INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Several genera of flagellated eubacteria show a behavior known as swarming when propagated on the surfaces of certain solid media (22, 29, 38, 67). Swarming involves flagellum-dependent surface translocation by groups of cells. Swarmer cells are generally longer and more flagellated than cells of the same species propagated in liquid media (swimmer cells). Swarmer cells move within an encasement of slime, a nondescript term for a complex mixture of polysaccharides, surfactants, proteins, and peptides, etc., that surrounds the colony. The morphology of swarmer cells, along with the extracellular slime, probably helps overcome surface friction, favoring rapid expansion of the swarmer colony. In some organisms, the swarmer cell state may be associated with pathogenesis (67, 70).

Although the swarming phenomenon has been described for several bacterial genera, very little is known about the signaling events or signal transduction mechanisms that lead to the production of swarmer cells. The discovery of swarming motility in the well-characterized bacteria Escherichia coli and Salmonella enterica serovar Typhimurium (39) allows us to use these organisms as model systems for the study of swarming behavior. We have exploited them in three ways in this study. First, extensive transposon mutagenesis was used to identify genes involved in swarming. Second, existing mutants were rationally chosen for testing swarming defects. Third, medium conditions were altered to test for specific chemical signals for promoting swarming.

S. enterica serovar Typhimurium was chosen for transposon mutagenesis because it is less fastidious than E. coli K-12 strains for swarming. Most laboratory strains of E. coli swarm best on media solidified with Japanese Eiken agar, while wild-type serovar Typhimurium can swarm on either Difco or Eiken agar. We have reported earlier that mutants of these two organisms with mutations in the chemotaxis pathway are defective in swarming (39). Here we report the isolation of a large class of conditional mutants that did not swarm on Difco agar but did so on Eiken agar. A large number of these mutants were defective in lipopolysaccharide (LPS) synthesis and many had mutations in genes controlling the physiological state of the cells, while some mutations mapped in putative two-component signaling genes. A swarming model integrating all the current information for S. enterica serovar Typhimurium and E. coli is presented.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, phage, and plasmids. Bacterial strains, phage and plasmids are listed in Table 1. S. enterica serovar Typhimurium strain SJW1103 (a fliC-stable derivative of serovar Typhimurium LT2) was the starting strain for all transposon mutagenesis experiments. For TnphoA mutagenesis, SJW1103 was made phoN by transducing phoN::Tn10dtet from MST308 to construct AT190. AT538 was constructed by transducing cheA::Tn10 from KK2051 into SJW1103. SJW2971 is an SJW1103 derivative with a point mutation in motB.

Media and chemicals. Swim medium consisted of 0.3% Difco Bacto agar and Gibco BRL L-broth (LB) base (20 g/liter). Swarming bacteria were propagated on either Eiken swarm medium (0.45% Eiken agar, Eiken broth base [20 g/liter], and 0.5% glucose) or Difco swarm medium (0.6% Difco Bacto agar, LB base [20 g/liter], and 0.5% glucose). Minimal swarm medium was made as described previously (3). Both swim and swarm plates were allowed to dry overnight at room temperature before use. Vibrio harveyi was grown in AB medium (34). Antibiotics used were kanamycin (50 µg/ml), ampicillin (100 µg/ml), tetracycline (20 µg/ml), and chloramphenicol (25 µg/ml). The phosphatase activity indicator XP (5-bromo-4-chloro-3-indolyl phosphate) was used at 40 µg/ml.

Transposon mutagenesis using TnphoA and Tn10dCm. TnphoA was delivered using Mud-P22 phage derived from MST2001 (Table 1). Induction of Mud-P22 and preparation of lysates were as follows (8). A 30-ml culture of MST2001 grown in LB-ampicillin to 5 × 10⁸ CFU/ml at 37°C was induced by the addition of 30 µl of a 2-mg/ml solution of mitomycin C and incubated overnight with shaking. Five milliliters of the overnight culture was treated with 0.5 ml of chloroform and vortexed for 1 min, and the lysate was centrifuged at 1,000 × g for 20 min. Chloroform treatment and centrifugation were repeated with the supernatant. TnphoA was delivered by mixing equal volumes of a 10² dilution of the lysate with an overnight culture of recipient bacteria. After incubation for 1 h at room temperature, 200 µl of the mixture was plated on LB-kanamycin with XP. Blue colonies are the result of insertions that generate in-frame PhoA fusions to membrane or periplasmic proteins (60). Insertions that do not form such fusions (indicated by white colonies) can be in either membrane or cytoplasmic protein genes.

