|Home | About | Journals | Submit | Contact Us | Français|
Eukaryotes often form symbioses with microorganisms. Among these, associations between plants and nitrogen-fixing bacteria are responsible for the nitrogen input into various ecological niches. Plants of many different families have evolved the capacity to develop root or stem nodules with diverse genera of soil bacteria. Of these, symbioses between legumes and rhizobia (Azorhizobium, Bradyrhizobium, Mesorhizobium, and Rhizobium) are the most important from an agricultural perspective. Nitrogen-fixing nodules arise when symbiotic rhizobia penetrate their hosts in a strictly controlled and coordinated manner. Molecular codes are exchanged between the symbionts in the rhizosphere to select compatible rhizobia from pathogens. Entry into the plant is restricted to bacteria that have the “keys” to a succession of legume “doors”. Some symbionts intimately associate with many different partners (and are thus promiscuous), while others are more selective and have a narrow host range. For historical reasons, narrow host range has been more intensively investigated than promiscuity. In our view, this has given a false impression of specificity in legume-Rhizobium associations. Rather, we suggest that restricted host ranges are limited to specific niches and represent specialization of widespread and more ancestral promiscuous symbioses. Here we analyze the molecular mechanisms governing symbiotic promiscuity in rhizobia and show that it is controlled by a number of molecular keys.
Studies of nitrogen-fixing symbioses began in Europe, largely on Northern hemisphere plants. Thus, in 1542, the German physician and botanist Leonhard Fuchsius (88) published drawings of nodulated legumes. During the 17th century, Malpighi (167), who worked mostly in Bologna, Italy, observed nodules on the roots of beans (Phaseolus vulgaris and Vicia faba, members of the Leguminosae). Almost 200 years later, the Russian botanist Woronin (280) noted that the nodules of Alnus glutinosa (Betulaceae) and Lupinus mutabilis (Leguminosae) were filled with minute bodies resembling bacteria. Although the observations that both legumes and nonlegumes possess nodules were historically important, the origin of nodules was controversial (86). Frank (85) found nodules on the roots of all healthy legumes in a study in Germany and demonstrated that incinerating soil prevented the nodulation of Pisum sativum. Hellriegel (120), and Hellriegel and Wilfarth (121) showed, in a study in Germany, that nodule formation results from external infection of Lupinus spp., P. vulgaris, P. sativum, Ornithopus sativa, Trifolium spp., and Vicia sativa. However, it was Beyerinck (13–17), working in The Netherlands, who furnished the first proof that bacteria provoke nodules; he did this by preparing pure cultures of nodule organisms from V. faba and using them to infect Faba beans growing in sterile soil (18). Prazmowski (204, 205), working in Poland, inoculated P. sativum with pure cultures and showed that the bacteria penetrate legumes via infection threads in root hairs. In fact, it is now clear that many diverse soil bacteria harbor symbiotic loci (28), including the genera Azorhizobium, Bradyrhizobium, Mesorhizobium, and Rhizobium. Collectively, they are called rhizobia.
Root hairs, which develop near the root apex, are unicellular extensions of root epidermal cells. Most root hairs emerge at the apical end of particular epidermal cells and elongate by polar growth of the tip (43). The tips of root hairs have high exocytotic and cell wall assembly activities (221). In the soil, they are extensively colonized by soil-borne microorganisms, including nitrogen-fixing bacteria. In the rhizosphere, young, growing root hairs play an important role in symbiotic recognition. Rhizobia and the Nod factors they secrete (for reviews, see references 42, 63, 115, 173, 240, and 256) stimulate the reorientation of root hair cell wall growth (47, 139), resulting in curled root hairs. Within these curled root hairs, Nod factors promote the formation of infection threads, and it is through these tubular structures that the bacteria enter most (but not all) plants (Fig. (Fig.11 and and2).2).
Work on the molecular basis of host specificity also began at the end of the last century. In a study in Germany, Hiltner (123) prepared aqueous, bacterium-free filtrates from mature P. sativum nodules and demonstrated that they contain a substance that induces root hair formation (Hai) and deformation of the root hairs (Had) in this plant. Many attempts to define the structure of these deformation factors followed (213), but these investigations yielded fruit only in 1990, when Lerouge et al. (154), working in France, showed that the substances responsible are N-acylated oligomers of N-acetyl-d-glucosamine. Since then, the Nod factor structures (the products of nod genes) of a number of rhizobia have been elucidated (see “Baroque decorations to the Nod factor core” below).
In other words, research on symbiotic nitrogen fixation was concentrated on European legumes, and although this has become less so with time, the focus on genera such as Medicago, Pisum, Trifolium, and Vicia has prevailed. Of course, model plants have advantages, but it is our contention that this bias toward European plants for historic reasons may have given a false impression of symbiotic specificity. What follows is an analysis of the molecular mechanisms governing broad host range (i.e., promiscuity) in rhizobia that may be regarded as a paradigm for other microbial symbioses. Readers are referred to the reviews by Irving et al. (128) and Downie and Walker (64) for more information on the role of the plant in nodulation.
Doubts that bacteria were faithful to the legume from which they were isolated also surfaced early. In experiments that are probably unrepeatable (Table (Table1),1), Bréal (25) induced nodule formation on Lupinus, Medicago sativa, and P. sativum by bacteria isolated from nodules of alfalfa (M. sativa). In similar, questionable experiments, Laurent (149) suggested that nodules could be produced on P. sativum by bacteria isolated from 36 species of plants. Perhaps the first reproducible data of this kind was obtained by Nobbe et al. (183, 184), who showed in experiments performed in Germany at the end of the 19th century that bacteria from P. sativum nodules were unable to nodulate plants of the tribes Genisteae and Hedysareae. Indeed, rhizobia which nodulate Lupinus, Medicago, Pisum, Vicia, etc., are now known to have restricted host ranges (Table (Table1).1).
The U.S. researcher J. K. Wilson was one of the first non-Europeans to publish extensive information on different symbiotic associations (279). He tested the host range properties of rhizobia isolated from 31 genera of legumes on 160 different legume species. All isolates, including those from Vicia spp., nodulated legumes of different tribes. The average number of plant species nodulated by a particular strain was 33%, but it ranged from 6% (for an isolate from Spartium scoparium) to 66% (for a slow-growing, and therefore probably Bradyrhizobium, strain isolated from Mucuna nodules). Indeed, four of Wilson's isolates nodulated more than half the legumes tested. These data are included in Table Table22 to show the patterns of nodulation provoked by broad-host-range rhizobia. Both Bradyrhizobium and Rhizobium species can be promiscuous, as shown in Table Table2,2, and broad-host-range rhizobia were even isolated from Albizia julibrissin (Mimosoideae). Isolates from P. vulgaris, which include R. etli and R. tropici, were shown to nodulate 17 and 18 different legume genera, respectively (122). It thus seems as if rhizobia associated with most tribes of the Leguminosae have various degrees of promiscuity. The principal exceptions to this rule are the more specialized rhizobia associated with the tribes Cicereae, Trifolieae, and Viceae, which have restricted host ranges (28).
