LysM domains mediate lipochitin–oligosaccharide recognition and Nfr genes extend the symbiotic host range

Abstract

Legume–Rhizobium symbiosis is an example of selective cell recognition controlled by host/non-host determinants. Individual bacterial strains have a distinct host range enabling nodulation of a limited set of legume species and vice versa. We show here that expression of Lotus japonicus Nfr1 and Nfr5 Nod-factor receptor genes in Medicago truncatula and L. filicaulis, extends their host range to include bacterial strains, Mesorhizobium loti or DZL, normally infecting L. japonicus. As a result, the symbiotic program is induced, nodules develop and infection threads are formed. Using L. japonicus mutants and domain swaps between L. japonicus and L. filicaulis NFR1 and NFR5, we further demonstrate that LysM domains of the NFR1 and NFR5 receptors mediate perception of the bacterial Nod-factor signal and that recognition depends on the structure of the lipochitin–oligosaccharide Nod-factor. We show that a single amino-acid variation in the LysM2 domain of NFR5 changes recognition of the Nod-factor synthesized by the DZL strain and suggests a possible binding site for bacterial lipochitin–oligosaccharide signal molecules.

Introduction

Multicellular organisms share common components and strategies that enable them to distinguish between self and nonself and to establish specific cell–cell interactions (Ausubel, 2005). We investigated such mechanisms in the plant nitrogen-fixing symbiosis. In this interaction, compatible bacteria (collectively called Rhizobium) recognize specific legume hosts and induce development of root nodules, which will host the bacteria and develop into symbiotic organs. The legume–Rhizobium symbiotic interaction is a sequential process where development of nodule primordia from dedifferentiated root cortical cells initiating a meristem, and a bacterial invasion process, run in parallel. In most legumes, the process starts with rhizobial attachment to plant root hair tips. Subsequent physiological and morphological responses result in root hair curling, entrapping bacteria in an infection pocket (Oldroyd et al, 2005; Miwa et al, 2006). Inwards growing infection threads are then formed by the plant and colonized by rhizobia. These infection threads act as rhizobial conduits from which the bacterial symbiont ultimately will be endocytosed into the plant nodule cells. In a fully compatible interaction, symbiosomes consisting of membrane-enclosed rhizobia differentiated into nitrogen-fixing bacteroids will form.

The developmental program leading to nodule organogenesis and infection is controlled by the legume through a suite of nodulin genes which are temporally and spatially regulated (Schultze and Kondorosi, 1998; Colebatch et al, 2004; Mergaert et al, 2006). In the model legume Lotus japonicus, perception of the rhizobial-synthesized lipochitin–oligosaccharide (Nod-factor) is required for initiation of the nodule developmental process through a signaling pathway encoded by seven genes, SymRK, Castor, Pollux, Nup133, Nup85, CCaMK and Cyclops (Kistner et al, 2005) and downstream of this pathway putative transcription factors encoded by LjNin (Schauser et al, 1999), LjNsp1 and LjNsp2 (Heckmann et al, 2006) are required for initiation of nodule organogenesis.

The ability to invade roots of leguminous plants and induce development of root nodules is shared by several bacterial species belonging to both the α- and the β-proteobacteria (Moulin et al, 2001). The relationship between the app. 18 000 legume species and their bacterial microsymbionts is nevertheless selective and this specificity led to the definition of cross-inoculation groups used to describe the symbiotic diversity of legume–Rhizobium symbiosis and to classify bacterial species. Among narrow host range interactions, Sinorhizobium meliloti (S. meliloti) and alfalfa, together with other Medics like Medicago truncatula, belong to one cross-inoculation group, whereas Mesorhizobium loti (M. loti) and Lotus spp. like L. japonicus, belong to another. Reciprocal nodulation between these two groups does not occur. This catalog of cross-inoculation groups provides a practical overview of legume–Rhizobium relationships although broad host range strains like NGR234 (Pueppke and Broughton, 1999) and Bradyrhizobium sp strains lacking the common Nod-factor biosynthetic genes (Giraud et al, 2007) are difficult to accommodate.

A deeper understanding of the classical symbiotic interaction involving Nod-factor signaling emerged from characterization of the two-way signal exchange. Rhizobial NodD proteins mediate host recognition by interacting with specific flavonoids or isoflavonoids exuded from host roots (Mulligan and Long, 1985; Spaink et al, 1989). Flavonoid-activated NodD promotes transcription of bacterial nod-genes involved in synthesis and secretion of lipochitin–oligosaccharides, called Nod-factors, required for initiation of nodulation (Mulligan and Long, 1985; Lerouge et al, 1990; Spaink et al, 1991). The ability of NodD to recognize the type of flavonoid exuded by the plant is one of the primary steps determining host range and expression of a constitutively active nodD transcriptional activator can extend the bacterial host range, bypassing flavonoid activation (Cardenas et al, 1995). For the rhizobial signal, the length of the Nod-factor carbohydrate moiety, the size and degree of saturation of the acyl chain and substitutions of the reducing and nonreducing glucosamine residues (Supplementary Figure 1) are characteristics for each species and these structural features determine whether the bacteria are able to infect legume plants and which plants are infected (Lerouge et al, 1990; Spaink et al, 1991; D'Haeze and Holsters, 2002 and references therein).

Analysis of loss-of-function mutants has previously shown that the L. japonicus LjNfr1 and LjNfr5 genes are required for the earliest cellular and physiological responses to M. loti and M. loti synthesized Nod-factor. Neither root hair deformation, Ca²⁺ spiking nor induction of nodulin gene expression was detected in these mutants (Madsen et al, 2003; Radutoiu et al, 2003; Miwa et al, 2006). The corresponding receptor kinase proteins, LjNFR1 and LjNFR5, were predicted to have a topology where single pass transmembrane domains anchor LysM-containing extracellular domains and intracellular serine/threonine kinase domains (Madsen et al, 2003; Radutoiu et al, 2003). Combining this prediction with the genetic evidence, a receptor complex composed of LjNFR1 and LjNFR5 was proposed to initiate signal transduction in response to Nod-factor (Radutoiu et al, 2003).

