Role of Symbiotic Auxotrophy in the Rhizobium-Legume Symbioses

Background Rhizobium leguminosarum bv. viciae mutants unable to transport branched-chain amino acids via the two main amino acid ABC transport complexes AapJQMP and BraDEFGC produce a nitrogen starvation phenotype when inoculated on pea (Pisum sativum) plants [1], [2]. Bacteroids in indeterminate pea nodules have reduced abundance and a lower chromosome number. They reduce transcription of pathways for branched-chain amino acid biosynthesis and become dependent on their provision by the host. This has been called “symbiotic auxotrophy”. Methodology/Principal Findings A region important in solute specificity was identified in AapQ and changing P144D in this region reduced branched-chain amino acid transport to a very low rate. Strains carrying P144D were still fully effective for N2 fixation on peas demonstrating that a low rate of branched amino acid transport in R. leguminosarum bv. viciae supports wild-type rates of nitrogen fixation. The importance of branched-chain amino acid transport was then examined in other legume-Rhizobium symbioses. An aap bra mutant of R. leguminosarum bv. phaseoli also showed nitrogen starvation symptoms when inoculated on French bean (Phaseolus vulgaris), a plant producing determinate nodules. The phenotype is different from that observed on pea and is accompanied by reduced nodule numbers and nitrogen fixation per nodule. However, an aap bra double mutant of Sinorhizobium meliloti 2011 showed no phenotype on alfalfa (Medicago sativa). Conclusions/Significance Symbiotic auxotrophy occurs in both determinate pea and indeterminate bean nodules demonstrating its importance for bacteroid formation and nodule function in legumes with different developmental programmes. However, only small quantities of branched chain amino acids are needed and symbiotic auxotrophy did not occur in the Sinorhizobium meliloti-alfalfa symbiosis under the conditions measured. The contrasting symbiotic phenotypes of aap bra mutants inoculated on different legumes probably reflects altered timing of amino acid availability, development of symbiotic auxotrophy and nodule developmental programmes.


Introduction
The largest input of available nitrogen into the biosphere comes from biological reduction of atmospheric N 2 to ammonium [3], with legume-Rhizobium symbioses responsible for much of this. These symbioses arise from infection of host plants, mainly of the legume family, by proteobacteria and results in root structures called nodules [4]. The symbiosis is initiated by flavonoids released from the plant roots which in return elicit the synthesis of lipochitooligosaccharide Nod factors by the rhizobial partner. Bacterial cells are trapped by curling root hairs and proliferate in an infection focus. This turns into an infection thread that grows into the root cortex where a zone of meristematic cells is newly induced. Here the bacteria are released from the infection threads into the plant cells by endocytosis and surrounded by a plant derived symbiosome membrane [5]. The whole structure develops into a nodule with plant cells filled with thousands of symbiosomes. Nodules can be classified as determinate or indeterminate, with those in the Inverted-Repeat-Lacking Clade (IRLC) of the legume subfamily Papilionoideae (such as Medicago, Pisum or Vicia) being indeterminate with a cylindrical shape and a persistent meristem at the tip of the nodule. Bacteria inside symbiosomes differentiate into N 2 fixing bacteroids and this is accompanied by dramatic increases in size, shape and DNA content [6]. The differentiation is caused by a family of plant peptides that share similarities with antimicrobial peptides and probably interfere with the bacterial cell cycle [7]. Nodules of the Milletioides clade (such as Phaseolus, Vigna or Glycine), form spherical determinate nodules with a transient meristem; bacteria do not undergo endoreduplication and therefore do not enlarge substantially. These bacteroids retain a normal DNA content and are largely viable after isolation from nodules [6]. However, other combinations also exist, including enlarged bacteroids in determinate nodules and non-enlarged bacteroids in indeterminate nodules, with the evolutionary newer form of enlarged bacteroids evolving at least 5 times independently in the legume family [8].
The shared metabolism between the symbiotic partners has been proposed to be a simple nutrient exchange of dicarboxylates for ammonium in the nodule [9,10,11], but amino acid transport by bacteroids has also been shown to be essential in pea (Pisum sativum) nodules [1]. R. leguminosarum bv. viciae strains contain two broad specificity amino acid ABC (ATP binding cassette) transporters, AapJQMP and BraDEFGC. AapJQMP consists of a solute binding protein (SBP) AapJ, two permease units AapQM and an ABC subunit dimer formed by AapP for ATP hydrolysis. The BraDEFGC complex consists of SBP BraC, permeases BraDE and ABC subunits BraFG. The aap bra double mutants, RU1357 and RU1722, formed pea bacteroids that appeared morphologically normal in electron micrographs and fixed nitrogen per bacteroid at wild-type levels, but the plants were nitrogen starved [1]. However, aap bra mutant bacteroid protein levels per plant were reduced to ,30% of wild-type. The reason for the nitrogen starved phenotype was later identified to be a limitation of branched-chain amino acid biosynthesis by developing bacteroids [2]. Branched-chain amino acid biosynthesis is transcriptionally reduced during bacteroid development and results in symbiotic auxotrophy where amino acids must be supplied by the plant.
Here we provide evidence that only a low rate of branchedchain amino acid transport is required to overcome symbiotic auxotrophy of R. leguminosarum bv. viciae strains in pea nodules. This shows that large quantities of amino acid are not needed and is consistent with symbiotic auxotrophy. Additionally we report the symbiotic phenotypes of an aap bra mutant in R. leguminosarum bv. phaseoli inoculated on French bean (Phaseolus vulgaris) forming determinate nodules and Sinorhizobium meliloti on alfalfa (Medicago sativa) forming indeterminate nodules.

