Negative impacts of excessive nitrogen fertilization on the abundance and diversity of diazotrophs in black soil with monocropping maize


 Background: Excessive nitrogen fertilizer input and low nitrogen fertilizer use efficiency in maize in China are serious ecological and economic problems, which might affect the procedures in the nitrogen cycle. To reveal the effects of long-term excessive nitrogen fertilization on diazotrophs in maize rhizosphere and bulk soil, we performed a long-term (five-year) N-input experiment (N rates from 0 to 300 kg N ha -1 ) in black soil maize in northeast China. The effect of N fertilizer application rates on the abundance, structure and compositions of diazotrophic community in both the bulk soil and rhizosphere of maize were investigated by Real-time quantitative PCR and high-throughput sequencing, and a structural equation model was constructed based on this study.Results: 1) Excessive N fertilization significantly reduced the abundance and diversity of diazotrophs. 2) The accumulation of Sphingobium was correlated positively with soil nitrate concentration and soil EC, and negatively with soil pH. The contrast correlation was found in Burkholderia . 3) Diazotrophs were enriched in maize rhizosphere, but the diversity and compositions of diazotrophic community were less affected by maize rhizosphere effect. 4) The enriched Bradyrhizobium and Methylobacterium in maize rhizosphere showed a significant positive correlation with of maize plant biomass. Conclusions: Our results suggest that through affecting soil pH, nitrate and EC values, long-term excessive N input increase Sphingobium accumulation and reduce the abundance of beneficial diazotrophs such as Bradyrhizobium and Burkholderia ，which contribute to the decreased nitrogen use efficiency.

Soil physicochemical properties in the rhizosphere and bulk soils of maize cultured for the long-term different N fertilizer application rates. Soil  The copy numbers of nifH gene and their correlation with soil properties The two-way ANOVA analysis showed that both N fertilization and rhizosphere effect had significant effects on the copy numbers of nifH gene (P < 0.05), but there was no significant interaction (P = 0.952) ( Table S2). The range of nifH gene copy numbers were 4.13-8.91 × 10 5 per g dry soil in rhizosphere soil, and 1.41-6.93 × 10 5 per g dry soil in bulk soil (Fig. 2). It can be seen that the copy numbers of nifH gene in rhizosphere soil was higher than that in bulk soil, and with the increase of N application, the copy numbers of nifH gene gradually decreases, and the excessive N application significantly reduced the nifH gene copy numbers (P < 0.05).
Pearson's correlation coefficients showed that nifH gene copy numbers in rhizosphere and bulk soil were positively correlated with AK and pH (P < 0.05), while negatively correlated with EC values and NO 3 − -N (P < 0.05) ( Table 2). In addition, AP in rhizosphere soil was positively correlated with nifH gene copy numbers (P < 0.05). Table 2 Pearson's correlation coefficients between total nifH gene copy numbers and soil physicochemical properties in the rhizosphere and bulk soils. Diazotrophic α-diversity and its correlation with soil properties After quality filtering and screening of amino acid sequences, a total of 375,772 high-quality nifH sequences were obtained from 32 soil samples. Among them, one sample with too few sequences (bulk soil, N04) was removed, and the remaining sample sequences were between 1882 and 39,170 (Table S3). Therefore, each sample was randomly sampled from 1882 sequences for subsequent analysis. According to 97% similarity, a total of 845 OTUs were kept and each sample contained about 88-240 OTUs with the coverage of 94.3-98.1% (Table S3). The rarefaction curves (Figs. S1a and b) indicated that the obtained sequences could objectively and accurately reflect the abundance and diversity of diazotrophs.
The α-diversity of diazotrophs shown in Fig. 3 demonstrated that all the α-diversity indices were similar between the bulk soils and the rhizosphere soils. While significant decrease of α-diversity was found at excessive N fertilization (N300). The two-way ANOVA analysis (Table S4) further confirmed the results in Fig. 3 that the abundance and diversity of soil diazotrophs were mainly affected by N fertilization rates (P < 0.05), while the rhizosphere had no significant effect (P > 0.05), and there was no significant interaction between the N levels and rhizosphere effect. The one-way ANOVA analysis evidenced no significant difference between the rhizosphere and bulk soil at all the N fertilization rates, and among the N0, N60 and N180 treatments, but the excessive N fertilization (N300) significantly (P < 0.05) decreased the α-diversity of diazotrophs.
Pearson's correlation coefficients showed that OTU number, Chao1 index and Shannon index of diazotrophic in all samples (rhizosphere and bulk soil) were positively correlated with pH, negatively correlated with EC values and NO 3 − -N contence (P < 0.05), while Simpson diversity index was contrary. In addition, chao1 index and shannon index in rhizosphere soil were positively correlated with AP, negatively correlated with OM, and OTU number was positively correlated with AK and negatively correlated with OM (Table S5).

