Genome-Wide Screening of Genes Required for Swarming Motility in Escherichia coli K-12

Abstract

Escherichia coli K-12 has the ability to migrate on semisolid media by means of swarming motility. A systematic and comprehensive collection of gene-disrupted E. coli K-12 mutants (the Keio collection) was used to identify the genes involved in the swarming motility of this bacterium. Of the 3,985 nonessential gene mutants, 294 were found to exhibit a strongly repressed-swarming phenotype. Further, 216 of the 294 mutants displayed no significant defects in swimming motility; therefore, the 216 genes were considered to be specifically associated with the swarming phenotype. The swarming-associated genes were classified into various functional categories, indicating that swarming is a specialized form of motility that requires a wide variety of cellular activities. These genes include genes for tricarboxylic acid cycle and glucose metabolism, iron acquisition, chaperones and protein-folding catalysts, signal transduction, and biosynthesis of cell surface components, such as lipopolysaccharide, the enterobacterial common antigen, and type 1 fimbriae. Lipopolysaccharide and the enterobacterial common antigen may be important surface-acting components that contribute to the reduction of surface tension, thereby facilitating the swarm migration in the E. coli K-12 strain.

Swarming is a flagellum-dependent form of bacterial motility that facilitates migration of bacteria on viscous substrates, such as semisolid agar surfaces, which has been observed for a variety of motile bacteria (13). In order to swarm, cells first differentiate into a specialized state (swarmer cells) characterized by an increase in flagellum number and the elongation of cells and then move as multicellular rafts across surfaces (10, 12, 13). This is in contrast to swimming motility, which represents individual cell motility in aqueous environments. In addition to the morphological changes, it is known that the swarmer cells produce extracellular materials (wetting agents), such as surfactants and exopolysaccharides, to increase surface wetness and thus facilitate movement (13). In view of these features of swarming, it is very likely that various cellular activities are involved in this type of motility.

Recent genome-scale approaches have disclosed that swarmer differentiation coincides with the regulation of a wide range of cellular activities. Wang et al., using DNA microarray analysis, demonstrated that surface-growing Salmonella enterica serovar Typhimurium cells had physiologies markedly different from those of Salmonella cells grown in broth (49). In addition, the proteome analysis by Kim and Surette indicated that the metabolic pathways in the swarmer cells of Salmonella were dynamically shifted compared with those in the swimmer cells (20). However, little is known about the genes that are required for the swarming phenotype in Escherichia coli. In the case of E. coli K-12 strains, swarming requires media solidified with Eiken agar (12). The requirement for this special agar is currently explained by the O-antigen-defective lipopolysaccharides (LPSs) of these strains and the particular wettability of the Eiken agar surface (13). However, the involvement of wetting agent-like materials in the swarming motilities of these strains has not been explored.

Recently, a set of single-gene knockout mutants of all the nonessential genes in E. coli K-12 (the Keio collection) was constructed (1). In the present study, we used this mutant collection to identify the genes required for swarming motility in E. coli and found that a wide variety of genes are implicated in this form of motility. In addition, we propose the cell surface components that are required for swarming motility in this strain, possibly as wetting agent-like materials. To our knowledge, this is the first report of a comprehensive analysis of swarming-related genes using a systematic, gene-deleted mutant collection.

MATERIALS AND METHODS

Single-gene knockout mutant collection (the Keio collection).

The E. coli K-12 strain W3110 was used as a wild-type strain for the swarming motility assay. In this study, we used mutants from a systematic, single-gene knockout mutant collection (the Keio collection) of all the nonessential genes of BW25113, a strain derivative of W3110. These mutants were created by Baba et al. (1) by replacing the open reading frame coding regions with a kanamycin resistance cassette, according to the method of Datsenko and Wanner (7). Glycerol stocks of the collection dispensed in a 96-well microplate format were maintained at −30°C and −70°C.

Screening of the swarming-associated genes.

