How rhizobial symbionts invade plants: the Sinorhizobium–Medicago model

Abstract

Nitrogen-fixing rhizobial bacteria and leguminous plants have evolved complex signal exchange mechanisms that allow a specific bacterial species to induce its host plant to form invasion structures through which the bacteria can enter the plant root. Once the bacteria have been endocytosed within a host-membrane-bound compartment by root cells, the bacteria differentiate into a new form that can convert atmospheric nitrogen into ammonia. Bacterial differentiation and nitrogen fixation are dependent on the microaerobic environment and other support factors provided by the plant. In return, the plant receives nitrogen from the bacteria, which allows it to grow in the absence of an external nitrogen source. Here, we review recent discoveries about the mutual recognition process that allows the model rhizobial symbiont Sinorhizobium meliloti to invade and differentiate inside its host plant alfalfa (Medicago sativa) and the model host plant barrel medic (Medicago truncatula).

The recent completion of the Sinorhizobium meliloti genome sequence, and the progress towards the completion of the Medicago truncatula genome sequence, have led to a surge in the molecular characterization of the determinants that are involved in the development of the symbiosis between rhizobial bacteria and leguminous plants. Aromatic compounds from legumes called flavonoids first signal the rhizobial bacteria to produce lipochitooligosaccharide compounds called Nod factors¹. Nod factors that are secreted by the bacteria activate multiple responses in the host plant that prepare the plant to receive the invading bacteria. Nod factors and symbiotic exopolysaccharides induce the plant to form infection threads, which are thin tubules filled with bacteria that penetrate into the plant cortical tissue and deliver the bacteria to their target cells. Plant cells in the inner cortex internalize the invading bacteria in host-membrane-bound compartments that mature into structures known as symbiosomes. The internalized bacteria then develop into bacteroids, a differentiated form that is capable of nitrogen fixation. During invasion and symbiosis, rhizobial bacteria can evade the host plant innate immune response. In this Review, we summarize and integrate the advances that have been made towards understanding the invasion of plant tissues by rhizobia, and the differentiation of the specialized bacterial and plant structures that facilitate nutrient exchange.

Invasion of plant roots

Although plant roots are exposed to various micro-organisms in the soil, their cell walls form a strong protective barrier against most harmful species. The early steps in the invasion of barrel medic (M. truncatula) and alfalfa (Medicago sativa) roots by S. meliloti are characterized by the reciprocal exchange of signals that allow the bacteria to use the plant root hair cells as a means of entry.

Initial signal exchange

Flavonoid compounds (2-phenyl-1,4-benzopyrone derivatives) produced by leguminous plants are the first signals to be exchanged by host–rhizobial symbiont pairs¹ (FIG. 1). Flavonoids bind bacterial NodD proteins, which are members of the LysR family of transcriptional regulators, and activate these proteins to induce the transcription of rhizobial genes^1,2. For example, the M. sativa-derived flavonoid luteolin stimulates binding of an active form of NodD1 to an S. meliloti ‘nod-box’ promoter, which activates transcription of the downstream nod genes³. S. meliloti has two other NodD proteins — NodD2, which is activated by as-yet-unpurified plant compounds, and NodD3, which does not require plant-derived compounds to activate gene expression from nod box promoters¹. The expression of NodD3 itself is controlled by a complex regulatory circuit². Any of these NodD proteins can provide S. meliloti with the ability to nodulate M. sativa¹. Flavonoids from non-host plants can inhibit the transcription of S. meliloti nod genes³.

Figure 1. The initial signalling dialogue between Sinorhizobium meliloti and Medicago truncatula.

a | The induction of rhizobial nod genes requires plant flavonoids¹,⁴. The nod gene products produce Nod factor (NF), which is initially perceived by the M. truncatula MtNFP receptor¹,⁴,¹⁷. b | Root hair curling and cortical cell divisions require many M. truncatula gene products⁴,¹³: MtNFP¹⁷; MtDMI1 (REF. ²⁰); MtDMI2 (REF. ²⁵); MtDMI3 (REFS ²²,²³); MtNSP1 (REF. ²⁷); MtNSP2 (REF. ²⁶); MtCRE1 (REFS ⁵⁸–⁶⁰); and MtNIN⁵⁸,⁶¹. MtLYK3/HCL is required for colonized curled root hair (CCRH) formation, but not for the induction of cortical cell divisions¹⁹ (P. Smit and T. Bisseling, unpublished data). The required rhizobial genes are boxed in brown and the required plant genes are boxed in light green.

Among the rhizobial genes induced by flavonoid-activated NodD proteins are several nod genes encoding enzymes that are required for the production of lipochitooligosaccharide Nod factors⁴. Bacterially produced Nod factors induce multiple responses required for nodulation of appropriate host plants, and are the best characterized of the signals that are exchanged between plant hosts and rhizobial symbionts¹. Nod factors consist of a backbone of β-1,4-linked N-acetyl-D-glucosamine residues, which can differ in number not only between bacterial species but also within the repertoire of a single species¹. Nod factors are N-acylated at the non-reducing terminal residue with acyl chains that can also vary between rhizobial species¹. The genes of the nodABC operon encode the proteins that are required to make the core Nod factor structure¹. The products of other nod genes (and noe and nol genes) make modifications to Nod factors that impart host specificity, including the addition of fucosyl, sulphuryl, acetyl, methyl, carbamoyl and ara-binosyl residues, as well as introducing differences to the acyl chain^1,4. Many rhizobial species produce more than one type of Nod factor, but it is not yet possible to predict the range of possible host plants from the Nod factor structure¹. See REFS ^1,4–7 for in-depth reviews of Nod factor production by bacteria, and perception and signalling by plants.

Nod factors initiate multiple responses that are essential for bacterial invasion of the plant host⁴. One of the earliest plant responses to the correct Nod factor structure is an increase in the intracellular levels of calcium in root hairs, followed by strong calcium oscillations (spiking)⁴ and alterations to the root hair cytoskeleton^8–11. These responses are followed by curling of the root hairs, which traps rhizobial bacteria within what is known as a tight colonized curled root hair (CCRH)^12,13 (FIG. 1). Simultaneously, Nod factor stimulates root cortex cells to reinitiate mitosis, a process that is dependent on the inhibition of auxin transport by rhizobia and plant flavonoids^8,14–16. These cells will form the nodule primordium, and give rise to the cells that will receive the invading bacteria¹⁴.

Multiple extracellular-domain-containing receptors are required for a complete plant response to Nod factor⁶ (FIG 2) and in the absence of a functional MtNFP (M. truncatula Nod factor perception) gene, M. truncatula cannot respond¹⁷. MtNFP is a member of the LysM family of receptors, and is required for root hair curling and the induction of a subset of transcriptional changes in response to Nod factor^17,18. Recently, RNA interference (RNAi) partial depletion experiments demonstrated an additional role for MtNFP during a later step in bacterial penetration of the root hair¹⁸ (infection thread formation; see below). Another LysM family receptor, encoded by MtLYK3/HCL, is known to be required for CCRH formation and fine-tuning of the plant responses to Nod factor during infection thread formation (P. Smit and T. Bisseling, unpublished data)¹⁹.

Figure 2. Downstream components of the Nod factor signal transduction system.

