Rhizobia are a diverse group of Gram-negative bacteria comprised of the genera Rhizobium, Bradyrhizobium, Mesorhizobium, Sinorhizobium and Azorhizobium. A unifying characteristic of the rhizobia is their capacity to reduce (fix) atmospheric nitrogen in symbiotic association with a compatible plant host. Symbiotic nitrogen fixation requires a substantial input of energy from the rhizobial symbiont. This review focuses on recent studies of rhizobial carbon metabolism which have demonstrated the importance of a functional tricarboxylic acid (TCA) cycle in allowing rhizobia to efficiently colonize the plant host and/or develop an effective nitrogen fixing symbiosis. Several anaplerotic pathways have also been shown to maintain TCA cycle activity under specific conditions. Biochemical and physiological characterization of carbon metabolic mutants, along with the analysis of cloned genes and their corresponding gene products, have greatly advanced our understanding of the function of enzymes such as citrate synthase, oxoglutarate dehydrogenase, pyruvate carboxylase and malic enzymes. However, much remains to be learned about the control and function of these and other key metabolic enzymes in rhizobia.
Bacteria in the genera Rhizobium, Sinorhizobium, Mesorhizobium, Bradyrhizobium and Azorhizobium (collectively called rhizobia) are Gram-negative heterotrophs with the unique ability to reduce nitrogen to ammonia in symbiotic association with a compatible legume host. Various aspects of carbon metabolism in this agronomically important group have been presented in previous reviews [1–5]. Recent studies with carbon metabolic mutants have begun to elucidate the metabolic significance of a number of tricarboxylic (TCA) cycle and anaplerotic enzymes in rhizobia, and this review summarizes these findings with an emphasis on the importance of these enzymes in symbiosis.
The development of the rhizobia-legume symbiosis has been reviewed [5–8]and what follows describes only a few general features of the interaction. During the formation of nitrogen-fixing root nodules rhizobia are converted from soil-dwelling saprophytes into intracellular symbionts, or bacteroids, which rely on the plant host to provide them with carbon substrates. The bacteroids are surrounded by a host-derived symbiosome membrane which mediates the exchange of metabolites between the microsymbiont and the plant [5, 8]. In quantitative terms the major metabolic exchange during symbiosis involves the bacteroids receiving reduced carbon from the plant in exchange for the nitrogen they fix. The dicarboxylic acids succinate and/or malate are produced in large quantities in the plant cells of the nodule and are the major carbon sources provided to the bacteroids [3, 5, 8–10]. In addition to dicarboxylic acids, other organic or amino acids may be used by some rhizobia during infection or by bacteroids under environmental stress [3, 5, 8, 11–17].
Nitrogenase, the bacteroid enzyme which catalyzes the reduction of atmospheric nitrogen, requires at least 16 ATP and 8 reducing equivalents per mol of ammonia produced. Because nitrogenase is oxygen-labile, a microaerobic environment is imposed by a gas permeability barrier in the nodule subcortex and the oxygen-binding protein leghemoglobin in the infected plant cells [7–9, 18]. In bacteroids, microaerobiosis triggers the synthesis of new respiratory chain components needed for a microaerobic metabolism and energy generation by oxidative phosphorylation, as well as the induction of genes encoding regulatory or structural proteins directly involved in nitrogen fixation [7, 18].
2Enzymes of the TCA cycle
In addition to energy generation the TCA cycle (Fig. 1) is used to produce precursors for the biosynthesis of amino acids, purines, pyrimidines and vitamins. The cycle has been intensively studied in Escherichia coli and Bacillus subtilis and a complex network of genetic and metabolic controls have been elucidated in these organisms [19–23]. The rhizobial oxygen-responsive regulators (FixLJ, FixK) which induce the synthesis of electron transport components needed for microaerobic respiration in bacteroids [7, 24]are homologs of the global regulator Fnr which performs a similar function in E. coli during anaerobic growth . However, control systems analogous to the E. coli ArcAB system, which directs the anaerobic repression of several TCA cycle genes , have not yet been encountered in rhizobia. Thus, although the details of TCA cycle regulation in rhizobia are lacking, a few broad generalizations from studies in other prokaryotes may apply, namely (i) the TCA cycle is regulated to ensure that energy and precursor generation match the needs imposed by growth in a particular environment , (ii) growth conditions determine which anaplerotic reactions (Section 3) are used to maintain TCA cycle function and (iii) paralogs, or products of homologous genes which perform related but non-identical functions in the same organism , exist for several TCA cycle and anaplerotic enzymes .
TCA cycle enzyme activities have been measured in bacteroids of many rhizobia (Table 1) and have been tentatively correlated with symbiotic efficiency [27, 28]. Enzyme activity data (Table 1) along with respirometric studies strongly indicate that a complete cycle is present in bacteroids [5, 29, 30]and in cells in culture [1, 2, 5]. The TCA cycle in bacteroids probably operates below its full aerobic potential because the microaerobic conditions in nodules limit the ability of the respiratory chain to oxidize reduced nucleotides, which may inhibit the activities of citrate synthase, isocitrate dehydrogenase (Section 2.1) and 2-oxoglutarate dehydrogenase (Section 2.2) [3, 5, 9, 31]. Nevertheless, studies with a variety of enzyme mutants described here support the importance of a complete TCA cycle in bacteroids.
