Genes induced in all three plant rhizospheres reflect general life in the rhizosphere and we consider these before examining responses specific to one plant (Additional file 7). They include elevated expression of rhiABC (pRL100169-171) and rhiI (pRL100164), previously described as rhizosphere-induced genes 7 , and the gene for autoinducer synthesis protein CinI (RL3378), which is involved in coordinating quorum-sensing regulation and biofilm formation (Figure 3). Quorum sensing is likely to be important in rhizosphere biology, where bacterial attachment is a key step in root colonization 7 . Expression of genes encoding an alternative aa₃ -type cytochrome c oxidase complex (RL3041-45) and a possibly associated cytochrome c (RL3046) were induced in the rhizosphere (Figure 3). This rhizosphere-induced cytochrome pathway, which is distinct from both the normal cytochrome aa₃ complex found in laboratory cultured bacteria and the high affinity cytochrome cbb₃ complex found in the N₂-fixing nodule form of rhizobia 8 , suggests a distinct redox environment in the rhizosphere. It may be that in the rhizosphere the level of available oxygen is lower than in shaken laboratory culture but higher than in the microaerophilic conditions found inside legume nodules.

Up-regulation of genes encoding C4-dicarboxylate transport protein, DctA (RL3424; Figure 1), and PEP carboxykinase, PckA (RL0037; Figure 2), reveals increased organic acid metabolism in the rhizosphere. Induction of pckA is required for gluconeogenesis and indicates sugar synthesis. R. leguminosarum represses pckA when grown on organic acids with added sugar 9 so while sugars are present in the rhizosphere (that is, based on induction of sugar transporters), central metabolism is almost certainly dominated by catabolism of organic acids. Soils are rich in organic acids and they are the main carbon sources in the tomato rhizosphere 10 . Mutations in both dctA (RL3424) and pckA (RL0037) decreased the ability of R. leguminosarum to compete in the pea rhizosphere as shown by RCIs of 0.65 and 0.57, respectively (Additional file 8).

The glyoxylate cycle was induced showing that short chain (C2) organic acids are catabolized (Figure 2). C1 metabolism is important based on the induction of NAD⁺-dependent formate dehydrogenase (RL4391-3) in all rhizospheres (Figure 2). Formate induced this operon in a laboratory culture of Rlv3841 (Additional file 9) and is a carbon source for autotrophic growth of S. meliloti 11 . Formate dehydrogenase (RL4391-3) requires a Mo cofactor and the gene encoding MoaA2 (RL2711), involved in molybdenum cofactor biosynthesis, showed elevated expression (Figure 2). Mutation of moaA2 (RL2711) resulted in a RCI of 0.73 in the pea rhizosphere (Additional file 8). In addition, in all the rhizospheres tested there was induction of an ABC transporter solute binding protein (SBP; RL3040) from the MolT (molybdate transporter) family (ABC families are according to Saier 12 ), which is likely to be part of an uptake system for molybdate (Figure 1; Additional file 10).

Aromatic compounds are important precursors or breakdown products of many plant compounds and can be used as a source of carbon by rhizosphere bacteria. Their presence in the rhizosphere is illustrated by induction of genes encoding transport systems for uptake of shikimate and protocatechuate. Shikimate is taken up by a multi-facilitator super-family (MFS) transporter (RL4709). Protocatechuate is imported by a TRAP transporter (pRL120499-pRL120500; Figure 1), which was identified by high level induction of pRL120498-500 in microarrays of cells grown in the presence of protocatechuate (Additional files 2 and 5). In the pea rhizosphere, mutation of pRL120500 led to a RCI of 0.72 (Additional file 8). Catabolism of aromatic compounds has also been shown to be important for Pseudomonas putida in the rhizosphere of Zea mays 13 .

One of the strongest general metabolic responses in the rhizosphere was induction of genes encoding proteins involved in catabolism of phenylalanine and tyrosine (RL1860-6; Figure 2). These genes were also induced in free-living cells grown on phenylalanine (Additional file 9). The presence of phenylalanine in the rhizosphere probably results from its important role as a precursor for lignin synthesis by roots. Mutation of two genes encoding enzymes on this phenylalanine breakdown pathway (RL1860 and RL1863; Figure 2) led to two of the largest reductions in pea rhizosphere competitiveness (RCIs of 0.42 and 0.45, respectively; Additional file 8).