AapQ site directed mutants with altered transport rates
Previous bioinformatic analysis suggested that a conserved Nterminal domain of the polar amino acid transporting permease proteins (PAAT family of ABC transporters) such as AapQ and AapM might differentially affect the uptake of amino acids [12]. Residue 53 of this domain (residue 144 in AapQ) is strongly correlated with the charge of the transported solute. Thus transporters of basic amino acids have asp or glu at residue 53, while acid transporters have a branched-chain amino acid and broad specificity transporters (e.g. AapQ) have proline. In order to investigate this, a number of site directed mutants were isolated in AapQ and screened for transport specificity by expression from a plasmid in the aap bra null mutant RU1722 (data not shown). Of these mutants P144D proved most interesting, so a single copy of AapQ P144D was generated by recombination into the chromosome of Rlv3841. This was achieved by replacing the VSp cartridge in RU1722 aapJQM::VSp braEF::VTc with a full copy of the aapJQM operon containing the mutated AapQ P144D, generating RU1976. The braEF::VTc null mutation is present in all the control strains so transport occurs exclusively via Aap. The uptake properties of RU1976 (aapQ P144D braEF::VTc) were compared to those of RU1721 (braEF::VTc), which contains a wild-type copy of Aap and the aap bra deletion mutant RU1722 (aapJQM::VSp braEF::VTc) ( Table 1). The maximum rate (V max ) of transport of small amino acids such as alanine was less severely affected (51% of control) by the AapQ P144D mutation than transport of larger amino acids such as glutamate (25% of control). The lower levels of amino acid uptake observed in RU1976 could also arise from a reduction in solute affinity (K m ). In order to investigate this, uptake rates were determined for glutamate, alanine, aspartate and leucine at varying concentrations in both RU1721 and RU1976. These were then used to calculate K m for amino acid uptake by these strains (Table 1). It is clear that the reduction in amino acid uptake in RU1976 is not due to reduced affinity of Aap for its solutes, indeed, the K m of AapQ P144D is slightly lower than wild-type. Rather, in each case, V max is lower in RU1976. Similar drops in transport rates were determined for all branched-chain amino acids (Table 2).

