Strains of polyyne-producing bacteria used in this study (Figure 1 and Table 1) were obtained from the Cardiff University Burkholderia culture collection (Mullins et al., 2020) and other recognised strain repositories [The Belgium Co-ordinated Collections of Microorganisms/Laboratorium voor Microbiologie, Universiteit Gent (BCCM/LMG); The Burkholderia cepacia Research Laboratory and Repository (BcRLR; Lipuma, 2010); Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures GmbH (DSMZ)] and stored at −80°C in Tryptone Soya Broth (TSB; Oxoid™) containing 8% (v/v) dimethylsulfoxide (DMSO; Sigma-Aldrich). Cultures were revived onto Tryptone Soya Agar (TSA; Oxoid™) in Petri dishes and incubated at 30°C for 24 h. Bacterial cultures were routinely streaked to single colonies on TSA to check for purity. Overnight liquid cultures were prepared by inoculating 5 mL of TSB with confluent growth from a fresh TSA plate, incubated at 30°C on a rocking platform (150 rpm) and used as bacterial inoculum in specialised metabolite induction, in vitro antagonism, and biocontrol assays.

Insertional mutants Burkholderia ambifaria BCC0191::ccnJ (Mullins et al., 2019) and Burkholderia gladioli BCC1697::cayA (Jones et al., 2021) with their respective fatty acyl-AMP ligase-encoding gene disrupted from the polyyne BGC were maintained as above, with the exception that 50 μg mL⁻¹ trimethoprim was included in the media. G. ultimum var. ultimum MUCL 16164 was obtained from BCCM/MUCL (Mycothèque de l’Université Catholique de Louvain) collection and grown on potato dextrose agar (PDA; Oxoid™) plates at 22°C. For long term storage, cultures were maintained on PDA slants at 4°C.

Polyyne BGC architecture figures were visualised using Clinker v0.0.21 (Gilchrist and Chooi, 2021), and metabolite structures created with ChemDraw Professional 16. BGC relatedness was then demonstrated by phylogenetic analysis. The amino acid sequence of the fatty acyl-AMP ligase present in every polyyne BGC was aligned using MAFFT v7.505 (Katoh and Standley, 2013), and a phylogenetic tree constructed with RAxML-NG v1.0.3 (Kozlov et al., 2019) using the LG + G8 + F model with 100 bootstraps.

Production of polyynes and other specialised metabolites were induced from bacteria by growing them at 22°C on two different media. Basal salts medium supplemented with glycerol (BSMG) as previously described (Mahenthiralingam et al., 2011; Webster et al., 2020a) and pea seed exudate medium (PEM; Mullins et al., 2021; Petrova et al., 2022). PEM was designed as a biomimetic medium to represent nutrient conditions during pea seed germination and was made as follows. Early Onward variety Pisum sativum seeds (approx. 100 g) were washed three times with deionised water, and then suspended in ultrapure water made up to 500 mL. Seeds were incubated in the dark with agitation (40 rpm on a rocking platform) for 2 d at 22°C. After incubation, seed exudate was removed and filtered twice. First with a Whatman® glass microfibre GF/D grade filter to remove seed coat material, and second with a Whatman® glass microfibre GF/A grade filter to obtain a clear seed exudate. The filtered seed exudate was diluted with ultrapure water at a 1:1 ratio and mixed with 1.5% (w/v) purified agar (Oxoid™) prior to autoclaving.

Detection of specialised metabolites was conducted according to the rapid screening method described previously (Webster et al., 2020a). In brief, bacterial strains were streaked onto 20 mL BSMG or PEM agar plates in duplicate, and incubated for 3 d at 22°C. Following incubation bacterial growth was removed from the agar surface, and a 20 mm diameter disc cut from the centre of the plate. The agar disc was placed into a 30 mL wide-mouth amber glass bottle with 0.5 mL dichloromethane and agitated on a rocking platform (40 rpm) for 2 h. Dichloromethane extracts were analysed by high performance liquid chromatography (HPLC) on a Waters® AutoPurification™ HPLC system fitted with a reversed-phase analytical column (Waters® XSelect CSH C18, 4.6 × 100 mm, 5 μm) and a C18 SecurityGuard™ cartridge (Phenomenex) in series. Detection of compounds was by absorbance at 210–400 by a photo-diode array detector (PDA). Mobile phases consisted of (A) water with 0.1% formic acid and (B) acetonitrile with 0.1% formic acid with a flow rate of 1.5 mL/min. Elution conditions were as follows: 0–1 min, 95% phase A/ 5% phase B; 1–9 min, gradient of phase A from 95 to 5% and gradient of phase B from 5 to 95%; 10 to 11 min, 5% phase A / 95% phase B; 11–15 min, 95% phase A / 5% phase B. Known specialised metabolites were identified by HPLC peak retention times and UV absorbance characteristics, and by referencing these to internal standards characterised by High Resolution Liquid Chromatography-Mass Spectrometry (LC–MS) and Nuclear Magnetic Resonance (NMR) as described (Mahenthiralingam et al., 2011; Song et al., 2017; Mullins et al., 2019; Webster et al., 2020a; Jones et al., 2021). Purified pyrrolnitrin (Sigma-Aldrich) and tropolone (Sigma-Aldrich) were used as additional standards to confirm HPLC detection and peak retention times for these specialised metabolites. Peak heights were calculated using MassLynx V4.1 software.¹

