V. Poinsot, Laboratoire des IMRCP, UMR5623 UPS/CNRS, 118 route de Narbonne, F-31062 Toulouse, France. Fax: + 33 5 61 55 81 55; E-mail: email@example.com
When the rhizosphere is nitrogen-starved, legumes and rhizobia (soil bacteria) enter into a symbiosis that enables the fixation of atmospheric dinitrogen. This implies a complex chemical dialogue between partners and drastic changes on both plant roots and bacteria. Several recent works pointed out the importance of rhizobial surface polysaccharides in the establishing of the highly specific symbiosis between symbionts. Exopolysaccharides appear to be essential for the early infection process. Lipopolysaccharides exhibit specific roles in the later stages of the nodulation processes such as the penetration of the infection thread into the cortical cells or the setting up of the nitrogen-fixing phenotype. More generally, even if active at different steps of the establishing of the symbiosis, all the polysaccharide classes seem to be involved in complex processes of plant defense inhibition that allow plant root invasion. Their chemistry is important for structural recognition as well as for physico-chemical properties.
Soil bacteria, collectively known as rhizobia, have the property of interacting with legume roots in nitrogen-starved environments, to form nodules where atmospheric dinitrogen can be fixed. The establishment of this symbiosis is based on a complex molecular dialogue, starting with secretion of flavonoids by the host plant roots. At least one of the roles of flavonoids is to activate the products of rhizobial regulatory nodD genes [1,2]. The interaction between the two partners seems to begin at a distance in the rhizosphere, even before rhizobia bind to the legume root hairs, and induce important morphogenic changes on host plant roots: a successful association provokes root hair deformation (Had phenotype) and afterwards root hair curling (Hac). Bacterial penetration (Inf) generally occurs via an infection thread (Thr) while cortical cells divide to generate the nodule primordium (Nod) . Bacteria are released at the tip of the infection thread. They are endocytosed by the plant cell and become surrounded by a plant membrane. In a later stage, the bacteria, still surrounded by the peribacteroid membrane of plant origin, differentiate into bacteroids expressing the nitrogenase (Nif) [4,5]. Two types of nodule developmental patterns are reported. In indeterminate nodules, after passing through the outer cortical cells, the infection threads have to reach the inner root cortex where the meristematic activity is initiated. Cell division originates in the pericycle and the so-formed persistent meristem divides constantly. Although the meristem remains uninfected, infection threads continuously invade the proximal cells. Consequently, all stages of differentiation are present in the same nodule and therefore the central tissue can be divided into adjacent zones representing successive stages of development . In the determinate nodule type, meristematic activity was temporary and initiated in the outer cortex. The bacteria mainly spread by the division of already infected outer cortical cells. Therefore, nodule cells in the central tissue are approximately in the same developmental stage.
These processes are partner-specific. This means that generally a given Rhizobium species can only nodulate a limited and defined range of legumes. One important key of the bacteria/plant interaction is a secreted rhizobial signal called Nod factor, the biosynthesis of which is induced by plant flavonoids . Nod factors are a family of lipochitooligosaccharides exhibiting mostly tetrameric or pentameric N-acetylglucosamine backbones, always substituted at the nonreducing end by a fatty acid. The Nod factors bear various substituents at both extremities, ranging from acetyl- to fucosyl-, and sometimes a sulfate at the nonreducing end. Whether structural variations involving these substitutions are mainly what control the host specificity remains controversial . Nod factors alone (without the bacteria) initiate the root hair deformations (Had+, Hac+) and in rare cases nodule primordia, but are not sufficient to produce the Nod or Inf phenotype. As discussed later, other bacterial carbohydrates, such as exopolysaccharides (EPSs) and capsular polysaccharides (KPSs) are necessary for establishing this kind of morphogenesis. More recent studies on polysaccharides suggest in some case lipopolysaccharides (LPSs) may play a role in the later stages leading to the nitrogen-fixing phenotype (Fix+) [9,10]. Rhizobial polysaccharides in plant–microbe symbiotic interactions are of continued interest because the exact mechanisms of symbiotic development are still unclear. Since the reviews published in 1999 and 2000 by Carlson et al.  and Spaink , respectively, – reporting principally on the biosynthesis of the surface polysaccharides – many papers have been published on mutations of genes coding for rhizobial polysaccharides and on their biological impact on a host/guest couple. It appears necessary to make an overview of the recently accumulated knowledge about the biological impact of the different bacterial polysaccharides during the establishment of symbiosis.
The most significant progress made in understanding the symbiotic role of EPSs, KPSs and LPSs produced by different rhizobial strains will be described here. As a conclusion, a short discussion will present a picture of the involvement of the different polysaccharide classes.
Relevance of exopolysaccharides (EPSs)
EPSs are abundant extracellular products accumulated on the cell surface and secreted into the cell's surroundings. They demonstrate a number of nonspecific functions such as protection against environmental stress, attachment to surfaces or nutrient gathering. However they also seem to play an active role in the nitrogen-fixing symbiosis occurring within the root nodules. Two major types of polysaccharides were found to be loosely attached to the surface of the outer bacterial membrane: the acidic EPSs released in the medium and the cyclic β-glucans. Cyclic β-glucans are thought to be found predominantly in the periplasmic space, although they are also secreted under certain circumstances. Despite this, cyclic β-glucans are usually treated as a separate category.
Since the microreview published in 1990 by Gray and Rolfe that focused only on the role of rhizobial exopolysaccharides in legume invasion processes , a detailed review of biosynthesis, structure and function has been written by Becker and Pühler .
EPS function seems to be related to the type of nodule ontogeny
The acidic EPSs are high molecular mass complex heteropolymers with repeating units ranging from seven to nine hexose residues. No general trend was found for their glycosidic linkage. They can be alpha, beta linear, or branched with side chains. They mostly contain noncarbohydrate substituents such as succinate, pyruvate or acetate and their acidic nature is due to the presence of uronic acids, pyruvate ketals and succinates.