The media used for the swarming assay was Luria-Bertani (LB) medium containing 0.5% (wt/vol) glucose and 0.6% (wt/vol) Eiken agar (Eiken Chemical Co., Tokyo, Japan), which was dispensed into OmniTray dishes (Nalge Nunc International, NY). Typically, these dishes, which are termed swarm plates, were air dried for 10 min before being used. Precultured bacteria grown in LB broth containing kanamycin (30 μg/ml) and 1.0% (wt/vol) agar in 96-well microplates were carefully inoculated with a 96-pin metal replicator (Copyplate; Tokken Inc, Chiba, Japan) onto the surfaces of the swarm plates. The plates were then wrapped with Saran Wrap to prevent dehydration. The plates were incubated at 33°C for 18 to 20 h. Swarming motility was assessed by examining the colony sizes and the branch-spreading patterns on the semisolid agar medium. Every mutant was classified into one of three categories: the strongly repressed phenotype, the moderately repressed phenotype, and the normal swarming phenotype (Fig. 1). The swarming assay was performed at least three times for each mutant in the Keio collection. The function of each swarming-associated gene was assigned by referring to the EchoBASE (http://www.ecoli-york.org/) (30) and EcoCyc (http://www.ecocyc.org/) (18) databases. Mutants exhibiting the strongly repressed swarming phenotype were tested for their swimming abilities. Each strain was inoculated on LB medium solidified with 0.3% (wt/vol) Eiken agar and incubated at 33°C for 15 to 16 h.

FIG. 1.

Swarming motility phenotypes. Bacterial cells were inoculated onto LB medium containing 0.5% (wt/vol) glucose and 0.5% (wt/vol) Eiken agar and incubated overnight at 33°C. Results are shown for the wild-type strain (WT; normal swarming type), the phoQ mutant (moderately repressed type), and the fliC mutant (strongly repressed type).

Electron microscopy.

The bacterial strains were grown overnight at 33°C on the swarm plates. The cells were taken up with toothpicks from the edges of the colonies and transferred into 10 mM ammonium acetate (pH 7.2). Five microliters of the cell suspension was placed on collodion membrane-coated grids, and the excess liquid was removed with filter paper. The cells on the grids were negatively stained with 1% phosphotungstic acid (pH 7.5) for 1 min and observed using a Hitachi H-800 transmission electron microscope at an accelerating voltage of 100 kV.

RESULTS AND DISCUSSION

Screening of the genes involved in swarming motility.

The swarming motility of E. coli K-12 strains can usually be observed on semisolid Eiken agar medium (Fig. 1). The genome of the E. coli K-12 strain comprises 4,390 open reading frames (1). The Keio collection contains 3,985 nonessential gene mutants, in each of which a kanamycin resistance cassette has been used to create a highly targeted single-gene disruption (1, 7). Mutants were screened for swarming defects on LB soft agar (0.6% agar) containing 0.5% glucose. On inspection of the mutant library, 294 gene mutants (7.4% of the nonessential genes) were found to express strong repression of the spreading and branching colony morphologies (the strongly repressed phenotype) (Fig. 1). In addition, another 510 mutants (approximately 13% of the nonessential genes) exhibited partially reduced colony spreading (the moderately repressed phenotype) (Fig. 1) (see Table S2 in the supplemental material).

Swimming ability was tested for the mutant displaying the strongly repressed phenotype, using LB solidified with 0.3% agar. Of the 294 mutants, 216 showed no significant defects in swimming motility, thereby indicating that the 216 genes were specifically associated with the swarming phenotype (Table 1). The remaining 78 mutants (the atpA, atpB, atpC, atpD, atpE, atpF, atpG, atpH, cheA, cheB, cheR, cheW, cheY, cheZ, cmk, dnaK, dsbA, fabH, fis, flgA, flgB, flgC, flgD, flgE, flgF, flgG, flgI, flgJ, flgK, flgL, flgN, flhA, flhB, flhC, fliA, fliC, fliD, fliF, fliG, fliI, fliJ, fliK, fliM, fliN, fliO, fliP, fliQ, fliS, folB, folP, galU, gmhA, hflD, motA, motB, mtn, pgm, priA, rcsC, rfaH, rho, rluD, rplA, rrmJ, sufC, tap, tolB, tolR, ubiH, ubiX, waaC, waaD, waaE, ycjW, yfgA, yhbG, yhcA, and yjeQ mutants) that were clearly defective in swimming ability had deletions in genes for flagellar function (the che, flg, flh, fli, and mot genes) and energy production (the ATP synthase and ubiquinone genes) (the function of each gene product is described in Table S3 in the supplemental material). For the other mutants with impaired swimming abilities, this impairment may be due to defects in flagellar function or energy production for active growth and motility.

