Geoderma 431 (2023) 116375 Contents lists available at ScienceDirect Geoderma journal homepage: www.elsevier.com/locate/geoderma Decreased CH4 emissions associated with methanogenic and methanotrophic communities and their interactions following Fe(III) fertiliser application in rice paddies Yihe Zhang a, Mengyuan Huang a, Kai Yu a, Yuxin Xie a, Yuxin Wang a, Jie Wu a, Fengwei Zheng a, Shuang Wu a, b, Shuqing Li a, b, *, Jordi Sardans c, d, Josep Peñuelas c, d, Jianwen Zou a, b a Jiangsu Key Laboratory of Low Carbon Agriculture and GHGs Mitigation, College of Resources and Environmental Sciences, Nanjing Agricultural University, Nanjing 210095, China b Jiangsu Key Lab and Engineering Center for Solid Organic Waste Utilization, Jiangsu Collaborative Innovation Center for Solid Organic Waste Resource Utilization, Nanjing Agricultural University, Nanjing 210095, China c CSIC, Global Ecology Unit CREAF-CSIC-UAB, Cerdanyola del Valles, Catalonia 08193, Spain d CREAF, Cerdanyola del Valles, Catalonia 08193, Spain A R T I C L E I N F O A B S T R A C T Handling Editor: Diego Abalos Rice paddies are important sources of atmospheric methane (CH4), where CH4 fluxes are determined by the balance of production and oxidation. While the application of Fe(III) fertilizer is a proven effective strategy for Keywords: mitigating CH4 emissions from rice paddies, little is known about the role of methanogen and methanotroph Fe(III) fertilisation communities and their interactions. In this study, we measured the CH4 flux in a rice paddy field experiment, CH4 emission where the responses of methanogenic and methanotrophic activities to Fe(III) fertiliser applied at medium (Fe-M) Rice paddies and high (Fe-H) levels were examined using real-time quantitative PCR coupled with co-occurrence network Methanogens Methanotrophs analysis after Illumina MiSeq sequencing of the mcrA and pmoA genes. Relative to the control without Fe(III) Co-occurrence networks fertiliser application, in the Fe-M and Fe-H treatments, seasonal CH4 emissions were significantly decreased by 65% and 62%, respectively. Fe(III) fertilisation significantly decreased the copy numbers and potential activity of methanogens, and had an opposite effect on methanotrophs. We found Methanothrix was the predominant genus more adaptive to grow in Fe-rich environments. Fe(III) fertilisation caused a shift in the methanotrophic com­ munity towards a predominance of Type II methanotrophs. Soil total Fe concentration, soil electrical conduc­ tivity, and different valences of iron were key factors in the composition of methanogenic and methanotrophic communities. Fe(III) fertilisation led to a less complex community taxonomic structure mostly attributed to fewer nodes and edges in both the co-occurrence and mutual exclusion networks of methanogen and methanotrophs. Our findings suggest that adding Fe(III) fertiliser in rice paddies could mitigate CH4 emissions by regulating both methanogens and methanotrophs. 1. Introduction Methane is mainly produced under anoxic conditions, where CH4 fluxes depend on the combined performance of methanogenic and Methane (CH4) is one of the most potent atmospheric greenhouse methanotrophic communities (Kong et al., 2019; Tokida et al., 2010; gases, with a 25-fold radiative forcing than carbon dioxide (CO2) over a Wang et al., 2018). In the past decade, many studies have been dedi­ 100-year period (IPCC, 2013). In agriculture, CH4 can be produced from cated to optimising agricultural practices for mitigating CH4 emissions fermentative digestion by ruminant livestock, stored manure, and from rice paddies, such as crop residue incorporation, fertilisers and waterlogging rice paddies. Estimated at 36 Tg yr− 1 globally, CH4 emis­ water regimes (Kong et al., 2019; Qin et al., 2010; Zou et al., 2004). In sions from rice paddies are one of the most important anthropogenic addition, it has been shown that a functional heavy metal-antagonist sources of atmospheric CH4 (IPCC, 2013). such as fertilisation with Fe(III) has the potential to mitigate CH4 * Corresponding author at: Jiangsu Key Laboratory of Low Carbon Agriculture and GHGs Mitigation, College of Resources and Environmental Sciences, Nanjing Agricultural University, Nanjing 210095, China. E-mail address: shuqingli@njau.edu.cn (S. Li). https://doi.org/10.1016/j.geoderma.2023.116375 Received 5 July 2022; Received in revised form 1 February 2023; Accepted 3 February 2023 Available online 11 February 2023 0016-7061/© 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/). Y. Zhang et al. Geoderma 431 (2023) 116375 emissions from rice paddies (Chen et al., 2017). Fertilisers containing hypothesized that the response of CH4 emissions to Fe(III) can be active Fe and Fe3+, which can act as electron acceptors, have been explained by the combined performance of methanogenic and meth­ shown to decrease CH4 emissions from rice paddies (Ali et al., 2008; anotrophic communities (i.e., abundance, composition, and activity). Arao et al., 2009; Gwon et al., 2018). Waterlogging can create an anoxic The main objective of this study is to examine the Fe(III) fertiliser- soil environment to favour Fe reduction coupled with organic matter induced decrease in CH4 emissions associated with CH4 production mineralization, which can reduce the transfer of electrons by meth­ and oxidation processes in terms of the abundance and community anogenesis and eventually decrease CH4 emissions (Fredrickson and structure of methanogens and methanotrophs. Gorby, 1996; Lovley and Phillips, 1987; Sun et al., 2016). Moreover, CH4 production can be suppressed by controlling the ratio of easily oxidiz­ 2. Materials and methods able organic matter to easily reducible iron. It is well documented that methanogenic archaea can be affected by Fe(III) fertiliser application in 2.1. Site description and field experiments paddy soils. For example, Fe(III) in oxides or available Fe(III) may serve as an electron acceptor to suppress the growth of methanogens and thus The field experiment over the 2019 rice-growing season was con­ influence the methane cycle (Bond and Lovley, 2002; Bowman et al., ducted in a paddy field on an ecological landscape in Xinghua city, 1994; Lovley and Phillips, 1987). The inhibition of methanogenesis has Jiangsu province, China (32◦ 96′ N, 120◦ 01′ E). In this rice paddy, Fe(III) been attributed to the competition between dissimilatory iron-reducing fertiliser had not been applied before the experiment. The cropping bacteria and methanogens (i.e., Methanocella and Methanosarcina) for regime of the experimental site was dominated by an annual paddy rice- the same substrate, such as hydrogen (Bodegom and Stams, 1999; winter wheat rotation system. The site has a typical monsoonal climate, Lovley, 1987). with an average annual rainfall of 1049 mm and annual mean temper­ On the other hand, aerobic interfaces of methanogenic environments ature of 14.8◦ C. The soil of the experimental site is classified as gleyic may provide habitats for methanotrophs in rice paddies, where CH4 can stagnic anthrosol, with a fine-scale soil texture (35.6% clay, 31.7% silt, be oxidized by methanotrophs. While the studies of Fe(III) fertiliser ef­ 32.7% sand), a bulk density of 1.09 g cm− 3, a pH of 7.15 to 7.26, and a fects on CH4 emissions have mostly focused on the CH4 production total nitrogen content of 1.15 g N kg− 1. process driven by methanogens in rice paddies (Chen et al., 2017; Zhang All experimental plots were managed in line with typical local et al., 2012), CH4 oxidation, as an important link in the methane cycle, cultivation practices (e.g., rice cultivar selection, tillage, water irriga­ has attracted little attention in the context of Fe(III) fertiliser application tion, nitrogen phosphorus and potassium fertiliser application regimes, in rice paddies. The distribution of Type I (Gammaproteobacteria) and and pest and weed control methods). The Fe(III) fertiliser used in this Type II (Alphaproteobacteria) methanotrophs is highly related to soil study was generally Fe(III) oxide-hydroxide hydrate, and the available physical and chemical properties, such as soil pH and salinisation, which iron content was 85%. The experimental treatment included Fe(III) need to be further examined following Fe(III) fertiliser application in fertiliser applied at three levels, namely, the control plots without Fe(III) rice paddies (Hanson and Hanson, 1996; Heyer et al., 2002; Ho et al., fertiliser application and ferric hydroxide fertiliser(Fe(OH)3) applied at 2011). Therefore, qPCR and Illumina MiSeq sequencing techniques rates of 4.0 t ha− 1 (Fe-M) and 8.0 t ha− 1 (Fe-H). A random block design would provide an opportunity to simultaneously examine the abun­ was adopted for the experimental plots with three replicates. Each field dance and community composition of