Plant phenotypes of P144D mutants on peas
In order to analyse if reduced transport by Aap triggers symbiotic auxotrophy, strain RU1976 and the control strains, RU1721 and RU1722 were inoculated onto pea plants. Dry weights of 6 week old pea plants inoculated with RU1976 and RU1721 were similar, while RU1722 dry weights were reduced as expected ( Table 2). Thus a ,60% reduction in the rate of branched-chain amino acid uptake in strain RU1976 compared to RU1721, did not significantly alter N 2 -fixation and presumably bacteroid formation. Consistent with the proposal of symbiotic auxotrophy only a low rate of branched-chain amino acid transport is needed for N 2 -fixation.
Transport phenotype of an aap bra mutant in R. leguminosarum bv. phaseoli The transport specificities and symbiotic phenotypes of R. leguminosarum bv. viciae strain A34 and its aapJQM::VSp braE::Tn-phoA mutant RU1357, have been previously described [1,13]. R. leguminosarum bv. phaseoli 4292 and R. leguminosarum bv. viciae A34 are isogenic strains that only carry different symbiotic (sym) plasmids. Strain 4292 nodulates Phaseolus while strain A34 nodulates Pisum, Vicia and Lens [14]. Mutants, aapJQM::VSp (RU1809), braE::TnphoA (RU1932) and aapJQM::VSp braE::Tn-phoA (RU1933) were isolated in R. leguminosarum bv. phaseoli 4292 as described in Materials and Methods. Transport of several amino acids by these strains confirmed that Aap and Bra have indistinguishable properties in R. leguminosarum bv viciae A34 and R. leguminosarum bv. phaseoli 4292 (Table 3; compare to [13]). Thus in strain 4292 glutamate transport is mainly facilitated via Aap while the branched-chain amino acid leucine is primarily transported via Bra. a-Aminoisobutyric acid (AIB), alanine and histidine are transported at similar rates by both transporters and GABA is exclusively transported via Bra. In the aap bra double mutant (RU1933) the transport of all solutes was reduced to background levels, except for alanine and histidine which under the assay conditions can still enter the cells via other transport systems [13]. It is important to note here, that branched-chain amino acid transport is abolished, which is essential to explain the previously reported phenotype of R. leguminosarum bv. viciae double transport mutants on pea plants [1,2]. R. leguminosarum bv. viciae cells are dependent on the provision of branched-chain amino acids by the host plant to develop into mature bacteroids and probably also for their persistence [2].

Plant phenotypes of aap bra mutants on beans
Growth and symbiotic performance of French bean (P. vulgaris) was measured at different time points. Plants at 8 weeks showed no significant differences when inoculated with the wild-type (RU2222), the aap mutant (RU1809) and the bra mutant (RU1932) ( Table 3). However, the aap bra double mutant, RU1933, produced significantly (p,0.001) reduced dry weights compared to wild-type RU2222 inoculated plants. Plants inoculated with RU1933 also showed signs of yellowing after 8 weeks although not as severe as uninoculated controls ( Figure 1). Nitrogen fixation rates, measured as acetylene reduction (AR), and nodule numbers were determined after 4, 5 and 6 weeks of plant growth. Because bacteroid protein levels per pea plant were reduced in aap bra mutants of R. leguminosarum bv. viciae [1,2] this was also measured in bean nodules after 4 weeks of growth. A strong reduction in nitrogen fixation rates (down to 36% of the wild-type control; Table 4) was accompanied by a reduced nodule number (down to 54%) after 4 weeks of plant growth (Table 4). In addition the level of acetylene reduction per nodule formed by RU1933 was significantly (p = 0.005) reduced after 4 weeks compared to levels in wild-type nodules (Table 4). Nitrogen fixation rates and nodule numbers were also markedly reduced on 5 and 6 weeks old plants inoculated with RU1933 compared to wild-type (Table 5), while acetylene reduction rates per nodule recovered. However, compared to 4 week old plants, nodule numbers on 5 and 6 week old plants were generally higher, which were also grown in bigger pots (see Materials and Methods).
At 5 and 6 weeks the reduced symbiotic performance of strain RU1933 (Table 5) can be explained by the reduced number of   nodules alone. However, at 4 weeks the reduction in nitrogen fixation rates is only partly explained by the reduced nodule number ( Table 4). The nitrogen fixation rate per nodule and bacteroid is reduced as well. TEM pictures of bacteroids formed by wild-type RU2222 and the aap bra mutant RU1933 showed remarkable differences ( Figure 2). Bacteroids of determinate French bean nodules usually show an accumulation of granules of the carbon storage compound poly-hydroxybutyrate (PHB). This accumulation was enhanced in the aap bra mutant background RU1933 (Figure 2). TEM micrographs were prepared from all the three experiments harvested after 4, 5 and 6 weeks and always showed identical results.   Properties of a Sinorhizobium meliloti 2011 aap bra transport mutant S. meliloti livHMGFK is homologous to R. leguminosarum braDEFGC [13], so for clarity we hereafter name S. meliloti livK as braC and livH as braD. The double mutants aapJ::VSp braC::mTn5 (LMB210) and aapJ::VSp braD::mTn5 (LMB211) were generated in the S. meliloti 2011 background as described in Materials and Methods. All amino acid transport was reduced to background levels in the aapJ braD mutant LMB211, which is very similar to R. leguminosarum strains (Table 6). Thus mutations in the SBP of the Aap and integral membrane proteins of the Bra abolish most amino acid transport in S. meliloti. However, strain LMB210 which contains a mutation in the gene coding for the SBP BraC, still transported isoleucine at a high rate. As previously described for R. leguminosarum bv. viciae 3841 an orphan SBP (BraC3) is able to interact with the Bra membrane complex and is very specific for the branched-chain amino acids and alanine [2]. A possible orthologue (SMc00078) of BraC3 is present in S. meliloti strain1021.

