In the present study, 284 PGPR strains from five Bacillaceae genera (Bacillus, Priestia, Cytobacillus, Neobacillus, and Gottfriedia) were obtained from the Plant-Associated Microbial collection that was assembled by Prof. Kloepper (1978) at Auburn University from various plant rhizospheres. Each of these PGPR strains were cryopreserved at −80 °C and were grown on Tryptic Soy Agar (TSA) for isolated colonies prior to evaluation as potential biological control agents. For genomic analysis, Illumina-generated draft genome sequences were analyzed, trimmed, and assembled using CLC Genomic Workbench¹. Genome quality was further evaluated by CheckM v1.1.3. (Parks et al., 2015), and 284 genome sequences with returned genome completeness of greater than 70% were included in this study. A total of 29 (out of 284) Bacillus strains with the greatest biocontrol potential were further sequenced at Nanopore for single-contig complete genome sequence. As originally collected from 1989 through 2014, all of these PGPR strains were classified as belonging to the Bacillus genus. Accordingly, the taxonomy of all 284 strains was initially evaluated by the top hit of average nucleotide identity (ANI) of whole-genome sequence at Microbial Genomes Atlas (MiGA) webserver (Rodriguez-R et al., 2018). Upon NCBI database deposition, strain identities were confirmed and annotated by the NCBI RefSeq database (Goldfarb et al., 2024), and all strain information was made publicly available at BioProject Accession: PRJNA1078443 and ID: 1078443. A spreadsheet summarizing the information for each strain evaluated in this study (including each NCBI BioProject Strain Accession Number) can be found in Supplementary Table 1. As cited in Supplementary Table 1, 27 of the strains have been mentioned or investigated in previous publications. For convenience, those references are listed here as well (Ran et al., 2012; Rao et al., 2014; Hossain et al., 2015, 2019; Nasrin et al., 2015; Ravu et al., 2015; Liu et al., 2016, 2017; Thurlow et al., 2019; Wang et al., 2019; Devkota et al., 2020; Afroj et al., 2021; Kumar et al., 2011; Olajide et al., 2022; da Silva et al., 2023; Yustiati et al., 2023; Romanenko et al., 2024; Sun and Shahrajabian, 2025). Notably, reclassification as proposed by Gupta et al. (2020), Patel and Gupta (2020) assigned clades within the Bacillus genus into multiple Bacillaceae genera. Accordingly, all 284 strains were taxonomically distributed into 15 species. Six species fell within non-Bacillus Bacillaceae including Priestia megaterium (65 strains), Cytobacillus firmus (5 strains), Neobacillus dretensis (17 strains), N. vireti (5 strains), N. niacini (5 strains), and Gottfriedia acidiceleris (6 strains). The remaining nine species were retained within the Bacillus genus: B. pumilus (17 strains), B. safensis (24 strains), B. altitudinis (14 strains), B. velezensis (50 strains), B. subtilis (6 strains), B. toyonensis (15 strains), B. thuringiensis (13 strains), and B. pseudomycoides (5 strains). Finally, “Bacillaceae (other)” representing a collection of 37 strains from less commonly observed species contributing < 5 strains each to the pool were evaluated. A phylogenetic tree was constructed using the 16S rRNA sequence of each type strain matching all 15 species obtained from EZbioCloud database (Yoon et al., 2017). For routine bacterial growth, the bacteria were cultured at 30 °C in TSA or tryptic soy broth (TSB) medium with shaking at 200 rpm.

