Research ArticleBiofilms

The extracellular matrix protein TasA is a developmental cue that maintains a motile subpopulation within Bacillus subtilis biofilms

See allHide authors and affiliations

Science Signaling  19 May 2020:
Vol. 13, Issue 632, eaaw8905
DOI: 10.1126/scisignal.aaw8905

Maintaining motile reserves

During biofilm formation, bacterial cells switch from a motile planktonic state to a matrix-producing, adherent state. Although bacterial biofilms are generally sessile, some, such as those formed by Bacillus subtilis, can spread to overtake and kill neighboring colonies of competitor species. Steinberg et al. found that a motile subpopulation of cells within B. subtilis biofilms was required for the biofilms to spread over foreign objects. This process required the matrix protein TasA, which stimulated a subset of cells within the biofilm to revert from a matrix-producing state to a motile state, thus ensuring that the colony could spread.


In nature, bacteria form biofilms—differentiated multicellular communities attached to surfaces. Within these generally sessile biofilms, a subset of cells continues to express motility genes. We found that this subpopulation enabled Bacillus subtilis biofilms to expand on high-friction surfaces. The extracellular matrix (ECM) protein TasA was required for the expression of flagellar genes. In addition to its structural role as an adhesive fiber for cell attachment, TasA acted as a developmental signal stimulating a subset of biofilm cells to revert to a motile phenotype. Transcriptomic analysis revealed that TasA stimulated the expression of a specific subset of genes whose products promote motility and repress ECM production. Spontaneous suppressor mutations that restored motility in the absence of TasA revealed that activation of the biofilm-motility switch by the two-component system CssR/CssS antagonized the TasA-mediated reversion to motility in biofilm cells. Our results suggest that although mostly sessile, biofilms retain a degree of motility by actively maintaining a motile subpopulation.


Bacteria in nature most often reside in multicellular differentiated communities of cells commonly referred to as biofilms (1). Compared to the planktonic (free-living) state, bacteria in biofilms are more protected from external insults, such as antibiotics and the host immune system, attach more firmly to their hosts, and have improved access to nutrients (29). Whereas biofilms formed by pathogenic bacteria are deleterious to the host (3, 10), biofilms formed by other bacteria can be beneficial to the host (1115).

Bacterial biofilms follow a robust and complex pattern of differentiation. One main property defining biofilms is the production of an extracellular matrix (ECM) that surrounds, connects, and protects the cells and allows them to adhere to each other or to a surface (1618). The main components of the bacterial ECM are exopolysaccharides, proteins, nucleic acids (19), and biogenic minerals (2022). Exopolysaccharides play a fundamental structural role in biofilms of different bacterial species (2328). Biofilms of Bacillus subtilis produce several exopolysaccharide polymers (29, 30) through the combined actions of enzymes encoded in the epsA-O operon. Mutants for this operon, and specifically the glycosyltransferase-encoding gene epsH, lack the exopolysaccharide component of the ECM and form colonies that are featureless, in contrast to the wrinkled colonies of wild-type cells (31). The major proteinaceous component of B. subtilis ECM is the protein TasA, encoded by the tapA-sipW-tasA operon (32, 33). TasA forms amyloidal fibers (32, 34, 35) that are attached to the cell wall and, in conjunction with other extracellular components, promote cell-cell adhesion (32, 36). Colonies formed by tasA deletion mutants are smaller and less structurally complex than those of wild-type strains (34).

Being energetically costly, ECM production is under tight regulation and activated only under appropriate conditions, such as when neighboring cells already produce ECM. In B. subtilis, two positive feedback mechanisms lead to increased ECM production: (i) the tyrosine kinase EpsAB, which is specifically activated by B. subtilis exopolysaccharides (37); and (ii) the disruption of flagellar rotation, which has been proposed to activate ECM production through a mechanosensory mechanism (38, 39). Motility is also energetically costly, involving the production of multiple protein components and energy investment in flagellar rotation. Thus, the expression of genes necessary for flagella structures, assembly, and rotation is all coordinated (4042). Only a subpopulation of cells expresses flagellar genes, which are activated by the alternative sigma factor D (σD) (41). Similarly to other bacterial species, B. subtilis was long thought to shut down flagellar motility when grown on high-friction surfaces (43, 44).

In an individual B. subtilis cell, a regulatory switch couples the activation of ECM production with the repression of motility (38, 39, 4560). This regulatory switch depends mainly on two master regulators that together control both motility and biofilm development—the homologous proteins SinR and SlrR (5459). During planktonic growth, SinR represses the expression of the ECM production operons epsA-O and tapA-sipW-tasA, as well as the expression of slrR (61, 62). Once the biofilm state is induced, SinI inhibits SinR activity by binding to it and preventing SinR from binding DNA, resulting in derepression of the ECM operons and slrR (61, 63, 64). In turn, SlrR binds to SinR, creating a heterodimer that represses the fla/che operon, which encodes key components of motility (59) and gene encoding autolysins, which are important for breaking down the cell wall so that cells can separate from one another (56). Thus, the same regulator, SinR, represses either the ECM operons or motility but not both simultaneously in the same cell, and therefore, the two transcriptional programs are mutually exclusive at the single-cell level. However, at the population level, other mechanisms may act to maintain cell heterogeneity and allow coexistence of the two subpopulations in the same biofilm.

We previously reported that in B. subtilis biofilms, flagellar motility is necessary for a growing colony to extend toward, engulfing and, eventually, eradicating a neighboring Bacillus simplex colony (65). Because the mechanism by which B. subtilis kills B. simplex involves molecules with poor diffusion, B. simplex survives better when inoculated next to immotile B. subtilis mutants (65). Thus, the outcome of the interaction between these two species is determined by whether cells in B. subtilis colony are motile, suggesting that the maintenance of motility with a biofilm is crucial for survival and successful competition with other bacteria.

In this study, we further examined biofilm motility in the absence of competition. We found that TasA acted as a signal that sustained biofilm motility in a subpopulation of cells within the colony. We also identified a previously unreported regulator of the motility switch, a two-component system coupling ECM production with motility. Our results imply that bacterial ECM proteins, similar to ECM proteins of multicellular organisms (6668), can serve as signals during biofilm development and sustain collective migration.


Flagellar motility enables B. subtilis biofilms to engulf objects on solid surfaces

We previously reported that the ability of a biofilm colony to engulf a nearby competitor depends on flagellar motility (65). Here, we examined the role of motility in biofilm development on high-friction surfaces. To test whether the engulfment behavior described by Rosenberg et al. (65) is an intrinsic property of the biofilm, occurring in the absence of a potential competitor, we assessed the ability of B. subtilis colonies to engulf artificial objects using discs made from either filter paper or cellulose acetate. We compared wild-type B. subtilis and two motility mutants: Δhag, which lacks the flagellar structural protein flagellin (Fig. 1A), and ΔmotAB, which lacks a motor subunit critical for flagellar rotation (fig. S1). Both mutant strains are defective in engulfing and eradicating a competitor colony (65). The overall population was monitored by phase microscopy, and the subpopulation expressing motility genes was identified by a green fluorescent protein (GFP) reporter driven by the hag promoter (Phag-GFP), located in the amyE locus. When a wild-type B. subtilis colony was inoculated near a cellulose acetate disc, the colony expanded and, after 36 hours, completely surrounded the disc (Fig. 1A and fig. S1). In case of a paper disc, the colony not only surrounded but also covered the disc, showing that the ability to extend and engulf is not a response to the presence of a competitor but is instead an inherent property of the biofilm. In this experimental system, neither motility mutant surrounded the discs or covered the paper disc by 36 hours (Fig. 1A and fig. S1).

