Ferrihydrite-mediated methanotrophic nitrogen fixation in paddy soil under hypoxia

Abstract Biological nitrogen fixation (BNF) by methanotrophic bacteria has been shown to play an important role in maintaining fertility. However, this process is still limited to aerobic methane oxidation with sufficient oxygen. It has remained unknown whether and how methanotrophic BNF proceeds in hypoxic environments. Herein, we incubated paddy soils with a ferrihydrite-containing mineral salt medium to enrich methanotrophic bacteria in the presence of methane (20%, v/v) under oxygen constraints (0.27%, v/v). The resulting microcosms showed that ferrihydrite-dependent aerobic methane oxidation significantly contributed (81%) to total BNF, increasing the 15N fixation rate by 13-fold from 0.02 to 0.28 μmol 15N2 (g dry weight soil) -1 d−1. BNF was reduced by 97% when ferrihydrite was omitted, demonstrating the involvement of ferrihydrite in methanotrophic BNF. DNA stable-isotope probing indicated that Methylocystis, Methylophilaceae, and Methylomicrobium were the dominant methanotrophs/methylotrophs that assimilated labeled isotopes (13C or 15N) into biomass. Metagenomic binning combined with electrochemical analysis suggested that Methylocystis and Methylophilaceae had the potential to perform methane-induced BNF and likely utilized riboflavin and c-type cytochromes as electron carriers for ferrihydrite reduction. It was concluded that ferrihydrite mediated methanotrophic BNF by methanotrophs/methylotrophs solely or in conjunction with iron-reducing bacteria. Overall, this study revealed a previously overlooked yet pronounced coupling of iron-dependent aerobic methane oxidation to BNF and improves our understanding of methanotrophic BNF in hypoxic zones.


Introduction
Industrial nitrogen fertilizer has widely been applied to enhance the nitrogen-limited primary productivity and pursue higher crop yields.However, excessive usage of nitrogen fertilizer and the low utilization efficiency inevitably give rise to a range of environmental hazards, including soil acidification, groundwater nitrate pollution, greenhouse gas (N 2 O) emission, and water eutrophication [1][2][3][4].Apart from industrial nitrogen fertilizer, biological nitrogen fixation (BNF) also plays a non-negligible role in maintaining soil fertility and was estimated to provide >45 kg N ha −1 for unfertilized paddy soil annually [5].Better insight into the mechanisms underlying BNF is of great significance when aiming to augment the soil BNF activity and reducing the usage of industrial nitrogen fertilizer and the environmental nitrogen burden.
The carbon and energy source of BNF in paddy soil generally come from soil organic matter or root exudates.The bioavailability of carbon and energy source often becomes a major limiting factor for BNF [6].Methane is another potential, but often overlooked carbon and energy source for BNF in paddy soil, with methane being released in a large amount to the atmosphere via the diffusion, ebullition, or rice aerenchyma transport [7].Methanotrophic BNF is not only able to function as a biofilter for the greenhouse gas methane but also supplies a large amount of nitrogen for plants.Boosting the methanotrophic BNF efficiency is beneficial for both agricultural sustainability and global climate balance.It was estimated that methanotrophic BNF is able to increase the N reserve by 30 kg N ha −1 over one rice growing season and contribute to 30% of the total nitrogen accumulation (i.e. 100 kg N ha −1 ) [8].Methanotrophic BNF has also been shown to account for 65% of total nitrogen fixation in rice root tissues and >33% of the new nitrogen accumulation in younger peatlands [9,10].These data suggest a considerable nitrogen input via methanotrophic BNF.
Aerobic methanotrophs (namely, methane-oxidizing bacteria (MOB)) are the most common drivers of methanotrophic BNF.MOB has been found to inhabit diverse ecological niches such as the rice root, the rhizosphere, and bulk soil [11,12].The abundance of MOB in paddy soil was shown to follow the order: root > rhizosphere > bulk soil [13].Such a distribution characteristic is likely due to a limited oxygen secretion and diffusion from rice roots because dissolved oxygen (DO) in the rhizosphere would gradually decrease to zero beyond a 5-mm distance from the root surface [14].Oxygen deficiency is presumed to be as a key constraint on methanotrophic BNF in the bulk soil, as aerobic methane oxidation (AMO) depends on oxygen.Unfortunately, most previous researches on methanotrophic BNF by MOB were focused on soil-free settings with a plenty of oxygen (>2%, v/v) in the headspace [15][16][17][18], whereas methanotrophic BNF by soil MOB suffering from strong oxygen constraint has as yet been unobserved.The widespread distribution of MOB in bulk soils necessitates a further exploration on whether MOB are able to perform methanotrophic BNF under hypoxia.
Recent studies have suggested that MOB were able to perform methane oxidation under hypoxia by shifting from oxygendependent respiration to Fe(III)-dependent respiration (i.e.AMO coupled to Fe(III) reduction, Fe-AMO) [19,20].A few of the MOB, including Methylosinus, Methylomonas, and Methylococcus, have been reported to catalyze Fe-AMO [21][22][23][24].The Fe-AMO process is ecologically significant as it has been suggested to predominate the methane oxidation process in iron-rich lake sediments, mitigating 40.3% of the methane emission [20].Yet, Fe-AMO is still not correlated with methanotrophic BNF, because it was characterized under ammonium-sufficient conditions where BNF seemed unlikely to have occurred [20,21].Since bioavailable nitrogen is quite low in natural niches, the Fe-AMO activity evaluation under a nitrogen-deficient condition is more environmentally relevant.For the Fe-AMO process, ferrihydrite is generally used as it is the most reactive among natural iron minerals [25].Ferrihydrite is relatively rich in the rhizosphere due to a rapid oxidation of Fe(II) [26].The oxidation of Fe(II) often forms a layer of iron plaque (IP) that mainly (80%) consists of ferrihydrite on the rice root surface [27], providing a natural condition for rhizosphere Fe-AMO.Such a widespread coexistence of abundant ferrihydrite, methane, and MOB in the rhizosphere inspired our speculation that methanotrophic BNF may extend from the root tissues to the rhizosphere/bulk soil via ferrihydrite-mediated Fe-AMO.Such a ferrihydrite-dependent methanotrophic BNF may dissolve the IPs on the root surface, accelerating soil iron mineral transformation/transport and root nutrient uptake (e.g.nitrogen and phosphorus).The resulting Fe(II) is likely to serve as a reductant for heavy metal detoxification (such as Cr(VI)) and for the degradation of organic contaminants via the Fenton reaction.
Therefore, this work aimed to explore the possibility, performance, and mechanism of ferrihydrite-mediated methanotrophic BNF under hypoxia in the paddy soil microcosm.Methane consumption, ammonia, and Fe(II) production were monitored to verify the dependence of methanotrophic BNF on Fe-AMO.Carbon ( 13 C) and nitrogen ( 15 N) isotope tracing were employed to evaluate the performance of such a novel BNF process.Nitrogen-fixing and/or iron-reducing MOB were identified via DNA stable-isotope probing (DNA-SIP) and sequencing.The metabolic features associated with BNF and ferrihydrite reduction and their coupling mechanism were analyzed by metagenomics.We were able to show that Fe-AMO enhanced methanotrophic BNF rate by 13-fold, contributing significantly (81%) to total BNF.Our findings suggest MOB play an appreciable role in linking methane, nitrogen, and iron cycling in hypoxic soils.

