An extensive literature search was conducted using the Web of Science database (http://apps.webofknowledge.com/). The data was collected from peer-reviewed publications from 2014 to 2021 for our literature survey and review. We search for terms including “rhizosphere”, “bulk”, “fungi”, “bacteria”, “microbial community”, “DNA extraction”, “PCR Amplification”, “16s”, “ITS”, “high-throughput”, “pyrosequencing”, and “Illumina”, etc. in the title, keyword, or abstract. We obtained a total of 861 data points based on 101 publications around the world ( Figure 1 ). Detailed information was given in Table S1 . The data from manipulation experiments conducted in laboratory settings were excluded. We exclusively employed 16S and ITS data, as detailed in Table S1 , for the purpose of this study. We examined the microbial community structure of bacteria and fungi by analyzing their relative abundance at the phylum level. The phylum level is often employed in rhizosphere and bulk soil research, and findings for the majority of microbial community analyses are commonly based on high-throughput sequencing technique.

The means (M), standard deviations (SD), and sample sizes (n) were obtained from both rhizosphere and bulk soils in each study. If only the standard error (SE) was provided, SD was calculated as SE multiplied by the square root of sample size. Missing standard deviations were estimated using the average coefficient of variation of datasets with known SDs (Chen et al., 2019). Data in tables were directly transferred to our dataset, while data in figures were extracted using Data Thief software, which is specifically designed for retrieving axis-related information from images (B. Tummers, DataThief III. 2006 https://datathief.org/). To ensure the accuracy and quality of the extracted data, only taxa with an average relative abundance > 2% were included. Taxa that were absent in any treatment or replicate of a study were excluded from the data extraction process (Dai et al., 2018).

The 16S and ITS data collected in this study were utilized for meta-analysis. The impacts of plant roots on soil microbial compositions were evaluated using the rhizosphere soil against the pairwise bulk soil (Chen et al., 2022). The natural log-transformed response ratio was previously described (Gurevitch et al., 2018):

Where ln R denotes the natural log of response ratio and is defined as the effect size, X_t , and X_c are the mean value of the rhizosphere and bulk soil, respectively, which are directly extracted from data of publications included in this meta-analysis. The ln R was further weighted by the pooled variance(v):

where SD_t and SD_c are the SD of rhizosphere and bulk soil, respectively, n_t and n_c are the sample size of the rhizosphere and bulk soil, respectively, which are directly extracted from data of publications included in this meta-analysis. The results obtained from both meta-analysis approaches were identical. We only reported the grand means and bias-corrected 95% CIs of effect sizes calculated by MetaWin because the 95% CIs from this approach enable us to assess the potential publication bias.

The total heterogeneity (Q_T) of the dataset (i.e. microbial structure and soil pH) was divided into within-group (Q_W) and between-group (Q_b) variations. The Q statistic has k-1 degrees of freedom and follows a Chi-square distribution, where k is the number of matched observations between rhizosphere and bulk soils. When the bias-corrected 95% CIs did not overlap zero, the effect size was considered substantially positive or negative at the = 0.05 level (Gurevitch et al., 2018). The lnR (effect size) was transformed and revealed as the percentage variation under rhizosphere relative to bulk soil:

where R denotes the response ratio. In our meta-analysis, the soil pH of rhizosphere was divided into five groups as follows: 5 (pH< 5), 6 (5 ≤ pH<6), 7 (6 ≤ pH<7), 8 (7 ≤ pH<8), and 9 (8 ≤ pH<9). The groups of soil pH were used to examine if there were any significant differences in effect size among the groups. The plant species were divided into maize, rice, vegetables, other crops, herbaceous, and woody plants. The means of effect size between the groups of soil pH or plant species were considered to be significant at α = 0.05 when bias-corrected 95% CIs were non-overlapping.

