Bacterial Diversity and Potential Functions in Response to Long-Term Nitrogen Fertilizer on the Semiarid Loess Plateau

Bacterial diversity and its functions are essential to soil health. N fertilization changes bacterial communities and interferes with the soil biogeochemical N cycle. In this study, bacterial community and soil physicochemical properties were studied in 2018 after applying N fertilizers (0, 52.5, 105, 157.5, and 210 kg N ha−1) for a long (2003–2018) and a short (2003–2004) duration in a wheat field on the Loess Plateau of China. Soil bacteria were determined using 16S rRNA Illumina-MiSeq®, and the prediction function was analyzed through PICRUSt. The study showed that N fertilizer significantly changed the diversity and abundance of bacterial communities. The phyla Proteobacteria, Actinobacteria, Acidobacteria, and Chloroflexi were most abundant, accounting for 74–80% of the bacterial community abundance. The optimum rates of N fertilizer application (N105) maintain soil health by promoting soil microbial diversity and abundance. The bacterial population abundance was higher after short-term N application than after N application for a long duration and lowest with the high N-fertilizer treatment (N210). High N enrichment led to more heterotrophic N-fixing microorganisms (Alphaproteobacteria), in which metabolism and genetic information processing dominated, while cellular processes, genetic information processing, metabolism, and organismal systems were the main functional categories under low N. The phyla Gemmatimonadetes, Actinobacteria, Bacteroidetes, and Chloroflexi were the key bacteria in the co-occurrence network. The genus Saccharimonadales of the superphylum Patescibacteria has a more significant impact under low N treatment. Long-term N fertilization affected the soil pH, NO3-N, and other physicochemical properties, and soil NO3-N was the highest indicator, contributing 81% of the bacterial community function under different N fertilizer treatments.


Introduction
Microorganisms are the most active part of the soil ecosystem as they transform and recycle soil nutrients [1] by the decomposition of soil organic matter [2]. Microbial community diversity and composition are indicators of soil fertility and health [3]. Microbial diversity is crucial for maintaining soil quality [4]; in particular, high bacterial diversity is essential for sustainable agriculture [5] because of their role in nutrient cycling, disease suppression, and plant growth [6].
The bacterial community is particularly sensitive to changes in the soil environment [7]. Therefore, the diversity and structure of bacterial communities can be used to assess soil quality [8] and changes in soil nutrients [9]. Researchers found that Proteobacteria, Actinobacteria, and Acidobacteria are the main bacterial phyla [10,11], and Agrobacterium, Alcaligenes, and Arthrobacter are the most important bacterial genera in the soil [12].
Extensive agricultural production with high inputs of chemical fertilizers has threatened soil quality [13]. Nitrogen (N) is the main component of nucleic acid and protein and also a key nutrient for all biological communities [14]. Therefore, N is the most extensively used fertilizer for farmland ecosystems; its input is expected to be 112 million tons in 2022 [15,16].
N fertilization can effectively increase soil N content, thereby promoting the growth and reproduction of soil microorganisms [17]. However, long-term N fertilization and high N levels could change microbial community structure and activity, which in turn affect N status in the soil and loss of microbial diversity [18]. Zhou et al. [19] showed that long-term N fertilizer application had turned a healthy bacterial-type soil to a fungal-type soil that was prone to soil-borne diseases. Soil N content also affects nitrogen-oxidizing, nitrogen-fixing, and denitrifying bacteria and other N cycle functional flora, as well as those in carbon and phosphorus cycles [20,21].
Researchers showed that long-term N addition reduced soil pH [22,23]. Soil microorganisms are sensitive to pH changes [24], so N fertilizer may modulate change in microbial activity by affecting soil pH [25]. Long-term N fertilization changes soil physicochemical properties [21] increases residual nitrate in the soil, which can cause N leaching or loss, thereby causing groundwater pollution [26] or increasing greenhouse gas emissions [27]. However, how long-term N fertilization (especially residual N in soil) affects the soil bacterial community diversity and function is poorly understood.
The demand for sustainable land use in agricultural systems raises the need to understand plant-soil-microbe interactions and how to improve soil ecosystems [28]. Long-term fertilization experiments play a pivotal role in providing bases for decision-making related to sustainable agriculture because of its rich, accurate, and reliable data and strong explanatory power [29]. Due to the complexities of soils, the variety of microorganisms, and the limited cultivable microorganisms, long-term localization experiments on microorganisms, especially in crop fields, are still limited.
In this study, changes in bacterial community diversity were examined in wheat fields of the Loess Plateau where different rates of N fertilizers were applied for a continuous 16 years (long-term) or for 2 years (short-term). Bacterial diversity was determined using high-throughput 16S rRNA Illumina-MiSeq ® (San Diego, CA, USA), and the prediction function was analyzed through PICRUSt analysis. The study was carried out based on the hypothesis that N fertilization could influence the soil microbial community structure, diversity, and potential functions by changing the soil physicochemical properties. The specific goals were (1) determine that continuous N fertilization over 10 years may affect the soil bacterial community, (2) characterize the bacterial community and prediction function related to N fertilization, and (3) confirm that N fertilization will change soil microbes by changing the soil physicochemical properties rather than just soil pH.

