INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Rhizobium tropici induces nitrogen-fixing nodules on several unrelated tropical legume plants, including species of Phaseolus, Leucaena, and Macroptilium (30). Bacterial respiration is essential for symbiotic nitrogen fixation in two ways: the ATP necessary for nitrogen fixation is derived from oxidative phosphorylation, and respiration removes oxygen, thereby preventing inactivation of nitrogenase by oxygen.

Rhizobia, like many other bacteria, possess branched respiratory chains with three or more terminal oxidases. The electrons derived from different sources are channeled to the quinone pool in the cytoplasmic membrane and from there are transferred directly to quinol oxidases or, via the cytochrome bc₁ complex and cytochrome c, to cytochrome c oxidases. The respiratory chains of Bradyrhizobium japonicum, Rhizobium leguminosarum, Rhizobium etli, and Azorhizobium caulinodans have been studied (1, 10, 18, 20, 28, 44). They all possess a cytochrome c oxidase of the aa₃ type, which is a major component of the respiratory chain in aerobiosis. This type of oxidase has also been described in R. tropici (14). Alternative cytochrome c oxidases or quinol oxidases may also contribute to aerobic respiration. In nodules, where oxygen levels are low, a cytochrome c oxidase of the cbb₃ type with a high affinity for oxygen facilitates respiration. The subunits of this oxidase are encoded by the fixNOQP genes, which have been described in most rhizobia (10, 18).

R. etli, like the related species R. tropici, can nodulate Phaseolus beans. Mutants of R. etli with increased respiration (33, 44, 45, 46) were isolated on the basis of their enhanced capacities to oxidize N,N,N',N'-tetramethyl-p-phenylendiamine (TMPD) by using the Nadi cytochrome oxidase test (29). This test is based on the reaction of TMPD and -naphthol in the presence of cytochrome c and cytochrome c oxidase to produce indophenol blue. Rhizobium mutants affecting formation of cytochrome bc₁, CycM, or cytochrome aa₃ are Nadi (10), whereas mutants with increased respiration via cytochrome c stain more strongly (Nadi⁺⁺). Two R. etli Nadi⁺⁺ mutants had increased respiration due to induction of the fixNOQP genes under free-living conditions (34, 46). The genes affected in both mutants are linked to the purine biosynthetic pathway, and it was proposed that the intermediary metabolite 5-amino-4-imidazolecarboxamide ribonucleotide could act as a negative effector of cytochrome cbb₃ production in R. etli (46). One of these mutants and two uncharacterized mutants increased the nitrogen content of Phaseolus vulgaris plants by 22 to 25% compared with the wild-type strain (33, 44, 45). Mutants of Sinorhizobium meliloti with increased respiration and symbiotic performance have also been described (54). The aim of this study was to isolate and characterize R. tropici mutants with enhanced respiration and symbiotic nitrogen fixation on P. vulgaris plants.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Microbiological techniques. The bacterial strains and plasmids used are listed in Table 1. R. tropici and Agrobacterium tumefaciens strains were grown at 28°C in TY medium (2) or Y minimal medium (43) supplemented with 10 mM ammonium chloride and 0.2% (wt/vol) succinate, glucose, galactose, or mannitol. Escherichia coli strains were grown at 37°C in L medium (31), to which maltose (0.5% [wt/vol]) was added for experiments involving detection of glycogen. Antibiotics were added as appropriate to the following final concentrations (micrograms per milliliter): ampicillin, 400; gentamicin (GEN), 10; kanamycin (KAN), 20; rifampin, 20; spectinomycin, 100; and tetracycline, 10. Sucrose, when present, was added at 5% (wt/vol).

