Ageratum conyzoides samples were collected from Nanning, Hechi, and Yulin cities in the Guangxi Zhuang Autonomous Region, southern China. Flowers, leaves, stems, and roots were selected for the isolation of microbial strains. The A. conyzoides sampled organs were cut into small fragments (1 cm in size), sterilized with 75% ethanol for 10 s followed by 2% sodium hypochlorite for 1 min, rinsed three times in sterile water, and homogenized. The samples were allowed to stand for 10 min; then, 100 μL supernatant was diluted to 10⁻¹, 10⁻², and 10⁻³ with sterilized water, and 100 μL of each diluent was spread on LB plates (Supplementary Table S1). After 2 d of incubation at 30°C, different morphotypes were picked up and re-streaked onto LB plates for further purifying a single, pure colony. Subsequently, purified colonies were cultured on LB medium agar (Supplementary Table S1) slants at 4°C and in 25% (v/v) sterile glycerol stock at −80°C for further experiments.

Colletotrichum fructicola is the dominant pathogenic species responsible for plum anthracnose (Huang et al., 2022). This species was selected as the preliminary control target- indicator. All bacterial isolates were screened for antagonistic ability using plate confrontation assay (Hong et al., 2022). Generally, the isolated strains were inoculated in LB liquid medium, and cultured on a shaker at 30°C. After cultured for 24 h at 200 rpm, bacterial concentration was adjusted to an OD₆₀₀ value of 1.0. In the plate confrontation assay, the bacterial suspension was streaked using an inoculation loop on the left side of the PDA plate 2 cm away from the edge. A mycelial plug (5 mm in diameter) of each phytopathogenic fungus from the 5-day culture was obtained using a sterile cork borer and placed on a PDA plate at a distance of 5 cm from the center of the inoculated strain line. Plates inoculated with only the phytopathogenic fungal plug served as the control group and were incubated at 28°C for 5 days. Treatments were performed in triplicate. The inhibition rate was calculated using the following formula:

I (%) = (r₀–r₁)/r₀ × 100

Where I is the inhibition rate, r₀ is the radius of the mycelia in the control group, r₁ is the radius of the mycelia in the dual-culture plate.

Bacterial VOC antifungal activity assays were performed using the two-sealed-plate method (Cortazar-Murillo et al., 2023). Briefly, a 200-μL aliquot of the overnight culture of each bacterial isolate (OD₆₀₀ = 1.0) was spread separately on a plate containing LB medium. The lids were removed and replaced with another PDA plate containing a plug of 5 mm diameter of fungal mycelia on the PDA. Both plates were immediately sealed with a double layer of parafilm to prevent volatile leakage and then incubated at 28°C for 5 days. Each bacterial isolate was tested for antifungal activity in triplicate. Three assays were set up without bacterial treatment (LB only) and used as controls. Antifungal activity was measured as the percentage of reduction in mycelial growth. The inhibition rates were calculated using the following formula:

Where I is the inhibition rate, d0 is the diameter of mycelial growth in the control, and d1 is the diameter of mycelial growth after exposure to bacterial VOCs.

To investigate whether the morphology of C. fructicola causing plum anthracnose was altered by the antagonistic bacteria, the mycelia of C. fructicola colonies in a 5-day dual culture or in the control medium were observed under a Nikon Ni-E microscope.

Potted seedlings of pearl plum, which was developed from the local wild resources of Tian’e County in Guangxi, was used in this experiment. The 2-year-old pearl plum seedlings were transplanted into pots and maintained under uniform field conditions. After growing new leaves, plants in a healthy growth state were selected and prepared for the assays. To prepare bacterial suspensions, each isolate was first grown in LB liquid medium at 30°C in a shaker with 180 rpm for 24 h. All cells were harvested by centrifugation at 5,000 rpm for 5 min and the supernatants were discarded. The pellets were washed, resuspended in 0.05% tween-20 solution, and adjusted to a population density of 1.0 × 10⁷ colony-forming units (CFU)/mL. To obtain the fungal inoculum, after growing on a PDA plate for 5 days, a 5-mm diameter mycelia plug of C. fructicola was inoculated in PDB liquid medium with a shaking speed of 180 rpm at 28°C in a shaker incubator for 3–4 days. The conidia were collected by filtration and centrifugation, and adjusted to a density of 1.0 × 10⁶ conidia/mL. First, the bacterial suspension was sprayed onto the leaves of the pearl plum seedlings and the control was sprayed with a suspension containing 0.05% tween-20. After the pear plum seedlings were kept in a greenhouse with relative humidity between 75 and 90% for 24 h, 10 μL of conidia suspension was inoculated on each leaf where a round filter paper sheet (0.5 cm in diameter) was placed for moisture. Tree seedlings were inoculated with each strain.