methanogens and methanotrophs plot was 4 m by 4 m in size. In accordance with the local fertiliser (Zhou et al., 2020), which would be helpful to gain insight into methane regime, two chemical N fertiliser, namely, compound N fertiliser (N: production and oxidation processes following Fe(III) fertiliser applica­ P2O5:K2O) and urea (46% N), were applied at a rate of 292 kg N ha− 1 per tion in rice paddies. plot, with 20% of the fertiliser applied as the base dose and the Microorganisms can interact in complex communities that affect remaining 80% applied as topdressing at three time points over the rice- microbial processes in the environment (Kaupper et al., 2021). Recently, growing season. Basal fertiliser was spread and then mixed with the soil co-occurrence networks have been used to map all the relationships by ploughing on June 16, 2019. between microorganisms within a community, and these interactions The average active and free iron concentrations ranged between 52% can be divided into three categories: joint evolution of organisms in the and 60%; Fe(III) fertiliser application rates in the Fe-M and Fe-H treat­ environment (positive relationship known as co-occurrence); opposite ments represent the local recommended medium and high levels for rice evolution (negative relationship known as co-exclusion), and the cultivation in Southeast China, respectively (Zhou et al., 2009). All plots absence of a relationship (a true lack of interaction or a null combination were flooded two days prior to rice sowing (15 June 2019) and remained of all interactions between microorganisms) (Gross, 2008; Karimi et al., flooded until 29 August 2019; thereafter, the plots were drained and 2017). Network analysis has been widely adopted to investigate the co- remained in the drainage stage until crop harvest. The cropping period occurrence patterns of methanogens and methanotrophs in soils (e.g., (from planting to harvest) was roughly June-September for the 2019 Fan et al., 2021; Kaupper et al., 2021). Specifically, the interaction rice-growing season. In the Fe(III)-fertilised plots, Fe(III) fertiliser network in the methanotrophic community became more complex together with crop residue was incorporated at a soil depth of 10–15 cm during recovery from desiccation-rewetting, and active anaerobic before flooding was initiated. Fe(III) fertiliser and crop residue were oxidation of methane (AOM) pathways compensated a fraction of CH4 adequately mechanically mixed prior to incorporation. produced during ongoing methanogenesis as identified by the network analysis (Kaupper et al., 2021; Fan et al., 2021). However, there is 2.2. Gas sampling and flux measurement limited knowledge about how Fe(III) fertilisation impacts methanogenic and methanotrophic groups involved in commonly co-occurring links The fluxes of CH4 were measured in situ using the static chamber-gas between networks. chromatograph (GC) method detailed in Kong et al. (2019). Before the To bridge this knowledge gap, we carried out a field experiment in fields were initially flooded, specially made aluminium flux collars (0.4 the 2019 rice-growing season under different application levels of Fe(III) m long by 0.4 m wide by 0.15 m tall) were permanently installed in each fertiliser. We measured the CH4 fluxes, methane production potential field plot to ensure the reproducible placement of the gas-collecting (MPP), and methane oxidation potential (MOP) following Fe(III) fertil­ chambers for gas sampling over the rice-growing season. The top edge iser application in rice paddies. The abundance and diversity of the of the collar had a groove that was filled with water to seal the rim of the methanogens and methanotrophs were analysed by real-time quantita­ chamber. The sampling chambers were made of opaque PVC materials tive PCR (qPCR) and Illumina MiSeq sequencing techniques, respec­ and were 0.4 m long by 0.4 m wide by 0.5 m tall. The chamber inside tively. Community network analysis was adopted to describe the specific was equipped with a circulating fan to ensure complete gas mixing, and functional composition of methanogens and methanotrophs. We it was wrapped with a layer of sponge and aluminium foil to minimise 2 Y. Zhang et al. Geoderma 431 (2023) 116375 air temperature changes inside the chamber during the sampling period. fertilisation and then twice or three times a week to examine seasonal For gas sampling, the rim of the chamber was placed into the groove of variations in soil properties (Table S1). Soil samples were collected from the collar, including any vegetation growth. Soil CH4 fluxes were the plough layer (0–20 cm) for each of the three replicates per treatment. measured once a week before N fertiliser application and then three All rhizosphere soil was gently scraped from the roots before placing the times a week. Gas samples were collected at 0, 5, 10, 15, and 20 min soil into a sterile sealable plastic bag, which was transported to the after chamber closure between 09:00 and 11:00 LST on each sampling laboratory in an ice box containing liquid nitrogen. The fresh soil sam­ day. Individual gas samples of 60 mL were collected from the headspace ples stored at 4◦ C were used to analyze soil physicochemical properties of the chamber using gas sampling bags and then transported to the after undergoing air drying at 25◦ C within one week of sampling. The laboratory for analysis by GC within a few hours. rest of the rhizosphere soil samples were stored at − 80◦ C to determine The gas samples were analysed using a modified GC (Agilent 7890A) microbial indicators. Prior to analysis, all soil subsamples were ground equipped with a flame ionisation detector (FID) (Qin et al., 2010; Zou to pass through a 2 mm sieve. The contents of soil NO–3-N and NH+ 4 -N et al., 2009). The oven and FID were operated at 55◦ C and 200◦ C, were measured using continuous flow analysis (Skalar SANplus analy­ respectively. The carrier gas (N2) flow rate was 30 mL min− 1. A ser, Amsterdam, Netherlands). The methods we used for determining nonlinear fitting approach based on the slope of the change in the soil physicochemical properties (pH, EC, TC, TN, C/N, and dissolved mixing ratio over time (measured at 0, 5, 10, 15, and 20 min) was organic carbon, or DOC) are described in Kong et al. (2019) and Yu et al. applied to determine the CH4 flux. The average and standard deviations (2021). The total Fe concentration was measured by digestion with HF- of the CH4 flux for each treatment were calculated from three replicates. HNO3-H2O2. The content of soil Fe2+ was measured using the ferrozine- Seasonal total CH4 emissions were determined as the sequential accu­ ultraviolet absorbance method (Stookey, 1970). Grain yield was mulation of the flux measurements between every-two adjacent tem­ measured at the physiological maturity stage by hand-harvesting all poral intervals (Zou et al., 2005). Yield-scaled CH4 emissions were plants in each 4 m × 4 m plot. Above-ground biomass, including rice calculated as the seasonal total of CH4 emissions divided by rice yield for straw and grain yield, was measured at harvest time by oven drying the each plot. samples at approximately 75◦ C to a constant weight. 2.3. Measurement of methane production and oxidation potentials 2.5. DNA extraction and real-time PCR assays In each experimental plot, rhizosphere soil samples (0–20 cm) were Total genomic DNA samples were extracted using MO BIO PowerSoil collected on 14 August 2019 and 2 October 2019 to investigate the ef­ DNA Isolation Kits (MO BIO Laboratories, Carlsbad, CA, USA) according fects of Fe(III) fertiliser on methane production potential (MPP) and to the manufacturer’s instructions. The concentration of DNA samples methane oxidation potential (MOP) during the waterlogging and was determined by Nanodrop (Thermo Scientific, USA). Then, DNA drainage stages. The 10 g soil samples used to measure the MPP and samples were used for quantitative polymerase chain reaction (qPCR) MOP were homogenized before being placed into bottles. The MPP was and Illumina MiSeq sequencing analysis. determined according to the method described by Singh et al. (2012) The copy numbers of the mcrA and pmoA genes were determined in a with minor modifications. Briefly, 10 mL of anoxic sterile water was StepOneTM real-time PCR system (Applied Biosystems, Germany) using added to 10 g of fresh soil samples in a 100 mL serum bottle to ensure the the primer pair mlas/mcrA-rev (Steinberg and Regan, 2008) and A189F/ soil was covered by a 10 mm water. To create an anaerobic environment, Mb661R (Costello and Lidstrom, 1999), respectively. The qPCR ampli­ we repeatedly purged the headspace of serum bottles with O2-free N2 gas fications were performed in a total volume of 20 µL using SYBR® Premix for 3 min to remove residual CH4 and O2, before incubating