Nitrogen fixation and diazotroph diversity in groundwater systems

Abstract

Biological nitrogen fixation (BNF), the conversion of N₂ into bioavailable nitrogen (N), is the main process for replenishing N loss in the biosphere. However, BNF in groundwater systems remains poorly understood. In this study, we examined the activity, abundance, and community composition of diazotrophs in groundwater in the Hetao Plain of Inner Mongolia using ¹⁵N tracing methods, reverse transcription qPCR (RT-qPCR), and metagenomic/metatranscriptomic analyses. ¹⁵N₂ tracing incubation of near in situ groundwater (9.5–585.4 nmol N L⁻¹ h⁻¹) and N₂-fixer enrichment and isolates (13.2–1728.4 nmol N g⁻¹ h⁻¹, as directly verified by single-cell resonance Raman spectroscopy), suggested that BNF is a non-negligible source of N in groundwater in this region. The expression of nifH genes ranged from 3.4 × 10³ to 1.2 × 10⁶ copies L⁻¹ and was tightly correlated with dissolved oxygen (DO), Fe(II), and NH₄⁺. Diazotrophs in groundwater were chiefly aerobes or facultative anaerobes, dominated by Stutzerimonas, Pseudomonas, Paraburkholderia, Klebsiella, Rhodopseudomonas, Azoarcus, and additional uncultured populations. Active diazotrophs, which prefer reducing conditions, were more metabolically diverse and potentially associated with nitrification, sulfur/arsenic mobilization, Fe(II) transport, and CH₄ oxidation. Our results highlight the importance of diazotrophs in subsurface geochemical cycles.

Introduction

Biological nitrogen fixation (BNF) is a process by which microorganisms convert N₂ into bioavailable nitrogen (N); it contributes 50% of the annual total available N in the biosphere, sustaining active nitrification and N loss processes [1]. BNF is mediated by three types of nitrogenases with similar structure and function, but different active site metal cofactors: Nif (molybdenum-iron [Fe]), Vnf (vanadium-iron), and Anf (Fe-Fe) nitrogenases [2]. The nifH gene encoding the nitrogenase Fe protein is used as a biomarker for detecting diazotrophic abundance and community composition in aquatic and terrestrial environments [3, 4]. Cyanobacteria are the most important diazotrophs in marine environments; these include Trichodesmium, Richelia, and three groups of unicellular Cyanobacteria (UCYN-A, UCYN-B, and UCYN-C) [5–7]. In terrestrial ecosystems, diazotrophs mainly belong to Pseudomonadota, Bacillota, and Actinomycetota [8, 9].

N₂ fixation has also been detected in diverse habitats such as oceans, estuaries, lakes, and forests through ¹⁵N₂ tracing (direct) and acetylene reduction assays (indirect) [4, 10, 11]. Oxygen levels [12, 13], N [5, 14], phosphorus (P) [15, 16], carbon (C) [3, 10], and trace metal concentrations [9, 17] are commonly implicated as the main influencing factors of N₂ fixation, which vary significantly across different habitats. In addition, diazotrophs may use Fe, S, and C as terminal electron acceptors [18], and the coupling of Fe(III)/sulfate reduction and N₂ fixation may be widespread in marine sediments and soils [3, 19–22]. Thus, diazotrophs appear to be habitat-dependent stable sources of N in both oligotrophic tailings and hypereutrophic estuaries [10, 23], and exploring their distribution, activity, and controlling factors is crucial for understanding the environmental effects and interconnected biogeochemical networks of BNF.

As the largest distributed freshwater reservoir worldwide, groundwater plays a central part in the hydrological cycle through discharge and recharge dynamics [24]. Groundwater systems are light-deprived and generally characterized by long water residence times and geochemical heterogeneity [25]. Anthropogenic activities, such as agricultural irrigation, livestock breeding, and domestic wastewater discharge increase the input of organic C, N, and other chemicals into groundwater systems, which can accelerate N cycling [26–29]. The mineralization of organic N is the main source of NH₄⁺ in groundwater [30]. Studies of the groundwater N cycle have mainly focused on tracing N and its transformation to NH₄⁺, NO₂⁻, and NO₃⁻ [31, 32]. Researchers have also investigated the distribution of microorganisms related to N cycling in groundwater at varying oxygen levels through metagenomic and metatranscriptomic sequencing [1]. Denitrification, anammox, and Fe(III) reduction coupled with ammonium oxidation (Feammox) appear to be the dominant pathways of N₂ production in aquifers [28, 29]. The vadose zone connects surface soil and groundwater and probably provides considerable N₂ to groundwater via denitrification or anammox [33–35]. Active N₂ generation has also been observed in anoxic groundwater [1, 36]. Together, these findings indicate excess N₂ and reducing substances in groundwater [32]. However, still is known about N₂ fixation in groundwater. Filling this N₂ fixation knowledge gap would provide a more comprehensive understanding of the N cycle in groundwater systems.

N₂ fixation rate measurements have rarely been conducted in field groundwater systems. GeoChip analysis and high-throughput sequencing detected nifH genes in groundwater, suggesting that BNF is an essential source of groundwater N [32, 37, 38]. Recently, many new nitrogenases and nitrogenase homologs have been found in groundwater-associated Elusimicrobia genomes using metagenomic methods [39]. However, none of these studies has analyzed the N₂ fixation and diversity of diazotrophs in detail. Current knowledge of N₂ fixation in groundwater systems is based primarily on the functional prediction of 16S rRNA genes or metagenomic sequencing. N₂-fixing strains affiliated with Mesorhizobium and Azospira have been isolated from groundwater [40, 41]; however, their N₂ fixation activity has yet to be verified. Thus, there may be many active N₂-fixing microorganisms in aquifers that are involved in or influence other biogeochemical processes. Nevertheless, convincing evidence for N₂ fixation activity in groundwater is still limited.

