This study was conducted using soil obtained from the park of Jinming campus of Henan University (longitude: 114.35°E, latitude: 34.80°N) in Kaifeng, China. Kaifeng has a temperate monsoon climate, with an annual average temperature of 14°C and an annual average precipitation of 650 mm. Rainfall mainly occurs in July and August every summer. The soil types mainly include fluvo-aquic soils, saline soil, aeolian sandy soil, and alluvial soils.

We selected two gardens to collect soil samples. One location is in the south of the campus and another is in the north (Supplementary Figure S1). In each garden, we collected samples of 9 species of plants. Only four species (Ligustrum lucidum, Forsythia viridissima, Oxalis corniculate, Veronica persica) were collected in both sampling locations. Finally, they were grouped into four vegetation types (arbors, shrubs, herbs, and bulk soil. For the detail information, please see Supplementary Table S1). No specific permissions were required for sample collection and the filed study did not involve endangered or protected species. The distance between two sampling points was more than 3 m. At first, the top litter layer was removed. Before sampling, the sampling tools were wiped with the original soil in the area near the collection location to minimize external interference as much as possible. Subsequently, while wearing disposable gloves, the soil was gently dug with a shovel and the fresh soil sorted to remove stones and to find fine roots (diameter ≤ 2 mm). Taking ginkgo as an example, 0–1 m away from the trunk is the area where fine roots are predominantly distributed, and the range for arbors and shrubs is within 5–20 cm underground. When herbaceous plant samples were collected, the whole plant was taken out as completely as possible to look for fine roots. After shaking off the loose soil, the soil adhering to the fine roots over a 1-mm layer was brushed off and collected as the rhizosphere soil sample. All the soil samples were put in sterile bags (stored in dry ice boxes), transported back to the laboratory within 2 h, and then stored at −80°C until DNA extraction. Bulk soil with no plant roots collected in the adjacent area, at a distance more than 5 m from the nearest sampling point was collected as the control soil sample.

A Power Soil DNA kit (MO BIO, Carlsbad, CA, United States) was used to extract the total DNA from soil samples. Afterward, a NanoDrop ND-1000 spectrophotometer (Thermo Scientific, Waltham, MA, United States) was used to quantify the DNA concentrations of samples. The extracted DNA was diluted to approximately 25 ng/μL with distilled water and stored at −20°C until use.

Rhizosphere and bulk soil bacterial and fungal community were tested by high-throughput sequencing techniques at IlluminaNovaSeq platform of MAGIGENE Company, Guangdong, China.¹ The V3–V4 hypervariable regions of bacterial 16S rRNA genes were amplified with the 338F and 806R primer set (Xu et al., 2016). And the ITS2 region of fungi was amplified using the ITS3F and ITS4R primer set (Toju et al., 2012; Supplementary Table S2). The Polymerase Chain Reactions (PCR) were implemented as follows: 3 min of denaturation at 95°C, followed by 30 cycles of 30 s at 95°C, 30 s at 55°C for annealing, and 45 s at 72°C for elongation, with a final extension at 72°C for 10 min. The reactions were carried out in 20-μL triplicate mixtures, each containing 4 μL of 5× FastPfu Buffer, 2 μL of 2.5 mM dNTPs, 0.8 μL of each primer (5 μM), 0.4 μL of FastPfu Polymerase, and 10 ng of template DNA (Toju et al., 2012; Xu et al., 2016).

The raw data sequences were processed and analyzed using QIIME2 (Bolyen et al., 2019) based on the workflow at https://qiime2.org. Briefly, to obtain the Amplicon Sequence Variant (ASV) table, quality control of the raw sequencing data was performed using the Deblur tool (Amir et al., 2017) and clustered based on 100% shared identity. The taxonomy of each bacterial phylotype was identified using the Greengenes release database (DeSantis et al., 2006) and the fungal taxonomy assignment was performed using the Sklearn-based taxonomy classifier with the dynamic Unite database from 10 October 2017.² Finally, we obtained 12,730,609 bacterial sequences in and 9,153,794 fungal sequences, with 98.5% classified into 104,824 distinct ASVs in bacteria and 13,233 in fungal distinct ASVs. To rarify all datasets for each sample to the same degree, 1,122,000 and 3,089,100 bacterial and fungal sequences, respectively, were selected randomly.