As expected, hosts also vary greatly in their ability to enter into symbiosis with different rhizobia. In Wilson's study (279), six plants formed nodules with at least 90% of the isolates (Table (Table2).2). At the other extreme, Hippocrepis sp. and Ornithopus sativa formed nodules with only one other isolate. In further studies of this kind, numerous legumes have been inoculated with large collections of rhizobia and the capacity of the legumes to nodulate have been evaluated. Again, promiscuity exists among diverse legumes (Table (Table3).3). Mimosoid legumes not only harbor broad-host-range rhizobia but also are capable of interacting with diverse bacteria; Leucaena leucocephala, for example, is nodulated by a moderately broad spectrum of symbionts. The fact that Azorhizobium caulinodans, a microsymbiont of Sesbania rostrata, has a very restricted host range (Table (Table1)1) gives the impression that Sesbania spp. have very defined rhizobial requirements. However, Wilson (279) showed that S. drummondii, along with another tree (Robinia pseudoacacia), had the broadest capacity to nodulate among the legumes he tested.
However, most broad-spectrum hosts known are annual and climbing legumes of the tribe Phaseoleae. These include Macroptilium atropurpureum, which has been widely used as a test plant for nodulation by uncharacterized rhizobia (27, 248); Lablab purpureus, the legume from which Rhizobium sp. strain NGR234 was isolated (269); and P. vulgaris (175). Equally, the other plants listed in Table Table33 are considered to have broad host ranges (185). The capacity of Vigna unguiculata to nodulate with many different rhizobia has been exploited in the search for nod genes (156, 157). Taken together, these data suggest that symbiotic promiscuity is widely dispersed in nature. It is not associated with a particular bacterial or plant taxonomic group and is not correlated with the growth habit of the legume. Broad-host-range legumes are also almost equally distributed between those that form indeterminate as opposed to determinate nodules (Table (Table3).3). Perhaps the only underlying similarity is that promiscuous bacteria and plants are found mostly in warm or tropical parts of the world.
Nodulation leads to the colonization of plant cells by invading bacteria. Although many host plants and effective rhizobia have the ability to enter into symbiosis with more than one partner, only certain combinations of symbionts result in the formation of nitrogen-fixing nodules. Ineffective associations lead to empty or nonfixing bacteroid-containing nodules. Specificity among compatible partners minimizes the chances of infection by pathogens and the formation of ineffective associations that are detrimental to both symbionts. Experimental evidence suggests that the progression of invasive rhizobia towards nodule primordia is challenged at various “doors” (Fig. (Fig.1).1). Codes contained in molecular signals open these checkpoints. During the initial phases of nodulation (root hair curling and bacterial entry), these codes are given by flavonoids and Nod factors. In both cases, NodD proteins are the chief interlocutors of molecular traffic in the rhizosphere.
Plants jettison surprisingly large amounts of organic matter into the soil, most of which supports the growth of rhizospheric microorganisms. By labelling whole plants with 14CO2, Helal and Sauerbeck (119) showed that 19% of the photosynthate was released as organic material into the rhizosphere. These compounds include carbohydrates, organic acids, vitamins, amino acids, and phenolic derivatives. Of these, flavonoids (2-phenyl-1,4-benzopyrone derivatives) are the most important from the symbiotic perspective. Although found throughout the plant kingdom, flavonoids specifically trigger the expression of the rhizobial genes required for nodulation (nod, nol, and noe). The inducing capacity varies with flavonoids and rhizobia; and in some cases flavonoids may inhibit induction (61, 79, 196, 197, 199). Other molecules, such as the betaines and erythronic and tetronic acids, may also act as inducers (90, 198).
Regulation of nod gene expression in rhizobia varies from strain to strain but is almost always mediated by NodD (for a review, see reference 237). NodD proteins belong to a family of LysR-like transcriptional regulators that bind to highly conserved 47-bp DNA motifs (nod boxes) found in the promoter regions of many nodulation loci (81, 227). Although NodD proteins bind to nod boxes even in the absence of an inducer, flavonoids are generally required for the expression of nod genes (81, 82, 101, 110). Thus, NodD acts both as a sensor of the plant signal and as an activator of transcription of nod loci (236). In R. leguminosarum bv. viciae, NodD proteins are localized in the cytoplasmic membrane (234), where the inducing flavonoid, naringenin, also accumulates (211). Direct binding of inducers to NodD has not been demonstrated, but point mutations in nodD affect the recognition of inducing molecules and cause an extension of the host range (30, 174). Furthermore, comparison of the NodD structure with various nuclear receptors has shown that the two types of proteins share conserved ligand-interacting domains located at the C-terminal end of their respective DNA-binding motifs (111).
Although nodD genes are ubiquitous in rhizobia, their symbiotic characteristics vary from one species to another. Some strains, such as R. leguminosarum bv. trifolii, have only one nodD gene, and in these cases, its mutation renders the strain incapable of nodulation (Nod−). In contrast, B. japonicum, Rhizobium sp. strain NGR234, R. meliloti, and R. tropici possess two to five copies of nodD (71, 105, 194, 269c). In R. meliloti, mutation of all three copies of nodD is required to abolish nodulation (124), whereas inactivation of nodD1 is sufficient to render strain NGR234 Nod− (213). nodD products of various Rhizobium species vary in that they respond to different sets of flavonoids (72, 249). Moreover, NodD homologues from the same strain may have different flavonoid preferences (112, 117, 198). In R. meliloti, NodD1 is activated when cells are supplied with a complex plant seed extract or the flavonoid luteolin; NodD2 is activated only with the complex extract, not with purified luteolin; and NodD3 apparently modulates the expression of nod genes even in the absence of any plant factor (180). Together with syrM, nodD3 of R. meliloti constitutes a self-amplifying positive regulatory circuit that is involved in the regulation of nod genes within the developing nodule (266). In contrast, NodD1 of strain NGR234 responds to a wide range of inducing molecules that include flavonoids known to be inhibitors in other rhizobia (e.g., vanillin and isovanillin) (4, 72, 155) and several estrogenic compounds (112). Transfer of the nodD1 gene of strain NGR234 to restricted-host-range rhizobia extends the nodulation capacity of the recipients to new hosts, including the nonlegume Parasponia andersonii (4, 11, 126). Although a number of correlations exist between the spectrum of flavonoids able to interact with NodD proteins and the breadth of the host range (11, 54, 126, 249, 250, 263, 269d), variations in the ability of these proteins to differentially sense inducing molecules (and regulate the expression of nod genes) are insufficient to explain the phenomenon of host specificity.