Soil sampling and physicochemical properties
A surface layer (top 20 cm) of bulk paddy soils was collected in October 2021 from a research farm (32 • 30 N, 119 • 25 E) of Yangzhou University, Jiangsu province, China, after the rice harvest [28].The meteorological data for the farm were the same as previously described [28].To get a sample representative of the typical paddy soil, five bulk soil subsamples in the paddy field were collected along an S-line track and were mixed to obtain a composite sample.The composite sample was transferred to the laboratory within 24 h and was divided into two parts.One part was air-dried and sieved through a 1-mm mesh, and soil physicochemical properties were determined according to standard methods [29].Brief ly, total carbon and total nitrogen were determined with an element analyzer (Vario EL Cube, Elementar, Germany); dissolved organic carbon (DOC) was determined on a total organic carbon analyzer (TOC-L CPH, Shimadzu, Japan); ammonia nitrogen was extracted with 3 M KCl for 1 h and was analyzed by the indophenol blue colorimetric method at 625 nm; total phosphorus was extracted by fusing the soil with sodium hydroxide and was quantified with the molybdenum blue method at 700 nm; total iron was extracted via digesting the soil with nitric acid-perchloric acid-hydrof luoric acid and was measured by an inductively coupled plasma optical emission spectroscopy (Avio 220 Max, PerkinElmer, USA); Fe(III) and Fe(II) were extracted with 0.5 M HCl and the supernatant/hydroxylamine hydrochloride reduced supernatant was determined by the o-phenanthroline colorimetric method at 510 nm; Fe(III) was calculated as the difference between total iron and Fe(II).Another part was preincubated at a 1:3 soil-to-ultrapure-water ratio for 1 week at room temperature to decompose organic carbon and nitrogen.The physicochemical properties of original soil and preincubated soil are summarized in Table 1.

