Maize (Zea mays L.) genotypes induce the changes of rhizosphere microbial communities

Plant–microbe interactions affect ecosystem function, and plant species influence relevant microorganisms. However, the different genotypes of maize that shape the structure and function of the rhizosphere microbial community remain poorly investigated. During this study, the structures of the rhizosphere microbial community among three genotypes of maize were analyzed at the seedling and maturity stages using high-throughput sequencing and bioinformatics analysis. The results demonstrated that Tiannuozao 60 (N) showed higher bacterial and fungal diversity in both periods, while Junlong1217 (QZ) and Fujitai519 (ZL) had lower diversity. The bacterial community structure among the three varieties was significantly different; however, fewer differences were found in the fungal community. The bacterial community composition of N and QZ was similar yet different from ZL at the seedling stage. The bacterial networks of the three cultivars were more complex than the fungal networks, and the networks of the mature stages were more complex than those of the seedling stages, while the opposite was true for the fungi. FAPROTAX functional and FUNGuild functional predictions revealed that different varieties of maize were different in functional abundance at the genus level, and these differences were related to breeding characteristics. This study suggested that different maize genotypes regulated the rhizosphere bacterial and fungal communities, which would help guide practices.


Introduction
The microbial community is a vital part of the ecological environment (Delmont et al. 2011). The interaction between plants and microbial communities is an essential link in the functioning of ecosystems (Vivanco et al. 2018) and can affect agricultural ecosystems (Teste et al. 2017). Many studies have shown that the rhizosphere is a crucial zone for plant-microbe interactions (Bulgarelli et al. 2015;Müller et al. 2016;Saleem et al. 2016;Zachow et al. 2014). The rhizosphere microbes can defend plants against pathogens through competition, antagonism, or interference with host immunity to ascertain a mutualistic association with the host (Bakker et al. 2013;Lu et al. 2018;Agler et al. 2016;Mendes et al. 2013;Mendes et al. 2011;Pozo and Azcón-Aguilar 2007;Raaijmakers and Mazzola 2016;Reinhold-Hurek et al. 2015;Ritpitakphong et al. 2016;Yu et al. 2019). They can promote plant growth by increasing nutrient availability, manufacturing plant hormones, enhancing tolerance to abiotic stresses, and adapting to environmental variations to enhance host immune functions (Etesami et al. 2014;Haney et al. 2015;Xu et al. 2015;Mansotra et al. 2015;Berg and Koskella 2018;Rolli et al. 2015). The structure and function of plant microbiome change with changes in stress and environmental stimuli (Gardener and Weller 2001;Ferrando and Scavino 2015;Santos-Medellín et al. 2017;Timm et al. 2018). Plants also affect microbial communities by producing completely different detritus and excretions (Cline and Zak 2016;Zhalnina et al. 2018) and have the capacity to vary soil microbiota by secreting bioactive molecules into the rhizosphere (Schlaeppi et al. 2014;Zgadzaj et al. 2016). Therefore, understanding the interaction between microbiota and plants has important agronomic significance.
Some studies have demonstrated that plant types change the microbial community under stable environmental conditions. For example, plant genotypes have specific effects on the wheat rhizosphere microbial community (Simonin et al. 2020). The investigation of the rhizosphere bacterial community of twelve rabbit-eye blueberry (RB) cultivars demonstrated that the rhizosphere of the plant cultivars affected bacterial association networks (Jiang et al. 2017). By analyzing the rhizosphere microbial community structure and activity of maize plants, the results indicated that the rhizosphere microbial community was relevant to the plant genotype (Hou et al. 2018). These studies demonstrated that plant genotype influences microbial community composition in controlled environments.
Maize (Zea mays L.) is one of the foremost versatile emerging crops with wider adaptability under varied agroclimatic conditions (Chandel 2018). It is, therefore, essential to understand the microbial communities of maize interrhizosphere soils. This study investigated the bacterial and fungal communities of three different maize cultivars using high-throughput sequencing methods. This study aimed to explore the composition, structure, and interactions of microbial communities from different maize cultivars. We speculated that specific differences exist among different genotypes of maize in regard to rhizosphere microorganisms. We speculated that the plant host regulated the differences and the changes were related to plant characteristics. We hope that this work can provide novel insights into the understanding of the microbiological community of maize.

