig. 2 and Additional file 7: Table S7). Therefore, a large majority of the microbes in the hadal habitat can potentially use oxygen as an electron acceptor for energy generation.
On the other hand, anaerobic respiration pathways were also found to be widely distributed, for which nitrate/nitrite or sulfate was the alternative electron acceptors. Dissimilatory nitrate reduction to ammonia pathway (DNRA), denitrification pathway, and sulfate reduction pathways were searched and found in a total of 145 MAGs (81%) (Fig. 2 and Additional file 7: Table S7). These MAGs covered 21 of the 26 phyla we identified in the hadal microbiome. Specifically, the genes for DNRA (nirB and nirD as markers), denitrification pathway (nirK and norC), assimilatory sulfate reduction pathway (sat, aprAB, and dsrAB), and dissimilatory sulfate reduction pathway (aprA- and aprB) were present in 64 MAGs (15 phyla), 77 MAGs (16 phyla), 135 MAGs (20 phyla), and 67 MAGs (15 phyla), respectively (Fig. 2 and Additional file 7: Table S7). Previous studies based on 16S rRNA gene sequencing analysis suggested that dissimilatory nitrate reduction and denitrification were the dominant reactions, whereas microbial sulfate reduction was negligible in the Challenger Deep sediment [13]. However, our analysis shows that the broad distribution of both assimilatory sulfate reduction and dissimilatory sulfate reduction points to the equal importance of sulfate reduction pathways for anaerobic respiration, if not otherwise more important. In addition, for microbes capable of anaerobic respiration, many were found to have the potential for different types of fermentation for degradation of organic matters (Additional file 17: Analysis of fermentation; Additional file 7: Table S7).
The enrichment of anaerobic or anaerobic respiration metabolism in microbial community was apparently driven by the local habitat conditions. Examples of the deep petroleum seep sediments and Guaymas Basin hydrothermal sediments illustrated MAGs enriched for anaerobic respiration in the presence of hydrocarbons, i.e., alkanes and aromatic compounds, or depleted oxygen levels in sediments [31, 42]. We further investigated the MAGs for the potential to carry out both aerobic and anaerobic respiration and found 136 MAGs out of 178 (~ 76%), affiliated with 20 phyla (Fig. 2 and Additional file 7: Table S7), were capable of facultative anaerobic metabolism. Compared to the microbial communities in deep petroleum seeps or Guaymas Basin hydrothermal habitat, the high proportion of facultative anaerobes in the Challenger Deep sediment, reflects the microbial metabolic versatility and the ability to adapt by the hadal microbes to endemic habitats or environmental disturbance.
Chemolithotrophy
Microbes are known to acquire their energy for growth and CO2 fixation from the oxidation of inorganic compounds, such as hydrogen (H2), hydrogen sulfide (H2S), ammonia (NH3), carbon monoxide (CO), and metals [43]. The nitrification gene markers were found in 53 MAGs (30%, 17 phyla) of the Challenger Deep sediment microbiome (Fig. 2 and Additional file 7: Table S7). The key enzyme in nitrite oxidation, nitrite oxidoreductase/nitrate reductase was present in 35 MAGs (20%; 13 phyla), suggesting that nitrification is an important approach for energy production for a substantial proportion of microbes in the hadal sediment biosphere. Other genes involved in nitrification pathways, PmoA-amoA, PmoB-amoB, and PmoC-amoC, were detected in 4 MAGs associated with 3 phyla (Additional file 7: Table S7). Hydroxylamine dehydrogenase gene (hao) was detected in 21 MAGs (12%; 12 phyla). Interestingly, the gene, hao, was not found in bathypelagic microbes capable of nitrification [44], reflecting the difference in microbial mediators between the bathypelagic and hadal zones in similar biogeochemical process.
