The exopolysaccharide of Rhizobium sp. YAS34 is not necessary for biofilm formation on Arabidopsis thaliana and Brassica napus roots but contributes to root colonization

Microbial exopolysaccharides (EPSs) play key roles in plant–microbe interactions, such as biofilm formation on plant roots and legume nodulation by rhizobia. Here, we focused on the function of an EPS produced by Rhizobium sp. YAS34 in the colonization and biofilm formation on non-legume plant roots (Arabidopsis thaliana and Brassica napus). Using random transposon mutagenesis, we isolated an EPS-deficient mutant of strain YAS34 impaired in a glycosyltransferase gene (gta). Wild type and mutant strains were tagged with a plasmid-born GFP and, for the first time, the EPS produced by the wild-type strain was seen in the rhizosphere using selective carbohydrate probing with a fluorescent lectin and confocal laser-scanning microscopy. We show for the fist time that Rhizobium forms biofilms on roots of non-legumes, independently of the EPS synthesis. When produced by strain YAS34 wild type, EPS is targeted at specific parts of the plant root system. Nutrient fluctuations, root exudates and bacterial growth phase can account for such a production pattern. The EPS synthesis in Rhizobium sp. YAS34 is not essential for biofilm formation on roots, but is critical to colonization of the basal part of the root system and increasing the stability of root-adhering soil. Thus, in Rhizobium sp. YAS34 and non-legume interactions, microbial EPS is implicated in root–soil interface, root colonization, but not in biofilm formation.


Introduction
Bacteria develop on plant roots as isolated cells, microcolonies and biofilms (Morris and Monier, 2003). Biofilms are assemblages of microbial cells that adhere to solid surfaces and are enmeshed in a self-produced extracellular matrix (Caldwell, 1995;Costerton et al., 1995). These structures provide cells with controlled living conditions that differ from the surrounding environment. The biofilm growth mode is a strategy for bacteria to invade their host in chronic versus acute infection (Furukawa et al., 2006).
The roles of EPS in the formation of biofilms on plant roots have principally been assessed using EPS-deficient or -overproducer mutants. However, in many studies, biofilms were seen on plant roots which were incubated in bacterial suspensions for short periods of time (Bianciotto et al., 2001;Bais et al., 2004;Walker et al., 2004;Matthysse et al., 2005;Timmusk et al., 2005;Fujishige et al., 2006). Moreover, the occurrence of EPS in bacterial root-biofilm matrix has never been seen by specific assays and on the roots of plants grown in soil and through longstanding plant-bacteria interactions. Yet, the development of root system and the soil matrix are environmental features that could influence the formation of biofilms on roots and the presence of EPS in the biofilm network.
In this study, we focused on the colonization of Arabidopsis thaliana and Brassica napus by an EPS-producing rhizobacterium Rhizobium sp. YAS34 (Alami et al., 2000), and the contribution of EPS synthesis to root colonization and biofilm formation using an EPS-deficient mutant of this strain. We have used confocal laser-scanning microscopy (CLSM) together with gfp tagging of bacterial cells and recognition-based selective probing of carbohydrates with lectins to target the EPS matrix. So far, lectins have never been used as reporters of carbohydrate and EPS in biofilms developed on plant roots and in the rhizosphere.
Rhizobium sp. YAS34 colonizes the whole-root system of A. thaliana and rapeseed plants as isolated cells, microcolonies and biofilms. This is the first evidence that Rhizobium forms biofilms on non-legume plant roots. We also show that at specific locations of the root system, the matrix of the Rhizobium sp. YAS34 biofilm is made of a bacteria-self-produced EPS. This bacterial EPS is not essential for biofilm formation on plant roots, but contributes to the colonization of specific zones in relation with nutrients availability.