Community Structure Analyses And Relative Abundance Of Diazotrophs
To assess the effects of N fertilization application and rhizosphere effect on the compositions of diazotrophic community, hierarchical cluster analysis (UPGMA), PCoA, NMDS, ANOSIM and PERMANOVA were performed based on OTU classification level. The results of hierarchical cluster analysis (Fig. 4a) showed that all the samples were divided into four clusters according to different N fertilizer application rates. The rhizosphere and bulk soil samples of the same N treatments were clustered together. Similar to the hierarchical cluster analysis, PCoA analysis based on bray-curtis distance also showed that all the samples were divided into four groups corresponding to the N fertilization rates in the first principal axis (42.24%) and the second principal axis (12.85%), and N300 was the treatment distantly separated from the other treatments, while the samples of rhizosphere and bulk soil were significantly separated in the fourth principal axis (6.33%) (Figs. 4b and S2a). The Stress value of NMDS analysis was 0.08, and the grouping situation was consistent with the PCoA result ( Fig. S2b). ANOSIM and PERMANOVA analysis showed the same results (Table S6) and demonstrated significant differences in the compositions of diazotrophic community among different N fertilizer application rates in all samples (P < 0.05), but no significant difference between rhizosphere and bulk soil (P > 0.05). Therefore, the differences of community compositions were mainly affected by N fertilizer application rates, especially by the excessive N fertilization.
The OTUs obtained from each sample were classified into different genera, and the 10 most abundant genera with relative abundance more than 2% were shown in Fig. 5a. Pseudomonas, Burkholderia, Azospirillum and Bradyrhizobium were the four most abundant dominant genera, accounting for 20.75-70.34% of the total nifH gene sequences. The results of heatmap analysis of these dominant genera showed that community compositions were mainly affected by N fertilizer application rates, and the relative abundance of the dominant genera at the N300 was far different from that in the other treatments (Fig. 5b).
The one-way ANOVA (Table S7) showed that excessive N fertilizer application significantly reduced the relative abundance of Burkholderia and Rhodobacter (P < 0.05). The relative abundance of  S4) showed that the compositions of diazotrophic community were mainly mediated by pH, while AP and AK contents also contributed to some extent. Pearson's correlation coefficients were conducted among the 10 dominant diazotrophic genera enriched in rhizosphere soil, soil properties and plant biomass (sum of shoot and root dry weights).
The results (Fig. S5, Table S8) showed that the tested soil variables were closely related to the most Based on the above results and relevant ecological knowledges, a SEM was constructed to further analyze the direct or indirect effects of N application and rhizosphere effect on the soil properties, diazotrophic absolute abundance and α-/β-diversities (Fig. 6). From the results of standardization, N application and the rhizosphere effect can affect the soil properties and the absolute abundance of diazotrophs directly. However, the α-diversity of diazotrophs was affected by soil properties indirectly mediated by different N fertilizer application rates, and the β-diversity of diazotrophs was mainly affected by N fertilizer application rates directly.