The genes for CO oxidation, sulfide oxidation, H2 oxidation, and iron oxidation were also detected in the Challenger Deep sediment microbiome. Carbon monoxide dehydrogenase (CODH; cox gene) was found in 92 MAGs (52%; 15 phyla) (Fig. 2 and Additional file 7: Table S7), indicating its broad distribution, and the importance of CO-oxidation as an energy source in the hadal habitat. Sulfide oxidation genes such as sqr and fccB were found in 33 MAGs (19%; 8 phyla), i.e., Acidobacteria, Actinobacteria, Bacteroidetes, Chloroflexi, Gemmatimonadetes, Planctomycetes, Thaumarchaeota, and Proteobacteria. For iron oxidation, we detected cyc2 gene (encoding Cytochrome c) in one MAG associated with Gemmatimonadetes, suggesting that Fe2+ can serve as an alternative electron donor for Gemmatimonadetes.
Microeukaryotic community and predominant fungal groups
Microbial eukaryotes in hadal sediment were the least studied and remained largely unknown. Our analysis found that Opisthokonta (Fungi) were the predominant eukaryotes in the Challenger Deep sediment, accounting for 87.42% of the total eukaryotic sequences. Ascomycota (41.73%), Basidiomycota (26.82%), and Mucoromycota (12.64%) were the top phyla, followed by Chlorophyta (6.61%), Rhodophyta (2.44%), Bacillariophyta (2.31%), Chytridiomycota (2.21%), Zoopagomycota (1.91%), Apicomplexa (0.96%), etc. (Fig. 3A). Our results revealed the presence of Stramenopiles, e.g., Blastocladiomycota, and Alveolata, e.g., Apicomplexa, two members of the super-group SAR (i.e., Stramenopiles, Alveolata, and Rhizaria), in the hadal sediment, albeit at low abundances. Rhizaria was, however, notably absent. Previous studies showed that the abundance of Alveolata and Stramenopiles decreased with depth in the water column of the Mariana Trench, whereas Opisthokonta had an inverse trend [45]. Our study expanded the findings into the trench sediment, illustrating the co-existence pattern of Alveolata, Stramenopiles, and Opisthokonta.
As the dominant eukaryotic group, fungi in the Challenger Deep sediment biosphere comprised six phyla, i.e., Zoopagomycota, Mucoromycota, Ascomycota, Basidiomycota, Chytridiomycota, and Blastocladiomycota, which can be further classified into twenty-seven classes (Fig. 3B). Fungi were known to be distributed globally in deep-sea water, including that in the Mariana Trench, as Basidiomycota and Ascomycota were the primary components of the deep-sea fungal community [46, 47]. Basidiomycota and Ustilaginomycetes were previously detected in deep-sea sediment using a cloning library method [48]. The capability of living in anoxia conditions may be crucial for fungi to adapt to the hadal sediment environments, as Ascomycetes, Basidiomycetes, and Chytridiomycetes were reportedly capable of fermentation and anaerobic growth in the deep ocean [49].
Metabolic functions of the fungal community in the hadal sediment biosphere
The details of gene contents and metabolic functions of the fungal community in both the water column and deep-sea sediment remained largely unknown, as early studies were based on 18S rRNA gene/ ITS amplicon sequencing [45, 47]. In the current work, we explored the metabolic potential of the fungal groups using the assembled metagenome.
Carbon metabolism
We investigated the potential of the hadal fungi to metabolize carbohydrates and peptides in the hadal sediment by looking into genes for carbohydrate-active enzymes (CAZYmes) and peptidases. In total, we detected 262 CAZYmes (88 family) and 703 peptidases (77 family) (Fig. 3B and Additional file 8: Table S8). Members of CAZyme included 79 GHs (glycosidases or glycosyl hydrolases), 8 PLs (polysaccharide lyases), 51 CEs (carbohydrate esterases), 83 GTs (glycosyltransferases), 14 AAs (auxiliary activities; associated with polysaccharide and lignin degradation), and 27 CBM (carbohydrate-binding modules). The major sources of CAZymes were Eurotiomycetes (number of genes n = 20), Agaricomycetes (n = 16), Sordariomycetes (n = 38), and Malasseziomycetes (n = 30). Particularly, key CAZyme genes encoding xylanase (GH10) were found in Agaricomycetes, and those encoding cellulase (GH5) were found in Agaricomycetes, Chytridiomycetes, and Microbotryomycetes (Additional file 8: Table S8).