Fig. 1 Flagellar motility promotes B. subtilis colony expansion and engulfment of foreign objects.

(A) Time-lapse images of the engulfment of cellulose acetate or filter paper discs placed 0.3 cm from the inoculation point by wild-type (WT) and Δhag colonies, both harboring the flagellin reporter Phag-GFP. Images show colony growth at the indicated time points after inoculation. For each strain, representative bright field and GFP fluorescence images are shown, as well as a magnified image of GFP fluorescence at the 36-hour time point. n = 9 for each strain. Scale bars, 2 mm. (B) Higher magnification representative images showing cells covering a filter paper disc at the indicated time points for a wild-type colony carrying both the Phag-GFP (motility, green) and PtapA-mKate (matrix, red) reporters. Three colonies per strain were examined, with at least 10 representative fields recorded per colony. Scale bar, 1 μm. (C) Quantification of filter paper disc engulfment by the parental wild-type and indicated mutant strains. Discs were placed at distances of 0.3 and 0.5 cm from the inoculation point. At the indicated times, the extent of engulfment was determined as the percentage of the disc circumference covered by bacterial cells. Disc coverage by at least 12 colonies per strain per condition is shown. The boxes indicate the lower and upper quartiles, and the central line indicates the median. Whiskers above and below the box indicate the 90th and 10th percentile. Outliers are shown as filled circles (•). The differences between all the mutants and WT and between ΔepsH and Δsrf strains and Δhag strain were significant. P < 0.05, as determined by analysis of variance (ANOVA) test, followed by Tukey’s honestly significant difference (HSD).

Because we observed that the population moving toward the disc during engulfment expressed Phag-GFP, we examined whether this was the only cell type moving. We repeated the experiment with a wild-type strain carrying both the Phag-GFP reporter to label the motile cells and the PtapA-mKate2 reporter to identify the matrix-producing cells. The motile GFP-positive population was moving toward the disc at 24 hours, and eventually, both populations were detected in the area of engulfment and on the disc (fig. S2). Moreover, when examined at a higher resolution, the two cell types were tightly associated, forming clusters that contained both cell types (Fig. 1B).

Because flagellar propulsion is not the only mode of bacterial motility, we examined whether other types of motility could also play a role in the observed phenomena. We inoculated biofilm colonies next to a paper disc at two distances (either 0.3 or 0.5 cm), followed them over time, and calculated the extent of disc engulfment (see Materials and Methods). We compared the effect of deleting hag to that of deleting gene encoding key enzymes for producing exopolysaccharides and surfactin [epsH and srfAA, respectively (69, 70)], both of which are involved in collective motility. We found that all the mutants were compromised in their ability to engulf the disc (Fig. 1C). After 48 hours, wild-type cells were already encircling the disc from the longer distance (0.5 cm) and almost completely covering it from the shorter distance (0.3 cm); after 72 hours, wild-type cells had completely engulfed the disc from both distances. On the other hand, none of the mutants encircled the disc completely after 72 hours, even starting from the shorter distance. At the longer distance, none had even started to engulf the disc at 48 hours. The defect in both the ΔepsH and Δsrf strains was more severe than that in the Δhag strain. In addition, we tested whether deletion of the matrix proteinaceous component TasA affected colony expansion and engulfment, because TasA is known to play a role in cell-cell adhesion during sliding motility (71). Similarly to other mutants we examined, we found that ΔtasA was compromised in its ability to engulf the disc. Together, these results suggest that motility is an important feature of sessile biofilm colonies and that at least two distinct modes of physical motility might be maintained during biofilm development.

The matrix protein TasA is required for flagellin expression

We repeated the engulfment experiment with wild-type and ΔtasA strains carrying a Phag-GFP reporter and confirmed that, similar to Δhag, ΔtasA could not engulf either paper or cellulose acetate discs (Fig. 2A). Moreover, in addition to the defect in engulfment, during the first 48 hours of development, the ΔtasA strain expressed less GFP (Fig. 2A and fig. S3). Because the different morphologies of the wild-type and ΔtasA colonies make visual comparison of GFP intensities difficult, we used flow cytometry to quantify Phag-GFP expression in the wild-type strain, in the ΔtasA mutant, and in the ΔepsH strain that is unable to produce the exopolysaccharide component of ECM, as well as in the ΔtasA ΔepsH double mutant (Fig. 2 B and C). In the wild-type strain, two subpopulations of cells were clear, one that overlaps the autofluorescence levels and one that strongly expresses Phag-GFP, in accordance with previous studies (53). Whereas ΔepsH showed only a slight reduction in the percentage of the cells expressing GFP, ΔtasA showed a notable reduction, with the double mutant completely losing the GFP-expressing population. The median value for fluorescence intensity of the GFP-expressing cell subpopulation remained similar in all strains. Thus, GFP levels in the expressing cells were the same, but the percentage of cells actually expressing GFP was lower in the ECM mutants. To confirm that the reduction in GFP expression did not result from growth rate differences between the wild-type and ECM mutant colonies (72), we examined GFP expression in shaking cultures, in which growth rates of the strains were comparable (Fig. 2D), and confirmed that they were consistent with those in the colonies (Fig. 2E).

Fig. 2 TasA induces expression of flagellar genes in biofilms.

(A) Time-lapse images of the engulfment of cellulose acetate or filter paper discs placed 0.3 cm from the inoculation point, by wild-type and ΔtasA colonies, both harboring the Phag-GFP reporter. Images show colony growth at the indicated time points after inoculation. For each strain, representative brightfield and GFP fluorescence images of one colony are shown, as well as a magnified image of GFP fluorescence at the 36-hour time. n = 9 colonies for each strain. Scale bars, 2 mm. (B) Representative bright field and fluorescence images of colonies of wild-type and the indicated ECM mutants harboring the Phag-GFP reporter, grown for 48 hours. Scale bar, 2 mm. (C) Flow cytometry analysis of cells from colonies in (B). Nonfluorescent colonies (gray) were used as a control (gray). The dashed vertical line indicates the autofluorescence signal used for gating. The solid line indicates the median of the gated population. Percentage of gated cells ± SD of three colonies is shown. ***P ≤ 0.001, compared to wild type, as determined by ANOVA, followed by Tukey’s HSD. n = 9 colonies for each strain. AU, arbitrary units. (D) Growth curves for wild-type and the indicated ECM mutants harboring the Phag-GFP reporter in shaking liquid cultures. Averages and SDs of six cultures for each strain are shown. (E) Phag-GFP fluorescence of the cultures in (D) at 10 hours of growth normalized to OD600. Fluorescence of six cultures at 10 hours of growth normalized to OD600 is shown. **P ≤ 0.01, as compared to wild type, as determined by ANOVA, followed by Tukey’s HSD.

Secreted TasA is required for flagellar motility in biofilms

TasA is a structural amyloid protein, supporting cell-cell adhesion, and, similar to other amyloids, forms fibers in vitro. We used a functional TasA-mCherry protein fusion reporter (73) to determine whether fibrillation of TasA was required for sustaining motility. Fibrillation of extracellular TasA was, although very rarely, observed (Fig. 3A, fig. S4, and movie S1). To understand the relationship between TasA and motility, we examined a strain in which the native tasA allele was replaced with a sequence encoding TasA-mCherry and which also carried the Phag-GFP reporter. We found that although the culture started as a nearly homogenous population of ECM-producing cells expressing TasA, over time, the daughter cells lost the TasA-mCherry signal and switched to the motile expression profile as indicated by GFP expression (Fig. 3B). No TasA fibers were observed during this switch, further suggesting that they are not required for the transition from ECM producer to motile cell.

