Competition for iron shapes metabolic antagonism between Bacillus subtilis and Pseudomonas marginalis

Abstract

Siderophores have long been implicated in sociomicrobiology as determinants of bacterial interrelations. For plant-associated genera, like Bacillus and Pseudomonas, siderophores are well known for their biocontrol functions. Here, we explored the functional role of the Bacillus subtilis siderophore bacillibactin (BB) in an antagonistic interaction with Pseudomonas marginalis. The presence of BB strongly influenced the outcome of the interaction in an iron-dependent manner. The BB producer B. subtilis restricts colony spreading of P. marginalis by repressing the transcription of histidine kinase-encoding gene gacS, thereby abolishing production of secondary metabolites such as pyoverdine and viscosin. By contrast, lack of BB restricted B. subtilis colony growth. To explore the specificity of the antagonism, we cocultured B. subtilis with a collection of fluorescent Pseudomonas spp. and found that the Bacillus–Pseudomonas interaction is conserved, expanding our understanding of the interplay between two of the most well-studied genera of soil bacteria.

Introduction

Interactions between bacteria are of huge relevance to how microbial ecology can be manipulated for biotechnological benefit. In natural environments, bioavailable iron is often scarce and because it is an essential mineral to virtually all life, many organisms have evolved to scavenge iron using high-affinity metal chelators [1]. Metal chelators are a main driver of antagonism between microorganisms both in sequestering metal ions from competing organisms and through bioactive toxicity, directly inhibiting or even killing competitors [2, 3].

The catecholate siderophore bacillibactin (BB) produced by members of the Bacillus genus is an example of an iron chelator that performs multiple functions. In Bacillus subtilis, it is the main iron acquisition metabolite with high affinity for ferric iron, Fe(III) [4]. BB is synthesized through a nonribosomal peptide synthetase pathway encoded by the dhbA-F operon and is exported to acquire Fe(III) before being taken up again by the FeuABC-YusV ABC transporter system [5-8]. Once intracellular, BB-Fe is hydrolysed by BesA to release iron ions for use in enzymatic reactions and in transcriptional regulation through the ferric uptake regulator Fur [7].

The dhbA-F operon also plays an essential role in Bacillus biofilm formation. The ΔdhbA mutant lacks dehydrogenation of (2S,3S)-2,3-dihydroxy-2,3-dihydrobenzoate to 2,3-dihydroxybenzoate (DHB) and therefore is unable to produce pellicle biofilms in the biofilm-promoting-defined medium MSgg due to an inability to perform extracellular electron transport with ferric iron captured by DHB [9]. Additionally, BB from Bacillus amyloliquefaciens has even been suggested to exhibit direct bioactivity against the plant pathogen Pseudomonas syringae pv. tomato, though without direct experimental evidence [10], and recent studies have demonstrated the antibacterial properties of BB isomers against Pseudomonas aeruginosa [11, 12].

Bacillus and Pseudomonas are bacterial genera with high environmental impacts. They are environmentally ubiquitous and often coisolated, especially within the context of crops. Here, they function as biostimulants and biocontrol agents, but also, in the case of Pseudomonas, as phytopathogens [13]. Both genera contain members capable of producing bioactive natural products that antagonize competing microbes [14-16], and both can alter the local microbiome [17, 18]. Such abilities have immense implications in areas such as agriculture, where microbiome composition is consistently correlated with plant health and disease outcome [19, 20], and therefore, it is considered as an environmentally friendly agricultural practice that lowers the utilization of chemical pesticides.

Pairwise interactions between bacilli and pseudomonads might therefore play a crucial role in determining plant growth, health, and local microbiome composition. Studies investigating such pairwise interactions are revealing more detail, outlining the molecular mechanisms, and determining the degree to which interactions and their effector molecules are conserved [21-25]. However, despite these advances, the large bioactive potential of Bacillus and Pseudomonas makes mapping the overarching interaction mechanisms challenging, and a consensus has not been reached on the general interactions between members of these two genera. Some studies investigated the role of iron and siderophores, identifying pyoverdine (Pvd) and pyochelin from Pseudomonas as informational cues for Bacillus velezensis [21], and pulcherriminic acid produced by B. subtilis as an iron sequestering agent in interactions with Pseudomonas protegens [26].

Here, we investigated the functional role of B. subtilis BB in antagonizing Pseudomonas. Using a macrocolony competition assay, we observed that B. subtilis DK1042 (hereafter, DK1042) was able to restrict the colony size of Pseudomonas marginalis PS92 (hereafter, PS92) but that the disruption of the BB biosynthetic pathway reversed the interaction, resulting in diminished growth of DK1042. By disrupting each gene in the dhb operon, we found that products of DhbA-F enzymes can be secreted and shared among B. subtilis cells. The interaction depended on iron in the media, but transcriptomic analysis revealed no iron starvation response in cocultures. Instead, the PS92 Gac/Rsm signalling system, herein, Pvd biosynthesis, was strongly downregulated transcriptionally and metabolically when cocultured alongside DK1042. Finally, we found that DK1042 could inhibit several fluorescent Pseudomonas spp., all in a BB-dependent manner.