In this study, we investigated the abundance, diversity, and community composition of native/active N₂-fixing bacteria in groundwater using reverse transcription PCR (RT-qPCR) and high-throughput sequencing by targeting nifH genes and transcripts. We also explored the geochemical factors controlling diazotrophs and the relationship between N₂ fixation and other biogeochemical processes based on metagenomic and metatranscriptomic analyses, and verified diazotroph activity in groundwater using ¹⁵N₂ tracing, single-cell resonance Raman (RR) spectroscopy with ¹⁵N stable isotope probing (SIP), and acetylene reduction assays.

Materials and methods

Sample collection and geochemical analysis

We collected groundwater samples in the Hetao Plain, Inner Mongolia, China (40°40–41°50’N, 106°45–107°90’E) (Fig. S1). The study region has an arid/semiarid climate, and its aquifers feature strong geochemical–microbial interactions involving N, C, S, arsenic (As), and Fe. Groundwater generally flows from oxidizing to reducing areas, potentially creating a redox gradient that supports multiple biochemical cycles [42]. The study area was described in detail in our previous study [32].

Groundwater samples (G1–G30) were collected directly from tube wells using Tygon tubing and a vacuum pump at depths of 15–100 m in Hangjinhouqi County, in the western part of the Hetao Plain. Before sampling, the wells were pumped slowly and uniformly (approximately 0.5 L/min) [43] for at least 20 min to obtain stable oxidation–reduction potential (ORP), temperature (T), and dissolved oxygen (DO). Water samples for determining N₂ fixation rates were subsampled into 1-L high-density polyethylene (HDPE) bottles (Nalgene, Rochester, NY, USA) that had been acid-washed and groundwater-rinsed three times; the bottled samples were kept in the dark in black plastic bags during collection and incubation. Groundwater biomass was collected by filtering approximately 10 L of water with 0.22 μm filters (Millipore, Burlington, MA, USA) for DNA extraction and 2500 L of water with tangential flow filtration using a Prep/Scale spiral-wound cartridge with a 0.22 μm molecular weight cutoff (Millipore) for RNA extraction. The standard filters were packaged into 50 mL sterile tubes (NEST Biotechnology, Wuxi, China) and then frozen for DNA extraction, qPCR, and nifH gene high-throughput sequencing. The cell suspension concentrated by tangential flow filtration was packed into 500 mL sterile nuclease-free HDPE bottles (Nalgene) and stored on dry ice. RNA was extracted for RT-qPCR and metatranscriptomic sequencing on the same day in the local Centers for Disease Control and Prevention (CDC) laboratory. The pipettes and centrifuge tubes used in the extraction were treated with 0.1% diethyl pyrocarbonate (Aladdin, Shanghai, China) to prevent RNase contamination. Groundwater samples for measuring dissolved organic carbon (DOC) were collected in 60 mL brown HDPE bottles (Nalgene) that were quickly acidified with 1% v/v HNO₃, sealed; samples were then stored on dry ice and determined within 3 days using a TOC analyzer (TOC-L; CPH, Shimadzu, Japan). Geochemical characteristics including T, pH, DO, ORP, and the concentrations of SO₄²⁻, NO₃⁻, NO₂⁻, NH₄⁺, Fe(II), and total iron (Fe_Tot), in water samples were measured in the field with a Hach spectrophotometers (HachDR890 and Hach40D, Loveland, CA, USA), according to the manufacturer’s instructions. Arsenic species were separated as described in our previous study [44].

Nucleic acid extraction, cDNA synthesis, and qPCR

DNA and RNA were extracted from all samples using the FastDNA SPIN (MP Biomedicals, Santa Ana, CA, USA) and RNeasy PowerSoil Total RNA (Qiagen, Hilden, German) kits, respectively, following the manufacturer’s protocols. DNA from N₂-fixing strains isolated from groundwater was extracted using the MiniBEST Bacteria Genomic DNA Extraction Kit v3.0 (TaKaRa Bio, Shiga, Japan). The quality and quantity of DNA/RNA were examined using a NanoDrop-2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and agarose (1%) (Biowest, Riverside, MO, USA) gel electrophoresis. The residual DNA in RNA was eliminated using Recombinant DNase I (TaKaRa Bio), and then cDNA was synthesized using a PrimeScript II First Strand cDNA Synthesis Kit (TaKaRa Bio). All DNA, cDNA, and RNA samples were stored at −80 °C for quantitative and sequencing analysis.

The abundance of nifH and 16S rRNA genes (V4 region) (DNA- and RNA-based) in groundwater samples was quantified (Applied Biosystems, Waltham, MA, USA); the primers and amplification conditions are described in Table S2. qPCR was performed in triplicate. Plasmids were extracted from Escherichia coli DH5α hosts using the MiniBEST Plasmid Purification Kit (TaKaRa Bio). Standard curves were constructed and quantified by 10-fold dilutions of the plasmids. The correlation coefficients (R²) for all the genes exceeded 0.95, and amplification efficiency in this study ranged from 92% to 97%.

PCR amplification, sequencing, and data analysis

The nifH genes in groundwater samples were amplified by PCR (T100, Bio-Rad, Hercules, CA, USA) with universal primers PolF (with an added unique 8-bp barcode sequence for sample sorting) and PolR [45]. The 16S rRNA genes in groundwater samples and N₂-fixer enrichments were amplified using primers 515 F and 806 R, targeting the 16S rRNA gene V4 region of microorganisms. The PCR reaction mixture and programs are described in detail in Table S2. The full-length 16S rRNA genes were amplified by PCR using primers 27 F/1492 R. The PCR products were detected using gel electrophoresis (1%) and purified using a MiniBEST agarose gel DNA extraction kit (TaKaRa, Bio).

The PCR products of the 16S rRNA genes were sent to Sangon Biotech (Shanghai, China) for sequencing using an Applied Biosystems 3730 DNA Analyzer and the sequences were deposited in the National Center for Biotechnology Information (NCBI) GenBank (accession nos: OQ561323-OQ561328). The nifH amplicons were paired-end sequenced (2 × 250 bp) on a NovaSeq platform (Illumina, San Diego, CA, USA) according to the standard protocol of Aikangjian (Wuhan, China). The 16S rRNA amplicons of groundwater and N₂-fixer enrichments were paired-end sequenced (2 × 250 bp) on a NovaSeq platform by Magigene Technology (Guangzhou, China).