Table 1. Specific activities of pyruvate dehydrogenase and TCA cycle enzymes in bacteroidsa
Enzyme and organism
(nmol min−1 mg protein−1)
aBacteroids were aerobically isolated and purified by sucrose or Percoll density gradient centrifugation.
Several pathways may be used to produce acetyl CoA for entry into the TCA cycle. Under aerobic conditions rhizobia rely mostly upon the pyruvate dehydrogenase complex (EC 188.8.131.52) [29, 32–34](Fig. 2). Multiple control mechanisms govern the synthesis and activity of this enzyme in E. coli[35, 36]but little is known about its regulation in rhizobia. In R. etli pyruvate dehydrogenase activity is markedly decreased with low culture oxygen availability and in bacteroids (Table 1), although this may result from the down-regulation of genes encoding enzymes for the synthesis of the catalytic cofactors thiamine pyrophosphate and lipoic acid [37–39]. In contrast to R. etli, pyruvate dehydrogenase activity is detectable in bacteroids of other Rhizobium species (Table 1).
Acetyl CoA may be derived from poly-β-hydroxybutyrate (PHB; Fig. 2), a polyester synthesized by bacteroids of many rhizobia . PHB may serve as a reserve of chemical energy [3, 41, 42]or as the product of an overflow pathway which consumes excess reductant in microaerobically respiring bacteroids [43, 44]. PHB is degraded under carbon-limited conditions in cultures and bacteroids [45–48]to produce acetyl CoA in the reaction catalyzed by β-ketothiolase (Fig. 2).
Acetyl CoA is also synthesized from acetate using acetate kinase in combination with phosphotransacetylase , or by acetyl CoA synthetase [50, 51](Fig. 2). All three activities are present in B. japonicum bacteroids and the in vitro kinetic properties of the acetyl CoA synthetase and acetate kinase suggest that both operate in the direction of acetyl CoA synthesis . It is interesting to note that acetyl phosphate, the product of acetate kinase, is important in regulating TCA cycle activity in E. coli, and a possible similar role for this metabolite in S. meliloti is currently being investigated .
2.1Metabolism of tricarboxylic acids via citrate synthase, aconitase and isocitrate dehydrogenase
The tricarboxylic acid portion of the TCA cycle consists of a three-step conversion of oxaloacetate plus acetyl CoA to 2-oxoglutarate (Fig. 1). In the first reaction citrate synthase (EC 184.108.40.206) condenses the acetyl group of acetyl CoA with oxaloacetate to form citrate (Fig. 1). Citrate synthases from Gram-negative bacteria are often feedback inhibited by 2-oxoglutarate and NADH and are subject to catabolite and anaerobic repression [53–55]. Little is known about the regulation of citrate synthase in rhizobia (see below), although its activity is high during growth on different carbon sources [56, 57]and in bacteroids (Table 1).
Two citrate synthase paralogs are present in R. tropici, one (pcsA) encoded on the symbiotic plasmid and the other (ccsA) on the chromosome (Table 2). The coding sequences of pcsA and ccsA are nearly identical, leading to the suggestion that pcsA may have arisen by duplication of ccsA. Nodules formed by ccsA or pcsA mutants are Fix+ (able to fix nitrogen; Table 2), which is perhaps not surprising since the mutants retain 70–80% of the wild-type citrate synthase activity in planta [58, 59]. Despite the residual activity both mutants are delayed in nodule formation on bean plants, with the ccsA mutant being more severely affected . Double mutants totally lacking citrate synthase activity form Fix− (non-nitrogen fixing) nodules devoid of bacteroids (Table 2), indicating a crucial role for the enzyme early in nodule development. The ccsA and pcsA ccsA (but not the pcsA) mutants are glutamate auxotrophs , suggesting that CcsA is required for the production of 2-oxoglutarate via the TCA cycle .
Table 2. TCA cycle and anaplerotic enzymes: cloned genes and phenotypes of existing mutants
Organism and enzyme
aWhere no genetic designation is provided, the gene encoding the enzyme has neither been cloned nor localized by mutagenesis.
bWhere no phenotype is given, no mutants exist. The mutant's ability to produce the indicated enzyme is listed first, with the superscripts signifying: −, no activity; +, approximately wild-type activity; red, reduced activity. The symbiotic phenotypes are listed next and correspond to: Ndv+, normal nodule development; Ndv±, delay in nodule formation and/or abnormal nodules formed; Fix−, nodules did not fix nitrogen based on plant phenotype or acetylene reduction assay; FixRed, significant reduction in nitrogen fixation; Fix+, nodules fixed nitrogen at approximately wild-type levels.
cK. LeVier and M.L. Guerinot, Dartmouth College, USA.
dIdentified only by comparison to homologs in other organisms.
In contrast to R. tropici, S. meliloti and B. japonicum appear to contain single genes encoding citrate synthase (Table 2). The rhizobial citrate synthases share over 70% deduced amino acid identity and have just slightly lower homology with the enzymes from other Gram-negative species. A S. meliloti citrate synthase mutant had the expected glutamate auxotrophy but in addition appeared to produce altered lipopolysaccharides, which could also explain its Fix− phenotype (Table 2).
Citrate synthase activity in Rhizobium sp. (Cicer) is markedly inhibited by NADH in vitro and, if the same phenomenon occurs in bacteroids, could result in the flow of carbon into PHB synthesis. Unfortunately, the effects of citrate synthase mutation on PHB synthesis have not been reported. The partitioning of acetyl CoA between the TCA cycle and PHB synthesis is also determined by the metabolic regulation of β-ketothiolase, the initial enzyme of the later pathway (Fig. 2). Purified β-ketothiolase from Rhizobium sp. (Cicer) and B. japonicumbacteroids is inhibited by CoASH, which may repress PHB synthesis when citrate synthase activity is high .