Site, Experimental Design
A mono-cropping spring wheat experiment field was built in Dingxi, Gansu Province, China (35 • 28 N, 104 • 44 E, elevation 1971 m a.s.l.). The average daily minimum temperature of the experimental field was −22 • C in January, and the maximum temperature was 38 • C in July during 2003 to 2018. The annual average precipitation, evaporation, and radiations were 390.7 mm, 1531 mm, and 5930 MJ m −2 , respectively. The field experimental soil was Calcaric Cambisoll [30], also known as Huangmian [31]. The soil is sandy loam, slightly alkaline (8.33 pH), with total N of 0.77 g kg −1 .
A randomized block design was used to arrange the treatments, and the plot area was 3 m × 10 m, repeated 3 times. Five treatments, i.e., non-N-fertilized control (N0), 52.5 (N52.5), 105 (N105), 157.5 (N157.5), and 210 (N210) kg N ha −1 were applied in the form of urea. In the first main plot, N fertilizer rates were applied from 2003 to 2018. In the second main plot, the N treatments (N52.5c, N105c, N157.5c, and N210c) were applied for two years in 2003 and 2004. Enough K content (221.42 mg kg −1 available K) was present in the soil (60 cm layer) of the experimental field, so no potassium fertilizer was applied. Each treatment applied the same amount of calcium superphosphate (105 kg P 2 O 5 ha −1 ), and all fertilizers were evenly spread on the entire subplot according to the treatment requirements at sowing. In all studied years, spring wheat (Dingxi 38, 187.5 kg ha −1 seeding rate, 20 cm row spacing) was planted in mid-March and harvested around late July of the same year.

Samples, Physicochemical and Bioinformatic Analysis
In 2018, soil samples were collected at the anthesis and grain filling period (14 DAA) of wheat at a depth of 0-20 cm in a zig-zag pattern and sieved (2 mm) after mixing. A part of the mixed soil was air-dried for physicochemical analysis, and the other was shipped using a sealed box with dry ice for molecular analyses.

Data Analysis
SPSS 23.0 software (Chicago, IL, USA) was used to analyze the differences between treatments (p < 0.05). Alpha diversity analysis (Shannon, Chao1, and PD_whole_tree diversity indexes) was used to assess community richness and homogeneity. Beta diversity analysis was used to analyze community compositions evaluated by principal coordinate analysis (PCoA). PICRUSt software was used to predict microbial gene functions [37]. The Kyoto Encyclopedia of Genes and Genomes (KEGG) was used to assign and compare functional counts and annotations [38]. Spearman's correlation and the Kullback-Leibler dissimilarity (KLD) measure was used to construct co-occurrence networks to describe the symbiosis mode in the soil bacterial network [39]. The co-occurrence network was conditioned by the Conet plug-in in Cytoscape software and visualized in Gephi software [40]. Multiple relationships between soil physicochemical properties and distribution of dominant phyla were evaluated by distance-based redundancy analysis (RDA). Indicator analysis was used to calculate the OTU in each sample, and then the indicator value was statistically analyzed (p < 0.05) to reveal the biological species or genera.