Nonetheless, NodD genes represent a molecular interface between the bacterium and the plant. However, other species-specific sensor-activator systems also contribute to the control of bacterial host range. For instance, nodV and nodW of B. japonicum are essential for the nodulation of M. atropurpureum, Vigna radiata, and V. unguiculata but contribute only marginally to the symbiosis with G. max (104). Modulation of two-component regulatory systems such as nodV and nodW is mediated by a series of phosphorylation steps. NodV is thought to be a membrane-bound protein that senses signals from the plant and transduces the signal to NodW, which in turn activates the expression of nod genes (230). Both in vitro and in vivo studies have confirmed that phosphorylation of NodW is induced by genistein and depends on both acetyl phosphate and its cognate kinase, NodV (159). Also, comparison of the biological activities of the wild-type and mutant proteins indicates that phosphorylation of NodW is essential for nod gene activation (159). A search for genes whose transcription is dependent on NodW led to the identification of two suppressor genes, nwsA and nwsB. Together, these genes also form a dual-component regulatory system (109). When overexpressed, nwsB can complement a nodW mutant as well as a nodD1 nodD2 double mutant, suggesting that in B. japonicum at least, alternatives to NodD- and flavonoid-sensing regulatory systems coexist (105).
nod gene expression is also subject to negative control. After initial induction by flavonoids, repression of several nodulation genes is required for optimal nodulation of M. sativa by R. meliloti (142) and of P. sativum by R. leguminosarum bv. viciae strain Tom (141). In R. meliloti, repression of several nod genes is controlled by NolR, a 13-kDa product that contains a helix-turn-helix (HTH) motif which is homologous to other regulators of the LysR family such as NodD and SyrM (143). In the dimeric form, NolR binds to conserved (A/T)TTAGN9A(T/A) target sequences found in the promoter regions it controls (46) and thus represses the expression of the nodD genes as well as those necessary for the synthesis of the core Nod factor structures (nodABC). However, it does not affect host-specific nod loci such as nodH. Interestingly, the absence of NolR repressor activity in R. meliloti stain 1021 is due to a single insertional mutation in the C-terminal coding sequence, which abolishes the DNA binding ability of the protein without affecting its HTH motif (45).
With a few exceptions, transcription of most inducible nod loci is repressed before rhizobia differentiate into functional bacteroids (235, 244), although transcription of nodD and nodE of R. leguminosarum bv. viciae in P. sativum nodules and nodA of A. caulinodans in stem nodules of S. rostrata still occurs (56, 235). R. meliloti is no exception, although NolR does not play a direct role in the down regulation of flavonoid-inducible nodulation genes in bacteroids (46). In B. japonicum, repression of nod gene expression by NolA is probably an indirect effect, mediated perhaps by nodD2 (91). B. japonicum nolA and R. meliloti nolR mutants retain the ability to nodulate their respective hosts, albeit at lower efficiency, suggesting that fine-tuning of nod gene expression is required for optimal nodulation. Similarly, a nodD2 mutant of strain NGR234 fails to repress the expression of the nodABCIJnolOnoeI operon after initial flavonoid induction and, in contrast to the wild-type strain, is unable to form nitrogen-fixing nodules on V. unguiculata and Cajanus cajan (73).
Sequence analysis showed that pNGR234a carries 19 sequences homologous to conserved NodD-dependent promoter elements (nod boxes), 5 of which control the expression of known nod operons (87, 195). No significant difference was found between the consensus sequence derived for nod boxes of pNGR234a and those published for other rhizobia (269d, 278). Interestingly, analysis of the promoter regions of the nodABCIJnolOnoeI, nodSU, nodZ, noeE, and nolL loci showed that nod boxes with up to 11 base mismatches (compared to the consensus sequence) are still functional and regulate gene expression in a NodD1-dependent manner (195). It is possible, however, that some of these nod boxes have higher affinities for certain NodD1-inducer complexes than do others.
Thus, a microsymbiont such as strain NGR234 has many possibilities for fine-tuning nod gene expression. First, NodD1 may interact with a broad spectrum of inducing molecules to produce complexes that may preferentially trigger transcription from certain nod boxes over others. Second, supplementary activators of transcription (e.g., SyrM1  and y4xI ) could, together with NodD1, form a regulatory network which exerts interlaced control over dispersed nodulation loci. Third, repressors such as NodD2 and NolR (73) could further modulate nod gene expression. Together, these systems control the expression of NGR234 nodulation genes in a host-specific way, possibly resulting in the secretion of different sets of molecular signals in response to the macrosymbiont. Nod factors are obviously the most important of these signals, and their synthesis depends upon the timely and coordinated expression of many nod genes.
In response to the release by host plants of appropriate inducers, rhizobia synthesize and secrete a family of lipochitooligosaccharides (LCOs) called Nod factors. The first step in Nod factor assembly is performed by an N-acetylglucosaminyltransferase encoded by nodC (95). Chain elongation by NodC takes place at the nonreducing terminus (134, 135, 170). Then the deacetylase NodB removes the N-acetyl moiety from the nonreducing terminus of the N-acetylglucosamine oligosaccharides (132, 253). Finally, an acyltransferase encoded by nodA links the acyl chain to the acetyl-free C-2 carbon of the nonreducing end of the oligosaccharide (51, 226). Although not essential, NodI and NodJ seem to be involved in the export of Nod factors (34, 74, 254).
However, recent work suggests that nodA and nodC are also components of host-specific nodulation (hsn) (225). NodA varies in its specificity for different fatty acid substrates, thus contributing to the host range. For instance, replacement of R. meliloti nodA by R. tropici nodA leads to the production of Nod factors acylated with vaccenic acid instead of C16:2 (51), whereas Bradyrhizobium sp. strain ANU289 NodA is unable to direct the transfer of the R. leguminosarum bv. viciae nodFE-dependent multiunsaturated fatty acids to the chitin oligosaccharide acceptor (222). NodC is also a determinant of the length of the Nod factor backbone and thus of host specificity (133, 135).
Properly expressed nodABC genes are all that is needed for the synthesis of acylated dimers to pentamers of N-acetyl-d-glucosamine that possess symbiotic activity on certain plants (3, 200, 252). Thus all other Nod factor substituents must play more subtle roles in nodulation, perhaps by permitting interaction with certain plants or by protecting Nod factors from degradation. In this sense, these substituents can be likened to baroque decorations—they enhance rather than support the basic structure (Table (Table4).4). By definition, the loci that control these additions are unique to one or a few rhizobia and are thus hsn genes.