A complete response of Medicago truncatula to Nod factor (NF) from Sinorhizobium meliloti requires multiple extracellular-domain-containing cell-surface receptors, including the LysM family receptors MtNFP and MtLYK3/HCL. DMI2 encodes a leucine-rich-repeat receptor kinase that is localized to the membrane and is required for tight root hair curling around the bacteria. Downstream of NF, DMI1 encodes a ligand-gated ion channel that localizes to the nuclear membrane. DMI3 encodes a Ca²⁺–calmodulin-dependent protein kinase that is required for the induction of cell division in the root cortex and for the transcriptional changes required for the establishment of the symbiosis. Two GRAS family transcriptional regulators, nodulation signalling pathway 1 (NSP1) and NSP2, are also required for Nod-factor-induced transcriptional changes. The response to NF also involves Ca²⁺ spiking.

Many downstream components of the Nod factor signal transduction system in leguminous plants have recently been identified^4–7 (FIG. 2). DMI1 (does not make infections 1) encodes a ligand-gated ion channel that localizes to the nuclear membrane, and null mutations in DMI1 eliminate the calcium spiking response to Nod factor^20,21. DMI3 encodes a Ca²⁺–calmodulin-dependent protein kinase (CCaMk) that is necessary for the induction of cell division in the root cortex and for the transcriptional changes that are required for the establishment of the symbiosis^22,23. Expression of a constitutively active allele of DMI3 induces these cell divisions and transcriptional changes, but does not induce root hair curling nor promote entry of bacteria, suggesting that these responses require other intermediates²⁴. DMI2 encodes a leucine-rich-repeat receptor kinase that is required for tight root hair curling around the bacteria²⁵. Two GRAS family transcriptional regulators, nodulation signalling pathway 1 (NSP1) and NSP2, are also required for Nod-factor-induced transcriptional changes^26,27. Additionally, calcium spiking and transcript induction are dependent on phospholipid signalling pathways²⁸. Transcriptional changes induced early in the nodulation programme, which are mainly due to the actions of Nod factor, have been characterized in several recent studies, and purified Nod factor can induce many of these signalling events in the absence of bacteria^2,8. However, although Nod factor is necessary for nodule formation and S. meliloti invasion, it is not the only bacterially produced effector that is required for these symbionts to enter plant tissues and colonize plant cells.

Infection thread development

In all but the most primitive rhizobial–host symbioses, the bacteria must be internalized by plant cells in the root cortex before they can begin to fix nitrogen¹. The bacteria penetrate these deeper plant tissues through the production of infection threads (FIG. 3). To form these structures, the bacteria must first become trapped at the root hair tip in a CCRH. Several S. meliloti mutants have been isolated that are inefficient in colonizing root hairs. Mutants that are unable to produce cyclic β-glucans cannot attach to root hairs effectively, a defect that might be due to a disrupted interaction with the host cell surface^29,30. A constitutively active mutant of the transcriptional regulator exoS and a null mutant of the transcriptional regulator exoR are both less efficient than wild-type bacteria at colonizing root hairs³¹. Both of these mutants overproduce succinoglycan, but this does not appear to be the cause of their symbiotic defect, which could be caused by the lack of flagella or other currently unidentified factors^31–33.

Figure 3. Root hair invasion by Sinorhizobium meliloti.

a | S. meliloti exoY³⁶,³⁹ and Medicago trunculata MtLIN⁴³ and MtNIN⁶¹ are required for infection thread initiation. b | S. meliloti exoH³⁹ and M. trunculata MtNFP¹⁸, MtLYK3/HCL¹⁹ (P. Smit and T. Bisseling, unpublished data), MtBIT1/ERN⁴⁴, MtNIN⁶¹ and MtCRE1 (REFS ⁵⁸–⁶⁰) are required for infection threads to extend to the base of the root hair cell. c | MtCRE1 (REFS ⁵⁸–⁶⁰), MtBIT1/ERN⁴⁴, MtRIT1 (REF. ⁶²) and MtSLI⁶⁴ are required for infection thread penetration into the underlying cell layers. The required rhizobial genes are boxed in brown and the required plant genes are boxed in light green.

Bacteria that have become trapped in a CCRH at the root hair tip, and can produce both Nod factor and a symbiotically active exopolysaccharide, induce the progressive ingrowth of the root hair cell membrane, resulting in bacterial invasion of interior plant tissue¹³. Efficient invasion occurs even if Nod factor and exopolysaccharide are supplied by separate S. meliloti strains that are co-inoculated and trapped together in the same CCRH³⁴. The tip of the developing infection thread is a site of new membrane synthesis, and is proposed to involve inversion of the tip growth that is normally exhibited by the root hair and to be the result of reorganization of cellular polarity¹³. Microscopic analysis of fluorescently tagged bacteria within infection threads indicates that only the bacteria at the tip of the ingrowing infection thread are actively dividing³⁵. Invasion appears to progress by continued bacterial proliferation at the tip and sustained induction of infection thread membrane synthesis.

S. meliloti produces the exopolysaccharides succinoglycan (also known as exopolysaccharide I, EPSI) and galactoglucan (EPSII), which facilitate infection thread formation^36,37. Succinoglycan is a polymer of an octasaccharide repeating unit modified with acetyl, succinyl and pyruvyl substituents³⁸ (BOX 1), is more efficient than galactoglucan in mediating infection thread formation on M. sativa and is the only exopolysaccharide produced by S. meliloti that can mediate the formation of infection threads on M. truncatula^36,37. An S. meliloti exoY mutant, which cannot produce any succinoglycan, can form CCRHs but initiates almost no infection threads — these mutants remain trapped in a microcolony at the tip of the root hair³⁹ (FIGS 3,4; TABLE 1).

Box 1. Sinorhizobium meliloti succinoglycan biosynthesis.

Sinorhizobium meliloti strain 1021 requires the exopolysaccharide succinoglycan to establish a functional symbiosis with Medicago truncatula and Medicago sativa^36,143. Bacterial mutants that are defective in succinoglycan production cannot initiate infection thread formation and yield nodules that are devoid of bacteria or bacteroids^36,143. Succinoglycan is a polymer of an octasaccharide repeating unit composed of one galactose and seven glucose residues with acetyl, succinyl and pyruvyl modifications^38,144 (see figure). Succinoglycan is produced in a high molecular weight (HMW) form, as a polymer containing hundreds of the octasaccharide repeating unit, and a low molecular weight (LMW) form that is composed of monomers, dimers and trimers of the repeating unit¹⁴⁵. Application of the purified trimer fraction can partially rescue infection thread formation of a succinoglycan-deficient mutant. However, the variability of invasion under these conditions suggests that succinoglycan must be applied within the colonized curled root hair to function effectively¹⁴⁵.

Many gene products are required for succinoglycan biosynthesis. ExoC, ExoB and ExoN are involved in the biosynthesis of the precursors UDP-galactose and UDP-glucose^146,147. ExoY initiates the synthesis of the repeating unit, by transferring the galactosyl residue to the lipid carrier on the cytoplasmic side of the inner membrane¹⁴⁸, and sugar transferases encoded by exoA, exoL, exoM, exoO, exoU and exoW subsequently elongate the octasaccharide backbone¹⁴⁸. ExoZ, ExoH and ExoV transfer succinyl (Succ), acetyl (Ac) and pyruvyl (pyr) groups, respectively, to the growing backbone^149,150. The mature repeating unit (linked to the lipid carrier) is flipped to the periplasmic side of the inner membrane, and then transferred to the growing polymer of succinoglycan by ExoP, ExoQ and ExoT^148,149,151. The high molecular weight fraction of this polymer can be further processed by two extracellular endo-glycanases, ExoK and ExsH, to the low molecular weight form of succinoglycan⁴⁰. UDP, uridine diphosphate.

