Introduction

Surface-associated bacteria can be found throughout ecology, from the cells adhering to our teeth to the dense bacterial colonies growing on plant roots. On the surface, cells form collectives that strongly compete for space. In doing so, cells often secrete antimicrobial compounds that antagonize competitors on the surface (e.g [1]). Cells can also work together and thereby promote collective growth or dispersal [2,3,4,5,6]. This is perhaps best illustrated during colony expansion, when cells rapidly colonize a surface in order to monopolize resources. In many bacterial species, colony expansion is facilitated by a coordinated movement of cells. Depending on the driving forces, this form of collective movement is referred to as swarming, gliding, or sliding [7]. For example, in the case of swarming, cells collectively express flagella that advance the colony boundary through propulsion. Similarly, bacteria can also slide over surfaces, which is passive form of surface movement that is driven by cell division and therefore occurs independently from cellular appendages like flagella. For example, in Bacillus subtilis, sliding motility results from cells forming filamentous bundles that push themselves away from the colony center as they grow [8].

Many forms of collective movement, including swarming and sliding, depend on the secretion of extracellular molecules [9,10,11,12]. For example, in the case of B. subtilis, sliding depends on the production and secretion of the lipopeptide surfactin, the bacterial hydrophobin protein BslA, and exopolysaccharides (EPS) [13, 14]. Without these molecules, colonies cannot spread. The secreted molecules carry out different functions. For example, surfactin acts as a surfactant and thereby lowers the surface tension, which facilitates lateral expansion. Through its hydrophobic properties [15, 16], BslA is assumed to have a similar biophysical effect. In contrast, EPS is required for the formation and expansion of filamentous bundles [8]. EPS might furthermore promote colony expansion by increasing the osmotic pressure within the colony [17].

Despite promoting surface expansion, cells that produce the above molecules do not necessarily gain a competitive advantage. For instance, when molecules are costly to produce and benefit other cells on the surface, producers might be exploited by close-by non-producers. This results in a so-called common goods problem, where the production of a common good, in this case a secreted molecule that promotes surface expansion, is selected against despite its beneficial effect on surface spreading. For example, in the case of B. subtilis colony biofilm growth, it was previously shown that EPS production promotes expansion, but is nevertheless selected against because EPS-deficient cells outgrow EPS-producing cells in competition for space [6].

There are several factors that can make it challenging to determine whether secreted molecules act as common goods. First, the fitness costs associated with molecule synthesis can differ between conditions. When molecule production is costly, i.e. lowering the rate of cell division, under specific conditions only (e.g. nutrient limitation), there is only a risk of exploitation under those conditions [18,19,20]. Second, the spatial scale at which molecules are shared may vary [21]. For instance, B. subtilis cells broadly share EPS molecules during both complex colony development and pellicle formation [22,23,24,25,26,27,28,29,30]. However, when it comes to sliding, EPS deficient mutants were shown to be excluded from the expanding colony edge [8], which suggests that EPS might not be broadly shared during sliding conditions. Cells can also directly limit the diffusion of secreted molecules themselves. For example, Drescher and colleagues [5] demonstrated that Vibrio cholerae cells limit the diffusion of chitinase by forming dense biofilms. In the absence of biofilm formation, chitinase production, which benefits surface growth by degrading chitin, is exploited by non-producing cells. Also, it was recently shown that the propagation of quorum-sensing signals can be limited by active signal uptake [31]. Third, secreted molecules may also provide direct benefits to the producer, thereby functioning as private or semi-common goods. For instance, in the budding yeast Saccharomyces cerevisiae, cells secrete an enzyme called invertase to hydrolyze sucrose. Although most monosaccharides produced by hydrolysis diffuse away, about 1% is immediately taken up by cell, thereby providing an immediate private benefit [32].

Given these challenges in evaluating potential common goods, it often remains unknown how molecules, secreted during collective growth, impact spatial competition between producing and non-producing cells. Here, we therefore systematically study how the secreted molecules, essential for sliding motility in B. subtilis, affect intraspecific spatial competition. We performed repeated competition experiments, using both knockout strains and inducible strains that differ in the production and secretion of extracellular molecules. We estimated the costs of molecule production and assessed whether knockout strains can be complemented by providing secreted molecules exogenously. Finally, we also used a spatial-explicit model to investigate the potential role of diffusion and private benefits of molecule production on the outcome of competition. Together, the experiments and model reveal the intricate extracellular biology that underlies surface competition, in which the costs of molecule synthesis, private benefits of molecule production, and the diffusion rate can all influence the outcome of competition.