The raw sequencing data were processed using a Galaxy-based pipeline in Ye Deng’s laboratory at the Research Center for Eco-Environmental Sciences, Chinese Academy of Science, China (http://mem.rcees.ac.cn:8080). Detailed processing descriptions are shown in Supporting Information S1. The raw sequencing data of nifH genes and 16S rRNA genes were deposited in the NCBI database under accession nos PRJNA937386, PRJNA996345, and PRJNA996367. A representative groundwater flow field (G1, G2, G27, and G30) was selected from the recharge areas (near a mountain, hypoxic conditions) to the discharge areas (anaerobic conditions) for metagenomic and metatranscriptomic analyses. Groundwater total DNA and RNA were sent to Beijing Novogene Bioinformatics Technology for metagenomic and metatranscriptomic sequencing using a HiSeq platform (Illumina). The raw sequences were deposited in the NCBI database (accession nos: PRJNA882225 and PRJNA884812). The raw sequences were trimmed, assembled, filtered for quality control, and annotated. Binning was conducted using MetaWRAP.26 v1.3 pipeline with the default parameters [46]. Further data processing is detailed in the Supporting Information S1 and our previous study [47].

Enrichment and isolation of N₂-fixer strains

Groundwater samples G1, G27, G28, G29, and G30, which were collected along a typical groundwater flow path [48], were used to enrich diazotrophs. First, a piece of filtered membrane was cut and transferred to 100 mL of sterilized enrichment medium (N-free Burk’s culture medium) containing (per liter) 20 g mannitol (BioFroxx; neoFroxx, Einhausen, Germany), 0.2 g KH₂PO₄, 0.8 g·K₂HPO₄, 0.2 g MgSO₄·7H₂O, 0.1 g CaSO₄·2H₂O (SCR), and trace amounts of Na₂MoO₄·2H₂O (Sigma-Aldrich, St. Louis, MO, USA) and FeCl₃ (Sigma-Aldrich). The initial pH of the medium was adjusted to 7.0 and then it was incubated at 30 °C and 120 rpm under anaerobic conditions (purged with N₂ for 40 min). After 5 days, 10 mL of the culture was removed and transferred into 100 mL of fresh Burk’s medium. The culture process was repeated three times, and N₂-fixer enrichments G1, G27, G28, G29, and G30 were cultured at 30 °C to determine the N₂ fixation activity. To isolate N₂-fixing strains, the enrichments were serially diluted 10-fold, and 50 μL cultures with 10⁻⁶ dilution were plated on LB solid medium (pH 7.0) containing 10 g tryptone (Oxoid, Basingstoke, UK), 5 g yeast extract (Oxoid), 10 g NaCl (SCR), and 15 g agar L⁻¹ (BioFroxx), and deionized water. All plates were incubated at 30°C for 5 days, and then single colonies were picked and subcultured on Burk’s medium three times in succession. The positive strains were preserved in 20% glycerol (SCR) at −80 °C for further experiments.

Characterization of N₂ fixation

Immediately after water sampling, N₂ fixation rates were measured at sampling stations in triplicate using the ¹⁵N-pulse tracing method following the incubation workflow (Fig. S2). Briefly, 4 mL ¹⁵N₂ (Cambridge Isotope Laboratories, > 98 atom% ¹⁵N₂, Lot #: I-24583/AR0664758) was injected into each HDPE bottle for incubation and then shaken gently for 15 min to dissolve tracer bubbles into the water samples [49, 50]. Then, the bottles were incubated at near in situ temperature ( ± 2 °C) in the dark. After approximately 16 h and 106 h, the incubation was stopped by filtration through precombusted (450 °C for 4 h) 25 mm glass fiber filters (GF75; Whatman, Maidstone, UK) to obtain particulate N (PN) and the culture for total dissolved N (TDN) was filtered through 0.2 µm syringe filters. Time zero (T0) control samples were processed as described above immediately after adding ¹⁵N₂ tracer [51]. PN and TDN concentrations were determined using the wet chemical oxidation method [52]. Then, the oxidized NO₃⁻ was measured using a chemiluminescence method [30, 53]. The δ¹⁵N of PN and TDN were analyzed using a GasBench-II (Thermo Fisher Scientific) connected to an isotope ratio mass spectrometer (Delta V Advantage; Thermo Fisher Scientific). We have tested the used ¹⁵N₂ gas bottle in this study for contamination with bioavailable non-N₂ ¹⁵N in previous studies [54, 55] as mentioned in Dabundo et al. [56]. Briefly, triplicate 2 mL ¹⁵N₂ gas and 10 mL surface seawater (frome South China and 0.2 μm filtration) were injected into 20 mL headspace vials, sealed with a septum stopper, and then shaken overnight. The δ¹⁵N of TDN was measured and compared with the δ¹⁵N of natural surface seawater samples (frome South China and 0.2 μm filtration). Values of δ¹⁵N TDN of the blank seawater and test seawater groups were 4.3 ± 0.2 ‰ and 4.2 ± 0.3 ‰, respectively, suggesting no contamination of the ¹⁵N₂ gas. Meanwhile, values of δ¹⁵N-NH₄⁺ and δ¹⁵N-NO_x⁻ in the control and experimental groups (T0) in different groundwater samples are shown in Table S3. In addition, the nitrogenase activity of the in situ groundwater was also characterized by the acetylene-reduction assays [57]. Detailed descriptions of the ¹⁵N analysis method, N₂ fixation rate calculation, and acetylene-reduction assays are provided in Supporting Information S2, S3.

