The culture of plant fungal pathogen (Fusarium proliferatum, Fusarium verticillioides, Fusarium oxysporum, Fusarium solani, Rhizoctonia solani, Alternaria alternata, Albifimbria verrucaria, Aspergillus flavus, Penicillium brasilianum, and Phytophthora cactorum) are used in this study. Details of the fungal plant pathogen cultures are shown in Supplementary Table 1. Stock cultures were maintained on potato dextrose agar (PDA) plates at 4°C. Pre cultures were established by transferring a stock agar plug containing mycelia onto fresh PDA plates and incubating for 6 days at 28°C.

The rhizosphere soil and root system were taken from healthy apple trees in the replanted orchard [replanted for 4–5 years on old apple orchards (more than 25 years old)] in the Northwest Loess region. The repeated cropping of the sampling orchard was serious, and there were dead trees. The basic information of the sampled sites are presented in Supplementary Table 2. The isolation of antagonistic bacteria refers to the method of Filippi et al. (2011), Fang et al. (2020), and Duan et al. (2021). Morphologically distinct colonies were replicated, and purified isolates were stored in cryogenic tubes. The bacterial strains screened for antagonistic activity toward plant fungal pathogen using dual culture technique refer to the method of Yu et al. (2011) and Duan et al. (2021).

The morphological identification was conducted based on the methods of Zhao et al. (2018). Biochemical identification was based on the methods of Dong and Cai (2001) and Vos et al. (2011). The GEN III MicroPlate™ test panel identification was based on the methods of Bochner (1989).

Genomic DNA was extracted from the obtained isolates using the Bacteria Genomic DNA Extraction Kit (Tiangen Corporation Ltd, Beijing, China) according to the manufacturer’s instructions (Sambrook and Russel, 2001). Molecular identification was performed on strain XNRB-3 using 16S ribosomal RNA gene (16S rDNA), DNA gyrase subunit A (gyrA), DNA gyrase subunit B (gyrB), and RNA polymerase subunit B (rpoB) gene sequence analysis. The primers and annealing temperatures are shown in Supplementary Table 3. Polymerase chain reaction (PCR) amplification, product purification, and sequencing refer to the method of Duan et al. (2021).

Sequences of 16S rDNA, gyrA, gyrB, and rpoB were aligned using maximum likelihood (ML) methods were performed for the datasets using RAxML-HPC2 on XSEDE (8.2.12) on the CIPRES website¹ (Stamatakis, 2014). Tree diagrams were created in FigureTree v1.4.3 and Adobe Illustrator CS6.

Genes related to the biosynthesis of lipopeptides, dipeptides and polyketides were detected by PCR using the primers listed in Supplementary Table 4. Amplification was performed with an Applied Biosystems 2720 Thermal Cycler (Applied Biosystems Inc., California, CA, United States). Following amplification, the PCR reaction mix (5 μL) was visualized by gel electrophoresis in a 1.5% agarose gel.

Seed culture of B. licheniformis XNRB-3 was prepared by inoculating a single colony into 100 mL of Luria-Bertani (LB) broth (tryptone 10 g, yeast extract 5 g, NaCl 10 g, pH 7.0) and incubated at 37°C overnight with agitation (180 rpm).

The cell-free culture filtrate (CFCF) of strain XNRB-3 was obtained according to the method of Abdelmoteleb et al. (2017). Bacteria were grown on the optimized liquid fermentation medium with constant shaking at 191 rpm for 24 h at 33 to obtain fermentation broth (FB), centrifugated at 10,000 rpm for 10 min at 4, and the supernatant (extracellular medium) was passed through a 0.22-μm Nylon66 microporous membrane to obtain the CFCF. The effect of strain XNRB-3 on Fusarium spore germination and hyphae refers to the method of Duan et al. (2021).

The ability of strain XNRB-3 to promote plant growth was tested with reference to the method in Supplementary Table 6. The FB of B. licheniformis XNRB-3 was centrifuged at 13,400 g for 10 min. The supernatant (extracellular medium) was removed and frozen at 4 for the analysis of the amino acids. Phytohormone extraction and quantitation were conducted using gas chromatography-mass spectrometry/selected ion monitoring (GC-MS/SIM) (Shimadzu, Japan). The specific method is mentioned in Supplementary Table 6.

A spectinomycin- and rifampin-resistant mutant of strain XNRB-3 (denoted XNRB-3^R) was obtained by inoculation of strain XNRB-3 into the LB medium containing gradually increasing concentrations of spectinomycin and rifampin (20, 50, 75, 100, 150, 200, 250, and 300 mg mL^–1; Supplementary Figure 4o) refer to the method of Yu et al. (2011). Malus hupehensis Rehd. seedlings were selected as the test material.