Based on the relative abundance of major fungal and bacterial phylum, we calculated Bray-Curtis dissimilarity for principal coordinate analysis (PCoA). PCoA was performed to visualize the microbial community structure of rhizosphere and bulk soil. Bray-Curtis similarity was calculated to construct triangular pairwise Bray-Curtis similarity matrix using PRIMER 7 (Plymouth Routines in Multivariate Ecological Research Statistical Software, v7.0.13, PRIMER-E Ltd, UK). ANOSIM and ADONIS tests were used to test the cluster significance of samples from the Bray-Curtis similarity distances, which were calculated based on the relative abundance of the major microbial groups. Taxonomic classification was performed using the Silva 16S rRNA and the UNITE fungal ITS reference database in all publications which we obtained. Most statistical analyses were performed in R (v4.0.1; http://www.r-project.org/).

Across all studies, the means of effect size of soil pH was -2%, indicating a significantly (P<0.05) negative effect in rhizosphere than bulk soils on soil pH ( Figure 2 ). In particular, when soil pH > 7, effect sizes of soil pH were significantly (P<0.05) lower than zero, showing that rhizosphere effects on soil pH were most in neutral or slightly alkaline soils. However, with a decline of soil pH, the effect sizes of rhizosphere also increased and became positive (+3%) under soil pH = 5, suggesting that bulk soil pH condition was the main factor for the rhizosphere effects on soil pH. Similarly, the effect size of rhizosphere on soil pH was also dependent on plant species, in which the effect sizes in rice (-1%) and vegetables (-4%) were significantly (P<0.05) lower than zero, while herbaceous and woody has non-significant effect sizes and maize even had a positive effects size (+2%).

Although the means of effect size of bacterial richness were not significantly different from zero, the effect size of bacterial richness in vegetables was significantly (P<0.05) negative (-29%) ( Figure 3 ). However, across all plant species, the means of effect size of bacterial Shannon index were significantly (P<0.05) negative (-4%). Specifically, the effect sizes of bacterial Shannon index in plant types of maize, rice, vegetables, herbaceous and other crops were significantly (P<0.05) lower than zero (-9% to -2%) while those in wood plants were non-significant.

Across all studies, the bacterial communities mainly consist of Proteobacteria (39%), Actinobacteria (14%), Acidobacteria (11%), Bacteroidetes (8%), Firmicutes (7%), Chloroflexi (6%), Verrucomicrobia (4%), Gemmatimonadete (4%), Planctomycetes (3%), and Nitrospirae (1%), with the relative abundance in the descending order ( Figure 4A ). Based on the composition of major phyla found in our study, PCoA of the Bray-Curtis distances was conducted to reveal that the beta-diversity of bacterial community was significantly separated between into rhizosphere and bulk soils (PERMANOVA, P< 0.0001) ( Figure 4B ). Specifically, the bacterial community structure of rhizosphere soils was also dependent on plant species (PERMANOVA, P< 0.0001) ( Figure 4C ). For example, the bacterial community structure in vegetable rhizosphere was significantly separated from those in maize, rice, and woody plants (PERMANOVA, P< 0.0001).

In total, there were two major phyla showing significantly (P<0.05) positive effect size of plant rhizosphere, including Bacteroidetes (+27%) and Proteobacteria (+24%) ( Figure 5 ). When the dataset was divided by plant species, the relative abundance of Bacteroidetes was significantly (P<0.05) enriched in the rhizosphere of maize (+34%), vegetables (+47%), woody plants (+38%), and other crops (+91%). The phylum of Proteobacteria was generally enhanced in the rhizosphere of all plant species except rice, such as maize (+20%), vegetables (+25%), herbaceous (+22%), and woody plants (+35%).

However, four bacterial phyla with average relative abundance over 5% had an overall negative effect size of plant rhizosphere across all plant species, although the reduction percentages varied, such as Acidobacteria (-28%), Actinobacteria (-7%), Firmicutes (-14%), and Chloroflexi (-12%) ( Figure 6 ). Specifically, in Acidobacteria, though all plant species had negative effect size, the negative effects of rhizosphere in vegetables was the lowest (-48%), significantly (P<0.05) lower than those in maize (-18%) and herbaceous plants (-22%). In Actinobacteria, all plant species, with the exception of maize (-4%) and vegetables (+32%), exhibited negative effect sizes ranging from -18% to -10%. In Firmicutes, both herbaceous (-27%) and woody plants (-33%) showed negative responses to rhizosphere effects, but those in maize (28%) and rice (52%) were contrarily enriched in the rhizosphere. In Chloroflexi, apart from herbaceous plants showing a positive effect size (+28%), the plant rhizosphere effects were consistently negative across all other plant types (ranging from -42% to -16%).