The N₂-fixer enrichment and isolate activity were detected using the ¹⁵N₂ assimilation method [11]. In brief, 1 mL N₂-fixer enrichments and bacterial suspensions (containing approximately 10⁷ CFU) were inoculated in 60 mL gas-tight serum bottles containing 30 mL fresh N-free Burk’s medium (sterile and purged with argon for 30 min). Then, 3 mL of headspace air was extracted and replaced with 3 mL ¹⁵N₂ gas (99 atom%, Lot no.210610; Reer Technology, Shanghai, China) using a gas-tight syringe. The bottles were placed upside down and incubated in a shaker at a constant temperature of 30 °C and 120 rpm for 48 h. After incubation, the N₂-fixer enrichment and isolate suspensions were centrifuged at 8000 × g for 20 min at 4 °C for biomass collection. The samples were dried after being frozen overnight at −80 °C and then placed in a vacuum freeze dryer (Biocool AB, Stockholm, Sweden) for 2 days. Dried samples (0.03–0.05 mg) and urea (0.15–0.20 mg) were placed in a tin capsule for ¹⁵N-enrichment analysis using an Isolink-Delta V elemental analyzer (Thermo Fisher Scientific). The abundance of ¹⁵N in the N₂-fixer enrichments and isolates (¹⁵N%) was calculated as follows [58]:

$${\mathrm{A}}_{\mathrm{t}}{}^{15}\mathrm{N}\%=\frac{\left({\mathrm{M}}_{\mathrm{urec}}\times {\mathrm{N}}_{\mathrm{urec}}\times {}^{15}\mathrm{N}{\%}_{\mathrm{urec}}+{\mathrm{M}}_{\mathrm{bacteria}}\times {\mathrm{N}}_{\mathrm{bacteria}}\times {}^{15}\mathrm{N}{\%}_{\mathrm{bacteria}}\right)}{\left({\mathrm{M}}_{\mathrm{urec}}\times {\mathrm{N}}_{\mathrm{urec}}+{\mathrm{M}}_{\mathrm{bacteria}}\times {\mathrm{N}}_{\mathrm{bacteria}}\right)}$$

Where A_t¹⁵N% is the percentage of ¹⁵N in total N (measured by IRMS), M_urea and M_bacteria are the weights of urea and sample in the tin capsule, and the ¹⁵N% in urea (¹⁵N%_urec) is 0.367% (natural abundance). The total N contents in urea (N_urea) and bacteria (N_bacteria) were 46.7% and 9.8%, respectively.

To demonstrate N₂-fixer activity more strongly, we used an acetylene-reduction assay, single-cell RR spectroscopy, and ¹⁵N₂ SIP [58, 59]. The cultivation methods and procedures are described in Supporting Information S3.

Statistical analysis

Principal component analysis (PCA) was performed to evaluate differences in geochemical parameters based on Bray‒Curtis dissimilarities. The influence of geochemical factors on diazotrophic community composition was determined using R v4.1.2 for Pearson correlation and Canoco5 [canonical correlation analysis (CCA)] software. Nonmetric dimensional scaling was used to evaluate differences in diazotrophic community composition based on the Bray‒Curtis dissimilarity matrix of operational taxonomic units using the R vegan package. Pearson correlations between gene abundances and geochemical characteristics in groundwater were computed using R v4.1.2 and Chiplot (https://www.chiplot.online/tvbot.html). Co-occurrence and coexpression networks of microbial interactions in the groundwater dataset were calculated using the R igraph package and visualized in Gephi v0.9.7 [60, 61]. Data were plotted and fitted using GraphPad Prism v9 (GraphPad Software, San Diego, CA, USA). Phylogenetic trees of all strains were constructed using MEGA X.

Results

Geochemical parameters of groundwater samples

The geochemical parameters of groundwater samples showed that DO concentrations ranged from 0.4 to 2.2 mg L⁻¹, and varied consistently with the ORP values (−168.8 to 88 mv) (Fig. S3 and Table S1). PCA divided the groundwater samples into two groups: oxidizing and reducing areas (Fig. S4A). Relatively high concentrations of electron donors, such as As(III), Fe(II), NH₄⁺, and DOC, accumulated in the reducing area (Fig. S3). The NO₃⁻ concentrations were higher in the oxidizing area than in the reducing area (p < 0.01) (Fig. S4B). SO₄²⁻ concentrations were also higher in the oxidizing area and were negatively correlated with As(III) (p < 0.05) and Fe(II) (p < 0.01) (Figs. S3C, S4B). Groundwater samples were either neutral or weakly alkaline (pH 7.4–9.4). A portion of the groundwater temperatures ranged from 11.2 to 17.7 °C and increased with ORP. Total N (TN) concentrations fluctuated between 2 and 8 mg L⁻¹ and were negatively correlated with SO₄²⁻ (Fig. S4B). These characteristics suggested a significant difference in geochemical conditions between oxidizing and reducing groundwater samples.

Quantification of nifH genes in groundwater samples

We assessed the abundances of the nifH genes in groundwater samples (Fig. 1, Table S4, and Table S5). At the gene level, the average abundance of nifH was 2.9 × 10⁷ copies L⁻¹ in the oxidizing area and 2.3 × 10⁷ copies L⁻¹ in the reducing area. The relative abundance of nifH (nifH/16S rRNA) genes was significantly elevated in the oxidizing area (p < 0.05) (Fig. 1A). Pearson correlation and linear regression analyses indicated that nifH/16S rRNA genes were positively correlated with DO (p < 0.05, r = 0.46). By contrast, the average abundance of nifH transcripts in the oxidizing area (1.2 × 10⁵ copies L⁻¹) was lower than that in the reducing area (3.2 × 10⁵ copies L⁻¹). The distribution of nifH/16S rRNA genes was consistent with the trend in transcript abundance (Fig. 1B). Correlation analysis showed that nifH/16S rRNA genes were negatively correlated with DO (p < 0.05, r = –0.65) at the transcriptional level, and Fe(II) was positively correlated with the expression of nifH genes (p < 0.001, r = 0.84) (Fig. S4B). PCA also showed differences in nifH/16S rRNA genes and their expressions under redox conditions (Fig. S4A).

