Introduction

Over 30 years ago, Henrichsen [23] identified six categories of bacterial surface motility based on the study of motility in hundreds of strains from 40 bacterial species belonging to 18 different genera: swimming, swarming, gliding, twitching, sliding, and darting. Of these phenomena, swimming in liquid medium has been studied mostly over the past three decades, especially in Escherichia coli and Salmonella enterica serovar Typhimurium. However, this form of motility is not restricted to liquid media, but also occurs when bacteria colonize medium with low agar concentrations (0.2–0.4%). In contrast to the individual movement of bacteria during swimming, swarming is considered as a bacterial group behavior associated with migration on semisolid surfaces. In the presence of extracellular slime (polysaccharides, peptides, surfactants, etc.), bacteria exhibit a flagella-driven movement on top of the agar (0.4–1.2%) enabling them to spread as a biofilm over the surface. In the laboratory, the concentration of the agar used to solidify the medium can be critical for swarming likely owing to a certain level of wetness required for movement. Association of cells in a group likely facilitates movement by increasing fluid retention. Swarming may be viewed as a specialized case of swimming on a surface [10, 22, 53].

The swarming phenomenon has been demonstrated in a growing number of diverse bacteria including members of Aeromonas, Azospirillum, Bacillus, Burkholderia, Chromobacterium, Clostridium, Escherichia, Proteus, Pseudomonas, Salmonella, Serratia, Sinorhizobium, Vibrio, and Yersinia [10, 53]. Swarming normally requires the differentiation of vegetative cells into a specialized cell type called swarmer cells. Surface contact (change in viscosity), physiological parameters, and chemical signals (nutritional status) provide stimuli that trigger swarmer cell differentiation. Although the pathways for signal integration are still poorly understood, it is believed that signals are sensed and transmitted by two-component regulatory systems, cytosolic regulators, or flagella leading to a complex network. The differentiated swarmer cells are often hyperflagellated and elongated. These cells move in groups or rafts organized parallel to their long axis to maximize cell–cell contact, colonizing the available surface. The migration front is preceded by a visible layer of slime-like extracellular material [5, 7, 10, 18]. As a result of the embedding in an extracellular slime matrix, population densities are normally extremely high [25].

Besides the differentiation into swarmer cells, wetting agents that reduce surface tension are also required for swarming in many organisms. Among the best studied are the glycolipids or lipopeptides produced by Bacillus subtilis (surfactin), Serratia (serrawettin), Pseudomonas aeruginosa (rhamnolipids) (for an overview: [46]). A role for the LPS-O-antigen in improving surface wettability or swarming was also suggested for S. enterica serovar Typhimurium, Proteus mirabilis, and Serratia marcescens [4, 27, 58]. Similarly, the capsular polysaccharide of P. mirabilis, also called colony migration factor, plays a key role by enhancing medium surface fluidity [21, 43]. In B. subtilis, extracellular protease activity is also important for surface movement [8]. Finally, it has been shown that P. mirabilis swarming is stimulated by peptides and amino acids, possibly generated by the three broad specificity proteases present in this bacterium [18].

Some recent publications have used a mutant screening approach to identify genes involved in swarming in E. coli K-12 [26] and P. aeruginosa PAO1 [40] or a microarray-based approach for genes involved in surface motility in S. enterica serovar Typhymurium [62]. Mutations in genes belonging to various functional categories were identified and underscore the notion that swarming is a complex type of motility requiring a wide variety of cellular activities. In E. coli K-12, different components of the cell surface were found among the identified genes including the lipopolysaccharide and enterobacterial common antigen. These compounds are thought to have a role in reduction of surface tension.

The involvement of quorum sensing (QS) regulation in swarming has been reported for both biosurfactant production and swarmer cell differentiation (reviewed by Daniels et al. [10]). Involvement of LuxIR-type systems producing N-acylhomoserine lactones (AHLs) in swarming is mainly reported for regulation of biosurfactant production like serrawettin production in Serratia liquefaciens and rhamnolipid production in P. aeruginosa [14, 38, 59]. Evidence also exists for a role of AHL-mediated regulation of biosurfactant production and swarming behavior in S. marcescens and Burkholderia sp. [25, 31, 45]. In Pantoea stewartii subsp. stewartii, mutation of the esaIR QS system was reported to affect bacterial adhesion, swarming, and biofilm formation [24, 29, 61]. Finally, a mutation in the yenI gene, encoding the LuxI-type AHL-synthase of Yersinia enterocolitica, affects both swimming and swarming [3].