2

Analysis of swarming defects. All mutants were screened on swim and Difco swarm media (six per plate) at 37°C to identify those that had functional flagella yet failed to swarm. Defects in flagellar function were identified on swim medium, where an absence of outward migration can be due to defects in synthesis of flagella (Fla), ability to rotate flagella (Mot), or chemotactic behavior (Che). Che mutants can be visually distinguished from Fla and Mot mutants by their fuzzy appearance at the site of inoculation, contrasted to the tight mound of cells formed by other two classes of mutants. The Che phenotype was confirmed by examination of the mutants in a drop of liquid under a light microscope, where they showed biased running or tumbling behavior.

Identification of mutant genes. Genomic DNA from TnphoA insertion mutants of interest was digested to completion with TaqI. DNA fragments were circularized with T4 DNA ligase and amplified by PCR using Vent polymerase (New England Biolabs) and appropriately designed primers that included phoA sequence. PCR products were cloned into pUC19 and sequenced using one of the amplification primers. DNA sequences linked to phoA were analyzed using the National Center for Biotechnology Information BLAST program (4), National Center for Biotechnology Information Microbiol Genome Blast databases (www.ncbi.nlm.nih.gov/Blast/unfinishedgenome.html), and Genome Sequencing Center Salmonella Sequencing Project data (www.genome.wustl.edu/gsc/).

luxS null mutant and activity. The luxS gene (listed as ygaG in the Salmonella typhi database [www.sanger.ac.uk/Projects/S_typhi/blast_server.shtml]) from S. enterica serovar Typhimurium SJW1103 was amplified by PCR and cloned into pUC19. luxS was disrupted by insertion of a cat gene (1.14-kb Klenow-filled XhoI fragment, originally isolated as an HhaI fragment from pCHL884 [Table 1]) into an EcoRV site and cloned into the SacI site of the suicide vector pKAS32. The resulting plasmid was conjugated from SM10pir into SJW1103, with selection for Cm^r. Chromosomal disruption of luxS in AT527 was confirmed by Southern analysis (78). AI-2 activity in cell-free culture fluids (CFCFs) was measured using the reporter strain BB170 as described previously (91) with the following modifications: 0.2 ml of CFCFs was combined with 1.8 ml of BB170 and incubated at 30°C with shaking for 5 h, and then 0.1 ml of the mixture was transferred to a scintillation vial and light production was measured in an SAI Technology integrating photometer (model 3000).

Testing of sensitivity to phage P22. A 0.1-ml portion of an overnight culture of the tester strain was mixed with 2.5 ml of soft agar (0.5%) and overlaid on 1.5% LB agar plates. Three microliters of a P22 lysate (HT12/4 int 103; 10¹¹ PFU/ml) was spotted on the lawn and incubated at 37°C overnight.

Complementation of swarming defects with LPS and surfactin. Five-microliter drops of solution containing S. enterica serovar Typhimurium LPS (3.5 mg/ml) or B. subtilis surfactin (10 mg/ml) were spotted directly on Difco swarm medium, allowed to dry, and inoculated with the tester strains. The plates were incubated overnight at appropriate temperatures (37°C for serovar Typhimurium and 30°C for E. coli [E. coli is nonmotile at 37°C]).

Flagellin estimation. Two hundred microliters of an overnight broth culture was spread on Difco swarm medium. After 5 h of incubation, cells were harvested in LB and the optical density at 600 nm (OD₆₀₀) of each sample was adjusted to approximately 0.7. Equal aliquots (30 µl) of whole-cell samples were boiled (5 min), electrophoresed on sodium dodecyl sulfate-10% polyacrylamide gels, and subjected to Western blot analysis using cross-reacting anti-Serratia marcescens flagellin antibodies (2). Flagellin bands were quantified using a Bio-Rad densitometer equipped with Multi Analyst software. A similar procedure was followed for broth-grown cells harvested at mid-log phase.

