Nodulation of groundnut by Bradyrhizobium: a simple infection process by crack entry

Authors

  • Fred C Boogerd,

    Corresponding author
    1. Department of Microbiology, Institute for Molecular Biological Sciences, BioCentrum, Vrije Universiteit, De Boelelaan 1087, 1081 HV Amsterdam, The Netherlands
    Search for more papers by this author
  • Diman van Rossum

    1. Department of Microbiology, Institute for Molecular Biological Sciences, BioCentrum, Vrije Universiteit, De Boelelaan 1087, 1081 HV Amsterdam, The Netherlands
    Search for more papers by this author
    • 1Universidade Católica de Portuguesa, Escola Superior de Biotecnologia, Rua Dr A.B. de Almeida, 4200 Porto, Portugal.


Corresponding author. Tel: +31 (20) 444 7194; fax: +31 (20) 444 7229; e-mail: fcb@bio.vu.nl

Abstract

Infection of legumes by rhizobia may occur by immediate intercellular penetration of root cells (crack entry) as an alternative mode to the more elaborate infection through infection threads. The intercellular spreading mode of infection is exemplified through a comprehensive description of root infection by Bradyrhizobium and nodule organogenesis in Arachis hypogaea (groundnut). The role of axillary root hairs and the processes of plant penetration and intercellular spreading, of internalization and intracellular multiplication, and of bacteroid differentiation are described. Then flavonoids and phytoalexins, Nod factors, lectins, and surface poly(oligo)saccharides pass in review. The roles of these various (macro)molecules in the chemical communication between the two symbionts are discussed. Attention is given to special features of groundnut nodules; the presence and functions of oleosomes and other bodies, the presence and functions of nodule lectins, and the evidence for the export of amides from the nodules are discussed. Finally, a speculative model for the groundnut infection process is presented.

1Introduction

1.1Cultivation and taxonomy of groundnut

Groundnuts were cultivated as early as about 1000 B.C. in Brazil and Peru, around their centre of diversity in the Mato Grosso, Brazil. They were cropped widely in the West Indies in pre-Columbian times. Dissemination of groundnut around the globe occurred presumably in the 16th and 17th centuries with the discovery voyages of the Spanish, Portugese, British, and Dutch. Magelhães carried (Peruvian) groundnut to the Far East around 1520 [1]. Groundnut is a (sub)tropical crop grown between 40°N and 40°S of the equator, since it has a high irradiation requirement. It grows in regions with an average annual rainfall of 500–1200 mm; it thrives best when over 550 mm of rain are received evenly distributed during the growing season. Groundnut grows best on well-drained sands or sandy loams [2].

Groundnut is a member of the Leguminosae family, subfamily Papilionoideae and tribe Aeschynomeneae. Taxonomically it groups together with Aeschynomene, Stylosanthes, and Discolobium, besides 22 other genera [3]. The genus Arachis counts 22 species, of which 9 are reported to be nodulated [1, 3]. Remarkably, A. hypogaea, the only cultivated species, is not known in its wild state [1]. A. hypogaea nodules were first noted by Poiteau as early as 1853 and A. glabrata nodules were depicted by Bentham in Martius' Flora Braziliensis (mid-19th century).

Groundnut possesses a high symbiotic nitrogen fixing capacity. The amount of nitrogen accumulated by groundnut is high compared to other tropical legumes [4]. Groundnuts are grown in large-scale, mechanised agricultural systems and in small-scale or subsistence systems. Pod yield in subsistence farming systems is much lower, at 780 kg ha−1 on average, than in large-scale, mechanised systems, with 2900 kg ha−1 on average [5]. Approximately 90% of the world groundnut crop is produced in developing countries [2].

1.2Crack entry deserves to be studied in depth

In general, two main types of nodules exist: determinate and indeterminate. Determinate nodules are spherical and develop from a non-persistent hemispherical meristem, while indeterminate nodules are elongated and cylindrical and develop by distal growth from a persistent apical meristem [1, 6, 7].

In essence, two modes of infection have been described: one by root hair entry and infection thread spreading and another by crack entry and intercellular spreading [8–10]. The infection thread mode is the best studied, and occurs with most temperate legumes yielding indeterminate nodules (Vicia, Trifolium, Pisum, and Medicago spp. among others) and with some (sub)tropical legumes yielding determinate nodules (Phaseolus, Glycine, Vigna, Macroptilium, Lotus, and Pueraria spp. among others). Root hair infections leading to determinate nodules occur by narrow infection threads as opposed to broad infection threads that lead to the formation of indeterminate nodules [7].

Crack entry/intercellular infection occurs in a few (sub)tropical legumes (Arachis, Sesbania, Stylosanthes, Neptunia, Aeschynomene). Comparatively few studies have been engaged to elucidate the molecular signals involved in the establishment of nodules by the crack entry/intercellular spreading mode of infection. Several of the phenomena that are essential in the infection thread infection mode do not occur with the intercellular spreading infection mode. In studying the latter mode, novel insights with regard to nodule morphogenesis may arise.

In all plant species investigated, the rhizobial mode of invasion and nodule morphogenesis is under control of the host plant. The host plant controls nodulation and also the mode of infection as may be concluded from the following pieces of evidence. (i) The nodulating or non-nodulating phenotype is conserved among taxonomically related legume species [4]; (ii) Nodule type bears botanical taxonomic relevance [11, 12]; (iii) One and the same rhizobial strain may infect different host legumes by different infection pathways, e.g., Macroptilium, Cajanus, and Vigna via infection threads (via root hairs) and Arachis via intercellular spreading, even resulting in different bacteroid morphology [13, 14]; (iv) The non-nodulating (mutant) phenotype in groundnut is determined by two duplicate recessive genes of the plant [15, 16]. Apparently, the host provides the genetic background for the mode of infection and nodule morphology.

It is therefore to be expected that taxonomically related leguminous species will show analogous nodule morphogenesis. This may aid the study of nodule morphogenesis in newly found symbiotic legume species. Moreover, investigation of the intercellular spreading mode, in being simple and sharing common features with infection processes in some nitrogen-fixing non-legumes [17], may yield insights that could provide a framework on how to extend nodulation to agronomically important plant species [18, 19]. However, it must be emphasized that not all non-legumes are invaded via crack entry, it is not the rule [20].

In this paper, we review the groundnut-Bradyrhizobium nodulation process in detail. We highlight nodule organogenesis and pay attention to various aspects during this developmental process such as molecular signalling during pre-infection, involvement of plant defence responses, the mode of bacterial root entry by cracks, as well as their intercellular spreading. Two underlying themes are woven through this review: (i) differences and resemblances between crack entry and infection thread invasion and (ii) crack entry viewed as a mild form of phytopathogenesis with a favourable course. With regard to this latter theme nodulation of leguminous plants with rhizobia has been referred to as a beneficial plant disease [21], (refined) parasitism [22, 23], and a plant defence reaction [24]. If such a point of view has merits (but see [25, 26]), it will apply even more to the crack entry mode of infection than to the infection thread mode of invasion.