Materials and methods

Strains and cultivating conditions

All B. subtilis strains (Table S1) used in this study are derivatives of a competence enhanced NCBI3610 (DK1042 [33]). A Δhag mutant lacking flagellin was used as “sliding wild type” since motile strains would swarm and sliding could not be investigated under the used conditions. To engineer mutant derivatives (see detailed description in Supplemental material and used primers in Table S2), a B. subtilis receptor strain was transformed with genomic DNA from a donor strain (method after [33]) or with the respective plasmid. The following antibiotic concentrations were used for respective resistant strains if appropriate: ampicillin 100 µg/ml (Amp), spectinomycin 100 µg/ml (Spec), kanamycin 5 µg/ml (Km), chloramphenicol 5 µg/ml (Cm), tetracycline 10 µg/ml (Tet), MLS: erythromycin 1 µg/ml + lincomycin 12.5 µg/ml.

Sliding competition assay

Sliding assays were performed as described previously [34]. Briefly, overnight cultures of B. subtilis were density normalized and if required, mixed in a 1:1, 1:10, or 10:1 ratio with a competitor strain. Two µl of the strain or mixture were spotted on semi-solid lysogeny broth (LB, Lennox, Carl Roth) supplemented with 0.7 % agar in 9 cm diameter plates that were dried 20 min prior and 10 min post inoculation. The plates were incubated at 37 °C for 24 h if not stated otherwise and sliding was evaluated by assessing the diameter of the expanded colony. Additionally, distribution of fluorescently labeled strains within the sliding disk was evaluated by detecting the fluorescence signal with an AxioZoom V16 fluorescence stereomicroscope equipped with a Zeiss CL 9000 LED light source, HE eGFP filter set (excitation at 470/40 nm and emission at 525/50 nm), HE mRFP filter set (excitation at 572/25 nm and emission at 629/62 nm), and an AxioCam MRm monochrome camera (Carl Zeiss Microscopy GmbH; for exact details of the instrument, see [28]). For image display, the brightness of all fluorescence images was adjusted in the same way and the background was subtracted using the program ImageJ with the rolling ball option (1100 pixels radius).

Determination of occupied area in the sliding colony

The area each strain occupies in the sliding colony was determined using fluorescence stereomicroscope images of the mixtures of different fluorescently labeled strains and the software ImageJ (detailed description in [34]). Briefly, the images were opened in ImageJ, separated by channel, and the scale was changed to pixel. After background removal, a defined threshold was applied to the image of the green and red channel to separate the fluorescence signal corresponding to the occupied area of the respective strain. The total area above the threshold was selected and measured as number of pixels. To compare both strains in the sliding colony, the ratio of the two areas was calculated. A summary of the area and percentage calculation was not possible since there was often at least a small overlap between the areas of the different strains, especially in the center of the sliding colony. For this analysis, original images were used.

Complementation assay

To complement sliding of mutants with externally supplied goods, cultures and plates were prepared as described above (see Sliding competition assay). Additionally, 2 µl of the respective compound was spotted 5 min before the mutant culture (TB532, TB534, TB536) on the same inoculation point. The respective mutant and wild-type cultures alone were used as controls. For surfactin complementation, 10, 1, 0.1, and 0.01 mg/ml of commercial surfactin (Sigma Aldrich; CAS# 24730-31-2) dissolved in methanol was used and pure methanol was spotted as control.

For BslA complementation, the lysate of a BslA-producing Escherichia coli BL21 strain (NRS4110 containing plasmid pNW1128 [16]) and an E. coli BL21 strain without the BslA-production plasmid was used. To obtain the E. coli lysate, the strains were grown overnight in LB medium and were afterwards inoculated 1:100 in autoinduction medium [35]. The NRS4110 culture was always supplemented with 100 µg/ml Amp. The cultures were grown for 7–8 h at 37 °C with 225 rpm shaking and cells were collected by centrifugation at 4000 × g for 15 min. The pellet was resuspended in 5 ml PBS buffer supplemented with 1 mM EDTA and cooled on ice for 10 min. The suspension was then sonicated using an Ultrasonic Processor VCX-130 (Zinsser Analytics, Frankfurt am Main, Germany) with ten repeats of a 10 s pulse of 45% amplitude. During sonication, the suspension was cooled on ice. To obtain the lysate, the suspension was centrifuged (5000 × g, 15 min), the supernatant was collected, filter sterilized, and stored at 4 °C.