In order to measure the difference between groups, the relative abundance of the top 10 microorganisms were logarithm transformed and then we used the LSD method of “agricolae” package for post-hoc test. To assess the abundance and diversity of microbial communities, the Shannon, Simpson, Chao1, and Observed species indices were calculated at the ASV level using QIIME2. The vegan package was used to calculate the β-diversity (Bray-Curtis and Jaccard distance) of bacterial and fungal communities in the arbor, shrub, and herbage rhizosphere soils, and bulk soil (Oksanen et al., 2020). Differences in diversity among the samples were analyzed using the Wilcoxon rank-sum test. Nonmetric multidimensional scaling (NMDS) ordinations was generated to distinguish the distribution of the samples based on Jensen-Shannon divergence (JSD), using the vegan package (Oksanen et al., 2020). Differences between communities were evaluated using Permutational Multivariate Analysis of Variance using the “Adonis” function in the vegan package.

To construct the phylogenetic networks of the 14 plant species, plastome sequences were downloaded from the GenBank database (Supplementary Table S3). Maximum Likelihood (ML) analysis was performed using CIPRES Science Gateway v3.3 (Miller et al., 2010) and RAxML v8.1.11(Stamatakis et al., 2008), with GTR + T + G as the optimal substitution model. The default parameter settings were used, except for the bootstrap iterations being set to 1,000. The phylogenetic distance between each species was calculated in the PAML program (Yang, 2007).

According to Stegen et al. (2013b, 2015), the ß-NTI and Bray-Curtis-based Raup-Crick metrics (RC-Bray) methods were jointly used assess the community assembly processes. β-NTI measures the deviation of the β-mean nearest taxon distance (β-MNTD) and the β-MNTD of the null model, and both were calculated using Phylocom v42 (Webb et al., 2008).

Traits regulating community assembly processes should be phylogenetically conserved (Stegen et al., 2012). Therefore, a phylogenetic signal analysis is required before calculating β-NTI. The relationship between phylogenetic distances of pairwise ASVs and the corresponding environmental conditions was evaluated using “mantelcorrelog” (Stegen et al., 2012), based on the phylogenetic distances calculated using the “cophenetic” function in the “picante” package in R v4.1.3 (R Foundation for Statistical Computing, Vienna, Austria). The Euclidean distance of each soil variable of pairwise ASVs was calculated and the abundance-weighted mean value obtained. Significant relationships within a short phylogenetic distance indicate that phylogenetic signals are also significant. | β-NTI | > 2 indicates a community that is dominated by deterministic processes (Stegen et al., 2012). Conversely, | β- NTI | < 2 indicates that stochastic processes, including DL, HD, and UD, are dominant in the community (Stegen et al., 2015; Tripathi et al., 2018; Feng et al., 2018b).

In the present study, the bacterial ASVs with relatively high abundances (i.e., > 0001%) were selected (3,000 ASVs in our study) for use in calculating the β-NTI and RCbray values (Shi et al., 2018; Feng et al., 2018a).

Co-occurrence network analyses were conducted based on a SparCC correlation matrix using the SpiecEasi package in R (Friedman and Alm, 2012; Kurtz et al., 2015). To enhance the reliability of the networks, the ASV table was filtered. We constructed four networks corresponding to the four types of samples, including the rhizosphere arbor, shrub, and herb soil, and bulk soil. We only retained ASVs present in more than 20% of the bacterial samples and more than 30% of the fungal samples for each sample type. For each group, the selected fungal and bacterial ASVs were used to jointly construct the microbial network, with 533, 401, 393, and 754 ASVs retained in the herb, shrub, and arbor rhizosphere soil, and bulk soil, respectively, for network construction.

