Introduction

It is well known that nitrous oxide (N2O) is a potent greenhouse gas and strongly catalyzes stratospheric ozone depletion [1]. Furthermore, recent reports have revealed that atmospheric N2O concentration has been constantly rising and reached 328 ppb in 2016 (121% of preindustrial levels), mainly due to anthropogenic intervention in the nitrogen cycle [2]. Given that agriculturally managed soils emit 4.3–5.8 Tg N2O yr−1, developing effective mitigation strategies is a key challenge for the agricultural sector [3].

Rice is one of the most widely consumed staple foods in the world and is cultivated on nearly 155 million ha [4]. Paddy fields are an unavoidable component of many agricultural ecosystems. Irrigated rice cultivation often includes water regimes that alternate between flooded and drained conditions throughout the crop cycle, producing optimal conditions for denitrification [5]. Notably, the use of mid-season drainage sharply increases the flux of N2O [6]. In addition, increasing N2O levels in the atmosphere are significantly correlated with high nitrogen (N) and organic fertilizer application rates in agricultural ecosystems [7], especially paddy fields [8]. Therefore, the reduced application of chemical fertilizers in paddy fields would be a route to significantly decreasing N2O emissions.

To achieve high yields of rice as well as decreasing N2O, sustainable alternatives to these inorganic fertilizers are being sought [9, 10]. Biochar is generated by thermal decomposition of organic biomass under low oxygen conditions and has attracted a lot of attentions from agronomists and soil scientists because of its soil-enhancing properties and resistance to biodegradation [12, 13]. Furthermore, its potential function of inhibiting greenhouse gas emissions has been verified in a series of field and incubation experiments [14,15,16]. There are some evidences that, when biochar is added, soil pH tends to high and then inhibit the emission of N2O [17, 18] due to certain physico-chemical properties [19]. In addition, factors such as the biochar source, the soil type, and the chemical form of the N fertilizer were found to affect the function of biochar decreasing N2O [20]. However, the main mechanism leading to decreased N2O emissions is still poorly understood [14, 20, 21].

Changes in soil properties can only explain the key mechanism of decreasing N2O with difficulty. Undoubtedly, denitrification is mediated by soil microorganisms called denitrifiers, which reduced nitrate (NO3−) under oxygen-depleted conditions to produce gases such as N2O and dinitrogen (N2). They are, therefore, widely identified as the initiators of nitrogen loss from soil to atmosphere [22]. Denitrification results from a series of biological enzymes catalyzing the reactions of the different reduction steps [23]. Among these, the key reaction is the reduction of NO2− to NO, catalyzed by two structurally different, but functionally equivalent nitrite reductases, namely, cytochrome cd1-containing reductase (nirS) and copper-containing reductase (nirK) [24]. This step causes dissolved N to become NO for the first time, and NO is then easily reduced to N2O [25]. In other words, the activity of nirS and nirK genes determines the initial amount of N2O, so these genes are widely used as molecular markers to investigate the micro-mechanism of N2O loss. NirS and nirK are thought to be mutually exclusive among denitrifying species and to represent two ecologically distinct denitrifying groups [26], with diversity and abundance responding differently to various ecology systems [27,28,29]. Research has been carried out to explore the correlation between the N2O flux and the diversity and abundance of denitrifiers, using denitrification enzyme activity (DEA) or potential denitrification activity (PDA) as indicators [30,31,32]. However, the DEA or PDA represented the whole of denitrification, rather than the known contribution of nitrite reductase. In this work, therefore, we have added nitrite reductase activity (S-NiR) as a measure of the activity of nirS and nirK genes. Based on this more accurate indicator, the determination of the diversity and abundance of gene function has enabled us to identify the micro-mechanism of N2O emission.

The purplish paddy soil in the southwest China was collected in this study [33]. We chose pot experiments as the main test method, in order to amplify the effect of a gradient reducing nitrogen application with biochar on nirS- and nirK denitrifier communities. Based on the discrepancy between nirS and nirK denitrifiers responding to different fertilizer treatments, the relationship between the N2O flux, nitrite reductase, and denitrifiers can be identified. We hypothesize that (1) the medium amount of biochar application decreases the N2O flux; (2) there is a significant species discrepancy between nirS- and nirK-denitrifier communities; (3) and the nirS-denitrifiers are mainly responsible for the N2O flux.