Common to all rhizospheres was induction of genes for uptake systems for inositol (IntA, RL4655) 8 14 and sorbitol/mannitol/dulcitol (MtlE, RL4218). Also elevated were genes encoding components of two previously uncharacterized systems. The first, RL3840, encodes a CUT1 (carbohydrate uptake transporter 1) family SBP likely to transport raffinose, melibiose and lactose based on 91% identity to SMb20931 from S. meliloti, whose expression was induced by these sugars 5 . The second, pRL110281, which encodes a PepT (peptide/opine/nickel transporter) family SBP, is clearly important in the pea rhizosphere since mutation of the gene led to a RCI of 0.44 (Additional file 8). The contiguous gene, pRL110282, encodes a product with putative α-N-arabinofuranosidase activity that could be responsible for removing arabinose subunits from arabinan. Based on this proximity, pRL110281 may import the arabinose polymer arabinan, or an oligosaccharide derived from it. Indeed, pRL110281 is unlikely to transport arabinose as its gene was not induced in laboratory cultures grown on arabinose (Additional file 9). Growth on arabinose did cause induction of genes encoding components of CUT2-family transporters, RL3615-6 and RL2377-8 (Additional files 9 and 10), neither of which was elevated in the rhizospheres tested.

Co-induction of transport systems and metabolic pathways provides additional evidence of the presence of a compound in the rhizosphere. The gene encoding mannitol dehydrogenase (MalK, RL4214), which converts mannitol to fructose, was elevated (Figure 2) along with those of a mannitol uptake system (MtlE, RL4218; Figure 1). Although intA (RL4655), encoding the myo-inositol transporter, was induced, genes for inositol catabolism were not. Induction of uptake genes may occur at lower substrate concentrations than for catabolic genes, and there are many examples in our data where catabolic genes were less induced than corresponding transport genes.

The genes encoding a CUT1 family ABC system (RL3860-2) were induced in all rhizospheres and although the solute specificity of this system is unknown, it clearly has a role in the pea rhizosphere as a mutation in RL3860 led to a RCI of 0.58 (Additional file 8). The transporter genes are surrounded by genes encoding predicted mandelate racemase/muconate lactonising proteins (RL3858, RL3864-66), a family of enzymes involved in breakdown of lignin-derived aromatic compounds, protocatechuate and catechol to intermediates of the citric acid cycle via the β-ketoadipate pathway. Although the transport genes were induced in all rhizospheres, of the genes encoding lignin breakdown enzymes only RL3864 was slightly elevated in the alfalfa rhizosphere (1.5-fold, P ≤ 0.05) and RL3866 in the pea rhizosphere (1.4-fold, P ≤ 0.05).

Examination of ABC transporters with unknown solute-specificity shows that 11 genes encoding CUT1 transporters were induced in plant rhizospheres (Figure 1). An increase of expression of CUT1 systems, which usually import oligosaccharides and their derivatives, is consistent with the presence in the rhizosphere of many different poly- and oligosaccharides. In addition, some members of the PepT class also transport oligosaccharides - for example, the α-galactoside (Agp) transporter (pRL110243; Figure 1) 15 . Genes for five PepT transporters were induced in these rhizospheres, with three of unknown solute-specificity induced only in the alfalfa rhizosphere (pRL90101, pRL120243 and pRL120609-10; Figure 1). This work supports the hypothesis that the large increase in the number of high-affinity ABC systems in rhizobia (and other α-proteobacteria) results from selective adaptation to the oligotrophic nature of soil and the rhizosphere.

