Microbial population structure along the water columns and sediments in the Diamantina and Kermadec trenches

Abstract

Background

Microbes are widespread from the marine surface to the hadal zones and play a significant role in global biogeochemical cycling. Physicochemical properties of hadal zone shift with depth, in turn influencing the distribution profiles, biogeochemical functions, and adaptative mechanisms of microbial communities in hadal trenches. However, the ecological functions and evolutions of microbial communities along the surface water down to the sediments in the Diamantina and Kermadec trenches have been rarely studied.

Results

Here, we provided a detailed metagenomic analysis of samples along the water columns (0–6553 m) and sediments (3060–9232 m) in the Diamantina and Kermadec trenches. The euphotic waters had a significantly higher ɑ-diversity than the deep-sea waters and sediments (p < 0.05, ANOSIM). Clear inter/intra-trench discrepancies of microbial communities along water layers appeared, with remarkable vertical connectivity exhibited in the Diamantina Trench (97.5%) than the Kermadec Trench (88.8%). Positive correlations among Proteobacteria, Bacteroidota, Actinobacteria, and Thaumarchaeota in seawaters and between Proteobacteria and Chloroflexi in sediments were revealed from the co-occurrence network. Niche-specific microbial groups showed distinct dominant metabolic pathways in carbon fixation, nitrogen, and sulfur cycles. Furthermore, we reconstructed 119 metagenome-assembled genomes (MAGs) of Rhodobacterales, and their notably low ratios of non-synonymous substitutions to synonymous substitutions (pN/pS, 0.23) and high carbon atoms per residue side chain (C-ARSC, 2.86) in deep-sea sediments suggested a pronounced selection critical for their survival.

Conclusions

We found a clear connectivity of microbial communities in vertical profile, and discrepancy existed between the Diamantina and Kermadec trenches; Rhodobacterales’ evolutionary adaptation related to genomic features (pN/pS and SNVs/kbp) in the deep-sea trench environments. These findings provided new insights into the community succession and potential adaption mechanism along the water columns to sediments in deep trenches.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12915-025-02490-7.

Background

Prokaryotes are widely distributed in the water columns and sediments [1], and play an essential role in the marine biogeochemical cycling [2]. The pelagic prokaryotes are generally divided into particle-attached (PA) and free-living (FL), together account for a large portion of the total biomass of microorganisms in the open ocean [3]. Microbial community structure shows horizontal heterogeneity in the open ocean and even greater variation along the vertical axis [4]. Throughout the water column, the euphotic waters exhibited different microbial communities from the deep-sea waters, shaped by gradients of temperature, sunlight, oxygen, and nutrients [5, 6]. Microbes originated from the surface water column had a significant impact on the structure and biogeography of deep-sea microbial communities [7, 8]. Sinking particles acted as a conduit for microbes to the deep sea, linking surface and deep-sea prokaryotic communities along the water columns [8, 9] via diverse processes, such as aggregation and disaggregation, degradation of particulate organic matter (POM), and trophic transfer [9, 10]. FL heterotrophic communities are also affected by chemotactically attracted to the dissolved organic matter (DOM) plumes released from the sinking POM [11]. The microbial community compositions in seawaters are also distinct from those in deep-sea sediments [12]. In addition, genes involved in carbon, nitrogen, and sulfur metabolisms revealed that heterotrophic processes are the most common microbial metabolisms in the seawaters, while chemolithoautotrophic metabolisms probably dominated at deep-sea sediments [12].

Hadal trenches are the deepest part of the ocean, characterized by extremely high hydrostatic pressure, low temperature, and isolated hydrotopographical conditions [13]. This implies that surface microbes attached to sinking particles are faced with increasing survival challenges when they are transported to the deep sea [14]. Furthermore, nutrient limitation drives the evolution of core genomic and proteomic traits [15], which probably shaped distinct Marinimicrobia lineages across different water layers and between waters and sediments [16]; the genomic properties and contents of microbes reflected the physical and chemical compositions of their environment as well as their microhabitat-specific adaptive traits [17]. By far, investigations of diversity, composition, and metabolic function of microbial communities had been conducted extensively [12, 18, 19], but limited information regarding the metagenomic comparisons of the microbial population shifts, ecological function along the water columns down to the sediments. Furthermore, few direct studies of the habitat-specific genomic traits, evolutionary histories and selection pressures of microbes in deep-sea trench water and sediments exist to understand their adaptations to extreme environments [20]. A few studies based on 16S rRNA analysis had showed the microbial community structure in the Diamantina Trench (DT) and Kermadec Trench (KT) [21, 22]. For example, microbial community structure of the water columns was significantly different from the benthic boundary layer (BBL) in the DT and KT [22]. However, the metagenomic investigation across the seawaters and sediments in different trenches was largely unexplored. Considering the geographic differences between the two trenches, different inter/intra trench-associated sinking particle connectivity would be expected.