The gram-negative nitrogen-fixing soil bacterium Rhizobium etli is the bacterial symbiotic partner of the common bean plant. R. etli CNPAF512 produces several QS signal molecules via the cinIR and raiIR quorum sensing systems [11, 47]. It was previously demonstrated that R. etli CNPAF512 swarms and colonizes the surface of YEM soft agar. For swarming migration to occur, the cin QS system is required, as R. etli cinI and cinR mutants are no longer able to move over this solid surface. In contrast, they form a regular colony at the inoculation point [10]. Recent data published by our laboratory revealed that the CinI-made long chain AHLs have an additional function besides high level induction of the cin and rai quorum sensing systems [9]. These molecules were shown to possess significant surface activity and possibly induce liquid flows as a result of gradients in surface tension at biologically relevant concentrations, pointing to a direct role of these molecules as biosurfactants during swarming.

Little is known about factors involved in swarming in rhizobia. Swarming has only been demonstrated for Sinorhizobium meliloti and R. etli. In S. meliloti, a fadD mutant, altered in a gene encoding a long-chain fatty acyl-CoA ligase, displayed multicellular swarming behavior [55]. In this work, we initiated research toward the genetic determinants involved in swarming in R. etli CNPAF512 by analyzing the swarming behavior of a subset of 700 mutants and investigated their symbiotic performance and a possible link with the cin QS system.

Methods

Bacterial Strains and Plasmids

The bacterial strains and plasmids used in this work are listed in Table 1. The mutant strains isolated during the screening for mutants affected in swarming are represented in Table 2. R. etli strains were cultured in complex TY-medium (0.3% Yeast extract, 0.5% Tryptone, 7 mM CaCl2) at 30°C [35], on YEM soft agar (0.75%) plates for swarming experiments [9], and TY soft agar plates (0.2%) for swimming experiments. Escherichia coli strains were routinely grown at 37°C in Luria–Bertani (LB) medium [49]. Agrobacterium tumefaciens NT1 was grown in TY or in AB medium at 30°C [47]. Antibiotics were supplied at the following concentrations: ampicillin 100 μg ml−1, gentamicin 30 μg ml−1, kanamycin 30 μg ml−1, spectinomycin 50 μg ml−1, nalidixic acid 30 μg ml−1, neomycin 60 μg ml−1, or tetracycline 1 μg ml−1 (R. etli) or 10 μg ml−1 (E. coli). For A. tumefaciens, kanamycin was supplied at 50 μg ml−1 and carbenicillin at 100 μg ml−1. Triparental conjugations were done as described previously [12].

Table 1 Strains and plasmids used in this study
Full size table
Table 2 Mutants with an altered swarming pattern
Full size table

DNA Techniques

Standard techniques were used for DNA manipulations [49]. Restriction endonucleases (Roche Diagnostics) and ligase (Invitrogen) were used according to the manufacturer’s instructions. Total DNA from R. etli CNPAF512 was isolated using the Puregene® Genomic DNA purification kit (Gentra Systems). Plasmids were isolated from E. coli using GFX™ Micro Plasmid Prep Kit (Amersham). DNA fragments were isolated from agarose gels using Qiaquick Gel Extraction kit (Qiagen) (>4 kb) or Minelute Gel extraction Kit (Qiagen) (<4 kb). Polymerase chain reactions (PCR) were performed in a Px2 Thermal Cycler (Thermo Electron). Amplified PCR fragments were recovered from the PCR mix with the QIAquick PCR purification kit (Qiagen). Primers were obtained from Eurogentec. DNA probes for Southern hybridization were labeled with digoxigenin-dUTP using a random-primed labeling kit (Boehringer Manheim). Genomic DNA was isolated, digested with restriction enzymes, separated by electrophoresis, and blotted to Hybond N membrane (Amersham) by standard techniques. Prehybridization and hybridization were carried out at 68°C. Signals were detected with a chemiluminescence detection kit (Boehringer Manheim).