Two basic types of fatty acids are N-linked to the terminal nonreducing sugar of the chitomeric core by NodA (Table (Table4):4): (i) the relatively common saturated or monounsaturated fatty acids (e.g., stearic and vaccenic acids) and (ii) the rarer, highly unsaturated compounds containing two to four double bonds as found in R. leguminosarum and R. meliloti strains containing nodEF (54, 252). NodF has homology to acyl carrier proteins, while NodE is a β-acetoacetylsynthase (21, 22, 49, 245). Transfer of R. meliloti nodEFGHPQ into R. leguminosarum bv. trifolii or bv. viciae confers on these strains the ability to nodulate M. sativa but prevents nodulation of the normal hosts, Trifolium repens and V. sativa, respectively (50, 70). Inactivation of R. leguminosarum bv. viciae nodE leads to the synthesis of vaccenic acid rather than the polyunsaturated fatty acids (C18:4) produced by the parent strain (252). Insertional inactivation of R. leguminosarum bv. trifolii nodE severely inhibits the nodulation of several Trifolium species and simultaneously enhances the nodulation of P. sativum and V. sativa (59, 251). Mutation of R. leguminosarum bv. viciae nodE renders the strain Nod− on V. sativa (32). Replacement of R. meliloti nodEF with R. leguminosarum bv. viciae nodEF leads to the synthesis of Nod factors containing a polyunsaturated C18 fatty acid side chain which resembles those produced by R. leguminosarum bv. viciae (52). Thus, nodEF and the unsaturated Nod factors they produce appear to be specializations necessary for nodulation of the legume tribes Trifolieae and Vicieae.
An additional sugar, which can be either arabinose or fucose, represents another type of decoration.
noeC and/or downstream genes are essential for arabinosylation of A. caulinodans Nod factors (171). In these Nod factors, the reducing terminus can be fucosylated, arabinosylated, or both (172). Perhaps this suggests that both arabinosylated and fucosylated Nod factors are necessary for nodule formation on the stems of Sesbania rostrata (163, 164). d-Arabinosylated Nod factors result in larger numbers of nodules on the roots of S. rostrata, whereas the presence or absence of l-fucose has no effect. In contrast, other hosts of A. caulinodans prefer fucosylated Nod factors (75).
Fucose is a frequent substitution, and its addition to Nod factors is encoded by nodZ in many rhizobia (162, 171, 208, 210, 261). Since all the usual Glycine max symbionts (B. elkanii, B. japonicum, R. fredii, and strain NGR234) secrete fucosylated Nod factors, the expectation that mutation of nodZ would prevent the nodulation of soybeans was strong. In fact, B. japonicum nodZ is required for nodulation of M. atropurpureum but not of G. max (261). Similarly, nodZ mutants of strain NGR234 produce nonfucosylated Nod factors and are unable to nodulate Pachyrhizus tuberosus but retain the capacity to nodulate soybeans (208). Since transconjugants of R. leguminosarum containing B. japonicum nodZ acquire the capacity to nodulate Macroptilium (but not to fix nitrogen) (162), it seems as if fucosylated Nod factors play a role in nodulation of only a few legumes. In A. caulinodans, NolK is involved in the synthesis of GDP-fucose, which is used by NodZ as the fucosyl donor in fucosylation of A. caulonidans Nod factors (171).
NodH and NoeE, the only two characterized sulfotransferases, are specific to the reducing terminus of Nod factors (68, 154, 209, 224, 239). Both enzymes use 3′-phosphoadenosine-5′-phosphosulfate as the sulfate donor. 3′-Phosphoadenosine-5′-phosphosulfate is synthesized by the enzymes encoded by nodPQ (242, 243). NodH-dependent sulfation of the reducing terminus at C-6 of R. meliloti Nod factors is a major determinant of host range in R. meliloti (154, 224). nodH mutants which fail to produce sulfated Nod factors were thought to completely lose the capacity to nodulate M. sativa but gain the ability to nodulate the nonhost V. sativa (49, 224). Although other data indicate that nodulation of alfalfa by some nodH mutants is strongly impaired rather than abolished (125, 186), these results confirm that optimal nodulation of M. sativa requires sulfated Nod factors whereas V. sativa does not tolerate them (70).
In contrast, NoeE is a fucose-specific sulfotransferase (113, 209). Mutation of strain NGR234 noeE blocks the nodulation of P. tuberosus, while its introduction into the closely related strain USDA257 extends the host range of R. fredii to include Calopogonium caeruleum (113). Surprisingly, these transconjugants do not acquire the ability to nodulate P. tuberosus (M. Hanin, unpublished results).
Unlike sulfate groups, acetyl residues may be found at both extremities of Nod factors, i.e., on C-6 of the reducing or nonreducing termini, as well as on the fucose (Table (Table4).4). Furthermore, acetate groups on the fucose can move from one α-hydroxyl group to another (12). Mutation of nodX provokes a drastic reduction in Nod factor production by R. leguminosarum strain TOM and the ability to nodulate primitive cultivars of P. sativum (48). In R. leguminosarum bv. viciae however, a fucosyl group added by NodZ can functionally replace the missing acetyl group in a nodX mutant (189). Although NodL is responsible for acetylation of C-6 of the nonreducing terminus in R. meliloti Nod factors (2), R. meliloti nodL mutants are impaired in their ability to elicit infection thread formation on M. sativa but form nodules with only a moderate delay (96). A similar phenotype was observed when the corresponding mutant of R. leguminosarum bv. viciae was used to inoculate V. hirsuta, while nodL nodF double mutants of R. meliloti are unable to penetrate their hosts (2). In strain NGR234, disruption of the flavonoid-inducible nolL gene results in the synthesis of NodNGR factors that lack the 3-O- or 4-O-acetate group (12). Interestingly, the nodulation capacity of the mutant NGRΩnolL is not impaired whereas transconjugants of USDA257 containing nolL of NGR234 produce acetylated Nod factors and nodulate the nonhosts C. caerulum, L. leucocephala, and Lotus halophilus (12). Acetylation of the fucose on Nod factors of R. etli also conditions efficient nodulation of some P. vulgaris cultivars and of the alternate host Vigna umbellata (44).
NodS is an N-methyltransferase (93, 94, 129), while NodU and NolO control carbamoylation at C-6 and C-3 (or C-4), respectively, on the nonreducing N-acetyl-d-glucosamine (57, 129, 131). Inactivation of nodS in A. caulinodans, strain NGR234, and R. tropici abolishes the nodulation of L. leucocephala and P. vulgaris (158, 276). Introduction of either nodS or nodU into R. fredii extends its host range to include Leucaena spp. (129, 144).
