Denitrification contributes to N2O emission in paddy soils

Denitrification is vital to nitrogen removal and N2O release in ecosystems; in this regard, paddy soils exhibit strong denitrifying ability. However, the underlying mechanism of N2O emission from denitrification in paddy soils is yet to be elucidated. In this study, the potential N2O emission rate, enzymatic activity for N2O production and reduction, gene abundance, and community composition during denitrification were investigated using the 15N isotope tracer technique combined with slurry incubation, enzymatic activity detection, quantitative polymerase chain reaction (qPCR), and metagenomic sequencing. Results of incubation experiments showed that the average potential N2O emission rates were 0.51 ± 0.20 μmol⋅N⋅kg–1⋅h–1, which constituted 2.16 ± 0.85% of the denitrification end-products. The enzymatic activity for N2O production was 2.77–8.94 times than that for N2O reduction, indicating an imbalance between N2O production and reduction. The gene abundance ratio of nir to nosZ from qPCR results further supported the imbalance. Results of metagenomic analysis showed that, although Proteobacteria was the common phylum for denitrification genes, other dominant community compositions varied for different denitrification genes. Gammaproteobacteria and other phyla containing the norB gene without nosZ genes, including Actinobacteria, Planctomycetes, Desulfobacterota, Cyanobacteria, Acidobacteria, Bacteroidetes, and Myxococcus, may contribute to N2O emission from paddy soils. Our results suggest that denitrification is highly modular, with different microbial communities collaborating to complete the denitrification process, thus resulting in an emission estimation of 13.67 ± 5.44 g N2O⋅m–2⋅yr–1 in surface paddy soils.


Introduction
Nitrous oxide (N 2 O) is not only a significant ozone-depleting substance (Ravishankara et al., 2009), but is also a well-known greenhouse gas with strong radiative forcing effects (Zumft and Kroneck, 2007). Since the industrial revolution, the concentration of N 2 O has increased at an annual rate of 0.25%, and the atmospheric N 2 O concentration has reached 331 ppb (Solomon et al., 2007). Moreover, the greenhouse effect of N 2 O is 310 times higher than that of the equivalent carbon dioxide (Lashof and Ahuja, 1990). Thus, N 2 O has received increasing attention owing to the environmental problems that it may cause.
Agricultural fields that receive substantial amounts of nitrogen fertilizers are typically known as N 2 O emission hotspots (Syakila and Kroeze, 2011). According to predictions, by 2030, agricultural soils will become the primary source of N 2 O emission, contributing 59% of the total N 2 O emission released into the atmosphere (Hu et al., 2015). Although nitrification, denitrification, and nitrate dissimilation to ammonium can all generate N 2 O, denitrification is the primary pathway for N 2 O emission in terrestrial ecosystems (Sanford et al., 2012;Butterbach-Bahl et al., 2013;Harris et al., 2021). Furthermore, in the denitrification process, N 2 O exists as an intermediate product and can be both generated and consumed, leading to the emission of N 2 O being regulated by multiple functional genes (Zumft, 1997) During denitrification, NO 3 − is successively reduced to NO 2 − , NO, N 2 O, and finally to N 2 (Zumft, 1997). Diverse phylogenetic denitrifying bacteria contain different functional genes, including napA, nirS, nirK, norB, nosZ I, and nosZ II genes, which encode enzymes that complete the denitrification process (Zumft, 1997). The napA gene encodes NO 3 − reductase, which catalyzes the reduction of NO 3 − to NO 2 − (Arnoux et al., 2003). NO 2 − and N 2 O reductions are considered the rate-limiting steps in denitrification (Zumft, 1997). NO 2 − reduction is catalyzed by NO 2 − reductases, including the nirS gene-encoded copper-containing NO 2 − reductase and the nirK gene-encoded cytochrome cd1-containing NO 2 − reductase (Zumft, 1997). Copper-containing or cytochrome cd1-containing NO 2 − reductases are functionally identical but have different structures and catalytic sites, and generally do not coexist in one bacterial species (Coyne et al., 1989). The NO reductase (NOR), encoded by the norB gene, is responsible for the reduction of NO to N 2 O (Braker and Tiedje, 2003). The N 2 O reductase (NOS) catalyzes the reduction of N 2 O, converting the greenhouse gas N 2 O into relatively harmless N 2 , thereby reducing its contribution to the greenhouse effect. NOS, encoded by either nosZ I or nosZ II genes, can complete denitrification or N 2 O reduction only (Zumft and Kroneck, 2007). The nosZ I and nosZ II genes generally do not coexist in the same bacteria, except in Thauera linaloolentis 47Lol T (Semedo et al., 2020).
Nitrous oxide is an intermediate product of the denitrification process, and N 2 O emission is typically associated with enzymes encoded by functional genes and the bacteria that harbor these genes during denitrification (Black et al., 2019). Denitrification is a complex process affected by multiple factors, including environmental parameters, the microbial composition of functional genes, and key enzymatic activities (Groffman, 2012). Rich et al. (2003) showed the community composition as well as environmental factors lead to the variation in denitrification rates in meadow and forest soils. The analysis of denitrification genes abundance can establish a stronger correlation with potential N 2 O emissions (Morales et al., 2010). For instance, the ratio of nir to nosZ can partly determine the extent of N 2 O emission in soils (Domeignoz-Horta et al., 2015) and lakes (Saarenheimo et al., 2015). However, the comprehensive mechanisms of denitrification and N 2 O emissions are yet to be elucidated.
In this study, we first investigated the dissolved N 2 O concentration in paddy water and calculated the N 2 O flux at the water-air exchange to provide background information regarding N 2 O emissions in paddy fields. Focusing on paddy soils, we performed 15 N isotope tracer experiments to determine the potential N 2 O emission rate and calculate the end-product ratio, i.e., N 2 O/(N 2 +N 2 O). The enzymatic activities of the NOR and NOS were measured using ELISA assay kits. Subsequently, qPCR and metagenomic analysis were performed to determine the abundance and community composition of key functional genes, respectively. The aim of this study is to demonstrate the contribution and underlying mechanism of denitrification to N 2 O emission in paddy soils and provide a more comprehensive understanding of the N 2 O emission process in paddy soils.