Figure 4. Infection thread failure can be caused by plant or bacterial defects.

Infection thread formation during invasion of Medicago truncatula or Medicago sativa by Sinorhizobium meliloti. a | An M. sativa root hair cell infected by S. meliloti wild-type bacteria, with a fully extended infection thread³⁶. b | An M. sativa root hair arrested at the colonized curled root hair (CCRH) stage during infection by an S. meliloti exoY mutant³⁶. c | A M. truncatula lin mutant arrested at the CCRH stage during infection by wild-type S. meliloti⁴³. d | An arrested CCRH, formed during infection of M. truncatula with an S. meliloti nodF nodL mutant¹⁹. e | An aberrant, aborted infection thread formed by wild-type S. meliloti on M. truncatula partially depleted of MtNFP mRNA by RNA interference (RNAi)¹⁸. f | An aberrant infection thread formed by an S. meliloti nodF nodE mutant on M. truncatula partially depleted for MtLYK3 by RNAi¹⁹. Parts a and b reprinted with permission from REF. 36 © (2000) The American Society for Microbiology. Part c and e reprinted with permission from REF. 43 © (2004) American Society of Plant Biologists. Parts d and f reprinted with permission from REF. 19 © (2003) American Association for the Advancement of Science.

Table 1.

Bacterial and plant genes that are involved in infection thread development

Sinorhizobium meliloti 1021 genotype	Medicago truncatula genotype	Infection thread phenotype	Comment
exoY mutant: succinoglycan-deficient39	Wild type	None initiated	Succinoglycan is necessary for infection thread initiation
Wild type	Mtlin mutant43	Very few initiated; remainder aborted	Implies LIN is a possible plant product required for infection thread formation
exoH mutant: succinoglycan lacks succinyl modification, and consequently is mostly in HMW form39,40	Wild type	Few initiated; remainder aberrant or aborted	Implies succinoglycan structural features are important for infection thread extension
NodFL double mutant: Nod factor lacks the acetyl substituent and carries an altered acyl chain19,42	Wild type	Few initiated; remainder aberrant or aborted	Correct Nod factor structure is needed for infection thread formation as well as other responses
Wild type	RNAi knockdown of Nod factor receptor MtNFP18	Aberrant	Implies a higher threshold of Nod factor perception or transduction is needed for infection thread formation than for other Nod factor responses
NodFE double mutant: Nod factor carries an altered acyl chain19	RNAi knockdown of secondary Nod factor receptor MtLYK3/HCL19	Aberrant	Suggests that secondary Nod factor receptor and Nod factor structural features interact to facilitate infection thread formation
Wild type	Mutant of putative transcription factor MtBIT1/ERN44	Aberrant or aborted	Implies the function of putative transcription factor ERN is required for infection thread extension
Wild type	Mutant of MtNIN61	Very few initiated; remainder aborted	Suggests reinitiation of cell division in the plant cortex is necessary for initiation and maintenance of infection threads.
Wild type	RNAi knockdown of putative cytokinin receptor MtCRE1 (REF. 58)	Aborted	Suggests re-initiation of cell division in the plant cortex is necessary for initiation and maintenance of infection threads
Wild type	Mutants of MtRIT1 (REF. 62) or MtSLI64	Extended to base of root hair cell; aborted in underlying cell layers	Implies RIT1 and SLI are possible plant products required for infection thread penetration

HMW, high molecular weight; RNAi, RNA interference.

An S. meliloti exoH mutant produces succinoglycan that lacks the succinyl substituent, and this mutant forms abortive, aberrant infection threads³⁹ (TABLE 1). Unlike the normal condition, aberrant infection threads are bulbous and winding, as if the bacteria have continued to proliferate even though the tip of the infection thread has halted its inward progression^36,39. In contrast to exoY-mutant-containing CCRHs, 70% of exoH-mutant-containing CCRHs initiate infection threads, but they all abort before reaching the base of the root hair³⁹. Interestingly, the succinoglycan polymer produced by an exoH mutant is mostly in the high molecular weight form because the lack of the succinyl substituent renders it refractory to cleavage by extracellular glycanases⁴⁰. It is not clear whether infection thread extension is dependent on the presence of the succinyl substituent of succinoglycan or on the production of low molecular weight (trimer, dimer and monomer) forms of succinoglycan. If succinoglycan has a role in signalling to the plant, perhaps the low molecular weight forms can interact more readily with the plant cell membrane, whereas the high molecular weight forms, which can be composed of polymers of hundreds of monomers, are sterically hindered by the plant cell wall. In a different host–symbiont pair, involving vetch (Vicia sativa) and Rhizobium leguminosarum, the structure of the bacterial exopolysaccharide appears to be less critical for infection thread extension⁴¹.

As mentioned earlier, in addition to succinoglycan, Nod factor is required for active and ongoing infection thread formation. Aberrant, abortive infection threads are formed when Nod factor is supplied in a structurally incomplete form, or if Nod factor perception is disrupted^18,19,42. An S. meliloti nodF nodL double mutant that produces Nod factor that lacks the acetyl substituent and has an altered acyl chain cannot effectively invade plant cells, and induces approximately 20% of colonized root hairs to form infection threads^19,42. Those infection threads that are formed are aberrant and ‘sac-like’, and abort before reaching the base of the root hair cell¹⁹ (FIG. 4 and TABLE 1). This is similar to the invasion phenotype of the exoH mutant described above³⁹.

In relation to the host plant, mutant isolation and RNA interference have been used to identify genes encoding proteins that are involved in infection thread formation. The M. truncatula lin mutant, when inoculated with wild-type S. meliloti, has an abortive infection thread phenotype that is similar to that of a wild-type plant inoculated with an S. meliloti exoY mutant^39,43 (TABLE 1). In both of these cases, infection arrests at the CCRH stage (FIG. 4). A deletion mutant of MtBIT1/ERN, which encodes a putative transcription factor involved in relay of the Nod factor signal, forms normal CCRHs but aberrant infection threads that cannot extend into deeper plant tissues⁴⁴. RNAi knockdown has been used to assess the role of Nod factor receptors in infection thread formation. M. truncatula plants that are partially depleted by RNAi for the primary Nod factor receptor MtNFP, although capable of root hair curling and initiation of cortical cell divisions, form aberrant infection threads¹⁸ (FIG. 4 and TABLE 1). RNAi depletion of the secondary Nod factor receptor MtLYK3/HCL (also known as the entry receptor) does not cause wild-type S. meliloti to form aberrant infection threads; however, it does cause an S. meliloti nodFE mutant (which produces Nod factor with a modified acyl chain) to form defective infection threads¹⁹ (FIG. 4 and TABLE 1). These results indicate that the Nod factor signal requirement for infection thread formation is more stringent and perhaps of greater amplitude than that required for CCRH formation and the initiation of cortical cell divisions.

