Nod factor internalization and microtubular cytoskeleton changes occur concomitantly during nodule differentiation in alfalfa

The symbiotic interactions between soil bacteria of the genera Rhizobium, Azorhizobium or Bradyrhizobium (here referred to as rhizobia) and plants of the Leguminosae family result in the formation of nodules, new organs in which the bacteria reduce nitrogen into ammonia which can subsequently be utilized by the plant.

Nodule organogenesis starts with a molecular dialogue between symbionts and takes place through a series of developmental stages. Rhizobia produce Nod factors (NFs), the synthesis of which is under the control of nodulation (nod) genes which are transcribed in the presence of plant flavonoids. Chemical studies have shown that NFs from all rhizobial species are lipochitooligosaccharides consisting of a backbone of N-acetylglucosamine residues which is decorated on its two terminal residues (reviewed by Dénarié et al., 1996; Long, 1996; Schultze and Kondorosi, 1996; Spaink, 1996). Each rhizobial species possesses a characteristic set of nod genes that specifies the length of the backbone and the nature of the decorations at both ends of the molecule, thus making the NFs specific for a given plant host (Roche et al., 1991). NFs are signal molecules involved in most of the early developmental responses which are elicited by the corresponding bacteria. Early rhizobia-dependent responses lead to various cellular events such as root hair induction and deformations, plant infection by means of tubular structures called infection threads and the formation of a nodule meristem whose activity ensures nodule growth (reviewed by Newcomb, 1981; Brewin, 1991; Roth and Stacey, 1991; Hirsch, 1992; Kijne, 1992).

In legumes such as alfalfa and vetch, the nodule meristem remains active for several weeks, thus leading to the formation of elongated indeterminate nodules comprising central and peripheral tissues. Histologically, central tissues are organized into five well-defined zones: the apical meristematic zone I, the prefixing (infection) zone II, the starch-rich interzone II-III, the nitrogen-fixing zone III and the proximal senescent zone IV. These central zones are surrounded by a parenchyma, vascular bundles, an endodermis and a cortex (Vasse et al., 1990). The differentiation of the central cells depends on whether the host cells are invaded or remain bacteria-free. Fully differentiated invaded cells are dramatically enlarged in size, highly polyploid and filled with nitrogen-fixing bacteroids (Truchet, 1978; Vasse et al., 1990). In contrast, non-invaded cells are of small size and their DNA content remains monosomatic (Truchet, 1978). Different developmental features characterize determinate nodules formed, for example, on the roots of siratro or soybean. In such nodules, meristematic activity is transient and the central cells differentiate almost simultaneously (Newcomb, 1981). As a result, round-shaped mature determinate nodules increase in size by cell enlargement and all central cells are more or less at similar stages of differentiation.

The plant cytoskeleton mediates several functions in living cells (reviewed by Seagull, 1989; Cyr and Palevitz, 1995). The cytoskeleton is dynamic in the sense that it continually rearranges by virtue of its two major components, microfilaments and microtubules which assemble and disassemble in response to extracellular and intracellular stimuli. This property can be correlated with fundamental cellular traits that the cytoskeleton is thought to regulate, such as, for example, cell division, cell shape and polarity, cell trafficking and the spatial organization of the cytoplasm (reviewed by Seagull, 1989; Goddard et al., 1994). An involvement of the plant cytoskeleton during early stages of nodulation has been suggested by studies showing that cytoskeletal reorganizations occur at the tip of root hairs treated with Nod factors (Allen et al., 1994) or in the root cortex of Vicia hirsuta either infected by its specific symbiont (Bakhuizen, 1988) or treated with Nod factors (Van Spronsen et al., 1995). Moreover, if one considers the general roles of the cytoskeleton in the context of symbiotic responses such as root hair induction and deformation (Ardourel et al., 1994), activation of cortical cells (Van Brussel et al., 1992; Ardourel et al., 1994) which enter the cell cycle but arrest in G₂ (Yang et al., 1994), oriented growth of an infection network in plant tissues (Bakhuizen, 1988; Ridge, 1992; Van Brussel et al., 1992), formation of nodulation-related division centers (Truchet et al., 1991) and the cell enlargement and DNA endopolyploidization typical of invaded central cells (Truchet, 1978), then it is reasonable to anticipate that variations in the cytoskeletal structure are likely to be involved in many of the symbiosis-related steps directing nodule development.