Soil Bacterial Alpha Diversity
Illumina MiSeq analysis showed that 1,124,670 filtered sequences (ranging from 33,244 to 49,879 reads per sample) and 901,159 reads were generated and clustered into 4613 OTUs, of which 353 OTUs, 25 genera, and 6 phyla were significantly different (Table S1). The Good's coverage index (>0.967) showed a reasonable amount of sequencing data and sufficient sequencing coverage. Therefore, OTUs were representing the entire microbial community library.
Alpha diversity indices indicated significant differences in bacterial populations among the treatments (Figure 1). The Chao1, PD_whole_tree, and OTU richness indexes were highest for the soil treated with N105, indicating that the N105 treatment resulted in higher bacterial population abundance. The Shannon index was lowest for N210 soil samples, suggesting that the higher N fertilizer lessened the bacterial population abundance. The Chao1 and OTU richness indexes of N52.5c, N105c, N157.5c, and N210c had no significant differences compared with N0, indicating that the bacterial population was not affected by short-term N and not supplying N after a short-term application. It was also observed that bacterial OTUs were most abundant in N105 and least in N210, and the order was N105 > N157.5 > N157.5c > N105c > N52.5 > N52.5c > N210c > N0 > N210 ( Figure S1).

Function Prediction
PICRUSt explored the first (Table 1) and second levels ( Table 2) of metabolic functions of soil bacterial communities. PICRUSt analysis identified seven functions in the first level of Kyoto Encyclopedia of Genes and Genomes (KEGG) ( Table 1). The first level of KEGG predictive function indicated that the metabolism and genetic information processing were the dominant functions, accounting for 42% and 18% of the total abundance of functional genes, respectively. Human disease was maximum under the highest-N treatment (N210). N105 treatment had the highest level of cellular processes, genetic information processing, metabolism, and organismal systems. The heatmap clustering showed the close segregation of N210 with N210c and of N157.5 with N105 ( Figure 3).
Random forest importance of species (variables) points showed that the genera Blastocatella, Gaiella, Chthoniobacter, and Flavisolibacter had the most importance of species (variables) points (>1) across soil samples ( Figure 5). Indicator analysis showed that genus Saccharimonadales of phylum Patescibacteria has a greater impact on the growth environment with N105 treatment ( Figure S4).

Physicochemical Properties and Bacterial Community
Soil indictors (pH, NH 4 -N, NO 3 -N, and AP) differed significantly among treatments ( Table 3). The soil pH decreased (8.09 to 7.81) with different N application rates, and pH at all N-fertilizer treatments was significantly lower than N0. N157.5 and N210 had significantly lower pH than the lower-N treatments (N52.5). Soil NO 3 -N and available phosphorus (AP) increased by 15-67% and 9-19%, respectively, with N treatment. Both the continuous long-term and short-term application of N increased the soil NO 3 -N, but the increase was greater after long-term application. The soil NH 4 -N increased with N application over the control, but significant increases were only observed at N157.5 and N210. Soil available phosphorus (AP) increased with N application over control, and the maximum AP was observed sat N105.
RDA analysis showed that 53.5% of the total variation was explained by axis 1 and 4.3% by axis 2 (Figure 6). The phyla Saccharibacteria, Proteobacteria, Bacteroidetes, Gemmatimonadetes, and Nitrospirae were clustered together and positively influenced by the soil NO 3 -N and NH 4 -N as well as negatively related with soil pH. Among the soil physiochemical properties, NO 3 -N explained 49.8% of the variation and was the most significant predictor with 80.7% contribution (Table S3). The phyla Verrucomicrobia, Acidobacteria and Chlorflexi was negatively related with NO 3 -N and NH 4 -N.   Low soil NO 3 -N and NH 4 -N decreased the phyla Proteobacteria and Bacteroidetes, and increased phyla Acidobacteria, Chloroflexi, Verrucomicrobia, and Planctomycetes (Table 4). Low soil pH decreased phyla Actinobacteria, Acidobacteria, Chloroflexi, and Verrucomicrobia and increased phyla Proteobacteria, Gemmatimonadetes, Bacteroidetes, and Nitrospirae.

