All bacterial and fungal strains used in this work are described in Supplementary Table S1. The bacterial strains were stored at −80°C using 50% (v/v) glycerol. Fungi were maintained on acidified potato dextrose agar plates (APDA) at 10°C and were sub-cultured regularly at room temperature (24°C ± 1°C). The B. velezensis isolates ES1-02 and EFSO2-04 are deposited in the National Collection of Industrial Food and Marine Bacteria (NCIMB), UK, with identification numbers NCIMB 15453 and NCIMB 15455, respectively.

Genes encoding LP synthesis, including srfAA, ituB, fenA, and fenD were screened by PCR method. The primers used for the PCR reactions, except ituB (Dunlap et al., 2019), were designed from the conserved gene region of the genes with information received from multiple sequence alignments of gene sequences obtained from GenBank using Clustal Omega (EMBL-EBI, Hinxton, Cambridge, UK). PCR reaction was performed using PEQSTAR Thermal Cycler (VWR International GmbH, Darmstadt, Germany). The PCR product was purified using a QIAGEN QIAquick PCR Purification kit (QIAGEN AB, Kista, Sweden), and then sequenced (Eurofins Genomics, Ebersberg, Germany). All primers used in this work are listed in Supplementary Table S2.

LC–MS analysis of LPs was performed on a 1290 UHPLC system (Agilent, Waldbronn, Germany) coupled to a Q-Exactive Plus Orbitrap mass spectrometer equipped with a heated electrospray ionization (HESI) source (Thermo Fisher Scientific GmbH, Braunschweig, Germany). LPs were separated on an Eclipse C18 column (Agilent, Waldbronn, Germany; 2.1 mm × 50 mm, 1.8 μm particle size). The column temperature was maintained at 40°C. Samples were dissolved in methanol and 5 µL of each sample was injected. Mobile phase A was 0.1% formic acid in water, and mobile phase B was 0.1% formic acid in acetonitrile. A constant flow rate of 0.3 mL/min was used, and gradient elution was performed as follows: 10%–50% B from 1 to 5 min, 50%–95% B from 5 to 15 min, isocratic at 95% B from 15 to 20 min, the system was returned to initial conditions from 95% B to 10% B from 20 to 22 min.

The HESI source was operated in positive ion mode, with a capillary voltage of 4.2 kV and an ion transfer capillary temperature of 360°C. The sheath gas flow rate and auxiliary gas flow rate were set to 60 and 20, respectively. The S-Lens RF level was 50%. The Q-Exactive Plus mass spectrometer was calibrated externally in positive ion mode using the manufacturers calibration solutions (Pierce, Thermo Fisher Scientific GmbH, Braunschweig, Germany). Mass spectra were acquired in MS mode within the mass range of 500–1700 m/z at a resolution of 70,000 FWHM (Full Width at Half Maximum) using an Automatic Gain Control (AGC) target of 1.0 × 10⁶ and 100 ms maximum ion injection time. Data dependent MS/MS spectra in a mass range of 200–2000 m/z were generated for the five most abundant precursor ions with a resolution of 17,500 FWHM using an AGC target of 3.0 × 10⁶ and 120 ms maximum ion injection time and a stepped collision energy of 15, 30 and 45. Xcalibur™ software version 4.3.73 (Thermo Fisher Scientific, San Jose, United States) was used for data acquisition and data analysis. Commercial standards of surfactin, fengycin and iturin A (all from Sigma-Aldrich, Seelze, Germany) were used as reference for LP identification in B. velezensis strains.