TABLE 1.

E. coli genes whose mutations caused strongly repressed swarming with no significant defects in swimming^a

Function	COG category(ies)	Gene(s)
Information storage and processing
Translation, ribosomal structure and biogenesis (3/170)	J	deaD, efp, truA
Transcription (11/288)	K	argP, cysB, envR, hipB, marA, rof, uxuR, ybdO, ycjZ, yfeR, yfeT
DNA replication, recombination and repair (6/237)	L	dps, fimB, idaB, ogt, xerC, xerD
Cellular processes
Cell division and chromosome partitioning (1/34)	D	sulA
Posttranslational modification, protein turnover, chaperones (11/133)	O	ahpF, grxB, hflC, hscA, htpG, htpX, osmC, ppiB, ppiD, tpx, trxB
Cell envelope biogenesis, outer membrane (20/219)	M	asmA, dgkA, etk, hlpA, mrcB, ompA, pal, rffH, tolC, tsx, waaB, waaF, waaG, waaI, waaJ, waaQ, wcaE, wecA, wecE, wecG
Cell motility and secretion (5/151)	N	fimA, fimD, fimF, fliR, tolQ
Inorganic ion transport and metabolism (15/253)	P	chaC, cvrA, feoA, feoB, fepA, fepB, fepD, fepG, fes, narH, ppk, trkA, yheL, yheM, yheN
Signal transduction mechanisms (9/187)	T	atoS, barA, cpxA, crp, cusR, evgS, gmr, hnr, ompR
Metabolism
Energy production and conversion (24/281)	C	ackA, acnA, aldA, aldH, cydD, cyoA, fdrA, fdx, hyaA, hyaB, iscU, lpdA, nuoF, rnfG, sdhA, sdhB, sdhC, sucA, sucB, sucD, ybdH, yeiA, ygfH, ynfG
Carbohydrate transport and metabolism (17/355)	G	fucU, galM, gatY, gpmI, mglB, mgsA, nagC, pgi, ptsH, ptsI, rpiA, treA, yhcH, yqaD, zwf
Amino acid transport and metabolism (12/390)	E	argG, argR, aroE, asnB, csdA, dapF, glnA, gmhB, metL, mpaA, oppC, yeeF
Nucleotide transport and metabolism (5/83)	F	guaB, ndk, purC, purK, pyrF
Coenzyme metabolism (6/144)	H	hemE, lipA, pabC, pdxJ, pdxY, ubiF
Lipid metabolism (1/96)	I	fadL
Secondary metabolite biosynthesis, transport and catabolism (4/95)	Q	entB, entE, entF, yrbE
Poorly characterized
General function prediction only (11/448)	R	cvpA, nlpI, prpD, trmE, wcaH, wzxE, ycdY, ydeE, yfeH, ygeD, ygfZ
Unknown (9/447)	S, U, V	sapF, ybaK, ybeD, ybiA, ydcZ, ydgA, yfgL, yneE, yqaA
No COG assignment (48/1,214)		arpB, dnaT, dsbB, envC, eutS, fadK, fimC, fimH, galE, gapC, iscS, macA, mdoC, mdtH, ninE, osmB, pldB, ppdC, prmB, rep, trpL, ubiC, ubiE, waaP, waaS, waaY, waaZ, wecB, wecC, wecD, wecF, yaiW, yccK, yccV, yciG, ydaT, ydbA, ydcX, ydfT, yehP, yfbJ, yfgJ, yfjN, yfjP, yigG, ymfA, yncH, yncN

^aGenes are classified into COG categories. The numbers in parentheses represent the number of swarming-related genes out of the number of genes belonging to each COG category. The function of each gene product is described in Table S1 in the supplemental material.

