Depth-related Variability of Biological Nitrogen Fixation and Diazotrophic Communities in Mangrove Sediments

Background: Nitrogen-xing microorganisms (diazotrophs) provide biological available nitrogen and play a pivotal role in nitrogen cycling of mangrove sediments. However, most studies on diazotrophs have been restricted to easily accessible surface sediments, while the diversity, structure and ecological function of diazotrophic communities at the in-depth prole of mangrove sediments are largely unknown. Here, we investigated how biological nitrogen xation vary with depth of mangrove sediments from the perspective of both NFR and diazotrophic communities. Results: Through acetylene reduction assay, nifH gene amplicon and metagenomic sequencing, we found that the nitrogen xation rate (NFR) increased but the diversity of diazotrophic community decreased with the depth of mangrove sediments. The structure of diazotrophic communities at different depths was largely driven by salinity, and exhibited a clear divergence at the partition depth of 50 cm. Agrobacterium and Azotobacter were specically enriched at 50-100 cm sediments, while aerobic diazotrophs such as Methylomonas had a higher abundance at 0-30 cm. Consistent with the higher NFR, metagenomic analysis indicated the elevated abundance of nitrogen xation genes (nifH/D/K) in deeper sediment layers, where nitrication genes (amoA/B/C) and denitrication genes (nirK and norB) became less abundant. Three metagenome-assembled genomes (MAGs) of diazotrophs from deep mangrove sediments indicated their facultative anaerobic and amphitrophic lifestyles as they contained genes for low-oxygen-dependent metabolism, hydrogenotrophic respiration, carbon xation and pyruvate fermentation. Together, this study determines and and our of the relationship between biological nitrogen xation limitation in


