Monoclonal antibodies were used as cytochemical markers to study surface interactions between endosymbiotic Rhizobium bacteroids from pea root nodules and the encircling peribacteroid membranes, which are of plant origin. Monoclonal antibodies that react with Rhizobium lipopolysaccharide (LPS) or with a plant membrane glycoprotein were used as markers for material from the bacteroid outer membrane or the peribacteroid membrane, respectively.

Membrane-enclosed bacteroids were isolated from nodule homogenates by sucrose gradient centrifugation, and the encircling peribacteroid membrane was released by mild osmotic shock treatment. Using an immunochemical technique (sandwich ELISA), it was shown that 1–5 % of the LPS antigen released into the peribacteroid fraction by mild osmotic shock treatment was physically associated with peribacteroid membrane through a detergent-sensitive linkage. This association could be visualized when freshly prepared peribacteroid material was immobilized on gold grids and examined by electron microscopy after dual antibody immunogold treatment and subsequent negative staining. The distribution of LPS antigen within infected nodule cells was also investigated by immunogold staining for thin sections of nodule tissue fixed in glutaraldehyde, and a close association between LPS antigen and peribacteroid membrane was often seen.

Nitrogen-fixing pea root nodules contain large numbers of Rhizobium bacteroids within the cytoplasm of infected cells of the nodule cortex. Each bacteroid is individually enveloped by a peribacteroid membrane (Robertson & Lyttleton, 1984), which shows many similarities to the plant cell plasma membrane (Brewin et al. 1985; Blumwald et al. 1985). The initial infection of the plant cell by Rhizobium involves endocytosis of bacteria from an infection thread, which itself represents an involution of the plant cell plasma membrane. Thereafter, the intracellular bacteria continue to grow and divide until the differentiated bacteroids occupy a large fraction of the cytoplasmic volume. New membrane material is supplied to the enlarging peribacteroid membranes by fusion of membrane vesicles derived from the Golgi apparatus (Brewin et al. 1985; Robertson et al. 1978 a, b), and every time the rhizobia divide there must be some mechanism for inducing a concomitant division of the peribacteroid membrane sac. It has been postulated (Robertson & Lyttleton, 1984) that surface interactions between peribacteroid membranes and the bacteroid envelope outer membranes could be responsible for division of the peribacteroid membrane. Similarly, adhesion between the plant cell plasma membrane and the surface of rhizobia within infection threads could provoke the endocytotic event that initiates intracellular infection.

We have recently isolated monoclonal antibodies (McAb) that react either with a glycoprotein component of the peribacteroid membrane (Brewin et al. 1985), or with lipopolysaccharide from the bacteroid outer membrane (Brewin et al. 1986). In the present study, these McAb were used as immunocytochemical markers to test whether or not there is any form of association between these two membrane systems, either in nodule homogenates or in nodule thin sections.

Fractionation of nodules

Pea nodules containing R. legutninosarum strain 3841 were homogenized at 4°C in a pestle and mortar in Tris-DTT buffer (50mM-Tris-HCl (pH7·5), 10mM-dithiothreitol) containing 0·5M-sucrose, and bacteroids still enclosed by peribacteroid membrane were prepared by fractionation through sucrose cushions in microcentrifuge tubes as described (Brewin et al. 1985). The membrane-enclosed bacteroids were then subjected to mild osmotic shock treatment (50 mM-Tris-DTT buffer without sucrose), centrifuged for 2min at 10000 g to remove the isolated bacteroids, and for a further 1 min to remove cell wall debris. The resulting supernatant, termed the peribacteroid fraction, was used directly for experiments.

In the membrane mixing experiments, R. legutninosarum strain 3624 (TOM-sZr; Brewin et al. 1986) was used to inoculate peas in order to prepare peribacteroid material that did not react with MAC57 antibody. In addition, the bacteroids isolated from pea nodules that had been inoculated with strain 3841 were purified by centrifugation through a sucrose gradient, washed in 50 mM-Tris-DTT buffer and sonicated as previously described (Brewin et al. 1985).

