Initiation of swarming in Bacillus subtilis as a resolution for the sorites paradox in bacteria

Significance

The sorites paradox shows how tricky it is to define when something changes—like a heap of sand losing grains one by one. Biology has a similar puzzle: When does a single bacterial cell become a swarm? We followed individual Bacillus subtilis cells on agar, identifying four clear stages toward swarming and using mutants to see which cellular processes matter most. This matters because swarming affects how bacteria survive, spread, and adapt. Understanding how single cells coordinate into powerful, collective groups reveals key principles of life and could help us manage harmful bacteria before they gain the full strength of a swarm.

Abstract

As a single cell grows and multiplies, at what stage does it constitute a swarm? Here, we provide an answer for Bacillus subtilis. We place single planktonic cells onto agar and monitor the biological and physical phases as they transition into an active, swarming cohort. We identify four distinct, nonoverlapping stages enroute to swarming. Using different mutants, we pinpoint the roles of specific cellular processes in the buildup of collective motion. Our results provide a detailed description of the transition from planktonic to a swarming lifestyle—a hallmark of bacterial adaptability to changing environments.

How many grains of sand constitute a heap? This is the famous sorites paradox—a single grain is not a heap, two grains are not a heap, and adding just one grain cannot change a nonheap into a heap. This leads to an infinite regression and a paradox that heaps do not exist. A biological analogue can be described in bacterial colonies—a single Bacillus subtilis cell, placed on agar, is not a colony. About 30 min later, it will divide to become two cells, which are still not a colony. Still, under appropriate conditions and within several hours, it is a swarming colony, vigorously moving and expanding.

For bacteria, such questions are not a matter of philosophical or linguistic sophistry, but fundamental to understanding the adaptive strategies that enable their persistence across diverse environments. Bacteria are known for their capacity to adopt distinct organizational states, commonly referred to as lifestyles. Well-characterized examples include the planktonic state, where individual cells move and survive independently; biofilms, in which multicellular aggregates adhere to each other and to surfaces; and swarming, a collective behavior in which thousands of cells migrate coordinately across surfaces, generating dynamic patterns of whirls and jets. Transitions between these states—where entire populations undergo coordinated shifts in both single-cell properties and colony-wide organization—represent complex and fascinating processes that enable adaptation to environmental cues and stresses. In this study, we focus on the relatively understudied transition from solitary planktonic cells to the collective swarming lifestyle.

Bacterial swarming is a complex biophysical phenomenon found among a variety of rod-shaped species, rendering this phenomenon a general and broadly relevant biological trait (1–8). The cells are powered by multiple flagella, secrete osmotic agents to extract or pump liquids from the substrates on which they grow and surfactants to reduce the surface tension of the liquid in which they move (9–12). Swarming cells differ from planktonic ones; it is known that, prior to swarming, they change their aspect ratio (2, 5), increase the number of flagella (13, 14), and in general show significant changes in gene expression patterns (2, 5, 13). Such properties, which are regulated by complex biological intracellular gene expression networks, have physical implications. Coherent cellular motion is a hallmark of bacterial swarming and is thought to be resulted by the elongated cell-shaped structure that induces cell–cell steric interactions and orientational alignment between individuals (8, 15–18). The local amount of liquid around a specific cell is correlated with the state of the flagellar bundle and determines whether it can move or not (19). In such fluidic environments, self-propulsion produces hydrodynamic interactions such as alignment and velocity synchronization that additionally control the collective migration (15, 20). Several reviews summarized the current understanding of the bacterial swarming phenomenon (1–2, 8), and differences compared to collective swimming of bacteria in liquids, both from the biological and physical perspectives (21).