Fig. 3 Secreted TasA maintains hag expression.

(A) Time-lapse images of a strain carrying PtapA-tapA-sipW-tasA-mCherry, taken using the CellASIC microfluidic platform. Images were taken at the fixed time intervals after loading as indicated. Red, mCherry. Images are representative of nine independent experiments. Scale bars, 10 μm. (B) Time-lapse images a strain carrying PtapA-tapA-sipW-tasA-mCherry and Phag-GFP, taken using the CellASIC microfluidic platform. Images were taken at the fixed time intervals after loading as indicated. Red, mCherry; yellow, GFP. White arrowheads indicate progeny of the cell marked with an arrow. Images represent one of three independent experiments. Scale bars, 10 μm. (C) Normalized Phag-GFP fluorescence in cultures treated with the indicated compounds as a function of each compound’s osmolarity or viscosity. Fluorescence at 10 hours of growth was normalized to the OD600 and then normalized to the fluorescence of a nontreated sample in the same experiment. Each point represents the average of at least three cultures; bars represent SD. (D) A representative colony of wild-type and the indicated mutants harboring the Phag-GFP reporter, grown for 24 hours. Representative brightfield and GFP fluorescence images are shown. n = 5 colonies for each strain. Scale bar, 2 mm. (E) Flow cytometry measurements of colonies in (B) and floating biofilms (pellicles) of the indicated strains. Data for two representative colonies (of seven analyzed) for each strain are shown. The dashed vertical line indicates the autofluorescence used for gating. (F) Flow cytometry analysis of colonies of the indicated strains, grown for 24 hours. Percentage of gated GFP-positive cells ± SD of nine colonies is shown. ***P ≤ 0.001 as compared to wild type or ΔtasA, as determined by Dunnett’s test.

ECM polysaccharides and amyloids change the physical properties of the cells’ microenvironment, such as osmolarity and viscosity, which, in turn, may influence biofilm development and the expression of biofilm genes (28, 7476). To test whether increased osmolarity or viscosity could restore flagellin expression in the ΔtasA strain, we added various substances that affect those properties to the culture and examined their ability to restore Phag-GFP expression. For each substance, we plotted GFP expression normalized to a nontreated sample, as a function of the measured osmolarity or viscosity (Fig. 3C), in the range that did not cause reduction of growth rate (fig. S5). In all cases, GFP expression remained the same as in the nontreated sample, regardless of the added substance. Because dextran and amino acids did not induce Phag-GFP expression, the effect of TasA was independent of its potential uses as a carbon and nitrogen source.

We next investigated the role of TasA and its secretion, which is mediated by the signal peptidase SipW (33, 77), in flagellin regulation. In sipW mutants, TasA is expressed but remains intracellular (77). Consistent with the hypothesis that TasA acts as an external signal, a sipW mutant had a severe defect in Phag-GFP expression both in biofilm colonies (Fig. 3, D and E) and floating biofilms (pellicles) (Fig. 3E). Overexpressing tasA significantly increased Phag-GFP expression in ΔtasA, and overexpressing the complete tapA-sipW-tasA operon did so even more efficiently (Fig. 3F). In contrast, neither SipW nor TapA restored the expression of Phag-GFP when expressed alone. In addition, coculturing ΔtasA with a complemented strain expressing the operon from a strong promoter (ΔtasA, Phs-tapA-sipW-tasA) increased Phag-GFP expression significantly compared with the monocultured ΔtasA control (fig. S6A). Purified TasA protein partially rescued Phag-GFP expression ΔtasA mutants under floating biofilm conditions (fig. S6B). Overall, these results suggest that TasA must be secreted and functional to activate motility within biofilms.

TasA inhibits the expression of flagellar genes

To systematically explore the regulatory role of TasA, we performed transcriptomic analysis, comparing transcription profiles of the wild-type and the ΔtasA strains. Gene-annotation enrichment analysis (using DAVID) identified three specific pathways affected by TasA deletion: flagellar assembly, chemotaxis, and nitrate assimilation. We found that in the absence of TasA, genes related to flagellar motility and chemotaxis were repressed and genes for nitrate assimilation were induced (Fig. 4, A and B, and data file S1). A similar pattern in gene expression changes was observed in a mutant lacking SipW (Fig. 4, C and D). These results further support a role of TasA as a secreted signal reactivating motility within the biofilm.

Fig. 4 Transcriptomic analysis of cellular pathways regulated by TasA.

(A) Volcano plot showing genes differently expressed in ΔtasA relative to wild-type cells. Genes belonging to functional categories identified by DAVID analysis are highlighted. (B) Log2 fold change (FC) in gene expression of the genes highlighted in (A) (marked with asterisk) and their neighboring genes. The genes within each functional category are ordered by their chromosomal location. (C and D) Log2 FC in expression of genes highlighted in (A), in ΔsipW relative to wild-type cells.

TasA stimulates reversal to motility after entering the biofilm state through the motility-biofilm switch

We next examined the relationship between TasA and Phag-GFP expression by single-cell analysis over time. To that end, we compared wild-type and ΔtasA cells harboring Phag-GFP as a reporter for flagellar motility and PtapA–cyan fluorescent protein (CFP) as a reporter for ECM production. Those strains were grown in liquid biofilm-inducing medium (MSgg), with shaking, and placed on agar pads for time-lapse microscopy (Fig. 5, A and B, and movie S2). We found that cells were expressing either the motility reporter, the ECM reporter, or neither but never both at the same time (Fig. 5A). Frequently, we observed that once cells in the chain stopped expressing ECM, their progeny activated motility after a few cell divisions. We next examined the activation of Phag-GFP expression in growing chains. In most chains of wild-type bacteria (72 of 77, 93.5%), at least one progeny cell activated motility during the experiment. In sharp contrast, this was the case in only about half of the analyzed chains (42 of 79, 53.2%) for the ΔtasA mutant strain (Fig. 5B and movie S3). Next, we measured the distribution of chain lengths at different time points after the transition to the agar pad. The final length distribution of ΔtasA cells showed a lower number of single cells (identified as chains of length ≤ 10 μm) and longer chains (median, 4.3 μm for the wild-type and 6.1 μm for ΔtasA), reflecting a lower number of chain-breaking events (Fig. 5C and fig. S7). In ΔtasA, a higher percentage of cells expressed the PtapA-CFP reporter (Fig. 5D) or the PtapA-GFP reporter (fig. S8). The longer chains in ΔtasA were the ones that expressed PtapA-CFP (Fig. 5E). Together, these results support the conclusion that in the absence of tasA, the switching rate from matrix producers to motile cells decreased and cells tended to stay longer in the nonmotile state.

Fig. 5 TasA stimulates the reversion to motility from the biofilm state.

(A) Time-lapse images of wild-type cells carrying PtapA-CFP (blue) and Phag-GFP (yellow). Images were acquired at the indicated time after inoculation. Arrows indicate one parental cell, and arrowheads indicate its progeny. n = 3 independent experiments. Representative fields are shown. Scale bars, 10 μm. (B) As in (A), but the wild-type strain was compared to the ΔtasA strain carrying the same reporters. n = 3 independent experiments. Representative fields are shown. Scale bars, 10 μm. (C) Chain length at 4.5 hours after inoculation, in wild-type and ΔtasA strains. At least 1800 chains for each strain were analyzed (at least 18 fields in n = 3 independent experiments). (D) Fraction of chain length expressing PtapA-CFP at 4.5 hours. At least 18 chains for each strain were analyzed from three independent experiments. Data represent the means with SD. ***P ≤ 0.001 as determined by Student’s t test. (E) Length of chains in which PtapA was expressed at 4.5 hours after inoculation, in wild-type and ΔtasA strains. At least 180 chains for each strain were analyzed from three independent experiments. (F) Flow cytometry of colonies of the indicated genotypes harboring the Phag-GFP reporter, grown for 48 hours. Three colonies for each strain are shown, and the same wild-type control is shown in both traces. The dashed vertical line indicated autofluorescence used for gating. (G) Ratio of cells expressing Phag-GFP above autofluorescence between the strains shown in (F); comparing the parental strain with a derivative carrying an additional tasA deletion. Average of gated cells in three colonies per strain is shown; error bars represent SD. *P ≤ 0.05 for the ΔtasA derivative strains compared to the parental strain, as determined by ANOVA test, followed by Tukey’s HSD.