Sample acquisition and physiochemical property determination
The soil to be tested was obtained from two paddy fields (22 • 55 16 N, 113 • 29 24 E; 22 • 54 24 N, 113 • 29 36 E) in May 2021 from Guangdong Province, China (Figure 1). Both paddy fields had been planted with rice for many years and the paddy fields were under water-logged conditions at the time of sampling. The dissolved oxygen (DO) in the paddy water was first determined using a portable multifunctional parameter meter equipped with a DO probe (HQ40D, HACH, Loveland, CO, USA). Quickly insert a 12.5 mL vial (Exetainer, Labco Limited, Lampeter, UK) below the surface of the paddy water until the vial was overflowing and there were no air bubbles inside. The cap was immediately sealed and 200 µL of saturated ZnCl 2 solution was added to stop microbial activities. The sample was in triplicate and stored at −4 • C. The dissolved N 2 O concentration of in situ water was determined within 24 h. The pH level was determined using a pH meter (Mettler Toledo S220, Greifensee, Switzerland). Salinity and temperature were measured using salinometers (ATAGO, Tokyo, Japan) and geothermometers (HG04-SYQX-2, Beijing, China), respectively.
Surface soil (0-10 cm) was sampled using a sterilized shovel, and each soil sample was mixed in triplicate. Paddy soil was divided into three parts: one for measuring potential rates of N 2 and N 2 O emissions, which was stored at 4 • C; another for DNA extraction and enzymatic activity measurement, which was kept at −80 • C; and the last part was stored at 4 • C for pH determination and analysis of dissolved inorganic nitrogen (DIN), including NO 3 − , NO 2 − and NH 4 + . DIN in the paddy soil was extracted using 2 M KCl with a soil extraction ratio of 1:5 (wt./vol.) before measurement (Bao, 2000). The DIN content was determined using the spectrophotometric method described by Wu et al. (2016) and Guan et al. (2017).

