Deterministic assembly of rhizosphere bacterial community enables predictable and stable suppression of Verticillium wilt via microbial homeostasis Authors: Jiamin Ai, Jiangying Huang, Chenyuan Wang, Xinrui Wang, Yiduo Zhao, Gehong Wei, Zhefei Li Published: 2026 ABSTRACT BackgroundThe rhizosphere microbiota plays essential roles in maintaining normal physiological function and preventing disease in plants. However, the mechanisms by which soil-borne pathogens break through the microbial defense line in the rhizosphere to infect the host are not well understood.ResultsIn this study, we observed that the relative abundance of Verticillium dahliae does not significantly differ between healthy and diseased rhizospheres, but the root-associated microbial community structure changes substantially. The invasion of pathogens shifts the assembly process of the microbial community from deterministic to stochastic, making it difficult to predict the disease resistance function of the rhizosphere microbial community. An increase in stochasticity promotes the rapid proliferation of opportunistic bacteria, which deplete carbon sources, disrupt the balance of the original microbial community, and reduce network stability. Consequently, the depletion of beneficial microbes leads to rhizosphere dysbiosis, while fungal pathogenic allies gain a competitive advantage, thereby facilitating the progression of Verticillium wilt. These results are further validated by constructing synthetic microbial communities (SynComs) using strains enriched in different ecological processes. Healthy plants maintain their health by recruiting beneficial microorganisms through deterministic processes. Moreover, the disease-suppressive efficacy of SynComs governed by deterministic processes was less affected by inoculation timing or concentration, making their functional outcomes more predictable. In contrast, the disease resistance levels of SynComs governed by stochastic processes are more sensitive to inoculation timing and concentration, making their function more unpredictable.ConclusionsOverall, our findings emphasize the potential of SynComs designed according to community assembly principles to combat soil-borne diseases, with deterministically assembled SynComs providing more durable disease resistance than stochastically assembled ones. Video AbstractSupplementary InformationThe online version contains supplementary material available at 10.1186/s40168-026-02432-7. Introduction Terrestrial plants host diverse and intricate microbial communities within soil regions closely associated with their roots. Niche-based theory suggests that deterministic processes predominantly govern community structure patterns [1]. A deterministic process refers to any ecological mechanism driven by nonrandom, niche-based factors, such as environmental filtering and various biological interactions (e.g., competition, facilitation, mutualism, and predation) [2, 3]. For example, plants can recruit specific beneficial microorganisms through root exudates to increase nutrient acquisition [4], improve tolerance to abiotic stresses [5], and prevent diseases by either inducing stronger immune responses in plants or directly combating pathogens [6]. While the beneficial impact of the plant microbiota is well-documented, some natural members of the resident microbiota can be harmful. These microbes can promote disease establishment and development by altering their relative abundances or forming partnerships with other invasive pathogens [7, 8]. Emerging evidence has highlighted the crucial role of the pathobiome in plant pathogenesis, suggesting that disease onset and progression are driven not only by individual pathogens but also by coordinated pathobiome interactions across multiple scales [9]. The concept of the pathobiome originates from studies of human diseases, where the disruption of the intestinal microflora that supports host health has caused ecological imbalance. This imbalance increases stochastic processes during microbial community assembly, allowing numerous potential pathogenic and opportunistic bacteria to invade, thereby increasing the risk of pathogen invasion and proliferation in the intestine [10, 11]. While most studies on microbial community dysbiosis—defined as changes in the composition and functional balance of the microbiome—have focused on human health, plants can also experience dysbiosis, which can impair their physiological functions and increase their susceptibility to pathogen invasion [12]. Numerous studies have shown that the rhizosphere microbial community structure significantly changes after pathogen infection. These changes are believed to be driven by the rearrangement of microbial communities induced by root exudates [13]. However, to date, few scholars have thoroughly explored whether these alterations are caused by plant disease development linked to the pathobiome. The plant holobiont can be conceptualized as a never-ending war between key beneficial microbes, such as the symbiome, and the allies of pathogenic microbes, such as the pathobiome [14]. Plants enhance their resistance to external pressures through the deterministic selection of beneficial microbes. When the new microbe configuration can no longer support resistance to disturbances, it may be replaced by the pathobiome, leading to changes in host health [15]. Compared with healthy microbiota, disease-associated microbiota exhibits greater stochasticity in its assembly. These stochastic variations increase the range of possibilities for the microbial compositions of diseased individuals [16]. Consequently, scientists have proposed the Anna Karenina principle (AKP) on the basis of Leo Tolstoy’s assertion and applied it to plant microbiomes: “All healthy microbiomes are similar; each disease-related microbiome is sick in its own way.” Arnault et al. [17] demonstrated that stochastic processes in the assembly of microbiota elicited by environmental stressors can lead to diseases through the AKP. True pathobiome implies a role for the altered microbiota in promoting disease development [18]. The use of microbiota transplantation methods and synthetic microbial communities (SynComs) has made it possible to test for a causal relationship between altered microbiota and symptoms [19]. In plants, the AKP effect may stem from alterations in pathogen-associated molecular pattern-triggered immunity (PTI) and effector-triggered immunity (ETI) responses, resulting in reduced immune control over microbiota composition; this reduction facilitates the random colonization of opportunistic bacteria and potential pathogens [20, 21]. The AKP has been described in animal [16], human [22], and coral microbiota [23]. However, how the rhizosphere microbiota affects plant health when driven by stochastic processes remains unclear. Astragalus membranaceus Bge. var. mongholicus (Bge.) Hsiao (hereafter, Astragalus) is a perennial herbaceous legume, and a traditional Chinese medicine widely used in clinical treatments. A major threat to the yield and quality of Astragalus is the high incidence of Verticillium wilt. Understanding the stability, complexity, and assembly mechanisms of rhizosphere microbial communities in both healthy and diseased Astragalus plants is crucial for understanding their role in the progression of Verticillium wilt. In this study, we employed a recently developed framework for phylogenetic bin-based null model analysis (iCAMP) to analyze community assembly mechanisms at a relatively fine taxonomic level and to quantify the relative contributions of stochastic versus deterministic processes to plant disease resistance [24]. The following hypotheses were proposed. (1) Compared with healthy microbial communities, the invasion of pathogens leads to a greater proportion of stochastic processes in microbial community assembly. This increased stochasticity promotes the invasion of pathogens by reducing the stability of the network. (2) The increase in stochasticity results in the significant enrichment of opportunistic bacteria in the rhizosphere, which then interact with pathogens to form a pathobiome that promotes disease progression. (3) Healthy plants maintain rhizosphere health by recruiting beneficial microbes through deterministic processes, and their functions are relatively more predictable. Therefore, we constructed synthetic microbial communities (SynComs) that represented different ecological processes in healthy and diseased samples and examined their effects on the incidence of Verticillium wilt through pot experiments. The results confirmed that deterministic processes favored the maintenance of root health, and the altered microbiota caused by pathogen invasion played a promoting role in inducing symptoms. Therefore, constructing SynComs on the basis of community assembly mechanisms was more predictable and could effectively control soil-borne diseases. Materials and methods Sample collection and processing Samples were collected from an Astragalus plantation in Tanchang County, Gansu Province, China (altitude 2307 m, 34° 28′ N, 104° 16′ E), during a severe outbreak of Verticillium wilt (Fig. 1). Plants were scored as “healthy” or “diseased” by integrated visual inspection of shoots and roots. Diseased plants exhibited classic wilt—lower-leaf chlorosis, drooping, and progressive necrosis—coupled with vascular browning visible in longitudinal taproot sections (Fig. S1a). Thirty plants (15 healthy, 15 diseased) were selected for the next experiments. The roots were gently shaken to remove excess soil before being transferred into tubes containing 1 × phosphate-buffered saline (PBS). The tubes were then vortexed for 5 min to dislodge soil particles from the roots. After the roots were removed, the tubes were centrifuged at 8000 rpm for 7 min to separate the soil from the solution. The supernatant was discarded, and the rhizosphere soil was collected for further analysis. The roots were then surface sterilized by immersing them in 75% ethanol for 1 min, followed by treatment with 1% sodium hypochlorite for 30 s, and then rinsing with sterile water. A total of 20 samples (2 compartments × 5 repetitions × 2 health states) were collected from the rhizosphere soil and roots. These samples were divided into two groups: the first group was subjected to microbial isolation, while the other group was stored at − 80 °C for subsequent DNA extraction.Fig. 1Assembly of synthetic communities (SynComs) and their effects on Verticillium wilt. Step 1: Collection of rhizosphere soil and root samples from healthy and diseased Astragalus plants. Step 2: Analysis of the microbial community composition of the rhizosphere soil and roots. Isolation and identification of plant-associated bacteria. Based on the enrichment patterns of different ASVs (healthy-associated vs. disease-associated) and community assembly mechanisms (deterministic vs. stochastic processes), the construction of the SynComs was guided. Step 3: Effects of SynComs inoculation timing on the incidence of Verticillium wilt. Step 4: Effects of SynComs inoculation