Peptidase genes were abundant in Endogonomycetes (n = 124), Agaricomycetes (n = 84), Sordariomycetes (n = 70), and Malasseziomycetes (n = 67) (Fig. 3B and Additional file 8: Table S8). Six peptidase groups, i.e., Aspartic (A), Cysteine (C), Metallo (M), Mixed (P), Serine (S), and Threonine (T) peptidases, were found in the fungal genomes. Among them, the most abundant was Metallo peptidase, consisting of 230 members that belong to 30 families, followed by Cysteine (n = 225; 21 families), Serine (n = 168; 19 families), Threonine (n = 43; 3 families), and Aspartic peptidases (n = 14; 3 families) (Fig. 3B and Additional file 8: Table S8). So, the hadal fungi have broad potentials in organic carbon cycling, capable of degrading a variety of carbohydrate and peptide substrates.
Nitrogen metabolism
. The hadal sediment fungi were found to possess the complete pathways for dissimilatory nitrate reduction and assimilatory nitrate reduction, and the partial pathway for nitrogen denitrification, but lacked the ability for nitrification or anammox (Fig. 3C). The key enzymes for nitrate reduction, namely NarGHI/NapAB and NirBD for dissimilatory pathway, and NR/NasAB and NIT-6 for assimilatory pathway, were found in several fungal groups, indicating dissimilatory nitrate reduction is an important function for the hadal fungi. Ammonia in hadal water could be generated from decomposition of nitrogenous organic matter [4]. However, it can be limited in hadal sediment. Therefore, the nitrate reduction by the sediment fungi can be an important source of ammonia for hadal sediment microbes.
The partial pathway for nitrogen denitrification was also found in the hadal sediment fungi, missing the NosZ gene that coverts nitrous oxide to nitrogen. Downstream of the NarGHI/NapAB genes (shared with the dissimilatory nitrate reduction pathway), the NirK and NorBC genes completed the pathway to covert nitrite to nitric oxide, and nitric oxide to nitrous oxide, which may be released into the hadal sediment environment (Fig. 3C). However, unlike the bacterial community, the fungal community appeared lacking the capability of nitrification, as some key enzymes were generally missing in nitrification and anammox pathways (Fig. 3C). Notably, denitrification pathway enzymes were found in Fusarium, Penicillium, and Aspergillus from the Challenger Deep sediment (Additional file 8: Table S8). Some Fusarium species in the deep-sea oxygen-limiting environments were reportedly capable of denitrification [50]. In other low-oxygen habitats, such as anaerobic marine sediment, salt-tolerant Penicillium and Aspergillus species were also identified to carry out the denitrification process [50]. These data illustrate the potential of the hadal fungi in nitrogen metabolism, possibly having a significant role in the ecological process in the hadal environment.
Sulfur metabolism
Sulfate reduction is one of the main anaerobic respiratory pathways that many microbes living in anaerobic conditions depend on. The hadal fungi in the Challenger Deep sediment were found to possess the complete pathways for assimilatory sulfate reduction, dissimilatory sulfate reduction, and sulfide oxidation (Fig. 3C). They were also found to contain partial pathway for the SOX system, but were incomplete for oxidation of thiosulfate. The data suggest that hadal sediment fungi—like hadal bacteria, could take part in sedimentary sulfur cycle, which has not been reported for fungal communities in deep-sea sediment.
The possession of complete sulfate reduction enzymes indicated the hadal fungi had the potential of using sulfate reduction for energy production when living under anaerobic conditions (Fig. 3C). A high copy number of genes for sulfate reduction enzymes, like Sat, CysC, CysH, and CysJ, were detected in the hadal sediment fungi (Additional file 8: Table S8). The sulfate adenylyltransferase gene (sat) for catalyzing the bidirectional reactions between APS and sulfate was widely distributed in members of different phyla, e.g., Ascomycota (4 classes), Basidiomycota (3 classes), Chytridiomycota (2 classes), and Mucoromycota (4 classes).