Discussion
A soil's bacterial community is a health indicator usually affected by multiple factors such as fertilization, agricultural practices, and environmental conditions [6,10]. Previous studies have shown that short-term application of N fertilization either leaves bacterial diversity unchanged or only shows temporary changes [23,41]. However, long-term N fertilization resulted in significant changes in bacterial community structure and diversity, resulting in a decline in bacterial diversity [42,43]. The present results show that the application of different rates of N fertilizer for 16 years significantly affected the soil physicochemical properties and the composition of the soil microbial community. However, the microbial diversity was lost as a consequence of long-term excessive N fertilizer application [43,44]. We found that the optimum rates of N fertilizer application maintain soil health by promoting soil microbial diversity and abundance, as bacterial population abundance was higher at the optimum N-fertilizer (N105) [45] and lowest at the high Nfertilizer treatment (N210). The bacterial population abundance was higher after short-term N application than after N application for a long duration. After two years of application of different N fertilizer rates and then not applying N fertilizer for several years (14 years) (N52.5c-N210c), the bacterial population abundance tends to be the same as in the non-fertilized control (N0).
In this study, the alpha diversity of soil bacteria was greatest with N105 treatment: N105 treatment resulted in greater bacterial population abundance, while the alpha diversity index at N210 treatment was lower than that of non-N control, indicating that higher N-fertilizer rate led a decrease in bacterial population abundance (Figure 1). The analysis of the taxonomic components of microorganisms showed that the phylum Proteobacteria (24.8%) was the most predominant bacterial, followed by Actinobacteria (20.1%), Acidobacteria (19.6%), and Chloroflexi (13%) (Figure 2, Table S2), and the similar trend was also observed by some previous studies [44,46]. N enrichment in soil affects the composition and function of soil microbial communities [47,48] such as increasing trophic taxa and reducing oligotrophic taxa of soil [49]. High N could even reduce microbial diversity and abundance [50,51]. The phylum Proteobacteria is considered as a copiotrophs [52], and therefore higher N treatment (N210) showed a high relative abundance of Proteobacteria. In contrast, higher N treatment (N210) lowered the relative abundance of Acidobacteria, because Acidobacteria are oligotrophs [53]. The bacteria of class Alphaproteobacteria are heterotrophic nitrogen-fixing microorganisms [54], and their relative abundance was highest with the highest N treatments (N210) due to higher residual soil TN [54]. In addition, there were similar soil physicochemical properties and bacterial community structure under N105 and N210c treatments with a relative higher abundance of phyla Acidobacteria, Planctomycetes, and Nitrospirae, suggesting that N had a significant residual effect on bacterial communities.
The PICRUSt function prediction analysis was used to determine the metabolic and functional capability [55,56]. Metabolism and genetic information processing were the dominant groups of functions, as revealed by PICRUSt analysis, accounting for 42% and 18% of total abundance of functional genes, respectively. The highest human disease function occurred with the highest N treatment (N210), while the most cellular processes, genetic information processing, metabolism, and organismal systems occurred at the medium N rate (N105), reflecting that the optimum-N fertilizer application makes soil healthier than the higher N rate. Functional genes indicate the relative abundance of certain types of microorganisms that are influenced by the biogeochemical cycles of major elements, i.e., C, N, and P [57]. Furthermore, the cellular processes of cell growth and death, environmental information processing of signaling molecules and interaction, metabolism of xenobiotics biodegradation and metabolism, and organismal systems of circulatory system were significantly greater, while glycan biosynthesis and metabolism and environmental adaptation were significantly lower with the highest N treatment (N210) compared with the non-N control.
The interactions among microorganisms such as mutual benefits or competition influence the nutrient cycle and soil health [58]. The symbiosis network between microorganisms reflects the function and stability of a biological community [59,60]. Co-occurrence network analysis analyzes the key interactions between microorganisms and their responses to the environment to explore the basic principles of the microbial communities of different soil environments and identify key groups assumed in the communities [61,62]. Proteobacteria and Actinobacteria are the main representative bacterial communities that have roles in biological carbon and nitrogen cycles [57]. This study observed that Gemmatimonadetes, Actinobacteria, Bacteroidetes, and Chloroflexi are the main components in the co-occurrence network.
The changes in microbial communities are mainly driven by soil characteristics such as soil pH, organic carbon, and N [63,64]. Fertilizer addition changes the microbial community composition and structure by effecting various soil physicochemical properties [47,65]. Soil microorganisms perceive changes in soil properties caused by N fertilizer and respond differently [23,66]. We found that the soil pH, NH 4 -N, NO 3 -N, and AP had significant differences between different N fertilizer treatments. N fertilization resulted in a decrease in soil pH from 8.09 to 7.81. In addition, after applying different rates of N fertilizer for two consecutive years, soil pH was still significantly lower than that of the non-N control after no application of N fertilizer for 14 years. Studies have shown that soil bacteria are most affected by pH [66]. Therefore, soil acidification by N addition is used to explain the observed decline in soil bacterial diversity [46,50,67].
The present study showed that the low soil pH decreased the abundance of phylum Actinobacteria and increased the abundance of Nitrospirae. These results were similar to previous findings indicating that the low soil pH decreased the abundance of Actinobacteria, Proteobacteria, and Bacteroides caused by soil acidification [68] and increased the abundance of Acidobacteria and Nitrospirae [69,70]. However, in contrast to previous results, the low soil pH decreased the abundance of Acidobacteria and increased the abundance of phyla Proteobacteria and Bacteroidetes, which might be due to the weakly alkaline soils in this field (pH ranged from 7.81 to 8.09). Since bacteria can tolerate a narrow pH range [68], they are more sensitive to pH. However, it has not been definitively confirmed whether the soil pH has a direct or an indirect effect on the bacterial community. soil NO 3 -N is one of the highly responsive indicators of the bacterial community function under different N fertilizer treatments (contribution of 80.7%) [23,66]. Bacterial communities are affected by multiple soil physicochemical properties such as NO 3 -N, NH 4 -N, and soil moisture.

