A soil-borne bacterium HK235 was isolated from the rhizosphere soil of tomato plants and deposited at the Korean Agricultural Culture Collection (KACC) under accession no. KACC 81044BP. The isolated bacterial strain was maintained on tryptic soy agar (TSA; BD Difco, Sparks, MD). Additionally, tryptic soy broth (TSB), Luria-Bertani broth (LB), malt extract broth (MB), and nutrient broth (NB) purchased from BD Difco were used for the liquid culture. The HK235 cells were suspended in 20% glycerol solution and kept at –20^°C until further use. For the in vitro antifungal activity assay, B. cinerea NY76 (accession no. KACC 48736) provided by the KACC was used and maintained on potato dextrose agar (PDA) medium. For the sporulation, this fungal pathogen was grown on a PDA medium at 20^°C for 6 days under darkness and then incubated under a 14 h light/10 h dark cycle of illumination for 4 days.

The genomic DNA of the HK235 strain grown in TSB at 30^°C for 48 h was extracted using QIAamp genomic DNA kits (Qiagen, Hilden, Germany), following the manufacturer’s instructions. The purity and concentration of the genomic DNA were determined with a 2,100 Bioanalyzer system (Agilent Technologies, Palo Alto, CA, USA). The whole genome of the HK235 strain was sequenced with a 20 kb SMRTbell library on the RS II sequencing platform (Pacific Biosciences, Menlo Park, CA, USA) using C4 chemistry with eight single-molecule real-time (SMRT) cells at Macrogen (Seoul, South Korea) (Chin et al., 2013). The cleaned reads were assembled de novo using Canu v1.7, and the result was polished with error correction by Pilon v1.23 (Pruitt et al., 2012; Seemann, 2014). All tools were run with default parameters unless otherwise specified. For the whole genome analysis, genome annotation was performed with the NCBI Prokaryotic Genome Annotation Pipeline (PGAP) v6.2 (Tatusova et al., 2016). Average nucleotide identities (ANI) with the currently available genome sequences of Chitinophaga spp.² were calculated using Oat 0.93 (Lee et al., 2016; Ciufo et al., 2018). The genome sequence of HK235 was deposited in the NCBI database with an accession code CP073766.

To examine the in vitro antifungal activity against B. cinerea, we investigated the mycelial growth and conidial germination of B. cinerea by treatment as previously described (Han et al., 2020). Briefly, a mycelial disk (5 mm in diameter) of B. cinerea was inoculated onto PDA medium containing different concentrations of culture filtrate (CF) or isolated active compound, and then, the radial growth of B. cinerea was investigated at 7 days post-inoculation (dpi). PDA plates containing 1% dimethyl sulfoxide (DMSO) were prepared as a negative control. For the inhibitory effect on the conidial germination of B. cinerea, microplate wells containing a conidial suspension (1 × 10⁵ conidia/mL of PDB) were treated with the CF and active compound, and then, after a 10 h incubation at 20^°C, the number of germinated conidia was counted by microscopic observation from a total of 100 conidia. All experiments were conducted twice with three replicates.