Antagonism assays for polyyne-producing bacteria were performed against a panel of susceptibility organisms: Pectobacterium carotovorum LMG 2464 (Gram-negative bacterium), Staphylococcus aureus NCTC 12981 (Firmicutes Gram-positive bacterium), Clavibacter michiganensis DSM 46364 (Actinobacteria Gram-positive bacterium), and Candida albicans SC 5314 (fungus) as described (Mahenthiralingam et al., 2011; Mullins et al., 2019; Webster et al., 2020b). In brief, polyyne-producing bacterial strains were grown overnight at 30°C in TSB, spotted (2.0 μL bacteria) onto BSMG or PEM agar plates and incubated at 22°C for 3 d. Polyyne-producing bacteria were then killed by chloroform exposure for 3 min, overlaid with susceptibility organism-seeded (0.4% [v/v] bacteria) half-strength iso-sensitest agar (Oxoid™) and the overlay plate incubated at 30°C or 37°C for 24 h. The diameter of the inhibition zone was then measured through the centre of the polyyne-producing bacterium. The mean inhibition zone was calculated from two plates per treatment.

Polyyne-producing bacterial strains were grown overnight at 30°C in TSB, spotted (5.0 μL bacteria) onto BSMG and PEM agar plates as four evenly spaced drops, allowed to dry and incubated at 22°C. After 24 h incubation, a 5.0 mm diameter plug of leading-edge growth of G. ultimum was placed in the centre of the four bacterial spots and the plates incubated again at 22°C for a further 48 h. The distance between the leading-edge of G. ultimum and the centre of each bacterial colony was measured, and the mean inhibition zone calculated for each treatment.

Globisporangium ultimum infested soil (a non-sterile potting mix) was prepared as previously described (Mullins et al., 2019). Briefly, a fresh PDA plate (90 mm diameter) was inoculated with a 5.0 mm diameter PDA plug of G. ultimum, and incubated at 22°C for 3 d. The agar was cut into 1.0 cm² squares, added to 120 g potting mix composed of 5-parts generic compost (Levington Advance Pot and Bedding M2; ICL UK) and 1-part sand (Horticultural Sharp Sand, Melcourt Industries Ltd), and incubated at 22°C for 3 d. The G. ultimum infested potting mix was combined with fresh potting mix to achieve a 0.25% (w/w) infested soil. Pisum sativum seeds were either planted in G. ultimum infested soil or control un-infested potting mix contained within recycled pipette tip boxes (Starlab (UK), Ltd.). Each treatment consisted of 16 pea seeds split between two boxes with 200 g potting mix per box.

Polyyne-producing bacterial seed coats were prepared by growing overnight liquid cultures in TSB at 30°C. Bacterial cultures were washed, resuspended, and concentrated in phosphate buffer solution (PBS). The concentrated bacterial suspension was diluted to allow optical density measurement on a spectrophotometer (Jenway 6300 Visible Spectrophotometer) and then adjusted to 5.0 OD₆₀₀ nm, equivalent to approx. 1 × 10⁹ cfu (colony forming units) mL⁻¹. Seeds were submerged in the bacterial suspension resulting in a standardised coating of approximately 10⁷ cfu ml⁻¹ per seed (as demonstrated in Mullins et al., 2019), immediately planted, and watered with 30 mL deionised water. Plants were grown at 22°C in propagators to maintain high humidity and placed in a Fitotron® plant growth chamber set at 16:8 h photoperiod and 40% relative humidity for 14 d. All bacterial strains were assayed for biocontrol potential in a minimum of 2 experiments (at least 2 × 16 seeds per treatment), along with uninoculated seeds (PBS only) and Burkholderia ambifaria BCC0191 coated seeds as controls in every experiment. Treatments were assessed over 12 independent experiments (Table 2) due to restrictions in growth chamber capacity.