The ability of B. licheniformis XNRB-3 to colonize roots was determined according to the method of Sanei and Razavi (2011), with some modifications. The M. hupehensis Rehd. seedlings were carefully uprooted from the substrate, their roots thoroughly washed in tap water without intentional wounding, and dipped in a bacterial (XNRB-3^R) suspension [1 × 10⁸ colony-forming units (CFU) mL^–1] for 10 min. For the control treatment, plants were treated similarly except that the roots were dipped in 10 mM MgSO₄.7H₂O. Plants were then transplanted (one per pot) into pots (AC140: the outer diameter is 12.5 cm, the inner diameter is 11 cm, and the height is 9.5 cm) filled with an autoclaved (121°C, 1 h, twice on consecutive days) soil mixture (180 mL vermiculite first, then 540 mL of soil), and finally covered with 180 mL of sterile vermiculite. There were four replicated plants for each treatment in a randomized complete block design. The experiment was repeated three times. Plants were incubated under greenhouse conditions. The air temperature during the experiment fluctuated between 18 and 33°C. Plants were watered as needed. To determine colonization of root tissue by the bacteria, plants were uprooted delicately from the pots and the root systems were thoroughly washed under running tap water, dried with sterile filter paper, and cut into 1-cm-long pieces. For each plant, samples of 2 g of root pieces were surface-deinfested in 1% NaOCl for 3 min, washed three times in sterile distilled water, and ground in 10 mL of 10 mM MgSO₄.7H₂O using an autoclaved pestle and mortar. Serial dilutions of the macerates were plated onto LB agar (LB supplemented with spectinomycin and rifampin at 300 mg mL^–1) and incubated at 37°C for 48 h. Then, bacterial colonies were counted, and the bacterial populations were expressed as CFU g ^–1 of fresh root tissue.

The protective function of strain XNRB-3 on the roots of M. hupehensis Rehd. seedlings was verified by Periodic Acid Schiff (PAS) staining according to the method of Shao et al. (2020) and Duan et al. (2021).

The ability of strain XNRB-3 to control ARD was investigated in a Fusarium−infested sterilized soil following the method described by Wu et al. (2019), with some modifications. Treatments included a sterilized soil inoculated with sterile distilled water as a negative control, a sterilized soil inoculated with Fusarium as a positive control, and sterilized soil inoculated with Fusarium and the strain XNRB-3. Each treatment included 15 M. hupehensis Rehd. seedlings. The spore suspension of Fusarium was first drenched into the sterilized soil, followed by a suspension of strain XNRB-3. The final concentration of Fusarium (10⁵ spores g^–1) and strain XNRB-3 (10⁸ CFU g^–1) was included in the growth substrates. Plant seedlings were then transplanted in the substrate trays (AC140), and then grown at 16 h light/8 h dark at 28°C. Plants were watered as required for plant growth and disease development. Each pot contained one M. hupehensis seedling, and all pots were arranged randomly with 15 replicates per treatment (Cachinero et al., 2002). The experiment was repeated three times. Disease severity was estimated over the course of 5 weeks starting 1 week after inoculation. The scoring criteria and calculation formula are presented in Supplementary Table 8.

The PDP was measured to estimate wilt severity and the ability of plants in different treatment groups to recover from the disease. To fulfill Koch’s postulates, isolates from discolored fibrous roots were also obtained from all dead plants at the end of the experiment. Isolation and identification of each organism were performed to the genus level. The growth substrate samples near to plant root rhizosphere were collected weekly after being transplanted and stored at –20°C. The population of Fusarium in the growth substrate samples were determined by real-time PCR.

In the present study, 15 carriers purchased from different Chinese commercial enterprises through Taobao were used as formulation carriers for strain XNRB-3. The selected properties of these carrier candidates are listed in Supplementary Table 10. The carriers were dried to a moisture content of 5% in an oven (Shandong, China) at 80°C for 24 h, finely ground in a hammer mill to pass through a 1-mm screen, and stored at room temperature for further study.

The effect of the carrier as a substrate on the survival of strain XNRB-3 was evaluated using plastic bottles equipped with a 0.22-μm filter membrane in the cap to allow for air exchange, which was in reference to the method of Wei et al. (2015), with some modifications. Then, 20 mL of fermentation broth of strain XNRB-3 was mixed with 100 g of each carrier (sterilization at 115°C for 30 min). Carriers treated with 20 mL of aseptic fermentation broth were used as the controls. Each treatment included three replications (three bottles). In addition, these plastic bottles were covered with black plastic bags to avoid the influence of light on the survival of strain XNRB-3. The bottles were stored at 25°C and periodically sampled at 0, 10, 20, 30, 60, 90, 120, and 180 days post-inoculation. The population of strain XNRB-3 in each carrier at each sampling time point was determined by the plate count method using LB agar.