For exopolysaccharide complementation, EPS was isolated from pellicle biofilms of a B. subtilis NCIB 3610 wild-type strain, a ΔtasA, and as control a Δeps mutant with slight modifications as previously described [36]. Briefly, four wells of a 24-well plate containing 2 ml biofilm promoting liquid MSgg medium [37] each were inoculated 1:100 with overnight culture of the respective strain and incubated at 30 °C for 2–3 d. The pellicle formed at the air-liquid interface was collected together with the medium, diluted 1:1 with PBS buffer, and vortexed. The pellicles were sonicated (Ultrasonic Processor VCX-130, Zinsser Analytics, Frankfurt am Main, Germany; 2 × 12 pulses of 1 s with 30% amplitude), 0.2 M NaOH was added to a final concentration of 0.1 M, and the samples were incubated at room temperature for 10 min with short periodic vortexing. Afterwards, the samples were chilled on ice for 5 min before adding cold 0.4 M HCl to a final concentration of 0.1 M. The samples were centrifuged (7000 × g for 15 min at 4 °C), the supernatant was collected and transferred to a three-fold volume of cold 96% ethanol and incubated for ca. 20 h at 4 °C. The precipitated EPS was collected by centrifugation (7000 × g for 15 min at 4 °C) and the pellet was dried overnight at 55 °C. After drying, the EPS was re-dissolved with deionized water in a 1:10 ratio, and NaCl was added to a final concentration of 0.5%. The precipitation, collection, and dissolving step was repeated once and the isolated EPS was filter sterilized and stored at 4 °C. The functionality of the BslA-containing lysate and the isolated EPS was successfully verified by testing them for biofilm formation.

Fitness assay

To determine the relative fitness under sliding conditions, strains with inducible gene constructs of epsA, bslA, and srfAA (TB875, TB873, TB977, respectively) were competed against the respective mutants (TB893, TB922, TB895, respectively). Therefore, the strains were density normalized and mixed in a 1:1 ratio in a reaction tube. This mixture was used to (a) determine the colony forming units (cfu) at the start by plating on antibiotic containing plates selective for each strain and (b) to inoculate a 50 ml Schott bottle containing 5 ml LB medium in a 1:100 dilution. The cultures were incubated at 37 °C and 225 rpm shaking for 6–6.5 h after which they were diluted 1:100, and grown again for 6–6.5 h under the same conditions. Following incubation, the final cfu of the two competing strains was determined as described above. Initial and final cfu were then used to calculate the relative fitness via the so called malthusian parameter after Lenski et al. [38] with r = (m inducible strain)/(m mutant) and m = ln [(final cfu)/(initial cfu)].

Mathematical model

In our experiments, we observed that for both BslA and EPS the outcome of competition deviated from a classical common goods dilemma: there were no apparent fitness costs for BslA and EPS production and the Δeps and ΔbslA knockout mutants did not exploit the wild type. Yet, we did observe that at least partially Δeps and ΔbslA mutants complemented each other during sliding motility. To determine what factors could potentially explain these competition outcomes, we decided to construe an agent-based model. We reasoned that the competitive advantage of the wild type over the Δeps and ΔbslA mutants might result from (i) private benefits that the wild type might obtain from molecule production or (ii) from limited diffusion of secreted molecules, which prevents knockout mutants from exploiting the wild type. In our model, we therefore explored the role of both factors.

In brief, we modeled a two-dimensional surface on which cells can grow. The surface consists of a hexagonal grid of 100 × 100 elements, which together are 100 mm in width, approximating the diameter of a Petri dish. Like for our experiments, at the start of growth, we assume resources are homogeneously distributed (Rinit) and cells occur at the surface center (inoculum is 10 mm in diameter). Cells consume resources (R) and divide at a constant rate: following our experiments, wild-type cells divide approximately every 45 min (rwt) and srf mutants have a slight ~3% fitness advantage per cell division (rsrf), as derived from the accumulative fitness costs quantified in the competition experiments using planktonic conditions (see Fitness assays), by correcting for the estimated number of generations within the competition assay. We assume cell division is impossible when resources are depleted (R < 1). We assume that the wild type secretes surfactin (M1), EPS (M2), and BslA (M3) production at a constant rate (d), scaled to the threshold concentration necessary for sliding (τ, see below), while the knockout mutants (ΔsrfAA, Δeps, ΔbslA) fail to produce one of the respective molecules. The secreted molecules degrade at a constant rate (δ) and diffuse in space (α). Since there is no evidence for active enzymatic degradation, we expect the secreted molecules to be stable and assume a half-life of approximately 12 h. As for the diffusion rates, we explored a range of parameter values (see below).

Following our experiments, in the model, colony expansion is only possible when the concentrations of surfactin, EPS, and BslA are sufficiently high. We implement this by allowing cells to expand to neighboring grid elements upon division when the local concentration exceeds a given threshold (M1,M2,M3 ≥ τ, threshold); thereby reflecting the passive form of sliding motility, which is driven by cell division and requires surfactin, EPS and BslA. If the concentration of any of these molecules is too low (min(M1,M2,M3) < τ), cells remain on the same grid element upon division, leading to vertical colony growth. We set the concentration threshold (τ), with respect to the signal production rate (d), such that wild type colonies expand as observed in our experiments. For BslA and EPS production, we also consider potential private benefits that may result from molecule production, which we implement as a reduced concentration at which cells can expand to neighboring grid elements upon division. For example, when wild-type cells have a 20% private benefit from secreting EPS, we assume these cells expand at a 20% lower EPS concentration compared to Δeps cells—thereby having a direct competitive advantage, despite having the same of cell division rate (note that we did not observe differences in cell division rates between wild type, Δeps and ΔbslA mutants).