them in a Ex Taq (Takara, China) with a reaction mixture of 10 µL of SYBR® dark environment at 30 ◦ C for approximately 72 h. Before gas sampling, Premix Ex Taq™, 0.4 μL of each primer (10 μmol/L), 0.4 μL of ROX the serum bottles were shaken vigorously to establish equilibrium be­ reference dye (50 × ), 2 μL of template DNA and 6.8 μL of sterile water. tween the liquid and gas phases. Then, gas samples (4 mL) were The amplified fragments for each gene were cloned in the pMD 18-T collected from the headspace with a 5 mL syringe, and an equal volume vector and sequenced. The standard curves for both genes were pre­ of N2 was injected back into the bottle afterwards. Gas samples were pared by using triplicate 10-fold dilutions of linear plasmid DNA. taken at intervals of 1, 60, and 120 h. Then, the gas samples were Amplification was performed in triplicate under the following cycling analysed for CH4 using a gas chromatograph. conditions: 30 s at 95 ◦ C; 40 cycles at 95 ◦ C, 55 ◦ C and 72 ◦ C for 5 s, 30 s, The methane oxidation potential (MOP) was determined by adding and 15 s, respectively; and a dissociation stage at 95 ◦ C for 15 s, 55 ◦ C for 10 g of soil and 10 mL of deionized water into a 150 mL serum bottle, 30 s, and 95 ◦ C for 15 s (Li et al., 2017). according to the method described in Gao et al. (2020) and Kim et al. (2016). Approximately 1.5 mL of CH4 (10 000 ppmv) was injected into 2.6. Illumina MiSeq sequencing each bottle using a 10 mL syringe and then dark-incubated and shaken at 120 rpm at 30 ◦ C for two days. Gas samples (4 mL) were collected as The structures of the methanogen and methanotroph communities described above, and an equal volume of air was injected back into the were assessed by Illumina MiSeq sequencing of the mcrA and pmoA bottles after each gas sampling. The concentration of CH4 was measured genes, respectively. The highly conserved mcrA gene encoding the every 2 h on the first day and every 3 h on the second day, as in Gao et al α-subunit of methyl coenzyme M reductase has been widely used for (2020). The MPP or MOP was calculated by the following formula: analysis and quantification of methanogen communities, and the pmoA gene encoding methane monooxygenase subunit A is used as the indi­ MPP or MOP (µg g− 1 d− 1) = W×(1−K×VWH/100),where K (µg g − 1 − 1 d ) is the cator for quantifying the methanotrophs from environmental samples slope of the linear curve for the accumulated CH4 concentration deter­ (Pereyra et al., 2010; Sonoki et al., 2013; Wasmund et al., 2009). The mined over an interval of three sampling times; V (L) is the volume of the mlas/mcrA-rev (Steinberg and Regan, 2008) and A189F/Mb661R serum bottle; W (g) is the weight of fresh soil; and WH (g) is the weight (Holmes et al., 1995) primers were set up for amplification of the mcrA of H2O per 100 g fresh soil. and pmoA genes, respectively. The Axyprep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) was used to purify the tar­ 2.4. Soil sampling and analysis geted bands (approximately 410 bp for mcrA and 478 bp for pmoA), and QuantiFluorTM-ST (Promega, Madison, WI, USA) was employed to Before rice sowing, soil samples from three replicate plots were quantify the targeted bands. For each sample, PCR was repeated in collected and mixed thoroughly to measure the soil physicochemical triplicate, prior to mixing the three PCR products. The PCR products of properties. In addition, soil samples were collected once a week before N each DNA sample were evaluated by 2% agarose gel electrophoresis. The 3 Y. Zhang et al. Geoderma 431 (2023) 116375 equimolar purified amplicons were pooled and paired-end sequenced (2 methanotrophic abundance and activities were fitted by a linear model × 250) on the Illumina MiSeq platform according to the standard pro­ with ordinary least squares (OLS). A t test was used to examine the tocols at Shanghai BIOZERON Biotechnology Co., ltd. (Shanghai, statistical significance of parameter estimates in the simulated OLS China). The Illumina MiSeq sequencing data are available in the NCBI model. Spearman correlation coefficients were employed to test the Sequence Read Archive (SRA) database under accession number relationships between soil properties and the relative abundance of PRJNA725635. genera in R. Statistical significance was determined at the 0.05 proba­ bility level. A permutational multivariate analysis of variance (PER­ 2.7. Illumina data analysis MANOVA) was conducted using the Adonis function in the R vegan package with 999 permutations to test the significance of independent According to Caporaso et al. (2010), we found three criteria for variables that accounted for the divergence in communities of metha­ demultiplexing and quantity filtering the raw mcrA and pmoA sequences nogens and methanotrophs (standard Mantel test). Redundancy analysis by Quantitative Insights Into Microbial Ecology (QIIME) (Version 1.17) (RDA) of OTUs was performed to summarize the variations in meth­ after sequencing was completed: (1) sequences with reads shorter than anogenic and methanotrophic communities that could be explained by 200 bp and quantity scores below 25 were discarded from further the soil properties using the ggbiplot package in R. The means of analysis; (2) sequences were clustered into operational taxonomy units different relative abundance of methanogenic genera from The Fe(III) (OTUs) at a 97% identity threshold; and (3) UCHIME was used to fertilised treatment (Xe) and control (Xa) groups were used to evaluate identify and remove chimeric sequences (Edgar et al., 2011). The mcrA effect size in the form of the natural log-transformed response ratio sequences were binned into species-level OTUs at 84% sequence iden­ (lnR). The standard deviations of both treatment and control groups tity, corresponding to 97% similarity based on the 16S rRNA gene by were included as a measure of variance: using UCLUST (Yang et al., 2014), and pmoA sequences that shared 87% Xe similarity, corresponding to 97% similarity based on the 16S rRNA gene, lnR = ln ( ) Xa were binned into species-level OTUs with UCLUST (Degelmann et al., 2010). The taxonomy of the mcrA and pmoA genes was analysed using where Xe and Xa are means in the Fe(III) fertilised treatment and control the Non-Redundant Protein Sequence Database. To calculate the α-di­ groups, respectively. The mean response ratio (RR++) is calculated from versity of the methanogen community, we used the abundance-based the individual lnR of each paired comparison between control and coverage estimator (ACE), Chao1 estimator, Shannon diversity and treatment groups with the weight of each lnR using a categorical random Good’s coverage. effect model (Zheng et al., 2020). Fe-M and Fe-H treatments were merged to assess the effects of the methanogen community composition 2.8. Construction of the microbial co-occurrence network on the paddy soil. The effects of different methanogenic genera were considered significant if the 95% CI did not overlap with the line lnR = Co-occurrence network analysis was performed based on the pro­ 0. portionally normalized OTU tables with Cytoscape (v3.8.2) software using the CoNet plug-in (v1.1.1). Prior to network construction, mcrA 3. Results and pmoA relative abundance matrices were grouped into a single ma­ trix so that links could be computed between OTUs of the same variant 3.1. CH4 fluxes (i.e., pairs of mcrA-mcrA or pmoA-pmoA) and between OTUs of different variants (i.e., mcrA-pmoA). To reduce false-positive predictions, the taxa The seasonal pattern of CH4 fluxes did not differ with Fe(III) fertiliser represented in above one-third of the samples were reserved for network application while being sensitive to the water regime (Fig. 1). During analysis. Copresence and exclusion links were identified with an initial flooding the CH4 fluxes remained at low levels within 35 days of ensemble-based approach, including two measures of correlation broadcast sowing. Thereafter, the CH4 fluxes increased to the largest (Pearson and Spearman) and dissimilarity (Bray-Curtis and Kullback- values in the following two–three weeks. The CH4 fluxes rapidly Leibler) to increase the robustness of the analysis (Faust and Raes, decreased because of the midseason drainage episode, followed by 2016). The interaction sign was used to distinguish between co-presence relatively low flux rates until harvesting. and exclusion links, which were considered undirected because of the A two-way ANOVA indicated that CH4 emissions were significantly nature of the correlation or dissimilarity measures used. Statistical sig­ affected by Fe(III) fertiliser, flooding or drainage episodes and their nificance was tested by obtaining the link- and measure-specific p value interaction (Table 1 and 2). The average seasonal CH4 fluxes from the as the mean of the permutation distribution