According to Gray and Rolfe , the most interesting observation made in the 1990s concerned the differences of EPS perception in the two basic nodule ontogenies (determinate or indeterminate) which appeared to exhibit different rhizobial exopolysaccharide requirements. The acidic EPS appeared to be essential for the establishment of nitrogen-fixing symbiosis on legumes developing an indeterminate type of nodule: Sinorhizobium meliloti/alfalfa, Rhizobium leguminosarum bv. viciae/Vicia sativa and bv. trifoli/trifolium, ssp. NGR234/Leucaena[15–20]. This is not the case for associations leading to determinate nodules type: Sinorhizobium fredii/Glycine max, R. leguminosarum bv. phaseoli (renamed R. etli)/Phaseolus[18,21,22]. One explanation for such differences could be that unidentified LPSs could complement the EPS deficiency in the determinate nodule formation. Finally, the first indications concerning the importance of the succinyl groups for root invasion were published .
To date, numerous works have been published, demonstrating different rhizobial exopolysaccharide requirements between the two basic nodule ontogenies. One objection could be that these were made comparing very different couples of symbionts. Therefore, in order to confirm the acidic EPS involvement only for indeterminate type of symbioses, Hotter and Scott studied an isogenic system, R. loti PN184, able to nodulate Lotus pedunculatus (determinate type) and Leucaena leucocephala (indeterminate nodulating legume) . Five Tn5 EPS-negative or EPS-altered mutants (exhibiting conserved LPS) are fully effective on a determinate nodulating host but ineffective on the indeterminate one. Stacey et al. gave one possible explanation for this , by pointing out that determinate nodule infection threads are broader than those for the indeterminate type, and that EPS could be a critical matrix component of the threads. One argument supporting this theory could be that, even if isolated bacteroids of S. meliloti still exhibit exo gene expression, in situ measurements revealed that EPS repression occurs after the invasion step . Nevertheless, Leigh and Coplin (1992) attributed this difference to the fact that in indeterminate nodules, bacteria have to spread out by means of continuous infection thread penetration in the new cortex cells, whereas in the determinate type, they spread by division of already infected cells . Consequently, EPSs might be relevant for the cortex cell penetration step.
Recently, Karr et al. examined the wild type strain Bradyrhizobium japonicum 2143 and soybean lectin-binding-deficient mutants . The authors demonstrated the suspected dependence of the EPS monosaccharidic composition on the carbon source used for growth. For example, when grown on arabinose, gluconate or mannitol, B. japonicum 2143 exhibits the EPS composition described earlier (Man/Glu/Gal/GalA 1 : 2 : 1 : 1) but cultured on malate, EPS became extremely enriched in Gal (novel EPS). The composition of capsular polysaccharide also seems to be affected by growth conditions, but not LPS. Interestingly, these variations in EPS composition impaired soybean lectin-binding ability but did not affect the nodulation or nitrogen fixation phenotype of 2143. In contrast, 1252 remains ineffective in nodulation, independent of the EPS composition. One can notice that the behaviour of 2143 points out, analogously to Hotter's observations , the uncritical role of EPS for such determinate nodulation type.
EPS function requires a structural recognition
Determining the role of S. meliloti polysaccharides, known to be rich in succinate substituents  and therefore called succinoglycans, has been the goal of many studies [17,23,26,27,29–33] (Fig. 1).
Succinoglycan. This is one of the best-understood symbiotically important exopolysaccharides and is required for the invasion of alfalfa roots by S. meliloti Rm1021, for which structure  and biosynthesis  have been determined (reviewed in ). Succinoglycan is a polymer of repeating octasaccharidic subunits (seven Glu and one Gal) bearing succinyl, acetyl and pyruvyl substituents. It can be detected in two size classes (high and low molecular mass). A tetramer was reported to partially restore the invasion by exo– mutants . Yang observed that infection threads are initiated in the curled root hairs by the exo– mutants but abort within the peripheral cells of developing nodules . S. meliloti's succinoglycan enables the infection of alfalfa by the S. meliloti exo mutant . Finally, Cheng and Walker (1998) constructed stable GFP-labelled rhizobia cells to accurately observe the impaired invasion in alfalfa roots of ineffective exo mutants . The product of the exoY gene is a galactosyl-1-P transferase, required for the first step of the succinoglycan synthesis. By observing the exoY mutant, this study demonstrated that succinoglycan is not required for S. meliloti to colonize curled root hairs or to induce the infection thread, but is necessary for its elongation through the nodule cells. Studying the exoZ mutant, coding for the acetylation of the succinoglycans, they determined that acetyl addition on succinoglycans markedly enhances the efficiency of both initiation and elongation of the infection threads. Finally, an exoH mutant, which produces a high molecular mass succinoglycan lacking succinyl groups, revealed the modest effect of succinylation on infection thread initiation, but its drastic effect on the elongation efficiency (Fig. 2).
One question remained open: are the other polysaccharides produced (KPS or EPS II) responsible for the rare infection thread initiation and sometimes elongation observed in these mutants? Indeed, it was demonstrated that EPS II  and KPS (thought at this time to be LPS)  can substitute for EPS I, although both were found to be less efficient for the invasion of alfalfa roots . This work also revealed that EPS I, K antigen, and EPS II promote distinct morphologies for the threads, suggesting that the three polysaccharides might have related but not identical mechanisms to promote the invasion, as discussed below.
To clarify the EPS structural requirement for an effective nodulation, Gonzalez et al. investigated the low molecular mass EPS of S. meliloti Rm1021 . Normally, this strain requires the presence of a succinoglycan (which it synthesizes by itself) to efficiently infect Medicago sativa root hairs. Nevertheless, two mutants (expR101/exoY, mucR/exoY), which synthesize EPS II but not succinoglycan, are symbiotically proficient and deficient, respectively. The difference between the two resides in the size of the EPS II. ExpR101/exoY generates a panel of compounds ranging from one or two disaccharide subunits up to high molecular mass material, as mucR/exoY produces only high molecular mass EPS II. Addition of the low molecular mass EPS II from expR101/exoY (at an amount of about 7 pmol per plant) allowed efficient formation of nitrogen-fixing nodules by the previously noninfective strain. Furthermore, this same fraction restored the nitrogen-fixing phenotype to the exoA/expA mutant deficient in both succinoglycan and EPS II synthesis. This particular EPS II exhibits an average size of 32 sugars (6500–8000 Da) consistent with the size of succinoglycan (32 sugar units, molecular mass 7000 Da), indicating that the plant may recognize both through a common mechanism. To date, no evidence for this perception mechanism has been found.