All strains were screened for their abilities to inhibit the growth of the root-associated plant-pathogenic oomycete, P. nicotianae in a plate-based assay. P. nicotianae was grown in a V8 agar medium (180 mL/L V8 juice, 2 g/L CaCO₃, and 15 g/L Bacto agar) while the Bacillaceae strains were grown on TSA. Assay plates were prepared using the V8 agar medium in which bacterial colonies were transferred into a well (diameter = 10 mm) containing TSA at the edge of the plate and the P. nicotianae was transferred as a plug to the center of the plate. The growth inhibition of P. nicotianae’s hyphae due to the presence of Bacillaceae colonies was recorded after a 7–10 days incubation. The inhibitory responses of Bacillaceae strains were classified as strong, weak, and no inhibition based on the measurement of the zone of inhibition (ZOI) and morphological changes of the Bacillaceae strain and P. nicotianae being evaluated. A strong inhibition was assigned for a clear zone of inhibition (ZOI) of 5–15 mm with complete elimination of P. nicotianae’s hyphae in the interface of Bacillus colonies and P. nicotianae, while no inhibition was assigned when the P. nicotianae’s hyphae spread over the Bacillus colonies with no observable ZOI. Exhibition of a less clear ZOI of 2–7 mm with substantial reduction of P. nicotianae’s hyphae was assigned as weak inhibition. Fifty-eight P. nicotianae-inhibitory Bacillaceae strains (all from Bacillus species) were further evaluated for antibiosis against F. oxysporum, F. graminearum, and R. solani. Each organism was assayed and evaluated in the same condition as described above for P. nicotianae, and the resulting antibiosis response was similarly classified as strong, weak, and no inhibition based on ZOI and morphological changes of the organisms evaluated.

Conservation of antibiosis activity expressed by various Bacillaceae species was calculated on the basis of the distribution of strong, weak, and no inhibition among strains within each species. A term “bioactivity index” accounting for such conservation of antibiosis activity was calculated using a weighted-average score of 1 for strong inhibition, 0.5 for weak inhibition, and 0 for no inhibition using the following equation:

As defined, bioactivity index of 1 indicates the highest expression of antibiosis activity, where all tested strains of a given species show strong antibiosis activity. Conversely, a 0 indicates no expression of antibiosis activity (i.e., no tested strains of a given species showed antibiosis activity).

Genome sequences of all 284 Bacillaceae strains were analyzed by antiSMASH v.5 (antibiotics and secondary metabolite analysis shell) to predict biosynthetic gene clusters (BGCs) and secondary metabolites. Predicted BGCs were further dereplicated based on respective BGCs from single contig complete genome sequences to eliminate duplicated and/or fragmented BGCs. To infer conservation in sequence and putative function, predicted BGCs were grouped into networks of clusters based on sequence similarities using BiG-SCAPE v.0.0.0r (Biosynthetic Gene Similarity Clustering and Prospecting Engine) (Navarro-Muñoz et al., 2020). Finally, network distances generated by BiG-SCAPE analysis were visualized and annotated using Cytoscape 2.8 (Smoot et al., 2011).

Representative strains from the five bioactive Bacillus species were selected for producing secondary metabolites that may be responsible for antibiosis activity. In order to produce secondary metabolites, the Bacillus strains were grown in Landy medium (glucose, 20 g/L, yeast 1 g/L, L-glutamic acid 5 g/L, KCl 0.5 g/L, MgSO₄ 0.5 g/L, KH₂PO₄ 1 g/L, L-phenylalanine 3 mg/L, MnSO₄ 5 mg/L, FeSO₄ 0.15 mg/L, CuSO₄ 0.16 mg/L, pH 7.0) for 72 h at 30 °C with constant agitation (175 rpm). To pellet cells, liquid cultures were subjected to centrifugation at 6,000 × g for 40 min. The pH of harvested cell-free supernatant was adjusted to 2.0 by dropwise addition of concentrated HCl with constant stirring. Following overnight incubation at 4 °C, the precipitate was collected by centrifugation at 6,000 × g for 50 min. The precipitate was extracted twice using 100% MeOH. The pooled MeOH extract was dried under a constant flow of N₂ (g), and the dried residue was redissolved in MeOH, filtered with a 0.2 μm Acrodisc syringe filter (Pall Corporation, Ann Arbor, MI), and stored at −20 °C until evaluated.

The antibiosis assay of the total extract of a representative strain from each of the five bioactive Bacillus species (i.e., B. velezensis JJ334, B. subtilis JM553, B. pumilus, B. altitudinis, and B. safensis) against P. nicotianae was carried out using a disk diffusion method (Wang et al., 2020). The assay was conducted using a V8 agar plate. Freshly grown P. nicotianae was transferred as a plug (diameter = 5 mm) to the center of the assay plate and allowed to grow for 72 h. A total of 10 μL of the total extract was added onto a sterilized filter disk (diameter = 6 mm) and then placed at the edge of the assay plate. The growth inhibition of P. nicotianae hyphae surrounding the filter disk was measured and recorded after 5–7 days of incubation at 25 °C. An example of antibiosis activity of total extracts of B. velezensis JJ334 against P. nicotianae, F. oxysporum, F. graminearum, and R. solani is shown in Supplementary Table 4.

