In June 2022, the soil was sampled in a long-term field experiment at Fengqiu Agricultural Ecological Experimental Station of Chinese Academy of Sciences (35°01′N, 114°32′E). The experiment has been conducted since 2010. The basic soil properties at the beginning of the experiment were shown in Table 1. This study collected soil samples from four fertilization treatments: NN (No straw and no nitrogen fertilization), CF (No straw and conventional fertilization), SM (Straw mulching and no nitrogen fertilization), SMCF (Straw mulching and conventional fertilization). NN and SM treatments were applied without nitrogen and only phosphorus and potassium fertilizers were applied. Conventional fertilization of CF and SMCF treatments indicated that both nitrogen, phosphorus and potassium fertilizers were applied. In all fertilizer treatments, the total amount of nitrogen (pure nitrogen) applied to wheat throughout the whole growth period was 210 kg ha^–1, and the phosphorus and potassium fertilizers were 157 kg ha^–1 for P₂O₅ and 105 kg ha^–1 for K₂O, respectively. Straw mulching was done by cutting corn straw into 20–50 mm strips after the previous season’s corn harvest, and then evenly covering the soil surface with straw after tilling (at a depth of about 15 cm). All the crop straw from each plot was returned to the same plot. The soil samples (0–20 cm depth) were collected using the five-point diagonal sampling method. Subsequently, the samples were put in sterile bags and temporarily stored in foam boxes with dry ice and taken back to the laboratory at the station within 2 h. Adequate dry ice was used to ensure that the soil samples were in an environment below 0°C. All the soil samples were passed through a 2 mm diameter mesh sieve to remove plant debris and stones, and the remaining fine root and straw residues were manually removed. The soil samples were stored at 4°C before constructing the micro world.

Soil samples were pre-incubated at 28°C for 7 days with moisture adjusted to 50% of water holding capacity. A 21-day microcosm experiment was established by placing 20 g (equivalent dry weight) of pre-incubated soil in 120 ml serum bottles. The following treatments were set up for each fertilized soil: (i) water only (CK), (ii) urea only (U), (iii) urea + PTIO (UP), and (iv) urea + DMPP (UD). Urea was added at 200 mg N kg^–1 dry soil and DMPP was added at 1.5% of urea N (Fan et al., 2019; Yin et al., 2021) and the concentration of PTIO was set to 100 μM, (Jung et al., 2014). Three replications were set up for all treatments and a total of 144 microcosms were established. Gas samples (20 ml) were collected from the top with a syringe at 1, 2, 3, 4, 7, 10, 14, and 21 days after incubation. N₂O concentrations were analyzed by gas chromatograph (Agilent 7890, Agilent Technologies, Santa Clara, CA, USA). To maintain aerobic conditions, all remaining bottles were vented for 30 min after each sampling and then resealed. Soil ammonium nitrogen (NH₄⁺-N) and nitrate nitrogen (NO₃^–-N) contents were determined by destructive collection of three bottles from each treatment on days 0, 7, 14, and 21 of incubation. A portion of the last sampled soil was stored at −80°C for DNA extraction.

Total DNA was extracted from 0.5 g of fresh Soil using the Fast DNA^® Spin Kit for Soil (MP Biomedicals, Inc.). The extracted DNA concentration and purity were examined by NanoDrop2000 (Thermo Fisher Scientific, Waltham, MA, USA). We used primers TAReuk454FWD1 (CCAGCASCYGCGGTA ATTCC)/TAReukREV3 (ACTTTCGTTCTTGATYRA) (Stoeck et al., 2010), Arch-amoA26F (GACTACATATTCTACACWGACT GGGC)/Arch-amoA417R (GGTGTCATATATTGGAGGCAACGT TGG) and amoA1F (GGGGTTTCTACTGGTGGT)/amoA2R(CCC CTCKGSAAAGCCTTCTTC) (Xia et al., 2011) amplified protists, AOA and AOB, respectively. AOA and AOB were analyzed on the qTOWER3/G amplification apparatus (Analytik Jena AG, Germany) with three replicates per sample. The standard curve is constructed from 8 series of diluents containing known copy number plasmids of the target gene. The reaction system for qPCR was 20 μl, including 1 μl of DNA template, 10 μl of TB Green^® Premix Ex Taq TM, 0.25 μl forward and 0.25 μl reverse primers, and 8.5 μl of sterilized double-distilled water. Negative control was performed using sterilized distilled water instead of the template DNA sample as the reaction template.