2Nodule organogenesis

2.1Crack entry

The infection by crack entry/intercellular spreading necessitates penetration of the bacterial microsymbiont through the epidermis. Epidermal breaches that allow such penetration occur at sites where lateral roots (Sesbania[27, 28], Arachis[29], Stylosanthes[30], Neptunia[31]) or adventitious roots (Aeschynomene[32], Sesbania[33, 34], Neptunia[31]) protrude, respectively, the root or stem epidermis and so provide penetration sites at the fissure. The formation of such a ‘crack’ was also observed with Parasponia (non-legume) infection by Bradyrhizobium: the rhizobia colonise the root surface, beneath which rapid outer cortical cell division takes place, leading to the formation of an epidermal dome and rupture of the epidermis [35]. Only in Mimosa scabrella, rhizobia were observed to enter between intact epidermal cells [36]. This was also the commonest entry observed with potato (Solanum tuberosum) after inoculation with (brady)rhizobial strains that enter legume hosts via a non-root hair entry mode [37]. Interestingly, alfalfa roots inoculated with Agrobacterium tumefaciens and Rhizobium trifolii, to which the R. meliloti nodulation genes were transferred, were unable to elicit curling of alfalfa root hairs, but were able to induce genuine nodules at a low frequency. These nodules harboured bacteria only in intercellular spaces, whereas they were completely devoid of infection threads and of intracellular bacteria [38]. Apparently, both mutant strains were able to penetrate alfalfa roots via crack entry and to spread intercellularly thereafter. R. leguminosarum strain 1020 represents a similar example of uncoupling of nodule morphogenesis from root hair infection and infection thread formation. This strain formed ineffective nodules on subterranean clover at lateral root emergence without root hair infection; neither marked curling nor infection threads were found in root hairs; after crack entry via breaks in the epidermis, the bacteria spread throughout the developing nodule in infection threads [39].

Promiscuous nodulation of Arachis (and Stylosanthes) has been attributed to the mode of infection via cracks in the epidermis. However, intercellular penetration would also lead to a (hyper)sensitive response by the plant; generally, no other bacteria than Bradyrhizobium seem to invade the intercellular spaces in groundnut. Up to now, only the broad host range Rhizobium sp. NGR234, which is able to nodulate dozens of legume genera, including groundnut [40], was suggested to enter groundnut roots via the direct intercellular route, because the bacteria were seen inside a nodular structure, while no infection threads were observed microscopically; however, the nodules formed were ineffective [41]. In any case, specificity of the symbiotic partner is also required with this type of infection; epidermal cracks are not sufficient to cause infection, since artificial wounding has never led to nodule formation (e.g. [42]). On the contrary, artificial wounding allows infection by pathogenic Agrobacterium, another member of the Rhizobiaceae.

Chandler studied determinate nodule morphogenesis in A. hypogaea[29]and Stylosanthes spp. [30]in detail and found the processes in both genera to be similar. We will here present the developmental process for A. hypogaea and include, in a comparative manner, relevant observations from Stylosanthes spp., S. rostrata, M. scabrella, Aeschynomene spp., Discolobium pulchellum, Neptunia spp., and Parasponia spp.

2.2Function of axillary root hairs

Root hairs play a pivotal role in the infection thread mode of infection. They provide the attachment sites for compatible bacteria. After binding, rhizobia penetrate the root hair and move to the nodule primordium via infection threads. The function of root hairs in the crack entry mode of invasion is not clear, because bradyrhizobia enter the plant at epidermal breaches. Root hairs on A. hypogaea tap and lateral roots are rare or entirely absent, while tufted clusters or rosettes of multicellular hairs are found in the lateral root axils [43, 44], as shown in Fig. 1. Such axillary root hairs were also found with Stylosanthes, some species of which also possess normal root hairs. Axillary hairs are long and thick-walled in A. hypogaea[29]. In Stylosanthes spp. the axillary hairs are thin-walled with areas of wall thickening [30]. The occurrence of multicellular root hairs also occurred at the infection sites of P. rigida[45]. In S. rostrata, axillary root hairs appear only after inoculation of the roots [27], or after incubation with Nod factors [46], whereas stem nodules are formed without axillary hairs [33]. Similarly, root hairs were observed neither in stem-nodulating Aes. afraspera[32]nor in hydroponically grown N. natans[31]or N. plena[47].

Figure 1.

Scanning electron micrograph of a rosette of multicellular root hairs in the axil of the lateral and primary (tap) root of groundnut (Arachis hypogaea) cultivar Red Spanish. The width of the graph corresponds to 1.8 mm. Reproduced from [43]with permission of Dr D.J. Brady.

Deformation and curling of axillary hairs upon inoculation occurs in Arachis and Stylosanthes[29, 30], but root hair curling is not a prerequisite of infection and nodulation [30]. Bradyrhizobium cells were even seen to be entrapped within ‘shepherd's crooks’ (which are always observed with the infection thread mode), but penetration of the root hair by bradyrhizobia in the crook and infection thread formation in the root hair never occurred [29, 30]. Nevertheless, root hairs appear to be essential for nodulation since non-nodulation is strongly associated with absence of axillary root hairs [48]. The suppression of axillary root hair growth by toxic levels of aluminium also prohibited nodulation [43]. Nodulation of groundnuts at the axils of lateral roots appeared to be dependent on the presence of emergent axillary root hairs, because nodulation failed to develop when well-developed root hairs were present at the time of inoculation [49].

Axillary root hairs in Stylosanthes show an unusual high cytoplasmic activity, as evidenced by the presence of many ribosomes, mitochondria, strands of endoplasmatic reticulum, Golgi bodies, and Golgi vesicles; Golgi vesicles are particularly numerous, indicating that those cells are secretive [30]. Bradyrhizobial cells were seen to be polarly attached to axillary root hairs [30]; such polar attachment also occurs to epidermal cells of M. scabrella[36]. Polar attachment to root hairs is also commonly observed between rhizobia and root hairs in the infection thread infection mode, just prior to infection thread inititation [8].

The axillary root hairs emerge from epidermal cells of the lateral root primordium. This primordium originates from pericycle cells opposite the protoxylem strand of the central stele. These initial processes take place within the primary root, for which reason groundnut nodules were sometimes termed endogenous nodules [1]. The axillary root hairs and the nodules, however, originate on the lateral root that is on its way out: nodules are attached to the lateral root and not to the primary root. The use of the term ‘crack entry’ meaning that the breach, caused by the lateral root pushing through the primary root epidermis, constitutes the infection site, is inaccurate. The actual infection site in groundnut is the site between an epidermal cell and the axillary root hair basal cell of the lateral root.

2.3Plant penetration and intercellular spreading

After entry, Bradyrhizobium cells occupy the space between epidermal and cortical cells. Beneath some, but not all, of the axillary root hairs basal cells become enlarged. Such enlarged basal cells are the first to become infected by some of the invading bradyrhizobial cells, while others continue to spread intercellularly. Successful infection is restricted to penetration sites where such large basal cells are found [29]. In summary, it appears that the presence of an emerging root hair together with an enlarged basal cell, co-occurring with the emergence of the lateral root, determines infectibility in groundnut. Interestingly, in M. scabrella subepidermal root hair-like structures were often observed [36], which might be analogues of such enlarged basal cells. Invaded large basal cells repeatedly divide and become smaller; these cells become incorporated in the nodule tissue.