under the bootstrap distri­ control plots were 33.33 mg m− 2 day− 1, 186% and 166% greater than bution, using 1000 iterations for each distribution. Probability values for those from the Fe-M and Fe-H treatments, respectively. Substantial CH4 different correlation or dissimilarity measures supporting the same link emissions occurred mostly during flooding episodes for all treatments. were merged using Brown’s method and corrected for multiple testing Relative to the control, in the treatment, Fe(III) fertiliser application using the Benjamini–Hochberg procedure. Finally, to reduce the detec­ significantly decreased yield-scaled CH4 emissions; these did not differ tion of false positives, we included only those links supported by at least between the Fe-M and Fe-H treatments (Table 2). two measures of correlation or dissimilarity and having an adjusted merged p value below 0.05. The correlation results were then imported into Gephi (v0.9.2), and the microbial co-occurrence network was 3.2. Soil physicochemical properties visualised using the Fruchterman Reingold algorithm. The network to­ pological parameters, including the degree of each node, clustering co­ Fe(III) fertiliser application significantly increased soil EC and pH (p efficient, network density and average path distance were calculated in < 0.05) over the rice-growing season (Fig. S1). Soil NH+ 4 -N showed a Gephi (v0.9.2). significant and positive relationship with CH4 fluxes (Fig. S1 and Table S1). The concentrations of total Fe and Fe(II) in the soil were 2.9. Statistical analysis measured to gain insight into the transformation of Fe during two typical periods of rice growth (Fig. S2). Relative to the control, soil total Fe and Two-way analysis of variance (ANOVA) was used to examine the Fe(II) concentrations were significantly greater in the Fe(III)-fertilised main and interaction effects of Fe(III) fertiliser treatment and flooding or treatments in the flooding episode (p < 0.05). However, this differ­ drainage stages on the abundance and activity of methanogens and ence was not pronounced in the late drainage stage of rice cultivation, methanotrophs. The correlations of CH4 fluxes with methanogenic and suggesting there were unique anoxic conditions favourable for the 4 Y. Zhang et al. Geoderma 431 (2023) 116375 Fig. 1. Seasonal dynamics of CH4 fluxes and cumulative CH4 emissions as affected by different levels of Fe fertiliser in the 2019 rice season. The control group corresponds to no Fe(III) fertiliser addition; the Fe-M treatment corresponds to a medium concen­ tration of Fe(III) fertiliser; the Fe-H treatment corresponds to a high concentration of Fe(III) fertiliser. Vertical bars indicate the standard errors of three replicates. Black arrows represent fertil­ isation time points. Letters a and b indicate significant differences at p < 0.05 level. Table 1 Two-way analysis of variance (ANOVA) of methane production potential (MPP), methane oxidation potential (MOP), and log copy numbers of mcrA and pmoA genes; df: degree of freedom, MS: mean square, and F: F value. *, **, and *** denote p < 0.05, p < 0.01, and p < 0.001, respectively. Source of variance df CH4 fluxes MPP MOP mcrA pmoA MS F MS F MS F MS F MS F Fe(III) fertiliser levels 2 1412.2 17.53*** 66.181 44.124*** 4.46 × 104 18.65*** 0.812 21.745*** 0.273 7.54** Flooding/Drainage 1 2063.4 25.61*** 23.029 15.354** 1.28 × 105 53.33*** 0.486 13.008** 3.446 95.023*** Fe(III) fertiliser levels × Flooding/Drainage 2 1317.4 16.35*** 12.545 8.364** 4.35 × 103 1.82 0.031 0.826 0.017 0.477 contrast to a negative correlation between the MOPs and CH4 fluxes (p Table 2 < 0.01; Fig. S3). Seasonal CH4 emissions (kg ha− 1), grain yield (t/ha), and yield-scaled CH4 (kg CH4 t− 1 yield) from rice paddies treated with different concentrations of Fe fertiliser regimes over the 2019 season. Control: no Fe fertiliser; Fe-M: medium 3.4. The response of methanogenic population size and community concentration of Fe fertiliser; Fe-H: high concentration of Fe fertiliser. Different composition to Fe(III) fertilisation letters (a, b) indicate a significant difference between treatments at p < 0.05. Treatments Seasonal CH4 emissions Grain yield Yield-scaled Both methanogenic population size and community composition CH4 were significantly affected by Fe(III) fertiliser application. The mcrA Flooding (kg Drainage (kg （t/ha) （kg CH4 t− 1 absolute abundance is significantly lower, by 67% and 86%, respec­ ha− 1) ha− 1) yield) tively, in the Fe-M and Fe-H treatments than in the control (p < 0.05) Control 170.92 ± 12.68 ± 2.23 a 10.26 ± 18.24 ± 2.61 a (Fig. S4a). In the methanogenic community, the top eight most abundant 22.32 a 0.62 a methanogenic genera were identified. Together, they account for almost Fe-M 57.78 ± 6.90b 5.84 ± 0.77 a 10.39 ± 6.20 ± 0.36b 85% of the relative abundance of all methanogenic groups. Methanothrix 0.60 a Fe-H 54.77 ± 3.29b 12.85 ± 2.03 a 10.03 ± 7.06 ± 1.07b (15.03%-34.46%), Methanocella (17.97%-25.95%), and Meth­ 0.78 a anobacterium (24.41%-29.17%) are the three most abundant genera (Fig. 3a). Fe(III) fertiliser application significantly altered the meth­ anogenic community composition. Specifically, the relative abundance conversion between Fe3+ and Fe2+ in rice paddies. of Methanothrix increased by 77%, and that of Methanocella and Meth­ anobacterium decreased by 32% (95% Cl, 37–22%) and 13%, respec­ 3.3. Methane production potential (MPP) and methane oxidation tively (Fig. 4a). Furthermore, both the Fe(III) fertiliser-increased and Fe potential (MOP) (III) fertiliser-decreased genera were significantly related to CH4 fluxes, but the regression line has a steeper slope (4.71) for the decreased A two-way ANOVA showed that the Fe(III) fertiliser and flooding or genera than for the increased genera (2.79) (ANOVA, p < 0.01, Fig. 4b). drainage episodes significantly altered the MPP and MOP, while their Redundancy analysis (RDA) and the standard Mantel test were used interaction, to a large extent, affected the MPP (Fig. 2 and Table 1). to investigate the relationship between the soil microbial composition of Relative to the control plots, the MPPs of the Fe-M and Fe-H treatments methanogens and methanotrophs (OTUs data) and soil properties. The were significantly decreased by 60% and 72% during the flooding period RDA biplot revealed that 95% of the total variability is explained by the and by 34% and 59% during the drainage period, respectively (Fig. 2a). first two axes of the RDA (Fig. 5a). In general, the MPPs, NH+ 4 -N, EC, In contrast, the MOPs were significantly greater in the Fe-M and Fe-H total Fe and Fe2+ concentrations show a significant relationship with the treatments than in the control (p < 0.05), suggesting that Fe(III) fertil­ methanogenic community (p < 0.01). The methanogenic communities iser application can facilitate soil CH4 oxidation. Moreover, a significant in the Fe-M and Fe-H treatments are separated from those in the control positive correlation was found between the MPPs and CH4 fluxes, in along the x-axis of the RDA plots. In the Fe-M and Fe-H treatments, the 5 Y. Zhang et al. Geoderma 431 (2023) 116375 Fig. 2. Methane production potentials (a) and methane oxidation potentials (b) under different levels of Fe(III) fertilisation treatments. Values are means ± SE (n = 3). Different letters (a, b, c) indicate a significant difference between treatments at p value < 0.05. Fig. 3. Relative abundance of methanogens (a) and methanotrophs (b) under different levels of Fe(III) fertilisation treatments at different water management stages. Relative abundance of methanogens and methanotrophs for the taxonomic levels of the genus. Values are means ± SE (n = 3). Fig. 4. (a) Top 8 most abundant methanogenic genera, other less abundant genera, and unclassified genera in response to Fe(III) fertiliser amendment during flooding episodes. Error bars represent 95% bias-corrected confidence intervals. Asterisks denote significant differences from the control plots (without Fe(III) fertiliser). * p < 0.05, ** p < 0.01, *** p < 0.001. (b) Dependence of CH4 fluxes on the relative abundance of methanogenic genera that responded positively and negatively to treatment analysed from (a). methanogenic composition is negatively related to MPPs but positively 3.5. The response of methanotrophic population and community related to EC, total Fe, and Fe2+ concentration. Indeed, the dominant compositions to Fe(III) fertilisation genera (Methanothrix and Methanocella) are both significantly related to the Fe(II) concentration, and MPPs (p < 0.05; Fig. 6a). Therefore, Fe(III) The results show a trade-off between the absolute abundances of fertiliser application stimulated the electrovalence of Fe, leading to mcrA and pmoA over the flooding and drainage episodes (Fig. S4). Fe(III) changes in methanogenic abundance