Motility Assays and Sequence Determination

Swarming assays were performed on YEM soft agar plates (0.75%) (containing 0.4 g/l MgSO4· 7H2O) as described in Daniels et al. [9]. For swimming, TY plates containing 0.2% agar were spot inoculated with bacteria previously grown on YEM agar plates. Each strain was tested in tenfold in three independent experiments. The plates were incubated at 30°C in a closed container before swimming behavior was checked. Images of representative surface motility plates were captured using a digital camera (Sony).

For the selected mutants, total DNA was digested overnight with the XhoI enzyme, leaving the transposon intact, religated, and transformed into E. coli. For each mutant, plasmid DNA of at least two independent colonies was submitted for sequencing. The DNA sequence of the genomic region flanking the 5′ end of the gusA gene was determined using the gusA primer RHI-628 (5′-CTCCTTAGCTAGTCAGGTACCG-3′), whereas primer AZO-497 (5′-GTCTGACGCTCAGTGGAAC-3′) was used for the region 3′ of the transposon. Southern hybridization using a gusA probe was performed to confirm insertion of the transposon in a single locus. The putative open reading frame (ORF)s located on the flanking sequences were identified by BlastX searches [2].

Determination of Quorum Sensing Super-regulation

To determine if AHL production was affected in the mutant strains, the A. tumefaciens tra reporter system was used. The reporter strain was grown overnight and diluted 40-fold into TY soft medium (0.75%). The mutants were spot inoculated in the middle of the agar plates on the surface and incubated for 3 days. The A. tumefaciens strain is deficient in its own AHLs, but exogenously produced AHLs can induce a traG-lacZ fusion, observable by a color reaction in the presence of the substrate X-gal [41]. AHL production was monitored by the examination of the diameter of the blue halo observed.

Expression Analysis

To investigate a possible regulation of the identified genes by AHLs, all selected transposon mutants were additionally mutated in the cinI gene. Inactivation of cinI was carried out with pCMPG8799. cinI mutants were selected by screening for loss of bacteriostatic activity toward the sensitive strain Rhizobium leguminosarum bv. viciae 248 and the loss of activation of the A. tumefaciens tra reporter system [11]. Mutants were confirmed by PCR analysis. Quorum sensing signal molecules were isolated from cultures of wild-type R. etli or A. tumefaciens NT1 carrying the cinIR locus on pFAJ4013 as described by Daniels et al. [9]. Overnight cultures of the mutants were diluted 40-fold in fresh medium containing the dissolved QS signal molecules extract (final concentration 1×) or a control. These cultures were incubated in microtiter dishes sealed with a breathable membrane (Greiner Bio-One) for 24 or 40 h after which expression was analyzed. Quantitative analysis of GusA activity was carried out in microtiter dishes with p-nitrophenyl-β-d-glucuronide (pNPG) as the substrate by the method of Miller. GusA activity was examined in VERSAmax microplate reader (Molecular Devices) [9].

Growth Analysis

Overnight cultures of R. etli strains were washed twice in 10 mM MgSO4, brought to an OD600 of 0.5, and subsequently diluted 6,000-fold in TY-medium. Three hundred microliters of the cultures, five repetitions for each condition, were introduced in the wells of sterile Honeycomb plates. The OD600 value was recorded every hour in a Bioscreen C Microbiology Workstation (Labsystems Oy, Zellik, Belgium) [37].

Plant Experiments

Seeds of Phaseolus vulgaris cv. Limburgse vroege were sterilized and germinated as previously described [36] and plants were inoculated with 100 μl of an overnight bacterial culture and incubated as described by Dombrecht et al. [13]. For each R. etli strain, at least ten plants were inoculated. Nitrogenase activity was determined by measuring acetylene reduction activity (ARA) of nodulated roots in a closed vessel 3 weeks after inoculation [13].