Immunochemistry

Rat monoclonal antibodies AFRC MAC64 and AFRC MAC57 (Brewin et al. 1985, 1986) react with a glycoprotein component of the peribacteroid membrane and a lipopolysaccharide from the bacteroid outer membrane, respectively. Immunoglobulin was purified from ascitic fluid derived from the corresponding cell lines by precipitation with 50 % saturated ammonium sulphate, followed by euglobulin precipitation (i.e. dialysis against low ionic strength buffer, 2mM-sodium phosphate, pH6’5). The purified immunoglobulin M (IgM) antibodies (≈10mgml-1 and more than 70% pure) were conjugated to biotin using the recommendations of the manufacturers (Amersham International).

ELISA assays

The ELISA method was derived from that of Engvall & Perlmann (1972). All incubation and washing steps were conducted at 4°C. Microtitre plates (Dynatech M129A, immulon) were coated with 50 μl antigen at an appropriate dilution in phosphate-buffered saline (PBS) containing 0·01 % (w/v) sodium azide. After incubation overnight, the plates were washed three times in cold tap water and uncoated binding sites were blocked non-specifically by further incubation for 1 h in the presence of 150 μd foetal calf serum (10% (v/v) in PBS-azide). After further washing, the wells were incubated for 1 h with 50 μl of appropriate probing antibody (MAC64 or MAC57 ascitic fluid, used at a final dilution in foetal calf serum of about 1 in 10000 or 1 in 50000, respectively). The wells were then washed and treated for 45 min with 50 μl of horseradish peroxidase conjugated to anti-rat IgG antibody (from Miles Laboratories, Slough), used at 1 in 2000 dilution in 10% (v/v) foetal calf serum. The wells were then washed for 3 min in PBS containing 0·05 % (v/v) NP40 detergent, followed by 10 washes in cold tap water before the addition of 150 μl of chromogenic substrate (0’1 mg ml-1 tetramethylbenzidine (TMB) in 0-lM-sodium acetate, pH 6’0, containing 0·002 % (v/v) H2O2). The reaction was left at room temperature for up to 1 h, stopped with 20 ·l of 2M-H2SO4, and screened at 450 nm using a Titertek Multiskan MC plate scanner.

For sandwich ELISA experiments (Fig. 1), microtitre plates were first coated by overnight incubation at 4°C with either MAC64 or MAC57 antibody (purified from ascitic fluid and appropriately diluted so as to saturate all available binding sites). Washing and blocking steps then proceeded as for normal ELISA, and subsequently 50 μl of peribacteroid material was added at appropriate dilutions in 10% (v/v) foetal calf serum in PBS-azide, with or without 0·5% (v/v) NP40. After incubation for 2h, the plates were washed three times in cold tap water, and then treated for 2h with 50/d of biotinylated antibody (MAC64 or MAC57) at appropriate dilution to saturate the reaction. After three more washes the wells were incubated for 40 min with 50 μl horseradish peroxidase conjugated to streptavidin (Amersham International) used at 1 in 1000 dilution in foetal calf serum in PBS. After incubation there was a 1 min wash with 0·05 % (v/v) NP40, followed by 10 washes in cold tap water before adding chromogenic substrate.

Electron microscopy

The peribacteroid fraction released from membrane-enclosed bacteroids by mild osmotic shock treatment was immediately mounted on carbon/parlodion-coated grids and sucrose was removed using the wick elution method (Webb, 1973). The grids, coated with 5—10μ1 of sample, were treated for 3 min using water as eluent and transferred immediately to blocking solution (20 mg ml-1 bovine serum albumin (BSA) in 10mM-Tris-HCl, pH7-4, containing 0·02% (w/v) sodium azide, 0·9% (w/v) NaCl, 0·5mgml-1 polyethylene glycol, 20 K) without being allowed to dry. Thereafter, the peribacteroid material immobilized on the grids was labelled using monoclonal antibodies and immunogold reagents as previously described for thin sections (Robertson et al. 1985), except that antibody incubations were for 8–16 h at 4°C and washing was performed using a stream of 3 ml buffer from a pipette (repeated three times over a period of 2min). For dual immunogold labelling experiments the material was first treated with the non-biotinylated antibody followed by colloidal gold (10 nm) anti-rat IgG and, secondly, with the biotinylated antibody followed by streptavidin conjugated to 15 nm gold particles. After a final wash with water, the samples were negatively stained using 2% (w/v) methylamine tungstate (Faberge & Oliver, 1974) and dried before being examined under a Siemens Elmiskop IA electron microscope.