Aconitase (EC 220.127.116.11) catalyzes the reversible isomerization of citrate and isocitrate (Fig. 1) and is encoded by acnB in E. coli, where its transcription is repressed by anaerobiosis [53, 63]. In E. coli an aconitase paralog, AcnA, functions in protecting the cell from oxidative stress .
An acnA homolog and its product have been studied in B. japonicum (Table 2) but this aconitase does not seem to participate in the TCA cycle. Instead, it appears that B. japonicum also encodes an AcnB-like enzyme which fulfils this function. This notion is supported by the facts that (i) acnA expression in B. japonicumand E. coliis induced during aerobic growth, (ii) E. coliand B. japonicumacnA mutants retain substantial aconitase activity and (iii) acnA mutants in either species are not glutamate auxotrophs, which in the E. coliacnA mutant results from the production of the aconitase encoded by acnB. The symbiotic properties of the B. japonicumacnA mutant were indistinguishable from those of the wild-type strain (Table 2), although one might expect a true aconitase knockout mutant to be symbiotically ineffective by analogy to the rhizobial citrate synthase and isocitrate dehydrogenase (see below) mutants. Further work is needed to confirm this and to demonstrate that B. japonicum also encodes an acnB homolog.
Isocitrate dehydrogenase (EC 18.104.22.168) catalyzes the oxidation of isocitrate to 2-oxoglutarate (Fig. 1). The activity of this enzyme is fairly constant during growth on different carbohydrate, amino acid or organic acid carbon sources [56, 57, 66–68]and high levels are present in bacteroids (Table 1). Rhizobia normally contain only NADP-specific isocitrate dehydrogenase , although an NAD-linked form occurs in some species [70, 71]. The substrate kinetic constants of NADP-isocitrate dehydrogenase from S. melilotiand B. japonicumare similar to those of the E. coli enzyme . In contrast to the E. coli isocitrate dehydrogenase, which is not allosterically regulated , oxoglutarate and NADPH inhibit the S. melilotiand B. japonicumenzymes, respectively. These forms of product inhibition could operate in bacteroids where the concentrations of reduced nucleotides and perhaps oxoglutarate are high [3, 31, 74].
The gene for the S. meliloti NADP-isocitrate dehydrogenase (Table 2) appears to encode the subunit of a monomeric form of the enzyme [72, 75]in contrast to the homodimeric isocitrate dehydrogenases produced by E. coli and Bacillus. Isocitrate dehydrogenase insertion mutants of S. meliloti were glutamate auxotrophs and were Nod+ but Fix− (Table 2). This phenotype may result from the inability of the mutant to synthesize 2-oxoglutarate, a co-substrate required for the symbiotically important reaction catalyzed by aspartate aminotransferase (Fig. 2) (Section 3.1). Glutamate auxotrophy is a common link between the citrate synthase and isocitrate dehydrogenase mutants, although it is interesting that S. meliloti mutants deficient in both isocitrate dehydrogenase and citrate synthase are Nod−, in contrast to the Nod+ Fix− phenotype obtained when only one of these enzymes is lacking (Table 2).
2.2Metabolism of dicarboxylic acids via 2-oxoglutarate dehydrogenase, succinyl CoA synthetase, succinate dehydrogenase, fumarase and malate dehydrogenase
The dicarboxylic acid portion of the TCA cycle consists of five reactions which regenerate oxaloacetate from 2-oxoglutarate. This pathway is initiated by the 2-oxoglutarate dehydrogenase complex (EC 22.214.171.124), which catalyzes the oxidative decarboxylation of 2-oxoglutarate to succinyl CoA (Fig. 1). Expression of 2-oxoglutarate dehydrogenase in E. coli is repressed by anaerobiosis [25, 54, 76]but, based solely on activity measurements, this does not appear to occur in R. leguminosarum bv. viciae . The low activity observed during growth of R. etli in oxygen-limited cultures and in bacteroids (Table 1) may result from a lack of cofactor biosynthesis, as discussed for pyruvate dehydrogenase (Section 1). The 2-oxoglutarate dehydrogenase from B. japonicum is markedly inhibited by NADH in vitro and, if this also occurs in bacteroids, may provide 2-oxoglutarate for the synthesis and accumulation of glutamate .
A 2-oxoglutarate dehydrogenase (SucA; Table 2) mutant of B. japonicum grew surprisingly well on malate or succinate as carbon sources and was even able to utilize glutamate to some extent. Because this phenotype differs from that reported for most other bacterial 2-oxoglutarate dehydrogenase mutants, it was suggested that a bypass pathway might be circumventing the metabolic block in the mutant . However, the presence of the γ-aminobutyrate bypass (Section 3.2) or the glyoxylate bypass (Section 3.4) could not be demonstrated in the mutant . Nodulation of soybean by the B. japonicum mutant was significantly delayed (Table 2) and infected cells contained a drastically reduced number of bacteroids. Interestingly, those mutant bacteria which did successfully infect and differentiate into bacteroids fixed nitrogen at near wild-type levels, indicating that the mutant was severely hampered during the early stages of the interaction but could generate sufficient energy to fix nitrogen without 2-oxoglutarate dehydrogenase activity , perhaps by using a yet unidentified bypass pathway in planta [77, 78].