In exploring parameter space, we specifically focus on the role of EPS and BslA in surface competition. We vary both the diffusion rates of EPS and BslA (10−8, 10−6, 10−4, 10−2 mm2 s−1) as well as their private benefits (0%, 10%, 50%, 90%). For surfactin, the diffusion coefficient (10−2 mm2 s−1) and private benefits (0%) were kept constant, consistent with its function as a common good during sliding motility. For each parameter condition, we simulate seven pairwise competitions: (1) WT-WT, (2) WT-Δeps, (3) WT-ΔbslA, (4) WT-ΔsrfAA, (5) Δeps-ΔbslA, (6) Δeps-Δsrf, (7) ΔbslA-Δsrf. Competition starts with an equal number of cells from each genotype, which can subsequently grow for 24 h, after which we quantified their abundance in the colony as well as the colony size. Second, we assessed the qualitative match between the outcome of competition in our simulations and our experiments: (1) WT-WT: strains should be equally abundant at end of competition; (2) WT-Δsrf: Δsrf mutant wins competition; (3) WT-Δeps: wild type wins competition; (4) WT-ΔbslA: wild type wins competition, but colonies are slightly smaller than those of the wild type; (5) ΔepsbslA: ΔbslA mutant wins competition and produce the smallest observed colonies; (6) Δepssrf: Δsrf mutant wins competition: (7) ΔbslAsrf: Δsrf mutant wins competition. We score the qualitative match between our simulations and experiments from 0 to 7, where 0 means that none of the competition outcomes were correctly predicted and 7 means that they were all correctly predicted. We determined the scores across parameter space using four replicate simulations per condition. For the simulations that best matched our experimental results, we also studied the spatial concentration of surfactin, EPS, and BslA at the end of colony growth. In the supplementary data, we provide the lists of both parameters (Table S3) and variables (Table S4); as well as a pseudocode description of our simulations. The C + + code for running the simulations is available on GitHub: https://github.com/jordivangestel/Bacillus-subtilis-surface-competition.

Results

Extracellular molecules have diverse effects on spatial competition

Sliding motility is driven by cell division, thereby occurring independently from flagella propulsion. During sliding, the colony grows as a flat circular disk that expands outwards over a surface. To assure that cells express sliding motility, without swimming or swarming, we exclusively examined B. subtilis strains that are deficient in flagellin production (Δhag). Thus, whenever we refer to a wild-type or mutant strain, this strain contains a hag deletion.

We started our analysis by examining the wild-type and non-producing mutant strains in isolation, by growing them for 24 h on a semi-solid surface. We consider mutants deficient in EPS, BslA, or surfactin (Δeps, ΔbslA, ΔsrfAA, respectively) production. After a short lag-phase, the wild-type colony slid over the surface at a constant rate, reaching a diameter of about 3 cm after 24 h (Fig. 1; see also [34]). In contrast, and in agreement with previous studies [14], the non-producing mutants were strongly impaired in colony spreading, reaching a diameter of ~1 cm only.

Fig. 1: Competition of sliding good producers and non-producers affects sliding expansion.
figure 1

Diameter of sliding colonies were recorded after incubation on semi-solid medium at 37 °C for 24 h. For competition assays, fluorescently labeled wild type (WT) and EPS, BslA and surfactin non-producers (Δeps, ΔbslA, ΔsrfAA, respectively) were mixed at 1:1 initial ratio (G—green fluorescent strain, R—red fluorescent strain). Black and gray bars indicate WT and mutants, respectively; striped and gridded bars denote mixture of WT and mutant, or two mutants, respectively. The box indicates the 25th–75th percentile; the line in the boxes represents the median. Single data values are represented as dots. An asterisks indicate a significant difference to the WT or WT control competition (unpaired two-sample t test with Welch Correction, p < 0.05, n = 6).

To perform pairwise competition experiments, strains were first mixed in a 1:1 ratio in a liquid culture before inoculating them on the semi-solid agar surface (qualitatively similar results are obtained when mixing 1:10 or 10:1, see Figs. S1 and S2; and Supplementary results). To distinguish strains on the surface, we labeled them with distinct constitutively-expressed fluorescent reporters (either GFP or mKate2) and used stereomicroscopy to monitor their distribution within the colony. Swapping the fluorescent reporters between strains had no effect on the outcome of competition (Fig. S3). We first examined how two wild-type strains, only differing in their fluorescent reporters, distribute themselves during colony growth. Strains strongly segregated in space, giving rise to more-or-less evenly distributed sectors growing out from the colony center, through the passive process of sliding motility (Fig. 2A; see also [34]). Since the wild-type strains are identical, besides their reporter gene, sliding motility was unaffected and the colony reached the same size as that of the wild types in isolation.