Fig. 1. Distribution of nifH gene abundances and their response to environment variables.

Comparison of (A) DNA-based and (B) RNA-based nifH/16S rRNA genes (%) between the reducing area and oxidizing area and linear correlations between nifH/16S rRNA genes and DO, ORP, Fe(II), and NH₄⁺ concentrations. The symbol * indicates significant differences between the reducing area and oxidizing area at p < 0.05. Solid lines represent linear regression modules. The shaded area represents the 95% confidence interval.

Diversity, community composition, and influencing factors of diazotrophs

Both DNA- and RNA-based nifH genes were sequenced to investigate the community composition of N₂-fixing bacteria in groundwater (Fig. 2). The Shannon indexes of DNA- and RNA-based diazotrophic communities were higher in the reducing area than in the oxidizing area (p < 0.05 and p < 0.01, respectively) (Fig. 2A), suggesting that redox conditions influence the diversity of groundwater-associated diazotrophs. However, gene and transcription levels influenced the community composition of diazotrophs more than redox conditions (Fig. 2B and Table S6). Phylum Pseudomonadota dominated N₂-fixers in groundwater. At the class level, Gammaproteobacteria, Alphaproteobacteria, and uncultured populations were the most abundant diazotrophs in nifH amplicons (Fig. 2C), metagenomes, and metatranscriptomes (Fig. S5). The amplicon and genome results consistently showed that Cyanobacteria are native diazotrophs in groundwater (Figs. 2C and S5A). Further investigation of nifH amplicons at the genus level indicated that Stutzerimonas, Pseudomonas, and uncultured microbial populations were abundant in all groundwater samples, whereas Klebsiella was significantly enriched in the RNA-based analysis (Fig. 2D). The higher contribution of the uncultured microbial populations at the transcriptional level suggested that there are many unknown groups of active N₂-fixing microbes in the underground environment.

Fig. 2. Diversity and community composition of native and active diazotrophs in groundwater samples.

A Comparison of the diversity index between oxidizing and reducing areas based on DNA and RNA levels (* and ** represent p < 0.05 and 0.01, respectively). B Nonmetric multidimensional scaling (NMDS) based on the Bray‒Curtis distance of diazotrophic communities. The relative abundances at the (C) class and (D) genus levels are based on the DNA/RNA level nifH gene amplicon sequencing.

Environment–microbe interactions between N₂ fixers and geochemical factors were characterized by CCA and Pearson correlation analyses (Figs. S6 and 3). CCA revealed that ORP, DOC, and NO₂⁻ (p < 0.05) were significantly correlated with the native diazotrophic communities, whereas active diazotrophs were potentially influenced by TN and NH₄⁺ concentrations (p < 0.05) (Fig. S6). Pearson correlation showed that some diazotrophic populations were positively or negatively correlated with Fe(II), As(III), and DOC (Fig. 3). The abundances of aerobic or facultative anaerobic diazotrophs such as Azospirillum, Rhizobium, Stutzerimonas, Pseudomonas, and Desulfuromonas were positively correlated with ORP, NO₃⁻, and TN (Fig. 3A). However, Desulfuromonas, Rhodopseudomonas, and Azoarcus transcript abundances were positively associated with Fe(II), As(III), and NH₄⁺ (Fig. 3B), suggesting that N₂ fixation might tightly interconnect with N, Fe, and As transformation. The abundances of Klebsiella, Rhizobium, Thiothrix, Sphingomonas, Citrobacter, Desulfomicrobium, and Rhodopseudomonas were positively related to DOC, which implied that DOC might be an important electron donor for diazotrophs in groundwater.

Fig. 3. Relationship between geochemical parameters and diazotrophic communities.

A DNA level. B RNA level. The symbol * indicates Pearson correlation with significance (*, **, and *** represent p < 0.05, 0.01, and 0.001, respectively).

Distribution, composition, and potential interconnected biogeochemical networks of the nitrogenase family in groundwater

Metagenomics and metatranscriptomics were obtained from groundwater samples G1, G2, G27, and G30 and 2.4–4.2 × 10⁷ and 2.3–3.4 × 10⁷ trimmed reads were obtained for the metagenomic and metatranscriptomic datasets, respectively (Table S7). For the potential N₂ fixation function of the microbial communities, the relative abundances of KEGG-annotated genes encoding Nif, Vnf, and Anf are shown in Fig. S7. Based on the metagenomic analysis, 17 (sub)nitrogenases were annotated in groundwater, Nif (nifA, nifU, nifV, nifB, nifK, nifD, nifH, nifZ, nifN, nifQ, nifW, nifT, nifHD1, and nifHD2) and Anf (anfG). N₂ fixation genes were more abundant in the oxidizing area than in the reducing area (Fig. S7A). Metatranscriptomic annotation results found 19 (sub)nitrogenases, including traditional Nif (nifA, nifU, nifL, nifD, nifE, nifK, nifB, nifN, nifV, nifH, nifX, nifQ, nifZ, nifW, nifT, nifHD1, and nifHD2), selective nitrogenase Anf (anfG), and Vnf (vnfH) (Fig. S7B). Contrasting the metagenomic results, the transcriptional nitrogenases in groundwater were mainly distributed in the reducing area, except for vnfH.

Pearson correlations between nitrogenase abundances and other critical functional genes of N, S, CH₄, Fe, and As metabolism processes suggested that N₂ fixation might be more positively linked to other biogeochemical cycles at the transcriptional level (Figs. 4 and S8). For example, most of the N₂ fixation genes were positively associated with those related to denitrification (nirS and norB) and dissimilatory SO₄²⁻ reduction (aprB, dsrA, and dsrB) at the DNA level (Fig. S8A and Table S9). By contrast, the transcriptional abundances of nitrogenase clusters were strongly correlated with nitrification (amoA, amoB, amoC, nxrA, and nxrB), denitrification (nirS), dissimilatory SO₄²⁻ reduction and oxidation (sat, aprA, aprB, dsrA, and dsrB), sulfide oxidation (fccb), CH₄ oxidation (pmoA, pmoB, and pmoC), Fe(II) transport (feoA and feoB), As(III) oxidation (aoxB), and As(III) pumping (arsB) (Fig. 4A and Table S10). Moreover, the co-occurrence/expression network analysis of metabolic modules also implied that active diazotrophs were associated more with other biogeochemical processes than native diazotrophs, consistent with the functional gene abundances described above (Figs. 4B and S8B).