and structure that affected CH4 fertilisation had a positive effect on pmoA absolute abundance during production potentials. flooding period and no significant effect difference during the drainage stage (Fig. S4b). A negative correlation (R2 = 0.68, p < 0.01) was found between the CH4 flux and pmoA absolute abundance across different levels of Fe(III) fertilisation over the rice-growing season (Fig. 6b). The overall distribution pattern and α-diversity of the 6 Y. Zhang et al. Geoderma 431 (2023) 116375 Fig. 5. Redundancy analysis (RDA) biplot of the relationship between methanogenic (a) and methanotrophic (b) communities and soil properties at the flooding stage over the 2019 rice-growing season. Treatments defined as in Fig. 1. methanotrophic community at the genus level were not significantly treatments the nodes are largely comparable, there are fewer network affected by Fe(III) fertilisation, flooding or drainage episodes, or their edges in the Fe-H treatment (for methanogens, 289 vs 329 in the control, interactions (Table S2). However, adding Fe(III) fertilisation altered the and for methanotrophs, 130 vs 144 in the control) (Fig. 7b; Table 4). relative abundance of methanotrophic genera. The methanotroph Moreover, the average degree, clustering coefficient, and network den­ community consisted of Type I genera, including Methylosarcina, Meth­ sity in the Fe-H treatment are lower than those in the control (Table 4). ylomonas, Methylobacter, Methylocaldum, Methylogaea, Methylomicrobium The most obvious difference was observed in the methanogenic mutual and Methylococcus, and Type II genera, including Methylocystis and exclusion networks (49 nodes and 47 edges vs 48 nodes and 109 edges in Methylococcus (Fig. 3b). Methylosarcina and Methylomonas of Type I the control). methanotrophs and Methylocystis of Type II methanotrophs were iden­ tified as the main methanotrophs, accounting for 72–82% of the total 4. Discussion abundance. Fe(III)-fertilised plots were dominated by the Methylocystis genus in Type II methanotrophs (44–50%), which were more abundant In agreement with previous studies, our results show the rice grain than in the control (33%). In contrast, the relative abundance of Meth­ yields were not significantly affected by Fe(III) fertiliser application ylomonas (Type I methanotrophs) was lower in the treatments than in (Table 2) (He et al., 2013; Xu et al., 2018). The effects of the two Fe(III) the control. Therefore, the application of Fe(III) fertiliser could have fertilisation treatments on CH4 emissions in paddy soils are not signifi­ altered the methanotrophic community composition, mainly by cantly different, which can be explained by the absence of the difference increasing Type II methanotrophs (Methylocystis genus) and decreasing in the concentration of soil total Fe and Fe2+. The seasonal CH4 emis­ Type I methanotrophs (Methylomonas genus). sions from the control totalled 184.73 kg CH4 ha− 1, and Fe(III) fertiliser Unlike for methanogens, for methanotrophs, the three treatments are application significantly decreased the seasonal CH4 emissions by clustered on the redundancy analysis biplot (Fig. 5b). MOPs, EC, pH, 62–65%. total Fe and Fe2+ concentration show a significant relationship with The mitigation of CH4 emissions by Fe(III) fertiliser amendment methanotrophic composition (p < 0.01), and are especially positively could be attributed to changes in the abundance and composition of correlated with Fe-M and Fe-H treatments. Furthermore, the pairwise methanogen and methanotroph communities. Generally, iron oxides correlation revealed that soil properties such as the EC and the elec­ and their biological reduction processes directly or indirectly affect trovalence of Fe contributed most to the difference in key methano­ anaerobic methanogenesis and methane oxidation processes in an trophic genera between treatments. anoxic environment (Beal et al., 2009; Weber et al., 2006). The addition of Fe(III) fertiliser can promote the reduction of Fe(III) and lead to a decrease in methanogen abundance (Figs. S2a and S4a). This inhibition 3.6. Microbial co-occurrence network is primarily explained by the competition between methanogens and dissimilatory Fe(III)-reducing bacteria (DIRB) for the common sub­ The integrated co-occurrence networks across methanogens and strates acetate and hydrogen, together with the direct inhibition of methanotrophs were compared between the control and Fe-H treat­ methanogenesis by amorphous Fe(OH)3 at some concentrations (Van ments in the flooding stage. The network of the control is more complex Bodegom et al., 2004). Recently, more attention has been paid to the than that of the Fe-H treatment, consisting of more nodes and edges (97 anaerobic oxidation of methane (AOM) coupled with alternative elec­ nodes and 854 edges in the control vs 95 nodes and 741 edges in the Fe- tron acceptors (AEAs) because of its potential activity of consuming CH4 H treatment) (Fig. 7, Table 3). In the control group, Methanocella and in different environments, such as freshwater sediments (Beal et al., Methylocystis are the most dominant genera in the methanogenic and 2009), peatlands (Reumer et al., 2018), tropical soils (Blazewicz et al., methanotrophic networks, respectively. However, in the Fe-H treat­ 2012) and rice paddies (Fan et al., 2019). However, Fan et al. (2020) ment, the proportion of Methanothrix is higher than that of the other reported that adding Fe3+ did not support AOM and even partially genera (Tables S3 and S4). Relative to the control, the Fe-H treatment suppressed the AOM rates. In this study, amendment with Fe(III) fer­ has a lower average path distance (2.167 vs 2.367 in the control), lower tiliser (Fe-M and Fe-H) significantly stimulated the absolute abundance average degree (15.6 vs 17.608 in the control), lower clustering coeffi­ of aerobic methanotrophs during the flooding stage (p < 0.05) by about cient (0.518 vs 0.633 in the control) and lower network density (0.166 62% and 68%, respectively. Moreover, we found soil salinity can induce vs 0.183 in the control) (Table 3). an increase in pmoA gene abundance in the Fe(III) fertilised treatment, Additionally, the co-occurrence networks of methanogens and because different methanotrophic genera may show distinct degrees of methanotrophs were also analysed (Fig. 7b). Although in both 7 Y. Zhang et al. 8 Fig. 6. Correlations between the CH4 emissions, the relative abundance of dominant methanogens (a) and methanotrophs (b), and soil physicochemical properties in the rice paddy soil shown as coloured covariance matrices. From porcelain white to navy blue, darker boxes represent stronger correlation coefficients; lighter boxes represent lower correlations. * p < 0.05, ** p < 0.01. Geoderma 431 (2023) 116375 Y. Zhang et al. Geoderma 431 (2023) 116375 Fig. 7. Microbial co-occurrence networks of the control and Fe-H treatments. (A) Integrated microbial co-occurrence networks across methanogens and meth­ anotrophs. Network construction is based on the proportion of OUT datasets. Nodes represent individual OTUs, and edges represent significant correlations. Network nodes indicating OTUs are depicted as spheres (red, mcrA; green, pmoA), and edges are coloured by interaction types (co-presence in red; mutual exclusion in black). The node size is proportional to the average degree of each taxon. (B) Separate microbial co-occurrence networks within methanogens and methanotrophs, respectively. Features of nodes and edges are the same as described in panel A. resistance to salt stress (Osudar et al., 2017). In our study, Type II the drainage stage (Fig. 3b). Thus, the superiority of Type II methano­ methanotrophs (e.g., Methylocystis) were more abundant and active, trophs over Type I methanotrophs at low-methane concentrations may while Type I methanotrophs (e.g., Methylomonas) were relatively sup­ also be a selective advantage in rice paddies, where growth is periodi­ pressed by Fe(III) fertiliser addition (Fig. 3b). Type II methanotrophs cally restricted because of fluctuations in the methane supply (Baani and exhibit lower methane yields than Type I methanotrophs, a fact attrib­ Liesack, 2008). uted to a serine pathway (Kalyuzhnaya et al., 2015). Han et al (2017) In contrast to methanotrophs being affected by soil salinity, metha­ revealed that Type II methanotrophs are resistant to relatively low salt nogens can be affected by Fe(III) oxides more directly. In our study we stress. Moreover, when the available substrate level is limited, Type II divided methanogenic genera into two groups according to their re­ methanotrophs can alter their survival strategy (Ho et al., 2013). sponses to Fe(III) fertiliser application (Fig. 4a). Among the positive Importantly, Type II methanotrophs, such as Methylocystis, strain SB2 response genera