Experimental site, crop varieties, and soil physicochemical properties
Experimental soils were collected at a depth of 15 cm from three major genotypes of corn and were set up in a complete randomized design in Qiqihar, Heilongjiang Province, China (123° 74′ 90.67″ E, 47° 40′ 43.17″ N; altitude: 146 m) on April 18, 2020, and September 13, 2020. The average annual precipitation in this area is 670.8 mm, and the soil at the study site is sandy loam soil. A total of 54 rhizosphere soil samples (9 treatments × 3 replicates per treatment × 2 periods) were collected from the Zea mays L rhizosphere. Blank soil physical and chemical properties are given in Supplementary Table S1. The volumetric method was used to measure the content of organic matter (SOC), available nitrogen (AN), and total nitrogen (TN). UV-Vis spectrophotometry was used to measure the content of total phosphorus (TP) and available phosphorus (AP). Total potassium (TK) and available potassium (AK) were determined by the inductively coupled plasma-atomic emission spectrometry (ICP-AES) method.
The different genotypes included foodstuff-type maize Fujitai519 (ZL), Junlong1217 (QZ), and Tiannuozao60 (N) in this study. The sample properties are shown in Supplementary Table S2. Rhizosphere samples were collected unbroken in a sterile bag within an ice-containing box and transported to the laboratory. The soil was removed of all impurities with tweezers, the rhizosphere soil was gently swept with a brush, and the samples were stored in a refrigerator at 4 °C (Barillot et al. 2013).

DNA extraction, PCR amplification, and illumina MiSeq sequencing
In this experiment, a total of 66 samples from two periods were used for sequencing. Microbial community genomic DNA was extracted from rhizosphere soil samples using the E.Z.N.A. ® soil DNA Kit (Omega Bio-Tek, Norcross, GA, US) according to the manufacturer's instructions. Distinct regions of bacterial 16S rRNA and fungal ITS genes were amplified using primers. The hypervariable region V3-V4 of the bacterial 16S rRNA gene was amplified with primer pairs 338F (5′-ACT CCT ACG GGA GGC AGC AG-3′) and 806R (5′-GGA CTA CHVGGG TWT CTAAT-3′) and the ITS gene was amplified with primer pairs ITS1F (5′-CTT GGT CAT TTA GAG GAA GTAA-3′) and ITS2R (5′-GCT GCG TTC TTC ATC GAT GC-3′) by an ABI GeneAmp ® 9700 PCR thermocycler (ABI, CA, USA).
PCR amplification of the 16S rRNA gene was performed as follows (Supplementary Table S3). The PCR product was extracted from a 2% agarose gel and purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) according to the manufacturer's instructions and quantified using a Quantus™ Fluorometer (Promega, USA). Purified amplicons were pooled in equimolar amounts and paired-end sequenced on an Illumina MiSeq PE300 platform (Illumina, San Diego, USA) according to the standard protocols by Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China).

Molecular biological analysis and statistical analysis
The raw 16S rRNA and ITS gene-sequencing reads were demultiplexed, quality-filtered by fastp version 0.20.0 (Chen et al. 2018), and merged by FLASH version 1.2.7 (Magoč and Salzberg 2011). Operational taxonomic units (OTUs) with a 97% similarity cutoff (Edgar 2013;Stackebrandt and Goebel 1994) were clustered using UPARSE version 7.1 (Edgar 2013), and chimeric sequences were identified and removed. The taxonomy of each OTU representative sequence was analyzed by RDP Classifier version 2.2 (Wang et al. 2007) against the 16S rRNA and ITS databases using a confidence threshold of 0.7.
Analysis of alpha diversity and beta diversity of the normalized dataset was performed using R (version 3.2.5, USA) and the Vegan package. Nonmetric multidimensional scaling (NMDS) was used to visualize the dissimilarities (Taguchi and Oono 2005). Differentially abundant features were analyzed using Metastat analysis in R (version 3.2.5, USA). Microbial diversity and the relative abundance of different microbial taxonomic levels were assessed for differences between different groups with one-way ANOVA, a t test, or a Kruskal-Wallis H test using SPSS software (SPSS, Chicago, IL, USA). Graphical representations were generated with GraphPad Prism 5 (GraphPad Software, Inc., La Jolla, CA, USA).