With an average relative abundance of less than 5%, three bacterial phyla showed negative effect size in general, including Gemmatimonadete (-29%), Nitrospirae (-42%), and Planctomycetes (-15%) ( Figure 7 ). The phylum Gemmatimonadete was significantly (P<0.05) depleted in the rhizosphere of all plant types except herbaceous plants and other crops, such as maize (-31%), rice (-51%), vegetables (-42%), and woody plants (-28%). Furthermore, the phylum Nitrospirae was significantly (P<0.05) depleted in the rhizosphere of all plant species (from -77% to -19%). Similarly, in Planctomycetes, the effect sizes of rhizosphere effects were all negative across all plant types (from -37% to -17%) except herbaceous plants. Last but not least, although Verrucomicrobia showed no significant response to the rhizosphere effect on average, maize had a significantly (P<0.05) positive effect size (+39%), while herbaceous plants had a negative effect size in the rhizosphere (-9%).

In fungal communities, the effect size of rhizosphere was less significant compared to bacteria. Neither Ascomycota nor Basidiomycota showed a significant (P<0.05) response to the rhizosphere effect ( Figure 8 ). However, the phylum Ascomycota enriched in the rhizosphere of maize (+30%) and vegetables (+32%), while the phylum Basidiomycota depleted in the rhizosphere of maize (-52%) and vegetables (-60%), and enriched in herbaceous plants (+22%).

Our meta-analysis revealed that both the effect sizes of bulk soil pH > 6 and the mean effect sizes of soil pH were significantly negative ( Figure 2 ). These findings imply that most plants tend to decrease the rhizosphere soil pH compared with the bulk soil across all studies. A previous study has reported that the release of H⁺ by roots is a dominant mechanism for plants to mobilize nutrients and maintain electrochemical potential on the root surface in slightly acidic, neutral, and alkaline soils (Marschner, 2012). Under the extreme soil pH condition (too low or too high), plant roots can mitigate the constraints, such as plant roots that could alleviate Al³⁺ or Fe³⁺ toxicity in acidic pH conditions, and also Fe or Mn deficiency in alkaline pH conditions (Philippot et al., 2013; Kuzyakov and Razavi, 2019). Thus, our results that the effect sizes of rhizosphere were significantly lower than zero in soil pH > 6 further suggested that plant roots could alleviate constraints under neutral or slightly alkaline conditions. Moreover, our result showed the effect sizes of rice and vegetables were significantly lower than zero, which implies that the roots of rice and vegetables tend to acidify the rhizosphere soil more severely than other plants. This is in line with the fact that rice root exudates organic acid decreasing rhizosphere soil pH, increasing amino acid availability, and promoting ammonia release and subsequent nitrification (Dohrmann et al., 2013; Di Salvo et al., 2018). Furthermore, legumes acidify the rhizosphere soil through excess cation uptakes during N₂ fixation and photosynthetic activity to alter cation-anion uptake ratios (Bolan et al., 1991; Rao et al., 2000; Rao et al., 2002). Similarly, it has been found that the rhizosphere soil pH of pak choi decreased by root exudates, which mainly consist of organic acids and amino acids (e.g., citric acid, ferulic acid, cinnamic acid, glutamic acid, alanine, and valine) (Kim et al., 2017; Jeon et al., 2018; Cai et al., 2019). As previous studies report most plant species tend to acidify the rhizosphere soil (Kuzyakov and Razavi, 2019), our results further provide substantial evidences to support that these pH variation may significantly contribute to the assembly of soil microbial community in plant rhizophere.