Other than the genetic and biochemical demonstrations of the requirement for succinoglycan and Nod factor, little is know about the way in which these factors mediate the extensive cytoskeletal remodelling that occurs during infection thread formation, or how the cytoskeleton might be involved in directing vesicular traffic to this site of new membrane synthesis. The cytoskeletal changes that occur during tip growth of polarized root hairs⁴⁵ and the early effects of Nod factor on the root hair cytoskeleton that lead to root hair deformation have been subject to detailed analysis^13,46,47. However, the cytoskeletal reorganizations involved in infection thread formation have not yet been thoroughly dissected. Inward movement of the root hair cell nucleus, in advance of the ingrowing infection thread, has been observed⁸. In these cells, a network of microtubules connects the nucleus to the tip of the infection thread and surrounds the infection thread itself ⁸. As the infection thread develops, the inside of this tubule is topologically outside the root hair cell and possesses a plant cell wall⁴⁸. Interactions between plant cell-wall material and other components of the infection thread matrix might play an important part in infection thread growth⁴⁸. Plant cell wall remodelling during root hair colonization and infection thread formation is another process that is not well understood, but is proposed to involve plant proteins that are exported to the cell wall and the infection thread matrix⁴⁸. Plant glycoproteins that are associated with the infection thread matrix become crosslinked, probably by reactive oxygen species (ROS) generated in the infection thread^49–51. One hypothesis is that solidification of the infection thread matrix, as a result of this crosslinking, generates a mechanical force that helps the ingrowth of the infection thread to proceed⁴⁸. The production of ROS by plants is often part of a defence response against invading pathogens and it can also serve as an activating signal for other plant defences (see below)⁵². One intriguing possibility is that the facilitation of infection thread progression by ROS is one way that rhizobial bacteria have adapted a plant defence response for their own benefit.

The correct balance of ROS in the infection thread might be necessary for infection thread progression, and an excess of ROS could be detrimental to the rhizobia within the infection thread. However, the level of stress experienced by bacteria in the infection thread due to ROS is unknown. Single mutants in any of the three S. meliloti catalases, one of the two superoxide dismutases (sodB) or in the global regulator of H₂O₂ protection, oxyR, are sensitive to ROS, but are not defective in symbiosis^53–56. However, double mutants in the genes encoding catalase B and catalase C (katB/katC) or catalase A and catalase C (katA/katC) are compromised in their ability to invade the host plant^53,54. Mutants in a second superoxide dismutase, sodC, a predicted sodM (Smc00911) or a secreted chloroperoxidase (Smc01944) have not yet been analysed⁵⁷. Further work is therefore required to elucidate how ROS, plant proteins and cell-wall materials interact with exopolysaccharides or other products that are secreted by the bacteria to influence development of the infection thread.

Targeting infection threads

Once an infection thread has penetrated to the base of a root hair cell the bacteria must induce new rounds of infection thread formation in each successive cell layer. Two important questions are: how is infection thread growth targeted to the cortical tissue layer; and, following the penetration of the cortex by an infection thread, what molecular processes mediate endocytosis of bacteria by plant cells into host-membrane-bound compartments?

The plant hormone cytokinin and the Nod-factor-dependent reinitiation of the cell cycle are involved in directing infection threads to the plant cortex. RNAi-mediated depletion of the M. truncatula putative cytokinin receptor/histidine kinase MtCRE1 results in a block in the reinitiation of cell division in the root cortical cells and the abortion of infection threads⁵⁸ (TABLE 1). A mutation in the Lotus japonicus orthologue of MtCRE1, LHK1 (BOX 2), causes infection threads to arrest or loop and wander in the plant epidermal tissue instead of penetrating straight into the deeper cell layers⁵⁹. It appears that this cytokinin receptor has an important role in initiating cortical cell division, as a constitutively active mutant of L. japonicus LHK1 induces spontaneous initiation of cell division in the absence of infecting bacteria⁶⁰. The induction of the putative transmembrane-domain-containing transcription factor MtNIN is dependent on MtCRE1⁵⁸. Mtnin mutants are particularly interesting because they have early signalling responses to Nod factor but are impaired in both induction of cortical cell division and infection thread development, suggesting that MtNIN might be involved in integrating these processes⁶¹ (TABLE 1). This is consistent with a model in which cortical cell divisions provide a signal for the initiation and maintenance of infection threads. Mtrit1 and Mtsli mutants are also defective in penetration of the infection thread into deeper cortical tissue, but the genes carrying these mutations have not yet been cloned^62–64 (TABLE 1).

Box 2. Mesorhizobium loti and Lotus japonicus: model system for determinate nodules.

Rhizobium–legume symbiosis research has progressed quickly using the Mesorhizobium loti–Lotus japonicus model system, in addition to the Sinorhizobium meliloti–Medicago truncatula model system as discussed in the text. The Nod factor signalling pathway in these two models is similar and has many conserved components (see table). However, what makes comparative study of both these systems most interesting is that they form different types of nodules, in which the nodule structure and the differentiation programme imposed on the symbiotic bacteria are different. M. truncatula and Medicago sativa are examples of indeterminate legumes, which form nodules that possess a persistent meristem, whereas L. japonicus forms determinate nodules, which lack a persistent meristem and in which all cells in the interior of the nodule proliferate, differentiate and senesce synchronously⁶⁵. Engineering of bacterial symbionts that nodulate the opposite type of host has determined that the plant controls the degree of differentiation and the ultimate fate of the bacteria⁶⁵. Differentiated bacterial symbionts isolated from L. japonicus nodules have not undergone genomic endoreduplication and can resume growth⁶⁵. However, fully differentiated bacteroids isolated from M. truncatula are unable to resume growth, perhaps because of effects of extensive genomic endoreduplication on the bacterial cell cycle⁶⁵.

Medicago trunculata protein	Lotus japonicus orthologue	Function
MtNFP	LjNFR5 (REF. 4)	Nod factor receptor
Unknown	LjNFR1 (REF. 4)	Nod factor receptor
MtDMI2 (REF. 25)	LjSymRK4	Leucine-rich repeat receptor-like kinase
MtDMI1 (REFS 20,21)	LjCASTOR/POLLUX152	Ion channel
Unknown	LjNUP133, LjNUP85 (REFS 153,154)	Nucleoporin
MtDMI3 (REFS 22,23)	LjSYM15/snf1 (REF. 155)	Ca2+–calmodulin-dependent protein kinase
MtNSP1 (REF. 27)	LjNSP1 (REF. 156)	GRAS family transcriptional regulator
MtNSP2 (REF. 26)	LjNSP2 (REF. 156)	GRAS family transcriptional regulator
MtCRE1 (REF. 58)	LjLHK1 (REF. 59)(not required for infection thread formation)	Cytokinin receptor or histidine kinase
MtNIN61	LjNIN157	Predicted transmembrane transcription factor

The earliest infection threads that penetrate the growing M. truncatula or M. sativa nodule must grow past the actively dividing cells in the developing nodule primordium¹⁴. Ultimately, cells adjacent to the initial primordium will give rise to a persistent nodule meristem that maintains a population of actively dividing cells and will continue to grow outward from the root for the life of the nodule^8,14. M. truncatula and M. sativa are examples of indeterminate legumes, which form nodules that possess a persistent meristem. Determinate legumes, such as L. japonicus (BOX 2), produce nodules without a persistent meristem and in which all cells in the interior of the nodule proliferate, differentiate and senesce synchronously⁶⁵. In developing indeterminate nodules, which we focus on here, infection threads grow through the developing meristematic zone to the underlying cell layers through which the wave of mitotic activity has already passed⁶⁵. These layers of invasion-competent cells have exited mitosis and become polyploid owing to cycles of genomic endoreduplication without cytokinesis^14,66. Plant cell endoreduplication is required for the formation of functional nodules, and is dependent on the degradation of mitotic cyclins by anaphase-promoting complex (APC), an E3 ubiquitin ligase^66,67. The advantages of cellular polyploidy are a higher transcription rate and a higher metabolic rate than cells with the normal 2n DNA content⁶⁸ — characteristics that might be important for the invaded plant cells to support bacterial nitrogen fixation (discussed later).