After 5 days, the leaves of the pearl plum trees were examined for symptoms of C. fructicola infection and disease severity based on the diameter of necrosis, and disease prevention efficacy was calculated according to the following formula:

where Pe is prevention efficacy, d0 is the diameter of necrotic lesions when the leaves were inoculated with conidia but without bacterial suspension, and d1 is the diameter of necrotic lesions that developed when leaves were inoculated with conidia added with bacterial suspension.

The antagonistic bacteria were grown in an LB liquid medium at 37°C in a shaker incubator at 180 rpm for 24 h. Genomic DNA was extracted using a TIANamp Bacteria DNA Kit (Tiangen Biochemical Technology Co., Ltd., Beijing, China).16S rDNA, gyrA, gyrB, and rpoB gene fragments of the isolates were amplified using the primers (Roberts et al., 1994; Yamamoto and Harayama, 1995; De Clerck et al., 2004; Wang et al., 2013) listed in Supplementary Table S2. The PCR amplification system consisted of 50 μL of 5 × PCR buffer 10 μL, 4 μL of dNTPs (2.5 mM), 1.5 μL of each primer (10 μM), 0.5 μL DNA polymerase, 1 μL DNA template and adequate ddH₂O to the final volume of 50 μL. The PCR cycle conditions are listed in Supplementary Table S2, and the PCR products were detected by 1.2% agarose gel electrophoresis. Amplified products were sequenced by Sangon Biotech Co. Ltd. (Shanghai, China). The resulting sequence analysis was performed via BLASTN alignment on NCBI, and the sequences were submitted to GenBank to obtain gene accession numbers. 16S rDNA gene sequences, gyrA, gyrB, and rpoB with high similarity were used to concatenate and construct phylogenetic trees using the maximum likelihood method with bootstrap analysis of 1,000 replicates in the MEGA X software.

Antagonistic bacterial strains were characterized according to morphological characteristics, such as colony margin, shape, and color, as well as physiological and biochemical characteristics, including Gram reaction, salt tolerance, contact enzyme, V-P test, methyl red (MR) test, citrate utilization, gelatin liquefaction, amylolysis, starch hydrolysis, and carbon source utilization, based on the methods of Berger’s Manual of Bacterial Identification (Holt et al., 1994) and the Common Bacterial Systems Identification Manual (Dong and Cai, 2001).

The antagonistic spectra of the bacterial isolates against 14 fungal phytopathogens, namely, Botryosphaeria dothidea, Neoscytalidium dimidiatum, Pseudofusicoccum violaceum, Rhizoctonia solani, Exserohilum sp., Peronophythora litchi, Fusarium sulawesiense, Phanerochaete sordida, Alternaria brassicae, Epicoccum sorghinum, Diaporthe phoenicicola, D. phaseolorum, C. fructicola, and Magnaporthe oryzae were determined by the plate confrontation assay as described above. The phytopathogens were collected and stored in a low-temperature refrigerator at our laboratory.

The LB, NA, PSA, and PDA culture media were tested to determine the impact of their composition on the production of VOCs and their involvement in antifungal activity. After growing in PDB, LB, NB, and PSB liquid medium at 30°C in a shaker with 180 rpm for 24 h, 200-μL aliquots of the overnight culture of the isolated bacterial strains were spread on the PDA, LB, NA, and PSA plates, respectively. The two-sealed-plate assay method described above was used for the preliminary assay, and the antagonistic activities of VOCs were compared. The strain with the highest inhibition rate was dual-cultured with C. fructicola in two-sealed plates for 24, 48, 72, 96, and 120 h to determine the optimum time for identification.