Most of the work on swarming bacteria has been done on the active swarm regimes, namely when the cells have already transitioned into the swarming stage and move vigorously [e.g., (13, 22)]. One of the main observations is that the dynamics, even in liquid niches, depends critically on cell density [e.g., (22–24)], so that fluctuations in bacterial surface coverage may alter the statistics of collective motion instantaneously, leading to a complex cell-clusters size distribution (25), also known as giant number fluctuations (25–29). Recent studies focused on variations from the wild-type (WT) strain, including variations in cell size (23, 30–31) and flagellar motility (32, 33), shedding light on the principle physical interactions underlying the dynamics of swarming, and how they depend on the physical properties of cells. The general goal was to obtain a phase diagram of the possible swarming regimes cells may obtain as a function of controlled parameters, such as cell shape, cell density, surfactant secretions, and motility (13, 16, 18, 22, 34). For example, it was found that the ~5:1 cell aspect ratio, commonly observed in nature, is optimal for swarming (31). The latter is particularly striking given the fact that a variety of bacterial species change their cell aspect ratio before starting to swarm, suggesting a possible selection mechanism in the evolution of the swarming phenomenon in bacteria.

A general theory capturing all phenomena related to bacterial swarming is still lacking. In particular, the coupling between the physical properties of swarming and its interaction with the underlying cellular process are poorly understood. This coupling is particularly manifested during the early stages of swarming initiation, in which bacterial reproduction has a pivotal role. Mukherjee et al. (35) indicated that cells that are artificially hyperactivated for flagellar number biosynthesis swarmed without a lag, suggesting that the transition to hyperflagellation was a limiting stage. Jeckel et al. (13) studied the dynamics of swarming B. subtilis and created a space-time phase diagram of an expanding colony. Among other things, they found an increase in the expression of surfactant and flagellar biosynthesis genes before expansion begins, and that the lag time—from inoculation to active swarming—depends on the initial number of cells in the inoculum. Beyond this, the very early stages of swarming initiated from single-cell inoculations have not been systematically studied experimentally, raising the question of “how immobile single cells aggregate and start moving?”

The main goal of the current paper is to quantitatively characterize and describe the initiation of the swarming phenomenon. We focus on a single planktonic B. subtilis cell, placed on agar, and monitor the biological and physical phases as it transitions into an active, swarming cohort. Different mutants are used to pinpoint the roles of specific cellular processes in the buildup of collective motion and swarming.

Results

The Planktonic to Swarm Transition Consists of Four Distinct Stages.

Depositing a highly diluted liquid culture (log-phase, i.e., planktonic cells) containing approximately 1,000 WT B. subtilis cells on an agar surface (0.5%), resulted in isolated individuals. Each isolated cell was allowed to grow independently into a small swarming colony. We begin by focusing on three key statistics, previously shown to be related to the buildup of swarming colonies: reproduction rate, surfactant secretion, and movement.

The average reproduction rate constitutes the most basic cellular characteristics as it depends on fundamental intracellular functions. In order to estimate it, we count the number of cells in a colony over time (Methods). Fig. 1 (red data) shows the mean number of cells in a colony as a function of time (log-scale). We identify a sharp transition in growth rate around 5 h postinoculation, indicating a fundamental change in doubling time from 38 min (same as in liquid culture) before the transition to 22 min after. We mark the time of this transition in growth-rate as t = 0. At this time, the number of cells in the colony is approximately 3,500.

Fig. 1.

Four stages of the planktonic to swarm transition. A single log-phase (i.e., planktonic) WT Bacillus subtilis cell deposited on an agar surface grows into a small swarming colony. Red symbols (circles): the number of cells, plotted on a logarithmic scale, as a function of time. The reproduction rate, which is the slope of the curve, jumps sharply from a doubling time of 38 min to 22 min. The transition point is marked t = 0. Green symbols (diamonds): surfactant secretion rate, observed using a fluorescent marker. Surfactant secretion starts abruptly around t = −30 min (~a single generation) and saturates at t = 0. Blue symbols (squares): The average speed increases from zero at t = 0 and saturates to the steady state swarming speed around t = 25 min (~a single generation). Overall, we identify four distinct stages: i) Proliferation, ii) Surfactant Secretion, iii) Motility Initiation, and iv) Expansion (active stage).

