Genes governing swarming in Bacillus subtilis and evidence for a phase variation mechanism controlling surface motility



Undomesticated strains of Bacillus subtilis, but not laboratory strains, exhibit robust swarming motility on solid surfaces. The failure of laboratory strains to swarm is caused by a mutation in a gene (sfp) needed for surfactin synthesis and a mutation(s) in an additional unknown gene(s). Insertional mutagenesis of the undomesticated 3610 strain with the transposon mini-Tn10 was carried out to discover genes needed for swarming but not swimming motility. Four such newly identified swarming genes are reported, three of which (swrA, swrB, and efp) had not been previously characterized and one of which (swrC) was known to play a role in resistance to the antibacterial effect of surfactin. Laboratory strains were found to harbour a frameshift mutation in the swrA gene. When corrected for the swrA mutation, as well as the mutation in sfp, laboratory strains regained the capacity to swarm and did so as robustly as the wild strain. The swrA mutation was an insertion of an A:T base pair in a homopolymeric stretch of eight A:T base pairs, and readily reverted to the wild type. These findings suggest that the swrA insertion and its reversion take place by slipped-strand mispairing during DNA replication and that swarming motility is subject to phase variation.


Swarming motility is a social form of motility that enables bacteria to travel rapidly and en masse atop solid surfaces (Henrichsen, 1972). Like swimming motility, swarming requires the production of functional flagella, but many factors appear to distinguish the two forms of movement. Whereas swimming occurs in liquid medium, swarming takes place on surfaces and in addition requires the production of a surfactant to reduce surface tension (Matsuyama et al., 1992; Gygi et al., 1995; Toguchi et al., 2000). Swarming cells dramatically increase the number of flagella on the cell surface and increase in size relative to their swimming counterparts (McCarter et al., 1988; Gygi et al., 1997). Importantly, microscopic observation of swarming cells reveals that locomotion appears to depend on direct cell-to-cell contact, and cells organize into multicellular rafts while actively moving (Morrison and Scott, 1966). A variety of bacteria have been discovered that exhibit swarming motility, but the genetic requirements of swarming and how it differs from swimming remain poorly understood (Harshey, 1994).

We have recently begun an investigation of swarming in the spore-forming bacterium Bacillus subtilis as a model system for studies of surface motility in Gram-positive bacteria. Previous genetic studies established that swarming motility in B. subtlis is flagellum based and depends on the production of the lipopeptide surfactant, surfactin (Kearns and Losick, 2003). Of particular interest was the finding that undomesticated (wild) strains of B. subtilis exhibited a robust swarming phenotype but laboratory strains were relatively immotile on surfaces, although completely unimpaired in their capacity to swim. We found that the failure of laboratory strains to swarm is due in part to the presence of a mutation in a gene (sfp) needed for surfactin biosynthesis (Nakano et al., 1992). However, the sfp mutation is not sufficient to explain the swarming defect of laboratory strains as the application of purified surfactin or the correction of the sfp mutation (as we show here) was unable to restore surface motility. These observations indicated that laboratory strains carry one or more additional mutations that contribute to the swarming defect and suggested that additional genetic determinants for swarming motility remained to be discovered in B. subtilis.

To discover such determinants, we carried out insertional mutagenesis of the undomesticated 3610 strain using the transposon mini-Tn10 and discovered four genes previously unidentified as being required for swarming but not swimming. Furthermore, we were able to completely account for the swarming defect in laboratory strains as due to mutations in sfp and the novel gene swrA. The swrA mutation reverted at a high frequency suggesting that B. subtilis swarming motility is subject to phase variation.


Discovery of previously uncharacterized genes involved in swarming motility

To discover genes involved in swarming motility, we carried out insertional mutagenesis in the undomesticated B. subtilis strain 3610 using the transposon mini-Tn10. Approximately 12 000 colonies from eight independent transposon libraries were screened for the inability to swarm over solid surfaces while remaining proficient for swimming in liquid media. A total of 131 mutants were isolated but many of these mutants contained identical transposon insertions and hence were likely to be siblings. Leaving these aside, our collection consisted of 27 different insertions located within 14 genes that abolished or impaired swarming motility (Fig. 1A; Table 1). That the transposon insertions were responsible for the observed defects in swarming was confirmed for representative insertions for all 14 genes (except for those involved in surfactin productin). This was carried out by phage SPP1-mediated transduction experiments in which the insertions and the mutations causing the swarming defect were found to be inseparable. Also, each of the genes identified by insertional mutagenesis was separately replaced in the chromosome of an otherwise wild-type strain of 3610 with an antibiotic resistance gene. In most cases, this resulted in a mutant with a swarming defect, confirming that the gene in question was indeed required for normal swarming behaviour. As discussed below, in some cases the effect of the transposon insertion was found to be the result of a polar effect on the expression of a downstream gene.

Figure 1.

Transposon insertions that impair swarming motility.
A. Chromosome map indicating sites of transposon insertions that impaired swarming but not swimming.
B. Phenotype of transposon insertions on swarm agar. Each image is a top down view of a swarm agar Petri plate that was inoculated centrally and incubated at 37°C for 18 h. The plates were filmed against a black background such that zones of bacterial colonization appear white and uncolonized agar appears black. WT indicates the wild-type 3610 strain.
C. Swarm expansion assay of swarming mutants: (closed circles) wild-type 3610 and (open triangles) comP::TnΩ1028. Some lines represent the behaviour of more than one mutant. The open diamonds represent the zones of consolidation behaviour exhibited by a mutant harbouring an insertion in rrnB-16S; similar results were obtained for insertions in yabR and ymfI. The open circles represent the non-swarming behaviour of a mutant harbouring an insertion in swrA; similar results were obtained for insertions in: cheC, cheD, efp, srfAA, srfAB, srfAC, swrB, swrC and ylxH. All values are the average of three replicas.