Here, we conducted a comparative metagenomic study to investigate the community structure, ecological function and evolutionary metrics of prokaryotes along the water columns and sediments in these two trenches to elucidate the community succession and evolutionary relationship in the microbial ecosystem of deep-sea trenches.

Results

Hydrographic conditions

A total of 31 samples including filters and surface layer of sediments were obtained from the DT and KT (Fig. 1). After trim and assembly, a total of 318,893 contigs were obtained with the largest contig length of 2,069,618 bp at FDZ177_PA (Additional file 1: Table S1). The temperature and salinity of seawaters exhibited different trends along the water columns (Additional file 2: Fig. S1). Salinity was the highest at the surface and gradually decreased with depth until 800 m, and then slightly increased until 5000 m. The NO₃⁻ and NO₂⁻ concentrations were increased with depth in both trenches. NH₄⁺ concentration was higher in the KT than DT except for 5000 m. The PO₄³⁻ concentrations peaked at 5000 m in the DT and at 800 m in the KT, respectively. Inorganic nutrient concentrations in the BBL were always higher at Stn. FDZ183 than those at Stn. FDZ177. As for the sediments, the highest concentration of NH₄⁺, TP and TOC appeared at Stns. FDZ176, FDZ177, and FDZ173 in the DT, respectively. The highest concentrations of NO₃⁻ appeared at Stn. FDZ146 in the KT (Additional file 1: Table S2).

Fig. 1.

Map of the sampling locations in the Diamantina (left) and Kermadec (right) trenches. CTD, conductivity-temperature-depth

Microbial community structure and environmental effects

For the community composition, bacteria predominated in all samples, except for the sediments at Stn. FDZ128. Sediments showed higher archaeal proportions than seawaters, with Stn. FDZ175 (38.36%) in the DT and Stn. FDZ128 (57.31%) in the KT (Fig. 2A). Pseudomonadales had the highest abundance in the DT_2000_FL and KT_800_FL with 91.54% and 44.92%, respectively. Pseudomonadales, Rhodobacterales, Sphingomonadales, Hyphomonadales, Alteromonadales, and Oceanospirillales were the major orders of Alphaproteobacteria and Gammaproteobacteria in two trenches. Thaumarchaeota and Alphaproteobacteria were abundant in the sediments and water column, respectively. Thaumarchaeota showed a significant correlation with NH₄⁺ (p < 0.05). Non-metric multidimensional scaling (NMDS) analysis showed that the microbial communities formed three distinct groups (i.e., euphotic waters, deep-sea waters and sediments; p < 0.05) (Fig. 2B). Significantly higher Simpson diversity was found in euphotic waters than in sediments (p < 0.05, Fig. 2C), and in euphotic waters than deep-sea waters in the KT (p < 0.05, Fig. 2C). The highest Chao1 was significantly higher in euphotic waters than in deep-sea waters (p < 0.01) and sediments (p < 0.05) in the DT. Moreover, sediments exhibited significantly lower Chao1 than both euphotic and deep-sea waters (p < 0.05) in the KT (Fig. 2C). CCA analysis demonstrated that depth, temperature and NO₃⁻ and NO₂⁻ (p < 0.05) were the key environmental parameters affected the microbial communities in seawaters (Additional file 2: Fig. S2A), whereas depth (p < 0.05) was critical to those in the sediments (Additional file 2: Fig. S2B) in both trenches.

Fig. 2.

Microbial community structure and diversity. Microbial community structure (A), non-metric multidimensional scaling (NMDS) plots (B), and community diversity index (C) in the Diamantina and Kermadec trenches. * p < 0.05, ** p < 0.01. EW and DW represented euphotic waters and deep-sea waters, respectively

Inter-/intra-trench variations and interactions of microbial communities

About 454 shared prokaryotic species were detected based on the calculation of the prokaryotic relative abundance in all samples and were observed to be concurrently present at different depths in both trenches (Fig. 3A). A total of 232 and 513 prokaryotic species were identified in both euphotic waters and deep-sea waters in the DT and KT respectively to show the intra-trench similarities of prokaryotic communities. Furthermore, the shared prokaryotic species among different seawater layers appeared to occur more frequently in seawaters than sediments (Fig. 3B, C). The prokaryotic communities were categorized into five groups: the euphotic zone (0 m and 200 m), the mesopelagic zone (800 m), the bathypelagic zone (2000 m, 2500 m and 3000 m) and the abyssopelagic zone (5000 m). The mesopelagic community in the two trenches contained 72.6–84.6% and 56.3–62.1% of surface-originated species, respectively. In the DT, about 97.5% and 98.6% of the surface PA and FL fractions were also appeared in the abyssopelagic communities, respectively. In the KT, their contributions were 88.8% and 76.4%, respectively (Fig. 3D).