The same transition mark is evident in two other measures. A slightly earlier event, starting at t-30 min, is observed in surfactant secretion (green data). This measurement is based on a fluorescently labeled mutant that glows upon surfactin gene expression (AE1336); expression reaches saturation at t = 0. In addition to the quantitative measurement shown in Fig. 1, surfactant has a clear visual manifestation in the three-dimensional structure of the colonies. Initially, the height of the colony increases, but around t = 0, the colony flattens out presumably due to the effect of surfactin (SI Appendix, Fig. S1A). Surfactin was previously shown to increase oxygen solubility, suggesting a plausible explanation for the increased reproduction rate (36). The third measurement pinpointing the planktonic to swarm transition involves movement; Fig. 1 (blue data). Measuring speeds using optical flow methods (8, 21), we find that cells begin to move at the transition (t = 0). Note these are the coarse-grained and not individual cell speeds. Speed increases gradually, reaching a maximum approximately 22 min later, corresponding to a single generation. All three measurements are summarized in Fig. 1.

Overall, we identify four distinct stages in swarm development (Fig. 1 and SI Appendix, Fig. S1A): i) Proliferation—This first stage lasts about 4.5 h, during which cells reproduce slowly (doubling time = 38 min) and do not move. ii) Surfactant synthesis—This second stage lasts approximately 30 min; cells maintain the same slow reproduction rate while beginning to secrete surfactants but still do not move. At the end of this stage, the colony reaches its maximal height (SI Appendix, Fig. S1A). iii) Motility Initiation—In this ~22-min stage, cells both secrete surfactant and increase their reproduction rate (doubling time = 22 min = 1 generation). The swirling motion characteristic of swarming develops gradually. iv) Colony Expansion (or the active stage)—Cells reproduce rapidly, secrete surfactant, and move at high speeds, resulting in colony expansion.

The division into four growth stages is the main result of this paper. In order to further characterize the biological and physical processes underlying each stage, as well as the transitions between them, we subsequently turn to describe specific properties of cells in the colony.

Quantifying the Surfactant Initiation Stage.

To characterize the role of surfactants, experiments were repeated with a surfactin-defective mutant (Δsrf strains DB1043-R_AE and DB1043-G_AE). Compared to the vigorous motion of the WT, Δsrf exhibits only two stages: proliferation, and a semiactive regime we term “poor swarming,” and does not change its reproduction rate during the experiment. SI Appendix, Fig. S1B shows that the average height increases monotonically and never collapses. In the latter phase, motion does not appear collective as many of the cells do not move at all, and the average speed is only 10% of the one obtained for an active WT swarm. In addition, the number of cells required to obtain poor swarming is significantly larger (~×7) than the one required to obtain active swarming in the WT. The inset of SI Appendix, Fig. S1B shows a visually observed difference between the WT (Inset of SI Appendix, Fig. S1A) and the Δsrf mutant (strain DB1043-R_AE). The camera settings, the light exposure, and the fluorescence brightness of single Δsrf and WT cells are identical. Hence, the significantly enhanced brightness of the surfactant deficient colony indicates that the colony forms thicker structures. The growth rate of the Δsrf strain is the same as in the WT proliferation stage.

Previous reports indicated that a surfactant deficient mutant could be extracellularly complemented by the presence of a surfactin proficient strain. To explore this observation at the level of individual cells, we initiated a colony from two single cells, one WT and one Δsrf, which were inoculated simultaneously and jointly at practically zero distance (WT in red and Δsrf in green strain DB1043-G_AE). An example is shown in Fig. 2 A and B (the two strains have similar growth rates so they maintain a nearly 1:1 ratio of populations). We find that the swarming phase in the mixed colony starts when the total number of cells is roughly twice the number required when WT is present alone. We infer this means that the amount of surfactin needed to initiate swarming is constant but only being synthesized by half the population in the mixture. Later, when the colony has transitioned to the swarming stage, the mixed colony exhibits a similar behavior to the one obtained for the WT-only. During swarming, the fraction of WT cells in the mixed colony is still approximately a half, and we find that the strains are well mixed (Fig. 2B) and both strains move and participate in swarming. To quantify these observations, we studied colonies initiated from a few mixed cells deposited randomly in the same proximity. The results show that initiation of swarming depends only on the number of WT cells in the mixed colony. In the active stage, on the other hand, as long as the fraction of WT cells in the mixed colony is larger than ~10%, the entire colony behaves similarly to a WT colony. In Fig. 2C we see, for the active stage, a practically vanishing spatial cross-correlation between the two strains for a variety of mixing ratios (except for below ~10%), and similar speeds for the two strains (Fig. 2 D–F). Importantly, as observed in previous experiments with both swimming and swarming bacteria (8, 21), speeds increase with density, which is the hallmark of collectivity. We conclude that a critical threshold of surfactin is needed to initiate swarming and that the necessary amount of surfactin is maintained afterward.

