The Effects of Microbial Inoculants on Bacterial Communities of the Rhizosphere Soil of Maize

The bacterial community of rhizosphere soil maintains soil properties, regulates the microbiome, improves productivity, and sustains agriculture. However, the structure and function of bacterial communities have been interrupted or destroyed by unreasonable agricultural practices, especially the excessive use of chemical fertilizers. Microbial inoculants, regarded as harmless, effective, and environmentally friendly amendments, are receiving more attention. Herein, the effects of three microbial inoculants, inoculant M and two commercial inoculants (A and S), on bacterial communities of maize rhizosphere soil under three nitrogen application rates were compared. Bacterial communities treated with the inoculants were different from those of the non-inoculant control. The OTU (operational taxonomic unit) numbers and alpha diversity indices were decreased by three inoculants, except for the application of inoculant M in CF group. Beta diversity showed the different structures of bacterial communities changed by three inoculants compared with control. Furthermore, key phylotypes analyses exhibited the differences of biomarkers between different treatments visually. Overall, inoculant M had shared and unique abilities of regulating bacterial communities compared with the other two inoculants by increasing potentially beneficial bacteria and decreasing the negative. This work provides a theoretical basis for the application of microbial inoculants in sustainable agriculture.


Introduction
Bacterial communities of rhizosphere soil are of vital importance to the growth of field crops and agricultural productivity [1]. Beneficial bacterial communities that are integrated into host plants contribute to the appreciating cycle of soil nutrients and high nutrient use efficiency [2,3]. The growth of beneficial bacteria and the reduction in pathogens result from the interaction between the rhizosphere and roots of their host plants [4], which can simultaneously promote the growth of crops and enhance induced systemic resistance in host plants against pathogens, soil-borne diseases, and other environmental stresses caused by abiotic factors [5]. Appropriate bacterial structure and functions, which are associated with microbial diversity, are the key drivers that can maintain the microbial ecosystem of agricultural soil and sustainable development of agriculture [6].
However, the bacterial structure and function have changed due to current unreasonable agricultural practices implemented by human beings, including intensive cultivation, years of continuous cropping, and overuse of chemical fertilizers [7,8]. Among them,

Preparation of Three Inoculants for Application of Field Experiment
Inoculant M was prepared by mixing the two bacterial strains screened above. The two strains were cultured in LB medium at 28 • C for 18-24 h, and they were mixed together at a ratio of 1:1 for application. Inoculant A was offered by Genliyuan Microbial Fertilizer Co. LTD (Hebei, China) and Inoculant S was provided by Otaqi Biological Products Co. LTD (Beijing, China). Inoculants A and S were commercial and patented products. Inoculant A mainly contained species of Actinomycetes, Bacillus, and Saccharomyces, as well as Agriculture 2021, 11, 389 3 of 18 some undescribed nitrogen-fixing bacteria and photosynthetic bacteria, while inoculant S contained not only living organisms but also some micro-nutrient such as Cu, Fe, Zn, Mn, and so on. However, detailed information of their composition was unknown. Inoculant A and Inoculant S did not need to be cultured beforehand, and they could be used directly according to the usage described in Table 1.