The HK235 strain was inoculated onto 400 mL of TSB medium in 2 L baffled Erlenmeyer flask and then incubated at 25^°C for 3 days with an agitation of 150 r/min. The resulting culture broth (5 L) was centrifuged at 10,000 × g for 30 min and passed through Whatman No.1 filter paper (Maidstone, United Kingdom). To isolate antimicrobial compounds, the CF was sequentially extracted with the same volume of n-hexane, ethyl acetate, and n-butanol. Each resulting layer was evaporated to dryness, yielding an n-hexane extract (0.1 g), ethyl acetate extract (0.5 g), n-butanol extract (6 g), and water extract (113 g). Based on an antifungal activity-guided bioassay against B. cinerea, the n-butanol extract was suspended in 1 L of water and then applied onto a Diaion HP-20 column (5 × 30 cm; Mitsubishi Chemical, Tokyo, Japan) with stepwise gradient elution of 0, 20, 40, 60, 80, and 100% aqueous acetone. The 80% aqueous acetone fraction (0.7 g) exhibiting the antifungal activity was subjected to reversed-phase flash column chromatography packed with LiChroprep RP-18 (40–63 μm; Merck, Kenilworth, NJ, United States), eluting with methanol at a flow rate of 5 mL/min. An active fraction (538 mg) was further purified with an LC-6AD high performance liquid chromatography system (Shimadzu, Kyoto, Japan) equipped with a Polaris C18-A column (250 × 21.2 mm, 5 μm; Agilent, Santa Clara, CA, United States). The column was eluted at a flow rate of 5 mL/min with 20–80% aqueous acetonitrile (containing 0.1% formic acid) at a linear gradient over a 50 min uninterrupted interval. The effluent was monitored with a SPD-M10Avp photodiode array detector (Shimadzu). Finally, the purified compound (37 mg) was obtained and kept at –20^°C until further analysis. The purified compound was analyzed with a Waters 515 high-performance liquid chromatography (HPLC) system equipped with a Phenomenex Luna C18(2) column (4.6 × 250 mm, 5 μm). The mobile phase consisted of solvent A (deionized water containing 0.1% formic acid) and solvent B (methanol) using a gradient elution as follows: 20–100% B at 0–15 min and 100% B at 15–30 min. The flow rate was 1 mL/min, and the effluent was monitored at 210 nm.

To identify the structure of the isolated compound, mass spectra were acquired with an Autoflex Speed TOF/TOF mass spectrometer (Bruker, Billerica, MA, United States), and 2,5-dihydroxybenzoic acid was used as a matrix. Spectra were obtained by positive ion detection and reflector mode. Monoisotopic masses were detected.

Amino acid composition analysis of the purified compound was performed with the Pico-Tag system (Waters, Milford, MA, United States) at the Korea Basic Science Institute (Daejeon, South Korea). The pure compound was hydrolyzed followed by labeling with phenylisothiocyanate (PITC) according to the Waters Pico-Tag protocol. The PITC-labeled hydrolysates were analyzed by a HPLC system equipped with a Waters Pico-Tag column (3.9 × 300 mm, 4 μm). The mobile phase consisted of solvent A (6% acetonitrile in 140 mM sodium acetate) and solvent B (60% acetonitrile in water) using the following gradient elution: 0–14% B at 0–9 min, 14–20% B at 9–9.2 min, 0–46% B at 9.2–17.5 min, and 46–100% B at 17.5–17.7 min. The flow rate was 1 mL/min, and the effluent was monitored at 254 nm.

To investigate the disease control efficacy against tomato gray mold, the HK235 CF was prepared at concentrations of 11, 33, and 100% by diluting it in distilled water. A purified compound was prepared by dissolving it in 1% DMSO. Chemical fungicide fludioxonil (50 μg/mL) and 1% DMSO were used as positive and negative controls, respectively. All samples contained 0.025% Tween 20 as a surfactant. Two-leaf stages of tomato seedlings (Solanum lycopersicum cv. Seokwang; Hungnong Seeds, Seoul, South Korea) were grown in a greenhouse at 25 ± 5^°C for 3–4 weeks. The treatments were applied onto the tomato seedlings by a spray method and then air-dried for 24 h. The treated tomato plants were inoculated with a conidial suspension (5 × 10⁵ conidia/mL) of B. cinerea by the spray method (Nguyen et al., 2022). The inoculated seedlings were incubated in a humidified chamber at 20^°C for 3–4 days. The disease control efficacy (%) was calculated based on the lesion as described previously (Ngo et al., 2021).

To evaluate the in vivo antifungal activity on fruits, strawberry fruits (Fragaria × ananassa cv. Seolhyang) were surface-sterilized using 0.01% (v/v) sodium hypochlorite for 1 min and then rinsed with running water. After air-drying, each fruit was punched by a 3 mm-diameter cork-borer and treated with CF and pure compound. The treated fruits were point-inoculated with 5 μL of the B. cinerea conidial suspension (1 × 10⁶ conidia/mL) and incubated in a humidified chamber at 20^°C for 5 days.