Electron Microscopy

A formvar (0.7%) carbon-coated grid (50 mesh, Cu-OD-841005, VECO®) was placed on the surface of a swarmer plate for 10 s; the grid was placed on a drop of phosphotungsten acid (0.25%, pH 7) for 30 s for staining, and excess liquid was drained. Samples were examined with a Philips® EM 208 S transmission electron microscope at 80 kV. Images were digitized using the SIS® image analysis system.

Results

Identification of Mutants with an Altered Swarming Behavior

Swarming in rhizobia is a relatively recently described phenomenon [9, 10, 55]. Besides the cinIR system, little is known about other genes involved in R. etli and the contribution of this form of motility to the symbiotic relationship with the host plant. To identify a first set of genes that are involved in the swarming behavior of R. etli CNPAF512, a group of 700 mutants were analyzed with respect to their swarming behavior. The 700 mutants in this library were previously selected in a screening of a mTn5gusA-oriV library in wild-type R. etli CNPAF512 for genes potentially involved in the symbiotic process (M. Moris, unpublished). Moreover, as these mutants were isolated based on differential GusA activity, the orientation of the gusA gene in these mutants enables us to investigate if the defect in swarming is related to quorum sensing-mediated gene regulation.

First, the mutants were grown overnight in microtiter dishes containing TY-medium. After washing the cultures with liquid YEM medium, 1.2 μl was spot inoculated on a swarm plate. After 3 days incubation, swarming behavior was monitored. The wild type and quorum sensing mutants, FAJ1328 and FAJ1329 (swarming positive), and FAJ4006, FAJ4013 (swarming negative) were included as controls for swarming conditions in every round of the screening.

Mutants, 183, showing no swarming behavior or a swarming behavior different from the wild type (e.g., swarming diameter, presence of extrusions, etc.) were selected. These mutants were retested after growth in microtiter dishes. From this analysis, 60 strains were selected for detailed analysis, and overnight cultures in 5 ml TY-medium were brought to equal optical densities and used to inoculate swarmer plates in triplicate. The latter step of the screening was repeated independently three times.

Finally, 24 mutants consistently showing no or an atypical (e.g., diameter, extrusions) swarming pattern in all repeats of the experiment were subjected to sequence analysis. The results are summarized in Table 2. Upon sequence analysis, mutants carrying the transposon in ypch00541 (CMPG8799) and ctaD (CMPG8790) were found to be present twice in the selection. The 22 different mutants were classified in three groups, based on the shape of the colony edge. Twelve mutants were identified that were completely swarming deficient (smooth edge; group 1), five mutants showed an atypical swarming pattern (no completely smooth edge with local extrusions; group 2), and five mutants displayed an intermediate swarming phenotype (less pronounced scalloping of the colony edge compared to the wild type; group 3). Representative pictures of each group are shown in Fig. 1.

Figure 1
figure 1

Swarming behavior of selected mutants. Representatives of the different groups are shown including the wild type (A) and a mutant designated as swarming positive (B). C–G represent different mutants classified as swarming negative in group 1 (C: CMPG8799; D: CMPG8788; E: CMPG8797; F: CMPG8791; G: CMPG8796). Three mutants showing local extrusions (classified in group 2) are represented in H to J (H: CMPG8787; I: CMPG8783: J: CMPG8786). Finally, CMPG8777 (K) and CMPG8781 (L) represent mutants classified in group 3.

Full size image

Swimming Behavior of Swarm Mutants

Genes not only involved in swarming but also important for swimming behavior were identified by performing swimming tests. Most mutants displayed a wild-type swimming pattern (Fig. 2A and B). As expected, mutants knocked-out in components of the flagellar system (CMPG8787, CMPG8786 see Fig. 2D, CMPG8785) were also defective in swimming, whereas CMPG8796 displayed a reduced swimming behavior as shown by the smaller swimming diameter. As the latter strain also displays a growth delay, we cannot exclude that the motility phenotype of this mutant is due to a more general defect. Therefore, to exclude that the swarming behavior is due to a growth defect, Bioscreen C curves of all mutants were obtained for growth (data not shown). Most strains displayed a wild-type growth pattern, three mutants displayed a small growth retardation (indicated as ± in Table 2), while three mutants (CMPG8796, CMPG8790, CMPG8777) were more significantly retarded. As this could also affect the outcome of the swarming assay, (e.g., by a defect in the energy supply as can be expected for CMPG8790 mutated in cytochrome C oxidase subunit 1), these three strains will not be further discussed.