Plant phenotypes of S. meliloti aap bra mutants on alfalfa
Medicago sativa plants inoculated with wild-type 2011 or the aapJ braD double mutant LMB211 were analysed for nitrogen fixation (AR) and dry weights after 4 weeks of growth. Neither nitrogen fixation rates nor dry weights differed significantly (p.0.05) between S. meliloti 2011 and LMB211, although uninoculated control plants showed strong signs of nitrogen starvation and a corresponding drop in dry weights (Table 7).
TEM pictures of wild-type 2011 and the aapJ braD mutant LMB211 bacteroids showed no obvious differences in develop-ment (data not shown) and flow cytometric characterisation of bacteroids also failed to demonstrate developmental alterations of bacteroids formed by LMB211 (data not shown).

Discussion
Pea plants inoculated with Rhizobium leguminosarum bv. viciae strains mutated in the two broad specificity amino acid ABC uptake systems, Aap and Bra develop a severe nitrogen starvation phenotype [1,2]. However, the very broad solute specificity of Aap and Bra made it difficult to determine which amino acids are responsible for this phenotype. Therefore, braC was mutated in an aap null mutant background and it was shown that strain RU1979 (DaapJ braC::VSpec) retains the transport of branched-chain amino acids. This is because of the presence of an orphan SBP (BraC3) which specifically binds branched-chain amino acids and interacts with the core Bra membrane complex. Strain RU1979, which only retains branched-chain amino acid transport, fixes nitrogen at wild-type rates on pea plants. In addition complementation analysis using heterologous ABC transport systems demonstrated that branched-chain amino acid transport alone rescues the starvation phenotype on pea plants. Microarray analysis of pea bacteroids showed that the pathways of branchedchain amino acid biosynthesis are transcriptionally down regulated leading to premature arrest of development in aap bra bacteroids [2].
Our first aim in this study was to determine how much branched-chain amino acid transport is needed for fully effective N 2 -fixation. It was already known that the single aap and bra mutants have no symbiotic phenotype so the wild-type rate of transport can be lowered by ,50% with no effect [1]. However, we had previously identified a region likely to be important in solute specificity in AapQ [12] and mutation of P144D reduced branch chain amino acid transport by ,60%. The absolute rates of branched-chain amino acid transport in RU1976 (aapQ P144D braEF::VTc) were 1-2 nmol min 21 mg protein 21 , while the aap bra double mutant strains have rates below 1 nmol min 21 mg protein 21 . This rate for RU1976 is roughly 10% of the maximum rates of branched-chain amino acid uptake measured in wild-type [2,13]. However, RU1976 provided enough branched-chain amino acids for developing bacteroids to achieve a wild-type rate of nitrogen fixation. This supports the hypothesis that R. leguminosarum bacteroids are symbiotic auxotrophs and do not require amino acids for catabolism, which would need a much higher rate of transport.
A leuD mutant of R. leguminosarum strain 3841 required leucine at 100 mM added to AMS glucose/NH 4 Cl cultures to overcome auxotrophy and restore a wild-type rate of growth. By comparison as sole nitrogen source 10 mM leucine was needed to obtain the same growth rate as NH 4 Cl.
Published microarray data shows transcriptional down regulation of branched-chain biosynthetic genes occurs in S. meliloti in symbiosis with alfalfa (M. sativa) and Bradyrhizobium japonicum with soybean (Glycine max) [15,16,17]. Therefore, the requirement for branched-chain amino acid transport was investigated in other legume-rhizobia interactions. The first interaction investigated is that between R. leguminosarum bv. phaseoli and French bean. This was done for two reasons. First, beans have determinate nodules, while peas have indeterminate nodules. Second, R. leguminosarum bv. viciae A34 and R. leguminosarum bv. phaseoli 4292 differ only in their Sym plasmid so the same aap bra mutations can be directly compared in pea and bean hosts.
