The bacterial and fungal strains used in this study are listed in Supplementary Table 1. Burkholderia arboris 1 (CGMCC No.16905), a strain isolated from Vernicia fordii (tung oil tree) in China, Botrytis cinerea B05.10, and Fusarium oxysporum f. sp. lycopersici were used in this study. Burkholderia arboris was cultured in nutrient agar (NA) medium at 28°C (Edwards et al., 1965), and F. oxysporum and B. cinerea were cultured in potato dextrose agar (PDA) medium at 28°C (Robbins, 1924). Escherichia coli were cultured in Luria-Bertani (LB) medium at 37°C (Elder, 1983). Kanamycin (Kan) was added to the medium at a final concentration of 50 μg/ml when screening for clones positive for E. coli, and at 500 μg/ml when screening for clones positive for B. arboris.

The Tn5 transposon mutant library was constructed as previously described (He et al., 2021). Briefly, a fragment of the kanamycin resistance gene (Kan) was amplified from plasmid pBBR1MCS-2 by PCR using primers MEKANA-F/R containing Tn5 mosaic end (ME) sequences. The PCR product was ligated into the pCE2 vector using a TA/Blunt-Zero Cloning Kit (Vazyme Biotech, Nanjing, China). After confirmation by sequencing, the Kan-ME fragment was transformed into B. arboris using an EZ-Tn5™ Transposase Kit (Lucigen, Middleton, WI, United States). Putative mutants were selected using 500 μg/ml kanamycin.

The plasmids used in this study are listed in Supplementary Table 1. The primer sequences used for plasmid construction are listed in Supplementary Table 2. Genomic DNA of B. arboris was extracted using a TIANamp Bacteria DNA Kit (Tiangen Biotech, Beijing, China). To knockout cobA, a fragment of the cobA gene in B. arboris genome was amplified by PCR. The PCR product was digested with BamHI and EcoRI restriction enzymes and then ligated into the suicide plasmid pJP5603 (Riedel et al., 2013), which was digested with the same enzymes. The ligation product was transferred into E. coli S17-1 λpir cells by heat shock. Positive clones were selected with 50 mg/ml kanamycin resistance and verified by sequencing the cobA insertion in the pJP5603-cobA plasmid. The resulting pJP5603-cobA plasmid was transferred into B. arboris competent cells with electroporation. Putative mutants were selected with 500 mg/ml kanamycin and verified by PCR with primers from 16S rDNA of B. arboris and the conserved R6K origin of replication (Riedel et al., 2013). The transcript levels of cobA in mutants were determined by qRT-PCR.

Tn5 insertion sites were identified using thermal asymmetric interlaced PCR (TAIL-PCR; Liu and Whittier, 1995). Arbitrary degenerated primers served as forward primers for all three rounds of PCR, and specific primers, including SP1, SP2, and SP3, were used as the reverse primers for primary, secondary, and tertiary rounds of PCR, respectively. The primer sequences are listed in Supplementary Table 2. The primary-round PCR product was 100-fold diluted and used as a DNA template for the secondary PCR reaction, which was 100-fold diluted and used as the template for the tertiary PCR reaction. Cycling conditions were as previously described.

Growth curves were generated as previously described (Xian et al., 2020). Bacterial suspension was centrifuged at 5,000 rpm for 5 min after overnight culture at 28°C. The pellet was resuspended in ddH₂O and adjusted to an OD₆₀₀ of 0.005 using fresh medium. Absorbance at OD₆₀₀ was measured with a microplate spectrophotometer (Multiskan™ FC, Thermo Fisher Scientific, Waltham, MA, United States) every 30 min for 36 h to obtain the growth curves. The growth rate, the doubling time and the lagging time were calculated.

Swimming ability was determined as previously described (Bahar et al., 2009). Bacterial strains were cultured overnight at 28°C until OD₆₀₀ reached 0.6. Thereafter, the bacteria suspension (2 μl) was point-inoculated on the center of a plate containing semi-solid LB broth with 0.3% agar, and the diameter of the swimming zone was measured 2 days after inoculation.

