The Structure and Diversity of Nitrogen Functional Groups from Different Cropping Systems in Yellow River Delta

The Yellow River Delta (YRD) region is an important production base in Shandong Province. It encompasses an array of diversified crop systems, including the corn–wheat rotation system (Wheat–Corn), soybean–corn rotation system (Soybean–Corn), fruits or vegetables system (Fruit), cotton system (Cotton) and rice system (Rice). In this study, the communities of ammonia oxidizer–, denitrifier– and nitrogen (N)–fixing bacteria in those cropping systems were investigated by Illumina Miseq sequencing. We found that Rice soil exhibited significantly higher diversity indices of investigated N–cycling microbial communities than other crop soils, possibly due to its high soil water content. Wheat–Corn soils had higher abundances of nitrification gene amoA and denitrification genes nirK and nirS, and exhibited higher soil potential nitrification rate (PNR), compared with Soybean–Corn, Cotton and Fruit soils. Consistently, redundancy analysis (RDA) showed that soil water content (SWC), electrical conductivity (EC), and total nitrogen (TN) were the most important influencing factors of the diversity and structure of the investigated N–cycling microbial.


Introduction
The Yellow River Delta (YRD) is one of the fastest growing deltas in the world, with great potential for land exploitation and utilization [1]. Soil salinity and alkalinity in the YRD has limited the land exploitation and agricultural development [2]. In order to alleviate the negative effects of salt-alkaline and to improve crop production, considerable nitrogen (N) fertilizers (about 1.6 × 10 5 ton/year) have been used in YRD [3]. However, only 27-40% of applied N is taken up by crops [4]. The retained N is lost in soil through nitrate leaching, runoff and N oxides emission [5], resulting in nitrate contamination and green house pollution [6][7][8]. Therefore, improving N use efficiency has become a significant concern for agricultural sustainable development in the YRD.
Soil N use efficiency is closely related to N-cycling associated microbial communities that perform N-fixation, nitrification and denitrification [9]. Nitrogen fixers convert nitrogen gas into biologically assimilable compounds that can be taken up by crops. In intercropping systems, the nitrogen activeness of soybean nodule bacterium can reduce N fertilizer input and consequently increase N use efficiency [10]. Ammonia-oxidizer and nitrifier oxidize ammonia into nitrite and diagonal lines and completely mixed as a single sample at each site. In total, corn-wheat rotation in 25 samples (Corn-Wheat, n = 5 sites), soybean-corn rotation in 25 samples (Soybean-Corn, n = 5 sites), cotton 20 samples (Cotton, n = 4 sites), fruit or vegetables in 20 samples (Fruit, n = 4 sites), and rice in 20 samples (Rice, n = 4 sites), were collected across the delta in June 2017. The corn-wheat rotation system, soybean-corn rotation system, cotton system, fruits or vegetables system, and rice system have been planted for approximately 7, over 20, 20, 7 and over 20 years, respectively ( Figure  1). Information on fertilizer utilization is shown in Supplementary Table 1. Samples were placed into an ice box and brought to the laboratory immediately after collection. One portion of the samples was air-dried at 25° C, grinded in a mortar and sieved through successively finer meshes to obtain a 2 mm fraction for the determination of soil pH, a 0.25 mm fraction for analysis of electrical conductivity (EC), total nitrogen (TN) and ammonium contents (NH4 + -N), nitrate concentrations (NO3 --N), available phosphorus (AP) and available potassium (AK), and a 0.149-mm fraction for measurement of effective cation exchange capacity (CEC) and soil organic matter (SOM). Other portions were frozen at -80° C until further DNA extraction. i.e., wheat-corn rotation (Corn-Wheat, sites 1-5), cotton (Cotton, sites 6-10), fruits or vegetables (Fruit, sites [11][12][13][14], rice (Rice, sites [15][16][17][18], soybean-corn rotation (Soybean-Corn, sites [19][20][21][22].