Very little is known about the early molecular mechanisms through which the host plant responds to rhizobial infection or NF treatment. The current hypothetical model proposes that NFs bind to plasmalemma-located receptors (Bono et al., 1995; Niebel et al., 1997), followed by subsequent signal transduction. Data indicate that Rhizobium meliloti NFs induce a depolarization of the plasma membrane potential (Ehrhardt et al., 1992; Felle et al., 1995), cytoskeletal changes (Allen et al., 1994) and calcium spiking (Ehrhardt et al., 1996) in alfalfa root hairs. In contrast, it has been shown that Rhizobium leguminosarum bv. trifolii NFs are internalized specifically into clover root hairs (Philip-Hollingsworth et al., 1997). In this study, we show that the microtubular cytosketon (MC) dramatically reorganizes in differentiating infected cells of both indeterminate and determinate nodules. In alfalfa nodules, MC changes initiate in the nodule zone where rhizobial Nod factors are internalized in infected cells and strongly correlate with symbiosis-specific cell differentiation traits. Thus, NF internalization, MC changes and cell differentiation are tightly coupled in the course of nodule development.

The list of bacterial strains and plant species used in this study is given in Table 1. Respective bacterial and plant growth conditions were as described by the references listed in Table 1. Spontaneous and NF- induced nodulation in alfalfa were achieved as described by Truchet et al. (1989) and Truchet et al. (1991), respectively.

Nodules harvested at different developmental stages depending on the nodulation type were processed for histological or ultrastructural observations as described by Vasse et al. (1990).

In situ visualization of the microtubular cytoskeleton was as follows. Nodules were fixed in 4% formaldehyde (prepared from paraformaldehyde) in microtubule stabilizing buffer (MSB) consisting of 60 mM Pipes, 10 mM EGTA, 1 mM MgCl₂, 0.1% Triton X-100 and 10% DMSO (pH 6.9, 30 minutes at room temperature) followed by a subsequent 30 minute fixation in 4% formaldehyde in phosphate- buffered saline (PBS) pH 7.4. After rinsing with PBS, the samples were infiltrated with sucrose up to 1 M in PBS, and cut into 8 and 50 µm thick sections at −20°C, using a MICROM HM500 M cryostat. Sections were deposited on poly-L-lysine-coated slides or in small containers, and tubulin immunolocalized by subsequent incubation with monoclonal anti-α-tubulin (Sigma T-5168) and an anti-mouse IgG-FITC antibody (Boehringer Mannheim, 821462). After staining in Evans blue (0.1% in PBS), to quench autofluorescence, the labeled sections were mounted in glycerol containing 1% 1,4-diazabicyclo- (2.2.2) octane as an anti-fading agent and 4,6-diamidino-2- phenylindole as a nuclear stain and viewed by laser confocal microscopy (Zeiss, LSM 410 Invert). Images were recorded either from a single focal plane with an average thickness of 0.5 µm or as an extended focus in which several confocal planes were superimposed. The gain and offset were chosen in such a way that all the 255 grey values were used resulting in 0 for the background (i.e. the slide without section) and 255 for the maximum fluorescence in the specimen. As a result of the chosen gain and offset, the scaling of the pseudo-color look up table is comparable for all images. Images were displayed, after background subtraction, as false colour images indicating increasing fluorescence intensity on a colour scale ranging from blue to red.