Plants produce antimicrobial agents (for example, phytoalexins) that bacteria must degrade or export. Plant-made antimicrobials such as halogenated hydrocarbons (for example, dichloroethane) could be dealt with by induction of RL4047 and RL4267, whose products may catalyze conversion of dichloroethane via chloroacetic acid to glycolate, with further degradation by the glyoxylate cycle (Figure 2). RL4267 shows 84% identity to a Xanthobacter autotropicus enzyme involved in 1,2-dichloroethane degradation 16 . Mutation of RL4267 resulted in a strain with reduced competitiveness in the pea rhizosphere (RCI = 0.47; Additional file 8). There have been descriptions of other haloalkanoate dehalogenases in Rhizobium sp. 17 , suggesting halogenated hydrocarbons may act as antimicrobials around roots.

In addition to metabolic detoxification, expression of the multi-drug resistance (MDR) family efflux pump encoded by rmrA (pRL90059) was elevated (Figure 1). RmrA is a membrane fusion protein whose role is typically to dock an inner membrane exporter to a TolC-like protein that spans the periplasm and outer membrane. An R. etli rmrA mutant produced 40% less nodules on bean roots and had increased sensitivity to phytoalexins, flavonoids and salicylic acid 18 . Another membrane fusion protein elevated in all rhizospheres is encoded by RL4274, a RND (resistance-nodulation-cell division) multi-drug exporter (Figure 1). The importance of this system in the pea rhizosphere is demonstrated by a mutant of RL4274 having a RCI of 0.57 (Additional file 8).

Maintaining the correct osmotic environment is important for bacteria in any situation. In all rhizospheres there was elevated expression of ndvA (RL4640). NdvA is responsible for export of cyclic β-1-2-glucan to the bacterial periplasm and important in rhizobia for hypoosmotic regulation 19 (Figure 1). Expression of RL1908, encoding a small-conductance mechanosensitive ion channel (MscS), was also elevated in the rhizospheres examined (Figure 1) and is important in osmotic homeostasis. This suggests that the test microcosm was mildly hypoosmotic.

Elevation of expression of genes involved in response to stress occurred in all rhizospheres (Figure 3). Mutation of RL3982 and RL4265 (msrB), which encode general- and oxidative-stress proteins, reduced pea rhizosphere competitiveness (RCIs of 0.52 and 0.55, respectively; Additional file 8). Some of the largest effects on ability to compete in rhizospheres were shown by mutation of genes encoding proteins of unknown function; the photo reaction centre (PRC) family protein RL0913 and flavoprotein RL3366 had RCIs of 0.42 and 0.43, respectively, in the pea rhizosphere (Additional file 8). Mutation of RL2946, encoding part of a two-component sensor regulator (Figure 3), led to a RCI of 0.59 in the pea rhizosphere (Additional file 8).

The largest class of genes induced only in legume rhizospheres were the nod genes (nodABCEFIJLMNO), required for synthesis and export of nodulation factors (Figure 3). This acts as an exquisite internal control since nodulation factors are specifically induced in response to secretion of flavonoids by legumes 3 . There is also a legume rhizosphere-specific transporter encoded by pRL90085 (Figure 1) shown to be important in the pea rhizosphere as mutation led to a RCI of 0.52 (Additional file 8). Although the solute is unknown, it is probably a monosaccharide, as pRL90085 is in the CUT2 family.

Increased expression of genes encoding enzymes of the glyoxylate cycle (RL0054, RL0866) only occurred in the pea rhizosphere. RL0054 (malate synthase) forms malate from glyoxylate and acetyl CoA while GlcF (RL0866) probably converts glycolate to glyoxylate (Figure 2). Thus, while C2 metabolism is elevated in all rhizospheres, it is particularly important in that of pea. Curiously, although the gene for isocitrate lyase (RL0761, aceA) was up-regulated in both alfalfa and sugar beet rhizospheres, indicating elevated C2 metabolism, expression of RL0054, encoding malate synthase, was only elevated in that of pea (Figure 2).

The gene for MFS transporter of tartrate (RL0996) was induced by three-fold or more in the pea rhizosphere (Figure 1; Additional file 7) while that for tartrate dehydrogenase (RL0995), which converts tartrate to oxaloglycolate for metabolism by the glyoxylate cycle, was only induced by legumes 20 (Figure 2). Mutation of RL0996, encoding the tartrate transporter, led to the largest effect on ability to compete in the pea rhizosphere (RCI = 0.35; Additional file 8). RL0996 was also induced 1.5-fold in the alfalfa rhizosphere, so although this falls below our two-fold cutoff, it suggests tartrate utilization may be important in legume rhizospheres (Additional file 8). However, tartrate may be more generally important as in Agrobacterium vitis the ability to utilize tartrate offered a selective advantage for growth on grapevine 21 .