Dissolved N 2 O analysis
The concentration of N 2 O was measured using static headspace gas chromatography (Xu et al., 2005). The water in the vial was replaced with 5 mL of He to achieve headspace, agitated vigorously for 15 min, and stored in the dark to attain gas-liquid equilibrium.
Frontiers in Microbiology 02 frontiersin.org Location of Guangdong Province and sampling site of paddy fields.
A gas chromatograph equipped with an electron capture detector (GC-2014C, Shimadzu, Tokyo, Japan) was used to measure the concentration of N 2 O (C G ) in the headspace. The dissolved N 2 O concentration (C L ) before gas replacement and the N 2 O (i.e., the net increase in N 2 O) were calculated by the following equations (Lin et al., 2016): where K 0 is the equilibrium constant (Weiss and Price, 1980); R is the ideal gas constant; T is the temperature at equilibrium; V G and V L are the gas and liquid volumes after He replacement, respectively; and C N2Oeq is the N 2 O concentration in equilibrium with the atmospheric concentration calculated based on Weiss and Price (1980). N 2 O saturation was calculated as follows: The estimated N 2 O flux through the water-air interface was calculated as follows: where K denotes the gas change rate calculated based on the method of Borges (2004).

Determination of potential N 2 production rate and N 2 O emission rate
The procedures for slurry incubation were modified from Thamdrup and Dalsgaard (2002) and described in detail by Xiang et al. (2023). A mixture of fresh soil and water was pre-incubated in the dark for 72 h at a weight-to-volume ratio of 1:7. The soil slurry was then flushed with He to eliminate N 2 O and create an anaerobic environment before being transferred to vials using a syringe. The vials were separated into two groups, one group (three vials) was for measuring the concentration of remaining 14 NO 3 − such that F n (the proportion of 14 NO 3 − in the total NO 3 − pool after adding 15 NO 3 − ) can be calculated, and the other group (12 vials) was injected with 15 NO 3 − ( 15 N 99.6%) to the final concentration of 100 µmol·L −1 . The vials containing 15 NO 3 − were incubated for T 0 (T 0 = 0 h) and T 2 (T 2 = 2 h) at in situ temperature in darkness, respectively, and 200 µL of saturated ZnCl 2 solution was injected immediately.
The 29 N 2 amounts (D 29 ) and 30 N 2 amounts (P 30 ) in the vials above were measured via membrane inlet mass spectrometry (HPR40, Hiden, Warrington, UK), and the potential N 2 production rate (R N2 ) was calculated as follows (Xiao et al., 2018;Wu et al., 2021): The N 2 O concentration in each vial was determined via a procedure similar to that used for determining the N 2 O concentration in the water sample (Xu et al., 2005). The potential N 2 O emission rates (R N2O ) was calculated based on the dissolved N 2 O concentrations in the vials at T 0 and T 2 (i.e., C L0 and C L2 , respectively) (Xiang et al., 2023).
The ratio of denitrification end-products was calculated as follows:

Measurement of enzymatic activity of NOR and NOS
The enzymatic activities of NOR (NO→N 2 O) and NOS (N 2 O→N 2 ) in the soils were determined using microbial NOR and NOS ELISA kits (Yilaisa Biotechnology Co., Ltd., Jiangsu, China).
Frontiers in Microbiology 03 frontiersin.org Briefly, 1 g of fresh soil, standards, and HRP-labeled detection antibodies were added based on the manufacturer's protocol. A microplate reader (BioTek Elx800, Winooski, VT, USA) was used to measure the absorbance at 450 nm. Enzymatic activity was calculated from the absorbance based on the standard curve.

DNA extraction and qPCR
Soil DNA was extracted using a Fast DNA Spin Kit for Soil (MP Biomedical, Irvine, CA, USA) based on the manufacturer's protocol. The concentration and purity of the extracted DNA were verified using a NanoDrop spectrophotometer (Thermo Scientific, Wilmington, DE, USA) and agarose gel electrophoresis, respectively. Primer sets 515F/806R (Caporaso et al., 2011), cd3af/R3cd (Throback et al., 2004), F1aCu/R3Cu (Hallin and Lindgern, 1999), nosZ 2F/nosZ 2R (Henry et al., 2006), nosZ IIF/nosZ IIR (Jones et al., 2013) were used to quantify the abundance of bacterial 16S, nirS, nirK, nosZ I, and nosZ II genes, respectively. qPCR was performed using the iQ5 Real-Time PCR system (Bio-Rad, Hercules, CA, USA). The amplification procedures are present in Supplementary Table 1 and have been described previously (Xiang et al., 2022). Plasmids, samples, and negative controls were prepared in triplicate and quantified simultaneously. The specificity of the amplified products was verified via melting curve analysis and agarose gel electrophoresis. Results of the amplification efficiency (90-110%) and correlation coefficient (R 2 > 97%) are shown.