The fucose group of Nod factors of B. japonicum USDA110, NGR234, and USDA257 is mostly 2-O methylated (Table (Table4).4). In strains NGR234, NoeI controls this function, and mutation of noeI leads to the production of fucosylated Nod factors that are not 2-O methylated (131). Introduction of noeI into R. loti allows the production of 2-O-methylated NodNGR factors. However, nodulation tests using L. leucocephala, Lotus japonicus, M. atropurpureum, and V. unguiculata and the NGRΔnoeI mutant have failed to identify a host which requires 2-O-methylated NodNGR factors (S. Jabbouri and M. Hanin, unpublished results).
Nod enzymes, which are responsible for the elaboration of Nod factors, are among the principal determinants of host specificity. Indeed, a number of them were called hsn determinants before the unified nod gene nomenclature was established. Most of them catalyze the adjunction of baroque decorations to the core Nod factor molecule (which is synthesized by NodC, NodB, and NodA); the effect of these different families of molecules on legumes is discussed below.
Although much information on the influence of Nod factor substituents on host range exists, no strict correlation can be drawn between the types of LCOs produced by rhizobia and the plants they nodulate. For example, although the Nod factors produced by R. etli and R. loti are identical, the two species have distinct host ranges (Phaseolus spp. and Lotus spp., respectively) (33). Also, the major Nod factors secreted by R. leguminosarum bv. trifolii are the same as two of the major LCOs produced by R. leguminosarum bv. viciae (188, 255), yet the two biovars nodulate distinct plants. Although the inability of NodD proteins to recognize the host flavonoids (and thus to activate nod gene expression) might be responsible for some of these phenotypes, these data emphasize the point that Nod factor structures alone cannot be used to predict host range. Furthermore, two rhizobia that nodulate the same plant may secrete different Nod factors. R. tropici and R. etli produce sulfated and acetylfucosylated Nod factors, respectively (Table (Table4),4), but both effectively nodulate P. vulgaris (202, 203). B. elkanii, B. japonicum, strain NGR234, and strain USDA257 have a number of common hosts (207), but their Nod factors vary considerably (Table (Table44).
The amounts of Nod factors released by rhizobia are also important in determining the host range. For instance, introduction of strain NGR234 nodD1 into R. meliloti increases Nod factor production by about twofold and permits the nodulation of V. unguiculata, a nonhost (214). Similarly, conjugation of the nodSU genes of strain NGR234 into R. fredii USDA257 not only extends the host range of the transconjugant to include L. leucocephala (144) but also vastly increases Nod factor production (Table (Table5).5). In contrast, the NGRΔnodSU mutant, which secretes only 1/10 the amount of Nod- factor produced by the wild type, is incapable of nodulating L. leucocephala (Table (Table5).5). One interpretation of these data is that higher levels of Nod factors (produced when a functional NodS or NodU is present) (129), not the presence of N-methylated or 6-O-carbamoylated LCOs, is necessary for nodulation of Leucaena.
Although the supporting evidence is lacking, it also seems likely that in addition to variations in the levels of Nod factors, different members and various proportions of the respective Nod factor families are secreted into the rhizospheres of specific legumes. There, only certain combinations of Nod factors, present at optimal levels, are capable of inducing deformation (Had) and curling (Hac) in homologous legumes (118, 148, 154, 169, 206, 213, 238, 252, 269a).
Cooperation among Nod factors and related signal molecules has also been demonstrated. Minami et al. (177, 178) examined the effects of LCOs on changes in the expression of early nodulin genes in G. soja roots and found that any combination of one active Nod factor with a nonspecific LCO was sufficient to induce mRNA expression of early nodulin genes such as Enod2. In contrast, both synthetic Nod factors and the chitin pentamer PACT, which is unable to provoke Had on G. soja, induced the transient accumulation of Enod40 mRNA. It would thus seem that not only are absolute levels of Nod-factors important but also the composition and relative proportions of the mixtures excreted by rhizobia are necessary for the induction of different components of the nodulation pathway (190).
The extent to which specific Nod factors bind to their respective plant receptors is still unknown. In Medicago, two putative Nod factor binding sites (NFSB) have been characterized (23, 181). Some properties of NFSB1 of M. truncatula suggest that it may also be involved in processes other than nodulation. First, its affinity and specificity for LCOs are low. Second, a similar site exists in particulate fractions of tomato roots (23). In contrast, NFSB2, which was isolated from the microsomal fraction of Medicago varia cell suspension cultures, is sensitive to proteases and shows high affinity for Nod factors (181). Covalent linkage of both the lipid and the chitooligosaccharide moieties, as well as O acetylation of the nonreducing sugars of Nod factors, is required for high-affinity binding to NFSB2 (108). In competition experiments in which the ligand specificity of NFSB2 was assayed by titrating 35S-labelled R. meliloti Nod factors against synthetic LCOs, those with short (C8) acyl chains were poor competitors compared to those with C16 and C18 fatty acid side chains (108). However, NFSB2 does not select for the presence of sulfate on the reducing sugar or for the number and/or positions of double bonds in the acyl chain. Since sulfated R. meliloti Nod factors are required for optimal nodulation of Medicago (see above) and as synthetic LCOs containing at least two double bonds are more active biologically than are monounsaturated compounds (53), it seems doubtful that these putative receptors are host-specific determinants of the Medicago-R. meliloti interaction.
Similarly, a root surface lectin of Dolichos biflorus has apyrase (Ca2+-dependent ATP diphosphatase) activity and binds Nod factors from both heterologous and homologous rhizobia (69). Reverse PCR-based techniques using mRNAs of L. japonicus and M. sativa root confirmed the presence of orthologues in these two legumes. Similar experiments were also performed with D. biflorus and Arabidopsis, leading to a second lectin in the former and a nonlegume lectin in the latter (223). Since much work still has to be done on these Nod factors binding lectin/nucleotide phosphohydrolases (LNPs), it is too early to pronounce on their exact symbiotic role. It is intriguing, however, that D. biflorus has two LNPs. Close orthologues of LNP1 are found only in legumes, while D. biflorus LNP2 is more closely related to apyrase sequences of nonleguminous plants (223).
Although Nod factors are bacterial products, they share many characteristics with plant hormones (213). One of these is that perturbations in the auxin-cytokinin balance mimic Nod factors in inducing meristematic activity of nodule meristems (128). Another is that the range of concentrations over which both hormones and Nod factors are active can vary by a millionfold. This suggests a desensitisation or an adaptation by the plant to increasing hormone concentrations (137). In turn, this implies that covalent and reversible modification to the “receptors” occurs; these suggestions are supported by the fact that the ethylene receptor is homologous to two-component chemoreceptors of bacteria (38). Perhaps this explains why the search for Nod factor receptors in direct Nod factor binding studies has been only moderately successful.