Methods

Culturing

The B. subtilis DK1042 (the naturally competent derivative of 3610) and P. marginalis PS92 were routinely cultured in lysogeny broth (LB; Lennox, Carl Roth, 10 g/l tryptone, 5 g/l yeast extract, and 5 g/l NaCl), LB supplemented with 1% (v/v) glycerol and 0.1 mM MnCl₂ (LBGM), or King’s B medium (KB; 20 g/l peptone, 1% glycerol (v/v), 8.1 mM K₂HPO₄, 6.08 mM MgSO₄·7H₂O) at 30°C. Colonies were grown on media solidified with 1.5% agar (w/v). FeCl₃, ammonium ferric citrate, and 2,2′-bipyridine (BPD) were added to autoclaved agar at varying concentrations. Antibiotics were added in the following final concentrations: 50 μg/ml gentamycin (Gm), 100 μg/ml ampicillin (Amp), 5 μg/ml chloramphenicol (Cm), and 10 μg/ml Kanamycin (Km).

Pairwise interactions on agar

DK1042 and PS92 were routinely spotted 5 mm apart on agar surfaces using 2 μl of overnight culture adjusted to an optical density at 600 nm (OD₆₀₀) of 1.0 in appropriate media. Prior to spotting, plates were dried for 30 min in a lateral flow hood and subsequently incubated at 30°C. This experimental setup was employed with all agar competition experiments, also with an in-house collection of fluorescent Pseudomonas spp. When spotting mixed cultures, each species was adjusted to an OD₆₀₀ of 1.0, and the two strains were then mixed in equal volumes. For mixed B. subtilis cultures, wild-type (WT) and mutant strains were mixed 1:1, 1:10, or 1:100, based on OD₆₀₀, before spotting 2 μl as detailed above.

Stereomicroscopy

Interactions for stereomicroscopy were assessed using DK1042 carrying amyE::P_hyperspank-mKate2 and PS92 carrying attTn7::msfGFP, which constitutively express fluorophores (mKate2 and msfGFP, respectively). Colonies were imaged with a Carl Zeiss Axio Zoom V16 stereomicroscope equipped with a Zeiss CL 9000 LED light source and an AxioCam 503 monochromatic camera (Carl Zeiss, Jena, Germany). The stereoscope was equipped with a PlanApo Z 0.5×/0.125 FWD 114 mm and filter sets 38 HE eGFP (ex: 470/40, em: 525/50) for msfGFP and 63 HE mRFP (ex: 572/25, em: 629/62) for mKate2. Exposure time was optimized for appropriate contrast.

Image processing and analysis were performed using FIJI (2.1.0/1.53f51) [27]. Contrast in fluorescence channels was adjusted identically on a linear scale. Colony area was measured by segmenting the colony of interest using Otsu’s algorithm [28] based on fluorescence (msfGFP or mKate2 for Pseudomonas or Bacillus, respectively).

Viable cell counting

Entire macrocolonies were harvested after 72 h incubation into 1 ml 0.9% NaCl, vortexed with 100 μl glass beads (r = 100 μm) for 10 min, and sonicated with a probe sonicator for 45 s in 1-s pulses (1 s on, 1 s off) at an amplitude of 25%. Homogenized colonies were diluted and spread on LB + 50 μg/ml gm (selecting for PS92) or LB + 5% NaCl (selecting for DK1042 and ΔdhbA).

Genetic modification

DK1042 mutants were created through homologous recombination using the B. subtilis single gene deletion library [29]. In brief, genomic DNA from a donor B. subtilis 168 mutant was extracted using a EURx Bacterial & Yeast Genomic DNA Purification Kit (EURex, Gdansk, Poland) following the manufacturer’s instructions, and 100–200 ng genomic DNA was added to DK1042 grown to OD₆₀₀ ~ 0.5 in 400 μl at 37°C in competence medium (80 mM K₂HPO₄, 38.2 mM KH₂PO₄, 20 g/l glucose, 3 mM tri-Na citrate, 45 μM ferric NH₄ citrate, 1 g/l casein hydrolysate, 2 g/l K-glutamate, 0.335 μM MgSO₄·7H₂O, 0.005% [w/v] tryptophan). Cells were incubated with DNA at 37°C for 3 h before 100 μl of the transformation mix was spread onto LB agar supplemented with kanamycin and incubated at 37°C for 16–24 h. Mutants were validated by colony PCR with primers designed to anneal 50–150 bp up- and downstream of the gene of interest (Table S1).

PS92 attTn7::msfGFP was constructed by conjugation with pBG42 as described by Zobel et al. [30]. In brief, cultures of Escherichia coli λpir CC118/pBG42 (donor—Gm^R), E. coli HB101/ pRK600 (helper, Cm^R), E. coli λpir CC118/pTNS2 (transposase-carrying, Amp^R), and PS92 (recipient) were started in LB medium supplemented with appropriate antibiotics and incubated overnight. Each culture was washed thrice in 0.9% NaCl before being mixed equally based on OD₆₀₀. The mix was spotted onto LB without antibiotics and incubated overnight at 30°C. Resultant colonies were resuspended in 0.9% NaCl and plated on Pseudomonas Isolation Agar (45.03 g/l; Millipore, Burlington, MA, USA) supplemented with Gm to select for positive Pseudomonas conjugants.