Consequently, three low molecular mass polysaccharides produced by S. meliloti are crucial for symbiosis establishment with M. sativa, namely succinoglycan, EPS II and K antigen (the active portion of which may have a weight of 5000–8000 Da). To check the efficiency for each one, Pellock et al. (2000) studied the infection initiated by different mutants or purified compounds using the previously described GFP-labelled S. meliloti strains (see § succinoglycan ). Even if all three polysaccharides are capable of achieving the Inf+ and Thr+phenotype, succinoglycan mediates highly efficient initiation and elongation of the threads, while KPS is more deficient in this last step and EPS II less efficient than succinoglycan and KPS for both steps. Modulation of these biological responses might indicate that the three polysaccharides act through related but not identical mechanisms. Strains producing only KPS or EPS II generate large quantities of aberrant infection threads resulting in the formation of bacterial pockets along the threads. Therefore, the authors suggested that these three polysaccharides might act as signalling molecules, directing cytoskeletal movements in the root hair cell, which would stimulate and maintain the infection thread, or modulate the plant defence response.
To close this section, it appears important to point out that the redundancy seen with exopolysaccharides on S. meliloti could also exist in other rhizobia and could lead to the erroneous interpretation that exopolysaccharides are not required for colonization of host legumes with determinate nodule meristems.
Other acidic EPS. With the exception of S. meliloti, R. leguminosarum biovars are unique model strains for examining the impact of EPS on the establishing of symbiosis, even if several EPS structures have been reported on other strains (e.g. B. japonicum, Bradyrhizobium elkanii) (Fig. 3).
Contrary to the conclusion of the work published by Philip-Hollingworth et al. [41,42], it has been demonstrated by O'Neill et al. that localization of the ester groups in the acidic EPS secreted by some R. leguminosarum biovars do not determine the host specificity of nodulation .
Nevertheless, indications that host plants indeed have structural requirements for EPSs to be biologically active were obtained from studies of some other rhizobia [13,20,34]. Coinoculation of purified Sinorhizobium sp. NGR234 EPS with Sinorhizobium sp. NGR234 exo mutant restored nodule development to Leucaena. R. leguminosarum bv. trifolii purified EPS also restored the Nod+ phenotype of R. leguminosarum bv. trifolii exo mutant on clover . Genetic evidence has indicated that the succinoglycan from S. meliloti can substitute for Sinorhizobium sp. NGR234 EPS in Leucena infection, probably due to their structural similarity .
Van Workum et al. used phase contrast microscopy to investigate the early symbiotic behavior of R. leguminosarum bv. viciae exo– mutants carrying the pXLGD4 plasmid containing the constitutive hemA–lacZ fusion . Coinoculations of nonisogenic EPS+ but Nod– strains with the full EPS– mutant pssD111 demonstrated that R. leguminosarum bv. trifolii ANU843 and R. etli CE3 EPSs restore nodulation on vetch, whereas S. meliloti Rm1021, Rm9000 and Sinorhizobium sp. NGR234 do not. This points to structural elements being essential for EPS biological activity. Indeed, the first three strains produce octasaccharide repeating units (with variable degrees of substitution by acetyl or hydroxybutanoyl groups or other small side chains) unlike the three others. Moreover, nodulation restoration kinetic studies led the authors to support Niehaus's theory (that EPSs are a determinant for infection thread induction) as well as Rolfe's hypothesis (that EPSs are a requirement for tip growth and deformation of the root hair, a role which was classically attributed to Nod factors).
EPSs modulate the plant defence response
Niehaus et al. (1993) found indications that S. meliloti EPS I (high molecular mass species), or a related compound, might suppress the M. sativa plant defence response . EPS I-deficient mutants failed to develop normal nodules, although they were able to penetrate the intercellular spaces of the root cortex and to induce infection thread-like structures. Finally, with a delay of up to five weeks after inoculation, the EPS I-deficient mutants were able to overcome the plant defence response as they induced infected nodules. To test this hypothesis, low molecular mass EPS I isolated from S. meliloti SU47 was added to alfalfa cell cultures exhibiting a strong medium alkalinization defence reaction provoked by yeast-elicitors . EPS I effectively suppressed this process.
A. Skorupska et al. investigated Tn5 mutants of R. leguminosarum bv. trifolii 24.1 in which insertions are located on two megaplasmids pRtb and pRtc and on the chromosome . All these bacteria with nonmucoid phenotype exhibit conserved LPS. The exo– mutants elicited two different types of clover root nodules, i.e. infected and not. For the first, the mutated genes seemed to control later stages of the infection. In fact, the progression of rhizobia into the root was quite normal except at the end of infection thread elongation, as enlarged and harboured degenerated bacteroids were observed. Plant cells contained bacteroids but no maturation into a Fix+ organ was observed. The second class of exo– mutants is defective in the early stages (no infection threads, no bacteroids). For the first class of mutants, the authors suggest that the absence of the EPS from R. leguminosarum bv. trifolii might induce in clover a plant defence response after the release of the first bacteria from the infection threads.
Rolfe et al. made similar conclusions after examining whether EPS quantity or quality is most important for the symbiosis . For this study, two exo mutants of R. leguminosarum bv. trifolii were investigated, one (pss1 mutant ANU437) produced only 0.3% of the parental strain EPS and deacetylated-acidic-oligosaccharides, and the other (exo– mutant ANU54) was disabled in EPS synthesis. This second strain did not form a capsule surrounding the cell, unlike the first one, and its growth was rapidly inhibited by phenolic phytoalexins. Both mutants induced slow root hair curling (Hac+), formed growing infection threads (Thr+) only on white clover and initiated small non-nitrogen-fixing nodules on both white and subterranean clovers. Under normal plant growth conditions, ANU437 bacteria were released into cortical plant cells but were Fix–. Mutant ANU54 caused thickening of the outermost layer of plant cells on the empty nodules as well as the deposition of polyphenolic material, indicating advanced plant defense processes, even if such compounds have also been reported in aborted wild-type bacteria infection.