Materials and Methods

Pot Experiments and Sampling

In the Southwest University greenhouse, a cultivation experiment with rice was established on March 10, 2017 with samples from the top 20 cm of soil at the National Monitoring Station of Soil Fertility and Fertilizer Efficiency on Purple Soils (30° 26′ N, 106° 26′ E, and 266.3 m above sea level) located in Beibei, Chongqing, China. The soil was a Cabhaplic Stagnic Anthrosol classified as a Regosol [34], derived from gray-brown purple sand shale parent materials in the Mesozoic Jurassic Shaxi Temple Group. It contained 13.95 g·kg−1 organic carbon (SOC), 121.52 mg·kg−1 available nitrogen (N), 264.25 mg·kg−1 available phosphorus (Olsen-P), and 206.32 mg·kg−1 available K and was at pH 7.86 prior to the start of the pot experiment.

The experiment included five fertilization treatments: (1) control (CK, no fertilizer); (2) chemical fertilizer (F, 1.20 g N pot−1 + 0.72 g P2O5 pot−1 + 0.96 g K2O pot−1); (3) chemical fertilizer +10 t·hm−2 biochar (LFB, 1.09 g N pot−1 + 0.69 g P2O5 pot−1 + 0.68 g K2O pot−1 + 26.67 g biochar pot−1); (4) chemical fertilizer +20 t·hm−2 biochar (MFB, 0.97 g N pot−1 + 0.67 g P2O5 pot−1 + 0.40 g K2O pot−1 + 53.34 g biochar pot−1); and (5) chemical fertilizer +40 t·hm−2 biochar (HFB, 0.74 g N pot−1 + 0.62 g P2O5 pot−1 + 0.16 g K2O pot−1 + 106.68 g biochar pot−1). Each treatment had six replicates and the 30 PVC pots (460 cm2 × 23 cm) were arranged in randomized blocks. Based on the traditional local agricultural management, the amounts of N, P2O5, and K2O were 0.2 g·kg−1, 0.12 g·kg−1, and 0.12 g·kg−1, respectively. The N, P, and K fertilizers used were urea (N 46%), calcium superphosphate (P2O5 12%), and potassium chloride (K2O 60%). Except for urea, the chemical fertilizers and biochar were applied as basal fertilizers. The proportions of the urea used as basal and supplementary fertilizer were 60 and 40%, respectively, applied on May 6 and June 1. The biochar used in the experiment was provided by Sichuan Jiusheng Agriculture Co., Ltd. It was manufactured using rape straw as the raw material to char under anaerobic pyrolysis conditions of 500 °C for 2 h.

Gas Collection and Analysis

Gas samples were collected by in situ static chamber [35]. The chambers were made of stainless steel, with a cross-sectional area of 0.36 m2 (60 × 60 cm) and a total height of 50 cm, including both layers (see Hou et al. [36] for detailed description). There was a bottom layer (50 cm diameter × 3 cm high) under each pot, and the static chambers were covered by thermal-protective coating to avoid high temperature. The first gas sample was collected immediately after the rice was transplanted and the collecting interval was 7–10 days. The operating requirements and interval time for gas collection was carried out in accordance with the method of Yang et al. [11]. Gas samples were stored in a 50-mL syringe, and N2O concentrations were analyzed using a gas chromatograph (Agilent GC-7890A, Agilent Science and Technology Ltd., USA).

Nitrite Reductase Measurement, Soil Sampling, and Analysis

The activity of nitrite reductase (S-NiR) was measured using a Nitrite Reductase Kit for soil (Comin Biotechnology Co. Ltd., Suzhou, China) according to the manufacturer’s instructions.

Three soil samples without visible stones or plant residues were collected randomly from each pot after harvest and thoroughly mixed. The soil was air-dried and passed through different sieve types according to experimental requirements. The basic physical and chemical characteristics were analyzed by standard methods [37,38,39].

DNA Extraction and Cloning

The genomic DNA of soils was extracted using a Fast DNA SPIN Kit for soil (MP Biomedicals, Illkirch, France) according to the manufacturer’s instructions. For amplification of nirS, the primer pair cd3aF [40] and R3cd [41] was used, while another primer pair 517F and 1055R was used for nirK-targeting polymerase chain reaction (PCR) [42]. The PCR was performed in a C1000™ thermal cycler (Bio-Rad, USA).