Analysis of Soil Properties
Soil pH and EC were measured in a 1:2.5 water suspension mixture using a glass electrode pH meter (PHS-2F, INESA, Shanghai, China). The soil water content (SWC) was calculated based on the losses of 20.00 g of fresh soil dried to a constant weight in an oven at 105 • C [27]. SOM was determined using the Walkley-Black method [28]. The contents of NH 4 + -N and NO 3 − -N in the soil were extracted with 2 M KCl and determined on a continuous segmented flow analyser (AutoAnalyzer III, SEAL Analytical, Fareham, UK). Soil TN was detected via the semi-micro Kjeldahl digestion method (Automatic Kjeldahl Apparatus K9860, Hanon, Jinan, China) [29]. Soil AP was estimated through extraction with sodium bicarbonate and measured on a spectrophotometer (UV -5500PC, Shjingmi, Shanghai, China) [30]. AK was extracted with 1 M ammonium acetate and measured on an Inductively Coupled Plasma-Optical Emission Spectrometer (iCAP-7000, Thermo Fisher, Waltham, MA, USA). The value of CEC was determined using a solution of barium chloride (ISO 11260-1994). To control the analytical quality, reference soil (GBW07413a) from the National Research Centre for the Certified Reference Materials of China was used in the detection processes of all indices [31]. The detected values of the standard samples fell within their accepted ranges, with a deviation of less than 5%.

Soil DNA Extraction and Real-Time Quantitative PCR
DNA was extracted from 0.25 or 0.5 g (Rice) of fresh soil using the DNeasy PowerSoil Kit (MO BIO Laboratories, Carlsbad, CA, USA) following the manufacturer's protocol. The quality of extracted DNA was examined on a Nano Drop ® Spectrophotometer ND-1000 (Thermo Fisher Scientific, Waltham, MA, USA), and it was then stored at −20 • C before amplification.
The copy numbers of the five N-cycling genes were determined by Real-time quantitative PCR using FTC-3000 (Funglyn Biotech Co., Richmond Hill, Ontario, Cananda). The abundance of AOA-amoA, AOB-amoA, nirK, nirS and nifH genes were quantified with Arch-amoAF/Arch-amoAR, amoA-1F/amoA-2R, FlaCu/R3Cu, cd3aF/R3cd, and pol-F/pol-R, respectively (Table 1). Real-time quantitative PCR was measured using the SYBR Premix Ex Taq TM kit (Takara Bio Inc., Dalian, China) in a 25 µL reaction system consisting of 12.5 µL 2× SRBR Premix Ex Taq TM , 2 µL forward and reverse primers (10 µM), 5 µL template DNA and 5.5 µL RNase-free H 2 O. The AOA-amoA gene was amplified under the following conditions: initial melting step at 95 • C for 30 s, 40 cycles of 95 • C for 10 s, 50 • C for 30 s, and 72 • C for 40 s. The AOB-amoA, nirS and nifH genes was amplified under the same conditions: initial melting step at 95 • C for 30 s, 40 cycles of 95 • C for 10 s, 55 • C for 30 s, and 72 • C for 30 s. The nirK gene was amplified as follows: initial melting step at 95 • C for 30 s, 40 cycles of 95 • C for 10 s, 60 • C for 30 s, and 72 • C for 30 s. Each sample had three replicates. Standard curves were created by seven-fold serial dilutions of plasmids containing the corresponding gene. The linear coefficient correlation (R 2 ) of the standard curves were 0.9998 for AOA-amoA, 0.9999 for AOB-amoA, 0.9999 for nifH, 0.9992 for nirK, 0.9992 for nirS.