Although infection threads may be more widespread in aeschynomenoid-type nodules than has hitherto been suspected [50], neither Arachis nor Stylosanthes ever show structures resembling infection threads in either root hairs or developing nodules [17, 29, 30, 42]. Instead, bradyrhizobia spread intercellularly in groundnut by separating cortical cells at the middle lamellae. In contrast, Rhizobium Br 3454 penetrating intercellularly in M. scabrella appears to progress through the primary cell wall rather than by separating the cells at the middle lamellae [36]. The bacteria spread further through the root cortex in an intercellular undefined matrix, which may consist of broken plant cell wall fragments and of bacterial exo- and capsular polysaccharides. A similar amorphous electron-dense matrix also surrounds rhizobia penetrating Parasponia[35, 45]and Mimosa[36]. At sites where intercellular infection takes place the adjoining plant cortical cells show a greatly increased activity: mitochondria become numerous and long strands of endoplasmatic reticulum are aligned in parallel with the cell wall.

In Stylosanthes, no enlargement of basal cells was observed, but basal cells were seen to be invaded after which they collapsed; bradyrhizobia penetrated into deeper layers of the cortex by this progressive invasion and collapse, rather than by intercellular spreading. Collapsed cells, however, appeared as intercellular spaces filled with Bradyrhizobium cells. Bradyrhizobial penetration through progressively collapsing infected cells towards the meristematic zone also occurs in stem-nodulating Aes. afraspera[32]and M. scabrella[36]. This invasion mode resembles a hypersensitive response more than a symbiotic interaction until meristematic cells start to be infected. Invasion via progressive collapse may be related to some recent observations in the infection thread mode of infection. Here a specific hypersensitive response might be part of the mechanism by which the plant controls infection; a substantial part of the infection threads in alfalfa inoculated with its symbiont R. meliloti were shown to abort in some root cortical cells which accumulated phenolic compounds and proteins associated with defence mechanisms and where both symbionts underwent necrosis [51]. Alternatively, a general non-specific mechanism for the induction of defence reactions may exist: the induction of plant defence responses could be decoupled from the expression of a hypersensitive reaction in Phaseolus vulgaris[52]. Ineffective (mutant) but also effective (wild type) soybean/Bradyrhizobium nodules exhibited chitinase and peroxidase activities in the cortex [53]. It has been noted that the levels of plant defence activities, e.g., chitinase and phytoalexin production, in Fix- nodules of soybean are well below those found in classical hypersensitive responses elicited by pathogens [24]. Thus, plant defence responses, including a limited (hyper)sensitive response, may regulate nodule development to some degree [24]. It would be interesting to measure such plant defence activities in tissues of crack entry species.

A mixed type of invasion has been shown to occur in several tropical legumes. In N. natans, bona fide infection threads were shown to develop following initial intercellular spreading [31]. A similar route of bacterial infection, i.e., intercellular spreading and intracellular dissemination via infection threads, might also occur in S. rostrata stem [54, 55]and root [27, 55]nodules. Such a mixed infection pattern has also been implicated in nodulation of Aes. americana roots [56], of Aes. fluminensis roots (but not stems) [50], and of N. natans[31], N. plena[47], and M. scabrella[36]roots. Bacteria were in those cases released from infection droplets at unwalled ends of infection threads [27, 31]. A further variation on the theme consists of intercellular spreading combined with subsequent formation of persistent nitrogen fixation threads, in which bacteria are confined by host cell wall material in structures resembling infection threads ([57], but see [35]). These features have been observed in Parasponia[58, 59], most examined species of Caesalpinioideae, some species of Papilionoideae, but no species of Mimosoideae[57, 60]. The thread-like nitrogen-fixing structures in this group of legumes are thought of as forerunners of most nodules found in familiar crop plants [60]. In the genus Chamaecrista, species can be found that at one extreme show bacteroids confined to peribacteroid membranes and at the other extreme bacteroids contained within persistent infection threads [61].

2.4Internalization and intracellular multiplication

After intercellular spreading of the rhizobial cells, cell walls of particular plant cells that will eventually internalize the bacteria are structurally altered and appear to become partially degraded, as in a cellulolytic process [29]. Cellulolytic activity has not yet been demonstrated in groundnut-nodulating Bradyrhizobia, but it is present in Rhizobium spp. [9], A. caulinodans[62], and Azoarcus spp. [63]. Interestingly, the endo- and exoglucanases detected in Azoarcus sp. strain BH72 were cell surface-bound; they possibly mediate a more localized digestion of plant cell wall polymers in comparison to phytopathogens that secrete cellulases [63]. Plant enzymes too could be involved in cell wall degradation. Fragments of cell wall may effect a (hyper)sensitive response, resulting in abortion of a number of infection attempts [8, 17].

Eventually, the plasma membrane is partly exposed to the bacteria in such invaginations (Fig. 2). Then the bacteria are internalized in host cells in an undefined matrix [29]. At this point, the bacteria have to cope with the plant cell turgor. During internalization, the electron-dense matrix surrounding the rhizobial cell is bound by a host-derived membrane [29]. Eventually, just a few groundnut plant cells may successfully become infected after internalization [8, 17]. Encapsulated bradyrhizobial cells multiply rapidly within the invaded plant cell. First they are enclosed in the same membrane envelope, but the envelope is later divided so that each cell becomes enclosed separately [29]. The invaded plant cell also divides rapidly, thereby distributing the endophytes like other cell organelles.

Figure 2.

Electron micrograph showing internalization of a Bradyrhizobium cell (r) by an Arachis hypogaea cortical cell. The invaginations of the cell wall expose the cell membrane. The rhizobia are encapsulated. Strands of endoplasmatic reticulum are visible near the invaginations (×20,000). Reproduced from [29]with permission of Dr M.R. Chandler.

As the entire nodule tissue originates from one or a few plant cell infections, the nitrogen fixing zone of the determinate nodule is filled with infected cells, sometimes separated by files of uninfected cells [8, 13, 29, 30, 64]. A central core without interspersed uninfected cells was also observed for stem nodules of Aes. afraspera[32]and Aes. indica[65], and for both root and stem nodules of D. pulchellum[66]. The presence of rays of uninfected cells has been interpreted as demarking different infection origins [29, 30]. Because there is neither much space in the infected cells for host cytoplasm nor almost any interspersed uninfected cells (Fig. 3), the question has been raised as to how high rates of nitrogen fixation are coupled to ammonia assimilation reactions in groundnut nodules [13]. In Aes. indica, plasmodesmata were numerous in the walls of infected cells, indicative of the ability to quickly transport products from one cell to another [65]. Recently, it has been suggested that the rays of uninfected cells in groundnut nodules are involved in assimilating ammonia and in transporting fixed nitrogen from the infected zone to the nodule parenchyma cells [67, 68].

Figure 3.

Scanning electron micrograph showing Arachis hypogaea nodule cells filled with spherical Bradyrhizobium strain 32H1 bacteroids (B). The bacteroids are located peripherally around the centrally located nucleus (N) and vacuole (V) (×1700). Reproduced from [13]with permission of Dr D. Sen.

2.5Bacteroid differentiation

Only after the A. hypogaea host cell ceases to divide, the rod-shaped bradyrhizobial cells differentiate into large, spherical bacteroids as shown in Fig. 3. Groundnut bacteroids are tightly arranged in the plant cytoplasm; the majority of the infected cells possess a central plant vacuole with the nucleus appressed to its side [13, 69]. The bacteroids are enclosed singly in peribacteroidal membrane sacs [13]. The bacteroids were first reported to be spheroplasts or protoplasts [70, 71], but it was later found that the (bacterial) outer membrane is shed and replaced by a new (bacteroid) outer membrane [72]. This is one of the first steps in bacteroid differentiation.