A R. leguminosarum bv. viciae sucA transposon mutant formed Fix− nodules on pea (Table 2) and excreted elevated levels of glutamate and 2-oxoglutarate when grown in culture . This later finding indirectly supports the hypothesis that glutamate accumulates in bacteroids as a result of 2-oxoglutarate dehydrogenase inhibition [31, 44, 79]. The sucA mutant produced several-fold higher activities of succinyl CoA synthetase and malate dehydrogenase in vitro , an effect also observed in a chemically induced 2-oxoglutarate dehydrogenase mutant of S. meliloti (Table 2). In these mutants the elevated succinyl CoA synthetase activity might act in the reverse direction to generate precursor quantities of succinyl CoA during growth on either succinate or glucose (Fig. 2), while the higher malate dehydrogenase activity could also serve as part of the biosynthetic route to succinyl CoA in the case of the R. leguminosarum mutant grown on glucose , or to increase the synthesis of 2-oxoglutarate by operating in the forward direction in the case of the S. meliloti mutant grown on succinate (Table 2 and Fig. 2). An apparent contradiction exists between the symbiotic phenotypes of the 2-oxoglutarate dehydrogenase mutants of S. meliloti (Fix−) and B. japonicum (Fix+) (Table 2) and the proposed existence of the γ-aminobutyrate bypass in these genera, since the bypass appears to exist in S. meliloti but not in B. japonicum (Section 3.2).
Of possible regulatory significance is the fact that in R. leguminosarum and B. japonicum the genes for 2-oxoglutarate dehydrogenase (sucAB), succinyl-CoA synthetase (sucCD) and malate dehydrogenase (mdh) are arranged contiguously in the order mdh sucCDAB[44, 77, 80]. This differs from E. coli, where mdh is encoded elsewhere on the chromosome and, most importantly, a cluster containing the 2-oxoglutarate dehydrogenase and citrate synthase genes [21, 76]is regulated from the succinate dehydrogenase (sdhCDAB) promoter located directly upstream [81, 82].
The hydrolysis of succinyl CoA to succinate (Fig. 1) is catalyzed by succinyl CoA synthetase (EC 126.96.36.199). A sucD insertion mutant of R. leguminosarum bv. viciae retained substantial succinyl CoA synthetase activity (raising the possibility of a paralog) but nevertheless formed Fix− nodules on pea (Table 2). This symbiotic phenotype was likely due to the polar effect of the sucD insertion on sucA (recall the sucCDAB gene order described above) which virtually eliminated 2-oxoglutarate dehydrogenase activity in the mutant . The sucD mutant was also impaired in amino acid uptake via the general amino acid permease, possibly as a result of increased glutamate excretion via this bidirectional system [44, 83]. Because of the residual succinyl CoA synthetase activity and the pleiotropic effects of the sucD insertion, conclusions about the requirement for this enzyme in symbiosis cannot be made.
Succinate dehydrogenase (EC 188.8.131.52) catalyzes the dehydrogenation of succinate to fumarate (Fig. 1) and, accordingly, nitrosoguanidine-generated mutants of S. melilotiand R. leguminosarum bv. viciae deficient in this enzyme exhibit little or no growth on succinate but grow well on malate or fumarate. The S. meliloti mutant was able to infect alfalfa but not differentiate into bacteroids (Table 2). Likewise, pea plants inoculated with the R. leguminosarum mutant formed white, Fix− nodules (Table 2). Bacteroids isolated from these nodules had substantial succinate transport activity indicating that the ability to metabolize succinate (Fig. 1) – as well as malate (Section 3.1) – is required for nitrogen fixation but not for infection [85, 86]. Thus succinate and malate are not ‘equivalent’ substrates for bacteroid metabolism as is sometimes assumed , perhaps because the rhizobial succinate dehydrogenase also presumably functions as part of the electron transport chain .
Fumarase (EC 184.108.40.206) catalyzes the hydration of fumarate to form l-malate (Fig. 1). Three distinct fumarases (fumA, fumB and fumC) are encoded in E. coli. FumA and FumB are termed class I fumarases and perform distinct metabolic roles in the TCA cycle under different growth conditions [21, 25]. FumC is a class II fumarase which participates not in the TCA cycle but rather in oxidative stress protection [21, 87]. Curiously, the sole fumarase encoded by B. subtilis is a homolog of the E. coli FumC but, unlike E. coli, fumC mutants of Bacillus exhibit defects in TCA cycle metabolism .
A fumarase gene cloned from B. japonicum had significant sequence identity to class II fumarases and was accordingly designated fumC (Table 2). A deletion mutant retained nearly 60% of the wild-type fumarase activity indicating that a second fumarase was present . The heat stability of the residual activity in the mutant was indicative of a class I (TCA cycle) fumarase, and indeed the fumC mutant was unaltered in its growth characteristics and in symbiosis (Table 2). From this, it is reasonable to assume that the residual activity results from a FumA and/or FumB paralog in B. japonicum. A fumarase knockout mutant might provide a tool for determining if the conversion of succinate to malate is important in bacteroids while avoiding a direct effect on the electron transport system, as may have occurred with the succinate dehydrogenase mutant described above.