Fig. 2: Structure and success of individual strains in competition assay sliding colonies.
figure 2

A Overlay of green and red fluorescent images from representative sliding colonies of competition assay from Fig. 1, with 1:1 initial ratio. Green text indicates a green fluorescent strain; red text indicates a red fluorescent strain. The scale is equal for all images and the scale bar represents 5 mm. B Ratio of occupied area of strain 1 versus strain 2 (in pixel2) of the sliding colonies from (A) obtained by quantitative image analysis using ImageJ. G indicates a green fluorescent strain; R indicates a red fluorescent strain. The box indicates the 25th-75th percentile; the line in the boxes represents the median. Single data values are represented as dots. An asterisks indicate a significant difference to an area ratio of 1 which is marked by a dark gray line (one-sample t test, p < 0.05, n = 5).

Next, we competed the wild type against each of the non-producing strains. When competing the wild type against the EPS or BslA non-producers (Δeps and ΔbslA respectively), colony expansion was indistinguishable from that of the wild type, i.e. after 24 h colonies were approximately 3 cm in diameter (Fig. 1). However, in contrast to the wild type, strains did not form radial sectors (Fig. 2A). The non-producing strains were instead confined to the center of the colony and entirely displaced by the wild type at the colony edge. When comparing the surface occupancy of the wild type with either that of the EPS or BslA non-producer, the wild type was 6 and 2 times more abundant, respectively (Fig. 2B, one-sample t test, pWT/Δeps = 2 ∙ 10−5, pWT/ΔbslA = 0.01, n = 6). This suggests that the EPS and BslA non-producer cells cannot exploit the wild type in competition for space and its associated resources. In contrast, the surfactin non-producer did seem to exploit the wild type: in competition with a surfactin non-producer (ΔsrfAA), the wild type was largely displaced at the colony edge and the surface occupancy was about three times smaller than that of the surfactin non-producer (Fig. 2, one-sample t test, p < 0.05, n = 5). Interestingly, colonies consisting of both wild type and surfactin non-producers were about 1 cm larger in diameter than wild-type colonies (Fig. 1, unpaired two-sample t test with Welch Correction: p = 0.043, n = 5). Since sliding motility is driven by cell division, these results suggest that surfactin non-producers have a significant growth advantage over the wild type and can advance colony expansion by exploiting surfactin production by the wild type.

Finally, we also competed each of the non-producers against each other. In all cases, colony growth was either partly or completely recovered when non-producers were grown together in comparison to their growth in isolation (Fig. 1). This suggests that at least to a certain degree secreted molecules are being shared. Complete recovery of impaired growth was only observed when ΔsrfAA was cultured with either Δeps or ΔbslA, and partial recovery was observed for the co-culture of Δeps and ΔbslA. Like for the wild type, the surfactin non-producer outcompeted both the EPS and BslA non-producers (Fig. 2, one-sample t test, p < 0.05, n = 5), further corroborating the idea that surfactin acts like a common good that can be exploited. When competing Δeps against ΔbslA, the BslA non-producer had a competitive advantage, occupying about twice as much space as the EPS non-producer after 24 h (Fig. 2, one-sample t test, p < 0.05, n = 5). This suggests that BslA is at least partly shared, enabling ΔbslA to exploit Δeps, even though it was unable to exploit the wild type.

Given that sliding is at least partly recovered for all pairwise combinations of non-producers, we conclude that surfactin and BslA can be shared between cells within the colony and EPS potentially as well. Yet, from the three secreted molecules, only surfactin fits the description of a typical common good: ΔsrfAA cannot slide by itself, but in combination with either the wild type or the other non-producers sliding is fully recovered with ΔsrfAA having a competitive advantage. In contrast, both Δeps and ΔbslA do not have a competitive advantage in competition with the wild type. To further disentangle what determines the outcome of competition, we next quantified the costs of molecule synthesis and assessed the degree by which molecules are shared.