Mass spectrometry imaging

Pairwise interaction colonies were grown on 10 ml agar plates incubated for 72 h before being excised, dried, and sprayed for matrix-assisted laser desorption ionization-mass spectrometry imaging (MALDI-MSI). Agar was adhered to MALDI IntelliSlides (Bruker, Billerica, Massachusetts, USA) using a 2-Way Glue Pen (Kuretake Co., Ltd, Nara-Shi, Japan). Slides were covered by spraying 1.5 ml of 2,5-dihydrobenzoic acid (40 mg/ml in ACN/MeOH/H₂O [70:25:5, v/v/v]) in a nitrogen atmosphere and dried overnight in a desiccator prior to MSI acquisition. Samples were then subjected to timsTOF flex using a Bruker Daltonik GmbH mass spectrometer for MALDI-MSI acquisition in positive MS scan mode with a 40 μm raster width and a mass range of 100–2000 Da. Calibration was performed using red phosphorus. The settings in timsControl were as follows: Laser: imaging 40 μm, Power Boost 3.0%, scan range 46 μm in the XY interval, laser power 70%; Tune: Funnel 1 RF 300 Vpp, Funnel 2 RF 300 Vpp, Multipole RF 300 Vpp, isCID 0 eV, Deflection Delta 70 V, MALDI plate offset 100 V, quadrupole ion energy 5 eV, quadrupole loss mass 100 m/z, collision energy 10 eV, focus pre TOF transfer time 75 μs, prepulse storage 8 μs. After data acquisition, data were analysed using SCiLS software (version 2021b Pro).

Images were prepared from data normalized to the root-mean square. Images from each m/z value were expanded to be of equal size by adding pixels with NaN values if necessary. Linear scaling of pixel intensity was performed across images displaying the same m/z values. NaN is presented as 0 in the final images.

Liquid chromatography mass spectrometry

B. subtilis DK1042 was inoculated into 20 ml LB in a 100 ml flask and grown overnight. The culture was adjusted to OD₆₀₀ = 1.0, and 1 ml was spread on each of 150 LBGM agar plates and incubated at 30°C for 72 h. Extraction and subsequent liquid chromatography mass spectrometry (LC–MS) analysis were performed as previously described [31]. LC–MS was performed on three independent rounds of culture.

Whole-genome sequencing and genome comparisons

Pseudomonas genomes were sequenced and assembled using a hybrid approach with Oxford Nanopore and Illumina sequencing. From a culture grown in LB overnight, genomic DNA was extracted using a EURx Bacterial & Yeast Genomic DNA Purification Kit following the manufacturer’s instructions. Short read library preparation and sequencing were performed at Novogene Co., Ltd on a NovaSeq PE150 platform (Illumina, CA, USA). Long-read library preparation was performed using an SQK-RBK004 Oxford Nanopore Rapid Barcoding kit. Libraries were sequenced on a fresh Oxford Nanopore MinION Mk1B 9.4 flow cell (Oxford Nanopore Technologies, Oxford, UK) for 12 h and base-called simultaneously with Guppy in MinKNOW using the default r941_min_hac_g507 model.

Genomes were assembled using Trycycler (V0.5.3) [32] when the long read sequencing depth was adequate for subsetting or Unicycler (V0.5.0) [33] in all other cases. Adapters were removed using Porechop (0.2.4 https://github.com/rrwick/Porechop), and reads were filtered with Filtlong (V0.2.1 https://github.com/rrwick/Filtlong). Default Trycycler and Unicycler workflows were applied to assemble genomes (see Supplementary Methods for details). Completeness and contamination were assessed using CheckM (V1.1.3) [34], and whole-genome taxonomy was determined with AutoMLST [35] and the TYGS database (V342) [36].

Assembled genomes were annotated with Bakta (V1.6.1) [37], and a gene presence/absence matrix was created with Panaroo (V1.3.2) [38] using its moderate mode and a CD-HIT amino acid identity threshold at 40%. The Panaroo presence/absence matrices were used to create the respective figure.

Transcriptomics

Colonies were collected with a 10 μl loop and placed in RNAse-free screw-cap tubes to sample either the interaction zone (~2 mm into each colony for pairwise interactions) or the entire colony (for monocultures). The collected biomass was solubilized in RNAProtect and frozen in liquid nitrogen before being stored at −80°C prior to RNA purification, for no more than 3 days.

RNA was extracted using a Qiagen RNeasy PowerMicrobiome RNA extraction kit (QIAGEN N.V., Venlo, The Netherlands) via combined bead beating and phenol–chloroform extraction (see Supplementary Methods for details). Ribosomal RNA depletion, library preparation, and sequencing were performed at Novogene Co., Ltd on a NovaSeq PE150 platform.