The ability of three B. ambifaria strains (BCC0191, BCC1237, and BCC1259) to colonise the root system of P. sativum was assessed. Seeds were coated with bacterial cell suspensions of approximately 1 × 10⁹ cfu mL⁻¹ and planted in potting mix as described above. Following 7 d of growth, the seedlings were removed and the first 2 cm segment of root from the seed was excised. This root section was macerated with a pestle in a 1.5 mL Eppendorf tube with 1 mL PBS and serially diluted in PBS. Serial dilutions were plated onto Burkholderia cepacia selective agar (BCSA, Oxoid™) and incubated overnight at 37°C to determine the B. ambifaria cfu per root section.

To assess soil persistence, 10 μL of approximately 1 × 10⁹ cfu mL⁻¹ of each B. ambifaria strain (BCC0191, BCC1237, and BCC1259) was added to 1 g of hydrated potting mix (50% water content) in 25 mL sterile universal bottles. Each treatment was replicated six times, and initially three replicates per strain were serially diluted in PBS to determine the recoverable cfu g⁻¹ at day 0. The remaining inoculated potting mix samples were incubated at 22°C for 7 d. Following incubation each 1 g of potting mix was serially diluted in PBS and plated onto BCSA to calculate the remaining cfu g⁻¹. Control samples (no added bacteria) were also set-up and treated as described above. No Burkholderia growth was identified on BCSA from all control samples.

The relative expression of the cepacin BGC across three B. ambifaria strains (BCC0191, BCC1237, and BCC1259) was determined by reverse-transcriptase (RT)-PCR. P. sativum seeds were coated with B. ambifaria and grown in potting mix as described above or on Whatman® filter paper grade 1 soaked in ultrapure water. Following 3 d of growth the seedlings were removed and 1 cm of root was excised, macerated, and pooled from three seedlings per treatment. Total RNA was extracted from the pooled samples using the FastRNA™ Pro Soil-Direct Kit (MP Biomedicals) and DNase I (RNase-free; New England Biolabs) treated according to the manufacturer’s protocol. RT-PCR was performed using the OneTaq® One-Step RT-PCR Kit (New England Biolabs) according to manufacturer’s protocols. PCR primers were designed to target one of the polyyne-associated desaturase genes in the cepacin BGC [ccnN; (Mullins et al., 2019)] and amplify a 514 bp product: Fwd: 5’-CTG TTC TGG GCA GGT ACG TT-3′ and Rev.: 5′-TGT CGT AGA AGT GGC AGT GG-3′. Thermal cycler conditions for RT-PCR were as follows: RT at 48°C for 15 min; initial denaturation at 94°C for 1 min; 35x cycles of denaturation at 94°C for 15 s, annealing at 60°C for 30 s, and extension at 68°C for 30 s; then a final extension at 68°C for 5 min. RNA extracted from B. ambifaria BCC0191 grown on BSMG and molecular grade water (Severn Biotech Ltd.) were used as positive and negative controls, respectively.

A systematic collection of 30 bacteria that all encoded polyyne BGCs (Mullins et al., 2021) was assembled and comprised: seven Burkholderia species (26 strains), Pseudomonas protegens (2 strains), and one strain of Trinickia caryophylli and Collimonas fungivorans (Table 1). To evaluate their ability to produce polyynes (Figure 1) and other specialised metabolites in vitro, they were grown on metabolite induction media and subjected to chemical analysis by HPLC (Webster et al., 2020a). Specifically, metabolite production after growth on a minimal medium, with glycerol as a carbon source (BSMG) known to induce specialised metabolites in Burkholderia species (Mahenthiralingam et al., 2011; Webster et al., 2020a), was compared to that seen on the novel biomimetic medium made from the exudate of pea seeds (PEM; Mullins et al., 2021; Figure 2). Of the 18 bacterial strains carrying the cepacin BGC, 15 of these strains produced cepacin on BSMG and PEM, while B. contaminans BCC0123 only produced detectable levels of cepacin on BSMG (Figure 2A and Table 1). B. ambifaria BCC0192 and B. vietnamiensis BCC1186 did not produce detectable levels of cepacin on either medium. Cepacin production levels were universally greater on BSMG compared to PEM, although the differences in metabolite production between the two media for a given strain varied considerably (Figure 2A). The composition of PEM was analysed using liquid chromatography-mass spectrometry by Lifeasible. ² A large fraction of identified compounds (35%) contained a glycerol moiety (glycerophosphocholines, monoacylglycerides, glycerophosphoethanolamines, glycerophosphates, glycerophosphoserines, and triacylglycerols; Supplementary Figure S1), and suggests that natural concentrations of glycerol-containing compounds found in PEM are important to induce bacterial metabolites similar to the effect of glycerol in BSMG (Mahenthiralingam et al., 2011).