For the penetration assay, the onion epidermis was used for microscopic observation of the penetration process of B. cinerea in the presence of the HK235 CF or pure compound solutions. After treatment of each sample, the epidermis was inoculated with 20 μL of the B. cinerea conidial suspension (1 × 10⁵ conidia/mL) and then incubated in a moist container at 20^°C for 24 h. The inoculated epidermis was stained with lactophenol cotton blue and photographed.

For the quantitative data of this study, experiments were performed independently at least two times with three biological replicates unless indicated. Data were subjected to one-way ANOVA, and the means of the treatments were separated by Duncan’s multiple range test (p < 0.05) using the R-software (Version 4.0.5). All values were expressed as the mean ± standard deviation, and the significant differences (p < 0.05) were indicated with different letters.

During the course of screening for antagonistic bacteria, we found that a HK235 strain exhibited a promising antifungal activity against B. cinerea (Figure 1A). The HK235 strain grown on TSA medium was observed to be yellowish, smooth and opaque, and rod-shaped with a size of 4.1–4.3 μm in length and 0.36–0.43 μm in diameter (Figure 1B). Based on the 16S rRNA gene sequence (1,353 bp) of the HK235 strain, phylogenetic analyses revealed that the HK235 strain belongs to the genus Chitinophaga and is most closely related to C. flava K3CV102501 with 99.48% similarity (Supplementary Figure 1). Furthermore, we found that the whole genome of the HK235 strain contains 6,324 coding DNA regions, 76 tRNAs, 18 rRNAs, 3 ncRNAs, 5 non-coding RNAs, and 18 pseudogenes (Figure 1C). Considering that Chitinophaga spp. produces various bioactive metabolites, we analyzed the HK235 whole genome by antiSMASH, PRISM, and RiPPMiner (Skinnider et al., 2017; Agrawal et al., 2021; Blin et al., 2021). The results showed that the HK235 genome contains 25 secondary metabolite biosynthetic gene clusters (BGCs) for 10 non-ribosomal peptides (NRPs), three ribosomally synthesized and post-translationally modified peptides (RiPPs), seven terpenes, and others (aryl polyene, resorcinol, aminoglycoside, and siderophore) (Supplementary Tables 1, 2). With the genome sequences of eight known Chitinophaga strains, the ANI dendrogram-heatmap revealed that the HK235 strain was closely related to C. flava GDMCC 1.1325 (GenBank accession No. GCA_003308995) with the highest ANI value of 93.24% at the genome level (Figure 1D). Therefore, we identified the HK235 strain as a Chitinophaga flava.

When the HK235 strain was grown in various liquid media such as LB, MB, NB, and TSB, the TSB medium supported a higher yield of cell mass of the HK235 strain (Figure 2). In the paper disk-agar diffusion assay using the HK235 CF derived from each culture medium, the HK235 CF from the TSB medium exhibited the most effective antifungal activity against B. cinerea (Figure 2). Therefore, we decided to use 3 days-old HK235 culture grown in TSB medium to isolate the active compound.

To isolate the active compound, the HK235 CF was sequentially partitioned with n-hexane, ethyl acetate, and n-butanol. Of the solvent layers, the in vitro antifungal activity was exclusively observed from the n-butanol layer at a concentration of 1,000 μg/mL (Figure 3). Therefore, the n-butanol layer was purified through a series of column chromatography based on antifungal activity-guided fractionation. The resulting purified compound had a MALDI-TOF mass spectrum showing only one series of [M + H]⁺ and [M + Na]⁺ ions at m/z 1,853 and 1,875, respectively (Figure 3). A doubly charged [M–2H]^2– ion at m/z 925 observed in an ESI mass spectrum of the purified compound also supported a molecular weight of 1,852 Da (Supplementary Figure 2). Considering that the compound showed the end absorption in the UV spectrum and a positive ninhydrin reaction, the compound is likely to be a peptide containing free amino groups. When the amino acid composition of the compound was analyzed, we found that it consisted of Ala, Leu, Ile, Val, Phe, Ser, Lys, and Arg (Figure 3). For further identification of the chemical structure, MALDI-TOF/TOF-MS analysis was performed, but we did not obtain a resulting peptide sequence (Supplementary Figure 3). Despite the failure to obtain a peptide sequence, all spectroscopic data including the ¹H, ¹³C, DEPT, COZY, HMBC, and HSQC NMR data (Supplementary Figures 4–9) were totally different from those previously reported. Although structure elucidation was not successful here, the results of the amino acid composition and spectroscopic analysis suggest that the purified active compound is likely to be a new antifungal molecule. Thus, we named the active compound isolated from the HK235 CF as chitinocin.

