Erwinia amylovora strain TS3128 was cultured on tryptic soy agar (TSA; pancreatic digest of casein 15 g/l, papaic digest of soybean 5 g/l, sodium chloride 5 g/l, agar 15 g/l; Difco, Detroit, MI, USA) medium in a 28°C incubator.

Full-blooming apple flowers (Hong-ro and Fuji cultivar) were collected from orchards in Chungbuk province. Five apple flowers were transferred into a 50 ml tube containing 20 ml of sterilized water and subjected to vortexing for 5 min. The flower suspension was diluted 10-fold in sterile water (SDW), and 100 μl of the dilution was spread on low-nutrient media, Reasoner's 2A agar medium (R2A; yeast extract 0.5 g/l, protease peptone 0.5 g/l, casamino acid 0.5 g/l, dextrose 0.5 g/l, soluble starch 0.5 g/l, sodium pyruvate 0.3 g/l, dipotassium phosphate 0.3 g/l, magnesium sulfate 0.05 g/l, agar 15 g/l; Difco), and TSA medium, and cultured at 28°C for 5 days. Microbial colonies with various morphologies and growth rates were randomly selected and cultured three times on R2A or TSA media to isolate 148 strains. Microorganisms were identified by reading the 16S rRNA sequence or the internal transcribed spacer sequence using the EzBioCloud database or the Basic Local Alignment Search Tool of the National Center for Biotechnology Information. All isolates were suspended in 10% glycerol and stored in a deep freezer (-80°C).

Murashige and Skoog minimal medium (180 µl, minimal media [MS]; glycine 2 mg/l, myo-inositol 100 mg/l, nicotinic acid 0.5 mg/l, pyridoxine HCl 0.5 mg/l, thiamine HCl 0.1 mg/l; Duchefa, Haarlem, the Netherlands) containing 1% of seven sugar sources (arabinogalactan, arabinose, galactose, glucose, sucrose, fructose, and sorbitol) were added to each well of a 96-well plate. A total of 148 isolates were individually suspended in MS medium and adjusted to an optical density (OD₆₀₀) of 0.2. Then, 20 μl of each suspension was added to MS medium supplemented with the respective sugar sources and incubated at 28°C with shaking at 120 rpm. The growth curve for each microorganism was generated by measuring the OD₆₀₀ every hour using a microplate reader (Hidex, Turku, Finland).

The antagonistic effect through the production of antimicrobial peptides or secondary metabolites by 10 microorganisms (5 bacterial and 5 yeast strains) selected through the growth assessment of sugar sources was confirmed using a double-layer agar assay. Microorganisms cultured in TSA medium at 28°C for 2 days were suspended in SDW and adjusted to an OD₆₀₀ of 0.2, and 5 μl of the microbial suspensions were dropped at intervals into the TSA medium and cultured at 28°C for 2 days. One milliliter of chloroform (Sigma Aldrich, Steinheim, Germany) was poured onto the lid of a Petri dish containing microbial colonies and placed inverted for 10 min to kill the microorganisms and the content of the cytoplasm to leak out. Microbial suspension (200 µl) prepared under the same conditions was mixed with 4 ml TSA soft agar (0.4% agar) and poured over the dead microbial colonies, allowed to harden, and then cultured at 28°C for 2 days. The antagonistic activity of the strains was evaluated by observing the formation of an inhibition zone around dead microbial colonies.

The crab apple (Malus prunifolia cv. Decobel) used in the experiment was stored in cold storage at 4°C and then moved to a greenhouse maintained at 25-30°C in April to May for bloom and use. When the pathogen and the four microorganisms selected from growth evaluation were cultured on TSA media and suspended in SDW to adjust the OD₆₀₀ to 0.2, the number of viable cells per ml was as follows: F2019-19, 3×10⁸ cfu/ml; F2020-09, 1×10⁶ cfu/ml; F2020-13, 1×10⁶ cfu/ml; F2020-15, 2×10⁸ cfu/ml; E. amylovora, 5×10⁸ cfu/ml. FireFighter-C, consisting of the four selected microorganisms, was prepared by mixing the microbial suspensions to obtain a final concentration of 1×10⁶ cfu/ml for bacteria (F2019-19 and F2020-15) and 2×10⁵ cfu/ml for yeast (F2020-09 and F2020-13). To prepare individual microbial inoculums, each microbe was suspended and diluted in SDW to obtain a final concentration of 1×10⁶ cfu/ml for bacteria and 2×10⁵ cfu/ml for yeast. Individual microbial inoculums and FireFighter-C (10 ml) were evenly sprayed on the flowers of the crab apples and stored at 27°C in a chamber maintaining saturated humidity. One day later, the pathogen was inoculated onto apple flowers by spraying a suspension of E. amylovora, which was prepared by suspending in SDW using bacterial cells cultured on TSA medium, once on the front of the flowers (5×10⁶ cfu/ml; approximately 80 μl). Five days after inoculation with the pathogen, the plants were removed from the chamber, and the number of diseased flowers was counted and recorded.