Metagenomic sequencing and gene annotation
The extracted DNA was sent to Genewiz (Suzhou, China) for library construction and shotgun sequencing. The raw sequencing reads of each sample exceeded 10 Gb and were trimmed using Trimmomatic (v 0.38) (Bolger et al., 2014) to obtain high-quality clean reads and ensure the accuracy of subsequent analysis. The clean reads were assembled to contigs using metaSPAdes (v 3.13.2) (Bankevich et al., 2012) with k-mer sizes of 21, 33, 55, 77, 99, and 127. The open reading frames (ORFs) of the contigs were predicted using Prodigal (v 2.6.3) (Hyatt et al., 2010) and were searched against KEGG (Aramaki et al., 2019) to obtain the corresponding KO numbers. The denitrification genes and corresponding KO numbers were as follows: napA, K02567; nirS, K15864; nirK, K00368; nosZ, K00376. The nosZ gene has only one KO number but includes nosZ I and nosZ II genes. We first obtained the protein sequence of nosZ gene and searched against the NCBI database. 1 The nosZ I and nosZ II genes were distinguished based on their top-10 hits. The protein sequences of the ORFs were annotated using Kraken 2 (v 2.0.8b) (Wood et al., 2019) and GTDB combined (Parks et al., 2018;Chaumeil et al., 2019). The normalized abundance of ORF for all samples, i.e., the expression levels measured on the transcripts per million (TPM) scale, were estimated using Salmon (v 1.1.0) (Patro et al., 2017). Raw metagenomic sequencing data were deposited in the NCBI under BioProject PRJNA957066.

Statistical analysis
Redundancy analysis (RDA) was performed using Canoco 5 (v 5.0) and linear regression analysis, and graphs were generated using Graphism (v 8.0). The spearman correlation analysis between physiochemical properties and the abundance of denitrification genes was conducted by SPSS (v 26.0). P-value less than 0.05 (p < 0.05) was considered statistically significant.

Physiochemical properties of paddy fields
The physiochemical properties of the paddy soils are shown in Table 1. The temperature at the time of sampling was relatively high, i.e., 31.3-36.7 • C. The DO of paddy water was low and fluctuated in the range of 4.89 ± 0.02 to 13.02 ± 0.20 mg·L −1 . The salinity value was two among all paddy water samples. The soil pH was weakly alkaline and its value ranged from 7.82 ± 0.03 to 9.34 ± 0.05. NH 4 + was the main existing form of soil DIN, followed by NO 3 − . The NH 4 + and NO 3 − concentrations varied from 3.64 ± 0.19 to 9.72 ± 0.03 and 0.16 ± 0.02 to 5.94 ± 0.31 mg·kg −1 , respectively. The NO 2 − concentrations were low in soil samples, with a minimum value of 0.38 ± 0.01 mg·kg −1 .

N 2 O concentration, N 2 O concentration, N 2 O saturation, and N 2 O flux of air-water exchange
Paddy water showed high N 2 O concentrations and saturations (Figure 2). The dissolved N 2 O concentration ranged from 123.65 ± 16.09 to 235.59 ± 5.59 nmol·L −1 , with minimum and maximum values indicated by samples S2 and S1, respectively (Figure 2A). Excluding dissolved N 2 O at the air-water equilibrium, the N 2 O concentration fluctuated between 117.56 ± 16.09 and 229.62 ± 5.59 nmol·L −1 (Figure 2B). The N 2 O dissolved in paddy water was supersaturated, with an average N 2 O saturation of 2588.13 ± 659.94% ( Figure 2C). The paddy fields were a net source of atmospheric N 2 O, and the maximum N 2 O flux from the paddy water was estimated to be between 367.28 ± 50.26 and 747.03 ± 18.19 µmol·m −2 ·d −1 (Figure 2D).