The transformation of a bacterium into a bacteroid should have dramatic consequences at the transcriptional, translational, and metabolic level. In free-living Bradyrhizobium 32H1, a correlation exists between relatively high DNA gyrase levels, low topoisomerase I levels, and expression of bacteroid-associated functions; higher DNA gyrase levels were observed in low-O2-grown cells [73]. It was suggested that relatively high negative supercoiling permits the expression of bacteroid-associated functions, such as nitrogenase [73]. A major regulatory role of oxygen in this developmental process is also indicated by the finding that 60% of the major proteins of the free-living (groundnut-nodulating) Bradyrhizobium strain 3G4b20, grown under carbon or oxygen limitation in chemostat cultures, were affected by oxygen [74]. Likewise, a shift from aerobic to micro-aerobic conditions in chemostat cultures of Bradyrhizobium strain 32H1 greatly reduced the extracellular polysaccharide level [75]. Remarkably, the levels of cell-associated cyclic β-1,6-1,3-glucans were essentially unchanged [75]. Lastly, a Bradyrhizobium NC92 mutant produced larger nodules than the wild type NC92, while nodulation was delayed and the specific nitrogenase activity was reduced relative to NC92 [76]. By contrast, the carbon costs and the efficiency of nitrogen fixation were similar for both strains. Further analysis might indicate if this phenotype is due to an impaired intracellular bacterial proliferation or to a limited number of successful transformations of bacteria into bacteroids.

During the transformation of a bradyrhizobial cell into a bacteroid a tremendous increase in volume occurs through swelling. Swelling was also observed in free-living Bradyrhizobium 32H1 incubated in succinate-containing medium [77]. Swelling is considered to be the cause for the very narrow peribacteroid space (<20 nm at some positions) [78]. The width of the peribacteroid space is considered important for nitrogen fixation as a widening of this compartment was observed in senesceing nodules and in nodules harbouring a Nod+Fix mutant strain of Bradyrhizobium NC92 [64]. Nodule functioning (nitrogen fixation) only commences when the bacteroids are fully differentiated [72]. This implies that rhizobial proliferation within the developing nodule requires the supply of all nutrients, including nitrogen, from the plant. At this developmental stage the bradyrhizobia are therefore only parasitic, not yet mutualistic (see [8]). In this respect, the peribacteroid membrane can be regarded as a physiological barrier shielding the microsymbiont from the ‘hostile’ plant cytoplasm.

After crack entry, intercellular spreading, and nodule organogenesis, finally a mature active nodule is established. Arachis nodules are of the typical aeschynomenoid-type, determinate, small (1–5 mm) and oblate, lacking lenticels, without interspersed uninfected cells in the infected tissue, and occurring in the axils of lateral roots [8, 12]. Other salient features are the absence of an endodermis-like layer separating the outer cortex from the parenchyma cortex (which is observed in most other nodules) and the dense packing of thick-walled cortical cells yielding a rather thin nodule cortex [13, 67]. The anatomy of a functional Arachis nodule is shown in Fig. 4. Since the xylem of the primary root is tetrarch, and the lateral roots arise from the xylem poles, the laterals with nodules arise in an orderly alignment of 4 vertical rows [44]. Aeschynomenoid-type nodules are common in the legume tribes Aeschynomeneae and Adesmieae[11], and Dalbergieae[79].

Figure 4.

Light micrograph of a transverse section of an Arachis hypogaea–Bradyrhizobium strain 32H1 nodule showing a vascular bundle (VB), the nodule cortex (C) and bacteroidal tissue (B). Reproduced from [13]with permission of Dr D. Sen.

3Molecular communication between the symbionts

Molecular communication between the micro- and macrosymbiont is a necessity for mutual recognition of the compatible symbiotic partners. Specificity in such identification is obligatory in order to exclude non-beneficial or parasitic (micro)organisms from entering the legume plant. Molecular signals that are implicated in legume–rhizobial communication fall into two classes, plant-derived and bacterial. The plant signals encompass flavonoids [80–82]and glycoproteins (lectins) [9]. The bacterial signals cover lipochitin oligosaccharides (LCOs) [25, 80, 81, 83, 84], also called Nod factors, and poly- and oligosaccharides (exo-, capsular, lipo-polysaccharides) [9, 85–87]. These different groups of signalling molecules and their relevance to the ArachisBradyrhizobium symbiosis, will be discussed in the four consecutive paragraphs. Along with flavonoids, the related topic of phytoalexins will be discussed.

3.1Flavonoids and phytoalexins

Legumes are known to produce aromatic compounds that are able to induce the rhizobial structural nod genes via regulatory NodD proteins [80, 82]. Most of these substances are collectively known as flavonoids; they include isoflavones, chalcones, flavonols, flavones, and anthocyanidins amongst others. Also coumarines and betaines can have nod-gene inducing activities. Besides nod gene induction, the flavonoids appear to have multiple roles during several stages of nodule [88]and plant [89]development. Flavonoid excretion by groundnut, however, has not been demonstrated yet, although coumarin may be among the exuded compounds [4]. Nevertheless, the expression of a nodA–lacZ fusion in Bradyrhizobium strain NC92 was induced by the addition of groundnut exudate and by the addition of the isoflavonoids genistein and daidzein [90]. Considering the promiscuity of groundnut, a large spectrum of inducing compounds might be anticipated. Another promiscuous legume, P. vulgaris, has indeed been shown to release nod gene inducers belonging to four different classes of flavonoids: flavonols, anthocyanidins, flavanones, and isoflavones [91, 92]. For Rhizobium NGR234, even the monocyclic aromatic compounds vanillin and isovanillin have been shown to activate the NodD1 protein [80].

An indirect indication for the presence of various flavonoids can be inferred from the fact that groundnuts produce several isoflavone phytoalexins (see below) via the phenylpropanoid pathway [93]; so groundnut at least has the capability to produce an isoflavone ring structure. Interestingly, it was found that most groundnut-nodulating Bradyrhizobium strains were able to use flavonoid degradation products as sole carbon and free energy sources [94].

Positive signals are necessary, but also negative signals must be dealt with. Incompatibility between plants and microrganisms is expressed as an array of plant defence responses, sometimes culminating in localized host-cell death (hypersensitive response) [95–98]. Such response then results in failure of the microorganism to multiply. Accumulation of phytoalexins is thought to play an important role in resistance [95, 99], although other active defence mechanisms may be equally important [95]. Many plants, including legumes [100], accumulate and excrete phytoalexins, which are low molecular weight antimicrobial compounds that are synthesized in response to challenge by microorganisms or abiotic agents [96]. Membranes are the usual targets of phytoalexin action [95]. Generally, a plant synthesizes phytoalexins that belong to no more than two or three classes [95]. As groundnuts are particularly prone to invasion by aflatoxin-producing Aspergillus flavus and A. parasiticus, and other fungi as well, antifungal phytoalexins have been extensively studied [101]. A plethora of phytoalexins is produced by groundnut belonging to three phytoalexin classes: stilbene, (iso)flavone, and aliphatic phytoalexins [101]. Unfortunately, phytoalexin production in response to inoculation of groundnut with (in)effective Bradyrhizobium spp. has not been studied so far.