Read abundances were determined using Kallisto (0.48.0) [39]. The genome of PS92 was annotated with Bakta (V1.6.1) [37], while the genome of B. subtilis DK1042 (Acc: CP020102 and CP020103) was retrieved from GenBank. Genomes were converted into a fasta-formatted transcript file with gffread (v0.11.7) [40] then combined into a Kallisto index.

RNA sequencing datasets were trimmed and filtered with FastP, and reads were quantified by mapping every sample to the combined index file (consisting of transcripts from both species), with 100 bootstrap replicates. Differential expression analysis was performed with DESeq2 (v1.34.0) [41] in Rstudio (2022.02.3-b492) [42] running R (4.1.1) [43] with the Tidyverse framework (1.3.1) [44].

Pathway analysis was performed by functionally annotating the transcriptome of each species using eggNOG-mapper V2 (online browser version) with default parameters [45]. Differentially expressed pathways were investigated using KEGG Mapper Reconstruct [46], and manually using the databases Subtiwiki [47] and The Pseudomonas Genome Database [48] as references.

Transcriptomic analysis was performed from three independent replicates for each sample.

Statistical analysis and reproducibility

Differences between group means were tested with analysis of variance (ANOVA), coupled with Tukey–Kramer’s post hoc analysis to determine significantly different means between groups. We adopt df_w to represent the degrees of freedom within groups (i.e. df_w = n—k, where n is the total number of samples and k is the total number of groups). One-way repeated measures ANOVA (Fig. 1B) was calculated using the three time points as within-group identifiers. The nonparametric Friedman test was used as data did not satisfy normality requirements as determined with Shapiro–Wilk’s normality test. Mean difference between two groups was compared using the nonparametric Wilcoxon rank-sum test. Multiple testing in Fig. 1C was corrected using the Benjamini–Hochberg method. For Figs 2B and C and 5, relative colony sizes of DK1042 and ΔdhbA grown on KB for 72 h (Fig. 1B) were included as control groups.

Figure 1B was created from paired groups (time points) of 80 independent datapoints (i.e. 80 pairs of co- and monoculture colonies) from two independent rounds of experiments. Each colony was followed over 72 h, and the subsequent ANOVA was adjusted for paired samples.

All reported P-values are adjusted for multiple testing when appropriate.

Results

Bacillibactin is involved in antagonism between DK1042 and PS92

Molina-Santiago and colleagues previously described that on solid KB medium (a growth medium for Pseudomonas-produced Pvd detection), DK1042 surrounds and restricts the growth of Pseudomonas chlororaphis PCL1606, unlike on various other media used in their experiments [24]. Therefore, we tested a soil-isolated Pseudomonas species (designated PS92) and revealed an interaction phenotype comparable to that of PCL1606. Since antagonism was apparent on KB medium, which induces Pvd production in Pseudomonas, and not on LB medium (Supplementary Fig. S1A), we hypothesized that siderophores from either DK1042, PS92, or both would be involved in this interaction. To evaluate the effect of BB in the interaction between DK1042 and PS92, we cultured fluorescently labelled versions of the two species together in close-proximity macro colonies on solid agar medium. Spots (2 μl) of each culture were placed 5 mm apart and grown at 30°C for 3 days (Fig. 1A and Supplementary Video S1). We compared WT DK1042 and the single gene disruption mutant (ΔdhbA) lacking the ability to produce DHB and BB.

By measuring the areas of colonies, we determined that the growth of PS92 cocultured next to DK1042 was restricted, and after 72 h, the area of monoculture colonies was approximately twice that of the area of coculture colonies (Fig. 1B; Friedman one-way repeated measured ANOVA, P = 1.67 × 10⁻²⁵, n = 60, df = 2). The ΔdhbA mutant did not surround nor restrict the growth of PS92 (P = .321, n = 80, df = 2), but rather was itself restricted ∼2-fold (P = 3.71 × 10⁻²⁸, n = 80, df = 2). The restriction in area is reflected in the number of colony-forming units from colonies in each interaction, showing a 16-fold reduction in cells in PS92 colonies grown next to DK1042 (Fig. 1C; Wilcoxon rank sum test, P = 1.08 × 10⁻⁵, n = 20, df = 18) and a 10³-fold reduction in cells from ΔdhbA-mutant colonies grown next to PS92 (P = 8.23 × 10⁻⁵, n = 17, df = 15). When coculturing the strains in liquid medium, we observed inhibition of PS92 in KB broth regardless of whether the coculture was with DK1042 or ΔdhbA. Neither DK1042 nor ΔdhbA was antagonized by PS92, demonstrating that the observed antagonism is specific to solid media (Supplementary Fig. S1D and E). We also observed that the interaction occurs on LBGM but not on MSgg, and that glycerol is likely implicated in the antagonism (Supplementary Fig. S2).