Cost of secreted molecule production

Since the benefits of molecule production during colony expansion can offset hidden costs, like reduced rates of cell division, we cannot use our sliding assay to measure such fitness costs directly. Instead, we therefore measured the fitness costs indirectly, by ectopically expressing the genes/operons underlying molecule production in liquid growth conditions, where there are no benefits of molecule production (using the same growth medium as for the sliding experiments). To this end, we first created strains with an IPTG-inducible promoter in front of the gene or operon responsible for production of the respective extracellular molecules (see “Materials and Methods”). Using these inducible gene constructs, we then determined what induction level best mimics molecule production in the wild type, by comparing sliding motility between the inducible strain and wild type for a range of induction levels. This was done separately for each inducible strain, resulting in different optimal induction level for each extracellular molecule (Fig. S4). Finally, we measured the relative fitness of the inducible strain by competing it against the associated non-producing knockout in batch culture, with and without induction (Fig. 3). Strikingly, only the induction of surfactin production (+IPTG) resulted in a reduced fitness, suggesting that in our medium conditions only surfactin production is costly (Fig. 3). These costs lead to a growth disadvantage and explain why ΔsrfAA has a competitive advantage over surfactin-producing strains during sliding. In contrast, and to our surprise, we did not observe any measurable fitness costs for EPS and BslA production, suggesting that these molecules are cost-free (i.e. no effect on the rate of cell division) and cannot be exploited.

Fig. 3: Surfactin is costly to produce under sliding promoting conditions.
figure 3

Relative fitness indicates the fitness of the IPTG-inducible strain (indicated on the x axis) relative to that of the associated knockout mutant, as determined from initial and final cfu counts (see Materials and Methods). When the inducible strain has a selective advantage, the relative fitness is >1, and if it has a selective disadvantage, it is <1. The competition was conducted both with (+IPTG) and without IPTG (-IPTG). IPTG results in the ectopic expression of either EPS, BslA, or surfactin. The box indicates the 25th-75th percentile; the line in the boxes represents the median. The dashed line indicates a relative fitness of 1. The asterisk indicates a significant difference to 1 (one-sample t test, test mean = 1, p = 6.9 ∙ 10−4, n = 6).

Complementation by the secreted molecules

Since colony expansion was (partly) recovered when mixing different non-producing mutants (Figs. 1 and 2), we know that secreted molecules are at least partially shared. Next, we investigated this directly, by exogenously adding EPS, BslA, or surfactin to mutant colonies (Δeps, ΔbslA, and ΔsrfAA respectively), and examining if and to what degree colony expansion was recovered. We isolated EPS, BslA and surfactin from different sources. For EPS, we extracted EPS-containing matrix from either wild-type biofilms or ΔtasA biofilms; these latter biofilms lack the main protein component in the biofilm matrix (i.e. TasA). As a control, we also collected matrix from Δeps biofilms, which lack EPS. BslA was isolated from the supernatant of a E. coli BL21 strain that was engineered to overproduce BslA. The quality and functionality of our extraction methods for both EPS and BslA were validated experimentally (see Fig. S5, S6). For surfactin, we simply used commercially available purified surfactin (Sigma Aldrich; CAS# 24730-31-2).

We supplemented colonies of non-producing mutants with different concentrations of external EPS, BslA, and surfactin and monitored their growth for 24 h. Δeps colonies did not show any recovery of colony expansion when adding EPS, while ΔbslA colonies showed a slight increase in size when adding BslA (Fig. 4A, p > 0.05 and Fig. 4B, pmutant = 1.2 ∙ 10−6, pBL21 = 9.4 ∙ 10−7, pWT = 7.6 ∙ 10−8, respectively; all unpaired two-sample t-test with Welch Correction, n = 6). However, this effect was strongly dependent on the BslA concentration: when adding 10-fold diluted BslA-containing supernatant no increase in colony size was observed (unpaired two-sample t-test with Welch Correction, pBslA1:10 = 0.21, n = 6). In contrast, we did observe full recovery of colony expansion when adding surfactin to ΔsrfAA colonies; at high surfactin concentrations colony expansion of ΔsrfAA colonies even exceeded that of the wild type (Fig. 4 C, p1 mg/ml = 0.049, p10 mg/ml = 6 ∙ 10−5, pWT-10 mg/ml = 6 ∙ 10−5; all unpaired two-sample t-test with Welch Correction, n = 6).

Fig. 4: Complementation of non-producer sliding with supplied molecules.
figure 4

Diameter of EPS (A), BslA (B) and surfactin (C) non-producers (Δeps, ΔbslA, ΔsrfAA respectively) incubated under sliding promoting conditions with supplied molecules or a control substance and wild type controls for 24 h. For EPS non-producer complementation (A), EPS isolated from wild type, tasA mutant, or eps mutant biofilms was used. For BslA non-producer complementation (B), BslA containing supernatant from an E. coli strain harboring a BslA production plasmid was used (+BslA); supernatant from an E. coli strain without the plasmid served as control (+BL21). For surfactin non-producer complementation (C), commercially available surfactin was used in different concentrations (0.01, 0.1, 1, and 10 mg/ml) dissolved in methanol, which was used as control (+MeOH). The box indicates the 25th-75th percentile; the line in the boxes represents the median. Single data values are represented as dots. Asterisks indicate significant differences.