The 216 swarming-associated genes were classified into clusters of orthologous groups (COGs) (1, 46), suggesting that they belong to a variety of functional categories (Table 1) (the function of each gene product is described in Table S1 in the supplemental material). The results imply that a wide variety of genes are associated with the swarming motility of E. coli. Compared to other swarming bacteria, E. coli K-12 strains are known to be relatively fastidious in their requirements for swarming. This is because the LPSs of the K-12 strains lack the O-antigen, and thus, the swarming ability of this bacterium is dependent upon the special Eiken agar that provides a particularly wettable surface. It might therefore be assumed that swarming in K-12 is more sensitive to any mutation affecting cell surface properties, thus resulting in a relatively large number of swarming-defective mutants. Additionally, it might be possible that due to polar effects in the case of genes arranged in operons, the current assessment of swarming motility results in an overestimation of the number of swarming-related genes. However, the available evidence suggests that E. coli K-12 cells deploy numerous and various cellular functions to facilitate migration over viscous surfaces and thereby expand their growing area. In the following sections, we focus on several functional groups of gene mutants with the strongly repressed swarming phenotype and no significant swimming defects, and we discuss the roles of the gene functions in swarming.

TCA cycle and glucose metabolism.

It has been suggested that swarming is a process that requires a large amount of energy (20). Consistent with this, there are several genes associated with energy production, such as those coding for ATP synthase, respiratory chain components, and tricarboxylic acid (TCA) cycle enzymes, with strongly repressed swarming motilities (Table 1; also see above). Of these, two TCA cycle enzymes, succinate dehydrogenase (Sdh) and 2-ketoglutarate dehydrogenase (Suc), are notable because these gene mutants displayed significant swarming defects but normal swimming motilities. The reactions catalyzed by these enzymes yield reducing power, which might supply a significant amount of energy required for swarming movement or increased production of cellular components for facilitating swarm migration (described below). With regard to this, Kim and Surette presented data indicating that the expression of these two enzymes as well as ATP synthase components was upregulated in the swarmer cells compared to that observed in the swimmer cells. From these findings, it seems likely that swarming is a more energy-consuming process than swimming. Recently, it was proposed that the entire TCA cycle is exploited during the swarm migration of Salmonella (20). However, our results demonstrated that certain TCA cycle enzymes, including isocitrate dehydrogenase (Icd), malate dehydrogenase (Mdh), and citrate synthase (GltA), were dispensable for swarming; this indicated that the total activity of the TCA cycle is not necessary for swarming in E. coli.

In Salmonella, the addition of glucose to LB agar medium is required for swarmer cell differentiation (12, 19, 21). This is probably because glucose metabolism provides the required energy for swarmer cells. In the case of E. coli, we observed the requirement of glucose for swarm migration (data not shown). Our results for E. coli suggested that glucose utilization is one of the key factors for the swarming motility of this bacterium. For example, several genes whose products are involved in the uptake of glucose (ptsH, ptsI) and in glycolysis (pgi, gpmI) are listed in Table 1. Furthermore, in addition to energy production, the requirement of the zwf gene product may suggest an alternative means by which glucose utilization plays important roles in swarming differentiation. The zwf gene encodes glucose-6-phosphate dehydrogenase, which catalyzes the conversion of glucose-6-phosphate into 6-phosphogluconolactone, which is the first step of the pentose pathway (50). The pathway may be important for swarming by supplying the precursors for the biosynthesis of various cellular constituents, particularly cell surface materials, such as LPS and the enterobacterial common antigen (ECA), that are required for swarming as described below.

Lipopolysaccharide and enterobacterial common antigen.

The medium used for testing bacterial swarming motility is generally solidified by the addition of agar to a final concentration of 0.5% to 2.0%. Owing to the lack of surface moisture, a semisolid surface on which the swimming motilities of individual bacterial cells are inhibited is created (13). Under such conditions, swarming bacteria are considered to produce extracellular wetting agents consisting of polysaccharides and surfactants, etc., that extract water from the agar and thereby increase their surface wetness; this, in turn, facilitates the flagellar-driven bacterial motility on the agar plates (13, 51). In the present study, a comprehensive analysis using the Keio collection demonstrated that the genes involved in the biosynthesis of two cell surface components, LPS and ECA, were required in bulk for E. coli K-12 swarming motility but they were not required for swimming motility. Here, we describe the results and discuss the roles of these amphipathic substances in the swarming motility of E. coli K-12.

The contribution of LPS, particularly its O-antigen, to swarming motility has been reported for several gram-negative bacteria (11, 29, 47, 49). In addition, it appears to be an indispensable component for the flagellum-independent surface translocation in Vibrio cholerae (4) and also for the social motility in Myxococcus xanthus (3). Presumably, LPS acts by increasing the wetness of both the bacterial cell surface itself and the surrounding environment. The outer-membrane-anchored form of LPS, through its outer extruded core oligosaccharide and O-antigen polysaccharide, confers hydrophilicity to the cell surface, whereas the extracellular form of LPS acts as a surfactant, reducing the surface tension and decreasing the frictional resistance between the agar and cell surface (13).