In the sulfide oxidation pathway, oxidation of sulfide to sulfite was catalyzed by DsrA/B, whose genes were detected in several classes of Basidiomycota and Ascomycota, like Malasseziomycetes and Sordariomycetes. Subsequently, sulfite was oxidized to APS by adenylylsulfate reductases. Dothideomycetes (belonging to Ascomycota) contained Apr-A gene for sulfite oxidization to APS. Agaricomycetes (belonging to Basidiomycota) had the adenylylsulfate reductase gene (glutathione), denoted as APR, for the same function. Finally, APS was oxidized to sulfate by the sat-coding enzymes, which was found in Ascomycota, Basidiomycota, Chytridiomycota, and Mucoromycota. The genes encoding taurine dioxygenase (tauD) for converting taurine to sulfite were found in Basidiomycota and Ascomycota, like Agaricomycetes, Dothideomycetes, and Eurotiomycetes. The genes encoding alkanesulfonate monooxygenase (ssuD) for converting alkanesulfonate to sulfite were found in Leotiomycetes (belonging to Ascomycota). These enzymes would produce sulfite from an organic sulfur substrate, which in turn would be fed into the sulfur metabolic pathways.
The hadal sediment fungi were also found to possess sox genes, indicating their potential for thiosulfate/sulfide oxidization. For example, Endogonomycetes of Mucoromycota contained soxC encoding sulfane dehydrogenase subunits for oxidizing thiosulfate to sulfate. Malasseziomycetes of Basidiomycota and Sordariomycetes of Ascomycota were found to have SQOR (eukaryotic sulfide quinone oxidoreductase) genes that are involved in the first step of hydrogen sulfide metabolism to produce sulfane sulfur metabolites. A previous study reported that fungi could feed sulfate to sulfate-reducing bacteria (SRB) [50]. Interestingly, we also found SRB, e.g., Desulfobacterales and Desulfuromonadales (within the class Deltaproteobacteria), and Nitrospirae (class) in the Challenger Deep sediment, which corroborates the evidence for the existence of sulfide oxidation reactions and the roles that fungi that play in sulfur cycling in the hadal sediment habitat. In addition, the hydrogen sulfide metabolisms possessed by the hadal sediment fungi would counter the accumulation of sulfide (H2S, HS−, and S2−) that was generated by SRB in the anaerobic environment. Thus, the discovery of the critical metabolic genes in sulfur metabolism implicated the important role of hadal fungal community in sulfur cycling and energy transformation in the trench environments.
Metavirome in the Challenger Deep sediment habitat
The existence and composition of viral community in the hadal sediment habitat were an open question remaining to be answered. Our deep metagenomics approach offered a new opportunity to look into the viral components for the deepest biosphere. Viral sequence reads and their taxonomic affiliations were determined using Kaiju by mapping to the viral references from NCBI-nr database. A total of fifteen major viral families were identified in the Challenger Deep sediment (Fig. 4A). dsDNA viruses were found most frequently, which were mainly affiliated with the order Caudovirales, also known as the tailed bacteriophages [52]. The predominant family was Myoviridae, accounting for 24.62% of the total viral sequences, followed by Siphoviridae (19.70%) and Podoviridae (9.40%). Notably, we detected a relatively high abundance of giant viruses in the Challenger Deep sediment, such as Mimiviridae (4.16%) and Phycodnaviridae (1.06%) (Fig. 4A, red star). Giant viruses were often missed or underestimated in early studies due to the technique used to capture viral particles via filtering [53]. Our approach is indifferent to the viral components of different particle sizes.
To investigate the diversity and evolution among the members of Caudovirales found in the Challenger Deep sediment, phylogenetic analysis was performed on the viral contigs, using a marker gene coding the terminase large subunit (TerL) [54]. Interestingly, for the two families, Myoviridae and Siphoviridae, there were separate branche(s) forming within each of them (Fig. 4B). They were tentatively named MT clade I and MT clade II for Myoviridae, and MT clade III for Siphoviridae. The new diversity of Caudovirales in the Challenger Deep sediment may reflect the differentiation of viral niche to microbial hosts living in the hadal habitat of high-hydrostatic pressure and oligotrophy.