Table 1. . Sites of transposon insertion that abolish or impair swarming motility.
insertion site
cheC Chemotaxis proteinTnΩ1064GTTATGACA
cheD Chemotaxis proteinTnΩ1045GGCTAAGTA
comP ComX sensor kinaseTnΩ1028TATACTAAC
efp (yqhU) Elongation factor PTnΩ1124CATTAAACA
rrnB-16S Ribosomal RNA-16STnΩ1146TAAAGATGA
srfAA Surfactin synthetaseTnΩ1043GATCGGGTG
srfAB Surfactin synthetaseTnΩ1044TGCTTAGTG
srfAC Surfactin synthetaseTnΩ1122TGCCCGCAT
swrA (yvzD) Novel geneTnΩ1026CACTAAGTC
swrB (ylxL) Novel geneTnΩ1107TACTTGAAG
swrC (yerP) Similar to multidrug efflux pumpTnΩ1113GGTCAAGCA
yabR Similar to ribosomal protein S1 mRNA binding domainTnΩ1078GGCTCAACC
ylxH Similar to ATPase involved in chromosome partitioning (minD)TnΩ1167TGTTTTAAC
ymfI Similar to short chain alcohol dehydrogenaseTnΩ1029AGCTCAGCA

Some of the insertions revealed genes that contributed to, but were not required for, swarming, as indicated by their intermediate phenotype. Thus mutants harbouring insertions in yabR and ymfI produced zones of consolidation that resembled terraces on the swarm agar surface (Fig. 1B). This pattern was produced because the mutants initiated swarming normally but the entire population periodically stopped moving, and then reinitiated surface migration after a pronounced lag period (Fig. 1C). Evidently, yabR and ymfI contribute to proper coordination among cells of the swarming population but are not absolutely required for surface migration.

A third insertion, which produced zones of consolidation (TnΩ1146), was found to disrupt one of the 10 B. subtilis genes that code for 16S ribosomal RNA (rRNA) (Jarvis et al., 1988). Southern blot hybridization using 23S rDNA as a probe revealed that the transposon was located in the rrnB operon (Fig. S1), which contains genes for 16S, 23S and 5S rRNAs, followed immediately downstream by an array of 21 tRNA genes (Green et al., 1985). However, it appears that the transposon insertion was polar on the expression of the downstream tRNA genes as an insertion/deletion mutant (DS408) in which the rrnB 16S and 23S rRNA genes were replaced with an antibiotic resistance determinant (which does not cause polar effects) was unimpaired in swarming motility (Fig. S1). The simplest interpretation of these results is that one or more rrnB-derived tRNAs are required for adequate translation of a gene(s) required for swarming but not for viability or swimming.

Some transposon insertions that were located within the fla/che operon also appeared to have an indirect effect on swarming motility. In B. subtilis, 31 genes involved in flagellum biosynthesis and chemotaxis are clustered in a single operon. At or near the extreme 3′ end of this operon is the gene (sigD) for the transcription factor σD, which is required for the expression of genes activated late in flagellum biosynthesis. Transposon insertions in three genes located upstream of sigD (the chemotaxis genes, cheC and cheD, and the previously uncharacterized gene ylxH) resulted in severe swarming defects but had little or no effect on swimming (Fig. 2A). The effects of the transposon insertions in cheC, cheD and ylxH appear to have been due, in whole or in part, to polar effects on the expression of sigD because null mutations of these genes created by replacement of their coding sequences with a drug resistance gene (which does not cause polar effects) had no effect on swarming (Kearns and Losick, 2003). Also, the swarming defects of cheC, cheD and ylxH insertions could be partially [in the case of the ylxH insertion (ylxH::TnΩ1167)] or completely bypassed by a construct in which a copy of the sigD gene was joined to a copy of the promoter for the fla/che operon (PflgB), and the resulting transcription fusion was inserted into the chromosome at the amyE locus (Fig. 2B). We conclude that cheC, cheD and ylxH are not required for swarming motility. Evidently, impaired synthesis of σD preferentially interferes with surface migration.

Figure 2.

Transposons that exhibit a polar effect on sigD abolish swarming motility.
A. The arrangement of genes in the fla/che operon and the sites of transposon insertions that blocked swarming motility. Twenty-two genes located between flgB and ylxH are omitted as indicated by the slashes. Grey inverted triangles indicate the sites of transposon insertions. Brackets indicate the sites at which chromosomal DNA has been deleted and replaced with an antibiotic-resistance determinant. Black arrows indicate promoters. Diagram is not to scale.
B. Some transposon insertions in the fla/che locus are polar on sigD. Each of the indicated transposon insertions introduced into strains harbouring a construct in which the flgB promoter region was fused either to the sigD gene (PflgB-sigD) or to the ylxL gene (PflgB-ylxL). The results of each cross were as follows: (–) no swarming (+) partial swarming, and (+ + +) wild type swarming. The following strains were used to generate the table: DS467, DS468, DS474, DS478, DS479, DS481, DS482, and DS483.