In the E. coli K-12 strain used in this study, swarming motility was significantly repressed in many of the LPS biosynthesis-related gene disruptants as shown in Table 2. The E. coli K-12 strain carries a mutation in the O-antigen biosynthesis gene locus that results in the inability to assemble mature O-antigen (26). The results also suggest that all the genes that were involved in core oligosaccharide synthesis, but not those involved in O-antigen synthesis, were required for swarming. The necessity of all the genes involved in the core oligosaccharide synthesis is presumably indicative of the fact that core oligosaccharide not only plays a role in conferring hydrophilicity to the cell surface but also serves as an acceptor for additional polysaccharides other than the O-antigen. In addition to the swarming defects, the disruption of the gmhA, waaE (hldE), waaD (hldD), and waaC genes led to impaired swimming motility (Table 2). The products of these genes are responsible for the construction of the inner core of LPS (34), thereby suggesting that the inner core may be necessary for flagellar assembly and function. Although GmhB is also a member of the inner-core-synthesizing pathway, the gmhB mutation did not result in swimming deficiency. This may be explained by the presence of an additional enzyme that partially compensates for the gmhB mutation as proposed by Kneidinger et al. (22).

TABLE 2.

Swarming and swimming of mutants in genes associated with biosynthesis of ECA and LPS^a

Biosynthesis group	Gene	Synonym	Status of indicated ability		Gene product and function
			Swarming	Swimming
ECA biosynthesis	wecA	rfe	−	+	Undecaprenyl-phosphate α-N-acetylglucosaminyl transferase
	wzzE	wzz	+	NT	Chain length modulation protein
	wecB	rffE	−	+	UDP-N-acetylglucosamine-2-epimerase
	wecC	rffD	−	+	UDP-N-acetyl-d-mannosaminuronic acid dehydrogenase
	rffG		±	NT	dTDP-glucose 4,6-dehydratase 2
	rffH		−	+	dTDP-glucose pyrophosphorylase 2
	wecD	rffC	−	+	dTDP-fucosamine acetyltransferase
	wecE	rffA	−	+	dTDP-4-oxo-6-deoxy-d-glucose transaminase
	wzxE	wzx	−	+	Lipid III flippase
	wecF	rffT	−	+	4-acetamido-4,6-dideoxy-d-galactose transferase
	wecG	rffM	−	+	UDP-N-acetyl-d-mannosaminuronic acid transferase
LPS biosynthesis Core	gmhB	yaeD	−	+	d,d-heptose 1,7-bisphosphate phosphatase
	gmhA	lpcA	−	−	d-sedoheptulose 7-phosphate isomerase
	waaE	rfaE	−	−	Fused heptose 7-phosphate kinase/heptose 1-phosphate adenyltransferase
	waaD	rfaD	−	−	ADP-l-glycero-d-mannoheptose-6-epimerase
	waaF	rfaF	−	+	Lipopolysaccharide core biosynthesis protein
	waaC	rfaC	−	−	Heptosyl transferase I
	waaZ	rfaZ	−	+	Protein involved in KdoIII attachment during core biosynthesis
	waaY	rfaY	−	+	Lipopolysaccharide core biosynthesis protein
	waaJ	rfaJ	−	+	UDP-d-glucose:(galactosyl)lipopolysaccharide glucosyltransferase
	waaI	rfaI	−	+	UDP-d-galactose:(glucosyl)lipopolysaccharide-α-1,3-d-galactosyltransferase
	waaB	rfaB	−	+	UDP-d-galactose:(glucosyl)lipopolysaccharide-1,6-d-galactosyltransferase
	waaS	rfaS	−	+	Lipopolysaccharide core biosynthesis protein
	waaP	rfaP	−	+	Core heptose phosphorylation protein
	waaG	rfaG	−	+	Glucosyltransferase I
	waaQ	rfaQ	−	+	Lipopolysaccharide core biosynthesis protein
O-antigen	wzzB		+	NT	Regulator of length of O-antigen component of lipopolysaccharide chains
	wbbL		+	NT	Putative creatinase
	wbbK		+	NT	Putative transferase
	wbbJ		±	NT	Putative transferase
	wbbI		+	NT	Lipopolysaccharide biosynthesis protein
	rfc		±	NT	O-antigen polymerase
	glf		+	NT	UDP-galactopyranose mutase
	rfbX		+	NT	Lipopolysaccharide PST transporter
	rmlC	rfbC	+	NT	dTDP-4-dehydrorhamnose 3,5-epimerase
	rmlA	rfbA	+	NT	dTDP-glucose pyrophosphorylase
	rmlD	rfbD	+	NT	dTDP-4-dehydrorhamnose reductase
	rmlB	rfbB	+	NT	dTDP-glucose 4,6-dehydratase
	waaL	rfaL	+	NT	O-antigen ligase