A similar trend of higher production on BSMG was also observed for the antifungal compound, pyrrolnitrin (El-Banna and Winkelmann, 1998) in B. ambifaria and B. contaminans BCC0123 (Figure 2A). Interestingly, strains B. ambifaria BCC0192 and B. contaminans BCC0123 also produced detectable levels of two other unknown metabolite peaks on BSMG (retention time (RT) = 6.83 min, UV absorbance = 323 & 336 nm; RT = 7.17 min, UV absorbance = 300 nm), and B. ambifaria BCC1259 produced only one of these extra HPLC peaks (RT = 6.83). It is possible that these peaks could be attributed to the presence of hydroxyquinolones and bactobolins, as B. ambifaria BCC0192 is known to have BGCs that encode for both these molecules, while B. ambifaria BCC1259 has only the hydroxyquinolone BGC (Mullins et al., 2019; Table 1).

A wider range of known specialised metabolites (toxoflavin, sinapigladioside, enacyloxin, caryoynencin, gladiolin and bongkrekic acid) were detected across the seven B. gladioli strains screened (Figure 2B). The production levels of the B. gladioli metabolites were also generally higher on BSMG than PEM (Figure 2B). The polyyne caryoynencin was detected in extracts from all seven B. gladioli strains and represented a universally induced and dominant metabolite when they were grown on BSMG and PEM (Figure 2B). Intriguingly, toxoflavin, a broad range phytotoxin and known antifungal compound (Furuya et al., 1997; Li et al., 2019) shown to be ubiquitous in B. gladioli (Webster et al., 2020a; Jones et al., 2021), was readily induced on BSMG, but absent in five strains, and present at very low levels in two strains when they were grown on germinating plant-mimicking PEM agar (Figure 2B). Similarly, the induction of the respiratory toxin bongkrekic acid (Anwar et al., 2017) and the antibiotic enacyloxin (Song et al., 2017; Jones et al., 2021) was also abrogated on the biomimetic PEM (Figure 2B). The antifungal isothiocyanate, sinapigladioside (Dose et al., 2021), from B. gladioli strain BCC1883 was induced on both BSMG and PEM. Two further uncharacterised B. gladioli metabolite peaks were also identified by HPLC and these displayed differential production on BSMG and PEM. One compound (RT = 7.46, UV absorbance = 281 nm) from B. gladioli BCC1770 was observed exclusively on BSMG, while the unknown compound (RT = 7.77, UV absorbance = 293, 308, 338, 363 & 392 nm) from B. gladioli BCC1713 was only observed on PEM (Figure 2B).

Specialised metabolites including polyynes were detected in vitro for B. plantarii, T. caryophylli and P. protegens on BSMG and PEM (Figure 2B), but no compounds including the polyynes, collimonins (Kai et al., 2018), were identified for C. fungivorans Ter331 (Table 1) under the conditions tested. Caryoynencin production by B. plantarii and T. caryophylli was induced at higher levels on BSMG than PEM (Figure 2B) in a similar manner to B. gladioli. B. plantarii also produced high levels of the antimicrobial compound, tropolone (Guo et al., 2019) on PEM, and an unidentified novel peak on BSMG (RT = 1.60, UV absorbance = 325 nm). Tropolone has been identified as a phytotoxin that causes bacterial seedling blight of rice caused by B. plantarii (Azegami et al., 1987).