Background
Mangroves are highly productive ecosystems with immense ecological values [1]. Their high productivity is greatly attributed to the high nitrogen-xing activity of diazotrophs in the mangrove sediments, which contribute about 40%-60% of the total nitrogen required by the mangrove ecosystem [2,3]. However, due to tidal uctuations and high denitri cation rates, mangrove ecosystems are considered nitrogen-limited [4]. It has been determined that nitrogen xation can primarily affect the nutrient status of sediments, since its products are the main source of nitrogen inputs in mangrove ecosystems [4,5]. Therefore, as the rate-limiting step of nitrogen cycling, nitrogen xation is particularly important to alleviate the nitrogen limitation of mangrove ecosystems [6,7].
Diazotrophic communities are biological engines to drive the atmosphere nitrogen into mangrove ecosystems [8]. The diversity of diazotrophs is dependent on the climate, vegetation types and environmental properties across different ecosystems [9]. In terrestrial ecosystems, proteobacterial members are the prevalent diazotrophs with a high phylogenetic diversity, and exhibit preference for gathering in speci c habitats [10]. Taking alphaproteobacterial diazotroph as an example, Azospirillum, Rhodobacter, Rhizobium and Bradyrhizobium prefer the habitats with warm climate and relatively large precipitation, such as paddy soil and eastern Inner Mongolia steppe [9,11]. Consistent with terrestrial ecosystems, the structure of marine diazotrophic communities also varies with environmental properties. For example, the cyanobacterium UCYN-A is broadly distributed in marine pelagic water and has a relative high biological NFR [12]. In contrast, anaerobic diazotrophs such as Chlorobium and Desulfovibrio prefer low temperature and appear uncommon in the surface ocean, but are preponderant in cold waters, especially in the Arctic [13]. Distinguishing from terrestrial and marine ecosystems, mangrove ecosystems act as the junctions between ocean and land, and speci c diazotrophs are expected in such tidal swamp ecosystems. Although few studies have shown that sulfate-reducing bacteria may be the main nitrogen xation group in mangroves [14], a systematic understanding of the diversity and ecological function of diazotrophic communities as well as their responses to environmental variations in the mangrove ecosystem is still lacking.
Our current understanding of diazotrophic community diversity and structure in the mangrove ecosystem is hitherto mainly limited to the sur cial layers (i.e., 0-25 cm) of sediment columns, where the density of microorganisms is high [15,16]. Yet, little is known about diazotrophic communities residing in the deeper (> 25 cm) sediment horizons. In fact, in-depth pro les of mangrove sediments have a gradient of resources and environmental properrties, which have been identi ed as the universal factors directly or indirectly in uencing the diazotrophic community in terrestrial ecosystems [17,18]. For example, the sensitivity of diazotrophs to pH was different. Bradyrhizobium members often have higher acid resistance than other diazotrophs, while Mesorhizobium has a wider pH tolerant range with the optimal pH from 6 to 8 [19]. Anaerobic environments are more suitable for the survival of most diazotrophs, while aerobic diazotrophs need to evolve some mechanisms to protect nitrogenase from damages by oxygen [17,20]. Ecological surveys further determine signi cant correlations of diazotrophic richness, diversity and community composition with soil moisture, C quality and quantity, N availability and the availability of trace elements (e.g., Mo, Fe and V) [18,21]. These previous ndings allow us to hypothesize that the deeper layers of mangrove sedments may contain diazotrophic communities that are specialized for their environments and fundamentally distinct from the sur cial communities. Meanwhile, the in-depth pro le of mangrove sediments provides a platform to investigate which physicochemical properties determine the depth-related distribution of diazotrophic communities in mangrove sediments.
From the view of ecological functions, biological nitrogen xation not only provides reactional materials for downstream processes of the nitrogen cycle, but also is a key process to alleviate nitrogen limitation, especially in mangrove ecosystems [22,23]. However, a central assumption of the progressive nitrogen limitation hypothesis is that, without changes in exogenous exchange of nitrogen in an ecosystem, increases in plant nitrogen uptake require increased soil nitrogen cycling rates [24]. This indicates that there exists a competition between plant nitrogen acquisition and microbiome-mediated nitrogen transformation processes [2,25]. One reason for this competition is that the product of both nitrogen xation and nitrate reduction (dissimilatory and assimilatory nitrate reduction to ammonium), NH 4 + , not only serves as the main form of nitrogen uptaken by mangrove roots, but also can be transformed by nitri cation and ammonia assimilation [26,27]. Therefore, nitrogen availability, driven by the balance among various nitrogen transformation processes, strongly regulates the ecological function of both terrestrial and aquatic ecosystems [28,29]. Locating at the transition between land and ocean [30], mangroves are characteristiced by nitogen limitation, but how the dynamic nitrogen transformation processes in uence the extent of mangrove nitrogen limitation is still a vacancy. To ll this gap, we investigated the in-depth functional pro le of the nitrogen transformation proceses (e.g. nitrogen xation and its downstream processes) in the mangrove ecosystem which is crucial for our better understanding of prevalent nitrogen limitation across wetlands.
In this study, we aimed to unveil how biological nitrogen xation vary with depth of mangrove sediments and identify the key factors governing the in-depth pro le of diaztrophic communities. In attempt to achieve these goals, we analyzed nitrogen xation rate, diazotrophic communities and their key functional genes at 10 depths (0-100 cm) of mangrove sediments through the acetylene reduction assay, nifH gene amplicon sequencing and metagenome sequencing. We also constructed draft genomes of key nitrogen-xing bacteria from these samples and determined potential metabolic pathways. This study reveals the depth-related variability of biological nitrogen xation and diazotrophic communities in mangrove sediments, and advances our understanding of nitrogen limitation mechanisms in mangrove ecosystems.

