MdUGT88F1 enhances plant resistance to Fusarium proliferatum f.sp. malus domestica MR5 via root exudate-mediated assembly of disease-suppressive rhizosphere microbiota

Abstract

Background

Apple replant disease (ARD) is a major threat to the sustainable development of China’s apple industry. It is primarily caused by the accumulation of phloridzin and the pathogen Fusarium proliferatum f.sp. malus domestica MR5 (Fpmd MR5). MdUGT88F1-mediated phloridzin biosynthesis is known to enhance disease resistance, but its role in shaping the rhizosphere microbiome and conferring resistance against Fpmd MR5 remains unclear. In this study, we used wild-type (WT) and MdUGT88F1 transgenic apple lines to systematically investigate the mechanism by which MdUGT88F1 regulates the rhizosphere microbiome to mitigate ARD.

Results

Compared with WT and MdUGT88F1-OE plants, MdUGT88F1-RNAi plants exhibited enhanced tolerance to ARD, as indicated by reduced disease severity, decreased abundance of Fpmd MR5 in the rhizosphere soil, and lower phloridzin content. Further greenhouse experiments demonstrated that the rhizosphere bacterial communities were triggered mainly by changes in community composition. Multi-omics joint analysis revealed that members of the family Bacillaceae with multiple plant growth-promoting traits were enriched in the MdUGT88F1-RNAi plant rhizosphere but only upon Fpmd MR5 invasion. MdUGT88F1-RNAi plants exhibited significantly higher exudation of D-tagatose, D-galactose, sucrose, 3-O-methyl-D-glucose, and maltitol. Interestingly, exogenous application of these compounds promoted the proliferation of Bacillus, enhancing plant resistance to Fpmd MR5. In vitro assays demonstrated that the recruited Bacillus significantly inhibited the hyphal growth and fumonisin B1 production of Fpmd MR5 and alleviated plant disease symptoms. We experimentally validated this observation by inoculating a synthetic microbial community (Bacillus velezensis, Bacillus mojavensis, Bacillus subtilis, Bacillus amyloliquefaciens, and Bacillus licheniformis) into replanted soil, which led to a significant reduction in pathogen Fusarium abundance and promoted plant growth.

Conclusion

Overall, these findings highlight that plant disease resistance is a complex trait driven by dynamic interactions among the host genetic background, rhizospheric microbial communities, and pathogens. Targeted modulation of the rhizospheric microbiome represents a potent “prebiotic” strategy. This approach can indirectly enhance plant disease resistance by fostering beneficial microbial activity in the rhizosphere. This study also provides a theoretical basis and practical solutions for the green control of ARD through prebiotics and synthetic microbial communities.

Download video file ^{(40.8MB, mp4)}

Video Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s40168-026-02416-7.

Introduction

The renovation of aging orchards in the main apple-producing regions in China (the Northwest Loess and the region around Bohai Gulf) is pivotal for overcoming the bottleneck of high-quality and sustainable development of the industry [13, 52]. Due to land policy restrictions, replanting in old orchards is unavoidable, leading to the widespread occurrence of ARD [20, 89]. ARD can result in weakened tree growth and photosynthesis, root necrosis, decreased chlorophyll content and stomatal conductance, reduced fruit yield and quality, and even the death of young trees [22, 60, 89, 102]. The main cause is the accumulation of soilborne pathogens (such as nematodes, fungi, and oomycetes) from residual roots and decaying organic matter in old orchards, which affect rhizosphere microorganisms and cause root infection [3, 38, 62, 78, 82, 83, 89]. In China’s primary apple-producing regions, the accumulation of rhizosphere phenolic compounds (e.g., phloridzin) and Fusarium spp. (i.e., Fusarium proliferatum, F. moniliforme, F. oxysporum, F. solani, and Fpmd MR5) was the main factor contributing to ARD occurrence [22, 97, 103]. Therefore, it is urgent to clarify how to control ARD through research on rhizosphere microorganisms.

Traditional strategies for controlling ARD, including agronomic practices, soil chemical fumigation, and the breeding of resistant rootstocks [37, 63, 89, 93], are gradually being phased out due to environmental concerns, high costs, and inefficiency [21, 37, 74]. In recent years, stimulating the soil’s suppressive capacity against certain pathogens and applying biocontrol agents to induce systemic resistance in plants have emerged as a widely recognized alternative for sustainable agricultural development [18, 74]. Beneficial rhizosphere microbes play essential roles in nutrient cycling, growth promotion, disease suppression, and mitigation of abiotic stresses through interaction with plants [15, 51, 58, 87]. Currently, extensive research has shown that particular microbial groups (e.g., Pseudomonas, Bacillus, Flavobacterium, and Trichoderma) can achieve specific disease inhibition by directly antagonizing pathogens, as well as confer broad-spectrum disease suppression by regulating the entire soil microbiome [32, 49, 68]. In practice, the application of bio-organic fertilizers or biocontrol agents like Trichoderma can directionally optimize the rhizosphere microbiome and effectively enhance plant disease resistance by enriching beneficial microorganisms, occupying pathogen niches, and inducing the production of secondary metabolites [18, 81]. However, the mechanisms governing soil suppressiveness and pathogen control are poorly understood. Unraveling how plants dynamically recruit beneficial microbes from the rhizosphere is essential to understanding microbiome-mediated disease suppression.

Plant host genes mediate plant disease resistance by regulating the quantity and composition of root exudates [32, 71], thereby modulating microbiome composition and function [48, 54, 100]. For example, in apples, the glycosyltransferase gene MdUGT88F1 catalyzes the conversion of phloretin to phloridzin [112], enhancing disease resistance by restricting pathogen growth and toxin biosynthesis, enhancing plant structural integrity, improving nitrogen metabolism, and promoting salicylic acid accumulation [111–113]. In tomato, mutations in the ethylene receptor gene altered the composition of rhizosphere bacterial communities [29] and affect the enrichment of bacteria and nematodes through the regulation of root exudates [14, 23]. Furthermore, plant can influence the structure and activity of rhizosphere communities and enhance host disease resistance through hormones such as salicylic acid, jasmonic acid, and ethylene [24], as well as root exudates [96], siderophores [33], and antimicrobial compounds [79]. However, the key plant genes regulating microbiome-mediated disease resistance remain poorly understood [100]. Therefore, deciphering how plant genetic factors enhance disease resistance by regulating root exudates and the microbiome is a key breakthrough for developing sustainable ARD control strategies.

Here, we hypothesized that MdUGT88F1 regulates phloridzin synthesis through glycosylation pathways, altering root exudation patterns to promote the growth or activity of specific microbes, thereby shaping the microbial community and contributing to resistance against Fpmd MR5. To test this hypothesis, we first constructed a specific rhizosphere microbial consortium using MdUGT88F1-RNAi plants, and then employed microcosm experiments to monitor the interaction dynamics between the pathogen and the microbiome following Fpmd MR5 inoculation. Simultaneously, we identified the differentially secreted compounds in the rhizosphere of MdUGT88F1-RNAi plants and assessed their direct and indirect effects on soil microbiome and disease suppressiveness. Finally, we isolated potential key microbial taxa responsible for disease suppression in the rhizosphere and further validated their inhibitory activity and ARD control efficacy through in vivo and in vitro assays. This study reveals the underlying mechanism by which plant metabolites recruit specific microorganisms to enhance disease resistance, providing a theoretical basis and new strategies for ARD control.

Methods

Plant material and growth conditions

Seeds of Malus hupehensis Rehd. were surface-sterilized (treated with 3% NaClO for 5 min followed by 70% ethyl alcohol for 1 min), stratified at 4 °C for approximately 30 days. In early March, seeds with visible bud emergence and cracked coats were sown in 50-cell seedling trays filled with sterile substrate (organic matter ≥ 45%, pH 5.8–6.5, EC 1.0–1.5 mS·cm⁻¹) after selecting seeds with visible bud emergence and cracked coats. After one month, uniformly grown seedlings were transplanted into sterile pots (32 cm × 25 cm). Following an additional 30 days of cultivation, disease-free plants with consistent growth were selected for further experiments [22].

A line with high regenerative capacity isolated from “Royal Gala” (Malus domestica) named GL-3 was used for genetic transformation [41]. In this study, all previously obtained MdUGT88F1 transgenic apple lines were used, including two pCambia2300-mediated overexpression lines (OE2-7 and OE3-4) and two pHellsgate2-mediated silencing lines (Ri-3 and Ri-6; [112]). In January 2023, healthy and pest-free apple plant stems were selected and inoculated onto Murashige and Skoog subculture medium containing 0.2 mg·L⁻¹ Indole-3-acetic acid (IAA) and 0.3 mg·L⁻¹ 6-BA (6-Benzylaminopurine). After one month, they were transferred to a rooting medium supplemented with 0.5 mg·L⁻¹ IAA and 0.5 mg·L⁻¹ IBA (Indole-3-butyric acid) for root induction. Following 45 days of culture, the plantlets were transplanted into plastic pots filled with a sterile substrate mixture (vermiculite/humus/perlite, 1:3:1, v/v/v) and grown in a light incubator for two months. Finally, well-grown and uniformly developed plants were selected for subsequent experiments.

Experimental soils

The replant soil was collected in October 2022 (autumn) from a 30-year-old apple orchard in Yangling District (34°20'N, 108°24'E), Shaanxi, China, when the average daily temperature was approximately 13 °C. This apple orchard was selected due to the presence of plants showing typical symptoms of replant disease in small localized plots, while the remaining trees in the orchards exhibited optimal vigor. The soil was classified as clay soil, containing 11.13 g·kg⁻¹ of organic matter, 0.70 g·kg⁻¹ of total nitrogen (N), 1.03 g·kg⁻¹ of total phosphorus (P), 18.57 g·kg⁻¹ of total potassium (K), 43.69 mg·kg⁻¹ of alkali-hydrolyzable nitrogen, 12.9 mg·kg⁻¹ of ammonia (NH₄⁺-N), 13.20 mg·kg⁻¹ of available phosphorus (AP), 169 mg·kg⁻¹ of available potassium (AK), and a pH of 6.2. The soil field capacity was 363.49 g·kg⁻¹. After removing surface debris on the ground, soil was collected from the root zone of trees displaying ARD symptoms, characterized by diminished plant vitality, leaf chlorosis, stunted twig growth, or mortality. Sampling was conducted at five points per orchard, within a 30 cm diameter area around the trunk at a depth of 10–20 cm. All experimental soils were sieved (< 4 mm) and homogenized thoroughly for the subsequent experiments. A portion of the bulk soil was stored moist in plastic-lined totes at 5 °C during the winter period and used for ARD bioassays in the greenhouse in the following spring (May 2023) [97]. Homogenized soil samples were then divided into two groups: natural soil (containing natural apple microbiota) and sterile soil, which was autoclaved twice at 121 °C for 60 min [18]. Soils were stored in a cool, ventilated, dry place for 7 days at 25 °C prior to their utilization in the experiments.