Anaerobic culturing. Swarm plates were stored in a Forma 1024 anaerobic glove box for at least 5 h to deplete the medium of oxygen. The oxygen-free plates were then inoculated with bacteria grown on swim plates and incubated anaerobically for 17 h. S. enterica serovar Typhimurium (tested on Difco and Eiken swarm media) was incubated at 35°C, while E. coli (tested on Eiken swarm media) was incubated at 26°C. An identical set of plates was incubated aerobically.

pH measurements. pH measurements were done on Eiken swarm medium, with and without 0.5% glucose, and with 0, 50, 100, and 200 mM MES [2-(N-morpholino)ethanesulfonic acid] buffer. In unbuffered media, an initial pH range of 5 to 8.3 was set by adjusting the pH with either HCl or NaOH. pH readings were taken using an Orion pH microelectrode (model 98-10) connected to a Corning pH meter (model 220). Readings were accurate within 0.05 pH unit. pH measurements of swarm medium were made by inserting the tip of the electrode approximately 1 mm into the agar. pH measurements within a swarming colony were made similarly, except that measurements were taken at approximately 2-mm intervals starting at the perimeter of the colony and moving towards the center. The efficiency of surface colonization under different pH conditions was estimated by measuring the colony radius every 30 min, starting 1 h after inoculation.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Analysis of transposon mutants of S. enterica serovar Typhimurium that are defective in swarming. Two different transposons were used for mutagenesis. TnphoA allows one to specifically examine insertions in membrane and periplasmic proteins (60). Tn10dCm was used in addition to avoid any insertional bias associated with use of a single transposon. Delivery of transposons and mutant analysis are described in Materials and Methods. Since swarming is dependent on functional flagella, all Fla mutants (lacking flagella) and Mot mutants (with nonfunctional flagella) were eliminated from further analysis.

Testing of known mutants for swarming defects. Since most of the isolated transposon mutations were recovered only once (Table 3), we thought it also prudent to test several defined mutations in both S. enterica serovar Typhimurium and E. coli that could be expected to affect swarming, based either on known requirements for these genes in other bacteria that show surface motility (e.g., fliL, lrp, pepQ, and flgN in Proteus mirabilis; lon in Vibrio parahaemolyticus; the luxR homologue sdiA [due to involvement of homoserine lactone autoinducer] in Serratia liquifaciens; and fim [due to requirement for fimbrae] in Myxococcus xanthus) or on their effects on physiological functions, including environmental sensing, signaling, gene activation, or possibly intercellular interactions, that might be reasonably expected to affect swarming. The data are presented in Table 4. No obvious defects were observed except for ntrA and CL79 in serovar Typhimurium, which showed a Dps phenotype. A Dps (for defective in progressive swarming) phenotype is one in which movement is restricted to the area of inoculation. This is likely the result of an absence of surface-active compounds that promote colony expansion (61, 69). ntrA (encoding sigma 54) is involved in nitrogen regulation but is also known to regulate other physiological functions, such as anaerobic metabolism (11) and flagellar regulation in Campylobacter coli and Vibrio cholerae (46, 48). The phenotype of CL79 is harder to understand since it carries a MudJ insertion in a non-open-reading-frame region of the virulence plasmid pSLT and since the absence of pSLT had no effect on swarming (Table 4). We note that a mutation in parB, which regulates, but is not essential for, partitioning of pSLT, also showed a Dps phenotype (Table 2). Defined LPS mutants are discussed separately below.

Effect of luxS disruption on swarming. In S. liquefaciens, a homoserine lactone autoinducer plays an important role in swarming (23) by controlling the production of the surfactant serrawettin (57). Recently E. coli and S. enterica serovar Typhimurium were shown to produce another type of autoinducer (AI-2), and a gene, luxS, homologous to the AI-2 synthase gene from V. harveyi was identified as being responsible for its synthesis (92, 93). To test whether AI-2 plays a role in serovar Typhimurium swarming, luxS was disrupted (see Materials and Methods). To ascertain that AI-2 synthesis was abolished in the resulting mutant, AT527 (Table 1), CFCFs were assayed using the V. harveyi reporter strain BB170 as described in Materials and Methods. CFCFs from the AI-2 positive (BB152) and AI-2 negative (DH5) control strains generated approximately 8,500 and 100 arbitrary light units, respectively, while those from wild-type serovar Typhimurium (SJW1103) and its luxS derivative (AT527) produced 23,000 and 100 arbitrary light units, respectively (data not shown). Thus, AI-2 production is indeed abolished in the luxS mutant AT527. However, swarming was unaffected in this mutant (Table 4), showing that luxS was not required for swarming under these conditions.