Site description and sampling
The sampling site is located at the Qi'ao Mangrove Wetland Park (22°26'12. Sediment physicochemical properties analysis NFR was measured by acetelene reduction assay [32]. Brie y, fresh sediment (10.0 g) was put into a 120-mL serum vial. The vials were sealed with rubber stoppers and 10% of the headspace was replaced with pure and fresh acetylene (C 2 H 2 ) before they were incubated in dark at 25°C. After incubation for 48 h, 200 µL headspace gas was taken out to measure the concentration of ethylene (C 2 H 4 ) by Agilent gas chromatograph (HP7890B, Agilent, USA) equipped with a ame ionization detector and a HP-PLOT MoleSieve5A capillary column (30.0 m × 530 µm × 50 µm) (Agilent, USA ), and He was used as a carrier gas [33].
Ammonia, nitrite and nitrate were determined by a multimode microplate reader (Varioskan LUX, Thermo Scientifc, USA) after extraction from 2.0 g fresh samples with 2 M KCl. Fully digestion method was used to extract total ions and AB-DTAP extraction method was used to extract the available ions from 0.5 g airdried sediments separately [34]. All of these trace element concentrations were determined by an inductively coupled plasma-optical emission spectrometer (ICP-OES, Avio 500, Perkin Elmer, Singapore). A sequential extraction protocol was used for ferrous and ferric ions from 0.5 g fresh sample [35], and iron content was measured by ICP-OES (Avio 500, Perkin Elmer, Singapore). Sediment moisture was measured by drying 10.0 g fresh sediments at 105°C to a constant weight. Sediment pH and salinity were measured with 2.0 g dry sediment in 1:2.5 (sediments/water) and 1:5 (sediments/water) suspension with a pH meter (SevenCompact210, Mettler-Toledo, USA) and a salinity meter (EUTECH SALT6+, Thermo Scienti c, USA).

Sediment microbial community DNA extraction
The sediment microbial community DNA was extracted from 5.0 g sediment using a modi ed sodium dodecyl sulfate extraction method [36]

PCR ampli cation of nifH genes and amplicon sequencing
The nifH gene was ampli ed using the speci c primer pair PolF (5'-TGCGAYCCSAARGCBGACTC-3') and PolR (5'-ATSGCCATCATYTCRCCGGA-3') with an expected fragment length of approximately 320 bp [37]. Both forward and reverse primers were tagged with an Illumina adapter sequence, a primer pad and a linker sequence. The reaction system for each sample was 50 µL, including 25 µL Phusion High-Fidelity DNA Polymerase (NEB, Inc., USA), 2 µL forward and reverse phasing primer, 5 µL DNA template and 16 µL RNase free Ultrapure water. The ampli cation was conducted in a BIO-RAD T100™ thermal cycler (Bio-Rad Laboratory, Hercules, USA) under the following conditions: initial denaturation at 94°C for 5 min, followed by 30 cycles of 94°C for 30 s, 55°C for 30 s and 72°C for 1 min, with a nal extension at 72°C for 10 min.
PCR products were then puri ed using AMPure XP Beads Kit (NEB, Inc., USA). Puri ed DNA was quanti ed by QuantiT™ dsDNA HS Reagent (Thermo Fisher Scienti c, Inc., USA) and diluted to a concentration of 2 nM before sequencing. Paired-End nifH amplicon sequencing was performed using an Illumina Hiseq 2500 sequencer (Illumina, Inc., CA, USA) at Personalbio Biotechnology Co., Ltd. (Shanghai, China).
The quality ltering and pre-processing of raw sequences were performed on the Linux and Galaxy pipeline (http://mem.rcees.ac. cn:8080/). The primers were rstly eliminated by Cutadapt [38]. The low quality sequences (quality score < 20) were removed by Trimmomatic, and then forward and reverse reads were combined using FLASH [39]. Combined sequences of < 285 bp and > 350 bp were eliminated, and sequences contained one or more ambiguous base(s) ("N") were also removed. The chimeras were identi ed and eliminated using UCHIME [40]. Framebot software was used to correct potential frameshifts caused by sequencing errors [41], and only DNA sequences that covered > 30% of reference nifH protein translations were retained for further analysis. Operational taxonomic units (OTUs) were clustered at a 95% cutoff [42] of similarity level with protein reference sequences by using Quantitative Insights into Microbial Ecology (QIIME) implementation of UPARSE [43]. Taxonomic assignments for nifH OTUs were carried out an 80% [15] identity cut-off by searching representative sequences against reference nifH sequences with known taxonomic information.