Methane-oxidizing bacteria enrichment and microcosm for ferrihydrite-mediated methanotrophic biological nitrogen fixation
The MOB enrichment was carried out for 160 days in a serum bottle reactor (total volume, 2.45 l; headspace, 1.22 l) (Fig. S1).Approximately, 330 g (dry weight) preincubated soil and 1 l nitrogen-free mineral salt solution (MSS, Table S1) were mixed in the reactors.Ferrihydrite was prepared using a previously described method and was amended to the reactors at an initial concentration of 10 mmol iron l −1 [30].The reactors were subsequently purged with N 2 (99.99%), aerated with CH 4 /N 2 (20%:80%), and incubated statically in the dark at 30 To demonstrate ferrihydrite-mediated methanotrophic BNF, the experimental treatment group (designated as CFeN) and four controls (FeN, CN, CFeNS 0 and N) were set up in 60-ml serum bottle reactors (Table S2).The control groups were used to investigate the roles of CH 4 , ferrihydrite, MOB, and soil organics, respectively.The CFeN group was prepared by mixing 4 ml soil slurry (∼1 g dry soil) from the parent reactor with autoclaved 5 ml MSS and 1 ml ferrihydrite (300 mmol l −1 in MSS).After the reactor was alternately filled with helium and was evacuated to make a hypoxic headspace (O 2 , 0.27% (v/v)), 10 ml CH 4 and 40 ml N 2 (99.99%) were injected into the headspace of evacuated reactors.The FeN group was prepared by injecting 50 ml N 2 into a methanefree headspace as other conditions were identical with the CFeN group.The CN group was conducted by omitting ferrihydrite and supplementing with 10 ml MSS.The N group was prepared by incubating the same amount of soil slurry with 10 ml MSS under a headspace of N 2 (50 ml) and omitting both ferrihydrite and CH 4 .A soil-free control (CFeNS 0 ) was also performed identically as the CFeN group except that no soil was added.All the microcosms were run in triplicate and were incubated for 50 days at 30 • C in the dark.Another batch of microcosms was conducted to confirm BNF by substituting 15 N 2 (purity of 99.99%) for 14 N 2 in CFeN, FeN, CN, and N.

Methane oxidation inhibition and Fe(III)-limitation experiments
To evaluate the dependency of nitrogen fixation on Fe-AMO, the CFeN group with dif luoromethane (CFeN + CF 2 H 2 ) and the CFeN group without ferrihydrite (namely, CN) were conducted to represent AMO inhibition and Fe(III) limitation, respectively.For details, 4 ml slurry (∼1 g dry soil) from the parent reactor on Day 160 was mixed with 5 ml N-free MSS and 1 ml ferrihydrite (300 mmol l −1 ) in a 60-ml reactor.The reactor was then alternately evacuated, filled with helium gas, and eventually filled with 10 ml CH 4 and 40 ml N 2 .CH 2 F 2 , an inhibitor of methane monooxygenase [31], was added to the headspace of reactors at a final concentration of 0.5% (v/v) to achieve an inhibitory effect.The 1 ml N-free MSS was substituted for ferrihydrite to make Fe(III) limitation.There were 18 reactors for each of the three groups (CFeN, CFeN + CH 2 F 2 , and CN).Every 10 days, three reactors from each group were randomly selected to ensure the representativeness and offered to withdraw the soil slurries for Fe(II) and NH 4 + -N analyses.Dual isotope labeling experiments were also conducted in another parallel batch of CFeN, AMO-inhibited, and Fe(III)-limited groups with 13 CH 4 (10 ml) and 15 N 2 (40 ml) in the headspace.The soil 13 C and 15 N abundances were determined at the end of the incubation.
The procedures for RNA extraction, cDNA synthesis by reverse transcription, and real-time quantitative PCR (qPCR) are provided in the Supplementary Information.

DNA stable-isotope probing and real-time quantitative PCR
To identify the microorganisms involved in Fe(III)-dependent methane oxidation and nitrogen fixation and dual-isotope labeled ( 13 CFe 15 N) and unlabeled ( 12 CFe 14 N) treatments were performed using 13 CH 4 (99.99 atom%) + 15 N 2 (99.99 atom%) and 12 CH 4 + 14 N 2 , respectively.For DNA-SIP, the batch conditions (10 ml methane and 40 ml nitrogen) were the same as mentioned above.All the reactors were subjected to a 50-day incubation at 30 • C in the dark.The soil slurries (∼10 ml) were then precipitated by centrifugation (8000 rpm, 15 min) to remove the supernatant.Total soil DNA was extracted by FastDNA Spin Kit for Soil (MP Biomedicals, USA) following the manufacturer's instructions.The extracted DNA (5 μg) from the 13 CFe 15 N and 12 CFe 14 N groups were dissolved in CsCl solution to reach a final buoyant density (BD) of 1.714 g ml −1 in 4.9-ml OptiSeal polyallomer tubes, respectively.The samples were centrifugated at 53 000 rpm for 48 h at 20 • C using an Optima XPN-100 Ultracentrifuge (Beckman Coulter, USA).After centrifugation, a syringe pump (LSP01-1A, Baoding Longer Precision Pump CO., Ltd, China) was applied to fractionate the obtained DNA gradients into 24 equal volumes (200 μl per fraction).The BD of each fraction was detected using a digital refractometer (AR200, Reichert, USA).Then, DNA in each fraction was precipitated with glycogen and two volumes of PEG (i.e.400 μl), purified with 70% ethanol, and eluted with 30 μl of TE buffer.The pmoA, mcrA, and nifH and the bacterial and archaeal 16S rRNA genes in each fraction were quantified via a real-time PCR system (LightCycler 96, Roche, USA) using the primer pairs A189F and mb661R, mlas-mod-F and mcrA-rev-R, polF and polR, Bac515F and Bac806R, and Arch519F and Arch915R, respectively.The qPCR primer sequences, reaction mixtures, and annealing temperatures are detailed in the Supplementary Information (Table S3).