Selection of MSDGT, an EPS-deficient mutant of Rhizobium sp. YAS34 affected in a glycosyltransferase gene
Using Tn5-based random mutagenesis in Rhizobium sp. YAS34 wild type (wt), the MSDGT mutant strain was isolated as small non-mucoid colonies on agar plates containing RCV medium supplemented with glucose as sole carbon source (Fig. 1B). Sequencing of the chromosomal regions flanking the transposon showed that the interrupted gene encoded a putative glycosyltransferase named gta (Accession Number EU184019). We identified a putative complete gene including start and stop codons and a ribosome binding site. This gene was expressed in MSDGT from the inducible P lac promoter of the plasmid pBBR1-MCS3 in order to complement the mutation. This complemented mutant showed a mucoid phenotype (Fig. 1C) identical to that of the wt strain ( Fig. 1A) while MSDGT transformed with the empty plasmid pBBR1-MCS3 grew as non-mucoid colonies (Fig. 1D).
The EPS production in the parental, mutant and complemented strains was quantified by ethanolprecipitation of cultures in stationary phase and expressed as lyophilized weight per litre. Rhizobium sp.
YAS34 wt produced up to 318 Ϯ 32 mg l -1 of EPS whereas only a small amount of material (3 Ϯ 3 mg l -1 ) could be isolated from cultures of the mutant using the same procedure. The complemented strain produced EPS (70 Ϯ 7 mg l -1 ) and this production was enhanced to 135 Ϯ 33 mg l -1 upon IPTG induction. Chemical structure of the isolated EPSs was confirmed by 1 H NMR (data not shown), which was in agreement with the previously described EPS structure (Villain-Simonnet et al., 2000a).
Our results show that the gta gene was correctly predicted as encoding a glycosyltransferase in the YAS34 genome and could be expressed from pBBR1-MCS3 under the control of Plac. Using BLAST homology searches, the gta gene product composed of 402 aa was found to be similar to a glycosyltransferase from Novosphingobium aromaticivorans (45% identity at the protein level) and Bradryrhizobium (43%). In the CAZy database (http:// www.cazy.org), such proteins are grouped in the GT4 family based on structural similarities, and mainly exhibit an alpha retention mechanism for sugar linkage. It is therefore likely that gta encodes a glucosyltransferase or a galactosyltransferase to add either glucose or galactose linked in alpha configuration in the monomeric structure of EPS from Rhizobium sp. YAS34 wt (B. Henrissat, pers. com.). We found that expression of this gene is essential for EPS production by this strain. Among the sequences of glycosyltransferases found most similar to the gta gene product in BLAST analysis, none have been shown to be functional in vivo. This result constitutes the first experimental evidence for in vivo activity of such glycosyltransferases.

Biofilm formation by Rhizobium sp. YAS34 wt and its mutant MSDGT
We tested the ability of Rhizobium sp. YAS34 wt and its mutant MSDGT to form biofilms in polypropylene tubes using a culture medium favourable for EPS synthesis with a crystal violet (CV) stain-based assay as described by O'Toole and Kolter (1998). First, YAS34 wt was compared with Pseudomonas aeruginosa (PAO1), which is a model organism for biofilm production. Levels of surfaceattached cells of strain YAS34 were similar to those of PAO1 (data not shown). This result indicates that strain YAS34 wt is able to form biofilm when grown statically in polypropylene tubes. The EPS-deficient mutant, MSDGT, showed identical levels of surface attachment to those seen in the parental strain (data not shown). Thus, EPS synthesis is not essential to that process, in our experimental conditions. However, the mutant could form biofilms at a slower rate than the parent strain and reached wt levels at the end of the experiment. Our results do not exclude that EPS could contribute to the kinetics of biofilm formation.

Evidence for EPS synthesis in Rhizobium sp. YAS34 wt colonies growing on agar
The growth of bacteria on agar surfaces better mimics the conditions that bacteria experience in habitats such as soil, where water availability is influenced by the solute and matric potentials (Chang and Halverson, 2003). Rhizobium sp. YAS34 wt forms mucoid colonies on agar medium ( Fig. 1A) in contrast to the EPS-deficient mutant that grows as small and rough colonies (Fig. 1B). To visualize the EPS matrix in these colonies by fluorescence microscopy, we modified the Thiery reaction (Thiery, 1967) generally used to localize carbohydrates in tissues, bacteria and soils (Erdos, 1986) by transmission electron microscopy. The substitution of thiocarbohydrazide and silver proteinate with a fluorescein thiosemicarbazide allowed fluorescent instead of electron-opaque staining of carbohydrates. This new method could allow fluorescent staining of an EPS for which no specific labelling is available. Linking of the fluorescent reagent with glycoconjugates of the cell membrane imaged the bacteria. Rhizobium sp. YAS34 wt mucoid colonies were stained with the modified Thiery reaction and examined in CLSM. Figure 2A shows an archetypical 3D structure of a biofilm with bacterial cells enmeshed in a glycoside-based matrix.