Endocytosis of rhizobia

When the bacteria reach the target tissue layer, the inner plant cortex, they must be internalized by a cortical cell and establish a niche within that cell. Each bacterial cell is endocytosed by a target cell in an individual, unwalled membrane compartment that originates from the infection thread⁴⁸. The entire unit, consisting of an individual bacterium and the surrounding endocytic membrane, is known as the symbiosome⁴⁸. In indeterminate nodules, a bacterial cell and its surrounding membrane divide synchronously before the bacteria differentiate into nitrogen-fixing bacteroids⁶⁹. New lipid and protein material targeted to the symbiosome membrane imparts a distinct biochemical identity to this compartment⁶⁹.

Both plant and bacterial defects can cause a failure of symbiosome formation. When the plant leucine-rich-repeat receptor kinase DMI2 is partially depleted by RNAi, the result is overgrown infection threads that fail to release bacteria into symbiosomes⁷⁰. In wild-type infections, the DMI2 protein is localized to the infection thread and symbiosome membranes, which is consistent with a role for this protein in release of bacteria from infection threads and symbiosome development⁷⁰. RNAi-mediated knockdown of the MtHAP2-1 gene, which encodes a predicted transcription factor, also prevents bacterial release from infection threads⁷¹. The M. truncatula nip mutant also produces overgrown infection threads and the bacteria fail to release from these infection threads⁷². The NIP gene has not yet been cloned, but it is likely that its gene product is involved in symbiosome development. Bacteria can also fail to release from infection threads owing to bacterial defects. The hemA mutant of S. meliloti, which has a primary defect in haem biosynthesis, is not released from infection threads and is not enclosed in symbiosomes⁷³. Several bacterial components that are important for symbiosis are known to require haem, such as the oxygen sensor FixL and cytochrome haem proteins, which could explain this phenotype^74,75.

Recent proteomic and immunolocalization studies have begun to define biochemical markers of the symbiosome membrane. These include previously identified nodule-specific proteins (ENOD8, ENOD16, nodulin 25 and nodulin 26), energy and transport proteins, bacterial proteins, and proteins predicted to be involved in folding, processing, targeting and storage⁷⁶. Based on the types of protein that have been extracted from the symbiosome membrane, Catalano et al. proposed that proteins are added to this membrane by multiple mechanisms, one of which is being targeted by syntaxin proteins to the symbiosome via the Golgi and membrane-bound vesicles⁷⁶. One particularly interesting protein that is predicted to be involved in the targeting of vesicles to the symbiosome membrane is the SNARE protein MtSYP132 (M. truncatula syntaxin of plants 132)⁷⁷. Syntaxins are involved in the docking of vesicles at the target membrane, and might be crucial for the development of the rhizobium-containing vesicle into a mature symbiosome⁷⁸.

As rhizobial bacteria are enclosed in the symbiosome membrane by endocytosis, it has been assumed that this compartment is a derivative of the lytic vacuole (the plant counterpart to the mammalian lysosome)^79,80. Consistent with this hypothesis, several enzymes that localize to the lumen of the soybean symbiosome, or the peribacteroid space, have an acidic pH optimum⁷⁹. However, parallels have also been drawn between the symbiosome and plant protein-storage vacuoles that are initially a non-lytic niche^79,81.

Bacteroid differentiation and survival

Once the bacteria have been engulfed within host cell membranes, they must survive within the symbiosome compartment and differentiate into the nitrogen-fixing bacteroid form. Both bacterial and plant factors are involved in these processes (FIG. 5).

Figure 5. Endocytosis of bacteria and bacteroid differentiation.

Bacterial endocytosis requires the Sinorhizobium meliloti hemA gene⁷³, the Medicago truncatula NIP gene⁷² and wild-type expression levels of the MtDMI2 (REF. ⁷⁰) and MtHAP2-1 (REF. ⁷¹) genes. S. meliloti lpsB⁸² and bacA⁸³ are required for bacterial survival within the symbiosome membrane. S. meliloti fixJ⁹⁸, M. truncatula MtSYM1 (REF. ⁶⁴), MtDNF1, -4, -5 and -7 (REFS ⁶²,⁶³), and pea (Pisum sativum) PsSYM13 (REF. ⁹⁵) are required for bacteroid differentiation. The S. meliloti nifHDK genes encode nitrogenase and are required for nitrogen fixation⁹⁸. The pea PsRUG4 gene encodes sucrose synthase and is required to support bacteroid nitrogen fixation¹¹⁵,¹¹⁶. The M. truncatula MtDNF3 and -6 genes are required for the maintenance of nitrogen fixation⁶²,⁶³. The required rhizobial genes are boxed in brown and the required plant genes are boxed in light green.

LPS and rhizobial survival

One of the most important defence mechanisms that Gram-negative bacteria use against the extracellular environment is the lipopolysaccharide (LPS), which is a major component of the outer membrane. LPS consists of a lipid A membrane anchor attached to a polysaccharide core that in turn is attached to an O-antigen repeating unit⁸². The S. meliloti bacA gene is required by S. meliloti for the production of lipid A of the appropriate structure and for survival within host cells^83,84 (BOX 3). The bacA mutant cells lyse soon after endocytosis and do not exhibit any morphological changes that are characteristic of bacteroids, such as cell elongation⁸³. BacA is required for modification of lipid A with a very-long-chain fatty acid moiety, and the fatty acid defect may be the cause of the bacA mutant symbiotic defect⁸⁴ (BOX 3). An S. meliloti lpsB mutant produces an LPS core with an altered sugar composition⁸² (BOX 3). Although it can invade the plant root, become enclosed within a symbiosome membrane and begin to elongate into the bacteroid form, it cannot complete the bacteroid differentiation programme and fix nitrogen⁸². The lpsB mutant symbiotic defect is probably a result of its compromised outer membrane; however, as these bacteria also appear to lack contact with the plant symbiosome membrane, it is possible that they fail to survive because they cannot interact properly with the host⁸².

Box 3. Host invasion parallels between rhizobia and animal pathogens.