Pathogen and growth conditions

Fpmd MR5, provided by Professor Zhiquan Mao (College of Horticulture, Shandong Agricultural University), was stored as a 22% glycerol stock culture. Conidia were produced by transferring mycelial plugs from one-week-old Potato Dextrose Agar (PDA) plates (Solarbio, Beijing, China) into liquid Potato Dextrose Broth (PDB) medium (Solarbio) and subcultured in the dark at 28 °C on a rotary shaker. After 1 week, the conidia were harvested by filtration through eight layers of sterile gauze, centrifuged at 8000 rpm for 8 min, and resuspended in sterile PDB medium to adjust the concentration. The spore suspension was then diluted, thoroughly mixed, and counted using a hemocytometer under a microscope. The final spore suspension concentration was adjusted to 5 × 10⁶ spores·mL⁻¹ [19].

Experimental setup

Evaluation of the tolerance of MdUGT88F1 overexpression and silenced transgenic apple lines to ARD

A greenhouse experiment was conducted to assess the tolerance of MdUGT88F1 transgenic apple lines and WT to ARD caused by Fpmd MR5. The experiment comprised five treatments: WT, MdUGT88F1-RNAi (Ri-3 and Ri-6), and MdUGT88F1-OE (OE2-7 and OE3-4) lines were planted in natural soil. Based on the severity of ARD, the natural soil was classified as severe (with a dry weight inhibition rate = 106.69% (P < 0.001, t = 29.130, df = 8, two-sided Student’s t-test)) (Supplementary Fig. S1a). The transgenic apple lines and WT plants were transferred to AC170 (15 cm × 11.5 cm × 9 cm) pots filled with natural soil (1.5 kg of soil per pot) on May 17, 2023. Each treatment consisted of three independent replicate plots, with each plot containing 20 plants, for a total of 60 plants per treatment and 300 plants in the entire experiment. Plants were maintained under glasshouse conditions (ambient temperature variation ranging from 25 °C to 30 °C; 70% relative humidity; 16/8-h photoperiod with 200 μmol·m⁻²·s⁻¹ photosynthetic photon flux density, PPFD) and watered regularly with sterile water. The position of the pots was randomized every 2 days within treatments. After 90 days, the glasshouse experiment was terminated. On the 90th day after transplanting, plant growth parameters (biomass, mineral elements, antioxidant system, photosynthetic characteristics, and root system architecture) of five randomly selected plants from each replicate were measured. The disease severity was quantified during the experiment using the disease intensity (DI), with the specific calculation method referenced from Duan et al. [22]. One value was obtained from each plot; thus, each treatment had three DI replicates.

The soil samples were collected following the protocol established by Deng et al. [18]. Samples were stored at − 80 °C for the quantification of pathogen density, Illumina MiSeq sequencing, metagenomic sequencing, and gas chromatography-mass spectrometry (GC–MS) metabolomics analysis. The soil samples used for bacterial isolation were stored at − 4 °C, and the remaining soil was sieved (50-mesh) for the determination of soil physicochemical properties [22].

Bioassay to assess soil disease suppressiveness

Since the MdUGT88F1-OE (OE) treatment exhibited lower resistance to ARD compared to the WT treatment, soils from the OE and MdUGT88F1-RNAi (RNAi) treatments were selected for subsequent microcosm experiments to investigate potential suppression mechanisms. In May 2024, we first evaluated the disease suppressiveness of RNAi soils using a previously established protocol [18]. Briefly, soils from the OE and RNAi pot treatments were collected. Five experimental treatments were designed as follows: (1) RNAi: soil from the RNAi treatment. (2) OE: soil from the OE treatment. (3) OE + SRNAi: a mixture of 90% OE soil and 10% 121 °C-autoclaved RNAi soil (w/w). (4) OE + 50RNAi: A mixture of 90% OE soil and 10% 50 °C heat-treated (1 h) RNAi soil (w/w). (5) OE + RNAi: a mixture of 90% OE soil and 10% untreated RNAi soil (w/w). Malus hupehensis Rehd. seedlings were transplanted into sterilized plastic pots (9 cm × 7.2 cm × 6 cm) containing 200 g of each of these soil mixtures. Each treatment was replicated three times in a completely randomized design, with each replicate consisting of 10 pots. Seedlings were grown in a controlled-environment chamber under a 16 h light: 8 h dark photoperiod at 28 °C for 30 days. Pots were irrigated with half-strength Hoagland’s solution every 3 days and deionized water as needed. Subsequently, all pots were inoculated with Fpmd MR5 [21] at a concentration of 10⁵ conidia·g⁻¹ artificial soil per pot. Finally, seedlings were planted, and the disease incidence rate was determined using the formula: Disease incidence rate (%) = (Number of diseased plants/Total number of plants) × 100 [22].

Pot experiment for Fpmd MR5 inoculation

In July 2024, we conducted a microcosm pot experiment to investigate the dynamic changes in the rhizosphere microbiome under Fpmd MR5 inoculation. Bulk soil collected from each pot of the OE and RNAi plants was used as the inoculum. To minimize the influence of endophytic microbes, Malus hupehensis Rehd. seeds were surface-sterilized and grown on a sterile substrate [58, 85]. In this experiment and all subsequent pot trials, Malus hupehensis Rehd. served as a model plant for the Malus genus, widely used to study physiological responses to biotic and abiotic stresses [21, 22, 88, 89, 97]. The OE and RNAi soil inocula were mixed with sterile soil at a 1:9 (v/v) ratio to prepare the growth substrate (Fig. 1b). Malus hupehensis Rehd. seedlings were then transplanted into sterilized plastic pots containing 200 g of the soil substrate. A total of 108 pots were used per soil type, with each treatment comprising three replicates in a completely randomized design, each containing nine pots. Seedlings were cultivated in a growth chamber under controlled conditions (16 h light: 8 h dark, 28 °C) for 30 days. Irrigation was performed every three days using half-strength Hoagland’s solution, supplemented with deionized water as needed. Subsequently, seedlings with established rhizospheres were transferred to new sterile pots containing sterile artificial soil. Half of these pots were inoculated with Fpmd MR5 at a concentration of 10⁵ conidia·g⁻¹ artificial soil. The seedlings were then grown until disease symptoms became apparent (Fig. 1b).

Fig. 1.

Overview of experimental design. a The glasshouse experiment was initiated in May 2023. Bulk and rhizosphere soils were sampled in August 2023. b Design of the pot experiment for the inoculation assay of Fusarium proliferatum f.sp. malus domestica MR5 (Fpmd MR5). c Effects of exudate compound classes on soil suppressiveness and chemotaxis. n, number of replicates per treatment

Soil substrates containing each soil type were collected at the initial time point into 2 mL sterilized microcentrifuge tubes, designated as “Bulk” samples. These included two treatments: BMO (sterile artificial soil mixed with OE soil inoculum) and BMR (sterile artificial soil mixed with RNAi soil inoculum). Rhizosphere samples collected before transplanting and pathogen inoculation were designated as “MRT0”, comprising: MO (rhizosphere derived from OE soil microbiome) and MR (rhizosphere derived from RNAi soil microbiome). Fourteen days post-pathogen inoculation, samples were designated as “MRT14” and categorized into four treatments: HMO (OE rhizosphere without Fpmd MR5 inoculation), FMO (OE rhizosphere inoculated with Fpmd MR5), HMR (RNAi rhizosphere not inoculated with Fpmd MR5), and FMR (RNAi rhizosphere inoculated with Fpmd MR5). Each treatment at each time point comprised three independent biological replicates, with samples exclusively taken from healthy plants. Roots from each replicate of MRT0 and MRT14 samples were collected and vigorously shaken to remove loosely attached soil. Rhizosphere soils were obtained by washing 3 g of roots from each sample in 30 mL sterile Phosphate-Buffered Saline (PBS). The suspensions were filtered through sterilized 0.45 μm membrane filters, collected in sterilized 5 mL tubes, flash-frozen in liquid nitrogen, and stored at − 80 °C until DNA extraction.

DNA extraction and quantitative polymerase chain reaction (qPCR) assay

Field soil samples from the pot experiments were chosen for subsequent DNA extraction using the SPINeasy™ DNA Pro Kit for Soil (MP Biomedicals, UK). The quality and quantity of DNA were determined using a NanoDrop ND-2000 spectrophotometer (NanoDrop, Wilmington, DE, USA). Abundances of bacteria, fungi, Fusarium oxysporum, Fusarium verticillioides, Fusarium proliferatum, Fusarium solani, Fpmd MR5, and Bacillus were determined with Eub338F/Eub518R, ITS1f/5.8 s [28], JR/JF, CHR/CHF, CR/CF, FR/FF (Duan et al., 2022a), MR5R/MR5F, and Bs16S1/Bs16SR [66] primers, respectively, following established protocols using a CFX96 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). The primers and thermal cycling conditions are presented in Supplementary Table S1. Thermal cycling conditions for each sample were run according to a standard procedure with three parallels, and the results were expressed as log₁₀ values (target copy number g⁻¹ soil).

Illumina MiSeq sequencing

Total DNA was extracted from 0.5 g of soil using a MagPure Soil DNA LQ kit (Magen, Guangdong, China) according to the manufacturer’s protocols. DNA concentration and integrity were measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and agarose gel electrophoresis, respectively. PCR amplification of the V3-V4 hypervariable regions of the bacterial 16S rRNA gene was conducted using universal primer pairs (forward primer, 343F-5′-GCTGCGTTCTTCATCGATGC -3′; reverse primer 798R-5′-AGGGTATCTAATCCT-3′) [69] and of the variable ITS1 region of the fungal ITS gene using universal primer pairs (forward primer, ITS1F-5'-CTTGGTCATTTAGAGGAAGTAA-3′, reverse primer ITS2-5'-GCTGCGTTCTTCATCGATGC-3′) [67]. The PCR products were purified using Agencourt AMPure XP beads (Beckman Coulter Co., USA) and quantified with the Qubit dsDNA assay kit (Life Technologies, ThermoFisher Scientific, USA). Purified amplicons were sequenced on an Illumina NovaSeq 6000 platform with two paired-end read cycles of 250 bases each (Illumina Inc., San Diego, CA; OE Biotech Company, Shanghai, China).

Raw sequencing FASTQ data were preprocessed using the cutadapt software for adapter trimming, followed by quality filtering, denoising, merging, and chimera removal using DADA2 [7] with the default parameters of QIIME2 [5]. Finally, the representative reads and the ASV abundance table were generated. The representative read of each ASV was selected using the QIIME 2 package [5]. All representative reads were annotated and blasted against the SILVA rRNA database [73] (version: r138.1) using q2-feature-classifier with the default parameters. The amplicon sequencing and analysis were conducted by OE Biotech Co., Ltd. (Shanghai, China).

