Application of mineral phosphorus fertilizer influences rhizosphere chemical and biological characteristics

ABSTRACT Soil microorganisms directly participate in phosphorus (P) cycle, which in turn affects crop yield and quality. However, how biological and chemical parameters respond to mineral P fertilizer application rates in rhizosphere soil remains unclear. Here, a pot experiment was conducted to investigate the effects of different mineral P fertilizer application rates (0, 60, 120, and 300 kg P2O5 ha−1) on soil chemical properties and bacterial community in maize rhizosphere in a Mollisol. With increasing P fertilizer rates, soil organic carbon concentration decreased, while soil Olsen P and inorganic P and organic P (Po) fractions concentrations increased. High-throughput 16S rRNA gene sequencing showed that the relative abundances of some beneficial bacteria (Anaerolinea, Anaeromyxobacter, Kaistia and Rhodobacter) significantly increased with reduced P fertilizer rates, and meanwhile those of other beneficial bacteria (Mesorhizobium, Microvirga, Rhizobacter and Skermanella) significantly decreased under high P fertilizer rates. Redundancy analysis (RDA) showed that labile Po was the main driver of the differences in bacterial community structure. Bacterial community diversity and composition significantly and indirectly affected maize yield via octacalcium-bound P, occluded Fe/Al-bound P, labile Po, and moderately labile Po. The results suggested that reducing P fertilizer rates can promote beneficial bacterial growth and increase bacterial community diversity.


Introduction
Phosphorus (P) is one of the essential and frequently limiting nutrients in aquatic and terrestrial ecosystems. The application of P fertilizers is a common practice to maintain and increase crop yields in agricultural ecosystems. Most P fertilizers are predominantly derived from the mining of P rock deposits (Powers et al. 2019). However, as a nonsubstitutable and nonrenewable resource, reserves of P rock deposits (especially high-grade P rocks) are finite all around the world and are mainly located in countries such as Morocco, China, and Algeria (Blackwell et al. 2019). It is forecasted that global P rock reserves will be exhausted within 100 years (Cooper et al. 2011). On the other hand, the low P fertilizer use efficiency (approximately 10%-15%) of most crops due to the high soil P fixing capacity (Zhu et al. 2018) leads to large amounts of legacy P that accumulate in cropland soil, which could increase the environmental risk of eutrophication of nearby aquatic ecosystems. Maize (Zea mays) is the third most important cereal crop after rice (Orzya sativa L.) and wheat (Triticum aestivum L.) worldwide (Thirunavukkarasu et al. 2017). However, the excess application of P fertilizers has been common to achieve high maize yields (Xin et al. 2017). It has been reported that P fertilizer use efficiency is lower in maize than in rice and wheat (Yu et al. 2021). Therefore, it is essential to reduce P fertilizer application rates and improve P fertilizer use efficiency in maize to conserve P rock resources, ensure environmental safety, and sustain agricultural productivity.
Microorganisms (e.g. bacteria) that can be mobilized to fix soil inorganic P into bioavailable forms through the release of organic acid anions and solubilization of precipitated P-metal complexes are key drivers of P cycling in ecosystems (Randall et al. 2019). In agroecosystems, fertilization regimes have a profound impact on soil microbial community structure, which could in turn provide feedback on the availability of soil P (Zhou et al. 2019). Previous studies have mainly focused on the effects of the different application rates of mineral P fertilizers on bacterial community structures in bulk soil (Lang et al. 2018;Chen et al. 2019). The rhizosphere, defined as the region of soil affected by living plant roots, is characterized by intense chemical and biological activities due to the existence of root exudates, which makes it a crucial zone for nutrient cycling and microbial and plant growth (Mahmud et al. 2021;Omotayo and Babalola 2021). Microbes associated with the rhizosphere are generally considered to play a remarkable role in controlling soil P transformation and dynamics (Zhou et al. 2019). On the other hand, the responses of soil microbial communities to P fertilizer inputs are highly dependent on the study regions and ecosystem types (Deng et al. 2021). Northeast China is one of the main maize cropping regions in the world, and maize production in the area accounts for 7.4% of the global total (Jiang et al. 2021). However, how microbial community structure in rhizosphere soil responds to different mineral P fertilizer rates under continuous maize cropping in Northeast China is still unclear, and the elucidation of this phenomenon would provide valuable information for the development of optimum fertilization patterns.
In the present study, we aimed to evaluate the impacts of different mineral P fertilizer application rates on bacterial community composition and diversity in the rhizosphere of a maize cultivar grown in a Mollisol in Northeast China in a pot experiment using Illumina MiSeq high-throughput sequencing technology. Moreover, the factors that determine bacterial community composition and diversity were also explored by examining the relationships between bacterial community structure and selected soil properties (e.g. pH, total P, Olsen P, inorganic P fractions, and organic P fractions). We hypothesized that soil bacterial community structure would be distinct under different P fertilizer rates and that an appropriate P fertilizer rate could promote beneficial bacterial growth and improve bacterial community diversity.