To determine whether TasA acted through the motility-biofilm switch, we examined the effect of known regulators of the motility-biofilm switch on motility activation by TasA. Deletion of either sinI or slrR, the master regulators that promote switching from motility to ECM production, resulted in higher numbers of motile cells (Fig. 5F), and tasA deletion did not further affect the number of motile cells (Fig. 5, F and G). In contrast, in the background of ΔdegU, which encodes a global regulator that induces matrix production independently of SinI/SlrR (49), tasA deletion caused an even greater reduction in the number of motile cells (fig. S9, A and B). In addition, although the membrane kinase KinD is known to sense the polysaccharide component of the ECM (72, 75), deletion of kinD had little or no effect on the response to TasA deletion, indicating that TasA acts independently of KinD (fig. S9C). Similar results were obtained with related kinases (KinA, KinB, and KinC) (78) that act upstream to Spo0A for matrix production (fig. S10).

Overall, we showed that TasA increased the switching rate of cells that are in the nonmotile state back to motility, and it did so through the motility-biofilm regulatory switch but independently of DegU transcriptional regulator or any single member of the KinA to KinD histidine kinases.

The two-component system CssR/CssS stimulates ECM production

To understand the molecular mechanism by which TasA promotes the switch to motility within biofilms, we followed ΔtasA colonies harboring Phag-GFP, for the emergence of suppressor mutants in which the level of expression of motility genes was restored.

When ΔtasA colonies were grown more than 5 days, protrusions began to appear at the edge of the colony (Fig. 6A). Many of these protrusions were highly fluorescent, indicating high expression from the hag promoter. Additional mutants were isolated from the circumference of the colony, in areas that showed high florescence but did not protrude. In total, 80 hyperfluorescent mutants were isolated from 141 ΔtasA colonies. Even after many passages, the strains retained this phenotype, indicating a genetic change (Fig. 6B). Testing of several representative mutants in a soft-agar (0.25%) swimming assay indicated that high expression of hag resulted in faster swimming (Fig. 6C).

Fig. 6 Spontaneous hypermotile mutants in ΔtasA colonies target regulators of the motility-biofilm switch.

(A) Overview of 25 representative ΔtasA colonies (of 141 colonies) harboring Phag-GFP reporter and grown for 5 days, showing fluorescent protrusions at the colony edge. Enlarged images of one typical ΔtasA colony with a protrusion (arrowhead) are shown. Scale bar, 2 mm. (B) An example of a colony formed by Mut310, one of 80 mutants isolated from protrusions as in (A). Scale bar, 2 mm. (C) Diameter of swimming ring in soft agar after 10 hours for wild-type, ΔtasA, ΔtasAΔeps, and representative mutants isolated from protrusions as in (A). Means ± SD of the indicated strains (four colonies per strain) and representative brightfield images of WT and Mut310 on swimming plates are shown. ***P ≤ 0.001 compared to WT, as determined by ANOVA, followed by Tukey’s HSD. Scale bar, 2 cm. (D) Mutations in the slrR locus that were found in hypermotile mutant strains Mut205, Mut340, Mut310, Mut316, and Mut331 isolated from colony protrusions. Mutations in additional genes were identified in the isolates listed in the table. Mutant strain names are shown in green with the parental strain name noted in parentheses. (E and F) Schematic model of SinR- and SlrR-mediated regulation of biofilm and motility genes. (E) During planktonic growth, the expression of both ECM operons and slrR is repressed by SinR. During the transition to the biofilm state, SinI represses SinR, causing the activation of the ECM operons and slrR. In turn, SlrR binds to SinR causing it to repress the motility genes. (F) In slrR mutants, as in the isolates found in this screen, SinR no longer represses the motility genes, and SinR can repress the ECM operons because it is no longer complexed with SlrR.

We sequenced the genomes of nine isolated mutants (Fig. 6D). Four of the nine mutants contained mutations in the slrR locus. Using polymerase chain reaction (PCR) and Sanger sequencing, we identified eight additional mutants in the slrR gene. Two isolates had a synonymous mutation in the biofilm master regulator sinR at a specific serine codon (Fig. 6D), previously found to affect the abundance of SinR, leading to changes in expression of ECM genes (79). In two other isolates, two different nonsynonymous mutations were found in the biofilm inducer remA. In both cases, mutations in remA appeared in combinations with additional mutations, raising the possibility that deletion of remA alone was not sufficient to fully restore motility to tasA mutant. Thus, most of the suppressor mutations appearing in the ΔtasA colonies were in known regulators of the motility-biofilm switch (Fig. 6, E to F).

One of the mutations identified in the suppressor mutant screen affected the two-component system CssR/CssS—isolate Mut328 carried a nonsynonymous mutation (K308E) in the histidine kinase domain of the membrane-bound histidine kinase CssS (fig. S11). The CssR/CssS two-component system was previously shown to participate in the response of B. subtilis to secretion stress (80, 81) but was not thought to be connected to either matrix production or motility. This suppressor mutation appeared together with a mutation in remA. To examine the possibility that CssR/CssS contributes to the motility-biofilm switch, we deleted the cssRS operon.

Flow cytometry analysis revealed that CssR/CssS regulated both motility and matrix expression. ΔcssRS colonies contained a larger motile subpopulation, reflected by an increase in Phag-GFP expression (Fig. 7A) and a reduction in PtapA-GFP expression (Fig. 7B). Because CssR/CssS was initially identified as a ΔtasA suppressor, we examined the combined effect of tasA and cssRS on motility. Deletion of tasA in the ΔcssRS background reduced the number of motile cells compared to ΔcssRS alone, but to a lesser degree than it did in the wild-type background (Fig. 7C), confirming a role for CssR/CssS in inhibiting motility. The original mutant isolated from the screen (Mut328) contained an additional mutation in the biofilm inducer remA. Therefore, we examined the relative contribution of the two mutations to motility in ΔtasA mutants. Both single mutations restored the expression of Phag-GFP, and the effect of ΔremA was stronger than that of ΔcssRS. However, neither deletion alone restored motility to that of the original isolate containing both mutations (Fig. 7D). Together, these results support the conclusion that remA alone was not sufficient for full restoration of motility and verified an independent role for CssR/CssS in the suppression of motility.

Fig. 7 The CssR/CssS two-component system represses the action of TasA on the motility-biofilm switch.