Amplicon sequencing and shotgun metagenome sequencing
DNA sample in fractions with the top three highest nifH and pmoA gene abundances from the treatments 13 CFe 15 N (heavy fractions) and 12 CFe 14 N (light fractions) were subjected to amplicon sequencing of bacterial 16S rRNA and nifH genes with the primers listed in Table S3.The sequencing was performed on an Illumina MiSeq platform at Beijing Novogene Bioinformatics Technology Co., Ltd, China.The raw reads were processed with QIIME2 for trimming and quality filtering.After being denoised and filtered at an abundance threshold of five reads, amplicon sequence variants (ASVs) were generated with the DADA2 algorithm.The representative sequence of each ASV was annotated against the Genome Taxonomy Database (GTDB).Shotgun metagenome sequencing was performed on an Illumina PE150 platform at Novogene Bioinformatics Technology Co., Ltd, China.Since a single heavy fraction of DNA was insufficient for metagenome sequencing, all the heavy fractions of DNA from the 13 CFe 15 N group were pooled as one composite DNA sample.The raw sequence reads were qualified using Trimmomatic v0.36 to remove low-quality reads and hostgenome contamination.A total of 147 685 624 (22.00 Gb) qualified reads were then de novo assembled using Megahit.Metagenomic binning was performed using MaxBIN2 and Metabat2.CheckM was used to estimate the completeness, contamination, and strain heterogeneity of the bins.Metagenome-assembled genomes (MAGs) were filtered at a threshold of completeness >80% and contamination <10%.Taxonomy annotation of the genomes and phylogenetic tree analysis were carried out using GTDB-Tk (r207) based on the GTDB.Gene annotation was performed with KOBAS 2.0 against the KEGG database with an E-value cutoff of 10 −5 .Functional genes involved in methane oxidation, BNF, and extracellular electron transfer (EET) for Fe(III) reduction were identified according to the method of Li et al. [20].

Analytic techniques
CH 4 in the headspace of reactors were determined by gas chromatography (GC) equipped with a f lame ionization detector (GC-2014, Shimadzu, Japan).The initial headspace oxygen (0.27% (v/v)) was measured by GC (7890B, Agilent, USA) with an electron capture detector.DO was determined by a microelectrode using the Sensor Trace Pro software (Unisense A/S).Fe(II) in the slurry was extracted with excess 0.5 M HCl and was measured by the o-phenanthroline colorimetric method at 510 mm [32].Ammonium was extracted with 3 M KCl for 1 h and was measured by the indophenol blue colorimetric method at 625 nm [33].The 15 N and 13 C abundances in the freeze-dried soil were measured by a stable-isotope ratio mass spectrometer (Isoprime-100, Elementar).The BNF rates were calculated using the formula (SW × TN/100 × ( 15 N c1 -15 N c2 )/ 15 N g × 100/MW/t) as previously described [17], where SW is the dried soil weight (1.00 g/reactor), TN is the average nitrogen content (%, w/w), and MW is the molecular weight (30) of 15 N 2 .The 15 N c1 and 15 N c2 represent the respective final and initial 15 N concentrations (atom% excess) in the soils, respectively.The 15 N g is the 15 N concentration (99.62 atom% excess) in the 15 N 2 gas, and t is the incubation time.The least significant difference test for geochemical parameters and microbial relative abundances in the DNA fractions were performed using the R software (version 4.3.1)package agricolae (v1.[3][4][5].