Conditions and Treatment Design of Field Experiment
The field experiment was conducted at the Institute of Plant Protection, Jilin Academy of Agricultural Sciences (Gongzhuling, Jilin Province, China; 43 • 31 52 N, 124 • 49 31 E, Figure 1) in 2018. Soil conditions of the field experiment are listed in Table S1, and the local climate was monsoon climate of medium latitudes. Twelve treatments, including three levels of nitrogen fertilizer and the three inoculants mentioned above, were used in this study (Table 1). Each treatment had four replications, and the area of every replication was 29.44 m 2 (6.4 m × 4.6 m). The maize seeds ('Jidan 558') were provided by the Biological Pesticide Laboratory, Institute of Plant Protection, Jilin Academy of Agricultural Sciences. All the seeds were sterilized in 10% H2O2 for 15 min, then washed in sterilized water three times [29], and then immersed in the different inoculants for 12 h. All seeds were sown on 30 April in a depth of 10 cm. Additionally, seeds were sown at a spacing between planting rows of 65 cm, and a spacing between plants of 23 cm. Thus, each replication had almost 150 maize plants. The maize seeds ('Jidan 558') were provided by the Biological Pesticide Laboratory, Institute of Plant Protection, Jilin Academy of Agricultural Sciences. All the seeds were sterilized in 10% H 2 O 2 for 15 min, then washed in sterilized water three times [29], and then immersed in the different inoculants for 12 h. All seeds were sown on 30 April in a depth of 10 cm. Additionally, seeds were sown at a spacing between planting rows of 65 cm, and a spacing between plants of 23 cm. Thus, each replication had almost 150 maize plants.

Sample Collection
Plant samples were collected by the quadrat method, in which a 2 m 2 (2 m × 1 m) quadrat was utilized three times in each replication. Bulk soil used for physiochemical detection was collected when the plant samples were dug out. Soil laid in the hole of plant roots and soil dropping from roots were considered as bulk soil. Rhizosphere soil samples were collected after plants were carefully dug out with roots and gently shaken to discard excess soil. Only soil without any aggregates was regarded as rhizosphere soil, which was adhering to the roots very closely [30]. Soil sample was mixed by all collected quadrats in each replication, and quartering was used to acquire the appropriate amount of soil sample for further analyses, from which 0.5 g rhizosphere soil of each replication was used for DNA extraction.

DNA Extraction and Polymerase Chain Reaction (PCR) Amplification
The MP DNA extraction kit (MP Biomedicals, LLC, Solon, OH, USA) was used for DNA extraction of rhizosphere soil samples according to the manufacturer's instructions. The V4-V5 region of the 16S rDNA gene was amplified from the bacterial DNA by PCR using barcode 515F (GTGCCAGCMGCCGCGG) and 907R (CCGTCAATTCMTTTRAGTTT) primers as described elsewhere [31]. The PCR amplification was tested by 1% agarose gel electrophoresis, colored by ethidium bromide for 40 min at 100 V.

Library Construction and Sequencing
The TruSeq ® DNA PCR-Free Sample Preparation Kit (Illumina, San Diego, CA, USA) was utilized in library construction, and the Qubit ® 2.0 Fluorometer (Life Technologies, Carlsbad, CA, USA) and qPCR were utilized to quantify the libraries. Then, the libraries were sequenced at the Illumina MiSeq platform described as Zhang et al. [32]. All sequence data were submitted to the Sequence Read Archive (SRA: SRP297881) and are freely available at the NCBI (BioProject: PRJNA685114).
Subsequently, all the effective tags were clustered through Uparse (Version 7.0.1001) (http://drive5.com/uparse/, accessed on 24 October 2020) [36]. The effective tags were clustered into the same OTUs when their identity was no less than 97%. The OTUs with the highest frequency were chosen to be representatives of OTU sequences. The OTUs that only had one sequence were removed from the dataset because these special OTUs could be caused by sequencing errors. To further explore their functions, a representative sequence of each OTU was assigned to a taxonomic level using the RDP classifier [37].