The final step in the TCA cycle regenerates oxaloacetate from malate and is catalyzed by malate dehydrogenase (EC 220.127.116.11; Fig. 1). In rhizobia, changes in malate dehydrogenase activity during growth on different substrates are modest [57, 66, 67, 90]and are similar to those observed in E. coli[53, 91]. In contrast to E. coli, growth under oxygen-limited conditions results in little or no decrease in malate dehydrogenase activity in rhizobia [37, 56]and activity is very high in bacteroids (Table 1).
Like the E. coli enzyme , malate dehydrogenase purified from B. japonicum bacteroids is inhibited by NADH, although its kinetic properties, subunit composition and deduced amino acid sequence are more similar to the B. subtilis enzyme (Table 2) [94, 95]. Upstream of both the B. japonicumand R. leguminosarum (Table 2) malate dehydrogenase genes are sequences resembling members of the AAA gene family. AAA gene family products regulate a variety of functions, including transcription, in diverse organisms . Interestingly, bacteroids formed by a B. japonicum mutant with a disrupted AAA-like gene produced substantially more malate dehydrogenase protein and activity in soybean nodules, and had up to 50% higher rates of nitrogen fixation compared to nodules formed by the parental strain . Although the B. japonicum malate dehydrogenase has a significantly higher affinity for malate as compared to malic enzyme [97, 98](Section 3.1), further increasing malate dehydrogenase activity appears to benefit symbiosis by generating more oxaloacetate for the TCA cycle or biosynthesis . Malate dehydrogenase mutants have not yet been isolated from rhizobia, either because mutation is lethal or because the activities of malic enzyme (Section 3.1) and pyruvate carboxylase (Section 3.3) compensate for the inactivated enzyme (Fig. 2), as occurs in B. subtilis malate dehydrogenase mutants . The apparent high flux of plant-provided malate through the TCA cycle almost assures that a malate dehydrogenase mutant, when available, will be Fix−. This prediction is less certain for A. caulinodans, which may also utilize lactate in planta [11–13]. Nitrogen fixation in this organism was, however, significantly reduced when malate dehydrogenase was selectively inhibited with exogenous 2-oxoglutarate .
3Enzymes catalyzing anaplerotic reactions
Because it is regenerated, a single molecule of oxaloacetate would be sufficient to maintain unlimited turns of the TCA cycle operating in a purely catabolic mode. However, rhizobia are no different from other organisms in using TCA cycle intermediates as anabolic precursors (Fig. 1). The withdrawal of cycle intermediates for anabolism would soon halt the cycle if these intermediates were not replaced, and rhizobia contain a variety of anaplerotic enzymes which function in this capacity (Fig. 2). Also discussed here are specialized anaplerotic pathways which may function in bypassing selected reactions of the cycle.
3.1Metabolism of four-carbon compounds: malic enzyme, aspartase and aspartate aminotransferase
Growth on dicarboxylic acids using the TCA cycle requires that rhizobia generate both oxaloacetate (using malate dehydrogenase; Section 2.2) and acetyl CoA from malate . The oxidative decarboxylation of malate to pyruvate is catalyzed by malic enzyme (Fig. 2). Rhizobia typically possess both an NADP-specific malic enzyme (EC 18.104.22.168) and an NAD-malic enzyme (EC 22.214.171.124) which also has some activity with NADP as a cofactor [57, 97–101].
Finan and coworkers have constructed and characterized NAD-malic enzyme (dme), NADP-malic enzyme (tme) and dme tme double mutants of S. meliloti. On alfalfa the dme and dme tme mutants formed small, Fix− nodules while the tme mutant was symbiotically unimpaired (Table 2). Measurements of malic enzyme activities and gene expression showed that only the NAD-malic enzyme was induced in bacteroids [102, 103]. Because the dme mutant bacteroids were enclosed in symbiosome membranes, NAD-malic enzyme activity appears to be important during the active phase of nitrogen fixation but not for infection . Thus it was proposed that the NAD-malic enzyme degrades four-carbon dicarboxylates to supply pyruvate to the TCA cycle, and the finding that acetyl CoA selectively inhibits the enzyme in vitro suggests that this metabolite is indeed the end-product of the NAD-malic enzyme/pyruvate dehydrogenase pathway in vivo . The NADP-malic enzyme may function mainly in generating NADPH for biosynthesis [102–104].
In rhizobia, NADP-malic enzyme has a higher affinity for malate in vitro relative to the NAD form. This difference is especially notable for the malic enzymes from B. japonicum[97, 98, 103, 105]whose separate activities in bacteroids are similar . Thus the different affinities of the malic enzymes for malate may dictate a distinct symbiotic role for each enzyme in this organism . It has been proposed that environmental factors alter the level of malate supplied by the plant to the bacteroids, with low malate supply favoring the provision of NADPH to nitrogenase via the high-affinity NADP-malic enzyme. Because the NAD form would be virtually inactive under these conditions, more malate would flow, via malate dehydrogenase, into the TCA cycle, consuming acetyl CoA and preventing PHB accumulation. With high malate supply, NADH generated by the NAD-malic enzyme would inhibit TCA cycle activity and favor PHB synthesis, resulting in a concomitant oxidation of a portion of the NADPH produced by the NADP-malic enzyme and making less reductant available for nitrogenase . Although environmentally induced changes in malate provision to the bacteroids have not been experimentally demonstrated, flow chamber experiments with B. japonicum bacteroids show a strong influence of malate supply and catabolism on PHB accumulation and nitrogen fixation [42, 105]. Because only the relative intracellular concentration of malate in B. japonicum bacteroids is known , relating the different kinetic properties of the malic enzymes to the levels of substrate likely to be available is not possible.