Intriguingly, stilbene and isoflavone phytoalexins are structurally and biosynthetically related to nod gene-inducing flavonoids. For plants in general, stilbene synthase (STS) is related to chalcone synthase (CHS) [102]and this also holds true for groundnut [103]. CHS is a key enzyme in the biosynthesis of chalcone, flavone, and isoflavone flavonoids. Both enzymes use the same substrates, malonyl-CoA and p-coumaroyl-CoA, and catalyse the same type of reaction. Based on sequence analysis of cDNA, it was proposed that the two groundnut proteins possess a common scaffold necessary for substrate binding and type of reaction [103]. The products of enzyme catalysis, however, are different: resveratrol for STS and naringenin chalcone for CHS. Resveratrol is then further converted to stilbene phytoalexins, while, likewise, naringenin chalcone may (putatively) be converted into nod-gene inducing flavonoids. STS likely evolved from CHS by a limited number of amino acid exchanges [102].

Importantly, another link between flavonoids and phytoalexins has been indicated for soybean. The isoflavones genistein and daidzein and the chalcone isoliquiritigenin induce resistance to the phytoalexin glyceollin, which is structurally related to these nod gene inducers, in several soybean-nodulating (brady)rhizobia [104, 105]. Since a common nod gene (nodD1D2YABC) deletion mutant also acquired glyceollin resistance induction of the nod genes is probably not required for the expression of the resistance. In this case, the actual basis of the resistance was not studied, although it was not due to degradation or detoxification of glyceollin. Isoflavone/chalcone-inducible resistance to phytoalexins might be ascribed an important role for the survival of (brady)rhizobia in the rhizosphere [104], apart from degradation of phytoalexins [99, 101].

3.2Nod factors

The common nodABC genes have been found in all Azorhizobium, Rhizobium, and Bradyrhizobium isolates studied so far; the host-specific nod (hsn) genes are necessary for the nodulation of a particular host plant [80]. After induction by plant flavonoids, the bacterial nod gene products catalyse the synthesis of lipo-chitin-oligosaccharides (Nod factors). In turn, Nod factors can induce a variety of root responses, including membrane depolarisation of root hair cells, root hair deformation, root hair curling, pre-infection thread formation, early nodulin gene transcription, increased biosynthesis of flavonoids, root cortical cell division, and sometimes even formation of genuine nodules [6, 23, 80, 81, 106]. Up to now, however, Nod factors that are induced upon inoculation of groundnut with Bradyrhizobium have not been described. In spite of the fact that neither root hair infection nor infection thread formation are involved in crack entry/intercellular spreading, Nod factors should still be engaged somehow, as indicated by the following observations (i–v): (i) Bradyrhizobium strain IRc 78 is able to induce nodules on cowpea, soybean, and pigeonpea via root hairs, and on groundnut via crack entry, whereas mutants of this strain, generated by Tn5 mutagenesis of a DNA-segment that hybridized to nod genes of R. meliloti, failed to elicit nodules on any of the four legumes [107]. (ii) Bradyrhizobium strain NC92 is able to nodulate siratro and pigeonpea through root hairs, and groundnut through crack entry, whereas a Tn5 mutant, supposed to contain an insertion in one of the common nodulation genes, was unable to nodulate all three hosts [14]. A transconjugant of this NC92 mutant complemented with a plasmid carrying the common nod genes of R. meliloti, was able to elicit effective nodules on siratro. Unfortunately, the complemented transconjugant has not been tested on groundnut [14]. In another study, Bradyrhizobium NC92 was found to nodulate siratro, cowpea, pigeonpea, mungbean, and groundnut, whereas a strain containing an insertional mutation in the nodB gene was unable to nodulate any of these plant [90]. (iii) Bradyrhizobium strain Rp501, isolated from Parasponia nodules formed after crack entry, also nodulates siratro and cowpea via root hairs, while mutants derived from this strain by Tn5 insertions in the common nod genes failed to nodulate all three host plants [108]. (iv) A. caulinodans enters its host Sesbania rostrata only via crack entry in both root and stem; at the same time, it has been demonstrated to produce Nod factors [46]upon induction by naringenin [109]. (v) Deformation and curling of axillary root hairs is observed accompanying crack entry [29, 30]. All in all, the common nod genes seem to be required for crack entry and intercellular spreading.

An intriguing question remains to be answered: what is the function of Nod factors in crack entry and intercellular spreading? It seems reasonable to exclude all functions specifically related to root hair infection and infection thread spreading as possible functions of Nod factors in groundnut invasion. Also induction of root cortical cell division is not deemed necessary until root cortex cells are invaded, because principally only one cell needs to be infected for a successful nodulation. After internalization of bradyrhizobia, however, plant cell(s) division starts. Two possible mechanisms, which were proposed for root hair entry species, as to how Nod factors might trigger cortical cell division can be mentioned. Firstly, an increased synthesis of flavonoids in response to Nod factors (the Ini response) may occur; the induction of this second wave of flavonoids might affect the normal endogenous hormone balance of the root and render the cortical cells susceptible to cell division [6, 7, 110]. Secondly, expression of the ENOD40 gene, which encodes a symbiosis-specific oligopeptide as a regulator of auxin action, may be induced by Nod factors [111, 112].

The interesting suggestion was made that Nod factors are simply a key to the legume door and simply permit rhizobia to enter nodules via infection threads formed in root hairs; they would not be necessary for subsequent nitrogen fixation: addition of Rhizobium NGR234 Nod factors, at inoculation, to NodABC mutant strains of NGR234 and B. japonicum permitted these mutants to enter the roots and to elicit (pseudo)nodules on Vigna unguiculata and Macroptilium atropurpureum, and on Glycine max., respectively [113, 114]; NGR234 Nod factors also allowed R. fredii to enter and nodulate Calopogonium caeruleum, a non-host [113]. If this hypothesis has general validity, Nod factors would not be needed to start nitrogen fixation in the groundnut nodule.

3.3Lectins

Lectins are sugar-binding (glyco)proteins other than enzymes or antibodies [115]. Generally, lectins in seeds or cotyledons are confined to membrane-bound vacuole-derived protein bodies and they may have protein storage or mobilization functions [116]. Their main function may be engagement in the plant's defence against microbial pathogens [116, 117]or they serve as determinants of recognition or host specificity [118–120]. For root hair infection, the importance of lectins has been convincingly shown for pea. Pisum sativum seed lectin (PSL) was shown to accumulate on pea root hair tips [121], while a transgenic Trifolium repens, being transformed with the psl gene, became nodulated by R. leguminosarum bv. viciae that could not nodulate the wild type white clover plant [122]. In transgenic clover, PSL is correctly processed and targeted to the tips of emerging and growing root hairs [123]. As transformation of white clover with an altered psl gene that encodes a protein that is unable to bind sugar, resulted in the loss of nodulating capacity of R.l. viciae, it was concluded that the sugar-binding activity of PSL is necessary for infection [124].

For groundnut, four different major lectins, present in various tissues, have been described so far: PNL, GL, ML, and PRAII. The major groundnut seed lectin (PNL) has been purified [125]. In contrast to other plant seed lectins, PNL agglutinated human blood cells only after neuraminidase-treatment which removes negatively charged sialic acid sugar residues from their surface; agglutination could be inhibited by both α- and β-galactosides. The disaccharide d-galactose-β-1,3-d-N-acetylgalactosamine is the hypothetical ligand for PNL [125]. PNL is present in all examined (4456) lines of A. hypogaea and in most other Arachis spp. as well [126]. The galactose-specific PNL consists of four identical subunits and its molecular weight was estimated to be 110 kD [125], although under certain conditions a dimeric form is found [127]. Crystals of PNL grown in the presence of N6-benzylaminopurine exist as a dimeric structure with the phytohormone bound to it; so a naturally occurring phytohormone can modify the quaternary structure of PNL by dissociation [128]. PNL is actually a mixture of several (6–9) isolectins, each of which is characterized by a different pI value; three isolectin groups were detected, but only one profile was predominant; the number of isolectins depends on groundnut genotype [126].