Here, we investigate the role of Nod-factor receptor genes in the biodiversity of cross-inoculation groups, and examine their role in specifying the legume's ability to distinguish between different rhizobial strains. We show that legume Nod-factor receptors are important components of the two-way signal and recognition processes determining the host range in plant-bacterial symbiotic interactions. We have further examined the selectivity of the predicted extracellular LysM domains of LjNFR1 and LjNFR5 in relation to Nod-factor recognition and the ability of rhizobial strains to induce root nodule development. We used domain swaps, amino acid substitutions and modeling to demonstrate that the LysM2 domain of NFR5 is important for Nod-factor perception, and present a model for the binding site.

Results

To understand the basis for host/non-host relations, we aimed to identify components required for extending the host range of wild-type legumes. Two different host plants, M. truncatula (Mt) and L filicaulis (Lf), were genetically transformed with L. japonicus (Lj) Nod-factor receptor genes, LjNfr1 and LjNfr5, and inoculated with rhizobial strains nodulating L. japonicus, but unable to nodulate M. truncatula and L. filicaulis wild-type plants. The range of bacterial strains normally able to induce a symbiotic developmental program in M. truncatula plants is limited to strains of S. meliloti, and L. filicaulis does not form nodules with the genetically modified R. leguminosarum biovar viciae DZL strain (Pacios-Bras et al, 2000, 2003), which, in contrast, induced fully infected although ineffective nodules on L. japonicus (Pacios-Bras, 2003 and this study). All the LjNfr1 and LjNfr5 gene constructs (Supplementary Table I) were transferred via A. rhizogenes into plant cells giving rise to transgenic roots (Stougaard et al, 1987; Hansen et al, 1989). Composite plants having both wild-type and transgenic roots were tested for their nodulation capacity. The rhizobial strains used to determine host range changes were M. loti, the symbiont of Lotus species and a modified R. leguminosarum bv. viciae DZL strain that contains a flavonoid-independent NodD activator, NodZ and NolL genes (Supplementary Table II). This strain synthesizes Nod-factors resembling Nod-factor synthesized by M. loti. Rhizobium strains that are not natural symbionts of M. truncatula, L. japonicus or L. filicaulis, respectively, were genetically modified with a constitutive NodD activator. This modification ensured Nod-factor synthesis independent of host-secreted flavonoids (Spaink et al, 1989).

Host range extension in wild-type M. truncatula

Introduction of L. japonicus LjNfr1+LjNfr5 into M. truncatula resulted in development of nodules on transgenic roots after inoculation with M. loti (Figure 1A and Supplementary Figure 2A and Supplementary Table III). Transgenic roots of composite plants (136/499 plants) formed on average 1.7 nodules per plant with M. loti (Table I). Inoculation of LjNfr1+LjNfr5-transformed M. truncatula roots with S. meliloti resulted in 5.8 nodules per transgenic root (43/56 plants). This demonstrated that both M. loti and S. meliloti were recognized by LjNfr1+LjNfr5-transformed M. truncatula roots (Figure 1A and B). Nodules were not observed on M. loti-inoculated transformed roots of M. truncatula carrying the empty vector or LjNfr1 and LjNfr5, separately (Table I; Supplementary Figure 2B). Additionally, inoculation with M. loti nodC∷Tn5, a modified strain unable to produce Nod-factors, did not result in nodule formation on LjNfr1+LjNfr5-transformed M. truncatula roots (Table I). These results demonstrated the involvement of both LjNFR1 and LjNFR5 in perception of M. loti Nod-factors and in the subsequent triggered nodule development in M. truncatula.

Figure 1.

Nodule development on LjNfr1+LjNfr5-transformed M. truncatula roots. (A) M. loti-induced nodule. (B) S. meliloti-induced nodule. (C) Enod12-GUS expression at different stages of M. loti-induced nodule during development. (D) Enod12-GUS expression in the M. loti-induced nodules (arrows), but not in the lateral roots (arrowheads). (E) Enod2-GUS expression in the M. loti-induced nodules (arrows) and lateral root primordia (arrowheads). (F) Section of M. loti-induced nodule. (G) Section of S. meliloti-induced and -infected nodule (H) Methylene blue staining of cleared M. loti-induced nodule. Arrows indicate peripheral vascular bundles. (I) Methylene blue staining of cleared lateral root. Arrow indicate central root vascular bundles. (J) Cell divisions in the inner cortex initiate nodule primordia (arrow) upon M. loti inoculation. Scale bars=100 μm.

Table 1.

Nodulation frequencies of transformed M. truncatula roots

Construct	Rhizobia	Fraction of nodulated plants	Nodules per total plants
Vector	Uninoculated	0/98	0
	M. loti	0/295	0
	M. loti nodC∷Tn5	0/20	0
	DZLa	0/50	0
	S. meliloti	58/69	5.0
LjNfr1	Uninoculated	0/20	0
	M. loti	0/59	0
	DZLa	0/29	0
	S. meliloti	37/43	6.9
LjNfr5	Uninoculated	0/18	0
	M. loti	0/79	0
	DZLa	0/30	0
	S. meliloti	34/43	4.5
LjNfr1+LjNfr5	Uninoculated	0/111	0
	M. loti	136/499	1.7
	M. loti nodC∷Tn5	0/51	0
	DZLa	22/42	3.4
	S. meliloti	43/56	5.8
aR. leguminosarum bv. viciae 5560 strain DZL.