UV-vis absorption spectra of Bacillus strain total extracts were evaluated for characteristic absorption features. To detect compounds at 220 nm, 20 μL of total extract were separated through a ZORBAX SB-C18 column (4.6 × 150 mm, 5 μm) for 60 min at a flow rate of 0.20 ml/min using an Agilent Infinity 1100 HPLC system (Santa Clara, CA). Solvent A (100% water) and B (100% acetonitrile), each containing 0.1% trifluoroacetate (v/v) were used with the following elution gradient for solvent B: 40% at 0 min, 55% at 15 min, 75% at 40 min, 100% at 60 min, and 75% at 77 min. To identify compounds in the total extract, 0.2 μL of the extract was separated by LC through an Acquity UPLC BEH C18 (2.1 × 50 mm, 1.7 μm) column and eluted onto a Thermo Fisher Exploris 120 orbitrap LC-MS (Milford, MA). The LC separation was run for 20 min with a column temperature of 40 °C and a flow rate of 0.20 mL/min. The mobile phase gradient was created using water (A) and acetonitrile (B), each containing 0.1% formic acid (v/v) such that solvent B was at 40% at 0 min and ramped to 100% at 14 min, and then back to 40% B at 16 min.

Ions were generated using both positive and negative ionization modes employing an electrospray ionization (ESI) source. In addition, fragment ions from the precursor ions were simultaneously produced in a high-stage MSⁿ analyzer. The identity of compounds from the total extract were initially confirmed by the parent ions generated by both positive and negative ionization modes. The mass spectra of each fragmented ion generated by the MSⁿ analyzer from the corresponding precursor ion produced in the positive ionization mode were used for the unambiguous identification of each compound. Further, characteristic fragment ions generated from lipopeptide core peptides were used as diagnostic ions for the identification of lipopeptide derivatives. For fengycin derivatives, two reporter fragment ions (A and B), generated by the cleavage of Orn2-Tyr3 (A) and Glu1-Orn2 (B) bonds from fengycin core peptide (Glu1-Orn2-Tyr3-Thr4-Glu5-Ala/Val6-Pro7-Gln8-Tyr9-Ile/Val10) were used as diagnostic ions for unambiguous identification (de Faria et al., 2011; Pathak et al., 2012; Volpon et al., 1999; Ait Kaki et al., 2020). For surfactin, the fragment ions generated by the cleavage of Glu1-Leu/Ile2 bond from the core peptide (Glu1-Leu/Ile2-Leu3-Val4-Asp5-Leu6-Leu/Ile7) and the remaining Glu1-fatty acid tail were used to determine the derivatives and length of fatty acid tail (Savadogo et al., 2011; Ma et al., 2016; Chen et al., 2017). Similarly, the fragment diagnostic ions generated by the cleavage of Asn-Tyr (278.11) and Asn-Tyr-Asn (392.15) fragments from the core peptide and fragmented fatty-acid tail were used for unequivocal identification of bacillomycin L derivatives (Dunlap et al., 2019).

Agilent Infinity 1,100 LC system was used for isolating bioactive compounds from the total extract of Bacillus strains. A total of 100 μL of the total extract was injected and eluted through Zorbax (Santa Clara, CA) SB-C18 column (4.6 × 150 mm, 5 mm) for 60 min, at a flow rate of 0.20 mL/min. The mobile phases were water (A) and acetonitrile (B), each containing 0.1% trichloroacetic acid (v/v) with the following gradient for solvent B: 40% at 0 min, 55% at 15 min, 75% at 40 min, 100% at 60 min, and 75% at 77 min. The compounds were detected at 220, 275, 375, and 450 nm by a diode array detector coupled with a full-spectrum (220–500 nm) analysis. Fifteen to 20 fractions were collected, pooled and concentrated from five consecutive runs. The purity of compounds in each fraction was evaluated by UV-vis absorption as well as by MS analyses.