ConcanavalinA specifically binds to the EPS produced by Rhizobium sp. YAS34 wt in biofilms
Based on the chemical structure of the EPS isolated from strain YAS34 wt (Villain-Simonnet et al., 2000a,b), the internal a-D-glucopyranosyle was targeted as a binding site for a lectin, concanavalinA (ConA) (Elgavish and Shaanan, 1997).
Turbidity measurements showed the specific interaction between this lectin and EPS from YAS34 wt. Addition of ConA to a suspension of EPS from strain YAS34 wt increased the turbidity by 75% (data not shown), which reflected binding of the lectin to the EPS chains. These aggregates dissociated when excess D-glucose, but not D-galactose, was added. Addition of excess D-glucose or D-galactose to the EPS suspension previous to that of the lectin, respectively, inhibited or permitted recognition.
Colonies of YAS34 wt, MSDGT and its complemented mutant were grown on an agar-solidified mineral medium supplemented with glucose, suitable to EPS synthesis. Strains YAS34 wt and MSDGT expressed a plasmid-born GFP. Cells of the complemented mutant were detected by direct optical transmission. Bacterial colonies were labelled with Alexa660-ConA (1 mg ml -1 ) and examined in CLSM ( Fig. 2B-E). In Rhizobium sp. YAS34 wt and the complemented mutant, the bacteria were entangled within a three-dimensional network stained with ConA that revealed the presence of an EPS matrix (Fig. 2B, C and E). Yellow zones resulted from the superimposition of GFP-tagged bacteria and ConA-tagged matrix ( Fig. 2B and C). Bacteria were also localized in areas that were not stained by ConA ( Fig. 2B and C), showing that within the same population, some YAS34 wt cells did not produce EPS. Neither fluorescent staining nor EPS matrixes were detected when MSDGT colonies were labelled with Alexa660-ConA and observed by CLSM (Fig. 2D).
Labelling of Rhizobium sp. YAS34 wt mucoid colonies with fluorescent ConA was inhibited in the presence of D-glucose, but not in the presence of D-galactose. Addition of D-glucose, but not D-galactose, to previously stained colonies with fluorescent ConA switched off the fluorescence (data not shown).
Altogether, macroscopic and microscopic data show that the matrix of Rhizobium sp. YAS34 wt biofilm contains an EPS, which is specifically recognized in situ by ConA.

Rhizobium sp. YAS34 colonizes A. thaliana and B. napus roots independently of EPS production
The ability of EPS-producing and EPS-deficient strains of Rhizobium sp. YAS34 to colonize A. thaliana was investigated in vitro and in natural soil. Population sizes attached to the root system were determined by plate counts of serial dilutions of crushed roots. Per gram of dry root weight, 1.8 ¥ 10 9 Ϯ 2.0 ¥ 10 9 cfu of Rhizobium sp. YAS34 wt (4.1 ¥ 10 7 Ϯ 1.6 ¥ 10 7 cfu per root system) and 2.7 ¥ 10 10 Ϯ 1.3 ¥ 10 10 cfu of mutant MSDGT (6.0 ¥ 10 7 Ϯ 2.9 ¥ 10 7 cfu per root system) were counted on A.
thaliana ecotype Columbia roots of plantlets grown in vitro for 27 days. Similarly, 1.2 ¥ 10 10 Ϯ 6 ¥ 10 10 cfu of YAS34 wt (4.2 ¥ 10 8 Ϯ 1.9 ¥ 10 8 per root system) and 1.2 ¥ 10 10 Ϯ 4 ¥ 10 10 cfu of MSDGT (3.0 ¥ 10 8 Ϯ 0.9 ¥ 10 8 cfu per root system) were counted on B. napus roots grown in vitro for 21 days. This corresponded to approximately 13 and 15 bacterial generations, respectively, on A. thaliana and B. napus roots. In vitro, the seeds were inoculated with a bacterial suspension. Therefore, the colonization occurred by adhesion of bacterial cells to young roots emerging from the seed. During growth, the root apex tends to diverge from the inoculation spot, driving a primary population that must utilize root exudates as sole carbon and energy source to persist on the root system. As bacteria do not grow on the medium used for the plant culture in the absence of plant roots (data not shown), the growth and development of bacterial cells are only depen-dent on root exudation. The fact that we observed similar numbers of EPS-producing and EPS-deficient bacteria colonizing roots indicates that EPS production is not necessary for root colonization in vitro.
In soil, 2.3 ¥ 10 7 Ϯ 6.6 ¥ 10 7 cfu of strain YAS34 wt and 0.9 ¥ 10 7 Ϯ 4.2 ¥ 10 7 cfu of mutant MSDGT were counted per gram of dry root matter on A. thaliana ecotype Columbia plants grown for 90 days. Similarly, 1.1 ¥ 10 8 Ϯ 0.6 ¥ 10 8 cfu of YAS34 wt and 2.9 ¥ 10 7 Ϯ 1.3 ¥ 10 7 cfu of MSDGT were counted per gram of dry root matter on B. napus grown for 33 days. The population of strains YAS34 wt and MSDGT, respectively, accounted for 8% and 4% of the total culturable bacteria isolated from A. thaliana roots and for 5% and 1.3% of the total culturable bacteria isolated from B. napus roots. This result shows that Rhizobium sp. YAS34 wt and its mutant MSDGT compete with other soil bacteria and colonize non-legume plants such