High-Throughput Sequencing and Data Analysis
The V3-V4 regions of bacteria (16S rRNA) and five N-cycling genes (amoA-AOA, ammonia oxidation archaea; amoA-AOB, ammonia oxidation bacteria; nirS and nirK, nitrite reduction bacteria; nifH, N-fixation bacteria) were amplified using different primer pairs targeting specific genetic regions or genes. The amplification primer pairs and thermal cycling conditions of bacteria 16S rRNA and functional N-cycling genes are listed in Table 1. Each 50 µL PCR reaction system was prepared using 10 µL 5x Buffer, 1 µL dNTP (10 mM), 1 U Phusion ® High-Fidelity DNA Polymerase, 1 µL (10 µM) each of forward and reverse primers, and 5-50 ng of template DNA supplemented with ddH 2 O. The PCR products were purified using the High Pure PCR Product Purification Kit (Roche Applied Science, Meylan, France). Purified PCR products were well mixed to establish sequencing libraries after further qualification, and the libraries were sequenced on an Illumina MiSeq platform (TinyGene Inc., Shanghai, China).
Raw data were filtered by Trimmomatic (V 0.32). Reads were truncated when the average quality score was less than 20 over a 50-base window. Trimmed reads shorter than 50 bp were also discarded. Then, forward and reverse reads were merged using FLASH with a 10 bp minimum overlap and a maximum mismatch density of 0.2. Mothur (V 1.39.5) was used to filter the reads that were ambiguous and homologous with the parameters: maxambig = 0, maxhomop = 8, minlength = 200 (for 16S rRNA, amoA-AOB, nifH, nirK and nirS)/150 (for amoA-AOA), maxlength = 580. Operational taxonomic units (OTUs) with a 97% similarity cutoff were clustered using UPARSE (V 7.1) and chimeric sequences were removed by UCHIME. The taxonomy of 16S rRNA was predicted against the SILVA 128 database using Mothur (V 1.39.5); the taxonomies of amoA, nifH, nirK and nirS were selected against the NCBI database. The differences in diversity and richness of the bacterial and chosen N-cycling functional groups amongst the five cropping systems were examined by using multiple independent sample non-parametric tests (Kruskal-Wallis H test). For β-diversity, to determine the differences of soil microbial community structure amongst different cropping systems, we performed Non-Metric Multidimensional Scaling (NMDS) using the Multiple Response Permutation Procedure (MRPP) from the library "vegan" in R (V 3.5.0).
To quantify the relationship between soil properties and bacteria and N-cycling functional genera, Spearman correlation coefficients were calculated using SPSS Statistics (V 19.0). We performed a redundancy analysis (RDA) based on the soil properties and functional N-cycling microbes at the genus level to determine the main environmental factors affecting the N-cycling processes in each cropping system using the Canoco software (V 5.0).
To investigate the relationship of dominant OTUs amongst the five cropping systems, Co-occurrence networks were performed. Firstly, we defined the "dominant" OTUs with relative abundances > 0.5% within at least one replicate in each cropping system. Then, the matrix was established based on the Pearson correlation coefficient using the relative abundance of each "dominant" OTU. The network was conducted through employing a threshold (Pearson correlation coefficient |r| > 0.85, p < 0.05). The network structure was visualized with Cytoscape (V 3.6.0). In the network structure, node values represent the relative abundances of OTUs, and the edge sizes indicate the Pearson correlation coefficients between each two OTUs. And the network parameters including the number of edges, the number of nodes, clustering coefficients, and network density were identified using the Network Analysis method.

Accession Numbers
Nucleotide sequence data were submitted to the National Centre for Biotechnology Information (NCBI) Sequence Read Archive (SRA) with accession number SRP152899.

Soil Physicochemical Properties
Differences between the soil physicochemical properties of the five cropping systems were shown in Table 2. Rice soils contained a significantly higher (p < 0.001) concentration of SWC, reaching 38.1 ± 4.4% compared to the other four cropping systems. This could be explained by the higher water content of rice habitats compared with other crops. The highest AP (71.7 mg kg −1 ) and AK (0.37 g kg −1 ) were found in Fruit soils, and significantly differed (p < 0.05) from others. There were no significant differences in the contents of SOM, TN, NH 4 + -N, NO 3 − -N, EC and CEC amongst the five cropping systems.