Van Workum et al. reported that R. leguminosarum bv. viciae exo– mutants failed to nodulate vetch normally and proposed that ethylene produced during the classical infection was responsible for root cell growth . By decreasing the ethylene production in V. sativa sp. nigra infected with the exo– mutants, the plant was able to generate nodules. The delay in nodulation ability of the exo– mutants could allow for ethylene accumulation, leading to inhibition of infection and nodulation. It is important to note that ethylene is a ubiquitous plant hormone known to play multiple roles in nodule development.
Other indications that EPS might prevent a plant defence response were given by the work of Parniske et al. [50,51]. Using B. japonicum 110spc4 exoB mutants ΔP5 and ΔP22 (producing structurally modified galactose-free EPS but conserved LPS), they demonstrated that in the early stages of interaction with Glycine max these mutants induced the accumulation of phytoalexins, low molecular mass antimicrobial compounds normally induced when the host plant is inoculated with pathogenic organisms such as Phytophthora megasperma. In soybean, the major phytoalexin is an isoflavonoid called glyceollin. After 72 h of incubation with exoB mutant, levels of glyceollin about 10 times higher were measured in the exudates of these roots compared to the roots inoculated with the wild type strain, apparently supporting the plant defence inhibition theory.
It remains unknown whether EPS acts as a signal for the plant defence receptors or as a mask of antigenic epitopes.
Cyclic glucans might be more important than expected
Initially, cyclic β-(1,2)-glucans were mostly studied in the extracellular medium of rhizobial cultures and their high level as cell-associated saccharides ignored. These are predominantly localized in the periplasmic compartment [19,53]. First isolated from Sinorhizobium meliloti, cyclic β-glucans consist of a neutral homopolymer of about 20 β-(1,2)-linked glucose residues – often substituted by phosphoglycerol, phosphocholin or succinyls – and probably play a passive role in the bacterial cell's adaptation to hypo-osmotic conditions in its surroundings .
Recently, a few publications have appeared on cyclic glucan structures and their symbiotic role, and extensive reviews have been written by Breedveld and Miller [53,54]. Concerning their structures, β-(1,2) is not the only possible linkage for cyclic glucans, as the glucan from Bradyrhizobium japonicum exhibits β-(1,3)- and β-(1,6)-linkages [55–57]. Consistent with a role for glucans in root attachment and hypo-osmotic adaptation, Gore and Miller reported a high level within bacteroids . This could be consistent with a need for rapid adaptation to osmotic variations in environments ranging from rhizosphere to infection thread and symbiosome. Nevertheless, their exclusively passive symbiotic role was questioned, as S. fredii and S. meliloti ndvA and ndvB mutants, affected in their cyclic β-glucan biosynthesis, were found to be ineffective. These formed only bacteroid-free pseudo-nodules and early aborting infection threads [56,59]. As complementation of ndvB mutants by the addition of exogenous cyclic β-(1,2) glucans was unsuccessful, it was concluded that their periplasmic location might be important . Interestingly, ndvB-like mutants of B. japonicum formed ineffective but bacteroid-containing nodules on soybean . Even if the ndvB mutant of S. fredii is not impaired in symbiosis with McCall and Peking soybean cultivars, indicating that β-(1,2) glucans are not required for nodule invasion , cyclic β-(1,3)-(1,6) glucans are important for symbiotic development involving B. japonicum and soybean, and might act as suppressors of a host defence response [61,62]. Miller et al. showed that cyclic glucans from B. japonicum USDA110 also elicit glyceollin and daizein production in soybean . In parallel, they showed that every nodule infected with these mutants exhibited an enhanced chitinase activity, whereas only 17% of the wild-type infected nodules do this. Even if chitinases seem to have important nondefensive functions in certain developing organs , these observations might indicate that in the later stages, defence processes still occur. Due to their structural analogy with fungal cell fragments of soybean pathogens, their possible elicitor activity for phytoalexin production was studied . They demonstrated that bradyrhizobial cyclic glucans compete with fungal fragments for binding activity with a putative elicitor. Recently, Mithöffer and coworkers showed that even if free-living wild-type B. japonicum cells showed low sensitivity to oxidative burst or glyceollins, mutant derivatives affected in their cyclic β-glucans exhibit an enhanced susceptibility .
Due to their structural analogy with cyclodextrins, their ability to form inclusion complexes with hydrophobic guest molecules was investigated. Incidentally, it has been shown that cyclic β-1,2 glucans increased the solubility of a legume flavonoid (inducer of nod genes). Such inclusion complexes could explain why the nodulation is more efficient when β-1,2 glucans are added to the symbiotic nodulation system  and could again attest the passive (not signal) role for these compounds.
Before closing this first section we would like to underline the importance of the work of Breedveld et al. (1993), which points out a practical problem often ignored . They generated Tn5 mutants of R. leguminosarum RBL5515 and studied more precisely one of them altered for EPS synthesis, the repeating unit of which lacks the terminal galactose. This mutant and three others were impaired in the infection stage. Nevertheless, they concluded that it was not possible to correlate the defect of a particular polysaccharide with the nodulation processes, because these alterations were accompanied by significant increases in the synthesis and secretion of cyclic β-1,2-glucans. Therefore, even if some of the authors mentioned above attested that the structures of polysaccharides not targeted by the mutation are conserved, they have not studied the conservation of their expression level. This could be important if such polysaccharides play a role in masking the rhizobia-faced surface from interactions with the plant defense mechanisms.
More about KPS involvement
Capsular polysaccharide surrounds the bacterium and constitutes a hydrated matrix which confers bacterial resistance to bacteriophages and to the dry conditions often encountered in the rhizosphere environment.
The first rhizobial capsular polysaccharides structures were described by Reuhs et al. on Sinorhizobium fredii USDA205 and Sinorhizobium meliloti (Table 1).