Co-occurrence network and functional prediction analysis
Phylogenetic molecular ecological networks (pMENs) from the three cultivars were constructed to investigate microbial interactions. The process was performed using the molecular ecological network analysis pipeline (MENA, http:// ieg4. rccc. ou. edu/ mena/ login. cgi) (Deng et al. 2012). The cooccurrence network is based on all genus levels. Correlation networks were visualized using Gephi software (Brughmans 2013). To visualize the associations in the network, we constructed a correlation matrix by calculating the possible pairwise Spearman's rank correlations (Junker and Schreiber 2010). FAPROTAX and FunGuild software (https:// github. com/ UMNFuN/ FUNGu ild) were used to predict the metabolic function of the microbial flora (Nguyen et al. 2016;Zhou et al. 2020) based on the literature on cultured representatives. All raw sequence data are accessible in the NCBI Sequence Read Archive (SRA) database under BioProject number PRJNA775859.

Analysis of alpha diversity and beta diversity of the microbial community
In total, 12 bulk soil (CK) and 54 rhizosphere soil samples were sequenced in the experiment; a profile of bacterial and fungal communities was achieved to reveal the microbial community structure of the maize rhizosphere. For the entire sampling set, a total of 3,787,073 raw bacterial sequences with an average length of 415 bp and 3,458,212 fungal sequences with an average length of 233 bp were identified using Illumina MiSeq analysis. After a series of preprocessing steps, 1,616,406 bacterial and 2,067,318 qualified fungal reads were classified into 6632 bacterial and 1671 fungal OTUs. All sampling efforts reached the saturation plateau using rarefaction analysis. They effectively covered the total extent of almost a majority of the bacterial and fungal diversity, consistent with 97 sequence similarities in the rank abundance curve approach ( Supplementary Fig. S1).
The alpha diversity of bacterial communities of different species was reflected by the Shannon index, Simpson index, and Chao 1 index. The results showed that the alpha diversity (Chao) at the seedling stage of the bacterial communities in CK was significantly lower than that in N and ZL (P ≤ 0.05). However, there was no significant difference in the Shannon and Simpson indices. There was a substantial difference in the alpha diversity at the maturity stage of bacterial communities (P ≤ 0.05) (Fig. 1a). For fungi, the alpha diversity (Simpson and Shannon) in the rhizosphere soil of all three genotypes of maize was completely different compared with CK at the seedling and maturity stages; Simpson index values of ZL and N were significantly different in both periods of the fungus (P ≤ 0.05) (Fig. 1d). Cultivar N showed higher bacterial and fungal diversity in both periods.
NMDS analysis was conducted based on Bray-Curtis distances to visualize the differences in community composition. The overall bacterial community composition was separated from before planting at the seedling and maturity stages (stress = 0.078; stress = 0.079). The bacterial community was separated into cultivar groups. However, it was not as distinct as CK (Fig. 1b,c). For fungi, four treatments were also clearly separated (stress = 0.069; stress = 0.080) (Fig. 1e,f), and bacterial communities were less distinct than Fig. 1 Alpha diversity and beta diversity of bacterial and fungal communities for three cultivars in two growth periods. a Simpson diversity, Shannon diversity, and the Chao1 diversity of the bacterial level at the seedling and maturity stage; and d Simpson diversity, Shannon diversity, and the Chao1 diversity of the fungal level at the seedling and maturity stage. NMDS analysis of the bacterial and fungal community composition at the genus level based on Bray-Curtis distance. b NMDS analysis of bacterial level at the seedling stage; c NMDS analysis of bacterial level at the maturity stage; e NMDS analysis of fungal level at the seedling stage; and f NMDS analysis of fungal level at the maturity stage. Significances between different groups were compared using Wilcoxon's test, with the results indicated on the top (P ≤ 0.05 is marked as *, P ≤ 0.01 is marked as **, and P ≤ 0.001 is marked as ***). The abscissa is the group name, and the ordinate is the average value of the index fungal communities (Fig. 1b,c,e,f). The results indicated that obvious differences existed in the microbial community structure of the rhizosphere among genotypes of maize.