Diversity of Bacteria and N-Cycling Functional Groups in the Five Cropping Systems
α-diversity: The coverage of all tested soil samples was more than 99% (Table 3), which indicated that there was a sufficient level of Illumina MiSeq sequencing to identify most diversity of soil samples. Rice soils had the highest OTUs of bacteria, followed by ammonia-oxidizer, denitrifier and N-fixing bacteria. The Chao and ACE indices showed a similar pattern to those of the OTUs, i.e., rice soils had the highest Chao and ACE indices of ammonia-oxidizer, denitrifier and N-fixing bacteria, whilst no significant differences in the Shannon index existed amongst the five cropping systems. β-diversity: The compositions of bacteria, ammonia-oxidizers and denitrifying bacteria in rice soils differed to those of other soils, and no overlap was present based on the NMDS analysis (Figure 2a-c). This was supported by the Multi Response Permutation Procedure (MRPP) analysis (Table 4), showing that the community compositions of soil bacteria, ammonia oxidizer amoA-AOA and denitrifier nirS significantly differed amongst the five cropping systems. The community composition of N-fixing bacteria in the five cropping systems had a large area of overlap (Figure 2d), indicating that there was no difference. Additionally, MRPP showed no significant differences in the community composition of N-fixing bacteria amongst the five cropping systems.   The A values are the chance corrected within-group agreements; p is the probability of a smaller or equal δ; * significance at the p < 0.05 level; ** significance at the p < 0.01 level.  The A values are the chance corrected within-group agreements; p is the probability of a smaller or equal δ; * significance at the p < 0.05 level; ** significance at the p < 0.01 level.

Composition and Difference in the Bacterial Community
A total of 55 phyla were identified in all tested soils. The phyla Proteobacteria, Acidobacteria, Chloroflexi and Gemmatimonadetes were the most abundant phyla (> 5% of all sequences in each cropping system) in the soil of Wheat-Corn, Cotton, Fruit and Soybean-Corn, with the range of 30.5-33.7%, 11.0-23.1%, 10.2-13.5% and 5.1-10.3%, respectively ( Figure 3A). Moreover, we found that the relative abundance of Bacteroidetes was significantly higher, whilst the Nitrospirae was significantly lower in Rice soils compared with other soils (Figure 4).

Composition and Difference in the Bacterial Community
A total of 55 phyla were identified in all tested soils. The phyla Proteobacteria, Acidobacteria, Chloroflexi and Gemmatimonadetes were the most abundant phyla (> 5% of all sequences in each cropping system) in the soil of Wheat-Corn, Cotton, Fruit and Soybean-Corn, with the range of 30.5%-33.7%, 11.0%-23.1%, 10.2%-13.5% and 5.1%-10.3%, respectively ( Figure 3A). Moreover, we found that the relative abundance of Bacteroidetes was significantly higher, whilst the Nitrospirae was significantly lower in Rice soils compared with other soils (Figure 4).

Abundance of N-Cycling Genes
The abundance of N-cycling genes, including AOA-amoA, AOB-amoA, nirK, nirS and nifH genes were determined. As shown in Figure 5, the log 10 of the five genes' copy number ranged from 5.8 to 7.2 g −1 soil for AOA-amoA, 6.5 to 7.2 g −1 soil for AOB-amoA, 5.1 to 6.1 g −1 soil for nirK, 6.5 to 7.3 g −1 soil for nirS, 5.8 to 7.3 g −1 soil for nifH, respectively. Wheat-Corn soils had significantly higher abundances of ammonia oxidizer amoA and denitrification nirK, whilst Rice soils had lower abundance of AOA-amoA and higher denitrification genes (nirK and nirS) abundance, compared to the Cotton, Fruit and Soybean-Corn soils. The abundances of nifH were significantly higher in Rice samples than those of the other four cropping system soils (p < 0.05).