For preparation of pre-cultures, 100 µL of a glycerol stock was inoculated into 10 mL of LB medium (5 g/L tryptone, 10 g/L NaCl, 10 g/L yeast extract) in 100 mL baffled shake flasks and cultivated overnight at 37 °C and 120 rpm using a shaking incubator (Newbrunswick™/Innova^® 44, Eppendorf AG, Hamburg, Germany). Subsequently, second precultures prepared with 25 mL mineral salt medium (MSM) in 250 mL shake flasks were inoculated with the first precultures to a starting optical density (OD₆₀₀) of 0.1 and incubated for approx. 16 h. The main cultivation was carried out in 50 mL MSM in 500 mL shake flasks representing a 10% filling volume which has been demonstrated to provide sufficient oxygen availability for an aerated bioprocess (Hoffmann et al., 2021). The main cultivations were inoculated with the second precultures to a starting OD₆₀₀ of 0.1. MSM was prepared according to Willenbacher et al. (2015) with slight modification. The medium contained 8 g/L glucose, 4.0 µM Na₂EDTA × 2 H₂O, 7 µM CaCl₂, 4 µM FeSO₄, 1 µM MnSO₄, 50 mM Urea, 30 mM KH₂PO₄, 40 mM Na₂HPO₄ and 800 µM MgSO₄. All cultivations were carried out in three biological replicates at 37°C and 0.4 g (orbit of 5.1 cm, shaking frequency of 120 rpm) in an incubation shaker (Innova 44®R, Eppendorf AG). Samples were taken at regular interval and the OD₆₀₀ measured using a spectrophotometer (WPA CO8000, Biochrom Ltf., Cambridge, UK). The samples were subsequently centrifuged at 7,164 g and 4°C for 10 min and cells were removed. The supernatants were stored at −20°C until further processing.

LPs were extracted according to Geissler et al. (2019). In brief, a threefold extraction of 2 mL of cell-free broth with 2 mL of a mixture of chloroform/methanol (2:1; v/v) was conducted. The pooled solvent layers were evaporated to dryness via a rotary evaporator (RVC2-25 Cdplus, Martin Christ Gefriertrocknungsanlagen GmbH, Osterode am Harz, Germany) at 10 mbar and 40°C. LPs were analyzed with an HPTLC system (CAMAG, Muttenz, Switzerland) according to a method described by Geissler et al. (2019) with little modification. In detail, dried samples were re-suspended in 2 mL of methanol and applied as 6 mm bands on HPTLC silica gel 60 plates (Merck, Darmstadt, Germany). A two-step development was performed for the simultaneous quantification of surfactin, iturin/bacillomycin and fengycin. The first development was performed with a mixture of chloroform/methanol/water (65:25:4; v/v/v) as the mobile phase, and a developing distance of 60 mm, while the second development was carried out using the mobile phase butanol/ethanol/0.1% acetic acid (1:4:1; v/v/v) over a migration distance of 60 mm as well. To detect and quantify surfactin and iturin/bacillomycin, plates were scanned after the first development, while fengycin was quantified after second development. All LPs were scanned at UV 195 nm. Surfactin, iturin and fengycin standards (Sigma-Aldrich, St. Louis, United States) were used for calibration.

Prior to data analysis, OD₆₀₀/cell dry weight (CDW) conversion factor was determined as described by (Geissler et al., 2019). CDW, glucose concentration, and LP (surfactin, iturin or bacillomycin, and fengycin) titers at sampling time point were plotted. The yield of biomass per substrate Y_X/S (g/g), product per biomass Y_P/S (g/g), growth rate (μ), and specific productivity q_P/X (g_product/g_CDW h) were determined with the equations below. Y_P/S, Y_P/X and q_P/X were determined at the maximum LPs (surfactin, iturin, bacillomycin or fengycin [P_max]) concentrations, while Y_X/S was determined at the maximum CDW. ΔS is the change in substrate between the time point at which a value was calculated and the beginning of cultivation (t₀). Δt is change in time (h) between the point of calculated value and t₀. The Y_X/S was determined at the maximum CDW, while Y_P/S and Y_P/X were analyzed at the maximum product concentration. For the calculation of growth rate, the time intervals between the start of cultivation and the time point of highest CDW during the exponential phase were used. The following analyses were done with equations reported by (Klausmann et al., 2021). Y X / S = X Δ S (1) Y P / S = P m a x Δ S (2) Y P / X = P m a x X (3) q P / X = P m a x X p m a x   ∗   Δ t (4) µ = ln x 2 − ln x 1 Δ t (5)

Antagonistic ability of each B. velezensis strain against Diaporthe spp. was evaluated using the dual culture technique. Therefore, a B. velezensis strain was inoculated at one side of an APDA plate and a 6 mm disc of an actively growing Diaporthe sp. was inoculated on the opposite side of the plate and incubated at 25°C. Diaporthe spp. inoculated on one side of the plate without co-inoculation of B. velezensis strains were used as control. All experiments were done in three biological replicates. Fungal growth was measured from day 5 to day 7. To determine the effects on Diaporthe spp. morphology caused by co-incubation with B. velezensis strains, mycelial samples taken from regions next to the inhibition zone were examined under light microscope. The mycelia taken from the edge of the growing fungal in the control plates were also examined.