Fig. 2.

Initiation of swarming in a mixed colony. (A) An example of a young colony composed of a 1:1 mix of WT (red) and Δsrf (green); at early stages the mixing is not homogenous. (B) Zoom-in at later stages of a similar mixed colony shows a homogenous distribution of the cells. (C) The spatial cross correlation between the strains is practically zero, indicating a homogenous spatial inhabitation. (D–F) High correlation between the speeds of the two strains in the mixed colonies, for a variety of surface coverages and a variety of ratios between the strains.

We have also examined the distribution of surfactant producing cells among the WT in a swarm population (based on single-cell fluorescence intensity of the strain labeled for surfactant production) and found that it is not uniform (SI Appendix, Fig. S2A—strain AE1336). In other words, some cells may contribute more surfactin than others. Our results indicate that, while surfactant is needed for both initiation of swarming and for the active stage, it is not mandatory that all cells secrete it.

Surfactant mutants were previously shown to swarm when purified surfactin was first spotted on a swarming plate and a high density of cells was inoculated within the surfactin zone (2). Adding pure surfactin (Sigma) to the surface prior to inoculation of a single cell of a Δsrf mutant however, flattened the growing microcolony but did not restore the microscopic swarming pattern until the cell number exceeded ~200,000 individuals, about 100-fold higher than required for WT swarming (surfactin was diluted up to 10⁶-fold). The Δsrf mutant microcolonies exhibited a similar initiation delay in the presence of spent supernatant containing surfactin taken from WT cells grown to high cell density in liquid. We conclude that adding surfactin exogenously is physiologically different than endogenous biosynthesis and that endogenous biosynthesis promotes earlier swarming initiation by allowing an initial buildup of a dense community that will support swarming.

Overall, our results indicate that surfactin is an essential factor for swarm initiation, without which cells are unable to move or spread. Surfactin secretion starts abruptly when the colony reaches a given size (in terms of cell number) and saturates within a single generation.

Role of ComQ.

Surfactin production is regulated by the Com quorum-sensing system in response to cell density. Each cell produces a peptide called ComX that is secreted and enzymatically processed by ComQ. The amount of processed ComX is thought to accumulate in proportion with cell density and function as a quorum signal. ComX levels are detected by a transmembrane histidine kinase called ComP, which phosphorylates the DNA binding response regulator ComA. Phosphorylated ComA binds to, and activates expression from, the promoter of the srfA operon. Further details can be found in (37).

To explore the role of quorum sensing during early events in swarming, a comQ mutant strain was tested [strain AE8661 (38)]. With the lack of comQ, surfactant production exists, but was dramatically reduced (SI Appendix, Figs. S2 B and C). As a result, we have found that the colony expands very slowly. The initiation is delayed until the number of cells is ~10 times larger compared to the WT. In the initiating experiments, starting the colony from a single cell, addition of pure surfactants from an external source (Sigma) to the developing ΔcomQ colony did not restore the behavior seen for the WT. However, additions of pure surfactants to a mature colony resulted in cells that swarm similar to the WT. Similar results were obtained if supernatant of WT cells is added instead.

We also tested a Com+ mutant (SI Appendix, Fig. S2D), which overexpresses ComP [strain AE8665 (38)]. This strain secrets surfactants at very early stages, starting when the colony is composed of a few tens of cells (SI Appendix, Fig. S2G). However, the colony does not transition to the swarming stage earlier than the WT, plausibly because the enormous amount of surfactant causes them to disseminate; SI Appendix, Fig. S2H shows the same field of view seen in SI Appendix, Fig. S2G after 2 min.