An insertion in the gene immediately downstream of the sigD gene, ylxL, also abolished swarming motility. In contrast to the behaviour of the insertions in cheC, cheD and ylxH, the effect of the insertion in ylxL could not be bypassed by the presence of the amyE::PflgB–sigD construct (Fig. 2B). The ylxL::TnΩ1107 insertion could be complemented, however, by a copy of the ylxL gene that had been fused to the promoter of the fla/che operon (PflgB) and inserted at the amyE locus. We conclude that the requirement for ylxL was not an indirect effect on the expression of sigD, and that the gene plays a direct role in surface motility. Interestingly, the ylxL::TnΩ1107 insertion could not be complemented by a construct in which the ylxL coding sequence and 500 base pairs of upstream DNA were inserted into the amyE locus (that is, without being joined to the flgB promoter) (data not shown). We infer from this observation that ylxL lacks its own promoter and that the gene is transcribed as part of the flagellum/chemotaxis operon, evidently representing the most downstream member of the operon.

Genes required for surfactin biosynthesis and export

Swarming motility was previously shown to require the production of the extracellular lipopeptide surfactant, surfactin, and therefore it came as no surprise that several of our insertions were located within the operon (srf) that governs surfactin biosynthesis (Fig. 1: srfAA, srfAB, srfAC; Nakano et al., 1991a; Kearns and Losick, 2003). In addition, the requirement for surfactin provided a simple explanation for the effect of two insertions located outside the srf operon, TnΩ1028 and TnΩ1150, that reduced the rate of swarming motility (Fig. 1C) and produced colonies with a perforated appearance (Fig. 1B). These insertions were located in comP, which encodes a sensor kinase that responds to the signalling peptide ComX by phosphorylating and thereby activating the response regulator ComA (Weinrauch et al., 1990; Magnuson et al., 1994). Among the targets of ComA∼P is the srf operon (Nakano et al., 1991b) and thus the decreased rate of swarming from the comP insertions could have been due to reduced levels of surfactin production. In keeping with this interpretation, the addition of purified surfactin restored wild-type levels of swarming motility to the comP mutants (data not shown) as it does for srf mutants (Kearns and Losick, 2003). We conclude that comP plays an indirect role in swarming by stimulating surfactin biosynthesis.

The involvement of surfactin in surface motility may also provide an explanation for the strong dependence of swarming on yerP (hereafter designated swrC). The swrC gene is predicted to encode a member of the AcrB-like family of multidrug resistance efflux pumps, which export a broad spectrum of hydrophobic substrates (Nikaido, 1998). A logical substrate for SwrC export is surfactin itself, and, indeed, swrC (yerP) had previously been identified as being required for resistance to the antibacterial activity of surfactin (Tsuge et al., 2001). SwrC-mediated export is not the only mechanism for surfactin secretion, however, as a mutant (DS441) in which the swrC gene was replaced by an antibiotic-resistance determinant failed to swarm and yet produced abundant surfactin on swarm plates. Why then was DS441 unable to swarm? We found that DS441 regained the capacity for surface motility when a second mutation (in srfAA) was introduced that blocked surfactin production (to create the double mutant DS516) and surfactin was exogenously provided to the swarm plates (Fig. 3). Our interpretation of these findings is as follows. The SwrC export pump contributes to the secretion of endogenously produced surfactin. In its absence, surfactin accumulates to excessive levels within the cells, and this intracellular accumulation somehow interferes with surface motility. The swrC srfAA double mutant (DS516) is able to swarm when surfactin is provided exogenously because extracellular surfactin is not deleterious to swarming and the srfAA mutation relieves the stress caused by the intracellular accumulation of surfactin. Consistent with our interpretation, we have found that a swrC mutant (DS441) readily gave rise to suppressor mutants (e.g. DS499) that are blocked in surfactin production and that, like DS516, exhibited surface motility only when surfactin was added to the swarm plate (Fig. 3).

Figure 3.

The requirement for swrC can be bypassed by a defect in surfactin biosynthesis. Swarm expansion assay depicting the behaviour of the srfAA mutant (RL2663; triangles), the swrC mutant (DS441; squares), the swrC suppressor (DS499; circles), and the swrC srfAA double mutant (DS516; diamonds) in the presence (closed symbols) and absence (open symbols) of 100 µg of surfactin. All values are the average of three replicas.

Newly identified swarming genes

Thus, from our original identification of 14 genes, we were left with a total of three previously uncharacterized genes (yvzD, ylxL and yqhU) that were strongly needed for swarming and in a manner that could not be readily attributed to an indirect effect or to an effect on surfactin production or export. Consistent with this interpretation, a srfAA mutant readily swarmed on LB fortified with 0.7% Eiken agar [an agar believed to have inherent surfactin- like properties; (Toguchi et al., 2000)], but mutants in yvzD, ylxL and yqhU remained non-swarming regardless of the type of agar used. The predicted products of two of the genes, yvzD (see below) and ylxL (see above), exhibited little or no similarity to other proteins in the databases and are hereafter designated swrA and swrB respectively. The yqhU gene has been annotated efp based on the sequence similarity to the gene for elongation factor P, which is involved in protein synthesis.

Laboratory strains are mutant for swrA

Armed with the knowledge that swarming by strain 3610 depends on several previously uncharacterized genes, we returned to the question of why laboratory strains are typically defective in surface motility. We were already aware that laboratory strains carry a mutation in a gene (sfp) required for production of surfactin. However, when laboratory strains that had been corrected for the sfp° mutation, such as EG371, an sfp+ derivative of the laboratory strain PY79, and DS281, an analogous sfp+ derivative of the laboratory strain 168, were inoculated on swarm agar, surfactin biosynthesis was visibly restored but the cells remained unable to swarm (Table 2). Interestingly, upon prolonged incubation (48 h), a population of cells emerged from non-swarming cells of EG371 and DS281 that were swarming-proficient and completely colonized the swarm agar plate. Isolates from these populations exhibited swarming behaviour that was similar to that exhibited by the undomesticated strain 3610 (Table 2). We conclude that the laboratory strains contain at least one genetic defect in addition to the sfp° mutation that renders them non-swarming and that this additional mutation(s) can readily revert or readily be suppressed by a second-site mutation.