(A) Flow cytometry measurements of colonies of wild-type and ΔcssRS strains harboring the Phag-GFP reporter, grown for 24 hours. Nonfluorescent control, gray; GFP, black (WT) and green (ΔcssRS). The dashed vertical line indicates autofluorescence used for gating. Percentage of gated cells ± SD of three colonies is shown. ***P ≤ 0.001, as determined by Student’s t test. n = 3 independent experiments. (B) As in (A) for WT and ΔcssRS strains harboring the PtapA-GFP reporter. Percentage of gated cells ± SD of three colonies is shown. ***P ≤ 0.001, as determined by Student’s t test. n = 3 independent experiments. (C) Ratio of the fractions of Phag-GFP–expressing cells (above autofluorescence) between the indicated strains, with and without an additional tasA deletion. Means and SD of nine colonies for each strain are shown. ***P ≤ 0.001, as determined by Student’s t test. n = 3 independent experiments. (D) Flow cytometry measurements of colonies (three for each strain) of the indicated genetic background harboring Phag-GFP reporter, grown for 24 hours. Percentage of gated cells ± SD of three colonies is shown in parentheses. ***P ≤ 0.001, compared to Mut328, as determined by ANOVA, followed by Tukey’s HSD. (E) K-means clustering of 401 genes that were differentially expressed in the ΔtasA and ΔcssRS transcriptomes relative to WT. n = 3 colonies for ∆tasA and ∆cssRS and n = 2 colonies for WT, all grown for 24 hours. (F) Log2 FC in gene expression of EPS production and sporulation-related genes in ΔtasA and ΔcssRS versus the WT parental strain. (G) STRING interaction network of genes that show differential expression behavior in the ∆tasA and ∆cssRS mutants. Genes without STRING interactions were omitted.

Comparison between the transcriptomes of ΔcssRS and ΔtasA mutants revealed a subset of genes regulated by both CssR/CssS and TasA (Fig. 7E; fig. S12, A and B; and data file S1). We identified two pathways differentially regulated by the two proteins, which could account for suppression of TasA by CssR/CssS. TasA repressed the Spo0A pathway, as reflected by the repression of motility genes (Fig. 4B) and the activation of matrix production and sporulation (Fig. 7F) in ΔtasA mutants. In contrast, those pathways were repressed in ΔcssRS (Fig. 7F), suggesting that CssR/CssS activated the Spo0A pathway. Last, we identified a subset of genes repressed in both mutants (Fig. 7G and fig. S13). Deletion of either tasA or CssRS led to a decrease in the transcription of several stress response genes belonging to the SigB regulon, including sigB itself (fig. S13 and data file S1). Because CssR/CssS is known to sense secretion stress (80), this could be a direct result of the high level of production and subsequent secretion of TasA (82). The expression of the SigB regulon was also repressed in the ΔsipW strain, which produces but does not secrete TasA. Therefore, the activation of stress genes by TasA could be related to secretion stress. Overall, our results indicate that CssR/CssS activates Spo0A, antagonizing the effect of TasA and increasing matrix production. Furthermore, secretion of TasA may have a previously undescribed role in activation of the stressosome, a signal integration hub that leads to activation of the alternative sigma factor B (σB). This sigma factor controls transcription of about two hundred genes involved in the general stress response and tolerance to antibacterial agents (83), and TasA secretion results in the increased expression of genes related to stress tolerance (fig. S13).


In this study, we described a role for motility in bacterial biofilms and identified events that maintain a motile subpopulation at the single-cell level. It has long been postulated that flagellar motility has to be suppressed during biofilm development and plays an important role only during the early stages for surface approaching and sensing, or in the very late stages, when biofilm dispersal commences (11). Numerous mechanisms by which motility is repressed simultaneously with the activation of ECM production have been described in various bacterial species. In B. subtilis, the presence of ECM was suggested to induce the transitioning of motile cells toward ECM production by a positive feedback loop (37). Flagella rotation is impaired in the ECM-rich environment, and this disturbance acts as a mechanosensory pathway stimulating ECM production (38, 39). Moreover, the vast majority of bacteria (with the exception of a few robust swarmers such as Proteus mirabilis and Paenibacillus vortex) is incapable of moving using flagellar motility on high-friction solid surfaces, such as those on which biofilms form (43, 44). Furthermore, deletion of the gene encoding flagellin, hag, has no notable effect on biofilm morphology in B. subtilis grown in isolation (65). For these reasons, the possible role of motility in biofilm development and function has been largely ignored.

We found that motility was essential for the ability of B. subtilis biofilms to engulf and kill colonies of B. simplex. The killing mechanism depends on a surfactant (surfactin) and protein toxins [Sporulation Delaying Protein (SDP) and Sporulation Killing Factor (SKF)] with poor diffusibility in agar (65, 84), and the ability to move toward and engulf the foreign colony enables B. subtilis to deliver these molecules more efficiently. We confirmed the ability of B. subtilis to move over high-friction surfaces and engulf objects placed in proximity to the colony, in a similar manner to the engulfment of B. simplex colonies (Fig. 1). In mutant strains with compromised motility, this ability was impaired. Flagella have been suggested to have a structural role in some bacterial biofilms (85); however, a mutant lacking the flagellar motor subunits MotA and MotB, which have intact flagella that cannot rotate, showed a similar phenotype to that of a mutant lacking the flagellar structural protein flagellin (fig. S1). On a surface, three major factors restrict flagellar motility: surface friction, surface tension, and lack of water (43). In biofilms, exopolysaccharides can attract water, and surfactants can reduce surface friction and tension. Thus, the ability to move on top of a high-friction surface might be enhanced by the formation of a biofilm.

In the natural ecosystems of many bacterial biofilms, such as B. subtilis in the soil, movement and engulfment of obstacles could aid the bacterium in outcompeting other species for access to nutrients and colonization of new niches. Although the majority of cells in a biofilm contributes to ECM production during biofilm development (53, 78, 86), it seems likely that motility would be maintained in a subpopulation of the biofilm cells to enable movement. We found that the biofilm colonies of mutants lacking the ECM protein TasA had a lower percentage of motile cells, resulting in a notable reduction in overall colony motility (Fig. 2). This suggests that TasA could be a factor that maintains the motile cell subpopulation within a B. subtilis biofilm. Single-cell analysis revealed a mechanism by which TasA could preserve the motile cell subpopulation. TasA acted locally and stimulated ECM-producing cells to switch back to the motile state, with ΔtasA cells remaining in the biofilm state much longer than wild-type cells (Figs. 3, A and B, and 4). Within the biofilm colony, the emergence of motile cells from preexisting matrix-producing chains would allow a more uniform distribution of motile cells throughout the biofilm compared to the emergence of two spatially segregated subpopulations. This could promote collective migration of biofilm cells because it requires colocalization of both flagellated cells and matrix-producing cells.

It was previously suggested that entry into the biofilm state is a stochastic process, whereas reverting back to motility is timed (87). The timer was suggested to be a passive production-dilution mechanism, in which the expression of transcription factors is stopped at a certain moment, and then they are diluted until reaching a low threshold that then allows motility to be reactivated. Here, we revealed an upstream layer of regulation—the TasA protein itself acting as an essential signal for exiting the biofilm state. Thus, the frequent switching from ECM production to motility observed by Norman et al. (88) could be explained by the presence of TasA in the growing chains. tasA deletion had no additional influence on the number of motile cells in cells lacking motility-biofilm switch master regulators and thus unable to enter the biofilm state. In addition, suppressor mutations that spontaneously appeared in ΔtasA colonies targeted known regulators of the switch (Fig. 5). Together, our findings suggest that TasA acts as a signal upstream of the motility-biofilm switch (fig. S14) to increase the switching from matrix production back to motility (movie S3).

Last, a nonsynonymous mutation was found in the histidine kinase domain of the histidine kinase CssS, part of the CssR/CssS two-component system. Because cssRS deletion increased the expression of motility genes and inhibited the expression of ECM genes, we suggest that the CssR/CssS two-component system is a previously unrecognized regulator of the switch that can antagonize the action of TasA (Fig. 6). It is tempting to speculate that CssR/CssS senses high amounts of secreted TasA and is a part of a discrete pathway, independent of KinA to KinD and DegU.