Microbial community composition of different maize cultivars
The soil bacterial sequences were assigned to 8 phyla (others < 0.01) in 95% of corn cultivar samples and included the phyla Actinobacteria, Proteobacteria, Acidobacteriota, Chloroflexi, Firmicutes, Bacteroidota, Gemmatimonadota, and Myxococcota. Actinobacteria accounted for the most significant proportion, 35.58-43.86% of all OTUs, while Proteobacteria and Acidobacteria were also the main bacteria in the soil, with amounts of 17.54-21.78% and 12.73-17.84%, respectively (Fig. 2a). There was a difference in relative bacterial abundance before and after planting, both at the seedling and maturity stages, with a significant increase in Actinobacteria and a substantial decrease in Proteobacteria in the three cultivar groups compared with CK ( Fig. 2a, Supplementary Fig. S2a). Some phyla, such as Actinobacteria, Acidobacteriota, and Gemmatimnadota, had significantly different relative abundances between cultivars. Actinobacteria differed among cultivars at the seedling stage but not significantly at maturity. Acidobacteriota of cultivar ZL differed from other cultivars at the seedling stage; Gemmatimnadota differed among cultivars at the maturity stage. In comparing different periods of the same cultivar, differences were found in Actinobacteria and Bacteroidota at both seedling and maturity stages (ANOVA. P < 0.05) (Fig. 2a,   Supplementary Fig. S2a). For fungi, compared with CK, the Ascomycota and Mortierellomycota in the cultivar groups showed a significant increase in both periods. In contrast, the Basidiomycota phylum showed a considerable decrease, most significantly in the N cultivar (ANOVA. P < 0.05) (Fig. 2b, Supplementary Fig. S2b).
The soil community composition at the genus level was also different among cultivars. In the top 50 genera, the bacterial community was dominated by Arthrobacter, Rubrobacter, and Blastococcus. Sphingominas, Microlunatus, and Rubrobacter had lower abundances in the CK group, while RB41, Paenibacillus, Streptomyces, and Pseudonocardia had lower abundances after planting. Comparison between seedling and maturity stages also revealed that Gaiella, Nocardioides, Pseudoncardia, and Blastococcus were generally less abundant at the seedling stage (Fig. 3a). There were 20 dominant species at the seedling stage and 22 dominant species at the maturity stage (average abundance > 1%). By comparing the dominant species of the two varieties during the same period, further analysis showed that the bacterial community composition of cultivars N and QZ was similar but different from that of cultivar ZL at the seedling stage. At the seedling stage, the abundance of Pseudonocardia and Blastococcus was significantly higher (P < 0.05) in N and QZ compared with that in ZL (Fig. 3a,c); Streptomyces and Nocardioides in cultivar N were significantly higher (P < 0.05) compared with those in ZL 9 (Fig. 3a,c). At maturity, Blastococcus and Microvirga were significantly higher (P < 0.05) in cultivar N than in cultivar QZ, while the relative abundance of Rubrobacter and RB41 was relatively higher (P < 0.05) in cultivar ZL (Fig. 3a,d).  Heatmap of the soil bacterial groups and fungal groups of the top 50 genera. c-f Significance test of differences between three different rhizosphere maize groups with dominant species based on Student's t test at the genus level (P ≤ 0.05 is marked as *, P ≤ 0.01 is marked as **, and P ≤ 0.001 is marked as ***) The fungal community composition at the genus level varied among cultivars at different periods. By analyzing the relative abundance of fungi at the genus level, the community was dominated by Tausonia, Gibberella, Pseudombrophila, Schizothecium and Mortierella (Fig. 3b). There were 12 dominant genera at the seedling stage and 14 dominant genera at the maturity stage (average abundance > 1%). Comparisons indicated that Gibellulopsis, Monondicty, Sporoirmia, and Tausonia decreased after planting, especially in cultivar N (Fig. 3b), while the abundance of Pseudombrophila increased significantly (P < 0.05) (Fig. 3b,f). Tausonia had a higher relative abundance in cultivar ZL, while Monocillium and Coprinopsis had a higher relative abundance in cultivar QZ (Fig. 3b). However, further analysis showed that the bacterial community composition of dominant genera in cultivars N and QZ was similar at the seedling stage (Fig. 3e). At maturity, the abundance of Rhizoctonia was significantly higher (P < 0.05) in cultivars (Fig. 3b); the relative abundances of Gibberalle and Chaetomium differed significantly (P < 0.05) between cultivars ZL and QZ (Fig. 3f). Compared with bacteria, the fungal differences were not particularly significant.