R. leguminosarum Rlv3841 has a chromosome and six plasmids designated pRL7-pRL12, with pRL10 containing most nodulation and nitrogen fixation genes 22 . Although pea rhizosphere-induced genes from different parts of the genome have been discussed above, many genes on pRL8 are specifically up-regulated in the pea rhizosphere (Figure 4; Additional file 6). Indeed, 37% (11 genes) of the 30 genes elevated by three-fold or more specifically in the pea rhizosphere (using both direct and indirect comparisons (Additional file 7)) are encoded on pRL8. With a threshold of up-regulation of two-fold or more (P ≤ 0.05), then 21 genes on pRL8 are pea rhizosphere-specific (15% of all genes on pRL8). By comparison, only three and two genes on pRL8 were up-regulated in alfalfa and sugar beet rhizospheres, respectively, and two genes were up-regulated in the legume rhizosphere. Since plasmid pRL8 is conjugative 22 , it can easily transfer between rhizobia. Consistent with its heavy bias to genes important in the pea rhizosphere, pRL8 shows little colinearity (< 5%) with other sequenced rhizobial genomes 23 . BLAST analysis shows that of its 142 genes, 25% are found only in R. leguminosarum bv viciae and a further 42% are specific to rhizobia or related α-proteobacteria.

Genes on pRL8 that are pea rhizosphere-specific include a molybdenum-containing xanthine dehydrogenase-like carbon monoxide dehydrogenase, CoxMSL (pRL80023-25), together with accessory protein CoxG (pRL80021). Nearby are genes for proteins that may be involved in maturation of this complex: proteins involved in molybdopterin biosynthesis (pRL80034 and pRL80033 encode MoaA and MoeA-like proteins, respectively) and CoxI (pRL80038), which is needed for insertion of a molydopterin cofactor into a xanthine dehydrogenase. However, while CoxMSL (pRL80023-25) may be able to catalyze CO conversion to CO₂ (Figure 2), in phylogenetic clustering these proteins form a separate clade from the biochemically characterized CO dehydrogenases, including one from Bradyrhizobium japonicum 24 . Mutation of pRL80021 (coxG) and pRL80023 (coxM) resulted in reduced competitiveness in the pea rhizosphere (RCIs of 0.44 and 0.73, respectively; Additional file 8). Since pRL80021 is up-regulated only in the pea rhizosphere, its mutation did not result in reduced competitiveness in the alfalfa rhizosphere (RCI = 0.97; Additional file 8).

Homoserine is abundant in pea root mucilage and can be utilized as a carbon source by R. leguminosarum 25 . Although the genes involved in catabolism of homoserine are uncharacterized in Rlv3841, pRL80071, which encodes a putative homoserine dehydrogenase (which catalyses conversion of homoserine to L-aspartate-semialdehyde), was specifically up-regulated in the pea rhizosphere (Figure 2).

Tryptophan is probably available in the rhizosphere 26 . The gene encoding N-formylkynurenine formidase (pRL80036) was six- to ten-fold elevated in the pea rhizosphere (Figure 4). It catalyzes release of formate and kynurenine from N-formylkynurenine, formed after the first step in tryptophan catabolism. The formate produced might be further metabolized to CO₂ by a NAD⁺-containing short-chain dehydrogenase encoded by pRL80037, whose expression is 2.5- to 3.5-fold elevated (Figure 4).