Nodulin gene expression in M. loti induced M. truncatula nodules

Plant genes known as nodulins are sequentially activated or upregulated after compatible rhizobial inoculation or Nod-factor application on host legume roots, and their expression serves as molecular markers for Nod-factor perception, nodule organogenesis and rhizobial infection (Schultze and Kondorosi, 1998; Stougaard, 2000). To confirm that M. loti-induced nodules on LjNfr1+LjNfr5-transformed M. truncatula roots indeed result from activation of the symbiotic developmental program, the expression of nodulin genes (Enod11, Enod12, Enod2, N6 and Nin) was monitored by quantitative RT–PCR or promoter-GUS activation. Enod11 and Enod12 mark the early symbiotic events in root epidermis and cortex, whereas in nodules their expression was correlated with the presence of infection threads (Journet et al, 2001). N6 expression was associated with plant cell preparation for rhizobial infection (Mathis et al, 1999). Nin regulates infection thread formation and nodule primordia initiation, whereas Enod2 was induced at a later time point upon infection, being expressed in the nodule parenchyma cells during development (Lauridsen et al, 1993; Vijn et al, 1995; Schauser et al, 1999; Marsh et al, 2007). MtEnod11, MtN6 and MtNin were upregulated in nodules induced by M. loti on LjNfr1+LjNfr5-transformed M. truncatula roots (Figure 2). Confirming this, the promoter-GUS reporter analyses showed stage- and tissue-specific activation of Enod12 and Enod2 nodulin gene promoters in M. loti-inoculated LjNfr1+LjNfr5-transformed M. truncatula roots. The Enod12 promoter was induced in root cortical cells where nodule primordia formed and showed a developmental activation pattern typical for indeterminate nodules (Figure 1C and D and Supplementary Figure 2C). The Enod2 promoter was activated in the young nodule primordia and lateral root meristems (Figure 1E and Supplementary Figure 2D). Activation of genes involved specifically in the nodule developmental program and bacterial infection demonstrated that M. loti inoculation induced a symbiotic developmental program in M. truncatula roots carrying LjNfr1+LjNfr5 genes.

Figure 2.

Expression of MtEnod11 (A), MtNin (B) and MtN6 (C) nodulin genes in roots and nodules of transformed M. truncatula. Transcript levels of MtEnod11, MtNin and MtN6 nodulin genes were determined by quantitative PCR. Induction of MtEnod11, MtNin and MtN6 in nodules formed upon S. meliloti inoculation of transformed M. truncatula control roots (left, A. rhizogenes empty vector), and in nodules formed upon M. loti inoculation of LjNfr1+Nfr5-transformed M. truncatula roots (right). Quantification of expression levels shows the induction of these nodulin genes in both M. loti and S. meliloti nodules compared to root tissue. Relative expression was determined as normalized ratios of the three nodulin genes and three housekeeping genes. Error bars represent the corresponding upper and lower 95% confidence intervals.

Ontogeny and anatomy of M. loti-induced M. truncatula nodules

Nodules induced by M. loti on LjNfr1+LjNfr5-transformed M. truncatula roots show morphological and anatomical features similar to S. meliloti-induced nodules, but with no evidence for bacterial endocytosis into the plant cells as seen for the natural symbiont (Figure 1A, B, F and G). Staining of vascular tissues and microscopical analyses of semi-thin sections revealed branched peripheral vascular bundles connecting the nodules with root vasculature (Figure 1H), distinguishing them from lateral roots that have a central vasculature (Figure 1I). The M. loti-induced nodules originated from inner cortical cells (Figure 1J) and developed opposite protoxylem poles (Supplementary Figure 2D). These features are histological hallmarks of indeterminate M. truncatula root nodules induced by S. meliloti (Vasse et al, 1990; Figure 1B and G). Infection threads were observed using confocal microscopy after inoculation of LjNfr1+LjNfr5-transformed M. truncatula roots with M. loti expressing an eGFP marker gene to visualize rhizobia. In contrast to the tubular appearance and root penetration of infection threads (Giovanelli et al, 2006) observed upon S. meliloti inoculation (3.9 infection threads/cm), most of the M. loti-induced infection threads were arrested inside the root hairs or ended in the epidermal cell layer with a swollen sac-like structure often observed when bacterial invasion is unsuccessful (Figure 3A and B). In few cases, some of these infection threads branched and progressed through the root cortical cell layers and reached the nodule primordia (Supplementary Figure 3A). However, no endocytosis of M. loti and therefore no symbiosome formation was observed (data not shown). On average, one M. loti-induced infection thread was counted per 3 cm of LjNfr1+LjNfr5-transformed M. truncatula root and 10% of the analyzed nodules contained branched infection threads. The histological characteristics of nodules, and the presence of infection threads demonstrated that M. loti was recognized as a symbiont by M. truncatula roots carrying LjNfr1+LjNfr5 genes. Hence, expression of LjNfr1 and LjNfr5 genes in M. truncatula confers a Nod-factor-dependent extension of host range.

Figure 3.

Microscopic analyses of Rhizobium infection phenotypes on LjNfr1+LjNfr5-transformed M. truncatula roots. (A) Infection thread induced by M. loti stopped in the epidermal layer (arrow). (B) Infection thread induced by S. meliloti penetrates the root (C) S. meliloti infected nodule showing mainly endocytosed bacteria. (D) DZL induced infection thread penetrating a nodule primordium. (E) DZL-induced infection threads invaded the nodule tissue in a similar manner as S. meliloti. (F) DZL induced infection threads (arrow) penetrate the nodule. Inset: close-up of the nodule tissue with DZL-induced infection threads spreading in-between the cells, ending with a sac-like structure (arrow). (G) Sac-like structure (arrow) formed in the epidermal cell layer at the end of DZL-induced infection thread. (H) DZL bacteria colonizing the nodule intercellular space (arrow) is not endocytosed in the plant cells. (I) S. meliloti endocytosis (arrow) from infection thread into the nodule cell. Scale bars=100 μm in panels A–F, 2 μm in panels G and H, and 1 μm in I. Panels A–E are confocal images showing rhizobia marked with GFP (green), (G–I) are electron micrograph images. Notice bacteria (labeled with GFP) are contained in the infection threads induced by all three Rhizobium strains.

Increased infection thread progression in DZL induced M. truncatula nodules

The LjNFR1- and LjNFR5-mediated nodulation of M. truncatula was further examined using an alternative microsymbiont, R. leguminosarum bv. viciae strain DZL, recognized by L. japonicus (Pacios-Bras et al, 2000, 2003) but not by M. truncatula (Table I). Similar to our observation using M. loti, transfer of LjNfr1+LjNfr5 extended the host range of M. truncatula to include the DZL strain and nodules developed on transgenic roots of 22/42 plants. Transgenic LjNfr1+LjNfr5 M. truncatula roots inoculated with DZL developed on average more nodules (3.4) than those inoculated with M. loti (1.7) (Table I).