Genetic techniques. Plasmids were transferred from E. coli DH5 to R. tropici by triparental matings with the helper plasmid pRK2013 (11). R. tropici CIAT899 was mutagenized with Tn5 using pJB4JI (3), selecting for mutants on Y-succinate medium supplemented with rifampin and KAN. Colonies were screened using the Nadi cytochrome oxidase test (29), which measures cytochrome c oxidase activity based on the reaction of N,N-dimethyl-p-phenylendiamine or TMPD and -naphthol, in the presence of bacterial cytochrome c and cytochrome c oxidase, to produce indophenol blue. Mutants with enhanced activity (Nadi⁺⁺) were isolated as colonies staining more strongly. The approximately 13-kb EcoRI fragment carrying the Tn5 insertion from the Nadi⁺⁺ mutant A554 was cloned into the EcoRI site in the multiple cloning site of partially digested pJQ200KS (39). The resulting plasmid (pIJ7796) was introduced into CIAT899, and selection was made for sucrose-resistant GEN-sensitive recombinants. A639 was one such recombinant and was confirmed by DNA hybridization to contain Tn5 in the appropriate location. The glgA deletion mutant A656 was generated by a reciprocal crossover, exchanging a deletion derivative of glgA with the Tn5 in A554. Plasmid pIJ7883 carries glgA in a 2.2-kb ClaI-SacI insert. An in-frame deletion of glgA was constructed by excising a 624-bp PstI fragment to form pIJ7884, and then the remaining 1.6-kb fragment was subcloned as a SalI-SacI fragment into pJQ200KS to form pIJ7899. This plasmid was conjugated into A554 and sucrose-resistant, GEN and KAN-sensitive recombinants were isolated. One such isolate, A656, was confirmed by DNA hybridization to have the appropriate pattern of DNA fragments. The fusions of the glycogen metabolism genes to the E. coli lacZ gene were generated by subcloning the following DNA fragments into pMP220 (47) in the correct orientation, resulting in the plasmids indicated in parentheses: 8-kb HindIII (glgP-lacZ; pIJ9160), 1.3-kb HindIII (glgB-lacZ; pIJ9147), 2.4-kb HindIII (glgC-lacZ; pIJ9014), 0.7-kb EcoRI-XbaI (glgA-lacZ; pIJ9016), 1.5-kb EcoRI-XbaI (pgm-lacZ; pIJ9015), 7.3-kb EcoRI-XbaI (glgA::Tn5-pgm-lacZ; pIJ9034), 1.0-kb PstI-EcoRI (glgAPstI-pgm-lacZ; pIJ9044), and 0.8-kb XbaI-PstI (glg-lacZ; pIJ9011) (Fig. 1B). -Galactosidase activity was assayed essentially as described by Rossen et al. (40) with cells grown in Y medium supplemented with glucose or galactose.

View larger version (23K): [in this window] [in a new window]   FIG. 1.   (A) Physical and genetic map of the R. tropici glycogen region and complementation of the Nadi2+ and low-EPS phenotypes of mutants A554 and A656 by different R. tropici cosmids. +, complementation; , no complementation. The open arrows represent the glycogen metabolism genes and the direction of their transcription. The broken lines indicate that the genes have not been sequenced completely. The position of Tn5 in mutant A554 is indicated by an inverted triangle, and the fragment of glgA deleted in mutant A656 is represented by a dotted segment. The dotted arrows indicate the putative transcriptional units. B, BamHI; E, EcoRI; H, HindIII; P, PstI; Sc, SacI. (B) Transcriptional fusions of the glycogen metabolism genes to E. coli lacZ. The direction of transcription is indicated by the arrows. The dotted segment of pIJ9044 indicates the region deleted. (C) Biochemical reactions involved in glycogen metabolism in A. tumefaciens (adapted from reference 49). The gene products indicated are proposed to catalyze the corresponding reactions.

DNA manipulations. Various operations were carried out according to the method of Sambrook et al. (41). DNA hybridizations were done at 65°C in 5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), and washes were done in 1× SSC-0.1% sodium dodecyl sulfate (SDS) at 65°C. Double-stranded DNA was sequenced with a Thermo Sequenase Dye Terminator cycle sequencing kit (Amersham) and the automatic sequencer ABI377 (Perkin-Elmer). The glgC, glgA, pgm, and glgX genes were sequenced on both strands; only parts of glgP and glgB were sequenced on both strands. In addition to the universal and reverse primers, the following primers were used: 5'-GAAGTCAGATCCTGGAAAACGGGAA-3' (to sequence one strand from the end of Tn5), 5'-GGATACGTCGGATTTCATGCCCTG-3' (to sequence the PstI deletion used to generate mutant A656), and 5'-CGTTGTCCTGGCAATAGGCA-3' (to complete the sequence of the glgX gene). The nucleotide sequence was analyzed with the Genetics Computer Group version 8.1 package.