Abundance of N-Cycling Genes
The abundance of N-cycling genes, including AOA-amoA, AOB-amoA, nirK, nirS and nifH genes were determined. As shown in Figure 5, the log 10 of the five genes' copy number ranged from 5.8 to 7.2 g −1 soil for AOA-amoA, 6.5 to 7.2 g −1 soil for AOB-amoA, 5.1 to 6.1 g −1 soil for nirK, 6.5 to 7.3 g −1 soil for nirS, 5.8 to 7.3 g −1 soil for nifH, respectively. Wheat-Corn soils had significantly higher abundances of ammonia oxidizer amoA and denitrification nirK, whilst Rice soils had lower abundance of AOA-amoA and higher denitrification genes (nirK and nirS) abundance, compared to the Cotton, Fruit and Soybean-Corn soils. The abundances of nifH were significantly higher in Rice samples than those of the other four cropping system soils (p < 0.05). Microorganisms 2019, 7, x FOR PEER REVIEW 14 of 25 One-way ANOVA was used to determine the difference in the abundance of nitrogen functional genes. Statistically significant differences are noted by different letters. In Figure 5A, different capital letters indicate significant differences of AOB amoA abundance, while different lowercase letters indicate significant differences of AOA amoA abundance. In Figure 5C, different capital letters indicate significant differences of nirK abundance, while different lowercase letters indicate significant differences of nirS abundance.

Potential Nitrification Rate (PNR)
The PNR of the different cropping system soils was shown in Figure 6. Soil PNR showed no significant differences in Wheat-Corn, Soybean-Corn, Cotton and Fruit. However, the soil PNR was lower in Rice soils (p < 0.05). Generally, PNR represents the ability of ammonia oxidizer (AOA and AOB) in the soil to transform NH4 + -N to NO2 − -N. Therefore, lower PNR values indicate a lower activity of soil ammonia oxidizer in Rice soils, compared to the other four cropping systems. in the five cropping systems: wheat corn rotation (Corn-Wheat), cotton (Cotton), rice (Rice), fruits or vegetables (Fruit), soybean corn rotation (Soybean-Corn). One-way ANOVA was used to determine the difference in the abundance of nitrogen functional genes. Statistically significant differences are noted by different letters. In Figure 5A, different capital letters indicate significant differences of AOB amoA abundance, while different lowercase letters indicate significant differences of AOA amoA abundance. In Figure 5C, different capital letters indicate significant differences of nirK abundance, while different lowercase letters indicate significant differences of nirS abundance.

Potential Nitrification Rate (PNR)
The PNR of the different cropping system soils was shown in Figure 6. Soil PNR showed no significant differences in Wheat-Corn, Soybean-Corn, Cotton and Fruit. However, the soil PNR was lower in Rice soils (p < 0.05). Generally, PNR represents the ability of ammonia oxidizer (AOA and AOB) in the soil to transform NH4 + -N to NO 2 − -N. Therefore, lower PNR values indicate a lower activity of soil ammonia oxidizer in Rice soils, compared to the other four cropping systems.
Microorganisms 2019, 7, x FOR PEER REVIEW 15 of 25 Figure 6. Soil PNR in the five cropping systems: wheat corn rotation (Corn-Wheat), cotton (Cotton), rice (Rice), fruits or vegetables (Fruit), soybean corn rotation (Soybean-Corn). One-way ANOVA was used to determine differences in soil PNR across the different cropping systems. Statistically significant differences are denoted by different letters.

Composition of N-Cycling Functional Communities
Of the genera in the well-identified AOA groups, Nitrososphaere from phylum Thaumatchaeota was the most abundant genus in the five agroecosystems, with a content of 25.3%-29.9%, followed by Nitrosopumilus (3.6%-16.1%) and Candidatus_Nitrosocosmicus (0%-3.4%) ( Table 5). For AOB members, Nitrosospira, Nitrosomonas, Nitrosovibrio and Roseateles were the four most abundant genera Figure 6. Soil PNR in the five cropping systems: wheat corn rotation (Corn-Wheat), cotton (Cotton), rice (Rice), fruits or vegetables (Fruit), soybean corn rotation (Soybean-Corn). One-way ANOVA was used to determine differences in soil PNR across the different cropping systems. Statistically significant differences are denoted by different letters.