Rhizobium sp. YAS34 forms biofilms on non-legume plant roots independently of EPS formation
Arabidopsis thaliana and rapeseed seeds were inoculated with Rhizobium sp. YAS34 wt or its mutant MSDGT expressing a plasmid-born GFP. Arabidopsis thaliana plants were grown in vitro and rapeseed in vitro and in non-sterilized field soil. As controls, axenic and noninoculated A. thaliana and rapeseed plants were grown in the same conditions. At various times of plant growth, plant roots were labelled with Texas Red (TR)-or Alexa660-ConA and observed in CLSM. Rhizobium sp. YAS34 wt and its mutant MSDGT showed a similar pattern of colonization on roots of A. thaliana and those of rapeseed grown in vitro. The bacteria were found at the surface of plant epidermal cells and in interstitial zones along the whole-root systems (basal, medium and apical parts) as isolated cells, microcolonies and three-dimensional patches of cells that fulfil the basic definition of biofilms (Figs. 3A-F and 4A-F). Several authors already reported the colonization (Alami et al., 2000;Fujishige et al., 2006) or the nodulation (Trinick and Hadobas, 1995) of non-legumes by Rhizobium species. However, this is the first visualization of a biofilm from Rhizobium formed on non-legume plant roots.
YAS34 wt showed similar profiles of colonization on rapeseed roots grown in vitro or in soil (Figs. 4A-C and 5A-C). This suggests that in vitro culture systems are valuable devices for investigating biofilm formation by Rhizobium sp. YAS34 wt on plant roots. The same observation was found for the mutant MSDGT except for the basal part of the root (Figs 4D-F and 5D-F). In this part of the root system, the mutant MSDGT colonized the rapeseed root as patches of cells when the plants were grown in vitro while isolated cells were frequent on the root plants grown in soil.
A very important result of this study is that EPS production by Rhizobium sp. YAS34 wt was not critical for biofilm formation either on solid surfaces (polypropylene tubes) or on Arabidopsis and rapeseed roots. Attachment to root cells is the first step of bacterial root colonization and formation of biofilms on root surface. This primary step is mediated by various processes and molecules secreted by both partners of the plant-microbe interaction (Rodriguez-Navarro et al., 2007). For instance, outer cell-surface proteins, capsular polysaccharides or cellulose have been shown to be involved in attachment of, respectively, Rahnella (Achouak et al., 1998) Pseudomonas, Azospirillum andAgrobacterium (Rodriguez-Navarro et al., 2007). In the case of Rhizobium sp. YAS34, we show that the mechanism for root cell attachment does not require EPS synthesis and must involve other processes similar to those described above.
Biofilms developed by Rhizobium sp. YAS34 on plant roots growing in vitro or in soil were less extended than those formed on roots subcultured and dipped in bacterial cell suspensions (Bianciotto et al., 2001;Bais et al., 2004;Walker et al., 2004;Matthysse et al., 2005;Timmusk et al., 2005;Fujishige et al., 2006). Yet, the context of root-bacteria interaction in liquid culture systems differs from the environmental conditions that rhizobacteria face in the soil and in the rhizosphere. Actually, water availability together with nutrients drastically influence the dynamics and the development of biofilms (Chang and Halverson, 2003).

ConcanavalinA binds to the EPS of Rhizobium sp. YAS34 produced in planta
For CLSM observations of roots, the tuning parameters were set as non-inoculated plant roots labelled with fluorescent derivatives of the lectin ConA (Figs 3D, 4D and 5D), and inoculated but unlabelled samples did not emit any fluorescence signal all along the root system (data not shown).
In our experiments, ConA did not bind to axenic roots (Figs 3G-I and 4G-I), root systems colonized by native soil rhizobacteria (Fig. 5G-I) or by mutant MSDGT (Figs 3D-F and 4D-F). ConA-labelled areas were exclusively visualized on the root system of plants inoculated by Rhizobium sp. YAS34 wt. As controls, the lectinlabelled roots giving rise to some fluorescence of ConAlabelled zones were incubated in a solution of D-glucose. The competition of D-glucose for which ConA has a high affinity totally switched off the detection of stained areas (data not shown).
Altogether, these results demonstrate that ConA binds to the EPS produced by Rhizobium sp. YAS34 wt on plant roots.