PCR-generated nirS and nirK fragments from soil DNA were excised from agarose gels and purified using a universal DNA Purification Kit (Tiangen Co. Ltd., Beijing, China) followed by molecular cloning with a pGM-T vector (Tiangen) into the E. coli strain DH5α. The 30 positive clones of nirS and nirK gene were randomly selected in each sample, respectively. In other words, each gene has 150 alternative clones from five samples. Screening for the correct size inserts was conducted via PCR with M13 forward and reverse primers (Tiangen Co. Ltd., Beijing, China). The correct size clones were sequenced, and BLAST was used to confirm the identity and validity of sequences. Typical nirS and nirK sequences, namely, operational taxonomic units (OTUs), were defined using 5% differences in the nucleotide sequences in the MOTHUR Program [43]. The typical nirS and nirK gene sequences were deposited in the GenBank database under the accession numbers nirK: MK521815-MK521827; nirS: MH316008 - MH316010, MH316012 - MH316013, MH316016 - MH316017, MH316019 - MH316020, MH316023, MH316025 - MH316029, MH316031 - MH316031, MH316037, MH316042 - MH316055.

PCR Amplification and Terminal Restriction Fragment Length Polymorphism

The nirS and nirK genes were amplified using the same primer pairs as the cloning. However, the forward primers were fluorescently labeled with carboxyfluorescein (FAM). PCR amplifications and thermocycling conditions were both same as for the cloning.

For each treatment, three individual PCR products were thoroughly mixed to avoid PCR artifacts and were gel-purified using Tiangen gel extraction kit (Tiangen Co. Ltd., Beijing, China) to get exact DNA fragments. For T-RFLP, the nirS and nirK PCR products were digested with HhaI and TaqI (New England BioLabs Co. Ltd., UK) endonuclease at 37 and 65 °C for 10 min and then inactivated at 60 and 80 °C for 20 min, respectively. T-RFLP profiles were generated by the Sangni Corporation (Shanghai, China) using an ABI Prism 3100 Genetic Analyzer. In order to assign the peaks of T-RFs to clone sequences, clones corresponding to different in silico T-RFs were also subject to T-RFLP analysis following the same procedure described above except that monoclones were used as the template for PCR.

Quantification of Gene Abundance

The abundances of nirS- and nirK-denitrifiers were estimated via quantitative PCR (qPCR). The primer pairs nirScd3aF/R3cd [41] and nirK876/1040R [44] were used for nirS and nirK gene amplifications, respectively. Three duplications were performed per sample. Thermal cycling, fluorescence data collection, and data analysis were conducted with the iCycler iQ5 thermocycler (Bio-Rad, USA) detection system according to the manufacturer’s instructions.

The standard curve was developed by methods described in the literature [44]. In order to test the potential inhibition of soil substance spiking in DNA, the soil DNA extracts were serially diluted and subject to qPCR analysis. Minimum inhibition was observed when the dilution was 20-fold. The efficiency of the reactions was 90% and 103% for nirS and nirK, respectively. The R2 values for the standard curves were 0.998 and 0.997, respectively.

Statistical Analysis

The physico-chemical properties and PCR values were analyzed using the IBM SPSS statistical software package version 23 (IBM Corporation, New York, USA) with one-way ANOVA and Pearson’s least significant difference (p < 0.05).

Size and relative abundances of T-RFs were quantified using Peak Scanner version 1.0 (Applied Biosystems, Inc., USA). The raw data was modified for noise and peak alignment and was presented in a data matrix [45]. Peaks with a fluorescence signal above 2% of the sum of all peak height signals in each sample were included in further analysis.

Based on the T-RFs, α-diversity (Shannon index, Margalef index and Pielou index) was calculated by the method of Atlas and Bartha [46]. The T-RFs composition of nir-denitrifiers was presented using Heatmap (HemI 1.0 version). Distance-based redundancy analysis (db-RDA) was chosen to explore the relationship between the community structure of nirK- and nirS-denitrifiers and soil environmental factors using Canoco 5 for Windows (Microcomputer Power, Ithaca, NY, USA) [47]. The clone sequences were input into MEGA 5.0. (version 5.0, www.megasoftware.net) [48], followed by alignment to construct the phylogenetic trees using neighbor-joining algorithm. The reliability of branching within the tree was tested with bootstrapping (1000 replicates), and the consensus tree was presented.