French bean plants inoculated with RU1933 (aap bra) became nitrogen starved, indicated by reduced dry weight, nitrogen fixation, and nitrogen fixation per nodule as well as yellowing and increased PHB accumulation in bacteroids. Unexpectedly, they produced fewer nodules on beans, in contrast to the increase in nodule formation on peas by equivalent aap bra mutants. Legumes usually increase nodule numbers when inoculated with fix reduced strains. The reduction in French bean nodules inoculated with RU1933 might result from limitation of branched-chain amino acids during infection or very early in bacteroid development leading to reduced success in nodule formation. Therefore, while peas and beans have different nodulation responses to aap bra mutants their different patterns of nodule development probably alter the timing of amino acid availability and the development of symbiotic auxotrophy.
In addition to reduced numbers of nodules elicited by RU1933 bacteroids had more PHB granules as visually scored from TEM pictures of 4, 5 and 6 week old nodules. Increased levels of carbon storage compounds in nodules, usually starch in the plant and PHB in bacteroids, are a typical feature of non-fixing or underperforming strains [1,18,19]. The plant provides carbon and reductant for nitrogen fixation, but the bacteroids of RU1933 may not use it efficiently. This suggests that in addition to less successful infection, nitrogen fixation by bean bacteroids is limited and leads to nitrogen starvation in the host plant. This might reflect, as for pea nodules, a general high demand for branched-chain amino acids [2], which are energetically expensive to make and are the most abundant amino acids in bacterial proteomes [20]. However, when inoculated with aap bra mutants, beans are less severely nitrogen starved than peas. For example French beans had dry weights reduced by 24% while peas were reduced by 59%, close to uninoculated controls [1].
Similar to R. leguminosarum Rlv3841 [2,21], branched-chain amino acid biosynthesis is down regulated in S. meliloti [15,16,22]. Thus an aap bra mutant was isolated in S. meliloti 2011 and its rates of branched-chain amino acid transport were below 1 nmol min 21 mg protein 21 . However, there were no differences in dry weight or acetylene reduction of alfalfa plants inoculated with mutant or wild-type. Inspection of TEM micrographs of bacteroids as well as FAC analysis of their size and DNA content did not reveal any differences between wild-type and mutant. Either the decrease in transcription of the branched-chain amino acid biosynthetic genes in S. meliloti 2011 bacteroids is not severe enough to produce symbiotic auxotrophy or residual transport rates by aap bra mutants are sufficient to support bacteroid maturation and persistence. The ability to reduce branched-chain amino acid transport to a very low rate in R. leguminosarum bv. viciae without inducing symbiotic auxotrophy is consistent with this. If the concentrations of branched-chain amino acids in alfalfa nodules are slightly higher than peas or the reduction in transcription of the genes for their biosynthesis is delayed or less severe, then symbiotic auxotrophy would not occur.
This study demonstrates that symbiotic auxotrophy is a phenotype right on the edge where it can be pushed in one direction or the other by small changes in conditions or gene expression. Perhaps the more important aspect of this is not whether a particular Rhizobium-plant association shows symbiotic auxotrophy for amino acids but, that bacteroids may shut down many aspects of metabolism as long as they retain the ability to transport essential nutrients from host plants. This emphasises the exquisite co-evolution of Rhizobium and legumes with the metabolism of bacteroids being organelle like.