Evidence for ferrihydrite-mediated methanotrophic biological nitrogen fixation in soil microcosm
Three successive cycles of batch incubation with methane and ferrihydrite were conducted to enrich iron-reducing methanotrophs (i.e.MOB) under hypoxia (DO, ∼0.1 mg l −1 ).During the enrichment, methane in the headspace was gradually consumed and Fe(II) increased from 18.68 ± 1.24 mmol l −1 on Day 0 to 28.95 ± 0.46 mmol l −1 on Day 50 (Fig. S1).The net Fe(II) addition (12.63 mmol) and methane consumption (2.77 mmol) gave a stoichiometric ratio of 4.56:1.This deviated from a theoretical ratio (Fe(III)/CH 4 = 8:1) under hypoxia [20,21], suggesting the possibility of other electron sinks (e.g.BNF).Moreover, the simultaneous enrichment of Fe(II)-oxidizing bacteria (see the DNA-SIP results below) could be also responsible for this deviation.Both methane consumption and Fe(II) generation were well reproduced in the next two batch cycles.Total NH 4 + -N accumulated to 0.56 ± 0.07 mmol l −1 during the enrichment period, exhibiting a similar trend like Fe(II) (Fig. S1).Moreover, the methane, Fe(II), and ammonium concentrations showed strong linear correlations between each other (R 2 = 0.85-0.99,P < .01)(Fig. S2).These correlations suggested that methane consumption, ammonium production, and ferrihydrite reduction occurred in a synergetic way.
To further investigate the occurrence of ferrihydrite-mediated methanotrophic BNF, the experimental group (CFeN) was compared with four controls (FeN, CN, CFeNS 0 , N) to assess the roles of methane, ferrihydrite, MOB, and the background iron/organics, respectively.The methane consumption for CFeN was significantly higher than those for CN and CFeNS 0 (Table 2), demonstrating ferrihydrite-enhanced methane utilization by MOB.Meanwhile, more Fe(II) and ammonium were produced in CFeN than in FeN, CN, and N groups (P < .05)(Table 2).The generated Fe(II) strongly correlated with the accumulated ammonium in the five groups (R 2 = 0.94, P < .001)(Fig. S2), indicating that ammonification or BNF was coupled to ferrihydrite reduction [34]. 15N 2 isotope tracing and the soil 15 N abundance ( 15 N/( 15 N + 14 N)) measurement were conducted to verify BNF.CFe 15 N exhibited a much higher soil 15 N abundance (0.565 ± 0.018 atom%) than those for Fe 15 N, C 15 N and 15 N (Table 2).By contrast, the CFe 14 N group showed no increase in soil 15 N abundance (0.369 ± 0.000 atom%), ruling out the possibility of nitrogen isotope fractionation [35].Hence, the enhanced 15 N 2 assimilation by the presence of both methane and ferrihydrite confirmed the occurrence of Fe(III)-mediated methanotrophic BNF.The slight increase (0.034 atom%) in the 15 N abundance in Fe 15 N might be driven by the degradation of soil organics [36,37].Compared to CFeN, the high Fe(II) and ammonium production with a low BNF activity in Fe 15 N suggested that Fe(III) reduction was preferentially coupled to ammonification of nitrogen-containing organics rather than BNF in this group.These results indicated BNF in CFeN was primarily attributed to methane and to a lesser extent to soil organics.It was therefore inferred that methane-fueled ferrihydrite reduction played a predominant role in BNF.

Dependence of biological nitrogen fixation on methane oxidation and Fe(III) reduction
To clearly demonstrate the dependence of BNF on methane oxidation and Fe(III) reduction, AMO-inhibited and Fe(III)-limited experiments were performed by adding CF 2 H 2 and omitting Table 2. Net methane, Fe(II) and ammonium changes as well as the soil 15 N abundances and BNF rates for different treatment groups after a 50-day incubation.

Group Methane change (μmol)
Ferrous ion increment (μmol) Ammonium production (μmol) 15 N abundance in soils (atom%)  ferrihydrite, respectively.As shown in Fig. 1A, the methane concentration in the CFeN group decreased significantly over time.By contrast, methane was hardly consumed in the CFeN+CF 2 H 2 group, confirming that AMO was responsible for methane consumption.Meanwhile, more Fe(II) and ammonium (18.69 and 0.18 mmol l −1 , respectively) accumulated in CFeN compared to those in CFeN+CF 2 H 2 (14.07 and 0.13 mmol l −1 , respectively) (Fig. 1B and C), verifying that AMO had led to Fe(III) reduction and ammonium production.The small Fe(II) increment (2.77 mmol l −1 ) in CFeN + CF 2 H 2 indicated that nonmethanotrophic processes (e.g.degradation of soil organics) contributed only a minor part to Fe(III) reduction.Isotope tracing with 13 CH 4 and 15 N 2 showed that the soil 13 C abundance ( 13 C/( 13 C + 12 C), 1.340 ± 0.008 atom%) for CFeN was higher than the natural background value (1.108 atom%) (Fig. 1D), suggesting the 13 CH 4 assimilation by soil microorganisms.Contrarily, no 13 CH 4 assimilation was detected for CFeN + CF 2 H 2 .Likewise, the soil 15 N abundance for CFeN (0.542 ± 0.045 atom%) was significantly higher than that (0.410 ± 0.004 atom%) of CFeN + CF 2 H 2 (P < .05)(Fig. 1D).Thus, the amendment of methane and ferrihydrite enhanced the BNF rate by 13-fold from 0.02 to 0.28 μmol N 2 g dw −1 d −1 (soil dry weight), which was close to that of moss-associated MOB (Table 3).By contrast, the BNF rate for the CFe 15 N + CF 2 H 2 group was only 0.05 μmol N 2 g dw This indicated that ferrihydrite-mediated methanotrophic BNF (0.24 μmol N 2 g dw −1 d −1 ) accounted for 81.1% of the total BNF, whereas the remainder (18.9%) was attributable to nonmethanotrophic BNF.The Fe-AMO based methanotrophic BNF rate was 48.6% of that (0.49 μmol N 2 g dw −1 d −1 ) for the rice root (Table 3) [10], likely due to the greater oxygen constraint on AMO in the soil.Based on these results, it was concluded that AMO rather than soil organics was mainly responsible for Fe(III) reduction and BNF.
After a 50-day incubation, 0.31 mmol l −1 of methane was consumed with simultaneous increases of 2.75 mmol l −1 of Fe(II) and 0.08 mmol l −1 of ammonium for CN, which was lower than those for CFeN (Fig. 1).This indicated that the endogenous Fe(III) in CN (4.10 mmol l −1 ) was not enough or available (e.g.iron oxide surface enveloped by soil components) to support an appreciable methane utilization, namely a Fe(III) limitation effect was achieved.At the RNA level, the CFeN group showed higher pmoA and nifH transcript abundances compared with CN (Fig. 2), agreeing with previous findings that Fe(III) fertilizer increased the potential activity of methanotrophs in rice paddies and that methane-fueled Fe(III) reduction promoted the expression of nitrogen-fixing enzymes in Methanosarcina barkeri [38,39].The enhanced gene transcription could be due to the reasons that (i) ferrihydrite provided the electron acceptor for Fe-AMO and (ii) ferrihydrite likely induced the oxygen generation by methanobactins to alleviate the oxygen constraint [40].The isotope tracing for the 13 C 15 N group showed the soil 13 C abundance increased from a background value (1.108 atom%) to 1.198 atom%, whereas the soil 15 N abundance increment (0.007 atom%) was negligible (Fig. 1D).The corresponding BNF rate for 13 C 15 N (0.01 μmol N 2 g dw −1 d −1 ) was only 3.5% of that for 13 CFe 15 N (∼0.28 μmol N 2 g dw −1 d −1 ) (Table 2).In other words, ferrihydritemediated BNF accounted for 96.5% of the total BNF in 13 CFe 15 N, confirming the dependence of BNF on ferrihydrite reduction.Taken together, these findings demonstrated for the first time that BNF was driven by Fe-AMO.