In contrast to the polyynes isolated from Burkholderia and Trinickia, elevated production levels of the newly described polyyne protegencin from P. protegens (Mullins et al., 2021) were observed on PEM when compared to BSMG (Figure 2B). The well-characterised P. protegens specialised metabolites pyoluteorin and 2,4-diacetylphloroglucinol (2,4-DAPG) (Neidig et al., 2011) showed differential production for both strains CHA0 and Pf-5 as follows (Figure 2B). Pyoluteorin production was 3-5-fold higher on BSMG compared to PEM, while 2,4-DAPG was not produced on BSMG but only detected at high concentrations on PEM agar (Figure 2B). In parallel to the response of B. gladioli, toxoflavin production in P. protegens was also diminished on PEM. The suppression of broadly toxic compounds such as toxoflavin and bongkrekic acid by PEM are important considerations if polyyne-producing bacteria are to be explored for use as biopesticides. It was therefore particularly encouraging to note that cepacin, caryoynencin, and protegencin were consistently induced and toxic metabolite production was suppressed on PEM, which mimics the pea plant root system (Figure 2).

To complement the detection of specialised metabolites by HPLC we investigated the in vitro antimicrobial activity (Webster et al., 2020b) of polyyne producing bacteria, again comparing BSMG and PEM agar growth media. Antagonistic activity (Figure 3A and Supplementary Figure S2) was examined against a panel of plant and human pathogens comprising an oomycete (G. ultimum), a fungus (Candida albicans), a Gram-negative bacterium (Pectobacterium carotovorum) and two Gram-positive bacteria (Staphylococcus aureus, Clavibacter michiganensis). All polyyne-producing bacteria possessed activity against G. ultimum when grown on BSMG, and overall, antagonism on BSMG was greater compared to the activity on PEM (Figure 3A). Growth on BSMG also induced greater antimicrobial activity against Ca. albicans compared to PEM. In contrast, S. aureus antagonism was similar or higher when polyyne-producing bacteria were grown on PEM compared to BSMG (Figure 3A). Inhibition of Pe. carotovorum varied between media and polyyne-producing species. B. ambifaria, B. contaminans and T. caryophylli exhibited stronger antagonism of the Gram-negative plant pathogen when grown on PEM, while B. vietnamiensis, B. diffusa and B. gladioli possessed greater Pe. carotovorum antagonism on BSMG. B. latens, C. fungivorans, and P. protegens lacked detectable Gram-negative activity on both media. Bioactivity against Cl. michiganensis was observed for B. gladioli, B. plantarii, P. protegens and surprisingly only one strain of B. ambifaria (BCC0192) on both BSMG and PEM. Interestingly, the caryoynencin-producing B. gladioli and B. plantarii strains possessed the greatest antimicrobial activity against all five susceptibility test organisms (Figure 3A), specifically the three plant pathogens (G. ultimum, Pe. carotovorum, and Cl. michiganensis).

To understand the influence of different strains and/or species, and distinct polyynes on the biological control ability of G. ultimum damping-off disease, a comparison of the polyyne-producing strains representing nine species was performed (Figures 3B,C). Biocontrol efficacy was assessed using an in vivo P. sativum (pea) germination assay challenged with G. ultimum damping-off disease, as applied to evaluate B. ambifaria strains (Mullins et al., 2019). The strains evaluated carried the BGC for one of the four characterised polyynes: cepacin, caryoynencin, collimonins, or protegencin. As described above all the strains, except C. fungivorans, had been demonstrated to produce their respective polyynes (Figure 2 and Table 1) and possess antimicrobial activity (Figure 3A) in vitro. A plant survival rate of 40% was defined as an arbitrary threshold for successful biological control efficacy, as this minimum protection was consistently achieved by the prototypic biopesticide strain B. ambifaria BCC0191 (Mullins et al., 2019). Of the seven B. ambifaria strains examined, only BCC0191 achieved P. sativum protection with mean survival rates of 40% or higher. While biocontrol was observed in the remaining six B. ambifaria strains, the P. sativum survival rate they achieved ranged between 3.1 and 35.4% (Figure 3B and Table 2). Despite the close evolutionary relationship between B. ambifaria and B. vietnamiensis as members of the B. cepacia complex (Jin et al., 2020; Mullins and Mahenthiralingam, 2021) and the ability of 6 B. vietnamiensis strains to produce cepacin in vitro (Figure 2A), B. vietnamiensis produced essentially no biocontrol of G. ulitimum in the pea protection assay (Figure 3B). All seven B. vietnamiensis strains achieved between 0 and 3.1% pea mean pea survival rates (Figure 3B). Other cepacin-producing species, such as Burkholderia diffusa and Burkholderia contaminans were more efficacious than B. vietnamiensis, but still conferred less than 40% P. sativum mean survival rates (Figure 3B).