Results

Basic Soil Parameters

The application of chemical and biochar fertilizers impacted on soil physico-chemical properties (Table 1). The impacts of fertilization treatments were significant on soil NO3−-N and NH4+-N. Chemical fertilizer alone significantly increased soil NO3−-N and NH4+-N, by approximately 51.23 and 3.02 units, respectively (Table 1). High and middle biochar addition fertilizations produced more significant impacts on the soil pH, soil organic carbon (SOC), total nitrogen (TN), and NO2−-N (Table 1). Chemical as well as biochar addition fertilization treatments all significantly increased contents of total potassium (TK) by 3.88 g kg−1 and 2.98–4.94 g kg−1, respectively. However, as the increasing biochar content in biochar addition fertilization treatments, the ability of increasing on TK was gradually weak.

Table 1 Physico-chemical properties in soil following addition of the gradient inorganic fertilizer with biochar regimes
Full size table

N2O Emissions and Nitrite Reductase Activity

Chemical and biochar addition fertilization treatments significantly altered N2O flux, especially HFB treatment (Fig. 1). Meanwhile, the effects of fertilization treatments on N2O flux were similar to S-NiR. Pearson correlation analysis indicates that the N2O flux was positive correlation with S-NiR (R2 = 0.925, p < 0.01). The N2O flux (R2 = 0.960, p < 0.01) and S-NiR (R2 = 0.925, p < 0.01) were both significantly correlated with NO2−-N, reaching an extremely significant level. In addition, pH, TN, and NH4+-N also significantly influenced N2O flux and S-NiR (Supplementary Table S1).

Fig. 1
figure 1

Soil N2O flux and S-NiR under chemical as well as biochar addition fertilization treatments. CK no fertilizer control, F chemical fertilizer, LFB chemical fertilizer + 10 t hm−2 biochar, MFB chemical fertilizer + 20 t hm−2 biochar, HFB chemical fertilizer + 40 t hm−2 biochar. Error bars indicate standard error (from n = 3 replicates) for each treatment. Different letters above the bar indicate significant differences among fertilizers (p < 0.05); letters in black stand for the S-NiR; letters in gray stand for the N2O flux

Full size image

Abundance of Nir-Denitrifiers

The copy numbers of nirK-denitrifiers was one order of magnitude higher than that of nirS-denitrifiers. Chemical fertilization significantly increased the copy numbers of nirS- and nirK-denitrifiers by 14.2 and 8.8%, respectively. Biochar addition fertilization treatments significantly increased the copy numbers of nirS-denitrifiers but significantly decreased that of nirK-denitrifiers. Notably, as the content of biochar decrease in mixed fertilization treatments, the impact of biochar addition fertilization treatments on the abundance of nirK-denitrifiers was more significant (Fig. 2).

Fig. 2
figure 2

Abundances of nirS- (a) and nirK-type (b) nitrate reductase genes (denitrifier abundance) under chemical as well as biochar addition fertilization treatments. CK no fertilizer control, F chemical fertilizer, LFB chemical fertilizer + 10 t hm−2 biochar, MFB chemical fertilizer + 20 t hm−2 biochar, HFB chemical fertilizer + 40 t hm−2 biochar. Error bars indicate standard error (from n = 3 replicates) for each treatment. Different letters above the bar indicate significant differences among fertilizers (p < 0.05)

Full size image

The abundance of nirS-denitrifiers was significantly related to NO2−-N (Table 2). The abundance of nirK-denitrifiers, however, was not correlated with any of the soil properties measured. In addition, the abundance of nirS-denitrifiers was related to soil S-NiR and N2O (Table 2).

Table 2 Correlation of nirS- and nirK-denitrifiers abundance with soil physico-chemical properties, N2O flux and S-NiR
Full size table

Community Structure of Nir-Denitrifiers

Based on the relative abundances of T-RFs of nirS- and nirK-denitrifiers, the characteristic indexes of α-diversity were calculated (Table 3). Obviously, the values of Margalef index nirK-denitrifiers were more than that of nirS-denitrifiers, further supported by Heatmap (Fig. 3). Chemical as well as biochar addition fertilization treatments all changed the α-diversity of nir-denitrifiers without statistic difference. The most impact on the α-diversity of nirS- and nirK-denitrifiers was from LFB and F treatments, respectively.