Fig. 4. Metabolic potential and genomic bins related to N₂ fixation and N/S/Fe/CH₄/As cycles in groundwater.

A Pearson correlation analysis between nitrogenase abundances and geochemical parameters/other essential functional genes of N, S, Fe, CH₄, and As cycling identified by the KEGG Orthologs database based on metatranscriptome. B The co-expression network of N₂ fixation and other biogeochemical processes were analyzed by the KEGG module database (Pearson correlations r > |0.6| and p < 0.05). C Taxonomic distribution and metabolic potential of the bacterial medium-quality metagenomic bins (completeness > 50% and contamination < 10%). For the taxonomy of the bins, “c” stands for class level, and “g” stands for genus level.

The metagenomic datasets produced 37 bacterial and two archaeal medium-quality metagenome-assembled genomes (MAGs) with 51.8 – 98.1% completeness, 0 – 8.5% contamination, and 33.4 – 72.5% GC content (Table. S11). Of these genomes, 19 bacterial MAGs spanning nine different phyla (Acidobacteriota, Bacteroidota, Desulfobacterota, Gemmatimonadota, KSB1, Methylomirabilota, Nitrospirota, Omnitrophota, and Pseudomonadota) carried critical genes involved in N₂ fixation (Fig. 4C). A keystone MAG (Thermodesulfovibrionia, bin.16) with a complete nitrogenase also possessed genes associated with denitrification (norBC), organic N to NH₄⁺ (gdhA), polysulfide to sulfide (hydAB), dissimilatory SO₄²⁻ oxidation and reduction (sat, cysCH, and dsrAB), Fe(II) transport (feoA and feoB), and As(V) detoxification (arsBC). In addition, two keystone MAGs (Thermodesulfovibrionia_bin 16 and Rhodothermia_bin 47) implied potential coupling between N₂ fixation and methanogenesis based on the KEGG module annotation (Fig. 4C).

Rates of near in situ N₂ fixation in field groundwater and N₂-fixer enrichment and isolates

N₂ fixation was active in both the oxidizing and reducing areas. For most incubation experiments of field groundwater, both Δδ¹⁵N (‰) and ¹⁵N concentrations of PN + TDN showed good linearity in the incubation time series R² > 0.964 (Fig. S9 and Table S12). N₂ fixation rates varied from 9.5 to 585.4 nmol N L⁻¹ h⁻¹, with a mean value of 165.0 nmol N L⁻¹ h⁻¹ (Fig. 5A). N₂ fixation activity was positively correlated with ORP (p < 0.01), NO₃^– (p < 0.001), TN (p < 0.01), and nifH/16S rRNA (p < 0.001) (Fig. S4B). After 40 days of incubation, nitrogenase activities of groundwater ranged from 7.0 to 69.7 nmol ethylene L⁻¹ d⁻¹ (Fig. S10). N₂-fixer enrichment of groundwater showed high activity, from 17.3 to 1728.4 nmol N g⁻¹ h⁻¹ (Fig. 5B). The highest N₂ fixation activity was found in the G1 N₂-fixer enrichment, in congruence with the field result. Six N₂-fixing strains, designated NF07, NF18, NF21, NF23, E1, and A3, were isolated from groundwater samples. BLAST analysis of the 16S rRNA gene sequences of these strains indicated that four belonged to Klebsiella, and the remaining two belonged to Ensifer and Rhodococcus (Fig. S11). All three of these genera were found in the 16S rRNA gene amplicon sequencing analysis of in situ groundwater and N₂-fixer enrichment (Fig. S12). The acetylene reduction assay showed that the nitrogenase activity rates of the strains ranged from 7.8 to 785.2 μmol L⁻¹ day⁻¹ (Figs. 5C and S13). The four active strains were selected to determine their ability to fix ¹⁵N₂ intracellularly using the ¹⁵N isotope tracer technique. The N₂ fixation rates ranged from 13.2 to 55.6 nmol N g⁻¹ h⁻¹ (Fig. 5B). In addition, ¹⁵N₂-induced shifts in the RR band of cytochrome c are a sensitive indicator of N₂ fixation [58]. Raman spectra of the four strains under ¹⁵N₂ incubation showed that the 1129 cm⁻¹ band (C–N stretch) at the blank shifted clearly to 1114 cm⁻¹ at NF07, NF18, NF21, and NF23, confirming the substitution of ¹⁴N with ¹⁵N in the C–N bond (Fig. 5D). These results demonstrated diazotroph activity in groundwater.

Fig. 5. N₂ fixation rates in field groundwater and N₂-fixer enrichments and isolates.

The verified function of N₂ fixation in the (A) field groundwater and (B) N₂-fixer enrichments and isolates (detected by the ¹⁵N₂ tracing incubation). The activities of N₂-fixer isolates were also analyzed using the (C) acetylene reduction method and (D) single-cell resonance Raman spectroscopy. ¹⁵N₂-induced shifts in the resonance Raman band of Cyt c are a sensitive and robust indicator of N₂ fixation. Single-cell Raman spectra from four isolates (D) indicate that the 1129 cm⁻¹ band (C-N stretch) shifted markedly to 1114 cm⁻¹ under ¹⁵N₂ incubation conditions, implying that it is the substitution of light ¹⁴N with heavier ¹⁵N in the C-N bond resulting in a decrease in the vibrational frequency of the C-N stretch. The pink shading indicates the ¹⁵N₂-induced shifts in the resonance Raman spectra.