Exopolysaccharide localization in planta: EPS contributes to a better colonization of roots under natural conditions
On A. thaliana and rapeseed plant roots, the EPS production was dependent on plant age. On young plantlet roots of A. thaliana and rapeseeds (7 day old), inoculated with Rhizobium sp. YAS34 wt and grown in vitro or in soil, high densities of cells embedded in a translucent matrix were seen. However, labelling with ConA did not reveal the presence of an EPS matrix produced by Rhizobium sp. YAS34 wt (data not shown). Exopolysaccharides and biofilms were frequently seen on mature A. thaliana and rapeseed plant roots, respectively, grown for 15 or 30 days in vitro or in soil. In both culture conditions, biofilm formation on roots was repeatedly seen at the base of the root system. Figures 3A, 4A and 5A show three-dimensional EPS networks (surface area near 30 ¥ 30 mm over a depth of 10 mm) that entangled GFP-expressing bacteria. At the base of the root system, microcolonies of Rhizobium sp. YAS34 wt were also found embedded in a three-dimensional EPS network (Fig. 6A) that enclosed autofluorescent soil organic matter (Fig. 6B). Transmission optical microscopy showed translucent mineral particles identified as crystal quartz that firmly adhered to these aggregates and resisted extensive washings of the root (Fig. 6C). Fluorescent lectins and CLSM were used to localize carbohydrates in various environments (Holloway, 1997;Neu and Lawrence, 1997;Langille and Weiner, 1998;Neu and Lawrence, 1999;Johnsen et al., 2000;Strathmann et al., 2002;Wigglesworth-Cooksey and Cooksey, 2005). However, this is the first report of the use of lectins to localize EPS in biofilms developed in the rhizosphere.
In biofilms, the possible roles for EPS-based matrix encompass protection against nutrient and water starvation, predation and chemical stress such as antimicrobial agents and reactive oxygen species (Wolfaardt et al., 1999). On plant roots, the composition, amount and spatial localization of root exudates and nutrient uptake are heterogeneous in time and in space (Martin, 1971;Walker et al., 2003). Environmental parameters could influence the formation of biofilms by Rhizobium sp. YAS34 at specific parts of the root.
Localization of Rhizobium sp. YAS34 wt and its EPS were similar on the roots of rapeseed grown either in vitro or in natural soil. However, the mutant MSDGT showed a slightly reduced colonization of the basal part of the root system grown in soil compared with in vitro. This observation is supported by a minor but significant difference (P < 0.05) in bacterial population sizes determined on the roots of soil-grown rapeseed plants colonized by Rhizobium sp. YAS34 wt and its mutant MSDGT. The basal part of the root system has higher nutrient influx rates than all other root zones Rubio et al., 2001). Models for water uptake by plant roots show that water uptake sites are mostly regions near the base of the root system, close to the soil surface, and near the root tips (Roose and Fowler, 2004). Exopolysaccharide appears to be important for colonization of Rhizobium sp. YAS34 wt at the base of the root system, where an EPS-rich matrix could act as a trap for bacterial nutrition during water and solutes fluxes resulting from the sucking force of the plant.
Smaller biofilm-like structures of surface area 10 ¥ 10 mm by a depth of 10 mm were observed throughout the whole-root system at the emergence of secondary roots and on root hairs (Figs 3B, C and 4B). The structure of bacterial aggregates on plant roots (Figs 4A and 6A) closely resembled that of microcolonies grown on an agar matrix where EPS producer bacteria coexisted with non- Fig. 6. Development of Rhizobium sp. YAS34 wt biofilm on the root of rapeseed plant grown in soil using CLSM. Images represent the projection of Z-sections through 17 mm (1 mm step). A. Overlay of green and red channels emissions. Excitation at 488 and 645 nm allowed the localization of GFP-expressing bacteria and autofluorescence of root tissues in green and EPS bound to Alexa660-ConA (emission > 640 nm) in red. Some Protozoa were found grazing GFP-tagged bacteria. Scale bar = 20.0 mm. B. Overlay of the same zone, superimposing the signal obtained with excitation at 548 nm and emission between 585 and 640 nm. This allows imaging of the autofluorescent organic matter (shown in blue). Superimposition of artificial colours is given in overlapping circles. Scale bar = 20 mm. C. Observation of the same zone in transmission mode at z = 6 mm. Mineral particles (indicated by an arrow) were stacked to the root tissues and embedded into micro-aggregates. Scale bar = 20 mm. producer ones (Fig. 2B and C). The development of a lateral root generates fissures and releases galactosides that favour colonization by rhizobacteria (Bringhurst et al., 2001). Exopolysaccharide production at root cracking and on root hairs is therefore seen as expected in theses locations, but is not necessary to root colonization.
A high density of Rhizobium cells was often localized as microcolonies and monolayers on the apical part of A. thaliana and B. napus roots (Fig. 3C), and as multilayers at the apex (Vicre et al., 2005). However, EPS was rarely detected on these parts of the root system (Figs 4C and  5C). Carbon-rich material collected as exudates is predominantly released from root tip regions (McDougal and Rovira, 1970;Griffin, 1976;Darwent et al., 2003). However, in spite of high densities of bacterial cells in this part of the root, no EPS could be detected. Plant root tips release exudates such as glycosides (Haggquist et al., 1984;Lynch and Whipps, 1990). However, the hypothesis that glucose exudates could inhibit EPS interaction with the ConA lectin is rather inconsistent with the short halflives of this compound in the rhizosphere (Coody et al., 1986;Kuzyakov and Jones, 2006). In liquid culture, Rhizobium sp. YAS34 wt starts to produce minute amounts of its EPS from the middle of exponential phase (data not shown). We therefore suggest that in short-term interactions and hence on the root tip, the physiological state of Rhizobium sp. YAS34 wt cells could be similar to an early exponential phase of microbial growth.
These results corroborate work from Foster and colleagues which reports that mucigel matrix is of microbial origin at the base and of plant origin on the younger parts of wheat root (Foster and Rovira, 1978;Foster, 1981).