Virus-host association network
To infer virus and microbial host interactions, we took the use of a combination of different methods as described [55]. Possible virus-host interactions were summarized for each viral family. As a result, members from 131 bacteria and archaea classes were associated with 20 viral families (Fig. 4C and Additional file 8: Table S9). The association network revealed many one-to-many relationships between viruses and hosts, and vice versa. Siphoviridae was connected with the largest number of hosts that belong to 95 classes. Notably, many of the hosts were uncultured microbes that were only inferred from their presence in metagenome sequences.
Unlike in epipelagic and mesopelagic ocean waters, where the most frequent hosts were Cyanobacteria and Alphaproteobacteria (mainly SAR11 [52]), the most frequent hosts in the Challenger Deep sediment were Firmicutes (mainly Bacilli) and Bacteroidetes, followed by Euryarchaeota (including uncultured marine group II/III and Diaforarchaea), unclassified Chloroflexi, and Alphaproteobacteria. A previous study reported that the dsDNA virus T7virus (belonging to Podoviridae) could infect Deltaproteobacteria, Gammaproteobacteria, Alphaproteobacteria, Firmicutes, and Cyanobacteria [55]. Our results indicated that the T7virus from the Challenger Deep sediment was also associated with an archaeal host, e.g., the Diaforarchaea group of Euryarchaeota.
Functional viromics
To investigate the gene contents and functions of the hadal sediment viruses, we annotated their contigs using the Clusters of Orthologous Groups (COGs) database [56]. While a large proportion of their ORFs had unknown functions, high abundance genes were found in the COG categories of “replication, recombination and repair,” “transcription,” “signal transduction mechanism,” “cell wall/membrane/envelope biogenesis,” etc. (Fig. 4D). Besides the genes for basic viral functions, a special group of virus-encoded genes can modulate the activities of the hosts upon infection. These so-called auxiliary metabolic genes (AMGs) were a means for viruses to manipulate host metabolism, like sulfur and nitrogen cycling.
We identified 40 putative AMGs from the Challenger Deep sediment viruses, having roles in carbon, sulfur, or nitrogen metabolism (Additional file 10: Table S10). Among them, auxiliary carbohydrate metabolic genes were the most frequent. By searching against the CAZymes database, 45 carbohydrate metabolism-related genes were identified, which included 31 GHs (glycosidases or glycosyl hydrolases), six PLs (polysaccharide lyases), five GTs (glycosyltransferases), two CBM (carbohydrate-binding modules), and one AA (auxiliary activities). The presence of frequent auxiliary carbohydrate metabolic genes provided evidence that the hadal sediment viruses may heavily manipulate carbohydrate metabolism of hosts, especially promoting carbohydrate degradation on hosts in an oligotrophic habitat. AMGs for carbohydrate metabolism were also the largest group in other marine virome dataset, like the Tara Oceans Viromes [55], which concentrated, however, on different pathways, like galactose metabolism and glycosyltransferases, reflecting the differentiated viral niche to hosts in dramatically different habitats.
For nitrogen metabolism, NirK that encodes nitrite reductase [57] was found in the hadal viruses, suggesting possible roles to enhance host nitrite reduction pathway and release of nitric oxide into hadal sediment. For sulfur cycling, the CysN/NodQ gene encoding ATP sulphurylase [58] that reduces sulfate to produce 3'-phosphoadenosine-5'- phosphosulfate (PAPS), the first step of assimilatory sulfate reduction, was found in the hadal viruses. Notably, both NirK and CysN/NodQ are found for the first time among AMGs for ocean viruses.