Antifungal activity was expressed as percentage growth inhibition and was calculated with the following equation. I n h i b i t i o n = R c   –   R i R c ∗ 100 % (6)

In this context, R _i is the distance (mm) from the point of inoculation towards the edge of the colony in dual culture, and R _c is the corresponding distance (mm) of mycelium determined in the control.

Agar plugs of approximately 500 mg were taken from the inhibition zones using a cork borer. To de-structure the agar and prepare for homogenization, agar plugs were frozen overnight at −20°C. Subsequently, 1 mL sterile distilled water was added to the agar and homogenized with a mechanical homogenizer (MICCRA MiniBatch D-9, MICCRA GmbH, Heitersheim, Germany) at 11,000 rpm for 2 min and afterwards mixed thoroughly by vortexing. Subsequently, LPs were extracted and quantitatively measured by HPTLC.

B. velezensis strain ES1-02 was co-cultivated with D. longicolla DPC_HOH20 on APDA plates as previously described. A control experiment was set up with ES1-02 cultivated alone on APDA plates. The experiments were set up in seven biological replicates. After 5 days of cultivation with a visible zone of inhibition between bacteria and fungi, the edge of the bacterial colony next to the inhibition zone was gently scraped with a sterile loop and immediately put in 40 µL lysis buffer (2% SDS, 50 mM Tris-HCl pH 6.8, 20 mM DTT). In the control experiment, samples were collected at the edge of the growing ES1-02 colonies after 5 days of cultivation and put into lysis buffer.

Bacterial samples in lysis buffer were subsequently heated for 5 min at 95°C and 700 rpm in a ThermoMixer^® (Eppendorf ThermoMixer^® C, Hamburg, Germany). After centrifugation at 20,000 g for 5 min, protein concentrations of the cleared supernatants were determined by the Bradford assay (Bradford, 1976) and adjusted to a protein concentration of 1 μg/μL using Tris-HCl (pH 8.5) and lysis buffer. In total, 28 µL of cleared lysate was reduced and alkylated (10 mM Tris-(2-Carboxyethyl)phosphine (TCEP), 40 mM chloroacetamide (CAA) for 30 min at room temperature in the dark and proteins were subsequently extracted by single-pot solid-phase-enhanced sample preparation (SP3; 1:1 mixture of SpeedBeads™ magnetic carboxylate modified particles 50 mg/mL; Cytiva; CAT No: 45152105050250 and 65152105050250) (Hughes et al., 2014). A volume of 10 µL of activated SP3 Bead-Mix and ethanol to a final concentration of 80% was added to the sample and incubated for 5 min at room temperature. Samples were mounted on a magnetic rack and briefly washed three times with 80% ethanol in 50 mM Tris/HCl (pH 8.5).

Proteins were digested on the SP3 beads in 0.2% sodium-deoxycholate (SDC), 50 mM ammonium bicarbonate using trypsin (Roche, Penzberg, Germany) and LysC (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) in a 1:100 and 1:200 protein to protease ratio, respectively. The protein digest was performed for 20 h at 36°C and 800 rpm. The digest was stopped by adding formic acid (FA) and acetonitrile (ACN) to the samples at final concentrations of 0.5% FA and 96% ACN. Samples were washed on a magnetic rack three times with 98% ACN and 0.05% FA. Peptides were eluted from the beads with water and lyophilized in a vacuum concentrator (Concentrator plus, Eppendorf, Hamburg, Germany). Peptides were solubilized in 0.1% trifluoroacetic acid (TFA) prior to LC-MS/MS analysis.