Exopolysaccharide synthesis by Rhizobium sp. YAS34 wt increases the stability of soil macro-aggregates
Brassica napus seeds were inoculated by Rhizobium sp. YAS34 wt and mutant MSDGT, and grown in soil for 33 days under controlled atmosphere. Inoculation of the two strains did not promote growth as non-inoculated control plants were found to have a similar mass of dry leaf and root (Fig. 7A). However, twice as much waterstable macro-aggregates were found in root-adhering soil inoculated with strain YAS34 wt compared with the EPSdeficient strain MSDGT or in non-inoculated control treatment (Fig. 7B). This result strongly supports a role for rhizobial polysaccharides in modifying the soil structure by cementing soil particles around the roots of nonlegumes (Alami et al., 2000). Together with the pictures of micro-aggregates stuck to the rhizoplane, these results show that the formation of biofilm by Rhizobium sp. YAS34 wt structures small root-adhering soil aggregates that resist more to water dispersion. These results definitely show a role of rhizobial polysaccharides in modify-ing the soil structure by cementing soil particles around the roots of non-legumes as hypothesized in Alami and colleagues (2000).
Soil structure determines the total volume of soil pores, their size distribution as well as their geometry and connectivity. The resulting properties of the soil and rhizosphere, such as aeration, resistance to root penetration, water reserve and therefore water and solute movements, are essential parameters that control plant growth. The stability of the soil structure is therefore one of the basic determinants of the soil and rhizosphere quality, if not of the ecosystem stability. Rhizobial EPS are essential in the successful symbiosis between rhizobia and legumes providing nitrogen to plants (Cheng and Walker, 1998;Fraysse et al., 2003). However, rhizobial populations persist owing to cells that do not participate in symbiosis but benefit from rhizodeposition for growth and survival (Bringhurst et al., 2001). In this study, we show that Rhizobium sp. YAS34 cells that produce EPS could be factors in promoting soil stability in the rhizosphere.
Rhizobium sp. YAS34 wt (Alami et al., 2000) and its EPSdeficient mutant MSDGT were labelled with GFP by triparental mating with Escherichia coli GM2163 (Marinus et al., 1983) carrying the pHC60 plasmid (Cheng and Walker, 1998) and E. coli LE392 carrying the helper plasmid pRK2013 (Figurski and Helinski, 1979) as described by Vicre and colleagues (2005). The constitutive expression of the GFP was stable for almost 3 months in soil, even in the absence of the antibiotic selection pressure (results not shown).

Construction of an EPS-deficient mutant of Rhizobium sp. YAS34
Mutagenesis in Rhizobium sp. YAS34 wt was carried out by random insertion of a Tn5 transposon from a non-replicative plasmid pRL1063a (Wolk et al., 1991). pRL1063a was transferred into YAS34 wt by triparental mating as described before and mutants by transposition were selected on TSBagar plates containing kanamycin at 50 mg ml -1 . Mutants were screened on RCV-agar plates supplemented with 2 g l -1 glucose to test for EPS production.
Genomic DNA from the YAS34 mutant was extracted, digested with ClaI (a restriction enzyme that does not cut within the Tn5 sequence) (NEB), ligated with T4 ligase (Roche Diagnostics) and transferred into E. coli DH5a (Hanahan, 1983). As the transposon carries an origin of replication (p15A), only the plasmids containing Tn5 and flanking regions from the YAS34 chromosome will replicate and maintain itself in E. coli. This plasmid was re-isolated and sequenced using unique primers OMP458 (5′-TACTAGATTCAATGCTATCAATTGAG-3′) and OMP459 (5′-AGGAGGTCACATGGAATATCAGAT-3′) directed outwards the Tn5 ends (van den Broek et al., 2003).

Complementation of MSDGT strain
Based on the sequence obtained for DNA flanking regions of the transposon in MSDGT, we identified a putative glycosyltransferase complete gene including start and stop codons and ribosome binding site. The gene sequence was submitted to GenBank under the Accession Number EU184019. Primers GlcTrXbaF (5′-CCCTATTCATCTAGAATGATAGGA CCAGATATG-3′) and GlcTrSacR (5′-ATATGAGAGCTCTTC ATCAAGTACCAGTCTGAA-3′) were designed to amplify this complete gene (small uppercase letters indicate start and stop codons and underlined sequence show putative rbs). Upon PCR, an expected 1.2 kb fragment was obtained and cloned in pBBR-MCS3 (Kovach et al., 1995) downstream of a Plac promoter using XbaI and SacI. The resulting plasmid, pBBR-gta, was transferred into MSDGT by triparental mating as described before using 15 mg ml -1 tetracycline for selection and allowing an IPTG-inducible expression of the recombinant gene.