DNA extraction and metagenomic shotgun sequencing of rhizosphere soil samples

Soil DNA was extracted and quantified as described above. Purified PCR products were quantified using Qubit 3.0 (Thermo Fisher Scientific). Metagenomic shotgun sequencing libraries were constructed and sequenced at OE Biotech Co., Ltd. (Shanghai, China). Three random DNA samples from each treatment were sequenced on the Illumina Novaseq 6000 platform using the PE150 sequencing strategy, with an average of 10 Gb per sample. Reads were trimmed and filtered using fastp (v 0.20.1), and then aligned against the apple (Malus domestica) genome to remove host-associated sequence contamination. MEGAHIT (v 1.2.9) was used to assemble the scaffolds (≥ 500 bp); then, prodigal (v 2.6.3) was used to predict the open reading frames (ORFs) (≥ 90 bp). The non-redundant gene sets were built for all predicted genes using MMSeqs2 (v 13.45111). The clustering parameters were 95% identity and 90% coverage. The longest gene was selected as the representative sequence of each gene set. Clean reads of each sample were aligned against the non-redundant geneset (95% identity) using salmon (v 1.8.0), and the abundance information of the genes in the corresponding sample was counted. The taxonomy of the species was assigned using the NR Library database, and the abundance of the species was calculated using the corresponding abundance of the genes. Gene functions were annotated using the KEGG, NR, and eggNOG databases with an e-value cutoff of 1e−5 using DIAMOND (v 0.9.10.111) [108]. Furthermore, contigs > 5 kb were processed with hmmscan (v 3.1) with default parameters to analyze the secondary metabolism ability. The sequencing and analysis were conducted by OE Biotech Co., Ltd. (Shanghai, China).

Soil physical and chemical properties analyses

The pretreatment of bulk soil samples and the determination of AP, AK, nitrate (NO₃⁻), NH₄⁺-N, microbial biomass carbon, P, and N, total organic carbon, organic matter, dissolved organic carbon, total N, total P, total K, alkali-hydrolyzable N, and pH were carried out by Nanjing Ruiyuan Biotechnology Co., Ltd (Nanjing, China). The soil field capacity was measured using the cutting ring method according to Wang et al. [90].

The activities of superoxide dismutase (SOD), peroxidase (POD), catalase (CAT), ascorbate peroxidase (APX), solid-urease (S-UE), solid-acid phosphatase (S-ACP), solid-sucrase (S-SC), solid-catalase (S-CAT), solid-β-Glucosidase (S-β-GC), solid-polyphenol oxidase (S-PPO), and dehydrogenase (DHA), and the content of H₂O₂ (hydrogen peroxide), O₂⁻ (superoxide anion radical), and malondialdehyde (MDA), were performed using detection kits based on the manufacturer’s instructions (Geruisi Bioengineer Company, Suzhou, China). The leaves were collected to examine the accumulation of H₂O₂ and O₂⁻ by histochemical staining using DAB (3,3′-diaminobenzidine) and NBT (Nitro Blue Tetrazolium), respectively. The populations of soil microbes (bacteria, fungi, and actinomycetes) were assessed using the dilution method of plate counting as described by Duan et al. [21].

The sample extraction and phenolic acid content analysis were performed in Nanjing Ruiyuan Biotechnology Co., Ltd (Nanjing, China). Briefly, 1 g of soil sample was digested with 1 M NaOH at 4 °C for 16 h, then 5 mL of 1 M NaOH was added to the precipitate and ultrasonicated for 30 min, and finally the addition of 5 mL of ethyl acetate for extraction. The supernatant was dried with nitrogen, then dissolved in 0.4 mL of methanol, passed through a 0.22-μm filter membrane, and the content of phenolic acids was analyzed using an Agilent LC–MS/MS 6420 system (Agilent 1260 infinity HPLC and Agilent 6420 triple quadrupole mass spectrometer) (Agilent Technologies, Wilmington, DE, USA).

Root exudate collection and GC–MS analysis

For root exudate collection, roots of intact plants were carefully washed to remove all remaining soil, and dead roots were removed using stainless steel tweezers. Each individual plant was then transferred to 100 mL of a hydroponic solution (1:10 soil: Milli-Q water mixture) in a controlled chamber for 48 h. This incubation period allowed the roots to recover from the stress of washing and prevented the collection of leaked root cell contents as a result of damage done by root washing and the rewetting of dry plant roots (as reviewed in [72]). After 48 h, each plant was rinsed and then transferred to a flask with 100 mL sterile Milli-Q water (placed on ice) and shaken at 60 rpm at 18 °C and ambient light for 2 h to collect root exudates. An 80 mL sample of each root exudate solution was filter-sterilized to 0.22 μm, lyophilized, and stored as powder at − 80 °C for further study [17]. The fresh roots were washed with deionized water, air-dried to remove the water adhering to them, dried at 60 °C for 72 h, and then their dry weight was determined [6].

Root exudates were extracted from the lyophilized samples with methanol. GC–MS analyses were performed as previously described [12], using an Agilent 8890–5977B Gas chromatograph–mass spectrometer. Gas chromatography was performed on a DB–5MS capillary column (30 m × 0.25 mm × 0.25 μm, Agilent J&W Scientific, Folsom, CA, USA). Before further analysis, the GC–MS results were first processed using MS–DIAL software (v4.24) to obtain the initial data matrix. Metabolites were identified based on multiple dimensions such as retention time, exact mass, and fragment ion mass spectrum. The LuMet–GC 5.0 database (untargeted GC–MS database from Lumingbio) was used for metabolite identification and analysis. Metabolite peak areas were normalized to that of the internal standard and expressed per unit root dry mass [6, 35]. The analysis was conducted by OE Biotech Co., Ltd. (Shanghai, China).

Assessing the impact of representative root exudate compounds on DI in RNAi and OE plants

We prepared two categories of differentially secreted root exudates based on their chemical composition: (1) amino acids + carbohydrates (ACs): l-alanine, l-leucine, l-valine, DL-xylose, and glycerol; (2) fatty acyls + carbohydrates (FCs): D-tagatose, D-galactose, sucrose, 3-O-methyl-D-glucose, and maltitol. The representative compounds for each category were selected based on three key criteria: (1) differential abundance: compounds exhibiting significantly higher or lower relative abundance in RNAi plants compared to OE plants (P < 0.05); (2) safety: compounds that are non-toxic to both humans and soil ecosystems; (3) availability: commercially available compounds that could be obtained in sufficient quantities at a reasonable cost for experimental purposes.

Aqueous solutions for both ACs (amino acids + carbohydrates) and FCs (fatty Acyls + carbohydrates) treatments were prepared by combining equal concentrations of each selected compound to achieve a final total concentration of 10 mM. Specifically: ACs solution (2 mM each): l-alanine, l-leucine, l-valine, DL-xylose, and glycerol; FCs solution (2 mM each): D-tagatose, D-galactose, sucrose, maltitol, and 3-O-methyl-D-glucose. Fifteen grams of natural soil was aliquoted into each well of 6-well plates. Plates were incubated in a growth chamber at 30 °C for 1 week to stabilize the soil microbiome. Each well received 1.5 mL of treatment solution twice weekly for 8.5 weeks (17 total applications). Three treatments were applied: (1) ACs solution (simulating OE-induced root exudates); (2) FCs solution (simulating RNAi-induced root exudates); (3) Sterile water (control). Each treatment comprised 18 biological replicates distributed across 3 plates, with randomized plate placement during incubation. The protective effect on Malus hupehensis Rehd. seedlings’ roots was assessed through Periodic Acid-Schiff (PAS) staining following established protocols [21].

To examine the influence of exudate compound classes on the development of soil suppressiveness, soil slurries were prepared from the treated soils and filter-sterilized as described previously [106]. Briefly, 5 g of soil from each treatment was mixed with 50 mL autoclaved water on an orbital shaker for 1 h. After settling for an additional hour, the soil slurry was obtained from the supernatant through filter paper filtration, which was then filter-sterilized using a 0.22-μm polytetrafluoroethylene (PTFE) membrane filter. Slurries were prepared from the ACs, FCs, control-treated soils, and mixtures of ACs and FCs-treated soils at ratios of 9:1, 5:5, and 1:9 (v/v). The artificial soil was a mixture of humus and sterile soil (horticultural grade, v/v 2:1). All artificial soil was autoclaved twice at 121 °C for 60 min [18]. In May 2024, Malus hupehensis Rehd. seedlings were transplanted into pots containing the autoclaved artificial soil mixture, supplemented with either 20 mL of soil slurry, filter-sterilized slurry, or autoclaved water. Each soil treatment consisted of three replicates within a completely randomized experimental design, with each replicate comprising 9 pots. After 2 weeks, the Fpmd MR5 strain was inoculated onto the soil as described above. All plants were grown in controlled glasshouse conditions and watered regularly with sterile water. Moreover, a 2-mL aliquot of each compound solution was added to the rhizosphere soil twice weekly throughout the experiment. Disease severity was quantified at the endpoint using the DI method.

Isolation and identification of culturable rhizosphere bacterial isolates

Bacteria were isolated from the rhizosphere soil of RNAi plants at the end of the experiment. Isolation and identification were conducted according to previously described protocols [21]. Briefly, soil samples were suspended in sterile distilled water and serially diluted. Soil suspensions were serially diluted and plated on tryptic soy agar (TSA). Distinct colonies were purified and screened for antagonistic activity against fungal plant pathogens using the dual-culture technique [21]. Of the 88 isolates tested (with three replicates per strain), 30 exhibited clear inhibitory activity against Fpmd MR5. These antagonistic strains were further characterized via 16S rRNA gene sequencing. For taxonomic identification, we performed BLASTn alignment of the 16S rRNA sequences against the SILVA database to determine the closest matches (see Supplementary Table S16 for detailed results).

The identified representative isolates were purified on TSA medium, and individual colonies were selected for validation via Sanger sequencing using the bacterial primers 27 F and 1492R. Purified strains were stored in 30% glycerol at − 80 °C for long-term preservation. The obtained 16S rRNA gene sequences were clustered into operational taxonomic units (OTUs) at a > 99% similarity threshold and submitted to the NCBI database for taxonomic classification. Finally, phylogenetic analysis was conducted using MEGA v.7.0.26, and the resulting tree was visualized with FigTree v1.4.3.

Disease suppression capacity of responsive isolates

Malus hupehensis Rehd. seedlings were cultivated in pots containing 200 g of sterile soil (prepared as described earlier) amended with 10% orchard-enriched (OE) soil. After 30 days, they were inoculated with isolated bacterial strains to assess their potential as biocontrol agents. Isolates were cultured in 100 mL of tryptic soy broth (TSB) for 48 h at 28 °C with shaking at 170 rpm. The cultures were then centrifuged, washed, and resuspended in sterile 10 mM magnesium sulfate (MgSO₄) solution. The single strain was inoculated into pots at an initial density of 10⁷ cells mL⁻¹ growth substrate. The control pots received only a 10 mM MgSO₄ solution. Each treatment consisted of 30 replicate plants. Approximately 20 mL of the bacterial suspension was evenly applied to the growth medium surface in each pot. Ten days after bacterial inoculation, plants were challenged with Fpmd MR5 at a final concentration of 10⁵ conidia·g⁻¹ substrate. Rhizosphere samples were collected 14 days following pathogen inoculation for subsequent analysis.

Plant growth promotion and Fpmd MR5 inhibition capacity of the isolated strains

The plant growth-promoting (PGP) potential of the isolated strains was assessed following the method of Duan et al. [21]. Key PGP traits evaluated included: phosphate solubilization, potassium solubilization, nitrogen fixation, ferric siderophore production, and cell wall-degrading enzyme activity (cellulose activity). Additionally, IAA concentrations in the fermentation broth of the isolates were quantified using high-performance liquid chromatography–tandem mass spectrometry (HPLC–MS/MS) (analyzed by Nanjing Ruiyuan Biotechnology Co., Ltd., Nanjing, China).