The VOCs emitted by the bacterial isoaltes were identified using headspace solid-phase microextraction coupled with gas chromatography and mass spectrometry (HS-SPME/GC–MS). The following treatment was used to prepare VOCs: each of the eight representative bacterial isolates was grown in LB liquid medium at 30°C in a shaker with 180 rpm for 24 h, then a 200-μL aliquot of the culture was transferred to a headspace vial with 100-mL volume containing 20 mL LB solid medium, and was air-dried on a clean bench. Then, the headspace vial was sealed with a silicone rubber spacer, and placed in a constant temperature incubator at 30°C for 3 days. A headspace vial containing the uninoculated medium was used as a control to remove naturally occurring volatile compounds from each treatment. Subsequently, the headspace vial was placed in a water bath adjusted to 60°C. After equilibration for 5 min, bacterial VOCs were collected with SPME fibers (50/30 μm DVB/CAR/PDMS, Supelco, Inc., Bellefonte, PA), which were inserted into the upper side of the headspace vial for 50 min. Then SPME fibers were injected into the GC port of a gas chromatograph coupled to a mass spectrometer (GC/MS, 7890A-5975CMSD, USA). VOCs were desorbed at 280°C for 5 min in the GC injector port interfaced with a mass detector. Helium gas was used as carrier gas (1.0 mL min⁻¹, constant flow) and a DB-5 MS capillary column (30 m length × 0.25 mm inner diameter × 0.25 μm film thickness) from Agilent Technologies (Santa Clara, CA, USA) was used as a stationary phase. Operational conditions were the following: initial oven temperature of 40°C for 5 min, increased to 120°C (4°C per min) for 3 min, and further increased to 280°C (30°C per min) for a run time of 10 min. The mass spectrometer was operated in the electron ionization mode at 70 eV with a source temperature of 250°C, and with a continuous scan from 35 to 450 m/z. Data processing was performed using MassHunter software (Agilent Technologies). VOCs were initially identified by comparing MS peaks with those in the National Institute of Standards and Technology (NIST) 2017 MS Library.

Plum and inoculum preparations were implemented as described in Section 2.4. Two inoculation methods were used in this study. The first method involved spraying the bacterial suspension for 24 h, followed by inoculation with conidia of C. fructicola (pre-treatment). In the second procedure, the leaves were inoculated with conidia of C. fructicola 24 h before spraying the bacterial suspension (treatment). Three treatments were used: (1) CK, C. fructicola alone, (2) C. fructicola + antagonistic bacterial isolates, and (3) antagonistic bacterial isolates + C. fructicola. Each treatment consisted of three plants with 15 leaves per plant. All pots were kept under greenhouse conditions at a temperature of 28°C and relative humidity between 75 and 90%, under a photoperiod regime of 16 h light /8 h dark.

After 5 days, the leaves of the pearl plum trees were examined for symptoms of C. fructicola infection and disease severity (percentage of infected plant area) based on necrosis symptoms. A severity score on a scale of 0 to 5 was based on the visual scale of necrosis on leaves using the modified method of Choub et al. (2021), where 0 = no sign of visible necrosis, 1 = necrotic area covering less than 5% of the total leaf surface, 2 = necrotic lesions covering 6–5% of the total leaf surface, 3 = necrotic lesions covering 16–25% of the total leaf surface, 4 = necrotic lesions covering 26–50% of the total leaf surface, and 5 = necrotic lesions covering more than 50% of the total leaf surface. The incidence rate, disease index, and control efficacy were calculated using the following formulas:

Software DPS 3.01(Tang and Zhang, 2013) was used for statistical analysis. Statistical significance between data groups was determined at p < 0.05 using analysis of variance (ANOV). Data shown are means ± SD (n = 3).

To screen the microbial resources for controlling pearl plum anthracnose, a total of 249 bacterial strains were isolated from different plant organs (stems, leaves, flowers, and roots) of the A. conyzoides collected samples. All isolated bacterial strains were screened for antifungal properties against C. fructicola on a dual culture assay. Among them, 27 strains showed varying antagonistic activities; furthermore, particularly, strain AH8 displayed significant antagonism, reaching an inhibition rate of 65.4% in the confrontation plate assay, and strain H16 showed the highest inhibition rate (73.1%) in the two-sealed-plate assay. The inhibition rate of the isolates ranged from 14.1 to 65.4% in the confrontation plate assay and from 13.4 to 73.1% in the two-sealed-plate assay (Figures 1, 2 and Supplementary Table S3). Of the 27 isolates, 9, 6, and 12 were obtained from the flowers, leaves, and stems, respectively (Supplementary Table S4). To determine whether the morphology of C. fructicola was changed upon exposure to the antagonistic bacteria, the mycelia at the front line facing the bacterial colony or the medium control were sampled and observed under a microscope. Microscopic observations showed evidence of morphological alterations in fungal mycelia when C. fructicola was dual-cultured with the antagonistic bacterial isolate (Figure 3), including severe deformation, swelling, and vacuolation of hyphae exposed to bacterial diffusible compounds (Figure 3D), or thin, crimped, branching, and shriveling of hyphae exposed to bacterial VOCs (Figure 3F), in contrast to hyphae without dual cultured with the isolate developed normally (Figure 4B).