Isolation of EPS from Rhizobium sp. YAS34 culture
The EPS-producing and EPS-deficient strains of Rhizobium sp. YAS34 were grown to stationary phase (OD600nm = 0.2) in TSB/10 and bacterial cells were removed by centrifugation at 6000 g for 15 min at 4°C. Supernatants containing the soluble polysaccharide were collected and 2.5 vols of icecold ethanol were added. The EPS precipitated in these conditions and was harvested by 30 min centrifugation at 6000 g at 4°C. Supernatants were discarded and the EPS was re-suspended in a minimal water volume to be frozen at -20°C. After lyophilization, the amount of EPS was weighted and expressed as mg per litre of culture.

Study of bacterial cell adhesion
To analyse the function of EPS in bacterial adhesion to solid surface and formation of biofilms, we used a CV stainingbased protocol (O'Toole and Kolter, 1998) developed on polypropylene tubes. Rhizobium sp. YAS34 wt and mutant MSDGT were grown statically to stationary phase in polypropylene tubes containing RCV medium supplemented with 2 g l -1 glucose. At an OD600nm of 0.2, 1 ml of 1% CV solution was added and tubes were incubated for 20 min. Tubes were emptied and washed three times with ultra-pure water. Stained biofilms were re-suspended in 1 ml of absolute ethanol and the OD590nm was measured in 96-well plates with 200 ml in each well. This OD value reflects the amount of surface-attached bacterial cells as described in O' Toole and Kolter (1998). Bacteria-free samples were stained using the same protocol and were taken as blanks.

Brassica napus and A. thaliana growth conditions
Seeds of B. napus cv. Drakkar and A. thaliana (Columbia ecotype) were surface-sterilized and sown in square plates (12 ¥ 12 cm or 25 ¥ 25 cm) containing sterile half-strength Hoagland nutrient solution (Arnon and Hoagland, 1940) solidified with Phytagel (Sigma, 7 g l -1 ) (Achouak et al., 2004). Each seed was inoculated with 5 ml of a Rhizobium sp. YAS34 suspension (about 10 6 cfu ml -1 ) grown to late exponential phase, washed twice with sterile KCl (0.85 g l -1 ) and then re-suspended in sterile ultra-pure water. On control plants, 5 ml of sterile ultra-pure water was added to the seeds. The plates were sealed with Micropore tape (3M, St Paul, MN, USA) and incubated vertically at 21°C with 16 h of light (100 mmol m -2 s -1 ) and 19°C for 8 h light off. Plantlets were harvested after 7, 15 and 21 days of culture for B. napus and 7, 15, 21 and 28 days for A. thaliana.
For plant culture in natural soil, we used the upper 30 cm of the soil profile of a typical Eutric cambisol (FAO-UNESCO classification) of INRA (Achouak et al., 2004). The last crop before sampling was wheat. Topsoil contained clay (17.5%), silt (53.0%) and sand (29.5%), organic carbon (1.35%) and nitrogen (0.12%). Water pH was 7.85 Ϯ 0.05. Soil was airdried, passed through 2 mm mesh sieve, hydrated to 13% of humidity and stored at room temperature for at least 10 days before bacterial inoculations and plant sowing. Soil was inoculated with Rhizobium sp. YAS34 cultures (wt and MSDGT mutant, 10 5 cfu g -1 of dry soil, final humidity 18%) and poured onto pots with a density of 1.2 (dry weight per volume of pot). Seeds of B. napus cv. Drakkar were surfacesterilized and sown in pots. Plantlets were grown in chambers with controlled atmosphere for 15, 21 and 33 days in the conditions described above. Seeds of A. thaliana ecotype Columbia were surfaced-sterilized, inoculated with 5 ¥ 10 3 cfu of the two YAS34 strains and germinated for 8 days on half-strength Hoagland medium solidified with 0.7% Phytagel. Plantlets were transferred to pre-inoculated soil and grown in chambers for 90 days under controlled atmosphere, 22°C with 8 h of light (120 mmol m -2 s -1 ) and 19°C for 16 h light off, 60-70% relative humidity.

Enumeration of bacterial populations developed on roots
Root-adhering bacteria were counted on the whole-root system. Three to five root systems per treatment were washed in 100 ml of ultra-pure water for roots of plantlets grown in natural soil and crushed. The bacterial populations were estimated by plating serial dilutions in KCl 0.85 g l -1 on TSB/10 agar containing tetracycline (15 mg ml -1 ) and counting of the cfu after 3 day incubation at 30°C.