In previous studies, BB production was reported to be linked to iron stress conditions [10]. As both LB and KB contain sources of natural iron from tryptone or yeast extract, low BB production could be expected on both types of media. LC–MS was used to compare BB production from DK1042 inoculated in liquid and solid LB and KB, revealing a 10-fold increase in BB concentration on solid KB compared to solid LB (Fig. 1D), while virtually undetectable in liquid media (Supplementary Fig. S3). The dependency on dhbA for antagonism and defence coupled with the exclusive production on solid KB suggests either a BB or DHB-mediated regulation in DK1042. However, a previous study showed that in B. subtilis biofilms, BB but not DHB is dispensable in air–liquid interface pellicle biofilms [9]. Thus, we tested whether BB or any precursor from the dhbA-F pathway influenced the pairwise interaction (Fig. 2A) and found all mutants unable to antagonize PS92 (Fig. 2B; one-way ANOVA with post hoc Tukey HSD, P < 2.23 × 10⁻¹⁰, n > 10, df_w = 99), suggesting that BB, and not an intermediate, must be present for antagonism.

Bacillibactin precursors are “public goods”

Since DHB has been shown to function as a siderophore in biofilms [7], we hypothesized that it could be used as a “public good” (here, a resource that can be produced by and shared among members of the population) and taken up by ΔdhbA mutant cells. Therefore, we mixed mutants disrupted in early biosynthesis (ΔdhbA and ΔdhbB), late biosynthesis (ΔdhbE and ΔdhbF), and mutants disrupted in both early and late biosynthesis (ΔdhbA and ΔdhbF) that should be able to complement each other through ΔdhbF supplying DHB to ΔdhbA, which then completes the synthesis of BB. Only the mixture of ΔdhbA and ΔdhbF was able to antagonize and restrict the growth of PS92 on solid KB medium (Fig. 2C; one-way ANOVA with post hoc Tukey HSD, P = 5.92 × 10⁻⁸, n > 10, df_w = 207), which not only demonstrates the ability of the mutants to complement each other, but also confirms that BB, and not DHB, is responsible for the antagonistic interaction between DK1042 and PS92. This was further corroborated by mixing DK1042 with ΔdhbA (unable to produce DHB) or ΔdhbF (unable to synthesize the nonribosomal peptide) using 1:1, 1:10, and 1:100 ratios, respectively, and spotted mixtures next to PS92 (Supplementary Fig. S4). Both mutants were rescued by the WT strain, even at a ratio of 1:100, and expansion of PS92 colonies was restricted independently of which mutant was used in the mixture or the ratio of the mixed strains.

Excess iron abolishes antagonism

Given that BB is the main siderophore of B. subtilis, we hypothesized that the antagonism was driven by iron in the media. To test this hypothesis, we cocultured DK1042 and PS92 on KB agar supplemented with varying concentrations of the ferrous iron chelator BPD (Fig. 3A) or with excess ferric iron either as ferric citrate, which is not transported into the cell via the BB and FeuABC-YusV machinery (Fig. 3B), or as FeCl₃ from which iron is readily bound to DHB and BB (Fig. 3C) [6, 9]. Curiously, sequestration of Fe(II) and addition of Fe(III) resulted in increased coculture colony size for the ΔdhbA mutant (one-way ANOVA resulted in nonsignificant differences between groups). Sequestering ferrous iron did not lead ΔdhbA to regain the ability to restrict PS92 size (one-way ANOVA with post hoc Tukey HSD, n > 4, df (residuals) = 29, P < .03), but, independently of concentration, both the addition of FeCl₃ and ferric citrate did (Fig. 3B and C), even though ΔdhbA should be unable to utilize FeCl₃ [49]. Similarly, the ΔbesA mutant, unable to release iron from BB, was also complemented by any iron-related change to the medium. However, the ΔfeuA mutant did not respond to FeCl₃ (as expected) but could be complemented by ferric citrate in a dose-dependent manner.

DK1042 heavily alters the PS92 transcriptome

To better understand the transcriptional landscape present in each strain as a result of cocultivation, we extracted total RNA from colonies cultured in monoculture or with an interaction partner (n = 3 for all samples). This allowed us to identify transcriptional differences between DK1042 and ΔdhbA both in monoculture and coculture, but also between each individual strain in monoculture compared to coculture to ascertain how a neighbouring partner alters the transcription of a focal strain (Supplementary Fig. S5). Transcriptomic analysis revealed that the PS92 transcriptome was heavily altered by DK1042 (1036 differentially regulated genes at log₂(FC) > |2| and adjusted P < .01) but not by ΔdhbA (9 differentially regulated genes). Approximately, 4%–8% of the transcriptomes of DK1042 and ΔdhbA (364 and 172 differentially regulated genes, respectively) were differentially regulated by the presence of PS92, but comparison of cocultures revealed that 1002 transcripts were differentially regulated between the two strains in coculture. In monoculture, 395 transcripts were differentially regulated between DK1042 and ΔdhbA.

ΔdhbA downregulates sporulation and biofilm formation

When coculturing DK1042 with PS92, genes related to sporulation and germination were upregulated in DK1042 (Fig. 4A), though not differentially regulated in the ΔdhbA mutant in coculture compared with its monoculture. However, comparing the levels of transcription between DK1042 and ΔdhbA reveals that lack of dhbA strongly downregulated sporulation and germination pathways, independently of whether it was cocultured with PS92.