Large-scale isolation of microbes from the Challenger Deep sediment habitat
To catalog and characterize microbes in the hadal sediment habitat, an intensive effort was made for the isolation of microbes from the Challenger Deep sediment. We adopted a protocol previously developed for the isolation of environmental microorganisms, which used serial dilutions of samples to allow the growth of “disadvantaged” microbes that otherwise would not be isolated [59]. To broaden the diversity of microbial isolates, we employed 24 different media and combined them with various culture conditions (Additional file 11: Table S11). Notably, the experimental conditions would enable facultative anaerobes but not strict anaerobes. As a result, more than two thousand individual isolates were obtained, for which 1089 were completed for 16S rRNA gene (for prokaryotes) or ITS amplicon (for eukaryotes) sequencing (Additional file 12: Table S12 and S13).
Bacterial isolates
The taxonomy of 1070 bacterial isolates was assigned based on 16S rRNA gene sequences using SILVA database (release 123) [60]. The bacterial isolates belonged to four phyla, i.e., Proteobacteria, Bacteroidetes, Actinobacteria, and Firmicutes, which were further categorized into 7 classes, 18 orders, 25 families, and 40 genera (Additional file 17: Fig. S2 and Additional file 12: Table S12). They matched the microbes represented by sequences in the metagenome. Halomonas, Pseudoalteromonas, and Idiomarina were the top genus, accounting for 32.77%, 22.23%, and 9.92% of the isolates, respectively (Fig. 5A). Nineteen bacterial isolates from five classes were candidates for new species, which had a 16S rRNA gene sequence with ≤ 97% identity to their closest references (Additional file 14: Table S14) [61].
We characterized representative isolates from each of the four phyla, i.e., MTFD_053 (Zunongwangia sp.), MTFD_039 (Microbacterium sp.), MTFD_075 (Halomonas sp.), MTFD_0511 (Salipiger sp.), MTFD_0323 (Pseudoalteromonas sp), and MTFD_035 (Paenibacillus sp.), which were visualized after incubation at elevated pressure (100 MPa) for 7 days. Some morphologic changes were observed with MTFD_039, MTFD_075, MTFD_0323, and MTFD_0511, when compared between their cells cultured at 0.1 and 100 MPa (Fig. 5B). Cases were found that sister cells were connected after cell divisions under culture at elevated pressure. These isolates exemplified the small fraction of culturable microbes in contrast to the community background in the Challenger Deep sediment illustrated by the MAG data, which complemented the lack of taxonomic details from the metagenomic analysis. The results suggested that they were piezotolerant, which are likely derived from the microbes that descended from the water column, contributing to the diversity and metabolic functions of the hadal sediment microbiome.
Microeukaryotic isolates
Nineteen microbial eukaryotes (fungi) were isolated from the Challenger Deep sediment (Additional file 13: Table S13 and Additional file 17: Fig. S4). Their taxonomy was assigned based on their ITS sequences using the NCBI ITS database. They belonged to nine fungal genera, namely Aspergillus, Penicillium, Acremonium, Microascus, Exophiala, Cladosporium, Gymnoascus, Purpureocillium, and Aureobasidium. They matched the eukaryotic microbes represented by sequences in the metagenome.
Representative isolates from each genus were further characterized after incubation at 0.1 and 100 MPa for 14 days. MT19_18 (Aureobasidium sp.) tended to form long pseudomycelia under 100 MPa (Fig. 5C). Some elongated cells contained multiple nuclei but did not have a diaphragm separating them. The filamentous fungi, i.e., MT20_2 (Exophiala sp.), MT20_5 (Aspergillus sp.), MT20_3 (Penicillium sp.), MT19-20 (Onygenaceae sp.), MT19_19 (Gymnoascus sp.), MT19_17 (Cladosporium sp.), MT20_1 (Purpureocillium sp.), and MT20_8 (Microascus sp.), have some morphologic variations induced by elevated pressure, such as hyphal swelling (MT19_19) and swollen conidia (MT20_3) (Fig. 5C). Conidiophore appeared normal for MT19_17 under 100 MPa compared to that under 0.1 MPa. MT20_1 (Purpureocillium sp.), MT19_19 (Gymnoascus sp.), and MT20_8 (Microascus sp.) may represent new species specialized in the Challenger Deep habitat, whose ITS sequences diverged from their closest known species (Additional file 13: Table S13).