^a−, strongly repressed; ±, moderately repressed; +, normal motility; NT, not tested.

In addition to LPS, swarming was also strongly repressed in most of the ECA biosynthetic gene mutants listed in Table 2. ECA is one of the cell surface components of enteric bacteria, and its carbohydrate portion is a linear heterosaccharide containing a trisaccharide repeat unit, →3)-α-d-Fuc4NAc-(1→4)-β-d-ManNAcA-(1→4)-α-d-GlcNAc-(1→ (38). Three forms of ECA have been reported: ECA_PG, ECA_LPS, and ECA_CYC. ECA_PG and ECA_LPS are the major and a minor forms of ECA in which the trisaccharide repeats are linked to the outer membrane phosphoglyceride and the core region of LPS, respectively (38). ECA_CYC is another water-soluble cyclic form of ECA (38), and in the E. coli K-12 strain, it has been demonstrated to consist of four uniformly repeating units (9). The structure and assembly of ECA are now fairly well understood due to exhaustive research efforts (17, 38, 39, 45); however, with the exception of its contribution to resistance against organic acids (2) and bile salts (36), little is known about the function of this molecule. Swarming has been reported for various species of enterobacteria (13). Although the requirement for ECA in swarming motility may be obvious in E. coli K-12 strains lacking the O-antigen, which is an important prerequisite for semisolid surface translocation in many species of gram-negative bacteria, the findings presented here indicate that ECA plays an important role, possibly as an additional wetting agent, in swarming motility in E. coli. Additionally, the findings provide a new insight into the ECA molecule itself and establish the importance of conserving ECA and its wide distribution.

With regard to the other polysaccharide materials, it is considered that capsular polysaccharide also contributes to a certain extent in the swarming motility of E. coli K-12. The respective genes are listed in Table 1 and include the cholanic acid capsule synthesis genes wcaE and wcaH. Furthermore, disruption of the wcaB, wcaF, and wcaM genes resulted in moderate swarming repression (see Table S2 in the supplemental material). Although the extracellular polysaccharide was reported to play a role in the social motility of M. xanthus (24, 27), the capsular polysaccharide is not generally considered to be crucial for swarming motility in other swarming bacteria. However, in the case of E. coli K-12 strains lacking the O-antigen, the assemblage of available extracellular polysaccharides may be necessary in order to facilitate swarm migration.

Type 1 fimbriae.

The type 1 fimbriae in E. coli are peritrichously expressed filamentous surface structures and are one of the virulence factors in uropathogenic E. coli (43). The following eight genes in the fim gene cluster are involved in the production of type 1 fimbriae: fimB, fimE, fimA, fimC, fimD, fimF, fimG, and fimH (43). Of these, the fimA (major fimbrial subunit), fimB (recombinase), fimC (periplasmic chaperone), fimD (outer membrane usher), fimF (adaptor), and fimH (adhesin) mutants exhibited strongly repressed swarming motilities but no significant repression of swimming motility (Table 1). The fimA, fimC, and fimD gene products are essential for constructing the fimbrial fiber, while the fimF and fimH gene products that are located in the fibrillar tip are not required for this. However, the fimF and fimH mutants were reported to have markedly reduced numbers of fimbriae per cell (40, 42).