When considering a developmental model for biofilm formation, it is tempting to speculate that the bacterial ECM is involved in the regulation of genetic programs in designated subpopulation of cells in the biofilm (74). It is clear that cell migration, tissue morphogenesis, and homeostasis in multicellular eukaryotes depend on cell-ECM interactions (66). We herein describe a similar role for an ECM protein in changing the decision-making processes of bacterial cells within a biofilm and demonstrate how this ECM-derived cue maintains an essential subset of motile cells within the biofilm population.

To conclude, we suggest that biofilms are motile—rather than sessile—entities. Numerous pathogenic biofilm-forming species, such as Pseudomonas aeruginosa, are thought to block flagellar motility while in a biofilm state (60). In this work, we present evidence of collective motility, playing a fundamental role in biofilm spreading, at least for the model organism B. subtilis. One worrying scenario is that biofilms of virulent bacteria can spread to various tissues in the body, relying on collaboration between the ECM producers and flagellated motile cells. Fuller understanding of this physiological property of bacterial biofilms may lead to the development of novel antibiofilm agents targeting biofilm collective motility.


Bacterial strains and strain construction

All experiments were performed with B. subtilis NCIB 3610 (31). Laboratory strains of B. subtilis (PY79) and Escherichia coli (DH5α) were used for cloning purposes (31). Lists of bacterial strains and primers are provided in table S1 and in (31, 51, 53, 55, 61, 62, 69, 73) and table S2, respectfully.

For cloning, linearized PCR products were first transformed into PY79 (89), and then the genomic DNA of the transformed strain was transformed to NCIB 3610 (90). Linearized plasmids were transformed directly to NCIB 3610. Deletion mutations were generated by long-flanking homology PCR mutagenesis (91). All suppressor mutants were verified by whole-genome sequencing.

Plasmids for the generation of GFP reporter strains or tapA-sipW-tasA overexpression construct were constructed as follows: (i) pYC121, which contains a functional GFP gene downstream to a cloning site and a chloramphenicol resistance gene (92), was used for the construction of amyE::PcomGA-gfp and amyE::PtapA-gfp reporter strains. PCR fragments were amplified from NCIB 3610 chromosomal DNA, using primers with the suitable restriction sites for ligation into the plasmids. (ii) hag promoter (41) was amplified with primers containing EcoRI and HindIII restriction sites and integrated into pYC121 upstream to GFP. The resulted plasmid was then cut with EcoRI and BamHI and integrated into pDR183 to create a lacA::Phag-gfp(mls) reporter strain. (iii) pIK76 was created by inserting the EcoRI and BamHI fragment of pDR111, which contains the hyperspank promoter upstream to a cloning site, the lacI repressor gene (41) into the plasmid pDR183 that enables integration into the lacA locus (93). pIK76 was used for the construction of tapA-sipW-tasA overexpression strain. PCR fragment, containing the tapA-sipW-tasA operon, ribosome binding sites, and SalI and SphI restriction sites, was integrated into pIK76. Plasmid pDFR6, containing the open reading frame of TasA without the signal peptide or the stop codon cloned into pET22b (Merck), was generated as described previously (33).

The ligated plasmids were then transformed into E. coli DH5α, and ampicillin-resistant colonies were selected and confirmed by sequencing. The GFP reporter or overexpression plasmids were then integrated into the neutral amyE or lacA locus of the laboratory strain PY79 by transformation, as described above, and selected for chloramphenicol, spectinomycin, or MLS [Macrolide (Erythromycin), Lincosamide and Streptogramin] resistance. Extracted genomic DNA of the transformed strains was transformed to NCIB 3610 as described above.


The strains were routinely manipulated in LB (Difco) or MSgg medium [5 mM potassium phosphate, 100 mM MOPS (pH 7), 2 mM MgCl2, 50 μM MnCl2, 50 μM FeCl3 for liquid medium or 125 μM FeCl3 for solid medium, 700 μM CaCl2, 1 μM ZnCl2, 2 μM thiamine, 0.5% glycerol, 0.5% glutamate, Threonine (50 μg ml−1), Tryptophan (50 μg ml−1), and PhenylAlanine (50 μg ml−1)] (31). MSgg plates [1.5% Bacto Agar (Difco)] were prepared on the day of the experiment and air-dried under a laminar flow hood for 40 min before inoculation.

For cloning purposes, LB or LB agar was used, with antibiotics at the following concentrations: ampicillin (100 μg ml−1; AG Scientific), kanamycin (10 μg ml−1; AG Scientific), chloramphenicol (10 μg ml−1; AMRESCO), tetracycline (10 μg ml−1; AMRESCO), spectinomycin (100 μg ml−1; Tivan Biotech), and erythromycin (1 μg ml−1; AMRESCO) + lincomycin (25 μg ml−1; Sigma-Aldrich) for MLS.

Biofilm assays

Cells from a single colony isolated on LB plates were grown to mid-logarithmic phase in a 3-ml LB broth culture (4 hours at 37°C with shaking). Then, a drop was spotted on solid MSgg medium. For engulfment experiments, colonies were inoculated next to a disc placed at a distance of 0.3 or 0.5 cm. Plates were incubated at 30°C for the time period indicated in the legend for each figure. For floating biofilms (pellicles), cells from mid-logarithmic phase were diluted 1:100 into 3 ml of liquid MSgg and grown in six-well plates, at 30°C, without shaking, for 24 hours.

Images of brightfield and GFP signal intensity for colonies were obtained with a Stereo Discovery V20 microscope with Objective Plan Apo S 0.5× FWD 134 mm or Apo S 1.0× FWD 60 mm (Zeiss) attached to an Axiocam camera. For microscopy, cells were resuspended in phosphate-buffered saline (PBS) and washed twice, and images were taken with an Axioplan 2 microscope (Zeiss) equipped with a high-resolution microscopy Axiocam camera, at ×100 magnification. Data were captured and analyzed using ZENpro AxioVision suite software (Zeiss, Oberkochenm, Germany).

Disc engulfment assays

The quantification of the extent of engulfment was performed as follows: for each colony, to calculate the percentage or the disc circumference covered by the biofilm cells, we measured the intensity of brightfield image at seven points, evenly spaced around the disc. The points were consistent for all the colonies analyzed, with minimal manual adjustments to compensate for false positive results resulting from the slightly irregular shape of the disc, when needed. We also measured the intensity of the opposite side of the colony as a positive control and the intensity of the agar plate next to the disc as background. After subtraction of background, a threshold level was unbiasedly calculated from the positive control measurement. Each measurement was scored as either negative (below the threshold) or positive (above the threshold). Last, for each colony, the fraction of positive measurements out of all measurement was calculated. The results presented are the distribution of measurements for at least 12 colonies for each strain and time point, over three independent experiments. Data were analyzed using AxioVision suite software (Zeiss).

Flow cytometry

B. subtilis biofilms were inoculated as described above and incubated for the time period indicated in the legend for each figure. Biofilms were then scraped from the plate surface and separated into single cells using mild sonication, as previously described in (9499). Samples were fixated in 4% paraformaldehyde (Electron Microscopy Sciences) and kept at 4°C until the measurement. Samples were measured using an LSR II cytometer (Becton Dickinson, San Jose, CA, USA) operating a solid-state laser at 488 nm. GFP intensities were collected by 505 LP and 525/50 BP filters. For each sample, 106 events were recorded and analyzed for GFP intensities. The autofluorescence level was determined in each experiment by measuring a biofilm sample from a nonfluorescent strain of the same genetic background. Then, the distribution of GFP intensities was analyzed using a custom MATLAB code.