concentration on the incidence of Verticillium wilt. Step 5: SynComs‒pathogen interaction analysis Assembly of synthetic communities (SynComs) and their effects on Verticillium wilt. Step 1: Collection of rhizosphere soil and root samples from healthy and diseased Astragalus plants. Step 2: Analysis of the microbial community composition of the rhizosphere soil and roots. Isolation and identification of plant-associated bacteria. Based on the enrichment patterns of different ASVs (healthy-associated vs. disease-associated) and community assembly mechanisms (deterministic vs. stochastic processes), the construction of the SynComs was guided. Step 3: Effects of SynComs inoculation timing on the incidence of Verticillium wilt. Step 4: Effects of SynComs inoculation concentration on the incidence of Verticillium wilt. Step 5: SynComs‒pathogen interaction analysis DNA extraction, amplification, and sequencing Total DNA was extracted from 0.5 g of soil using the FastDNA®SPIN Kit (MP Biomedicals, Solon, USA) following the manufacturer’s instructions. Root samples were pulverized in liquid nitrogen, and DNA was extracted using the MiniBEST Plant Genomic DNA Extraction Kit (Takara, Japan), followed by 16S rRNA and ITS amplicon sequencing at Novogene Technology Co., Ltd. (Beijing, China). The bacterial 16S rRNA gene was amplified using the primer set 515 F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 907R (5′-CCGTCAATTCCTTTGAGTTT-3′), which target the V4–V5 regions of the bacterial 16S rRNA gene. Then, ITS1F (CTTGGTCATTTAGAGGAAGTAA) and ITS1R (GCTGCGTTCTT CATCGATGC) were used to amplify the fungal ITS1 regions [25]. Raw sequencing data were processed through a series of steps: assembly and quality-filtering according to Nearing et al. [26], followed by removal of chimeric sequences using the UCHIME tool in USEARCH [27]. Sequence alignment for bacteria and fungi was performed using the Silva Database and Unite Database, respectively. All reads assigned to plant plastids (chloroplasts and mitochondria) were discarded from the dataset, and the remaining effective sequences were clustered into amplicon sequence variants (ASVs) at a 99% similarity. The bacterial 16S rRNA and fungal ITS sequencing data for all the samples in this study were submitted to the NCBI Sequence Read Archive database under BioProject accession numbers PRJNA1302709 and PRJNA1302716. Pathogenicity assessment of pathogenic fungi and isolation of plant-associated bacteria Pathogenic ASVs, which were significantly enriched in the diseased roots of Astragalus, were identified through ITS sequencing. The ASV sequence was compared with that of strains preserved in the laboratory (isolated from Astragalus). The strain showing the greatest similarity was selected to verify its pathogenicity under monocropping and sterile soil. Astragalus seeds were disinfected with 75% ethanol for 30 s, followed by immersion in a 1% sodium hypochlorite solution for 1 min. The seeds were then rinsed six times with distilled water and soaked for 48 h. Afterwards, the samples were placed on sterile, wet filter paper for germination. Seven germinated seedlings were planted in bags containing 500 g of either monocropping or sterile soil. After 7 days, the seedlings were thinned to 5 uniform plants per bag. Each plant was inoculated with 2 mL of a V. dahliae VD1 spore suspension sample (~106 spores per milliliter), and the disease index was calculated after 21 days. Bacteria were isolated from the rhizosphere soil of Astragalus using the method outlined by Oppenheimer-Shaanan [28], with detailed isolation procedures being provided in the Supplementary Materials. To confirm the presence of these bacterial isolates in the rhizosphere soil, similarity analysis was performed between the 16S rRNA sequences of the bacterial isolates and the ASV sequences (from the same genus). The V4–V5 hypervariable regions were extracted from the full-length Sanger sequences and aligned with ASV sequences using SnapGene software (v6.0). Isolates with V4–V5 sequences matching ASVs with > 97% similarity were considered representative of those ASVs [29]. Assembly of synthetic communities and their effects on Verticillium wilt The selection of strains for SynComs construction was primarily focused on rhizosphere-enriched ASVs rather than root-endophytic ones. This strategy was first informed by our community assembly analysis, which revealed that the transition from healthy to diseased states significantly altered the community assembly processes specifically in the rhizosphere. Furthermore, the rhizosphere is widely recognized as the primary defensive barrier against soil-borne pathogens and serves as the ecological reservoir for potential endophytes [30, 31]. Lastly, prioritizing rhizosphere-adapted species is more compatible with current agricultural practices (such as soil drenching or seed coating). Differential abundance analysis of bacterial ASVs was performed using the edgeR package in the R environment (version 4.4.2). To account for the compositional nature of microbiome data and variations in sequencing depth, raw counts were normalized using the Trimmed Mean of M-values (TMM) method. We employed a Generalized Linear Model (GLM) framework based on a Negative Binomial (NB) distribution to model the taxon counts, which effectively captures the inherent over-dispersion and high sparsity of the dataset. Specifically, robust dispersion estimation (estimateGLMRobustDisp) was applied to down-weight the influence of potential outliers and ensure stable statistical inference. Significantly enriched or depleted ASVs between the healthy and diseased groups were identified using the likelihood ratio test. ASVs were defined as significantly differential based on the criteria of an absolute log-fold change ∣logFC∣ > 1 and a False Discovery Rate (FDR) < 0.05. The R package “iCAMP” was used to analyze the community assembly mechanisms [24]. To construct the SynComs, isolated strains were sequence-matched against the differential ASVs, with a similarity threshold of > 97% used to identify candidate strains. The community assembly mechanisms of these isolated differential ASVs were further characterized using the iCAMP framework. Based on the enrichment patterns of the isolated differential ASVs and their corresponding ecological processes, four distinct SynComs were defined: (1) HD SynComs: ASVs enriched in the healthy rhizosphere and driven by deterministic processes; (2) HS SynComs: ASVs enriched in the healthy rhizosphere and driven by stochastic processes; (3) DD SynComs: ASVs enriched in the diseased rhizosphere and driven by deterministic processes; (4) DS SynComs: ASVs enriched in the diseased rhizosphere and driven by stochastic processes. Finally, three bacterial synthetic communities were constructed: HD SynComs (9 bacterial species), DD SynComs (6 bacterial species), DS SynComs (8 bacterial species), and a single strain of Paenibacillus vietnamensis XR-120 (HS) (Table S1). The bacterial strains (OD600 = 0.8) from each group were mixed in equal volumes to establish SynComs. The final application amount in the soil is approximately equivalent to ~2 × 10⁶ CFU per gram of soil. To investigate the impact of SynComs inoculation timing on the incidence of Verticillium wilt disease, a series of pot experiments were conducted. Early inoculation involved soaking seeds in SynComs, with no further inoculation at later stages (Fig. 1, Step 3). In contrast, late inoculation involved the application of SynComs to 7-day-old Astragalus seedlings. Both the bacterial solution immersion group and the control group were inoculated with 2 mL of sterile water on day seven. One week later, all the treatment groups were inoculated with 2 mL of V. dahliae spore suspension (1 × 10⁶ spores/mL). The plants were grown in a greenhouse at 25 °C under a 16-h light/8-h dark cycle. Disease index was recorded 21 days after pathogen inoculation. For detailed methods, see the Supplementary Materials. To identify which bacteria in SynComs were retained and functional during interactions between plants, V. dahliae, and the bacterial communities, rhizosphere and root samples were collected from the pot experiments and analyzed using full-length 16S rRNA sequencing. Near-full-length 16S rRNA amplicon sequencing and read assembly were performed on the PacBio platform at Novogene Technology Co., Ltd. (Beijing, China). Effects of different concentrations of synthetic communities on Verticillium wilt The bacteria were cultured in LB liquid medium to an optical density at 600 nm (OD600) of 0.8 (~1 × 10⁸ CFU·mL⁻1). After centrifugation and resuspension, the cell density was adjusted to ~2 × 10⁸ CFU·mL⁻1. Two inoculation concentrations were prepared for all SynComs: low concentration (~1.28 × 105 CFU per gram of soil): 20 μL of SynComs (~2 × 10⁸ CFU·mL⁻1) + 1980 μL of sterile water, and high concentration (~4 × 10⁶ CFU per gram of soil): 2000 μL of SynComs (~2 × 10⁸ CFU·mL⁻1). Five germinated seedlings were planted in planting bags containing 500 g of monocropping soil and inoculated with different SynCom concentrations (2 mL per plant) (Fig. 1, Step 4). After 7 days of growth, the plants were thinned to 5 uniform plants per bag. Each plant was inoculated with 2 mL of V. dahliae VD1 spore suspension (~106 spores per milliliter), and the disease index was recorded 21 days later. To assess the effects of different inoculation concentrations on the colonization of SynComs, full-length 16S rRNA sequencing was performed on rhizosphere soil and root samples collected during the pot experiments using the PacBio platform. To further investigate the effects of inoculation time and concentration on the resistance of HD SynComs + P. vietnamensis XR-120 (HS) in combination with Verticillium wilt, five treatment protocols were implemented in a pot experiment: (1) Control (Verticillium dahliae VD1): 2-week-old Astragalus seedlings were inoculated with V. dahliae VD1 spore suspension; (2) HS: 1-week-old Astragalus seedlings were inoculated with P. vietnamensis XR-120, followed by inoculation with V. dahliae VD1 spore suspension at 2 weeks of age; (3) HD_HS: 1-week-old Astragalus seedlings were inoculated with HD SynComs + P. vietnamensis XR-120 (v:v = 1:1), followed by inoculation with V. dahliae VD1 spore suspension at 2 weeks of age; (4) HD: 1-week-old Astragalus seedlings were inoculated with HD SynComs, followed by inoculation with V. dahliae VD1 spore suspension at 2 weeks of age; and (5) S_HS_HD: seeds were soaked in P. vietnamensis XR-120, after which 1-week-old Astragalus seedlings were inoculated with HD SynComs, after which they were inoculated with V. dahliae VD1 spore suspension at 2 weeks of age. The disease index was calculated 21 days after V. dahliae VD1 inoculation. Statistical analysis All the statistical analyses were conducted in the R environment (version 4.4.2; http://www.r-project.org/). The 'vegan' package was used to calculate α-diversity. For β-diversity analysis, principal coordinate analysis (PCoA) based on Bray‒Curtis distances was performed to describe the microbial community