Enzymatic activities of NOR and NOS and enzymatic activity ratios of NOR to NOS
The enzymatic activities related to N 2 O production and N 2 O reduction were measured (Figure 4). The results showed that the activity of NOR was relatively high, ranging from 413.49 ± 23.84 to 829.52 ± 18.34 U·g −1 (Figure 4A). The activity of NOS fluctuated between 85.19 ± 0.32 and 148.92 ± 3.16 U·g −1 in the paddy soils ( Figure 4B). The activity of NOR was 2.77 to 9.42 times than that of NOS, indicating that the production rate of N 2 O was higher than the N 2 O reduction rate at the enzymatic activity level (Figure 4C).
The napA gene, catalyzing the reduction of NO 3 − to NO 2 − , was abundant in Proteobacteria. The mean abundance of Proteobacteria of the napA gene was 21.22 ± 15.38 TPM (Supplementary Table 3), which constituted 69.58 ± 14.78% of Potential N 2 production rates (A), potential N 2 O emission rates (B), ratios of denitrification end-products (C), and potential denitrification rates (D) of paddy soils. the napA gene community ( Figure 6A). Additionally, unclassified bacteria constituted 13.39 ± 8.06% of the napA gene sequences, and the napA gene distributed in Desulfobacterota and Myxococcus for some samples (Figure 6A).
The enzymes encoded by the nirS and nirK genes contributed to the reduction of NO 2 − to NO. Similar to the composition of the napA gene, a large proportion of the nirS gene appeared in Proteobacteria (83.33-95.39%), and a small proportion of unclassified bacteria (3.34-13.31%) was detected in all paddy soils. The abundances of Proteobacteria and unclassified bacteria in the nirS gene sequences ranged from 9.85 to 59.86 TPM and from 0.71 to 3.47 TPM, respectively (Supplementary Table 4). Additionally, Chloroflexi (0.14-1.89 TPM) and Actinobacteria (0.06-0.33 TPM) were observed in all samples ( Figure 6B). The nirK gene was mainly distributed in unclassified bacteria (17.23-34.27%), Proteobacteria (25.05-49.96%), Actinobacteria (11.82-20.80%), and Chloroflexi Abundance of key functional genes for denitrification, including nir (nirS and nirK) (A), nosZ (nosZ I and nosZ II) (B), and gene abundance ratios (C) of paddy soils.
In general, Proteobacteria was the most typical group among the denitrification genes, and the relative abundance of classes in Proteobacteria based on the denitrification genes was analyzed (Supplementary Figure 2). Gammaproteobacteria was the dominant Proteobacteria in napA (Supplementary Figure 2A), nirS (Supplementary Figure 2B), and norB genes (Supplementary Figure 2D). Proteobacteria containing nirK (Supplementary Figure 2C), nosZ I (Supplementary Figure 2E), and nosZ II genes (Supplementary Figure 2F) were mainly composed of Alphaproteobacteria and Gammaproteobacteria.
The results of RDA showed that pH of the paddy soils significantly affected the community composition of most denitrification genes, including napA, nirS, nirK, nosZ I, and nosZ II genes (p < 0.05) (Supplementary Figure 3). In addition, the community composition of the nirS gene was affected by the NO 3 − concentration (p < 0.05) (Supplementary Figure 3A).