Experimental design and sampling
The pot experiment was started on 20 May 2018 in a glass greenhouse located at the experimental station of Jilin Agricultural University, Jilin Province, Northeast China (43°14ʹN, 125°15ʹE). The area has a temperate continental monsoon climate, with a mean air temperature of 4.9°C and average precipitation of 593.8 mm during the experiment. Composite soil samples (0-20 cm depth) were collected from maize fields near the glass greenhouse and sieved to 3 mm after removing stones and coarse plant debris. The experimental soil was classified as black soil (Mollisol in USDA Soil Taxonomy), with a pH of 6.28 and containing 19.4 g kg -1 organic carbon, 0.48 g kg -1 total P, and 13.8 mg kg -1 Olsen P. The maize cultivar used was LiangYu-99, with a growth period (from seedling emergence to maturity) of 129 d, which has been widely cultivated in this region.
The experiment comprised four P fertilizer rates, i.e. 0, 0.4, 0.8, and 2.0 g P 2 O 5 pot −1 , which were equivalent to 0, 60, 120, and 300 kg P 2 O 5 ha −1 (hereafter referred to as P0, P60, P120, and P300, respectively). The P fertilizer rate of 120 kg P 2 O 5 ha −1 corresponded to local farming practice, and the P fertilizer rate of 300 kg P 2 O 5 ha −1 was used to simulate the condition of soil P during yearly application at a rate of 120 kg P 2 O 5 ha −1 for approximately 2 or 3 years. In each treatment, 1.4 g N pot −1 and 0.6 g K 2 O pot −1 (equivalent to 210 kg N ha −1 and 90 kg K 2 O ha −1 , respectively) were also applied. The N, P, and K fertilizers used were urea, diammonium phosphate, and potassium chloride, respectively. Each treatment was arranged in a completely randomized design with three replications, and the total pot number was 12. The air-dried soil samples (15 kg) were thoroughly mixed with ground mineral N, P, and K fertilizers and then transferred to bottom-sealed cylindrical plastic pots (height: 39 cm, inner diameter: 29 cm, weight: 0.7 kg). All diammonium phosphate and potassium chloride and 30% of urea were applied as base fertilizers, and the remaining urea was applied at the jointing (30%) and heading (40%) stages of maize. Three maize seeds per pot were sown at a depth of 3 cm and thinned into one plant at the second leaf stage. Throughout the growth stage, soil moisture in each pot was monitored by microtensiometers (Nanjing Canglang Technology, Nanjing, China), and watering was periodically conducted to maintain soil moisture at 60%-70% of field water holding capacity. Pests and diseases were controlled by spraying insecticides and fungicides, and weeds were regularly removed by hand.
Maize grain was harvested at the physiological maturity stage on 8 October 2018. The grain yield was expressed as g pot −1 at a 14% moisture content. At the same time, the ear length, bald tip length, ear row number, kernel number per row, and hundred-kernel weight of all individuals were also recorded. The soil samples from the maize rhizosphere (0-5 mm away from the root surface) in each pot were separated by a shaking method (Zhang and George 2002), mixed together thoroughly, and then sieved through a 2 mm mesh that is appropriate to analyze soil chemical and biological parameters. The collected soil samples were divided into two equal parts: one part was stored at −80°C for DNA extraction, and the other part was air dried for the analysis of soil properties.