Table 1. K antigens of Sinorhizobium spp. Neu, 5-amino-3,5-dideoxynonulosonic acid (neuraminic acid); Pse, 5,7-diamino-3,5,7,9-tetradeoxynonulosonic acid (pseudaminic acid); Ac, acetyl group; α-OH-But, α-OH-butyryl group; Me, methyl group; Hex, any hexose; Kdx, any 1-carboxy-2-keto-3-deoxy sugar.
These polysaccharides contain a high proportion of 3-deoxy-d-manno-2-octulosonic acid (Kdo) and are structurally analogous to one subgroup of K antigens found in Escherichia coli. This explains why the term of K-antigen polysaccharide (KPS) is used to refer to the rhizobial capsular polysaccharides, even if it cannot be excluded that other types of capsular polysaccharides (non-K-antigenic) could be found in the future. Basically, all known KPSs have been described in Sinorhizobium species [33,67,68]. They differ by size range and contain a repetitive motif of a hexose linked with a Kdo or related 1-carboxy-2-keto-3-deoxy sugars such as pseudaminic and neuraminic acid (Table 1). Distinct polysaccharides can be produced by one strain, S. fredii USDA205, which produces two KPSs, one with galactose and one with 2-O-MeMan . Important facts, resulting from this work, from Reuhs et al. 1993 and Reuhs et al. 1995 [33,69], should be noted here: (a) the K-antigens are strain-specific antigens, whereas the EPS are conserved within a species; (b) the production of one or more KPS could be dependent on the growth conditions of cultured cells. Apigenin, a flavonoid inducer of the S. fredii nod genes, or host (soybean) root extract, causes the up-regulation of the secondary, 2-O-MeMan-containing-K-antigen produced by S. fredii USDA205; and (c) as they are only located directly around the bacterial membrane and not secreted, their symbiotic involvement occurs after the first contact between the microbial symbiont and the host. It is even possible that KPS mediates such contact.
Possible roles for KPS remain uncertain
The role of the K-antigen polysaccharide has only been studied extensively in the S. meliloti/alfalfa symbiosis. Apparently, it acts in two ways: one active (signal) and another passive (determinant-masking). First, K-antigen may be involved in the protection of the bacteria against natural legume defence products or microorganisms such as phages. Capsule mutants of S. meliloti Rm41 exoB (AK631) were found to be more resistant to one and less resistant to nine of the 12 phages investigated . This passive role could assign capsular polysaccharide as a host range determinant, the correct symbiont being resistant to the host plant defences, while the other rhizobia are not.
One of the difficulties in understanding the symbiotic role of KPS lies in the interconnected biosynthesis of LPS, EPS and KPS. Mutation in the exoB gene not only abolishes the EPS I and EPS II production, but also results in an altered LPS pattern . Moreover, it was shown recently by Kerest et al. that the biochemical pathways for KPS and LPS production also share common elements .
Interestingly, using EPS or KPS production-deficient S. meliloti mutants associated with direct infiltration of purified KPS into alfalfa roots, Becquart-de-kozak et al. demonstrated that S. meliloti K-antigen polysaccharides were able to induce transcript accumulation of alfalfa genes encoding enzymes of the isoflavonoid biosynthetic pathways . No quantification of these plant cell products was made, but it was concluded that bacterial KPS might have a significant role in the early recognition of S. meliloti by alfalfa root cells.
It is probable that KPS could play a role in the symbiosis between the indeterminate nodule-forming plant alfalfa and bacterium AK631. This is the only S. meliloti 41 mutant strain known to be deficient in EPS production but still able to invade alfalfa root nodules . An additional mutation of rkpZ results in Fix– nodules. This gene is known to affect the KPS molecular mass, making it much lower in size range. Rm1021, based on hydridation tests, is supposed to be deficient in rkpZ (rkpz0) . Conversely, complete genome sequence analysis revealed two rkpZ homologues. The activity of these genes remains under question. Wild type strain Rm1021 produces a symbiotically inactive KPS, as it does not substitute for missing EPS. It is not only the higher size range that causes the difference observed with AK631. Introduction of the rkpZ gene into Rm1021 resulted in only partial restoration of the Fix+ phenotype  of the Rm1021 exoB mutant. However, introduction of the complete pSymB megaplasmid of Rm41 results in an almost complete complementation, indicating that other pSymB genes might be involved. Indeed it was recently published by Kiss et al. that the rkp-3 gene region of AK631 located on pSymB coding for genes involved in the biosynthesis of KPS, which are strain specific, is absent in Rm1021 . These results suggest that K-antigen can substitute for EPS but that at least a change in its size from approximately 5000–7000 (for the active form in AK631) to around 20 000–25 000 (for the inactive one in AK631 rkpZ) makes it ineffective. In fact, the active role of K-antigen polysaccharide seems to affect the promotion of infection thread initiation and development [10,33,38] (see the succinoglycan section). Futhermore, the work of Pellock et al. using GFP-labelled S. meliloti AK631 mutants that lacked EPS and succinoglycan, – revealed that K antigen is able to initiate the infection thread even if three of four infection threads abort in the extension step . Based on all these investigations, KPS seems to play both an active and a more passive role regarding the plant defence reaction.
Understanding the role of LPS
LPSs are polysaccharides that are attached to the membrane by a lipidic part inserted into the bacterial phospholipid monolayer, the saccharidic part being oriented to the exterior. The general structure of this compound consists in an anchor named lipid A associated with a core polysaccharide, which can bears an O-antigen domain. Rough LPS contain only the two first domains, while all three domains constitute the smooth LPS. LPS are acidic polysaccharides found in the aqueous phase when extracted by the hot water/phenol method. However, they can be partitioned between the two phases, as for example the LPS from B. japonicum 61A123 .
Structures and biosynthesis of LPS were extensively reviewed by Noel and Duelli, Ruehs et al. and Price [9,69,77] (Fig. 4). Nevertheless, some recent and complementary works can be found hereafter.