Nano-LC-ESI-MS/MS experiments were performed on an Ultimate 3,000 nano-RSLC (Thermo Fisher Scientific, Germering, Germany) coupled to an Exploris 480 mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) using a Nanospray-Flex ion source (Thermo Fisher Scientific, Bremen, Germany). Peptides were concentrated and desalted on a trap column (5 mm × 30 μm, Thermo Fisher Scientific, Dreieich, Germany) and separated on a 25 cm × 75 µm nanoEase MZ HSS T3 reversed phase column (100 Å pore size, 1.8 µm particle size, Waters, Milforf, MA, United States of America) operated at constant temperature of 35°C. Peptides were separated at a flow rate of 300 nL/min using a gradient with the following profile: 2%–8% solvent B for 4 min, 8%–16% solvent B for 22 min, 16%–30% solvent B for 11 min, 30%–95% solvent B for 4 min, isocratic 95% solvent B for 4 min and 95%–2% solvent B for 5 min. Solvents used were 0.1% FA (solvent A) and 0.1% FA in ACN/H₂O (80/20, v/v, solvent B).

MS spectra (m/z = 300–1,500) were detected in the Orbitrap at a resolution of 60,000 (m/z = 200). Maximum injection time (MIT) for MS spectra was set to 50 ms, the AGC value was set to 3 × 10⁶. Internal calibration of the Orbitrap analyzer was performed using lock-mass ions from ambient air as described by Olsen et al. (2005).

The MS was operating in data dependent mode selecting the top 25 highest abundant peptide precursor signals for fragmentation (HCD, normalized collision energy of 30). For MS/MS analysis only undetermined charge states and charge states from 2-5 of considered; the monoisotopic precursor selection was set to peptides and the minimum intensity threshold was set to 7.5 × 10⁴. MS/MS scans were performed in the Orbitrap at a resolution of 15,000, isolation width was set to 1.6 Da. The AGC target was set to 7 × 10⁴, a maximum injection time (MIT) of 70 ms and a fixed first mass of 120 m/z. Dynamic exclusion was set to 60 s with a tolerance of 10 ppm (parts per million).

Raw files were first processed with MaxQuant (Cox and Mann, 2008). Protein identification in MaxQuant was performed using the integrated database search engine Andromeda (Cox et al., 2011). MS/MS spectra were searched against the B. velenzensis protein sequence database downloaded from UniProt (Bateman, 2019). Reversed sequences as decoy database and common contaminant sequences were added automatically by MaxQuant. Mass tolerances of 4.5 ppm for MS spectra and 20 ppm for MS/MS spectra were used. Trypsin was specified as enzyme and two missed cleavages were allowed. Carbamidomethylation of cysteine was set as a fixed modification and protein N-terminal acetylation and methionine oxidation were allowed as variable modifications. The “match between runs” feature of MaxQuant was enabled with a match time window of 1 minute and an alignment time window of 20 min. Peptide false discovery rate (FDR) and protein FDR thresholds were set to 0.01. Protein quantification in MaxQuant was performed by label free quantification (LFQ) with a LFQ minimum ratio count setting of one.

In a second data processing step, MaxQuant result files were fed into the IceR R-package, for further re-quantification using default settings (Kalxdorf et al., 2021). The pre-imputed IceR output file containing LFQ intensities was further normalized according to the median LFQ intensity of the sample and used subsequently for differential expression analysis.

Welch´s t-test, principal component analysis (PCA), hierarchical clustering and Volcano plots were performed using Perseus version 1.6.14.0 (Tyanova et al., 2016). Matches to contaminants (e.g., keratins, trypsin) and reverse databases identified by MaxQuant were excluded from further analysis. First, normalized LFQ values were log2 transformed. Only protein groups with at least 40% valid values in data set were considered for the statistical analysis. Remaining missing values were imputed by the “Replace missing values from normal distribution” function (Width = 0.3, Down shift = 1.8) implemented in Perseus drawing random numbers from a normal distribution representing low abundance intensity values. Significant changes in protein abundance were analyzed by a Welch’s t-test (two sided, S0 = 1) and corrected for multiple hypothesis testing using permutation-based FDR statistics (FDR = 0.05, 250 permutations). For subsequent hierarchical cluster analysis of the significant differential expressed proteins, imputed LFQ intensity values were removed. The mass spectrometry proteomics data was deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2022) partner repository with the dataset identifier PXD035016.