Four key factors including inoculation amount, pH, temperature, and rotating speed were selected to carry out the Box–Behnken test design with four factors and three levels. When the above fermentation conditions were optimized, except for the test single factor as a variable, the other conditions were unchanged, and the population number of strain XNRB-3 was determined for 180 d. Using Box–Behnken (Design-Expert 8.0.6) in the RSM to optimize the solid fermentation conditions of strain XNRB-3 (Anthony et al., 2009).

The BIO product was produced by Chuangdi Microbial Resources Co., Ltd. (Dezhou, China). The mixture of cow manure compost and wheat straw (1:2, w/w) was aerobically fermented with strain XNRB-3 for 6 days at 40°C, and the bacterial density was 5.0 × 10⁹ CFU g^–1. All operations are carried out under optimized fermentation conditions. The content of organic matter was 46.21%, total nitrogen was 2.36%, P₂O₅ was 1.49%, and K₂O was 3.03%.

In May 2017, the pot experiment was conducted at the National Apple Engineering Experiment Center of the Horticultural Science and Engineering College of Shandong Agricultural University and the State Key Laboratory of Crop Biology (117.156540 longitude, 36.164443 latitude). The collection and treatment of the experimental soil was conducted as described by Duan et al. (2021). The physicochemical properties of the tested soil were presented in Supplementary Table 11. There were four treatments: (1) 31-year-old orchard soil (CK1), (2) 31-year-old orchard soil fumigated with methyl bromide (CK2), (3) bacterial fertilizer carrier treatment (T1), and (4) XNRB-3 bacterial fertilizer treatment (T2). The addition amount of strain XNRB-3 and fertilizer carrier accounts for about 1% of the weight of the soil. The plants received unified watering and manure management and sample collection and management according to that of Duan et al. (2021).

To test the potential of strain XNRB-3 bacterial fertilizer to control ARD under field conditions, the field test was carried out in Wangtou Village, Laizhou City (Shandong China, Long: 119.814701, Lat: 37.095159). Physicochemical properties of the tested soil are presented in Supplementary Table 11. After the apple orchard was rebuilt, severe ARD occurred, the growth of fruit trees was weak, and the survival rate was less than 50%. In March 2020, 28-year-old trees were removed from the orchard, and the replanted orchard was simultaneously established. The apple seedlings used in the experiment were 2-year-old grafted seedlings. The rootstock and spike combination was from Yanfu 3/T337. The grafted seedlings had a stem thickness of about 10 mm and a fixed stem of 1.4 m. They were purchased from Laizhou Nature Horticultural Technology Co., Ltd. (Shandong, China). The row spacing of the plants is 1.5 m × 4 m, and the tree shape was pruned to a spindle shape. The production of strain XNRB-3 bacterial fertilizer was as described above.

The experiment consisted of four treatments: 28-year-old orchard soil (CK1), 28-year-old orchard soil fumigated with methyl bromide (CK2), bacterial fertilizer carrier treatment (T1), and XNRB-3 bacterial fertilizer treatment (T2). A planting hole of 80 cm³ was dug according to the row spacing, and the bacterial manure carrier and XNRB-3 bacterial manure were mixed with soil and backfill. The amount of application for each young tree was controlled at 1 kg, and 20 trees were treated for each. A new application was carried out at the beginning of the second spring. All indexes were measured on 15 July and 20 October in 2020 and 2021. The sample collection and management were conducted in accordance with that done by Sheng et al. (2020).

All statistical analyses were performed with the IBM SPSS 20.0 (IBM SPSS Statistics, IBM Corporation, Armonk, NY, United States). Different lowercase letters represent significant differences between treatments (one-way ANOVA, p < 0.05) according to Duncan’s multiple range test. The figures were plotted with Microsoft Excel 2013 (Microsoft Corporation, Redmond, WA, United States) and GraphPad Prism 7.0 (GraphPad Software, Inc., San Diego, CA, United States). The R statistical platform (v.4.1.1) was used for principal coordinates (PCoA) and cluster analysis (Duan et al., 2021).

The bacterium XNRB-3 was isolated from the root system of a healthy apple tree in a replanted orchard in Jincheng City, Shanxi Province. This bacterium had the strongest inhibitory effect on plant pathogenic fungi, with an inhibition rate of more than 70%. F. oxysporum and Alternaria alternata had the strongest inhibition rates, reaching 84.52 and 84.59%, respectively (Figure 1c).