Cell fractionation, protein gel electrophoresis, heme staining, and immunostaining. R. tropici cells grown on succinate were fractionated into membranes and soluble fractions essentially as described previously (9). Protein concentrations were estimated by the method of Bradford (5) using bovine serum albumin as a standard. The membrane and soluble fractions were suspended in loading buffer (124 mM Tris [pH 7.0] 20% [vol/vol] glycerol, 4.6% [wt/vol] SDS), incubated at 42°C for 20 min, separated by SDS-polyacrylamide gel electrophoresis (PAGE), and transferred to a nitrocellulose filter. Proteins containing covalently bound heme were stained by chemiluminescence as described by Vargas et al. (51).

Spectra and respiratory activities. For spectra, R. tropici cells were washed in fresh Y-succinate medium, diluted to an optical density of 0.05 at 540 nm, and grown at 30°C for 12 h. The cells were harvested by centrifugation, washed, and suspended to 30% (wt/vol) in 0.1 M phosphate buffer (pH 7.4) with 40% glycerol. Cytochrome spectra were recorded using an SLM Aminco Midan II spectrophotometer. The samples were reduced with a few granules of dithionite or oxidized with ammonium persulfate. Carbon monoxide difference spectra (reduced + CO minus reduced) were obtained by bubbling CO for 2 min through a dithionite-reduced cell suspension, and the spectra were recorded against a reduced sample. The spectra were measured at room temperature in 1.0-cm-diameter light path cuvettes. For the determination of oxygen uptake, the cells were harvested after 48 h of growth at 28°C in Y-succinate medium and resuspended in 1 ml of 25 mM potassium phosphate buffer (pH 7.0). The oxygen uptake was measured with a Hansatech electrode at 25°C after addition of 2 mM TMPD and 10 mM sodium ascorbate (final concentrations).

Glycogen determination. The glycogen content of R. tropici strains was determined as described by Krisman (22) using R. tropici cells grown for 3 days at 28°C in 40 ml of Y-glucose medium containing 3.5 instead of 10 mM NH₄Cl. The amount of glycogen present was calculated from the absorbance at 540 nm in relation to standard spectra recorded using rabbit glycogen (concentration range, 20 to 300 µg/ml) as a standard.

Quantitative analysis of EPS. The strains were grown for 4 days in Y medium supplemented with either glucose or galactose at 28°C. The cultures were centrifuged at 10,000 × g for 60 min, and the supernatant was collected, diluted 1:1 with distilled water, and centrifuged again in order to eliminate residual cells. The high-molecular-weight EPS (HMW-EPS) was precipitated from the supernatant by adding sodium chloride (to 0.3 M) and 2.5 volumes of ethanol. After 16 h at 4°C, the precipitated HMW-EPS was spooled with a glass rod and left to dry at 37°C in a preweighed petri dish. The ethanol-precipitable material was measured by weight difference. The structure of R. tropici CIAT899 EPS was previously determined (15).

Plant tests. P. vulgaris cv. Negro Jamapa plants (black beans) were grown in a greenhouse in Leonard jars with a nitrogen-free medium for an average of 45 days, as described by Vincent (52). At least four replicates were used per strain. Three parameters were measured: the dry weight of the aerial part of the plant, the number of nodules formed, and the dry weight of nodules per plant. The results were analyzed statistically by analysis of variance.