Amplicon sequence variants (ASVs) were generated from raw sequence data using DADA2 (Callahan et al., 2016). Briefly, after removing the primer and adapt sequences using “cutadapt” (Martin, 2011), the sequences were filtered and trimmed using “filterAndTrim” with the following parameters: maxN = 0, maxEE = c(2,2), truncQ = 2. Chimeric sequences were detected and removed using “removeBimeraDenovo” based on the “consensus” method. The taxonomy assignment for protist was performed with the “assignTaxonomy” function (minimal bootstrap value 50) based on the PR2 database (V4.14.0) (Guillou et al., 2012). Taxonomy of the amoA ASVs was assigned using BLASTn based on the nucleotide database in NCBI with the default algorithm parameters (expect threshold was 0.05, word size was 28, match/mismatch scores were 1, -2, gap costs was linear, and e-value was 0.00001) (Kõljalg et al., 2005). The unassigned ASVs were filter for further study. The ASVs of 18S rRNA assigned as Rhodophyta, Streptophyta, Metazoa, Fungi, and unidentified Opisthokonta were removed for protist community analysis. All high-throughput sequencing steps were performed by Guangdong Magi Gene Technology Co., Ltd. (Guangzhou, China).

One-way analysis of variance (ANOVA) in SPSS Statistics 21 was used to evaluate the effects of different treatments on N₂O emission and copy number of ammonia-oxidizing microorganisms. The “vegan” package in R v4.2.1 software was used to conduct principal coordinate analysis (PCoA) (Anderson and Willis, 2003). Pearson correlation analysis was conducted using R v4.2.1 software packages reshape2 and psych to examine the correlation between key taxa and soil physicochemical properties, as well as between key taxa. A heat map was generated using the “pheatmap” package.

The co-occurrence network analysis based on Spearman’s correlation matrix was calculated in the “Hmisc” package in R v4.2.1 (Wang et al., 2022). ASVs with a sum of relative abundance higher than 90% were selected, and then two correlations between all ASVs were calculated by combining protists (744 PASVs), AOA (105 ASVs) and AOB (67 BASVs) into one relative abundance table. To distinguish different microorganisms, we used PASV to represent the ASV of protists and BASV to represent the ASV of AOB. A correlation threshold higher than 0.6 and a P-value lower than 0.01 are considered to be stable co-occurrence relationships. Adjustments were made by using test corrections from the Benjamini-Hochberg procedure to reduce the chances of obtaining false positive results (Benjamini and Hochberg, 1995). The co-occurrence network was visualized and analyzed by “Gephi” software (Bastian et al., 2009). Subsequently, the connections within modules (Zi) and between modules (Pi) were calculated using the package “igraph” in R v4.2.1. According to the topological characteristics of nodes, node attributes were classified into four types, including: Module hubs, nodes with high connectivity within a module (Zi > 2.5 and Pi < 0.62). Connectors, nodes with high connectivity between two modules (Zi < 2.5 and Pi > 0.62). Network hubs, nodes with high connectivity throughout the network (Zi > 2.5 and Pi > 0.62), and Peripherals, nodes that do not have high connectivity both within and between modules (Zi < 2.5 and Pi < 0.62) (Deng et al., 2016). Moreover, according to Banerjee et al. (2018) and Shi et al. (2020), module hubs connectors, and network hubs were defined as keystone species in this study. Among them, the taxa of module hubs generally mediate interactions between species within the module and have significant implications for module function by regulating energy and material exchange within the module (Guimera and Nunes Amaral, 2005; Toju et al., 2018). Therefore, this study mainly focused on module hubs.

The SEM quantified the direct and indirect effects of soil physicochemical properties, protists, AOA, and AOB on PNR and N₂O. Soil physicochemical properties (soil NO₃^–-N and NH₄⁺-N) and microbial diversity (protists, AOA and AOB diversity) were represented by PCoA data of corresponding indicators. The maximum likelihood estimation method was applied to fit the model, and the parameters were consistent with relatively low chi-square and P > 0.05, goodness fit index (GFI) > 0.90 and root mean square error of approximation (RMSEA) < 0.05 (Grace and Keeley, 2006). All SEM analyses were performed using AMOS 21.