In principle, the chemical stability of Nod factors should influence the biological activity of the LCO mixtures secreted by rhizobia. In fact, chitinases and other enzymes rapidly cleave purified Nod factors in the host rhizosphere. The degradation products formed are only weakly active on their respective hosts (118, 258). Plant chitinases have been studied extensively in the context of plant-pathogen interactions. Among these are isoforms that are induced in symbiotic and/or pathogenic interactions (103, 192, 257, 273, 281). Distinct isoforms of chitinases and related hydrolases cleave specific β-1,4-linkages in the carbohydrate moiety of Nod factors (Tables (Tables66 and and7),7), resulting in the formation of acylated degradation products that are either di-, tri- or tetrameric. In general, leguminous and nonleguminous chitinases that belong to the same chitinase class have similar substrate specificities. Interestingly, the baroque decorations on certain Nod factors have a strong influence on the stability of these molecules (241, 258, 259) and modify the cleavage specificity of a given chitinase. For example, pentameric LCOs are hydrolyzed in two positions by an M. sativa class I chitinase, but only one of these positions is susceptible in sulfated Nod factors (Table (Table7).7). Furthermore, the sulfate group at the reducing end and the O-acetyl group at the nonreducing terminus of R. meliloti Nod factors dramatically increase the stability of these LCOs, rendering tetrameric molecules resistant to degradation by a large number of different chitinases (Table (Table6).6). Novel enzymes have been identified in Medicago, and their cleavage preferences are different from those of known chitinases (Tables (Tables66 and and7).7). One of them is an extracellular lipodisaccharide-releasing Nod factor hydrolase that cleaves all tetra- and pentameric R. meliloti Nod factors. Preincubation with R. meliloti Nod factors stimulated the activity of this enzyme in Medicago roots, suggesting that Nod factor inactivation is an early feedback response by the host plant.
However, two different types of observations suggest that chitinases may not play major roles in the determination of host specificity. First, the class I, II, III, V, and VI chitinases isolated from Nicotiana tabacum, as well as the class III and IV chitinases of Beta vulgaris and Daucus carota (all nonlegumes), degrade Nod factors in similar ways to that used by legume chitinases (Table (Table6).6). One interpretation of these data is that chitinases form part of a generalized plant response system rather than tailoring Nod factor degradation to specific legume-Rhizobium associations. A second important consideration is that root hydrolases take hours to degrade Nod factors (260), while the first responses of root hairs to inoculation by rhizobia (or to the presence of purified Nod factors) can be measured in seconds (64). Nevertheless, expression of a chitinase from Serratia marcescens by R. fredii USDA191 transconjugants resulted in delayed nodulation and a marked decrease in total nodule number on soybean cultivar McCall in comparison with the wild-type strain (146). What is undisputed, however, is that chitinases modulate the levels of Nod factors in the rhizosphere and, in so doing, help suppress defense responses stimulated by high concentrations of LCOs (231, 260). Since defense reactions also occur in aborted infection threads (273), it is possible that plant hydrolases help regulate Nod factor levels even after the bacterium has entered the plant. In other words, chitinases and related hydrolases cleave and so inactivate Nod factors. Local induction or down-regulation of these enzymes in leguminous roots would thus modulate local LCO concentrations. Since Nod factor levels seem to be host-range determinants, plant hydrolases might help tailor host specificity.
Although the mechanism of root hair deformation is poorly understood, it probably resembles that of root hair development (118). The tips of Nod factor-treated V. sativa root hairs swell before polar growth is induced. Induction of tip growth is preceded by local hydrolysis of the cell wall and followed by polar growth (118). Schiefelbein and Somerville (233) showed that the phenotype of root hair-specific mutants of Arabidopsis thaliana resembles root hair deformations caused by rhizobia. Thus, initiation and deformation of root hairs appear to be ontogenically related.
Nod factors are the signals required for entry of rhizobia into some legumes. Concomitant treatment of G. max and V. unguiculata roots with NodNGR factors and nodABC mutants of strain NGR234 or B. japonicum USDA110 permits these strains to nodulate and fix nitrogen on their respective hosts (215). NodNGR factors also allow the entry of R. fredii USDA257 into the roots of the nonhost Calopogonium caeruleum (215) and of the nodABC mutant of NGR234 into Macroptilium atropurpureum (213). Similarly, coapplication of purified A. caulinodans Nod factors and a nodA mutant of the same strain restores the formation of outer cortical infection pockets in Sesbania rostrata (56). These data leave little doubt that at least in some symbioses, Nod factors act like keys to legume doors. Interestingly, a recent report suggests that one of the “latches” that is opened by a Nod factor “key” has been found. A transposon-tagged mutant of Lotus japonicus, which is arrested at the stage of bacterial recognition, was identified (232). The mutation, which has been named nin (for nodule inception), is required for the formation of infection threads and is a transcription factor.
Once Nod factors have opened the root hair door to the invading rhizobia, additional bacterial signals appear necessary for continued development of the infection threads. This is clearly seen in variations of the Nod factor complementation experiments described above. These experiments work well if the strain deficient in Nod factor production is mutated in genes essential for Nod factor synthesis (nodABC) but not if the strain is mutated in nodD1. In this latter case, rhizobia bunch up in the curled part of the root hair and the infection thread fails to develop toward the root cortex, resulting in contortions that resemble a cerebellum (Fig. (Fig.2E).2E). It thus seems probable that products of genes other than those needed for the synthesis of Nod factors, but also under the control of NodD1, are required for the continued development of infection threads.
NodD proteins, with their ability to recognize different inducers, the complex mixtures of Nod factors as well as the various levels at which they are maintained, are not the only determinants of host specificity. Although Nod factors permit rhizobia to enter the outer door of the legume host and may play a role during nodule development (56, 267), additional “keys” are necessary for the formation of symbiotically proficient nodules. In fact, later steps of the infection process such as infection thread formation and propagation, as well as bacterial release into the cytoplasm of infected cells, require constituents of the cell wall and in some cases secretion of specific proteins (for a review, see reference 275).
Symbiotically relevant components of the rhizobial cell wall include extracellular polysaccharides (also known as exopolysaccharides) (EPS), lipopolysaccharides (LPS), capsular polysaccharides (CPS and KPS), and cyclic β-glucans (Fig. (Fig.3).3). Surface polysaccharides (SPS) form a complex macromolecular structure at the bacterium-plant interface. They accumulate on the surface of the prokaryote as a capsular layer but are also released into the extracellular space as bacterial slime.