Confocal microscopy on LjNfr1+LjNfr5-transformed M. truncatula roots inoculated with eGFP marked DZL bacteria (Figure 3D and E) revealed an invasion process similar to the early phases of S. meliloti invasion of wild-type M. truncatula (Figure 3B and C; Giovanelli et al, 2006). Most infection threads were tubular and arrested in root hairs or at the first epidermal cell layer. However, when infection threads were associated with nodule primordia, they progressed (Figure 3D), branched and invaded the nodule primordia (Figure 3E and F). Compared to M. loti, a larger number of infection threads were formed (1.6/cm), indicating a more effective interaction between DZL and the new host. Nevertheless, confocal, light and electron microscopy revealed that DZL infection threads end inside the nodules in expanded sac-like structures (Figure 3F–H and Supplementary Figure 3B). Most DZL bacteria colonized the nodule intercellular spaces (Figure 3H and Supplementary Figure 3B), and compared to S. meliloti inoculation (Figure 3I and Supplementary Figure 3C), no symbiosomes were observed.

Whereas M. truncatula control roots showed no root hair deformations upon inoculation with M. loti and DZL, the root hairs of LjNfr1-transformed M. truncatula roots showed a minor reorientation of root hair tip growth (Supplementary Figure 4A). The root hairs of LjNfr5-transformed M. truncatula roots curled and entrapped M. loti or DZL bacteria, without infection thread formation (Supplementary Figure 4B). However, the root hair response was not followed by the activation of Enod12 gene promoter in M. loti-inoculated transformed roots (Supplementary Figure 4C) showing that activation of this nodulin gene expression requires the presence of both LjNfr1 and LjNfr5 (Figure 1C and D and Supplementary Figure 2C). Expression of both LjNfr1 and LjNfr5 genes in M. truncatula under the control of L. japonicus promoters therefore ensured M. loti and DZL recognition, infection thread formation and nodule organogenesis, but the resulting nodules arrest in late development before endocytosis and symbiosome formation.

Host range extension in L. filicaulis

In contrast to L. japonicus, the close relative L. filicaulis was not infected by the DZL strain (Pacios-Bras, 2003; Figures 4A and 5A, B, G–I). As a first approach to determine the basis of this difference in DZL strain specificity between L. japonicus and L. filicaulis, we tested whether introduction of LjNfr1 and LjNfr5 genes into transgenic roots would enable L. filicaulis to develop nodules with DZL. At 7 weeks after inoculation, 6.5% of L. filicaulis plants transformed with LjNfr1+LjNfr5 had developed nodules with DZL on the transformed roots (23/349 plants), whereas no nodules were observed on transgenic L. filicaulis roots (243 plants) transformed with empty vector (Figures 4A, B and 5C, F, J). L. filicaulis plants transformed with LjNfr1 or LjNfr5 separately formed nodules on 0.65% (1/154) and 1.7% (3/180) of the transgenic roots, respectively (Figures 4B and 5D, E). These results show that both genes contributed to the changed specificity. Control L. filicaulis roots transformed with LjNfr1+LjNfr5 or empty vector (Figure 4C) were all nodulated by M. loti at comparable, normal efficiencies, showing that transgenic roots of L. filicaulis can be effectively nodulated and that concomitant transfer of LjNfr1 and LjNfr5 receptor genes does not perturb nodulation by the M. loti symbiont compatible with both L. filicaulis and L. japonicus. A reduction of the nodulation frequency of M. loti on L. filicaulis was observed after separate transfer of LjNfr1 or LjNfr5 possibly due to an imbalance between NFR1 and NFR5 receptors (Figure 4C). We conclude that LjNFR1 and LjNFR5 act in concert to confer perception of DZL Nod-factor molecules in L. filicaulis.

Figure 4.

Receptor-mediated nodulation response of L. japonicus and L. filicaulis. (A) Nodulation efficiency of A. rhizogenes transformed wild-type plants inoculated with either DZL or M. loti. (B) Nodulation frequency of DZL inoculated L. filicaulis roots carrying the AR12 empty vector, LjNfr1 or LjNfr5 individually, or LjNfr1+LjNfr5 together. (C) Nodulation frequency of M. loti inoculated L. filicaulis roots carrying Nfr1 or Nfr5 individually, Nfr1+Nfr5 together or the AR12 empty vector. (D, E) Schematic outline of the chimeric receptor genes composed of segments encoding L. filicaulis LysM domains (gray) inserted into L. japonicus Nfr1 and Nfr5 genes. (F) Complementation efficiency in Ljnfr1nfr5 double mutants transformed with either L. japonicus LjNfr1+LjNfr5 wild-type genes or chimeric FinG1+FinG5 genes and inoculated with either DZL or M. loti. (G) Complementation efficiency in Ljnfr1 mutants transformed with either LjNfr1 or chimeric FinG1 gene and inoculated with either DZL or M. loti. (H) Complementation efficiency in Ljnfr5 mutants transformed with either LjNfr5 or chimeric FinG5 gene and inoculated with either DZL or M. loti. The number of plants scored for nodulation is shown in parenthesis. Error bars represent the 95% confidence intervals.

Figure 5.

Nodulation phenotype of L. filicaulis and L. japonicus plants inoculated with M. loti and R. leguminosarum bv. viciae DZL. (A, B) Root segments showing the nodulation phenotype of L. japonicus (A) and L. filicaulis (B) when inoculated with M. loti. (C) Nodulation phenotype of L. filicaulis transformed with LjNfr1+LjNfr5 inoculated with DZL. (D–F) Nodules formed by DZL on L. filicaulis transformed with LjNfr1 (D), LjNfr5 (E) and LjNfr1+LjNfr5 (F). (G–J) Thin sections of L. japonicus (G, I) and L. filicaulis (H) or L. filicaulis transformed with L. japonicus Nfr1+Nfr5 (J) infected by M. loti (G, H) or DZL (I, J). Sections were stained with toluidine blue.