Rabbit anti-NF polyclonal antibody was prepared by Biocytex (Marseille, France) using R. meliloti purified NodRm-IV(S,C16:2) as immunogen. Immunserum was tested by ELISA according to standard procedures (Engvall and Perlmann, 1971). For immunocytochemistry several protocols were tested. Optimal results regarding labeling efficiency and preservation of cellular structure were obtained by fixing the specimen for 1 hour in 2% formaldehyde (prepared from paraformaldehyde) and 0.5% glutaraldehyde in 0.12 M cacodylate pH 7.2, or 2.5% glutaraldehyde in 0.2 M cacodylate pH 7.2 respectively for light and electron microscopy. After subsequent dehydration with ethanol and infiltration with LR white resin, polymerization was performed at low temperature under UV irradiation. For labeling, 1 µm semithin sections stuck onto poly-L-lysine-coated slides or ultrathin sections picked up on nickel grids covered with formvar were first incubated overnight at 4°C in 10% goat serum, 0.5% Triton X-100, 0.5% Tween 20 in Tris-buffered saline (TBS) to prevent non-specific binding of antibodies. For immunolocalization at the histological level, sections were incubated either overnight at 4°C or for 2 hours at 37°C with anti-NF immunserum diluted 1:50 in TBS. After rinsing in TBS, incubation was continued with anti-rabbit IgG antibody:1 nm gold (BioCell) for 1 hour at room temperature. The signal was amplified by using the BioCell silver enhancing kit. Sections were then stained briefly in 0.02% toluidine blue, mounted in DPX (BDH Laboratories Supplies) and observed by bright field or dark field microscopy with an Olympus Vanox light microscope. For immunolabeling at the ultrastructural level, sections were incubated as described above with the immunserum diluted 1:10 and an anti-rabbit IgG:20 nm gold was used as secondary antibody. Ultrathin sections were stained with uranyl acetate and lead citrate and observed with a Hitachi H600 transmission electron microscope.

The intensity of NF immunolabeling in different nodule zones was quantified by silver grain counting on labeled 1 µm thick semithin sections. The region of interest was identified under bright field optics at 100× magnification and outlined on the screen. The quantity of light reflected by each area was then measured under fluorescent epi- illumination using a computer-based image analysis system according to the procedure described by Blanchard et al. (1993) and converted into the number of grains per square µm. Background grain density measured on the resin but out of the section was finally subtracted from the grain density of each region to obtain the density per zone due to NF immunolabeling.

Speculating that cellular features occurring during nodule growth might reflect cytoskeletal changes (see Introduction), we first studied, in detail, the histological and ultrastructural differentiation and the cytoplasmic organization of central cells in alfalfa nodules. The terminology used in this study to describe the histological organization of alfalfa nodules and the ultrastructural stages in bacteroid differentiation is as described by Vasse et al. (1990).

Invaded cells enlarge isodiametrically in the distal part of zone II and become round shaped in proximal zone II, interzone II-III and zone III (Fig. 1A-C). Simultaneously, the clustering of cell organelles (plastids and mitochondria) and bacteroids at the cell periphery, shows that a modification in spatial organization takes place in zone II, particularly in the proximal part of this zone (Fig. 1B). Interestingly, as cells mature, the spatial positioning of bacteroids orientated perpendicular to the cell wall, whilst those in the center remain randomly distributed within the cytoplasm (Fig. 1C).

The positioning and the ultrastructural differentiation of cell organelles and of bacteroids were studied in more detail by electron microscopy. Both plastids (as undifferentiated proplastids) and rod-shaped mitochondria were found to be randomly distributed within the plant cytoplasm of bacteria-free cells observed either in the meristematic zone I or in the most distal border of zone II where the infection network develops (Fig. 2A). Progressive changes in ultrastructure and positioning of organelles occur after the release of bacteria along the prefixing zone II (Fig. 2B). First proplastids, and then mitochondria increase in size (Fig. 2B) and in length (Fig. 2C,D) and move to the periphery of the cytoplasm where they orientate parallel to, and in close contact with, the plasmalemma, particularly at intercellular spaces (Fig. 2E). Very short microtubules are often seen in the limited cytoplasmic space changes from a random orientation in the cells of the prefixing zone II (Fig. 1B,C), to a precise positioning which depends on bacteroid location in the interzone II-III and zone III. In the two latter zones, the bacteroids located at the cell periphery are between the organelles and the plasmalemma (Fig. 2F). In a similar way, actively dividing bacteroids (type I bacteroids; Vasse et al., 1990) and bacteroids which have ceased to divide and have started to differentiate (type II bacteroids; Vasse et al, 1990), also migrate to the cell periphery where they were seen randomly orientated in the cytoplasm (Fig. 2C,E).