Nodule segments containing R. leguminosarum strain 3841 were fixed in glutaraldehyde and osmium tetroxide, and embedded in LR white resin. Thin sections, after pre-treatment with saturated sodium metaperiodate-HCl to remove surface osmium tetroxide, were treated with antibodies as described previously (Brewin et al. 1985, 1986) and post-stained to enhance contrast, using saturated uranyl acetate (10min) and Reynold’s lead citrate (2min). Immunogold reagents were obtained from Jannsen Pharmaceutical and were EM grade.

Sandwich ELISA experiments

Following mild osmotic shock treatment of membrane-enclosed bacteroids, about 15—20 % of all the lipopolysaccharide (LPS) antigen reacting with MAC57 remained in the supernatant (peribacteroid) fraction after centrifugation. Using the experimental design outlined in Fig. 1, it was possible to investigate whether any or all of the released LPS antigen was associated with MAC64 antigen, i.e. the plant-derived glycoprotein found in the peribacteroid membrane. The data presented in Fig. 2A show that, when bacteroid membrane-derived LPS material was retained on an ELISA plate by immobilized anti-LPS antibody (MAC57), there was concomitant retention of plant-derived glycoprotein material, which could be detected by a coupled streptavidin-immunoperoxidase system involving biotinylated MAC64. Similarly, evidence for association between glycoprotein and LPS material could be obtained using immobilized MAC64 as the first antibody and biotinylated MAC57 in the second stage (Fig. 2D). However, if neither MAC57 nor MAC64 was immobilized on the microtitre plates before blocking with foetal calf serum, neither glycoprotein nor LPS was retained from the peribacteroid material, and there was no subsequent reaction either with biotinylated MAC64 or with biotinylated MAC57.

Preincubation of the peribacteroid material with detergent (0·5 % (v/v) NP40) for 1 h at 4°C eliminated the apparent association between plant glycoprotein and LPS material (Fig. 2A,D), without inhibiting the individual antibody-antigen reactions. Fig. 2B shows that, for the epitope of MAC64, which is probably reiterated within the glycoprotein molecule, detergent treatment did not eliminate the sandwich ELISA reaction involving immobilized MAC64 and biotinylated MAC64. However, in the case of the epitope of MAC57, Fig. 2C shows that detergent treatment eliminated the sandwich ELISA reaction involving immobilized MAC57 and biotinylated MAC57. Further experiments showed that this concentration of NP40 did not inhibit the MAC57-LPS reaction in dot immunoassays, simple ELISA assays, or on immunoaffinity columns (data not shown). The most likely interpretation of these results is that NP40 dissociated bacteroid membrane fragments into individual LPS molecules, each of which had only a single epitope for MAC57, and which therefore could not function in a sandwich ELISA system.

The proportion of all LPS antigen that was associated with pZwz-derived glycoprotein was estimated by measuring the proportion of LPS antigen withdrawn from a solution as a result of association with immobilized MAC64 antibody in a microtitre well. A dilution of peribacteroid material (2 μg protein ml-1) that was not saturating for the immobilized MAC64 was selected (Fig. 2D). After the normal incubation period, the peribacteroid material was withdrawn from the well and serial dilutions were transferred to a new microtitre dish and assayed by the ELISA assay using MAC57 ascitic fluid (1 in 50000 dilution). These measurements of LPS concentration were compared with those for a dilution series from a control sample that had not been exposed to immobilized MAC64 during the original incubation period. The differences between these two sets of measurements were very small and prone to relatively large error fluctuations. However, the best estimates were that between 1 and 5 % of the LPS antigen had probably been withdrawn as a result of association with MAC64 antigen.