Another galactose-binding lectin (GL) was detected first in nodules [129], but it is also present in other tissues. GL appeared to be similar to PNL in its molecular weight and amino-acid composition and ability to bind derivatives of galactose [127]. However, GL differs from the seed lectin PNL as the latter is non-glycosylated [127, 130]. GL contains the following sugars: mannose, galactose, fucose, xylose, and N-acetylglucosamine [130]. PNL and GL are likely to be encoded by different genes, because only four of the first ten N-terminal amino acid residues of GL correspond to those of PNL [130]. However, GL showed cross-reaction with the antibodies raised against PNL and also caused agglutination of human blood cells [127]. Thus previous enzyme-linked immunoassays and haemagglutination assays, most likely, have detected both galactose-binding lectins. Most of the galactose-binding lectins in young seedlings were in the cotyledons and hypocotyls [129, 131]. Levels of galactose-binding root lectins are also high initially, but they decline rapidly within 9–12 days after planting [129, 132, 133]. The galactose-binding levels also decline rapidly in all other plant tissues during the early stages of growth [132].

Another lectin was demonstrated in nodules: a mannose-specific glycoprotein (ML) [129]. ML is not present in seeds but was detected in large amounts in cotyledons and hypocotyls during the early stages of growth; in roots, lower levels of ML were found, although they were still 40-fold higher than those of GL [132]. Generally, the total ML content of plants is 25-fold larger than that of GL during the first two weeks after planting [132]. ML is quite different from PNL (and GL) in its properties: it differs in molecular weight, antigenic properties and amino-acid composition [127]. Accordingly, ML is not serologically related to PNL or GL [127].

Apart from PNL, GL, and ML, a fourth lectin is produced by groundnut roots, the glucose-specific PRAII [134, 135]. The inhibition pattern of its haemagglutinating activity with rabbit erythrocytes is markedly different from that of ML in showing a preference for β-linked sugars [136], whereas ML was shown to prefer α-linked sugars [127]. It is also different from PNL; for example, antiserum raised against PRAII does not react with PNL. Interestingly, groundnut root extracts showed rabbit erythrocyte-agglutinating activity, which is indicative of the presence of PRAII, starting from day four and remaining constant at least up to day 15 [134].

Finally, two glycosylated lectins, SL-I and SL-II, have been purified from the stems of 3–4-week-old groundnut plants [137]. They were distinct from PNL and PRAII; SL-I resembles ML, while SL-II is thought to be a new lectin [137].

With respect to possible functions of the four lectins present in groundnut roots the following data are relevant. Neither GL nor ML binds to any of a group of 11 (brady)rhizobia (7 effective Bradyrhizobium strains, 2 ineffective Rhizobium strains (NGR234 and ANU248), and 2 non-nodulating R. leguminosarum strains, using groundnut as a host); both lectins also failed to react with bacteroids of the effective Bradyrhizobium strain CB756 [132]. Neither did antibodies raised against GL bind to bacteroids derived from Bradyrhizobium strain 2290A [67]. Thus there is no support for GL and ML acting as symbiotic recognition determinants. Furthermore, the rapid decline in GL levels in groundnut tissues during the infectious period renders an involvement of GL in the early infection stages unlikely [129, 131, 132]. Thus, the biological functions of GL and ML in the early infection process have not been supported by experimental studies. The two other lectins, PNL and PRAII, possibly have functions that are related to their ability to bind to surface polysaccharides as will be discussed in the next paragraph.

3.4Surface polysaccharides and oligosaccharides

3.4.1EPS/CPS

Concerning the infection thread mode of invasion, the general view is that EPS/CPS is not required for the development of determinate nodules, but plays an essential role in the formation of indeterminate nodules [86, 138].

Concerning PNL, a biological function might be inferred from data demonstrating its ability to bind to surface polysaccharides of groundnut-nodulating bradyrhizobia. The effectively groundnut-nodulating Bradyrhizobium strains JLn(c) and RA-1 were shown to produce high molecular weight acidic, fibrillar polysaccharides and low molecular weight neutral glucans; they were found to be attached to cells and free in the culture medium as well [140]. The acidic heteropolymers consisted of the β-linked sugars glucose, galactose, glucuronic acid, mannose, and fucose, together with pyruvate and acetate as substituents [141]. In contrast to the neutral glucan, the acidic heteropolymer reacts with the groundnut lectin PNL and the acidic substituents pyruvate and acetate are likely to be involved in lectin binding [141]. Another lectin-binding acidic EPS molecule was reported to be produced by other groundnut-nodulating bradyrhizobia as well; it contained the α-linked sugars mannose, glucose, and galactose, and the acidic constituents pyruvate and acetate [142]. Exopolysaccharides were also found to be the major sites for PNL binding in the groundnut-nodulating Bradyrhizobia strains B.TG-3 and 5a, whereas only a minority was reported to be complexed with lipopolysaccharides [143]. The role of lectins in establishing a symbiosis may have evolved from the ability of lectins to agglutinate and immobilize bacteria as a defence reaction [117]; callus and cell suspension cultures of groundnut cotelydons were shown to secrete PNL into the medium [139]. In contrast, a group of 8 groundnut-nodulating bradyrhizobia failed to bind PNL, indicating that the interaction of bradyrhizobia with PNL is not a necessary event in the nodulation of groundnut [144]. It should be noted, however, that EPS/CPS formation strongly depends on the nutritional conditions [9, 118, 144].

Importantly, PNL has been shown to induce structural changes in groundnut-nodulating bradyrhizobia: cells displayed polarity in forming an extracellular polar outgrowth of the outer membrane opposite a polar periplasmic bay [145]. Such extracellular polar body consists of (lipo)polysaccharides, while the polar periplasmic bay is the place where presumably (lipo)polysaccharide synthesis takes place [146, 147]. In addition, PNL was found to enhance the incidence of extracellular polar body [145], although capsule formation also occurred without PNL as the culture grew older [140, 145]. Capsule development initiates at one pole of the cell and expands to engulf the entire cell [140]. Finally, purified PNL showed growth promoting effects on homologous Bradyrhizobium JLn(c) and RA-1 but not towards heterologous rhizobia [148].

3.4.2LPS

In general, for root hair entry species, defects in lipopolysaccharides more negatively affect the formation of determinate nodules than of indeterminate nodules [7]. The effects of rhizobial LPS deficiency in hosts that form indeterminate nodules range from no or little effect (Nod+Fix+) to inability to release bacteria from the infection threads (Nod+Fix) [7]. Symbiosis in hosts harbouring determinate nodules is more severely affected; an LPS mutant of R. leguminosarum bv. phaseoli could not be released from the infection threads into the nodules (Nod+Fix) [149]and LPS mutants of B. japonicum and A. caulinodans did not evoke nodules at all (Nod) [150, 151].

With respect to the role of LPS in the formation of determinate nodules of groundnut, experimental data are scarce. The lectin PRAII can be detected in groundnut from the fourth day onwards as soon as axillary root hairs emerge. Rabbit erythrocytes were demonstrated to bind to the axillary root hairs and to a certain population of root cortical cells, which indicates the presence of PRAII at these sites. PRAII was shown to interact specifically with LPS from the groundnut-specific Bradyrhizobium strain IGR92 [136].