Aspartase (EC 126.96.36.199) catalyzes the reversible deamination of aspartate to yield ammonia and fumarate (Fig. 2) and may thus play an anaplerotic as well as a biosynthetic role in prokaryotes . High concentrations of aspartate are present in nodules of some rhizobia-legume combinations [74, 100, 106]although whether plant-derived aspartate is available to bacteroids is less certain. The kinetic properties of the Bradyrhizobium sp. (Lupinus) bacteroid aspartase would appear to favor a biosynthetic rather than anaplerotic role . In R. leguminosarum, aspartase is present only in cells grown on aspartate or asparagine, suggesting that it functions in the degradation of one or both of these amino acids . Similarly, aspartase in R. etli functions in the degradation of asparagine after its conversion to aspartate by an inducible asparaginase . High aspartase and asparaginase activities in R. etliand Bradyrhizobium sp. (Lupinus) bacteroids indicate that some degradation of asparagine may occur in symbiosis, but whether these reactions actually function to generate aspartate for fumarate synthesis is not known. Bean nodules formed by aspartase-deficient R. etli mutants were Fix+ (Table 2), although aspartate aminotransferase (see below) could also catabolize aspartate and compensate for the lowered aspartase activity .
Aspartate aminotransferase (EC 188.8.131.52) catalyzes the reversible transamination of aspartate and glutamate to yield oxaloacetate (Fig. 2). These enzymes are ubiquitous and paralogs are common [113, 114]. Aspartate could potentially be catabolized by bacterial aspartate aminotransferases to produce oxaloacetate for the TCA cycle , as high activities are found in bacteroids of S. meliloti and several other rhizobia [66, 100, 116, 117].
Several distinct genes encoding aspartate aminotransferases have been cloned from S. meliloti although only one, the product of aatA, is substrate specific and required for nitrogen fixation (Table 2). The essential role of AatA in the S. meliloti-alfalfa symbiosis suggests that aspartate may be an important carbon source in planta . Catabolism of aspartate to oxaloacetate by aspartate aminotransferase also results in the production of glutamate (Fig. 2), which could be metabolized via 2-oxoglutarate (Section 2.2) or via the γ-aminobutyrate bypass (Section 3.2) . Bacteroids of S. meliloti are capable of aspartate transport and levels of this amino acid in the microsymbiont are less than 1% that in the nodule cytosol , indicating that if aspartate is transported into the bacteroids it is rapidly catabolized.
3.2Metabolism of 2-oxoglutarate via the γ-aminobutyrate bypass
γ-Aminobutyrate is often present in high concentrations in nodules [74, 100, 106, 120]or bacteroids [74, 106, 121]and can be used as a sole nitrogen and carbon source by many rhizobia in vitro [100, 122, 123]. Glutamate, the precursor of γ-aminobutyrate, is synthesized from 2-oxoglutarate as shown in Fig. 2. The γ-aminobutyrate bypass allows E. coli to metabolize endogenous oxoglutarate during anaerobic growth when 2-oxoglutarate dehydrogenase activity is repressed . A similar use of the bypass in rhizobia was discussed in Section 2. The lack of glutamate transport across the symbiosome membrane appears to preclude a role for the pathway in catabolizing host-supplied glutamate.
In the first step of the bypass glutamate is decarboxylated to γ-aminobutyrate by glutamate decarboxylase (EC 184.108.40.206; Fig. 2). Several workers have failed to detect glutamate decarboxylase activity in Bradyrhizobium[31, 77, 100]while others have reported high activities in bacteroids of this genus and in S. meliloti, where the enzyme is induced by glutamate in cultures . The inconsistency in detecting glutamate decarboxylase in rhizobia may be due to strain differences, culture growth conditions or enzyme extraction and assay methods, all of which have varied considerably in different studies.
The reversible conversion of γ-aminobutyrate to succinate semialdehyde is catalyzed by γ-aminobutyrate transaminase (EC 220.127.116.11) (Fig. 2), which has distinct preferences for keto acid acceptors in different rhizobia. The transaminase from B. japonicum has maximal activity with pyruvate or oxaloacetate and much lower activity with 2-oxoglutarate [31, 126], while the enzymes from cowpea Rhizobiumand S. melilotihave roughly the opposite preference for keto acid acceptor. If the γ-aminobutyrate bypass does in fact function in skirting an inhibited 2-oxoglutarate dehydrogenase reaction, the presence of oxoglutarate resulting from this inhibition might provide a physiological benefit for using it as a keto acid acceptor (Fig. 2).
The final step of the bypass is the oxidation of succinate semialdehyde to succinate via succinate semialdehyde dehydrogenase (EC 18.104.22.168; Fig. 2). Rhizobia contain either a single succinate semialdehyde dehydrogenase able to use either NAD or NADP as a cofactor , or separate NAD- and NADP-linked isoenzymes . Perhaps the best evidence for the γ-aminobutyrate bypass in rhizobia comes from studies with an S. meliloti mutant (presumably regulatory) having reduced levels of both NAD- and NADP-linked succinate semialdehyde dehydrogenases. This mutant grew poorly on glutamate and formed Fix− nodules on alfalfa (Table 2). Phenotypic revertants isolated from this mutant retained the lowered succinate semialdehyde dehydrogenase activities but produced elevated levels of glutamate dehydrogenase. This, in conjunction with the residual succinate semialdehyde dehydrogenase activity present in the revertants, may account for the ability of the revertants to catabolize glutamate (Fig. 2) and again form Fix+ nodules .