To examine the roles of fimbriae in swarming motility, we used electron microscopy to observe the expression of fimbriae and flagella in the wild-type strain and the fimA mutant. While the swimmer and swarmer cells of the wild-type strain possessed many flagella, almost all of these cells did not express fimbriae (Fig. 2A and C). This unexpected finding may suggest that swarming movement requires the participation of fimbrial genes but does not require the fimbrial fiber structure. In the swimming cells, there was no significant difference in flagellar number between the wild-type and the fimA mutant (Fig. 2A and B). However, when the cells are grown on the swarm plates, the number of flagella per cell in the fimA mutant was very small compared with that in the wild-type strain (Fig. 2C and D). From these observations, it seems most likely that the swarming defect in the fimA mutant is caused by a decrease in the number of flagella and not by the lack of fimbrial fiber. The reduced production of flagella in the fimA mutant might be due to defects in the induction of flagella expression in the fimA mutant during swarmer cell differentiation. Recently, Köhler et al. reported a similar observation in Pseudomonas aeruginosa, i.e., a mutation in the fimbrial subunit gene pilA caused complete swarming inhibition; however, there was no significant effect on swimming motility (23). Interestingly, the presence of pilus-like structures was not obvious in the swarmer cells of the P. aeruginosa strain. An alternate report by Latta et al. suggested that pili are expressed at specific times during the development of swarming colonies in Proteus mirabilis (25). Our findings that fimbrial genes are required for swarming but that no fimbriae were observed on the swarmer cells might suggest that the expression of fimbrial and flagellar genes is tightly controlled during swarmer cell differentiation in E. coli. A more detailed analysis is necessary for clarifying the roles of fimbrial genes in swarming motility.

FIG. 2.

Electron microscope images of bacterial cells negatively stained with 1% phosphotungstic acid. (A) Wild-type swimmer cells. (B) fimA mutant swimmer cells. (C) Wild-type swarmer cells. (D) fimA mutant cells grown on the swarm plate. Bars, 1 μm.

Signal transduction systems.

Flagella and chemotaxis systems have been demonstrated to play important roles in swarmer cell differentiation on semisolid agar surfaces. In addition to these factors, swarming motility in E. coli appears to require several two-component signaling systems consisting of sensor and regulator proteins; in most cases, a mutation in either the sensor or the regulator gene affected swarming motility. Of these, the Cpx-signaling system consisting of CpxA (sensor kinase) and CpxR (response regulator) is considered to respond to envelope stress (35). Additionally, the Cpx-signaling pathway was suggested to be involved in the regulation of adhesion-induced gene expression (32). With respect to flagellar gene expression, the Cpx system negatively regulates the level of motABcheWA gene expression (8). However, in our study, the cpxA gene mutation was observed to abolish swarming, whereas the cpxR mutant exhibited a normal swarming phenotype, thereby suggesting a specialized role for the cpxA product in swarming activity. The exclusive requirement for cpxA without cpxR was also reported for the regulation of HilA, an activator of invasion gene expression in S. enterica serovar Typhimurium (16, 31). These cases seem curious but may be due to cross-talk among two-component systems (14).

Chaperones and protein-folding catalysts.

The mutant for the gene encoding the molecular chaperone HscA exhibited a normal swimming but a remarkably repressed swarming phenotype (Table 1). HscA (Hsc66) is a DnaK-like chaperone that interacts with the iron-sulfur (Fe-S) cluster assembly protein IscU (15). Both the hscA and the iscU genes belong to the isc operon, encoding proteins involved in the biosynthesis of the Fe-S cluster (48). Analysis of our data revealed that the iscS and fdx genes in this operon were also associated with swarming (Table 1), suggesting that the Fe-S cluster, which is a cofactor incorporated into several proteins, is essential for swarming activity. Other molecular chaperons, including HtpX (a putative membrane-bound zinc metalloprotease) (41), HtpG (function unknown) (44), and the protein-folding catalysts PpiB and PpiD (peptidylproryl-cis-trans-isomerases) (6), were also implicated in swarming motility. As described above, swarming appears to involve a wide range of cellular processes. Therefore, it is reasonable to expect that chaperones and protein-folding catalysts are required for the correct folding and degradation of many swarming-related proteins.

Iron acquisition.