Strain XNRB-3 was cultured on an LB liquid medium at 37°C for 24 h and formed biofilms. According to Bergey’s Manual of Systematic Bacteriology (2nd edition) and the Common Bacterial Identification Manual, strain XNRB-3 matched the physiological and biochemical properties and morphological characteristics of B. licheniformis (Supplementary Figure 2 and Supplementary Table 12). Tests of the carbon source utilization and chemical sensitivity revealed that strain XNRB-3 can utilize 32 carbon sources and was sensitive to pH 6, 1% NaCl, 4% NaCl, and 8% NaCl conditions. The values of probability (PROB), similarity (SIM), and distance (DIST) were 0.689, 0.689, and 4.698, respectively, and strain XNRB-3 belonged to B. licheniformis, which was identified based on carbon utilization (Supplementary Tables 13, 14).

Approximately 555–938 bases were sequenced for rpoB and gyrA, 1,455 bases for 16S rDNA, and 1,136 bases for gyrB. The ML analysis of identities based on the four gene sequence alignments revealed that strain XNRB-3 had the highest homology with B. licheniformis (Supplementary Figure 3). In sum, strain XNRB-3 was identified as B. licheniformis.

The endophytic bacterium XNRB-3 possessed multiple PGP properties such as phosphate and potassium solubilization; nitrogen fixation; IAA, GA, and ABA production; ACC deaminase, ammonia, and amylase production; siderophore, cell wall-degrading enzyme (cellulase, pectinase, β1,3-glucanase, chitosanase, and protease), and antifungal activity against phytopathogens (Supplementary Tables 6, 15 and Supplementary Figures 4a–m).

An HPCE method was used to identify 20 Phenyl isothiocyanate (PITC)-derivatized amino acids, 17 of which were separated (Supplementary Table 16), including aspartic acid (Asp), threonine (Thr), serine (Ser), glutamate (Glu), proline (Pro), glycine (Gly), alanine (Ala), cystine (Cys), valine (Val), methionine (Met), isoleucine (Ile), leucine (Leu), tyrosine (Tyr), phenylalanine (Phe), histidine (His), lysine (Lys), and arginine (Arg).

To determine the mechanisms underlying the effects of strain XNRB-3, PCR was used to screen strain XNRB-3 for genes involved in the biosynthesis of antibiotics. Amplicons of the expected sizes detected included yndJ (involved in biosynthesis of Yndj protein), qk (involved in subtilisin synthesis), bamC (involved in bacillomycin synthesis), ituD (involved in iturin A synthesis), fen and fenD (involved in fengycin synthesis), and srf and srfAB (involved in surfactin synthesis) (Supplementary Figure 5).

The results of single-factor tests showed that the main carbon source affecting the growth of strain XNRB-3 was sucrose, the main nitrogen source was beef extract, and the inorganic salts were MgSO₄, KH₂PO₄, and KCl (Supplementary Figures 1d,g,h). According to the orthogonal test results (Supplementary Table 17), the organic salt that had the strongest effect on the growth of strain XNRB-3 was MgSO₄, followed by KH₂PO₄ and KCl. According to the K value, the optimal combination was F4I3E3 (MgSO₄ 1.0 g L^–1, K₂HPO₄ 1.5 g L^–1, and KCl 1.0 g L^–1). According to the results of the single-factor experiment (Supplementary Figure 1), the Plackett–Burman experiment was performed using the experimental factors and levels in Supplementary Table 18, and Minitab 17 Software was used for multiple regression analysis. The optimal equation with OD₆₀₀ as the response value was OD₆₀₀ = 0.4472 + 0.004009A – 0.004149B – 0.01018C – 0.01204D + 0.00649E – 0.01527F + 0.00465G + 0.001711H + 0.000272I. According to the size of the P-value, the key factor affecting OD₆₀₀ was A (temperature), followed by B (rotating speed), G (sucrose), H (beef extract), F (MgSO₄), I (KH₂PO₄), D (pH), C (liquid volume), and E (KCl) (Supplementary Tables 19, 20). The effect of A–H was highly significant (p < 0.01), the effect of F was significant (p < 0.05), and the effect of I–E was not significant. A Box–Behnken test was conducted with four factors (sucrose, beef extract, temperature, and rotating speed) and three levels (Supplementary Table 21). Design Expert 8.0.6 software was used to perform regression analysis and obtain the multiple quadratic regression equations: Y = 22.054 – 0.1142A – 0.0875B + 0.4358C – 0.5942D – 0.1175AB + 1.6175AC + 0.1075AD – 0.4825BC – 0.0975BD – 0.6425CD – 0.8941A² – 0.4016B² – 1.7641C² – 2.1366D², where Y is the inhibition zone diameter (mm), A is sucrose, B is beef extract, C is temperature, and D is rotating speed. An analysis of variance and a significance test of the regression model were conducted (Supplementary Table 22 and Table 1), the F-value of the regression model was 94.09 (p < 0.0001), indicating that the regression model was robust. The coefficient of determination of the model was R² = 0.9895 and R²_Adj = 0.9790, indicating a good fit. Therefore, the regression model could be used to analyze and predict the abundance of the strain XNRB-3. In the regression model, the primary term C, D and the secondary term AC, CD significantly affected the diameter of the inhibition zone (p < 0.0001); the significance of A, B, C, and D was the same based on the results of the Plackett–Burman test.