Mimosine (β-3-hydroxy-4 pyridone amino acid), a toxic amino acid related to tyrosine, is produced by the tree-legume leucaena, which is nodulated by Rhizobium sp. TAL1145. Rhizobium sp. TAL1145 has a specific ABC importer for mimosine (MidABC) and an aminotransferase responsible for its degradation (MidD) 27 . An ABC importer (encoded by pRL80060/pRL80063-4; Figure 1), which shows 44 to 79% identity to MidABC 27 , was induced in the pea rhizosphere. However, there is no protein in Rlv3841 with > 27% identity to the aminotransferase required for mimosine degradation. While the transporter encoded by pRL80060/pRL80063-4 is unlikely to transport mimosine, it may transport a similar amino acid. Expression of this system was also elevated in 7-day-old pea bacteroids 8 ; thus, it may have a role in the symbiotic interaction between Rlv3841 and pea.

Also elevated specifically in pea rhizospheres were pRL80026-30, which encode proteins belonging to the HAAT (hydrophobic amino acid transporter) ABC family (Figure 1). Despite the fact that this transporter has been annotated as a LIV (leucine, isoleucine, valine) system, it could transport one or more aromatic amino acid(s) or homoserine. Mutation of pRL80026 resulted in a RCI of 0.69 in the pea rhizosphere (Additional file 8).

Export of plant toxins is likely to be important for successful growth in the rhizosphere. The gene encoding the RND family exporter RL4274 was induced 3.5- to 2.8-fold in the rhizospheres of alfalfa and sugar beet and 135-fold (P ≤ 0.05) in pea (Figure 1). In addition, secDF2 (RL0680) encodes a membrane protein specifically induced to a very high level only in the pea rhizosphere (> 100-fold, P ≤ 0.05; Figure 1). SecDF homologues belong to the RND exporter superfamily and RL0680 may participate in metabolite rather than protein export. It is clear that SecDF2 has a key role affecting competitiveness in the pea rhizosphere since a mutant in this gene had a RCI of 0.45 (Additional file 8). In contrast, in the alfalfa rhizosphere the same mutant had a RCI of 0.97 (Additional file 8), showing that mutation of secDF2 has no significant effect on its ability to compete with Rlv3841in the alfalfa rhizosphere.

The mildly hypoosmotic nature of the general rhizosphere microcosm was revealed by the induction of genes encoding cyclic β-1-2-glucan exporter (NdvA, RL4640) and a mechanosensitive ion channel (MscS, RL1908) (Figure 1). In addition, the gene for a second MscS (RL1522) was specifically induced in the pea rhizosphere (Figure 1), presumably fine-tuning the osmotic response.

A dicarboxylate may be a key carbon source in the alfalfa rhizosphere since genes for the malonate transporter (MatC, RL0992; Figure 1) and enzymes for malonate metabolism to acetyl CoA (MatA, RL0990, and MatB, RL0991; Figure 2) were induced. Alternatively, the malonate present in the alfalfa rhizosphere may need to be detoxified as it can act as an inhibitor of succinate dehydrogenase.

Canavanine is a toxic amino acid analogue of arginine found in seeds and exudates of leguminous plants 28 . The canavanine exporter MsiA from Mesorhizobium tianshanense shows 97% identity with RL2856. In Rlv3841, expression of RL2856 was specifically elevated in the alfalfa rhizosphere (Figure 1). Canavanine comprises 0.6 to 1.6% of the dry weight of alfalfa seeds and restricts growth of bacteria 28 . MsiA (RL2856) was slightly induced in vetch bacteroids (3.2-fold, P = 0.07) 8 , suggesting that vetch releases some canavanine. MsiA is important for attachment to root hairs and survival in rhizospheres of canavanine-producing legumes 29 . The ability to deal with toxic canavanine may enable selection of MsiA-producing bacteria by these leguminous plants.

RL2720 encodes a CUT2 family SBP specifically induced in the alfalfa rhizosphere (Figure 1). From the up-regulation of expression of RL2720-2 (4- to 25-fold) in microarrays of cells grown in the presence of arabinogalactan (Additional file 9), this cluster of genes may encode an arabinogalactan transporter (Additional file 10). In addition, this gene cluster encodes two transketolases (RL2718-19). Expression of RL2719 was elevated (12-fold) in the alfalfa rhizosphere (Additional file 7), as was that of a short chain dehydrogenase (RL2725). The proteins encoded by these genes may have a role in arabinogalactan metabolism. These genes were not elevated in microarrays of cells grown on galactose or arabinose (Additional file 9), indicating a response only to the polysaccharide or to oligosaccharide break-down products (and not the monosaccharide constituents) of arabinogalactan.