were Methanothrix, whose relative abundance increased may contain two enzymes with low and high levels of affinity for in Fe(III) fertilised treatments. Importantly, this increase can be attrib­ adapting to various CH4 concentrations, and consume ethanol and ac­ uted to the direct interspecies electron transfer (DIET) (Rotaru et al., etate for growth (Im et al., 2011). In this study, the relative abundance of 2014), and Fe(III) oxides may serve as electron conduits to stimulate Methylocystis was greater in the treatments than in the control. More­ DIET (Cruz Viggi et al., 2014). However, an increase in the relative over, the relative abundance of Methylocystis increased by 33% and 16% abundance of Methanothrix indicates a decrease in other methanogenic in Fe-M and Fe-H treatments, respectively, compared to the control in genera that may have a stronger capacity for CH4 production (e.g., 9 Y. Zhang et al. Geoderma 431 (2023) 116375 Table 3 addition not only at the genus level but also at the OTU level, and this Topological parameters of the integrated microbial co-occurrence networks in response was further reflected by microbial co-occurrence networks. the control and Fe-H treatments. Network modelling allows incorporating the complex relationships Control Fe-H among microbes into an integrated pattern (Bascompte, 2007). There­ Nodes fore, we constructed a microbial co-occurrence network to evaluate the Number of nodes 97 95 effects of Fe(III) fertilisation on paddy soils and to explore the potential Percentage of mcrA nodes (%) 49.48 51.58 relationships among specific taxa (Fig. 7). Comparing the integrated co- Percentage of pmoA nodes (%) 50.52 48.42 occurrence networks between the control and Fe-H treatments showed Average degree 17.608 15.6 that Fe(III) fertiliser application reduced the network complexity Edges Number of edges 854 741 contrast of the control, which contained fewer nodes and edges Percentage of mcrA-mcrA edges (%) 38.52 39.00 (Table 3). The topological features of the control, such as average de­ Percentage of pmoA-pmoA edges (%) 16.86 17.54 gree, clustering coefficient, and network density, indicate a more con­ Percentage of mcrA-pmoA edges (%) 44.61 43.45 nected and robust network. This result suggests that the methanogenic Positive edges (%) 64.64 79.76 Negative edges (%) 35.36 20.24 and methanotrophic interactomes in the control are characterized by Average path distance (GD) 2.367 2.167 relatively higher metabolic exchange and competition among commu­ Network nity members, which likely increased their co-occurrence (Bissett et al., Clustering coefficient 0.633 0.518 2013; Faust and Raes, 2012). On the other hand, the average path dis­ Network density 0.183 0.166 tance in the Fe-H treatment is smaller than that in the control, which means faster communication between methanogens and methano­ trophs. However, in this case, local perturbations reached the whole Table 4 network quickly, possibly altering the system (Kitano, 2004; Zhou et al., Topological parameters of the separate microbial co-occurrence networks in the 2010). Thus, our results indicate that adding Fe(III) fertiliser can control and Fe-H treatment. decrease the co-occurrence of methanogens and methanotrophs, and Methanogens Methanotrophs methanogenic and methanotrophic communities can respond more Control Fe-H Control Fe-H quickly to environmental changes. By comparing the separate methanogenic and methanotrophic co- Nodes Number of nodes 48 49 47 46 occurrence networks between the control and Fe-H treatments, we Average degree 13.708 11.796 6.128 5.652 found that the network of methanogens and methanotrophs in the Fe-H Edges treatment contains fewer edges than that in the control. This result Number of edges 329 289 144 130 further proves that Fe(III) fertilisation can decrease the complexity of Percentage of positive edges (%) 66.87 83.74 63.89 70.77 Percentage of negative edges (%) 33.13 16.26 36.11 29.23 both methanogenic and methanotrophic networks. Methanocella and Average path distance (GD) 2.265 1.976 2.884 2.89 Methylocystis, belonging to methanogens and methanotrophs, respec­ Network tively are the network hubs in the control, while in the Fe-H treatment, Clustering coefficient 0.666 0.568 0.603 0.541 Methanothrix dominates the methanogenic network (Table S3). More­ Network density 0.292 0.246 0.133 0.126 over, the fewer edges in the Fe-H treatment (6.34%) compared to the control (12.76%) are mostly due to the lower proportion of mutual Methanocella and Methanobacterium). Methanocella are hydro­ exclusion among methanogens. This difference might reflect eased genotrophic methanogens known as rice cluster I and play a key role in competition among methanogens after Fe(III) fertilisation. According to CH4 production in rice soil (Lu and Conrad, 2005). Many H2-utilising Bond and Lovley (2002), methanogenesis can be inhibited because of methanogens can transfer electrons to poorly crystalline Fe(III) oxide, the diversion of electrons from CH4 production to Fe3+ reduction. Thus, and this can be one explanation for the inhibition of methanogenesis in it is possible that Fe, as a density- independent factor, competes with Fe(III)-rich ecosystems (Bond and Lovley, 2002; van Bodegom et al., methanogens for electrons, and then some special methanogens may 2004). In Fig. 4b, the steeper slope of the regression line of negatively lose their competitiveness in the system. responsive genera implies that the mitigation of CH4 emissions could be This study attempted to build a linkage between the activity, abun­ due to a decrease in the CH4 production efficiency of the predominant dance and composition of methanogens and methanotrophs and sea­ methanogens caused by Fe(III) fertiliser addition. Therefore, Fe(III) sonal CH4 emissions in response to Fe(III) fertiliser addition in a rice fertiliser-decreased CH4 emissions can also be achieved through the paddy in southeast China. However, although we studied the abundance change in methanogenic community composition. and composition of methanogens and methanotrophs and further Overall, according to our results, Methanothrix became the most explored the relationship between the CH4 fluxes and community obviously changed genus in the paddy soils and grew better in Fe-rich composition, the interaction between methanogens and methanotrophs environments. However, in this study, adding Fe(III) fertiliser remains unclear. In future studies, dominant genera and specific primers decreased the CH4 production efficiency from predominant methanogen of methanogens and methanotrophs should be detected after Illumina genera (e.g. Methanocella and Methanobacterium). This finding is MiSeq sequencing, to further quantify the absolute abundance and make consistent with the results of MPPs, showing that Fe(III) amendment community composition analysis more convincing. Importantly, in our suppressed the soil CH4 production capacity. On the other hand, the study, microbial co-occurrence networks were constructed to evaluate relative abundance of Methylocystis (Type II) was significantly and methanogens and methanotrophs’ response to Fe(III) fertilisation at the negatively related to CH4 fluxes (Fig. 6b). Methylomonas (Type I) varied OTU level. However, to explore and clarify the potential connection by with treatments and is significantly negatively correlated to MOPs, analysing microbial co-occurrence networks, more data on the compo­ showing that Type I methanotrophs’ activity was inhibited by Fe(III) sition of methanogen and methanotroph communities should be fertiliser addition. collected. Moreover, regarding the connection between methanogens Taken together, these results show the methanotroph community and methanotrophs, confirmatory experiments such as incubating plays an important role in CH4 oxidation. After Fe(III) fertiliser addition, dominant genera belonging to methanogenic and methanotrophic Type II methanotrophs were more adaptive and active, while Type I communities should be conducted to find the potential relationship methanotrophs were relatively suppressed. between them. Methanogens and methanotrophs responded to Fe(III) fertiliser 10 Y. Zhang et al. Geoderma 431 (2023) 116375 5. Conclusion Bowman, J.P., Sly, L.T., Nichols, P.D., Hayward, A.C., 1994. Erratum: Revised taxonomy of the methanotrophs: Description of Methylobacter gen. nov., emendation of Methylococcus, validation of Methylosinus and Methylocystis species, and a This study showed that the application of Fe(III) fertiliser had no proposal that the family Methylococcaceae includes only the group I methanotro. adverse effect on rice yield but significantly decreased CH4 emissions. Int. J. Syst. Bacteriol. 