Shotgun metagenome sequencing and data analysis
For sur cial (0-10 cm), middle (50-60 cm), and deep (90-100 cm) sediment samples, one microgram of DNA was used for metagenome sequencing library preparations combined with NEBNext® UltraTM DNA Library Prep Kit for Illumina (NEB, USA) as recommended by the manufacturer. Index codes were added to attribute sequences to each sample. The samples were puri ed (AMPure XP system), and the libraries were checked using Agilent 2100 Bioanalyzer (Agilent Technologies, CA) and quanti ed using real-time quantitative PCR. After cluster generation was performed on a cBot Cluster Generation System, pairedend reads (PE150) were performed on the Illumina platform. Low-quality (quality score ≤ 38; base N > 10 bp; the overlap length between adapter and reads > 15 bp) paired-end reads were ltered. The metagenomic assembly was performed using Megahit at default mode [44].

Metagenomic binning and metagenome assembled genome (MAG) annotation
Genome assembly and binning were performed according to the MetaWRAP pipeline [47]. The sequences were assembled with MEGAHIT (options: -mink 21 -maxk 141 -step 12) [44] to generate contigs. Genome binning of assembled contigs were done using MetaBAT2 [48] and MaxBin2 [49], and the resulting bins were consolidated with the Bin_re nement module. The consolidated bin sets were further improved by the Reassemble_bins module to generate MAGs. The quality of MAGs was evaluated with CheckM (v1.0.5; Table S8). MAGs were analyzed further if their completeness were more than 50% and their contaminations were below 10% (Supplementary Data 1). The abundance of each MAG was expressed as genome copies per million reads and calculated with Salmon [50]. Taxonomic assignments of MAGs were performed using the GTDB-Tk [51]. Gene prediction for MAGs was performed using prodigal (V2.6.2, default settings), and the predicted genes were further annotated using KAAS (KEGG Automatic Annotation Server) [52]. Additionally, we utilized a custom hmmer as well as the Pfam and TIGRFAM databases to search for key metabolic marker genes using hmmsearch and custom bit-score cutoffs [53].

Statistical analysis
Relationships between physicochemical characteristics, depth, diversity indices and relative abundance of diazotrophic communities were performed by linear regression analysis with GraphPad Prism (version 7.0). The other statistical analyses were conducted using R packages including vegan and ggplot2 (R Foundation for Statistical Computing, Vienna, Austria). The diazotrophic community dissimilarity based on Bray-Cutis distance matrices were evaluated by permutational multivariate analysis of variance (ADONIS) and analysis of similarities (ANOSIM). The contribution of environmental characteristics to diazotrophic community structure based on the Bray-Curtis distance matrices were assessed by analysis of variance (ANOVA). Additionally, we performed pairwise comparisons the functional genes involved in the nitrogen cycle of three selected sediment samples using STAMP (parameters: Fisher's exact test, Asymptotic-CC, Benjamini-Hochberg FDR), and screened out the functional genes with signi cant differences (p < 0.05).
We constructed structural equation model (SEM) to determine the direct and indirect contributions of sediment physicochemical properties, sediment depths, diazotrophic community richness and structure, to the in-depth pro le of NFR. Sediment moisture and salinity were chosen in SEM, since both of them were identi ed as the signi cant predictors of the diazotrophic community structure based on the redundancy analysis (RDA). SEM can partition direct and indirect effects that one variable might have on another, estimate and compare the strengths of multiple effects, and ultimately provide mechanistic information on the drivers of diazotrophic communities and NFR [54]. SEM analysis was performed via the robust maximum likelihood evaluation method using AMOS 22.0 (AMOS IBM, USA). The SEM tness was evaluated on the basis of a non-signi cant chi-square test (P > 0.05), the goodness-of-t index (GFI), the comparative t index (CFI), and the root mean square error of approximation (RMSEA).