Nucluotide sequence accession number. The DNA sequence was submitted to the EMBL database and has been assigned accession number AJ291603.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Isolation and characterization of an R. tropici Nadi⁺⁺ mutant. Tn5-induced mutants of R. tropici CIAT899 were screened on succinate minimal medium using the Nadi test (29). One mutant, A554, which stained darker blue (Nadi⁺⁺) than the wild type was chosen for detailed study because it had enhanced levels of symbiotic nitrogen fixation (see below). This mutant had a higher rate of TMPD oxidation than CIAT899 (21 versus 15 ng-atom of O₂/min/mg [dry weight] of cells). The Tn5 and flanking DNA from A554 was cloned on an EcoRI fragment into the suicide vector pJQ200KS (39) to form pIJ7796, and the Tn5 was recombined into CIAT899. The recombinant, A639, had the same Nadi⁺⁺ phenotype as A554, confirming that the Tn5 insertion was responsible for the phenotype. DNA hybridizations with DNA from A554 and A639 demonstrated that each carried only one copy of Tn5 and that it was inserted in the same position in both strains.

Cytochrome composition of A554. The Nadi⁺⁺ phenotype of A554 suggested that the mutation may affect the expression of respiratory chain components, so these were analyzed. Heme staining of soluble and membrane fractions of CIAT899 (Fig. 2) revealed a major soluble component of approximately 14 kDa, probably corresponding to periplasmic cytochrome c, and two membrane components of about 20 and 31 kDa, probably corresponding to CycM and cytochrome c₁. In other work, mutations affecting the genes encoding the cytochrome bc₁ complex have been shown to abolish the formation of the 31-kDa component (unpublished results). The heme staining of A554 is different from that of CIAT899 in that staining of the soluble 14-kDa component is less intense (Fig. 2A) and the membrane components of 31 (cytochrome c₁) and 20 (CycM) kDa stained slightly more strongly (Fig. 2B). This suggests higher levels of cytochrome c₁ and CycM in the mutant. Both of these cytochromes are essential for a positive Nadi reaction in Rhizobium strains (10). The increases in cytochrome c₁ and CycM could account for the enhanced Nadi staining in A554 compared with CIAT899.

View larger version (38K): [in this window] [in a new window]   FIG. 2.   C-heme stains of soluble (A) and membrane (B) proteins from R. tropici CIAT899 (lanes 1) and A554 (lanes 2) after SDS-PAGE. Each lane was loaded with 50 (A) or 30 (B) µg of protein. The molecular masses of the main bands are indicated and were estimated using prestained molecular mass markers (New England Biolabs) that do not show up on the heme-stained gel.

View larger version (11K): [in this window] [in a new window]   FIG. 3.   (A) Reduced minus oxidized difference spectra of R. tropici wild-type (CIAT899) and glycogen synthase mutant (A554) strains. (B) Reduced + CO minus reduced difference spectra of strains CIAT899 and A554. The numbers represent the wavelengths (in nanometers) at which the main peaks and troughs are detected. The cells were grown in Y-succinate minimal medium for 12 h. The protein concentrations were as follows: CIAT899, 15.2 mg/ml; A554, 17.0 mg/ml.

A554 has enhanced symbiotic performance. P. vulgaris cv. Negro Jamapa beans were grown under nitrogen limitation and inoculated with the wild-type CIAT899 or A554. Uninoculated control plants grew very poorly, and CIAT899 strongly stimulated growth (Table 2). The average dry weights of plants inoculated with the mutant A554 were 38 and 20% greater than those observed with CIAT899 in experiments conducted in two separate growing seasons (Table 2). The yield difference between plants inoculated with A554 and CIAT899 in the first experiment was significant at the 95% confidence level, although the difference in the second experiment was not significant at the 95% confidence level. However, using a two-way analysis of variance combining the data from both experiments, A554 was found to give a significantly enhanced yield compared with the wild type (confidence level, 98%).