3.3Metabolism of three-carbon compounds: pyruvate, phosphoenolpyruvate and propionyl CoA carboxylases, pyruvate orthophosphate dikinase and phosphoenolpyruvate carboxykinase
Carbon dioxide is a required nutrient for rhizobia and, like other bacteria, the carboxylating enzymes pyruvate carboxylase and/or phosphoenolpyruvate carboxylase are required for growth on three-carbon substrates. These enzymes use carbon dioxide to convert pyruvate and phosphoenolpyruvate, respectively, into oxaloacetate (Fig. 2) which is then used for biosynthesis or energy production via the TCA cycle .
Pyruvate carboxylase (EC 22.214.171.124) catalyzes the biotin-dependent carboxylation of pyruvate to form oxaloacetate (Fig. 2). The enzymes produced by Bacillus, Rhodobacter and Rhizobium are α4 homotetramers which are allosterically activated by acetyl CoA and inhibited by l-aspartate . Pyruvate carboxylase-negative mutants of R. leguminosarum bv. trifolii and bv. viciae do not grow on three-carbon substrates or carbohydrates [132, 133]but form fully functional nodules on their respective hosts (Table 2). In R. etli pyruvate carboxylase is expressed constitutively and is regulated by acetyl CoA and l-aspartate in vitro . The major factor affecting pyruvate carboxylase activity in R. etli is biotin availability [37, 131], which determines how much constitutively produced apoenzyme is converted to the active (biotinylated) holoenzyme. Gene transcription, but not enzyme activity, was detected in R. etli bacteroids from bean nodules, suggesting that the levels of biotin available to the microsymbiont are relative low in planta .
The symbiotic phenotypes of R. etli and R. tropicipyc mutants were indistinguishable from those of the parental strains, consistent with the results obtained with the R. leguminosarum mutants described above (Table 2). This shows that the carboxylation of pyruvate to form oxaloacetate has no indispensable role in symbiosis, although oxaloacetate may be anaplerotically produced from aspartate (Section 3.1) in the pyruvate carboxylase mutants . Because pyruvate carboxylase is required for growth on a wide variety of carbon sources [123, 131]it may be of importance to rhizobia living in the soil.
While S. meliloti and R. leguminosarum bv. trifolii have been reported to lack phosphoenolpyruvate carboxylase (EC 126.96.36.199) activity [66, 132, 134], the enzyme has been detected in B. japonicum, R. etli, R. tropici[37, 131]and A. caulinodans (M. Dunn, unpublished results). In these species, phosphoenolpyruvate carboxylase might reinforce or replace pyruvate carboxylase during growth on carbohydrates or other compounds metabolized via pyruvate .
Evans and coworkers demonstrated that bacteroids and free-living cells of rhizobia use the methylmalonyl CoA mutase pathway to convert exogenously supplied propionate to succinyl CoA, which is subsequently metabolized via the TCA cycle (Fig. 2). A key enzyme in this pathway is propionyl CoA carboxylase (EC 188.8.131.52), a biotin-dependent enzyme which converts propionyl CoA to methylmalonyl CoA (Fig. 2). The occurrence of propionyl CoA carboxylase in diverse rhizobia and the fact that propionyl CoA can be derived from many sources (e.g., the catabolism of odd-chain fatty acids or certain amino acids ) indicate that the enzyme could fulfil an anaplerotic role in rhizobia. We (M. Dunn and G. Araı́za) are currently exploring this possibility.
Pyruvate orthophosphate dikinase (EC 184.108.40.206) catalyzes the reversible, ATP-dependent conversion of pyruvate to phosphoenolpyruvate (Fig. 2). In E. coli an analogous enzyme, phosphoenolpyruvate synthase, is required for growth on three-carbon acids. Because phosphoenolpyruvate produced by pyruvate orthophosphate dikinase (or phosphoenolpyruvate synthase) is both an intermediate in gluconeogenesis and a precursor for biosynthesis, these enzymes fulfil both gluconeogenic and anaplerotic functions . In S. meliloti phosphoenolpyruvate synthase is absent and pyruvate orthophosphate dikinase activity is produced at very low levels. Furthermore, mutants are unaffected in their utilization of carbon substrates or in symbiosis (Table 2) making the function of pyruvate orthophosphate dikinase rather mysterious, at least under the growth conditions examined so far .
Phosphoenolpyruvate carboxykinase (EC 220.127.116.11) catalyzes the reversible, ATP-dependent decarboxylation of oxaloacetate to form phosphoenolpyruvate (Fig. 2). Its role in catalyzing the first step of gluconeogenesis has been demonstrated in several rhizobia [90, 138–142]although the symbiotic phenotype of mutants lacking the enzyme is host-dependent (Table 2). Although phosphoenolpyruvate carboxykinase in combination with pyruvate orthophosphate dikinase could be used to generate oxaloacetate from pyruvate [136, 143, 144], several lines of evidence indicate that it plays only a gluconeogenic role in rhizobia: (i) rhizobia contain pyruvate and/or phosphoenolpyruvate carboxylases, which function anaplerotically during growth on three-carbon glycolytic substrates, (ii) phosphoenolpyruvate carboxykinase gene transcription is severely repressed during growth on glucose but is markedly induced during growth on gluconeogenic substrates [138, 142, 145]and (iii) S. meliloti pyruvate orthophosphate dikinase mutants are symbiotically indistinguishable from the wild-type , and this enzyme would be required along with phosphoenolpyruvate carboxykinase for the anaplerotic synthesis of oxaloacetate from pyruvate.