Results
In-depth pro le of NFR and physicochemical characteristics in mangrove sediments We determined the NFR of 10 depths of the mangrove sediments using acetylene reduction assay (Additional le: Fig. S2). The NFR uctuated in the range of 0-0.20 nmol/(g*h), and the average NFR was 0.031 nmol/(g*h). Clearly, we observed a depth-related variability of NFR, which reached a maximum at the depth of 90-100cm. Compared to the sur cial layers (0-50 cm), the deeper layers (50-100 cm) showed higher NFR and a signi cantly (R 2 = 0.42, p < 0.05) postitive correlation between NFR and depth was detected in mangrove sediments, as revealed by a linear regression analysis (Fig. 1a).
In-depth pro le of physicochemical characteristics was also examined in mangrove sediments. Salinity in mangrove sediments ranged from 0.43‰ to 1.54‰, and increased with depth (Additional le: Fig. S2,  S3). Conversely, moisture of sediments, with an average of 52%, showed a consistent decreasing with depth (Additional le: Fig. S2, S3). pH and total iron concentration decreased from 0 cm to 50 cm, and then increased from 50 cm to 100 cm (Additional le: Fig. S3). However, the concentrations of NH 4 + , NO 2 − , NO 3 − , available Fe and Fe 3+ exhibited no momentous differences with depth (Additional le: Fig.   S3). The linear regression analysis showed that moisture had a negative correlation with depth (R 2 = 0.61, p < 0.05) (Additional le: Fig. S3d), but salinity had a positive correlation with depth (R 2 = 0.93, p < 0.0001) (Additional le: Fig. S3f). It is expected that such changes of physicochemical properies may in uence the in-depth pro le of diazotrophic communities in mangrove sediments.

In-depth pro le of diazotrophic communities in mangrove sediments
To investigate biotic factors contributing to the increased NFR with depth, we analyzed diazotrophic communities in mangrove sediments by sequencing nifH gene amplicons. From all samples, we obtained a total of 2,253,352 high-quality nifH sequences, and the nifH sequences were assigned into 974 operational taxonomic units (OTUs) and 59 genera after trimming (Additional le: Table S1). Notably, we observed the in-depth variation of diazotrophic communities in mangrove sedments. Both Shannon and Chao1 indices showed signi cant (R 2 = 0.47, p < 0.05) negative relationships with depth (Fig. 1a), highlighting the decrease of diazotrophic community diversity and richness with depth. Meanwhile, diazotrophic diversity indices showed negative relationships with NFR (Fig. 1b), namely, diazotrophic community diversity and richness decreased with the increasing of depth. Furthermore, principal coordinates analysis (PCoA) showed that diazotrophic microbial communities in mangrove sediments were well separated by depth, and 50 cm was identi ed as the partitioning depth based on the Bray-Curtis distance (Additional le: Fig. S4). Two nonparametric tests (ANOSIM and ADONIS) further veri ed such signi cant variations (p < 0.001) of diazotrophic communities between sediment layers above and below 50 cm. Together, these ndings determined the reduced diversity and varied structure of diazotrophic communities with depth of mangrove sediments.

Key diazotrophs contributing to increased NFR with depth
Taxonomic analysis of diazotrophic communities showed that bacteria (92.53%) dominated biological nitrogen xation in mangrove sediments, and a few archaea (such as Methanomicrobia belong to Euryarchaeota) (0.18%) had the potential to x nitrogen. At the phylum level, Proteobacteria was prevalent diazotrophs in mangrove sediments, which is consistent with terrestrial ecosystems [42]. Among Proteobacteria, Deltaproteobacteria occupied the largest proportion with an average relative abundance of 31%, followed by Gammaproteobacteria and Alphaproteobacteria (Additional le: Fig.  S5a). At the genus level, aerobic diazotrophs such as Methylomonoas and Heliobacterium had higher relative abundances in the sur cial layers (0-50 cm), while diazotrophic members represented by Agrobacterium, Desulfuromonas, Klebsiella and Azospira had higher relative abundances in the deeper layers (50-100 cm) (Additional le: Fig. S5b).
We performed Linear discriminant analysis Effect Size (LEfSe) on diazotrophic communities in mangrove sediments. Interestingly, diazotrophs presented manifest hierarchical clustering when 50 cm was set as the partitioning depth (Additional le: Fig. S6). The LEfSe results showed that, most of the diazotrophs enriched in sur cial sediments (above 50 cm) belonged to Proteobacteria, such as Burkholderiales, Nitrosomonadales, Desulfarculales, Myxococcales and Methylococcales. Conversely, Agrobacterium a liated to Alphaproteobacteria, Azotobacter a liated to Deltaproteobacteria, and Dehalococcoides a liated to Chloro exi appeared conspicuous clustering in sediments below 50 cm (Additional le: Fig.  S6). The diazotrophs speci c for deep sediments tended to be crucial for biological nitrogen xation in mangrove sediments because their higher relative abundances in deeper sediments corresponded with stronger NFR (Fig. 2b).