Symbiotic rhizobial bacteria and pathogenic Brucella species both form chronic infections of eukaryotic cells within a host-membrane-derived compartment¹⁵⁸. Brucella spp. are α-proteobacteria that are closely related to Sinorhizobium meliloti and invade mammalian cells, causing brucellosis¹⁵⁹. Both Brucella spp. and S. meliloti must gain entry to target host cells and prevent their own destruction by the lytic compartments of these cells. Bacterial production of lipopolysaccharide (LPS) with the appropriate structure (see main text) is important for survival of both Brucella spp. and S. meliloti within host cells. The S. meliloti lpsB mutant produces an LPS core with altered sugar composition (see figure) and the Brucella abortus pmm mutant produces a truncated LPS core^82,160,161. Neither mutant can survive within host cells and this might be because their outer membranes are compromised^82,160. These mutants are also sensitive to detergents and to antimicrobial peptides, such as polymyxin B, that resemble defence compounds produced by phagocytic cells and plant cells^82,162,163. LPS lipid A, with the appropriate structure, might also be important for the survival of S. meliloti and B. abortus within their respective hosts^83,158. The bacA gene encodes a predicted inner membrane protein with similarity to peroxisomal-membrane transporters of fatty acids and is required for the modification of lipid A with very-long-chain fatty acids in both S. meliloti and B. abortus (see figure)⁸⁴. However, S. meliloti mutants that are unable to synthesize this modified lipid A in the free-living form can form an effective symbiosis, raising the possibility that another function of BacA is required¹⁶⁴. The production of periplasmic cyclic β-glucans is required for Brucella spp. to survive within intracellular compartments and for S. meliloti to adhere to root hair cells^29,165. During host cell invasion by B. abortus, cyclic β-glucans prevent fusion of the bacteria-containing endocytic vacuole with the lysosome, by modulating lipid raft organization in host cell membranes¹⁶⁵. This suggests that cyclic β-glucans are critical for correct cell surface interactions of B. abortus with endocytic vacuole membranes and of S. meliloti with root hair cells.

Regulatory genes conserved between Brucella spp. and S. meliloti are also important for either invasion or persistence within their respective hosts. For example, the B. abortus bvrS/bvrR sensor kinase/response regulator system is required for intracellular survival^31,166. The B. abortus bvrS/bvrR system may be required for virulence, because it controls the expression of several outer membrane proteins and is involved in modulating the properties of bacterial lipopolysaccharide ^162,167. It has not been possible to construct null mutants of the S. meliloti homologues of these genes, exoS and chvI, indicating that they are essential^32,168. However, a mutant that produces a constitutively active ExoS sensor kinase is partially impaired in its ability to infect root hairs (see main text)^31–33. A Tn5 insertion mutant of another S. meliloti sensor kinase, cbrA, is also symbiotically defective¹⁶⁹. This mutant is sensitive to agents that stress the bacterial membrane, and this may be the cause of its symbiotic defect¹⁶⁹. The B. abortus homologue of cbrA, pdhS, is essential for viability and therefore it has not been possible to assess its requirement for pathogenesis¹⁷⁰. Panels a and c reprinted with permission from REF. 83 © (1993) Cold Spring Harbor Laboratory Press. Panel b reprinted with permission from REF. 82 © (2002) National Academy of Sciences. The scale bars represent 1 μm for a and c, and 0.21 μm for b.

Non-LPS factors and rhizobial survival

S. meliloti mutants that are defective in other aspects of physiology are also defective in symbiosis with the plant host. The sitA mutant invades plant cells at a reduced efficiency, but the bacteroids formed are senescent and cannot effectively fix nitrogen⁵⁵. The sitA gene encodes a component of a manganese transporter, and the mutant cannot grow in minimal medium without additional manganese⁵⁵. Although this mutant is sensitive to ROS, this is apparently not the cause of the symbiotic defect, suggesting that an Mn²⁺-requiring enzyme or intracellular Mn²⁺ homeostasis is important for symbiosis⁵⁵. Mutants in the pha gene cluster have a symbiotic phenotype that is similar to that of sitA mutants, with a reduced efficiency of infection and formation of senescent bacteroids⁸⁵. Interestingly, the product of the pha gene cluster is a K⁺-efflux system that is active under alkaline conditions⁸⁵. This suggests that mechanisms that cope with specific ion stress are required for the bacteria that invade, and to persist in, the intracellular environment. Mutants that are defective in global bacterial transcriptional responses under stress conditions are also defective in symbiosis. A mutant in rpoH1, encoding a putative σ³²-factor that is predicted to interact with bacterial RNA polymerase, appears to invade normally, but forms bacteroids that senesce prematurely⁸⁶. By contrast, a mutant in relA is defective at multiple stages in nodule formation⁸⁷; relA mutants lack the stringent response — a bacterial transcriptional response to starvation⁸⁷.

Plant control and support of bacteroid differentiation

The host plant exerts control over the survival of bacteria within the symbiosome, and must not only provide nutritional support and the correct micro-aerobic environment required for nitrogen fixation, but also provide specifications for part of the bacterial differentiation programme. In indeterminate nodules, the internalized bacteria and the symbiosome membrane divide concomitantly before bacteroid differentiation, whereas in determinate nodules, bacteria divide within the membrane compartment and form a small mass of cells⁸⁸. Surprisingly, plants that form indeterminate nodules impose a programme of genomic endoreduplication on the invading bacterial cells⁶⁵. Bacteroids within indeterminate nodules increase their DNA content and cell size, which might allow them to reach a higher metabolic rate to support nitrogen fixation^65,68. The intensive DNA synthesis that is required for endoreduplication within bacteroids requires a large quantity of dNTPs that must be supplied by ribonucleotide reductase (RNR)⁸⁹. For many bacterial species, performing DNA synthesis in both an oxygen-rich environment, such as the infection thread, and an oxygen-depleted environment, such as the symbiosome, would present a lethal problem. This is because most aerobes possess an oxygen-requiring class I RNR, whereas obligate anaerobes typically have an exquisitely oxygen-sensitive class III RNR⁹⁰. Some bacterial species, including rhizobia, have an adaptation that circumvents this problem — they possess a vitamin B₁₂-dependent class II RNR that functions independently of oxygen concentration. Significantly, when the class I RNR of Escherichia coli is substituted for the class II RNR from S. meliloti, survival within the nodule, but not free-living growth, is impaired (M. E. T. and G. C. W., unpublished data). Phosphate must also be obtained by bacteria during the invasion process. S. meliloti that is mutated in both the phoCDET and pstSCAB phosphate-transport systems forms small nodules that cannot fix nitrogen^91,92. However, these transporters are not expressed by mature bacteroids, suggesting that the requirement for phosphate declines after differentiation⁹².

Plants control bacteroid differentiation and survival in other ways that are in the process of being characterized. Bacteria fail to differentiate or senesce prematurely in several mutants of the host plants M. truncatula and pea (Pisum sativum), but none of the genes carrying these mutations has yet been cloned^{62–64,93,94}. In the pea sym13 mutant, symbiosome membranes are not closely associated with the bacteroid and the bacteroids senesce early⁹⁵. Immunological detection of plant glycoproteins that are still attached to bacteroids isolated from pea nodules suggests that there is a close physical interaction between the symbiosome and bacteroid membranes, at least in indeterminate nodules where a single bacteroid is enclosed within each symbiosome membrane^96,97. A close association between the bacteroid outer membrane and the symbiosome membrane might be important for nutrient exchange between the bacteria and the plant.