44, 375. https://doi.org/10.1099/00207713-44-2-375. Applying Fe(III) fertiliser significantly decreased the abundance and Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman, F.D., Costello, E.K., Fierer, N., Peña, A.G., Goodrich, J.K., Gordon, J.I., Huttley, G.A., Kelley, S.T., activity of methanogens while increasing the populations and activity of Knights, D., Koenig, J.E., Ley, R.E., Lozupone, C.A., Mcdonald, D., Muegge, B.D., methanotrophs. Moreover, where Fe(III) changed the composition of Pirrung, M., Reeder, J., Sevinsky, J.R., Turnbaugh, P.J., Walters, W.A., Widmann, J., methanogenic archaea and methanotrophic bacteria, there was a shift Yatsunenko, T., Zaneveld, J., Knight, R., 2010. QIIME allows analysis of high- throughput community sequencing data. Nat. Methods 7, 335–336. https://doi.org/ from Type I methanotrophs to Type II methanotrophs in their relative 10.1038/nmeth.f.303. abundance after Fe(III) fertiliser application. Suppression of CH4 emis­ Chen, Z., Tang, Y.T., Yao, A.J., Cao, J., Wu, Z.H., Peng, Z.R., Wang, S.Z., Xiao, S., sions from rice paddies was associated with the conversion between Baker, A.J.M., Qiu, R.L., 2017. Mitigation of Cd accumulation in paddy rice (Oryza sativa L.) by Fe fertilization. Environ. Pollut. 231, 549–559. https://doi.org/ Fe3+ and Fe2+, and soil salinity was attributed to Fe(III) fertilisation. The 10.1016/j.envpol.2017.08.055. application of Fe(III) fertiliser mitigated CH4 emissions by decreasing Costello, A.M., Lidstrom, M.E., 1999. Molecular characterization of functional and CH4 production from predominant methanogen genera and stimulating phylogenetic genes from natural populations of methanotrophs in lake sediments. Appl. Environ. Microbiol. 65, 5066–5074. https://doi.org/10.1128/ methane oxidation potentials. Our results suggest that a less complex aem.65.11.5066-5074.1999. microbial network structure within the Fe-H treatment can be associated Cruz Viggi, C., Rossetti, S., Fazi, S., Paiano, P., Majone, M., Aulenta, F., 2014. Magnetite with additional Fe(III) fertiliser, possibly reducing competition among particles triggering a faster and more robust syntrophic pathway of methanogenic propionate degradation. Environ. Sci. Tech. 48, 7536–7543. https://doi.org/ methanogens. 10.1021/es5016789. Degelmann, D.M., Borken, W., Drake, H.L., Kolb, S., 2010. Different atmospheric methane-oxidizing communities in european beech and norway spruce soils. Appl. Declaration of Competing Interest Environ. Microbiol. 76, 3228–3235. https://doi.org/10.1128/AEM.02730-09. Edgar, R.C., Haas, B.J., Clemente, J.C., Quince, C., Knight, R., 2011. UCHIME improves sensitivity and speed of chimera detection. Bioinformatics 27, 2194–2200. https:// The authors declare that they have no known competing financial doi.org/10.1093/bioinformatics/btr381. interests or personal relationships that could have appeared to influence Fan, L., Dippold, M.A., Ge, T., Wu, J., Thiel, V., Kuzyakov, Y., Dorodnikov, M., 2020. the work reported in this paper. Anaerobic oxidation of methane in paddy soil: Role of electron acceptors and fertilization in mitigating CH4 fluxes. Soil Biol. Biochem. 141, 107685 https://doi. org/10.1016/j.soilbio.2019.107685. Data availability Fan, L., Schneider, D., Dippold, M.A., Poehlein, A., Wu, W., Gui, H., Ge, T., Wu, J., Thiel, V., Kuzyakov, Y., Dorodnikov, M., 2021. Active metabolic pathways of anaerobic methane oxidation in paddy soils. Soil Biol. Biochem. 156, 108215 Data will be made available on request. https://doi.org/10.1016/j.soilbio.2021.108215. Faust, K., Raes, J., 2012. Microbial interactions: From networks to models. Nat. Rev. Acknowledgments Microbiol. 10, 538–550. https://doi.org/10.1038/nrmicro2832. Faust, K., Raes, J., 2016. CoNet app: Inference of biological association networks using Cytoscape [version 1; referees: 2 approved with reservations]. F1000Res. 5, 1–17. This work was supported by the Jiangsu Agriculture Science and https://doi.org/10.12688/F1000RESEARCH.9050.1. Technology Innovation Fund (JASTIF, CX(21)3007), National Natural Fredrickson, J.K., Gorby, Y.A., 1996. Environmental processes mediated by iron-reducing bacteria. Curr. Opin. Biotechnol. 7, 287–294. https://doi.org/10.1016/S0958-1669 Science Foundation of China (NSFC, 41877093, 41771323 and (96)80032-2. 41907072), and PADA. JS and JP were funded by Grant TED2021- Gao, D., Sheng, R., Whiteley, A.S., Moreira-Grez, B., Qin, H., Zhang, W., Zhan, Y., 132627B-I00 from MCIN, AEI/10.13039/501100011033 and the Euro­ Wei, W., 2020. Effect of phosphorus amendments on rice rhizospheric methanogens and methanotrophs in a phosphorus deficient soil. Geoderma 368, 114312. https:// pean Union NextGenerationEU/PRTR. doi.org/10.1016/j.geoderma.2020.114312. Gross, K., 2008. Positive interactions among competitors can produce species-rich communities. Ecol. Lett. 11, 929–936. https://doi.org/10.1111/j.1461- Appendix A. Supplementary data 0248.2008.01204.x. Gwon, H.S., Khan, M.I., Alam, M.A., Das, S., Kim, P.J., 2018. Environmental risk Supplementary data to this article can be found online at https://doi. assessment of steel-making slags and the potential use of LD slag in mitigating org/10.1016/j.geoderma.2023.116375. methane emissions and the grain arsenic level in rice (Oryza sativa L.). J. Hazard. Mater. 353, 236–243. https://doi.org/10.1016/j.jhazmat.2018.04.023. Han, D., Link, H., Liesack, W., 2017. Response of Methylocystis sp. strain SC2 to salt References stress: Physiology, global transcriptome, and amino acid profiles. Appl. Environ. Microbiol. 83 https://doi.org/10.1128/AEM.00866-17. Hanson, R.S., Hanson, T.E., 1996. Methanotrophic bacteria. Microbiol. Rev. 60, Ali, M.A., Lee, C.H., Kim, P.J., 2008. Effect of silicate fertilizer on reducing methane 439–471. https://doi.org/10.1128/mmbr.60.2.439-471.1996. emission during rice cultivation. Biol. Fertil. Soils 44, 597–604. https://doi.org/ He, W., Shohag, M.J.I., Wei, Y., Feng, Y., Yang, X., 2013. Iron concentration, 10.1007/s00374-007-0243-5. bioavailability, and nutritional quality of polished rice affected by different forms of Arao, T., Kawasaki, A., Baba, K., Mori, S., Matsumoto, S., 2009. Effects of water foliar iron fertilizer. Food Chem. 141, 4122–4126. https://doi.org/10.1016/j. management on cadmium and arsenic accumulation and dimethylarsinic acid foodchem.2013.07.005. concentrations in Japanese rice. Environ. Sci. Tech. 43, 9361–9367. https://doi.org/ Heyer, J., Galschenko, V.F., Dunfield, P.F., 2002. Molecular phylogeny of type II 10.1021/es9022738. methane-oxidizing bacteria isolated from various environments. Microbiology 148, Baani, M., Liesack, W., 2008. Two isozymes of particulate methane monooxygenase with 2831–2846. https://doi.org/10.1099/00221287-148-9-2831. different methane oxidation kinetics are found in Methylocystis sp. strain SC2. PNAS Ho, A., Kerckhof, F.M., Luke, C., Reim, A., Krause, S., Boon, N., Bodelier, P.L.E., 2013. 105, 10203–10208. https://doi.org/10.1073/pnas.0702643105. Conceptualizing functional traits and ecological characteristics of methane-oxidizing Bascompte, J., 2007. Networks in ecology. Basic Appl. Ecol. 8 https://doi.org/10.1016/j. bacteria as life strategies. Environ. Microbiol. Rep. 10.1111/j.1758- baae.2007.06.003. 2229.2012.00370.x. Beal, E.J., House, C.H., Orphan, V.J., 2009. Manganese- and iron-dependent marine Ho, A., Lüke, C., Frenzel, P., 2011. Recovery of methanotrophs from disturbance: methane oxidation. Science (80-). 325, 184–187. 10.1126/science.1169984. Population dynamics, evenness and functioning. ISME J. 5, 750–758. https://doi. Bissett, A., Brown, M.V., Siciliano, S.D., Thrall, P.H., 2013. Microbial community org/10.1038/ismej.2010.163. responses to anthropogenically induced environmental change: Towards a systems Holmes, A.J., Costello, A., Lidstrom, M.E., Murrell, J.C., 1995. Evidence that participate approach. Ecol. Lett. 16 https://doi.org/10.1111/ele.12109. methane monooxygenase and ammonia monooxygenase may be evolutionarily Blazewicz, S.J., Petersen, D.G., Waldrop, M.P., Firestone, M.K., 2012. Anaerobic related. FEMS Microbiol. Lett. 132, 203–208. https://doi.org/10.1016/0378-1097 oxidation of methane in tropical and boreal soils: Ecological significance in (95)00311-R. terrestrial methane cycling. J. Geophys. Res. Biogeosciences 117, 1–9. https://doi. Im, J., Lee, S.W., Yoon, S., Dispirito, A.A., Semrau, J.D., 2011. Characterization of a org/10.1029/2011JG001864. novel facultative Methylocystis species capable of growth on methane, acetate and Bodegom, P.M.V., Stams, A.J.M., 1999. Effects of Alternative Electron Acceptors and ethanol. Environ. Microbiol. Rep. 3, 174–181. https://doi.org/10.1111/j.1758- Temperature on. Chemosphere 39, 167–182. 2229.2010.00204.x. Bond, D.R., Lovley, D.R., 2002. Reduction of Fe(III) oxide by methanogens in the presence and absence of extracellular quinones. Environ. Microbiol. 