Contribution of denitrification to N 2 O emission
Soil and water in paddy fields are closely related but rarely investigated simultaneously. In this study, we determined the N 2 O emission capacity of both paddy soil and paddy water. The potential N 2 O emission rates fluctuated between 0.26 ± 0.15 and 0.90 ± 0.20 µmol·N·kg −1 ·h −1 (Figure 3B), which was higher than those of soils from Hebei Province (0.029 µmol·N·kg −1 ·h −1 ) (Zhao et al., 2019) and similar to those of sediments from the Chongming Dongtan wetland (0.21-0.84 µmol·N·kg −1 ·h −1 ) (Gao et al., 2022). Based on the potential N 2 O emission rate in surface paddy soils, the potential N 2 O emission by denitrification reached 0.31 ± 0.12 mol N 2 O·m −2 ·yr −1 , generating an emission estimation of 13.67 ± 5.44 g N 2 O·m −2 ·yr −1 in surface paddy soils (estimation method supplied in Supplementary material).
The concentration of N 2 O in the paddy water was high, i.e., 123.54 to 235.59 nmol·L −1 (Figure 2A). The concentration of N 2 O in the paddy water was approximately 18.17 times that in Indian estuaries (Rao and Sarma, 2013) and 7.72-14.72 times that in the Jinshui River (Zhao and Zhang, 2021). Excessive application of nitrogen fertilizer in paddy fields can promote the release of a significant amount of N 2 O (Gupta et al., 2021). As expected, supersaturation was observed in the paddy water, and its average N 2 O saturation (2588.13 ± 659.94%) was significantly higher than that of water from the Jinshui and Qi Rivers (Zhao and Zhang, 2021) and from the Shanghai River (770%) (Yu et al., 2013). The mean N 2 O flux (489.59 ± 147.28 µmol·m −2 ·d −1 ) in the paddy water were significantly higher than that in the upper Pearl River estuarine water (313 ± 150 µmol·m −2 ·d −1 ) (Lin et al., 2016) and the Shanghai city river (140 µmol·m −2 ·d −1 ) (Yu et al., 2013). Based on the N 2 O flux, the average N 2 O release reached 0.18 ± 0.05 mol·N 2 O m −2 ·yr −1 , which generated an emission estimation of 7.86 ± 2.37 g N 2 O·m −2 ·yr −1 in the paddy water (estimation method supplied in Supplementary material). The N 2 O released from the surface paddy soils contributed to the N 2 O dissolved in the paddy water and was further converted, and the remaining N 2 O was released from the paddy water into the atmosphere.

Abundance of denitrification genes and enzymatic activities related to N 2 O production and reduction in paddy soil
Microbial denitrification is a four-step reduction process catalyzed by different enzymes, which are mainly encoded by the napA, nirK/S, norB, and nosZI/II genes (Zumft, 1997). Both functional genes and enzymatic activities are vital to denitrification. N 2 O is an intermediate product of the denitrification process, and the relative contributions of N 2 O production and reduction determine whether N 2 O is released into the atmosphere as well as the amount released (Saarenheimo et al., 2015).
Regarding the abundance of the denitrification genes, previous studies pertaining to paddy soils (Zhao et al., 2019;Zhang et al., 2021) and wetland soils (Jiang et al., 2020) showed higher abundance of the nirS gene than the nirK gene. Similar to these studies, the abundance of nirS gene in the paddy soils was 1.39-19.71 times higher than that of nirK gene (Figures 5A, C). Furthermore, it was consistent with the better adaptation of the nirS gene to stable and watery environments compared with the nirK gene (Petersen et al., 2012). In addition, the abundance ratio of nirS to the 16S rRNA gene was positively correlated with the potential N 2 O emission rate (Supplementary Figure 4), as similarly observed in soils (Assémien et al., 2019;Zhao et al., 2019). Among the genes contributing to N 2 O reduction, the abundance of nosZ II gene was higher than that of nosZ I gene, and the abundance ratios of nosZ II to nosZ I genes ranged from 2.67 to 10.91 (Figures 5B, C). The ratios were consistent with the typically reported value ranging between 1.5 and 10 in different environments (Jones et al., 2013;Frame et al., 2014;Tsiknia et al., 2015). Additionally, previous studies demonstrated that the abundance of the nosZ II gene contributed significantly to the soil N 2 O sink capacity . Thus, nosZ II gene may be vital to N 2 O reduction. Considering genes related to N 2 O production and reduction, the nir/nosZ abundance ratios ranged from 0.94 to 5.17 (Figure 5C), which were similar to those in farmland soil (Zhao et al., 2019) and lake sediments (Saarenheimo et al., 2015). The ratios revealed an imbalance between N 2 O production and reduction, indicating that the potential of paddy soils to produce N 2 O is greater than reduce it at the genetic level (Domeignoz-Horta et al., 2015).
Microbial denitrification is an enzyme-mediated biochemical process; however, the activities of denitrifying enzymes have been disregarded in most researches (Morales et al., 2010;Petersen et al., 2012). Some studies showed that enzymatic activity affected sediment N 2 O emissions (Zheng et al., 2014;Su et al., 2019). NOR and NOS are enzymes that catalyze the production and reduction processes of N 2 O, respectively (Braker and Tiedje, 2003;Zumft and Kroneck, 2007). In our study, the NOR activity in paddy soil (413.19-829.52 U·g −1 ) was higher than that in Donghu sediment (274.70 U·g −1 ), while the NOS activity (85.19-148.92 U·g −1 ) was lower than that in Donghu sediment (188.73 U·g −1 ) (Zhang et al., 2022). NOR activity was higher than NOS activity, with the ratio of NOR to NOS ranging from 2.77 to 9.42 (Figure 4). This ratio was higher than the ratio detected in Donghu Lake sediments, i.e., 1.46 (Zhang et al., 2022), and that determined from riparian sediments, i.e., 0.33 (Su et al., 2019). These enzymatic activities agreed well with the genetic potential of N 2 O emissions ( Figure 5C) and potential N 2 O emission rates ( Figure 3C) in the paddy soils. The enzymatic activity ratios suggest that paddy soils have a higher potential to produce N 2 O than to reduce it at the protein level. Enzyme are proteins encoded by functional genes, in view of the fact that the gene abundance of nir was higher than that of nosZ gene (Figure 5C), the higher activity of NOR than NOS was more likely to be the result of gene expression.