Bioinformatic Analyses
To explore the differences in richness and diversity of bacterial communities based on inner samples among different groups, after rarefaction, the OTU numbers and alpha diversity, which consisted of the observed species, Shannon, Simpson, Chao1, ACE, Good's-coverage, and PD_Whole Tree indices, were computed by QIIME. On the basis of phylogenetics, the PD_Whole Tree index was utilized to compute Faith's phylogenetic diversity metric. R software (Version 3.6.0, R Foundation for Statistical Computing, Vienna, Austria) was used to draw the rarefaction curves based on graphics, plot, and RColor-Brewer packages.
To further explore the differences in bacterial communities among all samples based on either inner or outer comparisons of different groups, beta diversity was implemented. Unifrac distance metrics were computed by software QIIME (Version 1.7.0) on the basis of an unweighted pair group method with arithmetic mean (UPGMA). The differences among all treatments were demonstrated through PCoA (principal coordinates analysis) and NMDS (nonmetric multidimensional scaling). Thereafter, PCoA, PCA (principal components analysis), and NMDS diagrams came from the vegan, ade4, and ggplot2 packages of R software (Version 3.6.0). To achieve a better perspective into the clustering of bacterial communities, the weighted (taking changes of relative taxonomic abundance into consideration), unweighted UniFrac metrics, and Bray-Curtis distance were utilized for the calculation of beta diversity [39]. Metastats analysis was performed at different taxonomic levels, using the permutation test between groups based on R software (Version 3.6.0) (The R Foundation for Statistical Computing, Vienna, Austria).
Since the alpha and beta diversities were explored, the key phylotypes of all treatments in the CF group were further researched via heatmaps, LefSe (LDA effect size) analysis, and histograms [40]. Heatmaps were operated and clustered by representative bacterial statistics of RDA (redundancy analysis)-identified OTUs. Thereafter, LefSe analysis was implemented by LefSe software on Novogene Platform (Beijing Novogene Technology Co., Ltd., Beijing, China). Histograms were drawn on the basis of the relative abundance of the top 40 species [41].

Statistical Analysis
Raw data were initially preprocessed by Microsoft Excel 2016, and the analyses of variance (ANOVA) were implemented using IBM SPSS statistics 25.0 software (SPSS, Inc., Chicago, IL, USA). Kruskal-Wallis test was used to calculate the p-value in usual analyses based on relative abundance of different taxa in all treatments. Tukey test was used to calculate the p-value in the analysis of Bray-Curtis distance. Permutational multivariate analysis of variance (PERMANOVA), based on vegan, was used for the comparison of bacterial communities of different treatments. Wilcoxon rank-sum test was used to measure the p-value in LefSe analyses. All diagrams and plots were drawn using Origin 2018 (OriginLab Corporation, Northampton, MA, USA) and R (Version 3.6.0), and all tables were drawn directly using Microsoft Word 2016. All data are presented as means ± standard deviation.

Sequencing Results
Sequencing results of amplicon libraries contained samples from twelve treatments, provided 1,070,851 raw data, which was replaced by 1,067,020 after quality control with the high-quality reads' average length of 374 bp. All high-quality reads were assembled, and OTUs were clustered from all qualified tags to study the species diversity of the treatments (Table S2).

Dissimilarity of Bacterial Communities in Different Treatments
The comparison of the OTUs of the different inoculants combined with different nitrogen application rates illustrated that the CF group The variation in the OTUs in the CF group increased from CF to CF.M, and then decreased from CF.M to CF.A and CF.S. The D20N group had a similar tendency as the CF group in terms of using or not using inoculants, but the difference between these two groups was that CF.M had more OTUs than CF.A, while it was opposite in the D20N group. However, the performance of OTUs in the D40N group decreased from D40N to D40N.A (via D40N.M) and increased at D40N.S ( Figure 3B). Taking all of these results into consideration, although the bacterial communities showed different pattern in the D40N group compared with the CF and D20N groups, it was obvious that the diversity of bacterial communities tended to decrease with the utilization of inoculants except Inoculant M and Inoculant S, which were used in the CF and D40N groups, respectively. In the D20N group, the effect of inoculants on reducing the diversity of bacterial communities was weakened. Furthermore, the effects of Inoculant M on bacterial communities were the largest in the CF group based on the OTU richness.   The variation in the OTUs in the CF group increased from CF to CF.M, and then decreased from CF.M to CF.A and CF.S. The D20N group had a similar tendency as the CF group in terms of using or not using inoculants, but the difference between these two groups was that CF.M had more OTUs than CF.A, while it was opposite in the D20N group. However, the performance of OTUs in the D40N group decreased from D40N to D40N.A (via D40N.M) and increased at D40N.S ( Figure 3B). Taking all of these results into consideration, although the bacterial communities showed different pattern in the D40N group compared with the CF and D20N groups, it was obvious that the diversity of bacterial communities tended to decrease with the utilization of inoculants except Inoculant M and Inoculant S, which were used in the CF and D40N groups, respectively. In the D20N group, the effect of inoculants on reducing the diversity of bacterial communities was weakened. Furthermore, the effects of Inoculant M on bacterial communities were the largest in the CF group based on the OTU richness.