Isolation of mutants
The Rhizobium leguminosarum bv. viciae AapQ site directed mutant was generated as follows. AapJQMP was amplified using primers p495 and p441. The 5 kb PCR product was TOPO cloned into pENTR/D-Topo (Invitrogen, USA) and the CCG triplet coding for proline at position 144 within AapQ replaced by a GAC coding for aspartic acid (P144D) using the Quick Change site-directed mutagenesis kit (Stratagene, USA). The region was transferred by a Gateway LR reaction into the pJQ200SK derivative, pGW1, producing pRU1247. This plasmid was conjugated into RU1722 (aapJQM::VSp braEF::VTc) to generate the homogenote RU1976 by sucrose selection and loss of the spectinomycin marker. The correct mutation was confirmed by PCR.
Rhizobium leguminosarum bv. phaseoli mutants were generated as follows. Plasmid pRU410 [25], which contains an omega spectinomycin (VSp) cassette cloned into an MfeI deletion (3.2 kb) spanning aapJQM, was transformed into Escherichia coli S17-1 and conjugated into R. leguminosarum bv. phaseoli 4292. This plasmid is derived from pJQ200SK [26] which contains the sac genes. Transconjugants were spread on AMS agar plates [24] containing sucrose (10%) as a carbon source and spectinomycin resistant, gentamicin sensitive colonies were selected. One colony (RU1809) was confirmed to have the correct aapJQM::VSp genotype by PCR mapping using primers p418 and p419. As expected from the phenotype of aap mutants of R. leguminosarum bv. viciae, strain RU1809 did not grow on minimal medium with Lglutamate as the sole carbon and nitrogen source. To generate bra mutants, cosmid pBIO206A [25] (pIJ1427 braE::TnphoA) was transformed into S17-1 and conjugated into strain 4292 and RU1809. Colonies were selected for tetracycline and kanamycin, resistance, before the chaser plasmid pPH1JI (gentamicin resistant) was then conjugated into them to allow the selection of recombinants with loss of the tetracycline marker and resistance to kanamycin and gentamicin. Plasmid pPH1JI was then removed by conjugating pLAFR1 into all strains, and selecting for loss of the gentamicin marker while retaining resistance to kanamycin and tetracycline. The braE::TnphoA insertion was confirmed by PCR mapping using primers pIS50R and p439. The braE::TnphoA single and aapJQM::VSp braE::TnphoA double mutants were designated as RU1932 and RU1933, respectively. Strain 4292 containing pLAFR1 was designated RU2222.

Strains, plasmids, primers
Description Reference, source PCR product was cloned into pJET1.2 (Fermentas, Lithuania) resulting in plasmid pLMB111. Primers pr0165 and pr0166 were used in an inverse PCR to amplify the aapJ surrounding DNA from pLMB111 without the aapJ ORF. The resulting PCR product was digested with BamHI (underlined in primers) and religated forming plasmid pLMB112. An XbaI/XhoI fragment was then cloned from pLMB112 into pJQ200SK [26] to generate pLMB113. This vector was opened with BamHI and a VSp cassette was inserted producing pLMB118. This plasmid was used to generate aapJ::VSp homogenotes in the braC and braD mutant backgrounds to produce strains LMB210 and LMB211, respectively. The insertion was mapped from both sides using primers pOTforward and pr0240 or primer pr0241, respectively.

Transport assays
R. leguminosarum uptake assays were performed with 25 mM (4.625 kBq of 14 C) solute [13], using cultures grown in AMS with 10 mM D-glucose and 10 mM NH 4 Cl to an OD 600 of ,0.4. S. meliloti was grown in modified M9 medium with 10 mM D-glucose and 10 mM NH 4 Cl to an OD 600 of ,0.4.

Plant assays
French bean plants (Phaseolus vulgaris cv. Tendergreen) were surface sterilised and grown for 5 to 6 weeks in a sterile mixture of fine gravel and vermiculite (1:4) in 15 l pots with 7 seeds per pot in a controlled environment at 22uC and a 8:16 h day: night cycle as described earlier [29]. The 4 week experiments were performed in 8 l pots in a fine gravel and vermiculite mixture (1:4) with 7 seeds per pot in a controlled environment at 22uC and a 8:16 h day: night cycle. Pots were open to drain into a sterile autoclave bag and watered with nitrogen free rooting solution [30]. Medicago sativa cv. Europe plants were grown in 1 l pots in vermiculite with 7 seeds per pot for 4 weeks watered with nitrogen free rooting solution [30].
Acetylene reduction was measured as previously described [31]. Bean shoot dry weights were determined after 8 weeks, pea shoot dry weights after 6 weeks and alfalfa shoot dry weights after 4 weeks of growth.

Bacteroid assays
After counting and removing all nodules from the plants, nodules were crushed in 40 mM HEPES buffer and bacteroids harvested by differential centrifugation [32]. Bacteroid pellets were then broken in a FastPrep FP120 ribolyser (Bio101, Thermo) and protein levels determined following a standard Bradford assay.
Flow cytometric analysis of S. meliloti bacteroids was performed as described elsewhere [2].