Modularity of denitrification process and taxonomy groups of different denitrifying bacteria
By performing metagenomic sequencing, we achieved a more comprehensive understanding of the bacterial community composition in paddy soils, particularly that of denitrifying bacteria. Bacteria in the paddy soils investigated were mainly Proteobacteria, Actinobacteria, Chloroflexi, and Acidobacteria at the phylum level (Supplementary Figure 1); such microbial community composition has been similarly detected in paddy soils from Jiangxi Province . However, the dominant community composition varied for different denitrification genes (Figure 6).
Several genes including napA, nirS, nirK, and norB are closely related with the production of N 2 O (Zumft, 1997). Our results showed that microbial groups with different denitrification genes exhibited distinct taxonomic characteristics. Proteobacteria was the most abundant phylum for the napA gene in paddy soils, which was consistent with the results of sediments from the Pearl River Estuary (Wang et al., 2021). In addition, other napA-harboring bacteria were affiliated with Myxococcus and Actinobacteria ( Figure 6A and Supplementary Table 3). Previous studies based on high-throughput amplicon sequencing or clone libraries demonstrated that the nirS gene belonged to Proteobacteria in sediments , red soil (Ye et al., 2022) and paddy soil (Yoshida et al., 2009). We performed metagenomic analysis, which avoided primer preference, and discovered that the nirS gene predominantly existed in Proteobacteria, but also in a small proportion of unclassified bacteria, Chloroflexi, and Actinobacteria in all the paddy soils ( Figure 6A and Supplementary Table 4). The community diversity of nirS gene in this study was higher than that reported in other studies Liu et al., 2020;Ye et al., 2022). However, the community composition of the nirK gene was more diverse than that of nirS gene. The sequences of nirK gene were assigned to unclassified bacteria, Proteobacteria, Actinobacteria, and Chloroflexi ( Figure 6C and Supplementary  Table 5). Similarly, Proteobacteria and Chloroflexi were the dominant phyla of the nirK gene in sediments from the Pearl River Estuary (Wang et al., 2021). Owing to primer limitations, information regarding groups containing the norB gene is limited. In this study, norB gene was widespread in Proteobacteria, unclassified bacteria, Actinobacteria, Myxococcus, Acidobacteria, Planctomycetes, and Bacteroidetes (Figure 6 and Supplementary  Table 6). Moreover, Gammaproteobacteria contained high levels of both the nirS gene (9.22-57.61 TPM) and norB gene (16.58-42.58 TPM) ( Figure 7A). This result was similar to a previous finding where a high percentage of organisms with the nirS gene was discovered among bacteria that also contained the norB gene (Graf et al., 2014). However, the other groups of microorganisms did not contain the nirS gene but contained the norB gene (Figure 7). Many dominant phyla of the norB gene (including Actinobacteria, Planctomycetes, Desulfobacterota, Cyanobacteria, Acidobacteria, Bacteroidetes, and Myxococcus) do not have or have a relatively low abundance of the nirS gene because of the predominant distribution of the nirS gene in the Proteobacteria phylum (Figure 7).
The reduction of N 2 O, which is the only pathway for the N 2 O sink, is solely related to nosZ I and nosZ II genes (Jones et al., 2013). Although the nosZ I gene was predominately related to Proteobacteria (Green et al., 2010;Xiang et al., 2022), many previously overlooked phyla were identified in all paddy soils, including Actinobacteria and Myxococcus (Figure 6C and Supplementary Table 7). Thus, previous community studies based on high-throughput amplicon sequencing of nirS and nosZ I genes inevitably overlooked some denitrifying bacterial diversity. The nosZ II gene appeared in Proteobacteria, Acidobacteria, unclassified