Alpha Diversity
According to the results of the rarefaction curves, CF.M had the highest abundance of bacterial communities in the CF group ( Figure 4A), whereas D20N had the highest abundance of bacterial communities in the D20N group ( Figure 4B). When it came to the D40N group, the highest abundance occurred in D40N.S ( Figure 4C), which was consistent with the results of OTU numbers ( Figure 3B). The number of observed species was highest in D40N.S samples at 2449.25 ± 135.71, followed by D20N, D20N.M, and

Alpha Diversity
According to the results of the rarefaction curves, CF.M had the highest abundance of bacterial communities in the CF group ( Figure 4A), whereas D20N had the highest abundance of bacterial communities in the D20N group ( Figure 4B). When it came to the D40N group, the highest abundance occurred in D40N.S ( Figure 4C), which was consistent with the results of OTU numbers ( Figure 3B). The number of observed species was highest in D40N.S samples at 2449. 25  In addition, a trend was found that the diversity of bacterial communities declined in most inoculant-applying treatments in the different nitrogen application rate groups, compared with their own control (CF, D20N, and D40N) in the corresponding groups. Two exceptions were discovered: one was CF.M in the CF group, and the other was D40N.S in the D40N group, whose diversity of bacterial communities was enhanced by Inoculant M and Inoculant S, respectively. The results of alpha diversity were consistent with the statistics of the OTU numbers and their Venn diagrams (Figure 3). The performance of Inoculant M in the CF group (CF.M) was different among all treatments (p-value < 0.05, tested by DMRT). As a consequence, in order to further explore the effects of different inoculants on bacterial communities, we focused on CF group (CF, CF.M, CF.A, CF.S) in the subsequent analyses.  In addition, a trend was found that the diversity of bacterial communities declined in most inoculant-applying treatments in the different nitrogen application rate groups, compared with their own control (CF, D20N, and D40N) in the corresponding groups. Two exceptions were discovered: one was CF.M in the CF group, and the other was D40N.S in the D40N group, whose diversity of bacterial communities was enhanced by Inoculant M and Inoculant S, respectively. The results of alpha diversity were consistent