Since B. velezensis strains exhibit capability for producing a great range of LPs, the genetic potential of B. velezensis strains ES1-02 and EFSO2-04 was compared with reference strains QST713 and FZB42. Specifically, srfAA genes from the srfA operon encoding the surfactin-forming NRPS were amplified in all strains (Supplementary Table S3). Futhermore, since B. velezensis strains are capable of producing different variants of iturin-type LPs, the strains were screened for specific genes encoding different LPs of the iturin class, such as iturin A, bacillomycin L, and bacllomycin D. Therefore, a PCR technique targeting iturin gene (ituB) was used to predict the strain-specific nature of the iturin family of LP produced by each strain (Dunlap et al., 2019). In brief, amino acids in positions 1 to 3 of the peptide sequence within the LP are conserved in all iturin families, while amino acids at positions 4 to 7 may be variable (Dunlap et al., 2019). In the PCR approach, the binding site for a forward primer was located in the conserved third NRPS module (third amino acid: AA#3), which allowed integration of the D-asparagine for all three types of iturin (iturin A, bacillomycin D and L). In contrast, the binding sites for the reverse primers were different for each of iturin A, bacillomycin D and bacillomycin L and were located on variable AA#4 modules (iturin A–Gln, bacillomycin D–Pro, bacillomycin L–Ser). In this way, predictions were made about which type of the iturin family is encoded by each strain. In summary, ituB genes corresponding to bacillomycin L were detected in the isolates ES1-02 and EFSO2-04, while genes specifically encoding iturin A and bacillomycin D were amplified in reference strains QST713 and FZB42, respectively (Supplementary Table S3). With respect to fengycin, two genes (fenA and fenD) were targeted for amplification in the fenABCDE operon. While fenA was amplified in all strains, the amplification of the fenD gene was observed only in QST713 and FZB42.

While the detection of genes encoding for the production of LPs provides information about the genetic potential of strains, the production of these molecules needs to be determined by reliable methods such as mass spectrometric (MS) analyses. This provides information on the composition of the isoforms (congeners) and the length of the fatty acid chains, which are important properties for biological activities (Henry et al., 2011). The MS results showed that at least two LP types were co-synthesized in each strain (Table 1).

Further details on the LP detection showed that surfactin, composed of a mixture of C₁₂ to C₁₇ isoforms, was produced by all four B. velezensis strains (Table 1) with the most abundant isoforms being C₁₄ for both reference strains and C₁₅ for the isolates, respectively, and C₁₂ as the least abundant in all strains (Supplementary Figure S1A). However, the iturin family LPs produced by the strains differed in type and isoform composition. In detail, strains ES1-02 and EFSO2-04 produced bacillomycin L with fatty acid chain lengths from C₁₃ to C₁₇, while QST713 synthesised iturin A with C₁₂ to C₁₇, and FZB42 produced bacillomycin D isoforms from C₁₁ to C₁₇ (Table 1). Although the B. velezensis strains showed different iturin production, C₁₅ and C₁₃ were the most and the least abundant isoforms in all strains, respectively (Supplementary Figure S1B). Fengycin synthesis was not detected in the isolates ES1-02 and EFSO3-04, possibly due to shortened versions of the fen operon as indicated by the non-detection of the fenD genes (Supplementary Table S3). However, reference strains QST713 and FZB42 synthesized fengycin groups A, B, X and Y (Table 1) with varying isoform abundances (Supplementary Figure S1C, D).

In order to get insights about both the cell growth and LP production of the different B. velezensis strains, shake flask cultivations were performed for 60 h at an initial glucose concentration of 8 g/L. ES1-02 and EFSO2-04 reached maximum CDW values of 4.6 and 4.1 g/L, respectively, after 18 h, while QST713 reached comparable maximum CDW values of 4.6 g/L after a shorter cultivation time of 10 h. Strain FZB42, however, achieved a lower maximum CDW of 1.5 g/L after 14 h (Figure 1). Strains ES1-02, EFS02-04, and FZB42 showed growth limitation before glucose was completely consumed, indicating possible additional substrate limitation aside from glucose. Growth limitations associated with nitrogen, phosphate or trace element limitation have been reported in Bacillus (Mahmood, 1972; Hu et al., 1999; Hoffmann et al., 2021), and as such the impact on lipopeptide production should be a subject of future investigation.