Biofilm formation assay was conducted using the crystal violet staining method (Holmberg et al., 2009). Bacterial strains were cultured overnight until the OD₆₀₀ reached 0.5, and then transferred to 96-well cell plate at 100 μl per well and cultured at 28°C for 72 h without shaking. Thereafter, the wells were washed with sterile water twice and stained with 125 μl 0.1% crystal violet (w/v) for 30 min to quantitatively measure the attached biofilms. After removing the dye, wells were thoroughly dried at 37°C, and 1% sodium dodecyl sulfate (SDS) was added to dissolve the biofilm at 200 μl per well. The quantity of biofilm formed was determined at an absorbance of 570 nm using a Multiskan™ FC microplate spectrophotometer (Thermo Fisher Scientific).

Bacterial culture (300 ml) was centrifuged at 4,000 rpm for 20 min after culturing overnight, and the supernatant was mixed with an equal volume of ethyl acetate. The solution from the organic phase was transferred to a conical flask containing anhydrous Na₂SO₄ to remove residual moisture, and was finally filtered through a triangular funnel with cotton. Thereafter, the organic phase filtrate was transferred to a pre-weighed collecting bottle, and the ethyl acetate was evaporated with an IKA® RV 10 digital rotary evaporator at 42°C, 80 rpm under control of IKA® MVP 10 basic compact vacuum pump (IKA, Staufen, Germany). The crude extract was dissolved in 1 ml ethyl acetate.

For determination of antifungal activity of B. arboris bacteria, fungal mycelial disks and fungal spore suspension were used as described previously with slight modifications (Kumar et al., 2018). Fungal disks and B. arboris were co-cultivated on PDA medium with a distance of 30 mm from each other. The inhibitory zone was measured 6 days after co-cultivation. To prepare the fungal spore suspension, the fungal mycelia and spores on the agar plate were washed off with sterile ddH₂O and then filtered through two layers of sterile gauzes. The concentration of spore suspension was determined with a hemocytometer. For determination of antifungal activity of B. arboris extracts, fungal spore suspension (1.0 × 10⁵ CFU/ml, 500 μl) was evenly spread on PDA medium, and then 8 mm round holes were removed from the solid medium using a sterile punch, followed by the addition of 20 μl of melting PDA medium to each hole to seal the bottom. After cooling, 150 μl of crude B. arboris extract was added to the holes, and the diameters of the inhibition zones were measured 6 days later.

Hafk peptides with D-amino acid residues were synthesized by ChenPeptide Biotech (Nanjing, China). LPE (CAS # 53862–35-4) was purchased from Aladdin Ltd. (Shanghai, China). The antifungal activity of Hafk peptides and LPE was determined as described previously with slight modifications (Shi et al., 2010). Briefly, a stock solution of the Hafk peptides (100 mM) and LPE (100 mM) was prepared using ddH₂O and ethyl acetate, respectively. PDA medium was added to a 12-well cell plate at 2 ml per well, and then 50 μl of different concentrations of the peptide or LPE solution were evenly spread on the surface of the medium. Thereafter, fungal mycelial disks were placed at the center of the dried medium, and the diameters of fungal mycelial growth were measured 3 days later.