Mixing experiments

In order to investigate whether or not the observed association between bacteroid LPS and plant membrane glycoprotein material was simply the result of mixing these two components in the nodule homogenates, peribacteroid material was prepared from pea nodules that had been inoculated with R. leguminosarum strain TOM (3624). It has already been reported (Brewin et al. 1986) that LPS from bacteroids of this strain does not react with MAC57 antibody, and these results were confirmed (Fig. 3A-D). Testing serial dilutions of 3624 peribacteroid material in sandwich

ELISA assays, there was no evidence of any MAC57 antigen being present; the only positive signal was with a MAC64-biotinylated MAC64 sandwich (Fig. 3D). Hence this material could be mixed with a preparation of sonicated bacteroids derived from strain 3841 in order to look for evidence of subsequent association between MAC57 and MAC64 antigens as revealed by the sandwich ELISA assay. As expected, the sonicated 3841 bacteroid preparation only reacted positively in a sandwich assay involving MAC57 and biotinylated MAC57. When preparations of 3841 sonicated bacteroids and 3624 peribacteroid material were incubated together for 2 h at room temperature and 2h at 4°C no evidence for LPS-plant glycoprotein association was observed (Fig. 3E,F).

In another series of experiments the peribacteroid material derived from nodules containing strain 3841 bacteroids was frozen and thawed successively six times without any detectable effect on the degree of association between LPS and MAC64 antigens observed in the sandwich ELISA system. This provided further evidence that the observed association was not an artifact of the nodule homogenization procedure.

Dual-labelling experiments with immunogold

Using immunochemical systems coupled to 10 nm and 15 nm colloidal gold particles, it was possible to visualize by electron microscopy the physical interactions between peribacteroid membrane-derived glycoprotein and bacteroid-derived LPS, which had already been detected by the sandwich ELISA experiments described in Fig. 2. The peribacteroid material, released from membrane-enclosed bacteroids by mild osmotic shock treatment, was immobilized on gold grids and treated first with MAC64 and 10 nm colloidal gold conjugated to goat anti-rat IgG, followed by biotinylated MAC57 and 15 nm colloidal gold conjugated to streptavidin. This duallabelling system gave clear discrimination between the two types of membrane material and negative staining with methylamine tungstate also gave some indication of surface morphology for these membranes (Figs 4, 5).

LPS from the bacteroid outer membrane, which stained with MAC57 and immunogold, was usually seen either as fragments of membrane sheets (Figs 4, 7, 9) or more commonly as small vesicles, approximately 100 nm in diameter (Figs 6, 9), which were sometimes aggregated as chains (Fig. 8). In Fig. 4, an unusually well-preserved bacteroid outer membrane ghost is shown, in which the LPS takes the form of a latticework that protrudes from the surface of the bacteroid membrane. The peribacteroid membrane fragments, which stained with MAC64 and immunogold, characteristically had a complex multilayered structure (Figs 57) and were approximately 1–2μm in diameter. Sometimes these peribacteroid membrane fragments were joined together in a mosaic structure (not shown).

Immunogold labelling indicated that the large majority of peribacteroid membrane fragments were not associated with bacteroid membrane and vice versa (Figs 4, 5, 9). However, a small proportion of the immunogold label associated with bacteroid outer membrane fragments (perhaps 1 %) was closely associated with pbm material. (Examples of bom-pbm membrane associations are illustrated in

Figs 610). The most common form of association involved very small fragments or vesicles of bacteroid membrane (Figs 6, 7, 9) but, more rarely, pbm was entirely encircled by bacteroid membrane (Fig. 8), or bacteroid membrane and pbm were closely interspersed (Fig. 10).

Immunogold staining of nodule thin sections

The in situ distribution of LPS antigen within infected nodule cells was examined by immunogold staining of nodule thin sections in order to investigate possible interactions between material derived from the bacteroid outer membrane and the peribacteroid membrane. As expected, most of the gold label associated with MAC57 binding was distributed along the bacteroid surface (Fig. 11), but the LPS was patchy in distribution and often appeared to extend beyond the bacteroid outer membrane into the peribacteroid space. Clusters of LPS antigen were also found in close association with the peribacteroid membrane, even where this membrane was separated from the bacteroid outer membrane by a substantial peribacteroid space. More rarely, clusters of LPS antigen were seen in cytoplasmic membranes of infected plant cells and very occasionally in adjacent plant cell wall material (Fig. 11). Clusters of gold particles present in the plant cell wall after immunogold treatment with MAC57 were observed on three separate occasions during an extensive survey involving over 100 electron micrographs taken from 20 different grids and six different immunogold staining experiments. Immunogold staining of plant cell membranes or walls from uninfected cells was never observed with MAC57 antibody, and so the possibility of a cross-reacting antigen of plant origin would seem to be very unlikely.