Determination of soil chemical properties
Soil pH was tested by a PHS-3C pH meter (INESA Scientific Instrument, Shanghai, China) using a 1:2.5 (w/v) suspension in distilled water. SOC was determined via dichromate oxidation method (Lu 2000).
Soil total P and Olsen P were extracted by using concentrated H 2 SO 4 -HClO 4 digestion and 0.5 mol L −1 NaHCO 3 (pH 8.5), respectively (Olsen et al. 1954(Olsen et al. , 1982. Soil inorganic P (Pi) was sequentially fractionated using 0.25 mol L −1 NaHCO 3 (pH 7.5) for dicalcium-bound P (Ca 2 -P), 0.5 mol L −1 NH 4 OAc (pH 4.2) for Ca 8 -P, 0.5 mol L −1 NH 4 F (pH 8.2) for aluminum-bound P (Al-P), 0.1 mol L −1 NaOH-Na 2 CO 3 for iron-bound P (Fe-P), 0.3 mol L −1 Na 3 C 6 H 5 O 7 -1 g Na 2 S 2 O 4 -0.5 mol L −1 NaOH extraction followed by concentrated H 2 SO 4 -HClO 4 -HNO 3 digestion for occluded iron/aluminum-bound P (O-P), and the addition of 0.5 mol L −1 H 2 SO 4 for phosphorite bound P (Ca 10 -P) (Jiang and Gu 1989). Soil organic P (Po) was sequentially fractionated using 0.5 mol L −1 NaHCO 3 (pH 8.5) for LPo, 1.0 mol L −1 H 2 SO 4 followed by 0.5 mol L −1 NaOH extraction for MLPo, and concentrated HCl was added to acidify the 0.5 mol L −1 NaOH extract from the MLPo step for moderately resistant Po (MRPo); highly resistant Po (HRPo) was calculated by subtracting MRPo from the Po concentration in the 0.5 mol L −1 NaOH extract (Bowman and Cole 1978). All P concentrations in the above extractions were colorimetrically tested by ammonium molybdate and ascorbic acid (Murphy and Riley 1962). P efficiency indices, including partial factor productivity (PFP), P agronomic efficiency (PAE), P apparent utilization efficiency (PAUE), and P uptake efficiency (PUPE), were calculated according to the following formulas (Zhu et al. 2012;Ao et al. 2014;Roberts and Johnston 2015;Sandaña 2016): where Y1 is grain yield under the P treatment (kg), Y2 is grain yield under the no-P treatment (kg), G is the amount of applied P (kg), P1 is the amount of absorbed P in aboveground parts under the P treatment (kg), and P2 is the amount of absorbed P in aboveground parts under the no-P treatment (kg).

Soil DNA extraction
Soil total DNA was extracted from fresh soil samples (0.5 g) from each of three replicates using an E.Z.N.A™ soil DNA kit (Omega Biotek, Norcross, GA, United States) according to the manufacturer's protocols. The concentration of extracted DNA was determined with a NanoDrop 2000 UV-Vis spectrophotometer (Thermo Scientific, Wilmington, DE, United States), and the quality of DNA was evaluated by 1% agarose gel electrophoresis.

PCR amplification and high-throughput sequencing of bacterial 16S rRNA genes
The V3-V4 region of 16S rRNA was amplified by PCR, and the specific primers F338 (ACTCCTACGGGAGGCAGCAG) and R806 (GGACTACHVGGGTWTCTAAT) (Xu et al. 2016) were used. The PCR program were as follows: predenaturation for 5 min (95°C), 28 cycles of denaturation for 30s (95°C), annealing for 30s (55°C), extension for 60s (72°C), and a final extension for 7 min (72°C). The PCR system was 25 μL and contained the following components: forward primer (5 μm), reverse primer (5 μm), BSA (2 ng μL −1 ), 2x Taq plus Master Mix, and 1.0, 1.0, 3.0, 12.5, or 7.5 μL double distilled H 2 O. The PCR products from each sample were mixed, subjected to 2% agarose gel electrophoresis and purified with an AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA). According to preliminary quantitative results of electrophoresis, the Quantifluor™-ST blue fluorescence quantitative system (Promega, Madison, WI, USA) was used for detection and quantification of the PCR products, and then the corresponding proportion of each sample was mixed according to the sequencing requirements. The sequencing platform (Illumina, San Diego, CA, USA) at Beijing Allwegene Technology Co. Ltd. (Beijing, China) was used for sequencing. The raw sequences were submitted to the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) with accession number SRP277121.