Nitrous oxide emission rate was significantly affected by soil fertilization practices and nitrification inhibitors. For the urea (alone) treatments, the N₂O emission rate of the long-term no-nitrogen treatments (NN, SM) peaked on day 4 (Figures 1A, C), and the long-term nitrogen treatments (CF, SMCF) reached the maximum rates on day 2 (Figures 1B, D). Meanwhile, long-term nitrogen and straw treatment (SMCF) had the highest N₂O emission rate (Figure 1D). N₂O emissions ranged from 18.84∼64.32 pmol g^–1 h^–1 under different fertilization practices, thus, long-term nitrogen and straw application promoted N₂O emission by 3.4 times. The N₂O emission rate was reduced at different degrees by the application of nitrification inhibitors. In all soils, N₂O emission rates of U and UP treatments rapidly increased during the first week of incubation. During the whole incubation period, the UD treatment significantly decreased the N₂O emission rate, and the emission trend was essentially consistent with CK. The range of N₂O emissions under nitrification inhibitor treatments was 1.03∼64.32 pmol g^–1h^–1, which reduced N₂O emissions by 62.4 times. In contrast, nitrification inhibitors were much more effective on N₂O emissions than fertilizer treatments. Additionally, compared with U treatment, NO₃^–-N increased slowly in the UD treatment, and the UP treatment reduced NH₄⁺-N content (Supplementary Figure 1).

The effect of soil fertilization practices on PNR was significant only in the CK treatment, i.e., the long-term nitrogen application treatments (CF and SMCF) obviously increased PNR under the condition of no exogenous urea addition. Moreover, long-term nitrogen and straw application promoted PNR by 7.5 times. However, the addition of urea during incubation reduced the differences of soil fertilization practices. The U and UP treatments had the highest PNR in all soils. Compared with U treatment, the PNR was significantly increased in UP treatment on day 7 of NN and CF treatments, while there was a decreasing trend on the day 14 and 21. The PNR of UP treatment decreased with the extension of incubation time compared to the U treatment in SM treatment. The UD treatment significantly decreased PNR, with a trend consistent with CK or even obviously lower than CK (Table 2). PNR was reduced by 80 times after treatment with nitrification inhibitors. In general, the effect of fertilization practices on PNR was much lower than that of nitrification inhibitors.

To explore the effects of fertilization practices and nitrification inhibitors on the activity of ammonia-oxidizing microorganisms, we analyzed the changes of AOA and AOB gene abundance. For AOA, the highest gene abundance of all soil treatments was found in the NN treatment on day 7. Moreover, the UP treatment reduced AOA gene abundance in all soils. Unexpectedly, obviously higher AOA gene abundance was observed in UD and CK treatments than in U treatment (Figures 2A–D). For AOB, CF treatment had the highest gene abundance on day 7. It is noteworthy that the gene abundance of SMCF treatment was lower than that of CF treatment. The effect of nitrification inhibitors on the AOB gene abundance under different fertilization practices showed the same trend. AOB gene abundance was significantly improved in U treatment and obviously decreased in UD treatment for all soils (Figures 2E–H). It appeared that AOA was better suited to oligotrophic environments, while AOB was more active in nitrogen-rich environments.

We further analyzed the response of microbial community structure to fertilization practices and nitrification inhibitors. Soil fertilization practices and nitrification inhibitor treatments had different effects on the community structure of protist, AOA, and AOB. Notably, all microbial communities in NN and SMCF treatments showed obvious separation, suggesting that long-term nitrogen and straw application significantly changed the microbial community structure (Figure 3). For protists, CK treatment was significantly detached from other treatments, while there was no obvious variation in community structure between U, UP, and UD treatments. For AOA, significant changes in community structure occurred in the UP treatment. Compared with nitrification inhibitor treatments, different fertilization soils had a greater impact on AOB community structure. For example, the AOB community structure in soils with long-term non-nitrogen application was clearly apart from soils with long-term nitrogen application.