Beta Diversity
The results of the PCoA based on the unweighted Unifrac distances indicated that the bacterial communities of CF and CF.M were separate. Evident separations between the communities of CF.M and CF.A, CF.M and CF.S, CF and CF.A, and CF and CF.S exist ( Figure 5A). The highest variations in the microbiome of different treatments represented a strong separation between different utilizations of inoculants and their control, except that the communities of CF.A and CF.S were clustered very well. The results of the PCA, which were plotted on the basis of OTU levels, showed a similar trend to that of PCoA ( Figure 5B). NMDS analysis indicated that different microbial inoculants played an important role in shaping the bacterial communities in soil samples of the maize rhizosphere. The stress of NMDS analysis was 0.115, which is regarded as a good model in representing the differences among all treatments. There were high similarities in bacterial communities between the CF.A and CF.S samples, whereas they were both separated from CF.M and CF. The cluster of CF.M samples and CF samples were separated ( Figure 5C). The Bray-Curtis distance demonstrated that the CF.M samples had the highest variation among all samples. A trend was detected where the diversity of the bacterial community in the CF.M samples was enhanced compared with CF, while CF.A and CF.S had little variation between each other ( Figure 6). Interestingly, the bacterial communities, based on the Bray-Curtis distances, were highly similar between CF.A and CF.S (p-value = 0.9031, through Tukey test), except for CF and CF.M (p-value = 0.0396, through Tukey test), CF and CF.A (p-value = 7.21 × 10 −6 , through Tukey test), CF and CF.S (p-value = 1.80 × 10 −6 , through Tukey test), CF and CF.A (p-value = 0.0046, through Tukey test), and CF.M and CF.S (p-value = 0.0010, through Tukey test), which illustrated that the bacterial communities of these treatments were different (Table S3). Additionally, based on the results of PERMANOVA, CF.M, CF.A, and CF.S samples had significantly different bacterial communities from the CF samples. Moreover, the bacterial communities of CF.M were significantly different from those of CF.A and CF.S, respectively, while the last two were similar (Table S4). differences among all treatments. There were high similarities in bacterial communities between the CF.A and CF.S samples, whereas they were both separated from CF.M and CF. The cluster of CF.M samples and CF samples were separated ( Figure 5C). The Bray-Curtis distance demonstrated that the CF.M samples had the highest variation among all samples. A trend was detected where the diversity of the bacterial community in the CF.M samples was enhanced compared with CF, while CF.A and CF.S had little variation between each other ( Figure 6). Interestingly, the bacterial communities, based on the Bray-Curtis distances, were highly similar between CF.A and CF.S (p-value = 0.9031, through Tukey test), except for CF and CF.M (p-value = 0.0396, through Tukey test), CF and CF.A (p-value = 7.21 × 10 −6 , through Tukey test), CF and CF.S (p-value = 1.80 × 10 −6 , through Tukey test), CF and CF.A (p-value = 0.0046, through Tukey test), and CF.M and CF.S (p-value = 0.0010, through Tukey test), which illustrated that the bacterial communities of these treatments were different (Table S3). Additionally, based on the results of PERMANOVA, CF.M, CF.A, and CF.S samples had significantly different bacterial communities from the CF samples. Moreover, the bacterial communities of CF.M were significantly different from those of CF.A and CF.S, respectively, while the last two were similar (Table S4).

Inoculant M Mediated the Key Phylotypes of the Rhizosphere Microbiome of Maize
Since the effect of inoculant M on the bacterial community in maize rhizosphere soil was significantly different from that of inoculants A and S, and the control (CF), key phylotypes were explored to further understand the microbiome in maize rhizosphere