Table 3 α-diversity of the T-RFs of nir-denitrifier in soil following chemical as well as biochar addition fertilization treatments
Full size table
Fig. 3
figure 3

Relative abundances of nirS- (a) and nirK- (b) terminal restriction fragments (T-RFs) in soils under chemical as well as biochar addition fertilization treatments. CK no fertilizer control, F chemical fertilizer, LFM chemical fertilizer + 10 t hm−2 biochar, MFM chemical fertilizer + 20 t hm−2 biochar, HFM chemical fertilizer + 40 t hm−2 biochar

Full size image

The T-RFLP data suggested a different community structure of nirS- and nirK-denitrifier communities between the CK, F, LFB, MFB, and HFB samples (Fig. 3), respectively. This pattern was further supported by db-RDA ordination plots, with a long straight distance between each sample for both nirS- and nirK-denitrifiers, especially the CK and F fertilizer (Fig. 5).

For the nirS-denitrifiers (Fig. 3a), sequences with T-RFs of 19, 21, and 25 bp disappeared in chemical fertilization, with T-RFs of 98 bp only being present. The addition of biochar increased the T-RFs of nirS-denitrifiers. Among biochar addition fertilization treatments, the ability to increase the T-RFs of nirS-denitrifiers was the most significant in the HFB treatment. Furthermore, the LFB and MFB treatments both increased the relative abundance of a T-RF of 16 bp, while decreasing that of 21 and 23 bp. The HFB treatment increased the relative abundance of T-RFs of 25 and 26 bp. In addition, Pearson correlation analysis indicated that N2O and NO2−-N had a significant positive relationship to T-RFs of 36 bp and 153 bp, and the latter also had a significant positive relationship to S-NiR (Supplementary Table S2).

Chemical and biochar addition fertilization treatments all significantly changed the community structure of nirK-denitrifiers (Fig. 3b). The T-RF of 16 bp only presented in the biochar fertilization treatments and showed high relative abundance. Compared with the T-RFs composition of nirK-denitrifiers in CK, the F treatment significantly increased the sequences of 105, 147,161, 192, and 215 bp, whereas biochar addition significantly decreased them. In addition, Pearson correlation analysis indicated that N2O, S-NiR, and NO2−-N had a significantly positive relationship to the T-RFs of 101 bp and 154 bp (Supplementary Table S3).

Fertilization treatments completely changed the T-RFs’ composition of nirK-denitrifiers, whereas they changed the relative abundance of the T-RFs of nirS-denitrifiers. Cluster analysis indicated that the response of nirK-denitrifiers to biochar addition fertilization treatments was more discrepancy than that of nirS-denitrifiers.

Phylogenetic Analysis of Nir-Gene Sequences

Based on similar cultured and uncultured nirS-denitrifiers from the NCBI database, the 33 nirS OTUs from five fertilizer treatments divided into three clusters (Fig. 4a). Most sequences were identified as the Rhizobiales of the Alpha-proteobacteria and the Oxalobacteraceae and Burkholderiaceae families of the Beta-proteobacteria. The T-RFs of unknown sequences were not related to any culture bacteria. The sequence T-RFs sensitive to fertilization were scattered over the whole phylogenetic tree. In detail, sequences with T-RF 36 and 153 bp were identified as Paracoccus sp., and the latter was found in only one sequence from the HFB treatment; sequences with T-RF 158 and 98 bp were both identified as Bradyrhizobium sp.; sequences with T-RF 26 bp were identified as Herbaspirillum sp., and it was not clustered with T-RF 28 bp related to Thauera sp.; sequences with T-RF 101 bp were related to Ralstonia sp. and Azoarcus sp.

Fig. 4
figure 4figure 4

Phylogenetic trees of nucleic acid of PCR-amplified nirS (a) and nirK (b) genes. Trees were constructed by neighbor-joining algorithm based on 13 (nirK) and 33 (nirS) nucleic acid OTUs. Clusters in nirS and nirK trees were both defined based on taxonomic grouping of characterized denitrifying taxa and BLAST results. The symbols in brackets stand for: ●, no fertilizer control; ○, chemical fertilizer; ▲, chemical fertilizer + 10 t hm−2 biochar; ◆, chemical fertilizer + 20 t hm−2 biochar; ■, chemical fertilizer + 40 t hm−2 biochar. The numbers in front of the symbols denote the amount of sequences in the fertilizer. The string in brackets is the accession number in the NCBI

Full size image

The 13 nirK OTUs were divided into two clusters in the phylogenetic tree (Fig. 4b). Most of the sequence T-RFs were grouped into Alpha-proteobacteria (T-RFs 36, 154, and 433 bp), which were closely identified to known Bradyrhizobium (Rhizobiales, Alpha-proteobacteria). The T-RF of 101 bp was also related to Rhizobiales, but it was identified to Sinorhizobium. The T-RF of 215 bp was identified to Pannonibacter (Rhodobacterales, Alpha-proteobacteria). Although sequences with T-RF 105 bp cluster independent sub-clusters, it is closely related to Bradyrhizobium sp.