Fig. 3.

Connectivity of microbes in the Diamantina and Kermadec trenches. Overlap of prokaryotic species in multiple trenches (A), the Diamantina (B) and Kermadec (C) trenches. Proportion of the prokaryotes at different depth in each size fraction in the Diamantina and Kermadec trenches (D). The category of each prokaryotic species was defined as the depth where it was firstly detected in any size fraction, considering a directionality from surface to bathypelagic

The co-occurrence network based on the top 200 most abundant prokaryotic species revealed that positive correlations were predominant among different groups (Fig. 4). Modularity index > 0.4 suggested that the network has a modular structure (Additional file 1: Table S3). In seawaters, Proteobacteria, Bacteroidota, Actinobacteria, and Thaumarchaeota showed positive correlations (Fig. 4A, B); while in sediments, most of the positive correlations were observed between Proteobacteria and Chloroflexi, as well as between Proteobacteria and Planctomycetota (Fig. 4C).

Fig. 4.

Co-occurrence networks based on the top 200 species for microbes in the total euphotic waters (EW) (A), deep-sea waters (DW) (B), and sediments (C). The edges represent co-occurrence relationships consistent at the 0.6 correlation level, and the nodes represent microbial taxa

Major metabolic pathways of microbial community

Three nutrient cycling categories were analyzed, including carbon fixation, nitrogen and sulfur metabolism. Genes involved in three carbon fixation pathways were detected. The key genes involved in the Calvin cycle (pgk, gapA, ALDO, fbaB, fbp, fbaA, rpiA, prkB, tkt) showed a higher abundance (Welch’s t-test, p < 0.01) in deep-sea waters (Fig. 5A), and genes of pgk and tkt were affiliated with Alphaproteobacteria, Gammaproteobacteria, Actinomycetota, Pseudomonadota, and Flavobacteriales (Fig. 5B). Conversely, genes encoding enzymes catalyzing citrate to oxaloacetate in the rTCA pathway (ccsA, aclA, aclB, kor, fum) and genes associated with the 3HP/4HB pathway (3hpcs and 4hbcl) were generally more abundant in the sediments. The genes of korA, korB, ccsA, aclA, and aclB in the sediments were associated with Gemmatimonadota, Planctomycetota, Nitrospinota, Nitrospirota, Chloroflexota, and Pseudomonadota. Genes 3hpcs and 4hbcl found in the sediments were associated with Nitrososphaerota, Gammaproteobacteria, and Pseudomonadota. Additionally, FL also had a higher key gene abundance in these three carbon fixation pathways in the KT, but not observed in the DT (Fig. 5A).

Fig. 5.

Relative abundances of key genes were shown by z-score bubble plots in the carbon fixation, nitrogen and sulfur metabolic pathways (A), the data were normalized by column. The microbial taxa of typical genes and their relative abundance in each region. EW and DW represented euphotic waters and deep-sea waters, respectively

For nitrogen metabolism, genes involved in nitrogen fixation (nifD/H/K) were significantly higher in deep-sea waters than euphotic waters and sediments (Welch’s t-test, p < 0.01), and higher in PA than FL. Conversely, ammonia-oxidizing genes (amoA/B/C) were significantly more abundant in sediments than seawaters (Welch’s t-test, p < 0.01, Fig. 5A). Deep-sea waters and sediments in the DT had higher levels of nitrite reduction genes (nirB/D, nrfA/H, and narI/G/H) compared to the KT (Fig. 5A). Regarding sulfur metabolism, the DT sediments had more dissimilatory sulfate reduction and oxidation genes (dsr and apr), while deep-sea waters had the highest proportions of assimilatory sulfate reduction and SOX system genes.