Elimination of DHB and BB production affected several genes related to biofilm formation and sporulation in the epsA-O, tapA-tasA-sipW, spsA-L, cot and spoII, III, IV, V, and VI operons, an effect reported previously [9, 49]. In addition, the transcription level of the sinI gene was higher in monoculture for ΔdhbA compared with DK1042, though not in coculture. In coculture, the lack of dhbA upregulated the transcription of the biofilm repressor gene abrB.

Repression of sporulation and biofilm formation was further corroborated by the downregulation of genes involved in utilization of glutamine, manganese, and glycerol. Manganese and glutamine/glutamate utilizations are well known for their role in the Spo0A sporulation pathway that drives both biofilm formation and sporulation [50-53], while manganese homeostasis was very recently shown to be tightly interlinked with iron homeostasis [54].

The transcriptome of the ΔdhbA mutant did not display signs of iron starvation, with the exception of a slight repression in the transcription of fur when comparing transcription of monocultures. The dhb, feu, yusV, besA, and yvm genes were not differentially regulated in the ΔdhbA mutant compared with DK1042, except the dhbA transcript level, which was deleted in the genome of the mutant and therefore transcriptionally absent in the mutant. Additionally, genes involved in production of known bioactive secondary metabolites were not differentially regulated in Bacillus regardless of the mutation or coculture partner.

PS92 downregulates Gac/Rsm and iron acquisition in coculture with DK1042

When cocultured with DK1042, PS92 underwent downregulation of biosynthetic genes responsible for production of several secondary metabolites (Fig. 4B and Supplementary Dataset S1). All biosynthetic genes involved in Pvd, alginate (Alg), and viscosin production were repressed along with a Type VI secretion system (T6SS) subtype i3, while genes responsible for Type II and III secretion and reactive oxygen neutralization were upregulated. Similar changes have previously been observed in the closely related P. fluorescence SBW25 when removing the histidine kinase gacS (OrthoANIu similarity between SBW25 and PS92 = 99.06%) [55]. The Gac/Rsm pathway is a conserved two-component system found in many γ-proteobacteria and known to positively regulate biofilm formation and secondary metabolism in most pseudomonads and induce virulence factors in plant pathogenic species. Indeed, PS92 gacS (FNPKGJ_17835) was downregulated 4-fold in PS92 when cocultured with DK1042 but not when cocultured with ΔdhbA.

Although Pvd biosynthetic genes were downregulated, the iron starvation sigma factor-encoding gene pvdS (FPNKGJ_22055) had a higher transcript level in coculture with DK1042. Pvd is the primary siderophore in SBW25 [56], though other iron-chelating molecules are encoded in the genomes of both SBW25 [55] and PS92. A biosynthetic gene cluster encoding the production machinery for a corrugatin-like molecule (FNPKGJ_15100–15175) was similarly repressed (P.adj < .05), although not below the threshold of log₂(FC) < −2.

We found further evidence for the repression of the Gac/Rsm system in mass spectrometry imaging of cocultures, which showed that several molecules produced by genes under the control of the Gac/Rsm system were absent from cocultures with DK1042 but present in monoculture colonies and cocultures with ΔdhbA (Fig. 4C). The presence of viscosin in all cultures except coculture with DK1042 on KB prompted us to test if viscosin was the antagonistic molecule against ΔdhbA. In lieu of a PS92 ΔviscA gene deletion mutant, we obtained a Pseudomonas fluorescens SBW25 ΔviscA mutant with an identical interaction with DK1042, but found that ΔviscA was still able to antagonize ΔdhbA, similarly to the WT strain (Supplementary Fig. S6).

Antagonism depends on bacillibactin

Based on the transcriptome data for DK1042, we screened multiple gene disruption mutants for their antagonistic effect against PS92 (Fig. 5). Effects of mutations ranged from deficiency in the production of biofilm matrix (epsA-O, tasA) and secondary metabolites (srfAA, pksL, sfp, cypX, yvmC) to lacking sensor kinases (kinA-E) or being unable to undergo sporulation (spo0A). Only three mutants were significantly unable to antagonize PS92 (one-way ANOVA with post hoc Tukey HSD, n = 4, df_w = 212; Δsfp-DK1042 P = 1.32 × 10⁻⁶, ΔfeuA-DK1042 P = 1.29 × 10⁻⁵, and ΔglpK-DK1042 P = .019). Thus, neither sporulation nor biofilm formation are essential for antagonism, and likewise, typical secondary metabolites with antimicrobial properties such as surfactin and bacillaene had no effect on the restriction of PS92 colony area (although this was suggested from the mass spectrometry imaging). The inability of ΔfeuA to antagonize suggests that iron uptake by BB is essential for restriction of PS92 colony size, but the fitness of the ΔfeuA mutant was rather low even in monoculture (Supplementary Fig. S7), which could result in low metabolic output in general.