In our analysis, PCoA of the Bray-Curtis distances showed significant clustering of bacterial communities between rhizosphere and bulk soils, with decreased community diversity (Shannon index) observed in the rhizosphere soil across different plant species ( Figures 3 , 4B ). This is consistent with previous findings that diversity decreased from the bulk soil to the roots (Poole, 2017; Semenov et al., 2019). The significant contrast between the bulk soil and rhizosphere soil is a crucial factor contributing to variations in microbiota composition (Yan et al., 2017; Ren et al., 2020). For example, Fan et al. (2017) reported that the distance decay relationship (from the root surface to bulk soil) can reflect variations in microbial community composition. In addition, it has been suggested that the decrease in diversity from the bulk soil to roots could be attributed to root “rhizosphere effect” (Bulgarelli et al., 2012; Lundberg et al., 2012; Barajas et al., 2020; Attia et al., 2022; Santoyo, 2022). Specific microorganisms are commonly selected by plant roots to colonize the rhizosphere, which can attract beneficial microorganisms to improve nutrient acquisition and combat pathogenic taxa for plants (Dennis et al., 2010; Berendsen et al., 2012; Fan et al., 2017). Generally, the rhizosphere is a highly selective environment that can select microbiome through two distinct processes (Fan et al., 2017; Fan et al., 2018). The first process involves the general recruitment of microbes to the proximity of the root, whereas the second process involves the transition of microbes from external to internal occupancy in the root (Edwards et al., 2015). Molecular signals from plants, including components of root exudates and possibly cell wall or membrane proteins (Edwards et al., 2015; Edwards et al., 2018; Kuzyakov and Razavi, 2019), are involved in the selection of microbial communities. Notably, the DNA extraction protocols for metagenomic DNA, particularly in relation to rhizosphere and soil samples, may introduce inherent biases due to variations in rhizosphere soil sampling procedures and differences arising from DNA extraction methods. Despite the presence of disparities, our study represents a comprehensive synthesis of over 100 studies, thus mitigating any potential systematic biases.

Additionally, the results of the present meta-analysis showed that rhizosphere soil microbial community structure varies depending on plant species ( Figure 4C ). This is in agreement with previous studies that the plant species and the genotypes of individual plants can exert a profound influence on the composition of their associated microbial communities in the rhizosphere (Schweitzer et al., 2008; Lau and Lennon, 2012; Bever et al., 2013). However, our meta-analysis showed the phyla Proteobacteria and Bacteroidetes are consistently enriched in the rhizosphere ( Figure 5 ). This result suggests that the phyla Proteobacteria and Bacteroidetes are well-suited to the rhizosphere, which provides C-rich conditions for high metabolic activity, fast growth, and propagation (Pausch et al., 2013; Kuzyakov and Razavi, 2019). The phyla Proteobacteria and Bacteroidetes are generally considered r-strategists and copiotrophs that respond to labile C sources, and fast-growing microbiota with population opportunity fluctuations (Fierer et al., 2007; Peiffer et al., 2013). In contrast, our meta-analysis showed that the phyla Acidobacteria, Gemmatimonadete, Nitrospirae, and Planctomycetes are consistently depleted in the rhizosphere ( Figures 6 , 7 ). It has been previously reported that the phyla Acidobacteria were depleted in wheat rhizosphere under the field of North China Plain (Fan et al., 2017). Moreover, a previous study reported that the phyla Acidobacteria were enriched in the bulk soil, while depleted in the rhizosphere soil under the Central European grasslands and forests (Kaiser et al., 2016). Similar results have been revealed in Mexico’s agroecosystem, German grassland, and forest soil as well (Foesel et al., 2014; Embarcadero-Jimenez et al., 2016; Dawson et al., 2017). As a result, these phyla are extremely similar in rhizosphere soil across varied plant species (Ling et al., 2022). Similarly, Planctomycetes are more abundant in the rhizosphere than in bulk soil (Fierer et al., 2007). The phyla Gemmatimonadete, Nitrospirae, and Planctomycetes have been extensively detected as K-strategists (slow-growing microbiota) and oligotrophic that are adapted to survive when resource was limited or low substrate concentrations (Fontaine et al., 2003; Bernard et al., 2007; Blackburne et al., 2007), and generally considered to be enriched in the bulk soil with less energy and nutrients compared with the rhizosphere soil. Overall, our meta-analysis implied that the phyla Proteobacteria and Bacteroidetes enriched commonly in the rhizosphere of most plant species, while Acidobacteria, Gemmatimonadete, Nitrospirae, and Planctomycetes were depleted contrarily by plant rhizosphere effects. Notably, agricultural practices might make up the plant-associated microbial communities. Tillage can lead to shifts in microbial communities as anaerobic microorganisms are exposed to oxygen (Coleman-Derr et al., 2016; Dawson et al., 2017). Reduced or no-till farming can preserve anaerobic niches and maintain a different microbial community structure. Synthetic fertilizers often provide easily accessible nutrients, which can favor certain microbial taxa, while organic fertilizers release nutrients slowly, supporting a more diverse microbial community (Yang et al., 2017). Crop rotation and the diversity of crops planted in a field can provide different root exudates and organic matter, altering the nutrient availability and microbial community. This practice promotes a more diverse and dynamic microbial community (Visioli et al., 2018; Wang et al., 2020).