Relationships among sediment physicochemical characteristics, diazotrophic communities and NFR
To reveal the relationship between diazotrophic communities and environmental factors, we performed RDA to estimate the factors that had signi cant (p < 0.01) in uences on diazotrophic communities (Fig. 3a). The results showed that nitrate, ammonia, moisture and salinity were the important driving factors of diazotrophic communities in mangrove sediments. Specially, moisture and salinity were the only two environmental characteristics signi cantly related to depth (Additional le: Fig. S3d, f).Through all detected depths, Pearson's correlation analysis revealed that diazotrophic communities in terms of Chao 1 index and PCoA1 of diazotrophic communities were signi cantly (p < 0.05) negative with salinity (Additional le: Table S4, S5), suggesting that salinity and moisture were the main environmental factors driving the in-depth variation of diazotrophic communities in mangrove sediments.
To further evaluate the direct and indirect effects of depth, sediment properties (including moisture and salinity), and diazotrophic community richness and structure on NFR, we conducted SEM analysis based on known relationships among these observed variables (Fig. 3b). Consistent with the linear regression (Additional le: Fig. S3d, f) and RDA results (Fig. 3a), depth showed a directly positive effect on salinity and a directly negative effect on moisture, and salinity exerted signi cant effects on the diazotrophic community structure (Fig. 3a). Among all the observed variables in the model, only the diazotrophic community structure had a direct effect on NFR (Fig. 3c) although depth could indirectly in uence NFR by strongly affecting sediment salinity (Fig. 3b). Collectively, these results indicated that salinity-driven diazotrophic community structure played a critical role in determining the in-depth pro le of NFR in mangrove sediments.
In-depth pro le of biological nitrogen xation and its downstream processes in mangrove sediments Samples from three depths (M1: 0-10 cm, M2: 50-60 cm, M3: 90-100 cm) of mangrove sediments were selected for metagenomic sequencing analysis of N-cycling gene pro les across the sur cial, middle and deep sediments. We rst proposed an in-depth schema to illustrate metabolic potentials for various nitrogen cycling processes based on key N-cycling functional genes (Fig. 4a-c). Notably, a total of eight pathways consistently revealed depth-related variations in terms of functional gene abudances (Fisher's exact test, p < 0.05), including nitrogen xation, nitri cation, denitri cation, dissimilatory nitrate reduction to ammonium (DNRA), assimilatory nitrate reduction, ammonia assimilation, nitrate assimilation and organic N decomposition.
Consistent with the trend of diazotrophic activity (NFR), the gene clusters for nitrogen xing (nifH/D/K) increased in abundance with depth. Compared to the sur cial layer (M1: 0-10 cm), the abundance of nitrogen xation genes in deep sediment increased by 41.9% (Fig. 4c). Such increasing trend also occurred in ammonia assimilation and assimilatory nitrate reduction (Fig. 4). Particularly, the functional genes (nasA, narB and nirA) involved in assimilatory nitrate reduction remarkably increased 1.5, 17.6 and 9.3 times from the sur cial layer to the deep sediment (Fig. 4c). By contrast, the abundance of functional genes involved in nitri cation (aomA, amoB, amoC and hao), denitri cation (nirK, nirS, norB and norC), DNRA (nrfA, nrfH, nirB and nirD) and organic N decomposition (ureA, ureB and ureC) signi cantly (p < 0.05) decreased with depth (Fig. 4). Taking the rate-limiting process of denitri cation as example, the abundance of napA/B decreased by 21.5% from the sur cial layer to the deep sediment (Fig. 4c). Overall, these functional gene patterns showed that both biological nitrogen xation and its downstream processes in mangrove sediments underwent the depth-related variation with divergent trends.