Sequence data processing
The raw 16S rRNA gene sequencing reads were processed and analyzed using QIIME (http://qiime. org/tutorials/tutorial.html). Low-quality sequences with an average quality score of < 20 and shorter than 200 bp in length were removed. After quality filtering, the final sequence was subjected to OUT analysis, and OTU clustering (excluding a single sequence) was then carried out according to 97% sequence similarity by using UPARSE (http://drive5.com/uparse/). A representative sequence from each cluster was aligned using PyNAST. To obtain the taxonomy of each 16S rRNA gene sequence, the Ribosomal Database Project (RDP) classifier algorithm was used with a confidence threshold of 0.8 (http://rdp.cme.msu.edu/classifier/classifier.jsp). Based on the cluster file, the Shannon index and Chao index were calculated in MOTHUR (Schloss et al. 2009).

Statistical analysis
Significant differences in maize yield, phosphorus use efficiency, soil properties, Shannon index, and Chao index among treatments were compared using one-way analysis of variance (ANOVA) followed by Fisher's least significant difference (LSD) test at P < 0.05. Student's t test was used to evaluate the significance of differences in the relative abundances of soil bacterial community at the phylum level between P0 and P60, P120, and P300 at P < 0.05, P < 0.01, and P < 0.001. SPSS 20.0 software was used for statistical analysis.
To identify the genera showing significant differences between P0 and P60, P120, and P300, a t test was performed in Statistical Analysis of Metagenomic Profiles (STAMP) (version 2.1.3) (Parks et al. 2014). Partial least squares discriminative analysis (PLS-DA) was completed in the muma package of R software (version 3.5.0). The indicator species analysis was completed in the indicspecies package of R software (version 3.5.0), and the indicator value (indval) was greater than 0.5 (Dufrêne and Legendre 1997). As the response data were compositional and had a gradient length of 1.0 SD units, RDA based on Bray-Curtis distances at the OTU level was performed in Canoco (version 5.0) to determine the relationships between soil bacterial community and dependent variables (P fertilizer rates, SOC, soil pH, total P, Olsen P, Pi fractions, Po fractions). Soil variable effects were evaluated by variation portioning based on the simple effect and conditional effect for each soil variable, and ranking of the importance of soil variables was based on the significance of the displacement test (P < 0.05). Structural equation modeling (SEM) was performed to test the direct or indirect effect of soil bacterial community diversity and composition on maize yield in AMOS software (version 24.0). The nonsignificant paths were removed stepwise to acquire the most parsimonious model. The fitness of the final model was as follows: chi-square test (χ 2 ) = 12.9, degrees of freedom (df) = 8, P = 0.12, comparative fit index (CFI) = 0.95, Akaike information criteria (AIC) = 84.9, and root square mean error of approximation (RMSEA) = 0.16.

Soil properties, phosphorus use efficiency and maize yield
The ANOVA results showed that the concentrations of Olsen P and Pi and Po fractions in the rhizosphere soil all tended to increase with increasing P fertilizer rates. There were significant differences between P300 and other treatments for Olsen P, Ca 2 -P, Al-P, O-P, and Ca 10 -P (Table S1). In addition, we observed that the application of P fertilizer had no significant effect on soil pH value, while SOC concentration decreased significantly by 6.47%−7.96% under P120 and P300 compared with that under P0.
The ANOVA results showed that P efficiency indices (i.e. PFP, PAE, PAUE, and PUPE) all decreased in the order of P60 > P120 > P300 (Table S2). The maize grain yield was 9.10%−15.0% (P < 0.05) higher than that in the treatment without P fertilizer, while there was no significant difference among the treatments with P fertilizer at different application rates. The ear length of maize plants was 5.44%−8.97% (P < 0.05) greater under P60 and P300 than under P0 and P120, the ear row number was 11.9%−14.0% (P < 0.05) higher under P0 and P60 than P300, and the kernel number per row was 13.1%−13.9% (P < 0.05) higher under P300 than under P0 and P120.