Relationship between Nir-Denitrifiers Communities, Soil Properties, N2O, and S-NiR

The associations between environmental factors (soil physico-chemical properties, S-NiR, and N2O) and the community structure of nir-denitrifiers were calculated by db-RDA (Fig. 5). The changes of nirS-and nirK-denitrifier communities across the treatments were significantly related to soil SOC, NO3−-N, and TP, respectively (Table 4). For the nirS group, soil SOC had a higher correlation with its community structure compared to other physico-chemical properties. However, the nirK community was influenced by multiple physico-chemical properties. In addition, no matter nirS- or nirK-denitrifiers, the sample collected from HFB treatment had a high correlation with N2O.

Fig. 5
figure 5

Distance-based redundancy analysis (db-RDA) of correlation between community structures of nirS- (a) and nirK-denitrifiers (b) and soil environmental variables, N2O flux and S-NiR. The symbols on the plots stand for: ●, no fertilizer control; ○, chemical fertilizer; ▲, chemical fertilizer + 10 t hm−2 biochar; ◆, chemical fertilizer + 20 t hm−2 biochar; ■, chemical fertilizer + 40 t hm−2 biochar

Full size image
Table 4 Correlation of nirS- and nirK-denitrifiers community structures with soil physico-chemical properties, N2O flux and S-NiR
Full size table

Discussion

Effects of Biochar and Fertilizer Addition on Soil Physico-Chemical Properties

In this work, whether the amendments were chemical fertilization or biochar, the variation of pH was relatively narrow (range from 7.3–7.5) and was less than in previous studies [49, 50]. In the literature, pH changes following biochar application have resulted from the inherent differences in biochar sources and the varying buffering capacities of the soils [49]. However, in paddy soil, the water also acts as a considerable buffer [51], and a previous pot experiment adding biochar to paddy soil showed a pH range from 5.75 to 5.88, which was similar to our results [52]. The higher soil NO3−-N and NH4+-N values after chemical fertilizer application alone compared with biochar addition treatments (Table 1) confirmed the role of chemical fertilizer in contributing to dissolved organic N storage in soil. Krapfl et al. previously reported that, on increasing biochar addition content, SOC rose but TN was reduced [49], which was not completely in accordance with our results. In another previous study, Silber et al. reported that the K content of biochar is high and easily released, so adding it to soil significantly increased the content of total and available K [53]. However, high biochar content did not lead to high TK in our work. We hypothesize that high-content biochar addition has more effect of boosting rice to absorb K from soil.

Effects of Fertilizer Addition on N2O Emissions and S-NiR

Previous studies concerning the effect of biochar addition on N2O emissions showed variation according to conditions, with N2O emissions suppressed [54], promoted [55], or unaffected [56]. In this study, whether chemical or biochar addition fertilization treatments, N2O emissions were boosted, especially in the HFB treatment. Saarnio et al. reported that biochar increased soil moisture increasing N2O emissions, which was difficult to explain our result, because of flooded conditions in this work [55]. Under flooded conditions in previous studies, biochar addition had a positive effect on soil nitrate, which is the initial substrate of denitrification and easily reduced to nitrite [11, 57]. In our work, nitrite showed a significant correlation with S-NiR, which was a key stage in controlling denitrification [57]. In other words, nitrite has an important role indirectly influencing N2O. As for NH4+-N, it is worth remembering that it provides the substrate for nitrification, generating N2O as a byproduct [58]. Some previous studies also found an important role of NH4+-N in denitrification as well as pH and TN [59, 60]. As for the relationship between S-NiR and N2O, previous studies mostly regarded the abundance and community structure of nir-denitrifiers as markers to explain the relationship, either closely relating [58, 61] or having no correlation [62]. Until now, there has been no general conclusion as to whether abundance, community, or both plays the key role in generating N2O [59, 60]. The S-NiR was, therefore, aimed directly at showing the contribution of nir-denitrifiers to N2O.