Discussion

Diazotrophs inhabit groundwater systems

Direct evidence from our integrated ¹⁵N isotope tracing, acetylene-reduction assays, and RNA-based nifH gene analysis results indicated that many active diazotrophs inhabit groundwater, although other processes, including abiotic processes or biological denitrification and anaerobic ammonia oxidation, could potentially lead to the overestimation of N₂ fixation rates. However, natural groundwater isotopic fractionation reported in previous studies has not approached the isotopic increments measured in this study (Fig. S9 and Table S12) [62–64]. Our data support N₂ fixation in groundwater. In addition, ethylene production rates from acetylene-reduction assays were lower than the N₂ fixation rates determined through the ¹⁵N₂ tracing method. This may be attributed to the direct and/or indirect inhibitory effect of acetylene on N₂ fixation activity [65–70]. We also found that the average N₂ fixation rate was much higher than those reported for oceans and eutrophic estuaries such as the southern Indian Ocean [71], Arctic Ocean [72], North Pacific [73], North Atlantic Ocean [74], Taiwan Strait [75], South China Sea [76], eastern Arabian Sea [77], Kishon River [78], and Cochin Estuary [79], and was comparable to those reported for eutrophic sediments [4, 80, 81]. We hypothesize that excess N₂ [1, 36], diverse electron donors [82], and coupling with other biogeochemical processes [42] in groundwater are conducive to diazotrophic activity. The huge difference in N₂ fixation rates may be due to the higher PN concentrations (0.2–1 µM) in groundwater [30], compared to the open ocean (average, 0.2–0.3 µM) [83, 84]. Correspondingly, the diversity (Shannon index) of diazotrophs in the groundwater was higher than that in soils and marine surface water [85, 86]. Previous studies proposed that micro-hypoxic environments within particles and biofilms could be a suitable niche for diazotrophs under various redox conditions [1, 87, 88]. Thus, the patchy distribution of N₂ fixation rates in oxidizing and reducing conditions may have been caused by the heterogeneity of suspended particles, similar to marine environments [50, 89]. The lower abundance of diazotrophs in groundwater, with N₂ fixation rates similar to those of eutrophic estuarine sediments [10, 80, 81, 90], suggests that diazotrophs may have a more diverse and abundant energy source in groundwater.

A comprehensive set of nifH sequences obtained from amplicons, metagenomes, and metatranscriptomes revealed that groundwater-associated diazotrophs were predominately heterotrophic bacteria (Cluster I, aerobic or facultatively anaerobic) (Figs. 2C and S5) [91]. The richer DOC in groundwater may provide a favorable environment for heterotrophic diazotrophs [30, 44, 92]. Recent studies have found that heterotrophic bacterial diazotrophs were more abundant than cyanobacterial diazotrophs in marine environments, highlighting the importance of heterotrophic populations in N₂ fixation [93, 94], especially for Gammaproteobacteria and Alphaproteobacteria populations [95]. We found that populations in nifH amplicons, contigs, and N₂ fixation enrichments were mainly Gammaproteobacteria and Alphaproteobacteria (Figs. S5, S12). For example, the heterotrophic diazotrophs Klebsiella oxytoca and Ensifer adhaerens cultured from groundwater are Gammaproteobacteria and Alphaproteobacteria. Klebsiella is also an active heterotrophic diazotroph in estuarine environments [96]. Studies have suggested that heterotrophic diazotrophs are widely distributed in seawater through multiple metabolic pathways, such as denitrification and ammonification [95]. The results of our near in situ investigation and laboratory-scale cultures confirmed that active diazotrophs inhibit groundwater systems.

In conjunction with our recent findings, the highest near in situ rates of N₂ fixation and NH₄⁺ consumption were observed simultaneously in the oxidizing area (G1) [30]. Due to the lower rates of mineralization and DNRA [30], we propose that N₂ fixation is an essential source of N in NH₄⁺-depleted groundwater, promoting N transformation [28, 97]. N₂ fixation activity in groundwater samples with 211.2 μM NH₄⁺ implies that diazotrophs fix N₂ for reasons other than to meet their own N demand [98], similar to previous findings in aquatic environments [3, 19, 99] and for heterotrophic isolates [95]. Furthermore, there was evidence that N₂ fixation could balance the diazotrophic intracellular redox state by removing excess electrons and energy, serving a similar purpose to CO₂ fixation [3, 50, 98, 100]. This notion is supported by the positive correlation between diazotrophs and electron donors [Fe(II) and As (III)] found in this study. Thus, N₂ fixation in oxidizing groundwater with low NH₄⁺ concentrations sustains available N for microorganisms. In reducing areas with high NH₄⁺ levels, diazotrophs may regulate ideal redox conditions through N₂ fixation without synthesizing enzymes with expected functions (redox balancing) [95, 98]. In addition to organic N mineralization, DNRA, and surface input [30], future work on the sources of N in groundwater systems should consider the influence of N₂ fixation. Continued spatiotemporal variability studies are needed to comprehensively understand the ecological functions of N₂ fixation in aquifers.

Environmental implications of N₂ fixation in groundwater systems

Due to the varied hydrogeochemical conditions, diverse microbes could catalyze C, N, S, As, and Fe-related biogeochemical processes in groundwater [32]. Thus, N₂ fixation might be coupled with various catabolic processes, contributing to N flow in the groundwater microbiome [101]. Our study showed that N₂ fixation was closely linked to nitrification, denitrification, SO₄²⁻ reduction, sulfide oxidation, CH₄ oxidation, As(III) oxidation and pumping, and Fe(II) transport in a typical groundwater system.