Discussion
According to the plant biomass ( Fig. 1) in this study, 180 kg N ha − 1 was confirmed as the reasonable N fertilization for maize grown in the test soil [31], and excessive N fertilizer application (300 kg ha − 1 ) resulted in a significant decrease in plant biomass and in the diazotroph diversity (Fig. 3). This result was consistent with previous studies that plant biomass was not directly proportional to N fertilizer application rates [32,33]. However, the decreased biomass of maize under the excessive N fertilizer application is an interesting phenomenon, which was also reported in other plants [34,35].
According to the Leibig Law of minimum, the plant growth (biomass accumulation) is determined by the nutrient with the lowest concentration, so the addition of excessive N in the present study could not help the plant growth. The negative effects of excessive N on maize growth may be due to that the excessive N fertilizer is toxic for maize that fit the Shelford Law of tolerance. The toxic effect of excessive N fertilizer application might be from the modification of plant metabolism, or from modification of soil environments, such as the degradation of soil properties [34,35] and the alternation of soil microbial community that in turn could decrease the growth of maize. The present study offered evidence of alternation in soil properties and soil microbial communities caused by excessive N application, as well as the correlation among them.
Firstly, the results in the present study clearly evidenced the alternation in soil properties under the excessive N fertilization, which is similar to those in some other studies. Previous studies have shown that excessive N fertilizer input can lead to soil acidification and increase of soluble salt concentration [20]. Soil pH has been considered to be the main factor affecting microbial abundance and diversity [13,[36][37][38][39], and high EC values may increase the incidence of some crop diseases (such as maize  (Table 1) could affect the metabolism and nutrient absorption of maize, which presented optimal growth at pH 5.5-7.0 or pH 4.9-7.3 [43], and then decrease the growth of plant [34,35]. In addition, the acidic condition could enhance the solubility of Fe, Mn, Bo, Cu and Zn and decreased the solubility of Mo, Mg, Ca, K and S, which may cause the toxic effects of the heavy metal and reduce the availability of some mineral nutrients [44].
Secondly, the application of excessive N fertilizer decreased the α-diversity ( The different compositions of diazotrophic community were clearly demonstrated by the clustering analysis, in which the diazotrophic communities were separated according to the N fertilizer application rates; moreover, the communities from excessive N fertilization formed a cluster far distant with those from the other treatments (Figs. 4 and 5). So, the structure of diazotrophic community would be seriously changed by the N fertilizer application dose that exceeded a certain threshold. Among the ten most abundant genera (Fig. 5 Table 1) [45] and is sufficient to inhibit the biological N fixation, e.g. the diazotrophs can not fix N 2 in these soils.
Different from the six genera mentioned above, the relative abundances of Bradyrhizobium, Burkholderia, Rhodobacter and Sphingobium were significantly changed sequentially (Fig. 5, Table   S7): addition of N fertilizer decreased the relative abundance of Burkholderia, so it was most abundant in N0 treatment; while the abundances of Rhodobacter, Bradyrhizobium and Sphingobium were significantly enhanced in the treatments of N60, N180 and N300, respectively (Fig. 5, Table S7).
These changes demonstrated that different levels of N fertilizer application enriched distinct diazotrophic bacteria, which might be related to their adaption to the decreased pH values and increased EC values in soils caused by the fertilization (Table 1).