Interaction between EPS of Rhizobium sp. YAS34 wt and ConA
Stock solutions of ConA (Sigma and Molecular Probes, 1 mg ml -1 ) were prepared in 0.2 mm filtered phosphate buffer saline (10 mM, pH 6.8), supplemented with MnCl2 and CaCl2 (0.1 mM each, Sigma), and stored at 4°C.
Interaction between the lectin and the EPS was evidenced by monitoring the timecourse increase of turbidity at 600 nm of 1 ml solution of EPS (1 mg ml -1 ) upon increasing lectin concentrations (up to 1 mg ml -1 ) in water or in PBS. Then, a solution of D-glucose or D-galactose (Sigma, 25 mg ml -1 ) was added and the optical density was time-monitored.

Imaging the EPS produced by Rhizobium sp. YAS34 wt with ConA
Colonies of Rhizobium sp. YAS34 expressing a plasmid-born GFP were scrapped from the agar surface and deposited on a slide. One hundred microlitres of tetramethylrhodamine isothiocyanate (TRITC)-or Texas Red (TR-ConA) (Molecular probes) solution (1 mg ml -1 and 0.1 mg ml -1 ) in PBS containing CaCl2 and MnCl2 (0.1 mM each) were added and the slide was stored 15 min in the dark. The excess solution was then discarded and the slide-rinsed in 40 ml of PBS for 15 min. In some experiments, slides were incubated in glucose (25 mg ml -1 ) before or after the addition of ConA.
Alternatively, Rhizobium sp. YAS34 colonies grown inside agar-solidified medium were directly labelled by addition of TR-ConA solution (1 mg ml -1 ) and interaction during 12 h at 4°C. The plate was rinsed twice and incubated overnight with water at 4°C, then directly observed under the microscope.
Roots of plantlets grown in vitro were labelled on the plates by pouring the lectin solution (0.1 mg ml -1 in PBS, MnCl2, CaCl2) all over the roots. Alternatively, a whole plant was removed from the plate and incubated into 10 ml of lectin solution for 30 min. For plants grown in soil, the plants were harvested, gently hand-shacked and rapidly immerged first into 40 ml of water to remove non-adhering soil, then in 40 ml of TRITC-or Alexa660-ConA (0.2 mg ml -1 in PBS, CaCl2, MnCl2 buffer) for 1 h. In some cases, bovine serum albumin (grade II, Sigma, 0.2 g l -1 ) was added in the staining solution to saturate unspecific sites such as clays and reduce unspecific binding of ConA.
Excess labelling solution was discarded or the plant was removed from the labelling solution. The roots were washed twice with sterile water, and incubated for 1 h in water or PBS buffer. Plant roots were observed directly on the growthmedium surface with coverslip or mounted between a slide and a coverslip, with citifluor (Interchim) or water.
As controls, non-inoculated plantlets were labelled with the procedure described above and inoculated plantlets were labelled in the same way except the addition of the fluorescent lectin.
Plantlet roots for which numerous zones of bacteria and EPS had been colocalized were incubated in the presence of a glucose solution (20 mg l -1 ) for 15 min and rinsed with water. The plantlets were observed again in the same scanning conditions.

Modified Thiery reaction for labelling of carbohydrates
This procedure was adapted from the Thiery reaction (Thiery, 1967), used to detect glycopolymers having 1,2-diol by transmission electron microscopy. The thiocarbohydrazide-silver and proteinate reagents were replaced by a fluorescent thiosemicarbazide (Sigma, 0.02% in sodium acetate solution, pH 5.5) that could be visualized in CLSM.

Microscopic observations
An Olympus CLSM system equipped with krypton-argon lasers (488, 568, 647 nm lines) and oil objectives was used for microscopic observations. Lectin tagged with FITC, TRITC-, TR-or Alexa660 were excited with 488 (green channel), 568 and 647 nm wavelengths (red channel) respectively. 488 and 568 nm scanning were acquired separately. Emissions were observed with appropriate filters (510-560 nm on green channel, 585-640 nm or > 660 nm on red channel). Samples were examined in the fluorescence and in the transmission mode. Axenic plants labelled with the fluorescent lectin and Rhizobium sp. YAS34 unlabelled inoculated plantlets were used as control samples. In each experiment, these control samples were observed by CLSM in order to determine acquisition parameters leading to an absence of fluorescent signal (i.e. dark images) on the channel used for the detection of the fluorescent lectin. Observation of unlabelled plantlets gives information on the autofluorescence level of the sample. Observation of axenic plantlets labelled with the fluorescent lectin reports on the selectivity of the probe. At least six plantlets (three plantlets from two independent experiments) of each treatment (inoculated or non-inoculated) were observed along the whole-root material, in order to achieve a valid and representative analysis of the samples.