Untargeted metabolomic analysis of the bacterial sample was performed at Metware Biotechnology Co., Ltd.¹ as previously described (Oh et al., 2021). Bacterial samples were collected by centrifugation, and then a 500 μl solution (methanol: water = 4: 1, v/v) containing an internal standard was added to each bacterial sample and vortexed for 3 min. Samples were placed in liquid nitrogen for 5 min, dried on ice for 5 min, thawed on ice, and vortexed for 2 min. This freeze–thaw cycle was repeated thrice. Then they were centrifuged at 12,000 rpm for 10 min at 4°C, and 300 μl of supernatant was collected and kept at −20°C for 30 min, followed by further centrifugation at 12,000 rpm for 3 min at 4°C. Thereafter, 200 μl aliquot of the supernatant was collected for UPLC-QTOF/MS analysis. The analytical conditions were as follows: column, Waters ACQUITY UPLC HSS T3 C18 (1.8 μm, 2.1 mm × 100 mm); column temperature, 4°C; flow rate, 0.4 ml/min; injection volume, 2 μl; solvent system, water (0.1% formic acid), and acetonitrile (0.1% formic acid); gradient program, 95: 5 V/V at 0 min, 10: 90 V/V at 11.0 min, 10: 90 V/V at 12.0 min, 95: 5 V/V at 12.1 min, and 95:5 V/V at 14.0 min. Mass spectrometric analysis was performed using a quadrupole time-of-flight mass spectrometer in the positive and negative ion modes. The gas flow rate was set at 8 L/min; gas and sheath temperatures were 325°C; the fragmetor was 135 V; and the nebulizer was 40 V. The ESI voltage was 2,500 V in the positive-ion mode and 1,500 V in the negative-ion mode.

The original data file acquired by UPLC-QTOF/MS was converted into the mzML format using ProteoWizard software. Peak extraction, alignment, and retention time corrections were performed using the XCMS program. The “SVR” method was used to correct the peak area. Peaks with detection rate lower than 50% in each group samples were discarded. Subsequently, metabolic identification information was obtained by searching the laboratory’s self-built database, integrated public database, AI database, and metDNA. Differential metabolites between wild type and mutant B. arboris were determined at variable importance in projection (VIP) ≥ 1, p < 0.05 (Student’s t-test), and absolute log₂ fold change (FC) ≥ 1.0. VIP values extracted from the OPLS-DA results, which also contained score plots and permutation plots, were generated using MetaboAnalystR package in R software. The data were log transformed (log₂) and mean centered before OPLS-DA. A permutation test (200 permutations) was performed to avoid overfitting.

Hafk peptide in the bacterial supernatant was identified by LC–MS analysis. The bacterial supernatant (100 ml) was freeze-dried, dissolved in 4 ml 40% acetonitrile, and then subjected to LC–MS (6400 Series Triple Quadrupole LC–MS by Agilent). The standard sample is the synthesized Hafk peptide (Purity: 81.98%).

Burkholderia spp. are highly resistant to several antibiotics. Therefore, the sensitivity of B. arboris to common antibiotics was examined to facilitate genetic analysis of the species. The growth of B. arboris was suppressed by 500 μg/ml of Kan, 10-fold higher than the requirement for most bacteria; in contrast, streptomycin (Strep), tetracyclines (Tet), spectinomycin (Spe), and chloroamphenicol (Chl) did not inhibit the growth of B. arboris (Supplementary Figure 1). Subsequently, plasmids carrying Kan-resistant and mCherry fluorescent genes were transferred into competent B. arboris cells using electroporation (Supplementary Figure 2). Positive clones were selected using NA media containing 500 μg/ml of Kan, and all selected clones displayed pink color under bright field and fluorescent conditions (Figure 1). The pink color did not decrease after 20 generations of subculture. Overall, these results indicated that 500 μg/ml Kan can be used as a selective antibiotic for B. arboris. Transposon random mutagenesis technology was used to generate B. arboris mutant library. The Kan gene fragment (795 bp) was amplified from a plasmid (Supplementary Figure 3) by PCR, ligated to adapters containing a Tn5 mosaic end (ME, 19 bp), and the resulting Kan-ME fragments (833 bp) were introduced into B. arboris along with transposases (Figure 2A). Overall, 3,000 putative transposon insertion mutants were isolated from Kan-resistant plates.

A plate confrontation experiment was performed to identify B. arboris mutants with weak antifungal activity against Fusarium oxysporum. Wild type B. arboris exhibited high antagonistic activity against F. oxysporum (Supplementary Figure 4), whereas three out of 3,000 putative transposon insertion mutants showed reduced antagonistic activity (Supplementary Figure 5). Among the three, M464, with a little yellow color, displayed the weakest inhibitory effect against F. oxysporum, as evidenced by the small diameters of the inhibition zones (Figures 2B,C); therefore, it was selected for further analysis.