To clarify the taxonomic status of the 27 bacterial isolates, 16S rDNA of each isolate was amplified by PCR using the 16S universal primer pair 27F and 1492R. Gene fragments were sequenced and analyzed using BLAST at NCBI. The data showed that 25 isolates had high sequence identity (above 99%) with Bacillus spp., and were therefore affiliated with that genus, accounting for 92.6% of the total strains. The two other isolates belonged to genera Microbacterium and Pseudomonas. Furthermore, gyrA, rpoB, and gyrB from the 25 Bacillus strains and gyrB from the other 2 strains were amplified and sequenced. All sequences were deposited in GenBank and accession numbers were obtained (Supplementary Table S4). Phylogenetic tree construction based on the concatenated sequences of 16S rDNA, gyrA, rpoB, and gyrB from the 25 isolates and sequences derived from Bacillus type strains revealed that 14, 9, 1, and 1 strains were classified as B. velezensis, B. subtilis, B. altitudinis, and B. cereus, respectively (Figure 4). Phylogenetic analysis of two other strains, YB1 and H16 based on the 16S rDNA and gyrB sequences and sequences derived from the type strains, identified YB1 and H16 as Microbacterium phyllosphaerae and Pseudomonas monsensis, respectively (Figure 5).

Furthermore, a preliminary evaluation of the prevention of C. fructicola damage on pot plants by the 27 antagonistic bacterial isolates showed that most of the bacterial isolates were highly effective in preventing the disease. Among the isolates, XYAH1 and H16 showed the highest prevention efficacy, i.e., 98.4 and 92.0%, respectively (Supplementary Table S5). Therefore, eight representative isolates, B. velezensis XYAH1 and XYCJ12, B. subtilis XYAJ8 and AJ4, B. creceus XYDH11, B. altitudinis J1, P. monsensis H16, and M. phyllosphaerae YB1, were selected for further experiments.

The eight representative isolates were spread on an LB medium plate, placed at 28°C for 2–3 days, and their morphology was observed. Colonies of Bacillus spp. were creamy white, opaque, flat, rough, and wrinkled with irregular edges, and there were no obvious differences between them on LB plates (Supplementary Figure S1). A single colony of P. monsensis H16 was creamy white, round, convex, and smooth, with entire margins. The M. phyllosphaerae YB1 colony was beige-to-yellow, round, convex, and smooth, with entire margins (Supplementary Figure S1). Gram staining of Bacillus spp. was positive, while P. monsensis H16 and M. phyllosphaerae YB1 were Gram-negative (Supplementary Figure S2). Physiological and biochemical characterization revealed that Bacillus spp. grew on LB plates containing 10% NaCl or lower; Meanwhile, M. phyllosphaerae YB1 grew on LB plates containing 7% or less NaCk, and P. monsensis H16 grew on LB plates containing 5% NaCl or less (Supplementary Table S6). There were no apparent differences in the physiological or biochemical characteristics between B. velezensis and B. subtilis. Both effectively used citrate, sucrose, D-maltose, D-sorbitol, D-xylose, inositol, D-fructose, or D-glucose, but not α-lactose, D-galactose, dulcitol, or L- rhamnose. The contact enzyme, gelatin, starch, cellulose, and V-P reactions were positive, whereas the MR reaction was negative (Supplementary Figure S3 and Supplementary Table S7). B. cereus XYDH11 successfully used citrate, sucrose, D-maltose, D-sorbitol, inositol, D-fructose, or D-glucose, but not α-lactose, D-xylose, D-mannitol, D-galactose, dulcitol or L-rhamnose. Other characteristics are the same as B. velezensis and B. subtilis. Characteristics of B. altitudinis J1 are the same as B. cereus XYDH11, except MR and V-P reaction, D-maltose, D-sorbitol, inositol, and D-mannitol utilization. With regard to P. monsensis H16, this strain used citrate, sucrose, D-maltose, inositol, D-mannitol, D-fructose, and D-glucose, but not α-lactose, D-xylose, D-sorbitol, D-galactose, dulcitol or L-rhamnose; other characteristics are the same as B. cereus except MR reaction and starch hodrolysis. As for M. phyllosphaerae YB1, except starch hydrolysi, α-lactose, D-xylose, D-fructose, and D-glucose utilization other characteristics are the same as P. monsensis (Supplementary Figure S3 and Supplemental Table S7).