The interactions between temperature and sucrose, temperature and beef extract, and rotating speed and beef extract were the most significant; the interactions between rotating speed and sucrose, temperature and rotating speed, and the sucrose and beef extract on the diameter of the inhibition zone were relatively weak (Supplementary Figure 6). The optimal fermentation conditions for strain XNRB-3 predicted by the regression model were sucrose 21.03 g, beef extract 8.54 g, temperature 32.73°C, and rotating speed 191.46 rpm. The maximum theoretical value of the predicted inhibition zone diameter was 22.17 mm. To further verify the predicted value, three parallel experiments were performed using the optimized fermentation conditions. The diameter of the inhibition zone was 22.03 ± 0.24 mm, and the error from the theoretically predicted value (22.17 mm) was only 0.63%, indicating that the predicted value was a good fit with the measured value. Thus, the optimized model was reliable.

In summation, the optimal fermentation conditions for strain XNRB-3 were sucrose 21.03 g, beef extract 8.54 g, MgSO₄ 1.0 g, K₂HPO₄ 1.5 g, KCl 1.0, 1 L, pH 7.5, temperature 32.73°C, rotating speed 191.46 rpm, inoculation amount 5%, and filling volume 40%. Under optimal conditions, the OD₆₀₀ values (after dilution) of the strain XNRB-3 solution at different incubation times were measured and the growth curves were drawn (Supplementary Figure 1b).

A long and stable shelf life is one of the most important commercial characteristics of biocontrol products. Therefore, the population dynamics (an index of shelf life) of strain XNRB-3 using different candidate carriers were characterized in the LB medium. At the end of the storage period, the viable numbers of strain XNRB-3 in cow dung compost (CDC) and wheat straw were greater than 8.78 and 8.74 log CFU g^–1 dry carrier, respectively. CDC: wheat straw maintained the largest populations of strain XNRB-3 at a ratio of 1:2 (Figures 2M–O). Design Expert 8.0.6 software was used to perform regression analysis of the test data of Supplementary Tables 23, 24, and the multiple quadratic regression equation obtained was as follows: Y = 8.716-0.43A + 1.2608B + 0.6767C + 0.2692D – 0.52AB – 0.155AC + 0.145AD + 0.6125BC + 0.465BD – 0.1475CD – 0.9405A² – 2.0218B² – 0.7355C² – 0.4243D², where Y is population (E + 08), A is the inoculation amount; B is pH, C is temperature, and D is rotating speed.

The regression model was further analyzed by variance analysis and significance tests (Table 2). The F value of the regression model was 102.13 (p < 0.0001), indicating that the regression model was robust. The coefficient of determination of the model was R² = 0.9903 and R²_Adj = 0.9806, indicating a good fit. Therefore, the regression model could be used to analyze and predict the abundance of B. licheniformis XNRB-3. In the regression model, the effects of the first term A, B, C and the second term BC on the diameter of the inhibition zone were significant (p < 0.0001). The interactions between pH and inoculation amount, temperature and pH, and rotating speed and pH had the most significant effect on the population of strain XNRB-3 (Figures 2A–L). The optimal solid fermentation conditions for B. licheniformis XNRB-3 were predicted by the regression model: inoculation amount 20.25%, pH 8.25, temperature 39.23°C, and rotating speed 199.75 rpm. The predicted maximum theoretical value of the population of strain XNRB-3 was 8.91E + 08. To verify the predicted value, three parallel experiments were performed using the optimized fermentation conditions. The diameter of the inhibition zone was (22.03 ± 0.24) mm and the error from the theoretically predicted value (8.91E + 08) was only 0.86%, indicating that there was a good fit between the predicted value and the measured value and that the optimized model was robust.

In sum, the optimal solid fermentation component of B. licheniformis XNRB-3 was cow dung: wheat straw = 1:2, and the optimal conditions were inoculation amount 20.25%, pH 8.25, temperature 39.23°C, and rotating speed 199.75 rpm.