Comparative analysis reveals that the sugar beet rhizosphere is N-limited. Expression of genes encoding glutamine synthetase II (GlnII, RL3549), the NH₄ ⁺ transporter AmtB (RL4564) (Figure 1) and PAAT (polar amino acid transporter) family importer pRL120079 (Figure 1) was elevated, as they were in all N-limited microarrays (Additional files 7 and 9). As these experiments were deliberately conducted in N-free plant growth medium, it is not surprising that the sugar beet rhizosphere was N-limited. However, legume rhizospheres were not N-limited, indicating release of nitrogen into the rhizosphere, possibly specifically because rhizobia were present.

As useful information about how micro-organisms respond to the rhizosphere can be obtained by incubating bacteria in root secretions 6 , bacterial responses to constituent components were measured (Table 1). Pea root exudate was added to liquid-grown cells and microarray analysis showed 21 genes elevated by ≥ 3-fold (P ≤ 0.05), 18 of which were also elevated by growth in medium containing the flavonoid hesperetin (27 genes elevated ≥ 3-fold, P ≤ 0.05) (Additional file 6). Common induced genes, in addition to the nod and rhi gene clusters on pRL10 (nodABCEFIJLMNO and rhiIABC), include RL2418, encoding a CUT1 ABC transporter (and also induced three-fold or more in all rhizospheres and in the presence of acetoacetate, hydroxybenzoate and protocatechuate) (Additional files 6 and 9). Expression of RL2418 is clearly important for pea rhizosphere growth as a mutant showed much reduced competitiveness with Rlv3841 in the pea rhizosphere (RCI 0.43; Additional file 8).

Genes elevated under these two conditions (root exudate and hesperetin) form only a small proportion of those elevated in plant rhizospheres. One reason is that root secretions are very dilute and only induce genes responsive to low concentrations of bioactive compounds (for example, nod genes). Furthermore, growth in the rhizosphere involves a far more complex series of interactions, including attachment to roots, biofilm formation, contact with a complex array of plant macromolecules and cell-cell competition.

R. leguminosarum strains were grown either in tryptone yeast 30 or acid minimal salts as described 31 . Strains, plasmids and oligonucleotide primers are described in Additional file 11. Seeds of pea (Pisum sativa cv Avola), alfalfa (Medicago sativa) or sugar beet (Beta vulgaris) were surface sterilized and grown in 50 ml Falcon tubes containing autoclaved washed vermiculite and N-free rooting solution 32 . Growth was at 23°C with a 16-h/8-h light/dark cycle for 7, 14 or 21 days before inoculation with Rlv3841 (10³ or 10⁸ CFU). After 1, 3 and 7 days shoots were removed and roots vortexed (5 minutes) in 6 ml sterile water and 12 ml RNA Protect (Qiagen, Crawley, West Sussex, UK). Insoluble material was removed by filtration through four layers of sterile muslin and centrifugation (160 × g, 1 minute, 4°C). Bacteria were recovered by centrifugation (12,000 × g, 10 minutes, 4°C) and re-suspended (200 μL 10 mM Tris-HCl, pH 8).

Total RNA was extracted, quantified, amplified and hybridized to microarrays as described previously 8 . Results were analyzed using GeneSpring 7.2. (Agilent Technologies, Santa Clara, CA, USA) as described previously 8 . In brief, labeling, hybridization and scanning were as previously described, spot recognition was performed with Bluefuse (BlueGnome Limited, Cambridge, UK) and data were imported into GeneSpring 7.2 (Silicon Genetics, Redwood, CA, USA). The local background value was subtracted from the intensity of each spot and a Lowess normalization applied to the slide. Log ratios of expression and P-values were determined in GeneSpring 7.2. To establish a standard technique for sampling (that is, pea plant age on inoculation, harvest time dpi, inoculum size) preliminary experiments were performed (Table 1). Pea plant age was varied over 7, 14 and 21 days with bacterial harvest at 1 dpi (Additional file 1). Following inoculation of 7-day-old peas, the bacterial harvest was assessed at 1, 3 and 7 dpi (Additional file 2). The effect of Rlv3841 inoculum size (10³ versus 10⁸ CFU) was analyzed at 7 dpi of 7-day-old peas (E-MEXP-2854) (Table 1; Additional file 3). Standard conditions established were inoculation of 7-day-old plants with 10⁸ CFU followed by bacterial harvest at 7 dpi.