The latter results suggest that the Com system is required for a colony to initiate swarming properly and that alternatives (additions of surfactant or supernatant) restore swarming only when the colony is very large.

Motility Initiation Stage.

Next, we turn to the third stage, during which the entire colony transitions from an immotile group of cells to a collective swarm. To quantify the buildup of movement, we obtain the velocity field using optical flow analysis (Methods). Within approximately 22 min, corresponding to a single generation, the number of cells is doubled, the colony flattens to an average of 1.5 layers (SI Appendix, Fig. S1A), and reaches the active stage. Fig. 3 shows statistics for a ~20 min duration of this dynamic transition from nonmoving cells to an active swarm: the average speed increases, as well as the spatial correlation length. The final correlation length, about 5.5 µm, corresponds to the active-turbulent regime, is independent of the colony size. The kurtosis (centered, scaled 4th moment), decreases from about 12 to 3. A value of 3 indicates that the velocities of all cells are drawn from a Gaussian distribution. Large kurtosis indicates non-Gaussian tails and large variability in velocities because some cells are stationary while others move. At the end of this stage, the entire colony is moving, and no cells are observed to be stationary or stuck. Kindly see Movie S1 for more details.

Fig. 3.

Motility initiation. (A–C) Optical flow analysis of the motion. Within ~20 min, the number of cells doubles, while cells start moving collectively in coherent flows. Colors indicate the local speed (bar). (D–F) Statistical analysis of the transition stage. (D) The average cell speed as a function of colony area and time (nonlinear upper x-axis) shows a relatively smooth increment. (E) The spatial correlation increases threefold within a short time, then saturates, indicating the end of the transition into a collective state with a well-defined mesoscopic length scale. (F) The kurtosis (centered and scaled fourth-moment) in the velocity distribution decreases from 12 (heavy-tailed) to three (Gaussian), indicating uniform cellular motion with normal fluctuations.

Role of Flagella.

An increase in flagellar density on the cell surface has been reported to be essential to initiate swarming. To explore flagellar gene expression during initiation, we monitored the growth of microcolonies expressing yellow fluorescent protein (YFP) expressed under the control of the promoter of the flagellin gene hag (P_hag-yfp; strain AES1035) (Fig. 4A). Fluorescence was detected at the very early stages and preceded active motility (Fig. 4B). Moreover, flagellin expression of each cell increased linearly with colonial area until it reached a maximal value (Fig. 4C). We conclude that the flagellin gene expression increases prior to the initiation of swarming, consistent with the increase in flagellar basal body hook-number previously shown to be necessary to swarm (35).

Fig. 4.

Motility-related mutants. (A and B) Phase-contrast and fluorescence microscopy images of WT cells, with modification of a hag yfp-expression, showing that the cells express flagellin very early in the proliferation stage. (C) Expression of the hag protein in individual cells (averaged over ~250 cells; measured cells are not piled to avoid fake reading) obtained from fluorescently labeled mutants of the WT. The average expression (and error bars) is overlaid on the stages (SI Appendix, Fig. S1A), showing the gradual increase of the number of flagella up to some plateau as a function of the colony size and/or time. (D) Expression of the hag protein in individual cells (averaged over ~250 cells) obtained from fluorescently labeled mutants of Δsrf. Expression decreases with colony size and/or time. When the colony transitions to the poor-swarming stage, the expression increases again. (E) Expression of the hag protein in individual cells (averaged over ~250 cells) obtained from fluorescently labeled mutants of the Δsrf that are mixed with the WT (nonlabeled). The same gradual increase (as in the axenic WT), of the flagellin expression is obtained. (F) An active large WT swarming colony is expanding toward a much smaller WT colony that is still in its proliferation stage. Once coalesced, the smaller colony does not mix immediately with the larger one. (G and H). Young colonies grown from individual WT and IM cells, randomly deposited in the same proximity. The WT cells are labeled red and the IM are not labeled; (G) Phase-contrast image, showing both strains (without distinguishing between them). (H) Red fluorescence channel, showing only the WT. The strains do not mix as the IM strain does not migrate.