Identification of functional microorganisms through DNA stable-isotope probing
DNA-SIP was used to reveal the microorganisms which had assimilated the isotope 13 C and/or 15 N.The DNA samples from microcosms fed with 13 CH 4 + 15 N 2 and 12 CH 4 + 14 N 2 (namely  13 CFe 15 N-DNA and 12 CFe 14 N-DNA, respectively) were fractionated by CsCl gradient ultracentrifugation.The bacterial and archaeal 16S rRNA, pmoA, nifH, and mcrA genes in the fractions were quantified via qPCR.Compared with the 12 CFe 14 N-DNA fractions, the 13 CFe 15 N-DNA fractions showed a higher relative abundance of bacterial 16S rRNA genes at the heavy BD of 1.72 g ml −1 (Fig. 3A).This implied that certain bacteria had assimilated the isotope ( 13 C and/or 15 N).For the archaeal 16S rRNA and mcrA genes, 13 CFe 15 N-DNA and 12 CFe 14 N-DNA were distributed in a similar pattern across the entire BD with single peaks, which overlapped well around the light BD of 1.69 g ml −1 (Fig. S3).Thus, archaea were not active in isotope uptake.However, pmoA and nifH in 13 CFe 15 N-DNA fractions exhibited a major peak at the heavy BD of 1.72 g ml −1 , whereas these genes in 12 CFe 14 N-DNA fractions peaked at the light BD of 1.69 g ml −1 .Such a peak shift toward the heavy BD indicated that methanotrophs and diazotrophs were involved in isotope ( 13 C and/or 15 N) incorporation [41].Similarly, Geobacter had a higher abundance at the BD of 1.70 g ml −1 in the heavy fractions ( 13 CFe 15 N-DNA) than in the light fractions ( 12 CFe 14 N-DNA) (Fig. S3C), demonstrating that Geobacter had assimilated the isotope 13 C and/or 15 N, namely, Geobacter participated in iron reduction and/or nitrogen fixation using soil organic matter or 13 CH 4 -derived products as the substrate.This may account for the observed nonmethanotrophic BNF activity in the group CFeN+CF 2 H 2 .The heavy 13 CFe 15 N-DNA and light 12 CFe 14 N-DNA fractions containing the top three highest nifH abundances were subjected to 16S rRNA and nifH sequencing.Methylocystis (30.18% ± 2.09%) Figure 3. Normalized relative abundances of target genes in the gradient ultracentrifugation fractions of heavy 13 CFe 15 N-DNA and light 12 CFe 14 N-DNA (A) and microbial community compositions in the 13 CFe 15 N and 12 CFe 14 N groups based on 16S rRNA (B) and nifH (C) sequencing.dominated in the heavy 13 CFe 15 N-DNA fractions and was significantly more abundant than in light fractions (0.32% ± 0.04%) ( Fig. 3B).Methylocystis is a facultative Type II methanotroph and has a demonstrated capacity of methanotrophic BNF [42][43][44].Methylocystis was therefore considered to be a key diazotrophic methanotroph in our microcosms.Due to the enrichment in heavy fractions, unclassified Methylophilaceae (17.16% ± 3.55%) and Methylomicrobium (6.08% ± 3.57%) were also identified as putative methylotroph/methanotroph and/or diazotrophs.The family Methylophilaceae is associated with methylotrophs unable to grow on methane [45].The enrichment of Methylophilaceae in the heavy fractions suggested that it probably incorporated 13 CH 4 derived intermediates (e.g.methanol) and/or 15 N 2 .Some Fe(II)-oxidizing bacteria were also present in the light/heavy fractions, including Pseudomonas (4.10%/0.34%),Thiobacillus (2.32%/0.20%),Bradyrhizobium (0.10%/0.15%), and Hyphomicrobium [46][47][48][49].Moreover, diazotrophic Hyphomicrobium was enriched in heavy fractions (0.69% ± 0.16%) than in light ones (0.20% ± 0.01%).These suggested the produced Fe(II) from Fe-AMO could be reoxidized by these Fe(II)-oxidizing bacteria for 13 C/ 15 N incorporation, accounting for a lower stoichiometric ratio for Fe-AMO.