Microscopic observations of hyphal and spore morphology revealed that the control Fusarium mycelium was uniform in thickness and slender with fewer branches, spore structure was complete, and growth was strong (Figure 1aA–D). The mycelia treated with the CFCF were irregularly reticulated, uneven in thickness, shrunken (Figure 1aG–L), thinned (Figure 1aI,K), the cells had ruptured, and cell contents had overflowed (Figure 1aG,H,I,K), the spore cell wall was broken and deformed (Figure 1aE,H,I,K). Furthermore, FB, CFCF, and crude extract significantly inhibited the spore germination of Fusarium spores, and the spore germination rate decreased by more than 60% (Figures 1b,e). The antifungal metabolites produced by strain XNRB-3 were extracted from 3-day-old CFCF. As the concentration of extracellular metabolites increased, the inhibitory effect on the growth of Fusarium became more pronounced. At higher concentrations (25, 50, and 75%), the inhibition rate reached more than 60% (Figure 1d).

The results of the experiment conducted for the present study showed that a 25% concentration of strain XNRB-3 CFCF showed high activity against fungal pathogens. After GC-MS chromatographic detection and analysis, the main antibacterial substances with Area% > 0.62 and retention index RI > 700 were selected (Supplementary Figure 7 and Supplementary Table 25). Most substances identified from the components were organic acids and esters, alcohols, ketones, alkanes, and phenols (Figure 3). Among the 42 identified compounds, 16 pure compounds were purchased for individual testing of antifungal properties (Supplementary Figure 8). The nine pure compounds showed antifungal activity against pathogenic fungi (Supplementary Table 26). Among them, 2,4-di-tert-butylphenol and alpha-bisabolol had the strongest inhibitory effect on plant pathogenic fungi, especially Fusarium. At a concentration of 1000 μg L^–1, the inhibition rate was higher than 50% (Figures 4a,b). Of the nine pure compounds, the fresh weight and root growth of Arabidopsis plants compared with the control treatment (water and ethanol) were significantly enhanced by butanedioic acid, monomethyl ester (500 μg), and dibutyl phthalate (500 μg) (Figures 5a–d). Compared with ethanol treatment, the fresh weight, root length, number of primary roots, and number of secondary roots of Arabidopsis seedlings were increased by approximately 2.56, 1.42, 0.91, and 5.5 times, respectively, when treated with dibutyl phthalate; by 2.01, 4.08, 0.33, and 4.5 times, respectively, when treated with butanedioic acid and monomethyl ester; and by 1.89, 0.89, 0.71, and 10.5 times, respectively, when treated with 3-nonen-2-one. The above findings indicate that the main antifungal compounds produced by strain XNRB-3 play important roles in disease control and growth.

The phlorizin utilization efficiency by strain XNRB-3 in the MSM solution was high (Supplementary Figure 4q). After culture for 60 h, the degradation rate was 60.75, 64.79, and 68.83% when the amount of inoculum added was 1, 1.5, 2%, respectively. The ability of strain XNRB-3 to degrade phlorizin increased as the amount of inoculum added increased. Strain XNRB-3 could effectively utilize CA, BA, FA, and PHBA (Supplementary Table 27), and the degradation rates ranged from 45.65% to 69.20%. Strain XNRB-3 could also efficiently degrade phlorizin, CA, BA, FA, and PHBA in soil (Table 3). The content of phlorizin in the soil decreased to 2.9292 μg g^–1 at 9 days after inoculation. The content of BA and PHBA in the soil decreased to 18.4953 and 5.2882 μg g^–1, respectively. The FA and CA concentrations in the soil decreased to 13.3669 and 0.3785 μg g^–1 at 9 days after treatment, respectively, which was only 13.31 and 3.63% of the control concentration.

Strain XNRB-3 can form a thick biofilm in a static medium. After 14 days of planting, the population of strain XNRB-3 colonized on root tissue was approximately 7.30 × 10⁶ CFU g^–1. The population sharply decreased to 6.47 × 10⁴ CFU g^–1 after 35 days (Supplementary Figure 9b). The conidia and hyphae of Fusarium treated with strain XNRB-3 only attached to root epidermal cells (Supplementary Figures 10A–D). Fusarium hyphae invaded the cortical cells through the intercellular layer and then entered the vascular column, and a large amount of cell content (sticky substances and starch granules) was also observed in the cortical cells and vascular column (Supplementary Figures 9a, 10).