According to Banerjee et al. (2018), network hubs, module hubs, and connectors were defined as keystone species in the present study (Shi et al., 2020b). In addition, according to Guimerà and Nunes Amaral (2005), the z-scores (within-module degree) and c-scores (participation coefficient) of each node in the networks were calculated to identify the hubs and connectors. Based on the threshold values of the within-module degree (z-score) and participation coefficients (c-score) of nodes, nodes with a z-score > 2.5 and c-score > 0.6 were defined as network hubs. Nodes with a z-score > 2.5 and c-score < 0.6 were defined as module hubs, whereas nodes with a z-score < 2.5 and c-score > 0.6 were considered as connectors. Furthermore, nodes with a z-score < 2.5 and c-score < 0.6 were classified as peripherals. Shi et al. (2020b) has expounded in detail the particular role of each type of node in community networks.

Using the high throughput sequencing platform, 21,884,403 quality sequences were obtained from the 54 soil samples of four types; among them, 86,366 were identified at 100% similarity, being mostly bacteria and 11,188 were identified at 100% similarity, being mostly fungi. At the bacterial phylum level, Proteobacteria (28.8%), Actinobacteria (16.7%), Cyanobacteria (13.1%), Bacteroidetes (12.7%), Chloroflexi (9.7%), and Acidobacteria (5.8%) were dominant, accounting for more than 80% of all sequences (Figure 1) Ascomycota was the most common fungal phylum among the samples (Figure 1). At the fungal class level, Sordariomycetes (21.9%), Dothideomycetes (13.0%), Agaricomycetes (10.8%), Pezizomycetes (7.5%), and Leotiomycete (6.1%) were dominant. Dothideomycetes abundance was the highest in the herb rhizosphere (Figure 1).

In the case of bacteria, compare to those in all the rhizosphere soils, most high abundance microbiomes were significantly enriched in bulk soil, whereas Cyanobacteria and Bacteroidetes abundances in bulk soil were significantly lower (Figure 1; Supplementary Table S4). The abundances of phylum Cyanobacteria in the herbs’ rhizosphere soil were also significantly lower than that in arbors’ rhizosphere soil, but herbs had the most Bacteroidetes than the others (Figure 1; Supplementary Table S4). And in the case of fungi, the abundance of classes Leotiomycetes in arbor rhizosphere was significantly higher than those in the other soils. The abundances of Sordariomycetes, Pezizomycetes and Mortierellomycetes in bulk soils were the highest while the abundances of class Dothideomycetes and Leotiomycetes in bulk soils were the lowest among the different groups (Figure 1; Supplementary Table S5).

Although there were no obvious differences in fungal α diversity among the groups, there were significant differences in bacterial α diversity. The bacterial α diversity of herbs and bulk soils were significantly higher than those of shrubs and arbors. And there is no obvious difference in bacterial diversity between herbs and bulk soils, nor between shrubs and arbors (Supplementary Figures S2, S3).

According to the results of VPA, the explanation of plant type (1.1% for bacteria and 3.4% for fungi) for the differences between groups was higher than that of location (0.5% for bacteria and 1.7% for fungi; Supplementary Table S7). In addition, both the NMDS and Multivariate Welch ANOVA showed obvious differences between bacteria and fungus (Figure 2; Supplementary Tables S8, S9). According to the results of Multivariate Welch ANOVA, except for some comparisons (S. Herbage vs. N. Herbage, N. Bulk soil vs. N. Shrub, S. Bulk soil vs. N. Shrub, S. Bulk soil vs. N. Arbor, S. Bulk soil vs. N. Bulk soil, S. Bulk soil vs. S. Arbor and S. Arbor vs. S. Shrub), the bacteria in most comparisons had significant differences between groups (Figure 2A; Supplementary Table S8). While the majority of the differences of fungi among groups were significant, except for the differences between the bulk soil, the rhizosphere soil of shrubs and arbors in the south and between the bulk soil in the south and in the north (Figure 2B; Supplementary Table S9). Furthermore, Mantel test results showed that plant phylogeny had a strong influence on bacterial community structure, and not on fungal community structure, while plant vegetation types have a significant impact on the difference between groups of both bacteria and fungi (Supplementary Table S6).