Table 2. . The swrA gene is the site of a mutation and a reversion in laboratory strains.
  • a

    . presence (+ + +) or absence (–) of wild-type swarming behaviour.

PY79  sfp°
EG371PY79 sfp +
DS155EG371 sfp + spontaneous suppressor+ + +
168  sfp°
DS281168 sfp +
DS283DS281 sfp + spontaneous suppressor+ + +
3610  sfp + + + +
DS2153610 sfp + swrA::tet
DS231DS215 sfp + swrA::tet amyE::swrA3610+ + +
DS339DS215 sfp + swrA::tet amyE::swrAPY79
DS416DS215 sfp + swrA::tet amyE::swrADS155+ + +
DS340DS215 sfp + swrA::tet amyE::swrA168
DS456DS215 sfp + swrA::tet amyE::swrADS283+ + +

A clue to the location of one such mutation in laboratory strains came from transduction experiments with phage SPP1, in which a derivative of laboratory strain PY79 was the donor and strain 3610 was the recipient. These experiments indicated that a mutation that blocked swarming could be co-transduced with (18% co-transduction), and therefore was linked to a genetic marker (an antibiotic resistance determinant that had been inserted into a gene called yvzB) in the 10 o’clock region of the B. subtilis chromosome (Fig. 1A). Among our collection of 14 genes in which insertions caused swarming defects, only one, swrA, was linked to the transduced marker. To investigate whether laboratory strains harbour a mutant swrA gene, we cloned swrA including 500 base pairs of upstream DNA from the swarming-proficient strain 3610 (swrA3610) and introduced the gene into the amyE locus of the sfp+ derivatives of the laboratory strains PY79 (EG371) and 168 (DS281). In both cases, the presence of the amyE::swrA3610 construct was able to complement the defect in the laboratory strain and restore swarming to levels similar to that observed in the wild strain (Fig. 4, data not shown). These results suggest that PY79 and 168 are mutant for swrA and that the inability of laboratory strains to swarm is a consequence of defects in only two genes: sfp and swrA.

Figure 4.

Correction of the sfp mutation and complementation of the swrA mutation is sufficient to restore swarming behaviour to laboratory strains. Swarm expansion assay showing the behaviour of the wild type (3610; open circles), the sfp corrected laboratory strain (EG371; open squares), and the sfp corrected laboratory strain complemented with the swrA3610 allele (DS254; closed squares). All values are the average of three replicas.

As a further test of the idea that laboratory strains harbour a defect in swrA, we asked whether alleles of the gene from PY79 and 168 would function in 3610. To do this, we first constructed a null mutation of swrA in 3610 by replacing the swrA open reading frame with an antibiotic resistance determinant, creating strain DS215. The resulting mutation (swrA::tet) caused a severe swarming defect, similar to that exhibited by the transposon insertions (swrA::TnΩ1026, swrA::TnΩ1239 and swrA::TnΩ1248). We also demonstrated that the null mutation could be complemented by a cloned copy of the wild type gene using the amyE::swrA3610 construct described above (Table 2). The ability of amyE::swrA3610 to complement the swrA::tet mutation established that the effect of the null mutation was not due to a polar effect on the expression of a downstream gene. Next, we cloned swrA from PY79 (swrAPY79) and from 168 (swrA168) and introduced the genes into the amyE locus of DS215. Unlike swrA3610, neither swrAPY79 nor swrA168 was able to restore swarming to 3610 cells harbouring the swrA::tet mutation (Table 2), reinforcing the view that laboratory strains are mutant for swrA.

As indicated above, sfp+ derivatives of PY79 (EG371) and 168 (DS281) readily yield revertants or suppressor strains (such as DS155 and DS283 respectively) that swarm with high efficiency. We wondered whether these swarming-proficient derivatives were intragenic suppressors or revertants of the mutant swrA allele. Accordingly, we cloned swrA from DS155 and DS283 and introduced the genes into the amyE locus of DS215. The results show that both swrADS155 and swrADS283 fully complemented the swarming defect caused by the null mutation (Table 2). In toto, the results of the complementation experiments indicate that laboratory strains are mutant for swrA and that intragenic suppression or reversion had restored swrA to a functional state.

The swrA gene in laboratory strains contains a single base-pair insertion mutation

To investigate the nature of the mutation in swrA, we determined the sequence of the gene from the wild strain 3610, from the swarming-proficient suppressor strains, DS155 and DS283, and from the laboratory strains PY79 and 168. The swrA3610, swrADS155, and swrADS283 genes were identical to each other but differed from swrAPY79 and swrA168 at a single position: the presence of a single, A:T base pair insertion in a contiguous stretch of eight A:T base pairs (Fig. 5). It appears that swrA had acquired a single base pair insertion in laboratory strains and that the mutant gene had reverted to the wild-type (functional) sequence in strains that had regained the capacity to swarm.

Figure 5.