Growth and fluorescence measurements of shaking cultures

Cells from a single colony isolated on LB broth plates were grown to mid-logarithmic phase in a 3-ml LB broth culture (4 hours at 37°C with shaking) and were diluted to reach equal optical density (OD). Cells were rediluted 1:100 in 150 μl of liquid MSgg medium per well in a 96-well microplate (Thermo Fisher Scientific). Cells were grown with agitation at 30°C in a microplate reader (Synergy 2, BioTek), and the OD at 600 nm (OD600) and GFP fluorescence (485/20 and 528/20 filter set) were measured every 15 min.

Fluorescence microscopy

A single colony was collected into PBS, gently sonicated, and washed in PBS. Next, it was gently centrifuged, resuspended in 10 μl of 1-43 (Thermo Fisher Scientific) membrane stain (20 μg ml−1). The cells were then incubated for 10 min at room temperature, washed twice with PBS, centrifuged, and resuspended in 10 μl of PBS. Then, the cells were placed on a microscope slide and covered with a poly-l-lysine–coated coverslip (Sigma-Aldrich). Cells were visualized and photographed using an Axioplan 2 microscope (Zeiss) equipped with a high-resolution microscopy Axiocam camera, as required. Data were captured using AxioVision suite software (Zeiss).

Time-lapse fluorescence microscopy

Cells from a single colony isolated on LB plates were grown to mid-logarithmic phase in a 3-ml LB broth culture (4 hours at 37°C with shaking). For the strain carrying protein fusion TasA-mCherry and transcriptional reporter Phag-GFP, cultures were diluted 1:1000 in 16 ml of MSgg medium in 60-mm petri dishes and grown for 24 hours at 23°C standing cultures. Then, cultures were mixed and 150 μl was loaded into CellASIC B04 microfluidic plates (ONIX Microfluidic Platform, Merck). The Phag-GFP–expressing cells were trapped in the thinnest area of the plate (number 5, 0.7 μm), whereas the chained and thicker TasA-mCherry–expressing cells were trapped in wider areas (number 1, 2.3 μm; number 2, 1.3 μm).

For the strain carrying transcriptional reporters Phag-GFP and PtapA-CFP, cultures were diluted 1:1000 in MSgg medium and grown for 8 hours at 30°C with shaking. Then, cultures were diluted 1:10 in distilled water, and a 10-μl drop was placed on MSgg pads, incubated until all liquid was absorbed, covered with coverslips, and sealed completely with Valap [vaseline, lanolin, and paraffin, 1:1:1 (w/w/w)] to avoid evaporation.

Images were collected using a Nikon Ti-E with a Hamamatsu Flash 4.0 V2 camera, using the following excitation (Ex.) and emission (Em.) filters: GFP: Ex. 485/25, Em. 535/50; mCherry: Ex. 560/32, Em. 632/60; CFP: Ex. 434/17, Em. 470/28 to 25. Dichroic mirrors 69002bs, or 69008bs (Chroma), were used for GFP/mCherry and GFP/CFP imaging, respectively. Analysis was performed manually, using ImageJ Fiji distribution (100, 101).

An event of motility activation in a chain was defined as one or more of its cell progeny reaching intensity levels in the GFP channel above a threshold defined by nonfluorescent cells. Cells that were motile in the beginning of the time course were not considered for analysis. The number of chains in which an event occurred, or did not occur, during the time course was manually counted. Chain length measurements were performed by manually segmenting the cells at the beginning and the end of the time course. PtapA expression was defined as intensity levels in the CFP channel above a threshold defined by nonfluorescent cells. Length of segments in which PtapA was expressed was measured at t = 4.5 hours. For analysis of chains expressing PtapA-CFP reporter, a similar number of chains for the wild-type and ΔtasA strains at the beginning of the time lapse were counted and followed for about four generations of growth (1.15 ± 0.07 hours and 1.16 ± 0.04 hours for wild type and ΔtasA, respectively). Here, the total length of all cells and chains in the field of view was used as an approximation of cell number, because it is difficult to identify single cells inside a chain with confidence.

Osmolarity and viscosity measurements

Different weights of dextran, glucose, galactose, xylose, and polyethylene glycol, as indicated in the legend for each figure, were dissolved in the MSgg growth medium at the same concentrations that were used for the growth and fluorescence measurements (ranging from 0.2 to 10 weight %). The osmolarity of these polymer solutions was measured using an “Advanced Instruments Freezing Point Osmometer” (model 3300). The osmolarity was determined by subtracting the measured baseline corresponding to the MSgg solvent growth medium. The zero-shear viscosities of the polymer solutions were measured in a stress-controlled rheometer (TA Instrument DHR-3) using cone and plate geometry.

Complementation experiments

For complementation in trans, six-well plates separated into two compartments by Millicell Hanging Cell Culture Insert, PET 0.22 μm (Merck) were used. Each well contained 3 ml of MSgg medium. The strain to be complemented (ΔtasA) was inoculated below the filter, whereas the complementing strain was inoculated above. Medium with no bacteria above the filter served as control. After 24 hours, the filter was carefully removed, and the cells from the lower compartment were collected, washed twice in PBS, and analyzed by fluorescence-activated cell sorter (FACS).

For protein complementation, the medium was supplemented with purified TasA protein (final concentration of 20 μg/ml). TasA was expressed and purified as previously described (82) with some changes. BL21(DE3) E. coli competent cells were used for protein expression and purification. Induction of protein expression was performed in 500 ml of LB supplemented with ampicillin. The culture was incubated at 37°C until an OD600 of 0.7 to 0.8 was reached. Protein expression was induced with 1 mM isopropyl-β-d-1-thiogalactopyranoside (IPTG) and incubated overnight at 30°C with shaking to promote the formation of inclusion bodies. The next day, cells were pelleted by centrifugation (5000g for 15 min at 4°C) resuspended in buffer A [50 mM Tris and 150 mM NaCl (pH 8)] and then centrifuged again. Cell pellets were kept at −80°C until purification or processed immediately after 15 min. Cells were resuspended in buffer A, sonicated on ice in a Branson 450 digital sonifier (3 × 45 s, 60% amplitude) and centrifuged (15,000g for 60 min at 4°C). The pellet, mainly consistent of inclusion bodies, was recovered. The pellet was resuspended in buffer A supplemented with 2% Triton X-100 and incubated at 37°C with shaking for 20 min to ensure solubilization of cellular debris. The suspension was then centrifuged (15,000g for 10 min at 4°C), and the supernatant was discarded. The pellet was washed three times with buffer A, resuspended in denaturing buffer (50 mM Tris, 500 mM NaCl, and 6 M GuHCl), and incubated at 60°C with shaking overnight until complete solubilization of the inclusion bodies. The solution was clarified via sonication on ice (3 × 45 s, 60% amplitude) and centrifugation (15,000g for 1 hour at 16°C). Last, the inclusion bodies solution was then passed through a 0.45-μm filter before affinity chromatography. Recombinant TasA was purified using an ÄKTA Start fast protein liquid chromatography system (GE Healthcare). Soluble inclusion bodies were loaded into a HisTrap HP 5-ml column (GE Healthcare) previously equilibrated with binding buffer [50 mM Tris, 0.5 M NaCl, 20 mM imidazole, and 8 M urea (pH 8)]. The elution of the protein was performed isocratically with elution buffer [50 mM Tris, 0.5 M NaCl, 500 mM imidazole, and 8 M urea (pH 8)]. The eluted protein solution was loaded into a HiPrep 26/10 desalting column (GE Healthcare) to remove the urea and the imidazole, and the buffer was exchanged to 20 mM tris and 50 mM NaCl to perform the corresponding experiments.