Changes continued in interzone II-III. Here, we noted the exceptional elongation of both mitochondria and plastids, which accumulate very large starch granules in their stroma (Fig. 3A; see Vasse et al., 1990) and we confirmed our observation, made by light microscopy, that the spatial positioning of type III and type IV bacteroids (in interzone II-III and zone III, respectively;Vasse et al., 1990), indeed depends on their location (Fig. 3A, compare with Fig. 1C). Interestingly, the positioning of the outermost bacteroids often correlated with the presence of microtubules orientated parallel to, and often in close contact with, symbiosomes containing nitrogen-fixing type IV bacteroids (Fig. 3B). Such a distribution is progressively lost in proximal zone III (not shown), a nodule region where both plastids and mitochondria gradually lose their elongated form and return to being rod-shaped (Fig. 3C).

Immunolocalization of MC in alfalfa enabled the differentiation features described above to be correlated with variations in cytoskeletal structure. As expected, meristematic cells fluoresced strongly, while a loss in staining due to the progressive disorganization of both endoplasmic and cortical microtubules was observed in cell layers of the prefixing zone II (Fig. 4A-C). Such a MC disorganization continues throughout the entire zone II to the point where only very low level staining can be detected at the periphery of the most proximal cells of this zone (Fig. 4D).

A signal is suddenly recovered at the interzone II-III, a zone which differentiates from cell to cell in indeterminate nodules (Vasse et al, 1990). In this zone, a fluorescent band underlining the cell periphery without interruption (Fig. 4D), indicates that a cortical cytoskeleton has reformed. Moreover, short radial microtubular appendages are seen, which then develop centripetally from the cell wall into the cytoplasm of each invaded cell. As cells mature, the number and the length of these appendages increase significantly. In the cells of zone III, radial appendages are orientated in the cell cytoplasm like the spokes of a bicycle wheel (Fig. 4E), i.e. parallel to the individual outermost bacteroids (compare Figs 1C, 3A and 4E). These appendages which most likely correspond to those observed in close contact with symbiosomes (see Fig. 3B), do not extend to the central region of the invaded cells where bacteroids are randomly oriented in the cytoplasm (compare Figs 1C and 4E). Such an organization of the endoplasmic MC is progressively lost in the most proximal part of the nodule, where a new decrease in fluorescence is observed (Fig. 4F). It is worth mentioning that (i) MC changes described above are not related to the nitrogen-fixing capacity of nodules since similar variations were observed in equivalent zones of Fix⁻ alfalfa nodules elicited by a R. meliloti fixK mutant (Vasse et al., 1990; not shown), and (ii) no labeling was seen in controls where the primary antibody was omitted, thus confirming the specificity of the labeling (not shown).

To examine whether MC changes are general traits of nodule differentiation, we studied MC in various indeterminate and determinate nodules. Changes identical to those described above in alfalfa, i.e. transient microtubule depolymerization followed by a newly patterned reorganization, were observed in clover and vetch indeterminate nodules (not shown). A progressive MC disorganization was also seen in developing determinate siratro nodules (6-10 days after inoculation) from the outer meristematic cells, which fluoresced strongly, to the cells of the central region where labeling could hardly be detected (Fig. 4G). In this legume, MC reformed in the nitrogen-fixing cells of fully differentiated nodules. However the signal was restricted to the cell periphery (Fig. 4H), indicating that the cortical cytoskeleton, but not the endoplasmic cytoskeleton, reorganizes in determinate nodules.

To determine whether cytoskeletal changes are rhizobia-related, we studied the cytoskeleton in bacteria-free nodules induced either by a R. meliloti exoA mutant (Leigh et al., 1985) or by NFs purified from R. meliloti (Truchet et al., 1991) or following spontaneous development (NAR nodules; Truchet et al., 1989). All these nodule types have a uniform central tissue made of bacteria-free cells. In contrast to wild-type nodules, it was found that a predominantly cortical MC remained present throughout the differentiation of these three nodule types, indicating that MC changes require the presence of bacteria in nodules (Fig. 4I).