In the greenhouse test, the symptoms of Fusarium wilt appeared 7 days after the plant seedlings were transplanted into the infected soil. The incidence of Fusarium wilt increased rapidly over the next 28 days, and the disease severity index increased progressively over the experimental period and ultimately reached 4.00 ± 0.00; a reduction in disease progress was noted in plants treated with strain XNRB-3 (Table 4). In the fifth week, the disease index and relative control effect of plant seedlings inoculated with strain XNRB-3 were both stable at 33 and 50%, respectively. During the test period, plants inoculated with sterile distilled water remained healthy. The disease symptoms were the result of artificial infection with Fusarium, which was successfully re-isolated from inoculated plants at the end of the experiment. According to RT-PCR, the size of the Fusarium populations of the two treatments significantly increased after 21 days of planting in the infected soil. Compared with the CK treatment (5.50 × 10⁵ copies g^–1 soil), the size of the Fusarium population treated with strain XNRB-3 (2.37 × 10⁵ copies g^–1 soil) was significantly reduced (Supplementary Figure 9c). These results were consistent with the incidence of the aforementioned diseases.

The same results were obtained in the outdoor pot experiment. The biomass of apple plants treated with strain XNRB-3 (T2) in September was significantly higher than that of CK1. The plant height, ground diameter, fresh weight, and dry weight were increased by 88.07, 61.28, 181.35, and 140.44%, respectively, which were second only to the methyl bromide fumigation treatment (Supplementary Figure 11). Treatment with strain XNRB-3 also significantly enhanced the root growth of apple plants (Figures 5e,i,j,l,m). In September, the plants had grown considerably, and the root length, surface area, number of tips, and number of forks were significantly lower in CK1 than in CK2 and T2. The length, surface area, number of tips, and number of forks were 1.73, 2.62, 5.67, and 2.65 times higher in T2 than in CK1. The root respiration rate and the SOD, POD, and CAT activity increased from July to September in T2 and CK2 (Figures 5e–h,k,n). In September, the root respiration rate in T2 was 1037.98 μmol O₂⋅g^–1 FW⋅min^–1, which was 1.49 and 1.32 times higher compared with that observed in CK1 and T1. The activity of SOD, POD, and CAT was 1.50, 2.93, and 1.72 times higher in T2 than in CK1. The MDA content showed the opposite trend: the MDA content was 29.96, 60.35, and 78.13% lower in July, August, and September, respectively, in T2 compared with CK1.

The XNRB-3 bacterial fertilizer treatment (T2) significantly promoted the growth of young apple trees, and this difference became significant in October 2021 (Figures 6A,B). In July and October in 2020 and 2021, the biomass indicators were significantly lower in CK1 and T1 than in CK2 and T2. In October 2021, the plant height, ground diameter, number of branches, and average branch length increased by 38.10, 62.79, 49.17, and 176.00% in T2 relative to CK1, respectively, and the plant height, ground diameter, number of branches, and average branch length were 1.27, 1.25, 1.37, 1.77, 2.45, and 2.21-fold higher in T2 than in T1, respectively (Figures 6C–F).

In October of 2020 and 2021, the soil phenolic acid content was highest in CK1 and T1 and lowest in CK2 and T2 (Figures 7A,B). The total soil phenolic acid content was significantly lower in T2 than in CK1. In October 2020, the soil content of cinnamic acid, phlorizin, benzoic acid, ferulic acid, and p-hydroxybenzoic acid was 54.38, 74.02, 77.29, 54.70, and 71.83% lower, respectively, in T2 than in CK1. In October 2021, the soil content of cinnamic acid, phlorizin, benzoic acid, ferulic acid, and p-hydroxybenzoic acid was 87.23, 91.95, 89.80, 84.11, and 91.98% lower, respectively, in T2 than in CK1.

The activity of urease, phosphatase, sucrase, and catalase increased steadily in T1 and T2 in July and October in 2020 and 2021 relative to CK1. The soil enzyme activity decreased in the first year of the fumigation treatment; it then continued to increase, increasing most significantly in T2 (Figures 7G–J). In October 2020 and 2021, the urease activity was 1.65-fold and 1.64-fold higher, the phosphatase activity was 1.56-fold and 1.93-fold higher, the sucrase activity was 2.03-fold and 2.34-fold higher, and the catalase activity was 1.69-fold and 1.87-fold higher in T2 than in CK1, respectively. After 2 years of applying strain XNRB-3, the physical and chemical properties of plant rhizosphere soil were significantly improved. Compared with CK1, organic matter, total nitrogen, total phosphorus, total potassium, available potassium, available phosphorus, NH₄⁺-N, nitrate nitrogen, and soil pH increased by 228.54, 245.25, 242.29, 30.53, 327.89, 128.64, 56.05, 239.61, and 7.43%, respectively, in T1. Various nutrient indexes of the soil were also increased in T1, indicating that the addition of organic amendments could enhance the nutrient conditions of the soil (Supplementary Figure 12).