Regarding LP production, a surfactin titer of 97.4 mg/L after 18 h was measured for ES1-02, while EFSO2-04 produced an equal amount of surfactin as FZB42 after at 79.9 mg/L after 60 h. QST713 however produced 77.2 mg/L after 44 h. The highest amount of bacillomycin L was detected in ES1-02 (4.2 mg/L) after 44 h, while the highest amounts of iturin A and bacillomycin D were achieved in QST713 (18.5 mg/L) and FZB42 (6.4 mg/L) after 60 and 34 h, respectively. Fengycin biosynthesis by QST713 and FZB42 reached maximum of 62.8 after 44 h and 33.5 mg/L and 60 h, respectively.

To properly understand the performance of the strains, several data analyses were performed based on cell growth and LP production and the results summarized in Table 2.

Specifically, ES1-02 and EFSO2-04 showed comparable specific growth rates of 0.08 h^-1, while a higher specific growth rate of 0.15 h^-1 was determined for QST713 and a lower rate of 0.03 h^-1 for FZB42 (Table 2). Biomass yields on substrate Y_x/s were similar for ES1-02, EFSO2-04 and QST713. Similarly, ES1-02 showed a higher yield of surfactin on substrate (0.012 g/g) compared to QST713 (0.010 g/g) and FZB42 (0.010 g/g) but was inferior in terms of iturin/bacillomycin yield on substrate. With regards to specific productivities, higher surfactin values were recorded for QST713 and FZB42, while EFSO2-O4 was superior in bacillomycin/iturin.

Since the B. velezensis strains produced different LP combinations with varying amounts and relative isoform abundances, the applicability in terms of antifungal activity against phytopathogens was investigated. For this purpose, the soybean pathogens Diaporthe spp. were used as indicator strains. Using the dual in vitro antagonistic assay, distinct zones of inhibition were observed between the B. velezensis strains and the fungal pathogens Diaporthe spp. after 7 days of co-incubation at room temperature (Figure 2A).

In detail, B. velezensis strains ES1-02 and EFSO2-04 showed inhibitory efficiencies comparable to the commercial biocontrol strain QST713 with about 40% of inhibition against three of the test Diaporthe spp. (DPC_HOH2, DPC_HOH7 and DPC_HOH15). (Figure 2B). In contrast, higher inhibition activities were determined for strain FZB42 against all Diaporthe spp. compared to the other B. velezensis strains against the indicator pathogens (68.9% DPC_HOH2, 66% DPC_HOH7, 69.6% DPC_HOH15% and 50.4% DPC_HOH20). In more detail, strains ES1-01 and EFSO3-04 were the most effective against DPC_HOH2, while QST713 showed the best inhibitory effect against DPC_HOH2 and DPC_HOH15. Strain FZB42, however, demonstrated the highest efficacy against DPC_HOH15 (Figure 2B).

In order to evaluate the effects of the B. velezensis strains on the fungal pathogens, microscopic examination of the fungal mycelia at the edges of the inhibition zones was carried out. Morphological anomalies such as enlargement of hyphae, swellings, formation of bulbs, or complete disruption of mycelium were observed (Figure 3). In contrast, the mycelium from the control experiments, in which the fungi were not exposed to a B. velezensis strain, showed well-developed hyphae (Figure 3). In particular, swollen structures resembling vesicles were observed in indicator pathogens DPC_HOH7 and DPC_HOH2 exposed to B. velezensis isolates ES1-02 and EFSO2-04 as well as exposed to reference strain QST713, respectively (Figure 3). The co-cultivation approaches in this study thus show that an antifungal effect is already present under native cultivation conditions and that no purification of the LPs is necessary. This confirms the assumption that both reference strains FZB42 and QST713 as well as the B. velezensis isolates ES1-02 and EFSO2-04 presented here can have a beneficial effect in agriculture.