The efficiency by which ΔsrfAA colonies can be complemented by exogenously adding molecules, while the Δeps and ΔbslA colonies cannot or only partly, is consistent with our competition experiments (Fig. 2) where surfactin acts like a common good, while EPS and BslA do not. Our complementation experiment also corroborates that BslA is at least partly shared, since ΔbslA colony expansion is partly recovered when exogenously adding BslA (Fig. 4), similar to the recovery of colony expansion by co-culturing ΔbslA and Δeps (Fig. 2). Despite the fact BslA can be shared, in competition with the wild type, both ΔbslA and Δeps mutants show a strong competitive disadvantage. Based on Fig. 4, we speculate that there could be two potential causes for this competitive disadvantage: (1) BslA and EPS might show limited diffusion, which prevents ΔbslA and Δeps mutants from effectively exploiting the wild type, and/or (2) wild type cells might obtain direct private benefits from molecule production, which are inaccessible to ΔbslA and Δeps mutants, even when adding BslA or EPS exogenously.

Agent-based modeling recapitulates the observed complementation dynamic

Since we cannot examine the diffusion rate and private benefits of EPS and BslA production experimentally, we decided to examine the potential impact of these factors on the outcome of competition using an agent-based model. Following our experiments, we model surface competition during sliding motility, where colony expansion is driven by cell division (see “Materials and Methods”). The colony can only expand when the concentrations of surfactin, EPS and BslA are sufficiently high. For simplicity, we assume that wild-type cells secrete all three molecules at a constant rate. Cells furthermore incur a small fitness penalty (~3% lower division rate) for producing surfactin, as derived from the accumulative fitness costs shown in Fig. 3 (see also “Materials and Methods”). Molecules also degrade at a slow rate and diffuse in space. The molecule concentration determines whether the colony can expand (see “Materials and Methods”). Finally, we also assume that EPS and BslA producers can convey private benefits to producers, by increasing their propensity to expand. We model this by lowering the EPS and BslA concentrations at which EPS and BslA producers expand outwards. In our model, we manipulate both the degree by which molecules are being shared, i.e. diffusion rates (10−8, 10−6, 10−4, or 10−2 mm2 s−1), and their private benefits (0%, 10%, 50%, or 90%), and determine how these parameters affect the outcome of surface competition. For each parameter setting, we performed all pairwise competition experiments (Fig. 2): (1) WT vs. WT, (2) WT vs. Δeps, (3) WT vs. ΔbslA, (4) WT vs. ΔsrfAA, (5) Δeps vs. ΔbslA, (6) Δeps vs. Δsrf, (7) ΔbslA vs. Δsrf. We then scored the accuracy of our model predictions, both in terms of colony sizes and fitness values (see Materials and Methods): 0 means that none of the pairwise competition experiments were correctly predicted by our model; 7 means that all of them were correctly predicted.

Although diffusion affects competition, private benefits of EPS and BslA production were most important in predicting the outcome of competition (Fig. 5). The best match between our simulation and experimental data (see asterisk in Fig. 5) occurs for relatively high private benefits of EPS production (~90%) and relatively low but nonzero private benefits for BslA production (~10%). Similarly in agreement with our data, we find considerable variation between replicate simulations (e.g. Figure 1), which likely result from so-called founder effects that are commonly observed during (bacterial) surface expansion [39, 40]. Figure 6 shows the simulation results that most accurately match our experimental data (asterisk in Fig. 5). In accordance with our experimental findings, we find that the ΔbslA and Δeps mutants can partly complement each other, leading to recovery of colony expansion, where BslA is partly shared between cells. At the same time, the wild type readily outcompetes both ΔbslA and Δeps mutants, because wild-type cells profit from the private benefits of molecule production and cannot be exploited by either mutant (there are no fitness costs of BslA and EPS production and there is limited diffusion). The distribution of strains and the colony sizes after 24 h of colony expansion closely match our experimental findings (Fig. 2).

Fig. 5: Effects of private benefits and diffusion rates of EPS and BslA on outcome of competition.
figure 5

Smallest blocks show four attached squares (2 × 2), which represent replicate simulations; intermediate block show 4 × 4 smaller blocks that differ in diffusion rates; largest blocks contain 4 × 4 intermediate blocks that differ in private benefits. Asterisk shows what parameter conditions lead to model results that best recapitulate those of the experiments. For model description and parameters see “Materials and Methods” and Supplemental material.

Fig. 6: Modeling results of sliding motility as derived from agent-based simulations.
figure 6

Upper panel, outcome of pairwise competition between wild type (WT), surfactin non-producer (ΔsrfAA), EPS non-producer (Δeps) and BslA non-producer (ΔbslA). Lower panels, concentration of EPS, BslA and surfactin, normalized between the 0 (black) and highest (white) simulated concentrations (see also gradient bar in figure legend).