Figure 2
figure 2

Swimming behavior. Swimming behavior observed on 0.2% agar plates. Most mutants showed a wild-type swimming behavior (as shown in A and B), reaching the edge of the plates after approximately 3 days. Mutant CMPG8796 displayed a strongly reduced diameter of swimming after 3 days, while the mutants in genes involved in synthesis of flagella like CMPG8786 (D) displayed only growth at the inoculation point.

Full size image

Symbiotic Phenotype of Swarm Mutants

The swarmer mutants were evaluated for symbiotic nitrogen fixation capacity by measuring ARA values of inoculated plants. Eleven and nine strains displayed reduced or wild-type nitrogen fixation levels, respectively. Mutant CMPG8794 was Nod-. A summary of the results is presented in Table 2.

R. etli Swarmer Cells are Hyperflagellated

To obtain a more detailed view on the morphological differentiation associated with the swarming phenomenon in R. etli, we analyzed transmission electron micrographs of cells, directly taken from the edge of a wild-type R. etli-swarming colony. In R. etli, swarming coincides with hyperflagellation as found in other bacteria (Fig. 3). The hyperflagellation as obtained during swarming was not observed in bacterial cultures grown in broth (data not shown).

Figure 3
figure 3

Microscopic observation of swarmer cells. Electron micrographic picture of wild-type R. etli, isolated from the edge of a swarming colony and colored with phosphotungsten acid (0.25%, pH 7). The differentiated swarmer cells are hyperflagellated (bar, 1,000 nm).

Full size image

Relationship with Quorum Sensing

As cin mutants are no longer capable of swarming, and this phenotype can be restored by complementation, we determined if any of the mutants identified here is impaired in AHL production. Therefore, mutants were spot inoculated on TY soft agar plates containing the A. tumefaciens reporter strain. After 3 days of incubation at 30°C, the diameter of the blue halo, representing diffused AHL molecules produced by the mutant that activate the reporter, was compared with that observed for the wild-type strain. All strains, except mutant CMPG8788 (cinI), produced a blue halo indistinguishable from that of the wild type in different repeats of the experiment (data not shown). This indicates that the swarming phenotype is not caused by reduced production or secretion of AHL molecules and that no super-regulators of the R. etli QS systems were present, at least not operating in the conditions tested.

To determine whether any of the genes is regulated by cinI, the selected mutants were mutagenized in cinI by using pCMPG8799. Expression tests were carried out to determine the expression of the gusA fusions in the presence of QS signal molecule extract obtained from wild-type R. etli CNPAF512, from A. tumefaciens NT1 carrying cinIR on a plasmid and from A. tumefaciens NT1 negative control. Expression of none of the genes identified was found to be significantly affected by mutation of cinI (data not shown).

Discussion

It was previously shown that R. etli CNPAF512 swarms and promotes surface colonization of YEM soft agar plates [10]. To better understand the mechanism of swarming in R. etli, a first subset of mTn5gusA-oriV mutants of R. etli was analyzed for their swarming phenotype. Twenty-two different mutants showed no or an altered swarming behavior. These are summarized in Table 2. The effect of the transposon insertion is due to inactivation of the gene mentioned in the table or to a possible polar effect. As these 700 mutants were previously isolated based on differential expression patterns in micro-aerobic conditions or in the presence of nodule extract, the number of mutants identified here is probably not representative for the percentage of genes involved in swarming. Among the hitherto identified genes, genes encoding elements of the flagellar system and modifying the EPS matrix surrounding the bacteria were identified. Overall, mutants isolated here could be affected in the swarmer cell differentiation, in surface translocation, or both. Detailed microscopic observation of the bacteria will enable us to distinguish between these possibilities. As some of the mutants isolated also display a growth defect in rich medium, this might affect the outcome and timing of swarmer cell differentiation or surface translocation, leading to the observed phenotypes. Finally, as most mutants identified in this screening still await further confirmation, we cannot rule out at this stage that some of the observed effects are due to secondary site mutations.