Co-Ocurrence Network Analysis of amoA-AOA and nirS-Denitrifier
The first stage of nitrification involves the oxidation of ammonia to nitrite, which is catalysed by ammonia monooxygenase (AMO) [23]. The genes nirS and nirK, as marker genes of denitrifier [33], catalyse the reduction of nitrite to nitric oxide. Nitrification and denitrification are important activities in during the loss of ammonium-based fertilisers [34,35]. Based on MRPP analysis, the compositions of ammonia oxidizer amoA-AOA and nirS denitrifier were significantly different amongst the five cropping system soils (Table 4). A network analysis was therefore used to explore the differences of amoA-AOA and nirS-type denitrifier communities in the different cropping systems (Figures 7 and 8), to a better understand N-cycling functional community structures. by ammonia monooxygenase (AMO) [23]. The genes nirS and nirK, as marker genes of denitrifier [33], catalyse the reduction of nitrite to nitric oxide. Nitrification and denitrification are important activities in during the loss of ammonium-based fertilisers [34,35]. Based on MRPP analysis, the compositions of ammonia oxidizer amoA-AOA and nirS denitrifier were significantly different amongst the five cropping system soils (Table 4). A network analysis was therefore used to explore the differences of amoA-AOA and nirS-type denitrifier communities in the different cropping systems (Figures 7 and 8), to a better understand N-cycling functional community structures.  The network parameters of the two communities were shown in Supplementary Table 2. The node and edge of amoA-AOA in the Wheat-Corn and Cotton soils were higher than those of other cropping systems. The edges of nirS-denitrifier in Fruit soils were higher than those of other soils. However, the network density of amoA-AOA in Wheat-Corn soil was higher than that of other soils. The network parameters of the two communities were shown in Supplementary Table S2. The node and edge of amoA-AOA in the Wheat-Corn and Cotton soils were higher than those of other cropping systems. The edges of nirS-denitrifier in Fruit soils were higher than those of other soils. However, the network density of amoA-AOA in Wheat-Corn soil was higher than that of other soils. For the nirS-type denitrifier networks, the network density in Rice was higher than that of other soils.

Key Drivers of Bacterial and N-Cycling Functional Communities
In this study, Rice soils were well separated through RDA analysis (Figure 9), indicating that the composition of the chosen N-cycling functional groups in Rice soil differed to those of other soils. Moreover, RDA analysis showed that the environmental factors, SWC, EC, TN, AP and CEC were important and influential factors of bacterial and the chosen N-cycling functional groups (Table 6). Amongst them, SWC could explain 21.6-54.2% of the variation in the functional communities of different cropping systems, and was the most important influencing factor ( Table 6). were important and influential factors of bacterial and the chosen N-cycling functional groups ( Table 6). Amongst them, SWC could explain 21.6%-54.2% of the variation in the functional communities of different cropping systems, and was the most important influencing factor ( Table 6).   EC and TN were also related to the ammonia oxidizer AOA and AOB; EC and AP were related to denitrifier nirS and nirK; and CEC was related to nifH. EC explained 15.3% and 13.7% of the variation of ammonia oxidizer and denitrifier compositions, respectively. TN, AP and CEC were minor drivers of bacterial and N-cycling functional communities, which were related to ammonia oxidizer, denitrifier and nifH-type fixing bacteria, respectively. Table 6. Results for RDA testing effects of soil properties on the community structures of ammonia-oxidizers (amoA), denitrifying bacteria (nirS and nirK), N-fixing bacteria (nifH), and total bacteria (16S rRNA).

Soil Properties
Degrees of Explanation (%) SWC, AP, CEC, EC, TN, and NO 3 − -N represent the water content, available phosphorous, effective cation exchange capacity, electrical conductivity, total nitrogen, and nitrate contents, respectively. * significant at the p < 0.05 level; ** significant at the p < 0.01 level; NS represents no significant.