structures under different treatments. Beta-diversity was assessed based on Bray–Curtis distances calculated using the R package vegan. The significance was calculated using permutational multivariate analysis of variance (PERMANOVA), implemented via the adonis2 function in the R vegan package. The PERMANOVA was based on the Bray–Curtis distance matrix. We performed 999 random permutations to determine the p value. To further investigate the underlying mechanisms of microbial community dissimilarity, β-diversity was decomposed into two additive components: species replacement (Repl) and nestedness-resultant dissimilarity (Nest), using the beta.div.comp function in the adespatial package [32]. The R package “betapart” was employed to partition β-diversity into replacement and nestedness components. Network analysis was performed using robust Spearman’s correlation coefficients (|ρ|> 0.7) with a false discovery rate (FDR)-corrected p < 0.05 to identify significant correlations between microbial taxa [33]. Cytoscape 3.4.0 was used to visualize the network. Nodes were classified into four categories: module hubs (Zi > 2.5), network hubs (Zi > 2.5 and Pi > 0.62), connectors (Zi > 2.5 and Pi < 0.62), and peripherals (Zi < 2.5 and Pi < 0.62) [34]. Functional predictions of bacterial and fungal communities were performed using PICRUSt [35] and FUNGuild [36], respectively. Phylogenetic bin-based null model analysis (iCAMP) was used to infer community assembly mechanisms [24]. Phylogenetic trees were constructed using the neighbor-joining (NJ) method in MEGA v5 and annotated and visualized with iTOL software [37]. The ecological niche overlap index was calculated using the “spaa” package. Vulnerability was defined as the relative contribution of each node to global efficiency. The network’s vulnerability was represented by the maximum vulnerability among the nodes. The robustness index of the network was calculated by measuring the proportion of remaining nodes after the removal of 50% of the classification group [38]. The network stability of the microbial network was calculated by positive and negative cohesion [39]. The microbial life cycle strategies (r and K-strategists) were characterized based on the average 16S rRNA operon copy number of the microbial communities. The 16S rRNA operon (rrn) copy number for each ASV in the bacterial community was estimated using the rrnDB database on the basis of its closest relatives with known rrn copy numbers [40]. For each sample, the abundance-weighted average rrn copy number was calculated by first multiplying the estimated operon copy number by the relative abundance of each ASV and then summing this value across all ASVs in the sample [41]. Results Pathogen infection changed the structure of root-associated microbial communities Analysis of 16S and ITS amplicon sequencing revealed that the Shannon α-diversity was significantly lower in the diseased rhizosphere (DRS) than that in the healthy rhizosphere (HRS) (Fig. 2a; p < 0.05). However, no significant difference in bacterial diversity was observed between the healthy roots (HR) and diseased roots (DR), suggesting that the impact of Verticillium wilt on bacterial microbial diversity is more pronounced in the rhizosphere than in the root. PCoA based on Bray‒Curtis distances demonstrated that pathogen invasion induced significant shifts in the rhizosphere microbial community composition (Fig. 2b). β-Diversity decomposition analyses revealed that the compositional dissimilarities (replacement and nestedness) in the bacterial and fungal communities among all the samples were predominantly driven by species replacement processes (Fig. S1b). Verticillium dahliae infection induced significant (p < 0.05) alterations in the relative abundances of 8 bacterial phyla and 1 fungal phylum in the rhizosphere (Fig. S1c). Similarly, the roots experienced changes in the relative abundances of 3 bacterial and 1 fungal phylum. At the genus level, the most abundant bacterial genera in the healthy rhizosphere were Microbacterium (11.59%), Bacillus (10.22%), and Pseudomonas (6.36%) (Fig. 2c). Within the roots, Microbacterium (24.67%) and Pseudomonas (6.80%) remained dominant. In contrast, the most abundant bacterial genera in the diseased rhizosphere were Bacillus (16.14%), Pseudomonas (15.96%), and Lelliottia (11.01%), whereas the predominant genera within the roots were Microbacterium (14.45%) and Microvirga (7.73%). Additionally, pathogen infection led to the depletion of 176 and 50 bacterial ASVs, and the enrichment of 68 and 81 bacterial ASVs in the rhizosphere and roots, respectively (Fig. S1d). ITS amplicon sequencing revealed that Verticillium (23.20%) was significantly enriched in the diseased roots, but its relative abundance was not significantly different between the healthy and diseased rhizospheres (Fig. 2d). Among Verticillium, ASV_5 (17.57%) had the highest relative abundance, which was 98.13% similar to that of Verticillium dahliae VD1, which isolated from the root of diseased Astragalus (Fig. S2). The pathogenicity of V. dahliae VD1 was further verified. V. dahliae VD1 inoculation resulted in a greater disease index of Verticillium wilt in Astragalus in both natural and sterilized soil (Fig. 2e). Therefore, this fungus was used as an indicator in subsequent experiments. These findings suggest that the onset of Verticillium wilt is not solely determined by the proliferation of individual pathogens within the rhizosphere, but the disturbance of the rhizosphere microbiome is also an important driver.Fig. 2Root-associated microbial community composition. a α-Diversity indices of root-associated bacteria and fungi. Lines represent the mean and standard deviation. Statistical significance was determined using unpaired two-tailed t tests. *p < 0.05; **p < 0.01 and ***p < 0.001. b ASV-level PCoA of bacterial and fungal communities based on Bray–Curtis distances. The significance was calculated using permutational multivariate analysis of variance (PERMANOVA). c Comparison of bacterial community composition at the genus level. d Comparison of fungal community composition at the genus level. e Pathogenicity assessment of pathogenic fungi. DR, diseased root; HR, healthy root; DRS, diseased rhizosphere; HRS, healthy rhizosphere Root-associated microbial community composition. a α-Diversity indices of root-associated bacteria and fungi. Lines represent the mean and standard deviation. Statistical significance was determined using unpaired two-tailed t tests. *p < 0.05; **p < 0.01 and ***p < 0.001. b ASV-level PCoA of bacterial and fungal communities based on Bray–Curtis distances. The significance was calculated using permutational multivariate analysis of variance (PERMANOVA). c Comparison of bacterial community composition at the genus level. d Comparison of fungal community composition at the genus level. e Pathogenicity assessment of pathogenic fungi. DR, diseased root; HR, healthy root; DRS, diseased rhizosphere; HRS, healthy rhizosphere Assembly mechanism of the root-associated microbial community Analysis using the iCAMP model revealed that bacterial community assembly in healthy rhizosphere soil and roots was dominated by deterministic processes (Fig. S3). In contrast, the diseased rhizosphere bacterial and fungal communities were primarily shaped by stochastic processes. A total of 8043 bacterial ASVs observed in the rhizosphere were categorized into 132 bins (Fig. 3). In healthy rhizosphere communities, deterministic processes (homogeneous selection) dominated 50 bins, accounting for 76.83% of the relative abundance, whereas stochastic processes dominated 82 bins (23.17% of the relative abundance; dispersed restriction for 73 bins and drift for 9 bins). Conversely, bacterial community assembly in the diseased rhizosphere was mainly a stochastic process that dominated 106 bins (65.07% of the relative abundance; dispersal limitation for 33 bins, homogenizing dispersal for 14 bins, and drift for 59 bins). Deterministic processes dominated 26 bins (34.93% of the relative abundance; homogeneous selection accounted for 25 bins, and heterogeneous selection accounted for 1 bin). In healthy roots, deterministic processes (homogeneous selection) dominated 35 bins, accounting for 29.67% of the total bins and 72.08% of the relative abundance, with Proteobacteria being the predominant phylum (Fig. S4). Similarly, diseased roots exhibited a comparable pattern, with homogeneous selection dominating 31 bins, accounting for 27.97% of bins and 63.31% of the abundance, with Proteobacteria being the predominant phylum. The fungal communities were categorized into 62 rhizosphere and 43 root bins (Figs. S5 and S6). Notably, all the ASVs of Verticillium in the root samples were in bin 4, which was governed by dispersal limitation (Fig. S6). Within this bin, Verticillium accounted for 98.88% of the total abundance, with ASV_5 being the most abundant ASV. These findings indicated that V. dahliae infection increased stochasticity in microbial community assembly.Fig. 3Variations in ecological processes across different phylogenetic groups in healthy and diseased rhizospheres. a The phylogenetic tree, constructed on the basis of representative ASVs from each bin, is displayed at the center. The rings from the inside to the outside represent the main ecological processes of each bin in the healthy and diseased rhizospheres (b), the relative abundance of the main bacterial genera in each bin (c), and the relative abundance of each bin (d). The circular packing on the periphery shows the species composition of the bin where the SynComs’ ASVs are located Variations in ecological processes across different phylogenetic groups in healthy and diseased rhizospheres. a The phylogenetic tree, constructed on the basis of representative ASVs from each bin, is displayed at the center. The rings from the inside to the outside represent the main ecological processes of each bin in the healthy and diseased rhizospheres (b), the relative abundance of the main bacterial genera in each bin (c), and the relative abundance of each bin (d). The circular packing on the periphery shows the species composition of the bin where the SynComs’ ASVs are located Stochastic processes weakened the network stability and function of root-associated microbial communities The rhizosphere bacterial co-occurrence network consisted of 767 nodes and 7139 edges. Most network links (> 90%) were associated with Proteobacteria (50.85%), Actinobacteria (27.51%), Firmicutes (7.04%), and Bacteroidetes (5.57%) (Fig. 4a). Ecological process partitioning resulted in greater differentiation than phylum-level topological distributions did. ASVs governed by deterministic processes, stochastic processes, and ecological process transitions (the ecological process shifts between deterministic and stochastic processes) accounted for 26.86%, 36.38%, and 36.77%, respectively, of network topology. We separately extracted the networks for healthy and diseased rhizospheres from the complete bacterial rhizospheres network. The results revealed that compared with the