All statistical analyses were performed with the R software (version 4.2.2; R Core Team, 2020). A chi-square test was performed to statistically test the disease control effect according to biological control agents, followed by a Bonferroni post-hoc test.

A total of 148 isolates with different colony morphologies and growth times were obtained from apple blossoms, and as a result of identifying all microorganisms, 136 strains of bacteria and 12 strains of fungi (yeasts) were confirmed. Among these bacteria, Pseudomonas accounted for the largest proportion (30 species), followed by Curtobacterium (18 species), Bacillus (16 species), Pantoea (15 species), and Staphylococcus (12 species). The fungi used were Filobasium, Moesziomyces, Papiliotrema, and Sporidiobolus, all of which have been reported to be in the form of yeast states (Kot et al., 2021; Kwon-Chung, 2011; Mpakosi et al., 2022; Palmieri et al., 2021).

The ability of each microorganism to use seven sugar sources (arabinogalactan, arabinose, galactose, glucose, sucrose, fructose, and sorbitol) was evaluated by generating growth curves in a minimal medium containing each sugar. The ability to use arabinogalactan, which is known to play an important role in the initial colonization of flowers by E. amylovora, was given top priority. Among the microorganisms that preferred arabinogalactan, 10 strains that used the remaining six sugars more quickly than E. amylovora were selected (Fig. 1, Table 1).

To exclude strains that showed antagonistic activity against other microorganisms among the 10 selected microorganisms, an experiment was conducted using chloroform vapor to confirm the antagonistic activity through the production of antimicrobial peptides or secondary metabolites (Table 2). Strong antagonism was confirmed for F2019-49, F2019-65, and F2019-53 against each other and against F2019-19; therefore, they were excluded from the consortium candidates. F2019-52 was excluded because weak inhibition zones were created by F2019-64 and F2020-13. F2019-64 showed antagonism against F2019-19 and weak antagonism against F2019-65 and F2019-49. Therefore, it was excluded from the candidates because it is likely to produce antibacterial substances. F2019-48 and F2019-69 are yeastform fungi belonging to the genus Filobasidium and were excluded from the study because they contain pathogenic yeasts that can cause opportunistic infections in plants and humans. F2019-19 (Curtobacterium oceanosedimentum), F2020-09 (Sporidiobolus pararoseus), and F2020-13 (Papiliotrema laurentii) were selected as members of the microbial consortium. Therefore, we established a microbial consortium, FireFighter-C, for controlling fire blight on the flowers of apple trees, consisting of four microorganisms, including Pantoea agglomerans F2020-15, which has previously been shown to control fire blight through preemptive acquisition of arabinogalactan from flowers in a previous study (Lee et al., 2024). The four microorganisms that were finally selected do not have antagonistic effects on each other or on E. amylovora and have the ability to preempt the sugar source that the fire blight pathogen requires for colonization and infection (Supplementary Figs. 1, 2).

Individual microbial suspension and FireFighter-C, a mixture of four members, was first inoculated onto apple flowers, and then E. amylovora was inoculated 1 day later to test the effect of fire blight control on the flowers. Five days after inoculation with the pathogen, the number of diseased flowers was counted, and the ratio of diseased flowers to the total number of flowers was investigated (Fig. 2). Compared to the positive control (PC; 77.27%) inoculated with sterilized water instead of the biocontrol agent, F2019-19, F2020-13, and F2020-15 showed 43.83%, 80.68%, and 65.69% decreases in disease incidence, respectively. In contrast, F2020-09 showed a 4.17% increase in disease incidence. When FireFighter-C was applied, it showed the best effect, with an 89.75% reduction in the disease incidence rate compared with PC.