Among bacteria, Cyanobacteria and Firmicutes were significantly higher in the rhizosphere soil of arbor. Bacteroidetes were significantly higher in herbaceous plants, while bulk soil had the largest number of endemic bacterial phyla (Figure 3A). In particular, the bulk soil contains a variety of archaea that are less abundant in the rhizosphere of plants.

For fungus (Figure 3B), Archaeorhizomycetes and Mortierellomycetes were significantly higher in soil than in rhizosphere soil of all plants. Leotiomycetes, Glomeromycota, Tremellomycetes, Rozellomycota and Mucoromycota were significantly higher in arbor soil. The Rhizophydiomycetes, Basidiomycota and Glomeromycota in the rhizosphere soil of shrub were significantly higher. Rhizophlyctidomycetes, Ascomycota and Orbiliomycetes were significantly higher in the rhizosphere soil of herbage.

According to the Venn plot (Figures 3C, D), although the number of bacteria involved in the analysis was much higher than the number of fungi, the number of common fungi in the four samples was still much higher than the number of common bacteria. In addition, the shrub rhizosphere soil had a higher number of common bacteria with arbor rhizosphere soil than with herbage rhizosphere soil, with an opposite trend observed in the case of fungi. In general, the shrub rhizosphere soil had the fewest specialized microbes.

Bacterial community assembly in rhizosphere was dominated by stochasticity processes, and dispersal limitation was more prevalent in the rhizosphere than in the bulk soil, especially in the herb rhizosphere soil (Figure 4A). Both arbor and shrub rhizosphere soils were partially dominated by stochasticity processes. And in bulk soil, bacterial community assembly was more dominated by deterministic process (Figure 4B). In addition, deterministic processes dominated fungal community assembly processes across all samples. Among them, the assembly processes of shrub rhizosphere were the closest to those of the bulk soil, with relatively similar Normalized Stochasticity Ratio values (Figure 4B).

Individual networks were constructed for each of the four sample types (Figure 5A; Table 1). Compared with rhizosphere soil, bulk soil had a more complex microbial network, with the most positive and negative connections. The rhizosphere soils of different vegetation types had different core flora. There were 7, 5, 6 and 6 keystone species in the herb, shrub, and arbor rhizosphere soils, and bulk soils, respectively (Table 2). Among them, shrub had the fewest core bacteria. The node and edge numbers in the networks decreased in the order of bulk soil, and herbage, shrub, and arbor rhizosphere soils. Similar trends were observed in the number of positive correlations and negative correlations, and average connectivity (avgK). However, with regard to the ratio of positive to negative correlations and average clustering coefficient (avgCC), shrub rhizosphere had the largest values, excluding bulk soil. In addition, shrub rhizosphere has the lowest average path distance (GD), the lowest number of modules, and the highest Graph density. In addition, the network modularity increased in the order of herbs, shrub, and arbor rhizosphere soils, and bulk soil.

Network robustness was also examined. The results indicated a more stable microbial network in the bulk soil than in the rhizosphere soils. Nodes and edges were discarded in declining order of node betweenness. Therefore, we observed that the natural connectivity of networks in all vegetation types exhibited sharp slopes in all the samples excluding in the bulk soil, suggesting poor stability (Figure 5B). Furthermore, for each group, z-scores and c-scores were calculated for the nodes in the network to identify the keystone species.