Characterization of the gene region affected in A554. The DNA sequence of the 6.6-kb HindIII fragment that complemented the Nadi⁺⁺ phenotype of A554 (Fig. 1A) revealed the presence of four open reading frames transcribed in the same direction. Three adjacent HindIII fragments of 2.4, 1.3, and 8.0 kb (Fig. 1A) were also subcloned and partially sequenced, revealing two additional open reading frames also transcribed in the same direction. All six of the genes were found to encode proteins with sequences similar to those of the products of E. coli genes involved in glycogen metabolism (37): glgP (encoding glycogen phosphorylase), glgB (glycogen branching enzyme), glgC (ADP glucose pyrophosphorylase), glgA (glycogen synthase), pgm (phosphoglucomutase), and glgX (glycogen debranching enzyme). The roles of these six proteins in the synthesis and degradation of glycogen in E. coli are illustrated in Fig. 1C. In E. coli and many other bacteria, glycogen synthesis proceeds as follows (37). Phosphoglucomutase (Pgm) isomerizes glucose-6-P to glucose-1-P, which is used to form ADP-glucose in a reaction catalyzed by ADP-glucose pyrophosphorylase (GlgC). This activated form of glucose is polymerized via -1,4 linkages by glycogen synthase (GlgA). Subsequently, the branching enzyme (GlgB) forms -1,6-glycosidic linkages. Glycogen is degraded by debranching enzyme (GlgX), which removes the -1, 6 linkages, and by glycogen phosphorylase (GlgP), which degrades the -1,4 linkages, giving rise to glucose-1-P. The organization of the glycogen metabolism genes in R. tropici is similar to that reported for A. tumefaciens (48, 50), although the glgX gene was not described in A. tumefaciens.

(i) glgP. The R. tropici glgP-like gene was only partially sequenced. A 621-bp fragment showed 91% similarity at the protein level with A. tumefaciens GlgP. The 3' end of the R. tropici glgP gene was also sequenced (77% similarity at the protein level with A. tumefaciens GlgP in a 294-bp fragment), displaying an overlap between the predicted glgP termination codon (TGA) and the glgB initiation codon (ATGA), suggesting translational coupling (21).

(ii) glgB. The 5' and 3' ends of R. tropici glgB were sequenced, showing high similarity with A. tumefaciens glgB (66% in 285 bp and 89% in 405 bp at the protein level, respectively), which was proposed to encode glycogen branching enzyme based on sequence similarity with the E. coli gene (48).

(iii) glgC. The glgC gene is 26 bp downstream of glgB and is preceded by an A/G-rich region, part of which could constitute a ribosome-binding site. The deduced protein (420 residues) is homologous to GlgC from A. tumefaciens and E. coli (95 and 73% similarity, respectively).

(iv) glgA. The glgA gene is 3 bp downstream of glgC and does not have an obvious ribosome-binding site. The deduced protein (480 residues) is 90 and 66% homologous with GlgA from A. tumefaciens and E. coli, respectively. In E. coli GlgA, two important sites have been described, and they are conserved in the R. tropici glycogen synthase: the lysine at position 15, which forms part of the motif KXGG (where X represents any amino acid) and is involved in the binding of the ADP-glucose substrate (12), and the arginine at position 277 (lysine in the E. coli protein), which constitutes part of the proposed active site (13). DNA sequencing of a 3.5-kb EcoRI-BamHI fragment carrying part of the Tn5 cloned from A554 revealed that the transposon is inserted within glgA at a position corresponding to amino acid residue 304. This mutation was called glgA1::Tn5.

(v) pgm. A gene encoding a phosphoglucomutase homolog is included in the glycogen region of R. tropici, as is found in A. tumefaciens (50), but in E. coli it is elsewhere in the genome (27). The R. tropici pgm gene homolog is 4 bp downstream of the glgA stop codon and is preceded by a possible ribosome-binding site, GAGAGG, 11 bp from the ATG. The deduced Pgm protein (542 residues) is 88 and 50% similar to the A. tumefaciens and E. coli Pgms, respectively.

(vi) glgX. There is an R. tropici glgX homolog 211 bp downstream of pgm, and it is preceded by a possible ribosome-binding site, GGAAAG, 10 bp from the predicted ATG. The deduced protein (656 residues) is homologous to the predicted debranching enzymes of many bacteria, including E. coli (656 amino acids; 61% similarity) (53) (Fig. 4). R. tropici GlgX also shows 56% similarity with the Flavobacterium sp. isoamylase enzyme, which degrades -1,6 glycosidic bonds (23). This similarity is primarily around the four domains (indicated in Fig. 4) common to amylolytic enzymes (19).