EPS I and EPS II represent the two major classes of EPS synthesized by R. meliloti. EPS I members are polymers of an octasaccharide repeating unit of succinoglycan ranging in size from 106 to 107 Da for the high-molecular-weight (HMW) fraction (152). In contrast, the low-molecular-weight (LMW) fraction is formed of monomers, dimers, and trimers (277). Although mutants of R. meliloti Rm1021 incapable of synthesizing succinoglycan (EPS I) produce normal root hair curling on M. sativa, the nodules are devoid of bacteria and bacteroids and thus are ineffective (76, 151). Addition of the LMW fraction, specifically trimers of succinoglycan, to roots of alfalfa restored the nodule invasion capability of the exo mutant (5, 277). Interestingly, EPS I mutant strains of R. meliloti derepressed for the synthesis of another class of EPS are capable of forming normal nodules on alfalfa (99). EPS II members are polymers of modified glucose-(β-1,3)-galactose (99), and only small amounts (as little as 7 pmol per plant) of the purified LMW fraction are sufficient to restore nodule invasion by noninfective strains (102). Thus, although EPS II members can replace succinoglycan in nodule invasion, they are not required when wild-type EPS I is produced.
Fluorescence microscopy analyses show that nodule invasion is aborted at various stages in different exo mutants (40). Cells that lack ExoR, a negative regulator of exo gene expression (212), vastly overproduce succinoglycan and are unable to colonize curled root hairs and form infection threads. In contrast, a mutant with a mutation in exoY that is incapable of synthesizing succinoglycan (since it lacks the first enzyme in the biosynthetic pathway) (217) colonizes curled root hairs but forms few, very short infection threads. Although the exoH mutant that produces symbiotically dysfunctional succinoglycan forms infection threads longer than those produced by the exoY mutant, they never extend as far as the base of the root hair cell. Thus, in R. meliloti, succinoglycan could be regarded as a symbiotic signal (or key) required for opening the inner root hair door(s). Mutants that fail to produce the symbiotically active forms of these EPSs in adequate quantities are incapable of penetrating the adjacent plant cell and thus remain blocked or trapped in the infected root hair (40, 153). Also, since nodules induced by EPS I mutants of R. meliloti show pronounced symptoms of plant defense, it is possible that a molecule derived from the EPS I biosynthetic pathway functions as a suppressor of plant defense reactions (182). It has been postulated that in other plants, EPS of R. leguminosarum bv. viciae accelerates root hair curling and infection in such a way that invasion by rhizobia precedes the plant defense response (272). Similarly, it is thought that the wild-type cyclic β-glucans of B. japonicum (Fig. (Fig.3)3) may function as suppressors of host defense responses (20).
Formation of symbiosomes (which involves the release of rhizobia from infection threads into the cytoplasm of infected nodule cells) in plants other than M. sativa requires different components including EPS, LPS, CPS, KPS, and cyclic β-glucans (Fig. (Fig.3,3, Table Table8).8). Although EPS− mutants of M. loti are fully effective on Lotus pedunculatus, they provoke small, ineffective, tumor-like growths on L. leucocephala (127). Similarly, Exo− mutants of R. leguminosarum bv. trifolii (271), R. leguminosarum bv. viciae (24), and strain NGR234 (39) are ineffective on their respective hosts, T. repens, P. sativum, and L. leucocephala. In contrast, Exo− mutants of R. leguminosarum bv. phaseoli (58) and R. fredii (140), are fully effective on the determinate legumes P. vulgaris and G. max. One possible explanation of these data is that EPS is required for the formation of fully effective symbioses on plants that produce indeterminate nodules but not on legumes which form determinate nodules. However, Parniske et al. (192) showed that an EPS-defective mutant of B. japonicum which forms nitrogen-fixing nodules on G. max was ineffective on G. soja. Furthermore, a mutant of Rhizobium strain TAL1145 that is deficient in EPS synthesis still nodulates various hosts, independently of their nodule type (193).
A number of mutations that encode similar phenotypes but that occur in a variety of genes have been found (Table (Table8).8). For example, where the production of cyclic-β-(1→2)glucans, rhamnose-rich SPS, LPS, SG, β-(1→2)glucans, and KPS is impaired, Nod+ Fix− phenotypes are also observed. Different regulators appear to control the expression of these diverse genes, and the resultant phenotype depends on the plant. Quite often an Inf+ Fix− phenotype is observed, suggesting that these genes play similar roles. One possibility is that such surface components are necessary for or form part of the developing infection thread. Hence, a plant that could not itself supply the missing carbohydrate would give a Fix− phenotype while the rhizobial mutant in a plant that normally synthesizes the compound would have no effect.
Several strains of Rhizobium secrete symbiotically active proteins. Among these, the nodO product is required for nodulation of V. hirsuta by mutants of R. leguminosarum bv. viciae with nodFELMNT deleted (62), as well as for extension of the host range of an R. trifolii nodE mutant to include V. sativa (67). nodO encodes a Ca2+ binding protein that is thought to form cation-specific channels in membranes of leguminous plants (66, 264). Although NodO and NodS have distinct biochemical functions, a nodO homologue of Rhizobium sp. strain BR816 was shown to complement a nodS mutant of NGR234 for nodulation of L. leucocephala (269e), suggesting that secreted proteins may supplement some Nod factor deficiencies. Secretion of NodO is dependent on a C-terminal signal of about 24 residues (265) and on the prsDE genes, which encode two type I-like inner membrane proteins (77). In addition to NodO, at least three other proteins are secreted via this system, two of which (PlyA and PlyB) are glycanases involved in the processing of bacterial EPSs (78).
Sequencing the symbiotic plasmid of NGR234 revealed flavonoid-inducible genes encoding components of a type III secretion system (TTSS) (87). In a number of bacterial pathogens, TTSSs are induced upon contact with host cells (of plants or animals) and deliver virulence proteins directly into the eukaryotic cytosol (150). Strain NGR234, as well as several strains of R. fredii, was shown to excrete at least three to five proteins in a NodD1- and TTSS-dependent manner (145, 274). In USDA257, the nolXWBTUV cluster (corresponding to the nolX, rhcC1, nolB, rhcJ, nolU and nolV genes of NGR234) (274) regulates the nodulation of G. max in a cultivar-specific manner (168), whereas the TTSS of NGR234 profoundly affects the nodulation of various legumes such as P. tuberosus and Tephrosia vogelii (274). The absence of conserved nod box regulatory elements in the promoter regions of the nol and rhc operons indicates that NodD1-dependent transcriptional regulation of the TTSS genes is mediated by another factor. y4xI, an HrpG homologue which is under the control of a functional nod box, is thought to be the key intermediary in the regulatory cascade between flavonoids and activation of the TTSS machinery in NGR234 (274). The delayed induction of rhc genes in comparison with loci involved in the elaboration of the Nod factors suggests that the TTSS machinery is assembled after Nod factors have been elaborated and that protein export begins when intimate contact between bacteria and root hairs has been established. Thus, the most important role of TTSS-secreted proteins such as NolX and y4xL of NGR234 would occur after the bacteria have entered the plant, possibly during the development of infection threads.