The anchor. By its numerous fatty acids, lipid A allows the LPS molecule to attach to the outer phospholipid leaflet of the external microbial membrane. This anchor can vary among the different rhizobial species. Reuhs et al. reviewed the different known structures . In general, the lipid A moiety is a 2,3-diamino glucose or a glucosamine disaccharide unit, O-6 linked  to the core portion through a Kdo. Mainly four to six α-hydroxy fatty acids are carried by the glucosamino-disaccharide. Sometimes the reducing sugar is oxidized into amino gluconate, or is C1-phosphorylated. The second glucosamine can also be phosphorylated or substituted by a galacturonate moiety. The systematic presence of 27-OH:C28 fatty acids (or 27 hydroxyoctacosanoic acid) in all analysed preparations from rhizobia (except for Azorhizobium caulinodans) made it a chemotaxonomic marker for them [79,80]. The genus Mesorhizobium was analysed in 11 strains for its fatty acid pattern [81–83] and presents the particularity of containing a variable amount of unusual oxo-fatty acids.
Recently, Que and coworkers demonstrated in two joined reports that lipid A of R. etli LPS can in fact present far more structural variations than previously reported [84,85]. Using separation methods with higher resolution, they isolated and identified six distinct, but apparently biosynthetically related, lipid A components from the same CE3 strain. Variations in structural features were based on a 4′-branched galacturonic acid glucosamine disaccharide, the reducing end of which can be in an oxidized form or not. Four or five 3-OH:C14 and one 27-hydroxy butyryl:C28 fatty acids are present (Fig. 5).
The core moiety. The core region of rhizobial LPS is attached on one side to the lipid A and sometimes on the other side to the O-antigen by mild acid-labile Kdo sugars. Their structures were determined for the LPSs isolated from R. leguminosarum bv. phaseoli[86–88], R. trifolii ANU843 , R. etli, B. japonicum 61A101c  and B. elkanii, after mild acid hydrolysis and gel filtration or HPAEC purification. More recently, structures for S. fredii, S. meliloti and Rhizobium sp. NGR234 strains were established from sugar composition and HPAEC profiles of the core moiety . The critical point is to make sure that the core portion has been isolated and not an O-antigen (or other) repeating unit. The lack of O-antigens solves the problem in the study of the rough LPS. A simple purification of the entire LPS, and centrifugation after mild acid hydrolysis allows the elimination of the lipid A moiety and the isolation of the core region. A widely used alternative is to develop a mutant lacking the O-antigen portion.
All these studies, reviewed by Reuhs et al. and based on HPAEC profiles and the reaction with monoclonal antibodies, reflect that a common structural feature exists, at the core level, within the Sinorhizobium and the R. leguminosarum species although small variations are not excluded, even within one species .
O-antigen. The structures of the O-chain repeating units of R. leguminosarum bv. trifolii, Rhizobium tropicii CIAT 899 , R. etli CE3  and Mesorhizobium huakuii IFO15243T  are now completely known. Many others have been studied for their glycosyl composition, often to compare the wild-type structure with those synthesized by mutants showing an interesting phenotype. They can contain uronic acids, heptoses or Kdo (for R. leguminosarum) but are generally enriched with large proportions of various deoxy- and/or methyldeoxy-sugar residues. These researches have indicated interesting hydrophobic properties for the smooth LPS. The common features of the O-antigen domains are that they are very variable in structure even in the same species, from strain to strain, except for Sinorhizobium strains, which have structurally conserved lipopolysaccharides [71,94]. A strong antigenic activity is also characteristic because the LPSs carrying the O-chain are the dominant components to react with antisera.
LPS involvement in the establishment of symbiosis
Dazzo and coworkers demonstrated that R. trifolii LPS plays an important role by modulating infection thread development in white clover root hairs . Although LPS is a constitutive component of the bacterial membrane, it could be found in very low concentrations in growth media. Consequently, a putative role from a distance or in the early steps of symbiosis could be attributed to rhizobial LPS. From the host–symbiont pair involving white clover and R. leguminosarum bv. trifolii 0403, a bacterial LPS was purified which enabled S. meliloti to be effective (not normally a microbial symbiont species for white clover). The purified LPS alone was able, in a concentration-dependent way, to bind rapidly to root hair tips, infiltrate plant cell walls and stimulate the infection thread formation. One caveat should be made, however. Even if not yet described in R. trifolii, the soluble LPS could have been KPS, in which case the observations may have been misinterpreted as happened originally for the S. meliloti KPS.
Perotto et al. studied LPS mutants of R. leguminosarum strain 3841 presenting a lack, or structural modifications, of O-antigen  to experience how the host defence reaction could be modulated depending on the nature of the mutation. A putative role of LPS could be the suppression of defence reactions, but interestingly it was observed that these mutants were also impaired in bacteroid differentiation and presented so many abnormalities at this stage that despite their previous resistance to the defence reaction they developed into nonfixing bacteroids. Note that Jabbouri obtained Sinorhizobium sp. NGR 234 mutants lacking the smooth LPS and presenting a similar phenotype .
Albus et al. showed strong evidence for LPS involvement in inhibition of the host defense reaction. M. sativa cell cultures were incubated with a yeast elicitor  and a typical defense reaction was observed. An oxidative burst (elevation of the hydrogen peroxide concentration) as well as an alkalization of the medium were measured. Addition of the purified LPS produced by S. meliloti (the nodulating strain of M. sativa) inhibited the described reactions in a concentration-dependant way (Fig. 6).