In order to determine the impact of nitrification inhibitors on the symbiotic relationship of soil microorganisms, the co-occurring networks of protists, AOA, and AOB were constructed for CK, U, UP, and UD treatments, respectively, and the topological parameters of each network were calculated (Figure 4A). Compared to CK, the network nodes, edges, average degree, clustering coefficient and network density were increased for all urea-added treatments. Notably, all topological properties of UD treatment were reduced, resulting in a lower complexity of the network (Supplementary Table 1). Meanwhile, the direct microbial interaction showed the lowest positive correlation ratio (79.92%) under UD treatment. Nodes with higher within module connectivity (Module hubs, Zi > 2.5 and Pi < 0.62) were identified as key taxa in the network. These taxa may play an important role in maintaining the stability of network structure. There were no module hubs in CK treatment (Figure 4B), while UP and UD treatments had more key taxa than U treatment. The key taxa identified in the U, UP, UD treatments were mainly Cercozoa, Lobosa, Sagenista, Crenarchaeota and Proteobacteria (Supplementary Table 2). Moreover, the relative abundances of key taxa were also highly correlated with physicochemical properties associated with nitrogen transformation (Figure 5).

N₂O, PNR, NO₃^–-N, and NH₄⁺-N are indicators closely related to the nitrification process. However, in contrast to the other indicators, NH₄⁺-N was mainly negatively correlated with key species of microorganisms. Among the key taxa of the protists, the proportion of key taxa with significant positive correlations with N₂O, PNR and NO₃^–-N was 50%, while the proportion of negative correlations was only 25%. The key taxa of protists with positive correlations were PASV1204, PASV634, PASV23, and PASV393, classified at the phylum level as Cercozoa, Lobosa and Sagenista. For AOA, only ASV56 was strongly linked to the four physicochemical properties, but the correlation was mainly negative. For AOB, although there were only two key taxa, BASV1 was positively associated with both N₂O and PNR, and belonged to the Proteobacteria at the phylum level (Figure 5A). Additionally, the proportion of key species of AOA that were significantly negatively related to PASV1204, PASV634, PASV23, and PASV393 was 11%, and the proportion of positively correlated key species was 50% for AOB. For AOA, only ASV56 was significantly negatively linked to key species of protists, and BASV1 of AOB had the opposite pattern to ASV56 (Figure 5B).

The SEM was constructed to further clarify the relationship between soil properties, microorganisms, PNR and N₂O under different nitrification inhibitor treatments (Figure 6). In all treatments, AOA had a direct negative effect on protists, conversely, AOB had a direct positive impact on protists. In CK and U treatment, the influences of microorganisms on PNR and N₂O were mostly insignificant (Figures 6A, B). Microorganisms were more closely related to each other and to PNR and N₂O with the addition of nitrification inhibitors (Figures 6C, D). In terms of standardized total effects, the addition of nitrification inhibitors resulted in a greater contribution of AOA and AOB on protist compared to the U treatment. To be specific, the interaction between AOB and protists increased the contribution to N₂O after inhibiting AOA, the suppression of AOB enhanced the contribution of interaction between AOA and protists to PNR (Supplementary Figure 2). Overall, when AOA was suppressed, only N₂O was significantly affected by interaction between AOB and protist. The interaction between AOA and protists obviously affected both N₂O and PNR with high contributions when AOB was inhibited.

The highest N₂O emissions occurred in treatment with long-term nitrogen and straw application (Figure 1), which is consistent with recent studies (Xia et al., 2018). Long-term nitrogen application stimulates N₂O emissions by affecting biological processes such as nitrification and denitrification in soil (Wu et al., 2018; Bai et al., 2020). Meanwhile, straw application can alleviate the limitation of soil N₂O emission by providing carbon and nitrogen sources. In particular, straw addition enhances the availability of unstable carbon, which provided energy for the growth of N₂O-producing microorganisms (e.g., AOA and AOB) in the soil (Wu et al., 2020).