The inhibitory effect of strains on Fpmd MR5 mycelial growth was tested as described by Duan et al. [21]. Hyphae from the edge of the inhibition zone were collected with an inoculation needle and fixed in 2.5% glutaraldehyde for 24 h. Control samples underwent identical treatment. All samples were processed for scanning electron microscopy (SEM) by Chengdu Lilai Biotechnology Co., Ltd. (China).

Effects of phloridzin and isolated strains on the growth and fumonisin B1 production of Fpmd MR5

Effects of phloridzin on the growth and fumonisin B1 production of Fpmd MR5

Fumonisin B1 (FB1) production by Fpmd MR5 was determined using an enzyme-linked immunosorbent assay (ELISA) method, following the procedure of Jurado et al. [47] with minor modifications. Fpmd MR5 was inoculated into 250 mL Erlenmeyer flasks containing 50 mL of fumonisin-inducing liquid medium (malt extract 0.5 g·L⁻¹, yeast extract 1 g·L⁻¹, peptone 1 g·L⁻¹, KH₂PO₄ 1 g·L⁻¹, MgSO₄·7H₂O 0.3 g·L⁻¹, KCl 0.3 g·L⁻¹, ZnSO₄·7H₂O 0.05 g·L⁻¹, CuSO₄·5H₂O 0.01 g·L⁻¹ and fructose 20 g·L⁻¹). The flasks were inoculated with 1% (v/v) of a Fpmd MR5 spore suspension (5 × 10⁶ spores·mL⁻¹), and phloridzin was added at concentrations of 0, 50, 100, 200, 400, 800, and 1000 μmol·L⁻¹. The cultures were incubated in the dark at 28 °C on a rotary shaker at 180 rpm for 7 days. After incubation, the cultures were filtered through Whatman No. 1 paper (Whatman International, Ltd., Kent, UK) to collect the mycelia. The mycelia were washed twice with sterile saline solution and then dried at 60 °C to constant weight to calculate toxin production per unit dry weight. Meanwhile, the culture filtrates were collected, and the FB1 content was determined. The ELISA kits (MEIMIAN, MM-9513701) were purchased from Jiangsu Meimian industrial Co., Ltd. (Nanjing, China). All experiments were performed with three replicates.

The effect of different concentrations of phloridzin on the mycelial growth of Fpmd MR5 was determined using the mycelial growth rate method [46]. Phloridzin (purity ≥ 98%, purchased from Yuanye Bio-Technology Co., Ltd., Shanghai, China) was dissolved in anhydrous ethanol to prepare a 1.0 mmol·L⁻¹ stock solution. The solution was sterilized by filtration through Millex^®-GP 0.22 μm membrane filters (Millipore, Billerica, MA, USA) and stored in the dark at − 20 °C. Prior to use, care was taken to ensure complete evaporation of the anhydrous ethanol. Activated Fpmd MR5 cultures were used to prepare 5 mm mycelial plugs. The plugs were placed in the center of PDA plates amended with a series of phloridzin concentrations (0, 50, 100, 200, 400, 800, and 1000 μmol·L⁻¹). Each treatment was replicated three times. All plates were incubated in the dark at 25 °C. On the 5th day of incubation, the colony diameter (mm) was measured using the cross method. The mycelial growth rate was calculated as the ratio of the net growth diameter to the culture time.

Inhibitory effects of isolated strains on Fpmd MR5 growth and FB1 production

To verify the strain deletion effects of the synthetic microbial community (SynCom), the following treatment groups were established in this experiment: G0, sterile Nutrient Broth (Solarbio) medium without any Bacillus strains (blank control); G1, complete SynCom comprising MdAsB52, MdAsB213, MdAsB22, MdAsB202, and MdAsB115; G2, community with MdAsB52 deleted; G3, community with MdAsB213 deleted; G4, community with MdAsB22 deleted; G5, community with MdAsB202 deleted; G6, community with MdAsB115 deleted; G7, MdAsB52; G8, MdAsB213; G9, MdAsB22; G10, MdAsB202; G11, MdAsB115. The preparation of seed cultures and cell-free supernatants (CFS) for Bacillus strains followed the method described by Wang et al. [91]. Seed cultures of individual strains were mixed in equal volumes (1:1), and the final cell density of all treatment groups was uniformly adjusted to 10⁸ cells·mL⁻¹.

For each treatment, 1 mL of seed culture and 1 mL of CFS were separately added to Petri dishes containing PDA medium and evenly spread on the surface using a sterile spreader. Subsequently, mycelial plugs (5 mm in diameter) were excised from the margin of an actively growing Fpmd MR5 colony using a sterile cork borer and placed at the center of each plate. The plates were incubated at 25 °C for 7 days. Fungal colony diameters were measured using the cross-cross method, and the fungal growth inhibition rate was calculated relative to the blank control (G0) following the method described by Bertuzzi et al. [4].

The inhibitory effect on FB1 synthesis was assessed in vitro following the method of Jannat et al. [43] with minor modifications. Briefly, 500 µL of seed culture (or CFS) from each treatment and 500 µL of Fpmd MR5 spore suspension were simultaneously inoculated into 250 mL Erlenmeyer flasks containing 50 mL of fumonisin-inducing liquid medium. The flasks were incubated in the dark at 28 °C with shaking at 180 rpm for 7 days. After incubation, FB1 content was determined following the method described in the “Effects of phloridzin on the growth and fumonisin B1 production of Fpmd MR5” section. All experiments were performed with three replicates.

Evaluate the chemotactic response of candidate Bacillus and their ability to utilize different compounds

The chemotaxis assay was performed using the quantitative capillary chemotaxis assay method following the procedure described by Reyes-Darias et al. [76]. Briefly, bacterial cells were resuspended in chemotaxis buffer (HEPES, pH = 7.0) to an OD₆₀₀ of 0.04–0.05. Capillary micropipettes, sealed at one end and filled with root exudate compounds (0.01, 0.1, 1 mM) or buffer (control), were inserted into the bacterial suspension and incubated for 30 min. The contents were then transferred into centrifuge tubes containing sterile 0.9% sodium chloride solution. The suspensions were diluted and spread onto plates for colony counting. Additionally, different concentrations of root exudate compounds were co-cultured with Bacillus strains (OD₆₀₀ = 0.05) in medium containing 1/10 TSB. After incubation at 37 °C with shaking at 180 rpm for 12 h, the OD₆₀₀ of the bacterial cultures was measured [45].

Validation of the role of SynCom in ARD suppression

In May 2025, a 90-day outdoor pot experiment was conducted to validate the role of a SynCom in ARD suppression at the Horticultural Experimental Station of Northwest A&F University, Yangling, China (34°20′N, 108°24′E). The study comprised 3 treatments: 30-year-old orchard soil (CK, control), an inactivated SynCom suspension (T1), and an active SynCom suspension (T2). Each treatment was replicated 20 times using a completely randomized design. The SynComs comprised five bacterial isolates (MdAsB52, MdAsB213, MdAsB22, MdAsB202, and MdAsB115) mixed in equal proportions (1:1 ratio) at a final density of 10⁸ cells·mL⁻¹. Malus hupehensis Rehd. seedlings were inoculated with SynComs at 2 × 10⁸ CFU·g⁻¹ substrate [100]. All plants received standardized watering and nutrient management following established protocols [21]. After 90 days, growth parameters (e.g., plant height, ground diameter, fresh weight, and dry weight) were measured from three randomly selected plants per treatment. From these same plants, rhizosphere samples were collected for pathogen density quantification via qPCR and microbial community analysis through 16S rRNA amplicon sequencing.

Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) analysis

Total RNA was extracted using a Foregene Plant RNA Isolation Kit (Foregenes, Chengdu, China). RT-qPCR analysis was carried out as previously described by Gao et al. [30], using the primers listed in Supplementary Table S2.

Statistical analyses

One-way analysis of variance (ANOVA) was used to compare treatment means using IBM SPSS Statistics v20.0 (IBM Corp., Armonk, NY, USA). Post-hoc comparisons were performed using either Duncan’s multiple range test or Tukey’s HSD test, with statistical significance set at P < 0.05. All figures were prepared using Microsoft Excel 2013 (Microsoft Corp., Redmond, WA, USA) and GraphPad Prism v8.0.2 (GraphPad Software, San Diego, CA, USA). Bioinformatics analysis was conducted using OECloud tools at https://cloud.oebiotech.cn. To evaluate the effects of the prior microbiome specificity and Fpmd MR5 inoculation, four networks were individually constructed for each treatment of MRT14. The analysis was limited to OTUs present in at least three samples comprising at least 14 reads, and with a relative abundance ≥ 0.01% in each treatment. The edges in the co-occurrence network represent statistically significant (false discovery rate [FDR] < 0.01) Spearman correlations with correlation coefficient values above 0.8 (r² > 0.8), while nodes represent individual taxa. Networks were visualized using Gephi [100].

Results

ARD tolerance of MdUGT88F1-RNAi plants

We first evaluated the tolerance of WT and MdUGT88F1-OE (hereinafter "OE") and MdUGT88F1-RNAi (hereinafter “RNAi”) to ARD, which was mainly associated with the accumulation of rhizosphere phloridzin and Fusarium spp. (Fig. 1a). RNAi plants exhibited significantly reduced plant height and aboveground biomass compared to OE and WT plants, while there was no significant difference in ground diameter (Supplementary Fig. S1b, c, e, f; Duncan’s multiple comparison test, P < 0.05). This may be attributed to the suppression of vertical growth (manifested as dwarfism and shortened internodes) resulting from MdUGT88F1 gene silencing. On the other hand, RNAi plants exhibited significantly greater total root length, root surface area, number of root tips, number of forks, and root vitality compared to OE and WT plants (Fig. S1d, g–l; Fig. S2e; Duncan’s multiple comparison test, P < 0.05). Compared to OE and WT plants, RNAi plants exhibited significantly reduced accumulation of H₂O₂ and O₂⁻ in their leaves (Supplementary Fig. S2a–d; Duncan’s multiple comparison test, P < 0.05), along with a notably lower MDA content (Supplementary Fig. S2f; Duncan’s multiple comparison test, P < 0.05). Additionally, they maintained higher levels of antioxidant enzyme activity (Supplementary Fig. S2g–j; Duncan’s multiple comparison test, P < 0.05). Photosynthetic pigment analysis revealed significantly higher chlorophyll and carotenoid contents in RNAi plants compared to the OE and WT plants (Supplementary Fig. S3a–d; Duncan’s multiple comparison test, P < 0.05). Furthermore, RNAi plants exhibited a greater capacity for light energy capture and thermal energy dissipation, which enhanced photosynthetic performance by maintaining photosystem II (PSII) utilization efficiency under ARD conditions (Supplementary Fig. S3e–m; Duncan’s multiple comparison test, P < 0.05). Mineral nutrient uptake, impaired by ARD, was also sustained in RNAi plants, as evidenced by higher levels of N, K, Ca, Fe, and Mg in their leaves and the highest N and Fe contents in their roots compared to OE and WT plants (Supplementary Table S3). Notably, compared to OE and WT plants, RNAi plants exhibited a significant increase in rhizosphere bacterial and actinomycete populations as well as soil bacterial/fungal ratios, along with a marked reduction in fungal abundance and Fusarium density (Supplementary Fig. S4a–d, k–p; Duncan’s multiple comparison test, P < 0.05). Soil enzyme activities and physicochemical properties were also improved (Supplementary Fig. S4e–j; Supplementary Table S4; Duncan’s multiple comparison test, P < 0.05), while contents of phenolic compounds such as phloridzin were significantly reduced (Supplementary Table S5; Duncan’s multiple comparison test, P < 0.05). These changes collectively established a more favorable environment for RNAi plant growth and enhanced resistance to ARD.