Laboratory strains contain a frameshift mutation in swrA. DNA sequence is in lower case. Amino acid sequence is in upper case one letter code and is shown above the DNA sequence for the wild type (3610) reading frame and below the DNA sequence for the laboratory strain (PY79) reading frame. Bold text indicates the site of the homopolymeric A:T stretch and the one A:T base pair insertion found in laboratory strains. Boxed sequence indicates a putative ribosome binding site. Underlined sequence indicates the site of the 9 bp duplication generated by the indicated mini-Tn10 insertion. Bent arrows indicate the positions at which fusions to lacZ were created. Straight arrows above the DNA sequence indicate positions at which site directed mutagenesis was used to change putative start codons to the sequence indicated in parentheses. The straight arrow below the DNA sequence indicates the position at which site directed mutagenesis was used to change a stop codon to the sequence indicated in parentheses. The DNA sequence of swrA in the laboratory strain 168 was identical to that for PY79 and the DNA sequence of swrA from laboratory strains that had undergone reversion (DS155 and DS283) was identical to that for 3610.

Revisiting the swrA open reading frame

As swrA (yvzD) was originally annotated based on the sequence of the non-functional laboratory strain allele harbouring an A:T insertion mutation, we decided to revisit the nature of the swrA open reading frame. To do so, we separately fused lacZ to the wild-type gene (swrA3610) in all three possible reading frames upstream of the TnΩ1026 insertion site (Fig. 5). The results show that only one fusion supported the production of β-galactosidase, thereby establishing the ‘0 frame’ that corresponds to the swrA protein-coding sequence (Fig. 6A). Two potential translational start sites with putative ribosome binding sites were identified upstream of the functional (0 frame) lacZ fusion. Both putative start codons were replaced with conservative codons using site-directed mutagenesis and fused to the 0 frame lacZ reporter. Mutation of only one of these codons abolished expression from the downstream reporter, establishing that TTG was the swrA start site (Figs 5 and 6B). Based on these results, we conclude that swrA encodes a 117 amino acid long protein with no recognizable motifs and which is not significantly homologous to any protein in the databases.

Figure 6.

The swrA insertion in laboratory strains generates a frameshift mutation.
A. Analysis of the swrA open reading frame. DNA from the indicated swrA allele was fused to one of three translational lacZ fusion reporters in each of the potential swrA reading frames (see Fig. 5). Each image is a top down view of a colony grown overnight on LB agar containing X-gal. The blue colour indicates the production of β-galactosidase in constructs in which swrA and lacZ were in frame, and the absence of blue indicates that lacZ and swrA were not in the same frame. Site-directed mutagenesis was used to remove an intervening stop codon from the swrAPY79‘−1 frame lacZ ’ construct as indicated in Fig. 5. The following strains were used to generate the figure: DS242, DS301, DS306, DS336, DS338, and DS376.
B. Analysis of the swrA start codon. Site-directed mutagenesis was used to remove the putative TTG and GTG start codons from the swrA3610‘0 frame lacZ ’ construct as indicated in Fig. 5. As lacZ translation depended on swrA translation, the blue colour indicates that translation of swrA was not blocked by the indicated mutation whereas the absence of blue indicates that translation of swrA had been blocked by the indicated mutation. The following strains were used to generate the figure: DS375 and DS406.

The single A:T base pair insertion found in the swrAPY79 and swrA168 alleles is found within the swrA coding region and suggests that the mutation in laboratory strains causes a frameshift. In support of this interpretation, a translational fusion of swrAPY79 that was made to the 0 frame lacZ reporter did not support β-galactosidase synthesis (Fig. 6A). If the base pair insertion does indeed create a frameshift, the resulting shift is predicted to introduce in frame a stop codon immediately downstream of the insertion (Fig. 5). Therefore, to observe the effect of the frameshift mutation experimentally, a translational fusion of swrAPY79 was made to the ‘−1 frame’lacZ reporter and the stop codon was removed by site-directed mutagenesis (Fig. 5). The resulting construct did support β-galactosidase synthesis, confirming that the mutation in laboratory strains causes a frameshift (Fig. 6A).

The swrA frameshift mutation reverts with high frequency

A striking feature of the mutant swrA gene, as discussed above, is its capacity to readily revert. Now knowing that the mutation is a frameshift mutation, we sought to quantify the frequency of reversion through the use of a fusion of swrAPY79 to the ‘0 frame’lacZ reporter, which was inserted into the amyE gene of a laboratory strain to generate DS316. In this construct, lacZ was out of frame with the beginning of the swrA open reading frame because of the presence of the A:T basepair insertion. Swarm agar was inoculated with DS316 and then incubated for 2 days at 37°C. After incubation, cells were harvested from the plate, diluted and plated on LB agar containing X-gal to determine the frequency at which β-galactosidase production was restored. The results show that the deletion of a base pair within the swrA sequence occurred at a frequency of roughly 10−4, which is much higher than that observed for a typical point mutation (10−8−10−10, Cox, 1976). The addition of a base pair to this region occurred at a similar frequency (roughly 5 × 10−5). This frequency was determined by the use of a construct (DS522) in which the insertion of a base pair was necessary to place lacZ in the same open reading frame as swrA. The capacity to swarm played no role in the observed reverse and forward mutation frequencies as the DS316 and DS522 strains were incapable of swarming as a result of the presence of the sfp mutation found in laboratory strains.


Recent work has shown that domestication of B. subtilis in the laboratory has led to the loss of certain behaviours that are conspicuous in wild strains. For example, wild strains form robust and architecturally complex biofilms with spore-forming aerial projections that resemble fruiting bodies (Branda et al., 2001). Laboratory strains, in contrast, produce thin and fragile biofilms that are flat and morphologically uniform in appearance. Another striking difference between wild and laboratory strains is the capacity of the former to exhibit a robust swarming phenotype on agar surfaces. In previous work we established that the failure of laboratory strains to swarm effectively is due in part to a mutation in a gene (sfp) that is needed for the production of the surfactant, surfactin (Kearns and Losick, 2003). But failure to produce surfactin is not a sufficient explanation for the inability of laboratory strains to exhibit robust surface motility. A principal contribution of the present investigation is the discovery of a previously uncharacterized gene (swrA) involved in surface motility that accounts for the failure of laboratory strains to swarm. Laboratory strains in which the sfp and swrA mutations were simultaneously corrected swarmed as robustly as did wild strains.