It is well known that the roots of plant can enrich the some microbes via secretion of secondary metabolites, casts of roots and modification the rhizosphere microenvironments [46][47][48]. Previous studies had found that the diazotrophs in the rhizosphere soil of sorghum, soybean, maize was more abundant than that in the bulk soil [10, 27, 49,50]. Our study also showed that maize rhizosphere had a significant effect on the copy numbers of nifH gene, and the diazotrophs in rhizosphere soil were also more abundant than that that in bulk soil (Fig. 2, Table S2). In addition, NO 3 − -N in rhizosphere soil was lower than that in bulk soil (Table 1), which was due to the selective absorption of soil nutrients by plants [51]. The selectivity may be one of the reasons for the enrichment of diazotrophic in rhizosphere. The proliferation of more diazotrophic in the rhizosphere should help provide more nitrogen nutrients or other beneficial effects to the plant. However, there was no significant difference in the compositions of diazotrophic communities in rhizosphere and bulk soil, indicating that rhizosphere effect had little influence on the diazotrophic community in the tested soils, which was consistent with previous studies [10,17].
Based on our research, through the construction of structural equation model (SEM), it seems that the N fertilizer application rates and rhizosphere effect could directly lead to the variation of soil characteristics and diazotrophic abundance. Then, the variation of α-diversity of diazotrophic mediated by soil characteristics. Finally, α-diversity and N fertilizer application rates mediated the separation of β-diversity of diazotrophs in the tested soils (Fig. 6). The results revealed the influencing factors of diazotrophic community structure variation and the causal relationship among each of the variables, which provided a good idea for the study of subsequent variation mechanisms of microbial community structure.
In our study, excessive N fertilizer input led to a decrease in soil pH and an increase in EC values, which were significantly related to the abundance and diversity of diazotrophs (Tables 2 and S5).
Among them, pH was negatively correlated with Sphingobium, while positively correlated with Burkholderia ( Fig. S5), indicating that the low soil pH under excessive N application led to the accumulation of Sphingobium, while reducing the relative abundance of Burkholderia. Sphingobium were reported can degrade many kinds of polycyclic aromatic hydrocarbons (PAHs) and their derives [52], and Burkholderia was famous for its bioremediation and antifungal properties [53], suggested that excess N fertilizer despite increased the degradation ability of soil pollutants, but its resistance to pathogen was reduced, thus increasing the probability of occurrence of soil spread diseases.
Long-term excessive N fertilizer input also had a profound impact on soil nutrient content ( Table 1).
The variation of soil properties also had a significant impact on the compositions of diazotrophic community (Figs. S3 and S4, Table 3). In particular, the NO 3 − -N and EC values in soil were positively correlated with Sphingobium and negatively correlated with Burkholderia, which were agreed with the relationship between soil pH and these two diazotrophic bacteria, indicating the synergistic effect on the structure of diazotrophic community of high nitrate content, EC values and low pH caused by excessive N fertilizer input. On the contrary, the soil AP and AK negative correlation with Sphingobium and positive correlation with Burkholderia imply that the proportion of nitrogen, phosphorus and potassium in soil was also crucial for maintaining a normal diazotrophic community, which proved the importance of balanced application of NPK from the perspective of soil microbes [54].
The variation of soil microbial community structure has a feedback regulation effect on plant [1].
However, little is known about the feedback of plant from the change of diazotrophic community.
Based on our study, it was found that plant biomass was positively correlated with the relative abundances of Bradyrhizobium and Methylobacterium (P < 0.05, Fig. S5). Bradyrhizobium, which is widely distributed in soil, has important biological functions, including photosynthesis, symbiotic and free-living nitrogen fixation, denitrification and aromatic compound degradation, and plays an important role in the global nitrogen cycle. The multiple functions make it important for agricultural production [55][56][57]. In addition, Methylobacterium is also widely existed in the environment. It is a kind of diazotrophic bacteria that can produce plant growth hormone and can be classified as plant growth-promoting bacteria [59,60]. This indicated that the increase of biomass under normal N fertilization (N180) was more likely related to the enrichment of these two diazotroph. Based on the above analysis, Bradyrhizobium and Methylobacterium may be utilized to develop efficient microbial fertilizers.