diseased rhizosphere (DRS) bacterial network, the healthy rhizosphere (HRS) bacterial network was more complex and stable (Fig. 4b and Fig. S7a–c). Furthermore, deterministic ASVs demonstrated stronger network centrality, dominating the medium (4000–8000) and high (> 8000) betweenness centrality tiers (Fig. 4c–d). We compared the vulnerability of the HRS and DRS bacterial networks after removing either deterministic or stochastic ASVs. The results revealed that the role of deterministic ASVs in maintaining bacterial network stability was significantly greater in the HRS bacterial network than in the DRS bacterial network (Fig. 4e).Fig. 4Rhizosphere bacterial co-occurrence network and network properties of ASVs associated with different ecological processes. a The network is based on the entire rhizosphere bacterial community, with colors representing the distribution of phyla and ecological processes. b Co-occurrence network of healthy and diseased rhizosphere soil. c Rhizosphere network ZiPi analysis, with stacked plots on the right indicating the proportion of different ecological processes in the connectors. d Distribution of the betweenness centrality of different ecological process ASVs. e Vulnerability of the rhizosphere network before and after the removal of different ecological process ASVs to measure the importance of the different ecological process ASVs in contributing to the structure of the rhizosphere network. f The root network is based on the entire bacterial community, with colors representing the distribution of phyla and ecological processes. g Co-occurrence network of healthy and diseased roots. h Root network ZiPi analysis, with stacked plots on the right indicating the proportion of different ecological processes in the connectors. i Distribution of the betweenness centrality of different ecological process ASVs. j Vulnerability of the root network before and after the removal of different ecological process ASVs to measure the importance of the different ecological process ASVs in contributing to the root network structure. DR, diseased root; HR, healthy root; DRS, diseased rhizosphere; HRS, healthy rhizosphere Rhizosphere bacterial co-occurrence network and network properties of ASVs associated with different ecological processes. a The network is based on the entire rhizosphere bacterial community, with colors representing the distribution of phyla and ecological processes. b Co-occurrence network of healthy and diseased rhizosphere soil. c Rhizosphere network ZiPi analysis, with stacked plots on the right indicating the proportion of different ecological processes in the connectors. d Distribution of the betweenness centrality of different ecological process ASVs. e Vulnerability of the rhizosphere network before and after the removal of different ecological process ASVs to measure the importance of the different ecological process ASVs in contributing to the structure of the rhizosphere network. f The root network is based on the entire bacterial community, with colors representing the distribution of phyla and ecological processes. g Co-occurrence network of healthy and diseased roots. h Root network ZiPi analysis, with stacked plots on the right indicating the proportion of different ecological processes in the connectors. i Distribution of the betweenness centrality of different ecological process ASVs. j Vulnerability of the root network before and after the removal of different ecological process ASVs to measure the importance of the different ecological process ASVs in contributing to the root network structure. DR, diseased root; HR, healthy root; DRS, diseased rhizosphere; HRS, healthy rhizosphere Proteobacteria (62.31%) constituted most of the links in the root bacterial network (Fig. 4f). Of these links, ASVs governed by deterministic processes accounted for 46.23% of the root bacterial network topology. Compared with its diseased counterpart, the healthy root (HR) bacterial network exhibited greater complexity and stability (Fig. 4g and Fig. S7a–c). Additionally, ASVs governed by deterministic processes (44.29%) primarily served as connectors in the root bacterial network (Fig. 4h). In contrast to those in the bacterial rhizosphere network, ASVs governed by deterministic processes were enriched in the lower betweenness centrality tiers (< 2100) (Fig. 4i). Unlike bacterial rhizosphere networks, both deterministic and stochastic process ASVs were equally important in maintaining the stability of the HR bacterial network (Fig. 4j). Ascomycota (70.17%) represented most of the links in the fungal network (Fig. S8a). ASVs associated with stochastic processes contributed 73.01% and 62.80% to the topologies of the rhizosphere and root fungal networks, respectively (Fig. S8a). The networks of the healthy samples were more complex and stable than those of the diseased samples (Fig. S8b and Fig. S7d–f). ASVs driven by stochastic processes functioned as predominant connectors in both the rhizosphere and root fungal networks, accounting for 67.50% and 56.82% of network connectors, respectively, and were widely distributed across multiple betweenness centrality tiers (Fig. S8c–d). Targeted removal of ASVs driven by stochastic processes from healthy fungal networks significantly increased vulnerability (Fig. S8e), highlighting their critical role as structural determinants of fungal network stability. Additionally, pathogen invasion reduced positive cohesion but increased negative cohesion in the rhizosphere bacterial communities, suggesting that pathogen invasion disrupted synergistic relationships among microbial taxa (network nodes) in the bacterial rhizosphere soil (Fig. S7b). This disruption likely led to increased competition or inhibitory interactions. In root-associated bacterial communities, both positive and negative cohesion increased. In contrast, the rhizosphere fungal communities showed the opposite trend (Fig. S7b). Robustness and network stability analyses revealed a decrease in rhizosphere bacterial network stability following pathogen invasion (Fig. S7c). Therefore, cooperative interactions played an important role in maintaining the robust structure of healthy rhizosphere communities. Construction of a synthetic bacterial community and its effect on Verticillium wilt The use of SynComs would enable testing of ecological processes in maintaining plant health. Therefore, 204 bacterial strains taxonomically spanning 76 genera and 4 phyla were isolated from the Astragalus rhizosphere to construct SynComs (Fig. S9). Ten ASVs representing eight genera were cultured from 176 ASVs enriched in healthy rhizosphere communities, whereas 14 ASVs representing ten genera were enriched from 68 ASVs enriched in diseased rhizosphere communities (Fig. S10). On the basis of differential ASV enrichment samples and their associated ecological processes (deterministic or stochastic processes), three SynComs (healthy deterministic, HD SynComs; disease deterministic, DD SynComs; and disease stochastic. DS SynComs), and a single-strain treatment (Paenibacillus vietnamensis XR-120, denoted healthy stochastic, HS) was used for further studies (Fig. S10). Pot experiments revealed that the timing of SynComs inoculation significantly influenced the disease index of Verticillium wilt in Astragalus (Fig. 5a). HD SynComs consistently suppressed disease progression, regardless of the timing of inoculation. Early HS inoculation markedly reduced disease severity. Conversely, DD SynComs had no discernible effect on the disease index; thus, we excluded this treatment group from the subsequent analysis. Early inoculation of DS SynComs substantially increased the disease index. These findings suggested that the invasion of pathogens could drive stochastic assembly of rhizosphere bacteria, promoting the enrichment of specific ASVs, and then form pathobiome to promote the progression of the disease.Fig. 5Role of SynComs in the suppression of Verticillium wilt in Astragalus. a Influence of inoculation time on Verticillium wilt disease index. The preinoculation of SynComs is denoted as −1, and the late inoculation of SynComs is denoted as −2. Lowercase letters above the bars indicate significant differences (p < 0.05) between the different treatments (one-way ANOVA). b Effect of inoculation time on the colonization of SynComs. c Impact of SynComs inoculum concentration on disease resistance. Low inoculum concentration: 20 μL of SynComs (~2 × 10⁸ CFU·mL⁻1) + 1980 μL of sterile water, and high inoculum concentration: 2000 μL of SynComs (~2 × 10⁸ CFU·mL⁻1). d Impact of inoculum concentration on the colonization of SynComs. Colonization status was validated by mapping full-length 16S rRNA gene sequences (ASVs) from samples back to the reference sequences of individual SynCom members with 99% identity. The numbers indicate the count of unique strains whose 16S rRNA sequences (ASVs) were recovered from the samples with > 99% identity to the initial inoculum. e β-Diversity analysis of the HD and DS SynComs preinoculation treatment groups. Statistical significance was determined using unpaired two-tailed t tests. *p < 0.05; **p < 0.01 and ***p < 0.001. f β-Diversity analysis of different SynComs inoculation concentration treatment groups. g Bata dispersion analysis of different samples. h Bata dispersion analysis of different SynComs inoculation concentration treatment groups Role of SynComs in the suppression of Verticillium wilt in Astragalus. a Influence of inoculation time on Verticillium wilt disease index. The preinoculation of SynComs is denoted as −1, and the late inoculation of SynComs is denoted as −2. Lowercase letters above the bars indicate significant differences (p < 0.05) between the different treatments (one-way ANOVA). b Effect of inoculation time on the colonization of SynComs. c Impact of SynComs inoculum concentration on disease resistance. Low inoculum concentration: 20 μL of SynComs (~2 × 10⁸ CFU·mL⁻1) + 1980 μL of sterile water, and high inoculum concentration: 2000 μL of SynComs (~2 × 10⁸ CFU·mL⁻1). d Impact of inoculum concentration on the colonization of SynComs. Colonization status was validated by mapping full-length 16S rRNA gene sequences (ASVs) from samples back to the reference sequences of individual SynCom members with 99% identity. The numbers indicate the count of unique strains whose 16S rRNA sequences (ASVs) were recovered from the samples with > 99% identity to the initial inoculum. e β-Diversity analysis of the HD and DS SynComs preinoculation treatment groups. Statistical significance was determined using unpaired two-tailed t tests. *p < 0.05; **p < 0.01 and ***p < 0.001. f β-Diversity analysis of different SynComs inoculation concentration treatment groups. g Bata dispersion analysis of different samples. h Bata dispersion analysis of different SynComs inoculation concentration treatment groups Full-length 16S sequencing of rhizosphere and root samples from the pot experiments revealed that inoculation of different SynComs significantly altered the bacterial community structure in both compartments (Fig. S11a). In the HD