In July and October of 2020 and 2021, the number of soil bacteria in the XNRB-3 bacterial fertilizer treatment (T2) increased significantly, and compared with CK1, the number of soil bacteria was increased by 107.21, 45.13, 100.79, and 113.08% in T2, respectively (Figure 7C). There were significant differences in the number of soil fungi between different treatments. Compared with CK1, the number of soil fungi in CK2, T1, and T2 was significantly reduced, the number of soil fungi in T1 was significantly higher than that in T2, and the effect of T2 was similar to the fumigation treatment. In October 2020, 2021, the number of soil fungi in CK2 and T2 was reduced by 58.55, 50.26%, and 62.74, 52.83% compared with CK1, respectively (Figure 7D). The number of actinomycetes and the ratio of bacteria/fungi in the soil were significantly higher in T2 than in CK1 (Figures 7E,F).

The qPCR results showed that the abundance of Fusarium was significantly reduced in July and October of 2020 and 2021 in CK2 and T2 compared with CK1 and T1 (Figures 7K–N). The abundance of F. proliferatum, F. solani, F. verticillioides, and F. oxysporum was 50.43, 40.74, 25.96, and 22.20% lower in T2 relative to T1 in October 2020, and 60.43, 57.97, 28.69, and 40.31% lower in T2 relative to T1 in October 2021, respectively.

In the present study, AWCD was used as an indicator of soil microbial activity. The variation in AWCD with incubation time is shown in Figures 8I,L. Here, AWCD increased as the incubation time extended for all treatments in October 2020 and 2021. The soil with strain XNRB-3 (T2) had a higher AWCD value than soil in other treatments, signifying that the addition of strain XNRB-3 increased the activity of microorganisms. After 120 h, the AWCD values of T1 and CK1 were similar, and the AWCD value was lower in CK2, indicating that fumigation can inhibit the activity of microorganisms for a longer period.

Biolog-ECO plates have six categories of carbon sources: carbohydrates, carboxylic acids, amino acids, polymers, phenolic compounds, and amines (Guo et al., 2015). In October 2020 (Figures 8I–K), the OD value of three types of carbon sources (polymers, miscellaneous, carboxylic acids, and amino acids) increased in T2 relative to CK1, but did not reach significant levels except for miscellaneous. In contrast, amines/amines and carbohydrates were reduced in T2, but did not reach significant levels. The effect of strain XNRB-3 treatment (T2) on the utilization rate of the four substrate groups (polymers, miscellaneous, carboxylic acids, and amino acids) was stronger compared with the other treatments. The utilization rate of these four substrate groups was 6.11, 0.97, 15.41, and 10.24% higher in T2 than in T1, respectively. In October 2021 (Figures 8M,N), the changes in the utilization rate of microbial substrates under different soil treatments varied starting in 2020. Compared with CK1, the OD values of the two types of carbon sources (carbohydrates and carboxylic acids) increased significantly in T2, and the other four carbon sources (polymers, miscellaneous, amines/amines, and amino acids) also increased but did not reach significance levels. With the exception of miscellaneous, the OD values of amines/amines, carboxylic acids, carbohydrates, polymers, and amino acids increased by 40.79, 42.17, 24.45, 10.56, and 7.91% in 2021, respectively, compared with 2020. These findings indicate that the soil microorganisms treated by strain XNRB-3 mainly use polymers, carboxylic acids, and amino acids. The utilization rate of the six types of carbon sources in CK2 was significantly lower compared with the other treatments, denoting that fumigation greatly affected the microbial community in the soil.

A principal component analysis and cluster analysis demonstrated that the soil microbial community structure in T2 and CK2 significantly differed from that in CK1 (Figures 8A–H). The soil bacterial community structure in T2 differed significantly from that in other treatments (Figures 8G,H), and the soil fungal and bacterial communities of T1 and CK1 were similar (Figures 8C,D,G,H). The Margalef, McIntosh, Brillouin, Simpson, and Shannon indexes reflect the richness and diversity of soil microbial communities (Supplementary Tables 28, 29). The abundance of soil fungi and bacterial communities was significantly increased after adding strain XNRB-3. The diversity of soil fungi communities was significantly increased, the dominance index and carbon source utilization were significantly reduced, and the bacterial community showed the opposite pattern. The functional diversity of the microbial community was analyzed based on the data of 96-h Biolog Eco plates in 2020 and 2021, the functional diversity of the microbial community in T1 and T2 showed similar changes. The diversity of soil microbial communities was significantly increased in T1 and T2 than in CK1, which was consistent with the results of the T-RFLP analysis.

These results indicated that XNRB-3 bacterial fertilizer can also improve the soil microbial environment along with its biocontrol activity.