Conclusions
This study argues that excess N fertilizer input led to soil acidification, increased soluble salt concentration, and excess nitrate nitrogen, which directly led to a significant decrease in abundance The raw nifH gene sequencing reads were demultiplexed, quality-filtered by Trimmomatic and merged by FLASH [62], in which forward and reverse reads had the overlapping base length at least 10 bp, and the maximum mismatch ratio of overlap region was 0.2. Then the sequencing reads were analyzed using QIIME platform [63] to remove the low-quality sequences, including the quality score < 20, ambiguous nucleotides, or mismatched primer and barcode. After sorting the sequences of samples according to barcodes, barcode and primer sequences were also removed. The remaining sequences were translated into amino acid using the FunGene Pipeline [64]. The chimeras and translated protein sequences that did not match the nifH protein were then discarded [65], while the remaining high-quality sequences were clustered into operational taxonomic units (OTUs) at 97% similarity cutoff by UPARSE [66]. Representative sequences from each OTU were taxonomically classified by the BLAST algorithm-based search against the NCBI GenBank database (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

Data analysis
The software mothur [67] was used to count α-diversity of diazotrophs (OTU number, Chao1 A Mantel test was carried out to determine the correlation between the community composition of rhizosphere and bulk soil and soil properties. Only variables that had significant effects (p < 0.05) by the Mantel test were further used to perform a CCA analysis using Cannoco for Windows [68]. In addition, the mvpart package [69] of the R was used to construct a multivariate regression tree (MRT) to comprehensively explore main soil variables that best shaped the diazotrophic community composition.

Structural Equation Models
In order to quantify the importance of N fertilizcation rates and rhizosphere effect on soil properties and diazotrophic community, we constructed a structural equation model based on the current ecological knowledges. And theoretical model hypothesised that: 1) N fertilization rates and rhizosphere effect directly affect diazotrophs, respectively; 2) N fertilization rates and rhizosphere effect indirectly affect diazotrophs by changing soil properties; 3) β-diversity of diazotrophs was affected by the change of α-diversity. The variable of rhizosphere effect was created by assigning the value 1 to the rhizosphere soil and 0 to bulk soil. All the measured soil properties and β-diversity of the diazotrophs (based on bray-curtis distance matrix) were dimensionally reduced through NMDS, and the first axis of NMDS was used to represent the variance matrix of soil characteristics and βdiversity of the diazotrophs. Shannon index was used to represent α-diversity. The abundance of diazotrophs was expressed by nifH gene copy numbers. All variables were normalized by Z transformation (mean = 0, standard deviation = 1), and a covariance matrix of these variables was inserted into AMOS 24.0 (SPSS, Chicago, IL, USA) for SEM construction and analysis. Maximum likelihood estimation was used to fit the covariance matrix into the model [70]. Chi-square (P > 0.05), goodness of fit index (GFI > 0.90), and root mean square error of approximation (RMSEA < 0.05) were used to ensure the model adequately fit [71].

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and materials
The obtained sequences were submitted to the NCBI Sequence Read Archive (SRA) with accession number SRP253214.

Competing interests
The authors declare that they have no competing interests.

Funding
This work was financially supported by the National Nature    Abundance of the nifH gene in maize rhizosphere and bulk soils under long-term different N fertilizer application rates. Data are means ± standard error (n=3). All of the data were analyzed using a two-way analysis of variance, N: N fertilizer application rates; R: rhizosphere effect (bulk and rhizosphere). Different letters on the bars represent significant differences between all of the data by one-way anova (Duncan multiple-range test, P<0.05).   Abundance of the nifH gene in maize rhizosphere and bulk soils under long-term different N fertilizer application rates. Data are means ± standard error (n=3). All of the data were analyzed using a two-way analysis of variance, N: N fertilizer application rates; R: rhizosphere effect (bulk and rhizosphere). Different letters on the bars represent significant differences between all of the data by one-way anova (Duncan multiple-range test, P<0.05).  Abundance of the nifH gene in maize rhizosphere and bulk soils under long-term different N fertilizer application rates. Data are means ± standard error (n=3). All of the data were analyzed using a two-way analysis of variance, N: N fertilizer application rates; R: rhizosphere effect (bulk and rhizosphere). Different letters on the bars represent significant differences between all of the data by one-way anova (Duncan multiple-range test, P<0.05).        Significance levels are denoted with *P < 0.05, **P < 0.01, ***P < 0.001. Standardized total effects calculated by the SEM were displayed below the SEM. The low chi-square (χ2), nonsignificant probability level (P > 0.05), high goodness-of-fit index (GFI > 0.90), low Akaike information criteria (AIC), and low root-mean-square errors of approximation (RMSEA < 0.05) listed below the SEMs indicate that our data matches the hypothetical model.

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.