Metabolic potential of putative diazotrophic methanotrophs
Metagenomic sequencing and binning were conducted to predict the metabolic potential of Fe(III) reduction and BNF by methanotrophs.A total of 23 bins were gained and classified as "Bacteria."Fortunately, six high-quality MAGs (completeness >80% and contamination <10%) were recovered (Table S4).Among them, two MAGs associated with methane/methanol oxidation were identified as Methylophilaceae (Bin 5) and Methylocystis (Bin 6).The remaining four MAGs were classified as Methylotenera, Thermoanaerobaculi, Myxococcales, and Anaeromyxobacter.A genomewide phylogenetic tree was constructed to reveal the taxonomic positions of MAGs (Fig. 4A).The Methylophilaceae MAG (Bin 5, 88.1% completeness and 1.2% contamination) was phylogenetically clustered to Methylobacillus sp.MM3 with an orthologous average nucleotide identity (ANI) of 81.3% between them (Fig. 4A).For the Methylocystis MAG (Bin 6, 93.1% completeness and 1.7% contamination), the phylogenetically closest relative was Methylocystis echinoides with an ANI of 84.0%.This was lower than the threshold ANI (95%) demarcating different species [ 53], suggesting that it is possibly a novel Methylocystis sp.
Particulate methane monooxygenase and soluble methane monooxygenase, encoded by the pmoCAB operon and mmoXYBZDC operon, respectively, are two central enzymes for converting methane to methanol [54].pmoC and a complete set of gene encoding functions associated with conversion of methanol to CO 2 were detected in the Methylocystis MAG (Figs 4B and 5), supporting its identity as a methanotroph.Both pmoA and pmoB were missing, presumably due to the MAG incompleteness.In addition, the partially present RuMP pathway, the serine cycle assimilating formaldehyde into biomass, and the biosynthetic pathways of acetate, lactate, and pyruvate were detected in the MAG.Thus, methane-derived intermediates (methanol, etc.) from Methylocystis could support the coexistence of accompanying members.For the Methylophilaceae MAG, all the genes encoding enzymes for methanol oxidation, except mdh and mtdA and a complete RuMP pathway, were present in this MAG (Fig. 5).Therefore, Methylocystis and Methylophilaceae could cooperate syntrophically during Fe-AMO.
The Methylocystis and Methylophilaceae MAGs possessed all the necessary genes encoding the ribof lavin biosynthesis pathway (Fig. 5 and Fig. S4), suggesting that they had the potential of ferrihydrite reduction via ribof lavin-mediated EET.To support this statement, the filtrate of the CFeN treatment was analyzed by f luorescence spectrometry and differential pulse voltammetry (DPV).The spectra showed two excitation peaks at 370 and 445 nm and an emission peak at 525 nm (Fig. S5), demonstrating the presence of ribof lavin in the filtrate [55].DPV analyses showed that a redox-active species with a midpoint potential of −173 mV versus the standard hydrogen electrode was present in the filtrate of the CFeN soil slurry (Fig. S6), which could be assigned to ribof lavin [55].c-type cytochromes (c-Cyts) and conductive pili are well known to mediate electron transfer to Fe(III) [56,57].For example, Methylophilus (a genus of Methylophilaceae) was previously demonstrated to transfer electrons extracellularly to ferrihydrite, which was attributed to the mediation of c-Cyts and ribof lavin [50].Both Methylophilaceae and Methylocystis MAGs contained the genes encoding c-Cyts and pili.Thus, Methylophilaceae and Methylocystis might utilize c-Cyts, ribof lavin, and conductive pili for ferrihydrite reduction.Alternatively, the methanobactins of methanotrophs (e.g.Methylocystis) could also produce Fe(II) [40].A further study is needed to distinguish the effect of methanobactins on ferrihydrite reduction.
A nif gene cluster of the Methylocystis MAG was identified in a 11-kb nitrogen fixation region ( Fig. S7).The nif cluster contained 23 genes, including conserved genes nifH, nifD, nifK, nifB, nifE, and nifN, that have been shown to be essential for BNF.These nif genes were organized in an order identical with those of diazotrophic methanotrophs M. echinoides and Methylosinus sp.3S-1, but these were oriented in the opposite direction.Similarly, the genes encoding nitrogenase (nifK), activation proteins (nifZ, nifW, and nifU), nitrogenase cofactor synthesis (nifB, nifE, and nifQ), and genes with unknown functions (nifT) were present in the Methylophilaceae MAG [58].This indicated that Methylocystis and Methylophilaceae had the potential of BNF.The nif gene cluster was f lanked by a fix gene cluster encoding the electron bifurcating protein complex FixABCX (Fig. S7).As a result, Methylocystis sp.(Bin 6) is predicted to be using FixABCX to donate electrons to nitrogenase, as FixABCX was previously suggested to serve as an electron transport chain for BNF in diazotrophs (e.g.Azotobacter vinelandii) [59].
Based on the metagenomic results, a putative metabolic pathway linking methane oxidation to ferrihydrite reduction and BNF by methanotrophs could be identified (Fig. 5).Methane was oxidized to CO 2 where the required oxygen atoms were likely derived from the residual DO or water splitting by Fe(III)-methanobactins [40].The generated reducing equivalent NADH in this process carried electrons to NADH: quinone oxidoreductase and FixABCX.While one part of the electrons was transferred to the quinone pool by these two complexes, the other part was diverted to nitrogenases for BNF.Multi-heme cytochromes in the outer membrane accepted electrons from the quinone pool via Complex III in a final step and used them for direct or ribof lavin-mediated ferrihydrite reduction.Alternatively, ferrihydrite-released Fe(III) was likely reduced by methanobactins of MOB (e.g.Methylocystis) via water splitting [40], providing dioxygen for the initial methane oxidation/activation (i.e.2CH 4 + O 2 → 2CH 3 OH).