Flagellin upregulation in LPS mutants: rescue of swarming with external surfactin. Since a large number of the transposon mutants isolated in this study had mutations that mapped to the LPS biosynthetic pathway, we chose representative waa mutants affected in O-antigen synthesis (waaL mutant 41-11), core synthesis (waaC mutant), and core modification (waaK mutant) (Tables 2 and 3) for characterization. Flagellin upregulation was used as a marker for swarmer cell differentiation. When propagated on swarm medium, the LPS mutants had flagellin induction levels comparable to that of the wild type, suggesting that their inability to swarm on Difco medium was not due to signaling defects that interfered with regulating flagellar synthesis (Fig. 1). All LPS mutants showed normal motility on swim plates, suggesting that they did not harbor flagellar assembly defects. Elongation defects were not characterized, because this phenotype is not as dramatic in S. enterica serovar Typhimurium as it is in other organisms. No gross defects were obvious.

View larger version (20K): [in this window] [in a new window]   FIG. 1.   Flagellin levels in representative swarming mutants propagated on swarm medium. A wild-type strain (SJW1103 [SJW]) and the indicated mutants were grown on swarm medium for 5 h and harvested as described in Materials and Methods. Flagellin amounts in OD600-equalized aliquots of cells were analyzed by Western blotting followed by densitometry. Flagellin levels (arbitrary densitometry units) were normalized to 100 for the wild-type strain. Results from one experiment are shown. Absolute flagellin values differed in several repititions of the experiment, but the trend was very reproducible.

View larger version (119K): [in this window] [in a new window]   FIG. 2.   Rescue of swarming defects with external addition of surfactin and LPS. Five-microliter drops of solution containing surfactin or LPS were placed at the center of 0.5% Difco swarm medium, inoculated with S. enterica serovar Typhimurium (waaL and cheA) and E. coli (RP437) strains, and incubated overnight as described in Materials and Methods.

Precocious swarming phenotype of an LPS core mutant. Given that the majority of the swarming mutants harbored LPS defects (Tables 2 and 3), we also tested the swarming abilities of a collection of known LPS mutants obtained from the Salmonella Genetic Stock Center (Table 4). While most of these mutants also showed conditional defects, one mutant (the waaG mutant) swarmed earlier and faster than its wild-type parent on Difco swarm medium and secreted copious amounts of slime (Fig. 3). The growth rates of the waaG mutant and its wild-type parent were comparable. After 5 h of incubation, while the wild-type strain had just begun outward progression, the waaG strain had swarmed over half of the plate (Fig. 3A). This mutant can be described as a precocious swarmer (7). The swarming pattern of the waaG colony was unusual in that the cells swarmed in a monolayer (observed microscopically; the arrow in Fig. 3 points to the swarming front), in contrast to a wild-type colony, which is several layers thick. However, by the time the wild-type front had covered the plate, the waaG colony had grown denser than the wild-type colony and was very rough in appearance. The slime on the plate was so plentiful that the whole colony could slide off the plate as a mat upon tilting. (The circular pattern of dots near the edge of the 20-h waaG colony is not reproducible and is likely related to the rough nature of the mat.) This mutant was also able to swarm on minimal medium (Fig. 3B), which does not support swarming of wild-type S. enterica serovar Typhimurium (see below). On minimal medium, swarming was initiated with a delay of about 20 h. Interestingly, although a slime ring was clearly visible as a circular halo around the waaG colony propagated on minimal medium, the path of the swarming cells within this slime was highly branched. We were unable to determine if flagellin expression was constitutively upregulated in this mutant, because the cells aggregated easily in broth, lysed readily, and resisted Western blotting (likely due to the slime).