Interaction is conserved across fluorescent Pseudomonas isolates

To determine if the iron-related antagonism is specific to DK1042 and PS92, we cocultured DK1042 and ΔdhbA on KB agar with a collection of 16 fluorescent Pseudomonas soil isolates for which genomes had been sequenced (Fig. 6). Thus, we could qualitatively compare the interaction outcome with the genomic presence of biosynthetic pathways. Multiple isolates were restricted in colony size by DK1042 and not by ΔdhbA. Most isolates also seemed to influence the size and/or morphology of ΔdhbA, except XL272 (previously Pseudomonas stutzeri, now Stutzerimonas degradans), which is the only isolate lacking the Gac/Rsm-controlled biosynthetic machineries producing Pvd and Alg. A few isolates belonging to Pseudomonas corrugata and protegens subgroups were able to inhibit both DK1042 and ΔdhbA, which may be attributed to their potential for producing 2,4-diacetylphloroglucinol and pyoluteorin [57]. Strains from other P. fluorescens subgroups were all restricted in colony growth in the presence of BB-producing B. subtilis, demonstrating the conservation of this antagonism across species.

Discussion

In this study, we report a novel siderophore-mediated antagonism conserved between B. subtilis and multiple Pseudomonas spp. We show that the interaction is dependent on iron and BB, and that a lack of BB on KB agar results in both diminished offence and defence. The consequence of BB-mediated antagonism is a restriction of Pseudomonas colony spreading and a reduction both in transcription and production of several products related to the secondary metabolism-regulating Gac/Rsm system, which over time allows Bacillus to overgrow a Pseudomonas colony.

Our results show a strong implication of BB in the antagonism between DK1042 and PS92. Disrupting individual iron homeostasis-related genes diminished their antagonistic properties, and coculturing mutants in mixed cultures restored their properties. This indicates that the critical components of the DhbA-F pathway can be secreted and shared between members in a population as “public goods.”

Bacillus colony expansion in coculture is especially affected by iron availability (Fig. 3). Though the iron concentration added to the medium in these experiments was relatively high, the results suggest that the interaction, both defensive and offensive, is iron-dependent. Iron homeostasis, biofilm formation, and sporulation are tightly interlinked in B. subtilis, and DhbA is essential for structured biofilm formation and proper sporulation due to the high iron demands of these lifestyles [9, 49, 58]. In our coculture assays, lack of dhbA negatively affected multiple biofilm-related genes repressing biofilm formation and sporulation. This suggests that the ΔdhbA mutant did not form a biofilm or undergo sporulation either in coculture or in monoculture. The production of a matrix and the formation of spores as defensive measures in microbial ecology have received little attention [59] despite investigations into their relation to T6SS [24, 60]. Defence appears to take precedence for Bacillus in the present antagonism, and a biofilm- and sporulation-deficient mutant may be too antagonized by PS92 to produce the necessary effector for growth restriction. This could explain our observations with ΔfeuA which requires large quantities of supplemented iron for growing comparably to the wild type (Fig. 3B and C). Additionally, a biofilm-proficient strain lacking the antagonistic molecule may block PS92 growth through spatial competition [61-63], though it will be unable to restrict colony spreading chemically. Another possibility could be that BB is a stronger iron chelator than Pvd, sequestering Fe(III) away from Pseudomonas and thereby inducing iron starvation. When Bacillus is unable to produce BB, the tide turns, and Pvd from Pseudomonas sequesters Fe(III) away from Bacillus. Pvd production on KB has been shown in several pseudomonads to be negatively correlated with iron content, where the iron-binding Fur represses Pvd production through repression of pvdS [56]. This could explain how ΔdhbA, ΔbesA, and ΔfeuA are rescued by iron supplementation even in the form of ferric chloride (Fig. 3B). When iron is present, Pvd is not produced, and the Bacillus iron homeostasis mutants are not inhibited by iron competition. With this mechanism, strong iron starvation responses are likely in both Bacillus and Pseudomonas, but in ΔdhbA, iron homeostasis was not significantly upregulated in coculture with PS92 (Fig. 4A). On the other hand, pvdS in PS92 was significantly upregulated (Fig. 4B), which is indicative of iron starvation, but without the complementary upregulation of the rest of the Pvd biosynthetic genes. Thus, it is unclear whether the consequence of coculture is iron starvation, repression of the Gac/Rsm pathway, or a combination of both.

In P. aeruginosa, Gac/Rsm is known to positively regulate Pvd biosynthetic gene transcription trough PvdS [64, 65], and a similar phenotype has been described in P. fluorescens SBW25 [55]. Recently, both Pvd and, to a larger extent, pyochelin were shown to be sensed by Bacillus velezensis, instigating mobilization of secondary metabolites [21]. Although sensing of pyochelin occurs independently of iron and BB, Pvd was shown to increase the production of BB. PS92 does not contain a gene cluster encoding the synthetic machinery for any type of pyochelin, and thus, pyochelin sensing presumably does not occur in the interaction between DK1042 and PS92.