Effects of Fertilizer Addition on the Abundance of Nir-Containing Denitrifiers

Recently, more and more studies have reported that the size of nirS was always greater than that of nirK in paddy soils [27], which is opposite to our results. There was no doubt that inherent defects would impact the copy number of nirS and nirK, such as primer bias and multiple gene copies in a single cell. Yoshida et al. and Wang et al. [29, 63] also supported that the nirK-denitrifiers were more abundant than nirS-denitrifiers. It does not, however, necessarily indicate that nirK-denitrifiers contributed more to N2O than nirS-ones. From the Table 2, we found the abundance of nirS-denitrifiers and not nirK-denitrifiers had a significant correlation with S-NiR and N2O. Many previous studies have also arrived at this conclusion. Espenberg et al. found that N2O emission was more primarily related to nirS-denitrifiers than nirK-denitrifiers [61], a view also supported by Dong et al. [58]. In addition, Tao et al. indicated that the contribution to S-NiR was not in the abundance of nirK (R2 = 0.0261, p = 0.6153) but rather in that of nirS (R2 = 0.528, p = 0.0074) [60].

From Table 2, it is notable that soil properties have little significant effect on the abundance of the nirK-denitrifiers. Although many previous studies have indicated that nirK-denitrifiers are more sensitive to environmental changes than nirS-denitrifiers [64, 65], we guess that, over time, waterlogging and eutrophication would decrease the sensitivity of nirK-denitrifiers. With long-term irrigation regimes, soil properties were not only uncorrelated with the abundance of nirK-denitrifiers but also their diversity and community structure [66]. Lee et al. also reported that the nirS-denitrifier community, rather than the nirK-one, should be the optimal indicator of microbial nitrite reduction processes in an estuarine system [67].

From previous studies, compared with controls, organic fertilizers have boosted the abundance of nir-denitrifiers [58, 60]. However, in our experiments, biochar addition inhibited the abundance of nirK-denitrifiers and increased nirS-denitrifiers. We found that measured soil properties cannot influence the abundance of nirK-denitrifiers (see Table 2). In the literature, dissolved oxygen was not a routine soil property but an important factor determining denitrifier abundance [57]. Li et al. indicated that a soil micro-environment with low dissolved oxygen was more suitable for nirK-denitrifiers, whereas there was no spatially heterogeneous variation for the nirS-denitrifier [68]. Huanhuan et al. also found that nirK-denitrifiers prefer deeper soils [69]. In our study, biochar was an abundant pore material, which would have the effect of increasing dissolved oxygen [70]. This may possibly explain the decrease in abundance of nirK-denitrifiers. To summarize, biochar influenced the abundance of nirS-denitrifiers through physico-chemical properties, whereas it possibly influenced that of nirK-denitrifiers through pore structure.

Effects of Biochar and Fertilizer Addition on Nir-Denitrifier Community Structure

From the abundance of nir-denitrifier results, we have found that the abundance of nirS-denitrifiers was more sensitive to environmental change than that of nirK-denitrifiers. However, the community structure seems to show the opposite result. Both the Shannon index (Table 3) and the Heatmap result (Fig. 3) showed that the diversity of nirK-denitrifiers was more sensitive to the variation of fertilizers than that of nirS-denitrifiers. In relation to the discrepancy of environmental sensitivity between nir-denitrifiers’ abundance and community structure, Yin et al. have reported that the abundance of nirS-denitrifiers seems to be more sensitive than that of nirK-denitrifiers, while the community structure appears to be the opposite [30]. Certainly, the opposite result also has been reported. Yang et al. showed that the response of nirS-denitrifiers community to the gradient volume of N fertilization was more sensitive compared with that of nirK-denitifiers [59]. However, Luo et al. reported that N fertilizer had a clear effect on the community structure of nirK-denitrifiers but not the nirS-containing community [71]. Until now, few papers have examined the relationship between the nir-denitrifier community and gradient volumes of biochar addition. Consequently, we were unable to find any papers supporting our result that the LFB treatment obviously influenced the community structure of nirK-denitrifiers. We hypothesize that this phenomenon was correlated with the volume of dissolved oxygen carried from biochar into the soil.

Based on Figs. 3 and 4, we found that, with both nirS- and nirK-denitrifier communities, the dominant species were not playing a key role in controlling N2O emission and S-NiR activity, but minority species were. In previous studies, the denitrifiers belonging to Rhodobacterales and Rhizobiales were commonly present [28,29,30]. In our work, the nirS-denitrifiers Paracoccus (sequence T-RF 153 bp) and the nirK-denitrifiers Sinorhizobium (sequence T-RF 101 bp) seem to the key genera relate to N2O emission.