During the fractionation of nodule homogenates it was noted that approximately 15% of the lipopolysaccharide antigen from membrane-enclosed bacteroids was released into the peribacteroid fraction following mild osmotic shock treatment. A combination of immunogold and negative staining revealed that, in general, this released antigen was in the form of very small vesicles or fragments of bacteroid membrane (Figs 6, 7), although very occasionally larger structures and bacteroid outer membrane ghosts were seen (Figs 4, 8), presumably the result of damage to whole bacteroid cells during the osmotic shock treatment.

By entrapping bacteroid-derived LPS on ELISA plates using immobilized MAC57 antibody it was found (Fig. 2) that some of the LPS antigen present in the peribacteroid fraction was physically associated with a peribacteroid membrane-derived glycoprotein that reacted with biotinylated MAC64 antibody. This association was maintained in phosphate-buffered saline containing 10mM-EDTA (data not shown), but was dissociated by detergent treatment (Fig. 2). Although quantitative estimates are complicated by competitive binding effects between the various membrane species, and by variations in the specific activity of the biotinylated probes used in the sandwich ELISA assays, the best indications are that 1–5 % of the LPS antigen released as bom fragments into the peribacteroid fraction was in some way physically associated with peribacteroid membrane. These results were consistent with a visual examination by electron microscopy (Figs 610) of freshly prepared peribacteroid material, which showed that bom andpbm membrane fragments were often found in close association with each other.

The important question to be asked is whether this association between bom and pbm material was simply an artifact of the nodule homogenization procedure, or whether it represents an interaction that might have functional significance in terms of the physical interactions between bacteroids and peribacteroid membrane. The in vitro mixing of bacteroid and peribacteroid membranes did not result in any detectable intermembrane associations (Fig. 3E,F), and freezing and thawing of peribacteroid material did not affect the degree of association observed in sandwich ELISA experiments. However, the evidence from nodule thin sections (Fig. 11) indicated that LPS antigen could often be found in close association withpfwz, even in areas where the peribacteroid membrane and bacteroid cell surface membrane were not closely apposed. These data are reminiscent of earlier evidence by Bal et al. (1980) and Bal & Wong (1982) that the bacteroid outer membrane may be ‘sloughed off’ during the process of cell wall differentiation. However, in the present study (Fig. 11) the data indicate that LPS antigen may often be part of a chain-like structure that extends back from the pbm across the peribacteroid space and still remains anchored in the peribacteroid outer membrane. Perhaps these chains are related to the lattice-like arrangement of LPS antigen seen on the surface of bacteroid outer membrane ghosts by negative staining (Fig. 4).

A physical association between pbm and the bacteroid surface could account for the concomitant division of peribacteroid membranes with intracellular bacteroids observed by Robertson & Lyttleton (1984). Furthermore, as discussed by these authors, variations in the strength of the bom—pbm surface interactions in different legume species could account for the observed differences in the numbers of bacteroids contained within a single pbm envelope. In the case of Agrobacterium (Krens et al. 1985), the lipopolysaccharide fraction of the cell envelope is apparently involved in attachment to the plant cell surface (Banerjee et al. 1981), and the same may be true or Rhizobium bacteroids (Brewin et al. 1986). A morphological analysis of pea nodules induced by LPS mutants of Rhizobium could help to answer this question, and the molecular basis for the interaction between pbm and LPS antigen present in peribacteroid material could also be investigated biochemically using the sandwich ELISA technique described here (Figs 1, 2).

Some evidence was also obtained in the present study that LPS antigen could occasionally pass out of the confines of the peribacteroid space and circulate in the plant membrane system or even be deposited in the plant cell wall (Fig. 11). Perhaps LPS and other material from the peribacteroid space is engulfed by endocytosis of the peribacteroid membrane, or perhaps LPS in the form of membrane vesicles derived from the bacteroid surface actually fuses into the peribacteroid membrane (see Fig. 10). It is currently impossible to distinguish between these two possibilities, but in either case the results may indicate an interesting form of transport and communication between the bacteroid and the cytoplasm of the infected plant cell.