Our results suggest a Bacillus-mediated repression of gacS, a phenomenon with precedent from interaction between a Bacillus sp. isolate and P. syringae pv. Tomato DC3000 [66]. However, neither previous nor our experiments reveal if gacS repression is directly caused by DK1042 or by a sensory mechanism in PS92. Currently, there is no consensus on exactly which ligand triggers the Gac signal cascade [67], although it is expected that secreted products resulting from the activated pathway itself are at least partly responsible [68]. However, in Pseudomonas, GacS is activated and inhibited by the transmembrane receptors LadS and RetS, respectively, which in turn are activated by calcium ions (LadS) and mucin glycans or temperature (RetS) [69-71]. It is also hypothesized that the Gac/Rsm pathway functions via an alternative quorum sensing pathway, and that induction is achieved at higher cell densities [68]. This might explain why gacS was downregulated in PS92 spotted next to DK1042, where the colony is indeed restricted in size and at a lower cell density. Large-scale purification and chemical synthesis of BB were unsuccessful in our study; hence, we could not directly probe the antagonistic properties of BB. Siderophores have long been implicated in microbial interactions, particularly in iron-scarce environments. In infections, Pvd and enterobactin are essential for P. aeruginosa and E. coli, respectively, not only due to iron requirements in a complex environment, but also in polymicrobial interactions as interspecies competitive factors [72]. Siderophores are also thought to play a role in the biocontrol of plant pathogens by both Pseudomonas and Bacillus, but a thorough examination of the literature reveals that Bacillus and Pseudomonas siderophores have only rarely been directly implicated in interspecies interactions in planta [13]. Most evidence originated in the previous millennium where Pseudomonas siderophores proved to be determinants of microbial interactions through iron sequestration [73, 74]. Others have since inferred biocontrol potential from the ability to produce siderophores without experimentally determining the consequence of losing siderophore production [10, 75]. Our study systematically investigated the significance of BB and its effects on transcription and metabolite production in microbial interactions. Although we demonstrated the importance of BB on rich media, our study complements and expands the previously established role of BB in biocontrol. We found that Bacillus inhibition of colony expansion could be replicated qualitatively with multiple fluorescent Pseudomonas spp., and that disrupting dhbA abolished the antagonism (Fig. 6). The generality of this interaction against multiple pseudomonads is intriguing, in the potential for B. subtilis to compete against Pseudomonas spp. in soil. Although our P. marginalis isolate proved nonpathogenic against lettuce, cabbage, and spinach, many pseudomonads are pathogenic and could potentially be controlled by Bacillus producing BB. As BB production is conserved across many species of Bacillus, one would expect that other bacilli would display similar BB-dependent antagonism against Pseudomonas, though this remains to be experimentally validated. The lack of PS92 plant colonization meant that we were unable to determine if BB-mediated antagonism occurs in the phyllosphere as well as on agar surfaces. The medium specificity of this interaction is not surprising, as pairwise interactions have been found to always be context dependent [76]. We observed glycerol to possess some role in the antagonism, but only on solid medium (Supplementary Fig. S2). It is unclear why the antagonism was not present in the same form in liquid media (Supplementary Fig. S1D and E). Other studies have successfully transferred biocontrol properties from in vitro to in planta [77], but the medium dependency of the antagonism between DK1042 and PS92 might render it difficult to apply to agriculture. Regardless, understanding the molecular mechanism underlying the antagonism could allow the environment to be tailored specifically to the intended interaction. Future studies should investigate the apparent iron competition between BB and Pvd, as well as the hypothesis that BB represses Gac sensing in fluorescent pseudomonads.

Acknowledgements

The authors thank Lars Jelsbak and Morten Lindquist Hansen for contributing Pseudomonas soil isolates for this study, and the DTU Metabolomics Core for MSI and LC–MS instrumentation.

Author contributions

Mark Lyng and Ákos T. Kovács designed the study; Mark Lyng, Johan P.B. Jørgensen, and Scott A. Jarmusch collected and analysed data; Morten D. Schostag, Diana K.C. Aguilar, and Carlos N. Lozano-Andrade contributed with methodology or preliminary data; Mark Lyng and Ákos T. Kovács wrote first draft of paper with edits from all authors.

Conflicts of interest

None declared.

Funding

This project was funded by a DTU Alliance Strategic Partnership PhD fellowship, by the Danish National Research Foundation (DNRF137) for the Center for Microbial Secondary Metabolites, and the Novo Nordisk Foundation within the INTERACT project of the Collaborative Crop Resiliency Program (NNF19SA0059360) and for the “Imaging microbial language in biocontrol (IMLiB)” infrastructure grant (NNF19OC0055625).

Data availability

Analysis scripts and processed data have been deposited at Github. (https://github.com/marklyng/dhb_repository). Genome assemblies have been deposited at NCBI (accession numbers are listed in Table S1). Raw sequence reads (long- and short-gDNA reads and RNA) have been deposited at the Sequencing Read Archive (SRA; https://www.ncbi.nlm.nih.gov/sra) with BioProject ID PRJNA956831. Raw microscopy images have been deposited at Dryad (https://datadryad.org/; DOI https://doi.org/10.5061/dryad.vq83bk3zc). MSI data have been deposited at Metaspace [78] under project ID. https://metaspace2020.eu/project/lyng-2023. LC–MS data have been deposited at GNPS-MassIVE under MSV000092142.

References