Iron is an essential factor for the growth of bacteria; however, it is not readily available in natural environments. To obtain this metal, bacteria have developed multiple iron acquisition systems. In our study, mutations in most of the genes involved in the utilization of enterobactin (enterochelin), a well-known siderophore in E. coli, significantly influenced colony spreading by swarming motility (Table 1). In addition, other genes coding for iron uptake are included in the list of partially repressed swarming mutants (see Table S2 in the supplemental material). The results indicate that iron acquisition systems are indeed required for E. coli swarming on LB agar plates. The ent and fep genes listed are enterobactin biosynthesis and transport genes, respectively (5, 37). Mutations in these genes had no significant effects on the E. coli growth in liquid LB medium (1), supporting the contention that iron is not limiting in LB medium. However, our data may imply that E. coli cells growing on LB agar swarm plates are in an iron-starved state, which might promote swarmer cell differentiation. In this regard, McCarter and Silverman (28) have demonstrated that iron limitation is a second signal for the expression of the lateral flagella that is responsible for swarming motility in Vibrio parahaemolyticus. They reported that iron-regulated outer membrane proteins were also produced in V. parahaemolyticus cells grown on agar plates containing iron-rich heart infusion; this strongly suggests that the heart infusion agar-grown cells were iron deprived. A possible explanation for iron starvation on nutrient-rich agar media is based on the limitation of iron diffusion into cells on agar surfaces (28). Consistent with this explanation, it has been reported by Wang et al. that growth of Salmonella on swarming agar induced the expression of iron metabolism-related genes, which included genes for the biosynthesis and transport of enterobactin and the hydroxamate-dependent iron uptake system (49). Therefore, it may be presumed that subsistence of swarm plate-grown cells occurs in virtually iron-limited conditions even in nutrient-rich media, which may be an important factor in facilitating swarming motility. Additionally, this may be a common feature among various swarming bacteria.

In this study, we used a complete set of single-gene knockout mutants from E. coli K-12 for the screening of swarming-associated genes in this strain. The results illustrate that the production of energy and surface materials, such as LPS and ECA, significantly affects swarming motility on semisolid agar surfaces. Although E. coli K-12 swarming has been reported to be highly dependent on the wettability of the agar, the requirement of LPS and ECA suggests that wetting agents, which are bacterial products that reduce surface tension, are important in facilitating swarm migration in this strain, which is similar to that in other swarming bacteria. In particular, the biochemical properties of ECA have been extensively studied; however, the function of this molecule is largely unknown. Therefore, it is of interest that the present data suggest a novel function for ECA.

Another intriguing result of this study is that numerous genes, up to one-fifth of the genes on the genome, are involved in swarming motility. This leads us to infer that swarming is a highly organized mode of motility in viscous environments that requires various cellular functions in addition to the flagellar function. Among the swarming-associated genes, there are many genes whose functions are currently unknown. Studies of the roles of these uncharacterized genes that may be implicated in swarming would lead to an understanding of the functional properties of the gene products and new discoveries of the molecular mechanisms involved in swarming. Some of the swarming-associated genes are expected to be induced during swarmer cell differentiation. In this regard, a recent study by Wang et al. is notable in that nearly one-third of the genes on the Salmonella genome were demonstrated to be differentially regulated between surface and liquid growth (49). In viscous surface environments, a single bacterial cell would be unable to expand its growing space by the swimming mode. In order to overcome such stressful conditions, cells probably activate various metabolic and synthetic pathways to develop into swarmer cells and move as a group despite the high energy cost.

To date, it has been demonstrated that a variety of motile bacteria have the ability to swarm on semisolid agar, which may imply that bacterial cells frequently encounter viscous surfaces in their natural environments. It is considered that bacteria are capable of sensing environmental viscosity with their flagella and that they determine their mode of motility, swimming or swarming, depending on the perceived viscosity. For pathogenic bacteria, swarmer cell differentiation may represent one of the physiological states during the infection process. Thus far, swarming has been reported to be associated with virulence in several bacteria (10, 13, 33). Furthermore, some virulence factors were shown to be coregulated with swarmer cell differentiation. Although the involvement of swarming motility during the infection process remains to be elucidated, it is expected that bacterial pathogens would be exposed to viscous environments, such as mucus, during the early stages of infection; under these circumstances, they may preferentially differentiate into swarmer cells to increase their growing area. In the future, using genetically well-characterized E. coli and S. enterica serovar Typhimurium as model systems, a global network of gene expressions in viscous environments may be unraveled; this may contribute to our understanding of both bacterial signal transduction systems and pathogenic mechanisms within the host.