Finally, it must be emphasized that the functional difference between EPS, CPS, and LPS in bacteria might be only marginal in providing surface oligosaccharides that function as binding sites or signals; it has been concluded that the complexity, heterogenity, and structural plasticity of surface polysaccharides make it difficult to ascribe exact functions to particular cell-surface components involved in rhizobial infection [87].

3.4.3Cyclic β-1,6-β-1,3-glucans

The presence of cyclic β-1,6-1,3-glucans in Bradyrhizobium merits discussion. Bradyrhizobium 32H1 [152]and B. japonicum[152, 153], and two Bradyrhizobium spp. isolated from the nodules of tropical tree legumes [154]have been demonstrated to produce cyclic β-1,6-1,3-glucans. The low molecular weight neutral glucans described for the groundnut-nodulating Rhizobium strain ATCC 51466 [155]and Bradyrhizobium strains JLn(c) and RA-1 [140]are presumably similar cyclic glucans, although these glucans appear to have a somewhat higher molecular weight (3000 Dalton) than the cyclic β-1,6-1,3-glucans. In addition, a phosphocholine-substituted β-1,6-1,3-glucan was isolated from B. japonicum[153, 156]. The neutral cyclic β-1,6-1,3-glucans consist of 10–13 glucose residues and, thus, are smaller than the anionic cyclic β-1,2-glucans (17–24 sugar residues) produced by Agrobacterium and Rhizobium[157, 158]. They are localized in the periplasmic compartment of free-living B. japonicum USDA 110 [157]and they are also synthesized by bacteroids of the same strain in soybean root nodules [153, 159]. A mutant strain of B. japonicum defective in synthesis of β-1,6-1,3-glucans formed ineffective nodules on Glycine max[160]. Interestingly, the cyclic glucans of B. japonicum elicit the production of the phytoalexin glyceollin as well as the isoflavone daidzein; however, daidzein levels were higher than glyceollin levels, whereas the levels induced by a fungal glucan elicitor were reversed [88]. The bradyrhizobial glucans might modulate the levels of isoflavonoids produced by the host. It has been suggested that the bradyrhizobial cyclic β-1,6-1,3-glucans are functional analogs of the rhizobial cyclic β-1,2-glucans [157]: the cyclic β-glucans play a role during plant infection, perhaps by subtly suppressing the plant defence response [85, 88], and during adaptation to osmotic changes [85, 158].

4Groundnut nodule special features

Apart from the afore mentioned remarkable characteristics, the groundnut nodule has a number of other salient features that altogether make this symbiosis quite unique in appearance. In the three consecutive paragraphes, the presence and function of oleosomes (lipid bodies) and other types of bodies, the presence and function of nodular lectins, and the evidence in favour of amide export from this determinate nodule will be discussed.

4.1Oleosomes and other bodies in groundnut nodules

Ultrastructural studies of groundnut nodules have shown the presence of four types of bodies that are not commonly observed in tropical and temperate legume nodules: oleosomes, microbodies, dense bodies and proteinaceous bodies. Oleosomes (lipid bodies) are found in the host cytoplasm pressed against the peribacteroid membrane, microbodies in the close vicinity of the peribacteroid membrane, and dense bodies are located in the peribacteroid space and are attached to the bacteroidal outer membrane [69, 78]. A dense body is shown in Fig. 5. Oleosomes and microbodies, together with amyloplasts, are also present in the 2–3 layers of specialized cortical cells surrounding the infected zone [69]. These cell layers also exclusively feature electron-dense proteinaceous bodies at the interface of plasma membrane and cell wall. They may be involved in the bidirectional transport of gases and solutes during symbiosis [161].

Figure 5.

Electron micrograph of an Arachis hypogaea–Bradyrhizobium strain 32H1 symbiosome showing a dense body (D), the peribacteroid membrane envelope (pme), the bacteroid outer membrane (om), and the bacteroid inner membrane (im). Reproduced from [78]with permission of Dr A.K. Bal.

Oleosomes may occupy 4–5% of the cell area in mature nodules elicited by Bradyrhizobium 32H1 [69, 78]. In contrast to seed oleosomes, the nodule oleosomes serve not only as a carbon and energy reserve, they also show active lipolytic activity in effective nodules [69]. Accordingly, more oleosomes were found to accumulate in the ineffective nodules elicited by a Nod+Fix mutant strain than in effective nodules derived from NC92 wild type [64]. The dense bodies and microbodies are thought to be involved in active lipid catabolism [78]. Lipids are oxidised via the β-keto oxidation pathway. Acetyl-CoA is condensed with glyoxylate to form malate via malate synthase [162]. Fatty acids are not taken up by bacteroids [69]. Similarly, none of the Bradyrhizobium strains out of a group of 17 was able to use acetate as sole carbon and energy source for free-living growth, whereas 16 of those strains could use malate [94]; by inference, this implies that malate (succinate) produced from acetate in the peribacteroid space is a substrate for the bacteroids, not acetate. Lipid bodies were shown to be utilised under photosynthate stress caused by extending dark periods or detopping of plants, allowing for undiminished nitrogen fixation for up to 2 days [78, 162, 163]. Oleosomes have also been found in nodules of S. rostrata[164], the cool–temperate (sub)arctic legumes Lathyrus maritimus[165]and Oxytropis maydelliana[166], and Aes. fluminensis[50]. Preliminary data indicate that they might also be present in Cajanus cajan, Calopogonium mucunoides, and Macrotyloma axillare[167]. Thus, oleosomes are found in both determinate and indeterminate nodules, and in both annual and perennial legumes.

4.2Lectins in groundnut nodules

As mentioned before, nodules of groundnut contain lectins, whereas nodule lectins have not been detected in Glycine max., Vigna unguiculata, and Medicago sativa[129, 168]. Recently, however, promoter–gusA fusions for three lectin genes were found to be active in nodules of M. truncatula[169]and a lectin-like protein was detected in pea nodules [170]. The occurrence of lectins in groundnut nodular tissues is of considerable interest, because they may be essential for the development and proper functioning of the nodule. The mannose- and galactose-binding lectins (ML and GL, respectively), present inside the nodules, together amount to 3% of soluble nodule protein [127, 168]. As evidenced by light and electron microscopic immunochemistry, both lectins are widely distributed throughout the cortex and infected zones [168], although ML levels are on average 50-fold higher than GL levels [132]. In the nodule parenchyma, they were both associated with vacuoles; this suggests that they are involved in protein storage processes [67, 168]. In the infected region, GL was found in the central vacuoles of the plant cells, whereas ML was observed in the host cytoplasm and on the nucleus. Neither GL nor ML was located on the peribacteroid or bacteroid membrane. So they are unlikely to take part in host-symbiont recognition [132, 144, 168]. Another study, whilst confirming the role of GL as vacuolar storage protein, indicated that this lectin might play several additional roles. Six weeks after inoculation, GL was detected in the extracellular matrix between cells of the boundary layer and the nodule parenchyma, a location that corresponds to the tissue layer forming a barrier to oxygen diffusion [67]. The monoclonal antibody MAC-265, which recognizes carbohydrate epitopes of a pea glycoprotein known to occlude intercellular spaces, reacted with groundnut glycoproteins at the same site, i.e. in the nodule parenchyma adjacent to the boundary layer; this MAC 265-reactive glycoprotein, but not GL, was also a major component of intercellular spaces in the outer regions of the nodule parenchyma [67]. Interestingly, dense bodies present in the peribacteroid space were shown to react with antibodies against ML [168]and GL [67]. These may be the same dense bodies as observed in electron micrographs (see above paragraph).