Soil bacterial community composition and diversity
After removing the barcode, primer, and low-quality sequences, a total of 963,816 high-quality sequences were obtained from all 12 soil samples; the number of sequences varied from 37,227 to 137,087 per sample; and the length of sequences ranged from 239 bp to 450 bp, with the concentration between 400 bp and 440 bp (Table S3). The sequences were clustered into OTUs at the ≥ 97% similarity level, and 5,519 OTUs were obtained from all of the soil samples, ranging from 2817 to 3370 OTUs per sample. All OTUs could be classified into 41 phyla, 109 classes, 153 orders, 284 families, and 458 genera.
An indicator species is defined as a species that has an occurrence pattern related to the species richness of a larger biota. Indicator species can be used as ecological indicators for community types, environmental changes, and habitat conditions, and indicators can be selected according to their niche preferences to maintain or restore the ecological integrity of the ecosystem. For the different treatments, bacterial indicator species with the highest IndVal were OTU3501, OTU2173, OTU2267, and OTU1888 under P0, P60, P120, and P300, respectively. These OTUs were most closely related to Brachybacterium, uncultured Blastopirellula sp., uncultured Croceibacter sp., and Paenibacillus sp. (Table S5). The Shannon index was 2.31%−3.05% (P < 0.05) higher under the low (i.e. P0 and P60) than under the high (i.e. P120 and P300) P fertilizer application rates (Figure 2). However, the Chao index was not significantly different among the four treatments.
PLS-DA was performed to compare the similarities and distinctions of bacterial community structure among different mineral P fertilizer application rates. Comp1 and Comp2 explained 12.1% and 10.4% of the variation in bacterial community structure, respectively (Figure 3(a)). The bacterial community structure was similar between P120 and P300 on Comp1 but was different from that of P0 and P60. Moreover, bacterial community structure was also different between P0 and P60 on Comp2.

Relationships among bacterial community, soil properties, and crop yield
The Pearson correlation between bacterial community and soil properties showed that Shannon index was positively correlated with pH but negatively correlated with Olsen P, Ca 8 -P, Al-P, and LPo (Table S6). Chao index was negatively correlated with LPo. Comp1 was negatively correlated with Olsen P, Ca 8 -P, Al-P, O-P, and LPo. Comp2 was positively correlated with Olsen P, Al-P, O-P, LPo, MLPo, MRPo but negatively correlated with pH.
The RDA results showed that the examined soil properties explained 65.7% of the variation in bacterial community structure (Figure 3(b)). RDA1 and RDA2 explained 49.2% and 16.5% of the bacterial community structure, respectively. The low P (i.e. P0 and P60) and high P (i.e. P120 and P300) treatments produced differentiation along RDA1. LPo was a significant factor in shaping bacterial community structure and explained 24.6% of the total observed variance in bacterial community structure (Table S7).

Responses of soil properties, phosphorus use efficiency, and maize yield to mineral P fertilizer application rates
Our results implied that the application of excessive P fertilizer was not favorable to SOC accumulation (Table S1), which could be ascribed to the positive priming effect of P addition on SOC mineralization by stimulating microorganisms to decompose SOC for N acquisition and/or enhancing the activity of SOC viah degrading microorganisms following the application of P fertilizer (Mehnaz et al. 2019).
In addition, our results suggested that reducing P fertilizer application had no effect on maize grain yield but was beneficial for improving P use efficiency (Table S2) and preventing soil legacy P accumulation (Table S1). It is generally reported that crop yield does not respond to P application in soil with Olsen P concentrations that exceeded the critical Olsen P value for the optimal crop yield (Bai et al. 2013;Thuy et al. 2020). Thus, the nonsignificant response of maize grain yield to P fertilizer rates could be due to the soil Olsen P levels (31.9-76.7 mg kg -1 ) which were higher than the critical Olsen P concentrations for the optimal maize yield (18.0 mg kg -1 ) in the present study (Bai et al. 2013). The increase in P use efficiency with decreased P fertilizer application rates supported the notion that the efficiency of plant uptake of fertilizer P increases withincreasing soil P availability (Yu et al. 2021).

