Bacteria in nature are most often found in the form of multicellular communities known as biofilms (Kolter and Greenberg, 2006; Aguilar et al., 2007). When compared to planktonic (free-living) state, cells in biofilms attach more firmly to hosts, have better access to nutrients, and are more protected from environmental insults, including sterilizing agents, antibiotics, and the immune system, (Stewart, 1994, 2002; Costerton et al., 1999; Stewart and Costerton, 2001; Donlan and Costerton, 2002; Mah et al., 2003; Fux et al., 2005).

The model organism Bacillus subtilis is a genetically manipulatable robust biofilm former that resides in soil, where it colonizes plant roots and protects its hosts from fungi and other bacterial species (Vlamakis et al., 2013). A subpopulation of flagellated cells (Kearns and Losick, 2005) was demonstrated to exist both in B. subtilis floating biofilms, and in biofilms formed on the solid surfaces (Vlamakis et al., 2008). B. subtilis uses its' peritrichous flagella for both swimming and swarming (Mukherjee and Kearns, 2014). Similar to other bacterial flagella, B. subtilis flagella consists of a filament composed of subunits of the flagellin protein, encoded by the gene hag (LaVallie and Stahl, 1989), and a motor subunit that enables the flagellar hook and filament rotation, formed by the proteins MotA and MotB (Mirel et al., 1992). The direction of rotation of the flagella, and consequently the mode of swimming, is influenced by the chemotaxis machinery, which is governed by the two component system CheA-CheY (Mukherjee and Kearns, 2014). The flagellar genes are tightly regulated, and expressed only under specific conditions (Guttenplan et al., 2013).

In B. subtilis biofilms, motile cells are considered a minority (Vlamakis et al., 2008), while the majority of the cells are held together by the exopolymeric substances (EPS) matrix. The EPS is composed of exopolysaccharides synthesized by the products of the epsA-O operon, the BslA protein that forms a hydrophobic coat, and TasA—a functional amyloid (Vlamakis et al., 2013). The EPS microenvironment was suggested to limit flagellar motility (Cairns et al., 2013). However, EPS may also promote collective motility, by generating groups of motile cells.

We here explored the potential roles of flagellated cells and collective motility in floating biofilms. Natural isolates of B. subtilis are a robust model for pellicle formation, as they are capable of forming highly structured biofilms in the interphase between the liquid medium and the air (Mcloon et al., 2011). Motility, EPS production and chemotaxis were shown to contribute to the efficiency of pellicle formation (Branda et al., 2004; Holscher et al., 2015). Furthermore, in a genetic screen, it was shown that both motility and EPS promote pellicle formation together in its early stages (Kobayashi, 2007). This raises the possibility that migration of the biofilm cells to the water-air interphase may be collective, and rely on communication and collaboration between the EPS producers and motile cells.

To test this hypothesis, we follow the pellicle formation in B. subtilis under various conditions in real time, and analyze the potential contribution of motility genes to the dynamics of pellicle formation. We show that aggregates of EPS producers and motile cells are formed during pellicle formation. We demonstrate that these aggregates can exhibit vortex-like collective motion in liquid, and contribute to biofilm formation in the water-air interphase. Finally, we propose a mechanism that allows the formation of these large groups.

In order to better elucidate the various roles of motility in the biofilm state of bacteria, we examined floating biofilm formation in liquid. Wild-type B. subtilis cells are able to create a floating biofilm (pellicles) when grown in a standing liquid culture. The ability to form a pellicle was compared between the wild-type strain, and mutants in different components of the motility (hag and motAB) or chemotaxis (cheA and cheY) machineries (Figure 1). In shaking liquid cultures, the wild-type and the mutants had comparable growth (Figure 1A). The carrying capacity (as judged by maximal optical density), the lag-time and the growth rate were comparable to the parental wild type strain both in 30°C (Figure 1A) and in room temperature (data not shown). Moreover, CFU numbers during the stages prior to pellicle formation were comparable between all strains (Figure 1B). Next, time-lapse photography was then used to follow pellicle formation in the different strains in standing liquid cultures. While all strains eventually created a pellicle, its formation in mutant strains was significantly delayed (Figure 1C and Video S1). The extent of the delay in both motility and chemotaxis mutants was unexpectedly robust, and changes in the well area did not significantly affect it (Figure 2 and Video S1). The fixed relative contribution of the motile cells and the lack of response to the changes in the well area indicate that flagellar motility has an additional role, unrelated to reaching the liquid-air interphase, independent of surface parameters.

Consistently with the hypothesis that cells migrate using flagellar motility during pellicle formation, we noticed the appearance of turbid areas in the well before the formation of the floating biofilm. Importantly, in the wells of the wild-type, but not in those of the mutant strains, these turbid areas constantly changed their location between frames, implying collective motility of the cells during the 15 min long time-lapse (Video S1). Time-lapse images of a side-view of the wells clearly showed vortex-like movements of the cells before the pellicle was formed in the wild-type, but not in the motility or chemotaxis mutants (Figure 3 and Video S2). Between 10 and 20 h post-inoculation, the wild-type strain showed high levels of motion that were not seen in the motility or chemotaxis mutants. As the mutants were delayed in pellicle formation, it is possible that this collective movement is involved in accelerating the pellicle formation process. While vortex-like movement of bacteria was previously shown (Mendelson, 1999), this is the first evidence for the presence of collective movement in static liquid and for its connection to biofilm formation in water-air interphase. The aggregates may achieve a collective movement by cooperation between motile cells and matrix producers, a mechanism which was already shown to promote sliding motility on solid surfaces (van Gestel et al., 2015; Kovacs, 2016; Dragos and Kovacs, 2017).