Environmental implications
Microbial Fe(III)/Fe(II) cycling plays a fundamental role in biogeochemical reactions, and dissimilatory Fe(III) reduction (DIR) has been shown to be central to many environmental processes [60,61].The present work revealed the capability of Methylocystis and Methylomicrobium to function as new Fe(III) reducers, suggesting a diversity of MOB participating DIR in hypoxic paddy soil.The results also suggested that Fe-AMO is able to occur in nitrogendeficient environments.Although Fe(III) may be depleted by DIR, an alternation of oxic and anoxic conditions by wetting and drying was shown to regenerate Fe(III) to maintain Fe-AMO.The observed BNF driven by Fe-AMO indicates a novel coupling pathway of C, N, and Fe cycling.Nitrogen deficiency is often a limiting factor in soil methane oxidation [62].Fe-AMO coupled BNF is therefore able to alleviate nitrogen constraints, promoting methane oxidation.The fixed nitrogen provides an alternative explanation for unidentified nitrogen sources in soil systems [8] and suggests that iron application may be feasible to improve the soil fertility.

Conclusions
Our results demonstrate that methane oxidation coupled to ferrihydrite reduction was able to significantly (by 13-fold) enhance the BNF rate.Ferrihydrite-mediated methanotrophic BNF contributed 81% to the total BNF in hypoxic paddy soil.Methylocystis, Methylophilaceae, and Methylomicrobium were the predominant methanotrophs/methylotrophs that were responsible for Fe-AMO-coupled BNF.These microbes likely reduced ferrihydrite independently via ribof lavin and c-Cyts or cooperated with other iron-reducing bacteria (e.g.Geobacter).Future research should be devoted to exploring the in situ activity and distribution of this novel BNF across Fe(III)-rich rice paddies.We suggest that future research also comprehensively explore the microbial community structure characteristics of iron-mediated methanotrophic BNF in different regions and the interactions between Fe(II)-oxidizing bacteria and methanotrophs.A further understanding of the inf luencing factors (such as soil organic carbon, fertilization, and the alternation of wetting and drying) is imperative to understand and predict the contribution of iron-mediated methanotrophic BNF to nitrogen accumulation in paddy soils.

Figure 1 .
Figure 1.Kinetics of methane consumption (A), production of Fe(II) (B), and ammonium (C) in the CFeN, CFeN + CF 2 H 2 , and CN groups; the incorporated13 C and15 N abundances in paddy soils incubated under the dual-isotope-labeled condition for 50 days (D); data are mean ± SD of three independent replicates (n = 3).

Figure 2 .
Figure 2. Comparison of the pmoA and nifH mRNA transcripts in the CFeN, CFeN + CF 2 H 2 , and CN groups on Day 20 of incubation; data are mean ± SD of three independent replicates (n = 3).

Figure 4 .
Figure 4.A genome-wide phylogenetic tree showing taxonomic positions of recovered MAGs (A) and the genes for methane oxidation, nitrogen fixation, and electron transfer in the MAGs associated with Methylocystis and Methylophilaceae (B).

Table 1 .
Summary of the physicochemical properties of the collected paddy soil.

Table 3 .
Comparison of Fe(III)-mediated methanotrophic BNF with previous studies.
a Calculated from the fixed nitrogen content of 43.3 μmol N g dw −1 over 30 days.b Calculated by subtracting the nonmethanotrophic BNF rate (35 nmol N 2 g dw −1 h −1 ) from the BNF rate (49 nmol N 2 g dw −1 h −1 ) in the presence of light and methane.c Based on the net BNF (327.8 nmol N 2 g dw −1 ) in Hydrate Ridge Seep (Mat-774) at a depth of 3-6 cm in the presence of methane over 258 days.d Normalized to the bacterial dry weight.-denotes data not available.