It has been suggested that bacterial invasion of plant cells triggers nonspecific defense reactions and that successful pathogens overcome these defenses (89). Similarly, invading symbionts have probably evolved different strategies to lower host plant defenses. TTSS proteins and polysaccharides may contribute to this phenomenon. Some plants would perceive these compounds as part of the infection pathway and react to their presence by increased nodulation, as occurs in alfalfa inoculated with Exo+ R. meliloti and in T. vogelii, which favors strains of NGR234 with a functional TTSS. An alternate explanation is that these secreted substances function as suppressers of plant defense reactions, as proposed for various polysaccharides (20, 182). In contrast, P. tuberosus which is poorly nodulated by wild-type NGR234, produces many effective nodules when inoculated with TTSS mutants, suggesting that secreted proteins have detrimental effects (elicitors of defense reactions?) on certain hosts (274). Other rhizobia may have different keys to the inner door. Since apparently functional TTSSs are absent from R. leguminosarum and R. meliloti, other factors probably play similar roles. Some of these are undoubtedly polysaccharides, but it seems likely that a variety of keys exist. When they are used sequentially and in the correct combinations, infection thread development continues beyond the epidermal cells and a home for the invading rhizobia is constructed in the cortex of the root. Improper use of the keys results in abortion of the infection threads, as shown in Fig. Fig.22E.
Clearly, various components of nodulating rhizobia, which include the NodD proteins, the spectrum of inducers, the palette of Nod factor substituents, levels of Nod factors, polysaccharides, and secreted proteins, contribute to the control of host specificity. Although some of these may act synergistically (e.g., complementation of nodS mutants by NodO), most are part of a developmental process that leads to nodule invasion. Host plants are thus capable of controlling the successive steps of the rhizobial infection through checkpoints demanding specific keys. The absence of a suitable key does not necessarily result in a Nod− or Fix− phenotype, however. Most determinants of symbiotic specificity optimize the chances of forming effective nodules. Consecutive host specificity barriers are thus like hurdles of different heights that allow the host to select among many strains for the one best suited to its requirements.
Broad-host-range rhizobia most probably have keys which fit many legume doors. Experiments with two different but closely related bacteria, Rhizobium sp. strain NGR234 and R. fredii USDA257, and 452 different species of legumes showed that NGR234 and USDA257 form fully functional nodules on 136 and 66 species of legumes, respectively (207). Exactly which symbiotic determinants are shared by these two rhizobia is not known, yet both produce similar Nod factors. Those of USDA257 lack the baroque decorations encoded by nodS, nodU, nolO, nolL, noeE, and noeI of NGR234 (Table (Table4).4). Similarly, nodZ mutants of NGR234 still nodulate most plants tested, although the Nod factors lack all substituents carried on the fucose group (Table (Table4).4). In other words, it seems as if these additional Nod factor substituents extend an already broad host range but that they are not necessary for basic promiscuity.
If this logic reflects evolution, then ancestral Nod factors were unmodified N-acylated oligomers of N-acetyl-d-glucosamine, possessing simple, unsaturated fatty acid side chains. It should be remembered that polymers of N-acetyl-d-glucosamine, i.e., chitin, are abundant in nature where they form an important part of the cell wall of many fungi, including mycorrhizae. Vesicular arbuscular mycorrhizae establish symbiotic associations with a broad range of vascular plants (116, 246). The existence of mycorrhizal-like structures in fossils strongly suggests that fungal-plant associations developed before rhizobial symbioses (216). Since highly sensitive recognition systems for chitin oligomers are found in diverse plants, it is not unreasonable to suggest that evolution of rhizobial perception in legumes resulted in a perception apparatus that is specific to N-acylated chitin oligomers (i.e., Nod factors) (259).
Presumably, a bacterium producing ancestral Nod factors had the ability to nodulate many plants. Later, modifications followed which, in the case of NGR234, allowed it to nodulate even more plants. Acquisition of these broad-host-range genes was probably the result of lateral transfer. Concomitantly, Nod factor-cleaving chitinases coevolved in the legume host, thus directing selective pressure toward more stable Nod factors. In NGR234, this was provided through the acquisition of protective baroque decorations.
Morphological and molecular data suggest that both NGR234 and USDA257 have acquired the ability to nodulate three distinct subfamilies of plants, i.e., the Caesalpiniodieae, Mimosoideae, and Papilionoideae (207). In the Papilionoideae, this ability arose early and was maintained through to the Amorpheae, Robinieae, Indigofereae, Phaseoleae, and Desmodieae. Once acquired, the capacity to form symbioses with promiscuous rhizobia was only apparently lost in five tribes including the Cicereae, Vicieae, and Trifolieae. Since the microsymbionts of these plants generally secrete fewer, less highly modified Nod factors (albeit possessing polyunsaturated fatty acids), this implies that narrow host range is a specialization which developed for certain plants in restricted niches.
In other words, symbiotic promiscuity is probably ancestral to restricted host range. Support for this hypothesis comes from the observation that NGR234 and USDA257 also nodulate Parasponia andersonii (Ulmaceae) (207, 269). Parasponia spp. are small trees (up to 15 m high) which grow as pioneer plants in mountainous areas of Indonesia, Malaysia, and Papua New Guinea (1). Since Trinick isolated NGR234 in Papua New Guinea (268), this means that Parasponia-Rhizobium symbioses evolved in the same habitat. Most probably, the direct correlation that exists between solar energy input and species diversity (228) is driving evolution in tropical regions. Thus, ancestors of bacteria such as NGR234 could have been the first to intimately associate with legumes.
We thank C. Boivin, J. Dénarié, W. D'Haeze, J. A. Downie, M. Holsters, S. Jabbouri, T. Laeremans, K. Lindström, E. Martínez-Romero, G. Stacey, and J. Vanderleyden for helpful comments during the preparation of the manuscript. We are grateful to B. Relić for providing photographs of infected root hairs. Dora Gerber is thanked for her unfailing support of our work.
Financial assistance was provided by the Fonds National de la Recherche Scientifique and the Université de Genève.