Our observation that MC changes are rhizobia-dependent (see above) together with results showing that nod genes are transcribed in rhizobia that are still enclosed in infection threads (Schlaman et al., 1991), prompted us to examine whether MC changes could be correlated with the presence of NFs in central nodule cells. Therefore, rabbit polyclonal antibodies directed against purified NFs of R. meliloti were prepared and their reactivity verified by ELISA. Briefly, we found that coated NFs reacted with antiserum in a time- and zone III (Fig. 5D). This could be due to an accumulation in the vacuole of NFs released from bacteroids which undergo a degeneration process in the vacuole of infected cells in this region of the nodule (Truchet and Coulomb, 1973). Finally, and unexpectedly, labeling was seen on starch granules accumulating in amyloplasts and on the cell wall of endodermal cells (Fig. 5A and 5B). This latter labeling, which varied strongly between assays, could be due either to the recognition by the immune serum of epitopes localised at these sites and/or to non-specific labeling. Comparative studies between assays and controls omitting the primary antibodies (not shown), using the preimmune serum (Fig. 5E, compare with 5A), or where immune serum was exhaustively adsorbed with purified NFs prior to the staining procedure (Fig. 5F, compare with 5B) left open the two possibilites. However, in contrast to the labeling in infection threads and cell cytoplasm in zone II, it seems very unlikely that labeling on the endodermis accounts for the presence of NFs, since this tissue is totally bacteria-free (Vasse concentration-dependent manner (data not shown). Out of four sera tested, the one which gave the highest positive signal with a concentration of NFs as low as 10⁻⁹ M, was used for the experiments. We then controlled the specificity of the immune serum by inoculating axenic plants with R. meliloti strains known either to overproduce NFs (strain GMI6390, Roche et al., 1991), or to be unable to synthesize NFs (strain GMI6371, Roche et al., 1991). Immunostaining of plants 2 days post- inoculation showed that the NF-producing strain fluoresced strongly while no signal was detected on alfalfa inoculated with rhizobia which do not synthesize NFs (not shown). This result indicated clearly that the antibodies were specific for NFs and did not immunoreact with plant and bacterial surface components.

NFs were immunolocalized in wild-type alfalfa nodules elicited by the wild type R. meliloti RCR 2011 strain. By light microscopy, the strongest signal was observed in prefixing zone II, particularly associated with the infection threads, while a lower, but still significant labeling, was detected in the cytoplasm of the invaded cells (Fig. 5A-C). In contrast, low levels of staining were found in interzone II-III and zone III (Fig. 5A and 5B). This variation was confirmed by quantifying the density of gold particles per surface unit (µm²), in conditions where the highly reactive infection threads and large plastids were excluded from the scanned regions. We found in prefixing zone II, a higher density than in meristematic zone I and distal nitrogen-fixing zone III, respectively (Fig. 6). Detailed observations showed that NFs are internalized in the cytoplasm of the infected cells in zone II (Fig. 5C). There is also a constant, albeit discrete, level of immunolabeling associated with bacteroids at all stages of their differentiation (Fig. 5D). The labeling of bacteroids might account for the slight, but significant (t=2.8, P<0.025), increase in gold particle density detected in the cells of zone III as compared to the meristematic cells (Fig. 6). Moreover, a signal was also observed in the large vacuoles of the most proximal cells of et al., 1990). NF immunolocalization was confirmed at the ultrastructural level (not shown).

In this paper, we provide experimental evidence of NF internalization and MC architectural rearrangements during nodule differentiation. That similar changes are observed in different nodule types and depend on bacterial release indicate that cytoskeleton changes are symbiosis-specific and rhizobia- dependent. Our results lead to several conclusions and hypotheses concerning the general role that the cytoskeleton plays in mediating plant cell differentiation and suggest that NF internalization and MC changes are correlated during nodule maturation.