RT-qPCR analysis revealed significant downregulation of MdUGT88F1 and MdUGT88F4 gene expression in OE and WT plants (Supplementary Fig. S5b, c; Duncan’s multiple comparison test, P < 0.05). Intriguingly, in RNAi plants, reduced phloridzin biosynthesis enhanced disease resistance, as evidenced by the significantly lower DI and pathogen abundance compared to OE and WT plants (Fig. 2a, b; Duncan’s multiple comparison test, P < 0.05). Notably, a more pronounced increase in both DI and pathogen load was observed in OE plants compared to WT plants, without any significant growth-promoting effects (Supplementary Fig. S1b–l; Duncan’s multiple comparison test, P < 0.05). The abundance of Fpmd MR5 in the rhizosphere followed a trend similar to disease incidence (Supplementary Fig. S5a, Fig. 2b). Moreover, rhizospheric phloridzin contents were positively correlated with both pathogen density (Spearman, R = 0.8786, P < 0.0001) and DI (Spearman, R = 0.8853, P < 0.0001) (Supplementary Fig. S5d–e). Consistent with previous studies [104], the present study confirmed that phloridzin could simultaneously promote the mycelial growth and FB1 biosynthesis of Fpmd MR5 in a concentration-dependent manner (Supplementary Table S6; Supplementary Fig. S6).

Fig. 2.

Phenotypic and microbial responses of transgenic apple lines grown in replant soil. a DI (mean ± SD, n = 3) of transgenic apple lines and GL-3 grown in replant soil at 90 days post-transplantation. b Abundance of Fpmd MR5 in rhizosphere soil (mean ± SD, n = 3). c Phylogenetic diversity of the bacterial community (mean ± SD). Black dots represent individual replicate values. d Principal coordinate analysis (PCoA) of the bacterial community based on weighted UniFrac distances (P-values from PERMANOVA). e Top 20 bacterial families correlated with pathogen density, Bacillus abundance, or phloridzin levels (random forest model). Heatmap shows Spearman correlation coefficients (r). ‘MeanDecreaseGini’ denotes variable importance (higher values indicate stronger influence on the target). ANOVA with Duncan’s multiple range test; lowercase letters denote significant differences (P < 0.05). ‘ns’ = not significant (P > 0.05). WT, wild-type; OE2-7 and OE3-4 (OE), MdUGT88F1 overexpression lines; Ri-3 and Ri-6 (RNAi), MdUGT88F1 silencing lines

MdUGT88F1 was involved in microbe-mediated disease resistance

Microbial community profiling revealed significant compositional differences among treatments for both bacterial and fungal populations (Fig. 2d, PERMANOVA: F = 10.261, R² = 0.68, P = 0.001; Fig. S7a, PERMANOVA: F = 10.39, R² = 0.61, P = 0.001; Supplementary Note S1). However, sequencing of bacterial communities associated with bulk and rhizosphere soils revealed no significant differences in bacterial diversity across all treatments (Fig. 2c; Duncan’s multiple comparison test, P > 0.05). Volcano plot analysis revealed that under inoculation conditions, RNAi plant rhizosphere harbored a greater number of differentially abundant taxa compared to the WT rhizosphere (Supplementary Fig. S7b). Notably, the bacterial communities in the rhizosphere soil of RNAi plants exhibited more complex co-occurrence patterns, characterized by a greater number of nodes and edges and higher average degree (Supplementary Fig. S7c). Further taxonomic analysis revealed that under ARD conditions, the majority of responsive OTUs were enriched in RNAi plant rhizospheres (Supplementary Fig. S8a). Although these taxa displayed comparable relative abundances between OE and WT plants, they exhibited a sharp increase exclusively in RNAi plants (Supplementary Fig. S8b, two-sided Student’s t-test, P < 0.05). Random forest analysis identified the top five families potentially influencing rhizosphere pathogen abundance: Rhodanobacteraceae (MDG = 1.47), Comamonadaceae (MDG = 1.06), Rhizobiaceae (MDG = 0.93), Bacillaceae (MDG = 0.83), and Oxalobacteraceae (MDG = 0.65) (Fig. 2e). Notably, Bacillaceae abundance was negatively correlated with both pathogen density (Spearman, R = −0.7679, P = 0.0008) and phloridzin content (Spearman, R = −0.6607, P = 0.0073). Glasshouse experiments confirmed that Bacillaceae-related genera were the key responsive taxa exhibiting differential abundance (Supplementary Table S7). Pathogen inoculation significantly increased Bacillus abundance in the rhizosphere (Supplementary Fig. S8d, two-sided Student’s t-test, P < 0.05). Although no significant differences were observed in rhizosphere Bacillus abundance among plant lines, a higher Bacillus abundance was observed in RNAi plant rhizospheres (Supplementary Fig. S8c, Mann–Whitney U test, P < 0.01). Furthermore, rhizosphere Bacillus abundance was negatively correlated with both pathogen density (Spearman, R = − 0.6429, P = 0.0116) and phloridzin content (Pearson, R = −0.9260, R² = 0.8575, P < 0.0001) (Supplementary Fig. S8e, f).

Rhizosphere microbiota associated with OE and RNAi plants show apparent compositional and functional differences

After evaluating the disease suppressiveness of RNAi soil, our results indicated that this suppressiveness was microbiome-driven and could be transferred to conducive soils (Supplementary Fig. S11). To elucidate the contribution of rhizosphere microbiota to disease resistance, we employed metagenomics to explore microbiome differences between OE and RNAi plants. We found that the RNAi-associated microbiome differed significantly from the OE-associated microbiome both functionally (Fig. 3a, PERMANOVA: F = 4.74, R² = 0.32, P = 0.023; Supplementary Note S2) and taxonomically (Fig. 3b, PERMANOVA: F = 10.87, R² = 0.52, P = 0.001). The co-occurrence network analysis of the rhizosphere bacterial community in RNAi plants was consistent with the 16S rRNA gene amplification sequencing results (Fig. 3d). Notably, the relative abundances of Actinobacteriota, Gemmatimonadota, and Acidobacteriota were significantly lower in OE plants compared to RNAi plants (Fig. 3c, P = 0.011, t = 3.389, df = 7.363, two-sided Student’s t-test). Volcano plot analysis identified 305 microbial taxa with significantly higher abundance in the RNAi plant rhizosphere compared to OE under inoculation conditions (Supplementary Fig. S12a). Among these taxa, the Streptomycetaceae (specifically Streptomyces) (Supplementary Fig. S12b, P < 0.001, t = 15.024, df = 6.183, two-sided Student’s t-test) and Bacillaceae (specifically Bacillus) (Supplementary Fig. S12b, P < 0.001, t = 6.123, df = 10, two-sided Student’s t-test) were significantly enriched in RNAi plant rhizospheres (Supplementary Fig. S12c). Furthermore, the relative abundance of Actinobacteria (specifically Streptomyces) was negatively correlated with pathogen density, while showing a positive correlation with Bacillus abundance in the rhizosphere (Supplementary Fig. S12d).

Fig. 3.

MdUGT88F1-mediated differences in rhizosphere bacterial community composition based on metagenomic sequencing data. a PCoA indicating significant divergence in rhizosphere bacterial communities between OE and RNAi plants at 90 days post-transplantation (P-values from PERMANOVA). b Differences in functional gene composition of the rhizosphere microbiome between OE and RNAi plants (PCoA; P-values from PERMANOVA). c Differences in OE and RNAi plant rhizosphere bacterial communities in terms of relative phyla abundances (arrows indicate direction of change; two-sided Student’s t-test). d Co-occurrence network analysis of the top 500 OTUs from the rhizosphere of OE and RNAi plants under ARD conditions. OE2-7 and OE3-4 (OE), MdUGT88F1 overexpression lines; Ri-3 and Ri-6 (RNAi), MdUGT88F1 silencing lines. In c, *P < 0.05, **P < 0.01, ***P < 0.001

Community shifts of rhizosphere bacteria and fungi under Fpmd MR5 inoculation in pot experiments

To understand the mechanisms and processes underlying rhizosphere microbiota suppression of Fpmd MR5 inoculation, we directly inoculated Fpmd MR5 into apple plant rhizospheres and subsequently sequenced the rhizosphere microbial communities (Fig. 1b; Supplementary Note S3). We found that Fpmd MR5 inoculation resulted in significant DI differences between OE and RNAi plants (Fig. 4a, P = 0.049, t = 2.788, df = 4, two-sided Student’s t-test; Supplementary Fig. S13). Although Fpmd MR5 abundance increased in the rhizospheres of both genotypes, it remained significantly lower in RNAi than in OE plants, indicating enhanced microbial suppressiveness, consistent with the pot experiment results. Interestingly, although we did not observe significant differences in Fpmd MR5 abundance in bulk soil, after 14 days, its abundance in the RNAi rhizosphere was lower than that in the OE rhizosphere (Fig. 4b, P = 0.013, t = 4.286, df = 4, two-sided Student’s t-test). Before inoculation, there were no significant differences in bacterial phylogenetic diversity (PD) between RNAi and OE rhizosphere soils. However, bacterial alpha diversity increased in both treatments following inoculation. Despite this response, no significant difference in PD was observed between RNAi and OE plant rhizospheres inoculated with Fpmd MR5 at day 14 post-inoculation (Fig. 4c, Tukey’s test).

Fig. 4.