Purified bacillomycin L, fengycin, and surfactin were evaluated for bioactivity against P. nicotianae, and a representative fungus, F. oxysporum using a disk diffusion assay (Wang et al., 2020). The assay was conducted on a V8 agar plate wherein a freshly grown target pathogen was transferred as a plug (d = 5 mm) to the center of the assay plate and allowed to grow for 72 h. A total of 10 μL of the purified lipopeptide was added into a sterilized filter disk (d = 6 mm) and then transferred onto the edge of the assay plate. The growth inhibition of P. nicotianae’s hyphae surrounding the disk was measured and recorded after 5–7 days. In order to determine the inhibitory strength of purified bacillomycin L, fengycin, and surfactin against P. nicotianae and F. oxysporum, a quantitative bioassay was carried out using a 96-well microtiter-based plate assay as described by Noel et al. (2019) with minor modifications. Both P. nicotianae and F. oxysporum were grown in a diluted V8 medium (80 mL/L of V8 juice, 1 g/L of CaCO₃, and 6 g /L Bacto agar). Freshly grown plugs of target pathogens were macerated by passing through a 22-gauge needle attached to a 10 mL syringe and further homogenized by vortexing for 2 min. A total of 20 μL of homogenized culture macerate, 160 μL of diluted V8 broth (80 mL/L of V8 juice and 0.5 g/L CaCO₃), and 20 μL of lipopeptide with desired concentration were then loaded into a 96-well microtiter plate using wide-orifice tips. Plates were incubated for 2 days at 25 °C and the growth of the target organism was determined spectrophotometrically at 600 nm by a microtiter plate reader (Biotek Instruments, Highland Park, VT).

A library of 284 PGPR strains representing 15 species across five Bacillaceae genera was evaluated for antibiosis activity against a root-associated plant-pathogenic oomycete, P. nicotianae. Antagonism against P. nicotianae was classified as strong, weak, or non-inhibitory based on measurements of zones of inhibition (ZOI) as well as evaluation of morphological changes to P. nicotianae and the respective Bacillaceae strain. Defining characteristics of these levels of inhibition are given in section “Materials and methods,” and a representative antibiosis assay plate illustrating all three levels of inhibition was also documented (Supplementary Figure 1). Fifty-eight (20%) of the 284 strains exhibited strong inhibition, while 41 strains were weak inhibitors, and 185 showed no pathogen inhibition. All 58 of the strongly inhibitory strains belonged to Bacillus species. Across the 103 strains from the other four Bacillaceae genera (Priestia, Cytobacillus, Neobacillus, and Gottfriedia) none showed strong inhibition, and only four showed weak inhibition of P. nicotianae. Further, 55 out of the 58 strongly inhibitory strains belonged to five Bacillus species: B. pumilus, B. safensis, B. altitudinis, B. velezensis, and B. subtilis (Figure 1A). Interestingly, these five species are more closely phylogenetically related to one another as compared to the other species (Figure 1A), suggesting that the common factors contributing to antibiosis activity may be phylogenetically conserved. For each of these species, at least 40% of strains tested exhibited some level of inhibition (strong or weak): B. velezensis (74%), B. pumilus (100%), B. safensis (67%), B. subtilis (67%), and B. altitudinis (43%). Accounting for the overall percentage of inhibitory strains as well as the relative contribution of strong vs. weak vs. non-inhibitory strains, a bioactivity index (ranging from 0 to 1) was calculated (see section “Materials and methods”); these five species returned values of 0.39 (B. altitudinis), 0.58 (B. subtilis), 0.56 (B. safensis), 0.61 (B. velezensis), and 0.85 (B. pumilus) (Figure 1B). Inhibition was sparsely distributed among B. toyonensis, and B. thuringiensis strains, generating bioactivity indices of 0.27, and 0.23, respectively. Finally, all other species tested showed bioactivity indices ≤0.10: B. pseudomycoides (0), P. megaterium (0.02), C. firmus (0.1), N. drentensis (0.03), N. vireti (0), N. niacini (0), and G. acidiceler (0), and Bacillus (other) (0.05) (Figure 1B).