View larger version (57K): [in this window] [in a new window]   FIG. 4.   Protein sequence comparison of the glycogen debranching enzymes (GlgX) of R. tropici and E. coli (similarity, 61%; identity, 41%). The asterisks represent identical amino acid residues. The dots indicate conserved substitutions. Important residues conserved between R. tropici and E. coli proteins are marked with lines.

The glgA mutant lacks glycogen. The glycogen contents of strains CIAT899, A554 (glgA1::Tn5), and the complemented strain A554/pIJ7720 were analyzed. CIAT899 and the complemented strain had similar levels of glycogen (1.7 and 1.6 ng/10⁶ CFU, respectively), whereas A554 had essentially no glycogen (<0.01 ng/10⁶ CFU). This confirms that the transposon inserted in the glycogen synthase gene of A554 prevents glycogen formation and that the mutation is complemented by cosmid pIJ7720 carrying the glgA region.

Analysis of operon structure using lacZ fusions. Transcriptional fusions of all six genes were constructed using the lacZ reporter plasmid pMP220 (47) (Fig. 1B). The resulting plasmids were introduced into CIAT899, and the -galactosidase activity was measured (Table 3). Significant levels of activity were detected for the glgP-lacZ (pIJ9160) and pgm-lacZ (pIJ9015) gene fusions, whereas the glgB-lacZ (pIJ9147), glgC-lacZ (pIJ9014), and glgA-lacZ (pIJ9016) fusions had very low levels of activity. The glgX-lacZ fusion (pIJ9011) had -galactosidase activity, but the level observed was low. These results indicate that there is an operon composed of glgP, glgB, glgC, and glgA. The pgm gene appears to have a separate promoter, although the presence of only 4 bp between the predicted coding regions of glgA and pgm suggests that pgm is also expressed within the glgPBCA operon. Such a potential dual control of pgm expression seems reasonable in view of the fact that pgm has an important role in metabolism independent of glycogen biosynthesis (Fig. 1C). In A. tumefaciens, the glgP, glgB, glgC, glgA, and pgm genes are transcribed from a promoter located upstream of glgP (48). The A. tumefaciens pgm gene can also be transcribed as a shorter product from a promoter immediately upstream of an internal translation initiation codon within the pgm gene to yield a shorter Pgm protein (48). DNA sequence comparisons revealed an in-frame ATG within the R. tropici pgm gene in a context almost identical to the sequence found in A. tumefaciens. Therefore, pgm in R. tropici may be transcribed independently, as is seen in A. tumefaciens.

Analysis of a nonpolar glgA deletion mutant. To determine if the pleiotropic phenotypes of the glgA mutant A554 are due to polarity of the Tn5, an in-frame deletion mutation in glgA was constructed in vitro and recombined into the genome of R. tropici to form A656. This mutant was similar to A554 in that it had a Nadi⁺⁺ phenotype on Y-succinate plates and had increased levels of TMPD oxidation (23 ng-atom of O₂/min/mg [dry weight]) compared with CIAT899 (15 ng-atom of O₂/min/mg [dry weight]). Heme staining revealed that the cytochrome c content was essentially the same as that seen with A554, in that the content of cytochrome c₁ and CycM was increased and the soluble cytochrome c content was decreased (data not shown).