As is well known, N₂O is an important byproduct of nitrification or denitrification (Ma et al., 2012). The ammonia oxidation process driven by AOA and AOB is the initial and rate limiting step of nitrification (Hu and He, 2017; Lin et al., 2020). Our study confirmed that N₂O emission was slightly reduced when AOA was inhibited and significantly reduced when AOB was suppressed (Figure 1). It indicated that AOB may be the main driver of N₂O emission during the nitrification process in this study. Firstly, nitrification denitrification and incomplete NH₂OH oxidation are the causes of N₂O production by AOB in alkaline soils (Wu et al., 2018). Secondly, AOA seems to lack nitrification-denitrification ability due to the absence of NO reductase (Walker et al., 2010). Therefore, the effect of suppressing AOA on N₂O emissions was not obvious. Except for N₂O, PNR is also an important chemical index to characterize the nitrification of aerobic ammoxidation microorganisms, which can be used to quantify nitrification potential. Our study revealed that long-term nitrogen application significantly increased PNR, while PNR was reduced sharply after AOB was suppressed (Table 2). On the one hand, long-term nitrogen application provided sufficient substrate for nitrifying microorganisms and increased microbial activity (Wang et al., 2019), which in turn improved PNR. On the other hand, suppression of AOB reduced PNR probably owing to the fact that AOB is the major contributor to nitrification potential in neutral and alkaline soils (Ren et al., 2023). Thus, inhibition of AOB by DMPP indirectly suppressed PNR.

In this study, the long-term no nitrogen treatment had the highest AOA gene abundance (Figure 2A). This can be explained by the fact that AOA is dominant in nitrogen-limited environment (Wang et al., 2015), and long-term non-nitrogen soils provide favorable conditions for AOA to compete for ecological niches. In contrast to AOA, the peak of AOB gene abundance appeared in long-term nitrogen treatment (CF). It is due to that AOB is suited to survive in alkaline and nitrogen-enriched environments, and the application of nitrogen can provide high soil NH₄⁺-N concentration for AOB growth (Verhamme et al., 2011; Sun et al., 2023). Additionally, we found that AOB gene abundance was lower in SMCF treatment than in CF treatment (Figure 2H). The incorporation of straw increased soil C/N, thereby delaying the nitrification process (Elrys et al., 2021; Wang et al., 2023), and thus the decrease of AOB gene abundance is understandable.

The addition of nitrification inhibitors had different effects on AOA and AOB. Our results showed that PTIO significantly inhibited the AOA gene abundance (Figures 2A–D). This may be explained by the high sensitivity of AOA to PTIO (Martens-Habbena et al., 2015). Interestingly, AOA gene abundance was obviously increased after suppressing AOB in long-term nitrogen fertilization soils (Figures 2A–D). This was inconsistent with the results of most studies that revealed AOA populations were unchanged or decreased (Di and Cameron, 2011). Firstly, there may be substrate competition between AOA and AOB, and the suppression effect of DMPP may free AOA from the competition of AOB (Fan et al., 2019). Secondly, DMPP, as an organic compound, may serve as a useful C-substrate to promote AOA growth (Florio et al., 2014). Meanwhile, the application of DMPP strongly inhibited the AOB gene abundance (Figures 2E–H) in all soils. It was worth noting that AOA activity was not affected by DMPP (Figures 2A–D). In fact, the different responses of AOA and AOB to DMPP inhibition may be attributed to different nitrifying enzyme systems. The AOA amoA sequence is more similar to the gene encoding bacterial micromethane monooxygenase (pMMO), resulting in functional differences between the AMO of AOA and AOB (Jung et al., 2014). The nitrification inhibitor DMPP decreases AMO activity, whereas MMO is unaffected (Weiske et al., 2001). Therefore, DMPP specifically inhibited AOB activity.

In contrast to AOA and AOB, protists are significantly affected by nitrogen fertilizer rather than inhibitors. Protists are the predators of soil micro-food web and the response to nitrogen are sensitive (Zhao et al., 2021). To explore the effect of interactions between protists and AOA or AOB on nitrogen transformation, a symbiotic network between them was constructed. We found that the addition of urea increased the complexity of the soil microbiome network and caused closer network connectivity (Supplementary Table 1). As nitrogen addition increased the availability of resources and food in the soil, it enhanced the interactions within the microbiome to improve resource utilization efficiency (Shi et al., 2016; Yu et al., 2018). Remarkably, the network complexity was lower when AOB was suppressed by DMPP. This may be due to the key taxa of AOB was inhibited, which can maintain network stability and increase network complexity. It was also confirmed by the fact that AOB was not identified in the keystone species treated with UD (Figure 4).