The Kan fragment from the M464 mutant was amplified by PCR, and the resulting PCR product was electrophoresed on an agarose gel. As expected, 833 bp of PCR fragments were observed in the gel (Supplementary Figure 6), confirming that the Kan sequence were inserted into the M464 genome. Thermal asymmetrical interlaced PCR (TAIL-PCR) was performed to amplify the flanking sequence adjacent to the border of Kan gene to identify the insertion site of Kan sequence in the M464 genome. Three specific primers with high annealing temperatures were designed from the Kan gene: SP1, SP2, and SP3. TAIL-PCR was performed using nested specific primers in three subsequent rounds of reactions, together with six arbitrary degenerate (AD) primers (Figure 3A). A specific band of approximately 1,000 bp was obtained in the secondary rounds of PCR with the SP2 primer when the M464 template was amplified, but not the wild type B. arboris genomic DNA (Figure 3B). Two bands were obtained during the tertiary rounds of PCR using the SP3 primer (Figure 3B). Since the PCR products from the tertiary rounds of PCR are supposed to be smaller than those from the secondary round, we purified the small band (around 800 bp) and sequenced it using the primer SP3. After alignment with the B. arboris genome sequence using the nucleotide Basic Local Alignment Search Tool (BLAST), we found that Kan sequence was inserted in the cobA gene in B. arboris chromosome (Figure 3C). Subsequently, the insertion site of Kan was confirmed by PCR using specific primers (Figure 3D). Overall, these results suggest that the cobA gene is mutated in M464 mutants.

A cobA knockdown mutant was generated by site-specific mutagenesis to verify that cobA is important for the antifungal activity of B. arboris against F. oxysporum. To achieve this, cobA sequence was amplified from the B. arboris genome and ligated to the enzyme-digested mobilizable suicide plasmid pJP5603. Knockdown mutants were generated by homologous recombination and screened for resistance to high concentrations of Kan. Once the ∆cobA mutant had been verified, a confrontation plate test was performed to evaluate the antifungal activity of ∆cobA against F. oxysporum. The ∆cobA mutants showed significantly lower antifungal activity against F. oxysporum compared to the wild type, but higher than that of the M464 mutant (Figures 4A,B). Additionally, unlike M464, ∆cobA mutants did not show a yellow color, probably because of other mutations in M464. Nevertheless, deletion of cobA impaired the antifungal activity of B. arboris.

Furthermore, the potential mechanisms by which cobA mutation suppressed the antifungal activity of B. arboris were examined. First, we determined whether the cobA mutation affected the growth of B. arboris by comparing the growth curves of the wild type and ΔcobA mutant for 36 h, and found that the growth of ΔcobA mutant, including the doubling time and lagging time, had no substantial differences from that of the wild type (Supplementary Figure 7 A; Supplementary Table 3). Additionally, biofilm formation assay was performed, and it was observed that cobA mutation did not alter biofilm formation by B. arboris (Supplementary Figure 7 B). Moreover, swimming ability assay showed that the ΔcobA mutant had significantly higher swimming motility than wild type B. arboris (Supplementary Figure 7 C). Since ΔcobA mutant displayed reduced antifungal activity, it is unlikely that the increased swimming motility of the mutant was the major cause of the impaired antifungal activity. Overall, these results suggest that cobA mutation did not affect bacterial growth and physiological characteristics. Therefore, we hypothesized that the cobA may be involved in the production of antifungal compounds by B. arboris. To verify this hypothesis, the antifungal activities of metabolic crude extracts of the wild type and mutant B. arboris were evaluated. The crude extract from wild-type B. arboris significantly inhibited the growth of F. oxysporum, whereas that of ΔcobA mutant did not (Figure 4C), confirming that cobA is important for the production of antifungal compounds.