Kannenberg and Carlson identified another possible LPS role for the establishment of the nitrogen-fixing symbiosis . LPS produced by R. leguminosarum 3841 nodular bacteroids were compared with those produced by free-living bacterial cultures under low pressure of oxygen or not. The bacteroidal LPS and those produced under low pressure of oxygen have strong similarities in their physico-chemical behaviours and in their structures . Monitored by DOC/PAGE analysis, larger amounts of phenol-soluble LPS, that are smooth-type LPS (and almost no rough LPS) are produced. Their saccharidic domains present increased proportions of glycosyl-acetylation and methylation. Their portions of O-antigen were shown to be rich in 3,4-di-O-MeRha, 3-O-MeRha or fucose, enhancing (together with the presence of higher proportions of 27-OH-C28:0) the hydrophobic character of the LPS. This suggests that with less oxygen available the bacteria produce predominantly smooth forms of more hydrophobic LPS. Such conditions are encountered during bacteroid development, as the central infected tissues of the nodule constitute a low-oxygen environment for bacteroidal nitrogenase activity. To pursue this reasoning, it would be logical that to face these changes of environment during host plant penetration, bacteria synthesize higher proportions of more hydrophobic, smooth LPS as the presence of their O-antigen increases their hydrophobicity, and hence that of the cell surface. By its particular physico-chemical properties, this kind of LPS can allow the bacteria to adopt an optimal degree of interaction for a close contact with the host plant cell surfaces, due to the fairly high hydrophobicity of plant cell walls and membranes. It is possible to correlate this method of rhizobial adaptation for its survival (or a good bacteroidal activity) with the first results obtained by Jabbouri et al. . It has been shown that mutants lacking smooth LPS (provoking a loss of hydrophobicity of the bacterial cell membrane) infect Vigna nodules in the usual way but do not produce efficient bacteroids (Fix– phenotype). Without ignoring that LPS could play an active role in the later steps of symbiotic interactions, for example by inhibiting the host defence reaction, it is possible that this LPS (whose structure has been shown to be rhamnose-rich ) can simply allow the bacteria to adapt to the particular endosymbiotic conditions in the same way as R. leguminosarum 3841.
Impact of lipid A. The important role played by lipid A in symbiosis involving rhizobia and plants remains to be determined. To address the role of lipid A during symbiosis, it is necessary to work with mutations in the lipid A biosynthetic pathway. The construction of the relevant mutants relies on an in-depth understanding of the molecular genetics of lipid A biosynthesis, which in turn requires the structures of lipid A to be known.
The presence of an unusual C28 fatty acyl moiety in the structure of R. etli lipid A, and the fact that this anchor plays an antigenic role in the mammalian immune response, led Que and coworkers to hypothesize that lipid A plays an important role in eliciting the host plant defence [84,85].
It is interesting to note that such a length (C28) could insert nicely into a phospholipid bilayer, perhaps facilitating the phagocytosis of the bacteroid by the plant.
Impact of the core. The lpsB gene described by Lagares et al. [100,101] is involved in the LPS core biosynthesis. The lpsB mutant of Rm2011 was found to be less competitive on M. sativa (but remains Fix+) and displayed a Fix– phenotype on Medicago truncatula. Consequently, different host plants require different LPS structures for supporting wild-type symbiosis. Moreover, it was found very recently by Campbell et al., that the same mutation in Rm1021 resulted in complex Fix– phenotype on M. sativa.
Impact of the O-antigen. A hydrophobic character is given to the O-antigen by its systematic important proportions of deoxy- and O-methyl-deoxy-sugars [76,91,92,103–106]. In this field, the work of Gao et al. is interesting because they built a specific mutant of A. caulinodans (ORS571-oac2) with deficient rhamnose biosynthesis, resulting in arrested nodule development, blocked nodule penetration by ORS571-oac2 at the internalization stage and host defence symptoms . The immediate result is that the LPS PAGE-banding pattern showed a large increase in the electrophoretic mobility, interpreted as a lack of rhamnose in their O-antigenic part, resulting in a truncated O-antigen. This was paralleled by the fact that LPSs were water-soluble in the mutant while they were phenol-soluble in the wild type. In fact, despite its much larger sugar domain compared to rough LPS lacking the O-antigen domain, smooth LPS are more hydrophobic. This demonstrates that the core region is the most hydrophilic portion of the LPS. A recent study suggests that the LPS core acts as one of the main sites for binding calcium to the outer membrane, allowing it on the one hand to participate in the structural coherence of the bacterial membrane  and on the other to be a water accessible domain.
Ten years ago, Brink et al. investigated the role of O-antigen in R. leguminosarum ANU843, by studying mutants deficient in the O-antigen synthesis and in the production of some core region sugar residues . These elicited incompletely developed clover nodules with low bacterial populations and low nitrogenase activity. Complementation of the mutated regions showed restoration of the wild phenotype. They concluded that O-antigens are required for successful nodulation of clover and that subtle variation in their monosaccharidic composition could be tolerated. Similarly, mutants of R. etli lacking the LPS O-antigen domain are defective in infection [9,110].
More precisely, a recent paper  shows that R. etli mutant CE166, exhibiting a 60% quantitative deficit of its O-antigen bearing LPS and a total lack of the normal O-antigenic quinovosamine, forms nonfixing pseudonodules on Phaseolus vulgaris. This means that the O-antigen region is required in the latter phase of symbiotic steps. Indeed, these mutants progress normally into the infection thread until they are stopped ‘en route’ or at most after penetrating one or two cell layers. Surprisingly, a complemented mutant restored in its O-antigenic abundance, but not in the presence of quinovosamine, elicits fewer nodules more slowly, even though they are bigger than the normal ones and have a near-normal nitrogen-fixing activity. In fact the nitrogen-fixing activity per plant is higher with this complemented mutant than with the wild-type CE3. We notice that strain CE374  exhibits the same phenotype. Interestingly, in support of Noel's hypothesis , by comparison with wild-type CE3, it appears that the presence of the O-antigen is required to finalize the nodule infection, and subtle structural features confer a greater efficiency to the kinetics of infection for the wild-type bacteria, or in final nitrogenase activity for the complemented mutant. So, it would be interesting to investigate the possible presence of a plant–LPS binding protein, to clarify the plant behaviour in the later stages of the nitrogen-fixing interaction.
In contrast, other studies were done on S. meliloti and Rhizobium galegae indicating that O-antigen is not required for symbiosis. For S. meliloti, Clover and coworkers made a panel of mutants deficient in smooth LPS or strongly modified in their electrophoretic properties. These alterations appeared to have little effect on symbiosis but remain observable. It was not clear whether the small quantities of modified LPS remaining were at the origin of the activity, as one epitope on the molecule remained unmodified.
LPS structural modifications occur during the establishment of symbiosis
Generally, the rhizobial LPS structure is only partially known, particularly concerning the O-antigen section, which seems to play the most important role of the three LPS parts.