Taxonomic, genomic and proteomic traits and evolutionary metrics of MAGs

Across all metagenome-assembled genomes (MAGs), a total of 599,884 contigs were recovered, ranging from 2 to 2913 contigs in each MAGs with the average GC content of 57.03% (Additional file 1: Table S4). 1,401 MAGs which were medium quality (completeness ≥ 50%, contamination ≤ 10%) were obtained and affiliated with archaea (n = 23) and bacteria (n = 1378). Proteobacteria, Actinobacteriota, and Bacteroidota were the predominant bacterial phyla, while the Thermoproteota were the most prevalent archaeal phyla in all samples (Additional file 1: Table S5). To reduce annotation uncertainty raised from the high proportion of novel lineages, taxonomic summaries were generated at order or higher ranks. Rhodobacteriales was consistently detected across all samples and represented a dominant group of bacteria. Its wide distribution suggests an adaptative capability to diverse ecological niches, making it a suitable group for exploring potential functional and evolutionary strategies in different trench environments. Phylogenetic analysis of Rhodobacterales MAGs fell into distinct phylogenetic clades, the sediment-associated lineage stayed as in an independent clade from the water-associated clade (Fig. 6).

Fig. 6.

Phylogenetic tree of Rhodobacterales metagenomeassembled genomes (MAGs). The inter four rings (from inside to outside) show the the color codes of the various bacterial genus, as well as completeness, contaminations, and GC content of each MAG (detail for genus of each MAG in Additional file 1: Table S6). The next outer ring, the stacked columns indicate the relative abundance of MAGs at different habitats. The outermost ring to show the relative abundances of key genes in the sulfur metabolic pathways. Different colors were used to illustrate the various habitats: euphotic waters (EW) (green), deep-sea waters (DW) (blue) and sediments (grey)

Rhodobacterales MAGs exhibited clear genomic differentiation in different samples types and trenches. Significantly larger genome sizes of Rhodobacterales MAGs were in both euphotic waters and deep-sea waters than sediments in the DT (p < 0.01), and in deep-sea waters than in sediments in the KT (p < 0.01) (Fig. 7A). GC content also varied between depths and trenches. Rhodobacterales MAGs from euphotic waters had significantly lower GC content than deep-sea waters in the DT (p < 0.05). While seawater MAGs in the KT showed significantly higher GC content than that of the DT (p < 0.01) (Fig. 7B). In addition, the carbon atoms per residue side chain (C-ARSC) of Rhodobacterales MAGs decreased with increasing water depth in the DT (Fig. 7C), reflecting the low carbon availabilities in deep-sea habitats. As for protein elemental composition, the nitrogen atoms per residue side chain (N-ARSC) generally increasing with GC content increased (Fig. 7D). The ratio of non-synonymous substitutions to synonymous substitutions (pN/pS) (Fig. 7E) across the entire gene and the frequency of single nucleotide variants (SNVs/kbp) (Fig. 7F) exhibited considerable variability in the different sample types. Strikingly, the SNVs/kbp and nucleotide diversity in euphotic waters were lower than that in sediments (Fig. 7G), but the pN/pS higher in waters than that of sediments (Fig. 7E). pN/pS of dmdA (involved in dimethylsulfoniopropionate demethylation) and dddL (involved in dimethylsulfoniopropionate cleavage) genes was also higher in waters (Additional file 2: Fig. S3). A negative correlation between the pN/pS and SNVs/kbp was found in the Rhodobacterales populations inhabited all water depths in the two trenches (Fig. 7H).

Fig. 7.

The variances of genome size (A), GC content (B), carbon atoms per residue side chain (C-ARSC) (C) and nitrogen atoms per residue side chain (N-ARSC) (D), non-synonymous to synonymous mutation ratio (pN/pS) (E), single-nucleotide variants (SNVs/kbp) (F), nucleotide diversity (G), and pN/pS ratio in relation to SNVs/kbp (H) of all genes in reconstructed Rhodobacterales MAGs. Two-tailed Student’s t-test, * p < 0.05, ** p < 0.01

Discussion

Particles enhance intra-trench vertical connectivity

Bacteria occupied higher proportion in seawaters than in sediments in the both trenches, consistent with the finding in the Yap Trench [12]. The dominance of Pseudomonadales in the FL at 2000 m and 800 m likely reflects their opportunistic heterotrophic lifestyle, enabling them to exploit labile organic matter derived from sinking particles and benthic-pelagic coupling processes. The high abundance of free-living Pseudomonadales in deep water was in consistent with those found in the Northeast Atlantic, Mediterranean Sea, and North Pacific gyres, where this group thrives in DOM-rich conditions linked to particle degradation [7, 23]. Archaeal Thaumarchaeota dominated in sediments of all the trenches, might associate with NH₄⁺ [12, 19]. The ɑ-diversity of prokaryotic community in the euphotic zone was significantly higher than that in the deep seawaters and sediments, likely related to the carbon availability and oxygen usage [24, 25]. The most important impacting factors in the seawaters were sampling depth, temperature and NO₃⁻ and NO₂⁻ concentrations, as being found in the Mariana Trench [19].