The 58 strong inhibitors of P. nicotianae were further evaluated for antibiosis activity against three fungal pathogens: F. graminearum, F. oxysporum, and R. solani. Interestingly, strains from B. velezensis and B. subtilis exhibited strong antibiosis activity against all three fungal pathogens whereas strains from B. pumilus, B. safensis, B. altitudinis, B. toyonensis, and Bacillus (other) exhibited either weak inhibition or were non-inhibitory (Figure 2). Specifically, all strains from B. pumilus, B. safensis, and B. altitudinis exhibited no inhibition against both F. oxysporum and F. graminearum. Further, only weak inhibition was observed against R. solani. As shown in Figure 1A, B. velezensis and B. subtilis are phylogenetically closely related; similarly, B. pumilus, B. safensis, and B. altitudinis are closely related to one another. Due to their ability to exert strong antagonism against P. nicotianae and all three pathogenic fungi, we classified B. velezensis and B. subtilis as generalists for their broad-spectrum bioactivity against fungal and oomycete pathogens; conversely, we classified B. pumilus, B. safensis, and B. altitudinis as oomycete-inhibiting specialists.

We surmised that strains strongly inhibitory against P. nicotianae were likely to possess a conserved capacity for biosynthesis of secondary metabolites which would account for their antibiosis activity, and these BGCs would be absent from the non-inhibitory species. In order to evaluate this hypothesis, draft genome sequences of 284 Bacillus strains were analyzed for the presence of predicted BGCs using antiSMASH (v.5) (Blin et al., 2019). A total of 2,446 BGCs (average eight per strain) were predicted across the 284 genomes, including 1,279 known and 1,167 BGCs that do not have a known natural product. Based on the structural and chemical properties of predicted secondary metabolites, BGCs were grouped into six functional classes: NRPS, PKS, a hybrid of PKS and NRPS (PK-NRPS), ribosomally-synthesized and post-translationally modified peptides (RiPP), terpenes, and BGCs outside these five classes (other). Information regarding Bacillus species, strain IDs, anti-P. nicotianae bioactivity, and predicted BGCs (with corresponding secondary metabolites) are available in Supplementary Table 2.

Bacillus species affiliated with the strong inhibitory group generally tended to have a higher number of BGCs per strain (8.9–13.2) with B. velezensis carrying the largest number of BGCs per strain. By comparison, the sparsely inhibitory species group contained 9.0–10.1 BGCs per strain, and the non-inhibitory species group contained 4.2–8.4. Consistent with these observations, the correlation between the overall number of BGCs and bioactivity was strong (Pearson r = 0.80; p = 0.0001).

There was a striking distinction in the representation of NRPS clusters between strains from Bacillus species which averaged 2.67 clusters per strain versus species of the other Bacillaceae which were essentially devoid of them (average = 0.03 NRPS/strain) (Figure 3A). With respect to inhibition, a distinction was also observed between the strong-inhibition generalists (B. velezensis and B. subtilis) with an average 3.0–4.1 per strain, while the strong-inhibition specialists (B. pumilus, B. safensis, and B. altitudinis) only carried 1.9–2.3 NRPS per strain. As a contrast, PKS, RiPP, and terpene BGCs were evenly distributed across all of the Bacillaceae species evaluated, regardless of the level or breadth of inhibition exhibited. Accordingly, correlations between NRPS content and bioactivity were relatively strong (Pearson r = 0.63, p = 0.0069) while those between RiPPs and Terpenes were quite poor (Pearson r = −0.24, p = 0.3640; Pearson r = 0.01, p = 0.9620, respectively). The PKS and other BGCs occupied the middle, showing moderate correlations with bioactivity index (Pearson r = 0.42, p = 0.0953, and Pearson r = 0.58, p = 0.0153, respectively). Interestingly, the tightest correlation parameters between bioactivity index and BGC type were observed for the PKS-NRPS clusters (Pearson r = 0.79, p = 0.0002); however, one species from the strong inhibitory group (B. altitudinis) and one from the sparsely inhibitory group (B. toyonensis) carried no PKS-NRPS BGCs.