Using the standard assay conditions, the first method used to compare rhizospheres was microarray analysis of bacteria grown in a rhizosphere (that of pea, alfalfa or sugar beet) against glucose-grown laboratory cultures (leading to an indirect comparison between rhizospheres). Five independent biological replicates were used for pea and three for alfalfa and sugar beet (Table 1). A second method of analysis was comparison of samples isolated from two different rhizospheres (direct comparison) each with four independent biological replicates taken at 7 dpi of 7-day-old plants (Table 1; Additional files 4 and 5).

Gene expression for selected genes was confirmed by quantitative RT-PCR performed in triplicate using the QuantiTect SYBR Green PCR Kit (Qiagen) on an MJ Mini cycler MiniOpticon Real-Time PCR Detection System (Bio-Rad, Hemel Hempstead, Hertfordshire, UK) as previously described 33 . Primers are given in Additional file 11. The data were analyzed by the relative quantification method (comparative C_T method (ΔΔC_T)) to calculate the fold expression 34 35 . Expression levels were normalized against mdh and analyzed with REST 34 . The results from the two normalization procedures were not significantly different (Additional file 12).

R. leguminosarum bv viciae 300 (Rlv300, Str^s) is the parent of streptomycin-resistant (Str^r) derivative Rlv3841. Mutations were made in 46 up-regulated genes in Rlv300 using plasmid integration (leading to neomycin-resistant (Neo^r) mutant colonies) as described 8 (Additional file 8). To assess bacterial competiveness, 7-day-old pea or alfalfa seedlings (n = 8) were inoculated with a Neo^r mutant (10⁴ CFU) and wild-type Str^r Rlv3841 (10³ CFU). The 10:1 ratio of mutant to wild type was used as the wild type increases its frequency very effectively against strains with attenuated competitiveness 33 , enabling detection of subtle decreases in competitiveness. After 7 days growth, recovered bacteria were serially diluted and CFU scored on tryptone yeast agar with trimethoprim (10 μg ml^-1) and nystatin (50 μg ml^-1) with either Str (500 μg ml^-1; selects Rlv3841) or Neo (80 μg ml^-1; selects mutant in Rlv300 background) as described 33 . A nifH mutant was used as a negative control as it is only expressed in nodules and is unaffected in its ability to compete in the rhizosphere, while a thiM mutant (thiamine auxotroph) was used as a positive control because it is severely attenuated for competitiveness in the pea rhizosphere 33 . To establish a baseline competition between wild-type strains, where no selective advantage is expected, Rlv300 was inoculated at ten-fold excess over Rlv3841. Rlv3841 was recovered at 9.7 ± 0.28% (mean ± standard error of the mean, n = 51; Additional file 8), close to the theoretical 9.1%. Similarly, for the positive (thiM) and negative (nifH) controls, Rlv3841 was recovered from the rhizosphere at 71.6 ± 1.6% and 10.0 ± 0.9% (mean ± standard error of the mean), respectively. These data have been expressed as RCI = 1/[(Percentage Rlv3841 recovered versus mutant)/(Percentage Rlv3841 recovered versus Rlv300)]. The RCI for the thiM mutant was 0.14 (1/(71.7%/9.7%)) and for nifH was 0.97 (1/(10.0%/9.7%)). RCIs for all mutants are shown in Additional file 8 and range from 0.35 to 1.1. As the RCIs for even the most affected mutants were not as low as that for the thiamine auxotroph (thiM), it was concluded that although no mutant strain was crippled for growth in the rhizosphere, several were significantly reduced in ability to colonize the rhizosphere in competition with wild-type bacteria.