Abundance and Diversity of Chosen N-Cycling Functional Genes in Different Cropping Systems
In this study, we investigated the communities of soil bacteria, ammonia oxidizer, denitrifier and nitrogen-fixing bacteria by high throughput sequencing. High throughput sequencing reveals the presence of organisms containing these genes but their viability and metabolic activation have not been assessed. Nevertheless, the method has been more widely used to analyze microbial communities than cultivation-dependent method, as less than 1% of the bacterial taxa can be cultivated [36].
Here, we demonstrate that Wheat-Corn rotation soil had the most abundant nitrification gene amoA and relatively high PNR activity. Additionally, the soil had the highest microbial network density of amoA-AOA. Microbial network density reflects microbial interconnectivity, which likely affect microbial functions [37]. Collectively, the data suggest that Wheat-Corn rotation soil had higher ammonia oxidization activity than other cropping systems. This phenomenon could be caused by the application of N fertilizer, since the abundance of ammonia oxidizer positively correlated with ammonia availability in agricultural soil [38,39]. Consistently, Wheat-Corn rotation soil is supplemented with an average of compound fertilizer 600 kg h −1 year −1 and diammonium hydrogen phosphate 550 kg h −1 year −1 , which is higher than other soils (Supplementary Table S1). Moreover, Wheat-Corn rotation soil has high abundances of denitrification genes (nirK and nirS), suggesting that it contains a higher number of denitrification microbes. We reasoned that the NO 3 − produced by ammonia oxidizer and nitrite oxidizing bacteria, could promote the growth of denitrifier. Noticeably, a reduction in NO 3 − by denitrifier leads to N 2 O, which is a powerful greenhouse gas. Thus, the excessive application of N fertilizer increases the abundance of ammonia oxidizer and denitrifer, consequently enhancing NO 3 − leaching and N 2 O emission. Therefore, suppression of ammonia oxidizer would be an effective approach to increase nitrogen fertilizer use efficiency. We further reveal that AOB was more abundant than AOA in Wheat-Corn soils. AOB is favoured by inorganic fertilization and produces high N 2 O emission, whilst AOA is favoured by organic or slow-release fertilizer, yielding lower N 2 O emission [40]. Together, we deduce that Wheat-Corn soils have higher N 2 O emissions and consequently lower fertilizer use efficiency. The application of nitrification inhibitor and slow-release fertilizer should be taken into consideration in the Wheat-Corn cropping system to suppress ammonia oxidization and decrease AOB abundance, thereby decreasing N 2 O emission and improving N use efficiency.
Rice soil contained the most abundant nitrogenase gene nifH and denitrification genes (nirK and nirS), due to their high water content (discussed in Section 4.2). Rice soil also exhibited higher diversity of amoA-AOA and nirS communities than other cropping systems according to ACE and Chao analysis. However, the results contrasted those of microbial network density analysis, revealing that the nodes and edges of the networks of Rice soils were less than those of Wheat-Corn, Cotton and Fruit soils.
Since the number of edges reflect the degree of microbial diversity [41,42], the microbial network density analysis suggested simplified amoA-AOA and nirS communities in Rice and Soybean-Corn soils. The reason for these differences may result from the network density analysis only using "dominant" OTUs with a relative abundance > 0.5%. These data suggest that Rice soil has a higher diversity but fewer dominant species of amoA-AOA and nirS communities.
It should be realized that PCR primer design is based on current sequence databases and microorganism habitats, and that on primer sets can achieve a comprehensive coverage of natural sequences or unbiased amplifications [43]. In this study, the primer sets Arch amoA-F/Arch amoA-R for targeting AOA and amoA-1F/ amoA-2R for targeting AOB have limitations in terms of the comprehensive coverage of Nitrosotalea subclade 1.1 and Nitrosospira 8A clade, respectively. However, this may reduce the amplification bias of some minor populations of ammonia oxidizer members [44]. For nir genes, the conventional primers F1aCu/R3Cu and cd3aF/R3cd closely match the sequences of Cu-containing nitrite reductase (nirK) and cytochrome cd1 nitrite reductase (nirS) respectively [26,45]. The primers preferentially target subclasses of Proteobacteria [46] but fail to amplify previously unconsidered nir genes in newly discovered microorganisms [34]. In addition, the nifH primer set PolF/PolR can amplify three of the four clusters of nifH paralogs. The cluster IV/V of nifH that cannot be identified by PolF/PolR skews the diversity analysis and their sequences can be filtered before diversity analysis [47]. Moreover, the primer set PolF/PolR has lower template-specific bias in qPCR application than other primer sets, thereby more accurately evaluates nifH copy numbers [48]. The further development of clade-specific and habitat-specific primers is important to reveal the diversity and abundance of environmental microorganisms. Developing primers with a higher coverage and specificity of N-cycling genes in the saline-alkali soil of YRD remains a goal of further studies.