bacteria, Bacteroidetes, Myxococcus, Gemmatimonadetes, and Chloroflexi in all the paddy soils investigated (Figure 6D and Supplementary Table 8); similarly, these phyla of the nosZ II gene have been reported in sediments from Bohai Sea and Jiulong River (Dai et al., 2022). For the nosZ I and nosZ II genes, Gemmatimonadetes appeared only in the nosZ II gene (Supplementary Tables 7, 8), as detected in soils from Hebei Province (Zhao et al., 2019). In addition, the abundance of nosZ II was higher than that of nosZ I in many phyla, including Acidobacteria, Bacteroidetes, Myxococcus, and Chloroflexi (Supplementary Tables 7, 8), indicating that these phyla are more likely to be NosZ-II-type N 2 O-reducing bacteria.
NapA, nirS, and nosZ I genes were mainly affiliated with Proteobacteria, whereas nirK, norB, nosZ II genes were composed of different phyla (Figure 6). As the norB gene contributes to N 2 O production (Zumft, 1997), we further analyzed the abundance of denitrification genes in the dominant community of the norB gene. Gammaproteobacteria and many phyla, including Actinobacteria, Planctomycetes, Desulfobacterota, Cyanobacteria, Acidobacteria, Bacteroidetes, and Myxococcus, harbored the norB gene and did not have or had a relatively lower abundance of the nosZ gene (Figures 7A-E). This imbalance between N 2 O production and the reduction in these phyla may have contributed to the release of N 2 O from paddy soils. In addition, the abundance of the norB and nosZ II genes was relatively high in Acidobacteria, Bacteroidetes, and Myxococcus, whereas the abundance of napA, nirK, and nosZ I genes was low, and that of nirS gene was absent (Figures 7F-H). This suggests that Acidobacteria, Bacteroidetes, and Myxococcus ware more likely to be NosZ-II-type N 2 O-reducing bacteria that contain the norB gene but lack the nir gene. This was consistent with the discovery of bacteria that contain the nosZ gene and lack the nir gene, mainly in Bacteroidetes (Graf et al., 2014), and further extended the understanding that Bacteroidetes did not contain the nir gene but harbored the norB and nosZ II genes. For Alphaproteobacteria, the similar abundance distribution of the norB gene and the nosZ I gene indicate that they possess both N 2 O production and reduction potential (Supplementary Figure 5). It is inferred that their contribution to N 2 O emissions is likely smaller than that of Gammaproteobacteria. Results of cooccurrence analysis between different denitrification genes showed that denitrification in paddy soil was a highly modular and truncated process that caused an imbalance in the N 2 O production and reduction processes, thus resulting in N 2 O emission. However, more data regarding the distribution of denitrification genes in diverse microbial taxa are required to obtain a more comprehensive understanding of the modularity in denitrification.

Conclusion
This study provides valuable insights into N 2 O emission during paddy soil denitrification. The high potential for N 2 O emission was facilitated by the imbalance between N 2 O production and reduction processes in terms of gene abundance (nir/nosZ) and enzymatic activity (NOR/NOS). The nirS and nosZ II genes were abundant in paddy soils. Furthermore, the composition of denitrification genes demonstrated a highly modularized denitrification process in paddy fields. Gammaproteobacteria and other phyla, including Actinobacteria, Planctomycetes, Desulfobacterota, Cyanobacteria, Acidobacteria, Bacteroidetes, and Myxococcus, containing the norB gene without nosZ genes, may contribute to N 2 O emission from paddy soils. These results enhance our understanding of N 2 O emission during denitrification and provide a theoretical basis for mitigating greenhouse gas emissions in agricultural ecosystems.

Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in below: https://www.ncbi.nlm. nih.gov/, PRJNA957066.