Moreover, the responses of the phyla Actinobacteria, Firmicutes, Chloroflexi, and Verrucomicrobia to plant rhizosphere were highly dependent on plant species ( Figures 6 , 7 ). The phyla Actinobacteria was significantly depleted in the rhizosphere of rice, herbaceous, woody, and other crops, while enriched in that of vegetables. For example, it has been found that the phylum Actinobacteria dominated the wild beet rhizosphere (Zachow et al., 2014) and lettuce rhizosphere (Blau et al., 2019). Additionally, the phylum Actinobacteria has been shown to dominate in the pak choi rhizosphere compared with the bulk soil with or without Se application (Cai et al., 2019). In addition, we showed that the phylum Chloroflexi significantly depleted in the rhizosphere of maize, rice, vegetables, and woody, while enriched in that of herbaceous plants ( Figure 6 ). This is in agreement with a previous study reporting that the relative abundance of Chloroflexi increased from 2.7% to 8.0% in the rhizosphere compared to the bulk soil across 19 herbaceous plants (Dawson et al., 2017). We found that the phyla Firmicutes significantly enriched in the rhizosphere of maize and rice, while depleted in those of herbaceous, woody plants, and other crops. This is in line with previous studies observing that the phylum Firmicutes significantly increased in the rhizosphere compared with the bulk soil in maize fields at different rice growth stages (de Araujo et al., 2019; Li et al., 2019). Similarly, the abundance of Verrucomicrobia revealed opposite trends between the rhizosphere of maize and herbaceous plants ( Figure 7 ), which was greatly observed in previous studies (Coleman-Derr et al., 2016; Dawson et al., 2017; Yang et al., 2017; Visioli et al., 2018; Wang et al., 2020).

Additionally, the phyla Ascomycota and Basidiomycota were the dominant fungal communities having distinct responses to rhizosphere depending on plant species ( Figure 8 ), indicating that plant species differentiate their root microbiota in a species-specific manner (Berg and Smalla, 2009; Aira et al., 2010; Dawson et al., 2017). Such as, Ascomycota had a higher relative abundance in the rhizosphere of maize. In contrast, bulk soils had a higher abundance of Basidiomycota (Philippot et al., 2013). Commonly, plant species determine the structure of the rhizosphere soil microbial community as follows: firstly, the soil layer surrounding the roots promotes the growth of organotrophic microorganisms and initiates a shift in the soil microbiome through rhizodeposits and root cell wall features (Bulgarelli et al., 2013); secondly, the selection process that depends on the host genotype occurs close to the root surface (Reinhold-Hurek et al., 2015), fine-tuning the community profiles that thrive on the rhizoplane. Overall, the phyla Proteobacteria and Bacteroidetes are considered r-strategists that enriched in rhizosphere across all plant species, while Acidobacteria, Gemmatimonadete, Nitrospirae, and Planctomycetes are considered as K-strategists that depleted in rhizosphere across all plant species. Especially, Actinobacteria, Firmicutes, Chloroflexi, and Verrucomicrobia were selected in a species-specific manner from various plant species, thus revealing divergent abundance among different plants.