Discussion

Surface competition plays a critical role in the ecology of bacteria. Accordingly, bacteria often secrete molecules that facilitate surface spreading, enabling them to monopolize space in competition for resources with other bacteria that co-occur on a surface. B. subtilis secretes EPS, BslA, and surfactin to facilitate sliding motility, a rapid mode of surface spreading that is driven by cell division. Here, we reveal the complex extracellular biology of sliding motility, where EPS, BslA, and surfactin have a strongly different impact on surface competition. Although each of these molecules is secreted by cells and essential for sliding motility, only surfactin is costly to produce and diffuses broadly through the colony—thereby acting like a common good. The other two types of molecules are only partly shared, cost-free and likely provide direct benefits to the producer. These distinct functional consequences of molecule production impact both bacterial ecology and evolution.

Since surfactin acts like a common good, the wild type is outcompeted by ΔsrfAA during surface competition. We also observed that chimeric WT-ΔsrfAA colonies show more efficient sliding motility than wild-type colonies, which likely results from the fact that ΔsrfAA mutants divide quicker than the wild type and thereby accelerate colony expansion. Thus, the presence of cells that do not produce surfactin can benefit overall colony expansion. Similarly, MacLean et al. [41]. previously demonstrated that, in structured yeast populations, productivity during growth on sucrose was maximized when mixing cells that produce an enzymatic common good (invertase) with those that do not. Interesting, even in wild-type B. subtilis colonies, not all cells produce surfactin – due to heterogeneous gene expression – suggesting that even these colonies partly mitigate the costs of surfactin production through a phenotypic division of labor [8, 26, 42]. A genetic division of labor between wild-type and ΔsrfAA cells in chimeric colonies lies in extension of this [24]. It might therefore not be surprising that surfactin mutants are frequently isolated from soil samples that also contain surfactin producers [43]. It is however yet unclear how long wild type and surfactin mutants can stably co-existence in time (i.e. without displacement of the wildtype).

If surfactin production can so easily be exploited, what could prevent wild-type cells from being outcompeted by surfactin mutants in nature? Here, transcriptional regulation of the surfactin genes probably plays a crucial role: their expression is controlled through quorum sensing, using the strain-specific pherotype (i.e. distinct communication group) ComX [44, 45]. Quorum-sensing regulation has two possible effects. First, it prevents wasteful production of surfactin at low cell densities – when there is no benefit of producing surfactin. Second, surfactin could specifically be expressed when cells are surrounded by kin, similar or the same strain, which reduces the risk of exploitation [46, 47]. Finally, other ecological factors could play a role in preventing surfactin producers from being outcompeted. For example, besides facilitating sliding motility, surfactin also acts like an antimicrobial that kills other bacterial species [43, 48]. Surfactin producers might therefore gain an ecological advantage in competition with other species [48].

Although both EPS and BslA are secreted [15, 16, 27, 37], both the experiments and model suggest that these molecules are not broadly shared and provide direct private benefits to the producers. Consistently, a previous study of van Gestel and colleagues [6] showed that eps knockout mutants are excluded from the EPS-producing filamentous bundles at the colony edge during sliding motility. In contrast, under biofilm or pellicle growth conditions, EPS was previously shown to act as a common good that is broadly shared [22, 24, 26, 49, 50]. These different functional implications of EPS production in sliding motility and complex colony formation (e.g. biofilm and pellicle formation) show that one should be cautious in qualifying molecules as generic common goods: molecules can act as classical common goods in one condition and as private goods in another. Thus, rather than determining whether molecules act as common goods, a future challenge lies in understanding what factors determine the costs, shareability and private benefits of molecules in different ecological settings.

There are several factors that we did not study that could further influence the outcome of surface competition. For example, secreted molecules might alter gene expression and therefore lead to indirect phenotypic changes that impact competitive fitness. Zhang and colleagues demonstrated that matrix producing cells have a stronger response to quorum sensing signals, thereby altering their gene expression that could lead to indirect phenotypic benefits [51]. In our current model, these are lumped together under private benefits. Moreover, cells could directly influence the diffusion of secreted molecules and manipulate their private benefits. For example, siderophores secreted by Pseudomonas aeruginosa were shown to bind to amyloid fibrils, thereby limiting their diffusion and hence exploitation by non-producers [52]. By the same token, Psl exopolysaccharide secreted by P. aeruginosa anchor to the cell in early stages of biofilm formation [53]. Similar retention mechanisms might also play a role in the privatization of EPS during sliding motility.

In many ways, we are only at the beginning of exploring the extracellular biology of bacteria. Advances in single-cell microscopy, transcriptomics and molecule tracking will likely continue to push our understanding of surface-bound microbial communities [21, 54, 55]. In addition, we believe important challenges lie in developing laboratory settings that most accurately reflect the complex ecologies to which bacteria are exposed in nature [56].