The role of LysM domains in Nod-factor recognition

To investigate whether the LysM domains may bind Nod-factor in a structure-dependent fashion, we took advantage of the ability of closely related L. japonicus and L. filicaulis hosts to recognize M. loti and distinguish DZL. Chimeric genes encoding composite NFR1 and NFR5 proteins were constructed. In these composite receptors, LysM-containing domains of L. japonicus (ecotype Gifu) LjNFR1 and LjNFR5 were exchanged with corresponding domains from L. filicaulis LfNFR1 and LfNFR5, respectively. The chimeric genes were called Filicaulis in Gifu or FinG1 and FinG5, respectively (Figure 4D and E). To diminish possible interference from endogenously expressed NFRs (Figure 4B and C), we exploited the non-nodulating Ljnfr1, Ljnfr5 and Ljnfr1nfr5 double receptor mutants available in L. japonicus for in planta complementation. A. rhizogenes-induced transgenic roots were inoculated with DZL or the normal Lotus symbiont M. loti, which provided a positive test for the function of chimeric constructs. Parallel experiments were performed in which mutants were transformed with LjNfr1 and/or LjNfr5 genes (Figure 4F–H).

Both FinG1+FinG5 and LjNfr1+LjNfr5 constructs complemented double mutants effectively after inoculation with M. loti (Figure 4F), demonstrating that sufficient active proteins were synthesized from both FinG1 and FinG5. Complementation of Ljnfr1nfr5 mutants with FinG1+FinG5 resulted in a low nodulation efficiency with DZL, amounting to 18% of the efficiency obtained with LjNfr1+LjNfr5 genes (Figure 4F). To evaluate the cause of this significant change in specificity, Ljnfr1 and Ljnfr5 single mutants were transformed with FinG1 and FinG5, separately. The complementation efficiency of FinG1 in Ljnfr1 mutants inoculated with DZL was comparable to the efficiency of LjNfr1 (Figure 4G). On the other hand, FinG5 complementation efficiency in Ljnfr5 mutants inoculated with DZL was only 22% of the efficiency obtained with LjNfr5 (Figure 4H). This corresponds to the low efficiency obtained with FinG1+FinG5 in Ljnfr1nfr5 double mutants inoculated with DZL. Both FinG1 and FinG5 complemented as efficiently as LjNfr1 and LjNfr5 genes after inoculation with M. loti (Figure 4F–H). Taken together with results from transfer of L. japonicus LjNfr1 and LjNfr5 genes into L. filicaulis, these results clearly demonstrated that strain and Nod-factor specificity was provided by NFRs and that specificity could be changed by supplying variants of primarily NFR5 or NFR1 plus NFR5. Furthermore, domain swaps between L. japonicus and L. filicaulis proteins point at the LysM-containing domain of NFR5 as a major determinant in recognition.

Hybrid Nod-factor receptors

To examine the role of LysM modules in discrimination of Nod-factor, the amino-acid composition of LfNFR1 and LfNFR5 domains were compared to LjNFR1 and LjNFR5 proteins. Only four and three amino acids differ between the L. japonicus and L. filicaulis domains exchanged in the experiments described above (Figure 4D and E). L. japonicus NFR1 has Ile57, Ile171, Gln192 and Tyr213, whereas L. filicaulis NFR1 has Val, Leu, Arg and Asp, respectively. Ile/Val57 is located in the first, less conserved LysM domain, Ile/Leu171 and Gln/Arg192 are located in the third LysM and Tyr/Asp213 is positioned between the third LysM and the transmembrane domain. L. japonicus NFR5 has Lys30, Leu118 and Glu230, whereas L. filicaulis has Gln, Lys and Lys, respectively. Lys/Gln30 is located between the signal peptide and the first LysM domain, Leu/Lys118 is located in the second LysM domain, whereas Glu/Lys230 is in between the third LysM domain and the transmembrane domain.

For functional analysis, we constructed hybrid receptors combining the amino-acid variation by modification of FinG1 and FinG5. Coding capacities were changed by site-directed mutagenesis such that L. filicaulis amino acids were individually replaced with corresponding amino acids from L. japonicus. Three hybrid genes were obtained for both Nfr1 and Nfr5, and tested in transgenic roots of Ljnfr1 or Ljnfr5 mutant plants, respectively (Figure 6A and E). All six hybrid receptors were functional and complemented Ljnfr1 and Ljnfr5 mutants inoculated with M. loti effectively (Figure 6C and G). As no improvements or detrimental effects were detected, we infer that different combinations of amino-acid variations between L. japonicus and L. filicaulis receptors in these hybrids did not significantly perturb domain structures or diminish/enhance NFR1 and NFR5 activity. Substitution of Gln30 with Lys, and Lys230 with Glu, both amino acids located outside of NFR5 LysM domains, did not appreciably alter DZL complementation efficiency compared to plants transformed with FinG5 (Figure 6A and B). The third hybrid, where a basic hydrophilic Lys118 in the second LysM domain was replaced by a hydrophobic leucine residue, completely restored complementation efficiency with DZL to LjNfr5 levels (Figure 6A and B). The three NFR1 L. filicaulis to L. japonicus Val57 to Ile, Leu171 to Ile and Arg192 to Gln amino-acid replacements were tested without detecting any change in nodulation phenotype with DZL or M. loti (Figure 6E–G). We conclude that a Leucine/Lysine difference in amino-acid residue 118 of the second LysM module of NFR5 was largely responsible for the differential response of L. japonicus and L. filicaulis towards Nod-factor produced by DZL.

Figure 6.

Nodulation of Lotus japonicus Ljnfr1 and Ljnfr5 mutants transformed with hybrid receptor genes. (A, E) Schematic outline showing domains exchanged between L. japonicus (white) and L. filicaulis (gray) in domain swap constructs. Positions of amino-acid differences between L. japonicus and L. filicaulis that were analyzed in hybrid receptor constructs are indicated relative to LysM domains. The less conserved LysM1 of LjNfr1 is shaded light gray. (B–D) Complementation efficiency in Ljnfr5 mutants transformed with either LjNfr5 L. japonicus wild-type receptor gene, chimeric FinG5 gene, hybrid receptors where individual L. filicaulis amino acids were changed to the corresponding L. japonicus amino acid or the LjNfr5Leu118Ala gene. (F, G) Complementation efficiency in Ljnfr1 mutants transformed with either LjNfr1, chimeric FinG1 gene or hybrid receptors where individual L. filicaulis amino acids were changed to corresponding L. japonicus amino acid. Roots were inoculated with either DZL (B, F) or M. loti (C, G). The number of plants scored for nodulation is shown in parenthesis.