Versatile functions and adaptation strategies of diazotrophic MAGs
De novo assembly and binning of metagenomic sequencing data from three depths of mangrove sediments allowed the reconstruction of 3 archaeal and 64 bacterial MAGs (completeness > 50%, contaminated < 10%; Additional le: Supplementary Data 1). Given that metagenomic sequencing generated enormous data accompanied by tremendous undiscovered information, we inferred their potential physiological capabilities by annotating genes using the KAAS and TIFRFAM databases. Among all 67 high-quality and high-completion MAGs, three MAGs possessed genes for nitrogen xation (nifH/D/K), namely M2.bin.35, M2.bin.46 and M3.bin.42, which represented one Chloro exi and two Desulfurmonadales (Fig. 5, Additional le: Supplementary Data 2). Interestingly, these three MAGs consistently contained genes associated with other nitrogen cycling processes, such as ammonia assimilation and the complete DNRA pathway (Fig. 5

Discussion
Mangroves are considered as typical nitrogen-limited ecosystems [56]. Characterizing the biological nitrogen xation and diazotrophic communities is, therefore, crucial to fully elucidate the nutrient status and ecological functions of mangrove ecosystems. In this study, we systematically examined the in-depth pro le of NFR and diazotrophic communities across 10 depths of mangrove sediments. A prominent nding is that the diversity of diazotrophic communities decreased whereas NFR increased with the increase of sediment depth. Such depth-related variability of biological nitrogen xation could be further supported by our metagenomic sequencing analysis, which revealed an elevated abundance of genes related to biological nitrogen xation in deeper sediments. Moreover, the metagenomic binning and functional annotation of diazotrophic MAGs provide a genetic evidence for the functional versatility and adaptive strategies of diazotrophs under the low-oxygen and oligotrophic conditions of deeper sediments. These results provide novel insights into the depth-related variability of NFR and diazotrophic communities, and advance our understanding of the relationship between biological nitrogen xation and nitrogen limitation in mangrove ecosystems.
Our in-depth survey of mangrove sediments revealed a clear divergence of diazotrophic community composition at the partition depth of 50 cm. Although the diversity of diazotrophic communities was lower in deep sediments than in surface ones, the deep sediment-speci c diazotrophs, including Agrobacterium and Azotobacter, contributed to a higher NFR in deep sediments. There are two main reasons for this observation. First, the lifestyle of diazotrophs was related to nitrogen xation e ciency. Previous studies found that microaerophilic and anaerobic diazotrophs often exhibited a higher nitrogen xation e ciency than aerobic diazotrophs [57]. In our study, the dominant diazotroph in the deep sediment was Agrobacterium sp., which has been reported to be a typcial facultative anaerobe with the capability of anaerobic respiration in the presence of nitrate [58]. In line with this opinion, our MAG annotation indicated many genes related to low oxygen-dependent pathways, and determined a facultative anaerobic lifestyle of diazotrophs in deep mangrove sediments. Thus, the deep layer of mangrove sediments with lower oxygen concentrations could provide suitable conditions for microaerophilic/anaerobic diazotrophs to e ciently x nitrogen [59], which is well consistent with high NFR in the deep sediment. Second, decreasing oxygen and ammonium concentrations with depth may facilitate to ensure the high activity of nitrogenase. It has been reported that metalloproteins of nitrogenase were extremely sensitive to oxygen and irreversibly destroyed under high oxygen conditions [60]. For example, the activity of MoFe protein and Fe protein in Azotobacter was reduced in a few minutes when they were exposed to air [61]. Besides, high ammonium concentrations were well known to inhibit nitrogenase synthesis through regulating the transcription of nifA [55]. When exposed to high ammonium concentrations, the protein NifL in Azotobacter could inhibit the activity of nifA, and subsequently result in the inactivation of nitrogenase [62]. These ndings suggest that under the uctuated physicochemical gradients, the changes of nitrogenase activity and diazotrophic communities jointly contribute to the in-depth variability of biological nitrogen xation in the mangrove sediment.