Inoculant M Mediated the Key Phylotypes of the Rhizosphere Microbiome of Maize
Since the effect of inoculant M on the bacterial community in maize rhizosphere soil was significantly different from that of inoculants A and S, and the control (CF), key phylotypes were explored to further understand the microbiome in maize rhizosphere soil and the specific changes of bacterial communities caused by different treatments. From the results of the heatmaps, it was obvious that the variations in species were similar between CF.A and CF.S, but was significantly different from CF at both the phylum and genus levels ( Figure 7). Interestingly, the variation in the structure of the bacterial community in CF.M differed from that of CF.A, CF.S, and CF. The phenomenon was observed where the variation of key phylotypes in CF.M was between CF.A and CF.S, and CF at the phylum level ( Figure 7A). At the genus level, key phylotypes in CF.M distinguished them further ( Figure 7B). Different biomarkers were found in different treatments based on LefSe analysis ( Figure S1 and Table S5). At the phylum level, the biomarker of CF was Firmicutes, while the biomarker of CF.M was Proteobacteria, the biomarkers of CF.A were Actinobacteria and Gemmatimonadetes, and the biomarker of CF.S was Acidobacteria. When it came to genus level, the biomarkers of CF were Aeromonas and Acinetobacter. In contrast, the biomarkers of CF.M were Rhodanobacter and Chujaibacter, and the biomarkers of CF.A and CF.S were Gemmatimonas and unidentified Gammaproteobacteria. In addition, we selected the top 40 species shared in all treatments of the CF group at the genus level to investigate the differences in relative abundance among these treatments (Figure 8). The abundance of Pseudolabrys, Terracidiphilus, Granulicella, Phenylobacterium, Gemmatimonas, and Rhodanobacter were increased in CF.M, CF.A, and CF.S, compared with CF. Nevertheless, the abundance of Ralstonia, Xylophilus, and Comamonas were decreased in CF.M, CF.A, and CF.S (Table S6). Different biomarkers were found in different treatments based on LefSe analysis ( Figure S1 and Table S5). At the phylum level, the biomarker of CF was Firmicutes, while the biomarker of CF.M was Proteobacteria, the biomarkers of CF.A were Actinobacteria and Gemmatimonadetes, and the biomarker of CF.S was Acidobacteria. When it came to genus level, the biomarkers of CF were Aeromonas and Acinetobacter. In contrast, the biomarkers of CF.M were Rhodanobacter and Chujaibacter, and the biomarkers of CF.A and CF.S were Gemmatimonas and unidentified Gammaproteobacteria. In addition, we selected the top 40 species shared in all treatments of the CF group at the genus level to investigate the differences in relative abundance among these treatments (Figure 8). The abundance of Pseudolabrys, Terracidiphilus, Granulicella, Phenylobacterium, Gemmatimonas, and Rhodanobacter were increased in CF.M, CF.A, and CF.S, compared with CF. Nevertheless, the abundance of Ralstonia, Xylophilus, and Comamonas were decreased in CF.M, CF.A, and CF.S (Table S6).
Interestingly, we found that the relative abundance of the genus Dietzia was significantly (p-value < 0.05, through Kruskal-Wallis test) increased only in CF.M while the number of other three treatments was zero. Furthermore, the relative abundance of Rhodovastum was significantly (p-value < 0.05, through Kruskal-Wallis test) higher in Inoculant M than Agriculture 2021, 11, 389 12 of 18 CK whereas there was no significant difference between CF.A, CF.S, and CF (Table S7). When it came to Granulicella, the numbers had significant (p-value < 0.05, through Kruskal-Wallis test) differences between three inoculants treatments and CF, whereas there was no significant difference in relative abundance of Granulicella among these three inoculants. Additionally, both CF.A and CF.S had higher relative abundance of Gemmatimonas than CF.M (p-value < 0.05, through Kruskal-Wallis test), and that of CF.M was significantly higher (p-value < 0.05, through Kruskal-Wallis test) than CF (Figure 9). Interestingly, we found that the relative abundance of the genus Dietzia wa significantly (p-value < 0.05, through Kruskal-Wallis test) increased only in CF.M whil the number of other three treatments was zero. Furthermore, the relative abundance o Rhodovastum was significantly (p-value < 0.05, through Kruskal-Wallis test) higher i Inoculant M than CK whereas there was no significant difference between CF.A, CF.S and CF (Table S7). When it came to Granulicella, the numbers had significant (p-value 0.05, through Kruskal-Wallis test) differences between three inoculants treatments an CF, whereas there was no significant difference in relative abundance of Granulicel among these three inoculants. Additionally, both CF.A and CF.S had higher relativ abundance of Gemmatimonas than CF.M (p-value < 0.05, through Kruskal-Wallis test), an that of CF.M was significantly higher (p-value < 0.05, through Kruskal-Wallis test) tha CF (Figure 9).