Topological properties of the bacterial and fungal co-occurrence network
Co-occurrence networks were built to construct highthroughput sequencing data at the genus level. Network plots revealed that the three cultivars had different microbial co-occurrence network structures. The modular architecture was used as the coloring unit to visualize the ecological network of bacteria and fungi. Modules with less than 1% of nodes are shown in gray, with the node size proportional to the corresponding relative abundance. The same module represents the same trend of change. The red edges represent positive correlations, and the blue and green edges represent negative correlations.
The complexity of the co-occurrence network varied depending on both the variety and period. Different R 2 values for all groups indicate that the networks formed possessed scale-free properties ( Table 1). The network was looser at the seedling stage than at the maturity stage, suggesting that the network was more consistent at maturity (Fig. 4a). Fungal communities were compact at the seedling stage but loose at maturity (Fig. 4b). The network was dominated by red lines in the bacterial and fungal networks, indicating that positive interactions were higher than negative interactions. However, this was not the case for fungi in the mature stage of the N species (Fig. 4a,b).
Moreover, the average clustering coefficients, average path distance, and other parameters were also different between the twelve networks (Table 1). Comparing the three bacterial networks at the seedling stage, we discovered that the average path length (GD) of the ZL network was longer than those of the N and QZ networks, while the average connectivity and clustering coefficients (avgCC) of the N network were relatively higher. QZ had a higher average degree (avgK) and a shorter average path distance (GD) ( Table 1). However, in regard to maturity, the opposite was true (Fig. 4a, Table 1). For fungi, the network of the QZ cultivar was the tightest in the two periods (Fig. 4a, Table 1). Overall, network analyses indicated higher stability properties in the bacterial networks than in the fungal networks (Fig. 4a, Table 1).
Centrality is a concept commonly used in network analysis to express how each point in a network is related to other issues in the network. The nodes of maximum stress centrality were also different in all groups, such as Aridibacter, Phyllobacterium, and Pedomicrobium; none of these genera were found in the top 50 genera (Table 1). The degree of a node indicates the number of nodes in the network that are directly connected to that node, with higher connectivity indicating higher importance of the node in the overall network. The nodes with a maximum degree also varied in different cultivars, but the virtual nodes in the same cultivar in different periods always had consistency (Table 1). Different nodes may play distinctly different topological roles in a network, which is shown in Supplementary Table S4 in our experiment. Different cultivars had different key nodes at different periods. At the seedling stage, the connector found in cultivar N was Herbinix, and the module hubs were Bacillus, Gaiella, RB41, and so on. In cultivar QZ, Bacillus and Phyllobacteria were the module hubs. Halocella and Psychrobacillua (a kind of bacillus) were the module hubs in cultivar ZL (Supplementary Table S4). At maturity, the connector found in cultivar N was Rhpdpcytophage; the module hubs were Asanoa, Conexobacter, and Sporosarina. In cultivar QZ, Microbacterium and Nannocystis were the module hubs. Galella, Herpetosiphon, and Bdellourbrio were the module hubs and connectors in cultivar ZL (Supplementary Table S5). Similar results were found for the Z-P table of fungi, with different cultivars having different key nodes (Supplementary Tables S6, S7).
We constructed Spearman's analysis of the top 50 genera of well-functioning bacteria and pathogen-associated fungi to find the relationships between different bacteria and fungi (Fig. 4c). We discovered that Blastococcus RB41 was significantly positively (P ≤ 0.05) correlated with Bacillus and negatively correlated with Rhizoctonia; Gaiella was significantly negatively correlated (P ≤ 0.05) with Elin6055; and Microvirga was significantly negatively (P ≤ 0.05) correlated with Gribberella (Fig. 4c). Bacillus was the key node in both cultivars. In contrast, nodes in fungi with the maximum degree and stress centrality converged in the same cultivar; therefore, we compared the two and found more associations between key fungi than with Bacillus ( Supplementary Fig.  S3). Fungi and bacteria were in a state of mutual constraint and interdependence, and interestingly, pathogenic bacteria in fungi did not all show a negative correlation (Supplementary Fig. S4).