In order to compare their distribution among the bioactive Bacillus species based on their structural and chemical properties, NRPS clusters were further divided into four subtypes: lipopeptides, siderophores, other, and unknown. All five species belonging to the strongly inhibitory group had at least one and up to 2.7 lipopeptide BGCs per strain on average (Figure 3B). In contrast, lipopeptides were nearly completely absent from all other Bacillaceae species tested (i.e., the sparingly inhibitory and non-inhibitory species groups). Indeed, none were detected in any of the non-Bacillus Bacillaceae, B. pseudomycoides, or B. toynensis strains. Across the B. thuringiensis and Bacillus (other) strains tested the average number of lipopeptide clusters was 0.1 per strain. Siderophores were relatively common among species from strongly and sparsely inhibitory groups, while NRPS classified as “other” were almost completely absent from inhibitory species, and the distribution of unknown NRPS was sporadic. These data suggest that lipopeptides, almost exclusively produced by strains strongly antagonistic to P. nicotianae, may be substantial contributors to strong antibiosis activity. Interestingly, species from the generalist group contained two to three lipopeptides per strain while species from the specialist group contained only one lipopeptide per strain. This indicates that a diverse set of lipopeptides may produce a synergistic effect that contributes to the broad-spectrum antibiosis activity exhibited by generalists.

Biosynthetic gene clusters from all 284 Bacillus strains were further analyzed (BiG-SCAPE v.0.0.0r) (Navarro-Muñoz et al., 2020) for gene cluster similarities to determine the extent to which secondary metabolite biosynthesis is conserved. Similarity analysis showed that BGCs identified as encoding a common putative metabolite across generalists, specialists, and non-inhibitory strains tended to segregate into separate clusters corresponding to these groups (Figure 4). This was particularly striking among lipopeptide BGCs where the differences in the structure of the gene clusters were highly distinct between antibiosis-based species groupings. Only generalists carried BGCs with modules for synthesis of the fengycin core decapeptide (fenA – fenE). Interestingly, the fengycin BGC from B. velezensis also contained ituA, ituB, and ituC, the modules necessary for the synthesis of an iturin core heptapeptide (Figure 5A). Although Bacillus strains from specialist and non-inhibitory species also carried a gene cluster identified as having similarity to a fengycin BGC, none of these contained modules for core peptide synthesis, but only genes supporting the synthesis of a putative betalactone. Interestingly, the specific structures of this BGC from specialists on one hand and non-inhibitory Bacillus species on the other were distinct (Figure 5A). Similarly, generalist and specialist Bacillus species groups all carried a BGC for production of a surfactin-like lipopeptide (Figure 4); no such BGC was identified in any strains from non-inhibitory species. Invariably, generalists (B. velezensis and B. subtilis) carried a BGC identified as surfactin (Figure 4) characterized by three core genes with modules for the synthesis of a heptapeptide (srfAA – srfAC) (Figure 5B). A separate bi-lobed cluster was observed for specialists with high similarity scores (85%) for lichenysin, a surfactin-like lipopeptide (Figure 4). The typical structure for this BGC contained two core genes in addition to the three required for heptapeptide synthesis (Figure 5B). These data suggest that the two to three highly conserved lipopeptides produced by generalists may contribute to a broad antifungal/anti-oomycete activity while the single surfactin or surfactin-like BGC carried by specialists may only enable strong anti-oomycete activity. We have observed that some strains belonging to generalist or specialist species exhibited no (or weak) inhibition (see Figure 1A) even though they appear to carry lipopeptide BGCs. Although the present study does not address this issue, it is possible that these strains may ultimately be unable to produce one or more of these lipopeptides due to missing core genes, nonsense mutations, frameshift mutations, and/or altered gene regulation. Such phenomena have previously been observed and reported for lipopeptide production by among B. subtilis strains, some even co-isolated from the same soil samples (Kiesewalter et al., 2021).

It should also be noted that other clusters of BGCs that putatively generate metabolites with antimicrobial properties segregate along the lines of generalists versus specialists versus non-inhibitors as well. Many of these BGCs belong to NRPS, PKS-NRPS, or PKS classes. For example, unique bacillibactin BGCs were identified for each of the three groups, and unique bacilysin clusters were each observed for generalists and specialists while non-inhibitors appeared to lack such a cluster altogether. In a similar manner, a bacillaene BGC is found only in generalists, and zwittermicin BGCs were only observed in specialists. Notably, the PKS BGCs for macrolactin and difficidin were only observed in B. velezensis.