Additionally, the shrub has the highest positive cohesion and the lowest negative cohesion (Figure 5C), indicating more cooperation and less competition between microbes in the shrub rhizosphere than in those of other vegetation types.

Arbor, shrub, and herbs are the most common vegetation assemblages in city parks or campuses; therefore, city parks and campuses are ideal sites for investigating the impact of aboveground vegetation type on belowground soil microbial community structure, while eliminating the effect of climate and soil type. Arbors refer to trees with an upright trunk, usually 6–10 m high, with a trunk independent from the root, and a clear distinction between the trunk and the crown. Arbors also have strong vitality and are widely distributed. At present, arbors are basically found in all terrestrial biomes, including desert, Arctic, and other harsh environments (McBride and Douhovnikoff, 2012; Zhang et al., 2022). Conversely, shrubs are short plants (usually <6 m) with no obvious trunk and numerous branches near the ground, most of which are clustered. Shrubs are generally broad-leaved plants, and some coniferous plants, such as juniper, are shrubs (Lenard, 2008). Shrubs are widely distributed globally, mostly in the tropics and subtropics, and can also be found in arid regions (Xu et al., 2020). In China, shrubs are mainly spread in Zhejiang, Jiangsu, Anhui, Henan, etc., covering about 20% of the land area (Piao et al., 2009). Although there are many differences between arbors and shrubs, they have many similarities with regard to growth habit (Jingui et al., 2023). They are both perennial plants and can survive more than 3 years. Moreover, arbors and shrubs have large numbers of lignified cells, which are obviously different from herbs (Crivellaro et al., 2022). Herbs are usually short with stems that are soft and that break easily. In addition, many herbs are annuals, biennials or triennials, and their xylems are not developed and the vascular bundles do not have cambiums, so that they cannot grow continuously (Evans and Ortega, 2019). Herbs are generally adapted to warm and humid environments. However, herbaceous plants are very resilient, and can be found in hot and humid areas, as well as cold and dry areas. On the whole, the most obvious differences among trees, shrubs, and herbs are based on their physiological traits, biomass, and life span (Yuan et al., 2020).

In the present study, Pinus massoniana, Gingko biloba, Solanum nigrum, Ligustrum lucidum, Forsythia viridissima, Veronica persica, Punica granatum, Cercis chinensis, Bischofia polycarpa, Oxalis corniculata, Eriobotrya japonica, Euonymus japonicus, Lonicera maackii and Ophiopogon japonicus are grouped into arbors, shrubs, and herbs.

Due to differences in individual size, life history, or physiological function, bacteria and fungus exhibit distinct responses to aboveground plants. In the present study, we observed that soil bacterial α diversity was more sensitive to vegetation type than fungal α diversity. Bacteria are minute, propagate rapidly, and form spores, so that they are ubiquitous and easily dispersed (Foissner, 2006). Conversely, fungi proliferate mainly through budding and spore reproduction. Fungus can also form mycelia, with diverse functions (Cairney and Burke, 1996; Hodge, 2000). Fungal communities jointly form complex belowground networks, which drive the establishment of plant populations and communities, as well as soil nutrient turnover (Yang et al., 2022). According to Sheldrake (2020), fungi are regenerators, recyclers, and network builders that connect the world. Therefore, one or similar fungal species can be observed in different vegetation rhizosphere soils, so that we did not investigate the significant effect of vegetation type on fungal α diversity.