The results of the ARA measurements reveal that there is no direct link between the absolute nitrogen fixation capacity and swarming capacity. This raises the question about the possible role of swarming in the bacteria–plant interaction (reviewed by [10, 44]). In general, the phenomenon of swarming is often linked with biofilm formation, although the relationship between these phenomena appears to be rather dual. Recent research in P. aeruginosa revealed that early in biofilm formation, the extent of swarming, which depends on the conditions used, determines the structure of the biofilm formed. Moreover, the nutritional environment was found to influence the functions required for swarming including quorum sensing-mediated regulation [54]. Interestingly, attachment of bacteria to plants proceeds through a similar mechanism as observed for initiation of biofilm formation and the direct observation of bacteria adhered to plant cells often reveals multicellular assemblies variably described as microcolonies, aggregates, or cell clusters. These resemble each other in the fact that they all consist of bacteria embedded in a matrix on a surface [44].

The main role of surface motility in P. aeruginosa biofilm formation was proposed to be after the initial contact is established, when cells move over the surface, where they finally aggregate and form microcolonies [39, 42]. In these early steps of biofilm formation, bacteria undergo a variety of metabolic changes after attachment including differential expression of proteins involved in amino acid metabolism, membrane proteins involved in transport processes, and proteins necessary for the production of extracellular polymers and organelles involved in motility such as flagella and pili [51]. Metabolic differentiation in swarmer cells and genes specifically induced by surface growth are also reported for Salmonella [28, 63]. However, in other bacteria such as B. cepacia, no direct link between swarming and biofilm formation was observed, leading to the suggestion that swarming per se is not required for biofilm formation [25]. Moreover, a growing number of biosurfactants, important for surface translocation, are reported to be involved in biofilm dispersion, suggesting also a possible role for swarming in the dispersal of biofilms (reviewed by [6, 10]).

Evidence that surface movement is important during the colonization process, especially of distal parts of the roots, is provided by research on P. fluorescens F113 during the colonization of the alfalfa rhizosphere [50]. During the colonization process, P. fluorescens F113 undergoes phenotypic variation resulting in three phase variants, C, F, and S, with C representing the wild-type behavior. The F and S variants could swim faster and swarmed in conditions that did not allow swarming in the wild type. These variants preferentially colonized distal parts of the roots, which are not easily reached by the wild-type strain. The observation that P. fluorescens mutants in site-specific recombinases, responsible for the generation of subpopulations, are seriously affected in competitive rhizosphere colonization [33], demonstrates the importance of the generation of these specialized swarming subpopulations in colonization. As plant inoculation is carried out here with pure cultures directly inoculated on the developing root, we do not have information about the effect of mutants defective in swarming on the colonization of roots especially when they have to compete with wild-type bacteria. Careful microscopic observation of the distribution pattern of bacteria of mixed cultures and competition experiments might help to extend our understanding of the importance of swarming during symbiosis in R. etli.

The identification of a cinI mutant (CMPG8788) in the screening confirms previous observations that the cinI system is required for R. etli CNPAF512 swarming. As swarming is a flagellum-driven movement, the observation that three of the mutants identified here carry the transposon in genes important for accurate flagellum synthesis is not surprising. Motility, in general, is affected in these mutants, which is further confirmed by their defect in swimming behavior. CMPG8787 carries the transposon in the flbT gene, most similar to S. meliloti flbT (SMc03051) (85% amino acid identity). Study of the function of this gene has mainly been carried out for the corresponding gene in Caulobacter crescentus. In the latter bacterium, it was shown that the flbT gene product functions as a negative regulator of flagellin expression in the absence of flagellum assembly [32]. CMPG8785 carries the insertion in flgE, probably encoding the structural protein of the hook that connects the basal body of the flagellum with the filament. In general, the hook is important in bacterial flagella because it allows the synchronous rotation of several filaments driven by their motors in a bundle formed behind the cell (swimming) [48]. As this gene is predicted to be co-transcribed with flbT and other genes, the observed phenotypes of these two mutants can be due to polar effects on downstream genes in the operon. Finally, in mutant CMPG8786, the transposon is inserted in flaCch1, the first of four genes encoding putative flagellin subunits. The gene cluster flaA–flaB–flaD–flaC (Smc03037–Smc03040) is found at the corresponding position in the chemotaxis, flagellar, and motility region of the S. meliloti genome [56]. Generally, the flagellar filament consists of an assembly of about 20,000 flagellin subunits, whose molecular mass typically ranges from 40 to 60 kDa [30]. Mutational analyses revealed that in S. meliloti, flagellin A is the principal, absolutely essential subunit but that in addition, at least one of the secondary flagellin species is needed for assembling a functional filament [52]. No information on the importance of different potential fla genes in R. etli CNPAF512 is available, but in the R. etli CFN42 genome, additional genes encoding putative flagellin subunits have been annotated, and it would be interesting to test whether these play more specific roles in the adaptation of flagella toward different forms of motility like swimming and swarming.