Nodule development and nutrient exchange

Bacteria that have been enclosed within a functional symbiosome membrane, have been provided with a low oxygen environment and have completed the bacteroid differentiation programme can express the enzymes of the nitrogenase complex and begin to fix nitrogen⁹⁸. An oxygen-sensing bacterial regulatory cascade controls the expression of the nitrogenase complex and the microaerobic respiratory enzymes that are required to provide energy and reductant to nitrogenase⁹⁸. This cascade is induced by the presence of low oxygen tension within the differentiating bacteroid^89,99,100. The bacterial regulators include the oxygen-sensing two-component regulatory system FixL and FixJ, NifA, σ⁵⁴ and FixK⁹⁸. These regulators are responsible for many of the changes in gene and protein expression that have been detected during bacteroid differentiation². In general, the transcriptional changes in bacteroids are consistent with the downregulation of most metabolic processes and a concomitant increase in the expression of the gene products involved in nitrogen fixation and respiration². Respiratory activity provides nitrogenase with the 16 molecules of ATP and 8 electrons that are estimated to be required to reduce 1 molecule of N₂ to 2 molecules of NH₄⁺ (REF. ¹⁰¹).

The NH₄⁺ produced by nitrogenase can be secreted by bacteroids. Once secreted, it is thought to be taken up by the plant through NH₄⁺ channels that have been detected in the peribacteroid membrane and then assimilated^88,102. A complex amino-acid cycling system between plant cells and bacteroids might prevent the bacteroids from assimilating NH₄⁺, and allow NH₄⁺ to be secreted for uptake by the plant^{88,101,103,104}. Metabolite analysis and RNAi depletion experiments suggest that NH₄⁺ assimilation by plant cells occurs primarily through glutamine and asparagine synthetases^105–107. In L. japonicus, RNAi depletion of the next enzyme in the nitrogen assimilation pathway, glutamate synthase, results in a reduction of nitrogenase activity and the growth of stunted plants¹⁰⁸.

A constant carbon supply is required to provide metabolites and energy for bacteroid differentiation and nitrogen fixation. Polyhydroxybutyrate (PHB) granules, produced by S. meliloti during the invasion process, are degraded during bacteroid differentiation and appear to be used preferentially as a carbon source at this stage¹⁰⁹. Surprisingly, mutants in PHB synthesis (phbC) or degradation (bdhA) are symbiotically competent, indicating that alternate carbon sources are available to the bacteria during differentiation^109–111. In a mixed infection, phbC and bdhA mutants are less competitive for nodule occupancy than the wild type, suggesting that the ability to synthesize and use PHB might allow the bacteria to compete in situations of reduced carbon availability^109,111. Fixed carbon from the plant is provided to the bacteria in the form of dicarboxylic acids, such as malate, through the bacterial Dct uptake system^101,112. NAD⁺-malic enzyme activity, which produces pyruvate directly from malate, is required for nitrogen fixation by S. meliloti¹¹³. This suggests that the production of acetyl-CoA from malate using malic enzyme and pyruvate dehydrogenase, is important for the funnelling of carbon into the TCA cycle in bacteroids¹⁰¹.

Some plant proteins, in addition to glutamate synthase, have been shown to be required to support nitrogen fixation. Leghaemoglobins are plant-produced, oxygen-binding proteins that impart a red colour to functional root nodules, and which, it has long been proposed, can adjust the microaerobic environment within nodules¹¹⁴. It has recently been confirmed by RNAi that L. japonicus leghaemoglobin is required for nitrogen fixation and to maintain a microaerobic environment in the nodule¹¹⁴. Another required plant protein is sucrose synthase (SS), which has been characterized in pea using the SS1-defective rug4 mutant^115,116. Contrary to its name, sucrose synthase in pea nodules catabolizes sucrose to UDP-glucose and fructose, which is ultimately metabolized to malate and transported to bacteroids^101,105,115. The L. japonicus sulphate transporter Sst1 is also required for nitrogen fixation and has been detected in symbiosome membranes^117,118. Sulphate is required to form iron–sulphur clusters in nitrogenase subunits¹¹⁹. Transcript profiling and proteomic analysis of plant nodules have identified new candidates for RNAi that will almost certainly lead to the discovery of more plant factors required for nodulation².

Bacteria in indeterminate nodules might further benefit from the symbiosis by scavenging for nutrients in the ghost plant cells that are located in the root proximal section of the nodule, in which the original bacteroids have already senesced¹²⁰. These bacteria are released from infection threads into the decaying plant tissue and appear to proliferate. They do not differentiate into bacteroids or fix nitrogen and can probably resume a free-living lifestyle once the nodule has decayed¹²⁰.

Rhizobial evasion of plant innate immunity

It remains to be determined how the host plants tolerate such intimate contact with invading rhizobial bacteria without attacking the invaded tissues. Plants, as well as animals, can mount innate immune responses to microorganisms in response to the perception of conserved microbial factors (microorganism-associated molecular patterns or MAMPs)¹²¹. This usually results in the activation of signalling cascades, and the production of antimicrobial effectors that help the organism ward off microbial attack¹²¹. However, the plant perception of microorganisms differs from that of animals in the domain structure of receptor proteins and, in many cases, the epitopes perceived¹²¹. In the interaction between plants and rhizobial bacteria, some bacterial factors activate innate immune responses whereas other bacterial factors suppress these responses¹²². Defining how plants perceive the conserved microbial molecular patterns of rhizobia and respond to them is currently an active area of research.

MAMPs and host-specific elicitors

Microbial factors that induce innate immune responses in plants are classified as either general or host-specific elicitors⁵². General elicitors include flagellin, elongation factor Tu (EFTu), cold shock protein (CSP), chitin and LPS^123–125. These compounds are conserved across multiple groups of bacteria, allowing plants to perceive and respond to an epitope that is common to many bacteria^123–125. Plants respond to general elicitors with basal defence responses that can impart a measure of resistance to infection⁵². Basal defences include increases in extracellular pH, ethylene, ROS and reactive nitrogen species, phenolic compounds and changes in intracellular Ca²⁺, as well as cell signalling cascades and transcriptional changes⁵². Many plant species mount basal defence responses to the broadly conserved flg22 peptide found in bacterial flagellin¹²³. However, rhizobial bacteria, along with many other α-proteobacteria, such as Agrobacterium tumefaciens and Brucella spp., do not have the flg22 epitope as part of their flagellin sequence^126,127. Two other bacterial epitopes that induce plant defence responses are the elf18 peptide of EFTu and the csp15 peptide from bacterial CSP¹²³. These peptides are conserved in rhizobial bacteria; however legumes lack the ability to mount an innate response to them¹²³. Comparisons have been made between plant basal defence responses to pathogens and plant responses to rhizobial Nod factor⁵¹. One model, suggested by the production of ROS during infection thread formation, is that rhizobial effectors, such as Nod factor, provoke a defence reaction in host plants that is attenuated or modified by other rhizobial effectors, such as LPS (see below)^51,128,129.

Unlike general elicitors, most host-specific elicitors are effector proteins that are injected by pathogens into plant cells by bacterial type 3 secretion systems (T3SSs) and which cause a hypersensitive response and programmed cell death⁵². These responses are more extreme and cause more tissue damage than basal defence responses⁵². One example is the root pathogen Ralstonia solanacearum, which requires the GALA7 type 3 secreted protein to cause wilting disease on M. truncatula¹³⁰.