View larger version (64K): [in this window] [in a new window]   FIG. 5.   Immunostaining of phosphoglucomutase in A656 (glgAPstI). Soluble proteins from the glgA deletion mutant A656 and the wild type, (wt), CIAT899, were separated by SDS-PAGE and either stained with Coomassie blue (B) or immunostained with antiserum to phosphoglucomutase (A). The same amount of protein (10 µg) was loaded in each lane. The sizes (in kilodaltons) of the prestained standards (st) are indicated to the right of the gel.

glgA mutants have reduced EPS content. The glycogen synthase mutant A554 was originally identified following growth on succinate minimal medium plates, on which it grows normally. It also forms normal-size colonies on galactose or mannitol minimal medium or on complete medium (TY). However, on glucose or maltose the colony size was reduced significantly and the colonies appeared to have less EPS. In liquid glucose minimal medium, the mutants had a somewhat extended lag phase, but the growth rates were normal.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

It had been reported previously that some rhizobial mutants which had increased respiratory capacities had enhanced symbiotic nitrogen fixation (33, 44, 45), and it was our aim to identify mutants of R. tropici that had increased respiration and hence symbiotic nitrogen fixation. The mutant strain A554 was investigated in detail because it fulfilled both of these criteria, and it was somewhat unexpected to find that the mutation affected glycogen synthesis. Mutation of the glycogen synthase (glgA) gene of R. tropici induces pleiotropic effects in addition to the expected block of glycogen formation. It is not immediately obvious why mutation of glgA should also lead to (i) alteration in cytochromes, presumably causing increased ability to oxidize TMPD; (ii) decreased EPS production during growth on glucose (but not galactose or mannitol); and (iii) increased nodulation, apparently resulting in enhanced symbiotic performance with the Phaseolus bean. Indeed, we cannot be sure that these phenotypes are all direct consequences of the inability to form glycogen or which, if any, of these phenotypes causes the other. Thus, for example, we do not know if enhanced symbiotic performance is due to the absence of glycogen, increased respiratory potential, or an alteration in the level of HMW-EPS. It may be of particular significance that an R. etli mutant defective for storage of the other major source of stored carbon (poly--hydroxybutyrate) causes enhanced symbiotic performance (8).

In E. coli, glycogen metabolism is highly regulated at a metabolic level and is also under complex genetic control (37). Several metabolites have positive or negative control over glycogen gene expression. Cyclic AMP (cAMP) receptor protein-cAMP and ppGpp are the main positive regulators of the expression of the glgCAP operon in E. coli. cAMP receptor protein-cAMP also stimulates expression of another gene involved in glycogen biosynthesis, glgS, whose function is unknown (17). The stationary-phase sigma factor ^s (24) has a positive control over glgS. On the other hand, three negative effectors have been described: one that acts in cis (glgR) over the glgCAP operon, one that acts in trans (glgQ) over the glgBX and glgCAP genes (37), and the carbon storage regulator csrA, which negatively affects all glycogen metabolism genes (53) by specifically destabilizing mRNA (26).

It is possible that mutating glgA in R. tropici may result in a change in the levels of glucose phosphates and/or sugar nucleotides, such as ADP-glucose, and that this may affect aspects of metabolism other than glycogen synthesis. Interestingly, mutations of the S. meliloti cya3 gene encoding an adenyl cyclase-like protein also enhanced symbiotic effectiveness (42). In E. coli, intracellular UDP-glucose has regulatory effects via RpoS (4), and it is possible that some such regulatory effect could occur in R. tropici. Accumulation of sugar nucleotides or glucose phosphates might account for the pleiotropic phenotypes, but this needs to be tested experimentally. One way to test the effects of accumulation of ADP-glucose would be to generate a nonpolar glgC mutant. This should be defective for glycogen synthesis but would be predicted not to accumulate ADP-glucose (Fig. 1C) and so could be used to distinguish the effects of the absence of glycogen from possible accumulation of ADP-glucose. Glycogen and EPS biosynthesis are connected at the level of glucose-1-P (Fig. 1C), and the glucose-dependent repression of HMW-EPS biosynthesis in glgA mutants may be a consequence of increased levels of glucose phosphates, which could have some kind of inhibitory feedback effect (directly or indirectly) on enzymes of HMW-EPS biosynthesis. A key enzyme common to both pathways is phosphoglucomutase; when a mutation in this gene was first identified in A. tumefaciens and S. meliloti, the gene was originally called exoC (6) because of the strong effect on EPS formation. Later, the exoC gene was renamed pgm (48) when its function was more clearly understood. The observation that pgm is immediately downstream of glgA led us to suspect that the decreased level of HMW-EPS in the glgA Tn5 mutant was due to a polar effect on pgm, but we found no evidence for such an effect.