Shared populations between hadal trenches indicated the existence of stable, widespread microbial lineages capable of colonizing multiple ecological niches. This was likely caused by the sinking particulate organic matter, upwelling of deep seawaters, and circulation of currents and seawater masses in the deep ocean [26–28]. The vertical connectivity was higher between communities associated with the larger size fractions, likely due to their higher sinking rates. The intensive vertical connectivity from the surface to the deep seawaters might due to the colonization and transportation of surface microbes to the deep sea [7]. The DT exhibited a much stronger compositional vertical connectivity than the KT, because the latter had steep and deep walls, and the associated subduction processes lead to frequent seismic and volcanic activity, contributing to the geologically unstable nature [29]. More frequent similarities of prokaryotic communities exist along KT and DT water columns compared to sediments, possibly because sinking particles could be transported over long distances and in varying directions at different depths before reaching the seafloor [7]. Therefore, issues related particle size fractions should be taken into account for studying vertical connectivity of aquatic microbial communities.

Trench niche partitioning leads to metabolic specialization and versatility

Positive correlations predominated among all samples, indicating strong co-occurrence tendencies among taxa for microbes [30, 31]. Diverse metabolic potentials with regard to the carbon, nitrogen, and sulfur cycles were revealed, reflecting the lifestyle versatility of the microbial communities as well as the coupling of the different biogeochemical cycles. The Calvin cycle was the dominant carbon fixation pathway in the deep-sea waters, in consistent with increasing key cbbM gene abundance with depth in the meso- and bathypelagic layers of the Atlantic Ocean [32], supporting the notion that autotrophic carbon fixation is a common metabolic strategy in the deep ocean. Alphaproteobacteria and Gammaproteobacteria were widely recognized as important autotrophs contributing to primary production via the Calvin cycle, often inhabiting oligotrophic ocean waters [33]. Nitrospinota and Nitrospirota potentially contribute to the sustenance of prokaryotic heterotrophic carbon requirements. The 3-HP/4-HB cycle as a distinctive archaeal carbon fixation pathway was highly energy-efficient with tolerance to oxygen [34], and has been reported in Mariana [35] and New Britain Trench sediments [18], possibly derived by the autotrophic ammonia oxidizing Nitrososphaerota [36, 37], reflecting the coupling of carbon and nitrogen cycles.

Generally, FL groups were more involved in the carbon fixation processes, which was consistent with previous study [38]. Genes involved in nitrogen fixation were more abundant in PA than FL, consistent with non-cyanobacterial diazotrophs enriched in deeper and particle-associated niches [39], suggesting that nitrogen fixation in the deep ocean was tightly coupled to the localized micro-environments provided by PA. The high abundance of nitrification genes (i.e., amoA/B/C) in the sediments indicated that ammonia-oxidizing archaea were potential ammonia oxidizers in the sediments, corresponding to their high proportion in the microbial communities. DT sediments contained higher abundance of dissimilatory sulfate reduction and oxidation genes, reflecting sulfate reduction as an important energy metabolism resulted from adaptation to the extreme biosphere [40, 41]. Clear niche partitioning and depth/habitat-driven metabolic specialization were observed, for example, autotrophic carbon fixation (via Calvin cycle) dominated in the deep ocean, energy-efficient 3-HP/4-HB cycling prevailed in sediments. Nitrogen fixation in PA-associated communities, nitrification and sulfur cycling (especially dissimilatory processes) intensified in anaerobic trench sediments. These ecological strategies underscore the complexity of biogeochemical cycling in trench environments, might be shaped by environmental gradients such as nutrient availability and energy limitation.

Strong negative selection shapes the genomic adaptation of Rhodobacterales

Rhodobacterales, involved with both dimethylsulfoniopropionate (DMSP) demethylation and cleavage pathways, played an important role in deep-sea organic sulfur cycling [42]. They had a high proportion in seawaters than in sediments, because they were more concentrated in the water columns [43]. Rhodobacterales were mainly detected in the FL fractions of seawater with only four MAGs in sediments, thus the latter might be a subset of surface-derived populations that adapted to the sedimentary environment through versatile metabolic capabilities [44]. Gene abundance of dmdA and dddL, involved in the DMSP demethylation and cleavage, decreased with increase of sampling depths reflected the necessity of DMSP synthesis and storage to resist high pressure [43].