4.3Groundnut nodules export amides

Groundnut was reported to transport both amides and ureides; the non-protein amino acid 4-methyleneglutamine was identified as the major form of nitrogen in xylem from roots to shoot and leaves [171, 172], whereas allantoin and allantoic acid have been implicated in export from the nodules as well [173, 174]. However, the presence of 4-methyleneglutamine did not appear to be associated with nitrogen fixation [172, 175, 176], whereas ureides were undetectable or present at low levels [172, 175]; instead, asparagine was found to be the principal nitrogen product exported from nodules [175, 176]. Later this conclusion was essentially confirmed and extended [177]: in three species of groundnut (A. hypogaea, A. pintoi, and A. glabrata), ureide concentrations in xylem exudate (1–7% of xylem nitrogen) were much lower than in sap from the well-established ureide-exporter soybean (60–88% of xylem nitrogen). Moreover, these low ureide levels were not affected by groundnut species or cultivar, bradyrhizobial strain, plant size, growth rate, or stage of development, and importantly, were not related to nitrogen fixation. On the other hand, levels of asparagine accounted for more than 70% of xylem sap nitrogen; finally, of the total 15N2 recovered in xylem sap 90% was present in the form of asparagine, whereas less than 0.1% represented ureides [177].

In the genuine ureide-exporter soybean, numerous small interspersed uninfected cells in the central infected zone [178, 179]and the three innermost cortical layers [180]are involved in ureide synthesis; large peroxisomes and tubular endoplasmatic reticulum (ER) are present in these cells. In groundnut, interspersed uninfected cells are rare. Instead, arrays of uninfected cells are found occasionally. The presence of enlarged peroxisomes and tubular ER in such arrays of uninfected cells has twice been reported [68, 181]; in contrast, other authors reported that peroxisomes were invariably small and rare and that tubular ER was never present [182]. Moreover, the presence of tubular ER and peroxisomes does not necessarily imply extensive ureide synthesis; in the amide-exporter Robinia pseudoacacia, where peroxisomes and tubular ER are abundantly present in uninfected cells in the infected region, only 8% of total xylem-nitrogen consists of ureides [182].

Thus, it is now unambiguously proven that groundnut nodules export mainly asparagine during active nitrogen fixation and this tropical legume therefore harbours an amide-exporting determinate nodule type. Sesbania spp. and Aes. indica, two other legumes harbouring determinate nodules, were also found to export amides instead of ureides [183].

5A speculative model for the groundnut infection process

In Fig. 6, the differences and resemblances between the two modes of infection, crack entry/intercellular spreading and root hair entry/infection thread spreading, are summarized in a schematic representation of both infection processes. On the basis of what is known, we present the following speculative model that endeavours to connect the rather detached and scanty information concerning groundnut nodulation (Fig. 7). Since the experimental observations relevant to this model have already been presented, only some conspicuous features will be discussed here.

Figure 6.

Different stages of root hair entry/infection thread spreading and crack entry/ intercellular spreading. Left-hand views (L): root hair entry species (for instance, soybean); right-hand views (R): crack entry species groundnut. (A) Primary and lateral root sytems (L and R). (B) Normal root hairs (L) and axillary root hairs (R). (C) Root hair (L) and crack entry (R) infection. (D) Infection thread spreading (L) and intercellular spreading (R). (E) Central infected cores with small dispersed uninfected cells (L) and with an array of uninfected cells (R).

Figure 7.

Schematic representation of the postulated infection process leading to the groundnut–Bradyrhizobium root nodule symbiosis.

(i) Seedlings secrete flavonoids which induce compatible bradyrhizobia to produce and secrete Nod factors. The common nodABC genes, which are sufficient to synthesize the N-acylated glucosamine oligosaccharide backbones of Nod factors, are certainly required; it is not known which decorations are specifically required for nodulation.

(ii) Groundnuts secrete a variety of compounds that are engaged in the plant's defensive response to microorganisms. In contrast to (brady)rhizobia that are enclosed within infection threads, bradyrhizobia spreading intercellularly are continuously exposed to the plant's defence system and therefore need to protect themselves. The observed encapsulation of groundnut-nodulating bradyrhizobium strains might be a protection strategy; it may enable the bacterium to resist the plant's defence during the time of groundnut's greatest infectibility. Encapsulation may be perceived as the functional equivalent of the formation of infection threads. Bradyrhizobial encapsulation holds an interesting analogy to the way by which human pathogenic bacteria protect themselves through capsule formation from the human defence system. By way of encapsulation, pathogenic bacteria are able to evade the human defence response, at least for some time.

(iii) As the infection process proceeds, the ‘pathogenic’ bacterium needs to be recognized/labeled by the plant to allow for its internalization. Is a groundnut plant, in this regard, using a similar response strategy to a bradyrhizobial infection as humans do towards invading pathogens, employing lectins in a way observed in mammalian lectinophagocytosis? Lectinophagocytosis is a non-opsonic mechanism of phagocytosis in which lectin–carbohydrate interactions are essential [184]. Interestingly, since lectin–carbohydrate interactions are also important in the recognition and ingestion of bacteria by amoebae [185], it has been suggested that lectinophagocytosis of microorganisms by mammalian phagocytes represents a primitive system of host defence [186]. Opsonic phagocytosis via antibodies or complement represents a more elaborate (advanced) defence strategy to eliminate those bacteria that evade lectinophagocytosis [184].

We speculate that bradyrhizobia are internalized by one or a few plant cortex cells via a process reminiscent of lectinophagocytosis following (partial) cellulolytic degradation of cell wall components. Several lectins have been shown to be present in groundnut at the early infection stages. Moreover, some of them interact with compatible bradyrhizobial cells, thereby possibly providing the bacterium with a plant-derived protein coat. However, so far there are no experimental data showing the direct involvement of lectins in the internalization of bacteria in groundnut plant cells.

(iv) After internalization, simultaneous multiplication of host cell(s) and internal bacteria occurs. Finally, compartmentalization of the endosymbiont occurs within a specialised host lysosome, the symbiosome. The phagocytosis of bacteria into animal phagocytes normally results in fusion of phagosomes with lysosomes and subsequent digestion. However, some pathogenic bacteria replicate within a phagosome which does not fuse with lysosomes, whereas others are able to survive even after phagosome–lysosome fusion [187, 188]. Intriguingly, it has been proposed that bacteroids inhabit an environment which fulfils the definition of a lysosome [189]. Since the symbiosomes are morphologically different from the central lytic vacuole of plant cortex cells, they are thought to represent organ-specific modifications of lysosomes [189].

It is evident that the state of the art concerning the groundnut–Bradyrhizobium symbiosis is far from elaborate, but the contours of many interesting differences from and similarities with the root hair entry and infection thread spreading mode of infection are becoming visible. The speculative model here presented may serve as a working hypothesis that needs validation.

Acknowledgements

This work was supported by the Netherlands Government through the Directorate General for International Cooperation. We are grateful to W.H.O. Ernst, J. Mol, A.H. Stouthamer, and H.W. van Verseveld for helpful suggestions during preparation of the manuscript.

Ancillary