Microbial community traits of rhizosphere bacteria under Fpmd MR5 inoculation. a DI at 14 days post-inoculation (mean ± SD; n = 3). b Fpmd MR5 abundance in rhizosphere soils. c Phylogenetic diversity of the bacterial community (mean ± SD; n = 3). d Principal coordinate analysis (PCoA) of the bacterial community in MRT14 based on weighted UniFrac distances (P-values from PERMANOVA). e Volcano plot of differentially abundant bacterial OTUs in OE and RNAi lines following Fpmd MR5 inoculation (MRT14). f Co-occurrence network of the bacterial community in MRT14. Bulk: initial soil mixture before planting; MRT0: rhizosphere before Fpmd MR5 inoculation; MRT14: rhizosphere 14 days after Fpmd MR5 inoculation. HMO, OE microbiome (− MR5); HMR, RNAi microbiome (− MR5); FMO, OE microbiome (+ MR5); FMR, RNAi microbiome (+ MR5). ‘–MR5’: non-inoculated control; ‘ + MR5’: Fpmd MR5-inoculated. Black dots represent individual replicate values. ‘ns’ indicates no significant difference (two-sided Student’s t-test, P > 0.05). In b, *P < 0.05, **P < 0.01, ***P < 0.001. Different lowercase letters indicate significant differences (Tukey’s test, P ≤ 0.05)

Based on both weighted and unweighted UniFrac distances, PCoA revealed that temporal shifts in community composition were greatest along the first PCoA axis (Supplementary Fig. S14c, PERMANOVA: F = 23.82, R² = 0.91, P = 0.001; Supplementary Fig. S15b). Then, based on the unweighted UniFrac distance, variation among community origins was observed along the second PCoA axis (Supplementary Fig. S14b, PERMANOVA: F = 1.3, R² = 0.43, P = 0.001). To assess the impact of inoculation, we focused on the 14th day after inoculation with Fpmd MR5, revealing that both community origin and inoculation significantly influenced apple rhizosphere bacterial composition (Fig. 4d, PERMANOVA: F = 1.82, R² = 0.73, P = 0.001; Supplementary Fig. S14a, PERMANOVA: F = 4.74, R² = 0.41, P = 0.001). Under pathogen inoculation, the distance shift in rhizosphere bacterial communities was greater in RNAi than in OE (Supplementary Table S10). Volcano plot analysis revealed a greater number of differentially abundant taxa in RNAi rhizospheres compared to OE under inoculation (Fig. 4e). Furthermore, network analysis indicated that Fpmd MR5 inoculation increased the number of nodes and edges in both RNAi and OE rhizosphere bacterial communities; however, a more pronounced change was observed in RNAi communities (Fig. 4f). Further taxonomic analysis revealed that under pathogen inoculation, the majority of responsive OTUs were enriched in RNAi + Fpmd MR5 rather than in OE + Fpmd MR5 (Supplementary Fig. S15a). Although the relative abundances of these taxa were similar between OE and RNAi before inoculation with Fpmd MR5, a sharp increase was only observed in RNAi rhizospheres after inoculation (Supplementary Fig. S15b). Based on the results of the glasshouse pot experiment, genera associated with Bacillaceae were identified as key responsive taxa (Supplementary Table S11–S12). Random forest modeling highlighted Bacillaceae as key predictors of Fpmd MR5 suppression (Supplementary Fig. S16a). Fpmd MR5 inoculation significantly increased Bacillus abundance (Supplementary Fig. S16b, Mann–Whitney U test, P = 0.004). Although no significant difference was observed between OE and RNAi, a higher density of Bacillus was detected in RNAi rhizospheres (Supplementary Fig. S16b). Additionally, Streptomyces relative abundance was negatively correlated with Fpmd MR5 levels (Supplementary Table S11). However, Streptomyces abundance significantly decreased after Fpmd MR5 inoculation (Supplementary Fig. S16c, Mann–Whitney U test, P = 0.004).

Functional response of rhizosphere bacteria under Fpmd MR5 inoculation

To investigate the functional implications of microbial compositional shifts, we conducted whole metagenome sequencing of several samples (Fig. 1b). Genes retrieved from the metagenomic data were annotated using the eggNOG and KEGG databases. Our results demonstrated stronger functional and community responses in RNAi rhizospheres compared to OE (Supplementary Fig. S19), thereby validating the findings derived from 16S rRNA gene amplicon sequencing (Fig. 4). Specifically, under Fpmd MR5 inoculation, the direction of functional changes in the rhizosphere was similar between RNAi and OE, but the magnitude of change was greater in RNAi than in CF (Fig. 5, two-sided Student’s t-test, P ≤ 0.05). Among these pathways, three (eggNOG) and five (KEGG) pathways were significantly enriched in the rhizosphere of FMR compared to FMO, including carbohydrate transport and metabolism, defense mechanisms, membrane transport, and carbohydrate metabolism. Further secondary metabolites analyses using ANTISMASH also revealed significant enrichment of type II polyketide biosynthesis and polyketide sugar-unit pathways in the rhizosphere under FMR compared to FMO (Supplementary Fig. S20, two-sided Student’s t-test, P ≤ 0.05).

Fig.5.

Functional changes in rhizosphere bacterial communities in response to Fpmd MR5 inoculation. a Gene abundances based on the eggNOG database: Relative abundance (%) of annotated genes; Fold changes (pathogen-inoculated vs. non-inoculated); Heatmap of pathway abundances across treatments. b Gene abundances derived from the KEGG database: relative abundance (%) of annotated genes; fold changes (pathogen-inoculated vs. non-inoculated); Heatmap of pathway abundances (Z-score scale − 1.0 [blue] to 1.0 [red]). HMO, OE microbiome (−MR5); HMR, RNAi microbiome (−MR5); FMO, OE microbiome (+MR5); FMR, RNAi microbiome (+MR5). ‘−MR5’: non-inoculated control; ‘+MR5’: Fpmd MR5-inoculated. Asterisks (*) indicate significant enrichment in FMR relative to FMO (two-sided Student’s t-test, *, P ≤ 0.05). The hash symbol indicates a significant change in the +MR5 rhizosphere compared to its corresponding −MR5 rhizosphere (two-sided Student’s t-test, #, P ≤ 0.05)

MdUGT88F1-mediated suppressive microbiota was associated with varied root exudates

To elucidate the mechanistic basis underlying the distinct microbiota composition and elevated Bacillus abundance observed in RNAi plants, we investigated alterations in apple root exudate profiles. It is well established that root exudates connect plants with their associated microorganisms and drive the recruitment of suppressive microbial communities [100, 106]. RNAi plants exhibited significantly altered root exudation patterns (Fig. 6c, PERMANOVA: F = 15.00, R² = 0.41, P = 0.001; Supplementary Fig. S22c, PERMANOVA: F = 13.78, R² = 0.46, P = 0.001). Specifically, 18.33% of the detected metabolites were secreted at higher levels in RNAi plants (N = 60; Fig. 6a; Supplementary Table S13). The exudates of RNAi plants contained higher abundances of fatty acyls (P < 0.001, t = 30.641, df = 17.481, two-sided Student’s t-test) and organooxygen compounds (P < 0.001, t = 5.665, df = 15.550, two-sided Student’s t-test), but lower abundances of carboxylic acids and derivatives (P = 0.001, t = 4.327, df = 16.106, two-sided Student’s t-test) and organonitrogen compounds (P = 0.009, t = 2.914, df = 17.858, two-sided Student’s t-test) (Fig. 6b). In OE plants, l-alanine, l-leucine, l-valine, glycerol, and DL-xylose were more abundant in the root exudates (Fig. 6d, Supplementary Table S13, Supplementary Notes 4 and 5). In contrast, RNAi plants demonstrated particularly pronounced and consistent increases in the exudation of D-tagatose, D-galactose, sucrose, 3-O-methyl-D-glucose, and maltitol (Fig. 6d, Supplementary Table S13). KEGG enrichment bubble plot analysis revealed that the significantly enriched metabolic pathways in RNAi plants comprised galactose metabolism, phosphotransferase system (PTS), and carbohydrate digestion and absorption. The key metabolites that are part of these pathways were D-tagatose, D-galactose, and sucrose (Fig. 6e), suggesting their potential association with the recruitment of disease-suppressive microbiota and the activation of systemic resistance.

Fig. 6.

Impacts of MdUGT88F1 on root exudation profile at 90 days post-transplantation. a Volcano plot displaying differentially abundant root exudate metabolites in RNAi vs. OE plants (|log2FoldChange|> 0.263 and q < 0.05). Orange = overrepresented, blue = underrepresented, gray = unchanged metabolites in RNAi. b Differential accumulation of root exudate metabolite classes between OE and RNAi plants (two-sided Student’s t-test). Metabolite peak areas were normalized to the internal standard and expressed as relative abundance per gram root dry mass. c PCA of root exudate metabolites between OE and RNAi plants under ARD conditions (P-value from PERMANOVA). d Heatmap of exudate metabolites significantly reduced in OE vs. RNAi plants. Metabolites selected for validation are bolded. e Bubble plot of pathways upregulated in RNAi rhizospheres. In b, *P < 0.05, **P < 0.01, ***P < 0.001

Root exudatesare involved in disease suppression via the recruitment of disease-suppressing Bacillus species

Root exudate composition is known to influence rhizosphere microbiota and host health [29, 106]. To better understand the impact of differentially released compounds on soil microbiota, we conditioned soils with mixtures enriched or depleted in metabolites characteristic of RNAi plant exudates (Fig. 1c). For this purpose, we prepared two distinct exudate mimetics: FCs (fatty acids and carbohydrates), representing compounds significantly enriched in RNAi plant rhizospheres, and ACs (amino acids and carbohydrates), representing compounds that were reduced (Fig. 6d). Soil extracts were prepared from these preconditioned soils and used to inoculate Malus hupehensis Rehd. seedlings grown in sterilized sand–vermiculite mixtures, followed by inoculation with Fpmd MR5. Plants grown in FCs-conditioned soils exhibited enhanced resistance to Fpmd MR5 compared to those in control or ACs-conditioned soils. Although slight ruptures and deformations were observed in root epidermal and cortical cells, the internal tissue structure remained intact (Fig. 7a, Tukey's test, P ≤ 0.05; Supplementary Fig. S23a, c). Crucially, soil extracts subjected to bacterial filtration lost their ability to induce disease resistance (Fig. 7b, Tukey's test, P > 0.05; Supplementary Fig. S23a), demonstrating that the microbiome, rather than the exudate compounds themselves, triggered resistance in plants.

Fig. 7.

Apple root exudates mediate the recruitment of pathogen-suppressing Bacillus. a DI of Fpmd MR5-challenged apple plants grown on sterilized artificial soil inoculated with water or slurries of soils preconditioned with mixtures of ACs or FCs. b DI of Fpmd MR5-challenged apple plants grown on sterilized artificial soil inoculated with filter-sterilized water, ACs, or FCs. c DI of Fpmd MR5-challenged apple plants grown on sterilized artificial soil inoculated with water or FCs and ACs mixed in a ratio of 1:9, 5:5, or 9:1 (v/v). d DI of Fpmd MR5-challenged apple plants grown on sterilized artificial soil inoculated with water or filter-sterilized FCs and ACs mixed in a ratio of 1:9, 5:5, or 9:1 (v/v). e Abundance of Fpmd MR5 in rhizosphere soil inoculated with (left) water or slurries of soils preconditioned with mixtures of ACs or FCs, or (right) filter-sterilized water, ACs, or FCs. f Abundance of Fpmd MR5 in rhizosphere soil inoculated with (left) water or FCs and ACs mixed in ratios of 1:9, 5:5, or 9:1 (v/v), (right) water or filter-sterilized FCs and ACs mixed in the same ratios. g Numbers and heatmap illustrating the effect of culturable isolates on Fpmd MR5 pathogen growth inhibition in laboratory assays (mean, n = 3) and on disease suppression in a glasshouse experiment with apple plants relative to a non-Bacillus-inoculated control treatment (mean, n = 3). Bars show average ± SD of 3 replicates. Different lowercase letters indicate significant differences (Tukey’s test, P ≤ 0.05). “ns” denotes no significant difference (Tukey’s test, P > 0.05). Black dots represent individual replicate values

We further conducted soil transfer experiments to evaluate the required microbial inoculum for conferring resistance. To this end, different proportions (10%, 50%, and 90% w/w) of FCs-conditioned soil were mixed with ACs-conditioned soil before preparing soil suspensions. Incorporation of 50% or 90% FCs-conditioned soil conferred partial disease suppression (Fig. 7c, Tukey’s test, P ≤ 0.05; Supplementary Fig. S23b). In contrast, all filtered, microbe-free soil suspensions failed to provide disease suppression (Fig. 7d, Tukey’s test, P > 0.05; Supplementary Fig. S23b), with disease alleviation correlating with reduced pathogen density (Fig. 7e, f, Tukey’s test, P ≤ 0.05). Notably, RNAi exudates promoted a significantly greater increase in Bacillus abundance relative to OE exudates (Supplementary Fig. S24a, b, Tukey’s test, P ≤ 0.05). To test this directly, we compared the effects of FCs and ACs on apple rhizosphere microbiome composition under laboratory conditions. Conditioning native soil microbiomes with FCs significantly altered bacterial community structure (Supplementary Fig. S24c, PERMANOVA: F = 11.18, R² = 0.79, P = 0.005), leading to increased Bacillus relative abundance (Supplementary Fig. S24d, e; Supplementary Table S14). Collectively, these results suggest that MdUGT88F1 may influence soil microbiome-mediated disease resistance and the relative abundance of Bacillus by regulating the secretion of FCs. To test this hypothesis, the isolated Bacillus strains were co-cultured with root exudate compounds. The results showed that supplementation with FCs components significantly increased the OD₆₀₀ value of Bacillus, indicating that FCs may be utilized by Bacillus. Meanwhile, quantitative capillary chemotaxis assays revealed that five Bacillus strains exhibited a distinct dose-dependent chemotactic response to FCs components in the root exudates, while a slight repellent response was observed toward ACs components (Supplementary Fig. S25; Supplementary Table 15).