Located at the transition between ocean and land, mangrove sediments experienced the tidal uctuation day after day [63]. Probably due to the continuous scouring of sur cial sediments by tides and the deposition of salinity in depth [64,65], a continuous increase of salinity with depth of mangrove sediments was observed in our study. More interestingly, in agreement with the paddy soil [66] and salt marsh [67], salinity was identi ed as the most important environmental lter for shaping the diazotrophic community structure in mangrove sediments, as revealed by our RDA and SEM results. Such niche partition of diazotrophs across sediment depths may be closely tied to their salt tolerance and associated strategies. In the deep sediment layers with high salinity, we observed that the diazotrophic community was dominated by Azotobacter and Agrobacterium, which belong to more-salt-tolerant diazotrophs [68,69] and tend to replace less-salt-tolerant ones in deep sediments [70]. Moreover, these diazotrophs thriving in high-salinity sediments potentially apply the "low-salt-in" strategy to balance the osmotic potential of the cytoplasm [71]. This strategy was reported for accumulating low-molecular-weight organic compounds within the cell to maintain an osmotic equilibrium with the surrounding environment by excluding salt ions from the cytoplasm [72]. To support the adaptive strategy of diazotrophs to high salinity, we did observe that diazotrophic MAGs contained glycine betaine reductase and glucose/sorbosone dehydrogenase, which are well known osmolytes to balance the osmotic pressure created by the hypersaline habitat [73,74]. Collectively, our study highlights the role of salinity in controlling the in-depth structure of the diazotrophic communities and indicates the putative strategy of diazotrophs for high salinity tolerance in mangrove sediments.
Under nitrogen limited conditions, the nitrogen partitioning between mangrove species and sediment microorganisms altered the extent of nitrogen limitation in mangrove ecosystems [24,75]. Previous studies reported that mangrove species could acquire nitrogen from the sediments under both inorganic (nitrate and ammonium) and organic (e.g., urea, amino acid) forms [26,76]. However, the uptake of ammonium and nitrate by plants was down regulated under carbon limiting conditions [77]. Therefore, in deep mangrove sediments with low organic carbon content [78], the main form of nitrogen uptaken by mangrove species was assumed to be amino acids. In support of this assumption, our metagenomic sequencing analysis showed that with the increasing depth of mangrove sediments, both biological nitrogen xation and ammonia assimilation pathways were enhanced, whereas ammonia and organic nitrogen loss by denitri cation and organic nitrogen decomposition pathways was weakened. These results indicate that there were probably more amino acids for mangroves to take up, thus relieving nitrogen limitation in the deep layer of mangrove sediments. As the balance between nitrogen cycling processes determined whether nitrogen sink or source is occurring in mangrove ecosystems [79], we also inferred that the notable variation of nitrogen cycling processes with depth would result in the increased organic nitrogen burial and decreased N 2 O/N 2 emissions in deep sediments, which suggests a possible role of mangrove ecosystems as a potential nitrogen sink.

Conclusions
In summary, this study illustrates the depth-related variability of biological nitrogen xation in mangrove sediments from the perspective of both NFR and diazotrophic communities. The diversity of diazotrophic communities decreased with depth of mangrove sediments, but the NFR and nitrogen xation-related gene abundances increased. The salinity-driven structure of diazotrophic communities showed a clear divergence at the partition depth of 50 cm, and high abundances of Azotobacter and Agrobacterium at 50-100 cm sediments contributed to an elevation of NFR in deep mangrove sediments. Accompanied by such an elevation, our metagenomic sequencing analysis indicated that ammonia and organic nitrogen loss by denitri cation and organic nitrogen decomposition pathways was weakened with depth. These in-

Availability of data and materials
The nucleotide sequences and metagenomic data of microbial communities in mangrove sediments were deposited in SRA database under accession number PRJNA694572 and PRJNA698080. The authors declare that the primary data supporting the ndings of this study are available within this article and in the additional les. Extra data supporting the ndings of this study are available from the corresponding author upon request.

Ethics approval and consent to participate
This manuscript does not report data collected from humans or animals. Therefore, ethics approval and a consent to participate are not necessary.

Consent for publication
Not applicable.