The importance of Leu118 for recognition of the DZL Nod-factor was further examined by changing the coding capacity of LjNfr5 by site-directed mutagenesis. Leu118 was replaced by an alanine residue and the complementation efficiency of this gene construct (LjNfr5Leu118Ala) was compared to the efficiency of the wild-type gene construct in an Ljnfr5 mutant. When inoculated with M. loti there was no detectable difference in complementation efficiencies between the LjNfr5Leu118Ala gene and the wild-type LjNfr5 gene showing that Ala-substituted NFR5 was active. In contrast, a significantly reduced complementation efficiency of less than 15% was obtained with DZL (Figure 6D). This confirms the results obtained using FinG5 hybrids and provides independent evidence for the role of Leu118 in determining DZL Nod-factor specificity.

Homology model of the NFR5 LysM2 domain

A LysM domain structure from Escherichia coli membrane-bound MltD was solved by NMR (Bateman and Bycroft, 2000). We used this structural information to make a homology model of LjNFR5 LysM2 (Figure 7A and Supplementary data) and identified a possible Nod-factor-binding cleft containing several aromatic residues (Y114, F124, Y130 and F159) and a conserved Asp (D120). The variable L/K118 is positioned at the entry/exit of the suggested binding cleft and would therefore be in a favorable position for interacting with substituents at the reducing or nonreducing end of Nod-factor known to determine in vivo specificity (D'Haeze and Holsters, 2002). The engineered DZL strain decorates pentameric Nod-factors with an acetylated fucosyl at C6 of the reducing sugar moiety (R6), an acetyl at C6 (R5) and an 18:1 or 18:4 acyl chain at C2 (R1) of the nonreducing sugar (Pacios-Bras, 2003; Figure 7B). M. loti synthesizes pentameric Nod-factors with an acetylated fucosyl at C6 (R6) of the reducing sugar, a hydrogen at C6 (R5), a carbamoyl at C4 (R4) and a N-methylated 18:1 acyl chain at C2 (R1 and R2) of the nonreducing sugar (López-Lara et al, 1995; Figure 7C). The structural differences between the M. loti and DZL Nod-factors are thus substitutions at C2, C4 and C6 positions of the nonreducing moiety. Taken together with our model, this predicts an orientation of Nod-factor with the nonreducing end at the L/K118 entry/exit of the LjNFR5 LysM2 groove. We propose that the differences in these decorations at the nonreducing end influence affinity of the M. loti and DZL Nod-factors at the NFR5 LysM2 domain and that this affinity is also influenced by the amino acid present in position 118 of the LysM2. Supporting this notion, a change in strain specificity determined by the saturation of the acyl chain at C2 and the acetate at C6 of the nonreducing end has previously been reported (Spaink et al, 1991; Ardourel et al, 1994). It is now important to resolve the role of the binding site suggested here and the Nod-factor docking site suggested for the M. truncatula NFP LysM domain (Arrighi et al, 2006).

Figure 7.

Model of the L. japonicus NFR5 LysM2 domain and structure of Nod-factors produced by R. leguminosarum bv. viciae DZL and M. loti. (A) Surface representation of the model colored according to residue type: gray is hydrophobic, yellow is polar, red is acidic and blue is basic. The proposed Nod-factor binding site is highlighted and residues along this cleft with a plausible role in binding are labeled along with the important L118. (B, C) Differences in substitution of Nod-factor secreted by R. leguminosarum bv. viciae DZL (B) and M. loti (C). Numbering of Nod factor substitutions according to D'Haeze and Holsters (2002).

Discussion

The results presented show that LjNFR1 and LjNFR5 receptors act in concert as host determinants, transforming the non-hosts M. truncatula and L. filicaulis into hosts able to recognize and be infected by the L. japonicus symbionts M. loti or DZL and DZL, respectively. Recognition of these normally noncompatible bacteria triggered root cell dedifferentiation, redifferentiation and initiation of the developmental program leading to de novo nodule organogenesis. This extended NFR-mediated signal cascade depended on both Nod-factor presence and structure as shown with the M. loti nodC mutant in M. truncatula and DZL in L. japonicus mutants complemented with chimeric and hybrid receptors. We have further demonstrated the crucial role of the putative extracellular LysM-containing domains of the NFR1 and NFR5 receptor kinases that were responsible for perceiving the rhizobial signal in the epidermal root hairs (Radutoiu et al, 2003; Miwa et al, 2006). Specific recognition of rhizobial bacteria by NFR1 and NFR5 receptors relied on these LysM domains and the LysM2 domain of NFR5 had a major function in discrimination of M. loti and DZL Nod-factors. A single amino acid, Leu118 of the NFR5 LysM2 domain, was found to be largely responsible for the specificity. Interestingly, LysM2 of NFR5 is the most diverged LysM domain among NFR1 and NFR5 homologs found in other plant species (Madsen et al, 2003) and during evolution of plant symbiosis, this flexibility may have allowed specificity in Nod-factor perception to emerge. The functional role of the other LysM domains present in NFR1 and NFR5 was not revealed by our studies. The presence of three LysM domains in the LjNFR5, MtNFP and two conserved and one more variable LysM in the LjNFR1, MtLYK3, MtLYK4 receptors characterized so far (Limpens et al, 2003; Madsen et al, 2003; Radutoiu et al, 2003; Arrighi et al, 2006), suggests a mechanism involving more than one LysM domain in Nod-factor perception. The non-nodulation phenotype caused by amino-acid substitutions in the LysM1 domains of the MtNFP and MtLYK3 proteins supports this notion (Arrighi et al, 2006; Smit et al, 2007). Furthermore, we report here the requirement for both the LjNfr1 and LjNfr5 genes for nodulation of L. filicaulis by the DZL strain and for changing the non-host M. truncatula into a host for the Lotus microsymbionts M. loti and DZL.