In indeterminate nodules, MC changes occur in nodule central zones in which at least six major cellular changes are observed microscopically. The first cellular event which determines the individualization of the prefixing zone II is the penetration of meristem-derived cells by the growing part of the infection network. A number of studies either provide direct evidence or imply that the cytoskeleton is involved in infection thread growth: (i) microtubules are present at the original site of penetration (Ridge and Rolfe, 1985); (ii) microtubules interconnect the host cell nucleus to the growing infection thread tip (Bakhuizen, 1988); (iii) cytoskeleton-determined cytoplasmic bridges guide the growth of infection threads on their way down to the nodule primordium (Van Brussel et al., 1992; Bakhuizen, 1988); (iv) the cytoskeleton plays a major role in cell wall formation (reviewed by Seagull, 1989; Cyr, 1994), and infection threads are cell wall limited structures (Callaham and Torrey, 1981) and (v) an interesting model has been proposed in which Rhizobium infection through root hairs may result from the mobilization of normal root hair tip growth machinery, including the cytoskeleton, at the infection site (Ridge, 1992). Taken together, these and our data indicate that cytoskeletal integrity and functions are sufficiently preserved to sustain infection thread growth at the meristem- prefixing zone II interface of developing nodules.

The second cellular event is the release of bacteria in the cell cytoplasm which is restricted to the distal part of prefixing zone II (Vasse et al., 1990). The signals that trigger bacterial release are not known. However, the fact that release occurs exclusively at unwalled parts of infection threads (Roth and Stacey, 1989), has suggested the possibility of an enzymatic degradation of the thread cell wall either by enclosed rhizobia (Truchet and Coulomb, 1973) or by a mechanism which would involve the mobilization of the endoplasmic reticulum (Roth and Stacey, 1991) which is particularly abundant in the region of bacterial release (Truchet and Coulomb, 1973; this study). Our results open up another possibility. We hypothesize that the release might also be a consequence of discrete and local variations of the endoplasmic cytoskeleton in the direct vicinity of the growing tip of infection threads, i.e. at sites where the developing cell wall is not yet completely organized. As a consequence, some of the cytoskeleton-controlled functions, such as cell wall building (for example), could be locally modified, or slackened, to an extent and for a period of time sufficient to allow bacterial release through an opened door at infection extremities.

The third cellular event observed in prefixing zone II is that, once invaded, host cells enlarge isodiametrically in distal zone II, leading to a round-shaped morphology in the proximal cells of this zone. The cytoskeleton plays an active role in cell elongation and cell shape control by directing the ordered synthesis of cellulose microfibrils via a microtubule/plasma membrane/cell wall functional continuum (reviewed by Seagull, 1989; Giddings and Staehelin, 1991; Williamson, 1991; Cyr, 1994). Thus the isodiametric enlargement of the invaded cells in the distal part of zone II, should indicate that, in this region, the cortical array maintains its function in cell wall building necessary for cell enlargement. In contrast, the changes in both cell shape and morphology in zone II and the arrest in cell enlargement occurring in the central nodule tissues (this study) might result from the progressive lost in cytoskeletal integrity observed in the infected nodule cells. It might seem contradictory to hypothesize that in distal cells of zone II the cytoskeleton still functions efficiently in plant cell wall organization whilst infection thread cell wall formation is impaired. However, as stated above, we propose that local changes in the cytoskeleleton centered around the extremity of an infection branch and which are not discernable by microscopy, locally modify cytoskeleton functioning.

The fourth cellular event is that simultaneously to cell enlargement, endopolyploidy takes place gradually and exclusively in the invaded cells of prefixing zone II in indeterminate pea and alfalfa nodules (Truchet, 1978). The cytoskeleton is subjected to a series of specific rearrangements that are essential for karyokinesis and cytokinesis to proceed successively (reviewed by Seagull, 1989; Goddard et al., 1994). The fact that, in nodules, the impairment of these two cellular events occurs simultaneously to cytoskeletal disorganization raises the question of the mechanisms that are responsible for DNA amplification. Particularly relevant to our study are the results showing that drug-induced microtubule depolymerization is sufficient to initiate both DNA synthesis and the entry of human and animal fibroblast-like cells into the proliferative cycle (Crossin and Carney, 1981; Bershadsky et al., 1996) and that microtubule depolymerization is directly responsible for DNA synthesis (Crossin and Carney, 1981). It would be interesting to determine if a similar mechanism is involved in the establishment of the polyploid gradient in the prefixing zone II of indeterminate nodules.