SynComs treatment group, 8 strains were successfully colonized in the rhizosphere and root, whereas one strain failed to establish (Fig. 5b and Fig. S12). In the HS treatment group, the colonization of XR-120 in both the rhizosphere and root was significantly influenced by the timing of inoculation, with early inoculation notably improving its establishment. In the DS SynComs treatment group, 6 strains were successfully colonized in the rhizosphere and root, while 2 strains failed to colonize. Our results demonstrate that community assembly mechanisms significantly influence SynCom colonization and biocontrol efficacy (Fig. 5a). Strains in HD SynComs had a higher colonization rate at all inoculation timings, ensuring consistent disease suppression. Conversely, the establishment of stochastic-process strains (e.g., HS and DS SynComs) were highly inoculation time-dependent, relying on priority effects (Fig. S12). Early inoculation (seed soaking) facilitated niche pre-emption and enhanced stochastic-process strains colonization success. However, late inoculation (post-germination) resulted in competitive exclusion by indigenous soil microbiota, leading to poor establishment of stochastic-process strains and the loss of disease-modulating phenotypes. Resistance to Verticillium wilt caused by different concentrations of SynComs Pot experiments revealed that the inoculum concentration of SynComs significantly influenced the disease index of Verticillium wilt in Astragalus (Fig. 5c). Across all SynComs, HD SynComs consistently suppressed disease progression, regardless of the inoculation concentration. At low inoculum concentrations, HS and DS SynComs did not affect the disease index. Conversely, at high inoculum concentrations, HS significantly reduced the disease index, whereas DS SynComs notably increased it. Furthermore, full-length 16S rRNA sequencing of rhizosphere and root samples from the pot experiments revealed that inoculation of different SynComs substantially altered the bacterial community structure in both compartments (Fig. S11b). In the HD SynComs treatment group, 8 strains could colonize in the rhizosphere and roots, whereas one strain failed to establish (Fig. 5d and Fig. S13a). Strain XR-120 was not detected in either the rhizosphere or the root. In the DS SynComs treatment group, only 3 strains colonized both compartments, while 5 strains failed to do so. Therefore, strains assembled through deterministic processes exhibited higher colonization rates in both the rhizosphere and the root compared with those assembled under stochastic processes. Furthermore, regardless of the inoculation concentration, the cumulative relative abundance of HD SynComs was consistently higher than that of DS SynComs (Fig. S13b). Further investigation revealed that the inoculation of different combinations of HD SynComs and HS strains significantly affected resistance to Verticillium wilt (Fig. S14). All combinations of the HD SynCom and HS strains effectively decreased the disease index, with the S_HS_HD treatment group exhibiting the lowest disease index. Notably, a decrease in the proportion of HD SynComs within a combination increased the disease index. These findings highlighted that bacterial strains with deterministic processes enriched in the healthy rhizosphere could play a pivotal role in disease resistance to Verticillium wilt. Microbiome variability associated with disease Pathogen invasion led to an increase in β-diversity. To further investigate this issue, we compared the β-diversity between healthy and diseased samples under different treatments. Our results revealed that the β-diversity of the DRS was greater than that of the HRS (Fig. S15a). Additionally, this difference was more pronounced in the HD and DS SynComs early inoculation treatment groups (Fig. 5e). Analysis of β-diversity under different inoculation concentrations revealed that increasing the HD SynComs inoculation concentrations did not significantly affect their β-diversity, whereas the β-diversity of the DS SynComs increases significantly with higher inoculation concentrations (Fig. 5f). Moreover, the distance to the centroid for DRS was greater than that for HRS (Fig. S15b), with this difference being greater in the HD and DS SynComs early inoculation treatment groups (Fig. 5g). Under different inoculation concentrations, the distance to the centroid was not significantly correlated with disease severity in the HD SynComs treatment group, whereas a significant positive correlation was observed in the DS SynComs treatment group (Fig. 5h). These findings suggested that healthy rhizospheres maintained similar microbiomes that formed tight clusters in ordination space. However, pathogen invasion elevates the stochasticity in community assembly and thereby causes greater dispersion in the rhizosphere microbial composition of diseased plants. Ecological niches competition between SynComs and pathogens Niche overlap analysis demonstrated that HD SynComs and the HS had relatively strong niche overlap with Verticillium within healthy rhizosphere communities, whereas DS SynComs exhibited reduced overlap in diseased communities (Fig. 6a). Carbon source utilization analysis revealed that HD and DS SynComs exhibited similar types of carbon source utilization (Fig. 6b) and complementary carbon source utilization between HD SynComs and HS, whereas HD SynComs and Verticillium exhibited nested utilization patterns (Fig. 6c). Among the 31 carbon sources tested, HD SynComs (utilizing 20 sources) and Verticillium (utilizing 11 sources) exhibited a high degree of overlap (9 common sources) (Fig. 6d). Additionally, DS SynCom resulted in fast carbon source utilization rates (Fig. 6e). Life history strategy analysis demonstrated that health status significantly influenced community assembly rules. Specifically, diseased samples showed a significant enrichment of r-strategists (Fig. 6f), which possess higher average rRNA operon copy numbers and thrive through rapid growth in disturbed niches. Conversely, the healthy rhizosphere was dominated by K-strategists; these low-copy-number taxa are indicative of a stable, efficient, and mature community structure capable of maintaining ecological balance under competitive conditions (Fig. 7). Furthermore, plate confrontation assays demonstrated that HD SynComs significantly inhibited the growth of Verticillium (Fig. S16).Fig. 6Resource competition between SynComs and Verticillium dahliae. a Niche overlaps between SynComs and V. dahliae. Niche overlap was quantified using Levins’s index based on the genus-level taxonomic profiles from rhizosphere soil samples. Values represent the mean niche overlap between the constituent genera of each SynCom and the pathogen. b Principal component analysis (PCA) of carbon source utilization. The PCA plot illustrates the metabolic distances between SynComs and V. dahliae based on carbon source utilization profiles obtained from (Biolog EcoPlates). c Utilization of specific carbon sources by SynComs and V. dahliae. d Number of carbon sources overlapping between different treatment groups. This Venn diagram depicts the shared and unique carbon sources utilized by SynComs and V. dahliae, identifying potential areas of metabolic competition. e Utilization rate of test carbon sources by SynComs and V. dahliae. Error bars represent standard errors of the mean (SEM). f Life cycle strategies of the microbial communities in different samples. Life cycle strategies were inferred from the weighted average 16S rRNA operon copy number of the bacterial communities, predicted using the rrnDB database. Higher copy numbers represent a shift toward r-strategists (copiotroph, rapid growth), while lower values indicate a shift toward K-strategists (oligotrophe, higher resource-use efficiency). One-way ANOVA followed by Duncan’s Multiple Range Test (MRT) was employed to evaluate the differences between treatments. Lowercase letters above the bars indicate significant differences (p < 0.05) between the different treatments. Vel, Verticillium dahliae; HD, HD SynComs; DS, DS SynComs; HS, P. turicensis XR-120Fig. 7Anna Karenina Principle of the rhizosphere microbiome. Healthy rhizospheres exhibit high alpha diversity, low beta diversity, and relatively similar and stable community structures, with community assembly being driven primarily by deterministic processes. However, pathogenic invasion acts as an environmental stressor that disrupts this stability, elevates stochasticity in rhizosphere community assembly, and consequently initiates plant disease. Diseased rhizospheres are characterized by low alpha diversity, high beta diversity, and relatively dissimilar and unstable community structures, with community assembly driven primarily by stochastic processes. These findings demonstrate that pathogen invasion enhances the role of stochastic processes in community assembly, and enriches some microbes to form a pathobiome, which ultimately facilitates disease progression Resource competition between SynComs and Verticillium dahliae. a Niche overlaps between SynComs and V. dahliae. Niche overlap was quantified using Levins’s index based on the genus-level taxonomic profiles from rhizosphere soil samples. Values represent the mean niche overlap between the constituent genera of each SynCom and the pathogen. b Principal component analysis (PCA) of carbon source utilization. The PCA plot illustrates the metabolic distances between SynComs and V. dahliae based on carbon source utilization profiles obtained from (Biolog EcoPlates). c Utilization of specific carbon sources by SynComs and V. dahliae. d Number of carbon sources overlapping between different treatment groups. This Venn diagram depicts the shared and unique carbon sources utilized by SynComs and V. dahliae, identifying potential areas of metabolic competition. e Utilization rate of test carbon sources by SynComs and V. dahliae. Error bars represent standard errors of the mean (SEM). f Life cycle strategies of the microbial communities in different samples. Life cycle strategies were inferred from the weighted average 16S rRNA operon copy number of the bacterial communities, predicted using the rrnDB database. Higher copy numbers represent a shift toward r-strategists (copiotroph, rapid growth), while lower values indicate a shift toward K-strategists (oligotrophe, higher resource-use efficiency). One-way ANOVA followed by Duncan’s Multiple Range Test (MRT) was employed to evaluate the differences between treatments. Lowercase letters above the bars indicate significant differences (p < 0.05) between the different treatments. Vel, Verticillium dahliae; HD, HD SynComs; DS, DS SynComs; HS, P. turicensis XR-120 Anna Karenina Principle of the rhizosphere microbiome. Healthy rhizospheres exhibit high alpha diversity, low beta diversity, and relatively similar and stable community structures, with community assembly being driven primarily by deterministic processes. However, pathogenic invasion acts as an environmental stressor that disrupts this stability, elevates stochasticity in rhizosphere community