Rhodobacterales in the water columns and sediments formed distinct phylogenetic clades and occupied different ecological niches, as a result of environmental adaptations driven by distinct evolutionary selection pressures [45]. pN/pS values were well below 1 (indicating negative selection), suggesting that nonsynonymous mutations are selectively removed to maintain protein function. Rhodobacterales were more prevalent in seawaters, and their lowest pN/pS ratio and highest SNVs/kbp in sediments implied a stringent negative selection on their genes (including dmdA and dddL) and importance of maintaining the functional stability for survival in the challenging deep-sea conditions [46], indicating different evolutionary trends existed between samples types. The negative selection has been previously observed for natural comammox Nitrospira [47] and Thalassospira in deep-sea sediments [48], as well as Candidatus Methylomirabilis oxyfera in Mariana Trench [49]. C-ARSC and N-ARSC reflected the tolerance of proteins to mutations and the selective pressure, thereby potentially indirectly affecting diversity indices such as pN/pS and SNVs/kbp. C-ARSC of Rhodobacterales MAGs in sediments was higher than seawaters, and these MAGs from the DT seawater were negatively correlated with the increasing water depth, suggesting high carbon availabilities in the surrounding environment [16, 50]. Consequently, non-synonymous mutations were likely to disrupt the function and be negative selection during evolution, thus resulting in the low pN/pS, reflecting the mutual restraint relationship between the complexity of the protein structure and the conservation of its function [51]. Above evolutionary metrics among the sediments and seawaters were likely a result of an adaptation to the different physicochemical conditions [51]. These evolutionary metrics reflected the interplay between ecological processes and the evolution of microbes, and have been used in studies of different deep-sea extreme environments [48, 49].

Conclusions

This study revealed distinct microbial populations, metabolic and evolution processes across euphotic waters, deep-sea waters and surface sediments of the DT and KT. Inter/intra-trench variations of microbial communities was existed. The DT exhibited a remarkable vertical connectivity along the water columns than the KT, and particles enhance intra-trench vertical connectivity. trench niche partitioning leads to metabolic specialization and versatility involved in the carbon, nitrogen and surfur biogeochemical cycling. Strong negative selection shapes the genomic adaptation of Rhodobacterales, highlighting the interplay between ecological processes and the evolution of key bacteria in deep sea trench extreme environments, and shedding light on microbial adaptation in the subseafloor biosphere. Future studies should consider integrating metatranscriptomics, metaproteomics, and metabolomics with in situ measurements of nutrient concentrations, oxygen, and redox gradients to more precisely resolve the functional dynamics, ecological interactions, and adaptive evolutionary mechanisms of hadal microbial communities.

Methods

Sample collection and environmental factor measurement

The DT is located in the Indian Ocean about 1500 km west of Perth, Australia, and shaped by the geological breakup of the continents. It is approximately 520 km long, 70 km wide, and have a maximum depth of around 8047 m, extending in a northeast-southwest direction. The KT is located about 120 km off the northeastern coast of New Zealand in the Pacific Ocean and formed by the collision of plates. It is 1500 km long, with a mean width of 60 km, and the fifth deepest trench with a maximum depth of 10,047 m in the world [21]. Water column samples from five different depths were collected by conductivity-temperature-depth (CTD, SeaBird Electronics, Bellevue, WA, USA) in the DT (97°43′ E, 32°34′ S) and KT (176°25′ W, 29°55′ S) during cruise TS29 from Nov. 2022 to Mar. 2023 (Fig. 1). BBL samples with depths of 5,657 and 6,553 m were collected from Stns. FDZ177 (97°16′ E, 32°47′ S) and FDZ183 (101°51′ E, 34°23′ S) by the R/V “Fendouzhe” in the DT. As for sediments, five stations from the DT (3060 to 6553 m) and three stations from the KT (5,825 to 9,232 m) were sampled using pushcores, respectively. Seawater with volumes ranging from 8.8 to 41.5 L were pre-filtered through 200 µm mesh and then sequentially filtered through the 3 and 0.22 µm pore size polycarbonate filter (47 mm, EMD Millipore, Billerica, MA, USA) to collect the PA and FL microbes, respectively. The total of 31 samples including filters and surface layer of sediments were immediately stored at − 80 ℃ for further analysis. The in situ environmental parameters (temperature, salinity, and depth) were recorded with a CTD in the manned submersible the R/V “Fendouzhe”. The concentrations of inorganic nutrients (NO₃⁻ and NO₂⁻, NH₄⁺, and PO₄³⁻) in seawaters were analyzed using an auto-analyzer (QuAAtro, Blue Tech Co., Ltd., Tokyo, Jap). For the sediments, the concentrations of total organic carbon (TOC) were determined using an element analyzer (Elementar vario Macro cube, Germany) after over-drying at 105 °C [52]. Ammonia (NH₄⁺) and nitrate (NO₃⁻) were measured by analysis with a colorimetric auto-analyzer (SEAL Analytical Auto Analyzer 3, Germany) after 1 M HCl treatment [52]. The total phosphate (TP) of the sediment was measured by using the molybdate colorimetric method with a UV2450 (Shimadzu, Japan; Murphy and Riley) after digestion with nitric-perchloric acid [53].