The fifth cellular event concerns the major variations in the spatial distribution and the positioning of both organelles and bacteroids in the infected cells. The capacity of the cytoskeleton to mediate the movement and distribution of organelles in a cell is supported by several correlative studies (reviewed by Cole and Lippincott-Schwartz, 1995). For example, the distribution of organelles frequently parallel that of the cytoskeleton and the use of drugs which depolymerise microtubules or microfilaments results in modified spatial distribution of organelles (Yaffe et al., 1996). Our data are in agreement with these studies: the random distribution of cell organelles and bacteroids in the most distal cells of zone II is associated with an organized cytoskeleton; the progressive migration of organelles and bacteroids to the periphery of the cells in proximal zone II occurs simultaneously to endoplasmic MC disorganization. Moreover, our study suggests that the positioning of both cell organelles and bacteroids is microtubule-mediated. Thus, in zone III, the peculiar positioning of the outermost bacteroids, perpendicular to the cell wall, parallels that of radial microtubules originating from the cortical cytoskeleton and radiating centripetally to inner regions of the cytoplasm. In plants, a radial cytoskeleton pattern has been described in meristematic and elongating cells (Bakhuizen et al., 1985; Baluska et al., 1992), and in the syncytial stage during endosperm development (Brown et al., 1994). Generally, radial patterns are thought to originate from the nuclear periphery (Lambert, 1993) and to provide an anchor for the nucleus at its appropriate position during cell differentiation. Similar functions for the radial cytoskeleton in nitrogen-fixing cells are unlikely. In nodules, radial arrays obviously originate in the cell cortex, then elongate into the cell cytoplasm as cells mature and are not connected to other structures in the cytoplasm. Thus, we provide further evidence for the presence of microtubule organizing centers (MTOCs) in the cortex of plant cells (Liu et al., 1994; Panteris et al., 1995). The situation is different in mature determinate nodules in which the random distribution of large symbiosomes containing many bacteroids in the cell cytoplasm (Newcomb, 1981) is in agreement with our data showing that a cytoplasmic cytoskeleton does not reform in the invaded cells of nitrogen- fixing siratro nodules.

The sixth cellular event concerns the ultrastructural changes which characterize cell organelles and bacteroids in invaded cells. By correlating the present data with our previous results on bacteroid differentiation in alfalfa nodules (Vasse et al., 1990), it clearly appears that bacteroids (i) divide exclusively in the cells of distal zone II, where the MC is microscopically observed, (ii) stop dividing and elongate moderately in the second half of the proximal zone II where the MC disorganizes and (iii) increase dramatically in size in a cell to cell fashion (Vasse et al., 1990), correlating with the differentiation of the interzone II-III where cortical MC reforms. Interestingly, we found that cell organelles undergo similar variations in the same regions of the nodule, being small-sized in distal zone II, increasing moderately in size in proximal zone II and reaching their maximal size in interzone II-III, before returning to a smaller size in zone III. The mechanisms which control the division of cell organelles are poorly understood, although it is generally admitted that there is no link between the mechanism that divides the mother cell and the dividing organelle (reviewed by Warren and Wickner, 1996). Despite the difficulty in deducing a dynamic relationship simply from morphological features, our results strongly suggest that the MC plays a role in bacteroid and cell organelle division and elongation. This hypothesis implies that bacteroids are perceived as endogenous organelles. Microtubules mediate the distribution of cell organelles via membrane-binding microtubule-based motor proteins such as kinesin and cytoplasmic dynein (Yaffe et al., 1996). As bacteroids are enclosed by a plasmalemma-derived peribacteroid membrane (Roth and Stacey, 1989), we anticipate that the host cell senses intracellular bacteroids as normal cell organelles via the connection of cytoskeleton microtubular elements and associated proteins with the peribacteroid membrane.