To further confirm the presence and composition of macro-cell clusters in the early stages of pellicle formation, and to explore the mechanism that allows these clusters to form, the patterns observed during early static growth in the biofilm medium were carefully analyzed. Indeed, within 20 h, clear macro-clusters appeared in each well (Figures 4A–C). The clusters were held together by extracellular matrix polymeric substances, as their size and abundance was reduced dramatically both in a tasA mutant, lacking the proteinous component of the matrix, in an eps mutant unable to produce exopolysaccharides (Figure 4B), and in a double mutant for both extracellular matrix polymeric substances (Figures 4B,C). All the mutants failed to form a structured mature pellicle (Figure S1). Furthermore, a microscopic analysis of the TasA matrix protein showed that these aggregates were formed after the production of the ECM, and contained numerous motile cells (Figure 4A), while the surrounding medium contained isolated motile cells. Aggregation did not result from differential cell numbers (Figure 4D).

Cell to cell communication may promote aggregation. The production of autoinducer-2 (AI-2) signal molecules, mediated by the activity of the LuxS enzyme, in response to changes in cell density, was established in B. subtilis (Lombardia et al., 2006). This autoinducer has been shown to promote initial biofilm formation and cell aggregation in several Gram-positive bacterial species (Lombardia et al., 2006; Trappetti et al., 2011; He et al., 2015). Consistently, we found that a mutant defective in AI-2 signaling is dramatically impaired in early aggregation, as well as in pellicle formation, indicating a role of this signal molecule in the formation of floating biofilms by B. subtilis (Figure 5). The aggregation of a luxS mutant was restored in the presence of DPD (4,5-dihydroxy-2,3-pentanedione), the precursor of AI-2, in agreement with previous studies (Miller and Bassler, 2001; Bassler and Losick, 2006; Camilli and Bassler, 2006; Ng and Bassler, 2009; Figure 5A). In contrast, mutants in flagellar motility formed macro-clusters similarly to the wild-type (Figure S2), suggesting that the formation of the clusters depends on extracellular matrix and cell-cell communication, whereas their collective motion appears to depend on flagellar motility. These results indicated that EPS matrix and chemical communication play a role in aggregation.

Lastly, we wanted to confirm that the collective motion has a role in promoting biofilm formation. To test this hypothesis, we disrupted the macro-clusters during early stages of pellicle formation. This disruption caused a reproducible delay in pellicle formation, without impairing the number of viable cells (Figure 5B). The delay was especially pronounced in a mutant lacking the AI-2 communication signals (Figure 5B). As mutants in exopolysaccharides do not form an established pellicle (Branda et al., 2004), it was impossible to assess whether they can recover from the dissolution. These results are consistent with the hypothesis that cell to cell communication mediated by the AI-2 pheromone has critical importance in the stabilization of macro-clusters, consistent with the complete restoration of aggregation properties by exogenous AI-2 (Figures 5A,B, Table S3). Similarly, a matrix mutant in tasA had a defect in re-aggregation following physical disturbance of the aggregates (Figure S3).

In this study, we sought to gain better understanding of the role of motility in bacterial floating biofilms. Motile cells enable collective motion in the early stages of floating biofilm formation. This motion can be seen in wild-type strain, but is absent in the mutant strains. Furthermore, collective motility accelerates pellicle formation (Figures 2, 3 and Videos S1, S2). Both flagellar rotation and chemotaxis contribute to pellicle formation under different conditions, suggesting the demand for flagellar motility is robust. Furthermore, it was demonstrated that the collective motility was made possible by vortex-like motion (Videos S1, S2), which is absent in the motility/chemotaxis mutants. Most the location of the air–liquid interface by sensing oxygen gradient, as previously shown (Holscher et al., 2015). However, cells can either first adhere randomly to the water-air interphase as single cells to generate a pellicle as suggested previously (Holscher et al., 2015), or collectively and directionally migrate to the interphase. Our results show collective translocation of bacterial vortices toward this niche.

The collective translocation to the interphase involves the formation of multicellular aggregates and strongly depends on diffusible signal molecule, AI-2 (Figure 5). This universal signal molecule regulates cell density-dependent phenotypes in many bacterial species (Ng and Bassler, 2009), and was previously shown to promote aggregation in Gram-positive pathogens (Lombardia et al., 2006; Trappetti et al., 2011; He et al., 2015). Single-cell analysis of the aggregating cell clusters indicates that they are primarily composed of EPS matrix producers and motile cells. As both EPS and rotated flagella are required for this motion, it is reminiscent of swarming motility, a rapid multicellular bacterial surface movement powered by rotating flagella, over solid surfaces (Kearns, 2010). Interestingly, B. subtilis does not make vortices during swarming on solid (Kearns, 2010), but exhibit vortex-like motion under our conditions in liquid (Videos S1, S2). These results may suggest a need for large, localized groups of cells traveling in a common circular path (generating a “vortex”) in liquid, and a potential need of the founders of the pellicle to act as a group to assemble the biofilm efficiently. These vortices are achieved by cooperation and chemical communication between two distinct cell types, EPS producers and motile cells.

In nature, bacteria reside in complex communities and an efficient colonization of a preferred growth niche can provide a significant advantage over potential competitors (Elias and Banin, 2012). We suggest that the robust and efficient strategy for enhanced colonization of interphases by B. subtilis may provide an essential mechanism for its survival in complex multispecies environments.

NS, GR, AK-P and IK-G designed the experiments. NS, AK-P, IK-G, and GR performed the experiments. IK-G and NS contributed reagents. IK-G, AK-P, and NS wrote the manuscript. AK-P and GR contributed equally.