Responses of bacterial community composition to mineral P fertilizer application rates
In our study, we observed that the dominant bacterial phyla were Proteobacteria, Actinobacteria, Acidobacteria, and Chloroflexi (Figure 1(a)). A similar result was obtained by Silva et al. (2017), who indicated that Proteobacteria, Actinobacteria, Acidobacteria, and Chloroflexi were abundant bacterial phyla in triple-superphosphate-fertilized soil in the maize rhizosphere. P fertilizer application caused changes in the bacterial community composition at the phylum level. Compared with P0, P fertilizer significantly enhanced the relative abundance of Gemmatimonadetes (Table S4), which may be important in the decomposition of recalcitrant C compounds (Lipson and Schmidt 2004). Bacteroidetes abundance was significantly lower under P120 and P300 than under P0 (Table S4), and Bacteroidetes is commonly considered a eutrophic group with a high growth rate in a carbonrich environment (Eilers et al. 2010). It was suggested that appropriate P fertilizer application could contribute to the accumulation of soil nutrients and thus promote the growth of eutrophic bacteria.
At the genus level, Sphingomonas was the dominant genus under P0 and P60 in the maize rhizosphere soil (Figure 1(b)). Sphingomonaspituitosa is a potential phosphate-solubilizing bacterium that is beneficial to rice growth (Panhwar et al. 2014). Compared with P0, the relative abundances of Anaerolinea, Anaeromyxobacter, Rhodobacter, and Kaistia increased significantly under P60 ( Figure  S2). Anaerolinea was the main component of bacterial community involved in phosphorus removal (Kragelund et al. 2007). Su et al. (2015) also showed that the relative abundances of phosphatesolubilizing bacteria, such as Oxalobacteraceae, Klebsiella, Burkholderia, and Bacillus, under P fertilizer treatment were higher than those without P treatment. A large number of studies have proven that phosphate-solubilizing bacteria can not only increase the uptake of crop P and improve crop yield but also greatly improve the utilization rate of P fertilizer (Mander et al. 2012;Zhang et al. 2014). A large amount of insoluble P in soil may promote the reproduction of phosphate-solubilizing bacteria, which can improve the availability of P by secreting organic and inorganic acid substances, phosphatase, and hormones. The appropriate application of P fertilizer can promote the reproduction of phosphate-solubilizing bacteria in soil and improve the absorption of P by crops. Anaeromyxobacter can fix nitrogen in the paddy soil environment and fix and assimilate N 2 gas with nitrogenase (Masuda et al. 2020). The photosynthetic α-proteobacterium Rhodobacter capsulatus can reduce and fix atmospheric dinitrogen with molybdenum nitrogenase and iron-only nitrogenase (Demtröder et al. 2020). Kaistia shows cellulolytic activity . Thus, the relative abundance of potentially beneficial bacteria appeared to exhibit an increasing trend under an appropriate P fertilizer application rate.
Compared with P0, the relative abundances of Geobacter, Roseiflexus, Mesorhizobium, and Microvirga were significantly reduced under P120 and P300 ( Figure S2). Geobacter sulfurreducens in Deltaproteobacteria shows nitrogen fixation ability and promots bacterial Fe(III) reduction in paddy slurries upon flooding (Yi et al. 2013). Roseiflexus castenholzii was discovered to be a thermophilic filamentous anoxygenic phototroph and is able to develop photoheterotrophically in anaerobic conditions under light or in the presence of oxygen in the dark (Niedzwiedzki et al. 2010). Mesorhizobium loti fixs nitrogen for its host Lotus japonicus and prompts the host to grow without any other source of nitrogen (Quides et al. 2017). Microvirgavignae is a novel symbiotic nitrogenfixing Alphaproteobacterium obtained from a cowpea root nodule (Zilli et al. 2015). In addition, Skermanella, Rhizobacter, and Methylobacterium abundances decreased markedly under P120 and P300, respectively. Skermanella sp. acts as a biocontrol agent that plays a dynamic role in controlling rice leaf folder (Cnaphalocrocis medinalis Guenee) and pink stem borer (Sesamiainferens Walker) (Panneerselvam et al. 2018). Methylobacterium was shown to be a nitrogen-fixing microbial community component on a Leymus chinensis steppe (Zou et al. 2011). It may be suggested that excessive P fertilizer application leads to a decrease in the relative abundance of potential functional bacteria.