In this paper, we show that NFs are immunolocalized throughout the central tissues of alfalfa nodules from the prefixing zone II to the proximal part of zone III. The fact that the strongest signal is observed in infection threads is in agreement with previous studies indicating that nod genes are expressed in bacteria enclosed in infection threads (Schlaman et al., 1991; F. Maillet, G. Truchet and J. Dénarié, unpublished data). Moreover, in zone II, a specific signal is also detected in the cytoplasm of the infected plant cells indicating that NFs are internalized. Finally, a weak but reproducible labeling is associated with bacteroids, independent of their differentiation stage and histological location. This result is in contradiction with previous data indicating that in vetch and alfalfa nodules, nod genes are not expressed in intracellular bacteroids (Schlaman et al., 1991; F. Maillet, G. Truchet and J. Dénarié, unpublished data) and suggests that NFs are synthesized at a low level by bacteroids.

Our data and the recent results showing that R. leguminosarum bv. trifolii NFs are internalized in clover root hairs (Philip-Hollingsworth et al., 1997) indicate that NF internalization can take place in different tissues amenable to infection. At the moment, we do not know how NF internalization occurs. It might involve a simple diffusion process, or be receptor-mediated. Putative high-affinity NF receptors could be located in the plasma membrane (Bono et al., 1995; Niebel et al., 1997), a membrane which also surrounds infection threads and bacteroids (Hirsch, 1992; Kijne, 1992; see Fig. 2B). The possible coupling of NF reception and internalization is also supported by the recent report showing that purified R. leguminosarum bv. trifolii NFs are internalized into the root hairs of clover, the homologous host, but not of alfalfa, an heterologous host for this species (Philip- Hollingsworth et al., 1997). Finally, our observation that NFs are only internalized in the distal region of alfalfa nodules whilst they are synthesized by bacteroids throughout their entire differentiation process, suggests that the mechanism(s) directing NF internalization change(s) during nodule maturation.

Identifying the nature of the causal mechanism(s) resulting in cytoskeletal disorganization is a major challenge raised by this study. Our data suggest that NF internalization and MC disorganization are coupled events. Firstly, NFs are immunolocalized in the cytoplasm of invaded cells of the prefixing zone II, where the MC disorganizes, but not in the cytoplasm of cells in nitrogen-fixing zone III, where the cytoskeleton reforms. This result highlights the spatiotemporal correlation between the two processes. Secondly, the MC is not disorganized in bacteria-free alfalfa nodules indicating that MC disorganization requires the presence of bacteria. Thirdly, identical MC disorganization occurs in nodules of different types induced by rhizobial strains which are taxonomically distant (Martinez-Romero and Caballero-Mellado, 1996), but share the capacity to produce NFs belonging to the same chemical family and displaying similar biological activities (reviewed by Dénarié et al., 1996; Long, 1996; Schultze and Kondorosi, 1996; Spaink, 1996). From these considerations, it can reasonably be hypothesized that the causal mechanism(s) responsible for cytoskeletal disorganization in nodules is common to all rhizobia and that NFs, or a structural component of NFs, are (is) this common link.

In summary, this paper reports several correlations between NF internalization, MC changes and the particular differentiation of the invaded central cells in alfalfa nodules. We propose that, once internalized, rhizobial NFs are involved in cytoskeletal changes accompanying cell differentiation. A validation of this model requires a detailed study of the direct effects of purified NFs on cytoskeletal architecture.

We are very grateful to Fabienne Maillet and Jean Dénarié for providing purified R. meliloti Nod factors, Philippe Cochard for advice in using laser confocal microscopy and Michel Petitprez for help in gold label quantification. We thank very much D. Barker, P. Boistard and J. Dénarié for their constructive comments regarding the manuscript and C. Gough and D. Barker for English reviewing. A. C. J. T. was supported by BIOTECH and TMR postdoctoral fellowships and this work was funded by grants from the European Community BIOTECH Programme (PTP CT93-0400) and TMR Programme (FMRX-CT96-0039).