A striking feature of the swrA mutation is that it is an insertion of an A:T base pair in an eight-nucleotide-long stretch of A:T base pairs. This observation suggests that the insertion arose by ‘slipped-strand’ mispairing during DNA replication, as insertions and deletions of nucleotide base pairs is a hallmark of the copying of DNA across homopolymeric sequences of this kind (Levinson and Gutman, 1987). Slipped-strand mispairing involving repetitive DNA sequences is believed to be responsible for a wide range of phase variation events in microorganisms (Henderson et al., 1999). Consistent with a mechanism of phase variation, the swrA insertion mutation was found to revert to the wild-type sequence (that is, undergo a deletion of one of the A:T base pairs) at a much higher frequency than that of typical spontaneous mutation. Phase variation enables an organism to toggle readily between a state in which certain genes are expressed and a state in which the genes are inactive. It is easy to imagine how this would be beneficial to wild strains of B. subtilis as the capacity for surface motility could be advantageous in certain natural environments (i.e. growth on surfaces) but not in others (i.e. growth in liquid). We suppose that B. subtilis was inadvertently subjected to selection for the inability to swarm because colonies with the capacity to spread on agar surfaces would probably have been discriminated against in the laboratory. If so, then laboratory manipulation would have favoured the inactive (insertion-harbouring) phase state of swrA and further locked in a block in surface motility through selection for the sfp mutation. Our results also provide an explanation for reports of swarming by certain laboratory strains, such as 168 (Dixit et al., 2002). The high reversibility of the swrA mutation would have allowed for the ready selection of laboratory strains that had regained some measure of swarming motility, albeit restricted as the result of the presence of the sfp mutation.

Another contribution of the present investigation is the discovery in addition to swrA of three genes that are required for swarming but not swimming. These genes are swrB, swrC and efp. The swrB gene is predicted to encode a novel protein and shares a common promoter with a large cluster of genes dedicated to flagellum biosynthesis and chemotaxis. If swrB functions in a similar fashion to the genes with which it is co-transcribed, swrB may contribute to behavioural control or flagellum production. A role for swrB in flagellum assembly is particularly appealing as B. subtilis cells undergo a dramatic increase in the number of cell surface flagella during swarming motility.

The swrC gene, which encodes a membrane protein with similarity to the AcrB-like family of export pumps, was recognized previously as the site of a mutation that increases sensitivity to the autotoxic effects of surfactin, presumably by impairing surfactin export (Tsuge et al., 2001). Our present findings suggest that excessive intracellular accumulation of surfactin interferes with surface motility. This interference could be alleviated, and swarming restored, by the introduction of a second mutation that blocked surfactin production and by providing surfactin exogenously.

The discovery that disruption of the gene efp, which is predicted to encode translation elongation factor P (EF-P), abolishes swarming raises the possibility that translational regulation may play a role in establishing the swarming state. Elongation factor P acts as an enhancer of protein synthesis in vitro by stimulating peptide bond formation between particular combinations of amino acids during translation (Glick and Ganoza, 1975; Glick et al., 1979). Consistent with its biochemical activity, EF-P was found to be essential for growth in all organisms in which it had previously been studied (Aoki et al., 1997; Ramos et al., 1997). Depletion of eIF-5A, the EF-P homologue in Saccharomyces cereviseae (Kyrpides and Woese, 1998), resulted in a cessation of growth but only a partial reduction in total protein translation (Kang and Hershey, 1994). To reconcile the essential nature of eIF-5A with continued protein synthesis and the specificity of EF-P-stimulated peptide bond formation, it was suggested that eIF-5A might be responsible for initiating the translation of only a subset of mRNAs, some of which are required for viability in yeast. The situation in B. subtilis differs from that in yeast, as disruption of the bacterial gene had no measurable effect on growth but resulted in a severe non-swarming phenotype. In this case, EF-P conceivably enhances the translation of a subset of transcripts necessary for swarming motility and not viability. If so, swarming in B. subtilis could be an attractive system for exploring the mechanism of EF-P action.

Periodic migration and zones of consolidation, not found in wild-type B. subtilis swarming, are an integral part of the swarming behaviour of another bacterium, Proteus mirabilis (Rauprich et al., 1996). Therefore, it is interesting to note that disruptions in two genes yabR and ymfI and a polar effect on rrnB-derived tRNA genes generated zones of consolidation during B. subtilis swarming. The yabR gene is predicted to encode a protein with similarity to the mRNA binding domain of ribosomal protein S1 and is perhaps involved in transcript stability or translation efficiency (Subramanian et al., 1981; Schnier et al., 1982; Levin and Losick, 1994). Combined with the requirement for full expression of tRNA genes, it seems that translation at a high level is required to maintain unimpeded swarming motility. The role of ymfI, which is predicted to encode a short chain alcohol dehydrogenase, in periodic swarming is not clear.