Evolution experiment, whole-genome sequencing, and identification of mutations

Colonies of ΔtasA were inoculated as described above and incubated for 5 days. Fluorescent protrusions from the edge of the colony were isolated on LB for single colonies. Liquid cultures of single colonies were frozen and restreaked on LB, and the biofilm on solid surface assay was performed as described above for phenotype validation.

Genomes of nine representative strains, in addition to four ancestral strains, were sequenced. DNA was extracted using a DNeasy blood and tissue kit (QIAGEN). Libraries were generated using a Nextera XT DNA sample preparation kit (Illumina). Sequencing of 100–base pair single-read reads (50 cycles) was performed on an Illumina HiSeq 2500 sequencer (Illumina), using TruSeq rapid mode run reagent kits: two TruSeq Rapid SBS Kits HS (Illumina) and TruSeq Rapid SR Cluster Kit (Illumina).

Sequencing reads were aligned separately for each sample to the reference genome of B. subtilis NCIB 3610 (GenBank: NZ_CM000488) that was downloaded from National Center for Biotechnology Information. The reads were aligned using Novoalign 2.08.01 (Novocraft Technologies Sdn Bhd, with the default parameters and [−r Random]. Detection of mutations (mismatches and insertions) was performed by comparing the alignments of each sample to the alignments of the ancestor B. subtilis sample that was also sequenced. Genomic positions that consistently differed between both alignments (>70%) were recorded as mutations. Genomic positions with no aligned reads but with aligned reads in the ancestor sample were recorded as deletions. Mutations in additional strains were identified using PCR and Sanger sequencing.

Swimming assay

MSgg plates solidified with 0.25% agar were prepared a day before the experiment and dried overnight at room temperature. Cells from a single colony isolated on LB plates were grown to mid-logarithmic phase in a 3-ml LB broth culture (4 hours at 37°C with shaking) and diluted to reach equal OD. Then, a 1-μl drop was spotted on the center of the plate. Plates were incubated at 30°C for 10 hours. Images were obtained with a Nikon COOLPIX P510 camera (Nikon). Data were analyzed using AxioVision suite software (Zeiss).

RNA extraction and library preparation

Biofilms colonies were grown on MSgg solid medium for 24 hours, and colonies (n = 3) were collected for RNA extraction. The samples were snap-frozen in liquid nitrogen and stored at −80°C until extraction. Frozen pellets were lysed by incubation with lysozyme (20 mg/ml) for 10 min at 37°C, and RNA was extracted using TRI Reagent (Sigma-Aldrich, T9424) according to the manufacturer’s instructions. RNA levels and integrity were determined by Qubit RNA BR Assay Kit (Life Technologies, Q10210) and TapeStation, respectively. Next, samples were treated with TURBO DNase (Life Technologies, AM2238). A total of 5 μg of RNA was treated with Illumina Ribo-Zero rRNA Removal Kit (Illumina, RZMB123), according to the manufacturers’ protocols. RNA quantity and quality after depletion were assessed as described above. RNA-sequencing (RNA-seq) libraries were contracted with NEBNext Ultra Directional RNA Library Prep Kit (New England Biolabs, E7420) according to the manufacturer’s instructions. Libraries concentrations and sizes were evaluated as above and were sequenced as multiplex indexes in one lane using the Illumina HiSeq 1500 platform.

Bioinformatic analysis of RNA-seq data

For each of the sequenced libraries, we randomly down-sampled 4 million reads for analysis, using seqtk, to avoid saturation. Reads were trimmed from their adapter with cutadapt and aligned to the B. subtilis genome (subsp. subtilis str. NCIB 3610, NZ_CM000488.1) with Bowtie 2 version (102). The number of uniquely mapped reads per gene was calculated with HTSeq (103). Normalization and testing for differential expression were performed with DESeq2 version 1.16. A gene was considered to be differentially expressed if its normalized mean read count ≥ 50, fold change ≥ 2, and adjusted P < 0.05. Using these criteria, 300 genes were found to be differentially expressed in the ∆cssRS transcriptome and 357 in the ∆tasA transcriptome. A significant number of genes, 72, were found to be differentially expressed in both transcriptomes (P < 1.8 × 1010, hypergeometric test). DAVID (Database for Annotation, Visualization and Integrated Discovery) Bioinformatics (104) was used to performed gene annotation enrichment analysis.


Fig. S1. Flagellar motility promotes colony expansion and engulfment of foreign objects.

Fig. S2. Motile cells move forward during engulfment, followed by matrix producers.

Fig. S3. Expression of the flagellin GFP reporter in colonies over time.

Fig. S4. Formation of TasA-mCherry fibers in the microfluidics system.

Fig. S5. Osmotic pressure does not affect growth rate or restore hag expression in the ΔtasA mutant.

Fig. S6. Complementation of a tasA mutant by tapA-sipW-tasA expression in trans.

Fig. S7. ECM-expressing cells form chains in ΔtasA.

Fig. S8. TasA stimulates the reversal to motility by antagonizing tapA expression.

Fig. S9. TasA stimulates the reversal to motility independently of DegU or KinD.

Fig. S10. The effects of KinA to KinC on TasA-dependent motility expression.

Fig. S11. The CssR/CssS two-component system.

Fig. S12. Genes differentially expressed in both ΔtasA and ΔcssRS.

Fig. S13. Stress response genes regulated by TasA and CssR/CssS.

Fig. S14. TasA stimulates the transition from a matrix-producing state back to the motile state.

Table S1. Bacterial strains.

Table S2. Primers.

Movie S1. TasA fiber formation.

Movie S2. Induction of motility occurs after tasA expression.

Movie S3. TasA stimulates reversal to motility after entering the biofilm state.

Data file S1. Gene expression analysis.


Acknowledgments: We thank S. Rubinstein and S. Srinivasan from the School of Engineering and Applied Sciences, Harvard University for the osmolarity and viscosity measurements. We thank E. Oldewurtel for the help with the microscopy experiment. We thank Z. Bloom-Ackermann for the construction of a reporter strain. We thank R. Rotkopf from the Department of Life Sciences Core Facilities for guidance with the statistical analysis. We thank L. Chai (The Hebrew University) for the gift of purified TasA. Funding: The Kolodkin-Gal lab is supported by the Israel Science Foundation grant number 119/16, France-Israel grant number 3-13021 to I.K.-G. and S.v.T., Bi-institutional Kamin grant, Research Innovation Authority awarded 2018, Israel Ministry of Science–Tashtiot (Infrastructures) 123402, and Infrastructures for Life Sciences and Medical Science (123402). I.K.-G. is supported by an internal grant from the Estate of Albert Engleman and by a research grant from the Benoziyo Endowment Fund for the Advancement of Science and a recipient of Rowland and Sylvia Career Development Chair. Author contributions: N.S., D.R., R.J., K.Y.-G., and Q.H. performed the experiments. N.S., S.D., A.K.-P., R.J., R.H., and T.O. analyzed the data. S.v.T. and I.K.-G. contributed the reagents. J.C.-A. provided materials and reagents. N.S., G.R., and I.K.-G. designed the experiments. S.v.T. designed the microscopy experiments and image analysis. S.v.T., N.S., A.K.P., and I.K.-G. wrote the paper. Competing interests: The authors declare that they have no competing interests. Data materials availability: The RNA-seq data have been deposited into Gene Expression Omnibus repository ( with the accession number GSE138015. All other data needed to evaluate the conclusions in this paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article