Discussion
The bacterial community of rhizosphere soil is known to be associated with the status of agricultural soil: whether it is nutrient efficient [42], whether the elements are conveniently available for plants [43], whether it is sufficient for fertility [44], and whether

Discussion
The bacterial community of rhizosphere soil is known to be associated with the status of agricultural soil: whether it is nutrient efficient [42], whether the elements are conveniently available for plants [43], whether it is sufficient for fertility [44], and whether it is sensitive to pathogens [45]. Apart from reflecting the status of the soil, the bacterial community can influence and change the abiotic and biotic properties of soil [46,47] in return for the habitat (matter and energy) provided by their hosts [48]. With the negative variation in and destruction of bacterial communities resulting from unreasonable agricultural practices such as excessive use of nitrogen fertilizers [49], the physiochemical properties of soil have declined along with biotic factors [50]. To solve this issue, microbial inoculants have been focused on to mediate the microbiome adhering to the roots of host plants. Based on the urgent requirement of microbial regulators, inoculants with more specific and efficient abilities are being sought. It was found that these three inoculants modulated the bacterial communities to better structural and functional formations for maize production, compared with non-inoculant control. Moreover, among all results, the performance of Inoculant M in the CF group (CF, CF.M, CF.A, and CF.S) proved to be significantly different from that of Inoculants A and S based on OTU richness, species abundance, diversity analyses (alpha diversity and beta diversity), and key phylotypes analysis. This suggests that the regulatory effect of Inoculant M on the microbiome was unique to that of Inoculants A and S. Inoculant M is a promising modulator which can improve bacterial communities in maize rhizosphere soil for agricultural practice.
Combining the summaries of alpha (Table 2 and Figure 3) and beta diversity ( Figure 5, Figure 6, Tables S3 and S4), Inoculant M could shape the bacterial community into a differential structure compared by other two inoculants [51]. Previous studies have shown different effects of single bacterial strains and inoculants consisting of several (mostly no more than five) bacterial strains on microbial communities [52][53][54]; however, few have paid attention to the comparison between simple inoculants (mainly consisting of single, two, or three strains) and commercial inoculants referring to complex compositions, and between commercial inoculants themselves. Zhong et al. found that different inoculants led to different assemblies of the microbiome [55]. However, in this study, Inoculant A and Inoculant S led to similar bacterial communities. One of the conjectures was whether the formulae of Inoculants A and S were homologous [56]. The responses of bacterial communities and plants to the application of microbial inoculants are dependent on plant and bacterial genotypes as well [57]. Another hypothesis regarding Inoculant A and Inoculant S was that the formulas could have been different, but were rich enough or sufficiently complex that they provided more than the fundamental requirement of the soil, which might eventually result in a similar microbiome. This hypothesis needs to be tested further by comparison of complex inoculants. Putting the similarity between CF.A and CF.S aside, Inoculant M had unique effects on shaping bacterial communities in the study. To further understand the significantly different genera between different treatments, key phylotype analysis was implemented and discussed [58]. CF.M showed obvious different structure of bacterial community from CF and other two inoculants through heatmaps ( Figure 7A,B). From the LefSe analysis ( Figure S1 and Table S5), the genera whose LDA were larger than 4 were discussed as biomarkers of different treatments. It was reported that the biomarkers of CF, Aeromonas and Acinetobacter were severe pathogens [59,60]. When it came to the biomarkers of CF.M, CF.A, and CF.S, Rhodanobacter and Gemmatimonas were reported to have the ability to improve the circulation of nitrogen in soil. Little information about Chujaibacter could be found in the literature, but one investigation mentioned that it could survive in variable salinity conditions by degrading organic matter as a basis for utilizing N-acetylglucosamine [61]. Demonstrated by relative abundance statistics of the top 40 genera in all treatments ( Figures 7B and 8), many beneficial genera were increased by Inoculant M, Inoculant