To sum up, we have discovered four genes, referred to as swrA, swrB, swrC and efp, that are required for swarming but not swimming motility. Three of these genes were previously uncharacterized, and a fourth (swrC) was previously shown to play a role in resistance to the toxic effects of surfactin. A mutation in one of the genes (swrA) together with a mutation in a gene (sfp) needed for surfactin biosynthesis fully accounts for the inability of domesticated strains to exhibit a robust swarming phenotype. The swrA mutation is readily revertible, raising the possibility that swarming is subject to control by phase variation. A principal challenge of future work will be to elucidate the mechanism by which these genes enable B. subtilis to swarm in a coordinated manner on solid surfaces.

Experimental procedures

Strains and media

Bacillus subtilis PY79, 168, and 3610 were grown in Luria–Bertani (LB), 10 g of tryptone, 5 g of yeast extract, 5 g of NaCl per L broth or plate supplemented with 1.5% Bactoagar at 37°C. When appropriate, antibiotics were included at the following concentrations: 10 µg ml−1 tetracycline, 100 µg ml−1 spectinomycin, 5 µg ml−1 chloramphenicol, 5 µg ml−1 kanamycin and 1 µg ml−1 erythromycin plus 25 µg ml−1 lincomycin (MLS). When appropriate 80 µg ml−1 of Xgal (5-bromo-4-chloro-3-indolyl β-d-galactopyranoside) was added to the media. ‘Swarm agar’ plates (25 ml) containing LB fortified with 0.7% Bactoagar were prepared fresh and the following day were dried for 30 min in a laminar flow hood. Each plate was toothpick inoculated from an overnight colony and scored for swarming motility after 24 h incubation at 37°C. Plates were visualized with a Bio-Rad geldoc system and digitally captured using Bio-Rad Quantity One software.

Transposon mutagenesis

Transposition of mini-Tn10 was conducted within the undomesticated B. subtilis strain 3610 and mutants were selected for under swarming non-permissive conditions. The temperature sensitive vector carrying mini-Tn10, pIC333 (Petit et al., 1990; Steinmetz and Richter, 1994), was transduced using the phage PBS1 (Keggins et al., 1978) from lab strain RL2420 to the wild strain 3610 to generate DS1010, the host for transposition. To generate each transposon library, DS1010 was inoculated into 3 mls LB broth containing spectinomycin and incubated at 22°C in a roller drum for 14 h. The culture was then diluted 20-fold into fresh 2 ml LB broth containing spectinomycin and incubated at 22°C for 3 h before transferring to 37°C for another 4 h. Each culture was diluted and spread on LB plates fortified with 1.5% agar and spectinomycin and incubated overnight at 37°C. Transposition occurred at a frequency of 0.1% as measured by the number of spectinomycin resistant colonies relative to the total number of colony forming units.

Swarming mutant screen

Transposon mutants were subjected to a series of screens to identify mutants specifically defective in swarming motility. The mutants were inoculated, 25 at a time, onto LB containing spectinomycin fortified with 0.9% agar (an agar concentration which severely restricts but does not abolish swarming motility) and incubated at 30°C overnight. Putative swarming mutants were identified as small colonies and picked onto individual 30 mm diameter Petri plates containing 5 ml of swarm agar supplemented with spectinomycin and incubated at 30°C overnight. The mutants that remained unable to completely colonize the mini plates were then screened for MLS sensitivity. Occasionally, the entire temperature-sensitive plasmid will integrate into the genome in the absence of transposition and this event can be distinguished by the acquisition of MLS resistance carried on the plasmid (Steinmetz and Richter, 1994). Plasmid integration occurred at a 10% frequency and all MLS resistant colonies were dropped from the study. Swimming motility was assayed by growing cells in LB to late exponential phase and observing motile behaviour of cells microscopically. Mutants defective in swimming motility were dropped from the study. Finally, the mutant phenotype of all remaining transposon mutants was verified under the standard conditions for swarming motility (inoculation on LB swarm agar containing spectinomycin and incubated 24 h at 37°C).

Identifying transposon insertions

As the mini-Tn10 transposon had been engineered to carry an E. coli origin of replication, the transposon and flanking DNA was cloned into E. coli for sequencing (Steinmetz and Richter, 1994). Chromosomal DNA was purified from each mutant and 4 µg of DNA was digested in 20 µl volume with either HinDIII or EcoRI for 4 h at 37°C. The restriction reactions were extracted with phenol–chloroform and ligations were conducted in 200 µl overnight at 15°C. The ligations were phenol–chloroform extracted and transformed into E. coli DH5α by selecting for spectinomycin resistance. Each cloned transposon insertion was sequenced using primers ‘my050’ and ‘my051’ specific to either end of the transposon to sequence cloned flanking chromosomal DNA. Mini-Tn10 generates a 9 bp duplication at the site of insertion which served as a tag to identify the disrupted gene (Table 1).

Swarm expansion assay

Cells were grown to mid-log phase at 37°C in LB broth and resuspended in PBS buffer (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4 and 2 mM KH2PO4) containing 0.5% India ink (Higgins). Swarm agar was dried for 30 min in a laminar flow hood, centrally inoculated with 2 × 107 cells in 10 µl of the resuspension solution, dried for another 10 min and incubated at 37°C (standard conditions). The India ink sets into the agar and demarks the origin of the colony and swarm radii were measured relative to the origin. For consistency, an axis was drawn on the back of the plate and swarm radii measurements were taken along this transect. When appropriate, 10 mg ml−1 surfactin (Sigma) was dissolved in 20 mM NaOH and 10 µl was spotted in the centre of the swarm agar plate before inoculation.