assembly, and consequently initiates plant disease. Diseased rhizospheres are characterized by low alpha diversity, high beta diversity, and relatively dissimilar and unstable community structures, with community assembly driven primarily by stochastic processes. These findings demonstrate that pathogen invasion enhances the role of stochastic processes in community assembly, and enriches some microbes to form a pathobiome, which ultimately facilitates disease progression Discussion Understanding the ecological dynamics and processes that govern the assembly of soil–plant microbiomes and the steady-state balance of microbial communities is crucial for effectively managing soil-borne diseases. The results show that the occurrence of Verticillium wilt is attributed not only to the proliferation of pathogens in the rhizosphere but also to the formation of a pathobiome. The invasion of pathogens leads to an increased proportion of stochastic processes in rhizosphere bacterial community assembly. Increased stochasticity reduces community stability, facilitating the proliferation of opportunistic bacteria, which form a pathobiome to accelerate Verticillium infection. SynComs constructed on the basis of community assembly processes confirm the promoting effect of the altered microbiome on disease progression. In contrast, healthy plants maintain their health by recruiting beneficial microorganisms through deterministic processes. Building on these findings, we emphasize the potential of SynComs, constructed on the basis of community assembly process guidance, as a strategy for resisting soil-borne diseases. Rhizosphere homeostasis aiding plant defenses against pathogen infection Dysbiosis, defined as the disruption of a host’s ability to regulate its microbiota, leads to a decline in host health due to functional loss [17]. The primary characteristic of dysbiosis is a reduction in microbial community α-diversity, which can be explained by drastic changes in niche parameters such that only a few pathogens and opportunistic microorganisms can survive [42]. Verticillium invasion decreases α-diversity in the Astragalus rhizosphere, increases community stochasticity, and creates new niches that are chaotically colonized by opportunistic microorganisms, subsequently increasing β-diversity. These opportunistic microorganisms (such as the DS SynComs in this study) rapidly proliferate in the rhizosphere, depleting available resources and forming a pathobiome with pathogens, thereby accelerating disease progression. These results suggest that forming a pathobiome is a significant driving factor that profoundly affects the distribution patterns of microbial communities. This pattern is observed in both healthy and diseased field samples and in pot experiments with HD and DS SynComs, where the application of SynComs has more pronounced effects. Furthermore, no significant difference is detected in the relative abundance of Verticillium between healthy and diseased rhizospheres, suggesting that the disease outbreak may be driven by differences in microbial composition (microbiota dysbiosis), not just changes in pathogen abundance. As a powerful tool for investigating the complexity and stability of microbial communities, microbial network analysis has significantly enhanced our understanding of soil-borne diseases [43]. The bacterial co-occurrence networks in healthy rhizosphere soil displayed lower vulnerability and greater stability (robustness index and network stability) and complexity (network topology index and cohesion index) than their infected counterparts. The higher positive cohesion in the HRS suggests stronger nutritional and cooperative interactions between rhizosphere bacterial microorganisms, including commensalism, mutualistic symbiosis, syntrophic cooperation and cross-feeding [38]. In contrast, the higher negative cohesion in the DRS indicates a competitive or inhibitory relationship between microorganisms, which is likely driven by competition for limited resources such as water, nutrients, and living space. The reduced ratio of the absolute value of negative cohesion to positive cohesion in the DRS suggests a decline in network stability [39]. Rhizosphere microbiota dysbiosis leads to the reorganization of biological interactions within the microbial community, resulting in network disintegration and network complexity reduction [44]. The observed decrease in network complexity may be linked to a decline in the deterministic ratio, potentially due to the introduction of more random processes in the microbial community following successful pathogen invasion. Thus, the complex and stable microbial network in the rhizosphere is relatively resilient to the invasion of soil-borne pathogens. Deterministic process of the bacterial community contributing to plant health Elucidating the mechanisms that regulate microbial community assembly (MCA) has long been a central challenge in microbial community ecology [45]. Two types of processes—deterministic and stochastic processes—affect MCA and represent complementary components of the continuous ecological forces that shape the community structure [46, 47]. In soil–plant systems, the ecological assembly process is influenced by a complex interplay of abiotic and biotic factors [48]. Among the biological stressors, plant pathogen invasion plays a critical role in shaping MCAs. In the healthy rhizosphere of Astragalus, bacterial community assembly is driven primarily by deterministic processes; however, pathogen invasion alters this dynamic, making stochastic processes (e.g., drift) the dominant force in shaping bacterial community assembly. Studies have shown that the recruitment of beneficial microorganisms by plants through deterministic processes plays a crucial role in buffering against environmental stress [49]. In this study, Arthrobacter, Streptomyces, Novosphingobium, and Aminobacter are enriched in the healthy rhizosphere, which has been shown to promote plant growth by enhancing disease resistance, synthesizing growth-stimulating plant hormones, and inhibiting pathogenic microorganisms [50, 51]. In the co-occurrence network analysis, deterministic ASVs exhibit relatively high functional centrality in the rhizosphere bacterial network, dominating the moderate and high betweenness centrality tiers. Furthermore, the removal of deterministic ASVs from the HRS significantly increases network vulnerability. Increased stochasticity in the diseased rhizosphere facilitates the colonization of potential pathogens and opportunistic microorganisms, thereby reducing rhizosphere stability and promoting disease development. Our findings highlight the critical role of deterministic processes in the ability of the rhizosphere microbiome to resist soil-borne pathogen invasion. Consequently, the deterministic microbial filtration of the host positively influences the fitness of the plant holobiont. Effects of SynComs dominated by different ecological processes on resistance to Verticillium wilt The resistance of plants to external stressors can be enhanced through the deterministic selection of beneficial microorganisms. Compared to SynComs constructed through stochastic processes, HD SynComs—constructed through deterministic processes—exhibit efficient disease resistance and colonization, which were less affected by inoculation time and concentration. Additionally, the HD SynComs demonstrates direct inhibitory effects against pathogenic fungi, with a notable overlap in carbon source utilization patterns with those of the pathogens. In contrast, DD SynComs do not significantly impact the disease index of Verticillium wilt, suggesting that the deterministic recruitment mechanisms in diseased rhizospheres may be more complex. To deterministically recruit beneficial microbes to combat pathogens, plants employ a “cry for help” strategy during pathogen invasion [52]. In this study, Bacillus velezensis XS-164 within the DD SynComs exhibits direct inhibitory effects against Verticillium dahliae and significantly reduces the disease index in pot experiments (Fig. S17). Previous research has shown that during pathogen invasion, pathogens can recruit their allies to accelerate disease progression [53]. Specifically, Pseudomonas aeruginosa, a natural inhabitant of soil and a potential plant pathogen [54, 55], is identified in the DD SynComs. In the pot experiment, P. aeruginosa XS-134–8 is coinoculated with Verticillium dahliae, which significantly increases the disease index of Verticillium wilt (Fig. S17). Additionally, Bacillus (16.14%) and Pseudomonas (15.96%) are the most abundant bacterial genera in the diseased rhizosphere. We speculate that, given the interplay between plant “cry for help” strategies and pathogen recruitment, DD SynComs do not significantly impact the disease severity index of Verticillium wilt. Although stochastic processes are considered to play a significant role in shaping community structure, their relative importance compared with that of deterministic processes has been a subject of intense debate. Vellend et al. [56] defined neutral stochasticity in community ecology as random changes in community structure related to species identity and/or functional traits, and it can be driven by random variations in the order of colonization, such as priority effects [57]. In this study, the impacts of HS driven by stochastic processes on rhizosphere colonization and resistance to Verticillium wilt are significantly influenced by the inoculation sequence. Preinoculation increases the colonization of the rhizosphere by HS (P. vietnamensis XR-120) and significantly reduces the disease index of Verticillium wilt (Fig. 5a). Although high-concentration inoculation at a later stage significantly reduces the disease index, it does not increase the colonization of the rhizosphere by HS. This result may be because the resistance of HS to Verticillium wilt is not mediated by an increase in its own abundance to suppress pathogen growth, but rather by altering the composition of the rhizosphere microbiome. A study has been conducted on the effects of two bacterial strains (Pseudoalteromonas tunicata D2 and Pseudovibrio sp. D323) on the assembly and succession of microbial communities associated with the green macroalga Ulva australis. In the late successional stage, the abundance of strain D2 is close to zero; however, the community structure remains significantly different from that of the control [58]. This finding suggests that strain D2 has a legacy effect, influencing the assembly trajectory of the recipient microbial community through initial (inter) actions. Therefore, the priority effect initiated in the early stages of microbial community assembly may lead to microbial differences between hosts. Plate confrontation experiments show that bacteria recruited by deterministic processes can effectively inhibit pathogens, while the disease resistance of bacteria driven by stochastic processes is more difficult to predict. Therefore, constructing disease-suppressive SynComs based on strains that were recruited deterministically