DNA extraction and sequencing

Genomic DNA was extracted from 3 µm and 0.22 µm pore sized polycarbonate filters using a PureLink Genomic DNA Mini Kit (Invitrogen, Thermo Fisher Scientific, Corp., Carlsbad, CA, USA). Sediments DNA were collected (~ 0.5 g for each sample, homogenized) with the PowerSoil DNA Isolation Kit (MO BIO Laboratories, Inc., Carlsbad, USA), according to the manufacturer’s protocol. For each pushcore sediment, three replicate subsamples were taken for DNA extraction, and the extracted DNA was subsequently combined for metagenomic sequencing. The concentrations of obtained DNA were quantified using Qubit® 2.0 fluorometer (Life Technologies, USA). After the construction of the library, an Agilent 2100 bioanalyzer system (Agilent Technologies, Santa Clara, CA, USA) was used to detect the inserted size of the library. Sequencing was performed with an Illumina NovaSeq 6000 paired-end 150 bp reads platform (Novogene Co., Ltd., www.novogene.com). The quality filtering was achieved through removing the adapters, barcodes, and reads containing poly-N or those of low quality from the raw data with the FASTX-Toolkit (http://hannonlab.cshl.edu/fastx_toolkit) and Fastqc softwares (https://github.com/s-andrews/FastQC). Reads containing more than 40% of bases with quality scores below 15, or containing more than 5% ambiguous bases were considered low-quality.

Prokaryotic taxonomic assignment and functional annotation

In order to confirm the community composition, MetaPhlAn (v 4.1.0) was used to estimate microbial relative abundances by mapping metagenomic reads against a catalog of clade-specific marker sequences currently spanning the bacterial and archaeal phylogenies with the default marker gene database mpa_v30_CHOCOPhlAn_201901, marker-level analysis relied on the default presence threshold (–pres_th = 1.0) [54]. Ecological connectivity was referred to shared taxa, which were calculated by taxa presence/absence. The criteria for shared taxa between different prokaryotic communities was based on whether species present at one depth could be detected in the other depth. Following previous studies [7, 8], if one tax was detected in any of the 0 m samples, it was categorized as surface, but if one tax was first detected in mesopelagic zone but not in the previous depths (euphotic zone), it was categorized as mesopelagic, and so on.

The open reading frames (ORFs) of assembled contigs were identified by Prodigal (v 2.6.3) with the “-p meta” option [55]. Functional annotations of the assembled scaftigs (unigenes), especially involved in the Calvin cycle, rTCA, 3HP/4HB, N₂ fixation, nitrifcation, denitrification, nitrite reduetion, dissimilatory sulfate reduction and oxidation, assimilatory sulfate reduction, and sulfur-oxidation systempathways, were performed using Kyoto Encyclopedia of Genes and Genomes (KEGG) database (v 98.0) [56] via the Diamond (v 2.1.9.163) [57] using blastp with E-value < 10⁻⁶ parameters. Taxonomic annotation using CAT (v 5.0.3) [58] with default parameters based on NR database (v 20,231,125). The abundance of each gene was quantified using Salmon (v 1.10.3) and normalized to transcript per million (TPM) based on the gene length and sequencing depth. TPM values were further standardized using column-wise z-score normalization for visualization.