Untargeted metabolomics was performed to identify the antifungal compounds in B. arboris, using UPLC-QTOF/MS, and differential metabolites between B. arboris and the ΔcobA mutant were identified (Supplementary Table 4). Compared with wild type B. arboris, 17 compounds were identified in the ΔcobA mutant, among which three were downregulated (Table 1). The cobA gene, also known as cysG (Spencer et al., 1993; Woodcock et al., 1998), has been shown to be involved in sulfur and selenium metabolic pathways (Storbeck et al., 2009; Park et al., 2011; Moore et al., 2012). Accordingly, it was observed that several of the differentially expressed metabolites between B. arboris and the ΔcobA mutant contained these elements, indicating that cobA mutation affected the synthesis of the metabolites. The cobA gene also contributes to the biosynthesis of siroheme and vitamin B12 (Woodcock et al., 1998; Rodionov et al., 2003); however, the ΔcobA mutant did not show significantly reduced levels of heme and vitamin B12, which may be due to the detection sensitivity of the equipment or redundancy of cobA gene in B. arboris. Therefore, the antifungal activities of the three downregulated compounds in the ΔcobA mutant were evaluated.

The three downregulated compounds were dimethyl diselenide, lysophosphatidylethanolamine (LPE or lysoPE), and His-Ala-Phe-Lys (Hafk peptide). Dimethyl diselenide is volatile with a low boiling point of 57°C at 760 mmHg (Lide, 2007). However, we found that B. arboris did not inhibit the growth of F. oxysporum in a petri dish with separate compartments, suggesting that the antifungal compounds are not volatile (Supplementary Figure 8). Similarly, 10 and 100 mM of LPE did not inhibit the growth of F. oxysporum on PDA media (Supplementary Figure 9), indicating that both dimethyl diselenide and LPE were not the antifungal compounds produced by B. arboris. Overall, these results indicated that Hafk peptide may be the compound with antifungal activity.

To investigate whether the short peptide Hafk possesses antifungal activity, we first determined the stability of crude extracts from B. arboris suspensions. The crude extracts were heated for 30 min at 40, 50, and 60°C respectively, followed by a plate confrontation experiment to verify the antifungal activity against F. oxysporum (Supplementary Figure 10 A). Preheating at high temperatures did not reduce the antifungal activity of B. arboris. Additionally, the crude extract exhibited antifungal activity after treatment with proteinase K, which catalyzes the degradation of L-isoforms of amino acids (Supplementary Figure 10 B), indicating that the antifungal compounds produced by B. arboris are not peptides or amino acids with L-isoforms (Kannan et al., 2020). Therefore, we synthesized a Hafk peptide containing all the D-amino acids, and its antifungal ability against F. oxysporum was assessed in a 12-well cell plate. PDA media (2 ml) was added at the bottom of each well, and then 50 μl of different concentrations of Hafk peptide solution was spread evenly on the surface. A F. oxysporum mycelium disk was placed at the center of each well and the mycelial growth zone was measured 72 h after placement. Fusarium oxysporum growth was significantly inhibited by the addition of 10 and 100 mM Hafk peptide compared with the control group; however, concentrations lower than 10 mM did not significantly inhibit fungal growth (Figures 5A,B). To detect whether Hafk peptide was secreted by B. arboris, the crude extracts from bacterial supernatant were subjected to LC–MS analysis. Synthesized Hafk peptide was used as a standard. The chromatogram in standard solution was overlapped with that in the supernatant of B. arboris, suggesting that Hafk peptide is present in the supernatant (Supplementary Figure 11). To test whether the Hafk peptide antagonizes other phytopathogenic fungi, its antifungal activity against B. cinerea was evaluated. Consistent with the findings in F. oxysporum, 10 mM of Hafk peptide significantly inhibited the growth of B. cinerea (Figures 5C,D). Similarly, B. arboris inhibited the growth of B. cinerea in the confrontation assay (Supplementary Figure 12). Overall, these results indicated that Hafk peptide possesses antifungal activity.