In the β diversity analysis, the results of VPA showed that the explanation of vegetation type for the difference is greater than that of location, which is partially different from the findings of Vieira et al. (2020). In addition, Yang et al. (2019) found that the influence of plant factors on rhizosphere microbial community was greater than spatial factors in the forests of eastern China. In our study, the campus garden is subject to periodic management and frequent human activities, and this will reduce the difference of soil between the two gardens. On the other hand, with the strong effect of host plants, plants exhibited stronger effect than the locations. Additionally, given the low percentage of the variance between samples that explain both the type of plant and the sampling location, possibly, other unaccounted factors can help explain the differences between the samples. According to the results of Multivariate Welch ANOVA, we found that the differences between rhizosphere soils of different vegetation types were generally significant. Consistent with Fitzpatrick et al.’s (2018) study, which found the plant species showed strong effect on the rhizospheric communities. This is also partly similar to the findings in the study of the low Arctic tundra (Shi et al., 2015) which found the soil microbial community could be differed by vegetation types. This may be due to the fact that shrubs and herbs are often closely interlaced, and these interactions would reduce the difference between groups.

In the present study, bulk soil and rhizosphere soil bacterial communities showed contrasting assembly processes, even under different vegetation types. Bacterial community assembly in bulk soil was more dominated by deterministic process, whereas the rhizosphere bacterial community assembly was dominated by stochasticity and the construction of fungal communities was all dominated by deterministic processes (Figure 4). Similar phenomena were observed in a wheat field ecosystem, with deterministic factors playing a greater role in the assembly of nitrogen fixing bacteria communities in the bulk soil than in the rhizosphere soil (Fan et al., 2018). While Yang et al. (2018) reported a contrary phenomenon that deterministic processes played a more important role than stochastic processes in bacterial community assembly processes in Chinese grassland ecosystem. This may be because environmental filtration has a greater impact on the biogeographic pattern of bacteria. The Anna Karenina principle could explain why we observed that rhizosphere bacterial community assembly was dominated by stochastic process. According to the principle, healthy hosts have relatively stable microbial communities, which form close clusters in an orderly space, while various external stress factors undermine such stability, leading to more dispersed microbial communities (Zaneveld et al., 2017). Therefore, the rhizosphere filtering effect would lead to the cultivation of specific species and be accompanied by the Anna Karenina principal effect, so that a random process occurs.

Due to the strong filtering effect of plant roots, they harbor simpler communities than bulk soil, and in turn, less complex association networks. Consistent with our study, in farmland, Fan et al. (2018) observed that the network structure of nitrogen fixing microbial communities in rhizosphere soil was less competitive and more stable than that in bulk soil. In grasses, researchers found that rhizosphere networks had less nodes and edges, lower density, but had higher modularity, and greater positive links than bulk soil networks (Li et al., 2021). In a forest ecosystem, co-occurrence network analysis detected relatively higher network complexity and node connectivity in bulk soil than in the rhizosphere community (Jing et al., 2022; Wang et al., 2022). These studies indicated that less complex association networks were prevalent in rhizosphere. In addition, Ling et al. (2022) analyzed 557 pairs of published 16S rDNA amplification sequences from non-rhizosphere soil and rhizosphere soil of different ecosystems globally, and found that the rhizosphere had relatively reduced microbial diversity due to the selection of corresponding microbial populations from soil seed banks, thus forming a highly modular but unstable bacterial network in the rhizosphere. This indicate that less complex networks are not related to the community stability, it might be depending on the ecosystem type.

Yang et al. (2019) observed soil fungal communities could be strongly influenced by plant phylogenetic distance in forest ecosystems across Eastern China. In the present study, we also observed that rhizosphere bacterial communities were significantly correlated with plant phylogeny. In addition, in root microbiomes of multiple plant phyla (Yeoh et al., 2017), researchers observed that soil bacterial communities could be strongly affected by plant phylogeny at the small scale. The reason could be that, at very small scales, plants exert very strong effects on bacteria due to their lower interrelationships, and fungi, because of their hyphae, build highly connected networks at the local scale, so that plant phylogeny did not exhibit strong effects with regard to fungal community structure. Beside this, the effect of specific sampling quantity (Hermans et al., 2019) and soil depth (Chu et al., 2016) on the rhizosphere community were also reported, these two potential impacting factors will be tested in the future.