Nodes with high connectivity within the module (Zi > 2.5 and Pi < 0.62) were identified as key taxa. These taxa may function through energy and substance exchange within the module (Deng et al., 2016). When they are removed or disappeared from a particular ecosystem, it can lead to drastic changes in that ecosystem (Ikegami, 2005). In this study, the key species of protists after inhibitor addition mainly included PASV1204, PASV634, PASV23, and PASV393 (Supplementary Table 2), all of which were significantly positively correlated with N₂O, PNR and NO₃^–-N. We further classified the key species and determined that they belong to Cercozoa, Lobosa and Sagenista at the phylum level. However, only Cercozoa was the protist species that appeared consistently since suppressing AOA or AOB, which was classified as generalist in microbial networks (Barberán et al., 2012). It emphasized the important role in microbial symbiosis. Cercozoa belongs to the predatory protist in the functional classification. It generally affects soil nutrient mineralization by top-down regulation of microbial biomass and activity and by affecting the acquisition of soil nitrogen and carbon enzyme activities (Wei et al., 2021; Wu et al., 2022). Additionally, BASV1, belonging to Proteobacteria at the phylum level, is the only AOB key species identified after inhibiting AOA. It is strongly positively associated with N₂O, PNR, and NO₃^–-N, and key species of protist (Figure 5). It is well known that Proteobacteria is a key denitrification species with the potential to produce and reduce N₂O (Xiang et al., 2023).

For AOA, on the one hand, N₂O and PNR showed no significant changes or increased trends after AOA was inhibited. Meanwhile, we found that only ASV56 was significantly negatively correlated with both N₂O and PNR among the key species of AOA. On the other hand, only ASV56 was negatively linked to keystone species of protists. Furthermore, the SEMs of four treatments showed a significant negative correlation between AOA and protists. The standardized total effects also indicated high negative contributions of AOA to protists. Therefore, we speculated that among the key species of AOA, ASV56 may play a major role in microbial interactions in this study. Since the key taxa of AOA was not identified, the relevant functions were analyzed by NCBI similar sequence alignment (Supplementary Table 3). By comparison, the key AOA species may be Nitrososphaerota archaeon, which can obtain energy during ammonia oxidation and utilize carbon dioxide as a chemoautotrophic carbon source (Wei et al., 2023). The interactions between protists and AOA or AOB may provide substrates and motivation for the soil nitrification process and stimulate nitrogen transformation. Specifically, Cercozoa had a higher C/N than sub-trophic level organisms, and they excreted excess nitrogen making it available to other microorganisms (Jousset, 2017). Meanwhile, Proteobacteria and Nitrososphaerota archaeon can also use the nitrogen secreted by protists to promote PNR and the N₂O emission. Recent studies have found that the symbiotic relationship between Cercozoa and Proteobacteria was closely related to soil nitrogen transformation process (Xiao et al., 2022). The predator-prey interactions between Cercozoa and Thaumarchaeota increased nutrient turnover (Gu et al., 2023). These all support the above view.

The SEM analysis also provided a basis for the above argument. Protists were found to be the highly negatively correlated with AOA and significant positively associated with AOB. The addition of nitrification inhibitors resulted in a strong negative linkage between AOA and AOB. At the same time, standardized total effects indicated that AOA or AOB was closely related to protists. When one of AOA and AOB was inhibited, the other interacted with protists to increase contributions on N₂O or PNR. Since the fact that both AOA and AOB use ammonia nitrogen as energy substrates, there is a direct nutritional competition between them (French et al., 2021; Sun et al., 2022). When one of the microorganisms was suppressed, the interaction between the other and protists may be more closely. Thus, the addition of nitrification inhibitors may promote the interactions between ammonia-oxidizing microorganisms and protists, and the trophic interaction may stimulate nitrogen turnover (Gao et al., 2019). Therefore, the addition of nitrification inhibitors is a noteworthy agricultural strategy to improve the soil nitrogen cycle.