Since lipopeptides produced by Bacillus strains have been described as versatile weapons for biocontrol of plant diseases (Ongena and Jacques, 2008), further insights into LP production in the presence of pathogens are needed. Accordingly, the amounts of LPs in the inhibition zones were quantified. Therefore, the influence of the interaction of D. longicolla (DPC_HOH20) with the B. velezensis strains on the respective LP synthesis was determined. In this context, the surfactin accumulation in the inhibition zone of the isolates ES1-02 and EFSO2-04 showed 10-fold and 5-fold increase, respectively, compared to the control experiments, in which the B. velezensis strains were cultivated alone (Figure 4). Notably, no increase in surfactin accumulation by reference strains FZB42 and QST713 was observed when co-cultivated with the pathogenic indicator strain DPC_HOH20. Since the highest increase in surfactin accumulation in the inhibition zone was observed in ES1-02, the distribution of surfactin isoforms in the inhibition zone of ES1-02 was also examined and it showed increases in the relative abundances of C₁₆ and C₁₇ isoforms from 63% to 5% when ES1-02 was cultivated alone to 75% and 9% respectively in the inhibition zone when co-incubated with the D. longicolla indicator strain. A decrease in C₁₃ isoforms from 10% to 5% was also observed, while the abundances of the other isoforms remain relatively unchanged (Supplementary Figure S2).

In addition, a meaningful increase in bacillomycin L was determined in EFSO2-04, although to a lesser extent than in surfactin, by 1.5-fold (Figure 4). In contrast, the accumulation of bacillomycin D in the inhibition zone of FZB42 was slightly decreased compared to the strain cultivated alone. For fengycin, increased synthesis was detected in both QST713 and FZB42, with the higher induction of 1.5-fold recorded in QST713.

Surfactin has been previously described as a potent biosurfactant with various biological activities (Meena and Kanwar, 2015). Although its antimicrobial properties seem to occur mainly in the presence of synergistic metabolites (Lilge et al., 2022), the amount of surfactin indicates the biocontrol potential of a Bacillus strain. Accordingly, the B. velezensis strain ES1-02 with a 10-fold induction of extracellular surfactin accumulation during co-cultivation with the pathogenic indicator strain DPC_HOH20 (Figure 4) makes the strain interesting for a more detailed characterization. Accordingly, the bacterial adaptation was analyzed in more detail at the proteome level. To gain more insight into the molecular mechanisms involved in the response of B. velezensis to D. logicolla, proteome analyses were performed on strain ES1-02 cultured in the presence and absence of DPC_HOH20, respectively. Here, 100% quantification was achieved for 1,215 proteins with IceR (Supplementary Table S4). Principal Component Analysis (PCA) showed a clear separation between ES1-02 co-cultivated with D. longiclolla DPC_HOH20 and control experiments (Figure 5A).

A total of 148 (66 with higher abundance, 82 lower with abundance) proteins were found with significantly altered abundance between absence and presence of DPC_HOH20 (Figures 5B,C). A closer look at the protein abundances showed that about 23% of all proteins with significantly higher abundance associated with the production of bioactive secondary metabolites such as bacillaenes and polyketides (PRIDE dataset identifier PXD035016), for which extensive antimicrobial activities have been reported (Alenezi et al., 2021). In addition, about 11% of the proteins were found to be involved in the intracellular Bacillus oxidative, nitrosative and general stress response, as well as DNA repair, leading to non-specific, multiple and preventive resistance (Hecker and Völker, 2001). Interestingly, several proteins involved in fatty acid biosynthesis and modification (12%) were also found with higher abundance. This is reasonable in that fatty acid composition varies in response to environmental conditions and is important for cellular adaptation and might be a defence strategy regarding antagonistic co-cultivation (Diomandé et al., 2015; Joo et al., 2016).

In contrast, several proteins associated with the provision of phosphate (7%) and modification of the cell surface (7%), such as peptidoglycan-binding proteins and biosynthesis of teichoic/teichuronic acids and phospholipids, were detected at lower abundance. Interestingly, proteins associated with Fe-S cluster or binding these cluster were additionally determined with a lower abundance. Since, as described above, evidence for oxidative stress by correspondingly induced stress proteins could be detected, it can be assumed that this stress reaction is linked to a reduced biosynthesis of proteins with Fe-S clusters. In this way, defective proteins mediated by oxidized Fe-S clusters are avoided and further formation of radicals mediated by the Fenton reaction is reduced (Roche et al., 2013). In addition, proteins involved in amino acid import and biosynthesis (e.g., histidine, arginine and glutamate) (10%) and translation (10%) were detected with lower abundance, suggesting that B. velezensis cells localized close to phytopathogens reduce their metabolism. Table 3 summarizes the proteins that were most affected by co-cultivation with DPC_HOH20.