Strain construction

The genotypes of all strains used in this study are listed in Table S1. Laboratory strains were corrected to sfp+ by transforming with chromosomal DNA purified from OKB170 (Nakano et al., 1992) and selecting for MLS resistance. The MLS resistance marker in OKB170 is linked to the sfp+ allele and transformants in which sfp+ was co-transduced were identified by the ability to produce surfactin on swarm agar.

All insertion deletion mutations were generated using long flanking homology PCR (using primers indicated in Table S2) and transformed into competent cells of strain PY79 (Wach, 1996). DNA containing a tetracycline drug resistance gene (pDG1515) was used as a template for marker replacement (Guérout-Fleury et al., 1995). Mutations were transferred to the 3610 background using SPP1 mediated generalized transduction (Yasbin and Young, 1974).

Complementation and reporter constructs

All primers used in the construction of plasmids are listed in Table S2. To generate the swrA complementation constructs (pDP36, pDP40, pDP41, pDP53 and pDP54), a PCR product containing the swrA gene and upstream region was amplified from B. subtilis chromosomal DNA (from the indicated strain background) using the primers ‘swrAF’ and ‘swrAR’. The PCR product was cloned into the EcoRI and BamHI sites of plasmid pDG364, which carries a chloramphenicol resistance marker and a polylinker between two arms of the amyE gene (Karmazyn-Campelli et al., 1992).

To generate the swrA–lacZ translational fusions (pDP39, pDP47, pDP55, pDP56 and pDP58), a PCR product containing the 5′ region of swrA was amplified from B. subtilis chromosomal DNA (from the indicated strain background) using the primer ‘swrAlacZF’ and one of the following primers ‘swrAlacZ-1’, ‘swrAlacZ0’, or ‘swrAlacZ + 1’. The PCR product was then cloned into the BamHI and SalI sites of plasmid pDG1728, which carries a spectinomycin resistance marker and polylinker between the arms of the amyE gene (Guérout-Fleury et al., 1996). Polymerase chain reaction-based site-directed mutagenesis was conducted using the primer pairs ‘swrAstopCGA1’ and ‘swrAstopCGA2’, ‘swrAstartGTC1’ and ‘swrAstartGTC2’, ‘swrAstartTTA1’ and ‘swrAstartTTA2’, and ‘swrAstopCGA3’ and ‘swrAstopCGA4’, in conjunction with ‘swrAlacZ′ translational fusion primers and cloned to generate the mutated reporter constructs pDP66, pDP67, pDP68 and pDP95 respectively. All swrA complementation and swrA–lacZ constructs were verified by sequencing using the primer ‘swrAseq’.

To generate the sigD bypass construct pDP72, the promoter region of flgB was amplified by PCR from 3610 chromosomal DNA with primers ‘PflgBF’ and ‘PflgBR’ and digested with EcoRI and XhoI. The sigD gene was amplified from 3610 chromosomal DNA with primers ‘sigDF’ and ‘sigDR’ and digested with XhoI and BglII. The digested PflgB PCR fragment and the sigD PCR fragment were then simultaneously inserted into the EcoRI/BamHI sites of pDG364 by three-way ligation.

To generate the ylxL bypass construct pDP73, the promoter region of flgB was amplified from 3610 chromosomal DNA with primers ‘PflgBF’ and ‘PflgBR’ and digested with EcoRI and XhoI. The ylxL gene was amplified from 3610 chromosomal DNA with primers ‘ylxLF’ and ‘ylxLR’ and digested with XhoI and BglII. The digested PflgB PCR fragment and ylxL PCR fragment were then simultaneously inserted into the EcoRI/BamHI sites of pDG364 by three-way ligation.

All constructs were transformed into competent cells of strain PY79 and then transferred to the 3610 background using SPP1 mediated generalized transduction (Yasbin and Young, 1974).

SPP1 phage transduction

To 0.2 ml of dense culture grown in TY broth (LB broth supplemented after autoclaving with 10 mM MgSO4 and 100 µM MnSO4), serial dilutions of SPP1 phage stock were added and statically incubated for 15 min at 37°C. To each mixture, 3 ml of TYSA (molten TY supplemented with 0.5% agar) was added, poured atop fresh TY plates and incubated at 37°C overnight. The plate containing near confluent plaques was harvested by scraping into a 50 ml conical tube, vortexed and centrifuged at 5000 × g for 10 min. The supernatant was treated with 25 µg ml−1 DNase final concentration before being passed through a 0.45 µm syringe filter and stored at 4°C.

Recipient cells were grown to stationary phase in 2 ml of TY broth at 37°C. Cells (0.9 ml) were mixed with 10 µl of SPP1 donor phage stock. Nine ml of TY broth was added to the mixture and allowed to stand at 37°C for 30 min. The transduction mixture was then centrifuged at 5000 g for 10 min, the supernatant was discarded and the pellet was resuspended in the remaining volume. One hundred µl of cell suspension was then plated on TY fortified with 1.5% agar, the appropriate antibiotic and 10 mM sodium citrate.


We thank P. Zuber for the gift of OKB170 and J. E. González-Pastor for donation of EG371. We are grateful to K. Vernovsky for participating in the early stages of this work. We thank S. Ben-Yehuda, S. Branda, K. Carniol, J. Dworkin, R. Kolter, L. Turner and D. Rudner for helpful discussions and critical reading of the manuscript. This work was supported by National Institute of Health National Research Service Award GM66612 to D.K., by National Institute of General Medicine MBRS SCORE grant GM60654–03 to R. R. and by National Institute of Health grant GM18568 to R.L.

Supplementary material

The following material is available from

Fig. S1. Transposon insertions and an insertion/deletion mutation in the rrnB operon.

Table S1. Strains and plasmids.

Table S2. Primers.