Factor Affecting the Diversity and Composition of Bacterial and N-Cycling Functional Communities
According to RDA analysis (Table 6), SWC was the most important influencing factor for the variation of bacterial and targeted N-cycling functional communities. The habitat with high water content possesses more functionally diverse microbial communities, maintaining the activity of soil enzymes [49,50]. Additionally, high SWC can alleviate the negative effects of soil salinization on crop growth, as the irrigation water can push salts below the root zone [51].
SWC had a significant effect on the total bacterial community (p < 0.05) ( Table 6). Specifically, the relative abundance of bacterial phylum Bacteroidetes was positively related to SWC (Spearman's correlation coefficients = 0.469, p < 0.01), and the phylum was significantly higher in Rice soil than other crop type soils ( Figure 4). It is well recognised that phylum Bacteroidetes prominently inhabits water-based environments [52]. For diazotrophs, the abundance of nifH was higher in Rice soils than other cropping soils (p < 0.05) ( Figure 5B), and the SWC was positively related to N fixing bacterial community. This phenomenon could be explained by the higher SWC that can enhance the diffusive transport of organic substrates, which may indirectly increase the access of organic C substrates to N fixing bacteria [11]. Furthermore, SWC has a significant effect on denitrifier [53]. The abundances of denitrification nirK and nirS were higher in Rice habitat with high SWC. That is because the rice habitat with high SWC can provide an anaerobic environment for denitrifying microbe where the abundance and diversity of denitrifying genes remains relatively high [54].
Additionally, soil TN was an important factor driving the composition of ammonia oxidizer ( Table 6). The abundances of AOA and AOB were higher in Wheat-Corn soils with high fertilizer level. The results agreed with that abundances of AOA and AOB were positively related to ammonia availability in agricultural soil [38,39]. In addition, soil TN had no significant effect on the structure of N fixing bacterial communities in our study (p > 0.05) but could reduce nifH abundance. These data are consistent with that the activity of nifH community can be suppressed when nitrogen is abundant [55].
Here, we demonstrate that the studied N-cycling microbial communities were affected by the cropping strategy and soil properties, such as SWC and TN. These factors also influence N use efficiency [13]. The results support that optimizing soil water management and nitrogen fertilizer is a strategy to improve the N use efficiency in agroecosystem of YRD. Future studies should focus on the relationship between N-cycling microbial communities and N cycle activities, such as ammonium oxidation, denitrification and N fixation in the crop systems to reveal the contribution of N-cycling microbes to the N cycle.

Conclusions
This study has clarified the effects of different cropping systems in the YRD on the abundance and diversity of selected N functional genes, including the nitrification gene amoA, the denitrification genes nirK and nirS, and the nitrogenase gene nifH. We explored the contribution of soil properties to the abundance and diversity of these genes. We found that (1) Rice soil had significantly higher diversity indices in terms of soil bacterial and targeted N-cycling functional communities than other crop soils; (2) Rice soils with higher SWC had lower amoA abundance and soil PNR, indicating that Rice soils possess lower abundance and activity of ammonia oxidizer; (3) Wheat-Corn soil with higher level of fertilization had higher abundance of nitrification gene amoA and denitrification genes nirK and nirS, compared with Soybean-Corn, Cotton and Fruit soils; (4) SWC, EC, and TN were the most important influencing factors of soil bacterial and targeted N-cycling microbial diversity and structure. Additionally, considering that compound fertilizer 600 kg h −1 year −1 and diammonium hydrogen phosphate 550 kg h −1 year −1 were used in the Wheat-Corn soils, urgent actions are required to improve N fertilizer use efficiency and to reduce N loss, for example, optimizing soil water management and the content of nitrogen fertilizers.