3.4Metabolism of acetyl CoA via the glyoxylate bypass
Growth on fatty acids or acetate requires the glyoxylate bypass to circumvent the decarboxylating steps of the TCA cycle (Fig. 2) and allow the net assimilation of carbon from two-carbon precursors . In rhizobia, acetate may be converted to acetyl CoA as described in Section 2.
Isocitrate lyase (EC 18.104.22.168) catalyzes the cleavage of isocitrate to form succinate and glyoxylate (Fig. 2). Activity in rhizobia is highly induced during growth on acetate or oleate [68, 77, 146–148]but at most trace levels are found in bacteroids [68, 147, 148]. Furthermore, bacteroids isolated from carbon-starved nodules of Phaseolus vulgaris contained no detectable isocitrate lyase and nodule lipids were not utilized as a reserve carbon source by these bacteroids . In contrast, isocitrate lyase induction was observed in B. japonicum bacteroids isolated from detached nodules or from plants maintained in darkness . Thus, although free-living rhizobia utilize acetate or fatty acids via the glyoxylate bypass, this pathway does not normally operate in bacteroids for lack of isocitrate lyase activity.
The mechanism by which rhizobia partition isocitrate between the TCA cycle and the glyoxylate bypass has recently been addressed. In E. coli the very low substrate affinity of isocitrate lyase relative to isocitrate dehydrogenase requires that the latter enzyme be inactivated by phosphorylation in order to shunt isocitrate into the glyoxylate bypass . In contrast, the isocitrate Kms for the B. japonicum isocitrate lyase (62 μM) and isocitrate dehydrogenase (16 μM) differ by only four-fold , and the activity of the latter enzyme remains relatively constant regardless of the carbon source used for growth (Section 2.1). These and other data suggest that isocitrate is shunted to the glyoxylate cycle during growth on acetate due to the massive induction of isocitrate lyase and not by post-translational modification of the enzyme. Interestingly, B. japonicum isocitrate lyase activity is very low during growth on malate, which could explain its near absence in bacteroids .
The last reaction of the glyoxylate bypass (Fig. 2) is catalyzed by malate synthase (EC 22.214.171.124), which uses acetyl CoA and glyoxylate as substrates (Fig. 2). Unlike isocitrate lyase, malate synthase activity in rhizobia is high during growth on carbon sources other than acetate or oleate [68, 134, 146–148]. Low levels of malate synthase activity were detected in bacteroids isolated from pea , alfalfa and clover and substantially higher activities in bacteroids from bean, cowpea and soybean [68, 148]. Labeling studies with isolated B. japonicum bacteroids showed that a portion of exogenously supplied acetate was metabolized via malate synthase, with the remainder being oxidized in the TCA cycle [29, 30]. If malate synthase does produce malate in bacteroids, the origin of the glyoxylate used in the reaction does not appear to be isocitrate lyase (see above) but could potentially be provided by the plant host via ureide degradation [3, 149]. McDermott et al. have made the interesting observation that malate synthase activities are significantly higher in bacteroids isolated from most ureide-transporting plants as compared to amide-transporting species. Glyoxylate could also be produced by purine degradation in the microsymbiont , a possibility which remains to be explored.
Studies with carbon metabolic mutants have provided direct evidence for the importance of several TCA cycle and anaplerotic enzymes in rhizobia. Most of these enzymes are required for the normal growth of free-living rhizobia under many conditions, and several are also essential in bacteroids for generating energy or metabolic precursors from dicarboxylic acids. This latter group of enzymes includes citrate synthase, isocitrate dehydrogenase, 2-oxoglutarate dehydrogenase, succinate dehydrogenase and NAD-malic enzyme. Still lacking are definitive biochemical, physiological and mutant studies for several key enzymes including pyruvate dehydrogenase, malate dehydrogenase and aspartase. The apparent existence of paralogs for fumarase and aconitase has not allowed us to determine the role of these activities using the mutant strains available, and much remains to be learned about the distinct metabolic functions of these and other paralogs under different growth conditions or in symbiosis. Another important frontier in the study of carbon metabolism in rhizobia will be the discovery and characterization of global regulatory systems controlling carbon metabolism and integrating it with the rest of cellular physiology.
From a practical standpoint, the characterization of defined metabolic mutants allows us to define which reactions are dispensable and, more importantly, which might be beneficially enhanced by genetic manipulation. Although the metabolic engineering approach to rhizobial strain improvement will undoubtedly be an extremely complex undertaking (i.e., one cannot ignore the control imposed by the plant side of the interaction), studies such as those described here are laying the groundwork needed for such attempts.
I am grateful to Ismael Hernández-Lucas, Jaime Mora and Esperanza Martı́nez-Romero for helpful comments on the manuscript. I thank D. Day, D. Emerich, S. Encarnación, T. Finan, L. Green, M. Guerinot, T. McDermott, V. Romanov, H. Taboada, C. Tabrett and S. Tajima for discussions or for providing data prior to publication. Work in the author's laboratory supported was by grants from CONACyT (N9111-0954, 3232P-N9608 and 3309PB) and DGAPA-UNAM (IN202393, IN209697 and IN213095).