The observation that the HMW-EPS-reduced phenotype of glgA mutants is specific for growth on glucose but not other sugars, such as galactose, also argues against a simple polar effect on pgm. A pgm mutant of Agrobaterium lacks EPS on both galactose and glucose media, due to its inability to form UDP-glucose (49), and it is likely that a similar situation would occur in Rhizobium. The normal level of production of HMW-EPS by the R. tropici glgA mutant on galactose therefore implies that pgm is expressed in the glgA mutant. Our analysis of gene expression indicated that the pgm gene is expressed under a promoter separate from that of glgA, although its location immediately downstream of glgA would suggest it is also transcribed along with glgA.

A putative pgm mutant of R. tropici CIAT899 was reported previously (32). This mutant had reduced levels of EPS (although the reduction was not as great as in the R. tropici exo mutants characterized in the same work). The mutant was normal with respect to motility and a symbiotic phenotype on bean. This is different from S. meliloti or A. tumefaciens pgm (exoC) mutants, which lack EPS, have altered LPS, are non motile, and are defective for symbiosis or avirulent, respectively (6, 25). The reported phenotype of the R. tropici exoC mutant is similar to that of the glycogen synthase mutant described here. The insertion of Tn5 in the pgm gene was not demonstrated (32), and it was assumed to be a pgm (exoC) mutant based on complementation of an S. meliloti exoC mutant by a cosmid that carried other genes (32). We suspected that this previously described mutation might be in glgA and tried to construct a pgm mutant of R. tropici. A spectinomycin-resistant interposon (38) was cloned at two different sites of pgm in two different constructs. We integrated the pgm:: alleles on a sacB-based suicide-selection plasmid into the genome but could not generate a second recombination event to make a pgm mutant unless a complementing clone carrying pgm was introduced. This suggests that the pgm gene may be essential for growth in R. tropici. During the course of our work, we had tried to introduce the cloned glgA genes from R. tropici or A. tumefaciens into a pgm mutant of A. tumefaciens, and in both cases we failed to obtain transconjugants with the pgm mutant even though rates of conjugation into the wild-type strain were normal. This suggests that in an A. tumefaciens pgm mutant background, increased expression of glgA may be lethal. Possibly our inability to generate a pgm mutant of R. tropici is a related effect.

The increase in plant growth stimulated by R. tropici glgA mutants could in principle be due to enhanced efficiency of nitrogen fixation by bacteroids, increased nodulation, or both. The primary effect seems to be due to enhanced nodulation. The increased nodulation seems unlikely to be due to altered respiration, but this cannot be formally ruled out. A decrease in EPS formation could in principle affect nodulation, since EPS plays a signaling role during nodulation (35), and indeed, mutations affecting the levels of HMW-EPS in S. meliloti also enhanced symbiotic performance (42). Alternatively, it is possible that the inability to store glycogen in some way causes more efficient infection. However, in general, the plant tends to control the numbers of nodules formed, and this is related to the levels of fixed nitrogen supplied by nitrogen-fixing bacteroids to the plant. Thus, in plants inoculated with rhizobia defective for nitrogen fixation, nodulation is often increased. In view of such observations, one speculative model for the enhanced symbiotic performance of glgA mutants is that increased symbiotic performance may occur because the mutants are initially somewhat impaired in symbiotic nitrogen fixation. This might occur, for example, if the onset of nitrogen fixation is delayed. Alternatively, it could be envisaged that glycogen in bacteroids acts as a carbon store that buffers periods of deficiency in carbon supply from the plant. With glgA mutants, a lack of plant-supplied carbon could rapidly lead to cessation of nitrogen fixation, and this could in principle induce the plant to initiate the formation of additional nodules, and the net effect could be an overall increase in symbiotic performance. Clearly, experimental evidence would be required to test such models. However, the identification of such a bacterial mutant does suggest that there are opportunities for enhancing symbiotic nitrogen fixation in field environments.