Recognition of the normally noncompatible bacteria M. loti and DZL in M. truncatula induced infection thread formation, but endocytosis was not observed. Compared to M. loti, the engineered R. leguminosarum bv. viciae, strain DZL, induced an increased number of nodules and an infection process that progressed further through the cortical cells and invaded the nodule space of LjNfr1+LjNfr5-transformed M. truncatula roots. This difference may be due to the particular structural similarity of the DZL Nod-factor to the S. meliloti Nod-factor that differs only in the type of acyl chain and the acetyl-fucose decoration (Supplementary Figure 1). After the epidermal recognition ensured by L. japonicus receptors, perception of the DZL Nod-factor may function more effectively with endogenous M. truncatula signaling components than the M. loti Nod-factor. In M. truncatula, 17 genes encoding LysM receptor-like kinases similar to the LjNFR5 or LjNFR1 receptor proteins have been identified, 8 of them being expressed in roots and nodules (Arrighi et al, 2006) and the possibility that one or more of these putative receptors could promote infection thread progression and cortical invasion process of DZL, cannot be excluded. The MtLYK3 protein suggested to interact with the nonreducing end of S. meliloti Nod-factor and to promote infection thread progression in M. truncatula could be such an example (Smit et al, 2007). In addition to Nod-factor, specific surface carbohydrates are required for rhizobia to become efficient symbionts (Pellock et al, 2000; Laus et al, 2004; Mathis et al, 2005) and this may account for differences observed between M. loti and DZL. Like the DZL strain on transgenic M. truncatula, several lipopolysaccharide-deficient (lps) mutants of S. melitoti induce infection threads penetrating the core nodule tissue and a Fix⁻ phenotype (Campbell et al, 2003), but in contrast to the DZL strain, the lps mutants are usually endocytosed (Niehaus et al, 1998). The early arrest of infection thread growth observed in M. loti-inoculated transgenic M. truncatula resembles the phenotype reported for exopolysaccharide-deficient (exo−) S. meliloti mutants in wild-type alfalfa (Cheng and Walker, 1998). On the plant side, lectins were shown to be involved in host range presumably by recognizing bacterial exopolysaccharides (Diaz et al, 1989; van Rhijn et al, 2001). Transfer of the soybean Le1 gene into L. corniculatus resulted in a Bradyrhizobium japonicum-induced nodulation phenotype reminiscent of LjNfr1+LjNfr5-transformed M. truncatula roots inoculated with M. loti (van Rhijn et al, 1998). Our results are consistent with the notion that increased binding of B. japonicum bacteria would bring the local concentration of B. japonicum Nod-factor, which is similar but not identical to the M. loti Nod-factor, to pass the threshold required for perception by L. corniculatus (van Rhijn et al, 1998).

Our results provide circumstantial evidence that NFR1 and NFR5 are components of the same receptor complex or alternatively that balanced signaling through two independent NFR1 and NFR5 receptors is crucial. First, inefficiency of a receptor complex composed of NFRs from the two Lotus species in perception of DZL Nod-factor may explain the difference in nodulation efficiency of DZL and M. loti in LjNfr1+LjNfr5-transformed L. filicaulis (Figure 4B and C). Secondly, an imbalance between NFR1 and NFR5 may cause the reduced nodulation efficiency observed in M. loti-inoculated L. filicaulis plants transformed with LjNfr1 and LjNfr5, separately (Figure 4C). The observed reduced complementation efficiency of Ljnfr1 mutants transformed with LjNfr1 (Figure 4G) further implies that Nod-factor perception is particularly sensitive towards NFR1 expression levels or NFR1-mediated signaling. Our results demonstrate a higher nodulation frequency with the DZL strain in M. truncatula transformed with LjNfr1 and LjNfr5 than observed in similarly transformed L. filicaulis plants. This difference may suggest a higher level of (ineffective) complex formation between endogenous L. filicaulis receptors and LjNFR1 and LjNFR5 receptors expressed from transgenes in L. filicaulis than in the more distantly related M. truncatula. Although complex formation between NFR1 and NFR5 components would explain these results, a final verification of the Nod-factor receptor composition and its interacting partners await a detailed biochemical analysis. At the amino-acid sequence level there is certainly sufficient variation to account for differences in protein–protein interactions. NFR5 proteins of L. japonicus and L. filicaulis are 99% identical (591 out of 595 amino acids) compared to the 76% amino-acid identities between the corresponding L. japonicus NFR5 and M. truncatula NFP proteins (Madsen et al, 2003).

Adding further complexity, involvement of a third component is likely to account for the incomplete change in DZL specificity observed in the L. japonicus genetic background when complementing Ljnfr1nfr5 double mutants with FinG1+FinG5. The 18% nodulation efficiency in L. japonicus is significantly different from the absence of DZL nodulation seen in the L. filicaulis background. At present, the identity and function of this third component differing between L. filicaulis and L. japonicus is undescribed, but both a direct and an indirect involvement in Nod-factor recognition appears possible. A Nod-factor binding protein presenting the Nod-factor ligand at the receptor binding sites would serve a direct function. The high-affinity Nod-factor-binding sites identified in Medicago may represent such a function (Gressent et al, 1999). A more effective signal transduction component enabling nodulation of L. japonicus at a Nod-factor signaling level that was below the threshold in L. filicaulis would be an example of an indirect function. Similar differences in Nod-factor perception between closely related legume host plants have previously been observed between M. truncatula and alfalfa (Medicago sativa). These two Medicago species differ in their sensitivity towards unsulfated Nod-factor produced by S. meliloti nodH mutants and secondary components or threshold values were suggested to account for the observed difference (Wais et al, 2002).

The in planta assays used in this study measure NFR-mediated Nod-factor perception in its biological context and in addition to the factors mentioned above the results may also be influenced by the variable amounts of Nod-factor secreted by rhizobia and Nod-factor degradation by host chitinases or hydrolases (Ovtsyna et al, 2000). To promote a better understanding of the quantitative and structural requirement for Nod-factor function and ligand–binding site interactions we have modeled the NFR5 LysM2 structure. Based on the E. coli MltD LysM structure (Bateman and Bycroft, 2000), modeling of the NFR5 LysM2 domain predicts a cleft for Nod-factor binding and positions the L/K118 amino acid participating in the in vivo recognition of Nod-factor structure at the entry or exit of the groove. A similar homology model of NFR5 LysM2 based on the crystal structure of the B. subtilis ykuD protein supports our binding site prediction (Supplementary Figure 5). The most likely interpretation is a structure-dependent binding of Nod-factor at the LysM2 domain and this is in accordance with studies in both bacteria and plants. LysM domains from bacterial autolysin bind peptidoglycans and more recently the high-affinity chitin-binding protein CEBiP from rice, involved in perception of chitin oligosaccharide elicitors, was shown to carry two LysM domains (Kaku et al, 2006). Further characterization of these putative binding sites is a challenge for the future and in vitro biochemical investigations of the binding affinities and kinetics as well as high-resolution structures of the LysM domains will be required for a detailed understanding of Nod-factor receptor function during nodule organogenesis.