To examine the in vitro antifungal activity of the CF and chitinocin, we investigated the inhibitory effects on conidial germination and mycelial growth of B. cinerea. When PDB media were supplemented with CF, the inhibitory activity for conidial germination was observed in a concentration-dependent manner, and a decreased inhibition activity was observed also at 9 hpi compared to the 6 hpi in the treatment group with the same concentration (Figure 4). Similarly, chitinocin with a MIC value of 200 μg/mL exhibited a germination inhibition activity in a concentration-dependent manner (Figure 4). Besides conidial germination inhibition, when PDA media were supplemented with the CF and chitinocin, these treatments inhibited mycelial growth: 20% of the CF and 200 μg/mL of chitinocin completely inhibited mycelial growth (Figure 4 and Supplementary Table 3). To explore the antimicrobial activity spectrum of chitinocin, we measured the MIC values against each of the five phytopathogenic fungi and bacteria. Our results showed that chitinocin has a broad-spectrum antimicrobial activity against C. coccodes, Phytophthora infestans, Agrobacterium tumefaciens, Burkholderia cepacia, Erwinia amylovora, and Ralstonia solanacearum with MIC values of 50, 12.5, 6.3, 50, 50, and 6.3 μg/mL, respectively (Supplementary Table 3).

To investigate disease control efficacy, we treated the plants with samples that were derived from the HK235 culture prior to the inoculation of the fungal pathogen. After 3 dpi, we found that the 1-fold and 3-fold diluted CFs significantly reduced the disease development of tomato gray mold with control values of 69 and 40%, respectively, compared to the non-treatment control (Figure 5 and Table 1). When chitinocin was applied onto the tomato seedlings, the treatments at a concentration of 250, 500, and 1,000 μg/mL exhibited disease control efficacies with control values of 48, 71, and 92%, respectively (Figure 5 and Table 1). In addition, treatment with the sample from the BuOH extraction layer containing chitinocin also showed a disease control efficacy with control values of 72% at a concentration of 3,000 μg/mL (Table 1). Our results show that all samples reduced in a concentration-dependent manner the development of tomato gray mold. Similarly, we observed that the CF and chitinocin were effective in reducing the gray mold development on mature strawberries. The growth of B. cinerea on strawberries was remarkably inhibited by the treatment with the CF (5, 10, and 20%) and chitinocin (250, 500, and 1,000 μg/mL) on the strawberry fruits (Figure 5). Meanwhile, for the non-treatment control, we observed that the B. cinerea-infected strawberries were completely covered with the mycelia of B. cinerea and subsequently turned a gray color, softened, and rotted within a few days (Figure 5).

To examine an effect of CF and chitinocin on fungal penetration, onion epidermal cells were treated with the samples prior to inoculation of B. cinerea conidial suspension. From the non-treatment control, we observed that conidia and hyphae on the surface were stained with lactophenol cotton blue, and invasively growing hyphae remained colorless also (Figure 5). However, the treatment with 5% CF exhibited few hyphae and conidia on the onion surface and no penetrating hyphae (Figure 5). At a higher concentration of the CF, conidia failed to germinate. Similarly, chitinocin completely inhibited the germination of the B. cinerea conidia on the onion epidermal tissues at all the tested concentrations (Figure 5).