CMPG8799 is mutated in a gene encoding a probable polysaccharide export system protein. This protein contains the characteristics of the Poly_export Pfam family (PF02563) of periplasmic proteins involved in polysaccharide biosynthesis and/or export. The gene is most similar to S. meliloti SMc01794. As (secreted) polysaccharides are a component of the extracellular slime, important for the physical movement of the cells, and possibly also acting as a sink for the accumulation of signals involved in swarming and QS, the observed effect on swarming is not unexpected. However, no information is available at the moment about the precise polysaccharide(s) synthesized and/or secreted by the system involving this gene in R. etli CNPAF512.

In CMPG8783, the transposon is inserted in a gene designated plyA1 and encoding a polysaccharidase gene. These genes have been mainly studied in R. leguminosarum. Two similar glycanases, PlyA and PlyB, capable of degrading EPS and carboxymethyl cellulose, were described in R. leguminosarum. These genes are secreted via the PrsDE-encoded type I secretion complex that is also required for secretion of NodO and several other proteins [16, 17]. The plyA gene is located immediately upstream of prsDE and does not greatly affect EPS processing in the conditions tested unless the protein is overexpressed. PlyB is encoded elsewhere on the chromosome, and the plyB mutant is severely affected in the degradation of EPS, forms sticky colonies, and cultures were twice as viscous compared to the wild type [16]. In R. etli CFN42, three genes are designated plyA based on their similarity with the plyA gene of R. leguminosarum. Of these genes, plyA3 is also located upstream of the prsDE genes. In the mutant identified here, the transposon is located in plyA1. Given that this mutant also forms very sticky colonies on plate, the gene interrupted here might correspond to R. leguminosarum plyB. The fact that a mutation of this gene increases the stickiness of the matrix in which the bacteria are embedded explains the observed effect on swarming.

Finally, a number of genes involved in diverse metabolic pathways were identified. A number of these genes are related to or positioned in operons containing genes possibly involved in the metabolism of arginine and derivatives (including polyamines) or transport of these compounds (e.g., CMPG8781 and CMPG8791). In P. mirabilis, it was recently shown that the polyamine putrescine functions indeed as an extracellular signal for swarming, and mutants in speB (encoding SpeB, a protein involved in the conversion of agmatine to putrescine) display reduced swarming differentiation and migration [57]. The mechanism for the response to putrescine is currently unknown. However, as putrescine has been shown to form a complex with galacturonic acid in P. mirabilis LPS, this might contribute to efficient cell translocation [57]. Little is known about the role of different polyamines in rhizobia, although it was previously demonstrated that homospermidine is the polyamine present in highest concentration in all rhizobial strains tested [19]. In the related bacterium, A. tumefaciens, mutation of the att locus, in which most genes display the highest similarity with proteins involved in the transport of spermidine and putrescine, resulted in no or strongly reduced attachment to carrot suspension culture cells. This could be reversed by addition of conditioned medium [34]. Finally, peptides and glutamine have also been identified as being involved in P. mirabilis swarmer cell differentiation [1]. However, the positive effect of glutamine on P. mirabilis swarming has only been studied on minimal media.

In contrast to other bacteria, no obvious genes involved in biosurfactant production were identified in this screening. This is in agreement with recent research demonstrating that long chain AHLs, produced by CinI, possess significant surfactant activity. This fact may account for the swarming phenotype upon mutation of the cin system [9].