Materials and methods

Plant material, hairy root transformation and nodulation tests

The Ljnfr1 and Ljnfr5 loci are in L. japonicus Gifu ecotype background (Madsen et al, 2003; Radutoiu et al, 2003). Homozygosity for mutant alleles was confirmed by sequencing of PCR fragments spanning the point of mutation. Transgenic hairy roots on Lotus and Medicago truncatula plants (cultivar J5), were generated using the Agrobacterium rhizogenes strain AR12 or AR1193 (Stougaard et al, 1987; Hansen et al, 1989). For nodulation tests in Magenta containers (Sigma), seedlings with emerging hairy roots were planted in a 3:1 mixture of Leca (Optiroc):Vermiculite and supplemented with 70–80 ml of 1/4 strength B&D medium (Radutoiu et al, 2003). Nodulation and gene expression were assayed after 40–42 days of growth at 21/16°C and 16/8 h day/night regime. Medicago experiments: four composite plants were transferred to one Magenta box supplemented with 1 or 0.1 mM KNO₃; each plant was inoculated with 700 μl of Rhizobium inoculum, OD₆₀₀=0.01–0.02. Lotus experiments: after planting in Magenta boxes supplemented with 1 mM KNO₃, 2.8 ml inoculum (OD₆₀₀∼0.01–0.02) of NZP2235 or DZL was added per Magenta containing seven seedlings; to verify the presence and identity of DZL bacteria in nodules of L. filicaulis transformed with L. japonicus Nfr1+Nfr5, nodules were surface sterilized in hypochlorite (0.6–1.4% active chlorine) for 10 min, washed three times for 10 min in sterile water, crushed in sterile water and plated on YMB medium supplemented with appropriate antibiotics. An aliquot of the 3rd wash was plated as a sterilization control.

Rhizobial strains

M. loti NZP2235 and S. meliloti 2011 or 1021 (Supplementary Table II) were grown to high density in liquid YMB supplemented with tetracycline 2 μg/ml for S. meliloti at 28°C with shaking for 2 days before use. R. leguminosarum DZL was grown at 28°C on solid YMB supplemented with rifampicine 20 μg/ml, spectinomycine 100 μg/ml, tetracycline 2 μg/ml for 2 days and then transferred at high density to liquid YMB without selection and grown for 3–4 h at 28°C with shaking before use.

Microscopic analyses

Enod2-GUS and Enod12-GUS activity in transgenic roots was assayed according to Kosugi et al (1990). Whole-root and -nodule analyses were made by staining with 0.01% aqueous solution of methylene blue, followed by clearing with 2.4% sodium hypochlorite for 3 min and visualization under stereomicroscope. Sectioning of roots and nodules was carried out as described previously (van Spronsen et al, 2001). For infection thread screening, fresh, 1 cm long segments of transformed roots were observed using a Zeiss fluorescence microscope with FITC filters. Hand-cut sections of nodules were optionally stained with propidium iodide (0.1% (v/v), Fluka) in PEM buffer (PIPES 0.05 M, EGTA 1 mM, MgCl₂ 0.5 mM, pH 6.9) with 0.01% (v/v) Triton X-100, for 30 min at room temperature. Images were collected with the Zeiss LSM 510 Meta confocal microscope (488 nm excitation, 505–530 bandpass filter for GFP emission and 585–620 nm bandpass filter for propidium iodide emission). The number of root segments analyzed for presence of infection threads is presented in Supplementary Table IV. Electron microscopy was according to James and Sprent (1999).

Expression analyses

Total RNA was isolated using Trizol (Sigma-Aldrich) and RNA was treated with RQ1-DNAse (Promega). Transcript levels were determined by quantitative real-time RT–PCR (Radutoiu et al, 2003). All cDNA samples were tested for contaminating DNA using PCR primers specific for the LYK3 gene promoter. Primers for transcript amplification are listed in Supplementary Table V. For each sample, normalized relative ratios of nodulin genes and three independent housekeeping genes (Protein phosphatase2A, TIP41 and Tubulin beta-chain; Czechowski et al, 2005) were calculated using the Relative Quantification Software (Roche). The geometric mean of relative expression ratios for three biological, and three technical repetitions and corresponding upper and lower 95% confidence intervals were calculated (Vandesompele et al, 2002).

Construction of chimeric Nfr receptor genes

The FinG1 and FinG5 genes were based on the Nfr1 and Nfr5 complementation construct (Madsen et al, 2003; Radutoiu et al, 2003), which were modified using standard cloning techniques and transferred into the pIV10 integration vector (AM235368). The constructs were integrated into A. rhizogenes strain AR12 (Hansen et al, 1989). FinG1 was constructed as follows: a DNA fragment from position 4090 to position 4993 of the Nfr1 gene (AJ575246/AJ575247) was substituted by the corresponding fragment from L. filicaulis. FinG5 was constructed as follows: a DNA fragment from position 1267 to position 2102 of Nfr5 (AJ575254) was substituted by the corresponding fragment from L. filicaulis produced by PCR.

Construction of hybrid Nfr receptor genes

The GeneEditor™ in vitro Site-Directed Mutagenesis System (Promega) or a PCR-based method (Ke and Madison, 1997) were used for site-directed mutagenesis using FinG1 and FinG5 genes as starting constructs. Polymorphic amino acids between L. filicaulis and L. japonicus were replaced individually by changing the coding nucleotide triplets. All constructs and the transformed roots were checked by sequencing for the presence of the characteristic codon.

Supplementary Material

Supplementary Figure 1

Supplementary Figure 2

Supplementary Figure 3

Supplementary Figure 4

Supplementary Figure 5

Supplementary Tables

Supplementary Methods