View larger version (91K): [in this window] [in a new window]   FIG. 3.   Precocious swarming of waaG LPS core mutant. (A) A wild-type (wt) S. enterica serovar Typhimurium strain (SL3770) and its waaG mutant derivative (SL3769) were inoculated on Difco Bacto swarm medium and incubated at 37°C. Photographs were taken after 5 and 20 h of incubation. The arrow points to edge of the waaG swarm colony. (B) The waaG strain was inoculated on minimal swarm medium and incubated at 37°C. Photographs were taken after 24, 30, and 44 h. The arrow indicates edge of the slime ring. Swarming movement was confined to regions that appear as branches emanating from the center.

Flagellin induction in baeS, yfhK, and yojN mutants. To test if the three putative two-component signaling baeS, yfhK, and yojN mutants (Table 3) were defective in swarmer cell differentiation, flagellin levels in these cells on swarm medium were estimated as described for the waa mutants (Fig. 1). Compared to the wild type, all three mutants appeared defective in flagellin induction on swarm medium. A representative che mutant (cheA, encoding the central kinase) was also defective in flagellin induction, consistent with flagellar staining data of Che mutants reported earlier (16).

Signals for swarming. It is apparent from the results presented above that external surface conditions play an important role during swarming. At the very least, the milieu or slime elaborated by a swarmer colony must provide the aqueous environment within which cells must rotate their flagella in order to effect movement and allow swarm colony expansion. It is likely that the O antigen or surfactants contribute to such an environment. The slime may also be expected to harbor signals for swarmer cell development, and the two-component signaling systems could be involved in both slime generation and signal detection. Since known autoinducer systems are not required and little is known about the putative baeS, yfhK, and yojN two-component systems, we tested known chemotaxis signals since the che genes are essential for swarming (16, 39, 69). We also tested whether slowed flagellar rotation, the favored model for signaling in V. parahaemolyticus swarmer cell differentiation (62), was required for swarming in S. enterica serovar Typhimurium.

View larger version (22K): [in this window] [in a new window]   FIG. 4.   pH experiments. (A) SJW1103 was inoculated on Eiken swarm medium preset to different pH values, and colony radii were measured after overnight incubation. (B) SJW1103 was inoculated on Eiken swarm medium (with 0.5% glucose) preset to five different pH values (, 5.2; , 6.2; , 7.0; , 8.0; , 8.6). pH measurements were taken with a microelectrode 8 h after inoculation. Positive x-axis numbers indicate distance (in millimeters) within the swarming colony, and negative numbers indicate distance outside the colony. (C) Same as panel B, except with Eiken swarm medium without glucose. The initial pHs of the media were 6.2 (), 7.2 (), 8.0 (), and 8.7 (). (D) SJW1103 was inoculated on Eiken swarm medium without glucose and buffered with different MES concentrations (, 0 mM MES; , 50 mM MES; , 100 mM MES; , 200 mM MES). The pH at the center of inoculation was measured at the indicated times. The pH change displayed on the y axis is in tenths of a pH unit above the starting pH of 6.1. The results displayed in all panels were highly reproducible.

View larger version (140K): [in this window] [in a new window]   FIG. 5.   Effect of anaerobiosis on swarming. SJW1103 was grown on Difco swarm medium either anaerobically (left) or aerobically (right) as described in Materials and Methods and photographed after 6 and 20 h.

View larger version (18K): [in this window] [in a new window]   FIG. 6.   Flagellin levels in a Mot mutant in broth and in swarm medium. Cultures of the indicated strains were grown either in LB broth or on Difco swarm medium as described in Materials and Methods. Aliquots of OD600-equalized cells were analyzed by Western blotting using antiflagellin antibodies as described in Materials and Methods.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The overall picture emerging from our large-scale search for genes and signals governing swarming motility in S. enterica serovar Typhimurium is that important functions are performed by the LPS layer and the slime and that both known and putative two-component systems play a role in swarming. We have eliminated known autoinducer systems, specific amino acids, pH changes, oxygen, iron starvation, viscosity, and flagellar rotation as primary signals. We will consider an alternate signaling model below.