Previous papers stated that certain Herbaspirillum (Burkholderiales) strains and Azospirillum (Rhodospirillales) had the ability to emit N2O [72, 73]. Yamane et al. reported that, in pure culture, Paracoccus have a significant ability of emitting N2O [73]. From the above works, we have known that the NO2−-N had a significant correlation with N2O flux and nirS-denitrifiers, and the amount of NO2−-N and N2O emission both were significantly increased in HFB treatment. Previous studies have reported that Paracoccus denitrificans [74] and Paracoccus pantotrophus [39] can perform aerobic denitrification. Furthermore P. denitrificans can rapidly reduce NO2−-N and was able to grow aerobically [75]. This may explain why Paracoccus was related to N2O emission in the HFB treatment.

Notably, we found that the trend of abundance of nirK-denitrifiers was similar to the abundance of T-RF 101 bp, namely, Sinorhizobium (Figs. 2b and 3b). A previous study has reported that Sinorhizobium with the complete set of denitrification genes does not grow under anaerobic conditions with nitrate or nitrite as terminal electron acceptors [76], whereas under micro-oxic conditions it would use nitrate or nitrite as respiratory substrates [77]. In this work, combining the abundance of nirK-harboring Sinorhizobium (Figs. 1 and 3b), we guessed that the volume of oxygen in the HFB regime satisfied the demands of Sinorhizobium, and boosted it to use nitrate or nitrite as terminal electron acceptors, generating N2O emission.

As for phylogenetic analysis, species types of nirS-denitrifiers were more diverse than nirK-ones in paddy soil. Yoshida et al. also explored the difference of nirS- and nirK-denitrifiers in paddy soil and got the same result [29]. Furthermore, we have previously researched another soil, the calcareous purple paddy soil, and found that nirK gene was hardly amplified in it [78].

Correlation between Nir-Denitrifier Community Structure and Properties

There is no doubt that environmental factors have a considerable effect on shaping nir-denitrifiers community structure [79, 80]. Yin et al. indicated that the community structure of two types of nir-denitrifiers responds differently to denitrifying enzyme activity, and the nirK-denitrifiers community was the true regulators [30]. The opposite result is also recorded, in which the N2O emission was primarily related to the community structure of nirS-denitrifiers [61]. However, both the community structure of nirK- and nirS- denitrifiers were not related to S-NiR and N2O in our work. Our findings were in line with the results of Dandie et al. who observed that the differences in denitrifier community composition were not related to denitrification rates and N2O emissions [62].

As pointed out previously, soil NO3−-N and SOC can serve as metabolic substrates, directly or indirectly influencing the denitrifying bacterial communities [59, 61]. It has also been shown that changes of soil phosphorus content were related to a structural shift in the nir-containing denitrifiers community [81]. In addition, there has been indication that the community structure of P-dissolving bacteria was related to soil nitrogen [82]. Therefore, the cycles of phosphorus and nitrogen are not absolutely independent. However, the mechanism of phosphorus intervening in the activity of denitrifiers is not well known. In our work, we showed that phosphorus plays a key role in shaping nir-denitrifiers community, but the exact evidence cannot be presented. In the future, this will be a large area of the research needed to understand denitrification.

Conclusions

Compared with chemical fertilizer, although the LFB and MFB treatments not significantly decreased denitrifying S-NiR and N2O emission, the HFB treatment significantly increased the S-NiR and emit N2O in paddy soil. The abundance of nir-denitrifiers was responsible for S-NiR and N2O emission, rather than the variation of community structure. The NO2−-N was a key factor of controlling the abundance of nir-denitrifiers, S-NiR and N2O emission. Although the abundance of nirK-denitrifiers was more abundant than that of nirS-denitrifiers, S-NiR and N2O emission showed no significant correlation to the abundance of the former but to the later one. Analysis of T-RFLP and phylogenetic tree data showed that the biochar additions greatly affected the richness, diversity, and structure of nirS- and nirK-denitrifier communities, especially nirK-denitrifers. In addition, the nirS-denitrifier Paracoccus and nirK- denitrifier Sinorhizobium may be a key species of emitting N2O. Community structure was primarily influenced by SOC, NO3−-N, and TP. Our findings suggest that, although nirS- and nirK-denitrifiers both mediate nitrite reductase to influence N2O flux, their responds to gradient volume biochar addition were significant discrepancy.