Flagellin gene expression was also monitored in microcolonies of the swarming deficient Δsrf mutant; Fig. 4D (strain AES10179). In contrast to surfactant producing cells (strain AES1035), expression of YFP per cell in the surfactin mutant decreased with time and colony size and increased later only when the “poor swarming” regime was initiated. In contrast, a mixed colony composed of 10% Δsrf cells harboring the Phag-yfp reporter (AES10179), with 90% of WT cells (3,610; unlabeled), showed that flagellin expression in the Δsrf individuals increased with colony size in this mixed colony (Fig. 4E). We conclude that the deficiency in flagellin gene expression was due to the absence of surfactin, either responding to the chemical itself or a physical parameter mediated by surfactant activity.

Unlike the case for surfactin, previous reports showed that WT cells were unable to restore swarming to mutants unable to increase flagellar number. Here, we show that an active large WT swarming colony expands toward a much smaller WT colony that is still in the earlier proliferation stage. After the colonies coalesce, the smaller colony does not mix immediately with the larger one; it takes the smaller colony at least 10 more minutes to be evenly mixed within the larger one. The latter indicates that even when touching an active colony, the motility of the cells in the smaller colony is not sufficient to allow them to swarm, suggesting that WT cells can exclude other WT until those individuals have achieved the requisite flagellar density.

To explore the role of flagellar number in swarming, swarming initiation was studied in cells lacking SwrA. SwrA is part of the master activator of the 27 kb fla/che operon encoding the proteins required for basal body assembly. Cells lacking SwrA (strain DS215) have an insufficient number of flagella to swarm and here we find the cells do not initiate swarming such that colonial expansion is only due to mass increment. Conversely, cells that overexpress SwrA (strain DS526) are reported to hyperproduce flagella, and while the cells appeared to produce surfactin similar to WT in our assay, it transitions to the active stage much later, when the number of cells in the colony was approximately 5 times larger. A similar delay in transition was observed in an soa mutant (strain DS8885) that improves the promoter of fla/che operon expression in the absence of SwrA. We conclude that swarming initiation requires SwrA to increase flagellar number over time, but artificially high levels can be transiently inhibitory, at least at the level of microcolonies. Flagellar function is also important as a swrD mutant (strain DB183), defective in achieving maximal flagellar torque, shows no swarming even at the level of individuals in a microcolony.

To further understand the role of the flagella, we mixed the WT with an immotile strain (IM) that does not produce flagella at all (strain DS1677). Tiny colonies were initiated from two single cells, pairs of one WT and one immotile cell, which were inoculated at practically zero distance from one another. The results (Fig. 4 G and H) clearly show that the immotile strain does not mix with the WT; initiating a mixed colony of the two strains shows that the IM does not count toward the quorum needed for the initiation of swarming.

Varying the Agar Concentrations.

How unique is the division into four growth stages, and does one obtain the same number of 3,500 cells to initiate the processes? To answer these questions, swarm initiation was studied in different agar concentrations by increasing the agar concentration up to 1.0%. At agar concentrations below 0.5%, cells can swim through the porous substrate, while concentrations above 1.0% can induce biofilm formation or sporulation at early stages. A significant observation during swarm initiation was the impact of agar concentrations on colony height. Colony height decreased as a function of increasing agar concentration, a trend observed in both WT and Δsrf mutants. This reduction in colony height with higher agar concentrations can be attributed to the reduced availability of water, due to the increased agar content. Also, when studied as a function of time, increasing the agar concentration delayed the stages, where the proliferation rate was slightly reduced prior to transition, and the final speed was smaller as expected. However, despite the above differences, we have found that the number of stages as well as their timing, and the number of cells required to initiate the swarm (~3,500), all remained relatively constant across all agar concentrations (SI Appendix, Fig. S3). We suggest that the availability of water from the agar does not influence the quorum needed to initiate swarming. We also note that the number obtained using the direct counting method (3,500 in Fig. 1 red data) and the one obtained using the fluorescent intensity (~3,200 in SI Appendix, Fig. S1) are within the error bars (see similarities in all agar concentrations in SI Appendix, Figs. S3 C–E).