Caryoynencin-producing B. gladioli and B. plantarii strains were shown for the first time to be effective at protecting P. sativum seedlings from G. ultimum. Three of the seven B. gladioli strains (BCC1697, BCC1710, BCC1883) elicited mean survival rates of >40%, and strain BCC1697 achieved up to 75% (mean 58.8%) survival (Figure 3B and Table 2). Interestingly, despite producing lower levels of caryoynencin than the three effective strains in vitro on BSMG and PEM (Figure 2B), strain BCC1770 produced mean protection against damping-off disease in vivo that was just below the minimum threshold and therefore could also have potential as a biocontrol agent (Figure 3B and Table 2). Burkholderia plantarii BCC0777, a caryoynencin-producing strain, conferred a mean pea survival rate of 64.6%, similar to that of the best performing B. gladioli strains (Figure 3B). The high protection provided by B. gladioli and B. plantarii as a seed coat in this biocontrol model is significant since both species belong to the group of Burkholderia better known to cause disease symptoms in a range of plant species (Maeda et al., 2006; Jones et al., 2021). B. gladioli are well known as a causative agent of rot diseases in onions and mushrooms (Jones et al., 2021), and their protection of peas against G. ultimum suggests that these bacteria do not elicit these traits against a young, germinating plant.

The remaining strains examined in the in vivo biocontrol assay were selected because they encoded two further polyyne metabolites, the collimonins and protegencin. Despite evidence of in vitro production of protegencin from both P. protegens strains on PEM, neither of the strains demonstrated significant protection against G. ultimum (Figure 3B; mean pea survival rate < 4.2%). No collimonin production was detected from C. fungivorans Ter331 in vitro (Table 1) and no protection was observed in the biocontrol assay (Figure 3B).

Effective biological control of damping-off disease by G. ultimum was lost when either cepacin or caryoynencin production was abolished through the disruption of the polyyne BGC by insertional mutagenesis of the fatty acyl-ACP ligase gene. The B. ambifaria BCC0191::ccnJ cepacin mutant produced a mean pea survival rate of 31.3% (Table 2 and Supplementary Figure S3), corresponding to previous results observed for this biopesticidal strain (Mullins et al., 2019). Expanding on the finding that cepacin is a key anti-damping off protection agent (Mullins et al., 2019), the considerably reduced protection (25.0%, Table 2 and Supplementary Figure S3) elicited by the B. gladioli BCC1697::cayA mutant for the first time implicates that the polyyne caryoynencin can also offer biocontrol protection against G. ultimum in vivo. The plant protective effects of different polyynes have also been reported previously. For example, protegencin (alternatively named protegenin) from Pseudomonas protegens strain Cab57 (Murata et al., 2021) was attributed as a key bacterial metabolite in the protection of cucumber seedlings against Pythium damping-off disease. In addition, the antifungal and antioomycete properties of collimonins have also led them to being suggested as biocontrol agents for suppressing plant pathogens (Fritsche et al., 2014; Kai et al., 2018), although in vivo efficacy of C. fungivorans to protect peas against damping-off was not demonstrated in our study (Figure 3B). Furthermore, closely related Paraburkholderia species have been genetically engineered to express the polyyne BGCs, cepacin and caryoynencin (Petrova et al., 2022).

There is a history of Burkholderia species being used as commercial biopesticides with several Burkholderia strains being registered in the United States from 1992 and applied as soil treatments to control phytopathogenic fungi, damping-off and other plant diseases (Parke and Gurian-Sherman, 2001). Commercial biopesticide products were sold under the brand names Deny®, Blue Circle®, and Intercept® and contained mixtures of three Burkholderia strains, M36, M54 and J82 (Parke and Gurian-Sherman, 2001). However, these products were eventually withdrawn from the market after a 1999 US Environmental Protection Agency (EPA) risk assessment that resulted in a moratorium being issued on the registration of new biopesticidal products containing members of the Burkholderia cepacia complex or any bacteria with “affinities to a human opportunistic pathogen” until they could be proven safe (Parke and Gurian-Sherman, 2001). Biopesticide M36 was subsequently found to be a B. cenocepacia strain, while strains M54 and J82 (the original strain name of BCC0191 characterised herein; Mullins et al., 2019), were both B. ambifaria isolates (Payne et al., 2005). The exact mode of action of these commercial biopesticides was not proven at the time of their commercial use, but from our work it is clear that most of their anti-damping-off control properties are derived from the production of cepacin at the rhizosphere of germinating crops (Figure 3B; Mullins et al., 2019).