Role of LPS in swarming. All LPS mutants isolated in this study showed conditional swarming defects in that they were defective for swarming on Difco Bacto swarm medium but not on Eiken swarm medium. In an earlier study with S. marcescens (69), we reported that mutants unable to make the surfactant serrawettin displayed a similar conditional swarming defect (61). Given that Eiken agar can overcome surfactant defects, we surmised that the permissive quality of this agar may be due to some parameter (such as superior wettability, for example) that affects the surface. The rescue of a wide spectrum of S. enterica serovar Typhimurium mutants on Eiken swarm medium (Tables 2 to 4) argues against a specific chemical component in the medium that is responsible. This view is also supported by our inability to demonstrate a dialyzable constituent in the Eiken agar that promotes swarming on Difco Bacto agar (our unpublished results). It is more reasonable to imagine that all of the conditional mutants are ultimately altered either in their ability to elaborate wetting agents or in some property of the cell surface that hinders surface translocation. The ability of surfactin, which lowers surface tension and improves wettability, to support swarming of the LPS mutants isolated in this study (Fig. 2) is consistent with the notion that Eiken agar provides a more wettable surface.

Defects in putative two-component signaling systems. In contrast to the LPS mutants, swarming mutants with homology to two-component signaling systems (baeS, yfhK, and yojN mutants) were all defective in flagellin induction, similar to the defect in the known cheA signaling mutant (Fig. 1). Thus, multiple signaling systems appear to participate in swarmer cell differentiation. Preliminary results indicate that the ability of baeS, yfhK, and yojN mutants to swarm on Eiken agar is not due to induction of swarmer cell differentiation but likely is due to lower surface tension of this substrate that allows surface colonization without having to upregulate flagellar expression (data not shown).

Autoinducers and other signals in swarming. Since swarming is not initiated for several hours after inoculation, at which time the colony has reached a confluent cell density, it is reasonable to expect quorum sensing to play a role at some stage in the process. In gram-negative bacteria, two varieties of autoinducers have been identified as cell density signals or quorum sensors (93). Acyl-homserine lactones (HSLs or AI-1) are elaborated by the LuxI-LuxR family of proteins (LuxI is the HSL synthase and LuxR is the transcriptional activator protein that binds to the HSL). HSLs play a role in the swarming response of Serratia. The target genes regulated by this quorum-sensing system are involved in synthesis of the biosurfactant serrawettin (57). In S. liquefaciens, mutations in swrI (encoding HSL synthase) reduce swarming (23). It is important to note, however, that like the LPS mutations described in this study, mutations in the quorum-sensing pathway that produces serrawettin in both S. marcescens and S. liquefaciens do not interfere with swarm cell differentiation and motility but interfere only with swarm colony expansion (57, 69). Thus, serrawettin itself is not the signal for swarm cell differentiation in these organisms.

Model for cell density sensing and signaling during S. enterica serovar Typhimurium swarming. It is apparent that a large number of genes function to ensure external surface conditions (slime) that improve wettability and facilitate expansion of the swarming colony. It is possible that the nonswarming phenotype of mutants with mutations in various metabolic pathways (Table 3) is related to this function. In light of our elimination of many of the obvious candidates for signals, we suggest not only that the external slime plays a critical role in swarm colony expansion but also that polysaccharides in the slime constitute a signal for swarmer cell differentiation. A polysaccharide signal might act by altering external water activity, which may cause changes in physical characteristics of the membranes or periplasmic space where the chemoreceptors required for swarm signaling are located. The isolation of defects in the mdo genes (which synthesize branched glucans found in the periplasm), membrane functions, and several two-component signaling systems may be relevant in this regard. While this suggestion seems to implicate osmolarity as the signal, we note that the known osmolarity sensors EnvZ and Kdp are apparently not essential for swarming (Table 4). Changes in osmolarity (achieved by varying external NaCl or sucrose concentrations) are known to regulate flagellar expression through the EnvZ pathway in E. coli (83) but not to affect flagellar expression in serovar Typhimurium (51). Interestingly, the RcsB-RcsC signaling pathway is reported to influence flagellin expression in response to osmolarity in S. typhi (5). Given that we still do not know the molecular nature of the osmolarity signal in well-studied osmosensing pathways (103), it is difficult for us to suggest how the slime polysaccharide signal would operate. Our understanding of the structure of the periplasm is also very tenuous. It is possible that the polar localization of chemoreceptors (59) may position them to function as specialized osmosensors, detecting gradual rather than sudden shifts in water activity during swarming.