Nod factor and chitin

Chitin is another general elicitor that induces plant innate immune responses, such as the oxidative burst¹²⁴. This response helps plants defend themselves against fungal attack, as chitin is a component of fungal cell walls¹²⁴. Intriguingly, the chitin receptor of rice (Oryza sativa) is a LysM-domain-containing protein, as are the Nod factor receptors MtNFP and MtLYK3¹³¹. One possibility is that plant proteins containing extracellular, chitin-interacting LysM domains were co-opted in the legume lineage to recognize bacterially produced lipochitooligosaccharide Nod factors.

Negative feedback on nodulation

Plant defence responses have been implicated in restricting the number of nodules that form on a colonized root. After an initial round of nodule formation has begun, subsequent nodulation events are subject to autoinhibition¹³². Signalling that is antagonistic to the early responses to Nod factor is mediated by the plant hormones ethylene and jasmonic acid, which are involved in defence signalling in other plant–microorganism interactions^52,133. It has been proposed that autoinhibition is controlled by the plant by the abortion of infection threads¹³⁴. Late-initiating infection threads that abort are associated with necrotic cells, and with the accumulation of defence-related proteins and phenolic compounds in plant cells and cell walls¹³⁴. This type of cellular damage is also found near aborted infection threads and failed symbiosomes formed by M. truncatula lin, nip and sym6 mutants, and aborted infection threads formed on wild-type M. sativa plants by S. meliloti exoY mutants^{29,43,72,122,134–136}. M. truncatula plants inoculated with the exoY mutant also express several plant defence genes more strongly than plants inoculated with wild-type S. meliloti (K. M. J. and G. C. W., unpublished data). Another study established that plants depleted for a calcium-dependent protein kinase by RNAi over-express defence genes and have a reduced efficiency of nodulation¹³⁷. It remains to be determined whether plant defence responses are a cause of infection thread abortion or a consequence of the failure of infection thread progression, and how the absence of succinoglycan might influence these defence responses.

Exopolysaccharides and β-heptaglucoside

It has been proposed that the exopolysaccharide(s) produced by a rhizobial species are actively involved in suppress ing defence responses in the host plant¹²². In fact, soybean defence responses provoked by the β-heptaglucoside elicitor from the plant pathogen Phytophthora sojae can be suppressed by cyclic-β-glucans produced by the soybean symbiont Bradyrhizobium japonicum¹³⁸. As B. japonicum cyclic-β-glucans are effective antagonists of the soybean β-heptaglucoside receptor, the most likely mechanism for this suppression is competition for the same binding sites¹³⁹. Legumes are the only plant group that possess the β-heptaglucoside receptor and the ability to mount a defence response to this elicitor¹⁴⁰.

LPS

In most plant species, the response to bacterial LPS is different from the toxic shock that can be induced in animals¹⁴¹. Inoculation of a leaf with purified LPS from a plant pathogenic strain can often induce localized resistance in the treated tissue to subsequent infection by that strain¹⁴¹. In many cases, LPS pretreatment can prevent a plant hypersensitive response that would result in catastrophic tissue damage¹⁴¹. One possible explanation is that antimicrobials produced in the LPS-treated leaf tissue prevent the bacteria from proliferating enough to activate a hypersensitive response¹⁴¹. By contrast, plant cells in culture often respond to LPS fractions from bacteria with an oxidative burst and transcription of defence genes¹²⁵. For example, S. meliloti LPS core oligosaccharide can induce an oxidative burst in cultured cells of the non-host plant tobacco (Nicotiana tabacum)¹²⁸. However, the lipid A component of S. meliloti LPS can suppress both an oxidative burst and the expression of defence genes in cultured cells of its host plants M. truncatula and M. sativa^128,129. An exciting area of research is in determining which LPS epitopes from rhizobial bacteria and plant pathogens elicit or suppress defence responses on plants of different lineages. Interestingly, the LPS lipid A component of Brucella abortus, which has a very similar structure to that of S. meliloti, is much less toxic to animal cells than the lipid A of other bacterial species^84,142 (BOX 3).

Future perspectives

It is safe to say that the field of rhizobial–plant symbioses is booming. The genomes of multiple rhizobial species have been sequenced and the genomes of the model host plants M. truncatula and L. japonicus are nearing completion. The cloning of plant Nod factor receptors and signal transduction intermediates, and the dissection of early plant responses to Nod factor have opened a door between symbiosis research and the field of plant signal transduction. Plant hormones and signalling components are sure to have an even more prominent role in future symbiosis studies. Dissecting the plant signalling events that permit the subsequent steps of infection thread formation and bacterial endocytosis will be equally exciting. Another conceptual advance has been the recognition that rhizobia have many features in common with their mammalian pathogen cousins that also reside within intracellular, host-membrane-derived compartments. Despite the differences in the impact on the host of symbiotic and pathogenic bacteria, strategies that mammalian pathogens use to interact with their hosts and to survive within host cells should continue to provide clues as to how rhizobial bacteria accomplish the same mission.

Flavonoid

A 2-phenyl-1,4-benzopyrone derivative, produced by plants, that serves as a defence and signalling compound

Nod factor

A lipochitooligosaccharide compound that induces multiple responses that are required for nodulation of appropriate host plants

Symbiotic exopolysaccharide

A rhizobial secreted β-glucan that is structurally distinct for different species and that mediates infection thread formation

Infection thread

An ingrowth of the root hair cell membrane, populated with rhizobial bacteria, that progresses inward by new membrane synthesis at the tip

Symbiosome

A host-derived membrane compartment that originates from the infection thread housing a bacteroid

Bacteroid

A rhizobial bacterium that has been endocytosed by a plant cell and has elongated and/or branched, and has differentiated or is differentiating into a form that can perform nitrogen fixation

Nitrogen fixation

The reduction of atmospheric dinitrogen to ammonia

Colonized curled root hair

A root hair that has been induced by Nod factor to curl around a microcolony of rhizobial bacteria and entrap it

Auxin

A plant hormone (chiefly indole acetic acid) that regulates plant growth in a concentration-dependent manner

Nodule primordium

Dedifferentiated, proliferating tissue that develops in the plant cortex during nodule initiation

Cyclic β-glucan

A cyclized β-1,2 chain of 17–25 glucose residues produced by rhizobial bacteria, Brucella spp. and Agrobacterium spp. that localizes to the periplasm and that functions in osmotolerance and in interaction with host membranes

Succinoglycan

A symbiotic exopolysaccharide produced by S. meliloti, also known as EPS I, that mediates infection thread formation. An octasaccharide repeating unit modified with acetyl, succinyl and pyruvyl substituents that can be polymerized into a high molecular weight or a low molecular weight form composed of monomers, dimers and trimers

Galactoglucan

A second exopolysaccharide of S. meliloti (EPS II) that can mediate infection thread formation on M. sativa at a low efficiency and that is produced when an intact copy of the ExpR regulator is present. A disaccharide repeating unit modified with acetyl and pyruvyl substituents

Endoreduplication

Genomic replication without cytokinesis that results in greater than 2n DNA content within a cell

Indeterminate nodule

A nodule formed by plants of some clades of legumes that develops a continuously growing nodule meristem at the distal end and has zones of tissue at different stages of development

α-proteobacteria

A group of bacteria that contains several species able to persist within host-derived membrane-bound compartments in eukaryotic cells. Includes rhizobial bacteria and mammalian pathogens such as Brucella spp., Bartonella spp. and Rickettsia spp