With extensive specialised metabolite, genomic and taxonomic research on Burkholderia there has been renewed interest in repurposing these bacteria as biopesticides for several crops. This has included the use of new Burkholderia species or strains to control plant pathogens, such as Botrytis cinerea in grapevines (Esmaeel et al., 2020), Fusarium oxysporum in banana (Xu et al., 2020), Sporisorium scitamineum in sugarcane (Cui et al., 2020), and Rhizoctonia cerealis in wheat (An et al., 2022) via specialised metabolite production. In some cases, Burkholderia gladioli have also been found as endophytes in wild and ancient maize and shown to combat the fungal pathogen, Sclerotinia homoeocarpa (Shehata et al., 2016). Whereas Burkholderia sp. MSSP synthesizes 2-hydroxymethyl-chroman-4-one, demonstrating activity against Pythium, Phytophthora and Sclerotinia (Kang et al., 2004), and B. ambifaria strains inhibited phytopathogenic fungi through the emission of volatile organic compounds (Groenhagen et al., 2013). Alongside specialised metabolites, B. gladioli NGJ1 deploys a prophage tail-like protein secreted by a type III secretion system essential for mycophagy in Rhizoctonia solani (Swain et al., 2017).

B. ambifaria BCC0191 and B. gladioli BCC1697 also showed in vitro activity against the problematic wheat ‘take-all’ fungal pathogen (Palma-Guerrero et al., 2021), Gaeumannomyces tritici (Supplementary Figure S4). Recently, the mode of bacterial polyynes was demonstrated to inhibit a fungal-specific acetyl-CoA acetyltransferase in the first step of ergosterol biosynthesis (Lin et al., 2022). This indicates the potential use of polyyne-producing Burkholderia as biocontrol agents on a range of soil-borne fungal diseases that affect different crop species.

The comparative biological control experiments highlighted considerable variation in the biocontrol efficacy of the seven cepacin-producing B. ambifaria strains (3.1–50% mean pea survival rate; Table 2), with BCC0191 demonstrating the greatest protection against damping-off (Figure 3B). This led us to explore the potential reasons for intraspecies biocontrol variation, examining: (i) cepacin gene cluster expression at the rhizosphere, (ii) the rate of B. ambifaria root colonisation, and (iii) the persistence of the B. ambifaria strains within soil (a non-sterile potting mix) microcosms. Two B. ambifaria strains, BCC1259 and BCC1237, that performed poorly in biocontrol (Figure 3B), but exhibited similar bioactivity against G. ultimum on PEM (Figure 3A), were compared to the B. ambifaria biopesticide strain BCC0191. Both BCC1237 and BCC1259 exhibited biological control levels ≤12.5% in contrast to BCC0191 with a mean survival rate of 50.6% (Figure 3B and Table 2).

After 7 days of plant growth, B. ambifaria BCC0191 and BCC1237 exhibited similar root colonisation levels, with average counts of 4.6 × 10⁴ and 6.6 × 10⁴ colony forming units (cfu) root section⁻¹, respectively (Supplementary Figure S5B). In comparison, strain BCC1259 showed a consistently lower colonisation, with an average of 1.2 × 10⁴ cfu root section⁻¹ (Supplementary Figure S5B). Interestingly, cepacin production levels on PEM also showed that BCC1259 produced consistently lower amounts of the polyyne than both BCC0191 and BCC1237 (Supplementary Figure S5A). Variation in soil persistence was observed by comparing each strains’ viable count per gram of soil following a 7-day incubation within a soil microcosm (Supplementary Figure S5C). The less protective strains, B. ambifaria BCC1237 and BCC1259, had viable counts that were an average of 11.3% and 12.2%, respectively, of their initial inoculum, whereas the more bioprotective strain BCC0191 persisted in the soil over 7 days at a level of 27.6% of the initial inoculum. Overall, this demonstrated that BCC0191 can survive and compete in a mixed, non-sterile soil (potting mix) microbial community better than other strains of B. ambifaria (Supplementary Figure S5C). The induction and expression of the cepacin BGC by B. ambifaria directly in the rhizosphere was determined by RT-PCR targeting the desaturase gene ccnN (Mullins et al., 2019). This PCR method was applied to total RNA extracted from the rhizosphere following 3 days of plant growth. A RT-PCR amplicon correlating to the expression of the ccnN gene was observed in all three strains (Supplementary Figure S6) and demonstrated that the cepacin gene BGC was readily expressed by all B. ambifaria strains at the pea rhizosphere.