FAPROTAX functional and FUNGuild function predictions
FAPROTAX functional predictions were made for bacteria in the soil at different periods, with the highest abundance of bacteria related to chemoisotropy and aerobic-chemoheterotrophy (Fig. 5a, Supplementary Fig. S5), indicating that the microorganisms were closely related to plant metabolism. Microorganisms related to the degradation of aromatic compounds and cellulose decomposition were more abundant in N, microorganisms related to nitrogen fixation, and microorganisms related to lignin degradation were more abundant in QZ (Fig. 5a). In the seedling stage, ZL was closely related to nitrate reduction and nitrate respiration, and microorganisms related to photoautotrophy were less abundant at maturity Fig. 4 Ecological networks of fungal and bacterial communities and topological properties. a Ecological networks of bacterial communities. b Ecological networks of fungal communities. Every node signifies a genus. Nodes area unit colored by modules. Blue and green lines indicate positive relationships between two individual nodes, and red lines indicate negative relationships. c Spearman analysis of well-functioning bacteria and pathogen-associated fungi at the level of the first fifty genera (P ≤ 0.05 is marked as *, P ≤ 0.01 is marked as **, and P ≤ 0.001 is marked as ***) than at the seedling stage, a phenomenon in contrast to cultivar N (Fig. 5a, Supplementary Fig. S5). FUNGuild function predictions for fungi at both periods revealed that the QZ and N cultivars had fewer fungi associated with plant pathogens at maturity with QZ. Cultivars had the fewest fungi associated with plant pathogens at maturity (Fig. 5b,c).

Discussion
The interaction between plants and microorganism communities is a critical driving issue for ecosystem functions (Graham et al. 2014;Wu et al. 2019). Soil microorganisms play a significant role in the decomposition of organic matter (Collado et al. 2019), nutrient acquisition (Chibucos and Tyler 2009), and soil nutrient dynamics (Hou et al. 2018). In the study, the diversity and structure of the rhizosphere microbiota from different genotype maize were detected.
The diversity of the soil microbial community is a basic life characteristic of the soil ecosystem and is closely associated with changes in the ecosystem functions (Griffiths et al. 2000). By analyzing the α diversity and β diversity of the microbial community, the results demonstrated that the microbial composition of maize rhizosphere was different among the three genotypes, which was confirmed in both periods. The α diversity index of N was the highest (Fig. 1a,  d), and the α diversity index of ZL was the lowest in both periods. Normally, higher soil microbial diversity indicated that higher degree of stability of the ecosystem (Xun et al. 2021;Zhao et al. 2018). The rhizosphere microbial communities of different genotypes varied with different growth stages of plant (Sohn et al. 2021), which could be due to other factors (e.g., soil temperature, pH, etc.) not measured in this study (Pietikainen et al. 2005). These suggested that the host genotype contributed to a considerable portion of the variation in rhizosphere microbial diversity of maize, which was also confirmed by Peiffer et al. (2013).
Proteobacteria and Actinobacteria were the dominant phyla in rhizosphere of maize, which was consistent with previous generation-sequencing studies on maize cultivated soils (Kong et al. 2020). The soil community composition at the genus level was also different among different genotypes (Fig. 3). In the seedling stage, Blastococcus and Pseudonocardia were significantly higher (P ≤ 0.05) in N and QZ (Fig. 3c), and Pseudonocardia was reported to be mainly related to synthetic antibiotics (Caldera et al. 2019). In the maturity stage, Blastococcus, Pseudonocardia, and Microvirga were significantly higher (P ≤ 0.05) in N than in QZ, while RB41 was significantly higher (P ≤ 0.05) in ZL (Fig. 3d). The abundance of Blastococcus and RB41 in soil changed with salt stress (Kloepper et al. 1980;Wang et al. 2019). The results suggested that the abundance of genera in ZL was more variable and that the cause of the variation in the abundance of ZL may be related to salt stress. Blastococcus was positively correlated with Pseudonocardia, RB41, and Bacillus. Microviga was positively c Analysis of the plant pathogen of three cultivars in fungi based on Student's t test (P ≤ 0.05 is marked as *, P ≤ 0.01 is marked as **, and P ≤ 0.001 is marked as ***) correlated with Springomonas (Fig. 4c), probably because the association between the genera was in different patterns. The results indicated that different genotypes would lead to change in the abundance of genera by recruiting different microbes. The changes of fungal community was not significant between different cultivars (Fig. 3e, f), which suggested that genotype had a greater effect on the microbial community bacteria than fungi.