by the rhizosphere of healthy plants provides more stable and efficient disease resistance compared to those driven by stochastic processes. Although disease suppression was observed in the rhizosphere of Astragalus, it remains unclear whether these findings are applicable to other crops or to different plant tissues, such as the phyllosphere. This study focuses on a specific plant-microbe interaction and the roles of deterministic and stochastic processes; therefore, further research is needed to assess the generalizability of these results to other plant species and agricultural systems. The AKP (Anna Karenina principle) is used to identify and describe the increased stochastic processes in plant microbiome assembly, which are triggered by various external stress factors, ultimately leading to plant disease. In this study, early and high-concentration inoculation of DS SynComs significantly increases the disease index. One study shows that Verticillium dahliae releases the virulence effector VdAve1, which inhibits the colonization of most plant-associated beneficial bacteria in tomatoes and cotton [6]. However, some potentially pathogenic members, such as Pseudomonas corrugate, Ralstonia sp., and Serratia sp. exhibit adaptability [9, 59, 60], indicating that the pathobiome is formed during the successful infection of V. dahliae. Therefore, the formation of a rhizosphere pathobiome accelerates disease progression. However, dysbiosis can be both a cause and a consequence of stress or disease [61]. Based on the results of this study, we suggest that the formation of a pathobiome is more likely to be a cause of the disease. Only early and high-concentration inoculation of DS SynComs enhances their colonization in the rhizosphere, thereby increasing the disease index. This finding suggests that for DS SynComs to invade a stable rhizosphere community, they must do so through priority effects (early inoculation) or by increasing their concentration to outcompete local microorganisms. Additionally, studies have shown that the use of vancomycin to artificially disrupt gram-positive bacteria in the healthy tomato rhizosphere leads to dysbiosis, triggering the onset of bacterial wilt [62]. In the carbon source utilization assay, although the HD and DS SynComs exhibited similar substrate utilization patterns, the ecological strategies of the microbial communities in the healthy and diseased rhizospheres were markedly different. The healthy rhizosphere was predominantly enriched with slow-growing K-strategists, which contributed to community stability and functional maintenance. A study shows that plant growth-promoting microorganisms (PGPMs) can alleviate pathogen stress by driving the enrichment of K-strategists and enhancing microbial community stability [63]. In contrast, the diseased rhizosphere was significantly enriched with fast-growing r-strategists, which rapidly deplete resources, disrupt the compositions of the original microbiota, and accelerate disease progression. Previous studies have shown that pathogenic fungi secrete a variety of cell wall-degrading enzymes (CWDEs) to degrade plant tissues and release nutrients, thereby facilitating pathogen colonization and spread [64]. The r-strategists enriched in the diseased rhizosphere—such as the DS SynComs in this study—may efficiently exploit these newly available resources, proliferate rapidly, and competitively occupy ecological niches. This process may disrupt host control over the rhizosphere microbiota, increase the stochasticity of community assembly, and weaken the overall stability and functionality of the microbiome. Consequently, a pathobiome gradually forms, ultimately promoting disease progression. In this study, we observed a balance between the deterministic recruitment of beneficial rhizosphere microorganisms and pathogen-induced stochasticity. However, an unanswered question remains whether the recruitment of beneficial microorganisms can remain stable and effective under sustained pathogen pressure or throughout the plant’s entire growth cycle. Specifically, during prolonged exposure to pathogens, dysbiosis of the rhizosphere microbiota may intensify, potentially shifting the balance between plant-driven deterministic recruitment and pathogen-induced stochasticity. While our study highlights the importance of deterministic microbial recruitment in maintaining plant health, we also recognize that the long-term stability and efficacy of this mechanism require further investigation. Future research should focus on the long-term effects of repeated infections or sustained pathogen pressure on microbial community assembly. Such studies will help us evaluate the effectiveness of plant-driven recruitment mechanisms across different environments and temporal scales and further elucidate the potential mechanisms underlying the transition of microbial communities toward pathogen-dominated communities. Conclusions Elucidating the intrinsic relationship between microbial community assembly mechanisms and diseases can help optimize the use of rhizosphere microorganisms for the control of soil-borne diseases. Compared with healthy microbiomes, invasion of pathogens increases the proportion of stochastic processes in microbiome assembly, leading to reduced α-diversity and increased β-diversity in the diseased rhizosphere. Microbial communities in diseased individuals become increasingly dispersed. Moreover, the formation of pathobiome promotes symptom development. In contrast, healthy microbiomes are driven primarily by deterministic processes, with greater similarity between individuals. Healthy plants maintain their health by recruiting beneficial microorganisms through deterministic processes. Compared to SynComs constructed through stochastic processes, the disease resistance of the HD SynComs was less affected by the inoculation time and concentration, rendering their function more stable and predictable. On the basis of these findings, our results validate the potential of SynComs constructed based on community assembly processes to resist soil-borne diseases. Moreover, SynComs constructed through deterministic processes can stably contribute to disease resistance. Supplementary Information Supplementary Material 1. Table S1 Construction of SynComs. Fig. S1 Disease-associated restructuring of bacterial and fungal communities in the rhizosphere and roots of Astragalus. a Plants were classified as “diseased” or“healthy” based on visual inspection of both aboveground tissues and the internal root structures. b β-diversity decomposition analysis of bacterial and fungal communities. The total dissimilarity calculated based on the Jaccard-based dissimilarity index was partitioned into two distinct components: Replacement (Repl) and Nestedness (Nest). S: Similarity. c Differential phyla in rhizosphere and root samples. DR: diseased root; HR: heathy root; DRS: diseased rhizosphere; HRS: heathy rhizosphere. Statistical significance was calculated using unpaired two-tailed t-tests. *p < 0.05; **p < 0.01 and***p < 0.001. d Analysis of differential bacterial ASVs in the rhizosphere and roots of healthy and diseased Astragalus. Fig. S2 Identification of pathogenic fungi in diseased Astragalus samples. a Phylogenetic tree of Verticillium ASVs identified in diseased root samples. Phylogenetic tree of Verticillium ASVs based on the Neighbor-Joining (NJ) method. b Sequence alignment between ASV_5 and Verticilium dahliae VD1. HR: Healthy root; DR: Diseased root. Fig. S3 Quantitative assessment of the ecological assembly processes of bacterial and fungal communities. The relative importance of ecological processes driving bacterial and fungal community assembly was quantified using the iCAMP framework. HRS: Healthy rhizosphere soil; DRS: Diseased rhizosphere soil; HR: Healthy root; DR: Diseased root; HoS: homogeneous selection; HeS: heterogeneous selection; HD: homogenizing dispersal; DL: dispersal limitation; DR: drift. Fig. S4 Variations of ecological processes across healthy and diseased root. The circular visualization integrates phylogenetic relationships with ecological assembly processes and taxonomic information at the bin level. a The phylogenetic tree, constructed based on representative ASVs from each bin, is displayed at the center. The rings from the inside to the outside represent the main ecological processes of each bin in the healthy and diseased root (b), the relative abundance of the main bacterial genera in each bin (c), and the relative abundance of each bin (d). The outermost part is labeled with the bin and genus in which the SynCom ASVs are located. Fig. S5 Mechanism of rhizosphere fungal community assembly. The circular visualization integrates phylogenetic relationships with ecological assembly processes and taxonomic information at the bin level. a The phylogenetic tree, constructed based on representative ASVs from each bin, is displayed at the center. The rings from the inside to the outside represent the main ecological processes of each bin in the healthy and diseased rhizosphere (b), the relative abundance of the main bacterial genera in each bin (c), and the relative abundance of each bin (d). DRS: diseased rhizosphere; HRS: heathy rhizosphere. Fig. S6 Mechanism of root fungal community assembly. The circular visualization integrates phylogenetic relationships with ecological assembly processes and taxonomic information at the bin level. a The phylogenetic tree, constructed based on representative ASVs from each bin, is displayed at the center. The rings from the inside to the outside represent the main ecological processes of each bin in the healthy and diseased root (b), the relative abundance of the main bacterial genera in each bin (c), and the relative abundance of each bin (d). The ASV composition and relative abundance of the bin where the pathogenic fungus Fun_5 is located (e). DR: diseased root; HR: heathy root. Fig. S7 Co-occurrence network analysis. a Co-occurrence network topology index of bacterial networks. b The positive and negative cohesion of bacterial networks. c Robustness and stability of the bacterial networks. d Co-occurrence network topology index of fungal networks. e The positive and negative cohesion of fungal networks. f Robustness and stability of the fungal network. DR: diseased root; HR: heathy root; DRS: diseased rhizosphere; HRS: heathy rhizosphere. To facilitate a direct comparison across diverse units and scales, all network topological parameters were normalized to a dimensionless range of [0, 1] using Min-Max scaling for visualization in the radar plot. Fig. S8 The rhizosphere and roots fungal co-occurrence network and network properties of ASVs associated with different ecological processes. a The network is based on the entire fungal communities, with colors representing the distribution of phyla and ecological processes. b Co-occurrence network of healthy and diseased rhizosphere soil and root. c Rhizosphere and roots network ZiPi analysis, with stacked plots on the right indicating the proportion of different ecological process in connectors. d