4, 115–124. https://doi.org/10.1046/j.1462-2920.2002.00279.x. 11 Y. Zhang et al. Geoderma 431 (2023) 116375 IPCC, 2013. Climate Change 2013-The Physical Science Basis. Contribution of Working Steinberg, L.M., Regan, J.M., 2008. Phylogenetic comparison of the methanogenic Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate communities from an acidic, oligotrophic fen and an anaerobic digester treating Change. Cambridge University Press, Cambridge, New York. municipal wastewater sludge. Appl. Environ. Microbiol. 74, 6663–6671. https://doi. Kalyuzhnaya, M.G., Puri, A.W., Lidstrom, M.E., 2015. Metabolic engineering in org/10.1128/AEM.00553-08. methanotrophic bacteria. Metab. Eng. 29, 142–152. https://doi.org/10.1016/j. Stookey, L.L., 1970. Ferrozine–a new spectrophotometric reagent for iron. Anal. Chem. ymben.2015.03.010. 42 (7), 779–781. Karimi, B., Maron, P.A., Chemidlin-Prevost Boure, N., Bernard, N., Gilbert, D., Sun, H., Zhou, S., Fu, Z., Chen, G., Zou, G., Song, X., 2016. A two-year field measurement Ranjard, L., 2017. Microbial diversity and ecological networks as indicators of of methane and nitrous oxide fluxes from rice paddies under contrasting climate environmental quality. Environ. Chem. Lett. 15, 265–281. https://doi.org/10.1007/ conditions. Sci. Rep. 6, 1–11. https://doi.org/10.1038/srep28255. s10311-017-0614-6. Tokida, T., Fumoto, T., Cheng, W., Matsunami, T., Adachi, M., Katayanagi, N., Kaupper, T., Mendes, L.W., Lee, H.J., Mo, Y., Poehlein, A., Jia, Z., Horn, M.A., Ho, A., Matsushima, M., Okawara, Y., Nakamura, H., Okada, M., Sameshima, R., 2021. When the going gets tough: Emergence of a complex methane-driven Hasegawa, T., 2010. Effects of free-air CO2 enrichment (FACE) and soil warming on interaction network during recovery from desiccation-rewetting. Soil Biol. Biochem. CH4 emission from a rice paddy field: Impact assessment and stoichiometric 153, 108109 https://doi.org/10.1016/j.soilbio.2020.108109. evaluation. Biogeosciences 7, 2639–2653. https://doi.org/10.5194/bg-7-2639- Kim, G.W., Ho, A., Kim, P.J., Kim, S.Y., 2016. Stimulation of methane oxidation potential 2010. and effects on vegetation growth by bottom ash addition in a landfill final Van Bodegom, P.M., Scholten, J.C.M., Stams, A.J.M., 2004. Direct inhibition of evapotranspiration cover. Waste Manage. 55, 306–312. https://doi.org/10.1016/j. methanogenesis by ferric iron. FEMS Microbiol. Ecol. 49, 261–268. https://doi.org/ wasman.2016.03.058. 10.1016/j.femsec.2004.03.017. Kitano, H., 2004. Biological robustness. Nat. Rev. Genet. https://doi.org/10.1038/ Wang, C., Jin, Y., Ji, C., Zhang, N., Song, M., Kong, D., Liu, S., Zhang, X., Liu, X., Zou, J., nrg1471. Li, S., Pan, G., 2018. An additive effect of elevated atmospheric CO2 and rising Kong, D., Li, S., Jin, Y., Wu, S., Chen, J., Hu, T., Wang, H., Liu, S., Zou, J., 2019. Linking temperature on methane emissions related to methanogenic community in rice methane emissions to methanogenic and methanotrophic communities under paddies. Agr. Ecosyst. Environ. 257, 165–174. https://doi.org/10.1016/j. different fertilization strategies in rice paddies. Geoderma 347, 233–243. https:// agee.2018.02.003. doi.org/10.1016/j.geoderma.2019.04.008. Wasmund, K., Kurtböke, D.I., Burns, K.A., Bourne, D.G., 2009. Microbial diversity in Li, S., Song, L., Gao, X., Jin, Y., Liu, S., Shen, Q., Zou, J., 2017. Microbial abundances sediments associated with a shallow methane seep in the tropical Timor Sea of predict methane and nitrous oxide fluxes from a windrow composting system. Front. Australia reveals a novel aerobic methanotroph diversity. FEMS Microbiol. Ecol. 68, Microbiol. 8 https://doi.org/10.3389/fmicb.2017.00409. 142–151. https://doi.org/10.1111/j.1574-6941.2009.00667.x. Lovley, D.R., 1987. Organic matter mineralization with the reduction of ferric iron: A Weber, K.A., Achenbach, L.A., Coates, J.D., 2006. Microorganisms pumping iron: review. Geomicrobiol. J. 5, 375–399. https://doi.org/10.1080/ Anaerobic microbial iron oxidation and reduction. Nat. Rev. Microbiol. 4, 752–764. 01490458709385975. https://doi.org/10.1038/nrmicro1490. Lovley, D.R., Phillips, E.J.P., 1987. Competitive Mechanisms for Inhibition of Sulfate Xu, B., Wang, F., Zhang, Q., Lan, Q., Liu, C., Guo, X., Cai, Q., Chen, Y., Wang, G., Ding, J., Reduction and Methane Production in the Zone of Ferric Iron Reduction in 2018. Influence of iron plaque on the uptake and accumulation of chromium by rice Sediments. Appl. Environ. Microbiol. 53, 2636–2641. https://doi.org/10.1128/ (Oryza sativa L.) seedlings: Insights from hydroponic and soil cultivation. Ecotoxicol. aem.53.11.2636-2641.1987. Environ. Saf. 162, 51–58. https://doi.org/10.1016/j.ecoenv.2018.06.063. Lu, Y., Conrad, R., 2005. In situ stable isotope probing of methanogenic Archaea in the Yang, S., Liebner, S., Alawi, M., Ebenhöh, O., Wagner, D., 2014. Taxonomic database and rice rhizosphere. Science (80-) 309, 1088–1090. https://doi.org/10.1126/ cut-off value for processing mcrA gene 454 pyrosequencing data by MOTHUR. science.1113435. J. Microbiol. Methods 103, 3–5. https://doi.org/10.1016/j.mimet.2014.05.006. Osudar, R., Klings, K.W., Wagner, D., Bussmann, I., 2017. Effect of salinity on microbial Yu, K., Fang, X., Zhang, Y., Miao, Y., Liu, S., Zou, J., 2021. Low greenhouse gases methane oxidation in freshwater and marine environments. Aquat. Microb. Ecol. 80, emissions associated with high nitrogen use efficiency under optimized fertilization 181–192. https://doi.org/10.3354/ame01845. regimes in double-rice cropping systems. Appl. Soil Ecol. 160, 103846 https://doi. Pereyra, L.P., Hiibel, S.R., Prieto Riquelme, M.V., Reardon, K.F., Pruden, A., 2010. org/10.1016/j.apsoil.2020.103846. Detection and quantification of functional genes of cellulosedegrading, fermentative, Zhang, J., Dong, H., Liu, D., Fischer, T.B., Wang, S., Huang, L., 2012. Microbial reduction and sulfate-reducing bacteria and methanogenic archaea. Appl. Environ. Microbiol. of Fe(III) in illite-smectite minerals by methanogen Methanosarcina mazei. Chem. 76, 2192–2202. https://doi.org/10.1128/AEM.01285-09. Geol. 292–293, 35–44. https://doi.org/10.1016/j.chemgeo.2011.11.003. Qin, Y., Liu, S., Guo, Y., Liu, Q., Zou, J., 2010. Methane and nitrous oxide emissions from Zhou, J., Deng, Y., Luo, F., He, Z., Tu, Q., Zhi, X., 2010. Functional molecular ecological organic and conventional rice cropping systems in Southeast China. Biol. Fertil. Soils networks. MBio 1. https://doi.org/10.1128/mBio.00169-10. 46, 825–834. https://doi.org/10.1007/s00374-010-0493-5. Zhou, P., Pan, G.X., Zhang X.H., Song, X.Y., Li, L., 2009. An approach of C sequestration Reumer, M., Harnisz, M., Lee, H.J., Reim, A., Grunert, O., Putkinen, A., Fritze, H., and metigation inspired by ferric oxide from paddy soil. Journal of Agro- Bodelier, P.L.E., Ho, A., 2018. Impact of peat mining and restoration on methane Environment Science 28, 2590-2595 (in Chinese with English Abstract). doi: turnover potential and methane-cycling microorganisms in a northern bog. Appl. 10.3321/j.issn:1672-2043.2009.12.024. Environ. Microbiol. 84 https://doi.org/10.1128/AEM.02218-17. Zhou, G., Gao, S., Xu, C., Dou, F., Shimizu, K. yoshi, Cao, W., 2020. Rational utilization of Rotaru, A.E., Shrestha, P.M., Liu, F., Shrestha, M., Shrestha, D., Embree, M., Zengler, K., leguminous green manure to mitigate methane emissions by influencing Wardman, C., Nevin, K.P., Lovley, D.R., 2014. A new model for electron flow during methanogenic and methanotrophic communities. Geoderma 361, 114071. https:// anaerobic digestion: Direct interspecies electron transfer to Methanosaeta for the doi.org/10.1016/j.geoderma.2019.114071. reduction of carbon dioxide to methane. Energy Environ. Sci. 7, 408–415. https:// Zou, J., Huang, Y., Xi, Z., 2004. Carbon Dioxide, Methane, and Nitrous Oxide Emissions doi.org/10.1039/c3ee42189a. from. Advances 21, 691–698. Singh, A., Singh, R.S., Upadhyay, S.N., Joshi, C.G., Tripathi, A.K., Dubey, S.K., 2012. Zou, J., Huang, Y., Jiang, J., Zheng, X., Sass, R.L., 2005. A 3-year field measurement of Community structure of methanogenic archaea and methane production associated methane and nitrous oxide emissions from rice paddies in China: Effects of water with compost-treated tropical rice-field soil. FEMS Microbiol. Ecol. 82, 118–134. regime, crop residue, and fertilizer application. Global Biogeochem. Cycles 19, 1–9. https://doi.org/10.1111/j.1574-6941.2012.01411.x. https://doi.org/10.1029/2004GB002401. Sonoki, T., Furukawa, T., Jindo, K., Suto, K., Aoyama, M., Sánchez-Monedero, M.Á., Zou, J., Liu, S., Qin, Y., Pan, G., Zhu, D., 2009. Sewage irrigation increased methane and 2013. Influence of biochar addition on methane metabolism during thermophilic nitrous oxide emissions from rice paddies in southeast China. Agr. Ecosyst. Environ. phase of composting. J. Basic Microbiol. 53, 617–621. https://doi.org/10.1002/ 129, 516–522. https://doi.org/10.1016/j.agee.2008.11.006. jobm.201200096. 12