Pathogen suppressiveness of the responsive taxa

To functionally validate the microbiome analysis results, we isolated 88 distinct culturable bacterial strains from the rhizosphere soil of RNAi plants (Supplementary Fig. S26). Among these isolates, 30 strains significantly inhibited Fpmd MR5 growth in laboratory plate assays, causing characteristic morphological alterations in the pathogen, including irregularly reticulated mycelium, uneven thickness, shrinkage, twisting, swelling, thinning, and fragmentation with cellular content leakage, and spore wall deformation (Supplementary Table S16, Supplementary Fig. S27). Notably, all strains causing Fpmd MR5 growth inhibition possessed one or more PGP traits (phosphate solubilization, nitrogen fixation, potassium solubilization, siderophore production, cellulose degradation, and IAA biosynthesis) (Supplementary Table S16). Crucially, 27 of these strains (90%) significantly reduced disease severity in greenhouse-grown apple plants (Supplementary Table S16; Fig. 7f).

In this study, five Bacillus strains with plant growth-promoting (IAA > 10 μg·mL⁻¹) and antagonistic (inhibition rate > 50%; DI < 30%) properties were selected to construct a SynCom (Supplementary Fig. S28) for subsequent experiments. The selected strains included MdAsB52, MdAsB213, MdAsB223, MdAsB202, and MdAsB115. In vitro antagonistic assays showed that the original culture broth from all treatments completely inhibited the growth of the pathogen Fpmd MR5, while the CFS also exhibited antagonistic activity, suggesting that metabolites may contribute to the inhibitory effect. Among all treatments, the complete SynCom G1 exhibited the most pronounced inhibitory effect, with a mycelial growth inhibition rate of 79.74% and a 50.15% reduction in FB1 content compared to the control group (G0). Communities lacking any single strain showed intermediate inhibitory effects, while among the individual strains, MdAsB213 and MdAsB202 demonstrated relatively stronger antagonistic activity (Supplementary Table S17–S18; Supplementary Fig. S29). Pot experiments further confirmed that the SynCom could significantly mitigate ARD damage (Supplementary Fig. S30–S31). Compared to the T1 treatment (inactivated SynCom), the T2 treatment (SynCom) significantly promoted plant growth (Fig. S30a–e) and reduced Fusarium abundance in the rhizosphere (Supplementary Fig. S30f–j, Tukey’s test, P ≤ 0.05). Moreover, the SynCom significantly altered the structure of the rhizosphere bacterial microbiome (Supplementary Fig. S31a, PERMANOVA: F = 8.30, R² = 0.73, P = 0.003; Supplementary Table S19), leading to an increase in alpha diversity compared to the T1 treatment (Supplementary Fig. S31b, Tukey’s test, P ≤ 0.05) and to a significantly increased Bacillus relative abundance (Supplementary Fig. S31c–f, Tukey’s test, P ≤ 0.05). Network analysis indicated that the number of nodes and edges in the rhizosphere bacterial community increased following inoculation with the SynCom (Supplementary Fig. S31g).

Discussion

In China, the reconstruction of aging orchards in major apple-producing areas is often accompanied by a high incidence of soilborne diseases, which severely affects apple quality and yield [3, 20, 22, 82, 83]. In recent years, the use of biocontrol agents to enhance the pathogen-suppressive capacity of soil and induce systemic resistance in plants has emerged as an important strategy for promoting sustainable agricultural development [18, 74]. This study aimed to investigate the potential mechanisms underlying the resistance of RNAi plants against ARD. The results indicated that RNAi plants regulated the composition of root exudates (mainly including D-tagatose, D-galactose, sucrose, 3-O-methyl-D-glucose, and maltitol), thereby significantly enriching beneficial microorganisms such as Bacillus. The recruited Bacillus exhibited multiple PGP potentials and significantly inhibited the hyphal growth and FB1 production of Fpmd MR5. Furthermore, a SynCom composed of five Bacillus effectively reduced the abundance of pathogenic Fusarium in the rhizosphere, promoted plant growth, and alleviated ARD symptoms under pot conditions. In summary, this study reveals a mechanism by which MdUGT88F1 enhances plant resistance through regulating rhizosphere metabolites and reshaping the rhizosphere microbial community.

The rhizosphere is a critical zone for plant pathogen infection, and the assembly of its microbiome is jointly influenced by dispersal, drift, speciation, and plant selection [3, 18]. Rhizosphere microbial communities characterized by high diversity, beneficial microorganisms, and strong interactions are generally considered to have the potential to suppress pathogen colonization. The observation of lower pathogen abundance in disease-suppressive soils also provides evidence supporting the involvement of rhizosphere microbes in the suppression process [40, 49, 92, 102]. It has been suggested that agricultural practices which enhance soil biodiversity may strengthen the stability of microbial communities and their ability to resist pathogen invasion, a phenomenon consistent with the “diversity resistance” hypothesis [15, 59]. Notably, in this study, although the composition of the bacterial community changed significantly under different treatments, the MdUGT88F1 treatment did not significantly alter the overall diversity of the rhizosphere bacterial communities. This observation suggests that the suppression of soilborne diseases may depend more on the structural composition and functional traits of specific microbial groups than on an increase in overall diversity levels.

Many studies have suggested that specific suppressive rhizosphere microbial communities are capable of reducing disease incidence and Fusarium load, an effect potentially associated with intricate microbial interaction networks and distinct functional traits [28, 32, 95]. In this study, the bacterial communities in the rhizosphere soil of RNAi plants exhibited more complex co-occurrence patterns, characterized by a greater number of nodes and edges, and higher average degree. These topological features are consistent with those commonly observed in disease-suppressive soils, implying that the microbial community in RNAi soils was more interconnected, and its structure was associated with potentially stronger competitive ability against pathogens [95, 100]. Notably, correlation-based co-occurrence network analysis has inherent limitations. As emphasized by Freilich et al. [27], the co-occurrence patterns observed in this study do not represent definitive ecological interactions, but may reflect combined effects of environmental filtering, dispersal processes, and, to some extent, biological interactions (particularly positive non-trophic interactions). Therefore, further experimental validation (e.g., strain co-culture and synthetic microbial community inoculation) is still needed to verify the causal relationships between specific microbial interactions and soil disease suppressiveness.

Microbial secondary metabolites play an important defensive role in plant–pathogen interactions. When plants are infected by pathogens, genes involved in the biosynthesis of secondary metabolites, such as non-ribosomal peptide synthetase (NRPS) and polyketide synthase (PKS) genes, are upregulated in the rhizosphere microbial community [18], leading to the synthesis of small-molecule compounds with diverse biological activities [64, 65]. These metabolites can directly inhibit pathogen growth; for example, the polyketide antibiotic macrolactin produced by Bacillus exhibits broad-spectrum antimicrobial activity [105]. On the other hand, they may also indirectly enhance plant disease resistance by modulating microbial community structure [8, 32, 110]. Actinobacteria, particularly Streptomyces, may play a role in microbe-mediated disease resistance due to their relatively strong capacity for synthesizing secondary metabolites [14, 39, 65, 95]. In this study, we found that under ARD conditions, the roots of RNAi plants enriched potentially beneficial microorganisms, including Bacillaceae and Streptomycetaceae, which was accompanied by enhanced biosynthesis of PKS and other secondary metabolites. However, following infection by Fpmd MR5, the relative abundance of Bacillus in the rhizosphere increased, while that of Streptomyces decreased. This shift may be associated with the inhibitory effects of antimicrobial compounds produced by Bacillus on certain actinomycetes, including strains expressing PKS genes [2, 65]. Taken together, these observations suggest that secondary metabolites may play a functionally significant role in plant–microbe interactions and plant disease resistance; however, their specific mechanisms and causal relationships require further validation.

During the occurrence of ARD, phloridzin, as a phenolic compound, accumulates in the soil and can disrupt the plant’s antioxidant defense system, leading to the accumulation of reactive oxygen species (ROS), which in turn affects normal plant growth [88, 112]. Research indicates that in MdUGT88F1-RNAi plants, the activities of antioxidant enzymes (such as SOD, POD, and CAT) are enhanced, which helps alleviate the accumulation of ROS and maintain the relative stability of cell membrane structure [11]. Meanwhile, the PSII function in RNAi plants remains relatively well-preserved, suggesting a certain tolerance to damage to the photosynthetic apparatus [60]. In addition, certain pathogenic microorganisms (such as Erwinia amylovora, Valsa mali, and Fusarium verticillioides) possess the ability to degrade phloridzin, and their metabolites may exert toxic effects on plants, thereby weakening the host’s basal resistance [31, 94, 104, 112]. Previous studies have found that silencing MdUGT88F1 reduces phloridzin synthesis and activates defense mechanisms such as lignin accumulation, thereby enhancing plant resistance to Valsa mali [112]. The results of this study further support and extend the aforementioned findings. Under ARD conditions, we observed that downregulation of MdUGT88F1 reduced the accumulation of phloridzin in the rhizosphere, decreased the relative abundance of Fusarium, and correspondingly alleviated the severity of the disease. In vitro experiments further confirmed that phloridzin promotes the mycelial growth and FB1 synthesis of Fpmd MR5 in a concentration-dependent manner, suggesting that phloridzin may directly participate in regulating the growth and mycotoxin production of the pathogen by acting as a signaling molecule or nutrient source.