Methods

B. subtilis is the model organism for this study, all strains used are derivatives of the B. subtilis wild-type (WT) 3610. Strains were cultured in Luria-Bertani (LB) broth (25 g/L), preserved at −80 °C with 50% glycerol and revived on 2% hard agar plates (with LB). SI Appendix, Table S1 shows the strain details and characteristics.

Swarm Assay.

Overnight cultures were grown in 15 mL tubes containing 2 mL LB at 30 °C with 200 rpm shaking. Prior to the swarm experiment, 50 µL of overnight culture was recultured into 2 mL of fresh LB for 4 h to ensure logarithmic growth phase. Swarm plates are supplemented with 0.5% agar and 25 g/L of LB, aged 18 h at 22 °C and 50% relative humidity (RH). Prior to inoculation, the plates are dried for 8 min in a laminar flow hood.

For classic swarm assay, a 4 µL drop of culture is inoculated onto the center of the swarm plate, dried for 40 min at room temperature, then incubated at 30 °C and 95% RH until the swarm colony reaches 5 cm in diameter. For initiation experiments, fresh cultures are further serially diluted by a factor of 10⁴. A 20 µL drop was placed at the center of the agar plate, tilted 90° to form a trail, dried under laminar air flow for 10 min and incubated under the same conditions as classic swarm assay. Observations were conducted at various intervals to assess colony initiation and development. For mixed colonies, strains were prepared separately and mixed prior to serial dilution.

Counting the number of cells in a colony was performed by lifting an entire colony and placing it in a small Eppendorf tube containing 1 mL of liquid LB medium. The sample was then diluted and inoculated onto a counting plate (2% agar, 2.5 g/L LB). This procedure was repeated for a large number of plates, each incubated for a different period of time.

Microscopy and Data Analysis.

An epifluorescence microscope Axio Zen 16 equipped with Axiocam 506 mono camera, was used to track colony initiation and development. This system features a large field of view (1″ × 1″) and utilizes a small magnification (~10×) for observation to form small fields of view, leveraging the fluorescence properties of the colonies for tracking. The relative height of the colony (and number of cells within it) is quantitatively analyzed based on the fluorescent intensity data (39). Some variations were obtained when compared to the direct cell counting (e.g., see minor differences between Fig. 1 and SI Appendix, Fig. S1, and within SI Appendix, Fig. S3).

For high-resolution detailed observation, an Axio imager Z2 Zeiss optical microscope equipped with Andor Solis Neo camera was used. The latter microscope supports both phase-contrast and fluorescence imaging modes. This system is equipped with a splitting device (Optosplit II) that enables dual excitation and acquisition of two fluorescence fields at the same spatial and temporal field. Subsequent image analysis is conducted using custom made programs in MATLAB to analyze the dynamics and orientations of the cells. Optical flow analysis, particle tracking methods, collective vorticities, alignment of the cells and spatial- temporal correlations were analyzed, which are previously used in our studies (39). Note that the speed due to colony expansion is negligible, less than 0.1% of the reported final speeds. All data is available in Dropbox (40).

Discussion

When a single planktonic bacterial cell is removed from its liquid culture and placed on an agar surface, it must rapidly adapt to the new physical and chemical conditions. To do so, it activates a complex biochemical program that ultimately transforms the individual cell population into a coordinated swarm—an efficient and robust mode of surface spreading. This transition encompasses multiple layers of regulation, including intracellular control of gene expression, mechanical remodeling, and the secretion of molecules that alter the surrounding physical environment. These changes are essential for facilitating collective motility. Thus, the early stages of swarming rely on a tight interplay between biological regulation and physical processes.

In the lab, swarming is studied in massive groups of up to millions of cells, spanning over centimeter range. However, it is more likely that in nature, swarming is a localized phenomenon involving a much smaller number of cells. For instance, in natural conditions, B. subtilis sporulates, and once sporulated, there are no means to ensure group maintenance, so that a germinating spore is likely to be isolated. Therefore, understanding the conditions underlying the initiation of swarming from a small number of cells is of high relevance to natural setting.