Network analysis was performed to realize an integrated understanding of the bacterial community assembly rules reflecting ecological processes within the rhizosphere (Layeghifard et al. 2017), and offered new insights into the characteristics of advanced bacterial network structures and keystone populations (Barberán et al. 2012). QZ was more susceptible to changes in the external environment and less resistant to interference, ZL had an excellent buffering capacity for changes in environmental conditions (Zhan et al. 2021). The nodes with a maximum degree also varied in different genotype. In the seedling stage, the connectors found in N were all related to functions such as cellulose degradation and plant growth and development (Koeck et al. 2016;Zhang et al. 2020). The connector found in QZ was related to bacterial exopolysaccharides, which can provide the host plant with antioxidants and protection from corrosive pathogens (Supplementary Table S4) (Bouchotroch et al. 2000;Chi et al. 2019). The connector found in ZL was closely related to salt stress (Heng et al. 2019). Bacillus is a module hub in two cultivars. At maturity, the connectors were very different from seedlings (Supplementary Table S5). Blastococcus was less abundant in ZL than in the other two species, and was an important node (Fig. 3c, Supplementary Table S5). None of the nodes of maximum stress centrality were found in the top 50 genera (Table 1). Some nodes with essential roles in the network were not dominant genera, revealing the vital role of rare species in the network (Deng et al. 2016;Feng et al. 2017). However, for fungi, many nodes with the maximum degree overlape with stress centrality. Fungi and bacteria were in a state of mutual constraint and interdependence, and pathogenic in fungi did not all show a negative correlation with bacteria (Fig. 4c). The results suggested that different genotypes built their own unique microbial communities.
A variety of experimental studies have demonstrated the importance of microbial diversity for system functioning (Delgado-Baquerizo et al. 2016;Hector and Bagchi 2007;Wagg et al. 2014). The highest abundance of bacteria was related to chemoisotropy and aerobic chemoheterotrophy (Fig. 5a, Supplementary Fig. S5), indicating that microorganisms were closely related to plant metabolism. The variations in root exudates explained how these corn cultivars managed their rhizosphere bacterial ecosystems together with the core and cultivar-specific microbiota ( Fig. 5a, Supplementary Fig. S5) (Mendes et al. 2014). Ecological function predictions supported network analysis, confirmed a preference trend for plant-mediated microbial community change. Different maize genotypes all prefered their corresponding ecological functions and recruited their specific interrooted microbes according to their preferences (Fig. 5a, Supplementary Fig. S5). Bacteria associated with plant pathogens were more abundant at the seedling stage than at maturity (Fig. 5b, c). Several recent studies have proved that the surface pathogen infection can induce the assemblage of a plant-beneficial bacterial consortium in the root microbiome (Berendsen et al. 2018;Yuan et al. 2018). Therefore, the results indicated that plants secreted interroot secretions that were involved in a 'call for help' strategy and actively engaged their microbes to maximize their survival and growth when affected by external stresses, leading to an enrichment of beneficial bacteria that become essential members of the dominant network (Liu et al. 2019), and these also closely related to the characteristics of the plant.

Conclusions
In summary, the different genotypes of maize regulated the rhizosphere microbial community structure with secreting root exudates to build their unique microbial communities. Genotype had a greater effect on the bacterial community than fungi. The study further confirmed the pattern of genotype-induced microbial community assembly, and offered a framework for the more development of methods using plant microbiomes.