The rhizosphere, as a micro-domain environment directly regulated by plant root exudates, can account for up to 11% of fixed carbon input, forming an important “hotspot” for microbial activity [85]. Changes in the composition and abundance of root exudates can modulate the interactions among plants, pathogens, and the microbiome [23, 106]. These exudates not only provide carbon and energy for microorganisms [58, 96] but may also participate in the induction of plant resistance by recruiting specific beneficial microbial groups. For instance, substances such as long-chain organic acids, amino acids, phytohormones, coumarins, 3-hydroxyflavone, and riboflavin have been observed to exhibit similar functions in related studies [18, 24, 79, 106]. On the other hand, the accumulation of certain exudates, such as phloridzin, may be associated with the accelerated growth of pathogenic fungi (such as Fusarium), thereby exacerbating plant damage in contexts such as replant disease [104], Supplementary Table S6). In addition, the chemotactic effect of root exudates on beneficial microorganisms may influence their colonization efficiency in the rhizosphere, and chemotaxis itself is closely associated with functions such as nutrient acquisition, biofilm formation, and pollutant degradation [26, 98]. In this study, MdUGT88F1-RNAi plants significantly enriched microorganisms of the Bacillaceae family by reshaping the composition of root exudates (including D-tagatose, D-galactose, sucrose, 3-O-methyl-D-glucose, and maltitol), thereby enhancing resistance to Fpmd MR5. In vitro experiments showed that Bacillus exhibited a significant dose-dependent chemotactic response to the FCs in root exudates, while showing a slight repellent response to the avoidance compounds (ACs). Furthermore, the addition of FCs components significantly increased the OD₆₀₀ value of Bacillus, similar to the way lysine in apple rootstock root exudates attracts beneficial Bacillus [45]. Previous studies have indicated that these exudate components promote the colonization and function of beneficial bacteria through multiple mechanisms: D-tagatose can both disrupt pathogen metabolism [9, 10, 16] and serve as a chemotactic signal mediating the bidirectional recruitment between plants and Bacillus [10, 57] sucrose can activate signaling pathways to promote the colonization of Bacillus subtilis [84, 107]; 3-O-methyl-D-glucose may enhance the efficiency of microbial organic matter utilization by mimicking glucose [61]; maltitol can both promote the colonization and biofilm formation of B. velezensis and be synthesized by some Bacillus strains themselves [34, 56]. Notably, the pathogenic Fusarium exhibits a weak ability to utilize these enriched sugars and relies more on D-Xylose, glycerol, and amino acids (e.g., l-valine, l-leucine, and l-alanine) [42, 50, 53, 80, 101, 109], suggesting that changes in exudates may inhibit pathogen proliferation through nutrient competition. Furthermore, saturated fatty acids, such as palmitic acid, secreted by Bacillus, can directly inhibit the mycelial growth and spore germination of pathogenic fungi [11, 55]. In summary, MdUGT88F1-RNAi plants can modulate the composition of root exudates, thereby enriching beneficial Bacillus and potentially generating a synergistic effect through their antifungal functions, contributing to the ecological control of ARD.

The bidirectional interaction between plants and rhizosphere microbial communities is considered to have important implications for plant health and agricultural productivity [15, 87]. Typical symbiotic relationships, such as the symbiotic nitrogen fixation between legumes and rhizobia, support 40–50% of the global biological nitrogen fixation [86]. Mycorrhizal fungi can also positively influence plant growth by promoting nutrient uptake [89]. The composition and stability of rhizosphere microbial communities are closely linked to plant health and have therefore received widespread attention in agricultural research [58]. Previous studies have shown that plant species and genotypic differences can shape the structure of rhizosphere microbial communities to a certain extent [75, 100]. These microorganisms may participate in plant growth and disease resistance processes by secreting plant hormones, promoting nutrient uptake, and modulating plant immunity [25, 74]. Plants may also influence the formation of suppressive soil microbiomes by modulating the rhizosphere environment, thereby defending against pathogen infection [18, 64]. Under biotic stress conditions, the succession of rhizosphere microbial communities tends to develop towards a disease-suppressive state [77, 99]. In this study, RNAi plants were able to enrich microbial taxa such as Bacillus by regulating the components of root exudates and restore the original disease-suppressive function of the OE plant microbiome. This suggests that plants may indirectly enhance their resilience to environmental disturbances such as pathogen infection by regulating the pre-assembly mechanism of rhizosphere microorganisms. After inoculation with Fpmd MR5, the relative abundance of Bacillaceae in the rhizosphere microbial community increased. Pot experiments further demonstrated that the enriched Bacillus could reduce the abundance of rhizosphere Fusarium and alleviate ARD symptoms. Although the enrichment of Bacillus is associated with disease suppression to some extent, the synergistic effects of other microorganisms should also be considered.

As common functional microorganisms in microbial formulations, Bacillus can effectively colonize plant roots and may influence plant immune responses through niche competition, antimicrobial compound production, and induced systemic resistance [21, 25, 56, 74]. At the same time, they may also participate in processes such as nutrient solubilization and hormone synthesis, indirectly improving the soil environment and alleviating soilborne diseases [21, 25]. Numerous studies have demonstrated that various Bacillus, such as Bacillus subtilis, Bacillus pumilus, Bacillus amyloliquefaciens, and Bacillus megaterium, possess the potential for biological degradation of mycotoxins [4, 36, 43, 70]. Fusarium, especially Fusarium verticilliode, Fusarium proliferatum, and related Fusarium spp., produces FB1, FB2, and FB3, with FB1 being the most abundant and toxic among them [1]. In this study, through synthetic community depletion assays, it was observed that the complete Bacillus community (G1) exhibited a certain inhibitory effect on the hyphal growth and FB1 production of Fpmd MR5. The removal of any individual strain from the community resulted in varying degrees of reduction in this inhibitory effect, suggesting potential functional synergy among different strains within the community [44]. This finding still requires further validation under field conditions. Additionally, the study showed that the CFS derived from the complete Bacillus community could also inhibit pathogen growth and toxin accumulation to some extent, indicating that its metabolites may play a significant role in the antagonistic process. In summary, the reduction effect of Bacillus on FB1 may be achieved through multiple mechanisms, including inhibiting pathogen growth, interfering with toxin synthesis pathways, or directly degrading FB1. However, the specific mechanisms and regulatory details underlying these effects still require further investigation.

Conclusions

In summary, MdUGT88F1-RNAi plants can establish a disease-suppressive environment by reshaping bacterial community composition, with suppression primarily driven by broader changes in microbial community structure rather than solely by the abundance of introduced biocontrol strains (Fig. 8). Notably, members of the Bacillaceae family with PGP potential were selectively enriched in the rhizosphere under pathogenic fungi pressure. Soil pretreatment with “prebiotic” (FCs) facilitates the assembly of microbial communities capable of responding to biological disturbances in the apple rhizosphere. Experiments demonstrated that the targeted introduction of SynComs (Bacillus spp.) not only effectively promoted plant growth but also significantly reduced the abundance of pathogenic Fusarium, alleviating ARD. These findings advance our understanding of the complex interaction mechanisms between the rhizosphere microbiome and host plants and reveal the key mechanisms by which the MdUGT88F1 gene regulation drives disease suppression in the rhizosphere. Future research should focus on validating the effectiveness of SynComs in controlling ARD under field conditions and advancing their commercial potential. Additionally, exploring whether MdUGT88F1 affects the biosynthesis of plant metabolites, which may impact the composition and recruitment of endophytic bacterial communities, represents an interesting direction for future research.

Fig. 8.

A summary schematic of the role of MdUGT88F1 in plant immunity via root exudation-mediated recruitment of disease suppressive microbiota

Abbreviations

Fpmd MR5

Fusarium proliferatum MR5 f.sp. malus domestica

PDB

Potato dextrose broth

PDA

Potato dextrose agar

PBS

Phosphate-buffered saline

TSA

Trypticase soy agar

16S rRNA

16S ribosomal RNA gene

GC-MS

Gas chromatography–mass spectrometry

DI

Disease intensity

ARD

Apple replant disease

Qpcr

Quantitative polymerase chain reaction

KEGG

Kyoto Encyclopedia of Genes and Genomes

NR

Non-redundant protein database

eggnog

Evolutionary Genealogy of Genes: Non-supervised Orthologous Groups

SOD

Superoxide dismutase

POD

Peroxidase

CAT

Catalase

APX

Ascorbate peroxidase

MDA

Malondialdehyde

S-UE

Solid-urease

S-ACP

Solid-acid phosphatase

S-SC

Solid-sucrase

S-CAT

Solid-catalase

S-β-GC

Solid-β-Glucosidase

S-PPO

Solid-polyphenol oxidase

DHA

Dehydrogenase

PAS

Periodic Acid-Schiff

PCoA

Principal component analysis

ITS

Internal transcribed spacer

SD

Standard deviation

ANOVA

One-way analysis of variance

IAA

Indole-3-acetic acid

AP

Available phosphorus

AK

Available potassium

SEM

Scanning electron microscopy

SynCom

Synthetic microbial community

PGP

Plant growth-promoting

RT-qPCR

Reverse transcription-quantitative polymerase chain reaction

PD

Phylogenetic diversity

OUT

Operational taxonomic unit

ROS

Reactive oxygen species

WT

Wild-type

N

Nitrogen

P

Phosphorus

K

Potassium

NH₄⁺ -N

Ammonia

TSB

Tryptic soy broth

ELISA

Enzyme-linked immunosorbent assay

CFS

Cell-free supernatants

PKS

Polyketide synthase

Authors’ contributions

Fengwang Ma and Chao Li: Conceptualization, Validation. Yanan Duan and Ziqing Ma: Methodology, Formal analysis, Data curation, Writing - original draft. Yiwei Jia and Chao Yang: Writing-Review & Editing. Yiting Liu and Zhijun Zhang: Software, Visualization, Investigation. Mao Zhiquan and Xiaoqing Gong: Supervision. All authors have discussed the results, read and approved the contents of the manuscript.

Funding

This research was supported by the National Natural Science Foundation of China (32302476 and U24A20414), the earmarked fund for the China Agriculture Research System (CARS-27), the Xinjiang Key Research and Development Project (2023B02018), and the Taishan Scholar Funded Project (NO.ts20190923).

Data availability

All data required to evaluate the conclusions of the paper are available in the paper and/or the Supplementary Materials. All sequence data derived from experiments conducted are deposited in GenBank under accession numbers, and PV849104 to PV849133. The raw sequence data for the 16S rRNA gene and metagenomics of all samples were submitted to the NCBI Sequence Read Archive database (https://www.ncbi.nlm.nih.gov/): (1) The rhizosphere and bulk soil of MdUGT88F1 transgenic and wild-type apple lines under the BioProject: PRJNA1233871 (https://www.ncbi.nlm.nih.gov/sra/PRJNA1233871) and PRJNA1233884 (https://www.ncbi.nlm.nih.gov/sra/PRJNA1233884). (2) The pot experiment for Fpmd MR5 strain inoculation under the BioProject: PRJNA1233958 (https://www.ncbi.nlm.nih.gov/sra/PRJNA1233958). (3) The rhizosphere soil bacteria treated with root exudates compounds and synthetic communities under the BioProject: PRJNA1233940 (https://www.ncbi.nlm.nih.gov/sra/PRJNA1233940). The datasets on metabolomics generated and analyzed during the current study are available at MetaboLights under the accession number MTBLS10213 (https://www.ebi.ac.uk/metabolights/). Comment: Until the data release on MetaboLights, the data is provided via nexcloud and can be assessed at the following links: https://www.ebi.ac.uk/metabolights/reviewer1eeed12e-36ab-4aa5-8473-248913c935a8. Additional data related to this paper can be requested from the authors.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.