Distribution of betweenness centrality of different ecological process ASVs. e Vulnerability of the rhizosphere and root network before and after removal of different ecological process ASVs to measure the importance of the different ecological process ASVs in contributing to the rhizosphere and root network structure. DR: diseased root; HR: heathy root; DRS: diseased rhizosphere; HRS: heathy rhizosphere. Fig. S9 Phylogenetic characterization of bacterial isolates from the Astragalus rhizosphere. The Neighbor-Joining (NJ) phylogenetic tree was constructed based on the 16S rRNA gene sequences of bacterial strains isolated in this study. Fig. S10 Strategic selection and characterization of bacterial isolates for SynCom construction. The bacterial isolates correspond to the differential ASVs, as well as their Bins and ecological processes. The stacked bar plot quantifies the relative importance of different ecological processes for each Bin. The far-right panel explicitly lists the Genus of each ASV/isolate. Fig. S11 Effects of inoculation of SynComs on bacterial community composition in rhizosphere and roots. a Effects of different inoculation time on bacterial community composition. b Effects of different inoculation concentrations on bacterial community composition. Differences among groups were evaluated using PERMANOVA (adonis2, 999 permutations). Fig. S12 Colonization of synthetic bacteria in rhizosphere and root under different inoculation time treatments. Lowercase letters above the bars indicate significant differences (p < 0.05) between the different treatments (one-way ANOVA). Fig. S13 Colonization of SynComs strains. a Colonization of SynComs bacteria in rhizosphere and roots under different inoculation. b Cumulative relative abundance of HD and DS SynComs under different inoculation concentrations. One-way ANOVA followed by Duncan’s Multiple Range Test (MRT) was employed to evaluate the differences between treatments. Lowercase letters above the bars indicate significant differences (p < 0.05) between the different treatments. Fig. S14 Effects of HD SynCom and HS strain combinations on Verticillium wilt resistance. One-way ANOVA followed by Duncan’s Multiple Range Test (MRT) was employed to evaluate the differences between treatments. Lowercase letters above the bars indicate significant differences (p < 0.05) between the different treatments. Fig. S15 The β-diversity differences of healthy and diseased rhizosphere samples collected in the field. a β-Diversity analysis of field samples. b Bata dispersion analysis of different samples. DRS: diseased rhizosphere; HRS: heathy rhizosphere. Fig. S16 Antibacterial activity assay of HD SynComs. Control: Verticillium dahliae VD1; HD: HD SynComs; HS: P. vietnamensis XR-120. Lowercase letters above the boxes indicate significant differences (p < 0.05) between the different treatments (one-way ANOVA). Fig. S17 Effects of isolates XS-134-8 and XS-164 on the disease index of Astragalus Verticillium wilt. Lowercase letters above the boxes indicate significant differences (p< 0.05) between the different treatments (one-way ANOVA). Supplementary Material 1. Table S1 Construction of SynComs. Fig. S1 Disease-associated restructuring of bacterial and fungal communities in the rhizosphere and roots of Astragalus. a Plants were classified as “diseased” or“healthy” based on visual inspection of both aboveground tissues and the internal root structures. b β-diversity decomposition analysis of bacterial and fungal communities. The total dissimilarity calculated based on the Jaccard-based dissimilarity index was partitioned into two distinct components: Replacement (Repl) and Nestedness (Nest). S: Similarity. c Differential phyla in rhizosphere and root samples. DR: diseased root; HR: heathy root; DRS: diseased rhizosphere; HRS: heathy rhizosphere. Statistical significance was calculated using unpaired two-tailed t-tests. *p < 0.05; **p < 0.01 and***p < 0.001. d Analysis of differential bacterial ASVs in the rhizosphere and roots of healthy and diseased Astragalus. Fig. S2 Identification of pathogenic fungi in diseased Astragalus samples. a Phylogenetic tree of Verticillium ASVs identified in diseased root samples. Phylogenetic tree of Verticillium ASVs based on the Neighbor-Joining (NJ) method. b Sequence alignment between ASV_5 and Verticilium dahliae VD1. HR: Healthy root; DR: Diseased root. Fig. S3 Quantitative assessment of the ecological assembly processes of bacterial and fungal communities. The relative importance of ecological processes driving bacterial and fungal community assembly was quantified using the iCAMP framework. HRS: Healthy rhizosphere soil; DRS: Diseased rhizosphere soil; HR: Healthy root; DR: Diseased root; HoS: homogeneous selection; HeS: heterogeneous selection; HD: homogenizing dispersal; DL: dispersal limitation; DR: drift. Fig. S4 Variations of ecological processes across healthy and diseased root. The circular visualization integrates phylogenetic relationships with ecological assembly processes and taxonomic information at the bin level. a The phylogenetic tree, constructed based on representative ASVs from each bin, is displayed at the center. The rings from the inside to the outside represent the main ecological processes of each bin in the healthy and diseased root (b), the relative abundance of the main bacterial genera in each bin (c), and the relative abundance of each bin (d). The outermost part is labeled with the bin and genus in which the SynCom ASVs are located. Fig. S5 Mechanism of rhizosphere fungal community assembly. The circular visualization integrates phylogenetic relationships with ecological assembly processes and taxonomic information at the bin level. a The phylogenetic tree, constructed based on representative ASVs from each bin, is displayed at the center. The rings from the inside to the outside represent the main ecological processes of each bin in the healthy and diseased rhizosphere (b), the relative abundance of the main bacterial genera in each bin (c), and the relative abundance of each bin (d). DRS: diseased rhizosphere; HRS: heathy rhizosphere. Fig. S6 Mechanism of root fungal community assembly. The circular visualization integrates phylogenetic relationships with ecological assembly processes and taxonomic information at the bin level. a The phylogenetic tree, constructed based on representative ASVs from each bin, is displayed at the center. The rings from the inside to the outside represent the main ecological processes of each bin in the healthy and diseased root (b), the relative abundance of the main bacterial genera in each bin (c), and the relative abundance of each bin (d). The ASV composition and relative abundance of the bin where the pathogenic fungus Fun_5 is located (e). DR: diseased root; HR: heathy root. Fig. S7 Co-occurrence network analysis. a Co-occurrence network topology index of bacterial networks. b The positive and negative cohesion of bacterial networks. c Robustness and stability of the bacterial networks. d Co-occurrence network topology index of fungal networks. e The positive and negative cohesion of fungal networks. f Robustness and stability of the fungal network. DR: diseased root; HR: heathy root; DRS: diseased rhizosphere; HRS: heathy rhizosphere. To facilitate a direct comparison across diverse units and scales, all network topological parameters were normalized to a dimensionless range of [0, 1] using Min-Max scaling for visualization in the radar plot. Fig. S8 The rhizosphere and roots fungal co-occurrence network and network properties of ASVs associated with different ecological processes. a The network is based on the entire fungal communities, with colors representing the distribution of phyla and ecological processes. b Co-occurrence network of healthy and diseased rhizosphere soil and root. c Rhizosphere and roots network ZiPi analysis, with stacked plots on the right indicating the proportion of different ecological process in connectors. d Distribution of betweenness centrality of different ecological process ASVs. e Vulnerability of the rhizosphere and root network before and after removal of different ecological process ASVs to measure the importance of the different ecological process ASVs in contributing to the rhizosphere and root network structure. DR: diseased root; HR: heathy root; DRS: diseased rhizosphere; HRS: heathy rhizosphere. Fig. S9 Phylogenetic characterization of bacterial isolates from the Astragalus rhizosphere. The Neighbor-Joining (NJ) phylogenetic tree was constructed based on the 16S rRNA gene sequences of bacterial strains isolated in this study. Fig. S10 Strategic selection and characterization of bacterial isolates for SynCom construction. The bacterial isolates correspond to the differential ASVs, as well as their Bins and ecological processes. The stacked bar plot quantifies the relative importance of different ecological processes for each Bin. The far-right panel explicitly lists the Genus of each ASV/isolate. Fig. S11 Effects of inoculation of SynComs on bacterial community composition in rhizosphere and roots. a Effects of different inoculation time on bacterial community composition. b Effects of different inoculation concentrations on bacterial community composition. Differences among groups were evaluated using PERMANOVA (adonis2, 999 permutations). Fig. S12 Colonization of synthetic bacteria in rhizosphere and root under different inoculation time treatments. Lowercase letters above the bars indicate significant differences (p < 0.05) between the different treatments (one-way ANOVA). Fig. S13 Colonization of SynComs strains. a Colonization of SynComs bacteria in rhizosphere and roots under different inoculation. b Cumulative relative abundance of HD and DS SynComs under different inoculation concentrations. One-way ANOVA followed by Duncan’s Multiple Range Test (MRT) was employed to evaluate the differences between treatments. Lowercase letters above the bars indicate significant differences (p < 0.05) between the different treatments. Fig. S14 Effects of HD SynCom and HS strain combinations on Verticillium wilt resistance. One-way ANOVA followed by Duncan’s Multiple Range Test (MRT) was employed to evaluate the differences between treatments. Lowercase letters above the bars indicate significant differences (p < 0.05) between the different treatments. Fig. S15 The β-diversity differences of healthy and diseased rhizosphere samples collected in the field. a β-Diversity analysis of field samples. b Bata dispersion analysis of different samples. DRS: diseased rhizosphere; HRS: heathy rhizosphere. Fig. S16 Antibacterial activity assay of HD SynComs. Control: Verticillium dahliae VD1; HD: HD SynComs; HS: P. vietnamensis XR-120. Lowercase letters above the boxes indicate significant differences (p < 0.05) between the different treatments (one-way ANOVA). Fig. S17 Effects of isolates XS-134-8 and XS-164 on the disease index of Astragalus Verticillium wilt. Lowercase letters above the boxes indicate significant differences (p< 0.05) between the different treatments (one-way ANOVA).