Responses of bacterial community composition and diversity to mineral P fertilizer application rates
As expected, we demonstrated that the bacterial diversity was higher under the low P fertilizer application rates (P0 and P60) than under the high P fertilizer application rates (P120 and P300) ( Figure 2). However, Tan et al. (2013) found that as the amount of P fertilizer increased, soil bacterial diversity also increased under different P application rates. The increase in soil nutrients, especially the content of organic matter, is the main factor in the increase in bacterial diversity after P fertilizer application (Zhong and Cai 2007). This might be mainly due to the influence of the P source on bacterial diversity, which may vary. Soil microorganisms show differences in ecological resistance, adaptability and redundancy, and their responses to the external environment are diverse.
This study emphasized that there is a significant difference in bacterial community structure between the low P fertilizer application rates and the high P fertilizer application rates (Figure 3(a)). This result agreed with the finding of Chen et al. (2014), who confirmed that microbial community structure changed with different P fertilizer application rates. However, Shi et al. (2012) stated that after the long-term application of different P fertilizers, soil bacterial community structure did not change significantly and showed that although the content of soil Olsen P was low, the contents of soil C and N were still the main limiting factors for soil microorganisms. This discrepancy may be mainly due to the different methods applied, as molecular methods are sensitive and limited. The work of Shi et al. (2012) was based on phospholipid fatty acid profiles (PLFA), while this study was based on high-throughput sequencing.

Controls of soil properties on bacterial community structure under different mineral P fertilizer application rates
In this study, we found that LPo was negatively correlated with Shannon and Chao indices (Table S6), and LPo was also a significant factor regulating bacterial community structure (Figure 3(b), Table S7). The significant influence of LPo on soil microbial community composition has also been observed in a previous study (Deforest and Scott 2010). However, many studies have found that soil pH is the main factor driving microbial community composition (Fierer and Jackson 2006;Ragot et al. 2016), and P fertilizer affects soil microbes in various ways such as increasing nutrient availability and changing soil pH and osmotic potential (Huang et al. 2016). This might be due to the pH, which was observed to be stable and showed no statistical difference among mineral P fertilizer application rates in this study, and the effect on the bacterial community structure was not obvious. On the other hand, we found that bacterial community had a significant direct influence on P forms ( Figure S3). Crop root exudates can form a concentration gradient in the rhizosphere soil that promots the enrichment or reduction of functional microorganisms related to P transformation within the spatial gradient. Microorganisms directly participated in the mutual transformation of different forms of P by releasing organic and inorganic acids into the soil (Lopez et al. 2011).
In this study, RDA showed that bacterial community structure under different mineral P fertilizer application rates was influenced by distinct mineral P components (Figure 3(b)). Similarly, Wang et al. (2018) considered the fertilizer regime to be the main factor that affected soil bacterial community, followed by soil properties, and Olsen P. Zhao et al. (2014) showed that soil bacterial community structure was more responsive to the fertilizer regime than soil properties (soil organic matter, total P, total K, and available K) and the sampling time in a rice-wheat cropping system. Grover (2004) also found that P fertilizer increased the content of soil Olsen P, thereby changing soil microbial community structure and promoting the growth of specific microorganisms. These results suggested that P input is a determining factor in the development of soil microbial community structure in agricultural systems.

Conclusions
This study revealed that reducing P fertilizer rates could reduce soil P accumulation and SOC concentration while increase P use efficiency, soil Olsen P, and Pi and Po fractions concentrations. P fertilizer application significantly affected soil microbial community structure, and LPo was the most crucial factor in altering bacterial community structure. Reducing the P fertilizer rate (P60) significantly increased the relative abundances of beneficial bacteria (i.e. Anaerolinea, Anaeromyxobacter, Kaistia, and Rhodobacter), while high P fertilizer application rates significantly decreased the relative abundances of beneficial bacteria (i.e. Mesorhizobium, Microvirga, Rhizobacter, Skermanella). Bacterial community diversity and composition had significant and indirect effects on maize yield mediated by Ca 8 -P, O-P, LPo, and MLPo. The above results suggested that reducing P fertilizer rates increased bacterial community diversity and significantly altered bacterial community composition, with an especially strong effect on beneficial bacterial reproduction. We recommend 60 kg P 2 O 5 ha -1 of mineral P fertilizer as the optimal application rate for spring maize cropping in the Mollisol region of northeast China.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This study was supported by the Special Funds Project for Central Government Guides Local Science and Technology Development in Jilin Province (Grant number 202002013JC) and the National Natural Science Foundation of China (41471196).