We are indebted to B. Wells and G. J. Hills for advice on negative staining techniques, to T. Bisseling (Wageningen) for useful discussions and Anne Williams for typing the manuscript.

Bal
,
A. K.
,
Shantharam
,
S.
&
Verma
,
D. P. S.
(
1980
).
Changes in the outer cell wall of Rhizobium during development of the root nodule symbiosis in soybean
.
Can. J. Microbiol
.
23
,
573
582
.
Bal
,
A. K.
&
Wong
,
P. P.
(
1982
).
Infection process and sloughing off of rhizobial outer membrane in effective nodules of lima bean
.
Can. J. Microbiol
.
28
,
890
896
.
Banerjee
,
D.
,
Basu
,
M.
,
Choudhurry
,
I.
&
Chatterjee
,
G. C.
(
1981
).
Cell surface carbohydrates of Agrobacterium tumefaciens involved in adherence during crown gall tumour initiation
.
Biochem. biophys. Res. Commun
.
100
,
1384
1388
.
Blumwald
,
E.
,
Fortin
,
M. G.
,
Rea
,
P. A.
,
Verma
,
D. P. S.
&
Poole
,
R. J.
(
1985
).
Presence of host-plasma membrane type H+-ATPase in the membrane envelope enclosing the bacteroids in soybean root nodules
.
Pl. Physiol
.
78
,
605
612
.
Brewin
,
N. J.
,
Robertson
,
J. G.
,
Wood
,
E. A.
,
Wells
,
B.
,
Larkins
,
A. P.
,
Galfre
,
G.
&
Butcher
,
G. W.
(
1985
).
Monoclonal antibodies to antigens in the peribacteroid membrane from Rhizobium-induced root nodules of pea cross-react with plasma membranes and Golgi bodies
.
EMBOJ
.
4
,
605
611
.
Brewin
,
N. J.
,
Wood
,
E. A.
,
Larkins
,
A. P.
,
Galfre
,
G.
&
Butcher
,
G. W.
(
1986
).
Analysis of lipopolysaccharide from root nodule bacteroids of R. leguminosarum using monoclonal antibodies
.
J. gen. Microbiol. 132 (in press)
.
Engvall
,
E.
&
Perlmann
,
P.
(
1972
).
Enzyme-linked immunosorbent assay, ELISA
.
J. Immun
.
1
,
129
135
.
Faberge
,
A. C.
&
Oliver
,
R. M.
(
1974
).
Methylamine tungstate, a new negative stain
.
J. Microsc
.
20
,
241
246
.
Krens
,
F. A.
,
Molenduk
,
L.
,
Wullems
,
G. J.
&
Schilperoort
,
R. A.
(
1985
).
The role of bacterial attachment in the transformation of cell-wall-regenerating tobacco protoplasts by Agrobacterium tumefaciens
.
Planta
166
,
300
308
.
Robertson
,
J. G.
&
Lyttleton
,
P.
(
1984
).
Division of peribacteroid membrane in root nodules of white clover
.
J. Cell Sci
.
69
,
147
157
.
Robertson
,
J. G.
,
Lyttleton
,
P.
,
Bullivant
,
S.
&
Grayston
,
G. E.
(
1978a
).
Membranes of lupin root nodules. I. The role of Golgi bodies in the biogenesis of infection threads and peribacteroid membranes
.
J. Cell Sci
.
30
,
129
150
.
Robertson
,
J. G.
,
Warburton
,
M. P.
,
Lyttleton
,
P.
,
Fordyce
,
A. M.
&
Bullivant
,
S.
(
1978b
).
Preparation and properties of peribacteroid membranes and bacteroid envelope inner membranes from developing lupin nodules
.
J. Cell Sci
.
30
,
151
174
.
Robertson
,
J. G.
,
Wells
,
B.
,
Brewin
,
N. J.
&
Williams
,
M. A.
(
1985
).
Immunogold localisation of cellular consituents in the \egyme—Rhizobium symbiosis
. In
Oxford Reviews of Plant Molecular and Cell Biology
, vol.
2
(ed.
B. J.
Miflin
), pp.
69
-
89
. Oxford: OUP.
Webb
,
M. J. W.
(
1973
).
A method for the rapid removal of sugars and salts from virus preparations on electron microscope grids
.
J. Microsc
.
98
,
109
111
.