Genome binning, annotation and evolutionary metrics

High-quality reads from each sample were denovo assembled using MEGAHIT (v 1.2.9) [59] with parameters “–k-min 21 –k-max 141 –k-step 10”. All the taxonomic and functional annotations were based exclusively on the quality reads. Genome binning was performed using CONCOCT (v 1.0.0), MaxBin (v 2.2.6), and MetaBAT (v 2.12.1) by MetaWRAP pipeline (v 1.3.2) [60]. CheckM (v 1.1.3) [61] was used to assess the completeness, contamination of the retrieved MAGs. The relative abundances of the MAGs across the samples were estimated using CoverM (v 0.7.0) (https://github.com/wwood/CoverM). The N/CARSC ratio was used to estimate the nitrogen and carbon biosynthetic investment required for amino acid production encoded by the genomes [62]. The python script “get_gc_and_narsc.py” (https://github.com/faylward/pangenomics/) was used to calculate GC content and N/CARSC of the predicted genome proteins. Taxonomy of the MAGs was assigned using GTDB-TK (v 2.4.0) with reference to GTDB database (r220) [63]. The 120 single-copy genes of 119 MAGs from Rhodobacterales (completeness > 70% and contaminnation < 10%) were identified and aligned by GTDB-tk (v 2.4.0) and trimAl (v1.4.rev15) [64] with the auto options, respectively. The maximum-likelihood (ML) phylogenetic tree was inferred based on this alignment using IQ-TREE (v 2.3.4) [65] with the best-fit model (LG + F + I + R5) selected by ModelFinder [66] and 1,000 ultrafast bootstrap iterations using UFBoot2 [67], g_NORP110 designated for tree rooting. The tree was annotated and visualized by iTOL (v 6) [68] (https://itol.embl.de/).

Evaluation of selective pressure on the coding sequences would allow us to assess potential adaptive evolution to the hadal trench environment. Nucleotide metrics including pN/pS, nucleotide diversity and SNVs/kbp were calculated from these mappings using the profile module of the inStrain program (v 1.9.0) [69]. To estimate the impact of the changing sequencing coverage on SNVs/kbp, the correlation between SNVs/kbp and the coverage of all genomes were analyzed [51].

Statistical analysis

The NMDS based on the Bray–Curtis similarity index, were calculated with PAST (v 3.21) to show the distribution pattern of prokaryotic communities. An analysis of similarities (ANOSIM), based on the relative abundance of prokaryotes community, was conducted with PAST (v 3.21) [70] to test whether there was a significant difference in the microbial community among the sampling sites. The gradient length of the first axis from detrended correspondence analysis (DCA) was greater than 3, thus canonical correspondence analysis (CCA) was used to analyze the environmental effects on prokaryotes community [71]. Venn plots were constructed to display shared order among sampling sites via R package “limma” (v 3.50.3) [72]. Network analysis was conducted to explore the co-occurrence patterns within/between different taxonomy groups among the top 200 most abundant microbial species. A similarity matrix was generated by “psych” package [73] in R v3.5.3, statistically significant correlations (Spearman’s |r|> 0.6 and FDR-adjusted p < 0.05) were further visualized with Gephi (v 0.9.3) [74]. The Welch’s t-test with the Benjamin-Hochberg false discovery rate (FDR) correction was performed to compare the different gene abundance. Two-tailed Student’s t-test was performed to analysis the evolutionary metrics.

Abbreviations

EW

Euphotic waters

DW

Deep-sea waters

PA

Particle-attached

FL

Free-living

POM

Particulate organic matter

DOM

Dissolved organic matter

DT

Diamantina Trench

KT

Kermadec Trench

BBL

Benthic boundary layer

CTD

Conductivity-temperature-depth

TOC

Total organic carbon

NH₄⁺

Ammonia

NO₃⁻

Nitrate

TP

Total phosphate

ORFs

Open reading frames

KEGG

Kyoto Encyclopedia of Genes and Genomes

TPM

Transcript per million

MAGs

Metagenome-assembled genomes

pN/pS

The ratios of non-synonymous substitutions to synonymous substitutions

SNVs/kbp

Single nucleotide variants

CCA

Canonical correspondence analysis

C-ARSC

Carbon atoms per residue side chain

N-ARSC

Nitrogen atoms per residue side chain

DMSP

Dimethylsulfoniopropionate

Authors’ contributions

HL and HJ collected the samples. HJ contributed reagents, materials, and analysis tools. PW, HL, YX and FW performed the experiments. YX contributed to the data analysis. YX wrote the paper. HJ, HL and YZ contributed to the manuscript revision.

Funding

This work was supported by the Innovational Fund for Scientific and Technological Personnel of Hainan Province (KJRC2023C37), the National Natural Science Foundation of China (NSFC41776147), the Hainan Provincial Natural Science Foundation of China (424MS115; 424QN341), a project of the Academy-Locality Science and Technology Cooperation of Sanya City, China (2014YD05), and the National Key R&D Program of China (2022YFC2805505; 2022YFC2805400).

Data availability

The raw sequence data for metagenomics have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) under accession number PRJNA1187693, and genomic data have been submitted to the NCBI Genome under BioProject accession number PRJNA1297678.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.