Overall, this comparative evaluation of phenotypic traits demonstrated that multiple factors, in addition to polyyne production, play a role in the biological control efficacy of B. ambifaria strains. Successful rhizosphere colonisation, and the ability of the bacterium to persist and compete with the soil microbial community are clearly important attributes for an effective biocontrol agent. It is probable that the combination of these factors along with other qualities, including production of other antimicrobials (Mullins et al., 2019; Webster et al., 2020a), allows B. ambifaria BCC0191 to be a more efficient biocontrol agent in the G. ultimum damping-off assay than other strains of B. ambifaria. Previously, it has been shown that many Pseudomonas species are effective biocontrol agents due to their ability to colonise the plant surface (spermosphere, rhizosphere and phyllosphere) and the endosphere (Lugtenberg et al., 2001). These Pseudomonads not only protect plants by producing bioactive metabolites but can also use plant exudates with high growth rates, allowing them to compete with other microorganisms for space and nutrients in the plant environment (Lugtenberg et al., 2001). For example, the biocontrol ability of P. fluorescens strain 54/96 in the control of Pythium damping-off disease is based on their capacity to colonise plant tissues, exhibit high growth rates and outcompete the pathogen for limited plant nutrients and infection sites (Ellis et al., 1999). Colonisation of plant roots by bacterial endophytes through crack entry between adjacent cells, during emergence of lateral roots is well established (Webster et al., 1998).

Given the lack of studies on Burkholderia biocontrol agents since the EPA risk assessment moratorium in 1999, there has been a comparative dearth of investigation into the mode of action and efficacy of strains within this genus compared to Pseudomonas and Bacillus where multiple strains and metabolites have been studied (Dimkić et al., 2022). The exploration of Burkholderia species has rightly focused on their pathogenic traits over the past two decades, however, multiple studies are now beginning to uncover the mechanisms and specialised metabolites that underpin their biological control proficiency (Mullins et al., 2019; Murata et al., 2021). However, it was unknown whether other Burkholderia strains or bacterial species encoding cepacin, caryoynencin and other polyyne BGCs were capable of similar protective biopesticidal roles. Our strain panel represented nine species and four polyyne BGCs, and in vitro microbial antagonisms was systematically compared to their efficacy in a pea biological control model with G. ultimum damping-off disease.

Considerable differences were observed in biocontrol proficiency between evolutionarily close species of B. ambifaria and B. vietnamiensis despite equivalent in vitro microbial antagonism and cepacin production (Figures 2, 3). In contrast to its widely characterised plant and human pathogenic nature (Jones et al., 2021), B. gladioli was uniquely shown to be an efficacious anti-damping off agent when interacted with germinating plants (Figure 3B), with the polyyne caryoynencin demonstrated to be a key metabolite in mediating G. ultimum biocontrol. Moreover, despite the protective B. gladioli strain being shown to be capable of producing broad-host range toxins such as toxoflavin and bongkrekic acid in vitro, these metabolites were suppressed on the biomimetic PEM medium, and the excellent pea survival rate suggested that toxic metabolites or pathogenic factors were not being deployed during the pea rhizosphere colonisation. Overall, the protective phenotype of caryoynencin, associated with rhizosphere colonisation of germinating peas exhibited by B. gladioli and B. plantarii in the biocontrol model, contrasts to the historical perspective of these bacteria as plant pathogens (Maeda et al., 2006; Jones et al., 2021). The beneficial interactions observed in this study suggests that the plant pathogenic lifestyle of B. gladioli and B. plantarii is likely host specific and opportunistic dependent on multiple factors such as host age and damage.

Our study demonstrated that certain Burkholderia strains with the capacity to produce polyynes act as optimal biological control agents for damping off-disease when applied as a seed coat to peas. This aligns to their successful commercial use and sheds light on their mechanism of action as biopesticides capable of protecting peas, maize, and other crops species (Parke and Gurian-Sherman, 2001). Interestingly, Pseudomonas protegens lacked noticeable biocontrol activity against G. ultimum on peas despite in vitro antagonism (Figure 3) and the observed protection mediated by protegencin in the biocontrol of G. ultimum in a cucumber model (Murata et al., 2021). These differences suggest host-specific interactions, as well as abilities such as rhizosphere colonisation rate, persistence, antimicrobial arsenal, and the ability to compete within the soil-associated microbial community play additional roles in successful biological control. Uncovering these factors and utilising them towards sustainable control of crop damping-off diseases is vital for food security going forward.