Lipopeptides were produced by and extracted from five strains, each representing a Bacillus species with strong P. nicotianae inhibitory activity, including, B. velezensis JJ334, B. subtilis JM553, B. pumilus JJ1622, B. safensis JJ1244, and B. altitudinis JJ1138. Each metabolite extract was evaluated by UV-vis and LC elution profile (see Supplementary Figure 4). Consistent with the structure and predicted products of its BGCs, mass spectrometric screening of B. velezensis strain JJ334 extracts showed the production of bacillomycin L (an iturin), fengycin, and surfactin. Fengycin and surfactin were identified in extracts of the B. subtilis strain, JM553. This also was consistent with the fengycin BGC observed across B. subtilis strains which contained core genes for only fengycin production but not an iturin. Interestingly, only surfactin was identified in B. pumilus JJ1622, B. safensis JJ1244, and B. altitudinis JJ1138 extracts. This was consistent with the production of surfactin-like (lichenysin) compound predicted by antiSMASH. Fractionation of B. velezensis JJ334 extracts by HPLC followed by LC-MS analyses of lipopeptide fractions showed that bacillomycin L derivatives eluted between 1.5 and 2.8 min., while fengycin and surfactin derivatives eluted from 8.0 to 10.0 min and 13.4 to 17.4 min, respectively (Figure 6A). Mass spectrometric analyses identified the presence of six derivatives of bacillomycin L, 10 of fengycin, and eight of surfactin in the purified lipopeptide fractions (Figure 6B).

Extracts corresponding to the purified fraction of each lipopeptide were analyzed by LC-MSⁿ employing positive and negative ionization modes (see section “Materials and methods”). Each lipopeptide compound was confirmed by the fragmented ions generated by MSⁿ analyses (see Supplementary Tables 3–5 and Supplementary Figures 6–8). Bacillomycin L (six) and surfactin (eight) derivatives were identified each varying by the carbon chain length of the fatty-acid side chains. For both lipopeptides, derivatives were detected in both positive and negative ion modes as protonated, deprotonated, double-protonated, and/or sodium-adduct forms (Table 1). Thirteen fengycin derivatives were identified as [M+H]⁺, [M–H], and [M+2H]²⁺ ions. Four were fengycin A and two were fengycin A2 where the amino acid at position six in the core peptide was Ala, and either Ile or Val occupied position 10, respectively. Three were fengycin B and three were fengycin B2 where the amino acid at position six in the core peptide was Val, and either Ile or Val occupied position 10, respectively. Here as well, derivatives within fengycin A, A2, B, and B2 varied by fatty acid carbon chain length or amino acid in the core peptide (Table 1). There were insufficient data to identify a 13 fengycin, but its ion m/z values were consistent with C21 fengycin B or C22 fengycin B2. To our knowledge, this diversity of production of lipopeptide derivatives from a single Bacillus strain has not previously been observed.

Fractions containing bacillomycin L, fengycin, and surfactin were evaluated for their ability to inhibit the growth of P. nicotianae, F. oxysporum, F. graminearum, and/or R. solani by a disk diffusion method. Interestingly, each lipopeptide fraction exhibited inhibition against all four target pathogens (Supplementary Figure 5). Considering the production of all three lipopeptides by B. velezensis strains Each lipopeptide was further evaluated for inhibition against P. nicotianae and F. oxysporum by a microtiter plate assay. The maximum inhibition of P. nicotianae growth produced by bacillomycin L and fengycin was observed at concentrations of 50 μg/mL each (Figure 7A), where 56% and 59% growth inhibition were detected, respectively. For surfactin, a similar level of P. nicotianae growth inhibition (∼40%) was observed at concentrations ranging from 12.5 to 50 mg/mL. Growth inhibition of F. oxysporum was also observed for all three lipopeptides (Figure 7B). Bacillomycin L and fengycin showed similar inhibitory potency against this organism, with significant growth inhibition observed at lipopeptide concentrations 25 mg/mL and producing 85% and 87% growth inhibition at 100 mg/mL, respectively. The maximum inhibitory effect of surfactin (∼40%) against F. oxysporum was observed at a concentration of 50 μg/mL.