In this work we have found that vigorous bacterial swarming requires several cellular constituents that build up following specific stages that appear in order. Swarming will be initiated from a single cell only once several conditions and cellular changes are fulfilled. First, a constant production of surfactin is needed, and not simply a given initial amount, suggesting that the Com system must be active (but not over expressed). Second, a sufficient number of flagella is required for each cell. The flagella must function properly producing sufficient torque to overcome friction with the substrate. Finally, it is mandatory to be in a sufficiently moist niche.

Our study raises questions regarding the production and use of shared resources. All cells need to produce flagellin in order to swarm. This is evident, for example, in the spatial demixing of WT and immotile cells. However, surfactin seems to be a shared resource, possibly because it is an extracellular property. The distribution of surfactin production rates across WT cells is not uniform (SI Appendix, Fig. S2A). Moreover, it is enough to have only 10% surfactin producing cells to facilitate collective motion and swarming (Fig. 2). As producing surfactin consumes resources, it is not clear why the nonproducing Δsrf does not out-compete the WT and overtakes the colony, eventually terminating swarming. It is possible that over longer time scales this may indeed happen, but alternatively, the colony may have other means to discourage cheaters (41, 42). It is also interesting to note that endogenous vs. exogenous insertion of surfactin yields a different result. The latter may stem from the fact that the purchased surfactin (Sigma, diluted up to 10⁶-fold) is dissolved in ethanol that might affect the cells. Also, as shown, endogenous surfactin is generated at specific times and only after the cells have reached some density/quantity (which is independent of the agar concentration), while the exogenous is typically given at a specific stage and specific amount (although varied); we believe the colony must first build-up and reach a specific density so that the cells will not be too sparse. The latter is supported by the results obtained in the Com+ strain where the large amount of surfactin produced at early stages does not provide the normal swarm pattern obtained for the WT.

Alternatively, we note that a physiological difference between endogenously produced and exogenously added surfactin has been previously reported. Specifically, cells mutated for SwrC (YerP), a multidrug efflux pump thought to remove surfactin from the membrane (43), were defective in swarming despite the presence of surfactin on the plate (44). Moreover, swarming was restored to SwrC by mutation of surfactin biosynthesis, provided that surfactin was provided exogenously (44). Thus, intracellular accumulation of surfactin inhibited motility, but exogenous addition did not and facilitated WT swarming behavior. We infer that surfactin accumulates in the membrane early during synthesis and motility is transiently inhibited until SwrC is produced and/or otherwise activated. While motility inhibition may seem inconsistent with surfactin proficient cells’ earlier initiation of swarming, it is possible that transient surfactin accumulation has additional effects on cell physiology that are important as a colony grows from a single cell.

To summarize, the transition into swarming is facilitated by complex processes that occur both at the single-cell and collective scales. The transition is relatively sharp and occurs in several sequential states, with each stage lasting about a single generation. First, flagella are required at the single cell level, and their build-up sets a temporal limit. Then, more importantly, surfactant has to accumulate at sufficient amount at the group level to initiate the transition. We find that the two traits are linked, as surfactin mutants do not express flagella at high levels. We also find that either having too much or too little of each trait reduces the efficiency of the transition to swarming, suggesting that cells need to carefully balance these traits through appropriate regulatory pathways, as observed. Under similar conditions, the number of cells at which the transition occurs may vary between colonies, as indicated by the error bars in Fig. 1. The average colony size may further depend on environmental conditions such as temperature, available nutrients, humidity, and more. Still, for the specific case of B. subtilis growing on agar plates, we have provided a resolution for the sorites paradox; so how many bacteria make a swarm? Roughly 3,500.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Microscopic motion of cells during approximately 22 min of transition (agar concentration = 0.5%). Optical flow analysis: vectors indicate the local direction of motion, and arrow length and color represent the speed magnitude (as in Fig. 3). The movie is shown in real time. The field of view is approximately 130 μm.

Data, Materials, and Software Availability

All data have been deposited in Dropbox (40).