There is evidence that changes in the bacterial environment provoke structural alterations on the surface polysaccharides even within the roots [115–117]. But it is important to note that introduction into the growth media of nod gene inducers also has an important effect. Reuhs and coworkers observed on S. fredii USDA205 that the LPS O-antigen is mainly changed in carbohydrate composition and in mass range . Moreover, important genes involved in the biosynthetic pathway could be located on nod regions of rhizobial symbiotic plasmid , that would enable the bacteria to adapt their external structure before their first infective interaction. The LPSs are not the only surface polysaccharides affected by the induction. Effectively, KPS seems to present a much higher proportion of 2-O-methyl mannose-Kdo motif on S. fredii USDA205 .
Curiously, Noel et al. made opposite observations. Even if growth of R. etli CE3 in the presence of exudates from Phaseolus vulgaris resulted in a modified LPS that no longer reacted with monoclonal antibody JIM28, they concluded that these exudate-induced LPS modifications did not require the Sym plasmid .
Nodular polysaccharide (NPS)
At the beginning of 1992, Streeter et al. published a major paper concerning symbiotically induced EPS . They demonstrated that certain B. japonicum strains produce a polysaccharide (NPS) in soybean root nodules, that probably accumulates within the peribacteroid space inside the plant. Sugar composition of the NPS differs from the EPS composition obtained in same strain lab cultures.
In the early 1980s, based on several characteristics including EPS composition, B. japonicum was divided into two classes. The first taxonomic group remains B. japonicum and the other was renamed B. elkanii. This distinction is important for our current purposes, because the two groups differ in NPS production. In cultures B. japonicum excretes EPS composed of glucose, mannose, galactose, 4-O-methylgalactose and galacturonic acid, and in nodules NPS composed of rhamnose, galactose and 2-O-methylglucuronic acid (the last sugar has not been previously reported in bacterial polysaccharides). In contrast, both in cultures and in nodules, B. elkanii synthesizes EPS (NPS), which is only composed of rhamnose and 4-O-methylglucuronic acid. It is interesting to note the very different EPS composition of the two groups, but the similarity of their NPSs. To assess the bacterial origin of these NPSs, the influence of bacterial and host plant genotypes on NPS formation and composition has been studied. Three observations were made. First, the sugar combination is unknown for plant polysaccharides but quite usual for bacterial ones. Second, the NPS composition of B. japonicum is independent of plant genotype (using a comparison of four soybean cultivars). Third, the NPS is localised inside the symbiosome membrane, making this polysaccharide a probable bacterial one. An et al. published the precise structure of these NPSs  (Fig. 7).
Guentas et al. published a very intriguing paper . By reculturing a nonidentified Rhizobium strain (called Rhizobium sp. B) isolated from atypical nodules on alfalfa (often infected by S. meliloti), they identified an EPS that presents similarities with the NPS described above. In fact this EPS exhibits 2-deoxy-β-d-GlcpA and α-l-Rhap, which could be compared to the 2-MeO-β-d-GlcpA and β-l-Rhap of B. elkanii EPS or to the 4-MeO-β-d-GlcpA and α-l-Rhap of B. japonicum NPS. As it was not described whether the bacteroids were present or not during the culture stage, we cannot exclude the possibility that this EPS is produced under induction, similar to NPS.
EPSs seem to be critical constituents only for the invasion leading to indeterminate, and not determinate, nodule types. Even if this observation was recurrent, no explanation was given. Several putative roles have been considered for EPS. Apparently, as well as cyclic glucans, they seem to have a specific signal role in the root invasion process inhibiting the plant defense response [20,47–50]. A recent trend demonstrates a structural requirement for such role. In certain cases, EPS of two different strains or different EPSs produced by one strain can substitute for each other during invasion if they exhibit similar structures [39,40,122]. Some caveats for this could be firstly that the expression level, and probably the precise composition, of the EPS are culture condition-dependent [28,123], questioning the validity of lab experiments compared to natural conditions. Secondly, that even if the polysaccharide synthesis is coded in the chromosome, some modifications could be introduced by symbiotic genes occurring only under the direct or indirect induction of the promoters by plant flavonoids [10,99]. However, to our knowledge, comparative studies carried out on rhizobial EPS composition, cultured with or without nod-box inductors, have been poorly documented . This is not anecdotal, as we can consider that the newly characterised nodular exopolysaccharides NPS  are induction-dependent de novo synthesized, or modified bacterial EPS.
A new approach could perhaps be, after the structural investigation, the study of the physico-chemical properties of such compounds associated with their effect on membrane stability exposed to different environments. When produced, NPSs exhibit a particularly hydrophobic character, and they look suspiciously like LPS O-antigens. Similarly to Kannenberg [98,116] we recently demonstrated that lipopolysaccharides are modified when cultured under flavonoid induction (environmentally or within cultures), as the normally produced rough LPS became a quite hydrophobic deoxysugar-rich smooth LPS . However, the specific active LPS roles appear in the later stages of the establishment of symbiosis, and exhibit two apparently distinct functions. Firstly they facilitate the infection thread penetration into the cortical cells and obtaining the Fix+ phenotype. But why do the bacteria need to increase the hydrophobicity at their surface? One possibility could be that they have to mimic a plant-like interface necessary for setting up the symbiosome where, by an endocytotic process, a plant membrane surrounds the bacteroid assuming a material exchange. The high hydrophobicity could be important in adhesion behaviours, allowing interaction and material exchange between the bacterial and the peribacteroid membranes. Such intimate interaction might be necessary for maintaining the bacterial multiplication within the symbiosome in indeterminate nodules . These processes are poorly documented for the rhizobia–legume symbiosis. Nevertheless, recently a very interesting parallel was made by Lerouge and Vanderleyden between the animal–microbe (well studied) and the plant–microbe interactions, supporting such a hypothesis . A second possibility is the involvement of the LPSs in the bacteroidal differentiation processes, where the bacteria must undergo dramatic morphological changes (spectacular enlargement and even distortion into Y shapes) . Such smooth hydrophobic polysaccharides could confer a certain plasticity to the bacterial membrane.