The biocontrol efficacy of eight representatives against anthracnose of pearl plum was determined under greenhouse conditions. The results showed that all representatives tested had high prevention effectiveness in the pre-treatment, in which case, the prevention efficacy of B. subtilis XYAJ8 and B. altitudinis J1 reached 100%, and the prevention efficacy of all other bacterial strains was no less than 85%. The disease index and incidence rate of anthracnose on pearl plum leaves in the treatments with the eight tested representatives were significantly lower than those in the control group. Especially were the disease index and incidence rate of leaves in the treatment with B. velezensis XYAH1. As a result, in the treatment, B. velezensis XYAH1 was the best with a control efficacy of 98.8%; conversely, M. phyllosphaerae YB1 showed the poorest control efficacy at 42.0%, whereas the others did not significantly differ in efficacy, ranging from 73.3 to 80.2% (Figure 6 and Supplementary Table S8).

The selected eight bacterial strains showed diverse inhibitory activities against the 15 fungal phytopathogens tested (Figure 7). Among the tested bacterial strains, Bacillus spp. exhibited relatively better inhibitory effects, whereas P. monsensis and M. phyllosphaerae showed weaker inhibition levels of all tested phytopathogens in the confrontation assay. These results indicated that the Bacillus spp. isolates exhibited broad-spectrum antifungal activity. In addition, except B. cereus XYDH11, most of the Bacillus spp. isolates displayed a high inhibitory effect on Magnaporthe oryzae, with inhibition rates ranging from 88.7 to 95.0%, followed by Peronophythora litchii, with inhibition rates ranging from 55.8 to 73.4%. However, B. cereus XYDH11, P. monsensis H16, and M. phyllosphaerae YB1, had no antagonistic effects on Neoscytalidium dimidiatum, and M. phyllosphaerae YB1 had no antagonistic effect on P. litchii.

Volatile compounds produced by the eight bacterial strains B. velezensis XYAH1 and XYCJ12, B. subtilis XYAJ8 and AJ4, B. creceus XYDH11, B. altitudinis J1, P. monsensis H16, and M. phyllosphaerae YB1, were identified and quantified using GC/MS following headspace solid-phase microextraction (HS-SPME). The data showed that there was no significant difference in the inhibitory effects of strains H16, J1, XYDH11, and XYAJ8 grown in the four different media (Supplementary Table S9). Further, the mean inhibition rate of the eight bacteria grown on LB plates (41.7%) was the highest, whereas the mean inhibition rates on NA, PSA, and PDA plates were 33.5, 31.4, and 33.8%, respectively. Therefore, LB was selected for the analysis of bacterial VOCs. P. monsensis H16, which showed the highest inhibition rate, was dual-cultured with C. fructicola in two-sealed plates for 24, 48, 72, 96, and 120 h, and the results confirmed that 24 h was the best time (data not shown).

Based on the treatment and bacterial volatile-profile analysis, 47 compounds were identified after analyzing the uninoculated sterile medium, and each strain was cultured in the medium using the NIST reference database comparison with similarity above 90% (Supplementary Table S10). The eight selected bacterial isolates with diverse antifungal activity against C. fructicola indicated different chemical classes, including ketones, alkanes, alkenes, alcohols, pyrazines, and phenols; however, different strains of the same species did not have the same composition, such as B. velezensis XYAH1 and XYCJ12. Additionally, VOCs in the ketone chemical group were present in the volatile profiles of most bacterial isolates and were generally among the most abundant chemical groups (Table 1). The most common VOCs produced by the Bacillus isolates were 2-heptanone, 6-methyl-2-heptanone, 4-ethyl-decane, 2-decanone, 2-dodecanone, 2-heptadecanone, 2-heptanone, 2-hexadecanone, 2-nonanone, 2-tridecanone, butylated hydroxytoluene, D-limonene, dibutyl phthalate, and benzyl alcohol. The abundant VOCs methoxy-phenyl-oxime and dimethyl-silanediol produced by B. subtilis AJ4, were distinguishable from the VOCs produced by all other Bacillus isolates. In P. monsensis H16, volatile compounds especially ketones, were the most abundant and varied compounds particularly 2-nonanone, 2-undecanone, and tridecanone. In contrast, M. phyllosphaerae YB1 mainly contains benzyl alcohol, 4-ethyl-decane, phenylethyl alcohol, and cetene (Supplementary Table S10).