Conclusions
N fertilization for 16 years significantly affected the soil NO 3 -N, thus changing the diversity and abundance of bacterial communities. The phyla Proteobacteria, Actinobacteria, Acidobacteria, and Chloroflexi accounted for 74-80.3% of the bacterial community abundance. N fertilization at high rates reduced bacterial community diversity and led to more heterotrophic N-fixing microorganisms (Alphaproteobacteria), in which metabolism and genetic information processing dominated. The relative abundance of Proteobacteria was increased while Acidobacteria was reduced, and the abundance of heterotrophic N-fixing microorganisms increased due to greater residual soil TN. On the other hand, the optimum rate of N fertilizer enhanced the soil health by promoting soil microbial diversity and abundance, and the genus Saccharimonadales of phylum Patescibacteria was the biological genus that had the most impact on its growth environment.
Supplementary Materials: Supplementary data to this article can be downloaded at https://www. mdpi.com/article/10.3390/microorganisms10081579/s1, Table S1: Detailed sequencing depth results of soil samples; Table S2: Relative abundance of soil bacterial taxonomic composition at phyla level for all samples; Table S3: Redundancy analysis results of soil physiochemical properties affecting the distribution of dominant phyla and clustering of soil samples; Figure S1: Operational taxonomic unit Venn analysis of unique and shared bacteria 16S rRNA as affected by N fertilization; Figure