A, and Inoculant S, such as Pseudolabrys, Terracidiphilus, Granulicella, Phenylobacterium, Gemmatimonas, and Rhodanobacter. Among them, Pseudolabrys, Terracidiphilus, Granulicella, and Phenylobacterium were found to have positive correlations with solubilizing phosphate in soil. Pseudolabrys had been reported to secrete naphthol-AS-BI-phosphohydrolase [62], Terracidiphilus and Phenylobacterium can both secrete alkaline phosphatase (ALP) [63,64], and Granulicella can produce ALP, acid phosphatase (ACP), and naphthol-AS-BI-phosphohydrolase simultaneously [65]. With all the enzymes mentioned above, the process of solubilizing phosphate can proceed smoothly. Additionally, Park et al. revealed that Gemmatimonas can denitrify and break down lignin and cellulose [66].
Rhodanobacter was found to participate in the process of denitrification by Van et al. [67] as well. The genera mentioned above were almost all beneficial bacteria associated with nutrient uptake, plant growth-promotion, and denitrification, which was partly consistent with the results of the LefSe analysis. Meanwhile, some negative bacteria (i.e., potential plant pathogens), including Ralstonia [68], Xylophilus [69], and Comamonas [70], were decreased by the three inoculants. Except for the common variations among three inoculants and CF, some special differences were explored in genera Dietzia and Rhodovastum, whose relative abundances were significantly (p-value < 0.05) increased only by Inoculant M (Figure 9). Bharti et al. found that Dietzia could promote the growth of wheat and protected wheat from salt stress by secreting various enzymes and other molecule organics [71]. Rhodovastum was reported to be a photo-organotrophic bacterium, which was regarded as a beneficial bacterium to plants [72]. Inoculant M modulated the key phylotypes of the microbiome not only by improving the beneficial bacteria as with Inoculant A and Inoculant S, but also enhanced some advantageous bacteria, whose variations were unique to the other two inoculants. This suggests that Inoculant M has unique functions in mediating bacterial communities of maize rhizosphere soil, which makes Inoculant M potentially applicable in maize production. It should be pointed out that the use of microbial inoculants will not cause an increase in production cost. As we all know, cost control is an important part of agricultural production, and the cost of implementing the technology is the basis for its application. The maize yields of CF.M, CF.A, and CF.S were all significantly higher (p-value < 0.05, tested by DMRT) than CF. Furthermore, the maize yield of CF.M was significantly higher (p-value < 0.05, tested by DMRT) than CF.A and CF.S, while there was no significant difference between CF.A and CF.S (Table S8). Nevertheless, the cost of treatment is hard to obtain, since the Inoculant A and Inoculant S were provided freely by corresponding companies for scientific research. We can only calculate the cost of Inoculant M, which is no more than 750 rmb ha −1 . Detailed cost accounting of using these microbial inoculants is currently needed, which will provide better application potential of this technology.

Conclusions
In this study, all three inoculants were able to shape the bacterial communities of maize rhizosphere soil into improving assemblies by increasing potentially beneficial bacteria and decreasing the harmful bacteria, as compared to the non-inoculant control. In particular, Inoculant M showed shared and unique abilities to modulate bacterial communities compared with the other two inoculants, proving that Inoculant M is promising for application in agricultural practices in the future. This study provides data support for the mediation of the microbial community of maize rhizosphere soil by microbial inoculants and a theoretical basis for the application of microbial inoculants in the green, healthy, and sustainable development of agriculture. This article focused on the effects of different inoculants on bacterial communities of maize rhizosphere soil, moreover, the effects of these inoculants on fungal communities and nematode communities should be further researched.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/agriculture11050389/s1, Figure S1: LDA (Linear Discriminant Analysis) plot of LefSe anlysis among different treatments. Table S1: Soil conditions of the field experiment, Table S2: Statistics of the sequencing results, Table S3: Bray-Curtis results, Table S4: PERMANOVA results of bacterial communities treated by different treatments, Table S5: Statistic results of LefSe analysis, Table S6: Relative abundance of top40 genera, Table S7: OUT table, Table S8: The yields of maize in different treatments.