N₂ fixation and nitrification are the opposite of NH₄⁺ transformation. In this study, N₂ fixation was positively associated with nitrification and comammox based on metatranscriptomic analysis (Fig. 4A, B), congruent with the observation that nitrification potentially counteracted the inhibition of NH₄⁺ on diazotrophs, sustaining N₂ fixation activity [99, 102]. This finding supported our recent finding that the highest nitrification rates were present in NH₄⁺-deficient groundwater samples (G1 and G2) [30] with more active N₂ fixation (Fig. 5A). In addition, oxygen consumption by nitrifiers supports anaerobic microbial metabolism [1], which may be a survival pathway for diazotrophs. Aerobic methanotrophs may also contribute to oxygen depletion. Some aerobic methanotrophs also fix N₂. For example, genus Methylocella in the nifH amplicons in our study was found with active CH₄ consumers widespread in terrestrial environments [103, 104]. We also recovered a medium-quality MAG affiliated with Gemmatimonadota that was predicted to fix N₂ and oxidize CH₄ in the G27 groundwater sample (reducing aquifer); Gemmatimonadota has been identified as a fourth phylum potentially capable of aerobic methanotrophy [105]. Besides methanotrophs, diazotrophs found in the current study, such as Rhizobium and Bradyrhizobium, can utilize the methanol produced by methanotrophs [106]. Together, these findings may explain the correlation between N₂ fixation and CH₄ oxidation in groundwater (Fig. 4A, B) [106].

The positive influence of organic matter (OM) on N₂ fixation was probably due to the supplementation of energy for heterotrophic diazotrophs [107]. We also found a significant influence of DOC on diazotrophic communities. In aphotic conditions, the addition of DOM such as free amino acids, glucose, and ATP stimulated N₂ fixation in enrichment experiments [50, 89]. In addition to OM metabolism, other biological processes such as SO₄²⁻ reduction [2] and S oxidation [23], may also drive N₂ fixation in groundwater. Putative sulfate-reducing bacteria, including Desulfomicrobium, Desulfuromonas, and Pseudomonas [14, 108], were identified as active diazotrophs in the present study. ATP production by SO₄²⁻ reduction could partially support N₂ fixation [20, 109]. In addition, diazotrophs have been found to use S as a terminal electron acceptor/donor [18, 23]. These factors account for the positive correlation between N₂ fixation and SO₄²⁻ reduction and sulfide oxidation in our results and similar previous findings in benthic and estuarine environments [21]. A recent study found that As-dependent N₂ fixation was also widespread in high-As tailings and soils [101]. In our study, diazotrophs were positively correlated with As(III)-oxidizing and As(III)-pumping microbes (aoxB and arsB genes) in groundwater (Fig. 4) characterized by high As concentrations [47, 110]. In addition, members of Hydrogenophaga, an active diazotroph found in this study (Fig. 2D), may participate in As(III) oxidation [111], further refining the coupling between As and N cycling in high-As groundwater [32].

In this study, N₂ fixation was closely related to dissimilatory Fe(III) reduction. We found a positive correlation between N₂ fixers (abundance and community composition) and Fe(II) concentrations or Fe(II) transport genes (Figs. 1B, 3, 4A, and S6). Some diazotrophs identified in this study have been found with Fe(III) reduction, including Azotobacter, Bacillus, Desulfuromonas, Pseudomonas, and Klebsiella [112–114]. One MAG (Thermodesulfovibrionia, bin 16) found by metagenomic binning may be capable of Fe, S, and As-dependent N₂ fixation according to the functional gene or module analysis (Fig. 4C), suggesting that diazotrophs in groundwater also use Fe(III)/ SO₄²⁻/As(III) as electron acceptors [115]. Thermodesulfovibrionia was also found in nifH gene contigs in the metagenome and metatranscriptome (Fig. S5). Thermodesulfovibrionia has been reported to dominate reducing NH₄⁺-rich groundwater and may be involved in N turnover and Fe(III) reduction [42, 116]. During organic matter decomposition, Fe(III) is usually reduced before sulfate, yielding 228 kJ/mol energy per reaction [117, 118]. Relationships between N₂ fixation rates and Fe(III) reduction and Fe(II) concentrations have also been found in pure cultures (from our unpublished data), as well as in coastal sediments, soils, and wetlands [10, 22, 81], perhaps because Fe(III)-reducing bacteria can fix N₂ directly and provide energy for N₂ fixation using H₂ or organic C as electron donors [20, 117].

Overall, our results indicate that diazotrophs were active in groundwater and may interact positively with other microorganisms. N₂ fixation may be coupled with other elemental cycles in groundwater both among microbial species and as an intra-species process. Further studies of the correlated mechanisms of N₂ fixation and C/N/S/As/Fe metabolism in groundwater are warranted. The contribution of N₂ fixation to groundwater systems globally should also be explored. It should be possible to culture novel diazotrophs and nitrogenases from groundwater.

Conclusion

Our study demonstrates the activity, abundance, and distribution of diazotrophs in groundwater, indicating that N₂ fixation was sufficiently active and abundant to affect biogeochemical cycling in groundwater systems. The majority of nifH communities in amplicons, metagenomes, and metatranscriptomes were aerobic or facultative anaerobic diazotrophs, supporting the expression of nifH genes throughout varied redox groundwater conditions. Moreover, a tight correlation was found between N₂ fixation and various biogeochemical cycles, such as nitrification, SO₄²⁻ reduction, sulfide oxidation, Fe(II) transport, CH₄ oxidation, As(III) oxidation, and As-resistant. Here, we identified groundwater systems as neglected habitats for diazotrophs and provided insight into N₂ fixation with distinct geochemical factors and N/CH₄/S/As/Fe cyclings. Further understanding of the synergistic mechanisms of N₂ fixation and other biogeochemical processes is necessary.

Author contributions

Conceptualization: PL, XL, and HW. Sample collection and methodology: HW, XL, L-LH, PL, YW, and ZJ. Analysis and writing-original draft preparation: XL and HW. Writing-review and editing and funding acquisition: PL, LC, and S-JK. ¹⁵N analysis: L-LH and S-JK. Single-cell resonance Raman-¹⁵N₂ SIP analysis: KY and LC.

Data availability

The raw data have been submitted to the NCBI database under BioProject numbers PRJNA937386, PRJNA996345, PRJNA996367, PRJNA882225, and PRJNA884812.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41396-023-01513-x.